p

DevForce

Developer Guide
Version 5.2.5

IdeaBlade DevForce Developers Guide

Contents

Contents
DevForce, Enterprise Applications, and the ADO.NET Entity Framework .............................. 8
The Problem ..........................................................................................................................................................9 Object Mapping Technology ............................................................................................................................... 10 The Microsoft ADO.NET Entity Framework ...................................................................................................... 10

Using DevForce with the Entity Framework ...................................................................................... 13
Advantages of Using DevForce ........................................................................................................................... 14

DevForce in More Detail ....................................................................................................................... 16
Advantages of Using DevForce (Revisited) ........................................................................................................ 16 More DevForce Advantages ................................................................................................................................ 28

Conclusion .............................................................................................................................................. 34

Getting Started.............................................................................................................................. 35
Installation ........................................................................................................................................................... 35 DevForce Start Menu ........................................................................................................................................... 35

The “NorthwindIB" database .............................................................................................................. 37 Development Process ............................................................................................................................. 37

Hello, DevForce ........................................................................................................................... 41
DevForce Application Architecture - The Big Picture ........................................................................................ 41 DevForce and the ADO.NET EntityModel ......................................................................................................... 42 Your First DevForce Application: a Walk-Through ............................................................................................ 45 Building the Domain Model ................................................................................................................................ 45 Add a User Interface ............................................................................................................................................ 66 Add Unit Tests ..................................................................................................................................................... 68 Add a WinForm UI .............................................................................................................................................. 75

Understanding the App.Configs ........................................................................................................... 89
Information Flow Between the App.Configs ....................................................................................................... 91

Monitoring Activity ............................................................................................................................... 92
Appendix: Listings of Sample App.Config Files ................................................................................................. 94 Appendix: Probing Sequence for the App.Config File ........................................................................................ 95

Class Libraries.............................................................................................................................. 96
Important Namespaces ......................................................................................................................... 96 The IdeaBlade.EntityModel.Entity ...................................................................................................... 97 Finding Help on DevForce .................................................................................................................. 100
XML Documentation ......................................................................................................................................... 100 IntelliSense ........................................................................................................................................................ 100 The Object Browser ........................................................................................................................................... 102 Class View ......................................................................................................................................................... 103 Class Diagram.................................................................................................................................................... 103

Business Object Mapping .......................................................................................................... 105
Introduction ......................................................................................................................................... 105
Overview of the ADO.NET Entity Model ......................................................................................................... 106 Working with the IdeaBlade DevForce Object Mapper .................................................................................... 106

Object Mapper Walk-Through .......................................................................................................... 106

2|P age

IdeaBlade DevForce Developers Guide

Contents

Exiting The Object Mapper ............................................................................................................................... 117 The Object Mapper Menus ................................................................................................................................ 118 Injected Base Types ........................................................................................................................................... 119 The Name Pluralizer: Fixing the Pluralization in Type Names ......................................................................... 121 Mapping a Web Service..................................................................................................................................... 124

Notes on the Generated Code ............................................................................................................. 127 Multiple Datasources .......................................................................................................................... 131
DataSourceKeys ................................................................................................................................................ 132

Appendix: Many-to-Many Associations in the Entity Framework................................................. 133

Property Interceptors ................................................................................................................. 140
Named vs. Unnamed Interceptor Actions .......................................................................................................... 141 Interceptor Chaining and Ordering .................................................................................................................... 142

Business Object Persistence....................................................................................................... 156
Note: Code Snippets in This Document............................................................................................................. 158

Object Persistence Overview .............................................................................................................. 158
The Big Picture .................................................................................................................................................. 158 DevForce and the ADO.NET EntityModel ....................................................................................................... 160 Support for POCOs (Plain Old CLR Objects) ................................................................................................... 162 Persistence Management Capabilities ............................................................................................................... 163

Entity Queries and Entity Navigation ............................................................................................... 169
Entity Queries .................................................................................................................................................... 169 Entity Navigation ............................................................................................................................................... 192 The Null Entity .................................................................................................................................................. 202

Asynchronous Communication with the Business Object Server ................................................... 203
Asynchronous Queries ....................................................................................................................................... 203 IAsyncResult Asynchronous Pattern ................................................................................................................. 206 Asynchronous Fulfillment of Navigation Property Queries .............................................................................. 206 Canceling Pending Operations........................................................................................................................... 206

The EntityListManager ...................................................................................................................... 207 Entity Caching ..................................................................................................................................... 210
All Business Objects are Cached ....................................................................................................................... 210 Queries, Navigation, and the Cache ................................................................................................................... 213 Query Workflow ................................................................................................................................................ 215 Query Strategy ................................................................................................................................................... 217 Span Queries ...................................................................................................................................................... 226 Cached Entity Lifespan...................................................................................................................................... 228 Saving the Cache Locally .................................................................................................................................. 228 The TraceViewer: Watch What Data Is Being Loaded, and How ..................................................................... 229

Creating Business Objects .................................................................................................................. 242
When Not to Create ........................................................................................................................................... 242 The Business Object Create Method ................................................................................................................. 242 Auxiliary Business Object Class Methods ......................................................................................................... 247 Adding and Removing Related Objects using Add() and Remove() ................................................................. 247 Business Object Creation Review...................................................................................................................... 249

Saving Business Objects...................................................................................................................... 250
EntityState of an Object ..................................................................................................................................... 250 Undo .................................................................................................................................................................. 250 Validation .......................................................................................................................................................... 250

3|P age

IdeaBlade DevForce Developers Guide

Contents

Temporary Id Fix-up ......................................................................................................................................... 251 Life Cycle Events .............................................................................................................................................. 251 Saves and Transaction Management .................................................................................................................. 254 Re-query After Save .......................................................................................................................................... 255 When Save Fails ................................................................................................................................................ 255 Data Source Concurrency .................................................................................................................................. 258 Saving the “Dependency Graph” ....................................................................................................................... 263 Dependency Graph Retrieval ............................................................................................................................. 266 Workflow For a Save ......................................................................................................................................... 268 Saving the Cache to a Local Disk File ............................................................................................................... 270 XML Serialization of Business Objects ............................................................................................................. 271

Business Object Persistence – Advanced .................................................................................. 274
Getting Information About an Entity Type with GetEntityMeta() ................................................ 275 Access Both Local and Remote Data Sources In the Same N-tier Application ............................ 276 Stored Procedure Queries................................................................................................................... 277
SQL Server Stored Procedure Queries .............................................................................................................. 278

Stored Procedure Entity Navigation.................................................................................................. 281 Forced Re-fetch ................................................................................................................................... 282 Lost Connection During Query .......................................................................................................... 283 Query Cache ........................................................................................................................................ 283
EntityManager.RemoveEntities Overload Preserves Query Cache ................................................................... 284

MergeStrategy In More Detail ........................................................................................................... 285 The EntityManager.AttachEntity Method........................................................................................ 289 Filtering Queries .................................................................................................................................. 291 Query Inversion in More Detail ......................................................................................................... 293 Transactional Queries ......................................................................................................................... 297 DevForce and Data Sources – Deep Dive .......................................................................................... 297
The Object Mapper and Manually Added or Modified Keys ............................................................................ 299 DataSourceKeys, DataSourceKeyResolvers, and DataSourceExtensions ......................................................... 299 EntityManagers and DataSourceExtensions ...................................................................................................... 299 Tenant Extensions.............................................................................................................................................. 302 Multi-Part Extensions ........................................................................................................................................ 303 Extensions and EntityServers ............................................................................................................................ 304 Dynamic DataSourceKeys and the DataSourceKeyResolver ............................................................................ 304

Multiple Application Environments .................................................................................................. 307 Multi-Level Undo with Checkpoints.................................................................................................. 307 Multiple EntityManager Instances .................................................................................................... 309 Multi-Threading in a DevForce App ................................................................................................. 310 Batching Asynchronous Tasks ........................................................................................................... 312 Service Oriented Architecture ........................................................................................................... 314 POCO Support in DevForce............................................................................................................... 315
Examples of POCO Classes ............................................................................................................................... 316 Examples of a POCO Service Provider Class .................................................................................................... 318

4|P age

IdeaBlade DevForce Developers Guide

Contents

Example of a Client-Side Class Containing Extension Methods for the EntityManager ................................... 320 Obtaining an EntityAspect Property on Your POCO Object ............................................................................. 321 Data Contract Serializer (DCS) versus .NET Data Contract Serializer (NDCS) .............................................. 322 POCO Save mechanisms ................................................................................................................................... 326 Summary – Things to Remember When Using POCOs in Your DevForce App .............................................. 329

Validation Through Verification ............................................................................................... 330
DevForce Verification ....................................................................................................................................... 331

Getting Started .................................................................................................................................... 332
Validation-Related Settings In the Object Mapper ............................................................................................ 332 Generated Property Code ................................................................................................................................... 334 Impact of Verifiers on the User Interface – A Caution ...................................................................................... 338

Now That You‟ve Been Initiated (and Before We Enter the Forest): A Quick Overview of the Mechanics............................................................................................................................................. 339 Verification Types Overview .............................................................................................................. 340
Main Verification Classes.................................................................................................................................. 340 Verifiers ............................................................................................................................................................. 341 VerifierResult .................................................................................................................................................... 344 Triggers.............................................................................................................................................................. 347 VerifierEngine ................................................................................................................................................... 348 PropertyValueVerifiers ...................................................................................................................................... 350

Verification Deep Dive ........................................................................................................................ 355
Verifiers ............................................................................................................................................................. 355 Verifier Result ................................................................................................................................................... 359 Triggers.............................................................................................................................................................. 362 VerifierEngine ................................................................................................................................................... 370

Invoking Verification .......................................................................................................................... 375
Instance Verification .......................................................................................................................................... 376 Trigger Verification: Preset and Postset ............................................................................................................ 377 Monitor Execution with the VerifierBatchInterceptor ....................................................................................... 381

Verification and WinForms User Interfaces ..................................................................................... 382
UI Lockup .......................................................................................................................................................... 382 Improving the User‟s Experience ...................................................................................................................... 384

DevForce Silverlight Apps ......................................................................................................... 386
Overview - What is DevForce Silverlight? ........................................................................................................ 386 Creating a DevForce Silverlight Application .................................................................................................... 386 Silverlight Deployment Steps ............................................................................................................................ 387 Questions and Answers...................................................................................................................................... 387 Troubleshooting ................................................................................................................................................. 389

WinForm User Interfaces .......................................................................................................... 393
UI Data Binding ................................................................................................................................... 394
NET Data Binding ............................................................................................................................................. 394 NET v. DevForce WinClient UI Data Binding for WinForms .......................................................................... 395 Data Binding with DevForce WinClient UI Designers For WinForms ............................................................. 397 DevForce WinClient Data Binding Architecture ............................................................................................... 399 Nested Property Paths ........................................................................................................................................ 413 Data Binding to Data Objects of Any Type ....................................................................................................... 418 When to Use .NET Data Binding Instead .......................................................................................................... 421 When Not to Use Data Binding at All ............................................................................................................... 422

UI Architecture .................................................................................................................................... 423

5|P age

IdeaBlade DevForce Developers Guide

Contents

Nested Property Paths ........................................................................................................................................ 423 The BindableList(of T) ...................................................................................................................................... 423 EntityPropertyDescriptors ................................................................................................................................. 439

UI Designers ......................................................................................................................................... 443
BindingManagerDesigners ................................................................................................................................ 443

More on Third-Party WinForm Control Suites ............................................................................... 459
Developer Express “DXperience” ..................................................................................................................... 459 Infragistics “NetAdvantage” .............................................................................................................................. 460

DataBinders ......................................................................................................................................... 460 Troubleshooting ................................................................................................................................... 461
Third-Party Control Suites ................................................................................................................................. 461 UI Performance Tuning ..................................................................................................................................... 462 Large BindingSource loads are Slow ................................................................................................................. 463

DevForce WinClient Assemblies for WinForm Support ................................................................. 463

Web Applications........................................................................................................................ 465
The DevForce ASPDataSource Component ...................................................................................................... 465 Using the ASPDataSource in Development ...................................................................................................... 465 Overridable Methods for Select, Update, Insert, and Delete ............................................................................. 465 The EntityAdapterManager Class ...................................................................................................................... 466 The Configure Data Source Wizard ................................................................................................................... 467 Parameter Collection Editor .............................................................................................................................. 467 Retrieving Schema Information ......................................................................................................................... 468 Third Party Support ........................................................................................................................................... 468

Business Object Server............................................................................................................... 469
Business Object Server Architecture ................................................................................................................. 469 EntityService Startup and Shutdown ................................................................................................................. 472 EntityServer Startup and Shutdown ................................................................................................................... 473 Remote Service Method Call (RSMC) Methods ............................................................................................... 473 Push Notification ............................................................................................................................................... 475

BOS Hosting Details ............................................................................................................................ 476
The DevForce Client ......................................................................................................................................... 478

Vista Setup ........................................................................................................................................... 479
Vista setup requirements for the ServerConsole or ServerService .................................................................... 479 Vista setup requirements for IIS ........................................................................................................................ 479

Troubleshooting ................................................................................................................................... 480
Worked in 2-Tier, Strange Errors in n-Tier ....................................................................................................... 480

Disconnected Applications......................................................................................................... 482
Running Offline ................................................................................................................................................. 483 Securing Offline Data ........................................................................................................................................ 492

Security ....................................................................................................................................... 497
Authentication ................................................................................................................................................... 497 Authorization ..................................................................................................................................................... 500 Encryption ......................................................................................................................................................... 502 ASP.NET Security Integration .......................................................................................................................... 502

Deployment ................................................................................................................................. 506
Document Overview ............................................................................................................................ 507

6|P age

IdeaBlade DevForce Developers Guide

Contents

DevForce And the App.Config File.................................................................................................... 507
Creating and Editing a Configuration File ......................................................................................................... 508 IdeaBlade DevForce Configuration Editor ........................................................................................................ 508 DevForce Elements in App.Config .................................................................................................................... 510 Configuration File Location .............................................................................................................................. 511 Client and Server Versions of App.Config ........................................................................................................ 512 Probing in DevForce .......................................................................................................................................... 513

Data Server Deployment ..................................................................................................................... 516 Deploying a DevForce Silverlight Application.................................................................................. 516
Deploying to IIS Version 6 ................................................................................................................................ 517 Deploying to IIS Version 7 ................................................................................................................................ 519 Troubleshooting ................................................................................................................................................. 522 Resources ........................................................................................................................................................... 523

Deploying a DevForce WinClient Application.................................................................................. 523
Overview ........................................................................................................................................................... 523 Deploying a Single-Tier WinClient Application ............................................................................................... 526 Deploying Two-Tier (Client-Server) WinClient Applications .......................................................................... 526 Deploying N-Tier (Smart-Client) Applications ................................................................................................. 527 Building Blocks ................................................................................................................................................. 527

Troubleshooting ......................................................................................................................... 547
General Troubleshooting .................................................................................................................... 547 Troubleshooting Silverlight Apps ...................................................................................................... 548 Contacting Support ............................................................................................................................. 551
Identifying your DevForce version .................................................................................................................... 551

Upgrading Your Software .................................................................................................................. 553

7|P age

IdeaBlade DevForce

DevForce and the Entity Framework

DevForce, Enterprise Applications, and the ADO.NET Entity Framework
DevForce, Enterprise Applications, and the ADO.NET Entity Framework
The Problem Object Mapping Technology The Microsoft ADO.NET Entity Framework

Using DevForce with the Entity Framework
Advantages of Using DevForce

DevForce in More Detail
Advantages of Using DevForce (Revisited) More DevForce Advantages

Conclusion

DevForce is a framework for building and operating multi-tier, data-driven enterprise applications. By “enterprise application” we do not mean simply a big application, or an application for a big company. Rather, we refer to an application with the following specific characteristics:      Its users devote many hours to its use, performing task essential to conducting the organization‟s business. It requires a rich and responsive graphical user interface, dense with sophisticated controls User interactions are complex; task and context switching is common. It presents data that are complex in themselves, and deeply interrelated. The data are stored centrally and shared with other users.

Supply chain, customer relationship management (CRM), and asset tracking applications are typical examples. User productivity is critical. That puts a premium on the application‟s ability to provide a highly responsive, richly featured user experience – the kind of experience typical of a desktop application running directly on a client machine. We expect people to get work done at any time from anywhere. Those people may be employees or they may be valued partners. In either case, security matters. Accordingly, we often need to deploy and operate enterprise applications over a wide area network – preferably over the internet – with undiminished productivity and security. DevForce is especially suited to building and running applications that require a rich user experience delivered to remote, Internet-connected clients. While DevForce contributes at all levels of the enterprise application architecture stack, its Object Relational Mapping (ORM) technologies and object-oriented approach to data management draw most of the attention. Microsoft has stepped into this arena with the Language Integrated Query (LINQ) and the ADO.NET Entity Framework, both released with version 3.5 of the .NET framework. The Entity Framework is a robust ORM solution; the developer can retrieve data as “entities” by writing “LINQ to Entities” statements in her preferred .NET programming language. DevForce delegates to the Entity Framework the mapping between object and relational database schemas, as well as the database persistence operations (queries and saves). These are important and challenging tasks that the Entity Framework handles well.

8|P age

IdeaBlade DevForce

DevForce and the Entity Framework

There is much more to an application than how it handles raw data. There is the business object layer that encapsulates the data and governs those data with business rules. There are higher layers that address the application workflow and user experience. All of this is outside the purview of the Entity Framework. If we concentrate only on data management, we still find enterprise application requirements untouched by the Entity Framework. Chief among them are:      Central services, Internet connectivity, distributed transactions, performance, security, scalability, and Silverlight support - needs best met with an intelligent middle-tier server. Highly responsive client UI‟s that exploit caching to avoid redundant, slow trips across the wire. Object models mapped to multiple data repositories Objects mapped to Web and WCF service data sources Proper support for a business object layer with business rules.

DevForce satisfies these requirements even as it relies on the ADO.NET Entity Framework for basic ORM and query facilities. The key components of DevForce include:     the Entity Manager, which includes a queryable client-side cache; the Business Object Server (BOS) for services in a middle tier; a provider for the LINQ language that permits LINQ queries to be used with both the client-side cache and remote data sources the Object Mapper which extends the ADO.NET Entity Framework designer and generates DevForce entity code.

This chapter explores the key data management issues for .NET enterprise application developers. It introduces the LINQ and the ADO.NET Entity Framework, explaining what they do and where they leave off. It then describes how DevForce fills in the critical gaps.

The Problem
Every business application is an extended dialogue between a user and the business objects that fulfill the application‟s purpose. Those business objects are behavioral objects first and foremost. They are the embodiment of the customer stories that describe what the application does and how it does it. A few behaviors may be stateless; financial calculations come to mind. But there is usually data somewhere in those business objects. An order has a customer and a delivery date and line items describing quantities of goods sold for a price. There is no escaping the data aspect of business objects and all of that data must be managed. While the application is running, the data are held in session in some form. In an object-oriented system they are held in fields and exposed as properties of a class instance. But because the data are long-lived – longer-lived than any one session – they have to be saved between sessions. And because we share our data with others, we have to save the data in permanent storage accessible over a network. Shuttling data between storage and the application session is one of those necessary but “dirty” jobs, a job completely unrelated to the application‟s purpose. Developers long ago discovered three data management problems. First, the way we store data is not the way we use data in an application. Money, for example, is both an amount and a currency (dollars, euros). The two aspects require separate slots in storage; from the application perspective, it‟s just one thing: money. An “order” in the context of an application session may be seen as one “thing” with a customer, a shipper, line items, etc. When we store that order in a relational database, the order, customer, shipper and line are five different things. So the best representation of stored data often is not the best representation for session data. Second, session data are governed by rules. We must know the customer for an order before we can deliver the ordered goods. The date of the order should precede the delivery date. Some other part of the application may need

9|P age

IdeaBlade DevForce

DevForce and the Entity Framework

to be alerted when the order is actually delivered. The application is more maintainable and easier to understand when the rules (behavior) and the data are bound together as “business objects” or “entities”. Such rules are largely irrelevant when the data are tucked safely away in storage. Third, there are many mechanical matters surrounding saving and retrieving data that have nothing to do with the application‟s purpose such as opening and closing connections, composing SQL, detecting concurrency violations, converting raw data into Data Transfer Objects, and managing transaction boundaries. Getting the application dialogue right is hard enough without these distractions. Yes, the application still has to ask for data and stow them away. But there should be a way to express our intent simply and entirely in terms of the application entities. Ordinary operations should make no mention of databases, connections, tables, or columns. The profound differences between stored data and session data lead developers to expend enormous energy moving and translating between stored and session representations. This is wasted energy from the perspective of the application customer who could not care less about our implementation problems. It is also wasted energy from the developer‟s perspective because this problem has been solved by Object Mapping technology.

Object Mapping Technology
An object mapping technology maintains two views of the data. There is a conceptual model for representing the data within the entities used by the application and there is a storage model that defines how the data are stored in the repository. These two models have completely different characteristics, as we have seen. The conceptual model could include a conceptual order, an order entity, as it is understood by the application. The storage model describes how the order entity‟s data values are held in the data repository. If the repository is a relational database, many of the order entity data values – its state – are likely held in columns of a table. The value of a DeliveryDate property of an Order entity might be stored in the [DeliveryDt] column of an [OrderHeader] table row. The correspondence between the conceptual order entity and the table row is obvious and strong in this example. Even so, the correspondence is not literal; there is Order and DeliveryDate on one side; OrderHeader and DeliveryDt on the other. Therefore, the object mapping technology maintains a “map” of the correspondence between entities of the conceptual model and the table rows in the storage model so that it can transform one representation into the other. The Order entity has a related Customer entity and related OrderDetail entities. These additional entities might correspond to Company and OrderLineItem tables in a relational database. Relational databases objects don‟t have relationships. They have foreign key constraints that imply these relationships. Accordingly, the object mapping technology also maintains a map of the associations between entities and the foreign key constraints in the database. The map records the pairing of the relationship between Order and Customer with the foreign key constraint between the OrderHeader and Company tables. This order example is especially simple. Other mappings could be enormously complex, with values changing shape (type), entities splitting among multiple tables, and relationships weaving through intermediate association tables. Without an object mapping facility, the application developer would have to be constantly aware of these correspondences as she wrote instructions to retrieve and save application data. Small changes in the actual storage schema or in the application entity model could easily break the code in a hundred places. Without an object mapping facility, the application would become vulnerable and brittle as it grew and aged. Productivity would fall as developers devoted increasing effort to keeping the conceptual and the storage models aligned.

The Microsoft ADO.NET Entity Framework
The Microsoft ADO.NET Entity Framework is one Object Mapping technology to consider. DevForce builds upon the Entity Framework, so we introduce the Microsoft technology here before explaining DevForce‟s added value.

10 | P a g e

IdeaBlade DevForce

DevForce and the Entity Framework

Read more about the Entity Framework at http://msdn2.microsoft.com/en-us/library/bb399572(VS.90).aspx

Entity Data Model (EDM)
The Entity Framework supports an Entity Data Model (EDM) that describes data from the application perspective. The EDM does not include the actual business object classes that contain those data; rather it defines certain of the data and data relationships within those classes in an implementation-agnostic language of its own. Concretely, the EDM is an XML schema file that defines a conceptual data model. That schema is accompanied by two other XML schema files: one describing how the data are stored (the storage model) and another that maps the conceptual model to the storage model. The Entity Framework uses this chain of descriptions to move data between the data-laden objects in memory and the actual data repositories. For this to work at runtime, the conceptual schema (the EDM proper) refers to entity classes of the application while the storage model gets matched up, via configuration, with a real database running on a server somewhere.

The Entity Data Model Designer
Most developers prefer to use a tool to work with XML rather than edit XML by hand. EDM XML is dense and forbidding so a tool is a practical necessity. The ADO.NET Entity Data Model Designer is a Visual Studio design tool that provides the developer with a graphical, drag-and-drop EDM design experience. The designer enables simultaneous development of all three related schemas – the conceptual, storage, and mapping schemas. Most applications are predicated on a pre-existing database. This database cannot be ignored; the conceptual model must ultimately come to terms with it. Most developers find it convenient to confront this fact early and will prefer to generate the conceptual data model and associated schemas using the Entity Data Model Wizard. The wizard produces the EDM schemas which then can be viewed and edited in the designer.

Entity Object Layer
The Entity Framework business object layer consists of the classes that implement the application business objects. The Entity Framework includes an entity class generator that uses the EDM to produce class code that defines the business object data fields and their accessor properties. It also generates the navigation properties that enable the application to traverse from one object to its related objects (e.g. from an order to its customer). The EDM describes only the business object data and their relationships. The Entity Framework knows nothing about the business object behavior that applies to the data so there is no business logic in the generated code. The application developer writes business logic separately in a companion class file. The two files – the developer‟s business logic file and the generated object data management file – combine to form a single definition of the business object, the business object class. Technically, each file defines a .NET partial class. The compiler knits the two together, resulting in the complete business object class.

Entity Persistence
The Entity Framework includes components responsible for moving business object data between the application and the database.

11 | P a g e

IdeaBlade DevForce

DevForce and the Entity Framework

The ObjectContext is the most visible of the components. The application uses ObjectContext to retrieve, hold, and save entities. The ObjectContext maintains a cache of all the entities it manages. The developer writes queries and submits them to the ObjectContext, which retrieves the selected entities and adds them to its cache before returning them to the caller. The developer creates new objects and adds them to the ObjectContext. The ObjectContext tracks changes – adds, modifications, deletes – to entities in its cache. A save command tells the ObjectContext to write the changed entities to the database. The Entity Framework handles all of these relational database persistence operations without troubling the developer with details. The Entity Data Model and a few guiding parameters are all it needs.

LINQ to Entities
Earlier we described three problems for the developer who needs to represent data in the application as business objects. The third problem was how to retrieve and save business objects using a language that hid the underlying mechanisms and stayed true to the entity-oriented paradigm. While the mechanics of saving business object data are challenging, it has never been difficult for developers to express their intent. It is usually sufficient to tell some service class to “save” and the service knows what to do. Getting data is a different story. It is not easy to say precisely which data you want, and in what form, using a general purpose programming language. It‟s harder still to write queries in a strongly-typed manner and stay within an entity-oriented paradigm. Until recently, object mapping vendors offered their own “object query languages” (OQLs) which were, in fact, merely special purpose classes with strangled interfaces. OQL queries were clumsy to write and repugnant to read. With its release of the .NET 3.5 Framework, Microsoft added new language facilities for finding and accessing data in a general purpose, object-oriented way, without exposing the details of data storage and retrieval. Chief among the new features is LINQ, an abbreviation of Language Integrated Query. A LINQ query looks much like an SQL query. Most programmers have long experience with SQL so, while SQL itself may be tortured, most programmers are accustomed to it and find LINQ expressions familiar:

C#

IQueryable<Product> products = from prod in anObjectContext.Products where prod.ReorderLevel > 100 select prod; foreach (Product aProduct in products) {…}

VB LINQ defines a set of query operators for interrogating arbitrary sources of data. Anything that can be enumerated can be queried with a LINQ expression. We can use LINQ to select items from a list, nodes from an XML file, file names from a file folder, or records from a database. LINQ itself does not know how to do any of these things. LINQ defines the query operators and patterns for writing query expressions. The operators and expressions are meaningless until they are married to an implementation that is specific to a domain. Thus there is a LINQ implementation for querying in-memory objects (LINQ to Objects), an implementation for querying XML structures (LINQ to XML), an implementation for querying relational databases (LINQ to SQL), and so on. Microsoft provides some of these implementations but third parties can develop their own and Microsoft encourages them to do so. The LINQ facility provides the expressiveness we need for querying entities. What we need is a LINQ implementation that supports an object mapping technology. Microsoft‟s LINQ to Entities is that implementation for the Entity Framework.

12 | P a g e

IdeaBlade DevForce Entity SQL

DevForce and the Entity Framework

The Entity Framework supplements LINQ to Entities with its own query language called Entity SQL. Entity SQL is a storage-independent dialect of SQL that works directly with the conceptual model. An Entity SQL query refers to entities, properties, and associations (e.g. Order and Order_Customer) rather than the database elements in the storage model. The particulars of data storage remain hidden in the object-oriented data design. Entity SQL queries are strings as seen in this example: C#
string queryString = @"SELECT VALUE Product FROM Products “ + AS Product WHERE Product.ReorderLevel > 100"; ObjectQuery<Product> products = new ObjectQuery<Product>(queryString, anObjectContext); foreach (Product result in products) {…}

VB One significant drawback: Visual Studio will not detect even simple mistakes because the query string won‟t be evaluated until runtime.

Using DevForce with the Entity Framework
Microsoft‟s ADO.NET Entity Framework is a solid foundation for object relational mapping and relational database persistence operations. LINQ to Entities is a huge advance over SQL string commands and proprietary object query languages. We covered this same territory in our earlier, .NET 2.0 version of DevForce; we are pleased now turn over some of these responsibilities to the Entity Framework for applications built on the .NET 3.5 platform. DevForce provides an alternative Entity Data Model editor, the DevForce Object Mapper, which is used for four main reasons:     to augment the EDM schemas with DevForce-specific XML to generate the DevForce business object classes which extend the Entity Framework classes to work with a tabular interface that is more productive for larger (>20 class) object models for more granular control over the generated class and property code

The Object Mapper plugs into Visual Studio and the developer can switch freely between the Object Mapper and the Entity Framework designer, choosing the one that is most productive for the task at hand. DevForce relies upon the Entity Framework for the persistence operations that target relational databases. The Entity Framework prepares and issues the actual vendor SQL. The Entity Framework issues all insert, update and delete commands and employs optimistic concurrency techniques to detect collisions between updates of the same object by different users.

13 | P a g e

IdeaBlade DevForce

DevForce and the Entity Framework

Advantages of Using DevForce
The ADO.NET Entity Framework does a good job of handling relational database mapping and persistence operations for client / server applications. However, most enterprise applications need better data management and better support for developing the business objects that encapsulate the relational data. DevForce provides essential improvements in such critical areas as:        Infrastructure for multi-tier applications Security Client application performance Model design and code generation Multiple data sources Web Services Intermittently connected and offline apps

We summarize each point in the balance of this section.

Infrastructure for Multi-Tier Applications
The ADO.NET Entity Framework only supports a 2-tier architecture in which the client machine speaks directly to a relational database server. This won’t work for many enterprise applications, especially those that      Connect to servers over the Internet, a wireless, or a wide area network. Require rigorous security. Must scale to support many users, especially external partners and customers. Offer applications On-Demand (Software-as-a-Service). Will deploy as a Silverlight application in a browser.

Such applications require the performance, security and scalability of an intelligent middle tier server that mediates between client machines and such server-side resources as databases and web services. DevForce implements an end-to-end, multi-tier (n-tier) architecture whose middle tier component is called the “Business Object Server” (BOS). DevForce is the only way to bring n-tier capabilities to LINQ- and Entity Framework-based applications.

Security
ADO.NET Entity Framework has no intrinsic security features. Because of it two-tier approach, the security burden falls entirely on the network and the database. That may be sufficient for simple applications with few users who are always connected within the company LAN. But we will need a better answer when authentication and authorization schemes become fine grained and application specific, when the number of users grows, and when some of those users are reaching in from outside the company walls. The DevForce n-tier solution supports a rich variety of standard and custom authentication techniques and provides encryption and authorization points on both client and server.

Client Application Performance

14 | P a g e

IdeaBlade DevForce

DevForce and the Entity Framework

Data access is the number one performance killer. Large volumes of data are deadly. Frequent trips to the server are worse. And it‟s really bad if the UI freezes while waiting for data. Responsiveness and user productivity improve dramatically when we eliminate unnecessary trips, reduce the size of data traveling over the wire, and retrieve data asynchronously. None of this is easy to implement. The ADO.NET Entity Framework is a purely 2-tier architecture in which the client talks SQL to the database, a chatty conversation with few means to shrink the data. It doesn‟t remember previous queries, we can‟t query its primitive entity cache, and we can‟t query asynchronously. A DevForce application deployed in n-tier mode represents business object data in a compact form and compresses the data before sending it resulting in smaller payloads over the wire. Smaller payloads, faster app. Most applications ask for the same data over and over. DevForce has a query-able entity cache and a query cache. We can ask the entity cache any question, including questions we‟ve never asked before. The query cache remembers previous database queries so repeated questions don‟t cause redundant server visits. In fact, we use DevForce to Entities, a LINQ-based query language, to pose questions that can search the cache, search the data source, or search both as we wish. Finally, DevForce offers asynchronous queries that can hide the actual cost of a remote query as perceived by the end user. The UI continues to function and we can occupy the user‟s attention with an initial set of data while the balance is retrieved in background.

Model Design and Code Generation
The ADO.NET Entity Framework design tools and code generation are not as strong as they need to be for enterprise-scale applications. The drag-and-drop designer becomes unwieldy with modestly sized domain models; class diagrams with more than 20 objects are almost impossible to read or manage. The developer has little control over the mapping and the generated code doesn‟t support common business behavior scenarios such as validation, property-level security, value and message localization, and change auditing. The DevForce Object Mapper can read and write EF schemas. Its utilitarian interface targets medium to large models (20 to 2,000 entity types). It‟s easy to determine how data are exposed as classes and properties and it‟s easy to grow the model as requirements change. It can generate classes from storage schema but it also tolerates conceptual class development in advance of storage mapping. The generated code is designed for augmentation with business logic so we can build business objects instead of property bags.

Multiple Data Sources
The ADO.NET Entity Framework supports just one database per Entity Data Model. But many application data models draw from data storied in multiple data sources. In a supply chain application, orders may be stored in an inventory database while ledger entries are captured in an accounting database. Orders and ledger entries have keys that, conceptually, enable navigation between them even though a cross database query is not technically possible. In DevForce we can define a single model that holds both orders and ledger entries and the code generator can produce “navigation properties” for seamlessly navigating between them. DevForce handles the SQL for simulating the cross database “join”. Order and ledger updates must be saved transactionally. DevForce can perform such distributed transaction; the Entity Framework, knowing only one database, cannot.

Web and WCF backed Business Objects
Web services and WCF services are increasingly important sources of application data both as front ends to legacy databases and as the preferred modality for accessing Internet resources.

15 | P a g e

IdeaBlade DevForce

DevForce and the Entity Framework

The ADO.NET Entity Framework can only map entities to relational databases. DevForce can map business objects to relational databases, web services, and WCF services, all in a single consolidated model.

Intermittently Connected and Offline Applications
ADO.NET Entity Framework applications are vulnerable to temporary connection failures. There is no effective way to recover from a query or save that fails because the connection or server is unavailable. There is no intrinsic solution to “the airplane problem” – the application that must be able to launch and run offline as when working while in flight. DevForce applications have the means to survive transient connectivity and to thrive offline.

DevForce in More Detail
We highlighted the most significant DevForce differences in the previous section. Here we explain them in greater detail and cover some of the other important DevForce features that improve application design and developer productivity.

Advantages of Using DevForce (Revisited)
Multi-Tier Applications
The ADO.NET Entity Framework is a client / server technology. It’s ObjectServices component, which is responsible for querying and saving data to the database, executes in the same process as the client business object layer. Database SQL commands and raw data flow over the wire.

This works just fine when there are relatively few clients, all connected to a secure, high speed LAN. Performance becomes a serious problem when the traffic goes up or when going over a wide area network. There‟s a lot of back-and-forth talk when SQL passes over the network and the data are verbose. With reduced bandwidth and increased latency, those frequent roundtrips for data that no one noticed before become serious problems and the user experience slows to a crawl. Furthermore, in order for a two-tier application to work over the internet, you would have to expose your database directly to the world. This opens up the possibility of someone stealing the connection string and browsing or changing your database without authorization.

16 | P a g e

IdeaBlade DevForce The DevForce n-Tier Solution

DevForce and the Entity Framework

The DevForce n-tier solution, with its “Business Object Server” (BOS) deployed in a middle tier, overcomes all of these obstacles.

The ADO.NET Entity Framework has relocated from the client arena to the Business Object Server where it now functions purely as an object mapping technology, translating persistent data between entity and storage representations. The client application hosts the DevForce Entity Manager, a component responsible for holding business objects in cache and communicating with the BOS. The business objects and the Entity Manager itself are completely decoupled from the ADO.NET Entity Framework. There are no references on the client to any of the Entity Framework assemblies. Nor do clients talk to the database. Instead, the Entity Manager sends commands to the BOS and receives business objects in return. Commands may be expressed in a variety of formats including the new LINQ to DevForce query language. The BOS translates a LINQ to DevForce query into a LINQ to Entities query and submits it to the Entity Framework. The Entity Framework returns simple entities to the BOS which forwards them to the client. DevForce on the client turns them into business objects and caches them in the Entity Manager. The BOS and the client DevForce Entity Manager exchange data in a serialized binary form that passes easily through firewalls and over the Internet. The BOS compresses the data before sending them to the client. These smaller payloads reduce network traffic and improve client performance. The BOS is effectively stateless. It retains no essential information about client sessions between requests. Each client request resolves to a method call running on a new thread; the call holds onto entity data just long enough to fulfill the request after which it is discarded. Such statelessness makes it easy to distribute requests among multiple BOS servers for scalability and fault tolerance.

Remote Services
Some applications require services that must execute in a centrally hosted environment, perhaps because they involve proprietary logic or because they crunch volumes of data that would swamp the network if transmitted to clients. A client can make a “remote service call” to the BOS, which will invoke custom server side methods to perform or delegate these hosted services. The BOS can watch for server-side events such as data updates or network notifications, and publish corresponding events to subscribing clients through its “push” service.

17 | P a g e

IdeaBlade DevForce DevForce Silverlight 1

DevForce and the Entity Framework

Features described in the section are included with the DevForce Silverlight product. Microsoft Silverlight enables deployment of .NET applications within a browser. There is no application to install, no client footprint, and no compromise of the client machine‟s security. The door is open to deliver applications to consumers and locked-down enterprise environments securely. Data access remains a challenge. Data-driven Silverlight applications need access to the same data as their desktop equivalents. A Silverlight application can only reach data resources over the Internet and, as we‟ve seen, the ADO.NET Entity Framework cannot move data over the Internet. But a DevForce Silverlight application can. In 2009, IdeaBlade will release “DevForce Silverlight” supporting SilverLight applications that are based on the same rich object model deployed in DevForce WPF smart client applications.

In Summary
With the DevForce n-tier capability,      The Entity Framework becomes an n-tier platform Business object data can travel through firewalls and over the Internet Data are compressed and encrypted for fast, secure transport The client can request non-data services to be executed on the server and subscribe to server events. A software vendor can offer software-as-a-service to its Internet customers.

With the DevForce Silverlight product, you get all of the above capabilities in a tool that permits you to develop Silverlight applications that use the Entity Framework in the same way that WPF Windows clients do.

Secure Services
The Entity Framework only supports a two tier architecture in which the client talks directly to the database. There are not intrinsic capabilities for authenticating users, authorizing access, or encrypting data. This architecture relies entirely on coarse grained network and database measures to secure the application and requires extra care to protect the client machine from theft or intrusion. This level of security is not good enough in many environments. There may be tough corporate or legal mandates to protect sensitive data from unauthorized access. A client machine could fall into mischievous hands. Any .NET program is easily disassembled. A determined malefactor could discover the client-side application security measures, develop counter measures, and attempt unauthorized persistence operations.

Connection Security
The trouble begins with the database connection string. In a two-tier world, each client must provide the Entity Framework ObjectContext with a database connection string before it can access the database. The database is easily compromised if the string contains a user and password. Encrypting the string until the moment of use certainly helps – if you remember to do so – but still amounts to security-by-obfuscation. It is much safer to rely on the operating system to authenticate the user to the database via the Security Support Provider Interface (SSPI) as when the MS SQL Server connection string specifies “Integrated Security=SSPI;”.

1

DevForce Silverlight and DevForce WinClient are separate IdeaBlade products. They are combined in the DevForce Universal product.

18 | P a g e

IdeaBlade DevForce

DevForce and the Entity Framework

Moreover, each database connection is unique, defeating the performance advantage of connection pooling. This technique works but there are problems. The IT management burden grows heavy when there are many application users scattered across a widespread corporate network. New users must be added both to network directories and to the database‟s own list of authorized users. Departing employees should be removed from all directories. The application administrator rarely maintains the network and database logons so there are communications breakdowns that lead to mistakes. In a DevForce n-tier deployment, the Business Object Server (BOS) stands between the client and the database. The client must login to the BOS before the BOS makes any requests on the client‟s behalf. After login every transmission from client to server is accompanied by an encrypted session token that identifies the client. NT Authentication and impersonation are viable alternatives for LAN users and can be combined with alternative login mechanisms when users access the application from outside the corporate network. Clients no longer access the database directly. They don‟t hold a connection string nor issue vendor SQL calls. They don‟t know where the data physically reside. Instead they ask the BOS to fetch and save data on their behalf and only commands and object data travel over the wire. The BOS, running on a secure machine, connects to the database with its own private connection string. The BOS performs all database operations.

Authorization
The ADO.NET Entity Framework has no authorization mechanisms. In most cases, the application relies upon authorization settings in the database – settings which operate crudely at table levels and do not reflect more detailed business rules. Application-specific authorizations can only be enforced in the client. The ability to limit order approval or restrict access to a patient record depends entirely on business logic executing in the client. With the DevForce BOS in place every query and save operation is subject to inspection. The BOS invariably calls certain customizable secured operation methods, passing along the client‟s Principal so each method can identify the client user and his assigned roles. The method can determine if the user is allowed to perform the requested operation and what action to take if permission is denied. Every step in this process, from login to security check can be tailored to meet the particular needs of the application. There is nothing that client can do to thwart these measures. The BOS will execute them like clockwork and the client has no access to the server, no ability to inject malicious code.

Encryption
The developer is free to engage the kind of encryption that is most appropriate. SSL is typical but other methods can be inserted in the pipeline. DevForce prefers to use Microsoft Windows Communication Foundation (WCF) for client-to-server communications; the WCF security-related configuration options are all available.

Client Performance with DevForce Caching
Fulfilling a request for data with a trip to the database is thousands of times slower than satisfying the same request from local memory. The trip is longer still when the database resides across the network. That‟s why responsive, data-intensive business applications cache entity data locally. If we‟ve asked for the data before, we should not have to ask for the same data again – at least not immediately. The ADO.NET Entity Framework caches entities and can look up an object by key rather than go to the database. That can be a big time saver – unless the entity isn‟t in cache! The Entity Framework returns null if it can‟t find the

19 | P a g e

IdeaBlade DevForce

DevForce and the Entity Framework

object. Maybe the entity doesn‟t exist. Maybe it just hasn‟t been retrieved yet. The Entity Framework can‟t tell. We‟d have to query for the object – or explicitly load it – to be sure. Unfortunately, we can‟t simply query the Entity Framework cache directly. All Entity Framework queries (and loads) reach across the net to the database – even repeat queries issued mere moments ago. Do applications ask the same question twice? Yes they do. Users are always cycling among several active tasks; each time they return to a task underway, the application re-issues a query. The developer might take pains to cache such queries herself. But that‟s an arduous and error prone pursuit best left to the DevForce framework.

DevForce Caching and LINQ to DevForce
The DevForce Entity Manager maintains a query-able, client-side entity cache. By “query-able” we mean that we can always apply a LINQ to DevForce query to the in-memory cache. LINQ to DevForce is a LINQ implementation that enables queries to both the entity cache and to remote data sources. Let‟s look at an example. We want to see the orders of star sales rep, “Nancy Davolio.” We compose a LINQ query that searches for orders of the rep who‟s first name = “Nancy” and whose last name = “Davolio.” The first time we run it, the Entity Manager realizes that the query is new and sends the query over the wire to the BOS. The results come back after a fraction of a second or several seconds, depending upon the amount of data, the load on the database, and the speed of the network. A minute later we ask for Nancy‟s orders again. The Entity Manager recognizes the repeat query and looks only in the local cache. It returns with the results immediately. Behind the scenes the DevForce Entity Manager maintains both a cache of entities and a cache of queries. The query cache is the memory of queries run against the database. When DevForce executes a LINQ to DevForce query it checks this query cache first. If it finds the query it assumes the query can be satisfied by the entity cache. It then translates the LINQ expression tree into search operations against that cache. The developer can inspect, add, remove, clear, and update the contents of both the entity and query caches.

Responsiveness with Asynchronous Queries
Responsiveness is subjective. The application is fast or slow if the user thinks it is. Users worry if the application freezes for more than a second. A prolonged delay when the application launches or a heavy screen loads is a common cause for complaint. Initialization queries or big data transfers are often the source of the problem. You can alleviate the pain by fetching the data in background with asynchronous queries. The Entity Framework does not support asynchronous queries. DevForce does. It is easy to fire off a series of async queries before displaying a form on screen. The form appears immediately and fills as the data arrive. Some entities are more volatile than others. The list of provincial and city tax rates is probably constant during a particular session. Inventories, on the other hand, are changing constantly and screen full of quantities on hand should probably be refreshed every few minutes (or seconds perhaps). DevForce async queries on a timer can keep that screen current without stalling the UI while the application polls for changes. There is always the danger of a runaway query – the query that pulls down so much data that it either freezes the UI for agonizing minutes or times out. Fortunately, it‟s easy to use the LINQ extension method Take() to pull down sequential sections of a collection of entities. The following query, for example, will bring down the first 100 customers, ordered by the name of their company:

C#

var query =_mgr.Customers.OrderBy(c => c.CompanyName).Take(100);

20 | P a g e

IdeaBlade DevForce

DevForce and the Entity Framework

We can make the application appear extremely fast by combining a Take() query that requests a small set of data with an asynchronous query that requests a larger set. Suppose, for example, that the user requests several thousand orders. We don‟t know for sure he‟ll do so, but we‟ve seen it before. So we take defensive measures. We first compose the user‟s order query in the usual manner. We then suffix it with a call to Take() that limits the request to a safe maximum of 3,000 orders. We submit this one as an asynchronous query because we know from experience that it will take several uncomfortable seconds to return. We follow immediately with the same query, also suffixed with a call to Take(), this time limited to 100 orders. This one we submit synchronously2; we‟re willing to wait a half second for this one. It returns as a list and we present the first 100 orders. The original request for 3,000 eventually arrives; the call-back method fills the list. On screen, the order grid magically grows from 100 to 3,000. The user is delighted. Note that there is also a Skip() extension method that can be used if you want something other than the first n members of an ordered result set. The following query will bring down the next 100 customers: C#
var query = _mgr.Customers.OrderBy(c => c.CompanyName).Skip(100).Take(100);

Model Design and Code Generation
Architects are increasingly convinced that we should design business objects with a blind eye to the way their inner state are stored. Our job is to interpret the user stories, to tease out the logic and data necessary to support those stories. A business object model gradually emerges and from that model we later discover the storage scheme that fits best. This approach is called Behavior Driven Development (BDD) because it encourages us to start from the required application behavior and work toward the implementation rather than leap directly to data design (as most of us old folks have done our entire careers). If a user story says “the order date must precede the delivery date”, it is clear we‟ll need two date fields. When the story says “the user enters an order date” and “the user enters a delivery date”, we will know enough to give our Order class properties to get and set these dates. On the other hand, our Order class won‟t have an “approval date”, a “credit checked date”, a “status changed date” or any other date unless another user story calls for them. No peeking to see if these fields are in the Order table! We won‟t worry just yet about how or where the order and delivery dates are stored. BDD says we should wait to the “last responsible moment” before committing to a storage scheme. Meanwhile, we can code and test our Order class now. As storage blindness is rarely possible in real life, we should at least hang a curtain to hide the storage details – and peer behind that curtain as little as possible. The Entity Data Model helps by separating the conceptual data model from the storage schema. There is no mistaking the fact that the conceptual data model remains a data model – well short of a business object layer whose members combine behavior and state. Moreover it exists for one reason only: so that we can move values between business objects and storage when that time inevitably arrives. So it is actually a model of the state within the business objects rather than a model of the business objects. Nonetheless, we should be able to maintain the pretense that our state is purely conceptual and could be moved to any form of storage. We only commit to a storage scheme when we‟re in a different frame of mind. This kind of design separation is extremely difficult to accomplish by hand. There is a lot of tedious programming for each business object, most of it concerning access to the fields of persisted data. An object mapped to a table row
2

In Silverlight applications all queries must be asynchronous, so in that case we will have to do both of our queries – the larger one and the smaller one – asynchronously. In a Silverlight app, we might choose to tie up the user interface of our application by other means (such as displaying a child window) while waiting for the smaller query to return.

21 | P a g e

IdeaBlade DevForce

DevForce and the Entity Framework

of twenty columns could yield a couple of pages of code. The slightest change to the storage schema necessitates a revision of this code. We won‟t do it without adequate tools and code generation. We would simply lack the patience and discipline.

ADO.NET Entity Framework Development
The ADO.NET Entity Framework takes a stab at the appropriate tooling and code generation. There is a Visual Studio EF designer that presents a visual canvas upon which to draw entity classes, the relationships among them, and the mapping to the storage schemas. The EF code generator produces a partial class file with properties to access persisted data fields. It also inscribes navigation properties that return related entities; the Order.Customer property returns the Customer object associated with a given Order instance. Because a business object is more than data and needs more logic – more behavior – than just data access properties, the generated class file needs a companion partial class file. The design tool can‟t generate the companion file – only the developer knows what belongs there. The developer creates this file and pours her custom business object behavior into it. The compiler combines the two files, yielding a complete class with both business logic and data management capabilities. This two-part, “bicameral” file structure is effective in keeping developer and generated code in separate rooms. In principle the generator can be run repeatedly – rearranging the generated code “room” – without disturbing the furniture in the developer‟s room.

Weaknesses
The Entity Framework designer and generator fall short in several critical respects:    The designer does not give the developer adequate control over the generated code The generated properties are not adequately extensible, limiting the developer‟s ability to abstract out the business logic shared across business objects. The code generator blocks introduction of “base” classes into the inheritance hierarchy, limiting the developer‟s ability to inherit common business object behavior.

Designer Woes
We could write the properties by hand. But we‟d like to use the Entity Framework Designer to generate the code for us so that the property code conforms to standard and includes all the property interceptor calls it should have. Unfortunately, the Entity Framework Designer won‟t generate entity code without validating the storage model and the mapping schema – which don‟t yet exist. The DevForce Object Mapper can generate the tedious persistent data accessor property code with the conceptual model alone. It doesn‟t need the storage or mapping schemas which we can fill in later. The developer should be able to build the conceptual data model without first committing to a storage or mapping specification. When the developer determines that the application data requirements are sufficiently well known to warrant database schema design, she can add the storage and mapping schemas “just in time.” Unfortunately, the EF insists that every conceptual entity be mapped. It refuses to “validate” the model when there is no mapping and it won‟t generate code for an un-validated model. The developer should work on just those business objects that are pertinent to the user story. Who cares if the database has hundreds of tables when we only need five business objects. Sadly, the EF designer is unforgiving. There is no going back once the developer has selected his tables and generated her model. She will have to edit the XML to add the sixth, seventh, and eighth objects.

22 | P a g e

IdeaBlade DevForce

DevForce and the Entity Framework

Indeed, there are a great number of everyday mapping activities that can only be accomplished by dipping into the raw XML.

Rigid Code Generation
The Entity Framework code generator grants the programmer only limited control over the generated class code. For example, it emits public properties for all mapped data values, even those you don‟t want exposed. And it always generates properties with both getters and setters. This is a reasonable default but is not desirable in every case. The primary key value is usually immutable; its property should be read only if it can be read at all.

Anemic Data Properties
The bicameral approach works fine when we can locate the business logic in the developer‟s custom partial class file. It‟s easy to put calculations and workflow rules there when they concern the entire object. For example, this is the place to augment the order object with an InvoiceTotal property that sums the cost of all item details. But a great deal of business logic is only effective when it executes inside the data access properties – and these properties reside in the generated file. Suppose we want to constrain the transition from one order status value to another; perhaps the status proceeds from “new” to “approved” to “shipped” to “delivered”. We should reject any attempt to transition directly from “new” to “shipped”. Maybe we should block unauthorized users from changing the status at all. The critical place to catch validation and security violations is inside the OrderStatus property itself. The EF did not generate the OrderStatus property with these capabilities. We cannot add them to the generated property code ourselves; the designer will overwrite our change the next time we use it – as we surely will in response to changing application requirements. The generated code must have adequate extension points – mechanisms that enable the developer to inject behavior into the properties without touching the code itself. Unfortunately, the Entity Framework generates anemic property accessors. Here is another example:
[EdmScalarPropertyAttribute()] public string SocialSecurityNumber { get { return _socialSecurityNumber; } set { OnSocialSecurityNumberChanging(value); ReportPropertyChanging("SocialSecurityNumber"); _socialSecurityNumber = value; ReportPropertyChanged("SocialSecurityNumber"); OnSocialSecurityNumberChanged(); } }

The “getter” is not extensible. It simply returns the social security number field value. What if the user is not authorized to view that number? There is no way to block the attempt to read this value or to mask it so the user sees only a safe portion of it (e.g., the last four digits). The “setter” has a few extension points. There are reporting methods that could alert the application to changes. The ReportPropertyChanging and ReportPropertyChanged methods defer to an Entity Framework ChangeTracker object that monitors current and original property values. It could be useful to a watching application component (e.g., for data binding support). There are the partial methods, OnSocialSecurityNumberChanging and OnSocialSecurityNumberChanged, with which the developer can implement some limited logic specific to Social Security Numbers. Observe that the incoming value cannot be transformed before it reaches the field; we can complain (i.e., throw an exception) but we cannot heal.

23 | P a g e

IdeaBlade DevForce

DevForce and the Entity Framework

We are out of luck if we need generalized property logic that works across multiple properties. We shouldn‟t have to manually implement an On…Changing or On…Changed method for every property we want to validate. We should have a model-wide solution to validating changes that centralizes validation rules and manages them as resources … as we do in DevForce. And remember: validation is but one example of logic we could manage as metadata and introduce dynamically into the property.

Missing Inheritance
The Entity Framework supports inheritance hierarchies but only if each class in the hierarchy is mapped to a physical database table. The only base class that isn‟t mapped is the Entity Framework‟s own Entity class. There is no room to insert a class into the hierarchy that provides pure behavior. This is a serious omission. Years of real world application building confirm the wisdom and necessity of at least one base class that provides behavior that all business objects have in common. This is the application model base class, not Microsoft‟s or IdeaBlade‟s. Such a class could      Manage persistent auditing fields such as LastModifiedBy and LastModifiedDate. Generate separate audit trail objects during save. Implement data binding interfaces such as IDataErrorInfo. Cache broken validation rules. Provide access to the application‟s Dependency Injection or Service Locator facilities.

It is not uncommon to introduce similar classes elsewhere in the hierarchy. We might want an Inventory class in support of several distinct types of inventory, each mapped to its own table; we shouldn‟t have to have an Inventory table too.

DevForce Design and Code Generation
The DevForce Object Mapper and code generation address each of these deficiencies. The Object Mapper is a Visual Studio plug-in accessible from the tools menu. It presents classes and mappings in the grid format familiar to DevForce developers today. It surrenders to the Entity Framework all of the drag-anddrop finery of pretty boxes arranged on a stylish canvas. It favors a proven utilitarian approach that grants the developer unfettered access to large object models. The developer can add classes that are not yet mapped to a database table (or web method) and generate the entity classes. These classes can‟t be persisted until they are mapped. But they can be elaborated to support user stories and they can be tested. Call it impure if you must but it is a huge time saver to generate part of the conceptual model from existing database tables or web methods. You can do so incrementally. If you only need five objects, that is all you map. It‟s easy to come back later to generate additional storage-backed business object classes. The developer can specify an abstract base class that will never have a corresponding member in a data repository and insert this class anywhere in the business object class hierarchy. It‟s easy to set an application base class from which all new business objects derive by default. Like the Entity Framework, DevForce generates a partial class file covering the persistent data, leaving the developer to write custom business logic in a companion file. But DevForce gives the developer better control over the generated code. For example, using the DevForce Object Mapper she can     Decide which properties to make public and which to hide Make any property read only Include or exclude DevForce value verification Impose a required-value requirement on a property mapped to a nullable column

24 | P a g e

IdeaBlade DevForce Property Interceptors

DevForce and the Entity Framework

DevForce provides a mechanism to intercept and either modify or extend the behavior of any .NET property. This interception is intended to replace, and expand upon, the technique of marking properties as virtual and overriding them in a subclass. This facility is a lightweight form of what is termed “Aspect-Oriented Programming”. Interception can be accomplished either statically, via attributes on developer-defined interception methods, or dynamically, via runtime calls to the „current‟ instance of a PropertyInterceptorManager. Attribute interception is substantially easier to write and should be the default choice in most cases. You can learn about property interceptors in the chapter “Property Interceptors” in this Developer Guide.

Multiple Data Sources
An Entity Data Model maps all of its entities to tables in a single database. This is unrealistic for the many enterprise applications whose conceptual data models integrate information from multiple resources. It‟s not uncommon for an application to draw upon data resident in three or four different databases. Consider, for example, a custom ERP application that keeps order information in one database and accounting information in a separate database under the control of a third party accounting package. Business requirements are such that a contract for a new order stimulates a cascade of credits and debits. The ledger entries refer back to the order number and it must be possible to navigate from an order to its entire ledger history. Both databases must be updated when saving the new order. It would be a catastrophe if the order was added but not the ledger entries. We require a distributed transaction which means that the changes to both databases must either all succeed or all fail. This is an extremely difficult scenario for the Entity Framework. The two databases require two Entity Data Models and two sets of entity classes. An EF ObjectContext can only manage entities from a single model so we‟ll need at least two ObjectContexts at runtime. Our scenario calls for the ability to navigate from an order to ledgers and from a ledger entry back to an order. That will be tricky because the related objects live in different ObjectContexts. Entity instances don‟t know about other ObjectContexts so an order won‟t know which ObjectContext holds its companion ledger entries. The developer will have to create some clever infrastructure to make this work. Saving changes to orders and ledger entries is no picnic either. We have to save orders and ledger entries separately. The developer will have to be aware of the issue, set up a distributed transaction, and make sure that the Entity Framework properly enlists both save operations in that transaction. By contrast, DevForce supports multiple data sources. The Order and LedgerEntry classes can reside in a single model so there is no need for a separate interface assembly. DevForce will generate the navigation properties to walk from order to ledger entry and back again. And DevForce takes care of setting up the distributed transaction and enlisting the save operations within that transaction. Our example looks a bit like this:

25 | P a g e

IdeaBlade DevForce

DevForce and the Entity Framework

We see that DevForce is relying on the Entity Framework for object mapping and persistence operations while shielding the developer from unpleasant implementation complexities. The critical factor is the introduction of the DevForce business object model as a construct separate from the Entity Framework‟s own conceptual data model. In effect, DevForce provides a higher level abstraction over the Entity Framework object mapping abstraction.

Web Service Data Sources
The Entity Framework only works with relational data. It can only communicate with a relational database server. Not all data can be reached via a relational database server. Sometimes the data are locked up in a legacy nonrelational database. Sometimes the database is guarded by corporate IT or walled in behind a vendor‟s proprietary API. We may have to access such data sources through a web service. Our application may have to reach outside the corporate walls to access data from external sources. Tax rates, credit scores, geographical data, and zip codes are some of the external resources our application might expect to acquire. Most of these resources are already exposed as web services and those that aren‟t can be wrapped in a web or WCF service by a moderately skilled developer. Web service data are every bit as resistant to object oriented treatment as relational data. There is the same disconnect between the storage model and the conceptual model. Our object mapping technology should support entity classes backed by web service data. The upper application layers should be not be reminded constantly of the underlying storage technology. Entities of a DevForce conceptual model can be mapped to Web and WCF Service methods as illustrated here:

Lost Connections and Offline Applications
The Entity Framework‟s ObjectContext must always be able to connect to the database. If the application cannot connect, any query will throw an exception. The ObjectContext cache is unstable and unusable while the connection is broken so it is dangerous to continue even if something appears to work. Many applications operate in environments with unreliable connectivity. Mobile applications and wireless laptops are vulnerable to sudden outages. Without DevForce, the developer must work hard to protect against connectivity failures. The user‟s pending changes could all be lost. A DevForce application can be immune to these problems. The application can recover from an outage and continue to process queries against the cache alone until the connection is restored. The cache preserves unsaved changes, including newly added objects, so the user can continue working, albeit constrained to the world of entities presently in cache. A DevForce application can encrypt and save the entity cache to a local file with just a few commands. Later, with a few more commands, the application restores the cache from that file. A bullet-proof application might

26 | P a g e

IdeaBlade DevForce

DevForce and the Entity Framework

automatically store a user‟s pending changes locally every few minutes “just in case.” If the application crashes or the battery dies, the user could re-launch later and recover her work. We use this same mechanism to develop applications that operate offline intentionally. The user pre-loads the cache, preserves the cache locally, shuts down, re-launches while disconnected, does work, saves that work locally, and finally saves the pending changes to the database when reconnected. Someone may have saved changes to the same data while this user was offline. It happens while online too but the risk is greater when the time from change to save is prolonged. The response is essentially the same: DevForce detects the concurrency violation and the application resolves it, perhaps with the user‟s help.

27 | P a g e

IdeaBlade DevForce

DevForce and the Entity Framework

More DevForce Advantages
We‟ve seen the DevForce capabilities that are most critical for enterprise application development. There are other ways in which DevForce improves upon the ADO.NET Entity Framework. They may not be as critical in the majority of applications but they can significantly enhance developer productivity and code quality and are worthy of comment here.

Entity LifeCycle Events
The business object and upper application layers often need to know what the persistence layer is doing. The Entity Framework functions silently most of the time. It raises a SavingChanges event but won‟t tell you when the save operation succeeds or what entities were saved. There is no easy way of knowing when it reaches out to the database or returns with data. DevForce provides pre- and post- events or interception points for all significant moment in the “life-cycle” of an entity. Client side events include Creating and Created, Fetching and Fetched, Saving and Saved, Deleting and Deleted, Removing and Removed. There are also life-cycle extension points on the server-side (BOS) . These include the ServerSaving and ServerSaved methods so that developers can add custom processing immediately before and immediately after the save transaction. The ServerSaving method has access to the entities to be saved; the method can manipulate these entities, add to them, and remove them, before turning the final list over to DevForce for the save operation. The ServerSaved method knows if the transaction succeeded or failed and can invoke another server-side process as appropriate. Such a process might send a message to another service running in the hosted environment.

Lazy Load by Default
(The material in this section applies to DevForce WinClient but not to DevForce Silverlight, where all data retrieval is asynchronous.) DevForce navigation properties return a result if possible. The expression Order.Customer returns the order‟s customer if it has one. If the customer is already in cache, DevForce returns it. If the customer is not in cache, DevForce fetches it from storage. The behavior is the same if the navigation property returns a collection. The expression Order.OrderDetails returns the order‟s line items, retrieving them from storage if they are not found in cache. The Entity Framework takes a contrary approach. The navigation property it generates for Order.OrderDetails returns an empty list if the line items are not already in cache. The EF design team seems to have succumbed to architectural Puritanism. OO guidelines say a property should return quickly. A database query is not a fast operation. Therefore the team reasoned it should return nothing rather than return what the caller clearly expected: the list of line items. We agree with the rule in general. But we can think of no use case in which returning an empty list from Order.OrderDetails is the right thing to do. It only punishes the caller who will now have to write several lines of defensive code to satisfy the guardians of OO propriety. IdeaBlade decided to break the rule and provide useful behavior.

The Null Entity Pattern
DevForce scalar navigation properties always returns a business object. The expression Order.Customer always returns a customer object. Of course the returned customer is the order‟s real customer entity if the order actually has a customer. If the order doesn‟t have a customer, DevForce returns a placeholder object called the Null Entity.

28 | P a g e

IdeaBlade DevForce

DevForce and the Entity Framework

The same navigation property if generated by the Entity Framework would have returned null. Null values greatly complicate the developer‟s life. She has to be on constant alert for null reference exceptions. Data binding to a property that can return null is pure hell. A null reference exception thrown during data binding results in an ugly red bullet on screen and an error message that baffles the poor user. A customer null entity has all the properties of a real customer. The programmer can distinguish a null entity from the real thing when she has to but she doesn‟t have to litter the code with null value tests. Data binding survives nicely; a UI widget bound to a null entity displays a conveniently vacant value of the developer‟s choosing. The null entity pattern spares developers many hours of pain both in writing and reading code.

Proper Merge Strategies
When the Entity Framework fetches data from the database it must decide how to merge those data into its cache. What happens if the retrieved entities match entities already in the cache? What if some of those entities have pending unsaved changes or are scheduled for deletion? By default the EF only adds unmatched entities. That leaves modified entities untouched. But it also means that stale data are not refreshed. Inventory levels won‟t be updated. The user won‟t know about depletions or replenishments unless she is “lucky” enough to try saving a change to one of the adjusted products; the save will fail with a concurrency exception and she‟ll know to refresh the inventory level. An EF query with the “overwrite” option with refresh the unmodified inventory level – and wipe out the user‟s pending changes to other inventory objects. An EF query with the “preserve changes” option seems to do the right thing. It updates the unmodified inventory level and preserves the user‟s changes. Unfortunately, it obscures the fact that the changed inventory item is out of sync with the database. Suppose there was one item left in stock when the user fetched the inventory level. The user allocates it to her customer. Meanwhile, a different user sold the item to his customer, reducing the stock level to zero. After this user refreshes her cache with “preserve changes” she still believes there is one item in stock. There is not indication otherwise. She saves, intending to sell the item to her customer. The save succeeds and now the same item has been silently sold to two different customers? The DevForce offers equivalents to the EF merge strategies; they have their place. But the DevForce “preserve changes” option also preserves the pending concurrency conflict. The other user sold the item first and DevForce will prevent her from selling it twice.

The DevForce Verification Engine
DevForce provides a robust “Verification Engine” for validating the correctness of business objects. The developer can code custom verification rules and apply rules to objects by decorating properties with attributes, specifying the rules programmatically in the business object, or by reading them from metadata and adding them to the engine at runtime. While the application could suspend business object validation until just before save, user‟s prefer to be alerted immediately when they enter invalid data. Validation should be performed in the business object rather than the UI. Business object properties should validate proposed values as those values are conveyed from the UI to the object. DevForce supports this approach by inscribing calls to the Verification Engine inside the property setters.

Entity Metadata
The developer sometimes needs to know aspects of the conceptual data model itself. For example, she might need to iterate over all the child relationships of an order without knowing what those relationships are in advance. The metadata about such features of the model are often hard or impossible to find in the Entity Framework. DevForce records these features in metadata objects that can be easily reached programmatically through the

29 | P a g e

IdeaBlade DevForce

DevForce and the Entity Framework

EntityMetadataStore class. See the section “Getting Information About an Entity Type with GetEntityMeta()” in the Object Persistence chapter for detail.

Eager Entity Loading
By default, a query only returns the entities we ask for. If we query for orders, we get orders – and not the other objects related to those orders such as the customers, shipping addresses, line item details, and the product catalog. That‟s usually a good thing. Why suffer the performance cost of fetching related objects if we won‟t need them? With “lazy load” we can get a related object as we need it, when we need it, if we need it. In many scenarios we know we need the related objects immediately. Suppose our application presents the user with a list of orders. There is a grid beneath the list that displays the order details associated with the currently selected order. Clearly we need both the orders and their details at the same time. But if we stick with “lazy loading”, we‟ll see a flurry of tiny database requests as the grid calls Order.OrderDetails for each order in every displayed row. Performance will stink. Fortunately, in DevForce we can “eagerly load” the related objects by adding one or more “spans” to the query. When we add a span that specifies the relationship between Order and OrderDetail, the query engine fetches and caches the order details at the same time that it fetches and returns the selected orders. The grid‟s subsequent calls to Order.OrderDetails are satisfied quickly from the entity cache; there will be no extra trip to the server. The Entity Framework‟s LINQ to Entities syntax has a comparable feature called an “include”. We can add one or more “include” statements to eagerly load related entities. Unfortunately, there is no way to manage the includes of a LINQ to Entities query; there is no way to discover if it contains an include, no way to remove an include if it is not wanted. Moreover, an include instruction is a string, which means it cannot be type checked. In contrast, the DevForce programmer can inspect a LINQ to DevForce query for spans and add or remove them at will.

Dynamic Data Source Configuration
Data source connection management is unexpected chore. It seems simple at first: record the connection in configuration file and get out of the way. But, for many applications, the connections proliferate and the rules about who gets which connection become complex. The Entity Framework isn‟t much help in this department in part because it does not contemplate a world of multiple databases. But DevForce can help you tame the complexity. Two common scenarios illustrate the problem. In typical Enterprise development cycles, an application advances through a sequence of “environments” that begin with “Dev” and proceed through “QA”, “Stage”, and “Production”. The executables are the same but the data source connection information changes at each step. It should be easy “flip a switch” and re-point the application to the database (or set of data sources) that are appropriate for the targeted environment. In some On-Demand applications, each tenant has its own database or data source set. Financial institution „A‟ has its database, „B‟ has theirs, and so on. Users launch a common application front end. When they enter their credentials, the login module identifies the user‟s company and determines the corporate database that is correct for that user‟s session. In both illustration, the data source schemas are the same across all session; what changes from session to session is that actual database used. The data model mapping schema associates each entity type with a home storage schema. That schema has a symbolic name, the DataSourceKey. We know the storage schema at design time. We know the DataSourceKey at design time. But we don‟t won‟t know the actual data source to access until runtime. That‟s when we‟ll use the DataSourceKey to locate the appropriate connection string and hook up to a real data source.

30 | P a g e

IdeaBlade DevForce

DevForce and the Entity Framework

By default, DevForce looks for the connection string in an XML configuration file, expecting to find a dataSourceKey node identified by the DataSourceKey name. The connection string should be an element within that node. Continuing our first example, we might locate any one of four connection strings depending upon the environment. We don‟t want four separate configuration files. So instead, DevForce lets us maintain multiple connection strings for each DataSourceKey. It differentiates among them by means of a DataSourceKeyExtension, an extra bit of string associated with the DataSourceKey name. Now we can record as many connection strings as we need for any conceptual data source by creating distinct nodes uniquely identified by the both key name and extension. Nodes that share the same key name refer to the same conceptual data source; the extension tells us which concrete data source to use at runtime. We control runtime behavior by telling the client-side Entity Manager which extension to use. If we‟re running in the “QA” environment, we‟ll specify a “QA” extension. If the application entities map to conceptual databases “Alpha” and “Beta”, the application will connect to the concrete databases identified by “Alpha_QA” and “Beta_QA”. When we run in production we switch to the “Prod” extension and the application now connects to databases identified by “Alpha_Prod” and “Beta_Prod”. Notice that databases travel in sets. There is the “QA” set and the “Prod” set. We can use this same technique to support multi-tenant applications that store customer data in separate databases – an approach often mandated by financial clients. An “Acme” client session runs against the “Alpha_Acme” and “Beta_Acme” databases. The “Baker” client runs against the “Alpha_Baker” and “Beta_Baker” databases. The DevForce configuration file may not be the best place to store the connection information. In our second “On Demand” scenario, we could be adding new application tenants frequently. Rather than update the configuration file every time, we write a DataSourceKeyResolver to calculate and locate connection information based on key name and extension.

Custom Key Generation
Every entity must have a unique Entity Key so that the framework (a) can distinguish one entity from another and (b) recognize when two apparently distinct object instances actually represent the same thing. The Entity Key is the conceptual equivalent of a primary key in a database table row. Like a primary key, it can be a single value (e.g., an integer Id) or a composite key (e.g. as when a line item‟s key consists of it parent Order and Product ids). A newly created entity must have a unique key before it can be added to the cache; this is true whether we add the entity to the Entity Framework ObjectContext or to the DevForce Entity Manager. Sometimes we can create the key on the spot. It‟s easy if the key is a Guid or some other globally unique value that can be determined by the client alone. It‟s not easy if we must construct the key based on values acquired from a remote source. That‟s the more usual case. The key could be mapped to an auto-incrementing column in the object‟s home table. It could be generated by incrementing a counter stored in a separate database table (e.g., a NextId table).

Identity Column Keys
The Entity Framework supports the attributing of a column described in the storage model (SSDL) section of the Entity Data Model with the StoreGeneratedPattern enumeration. This lets the EF know that the back-end data store will generate a value for a column upon insert (or upon both insert and update) so that when an entity containing such a column is persisted the EF knows, post-save, to read the new value from the back-end data store and update the entity in the EF cache. The EF supports three states for StoreGeneratedPattern: None (the default), Identity, and Computed. Columns flagged with StoreGeneratedPattern=Identity are those updated only upon insert. Columns flagged with StoreGeneratedPattern=Computed are updated upon both insert and update.

31 | P a g e

IdeaBlade DevForce

DevForce and the Entity Framework

DevForce supports the StoreGeneratedPattern=”Identity” setting, extending its capabilities to encompass entities in the DevForce client-side cache. These entities need primary key values immediately upon creation, though they may not be persisted until much later. DevForce gives such entities a temporary primary key upon creation so they can be referenced client-side without any trip to the data source. Upon saving, their value is updated in the client-side cache to the value generated on the server. The foreign key values in other entities that reference the targeted entity are also updated to reflect the new, server-generated primary key value of the target entity. The Entity Framework can generate the new key for you if the key is a single valued integer key mapped to a SQL Server identity column. The EF can‟t set the object‟s permanent key; that won‟t happen until the newly created object is saved and even then it will be the database, not the application, that determines the key. So the EF assigns a temporary key and refers to that key when it adds related entities to the new object‟s graph. For example, upon creating a new Order, the EF assigns it a temporary key (e.g., “-1”). When we add a new OrderDetail to that Order, EF inserts “-1” into the hidden foreign key field of the OrderDetail that links the detail to the parent order. When the application saves these new entities, the EF acquires the permanent ids from SQL Server and updates the objects accordingly. Continuing our example, the EF learns that the new Order‟s primary key is “123” and updates the order‟s id. It also takes a critical second step: it finds all associated OrderDetails and updates their “ParentOrderId” column values from “-1” to “123”. “Id Fix-up” is our name for this propagation of permanent ids to related objects. Only then does it try to save the fixed-up OrderDetails.

Coping with Custom Keys
Many applications are tethered to an existing database with it‟s legacy primary key scheme. They can‟t use Guids. The key may be a simple integer acquired from a counter table named NextId. It might be a semantic key that combines the counter value with “meaningful characters”; maybe the order key includes the state and fiscal year as in “FY07-0270-CA”. We‟ll have to write the logic ourselves. When we create the order, we read the current counter from the NextId table, bump it for next time, calculate our key, set the Order‟s key – and then we‟re ready add the entity to the ObjectContext. It‟s a pain but it‟s manageable for a continuously connected application. It‟s much harder if we must support an application that can operate offline. We won‟t always be able to reach the NextId table so we can‟t always calculate the permanent keys immediately. We‟ll need a custom temporary key and Id Fix-up scheme. DevForce can do all of this for you. You write a custom Id calculation class that conforms to a DevForce interface. DevForce discovers the class and manages key creation, temporary keys, and Id Fix-up during the save. Of course it works even when your application runs offline.

Declarative Concurrency Column Management
Many applications must guard against the possibility that two different users will unknowingly edit and save the same entity simultaneously. Without some kind of checking, the last person to save wins. If I sell a particular item and you sell the same item, we will have sold the same item twice although the database will show only that you sold it. I could have put a database lock on the item record, thus preventing you from reading and editing it. Such “pessimistic locking” harms performance and leads to troubling lock-out scenarios. Neither DevForce nor the Entity Framework supports such a physical locking scheme. The Entity Framework relies on “optimistic concurrency” techniques to detect and resolve concurrent access conflicts. Optimistic concurrency assumes that two users rarely wrestle over the same record and therefore allows all users to access records freely. If two users, such as you and I, try to update the same record, it detects the conflict and terminates the second save; it informs the second client be raising a concurrency exception.

32 | P a g e

IdeaBlade DevForce

DevForce and the Entity Framework

The Entity Framework implements optimistic concurrency by comparing the value of a concurrency column in the pending record with the value of that column in the stored record. If the values are the same, the pending record can be saved. If the values are different, the pending record is out of sync with the stored record; the framework assumes a concurrency conflict and throws the exception. This technique works so long as the concurrency value is changed after each successful save. Who is responsible for that change? The Entity Framework says that you are. You are fortunate if the database table has an update trigger that can do it. Otherwise, you have to write the code that updates the concurrency column and you have to remember to call it at the right moment. DevForce can handle the concurrency column update for you. In the DevForce Object Mapper you declare the concurrency column (or columns) and pick a method from a list of concurrency column update methods. DevForce will call that method at the appropriate time. Yes, you can extend the list with a custom method.

Undo and Checkpointing
The Entity Framework lets you accept all entities with pending changes (thus disguising a discrepancy between data in session and data in storage!) but won‟t let you roll back changes – either individually or collectively – without many lines of programming. “Undo” is a one line command in DevForce. There is no progressive undo capability in Entity Framework. With the DevForce “Checkpointing” facility, the application can roll back the state of the entire entity cache to any one in a sequence of “checkpoints” or snapshots. Wizards put this feature to good use. Each step forward through the wizard can be marked. In the current step the user might add new entities, modify or delete existing entities, and retrieve more from the database. If the user then cancels the current wizard page and retreats a step, the application can discard all of these changes and restore the state of both the entities and the cache to the marked state with a single command.

Sandbox Editors
Sandbox editors are a convenient alternative – or compliment – to checkpointing. Imagine that customer “Jim” calls to adjust one of his orders. You find the order in the list and open it in an editor and begin working on it. You‟re in the midst of changing deliver addresses, order items, billing information, etc. Suddenly, premium customer “Sally” calls you with an urgent request for a new order that you must enter right now. Jim kindly agrees to complete his changes later. You begin Sally‟s order in a second order editor. You are half way through Sally‟s order when Jim calls you back. He says “never mind, that order we were changing is just fine the way it was.” You switch briefly over to Jim‟s order and discard all changes simply by shutting down the order editor. You return to Sally‟s order editor, complete it, and save. There are two distinct orders in flight in this example. Each has its own set of entities some of which may overlap (e.g., the list of shippers) although most do not. With DevForce, you can create separate Entity Managers – with separate caches – and maintain these editor sets separately, each in their own “sandbox”. The entities in the “Jim” Entity Manager are isolated from the entities in the “Sally” Entity Manager and all of these entities are isolated from the list of orders held in the application‟s main Entity Manager. Now imagine that this scenario takes place off line. There is no access to the database. That still works in DevForce because you can easily pass copies of entities from one manager to the next without going to the database. You might even prefer this approach when connected if the performance of your application is at a premium and bandwidth is poor.

Managed Lists
Keeping lists of entities up-to-date is a recurring application problem. It‟s the holiday season as I write this so let‟s imagine we‟ve written Santa‟s inventory tracker. The tracker displays a list of undelivered packages on Santa‟s dashboard. As each package finds its intended child, the list should grow shorter.

33 | P a g e

IdeaBlade DevForce

DevForce and the Entity Framework

An elf in the back of the sleigh is updating package information on a separate screen, marking each one “delivered” as it drops down the chimney. Santa sees the same dashboard on the console monitor because he and the elf are cabled together. What makes the list shrink when the elf marks the package delivered? Traditionally, we‟d have written the logic ourselves. But there is a problem: the elf‟s module doesn‟t know about the list displayed on the dashboard. So it‟s not as easy as remembering to remove an item from UndeliveredList when the elf clicks the “Delivered” button. We‟ll probably need some kind of cross module event scheme. DevForce can handle this for us automatically with its managed list feature. Let the two modules share the same Entity Manager, let the list be governed by this manager, give the list the appropriate predicate – “keep item if not delivered”- and the list takes care of itself.

Conclusion
IdeaBlade has been in this arena since the early days of .NET. The DevForce product has long offered most of the capabilities described in this paper including the multi-tier ORM, client-side caching, and code generation. The Microsoft ADO.NET Entity Framework is a solid contribution to the field and its very existence confirms the widespread need for an infrastructure like DevForce. But the Entity Framework by itself cannot fulfill the needs of many enterprise applications. The productivity isn‟t quite there. The generated code lacks essential support for business object development. Its two-tier architecture limits the application‟s ability to reach a distributed user community with the required performance and security. With DevForce, developers can quickly realize the potential of an object-oriented, multi-tier, enterprise application connecting hundreds or thousands of users.

34 | P a g e

IdeaBlade DevForce

Getting Started

Getting Started

Getting Started
Installation DevForce Start Menu

The “NorthwindIB" database Development Process

This section offers a brief overview of how to get started with IdeaBlade. Topics covered in this chapter are:

Topic Installation DevForce Start Menu NorthwindIB database

Description Brief introduction and pointer to pertinent sources. Tools and information accessible from the Windows Start Menu. Many examples make use of the tutorial NorthwindIB database, which is based on Microsoft's blueprint NorthwindEF database. Best Practices indicators and typographical conventions used in this guide.

Documentation Conventions

Installation
The separate DevForce Installation Guide covers the installation and upgrade process in detail and also contains a troubleshooting section. The DevForce Release Notes contain version specific information that you may need for certain upgrades.

DevForce Start Menu
Installation adds an “IdeaBlade DevForce” folder to your Start menu. At the moment it looks like this:

35 | P a g e

IdeaBlade DevForce

Getting Started

Documentation Menu Item Developers Guide DevForce Help Installation Guide Learning Units Release Notes

Description The document you‟re reading now. The technical help covering the DevForce assemblies, types, and type members. How to install and upgrade DevForce. Includes Troubleshooting tips. Scripts and solutions for hands-on walk-thrus of DevForce product features and applications. Documentation of new features, enhancements to existing features, bug fixes, upgrade issues, and everything else you need to know when upgrading your copy of DevForce.

36 | P a g e

IdeaBlade DevForce

Getting Started
Applies to DevForce Silverlight No

Tools Menu Item Assembly Binding Redirector

Description Discovers third-party control suites on your machine; compares the names and version numbers of the associated DLLs with the versions supported by DevForce; suggests redirections to permit you to use your installed versions with DevForce; and writes machine.config statements to implement the selected redirections. Edits an IdeaBlade.ibconfig file governing your application‟s deployment. DevForce application deployment is its own chapter in this guide. Installs the NorthwindIB sample database. This is a SQL Server database (not a database server) that is used in many of the Learning Units that accompany the product. Quick „n dirty tool that facilitates testing a distributed app. Creates folders for client- and server-side assemblies and configuration files, and otherwise facilitates running the DevForce EntityService in a separate process from the client side app and EntityManager. Facilitates replacing your current product key with a new one (in the case of upgrade, etc.) When you elect installation of Windows Forms support (the default), DevForce adds a number of visual design components to the Visual Studio 2008 Tool Box. Sometimes VS won‟t accept our automated attempt to install these tools. You may remove one or more from the VS tools accidentally. You may acquire a 3 rd party control suite for which we have a dedicated visual component. This installer will help you (re)install these components. Tool for listening to logged activity from a running DevForce application.

Config Editor

Yes

Database Installer

Yes

N-Tier Configuration Starter Product Key Updater Tool Box Installer

No

Yes No

Trace Viewer

Yes

The “NorthwindIB" database
NorthwindIB is a sample database that is referenced by the DevForce Tutorials and other documentation. It differs from its source, the Microsoft NorthwindEF database, in several significant respects while retaining a recognizable parentage. The data are mostly the same. The differences between NorthwindIB and NorthwindEF are detailed in a text file “NorthwindIB_DifferencesFromEF.txt” which installs in the DevForce installation directly alongside the NorthwindIB.MDF database file. We recommend installing or upgrading your copy of "NorthwindIB" so you can follow along with our tutorials and code samples. The normal installation process tries to add this to your SQL Server databases but it may fail to do so for any number of reasons. We also update this database from time to time in order to support new example code that illustrates DevForce features. Please see the Installation Guide for instructions on how to install or upgrade this database.

Development Process
DevForce promotes a distinctive process for development of distributed, object-oriented, enterprise applications.

37 | P a g e

IdeaBlade DevForce
The “object-oriented” and “distributed” parts may seem a little foreign to some.

Getting Started

The object-oriented approach to data means thinking in terms of Business Objects and Object Persistence rather than retrieving, inserting and updating data records. This becomes so obvious and easy in DevForce that, in a few days, you stop thinking in terms of fields and joins and you may even forget how to use ADO. The “distributed” aspects don‟t surface until well down the road and, because it‟s easy to re-configure the application for multiple physical tiers, there is no cost to delaying awareness of multi-tier considerations. Many developers do the lion‟s share of their work in a one-tier physical model in which all components of the system – even a test database – reside on a single physical machine. You may prefer to access test data on an independent server in which case your development experience is not that much different than good-old client / server. The following sections summarize the stages in a typical development process. The summary highlights the end-toend influence of the DevForce infrastructure.

Database Schema Implementation
You either have a DBA or you are the DBA. If you are the DBA, you‟ve always been in control. If you have a DBA, you‟re going to have to coordinate with her. Fortunately, she can remain the only one who touches the database. You may recommend schema changes but nothing in DevForce requires a schema change. There may be a tussle over stored procedures. The DBA may want you to use them. You can. But your life will be much better without them in most cases. DevForce assumes you are starting from an existing database schema. Theory says the object model should dictate the storage schema. This is highly desirable, especially in the early design phases. However, once your application(s) have settled in, the “Model Driven Architecture” (MDA) approach becomes academic. The database is what it is and you may change it only at increasing cost. DevForce does not provide an MDA tool. Neither does it interfere with MDA. It picks up where MDA leaves off. Of course your schema doesn‟t stand still either. DevForce adapts to those changes without imposing any of its own.

Object Mapping
The application architect or senior developer uses a combination of the Entity Framework‟s Entity Data Model (EDM) Designer and the DevForce Object Mapper to configure the map business object classes to data source objects. While “data source objects” can reside in databases, web service methods, or message queues, most enterprise application data are stored in relational databases. Accordingly, most object mapping is between business object classes, AKA entities, and tables, views, or stored procedures in a relational database. You cycle around and around from schema design to database schema change to object mapping to redesign. You return repeatedly to the Object Mapper, knowing that it adapts to change while your business object layer rides above the mapped classes, insulating the UI layers of your application from adverse consequences of those changes. The DevForce Object Mapper and the Entity Framework EDM designer co-exist happily: neither interferes with the other‟s work.

Business Logic Elaboration
That business object layer is the locus of business logic development. You begin with a collection of use cases and test cases that explore the persisted features of the Business Objects. You might read simple entity properties and explore the network of relations among the entities, perhaps printing results to the console.

38 | P a g e

IdeaBlade DevForce

Getting Started

Business Object creation and comparison methods are next, following the IdeaBlade recommended patterns and using a few simple methods of the Entity class at the root of all Business Objects. Note that we‟re not writing UI here. We‟re exploring and enriching the business objects independently of any particular user experience. Our changes go in the developer partial classes that are initially generated by the Object Mapper and subsequently left entirely alone. Slowly we begin to add rules and to verify those rules. The ship date must be after the order date, for example. Only a user with administrative rights can change a salary. The developer inscribes these rules in the custom entity classes that comprise the business logic layer. The developer doesn‟t spend much time thinking about how to push and pull entities from their persistent homes in data storage. That‟s the job of the EntityManager, guided by the object map. Developers no longer embed SQL commands for reading or writing data to storage. Instead they invoke strongly-typed LINQ queries that return business objects (entities). They write business logic that references object properties, not data fields.

Nonpersisted Classes
Business objects have state and long-lasting identity. They are stored (persisted) for an extended time. Not everything is a business object. There are transient objects and Singleton classes with no state to store such as temporary collections (e.g. list of user-selected products). calculation classes (e.g., a ROI calulator). helper classes to which similar Business Objects delegate common functionality (e.g., audit logging). APIs to external applications. Because these are not "Business Objects" - they do not carry state (or least what state they have is not stored in the database). They are not mapped and they fall outside the purview of DevForce Object Persistence. Such classes are written in the normal fashion and will be collected in one or more separate application projects.

Application Control Classes
The developers add the control classes that manage navigation and work flow. They enable the graphic designer‟s buttons and menus and tie the forms to other methods in the business objects.

Deploy
The application is completed in record time. There have been mock deployments to development, test, and staging environments. The production deployment is no different. The application code itself is identical, whether it is installed on the server or deployed to a client; the only difference is the configuration of their respective IdeaBlade Configuration files which are now separately edited for the production environment. There are packages of files: a server package and a client package. The server package typically is a Microsoft Install file (MSI). This file is unpacked into the designated directories of one or more servers and each is launched. The Business Object Server monitors the health and activity of each server application. DevForce WinClient .NET‟s “ClickOnce” publishing is the easy way to build and distribute the client package.

39 | P a g e

IdeaBlade DevForce

Getting Started

ClickOnce puts the package (a collection of files) on a web server and makes it accessible through a web page. PC users with .NET runtime installed navigate by browser to the page and clicking an install button. ClickOnce downloads and installs the client application in the user‟s personal directory. It then launches automatically. This install happens once. Developers upgrade the application and publish revised client packages. ClickOnce detects the upgrade and downloads the new version seamlessly. There is no danger of the application executables and class libraries colliding with those of other application. Nor will an install (or uninstall) of a different application disturb this one. .NET has corrected the DLL nightmare with proper versioning. .NET applications don't have to use the Windows registry either. DevForce Silverlight In a Silverlight application, the XAP file is your client package, and Visual Studio has fortunately built it for you. You‟ll need to be sure to place the XAP file in the ClientBin folder of the web site which is hosting the application. This site will typically also host the Business Object Server (deployed as the server package above) but this is not a requirement. If you do host the BOS and Silverlight application from different web sites you‟ll need to be sure to also deploy a file named “clientaccesspolicy.xml”. More information on this is available later in this document.

40 | P a g e

IdeaBlade DevForce

Hello DevForce

Hello, DevForce

Hello, DevForce
DevForce Application Architecture - The Big Picture DevForce and the ADO.NET EntityModel Your First DevForce Application: a Walk-Through Building the Domain Model Add a User Interface Add Unit Tests Add a WinForm UI

Understanding the App.Configs
Information Flow Between the App.Configs

Monitoring Activity
Appendix: Listings of Sample App.Config Files Appendix: Probing Sequence for the App.Config File

Creation means finding the new world in that first fierce step with no thought of return. David Whyte, “Statue of Buddha”

Don‟t look back. All change, all creation, is attended first by grief for what is lost followed by the clarity in moving on with no thought of return. DevForce is not magic and you‟re unlikely to build an enterprise application over night. But you can build a good application that you‟re proud of in reasonable time. Once you lay to rest your old habits and have grieved for them awhile, the new path will embrace you and, in spare moments, you may wonder how you ever did it that old way. “But I‟m so happy in my comfortable way. What if things go wrong? That DevForce thing is just a little intimidating.” This chapter should ease you across the threshold, highlighting some of the more prominent DevForce features along the way.

DevForce Application Architecture - The Big Picture
A DevForce application relies upon a layered architecture for data access. At one end is a data source – typically a relational database. At the other end is the user interface which works with business objects in a business object model. There are several components in the middle.

41 | P a g e

IdeaBlade DevForce

Hello DevForce

Figure 1. Application Components in a DevForce Application

One of them, called an EntityServer, moves data (and data requests) between the ADO.NET Entity Framework and DevForce business objects. If the back-end data store is a relational database, the EntityServer leaves the direct communication with the data store to the ADO.NET Entity Framework. However, if the back-end data store is a web service, the DevForce EntityServer handles the job, since that capability does not exist within the Entity Framework. The EntityServer has a copy of the application‟s business object model so that it can instantiate DevForce business objects server-side if need be. However, for most operations (such as simple data retrievals), it forwards to the client-side EntityManager the data required for hydrating DevForce business objects there, without ever instantiating DevForce business objects on the server. The data is packaged and passed in a highly efficient format and process. The ADO.NET Entity Data Model includes the mapping information necessary to translate between locations in a relational data source and the corresponding persistent fields in the ADO.NET business entities. The EntityServer (besides handling those jobs against web services), mediates between the Entity Framework and the DevForce EntityManager that manages the client-side cache used by your application. The second important DevForce component is the EntityManager. The EntityManager takes instruction from the higher levels of the application such as the UI, and forwards UI requests for entities to the EntityServer. The EntityManager puts the received entities – obtained from whatever source by the EntityServer -- into its entity cache and makes them available to the UI. End users review the entities and make changes through the UI. The UI signals the EntityManager to save the changes. It dutifully forwards the changed entities to the EntityServer which communicates with the appropriate component to commit the data into persistent storage.

DevForce and the ADO.NET EntityModel
Visual Studio‟s ADO.NET Entity Data Model wizard creates an EDMX file which contains descriptions of a conceptual data schema (the object model), an actual data store schema (the database model), and the mappings between the two. It also renders the object model in code in a file named <ModelName>.Designer.cs (or .vb). 42 | P a g e

IdeaBlade DevForce

Hello DevForce

The developer‟s first step in building the object model for her application will consist in creating an entity model in an EDMX file. Typically she will use the Visual Studio Entity Data Model wizard to create the initial version of the EDMX file and the corresponding generated code file. After that, she will work with some combination of the Visual Studio Entity Model Designer and direct XML coding in the EDMX file, depending upon her preferences and whether she needs to use features in her model that are not supported by the Entity Model designer. The second step will be to create a Domain model using the DevForce Object Mapper. This model is so named because it will be composed of one or more Entity Models persisted in .EDMX files. The DevForce Object Mapper will alter the .EDMX file by adding additional elements and attributes. These added features are ignored, and left undisturbed by, the ADO.NET Entity Data Model Designer. Because of this, the developer can move back and forth between the Visual Studio Entity Model Designer and the DevForce Object Mapper without fear of either disturbing the other‟s work. There is, by intent, some overlap in the the functionality of the DevForce Object Mapper and ADO.NET Entity Data Model Designer. However, our intial work on the DevForce Object Mapper has been focussed on providing needed or useful capabilities that are either not present, or are difficult to work with, in the Entity Data Model Designer. Our goal is to make it as convenient as possible for you to work with your model. We mentioned that the Entity Data Model wizard and designer, in addition to altering the .EDMX file, generate the classes that comprise the compilable manifestation of the object model. From the Object Mapper‟s enhanced version of the .EDMX DevForce generates two sets of classes. The first is essentially the same Entity Framework model generated by the Visual Studio tools. This version of your object model will be deployed to the logical middle tier of your application, where it is used by the ADO.NET Entity Framework for creating objects of the type that it understands. The second version of the object model generated by the DevForce Object Mapper is a DevForce version consisting of business classes that inherit from IdeaBlade.EntityModel.Entity. As previously mentioned, we refer to this version of the model as the Domain model. The Domain model is “persistence ignorant”: unlike the Entity Framework model, it has no knowledge whatsoever of the back-end datastore or the mapping between that and its objects. In an n-tier deployment, it is the only model that is deployed client side. The client needs no connection information for back-end datasources. The DevForce Domain Model is the only client-side model your application will use. It is, however, also deployed server-side; and it‟s scope of operation is synonymous with the bracketed area labelled “DevForce Business Objects” in Figure 1. The Domain Model is a consumer of Entity Data Models, whose .edmx files typcially define the lion‟s share of its content. Server-side, DevForce delegates to the Entity Framework the jobs of communicating with the database(s) to perform persistence operations including data retrieval and saving. The Entity Framework, in turn, uses the compiled versions of the Entity Data Models, as well as connection information typically stored in an app.config file, to do its work. In DevForce, all direct communications with back-end data sources are considered, logically, as server-side operations, which they will literally be in an application deployed across three or more physical tiers. The application components that facilitate such communications, including the Entity Framework, Entity Data Model, and DevForce EntityServer are considered server-side components, and are kept logically separate from client-side components such as the DevForce EntityManager and the client application. It is perfectly possible to deploy both the logical client-side components and the logical server-side components to the client machine, and this is often the configuration used for much of the development work even on enterprise applications. When all application components including the database server are deployed on a single physical machine, you have a “single-tier deployment”. When all application components except the database server are deployed on a single physical machine, and the latter is deployed to a remote machine, you have what is known as a “client-server” application. When client-side application components are deployed on a separate machine from server-side application components, this is typically referred to as “n-tier” deployment, even if the database server resides on the same machine as the application server (e.g., the DevForce BOS).

43 | P a g e

IdeaBlade DevForce

Hello DevForce

However, since the strongest application security, widest availability, greatest scalability, and easiest deployment are all associated with n-tier physical deployment, we figure it‟s much to your benefit to write your application from the beginning to permit that, and we make it as easy for you as we can. What are the Parts of Your Business Model, and Where Are the Parts Deployed? The ADO.NET Entity Data Model and the DevForce Domain Model each have representations in both XML and in .NET code. The representation of the Entity Data Model (EDM) in XML is a file with the extension .EDMX. Visual Studio includes a code generator that creates a corresponding file of .NET code. This file has the same name as the .EDMX file, but an extension of “designer.cs”. It is stored by Visual Studio subordinate to the .EDMX file in the Entity Data Model project. The representation of the DevForce Domain Model in XML consists of a file with the extension .IBEDMX; and one or more of the Entity Data Model (.EDMX) files just discussed. The .IBEDMX file mostly acts as a catalog of the Entity Data Models that contain, in XML, the detailed specifications of entities, properties, associations, tables, columns, relationships, and mappings. Both the DevForce Object Mapper and the Visual Studio Entity Data Model Designer read from and write to the .EDMX files. The tools cooperate completely, fully respecting each others‟ work, and may be used in any order. Using the specifications stored in the .IBEDMX and .EDMX files, the DevForce Object Mapper generates a file of .NET code which has the same name as the .IBEDMX file, but an extension of “designer.cs”. This generated code file is stored by the Object Mapper subordinate to the .IBEDMX file in the Domain Model project. The Object Mapper also generates “developer partial class” files for each entity in the Domain Model. These files are named “<EntityName>.cs” and are generated into the Domain Model project.

44 | P a g e

IdeaBlade DevForce

Hello DevForce

Your First DevForce Application: a Walk-Through
With that information, we‟re ready to walk through the process of building a simple DevForce-based application. The process consists of the following steps: 1. 2. 3. 4. Obtain or create at least one ADO.NET Entity Data Model Create a new Domain Model using the DevForce Object Mapper Add the Entity Data Model to the Domain Model Adjust the Entity Data Model as desired; e.g., rename classes and properties, designate concurrency columns, etc. 5. Save the Domain Model, which results in code being generated for the types in the model 6. Optionally, add additional Entity Data Models and repeat 7. Optionally, add entities backed by web-services 8. Add custom business logic to the entity classes 9. Add Unit Tests a. Add References b. EmployeeTest First Look c. Get a Test Employee d. Run the Test e. Accumulating Test Results f. Lessons Learned 10. Create the UI a. Unaided .NET Winforms b. .NET Winforms Using DevForce V3 UI tools c. WPF Let‟s get started!

Building the Domain Model
1. We‟ll begin our walk-through by creating a blank Visual Studio solution named “DevForce01”:

45 | P a g e

IdeaBlade DevForce

Hello DevForce

To that we‟ll add an empty project into which we‟ll subsequently put a newly created Entity Data Model. If you know you‟ll be starting with an existing solution that already includes one or more Entity Models, you can skip ahead to the section, “Build Your Domain Model Using the DevForce Object Mapper”. 2. Add a new Class Library project to your blank solution. Name this project “ServerModelNorthwindIB”. We‟re naming it this because it will house an Entity Data Model, which would only be deployed serverside in an n-tier deployment; and because that Entity Data Model will be based upon the NorthwindIB database. Delete the Class1.cs file that gets created by default in the new project. You will not use it. Add a New Item, an ADO.NET Entity Data Model, to the project using the Entity Data Model wizard. Name your model ServerModelNorthwindIB.edmx.

3. 4.

a. b.

On the Choose Model Contents dialog, select “Generate from database”. On the Choose Your Data Connection dialog, create or select the NorthwindIB data connection. Rename the connection settings key name for the App.Config file to “ServerModelNorthwindIBContext”. (That name will end up being used for the .NET ObjectContext that will be generated by the Entity Framework.)

46 | P a g e

IdeaBlade DevForce

Hello DevForce

c.

On the Choose Your Database Objects dialog:  Uncheck the Stored Procedures and Views.  Expand the Tables node and make sure only the following tables are checked:  Customer  Employee  Order  OrderDetail  Product  Supplier

47 | P a g e

IdeaBlade DevForce

Hello DevForce

 Rename the Model Namespace to “ServerModelNorthwindIB”  Click <Finish>. 5. Visual Studio will create an Entity Data Model using the settings you specified, and will open it in its graphical editor. Save the file without making any changes, then inspect it in the graphical editor. Note the associations (“relationships” in database parlance) among the various entities. Order, for example, has associations to the Customer, Employee, and OrderDetail entities. It‟s on the “many” end of 1-to-many associations with Customer and Employee; it‟s on the “1” end of a similar association with OrderDetails. Notice the corresponding Navigation Properties that were generated by the Entity Data Model wizard: Customer, Employee, and OrderDetails.

48 | P a g e

IdeaBlade DevForce Note also that the Employee entity has a relationship to itself, representing the reporting relationship among NorthwindIB Employees. The Entity Data Model wizard modelled the relationship and generated navigation properties, but lacking much understanding of what the entities on the two ends of that relationship represent, simply named them “Employee” and “Employee1”. At the same time it created navigation properties in the Employee type named “Employee1” and “Employee2”. “Employee1” returns a collection of Employees (those who report to any current Employee); whereas “Employee2” returns a single Employee (the current Employee‟s manager). All very confusing!

Hello DevForce

You could, at this point, do considerable further work on your Entity Data Model. For example, you might:   rename entities, entity properties, and entity sets, to fix pluralization, or for other reasons; create new entities; 49 | P a g e

IdeaBlade DevForce      

Hello DevForce

describe inheritance relationships among the entities; create new associations (relationships) between the entities; define complex types within the entities; create and map function imports for stored procedures; map entities to stored procedures for CRUD (Create, Retrieve, Update, Delete) operations; and do other operations.

But let‟s leave it alone for now, and keep moving. Note that we could also have included Stored Procedures and Views in our Entity Data Model while creating it with the wizard. We didn‟t in order to keep things simple for the purpose of this beginning tutorial. But DevForce supports everything you can do in your Entity Data Model.

Build Your Domain Model Using the DevForce Object Mapper Now that we have an Entity Data Model to work with, we‟ll create our Domain Model. 1. From the Visual Studio main menu, select Tools / DevForce Object Mapper. DevForce, finding no domain model already in the solution, displays a node named (New Model) in the navigational tree control. Select the (New Model) node in the tree control. Observe the Project Settings:

2.

DevForce has picked a project as the target for the generated domain model, but let‟s say we want to put the generated DevForce class files into their own project. Click the <New project…> button. The following dialog displays:

50 | P a g e

IdeaBlade DevForce

Hello DevForce

We‟ll accept the defaults here, letting the Object Mapper create a new Windows Class Library project named “DomainModel”. A directory of the same name will be created at the Location shown to house the generated files. Note also the checkbox labelled “Create Silverlight Domain Model project”:

This checkbox will only appear if you have a DevForce Universal or DevForce Silverlight license (but not a DevForce WinClient license). In Universal it will default to being unchecked, as shown. In DevForce Silverlight, it will default to being checked. For now we‟ll leave it as is, directing the Object Mapper not to generate any Silverlight files. 3. Observe the “Domain Model Settings”.

When the Object Mapper saves the domain model and generates code, it will use the Namespace (“DomainModel”) shown for the generated code, and will also name the EntityManager container (“DomainModelEntityManager”) as shown. You‟ll use the EntityManager for retrieving data into your local cache, for saving changes, and for many other data persistence operations.

51 | P a g e

IdeaBlade DevForce 4. Now observe the “Save-Related Settings”.

Hello DevForce

By default, the Object Mapper will, when you first save your work, do all of the following:  Generate a code file, <DomainModel Name>.<Entity Data Model Name>.Designer.cs (or .vb); e.g., “DomainModel.ServerModelNorthwindIB.Designer.cs”. This code file contains partial classes for all business entities defined in the DomainModel. Create an app.config file in the domain model project, and store configuration information (such as connection strings) there. Generate a handler for the post-build event of any executable project that references the DomainModel assembly. This handler will make sure that  the ideaBlade.configuration section of the app.config in that project gets updated at build time to reflect any new information in the copy of app.config that resides in the domain model project; and the assembly containing the Entity Data Model gets copied to the executables folder. 3

 

All .NET code will be generated in the language you select, C# or VB. If the “Create developer classes” checkbox is checked, the Object Mapper will also generate developer partial classes for each of the entities in your model. We‟ll discuss these more later. 6. From the (right-click) shortcut menu associated with the domain model (New Model) node in the tree control -- or from the Model option on the main menu -- select Add Entity Model.

3

If the developer has elected to have the .SSDL, .CSDL, and.MSL files generated by Visual Studio from the Entity Data Model stored as loose files rather than embedded resources, those will also get copied to the executables folder.

52 | P a g e

IdeaBlade DevForce DevForce will find the Entity Data Model in your solution and suggest that as the model to add:

Hello DevForce

Click <Open> to affirm that selection. DevForce will display a node for the ServerModelNorthwindIB Entity Data Model as a child of the (New Model) domain model.

You can add as many Entity Data Models as you like to your domain model. 7. Select the ServerModelNorthwindIB.edmx node in the tree control. Observe the settings in the details pane.

53 | P a g e

IdeaBlade DevForce

Hello DevForce

The meanings and uses of these settings are described in detail in the “Object Mapping” topic document in the DevForce Learning Resources. Please refer to that for details. For our purposes here, just accept all the default settings. 8. We‟re going to do some further work on our model, but before we do, let‟s save our work. Do so by clicking the “Save Domain Model” button on the toolbar (with a diskette icon), or by clicking the Save option on the File menu, or by clicking the Save option on the shortcut menu associated with the (New Model) node. A new project named for your DomainModel is created in the current solution. In that project your domain model is generated. Both the project file and the domain model files have been placed in a directory to house them. The Location requested is for the project directory. By default that directory will be created immediately within the directory where the existing solution resides. In addition to the above, the Object Mapper created a Solution Folder and moved the Domain Model project into it. You may, if you like, choose to move the project containing the Entity Data Model into that folder as well; but that‟s up to you. The Object Mapper has also generated a “designer” code file, DomainModel.ServerModelNorthwindIB.Designer.cs, and placed it subordinate to DomainModel.ibedmx, the XML file representing the domain model. DomainModel.ServerModelNorthwindIB.Designer.cs plays a role relative to the domain model similar to that played by the ServerModelNorthwindIB.Designer.cs file in relation to the entity data model, ServerModelNorthwindIB.edmx. That is, it contains the generated C# or VB code that represents the blueprint for the runtime object model. Note that DomainModel. ServerModelNorthwindIB.Designer.cs contains the domain model (consisting of DevForce entities and related classes), whereas ServerModelNorthwindIB.Designer.cs contains the entity data model (consisting of ADO.NET EntityObjects and related classes). If we had checked the “Create developer classes” checkbox (see at right), then the Object Mapper would also have generated one additional file for each entity in our model. Each file would contain a single partial class for the entity specified in the file name. Let‟s do that now to see the additional files.

If you closed the Object Mapper, re-open it. Check the “Create developer classes” checkbox and then save the domain model again.

54 | P a g e

IdeaBlade DevForce

Hello DevForce

The Object Mapper has now generated “developer” partial classes (in individual files) for each of the types in the business object model. These are Customer.cs, Employee.cs, Order.cs, and so forth. Once generated, these classes won‟t be overwritten by the Object Mapper. They‟re designed for your custom code.

The Object Mapper also generated into the DomainModel project an app.config file that contains (among other things) connection information for the data sources for all Entity Models contained in the Domain Model. In an n-tier app, this connection information will not be deployed to the client machine, but only to the machine with the DevForce Business Object Server (BOS).

Examining and Editing the Contents of the Entity Data Model in the Object Mapper 9. Reopen the Object Mapper if it is closed, and double-click the ServerModelNorthwindIB.edmx node in the tree control, to expand it. You should see (hierarchically just below it) the container node for the ServerModelNorthwindIB Entity Data Model, called ServerModelNorthwindIBContext. Note that this container has the name that you specified, while creating the Entity Data Model in the EDM wizard, for saving the connection information.

10. In the upper part of the details pane you should see the names of each type that you selected for inclusion in ServerModelNorthwindIB. Note that both the type names and the corresponding Entity Set Names are the same as the names of the tables on which they were based. But really, we‟d like the type names to be singular and the Entity Set Names to be plural (e.g., the Customer type lives in the Customers entity set). We‟ll address that presently. 11. In the lower part of the details pane see the property details for the type selected in the upper part. Select the Order type in the upper partition of the details pane. Scroll through and note the many pieces of information available about the Order object‟s properties that are visible or modifiable on the Simple Properties tab. 12. Select the Associations tab and note the associations discovered in the Entity Data Model. (These, in turn, were created there because of the discovery of foreign key relations in the NorthwindIB database.)

55 | P a g e

IdeaBlade DevForce

Hello DevForce

We see associations between Order and Customer, Order and Employee, and Order-OrderDetails. 13. Select the Navigation Properties tab. These properties were discovered in the Entity Data Model, where they were generated as a consequence of the discovered relationships among the database tables. The Entity Data Model designer named the navigation properties according to the name of their source table. But again, we might well be a bit uncomfortable with the generated names, and wish that navigation properties that return a collection had plural names, and ones that return a single object had singular names. 14. We‟ve already noted the navigation properties on the Employee entity (which arose from the Employee table‟s relationship to itself), and their confusing names. We could rename those properties now, but first let‟s address the pluralization issues in the names in the model in a global way. 15. In the tree control pane, select the ServerModelNorthwindIB.edmx node. Click the <Name Pluralizer> button in the Entity Model Settings section. That launches the following model dialog:

16. Examine the default settings. By default, this tool will make Entity Set and Collection Navigation Property names plural, but will make Entity Type names and Scalar Navigation Property names singular, regardless of what they are now. Accept the default settings and click OK.

56 | P a g e

IdeaBlade DevForce

Hello DevForce

17. Select the ServerModelNorthwindIBContext node in the tree control again and observe that the Entity Set Names are now plural. Look at the Navigation Properties for the Order type and observe that the property for OrderDetails is now named in the plural, whereas the others, which do not return collections, have been left singular.

18. Examine the navigation properties for the Employee entity. Running the Pluralizer changed them from “Employee1” and “Employee2” to “Employee1s” and “Employee2”. Now it‟s easy to see which one returns the collection, and which returns a scalar. Rename “Employee1s” to “DirectReports” and “Employee2” to “Manager”.

19. Let‟s make one more manual name change. On the Simple Properties tab, change the name of the “Freight” property to “FreightCost”. (Press ENTER to complete the change.) You can change the name of any property, or any entity. 20. Save the Domain Model. Whenever you save again after having done so previously, the following things happen: a. The code in the file DomainModel.ServerModelNorthwindIB.Designer.cs gets overwritten;

57 | P a g e

IdeaBlade DevForce b. c. The individual files containing the partial “developer” classes are left alone; and,

Hello DevForce

Depending upon the nature of your changes, either or both of the DomainModel.ibedmx and app.config files may be updated.

Build and Add a Second Entity Model to Your Domain Model
[If you already know that your domain model will only get data from a single datasource, or you just want a more streamlined introduction to DevForce, feel free to skip this section. If you’re following along in Visual Studio, note that the following discussion provides less step-by-step detail than the previous one.] 1. Now add a second Entity Model to the existing Domain Model. Our second Entity Model is based on the Adventureworks2000 database from Microsoft, but uses only the tables Address, CountryRegion, Department, Employee, and StateProvince.4 We‟ll name it ServerModelAw2000. Our ServerModelAw2000 model looks something like the following when viewed in the Visual Studio Entity Model Designer:

2.

Note that the Employee entity for Adventureworks has an association to itself, just as the Employee entity in NorthwindIB did. So it naturally has the same naming issues that we saw in that model for the navigation properties that result from the recursive relationship. We‟ll fix them in the same way!

4

Note that the selection of Adventureworks2000 as the data source for our second Entity Data Model example was driven much more by its likely familiarity and availability to you, the reader, than by any actual relevance it has as an extension to NorthwindIB in a practical domain model. A more likely real-world scenario would be (as an example) one in which product inventory information resides in one database and accounting information in another, with both types of information being required by a target application.

58 | P a g e

IdeaBlade DevForce

Hello DevForce

3.

Select any other entity in the diagram and inspect its properties. Each Entity Set Name has been defaulted to the same name as that used for the entity (e.g., the Entity Set for the Address type is named “Address”).

Don‟t do anything about the names at this time: we can address them more easily in the DevForce Object Mapper (as we‟ve already seen). Close the Entity Data Model based on the AdventureWorks2000 database.

Add the Adventureworks Entity Model to the DevForce Domain Model
4. From the Visual Studio menu, select Tools / DevForce Object Mapper. The Object Mapper will launch with the existing domain model loaded.

5.

Right-click the DomainModel.ibedmx node and select “Add Entity Model” from the shortcut menu. In the File Open dialog, navigate to the ServerModelAw2000.edmx entity model, as necessary. When you open it, you‟ll see the following message:

59 | P a g e

IdeaBlade DevForce

Hello DevForce

This message results from the fact that the new Entity Model contains a type, “Employee”, whose name is already being used in the domain model. (Remember, the ServerModelNorthwindIB model also contains an Employee type.) Respond by clicking the <Yes> button to allow the Object Mapper to rename the incoming class to resolve the conflict. 6. Find and select the ServerModelAw2000Context node in the Object Mapper and inspect the imported model. Note the following: a. The Employee from AdventureWorks was renamed to “Employee1” to resolve the conflict with the Employee from NorthwindEF. All Entity Set Names are singular, as currently defined in the Entity Model (.edmx) file.

b.

7.

Double-click the Employee1 type in the upper right pane of the Object Mapper (titled “ServerModelAw2000Context”) and rename it to “EmployeeAw2000”. Note that all of these names are maintained in the Entity Model (.edmx) file, so changes will be written to that file, and will be recognized by the Visual Studio Entity Data Model Designer. In the tree control, re-select the “ServerModelAw2000.edmx” node. In the Entity Model Settings pane, click the <Name Pluralizer> button. You‟ll see a dialog like the one you saw previously:

8.

9.

And again, if you accept all of the default, the Object Mapper will change    Entity Sets names so that they are plural Entity Type names so that they are singular Navigation properties that return a single object so that they are singular

60 | P a g e

IdeaBlade DevForce  Navigation properties that return a collection of objects so that they are plural.

Hello DevForce

Click <OK> to accept the defaults and allow the Object Mapper to fix up names in your model. 10. Perform the same fixup on the navigation properties for the EmployeeAw2000 entity that you did for the Employee entity in the NorthwindIB model. Rename the “Employee1s” property to “DirectReports”; rename the “Employee2” property to “Manager”.

11. Again select the ServerModelAw2000Context node in the tree control. Note the new Entity Set Names. All are plural and look good except the name for the EmployeeAw2000 type. It was left unchanged because the Object Mapper noticed that the type name and EntitySet name were different to begin with. Double-click the Entity Set Name “Employees1” and change it to “EmployeesAw2000”.

Save the Enhanced Domain Model
12. Select File / Save from the Object Mapper menu. The Object Mapper generates a second “designer” class file under the DomainModel.ibedmx file for the new AdventureWorks2000 entities.

Since the “Create developer classes” checkbox is still checked, it also generates developer partial classes for the new entities: Address, CountryRegion, Department, EmployeeAw2000, and StateProvince. All entities from both data sources are now part of the same domain model.

61 | P a g e

IdeaBlade DevForce

Hello DevForce

...and your solution tree should something like that shown at right. Note that we‟ve moved both the ServerModelNorthwindIB and ServerModelAw2000 projects into the DomainModel Folder so that all parts of our business model are there. The organization of the various model projects under the “DomainModel Folder” solution folder is, of course, a matter of preference and not a necessity. It does, however, make it easy to collapse the model and all of its parts in the solution tree when you want to concentrate on the front-end (or some other) project.

Add Business Logic
Since we elected to generate “developer partial classes” for each entity in our Domain Model, those entities are each now represented by two partial classes – the elective one (stored in a stand-alone file, e.g., Employee.cs), and one in the “designer” code file associated with the ibedmx (e.g., DomainModel.ServerModelNorthwindIB.Designer.cs.) The partial class in the designer file is subject to frequent regeneration by the Object Mapper, and as such isn‟t the place to put custom logic – or to make any sort of manual changes The partial class in the stand-alone file, on the other hand, is expressly designed as the venue for such customization. Let‟s add some custom logic to the Employee partial class in Employee.cs.

62 | P a g e

IdeaBlade DevForce 1.

Hello DevForce

View the code in Employee.cs (or .vb), and find the Suggested Customizations region. Just before the #endregion statement (#End Region in VB), add the following code:

C#

/// <summary> /// Age as of today /// </summary> public int Age { get { if (null == this.BirthDate) return 0; DateTime oBirthDate = (DateTime)this.BirthDate; DateTime oToday = DateTime.Today; int oAge = oToday.Year - oBirthDate.Year; if (oBirthDate.AddYears(oAge) > oToday) oAge--; if (oAge < 0) return 0; else return oAge; } }

VB

''' <summary> ''' Age as of today ''' </summary> Public ReadOnly Property Age() As Integer Get If Nothing Is Me.BirthDate Then Return 0 End If Dim oBirthDate As Date = CDate(Me.BirthDate) Dim oToday As Date = Date.Today Dim oAge As Integer = oToday.Year - oBirthDate.Year If oBirthDate.AddYears(oAge) > oToday Then oAge -= 1 End If If oAge < 0 Then Return 0 Else Return oAge End If End Get End Property

This code defines a calculated property, Age, which returns the Employee‟s current age by calculating it from their birthdate. 2. C# Below the Age property, add a second one:
/// <summary> /// Total revenue for this Employee's orders /// </summary> public double TotalOrderRevenue { get { double revenue = 0; foreach (Order aOrder in this.Orders) { foreach (OrderDetail aOrderDetail in aOrder.OrderDetails) { revenue += aOrderDetail.Quantity * Convert.ToDouble(aOrderDetail.UnitPrice) * aOrderDetail.Discount;

63 | P a g e

IdeaBlade DevForce
} } return revenue; } }

Hello DevForce

VB

''' <summary> ''' Total revenue for this Employee's orders ''' </summary> Public ReadOnly Property TotalOrderRevenue() As Double Get Dim revenue As Double = 0 For Each aOrder As Order In Me.Orders For Each aOrderDetail As OrderDetail In aOrder.OrderDetails revenue += aOrderDetail.Quantity * Convert.ToDouble(aOrderDetail.UnitPrice) * aOrderDetail.Discount Next aOrderDetail Next aOrder Return revenue End Get End Property

This property uses navigation properties on the Employee and Order entities, automatically generated by the Entity Framework and carried through into the DevForce entities, to roll up the revenue from each line item of each order written by the Employee. 3. Let‟s add one more bit of custom logic to see how we can modify the behavior of even those properties that map directly to a back-end datasource and were therefore written into the generated class in DomainModel.ServerModelNorthwindIB.Designer.cs. We‟ll make a simple adjustment: we‟ll convert the Employee‟s LastName value to upper-case for display purposes while allowing entered or changed values to retain whatever capitalization the end user entered. In other words, we want a value stored as “Davolio” in the back-end datasource to be returned as “DAVOLIO” when we ask for it from the Employee object. First let‟s have a look at that generated code. There doesn‟t appear to be much there, but it has an amazing amount of flexibility built into it, behind the scenes: C#
#region LastName property /// <summary>Gets or sets the LastName. </summary> [Bindable(true, BindingDirection.TwoWay)] [Editable(true)] [Display(Name="Last Name", AutoGenerateField=true)] [IbVal.ValidateProperty] [IbVal.StringLengthVerifier(MaxValue=30, IsRequired=true)] [IbCore.MaxTextLength(30)] [MsSer.DataMember] public String LastName { get { return LastNameEntityProperty.GetValue(this); } [System.Diagnostics.DebuggerNonUserCode] set { LastNameEntityProperty.SetValue(this, value); } } #endregion LastName property

VB

#Region "LastName property"

64 | P a g e

IdeaBlade DevForce
'''<summary>Gets or sets the LastName. </summary> <Bindable(true, BindingDirection.TwoWay)> _ <Editable(true)> _ <Display(Name:="Last Name", AutoGenerateField:=true)> _ <IbVal.ValidateProperty> _ <IbVal.StringLengthVerifier(MaxValue:=30, IsRequired:=true)> _ <IbCore.MaxTextLength(30)> _ <MsSer.DataMember> _ Public Property LastName() As String Get Return LastNameEntityProperty.GetValue(Me) End Get <System.Diagnostics.DebuggerNonUserCode> _ Set LastNameEntityProperty.SetValue(Me, value) End Set End Property #End Region

Hello DevForce

An important part of this flexibility is provided by DevForce‟s support for methods known as property interceptors. Let‟s implement one now. 4. C# Enter the following code below the TotalRevenueCost property that you most recently added:
[AfterGet(EntityPropertyNames.LastName)] public void UppercaseLastName(PropertyInterceptorArgs<Employee, String> args) { var lastName = args.Value; if ( !String.IsNullOrEmpty(lastName)) { args.Value = args.Value.ToUpper(); } }

VB

<AfterGet(EntityPropertyNames.LastName)> _ Public Sub UppercaseLastName(ByVal args As PropertyInterceptorArgs(Of Employee, String)) Dim lastName = args.Value If Not String.IsNullOrEmpty(lastName) Then args.Value = args.Value.ToUpper() End If End Sub

UppercaseLastName() simply converts the current value of the LastName property (passed to the method in args.Value) as desired, and the work is done. The [AfterGet] attribute with which the public method UppercaseLastName() is decorated tells DevForce to call that method during any get operation for the designated property, just after the raw value is retrieved from the local instance of the object. The static5 EntityPropertyNames.LastName property, included in the attribute, simply returns the string-valued name of the LastName property. (The EntityPropertyNames class was automatically created by DevForce during Object Mapper code generation so you don‟t have to hard-code property names and risk misspelling them, in this and other contexts.) You can call the method anything you like (since the [AfterGet] attribute defines its role), but the signature does requires the args parameter, which is an instance of IPropertyInterceptorArgs. The version used here employs the generic version of IPropertyInterceptor, which fully specifies the type of both the property and its containing entity,

5

“Shared” in VB

65 | P a g e

IdeaBlade DevForce

Hello DevForce

so that you need not cast Args.Value within the method code. The compiler already knows (in this example) that it‟s a string.

Add a User Interface
Let‟s implement a quick console app to use as a front-end for our application so we can see the results of our custom work on the Employee class. We‟re choosing a console app as a UI, for now, simply for its simplicity. Later in this example we add a Winforms UI; and then a Silverlight UI. A WPF UI is yet another available option. Examples of all four UI types are included in the Learning Units that install with DevForce. 1. 2. 3. 4. C# Compile your DomainModel project. Add a new project, a Console Application, naming it “Console01”. To Console01, add references to IdeaBlade.EntityModel and to the DomainModel project. Add the statements shown in bold red below to the Main() method in Program.cs so that it looks as follows:
using using using using System; System.Collections.Generic; System.Linq; System.Text;

using IdeaBlade.EntityModel; using DomainModel; namespace Console01 { class Program { static void Main(string[] args) { GetEmployees(); } private static void GetEmployees() { var query = _manager.Employees; foreach (Employee anEmployee in query) { Console.WriteLine("Last Name = " + anEmployee.LastName); Console.WriteLine("\tBirth date = " + anEmployee.BirthDate.ToString()); Console.WriteLine("\tAge = " + anEmployee.Age); Console.WriteLine(String.Format("\tTotal Order Revenue: {0:C}", anEmployee.TotalOrderRevenue)); Console.WriteLine(); } PromptToContinue(); } private static void PromptToContinue() { Console.WriteLine(); Console.WriteLine("Press ENTER to continue..."); Console.ReadLine(); } #region Private Fields static DomainModelEntityManager _manager = DomainModelEntityManager.DefaultManager; #endregion }

66 | P a g e

IdeaBlade DevForce
}

Hello DevForce

VB

Imports IdeaBlade.EntityModel Imports DomainModel Module Module1 Sub Main() GetEmployees() End Sub Private Sub GetEmployees() Dim query = _manager.Employees For Each anEmployee As Employee In query Console.WriteLine("Last Name = " & anEmployee.LastName) Console.WriteLine(vbTab & "Birth date = " & anEmployee.BirthDate.ToString()) Console.WriteLine(vbTab & "Age = " & anEmployee.Age) Console.WriteLine(String.Format(vbTab & "Total Order Revenue: {0:C}", anEmployee.TotalOrderRevenue)) Console.WriteLine() Next anEmployee PromptToContinue() End Sub Private Sub PromptToContinue() Console.WriteLine() Console.WriteLine("Press ENTER to continue...") Console.ReadLine() End Sub #Region "Private Fields" Private _manager As DomainModelEntityManager = DomainModelEntityManager.DefaultManager #End Region End Module

5.

Make Console01 the Startup Project for your DevForce01 solution. Compile and then run the app. You should see output similar to the following:

67 | P a g e

IdeaBlade DevForce

Hello DevForce

Note that, for each Employee, the LastName is being converted to upper case; the Age is being computed properly from the birth date; and the Total Order Revenue is being rolled up across all Orders and their line items!

6.

Press the ENTER key to end the app.

Add Unit Tests
We sheepishly dodged the “Test First!” methodology. But we‟re not going to skip the tests altogether. We‟re going to lay them in right now. Do write unit tests as you go!

Unit Testing with Visual Studio Team Test
Do you have Team Test? Look for “Test” among the Visual Studio menus. If it‟s there, you‟ve got Team Test; if it isn‟t, you don‟t. For now you should skip ahead. Before you do, think about how you will test your application. If you can‟t afford Team Test, you might consider NUnit. It‟s solid and it‟s free. While it won‟t generate the template tests (we‟ll see Team Test do that shortly), it is otherwise remarkably similar to testing with Team Test. It‟s easy to get started with the new Visual Studio Team Test. 1. Open Employee.cs (or .vb) in code view, and right-click in the Code View window. 2. Pick “Create Unit Tests …” The “Create Unit Tests” Dialog appears.

68 | P a g e

IdeaBlade DevForce

Hello DevForce

3. Un-check the DomainModel assembly. 4. Expand the DomainModel.Employee node, and select the following items for testing:   Employee() (the parameterless constructor) Age (one of our calculated properties)

5. For “Output Project”, accept “Create a new Visual C# [or VB] Test Project”...

6. Click [OK] 7. Enter “DomainModel.Test” as the project name, when asked, and click <Create>. After some grinding in the background, you will see the DomainModel.Test project added to the solution, with references to the DomainModel and the necessary IdeaBlade assemblies. You will also see an EmployeeTest.cs (or .vb) tab and an AuthoringTests.txt tab. AuthoringTests is a non-functional introduction to testing. It‟s well worth reading at some point, but for now, just close it.

69 | P a g e

IdeaBlade DevForce

Hello DevForce

Configure the Test Project
We‟re almost ready to run our test, but we have just a couple of configuration changes to make the test project ready to run. 1. Make sure the following using [C#] or Imports [VB] statements at the top of the Employee.Test.cs (or .vb) file:

C#

using using using using using

DomainModel; Microsoft.VisualStudio.TestTools.UnitTesting; System.Collections.Generic; System.Linq; IdeaBlade.Core;

VB

Imports Imports Imports Imports Imports

DomainModel Microsoft.VisualStudio.TestTools.UnitTesting System.Collections.Generic System.Linq IdeaBlade.Core

2. 3.

Build or rebuild the entire solution. Set LocalTestRun.testrunconfig to deploy the app.config to the test output directory. You can do this as follows: a. On the Visual Studio Test menu, select Edit Test Run Configurations and then Local Test Run (localtestrun.testrunconfig). On the localtestrun.testrunconfig dialog, select Deployment. Click <Add File...>. Change the Objects of type select to “All files(*.*)”, and find the app.config file in the folder for the solution‟s executable project (DevForce01\Console01), select it, and click the <Open> button. The localtestrun.testrunconfig dialog should then appear as follows:

b. c.

70 | P a g e

IdeaBlade DevForce

Hello DevForce

Note that there is no need to set the ConfigFileLocation in code (e.g., in one of the test initializer methods), because, as in any DevForce app, DevForce automatically searches the main application folder (BaseAppDirectory) for an app.config file. It will therefore find the one deployed in the test deployment folder (TestDeploymentDir).6 4. Next we need to ensure that the server model assembly will be deployed to the test result directory. Click <Add File...> again. Navigate to the bin\debug folder under the project folder for your Entity Data Model project (the ServerModelNorthwindIB folder, in our case). Select the ServerModelNorthwindIB assembly:

6

You can find detailed information about how DevForce finds configuration in the appendix to this chapter entitled “Probing Sequence for the App.Config File”.

71 | P a g e

IdeaBlade DevForce

Hello DevForce

5.

If you have elected to have the Entity Data Model artifacts stored as loose files rather than to be embedded in the server model assembly (as is the default), then you will also need to add the three components of the ServerModelNorthwindIB model (the .ssdl, .csdl, and .msl files). You will find these in the same folder where you found the model assembly. On the localtestrun.testrunconfig dialog, click <Close> and respond Yes when prompted whether to save changes to the testrunconfig.

6.

EmployeeTest First Look
The action is in EmployeeTest.cs (or .vb). It is a little forbidding at first but we‟ll clear that up in a few quick steps. 1. Ignore the TestContext property (perhaps by wrapping in its own region). 2. Close the “Additional Test Attributes” region if it‟s open – we don‟t need it yet. 3. Find the EmployeeConstructorTest (). Rework it to look like the following:

C#

/// <summary> ///A test for Employee Constructor ///</summary> [TestMethod()] public void EmployeeConstructorTest() { Employee target = new Employee(); string desiredTypeName = "Employee"; string actualTypeName = target.GetType().Name; Assert.IsTrue(actualTypeName == desiredTypeName, "Created type, '" + actualTypeName + "', is incorrect. It should be '" + desiredTypeName + "'."); } ''' <summary>

VB

72 | P a g e

IdeaBlade DevForce

Hello DevForce

'''A test for Employee Constructor '''</summary> <TestMethod()> _ Public Sub EmployeeConstructorTest() Dim target As New Employee() Dim desiredTypeName As String = "Employee" Dim actualTypeName As String = target.GetType().Name Assert.IsTrue(actualTypeName = desiredTypeName, "Created type, '" & actualTypeName & "', is incorrect. It should be '" & desiredTypeName & "'.") End Sub

That leaves just the Age property that we added. The following test method is automatically generated for that: C#
/// <summary> ///A test for Age ///</summary> [TestMethod()] public void AgeTest() { Employee target = new Employee(); // TODO: Initialize to an appropriate value int actual; actual = target.Age; Assert.Inconclusive("Verify the correctness of this test method."); }

VB

'''<summary> '''A test for Age '''</summary> <TestMethod()> _ Public Sub AgeTest() Dim target As Employee = New Employee ' TODO: Initialize to an appropriate value Dim actual As Integer actual = target.Age Assert.Inconclusive("Verify the correctness of this test method.") End Sub

The newly created Employee won’t have an Age property value that’s of much interest. Let’s fix up the test so it uses an actual employee in our test database.

Get a Test Employee
We will want to be careful to keep the test data always in a known state. If we make changes, we‟d better restore the original data even if our tests fail. We should have a back up of the database just in case. One of the cardinal testing rules is that there should be no dependencies among the tests. That means that tests can be run independently and in any order. Accordingly, if we make changes during a test, we must restore the original data immediately after the test. We must make sure we do so even if the test fails. Do restore the test data source after every test Do maintain a back up of the database just in case.

73 | P a g e

IdeaBlade DevForce

Hello DevForce

Our example uses the NorthwindIB database as its data source. It has well-known data and we‟re not going to make any changes just yet, so we‟re fine. 1. We‟re not going to be purists now – this is just a tutorial – so we‟ll be sloppy with our first test. Let‟s rewrite it so it looks like this:

C#

/// <summary> ///A test for Age ///</summary> [TestMethod()] public void AgeTest() { DomainModelEntityManager manager = DomainModelEntityManager.DefaultManager; Employee target = manager.Employees .Where(e => e.LastName.ToLower() == "Davolio".ToLower()).ToList().First<Employee>(); int actual = target.Age; int approxAge = System.DateTime.Now.Year - target.BirthDate.Value.Year; Assert.IsTrue(approxAge - actual <= 1);}

VB

''' <summary> '''A test for Age '''</summary> <TestMethod()> _ Public Sub AgeTest() Dim manager As DomainModelEntityManager = DomainModelEntityManager.DefaultManager Dim target As Employee = manager.Employees _ .Where(Function(e) e.LastName.ToLower() = "Davolio".ToLower()).ToList().First() Dim actual As Integer = target.Age Dim approxAge As Integer = Date.Now.Year - target.BirthDate.Value.Year Assert.IsTrue(approxAge - actual <= 1) End Sub

Run the Test
1. 2. Open the Test View from the menu Test ► Windows ► Test View. Run the selected AgeTest test by selecting it and clicking the <Run Selection> button, as shown at right.

A “Test Results” window appears at the bottom of Visual Studio. The test passes!

74 | P a g e

IdeaBlade DevForce

Hello DevForce

Just FYI, in case you‟re new to Team Test: you can also run the tests in debug mode so you can insert breakpoints. Instead of starting the test with the <Run Selection> button, to run in debug mode, right-click the test and select Debug Selection.

Accumulating Test Results
Team Test keeps track of every test run. You can review them from within the “Test Results” window shown immediately above.

When you examine your solution directory in Windows Explorer, you will find the TestResults directory where these results are stored. You don‟t want this directory under source control and you definitely want to clear it out periodically. Prune or delete at your leisure.

Lessons Learned
We have the foundation for testing the logic we add to our business entities. It didn‟t take long to set up a test environment. Now it‟s just a matter of keeping it up. There is far more to learn about testing than we can cover in this Guide. Check the “Suggested Reading” chapter in the Concepts Manual for our recommendations. There is no end to the information available on the web.

Add a WinForm UI
In Visual Studio 2008, you have many options for a UI. You can use Winforms, WebForms, WPF, or Silverlight. If you choose Winforms, you have the option to use the UI-related assemblies of DevForce to grease the wheels. We‟ll see what that looks like now.

Add a Windows Forms Application Project
1. In the Solution Explorer, right-click the solution node and select a new Windows Forms Application. Name the project “WinForms01”. Set a reference in the WinForms01 project to the DomainModel project so our form can see the entities in the domain model. Set references to IdeaBlade.Core and IdeaBlade.EntityModel and so you can use the DevForce types you‟ll need. Finally, set a reference to the .NET assembly WindowsBase, as it contains an interface used by the DevForce BindableList<T> type that you‟ll use in data binding.

2.

75 | P a g e

IdeaBlade DevForce

Hello DevForce

3.

In the designer, select Form1. In the Properties window, change the Text property of the Form from “Form1” to “Employees”. In the Solution Explorer, rename the file containing the form from “Form1.cs” (or .vb) to “EmployeeForm.cs” (or .vb). Say yes when asked whether you want to rename references to the form in the project:

4.

Find the BindingSource control in the Toolbox (“All Windows Forms” group) and drag two of them on to the form. Name one of them “_ordersBindingSource” and the other “_employeesBindingSource”. (We‟re using the underscore prefix “_” for elements that will be scoped at the class level in our form, as these will.) Drag a BindingNavigator from the ToolBox to the form, positioning it along the top edge. Name that “_employeesBindingNavigator”. Find the “IdeaBlade DevForce” group in the Toolbox. (Remember, it won‟t be there unless you installed the DevForce WinForms UI support when you installed DevForce!) Drag a ControlBindingManager and then a DataGridViewBindingManager on to the form. Rename the ControlBindingManager “_employeesControlBindingManager”; rename the DataGridViewBindingManager “_ordersDataGridViewBindingManager”. In the component tray, select the _employeesControlBindingManager. Then, in the Properties window, assign the _employeesBindingSource to the BindingSource property for the _employeesControlBindingManager.

5.

6.

7.

76 | P a g e

IdeaBlade DevForce 8. Similarly, assign the _ordersBindingSource to the BindingSource property for the _ordersDataGridViewBindingManager.

Hello DevForce

9.

In the component tray, select the _employeesControlBindingManager. Next, right-click the smart tag that shows up at its upper right corner, and choose Autopopulate Controls. In the dialog “Bind to which object type?”, select the DomainModel assembly and the Employee type, then click <Ok>. You‟ll see the DevForce “Configure Databindings” designer:

10. From the Properties tree control, drag the following properties onto the Autopopulation tab: LastName, FirstName, BirthDate, Age, and Photo. 11. Click the Naming Conventions button and add an underscore in front of the default text of “{0}{1}” to make it “_{0}{1}”. Click the <Update Sample> button if you want to see what it will look like. Then click <OK>.

77 | P a g e

IdeaBlade DevForce

Hello DevForce

12. On the “Configure DataBindings” dialog, click <OK> to close it and autopopulate the form with controls for your selected properties. Rearrange the controls as desired. We‟ll move the PictureBox for the Photo to the right of the other controls, and delete its label. 13. Now select the _ordersDataGridViewBindingManager, click the Smart Tag, select “Configure Databindings”, select DomainModel as the assembly containing your target types, and select Order as the target type. Into the grid, drag the properties OrderDate, RequiredDate, ShippedDate, and FreightCost. 14. Expand the Customer property in the tree control and drag the CompanyName property of the Customer onto the grid. Drag it upward to make it the top row. Edit the value in the “Column Title” column from “Customer Company Name” to “Company”. 15. Click <OK> to close the dialog. Whoops! The designer warns you that you haven‟t linked your selected properties to a grid (see picture below). Response that <No>, you don‟t want to exit just yet.

16. Click the <Create Grid> button, and name the grid to be created “_ordersDataGridView”. Click <OK>, then click <OK> on the main dialog again. (If you get prompted asking whether to enlarge the form to accommodate the new control, say yes.) The designer configures a DataGridView control and plops it on to your form. Position and size it as desired. Your form should now look something like the following:

78 | P a g e

IdeaBlade DevForce

Hello DevForce

17. There‟s more configuration we could do in the designer, but we‟ll choose to do it in the “code behind” for our form, instead. Make that look as follows:

C#

using using using using using using using using

System; System.Collections.Generic; System.ComponentModel; System.Data; System.Drawing; System.Linq; System.Text; System.Windows.Forms;

using IdeaBlade.EntityModel; using IdeaBlade.Util; using DomainModel;

namespace WinForms01 { public partial class EmployeeForm : Form { public EmployeeForm() { InitializeComponent(); this.Load+=new EventHandler(EmployeeForm_Load); } private void EmployeeForm_Load(object sender, EventArgs e) { ConfigureBindingSources(); ConfigureBindingNavigators(); ConfigureBindingManagers(); ConfigureHandlers(); LoadData(); } private void ConfigureBindingSources() { _employeesBindingSource.DataSource = _employees;

79 | P a g e

IdeaBlade DevForce
_ordersBindingSource.DataSource = _orders; } private void ConfigureBindingNavigators() { _employeesBindingNavigator.BindingSource = _employeesBindingSource; } private void ConfigureBindingManagers() { _employeesControlBindingManager.BindingSource = _employeesBindingSource; _ordersDataGridViewBindingManager.BindingSource = _ordersBindingSource; } private void ConfigureHandlers() { _employeesBindingSource.CurrentChanged += new EventHandler(_employeesBindingSource_CurrentChanged); } void _employeesBindingSource_CurrentChanged(object sender, EventArgs e) { Employee currentEmployee = (Employee)_employeesBindingSource.Current; _orders.ReplaceRange(currentEmployee.Orders); } private void LoadData() { _employees.ReplaceRange(_manager.Employees); } #region Private Fields DomainModelEntityManager _manager = DomainModelEntityManager.DefaultManager; BindableList<Employee> _employees = new BindableList<Employee>(); BindableList<Order> _orders = new BindableList<Order>(); #endregion } }

Hello DevForce

VB

Imports Imports Imports Imports Imports Imports Imports Imports

System System.Collections.Generic System.ComponentModel System.Data System.Drawing System.Linq System.Text System.Windows.Forms

Imports IdeaBlade.EntityModel Imports IdeaBlade.Util Imports DomainModel

Partial Public Class EmployeeForm Inherits Form Public Sub New() InitializeComponent() AddHandler Load, AddressOf EmployeeForm_Load End Sub Private Sub EmployeeForm_Load(ByVal sender As Object, ByVal e As EventArgs) ConfigureBindingSources() ConfigureBindingNavigators() ConfigureBindingManagers() ConfigureHandlers()

80 | P a g e

IdeaBlade DevForce
LoadData() End Sub Private Sub ConfigureBindingSources() _employeesBindingSource.DataSource = _employees _ordersBindingSource.DataSource = _orders End Sub Private Sub ConfigureBindingNavigators() _employeesBindingNavigator.BindingSource = _employeesBindingSource End Sub Private Sub ConfigureBindingManagers() _employeesControlBindingManager.BindingSource = _employeesBindingSource _ordersDataGridViewBindingManager.BindingSource = _ordersBindingSource End Sub Private Sub ConfigureHandlers() AddHandler _employeesBindingSource.CurrentChanged, AddressOf _employeesBindingSource_CurrentChanged End Sub

Hello DevForce

Private Sub _employeesBindingSource_CurrentChanged(ByVal sender As Object, ByVal e As EventArgs) Dim currentEmployee As Employee = CType(_employeesBindingSource.Current, Employee) _orders.ReplaceRange(currentEmployee.Orders) End Sub Private Sub LoadData() _employees.ReplaceRange(_manager.Employees) End Sub #Region "Private Fields" Private _manager As DomainModelEntityManager = DomainModelEntityManager.DefaultManager Private _employees As New BindableList(Of Employee)() Private _orders As New BindableList(Of Order)() #End Region End Class

Make “WinForms01” the start-up Project
18. In the Solution Explorer, right-click the Winforms01 project node and select Set As Startup Project.

Run It!
19. Build and run your app.

81 | P a g e

IdeaBlade DevForce

Hello DevForce

Please note: The code solutions that accompany this document in the Learning Resources reflect the work to this point, with the WinForm user interface.

Add a WPF UI
Please see the Learning Units that install with DevForce for examples of a DevForce app with a WPF user interface.

Create a Silverlight App with DevForce
Features described in the section are included with the DevForce Silverlight product. Re-open the Object Mapper and check the “Generate Silverlight Projects” checkbox:

82 | P a g e

IdeaBlade DevForce

Hello DevForce

Doing so causes a ComboBox for the project name to appear, and a button to create a new project. Click the <New Project…> button to see the Create Project dialog:

83 | P a g e

IdeaBlade DevForce

Hello DevForce

We‟ll accept the default name for the project, “DomainModelSL”, by clicking <OK>. Then we will again save our Domain Model. After closing the Object Mapper, we see that it has created a new project named “DomainModelSL”.

Now that we have our Silverlight copy of the DomainModel, it‟s time to add the executable project. .NET provides a Silverlight Application project template, but that‟s not what you want:

84 | P a g e

IdeaBlade DevForce

Hello DevForce

Why not? Because we‟ve provided a DevForce-aware template that will get you there a lot faster, and with much less trouble. Under the Visual C# (or VB) project types, find the DevForce section and select the “DevForce Silverlight Application” template. We‟ll name our project “DevForceSilverlight01”:

Visual Studio creates a Silverlight project of the specified name. It also creates a web project to host the Silverlight app, naming it “<appname>Web”, or in our example, “DevForceSilverlight01Web”.

85 | P a g e

IdeaBlade DevForce

Hello DevForce

Actions of the DevForce Silverlight Application Project Template The “DevForce Silverlight Application” template, like the standard Visual Studio Silverlight Application template, creates a Silverlight application project and a web application project. But it also provides considerable DevForcespecific functionality to make it easy to host a BOS in the same web application project. In the web application project, it does all of the following:  includes two WCF service files for the BOS, EntityService.svc and EntityServer.svc;  includes a Global.asax showing how to check the IdeaBlade configuration information and to enable support for a trace viewer;  provides a web.config file that contains WCF ServiceModel configuration information for the BOS (the EntityService and EntityServer services), as well as a stub ideaBlade.configuration section;  configures the Default.aspx file so that it contains the Silverlight control, and skips the creation of Silverlight test pages;  creates a log folder to hold the debuglog.xml file  includes references to necessary IdeaBlade assemblies; and  selects the “specific port” setting and specifies 9009 as the default port. Using port 9009 is not a DevForce requirement, but it‟s a handy open port that can be used during development; and using a specific port rather than an auto-assigned one makes communicating with the BOS easier. In the Silverlight application project as well, the DevForce Silverlight Application project template includes references to needed IdeaBlade assemblies. The assembly and namespace names for both projects are set to the same value. This is necessary if you plan to place the Domain Model in the web application project and the linked (Silverlight) domain model in the Silverlight application project. Using the same assembly name and namespace allows DevForce seamlessly to transmit entities between the .NET and Silverlight environments. If your domain model will not be in these projects but in a separate

86 | P a g e

IdeaBlade DevForce

Hello DevForce

assembly, then that assembly and the assembly holding the linked SL model must have the same name and namespace. Which Startup Project? The DevForce Silverlight Application template, like the standard one, designates the web project as the Startup Project. That‟s important. If the Silverlight project is designated as the Startup Project, the application will start, and you‟ll probably see the start page; but the app, in that circumstance, is running under the file system instead of being served by a web server. When running under the file system it can‟t access a service such as the DevForce Business Object Server; so you will not, for example, be able to retrieve data. You can easily tell how your app is being hosted by looking at the browser‟s address bar. If it‟s being served by a web server, the URL in the address bar will start with http://.

If it‟s running from the file system, the URL in the address bar will look like a file path (and you‟ll see no evidence of retrieved data where you otherwise would have).

87 | P a g e

IdeaBlade DevForce

Hello DevForce

Resuming the Walk-Through... Launching the project, you should see the following display in your browser. (We‟ve made our default browser Mozilla Firefox to demonstrate the browser independence of Silverlight!) The text you see is displayed in a Silverlight TextBlock control hosted by Page.xaml in the Silverlight project. You‟re ready immediately to start building meaningful functionality into that page – including creating data bindings to your DevForce entities. See the Learning Units for samples of Silverlight / DevForce applications!

Specifying a Target Browser If you have more than one Silverlight-compatible browser installed on your computer, you can specify in which browser you would like your Silverlight app to launch.

88 | P a g e

IdeaBlade DevForce

Hello DevForce

Find the start page for your app (Default.aspx as generated by the project template ), right-click it, and select Browse With... Then select the desired browser and click the <Set as Default> button. You may then click <Browse> to launch the app, or <Cancel> if all you wanted to do was to change the setting. Either way, subsequent launches of the application will occur in the browser you specified.

This concludes our walk-through of the setup of a DevForce Silverlight application. To see more detailed sample Silverlight / DevForce applications, please consult the Learning Units that install with DevForce.

Understanding the App.Configs
You will soon discover that your Entity Framework / DevForce app includes many app.config files. Each has its necessary and particular role, and there is a flow of information between them. In this section we‟ll explain those roles and information flows.

89 | P a g e

IdeaBlade DevForce

Hello DevForce

The sample Visual Studio solution at right includes three copies of app.config, one in each of the following locations: 1. 2. 3. in the project for the Entity Data Model (#1); in the project for the DomainModel (#2); and in the executable project (#3)

We‟ve listed them in the above order because it reflects the flow of information between them. (We‟ll provide more detail on that momentarily.)

The App.Config in the Entity Data Model project (#1 in the picture) typically gets there by being generated by the Visual Studio Entity Data Model designer. It contains a configuration section with a connectionStrings element. For a sample, see Listing 1 in the Appendix “Listings of Sample App.Config Files” at the end of this chapter.

The app.config in the DomainModel project (#2) typically gets there by being generated by DevForce. It contains, most importantly, an ideaBlade.configuration section with edmKeys that include connection information and other settings related to specific data sources; and other settings that control or reflect such things as application logging behavior, location of the Business Object Server, etc. The connection information included in the edmKeys usually originates from the EDM project‟s app.config (#1), being copied from there by DevForce. However, the developer can add to it manually or using the DevForce Configuration Editor. 7 The app.config file associated with the Entity Data Model enables the Entity Data Model designer to find the EDM‟s datasource. As such, it is essential to the design-time utility of that designer. The app.config file associated with the Domain Model is updated at design time, but other than to be available for update, its contents have no design-time function. The same is true for the app.config associated with the executable project. The latter, however, becomes the (only) copy of app.config used at runtime. It is actually copied and renamed by Visual Studio to reflect the name of the assembly; for example, if the UI project above is used to create an assembly named UI.exe, then Visual Studio will create a copy of the app.config file in that UI project and name the copy UI.exe.config. The latter is the version used at runtime by the .NET framework.

7

“Config Editor” under IdeaBlade DevForce / Tools on the Windows Start menu.

90 | P a g e

IdeaBlade DevForce

Hello DevForce

Information Flow Between the App.Configs
The typical flow of information between the copies of app.config is as follows. 1. Information is written to the EDM app.config (#1), usually by the EDM designer. 2. The DomainModel app.config (#2) is updated using information from one or more EDM app.configs. This update occurs either upon a Save in the DevForce Object Mapper, or when the developer responds “Yes” to the following prompt, which is presented after the developer saves a change to the Entity Data Model8:

3.

The app.config associated with the executable project (#3) is updated using information from the DomainModel app.config. This occurs at build time 9 when DevForce discovers a mismatch between the information in the ideaBlade.configuration section of the executable project app.config and the corresponding information in the DomainModel copy (#2). Before updating app.config #3 DevForce prompts you as follows:

Note that, except for updating the ideaBlade.configuration and configSections elements of the executable project‟s app.config (#3), DevForce leaves the app.config alone. Thus, it can contain any other elements and information needed by the app; those will be left undisturbed.

8 9

DevForce watches the Visual Studio IDE for such a change, and responds when it occurs by presenting this prompt. Specifically, when the executable project is built. This is performed by a Build Event handler attached by DevForce to the executable project.

91 | P a g e

IdeaBlade DevForce

Hello DevForce

The flow of information between app.configs 1, 2, and 3 is summarized in the graphic at right. Content from the EDM app.config (#1) flows to the DomainModel app.config (#2). Content from the DomainModel app.config (#2) flows to the executable project app.config (#3). There is no “back flow” of information; and unrelated app.config content is preserved at each stage during updating.

Monitoring Activity
What is actually happening as we run the applications? When is it asking for data? What does the SQL look like?

SQL Profiler
We can always monitor activity on the SQL Server using SQL Profiler. Here we assume SQL Server 2005. Launch the SQL Server Management Console. Select “Tools ►SQL Server Profiler” from the menu. Select “File ►New Trace ..” from the Profiler menu and connect to your database server Click [Run] on the [Trace Properties] dialog. Return to Visual Studio and re-run the application [F5]

92 | P a g e

IdeaBlade DevForce The trace window fills, showing us exactly how we’re hitting the database.

Hello DevForce

DevForce DebugLog
We may want to supplement the SQL Profiler with a DevForce tool that helps us see both the database activity and the other application activity that surrounds it. DevForce applications generate a trace log10 every time you run the application. The log appears in the executable directory; its default name is “DebugLog.xml”11. Open Windows Explorer. Navigate to the ..\bin\debug directory under the UI directory. Launch DebugLog.xml. The log appears in a browser window.

Each row speaks of some event during the life of the last application run. You‟ll see database access events among other event occurring from the start of the application until it shuts down. You can launch the DebugLog while the application is running and refresh the browser from time to time to see how the log is progressing as you move through the application.

DevForce TraceViewer
The DevForce TraceViewer affords a friendlier and more dynamic look at logged activity. You can launch the TraceViewer from the IdeaBlade DevForce/Tools menu. It can also be linked directly into your application. See the Object Persistence chapter for details.

10 11

You can turn it off or filter it. There are companion .css and .xslt files in that directory as well so that the log displays in the browser nicely. You can rename the log in the App.Config file.

93 | P a g e

IdeaBlade DevForce

Hello DevForce

Appendix: Listings of Sample App.Config Files
Listing 1. App.Config associated with the Entity Data Model

XML

<?xml version="1.0" encoding="utf-8"?> <configuration> <connectionStrings> <add name="ServerModelNorthwindIBContext" connectionString="metadata=res://*/ServerModelNorthwindIB.csdl|res://*/ServerModelNorthw indIB.ssdl|res://*/ServerModelNorthwindIB.msl;provider=System.Data.SqlClient;provider connection string=&quot;Data Source=.;Initial Catalog=NorthwindIB;Integrated Security=True;MultipleActiveResultSets=True&quot;" providerName="System.Data.EntityClient" /> </connectionStrings> </configuration>

Listing 2. Copy of app.config associated with the DomainModel

XML

<?xml version="1.0" encoding="utf-8"?> <configuration> <configSections> <section name="ideablade.configuration" type="IdeaBlade.Core.Configuration.IdeaBladeSection, IdeaBlade.Core, Version=5.1.0.0, Culture=neutral, PublicKeyToken=287b5094865421c0" /> </configSections> <ideablade.configuration version="5.00" updateFromDomainModelConfig="Ask"> <logging logFile="DebugLog.xml" /> <objectServer isDistributed="false" remoteBaseURL="http://localhost" serverPort="9009" serviceName="EntityService" /> <edmKeys> <edmKey name="Default" connection="metadata=res://ServerModelNorthwindIB/ServerModelNorthwindIB.csdl|res://Se rverModelNorthwindIB/ServerModelNorthwindIB.ssdl|res://ServerModelNorthwindIB/ServerMo delNorthwindIB.msl;provider=System.Data.SqlClient;provider connection string=&quot;Data Source=.;Initial Catalog=NorthwindIB;Integrated Security=True;MultipleActiveResultSets=True&quot;" containerName="ServerModelNorthwindIB.ServerModelNorthwindIBContext"> <probeAssemblyNames> <probeAssemblyName name="DomainModel" /> <probeAssemblyName name="ServerModelNorthwindIB" /> </probeAssemblyNames> </edmKey> </edmKeys> </ideablade.configuration> </configuration>

94 | P a g e

IdeaBlade DevForce
Listing 3. Copy of app.config in the project for the executable

Hello DevForce

XML

<?xml version="1.0" encoding="utf-8"?> <configuration> <!—The following section is left alone by DevForce when it updates the app.config in the executable project --> <DeveloperAddedSection> <ArbitraryElementNo1 /> <ArbitraryElementNo2 /> </DeveloperAddedSection> <connectionStrings> ...snip </connectionStrings> <configSections> ...snip </configSections> <ideablade.configuration version="5.00" updateFromDomainModelConfig="Ask"> …snip </ideablade.configuration> </configuration>

Appendix: Probing Sequence for the App.Config File
Whenever a .NET application attempts to exercise any aspect of the DevForce API that requires DevForce configuration information and such configuration information has not already been found and placed into memory, DevForce will search for a configuration file that contains this information. Its probing path is as follows: 1. If the ConfigFileLocation property of the IdeaBladeConfig object has been set in the executing code, DevForce will search the indicated location for a file named or matching “*.exe.config”, “web.config” or “app.config”, in that order. Note that files matching “*.dll.config” will not be found. If the ConfigFileLocation property was not set, or a suitable config file was not found in the indicated location, then DevForce will look to see if the IdeaBladeConfig.ConfigFileAssembly property has been set. If so, it will look for an embedded resource named “app.config”. Next, the BaseAppDirectory is searched for a file named or matching “*.exe.config”, “web.config” or “app.config”, again in that order. This property is read-only and determined at run time, and is usually the directory from which the entry assembly was loaded. However, in a test project executable, which has no entry assembly, the AppDomain.BaseDirectory is used. In MSTest, this will be the TestContext.TestDeploymentDir. DevForce next searches for an embedded resource named “app.config” in the entry assembly. This is ignored in the case of a test project because the entry assembly is null.

2.

3.

4.

Finally, DevForce will look for an embedded resource named “app.config” in an assembly named “AppHelper”. This search is done for reason of backward compatibilility with DevForce applications that use the AppHelper project formerly generated by the DevForce Object Mapper.

95 | P a g e

IdeaBlade DevForce

Class Libraries

Class Libraries

Class Libraries
Important Namespaces The IdeaBlade.EntityModel.Entity Finding Help on DevForce
XML Documentation IntelliSense The Object Browser Class View Class Diagram

This chapter takes you on a brief tour of the DevForce libraries. The DevForce installer put a number of assembly DLLs in your installation directory and in the global assembly cache (GAC, pronounced “gack”). You do not have to put DevForce DLLs in the GAC on your application client machines. In most cases your installation will likely be via XCOPY or ClickOnce. You should turn to the Reference Help for the minute details about the DevForce classes, interfaces, methods, properties and events. That effort can be like looking through a keyhole. Here you can learn a bit about the room you‟re seeing through that keyhole. While an assembly can host multiple namespaces and a DLL can host multiple assemblies, each DevForce namespace is almost invariably in its own assembly and each assembly is in its own DLL. Thus, as our tour proceeds namespace-by-namespace, we are also walking from one assembly and DLL to the next. In this chapter we‟ll also give you some ideas about how to use Visual Studio to get to this information quickly.

Important Namespaces
All DevForce namespaces have an “IdeaBlade” prefix. The discussion below elides that prefix for brevity so, when we refer to “EntityModel”, we mean “IdeaBlade. EntityModel”. See the DevForce Reference Help or Consolidated Help for much more detail on the content of these namespaces.

96 | P a g e

IdeaBlade DevForce
DevForce Namespace
EntityModel …

Class Libraries
Summary Includes critical modeling and persistence classes including Entity, EntityManager, EntityQuery, EntityRelation, QueryCache, QueryStrategy, and many others. Includes classes for modelling, querying, and persisting Entities backed by web services. Includes classes to assist with databinding, encryption, debugging, i/o, localization, tracing, and other common development tasks. Includes classes related to DevForce‟s verification (validation) facilities. Includes the TraceViewer, a utility that provides a real-time display of DevForce tracing messages; and the TraceViewerForm, a window that provides an immediately display surface for trace messages from a DevForce app. Includes classes used for specific operations performed directly against relational databases. Can only be used from code executing on the middle-tier server.

EntityModel.WS … Core …

Validation … DevTools.WinTraceViewer …

Ado …

The IdeaBlade.EntityModel.Entity
The IdeaBlade.EntityModel.Entity is the base class of all of persistable business entities in your DevForce domain model. Instances of Entity are not created directly. Entity objects are managed and cached by a EntityManager. You'll use an EntityManager to create, retrieve and save your entities. The EntityManager will also handle serialization and transfer of entities to a distributed Object Server. When working with entity classes, basic properties for your business objects – those that map to columns of a database, for example -- are auto-generated for you by the IdeaBlade DevForce Object Mapping Tool. You focus on creating additional custom properties and methods to support business logic and rules.

The EntityAspect Property
Every instance of an entity has an EntityAspect property that is inherited from IdeaBlade.EntityModel.Entity. C# or VB
aCustomer.EntityAspect

Through the EntityAspect you have access to a rich set of (a) properties that define the entity‟s state, and (b) persistence-related methods. These are listed below: method names terminate with parentheses. See the DevForce reference help (IdeaBlade DevForce / Documentation / DevForce Help from the Windows Start Menu) for more detail.

97 | P a g e

IdeaBlade DevForce
Table 1. EntityAspect Members

Class Libraries

Member AcceptChanges() AddToManager() ClearErrors() Delete() Entity EntityGroup EntityKey EntityManager EntityMetadata

Function Accepts all changes to this Entity, returning the EntityState to Unchanged. Adds a newly created entity to its associated EntityManager. Clears any error messages on this Entity. Marks this Entity for deletion; the EntityState becomes "Deleted".

The EntityGroup that this Entity belongs to. An EntityGroup is a container that holds cached entities of a particular type. The EntityKey for this entity. Represents the primary key for an Entity. The EntityManager that manages this entity. The EntityMetadata for this Entity. The available EntityMetadata includes all of the following properties: Member Description

CanQueryByEntityKey Gets whether primary key queries are allowed. ComplexTypeProperties Returns a collection of EntityProperties that describe complex object properties for entities of this type. ConcurrencyProperties Returns a collection of EntityProperties that are concurrency properties for entities of this type. Returns a collection of DataEntityProperties for entities of this type. Gets the data source key name. The default EntitySetName for entities of this type. Returns a collection of EntityProperties that belong to entities of this type. Gets the Type of the entity. Returns whether this metadata describes a ComplexObject. Returns a collection of EntityProperties that are keys for entities of this type. Returns a collection of DataEntityProperties for entities of this type.

DataProperties

DataSourceKeyName DefaultEntitySetName EntityProperties

EntityType IsComplexType

KeyProperties

NavigationProperties

98 | P a g e

IdeaBlade DevForce

Class Libraries

Public Methods Name CreateEntity Description Creates a new entity of the type describe by this metadata item.

GetDefaultValue

EntitySetName EntityState Equals() FindRelatedEntities() ForcePropertyChanged()

The name of the Entity Framework EntitySet containing this entity. An enum signifying that the entity is Detached, Unchanged, Added, Deleted, or Modified. As defined on System.Object. Finds any cached entities related to this entity by a specified EntityRelationLink. Forces a PropertyChanged event to be fired. This method should only be needed in situations where changes to calculated fields or other properties not backed by an EntityProperty must be made known. Returns the EntityProperty corresponding to the specified property name. As defined on System.Object. Returns all related entities via a specified EntityRelationLink. Differs from FindRelatedEntities() in that it will retrieve the related entities from the back-end data source of if necessary. Generic version of GetRelatedEntities() Returns a related entity via an EntityRelationLink. Generic version of GetRelatedEntity() As defined on System.Object. Determines whether this entity has any pending changes. Returns a Boolean. Gets a boolean value indicating whether there are errors in this entity. Returns whether the current instance is a null entity. (The EntityManager will return a NullEntity instead of a null value when a requested entity is not found.) Returns whether the current instance is a null entity or a pending entity. Returns whether the current instance is a pending entity. (The EntityManager will return a PendingEntity instead of a null value when a requested entity is being queried asynchronously and has not yet been returned.) Rejects (rolls back) all changes to this Entity since it was retrieved or had Entity.AcceptChanges called on it.

GetDataProperty() GetHashCode() GetRelatedEntities()

GetRelatedEntities<>() GetRelatedEntity() GetRelatedEntity<>() GetType() HasChanges() HasErrors IsNullEntity IsNullOrPendingEntity IsPendingEntity

RejectChanges()

RemoveFromManager()

Removes the entity from the EntityManager cache.

99 | P a g e

IdeaBlade DevForce
SavingErrorMessage SetAdded()

Class Libraries
Gets the description of any error that occured during the most recent save of this entity. Forces this entity into the EntityState of Added. You will usually have no reason to call this method from application code. The EntityState is automatically set to Added by the framework when a new entity is added to an EntityManager. Forces this entity into the EntityState of Modified. You will usually have no reason to call this method from application code. The EntityState is automatically set to Modified by the framework when any EntityProperty of the entity is changed. A string representation of this object. Gets the IdeaBlade.Validation.VerifierEngine shared by all entities within the same EntityManager.

SetModified()

ToString() VerifierEngine

Finding Help on DevForce
Refer to the Reference Help (available from the IdeaBlade DevForce / Documentation menu option) for information on DevForce classes. For the detail help on individual types and their members, you can launch the Reference Help from the Start Menu ► All Programs ► IdeaBlade DevForce ►Reference Help. There are some handy ways to get type and member level help within Visual Studio itself. These techniques are great for exploring .NET and your own code as well as DevForce. IntelliSense Object Browser Class View Class Diagram These are discussed in more detail below.

XML Documentation
These techniques are only as good as the XML documentation embedded in the code.

IntelliSense
You probably know that if you hover the mouse over a class or method for a moment, you‟ll see a tool tip that gives some brief syntactical information.

IntelliSense can do much more. A full discussion is out of scope but we thought it worth mentioning a few of the tricks we use all of the time.

100 | P a g e

IdeaBlade DevForce Learn the key accelerators.
You‟ll find many of the IntelliSense accelerator keys listed in the Visual Studio Edit Menu.

Class Libraries

We find List Members and Parameter Info to be the most informative short-cuts.12    Press the accelerator key combination for List Members within an identifier and IntelliSense displays its description straight from the <summary/> tag of its XML documentation. This works for a member, too. Enter the dot („.‟) to the right of an identifier and again you get an open list. IntelliSense scrolls to the most recently used option if there is one. Clicking on the option reveals its description. Press the accelerator key combination for Parameter Info while your cursor is in the parameter list and you‟ll see the method overloads and a description of the nearest parameter if the developer provided the XML documentation

Use the up and down arrow to scroll among the overloads.

12

Word completion is a big help too although that‟s more about saving keystrokes than discovery. Type just a few letters of the desired identifier, then signal work completion; VS fills in your choice.

101 | P a g e

IdeaBlade DevForce

Class Libraries

The Object Browser
From the menu select View ► Object Browser (or press the accelerator key combination shown adjacent to that menu option). Visual Studio adds an “Object Browser” tab showing all of the assemblies referenced by any project in the solution. There are three panes: a tree view of the assembly, list of member of the selected type, and the XML documentation.

You can expand any node in the tree view to see its inheritance chain and examine any type within that chain:

Search
Best of all, you can search for a word (or part of a word) among your referenced assemblies and the .NET framework assemblies.

102 | P a g e

IdeaBlade DevForce

Class Libraries

In this example, we browse just among the solution assemblies and search for the word “manager”. After we click the green search button, the result might look like this:

We can browse around in this filtered set as before. We click the “Clear Search” icon when we‟re done.

Class View
Class View is just like the Object View but narrows the scope to just the assemblies we can edit in our Solution. You can open the Class View from the menu (View ► Class View).

Class Diagram
With both the Object Browser and Class View we can examine the assemblies but we cannot change their XML documentation. We can only change the XML documentation of the code we‟re editing. As with previous versions of Visual Studio we can enter the XML comments within the Code View. Visual Studio 2005 introduced the Class Diagram which promises to be the means for programming visually. You can open a class diagram in Visual Studio 2008 using the View Class Diagram button on the Class View toolbar. The Class Diagram displays a lovely design service onto which we can drag our existing classes and mark associations among them like so:

103 | P a g e

IdeaBlade DevForce

Class Libraries

We can add classes and add members and the Class Diagram will stub in the code which we later flesh out by switching back to Code View. There are no round-tripping problems because the Class Diagram is just another view on the exact same code; there is no intermediate format that must be translated as we shift from one perspective to the other. This will be a wonderful documentation tool as soon as Microsoft smoothes some of the rough edges. You soon discover that you‟re wrestling with positioning the blocks and the lines that connect them. Just as you have it about right, you touch something and all of the graphics scramble to odd and inconvenient positions. We are not enamored of the code generation either. It generates a nice stub but we can do better with code snippets and, really, how hard is it to write a stub method or property? If you like it, by all means use it.

104 | P a g e

IdeaBlade DevForce

Business Object Mapping

Business Object Mapping

Business Object Mapping
Introduction
Overview of the ADO.NET Entity Model Working with the IdeaBlade DevForce Object Mapper

Object Mapper Walk-Through
Exiting The Object Mapper The Object Mapper Menus Injected Base Types The Name Pluralizer: Fixing the Pluralization in Type Names Mapping a Web Service

Notes on the Generated Code Multiple Datasources
DataSourceKeys

Appendix: Many-to-Many Associations in the Entity Framework

The business object model consists primarily of “business objects”, the programmatic constructs that represent entities in the domain of the application. In most cases, business objects represent entities in the “real world” such as customers, employees, orders, and products. Business Object Model development proceeds in the following steps:   The developer uses the ADO.NET Entity Data Model Designer, or XML hand-coding, or a likely combination of the two, to develop one or more Entity Data Models that describe conceptual entities and map them to a back-end storage schema. The developer uses the DevForce Object Mapper to create a “Domain Model” that combines one or more Entity Data Models into a single logical model. The resulting Domain Model encompasses entities from all source Entity Models, uniting them in a single structure that permits relationships to be defined amongst them and (at runtime) updates to be handled transactionally across them. The developer uses the Object Mapper to enhance and adjust the Entity Models in a variety of ways. The changes made to the source Entity Model(s) by the Object Mapper do not conflict with the demands of the Entity Data Model Designer, so the developer can work with confidence on her business model using either or both tools, throughout the development life cycle. The Object Mapper then generates DevForce Entity class files from the Domain Model, just as the Entity Data Model Designer generates class files from the Entity Model(s). These class files define the specification for the runtime object model.

This chapter covers addresses the building of the Business Object Model. Another chapter, “Object Persistence”, describes the use of that model in persistence operations such as queries, creates, updates, deletes and saves.

Introduction
The ADO.NET Entity Framework provides a mapping scheme for declaring how its persistence mechanisms should translate between data in a data source and data in an in-memory business object state. The Entity Framework, however, provides only for operation in client-server mode. Its object model includes detailed information about the

105 | P a g e

IdeaBlade DevForce

Business Object Mapping

schema of the back-end data sources, and the Entity Framework depends upon locally available connection information to carry out its persistence operations with those data sources. The DevForce Composite Object Model, by contrast, is entirely free of knowledge or information about back-end data sources, and so can be deployed to client machines without compromising the security of the persistent data stores (which may be relational databases or web services). Furthermore, the DevForce persistence framework that uses the Domain Model provides and manages a local cache in which data can be stored during application sessions, permitting extensive and complex data maintenance work independent of the back-end data stores. In short, DevForce extends the capabilities of the Entity Framework from client-server to n-tier; and in the process, provides all the benefits of n-tier deployment. These include:       client machines that are more secure; vastly better performance over low and moderate speed connections (such as typical internet connections); wider application reach (because of the reduced connection performance requirements); fuller use of client-side computing power and corresponding reduced server-side hardware demands; server statelessness and the concomitant ease of server scaling; and disconnected operation.

Because the DevForce Composite Object Model integrates multiple Entity Models, you business model can – unlike in the Entity Framework – span multiple back-end databases. The DevForce model, again unlike the EF Model, also supports business objects that are sourced from web services. Objects derived from all of these sources are integrated into a single model, in which they are treated as equal members, have relationships, and participate in single, integrated transactions.

Overview of the ADO.NET Entity Model
In the ADO.NET Entity Model you declare, for example, that anEmployee.FirstName, maps to the [FirstName] column of [Employee] table records in a relational database. The Entity Framework‟s persistence mechanism can take it from there. When it retrieves employee data from the data source, it will know how to find the first name value and copy it into the EF version of the Employee business object. When it saves the Employee business object, it will know how to extract the first name value and put it where it belongs in the data source. The DevForce persistence framework maps DevForce Entities to the corresponding Entity Framework objects. But that is all happily behind the scenes for you. All you need to know is that DevForce uses the services of the Entity Framework to perform the direct communications with back-end databases, and that, in order to do so, it knows how to map its own entities to the EF ones. In your development world, you can happily ignore the latter; you will code to, and in your application use, the DevForce entities exclusively.

Working with the IdeaBlade DevForce Object Mapper
The IdeaBlade DevForce Object Mapping (OM) tool, AKA “Object Mapper”, is a graphical designer for creating and maintaining domain model from the starting point of one or more ADO.NET Entity Data Models. The Object Mapper plugs directly into Visual Studio during DevForce installation. Note that we do not address the creation or modification of the Entity Model using the Visual Studio Entity Data Model designer or XML hand-coding. Those are topics for which Microsoft provides abundant documentation. If you can put it in the Entity Model and make it work with the Entity Framework, we support it.

Object Mapper Walk-Through
This topic is arranged as a structured walk through the tool during which we create and elaborate a domain model containing a collection of related business object entities based on tables in the NorthwindIB database.

106 | P a g e

IdeaBlade DevForce
1.

Business Object Mapping

Begin with an existing Visual Studio solution containing at least one project with an ADO.NET Entity Data Model (EDM) in it. The solution shown below has an EDM based on the NorthwindIB database that ships with DevForce. This is a modified version of the NorthwindEF database distributed by Microsoft. 13

2.

Launch the Object Mapper from the Visual Studio menu:

The Object Mapper launches with no Entity Data Model loaded in this first-time use:

13

A discussion of the differences between NorthwindIB and NorthwindEF, and the reasons for them, is included as an Appendix to this chapter.

107 | P a g e

IdeaBlade DevForce

Business Object Mapping

3.

Click on the Domain Model name – “(New Model)” -- in the tree control to select it. Now inspect the properties of the Domain Model in the Detail pane.

108 | P a g e

IdeaBlade DevForce

Business Object Mapping

Domain Model Project. This will display the path to the folder for the Domain Model project and files. Use the <New project…> button to create a new Domain Model project; otherwise select an existing project from the dropdown list. Create Silverlight Domain Model Project checkbox. Check this box if you wish to create a Silverlight project. (See the chapter, “DevForce Silverlight Apps” for more detail.) This option is only available in the DevForce Silverlight and DevForce Universal products; not in DevForce WinClient. Namespace. Code for the Domain Model will be generated into the namespace shown in the Settings area. By default, the namespace will be “DomainModel”. You can change this if you prefer a different name for the namespace. Entity Manager Name. By default, the container for the Domain Model‟s collections of objects will be “DomainModelEntityManager”. You can change this as well. You will be able to reference its collections in LINQ and elsewhere as follows: C# VB
DomainModelEntityManager _em1 = DomainModel.DomainModelEntityManager.DefaultManager;

Dim _em1 As DomainModelEntityManager = DomainModel.DomainModelEntityManager.DefaultManager

Generate Code after save. This option determines whether the the Object Mapper will generate code when you save your Domain Model. You can selectively turn on or off the generation of two types of code:  Create Developer Partial Class Files If this option is checked, the Object Mapper will generate a developer partial class in a stand-alone file for each entity in the model for which it does not find such a file already in existence. It will never overwrite an existing partial class developer file. Update model project‟s app.config Determines if the app.config file (written in XML) will be generated or updated.

.NET Language. This option determines the .NET language in which the class files will be generated. You can choose C# or VB. Copy Entity Model Artifacts. This option determines whether the Object Mapper will generate statements in the DomainModel project‟s post-build event handler to move necessary DLLs and Entity Data Model components to the executables directory for the solution. Without these statements you will have to move the files yourself whenever you change and rebuild the model. File Name. This setting displays the file name (path) to the folder where the Domain Model (.ibedmx) file is (or will be) saved. 4. Add an Entity Model to your Domain Model.

109 | P a g e

IdeaBlade DevForce
Select Model / Add Entity Model to add an existing Entity Model to your Domain Model.

Business Object Mapping

You will be presented with an Open File dialog. If there is an Entity Data Model already in your Visual Studio solution, the Object Mapper will find it. Otherwise, or if your solution has multiple Entity Data Models and you wish to select a different one than the Object Mapper selected, browse to the desired Entity Model file. Click <Open> when the desired EDM is selected in the dialog to add it to your domain model.

110 | P a g e

IdeaBlade DevForce
5. Inspect the Properties of the Entity Model:

Business Object Mapping

File Name. This setting displays the file name (path) to the Entity Model (.edmx) file. This is the location to which you browsed when adding the Entity Model to the Domain Model. Data Source Key Name. The data source key name is the symbolic name that will be used by DevForce for this Entity Model, and by extension, the database to which it is bound. Namespace. Code for the Entity Model will be generated into the namespace shown. The namespace shown is the one currently encoded into the Entity Model (.edmx) file. If you have not previously changed it in the Object Mapper, it is the namespace you selected when creating the Entity Model, or which you later changed in the Entity Model. You can change this name if you prefer. Container Name. The container for the entity model‟s collections of objects is a System.Data.Object.ObjectContext whose name is encoded into the Entity Model (.edmx) file. Note that you are unlikely to use this container directly in your DevForce app, since you will be working through the Domain Model‟s DomainModelEntityManager container and its collections. C# VB
DomainModelEntityManager _em1 = DomainModel.Manager.DefaultManager;

111 | P a g e

IdeaBlade DevForce

Business Object Mapping

Injected Base Types. This button launches a dialog to help you define base classes to sit in the inheritance hierarchy of your business type classes. This will be discussed below. Name Pluralizer. This button launches a tool to help you save a great deal of time making the pluralization of your type names orderly and sensible. This will be discussed below. Verification Interceptors. These options give you control over which verification-related interceptors are coded into your generated property definitions. You can choose among the options shown below:

Verification interceptors are discussed in Chapter 7, “Validation Through Verification,” of this Developer Guide. 6. Click on the ServerModelNorthwindIbContext node in the navigational tree control. When you‟re positioned in the tree on a container node such as this one, you have a comprehensive, quick reference view of the entity types and their properties:

112 | P a g e

IdeaBlade DevForce
For each Entity type you see:     7.

Business Object Mapping

a read-only IsModified property indicating whether the type definition has been modified in the current Object Mapper session the name of the Entity Set that will provide instances of the type whether the type is abstract the type‟s Base type, if it inherits from another type.

Click in the tree control on one of the individual types. You‟ll see only the properties and associations for that type:

8.

Select the Navigation Properties tab in the detail pane. This shows you the properties defined in the ServerModelNorthwindIB Entity Model for navigating from the Order object to its related entities. The Order has a Customer who placed it, an Employee who wrote it, a collection of line items (which are OrderDetail objects), and, potentially, a related InternationalOrder. These properties were generated by the EDM Designer in response to its discovery of relationships in the targeted database schema.

113 | P a g e

IdeaBlade DevForce
9.

Business Object Mapping

Select the Associations tab. You will see the background information about the associations defined in the model (which led to the generation of the navigation properties you just examined).

A foreign key constraint, FK_Order_Customer, found in the database, relates the Order and Customer tables in a many-to-1 relationship. (One customer can place many Orders.) Another, FK_Order_Employee, relates the Order and Employee tables in a many-to-1 relationship. (One Employee can write many Orders.) A third, FK_OrderDetail_Order, relates the Order and OrderDetail tables in a 1-to-many. (One Order can have many details – line items).

114 | P a g e

IdeaBlade DevForce

Business Object Mapping

10. Return to the Simple Properties tab and examine the detail on the properties of the Order Entity:

On this grid, you can:  Rename a property;  Add it to, or remove it from, the type‟s primary key;  Observe its datatype ;  Make it nullable or non-nullable;  Designate it as a column to be considered when checking for concurrency conflicts, and ask DevForce to automatically update it for you as needed, according to a variety of strategies;  Change Getter and Setter Access levels

Continuing to the right on the Simple Properties grid, you can:       Set the property‟s Getter and Setter access types; Turn generation of Validation Attributes on or off; Set the Verification Setter Mode; Observe the property‟s maximum allowed length, for data types for which that property is defined; Observe whether the property‟s length is fixed; Observe the property‟s Precision and Scale.

115 | P a g e

IdeaBlade DevForce

Business Object Mapping

Detail on the Property Settings
Many of the property settings are self-explanatory, but a few need further explanation:

Concurrency Strategy
Six strategies are defined, as shown here:

These are discussed in detail in Chapter 6, “Object Persistence”, in the “Basic Mechanics of Concurrency Detection” section.

Getter Access
The Getter Access setting determines with what access type the property getter will be generated. The options are: Public, Protected, Internal, Private, and Protected Internal.

Setter Access
Setter Access options are similar to those for the Getters, with the additional option of None (which causes the setter not to be generated at all).

Generate Validation Attributes
This option controls whether Validation attributes are included in generated property definitions.

Verification Setter Mode

116 | P a g e

IdeaBlade DevForce

Business Object Mapping

The Verification Setter Mode controls what verification-specific interception points will occur for the generated properties. Verification interceptors are in Chapter 7, “Validation Through Verification”.

Exiting The Object Mapper
If you attempt exit the Object Mapper with unsaved changes, you‟ll see the following dialog:

Clicking <Yes> will save the domain model with a default name and location, those being:   Name. The domain model will be given the same name as the (first) Entity Data Model added; and Location: The domain model will be saved into the same project as the Entity Data Model most recently added.

If you wish to exert explicit control over where the domain model is saved and what it is named, click <Cancel> when you see the above dialog, then select the domain model node in the navigation pane of the Object Mapper dialog (which will have the title “(New Model)” if the domain model has not previously been saved):

Using the ComboBox labeled Domain Model Project, you can specify into which of your solution‟s projects the domain model will be generated, or you can click the <New project…> button to create a new project just for the domain model.

117 | P a g e

IdeaBlade DevForce

Business Object Mapping

By default, a new project will be named “DomainModel”. Note that the domain model itself, encoded in a file with extension .ibedmx, will always be named the same as the folder into which it is generated. When you create the new project, a solution folder of the same name will be created, and both the domain model project and all existing Entity Data Model projects that are part of the domain model will be moved into that solution folder. The domain model itself will be generated and saved only when you so order.

The Object Mapper Menus
Menu Options

The File menu contains options related to the Domain Model file as a whole.

The Model menu contains options related to operations on the Entity Model components of the domain model. Many of these options are also available on right-click shortcut menus associated with nodes of the navigation tree.

118 | P a g e

IdeaBlade DevForce

Business Object Mapping

The View menu contains a toggle option to suppress or display the navigation window (tree control).

The Help menu contains an About option from which you can learn the version number and license type associated with the copy of DevForce you have installed.

Injected Base Types
In the Detail pane for an Entity Model we previously noted the presence of a button to launch something called “Injected Base Types...”.

Clicking this button launches the following dialog:

119 | P a g e

IdeaBlade DevForce

Business Object Mapping

This dialog permits you to introduce base types between levels in your inheritance hierarchy. The most common operation will be to create a base entity that inherits from IdeaBlade.EntityModel.Entity, and to make it the base type for all business object types in the model. This you would do as follows: 1. 2. Click the <Add> button. In the “Add Injected Base Type” window, type the desired name of your base entity, and accept the inheritance default of “Entity”.

3.

Click OK. You‟ll see the following dialog:

4.

Now select the row created for BaseEntity and click <Set Default>.

120 | P a g e

IdeaBlade DevForce

Business Object Mapping

BaseEntity becomes the default base class for all your business object types.

You can all other base types and assign any subset of your business object types to inherit from them. Note, however, that you can insert at most one base type between each pair of concrete types. Thus, you can have Order inherit from BaseEntity and International Order inherit from NonDomesticOrder; but you can‟t use the Injected Types dialog to make Order inherit from OrderBase and OrderBase inherit from BaseEntity. If you need to do that, do the following: 1. 2. 3. Define base types BaseEntity and OrderBase. Make BaseEntity the default type. After close the Injected Types dialog, set Order to inherit from OrderBase.

4. 5.

Save and generate the model. Manually change the code for the OrderBase class to make OrderBase inherit from BaseEntity. (Note that you will have to re-do this last step if you regenerate OrderBase.)

We stopped short of providing facilities to define such multi-level inheritance stacks in the Injected Types dialog in order to keep the UI simple.

The Name Pluralizer: Fixing the Pluralization in Type Names
In the Detail pane for an Entity Model we previously noted the presence of a button to launch something called the “Name Pluralizer”.

121 | P a g e

IdeaBlade DevForce

Business Object Mapping

The Name Pluralizer can also be launched from the main menu via Edit / Name Pluralizer, or from the context menu associated with the Entity Model node in the navigational tree control:

Launching the Name Pluralizer by any of these mechanisms presents you with the following modal dialog:

We‟ll get to the mechanics of this tool in a moment, but let‟s get a little background first. The Entity Model Designer, by default, always names the Entity Sets and Entity the same as the table on which they are based. In the NorthwindIB database, tables were named in the singular (“Customer” rather than “Customers”,

122 | P a g e

IdeaBlade DevForce

Business Object Mapping

etc.), so both the Entity Types and the Entity Sets are named in the singular as well. We‟re going to want the Entity Set names to be plural, as we find it more natural to refer to the “Customers” collection, which contains individual “Customer” entities. The EDM Designer also makes no attempt to address the number (singular or plural) of the navigation properties. It simply names them the same thing as the corresponding Entity. You may want to do better than that. Most developers want the navigation properties that return a collection to have plural names (“OrderDetails” instead of “OrderDetail”), and the navigation properties that return a single object to have singular names (“Customer”, “Employee”, “InternationalOrder”). Now let‟s take a second look at the Name Pluralizer dialog, first shown above. The Name Pluralizer will fix our pluralization problems with both Entity Set names and Navigation Property names at one stroke. You can change them in either direction, but for most developers, the default settings will be perfect. Regardless of what their names are now, Entity Sets and Navigation Properties that return a collection will end up with plural names; Entity Types and Navigation Properties that return a single object with singular ones. (The <Reset Defaults> button will always re-establish the settings you see here.)

You just click <OK> or <Apply> to perform the work. (The <Ok> and <Apply> buttons have the standard Windows behaviors: both perform the indicated operations, but <Ok> will also close the dialog, whereas <Apply> will leave it open.)

123 | P a g e

IdeaBlade DevForce

Business Object Mapping

Note the pluralization of the Entity Set and Navigation Property names after applying the Pluralizer. This will save you a lot of time, particularly if your model is large!

Mapping a Web Service
The process of adding a web service to your domain model will result in a service reference being added to the project that is selected at the time you add the web service. Since such a service reference should not be included in a project destined for client-side deployment, you should, before launching the Object Mapper to add a web service to your domain model, select such a project. If you don‟t already have a suitable project targetted for server-side-only deployment, create one. To add web service entities to your domain model, begin by selecting File / Add Web Service from the Object Mapper menu:

124 | P a g e

IdeaBlade DevForce

Business Object Mapping

In the resultant dialog, enter the URL of the desired web service:

125 | P a g e

IdeaBlade DevForce

Business Object Mapping

The web service returns information about what Services and Operations it provides:

A .NET class will be generated to facilitate your access of this web service. Change the namespace that will be used for the code in this class if desired:

This will result in an additional EntityModel node in the navigation pane, and a new container (ObjectContext) with classes corresponding to the output of the web service. Such classes will become types in your DomainModel, peers of those generated from database tables.

126 | P a g e

IdeaBlade DevForce

Business Object Mapping

Notes on the Generated Code
The DevForce Object Mapper generates a great deal of code into its designer class file. Let‟s have a look at the code for generated properties.

Data Properties
First, let‟s consider the generated code for a simple property. We‟ll look at the CompanyName property of the Customer object in the NorthwindIB database. Here is the complete generated code for this property:

C#

#region CompanyName /// <summary>Gets or sets the CompanyName. </summary> [Bindable(true, BindingDirection.TwoWay)] [Editable(true)] [Display(Name="Company Name", AutoGenerateField=true)] [IbVal.ValidateProperty] [IbVal.StringLengthVerifier(MaxValue=40, IsRequired=true)] [IbCore.MaxTextLength(40)] [MsSer.DataMember] public String CompanyName { get { return CompanyNameEntityProperty.GetValue(this); } [System.Diagnostics.DebuggerNonUserCode] set { CompanyNameEntityProperty.SetValue(this, value); } } #endregion CompanyName property

VB

#Region "CompanyName property" '''<summary>Gets or sets the CompanyName. </summary>

127 | P a g e

IdeaBlade DevForce
<Bindable(true, BindingDirection.TwoWay)> _ <Editable(true)> _ <Display(Name:="Company Name", AutoGenerateField:=true)> _ <IbVal.ValidateProperty> _ <IbVal.StringLengthVerifier(MaxValue:=40, IsRequired:=true)> _ <IbCore.MaxTextLength(40)> _ <MsSer.DataMember> _ Public Property CompanyName() As String Get Return CompanyNameEntityProperty.GetValue(Me) End Get <System.Diagnostics.DebuggerNonUserCode> _ Set CompanyNameEntityProperty.SetValue(Me, value) End Set End Property #End Region

Business Object Mapping

See the chapter “Property Interceptors” for a detailed discusion of how you can write code to intercept Gets and Sets to change the delivered or received value, perform security checks, or perform any desired related operation. There is even more information in the CompanyNameEntityProperty that we have so far described. As it so happens, EntityProperty has two subclasses, DataEntityProperty and NavigationEntityProperty, which contain additional information. Since CompanyName isn‟t a navigation property, but rather a simple data property, CompanyNameEntityProperty is generated into the designer code as a DataEntityProperty. That has the following members:

As you can see, the information you have available about the CompanyName property to your interceptor methods is quite rich indeed. In addition to the things we‟ve seen before, you have the property‟s default value, and you can tell if its value is autoincremented, if it is a complex type, if it is designated as a property to be checked for the determination of data concurrency, and if it belongs to its containing object‟s key.

Navigation Properties
Now let‟s look at the definition for a navigation property. These, you may recall, are generated when relations are defined between types. The Customer type, for example, is involved in a one-to-many relation with the Order type: a given Customer can place many Orders. So the DevForce Object Mapper generated an Orders property into the Customer class: C#
#region Orders property /// <summary>Gets the Orders. </summary> [Bindable(false)]

128 | P a g e

IdeaBlade DevForce

Business Object Mapping

[Display(AutoGenerateField=false)] [MsSer.DataMember] [IbEm.RelationProperty("FK_Order_Customer", IbEm.QueryDirection.ToRole2)] public IbEm.RelatedEntityList<Order> Orders { get { return OrdersEntityProperty.GetValue(this); } } #endregion Orders property

VB

#Region "Orders property" '''<summary>Gets the Orders. </summary> <Bindable(false)> _ <Display(AutoGenerateField:=false)> _ <MsSer.DataMember> _ <IbEm.RelationProperty("FK_Order_Customer", IbEm.QueryDirection.ToRole2)> _ Public ReadOnly Property Orders() As IbEm.RelatedEntityList(Of Order) Get Return OrdersEntityProperty.GetValue(Me) End Get End Property #End Region

The code for the Orders property has many similarities to the CompanyName property we previously examined, but some important differences as well. Whereas CompanyName returned a simple string type, Orders returns a RelatedEntityList<Order>. It has an attribute that flags it as a RelationProperty (the term is synonymous with “navigation property”), and which specifies the relation type (FK_Order_Customer) that connects Order to Customer and indicates that Order is the child in that relation. Note that a ChildrenReference<Order> property is also defined in the generated code. This property, named Orders_Reference, allows you, among other things, to examine the navigation property before the fact to determine if it returns a scalar value or a list. The name “Orders” sure looks like something that returns a collection, but that just a happy consequence of the way the modeller named it, and not an easy or reliable way to make the determination! Since Orders is a navigation property, its corresponding EntityProperty is generated as a NavigationEntityProperty. The following information is available on such a property, above and beyond the information we say previously that belongs to all EntityProperties:

Note that you can tell which side of a relation the type returned by the property is on (QueryDirection), and which relation is involved (RelationName).

Options for Getting and Setting Property Values
Note that you have two distinct ways of requesting or setting the value of a given property, shown below. All of the syntaxes shown result in operations that are routed through the property‟s getter or setter logic14:

14

See the “Property Interceptors” chapter to learn how to bypass the getter and setter logic for those cases where you need to.

129 | P a g e

IdeaBlade DevForce
Get
  anEmployee.LastName Employee.LastNameEntityProperty.GetValue(anEmployee)

Business Object Mapping

Set
  anEmployee.LastName = “Jones” Employee.LastNameEntityProperty.SetValue(anEmployee, “Jones”)

Another overload of GetValue() permits you to access the several different versions of value of the property:

Current is the value you retrieve using the two Get syntaxes first shown. Original is the value as last retrieved from the data source, before any local changes (if any) were made. Just before a change is made to the Current value of an entity, it has a Proposed value which may or may not be allowed depending upon setter logic. The Default value depends upon the entity‟s EntityState. For example, the Default value is the Current value for an entity in any of the following EntityStates: Added, Modified, or Deleted. For an entity in the states EntityState.Detached or IEditableObject.Edit, the Default value is the Proposed value.

EntityManager.GetEntityGroup
In the cache, entities of a single type are stored in a container called an EntityGroup. You probably won‟t find much direct need of this container, but it does raise a few low-level events that can be useful in very specific situations, those being:  EntityPropertyChanging  EntityPropertyChanged  EntityChanging  EntityChanged The first two fire when a property is changed, and are specific to that property; the last two fire when anything about an entity changes (including a property). If you‟re alert, you may note that those occurrences seem pretty well covered by the corresponding interceptors in the property setters that we just finished discussing: Event on EntityGroup EntityPropertyChanging EntityPropertyChanged EntityChanging EntityChanged Corresponding Interceptor methods generated into the Entity class Before{PropertyName}Set After{PropertyName}Set BeforeSet AfterSet

And you‟d be right: for most things you need to be in response to a change in an entity, the interceptor methods are the vehicle of choice. But the EntityGroup events offer one advantage over the latter: because they are implemented

130 | P a g e

IdeaBlade DevForce

Business Object Mapping

as event handlers, they can be added and removed dynamically at runtime. If you need to do something conditioned upon runtime circumstances, they‟ll be just the ticket. You can get an instance of the EntityGroup for a type (we‟ll use Customer again) from an EntityManager as follows: C# EntityGroup customerGroup = anEntityManager.GetEntityGroup(typeof(Customer)) ; VB You can also get the EntityGroup associated with a particular entity: C# aCustomer.EntityAspect.EntityGroup

VB

Multiple Datasources
Some business object models unite data from multiple data sources. Order information, for example, might reside in both an application-specific database and in an accounting database. We might need to read from both. We might need to post to both. We can accommodate these requirements within a single business object model 15. We began the mapping session illustration by supplying a database connection string. We then examined the database objects, picked a few, mapped them, and generated their classes.

15

The data sources can be a mix of databases and web services. We illustrate this discussion with database data sources.

131 | P a g e

IdeaBlade DevForce

Business Object Mapping

DataSourceKeys
We neglected to mention that the concrete business objects we declared were attached permanently to a particular DataSourceKey. We associated that key with a database accessible via the connection string. Let’s revisit the moment after setting that connection string.

Notice the DataSourceKey property associated with the Entity Data Model. In the screen shot, above, the property has the name “Default”, which is the default value given to this property by the Object Mapper. But this can be renamed as desired by the developer. In the Entity Data Model, the DataSourceKey name is stored as a DataSourceKey attribute of the Schema element within the ConceptualModel. This tells you that there is a one-to-one mapping between a DataSourceKey and a particular Entity Data Model (EDM). Since a single DevForce DomainModel can encompass multiple EDMs, there is then a one-to-many relationship between a DomainModel and its DataSourceKeys. In the app.config file, the name of a DataSourceKey is stored in a name attribute of an edmKey element. In that file, the Object Mapper generates one edmKey for each Entity Data Model included in the DomainModel. Each edmKey, besides having a name, also has a connection attribute whose value includes both the location of the entity model artifacts (.csdl, .ssdl, and .msl files) and the connection string for the database. When the Object Mapper generates its version of the app.config initially, this connection string is brought over intact from the app.config file associated with the underlying Entity Data Model.

132 | P a g e

IdeaBlade DevForce

Business Object Mapping

You might assume, at first blush, that there is necessarily exactly one edmKey for each DataSourceKey. But that isn‟t the case. While there can be, at runtime -- at a given moment -- only one physical data source associated with a given Entity Data Model, DevForce provides you with the flexibility to assign different physical datasources to a given EDM at different times. For example, you might have Development, Test, and Production versions of a given database. All have the same schema, but contain different data. You can decide for a given runtime session of the application which database should be used to supply data to Entity Data Model XYZ. You can even switch the datasource out dynamically while running! To use this capability, you add additional edmKey elements manually to the app. config, so that it will end ups with multiple edmKeys for at least some of the Entity Data Models that comprise your Domain Model. Because of this capability, the actual formal relationship between Entity Data Models and edmKeys is one-to-many. On the other hand, there should be exactly one edmKey for each physical database that can contribute data to a given DomainModel at runtime. Thus the relationship of edmKeys to physical databases is 1-to1. The relationship between these various elements is summarized in the following sidebar and diagram:

Models, Keys, and Data Sources
The diagram at right summarizes the relationships between models, keys, and data sources. One DomainModel may be associated with many Entity Data Models (EDMs). A given EDM is associated with a single DataSourceKey. A given EDM / DataSourceKey may be associated with many Entity Types. A given EDM / DataSourceKey may be associated with many edmKeys (in the app.config file). Each edmKey represents a single physical data source.

The DataSourceKey represents a schema: for a given EDM, the DataSourceKey identifies the Datasource schema to which the EDM‟s conceptual model is mapped. Every business object has a DataSourceKeyName property, defined in the (Entity) class that was generated by the DevForce Object Mapper to contain the object‟s blueprint.Any data source used at runtime to supply data for that business object must have the same schema as the data source to which that business object type is mapped in the Entity Data Model.

Appendix: Many-to-Many Associations in the Entity Framework
In this appendix we examine alternative ways of modeling many-to-many associations in Entity Data Models. Specifically, we‟ll look at three different models, all based on the NorthwindIB database 16, that link Employees and Territories in a many-to-many relationship. The central differences between the three relate to how the linking entity in the many-to-many association is modeled. Accordingly, we describe them with reference to that entity:  Exposed Linking Entity With Payload
16

NorthwindIB is our IdeaBlade version of the NorthwindEF sample database distributed by Microsoft. We‟ve made a variety of changes to permit us to illustrate different capabilities of DevForce and the Entity Framework, but the two database remain substantially similar.

133 | P a g e

IdeaBlade DevForce
  Non-Exposed Linking Entity With No Payload Exposed Linking Entity With No Payload

Business Object Mapping

We‟ll see how each of the three situations is modeled in the Entity Framework, and discuss some pros and cons.

Introduction
“Payload” is the term used by the Entity Framework designers to describe columns in a linking entity other than the foreign keys to the two external items that the linking entity connects. In the database diagram below, EmployeeTerritory is the many-to-many linking table between the Employee and Territory tables. It has two foreign keys, EmployeeID and Territory ID, which link it to those tables (in many-to-1 relationships). But it has also has ID and RowVersion columns which, whatever their business function, constitute payload in the linking entity.
Figure 2. Linking Table with Payload

The following database diagram, on the other hand, depicts a linking table, EmployeeTerritoryNoPayload, which has only the two foreign key columns.
Figure 3. Linking Table with no Payload

134 | P a g e

IdeaBlade DevForce

Business Object Mapping

As it so happens, the Entity Data Model (EDM) Wizard in Visual Studio, which is launched when you add an ADO.NET Entity Data Model item to a project, generates very different conceptual models from these two sets of table schemas. Let‟s have a look.

Exposed Linking Entity With Payload
If you run the EDM wizard against the set of tables in Figure 1 and do a bit of renaming on the Navigation Properties to make their function clearer, you get an EDM that looks like the following:

135 | P a g e

IdeaBlade DevForce

Business Object Mapping

Notice the many-to-1 associations17 between EmployeeTerritory and Employee, and between EmployeeTerritory and Territory. Taken together, they define a many-to-many association between Employee and Territory, but that association is not explicit. Notice also that the EDM wizard created a navigation property in the Employee entity to return the employee‟s collection of associated EmployeeTerritory objects. It named that property EmployeeTerritory; we renamed it to “EmployeeTerritories” to make clearer that it returns a collection. The wizard created a corresponding navigation property in the Territory entity, which we also renamed. If we want the collection of Territory entities that are associated with a particular Employee entity, we‟ll either have to iterate through its collection of EmployeeTerritory entities, grabbing the Territory associated with each and setting it aside in a list; or write a query to retrieve them. Whatever operation we choose to use to compile the list we can of course embed in a property or method of our Employee class to give us convenient access to the desired associates.

Non-Exposed Linking Entity With No Payload
If you run the EDM wizard against the set of tables in Figure 2 and again do a bit of renaming on the Navigation Properties, you get an EDM that looks like the following:

Good heavens: what happened to the linking entity, EmployeeTerritory?

17

If you‟re new to the Entity Framework and/or to object modeling, just be aware that what are called “relationships” between tables in a database are referred to as “associations” between the corresponding objects in an Entity Framework conceptual model. For practical purposes the terms “relationship”, “relation”, and “association” frequently get used more or less interchangeably in discussions of these object models, even though, technically, in UML terminology, an association is just one subtype of relationship.

In this paper I‟ve used the term “relationship” when speaking of tables in a database, and “association” when speaking of entities in an object model. I attempt no finer distinction than that.

136 | P a g e

IdeaBlade DevForce

Business Object Mapping

As it turns out, the EDM code generator decided that its function was entirely to associate the Employee and Territory entities, and that it therefore needed no explicit presence in the model. Instead, it chose simply to describe the many-to-many association between Employee and Territory. It also created a navigation property in the Employee entity to return the collection of associated Territories. Since it doesn‟t attempt meaningful pluralization, it named this property “Territory”; we renamed it to “Territories”. Corresponding, it created a navigation property in the Territory entity to return the related Employees. This, of course, it named “Employee”; we renamed it to “Employees”.

The Issues
While the invisible linking entity used for the payload-free linking table has its attractive aspects, it also means that you must necessarily live with two different ways of modeling many-to-manys in your Entity Data Model. You can‟t always live without a payload in your linking entity. Consider, example, an Order entity that links Sales Representatives to Customers in a many-to-many association. The Order is important in its own right, and is likely to carry a great deal of important informational baggage. It certainly must be exposed in your business model, whatever its function as a linking entity. The other issue is that “pure”, payload-free linking entities sometimes, over their lifetimes, need to grow a payload. You may find that you wish to record certain pieces of information about the association itself. When was said Territory assigned to said Employee? Who made the assignment? Why was the assignment made? The moment you add payload, you will have to change your model, because the linking entity can, by the rules of the ADO.NET Entity Data Model, no longer remain unexposed. Furthermore, the explicit many-to-many association it formerly defined is no longer supported. So you will have to rewire that many-to-many association as a pair of many-to-1s. This isn‟t terribly hard if you know what you‟re doing with the EDM, but we would certainly advise you to practice the job offline, in a small test model, before you try it on your real EDM. And make a backup of the latter before you start hacking away at it. It‟s all too easy to hose it up, at which point you will have no choice but to don your swamp boots and head bravely into a mosquito-infested swamp of XML.

Advantages of Standardizing on Linking Entities Having Payload
It often happens, as a business model evolves, that linking entities which began life as pure utililitarian connectors come to need additional properties to describe their state satisfactorily, and therefore to need a payload. If that‟s what the business model demands, one doesn‟t want to create a disincentive for adding such a payload when the day comes that it is discovered to be needed. Knowing that adding even a single extra column to a linking entity will force us to make a non-trivial18 change to our model might make us think twice about adding a column we really need. To head this off at the pass, we might make the design decision that our linking entities should always be exposed in the model, from day one. Then adding a column to a table, and a corresponding property to an entity in our conceptual model, will be a very simple operation. We might also prefer that every one of our tables and every one of our entities have an arbitrary, single-column primary key. When all of our entities, linking or otherwise, have the same kind of key, we work with all of them in exactly the same way, and we don‟t have to explain why a certain entity (which formerly didn‟t have, but now does have, plenty of payload columns) has a two-property primary key when the rest have single-property primary keys. “Is there a reason for that?” a new developer on our team asks. “Sure,” we answer. “It‟s historical.” No thanks! Applying the inclination for single-column primary keys to linking entities means you automatically also get linking entities that have payload and are explicitly exposed in your model. Thus, you gain two aspects of standardization for the price of one: (1) consistent primary key styles and (2) consistent modeling of linking entities and many-tomany associations across your model and down through time.

18

all in the eye of the beholder, of course

137 | P a g e

IdeaBlade DevForce Wait a Minute: What About the Navigation Properties?

Business Object Mapping

If you‟re very alert you may have noted that pursuing the above-discussed key and linking-entity-exposure preferences does leave us without one thing we get from the EDM‟s default treatment of payload-free linking entities: the many-to-many navigation properties. Isn‟t that a big disadvantage? Now we now have to write them ourselves? It‟s true, we do, so let‟s see how hard it is. In a DevForce application, we end up writing all our code against the Entity classes generated by DevForce rather than those generated by the Entity Framework (which, strictly speaking, are System.Data.Objects.DataClasses.EntityObjects rather than IdeaBlade.EntityModel.Entities). So if I‟m going to write a Territories property to return the Territories associated many-to-many with an Employee, I‟ll do it in the Employee partial class generated by DevForce in a standalone Employee.cs (or .vb) file. Here‟s what the property looks like: C#
public List<Territory> Territories { get { var query = this.EntityManager.EmployeeTerritories .Where(et => et.Employee.EmployeeID == this.EmployeeID) .Select(et => et.Territory) .Distinct(); return query.ToList(); } }

VB

The next 30 of these look pretty much like this one: I just substitute the appropriate entity types and key properties.

Exposed Linking Entity With No Payload
For our final exercise, let‟s suppose, hypothetically, that you‟re one of those wrong-headed people who disagree with me and happen to like, for linking entities, multi-column primary keys consisting of the two foreign keys. But let‟s also propose that you are persuaded by the utility of modeling a payload-free linking entity in a manner that allows it to acquire payload later with a minimum of upheaval in your model. Is there some way you can get the payload-free linking entity to show up, and with the same many-to-1 associations coming out of it that it will have to have later when it does have a payload? Yeah, okay, if the answer were “no”, I probably wouldn‟t have brought it up. Here‟s a picture of such a model:

138 | P a g e

IdeaBlade DevForce

Business Object Mapping

The unfortunate thing is that this is not an option provided by the EDM designer. You‟ll have to do a bit of twiddling. The dirty details would make this article too long and take it off course, but here‟s a prescription by which you can figure them out yourself. You can do everything below, except the final step, in the EDM designer: 1. Create a linking entity that has a payload (anything!) and build a model that uses it. The best model for this purpose is probably one that contains exactly three entities, like the ones we‟ve looked at in this article. 2. Remove the payload columns from the linking entity‟s backing database table. 3. Update the EDM using its “Update Model from Database” option. 4. Delete the properties corresponding to the payload column or columns that you removed. 5. Add the two foreign key properties to the conceptual in the EDM as explicit scalar properties. 6. Designate the two foreign key properties as primary keys. 7. Flesh out the table mappings for the two newly added properties, and for the two many-to-1 associations. 8. Examine the Before and After model to see what‟s different. If you‟re free to change the database on which the model is built, you might be able to use the above technique directly on your actual development model. Otherwise you‟ll have to spend enough energy on the last step above to be able to reproduce the result in your model. It‟s not rocket science, but it‟s not quite falling off a log, either.

139 | P a g e

IdeaBlade DevForce

Property Interceptors

Property Interceptors

Property Interceptors
Named vs. Unnamed Interceptor Actions Interceptor Chaining and Ordering

DevForce provides a mechanism to intercept and either modify or extend the behavior of any .NET property. This interception is intended to replace, and expand upon, the technique of marking properties as virtual and overriding them in a subclass. This facility is a lightweight form of what is termed “Aspect-Oriented Programming”. Interception can be accomplished either statically, via attributes on developer-defined interception methods, or dynamically, via runtime calls to the „current‟ instance of the PropertyInterceptorManager (described later). Attribute interception is substantially easier to write and should be the default choice in most cases.

Attribute Interception
DevForce supplies four attributes that are used to specify where and when property interception should occur. These attributes are
IdeaBlade.Core.BeforeGetAttribute IdeaBlade.Core.AfterGetAttribute IdeaBlade.Core.BeforeSetAttribute IdeaBlade.Core.AfterSetAttribute

Under most conditions these attributes will be placed on methods defined in the custom partial class associated with a particular DevForce entity. For example, the code immediately below represents a snippet from the autogenerated Employee class. (Generated code) C#
public partial class Employee : IdeaBlade.EntityModel.Entity { public String LastName { get { return LastNameEntityProperty.GetValue(this); } set { LastNameEntityProperty.SetValue(this, value); } } Partial Public Class Employee Inherits IdeaBlade.EntityModel.Entity Public Property LastName() As String Get Return LastNameEntityProperty.GetValue(Me) End Get Set(ByVal value As String) LastNameEntityProperty.SetValue(Me, value) End Set End Property

VB

Property interception of the get portion of this property would be accomplished by adding the following code fragment to a custom Employee partial class definition: (Developer code)

140 | P a g e

IdeaBlade DevForce
C#

Property Interceptors

[AfterGet(EntityPropertyNames.LastName)] public void UppercaseLastName(PropertyInterceptorArgs<Employee, String> args) { var lastName = args.Value; if ( !String.IsNullOrEmpty(lastName)) { args.Value = args.Value.ToUpper(); } } <AfterGet(EntityPropertyNames.LastName)> _ Public Sub UppercaseLastName(ByVal args As PropertyInterceptorArgs(Of Employee, String)) Dim lastName = args.Value If Not String.IsNullOrEmpty(lastName) Then args.Value = args.Value.ToUpper() End If End Sub

VB

DevForce will insure that this method is automatically called as part of any call to the Employee.LastName „get‟ property. The “AfterGet” attribute specifies that this method will be called internally as part of the „get‟ process “after” any internal get operations involved in the get are performed. The effect is that the LastName property will always return an uppercased result. For the remainder of this document, methods such as this will be termed interceptor actions. The corresponding „set‟ property can be intercepted in a similar manner.

C#

[BeforeSet(EntityPropertyNames.LastName)] public void UppercaseLastName(PropertyInterceptorArgs<Employee, String> args) { var lastName = args.Value; if ( !String.IsNullOrEmpty(lastName)) { args.Value = args.Value.ToUpper(); } } <BeforeSet(EntityPropertyNames.LastName)> _ Public Sub UppercaseLastName(ByVal args As PropertyInterceptorArgs(Of Employee, String)) Dim lastName = args.Value If Not String.IsNullOrEmpty(lastName) Then args.Value = args.Value.ToUpper() End If End Sub

VB

In this case we are ensuring that any strings passed into the „LastName‟ property will be uppercased before being stored in the Employee instance ( and later persisted to the backend datastore). Note that, in this case, the interception occurs “before” any internal operation is performed. In these two cases we have implemented an „AfterGet‟ and a „BeforeSet‟ interceptor. BeforeGet and AfterSet attributes are also provided and operate in a similar manner.

Named vs. Unnamed Interceptor Actions
The property interception code snippets shown above were all examples of what are termed „Named‟ interceptor actions, in that they each specified a single specific „named‟ property to be intercepted. It is also possible to create „Unnamed‟ interceptor actions that apply to all of the properties for a specific target type. For example, suppose that the following code were implemented in the Employee partial class:

141 | P a g e

IdeaBlade DevForce
C#

Property Interceptors

// Note that no parameter follows the BeforeSet attribute [BeforeSet] public void BeforeSetAny(IPropertyInterceptorArgs args) { if (!Thread.CurrentPrincipal.IsInRole("Administrator")) { throw new InvalidOperationException("Only admistrators can change data"); } }

VB

' Note that no parameter follows the BeforeSet attribute <BeforeSet> _ Public Sub BeforeSetAny(ByVal args As IPropertyInterceptorArgs) If Not Thread.CurrentPrincipal.IsInRole(“Administrator”) Then Throw New InvalidOperationException(“Only admistrators can change data”) End If End Sub

The result of this code would be that only those users logged in as administrators would be allowed to call any property setters within the Employee class. A similar „set‟ action might look like the following: C#
[AfterSet] public void AfterSetAny(IPropertyInterceptorArgs args) { LogChangeToEmployee(args.Instance); }

VB

<AfterSet> _ Public Sub AfterSetAny(ByVal args As IPropertyInterceptorArgs) LogChangeToEmployee(args.Instance) End Sub

This would log any changes to the employee class. Later in this document we will also describe how to define interceptors that apply across multiple types as well as multiple properties within a single type.

Interceptor Chaining and Ordering
Any given property may have more than one interceptor action applied to it. For example:

C#

[AfterGet(EntityPropertyNames.LastName)] public void UppercaseLastName(PropertyInterceptorArgs<Employee, String> args) { /// … do something interesting } [AfterGet(EntityPropertyNames.LastName)] // same mode (afterGet) and property name as above public void InsureNonEmptyLastName(PropertyInterceptorArgs<Employee, String> args) { // … do something else interesting } [AfterGet] // same mode as above and applying to all properties on employee. public void AfterAnyEmployeeGet(PropertyInterceptorArgs<Employee, Object> args) { // … global employee action here } <AfterGet(EntityPropertyNames.LastName)> _ Public Sub UppercaseLastName(ByVal args As PropertyInterceptorArgs(Of Employee, String))

VB

142 | P a g e

IdeaBlade DevForce
''' … do something interesting End Sub

Property Interceptors

<AfterGet(EntityPropertyNames.LastName)> _ Public Sub InsureNonEmptyLastName(ByVal args As PropertyInterceptorArgs(Of Employee, String)) ' … do something else interesting End Sub <AfterGet> _ Public Sub AfterAnyEmployeeGet(ByVal args As PropertyInterceptorArgs(Of Employee, Object)) ' … global employee action here End Sub

In this case, three different interceptor actions are all „registered‟ to occur whenever the Employee.LastName property is called. To execute these actions, the DevForce engine forms a chain where each of the „registered‟ interceptor actions is called with the same arguments that were passed to the previous action. Any interceptor can thus change the interceptor arguments in order to change the input to the next interceptor action in the chain. The „default‟ order in which interceptor actions are called is defined according to the following rules. 1) 2) 3) 4) 5) Base class interceptor actions before subclass interceptor actions. Named interceptor actions before unnamed interceptor actions. Attribute interceptor actions before dynamic interceptor actions. For attribute interceptor actions, in order of their occurrence in the code. For dynamic interceptor actions, in the order that they were added to the PropertyInterceptorManager.

Because of the rigidity of these rules, there is also a provision to override the default order that any interceptor action is called by explicitly setting its „Order‟ property. For attribute interceptors this is accomplished as follows:

C#

[BeforeSet(EntityPropertyNames.LastName, Order=-1.0)] public void UppercaseLastName(PropertyInterceptorArgs<Employee, String> args) { … } <BeforeSet(EntityPropertyNames.LastName, Order:=-1.0)> _ Public Sub UppercaseLastName(ByVal args As PropertyInterceptorArgs(Of Employee, String)) … End Sub

VB

The „Order‟ property is defined as being of type „double‟ and is automatically defaulted to a value of „0.0‟. Any interceptor action with a property of less that „0.0‟ will thus occur earlier than any interceptors without a specified order and any value greater that „0.0‟ will correspondingly be called later, and in order of increasing values of the Order parameter. Exact ordering of interceptor actions can thus be accomplished.

Multiple attributes on a single interceptor action
There will be cases where you want a single interceptor action to handle more than one property but less than an entire class. In this case, it may be useful to write an interceptor action similar to the following:

143 | P a g e

IdeaBlade DevForce
C#

Property Interceptors

[BeforeSet(EntityPropertyNames.FirstName)] [BeforeSet(EntityPropertyNames.LastName)] [BeforeSet(EntityPropertyNames.MiddleName)] public void UppercaseName(PropertyInterceptorArgs<Employee, String> args) { var name = args.Value; if ( !String.IsNullOrEmpty(name)) { args.Value = args.Value.ToUpper(); } } <BeforeSet(EntityPropertyNames.FirstName), BeforeSet(EntityPropertyNames.LastName), BeforeSet(EntityPropertyNames.MiddleName)> _ Public Sub UppercaseName(ByVal args As PropertyInterceptorArgs(Of Employee, String)) Dim name = args.Value If Not String.IsNullOrEmpty(name) Then args.Value = args.Value.ToUpper() End If End Sub

VB

The EntityPropertyNames class
In all of the previous examples we have shown „Named” attributes specified with the form “EntityPropertyNames.{PropertyName}. This is a recommended pattern that ensures type safety. However, the following two attribute specifications have exactly the same effect:

C#

[BeforeSet(EntityPropertyNames.FirstName)] // or [BeforeSet(“FirstName”)]

VB The „EntityPropertyNames‟ reference is actually to an inner class that is automatically generated inside each of the DevForce Entity classes. Its primary purpose is to allow specification of property names as constants. Note that the EntityPropertyNames class is defined as a partial class so that developers can add their own property names to the class for any custom properties that they create.

PropertyInterceptorArgs and IPropertyInterceptorArgs
Interceptor actions get all of the information about the context of what they are intercepting from the single interceptor argument passed into them. This argument will obviously be different for different contexts; i.e. a set versus a get action, a change to an employee versus a company, a change to the FirstName property instead of the LastName property. Because of this there are many possible implementations of what the single argument passed into any interceptor action might contain. However, all of these implementations implement a single primary interface: IPropertyInterceptorArgs. Every interceptor action shown previously provides an example of this. In each case, a single argument of type PropertyInterceptorArgs<Employee, String> or of type IPropertyInterceptorArgs was passed into each of the interceptor methods. In fact, the type of the „args‟ instance that is actually be passed into each of these methods at runtime is an instance of a subtype of the argument type declared in the methods signature. For any interceptor action defined on a DevForce entity, the actual args passed into the action will be a concrete implementation of one of the following classes.

DataEntityPropertyGetInterceptorArgs<TInstance, DataEntityPropertySetInterceptorArgs<TInstance,

TValue> TValue>

144 | P a g e

IdeaBlade DevForce NavigationEntityPropertyGetInterceptorArgs<TInstance, NavigationEntityPropertySetInterceptorArgs<TInstance,
TValue> TValue>

Property Interceptors

The boldfaced characters above indicate whether we are providing interception to a get or a set property, as well as whether we are intercepting a data or a navigation property. In general, you can write an interception method with an argument type that is any base class of the actual argument type defined for that interceptor. If you do use a base class, then you may need to perform runtime casts in order to access some of the additional properties provided by the specific subclass passed in at runtime. These subclassed properties will be discussed later. The entire inheritance hierarchy for property interceptor arguments is shown below: Assembly Where Defined IdeaBlade.Core Property Interceptor Arguments
IPropertyInterceptorArgs IPropertyInterceptorArgs<TInstance, TValue> PropertyInterceptorArgs<TInstance, TValue> DataEntityPropertyInterceptorArgs<TInstance, TValue>

IdeaBlade.EntityModel

DataEntityPropertyGetInterceptorArgs<TInstance, TValue> DataEntityPropertySetInterceptorArgs<TInstance, TValue> NavigationEntityPropertyInterceptorArgs<TInstance, TValue> NavigationEntityPropertyGetInterceptorArgs<TInstance, TValue> NavigationEntityPropertySetInterceptorArgs<TInstance, TValue>

The generic <TInstance> argument will always be the type that the intercepted method will operate on, known elsewhere in this document and the interceptor API as the “TargetType”. The <TValue> argument will be the type of the property being intercepted. i.e. „String‟ for the „LastName‟ property. Note that the interceptor arguments defined to operate on DevForce entities break into multiple subclasses with additional associated interfaces based on two primary criteria. 1) Is it a „get‟ or a „set‟ interceptor? a. „get‟ interceptor args implement IEntityPropertyGetInterceptorArgs b. „set‟ interceptor args implement IEntityPropertySetInterceptorArgs 2) Does it involve a „DataEntityProperty‟ or a „NavigationEntityProperty‟?. a. „DataEntityProperty‟ args implement IDataEntityPropertyInterceptorArgs b. „NavigationEntityProperty‟ args implement INavigationEntityPropertyInterceptorArgs

The API for each of the interfaces above is discussed below.

IPropertyInterceptorArgs
The root of all property interceptor arguments is the IPropertyInterceptorArgs interface. Its properties will be available to all interceptors.

145 | P a g e

IdeaBlade DevForce
C#
public interface IPropertyInterceptorArgs { Object Instance { get; } Object Value { get; set; } bool Cancel { get; set; } Action<Exception> ExceptionAction { get; set; } object Tag { get; set; } object Context { get; } }

Property Interceptors

VB

Public Interface IPropertyInterceptorArgs ReadOnly Property Instance() As Object Property Value() As Object Property Cancel() As Boolean Property ExceptionAction() As Action(Of Exception) Property Tag() As Object ReadOnly Property Context() As Object End Interface

In general the most useful of these properties will be the „Instance‟ and „Value‟ properties. The „Instance‟ property will always contain the „parent‟ object whose property is being intercepted. The „Value‟ will always be the value that is being either retrieved or set. The „Cancel‟ property allows you to stop the execution of the property interceptor chain at any point by setting the „Cancel‟ property to „true. The „ExceptionAction‟ property allows you to set up an action that will be performed whenever an exception occurs anywhere after this point in the chain of interceptors. The „Tag‟ property is intended as a general purpose grab bag for the developer to use for his/her own purposes. The „Context‟ property is used for internal purposes and should be ignored.

An example of using the ExceptionAction and Cancel is shown below: C#
[AfterSet] public void BeforeSetAny(IPropertyInterceptorArgs args) { // Do not let any setters throw an exception // Eat them and log them, and cancel the remainder of the set operation. args.ExceptionAction = (e) => { LogException(e); args.Cancel = true; }; } <AfterSet> _ Public Sub BeforeSetAny(ByVal args As IPropertyInterceptorArgs) ' Do not let any setters throw an exception ' Eat them and log them, and cancel the remainder of the set operation. args.ExceptionAction = Function(e) AnonymousMethod1(e, args) End Sub Private Function AnonymousMethod1(ByVal e As Object, ByVal args As IPropertyInterceptorArgs) As Object LogException(e) args.Cancel = True Return Nothing End Function

VB

146 | P a g e

IdeaBlade DevForce

Property Interceptors

Generic IPropertyInterceptorArgs
The following is a generic version of IPropertyInterceptorArgs where both the Instance and Value properties are now strongly typed; otherwise it is identical to IPropertyInterceptorArgs. C#
public interface IPropertyInterceptorArgs<TInstance, TValue> : IPropertyInterceptorArgs { TInstance Instance { get; } TValue Value { get; set; } bool Cancel { get; set; } Action<Exception> ExceptionAction { get; set; } object Tag { get; set; } object Context { get; } }

VB

Public Interface IPropertyInterceptorArgs(Of TInstance, TValue) Inherits IPropertyInterceptorArgs ReadOnly Property Instance() As TInstance Property Value() As TValue Property Cancel() As Boolean Property ExceptionAction() As Action(Of Exception) Property Tag() As Object ReadOnly Property Context() As Object End Interface

IEntity PropertyInterceptorArgs and subclasses
Whereas the interfaces above can be used to intercept any property on any object, the argument interfaces below are for use only with DevForce specific entities and complex objects. Each interface below provides additional contextual data to any interceptor actions defined to operate on DevForce entities. The most basic of these is simply the idea that each property on a DevForce entity has a corresponding “EntityProperty” ( discussed elsewhere in this guide).

C#

public interface IEntityPropertyInterceptorArgs : IPropertyInterceptorArgs { EntityProperty EntityProperty { get; } }

VB

Public Interface IEntityPropertyInterceptorArgs Inherits IPropertyInterceptorArgs ReadOnly Property EntityProperty() As EntityProperty End Interface

An example is shown below:

147 | P a g e

IdeaBlade DevForce
C#

Property Interceptors

[AfterSet] public void AfterSetAny(IPropertyInterceptorArgs args) { var entityPropertyArgs = args as IEntityPropertyInterceptorArgs; if ( entityPropertyArgs != null) { Log(“The “ + entityPropertyArgs.EntityProperty.Name + “ was set to the value: “ + args.Value.ToString()); } } <AfterSet> _ Public Sub AfterSetAny(ByVal args As IPropertyInterceptorArgs) Dim entityPropertyArgs = TryCast(args, IEntityPropertyInterceptorArgs) If entityPropertyArgs IsNot Nothing Then Log(“The “ + entityPropertyArgs.EntityProperty.Name + “ was set to the value:= “ + args.Value.ToString()) End If End Sub

VB

The next two interfaces provide additional context based on whether the interceptor action being performed is a „get‟ operation or a „set‟ operation. For a get operation, IdeaBlade entities have a concept of possibly multiple versions, i.e. an original, current, or proposed version, of an entity at any single point in time. It may be useful to know which „version‟ is being retrieved during the current action. Note that the version cannot be changed. C#
public interface IEntityPropertyGetInterceptorArgs : IEntityPropertyInterceptorArgs { EntityVersion EntityVersion { get; } }

VB

Public Interface IEntityPropertyGetInterceptorArgs Inherits IEntityPropertyInterceptorArgs ReadOnly Property EntityVersion() As EntityVersion End Interface

For a set operation, IdeaBlade has as part of its underlying implementation of any property the idea of possibly validating ( verifying) the incoming data. The VerificationSetterOptions property of any implementation of IEntityPropertySetInterceptorArgs provides the ability to determine whether a validation has or will be called as well as allowing any „BeforeSet‟ code to actually change how the verification will occur.
public interface IEntityPropertySetInterceptorArgs : IEntityPropertyInterceptorArgs { VerificationSetterOptions VerificationSetterOptions { get; set; } }

An example: C#
[AfterSet] public void BeforeSetAny(IEntityPropertySetInterceptorArgs args) { // turn off validation args.VerificationSetterOptions = VerificationSetterOptions.None; } <AfterSet> _ Public Sub BeforeSetAny(ByVal args As IEntityPropertySetInterceptorArgs) ' turn off validation args.VerificationSetterOptions = VerificationSetterOptions.None End Sub

VB

148 | P a g e

IdeaBlade DevForce

Property Interceptors

The DevForce EntityProperty is an abstract class with two concrete subclasses; a DataEntityProperty and a NavigationEntityProperty ( discussed elsewhere in this guide). The next two IEntityPropertyInterceptorArgs subinterfaces allow access to instances of one or the other of these depending on whether the property being intercepted is a data or a navigation property. C#
public interface IDataEntityPropertyInterceptorArgs : IEntityPropertyInterceptorArgs { DataEntityProperty DataEntityProperty { get; } } Public Interface IDataEntityPropertyInterceptorArgs Inherits IEntityPropertyInterceptorArgs ReadOnly Property DataEntityProperty() As DataEntityProperty End Interface

VB

C#

public interface INavigationEntityPropertyInterceptorArgs : IEntityPropertyInterceptorArgs { NavigationEntityProperty NavigationEntityProperty { get; } } Public Interface INavigationEntityPropertyInterceptorArgs Inherits IEntityPropertyInterceptorArgs ReadOnly Property NavigationEntityProperty() As NavigationEntityProperty End Interface

VB

IPropertyInterceptorArgs Type Coercion
One of the first issues that a developer will encounter with writing interceptor actions that handle more than one property is that it becomes difficult or impossible to use a concrete subtype as the argument to the interceptor. For example, imagine that we wanted to write a single action that handled two or more very different properties each of a different type: This could be written as follows:

C#

[BeforeSet(EntityPropertyNames.HireDate)] // hire date is of type datetime [BeforeSet(EntityPropertyNames.FirstName)] // firstname is of type string public void StrangeAction(IPropertyInterceptorArgs args) { var emp = (Employee) args.Instance; var entityProperty = ((IDataEntityPropertyInterceptorArgs) args).EntityProperty; .. do some very baroque operation with emp and entityProperty } <BeforeSet(EntityPropertyNames.HireDate), BeforeSet(EntityPropertyNames.FirstName)> _ Public Sub StrangeAction(ByVal args As IPropertyInterceptorArgs) Dim emp = CType(args.Instance, Employee) Dim entityProperty = (CType(args, IDataEntityPropertyInterceptorArgs)).EntityProperty .. do some very baroque operation with emp and entityProperty End Sub

VB

But ideally we would prefer to write it like this, in order to avoid performing a lot of superfluous casts:

149 | P a g e

IdeaBlade DevForce
C#

Property Interceptors

[BeforeSet(EntityPropertyNames.HireDate)] // hire date is of type datetime [BeforeSet(EntityPropertyNames.FirstName)] // firstname is of type string public void StrangeAction(DataEntityPropertySetInterceptorArgs<Employee, Object> args) { // no casting var emp = args.Instance; var entityProperty = args.DataEntityProperty; .. some very baroque operation } <BeforeSet(EntityPropertyNames.HireDate), BeforeSet(EntityPropertyNames.FirstName)> _ Public Sub StrangeAction(ByVal args As DataEntityPropertySetInterceptorArgs(Of Employee, Object)) ' no casting Dim emp = args.Instance Dim entityProperty = args.DataEntityProperty .. some very baroque operation End Sub

VB

The problem is that, according to the rules of inheritance, the two concrete classes that this method will be called with:
Type 1: DataEntityPropertySetInterceptorArgs<Employee, String> Type 2: DataEntityPropertySetInterceptorArgs<Employee, DateTime>

…do not inherit from:
Type 3: DataEntityPropertySetInterceptorArgs<Employee, Object>

In fact, the only class or interface that they do share is:
IPropertyInterceptorArgs

So in order to allow this construction, DevForce needs to “coerce” each of „Type1‟ and „Type2” into „Type3” for the duration of the method call. Because DevForce does do this, any of the following arguments are also valid:
Type 4: DataEntityPropertySetInterceptorArgs<Entity, Object> Type 5: DataEntityPropertySetInterceptorArgs<Object, Object> Type 5: PropertyInterceptorArgs<Employee, Object> … etc.

The basic rule for the type coercion facility is that any concrete type can be specified if its generic version is a subtype of the generic version of the actual argument type that will be passed in.

PropertyInterceptor Attribute Discovery
In general, any interceptor method declared within a DevForce entity and marked with a property interceptor attribute will be automatically discovered before the first property access. PropertyInterceptors will most commonly be defined within the developer-controlled partial class associated with each entity. Property interceptors can also be defined on any base class and these will also be discovered automatically. In order to reduce the surface area of any entity class, a developer may not want to expose the property interceptor methods directly on the surface of his or her class. To facilitate this, DevForce will also probe any public inner classes of any class and will locate any property interceptors defined there as well. Example:

150 | P a g e

IdeaBlade DevForce
C#
public partial class Employee : IdeaBlade.EntityModel.Entity {

Property Interceptors

// internal class just for property interceptors public class PropertyInterceptorsDefinitions { [BeforeGet(Employee.EntityPropertyNames.LastName)] public void LastNameInterceptor(IEntityPropertyInterceptorArgs args) { … } [AfterSet] public void LoggingInterceptor(IEntityPropertyInterceptorArgs args) { … } }

VB

Partial Public Class Employee Inherits IdeaBlade.EntityModel.Entity ' internal class just for property interceptors Public Class PropertyInterceptorsDefinitions <BeforeGet(Employee.EntityPropertyNames.LastName)> _ Public Sub LastNameInterceptor(ByVal args As IEntityPropertyInterceptorArgs) … End Sub <AfterSet> _ Public Sub LoggingInterceptor(ByVal args As IEntityPropertyInterceptorArgs) … End Sub End Class

One important note: property interceptor methods defined on a class directly may be either instance or static methods; whereas property interceptors defined on an inner class (or anywhere other than directly on the entity class) must be static methods. In the event that a developer wants to completely isolate his interception methods in another non-entity-based class, then discovery will not occur automatically. In this case, the DiscoverInterceptorsFromAttributes(Type targetType) method on the PropertyInterceptorManager class may be used to force discovery of any specified type and all of its base types. Attribute interceptors that are declared outside of the classes to which they apply must be further qualified via the “TargetType” property as shown below: C#
public class UnattachedInterceptor { [AfterSet(User.EntityPropertyNames.Name,

TargetType=typeof(User)]

public void LoggingInterceptor(IEntityPropertyInterceptorArgs args) { … } }

VB

Public Class UnattachedInterceptor <AfterSet(User.EntityPropertyNames.Name, TargetType:=GetType(User)> _ Public Sub LoggingInterceptor(ByVal args As IEntityPropertyInterceptorArgs) … End Sub End Class

151 | P a g e

IdeaBlade DevForce
Alternative PropertyInterceptor Attribute Method Signatures

Property Interceptors

While the property interceptor methods described previously allow a great deal of control over the entire interception process, there are times when this is overkill. Sometimes all you really want is to do is modify or inspect the incoming or outgoing values. In these cases, a simplified signature for an interception method is also provided. For example the following standard interceptor action: (Developer code) C#
[AfterGet(EntityPropertyNames.LastName)] public void UppercaseLastName(PropertyInterceptorArgs<Employee, String> args) { var lastName = args.Value; if ( !String.IsNullOrEmpty(lastName)) { args.Value = args.Value.ToUpper(); } } <AfterGet(EntityPropertyNames.LastName)> _ Public Sub UppercaseLastName(ByVal args As PropertyInterceptorArgs(Of Employee, String)) Dim lastName = args.Value If Not String.IsNullOrEmpty(lastName) Then args.Value = args.Value.ToUpper() End If End Sub

VB

can also be written as (Developer code) C#
[AfterGet(EntityPropertyNames.LastName)] public String UppercaseLastName(String lastName) { if ( !String.IsNullOrEmpty(lastName)) { return lastName.ToUpper(); } else { return String.Empty; } } <AfterGet(EntityPropertyNames.LastName)> _ Public Function UppercaseLastName(ByVal lastName As String) As String If Not String.IsNullOrEmpty(lastName) Then Return lastName.ToUpper() Else Return String.Empty End If End Function

VB

In general, any property interceptor action that only inspects or modifies the incoming value without the need for any other context can be written in this form. In fact, if the action does not actually modify the incoming value, the return type of the interceptor action can be declared as void.

Dynamic Property Interception and the PropertyInterceptorManager.
Property interceptors can be added and removed dynamically by making use of the PropertyInterceptorManager and the PropertyInterceptor classes. Their API‟s are shown below:

152 | P a g e

IdeaBlade DevForce
C#

Property Interceptors

public sealed class PropertyInterceptorManager { public static PropertyInterceptorManager CurrentInstance { get; set; } public void DiscoverInterceptorsFromAttributes(Type targetType) public void AddAction(PropertyInterceptorAction interceptorAction) public bool RemoveAction(PropertyInterceptorAction interceptorAction) public IList<PropertyInterceptorAction<TArgs>> GetActions<TArgs>(Type targetType, String targetName, PropertyInterceptorMode mode) where TArgs : class, IPropertyInterceptorArgs }

VB

Public NotInheritable Class PropertyInterceptorManager Private privateCurrentInstance As PropertyInterceptorManager Public Shared Property CurrentInstance() As PropertyInterceptorManager Get Return privateCurrentInstance End Get Set(ByVal value As PropertyInterceptorManager) privateCurrentInstance = value End Set End Property public void DiscoverInterceptorsFromAttributes(Type targetType) public void AddAction(PropertyInterceptorAction interceptorAction) public Boolean RemoveAction(PropertyInterceptorAction interceptorAction) public IList(Of PropertyInterceptorAction(Of TArgs)) GetActions(Of TArgs)(Type targetType, String targetName, PropertyInterceptorMode mode) where TArgs : class, IPropertyInterceptorArgs End Class

C#

public class PropertyInterceptorAction<TArgs> : PropertyInterceptorAction where TArgs : class, IPropertyInterceptorArgs { public PropertyInterceptorAction(Type targetType, String targetName, PropertyInterceptorMode mode, Action<TArgs> action); public PropertyInterceptorAction(Type targetType, String targetName, PropertyInterceptorMode mode, Action<TArgs> action, Double order, String key); public Type TargetType { get; } public String TargetName { get; } } public PropertyInterceptorMode Mode { get; } public String Key { get; } public Double Order { get; } public Type ArgsType { get; } public Type InstanceType { get; } public Type ValueType { get; } public PropertyInterceptorAction<TArgs> ConvertTo<TArgs>() where TArgs : class, IPropertyInterceptorArgs; }

153 | P a g e

IdeaBlade DevForce
VB

Property Interceptors

Public Class PropertyInterceptorAction(Of TArgs As {Class, IPropertyInterceptorArgs}) Inherits PropertyInterceptorAction public PropertyInterceptorAction(Type targetType, String targetName, PropertyInterceptorMode mode, Action(Of TArgs) action) public PropertyInterceptorAction(Type targetType, String targetName, PropertyInterceptorMode mode, Action(Of TArgs) action, Double order, String key) public Type TargetType {get;} public String TargetName {get;} End Class public PropertyInterceptorMode Mode {get;} public String Key {get;} public Double Order {get;} public Type ArgsType {get;} public Type InstanceType {get;} public Type ValueType {get;} public PropertyInterceptorAction(Of TArgs) ConvertTo(Of TArgs)() where TArgs : class, IPropertyInterceptorArgs

Since there is no public constructor for the PropertyInterceptorManager class, the only instance available to the developer is via the „CurrentInstance‟ property. This property will always have a value. The current instance is the container for all currently „registered” interceptor actions. PropertyInterceptorActions can be created via the PropertyInterceptorAction class and added to the PropertyInterceptorManager.CurrentInstance as shown below: (Developer code) C#
var piAction = new PropertyInterceptorAction<PropertyInterceptorArgs<Employee, String>>( typeof(Employee), Employee.LastNameEntityProperty.Name, PropertyInterceptorMode.BeforeGet, (args) => args.Value = arg.Value.ToUpper); PropertyInterceptorManager.CurrentInstance.AddAction(piAction); 'INSTANT VB TODO TASK: Assignments within expressions are not supported in VB.NET 'ORIGINAL LINE: var piAction = New PropertyInterceptorAction(Of PropertyInterceptorArgs(Of Employee, String))(typeof(Employee), Employee.LastNameEntityProperty.Name, PropertyInterceptorMode.BeforeGet, (args) => args.Value = arg.Value.ToUpper); Dim piAction = New PropertyInterceptorAction(Of PropertyInterceptorArgs(Of Employee, String))(GetType(Employee), Employee.LastNameEntityProperty.Name, PropertyInterceptorMode.BeforeGet, Function(args) args.Value = arg.Value.ToUpper) PropertyInterceptorManager.CurrentInstance.AddAction(piAction)

VB

Interceptor actions can be removed in a similar manner. This mechanism also allows the application of an interceptor action to a base class that is then, in turn, applied to all of its subclasses. As a somewhat contrived example, you might want to completely disable all setters in an application via a call like this: C#
var piAction = new PropertyInterceptorAction<PropertyInterceptorArgs<Employee, String>>( typeof(Object), // everyone inherits from object null, // no property name PropertyInterceptorMode.BeforeSet, (args) => throw new Exception(“No sets allowed”); PropertyInterceptorManager.CurrentInstance.AddAction(piAction); 'INSTANT VB NOTE: This code snippet uses implicit typing. You will need to set 'Option Infer On' in the VB file or set 'Option Infer' at the project level: Dim piAction = New PropertyInterceptorAction(Of PropertyInterceptorArgs(Of Employee,

VB

154 | P a g e

IdeaBlade DevForce

Property Interceptors

String))(GetType(Object), Nothing, PropertyInterceptorMode.BeforeSet, Function(args) throw New Exception(“No sets allowed”); PropertyInterceptorManager.CurrentInstance.AddAction(piAction)

EntityProperties and Property Interceptors
Within a DevForce application, every property interceptor has a GetterInterceptor and a SetterInterceptor property. These properties can also be used to modify the property interceptor actions associated with that property. Under the covers this is going through the PropertyInterceptorManager mechanism described above, but the syntax is often simpler. For example:

C#

Employee.AddressEntityProperty.SetterInterceptor.AddAction( PropertyInterceptorTiming.Before, args => args.Value = AddZipCode(args.Value)); 'INSTANT VB TODO TASK: Assignments within expressions are not supported in VB.NET 'ORIGINAL LINE: Employee.AddressEntityProperty.SetterInterceptor.AddAction(PropertyInterceptorTiming.Before, args => args.Value = AddZipCode(args.Value)); Employee.AddressEntityProperty.SetterInterceptor.AddAction(PropertyInterceptorTiming.Before, Function(args) args.Value = AddZipCode(args.Value))

VB

PropertyInterceptor keys
Every property interceptor action has a key that can either be specified via an optional attribute property or dynamically when the action is first created. If no key is defined, the system will automatically create one. This key will be used to identify an action for removal. The PropertyInterceptorManager.RemoveAction(interceptorAction) attempts to find an interceptor that matches the one passed in. This match requires that the TargetType, TargetName, Mode, and Key be the same between the two interceptor actions.

Mechanics of Property Interception
Property interception within DevForce is accomplished by dynamically generating compiled lamda expressions for each interceptor action. DevForce interceptors are discovered (but not compiled) as each entity class is first referenced. Runtime compilation of each property interceptor occurs lazily the first time each property is accessed. After this first access, the entire code path for each property access is fully compiled. Properties that are never accessed do not require compilation. The addition or removal of interceptor actions after they have been compiled does require a new compilation the next time the property is executed. This happens automatically. Errors encountered during the compilation process will thus appear when a property is accessed for the first time. These exceptions will be of type PropertyInterceptorException and will contain information on the specific method that could not be compiled into a property interceptor action. These are usually a function of a PropertyInterceptorArgs parameter type that is not compatible with the property or properties being accessed.

155 | P a g e

IdeaBlade DevForce

Business Object Persistence

Business Object Persistence

Business Object Persistence
Note: Code Snippets in This Document

Object Persistence Overview
The Big Picture DevForce and the ADO.NET EntityModel Locating the Physical Data Source with a Key Support for POCOs (Plain Old CLR Objects) Persistence Management Capabilities Retrieving business objects The Entity Cache Business objects in motion Creating new business objects Saving and undoing business object changes Offline Support Application Security Business Object Security N-Tier Architecture Three-Tier Deployment Model Choice by Configuration Conclusion

Entity Queries and Entity Navigation
Entity Queries Query v. Method Syntax LINQ The DevForce Predicate Builder Example: Simulate an In() Clause Condition on a Distantly Related Entity The PredicateDescription Class Example: Given a Collection of Parent Entities, Retrieve the Related Children PassthruESQL Queries Remote Service Method Call (RSMC) Entity Navigation Parent-Child Navigation properties Navigation Properties in Silverlight Deferred Retrieval Proactive Data Loads Missing objects The Null Entity

Asynchronous Communication with the Business Object Server
Asynchronous Queries IAsyncResult Asynchronous Pattern Asynchronous Fulfillment of Navigation Property Queries Canceling Pending Operations

The EntityListManager Entity Caching
All Business Objects are Cached Entity Ancestry and Organization of the Cache

IdeaBlade DevForce

Business Object Persistence

Business objects are unique in each cache Entities in Lists Business object proper, not the business object graph Queries, Navigation, and the Cache Query Cache Primary key queries “Object Not Found” and the Null Entity Cache use when disconnected Modifications Stale Entity Data Fetch Life Cycle Events Query Workflow Query Strategy Fetch Strategies MergeStrategies InversionMode Pre-Defined QueryStrategies Custom QueryStrategies DefaultQueryStrategy When to Use The Different QueryStrategies Making a One-Time Change to the QueryStrategy With Which a Given Query Is Run Span Queries Performance Details Cached Entity Lifespan Saving the Cache Locally The TraceViewer: Watch What Data Is Being Loaded, and How Using the Trace Viewer Stand-Alone Embedding the Trace Viewer in Your Application Embedding the WPFTraceViewer in Your WPF App Embedding the WinTraceViewer in Your WinForms App Getting Generated SQL to Display in the TraceViewer Using the Trace Viewer With a Silverlight App

Creating Business Objects
When Not to Create The Business Object Create Method Generating unique identifiers GUIDs Sql Server Identity Ids Custom id generation Ids in mapping objects Creating a valid business object Auxiliary Business Object Class Methods CompareTo() ToString() Adding and Removing Related Objects using Add() and Remove() Business Object Creation Review

Saving Business Objects
EntityState of an Object Undo Multi-level Undo Validation Temporary Id Fix-up Life Cycle Events Client-Side Life Cycle Events Saves and Transaction Management

IdeaBlade DevForce
Distributed Transactions Re-query After Save When Save Fails SaveChanges() Exceptions EntityManagerSaveException SaveResult Alternatives to Default SaveChanges Exceptions Data Source Concurrency Saving the “Dependency Graph” Association Types Compositions Save the Root Entity Saving Event Handler Composition Business Rules Concurrency Violations Dependency Graph Retrieval Workflow For a Save Saving the Cache to a Local Disk File XML Serialization of Business Objects

Business Object Persistence

Note: Code Snippets in This Document
The code snippets in this document are duplicated in accompanying C# and VB code solutions. After installing DevForce, you will find these solutions in the _TopicDocumentSnippets folder under the Business Object Persistence topic in the Learning Resources. The captions associated herein with the snippets reflect the corresponding method names in the code solutions. You will find these methods in the Program.cs or Main.vb files, respectively, for the C# and VB solutions.

Object Persistence Overview
In previous chapters you‟ve seen how object mapping declares relationships between business objects and remote data sources. You learned that it generated classes for each business object as well as some helper classes such as EntityRelations. The collection of these classes constitutes the application‟s business object model. In this chapter we describe how the DevForce persistence scheme works with the business object model. You will learn that instances of the business object class (AKA the entity class) are held in a container called the entity cache. This cache belongs to and is managed by an instance of the DevForce EntityManager class. You will discover that an EntityManager instance is rich in capabilities that go beyond retrieving and saving business objects. We‟ll introduce them here and elaborate on a few of them in subsequent sections. By the end of the chapter, you will appreciate that the EntityManager class is one of the most important and useful classes in the DevForce framework.

The Big Picture
A DevForce application relies upon a layered architecture for data access. At one end is a data source – typically a relational database. At the other end is the user interface which works with business objects in a business object model. There are several components in the middle.

IdeaBlade DevForce

Business Object Persistence

Figure 4. Cross-tier flow of data and business objects.

One of them, called an EntityServer, moves data (and data requests) between the ADO.NET Entity Framework and DevForce business objects. If the back-end data store is a relational database, the EntityServer leaves the direct communication with the data store to the ADO.NET Entity Framework. However, if the back-end data store is a web service, the DevForce EntityServer handles the job, since that capability does not exist within the Entity Framework. The EntityServer has a copy of the application‟s business object model so that it can instantiate DevForce business objects server-side if need be. However, for most operations (such as simple data retrievals), it forwards to the client-side EntityManager the data required for hydrating DevForce business objects there, without ever instantiating DevForce business objects on the server. The data is packaged and passed in a highly efficient format and process. The ADO.NET Entity Data Model includes the mapping information necessary to translate between locations in a relational data source and the corresponding persistent fields in the ADO.NET business entities. The EntityServer (besides handling those jobs against web services), mediates between the Entity Framework and the DevForce EntityManager that manages the client-side cache used by your application. The EntityServer is an important component and you should understand its role in the object persistence process. That said, you will seldom see or deal with it directly. The second important DevForce component is the EntityManager. The EntityManager takes instruction from the higher levels of the application such as the UI, and forwards UI requests for entities to the EntityServer. The EntityManager puts the received entities – obtained from whatever source by the EntityServer -- into its entity cache and makes them available to the UI. End users review the entities and make changes through the UI. The UI signals the EntityManager to save the changes. The EntityManager dutifully forwards the changed entities to the EntityServer which communicates with the appropriate component to commit the data into persistent storage.

IdeaBlade DevForce

Business Object Persistence

DevForce and the ADO.NET EntityModel
Visual Studio‟s ADO.NET Entity Data Model wizard creates an EDMX file which contains descriptions of a conceptual data schema (the object model), an actual data store schema (the database model), and the mappings between the two. It also renders the object model in .NET code in a file named <ModelName>.Designer.cs (or .vb). The developer‟s first step in building the object model for her application will consist in creating an entity model in an EDMX file. Typically s/he will use the Visual Studio Entity Data Model wizard to create the initial version of the EDMX file and the corresponding generated code file. After that, he will work with some combination of the Visual Studio Entity Model Designer and direct XML coding in the EDMX file, depending upon his preferences and whether he needs to use features in his model that are not supported by the Entity Model designer. The second step will be to create a Domain model using the DevForce Object Mapper. This model is so named because it will be composed of one or more Entity Models persisted in .EDMX files. The DevForce Object Mapper will alter the .EDMX file by adding additional elements and attributes. These added features are ignored, and left undisturbed, by the ADO.NET Entity Data Model Designer. Because of this, the developer can move back and forth between the Visual Studio Entity Model Designer and the DevForce Object Mapper without fear of either disturbing the other‟s work. There is, by intent, some overlap in the the functionality of the DevForce Object Mapper and ADO.NET Entity Data Model Designer. Over time, this overlap may increase as we subsume additional aspects of the Entity Model Designer‟s functionality. Our goal is to make it as convenient as possible for you to work with your model. However, our intial work on the DevForce Object Mapper has focussed on providing needed or useful capabilities that are either not present, or are difficult to work with, in the Entity Data Model Designer. We mentioned that the Entity Data Model wizard and designer, in addition to altering the .EDMX file, generates the classes that comprise the compilable manifestation of the object model. From the Object Mapper‟s enhanced version of the .EDMX, DevForce generates two sets of classes. The first is essentially the same Entity Framework model generated by the Visual Studio tools. This version of your object model will be deployed to the logical middle tier of your application, where it is used by the ADO.NET Entity Framework for creating objects of the type that it understands. The second version of the object model generated by the DevForce Object Mapper is a DevForce version consisting of business classes that inherit from IdeaBlade.EntityModel.Entity. As previously mentioned, we refer to this version of the model as the Domain model. The Domain model is “persistence ignorant”: unlike the Entity Framework model, it has no knowledge whatsoever of the back-end datastore or the mapping between that and its objects. In an n-tier deployment, it is the only model that is deployed client side. The client needs no connection information for back-end datasources. For those familiar with DevForce Classic (mated with .NET 2.0): the Entity Framework model essentially takes over the function handled in DevForce Classic by the .ORM file. Both contain knowledge of the data source and mapping information. A copy of the assembly containing the Domain model is also deployed server-side in an n-tier deployment. Architecture of the DevForce Business Object Class The (partial) inheritance hierarchy for a DevForce business class is as follows:

IdeaBlade DevForce
Figure 5. Inheritance Hierarchy for a DevForce Business Class

Business Object Persistence

The class for a business type is generated as one or two partial classes. In the partial classes labelled in the picture as DevForce-controlled, the essential data structure of the type is defined. This partial class is driven by settings in the domain model and gets regenerated whenever the develop instructs the DevForce Object Mapper to regenerate code. Thus it should never be modified by the developer. All DevForce-controlled partial classes for types originating from a given Entity Data Model are generated into a single file, named <DomainModelName>.<EntityModelName>.Designer.cs (or .vb, if generated in Visual Basic rather than C#). For example, the code file for the ServerModelNorthwindIB Entity Data Model of a domain model named DomainModel generated in C# would be named DomainModel. ServerModelNorthwindIB.Designer.cs, as shown at right. If the domain model includes multiple Entity Models, one such code file will be generated for each model, as shown at left. The partial class described in Figure 5 as “Developer-controlled” is optional, and can be generated by the Object Mapper in a one-time operation, or hand-coded by the developer. In either case, once it exists, the Object Mapper will not overwrite or modify it. The developer-controlled partial class is the developer‟s workshop, where he can add custom properties, methods, and events, as well as create property interceptors 19 to change the getter/setter behavior of properties defined in the DevForce-controlled partial class. If the developer asks the Object Mapper to generate developer partial classes, it will generate one such class for each type in the domain model. Each such partial class will be generated into its own file, which bears the name of the type. You can see this at right.

Again, these files are generated by the Object Mapper only when they do not already exist, and are not touched subsequently. Thus the developer can safely add her own code to this file without fear that it will be overwritten.

19

See the Developers Guide chapter on “Property Interceptors”

IdeaBlade DevForce

Business Object Persistence

If you are already familiar with the Entity Framework, you will note that DevForce code generation proceeds according to the same pattern used by the Entity Framework. The Entity Framework also generates partial classes for each type in a model, and all into a single class. It does not generate partial classes for developer work, but does permit the developer to create and maintain such partial classes.

Modifying the behavior of a generated property
DevForce provides a mechanism to intercept and either modify or extend the behavior of any .NET property, including, of course, those generated into the DevForce-controlled partial business classes. You can accomplish virtually any desired behavior modification of property getters or setters via this interception mechanism. The mechanism replaces, and expands upon, the technique of marking properties as virtual and overriding them in a subclass. This facility is a lightweight form of what is termed “Aspect-Oriented Programming”. You can find detail about this in the chapter, “Property Interceptors”.

Locating the Physical Data Source with a Key
How does the EntityServer locate the physical storage to use? You learned earlier that every business object – every concrete entity – is mapped to a particular data source. That data source is identified symbolically by a data source key. That key is compiled into the entity and cannot be changed at run-time. The EntityServer has a copy of the business object model so it knows the data source key for each kind of business object. But the key is purely symbolic. It does not contain the location of a physical data source nor can it determine how to connect to such a data source. It does not contain a database connection string, for example. Fortunately, the EntityServer also has a private copy of the application configuration file. It can use the data source key to find in that file the physical data source configuration information it needs such as the connection string for the physical data source it should use. This is all we need to know for the moment to assure ourselves that a DevForce application actually can move data between a physical data source and business objects in the client application. We turn next to the EntityManager which is the keeper of business objects on the client side.

Support for POCOs (Plain Old CLR Objects)
DevForce supports POCOs: instances of such objects can be queried, cached, updated, and saved just like DevForce entities. Consider Figure 6. DevForce EntityManager Support for POCOs. The business entity you saw diagrammed earlier in Figure 5. Inheritance Hierarchy for a DevForce Business Class is now shown wrapped by a DevForce EntityWrapper. Alternatively, a POCO is wrapped. The abstraction of the EntityWrapper permits the DevForce EntityManager to work with either type of object.

IdeaBlade DevForce
Figure 6. DevForce EntityManager Support for POCOs

Business Object Persistence

POCOs are discussed in detail later in this chapter (“POCO Support in DevForce”).

Persistence Management Capabilities
In this section we introduce the most important capabilities of the EntityManager. Some topics deserve extended attention and are discussed more fully in later chapters but you‟ll get a preview here of how DevForce persistence management can            retrieve business objects from data sources manage them in its cache move business objects across the Internet create new business objects save additions, changes, and deletions to a data source restore pending changed and deleted objects to their retrieved state continue to function when disconnected (even in Silverlight!) preserve cache contents temporarily in local storage log in and log out of the central server ensure business object security, and exploit an n-tier architecture.

Retrieving business objects
DevForce applications deal in business objects. Accordingly, the DevForce retrieval mechanisms return business objects. There are two such mechanisms: entity queries and entity navigation. An entity query hunts for objects with attributes that match specified search criteria. Suppose you need a list of employees over 40. In DevForce you could express this criterion in a LINQ-To-DevForce query which could be

IdeaBlade DevForce

Business Object Persistence

enumerated over to return a collection of Employee entities that happens to include sales representative “Nancy Davolio.” Entity navigation involves traversing from one kind of business object to another along a relation between them. You can navigate from a sales order to “Nancy” with an expression such as anOrder.SalesRep. This returns an Employee entity. Entity navigation can return a collection of entities as well. The expression aSalesRep.Orders returns the orders assigned to this employee sales rep. The orders are returned in special kind of generic list whose contents are managed by the EntityManager, a feature you‟ll find especially useful in your UI. The section “Entity Queries and Entity Navigation” offers greater detail.

The Entity Cache
A EntityManager caches business objects both for performance and to enable offline operation of the application. Each instance of EntityManager has its own cache of entities. Entities enter the cache in one of three ways:    from a data source as a result of entity query or entity navigation by creation as new entities by import from another EntityManager or outside source

Most entities enter the cache from a data source. Standard entity queries and entity navigations check the cache first to see if the desired objects are present; they resort to the data source only if the objects are not found 20. This behavior is usually desirable as it improves performance. The risk is that the entities in the cache become stale. The programmer can, at his election, by-pass the cache and query the database directly (the query results still end up in the cache). There are a host of other options which are addressed in the section “Entity Queries and Entity Navigation”. After a successful query, the cache holds the root business objects of the result. If you searched for employees, the cache will hold employee entities. The cache may hold other related entities as well. But it may not, and you shouldn‟t assume that it holds the entire business object graph of an employee after retrieving that employee. For example, after querying for “Nancy Davolio”, she is in the cache, but the Orders for which she is responsible as Sales Rep probably are not. A cache holds at most one copy of a business object. Recall that a business object has a unique identity implemented as a unique primary key. There is only one Employee in the application universe with an Id = 42. If follows that there can only be one Employee in the cache with Id = 42.21 Finally, entities stay in the cache until the application terminates or they are removed explicitly. If your application retrieves a great deal of data, you may have to take steps to prevent cache overflow, and a variety of facilities are provided to assist with this. However, for most applications this never even becomes an issue. We‟ll have more to say about caching in the coming pages.

20

DevForce keeps a cache of query objects for use in determining whether requested objects are already in the cache; we‟ll cover this in more detail later.

21

An application can actually have more than one EntityManager instance, though this is a needed only in sophisticated
applications and for special purposes. Each EntityManager instance will have its own cache, and each cache can contain an instance of any given business object. But every entity instance knows its own EntityManager. If we ever encounter two Employee entities with Id = 42, we can ask them “who is your EntityManager?” For more information on the use of multiple EntityManagers, see the section “Multiple Entity Manager Instances” under “Advanced Business Object Concepts”. For the balance of the current discussion, we will assume the application uses just one EntityManager instance.

IdeaBlade DevForce Business objects in motion

Business Object Persistence

The EntityServer and EntityManager exchange data in a highly optimized manner. Because of our efficient, automatic, and as-needed dehyration and rehydration of objects, as well as our seamless interaction with the Microsoft Entity Framework and its objects, your experience of the exchange of data between logical tiers is that it is simply DevForce business objects that are moving back and forth. A DevForce business object sent from the EntityManager to the EntityServer, or vice versa, is, in all important respects, exactly the same object when it arrives as when it left. It is of the same type, with the same persistable field values, properties, methods, and events. In practical effect, the entire object has traveled over the network; it is a “mobile business object.” There are two important implications.  Developers write one business object class with the full capacity to execute on either the client or server as required. They don‟t write one kind of object for the server and a different one for the client. They write one class, period. The application can be deployed on one physical tier, two tiers, three, or n-tiers – without recoding.

We guarantee complete object fidelity for cross-process or cross-machine communication, achieving this through a combination of storage format, serialization methods, transport mechanisms, and data merge facilities.

Creating new business objects
The developer can make a new entity by invoking either a constructor or a factory method that returns an instance of the business object. For most circumstances we recommend the latter technique, since it permits you fully to control the details of the instantiation (such as initialization of required properties). You write the factory method, and typically call it Create, making it a public static (Shared in VB) method of the business object‟s developer-controlled partial class; e.g., Employee.Create(). There are four steps to the typical Create method:     Get a prototype instance of the new entity from the EntityManager Give the prototype a unique identity Initialize some of its values Add the completed prototype to the EntityManager‟s cache

We explore these steps in the section “Creating Business Objects”.

Saving and undoing business object changes
Adding, changing, and deleting are operations affecting business objects in a EntityManager cache only. They are purely local modifications. They have no effect on the database and are invisible to other application users. The developer updates the database by telling the EntityManager to save changes. The developer can save an individual entity, an arbitrary list of entities, or all entities with pending changes. The wise developer will validate the business objects before saving them. The developer can also undo the changes in which case the affected business objects are restored to their state when last retrieved.22 We explore these summary remarks with greater depth in the section “Saving Business Objects”.
22

DevForce also provides a facility known as “checkpointing” that provides a transaction facility for operations in the local cache. Checkpointing gives you the ability to undo changes back to a specified state, perhaps not so far back as the state when retrieved from the data source. The utility of “checkpointing” is most apparent in the UI so we cover it in the WinForm User Interfaces chapter in the topic “Multi-Level Undo with Checkpoints”.

IdeaBlade DevForce Offline Support

Business Object Persistence

A client application can lose its connection to the central servers. The interruption may be brief, sudden, and unexpected, as when a mobile device loses its signal; or it may be voluntary and last for hours, as when the user runs the application offline on an airplane. An application which is susceptible to connection failures is called a “partially connected” or “intermittently connected” application. A DevForce smart client application can operate when disconnected -- whether suddenly and unexpectedly or on purpose -- for any length of time. It can be shut down and re-started without skipping a beat. While disconnected, the application can still create new objects and modify or delete cached entities. Such changes accumulate in the cache until the application reconnects and performs a save. All it takes is a little programming using some simple DevForce EntityManager features. Step #1: Manage the connection. The developer can control voluntary connection to the host and respond to unexpected disconnects with the help of a small number of EntityManager properties, methods, and events. Step #2: Save a copy of the cache locally. The typical sequence is: 1. 2. 3. 4. 5. Fill the cache with entities that will be needed while running disconnected. Disconnect and continue running. Save the cache to the client‟s local storage (e.g., a file) just before exit. Shut down. On re-launch, restore the cache from the client copy.

All pending changes are preserved across the two sessions. See the “Saving the Cache Locally” section of the “Business Object Caching” chapter to learn more.

Application Security
We‟ll devote a later chapter to securing your application, so we‟ll just mention the topic briefly in this overview. Application security has three aspects:    Confidentiality Authentication Authorization

Confidentiality – A secure application guards against eavesdroppers intercepting and reading traffic flowing between client and host. DevForce supports a variety of encryption measures including standard SSL. They are discussed in the Security chapter of this Developers Guide. Authentication – A secure application employs an authentication scheme to ensure that both parties to a connection are who they claim to be. In a smart-client context there are two authentication burdens: (1) the server must confirm and remain confident it is talking to a real, authorized client and (2) each client must be confident it is conversing with an authentic server. DevForce has mechanisms to support both kinds of checks. Authorization –The EntityManager‟s Login method stamps the client-side application thread with a Principal object representing the authenticated user. This Principal has an IsInRole method that returns true if the user participates in a named role passed to it. The developer has total flexibility in determining the implementation of the login method, the IPrincipal object returned from it, and the definition and usage of the role scheme.

IdeaBlade DevForce

Business Object Persistence

For its own part, DevForce maintains a tamper-proof SessionBundle object that is used to authenticate every transaction between the EntityManager and EntityServer.

Business Object Security
A secure application prevents improper access to data in the data source. The first step is to remove connection strings from the client. Connection strings have database addresses and passwords. There is no disguising or hiding them on the client. They belong in a safe place on the server. The EntityManager doesn‟t connect to the data source so it doesn‟t need connection strings. It tells the EntityServer which data source to use by sending a symbolic data source key. The key is just a name. The EntityServer knows how to use the key to find and connect to the data source. No process on the client side can use it. A secure application provides more fine-grained security than just whether or not the client can access the data source. It should prevent certain users from retrieving certain business objects. It should discriminate among users in determining which kinds of data source update are permitted. The screening could be at any level of detail from, say, the tables, down to a single column of a particular record.

Spoofing
In n-tier applications, whether browser-based or smart client, there is always a risk that some process pretending to be a valid client will attempt access the database in an unauthorized way. A good security design assumes that the client process -- because it cannot be physically secured -- will be compromised. While it may not be possible to fully protect the client, you can secure the host by deploying the DevForce Business Object Server (BOS) which includes the full-scale version of the EntityServer. The BOS will run special security methods whenever the client attempts to reach the server. As discussed above, the EntityServer includes ServerFetching, ServerFetched, ServerSaving, and ServerSaved events. You can handle the ServerFetching event to intercept data retrieval requests and the ServerSaving event to intercept save requests, in each case before-the-fact, to make sure the authenticated user has rights to do what she is requesting. These handlers run server-side, and no client can prevent the server from invoking them. Furthermore, your handlers can delegate their work to other methods that exist in libraries only deployed to the server. No hacker can examine the latter, so your application can be made safe from disassembly and spoofing. Finally, DevForce business objects can be digitally signed before transmission to the client. A rogue client cannot order the server to update the data source with an imposter entity.

N-Tier Architecture
We discussed n-tier architecture at the beginning of this chapter. “The Big Picture” topic described three data management tiers: 6. 7. 8. Data source(s) on the data tier
EntityServer(s) as the data access tier EntityManager within a client tier

You can run all three logical tiers on the client machine if you have a totally stand-alone application. This is the preferred choice for most development work because it eliminates the complexities of coordinating with other people, software, and hardware. When a data-driven application is deployed for production use, the database must reside on a central tier so that many users can share the data efficiently. If, with the database so deployed, you put both the EntityManager and

IdeaBlade DevForce

Business Object Persistence

an EntityServer in the same process running on a client PC, you have the ever popular two-tier, “client/server” model. This simplifies the exchanges between an EntityServer and the EntityManager. The two components don‟t have to communicate over a process boundary, so in a DevForce application deployed thusly, a light-weight version of the EntityServer reads and writes directly to the EntityManager cache. An EntityServer running under such circumstances cannot provide any meaningful security or monitoring services. It serves simply a data access abstraction – a job it does very well.

Three-Tier Deployment
Enterprise-grade applications will deploy the logical tiers on three separate physical tiers: a database server, an application server, and PC client machines. The application server hosts the Business Object Server (BOS) which runs multiple instances of a more muscular version of the EntityServer. This three-physical-tier deployment provides some remarkable advantages over the two-tier model. You get: Improved performance over connections slower than a local area network (e.g., the internet). The slow, heavy work takes place between the BOS and the database over a fat, fast pipe. Communications and data passing between the client and the middle tier are concise, compact, and highly optimized. Application Reach -- Because the application can be on-line wherever there is an Internet connection and without resort to VPN, it can be deployed and used by larger numbers and with reduced system requirements. Whereas SQL commands and result sets – the raw data exchanged between a database and a client-side access layer – cannot flow over web protocols, DevForce‟s business objects can. Security is much tighter. We covered earlier the many layers of security available with the BOS in place. Scalability. It is impractical to maintain live connections for each client when the number of simultaneous users becomes large. The tipping point appears to be around one hundred. An EntityServer running on a central server can pool connections to the data sources and serve many clients simultaneously. The server is stateless – there is no need for session awareness – so fail-over and load balancing are easy options. The BOS monitoring console provides detailed data and global insight into the use (and abuse) of the application.

Model Choice by Configuration
One-tier? Two-tier? Three-tier? You don‟t have to make the choice right away. You write our code pretty much the same way no matter what the model. In general, you don‟t have to think about which code is executing where, or by what route our business objects arrived in cache. For the most part, you write code as if every aspect of the application takes place inside your development PC. When you are ready to deploy to a multi-tiered environment, you set a few values in the XML application configuration file (App.Config) and build some set-up projects.

Conclusion
We just took a high-level view of the persistence management landscape. Some of the key points were:     The EntityManager is perhaps the most important component in the DevForce framework. It is the client application‟s gateway to the remote data. The EntityManager holds and manages an entity cache of business object instances and makes them available to the application UI. All entities within a cache are unique; no two entities can have the same primary key. You can fetch entities into the cache from remote data sources using entity queries and entity navigation.

IdeaBlade DevForce
    

Business Object Persistence

Entity navigation returns a collection whose contents are managed dynamically by a EntityManager. You can create, modify, delete, remove, and save cached entities. These actions raise “Life Cycle” events to which you may subscribe. Entities in a cache can come from many different data sources. Each data source is identified by its data source key. Each entity belongs to just one data source. A smart-client application can run off-line. An EntityServer handles the data access and object map translation chores for each of the application data sources. It exchanges business objects with one or more EntityManager instances on individual client machines. A Business Object Server (BOS) running on a central host provides enterprise-grade security, scalability, data integrity, performance, and application monitoring.

The following sections and chapters delve deeper into the features introduced here.

Entity Queries and Entity Navigation
Entity queries and entity navigation are the two mechanisms for retrieving business objects from a data source. Both deposit business objects into the EntityManager‟s cache. You use entity queries to get started in a work flow. In response to a question like “What orders were placed last month?”, they return Order entities. If your query asks “Which employees were hired last year?”, you get Employee entities. The results of entity queries are root objects. Once you have a root object, your subsequent queries are often about entities related to the root object. Given employee “Sally”, you start exploring her object graph by looking for her address, her manager, her orders, etc. You traverse Sally‟s object graph using entity navigation and it has its own simple and intuitive syntax. Most applications require surprisingly few entity queries. Once you have a list of orders or employees that interest you, you‟re likely to settle in and poke around using entity navigation. It is common for applications to show 10 or 20 times as many entity navigations as entity queries. Since we can‟t navigate anywhere until we have some root entities in hand, let‟s start with entity queries.

Entity Queries
Use an EntityQuery when you want to retrieve a set of business objects that satisfy selection criteria - the set of employees who were hired last year, for example. Entity queries come in many flavors. Some of them are linguistically independent of any particular data source; some are specialized to a particular data source. Some can query the data source and the entity cache at the same time; some can only query the data source23. EntityQueries, like .NET ObjectQueries, are enumerable, and so can be executed in a variety of stepwise ways. Consider, for example, the following query:
Code Snippet 1. BasicQuerySyntaxQuery

C#

var customersQuery = from cust in _Em1.Customers where cust.ContactTitle == "Sales Representative" orderby cust.CompanyName

23

This means this kind of query can be used only when the application is connected to the data source; such queries can‟t run when the application is off-line.

IdeaBlade DevForce
select cust;

Business Object Persistence

VB

Dim customersQuery = From cust In _em1.Customers _ Where cust.ContactTitle = "Sales Representative" _ Order By cust.CompanyName _ Select cust

This can also be written in method-based syntax24 as
Code Snippet 2. BasicMethodSyntaxQuery

C#

var customersQuery = _Em1.Customers .Where(c => c.ContactTitle == "Sales Representative") .OrderBy(c => c.CompanyName) .Select(c => c);

VB

Dim customersQuery = _em1.Customers _ .Where(Function(c) c.ContactTitle = "Sales Representative") _ .OrderBy(Function(c) c.CompanyName) _ .Select(Function(c) c)

Or just,
Code Snippet 3. MethodSyntaxShortForm

C#

var customersQuery = _Em1.Customers .Where(c => c.ContactTitle == "Sales Representative") .OrderBy(c => c.CompanyName);

VB

Dim customersQuery = _em1.Customers _ .Where(Function(c) c.ContactTitle = "Sales Representative") _ .OrderBy(Function(c) c.CompanyName)

Each of these returns an IdeaBlade.EntityQuery.EntityQuery<Customer>. If you choose to type your variable to hold the query‟s return value explicitly as a DevForce EntityQuery<T>, the statement becomes the following:
Code Snippet 4. QueryWithExplicitlyTypedReturnValue

C#
24

IEntityQuery<Customer> customersQuery = _em1.Customers

Query-based syntax looks a great deal like SQL and is, for that reason, attractive to many developers, especially those new to LINQ. At IdeaBlade we tend to prefer the more regularly structured and comprehensive method-based syntax for most queries, so you will see most of our sample queries in that format. Be assured, however, that you may write your LINQ queries in the syntax you prefer! We discuss the two syntaxes more in the section “Query v. Method Syntax”, in this document.

IdeaBlade DevForce
.Where(c => c.ContactTitle == "Sales Representative") .OrderBy(c => c.CompanyName);

Business Object Persistence

VB

Dim customersQuery As IEntityQuery(Of Customer) = _ _em1.Customers _ .Where(Function(c) c.ContactTitle = "Sales Representative") _ .OrderBy(Function(c) c.CompanyName)

The following query retrieves only a single Customer entity is retrieved from the data source into the local cache. If no Customer matches the stated criterion, DevForce returns the Null Entity Customer:
Code Snippet 5. RetrieveFirstCustomer

C#

Customer firstCustomer = _em1.Customers .Where(c => c.ContactTitle == "Sales Representative") .OrderBy(c => c.CompanyName) .FirstOrNullEntity();

VB

Dim firstCustomer As Customer = _em1.Customers _ .Where(Function(c) c.ContactTitle = "Sales Representative") _ .OrderBy(Function(c) c.CompanyName) _ .FirstOrNullEntity()

The addition of a call to extension method ToList() forces DevForce to execute the query immediately:
Code Snippet 6. ForceImmediateExecution

C#

ICollection<Customer> customersQuery = _em1.Customers .Where(c => c.ContactTitle == "Sales Representative") .OrderBy(c => c.CompanyName) .ToList();

VB

Private Sub ForceImmediateExecution() Dim customersQuery As ICollection(Of Customer) = _ _em1.Customers _ .Where(Function(c) c.ContactTitle = "Sales Representative") _ .OrderBy(Function(c) c.CompanyName) _ .ToList()

The call to ToList(), because it demands a complete set of pointers to the retrieved matching customers, forces the complete query to be executed. Below is a DevForce DebugLog listing for a test that first issued a First() call like the one we just considered, then a call to ToList(). We‟ve removed some of the columns included in the log so the table won‟t be quite so wide, but note the highlighted “Fetch … value” messages. The first one, when delivered to the EntityFramework, will be translated into a SQL query that returns a single record; the second will be translated into a SQL query that returns all of the matching customers.

IdeaBlade DevForce

Business Object Persistence

If you want to see the SQL generated by the EntityFramework to process your query, find the appropriate edmKey in your App.Config file and add a logTraceString attribute set to “true”:

This will result in output like the following. (Again, some columns were omitted to reduce the table width for inclusion here.) Note the generated SQL statements:

IdeaBlade DevForce

Business Object Persistence

In between the two extremes of asking a query object for its first element and asking it to dump its contents ToList() are many possibilities, such as using it in a foreach loop:

C#

IEntityQuery<Customer> customersQuery = _em1.Customers .Where(c => c.ContactTitle == "Sales Representative") .OrderBy(c => c.CompanyName); foreach (Customer aCustomer in customersQuery) { // All customers are retrieved at the start of the loop }

VB

Dim customersQuery As IEntityQuery(Of Customer) = _em1.Customers.Where(Function(c) c.ContactTitle = "Sales Representative").OrderBy(Function(c) c.CompanyName) For Each aCustomer As Customer In customersQuery ' All customers are retrieved at the start of the loop Next aCustomer

Code Snippet 7. ForceRetrievalUsingForEach

The foreach loop returns references to the retrieved Customers one at a time, but it does so from a collection of those references which must be obtained up front. Thus, as soon as the first iteration of the loop is executed, the

IdeaBlade DevForce

Business Object Persistence

entire set of Customers is retrieved to the local cache, and a collection of references to them is assembled. The debug log will show only a single query:

On the other hand, the following query results in exactly five (5) entities being retrieved from the data source:
Code Snippet 8. QueryWithSkipAndTake

C#

IEntityQuery<Customer> customersQuery = _em1.Customers .Where(c => c.ContactTitle == "Sales Representative") .OrderBy(c => c.CompanyName) .With(QueryStrategy.DataSourceOnly); ICollection<Customer> customers = customersQuery.Skip(5).Take(5).ToList();

VB

Dim customersQuery As IEntityQuery(Of Customer) = _em1.Customers _ .Where(Function(c) c.ContactTitle = "Sales Representative") _ .OrderBy(Function(c) c.CompanyName) _ .With(QueryStrategy.DataSourceOnly) Dim customers As ICollection(Of Customer) = customersQuery.Skip(5).Take(5).ToList()

Note the use of the DataSourceOnly QueryStrategy. That‟s often important when using Skip(). You can learn why in the section of this chapter on FetchStrategies.

The With() Extension Method
DevForce provides an extension method, With(), that permits you to substitute a different QueryStrategy, a different target EntityManager, or both, on an existing query. The original query will be left unaltered. When a call to With() is chained to a query, the result may be either a new query or a reference to the original query. Normally it will be a new query, but if the content of the With() call is such that the resultant query would be the same as the original one, a reference to the original query is returned instead of a new query. If you ever want to be sure that you get a new query, use the Clone() extension method instead of With(). With() avoids the overhead of a Clone() when a copy is unnecessary. You can pass null arguments to With(). When a query has a null EntityManager assigned, it uses the DefaultManager. When a query has a null QueryStrategy, it uses the DefaultQueryStrategy of the assigned (or default) EntityManager. See the code below for more detail on the possibilities.
Code Snippet 9. QueriesWithWITH

IdeaBlade DevForce
C#
IEntityQuery<Customer> query0 = _em1.Customers .Where(c => c.CompanyName.ToLower().StartsWith("a")); query0.QueryStrategy = QueryStrategy.DataSourceOnly;

Business Object Persistence

// Use With() to run the existing query against a different EntityManager: DomainModelEntityManager em2 = new DomainModelEntityManager(); List<Customer> customers = new List<Customer>(query0.With(em2)); // The next two examples use With() to run the query with a different QueryStrategy. // The With() call in the right-hand side of the following statement // specifies a query that is materially different from query0, in // that it has a different QueryStrategy associated with it. // Accordingly, the right-hand side of the statement will return // a new query: IEntityQuery<Customer> query1 = query0.With(QueryStrategy.CacheOnly); // Because the content of the With() call in the right-hand side // of the following statement doesn't result in a modification // of query0, the right-hand side will return a reference to // query0 rather than a new query. IEntityQuery<Customer> query2 = query0.With(QueryStrategy.DataSourceOnly); // If you want to be certain you get a new query, use Clone() // rather than With(): EntityQuery<Customer> query3 = (EntityQuery<Customer>)query0.Clone(); query3.QueryStrategy = QueryStrategy.DataSourceOnly; // Change both the QueryStrategy and the EntityManager IEntityQuery<Customer> query4 = query0.With(em2, QueryStrategy.CacheOnly); // You can pass null arguments to With(). When a query has a null EntityManager, // assigned, it uses the DefaultManager. When a query has a null QueryStrategy, // it uses the DefaultQueryStrategy of the assigned (or default) EntityManager. // Run the query against the default EntityManager, using its default QueryStrategy: IEntityQuery<Customer> query5 = query0.With(null, null); // When you pass a single null to With, you must cast it to the appropriate // type so the compiler know's which single-parameter overload you mean to use: // Run the query against the default EntityManager, using the base query's // assigned QueryStrategy: IEntityQuery<Customer> query6 = query0.With((DomainModelEntityManager)null); // Run the query against the assigned EntityManager, using that EntityManager's // default QueryStrategy: IEntityQuery<Customer> query7 = query0.With((QueryStrategy)null);

IdeaBlade DevForce
VB

Business Object Persistence

Dim query0 As IEntityQuery(Of Customer) = _em1.Customers.Where(Function(c) c.CompanyName.ToLower().StartsWith("a")) query0.QueryStrategy = QueryStrategy.DataSourceOnly ' Use With() to run the existing query against a different EntityManager: Dim em2 As New DomainModelEntityManager() Dim customers As New List(Of Customer)(query0.With(em2)) ' The next two examples use With() to run the query with a different QueryStrategy. ' The With() call in the right-hand side of the following statement ' specifies a query that is materially different from query0, in ' that it has a different QueryStrategy associated with it. ' Accordingly, the right-hand side of the statement will return ' a new query: Dim query1 As IEntityQuery(Of Customer) = query0.With(QueryStrategy.CacheOnly) ' Because the content of the With() call in the right-hand side ' of the following statement doesn't result in a modification ' of query0, the right-hand side will return a reference to ' query0 rather than a new query. Dim query2 As IEntityQuery(Of Customer) = query0.With(QueryStrategy.DataSourceOnly) ' If you want to be certain you get a new query, use Clone() ' rather than With(): Dim query3 As EntityQuery(Of Customer) = CType(query0.Clone(), EntityQuery(Of Customer)) query3.QueryStrategy = QueryStrategy.DataSourceOnly ' Change both the QueryStrategy and the EntityManager Dim query4 As IEntityQuery(Of Customer) = query0.With(em2, QueryStrategy.CacheOnly) ' You can pass null arguments to With(). When a query has a null EntityManager, ' assigned, it uses the DefaultManager. When a query has a null QueryStrategy, ' it uses the DefaultQueryStrategy of the assigned (or default) EntityManager. ' Run the query against the default EntityManager, using its default QueryStrategy: Dim query5 As IEntityQuery(Of Customer) = query0.With(Nothing, Nothing) ' When you pass a single null to With, you must cast it to the appropriate ' type so the compiler know's which single-parameter overload you mean to use: ' Run the query against the default EntityManager, using the base query's ' assigned QueryStrategy: Dim query6 As IEntityQuery(Of Customer) = query0.With(CType(Nothing, DomainModelEntityManager)) ' Run the query against the assigned EntityManager, using that EntityManager's ' default QueryStrategy: Dim query7 As IEntityQuery(Of Customer) = query0.With(CType(Nothing, QueryStrategy))

The FirstOrNullEntity() ExtensionMethod
LINQ to Entities provides First() and FirstOrDefault() extension methods on queries. First() returns the first item in a collection meeting the query criteria; FirstOrDefault() returns that, or if no items meet the criteria, the default value for the target type. For integer target types, FirstOrDefault() returns a zero; for string types, it returns an empty string. For complex types or other types that have no default, it returns a null. DevForce adds a FirstOrNullEntity() extension method that can be used when you are querying for target types that inherit from IdeaBlade.EntityModel.Entity. If no entity meets the specified criteria, FirstOrNullEntity() returns the

IdeaBlade DevForce

Business Object Persistence

DevForce NullEntity for the target type. The NullEntity is a non-saveable, immutable, syntactically correct instance of an entity represents “nothing there” but will not trigger an exception.

The ToQuery () ExtensionMethod
Every IdeaBlade.EntityModel.Entity has a ToQuery() extension method that returns an IEntityQuery<T> where T is an Entity type. This IEntityQuery<T> specifies the Entity on which it was based using its EntityAspect.EntityKey, and can be extended to perform various useful operations. Consider, for example, the following statements:
Code Snippet 10. UsingToQueryPt01

C#

Customer aCustomer = _em1.Customers.FirstOrNullEntity(); var query = aCustomer.ToQuery<Customer>() .Include(Customer.PathFor(c => c.Orders)); query.With(QueryStrategy.DataSourceOnly).ToList();

VB

Dim aCustomer As Customer = _em1.Customers.FirstOrNullEntity() Dim query = aCustomer.ToQuery().Include(Customer.PathFor(Function(c) c.Orders)) query.With(QueryStrategy.DataSourceOnly).ToList()

Here, from a Customer entity, we have created a query that will retrieve that same Customer. We have then extended with a call to Include() it to create a span query that will also retrieve all of that Customer‟s associated Orders. We do not otherwise have so convenient a way to accomplish this goal. The ToQuery() extension method is also provided on any IEnumerable<T> collection, when T is an Entity. Thus you can turn an arbitrary list of Customers into a query that will return the same set of Customers. The Where()

IdeaBlade DevForce

Business Object Persistence

clause on the resultant query will specify a series of OR‟d key values. For example, consider the following statements:
Code Snippet 11. UsingToQueryPt02

C#

List<Customer> customers = _em1.Customers .Where(c => c.CompanyName.ToLower().StartsWith("a")).ToList(); var query2 = customers.ToQuery<Customer>();

VB

Dim customers As List(Of Customer) = _em1.Customers _ .Where(Function(c) c.CompanyName.ToLower().StartsWith("a")).ToList() Dim query2 = customers.ToQuery()

Placing query2 in a watch window reports its value as the following: {value(IdeaBlade.EntityModel.EntityQueryProxy`1[DomainModel.Customer]).Where(t => ((((t.CustomerID = 785efa04-cbf2-4dd7-a7de-083ee17b6ad2) || (t.CustomerID = b61cf396-206f-41a6-9766-168b5cbb8edd)) || (t.CustomerID = f214f516-d55d-4f98-a56d7ed65fd79520)) || (t.CustomerID = 256d4372-baa7-4937-9d87-d9a4e06146f8)))} The first query evidently placed four Customers in the customers list; the query returned by ToQuery() specifies those four by their (GUID) key values.

Other Query Types
In addition to the EntityQuery, DevForce provides the PassthruESQLQuery and StoredProcQuery types for querying using Entity SQL and stored procedures, respectively. Like the EntityQuery, these types implement DevForce‟s IEntityQuery interface.

Code Snippet 12. PassThruEsqlQuery

C#

PassthruEsqlQuery query = new PassthruEsqlQuery(typeof(Employee), "SELECT VALUE e FROM Employees AS e Where e.EmployeeID < 10"); IEnumerable results = _em1.ExecuteQuery(query);

VB

Dim query As New PassthruEsqlQuery(GetType(Employee), _ "SELECT VALUE e FROM Employees AS e Where e.EmployeeID < 10") Dim results As IEnumerable = _em1.ExecuteQuery(query)

Code Snippet 13. StoredProcQuery

C#

QueryParameter param01 = new QueryParameter("EmployeeID",1); QueryParameter param02 = new QueryParameter("Year",1996); StoredProcQuery query = new StoredProcQuery(typeof(Order)); query.Parameters.Add(param01); query.Parameters.Add(param02); // Note that a FunctionImport must be defined in the Entity Model query.ProcedureName = "OrdersGetForEmployeeAndYear"; _em1.ExecuteQuery(query);

IdeaBlade DevForce
VB
Dim param01 As New QueryParameter("EmployeeID", 1) Dim param02 As New QueryParameter("Year", 1996) Dim query As New StoredProcQuery(GetType(Order)) query.Parameters.Add(param01) query.Parameters.Add(param02)

Business Object Persistence

' Note that a FunctionImport must be defined in the Entity Model query.ProcedureName = "OrdersGetForEmployeeAndYear" _em1.ExecuteQuery(query)

The Query Object return type
An entity query returns one and only one kind of thing. That kind of thing is always an entity type declared in the business object model. The query developer must identify that entity type and ensure that the substance of the query actually will return such entities. Although the query returns only one kind of entity, it may populate the entity cache with other kinds of entities. You‟ll see just how useful this can be when we discuss span queries and query inversion.

The Fetch and Merge
The EntityManager evaluates the query and searches for suitable entities either in the cache, in the data source, or in both. Where it looks for entities and what it does with the ones it finds are determined by a QueryStrategy object which we will cover in the “Caching” topic below.

Query v. Method Syntax
The following LINQ query is written in the syntax known as “query syntax”, “query comprehension syntax”, or just “comprehension syntax”:
Code Snippet 14. BasicQuerySyntaxQuery (Repeated)

C#

var customersQuery = from cust in _Em1.Customers where cust.ContactTitle == "Sales Representative" orderby cust.CompanyName select cust;

VB

Dim customersQuery = From cust In _em1.Customers _ Where cust.ContactTitle = "Sales Representative" _ Order By cust.CompanyName _ Select cust

This can also be written in method-based syntax as
Code Snippet 15. BasicMethodSyntaxQuery (Repeated)

C#

var customersQuery = _Em1.Customers .Where(c => c.ContactTitle == "Sales Representative") .OrderBy(c => c.CompanyName) .Select(c => c);

IdeaBlade DevForce
VB

Business Object Persistence

Dim customersQuery = _em1.Customers _ .Where(Function(c) c.ContactTitle = "Sales Representative") _ .OrderBy(Function(c) c.CompanyName) _ .Select(Function(c) c)

At IdeaBlade we mostly prefer the method-based syntax as a general rule. The capabilities available in methodbased syntax are substantially a superset of those available in query syntax, so when using query syntax you may be forced into concatenating method-based clauses anyway to get what you want, as in the following:
Code Snippet 16. MixedQueryAndMethodSyntax

C#

ICollection<Customer> customers = (from cust in _em1.Customers orderby cust.CompanyName select cust) .ToList();

VB

Dim customers As ICollection(Of Customer) = _ (From cust In _em1.Customers _ Order By cust.CompanyName _ Select cust) .ToList()

Having said that, there are a few things that are arguably a bit easier or more natural to do in query syntax 25, and of course there are simply personal preferences. So use what you like!

LINQ
The typical data-oriented approach to retrieving objects relies upon a specialized query language such as SQL. SQL is a powerful query language requiring considerable sophistication and experience to use properly. But there are pitfalls to using SQL and several good reasons to prefer LINQ to SQL queries, including:     Object orientation Compile time checking Query portability Query manipulation

LINQ is a vast subject and is, for the most part, beyond the scope of this document. A web search on “LINQ” will provide you with an abundance of excellent resources for learning about LINQ. It suffices to say here that our implementation of LINQ -- LINQ to DevForce -- permits the same query to be used against a local cache or a back-end datasource supported by Microsoft‟s LINQ to Entities. You can specify, by means of a QueryStrategy property on the query object, just what you want its target data store or data stores to be; or you can let DevForce apply sensible defaults which work well for the majority of cases.

25

Joseph and Ben Albahari, in a fine discussion of LINQ, opine that query comprehension syntax “is much simpler for queries that involve any of the following:   A let clause for introducing a new variable alongside the iteration variable SelectMany, Join, or GroupJoin, followed by an outer iteration variable reference”

See their excellent book C#3.0 In a Nutshell, O‟Reilly Media Inc., 2007, p.285

IdeaBlade DevForce The DevForce Predicate Builder

Business Object Persistence

The time comes when you want to construct a LINQ “Where” clause programmatically. It should be easy. It turns out to be more challenging … until you use the DevForce PredicateBuilder. (You will find this class in the IdeaBlade.Linq namespace, in either the IdeaBlade.Linq or IdeaBlade.Linq.SL [for Silverlight] assembly.) Imagine a product search interface. The user can enter words in a “Name Search” text box. Your program should find and display every product that contains any of the words entered by the user. You don‟t know how many words the user might enter. What do you do? The solution would be easy if you knew the user would enter exactly one word.
Code Snippet 17. ProductsWithNamesThatContainSpecifiedString

C#

var word = "Sir"; var q = _em1.Products .Where(p => p.ProductName.Contains(word)); var results = q.ToList();// returns 3 Northwind products

VB

Dim word = "Sir" Dim q = _em1.Products _ .Where(Function(p) p.ProductName.Contains(word)) Dim results = q.ToList() ' returns 3 Northwind products

Of course you don‟t know how many words the user will enter. You want to be prepared for more than one so you write this too-simple helper method that returns an array of words from the text entered in the text box:
Code Snippet 18. GetWords

C#

private IEnumerable<String> GetWords(string phrase) { return phrase.Split(new[] {' '}, StringSplitOptions.RemoveEmptyEntries); }

VB

Private Function GetWords(ByVal phrase As String) As IEnumerable(Of String) Return phrase.Split(New Char() {" "c}, StringSplitOptions.RemoveEmptyEntries) End Function

Now all you have to do is replace the “Where” clause with a sequence of OR clauses. You‟ll want to construct it by iterating over the words. Go ahead and write it. We‟ll wait... Having trouble? I‟ll give you the user‟s input: “Sir Cajun Louisiana”. Did that help? You will probably come up with something like the following:

IdeaBlade DevForce
C#
var q = _em1.Products .Where(p => p.ProductName.Contains("Sir") || p.ProductName.Contains("Cajun") || p.ProductName.Contains("Louisiana") ); var results = q.ToList(); // returns 6 Northwind products

Business Object Persistence

VB

Dim q = _em1.Products.Where(Function(p) _ p.ProductName.Contains("Sir") _ OrElse p.ProductName.Contains("Cajun") _ OrElse p.ProductName.Contains("Louisiana")) Dim results = q.ToList() ' returns 6 Northwind products

Code Snippet 19. ProductsWithNamesThatContainSpecifiedStrings

This is ultimately what the lambda expression must look like. Of course you cannot demand that the user enter exactly three words any more than you can insist she enter exactly one. You want to construct the lambda dynamically based on the actual number of words entered. Sadly, there is no obvious way of constructing a lambda expression dynamically. But the DevForce PredicateBuilder can help you build predicates dynamically.

What‟s a “predicate”?
A “predicate” is a function that evaluates an expression and returns true or false. The code fragment... C# /VB p.ProductName.Contains(“Sir”)

...is a predicate that examines a product and returns true if the product‟s ProductName contains the “Sir” string. The CLR type of the predicate in our example is: C# VB
Func<Product, bool>

Func(Of Product, Boolean)

Which we can generalize to: C# VB
Func<T, bool>

Func(Of T, Boolean)

IdeaBlade DevForce

Business Object Persistence

We almost have what we want. When the compiler sees an example of this kind of thing, it immediately resolves it into an anonymous delegate. But we don‟t want the delegate. We need a representation that retains our intent and postpones the resolution into a delegate until the last possible moment; because before we get that delegate, we may want to build a more complex expression. What we need is an expression made up of Func<T, bool>‟s: C# VB
Expression<Func<T, bool>>

Expression(Of Func(Of T, Boolean))

As it so happens, this is exactly what the DevForce “Where” extension method demands:

C#

public static IEntityQuery<T> Where<TSource>( this IEntityQuery<T> source1, Expression<Func<T,bool>> predicate)

VB

public static IEntityQuery(Of T) Where(Of TSource) _ (Me IEntityQuery(Of T) source1, Expression(Of Func(Of T,Boolean)) predicate)

The methods of the static IdeaBlade.Linq.PredicateBuilder class take things even a step farther: they permit us to combine two or more Predicate Expressions into a single Predicate Expression that we can pass to that Where() method. Let‟s stick with the example and see one of those PredicateBuilder methods in action. Let‟s first write a little method to produce an IEnumerable of Predicate Expressions, one expression for each string in a collection of strings:
Code Snippet 20. ProductNameTests

C#

private IEnumerable<Expression<Func<Product, bool>>> ProductNameTests(IEnumerable<String> words) { foreach (var each in words) { var word = each; yield return p => p.ProductName.Contains(word); } }

VB

Private Function ProductNameTests(ByVal words As IEnumerable(Of String)) _ As IEnumerable(Of Expression(Of Func(Of Product, Boolean))) Dim expressions As New List(Of Expression(Of Func(Of Product, Boolean)))() For Each [each] In words Dim word = [each] ' include this statement so *each* is evaluated at each iteration expressions.Add(Function(p) p.ProductName.Contains(word)) Next [each] Return expressions End Function

IdeaBlade DevForce

Business Object Persistence

The result is an IEnumerable of Predicate Expressions about the Product entity. The body is an iterator that returns a Predicate Expression for each word. That expression is exactly the same as the first predicate we wrote when we knew only one word. If we give it the three-word input in our example, we‟ll get an IEnumerable of three Predicate Expressions, each looking for one of the words in the product‟s ProductName. We‟re want to OR these Predicate Expressions together so we will use a static method of PredicateBuilder named, well, Or():

C#

public static Expression<Func<T, bool>> Or<T>( params Expression<Func<T, bool>>[] expressions)

VB

public static Expression(Of Func(Of T, Boolean)) Or(Of T) _ (params Expression(Of Func(Of T, Boolean))() expressions)

You see it takes an array (a params array to be precise) of Predicate Expressions. We will convert the output of our ProductNameTests into an array before giving it to this PredicateBuilder method. The final code looks like this:
Code Snippet 21. PredicateBuilder01

C#

var words = GetWords("Sir Cajun Louisiana"); var tests = ProductNameTests(words).ToArray(); if (0 == tests.Length) return; var productNamePredicate = PredicateBuilder.Or(tests); var q = _em1.Products.Where(productNamePredicate); var results = q.ToList(); // returns 6 Northwind products

VB

Dim words = GetWords("Sir Cajun Louisiana") Dim tests = ProductNameTests(words).ToArray() If 0 = tests.Length Then Return End If Dim productNamePredicate = PredicateBuilder.Or(tests) Dim q = _em1.Products.Where(productNamePredicate) Dim results = q.ToList() ' returns 6 Northwind products

To summarize the steps we‟re taking: 1. Split the user‟s search text into separate words 2. Generate an array of Predicate Expressions that look for each word in the ProductName 3. Skip the query if there are no clauses … because there are no words 4. Ask “PredicateBuilder.Or” to combine the tests into a single Predicate Expression 5. Run it to get results.

PredicateBuilder Methods
There are seven methods of interest:

IdeaBlade DevForce

Business Object Persistence

Method Or Or And And True False Not

Syntax by example p1.Or(p2) PredicateBuilder.Or(p1, p2, p3 .. pn) p1.And(p2) PredicateBuilder.And(p1, p2, p3 .. pn) PredicateBuilder.True() PredicateBuilder.False() PredicateBuilder.Not(p1)

“p” = Predicate Expression, Expression<Func<T, bool>>. All expressions must be of the same type (e.g., Product).

Examples
Here are some examples using the PredicateBuilder methods:
Code Snippet 22. PredicateBuilderMiscExamples

C#

Expression<Func<Product, bool>> p1, p2, p3, p4, bigP; // p1 p2 p3 p4 Sample = p => = p => = p => = p => predicate expressions p.ProductName.Contains("Sir"); p.ProductName.Contains("Cajun"); p.ProductName.Contains("Louisiana"); p.UnitPrice > 20;

bigP = p1.Or(p2); // Name contains "Sir" or "Cajun" bigP = p1.Or(p2).Or(p3); // Name contains any of the three bigP = PredicateBuilder.Or(p1, p2, p3); // Name contains any of the 3 bigP = PredicateBuilder.Or(tests); // OR together some tests bigP = p1.And(p4); // "Sir" and price is greater than 20 // Name contains "Cajun" and "Lousiana" and the price is greater than 20 bigP = PredicateBuilder.And(p2, p3, p4); bigP = PredicateBuilder.And(tests); // AND together some tests // Name contains either “Sir” or “Louisiana” AND price is greater than 20 bigP = p1.Or(p3).And(p4); // bigP = PredicateBuilder.Not(p1); // Name does not contain "Sir" bigP = PredicateBuilder.True<Product>().And(p1);// same as p1 bigP = PredicateBuilder.False<Product>().Or(p1);// same as p1

IdeaBlade DevForce

Business Object Persistence

// Not useful bigP = PredicateBuilder.True<Product>().Or(p1);// always true bigP = PredicateBuilder.False<Product>().And(p1);// always false

VB

Dim p1 As Expression(Of Func(Of Product, Boolean)), p2 As Expression(Of Func(Of Product, Boolean)), p3 As Expression(Of Func(Of Product, Boolean)), p4 As Expression(Of Func(Of Product, Boolean)), bigP As Expression(Of Func(Of Product, Boolean)) ' Sample predicate expressions p1 = Function(p) p.ProductName.Contains("Sir") p2 = Function(p) p.ProductName.Contains("Cajun") p3 = Function(p) p.ProductName.Contains("Louisiana") p4 = Function(p) p.UnitPrice > 20 bigP = p1.Or(p2) ' Name contains "Sir" or "Cajun" bigP = p1.Or(p2).Or(p3) ' Name contains any of the three bigP = PredicateBuilder.Or(p1, p2, p3) ' Name contains any of the 3 bigP = PredicateBuilder.Or(tests) ' OR together some tests bigP = p1.And(p4) ' "Sir" and price is greater than 20 ' Name contains "Cajun" and "Lousiana" and the price is greater than 20 bigP = PredicateBuilder.And(p2, p3, p4) bigP = PredicateBuilder.And(tests) ' AND together some tests ' Name contains either "Sir" or "Louisiana" AND price is greater than 20 bigP = p1.Or(p3).And(p4) bigP = PredicateBuilder.Not(p1) ' Name does not contain "Sir" bigP = PredicateBuilder.True(Of Product)().And(p1) ' same as p1 bigP = PredicateBuilder.False(Of Product)().Or(p1) ' same as p1 ' Not useful bigP = PredicateBuilder.True(Of Product)().Or(p1) ' always true bigP = PredicateBuilder.False(Of Product)().And(p1) ' always false

Observations Regarding the PredicateBuilder Methods
Notice that one each of the Or(), And(), and Not() methods are Predicate Expression extension methods; they make it easier to compose predicates from a number of Predicate Expressions known at design time. Put a breakpoint on any of the “bigP” lines and ask the debugger to show you the result as a string. Here is the Immediate Window output for “bigP = p1.Or(p3).And(p4);”:
{p => ((p.ProductName.Contains("Sir") || p.ProductName.Contains("Louisiana")) && (p.UnitPrice > Convert(20)))}

The True() and False() methods return Predicate Expression constants that simply help you jumpstart your chaining of PredicateBuilder expressions. Two of the combinations – True()…Or() and False()…And() -- are not useful.

IdeaBlade DevForce

Business Object Persistence

Example: Simulate an In() Clause Condition on a Distantly Related Entity
Consider the following query:
Code Snippet 23. PredicateBuilderForInClause()

C#

var employeeTerritoriesQuery = _em1.EmployeeTerritories .Where(et => et.Employee.Orders.Any(o => o.Customer.City == "Albuquerque" || o.Customer.City == "Frankfurt" || o.Customer.City == "London" || o.Customer.City == "Rio de Janeiro" || o.Customer.City == "Sao Paulo"));

VB

Dim employeeTerritoriesQuery = _em1.EmployeeTerritories _ .Where(Function(et) et.Employee.Orders _ .Any(Function(o) o.Customer.City = "Albuquerque" OrElse _ o.Customer.City = "Frankfurt" OrElse _ o.Customer.City = "London" OrElse _ o.Customer.City = "Rio de Janeiro" OrElse _ o.Customer.City = "Sao Paulo"))

We have, in essence, placed an In() condition on the Customer for any Order associated with the Employee that is associated with the EmployeeTerritory entities we want to retrieve. Of course, In() isn‟t support by the version of LINQ in .NET 3.5, so we had to code it the hard way. Still, it works, so we‟re happy until we realize that we need to use such a query in a situation where we don‟t know until runtime what cities – or how many cities – our end user will want to match. We need to let that user pick the cities from a list, or even type their names in freeform. For this, we‟ll need the PredicateBuilder, as shown in the version of the query below. This version uses a string array of city names as input to the query. We stuff that array in a code statement here, but it could, of course, be populated by user input in the user interface.
Code Snippet 24. PredicateBuilderForInClause()

C#

… string[] targetCities = _ { "Albuquerque", "Frankfurt", "London", "Rio de Janeiro", "Sao Paulo" }; IEnumerable<Expression<Func<EmployeeTerritory, bool>>> tests = CustomerCityNameTests(targetCities.ToArray()); var cityNamePredicate = PredicateBuilder.Or(tests.ToArray()); IEntityQuery<EmployeeTerritory> search3 = _em1.EmployeeTerritories.Where(cityNamePredicate); search3.ToList(); … private IEnumerable<Expression<Func<EmployeeTerritory, bool>>> CustomerCityNameTests(IEnumerable<String> words) { foreach (var each in words) { var word = each; // must include this statement so *each* is evaluated at each iteration yield return et => et.Employee.Orders.Any(o => o.Customer.City == word); } }

IdeaBlade DevForce
VB

Business Object Persistence

… Dim targetCities() As String = _ {"Albuquerque", "Frankfurt", "London", "Rio de Janeiro", "Sao Paulo"} Dim tests As IEnumerable(Of Expression(Of Func(Of EmployeeTerritory, Boolean))) = _ CustomerCityNameTests(targetCities.ToArray()) Dim cityNamePredicate = PredicateBuilder.Or(tests.ToArray()) Dim search3 As IEntityQuery(Of EmployeeTerritory) = _ _em1.EmployeeTerritories.Where(cityNamePredicate) search3.ToList() … Private Function CustomerCityNameTests(ByVal words As IEnumerable(Of String)) _ As IEnumerable(Of Expression(Of Func(Of EmployeeTerritory, Boolean))) Dim predicateExpressions = _ New List(Of Expression(Of Func(Of EmployeeTerritory, Boolean)))() For Each [each] In words Dim word = [each] ' must include this statement so *each* is evaluated at each iteration predicateExpressions.Add( _ Function(et) et.Employee.Orders.Any(Function(o) o.Customer.City Is word)) Next [each] Return predicateExpressions End Function

The PredicateBuilder versions retrieves exactly the same set of entities into the cache as the hard-coded version.

The PredicateDescription Class
So far, so good: but what about when you need to build a filter for a query dynamically? For example, suppose the filter criteria, including the search field and operator, are user-controlled (e.g., obtained from UI controls). With the facilities you‟ve seen so far, you don‟t have a good tool. Enter the PredicateDescription. Here are some examples:  Create two filters. The snippet below comprises two statements, each of which uses PredicateBuilder.Make(Type type, string propertyName, FilterOperator filterOp, object value) to create a PredicateDescription representing a single predicate (filter criteria):

Code Snippet 25. PredicateDescriptions01

C#

PredicateDescription p1 = PredicateBuilder.Make(typeof(Product), "UnitPrice", FilterOperator.IsGreaterThanOrEqualTo, 24); PredicateDescription p2 = PredicateBuilder.Make(typeof(Product), "Discontinued", FilterOperator.IsEqualTo, true);

VB

Dim p1 As PredicateDescription = PredicateBuilder.Make(GetType(Product), "UnitPrice", _ FilterOperator.IsGreaterThanOrEqualTo, 24) Dim p2 As PredicateDescription = PredicateBuilder.Make(GetType(Product), "Discontinued", _ FilterOperator.IsEqualTo, True)

Create a filter query, ANDing the two filters.

IdeaBlade DevForce

Business Object Persistence

This snippet uses PredicateBuilder.FilterQuery(IQueryable baseQuery, IPredicateDescription predicateDescription) to create a new, filtered query from a base query. The new query can then be executed by an EntityManager as usual:
Code Snippet 26. PredicateDescriptions01

C#

var query = PredicateBuilder.FilterQuery(_em1.Products, p1.And(p2)); var results = _em1.ExecuteQuery<Product>((IEntityQuery<Product>)query); // The above query is the same as: //var queryb = _em1.Products.Where(p => p.UnitPrice > 24 && p.Discontinued);

VB

Dim query = Dim results ' The above 'var queryb

PredicateBuilder.FilterQuery(_em1.Products, p1.And(p2)) = _em1.ExecuteQuery(Of Product)(CType(query, IEntityQuery(Of Product))) query is the same as: = _em1.Products.Where(p => p.UnitPrice > 24 && p.Discontinued);

Now let‟s accomplish the same thing in a slightly different manner. Multiple predicates can be And‟ed and Or‟ed together to form a CompositePredicateDescription.  Create a composite filter from two individual filters.

C#

PredicateDescription p1 = new PredicateDescription(typeof(Product), "UnitPrice", FilterOperator.IsGreaterThanOrEqualTo, 24); PredicateDescription p2 = new PredicateDescription(typeof(Product), "Discontinued", FilterOperator.IsEqualTo, true); // And the two filters. CompositePredicateDescription p3 = p1.And(p2);

VB

Dim p1 As New PredicateDescription(GetType(Product), _ "UnitPrice", FilterOperator.IsGreaterThanOrEqualTo, 24) Dim p2 As New PredicateDescription(GetType(Product), _ "Discontinued", FilterOperator.IsEqualTo, True) ' And the two filters. Dim p3 As CompositePredicateDescription = p1.And(p2)

Code Snippet 27. CompositePredicateDescription

We can‟t use the CompositePredicateDescription directly in or on a query. Instead we, must first convert it into a Lambda expression. We can then use that in the query:  Create a lambda expression, and use that in a Where clause.

Code Snippet 28. CompositePredicateDescriptionToLambda

C#

using System.Linq.Expressions; ... var exprFunc = (Expression<Func<Product, bool>>)p3.ToLambdaExpression(); var filterQuery = _em1.Products.Where(exprFunc);

IdeaBlade DevForce
var results = _em1.ExecuteQuery(filterQuery);

Business Object Persistence

VB

Imports System.Linq.Expressions ... Private exprFunc = CType(p3.ToLambdaExpression(), Expression(Of Func(Of Product, Boolean))) Private filterQuery = _em1.Products.Where(exprFunc) Private results = _em1.ExecuteQuery(filterQuery)

A PredicateDescription can always be instantiated from its constructor, and AND‟d or OR‟d with another PredicateDescription to form a CompositePredicateDescription. The method ToLambdaExpression() can be used to turn any predicate description into an expression which can be used in a standard LINQ Where clause.

Example: Given a Collection of Parent Entities, Retrieve the Related Children
As a further example of the use of the PredicateBuilder and PredicateDescription types, let‟s consider the following scenario: you want to retrieve a set of Orders related to an arbitrary collection of Customers. In fact, you‟re going to let your end user select the Customers whose Orders she wants to see. You won‟t know until runtime. Here‟s the code:
Code Snippet 29. GetRelatedChildrenOfParentCollection

C#

// Start with a list of customers that you‟ve populated however you see fit, // perhaps from end-user input. Here, we‟ll arbitrarily populate one as follows: List<Customer> customers = _em1.Customers.Where(c => c.Country == "USA").ToList(); customers.AddRange(_em1.Customers.Where(c => c.Country == "Brazil").ToList()); // From the list of customers, create an IEnumerable<PredicateDescription> var predicates = customers.Select(c => new PredicateDescription(typeof(Customer), "CustomerId", FilterOperator.IsEqualTo, c.CustomerID)); // Convert that IEnumerable<PredicateDescription> to an array, and feed the array // to the PredicateBuilder‟s Or() method to get a CompositePredicateDescription var customerPredicate = PredicateBuilder.Or(predicates.ToArray()); // Convert the CompositePredicateDescription to a LambdaExpression; pass the // LambdaExpression to the Where clause of a Customer query; and project out // the related Orders using a SelectMany() call. Execute the query to retrieve // the desired Orders! var exprFunc = (Expression<Func<Customer, bool>>)customerPredicate.ToLambdaExpression(); var ordersQuery = _em1.Customers.Where(exprFunc).SelectMany(c => c.Orders); var orders = _em1.ExecuteQuery(ordersQuery);

VB

' Start with a list of customers that you‟ve populated however you see fit, ' perhaps from end-user input. Here, we‟ll arbitrarily populate one as follows: Dim customers As List(Of Customer) = _ _em1.Customers.Where(Function(c) c.Country = "USA").ToList() customers.AddRange(_em1.Customers.Where(Function(c) c.Country = "Brazil").ToList()) ' From the list of customers, create an IEnumerable(Of PredicateDescription) Dim predicates = customers.Select(Function(c) New PredicateDescription(GetType(Customer), "CustomerId", FilterOperator.IsEqualTo, c.CustomerID))

IdeaBlade DevForce

Business Object Persistence

' Convert that Ienumerable(Of PredicateDescription) to an array, and feed the array ' to the PredicateBuilder‟s Or() method to get a CompositePredicateDescription Dim customerPredicate = PredicateBuilder.Or(predicates.ToArray()) ' Convert the CompositePredicateDescription to a LambdaExpression; pass the ' LambdaExpression to the Where clause of a Customer query; and project out ' the related Orders using a SelectMany() call. Execute the query to retrieve ' the desired Orders! Dim exprFunc = CType(customerPredicate.ToLambdaExpression(), _ Expression(Of Func(Of Customer, Boolean))) Dim ordersQuery = _em1.Customers.Where(exprFunc).SelectMany(Function(c) c.Orders) Dim orders = _em1.ExecuteQuery(ordersQuery)

PassthruESQL Queries
DevForce supports queries in Entity SQL (ESQL) with its PassThruEsqlQuery() method.
Code Snippet 30. EsqlBasic

C#

var query = new PassthruEsqlQuery(typeof(Customer), "SELECT VALUE c FROM Customers AS c Where c.Country == 'Brazil'"); var result = query.With(_em1).Execute().Cast<Customer>();

VB

Dim query = New PassthruEsqlQuery(GetType(Customer), _ "SELECT VALUE c FROM Customers AS c Where c.Country == 'Brazil'") Dim result = query.With(_em1).Execute().Cast(Of Customer)()

As you can see, PassThruEsqlQuery() requires the Entity type to which you want references returned and the ESQL query string. Here‟s an ESQL query that takes a parameter, “bonus”, which we‟ll give a value of 2000:
Code Snippet 31. EsqlWithParameter

C#

var param = new QueryParameter("country", "Brazil"); var paramEsql = new ParameterizedEsql( "SELECT VALUE c FROM Customers AS c Where c.Country > @country", param); var query = new PassthruEsqlQuery(typeof(Customer), paramEsql); var result1 = query.With(_em1).Execute().Cast<Customer>(); Console.WriteLine("Retrieved {0} Customers from {1}", result1.Count(), param.Value.ToString()); // Retrieved 75 Customers from Brazil param.Value = "Germany"; var result2 = query.With(_em1).Execute().Cast<Customer>(); Console.WriteLine("Retrieved {0} Customers from {1}", result2.Count(), param.Value.ToString()); // Retrieved 46 Customers from Germany

VB

Dim param = New QueryParameter("country", "Brazil") Dim paramEsql = New ParameterizedEsql( _ "SELECT VALUE c FROM Customers AS c Where c.Country > @country", param) Dim query = New PassthruEsqlQuery(GetType(Customer), paramEsql) Dim result1 = query.With(_em1).Execute().Cast(Of Customer)() Console.WriteLine("Retrieved {0} Customers from {1}", result1.Count(), _ param.Value.ToString()) „Retrieved 75 Customers from Brazil param.Value = "Germany"

IdeaBlade DevForce

Business Object Persistence

Dim result2 = query.With(_em1).Execute().Cast(Of Customer)() Console.WriteLine("Retrieved {0} Customers from {1}", result2.Count(), _ param.Value.ToString()) „Retrieved 46 Customers from Germany

Note that the value of the parameter can be changed and the same query re-executed, returning different results. When you use Entity SQL, you‟re responsible for formulating a query string that constitutes a valid query. If you goof, you won‟t know until you run it. A PassthruEsqlQuery will not interrogate the local cache26. It goes directly to the Entity Data Model to which the application must be connected when the query is issued. The EntityServer will throw an exception if it cannot convert the result set into objects of the target entity‟s type. We highly recommend a try/catch around your passthru query call.

Remote Service Method Call (RSMC)
DevForce offers a Remote Service Method Call (RSMC) facility that enables a client-side caller to invoke an arbitrary static method of a class accessible to the DevForce Business Object Server (BOS). The method can return any kind of serializable object27: a list, a custom object, a list of custom objects, etc. The client calls EntityManager.InvokeServerMethod() with the appropriate arguments: typically a class name, method name, and arguments for the method. An EntityServer instance in the BOS runs a security check and (if passed) invokes the requested method. The BOS serializes the result and transmits it back to the requesting EntityManager which presents the object to the caller after deserialization. It is up to the caller to make sense of this object. There is no restriction on what the remote method does or how it does it. The object returned must be serializable and – like business objects – must be of the same type on both client and server. The RSMC mechanism ensures that remote method callers go through the same security checks as the other EntityManager query methods. An asynchronous version of the Remote Service Method Call is also provided. It‟s perfect for any timeconsuming, server-based operation whose results are not needed immediately for continued work in the client application. The asynchronously RSMC can, for example, be used to load huge and even unrelated collections of data from the backend data store to the local cache without freezing the UI. The end user continues productive work while the data is being loaded; and then subsequently enjoys extremely crisp response in all aspects of the client application that depend upon the data that was loaded, which is now available directly from the local cache.

Entity Navigation
Entity navigation is a convenient syntax for accessing data from related business objects. Consider these familiar scenarios:    Get all of a particular sales rep‟s orders. Find the employee‟s home address Calculate the sales tax for an order

26 27

We can extend some Passthru queries to search the cache. See “Advanced Business Object Concepts.” RPC is not an “entity query” facility because it is not required to return entities.

IdeaBlade DevForce

Business Object Persistence

In each instance, we want information (orders, address, sales tax table for the ship-to-address) related to a single entity (salesrep, employee, order). The desired information exists somewhere in the entity‟s business object graph – the network of other entities that are related to our primary entity. In DevForce, you can begin with an entity – arbitrarily designated the “root entity” – and traverse its relations to reach other entities, both near and far. We call this “navigating the graph.” All you do is write a simple navigation property expression such as myOrder.Customer. Observe that the navigation property syntax, myOrder.Customer, looks just like one of the entity‟s simple properties, myOrder.ShippedDate. The key difference is that it returns an entity (Customer) rather than a value (DateTime). Entities have properties so you can write myOrder.Customer.Name. They have navigation properties so you can walk further along the graph to the HeadquartersAddress entity where you‟ll find the headquarters city:
myOrder.Customer.HeadquartersAddress.City

Parent-Child Navigation properties
So far we‟ve considered only navigation properties that return a single entity. Navigation properties can return many entities. The myOrder.OrderDetails navigation property, for example, returns the many line items of a single order. Navigation properties that return multiple entities are invariable parent-child properties. The property belongs to the parent entity such as Order and it returns child entities such as OrderDetail entities. The navigation property returns child entities in a RelatedEntityList<T> collection. The Order.OrderDetails property returns its OrderDetail children in a concrete collection, RelatedEntityList<OrderDetail>.

A brief example
I am writing a program in C#. I write and run the following statements and learn that there are three line items in the collection owned by anOrder:
Code Snippet 32. NavigationBasic

C#

Order anOrder = _em1.Orders.FirstOrNullEntity(); List<OrderDetail> lineItems = new List<OrderDetail>(anOrder.OrderDetails); Console.WriteLine("lineItems.Count = {0}", lineItems.Count);

VB

Dim anOrder As Order = _em1.Orders.FirstOrNullEntity() Dim lineItems As New List(Of OrderDetail)(anOrder.OrderDetails) Console.WriteLine("lineItems.Count = {0}", lineItems.Count)

We decide to increase the quantity ordered for the first OrderDetail as follows.

C#

OrderDetail firstItem = lineItems[0]; firstItem.Quantity = 10;

IdeaBlade DevForce
VB
Dim firstItem As OrderDetail = lineItems(0) firstItem.Quantity = 10

Business Object Persistence

Navigation Properties in Silverlight
Because all data retrieval and save operations in Silverlight are required to be asynchronous, navigation properties return their results to callback methods. Consider the following code:

Code Snippet 33. NavigationBasicAsynchronous

IdeaBlade DevForce
C#

Business Object Persistence

public void NavigationBasicAsynchronous() { _em1.UseAsyncNavigation = true; IEntityQuery<Order> query = _em1.Orders.Where(o => o.OrderID == 10248); _em1.ExecuteQueryAsync<Order>(query, GotOrder, null); PromptToContinue(); } private void GotOrder(EntityFetchedEventArgs<Order> args) { if (args.Error != null) { Console.WriteLine(args.Error.Message); } else { // Retrieve a single related entity using a scalar navigation property Order targetOrder = (Order)args.Result.ToList()[0]; Console.WriteLine("Order: {0}", targetOrder.OrderID.ToString()); targetOrder.Customer.PendingEntityResolved += new EventHandler<PendingEntityResolvedEventArgs>( Customer_PendingEntityResolved); Customer aCustomer = targetOrder.Customer; Console.WriteLine("Customer (from GotOrders): {0}", aCustomer.CompanyName); // Retrieve a collection of related entities using a collection navigation property targetOrder.OrderDetails.PendingEntityListResolved += new EventHandler<PendingEntityListResolvedEventArgs<OrderDetail>>( OrderDetails_PendingEntityListResolved); } } void Customer_PendingEntityResolved(object sender, PendingEntityResolvedEventArgs e) { Customer customer = (Customer)e.ResolvedEntity; Console.WriteLine("Customer (from Customer_PendingEntityResolved): {0}", customer.CompanyName); } void OrderDetails_PendingEntityListResolved(object sender, PendingEntityListResolvedEventArgs<OrderDetail> e) { Console.WriteLine("OrderDetails retrieved: {0}", e.ResolvedEntities.Count); } private void PromptToContinue() { Console.WriteLine(); Console.WriteLine("Press ENTER to continue..."); Console.ReadLine(); }

IdeaBlade DevForce
VB

Business Object Persistence

Public Sub NavigationBasicAsynchronous() ResetEntityManager(_em1) _em1.UseAsyncNavigation = True Dim query As IEntityQuery(Of Order) = _em1.Orders.Where(Function(o) o.OrderID = 10248) _em1.ExecuteQueryAsync(Of Order)(query, AddressOf GotOrder, Nothing) PromptToContinue() End Sub Private Sub GotOrder(ByVal args As EntityFetchedEventArgs(Of Order)) If args.Error IsNot Nothing Then Console.WriteLine(args.Error.Message) Else ' Retrieve a single related entity using a scalar navigation property Dim targetOrder As Order = CType(args.Result.ToList()(0), Order) Console.WriteLine("Order: {0}", targetOrder.OrderID.ToString()) AddHandler targetOrder.Customer.PendingEntityResolved, _ AddressOf Customer_PendingEntityResolved Dim aCustomer As Customer = targetOrder.Customer Console.WriteLine("Customer (from GotOrders): {0}", aCustomer.CompanyName) ' Retrieve a collection of related entities using a collection navigation property AddHandler targetOrder.OrderDetails.PendingEntityListResolved, _ AddressOf OrderDetails_PendingEntityListResolved 'Console.WriteLine("OrderDetails retrieved: {0}", targetOrder.OrderDetails.ToList().Count) End If End Sub Private Sub Customer_PendingEntityResolved(ByVal sender As Object, ByVal e As PendingEntityResolvedEventArgs) Dim customer As Customer = CType(e.ResolvedEntity, Customer) Console.WriteLine("Customer (from Customer_PendingEntityResolved): {0}", customer.CompanyName) End Sub Private Sub OrderDetails_PendingEntityListResolved(ByVal sender As Object, ByVal e As PendingEntityListResolvedEventArgs(Of OrderDetail)) Console.WriteLine("OrderDetails retrieved: {0}", e.ResolvedEntities.Count) End Sub Private Sub ResetEntityManager(ByVal em As EntityManager) em.Clear() em.UseAsyncNavigation = False End Sub

In the method‟s first statement we set the UseAsyncNavigation property of the EntityManager to true. This step would be unnecessary in a Silverlight application, as true is the default setting for that property in that environment. But the above code could run in both Silverlight and non-Silverlight environments. Now consider the statements that retrieve the Order. For a couple of reasons, we can‟t simply say this… C# VB
Order anOrder = _em1.Orders.FirstOrNullEntity();

Dim anOrder As Order = _em1.Orders.FirstOrNullEntity()

IdeaBlade DevForce

Business Object Persistence

…firstly, because the attempt to execute the above statement would fail in a Silverlight app with a message to the effect that “Queries in Silverlight must be executed asynchronously.” But in fact it also is not possible at present to execute asynchronously immediate execution queries (of which any query ending with a call to FirstOrNullEntity() is an example). So to get our single Order, we need to submit a query with a condition that retrieves the desired Order, as you saw in the main snippet. That query must, of course, also be submitted asynchronously, and a callback method provided to process the results.

C#

... IEntityQuery<Order> query = _em1.Orders.Where(o => o.OrderID == 10248); _em1.ExecuteQueryAsync<Order>(query, GotOrders, null); ... } private void GotOrder(EntityFetchedEventArgs<Order> args) { if (args.Error != null) { Console.WriteLine(args.Error.Message); } else { // Retrieve a single related entity using a scalar navigation property Order targetOrder = (Order)args.Result.ToList()[0]; Console.WriteLine("Order: {0}", targetOrder.OrderID.ToString()); } }

VB

... Private IEntityQuery(Of Order) query = _em1.Orders.Where(Function(o) o.OrderID = 10248) _em1.ExecuteQueryAsync(Of Order)(query, GotOrders, Nothing) ... private void GotOrder(EntityFetchedEventArgs(Of Order) args) If args.Error IsNot Nothing Then Console.WriteLine(args.Error.Message) Else ' Retrieve a single related entity using a scalar navigation property Dim targetOrder As Order = CType(args.Result.ToList()(0), Order) Console.WriteLine("Order: {0}", targetOrder.OrderID.ToString()) ... End If End Sub

In this case, since we‟re using the primary key to fetch our Order, we know that args.Result will contain at most one entity; so we simply cast it into an Order and proceed. To get the Customer related to that Order (refer back to the full snippet), we set up a handler for the PendingEntityResolved event of the Customer navigation property, targetOrder.Customer. Then to initiate the asynchronous retrieval of that customer, we reference it in a code statement:

C# VB

Customer aCustomer = targetOrder.Customer;

Dim aCustomer As Customer = targetOrder.Customer

IdeaBlade DevForce

Business Object Persistence

We included a call to Console.WriteLine() immediately following the above statement just to show that the desired Customer simply isn‟t going to be available at that point. The statement will write out a blank for the Customer‟s CompanyName. Where we will get results is in the Customer_PendingEntityResolved handler:

C#

void Customer_PendingEntityResolved(object sender, PendingEntityResolvedEventArgs e) { Customer customer = (Customer)e.ResolvedEntity; Console.WriteLine("Customer (from Customer_PendingEntityResolved): {0}", customer.CompanyName); }

VB

Private Sub Customer_PendingEntityResolved(ByVal sender As Object, _ ByVal e As PendingEntityResolvedEventArgs) Dim customer As Customer = CType(e.ResolvedEntity, Customer) Console.WriteLine("Customer (from Customer_PendingEntityResolved): {0}", _ customer.CompanyName) End Sub

Collection Navigation Properties in Silverlight
For navigation properties that return a collection, DevForce provides a PendingEntityListResolved event, similar to the PendingEntityResolved event we‟ve just discussed:

C#
...

private void GotOrder(EntityFetchedEventArgs<Order> args) { // Retrieve a collection of related entities using a collection navigation property targetOrder.OrderDetails.PendingEntityListResolved += new EventHandler<PendingEntityListResolvedEventArgs<OrderDetail>>( OrderDetails_PendingEntityListResolved); } } void OrderDetails_PendingEntityListResolved(object sender, PendingEntityListResolvedEventArgs<OrderDetail> e) { Console.WriteLine("OrderDetails retrieved: {0}", e.ResolvedEntities.Count); }

VB

Private Sub GotOrder(ByVal args As EntityFetchedEventArgs(Of Order)) ... ' Retrieve a collection of related entities using a collection navigation property AddHandler targetOrder.OrderDetails.PendingEntityListResolved, _ AddressOf OrderDetails_PendingEntityListResolved End If End Sub

When we run the full snippet, the code displays the following results in the Console window:

IdeaBlade DevForce

Business Object Persistence

The output line “Press ENTER to continue..” comes from the utility method PromptToContinue(), which executes synchronously and immedately. Then we see reflected back the OrderID of the retrieved Order; the non-existent CompanyName of the not-yet-retrieved, related Customer; the CompanyName of the Customer written after its retrieval by the Customer_PendingEntityResolved callback method; and the display of OrderDetails retrieved, written by the OrderDetails_PendingEntityListResolved method.

Using An Anonymous Method for Navigation Property Callback
If you‟re working in C#, you can also use inline, anonymous methods for your ExecuteQueryAsync() callbacks:
Code Snippet 34. NavigationBasicAsynchronousAnonymousCallback (C# only)

C#

public void NavigationBasicAsynchronousAnonymousCallback() { _em1.UseAsyncNavigation = true; IEntityQuery<Order> query = _em1.Orders.Where(o => o.OrderID == 10248); _em1.ExecuteQueryAsync<Order>( query, // IEntityQuery<Order> (args) => { // AsyncCompletedCallback Console.WriteLine("Order: {0}", // " ((Order)args.Result.ToList()[0]).OrderID); // " }, // " null // UserState object ); PromptToContinue(); }

These are handy when the logic to be included in the callback isn‟t too involved. VB.NET doesn't support multistatement lambda expressions or anonymous methods.

Deferred Retrieval
When does the EntityManager fetch myOrder‟s line items from the data source? We might have written DevForce to fetch them automatically when it fetched myOrder. But if DevForce were to get the line items automatically, why stop there? It could get the customer for the order, the sales rep for the order, and the products for each line item. Those are just the immediate neighbors. It could get the customer‟s headquarter address, the sales rep‟s address and manager, and each product‟s manufacturer. If it continued like this, it might fetch most of the database. Retrieving the entire graph is obviously wasteful and infeasible. How often do we want to know the manager of the sales rep who booked the order? Clearly we have to prune the object graph. But where do we prune? How can we know in advance which entities we will need and which we can safely exclude?

IdeaBlade DevForce

Business Object Persistence

We cannot know. Fortunately, we don‟t have to know28. We keep it simple. We use an entity query to get the root entities (such as myOrder). Then we use entity navigation to retrieve neighboring related entities as we need them. This just-in-time approach is called deferred retrieval (also known as “lazy instantiation”, “lazy loading”, “Just-InTime [JIT] data retrieval”, and so on).

Proactive Data Loads
Having established that the DevForce default is deferred retrieval, we hasten to add that there are many circumstances when it absolutely makes sense to load data before it is specifically needed to satisfy some demand of the application. Filling a large data grid is an excellent example of such a situation. Suppose you‟re filling a grid with Orders – lots of them – and that for each Order you also wish to display the name of the Customer who placed it, the Sales Representative who wrote it, and the Shipping Company that will deliver it. With deferred retrieval, filling a single row of the grid would require three extra trips to the data source – one each for a Customer, Employee, and Shipper entity -- above and beyond the one that got all of the Orders to begin with. If the grid were populated with a thousand Orders, there would be three thousand separate (and unnecessary) trips to the data source to retrieve the related entities. You can well imagine that this might negatively impact your application‟s performance. For circumstances like these where there is an obvious impending need for a great deal of related data, you can add Include() clauses to your data retrieval query to bring back the related data at the same time your retrieve the root data. The following example retrieves selected Customers and a graph of related data: the Customers‟ Orders, the OrderDetails for those Orders, the Products referenced in the OrderDetails, the Suppliers of those Products, and the SalesRep who wrote the Orders:
Code Snippet 35. NavigationSynchronousPreload

C#

IEntityQuery<Customer> query = _em1.Customers.Where(c => c.Country == "France") .Include("Orders") .Include("Orders.OrderDetails") .Include("Orders.OrderDetails.Product") .Include("Orders.OrderDetails.Product.Supplier") .Include("Orders.SalesRep"); _em1.ExecuteQuery<Customer>(query); // _em1.ExecuteQuery(query); // accomplishes the same thing // query.ToList(); // accomplishes the same thing

VB

Dim query As IEntityQuery(Of Customer) = _ _em1.Customers.Where(Function(c) c.Country = "France") _ .Include("Orders") _ .Include("Orders.OrderDetails") _ .Include("Orders.OrderDetails.Product") _ .Include("Orders.OrderDetails.Product.Supplier") _ .Include("Orders.SalesRep") _em1.ExecuteQuery(Of Customer)(query) ' _em1.ExecuteQuery(query) // accomplishes the same thing ' query.ToList() // accomplishes the same thing

Proactive Data Loads in Silverlight
In Silverlight apps, where all data retrieval must be asynchronous, the benefits of preloading data are even more general. In the following snippet, we preload, using a span query, a large object graph for each of a group of

28

We don‟t have to know if we can be certain of continuous connection to the data source. If we expect the application to run offline, we‟ll have to anticipate the related entities we‟ll need and pre-fetch them. We‟ll get to this issue later.

IdeaBlade DevForce

Business Object Persistence

Customers who meet a specified condition. Having done so, all of our subsequent queries for entities can be cacheonly and synchronous:
Code Snippet 36. NavigationAsynchronousPreload

C#

public void NavigationAsynchronousPreload() { ResetEntityManager(_em1); _em1.UseAsyncNavigation = true; IEntityQuery<Customer> query = _em1.Customers.Where(c => c.Country == "France") .Include("Orders") .Include("Orders.OrderDetails") .Include("Orders.OrderDetails.Product") .Include("Orders.OrderDetails.Product.Supplier") .Include("Orders.SalesRep"); _em1.ExecuteQueryAsync<Customer>(query, GotCustomers, null); PromptToContinue(); } private void GotCustomers(EntityFetchedEventArgs<Customer> args) { if (args.Error != null) { Console.WriteLine(args.Error.Message); } else { DisplayCacheContents(); } } private void DisplayCacheContents() { Console.WriteLine("Contents of Cache"); Console.WriteLine("-----------------"); Console.WriteLine("Customers: {0}", _em1.Customers.With(QueryStrategy.CacheOnly).Count()); Console.WriteLine("Employees: {0}", _em1.Employees.With(QueryStrategy.CacheOnly).Count()); Console.WriteLine("Orders: {0}", _em1.Orders.With(QueryStrategy.CacheOnly).Count()); Console.WriteLine("OrderDetails: {0}", _em1.OrderDetails.With(QueryStrategy.CacheOnly).Count()); Console.WriteLine("Products: {0}", _em1.Products.With(QueryStrategy.CacheOnly).Count()); Console.WriteLine("Suppliers: {0}", _em1.Suppliers.With(QueryStrategy.CacheOnly).Count()); }

VB

Public Sub NavigationAsynchronousPreload() ResetEntityManager(_em1) _em1.UseAsyncNavigation = True Dim query As IEntityQuery(Of Customer) = _em1.Customers _ .Where(Function(c) c.Country = "France") _ .Include("Orders") _ .Include("Orders.OrderDetails") _ .Include("Orders.OrderDetails.Product") _ .Include("Orders.OrderDetails.Product.Supplier") _ .Include("Orders.SalesRep") _em1.ExecuteQueryAsync(Of Customer)(query, AddressOf GotCustomers, Nothing) PromptToContinue() End Sub Private Sub GotCustomers(ByVal args As EntityFetchedEventArgs(Of Customer)) If args.Error IsNot Nothing Then Console.WriteLine(args.Error.Message) Else DisplayCacheContents() End If End Sub

IdeaBlade DevForce
Private Sub DisplayCacheContents() Console.WriteLine("Contents of Cache") Console.WriteLine("-----------------") Console.WriteLine("Customers: {0}", _ _em1.Customers.With(QueryStrategy.CacheOnly).Count()) Console.WriteLine("Employees: {0}", _ _em1.Employees.With(QueryStrategy.CacheOnly).Count()) Console.WriteLine("Orders: {0}", _ _em1.Orders.With(QueryStrategy.CacheOnly).Count()) Console.WriteLine("OrderDetails: {0}", _ _em1.OrderDetails.With(QueryStrategy.CacheOnly).Count()) Console.WriteLine("Products: {0}", _ _em1.Products.With(QueryStrategy.CacheOnly).Count()) Console.WriteLine("Suppliers: {0}", _ _em1.Suppliers.With(QueryStrategy.CacheOnly).Count()) End Sub

Business Object Persistence

Here is the output of the above method:

Missing objects
Every order should have a shipping address. What if it doesn‟t? Will myOrder.ShippingAddress.City throw an exception? Will we have to wrap every entity navigation in a giant try/catch block? Will it return null? Will we have to follow every entity navigation with a test for null? That might be worse than catching an exception. Fortunately entity navigation neither returns a null nor throws an exception. Instead, when the EntityManager discovers there is no shipping address, it returns the Address Null Entity.

The Null Entity
The null entity is a sentinel object that looks and behaves, for the most part, like a real entity instance. Every entity class defines its own “null entity” instance. When a query such as anEntityManager.DiscontinuedProducts must return an entity and it has no valid entity instance to return, it returns a null entity of the requested type instead. When a navigation property should return a related entity instance and there is no such instance, it will return a null entity instead. This is far better than returning a null (Nothing in VB). The caller can‟t do a thing with null and may even crash.

IdeaBlade DevForce

Business Object Persistence

The null entity, on the other hand, has the properties of a real entity instance. For example, it can report its type and the EntityManager that owns it29. All cached entities answer to IsNullEntity; only a null entity replies true. Most of its properties return runtime safe but semantically “empty” values that can be displayed in a UI. If anEmployee is a null entity, for example, the expression anEmployee.FirstName returns an empty string. The navigation property anEmployee.Orders returns an empty IList<Order>. The navigation property anEmployee.HomeAddress returns the Address null entity. This means we can write a long expression such as anEmployee.HomeAddress.State.Name without throwing an exception. In this case the Address null entity‟s State navigation property returns a State null entity whose Name property returns an empty string. The null entity cannot be changed, deleted, or saved. But the savvy developer can redefine a null entity‟s default property responses by overriding the UpdateNullEntity() method in the entity‟s Developer class30. She could change the Address.City property, for example, so that it returns the string “<unknown>”.

Asynchronous Communication with the Business Object Server
The EntityManager now supports asynchronous versions of methods which communicate with the BOS. These methods include:         LoginAsync LogoutAsync ExecuteQueryAsync ExecuteQueryAsync<T> SaveChangesAsync ForceIdFixupAsync RefetchEntitiesAsync InvokeServerMethodAsync

Asynchronous communication with the BOS is considerably more complicated than synchronous; but alas, it is the law of the land in Silverlight applications. So, for those many of you who are working in that environment, we are addressing the topic here, rather than in the Business Object Persistence – Advanced document. The EntityManager supports a hybrid of the .NET event-based asynchronous pattern31 for these asynchronous methods. We refer to it as a “hybrid” because corresponding events have not been defined for these methods. So instead of subscribing to an event to receive notification about the completion status, you can instead pass a methodspecific callback as part of the call. You can identify this hybrid pattern by the OperationNameAsync naming convention. See, for example, the code below to submit a query asynchonously.

Asynchronous Queries
You‟ve seen asynchronous queries earlier in this document, but here we revisit them with a slightly more formal treatment, and in the context of other asynchronous communications with the Business Object Server.

29 30

Like real cached entities, null entities must belong to a EntityManager and, in fact, are created by a EntityManager This method is inherited from the root business object class, Entity.

31

The standard .NET “Event-based Asynchronous Pattern” is described in described in an article at this URL:
http://msdn2.microsoft.com/en-us/library/wewwczdw(en-US,VS.80).aspx

IdeaBlade DevForce

Business Object Persistence

The following code defines an EntityQuery and launches it asynchronously, assigning the result set to a list in the operation‟s callback method:
Code Snippet 37. BOSCom_AsyncQuery

C#

private void BOSCom_AsyncQuery() { ResetEntityManager(_em1); var query = new EntityQuery<Customer>() .Where(c => c.Country == "Denmark"); int token = 1; _em1.ExecuteQueryAsync<Customer>(query, QueryCompletedCallback, token); } private void QueryCompletedCallback(EntityFetchedEventArgs<Customer> args) { var resultList = args.Result; Console.WriteLine("Query returned {0} entities", resultList.Count()); }

VB

Private Sub BOSCom_AsyncQuery() ResetEntityManager(_em1) Dim query = New EntityQuery(Of Customer)().Where(Function(c) c.Country = "Denmark") Dim token As Integer = 1 _em1.ExecuteQueryAsync(Of Customer)(query, AddressOf QueryCompletedCallback, token) End Sub Private Sub QueryCompletedCallback(ByVal args As EntityFetchedEventArgs(Of Customer)) Dim resultList = args.Result Console.WriteLine("Query returned {0} entities", resultList.Count()) PromptToContinue() End Sub

As you‟ve seen previously, in C#, you have the additional option of passing a lambda expression for the callback instead of defining a separate method:
Code Snippet 38. BOSCom_AsyncQueryLambda

C#

private void BOSCom_AsyncQueryLambda() { var query = new EntityQuery<Customer>().Where(c => c.Country == "Denmark"); int token = 2; _em1.ExecuteQueryAsync<Customer>( query, (args) => { var resultList = args.Result; Console.WriteLine("Query returned {0} entities", resultList.Count()); }, token); PromptToContinue(); }

IdeaBlade DevForce
The signature for the above queries is as follows: C#

Business Object Persistence

public void ExecuteQueryAsync<T>( IEntityQuery<T> query, AsyncCompletedCallback<EntityFetchedEventArgs<T>> userCallback, object userState );

VB

public void ExecuteQueryAsync(Of T)( _ IEntityQuery(Of T) query, _ AsyncCompletedCallback(Of EntityFetchedEventArgs(Of T)) userCallback, _ Object userState _ )

You can run multiple ExecuteQueryAsync operations simultaneously. The final parameter, userState, is a unique object created by the developer to identify the async query. When a query completes, the UserState is returned to the caller as part of the EntityFetchedEventArgs argument so she can distinguish one query from another. The UserState object can be as simple as an integer, or it can be an arbitarily complex custom type.

Completed Queries
The EntityFetchedEventArgs parameter passed into an async query‟s callback method contains the following members:

Property
Cancelled

Description True if the query was canceled. Cancellation of an async operation can be ordered by a call to EntityManager.CancelAsync(), which takes a UserState object as a parameter. Such cancellation only succeeds if the order is received in time. Note that UserState objects must be unique across all async operations, whether queries, saves, logins, or other. An IList containing every entity added to or modified in the EntityManager cache. An Exception object, of an exception was thrown during the async operation. True if the operation completed successfully. True if the operation was executed synchronously and completed successfully. A query, for example, will execute synchronously if DevForce determines that it can be satisfied entirely from the EntityManager cache.

ChangedEntities Error IsCompleted IsCompletedSynchronously

IdeaBlade DevForce
Query Result UserState

Business Object Persistence
The IEntityQuery<T> object used in the asynchronous operation. An IEnumerable<T> of returned objects. The object passed in the async call uniquely to identify the operation.

IAsyncResult Asynchronous Pattern
For those needing additional control over their asynchronous operations, the EntityManager also supports the IAsyncResult asynchronous pattern through an explicit implementation of the IEntityManagerAsync interface. You will need to cast an EntityManager to this interface in order to use methods following this pattern. In the IAsyncResult pattern an asynchronous operation is implemented as two methods named BeginOperationName and EndOperationName to begin and end the asynchronous operation "OperationName". More information on using this interface is available in the IdeaBlade DevForce Reference Help, available from the IdeaBlade DevForce Windows Start menu.

Asynchronous Fulfillment of Navigation Property Queries
DevForce returns data for navigation properties (such as Order.Customer or Order.OrderDetails) by issuing queries. Explicit queries in your DevForce app can be written using the asynchronous method calls detailed above, but control over the fulfillment of navigation properties must be exercised in a different manner. The EntityManager now has a boolean UseAsyncNavigation property that can be set to specify that navigation properties should be fulfilled using asynchronous queries. When reference is made to a navigation property, DevForce returns either an entity (if the property is scalar) or a RelatedEntitiesList<T> (for collection properties). Entities now have an IsPendingEntity property; When EntityManager.UseAsyncNavigation is set to true, the entities initially returned for scalar properties, and the RelatedEntityLists returned for collection properties, will have a “pending” state until the asynchronous query issued for their fulfillment actually returns data. This state can be diagnosed with one of the following properties:   Entity.IsPendingEntity RelatedEntityList<T>.IsPendingEntityList

Entities and RelatedEntityLists also now have events that fire when the data for pending entities is returned. These are:   Entity.PendingEntityResolved RelatedEntityList<T>.PendingEntityListResolved

Handlers can be attached to these event to perform actions when the data for pending entities becomes available to your app.

Canceling Pending Operations
We may attempt to cancel an asynchronous peration by calling EntityManager.CancelAsync(). We identify the operation to cancel by passing in its identifying UserState. The operation stops at the next safe breaking point before the operation finishes, if such a breaking point exists, and then invokes the callback method. The caller can confirm that the query was successfully canceled by checking the Cancelled parameter of the EventArgs object; it should read true.

IdeaBlade DevForce

Business Object Persistence

The EntityListManager
Instances of IdeaBlade.EntityModel.EntityListManager<T> watch the DevForce cache for changes and add entity references to designated lists if such changes meet developer-defined rules. Consider the following code:
Code Snippet 39. SetUpEntityListManager

C#

var filter = new Predicate<Employee>( delegate(Employee anEmployee) { return anEmployee.City == "London"; }); _employeeEntityListManager = new EntityListManager<Employee>(_em1, filter, null); bool refreshListWhenPlacedUnderManagement = true; _employeeEntityListManager.ManageList(_salesReps, refreshListWhenPlacedUnderManagement);

VB

Dim filter = New Predicate(Of Employee)( _ Function(anEmployee As Employee) anEmployee.City = "London") _employeeEntityListManager = New EntityListManager(Of Employee)(_em1, filter, Nothing) Dim refreshListWhenPlacedUnderManagement As Boolean = True _employeeEntityListManager.ManageList(_salesReps, refreshListWhenPlacedUnderManagement)

This code sets up an EntityListManager to watch the cache for changes to Employees, or the insertion of new Employees. If any changed or new Employee is found to be based in London, a reference to that Employee will be added to the _salesReps list. At the same time, _employeeEntityListManager will inspect all items in the _salesReps list to see that they meet the specified rule about London. The only requirements for _salesReps are that it   implement System.Collections.IList; and contain instances of IdeaBlade.EntityModel.Entity.

A single EntityListManager can manage as many different lists as you wish. To put _employeeEntityListManager in charge of additional lists, you would simply invoke its ManageList() method again for each desired list:

C#

_employeeEntityListManager.ManageList(_telecommuters, false); _employeeEntityListManager.ManageList(_fieldAgents, false);

VB

_employeeEntityListManager.ManageList(_telecommuters, False) _employeeEntityListManager.ManageList(_fieldAgents, False)

Of course, it only makes sense to do this when the same inclusion criteria apply to each targetted list. In additions to changes to the cache, changes to a managed list trigger action by the managing EntityListManager. Thus, any of the follows statements will cause _employeeEntityListManager to examine the current contents of the cache and add references to all London employees to the _salesReps list:

C#

_salesReps.Add(anEmployee); _salesReps.Remove(anEmployee); _salesReps.Clear();

VB

_salesReps.Add(anEmployee)

IdeaBlade DevForce
_salesReps.Remove(anEmployee) _salesReps.Clear()

Business Object Persistence

In the case of the statement _salesReps.Clear(), you will not end up with an empty list unless you first remove _salesReps from the list of lists being managed by employeeEntityListManager. Removing an entity that the rule says should be included also will not result in the entity disappearing from the list. The EntityListManager will just put it right back! In general, beware of making manual changes (adds or removals) to the set of items contained in a managed list.

EntityListManagers and The NullEntity
One exception occurs when you want the NullEntity for the type contained in a list to be included. NullEntities are singletons and do not reside in the cache, so there is no way that an EntityListManager will ever find one there to add a reference to! If you want the NullEntity in a managed list, you should manually add it. The ListManager will not remove it.

EntityListManagers and Duplicates
The EntityListManager will not eliminate duplicates from a list. For example, suppose you direct the following statement against a list, _salesReps, that is already being managed to include Employees based in London:

C# VB

_salesReps.ReplaceRange(_entityManager.Employees.Where(e=>e.City == "London"));

_salesReps.ReplaceRange(_entityManager.Employees.Where(e=>e.City == "London"))

You will end up with duplicate references to each of the London employees!

EntityListManagers and Performance
EntityListManagers do create a certain amount of overhead, so be judicious in their use. It is also possible to narrow their scope of what they must monitor more than we did in our examples above. We instantiated our EntityListManager as follows:

C#

var filter = new Predicate<Employee>( delegate(Employee anEmployee) { return anEmployee.City == "London"; }); _employeeEntityListManager = new EntityListManager<Employee>(_entityManager, filter, null);

VB The third argument, which we left null, is an array of EntityProperty objects. By leaving it null, we told the manager to submit any added or modified Employee to the test encoded in the filter Predicate. Suppose that, instead, we pass a list of properties of the Employee to this argument:

C#

new EntityListManager<Employee>(_entityManager, filter, new EntityProperty[]{Employee.CityEntityProperty});

IdeaBlade DevForce
VB

Business Object Persistence

_employeeEntityListManager = New EntityListManager(Of Employee)(_em1, filter, _ New EntityProperty() {Employee.CityEntityProperty})

Now the EntityListManager will apply its test (about City being equal to London) only to an Employee whose City property, specifically, was modified. If you simply change only the Birthdate of an Employee already in the cache, the rule will not be evaluated. It can, after all, be safely assumed that said Employee would already be in the lists being managed if the value in its City property were “London”.

Coding More Involved Rules
In the examples above we passed an anonymous delegate to the constructor of the Predicate filter. That‟s great for simple rules, but you can declare the predicate separately if you need to do something more involved. This also gives you a chance to name the rule, which can make your code more readable. Here‟s a simple example:
Code Snippet 40. SetUpEntityListManagerWithNamedDelegate

C#

private void SetUpEntityListManagerWithNamedDelegate() { // Identify Customer currently being edited by some process; // this is a stand-in. _currentCustomer = _em1.Customers.FirstOrNullEntity(); EntityListManager<Order> orderEntityListManager = new EntityListManager<Order>(_em1, FilterOrdersByDate, new EntityProperty[] { Order.OrderDateEntityProperty, Order.CustomerEntityProperty } ); } /// <summary> /// This rule gets the 1996 Orders for the current Customer /// </summary> /// <param name="pOrder"></param> /// <returns></returns> Boolean FilterOrdersByDate(Order pOrder) { return (pOrder.OrderDate.Value.Year == 1996 && pOrder.Customer == _currentCustomer); }

VB

Private Sub SetUpEntityListManagerWithNamedDelegate() ' Identify Customer currently being edited by some process; ' this is a stand-in. _currentCustomer = _em1.Customers.FirstOrNullEntity() Dim orderEntityListManager As New EntityListManager(Of Order)(_em1, _ AddressOf FilterOrdersByDate, New EntityProperty() { _ Order.OrderDateEntityProperty, Order.CustomerEntityProperty}) End Sub ''' <summary> ''' This rule gets the 1996 Orders for the current Customer ''' </summary> ''' <param name="pOrder"></param> ''' <returns></returns> Private Function FilterOrdersByDate(ByVal pOrder As Order) As Boolean Return (pOrder.OrderDate.Value.Year = 1996 AndAlso

IdeaBlade DevForce
pOrder.Customer.Equals(_currentCustomer)) End Function

Business Object Persistence

Entity Caching
There are at least three good reasons to cache business objects: 1. 2. 3. The connection to the server may break during a session Writing business object changes directly to the data source is impractical and often unwise. In real life applications, the same entities are retrieved repeatedly; it wastes time and resources to bother the server with redundant requests for the same entities.

Each DevForce EntityManager has its own, private entity cache that:         holds all retrieved and newly created entities; is searchable by query and object navigation; tracks cached entity changes, deletions and additions; insulates the developer from cache mechanics; enables the developer to control how entities are fetched and merged into the cache; raises events when entities are fetched, changed, deleted, or added; permits the developer to manipulate the cache when necessary; can be persisted to and retrieved from client storage.

All Business Objects are Cached
We always create or retrieve a business object into the cache of a particular Entity Manager instance. Every business object can report to which Entity Manager instance it belongs. The developer can rummage around in its cache discovering and manipulating the business objects therein.

Entity Ancestry and Organization of the Cache
The business object developer class inherits from IdeaBlade.EntityModel.Entity. In DevForce 3.x, and in earlier versions of DevForce, Entity inherited from System.Data.DataRow; but for several reasons, chief among which was our desired to make our object model cross-compatible with Silverlight (which does not support DataSets), we have replaced the DataSet and its component parts with our own set of storage classes. Entity now lies at the base of the business object inheritance tree, inheriting nothing, though it does implement several interfaces. These include:     IdeaBlade.EntityModel.IEntityBase; System.ComponentModel.IEditableObject; System.ComponentModel.INotifyPropertyChanged; and SystemRuntime.InteropServices.IComparable.

Previously an Entity was a DataRow, and as such resided in a System.Data.DataTable, which in turn resided in a System.Data.DataSet. Now an Entity lives in an IdeaBlade.EntityModel.EntityGroup which lives within an EntityGroupCollection. You may find that you rarely need to interact directly with an EntityGroup or

IdeaBlade DevForce

Business Object Persistence

EntityGroupCollection; and virtually all of the metadata you will ever need about an entity can be accessed through the Entity‟s EntityAspect.EntityMetaData property. Public properties and methods of that include the following:

Member Type Property Property

Name EntityType IsComplexType

Function Gets the Type of the entity Returns whether this metadata describes a "ComplexObject"

Property Property

DataSourceKeyName DefaultEntitySetName

Gets the data source key name. The default EntitySetName for entities of this type.

Property

EntityProperties

Returns a collection of EntityProperties that belong to entities of this type. Returns a collection of DataEntityProperties for entities of this type. Returns a collection of DataEntityProperties for entities of this type. Returns a collection of EntityProperties that are keys for entities of this type. Returns a collection of EntityProperties that are concurrency properties for entities of this type. Returns a collection of EntityProperties that describe complex object properties for entities of this type. Gets whether primary key queries are allowed.

Property

DataProperties

Property Property Property Property

NavigationProperties KeyProperties ConcurrencyProperties ComplexTypeProperties

Property

CanQueryByEntityKey

Method Method

CreateEntity() GetDefaultValue(Type pType)

Creates a new entity of the type describe by this metadata item. Returns the default value of a type: usually '0' or null for any data type. Note that this is subtly different from the TypeFns.GetDefaultValue method in that it returns Today for a default date time.

Business objects are unique in each cache
DevForce persistence management ensures that each business object appears at most once in a particular EntityManager cache. No matter how many times the employee “Nancy Davolio” is read into the cache, she appears at most once. Within the application, a reference to any “Nancy Davolio” employee object is a reference to

IdeaBlade DevForce

Business Object Persistence

the same one employee object. If we change her first name to “Sue”, she becomes “Sue” everywhere in the session unless … … unless there is more than one EntityManager instance32. Each EntityManager instance maintains its own independent cache. The “Nancy Davolio” retrieved into EM1 is not the same object as the “Nancy Davolio” retrieved into EM2, even though they are both mapped to the same row in the Employee table of the database. Changes to a copy of a business object in one cache are invisible to other copies in other caches both in this client and in all other clients. Changes become visible to other caches only after the object is saved to the data source and re-fetched to those caches.

Entities in Lists
Entities in lists are always references to entities in the EntityManager‟s cache. This is true whether the EM maintains the list or you maintain the list. In general we prefer to work with only one list of entities of a particular type. But it may be useful to have two such lists that are a little different. For example, one list could hold all employees of the company while the second list holds the subset of those employees who are managers. Both lists contain references to the same employee instances in cache but they are very different lists. If we change the Employee „A‟ who happens to be a manager, we are also changing the Employee „A‟ in the general employee list. They are the same Employee „A‟. If follows that if the PM re-fetches a clean copy of Employee 'A' from the data source, the pending changes will disappear for all viewers of Employee „A‟ whether they are looking at „A‟ in the first list or in the second list.

Business object proper, not the business object graph
When speaking of a business object held in cache, we may easily lose sight of what we mean by a “business object.” We distinguished earlier between the “business object proper”, which encapsulates the simple, scalar values stored in the object‟s base table, and the “business object graph” which embraces the entire network of other business objects to which it is related. For example, the simple Employee properties such as “FirstName” and “LastName” access data values that are stored in the Employee table; these are properties of the employee business object proper. The “HomeAddress” navigation property, on the other hand, delivers a related business object, the employee‟s home address. The data values of the address come from a different table (Address) and “belong” to the address business object proper, not the employee per se. An EntityManager instance retrieves and holds business objects proper, not their graphs. Objects in the graph of a particular business object may be in the cache. Or they may not. They don‟t enter the cache simply by virtue of being in another object‟s graph. The employee‟s home address object will not enter the cache just because we retrieved the employee object. It will enter the cache after we execute an expression such as anEmployee.HomeAddress.

32

Multiple EntityManagers have their place but most applications will need only one. Multiple EMs are covered in “Advanced Business Object Concepts”.

IdeaBlade DevForce

Business Object Persistence

Queries, Navigation, and the Cache
We‟ve covered entity queries and entity navigation. Although entity queries make explicit reference to the EntityManager, we learned that entity navigation is also performed by the EntityManager. Here we explain how the EntityManager processes both explicit entity queries and the implicit queries inside entity navigation syntax. We will see that EM query processing is guided by a query strategy. When following the default, “normal” strategy, the EM tries first to satisfy a query from data in its cache; it reaches out to the data source only if it must.

Query Cache
When a EntityManager begins to process a normal query, it checks its query cache to see if it has processed this exact query before. The query cache holds queries and is not the same as the entity cache which holds objects and is what we usually mean when we refer to “the cache.” If the EntityManager finds the query in the query cache, it assumes that the objects which satisfy the query are in the entity cache; accordingly, it satisfies the query entirely from the cache without consulting the data source. A one-to-many entity navigation, such as from employee to the employee‟s orders, is translated implicitly to an entity query language (OQL) query that also enters the query cache. The next time the application navigates from that same employee to its orders, the EntityManager will recognize that it has performed the query before and look only in the cache for those orders. The query cache grows during the course of a session. Certain operations clear it as one of their side-effects; removing an entity from the cache is one such operation. The developer can also clear the query cache explicitly. We just said that the EntityManager searches the query cache for an exact match of the current query, but that was really a “little white first approximation.” Actually, the EntityManager does better than that: it searches either for an exact match, or for an unrestricted query returning the same type. If, for example, you have previously retrieved “all Customers” and now ask for “Customers from Canada”, your new query will be satisfied from the cache.

Primary key queries
A query for business objects by primary key may be resolved entirely in the cache. If we search 33 for the employee with Id = „1‟ the EntityManager will try to find it in the cache and, if not found there, will only then look for it in the data source. The EntityManager treats navigation along a one-to-one relationship, such as from Employee to HomeAddress, as a primary key query. Navigation in the parent direction along a one-to-many relationship, such as from an OrderDetail to its parent Order, is also a primary key query.

“Object Not Found” and the Null Entity
When we search for an entity and do not find it, the EntityManager, rather than returning a null that may cause an exception in your application, returns a “sentinel” object called the Null Entity. Such a sentinel behaves much like a real entity of the sought-for type except that it can‟t be changed, deleted, or saved. Every business object class defines its own null entity. See “The Null Entity” elsewhere in the section on queries and navigation.

33

If we use the default QueryStrategy; we are just about to discuss QueryStrategy so bear with me.

IdeaBlade DevForce Cache use when disconnected

Business Object Persistence

When the EntityManager “knows” it is disconnected from the server, it will satisfy a navigation, or a query submitted with the Normal QueryStrategy, from the cache alone; it will not attempt to search the data source. If a sought-for object is not in the cache, the EntityManager will return the Null Entity for objects of that type. The EntityManager raises an exception if it discovers during query processing that it can‟t reach the data source; see the “Lost Connections” topic in the “Advanced Business Object Concepts” section below.

Modifications
Each business object carries a read-only EntityState property that indicates if the object is new, modified, marked for deletion, or unchanged since it was last retrieved. It bears repeating that our local modifications affect only the cached copy of a business object, not its version in the data source. The data source version won‟t be updated until the application tells the EntityManager to save the changed object. It follows that the data source version can differ from our cached copy either because we modified the cached copy or because another user saved a different version to the data source after we retrieved our copy. It would be annoying at best if the EntityManager overwrote our local changes each time it queried the data source. Fortunately, in a normal query, the EntityManager will only replace an unmodified version of an object already in the cache; our modified objects are preserved until we save or undo them.

Stale Entity Data
All of this is convenient. But what if another user has made changes to a cached entity? The local application is referencing the cached version and is unaware of the revisions. For the remainder of the user session, the application will be using out-of-date data. The developer must choose how to cope with this possibility. Delayed recognition of non-local changes is often acceptable. A list of U.S. States or zip codes is unlikely to change during a user session. Employee name changes may be too infrequent and generally harmless to worry about. In such circumstances the default caching and query behavior is fine. If concurrency checking is enabled and the user tries to save a changed object to the data source, DevForce will detect the collision with the previously modified version in the data source. The update will fail and DevForce will report this failure to the application which can take steps to resolve it. Some objects are so volatile and critical that the application must be alert to external changes. The developer can implement alternative approaches to maintaining entity currency by invoking optional DevForce facilities for managing cached objects and forcing queries that go to the data source and merge the results back into the cache. The facilities for this are detailed in the section “Query Strategy” further on in this chapter.

Fetch Life Cycle Events
DevForce raises the client-side Fetching event prior to performing a query and raises the client-side Fetched event just before returning query results. We can listen to either or both by attaching a custom handler.

IdeaBlade DevForce

Business Object Persistence

The Fetching event provides the query object. Our handler can examine the object (it implements IEntityQuery) and choose to let the query through, modify it first, or cancel it. If we cancel the query, the Entity Manager method returns as if it found nothing34. The Fetched event fires just before the query method returns. Entities have been fetched and merged into the cache. The event arguments include the list of entities that came from the data source. There might be none if the query found nothing or was satisfied entirely from the cache. It could include entities of the target entity type – the kind we expected returned from the query. It could include entities of other types as is likely if this is a span query or if the query provoked query inversion35. As previously discussed, there are corresponding server-side events named ServerFetching and ServerFetched.

Query Workflow
Putting these points together, we can construct a schematic workflow for normal 36 DevForce entity queries and entity navigation when the application is connected to the Business Object Server (BOS) running on its own physical tier.

Table 2. Entity Query and Navigation Workflow When QueryStrategy = Normal

Component Client Tier – Application Code Client Tier – EntityManager

Action The client application requests a particular set of entities (the “desired entities”) either by entity query or by entity navigation Raises Fetching event. Listeners can see the query and, optionally, cancel the query. Checks if it can satisfy the query with the entities in the client-side cache. If so, it returns them immediately; end of workflow. If not, the EntityManager sends the query along with authentication information to the Business Object Server (BOS) on the middle tier. It may modify the request before sending to the BOS if it can determine that some of desired entities are already in the client side cache.

Middle Tier - Business Object Server

The BOS authenticates the client (the currently logged in “user”) and runs any developer-specified security checks in the ServerFetching handler. If security

34

If the method returns a scalar entity, it yields the return entity type‟s Null Entity; otherwise, it returns a null entity list. Beware of canceling an entity navigation list query method Span queries are later in this section. We cover “Query Inversion” in the “Advanced Business Object Concepts”. The workflow is different in a few places when we use a different QueryStrategy. See the “QueryStrategy” topic under “Advanced Business Object Concepts”.

35 36

IdeaBlade DevForce
checks fail, it raises a security exception and sends this back to the client tier. Middle Tier - Business Object Server If the data source is a relational database: Having passed security checks, the BOS converts the query into one or more LINQ-toEntities queries in the form expected by the ADO.NET Entity Framework. If a relational database is the data source, the Entity Framework converts the LINQ to Entities query into one or more SQL queries and submits them to the data source query mechanism. Data source – Data Source Middle Tier - Business Object Server If the data source is a web service: The BOS converts the query into appropriate web service calls and submits them against the targeted service.

Business Object Persistence

The data source performs the query or queries and returns one or more result sets back to the Business Object Server. If the data source is a relational database: The Entity Framework converts the result sets returned from the data source into ADO.NET entities and delivers them to the EntityServer. If the data source is a web service: The EntityServer converts the result sets returned from the data source into entities.

Middle Tier – Business Object Server

The EntityServer repackages the entities obtained from the data source into a format that can be transmitted efficiently. It then ships the entity data to the client side application. After transmission, the BOS allows the server‟s local copy of the entities to go out of scope and the garbage collector reclaims them. This enables the BOS to stay stateless.

Middle Tier – Business Object Server

IdeaBlade DevForce
Client Tier – EntityManager Client Tier –EntityManager: Compares fetched entities to entities already in the cache. Adds new entities to the cache. Replaces matching cached entities that are unmodified (in essence refreshing them). Preserves cached entities with pending modifications because the query strategy is normal. Client Tier –EntityManager: Reapplies the original query to the cache to locate all desired entities. Client Tier –EntityManager: Raises the Fetched event. Listeners can examine the list of entities actually retrieved from the data source. Client Tier –EntityManager: Returns the desired entities to the application. Client Tier – Application Code Client Tier – Application Code: The entities are available for processing.

Business Object Persistence

The application developer may proceed blissfully unaware of all this effort.

Query Strategy
When the EntityManager performs a query, it follows a query strategy. That strategy determines several things, chief among them these:   the source of the data returned in a query; how data obtained from a source external to the EntityManager cache is merged with existing data in the cache; and how issues related to satisfaction of the query from the cache are handled.

The QueryStrategy is a settable property of the query itself:
Code Snippet 41. QueryStrategyAssortedSyntaxExamples

C#

EntityQuery<Order> query01 = _em1.Orders; query01.QueryStrategy = QueryStrategy.DataSourceThenCache;

VB

Dim query01 As EntityQuery(Of Order) = _em1.Orders query01.QueryStrategy = QueryStrategy.DataSourceThenCache

In addition, every EntityManager has a DefaultQueryStrategy that is used whenever you do not explicitly specify the query strategy you want to use with a particular query. You can also change this default:

IdeaBlade DevForce
C# VB
_em1.DefaultQueryStrategy = QueryStrategy.Normal;

Business Object Persistence

_em1.DefaultQueryStrategy = QueryStrategy.Normal

Entity navigation (e.g., myEmployee.Orders) is implemented with relation queries governed by the DefaultQueryStrategy. In addition, any query whose QueryStrategy property has a value of null will be executed with the DefaultQueryStrategy for the EntityManager underwhich it is run. The QueryStrategy object has four properties: FetchStrategy, MergeStrategy, InversionMode, and TransactionSettings. The FetchStrategy controls where DevForce looks for the requested data: in the cache, in the datasource, or in some combination of the two. The MergeStrategy controls how DevForce resolves conflicts between the states of objects which, although already in the cache, are also retrieved from an external source. The InversionMode controls whether DevForce attempts to retrieve objects that are referenced in the query but are not the target type (e.g., the query “give me all Customers with Orders in the current year” will return references to Customer objects, but must process Order objects along the way). The TransactionSettings object permits you to control the TimeOut and IsolationLevel associated with a query, and also whether and how to use the Microsoft Distributed Transaction Coordinator. There are five static (Shared in VB) properties in the IdeaBlade.EntityModel.QueryStrategy class that return the five most common combinations of a FetchStrategy, a MergeStrategy, and an InversionMode. These will be named and discussed momentarily, but are much easier to understand after examining the available FetchStrategy, MergeStrategy, and InversionMode options.

Fetch Strategies
Five FetchStrategies are available in DevForce:

Table 3. FetchStrategies

Strategy
CacheOnly

Action Apply this query against the cache only, returning references only to entities already there. Do not consult the data source. (Note that this query leaves the cache unchanged.) Retrieve matching entries from the datasource into the entity cache. Return references only to those entities retrieved from the the data source. A result set returned from a query using this FetchStrategy would not include locally added entities that had not yet been persisted to the data source. First retrieve matching entries from the datasource into the entity cache. Discard all references to entities retrieved in this step. Resubmit the same query against the updated cache. Return references only to entities matched by this second, CacheOnly query.

DataSourceOnly

DataSourceThenCache

DataSourceAndCache

First retrieve matching entries from the datasource into the entity cache. Retain references to entities retrieved in this step.

IdeaBlade DevForce

Business Object Persistence
Resubmit the same query as CacheOnly. Combine (union) the references obtained in this second, CacheOnly query with those obtained in the data source retrieval step.

Optimized

Check the query cache to see if the current query has previously been submitted (and, if necessary, inverted) successfully. If so, satisfy the query from the entity cache, and skip the trip to the datasource. If the query cache contains no query matching or encompassing the current query, then determine if all entities needed to satisfy the query correctly from the cache can be retrieved into the cache.37 If so, apply the DataSourceThenCache FetchStrategy. Otherwise, apply the DataSourceOnly FetchStrategy.

Operation of the FetchStrategies When the Client is Disconnected from the Data Source
If the client is disconnected from the data source, the DataSourceOnly, DataSourceThenCache, and DataSourceAndCache strategies will throw an InvalidOperationException. The Optimized strategy will behave as a CacheOnly query. It will not throw an exception, even if no matching query exists in the query cache.

MergeStrategies
A MergeStrategy comes into play whenever DevForce discovers that an entity retrieved from an external source already exists in the entity cache. (The two versions are recognized as the same entity because of matching type and primary key value.) The MergeStrategy determines how DevForce will resolve any conflict found in the two instances of the entity.38 DevForce supports five different MergeStrategies: PreserveChanges, OverwriteChanges, PreserveChangesUnlessOriginalObsolete, PreserveChangesUpdateOriginal, and NotApplicable. Their meanings are shown in Table 4. When reviewing the table, remember that, for every cached DevForce entity, two states are maintained: Original and Current. The Original state comprises the set of values for all properties as they existed at the time of the last retrieval from, or save to, the datasource. The Current state comprises the set of values for the object‟s properties as the end user sees them. That is, the Current state values reflect any local changes that have been made since the entity was retrieved, or last saved. When an entity is persisted, it is the values in its Current state that are saved.
Table 4. MergeStrategies

Strategy
PreserveChanges OverwriteChanges PreserveChangesUnless OriginalObsolete

Action when cached entity has pending changes
Preserves the state of the cached entity. Overwrites the cached entity with data from the data source. Sets the EntityState of the cached entity to Unchanged. Preserves the values in the Current state of the cached entity, if its Original state matches the state retrieved from the datasource.

37 38

See the discussion on query inversion for more detail. Conflicts are diagnosed by comparing the values in the entity‟s designated Concurrency column.

IdeaBlade DevForce

Business Object Persistence
If the state as retrieved from the datasource differs from that found locally in the Original set of property values, this indicates that the entity has been changed externally by another user or process. In this case (with this MergeStrategy), DevForce overwrites the local entity, setting the values in both its Current and Original states to match that found in the datasource. DevForce also then sets the EntityState of the cached instance to Unchanged.

PreserveChangesUpdateOriginal

Unconditionally preserves the values in the Current version for the cached entity; and also updates the values in its Original version to match the values in the instance retrieved from the datasource. This has the effect of rendering the local entity savable (upon the next attempt), when it might otherwise trigger a concurrency exception. This merge strategy must be used – and may only be used – with the CacheOnly fetch strategy. No merge action applies because no data is retrieved from any source outside the cache.

NotApplicable

We drill deeper into the topic of merge strategies in the section “MergeStrategy In More Detail” much later in this chapter. We suggest you defer reading that at least until you‟ve completed this section on Query Strategy – so you don‟t miss the big picture.

InversionMode
Query inversion applies to queries which: a) are directed against a data source, and

b) though returning references to instances a single business object type, or a scalar simple type, must process other types in order to acquire the result. For example, the query “get me all Customers with Orders in the current year” will return references to Customer objects, but must first examine many Order objects in order to return the correct set of Customers. The query “give me the count of Customers located in Idaho” will return an integer, but must examine the Customer collection in the data source. Query inversion is the process of retrieving those non-targeted objects that are nonetheless necessary for correct completion of a query. The most fundamental reason for doing query inversion is so that the query can be applied against a pool of data that combines unpersisted local data with data that exists in the datasource. This is, after all, what your end user normally wants: query results based on the state of the data as she has modified it. The only place that combined pool of data can exist, prior to persisting changes, is the local cache. Therefore the query must ultimately be applied against the cache; and that operation, if it is to return correct results, requires the cache to contain all entities that must be examined in the course of satisfying the query. So to satisfy the query “get me all Customers with Orders in the current year”, the cache must contain not only the Customers to which references will be returned, but also all extant current-year Orders, so we can know which Customers those are. A handy side-effect of inverting queries is that the same query, if resubmitted during the same application session, can be satisfied entirely from the cache, without requiring another trip to the datasource. Another results from the fact that there is a reasonably good statistical chance that the related objects needed for satisfaction of the query will also be referenced in other ways by the application. In this very common scenario, the effect of the extra data retrieved is to improve client-side performance by eliminating the need for separate retrieval of the related objects.

IdeaBlade DevForce

Business Object Persistence

Note that the end result of a query inversion process is very similar to that which occurs when the .Include() method is used in a query. Both processes result in the retrieval and local storage of objects that are related to a set of root objects that are the primary target of a particular query. Four InversionModes are available in DevForce for a query:
Table 5. InversionModes

Strategy
On

Implicit Instuctions to DevForce Attempt to retrieve, from the datasource and into the cache, entities other than the targetted type which are needed for correct processing of the query. If this attempt fails, throw an exception. Do not attempt to retrieve entities other than the targetted type into the cache. Attempt to retrieve, from the datasource and into the cache, all entities other than the targetted type which are needed for correct processing of the query. However, if this attempt fails, just retrieve the entities of the directly targetted type, and do not throw an exception. Don‟t attempt to invert the current query; but act as if it were successfully inverted (if it needed to be). You (the developer) should only use this InversionMode when you are prepared to guarantee, on your own, that the entity cache contains (or will contain, after the DataSource portion of the query operation) all the necessary related objects to return a correct result if submitted against the cache. Normally you would make good on this guarantee by performing other data retrieval operations (prior to the one in question) to retrieve the necessary related data; or by including calls to the Include() extension method in the current query, sufficient to retrieve the necessary related data.

Off Try

Manual

The default InversionMode is Try, and this will likely be your choice for most queries. You should use On only if your application absolutely depends upon the related entities being brought into the cache by your query, and you should include exception handling in case the strategy fails. Choose the Off setting if you only want the targeted entries retrieved into the cache. Be sure you choose a compatible FetchStrategy. For queries that DevForce can successfully invert, the InversionModes of Try and On will yield the same end state: the query will be cached, and all related objects necessary to permit future satisfaction of the query entirely from the cache will be assumed to be present in the cache. If you use the InversionMode of Manual properly – that is, you take care to see that the necessary related objects get retrieved into the cache by some means or another before the query is submitted – then it, too, will produce the same ending state as the Try and On settings.

Queries That Cannot Be Inverted
The following types of queries cannot be inverted:

IdeaBlade DevForce

Business Object Persistence

A query that returns a scalar result. This includes all aggregate queries (Count, Sum, Avg, etc.). 39

C# VB

var query02 = _em1.Orders.Select(o => o.FreightCost).Sum();

Dim query02 = _em1.Orders.Select(Function(o) o.FreightCost).Sum()

A query whose return type is a single element. These include queries that call .First(), .Last(), and .Single()

C# VB

var query03 = _em1.Products.OrderByDescending(c => c.ProductName).FirstOrNullEntity();

Dim query03 = _em1.Products.OrderByDescending(Function(c) c.ProductName) _ .FirstOrNullEntity()

A query whose return type is different from the type contained in the collection first referenced.

C#

var query04 = _em1.Customers .Where(c => c.Country == "Argentina") .SelectMany(c => c.Orders);

VB

Dim query04 = _em1.Customers _ .Where(Function(c) c.Country = "Argentina") _ .SelectMany(Function(c) c.Orders)

” much later in this chapter. Again, we suggest you defer reading that at least until you‟ve completed this section on Query Strategy.

Pre-Defined QueryStrategies
As mentioned previously, every QueryStrategy combines a FetchStrategy, a MergeStrategy, and a InversionMode. Since there are five FetchStrategies, five MergeStrategies, and four InversionModes, there are potentially 100 versions of QueryStrategy, even keeping the TransactionSettings constant. However, in practice, a much smaller set of QueryStrategies suffices for the great majority of purposes. DevForce has identified five of them as being of particular significance, enshrining them as static (Shared in VB) properties of the QueryStrategy class. These predefined QueryStrategies combine FetchStrategy, MergeStrategy, and InversionMode strategies as shown in Table 6.

39

Note that this group includes the example mentioned earlier in this discussion: “Give me the count of Customers located in Idaho.”

IdeaBlade DevForce
Table 6. Fetch and merge strategies of the common query strategies

Business Object Persistence

Query Strategy
Normal CacheOnly DataSourceOnly DataSourceThenCache DataSourceOnlyWithQueryInversion

Fetch Strategy
Optimized CacheOnly DataSourceOnly DataSourceThenCache DataSourceAndCache

Merge Strategy
PreserveChanges (Not Applicable) OverwriteChanges OverwriteChanges OverwriteChanges

InversionMode
Try (Not Applicable) Off Try On

Here‟s how you assign a pre-defined QueryStrategy:

C# VB

query04.QueryStrategy = QueryStrategy.DataSourceThenCache;

query04.QueryStrategy = QueryStrategy.DataSourceThenCache

Custom QueryStrategies
As just noted, only five of the possible combinations of a FetchStrategy and a MergeStrategy are covered by the named QueryStrategies. What if you want one of the other combinations? You can create your own QueryStrategy by supplying the fetch and merge strategy enumerations to its constructor. The result is a new immutable QueryStrategy instance40.

40

Immutable meaning that we can get the component fetch and merge strategies but we cannot reset them.

IdeaBlade DevForce
Here‟s an example of the creation and assignment of a custom QueryStrategy:

Business Object Persistence

C#

QueryStrategy aQueryStrategy = new QueryStrategy(FetchStrategy.DataSourceThenCache, MergeStrategy.PreserveChanges, QueryInversionMode.On);

VB

' Creating a custom QueryStrategy Dim aQueryStrategy As New QueryStrategy(FetchStrategy.DataSourceThenCache, _ MergeStrategy.PreserveChanges, _ QueryInversionMode.On)

DefaultQueryStrategy
We mentioned earlier that the DevForce EntityManager has a DefaultQueryStrategy property that can be used to shape the fetch and merge behavior of queries where the QueryStrategy is not explicitly specified. The default setting for the EntityManager‟s DefaultQueryStrategy is QueryStrategy.Normal. If you leave this setting at its default value, and in an individual query do nothing to countermand the default settings, then the FetchStrategy of Optimized will be used in combination with the MergeStrategy of PreserveChanges. If for some reason you wanted a EntityManager where the default QueryStrategy would always involve a trip to the data source, you could assign a different QueryStrategy, such as DataSourceOnly, to the PM‟s DefaultQueryStrategy property. For a given query, you could still use any desired QueryStrategy by explicitly specifying a different one.

When to Use The Different QueryStrategies
For most users, most of the time, the DevForce defaults are perfect:   Satisfy a query from the entity cache whenever possible; When a trip to the data source is found necessary, resolve any conflicts that occur between incoming data and data already cache by giving the local version priority; and Perform query inversion as needed; if needed and undoable, revert to a DataSourceOnly FetchStrategy.

Your choice of a non-default strategy can be driven by a variety of things. For example, suppose your application supports online concert ticket sales. Your sales clerks need absolutely up-to-date information about what seats are available at the time they make a sale. In that use case, it will be essential to direct your query for available seats against the data source, so a FetchStrategy of DataSourceOnly might be in order. In code to handle concurrency conflicts, one might need a QueryStrategy with a MergeStrategy of PreserveChangesUpdateOriginal to make an entity in conflict savable. (The data source version of the conflicted entity would only be retrieved and used to partially overwrite the cache version after the concurrency conflict had been resolved by some predetermined strategy.) You can and will think of your own reasons to use different combinations of FetchStrategy, MergeStrategy, and InversionMode. Just ask yourself, for a given data retrieval operation, whether the data in the cache is good enough, or you need absolutely current data from the data source. Then ask yourself how you want to resolve conflicts between data already cached and duplicate incoming data. Then consider the process DevForce will use to satisfy the query and make sure it will have the data it needs to give you a correct result. DevForce gives you the flexibility to set the behavior exactly as need it.

IdeaBlade DevForce

Business Object Persistence

Making a One-Time Change to the QueryStrategy With Which a Given Query Is Run
You may find yourself with an existing IEntityQuery object that you don‟t want to disturb in any way, but which you would like to run with a different QueryStrategy for a specific, one-time purpose. DevForce provides an extension method, With(), that permits you to do this. 41 When a call to With() is chained to a query, the result may be either a new query or a reference to the original query. Normally it will be a new query, but if the content of the With() call is such that the resultant query would be the same as the original one, a reference to the original query is returned instead of a new query. If you ever want to be sure that you get a new query, use the Clone() extension method instead of With(). With() avoids the overhead of a Clone() when a copy is unnecessary.
Code Snippet 42. QueryStrategyWithAndCloning

C#

IEntityQuery<Customer> query00 = _em1.Customers .Where(c => c.CompanyName.ToLower().StartsWith("a")); query00.QueryStrategy = QueryStrategy.DataSourceOnly; // The With() call in the right-hand side of the following statement // specifies a query that is materially different from query0, in // that it has a different QueryStrategy associated with it. // Accordingly, the right-hand side of the statement will return // a new query: IEntityQuery<Customer> query01 = query00.With(QueryStrategy.CacheOnly); // Because the content of the With() call in the right-hand side // of the following statement doesn't result in a modification // of query0, the right-hand side will return a reference to // query0 rather than a new query. IEntityQuery<Customer> query02 = query00.With(QueryStrategy.DataSourceOnly); // If you want to be certain you get a new query, use Clone() // rather than With(): EntityQuery<Customer> query03 = (EntityQuery<Customer>)query00.Clone(); query03.QueryStrategy = QueryStrategy.DataSourceOnly;

VB

Dim query00 As IEntityQuery(Of Customer) = _em1.Customers.Where(Function(c) c.CompanyName.ToLower().StartsWith("a")) query00.QueryStrategy = QueryStrategy.DataSourceOnly ' The With() call in the right-hand side of the following statement ' specifies a query that is materially different from query0, in ' that it has a different QueryStrategy associated with it. ' Accordingly, the right-hand side of the statement will return ' a new query: Dim query01 As IEntityQuery(Of Customer) = query00.With(QueryStrategy.CacheOnly) ' Because the content of the With() call in the right-hand side ' of the following statement doesn't result in a modification ' of query0, the right-hand side will return a reference to ' query0 rather than a new query. Dim query02 As IEntityQuery(Of Customer) = query00.With(QueryStrategy.DataSourceOnly) ' If you want to be certain you get a new query, use Clone() ' rather than With():

41

Our topic here is QueryStrategy, but in fact some overloads of the With() method also (or alternatively) permit you to make a one-time change to the EntityManager against which the query will be run.

IdeaBlade DevForce

Business Object Persistence

Dim query03 As EntityQuery(Of Customer) = CType(query00.Clone(), EntityQuery(Of Customer)) query03.QueryStrategy = QueryStrategy.DataSourceOnly

Span Queries
A EntityManager query method always returns entities of a single type, the return type identified in the query object. But what about entities related to the returned entities? When do we get those? Consider a query for second quarter orders. We display them in a grid with their customer names and order totals.

The Order entities entered the cache when we processed the query. Not so the Customer and the OrderDetail entities that we need to calculate the order total. The EntityManager gets these entities only when we ask for them explicitly. Such delayed fetching we called deferred retrieval. The grid control binding calls an Order property each time it fills a cell. The “Customer” and “Order Total” columns are bound to two properties that resolve to two relation queries, one for Customer entities and one for OrderDetail entities. This means the grid control invokes two relation queries for each and every row. There are three rows showing in the screen shot so there will be six queries, each one requiring a round trip to the data source. In other words, filling this grid requires six trips to the data source. Now suppose that we had an excellent quarter and placed a thousand orders. The user clicks the “Customer” column caption, causing the grid to sort by customer. The sort requires examination of every one of those thousand orders. Most grids will fire every visible property on every examined row. That could mean two thousand separate trips to the server: one thousand fetches of customers and one thousand fetches of order details. The UI will stall for ten uncomfortable seconds and then return to its familiar crisp responsiveness. Subsequent sorts and scrolling are fast; all of the entities are now in cache so there are no trips to the data source42. But those ten seconds felt like an eternity. The problem wasn‟t the ten seconds; it‟s that they occurred when the user thought they should not. She expected the search for orders take some time; maybe not ten seconds but she expected a pause of some length. On the other hand, she expected the sort to happen immediately. When it didn‟t, she thought there was something wrong with the application. Is the sort delay necessary? Of course not! The program cannot anticipate needing the related data and so it fetches entities inefficiently. We know better. When we grab the thousand orders, we can fetch their customers and order details at the same time. Not every Customer in the data source. Not every OrderDetail entity either. We only need the customer and order details that are related to those thousand second quarter orders. We should get them all at once, not piecemeal as we scroll or sort the grid. Span queries to the rescue. We can add span instructions to our query so that the EntityManager gets the related entities when it gets the orders. A span query instruction describes a path along the root entity graph to a particular entity type know as the span target.
42

The volume of data is not the issue. We might think that we‟d improve performance if we used a view that summed the OrderDetails on the server. We‟d get one value per row instead of having to bring down the details and sum them locally. When we try this, we observe no improvement whatsoever. The delays were due entirely to the round-tripping, not the data volume nor the summations.

IdeaBlade DevForce

Business Object Persistence

Of course, a EntityManager returns references to the root objects when it executes a span query. At the same time it fetches every span target entity related to any of the returned root entities and puts them in the cache. We‟ll need two of spans for our example. There is a simple syntax for spanning to the immediate neighbors of the query‟s result entity type:

C#

var query = _em1.Customers .Include("Orders")

VB

Dim query = _em1.Customers _ .Include("Orders")

We can span to entities farther away on the Order business object graph also as we might do if we were displaying product name in the Order‟s OrderDetails grid.
Code Snippet 43. NavigationSynchronousPreload (repeated)

C#

var query = _em1.Customers .Include("Orders") .Include("Orders.OrderDetails"); .Include("Orders.OrderDetails.Product"); .Include("Orders.OrderDetails.Product.Supplier"); .Include("Orders.SalesRep")

VB

Dim query = _em1.Customers _ .Include("Orders") _ .Include("Orders.OrderDetails") _ .Include("Orders.OrderDetails.Product") _ .Include("Orders.OrderDetails.Product.Supplier") _ .Include("Orders.SalesRep")

Again, span queries don‟t change the list of entities to which references are returned from the query. The caller still receives the same thousand orders. But before returning the orders, the span query processing fetches the related entities and merges them into the cache. When the grid cells call upon Order properties to return customers or calculated order totals, those properties will find the pertinent entities waiting in cache. The main order query is a little slower because there are more entities retrieved. The user won‟t notice; she expected the search to take a beat or two. The first sort is instantaneous; she is thrilled.

Performance Details
While spans greatly reduce the number of queries submitted to the database, they do not, of course, eliminate them altogether. Each span resolves to a separate query and each of these span queries necessitates a separate trip to the database. Thus, if our we added three spans to an Order query, there would be four queries (one for the Orders, one for the related type referenced in each of the spans) and four trips to the database. But these four trips -- as our previous discussion has illustrated – might well replace thousands of trips required in the absence of spans. In an n-tier deployment using the Business Object Server (BOS), the picture is even rosier. In that configuration, the client submits the entire request, including spans, in a single transmission to the BOS. It is the BOS that makes the four trips to the database. When the BOS has a fast, fat pipe to the database - as it should – those four trips are very quick indeed. The BOS then combines the results from its queries against the database into a single package that it ships back to the client. There has been only one trip across the “slow” connection between client and server!

IdeaBlade DevForce

Business Object Persistence

Note also that the total loads on the EntityServer and database are reduced when each client is making efficient data requests using spans. Thus, every individual client benefits from the improved efficiency of the other clients. Performance matters ... but not all time and effort spent optimizing performance returns equal results. We strongly advise instrumenting your queries during development and testing to identify performance hotspots. Then optimize where it really matters.

Cached Entity Lifespan
Entities stay in the cache until the application terminates or they are removed. There is no garbage collection. We may need to purge the cache of unwanted entities if    we accumulate a large volume of entities during a user session a session might last a long time – days, weeks, etc. the cache contents will be saved and later restored from local storage.

The programmer has many “remove” options including the ability to remove a single entity, a list of entities, entities of a particular type, and all entities with a specified EntityState. “Removal” and “deletion” are not the same thing. “Remove” means “remove the entity from the cache.” There are no data source implications. The entity is simply no longer in the cache; it is as if we had never fetched it. “Delete” means “schedule the entity for deletion from the data source.” The entity remains hidden in the cache, waiting for the moment when we send a delete request to the data source. That moment arrives when we “save” the deleted entity. Once saved (that is, deleted from the data source), the object is removed from the cache. New entities are removed from the cache immediately when deleted; they were never in the data source so there is nothing there to delete, nothing to schedule.

Saving the Cache Locally
An EntityManager can save its cache locally. This feature is useful in many scenarios including these two:   The application must be able to run offline for extended periods. It must be possible to exit the application and launch it again later while still disconnected. The developer is worried that the user may accumulate many changes for a long time without saving to the data source. The application would snapshot the changes periodically in case the application goes down. But many of the modified business objects won‟t pass data source validity checks or won‟t satisfy business rules for permanent business objects. They can‟t be saved to the data source.

In the first case the application can‟t reach the data source and in the second its access is blocked. The application needs a local option. The application can tell the EntityManager to serialize its object cache as an XML stream and save the stream to a file on the client‟s file system. Variations on the theme enable encryption of the stream and filing to isolated storage or other arbitrary destinations. On command or when the application is re-launched, the application can locate the file and restore its contents to the EntityManager‟s cache. The developer can choose to completely replace the target cache or merge the saved cache objects into it; in a merge, objects from the saved cache replace corresponding objects in the target cache. The pool of temporary ids maintained by the developer‟s custom implementation of IIdGenerator is also saved and restored. The process preserves pending business object changes – additions, modifications, deletes. When the application next obtains a server connection, it can synchronize local objects with the central data source. It can refresh local

IdeaBlade DevForce

Business Object Persistence

unmodified copies of business objects that have been changed by other users. It can save local pending changes, relying upon DevForce optimistic concurrency checking to prevent overwriting other users‟ changes. If the developer expects the application to operate offline, she should prep the cache by retrieving the business objects the user is likely to need before disconnecting and saving the cache locally. While disconnected, queries and object navigation can only access objects already in cache.

The TraceViewer: Watch What Data Is Being Loaded, and How
Sometimes you may not be aware of what data is being loaded during particular processes. In this, the DevForce TraceViewer can be extremely helpful. It monitors all communications with the business object server, providing a real-time log of same. There are two different ways to use the Trace Viewer:   Stand-alone, and Embedded in your application.

To use the Trace Viewer in stand-alone mode, you will typically launch it from the Windows Start Menu for DevForce:

You can use the Trace Viewer in this mode with no change to your application code, but only if run your application in n-tier mode, with the Business Object Server running in a separate process from the client application. You can also use the stand-alone Trace Viewer without running n-tier if you are willing to add a single line of code to your application. Embedding the Trace Viewer in your application requires a couple of minor (and isolated) changes to your application code, but offers greater convenience – you can set it to begin working automatically whenever you start the app – and it does not require that the Business Object Server be launched in a separate process. For our own development work, in non-release versions of our applications, we often use the Trace Viewer this way. We‟ll detail both approaches in the following material.

Using the Trace Viewer Stand-Alone
To use the Trace Viewer stand-alone, launch it from the Windows Start Menu entry shown in the screen shot above. It will display a dialog window like the following:

IdeaBlade DevForce

Business Object Persistence

Once launched, the Trace Viewer makes periodic attempts to connect with a TracePublisher. It will find a Business Object Server instance once one is running.

Using the Stand-Alone Trace Viewer While Running N-Tier
To see the server activity instigated by your application without making any code changes, you‟ll need to launch the app in n-tier mode. You can do that easily, on a single development machine, using the NTier Configuration Starter utility which you will also find on the IdeaBlade DevForce / Tools menu. You can find step-by-step instructions for working with the N-Tier Configuration Starter in the Deployment topic document, in the section “N-Tier Configuration Starter”.

Once your application in running n-tier, you‟ll see communications with the BOS logged as follows:

IdeaBlade DevForce

Business Object Persistence

The activity logged just above resulted from execution of the following method in an app: C#
private void DoIt() { _mgr.Customers.ToList(); _mgr.Orders.ToList(); _mgr.Products.ToList(); _mgr.Suppliers.ToList(); _mgr.Employees.ToList(); }

VB

Private Sub DoIt() _mgr.Customers.ToList() _mgr.Orders.ToList() _mgr.Products.ToList() _mgr.Suppliers.ToList() _mgr.Employees.ToList() End Sub

The method simply fires off five queries that must hit the server to get their data.

Using the Stand-Alone Trace Viewer While Running “Single-Tier” (Client and Server in the Same Process)
To see activity in the stand-alone Trace Viewer when running in single-tier, development mode, you must add one line of code to your application:

C#

TracePublisher.LocalInstance.MakeRemotable();

IdeaBlade DevForce
VB
TracePublisher.LocalInstance.MakeRemotable()

Business Object Persistence

Once your app has made the above call to MakeRemotable(), it begins functioning as a TracePublisher, doing so on a default port and default service name that matches the defaults on the Trace Viewer. You can also make it a publisher on a different port and with a different service name, but then you will need to change the settings on the Trace Viewer to listen on the specified channel.

C# VB

TracePublisher.LocalInstance.MakeRemotable(9010, "MyClientService");

TracePublisher.LocalInstance.MakeRemotable(9010, " MyClientService")

For most uses, you probably won‟t find it necessary to change the port or service name. Note that when a DevForce app is deployed n-tier, separate sets of messages are published server-side and clientside. (These messages end up in the server- and client-side debug logs, as well as in any Trace Viewers that are listening for them.) When running single-tier, messages written by the (logically server-side) EntityService (which in single-tier mode runs inside the same process as the client application) will be published along with messages from the logical client-side. You‟ll see everything. Here are the messages captured by the stand-alone Trace Viewer after adding the MakeRemotable() call and running single-tier:

Note that including the call to MakeRemotable() and running the stand-alone Trace Viewer is the only way to use the Trace Viewer with a (single-tier) console app. The options for embedding the TraceViewer (described below) require WPF or WinForm applications.

Embedding the Trace Viewer in Your Application
For convenience during development, you may prefer to embed the Trace Viewer in your application. This requires adding a reference to the TraceViewer executable and adding a line or two of code.

IdeaBlade DevForce

Business Object Persistence

There are actually two different implementations of the TraceViewer within DevForce: one for WinForm apps and one for WPF apps. The names of their executables are as shown below:

Target Application Type WinForm App WPF App

Executable WinTraceViewer.exe WPFTraceViewer.exe

To use either TraceViewer in your app, you must first set a reference (in your app‟s UI project) to the executable file where it lives. Both versions of the TraceViewer are deployed to the DevForce installation folder, usually C:\Program Files\IdeaBlade DevForce.

Embedding the WPFTraceViewer in Your WPF App
To add the reference to the WPF TraceViewer, for example, right-click the references node in your desired UI project, and select Add Reference. On the Add Reference dialog, select the Browse tab, then browse to the file and click OK:

Here‟s some code for the startup window of a simple WPF app. The code launches the WPF TraceViewer during initialization, and includes a button click handler that launches a query for some data:

IdeaBlade DevForce
C#
namespace Wpf01 { /// <summary> /// Interaction logic for Window1.xaml /// </summary> public partial class Window1 : Window { public Window1() { InitializeComponent(); SetUpTraceViewer(); } private void SetUpTraceViewer() { WPFTraceViewer tv = new WPFTraceViewer(); tv.Show(); }

Business Object Persistence

private void _loadDataButton_Click(object sender, RoutedEventArgs e) { List<Customer> customers = _em1.Customers.Where(c => c.Country == "Brazil").ToList(); _outputTextBlock.Text += String.Format("Customer retrieved: {0}\n", customers.Count); } #region Private Fields DomainModelEntityManager _em1 = new DomainModelEntityManager(); #endregion Private Fields } }

VB

Public Class Program Public Shared Sub main() #If DEBUG Then Dim tv As New IdeaBlade.DevTools.TraceViewer.TraceViewerForm() tv.Show() #end If Application.Run(MainForm) End Sub End Class

Here is the display that results:

IdeaBlade DevForce

Business Object Persistence

The TraceViewer logs all operations against the Entity Server, so you can use it to see exactly what data loading operations result from actions performed in the user interface. In this app, additional clicks of the <Load Data> button result in no further activity against the Entity Server, since the desired Customers, once retrieved into the cache, can be accessed there thenceforward.

Embedding the WinTraceViewer in Your WinForms App
To add the reference to the WinTraceViewer, right-click the references node in your desired UI project, and select Add Reference. On the Add Reference dialog, select the Browse tab, then browse to the file and click OK:

IdeaBlade DevForce

Business Object Persistence

Here‟s some code for the startup program of a simple WinForm app that launches the WinForms TraceViewer during initialization:

IdeaBlade DevForce
C#
static class Program { [STAThread] static void Main() { Application.EnableVisualStyles(); Application.SetCompatibleTextRenderingDefault(false);

Business Object Persistence

IdeaBlade.DevTools.TraceViewer.TraceViewerForm tv = new IdeaBlade.DevTools.TraceViewer.TraceViewerForm(); tv.Show();
Application.Run(new _customerForm()); } }

VB

The main method, after launching the TraceViewer, launches _customerForm as the startup form:

IdeaBlade DevForce

Business Object Persistence

The handler for the button‟s click event launches a query for some Customers:

C#

private void _loadDataButton_Click(object sender, EventArgs e) { _customers.ReplaceRange(_em1.Customers .Where(c => c.Country == "Brazil")); }

VB When you click the button, you see activity logged in the TraceViewer:

The TraceViewer logs all operations against the Entity Server, so you can use it to see exactly what data loading operations result from actions performed in the user interface. In this app, additional clicks of the <Load Data> button result in no further activity against the Entity Server, since the desired Customers, once retrieved into the cache, can be accessed there thenceforward.

IdeaBlade DevForce Getting Generated SQL to Display in the TraceViewer

Business Object Persistence

By default, both TraceViewers (WPF and WinForms) show queries in a LINQ-like representation:

The above, for example, is an unrestricted query for entities of type Employee. You can, however, elect to see the SQL generated server-side by the Entity Framework. To do that, you must change the logTraceString setting in the applicable app.config file43 to true. Note that logTraceString is an attribute of a particular edmKey (which represents a single data source).

This results in a display like the following:

The TraceViewer can be invaluable in troubleshooting performance problems. These are often caused by inefficient data retrieval (such as loading a data grid where each rows triggers several trips to the server to pick up related objects that were not pre-loaded).

Using the Trace Viewer With a Silverlight App
For Silverlight applications, the EntityService automatically runs in a separate process from the client, so a standalone TraceViewer will automatically pick up the server messages associated with your app (assuming you haven‟t changed the default service name and port). If you wish to see client-side Trace messages, you will need to embed a UserControl into your Silverlight front end. Here is the XAML for the control...

43

In a development app with all parts running on a single machine, choose the App.Config file in the AppHelper project.

IdeaBlade DevForce
XAML

Business Object Persistence

<UserControl x:Class="DevForceSilverlightApp.TraceWindow" xmlns="http://schemas.microsoft.com/winfx/2006/xaml/presentation" xmlns:x="http://schemas.microsoft.com/winfx/2006/xaml" xmlns:data="clrnamespace:System.Windows.Controls;assembly=System.Windows.Controls.Data" Width="Auto" Height="Auto"> <Grid x:Name="LayoutRoot" Margin="20,20,20,20" > <data:DataGrid x:Name="_dataGrid" HorizontalAlignment="Left" VerticalAlignment="Top" AutoGenerateColumns="True" MinWidth="250" MinHeight="100" Background="#FFB5BAB5" Margin="0,0,20,0" IsReadOnly="True" /> </Grid> </UserControl>

...and here is the code behind:

C#

using using using using

System; System.Collections.ObjectModel; System.Windows.Controls; IdeaBlade.Core;

namespace DevForceSilverlightApp { /// <summary> /// Sample trace subscriber. You can drop the TraceViewer UserControl onto a page /// to display tracing information from the Silverlight application in a grid. /// </summary> /// <remarks> /// To use the TraceSubscriber: 1) listen for its Publish event, and 2) call StartSubscription() /// to have tracing messages sent to you. You can also call StopSubscription() /// to temporarily or permanently stop receiving messages. /// </remarks> public partial class TraceWindow : UserControl { public TraceWindow() { InitializeComponent(); _messages = new ObservableCollection<TraceMessage>(); _subscriber = new TraceSubscriber(); _subscriber.Publish += new EventHandler<PublishEventArgs>(_subscriber_Publish); _subscriber.StartSubscription(); _dataGrid.ItemsSource = _messages; } private void _subscriber_Publish(object sender, PublishEventArgs e) { _messages.Add(e.TraceMessage); if (_dataGrid.Columns.Count > 0) {

IdeaBlade DevForce
} } TraceSubscriber _subscriber; ObservableCollection<TraceMessage> _messages; } }

Business Object Persistence

_dataGrid.ScrollIntoView(e.TraceMessage, _dataGrid.Columns[0]);

VB

Be sure to change the namespace in both the XAML and the code to match your app! In the following, we have embedded the above TraceWindow UserControl in another UserControl:

XAML

<UserControl x:Class="DevForceSilverlightApp.ConsoleUserControl" xmlns="http://schemas.microsoft.com/winfx/2006/xaml/presentation" xmlns:x="http://schemas.microsoft.com/winfx/2006/xaml" xmlns:local="clr-namespace:DevForceSilverlightApp" > <Grid x:Name="LayoutRoot" Background="White"> [... snip] <ScrollViewer x:Name="_traceWindowScrollViewer" Grid.Row="1" Margin="0,0,20,0"> <local:TraceWindow /> </ScrollViewer> [... snip] </Grid> </UserControl>

That‟s enough to get it to display client-side trace messages written by DevForce. We can add our own trace messages as follows...

C# VB

IdeaBlade.Core.TraceFns.WriteLine("Hello world!");

IdeaBlade.Core.TraceFns.WriteLine("Hello world!")

IdeaBlade DevForce

Business Object Persistence

...resulting in the following output (the TraceWindow control contains the DataGrid):

Creating Business Objects
In this short chapter we discuss business object creation in a bit more detail. We‟ll explain why and when the developer must write her own creation method and what minimal steps are essential to its implementation. We delve into the special challenge of creating unique business object identifiers and how DevForce supports this process. We mention also two other custom class methods, CompareTo() and ToString(). We may want to add them while writing the creation method.

When Not to Create
A business object class needs a create method only if the application can add new business objects of its type. This is not so in a surprising number of cases. For example, we don‟t add states to the USA very often. Our application may want access to these states as business objects but it is unlikely to need to add new ones (or change existing states).

The Business Object Create Method
Most applications will add new instances to many of its business object classes. The developer must write a Create method for each of these classes and call it whenever she wants a new object.

IdeaBlade DevForce

Business Object Persistence

For technical reasons, we must acquire new instances via a class method rather than by means of a constructor. The expression emp = new Employee(…) is always invalid; instead it must look something like emp = Employee.Create(…). Most Create method implementations return a single business object after following these four steps: 1. 2. 3. 4. Ask the EntityManager for a prototype of the new business object Give the prototype a unique identity Fill in some of its initial values (optional) Add the completed prototype to the EntityManager„s cache

Why can‟t DevForce take care of this for us? Because steps 2 and 3 require application-specific know-how that DevForce can neither discover nor supply. Step #2 concerns the identity of the object. DevForce requires that every business object have a unique identity. Identity is captured in the object‟s primary key which is composed of one or more identifiers. There is no way for DevForce to know how identifiers are determined. While it can discover that a particular database table‟s key is a single integer field, this fact is insufficient to generate an identifier. The integer could come from anywhere. Step #3 concerns the validity of the object. It is generally a good idea to maintain an object in a valid state. This isn‟t always possible but it is a useful goal and the Create method is a place to start. Of course DevForce is ignorant of application business rules so if there is to be any object initialization it is up to the developer to code it here.

Generating unique identifiers
Unless the primary key is an Identity column, DevForce doesn‟t know how to generate an object‟s primary key identifier(s) so it cannot deliver new business objects on its own. That is why the EntityManager provides a prototype in Step #1 that is not yet in the cache. Once the developer sets the primary key‟s identifier(s) in the prototype, the prototype may be added to the EntityManager‟s cache and become a business object accessible to the application. It may still be invalid from a business perspective but it is programmatically acceptable to DevForce. A business object‟s key must be unique not only within the context of the current user session but across the application domain. We have to make sure the key we assign to a new employee object cannot also be assigned to a different employee object by someone else.

GUIDs
GUIDs (globally unique identifiers) make great identifiers (aka “ids”) because they are easy to mint, are nearly certain to be unique, and can be generated locally, independent of any external resource. If we are in complete control of the database schema design, GUIDs are the way to go. The MS SQL uniqueidentifier data type is the database analog for a GUID. When we need a new GUID, we ask .NET to compute one for us, assign it to the prototype‟s identifier data member, and move on to the next step in object creation. GUIDs have two disadvantages:   GUID values are long and obscure. Users find them difficult to type correctly and difficult to remember. At 16 bytes, the GUID is large compared to other data types such as 4-byte integers. Database indexes built using GUID keys may be relatively slower than indexes using an integer key.

IdeaBlade DevForce

Business Object Persistence

In our experience, striving for meaningful identifiers leads to disappointment and failure; we strongly council against using identifiers with semantic content. If you disagree, you may regard these additional GUID properties as disadvantageous:   GUID values are random and cannot accept any patterns that may make them more meaningful to users. There is no way to determine the sequence in which GUID values are generated. They are not suited for applications that depend on incrementing key values.

If GUIDs work for you, you may skip the next section on custom id generation. Unfortunately, few of us have this option. We are usually given a database that we cannot change. We‟re not allowed to replace all table ids and all foreign key columns with 16 byte integer GUIDs. We have to conform to the existing key schemes which impose both the identifier data types and the manner of their generation.

Sql Server Identity Ids
DevForce detects Sql Server Identity columns and generates ids automatically. Its approach is almost exactly the same as “custom id generation” which we‟ll discuss next. The key difference is that the id “seed” value – the source of the next available id – is maintained by Sql Server rather than in a custom id seed table (e.g., “NextId”). There are special consideration when using the DevForce Identity Id generator which are covered below.

Custom id generation
Custom id generation almost always requires access to some external resource, some application-specific logic for deriving new ids, and additional logic to increment the resource. Suppose our application uses integer keys for all of its tables. The database has a special “NextId” table that holds the next integer id. To get a new id, a server-side process could quickly lock that table, grab the id, update the table to hold a new next id, and free the table. This is just one among thousands of ways applications generate ids. The commonality is the external resource, the functional equivalent of the NextId table, without which we could not be sure of generating identifiers that are unique within the application domain. The developer must write the code that reads the resource, calculates ids, and updates the resource. If only it were this simple. Remember that we are describing a smart client application in which new object creation begins on the client machine. The client machine could be disconnected and thus unable to reach a NextId table or some other external source of permanent ids. We still want to be able to create new objects while disconnected. We know that we will have to connect to that external resource to get permanent ids and store the new objects in the database. In the interim, we must finesse the situation and use locally generated, temporary ids until we can reconnect and replace them with permanent ids. For example, since our permanent ids are always positive integers, we could use negative integers for temporary ids, acquiring them by decrementing a client-side counter. We assign temporary ids (however generated) to the new objects and to the foreign keys of the objects that reference them. At some point when we‟re sure we‟re connected, we run around to all the locations with temporary ids and replace them with permanent ids.

Id Fix-up
Just before we save objects back to the database is a good time to attempt this fix-up because (a) we must be connected to save and (b) we must fix all locally modified objects before saving any of them in case one such object has a reference to a temporary id.

IdeaBlade DevForce
The DevForce id generation facility can help. In essence:   

Business Object Persistence

The developer writes an id generation class that conforms to the DevForce IIdGenerator interface. This class will handle id generation for every class of business object in the data source. The developer implements the prescribed methods in the id generation class that provide temporary and permanent ids. Back in the business object creation method, the developer invokes the EntityManager.GenerateId(…) method which assigns a temporary id to the new object prototype. A typical call looks like: pm.GenerateId(protoEmp, Employee.IdEntityColumn). The EntityManager attempts a save The DevForce framework tells the developer‟s id generation class to give it the map of temporary ids to permanent ids. The framework runs around the cache, replacing temporary ids with permanent ids.

  

Foreign Key Fix-Up
The framework replaces temporary ids in entity properties that are connected to the generated id column by a relation. The generated Id column in this example is the Employee.IdEntityColumn. Suppose there is a relation defined in the model between Order.SalesRepId and EmployeeId, but that no relation is defined between Customer.SalesRepId and EmployeeId. This is a critical omission, as you will see. We create a new employee, myEmployee. The Id generator gives him a temporary id value of –1. During the application session myEmployee is assigned to myOrder:
myOrder.SalesRep = myEmployee

But no relation was defined between Employee and Customer, so there is no Customer.SalesRep property44 to which myEmployee can be assigned directly. Nevertheless, the determined developer stuffs the EmployeeId value directly into the Customer.SalesRepId property.
myCustomer.SalesRepId = myEmployee.Id

This is a bad practice and should be avoided. The absence of the myCustomer.SalesRep property should have been a warning that a critical relation was missing. See what happens: The user saves and the fix-up begins.    The value of myEmployee.Id is updated to its permanent value, 301.
myOrder.SalesRepId is fixed up to 301 (since there is a relation back to Employee.Id) . myCustomer.SalesRepId stays stuck with id = –1. (There was no relation from myCustomer.SalesRepId back to Employee.Id so the PM didn‟t know to replace the SalesRepId.)

Not good! In most cases the end result of all this would be an errant foreign key value persisted to the data source. If, however, the data source did have the necessary foreign key constraint (but the related relation had been deleted from the model), the result of attempting to persist the errant foreign key value would be a foreign key constraint exception. That might appear to reflect a PersistenceOrder problem (e.g., saving a child before saving its new parent) when in fact it is not. Important: Map all of the relations.

44

At least, there would be no such property generated by the DevForce Object Mapper

IdeaBlade DevForce
Sample Id Generator

Business Object Persistence

DevForce ships with source for example id generator classes that you can either use directly or adapt for your application. It‟s now easy to see why we prefer GUIDs. We can use .NET‟s free GUID generator while disconnected because it works locally without resort to an external resource. GUIDs are globally unique so the ids we create are fine as permanent ids. All of the complexity disappears. The 16-byte cost of GUIDs is usually worth it. Use GUIDs if you can.

Ids in mapping objects
We cannot leave this subject without observing that some business objects do not use generated ids. Mapping objects relate one kind of business object to another in a many-to-many relationship. OrderDetails in the IdeaBladeTutorial database is one such business object. In addition to carrying information about a particular purchase item such as quantity and price, it relates Orders to Products in a many-to-many relationship. Orders have many items each associated with a particular product being purchased. A given product will appear as a purchased item on many different orders. A mapping object‟s primary key is typically a combination of the ids from the two objects it relates. The key of an OrderDetail object is comprised of its parent order‟s id and the id of the product being sold. It is an {OrderId, ProductId} tuple. In such cases, the ids that form the primary key are not generated within the create method but rather passed into the method in its parameter list. The create method for OrderDetail would include an order object and a product object among its parameters. Inside the Create() method, we would extract their ids and set the OrderDetail prototype‟s primary key accordingly. OrderDetail happens also to be the detail object in a master/detail relationship. The id of the master object is often one of the identifiers in the detail object‟s key and it will usually be passed into the method in one of its parameters.

Creating a valid business object
EntityManager delivers a prototype in step #1 of the create method. DevForce ensures that the prototype has a nonnull value for every object member that is mapped to a non-nullable field in the database. This “assistance” is often helpful but it may be wrong. Suppose business rules demand that every employee have a hire date and that hire dates must be later than the company‟s incorporation. The HireDate field in the database is mandatory so the prototype carries a default value for the corresponding object data member. The developer should make no assumption about this value other than that it is a valid date from the perspective of the database. It could be anything and might well be a date prior to the founding of the company. The hire date is probably unknown when the object is created. The developer may choose to wait until it becomes known in which case the prototype default value will suffice for awhile. The object can‟t be saved but we will have time to get a valid hire date from the user before we save it. Alternatively, the developer may decide that a particular date, such as today‟s date, makes a good initial hire date. She will initialize the prototype‟s hire date accordingly, here in the Create method. The lesson: strive to make the new object as valid as possible by setting appropriate initial values in this step of the creation method. It is often helpful to add parameters to the Create method so the caller can pass in appropriate initial values. None of this is required but it is good practice.

IdeaBlade DevForce

Business Object Persistence

Auxiliary Business Object Class Methods
While we‟re adding a new object creation method inside the business object class, it‟s a good time to mention two other useful methods: CompareTo() and ToString().

CompareTo()
When DevForce sorts a collection of business objects it often looks to the class CompareTo() method to determine which of two objects sorts before the other. Business objects inherit a CompareTo() method from the root class of all business objects, Entity. It‟s rarely what we want; the results are arbitrary and unpredictable. We should override it with a comparison that is useful. A CompareTo()for the Employee class might compare employee first and last names.

ToString()
It is common for both DevForce and .NET to invoke an object‟s ToString() method. An object‟s default ToString() returns the object‟s class name. This is rarely useful. For example, anEmployee.ToString() might return “Tutorial.Entities.Employee”. We should override the Employee ToString() method so that it returns something useful like “Nancy Davolio”. Many classes, not just business object classes, should have their own ToString() methods.

Adding and Removing Related Objects using Add() and Remove()
Navigation properties that return a collection (e.g., anEmployee.Orders) have Add() and Remove() methods.

Add()
The Add() method takes a parameter of the type contained by the collection (e.g., an Order).
Code Snippet 44. AddUsingAdd()

C#

Order anOrder = new Order(); anOrder.OrderDate = DateTime.Today; anOrder.FreightCost = Convert.ToDecimal(999.99); anEmployee.Orders.Add(anOrder);

Dim anOrder As New Order() anOrder.OrderDate = Date.Today anOrder.FreightCost = Convert.ToDecimal(999.99) anEmployee.Orders.Add(anOrder)

Invoking Add() adds the supplied item to the collection. If the relation between the parent and child types is 1-tomany and the supplied item is currently associated with a different parent, then Add() simultaneously removes it from the corresponding collection of the other parent. 45

45

The equivalent result on table rows in a relational database is that the child entity‟s foreign key value is changed.

IdeaBlade DevForce

Business Object Persistence

Note that, in the above snippet, we did not need to set the SalesRep property of the new Order to the Employee whom we wanted to become its parent:

C# VB

// anOrder.SalesRep = anEmployee;

//don't need this; Add() will handle it

' anOrder.SalesRep = anEmployee

'' don't need this; Add() will handle it

Invocation of the Add() method on anEmployee.Orders produced the equivalent result.

Remove ()
Remove() also takes a parameter of the type contained by the collection. It dissociates the indicated instance from the collection‟s parent46.

C# VB

anEmployee.Orders.Remove(anOrder);

anEmployee.Orders.Remove(anOrder)

Note that while Remove unassigns the Order from the target Employee, removing it from the collection returned by the navigation property, it does not remove it from the cache or mark it for deletion. If you want the Order removed from the cache or deleted from the back-end datastore, you must order those actions separately by calling the Order‟s EntityAspect.Remove() or EntityAspect.Delete() methods, as appropriate.

Add() and Remove () on Many-to-Many Navigation Properties
You can also use Add() and Remove () on many-to-many navigation collections generated by the Entity Data Model. You get these in your Entity Data Model when two entities are linked by a many-to-many linking table that has “no payload”; that is, no columns other than the two foreign keys (which also form a composite primary key). 47 An example would be an Employee linked to a Territory by means of an EmployeeTerritory table whose composite primary key consists of the two foreign keys EmployeeId and TerritoryId, and which has no other columns. When you have such an association, invoking Add() on the many-to-many navigation property creates (in the EntityManager cache) the necessary linking object in the EntitySet for the linking objects48. Remove() marks as deleted the linking object that formerly connected the two entities in the many-to-many relationship. Both changes – the insertion of a new linking object or the deletion of an existing one – are propagated to the back-end data store upon the execution of SaveChanges() on the governing EntityManager.

46

Speaking again of the equivalent result on table rows in a relational database, the child entity‟s foreign key value is set to null.
47 48

See the appendix “Many-to-Many Associations in the Entity Framework” in the Object Mapping chapter for more information. Note that those objects are not exposed in the conceptual model, and are never manipulated directly by you.

IdeaBlade DevForce

Business Object Persistence

Adding and Removing Items in Custom-Coded Many-to-Many Navigation Properties
You can (and probably will) also have in your model many-to-many associations involving linking entities that do have payload. (For example, in the NorthwindIB database, Order links Employees (who act as sales reps) to Customers in a many-to-many relationship.) For these cases, you should add and remove elements to the m-to-m collection (e.g., anEmployee.Customers) by inserting or deleting instances of the linking entity. Since that linking entity is probably significant in its own right (again consider an Order), it likely has properties that need their values set at creation time in any case. For example, the following code will have the indirect effect of adding a new Customer to the Customers collection of anEmployee, but only if the Order being added is for a Customer with which anEmployee is not already linked through some other Order. Otherwise, aCustomer is already in anEmployee‟s Customers collection.

C#

// May add a Customer to anEmployee‟s Customers collection anOrder = Order.Create(_entityManager, aCustomer, anOrderDate); anEmployee.Orders.Add(anOrder);

VB

' May add a Customer to anEmployee‟s Customers collection anOrder = Order.Create(_entityManager, aCustomer, anOrderDate) anEmployee.Orders.Add(anOrder)

Similarly, the following code will have the indirect effect of removing aCustomer from the Customers collection of anEmployee, but only if anEmployee has no other Orders for aCustomer. If she does, then aCustomer will remain in her Customers collection.

C#

// May remove a Customer from anEmployee‟s Customers collection anOrder.EntityAspect.Delete();

VB

' May remove a Customer from anEmployee‟s Customers collection anOrder.EntityAspect.Delete()

Business Object Creation Review
Developers will write a creation method for each business object class that can add new objects. That method will return a new business object after it     gets a prototype from the EntityManager assigns an id to the prototype (optionally) sets certain prototype values to satisfy minimum standards of validity adds the prototype to the EntityManager

If we have to generate custom ids for our business objects, we probably will write and register an id generation class that conforms to the DevForce IIdGenerator interface.

IdeaBlade DevForce

Business Object Persistence

Saving Business Objects
Add, change, and delete operations only affect entities in a EntityManager cache. They are not written to the data source nor are they visible to other application users until the application tells the EntityManager to save them. Alternatively, the application can undo the changes rather than save them. If the application decides to save, it issues one of the overloads of EntityManager.SaveChanges() that can save an individual business object, an arbitrary list of objects, or all entities with pending changes. Saves are always transactional in DevForce. If concurrency checking is enabled, DevForce will confirm that entities being saved have not been modified or deleted by another process since they were last retrieved. This chapter elaborates on each of these points.

EntityState of an Object
Unmodified entities are never saved. Attempts to save them are ignored. The application can determine if a particular object is new, modified, marked for deletion, or unmodified by examining its EntityState property which returns one of the corresponding EntityRowState enumerations. The application can also query the cache for all entities that are in one particular EntityState or specific combination of EntityStates and submit them together for save.

Undo
Modified business objects don‟t have to be saved. The application can undo changes made to a single object or a list of objects in the cache. This is a single level undo. Undoing a pre-existing object, whether changed or marked for deletion, restores it to its state when last retrieved from the data source 49; its EntityState becomes “unmodified.” Undoing a newly created object deletes it immediately and removes it from the cache. There is no undo of an undo.

Multi-level Undo
The EntityManager provides “Checkpoint” methods that facilitate implementation of applications that need multilevel undo. The utility of “checkpointing” is most apparent in the UI so we cover it in the WinForm User Interfaces chapter in the topic “Multi-Level Undo with Checkpoints”.

Validation
The wise developer will validate business objects before saving them. Many developers perform validity checks in the presentation layer. Some checks in the UI make sense especially when they provide crisp and immediate user feedback. But good design keeps most validation logic out of the presentation layer and delegates it to the business object. Here are four good reasons:

49

Technically, undoing a modified entity sets the “current” version of the entity to its “original” version. Entity versions are covered in “Advanced Business Object Concepts”.

IdeaBlade DevForce

Business Object Persistence

As the application evolves there are likely to be multiple screens – even multiple UIs – updating the same business object. There is high risk that they will perform validation differently and omit essential checks if each handles its own validation. The object may be changed by a batch program or by a web service. We need to perform the same validations in these modes as we do in a graphical interface. Cross-field and cross-record checks in the UI can create deadlocks and recursion problems. It‟s easier to apply rules such as “the birth date comes before the hire date” and “orders weighing more than 100 pounds must be shipped by ground” after the user presses a button rather than try to enforce them while the user is typing. It‟s easier to break up or combine forms in an interface if you don‟t also have to juggle the validation code to match.

 

DevForce offers extensive facilities for defining and executing validation logic. See the chapter, “Validation Through Verification”.

Temporary Id Fix-up
Initiation of any save operation causes the EntityManager to attempt to replace temporary ids with permanent ids. Subsequent success, failure, or cancellation is immaterial. The act of saving launches the fix-up process. The fix-up process was covered above, in the section “Id Fix-up”. Be sure you understand the fix-up process as detailed in that section.

Life Cycle Events
Creation, retrieval, modification, removal, deletion, and save are key moments in a cached entity‟s life cycle. DevForce raises events on these occasions. The developer can subscribe and react accordingly.

Client-Side Life Cycle Events
Client-side life cycle events on the EntityManager include Fetching, Fetch, Saving, and Saved. These are summarized in Table 7:
Table 7. EntityManager Life-Cycle Events

Event Fetching Fetched Saving Saved

Typical Uses of the Corresponding Event Handler Modify the query being submitted, or refuse the request for data. Modify the objects that were returned by the query Modify the object submitted for saving, or refuse the request to perform inserts and/or updates. Modify the saved object (which might be different from the object submitted for saving by virtue of triggers that were fired on the back end to modify the latter after it was saved).

The EntityManager raises a Fetching event shortly after the application initiates a data retrieval operation. It raises a Fetched event if any entities are retrieved successfully. We can add our own event handlers to these events. The Fetching event provides the handler with a copy of the query object that the caller proposes to submit. The event handler can scrutinize the query object, modifying it or rejecting the query entirely if security or other considerations make that the appropriate response.

IdeaBlade DevForce

Business Object Persistence

The EntityManager raises the Fetched event if any entity is retrieved. The handler receives a list of the entities that were retrieved. The EntityManager raises a Saving event shortly after the application initiates a save. It raises a Saved event if any entities are saved successfully. We can add our own event handlers to these events. The Saving event provides the handler with a list of entities that the caller proposes to save. It will calculate that list if the method parameters do not prescribe the list50. The event handler can scrutinize the list, invoke validation methods on selected entities, clean up others (e.g., clear meaningless error conditions), add additional entities to the list, and even exclude entities from the list. Lastly, it can cancel the save. The EntityManager raises the saved event if any entity is saved. The handler receives a list of the entities that were saved successfully. In transactional saves, either every entity in the save list is saved or none of them are. In DevForce, saves are always transactional, even across disparate back-end data sources.

Server-Side Life Cycle Events
Server-side life cycle events on the EntityServer include ServerFetching, ServerFetched, ServerSaving, and ServerSaved. These are summarized in Table 8.
Table 8. PersistenceServer Life-Cycle Events

Event ServerFetching ServerFetched ServerSaving ServerSaved

Typical Uses of the Corresponding Event Handler Modify the query being submitted, or refuse the request for data. Modify the objects that were returned by the query Modify the object submitted for saving, or refuse the request to perform inserts and/or updates. Modify the saved object (which might be different from the object submitted for saving by virtue of triggers that were fired on the back end to modify the latter after it was saved).

These events provide the developer with the opportunity to do perform server-side, before-the-fact and after-the-fact operations on both queries and saves. The EntityManager, which resides client side, provides corresponding client side events: Fetching, Fetched, Saving, and Saved. The developer thus has complete flexibility to perform centralized processing on data retrievals and updates, client-side or server-side, as her use case dictates. For those familiar with DevForce Classic, the EntityServer.ServerFetching event replaces the DevForce Classic PersistenceServer‟s QuerySecurityCheck event. Similarly, EntityServer.ServerSaving replaces PersistenceServer.SaveSecurityCheck.
ServerFetching will have access both to the submitted query object and to an IPrincipal representing the authenticated user who made the request. ServerFetched, ServingSaving, and ServerSaved will also have

access to that same IPrincipal, but instead of a query object they will have access to the full collection of DevForce entities being retrieved or updated. Thus, the ServerFetched, ServingSaving, and ServerSaved event handlers will make use of the copy of the Domainmodel assembly that has been deployed server-side.

50

SaveChanges() with no arguments, for example, is a blanket request to save every changed entity in cache.

IdeaBlade DevForce
Implementing Server-Side Life-Cycle Event Handlers

Business Object Persistence

Unlike the client-side life-cycle events, the server-side events are handled by providing, for each, a class that implements an appropriate interface. These interfaces reside in the IdeaBlade.EntityModel.Server assembly (which must, of course, be referenced by the project that contains the life-cycle handlers), and are as shown in the table at right.

Event ServerFetching ServerFetched ServerSaving ServerSaved

Interface IEntityServerFetching IEntityServerFetched IEntity ServerSaving IEntity ServerSaved

Once you have provided an implementation of the desired interface, you must attend to two additional steps to ensure that the server-side methods can be found and used by DevForce: 1. 2. Make sure that the assembly containing the implementations is listed as a top-level probe assembly in the app.config file; and Make sure that said assembly is deployed to the appropriate location at build time.

Here‟s an excerpt from an app.config file that lists an assembly named “Server” as a top-level probe assembly:

XML

<ideablade.configuration version="5.00" updateFromDomainModelConfig="Ask" loginManagerRequired="true"> <probeAssemblyNames> <probeAssemblyName name="Server" /> </probeAssemblyNames>

Here‟s a post-build event that ensures that the Server assembly will be deployed to the executables folder in a singlemachine development environment:

IdeaBlade DevForce

Business Object Persistence

Saves and Transaction Management
EntityManager save methods can save a single business object, a list of objects, all objects in a particular

modified state (e.g., “new”), or all entities with pending changes. Recall that modified objects include additions, updates, and deletes. Deleted records are actually marked for delete and must be “saved” to be deleted from the data source.
EntityManager saves are transactional by default. When the developer saves more than one entity at a time,

DevForce processes them together as a single unit of work. Either every save succeeds, or they are all rolled back. Behind the scenes, DevForce causes the necessary INSERT, UPDATE, and DELETE statements to be wrapped within “Begin Transaction” and “Commit Transaction” or “Rollback Transaction” statements. If all succeed the transaction is committed. If any fail, the data source is restored to its pre-transaction condition51. The application relies upon the data source manager to provide two key benefits throughout the transaction: Consistency - simultaneous queries and change requests cannot collide with each other, and users must never see or operate on data that is in mid-change. In a multi-user environment the data source manager must prevent simultaneous queries and data modification requests from interfering with each other. This is important because if the data being processed by a query can be changed by another user's update, the results of the query may be ambiguous. Recovery - in case of system failure, data source recovery is complete and automatic.

51

We cover save failures in topic coming up soon.

IdeaBlade DevForce

Business Object Persistence

SQL defines different degrees of consistency enforcement called “isolation levels”. Each database vendor has a different default isolation level and a proprietary syntax to change it. The developer is responsible for setting the database isolation level and all other global database behavior options. Such settings may be made in the database itself or with proprietary information embedded in the connection string. Consult the database vendor‟s documentation.

Distributed Transactions
DevForce can provide transactional integrity when saving entities to two or more data sources. These data sources can be of different types from different vendors. Their data source managers must support the X/Open XA specification for Distributed Transactions52. The developer instructs DevForce to use the .NET Enterprise Services (AKA, COM+) Distributed Transaction Coordinator (DTC) to handle transaction management. DTC performs a two phase commit. In the first “prepare” phase all parties to the transaction signal their readiness to commit their parts of the transaction. In so agreeing they must guarantee that they can complete their tasks even after a crash. If any participant does not agree, all parties roll back the transactions they have underway. If all participants agree, the transaction moves into the second, “commit” phase in which the parties actually commit their changes. If the transaction is successful, the entities are re-queried.

Re-query After Save
DevForce immediately re-queries the entity after inserting or updating it successfully. Re-query is essential because the insert or update may provoke a data source trigger that modifies the data source object. We often use a trigger to update an optimistic concurrency column. A database-maintained timestamp is another common example. In such cases, the row in the database is no longer exactly the same as the row we wrote. The EntityServer must update the entity and then send it back to the client‟s EntityManager. The revised entity re-enters the cache, replacing its original; its EntityState becomes Unchanged.

When Save Fails
The EntityManager.SaveChanges() method overrides all return a SaveResult object.
SaveResult.Ok returns “true” if the save was entirely successful. If the save was cancelled in a ServerSaving handler, SaveResult.WasCancelled will return “true” and SaveResult.Ok will return “false”. If the save failed for any reason, the PM throws an EntityManagerSaveException.

This is the default behavior. You can change that behavior – indeed, SaveChanges() used to behave differently in versions prior to 3.1.3 – but we recommend that you stay with this default. It follows that you prepare your code to catch and analyze an exception. You will find the information you need in the exception, including the SaveResult object that SaveChanges() would have returned. Always handle SaveChanges exceptions.  Do wrap every call to EntityManager.SaveChanges() in your own custom Save method.

52

At this writing, databases are the only DevForce supported data sources that support the X/Open XA protocol.

IdeaBlade DevForce

Business Object Persistence

 Do wrap every SaveChanges in a Try/Catch and analyze the exception when thrown. Here‟s a code fragment showing a Save method that matches our recommendation:
Code Snippet 45. WhenSaveFails()

C#

internal void SaveAll() { try { using ( new WaitCursor(Page.ParentForm) ) { MainEm.Manager.SaveChanges();// Save everything } DisplaySaveOk(); } catch ( EntityManagerSaveException saveException ) { ProcessSaveFailure(saveException); } catch { throw; // re-throw unexpected exception } }

VB

Friend Sub SaveAll() Try _em1.SaveChanges() ' Save everything DisplaySaveOk() Catch saveException As EntityManagerSaveException ProcessSaveFailure(saveException) Catch Console.WriteLine("While saving, an exception not of type " + _ "EntityManagerSaveException was thrown.") End Try End Sub

The serious failure interpretation and recovery work is in the ProcessSaveFailure method which is custom code that we write. The information we need is in the EntityManagerSaveException instance passed as a parameter to the method.

SaveChanges() Exceptions
The EntityManager raises a EntityManagerSaveException if the save is canceled (e.g., you cancel it in your Saving event handler) or if there is any kind of exception. The EntityServerError handler gets the first crack at the exception. If there is no handler or it doesn‟t handle the exception, the PM throws it again, now in the context of the SaveChanges() call.  We recommend that you do not handle save exceptions in the EntityServerError; leave that to the code near your SaveChanges() call that catches and interprets save failures. You‟ll find examples of this recommendation in the Funhouse in the ApplicationController and EmployeePageController classes.

EntityManagerSaveException
The EntityManagerSaveException inherits from EntityServerException, supplementing that base class with information pertaining to the save.

IdeaBlade DevForce

Business Object Persistence

That information includes an instance of SaveResult such as would have been returned from SaveChanges(). We‟ll discuss that in a moment. First we‟ll get a rough idea of what went wrong by looking at the exception‟s Failure Type.

FailureType Connection Data Concurrency Other

Description The Entity Manager could not reach the data source. There might be a network connection problem or the data source itself could be down. The data source threw an exception such as a referential integrity violation. There was a concurrency conflict. Could be anything but usually the cause is that the save was canceled by the Saving event handler. Check the SaveResult.Canceled.

Once we‟ve learned the category of failure we can decide how to handle it. We can look to the precipitating exception itself to further refine our response. When the failure type is anything but Connection, we‟ll likely want to examine the SaveResult to learn about which entities were affected and how.

SaveResult
Among its contents are:    SaveResult.Canceled which is true if the save was canceled while handling the Saving event. The precipitating exception, whether from an attempt to connect to the data source or an exception from the data source itself such as a concurrency conflict or referential integrity violation. A list of the entities that were not saved called EntitiesWithErrors. In practice, this will always be a list of one -- the first entity to fail -- since saves are transactional. These entities remain in the cache and retain exactly the values and setting they had before the save attempt.

Alternatives to Default SaveChanges Exceptions
In prior versions, the PM threw an exception only if there was a problem connecting to or exchanging data with the database. Database exceptions, such as concurrency violations or referential integrity violations, were returned to the user in an instance of SaveResult. There are two problems with that approach: 4. 5. Many developers neglected to check the SaveResult and did not realize that the save had failed. It was difficult to anticipate which kinds of problems would appear in the SaveResult and which would cause an exception.

IdeaBlade DevForce

Business Object Persistence

Data Source Concurrency
A multi-user application must decide how to resolve the conflict when two users try to update the same data source entity53. Consider the following: 1. 2. 3. 4. I fetch the Employee with Id = 42 You fetch the Employee with Id = 42 You change and save your copy of the Employee with Id = 42 I try to save my copy of the Employee with Id = 42

Is this really going to happen? There is always a risk that another client or component will change the data source entity while we are holding our cached copy of it. The risk grows the longer we wait between the time we fetch the entity and the time we save or refresh it. In offline scenarios, the time between fetch and update can be hours or days. There could be a great many concurrency conflicts waiting to happen. If I save my copy now, should it overwrite the one you saved? If so, we‟ve chosen “last-in-wins” concurrency checking. My copy replaces your copy; your changes are lost. This is the default in DevForce but we strongly recommend that you adopt another type of concurrency control. Permitting one user to blithely overwrite changes that another user made can be dangerous or even fatal. There is an enormous literature on this subject of “concurrency checking.” The coping strategies are many and complex.

Basic Mechanics of Concurrency Detection
DevForce defers to the ADO.NET Entity Framework‟s mechanism for detecting concurrency conflicts at the time an update is submitted to the back-end data source. The Entity Framework (EF) permits the developer to designate, in the Entity Model, one or more columns of a type‟s data source table as concurrency columns. When a client application submits an update order against such a model to the EF, the EF prepares a SQL Update statement. To that statement it adds a WHERE clause that ensures that all columns designated as a concurrency columns have the same value they did when the record was last retrieved by the submitting application. (In other words, they have not been changed in the meantime by another user.) If that proves not to be the case, the exception thrown by the back-end data source will be propagated back down the application‟s calling chain. It is the developer‟s responsibility to ensure that concurrency columns that should change upon an update do change. DevForce makes that considerably easier by providing, in the Object Mapper, a mechanism for automatically updating the value of a given column-based property upon any other change to the record. The Object Mapper offers six Concurrency Strategies that can be applied to a given property:

53

The “data source entity” is the term of convenience we use to describe the data in the data source that map to a corresponding entity in cache. The data source entity may be a single row in a database table as when an Employee cached entity maps to a row in an Employee table. Alternatively, the data may be scattered in many places in some other kind of data source. We have no clue as to the actual location of data behind a Web service entity.

IdeaBlade DevForce
Concurrency Strategy

Business Object Persistence
Instruction to DevForce (Action to Perform Whenever the Entity Is Updated) Replace existing value of the property with a new GUID value. Replace existing value of the property with the current Date/Time. Increment the existing value of the property by 1. Find, in one of the data source‟s probe assemblies, a class that implements the IConcurrencyStrategy interface, and call its
SetNewConcurrencyValue()

AutoGuid AutoDateTime AutoIncrement Server Callback

method, passing that method the Entity and the ConcurrencyProperty as parameters. Said method must be written to update the ConcurrencyProperty as appropriate. Client Just include this column in the concurrency test. The client application will take responsibility for seeing that it is properly updated whenever the entity is modified. Do not use this property to test concurrency.

None

Note that some of the strategies only apply to properties of specific types: clearly we cannot force a GUID value into an integer property, or a DateTime value into a boolean property, and so forth. It remains the developer‟s responsibility to handle any concurrency exception thrown by the back end.

One Concurrency Column, or Many?
Since the Entity Framework permits you to designate any number of columns as concurrency columns, it may be tempting simply to designate them all.54 That‟s one way of making sure that, if anything in the record has been changed by another user since you got your copy, a concurrency conflict will be diagnosed. This may be your only alternative if you have no design access to the database, but be aware that there will be a performance impact. Every update will be accompanied by a flurry of activity comparing values. As with other performance issues, you should do some upfront testing to determine whether the performance impact is unacceptable, or even significant. If you do have design access to the database, or you‟re fortunate enough to inherit a database already designed the way you want it, it‟s generally a better alternative to provide a single column that is guaranteed to be updated whenever anything else is, and to use that as your sole determinant of a concurrency conflict. A simple integer column that is incremented each time the record is updated will do quite nicely; you can also use a GUID,

54

You can, of course, safely omit the primary key.

IdeaBlade DevForce

Business Object Persistence

timestamp, or any other type and methodology that guarantees that the value will change in a non-cyclical way. As you have seen, DevForce makes it easy for you to make a column auto-updating.

Concurrency and the Object Graph
A large part of the complexity revolves around the scope of concurrency checking. Have I changed an order if I add, change or delete one of its OrderDetail items? If I change the name of a customer, have I changed its orders? These considerations have to do with concurrency control of the business object graph. DevForce does not support graph concurrency directly. DevForce supports single-table, “business object proper” concurrency control. The developer can achieve the desired degree of graph concurrency control by employing single-table checking within a properly conceived, transactional concurrency plan. It doesn‟t have to be wildly difficult. In brief, the developer adds custom business model code such that     Added, changed, or deleted children entites always modify their parents. An application save attempt always includes the root entity of the dependency graph. During a save attempt, the root entity ensures that its children are included in the entity-save list. These children include their children.

Handling concurrency conflicts in these situations is discussed further in the section “Concurrency and Dependent Entries.” For now we return to concurrency checking of single business objects.

Pessimistic versus Optimistic Concurrency Checking
There are two popular approaches to concurrency checking: pessimistic and optimistic. In pessimistic concurrency, we ask the data source to lock the data source entity while we examine it. If we change it, we write it back to the data source. When we are done looking at it or updating it, we free the lock. While we have it locked, no one else can see or use it. This approach holds up other users trying to reach the object we hold. It gets worse if we need many objects at once. There are potential deadlocks (I grab A, you grab B, I want B too, but can‟t get it, so I wait. You want A also, but can‟t get it, so you wait. We both wait forever). There are more complicated, less draconian implementations to this approach but they amount to the same punishing performance. Under optimistic concurrency, we don‟t lock the table row. We bet that no one will change the source data while we‟re working with it and confirm our bet when (and if) we try to update the data. The mechanism works as follows. We fetch a copy of the table row and turn it into a business object. We work with this copy of the data source entity. We may decide to update the entity or mark it for deletion. When we save an altered entity, the business object server converts our intention into a data source management command. That command, in the process of updating or deleting the supporting table row, confirms that the row still exists and has not changed since we fetched it. If the row is missing or has changed, the command fails and it‟s up to the application to figure out what to do about it. Changes are comparatively rare so we have reason to be optimistic that the row will be exactly as we last found it.

Resolving Concurrency Collisions
Our optimism is usually rewarded. Occasional disappointment is inevitable. Eventually, we will encounter a conflict between our cached entity, with its pending changes, and the newly-updated data source entity. We will want to resolve that conflict one way or the other. The possible resolutions include:   Preserve the pending changes and ask the user what to do. Abandon the pending changes and re-fetch the entity.

IdeaBlade DevForce
 

Business Object Persistence

Arrange for the cached entity to become the current entity while preserving the pending changes Compare the cached entity with the current data source entity and merge the difference per some business rules or as guided by the user.

The first choice is the easiest place to start. We do nothing with the entity and report the problem to the user. The cached entity cannot be saved. We leave it up to the user to decide either to abandon the changes (option #2) or push them forward (options #2 and #3). The remaining options involve re-fetching the entity from the data source. They differ in what they do with the entity retrieved – a difference determined by the MergeStrategy55 and how we use it.

C# VB

aManager.RefetchEntity(anEntity, aMergeStrategy);

aManager.RefetchEntity(anEntity, aMergeStrategy)

OverwriteChanges

The second choice uses the OverwriteChanges strategy to simply discard the user‟s changes and update the entity to reflect the one current in the datasource. While unmatched in simplicity, it is almost the choice least likely to satisfy the end user. If this is the only option, we should have the courtesy to explain this to the user before erasing her efforts.
PreserveChangesUpdateOriginal

The third choice makes the cached entity current by re-fetching with the PreserveChangesUpdateOriginal strategy. This strategy causes the cached entity to trump the current datasource entity with a little trickery. The refetch replaces the cached entity‟s original version56 with the values from the current data source entity but it preserves the cached entity‟s current version values, thus retaining its pending changes. The cached entity‟s original concurrency column value now matches the concurrency column value in the datasource record.
Code Snippet 46. CurrentAndOriginal()

C#

// the current value of the property in the cached entity Employee.FirstNameEntityProperty.GetValue(anEmployee, EntityVersion.Current); // the value from the datasource when most recently retrieved Employee.FirstNameEntityProperty.GetValue(anEmployee, EntityVersion.Original);

VB

' the current value of the property in the cached entity Employee.FirstNameEntityProperty.GetValue(anEmployee, EntityVersion.Current) ' the value from the datasource when most recently retrieved Employee.FirstNameEntityProperty.GetValue(anEmployee, EntityVersion.Original)

The effect is as if we had just read the entity from the datasource and applied the user‟s changes to it.
55 56

We discuss merge strategies in “Advanced Business Object Concepts”. We cover entity versions in “Advanced Business Object Concepts”.

IdeaBlade DevForce

Business Object Persistence

If we ask the persistence layer to save it now, the datasource will “think” that we modified the most recently saved copy of the entity and welcome the changed record. This option is much like “last one wins” concurrency with a crucial difference: it was no accident. We detected the concurrency collision and forced the issue in accordance with approved business rules. The PreserveChangesUpdateOriginal strategy works only if the entity is governed by optimistic concurrency. If the entity lacks a concurrency column, the refetch uses the OverwriteChanges strategy instead. Of course we wouldn‟t be talking about concurrency resolution if there were no concurrency columns.
PreserveChangesUpdateOriginal with Merge

The fourth possibility begins, like the third, with a re-fetch governed by the PreserveChangesUpdateOriginal strategy. This time we don‟t forcibly save the cached entity. We execute business logic instead which compares the current and original versions, column by column, deciding whether to keep the locally changed value (the “current” value) or the datasource value (now tucked inside the “original” value). Such logic can determine if and when the cached entity‟s values should prevail. The logic may be entirely automatic. Alternative, the program could present both versions to the user and let her decide each difference.

Concurrency and Dependent Entities
What if a bunch of entities are mutually dependent? Suppose we have an order and its details. User „A‟ adds two more details and changes the quantity on a third. She deletes the fourth detail and then saves. In many applications, an order is never less than the sum of its parts. The order and every one of its details must be treated as a unit at least for transactional save purposes. We will describe this network of dependency as a “Dependency Graph”. DevForce does not offer native support for dependency graphs and its concurrency conflict detection and resolution features target single entity, “business object proper” concurrency only. We are about to consider how you can extend DevForce concurrency checking for dependency graphs. We‟ll talk more about dependency graphs in general later in this section.

Detection
Continuing our story and standing at an Olympian distance with an all knowing eye, we see that User „B‟ changed the fifth order detail and saved before User „A‟ tried to save her changes. User „A‟ didn‟t touch the fifth order detail. She won‟t know about the change because there will be no concurrency conflict to detect; she can‟t detect a concurrency conflict unless she save the fifth order detail and she has no reason to do so. If this worries you (it worries me), you may want to establish business rules that detect concurrency violations for any of entity in a dependency graph. A good approach is to   Identify the root entity of the graph (Order) and Ensure that a change to any node in the graph (OrderDetail) causes the root entity to change.

User „B‟s change to the fifth detail would have meant a change to the order. User „A‟s changes also modified the order. User „A‟s save attempt will provoke a concurrency violation on the root entity, the order.

IdeaBlade DevForce
Resolution

Business Object Persistence

Now that User „A‟ has learned about the violation, what can she do? There is no obvious problem. Neither „A‟ nor „B‟ changed the order entity itself so there are not differences to reconcile. There is only the tell-tale fact that their concurrency column values are different. It doesn‟t seem proper to proceed blithely, ignoring the violation and proceeding as if nothing happened. User „A‟ should suspect something is amiss in the details. The application should re-read all details, even those the user didn‟t change. It should look for diffences at any point in the graph and only after applying the application-specific resolution rules should it permit the entire order to be saved again. What are those resolution rules? We suggest taking the easiest way out if possible: the application should tell the User „A‟ about the problem and then throw away her changes. There must be something fundamentally wrong if two people are changing an order at the same time. In any case, the complexity of sorting out the conflict and the risk of making a total mess during “reconciliation” argue for a restart. If you can‟t take the easy way out – if you have to reconcile – here are a few pointers. It is probably easiest to use a temporary second EntityManager for the analysis. A single EntityManager can only hold one instance of an entity at a time and we need to compare two instances of the same entity. This is manageable if there is only one entity to deal with – we‟ve seen how to use the current and original versions within each entity to carry the difference information. This trick falls apart when we are reconciling a dependency graph. Instead we‟ll put User „A‟s cached order and its details in one manager and their dopplegangers from User „B‟ in another. The author thinks it is best to import User „A‟s order and details into the second manager and put User „B‟s version into the main manager by getting them with the OverwriteChanges strategy. This seems counter-intuitive but there are a couple of good reasons.   We can ImportEntities into the second manager without logging it in. We‟d have to log in the second manager before we could use it to get GetEntities. This is not hard, but avoiding it is even easier! The application probably should favor User „B‟s order; if so that order will be in the main manager where it belongs.

Show some restraint
The order‟s entire object graph is not its dependency graph. The order may consist of details but that may be as far as it goes. For example, every detail is associated with a product. If User „B‟ changed the fifth detail‟s product name or its color, should this provoke a concurrency conflict? It User „C‟ updated the order‟s customer name, does that mean all orders sold to that customer must be scrutinized for concurrency collisions. Most businesses will say “no.”

Saving the “Dependency Graph”
The DevForce relations between entity classes are indicative of associations among those classes. These associations define an object graph which may cast a wide net over the data source data. In this section we consider a portion of that object graph in which a change to one node requires a change to another node. Such nodes form a sub-graph of mutual dependency we could call a “dependency graph”. Let‟s look a little closer.

IdeaBlade DevForce Association Types
Associations come in a variety of strengths:

Business Object Persistence

Type Association

Description A simple association is typically read as a “Has a” relationship. An Address has a State or a Part has a Color. The two ends have independent lifetimes. A change to the city or the name of the part does not alter the state or the color.57 An aggregation implies a stronger, “Owns a” relationship. A Company owns its employees. The two ends still have independent lifetimes. There is still a law against slavery and the employee may transfer to another company. Yet the bond between Company and Employee is stronger than between Part and Color. There are ramifications to the making and breaking of ties. A composition is a “whole / part” relationship in which the whole is said to “consist of” or “be made up of” the parts. An Order is substantially made up of its detailed items. This is the strongest bond. The lifetimes of the two ends are closely tied. If the order disappears, its details disappear with it. Adding, changing, or deleting details alters the parent order.

Aggregation

Composition

DevForce itself has no mechanism for distinguishing among these association types. In fact, DevForce treats its relations as the simplest association. It makes no assumption about the consequences for related entities of any alterations to either parent or child. It is not clear that there is a meaningful programmatic distinction between Association and Aggregation. There will be more business rules surrounding an Aggregation but business rules always require custom coding so the difference is one of degree rather than of kind. The relevant fact in this context is that parent and child may be modified independently. Yes, we must adjust the child if we delete the parent. There may be constraints and consequences to joining and separating parent and child. There can be side-effects of altering parent or child data unrelated to their bond. But, in general, we don‟t require a modification in one to effect a modification of the other.

Compositions
There are systems that explicitly support the Composition distinction. If you mark a relationship as a composition, the system will implement it differently. The parts (children) in a composition will be contained by the whole (parent) and they may only be accessed through the parent. If you marked an Order‟s OrderDetails property as a composition, the only way to obtain details would be through this property. OrderDetails fetched through any other mechanism would be different objects than the conceptually same entities fetched with the OrderDetails property. We think that is a rare and extreme position which is more trouble than it is worth. The developer can program to it when it occurs but DevForce does not encourage the practice with any means of its own. No mechanism is provided to mark the OrderDetails navigation property as a “Composition”. But this is not to diminish the importance of the Composition bond. In many applications, we should consider the Order modified if we add, change, or delete one of its OrderDetail entities. If we delete the Order, we almost certainly intend to delete its details as well.

57

They may become incompatible – as when the change to city moves the address to a different state or the part turns out to be colorless – but compatibility is a matter for business rules unrelated to the fundamental nature of the association.

IdeaBlade DevForce

Business Object Persistence

This is precisely the behavior sought by systems with native support for composition. But we can achieve the same effect in DevForce. It is not hard work, although it requires some care. The reward is flexibility. Each application has its own requirements.We can offer only a brief outline of the main points here.     Our application “save” operation concentrates on the root entity (or enties) of the dependency graph. We implement a Saving handler to invoke composition business rules of the entities. We add the composition business rules to the business object, wrapped in a method the Saving handler can call. We provide for intelligent concurrency resolution to detect and manage the collision of our changes with changes by other users.

Save the Root Entity
This step is irrelevant if our save operation calls one of the “Save all” methods of the EntityManager. The “Save all” methods saves every changed entity in the cache. Our Saving handler must be clever because it might encounter a child entity before its parent. It may not learn of the parent at all; the child entity will be responsible for modifying its parent and including that parent in the list of entities to save. On the other hand, if we choose to save a particular set of entities – the current order and its graph for example – it may be convenient to compose the “save list” entirely of root entities – orders in this case. We will see in a moment that compositional business rules ensure that (a) the root entity is in an altered state and (b) its modified dependent objects are also in the save list.

Saving Event Handler
Remember, the EntityManager raises the Saving event whenever it is ready to write entities to the host data sources. We were going to write a Saving handler anyway. We should validate every business object just before saving it to make sure it is safe to persist. The best approach is to write a Saving event handler that iterates over the list of entities-to-save (the “save list”), calling a validate method on each. We might as well extend this approach to call a PrepareSave method instead that both validates and enforces composition business rules.

Composition Business Rules
We may have any number of composition business rules. One of them must ensure that, if a child is on the save list, its parent is also on the list. That‟s easy because we can always add (and remove) items from the save list within the Saving event handler. Another composition rule must ensure that any change to a child entity modifies its parent. That‟s necessary because DevForce will only save an entity that has been changed. It is not sufficient merely to add the parent to the save list; we must make sure it is in an altered state.
Code Snippet 47. EnforcingConcurrencyCheckingOnAConceptualEntity()

C#

anOrderDetail.UnitPrice *= 1.1M; Order parentOrder = anOrderDetail.Order; if (parentOrder.EntityAspect.EntityState == EntityState.Unchanged) { parentOrder.EntityAspect.SetModified(); }

VB

IdeaBlade DevForce

Business Object Persistence

anOrderDetail.UnitPrice *= 1.1D Dim parentOrder As Order = anOrderDetail.Order If parentOrder.EntityAspect.EntityState = EntityState.Unchanged Then parentOrder.EntityAspect.SetModified() End If

Concurrency Violations
We always use transactional saves. We‟ve taken steps to ensure that all members of the “dependency graph” – the order and all of its details, for example, - are part of the same save list and are slated for persistence as a single transaction. When DevForce persistence layer detects a concurrency violation, it terminates the transaction and returns the offending entity as we learned earlier. Chances are there will be more than one entity in the transaction that is in potential concurrency conflict with its corresponding object in the data source. The end user will be most unhappy if we walk her through each entity one by one. We should resolve the concurrency conflicts of all entities in the dependency graph in a single shot. While the exact details will be application specific, they will be some variation on the techniques you learned for resolving conflicts of individual entities.

Dependency Graph Retrieval
Many large DevForce applications use multiple EntityManagers (PM) to maintain separate editing contexts. They often need to transfer entire entity graphs between PMs. For example, an application might have a main PM to hold lists of entities. One list might hold SalesReps and the application could display that list in a grid. Double clicking on one SalesRep row launches a popup editor for the selected sales rep. Double clicking a different SalesRep row launches a second popup editor for the rep in that row. The user can make changes to the first rep, switch to the second editor and make changes to that rep, go back to the first editor, make more changes, and save them. The second editor (and the rep it edits) remains open, and its pending changes are not saved. The user may decide to cancel the second editor, discarding the changes; of course the changes to the first rep have been saved independently. To implement this scenario, we recommend that each editor have its own PM which constitutes an “editing context” that is independent both of the main PM and of other editors. When the application launches an editor, it populates the editor‟s PM with the selected SalesRep from the grid. Because that SalesRep is in the main PM, the application will likely transfer (import) the selected SalesRep into the editor‟s PM. At this moment, the rep in the editor PM is a clone of the rep in the main PM. After save, the application might export the saved rep back into the main PM where it now displays in the grid in its post-save glory[1]. Notice that we mentioned only the transfer of a single entity – the “root entity” – between the PMs. In practice, we often want to transfer both the root entity and many of its related entities. We might transfer the sales rep and his order information (Orders, OrderDetails) as well so that the entire “entity graph” can be edited in a single context. Heretofore, the developer would have to implement the logic to gather up the entities in the graph before transferring them, a task that could require a sophisticated knowledge of the DevForce Object Query Language. Now she can use the DevForce “span” technology to compose a single query-like statement to do the job. The following example implements the scenario described above:

IdeaBlade DevForce
C#

Business Object Persistence

// Copy selected Employees and their Orders, OrderDetails, and Products // from one PM to another. private void GetGraph_OneRootOneSpan() { DomainModelEntityManager em1 = new DomainModelEntityManager(); DomainModelEntityManager em2 = new DomainModelEntityManager(); int employeeID = 1; var targetedEmployeesQuery = em1.Employees .Where(e => e.EmployeeID == employeeID); int targetedEmployeesCount = targetedEmployeesQuery.Count(); if (targetedEmployeesCount != 1) { Console.WriteLine("Unable to retrieve Employee with EmployeeID == {0}", employeeID); PromptToContinue(); return; } // FindEntityGraph() operates against the cache only: it does not retrieve // entities into the cache. So let's retrieve the desired entities... Console.WriteLine("Retrieving Emp-Orders-OrderDetails-Products..."); List<Employee> employees = targetedEmployeesQuery .Include("Orders.OrderDetails.Product").ToList(); Employee anEmployee = employees[0]; // Create roots list and add the employee. List<Entity> roots = new List<Entity>(); roots.Add(anEmployee); // Add span(s). List<EntitySpan> spans = new List<EntitySpan>(); EntitySpan aSpan = new EntitySpan(typeof(Employee), EntityRelations.FK_Order_Employee, EntityRelations.FK_OrderDetail_Order, EntityRelations.FK_OrderDetail_Product); spans.Add(aSpan); // Get entity graph for entities in roots EntityState entityState = EntityState.Unchanged; IList<Object> entityGraph = em1.FindEntityGraph(roots, spans, entityState); Console.WriteLine("{0} entities collected by FindEntityGraph.", entityGraph.Count ); // Import graph into a second EntityManager em2.ImportEntities(entityGraph, MergeStrategy.OverwriteChanges); }

VB

Private Sub GetGraph_OneRootOneSpan() Dim em1 As New DomainModelEntityManager() Dim em2 As New DomainModelEntityManager() Dim employeeID As Integer = 1 Dim targetedEmployeesQuery = em1.Employees.Where(Function(e) e.EmployeeID = employeeID)

IdeaBlade DevForce

Business Object Persistence

Dim targetedEmployeesCount As Integer = targetedEmployeesQuery.Count() If targetedEmployeesCount <> 1 Then Console.WriteLine("Unable to retrieve Employee with EmployeeID == {0}", employeeID) PromptToContinue() Return End If ' FindEntityGraph() operates against the cache only: it does not retrieve ' entities into the cache. So let's retrieve the desired entities... Console.WriteLine("Retrieving Emp-Orders-OrderDetails-Products...") Dim employees As List(Of Employee) = targetedEmployeesQuery.Include("Orders.OrderDetails.Product").ToList() Dim anEmployee As Employee = employees(0) ' Create roots list and add the employee. Dim roots As New List(Of Entity)() roots.Add(anEmployee) ' Add span(s). Dim spans As New List(Of EntitySpan)() Dim aSpan As New EntitySpan(GetType(Employee), EntityRelations.FK_Order_Employee, EntityRelations.FK_OrderDetail_Order, EntityRelations.FK_OrderDetail_Product) spans.Add(aSpan) ' Get entity graph for entities in roots Dim entityState As EntityState = entityState.Unchanged Dim entityGraph As IList(Of Object) = em1.FindEntityGraph(roots, spans, entityState) Console.WriteLine("{0} entities collected by FindEntityGraph.", entityGraph.Count) ' Import graph into a second EntityManager Console.WriteLine() Console.WriteLine("Entities imported to second EntityManager:") Console.WriteLine("------------------------------------------") em2.ImportEntities(entityGraph, MergeStrategy.OverwriteChanges) DisplayCacheContents(em2) PromptToContinue() End Sub

Workflow For a Save
Let‟s put most of these ideas together along with our other knowledge of DevForce business objects and look at a schematic workflow for saving all pending changes to a single database.
Table 9. Transactional Save Workflow in an N-Tier Deployment

Component Client Tier – Application

Action The client application adds, modifies and deletes any number of business objects

IdeaBlade DevForce
Code on the client.

Business Object Persistence

The client application asks a EntityManager to save all pending changes. Client – EntityManager Makes a save list of the new, modified, and deleted entities in cache. Fires the Saving event. Assume that application listener okays the save. Connects to the data source and authenticates the user. Assume success. If there are any temporary ids, the PM sends them to the BOS for fix-up. Middle Tier – Business Object Server Builds map of data source-generated ids (e.g., for Identity columns). Calls method on instance of developer‟s id generation class with remaining temporary. This method returns a map of temporary-to-permanent ids which the BOS returns to the client tier. Uses the temp-to-perm id map to replace all temporary ids. Transmits the save list to the BOS. Middle Tier – Business Object Server First the Saving event. This can be used to perform security checks on each entity in the save list. If a security check fails, an exception can be thrown back to the EntityManager (or any other desired action taken.) Workflow ends. Otherwise… Constructs a batch of insert, update, and delete operations, adjusted for optimistic concurrency checking as required. Arranges them by type per the prescribed PersistenceOrder. Middle Tier – Business Object Server If the data source is a relational database: Forwards them to the Entity Framework for execution. Data Tier - Data Source Middle Tier – Business Object Server If the data source is a web service: Converts the requests to the approprate web service calls and submits them to the web service.

Client –EntityManager

Performs the persistence operations. If there are no failures, it commits them; if there is a single failure, it rolls them all back. If the transaction failed, returns to the EntityManager the identity of the culprit entity and the exception raised by the data source. The EntityManager stores this information in the SaveResult and returns to the client application. Workflow ends. Otherwise… The transaction succeeded. The BOS re-queries the database(s) for all of the inserted and modified entities that are sourced in databases, thus capturing the effects of triggers that fired during save. Converts the (potentially) revised data into entities and sends them to the client side EntityManager. The server‟s local copy of the entities go out of scope and the garbage collector reclaims them. This enables the object server to stay stateless.

Client Tier – EntityManager

Replaces cashed entities with updates from BOS. They are marked “unchanged” because they are now current. Raises the Saved event with list of saved inserted and modified entities.

Client Tier – Application

The application resumes.

IdeaBlade DevForce
Code

Business Object Persistence

Saving the Cache to a Local Disk File
EntityManager has a property, CacheStateManager, that can be used to capture, save, and restore the contents of the EntityManager‟s cache to a local disk file. The following statements save the contents of the cache managed by EnttyManager _mgr:
Code Snippet 48. SaveRestoreCacheToDisk()

C#

string cacheFilePath = System.IO.Directory.GetCurrentDirectory() + "EntityCacheState.bin"; _em1.CacheStateManager.SaveCacheState(cacheFilePath);

VB

Dim cacheFilePath As String = System.IO.Directory.GetCurrentDirectory() & "EntityCacheState.bin" _em1.CacheStateManager.SaveCacheState(cacheFilePath)

These statements restore the contents of the EntityCacheState file "C:\_DevForceCache.dat" to the EntityManager‟s current cache:

C# VB

_em1.CacheStateManager.RestoreCacheState(cacheFilePath);

_em1.CacheStateManager.RestoreCacheState(cacheFilePath)

When called using the overload above, the restore operation using RestoreStrategy.Normal. That RestoreStrategy restores the data in the cache state file using a MergeStrategy of PreserveChanges (see the discussion of this elsewhere in this chapter); it also restores the DefaultSaveOptions and DefaultQueryStrategy saved in that file, overwriting the current values for those properties. When RestoreStrategy.Normal doesn‟t meet your needs, you can restore using a custom RestoreStrategy:

C#

bool restoreSaveOptions = true; bool restoreQueryStrategy = false; RestoreStrategy aRestoreStrategy = new RestoreStrategy( restoreSaveOptions, restoreQueryStrategy, MergeStrategy.OverwriteChanges); _em1.CacheStateManager.RestoreCacheState(cacheFilePath, aRestoreStrategy);

VB

Dim restoreSaveOptions As Boolean = True Dim restoreQueryStrategy As Boolean = False Dim aRestoreStrategy As New RestoreStrategy(restoreSaveOptions, restoreQueryStrategy, MergeStrategy.OverwriteChanges) _em1.CacheStateManager.RestoreCacheState(cacheFilePath, aRestoreStrategy)

CacheStateManager also includes a method, GetCacheState(), which returns the state of the cache as a serializable in-memory object:

IdeaBlade DevForce

Business Object Persistence

C# VB

EntityCacheState cacheState = _em1.CacheStateManager.GetCacheState();

Dim cacheState As EntityCacheState = _em1.CacheStateManager.GetCacheState()

This can be used in a variety of ways; for example, in a server-side method called from the client using EntityManager.InvokeServerMethod() or InvokeServerMethodAsync(), you could fill an EntityManager‟s cache with any arbitrary collection of data, capture that in an EntityCacheState, and return that EntityCacheState to the client where it could be restored using another overload of RestoreCacheState:

C# VB

_em1.CacheStateManager.RestoreCacheState(cacheState);

_em1.CacheStateManager.RestoreCacheState(cacheState)

You can also, of course, encrypt the cache state before saving it to local storage.

XML Serialization of Business Objects
IdeaBlade entities can be serialized as XML for any number of purposes, including exposing these entities to a Web Service, or as the first step in some XSLT transform for reporting or further processing. Please see Microsoft‟s WCF documentation for detailed examples of exposing objects to Web Services via data contract serialization. All entities generated by DevForce are marked with WCF DataContract attributes; and all public properties of these entities are marked with a DataMember attribute. Serialization using standard WCF via either the   System.Runtime.Serialization.DataContractSerializer, or the System.Runtime.Serialization.NetDataContractSerializer

is supported. One of the big issues with XML Serialization when serializing object graphs (objects that are connected to other objects, ad-infinitum) has to do with the depth of the object graph that should be serialized. Without some mechanism to control the depth of the graph, the serialization of a single entity might result in hundreds or even thousands of related entities being serialized. DevForce controls this by only serializing entities that are present within an EntityManager‟s cache at the inception of serialization and are navigable from the directly serialized entities. (Think of this as all entities that are available via a CacheOnly query) This allows fine-grained control over what will be serialized. Any relation properties that would return an entity or entities that are not in the cache will instead serialize the property value either as a null entity (for scalar properties), or as an empty collection (for collection properties). The examples serializes two employees along with all related Orders and their line items (OrderlDetails):
Code Snippet 49. SerializeBusObjects()

IdeaBlade DevForce
C#

Business Object Persistence

private void SerializeBusObjects(string serializerName) { var employees = _em1.Employees .OrderBy(e => e.LastName).Take(2).Include("Orders.OrderDetails").ToList(); var aMemoryStream = new MemoryStream(); var aXmlDictionaryWriter = System.Xml.XmlDictionaryWriter.CreateTextWriter(aMemoryStream); if (serializerName == "NetDataContactSerializer") { SerializeWithNetDataContractSerializer(employees, aXmlDictionaryWriter); } else { SerializeWithDataContractSerializer(employees, aXmlDictionaryWriter); } aXmlDictionaryWriter.Flush(); aMemoryStream.Position = 0; string result = StreamFns.ToString(aMemoryStream); } private static void SerializeWithNetDataContractSerializer(List<Employee> employees, System.Xml.XmlDictionaryWriter aXmlDictionaryWriter) { NetDataContractSerializer aNetDataContractSerializer = new NetDataContractSerializer(); aNetDataContractSerializer.WriteObject(aXmlDictionaryWriter, employees); } private static void SerializeWithDataContractSerializer(List<Employee> employees, System.Xml.XmlDictionaryWriter aXmlDictionaryWriter) { DataContractSerializer aDataContractSerializer = new DataContractSerializer( typeof(Employee), new Type[] { typeof(List<Employee>) }, int.MaxValue, false, true, null); aDataContractSerializer.WriteObject(aXmlDictionaryWriter, employees); }

VB

''' <summary> ''' DataContract serialization ''' </summary> Private Sub SerializeBusObjects(ByVal serializerName As String) Dim employees = _em1.Employees.OrderBy(Function(e) e.LastName).Take(2).Include("Orders.OrderDetails").ToList() Dim aMemoryStream = New MemoryStream() Dim aXmlDictionaryWriter = System.Xml.XmlDictionaryWriter.CreateTextWriter(aMemoryStream) If serializerName = "NetDataContactSerializer" Then SerializeWithNetDataContractSerializer(employees, aXmlDictionaryWriter) Else SerializeWithDataContractSerializer(employees, aXmlDictionaryWriter) End If aXmlDictionaryWriter.Flush() aMemoryStream.Position = 0 Dim result As String = StreamFns.ToString(aMemoryStream) Console.WriteLine("Two employees, serialized using the {0}..." & vbLf, serializerName) Dim abbreviatedResult As String = result.Substring(0, 500) & vbLf & vbLf & "...[snip]..." & vbLf & vbLf & result.Substring(result.Length - 201) Console.WriteLine(abbreviatedResult) PromptToContinue() End Sub

IdeaBlade DevForce

Business Object Persistence

Private Sub SerializeWithNetDataContractSerializer(ByVal employees As List(Of Employee), ByVal aXmlDictionaryWriter As System.Xml.XmlDictionaryWriter) Dim aNetDataContractSerializer As New NetDataContractSerializer() aNetDataContractSerializer.WriteObject(aXmlDictionaryWriter, employees) End Sub Private Sub SerializeWithDataContractSerializer(ByVal employees As List(Of Employee), ByVal aXmlDictionaryWriter As System.Xml.XmlDictionaryWriter) Dim aDataContractSerializer As New DataContractSerializer(GetType(Employee), New Type() {GetType(List(Of Employee))}, Integer.MaxValue, False, True, Nothing) aDataContractSerializer.WriteObject(aXmlDictionaryWriter, employees) End Sub

IdeaBlade DevForce

Business Object Persistence - Advanced

Business Object Persistence – Advanced

Business Object Persistence – Advanced
Getting Information About an Entity Type with GetEntityMeta() Access Both Local and Remote Data Sources In the Same N-tier Application Stored Procedure Queries
SQL Server Stored Procedure Queries

Stored Procedure Entity Navigation Forced Re-fetch Lost Connection During Query Query Cache
EntityManager.RemoveEntities Overload Preserves Query Cache

MergeStrategy In More Detail The EntityManager.AttachEntity Method Filtering Queries Query Inversion in More Detail Transactional Queries DevForce and Data Sources – Deep Dive
The Object Mapper and Manually Added or Modified Keys DataSourceKeys, DataSourceKeyResolvers, and DataSourceExtensions EntityManagers and DataSourceExtensions Tenant Extensions Multi-Part Extensions Extensions and EntityServers Dynamic DataSourceKeys and the DataSourceKeyResolver

Multiple Application Environments Multi-Level Undo with Checkpoints Multiple EntityManager Instances Multi-Threading in a DevForce App Batching Asynchronous Tasks Service Oriented Architecture POCO Support in DevForce
Examples of POCO Classes Examples of a POCO Service Provider Class Example of a Client-Side Class Containing Extension Methods for the EntityManager Obtaining an EntityAspect Property on Your POCO Object Data Contract Serializer (DCS) versus .NET Data Contract Serializer (NDCS) POCO Save mechanisms Summary – Things to Remember When Using POCOs in Your DevForce App

IdeaBlade DevForce

Business Object Persistence - Advanced

Getting Information About an Entity Type with GetEntityMeta()
The instance method GetEntityMetadata() on the EntityMetadataStore type returns an EntityMetadata object that is rich with information about a specified entity type:

C#

EntityMetadata employeeEntityMetaData = EntityMetadataStore.Default.GetEntityMetadata(typeof(DomainModel.Employee));

VB The EntityMetadata objects provides the following members:

The table below provides an explanation for key members:

Property Property Property Method Property Property

CanQueryByEntityKey ComplexTypeProperties ConcurrencyProperties CreateEntity() DataEntityProperties DataSourceKeyName

Gets whether primary key queries are allowed. Returns a collection of EntityProperties that describe complex object properties for entities of this type. Returns a collection of EntityProperties that are concurrency properties for entities of this type. Creates a new entity of the type described by this metadata item. Returns a collection of DataEntityProperties for entities of this type. Gets the name of the Data Source Key associated with this type.

IdeaBlade DevForce
Property Property Property Property Property DefaultEntitySetName EntityProperties EntityType KeyProperties NavigationProperties

Business Object Persistence - Advanced
Gets the default EntitySetName for entities of this type. Returns a collection of EntityProperties that belong to entities of this type. Gets the type of the entity. Returns a collection of EntityProperties that are keys for entities of this type. Returns a collection of NavigationEntityProperties for entities of this type.

Such metadata can be useful in many situations. For example, suppose you wish to dynamically populate a form with bound controls for the properties of a type. You could easily get the list you need from the EntityMetadataStore.

Access Both Local and Remote Data Sources In the Same N-tier Application
An application may need to persist volatile data to a centrally hosted database and have the ability to simultaneously access comparatively static data on a local database. Field technicians who service complex machine parts may need ready access to voluminous parts catalogs and repair manuals. The catalog and repair data don‟t change often . They may be stored in a database on the tech‟s laptop. On the other hand, the central office needs to monitor the technicians rounds and dispatch him to new client sites. There could be significant exchange of information between the dispatch center and the remote technician. A DevForce program should be able to provide access to both the remote and local database in an n-tier deployed application. One of the EntityManager constructors facilitates construction of such an application. C#
public EntityManager( bool pShouldConnect, String pDataSourceExtension, PersistenceServiceOption pPersistenceServiceOption)

VB

The caller sets the third parameter to the value of a PersistenceServiceOption enumeration that indicates how the how the EntityManager‟s PersistenceService should be configured. C#
public enum PersistenceServiceOption { /// <summary> /// Use the Ibconfig file [remoting][remotePersistenceEnabled] node. /// </summary> UseDefaultService = 0, /// <summary> /// Use a local service - Service will run in process with the client /// The Ibconfig file [remoting][remotePersistenceEnabled] node is ignored. /// </summary> UseLocalService = 1, /// <summary> /// Use a remote service as defined in the Ibconfig file [remoting] node.

IdeaBlade DevForce

Business Object Persistence - Advanced

/// The Ibconfig file [remoting][remotePersistenceEnabled] node is ignored. /// </summary> UseRemoteService = 2 }

VB

After configuration, the EntityManager will connect either to the remote database or to the local database. A specific PM cannot switch between the two modes. But the application can have more than one EntityManager and bridge the two at convenient moments – which is exactly how we‟d approach the scenario described above.

Stored Procedure Queries
We broached the subject earlier of the occasional need to use a stored procedure to query for business objects. The need arises most frequently when we require the entities resulting from an extraordinarily complex query involving large volumes of intermediate data that are not themselves required on the client. One might imagine a multi-step query that touched several tables, performed multi-way joins, ordered and aggregated the intermediate results, and compared values with many thousands of records, all so as to return a handful of qualifying results. All of the other data were needed only to satisfy the query; the user won‟t see any of them and there is no point to transmitting them to the client. This is a clear case for a stored procedure because we can and should maximize performance by performing all operations as close to the data source as possible. Chances are that the entities returned by the stored procedure are entities we already know. That procedure could be just an especially resource-consuming query for Order entities that we retrieve and save in the usual way under normal circumstances. The Stored Procedure Query is perfect for this situation. We define such a query, identify Order as the query return type, and turn it loose on the database. We accept the sproc-selected Order objects and work with them in our typical merry way. Note that a stored procedure query, by its nature, must be executed by the database: we can‟t run it against the entity cache58. So we may not invoke it while the application is running offline.

Accessing Related Entities Via Navigation Properties On Entities Retrieved Using Stored Procedure Queries
When using a stored procedure query, the Entity Framework handles the retrieval of information about related entities differently than it does for normal queries. In the normal case, foreign key values are retrieved and retained with the returned entities. These foreign key values are not exposed as public properties on the returned entities, but they‟re present under the covers. For entities retrieved via stored procedures, the EF does not have sufficient information reliably to identify foreign keys, and so does not retrieve values for any. Recall that in the Entity Framework – in contrast to the behavior DevForce -- all related entities must be retrieved by explicit command. When such command is given, EF always returns the related entities. But for parent entities that were retrieved using stored procedures, it necessarily uses a different (and less performant) process to get the related entities than for entities retrieved using ordinary queries. That is made necessary by the lack of foreign key values on the parent entities.

58

There is an advanced technique for applying a stored procedure query to the cache that we cover briefly in “Advanced Business Object Concepts.”

IdeaBlade DevForce

Business Object Persistence - Advanced

DevForce, in contrast to the EF, retrieves related entities automatically; all you need do is to make reference to them. However, in the case of entities retrieved via stored procedure queries, we had to make a tough call. One choice was to retrieve foreign key values automatically during any stored procedure query. That would produce the simplest and most intuitive behavior on the client: for all entities, retrieval of related entities referenced through navigation properties would be automatic. But of course there was a problem: each foreign key requires an additional round trip to the database from the object server; and there is, of course, a performance price for this. We elected to make the default the more performant choice: unless you ask for them explicitly, we do not retrieve the foreign key values during stored procedure queries. In consequence, by default, references to navigation properties on such entities will return Null Entities. If you know you will need the related entities for entities retrieved using a stored procedure proc, you can get them via the ShouldLoadEntityRefs property on the StoredProcQuery. If you set this property to true – the default is false -- all foreign key properties on the entity are looked up during the initial query, and references to related entities will return the proper entities.

SQL Server Stored Procedure Queries
Suppose your data source table includes a stored procedure named “SalesByYear”. It is defined as follows:

TSQL

ALTER procedure "SalesbyYear" @Beginning_Date DateTime, @Ending_Date DateTime AS SELECT OrderSummary.ShippedDate, OrderSummary.id, "Order Subtotals".Subtotal, DATENAME(yy,ShippedDate) AS Year FROM OrderSummary INNER JOIN "Order Subtotals" ON OrderSummary.Id = "Order Subtotals".OrderSummaryId WHERE OrderSummary.ShippedDate Between @Beginning_Date And @Ending_Date

When included among the items imported into an Entity Data Model, this results in the following Function element in the schema (SSDL) section of the Entity Model file:

XML

<Function Name="SalesbyYear" Schema="dbo" Aggregate="false" BuiltIn="false" NiladicFunction="false" IsComposable="false" ParameterTypeSemantics="AllowImplicitConversion"> <Parameter Name="Beginning_Date" Type="datetime" Mode="In" /> <Parameter Name="Ending_Date" Type="datetime" Mode="In" /> </Function>

To make this convenient available for calling directly off of our EntityManager (as you would equally have to do to make it available on the ADO.NET ObjectContext), you must add a FunctionImport element in the conceptual model (CSDL) section of the Entity Model:

XML

<FunctionImport Name="GetSalesByYear" EntitySet="SalesByYearResults" ReturnType="Collection(IdeaBladeTest1Model.EF.SalesbyYear)"> <Parameter Name="Beginning_Date" Type="DateTime" Mode="In" /> <Parameter Name="Ending_Date" Type="DateTime" Mode="In" /> </FunctionImport>

IdeaBlade DevForce

Business Object Persistence - Advanced

This will cause a C# or VB method to be generated in your EntityManager class by the name you specified, “GetSalesByYear”. Note that the FunctionImport element also specifies the EntitySet into which results returned by the stored proc will be housed: “SalesByYearResults”; and the return type of the method, which will be a collection of SalesByYear entities.

The SalesByYear Entity type must be defined in your conceptual model:

XML

<EntityType Name="SalesbyYear" Abstract="false" ib:PrevName="SalesbyYear"> <Key> <PropertyRef Name="ShippedDate" /> </Key> <Property Name="ShippedDate" Type="DateTime" Nullable="false" /> <Property Name="id" Type="Int64" Nullable="false" /> <Property Name="Subtotal" Type="Decimal" Nullable="false" Precision="19" Scale="4" /> <Property Name="Year" Type="String" Nullable="false" MaxLength="4" /> </EntityType>

The method specified in the conceptual model in the FunctionImport element must be mapped to the Function element in the SSDL that represents the stored procedure. That mapping must, of course, be specified in the mapping (MSL) section of the Entity Model:

XML

<FunctionImportMapping FunctionImportName="GetSalesByYear" FunctionName="IdeaBladeTest1Model.EF.Store.SalesbyYear" />

Having done all of that in your Entity Model, you can now use the resultant method as shown following two examples:

C#

_em1 = new IdeaBladeTest1Entities(); [TestMethod] public void StoredProcQuery() { DateTime dt1 = DateTime.Parse("1/1/1990"); DateTime dt2 = DateTime.Parse("1/1/2000"); var results = _em1.GetSalesByYear(dt1, dt2); }

[TestMethod] public void StoredProcQuery2() { DateTime dt1 = DateTime.Parse("1/1/1995"); DateTime dt2 = DateTime.Parse("12/31/1996"); var results = _em1.GetSalesByYear(dt1, dt2).Where(s => s.Subtotal > 2500); }

IdeaBlade DevForce
VB

Business Object Persistence - Advanced

The method is simply called on the EntityManager with appropriate parameters. It returns an IEnumerable<SalesByYear>, which can be subjected to qualifying filters as you see in the second example above. Below is the Generated code in the domain model designer code file for the GetSalesByYear() method:

C#

#region GetSalesByYear StoredProcQuery /// <summary> /// Constructs and executes the <see cref="T:IdeaBlade.EntityModel.StoredProcQuery"/> /// associated with the given stored procedure. /// </summary> public IEnumerable<IdeaBladeTest1Model.SalesbyYear> GetSalesByYear( Nullable<DateTime> Beginning_Date, Nullable<DateTime> Ending_Date) { StoredProcQuery query = GetSalesByYearQuery(Beginning_Date, Ending_Date); return this.ExecuteQuery<IdeaBladeTest1Model.SalesbyYear>(query); } /// <summary> /// Constructs and returns the <see cref="T:IdeaBlade.EntityModel.StoredProcQuery"/> /// associated with the given stored procedure. /// </summary> public StoredProcQuery GetSalesByYearQuery( Nullable<DateTime> Beginning_Date, Nullable<DateTime> Ending_Date) { QueryParameter Beginning_DateParameter; if (Beginning_Date.HasValue) { Beginning_DateParameter = new QueryParameter("Beginning_Date", Beginning_Date); } else { Beginning_DateParameter = new QueryParameter("Beginning_Date", typeof(DateTime)); } QueryParameter Ending_DateParameter; if (Ending_Date.HasValue) { Ending_DateParameter = new QueryParameter("Ending_Date", Ending_Date); } else { Ending_DateParameter = new QueryParameter("Ending_Date", typeof(DateTime)); } StoredProcQuery query = new StoredProcQuery(typeof(IdeaBladeTest1Model.SalesbyYear), "GetSalesByYear", Beginning_DateParameter, Ending_DateParameter); return query; } #endregion GetSalesByYear StoredProcQuery

VB

For the record, here‟s an alternative way to invoke your stored procedure:

IdeaBlade DevForce
C#

Business Object Persistence - Advanced

[TestMethod] public void StoredProcQuery3() { DateTime dt1 = DateTime.Parse("1/1/1996"); DateTime dt2 = DateTime.Parse("12/31/1998"); StoredProcQuery query = new StoredProcQuery(typeof(SalesbyYear)); // Note that a FunctionImport must be defined in the Entity Model query.ProcedureName = "GetSalesByYear"; query.Parameters.Add(new QueryParameter("Beginning_Date", dt1)); query.Parameters.Add(new QueryParameter("Ending_Date", dt2)); var results = _em1.ExecuteQuery<SalesbyYear>(query); }

VB

Stored Procedure Entity Navigation
Dot Navigation is a bit tricky for business objects that are defined by a stored procedure (sproc entities). If the source class is a sproc entity, the tool can implement the Source.Target navigation property if the target class is a table or view entity. Unfortunately, there is no obvious way to automatically generate the implementation if the target is also a sproc entity. Consider an example. Suppose the source is Customer and the target is Order and both are mapped to stored procedures. In principle we could map the Customer to Order by creating a relation that joins Order.CustomerId to Customer.Id59. We tell the tool “implement this!” Unfortunately, the Object Mapper must give up immediately. The tool knows the signature of the base stored procedure but has no idea how the sproc actually responds to different parameter values. Therefore, it can not invoke the Order‟s underlying stored procedure such that the sproc returns all orders for a given customer. That operation may not even be possible. A developer can interpret the stored procedure well enough to know what call (if any) would do the job. Accordingly, the developer may choose to implement a Customer.Orders property within the custom logic of the Customer class, using a stored procedure query.
The same conundrum confronts us when we devise a relation heading the other direction, from any business object entity to a stored procedure entity. Once again, the Object Mapper does not know how to call the stored procedure so that it returns the objects expected by the source entity type.

Table 10 summarizes the situation.

Table 10. Who writes the navigation property involving a sproc entity.

Navigation property Source Entity Type Sproc Sproc Any type Target Entity Type Table or View Sproc or Web Service Sproc

Relation Written By Tool Developer

59

In fact you can‟t do this within the Object Mapper for reasons we are now discussing.

IdeaBlade DevForce

Business Object Persistence - Advanced

Forced Re-fetch
There are a number of methods that help us re-fetch specific entities from their data sources. Among them are EntityList.ForceRefetch and EntityManger.RefetchEntities<T>. They assume the OverwriteChanges merge strategy but we can give them any of the other merge strategies.
OverwriteChanges replaces the cached entities, overwriting our pending changes. We often want to (a) keep

pending changes but (b) refresh copies of unmodified entities. The PreserveChanges… strategies can help us achieve our purpose.

Table 11. PreserveChanges… strategies in a forced re-fetch

Strategy
PreserveChanges PreserveChangesUnless OriginalObsolete PreserveChangesUpdateOriginal

Description Replace unchanged entities but keep changed entities as they are. Replace unchanged entities and changed entities that are obsolete (i.e., that would fail an optimistic concurrency check if saved now). Replace unchanged entities. Keep changed entities and make them current if they are obsolete by updating their original versions.

Custom Navigation property with Forced Re-fetch
Navigation properties execute according to strategy prescribed by the EntityManager. DefaultQueryStrategy. The default is Normal. We can change it dynamically but the Normal strategy is the best default choice for most applications so let‟s assume we leave it that way. The first time we call the navigation property the PM will get the entities from the data source and put them in the cache. The next time, and every subsequent time, the navigation property will look in the cache first and find the entities there. So during the entire user session these entities may never be refreshed. This is great for a list of states but not so great for more volatile entities such as theater seats. Some developers will be tempted to override the navigation property to get fresh data from the data source every time. The following is a typical example that strives to keep the Customer.Orders ultra-current:

C#

public override ReadOnlyEntityList<Order> Orders { get {return base.Orders. ForceRefetch(MergeStrategy.Overwrite);} }

VB

Public Overrides ReadOnly Property Orders() As _ IdeaBlade.Persistence.ReadOnlyEntityList(Of Order) Get Return MyBase.Orders.ForceRefetch(MergeStrategy.Overwrite) End Get End Property

IdeaBlade DevForce

Business Object Persistence - Advanced

Performance is likely to be terrible. Entity properties fire frequently and sometimes unexpectedly. Properties should return quickly. This one goes to the data source every time. Not good. The intention is laudable and we can make this work. One approach is to remember the last time we invoked this method. If we just did it, return with the most recently fetched list. If we did it “too long ago”, force the re-fetch.

Lost Connection During Query
What if the EntityManager can‟t reach the data source when processing a query60 either because of a network connection problem or because the data source is unavailable? This is a non-issue for the CacheOnly query but applies to all other fetch strategies. The PersisenceManager responds differently depending upon whether or not it knows that the connection is broken before attempting the query. If it knows it is disconnected, its behavior is simple: treat every query as a CacheOnly query. This is consistent with the general principle that writing code for a disconnected application should be as easy as possible. We shouldn‟t have to write a lot of special case logic once we have acknowledged that the application is off-line.

Unexpected loss of connection
When the EntityManager believes it is connected, it will attempt to search the database once the cache proves inadequate. If in fact it is not connected or the connection is broken during the search, the EntityManager will and then raise an event. If the application doesn‟t handle the event, it throws an exception. If the EntityManager “believes” it is connected and discovers that it can‟t reach the data source while processing the query, it will take the following steps in sequence. 6. 7. 8. change its internal state to “disconnected” raise An EntityServerError event throw An EntityServerException unless the event says it handled the problem.

We can and should supply the EntityManager with An EntityServerError event handler. Our handler can quickly tell that the cause is a connection problem. It can distinguish between network connection failure and data source unavailability. If we know what to do, we can do it and signal that we‟ve handled it; the EntityManager won‟t throw an exception. If we don‟t handle the event or don‟t signal that we‟ve handled it, the EntityManager will throw An EntityServerErrorException.

Query Cache
DevForce caches queries to improve performance 61. Consider a query for employees with FirstName = “Nancy”. The QueryStrategy is Normal which means the fetch strategy is CacheThenDataSource. When we execute this query in an empty EntityManager, there will be a trip across the network to fetch the entities from the data source. We get back “Nancy Davolio” and “Nancy Sinatra”. If we execute the query again, the EntityManager satisfies the query from the entity cache and returns the same result; it does not seek data from the data source. During the first run the EntityManager stored the query in its Query Cache62. The second time it found the query in the Query Cache and thus knew it could use apply the cache to the query instead.

60 61

This analysis applies to both entity query and entity navigation. This analysis applies to both entity queries and entity navigation. Both use CacheFirstThenDataSource fetch strategy by default.

IdeaBlade DevForce

Business Object Persistence - Advanced

If we change “Nancy” to “Sue” and run the query again, we get back just “Nancy Sinatra”. If we change “Sally Wilson” to “Nancy Wilson” and run it again, we‟ll get the principals of a strange duet. So far, everything is working fine. Meanwhile, another user saves “Nancy Ajram” to the data source. We run our query again and … we still have just a duet. The EntityManager didn‟t go to the data source so it doesn‟t find the Lebanese pop star. Such behavior may be just fine for this application. If it is not, the developer has choices. She can:    use a QueryStrategy with a different fetch strategy that looks at the database first. clear the query cache explicitly by calling EntityManager.ClearQueryCache clear the query cache implicitly by removing any entity from the entity cache

EntityManager.RemoveEntities Overload Preserves Query Cache
When we remove an entity from a EntityManager‟s entity cache, DevForce automatically clears the PM‟s entire query cache. That‟s right – it erases the EntityManager‟s memory of all the queries it has performed. Suppose we frequently query for employees hired this year. If we issue this query twice. The first query fetches the employees from the database; the second retrieves them from the cache. The second query is almost instantaneous. Then we remove an unrelated entity such as a Customer or an Address. We query again. Instead of reading from the cache as it did before, the PM goes back to the database for these employees. Seems unfair, doesn‟t it? But it‟s the safe thing to do. If we issue the same query multiple times, we expect the same results every time. We expect a different result only if data relevant to our query have changed. The EntityManager will search the local cache instead of the database only if it “believes‟ that all essential information necessary to perform the query are resident in the cache. If it “thinks” that the cache has been compromised, it should go back to the data source to satisfy the query. Removing an entity compromises the cache. For sure it invalidates at least one query – the query that fetched it in the first place. But is that the only invalidated query? The EntityManager does not know. So it does the safe thing and forgets all queries. You and I know (or we think we know) that removing a Customer or Address has no bearing on employees hired this year. The EntityManager is not so sure. There are circumstances when (a) we have to remove an entity and (b) we are certain that no queries will be adversely affected. For example, our query may return entities which we‟ve marked as inactive. We never want inactive entities in our cache but, for reasons we need not explain here, we have inactive entities in the cache. We want to remove those entities. Being inactive they cannot possibly contribute to a correct query result. Unfortunately, removing those entities clears the entire query cache. The EntityManager will satisfy future queries from the database until it has rebuild its query cache. This is not a problem if we rarely have to purge inactive entities. But what if we have to purge them after almost every query63? We will never have a query cache and we will always search the database. The performance of our application will degrade

62

The PersistenceManager stores the query in the query cache when (a) the query is successful and (b) it searched the data source (not just the cache). This is not a rare scenario.

63

IdeaBlade DevForce

Business Object Persistence - Advanced

Fortunately, there is now a RemoveEntities signature that can remove entities without clearing the query cache. In the full knowledge of the risk involved, we can call
EntityManager.RemoveEntities(entitieToRemove, false)

The “false” parameter tells the PM that is should not clear the query cache. Remember: removing an entity and deleting it are different operations. Removing it from the cache erases it from client memory; it says nothing about whether or not the entity should be deleted from its permanent home in remote storage. “Delete”, on the other hand, is a command to expunge the entity from permanent storage. The “deleted” entity stays in cache until the program can erase it from permanent storage.

MergeStrategy In More Detail
The discussion here expands upon that in the section ”Inversion Mode” earlier in the basic topic document for Business Object Persistence. It is provided as a supplement for a deeper understanding of the topic. What happens during the merge of a data source entity and a cached entity depends upon the answers to three crucial questions: 1. 2. 3. Is the entity current or obsolete? How has it changed? Is the entity represented in the data source?

Is the entity current or obsolete relative to the data source?
We compare the cached entity‟s concurrency column property value to that of its data source entity. If the two are the same, the cached entity is current; if they differ, the cached entity is obsolete. As it happens, the cached entity has two concurrency column property values, a current one and an original one. The value of the concurrency column in the current version is meaningless. It‟s the value of the concurrency column in the original version that counts. Every DevForce entity has an original version and a current version of its persistent state. We can get to one or the other by means of a static GetValue() method defined on the EntityProperty class. For example, the following code gets the original value (as retrieved from the database) for the RequiredDate property of a particular Order instance:

C#

DomainModelEntityManager mgr = DomainModelEntityManager.DefaultManager; anOrder = mgr.Orders.Where(o => o.OrderID == 10248); Datetime reqdDate = Order.RequiredDateEntityProperty.GetValue(anOrder, EntityVersion.Original);

VB Both of the following statements get the current value for the same property: C#
reqdDate = Order.RequiredDateEntityProperty.GetValue(anOrder, EntityVersion.Current); reqdDate = anOrder.RequiredDate; // same as above (but simpler!)

VB

IdeaBlade DevForce

Business Object Persistence - Advanced

Again, DevForce and the Entity Framework determine if our cached entity is current or obsolete based on the original version of the property value.

How has it changed?
The merge action depends upon whether the entity was added, deleted, or changed since we set its original version. The entity‟s EntityState property64 tells us if and how it has changed.

Is the entity represented in the data source?
If there is a data source entity that corresponds to the cached entity, we may use the data from data source entity to change the cached entity in some way. If we don‟t find a matching data source entity, we have to decide what to do with the cached entity. Maybe someone deleted the data source entity in which case we might want to discard the cached entity. If we, on the other hand, we want to save the cached entity, we‟ll have to insert it into the data source rather than update the data source.

Merging when the entity is in the data source
We‟ll look at each strategy and describe the outcome based on (a) whether or not the cached entity is current and (b) the entity‟s EntityState. If the entity is Unchanged, we always replace both its original and current versions with data from the data source entity. Our remaining choices are evident in the following table.
Table 12. Merge strategy consequences for a changed cached entity that exists in the data source.

Merge Strategy
PreserveChanges

Current Y N Y or N Y N

Added NC NC OW ---OW NC

Deleted NC NC OW NC OW NC

Detached NC NC OW NC OW NC

Modified NC NC OW NC OW NC

Post Current Y N Y Y Y Y

OverwriteChanges PreserveChangesUnless OriginalObsolete

PreserveChangesUpdateOriginal

Y or N

NC = No change; preserve the current version values of the cached entity OW = Overwrite the cached entity‟s current version values with data from the data source entity Post Current = „Y‟ means the cached entity is “current” relative to the data source after the merge.

There are important artifacts not immediately observable from this table. The entity‟s EntityState may change after the merge. It will be marked Unmodified after merge with OverwriteChanges. It will be marked Unmodified after merge with PreserveChangesUnlessOriginalObsolete if the entity is obsolete. Note that deleted and detached entities are resurrected in both cases.

64

The possible values are Added, Deleted, Detached, Modified, and Unchanged. See “Data Row State” in the glossary.

IdeaBlade DevForce

Business Object Persistence - Advanced

An added cached entity must be deemed “obsolete” if it already exists in the data source 65. We will not be able to insert that entity into the data source; we‟ll have to update the data source instead. The PreserveChangesUpdateOriginal strategy enables us to force our changes into the data source even if the entity is obsolete. An added entity merged with PreserveChangesUpdateOriginal will be marked Modified so that DevForce knows to update the data source when saving it. These effects are summarized in the following table:
Table 13. EntityState after merge.

Merge Strategy
PreserveChanges OverwriteChanges PreserveChangesUnless OriginalObsolete PreserveChangesUpdateOriginal A = Added, D = Deleted,

Current Y or N Y or N Y N Y or N

Added A U --U M

Deleted D U D U D

Detached Dt U Dt U Dt

Modified M U M U M

Dt = Detached,

M = Modified,

U = Unchanged

The merge may change the original version of a changed cached entity to match the data source values.    
PreserveChanges never touches the original version.

The original version is always changed with the OverwriteChanges strategy. It is reset with the PreserveChangesUnlessOriginalObsolete strategy if (and only if) the entity is obsolete..
PreserveChangesUpdateOriginal updates the original version (but not the current version!) if the

entity is obsolete. This step ensures that the cached entity appears current while preserving the pending changes. These effects are summarized in the following table:
Table 14. Merge strategy effect on the original version of the cashed entity.

Merge Strategy
PreserveChanges OverwriteChanges PreserveChangesUnless OriginalObsolete PreserveChangesUpdateOriginal

Current Y or N Y or N Y N Y or N

Added NC OW ---OW OW

Deleted NC OW NC OW OW

Detached NC OW NC OW OW

Modified NC OW NC OW OW

Merging when the cached entity is not in the data source
We begin by considering cached entities that are unchanged. If the query applied to the cache returns an unchanged entity, „X‟, and the query applied to the data source did not return its mate, we can safely assume that „X‟ was deleted after we fetched it. We can remove „X‟ from the cache. We turn next to changed cached entities where we must distinguish between a query that tests only for the primary key and one that tests for something other than the primary key.

65

The entity exists in the data source if the query returns an object with a matching primary key. If we think we created Employee with Id=3 and we fetch one with Id=3, someone beat us to it and used up that Id value. Our entity is obsolete.

IdeaBlade DevForce

Business Object Persistence - Advanced

If the query tests for anything other than the primary key, we can draw no conclusions from the fact that a cached entity was not found in the database. For what does it mean if we have an employee named “Sue” in cache and we don‟t find her in the data source? Perhaps someone deleted her from the data source. Maybe someone merely renamed her. Maybe we renamed her. The combinations are too many to ponder. On the other hand, if we query for Employee with Id = 3 and we don‟t find that employee in the data source, we can be confident of a simple interpretation66. A business object must have unique identity so if it isn‟t there, either it was never there or it has been deleted. What happens next depends upon the EntityState of the cached entity and the merge strategy.  DevForce recovers gracefully when it attempts to save an entity marked for deletion and it can‟t find the data source entity to delete so the merge can leave this cached entity alone. It can also skip over the detached entities.
PreserveChanges forbids merge effects on changed entities. The entity stays put in the cache. OverwriteChanges takes the data source as gospel. If the cached entity‟s EntityState is Modified,

 

there should be an existing data source entity. There is not, so DevForce assumes the data source entity has been deleted and the cache should catch up with this reality. It removes 67 the entity from the cache. On the other hand, if the cached entity is new (Added), we don‟t expect it to be in the data source. The entity remains “as is” in the cache, a candidate for insertion into the data source.   In sum:
Table 15. Merge strategy consequences for a changed cached entity that does not exist in the data source.

PreserveChangesUnlessOriginalObsolete behaves just like OverwriteChanges. PreserveChangesUpdateOriginal strives to position the entity for a successful save. It must intervene to enable data source insertion of a modified entity by changing its EntityState to Added68.

Merge Strategy
PreserveChanges OverwriteChanges PreserveChangesUnlessOriginalObsolete PreserveChangesUpdateOriginal A = Added, M = Modified, R = Removed

Added A A A A

Modified M R R A

DataSourceOnly Subtleties
We may get a nasty surprise if we use a DataSourceOnly or DataSourceThenCache query with other than the OverwriteChanges merge strategy. Consider the following queries using the PreserveChanges merge strategy. Suppose we hold the “Nancy” employee in cache. We change her name to “Sue” and then search the database for all Employees with first names beginning with „S‟. We will not get “Sue” because she is still “Nancy” in the database. Suppose we search again but this time we search for first names beginning with „N‟. This time we get “Sue”. That will confuse the end user but it is technically correct because the “Sue” in cache is still “Nancy” in the database 69.
66

DevForce confirms that the primary key has not changed. While it is good practice to use immutable keys, it is not always so. If the primary key has been changed, DevForce leaves the cached entity alone. Removal from the cache is just that. The entity disappears from cache and will not factor in a save. It does not mean “delete” which requires DevForce to try to delete the entity from the data source. It is an action neutral to the data source.. An update would fail because there is no data source entity to update. DataSourceThenCache will produce the same anomaly for the same reason: the database query picks up the object in the database as “Nancy” but preserves the modification in cache which shows her as “Sue”.

67

68 69

IdeaBlade DevForce

Business Object Persistence - Advanced

The EntityManager.AttachEntity Method
Those of you who write tests and don't want those tests to touch the database will appreciate this method. Here is its signature: C#
AttachEntity(object entity)

As you know, you sometimes need to write tests which rely upon interaction with the EntityManager. You want to populate a disconnected EntityManager with a small collection of hand-rolled stub entities. While such tests are integration tests because they rely on a dependency, we still want to make them easy to write and we want them to be fast. That means we don't want a trip to a database when we run them; we shouldn't need to have a database to run them. I usually start by creating a test-oriented, disconnected EntityManager ... which can be as simple as the following: C#
var testManager = new EntityManager(false /* disconnected */ );

The easiest way to get a stub entity is to "new" it up, set some of its properties, give it an EntityKey, and dump it in our testManager. When we're done it should appear there as an unchanged entity ... as if you had read it from the datastore. The catch lies in the answer to this question: "How do I add the entity to the manager?" In the absence of AttachEntity() method, you would have to use EntityManager.AddEntity(). But after AddEntity, the EntityState of the entity is always "Added". You want a state of "Unchanged" so you have to remember to call AcceptChanges (which changes the state to "Unchanged"). That's not too hard. Unfortunately, it gets messy if the key of the entity is auto-generated (e.g., mapped to a table whose id field is auto-increment) because DevForce automatically replaces your key with a temporary one as part of its auto-id-generation behavior. We could explain how to work around this, but what was really needed was a simple way to simulate the result of retrieving an entity. That's why we created the AttachEntity() method. Here's the XML documentation for AttachEntity: Adds a detached entity to this EntityManager in an Unmodified state. Throws an exception if an entity with the same key already exists in the manager of if the specified entity is not in a detached state.

Let us elaborate here and compare it to some similar methods by calling out some facts about the following code fragment:
C#
theEntityManager.AttachEntity(object theEntity)

  

theEntity‟s EntityKey (“the key”) must be preset prior to the attach operation (which will not touch the key). An exception is thrown if an entity with that key is already in the cache. After attach, theEntity is in an “Unchanged” EntityState (“the state”).

IdeaBlade DevForce
    

Business Object Persistence - Advanced

 

theEntity is presumed to exist in the persistent store; a subsequent change and save will translate to an update statement. After a successful attach, a reference to theEntity is a reference to the entity with that key in the manager‟s EntityCache. Contrast this with the effect of anEntityManager.Imports(new [] {anEntity})” as discussed below. theEntity must be in the “Detached” state prior to the operation. An exception is thrown if theEntity is other than in “Detached” state prior to the operation. After attach, related entities are implicitly associated with theEntity automatically; for example, if anOrder with Id==22 is attached and there are OrderDetails with parent OrderId==22, then after the attach, anOrder.OrderDetails returns these details and any one of them will return „anOrder‟ in response to anOrderDetail.Order. The sequence of attachments is not important; OrderDetails may be added prior to the parent Order. Attach has no effect on theEntityManager‟s QueryCache.

AddEntity behaves the same way as AttachEntity except as follows:    After add, theEntity is in an “Added” state theEntity is presumed to be new and to be absent from in the persistent store; a save will translate to an insert statement. If the key for this type is auto-generated (e.g., backed by an auto-increment field in the database), the existing key will be set to a generated temporary key, replacing the prior key value.

The following is true regarding detaching anEntity:     After detach, anEntity enters the “Detached” state no matter what its prior state. Detaching an Order does not detach its child OrderDetails; they remain “orphaned” in the cache. The sequence of detachments is not important; an Order may be detached prior to detaching its child OrderDetails. Detach has no effect on theEntityManager‟s QueryCache.

EntityManager.Imports is another way of populating an EntityManager with a collection of entities that may have come from anywhere (including hand-rolled). Here's how you might "import" a single stub entity: C# theEntityManager.Imports(new [] {theEntity}) ;

Imports differs from AttachEntity in that:     It requires a MergeStrategy to tell it what to do if an entity with the same key as "theEntity" already exists in the cache. It merges "theEntity" into the cache based on the MergeStrategy It makes a clone of "theEntity" and adds that clone to the EntityCache ... unless "theEntity" happens to already be in the cache in which case it is ignored ... which means that Using our example and assuming that "theEntity" was not already in the manager, the entity instance in the cache is not the same as the entity instance you imported, although their keys are equal; the following is true: C#
theEntity != theManager.FindEntity(theEntity.EntityAspect.EntityKey)

A "clone" is a copy of an entity, equivalent to calling the following:

IdeaBlade DevForce
C#

Business Object Persistence - Advanced

((ICloneable)theEntity).Clone();

This is a copy of the entity, not of its related entities.

Filtering Queries
DevForce provides an extension method, Filter(), that can be used to superimpose one or more independently defined filter conditions upon an existing query. Filter() differs from Where() in that it can apply a condition defined independent of the targetted query. Filter()‟s primary motivating use case is the need to apply server-side filters to submitted queries in a handler for the Server.Fetching event; though it is perfectly possible to use it in other contexts. For example, suppose your application‟s database includes data for customers worldwide, but that a given Sales Manager only works with data for customers from his region. Instead of baking the region condition into every query for Customers throughout your application, you could implement a ServerFetching handler that imposes the condition upon any query for customers made while that Sales Manager is logged in. The usefulness of Filter() becomes even more apparent when you need to apply filters in a global way for more than one type. There are four overloads of Filter(), two of which are generic, and two of which are not. Each pair includes one overload that takes a Func<T> and another that takes an EntityQueryFilterCollection (each of whose members is a Func<T>). The generic versions normally get used client-side, because they normally operate upon an EntityQuery<T>, whereupon.NET uses type inference to get T and route the call through the generic signature. The non-generic versions are necessary because, server-side, DevForce has access only to an EntityQuery, not an EntityQuery<T>; that being a consequence of the .NET constraint that generic types can‟t be passed in event arguments. Let‟s look at some examples:

C#

var query = _em1.Territories.Where(t => t.Id > 100); var newQuery = query.Filter((IQueryable<Territory> q) => q.Where(t => t.Description.StartsWith("M")));

In this example we have used the overload of Filter which is non-generic, and which takes as its argument a Func delegate. Said delegate takes an IQueryable<T> -- essentially a list of items of type T – and returns an IQueryable<T>. The IQueryable<T> that goes in is the one defined by the variable query, defined as

C#

_em1.Territories.Where(t => t.Id > 100)

The one that comes out is the one that went in minus those Territories whose Description property value begins with the letter “M”. In the first example, above, our filter applies to the query‟s root type, Territory. We aren‟t limited to that: we can also apply filters to other types used in the query. Consider the following:

IdeaBlade DevForce
C#

Business Object Persistence - Advanced

var q1 = _em1.Customers.SelectMany(c => c.OrderSummaries .Where(o => o.ShipCity.StartsWith("N")) ); var q1a = q1.Filter((IQueryable<OrderSummary> q) => q.Where(o => o.Freight > maxFreight));

The root type for this query is Customer, but the query projects OrderSummaries as its output, and it is against OrderSummaries that we apply our filter. Again we use the non-generic form of Filter; and again, the overload that takes a Func<T> argument. This time the filter imposes a condition upon the values of the OrderSummary.Freight property. Without the filter we would have retrieved all OrderSummaries having a ShipCity whose name begins with “N”; with the filter, not only must the name begin with “N”, but the Freight property value must exceed the value maxFreight. Let‟s look at another example of filtering one some type other than the query‟s root type:

C#

var q1 = _em1.Customers.Where(c => c.OrderSummaries.Any(o => o.ShipCity.StartsWith("N"))); var q1a = q1.Filter((IQueryable<OrderSummary> q) => q.Where(o => o.Freight > maxFreight));

In the absence of the filter, the above query would retrieve Customer objects: specifically, Customers having at least one Order whose ShipCity begins with the letter “N”. The filter potentially reduces the set of Customers retrieved by imposing an additional condition on their related OrderSummaries (again, on the value of their Freight property). Now let‟s look at a use of Filter() involving conditions on more than a single type.

C#

var eqFilters = new EntityQueryFilterCollection(); eqFilters.AddFilter((IQueryable<Customer> q) => q.Where(c => c.Country.StartsWith("U"))); eqFilters.AddFilter((IQueryable<OrderSummary> q) => q.Where(o => o.OrderDate < new DateTime(2009, 1, 1))); var q0 = _em1.Customers.Where(c => c.OrderSummaries.Any(o => o.ShipCity.StartsWith("N"))); var q1 = q0.Filter(eqFilters);

In the above snippet, we instantiate a new EntityQueryFilterCollection, to which we then add two individual filters, each of which is a Func<T>. The first filter added imposes a condition on the Customer type; the second imposes a condition on the OrderSummary type. Note that we could now apply these filters to any query whatsoever. If the targetted query made use of the Customer type, the condition on Customers would apply; if it made use of the OrderSummary type, the condition on OrderSummaries would apply. If it made use of both, as does our example q0, both conditions would apply. A filter is also applied directly to any clause of a query that returns its targetted type. Thus, the effect of the two filters defined above, applied against query q0, is to produce a query that would look like the following if written conventionally:

C#

var q0 = _em1.Customers .Where(c => c.Country.StartsWith("U")) .Where(c => c.OrderSummaries .Where(o => o.OrderDate < new DateTime(2009, 1, 1)) .Any(o => o.ShipCity.StartsWith("N")));

IdeaBlade DevForce

Business Object Persistence - Advanced

Query Inversion in More Detail
The discussion here expands upon that in the section “InversionMode” earlier in this chapter. It is provided as a supplement for a deeper understanding of the topic.

Interaction of the FetchStrategy and the InversionMode
Consider the query shown below. For this query, we have custom-baked a QueryStrategy so we can experiment with various FetchStrategies and InversionModes. The collection against which the query is directed is _Em1.Customers; but then it uses the SelectMany() method to project Order objects into the result set. Since its return type is different from the type contained in the collection first referenced, the query is non-invertible. var query = _Em1.Customers .Where(c => c.CustomerID == "CONSH") .SelectMany(c => c.Orders); QueryStrategy aQueryStrategy = new QueryStrategy(FetchStrategy.DataSourceThenCache, MergeStrategy.PreserveChanges, InversionMode.On); query.QueryStrategy = aQueryStrategy; foreach (Order anOrder in query) { System.Diagnostics.Debug.WriteLine(anOrder.OrderDate.ToString()); } Assert.IsTrue(query.ToList().Count > 0, "should return orders"); VB In our initial run, we have the InversionMode set to On. Because DevForce is unable to invert the query, a QueryInversionServerException is thrown, with the following message: This query is not automatically invertible and cannot be executed unless either its QueryInversionMode is set to 'Manual' or its FetchStrategy is set to Optimized, DataSourceOnly or CacheOnly. If we change the InversionMode to Try and rerun the query, it runs without an exception, but the Assert test fails, because no Orders were included in the result set. Why? Because changing the InversionMode from On to Try didn‟t alter the fact that the query couldn‟t be inverted; it just told DevForce not to worry about that fact. The result set returned with a FetchStrategy of DataSourceThenCache is only that obtained in a final query against the cache, after entities retrieved from the data source have been placed there. Since the query was not invertible, no Customer objects were retrieved into the cache, and that final query returns an empty result. Suppose now we set the FetchStrategy to DataSourceAndCache. Now references to the Order objects retrieved from the data source are included in the result set. A second application of the query, this time against the cache, may or may not pick up additional Orders70. But in any event, the final result set will contain references to the in-cache Orders that are linked to the specified Customer. This will be true even if, at the end of the process, there are still no Customer objects in the cache!
70

C#

It will pick up additional Orders if there are Orders in the cache that are (a) linked to Customer “CONSH”, and (b) either do not exist in the data source, or are not linked to Customer “CONSH” in the data source

IdeaBlade DevForce

Business Object Persistence - Advanced

When a query cannot be inverted, a FetchStrategy other than DataSourceThenCache should be used. Table 16 shows the combinations of FetchStrategy and InversionMode that lead to exceptions. Note that these exceptions are designed to prevent you from receiving query results that, although they may look perfectly valid, are not!
Table 16. FetchStrategy x InversionMode - Exception Behavior

FetchStrategy CacheOnly DataSourceOnly DataSourceOnly DataSourceOnly DataSourceOnly DataSourceThenCache DataSourceThenCache DataSourceThenCache DataSourceThenCache Optimized Optimized Optimized Optimized DataSourceAndCache DataSourceAndCache DataSourceAndCache DataSourceAndCache

InversionMode NA On Try Off Manual On Try Off Manual On Try Off Manual On Try Off Manual

QueryInversionServerException Never If query requires inversion and cannot be inverted Never Never Never If query requires inversion and cannot be inverted If query requires inversion and cannot be inverted If query requires inversion Never If query requires inversion and cannot be inverted Never Never Never If query requires inversion and cannot be inverted Never Never Never

Only queries that either have been inverted or do not require inversion are saved in the query cache.

Turning a Non-Invertible Query on Its Head
Note that the previous query (for Orders placed by Customer “CONSH”) can be rewritten as follows:
var query = _Em1.Orders .Where(o => o.Customer.CustomerID == "CONSH");

C# VB

This form of the query, unlike the other one, is invertible.

A Special Case: Using the Skip() Method on an EntityQuery
The query below uses the DataSourceOnly QueryStrategy in combination with a call to Skip().

IdeaBlade DevForce

Business Object Persistence - Advanced

C#

EntityQuery<Customer> customersQuery = _Em1.Customers .Where(c => c.ContactTitle == "Sales Representative") .OrderBy(c => c.CompanyName); customersQuery.QueryStrategy = QueryStratey.DataSourceOnly; // <--note! ICollection<Customer> customers = customersQuery.Skip(5).Take(5).ToList();

VB

You can easily get results that are not what you would expect if you do not specify the QueryStrategy when using Skip. Suppose we omitted the statement in the above example that specifies the QueryStrategy:

C#

EntityQuery<Customer> customersQuery = _Em1.Customers .Where(c => c.ContactTitle == "Sales Representative") .OrderBy(c => c.CompanyName); ICollection<Customer> customers = customersQuery.Skip(5).Take(5).ToList();

VB In the above case, DevForce would use the EntityManager‟s default QueryStrategy, which (unless you had changed it) would be QueryStrategy.Normal. Recall that QueryStrategy.Normal uses a FetchStrategy of DataSourceThenCache, and that the latter returns a list of references obtained in a final, cache-only query. So here‟s the flow of events for the above query. (Assume an empty cache as a starting point.) 1. 2. 3. 4. Query is submited to the EntityManager. EntityManager checks the query cache to see if query has been submitted before. It finds that it has not. EntityManager submits query against the data source, which returns five Customers, which are placed in the cache. EntityManager submits the query again, this time against the cache (so that it will incorporate any Customers who have been added locally but have not yet been saved to the data source).

The second query, against the cache, skips the five Customers it finds there, and upon attempting to take the next five, discovers that there are no more. It therefore returns 0 Customers. Although this isn‟t, technically, a case of a failed query inversion, the result and the reason for it are clearly similar to that. The only real advice here is that, if you‟re using Skip(), you should either use a FetchStrategy of DataSourceOnly, or make good and certain that you understand FetchStrategies in detail.

DataSourceThenCache Versus DataSourceAndCache
The distinction between the DataSourceThenCache and DataSourceAndCache strategies is subtle but important in the case of queries that must process non-targeted types and are therefore subject to query inversion. Suppose you were to submit the following query:

IdeaBlade DevForce
C#

Business Object Persistence - Advanced

var query = em1.Customers .Where(c => c.Orders .Any(o => o.OrderDate.HasValue == true && o.OrderDate.Value.Year == 1997));

VB This query targets Customers but must process Orders to find the correct set of Customers. DevForce would have no difficulty inverting this query, but suppose you submitted it with an InversionMode of Off and a FetchStrategy of DataSourceThenCache. The InversionMode setting would mean that only Customer objects were retrieved into the cache: no Order objects. “Great!” you say. “That‟s all I wanted: Customers.” But even though you have the desired Customers in your cache, you don‟t yet have references to them. How does DevForce get these references? Because of the FetchStrategy you specified, DevForce now resubmits your query, this time against the cache; and the set of references to Customer objects that it will return will be entirely determined by the Customers that meet the query criteria when the query is resubmitted against the cache. But wait! There is no guarantee that the cache contains the same Order objects that were found in the data source; it will, in fact, contain no Order objects at all unless some other, unrelated operation that was previously executed caused some to be retrieved. Therefore the set of Customers found by the query when submitted against the cache may be very different from the set found when it was submitted against the data source. Indeed, the set may be empty. You may get references to no Customers or some Customers, but there is no guarantee, and indeed little likelihood, that you‟ll get references to all of the Customers retrieved by your query from the data source. If, on the other hand, you submitted your query with a FetchStrategy of DataSourceAndCache, you‟ll get want you wanted: all Customers in the data source who meet your conditions, as well as all Customers that exist only in your local cache that meet those conditions. With that FetchStrategy, DevForce performs a union of the references obtained by the two query submissions. The DataSourceAndCache FetchStrategy does have some drawbacks which we‟ll discuss momentarily. Generally speaking, it is the appropriate FetchStrategy only in the following circumstance: 1. 2. Your query will use related objects; You want to include in the result set references to entities that exist in the cache but which have not yet been persisted to the database; DevForce can‟t invert the query; and You can‟t write an equivalent query that is invertible.

3. 4.

The reason that DataSourceThenCache is the preferred FetchStrategy for other circumstances is that, under certain circumstances, DataSourceAndCache can produce confusing results. Suppose you have some Customer objects in the cache, including Customer XYZ, and you submit a DataSourceAndCache query for Customers with Orders in the current year. Customers meeting this condition are fetched from the data source into the cache, and merged there with Customers already residing in the cache with a MergeStrategy of PreserveChanges. Meanwhile DevForce hangs on to a list of references to the objects just fetched. Now it so happens that Customer XYZ, who was in the cache already, had (during the current application session) just cancelled their one and only order for the current year. The Order was marked for deletion, but this change had not been committed to the database when the query was submitted. So, based on the state of data in the data source, Customer XYZ met the query conditions and was retrieved, and a reference to their object in the cache was included in the set returned by the query against the datasource. DevForce continued on, resubmitting the query against the cache. This time Customer XYZ did not make the cut because, according to the data in the cache, they did not have a current year Order. No reference to their in-cache object was included in the list of pointers resulting from the query against the cache.

IdeaBlade DevForce

Business Object Persistence - Advanced

But DataSourceAndCache, DevForce then performed a UNION of the references obtained in the query against the data source and those obtained in the query against the cache. A reference to cached Customer XYZ therefore ended up in the result set returned by the query. Your app happily filled a datagrid with the returned Customers, and there sat Customer XYZ, even though they (quite visibly) did not have an order in the current year! Can a phone call from your end user be far away? The DataSourceThenCache FetchStrategy, by contrast, would have retrieved, from the data source and into the cache, whatever data met the query conditions. It would then have submitted the query against the cache, and only the Customers meeting the specified condition in that final query would have been included in the returned result set. Customer XYZ, having been found to have no current year Order, would have been excluded, properly.

Transactional Queries
DevForce query requests are atomic: the developer can issue only one (synchronous) query request at a time. But when the request resolves into multiple SQL queries, they can all be performed together within the same transaction. Individual query requests resolve into several SQL queries when the query has includes that fetch related objects or when the query includes one or more sub-queries and “query inversion” is turned on. When the root query is performed transactionally, both the main select and the selection of related entities occur within transactional boundaries.

DevForce developers can set the transaction isolation level on individual commands
Developers can set the transaction isolation level for individual queries and saves.

Implementation
There is a TransactionSettings class and a TransactionSettings property on both the SaveOptions and QueryStrategy classes. The TransactionSettings class provides the ability to dynamically set:    whether or not to use the Microsoft Distributed Transaction Coordinator the Transaction Isolation level of a Save or Query. This provides in effect the Transaction timeout to be applied to a Save or Query

Note: For the current version Transaction isolation levels and timeouts can only be applied if the DTC is turned on. Note: The Default Transaction Isolation Level for Saves is “Serialized”; for queries it is “ReadCommitted”. Note: Non-locking queries can be implemented by setting the TransactionSettings.IsolationLevel to “ReadUncommitted”.

DevForce and Data Sources – Deep Dive
There are potentially many data sources at play in a DevForce application. Data sources can be databases or web services. The DevForce Object Mapper does not do design-time access to databases. Any design-time access of a database is initiated by Visual Studio‟s Entity Data Model Designer during your design session using that tool. That designer associates an app.config file with the Entity Data Model (.edmx) file, placing it in the same project as the latter. That app.config contains connection information to the database used by the EDM designer for its design work. When you direct the DevForce Object Mapper to generate code, it creates (or updates) an instance of app.config in the Visual Studio project where it stores the DomainModel (.ibedmx) file. In creating an edmKey element in the

IdeaBlade DevForce

Business Object Persistence - Advanced

app.config for an Entity Data Model‟s database, the Object Mapper simply copies the connection information found in the EDM‟s app.config. Listing 4 shows the XML Schema element from an Entity Data Model (.edmx) file. Typically, this element is generated initially by the EDM Designer, then modified slightly the DevForce Object Mapper. Note namespace and two attributes prefixed with “ib”. These were written into the .edmx file by the DevForce Object Mapper. They are respected by the EDM Designer, however, and it will not overwrite them even if you use it to generate fresh EDM code later.
Listing 4. DataSourceKey attribute in the Entity Data Model (.edmx) file <Schema Namespace="ServerModelNorthwindIB" Alias="Self" xmlns="http://schemas.microsoft.com/ado/2006/04/edm" XML xmlns:ib="http//www.ideablade.com/schemas/edmx" ... ib:DataSourceKey="Default" ib:LastModTs="7/3/2008 12:54:54 PM">

When you save your work in the DevForce Object Mapper , it writes (subject to your okay) a similar connection string into the app.config file that it saves in the DomainModel project. It writes this information as part of an edmKey (for relational database sources) or a wsKey (for web service sources). Listing 5 shows the XML statement written by the Devforce Object Mapper into its app.config for the same object model just referenced:
Listing 5. Data source identifier (edmKey) for run-time operations (written to the app.config file)

XML

<ideaBlade.configuration version="5.00" updateFromDomainModelConfig="Ask" ...
<edmKeys> <edmKey connection="metadata=res://ServerModelNorthwindIB/ServerModelNorthwindIB.csdl|res://Serve rModelNorthwindIB/ServerModelNorthwindIB.ssdl|res://ServerModelNorthwindIB/ServerModelNor thwindIB.msl;provider=System.Data.SqlClient;provider connection string=&quot;Datasource=.;Initial Catalog=NorthwindIB;Integrated Security=True;MultipleActiveResultSets=True&quot;" containerName="ServerModelNorthwindIB.ServerModelNorthwindIBContext" logTraceString="false" name="Default" tag=""> <probeAssemblyNames> <probeAssemblyName name="DomainModel" /> <probeAssemblyName name="ServerModelNorthwindIB" /> </probeAssemblyNames> </edmKey> </edmKeys>

... </ideaBlade.configuration>

Observe the connection information for the Entity Data Model and its datasource, and the DataSourceKeyName, stored as the name attribute of the edmKey. DataSourceKeys originally written by the Object Mapper into the app.config file, such as the one just shown, may subsequently be altered or removed manually by the developer. Other DataSourceKeys may also be added manually. Why might a developer alter or add a DataSourceKey in app.config? The most common reason would be that he wants to add keys that point to multiple variations of a particular Datasource (e.g., Development, Test, Production).

IdeaBlade DevForce

Business Object Persistence - Advanced

The Object Mapper and Manually Added or Modified Keys
If your Domain Model project includes an existing app.config file and you add or modify Entity Data Models or their DataSourceKey names inside the Object Mapper and save your work, the Object Mapper will ask if you want it to update the app.config. Basically this update consists of modifying the edmKeys in the app.config file. The Object Mapper will only modify edmKeys in the app.config file with names that match those that it displays in its designer, and only if you accept its offer to do the update.

DataSourceKeys, DataSourceKeyResolvers, and DataSourceExtensions
A DataSourceKey is a symbolic representation of a data source used by the DevForce EntityManager and associated with the Entity objects it retrieves, updates, and creates. Every Entity has a “DataSourceKeyName” attribute that identifies its symbolic Datasource71. This key name is hard-coded into the business class at the time the latter is generated by the Object Mapper. Recall that a DomainModel, and therefore an EntityManager can access multiple data sources. A given EntityManager might, for example, access a SQL Server database, an Oracle database, and a web service, mapping business classes from each and joining all into a single transactional unit. Each of those three datasources gets a distinct DataSourceKey, and entities generated from each of them get assigned the name of that DataSourceKey. But what if you need multiple versions of those three data sources? For example, you might have Development, Test, Stage, and Production versions of the same three-datasource set. The data sources in all four versions would have the same schemas, but different content. For example, data in the development and test data sources might be “scrubbed” so as to eliminate security issues during relatively unprotected use; data in the development data sources might be lightweight compared to that in the Test data sources; and so forth. All four versions of a given database (schema) in a set of data sources would be identified with the same DataSourceKey and all would map to the same set of business classes, so that an application consuming their data would be indifferent to which of the physical instances of that schema it accessed in any given launch. DevForce uses a string called a DatasourceExtension to discriminate between alternative instances of a given data schema. You supply these extensions in the edmKeys (and possibly in wsKeys) that you configure in the app.config file, by adding them to the name attribute of the edmKey or wsKey, separating them from the DataSourceKey Name by an underscore character (“_”). At runtime, to obtain the data required for business objects of a designated type (e.g, Employees), DevForce connects to an actual data source by consulting a DataSourceKeyResolver. The DataSourceKeyResolver combines the DataSourceKeyName associated with the desired Entity type with a DataSourceExtension (supplied by the requesting EntityManager) and returns a DataSourceKey object. That DataSourceKey object contains all the information required to connect to an actual, deployed data source.

EntityManagers and DataSourceExtensions
Every EntityManager gets associated at instantiation with a “DataSourceExtension”. You can see this clearly in the following code statement which uses an overload of the EntityManager constructor that specifies the extension explicitly (as “Development”):

C#

DomainModelEntityManager mgr = new DomainModelEntityManager(true, "Development");

71

You can find this on an entity instance as its EntityAspect.EntityMetadata.DataSourceKeyName property

IdeaBlade DevForce

Business Object Persistence - Advanced

The extension determines which version – e.g., Development, Test, Stage, or Production – of the data source(s) actually gets accessed by the EntityManager. Expressed another way: the “Extension” identifies a collection of one or more data sources, each the repository of a set of tables or web services that map to business object 72 classes, which will be accessed by a given EntityManager. In the illustration below, all four of the DS#1 data sources would have the same DataSourceKeyName. The same could be said for the DS#2 and DS#3 data sources. On the other hand, the set of data sources accessed by a single EntityManager would comprise a DS#1, a DS#2, and a DS#3. But which copy of DS#1, a DS#2, and DS#3 should be used? That would be determined by the DataSourceExtension with which the EntityManager was associated at instantiation.

Now let‟s look at DataSourceKey names and extensions as they appear in edmKeys and wsKeys in an App.config file. Listing 6 is an excerpt from an app.config file containing multiple DataSourceKeys with different key names and extensions. For clarity, we‟ve made sure the name attribute is the first attribute listed for the the <edmKey> element. Note that each DataSourceKey, in addition to containing a connection string, also includes probe assembly names for assemblies that hold auxiliary classes for id generation, authentication, event handling, and the like. 73
Listing 6. Extract of app.config file with multiple DataSourceKeys

XML

<edmKeys> <!-- Production databases --> <edmKey name="NorthwindIB_Release" connection="metadata=res://ServerModelNorthwindIB/ServerModelNorthwindIB.csdl|res://Serve rModelNorthwindIB/ServerModelNorthwindIB.ssdl|res://ServerModelNorthwindIB/ServerModelNor thwindIB.msl;provider=System.Data.SqlClient;provider connection string=&quot;Data Source=ProductionDBMS_A;Initial Catalog=NorthwindIB;Integrated Security=True;MultipleActiveResultSets=True&quot;" containerName="ServerModelNorthwindIB.ServerModelNorthwindIBContext" logTraceString="false" tag=""> <probeAssemblyNames> <probeAssemblyName name="DomainModel" /> <probeAssemblyName name="ServerModelNorthwindIB" /> </probeAssemblyNames> </edmKey> <edmKey name="Aw2000_Release" connection="metadata=res://ServerModelAw2000/ServerModelAw2000.csdl|res://ServerModelAw20 00/ServerModelAw2000.ssdl|res://ServerModelAw2000/ServerModelAw2000.msl;provider=System.D ata.SqlClient;provider connection string=&quot;Data Source=ProductionDBMS_B;Initial

72

The default extension, incidentally, is no extension at all. If you create a new DevForce DomainModel and let the Object Mapper write the edmKey entry into the configuration file, the DataSourceKey will be entered with a name of “Default”, without an extension. It may also contain a <tag>, where you can put any sort of string-value custom information you desire. At runtime you can access the information placed there via the Tag property of a DataSourceKey object -- which you can get from the DataSourceKeys collection of a DataSourceKeyResolver object.

73

IdeaBlade DevForce

Business Object Persistence - Advanced

Catalog=AdventureWorks2000;Integrated Security=True;MultipleActiveResultSets=True&quot;" containerName="ServerModelAw2000.ServerModelAw2000Context" logTraceString="false" tag=""> <probeAssemblyNames> <probeAssemblyName name="DomainModel" /> <probeAssemblyName name="ServerModelAw2000" /> </probeAssemblyNames> </edmKey>

<!-- Development databases --> <edmKey name="NorthwindIB_Development" connection="metadata=res://ServerModelNorthwindIB/ServerModelNorthwindIB.csdl|res://Serve rModelNorthwindIB/ServerModelNorthwindIB.ssdl|res://ServerModelNorthwindIB/ServerModelNor thwindIB.msl;provider=System.Data.SqlClient;provider connection string=&quot;Data Source=DevelopmentDBMS_A;Initial Catalog=NorthwindIB;Integrated Security=True;MultipleActiveResultSets=True&quot;" containerName="ServerModelNorthwindIB.ServerModelNorthwindIBContext" logTraceString="false" tag=""> <probeAssemblyNames> <probeAssemblyName name="DomainModel" /> <probeAssemblyName name="ServerModelNorthwindIB" /> </probeAssemblyNames> </edmKey> <edmKey name="Aw2000_Development" connection="metadata=res://ServerModelAw2000/ServerModelAw2000.csdl|res://ServerModelAw20 00/ServerModelAw2000.ssdl|res://ServerModelAw2000/ServerModelAw2000.msl;provider=System.D ata.SqlClient;provider connection string=&quot;Data Source= DevelopmentDBMS_B;Initial Catalog=AdventureWorks2000;Integrated Security=True;MultipleActiveResultSets=True&quot;" containerName="ServerModelAw2000.ServerModelAw2000Context" logTraceString="false" tag=""> <probeAssemblyNames> <probeAssemblyName name="DomainModel" /> <probeAssemblyName name="ServerModelAw2000" /> </probeAssemblyNames> </edmKey> </edmKeys> <wsKeys> <wsKey url="http://api.google.com/search/beta2" endpointName="GoogleSearchPort" name="GoogleSearch" tag=""> <probeAssemblyNames> <probeAssemblyName name="DomainModel" /> </probeAssemblyNames> </wsKey> </wsKeys>

In the above excerpt from an app.config file, edmKeys are present for two databases (NorthwindIB and Adventureworks2000). Two versions (Development and Release) are maintained of these databases. A wsKey is present for a Google web service: the same service is used for Development and Production. Instantiating an EntityManager and specifying a DataSourceKey Extension of “Release”...

IdeaBlade DevForce

Business Object Persistence - Advanced

mPersMgr = new DomainModelEntityManager(true, "Release");

…would cause all data accesses for entities based on the databases to go against the sources named in the edmKeys that have the suffix “_Release” in their name attribute. For example, data for classes mapped to tables in the NorthwindIB database would be retrieved from the copy of that database running on the ProductionDBMS_A instance of SQL Server; data for classes mapped to the AdventureWorks2000 database would be retrieved from the copy of that database running on the ProductionDBMS_B instance of SQL Server; and data for classes mapped to the Google web service would be accessed via the service addressable at the URL http://api.google.com/search/beta2. Were an EntityManager to be instantiated with the extension “Development”, different copies of the two databases would be accessed. Note the following points: 1. 2. The DataSourceKey Names and DataSourceKey Extensions are case insensitive. In the name attribute of the edmKey element, the Datasource Extensions are always preceded by an underscore “_” character.

If you wished to establish one or the other set of databases as the default – say, the Development versions – then you could include in the <edmKeys> section of the app.config an additional pair of edmKeys with no extensions specified in their names, as shown below. The information in these keys would be used by any EntityManager instantiated with no DataSourceExtension specified. (This time, for brevity, we‟ve snipped out the detail for the connection attribute value.)
Listing 7. Extract of app.config file with multiple DataSourceKeys

XML

<!-- Default databases --> <edmKey name="NorthwindIB" connection="..." containerName="ServerModelNorthwindIB.ServerModelNorthwindIBContext" logTraceString="false" tag=""> <probeAssemblyNames> <probeAssemblyName name="DomainModel" /> <probeAssemblyName name="ServerModelNorthwindIB" /> </probeAssemblyNames> </edmKey> <edmKey name="Aw2000" connection="..." containerName="ServerModelAw2000.ServerModelAw2000Context" logTraceString="false" tag=""> <probeAssemblyNames> <probeAssemblyName name="DomainModel" /> <probeAssemblyName name="ServerModelAw2000" /> </probeAssemblyNames> </edmKey>

Tenant Extensions
Extensions are also a good way to segment data sources by client in a “multi-tenant application”. Multi-tenant applications are typical of Application Service Provider (ASP) scenarios in which each customer‟s data is managed in isolated data sources. When the user logs in, the application identifies the user‟s parent customer and knows which set of Datasources is appropriate for that user. The application can then instantiate an EntityManager that draws upon just those data sources.

IdeaBlade DevForce

Business Object Persistence - Advanced

The “DataSourceExtension” is the ideal representation for a customer-specific data source set as in this depiction of a three-tenant scenario with customers “A”, “B”, and “C”:

Multi-Part Extensions
DataSourceExtensions may have multiple parts, permitted an even more sophisticated scheme for selected a data source instance at runtime. Consider the following edmKeys in an app.config file (connection value and probe assembly section removed for brevity):

XML

<edmKeys> <!-- Production databases --> <edmKey name="Acmetest2_SQLSRVR_OLE1" connection="... " containerName="ServerModelAcmeTest.ServerModelAcmeTestContext" logTraceString="false" tag=""> ... </edmKey> <edmKey name="Acmetest2_SQLSRVR_OLE2" connection="..." containerName="ServerModelAcmeTest.ServerModelAcmeTestContext" logTraceString="false" tag=""> ... </edmKey> <edmKey name="Acmetest2_SQLSRVR" connection="... " containerName="ServerModelAcmeTest.ServerModelAcmeTestContext" logTraceString="false" tag=""> ... </edmKey> <edmKey name="Acmetest2" connection="..." containerName="ServerModelAcmeTest.ServerModelAcmeTestContext" logTraceString="false" tag=""> ... </edmKey> </edmKeys>

Note that the first two edmKey names contain two underscores. These delimit multi-part DataSourceExtensions. Were you to instantiate an EntityManager as follows: C#
mPersMgr = new DomainModelEntityManager(true, "SQLSRVR_OLE1");

…you would get the database identified in the first edmKey in the above excerpt. On the other hand, if you wrote this statement: C#
mPersMgr = new DomainModelEntityManager(true, "SQLSRVR_FOO");

IdeaBlade DevForce

Business Object Persistence - Advanced

…then the DataSourceKeyResolver, being unable to locate an edmKey with both parts of the extension matching, would resolve the database to the one identified with the third edmKey, named “AcmeTEst2_SQLSRVR”. It would do so because it finds a match on the first part of the extension. If the DataSourceKeyResolver can find no edmKey with an extension that matches at least on the first part of the extension submitted, it will throw an exception. Thus, the following statement, containing a misspelled first part of the extension, will result in an exception. It will find no matching set of extensions; no matching first extension; and will not default to the extensionless key “AcmeTest2”: C#
mPersMgr = new DomainModelEntityManager(true, "SQLSVRR_OLE1");

Extensions and EntityServers
Let‟s stick with the multi-tenant, ASP scenario for awhile. When the application client determines the customer, it creates an EntityManager dedicated to the data sources applicable to that customer by including the customer‟s “DatasourceExtension” name in the constructor. C#
msManager = new DomainModelEntityManager(true, "A"); // Connect to customer "A"

Now the client application tries to login or fetch entities with this EntityManager. The EntityManager contacts the EntityService. The EntityService checks among its EntityServers for one that is associated with extension “A”. It doesn‟t find one so it creates a new EntityServer instance for extension “A” and adds it to its collection. This EntityServer now serves every EntityManager presenting the “A” extension. When the EntityService encounters EntityManagers with unknown extensions – “B” and “C” for example –, it creates more EntityServers. The three-tenant scenario could look like this:

Dynamic DataSourceKeys and the DataSourceKeyResolver
Every entity has a “DataSourceKeyName” which identifies its symbolic data source. There should be at least one real data source somewhere that holds the data source object to which the entity is mapped. The “DataSourceKeyName” helps DevForce find it. The DataSourceKeyName property of the entity reveals this name; for example: anEmployee.EntityAspect.EntityMetadata.DataSourceKeyName. DevForce connects an actual data source at runtime by asking a DataSourceKeyResolver for the DataSourceKey that corresponds to the key name.

IdeaBlade DevForce

Business Object Persistence - Advanced

To be more precise, it returns an object that implements IDataSourceKey. There are three implementations of this interface at the moment, the EdmKey, the WsKey and the ClientEdmKey. EdmKey and WsKey provide access and management information for relational database and web service data sources, respectively. ClientEdmKey has no dependency on the Entity Framework on its data sources. A DataSourceKeyResolver has a single method, GetKey(KeyName, KeyExtension) to do get this key. The KeyName is the symbolic data source name that we see inscribed in the entity‟s DataSourceKeyName property. The KeyExtension, as we have seen, is an optional string for differentiating among multiple keys each referring to a distinct runtime data source. DevForce uses its own DefaultDataSourceKeyResolver unless we provide an alternative. The default version looks for a key in the IdeaBlade section of the application configuration file, App.config; it knows how to find the requested key definition in the configuration file‟s XML and turn it into the appropriate kind of DataSourceKey. The App.config is a fixed file that must reside in a known place. That means the key information must be comparatively static as well. True, the configuration file does not have to be compiled into the application74. DevForce will prefer a loose version of the file in the executable‟s directory. We make our change, drop it into the executable‟s directory, and DevForce will prefer that version over any other. No re-compilation or major re-deployment required. We may need more flexibility or security than the configuration file affords in situations such as the following:     The connection facts change periodically and we can‟t count on redeploying the updated configuration file. The connection facts are different for different users of the application. The connection facts must not reside in a text file; they must be delivered to the application at runtime after authenticating the user. You are having trouble deploying a loose configuration file on IIS.

Custom DataSourceKeyResolver
Fortunately, it is easy to write a custom DataSourceKeyResolver that does exactly what you want it to do. Pick a project to hold your key resolver, e.g. DomainModel If in an assembly not already being probed by DevForce, add a top-level probe assembly tag to App.config so DevForce can find it. XML <edmKey connection="metadata=res://ServerModelNorthwindIB/ServerModelNorthwindI B.csdl|res://ServerModelNorthwindIB/ServerModelNorthwindIB.ssdl|res://S erverModelNorthwindIB/ServerModelNorthwindIB.msl;provider=System.Data.S qlClient;provider connection string=&quot;Data Source=.;Initial Catalog=NorthwindIB;Integrated Security=True;MultipleActiveResultSets=True&quot;" containerName="ServerModelNorthwindIB.ServerModelNorthwindIBContext" logTraceString="false" name="Default" tag=""> <probeAssemblyNames> <probeAssemblyName name="DomainModel" /> <probeAssemblyName name="ServerModelNorthwindIB" /> </probeAssemblyNames> </edmKey>
74

It is compiled into the application by default as an embedded resource of the AppHelper.dll.

IdeaBlade DevForce

Business Object Persistence - Advanced

Add the following references to that project:
IdeaBlade.EntityModel IdeaBlade.Core IdeaBlade.EntityModel.WS // if creating WsKeys

Write a class that implements IDataSourceKeyResolver. Decorate the class with the SerializableAttribute ([Serializable] in C#, in VB).
<Serializable()>_

Implement your version of GetKey(KeyName, KeyExtension) to handle the keys you want to manage. Return null (Nothing in VB) if you want the DefaultDataSourceKeyResolver to determine the key.

C#

using IdeaBlade.EntityModel; using IdeaBlade.Core; namespace AppHelper { [Serializable] class MyDataSourceKeyResolver : IDataSourceKeyResolver { public IDataSourceKey GetKey(string keyName, string keyExtension, bool onServer) {

if (!onServer) {return null;}
Console.WriteLine("Shot ya with my resolver"); // Demo code. // Build your own ClientEdmKey starting with the following // return new MakeClientEdmKey(keyName, theConnectionString) return null; // Didn't build key; DefaultDataSourceKeyResolver takes over } } }

VB
Imports IdeaBlade.EntityModel Imports IdeaBlade.Core <Serializable()> _ Public Class MyDataSourceKeyResolver : Implements IDataSourceKeyResolver Public Function GetKey(ByVal keyName As String, _ ByVal keyExtension As String, _ ByVal onServer As Boolean ) _ As IDataSourceKey Implements IDataSourceKeyResolver.GetKey

If !onServer Then Return null
Console.WriteLine("Shot ya with my resolver") ' Demo code. ' Build your own ClientEdmKey starting with the following ' Return New MakeClientEdmKey(keyName, theConnectionString) Return Nothing ' Didn't build key; DefaultDataSourceKeyResolver takes over End Function End Class

THE NET RESULT OF KEY LOOKUP MUST DELIVER A KEY ON BOTH CLIENT AND SERVER. However, in n-tier, the client should not provide connection info in the key. Note that GetKey receives a boolean onServer parameter that indicates whether GetKey() is operating on the server or client.

IdeaBlade DevForce

Business Object Persistence - Advanced

Multiple Application Environments
Many IT shops prescribe separate Development, QA, Test, Stage, and Production environments. Each version of the application works its way through a testing gauntlet from the developer environment to ultimate production release. Suppose our application refers to a database data source called “default”. Its data source key is “default”75. The application will use this key at runtime to find a data source configuration in the application configuration file (IdeaBlade.ibconfig). The data source configuration is very simple for the development environment. The development deployment puts all tiers on the PC. The “default” development configuration‟s connection string points to a database on the PC. The QA environment, on the other hand, has a 3 tier deployment with separate machines for client, business object server, and database. This requires many changes to the “default” configuration including a different connection string that points to the QA database. We really need a separate “default” configuration for QA. In fact, we need five “default” configurations in the application configuration file. The symbolic data source, “default”, doesn‟t change as we cross environments. The business objects associated with the “default” data source should be indifferent to configuration differences. The executing environment, on the other hand, has to know which of the “default” configuration to use. DevForce provides data source key extensions to help distinguish the five “default” data source configurations. By convention, the data source configuration name is the data source key name followed optionally by an underscore “_” and data source key extension. In our example, the configurations could be named “Default_Development”, “Default_QA”, etc. When the application launches, it determines its runtime environment and then tells the EntityManager to connect to its data source(s) using the extension to find the appropriate data source configuration information 76. If we execute in development, we initialize the PM with “Development” and it adds the “_Development” suffix to “default”. If the EntityManager (and, later, the EntityServer) cannot find a data source configuration named “Default_Development”, it will look for one named “default” before giving up.

Multi-Level Undo with Checkpoints
Many applications could benefit from a robust, cross-entity undo feature. Simple dialogs, for example, may present opportunities to modify several entities perhaps of different types. If the user cancels, we have to reverse all those changes and restore the world to its pre-dialog state. There is a lot of booking to do if we want to handle this manually. Imagine a more complex case, a “New Account Wizard” in which the user steps through a series of screens, adding a customer account, an address, some contacts, etc. In each step the user creates or modifies at least one business object but often many more and of different entity types. The user may need to back up a step or two, discarding changes page by page. DevForce WinClient applications can set a “checkpoint” at each step and “roll-back” to an earlier step if the user clicks “Cancel” or “Back.” When the Wizard opens, we call “BeginCheckpoint” just before presenting the first page. DevForce WinClient starts recording the user‟s changes as they affect entities in or entering the EntityManager cache. Such changes could include:

75 76

“default”, not coincidentally, is the DevForce Object Mapper‟s default data source key name for the first data source. Entities in the PM may map to more than one data source. The PM will suffix each data source key name with the same extension.

IdeaBlade DevForce
        Fetched entity Created entity Modified entity Entity undo Deleted entity Removed entity Re-attached entity Entity merged into the PM from another PM or EntitySet

Business Object Persistence - Advanced

If the user cancels a Wizard step, we call RollBackCheckpoint, and the EntityManager restores its entity cache to the state when we began the checkpoint. We can maintain a stack of checkpoints. We call BeginCheckpoint for step one; the user makes some changes and proceeds to step 2 where we call BeginCheckpoint again. The EntityManager records a boundary in the checkpoint log and increases the checkpoint “depth” by one. Now we can rollback either to the first or the second checkpoint, depending upon how much activity we want to discard. On the other hand, if the user presses “Ok” and completes the Wizard successfully, we can save all modified entities and prevent rollbacks prior to the save point by calling EntityManager.SaveChanges(). We don‟t have to save the changes to close a checkpoint “session.” If we call ClearCheckPoints(), the EntityManager discards the checkpoint log and stops logging entity cache activity. The user‟s changes are still pending in cache – they are not in the database - but we have erased our checkpoints and can no longer rollback. We might think of a DevForce WinClient checkpoint as an in-memory transaction along the lines of the more familiar database transaction. The BeginCheckpoint, RollbackCheckpoint, and SaveChanges correspond approximately to “Begin Transaction”, “Rollback Transaction”, and “Commit Transaction.” We can nest checkpoints just as we nest database transactions. We can rollback to any pending checkpoint as we can rollback to any pending transaction depth. Why is there a ClearCheckPoints() method but no Commit() method? There is no “Commit” because we feared a potentially fatal confusion. “Commit,” for most of us, implies a degree of permanence that an in-memory transaction cannot match. We expect a durable modification of the database after a database “commit”. In contrast, entity changes are still pending and tenuous after ClearCheckPoints(); they will be lost if the application terminates before we persist them explicitly with SaveChanges(). We might regard a “checkpoint” as a kind of “snapshot”. A checkpoint differs from the everyday meaning of “snapshot” in one key respect: a “checkpoint” records changes to the entity cache; a snapshot would record the entire cache. Checkpoints are comparatively lightweight. The snapshot of a large entity cache could hold thousands or millions of entities while the equivalent checkpoint held only a few. A snapshot might be too large to hold in memory; a checkpoint is compact and easily held in memory. Checkpoints are efficient, especially for the SaveChanges and ClearCheckPoints operations; the EntityManager just throws away the log. Rollback is a bit more expensive because the EntityManager must reverse the logged changes; in most cases rollbacks are rare and the changes are few. The new EntityManager checkpoint signatures are:

IdeaBlade DevForce
Method
BeginCheckpoint() RollbackCheckpoint() RollbackCheckpoint(int pCount) RollbackCheckpoints() ClearCheckpoints()

Business Object Persistence - Advanced
Description Start “checkpointing” (first call) or add a new checkpoint level (subsequent calls). Returns the new checkpoint depth. Rollback one checkpoint. Restores the entity cache to its state one checkpoint ago. The method returns the new checkpoint depth. Rollback “pCount” number of checkpoints ago; returns the new depth. Rollback all checkpoints and stops checkpointing. Restores the entity cache to the state prior to the first checkpoint. Stop checkpointing and discard the checkpoint log. The entity cache remains in its current state. Roughly equivalent to a “commit”.77 True if the EntityManager is maintaining checkpoints. Integer of the current checkpoint depth. The first BeginCheckpoint is depth one; each subsequent call increases the depth by one and each RollbackCheckpoint() reduces it by one. The depth is zero when checkpointing stops.

IsCheckpointing GetCheckpointDepth()

A few points of additional interest:    The scope of a checkpoint session is a single EntityManager.
EntityManager.Clear(), like SaveChanges, clears the checkpoints and stops checkpointing.

The checkpointing records changes to entity persistable state contained in the fields mapped to data source columns. Checkpointing does not capture or restore data in custom fields that you may have added to your business object‟s custom class. At this writing, the checkpoints are not included in the EntityManager‟s EntitySet nor are they accessible directly as data. Therefore, we cannot preserve checkpoints when the application terminates and restore them when we re-launch.

Multiple EntityManager Instances
Most applications only need a single EntityManager instance. A EntityManager instance can hold every entity we need in a single cache – even entities that persist to different data sources. Accordingly, when we write “EntityManager” we mean an instance; we say “EntityManager class” when referring to the class rather than the instance. We can create new instances and there are scenarios for which this is useful. Perhaps we have a long-running query or series of queries that should run in a background thread without blocking the UI. Maybe we want to poll for changes to a set of entities or be on the look-out for certain conditions in the database. Our implementation should use a different EntityManager in the background thread so as not to conflict with the main manager in the UI thread. When the background process completes, the call-back method can pause the UI, import data from the second manager, alert the user, and resume the UI.

77

Note that clearing a single checkpoint, or any number less than all of them, is not supported. Changes subsequent to a given checkpoint may depend upon changes made after earlier checkpoints. It is therefore not possible to support the “commit” of changes made since a more recent checkpoint while still permitting rollbacks to earlier checkpoints. If you clear checkpoints, you must clear them all.

IdeaBlade DevForce Life with two PMs

Business Object Persistence - Advanced

Each EntityManager has its own entity cache. An entity instance in one cache is not the same as an entity instance in another cache even when the two instances have identical primary keys. Entities with duplicate keys cannot exist within a single cache. There can be only one Employee object with Id = 42 in a given cache. However, after reading Employee #42 into PM „A‟ and PM „B‟, there are two Employee objects in the application that have Id = 42. This is fine as long as we are aware of it. Think of the two EntityManager caches as separate clients. When „A‟ changes Employee #42, this has no immediate effect on „B‟s copy of Employee #42. If „A‟ saves the changes, „B‟s copy is no longer current with respect to the data source. If „B‟ then makes changes and tries to save, „B‟ gets a concurrency violation. These rules apply whether „A‟ and „B‟ are two end users on different PCs or two EntityManagers in the same application.

Miscellaneous observations
Different EntityManagers do not interact. It is possible – and useful – to import entities from one PM to the other.

Logging In a Second EntityManager Based on the Credentials of Another EntityManager
The ability to create a second EntityManager that is logged with the same credentials as the first facilitates scenarios in which the application creates a second context for editing. Changes in this second context are isolated from the main context and can be saved or canceled without unintended effects on entities in the primary PM. The EntityManager has a copy constructor for this purpose. C#
EntityManager Pm2 = new EntityManager(Pm1);

VB

The new EntityManager (Pm2) will have the same settings and credentials as its prototype (Pm1) but without any data. Its login state will be the same as its prototype. The second PM must connect to the database in order to save changed entities. It can only connect if it is logged in. Without the ability to login the second PM at its creation, we would have to preserve the users original credentials in some “safe” place in memory. This would be both inconvenient and discomforting, as one can never be quite certain that a rogue module can be prevented from acquiring those credentials and misusing them. It is best to forget about them as soon as possible. The new constructor permits you to do so.

Multi-Threading in a DevForce App
Let‟s begin our discussion of multi-threading with a definition of thread-safety:
For a class to be thread-safe, it first must behave correctly in a single-threaded environment. If a class is correctly implemented, which is another way of saying that it conforms to its specification, no sequence of operations (reads or writes of public fields and calls to public methods) on objects of that class should be able to put the object into an invalid state, observe the object to be in an invalid state, or violate any of the class's invariants, preconditions, or postconditions. Furthermore, for a class to be thread-safe, it must continue to behave correctly, in the sense described above, when accessed from multiple threads, regardless of the scheduling or interleaving of the execution of those threads by the

IdeaBlade DevForce

Business Object Persistence - Advanced

runtime environment, without any additional synchronization on the part of the calling code. The effect is that operations on a thread-safe object will appear to all threads to occur in a fixed, globally consistent order. The relationship between correctness and thread safety is very similar to the relationship between consistency and isolation used when describing ACID (atomicity, consistency, isolation, and durability) transactions: from the perspective of a given thread, it appears that operations on the object performed by different threads execute 78 sequentially (albeit in a nondeterministic order) rather than in parallel.

The DevForce EntityManager is safe for multithreaded read operations. If you attempt writes to a single EntityManager from multiple threads, you must synchronize the write operations yourself. For us to make the EntityManager thread-safe for write operations would require that we make thread-safe every method therein – and every method of the business objects it manages, including property setters. This would increase the EntityManager‟s complexity – and degrade its performance – significantly. Every user of the EntityManager, and every use thereof, would incur the performance penalty, whether such users and uses required thread-safety or not. At least 90% of the use cases that people submit to us for multi-threading involve retrieving data while other operations proceed. For this we have provided Asynchronous Queries. You call the EntityManager‟s GetEntitiesAsync() method, and we take care of putting the data retrieval operation on a separate thread so that the rest of your application can continue processing. Any number of such asynchronous queries can be launched simultaneously. You can read about asynchronous queries in the DevForce Developers Guide (in the chapter on Object Persistence), and see sample code in the Asynchronous Queries instructional unit that is shipped with the product. Does this mean that you can‟t do multi-threading (other than by using Asynchronous Queries) in a DevForce application? No, it does not. It just means that you should never share a single EntityManager, or any of the entities it manages in its cache, across multiple threads. Let us repeat:   Never share a EntityManager across more than one thread. Never share entities from a given EntityManager in more than one thread.

Note that the problems that occur with multi-threaded applications are, by their very nature, timing-dependent and difficult to diagnose, reproduce, and test for. Your multi-threaded process can work successfully for long periods of time, then fail catastrophically when two or more inconsistent changes happen to be made simultaneously. You should definitely not count on this failure occurring at a convenient time! If You‟re Determined To Do Multi-Threading… Be sure you really need multiple threads. Remember, if all you want to do is fetch data asynchronously, you will be fully satisfied with Asynchronous Queries. Don‟t mess around with multi-threading if this is all you want to do. Use caution when writing any multi-threaded app. Don't be lulled into a false sense of confidence just because it is easy to spawn a BackgroundWorker in .NET 2.0. Multi-threading is still hard. The BackgroundWorker made the syntax easy: it did not make good multi-threaded design easier! If you‟re new to multi-threaded programming, work with someone who has significant prior experience doing it, if at all possible. If you can‟t arrange that, do some serious reading and study on the topic before attempting it on a critical application.    If your multi-threaded aspirations involve DevForce business objects: Use a different EntityManager in each thread. Such EntityManagers can do anything a normal EntityManager can do; they can fetch (both synchronously and asynchronously), save, and so forth. Never use EntityManager.DefaultManager when multi-threading – the DefaultManager is “global” to the AppDomain and will be shared among any threads in which it is used.

78

Excerpted from ”Characterizing Thread Safety” by Brian Goetz, available on the web at:
http://www-128.ibm.com/developerworks/java/library/j-jtp09263.html

IdeaBlade DevForce

Business Object Persistence - Advanced

Never communicate entities across thread boundaries. If the caller must know about some entities, send a list of PrimaryKeys across the thread boundary in a call back. Alternatively, you could bury EntitySets in a call back to serialize copies of entities across the thread boundary.

Batching Asynchronous Tasks
DevForce includes two classes, the AsyncSerialTask and the AsyncParallelTask, that permit you to define and execute asynchronously, in series or in parallel, a collection of linked actions. Each method uses a single callback to handle all processing results, and each provides the ability to specify an ExceptionHandler to provide a single point of error handling.

AsyncSerialTask
The AsyncSerialTask provides you with a mechanism to define a sequence of linked actions, each of which can be performed synchronously or asynchronously. To use the feature, you first create the root task in the sequence, and then add actions to it, until you have a sequence ready for launch via the Execute method. The AsyncSerialTask allows you easily to link a series of actions, passing the output from the previous action as input to the next action. Without the AsyncSerialTask, you would need to issue each asynchronous action separately and in the handler for the completed action launch the next action in the sequence. The AsyncSerialTask takes care of this housekeeping for you. It allows you to pass an argument into the task sequence when execution begins, and to specify a single handler when the entire sequence completes. You can specify an ExceptionHandler to provide a single point of error handling. Note that the entire sequence is not executed as a group on a worker thread. Instead, as each action is serially executed, if the action is asynchronous then a worker thread is started for it; when the action completes its results are returned back to the main thread, which then continues with the next action in the sequence. Note that if you add only synchronous actions and functions to the AsyncSerialTask the entire sequence will execute synchronously. Use the AsyncParallelTask rather than the AsyncSerialTask if you can execute all actions simultaneously.

C#

public void SampleAsyncTask() { DomainModelEntityManager mgr = new DomainModelEntityManager(); // Let's take a series of "actions" all performed synchronously. // Login - if ok, then: // - Run a query for customers // - Modify the retrieved data // - Save changes // It might look like this: if (mgr.Login(new LoginCredential("demo", "demo", "earth"))) { var customers = mgr.Customers.Where(c => c.Country == "USA").ToList(); customers.ForEach(c => c.Country = "US"); SaveResult sr = mgr.SaveChanges(); Debug.Assert(sr.Ok); } // // // // // // Now assume that some of these actions should be done asynchronously. In Silverlight, any actions which go to the BOS - such as query and save must be performed asynchronously. The AsyncSerialTask let's you group a series of actions to be performed together. Without this, you would need to issue each async call separately, and in the handler for the completed action fire off the next action. The AsyncSerialTask does

IdeaBlade DevForce
// this for you. // // // // // // // //

Business Object Persistence - Advanced

The same actions with the AsyncSerialTask: - Login asynchronously - When login completes, run an async query to retrieve customers - When the query completes, modify the retrieved data - Save these changes asynchronously (You should also use an ExceptionHandler to trap errors, but we've removed that to make this sample a bit easier to read.) Here's how you might build this task:

AsyncSerialTask.Create("ASimpleTask") .AddAsyncLogin(mgr, new LoginCredential("demo", "demo", "earth")) .AddAsyncQuery(loginArgs => mgr.Customers.Where(c => c.Country == "USA")) .AddAction(fetchArgs => { var customers = fetchArgs.Result; customers.ForEach(c => c.Country = "US"); }) .AddAsyncSave(mgr) .Execute(null, (completionArgs) => { SaveResult sr = completionArgs.Result.Result; Debug.Assert(sr.Ok); }); }

AsyncParallelTask
The AsyncParallelTask allows you to create a set of asynchronous actions, execute them in parallel, and provide a single callback to handle all processing results. To use the feature, you first create a task, and then add asynchronous actions to it until you have a set ready for launch via the Execute method. In the absence of the AsyncParallelTask you would need to issue multiple asynchronous method calls and provide handlers for each. Instead, the AsyncParallelTask takes care of much of this housekeeping for you. It allows you to pass an argument to each action in the task, and to specify a single handler when the entire task completes. You can also specify an ExceptionHandler to provide a single point of error handling. Each action is executed on a separate worker thread. The completion action is called on the main thread once all actions have completed. If you've specified a callback for an asychronous action, that callback will also be called on the main thread. The Execute call returns immediately after starting all of the specified parallel actions. Use the AsyncSerialTask rather than the AsyncParallelTask if you need to link the outputs from one action to the inputs to the next, or to mix asynchronous and synchronous actions.

IdeaBlade DevForce
C#
public void SampleAsyncTask() {

Business Object Persistence - Advanced

DomainModelEntityManager mgr = new DomainModelEntityManager(); // Let's take a few "actions" performed asynchronously: // - Run a query for all customers // - Run a query for all employees mgr.ExecuteQueryAsync<Customer>(mgr.Customers, cb => { if (cb.Error != null) { Debug.WriteLine(cb.Error.Message); } else { cb.Result.ForEach(c => Debug.WriteLine(c.CompanyName)); } }, null); mgr.ExecuteQueryAsync<Employee>(mgr.Employees, cb => { if (cb.Error != null) { Debug.WriteLine(cb.Error.Message); } else { cb.Result.ForEach(e => Debug.WriteLine(e.LastName)); } }, null);

// Since these async actions both essentially run in parallel, let's // combine them into a single task: AsyncParallelTask.Create() .AddExceptionHandler(args => Debug.WriteLine(args.Exception.Message)) .AddAsyncQuery(1, x => mgr.Customers) .AddAsyncQuery(2, x => mgr.Employees) .Execute(cb => { ((EntityFetchedEventArgs <Customer>)cb.CompletionMap[1]) .Result.ForEach(c => Debug.WriteLine(c.CompanyName)); ((EntityFetchedEventArgs <Employee>)cb.CompletionMap[2]) .Result.ForEach(e => Debug.WriteLine(e.LastName)); }); }

Service Oriented Architecture
We are sometimes asked whether DevForce is a Service Oriented Architecture (SOA). DevForce applications can be SOA in three respects. First, DevForce applications are .NET applications, which means it is very easy to build web services into the application. Second, you can map a DevForce business object to a Web service, which means that web service entities are first class entities like table, view, and stored procedure entities. Third, we can expose all or part of the business object model as a web service, which means external applications and non-.NET clients can take advantage of the hard work we put into our model 79. On the other hand, SOA is easy to abuse. It does not belong everywhere and it is an especially unfortunate choice for cross-tier data transfers in an n-tier application. The difference between n-tier and Service Oriented Architectures are vitally important and worth at least some discussion such as we‟re about to have now.

79

Web service entities and Web service wrappers for the business object model are in the alpha bits at this writing. Please contact IdeaBlade for more recent information about these important features.

IdeaBlade DevForce

Business Object Persistence - Advanced

SOA design emphasizes loose coupling and a contract between a client and a service. The client should know as little as possible about the service internals and engage with the service only through a message-like interface having a simple protocol. The interface should be course grained, meaning that we expect to get a lot done with each service method call and we don‟t over-task the interface with fine details. We are often aware of the boundary between the client and service. The services of an SOA application do not belong to that application. In principle, the services are designed independently of any particular SOA application and could be accessed by any authorized client. An n-tier application, by contrast, has logical layers that are tightly coupled. The layers tend to have many, fine grained interface points. The layers are designed to work together as a single, operating whole. It is a secondary benefit if a tier can serve another application through the same interface. SOA proponents emphasize the importance of the contract between client and server. But SOA can only ensure the consistency of the interface points. It can‟t ensure that the semantics implied in the interface are actually the same on both sides of the fence. A program manager on Microsoft‟s CLR team made an analogous point in a commentary on the difficulty of choosing between defining classes and interfaces:
I often hear people saying that interfaces specify contracts. I believe this is a dangerous myth. Interfaces, by themselves, do not specify much beyond the syntax required to use an object. The interface-as-contract myth causes people to do the wrong thing when trying to separate contracts from implementation, which is a great engineering practice. Interfaces separate syntax from implementation, which is not that useful, and the myth provides a false sense of doing the right engineering. In reality, the contract is semantics, and these can actually be nicely expressed with some implementation.[emphases ours] Krzysztof Cwalina, [Framework Design, 80]

N-tier applications can impose much stricter contracts than SOA applications. They can enforce common semantics by requiring both sides to implement the contract by using the same object classes. A DevForce application forces the type on the server tier to be exactly the same as the type on the client tier. This is known as “type fidelity”. Hiding implementation details is as essential to n-tier design as it is to SOA design; each layer should know as little as possible about the design and works of the other layers. But an n-tier application can and should impose cross-tier requirements if these help realize application objectives. For example, we can require that the data access tier communicate with a UI tier via .NET remoting rather than Web services if this makes the application less complex and perform better; such objectives may matter far more to the customer right now than exposing the business object model as a service80. An n-tier application can also be an application with a Service Oriented Architecture. The application may implement some number of features by invoking a Web service or by embedding a Web service within an object wrapper. The tier may expose some of the application‟s own functionality as Web services. In such cases, the application is communicating externally via the service. Cross-tier interactions, on the other hand, are communications within the application. Let not blind obedience to SO orthodoxy triumph over rational choice.

POCO Support in DevForce
In addition to objects based on classes inheriting from IdeaBlade.EntityModel.Entity and generated by the DevForce Object Mapper, you can now use Plain Old CLR Objects (POCOs) with DevForce. The class for your object must be deployed on both the client and the server and must be contained in one of the assemblies routinely searched by

80

This decision does not prevent us from exposing the business object model as a Web service later. Even then we could retain our remoting interface within the application.

IdeaBlade DevForce

Business Object Persistence - Advanced

DevForce (e.g., the one with the DomainModel) or in an assembly explicitly mentioned in the Probe Assemblies section of the app.config (and/or web.config) file at the top level (rather than in an EdmKey). An additional class (or classes), a “POCO service provider”, must be supplied server-side, marked with special attributes. This class will contain methods to perform the server-side data retrieval for your custom objects. The names for the data retrieval methods may either conform to flexible naming conventions (to be discussed below), or they can be named anything you desire and marked with an attribute that specifies that their exact name must be used in the client-side data retrieval statement. For convenience, you will probably also want to deploy a client-side class containing extension methods for the EntityManager, so that you can refer to your POCO objects in LINQ queries in a manner similar to that which you use with your DevForce Entities. If the objects have an identified (primary) key value, they will operate fully as first-class citizens in the DevForce local cache. That means, among other things, that they can be:  Stored in the cache;  Queried against and retrieved from the cache;  Updated in the cache;  Created in the cache; and of course  Saved from the cache. Objects that do not have an identified key can still be retrieved, but will not be stored in the DevForce cache. DevForce‟s support for POCO objects follows Microsoft RIA (Rich Internet Application) standards, and uses RIA attributes and naming conventions, to the extent that those are available. As those standards and facilities evolve or are fleshed out, the implementation in DevForce will be enhanced or migrated to maintain maximum compatibility with the RIA standards.

Examples of POCO Classes
Here is a class representing a State of the United States. At runtime, you might, for example, creating instances of this class using data from an XML data file deployed on the server:

IdeaBlade DevForce
C#
using using using using System; System.Collections.Generic; System.Linq; System.ComponentModel.DataAnnotations;

Business Object Persistence - Advanced

namespace AppWithPocos.Pocos { public class State { public State() { } //[Key] public string Abbrev { get { return _abbrev; } set { _abbrev = value; } } public string Name { get { return _name; } set { _name = value; } } public bool Lower49 { get { return _lower49; } set { _lower49 = value; } } public long Population { get { return _population; } set { __population = value; } }

#region Private Fields string _abbrev; string _name; bool _lower49; long _population; #endregion Private Fields } }

IdeaBlade DevForce

Business Object Persistence - Advanced

Examples of a POCO Service Provider Class
This POCO Service Provider class can be named anything – and you may have many such classes – but must be flagged with the attribute [EnableClientAccess]. All such classes will be deployed and used server-side only. C#
using using using using using using using using using using using using using using System; System.Collections.Generic; System.Linq; System.Web; System.ComponentModel; System.Text; IdeaBlade.EntityModel; IdeaBlade.Core; System.Reflection; IdeaBlade.Core.DomainServices; AppWithPocos.Pocos; System.Xml; System.ComponentModel.DataAnnotations; System.Runtime.Serialization;

namespace AppWithPocos { [DataContract] [KnownType (typeof(List<State>))] [EnableClientAccess] public class PocoServiceProvider { public PocoServiceProvider() { } #region State //[RequiresAuthentication] // Uncomment this attribute will require "real" authentication public IEnumerable<State> GetStates() { IEnumerable<State> states = ReadStatesData("states.xml"); return states; } private static IEnumerable<State> ReadStatesData(string fileName) { // Create an isntance of XmlTextReader and call Read method to read the file XmlTextReader textReader = new XmlTextReader(fileName); textReader.Read(); List<State> states = new List<State>(); while (textReader.Read()) { State aState = new State(); textReader.MoveToElement(); if (textReader.Name == "State") { aState.Abbrev = textReader.GetAttribute("Abbrev").Trim(); aState.Name = textReader.GetAttribute("Name").Trim(); aState.Lower49 = Convert.ToBoolean(Convert.ToInt32(textReader.GetAttribute("Lower49"))); states.Add(aState); } } return (IEnumerable<State>)states; }

IdeaBlade DevForce
#endregion State } }

Business Object Persistence - Advanced

Note the “GetStates()” method. This method retrieves the data requested in a query (which is usually submitted from the client). It follows a naming convention similar to the one supported under RIA services, where the name consists of a prefix that is one of a number of synonyms for “retrieve” (here, “Get”) and a suffix that is the pluralized name of a POCO type (here, “States”). The prefixes may be any of the following: Get Fetch Find Query Retrieve Select

When the above naming convention is used, a query can be constructed client-side with an expression such as the following: C#
new EntityQuery<State>("States", anEntityManager);

Alternatively, the method can be adorned with the QueryAttribute. C#
[Query] public IEnumerable<State> ReturnAllStates() { IEnumerable<State> states = ReadStatesData("states.xml"); return states; }

In that case, when called client-side the exact name used for the query must be supplied: C#
new EntityQuery<State>("ReturnAllStates", anEntityManager);

Query Methods with Parameters
Methods with parameters are supported as well. For example, an additional overload to the GetStates() method above that only returns states with a population size greater than a size passed in might be written as shown below. (Naturally, this would require that the internal ReadStatesData() method be modified as well.)

C#

public IEnumerable<State> GetStates(long minPopulationSize) { IEnumerable<State> states = ReadStatesData("states.xml", minPopulationSize); return states; }

IdeaBlade DevForce

Business Object Persistence - Advanced

On the client side, parameters may be specified by using one of the EntityQuery.AddParameter() overloads. The following snippet calls the parameterized GetStates() method to return just those states with a population greater than one million people: C#
var query = new EntityQuery<State>("States", anEntityManager); query.AddParameter(1000000); // see

Any number of query parameters are permitted, and the standard .NET overload resolution rules apply. This means that the order and type of the parameters are checked to find appropriate matches, with type coercion occurring as required.

When to Use Parameterized Query Methods Rather Than LINQ Expressions
The following two queries will return the same results: C#
var query1 = new EntityQuery<State>("States", anEntityManager); query.Where(state => state.Population > 1000000);

var query2 = new EntityQuery<State>("States", anEntityManager); query.AddParameter(1000000);

Furthermore, in both cases, the restriction to those states with greater than one million population occurs on the server, not the client. So the question arises: is one to be preferred over the other? The answer usually depends upon how the server-side method itself is implemented. In general, unless the server-side method can internally use the query parameter to restrict its own query against some back-end data store, query parameters have no advantage over LINQ query restrictions. In fact, LINQ queries are far more flexible and intuitive to work with under most circumstances. Nevertheless, there will be cases where a back-end data store‟s ability to optimize some queries will yield sufficient performance improvement to justify the use of query parameters. For example, consider the Windows file system‟s ability to search for files, given a path and wildcards. While the same result could be accomplished via a server-side method that returned all of the files in the file system and then iterated over them to locate a desired set of files, it would likely be faster to call the file system directly with the path and wildcard restrictions.

Example of a Client-Side Class Containing Extension Methods for the EntityManager
The following class, deployed client-side, can provide further help in putting your POCO objects on an equal footing with your DevForce-generated objects. First, look at the class; then below, we‟ll show what its presence permits in terms of data retrieval syntax: C#
using System; using System.Collections.Generic; using System.Linq; //using System.Web; using System.Diagnostics; using System.Text; using System.Linq.Expressions;

IdeaBlade DevForce
using IdeaBlade.EntityModel; using IdeaBlade.EntityModel.Extensions; using IdeaBlade.Core; //using AppWithPocos.Pocos; using DomainModel; using AppWithPocos.Pocos; namespace AppWithPocos { public static class EmExtensions {

Business Object Persistence - Advanced

#region State public static EntityQuery<State> States(this EntityManager em) { return new EntityQuery<State>("States", em); } #endregion State } }

Because there are now static property named “States” and “Foos” available that return EntityQuery<State> and EntityQuery<Foo>, respectively, you can now order the retrieval of States and Foos with the following syntax: C#
_mgr.ExecuteQueryAsync<State>(_mgr.States() .Where(s => s.Lower49, GotStates, null);

This is very similar to what you do with DevForce entities, except that, since States() is an extension method, so you will have to include the parentheses -- _mgr.States() -- which you do not have to do when referencing the generated properties of the EntityManager that return EntityQuery<T>. Note that you can extend the queries with clauses with the full set of LINQ extension methods -- such as Where(), used above -- just as you can do with ordinary queries. 81

Obtaining an EntityAspect Property on Your POCO Object
Objects from DevForce-generated Entity classes have an EntityAspect property through which a number of important operations and useful pieces of metadata can be obtained. Custom objects can benefit from the same facilities by one of two methods:  Implementing a very simple interface, IHasEntityAspect, or  Wrapping them at runtime The latter method preserves their DevForce-ignorant POCO status, but will not provide equal performance with the former method. Getter and Setter can be placeholders in the IHasEntityAspect implementation: C# amespace AppWithPocos.Pocos {

81

As of DevForce 5.1.1, the Include() syntax on a POCO entity query is not yet implemented. The call will compile but will not yet do anything. This will be corrected in a later release.

IdeaBlade DevForce
public class State:IHasEntityAspect {

Business Object Persistence - Advanced

…[snip]

#region IHasEntityAspect Members

public EntityAspect EntityAspect { get; set; } #endregion

To wrap an Entity at runtime for the purpose of getting at the EntityAspect property, call the static method EntityWrapper.Wrap(): C#
var stateWithAspect = EntityWrapper.Wrap(aState);

Having done that, you can then ask questions like the following… C#
Bool isModified = stateWithAspect.EntityAspect.IsModified();

…and of course, use any of the other facilities of EntityAspect. (Those are documented in the Developers Guide chapter “Class Libraries”.)

Data Contract Serializer (DCS) versus .NET Data Contract Serializer (NDCS)
The .NET framework includes several important serialization classes (which are described, for reference, in the callout at the end of this section). For the purpose of your DevForce apps, two of these are of particular importance: the Data Contract Serializer (DCS) and the .NET Data Contract Serializer (NDCS). The NetDataContractSerializer (NDCS) is used by all types of .NET apps except Silverlight apps, for which only the DataContractSerializer (DCS) is available. Since the DCS requires more from you in the way of data annotations than the NDCS, and since some of you will be writing business object classes that you will want to use in both Silverlight and WFP or WinForm apps, we have provided an option in DevForce to force the use of the DCS. By setting this option you can be sure that you business object classes will work properly in both environments (Silverlight v. non-Silverlight). With DCS you may need to attribute your POCO classes with so they will be serializable. The need to be so arises when your POCO class does not have a property attributed as a key. Here‟s a POCO class that does have key:

IdeaBlade DevForce
C#
public class State { public State() { }

Business Object Persistence - Advanced

[Key]
public string Abbrev { get { return _abbrev; } set { _abbrev = value; } } public string Name { get { return _name; } set { _name = value; } } }

VB

When your POCO class is defined with a key, DevForce assumes that it will need to be serialized at some point and performs the necessary type inference. When your class is defined without a key, DevForce makes no such inference, and you will have to let it know explicitly that it needs to be serialized. You can do this in two ways. The preferred way for most usages is to attribute the class with the DevForce DiscoverableType attribute: C#

Using IdeaBlade.EntityModel; [DiscoverableType(DiscoverableTypeMode.KnownType)]
public class State { public State() { } public string Abbrev { get { return _abbrev; } set { _abbrev = value; } } public string Name { get { return _name; } set { _name = value; }

IdeaBlade DevForce
} }

Business Object Persistence - Advanced

VB

A second method is to implement the (empty) interface IKnownType: C#

Using IdeaBlade.EntityModel;
public class State { public State() } public string Abbrev { get { return _abbrev; } set { _abbrev = value; } } public string Name { get { return _name; } set { _name = value; } } }

: IKnownType

{

The first method is preferable, but the second method can be useful if you will have many classes that inherit from a base class, and you wish only to “mark” the base class.

IdeaBlade DevForce

Business Object Persistence - Advanced

.NET Serialization Classes (for reference)

The following is taken from Martin Konicek‟s discussion at http://coding-time.blogspot.com/2008/03/serialize-object-graph-to-xml-innet.html . XmlSerializer  Cannot serialize circular references.  If more objects point to the same object, its copies are created in the xml for each of these references.  Has to know all types that could be encountered during serialization in advance - throws an exception on unknown type. Known types are passed in the constructor to XmlSerializer or marked by XmlIncludeAttribute.  Generates simple readable xml.

DataContractSerializer  Has to know types in advance - like XmlSerializer.  Can serialize circular references - construct with preserveObjectReferences = true  Used by WCF.

NetDataContractSerializer - better!  Serializes object graph properly including circular references.  Doesn't have to know types in advance.  However, MSDN states that it can be only used in .Net <-> .Net communication, which is ok also for storing object in a file.  Generates simple readable xml.

BinarryFormatter  Works well, like NetDataContractSerializer.  Disadvantage is that it serializes to binary format, which make its unusable e.g. for saving to a file that user could later edit.  The output is smallest thanks to the binary format.

SoapFormatter  Deprecated. Cannot serialize generic collections

All serializers need the type to be serialized marked by SerializableAttribute.

Attributing a Type Where the Signature of Any DataMember-Attributed Member Indicates a Different Type Than Is Actually Returned
The KnownType attribute is used by the DCS to mark up a type with information about any of its members marked with a [DataMember] attribute where the declared type of the member is different from the actual type of the member. Suppose, for example, that your State class includes a Cities property which is typed as an object but actually returns a List<Cities>. This will confuse the serializer and make it serialize the property improperly. To enable the property to be serialized properly, you must attribute it as follows:

IdeaBlade DevForce
C#

Business Object Persistence - Advanced

[DataContract] [KnownType(typeof(List<City>)]
public class State { public State() {} [... snip] public object Cities { get { return _cities; } set { _cities = value; } } #region Private Fields List<City> _cities; #endregion Private Fields }

POCO Save mechanisms
DevForce provides two different mechanisms for saving POCO objects. These will be referred to as either adapterbased or convention-based implementations. By default, IdeaBlade will attempt to locate an adapterbasedimplementation; if unsuccessful, it will then look for a convention-based implementation. The implementation mechanism can also be specified explicitly by setting the PocoSaveMode property of the SaveOptions instance passed into the EntityManager.SaveChanges call:
var so = new SaveOptions(); so.PocoSaveMode = PocoSaveMode.UseEntitySaveAdapter; myEntityManager.SaveChanges(so);

C#

IdeaBlade DevForce

Business Object Persistence - Advanced

The PocoSaveMode is an enumeration with the following structure: C#
[/// <summary> /// Determines how to discover any custom server side POCO save methods. /// </summary> public enum PocoSaveMode { /// <summary> /// Use an EntitySaveAdapter subclass if found, otherwise use save methods discovered via convention. /// </summary> Default = 0, /// <summary> /// /// </summary> UseEntitySaveAdapter = 1, /// <summary> /// /// </summary> UseMethodsDiscoveredViaConvention = 2 }

Adapter-Based Saves
The adapter-based mechanism requires the existence of a server-side class that inherits from the IdeaBlade.EntityModel.Server.EntitySaveAdapter. If such a class is found, a single instance of this class will be created for each SaveChanges call, and the appropriate methods corresponding to each insert, update, or delete will be called for each entity to be saved. A single null-parameter constructor is required. Note that a single insert, update, or delete method handles saves for every entity type, so if save logic needs to be different for different types, the type of the entity passed in will need to be inspected and the execution branched appropriately. The following bare-bones version of an EntitySaveAdapter implementation shows the methods you have available to override:

IdeaBlade DevForce
C#

Business Object Persistence - Advanced

public class PocoSaveAdapter : EntitySaveAdapter { public PocoSaveAdapter() { } public override void BeforeSave( System.Collections.IEnumerable entities, SaveOptions saveOptions) { } public override void AfterSave() { } public override void InsertEntity(object entity) { } public override void UpdateEntity(object entity, object originalEntity) { } public override void DeleteEntity(object entity) { } }

Convention-Based Saves
The convention-based mechanism requires the existence of a server-side class with either the [DiscoverableType(DiscoverableTypeMode.ServerService)] attribute or alternatively (for RIA services compatibility), the [EnableClientAccess] attribute.82 This class will include insert, update, and delete methods named according to the conventions defined below. It must also include a single null-parameter constructor. If such a class is found, a single instance of it will be created for each SaveChanges call, and the appropriate methods corresponding to each insert, update, or delete will be called for each entity to be saved. For each type for which a persistence operation is provided, up to three methods must be written. The name of the method must begin with one of the prefixes defined below. The method must also conform to the signature defined below. If you do not expect to save objects that require one or more of these signatures, then they do not have to be implemented. In other words, if you never plan to delete “Foo” type objects, then you do not need to implement the Delete method. Insert Method   Prefixes: Insert, Add, or Create Signature: void InsertX(T entity), where T is an entity Type

Update Method   Prefixes: Update, Change, or Modify Signature: void UpdateX(T current, T original), where T is an entity Type

Delete Method  
82

Prefixes: Delete or Remove Signature: void DeleteX(T entity), where T is an entity Type

DevForce recognizes either attribute, as a convenience for those porting their app from RIA services

IdeaBlade DevForce

Business Object Persistence - Advanced

Below is an example implemention showing the use of the required conventional method names and signatures. InsertFoo could equally be named “AddFoo” or “CreateFoo”; UpdateFoo could be named “ChangeFoo” or “ModifyFoo”; and so forth. C# [DiscoverableType(DiscoverableTypeMode.ServerService)] // or [EnableClientAccess]
public class PocoSaveAdapterUsingConventions { public PocoSaveAdapterUsingConventions() { } public void InsertFoo(Foo entity) { // insert logic for any „Foo‟ entities } public void InsertBar(Bar entity) { // insert logic for any „Bar‟ entities } public void UpdateFoo(Foo current, Foo original) { // update logic for any “Foo‟ entities } public void UpdateBar(Bar current, Bar original) { // update logic for any “Bar‟ entities } public void DeleteFoo(Foo entity) { // update logic for any “Foo‟ entities } public void DeleteBar(Bar entity) { // update logic for any “Bar‟ entities } }

Summary – Things to Remember When Using POCOs in Your DevForce App
        Be sure to add the assembly that contains the POCOs as a top-level ProbeAssembly (i.e., not a in the ProbeAssemblies section of an EdmKey element) in the app.config or web.config. POCO classes must be included in both the server- and client-side assemblies. Put them in the server-side project and link them into the client-side project. Include a POCO ServiceProvider class server-side. Include EntityManager extensions client-side. You can use all forms of LINQ query against these objects. If your POCO class has a primary key (designated with the [Key] attribute, instances of it will be stored in the EntityManager cache, and can be updated, deleted, or inserted there. If your POCO class has no key, instances can be retrieved but will not be placed in the EntityManager cache. Implement IHasEntityAspect to get EntityAspect property and associated functionality on your “almost POCO” entity.

IdeaBlade DevForce

Validation Through Verification

Validation Through Verification

Validation Through Verification
DevForce Verification

Getting Started
Validation-Related Settings In the Object Mapper Generated Property Code Impact of Verifiers on the User Interface – A Caution

Now That You‟ve Been Initiated (and Before We Enter the Forest): A Quick Overview of the Mechanics Verification Types Overview
Main Verification Classes Verifiers VerifierResult Triggers VerifierEngine PropertyValueVerifiers

Verification Deep Dive
Verifiers Verifier Result Triggers VerifierEngine

Invoking Verification
Instance Verification Trigger Verification: Preset and Postset Monitor Execution with the VerifierBatchInterceptor

Verification and WinForms User Interfaces
UI Lockup Improving the User‟s Experience

“Validation” is the process of evaluating input and judging it valid or invalid. Such evaluation subjects the input to a battery of “validation rules” that evaluate the input in the appropriate context. For example, if the user enters a “committed delivery date” we might want to ensure that:    the committed delivery date is reasonable in the abstract, e.g., occurs in the future; it is possible to deliver on that date given the availability of the desired products, and the currently selected shipping method, and whether there is enough time to prepare the goods for shipping; the order is “shippable”, e.g. the customer‟s credit has been verified, the address is legitimate, and the total is within the limits authorized for this user.

Clearly such rules can be complex, involving not only the input value itself but also the state of the target object (the order), facts about related objects (customer, shipper, product), and aspects of the environment during the validation (time of day, the user‟s role).

330 | P a g e

IdeaBlade DevForce

Validation Through Verification

User input validation gets most of the attention but we need to validate programmatic inputs as well. That delivery date could as easily be set by business logic or a web service request as it is by a wayward click on a calendar control. The rules are the same for everyone, human and machine. Validation is hard to do well especially as the application grows and validation criteria change. Common failings include:      Missing and incorrect validity checks Inconsistent checking Failure to validate at the right times Poor communication with end-users Inadequate mechanisms for correcting mistakes.

Enterprise application developers are looking for a robust validation system that operates consistently and reliably across a large application. Robust validation cuts both “vertically” and “horizontally”:  We validate “vertically” when we validate several times in multiple layers of the application. We want to validate in the client UI layer so we can give immediate feedback to the user. We may need to validate again when we save, even though the objects we save are no longer on screen. We may even need to validate again on the server side to protect against misadventure coming from outside the relative safety of the hosted environment. We validate “horizontally” when we apply the same mechanisms uniformly across all modules of the application. If the user can set the delivery date on any of several screens, the same rules ought to apply – unless, of course, there is something special about a particular screen.

DevForce Verification
“Verification” is IdeaBlade‟s answer to the challenges of validation. “Verification” is a collection of interoperating validation components that are both easy to use and capable of handling sophisticated scenarios. The developer can:  Write rules of any complexity. The developer can draw upon pre-defined rules (required value, range check, field length) or write custom rules of any complexity, including rules that compare multiple fields and span multiple objects. Generate validity checking into business objects automatically via the DevForce “Object Mapper. Validate any kind of object, not just objects that derive from base business classes. Trigger validity checking at any time such as upon display, before save, or when setting properties. The engine can fire “pre-set” to block critically errant data from entering the object or fire “post-set” to accommodate temporarily invalid values. The UI can inspect the engine for rules of interest, fire them, and adjust the display accordingly. It could color a text box, for example, or hide a portion of the form until applicable criteria were met. Display a localized message in the UI without special programming. The UI could display just the “validation failed” message but it might also show warnings or “ok” messages and it might supplement the message be re-directing the application focus to the offending object and property. Each rule returns a rich, extensible object with all the information necessary for the developer to deliver a helpful response. Discover rules in the code or retrieve them at runtime from a central store. The engine automatically discovers rules in the code and can acquire rules defined externally in configuration XML, a database, or some other store of rules. The application can inspect, add, and remove rules at any time. Leverage rules inheritance. Rules defined in base classes propagate to their derived classes where they are “inherited” or overridden.

  

331 | P a g e

IdeaBlade DevForce

Validation Through Verification

Adjust validation behavior based on a custom validation context. The developer must have the flexibility to convey custom information to the validation process to cope with the variety of situational factors that arise in real applications. Inspect and intervene as the engine validates. The application can monitor the engine‟s progress and interrupt, modify, or terminate a validation run at any point.

“Verification” Versus “Validation”
The DevForce validation mechanism is called “Verification” and all of its components are named with some variation on this word. We mean to try neither your patience nor your vocabulary. We would call our offering “validation” if we could. However, Microsoft uses the term “validation” throughout .NET. It appears in Windows Presentation Foundation (WPF) and Windows Workflow Foundation (WWF) namespaces and in the Enterprise Library as well. Microsoft also uses the following class names:
ValidationError, ValidationErrorCollection, ValidationManager, ValidationResult, ValidationRule, ValidationStatus, ValidationType

IdeaBlade is integrating DevForce with Microsoft‟s WPF and WWF. You are likely doing the same. We will all become confused if we cannot easily distinguish among the same or very similar names. So “Verification” it is. We will continue to say “validation” when we speaking in general terms; we will use the term “verification” (and its variants) when we refer specifically to the DevForce classes located in the IdeaBlade.Verification namespace.

Whither WPF Validation?
We can‟t leave this digression without a parting comment about validation in Microsoft‟s Windows Presentation Foundation. WPF validation concentrates on presentation of validation results within a WPF user interface. This is a vital aspect of any validation strategy. At present, most applications punish the user for the developer‟s own design failings. We need better UIs and better means to guide users rather than humiliate them. DevForce‟s “verification” concentrates on the validation process. It complements WPF by producing the rich validation results necessary to deliver an effective user experience. We will address the integration of these mechanisms in a separate document.

Getting Started
The easiest point of entry to DevForce verification is through the DevForce Object Mapper. We‟ll assume that you are familiar with the Object Mapper and have an existing, working application with its own business object model. See the “Hello DevForce” topic document in the DevForce Learning Resources under “Introduction to DevForce” for a walk-through of the Object Mapper.

Validation-Related Settings In the Object Mapper
1. 2. Launch the Object Mapper in Visual Studio. Select an Entity Model node in the tree control and drop down the Verification Interceptors ComboBox in the properties pane:

332 | P a g e

IdeaBlade DevForce

Validation Through Verification

3.

Note the default setting of “Both”. This means that a check will be made for applicable verifiers before and after the value of any property of any class in your model is set (which is to say, before and after a proposed value is actually pushed into the business object). However, this default setting setting can be overridden for any individual property using the ComboBox in the “Verification Setter Mode” column of the property grid:

333 | P a g e

IdeaBlade DevForce
4.

Validation Through Verification

Return to the Entity Model dialog by selecting the Entity Model node in the navigation pane. This time note the CheckBox labled “Generate Validation Attributes” and the RadioButtons labeled “DevForce” and “.NET”. The CheckBox determines whether properties generated into your object model will be decorated with any sort of validation attribute; the RadioButtons determine whether the attribute used will be one of those defined in the .NET framework, or DevForce‟s validation attribute

Generated Property Code
Here we‟ll show you the three basic styles of validation-related attribute decoration that you can select in the Object Mapper.

DevForce Validation Attributes
Here is the FirstName property of an Employee object as generated with the settings shown above:

334 | P a g e

IdeaBlade DevForce

Validation Through Verification

C#

VB

IbVal is an alias for the IdeaBlade.Validation namespace, defined at the top of the code file. The IbVal.ValidateProperty attribute tells DevForce to check for verifiers when this property gets set, and run any whose settings define them as appropriate. The IbVal.StringLengthVerifier sets a maximum length on the (text) value, and its IsRequired argument declares the property non-nullable. The IbCore.MaxTextLength attribute is used only with WinForm user interfaces, by the DevForce BindingManagers. It extends a legacy established in DevForce Classic.

.NET Validation Attributes

335 | P a g e

IdeaBlade DevForce

Validation Through Verification

Here is the generated code that results from changing the RadioButton selection to “.NET”:

C#

VB

Note that the non-nullability (i.e., Required) and string length constraints are specified a bit differently. DevForce can make use of either style of validation attribute. Its own versions provide richer capabilities than the .NET counterparts, but if you need your code to use the .NET attributes for reasons of your own, DevForce cooperates.

Neither DevForce Nor .NET Validation Attributes
Here is the generated code we get after unchecking the “Generate Validation Attributes” CheckBox:

336 | P a g e

IdeaBlade DevForce

Validation Through Verification

C#

VB

DevForce still indicates that validation should be run against the property, so that if you define any custom verifiers using other mechanisms (to be discussed shortly) they will get exercised; but there is now no indication from the property attributes that the property either requires a value, or is limited to any particular length.

337 | P a g e

IdeaBlade DevForce

Validation Through Verification

Impact of Verifiers on the User Interface – A Caution
As you‟ve just seen, DevForce automatically generates verifiers to enforce constraints implicit in column definitions of the backing datastore. In our example, we saw that, because the FirstName column of the Employee table behind the Employee object class is defined as non-nullable and 30 characters wide, DevForce generates a verifier to impose corresponding constraints on the Employee.FirstName property. Suppose our application displays Employee information on a form, as shown below. Suppose our end user deletes the contents of the Employee “First Name” textbox to start. She attempts to move out of the field … but can‟t. Instead, she sees this:

She can‟t change any value on the form. She can‟t move to a different Employee. She can‟t even close the form! The user must enter something in the “First Name” textbox. What happened?     DevForce generated a validation rule, known in DevForce as a Verifier. It put code in our business object that caused validation to be invoked when the user tried to set the value. The validation failed when the user cleared the “First Name” control, since the value for that property is Required. The form responded by locking the text box and displaying an “error bullet” with an informative message.

This behavior can be addressed in various ways – mainly quite simple and straightforward. We address the topic further as it pertains to WinForms in the section “Verification and WinForms User Interfaces” later in this document. But that‟s all another discussion, so on with our discussion of the basics!

338 | P a g e

IdeaBlade DevForce

Validation Through Verification

Now That You‟ve Been Initiated (and Before We Enter the Forest): A Quick Overview of the Mechanics
What you‟ve seen thus far in this document are verifiers defined with property attributes. You get these for free with DevForce. Simply by flipping a switch in the Object Mapper (actually, it‟s set ON by default!) you get all of the basic constraints defined on column values by the backing database carried through to your business classes in the code that the Object Mapper generates. This, however, is by no means the extent of the validation capabilities that DevForce provides. We‟ll describe the Verification system in detail later in this document, but first, so you don‟t lose the forest for the trees, let‟s take a moment for a high-level overview of how the system works, what you can do with it, and how you do it.

Creating a Verifiers Collection for a Type
When one of your business types is first instantiated during an application session, a VerifierEngine is invoked to discover all verifiers applicable to that type. Some of these verifiers are defined using property attributes, as you‟ve seen; but verifiers may also be defined in .NET code -- in very flexible and powerful ways! – and even in XML in the app.config file. The VerifierEngine discovers all of the verifiers, however encoded, and creates an in-memory collection of them. They are then available for it to call upon whenever needed.

Execution Modes for a Verifier
Each of the verifiers has its own properties which tell the VerifierEngine when it should be run. For example, you can define a verifier so that it runs before a proposed new property value is pushed into the business object (the default for the generated verifiers); or only after; or both. You also want most verifiers to run whenever an entire instance of a type is being validated. To specify these things, you specify the ExecutionModes for a verifier. Your selections are:        OnPostSetTriggers OnPreSetTriggers Instance InstanceAndOnPostSetTriggers InstanceAndOnPreSetTriggers All Disable

PreSet triggers run before the proposed new value is pushed into the business object; PostSet triggers run after. Although it is desirable never to allow invalid values into a business object in the first place (hence the default ExecutionMode of OnPreSetTriggers for generated verifiers), sometimes you can only know that a value is invalid by comparing it to the value of some other property of the same or another object. In that event, you must allow the new value to be pushed into the business object, because the fix for an invalid value may actually be to change the value of something else.

Defining Triggers for a Verifier
Consider, for example, a constraint that specifies that the HireDate for an Employee must be later than the Employee‟s BirthDate. That seems a pretty reasonable requirement, but suppose when the user enters a HireDate of today, this rule is found to have been violated. But the Employee really was hired today, and the problem is that the BirthDate previously entered for the Employee was in error, specifying the year 2015 when it should have said 1985. If we prevent the new HireDate value from being entered into that property, we‟ll subject the user to a lot of unnecessary work. She‟ll have to clear the new HireDate, even though it is entirely correct, and then go fix the BirthDate value, and then come back and re-enter the new HireDate value, even though she got it right the first time.

339 | P a g e

IdeaBlade DevForce

Validation Through Verification

Users don‟t have a lot of tolerance for this sort of tom-foolery! It‟s confusing, or irritating, or often both, in succession. A verifier that applies to BirthDate doesn‟t have to be triggered by a change to that property. You can, instead – or additionally – specify that you want it called when the value of HireDate is changed. You can even set it to be triggered by a change to a property of some other type; and not just any instance of that type, but one that happens to be specifically related. For example, suppose you have a rule that says that the OrderDate on an Order taken by Employee X must be greater than or equal to that Employee‟s HireDate. You could define this rule (as a verifier) on the Employee type, but specify that it should be triggered not only by a change to Employee.HireDate, but equally by a change to the OrderDate property of any Order written by that Employee!

Pre-Defined Verifiers
DevForce ships with a number of pre-defined verifiers that can be subclassed in flexible ways: examples include the DateTimeRangeVerifier, the PhoneNumberVerifier, the NamedRegexPatternVerifier, the StringLengthVerifier, the RequiredValueVerifier, and a number of others. In addition, it defines a generic DelegateVerifier<T> that provides unlimited verification capabilities: you can use it to define any sort of rule you need or can imagine.

Fast Track to Verification for the Prudently Impatient
You‟ll find examples of all of these types of verifiers in the sample code solutions included in the Validation area of the Learning Resources. If you want to see the code right away, feel free to take a side trip now and jump in. There‟s a decent chance you‟ll be able to figure out how to do most of what you need to do from what you‟ve already read and from the samples there. If not, come back and read some more.  For our purposes here, it‟s time to dig into the details of the Verification facilities.

Verification Types Overview
Most Verification classes are defined under the IdeaBlade.Validation namespace and deployed in a single DevForce class library, IdeaBlade.Validation.dll.83 This section is a guide to the key Verification constructs in that library.

Main Verification Classes

83

There are some legacy classes having to do with WinForm support that are defined in IdeaBlade.Verification.dll. You will not need these if doing WPF or Silverlight applications.

340 | P a g e

IdeaBlade DevForce

Validation Through Verification

These are the main classes, the classes at the heart of the DevForce Verification paradigm. Type
VerifierEngine

Description A VerifierEngine maintains a list of Verifier instances in its VerifierCollection and executes them at the appropriate times, accumulating their VerifierResults in its VerifierResultCollection. A Verifier should not run independently. Rather it should be evaluated by a VerifierEngine and the caller should reap the results from the engine‟s Execute method when it finishes. A Verifier validates the state of an object. This is the abstract base class for a family of verifiers described below. execution produces a VerifierResult object. This object, in addition to signaling validation success or failure, contains detailed information about the outcome and the context of the verifier‟s execution.
Verifier

Verifier VerifierCollection VerifierResult VerifierResultCollection

TriggerItem

A TriggerItem identifies something like an “event”. When the VerifierEngine executes in the context of this “event”, it evaluates all verifiers attached to that TriggerItem. A property is the most commonly encountered TriggerItem. The setting of that property is the associated “event”. The engine looks for all verifiers that are attached to that property “in the right way” and executes them.

TriggerLink

Triggered verifiers have one or more TriggerLinks, each of them connecting the verified object to a TriggerItem. The TriggerLink specifies both the TriggerItem and a “path” back to the verified object. In the case of a simple property it is the very short path from the property to the instance as in the path from Employee.FirstName to Employee. Developers can write complex paths that “navigate” from a triggering “event” on an object that is far removed from the object being validated. For example, changing a customer‟s credit limit property (the trigger) could stimulate verification of all outstanding orders (the verified objects) related to (the path to) that customer.

The Verification types are all interrelated, but we separate them by category for explanatory purposes. Verifiers VerifierResult Triggers VerifierEngine PropertyValueVerifiers The Verifier class and its supporting types.
VerifierResult

and related types. and related types.

TriggerItem, TriggerLink,

The VerifierEngine and its supporting types Describes the pre-defined verifiers for the most commonly encountered validation cases.

Verifiers
A Verifier validates the state of an object. A Verifier can only run after it is instantiated by a VerifierEngine.

341 | P a g e

IdeaBlade DevForce

Validation Through Verification

The Verifier class is the abstract base class for derived verifiers that are attuned to specific validation tasks; we cover these derived classes in a separate section below. The following table highlights significant members of the Verifier class.

Class Member
AddTrigger, AddTriggers ApplicableType DefaultSortOrder Description ExecutionModes GetDisplayName

Description Adds a TriggerLink to the verifier. The verifier subsequently responds to “events” associated with the TriggerItem in that TriggerLink. The verifier validates objects of this type Static property that reveals the SortOrder given to new verifiers by default. The description of the verifier as displayed to the user. The situations in which the verifier should run. The value combines flags from the VerifierExecutionModes enumeration. Returns a name for the member of a type as it should be displayed to the user. “First Name” might be the display name for the FirstName property of an Employee. The method is often used to construct the Description from a message template. The integer position of this Verifier in its engine‟s list. Runs the method that indicates if this Verifier applies to a given instance when run in a particular context. Returns a VerifierApplicabilityresult. Set to one of the VerifierOnErrorMode enumerations (Stop, Continue) that tells the engine whether it should stop or continue verifying if this verifier produces an errant VerifierResult. Continue is the default. Removes a TriggerLink from the verifier. The verifier no longer responds to “events” associated with the TriggerItem in that TriggerLink. The engine executes verifiers in sorted order. The engine sorts verifiers first by this SortValue and, when those are the same, by the order in which the verifiers were added to the engine (InitializationOrder). Thus, the developer can influence verifier processing order by setting this SortValue. Returns the TriggerLinks attached to this verifier. Configuration data for the Verifier. Every Verifier is created with a VerifierArgs instance, either explicitly or implicitly. The engine to which the verifier is attached.

InitializationOrder IsApplicable

OnErrorMode

RemoveTrigger, RemoveTriggers

SortValue

TriggerLinks VerifierArgs VerifierEngine

342 | P a g e

IdeaBlade DevForce
Class Member
Verify

Validation Through Verification
Description For Internal Use. This method implements the verifier‟s core validation test and can be overridden by a derived class. The method is public so that a VerifierEngine can call it at the appropriate time. Developers cannot call it directly – it will throw an exception.

The following is list of types that are closely related to Verifier. The list is (mostly) alphabetical to make it easier to locate a type and for lack of more compelling organizational principle. Type
ApplicabilityConstraint(Of T)

Description that determines if a Verifier applies to a particular object given the current TriggerContext and VerifierContext. It returns a VerifierApplicability object. T is the type of the verified object.
Delegate

The developer can invoke this constraint directly by calling Verifier.IsApplicable.
DelegateVerifier(Of T)

A harness for a custom verifier that validates an object of type T. The developer can build almost any kind of verifier with an instance of this class. The developer writes the validation test inside a VerifierCondition delegate (hence the name) and includes a reference to the delegate method in the DelegateVerifier constructor. This verifier can be configured with triggers in the same way as all other verifiers.

DelegatePropertyValueVerifier (Of T)

A harness for a custom property verifier that validates an object of type T. Used to build a verifier triggered by a single property with the purpose of evaluating the proposed or actual value of that property. The developer implements the validation test inside a method that conforms to the ValueVerifierCondition delegate and passes a reference to the method in the DelegatePropertyValueVerifier constructor. The verifier behaves like any of the predefined PropertyValueVerifier classes described below.

PropertyValueVerifierAttribute

Each of the PropertyValueVerifiers can be specified declaratively by decorating a property with the corresponding PropertyValueVerifierAttribute. An object passed to a Verifier when it is triggered by a TriggerItem. The object provides the Verifier with information about what triggered it. Trigger classes are covered separately below.
Delegate that determines if a value passes its verifier VerifierResult. T is the type of the verified object.

TriggerContext

ValueVerifierCondition(Of T)

test. It returns a

The developer can supply such a delegate as an argument to the constructor of a DelegatePropertyValueVerifier.
Verifier

Abstract base class for a family of Verifiers. A Verifier validates the state of an object and returns a VerifierResult containing detailed information about the validation outcome and the context of the verifier‟s execution.

343 | P a g e

IdeaBlade DevForce
Type
Verifier(Of T) VerifierApplicability

Validation Through Verification
Description Strongly typed abstract subclass of Verifier where T is the type of the verified object. The Verifier determines if it applies to an object it is validating based on an ApplicabilityConstraint. That constraint method returns an object of this type which contains both a VerifierApplicabilityCode and an optional message. An enumeration of result codes emerging from evaluation of an ApplicabilityConstraint. These args carry configuration data for a Verifier. Every Verifier is created with a VerifierArgs instance, either explicitly or implicitly; every Verifier retains a reference to that instance. This is also the base class for a family of VerifierArgs classes, each strongly typed to fit closely with it corresponding Verifier class. The ListVerifier has its ListVerifierArgs for example. Abstract base class for a family of Attribute classes that enable declaration of a Verifier by decoration with an attribute. For example, we can declare that the FirstName property has a StringLengthVerifier by adorning it with the StringLengthVerifierAttribute. A collection of Verifier instances. The collection implements many of the features of List<Verifier> and, most importantly, many Find overloads to facilitate extraction of Verifier subset collections. that implements a validation test on an object of type T. This is the beating heart of the developer‟s custom DelegateVerifier.
Delegate

VerifierApplicabilityCode VerifierArgs

VerifierAttribute

VerifierCollection

VerifierCondition(Of T) VerifierContext VerifierOnErrorMode

The VerifierEngine executes a Verifier in a particular context and makes this context available to the Verifier as it executes. The enumeration (Stop, Continue) that tells the engine whether it should stop or continue verifying if this verifier produces an errant VerifierResult. The developer can set the Verifier.OnErrorMode to a value from this enumeration. The exception thrown when the Verifier itself fails to execute properly, i.e. when the Verifier throws an exception; that exception is included in the InnerException. Not to be confused with the VerifierResultException. A flag enumeration (Disabled, Instance, OnPostsetTriggers, OnPresetTriggers) that describes the situations in which a Verifier can run. The VerifierEngine, while executing in one of these situations, runs the verifiers that have a matching ExecutionModes flag. The developer can set a Verifier to run in multiple situations by setting its ExecutionModes to a combination of these flags constructed by “or”ing them together. The VerifierExecutionModes enumeration exposes several of the most common combinations (e.g. All which translates to Instance | OnPostsetTriggers | OnPresetTriggers ).

VerifierException

VerifierExecutionModes

VerifierResult

344 | P a g e

IdeaBlade DevForce
84

Validation Through Verification

Verifier execution produces a VerifierResult object . This object, in addition to signaling validation success

or failure, contains detailed information about the outcome and the context of the verifier‟s execution.

84

We may say casually that a Verifier returns a VerifierResult but this is not strictly correct and might mislead the developer into improper use of verifiers. While it is true that the Verifier.Verify method returns a VerifierResult, that method executes only one part of the Verifier‟s validation logic and should not be called by developer code. The Verifier should be executed by a VerifierEngine which stores the VerifierResult in its own results collection. The developer retrieves those results from the engine.

345 | P a g e

IdeaBlade DevForce

Validation Through Verification

The following table highlights significant members of the VerifierResult class. Class Member
Description IsOk ResultCode TargetInstance TriggerContext Verifier VerifierContext

Description Description of this result. The text is usually fashioned in a form suitable for display when the result is “not Ok”. True if the result has one of the “Ok” VerifierResultCodes. The VerifierResultCode for this result. The object that was verified. The TriggerContext in which the verifier was executed. The Verifier whose execution produced this result. The VerfifierContext in which the verifier was executed.

The following are the important types that are most closely related to VerifierResult. Type
VerifierResultCode

Description Enumeration summarizing the result of verification in a single value. While there are several codes, each is a flavor of a binary outcome: success (Ok) or failure (Error). The codes at this writing are: Error, ErrorInsufficientData, Ok, OkNotApplicable, OkWarning.

VerifierResultCollection

The VerifierEngine accumulates a collection of VerifierResult instances which it returns as a VerifierResultCollection from its Execute method. The collection implements most of the features of Collection<VerifierResult> and, importantly, many Find overloads to facilitate extraction of VerifierResult subset collections. The caller of the VerifierEngine may want to throw an exception if it detects an errant VerifierResult. The VerifierResultException is a strongly-typed exception for this purpose; it can report the initial errant VerifierResult as well as a VerifierResultCollection of other results that may be useful to a handler of the exception. The Entity.BeforeSetValue and Entity.AfterSetValue methods are examples of VerifierEngine callers that throw VerificationResultExceptions. The Verifier and the VerifierEngine do not themselves throw this exception; they merely report errors by providing VerifierResults. Do not confuse the VerificationResultException with the VerificationException. A Verifier or VerifierEngine will throw a VerificationException when the verifier fails to execute properly. Improper verifier execution is not the same as an invalid object condition.

VerifierResultException

346 | P a g e

IdeaBlade DevForce
Type
VerifiersErrorsResource.resx

Validation Through Verification
Description A resource file of predefined error message templates. This resource file contains the message templates for constructing VerifierResult descriptions. The developer can substitute a different .NET ResourceManager that governs a wider set of message templates and resources files for different locales; the developer‟s main resource file must contain definitions for all of the message names defined in the VerifiersErrorsResource.resx. The DevForce distribution includes this resource file as a starting place for the developer‟s own resource file (and satellite translation files).

Triggers
Evaluation of a Verifier may be triggered by one or more “events”. “Events” is in quotes because the mechanism, while it feels like an event, does not use the .NET event. The exact mechanism is introduced here and covered more extensively elsewhere in this document. Setting a property is likely the most commonly encountered trigger. Setting Employee.FirstName, for example, could trigger evaluation of a Verifier that checked if the FirstName string value is present and not longer than thirty characters. The Verifier that checks the FirstName string length can be evaluated independently of any trigger. It could be evaluated during validation of an Employee instance85. But we often want to verify the value the moment the user enters the text. Accordingly, the developer attaches a trigger to that Verifier – a trigger bound to the Employee.FirstName property.

TriggerItem
DevForce represents the triggering Employee.FirstName property as a TriggerItem. A TriggerItem is little more than a .NET Type and the name of some member on that type. If a TriggerItem represents a property, the member name is the property name.

TriggerLink
It will not always be enough just to know the TriggerItem. We may have to find our way back from the trigger to the object being verified. This is easy when the TriggerItem refers to a property of the object being verified. If the trigger is Employee.FirstName and the Verifier targets the Employee object, it is obvious that “the way back” from the property to Employee involves no effort at all: the triggering object and the verified object are the same. On the other hand, we may want to evaluate the verifier when a value changes on some different object. For example, we may want to verify that an Order‟s total price is still valid if the price of any of its OrderDetail items goes up. The OrderDetail is not the same object as the Order we need to verify. The TriggerLink provides the path from the OrderDetail whose price changed to its parent Order which must be verified.

85

Evaluation in this situation is called “Instance Verification”.

347 | P a g e

IdeaBlade DevForce

Validation Through Verification

The TriggerLink holds both the end point (the TriggerItem for OrderDetail.Price) and the method to navigate from the trigger object (OrderDetail) to the object to verify (Order). This method is called the TriggerTargetNavigator. We‟ll cover all of this in greater depth later; for now we look at the classes and other types involved in triggering execution of a Verifier. Type
TriggerContext

Description An object passed to a Verifier when it is triggered by a TriggerItem. The object provides the Verifier with information about what triggered it. A TriggerItem identifies something like an “event”. A A TriggerItem is defined by the Type of the triggering object and the name of a member on that type that does the triggering.

TriggerItem

TriggerLink

A TriggerLink specifies both the TriggerItem and a path back to the verified object. The path is implemented by a TriggerTargetNavigator method. That method is null when the triggering object and the object to be verified are the same as they are when we trigger a verifier for Employee.FirstName when the user sets that property. The navigator could be a.NET property path (e.g. a PropertyDescriptor). It could also be a custom method capable of bridging the two object types; see TriggerTargetNavigator.

TriggerTargetNavigator TriggerTiming

for “navigating” from a TriggerItem to the object being verified. See TriggerLink.
Delegate

An enumeration available within the TriggerContext. It indicates when a verifier was “triggered”. There are two choices: Preset and Postset. Properties are the most common triggers so a TriggerTiming typically indicates whether the verifier was evaluated before or after the property was set.

VerifierEngine
Verifiers do not execute themselves86. They are executed by a VerifierEngine instance. Each engine maintains a list of Verifier instances and evaluates them at the “appropriate” times based on a variety of factors that include (but are not limited to) properties of the verifiers themselves.

86

You should not call the Verifier.Verify method directly even though it is public. That method performs some – but not all ! – of the validation work and will throw an exception if called outside a VerifierEngine.

348 | P a g e

IdeaBlade DevForce

Validation Through Verification

The details of the engine are covered elsewhere in this chapter. Here are the types most relevant to understanding it. Type
VerifierCollection

Description The engine maintains a collection of Verifier instances, accessible via one of the GetVerifier method overloads. The collection implements most of the features of List<Verifier> and, importantly, many Find overloads to facilitate extraction of Verifier subset collections. The VerifierEngine executes a Verifier in a particular context. The engine creates an instance of a VerifierContext before every validation run (a “batch”) and makes it available to the verifiers in that run. Each verifier can both see and modify the context. The developer can activate a VerifierBatchInterceptor delegate method that can see and modify the context. The context includes a great deal of useful information including a reference to the engine itself, the BatchId of the engine‟s current validation run, the VerifierResultCollection of results accumulated so far, the currently executing Verifier, and a CustomContext object supplied by the developer.

VerifierContext

VerifierEngineCreatedEventArgs

provided to a VerifierEngine.VerifierEngineCreated event handler. The VerifierEngine class raises this static event after creating a new VerifierEngine instance. The developer can attach a handler to consistently configure every new instance.
EventArgs

VerifierEngine. PropertyNameTranslator

Delegate method that takes a type and a string (presumed to be the property name) and returns the string that will be injected into the message produced by the verifier. A verifier description or message should appear in the user‟s preferred language. The message templates can be localized but they often have a placeholder for the property name. The “{0}” in the message “{0} is required” will be filled by a property name at runtime. This name should be localized as well. The VerifierEngine.PropertyNameToDisplayNameTranslator property takes such a delegate. Delegate method called by the VerifierEngine after every verifier evaluation in a batch and once more at the end of the batch. A resource file of predefined error message templates. The exception thrown when the Verifier itself fails to execute properly within the engine, i.e. when the Verifier throws an exception; that exception is included in the VerifierException.InnerException. Not to be confused with the VerifierResultException. A flag enumeration (Disabled, Instance, OnPostsetTriggers, OnPresetTriggers) that describes the situations in which a Verifier can run. The VerifierEngine, while executing in one of these situations, runs the verifiers that have a matching ExecutionModes flag.

VerifierEngine. VerifierBatchInterceptor VerifiersErrorsResource VerifierException

VerifierExecutionModes

349 | P a g e

IdeaBlade DevForce
Type
VerifierOnErrorMode

Validation Through Verification
Description The developer can set the Verifier.OnErrorMode to a value from this enumeration (Stop, Continue). The value tells the VerifierEngine whether it should continue (the default) or stop verifying if this verifier reports that its validation failed. The result of executing a Verifier. It contains detailed information about the outcome and the context of the verifier‟s execution The VerifierEngine accumulates a collection of VerifierResult instances which it returns as a VerifierResultCollection from its Execute method. The collection implements most of the features of Collection<VerifierResult> and, importantly, many Find overloads to facilitate extraction of VerifierResult subset collections. Args of the VerifierEngine.VerifiersChanged event, raised when a verifier is added to or removed from the engine or a trigger is added to or removed from a verifier already held by the engine. Enumeration of the types of changes reported in
VerifiersChangedEventArgs

VerifierProviderAttribute VerifierResult VerifierResultCollection

VerifiersChangedEventArgs

VerifiersChangedType

PropertyValueVerifiers
The class diagram for Verifier and its derived classes as of this writing looks like this:

Most of the verifier classes are PropertyValueVerifiers. A PropertyValueVerifier tests a property value. Technically, it is a Verifier attached to single TriggerItem which is a property on the object being validated.

350 | P a g e

IdeaBlade DevForce

Validation Through Verification

The value to test may be the proposed property value (prior to the property set) or the current value (after the property was set). Many application validations are property validations and most of these resolve into some variation of just a few kinds of verifier: required, range or length, and membership in a list.

Attribute Classes
Verifiers can be prescribed programmatically and added to the VerifierEngine at runtime. It is sometimes convenient to prescribe them programmatically by adorning properties with attributes. The DevForce includes a number of PropertyValueVerifierAttribute classes to facilitate this approach. The DevForce Verification library covers many of these verifiers and attributes; of course you can easily extend them or write your own.

Null property values
We have to check for null before we can test a property value. In many cases, null values are not permitted. Rather than oblige the developer to specify both a RequiredValueVerifier and the verifier of interest, all PropertyValueVerifiers include an IsRequired parameter; the base, abstract PropertyValueVerifier evaluates IsRequired before handing the value on to the derived verifier classes. The outcome of the test is often arbitrary in the face of a null value; is a null BirthDate before or after the minimum date in a range check? You should be sure you know how the verifier handles nulls. The following table highlights significant members that are specific to the PropertyValueVerifier class. Type
DisplayName

Description The displayable name of this verifier; this is typically the display name for the property it verifies. See also the Verifier.GetDisplayName method. Returns the value of this property as it currently is in the object being verified. This value could be compared to the proposed value if the verifier is executing in a “preset” context. Returns true if a property value is required (if it cannot be null). The .NET PropertyDescriptor for the property it verifies. Returns the strongly type PropertyVerifierArgs that configure this verifier.

GetPropertyValue

IsRequired PropertyDescriptor TypeVerifierArgs

351 | P a g e

IdeaBlade DevForce
The following are types closely related to this class and its derived classes. Type
DelegatePropertyValueVerifier (Of T)

Validation Through Verification

Description The foundation of a custom property verifier. The developer implements the validation test inside a ValueVerifierCondition delegate of his own devising. It behaves otherwise like any of the predefined property verifiers. A Regex expression for use with the RegexVerifier. You can use one of the pre-named static patterns or create your own “named” Regex pattern. Verifiers that apply to a single property of an object. The DevForce predefined PropertyValueVerifiers, as of this writing, are:
DateTimeRangeVerifier DecimalRangeVerifier DelegatePropertyValueVerifier(Of T) DoubleRangeVerifier Int32RangeVerifier Int64RangeVerifier ListVerifier RangeVerifier RegexVerifier RequiredValueVerifier StringLengthVerifier

NamedRegexPattern

PropertyValueVerifier

PropertyValueVerifierAttribute

Each of the PropertyValueVerifiers can be specified declaratively by decorating a property with the corresponding PropertyValueVerifierAttribute.
DateTimeRangeVerifierAttribute DecimalRangeVerifierAttribute DelegatePropertyValueVerifierAttribute DoubleRangeVerifierAttribute Int32RangeVerifierAttribute Int64RangeVerifierAttribute RangeVerifierAttribute RegexVerifierAttribute RequiredValueVerifierAttribute StringLengthVerifierAttribute

RangeVerifier(Of T)

A generic range Verifier where T is the type of value tested (not the type of the verified object). A range verifier accepts arguments specifying minimum and maximum (either optional) and whether the range includes or excludes either end point.
Delegate that determines if a value passes its verifier VerifierResult. T is the type of the verified object.

ValueVerifierCondition(Of T)

test. It returns a

The developer can supply such a delegate as an argument to the constructor of a DelegatePropertyValueVerifier.

352 | P a g e

IdeaBlade DevForce
Adding Attributed Verifiers to Generated Properties

Validation Through Verification

Previously you saw an example of the code generated by the Object Mapper for a string-valued property, FirstName. The property definition was decorated with a StringLengthVerifier :

C#

[StringLengthVerifier(MaxValue=30, IsRequired=true)] public String FirstName { ... }

VB

<StringLengthVerifier(MaxValue:=30, IsRequired:=True)> _ Public ReadOnly Property FirstName() As String ... End Property

To add an attributed verifier to a custom property defined in your developer partial class, you would simply add the appropriate attribute – such as the StringLengthVerifier attribute shown above – to the property definition. Clearly you can‟t do the same for properties defined in the designer code file generated by the DevForce Object Mapper. That file, and the code in it, “belongs” to the Object Mapper, which reserves the write to overwrite it whenever ordered to do so. Nevertheless, you can still apply your own attributed verifiers to generated properties. You do this by means of a “buddy” class that partners with your business class and contributes additional metadata to it. In the example below, we‟ve added such a buddy class to the developer partial class file for the Customer type, Customer.cs. In the buddy class, we‟ve decorated the static property CompanyName with the StringLengthVerifier, assigning our own MaxValue, which is more restrictive than the one generated for CompanyName in the designer code file.

353 | P a g e

IdeaBlade DevForce

Validation Through Verification

C#

...

using IbVal = IdeaBlade.Validation; using DataAnnot = System.ComponentModel.DataAnnotations;
namespace DomainModel {

[DataAnnot.MetadataType(typeof(CustomerMetadata))]
public partial class Customer : IdeaBlade.EntityModel.Entity { ... } /// <summary> /// The buddy class for Customer /// </summary> public class CustomerMetadata { /// <summary> /// Override CompanyName to make it required. /// </summary>

[IbVal.StringLengthVerifier(MaxValue = 10, IsRequired = true)]
public static string CompanyName; } }

VB

... Imports IbVal = IdeaBlade.Validation Imports DataAnnot = System.ComponentModel.DataAnnotations Namespace DomainModel

<DataAnnot.MetadataType(GetType(CustomerMetadata))> _
Partial Public Class Customer Inherits IdeaBlade.EntityModel.Entity ... End Class ''' <summary> ''' The buddy class for Customer ''' </summary> Public Class CustomerMetadata ''' <summary> ''' Override CompanyName to make it required. ''' </summary>

<IbVal.StringLengthVerifier(MaxValue := 10, IsRequired := True)> _
Public Shared CompanyName As String End Class End Namespace

Important!! Note, in order that the Verification engine should be aware of the Customer type‟s buddy class, that we have decorated the Customer class with the MetadataType attribute from the System.ComponentModel.DataAnnotations namespace. Don‟t forget to do that: otherwise the verifiers defined in your buddy class will not be enforced!

354 | P a g e

IdeaBlade DevForce

Validation Through Verification

Verification Deep Dive
Now that we have toured the Verification types we are ready to look more closely at the major types.

Verifiers
We use an instance of the DevForce abstract Verifier class to implement a validation rule. Verifier = Validation Rule A verifier‟s primary task is to render judgment on the validity of an object. It isn‟t suppose to change the object, just evaluate it and pronounce the object valid or invalid. That‟s a big job – too big for any one verifier instance 87. So we create lots of verifiers each of which limits itself to evaluating one aspect of an object such as the string length of a single property. Each verifier produces a VerifierResult object which, at its most basic, (a) indicates success ( Ok) or failure (Error) and (b) provides a message (the VerifierResult.Description) for display to a user. An object is “valid” if the accumulated results of individual verifiers are all “ok”. The DevForce Verification library contains several predefined Verifier subclasses88 as well as several higher level abstract classes that allow developers to construct their own verifiers. Verifiers don‟t execute on their own. They have to be evaluated by a VerifierEngine which means we have to tell the engine about them by registering configured instances of some verifier class with the engine. While we can register verifier instances programmatically, it is often more convenient to let the VerifierEngine discover them – a process we‟ll get to when we consider the engine in detail. For now we‟ll talk about registration as if we always took an active hand in it. Each verifier has an ApplicableType which is the type of object that the verifier can verify. Verifiers with an ApplicableType of a .NET base type are presumed to be applicable to all subclasses of that base type. The VerifierEngine ensures that verifiers registered for a base class are propagated automatically the verifier collection of all derived types. Imagine that you had an abstract class called Produce and a bunch of subclasses – Carrot, Apple, Potato. When you attach a verifier to Produce.Name, that same verifier applies to Carrot.Name, Apple.Name, and 89 Potato.Name .

Verifier Execution
A verifier cannot be executed until it has been added to a VerifierEngine. An individual verifier instance can be attached to only one VerifierEngine at a time. Verifiers are executed in the order that they were added to the VerifierEngine. It is possible to modify the order by setting the SortValue property on each verifier. A VerifierEngine runs in one of three “Execution Modes” at a given time. How we call it determines the mode. Instance Verification

87 88 89

While it is possible to write a single super verifier that does it all, it would be unwise to do so. See the class diagram above. It is the same verifier even if Potato.Name overrides Produce.Name. The developer can remove or replace the propagated verifier for Potato.Name by manipulating the Potato verifiers after they have been built. Carrot.Name and Apple.Name will be unaffected.

355 | P a g e

IdeaBlade DevForce
Preset Trigger Verification Postset Trigger Verification

Validation Through Verification

We cover these modes in detail in the “Invoking Verification” section. The point to note here is that the engine will only evaluate the verifiers that are configured to run in a compatible execution mode. Thus, if the engine is running in “Preset Trigger” mode and the verifier‟s ExecutionModes = InstanceAndOnPostsetTriggers, the verifier will not be evaluated; it will be evaluated when the engine runs in either instance or postset trigger mode. When a VerifierEngine evaluates a verifier it calls two verifier methods: IsApplicable and Verify. C#
public virtual VerifierApplicability IsApplicable( Object pItemToVerify, TriggerContext pTriggerContext, VerifierContext pVerifierContext); public abstract VerifierResult Verify( Object pItemToVerify, TriggerContext pTriggerContext, VerifierContext pVerifierContext);

Visual Basic
Public Overridable Function IsApplicable( _ ByVal pItemToVerify As Object, _ ByVal pTriggerContext As TriggerContext, _ ByVal pVerifierContext As VerifierContext) _ As VerifierApplicability Public MustOverride Function Verify(ByVal pItemToVerify As Object, _ ByVal pTriggerContext As TriggerContext, _ ByVal pVerifierContext As VerifierContext) _ As VerifierResult

Observe that engine calls both methods with the same inputs Parameter
pItemToVerify pTriggerContext

Description The object instance to verify. Its type will be the same as or a descendent of the ApplicableType of the verifier. A TriggerContext object that describes how the verifier was triggered – a topic covered elsewhere in this chapter. Note that this value is null (Nothing in VB) when the verifier was not triggered (i.e., during “instance verification”).

pVerifierContext

IsApplicable
The execution cost of some verifiers may be high. We don‟t want to pay that cost if the verifier does not apply in the present circumstances. The developer can specify an IsApplicable method to short-circuit unnecessary verifier evaluation. For example, most validations are irrelevant if the object is marked for delete. We might test for that in our IsApplicable method. The VerifierEngine calls the verifier‟s IsApplicable method first. The IsApplicable method returns a VerifierApplicability object with a VerifierApplicabilityCode. If the code is Yes the engine continues evaluating the verifier. If the code is anything else, the engine stops evaluating, prepares a VerifierResult for this verifier, and moves on to the next verifier.

356 | P a g e

IdeaBlade DevForce

Validation Through Verification

The VerifierResultCode of the prepared VerifierResult will be an “ok” code (VerifierResultCode.OkNotApplicable) if the VerifierApplicabilityCode is No. It will be an “error” code (VerifierResultCode.ErrorInsufficientData) if the VerifierApplicabilityCode is InsufficientData. An applicability test is rarely needed. Accordingly, the base IsApplicable implementation in the abstract Verifier class simply returns VerifierApplicability.Yes.

VerifierContext
The VerifierEngine provides both the IsApplicable and the Verify methods with a VerifierContext defined as follows: This context gives each Verifier information about its calling and executing environment including the engine‟s progress during this particular execution. The context values change over the course of the verification. The VerifierEngine will change them. Each Verifier can change them too.

Initializing the VerifierContext
The initial VerifierContext is either a context created by the engine or a context provided by the code that puts the engine to work. Such code calls one of the engine‟s Execute methods. There are a number of signatures, as we‟ll see later in this chapter, and many of them take a VerifierContext. If the caller provides the context, it will have instantiated the context with this constructor: C#
VerifierContext(VerifierOnErrorMode pOnErrorMode, Object pCustomContext)

Visual Basic
New (ByVal pOnErrorMode As VerifierOnErrorMode, ByVal pCustomContext As Object)

The VerifierOnErrorMode is an enumeration with two values - Stop and Continue – meaning “Stop verifying if you encounter an error” and “keep verifying until there are no more verifiers to evaluate”90. The “CustomContext” can be any kind of object. It is a mechanism to enable the calling code to communicate situational information to the verifiers that know how to interpret that information. If the caller does not provide a VerifierContext, the VerifierEngine constructs one from its own resources: the VerifierEngine.DefaultOnErrorMode and the VerifierEngine.DefaultCustomContext. The application could set these defaults when it creates the engine instance; it can revise them at will.

Features of the VerifierContext
Here is the interface of the VerifierContext C#
public class VerifierContext { public Int64 BatchId { get; } public VerifierOnErrorMode OnErrorMode { get; set; } public Object CustomContext { get; set; }
90

An individual verifier can terminate the batch even if the OnErrorMode is Continue.

357 | P a g e

IdeaBlade DevForce
public public public public public }

Validation Through Verification

VerifierResultCollection VerifierResults { get; } object BatchContext { get; set; } bool EndOfBatch { get; } Verifier Verifier{ get; } VerifierEngine VerifierEngine { get; }

Visual Basic
Public Class VerifierContext Public ReadOnly Property BatchId() As Int64 public Property OnErrorMode() As VerifierOnErrorMode public Property CustomContext() As Object public ReadOnly Property VerifierResults() As VerifierResultCollection public Property BatchContext() As Object public ReadOnly Property EndOfBatch As Boolean public ReadOnly Property Verifier As Verifier public ReadOnly Property VerifierEngine As VerifierEngine End Class

Let‟s walk through them quickly. Calling the engine‟s Execute method initiates a new verification “batch” that lasts for the duration of the method‟s execution. The engine assigns the batch a unique BatchId. As stated earlier, OnErrorMode returns an enumeration with two values - Stop and Continue – meaning “Stop verifying if you encounter an error” and “keep verifying until there are no more verifiers to evaluate”. A verifier can change this value at any time. We‟ve already met the CustomContext containing an arbitrary object defined by the developer and made available either when the engine was called or through its DefaultCustomContext property. Observer that the object can be reset at any time during the batch. The VerifierEngine adds each verifier‟s VerifierResult to the VerifierResultCollection in the context. Verifiers can see prior results and take action accordingly 91. The BatchContext is a means of accumulating and communicating execution state within the batch. It starts null (Nothing in VB). Any verifier can change it, perhaps depositing useful information for downstream verifiers. The EndOfBatch starts false. The VerifierEngine will set it to true after it evaluates the last verifier in the batch. This flag is intended for use by a VerifierEngine.BatchInterceptor, a delegate method called by the engine after it evaluates each verifier – and once more at the end of the batch when it sets this EndOfBatch flag to true. The interceptor could perform “batch cleanup” when it sees the flag set true. The VerifierEngine records the most recently evaluated verifier in the context‟s Verifier property. This property is aimed at the VerifierEngine.BatchInterceptor which may need to take some action after the engine evaluates a particular verifier. The VerifierEngine also registers itself in the context‟s VerifierEngine property. Verifiers don‟t need this – they know to which engine they belong. The VerifierEngine.BatchInterceptor does not know what engine is running; it can find out by looking at this context property.

Custom Verifiers

91

They can even manipulate the VerifierResultCollection itself; one hopes they are prudent in doing so.

358 | P a g e

IdeaBlade DevForce

Validation Through Verification

The Verification library comes with many predefined verifiers that cover the majority of cases. Of course you have to be able to create your own – and you can do so easily. Keep reading and you will see examples. You can find these same examples, in context, in the Learning Unit on Verification that ships with DevForce.

Verifier Result
We expect a verifier to render a binary decision most of the time. It‟s usually a pass / fail test. Accordingly, every verifier returns a VerifierResult with an IsOk property. Either it is or it isn‟t. More nuanced information is also available but there is always a firm yes or no. If the validation failed we probably want to display a message to the user 92 explaining how it failed. The VerifierResult.Description contains the message prepared by the Verifier – a message that may have been translated into the local language and culture. The VerifierResult.Description comes from the Verifier.Description by default. The phrase “First Name cannot exceed 30 characters” serves well both as the description of the Verifier and the message to the user when the entered text exceeds 30 characters.

Customizing the Description
When this is not satisfactory, the developer can customize the message.

Sub-class the Verifier and override the Description property
In this example in which the author wants to drive home the point about keeping the birth date reasonable. The DateTimeRangeVerifier would be fine if not for the message. So the author fills out the DateTimeRangeVerifier and then overrides the Description property. C#
/// <summary>Default Ctor,</summary> /// <remarks> /// BirthDate is not required, /// must be on or after global min date (<see cref="M:MinBirthDate"/>), /// and before today. /// </remarks> public BirthDateRangeVerifier() : base(typeof(Employee), // Type of the object being verified Employee.BirthDateEntityProperty.Name, // Property trigger false,// Non-null value is not required MinBirthDate, true, // starting min date (inclusive) DateTime.Today, false) { } // ending max date (exclusive) public override string Description { // ToDo: Localize get { return "Must be born after " + MinBirthDate.Year.ToString() + "; No time travellers allowed!"; } } }

VB

92

Or perhaps to a log file if we are validating outside of a user interface.

359 | P a g e

IdeaBlade DevForce Create a Delegate Verifier

Validation Through Verification

One of the easiest ways to create a new verifier is to create an instance of one of the delegate verifiers as we showed above. The description is one of the parameters in their constructors. C#
public DelegateVerifier(String pDescription, VerifierCondition<T> pVerifierCondition) public DelegatePropertyValueVerifier(String pDescription, String pPropertyName, bool pIsRequired, ValueVerifierCondition<T> pVerifierCondition)

Visual Basic
' DelegateVerifier Constructor Public Sub New (ByVal pDescription As String, _ ByVal pVerifierCondition As VerifierCondition(Of T)) ' DelegatePropertyValueVerifier Constructor Public Sub New (ByVal pDescription As String, _ ByVal pPropertyName As String, _ ByVal pIsRequired As Boolean, ByVal pVerifierCondition _ As ValueVerifierCondition(Of T))

Sub-class the Verifier and override the Verify() method
The Verifier.Verify method returns the VerifierResult picked up by the VerifierEngine. This gives you complete control over the VerifierResult.Description which you can construct dynamically. This is Verify‟s signature93. C#
public abstract VerifierResult Verify(Object pItemToVerify, TriggerContext pTriggerContext, VerifierContext pVerifierContext);

Visual Basic
Public MustOverride Function Verify(ByVal pItemToVerify As Object,_ ByVal pTriggerContext As TriggerContext, _ ByVal pVerifierContext As VerifierContext) As VerifierResult

Localization and Internationalization
There are two mechanisms for localizing the messages reported through the VerifierResult 1. 2. Use resource files for the message templates Translate the property name that is injected into the template.

For example, the basic message template, “{0} is required”, is ready to use as “PropertyRequired” message. At runtime we plug “First name” or “Last name” or whatever into the slot reserved by “{0}”. If the application will be used by non-English speakers, we‟ll want to translate the template and we‟ll want to translate the property names.

93

Although the method is public and it would seem that you can instantiate all of its parameters, you cannot call it yourself; you will get an exception if you try. This is deliberate; DevForce can ensure proper verifier execution only within a VerifierEngine.

360 | P a g e

IdeaBlade DevForce Message Templates

Validation Through Verification

DevForce ships with standard error and warning message templates. The developer can replace them with a completely custom version. The developer creates the .NET resource files for each language 94. The only requirement is that at least the default file has an entry for all of the DevForce template keys. A copy of the DevForce verification resource file is available from IdeaBlade as a starting point for customization. Visual Studio generates a strongly-typed ResourceManager class to support these custom files. The developer sets the ErrorsResourceManager property of each new VerifierEngine to this ResourceManager as shown. C#
VerifierEngine engine = new VerifierEngine(); engine.ErrorsResourceManager = myResourceManager;

Visual Basic
Dim engine As VerifierEngine = New VerifierEngine() engine.ErrorsResourceManager = myResourceManager

“Property Names”
Most common verifiers apply to a single property and inherit from the PropertyValueVerifier. Their message templates have a slot for the property name and the verifier knows how to fill that slot with the property name after it has been translated. The key to the process is the PropertyNameTranslator. The VerifierEngine has a PropertyNameToDisplayNameTranslator property that takes a PropertyNameTranslator delegate defined as follows C#
public delegate String PropertyNameTranslator( Type pType, String pPropertyName);

Visual Basic
Public Delegate Function PropertyNameTranslator( _ ByVal pType As Type, ByVal pPropertyName As String) As String

The expected implementation takes a type-and-string (e.g. Employee and “FirstName”) and turns it into a translated string. Note that type-and-string also defines a TriggerItem. As with TriggerItem, the string is typically the name of a member of the target type … but it doesn‟t have to be. With that background we are ready to proceed.

94

The .NET practices for localization are beyond the scope of this document.

361 | P a g e

IdeaBlade DevForce

Validation Through Verification

All predefined PropertyValueVerifier subclasses within the Verification library observe the following protocol when preparing a “property name” for insertion into the template: If the engine has a PropertyNameToDisplayNameTranslator, that method is used to translate the property name. If there is no translator, the verifier tries the value of PropertyValueVerifier.DisplayName. If DisplayName is null, the verifier looks for a .NET DescriptionAttribute adorning the object property. It there is no such attribute, the verifier uses the property name. This same protocol can be used within a custom verifier, even one multiple slots for multiple property names and values. The translator is not limited to translating property names.

Triggers
Evaluation of a Verifier may be triggered by one or more “events”. “Events” is in quotes because the mechanism, while it feels like an event, does not use the .NET event. The exact mechanism is introduced here and covered more extensively elsewhere in this document. Setting a property is the most commonly encountered trigger. Setting Employee.FirstName, for example, could trigger evaluation of a Verifier that checked if the FirstName string value is present and not longer than thirty characters. The Verifier that checks the FirstName string length can be evaluated independently of any trigger. It could be evaluated during validation of an Employee instance95. But it is often a kindness to the user if we validate the first name text at the moment she enters it rather than wait for the entire Employee object to be evaluated. Accordingly, the developer attaches a trigger to that Verifier – a trigger bound to the Employee.FirstName property. Property validation of this kind - a property Verifier with an attached property trigger - is extremely popular. It is so popular that DevForce provides the PropertyValueVerifier96 and a host of derived verifiers to make it easy to specify property validation. One approach is to adorn a property with one of the attribute-based versions of the PropertyValueVerifier as we do for the FirstName property in the following example. C#
/// <summary>Gets or sets the FirstName.</summary> [StringLengthVerifier(MaxValue=30, IsRequired=true)] public virtual System.String FirstName { // …

Visual Basic
''' <summary>Gets or sets the FirstName.</summary> <StringLengthVerifier(MaxValue:=30, IsRequired:=True)> _ Public Overridable ReadOnly Property FirstName() As System.String Get '…

A VerifierEngine discovers the attribute and the FirstName property it adorns and then adds a StringLengthVerifier, triggered by the FirstName property, to its list of verifiers.

95 96

Evaluation in this situation is called “Instance Verification”. The PropertyValueVerifier and its kin are covered below.

362 | P a g e

IdeaBlade DevForce

Validation Through Verification

Something similar happens when we add the Verifier programmatically to a list of verifiers that we later add to a VerifierEngine. C#
// Add FirstName StringLengthVerifier to a list of verifiers. verifiers.Add(new StringLengthVerifier( typeof(Employee),"FirstName", true, 1, 30));

Visual Basic
' Add FirstName StringLengthVerifier to a list of verifiers. verifiers.Add(New StringLengthVerifier( _ GetType(Employee), "FirstName", True, 1, 30))

Behind the scenes, DevForce constructs a Verifier that can validate the Employee.FirstName property and arranges for that Verifier to be evaluated when someone tries to set the Employee.FirstName property97. That “arrangement” is the trigger.

Adding Triggers Explicitly
The predefined PropertyValueVerifiers and their corresponding attribute versions each add a property trigger to a property verifier implicitly (which is to say, “automatically”). When you create your own verifiers, you may want to add one or more triggers yourself. These you must add explicitly. It is easy to do with the EntityPropertyDescriptors generated by the Object Mapper98 as we see in this example:
hireDateVerifier.AddTrigger( EntityPropertyDescriptors.Employee.HireDate )

You can add them by string name too.
hireDateVerifier.AddTrigger(“HireDate”)

This isn‟t type safe and it assumes that the trigger property is a property of the object to be verified as is usually the case. You can specify the trigger type if you want to do so 99. C#
hireDateVerifier.AddTrigger(typeof(Employee), “HireDate”);

Visual Basic
hireDateVerifier.AddTrigger(GetType(Employee), “HireDate”)

You may go far with just this much understanding of triggers. On the other hand, you may find you need to dig deeper and then you‟ll want to know about TriggerItem and TriggerLink.

TriggerItem
The TriggerItem represents the triggering Employee.FirstName property.
97 98 99

It will also be evaluated when the program validates the Employee object (that is, during “Instance Verification”). You can extend them to include your custom properties. Or if you have to do so for reasons that will become clear below.

363 | P a g e

IdeaBlade DevForce

Validation Through Verification

The Employee.FirstName property serves two roles in our example. It is both the value that is validated by the verifier and it is the “thing” that can trigger the verifier. We have to distinguish between the two. At the moment, we are interested in the property only in its second role – in its capacity as a trigger. Imagine that the verifier didn‟t look at the first name. Imagine that it performed some other Employee validation such as checking to see if the person is old enough to be an Employee. We could still trigger this verifier every time the user touched the FirstName property. The FirstName property serves in the second role, as trigger, even though it plays no role at all in the validation. A TriggerItem is little more than a .NET Type and a string called the MemberName. The string is almost always the name of some member on that type. If TriggerItem represents a property, the MemberName is the property name. While most TriggerItems are properties, it should be clear that we can represent almost any member of a Type as a TriggerItem. We could trigger evaluation of a Verifier with a method as easily as a property. In fact, the MemberName could be an arbitrary string that is not an actual member of the type.

TriggerContext
In the course of evaluating a Verifier, the VerifierEngine calls methods on that verifier. Remember, the VerifierEngine calls these methods. You do not. These methods include: C#
public VerifierApplicability IsApplicable( Object pItemToVerify, TriggerContext pTriggerContext, VerifierContext pVerifierContext); public VerifierResult Verify( Object pItemToVerify, TriggerContext pTriggerContext, VerifierContext pVerifierContext);

Visual Basic
Public Function Verify( _ ByVal pItemToVerify As Object, _ ByVal pTriggerContext As TriggerContext, _ ByVal pVerifierContext As VerifierContext) As VerifierResult Public Function IsApplicable( _ ByVal pItemToVerify As Object, _ ByVal pTriggerContext As TriggerContext, _ ByVal pVerifierContext As VerifierContext) As VerifierApplicability

Notice that the second parameter is a TriggerContext. The TriggerContext provides the verifier with vital information about how the verifier was triggered.

364 | P a g e

IdeaBlade DevForce

Validation Through Verification

The engine does not have to be triggered to evaluate the verifier. It could evaluate an entire instance without prompting by a trigger100. The TriggerContext is null in this situation – a fact the verifier may use to establish that it was not triggered. The following table highlights the key elements of a TriggerContext: Class Member
ProposedValue

Description Typically a value entered by the user. The value is not yet committed; if the trigger is a property, the property has not yet been set to this value. The ProposedValue is meaningful only when the trigger‟s Timing is Preset. By convention, the verifier evaluates the proposed value. If the value is invalid (per the verifier), the triggering property should discard the proposed value and leave the current property value intact. The Entity.BeforeSetValue observes this convention.

Timing

One of the TriggerTiming enumerations (Preset, Postset) that indicate whether the validation occurs before the triggering property performs its task (Preset) or after it has already performed its task (PostSet). The ProposedValue is meaningful only when the Timing is Preset. The TriggerItem that inspired the VerificationEngine to evaluate the Verifier. The object that pulled the trigger. The Employee instance is the TriggerItemInstance in an Employee.FirstName trigger.

TriggerItem TriggerItemInstance

TriggerLink
We have neglected the TriggerLink to this point, conveniently confining our attention to the TriggerItem. As it happens, the TriggerItem alone is insufficient if we are to support a robust validation system. The TriggerItem tells us what kind of object triggered a Verifier. Now we have to find a way back from the object trigger to the object being verified. This is easy when the TriggerItem refers to a property of the object being verified. If the trigger is Employee.FirstName and the Verifier targets the Employee object, it is obvious that “the way back” from the property to Employee involves no effort at all: the triggering object and the verified object are one and the same. We wouldn‟t bother with such minutia unless we had grander plans – and we do. We would like to trigger evaluation of a Verifier when something happens much farther away. Let‟s change our example from Employee to Order. Suppose the user increases the quantity of an item on Order, a change that typically increases the total price of the order. Imagine that there is a verifier on the Order that constrains the total allowed amount of any order to a maximum amount, an amount calculated per a rule that factors the role of the user entering the data and the Customer‟s credit limit. This verifier sits on the Order class.

100

This is called “instance verification”.

365 | P a g e

IdeaBlade DevForce

Validation Through Verification

We could wait until we validated the entire order before evaluating this verifier. If the change broke the limit, we‟d tell the user. But it might be better to tell the user right away. It might be better if the change to the OrderDetail.Quantity property triggered the Order verifier immediately.
OrderDetail.Quantity is not a property of Order. It is one hop away, on the navigation path from OrderDetail to Order. In other words, to make this trigger work, the VerifierEngine must be able to follow the path from the triggering change in Quantity to OrderDetail and from there to Order.

Enter the TriggerLink. The TriggerLink includes both the TriggerItem and a method that can navigate from the triggering object to the object to verify, a method known as the TriggerTargetNavigator. In our order example, the navigator could be the method that implements the nested property path from OrderDetail to Order.

Verifiers and TriggerLinks
We observed earlier that specifying a StringLengthVerifier for FirstName, simultaneously specifies the Employee.FirstName as its TriggerItem. It turns out that we are actually attaching a TriggerLink to the Verifier; the Employee.FirstName is the TriggerItem contained within that link whose other half is the navigator to Employee. When we use any of the PropertyValueVerifiers, we implicitly create a verifier attached to a TriggerLink that refers to the chosen property as its TriggerItem. The DevForce syntax hides the hook-up to make creating the verifier easy. Easy things should be easy. But hard things should be possible – and a full appreciation of what is actually happening can open our eyes to more complex scenarios. Let‟s take a look at some syntax for adding a TriggerLink to a Verifier explicitly. First, the simple case: C#
TriggerLink aLink = new TriggerLink( new TriggerItem(typeof(Employee), "FirstName"), // TriggerItem null, false); // Navigation aStringLengthVerifier.AddTrigger(aLink);

Visual Basic
Dim aLink As New TriggerLink( _ New TriggerItem(GetType(Employee), "FirstName"), _ Nothing, False) aStringLengthVerifier.AddTrigger(aLink)

The TriggerItem consists of a Type and a property name, just as we expect. The navigator is null (Nothing in VB) because there is no navigation necessary from the object that triggers the verifier to the object that is verified – they are the same object. The third Boolean parameter is false because the link does not return a collection and therefore cannot return “multiple targets”. The meaning of this mysterious option will become clear shortly. We would never actually add a simple property trigger this way. There is no reason to specify the TriggerLink or even the triggering object‟s type. There is no navigator and the type of the trigger is the same as the type of the verifier. Instead we would write, in both C# and VB,
aStringLengthVerifier.AddTrigger("FirstName")

Now look at the second case involving Order and OrderDetail: C#
TriggerLink aLink = new TriggerLink(

366 | P a g e

IdeaBlade DevForce

Validation Through Verification

new TriggerItem(typeof(OrderDetail), "Quantity"), // TriggerItem "Order", false); // Navigation orderTotalPriceVerifier.AddTrigger(aLink);

Visual Basic
Dim aLink As New TriggerLink( _ New TriggerItem(GetType(OrderDetail), "Quantity"), _ "Order", False) orderTotalPriceVerifier.AddTrigger(aLink)

This time we have a navigator. The navigator is indicated by the Order property, a property of OrderDetail that returns the Order instance to verify. Apparently DevForce can convert a nested property path into a TriggerTargetNavigator.

How It Works
Here in schematic form is how Verifiers, TriggerItems, and TriggerLinks come together under the control of a VerifierEngine when the triggering object and the verified object are different. Something in the trigger property implementation tells the VerifierEngine to verify101, supplying it with the means to identify the TriggerItem. The VerifierEngine finds a TriggerLink for that TriggerItem and also the Verifier to which that TriggerLink is attached. The VerifierEngine extracts the TriggerTargetNavigator and calls it, passing the trigger object as a parameter. The trigger object is the OrderDetail in our example. The navigator returns the object to verify (the Order). The VerifierEngine confirms that the object to verify is of the correct type (i.e., it matches the Verifier.ApplicableType). The VerifierEngine executes the Verifier, passing the trigger information (a TriggerContext) as one of the parameters.

Triggering Multiple Verifiers
We said that the “OrderTotalPrice” verifier consults the customer‟s credit limit when determining if the total price of the order is valid. If the user changes the customer‟s credit limit, she could render the order valid or invalid. Not just one order either. She could change the validity of all of the customer‟s outstanding orders. We might want to draw attention to this by adding a Customer.CreditLimit trigger to the “OrderTotalPrice” verifier. Here‟s some syntax: C#
TriggerLink aLink = new TriggerLink( new TriggerItem(typeof(Customer), "CreditLimit"), // TriggerItem "Customer.Orders", // Navigation true); // true = returns multiple targets orderTotalPriceVerifier.AddTrigger(aLink);

Visual Basic
Dim aLink As New TriggerLink( _ New TriggerItem(GetType(Customer), "CreditLimit"), _ "Customer.Orders", _
101

We‟ll investigate how to engage the VerifierEngine in just a few moments.

367 | P a g e

IdeaBlade DevForce
True) orderTotalPriceVerifier.AddTrigger(aLink)

Validation Through Verification

Note that this time the third argument of the TriggerLink constructor is True. We had to add an additional argument to signal that this TriggerLink could return multiple objects to verify102. The VerifierEngine will execute the “OrderTotalPrice” verifier for each of the customer orders. If there are twenty customer orders, there will be twenty VerifierResults.

TriggerLinks and Performance
We typically don‟t worry about how long it takes to set a property. Now that we‟ve introduced triggers that can provoke a series of verifications, we should pause and reflect. The navigator in this last example invoked the Orders property of a Customer instance. That Customer may have thousands of orders, none of them in the entity cache. Calling Customer.Orders in this situation usually means a trip to the data store. The UI could stall noticeably while DevForce runs out to the server to fetch the orders. The developer must be aware of this possibility if she is going to write fancy triggers like this one. She may want to confine it to entities in the cache or only retrieve the orders that are still open 103. The Customer.Orders property can‟t be changed. Fortunately, there is an alternative.

TriggerTargetNavigator Delegate
In our previous TriggerLink examples we specified the navigator with a nested property path. We could have used a TriggerTargetNavigator delegate, defined as follows.
public delegate Object TriggerTargetNavigator(Object pInstance); Public Delegate Function TriggerTargetNavigator(ByVal pInstance As Object) _ As Object

It‟s a simple method that takes one object – the triggering object – and returns another object – the object to verify104. Here is the same TriggerLink, rewritten to use a TriggerTargetNavigator delegate method called “aCustomerOrdersNavigator”. C#
TriggerLink aLink = new TriggerLink( new TriggerItem(typeof(Customer), "CreditLimit"), // TriggerItem aCustomerOrdersNavigator, // Navigation delegate true); // true = returns multiple targets orderTotalPriceVerifier.AddTrigger(aLink);

Visual Basic
Dim aLink As New TriggerLink( _ New TriggerItem(GetType(Customer), "CreditLimit"), _ aCustomerOrdersNavigator, _ True)

102

If we said False, the link would return a single target object – a collection of Order. The “OrderTotalPrice” verifier applies to a single Order instance, not a collection. There is a type mismatch between the verifier and the (collection) object returned from the TriggerLink; the VerifierEngine will raise a VerifierException indicating that the verifier‟s execution failed. There is no point in verifying closed orders. Remember that this object can be a collection of objects. The boolean TriggerLink.ReturnsMultipleTargets property tells the VerifierEngine whether to verify the items in the collection individually (true) or as a single object (false).

103 104

368 | P a g e

IdeaBlade DevForce
orderTotalPriceVerifier.AddTrigger(aLink)

Validation Through Verification

This method could use DevForce persistence operations to do the navigation but it doesn‟t have to. It can have any implementation that returns objects that match the verifier‟s target object type, the value of Verifier.ApplicableType.  Do not use an asynchronous delegate. Validation is not workflow. Validation is an inherently synchronous operation and the VerifierEngine is not thread safe. The navigator must return an object to verify; the application must pause until that object becomes available.

PropertyDescriptor Syntax
We have shown the TriggerItem in its “native form” as a .NET Type and a member name. The PropertyDescriptor alternative may be easier to enter, easier to read, and is certainly more type-safe because the developer does not have to code the member name as a string. Here‟s how to add the simple property trigger in a single statement using the PropertyDescriptor notation in either C# or Visual Basic:
aStringLengthVerifier.AddTrigger(Employee.PathFor(e=>e.FirstName))

Here‟s how to add the TriggerLink with PropertyDescriptor notation. C#
TriggerItem item = new TriggerItem(typeof(Customer), Customer.CreditLimtEntityProperty.Name); orderTotalPriceVerifier.AddTrigger( new TriggerLink(item, // TriggerItem Customer.PathFor(c=>c.Orders), // Navigation True)); // true = returns multiple targets Dim item As New TriggerItem(typeof(Customer), Customer.CreditLimtEntityProperty.Name) orderTotalPriceVerifier.AddTrigger( new TriggerLink(item, „ TriggerItem Customer.PathFor(c=>c.Orders), „ Navigation True)) „ true = returns multiple targets

VB

Non-Property Triggers
We tend to discuss triggers as if they were always property triggers. They usually are. But they don‟t have to be. It takes a TriggerItem to trigger verification. The TriggerItem consists of a Type and a String called the MemberName. The MemberName could be any string. Usually it is a property name but it need not be. It could be a method name. It could be a string with no intrinsic meaning at all. The VerifierEngine uses the “type-and-string” to find verifiers to evaluate. It is as if the engine had a dictionary of TriggerItems, each leading to a TriggerLink and each link leading to a Verifier105. The “reality” of the MemberName is irrelevant from this perspective. Any block of code can trigger verification. All it has to do is call a VerifierEngine in a trigger-like way as discussed in the section “Invoking Verification”.
105

Actually, a TriggerItem could lead to multiple TriggerLinks and each of those links could be attached to multiple Verifiers. A single TriggerItem can launch an avalanche of verifications.

369 | P a g e

IdeaBlade DevForce

Validation Through Verification

The DevForce Object Mapper generates property setter code that calls a VerifierEngine in a trigger-like way. You do the same when you write your own custom settable properties. You could put the same call logic inside a method. For example, you might trigger Order verification inside methods that add or remove OrderDetail items so that you can immediately test the effect of adds and deletes on the total price of an order. TriggerTiming (Preset, Postset) is a convention that you should follow but can adapt to your purpose. Your AddOrderDetail method could trigger verification in a Preset manner before adding the new item106. If validation fails, the method could discard the item before it did any harm.

VerifierEngine
The VerifierEngine is the primary entry point for verification services. An application may have any number of VerifierEngines although most will only need one. Each VerifierEngine contains a list of verifiers and a set of methods that allow collections of these verifiers to be evaluated sequentially against an instance of a .NET class. The verified object could be a DevForce business object but it doesn‟t have to be. The object can be of any concrete type. Each verifier produces a VerifierResult. The engine accumulates these results in a VerifierResultCollection as it proceeds and returns the entire collection as its own result.

Adding Verifiers to a VerifierEngine
Verifiers can be added to a VerifierEngine in two ways:

The engine can discover them automatically by inspecting the .NET types for verifier attributes. The developer can add them programmatically. The application can combine these methods. We got a taste of verifier discovery in the “Getting Started” started section. We‟ll cover it in more depth shortly. Programmatic management of an engine‟s verifiers is straightforward via the AddVerifier and RemoveVerifier methods. Verifiers can be added or removed from a VerifierEngine at any time. The engine raises a VerifiersChanged event when verifiers are added or removed. available on the verifier engine and will inform any subscriber of the addition or removal of any verifier 107.

Verifier Discovery
The VerifierEngine always discovers verifiers in the types it is asked to verify 108. When a VerifierEngine attempts to verify an instance of a type it has not seen before, it probes the type reflectively, looking for verifiers. The probing strategy is as follows. Start with the most senior base class in the type‟s inheritance chain.
106 107 108

You must supply a ProposedValue. It can be any kind of object such as the item to be added. It could be null. The event is also raised when triggers are added or removed from a verifier that has been registered in the engine. Automatic discovery is not always a good thing, and developers can disable an engine‟s automatic discovery. An engine with automatic discovery disabled can still perform discovery when asked to do so.

370 | P a g e

IdeaBlade DevForce

Validation Through Verification

Look for instances of the VerifierAttribute109 class on members of that base class. These define the “attributed verifiers”. Look for a static method decorated with the VerifierProviderAttribute110; Such a method must take a single parameter of type object – this is the “VerifierProviderContext” – and it must return an IEnumerable(Of Verifier). The engine calls the VerifierProvider and adds the Verifier instances returned by that method to its list of verifiers for the base type. Find the next class in the type‟s inheritance chain and return to step #2. Stop when have descended to the type that initiated the discovery process. We have seen the attribute verifiers earlier. A VerifierProvider might look like this: C#
#region Verification #region GetVerifiers Method /// <summary>Get Verifiers.</summary> /// <param name="pVerifierProviderContext">Context in which these Verifiers are retrieved.</param> /// <returns>The verifiers.</returns> [VerifierProvider] public static IEnumerable<Verifier> GetVerifiers(Object pVerifierProviderContext) { List<Verifier> verifiers = new List<Verifier>(); verifiers.Add(GetHireDateRangeVerifier()); verifiers.Add(new BirthDateRangeVerifier()); verifiers.Add(GetBornBeforeHiredVerifier()); verifiers.Add(GetPhoneNumberVerifier(Employee.HomePhoneEntityProperty)); return verifiers; } #endregion #region Hire Date Verifier /// <summary>Get a GetHireDateRangeVerifier.</summary> /// <remarks> /// Demonstrates building a highly focused verifier /// by encapsulation a standard verifier /// and its configuration. /// </remarks> private static Verifier GetHireDateRangeVerifier() { Verifier v = new DateTimeRangeVerifier( typeof(Employee), // Type of the object being verified Employee.HireDateEntityProperty.Name, // Property trigger false, // Non-null value is not required MinHireDate, true, // starting min date (inclusive) MaxHireDate, false); // ending max date (exclusive) return v;

109

DevForce provides a number of common verifiers in attribute form all of which descend from VerifierAttribute. The developer can add custom VerifierAttribute subclasses just as he can add custom Verifiers. Actually, there can be more than one such method in the class and the VerifierEngine will call each one.

110

371 | P a g e

IdeaBlade DevForce
} private static DateTime MinHireDate { get { return new DateTime(1990, 1, 1); } } private static DateTime MaxHireDate { get { return DateTime.Today.AddMonths(1); } } #endregion #region BirthDateRangeVerifier inner class

Validation Through Verification

/// <summary>Get the minimum BirthDate allowed.</summary> private static DateTime MinBirthDate { get { return new DateTime(1900, 1, 1); } } /// <summary>BirthDate Range Verifier</summary> /// <remarks> /// Illustrates changing the error messaging for a particular property. /// Have to subclass to take control of the messaging. /// Here the message is statically known so we override /// <see cref="M:Description"/>; /// if it were dynamic or if /// <see cref="T:DateTimeRangeVerifier"/> constructed the /// message dynamically, we would have overridden /// <see cref="M:VerifyValue"/> and manipulated the /// message while creating the <see cref="T:VerifierResult"/>. /// </remarks> private class BirthDateRangeVerifier : DateTimeRangeVerifier { /// <summary>Default Ctor,</summary> /// <remarks> /// BirthDate is not required, /// must be on or after global min date (<see cref="M:MinBirthDate"/>), /// and before today. /// </remarks> public BirthDateRangeVerifier() : base( typeof(Employee), // Type of the object being verified Employee.BirthDateEntityProperty.Name, // Property trigger false,// Non-null value is not required MinBirthDate, true, // starting min date (inclusive) DateTime.Today, false) { } // ending max date (exclusive) public override string Description { // ToDo: Localize get { return "Must be born after " + MinBirthDate.Year.ToString() + "; No time travellers allowed!"; } } } #endregion #region Born Before Hired Verifier /// <summary>Get a BornBeforeHiredVerifier.</summary> /// <remarks> /// Demonstrates comparing two property values

372 | P a g e

IdeaBlade DevForce

Validation Through Verification

/// by creating an instance of a /// <see cref="T:DelegateVerifier{TVerifiedObject}"/>. /// </remarks> private static Verifier GetBornBeforeHiredVerifier() { // ToDo: localize description string description = "Must be born before hired."; DelegateVerifier<Employee> v = new DelegateVerifier<Employee>(description, BornBeforeHiredCondition); v.AddTriggers(Employee.BirthDateEntityProperty.Name, Employee.HireDateEntityProperty.Name); v.ExecutionModes = VerifierExecutionModes.InstanceAndOnPostsetTriggers; return v; } /// <summary> /// The <see cref="T:VerifierDelegate{TVerifiedObject}"/> /// for the <see cref="M:GetBornBeforeHiredVerifier"/>. /// </summary> private static VerifierResult BornBeforeHiredCondition( Employee pEmp, TriggerContext pTriggerContext, VerifierContext pVerifierContext) { if (pTriggerContext != null && // We are not checking the proposed value because don't expect to call it preset pTriggerContext.Timing == TriggerTiming.Preset) { throw new VerifierException("BornBeforeHired verifier not implemented for Preset"); } return new VerifierResult(pEmp.BirthDate < pEmp.HireDate); } #endregion #region Phone Number Verifier /// <summary>Get a GetPhoneNumberVerifier.</summary> /// <remarks> /// Encapsulates a standard RegexVerifier, subclassed so the description can be customized. /// </remarks> private static Verifier GetPhoneNumberVerifier(EntityProperty pPhoneEntityProperty) { return new PhoneNumberVerifier( pPhoneEntityProperty.EntityType, // Type of object being verified pPhoneEntityProperty.Name, // Trigger false, // Non-null value is not required NamedRegexPattern.USPhone); // Regex pattern to use } private class PhoneNumberVerifier : RegexVerifier { public PhoneNumberVerifier(Type pApplicableType, string pPropertyName, bool IsRequired, NamedRegexPattern pattern) : base( pApplicableType, pPropertyName, IsRequired, pattern ) { }

373 | P a g e

IdeaBlade DevForce

Validation Through Verification

public override string Description { get { return base.Description + " including area code [e.g., (206)555-1212, 206-555-1212, or 206.555.1212]."; } } } #endregion #endregion

VB

VerifierProviderContext
Observe that a VerifierProvider method has an object parameter called the “VerifierProviderContext”. This is an arbitrary object, open to the developer‟s imagination. The VerifierEngine will pass it along to each provider. The engine acquires this context object in one of two ways: From the VerifierEngine.DefaultVerifierProviderContext which the developer must have initialized before the engine starts its discovery process. As the second argument to VerifierEngine.DiscoverVerifiers(Type, Object). This is a method that forces verifier discovery for the given type. The VerifierProviderContext object could be anything. It could be a pre-calculated list of verifiers for the type. It could include the VerifierEngine itself so that the VerifierProvider can inspect and manipulate the other verifiers for this type. The application must set the VerifierEngine‟s DefaultVerifierProviderContext or call its DiscoverVerifiers method early. The engine starts auto discovery as soon as it receives a request to verify an instance of a type. That discovery could fail or populate the engine with the wrong verifiers if the developer doesn‟t make these calls first.

Recommended Verifier Loading Approach
We recommend that most applications rely on automatic discover to build up a VerifierEngine‟s list of verifiers. It is ok to add or remove verifiers from a VerifierEngine programmatically outside of the class being verified but you should have a good reason for the extra and unexpected complexity. Some business requirements call for configurable validation rules. Verifiers can be represented in metadata, saved to storage, retrieved when the application starts, and plugged in to a VerifierEngine.

Configuring New VerifierEngines Consistently

374 | P a g e

IdeaBlade DevForce

Validation Through Verification

While most applications will have only one VerifierEngine, there are good use cases for having two or more. Wherever there are multiple engines there arises the need to ensure that they are all configured consistently and appropriately. We don‟t want a rogue programmer blithely instantiating new engines that lack a DefaultVerifierProviderContext or are missing some other critical setting. The application can attach a handler to the static event, VerifierEngineCreated, on the VerifierEngine class. The event is raised whenever there is a newly created engine. The new engine is passed in the VerifierEngineCreatedEventArgs so that the handler can configure it.

Invoking Verification
Verifiers do not execute themselves nor can they be executed on their own. They must belong to a (single) VerifierEngine and rely on that engine to make them do their validation work.

A VerifierEngine doesn‟t verify on its own either. Something has to tell it to verify. DevForce shouldn‟t perform any operation unless it is asked to do so. Verification is a potentially costly operation. Perhaps as important, DevForce would not know what to do when it was done verifying. Only the application developer can know when to verify and what to do with the results. DevForce does provide an easy way to automate trigger verification of the properties of business objects. The developer simply launches the Object Mapper and turns Verification on 111. The Object Mapper generates “setter” code to call the VerifierEngine at the appropriate time. It is still up to the developer to invoke verification at other key moments in the application such as: Verification of entities just before they are saved. Trigger verification of custom, settable properties of business objects. Verification upon business object fetch or merge. Trigger verification of non-business objects. Fortunately, there are .NET events for all of the key business object moments and trigger verification of nonbusiness objects looks just like trigger verification of business objects. In every case, the developer calls one of the VerifierEngine.Execute overloads. The public Execute methods available at this time fall into three “Execution Modes”: Instance Verification Preset Trigger Verification Postset Trigger Verification We‟ll examine each mode in this following segments. We‟ll learn how calling the VerifierEngine‟s Execute method determines whether it will perform instance, preset, or postset verification. Before we do, it is important to remember that we do not call Verifiers; the VerifierEngine does that. When we tell it to execute in one of the three modes, it will iterate over its internal list of registered verifiers, evaluating each verifier that is enabled for the current mode. A Verifier will only be evaluated if its Verifier.ExecutionModes matches the current mode! For example, if a verifier‟s ExecutionModes = VerifierExecutionModes.Disabled, the verifier won‟t be evaluated at all, no matter how we call the VerifierEngine.

111

We saw how to do this in the “Getting Started” section.

375 | P a g e

IdeaBlade DevForce
Keep this in mind as you review the scenarios below.

Validation Through Verification

Instance Verification
The following are the VerifierEngine.Execute overloads for instance verification: Execute Overload Instance Verification
1 2 3 Execute(object pInstance) Execute(object pInstance, VerifierContext pVerifierContext) Execute(object pInstance, IEnumerable<Verifier> pVerifiers, VerifierContext pVerifierContext)

Description

Validate an instance within the default VerifierContext Validate an instance within a particular
VerifierContext

Validate an instance with just the given list of 112 Verifiers . Validate within a particular VerifierContext.

The “Instance Verification” Execute overloads validate an entire instance. The VerifierEngine finds the Verifiers for the instance type keeps only those with the Instance flag set in their Verifier.ExecutionModes sorts them in execution order113 and evaluates them sequentially.

VerifierContext
Every verifier receives a VerifierContext object during its evaluation. The simplest Execute, which accepts only the object to verify, passes along a VerifierContext constructed by the VerifierEngine. The other signatures take a custom VerifierContext argument which the engine modifies before handing to the verifiers. One of the signatures lets you specify which verifiers the VerifierEngine should evaluate. These verifiers must be registered with the VerifierEngine and their Verifier.ApplicableType must match the type of the verified object.

When and Where to Verify an Instance
The business requirements dictate when and where to verify an instance. Many applications provide the ability to validate an entity at any time and then ensure that every entity passes validation before it can be saved. Accordingly, this author recommends: Prepare business objects for instance verification     
112 113

Generate a BaseEntity in the Object Mapper Make all business objects inherit from this BaseEntity Write a VerifyInstance method in that BaseEntity

Verify instances in your handler of the EntityManager.Saving event Make sure you have such a handler on every EntityManager Iterate through the entities to be saved, calling VerifyInstance on each one

All of the verifiers must have been registered with this engine or else the Execute method returns an exception. Verifiers are sorted by Verifier.SortValue ; ties are broken by the order in which they were loaded into the engine (Verifier.InitializationOrder).

376 | P a g e

IdeaBlade DevForce
   Accumulate the VerifierResults from each call Cancel the save if there are any VerifierResults. Report these results to the user.

Validation Through Verification

These are basic techniques taught in the DevForce tutorials, demonstrated in DevForce movies, and incorporated (albeit in enriched form) in DevForce reference applications such as Cabana. Here is a simplified example of a VerifyInstance method. C#
/// <summary>Validate object for all instance Verifiers.</summary> protected virtual VerifierResultCollection VerifyInstance() { return this.VerifierEngine.Execute(this); }

Visual Basic
''' <summary>Validate object for all instance Verifiers.</summary> Protected Overridable Function VerifyInstance() As VerifierResultCollection Return Me.VerifierEngine.Execute(Me) End Function

Observe that each instance has access to a VerifierEngine; this is the VerifierEngine that belongs to its EntityManager.

Trigger Verification: Preset and Postset
Should we validate a value before we set the property or after we set the property? There is no universally correct answer to this question.

Preset Triggers
Some bad values should never enter the object. If the object property concerned the dosage level of a drug, we‟d want to prevent entry of an invalid value. Ten thousand milligrams of something could be fatal. We have to block that at the moment of data entry. We don‟t want the user to be able to move until the problem is corrected. We certainly don‟t want that dosage to appear in the business object ever – not even in cache. This is the right place for preset trigger verification. In preset verification, the VerifierEngine receives a “proposed value” from the caller. The engine creates a TriggerContext with TriggerContext.Timing set to TriggerTiming.Preset. It embeds the proposed value in the TriggerContext.ProposedValue. Then it makes calls on the verifier(s) linked to the trigger, passing in this TriggerContext so that the verifier (a) knows how it was triggered and (b) the value it should test. By convention, the code that asks for preset trigger verification should examine the VerifierResultCollection returned from the engine before doing anything more with the proposed value. If the results collection contains an errant result – if VerifierResultCollection.AreOk is false – the code should discard the proposed value.

377 | P a g e

IdeaBlade DevForce

Validation Through Verification

The following are the VerifierEngine.Execute overloads for “preset” trigger verification: Preset Trigger Execute Signatures
1 Execute(object pTriggerItemInstance, String pMemberName, Object pProposedValue)

Description Perform “preset” validation of verifiers on all objects that are linked to the pMemberName property of the pTriggerItemInstance. The property will be set to the pProposedValue unless the validation fails. Verifiers receive the default VerifierContext. We say that the property caused a “preset trigger validation”

2

Execute(object pTriggerItemInstance, String pMemberName, Object pProposedValue, VerifierContext pVerifierContext)

Perform “preset” validation of verifiers on all objects that are linked to the pMemberName property of the pTriggerItemInstance. Verifiers receive the given VerifierContext. Perform “preset” validation of verifiers on all objects that are linked to the given PropertyDescriptor which translates to a property of the pTriggerItemInstance. Verifiers receive the default VerifierContext. Perform “preset” validation of verifiers on all objects that are linked to the given PropertyDescriptor which translates to a property of the pTriggerItemInstance. Verifiers receive the given VerifierContext. Perform “preset” validation of verifiers on all objects that are linked to the given TriggerItem. Verifiers receive the default VerifierContext. Perform “preset” validation of verifiers on all objects that are linked to the given TriggerItem. Verifiers receive the given VerifierContext.

3

Execute(object pTriggerItemInstance, PropertyDescriptor pDescriptor, Object pProposedValue)

4

Execute(object pTriggerItemInstance, PropertyDescriptor pDescriptor, Object pProposedValue, VerifierContext pVerifierContext)

5

Execute(object pTriggerItemInstance, TriggerItem pTriggerItem, Object pProposedValue)

6

Execute(object pTriggerItemInstance, TriggerItem pTriggerItem, Object pProposedValue, VerifierContext pVerifierContext)

Setting a Preset Trigger
The natural place to trigger a preset validation is inside the setter of the property, before writing the incoming value into the object. The code should provide the incoming value as the “ProposedValue” parameter. The code may include a VerifierContext if it will help the triggered Verifier do its job but the context is optional and may be null. The VerifierEngine provides the Verifier with a TriggerContext object that (a) alerts the Verifier to the fact that it was triggered and (b) provides the contextual information it needs to do its evaluation, including the proposed value in this preset case. See the “TriggerContext” section for more information. If the verification fails – if any preset Verifier produces an errant VerifierResult – the property must do something. The .NET framework development guidelines suggest that it should throw an exception. There is a VerifierResultException114 for this purpose.

114

Its constructor accepts a VerifierResultsCollection parameter that handlers can interpret and present intelligently.

378 | P a g e

IdeaBlade DevForce
A VerifierResultException should not terminate the application.

Validation Through Verification

DevForce handles the exception gracefully when it occurs during data binding; see the “Verification and WinForms User Interfaces” section. The developer must handle a VerifierResultException thrown outside of data binding.

Entity.BeforeSetValue
The Object Mapper generates an Entity.BeforeSetValue method that adheres to this recommendation precisely115. The method is virtual; developers can override it in a base entity class if they want different behavior or if they want to augment it with other behavior such as error logging.

Postset Triggers
“Life and death” properties are relatively rare. It is usually ok if the property value is invalid while the user is working with the object. We want the user to know the value is invalid. We want to block every attempt to save invalid data. But we can tolerate bad values for a while. For example, the employee‟s home city may be a required value. We may not be able to save the employee record until we have a complete and valid home address. We want the application to tell us about the omission in time to correct it. On the other hand, it isn‟t going to harm anything if it stays blank while the user is entering new employee information. If the user mistakenly enters the wrong city, she should be able to clear it. She may not know the name of the correct city; it is better to leave the city blank than to leave the incorrect city in place. This is fine as long as we prevent the user from saving the address. Summarizing the requirement: Permit entry of an invalid value but advise the user of that fact. Prevent saving of an object with an invalid value and tell the user about that. The rule – manifested in the Verifier - is the same in both cases. How we validate and what we do with the result depends upon the context. We covered the second scenario - block the save – when we discussed “instance validation” above. We want “postset” triggered validation to handle the first scenario. “Postset” means that the property has already been set with the incoming, invalid value from the user by the time we validate. There is no “proposed value” to worry about. We still want to validate the (now current) property value and tell the user if there is a problem.

115

The appendix discusses the implementation of BeforeSetValue in detail.

379 | P a g e

IdeaBlade DevForce

Validation Through Verification

The following are the VerifierEngine.Execute overloads for “postset” trigger verification: Postset Trigger Execute Signatures
1 Execute(object pTriggerItemInstance, String pMemberName)

Description Perform “postset” validation of verifiers on all objects that are linked to the pMemberName property of the pTriggerItemInstance. Verifiers receive the default VerifierContext. We say that the property caused a “postset trigger validation”

2

Execute(object pTriggerItemInstance, String pMemberName, VerifierContext pVerifierContext)

Perform “postset” validation of verifiers on all objects that are linked to the pMemberName property of the pTriggerItemInstance. Verifiers receive the given VerifierContext. Perform “postset” validation of verifiers on all objects that are linked to the pMemberName property of the pTriggerItemInstance. Verifiers receive the given VerifierContext. Perform “postset” validation of verifiers on all objects that are linked to the given TriggerItem. Verifiers receive the given VerifierContext.

3

Execute(object pTriggerItemInstance, PropertyDescriptor pPropertyDescriptor, VerifierContext pVerifierContext)

4

Execute(object pTriggerItemInstance, TriggerItem pTriggerItem, VerifierContext pVerifierContext)

As always we must tell the VerifierEngine to perform verification. The VerifierEngine will give the triggered Verifier a TriggerContext just as it did for the preset trigger but this time there will be no proposed value; the verifier may have to fish the value out of the object. That shouldn‟t be hard. The Verifier typically knows the property it verifies and this property is usually the same property that triggered verification. A “First Name” StringLengthVerifier that is triggered by input of first name text will know how to examine the FirstName property of the Employee instance it verifies. DevForce removes the guess work if the Verifier inherits from PropertyValueVerifier (as StringLengthVerifier does). Every subclass of PropertyValueVerifier has a virtual VerifyValue method that receives both the instance to verify and the value to verify. It is slightly trickier if the instance triggering the verifier is different from the object instance verified. We encountered such a case when we considered a “TotalPriceVerifier” on Order that is triggered by a change to the price of one of its OrderDetails. Fortunately, the Order‟s “TotalPriceVerifier” can use the TriggerContext.TriggerItem.MemberName (“UnitPrice”) to dig the changed price value out of the TriggerContext.TriggerItemInstance (the OrderDetail instance). Relatively few verifiers involve such circuitous triggering. The vast majority of verifiers are PropertyValueVerifiers whose triggering and verified instances are the same object. Which leaves us with the small problem of invoking the VerifierEngine at the right time. As this is a postset trigger, we should call the engine immediately after the line that pushes the incoming value into the trigger object.

Entity.AfterSetValue

380 | P a g e

IdeaBlade DevForce

Validation Through Verification

That is what the Object Mapper does when it inscribes an Entity.AfterSetValue method into the generated property code 116. What happens if the verification fails? We invoked the verifier for a reason, presumably to alert the user to a problem. The AfterSetValue throws a VerifierResultException just as the BeforeSetValue does. DevForce and .NET handle this just fine if the exception occurs within data binding. The developer must handle the exception if it occurs anywhere else.
AfterSetValue is virtual so developers can override it in a base entity class if they want different behavior. We‟ll

consider an alternative implementation in the “Verification and WinForms User Interfaces” section. Remember that you can delay telling the user about invalid input and rely upon instance verification to catch it just before save. You won‟t need postset triggers if you go this route.

The Role of the Object Mapper
As we just noted, the Object Mapper includes the Entity.BeforeSet and the Entity.AfterSet methods in the code it generates for properties unless you specify otherwise. It also generates an Args parameter for those methods that specifies whether verification should be invoked preset and postset. By default, it is invoked in both situations.

Writing Verified Custom Business Object Properties
Developers often write custom business object properties. Such properties are usually ReadOnly, which is to say, they have a getter but no setter. Trigger validation is a non-issue if there is no setter. When the developer needs to write a settable property, her code probably should parallel the code generated by the Object Mapper.

Monitor Execution with the VerifierBatchInterceptor
Some applications need to monitor the progress of a VerifierEngine‟s execution and intervene at certain points. The VerifierEngine.BatchInterceptor is the way to do it. The engine calls the interceptor after evaluating each Verifier giving all of the visibility and opportunity it needs. An interceptor is a method that conforms to the VerifierBatchInterceptor delegate signature: C#
public delegate VerifierOnErrorMode VerifierBatchInterceptor( Object pInstance, TriggerContext pTriggerContext, VerifierContext pVerifierContext);

Visual Basic
Public Delegate Function VerifierBatchInterceptor( _ ByVal pInstance As Object, _ ByVal pTriggerContext As TriggerContext, _ ByVal pVerifierContext As VerifierContext) As VerifierOnErrorMode

Because the interceptor‟s parameters are the same as the parameters of the Verifier methods, IsApplicable and Verify(), it has the same visibility into the verification process as they do.
116

The appendix discusses the implementation of AfterSetValue in detail.

381 | P a g e

IdeaBlade DevForce
The interceptor can

Validation Through Verification

see the last verifier evaluated by looking at the VerifierContext.Verifier. review and edit the accumulating VerifierResultCollection by looking at the VerifierContext.VerifierResults. terminate the current batch at any time by returning VerifierOnErrorMode.Stop. post-process the VerifierResults when the batch is done – because the engine will call it one last time with the VerifierContext.EndOfBatch flag set true. The following example shows how one could use an interceptor to curb run-away validations. In this case, it terminates the batch on the third error: C#
… VerifierEngine engine = new VerifierEngine(); engine.BatchInterceptor = MyBatchInterceptor; … private VerifierOnErrorMode MyBatchInterceptor( Object pInstance, TriggerContext pTriggerContext, VerifierContext pVerifierContext) { if ( pVerifierContext.VerifierResults.Errors.Count > 2 ) { pVerifierContext.VerifierResults.Add( new VerifierResult(false,"More than 2 errors encountered")); return VerifierOnErrorMode.Stop; } else { return VerifierOnErrorMode.Continue; } }

Visual Basic
… Dim engine As New VerifierEngine() engine.BatchInterceptor = AddressOf MyBatchInterceptor … Private Function MyBatchInterceptor( _ ByVal pInstance As Object, _ ByVal pTriggerContext As TriggerContext, _ ByVal pVerifierContext As VerifierContext) As VerifierOnErrorMode If pVerifierContext.VerifierResults.Errors.Count > 2 Then pVerifierContext.VerifierResults.Add( _ New VerifierResult(False,"More than 2 errors encountered")) Return VerifierOnErrorMode.Stop Else Return VerifierOnErrorMode.Continue End If End Function

Verification and WinForms User Interfaces
Now that the application is detecting invalid data and throwing exceptions, we had better think about how we want to handle those exceptions and tell the user what is going on.

UI Lockup
The UI is going to lock up the moment the user enters an invalid value into a verified UI control. That is any data entry control: TextBox, DataPicker, ComboBox, etc. The user will not be able to leave that control until she enters a value that passes validation – not even to close the form.

382 | P a g e

IdeaBlade DevForce

Validation Through Verification

In this illustration, the user cleared the “Last Name”. The last name is required. The form displays an error bullet and prevents the user from moving out of the textbox.

How does the user recover? If this were a grid, she could press the [Esc] key; it is “standard” for grid controls to restore the previous value when the user presses “escape.” How many users know that? In any case, this TextBox is not in a grid and pressing [Esc] does nothing but ring an annoying bell. The user can press the standard key chord for “undo”: Ctrl+Z. How many users know that? No, the most users will just keep entering new values until they find one that lets them out of the field. Needless to say, a UI should apply the “lock up” enforcement technique sparingly. In the author‟s opinion, it makes sense only for a value the user must know and is sure to know a value that must be correct immediately and at all times. Dosage of a dangerous prescription drug would fit this bill. Few other properties qualify.

Unlock the UI with AutoValidate
Recall that the DevForce Entity.BeforeSetValue and Entity.AfterSetValue methods raise a VerifierResultException when the property fails validation. This exception bubbles up and out of the property setter.117 Data binding traps the exception118 and responds by locking up the form. Fortunately, WinForms .NET 2.0 makes it easy to change this response. The key is the System.Windows.Forms.UserControl.AutoValidate property which takes one of the System.Windows.Forms.AutoValidate enumerations. AutoValidate
Inherit

Description Do what the parent UserControl does. The parent is the UserControl that contains this UserControl. This is the default for new UserControl instances. If there is no parent, the value is the default, EnablePreventFocusChange.

EnablePreventFocusChange EnableAllowFocusChange Disable

Prevents the user from leaving the control until the value passes validation. Validate but permit the user to leave the control if validation fails. Does not validate. Generally not a good choice.
119

Inherit is the default value for all new UserControls . Inherit means that the UserControl is governed by the AutoValidate setting of its parent UserControls, the UserControl that contains it.

117 118

Thanks to the System.Diagnostics.DebuggerNonUserCodeAttribute that decorates the setter. During the data binding Validate event raised when the user attempts to leave the TextBox.

383 | P a g e

IdeaBlade DevForce

Validation Through Verification

The outer UserControl, typically a Form, doesn‟t have a parent so it is governed by the EnablePreventFocusChange setting. If we never change the AutoValidate property on any UserControl, our application is governed by the setting in the Form which, as we have seen, is EnablePreventFocusChange, the setting that locks up the form. All UserControls within the Form are inheriting this behavior. If we change the Form‟s AutoValidate property to EnableAllowFocusChange, the widgets on the Form will no longer lock up when the setter throws an exception. Neither will widgets on the contained UserControls because they inherit the parent Form‟s setting. So the quick answer to UI lockup: Change the Form‟s AutoValidate property to EnableAllowFocusChange C#
this.AutoValidate = System.Windows.Forms.AutoValidate.EnableAllowFocusChange; // Can move

Visual Basic
me.AutoValidate = _ System.Windows.Forms.AutoValidate.EnableAllowFocusChange ' Can move

Improving the User‟s Experience
EnableAllowFocusChange and Preset Triggers
AutoValidate.EnableAllowFocusChange works great for property verifiers governed by preset triggers.

The user can move out of the TextBox. Yet she can still see the error bullet protesting the lack of a “last name”.

The TextBox remains cleared so we can see that there is a problem – or rather that there was a problem, that our intent to clear the name was invalid. The LastName property itself was never actually changed. A preset trigger prevents the property setter from updating the object. At the moment there is a discrepancy between the business object property value and the corresponding widget control display property on screen 120. We can see reveal the discrepancy and cure it by scrolling off of the “Nancy” employee and then returning to her. The TextBox refreshes with her current LastName property value which remains “Davolio”.

119

UserControl is the base class for developer designed screens. System.Windows.Form inherits from UserControl. Individual “UI widgets” such as TextBox do not inherit from UserControl.

120

We could set the DevForce BindingDescriptor.CancelEditOnError for the binding to LastName to true; this would immediately restore the TextBox‟s display of the original value. The author dislikes that choice because it obscures what the user was trying to do by replacing the user‟s data entry. She sees a warning about a problem that is no longer the problem.

384 | P a g e

IdeaBlade DevForce EnableAllowFocusChange and Postset Triggers

Validation Through Verification

The behavior is different for verifiers evaluated in response to postset triggers. If we had a LastNameRequiredVerifier and set its ExecutionModes to InstanceAndOnPostsetTriggers, the LastName property value would be empty, just as it appears in the TextBox. A postset trigger causes validation after the property has been set with the “proposed value.” We can confirm this by scrolling off of the “Nancy” employee and then returning to her. The TextBox remains blank. The current LastName property value is empty. However, we are no longer aware of the latent validation error. Our application does not validate the Employee upon display … and that might be a user experience problem121. At least it is not a data integrity problem – or doesn‟t have to be. We must assume that the application follows our advice and ensures that every entity must survive “instance verification” before it can be saved. We further assume that the application has some mechanism to display errant entities and their problems. Perhaps a simple MessageBox will do. This Employee will not survive validation, will not be saved, and the user will be told why.

Questionable User Experience
This approach may be viable if little time can pass between data entry and instance verification. Some applications attempt a save whenever the user moves off the current screen. The user will never lose sight of the LastName error bullet and the save effort will reveal all latent problems with this employee. Many applications delay save and allow the user to move around among entities with pending changes. That‟s how our tutorial works. Users can make a change to “Nancy”, scroll to “Andrew” and make changes to him, then scroll back to “Nancy” to continue her updates. In this kind of workflow, the user may not remember that there is a problem with the “Nancy” object for minutes or hours. When the application finally tells the user about this problem, the mental context is long gone and the application will be perceived to be “unfriendly”. There is another, potentially greater risk. The user may make a critical business decision base upon what is visible on the screen. That data could be in error. The user won‟t know it if she scrolled off and then back on to the record. If this risk is serious, the application must behave differently whenever the UI displays a new object – a new Employee in our example.

Instance Verification Upon Display
One approach would be to perform instance verification whenever the currently displayed object is changed.

121

We could write code to perform “instance validation” whenever the Employee changed. We could capture the VerifierResults and display them as well as light up bullets next to each widget. The code is not hard to write but it‟s not utterly trivial either. We‟ll describe an approach that achieves something of that effect using a different technique.

385 | P a g e

IdeaBlade DevForce

DevForce Silverlight Apps

DevForce Silverlight Apps
Features described in the section are included with the DevForce Silverlight product.

DevForce Silverlight Apps
Overview - What is DevForce Silverlight? Creating a DevForce Silverlight Application Silverlight Deployment Steps Questions and Answers Troubleshooting

Overview - What is DevForce Silverlight?
DevForce Silverlight allows you to deliver line of business applications in the browser with the kind of responsiveness users expect from a desktop application. Developed for Microsoft Silverlight, the browser plug-in which powers rich application experiences, it allows you to leverage your existing DevForce experience with new tools and techniques to build serious applications. A few things to note about Silverlight, and thus about DevForce Silverlight:  Silverlight is inherently n-tier. The client application executes in a sandbox on the browser, and must communicate with a service to retrieve and save data. The DevForce Silverlight Business Object Server (BOS) provides that service, and allows you quickly to have a Silverlight application retrieving and saving to a database, using the domain model and business objects you're already familiar with. Silverlight is inherently asynchronous. To avoid blocking the browser, Silverlight requires that all service communications be performed asynchronously. This can be a bit challenging at first, but DevForce Silverlight provides an asynchronous API very similar to the standard synchronous API, plus additional features to make asynchronous programming as easy as possible.

In DevForce Silverlight, you have the EntityManager to hold your client-side entity cache and communicate with the BOS, just like you would in a standard DevForce application. The Domain Model is actually shared between the two environments, and DevForce handles the movement of your business objects between tiers. You use the standard EntityQuery syntax to build true LINQ queries, which can be directed against a back-end data source or against the local DevForce cache. Your queries run asynchronously against back-end data sources, or synchronously against the local cache. Key to it all is the shared domain model. The domain model used by the Silverlight application is the same domain model used on the server, or in any .NET DevForce application: not an anemic object model with an unfamiliar API. You can add business logic - via custom methods and properties, DevForce property interceptors, and DevForce verification - to your shared domain model. You can also choose to deploy logic which is applicable to the client-side or server-side only.

Creating a DevForce Silverlight Application
You can use several different approaches to create a Silverlight application with DevForce:

386 | P a g e

IdeaBlade DevForce
1. Use the DevForce Silverlight application template.

DevForce Silverlight Apps

You can find this project template by choosing File - New Project' or 'File - Add - New Project' in Visual Studio. The template is in the DevForce folder under both the Visual C# and Visual Basic project types. Its use will result in the creation of both the Silverlight and web application projects for your DevForce Silverlight application. From here you can work on UI and domain model features, or reference already created projects. You'll use the DevForce Object Mapper to create the domain model and the "linked" Silverlight domain model. 2. Use the DevForce Object Mapper. In this approach, you start by creating a new model (or opening an existing one). In the "Project Settings" for the domain model you'll see a check box labelled “Create Silverlight Domain Model Project”; you can use this as a toggle to choose an existing Silverlight project in your solution, or to create a new one. Select “New project” and choose "Silverlight Application" as the project type in the resulting dialog. This will use the DevForce Silverlight application template to create the Silverlight and web application projects. These new projects will also be set to the selected value in the corresponding “Domain Model Project” and “Silverlight Project” dropdowns. You can then continue working in the Object Mapper as usual. 3. Use the standard Silverlight application template. The standard Visual Studio template for a Silverlight application will create both the Silverlight and web application (or web site) projects. If you want the web application to host your BOS, you will need to do the following:    Add EntityService.svc and EntityServer.svc files to the project; Add all necessary IdeaBlade references; and Modify the web.config to include the appropriate settings for the BOS.

You can find samples of the EntityService.svc, EntityServer.svc, and web.config files in the DevForce installation LearningResources10_Deployment\Snippets\IIS Files folder installed by DevForce. You'll need to use the Object Mapper to create the domain model and the "linked" Silverlight domain model.

Silverlight Deployment Steps
Please see the Deployment topic document (in the Deployment section of the Learning Resources) for detailed instructions and information about deploying Silverlight apps.

Questions and Answers
1. What is the "linked" Silverlight domain model? In order to provide a single "shared" domain model which can be used between application tiers, DevForce Silverlight creates two versions of the model - one compiled with .NET assemblies and one compiled with Silverlight assemblies. These two versions actually reference the same code files, and use the "linked" file feature of Visual Studio so that only a single copy of any file is required. The Object Mapper will perform this linkage for you: it will generate the domain model files into the .NET project, and then create links to these files in the Silverlight project. The result is that the domain model is available to both environments: a Customer class is the same whether it‟s defined in the client Silverlight application, or the server domain model assembly. One additional requirement also ensures that the types in your domain model can be shared: both the namespace and assembly names must be the same for the two assemblies holding the domain model. The Object Mapper also does this for you, so in most cases you don't need to be concerned with the implementation details.

387 | P a g e

IdeaBlade DevForce
2.

DevForce Silverlight Apps

Why is there an app.config in my Silverlight application, since Silverlight doesn't support configuration files? And why is it an embedded resource? DevForce Silverlight, like any DevForce application, requires configuration information when starting. To get that configuration information in DevForce Silverlight you should ensure that a file named app.config is located in the Silverlight application project and marked as an embedded resource. DevForce, via the Object Mapper and build-time utilities, will automatically create this file and embed it for you, and keep it up to date, so there's usually nothing for you to do; just don't delete the file. Probing for configuration in DevForce Silverlight follows the same probing logic, where applicable, as in a standard DevForce application. 122

3.

Where is the debug log? Unfortunately, a "client-side" debug log is not currently provided in the beta release of DevForce Silverlight. A debug log is generated on the BOS server, but it contains the usual server-side messages. A logging or tracing facility will be added in a future release.

4.

Do I have to host the BOS from IIS? And must it be the same web site that's serving the Silverlight application? You can still host the BOS from either the console (ServerConsole.exe) or Windows Service (ServerService. exe) in DevForce Silverlight. You can also host the BOS from a different web site than the Silverlight application. In both scenarios you need to ensure that a policy file is in place to avoid getting a cross-domain access error. You'll find a sample clientaccesspolicy.xml file in the LearningResources10_Deployment\Snippets\Silverlight folder installed by DevForce, along with a readme explaining how to deploy the file.

5.

Can a single BOS support both Silverlight and .NET client applications at the same time? Unfortunately it cannot, at this time. Currently, a flag in the config file named “clientApplicationType” determines whether the BOS will communicate with Silverlight or standard .NET client applications. This flag is global to the BOS. This restriction may be removed in a future release.

6.

How can I bind anonymously typed objects in my Silverlight application? The DynamicTypeConverter converts anonymously-typed objects to dynamically-typed objects for binding in Silverlight applications. Use the Convert(IEnumerable) method to convert one or more instances of an anonymous type to corresponding instances of a DevForce dynamic type. A DevForce "dynamic type" is a System.Type created dynamically at runtime. Generally the primary use for this conversion is in Silverlight applications, which do not support data binding to anonymous types. Projection queries are one common example in which return data will be anonymously-typed.

7.

How can I customize the communications channel to the BOS? For example, I need to set higher timeout values and add security. The default configuration used by DevForce uses HTTP binding, binary encoding, and a MaxReceivedMessageSize set to maximum value (2G), with all other attributes defaulting. To override the DevForce defaults you can add a ServiceReferences.ClientConfig file to your Silverlight application. If found, DevForce will use this file to configure communcations. A sample ServiceReferences.ClientConfig is provided with the DevForce installation, in the Deployment\Silverlight folder. Unless you‟re familiar with these files, it‟s best to copy the sample into your project and customize that. To use - include the file in the Silverlight application project, and mark it as “Content”. Remember that both the client and server configurations must be compatible for communications to

122

This logic is documented in an appendix to the “Hello DevForce” chapter of this Developers Guide, entitled “Probing Sequence for the App.Config File”.

388 | P a g e

IdeaBlade DevForce

DevForce Silverlight Apps

succeed, so you will likely need to modify your web.config file also. The “Deployment\Samples N-tier config files” folder contains samples showing different communications configurations.

Troubleshooting
1. You attempt to Connect to the BOS from the Silverlight client and receive the exception "An error occurred while trying to make a request to URI 'http://localhost:9009/EntityService.svc'" Connection errors can have many causes, but the first thing to check, especially in a new application using the ASP.NET Development Server, is that the Silverlight application is actually "served" by the web application. You can see this by looking at the address bar in the browser. If it doesn't start with "http://" then the application is instead loading from the file system. Why is this a problem? Because, for security reasons, a Silverlight application cannot make service requests unless served by a web server. In DevForce Silverlight this means that the application cannot connect to, or make other requests of, the BOS; thus, data cannot be retrieved from or saved to the back-end data source. The problem is easily remedied by ensuring that the web application project is always the startup project in your solution. 2. "No license found after probing all assemblies in the config file - Check for valid probeAssemblyNames in the config file." Possibly seen when double-clicking the “Error on page” icon in Internet Explorer and viewing the detailed error message. The probeAssemblyNames in the app.config embedded in the Silverlight application must be fully qualified assembly names. If not, since Silverlight is not able to load partial assembly names, no assemblies can be "probed" and no license found. DevForce will ensure the probeAssemblyName is correct if you set the updateFromDomainModelConfig setting in the file to either "Ask" or "Yes". This synchronization takes place at build time. The fully-qualified assembly name might look something like this: XML <probeAssemblyName name="FirstSilverlightApp, Version=1.0.0.0, Culture=neutral, PublicKeyToken=null" />

Probed assemblies are used by DevForce not only for validation of the product license, but also to determine the location of the domain model classes and for custom interface implementations. 3. "*** License violation *** - 'Distributed BOS' not supported with the current license: StandardEF" You must have a license for DevForce Silverlight in order to develop Silverlight applications with DevForce. The Silverlight samples in the Learning Units were created with an SL license key and you'll be able to run the samples as long as you don't regenerate the domain model. Once you regenerate the model with your license key, the sample may stop working due to the license violation. 4. I get the following exception when trying to fetch: "Unable to locate type: XX.YY" This not-so-friendly message may be caused by a type name mismatch between your .NET and Silverlight domain model assemblies. DevForce will seamlessly transmit entities between the SL and BOS tiers, but it does this using what is essentially a "shared" domain model. DevForce expects to see entities having the same fully-qualified type name, for instance "DomainModel.Customer, DomainModel, Version=1.0.0.0, Culture=neutral, PublicKeyToken=null", in both the .NET and Silverlight assemblies holding the model. This is why DevForce attempts to keep the assembly and namespace names in sync between the two projects, since without this type name equality, entities cannot move between tiers. This restriction will likely be

389 | P a g e

IdeaBlade DevForce

DevForce Silverlight Apps

removed in later releases of DevForce Silverlight. To fix the exception, ensure that the assembly and namespace names of the two projects containing the domain model are identical. 5. Why aren't my breakpoints working? This has nothing to do with DevForce, but we run into it from time to time. Double-check the Web properties on the web application project, and ensure that both ASP.NET and Silverlight debuggers are checked. 6. Your application was running initially and then crashes after a few minutes with an exception message such as: “Object reference not set to an instance of an object.. ---> System.NullReferenceException: Object reference not set to an instance of an object”. You may have encountered a problem that occurs when the IIS application pool has recycled. One of the best ways to insure this does not happen is to create a new application pool that does not recycle on a time limited basis and then assign your application to that pool. 7. Your application had been running and then crashes after you make a change to one or more of the files in the application directory. The exception includes this message: “Could not load file or assembly 'App_Web_........”. You may have encountered a problem that occurs when files in the application folder no longer match the compiled version located in the “Temporary ASP.NET Files” folder. You can force a rebuild of your application by deleting the “bin” folder and then replace it with a copy or by running the “aspnet_compiler.exe” command with the “-c” switch. You can find the command by first browsing to the folder “%SystemRoot%\Microsoft.NET\Framework\” and then open the v2.0.xxxxx subfolder (the numbers after v2.0 can vary) . Here is an example using the virtual directory name of the application: aspnet_compiler –v /MyApp -c 8. FIPS Compliance If your Silverlight application will be served from a web server on which FIPS (Federal Information Processing Standards) compliance is enforced, you will need to make the following changes to both the web.config and startup pages. In the web.config, you must set debug to false when FIPS is enabled. This is true even during development: you cannot set debug to true with FIPS enabled!

XML

<system.web> <!-Set compilation debug="true" to insert debugging symbols into the compiled page. Because this affects performance, set this value to true only during development. --> <compilation debug="false">

In the startup page (normally default.aspx), you cannot use <asp:ScriptManager> or any controls that rely on it since it generates a FIPS error. Therefore, you need to use html or javascript to start the Silverlight application. The example below is using html which will work in most non-IE browsers such as Firefox:

XML

<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd"> <html xmlns="http://www.w3.org/1999/xhtml"> <head>

390 | P a g e

IdeaBlade DevForce

DevForce Silverlight Apps

<title>DevForceSilverlightApp</title> <style type="text/css"> html, body { height: 100%; overflow: auto; } body { padding: 0; margin: 0; } </style> </head> <body> <object data="data:application/x-silverlight," type="application/xsilverlight"> <param name="source" value="ClientBin/DevForceSilverlightApp.xap" /> </object> </body> </html>

The value in red is the location of your xap file relative to the location of the startup page. If you use an .html page(ex: index.html) instead of an .aspx page, you will need to delete:

XML

<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">

in order to be compatible with Firefox.

9.

How to tell if the BOS is running. You‟ve received an error message from your client application stating, “The remote server returned an error: Not Found”. This is a communications error which occurs when the WCF client application is unable to complete a handshake with the server. There are, unfortunately, a myriad of reasons why this might occur, but one of the first things to check is if the service is actually running. You can do this easily: open the web browser and point it to the URL which the client application is using. For example, if the client app.config contains this...

C#

<objectServer isDistributed="true" remoteBaseURL="http://localhost" serverPort="9009" serviceName="EntityService.svc" />

...then open the web browser to http://localhost:9009/EntityService.svc. If the service is running, you will see a “Service description” page generated by WCF. If, instead, you see a page showing error information, then you know the service cannot be started and that your application will be unable to run. Usually the error message on the page has helpful diagnostic information. 10. Known Issues (Silverlight-Only)

391 | P a g e

IdeaBlade DevForce

DevForce Silverlight Apps

The Copy Local property on DevForce references in the web project must be set to true for the apps to run properly. This setting is required to allow the DevForce WCF services (defined in the *.svc files) to be compiled correctly. If not set, the services will not start, the client application will be unable to connect to the server, and you will see an error message as follows: The remote server returned an error: NotFound. If the service is unavailable, then also make sure that the endpoint bindings match between client and server. When you begin your Silverlight solution using the DevForce Silverlight Application project template, several DevForce assemblies are added as references to the web project; and for all, CopyLocal is set to true. However, if you manually add or modify references, you may see that the property is initially set to false (which is the Visual Studio default). Always check this property when you see the above error.

392 | P a g e

IdeaBlade DevForce

WinForm User Interfaces

WinForm User Interfaces
Features described in the section are included with the DevForce WinClient product, and apply to developers working with WinForm (not WPF) user interfaces. The features and facilities discussed in this chapter do NOT apply to Silverlight application development using the DevForce Silverlight product.

WinForm User Interfaces
UI Data Binding
NET Data Binding NET v. DevForce WinClient UI Data Binding for WinForms Data Binding with DevForce WinClient UI Designers For WinForms DevForce WinClient Data Binding Architecture Nested Property Paths Data Binding to Data Objects of Any Type When to Use .NET Data Binding Instead When Not to Use Data Binding at All

UI Architecture
Nested Property Paths The BindableList(of T) EntityPropertyDescriptors

UI Designers
BindingManagerDesigners

More on Third-Party WinForm Control Suites
Developer Express “DXperience” Infragistics “NetAdvantage”

DataBinders Troubleshooting
Third-Party Control Suites UI Performance Tuning Large BindingSource loads are Slow

DevForce WinClient Assemblies for WinForm Support

DevForce WinClient includes specific support for building WinForm user interfaces. ControlBindingManagers are provided to centralize all bindings to a particular business object type on a Form or UserControl. A special subclass of the .NET BindingList<T> class, the BindableList<T>, provides bi-directional binding refresh, facilitates sorts, and can be configured for automatic update as the contents of the DevForce WinClient local cache change. BindingDescriptors, DataConverters, and ViewDescriptors encapsulate your specifications for UI databinding behavior and facilitate reuse and consistency in your databindings across your user interface. UI designers ease and speed the layout of the UI view and the setup and configuration of data bindings. We‟ll detail all of these classes and facilities in this chapter.

393 | P a g e

IdeaBlade DevForce

WinForm User Interfaces

UI Data Binding
A primary concern of any UI is the movement of data between a UI control property such as the Text property of a TextBox and a corresponding value in a data item such as the FirstName property of an Employee object. We want to display “Nancy” in the TextBox when we she becomes the current Employee. We want to update her Employee object when the user changes her name to “Sally”. We could write the code to do this ourselves. We could fill the TextBox when “Nancy” becomes the current Employee. We could subscribe to the TextBox‟s Leave event and, in our handler, pull “Sally” from the TextBox to set the Employee‟s FirstName property. This is called “imperative” programming. It is tedious and error prone and difficult to refactor when we want to change the process or abstract it from the form. There are times when it is the right approach, but there should be a an easier and safer way for 90+% of cases … and there is. It‟s called “UI Data Binding”. “UI Data Binding” describes the mechanism by which UI control properties are “bound” to data item properties. “Binding” in this context means that data values are exchanged between the UI control property and the data item property in response to particular events recognized by the Data Binding infrastructure. Some events trigger the setting of the UI control property; some trigger the setting of the data item property. The exchanges happen automatically. We don‟t have to write the transfer code. Our job as programmers is to declare the mapping between each UI control property and a corresponding property of the data item. We map once and the infrastructure executes according to our plan. This is called “declarative” programming. There are objects galore in object-oriented programming and it‟s often difficult to follow what object we‟re talking about at any given moment. We‟ll follow Brian Noyce‟s convention of referring to the data bound object – the source and temporary repository of data displayed in the UI – as the data item. We say “data item”, not “business object”. Our data item examples usually are business objects but they don‟t have to be. We can bind to any application object including the parent form or control.

NET Data Binding
.NET itself provides the basic DataBinding infrastructure. That infrastructure was greatly improved in .NET 2.0 as were the design tools to exploit it. DevForce WinClient builds on .NET Data Binding in ways we will explore later in this section. Most DevForce WinClient developers working with WinForms never bother learning raw .NET Data Binding because the DevForce WinClient extensions and adaptation make binding much easier and consistent. Nonetheless, there will be times when a solid understanding of native .NET Data Binding is helpful or even essential123. In this chapter we will cover just a few of those concepts and techniques and, at that, only in the context of the use cases that require their use or explanation. Here are a few resources for learning more about DataBinding in as it comes out of the .NET box. [Noyes] Noyes, B., 2006, DataBinding with Windows Forms 2.0, Boston, Addison Wesley. At last a book dedicated to Data Binding in .NET 2.0. Brian Noyes presents the definitive account of the subject in a clear, concise, professional, and enjoyable read. You need this book; and on this topic, you need no other.

123

Most notably when we must bind to non-data properties of a control (e.g., background color) or bind to .NET and third-party controls not yet supported by DevForce WinClient.

394 | P a g e

IdeaBlade DevForce

WinForm User Interfaces

[MacDonald] MacDonald, M., 2006, Pro .NET 2.0 Windows Forms and Custom Controls, Berkeley: Apress. This detailed and lively examination of Windows Forms construction in NET 2.0 is the best WinForms resource so far. There is plenty of meat – over 1,000 pages - garnished with hard-to-find tips. The chapters include coverage of tool strips, DataGridView, .NET data binding, sound and video, threading, and interface styles. The appendix on ClickOnce is a bonus. [Petzold] Petzold, Charles, 2006, Programming Microsoft Windows Forms, Redmond, Microsoft Press. An excellent and approachable introduction to the Windows Forms features new in .NET 2.0. Strips away the crud generated by the .NET designers so we can see the bare bones, sinews, and muscles. Other books do more but they‟re also huge; Petzold‟s book is spare and focused.

NET v. DevForce WinClient UI Data Binding for WinForms
There remain serious shortcomings in the base .NET implementation and there are opportunities to wrap and extend .NET Data Binding that improve the development experience and facilitate proper separation of controller and view logic. DevForce WinClient UI Data Binding for WinForms aims to overcome the shortcomings and provide the helpful extensions. The main points are: Issue Bugs DevForce WinClient Solution There are a great many traps in the base .NET implementation. Actual data binding behavior isn‟t exactly as documented. Events don‟t fire when they are supposed to or don‟t fire at all. Data aren‟t written to the control or data item as prescribed. Some controls don‟t follow the rules. We catch and work around many of them so you don‟t have to. Many desirable behaviors only become available when the data item implements the appropriate interfaces. DevForce WinClient business objects do. The DevForce WinClient binding collection, BindableList(Of T), fills in many of the gaps. .NET and third-party UI controls are wildly inconsistent both in their property names and in their support for the data binding infrastructure. The unwary developer is in for a long and frustrating voyage of discovery if she sets out all on her own. Books and tutorials help. But wouldn‟t it be easier to let IdeaBlade do the grunt work? We‟ve encapsulated much of the bizarre and “nuanced” behavior inside consistent and simple DevForce WinClient APIs and DevForce WinClient Control Binders. You can always fiddle with the controls directly and even write your own Control Binders but DevForce WinClient‟s default behavior is usually just what you want. Text book examples always show binding to simple object properties like FirstName. What about anEmployee.Address.State.Abbreviation? You don‟t see that one often – for good reason. It doesn‟t work. Oh, it works some of the time. But never in grids, and the breakdowns are difficult to predict. The DevForce WinClient DataBinding collections take care of this important problem.

Missing implementations Inconsistency

Nested Properties

395 | P a g e

IdeaBlade DevForce

WinForm User Interfaces

Issue Grids

DevForce WinClient Solution The .NET DataBinding infrastructure applies to whole UI controls. It works well with loose controls – controls bound to a property of a single data item. All bets are off for behavior inside a UI control. Container controls – grids especially –observe different Data Binding rules that are imperfectly applied. Nested properties, for example are not supported. DevForce WinClient grid binding managers for WinForms strive to overcome these deficiencies for each of the supported grids, including the .NET DataGrid and DataGridView, the DeveloperExpress XtraGrid, and the Infragistics UltraGrid. Sometimes we need a UI control to respond to something in the data item that is not expressed – or not expressed appropriately – as a public property. Our first instinct is to add the needed property to the data item. That may not be possible; we can‟t add a property to a class we don‟t own (e.g., a .NET class). It may not be wise to add the property even if we could; proper delegation of responsibility tells us that we should not add a purely UI property (such as an image) to a business object class. DevForce WinClient provides the means to enrich the data item with bindable “dynamic properties” without altering the data item class itself. .NET 2.0 made it easier to add format and parse logic to a binding but the approach remains crude and does not encourage abstraction. Nor does .NET 2.0 provide an easy answer to the need to change the format and parse behavior dynamically in accord with business or application rules. You have to wire that up yourself. .NET does little to help you organize your data bindings logically. The standard approach is to pile data bindings onto the form with no apparent concern for how groups of bindings serve a common purpose. We can discover all bindings associated with a BindingSource with the expression:
aBindingSource.CurrencyManager.Bindings

Dynamic Properties

DataConverters

Disorganization

But this is an after-thought at best and the collection returned cannot be manipulated directly – we can‟t add or remove items for example. The DevForce WinClient binding managers for WinForms provide an explicit mechanism for organizing bindings around a specific collection of data items and for a particular purpose. This is the foundation for later refactorings that facilitate maintenance and testing. Reuse Static analysis of an application with Data Binding reveals an immense amount of duplicate code. The CompanyName property, for example, may appear thirty or more times over as many screens. In most cases it is bound anew each time with an exact duplicate (one hopes) of the other twenty-nine parameters. The DevForce WinClient infrastructure for Winforms – ViewDescriptors in particular – provide the foundation for a clean refactoring to reusable and testable data bindings.

Most DevForce WinClient developers succeed admirably without ever dropping down to raw .NET Data Binding. Nonetheless it is vital to understand that we can do so comfortably – and probably will do so – without rocking the DevForce WinClient Data Binding boat in the slightest.

396 | P a g e

IdeaBlade DevForce

WinForm User Interfaces

Data Binding with DevForce WinClient UI Designers For WinForms
DevForce WinClient UI Designers for WinForms are the easy way to bind data items to loose controls and grids. The designers can also deposit controls on the “form” canvas, giving them names that conform to our preferred conventions. The control population feature alone can save hours of tedium. The UI Designers may be found on the “IdeaBlade DevForce WinClient” tab of the Visual Studio Toolbox. A typical example looks like so:

 Drag the ControlBindingManager tool onto the canvas. A small icon representing an instance of the ControlBindingManager class appears in the component tray beneath the canvas.  Open its context menu (right-click with mouse) and see some choices124.

 Launch “Autopopulate”  Pick “Employee” as the entity class to bind. The Designer offers a list of Employee public properties. We select a few until it looks like so:

The designer has suggested some controls and control names for us. We push “Ok”. The labels and controls appear on the canvas. We do some cleanup: erase the photo label, re-label the manager, re-size and re-locate the image.

124

The details of this scenario are covered at a sane and leisurely pace in other documentation.

397 | P a g e

IdeaBlade DevForce

WinForm User Interfaces

We go back to the toolbox, open the “All Windows Forms” folder, and drag a BindingSource component on to the canvas. We examine the controlBindingManager1‟s property sheet, find “BindingSource”, and set it to the new instance, bindingSource1. Our Visual Studio design view now looks like this.

At this point we have controls governed by a ControlBindingManager which looks for Employee objects in a BindingSource. We don‟t have any employees in that source yet. We‟ll solve that next. We double click the form; Visual Studio hooks up the form‟s Load event to a Form_Load handler template. We add a single line to that handler that asks the default EntityManager to get every Employee and store the resulting list into the data source of our BindingSource instance:
bindingSource1.DataSource = DomainModelEntityManager.DefaultManager.Employees.ToList(); // C# bindingSource1.DataSource = _ DomainModelEntityManager.DefaultManager.Employees.ToList() ' VB

We compile, run, and … there‟s our form.

We can now circle back and add more stuff, some navigation, some buttons, and pretty soon we have an Employee editor. DevForce WinClient offers a variety of UI Designers for WinForms that work more or less this same way. Each UI Designer is a .NET component that runs within the Visual Studio development environment and is accessible from the Visual Studio toolbox. A UI Designer generates .NET source code directly into the “Form1.Designer” class file which is the companion to the “Form1” class file we‟ve been modifying. 125. Here‟s a peek at the code generated in C#
… // // controlBindingManager1

125

These two files together define the entire “Form1” class. Each file defines a “Partial” class meaning that it contains a part of the definition of the class. The compiler assembles all the partial class files together into the finished class. “Partial Classes” is a .NET language feature introduced in .NET 2.0.

398 | P a g e

IdeaBlade DevForce

WinForm User Interfaces

// this.controlBindingManager1.BindingSource = this.bindingSource1; this.controlBindingManager1.BoundType = typeof(Entities.Employee); this.controlBindingManager1.Descriptors.Add(new IdeaBlade.UI.WinForms.ControlBindingDescriptor(this.mFirstNameTextBox, typeof(Entities.Employee), "FirstName")); this.controlBindingManager1.Descriptors.Add(new IdeaBlade.UI.WinForms.ControlBindingDescriptor(this.mLastNameTextBox, typeof(Entities.Employee), "LastName")); …

The generated code is not pretty but it works. Usually we just leave it alone unless we wish to learn how DevForce WinClient writes data binding code. This snippet will make more sense when we cover the DevForce WinClient data binding architecture for WinForms below. There will come a time when we have to dig in. The UI Designers are fine for quick, one-off screens that can be defined statically at design time. On the other hand, we‟ll write the code ourselves if we must change or add data bindings on the fly such as when a control toggles from read-write to read-only. We may discover that certain control logic appears repeatedly and want to re-factor. These and other reasons compel us to learn more about DevForce WinClient data binding for WinForms. It is not hard. We have the UI Designer output as a guide. We can mix generated and custom code as we deem appropriate. So let‟s leave Designers behind for now to explore the inner workings of DevForce WinClient data binding.

DevForce WinClient Data Binding Architecture
In this chapter we unveil the DevForce WinClient data binding architecture that implements the MVC pattern. That architecture pursues certain goals:    Organize bindings around their datasources. Strive for order, consistency, and simplicity in the face of heterogeneous controls and objects. Promote code re-use and easy maintenance.

We‟ll hearken back to these goals as we explain what may seem initially a complex implementation model.

High-Level View
Figure 7 shows a high-level view of the DevForce WinClient WinForm data binding architecture. At the extreme left side of the figure you see the data source – which for the purpose of data binding, is a business object (no matter where the business object got its data from). At the extreme right side is the data target: a UI control like a TextBox, ComboBox, or DataGrid.

399 | P a g e

IdeaBlade DevForce
Figure 7. DevForce WinClient Data Binding Architecture : High-Level View

WinForm User Interfaces

DevForce EF WinForm Databinding Architecture High-Level View
Formatting

Data Source (Business Object)

DataConverter

DataBinder

Data Target (UI Control)

Validation

Parsing

A property may not have the same data type in the business object data source that it needs in the UI control. A DateTime or numeric property, for example, may be targetted for display and edit in a TextBox, which only understands string values. And even if the UI control designated to display the business object property accomodates the property‟s data type, the developer may want to permit (or require) it to be entered in a different form in the control. A social security number, for example, might be stored as 123456789 in the business object, but be entered as 123-45-6789 in the control. Because the data may need to take different forms in the source and target, data transformations must take place as it travels between the two. The transformation (if any) that takes place as data moves from the source to the target is known as formatting; the transformation that takes place as data moves from the target to the source is known as parsing. Closely related to parsing is validation. Parsing ensures that the entered value can be transformed into the data type required by the data source; validation ensures that the transformed value conforms to business rules. DevForce WinClient WinForm data binding interposes two objects between the source and the target: a DataConverter and a DataBinder. The DataConverter encapsulates information about formatting, parsing, and validation requirements. The DataConverter knows the data type of the value it will receive from (and deliver to) the data source; it knows how to transform that value into a variety of anticipated types and formats; and it knows those aspects of the data validity requirements that are appropriate for enforcement at the point-of-entry. The DataBinder, on the other hand, knows about a specific UI control and its requirements. The DataBinder uses information from the DataConverter to configure the UI control. It also tells the DataConverter what type it needs, leaving it to the latter to deliver the requested type. You‟ll learn more about DataConverters and DataBinders as you proceed through this chapter. Right now, let‟s zoom in just a bit to look at some additional details of the DevForce WinClient DataBinding architecture.

DevForce WinClient WinForm Data Binding Architecture – Zooming In
Figure 8 depicts the key components of the DevForce WinClient binding architecture. A BindingManager collects BindingDescriptors, each of which describes the binding of a single property to a single UI control (which might stand alone, or be embedded in a DataGrid column). The BindingManager receives data from a BindableList (often an EntityList) which collects instances of a single type (usually a business object type). BindingDescriptors

400 | P a g e

IdeaBlade DevForce

WinForm User Interfaces

encapsulate specifications about an object type and property; a UI control; and desired UI behavior (controlled by the DataConverter).
Figure 8. DevForce WinClient Data Binding Architecture: Zooming In

The ViewDescriptor, you will note, marries the behavioral specs encapsulated in a DataConverter (which are specific only to a particular simple data type, like a String or a DateTime) to a specific property from a specific type -- usually a business object type. You can, in other words, use it to fix the UI behavior of the Employee.BirthDate, or any other specific property, from your business model. You can then reuse the ViewDescriptor as often as desired to present consistent property-specific UI behavior across your entire application, regardless of the number of different contexts in which that property needs to be exposed. You have but one specification to create and maintain. That‟s the quick summary of DevForce WinClient databinding. Now let‟s explore it in more detail.

DevForce WinClient Data Binding Definition Process
At the macro level data binding definition is a three step process Set up one of the kinds of DevForce WinClient Binding Manager Add binding descriptors to that manager Create a BindingSource and assign it to the manager At run-time, we set the DataSource of the BindingSource and fill that DataSource with data items – probably business objects. We may subsequently tune the bindings as the application and session activities require.

The ControlBindingManager
We‟ll delve into these concepts by beginning with the ControlBindingManager. We rely on a ControlBindingManager to help us manage the UI controls on a form or other container (user control, panel, etc.) that displays a single data item. That data item could be an Employee object, one among a list of Employee objects selected for display or editing. We can work with only one Employee data item at a time in such a container. That one data item is the “current” data item. A dialog devoted to editing a single employee typically features several “loose controls” layed out in some meaningful pattern across the canvas of the “form.” Each such control can display a single value of a single property of the “current” employee. The TextBox is a good example of a loose control. Its Text property can display a string associated with a single property of one object – an employee first name for example.

401 | P a g e

IdeaBlade DevForce

WinForm User Interfaces

A ComboBox is also a loose control. It displays a list of potential values in its drop-down but only one of those values belongs to the current data item and it is bound to the value accessed by a single property of the current data item. For example, the value of the State property of the current employee‟s home address, when bound to a ComboBox, appears as the selected item in the list of fifty states presented in that ComboBox. The ControlBindingManager manages the entire collection of Employee data item bindings to these loose controls. It can manage bindings to any number of controls, but all must be bound to employee objects in a single BindingSource, the BindingSource of Employee objects edited in this “form.” The managed controls may all belong to the same vendor control suite (e.g., .NET), or may belong to different vendor control suites. Specifically, you can mix and match .NET controls with Developer Express or Infragistics controls (or other controls for which you have written custom binders).

Grid Binding Managers
A grid is a special kind of container control with its own data source. It is a tabular container of other controls, one control per column. Each row represents a single object in the data source. Each table cell displays a property of the row object in the column‟s control.

The value in a cell can be data bound to a property in a data item just as with loose controls. The collection of all such bindings falls under the care of a DevForce WinClient binding manager. The ControlBindingManager is not a candidate for this job. There are a number of peculiarities to the grid binding process that necessitate a distinctive kind of binding manager, a grid binding manager.      A grid control displays many data items simultaneously. Accordingly, the grid data binding mechanism must accommodate binding to properties of more than one data item at a time. The .NET grids we‟ve seen require that all column controls belong to the suite offered by the grid vendor. We can‟t mix .NET controls with DeveloperExpress controls in either a .NET or DevEx grid. Our ability to bind to a grid is often constrained. Outside the grid, we may choose from a wide spectrum of the vendor‟s controls. Inside the grid we are limited to just a few types of column control. 126 Column controls may look like their loose control cousins but they are usually crippled in annoying and peculiar ways. The data binding rules inside a grid differ from the data binding rules for controls outside the grid. There are different binding events, different format and parsing behaviors, and different phases in the value editing cycle. Grid architectures and APIs differ markedly from one grid vendor to the next.

Consequently grid data binding is considerably more challenging than data binding of simple controls and it is infeasible to write a single binding manager that can handle every grid of every vendor. DevForce WinClient offers multiple grid binding managers, one per supported control suite. There are four such managers at this writing:
126

Some grid vendors enable us to develop new column control types that incorporate controls otherwise available only in loose form. The .NET DataGridView is especially flexible in this regard. We can extend DevForce WinClient to accommodate these new column controls.

402 | P a g e

IdeaBlade DevForce
   
DataGridBindingManager (for the now deprecated .NET 1.1 grid) DataGridViewBindingManager (.NET 2.0 grid) XtraGridBindingManager (Developer Express “DXperience XtraGrid”) UltraGridBindingManager (Infragistics‟ “NetAdvantage UltraGrid”).

WinForm User Interfaces

While the differences between the ControlBindingManager and these grid binding managers are important, the fundamental concepts are almost identical for all of them – a consistency intended by DevForce WinClient‟s data binding abstraction. Therefore, we can continue to explore the DevForce WinClient data binding architecture using the ControlBindingManager as our model with only occasional nods to the specific character of a grid binding manager.

A Guided Walk Through the DevForce WinClient DataBinding Architecture
In the next few pages we‟ll take a stroll through the menagerie of the DevForce WinClient Data Binding Architecture, identifying the creatures we encounter as we walk through the process of building a “form”.

The ControlBindingManager Revisited
We just met the ControlBindingManager. It manages a collection of bindings between UI controls and the current data item in a collection of bound data items. A ControlBindingManager can only manage bindings to data items of a single bound type. The data items may be derived from that type (as we‟ll discuss later) but they must “belong” to the same common type. One of our first steps when creating a binding manager is to specify its type 127. We can set the ControlBindingManager‟s BoundType property as in
mEmployeeCbm.BoundType = typeof(Model.Employee);

Or we can specify it in the manager‟s constructor
mEmployeeCbm = new ControlBindingManager(typeof(Model.Employee));

Do set the BoundType early in the definition of the binding manager; it is not something to change later.

BindingDescriptorCollection
We‟ve said that the ControlBindingManager manages data bindings. It would be more precise to say that it manages a collection of BindingDescriptors, each of which is responsible for a data binding. We can access that collection of descriptors via the manager‟s Descriptors property as in
ControlBindingDescriptorCollection descriptors = mPageCbm.Descriptors;

We can add and remove descriptors from this collection at will. So what is a descriptor?

BindingDescriptor
A binding descriptor defines the binding between a specific UI control on the form and a particular property of a data item.

127

The type need not be a DevForce WinClient business object. It can be any object in any referenced assembly.

403 | P a g e

IdeaBlade DevForce

WinForm User Interfaces

We‟ll unpack the components of a binding descriptor in a moment; here‟s a preview to get the sense of how we could create one.
ControlBindingDescriptor aDescriptor; aDescriptor = new ControlBindingDescriptor( mFirstNameTextBox, typeof(Employee), "FirstName");

We see the essential ingredients are:    a control the type of the data item the name of the data item property

We could add this descriptor to the ControlBindingManager‟s collection.
descriptors.Add(bd);

We can create a descriptor and add it to the collection in a single line.
aDescriptor = descriptors.Add(mFirstNameTextBox, "FirstName");

Observe that:   We omit the data item type because it must be the same as the bound type of the ControlBindingManager to which this descriptor collection belongs. The Add method returns the binding descriptor; we may want to adjust some of its properties before adding another.

In words, we have a ControlBindingManager managing multiple data binding definitions that relate properties of UI controls to properties of an object of a specific bound type.

Architecture Summary #1
Our Data Binding Architecture diagram at this point looks like this:

We‟ve mapped a group of UI controls to the properties of a single bound type, all under the care of binding manager. We‟re at roughly this stage when we emerge from the UI Designer. Next we associate this mapping with a container of data items, the BindingSource.

BindingSource
Data binding in action involves marshalling data between UI control properties and the properties of actual data item instances. A ControlBindingManager, defined for Employee objects, at some point must be given some real employee objects to bind. Those objects are, collectively, the datasource for the binding manager. The ControlBindingManager needs access to this datasource but we don‟t specify it directly128. Instead we specify a BindingSource that, in turn, holds our datasource.

128

We used to in .NET 1.1, before the advent of the BindingSource.

404 | P a g e

IdeaBlade DevForce
We can set the binding manager‟s BindingSource through a property:
mEmployeeCbm.BindingSource = mEmployeeBindingSource;

WinForm User Interfaces

We might have set it when we created the ControlBindingManager as in
mEmployeeCbm = new ControlBindingManager(mEmployeeBindingSource);

A BindingSource must contain objects of the same type, just like a binding manager. Therefore, the newly constructed mEmployeeCbm can infer its BoundType from its BindingSource. .NET 2.0 introduced the BindingSource as the preferred means to maintain collections of data items. The BindingSource is a smart collection that does far more than hold data items. For example, it keeps a pointer to the “current” data item and provides methods for moving the pointer up and down the list; each move causes a new item to become “current” and, in the process, raises a host of events leading to an update of the UI. The programmer can sit back and enjoy the show. We access the collection of data items in the BindingSource by means of its DataSource property. A DevForce WinClient application should set the BindingSource.DataSource as soon as possible, even if there are no items in that DataSource. One approach would be to set its DataSource during construction. Consider the following:
mEmployeeEntityList = new EntityList<Employee>(); mEmployeeBindingSource = new BindingSource(mEmployeeEntityList, null); mEmployeeCbm = new ControlBindingManager(mEmployeeBindingSource);

We are    Creating an empty EntityList to hold the Employee data items Creating a BindingSource based on this EntityList; the BindingSource discovers its bound type from the EntityList129. Creating a new ControlBindingManager with an Employee BoundType and a BindingSource called mEmployeeBindingSource.

This is not a typical initialization sequence but it makes the essential points about the relationships among DataSource, BindingSource, and binding manager.

BindableList DataSource
A DevForce WinClient binding manager requires a list datasource whose contents are all of the same type (or derived from the same type) as the binding manager‟s bound type. We must provide a list datasource. If we only need to bind to one data item, we‟ll turn it into a one item list as in
mEmployeeEntityList = new EntityList<Employee>(new Employee[] { anEmployee });

The datasource list should be of a type derived from the DevForce WinClient BindableList<T>130 in order to take full advantage of DevForce WinClient Data Binding features such as nested and dynamic properties.
EntityList<T> is a typical choice in DevForce WinClient applications because it both derives from BindableList<T> and holds DevForce WinClient business objects.

129

The “null” parameter means there is no DataMember which, in turn, means that the BindingSource finds its contents in the DataSource. We‟ll consider the “DataMember” when we discuss BindingSource chaining in Master / Detail scenarios. If we don‟t supply such a list, DevForce WinClient will wrap the list in a BindableList<T> where T is the type of object in the list. We cover BindableList<T> in detail elsewhere in the chapter.

130

405 | P a g e

IdeaBlade DevForce Filling the DataSource

WinForm User Interfaces

We don‟t have to fill the datasource with data items while we‟re setting up the bindings. In fact, this is rarely a good idea. We generally don‟t know which items to display until after the application starts, the “form” is loaded, and we‟ve given the user the chance to tell us what he wants to see. Suppose the user filled in a filter form that we resolved into a query. We might see code such as.
// Get selected employees and sort them. mEmployeeEntityList.ReplaceRange(pm.GetEntities<Employee>(filterQuery); mEmployeeEntityList.ApplySort("FullName", ListSortDirection.Ascending, false);

Note that we used the ReplaceRange method. Do not re-assign the EntityList as in
// Wrong!!! Don‟t do this !!!. mEmployeeEntityList = pm.GetEntities<Employee>(filterQuery);

This code changes the mEmployeeEntityList object reference. Meanwhile, the BindingSource.DataSource is still holding on to the previous entity list. The UI will show the Employees in that old list rather than the ones we fetched into mEmployeeEntityList! Do not re-assign the BindingSource.Datasource as in
// Wrong!!! Don‟t do this !!!. mEmployeeBindingSource.DataSource = pm.GetEntities<Employee>(filterQuery);

This “works” in the sense that the UI will show the newly retrieved employees. But again we‟ve changed the collection object that was the BindingSource‟s DataSource. This change can cause massive disruption to the data bindings. This can cause DevForce WinClient to break and remake all of the bindings when it is managing the BindingSource (as it is in our example). If the binding manager is a grid binding manager, the end user‟s column adjustments (columns widened or moved, for example) could be lost. The effects are rarely fatal but they are disruptive and annoying to the users. Do use ReplaceRange()

Architecture Summary #2
Our Data Binding Architecture diagram at this point looks like this:

This is where we are after we‟ve mapped controls to a bound type, added a BindableList, and populated that list with business entities using the EntityManager. We could be looking at a filled in form right now.

406 | P a g e

IdeaBlade DevForce BindingDescriptors Revisited
BindingDescriptors bind to the “data” property of a control

WinForm User Interfaces

The acute observer may notice a slight discrepancy between the characteristics of a .NET Data Binding and those of a binding descriptor. A .NET data binding can bind a data item property to any public property of the control. We usually bind to the control‟s “data” property.
Text is the “data” property of a TextBox. For example, mFirstNameTexBox.Text returns a string bound to the FirstName property of an Employee data item

A DevForce WinClient binding manager assumes that you want to bind the “data” property of the control so we don‟t have to specify it131. This saves both time and heartache because the name of the “data” property is not always the same from one control to the next.

Binding to non-data properties of a control
.NET data binding itself supports binding to almost any UI control public property. We can bind to the background color of a TextBox if we want to. Of course this presupposes that there is corresponding property of the data item that returns a color appropriate for this purpose. Such properties are rare in most application objects, and even rarer still in business objects. Rare or not, we may want to bind to one of these other control properties. The control‟s ReadOnly or Enabled property is a common target for data binding. Also, as we‟ll learn, while the data item may not have an appropriate bindable property, we can simulate such a property with the DevForce WinClient “dynamic property” feature. Fortunately, we can have both .NET data bindings and DevForce WinClient data bindings in the same “form” without conflict. If our Employee had an OutOfRangeBirthDateColor property, we could add a line to our “form” initialization method such as:
mBirthDateDateTimePicker.DataBindings.Add( "BackColor", EmployeeBindingSource, "OutOfRangeBirthDateColor");

DataConverters
A data converter massages data on route between the object property and the control property. It does so by means of the .NET data binding Format and Parse events. The Format event fires just as the object property data are about to be handed to the control property. We handle the event if we care how our data are displayed. We have many choices as illustrated in these examples:

The Parse event fires when data will flow in the reverse direction, that is, just as the changed input in the control is about to be handed to the object property.

131

In fact, we couldn‟t specify that property if we want to do so. A binding manager can only bind to the “data” property of a control, as we will shortly discusss.

407 | P a g e

IdeaBlade DevForce

WinForm User Interfaces

Parsing the input means converting it into a format acceptable to the data type of the receiving property. The parsing process is closely related to the formatting process. If we used a regex expression to format our data, we probably could use the same regex expression to control and parse the user‟s input. Were we confronted with input as shown below, we would accept the first, reject the second, and, depending on the sophistication of our parsing logic, might accept the third.

After successful parsing, we could pass the parsed product directly to the property, confident that we would not crash the application due to data type incompatibility. The property then should determine if the input is acceptable from a business standpoint. In other words, having parsed the input, we should now validate it. In principle, the object alone should validate the data. In practice, the user experience is terrible if there is a lengthy delay between user input and validation. Such delays are among the principle frustrations of browser-based applications.

Benefits of DataConverters
DataConverters abstract the management of formatting and parsing operations into an object. This abstraction works for loose controls and for grid columns; you don't have to approach them differently. The abstraction works for controls of different types and from different vendors. There is a great deal of variation in the databinding syntax and behavior among UI controls; events often don't fire (or fire properly). In DevForce WinClient, we have gone to considerable effort to discover and do the best thing for each control. The DataConverter unifies separate but related concepts of Editability, Display (format), Parsing (valid for a type), and Validation (valid from a business rules perspective). DataConverter objects can be pre-configured and held in a library prior to their use in any actual data binding. By contrast, if you hook up the events yourself (or add them to the data binding expression), you must wait until you have such a binding to do so. With a DataConverter, you configure and hold it; you retrieve and apply it when and where appropriate. By applying the behavioral specifications encapsulated in DataConverters to specific properties of specific business objects – for which marriage DevForce WinClient supplies the ViewDescriptor object -- you can pre-define UI behavior on a property-by-property basis and then apply it with absolute consistency across your entire UI – not matter how complex that UI becomes. The point of all of this is to facilitate development of scalable UI's that are consistent throughout and easy to maintain.

The base DataConverter class
There is a base DataConverter class. We can assign an instance of it to almost any ViewDescriptor. It is very flexible. It can format and parse for properties returning most data types. It works with virtually all of the single value UI controls. The base DataConverter class is also pretty limited. Although we can set editability, we are obliged to accept its default formatting and parsing behavior and there is no validation option. It is, after all, only a base class.

408 | P a g e

IdeaBlade DevForce Many standard data converters

WinForm User Interfaces

DevForce WinClient ships with a large number of specialized data converter classes that inherit from DataConverter. Among them are BooleanConverter, DateConverter, DateRangeConverter, NumericConverter, NumericRangeConverter, RegexConverter, TextConverter, … the list goes on.

The converter‟s object property data type
As the names suggest, these converters translate to and from a particular object property data type, the converter‟s “base data type”. When we add a data converter to a ViewDescriptor, the converter‟s base data type must be compatible with the data type of the ViewDescriptor‟s object property. For example, a binding to the employee BirthDate or HireDate properties would likely use a data converter with a DateTime base type. The program might crash if we used a text data converter.

The converter‟s control property data types
A rich data converter can format to and parse from a variety of control property data types. A converter that supports many control property types can support many controls. The DataConverter class is especially accomplished in this respect. A data converter can only be used with a compatible control. If we bind BirthDate to a TextBox, we must use a converter that can translate between DateTime and string because that‟s what the TextBox control‟s display property expects. If we bind to a Calendar control that takes a DateTime input, the converter should be able to pass the date straight through. The developer of a data converter anticipates the controls we might want to bind to and provides suitable conversions from the base data type.

Control-specific data converters
A data converter may be able to support many controls or it may be dedicated to just a few. It might specialize in a single control or a small family of controls to take advantage of the special properties of those controls. The ListConverter is an example of a control-specific data converter. It specializes in the controls, such as ListBox and ComboBox, that let the user pick one entry from a list of choices. These list controls have distinctive properties such as Datasource, DisplayMember, and ValueMember. The ListConverter recognizes these properties and exposes them to the developer for configuration. The developer sets them, statically or dynamically, and leaves it to the ListConverter to manage the user interaction.

Custom data converters
Developers can create their own data converters. The easiest way is to sub-class an existing converter and modify it to suit our needs. Alterations could include new validation checks, different formatting, more robust parsing, or better control over the visual aspects of the bound UI controls. A more challenging approach is to write an entirely new data converter, either by sub-classing from DataConverter or, bolder still, by implementing the IDataConverter interface.

Data converters and re-usable code
This discussion highlights another way the DevForce WinClient data binding architecture promotes re-usable code. We shouldn‟t have to write individual custom event handlers for each binding when we can see commonalities across a wide range of cases. DevForce WinClient captures those commonalities in the data converter. As we add new binding descriptors (or ViewDescriptors) to our UI, we‟ll use one or another flavor of data converter repeatedly,

409 | P a g e

IdeaBlade DevForce

WinForm User Interfaces

configuring each converter instance to meet the particular requirements of the object properties to which we will bind.

ViewDescriptors
A ViewDescriptor marries a DataConverter to a specific property of a specific business object:

The ViewDescriptor describes binding behavior in a completely portable manner for its target property. While a binding descriptor is tied to a particular control on one form (or to a single column of a particular grid), a ViewDescriptor is independent of any UI control and can be reused across any number of data bindings to any number of different UI controls. For a given business object property, you need only describe the desired behavior once.

ViewDescriptor Catalogs
With ViewDescriptors we can programmatically describe how a specific property of a business object type should be rendered by the UI before we layout a single screen. We can standardize the visual treatment of objects and object properties.

Inside the ViewDescriptor
The ViewDescriptor captures four pieces of view information:     the type of the object to bind the property path from that type which leads to the property that will be bound the display name for this property the format, parse, and validation information (which is encapsulated in a DataConverter)

ViewDescriptor‟s Property Path Digression: hiding the ViewDescriptor
Recall earlier that we can define a binding descriptor without explicitly referencing the embedded ViewDescriptor. We showed a constructor with parameters for specifying the control, object type, and property path.

410 | P a g e

IdeaBlade DevForce

WinForm User Interfaces

The data converter is so important that DevForce WinClient exposes it both as a property of the binding descriptor and as a parameter of a binding descriptor constructor overload.

C#

TextConverter aDataConverter = new TextConverter(IdeaBlade.UI.Editability.Optional, 40); ControlBindingDescriptor aControlBindingDescriptor = new ControlBindingDescriptor(mLastNameTextBox, typeof(Employee), "LastName", aDataConverter); | | | | Control ObjectType Property Path DataConverter Dim aDataConverter As New TextConverter(IdeaBlade.UI.Editability.Optional, 40) Dim aControlBindingDescriptor As New ControlBindingDescriptor( _ mLastNameTextBox, GetType(Employee), "LastName", aDataConverter) | | | | Control ObjectType Property Path DataConverter

VB

Internally, DevForce WinClient derives a display name from the property path so that all elements of the ViewDescriptor are complete. It is as if the model looked like this:

To create the binding descriptor using an explicit reference to a ViewDescriptor, you would instantiate your ViewDescriptor and use a different overload of the ControlBindingDescriptor constructor:

C#

ViewDescriptor aViewDescriptor = new ViewDescriptor(typeof(Employee), "LastName", aDataConverter); ControlBindingDescriptor aControlBindingDescriptor = new ControlBindingDescriptor(mLastNameTextBox, aViewDescriptor); | | Control ViewDescriptor Dim aViewDescriptor As New ViewDescriptor(GetType(Employee), "LastName", aDataConverter) Dim aControlBindingDescriptor As New ControlBindingDescriptor( _ mLastNameTextBox, aViewDescriptor) | | Control ViewDescriptor

VB

The ViewDescriptor constructor used above does not call for the display name, but other overloads are available that do, and the DisplayName property, along with others, can of course be set after the ViewDescriptor is instantiated.

Formatting and Parsing
The careful reader observes that the binding instruction specifies the name of a property of a data source object type (the “FirstName” property of an Employee) and the name of a property of a specific UI control (the “Text” property of a TextBox). .NET uses reflection at run-time to discover the actual properties involved. See [Petzold, 262 ff]

411 | P a g e

IdeaBlade DevForce

WinForm User Interfaces

Take note! The compiler is unable to help us with standard .NET data binding because it cannot know until runtime if those properties actually exist for the data source or the UI Control. For now we shall keep with the fiction that we are binding to the properties themselves.

BindingSource - Binding to a List of Objects
Returning to our example, we can load the DataSource of the BindingSource with Employee objects and redefine the binding like so:
myTextBox.DataBindings.Add("Text", myBindingSource, “FirstName”)

CurrencyManager – Positioning within the list
At any given moment, all such data bindings channel data from a single data object within that DataSource, the object that the BindingSource says is Current. A typical UI would have some mechanism for moving the position up and down within the list. The .NET BindingNavigator is a convenient choice. We can bind it to the same BindingSource:
myNavigator.BindingSource = myBindingSource

Clicking the buttons moves us an object at a time within the Employees of the DataSource.

Format and Parse
The Format event is fired in two ways: (1) when the data object becomes current so that the binding can initialize the control and (2) when there is a change to the bound property of the current data object (we‟ll see how that works shortly). Each Format event handler receives a value from the bound data object property, formats the value, and passes the formatted value on to the waiting control. The Parse event is fired after user input. From time to time we have discovered specific UI controls that do not fire the event properly. We do our best to insulate you from such aberrant behaviors. when the user changes focus to another control. The Parse event receives data from the control, parses it, and passes the parsed value on to the data object property if all went well. The following diagram summarizes these flows.

Data Binding v. Persistence
In most UIs we bind controls to business objects. It's important to remember that changing an in-memory business object and updating the persistent data source are different operations.

412 | P a g e

IdeaBlade DevForce

WinForm User Interfaces

When the user changes the employee‟s first name, data binding will update the employee entity‟s FirstName property. Neither .NET nor DevForce WinClient update the corresponding “FirstName” column for that employee‟s row in the Employee database table. Not automatically. The persistent data source remains ignorant of all our inmemory changes until we tell the DevForce WinClient EntityManager to save the modified entity.

Nested Property Paths
We can traverse a business object graph by navigating along the path of its relations to adjacent business objects as we learned in the section “Business Objects and Persistence.” The expression anEmployee.HomeAddress.City begins with an employee object, emp, and follows its HomeAddress relation property to an address object before invoking the City property of that address. The syntax is as graceful as anEmployee.FirstName which invokes a property of the emp object proper. We refer generally to an expression such as anEmployee.HomeAddress.City as a “nested property path” and to City as a “nested property.”

Formatting
Formatting is the process by which data is transformed on its way from a datasource to a UI control.

Datasource

Formatting

UI Control

We introduced this topic in the last chapter. Let‟s revisit it again to understand the problem and learn how DevForce WinClient solves it with data converters.

Formatting in Native .NET
In native .NET, some formatting can be done in the constructors for a System.Windows.Form.Binding. In the statement below, the date formatting string “d” is passed into the Binding‟s constructor: C# VB
Binding aBinding = new Binding("Value", mEmployeesBS, "BirthDate", true, DataSourceUpdateMode.OnValidation, System.DateTime.Today, "d"); Dim aBinding As New Binding("Value", mEmployeesBS, "BirthDate", _ True, DataSourceUpdateMode.OnValidation, System.DateTime.Today, "d")

The signature of that particular overload is as follows:

C#

public Binding(string propertyName, object dataSource, string dataMember, bool formattingEnabled, System.Windows.Forms.DataSourceUpdateMode dataSourceUpdateMode, object nullValue, string formatString) Public Sub New(ByVal propertyName As String, ByVal dataSource As Object, ByVal dataMember As String, ByVal formattingEnabled As Boolean, ByVal dataSourceUpdateMode As System.Windows.Forms.DataSourceUpdateMode, ByVal nullValue As Object, ByVal formatString As String)

VB

Another overload adds a formatInfo parameter that takes a System.IFormatProvider:

413 | P a g e

IdeaBlade DevForce

WinForm User Interfaces

C#

public Binding(string propertyName, object dataSource, string dataMember, bool formattingEnabled, System.Windows.Forms.DataSourceUpdateMode dataSourceUpdateMode, object nullValue, string formatString, System.IFormatProvider formatInfo) Public Sub New(ByVal propertyName As String, ByVal dataSource As Object, ByVal dataMember As String, ByVal formattingEnabled As Boolean, ByVal dataSourceUpdateMode As System.Windows.Forms.DataSourceUpdateMode, ByVal nullValue As Object, ByVal formatString As String, ByVal formatInfo As System.IFormatProvider)

VB

The IFormatProvider provides a mechanism for retrieving an object to control formatting. Binding also provides a Format event that can be handled. The handler can perform just about any desired transformation of the incoming data:

C#

public void SetBindings() { Binding aBinding = new Binding("Text", mProductsBS, "UnitPrice"); aBinding.Format += new System.Windows.Forms.ConvertEventHandler(DecimalToCurrencyString); mUnitPriceTextBox.DataBindings.Add(aBinding); } private void DecimalToCurrencyString(object sender, System.Windows.Forms.ConvertEventArgs e) { if (e.DesiredType == typeof(string)) { e.Value = (System.Convert.ToDecimal(e.Value)).ToString("c"); } }

VB

Public Sub SetBindings () Dim aBinding As New Binding("Text", mProductsBS, "UnitPrice") AddHandler aBinding.Format, AddressOf DecimalToCurrencyString mUnitPriceTextBox.DataBindings.Add(aBinding) End Sub Private Sub DecimalToCurrencyString(ByVal sender As Object, ByVal e As System.Windows.Forms.ConvertEventArgs) If e.DesiredType Is GetType(String) Then e.Value = (CDec(e.Value)).ToString("c") End If End Sub

Formatting with DevForce WinClient DataConverters
DevForce WinClient DataConverters provide several different options for performing required formatting on a property value. At the most basic and straightforward end, converters for DateTime and Numeric types 132 provide a FormatString property that uses the formatting strings pre-defined in .NET. Table 17 shows the .NET formatting strings for DateTime types; Table 18 shows the .NET formatting strings for numeric types. In addition to the formatting strings shown in the table, you can define custom formatting strings for both DateTimes and numerics.
Table 17. Formatting Strings for DateTime Types (from the .NET Framework Help File) Format specifier Name Description

132

DateConverter, DateRangeConverter, NumericConverter, and NumericRangeConverter

414 | P a g e

IdeaBlade DevForce

WinForm User Interfaces

d

Short date pattern Long date pattern Short time pattern Long time pattern Full date/time pattern (short time) Full date/time pattern (long time) General date/time pattern (short time) General date/time pattern (long time) Month day pattern RFC1123 pattern

Displays a pattern defined by the DateTimeFormatInfo.ShortDatePattern property associated with the current thread or by a specified format provider. Displays a pattern defined by the DateTimeFormatInfo.LongDatePattern property associated with the current thread or by a specified format provider. Displays a pattern defined by the DateTimeFormatInfo.ShortTimePattern property associated with the current thread or by a specified format provider. Displays a pattern defined by the DateTimeFormatInfo.LongTimePattern property associated with the current thread or by a specified format provider. Displays a combination of the long date and short time patterns, separated by a space.

D

t

T

f

F

Displays a pattern defined by the DateTimeFormatInfo.FullDateTimePattern property associated with the current thread or by a specified format provider.

g

Displays a combination of the short date and short time patterns, separated by a space.

G

Displays a combination of the short date and long time patterns, separated by a space.

M or m

Displays a pattern defined by the DateTimeFormatInfo.MonthDayPattern property associated with the current thread or by a specified format provider. Displays a pattern defined by the DateTimeFormatInfo.RFC1123Pattern property associated with the current thread or by a specified format provider. This is a defined standard and the property is read-only; therefore, it is always the same regardless of the culture used, or the format provider supplied. The property references the CultureInfo.InvariantCulture property and follows the custom pattern "ddd, dd MMM yyyy HH:mm:ss G\MT". Note that the 'M' in "GMT" needs an escape character so it is not interpreted. Formatting does not modify the value of the DateTime; therefore, you must adjust the value to GMT before formatting. Displays a pattern defined by the DateTimeFormatInfo.SortableDateTimePattern property associated with the current thread or by a specified format provider. The property references the CultureInfo.InvariantCulture property, and the format follows the custom pattern "yyyy-MM-ddTHH:mm:ss".

R or r

s

Sortable date/time pattern; conforms to ISO 8601 Universal sortable date/time pattern

u

Displays a pattern defined by the DateTimeFormatInfo.UniversalSortableDateTimePattern property associated with the current thread or by a specified format provider. Because it is a defined standard and the property is read-only, the pattern is always the same regardless of culture or format provider. Formatting follows the custom pattern "yyyy-MM-dd HH:mm:ssZ". No time zone conversion is done when the date and time is formatted; therefore, convert a local date and time to universal time before using this format specifier. Displays a pattern defined by the DateTimeFormatInfo.FullDateTimePattern property associated with the current thread or by a specified format provider. The time displayed is the universal time, rather than the local time, equivalent to the DateTime value.

U

Universal sortable date/time pattern Year month pattern Unknown specifier

Y or y

Displays a pattern defined by the DateTimeFormatInfo.YearMonthPattern property associated with the current thread or by a specified format provider.

Any other single character

Table 18. Formatting Strings for Numeric Types (from the .NET Framework Help File) Format Name Description

415 | P a g e

IdeaBlade DevForce

WinForm User Interfaces

specifier C or c Currency The number is converted to a string that represents a currency amount. The conversion is controlled by the currency format information of the NumberFormatInfo object used to format the number. The precision specifier indicates the desired number of decimal places. If the precision specifier is omitted, the default currency precision given by the NumberFormatInfo is used. This format is supported for integral types only. The number is converted to a string of decimal digits (0-9), prefixed by a minus sign if the number is negative. The precision specifier indicates the minimum number of digits desired in the resulting string. If required, the number is padded with zeros to its left to produce the number of digits given by the precision specifier. The number is converted to a string of the form "-d.ddd…E+ddd" or "-d.ddd…e+ddd", where each 'd' indicates a digit (0-9). The string starts with a minus sign if the number is negative. One digit always precedes the decimal point. The precision specifier indicates the desired number of digits after the decimal point. If the precision specifier is omitted, a default of six digits after the decimal point is used. The case of the format specifier indicates whether to prefix the exponent with an 'E' or an 'e'. The exponent always consists of a plus or minus sign and a minimum of three digits. The exponent is padded with zeros to meet this minimum, if required. The number is converted to a string of the form "-ddd.ddd…" where each 'd' indicates a digit (0-9). The string starts with a minus sign if the number is negative. The precision specifier indicates the desired number of decimal places. If the precision specifier is omitted, the default numeric precision given by the NumberFormatInfo is used. The number is converted to the most compact of either fixed-point or scientific notation, depending on the type of the number and whether a precision specifier is present. If the precision specifier is omitted or zero, the type of the number determines the default precision, as indicated by the following list. Byte or SByte: 3 Int16 or UInt16: 5 Int32 or UInt32: 10 Int64 or UInt64: 19 Single: 7 Double: 15 Decimal: 29 Fixed-point notation is used if the exponent that would result from expressing the number in scientific notation is greater than -5 and less than the precision specifier; otherwise, scientific notation is used. The result contains a decimal p