You are on page 1of 22

A Shallow Introduction to the K Programming Language || kuro5hin.


create account | help/FAQ | contact | links | search | IRC | site news Everything Diaries Technology Science Culture Politics Media News Internet Op-Ed Fiction Meta MLP We need your support: buy an ad | premium membership

A Shallow Introduction to the K Programming Language
By jjayson in Technology Thu Nov 14, 2002 at 05:58:07 AM EST Tags: Software (all tags) About two years ago I was introduced to a programming language that I really didn't like: it didn't have continuations, I didn't see any objects, it had too many operators, it didn't have a large community around it, it was strange and different, and it looked like line noise, like Perl, and I don't like Perl. However, I gave it a try. I had to learn that continuations may not be there, but first-class functions are; it may not have a normal object system, but that is because the language doesn't need it and gets it power by cutting across objects; all the operators are the functions that make up its standard library; it's community may not be large, but it is incredibly intelligent; it only looks strange until you understand its concepts; and well, it will always look like line noise, but you will stop caring because this also make the concise code easier to read. K has since become my language of choice.


Managed Hosting VoxCAST Content Delivery Raw Infrastructure

Login Introduction and Terminology K is a high level system programming language available for Windows, Solaris, Linux, and maybe a few other Unix flavors upon request. It is mixture of APL, a functional programming language, and many unique features not found elsewhere. The K system is produced by Kx. They produce the language and one of the fastest commercial relational database available, KDB — written in K. You can get copies of the interactive interpreter free for educational, research, or toy purposes, and there is a non-interactive, runtime-only interpreter that is strictly free. I have heard rumors that there would be interest in opening up the language if there would be enough community response around it. Contrary to some opinion, K is a real-world language, made for real use and sold as a real product. It is not a toy language. It is not a Turning tarpit language. It is not intentionally obfuscated or made to be intentionally strange, like Befunge or Intercal. K was originally developed by Arthur Whitney, who was a very Make a new account
Username: Password:

Note: You must accept a cookie to log in.

Advertising by Blogads
Advertise here

Poll Would you try K? Yes No

51% 25%

If if didn't look so ugly 22% Votes: 108

1 of 22

A Shallow Introduction to the K Programming Language ||

influential member of the APL community. Before K, Arthur was asked to write the Dr. Iverson's APL-successor, the J language, but he decided to go his own way. This isn't his first language either; he is also the creator of the Morgan Stanley financial language A+. From 1994 to 1997 the Union Bank of Switzerland purchased exclusive rights for the use of K, but now it is available to all of us. One of the large draws of K is the extreme programmer productivity it offers, its incredibly fast execution speeds, and its very small executable units. One of the proofs of these claims is Brian Kernighan's Experiments with Scripting and User-Interface Languages, in which K easily outperforms other favorite languages. Another example of these claims met is how KDB outperforms Oracle on TPC benchmarks in both query speed and data storage size while essentially being written by one person. Also, K code is very dense, and it is typical to see 100:1 code size reduction when migrating to from C to K. I have heard of almost a 1000:1 reduction when a project moved from Java and SQL to K and KSQL (KDB's query language). K is an exceptional language for dealing with mathematical analysis, financial prediction, or anything that handles bulk data. I have used it for a document search engine and other computational linguistic tasks. K doesn't work very well when the problem is not vectorizable and scalars are the largest datum to operate on. Often though, algorithms can be written in either method. However, some problems have been almost impossible to vectorize, such as Ackermann's function, and are not good to attempt in K. K has bindings to other popular language such as C, Java, VisualBasic, and Excel. There has also been work done on bindings to Python and Mozilla's XUL. K's builtin interprocess communication and binary format for objects is very simple and documented so making other systems interact with K is often equally simple. Even though K is an interpreted language, the source code is somewhat compiled internally, but not into abstract machine code like Java or Python. The interpreter can be invoked interactively or non-interactively, and you can also compile a source file down to a binary file for product distribution, if you wish. One of the hardest things for many people to get over at first is the way K looks. Even the strongest K enthusiast will freely admit that K tends to look like line noise. I came from a Scheme background and already had my eyes hurt a couple times trying to learn Perl. When my roommate introduced me to K I ridiculed it and couldn't imagine ever wanting to use it. However, after a few sessions, my eyes began to relax. Unlike Perl, the syntax is extremely regular and there is almost no syntactic sugar or special cases. K is even translatable into English; there is a program that will take a K expression and produce its English translation. To allow this translation each

Results | Other Polls Related Links Scoop available for Windows, Solaris, Linux Kx KDB Arthur Whitney A+ Experiment s with Scripting and User-Interface Languages K easily outperforms other favorite languages Ackermann' s function English translation Nobody Expects the Spanish Inquisition Also by jjayson

2 of 22

A Shallow Introduction to the K Programming Language ||

operator is given is a short English name and sometimes small combinations of operators are also given common names. This is similar to building a vocabulary in a natural language. You will begin to look at larger segments of code and abstract away the actual details of what happens to the data, while understanding the higher-level concepts and transformations more easily. To make even stronger connections with linguistics, grammatical terms are used to describe K. Operators are called verbs, and data is called nouns. There are also operators that modify other operators (these will be described later) that are called adverbs. Stringing some nouns, verbs, and adverbs together will produce clauses and sentences. This dialogue has been inherited from APL and is often abandoned for the more commonplace names of operators, functions, and variables. Comments and Conventions When small code segments are presented, I will use the conventions of the language and interpreter for comments and display. Forward-slash (/) is special when it comes at the beginning of a line or it has space to the left of it, then it starts a comment that lasts until the end of the line. Some people will also disguise comments as strings, since that will not effect the results of a computation. The interpreter prompts the user with two spaces to signify that it is ready to accept input. It will print all of its output without any leading space. I will do the same, showing interpreter input preceded by two spaces and interpreter output unindented. K tends to favor simplicity over sugar. One thing that may confused people easly is K has no precedence rules. Everything is parsed from right to left. This makes it easy to see what happens next in a computation, but it also takes a little getting used to. For example, 3*2+1 in K will produce 9, instead of the more usual 7 in other language. This is always the case, except when a set of parentheses is come across. What ever inside the parentheses is first evaluated then normal evaluation resumes. People will frequently play with the order of their statements to reduce the number of characters in the expression, instead of placing parentheses around it. Values There are four simple types of values in K -- integer, floating point, character, and symbol; a special null type for the singular value of _n; and two composite types -- dictionaries and lists. The composite types are containers that others values may be stored and retrieved from. Lists are further classified as either homogenous or heterogeneous, and all homogenous lists carry the type of value they contain around with them. A list where all the simple elements are of the same type are called vectors. K is optimized to handle these homogenous lists and vectors since they appear in programming so frequently.

3 of 22

A Shallow Introduction to the K Programming Language ||

Lists are represented in different ways. The most general case is surrounded by parentheses with each element separated by a semi-colon. If the list is homogenous the parentheses and semi-colons will be elided and you will simply see the elements separated by a space. If all the elements are characters then the lists will be a string surrounded by double-quotes. To index a list you use brackets ([]) with multiple indices allowed. To index a multidimentional list use a semi-colon to separate the dimensions, but leaving a dimension blank selects all (this the same as using _n as the index). A symbol is an interned string that the interpreter uses for variable lookup. They are created by the backtick followed by any legal variable name: a letter or underscore followed by any combination of letters, numbers, underscores, and dots -- but no more than two consecutive dots (since they have a special meaning that will re explained later). If you would like the symbol to not follow a legal variable name patters, you may enclose it in double quotes. Examples:
(1; 2.3; 4; 5.6) 1 2 3 4 5 1.2 3.4 5.6 "quack" (1 2; 3.4 5.6; "meow") (1;2 3;4 5 6) / / / / / / heterogeneous list of integers and floats homogeneous list of integers homogeneous list of floats homogeneous list of characters a list of lists a vector of integers

"abcdefghijlkmnopqrstuvwxyz"[14 8 13 10] "oink" ("qwerty";"poiuy";"asdf";"jhgfdsa")[;3] / slicing or projection "ruff" `rusty `""

Verbs K is a very small language, so it needs to make the most of everything. Unlike APL, It uses only ASCII, so it already starts off with not too many characters to use and abuse. To allow for compact code operators are overloaded with two cases: a monadic (one argument) and diadic (two argument) use. Sometimes these uses are related and sometimes they are not. And each case the operator sometimes has slightly different behavior depending on the type or domain of the arguments. Examples:
!4 0 1 2 3 5!3 2 2!1 2 3 / rotate: spins right back-to-front left number of positions. 3 1 2 ,2 ,2 1,2 / join: forms one list of the left and right argument / enlist: a one item list containing only 2 / enumerate: a list of integers from 0 to x-1 / mod: the residue of the left modulus the right

4 of 22

A Shallow Introduction to the K Programming Language ||

1 2 |1 2 3 / reverse: reverses a list 3 2 1 0|1 1 &1 2 3 4 / where: returns the number of units specified 0 1 1 2 2 2 3 3 3 3 &0 1 1 0 0 1 / where: an important use to get the indices of the 1s 1 2 5 0&1 0 4<5 / less-than: predicate of "is the left smaller than the right" 1 <7 4 9 / grade-up: sorts indices in ascending order 1 0 2 =1 0 1 0 0 3 0 1 3 1 / group: groups all indices of same value (0 2 7 9 1 3 4 6 5 8) 2=0 1 2 3 4 0 0 1 0 0 ?1 0 1 0 0 3 0 1 3 1 1 0 3 5 / equals: compares values / unique: all unique elements in order seen / min: the minimum also boolean AND / max: the maximum also boolean OR

1 0 1 0 0 3 0 1 3 1?3 / find: the first indice of the right in left

Variables and Bindings K is a dynamically, strongly typed language and, variables are not declared, but they come into existence when you assign a value to it. This can be done anywhere, even in the middle of an expression since there is no distinction between statements or expressions. If you try to read a value from a variable that has not yet been assigned to, you will raise an error. There are also no pointers. In true functional style, when you assign to a variable a deep copy of the value is made (K does this lazily, though). Assignment is done via the colon and it is read as "gets" or "is." As a special case, when an assignment is the last thing in an expression, null is returned (this helps prevent cluttering up the display log). You can force the return of a value from an assignment statement by using a case of the monadic colon. Examples:
a:"moo" b:!10 :c:b effect c 0 1 2 3 4 5 6 7 / a gets the string "moo" / b gets enumerate 10 (integer list from 0 to 9) / c gets the value of b, but changes to b do not 8 9

:h:(g*2),g:1+2 / h get g times 2 join g, where g gets 1 plus 2 6 3 g 3

User-defined Functions Braces ({}) are used to create functions; they are the

5 of 22

A Shallow Introduction to the K Programming Language ||

equivalent of lambda in Lisp. Often they are used then the resulting function is assigned to a variable, but sometimes not. To assist in making the code compact, if a function requires three or fewer arguments, K will allow you to implicitly use x, y, and z as the arguments. If you need more than three arguments you must declare them all. All functions return the value of the last executed statement, even if that statement return null. To call a function you use brackets (just like a list index). If you do not supply an argument when calling a function, it will project. Examples:
pyth:{_sqrt(x*x)+y*y} / notice the two implicit arguments pyth[30;40] 50.0 pyth[3 15;4 20] 5 25.0 / cute huh.

dist:{[x1;y1;x2;y2] _sqrt((x2-x1)^2)+(y2-y1)^2} dist[1;1;4;5] 5.0 :d:dist[1;1] / project or curry the first two arguments {[x1;y1;x2;y2] _sqrt((x2-x1)^2)+(y2-y1)^2}[1;1] d[7;9] 10.0 :e:dist[1;;4] / project on first and third argument {[x1;y1;x2;y2] _sqrt((x2-x1)^2)+(y2-y1)^2}[1;;4] e[2;6] 5.0 inc:1+ inc 8 9

System Functions and Variables After running out of punctuation Arthur made system function. Every symbol beginning with an underscore is reserved for either a system variable or system function. System functions use infix, like their less readable cousins, but like user defined functions they cannot be overloaded with monadic and dyadic cases (in the next version of K this will be changed and users will be able to define infix functions and overload them with n-adic cases). Examples:
3_draw 5 2 2 4 / list of 3 random numbers from 0 to 4

2_draw 0 / list of 2 random real numbers from 0 to 1 0.2232866 0.9504653 4_draw-4 0 to 3 2 0 1 3 4 (29 51 39 33 / deal: list of 4 random nonrepeating numbers from

13_draw-52 / deal a deck of cards into four piles 27 10 0 23 3 28 5 24 16 40 8 22 20 36 47 18 31 26 11 44 37 38 9 13 42 34 50 21 6 19 46 48 45 14 43 2 49 4 25 41 30 35 7 32 17 1 12 15) / binary search through list returning index

1 3 4 5 7 9_bin 4

6 of 22

A Shallow Introduction to the K Programming Language ||

2 1 3 4 5 7 9_binl 2 4 6 / binary seach for a list of numbers 1 2 4 16_vs 543212 8 4 9 14 12 5 3 2_vs 21 3 1 1 5 3 2_sv 3 1 1 21 _host` -815566008 / returns ip address as integer / vector from scalar: changes base to 16 / also does variable change of base / scalar from vector: the inverse

256_vs _host` / presentation form 207 99 115 72

Adverbs This is where K starts to set itself from apart from most of the common programming languages in use today. You rarely write loops in K (KDB is 100% loop-free), instead you use adverbs. An adverb modifies a function, returning another function, changing the ways it operates over its arguments and what it does with it's return values. Here is a small selection of adverbs' usages (there are other uses that are not covered here). Over (/) modifies a diadic function and will apply the function down a list, collection the result. Converge (/) modified a monadic function and will continually apply the function to the previous result until either the initial value or the result of the preceding value is returned. Scan (\) will apply the function down a list, collection all intermediate results (this is sometimes called trace). There is a trace analog to all usages of over. Each (') will apply the function down lists of the same length (equal to the valence of the function). Each-right (/:) will hold the left argument of the function and apply the function down the list of right arguments. Examples:
+/1 2 3 4 10 +\1 2 3 4 +/ 1 3 6 10 |/5 3 7 4 2 7 ,/(1 2;(3 4;5);6) (1;2;3 4;5;6) / join-over: (1 2),(3 4;5),6 / plus-scan: returns the intermediate values of / plus-over (sum): is similar to 1+2+3+4

/ max-over: compares all items like 5|3|7|4|2

,//(1 2;(3 4;5);6) / flatten: explained below 1 2 3 4 5 6 3 4_draw'-3 -4 (1 2 0 2 0 1 3) 2_draw/:10 100 (7 7 45 91) / draw-each: (3_draw-3),(4_draw-4)

/ draw-right-each: (2_draw 10),(2_draw 100)

7 of 22

A Shallow Introduction to the K Programming Language ||

Let me break down flatten for you. Join is a function that takes two values. Over modifies join and the result is a function that now takes one argument (a list) and joins all the elements of that list. Looking at the above example, when join was given a nested list, it simple stitched all the elements into a single list, removing one level of depth. However, we would like to remove all level of depth. We want join-over again to remove another level of depth, and join-over again to remove another level of depth, until the list is completely flat. This is what converge will do for us. The second application of over takes this monadic function, join-over, and continually applies it until the result no longer changes. If we didn't want to flatten the topmost list, but instead we had a list of lists to flatten, we would only want to apply flatten to each element of the top-level list. That is what flatten-each (,//') would do. If we wanted to keep the top two levels of depth we would write flatten-each-each (,//'') and this is where things start too look like line noise. Conditionals Although rarely used, there are a few conditionals; most often used is the colon. It is similar to cond in Lisp: it takes pairs of arguments and an optional final argument. The first argument of each pair is tested for truth (0 is false, all other integers are true, anything besides an integer is an error). If it is true then the result of evaluating the second of the pair is returned. If it is false then the next pair is tested. If all the pairs have been exhausted, then the final argument is evaluated and the result returned. If there is no final argument and all the conditions are false, then null is returned. Examples:
:[0;"true";"false"] "false" s:{:[x>0;"+"; x<0;"-"; "0"]} / returns the sign of x or 0 s 4 "+" s -3 "-3"

Naive Primality Test As in Nobody Expects the Spanish Inquisition, I will conclude the basics with a naive primality predicate and try to explain it.
isprime:{&/x!/:2_!x} x isprime 14 0 isprime 7 1 / min over x mod right-each 2 drop enumerate

Analyzing from right to left. We create a list of all integers from 0 to x (exclusive), then we remove the first two elements (2_), so we are left with a list from 2 to x (exclusive). Next we determine the residue of x and each of the numers in the list. Finally, we calculate the minimum of the residues. If the number is prime, then the lowest residue will be 1 and be

8 of 22

A Shallow Introduction to the K Programming Language ||

considered true. If the number was composite there will be a 0 residue for some value.
!14 0 1 2 3 4 5 6 7 8 9 10 11 12 13 2_!14 2 3 4 5 6 7 8 9 10 11 12 13 14!/:2_!14 0 2 2 4 2 0 6 5 4 3 2 1 &/14!/:2_!14 0

The K-tree Here is a quick, small look at one of the unique elements of K, called the K-tree. It falls nicely into the K philosophy of keeping things simple and powerful. The K-tree can be used for modularization of programs, as a scoping mechanism, for GUI design, and as a rudimentary object system. All variables exist somewhere on the K-tree. To reference a variable you separate the name of the variable and each branch name with periods. The root of the tree is an empty symbol. For example, you might have a fully qualified variable named .branch.subbranch.variable. A branch is really just a dictionary. If I were to assign a dictionary that contained symbols sym and bol to the variable then you would be able to access and It goes the other way, too. If you were to create the variables .dict.ion and .dict.ary then the variable .dict would be a valid dictionary. This makes the language very reflective since you can now modify variables locations, scopes, and manually manipulate extents. Whenever you are in the K environment you are always running in a branch of the K-tree. When the K interpreter starts it places you in the .k branch. You can move around the tree with the directory (\d) command. For example \d .tw.ig would create the new branch tw then create a subbranch ig. You can inspect all the variables in a branch with the list variables (\v) command. Often a script will begin with a change directory command and then define all of its variables in that (and maybe a few more) branches. This effectively uses the K-tree as a module system.
\d .test \d ement \d .test.ement new:`small old:`big \v new old \d ^ \d .test \v ement / create a new directory off the root / create a sub-branch / show the current directory / put some values in the directory / list the contents o / back up one directory in the tree

9 of 22

A Shallow Introduction to the K Programming Language ||

ement .((`new;`small;) (`old;`big;)) `small ement.old `big ement`old `big

/ inspect the value of ement

/ fully-qualified / partially qualified / index like an array

Other Unique Elements of K The K philosophy itself has some unique insights on how to treat data and the appropriate way to organize data for efficient processing, however within the language there are some features that you will not find in any other language: K uses the tree to hold an attribute structure. These attributes are used for documentation strings, GUI representations, for other functionality in K, and for whatever you decide to use them for. Some of the programming environments for K make entensive use of attributes to store information about where functions are defined and other administrative information. K also has a the concept of dependencies and triggers. They are used for efficient, demand-based calculation of data and executing callback code whenever a value is changes (this is how the timer system in K works). K will keep track of out of date dependencies for you and only do the minimal amount of work necessary to update values. K has a unique, declarative GUI subsystem that is based around the K tree and triggers. It takes a minimal amount of work to put a GUI on your application. K takes the approach that you show variables and changes made to that variable on the screen are immediately reflected in the program without any work by you. K's interprocess communication (IPC) and network communication systems is also based around callbacks and message handlers. There are simple K primitives that will ship entire K data structures around for you, or you may do it yourself. The goal of the IPC system is like the goal with rest of K, make it fast, simple, powerful, and highly useful. In a time when programming languages are lacking in originality and are not bringing new ideas to the table, K succeeds where others fail. But K is not just a research language, not appropriate for real-world use. While the community may be small some of the users of the language are very big. With recent implementations by Island ECN and the US government, this looks to only be getting bigger, too. The next version of the language will fix many of the nagging holes and annoyances and take away the line noise factor that has pushed many away. Here are some closing simple segments of K that are

10 of 22

A Shallow Introduction to the K Programming Language ||

informative on how the K way of approaching a problem may be different:
cut:{1_'(&x=*x)_ x:" ",x} cut "Scoop ate my spaces" ("Scoop" "ate" "my" "spaces") fibonacci:{x(|+\)\1 1} fibonacci 5 (1 1 2 1 3 2 5 3 8 5 13 8) first:*: euclid:{first(|{y!x}\)\x,y} / Euclid's Algorithm euclid[24;40] (24 40 16 24 8 16 0 8)



A Shallow Introduction to the K Programming Language | 162 comments (156 topical, 6 editorial, 0 hidden) Interested about deep copying (4.25 / 4) (#6) by xriso on Thu Nov 14, 2002 at 04:08:45 AM EST If it is a completely data-oriented language (i.e. no pointers, references, etc.), does this mean that GC can be completely done using reference counts when a lazy copying system is used? This would seem to get over the problem of excessive memory copying (something I hate oh so much), yet also prevent having to deal with memory management. A very interesting language, it seems! +1F -*** Quits: xriso:#kuro5hin (Forever) exactly by jjayson, 11/14/2002 04:50:17 AM EST (5.00 / 2) Interesting question, actually by Simon Kinahan, 11/14/2002 06:34:38 AM EST (5.00 / 3) No iraq, +1 (3.00 / 2) (#8) by dreancha on Thu Nov 14, 2002 at 04:21:14 AM EST ... and you get extra credit for mentioning befunge :)

+1 FP (2.50 / 2) (#9) by dvchaos on Thu Nov 14, 2002 at 04:33:25 AM EST It's a well written 'clean' article. and is something different too. - anonymous proxy server!

11 of 22

A Shallow Introduction to the K Programming Language ||

Perl (2.50 / 2) (#11) by carbon on Thu Nov 14, 2002 at 05:05:40 AM EST Unlike Perl, the syntax is extremely regular and there is almost no syntactic sugar or special cases. Actually, Perl5 has quite regular syntax: it just has a lot of a) special variables, all of which are listed and follow the pattern of having a single non-alphanumeric identifier, and b) having extremely weird syntax. Perl is one of my personal favorite languages, right up there with C++, so I suppose this is just my bias speaking. :-) Is there anything in particular you found contradictory about Perl's syntax?

Wasn't Dr. Claus the bad guy on Inspector Gadget? - dirvish no it doesn't by jjayson, 11/14/2002 05:19:26 AM EST (2.33 / 3) Bah by carbon, 11/14/2002 06:05:29 AM EST (none / 0) Bah? by mishmash, 11/14/2002 08:26:20 AM EST (none / 0) Funny by carbon, 11/14/2002 09:41:32 PM EST (none / 0) and my dad could beat your dad... by sholden, 11/14/2002 07:26:05 AM EST (none / 0) umm... by pb, 11/14/2002 02:59:56 PM EST (none / 0) !(descended -> baggage) by sholden, 11/14/2002 07:43:38 PM EST (none / 0) That is kind of what Peter is saying... by jjayson, 11/15/2002 02:17:01 PM EST (none / 0) It did by sholden, 11/15/2002 06:04:50 PM EST (none / 0) Design purity by mishmash, 11/14/2002 10:16:10 AM EST (4.00 / 1) Arrogance by yooden, 11/14/2002 11:01:57 AM EST (none / 0) Its true by Peaker, 11/14/2002 11:54:38 AM EST (none / 0) So? by bodrius, 11/14/2002 02:40:43 PM EST (none / 0) A note by carbon, 11/14/2002 09:49:08 PM EST (none / 0) A Follow-up Note by yooden, 11/15/2002 03:54:44 AM EST (none / 0) Hey now by ph317, 11/14/2002 02:36:23 PM EST (4.00 / 1) well... by tin, 11/15/2002 11:11:21 PM EST (none / 0) perl and c++ by jefu, 11/14/2002 12:03:25 PM EST (none / 0) my three cents by skyknight, 11/14/2002 01:54:35 PM EST (none / 0) I keep hearing this, and it's bullshit by tzanger, 11/14/2002 04:00:59 PM EST (5.00 / 3) right only? by jefu, 11/14/2002 08:05:09 PM EST (none / 0) oops by jefu, 11/14/2002 08:32:12 PM EST (none / 0) You have some good points by tzanger, 11/14/2002 09:03:05 PM EST (none / 0) Eh? Bullshit? by skyknight, 11/15/2002 05:18:53 PM EST (none / 0) Discipline: that was my whole point by tzanger, 11/15/2002 07:27:13 PM EST (none / 0) Analogy by skyknight, 11/15/2002 07:50:12 PM EST (4.00 / 1) Auto analogies never work right by tzanger, 11/16/2002 10:55:47 AM EST (none / 0) Examples, please. by Shpongle Spore, 11/15/2002 06:28:36 PM EST (none / 0) Overload +? by der, 11/14/2002 02:43:57 PM EST (5.00 / 1) Or like Java... by ekj, 11/20/2002 12:17:26 PM EST (none / 0) overloading "+" by Shpongle Spore, 11/15/2002 06:42:43 PM EST (none / 0) much ado over loading by jefu, 11/20/2002 09:19:35 PM EST (none / 0) What's good enough? by Shpongle Spore, 12/04/2002 02:27:49 PM EST (none / 0) perl is driven by natural language constructs by Dr. Zowie, 11/14/2002 03:15:52 PM EST (5.00 / 1)

12 of 22

A Shallow Introduction to the K Programming Language ||

Ah - regular expressions. by static, 11/14/2002 06:07:57 PM EST (none / 0) some contradictions in the article though... by tin, 11/15/2002 10:34:46 PM EST (none / 0) +1 FP (4.00 / 1) (#13) by daragh on Thu Nov 14, 2002 at 05:37:43 AM EST Great article, well written. I love hearing about esoteric and new programming languages. I just have to persuade the powers that be to let me use them... No work. Nice article, some questions. (5.00 / 2) (#14) by a2800276 on Thu Nov 14, 2002 at 05:48:53 AM EST I came from a Scheme background Somehow, I knew this instinctively from the point that I'd finished reading the first couple of sentences :) So, if you don't mind my asking: What exactly is a continuation? I've looked around, but haven't found anything in google or foldoc that satisfies my curiosity. I take it that it's a rather academic description of a language feature. Also, and this may sound trollish but isn't meant to be: what is APL still used for except in really archaic systems?
&1 2 3 4 / where: returns the number of units specified 0 1 1 2 2 2 3 3 3 3 &0 1 1 0 0 1 / where: an important use to get the indices of the 1s 1 2 5

I don't think I understand the concept of the "where" function. What does it do? In the second example it lists the positions in the list that contain 1s, but in the first example? Ok, I think I just figured it out: Each index is listed x number of times where x is the value at the index-th position in the list. I think. You might want to clarify that... I also got kindof lost in the system function section: the first argument gets "attached" (i.e. no spaces) to the function, correct? 1_draw 5. But what is the difference when further attributes are a.) seperated by a space (2_draw 5), seperated by a dash (4_draw-4) and c.) not seperated at all:
1 3 4 5 7 9_binl 2 4 6 / binary seach for a list of numbers 1 2 4

(This one I definately don't understand) All in all great article! Could maybe have profited a little by being cut into pieces and expanded, but ...

spacing issues, _bin, and _binl by jjayson, 11/14/2002 06:21:54 AM EST (5.00 / 3) Ok, I see where I lost you.... by a2800276, 11/14/2002 07:17:46 AM EST (none / 0) Continuations by TuringTest2002, 11/14/2002 06:40:19 AM EST (4.00 / 1) Continuations... by mirleid, 11/14/2002 06:45:39 AM EST (5.00 / 3) a good introduction to continuations in scheme by sesquiped, 11/14/2002 03:44:42 PM EST (5.00 / 1) Proof? (5.00 / 11) (#18) by obsidian head on Thu Nov 14, 2002 at 06:37:22 AM EST

13 of 22

A Shallow Introduction to the K Programming Language ||

One of the proofs of these claims is Brian Kernighan's Experiments with Scripting and User-Interface Languages, in which K easily outperforms other favorite languages. I read the entire thing. Where does it mention K?

Look at the next link by jjayson, 11/14/2002 12:48:17 PM EST (5.00 / 1) More conspiracy by jtra, 11/14/2002 02:05:08 PM EST (5.00 / 2) Ah yes. by pb, 11/14/2002 02:43:50 PM EST (5.00 / 1) All my hard work, stolen! by Cal Bunny, 11/14/2002 02:58:31 PM EST (4.00 / 6) Stop your complaining. by jjayson, 11/14/2002 03:01:13 PM EST (3.66 / 3) Cal Bunny.. by infinitera, 11/14/2002 06:19:50 PM EST (5.00 / 2) min/max and/or (4.50 / 2) (#22) by Karellen on Thu Nov 14, 2002 at 07:21:08 AM EST if | is both a `boolean or' and `max' operator, I can see how that works for `0|1', but how does it work for `0|-1'? Boolean or should be true (non-zero), but the max of those two is zero. How does that work?

Excellent question which reveals the heart of K by WilliamTanksley, 11/14/2002 11:45:54 AM EST (5.00 / 1) _vs returning vectors of the same length by jjayson, 11/14/2002 12:46:24 PM EST (5.00 / 1) only boolean when 0s and 1s by jjayson, 11/14/2002 01:10:36 PM EST (5.00 / 1) I didn't think you meant it was a bitwise op... by Karellen, 11/14/2002 07:34:00 PM EST (none / 0) It is far simplier than that (and not-equal) by jjayson, 11/14/2002 08:05:11 PM EST (none / 0) You meant `happen to mimic or and and', right? by Karellen, 11/15/2002 05:14:27 PM EST (none / 0) you have the right idea. by jjayson, 11/15/2002 05:33:05 PM EST (none / 0) thanks for the arcticle (3.00 / 2) (#24) by mreardon on Thu Nov 14, 2002 at 08:08:53 AM EST You've got me interested. I want to investigate more. Cheers.

Searching by nklatt, 11/14/2002 01:55:21 PM EST (3.66 / 3) It's not Free (3.28 / 7) (#25) by yooden on Thu Nov 14, 2002 at 08:10:44 AM EST

I have heard rumors that there would be interest in opening up the language if there would be enough community response around it. Please come back when it's done. I use a Free OS and Free apps, why would I want to spend time on a non-Free language that may or may not be opened up (whatever that means) sometimes?

14 of 22

A Shallow Introduction to the K Programming Language ||

there... by Work, 11/14/2002 11:40:22 AM EST (4.50 / 4) Your Loss by yooden, 11/14/2002 11:57:38 AM EST (3.85 / 7) Not only that by Work, 11/14/2002 01:30:54 PM EST (2.66 / 3) Nonsese by bodrius, 11/14/2002 02:54:25 PM EST (5.00 / 1) You don't have to talk about Software by yooden, 11/15/2002 04:13:59 AM EST (none / 0) it's pointless because it's low probability. by Shren, 11/14/2002 06:40:43 PM EST (2.00 / 1) GPLnixed=lowprobability by sugarkandimt, 11/14/2002 11:15:38 PM EST (4.00 / 1) Probability by yooden, 11/15/2002 05:10:14 AM EST (none / 0) yes by jjayson, 11/15/2002 05:16:28 AM EST (none / 0) Microsoft vs. GPL by yooden, 11/15/2002 05:27:42 AM EST (4.00 / 1) Reason the GPL has never been tested by pwhysall, 11/15/2002 06:10:30 AM EST (none / 0) because it doesn't need to be 100% open by jjayson, 11/14/2002 01:08:53 PM EST (none / 0) It sure does by yooden, 11/15/2002 05:18:46 AM EST (none / 0) Sounds like... (4.00 / 1) (#27) by IPFreely on Thu Nov 14, 2002 at 08:29:18 AM EST it didn't have continuations, I didn't see any objects, it had too many operators, it didn't have a large community around it, it was strange and different, and it looked like line noise, From all of that, I thought he was talking about APL. Continuations? Who needs 'em. you can write a whole program on one line. Line Noise? It has its own character set, it doesn't use the normal ASCII characters. Objects? No, it uses Vectors and Matrices. Too many Operators? Plenty of them! I actually liked APL for a while. It was special purpose, but it was very powerful in that purpose.

functional programming (3.00 / 4) (#28) by ueluei on Thu Nov 14, 2002 at 08:42:45 AM EST I prefer languages as Haskell. The syntaxis can be very compact while the language is beautiful.

speed by jovlinger, 11/15/2002 07:32:59 PM EST (none / 0) minor nit (3.00 / 1) (#29) by karb on Thu Nov 14, 2002 at 09:15:42 AM EST I think it's "turing tar-pit" not "turning tarpit". Incidentally, it didn't know what either was until I checked out the venerable jargon file. Excellent article, as well. Thank you. -Who is the geek who would risk his neck for his brother geek? Wow (5.00 / 4) (#31) by p3d0 on Thu Nov 14, 2002 at 10:59:43 AM EST

15 of 22

A Shallow Introduction to the K Programming Language ||

I must say, that primality test is very cool, if not very efficient. This is one thing that bothers me a bit about functional languages: it's hard to tell what is efficient without fully and completely comprehending what is going on. In procedural programs, you can look for loops and procedure calls, and there can be reasonable code snippets that have neither. Furthermore, simple loops (without breaks/continues) can also be understood from an efficiency point of view. FP and OOP both spoil this, because there aren't many realistic snippets of code with no operators/virtual method calls, and these things are a brick wall when it comes to understanding performance. You simply must understand exactly what the invoked operator/method will do, and that greatly increases the amount of the program one must understand in order to evaluate a program's efficiency. Since the programmer's mental capacity is the bottleneck in software development, anything that can reduce the programmer's mental load is helpful. Paradigms like FP and OOP reduce the mental load for algorithms and information hiding (respectively), but for performance, I haven't seen anything that beats the procedural paradigm. -Patrick Doyle My comments do not reflect the opinions of my employer. the K methodology for developing software by jjayson, 11/14/2002 01:04:37 PM EST (5.00 / 1) Always good advice. [nt] by p3d0, 11/14/2002 01:42:40 PM EST (none / 0) maintainability more important than efficiency? by br14n, 11/14/2002 01:51:43 PM EST (4.00 / 1) yes by expostfacto, 11/15/2002 04:01:40 PM EST (none / 0) maintainability vs efficiency by Robb, 11/17/2002 01:24:12 PM EST (none / 0) Maintainability is always more important by p3d0, 05/18/2003 01:25:09 PM EST (none / 0) Premature optimization is the root of all evil by jefu, 11/14/2002 07:45:41 PM EST (4.50 / 2) Agreed by p3d0, 11/15/2002 07:49:25 AM EST (none / 0) Great language, hope it gets opened (none / 0) (#33) by WilliamTanksley on Thu Nov 14, 2002 at 11:30:18 AM EST If K were open, I'd use it a lot. If someone were to write an approachable tutorial to K, I'd teach it to my friends. As things are, in order to learn K I first tried to learn APL (since my university library had a book on it), then learned J (the online books are decent, "J for C programmers" was especially helpful), then went on to learn K. I love K. It's stunning, awesome. I like J, but it's so complex and so much less natural to use. But I'm almost convinced that I wouldn't have figured out K without first learning J. That's a waste. And that's why we need a good K tutorial. K is the logical replacement for the spreadsheet. -Billy

learning J first? by jjayson, 11/14/2002 01:02:43 PM EST (5.00 / 1) Thank you for writing this tutorial. by WilliamTanksley, 11/14/2002 03:11:45 PM EST (5.00 / 1) k tutorials by sa, 11/17/2002 08:35:21 AM EST (none / 0) You mentioned your website... by WilliamTanksley, 11/14/2002 03:14:28 PM EST (none / 0) Down right now, but... by jjayson, 11/14/2002 03:28:05 PM EST (none / 0) website by jefu, 11/14/2002 07:28:41 PM EST (none / 0)

16 of 22

A Shallow Introduction to the K Programming Language ||

It will be in my diary... by jjayson, 11/15/2002 02:09:19 PM EST (none / 0) J vs. Scheme and write-only languages by booyeah451, 11/16/2002 10:10:21 PM EST (none / 0) Amusing google result (4.00 / 1) (#36) by jbm on Thu Nov 14, 2002 at 11:48:23 AM EST Results Note: I have used K from C foreign function calls. It has nice to have a small language work without any leaks or bugs popping up.

agreed by jjayson, 11/14/2002 12:59:00 PM EST (none / 0) Oh, K! (5.00 / 3) (#40) by jefu on Thu Nov 14, 2002 at 12:22:55 PM EST APL was one of the first programming languages I actually got paid for using and I quite liked it. Once you got used to ("You never learn anything, you just get used to it.") the syntax and the odd character set, you could do array operations quickly and very easily. There was a story (folklore?) that some folks at Honeywell or CDC or wherever had an apl interpreter that would compile the operators on the fly down to parallel/vectorized form (not hard to do when there are all those vectors flying around) and run the vectorized form, so people could interactively play with algorithms at supercomputer speeds. Very cool. Even if only folklore the fact that it is so clearly possible with APL makes it cool. Indeed, after some experiments with a Connection Machine, I was surprised that there was no APL* for the CM4. I like languages tailored for specific purposes. I still keep a copy of snobol around (for some things its still unbeatable), as well as icon, aplus, prolog and so on. I dont have a copy of Cobol, but if the right problem calls, I may have to get one and install it. I'll take a serious look at K, but there is already aplus for linux (uses the apl character set), a very good apl that I'm using for some kinds of problem, so until K is open and provides some nice advantages, I may just sit with aplus. One minor nit - you define :

clearly meaning : sqrt ( x2 + y2 ) and I suspect that K groups to the right (as does apl) so that it works, but I'd also suggest that, even for many apl'ers :
pyth:{_sqrt (x*x)+(y*y) }

might be just that wee bit clearer, at the relatively minor cost of a pair of parens. But perhaps its just my enfeebled brain complaining again.

the point is... by Shren, 11/14/2002 06:33:12 PM EST (5.00 / 1) parallel execution by sa, 11/17/2002 08:22:05 AM EST (none / 0) spaces in code... by jjayson, 11/15/2002 02:07:40 PM EST (none / 0) 100:1? Conditions, please? (4.00 / 1) (#68) by pla on Thu Nov 14, 2002 at 04:35:54 PM EST

17 of 22

A Shallow Introduction to the K Programming Language ||

I find this hard to believe. Even C++, as bloated as it can sometimes get, still has a fairly close relation between its atomic statements and a corresponding sequence of assembly statements. Even with hand-coded assembly, getting better than 4:1 improvement (size-wise, of course) over C takes some serious effort. Then again, I have to admit that the vast majority of programs I download weigh in the megabytes, whereas somehow my own code, even doing similar tasks, almost always comes in under 100k (and I even statically link my executables whenever possible, so that should *ADD* to their size). Since you mention that they only have an *interpreter* available, do you mean to compare source-code sizes? In that case, I would agree the 100:1 claim sounds more reasonable, but only because most people do not write C in the unintelligible-but-valid shorthand C makes possible (check out some of the past obfuscated code contest winners, for an example of what I mean).

source code and more by jjayson, 11/14/2002 04:56:41 PM EST (none / 0) Source code size by ENOENT, 11/15/2002 11:39:16 AM EST (none / 0) Insane - or genius... by joto, 11/15/2002 01:12:39 PM EST (none / 0) code clarity by werner, 11/16/2002 06:45:10 AM EST (none / 0) You lack a perspective by jjayson, 11/15/2002 02:03:53 PM EST (none / 0) Perspective by ENOENT, 11/15/2002 03:54:08 PM EST (5.00 / 1) The syntax is simple, just not compared to Scheme. by jjayson, 11/16/2002 08:26:13 PM EST (none / 0) syntax matters by sa, 11/17/2002 11:16:08 AM EST (none / 0) error checking by toadi, 11/15/2002 08:27:29 AM EST (none / 0) Great Article (none / 0) (#85) by mvsgeek on Fri Nov 15, 2002 at 02:13:49 AM EST I'm not sure if I can find a use for the language but it was definately entertaining if only because of it's esotericism. - mvsgeek How well does it... (none / 0) (#101) by steveftoth on Fri Nov 15, 2002 at 02:19:18 PM EST How well does K integrate into the world of say Win32 applications? Can you do everything that you could do with only C/C++ in K? Mostly this is a problem of bindings, being able to call in and out of the C libs that Windows provides. What about Unix? Is it good for graphical applications or more text based ones? Are there any comercial apps written in K besides the kdb?

kinda, yes, yes by jjayson, 11/15/2002 02:51:21 PM EST (none / 0) SO... by steveftoth, 11/15/2002 04:00:30 PM EST (none / 0) gui by sa, 11/17/2002 07:57:42 AM EST (none / 0) Sweet Jesus, No! (5.00 / 1) (#102)

18 of 22

A Shallow Introduction to the K Programming Language ||

by egg troll on Fri Nov 15, 2002 at 02:24:40 PM EST I hope to God that this and this are intenionally obfuscated pieces of K code. Otherwise, this language is even more hideous than Perl! He's a bondage fan, a gastronome, a sensualist Unparalleled for sinister lasciviousness. sadly, no ;) by jjayson, 11/15/2002 02:41:32 PM EST (none / 0) xhtml vs. xml by E7L3, 11/17/2002 07:15:19 AM EST (none / 0) k people seem too proud of kdb (none / 0) (#105) by expostfacto on Fri Nov 15, 2002 at 03:40:01 PM EST maybe it's because it's the only "real" application done in K... it's just laughable to compare this with Oracle. comparison with gadfly (a python database) would be more apt. as mysql has proven for many years, using a variant of sql with extremely limited isolation level support does not make you a RDBMS. as mysql has also proven, it does make you liable to make sweeping claims of speed for oversimplified queries. :/ this is right up there with monty claiming "real programmers don't need referential integrity": The main reason [for kdb speed] is that kdb tables are "inverted", i.e. the data in each column is stored together, instead of in the row orientation used by most rdbms's. The column data is also organized for optimal processing speed. The result is an ultra high performance database. the reason most databases store rows together is SQL is a ROW ORIENTED LANGUAGE. (duh?) so yeah, I imagine kdb is great for computing aggregates of a single column... but i strongly doubt it's faster than [insert real database of choice] in the general case. discussion of enterprise features like triggers, in-database procedural languages, views, etc. is noticably absent on the site. i'm sure K is a great language, but knock if off with the ludicrous claims for kdb, please. -Carnage Blender: over 50 million battles served quick and dirty relational semantics by sa, 11/16/2002 01:06:03 PM EST (none / 0) relational operations (correction) by sa, 11/16/2002 01:14:53 PM EST (none / 0) relational operations (faster, more) by sa, 11/20/2002 03:37:40 PM EST (none / 0) oops by sa, 11/20/2002 03:39:49 PM EST (none / 0) faster group by by sa, 11/17/2002 08:55:49 AM EST (none / 0) think more, post less by sa, 11/17/2002 09:39:09 AM EST (none / 0) KDB by jjayson, 11/16/2002 08:04:32 PM EST (5.00 / 2) reply by expostfacto, 11/17/2002 06:00:04 PM EST (5.00 / 1) kdb by sa, 11/17/2002 07:48:27 PM EST (5.00 / 1) the uptick problem by sa, 11/17/2002 07:53:33 PM EST (none / 0) "real databases" by expostfacto, 11/17/2002 11:56:55 PM EST (none / 0) btw, re namespaces by expostfacto, 11/18/2002 12:15:49 AM EST (5.00 / 1) transaction logging by sa, 11/18/2002 07:22:39 AM EST (5.00 / 1) please by expostfacto, 11/18/2002 10:30:01 AM EST (none / 0) no, you don't get it. by sa, 11/18/2002 10:54:10 AM EST (none / 0) every checkpoint requires writing a new copy? by expostfacto, 11/18/2002 12:30:52 PM EST (none / 0)

19 of 22

A Shallow Introduction to the K Programming Language ||

checkpoint frequency by sa, 11/18/2002 03:43:11 PM EST (none / 0) kdb whitepapers by sa, 11/18/2002 10:28:51 AM EST (none / 0) that is great thinking (none / 0) (#106) by expostfacto on Fri Nov 15, 2002 at 03:48:03 PM EST ... but only when you are the only one who ever looks at your code. when you are part of a team who each has his own "way to do it," perl is an absolute nightmare. many, many years of research into SW engineering show that standards and conventions help keep you sane when working with others. the fact that perl standardizes much less at the language level than most is a barrier to good practice, not an aid to it. when it comes to the kitchen-sink school of language design, larry wall makes stroustroup look like an amateur. -Carnage Blender: over 50 million battles served sorry by expostfacto, 11/15/2002 03:50:26 PM EST (none / 0) My Confusion (none / 0) (#111) by glassware on Fri Nov 15, 2002 at 04:12:09 PM EST What is the argument for using more and more complicated operators? What is the argument for re-using the same operator in multiple conditions to mean different things? What is the argument for eliminating whitespace, and using a comment marker which is also an operator? Perhaps the ! symbol means "produce a list of all the numbers from 0 to X". Perhaps it can be reused to mean different things in different cases. Why not give it a name that indicates what it means in each case? Would typing "factor(x)" instead of "!x" break your wrist or give you carpal tunnel? Why are we trying to redefine the notion of what the symbol means in the first place? It takes time to learn what any symbol means. I applaud languages like Java that understand it is worthwhile to keep the syntax of other commonly used languages, even if they are not perfect, because it lessens the learning curve for new programmers. Yet, language developers seem to think that they can obtain a massive benefit by somehow discarding all their knowledge and starting over. I have read some very well-written critiques of the microprocessor industry that speak along these same lines. Each company seems to think they can design a new microarchitecture that will be unique and compelling. After spending a huge sum of money, they build a CPU that does one specialized task really well, and requires programmers to relearn and rewrite everything. Is that really worth the investment they made? As for K, I am not interested in learning languages that make code harder to read. Perhaps in the world you people live, code speed is 100% of your goal; but in my world, I produce volumes and volumes of code, none of which needs to be particularly fast. On a daily basis, I write a new program, read through an old program to effect a fix, and need to acquire knowledge of someone else's program. As I see it, the prime usefulness of this K language is to prevent users from inserting comments into their code by stacking it so densely that explanations of the author's thought processes are wrung out. More power to you. I prefer to have a thoughtful comment every three lines of code, which in K seems to compress down to about two characters worth of "line noise".

20 of 22

A Shallow Introduction to the K Programming Language ||

syntactic compression and lexical density by sa, 11/16/2002 02:02:36 PM EST (none / 0) hi by jjayson, 11/16/2002 07:10:54 PM EST (none / 0) the little things by jjayson, 11/16/2002 07:45:14 PM EST (none / 0) Excel is a language???? (none / 0) (#133) by ThreadSafe on Sun Nov 17, 2002 at 12:22:22 AM EST aye? Make a clone of me. And fucking listen to it! - Faik questions (none / 0) (#134) by Josh A on Sun Nov 17, 2002 at 05:42:56 AM EST Is there more than one K mailing list? How do we join it/them? Where can we find actual information about this language, as opposed to the non-information on the site? Do you think if the source was opened up, we might have an [insert nonsupported OS, like OS X] version and an apache module to embed K in html pages anytime soon?
--Thank God for Canada, if only because they annoy the Republicans so much. – Blarney

more links by jjayson, 11/18/2002 12:36:09 AM EST (none / 0) where is the 'strictly-free' implementation? (none / 0) (#146) by dawandg on Sun Nov 17, 2002 at 11:24:11 PM EST Ok, so I'm just about (85-90%) convinced that this isn't a joke. And as it happens, I'm trying to spec out a project where this kind of speed would be very helpful. But jjayson mentions a strictly free version, and I can't find it. Does anyone know where this is? Obviously I can start playing with the licensed version, but I don't want to commit unless I can be sure of the terms.

hmmm... by jjayson, 11/18/2002 12:29:18 AM EST (none / 0) A Shallow Introduction to the K Programming Language | 162 comments (156 topical, 6 editorial, 0 hidden) Display: Sort:

All trademarks and copyrights on this page are owned by their respective companies. The Rest © 2000 - Present Inc. See our legalese page for copyright policies. Please also read our Privacy Policy.

21 of 22

A Shallow Introduction to the K Programming Language || is powered by Free Software, including Apache, Perl, and Linux, The Scoop Engine that runs this site is freely available, under the terms of the GPL. Need some help? Email My heart's the long stairs.

create account | help/FAQ | mission | links | search | IRC | YOU choose the stories!

22 of 22