# The Little Schemer

F o u r t h E d i t i o n
Daniel P Friedman and Matthias Felleisen
Foreord bf G J. s
The Ten Commandments
The First Commandment
When recurring on a list of atoms, /at, ak
two questions about it: [au//? /at) and else.
When recurring on a number, a, ak two
questions about it: [ten? a) and else.
When recurring on a list of
S
-expressions, /,
ak three question about it: [au//?/), [atam?
[ car /)), and else.
The Second Commandment
Use cons to build lists.
The Third Commandment
The Fifh Commandment
When building a value with + ,always use
0 for the value of the terminating line, for
adding 0 does not change the value of an
When building a value with x, always use
1 for the value of the terminating line, for
multiplying by 1 does not change the value
of a multiplication.
When building a value with LÜÏÛ_ always
consider [) for the value of the terminating
line.
When building a list, describe the frst typi- The Sixth Commandment
cal element, and then ceas it onto the natu­
ral recursion.
The Fourth Commandment
Always change at leat one argument while
recurring. When recurring on a list of atoms,
/at,use [cdr /at). When recurring on a num­
ber, a,use [su/1 a). And when recurring on
a list of
S
-expressions, /,use [car/)and [cdr
/)if neither [au//? I) nor [atam? [car/)) are
true.
It must be changed to be closer to termina
tion. The changing argument must be tested
in the termination condition:
when using cdr, test termination with au//?
and
when using su/1, test termination with
ten?.
S
implify only ater the function is correct.
The Seventh Commandment
Recur on the su/par|s that are of the same
nature:
• On the sublists of a list.
• On the subexpressions of an arithmetic
expression.
The Eighth Commandment
Use help functions to abstract from represen­
tations.
The Ninth Commandment
Abstract common patterns with a new func­
tion.
The Tenth Commandment
Build functions to collect more than one
value at a time.
The Five Rules
The Law of Car
The primitive car is defned only for non­
empty lists.
The Law of Cdr
The primitive cdr is defned only for non­
empty lists. The cdrof any non-empty list
is always another list.
The Law of Cons
The primitive caas takes two arguments.
The second argument to ceasmust be a list.
The result is a list.
The Law of Null?
The primitive au//?is defned only for lists.
The Law of Eq?
The primitive eq' takes two arguments.
Ea must be a non-numeric atom.
The Little Schemer
Fourth Edition
Daniel P. Friedman
Indiana Uni1ersity
Bloomington, Indiana
Matthias Felleisen
Rice Uni1ersity
Houston, Texas
Ür8wÍBgS by Üu8B0 ÜÍbby
ÎOr8wOrd by Û8r8Íd J. buSSm8B
The MIT Press
Cabridge, Masahusetts
London, England
Original edition published as The Little LISPer. © 1986, 1974 by Scientifc Research
Associates.
First MIT Press Edition, 1987.
© 1996 Massachusetts Institute of Technology
electronic or mechanical means (including photocopying, recording, or information
storage and retrieval) without permission in writing from the publisher.
This book was set by the authors and was printed and bound in the United States
of America.
Fiedman, Daniel P.
The little schemer / Daniel P. Fiedman and Matthias Felleisen; drawings by
Duane Bibby; foreword by Gerald J. Sussman. -4 ed. , 1st MIT Press ed.
p. em.
Rev. ed. of: The little LISPer. 3rd ed. @1989.
Includes index.
ISBN 0-262-56099-2 (pbk: alk. paper)
1. Scheme ( Computer program language) 2. LISP ( Computer program language)
I. Felleisen, Matthias. II. Fiedman, Daniel P. Little LISPer. III. Title.
QA76. 73. S34F75 1996
005. 13'3-dc20
95-39853
CIP
To Mar, Helga, and our childrn
( (Contents)
(Foreword ix)
(Preface xi)
((
1
. Toys) 2)
((2. Do It, Do It Again, and Again, and Again ... ) 14)
((3. Cons the Magnifcent) 32)
((4. Numbers Games) 58
)
((5. *Oh My Gawd*: It's Fll of Stars) 80)
((7. Fiends and Relations) 110)
((8. Lambda the Ultimate) 124)
((9 .... and Again, and Again, and Again, ... ) 148)
((
10. What Is the Value of All of This?) 174)
(Intermission 192)
(Index 194))
IOrCVOrd
This foreword appeared in the second and third editions of The Li ttle
LISPer. We reprint it here with the permission of the author.
In 1967 I took an introductory course in photography. Most of the students (including me) came
into that course hoping to learn how to be creative-to take pictures like the ones I admired
by artists such a Edward Weston. On the frst day the teacher patiently explained the long
list of technical skills that he wa going to teach us during the term. A key was Ansel Adams'
"Zone System" for previsualizing the print values (blackness in the fnal print) in a photograph
and how they derive from the light intensities in the scene. In support of this skill we had
to learn the use of exposure meters to measure light intensities and the use of exposure time
and development time to control the black level and the contrast in the image. This is in turn
supported by even lower level skills such as loading flm, developing and printing, and mixing
chemicals. One must learn to ritualize the process of developing sensitive material so that one
gets consistent results over many years of work. The frst laboratory session was devoted to
fnding out that developer feels slippery and that fxer smells awful.
But what about creative composition? In order to be creative one must frst gain control
of the medium. One can not even begin to think about organizing a great photograph without
having the skills to make it happen. In engineering, as in other creative arts, we must learn
to do analysis to support our eforts in synthesis. One cannot build a beautiful and functional
bridge without a knowledge of steel and dirt and considerable mathematical technique for using
this knowledge to compute the properties of structures. Similarly, one cannot build a beautiful
computer system without a deep understanding of how to "previsualize" the process generated
by the procedures one writes.
Some photographers choose to use black-and-white 8x10 plates while others choose 35mm
slides. Each has its advantages and disadvantages. Like photography, programming requires a
choice of medium. Lisp is the medium of choice for people who enjoy free style and fexibility.
Lisp was initially conceived as a theoretical vehicle for recursion theory and for symbolic
algebra. It has developed into a uniquely powerful and fexible family of software development
tools, providing wrap-around support for the rapid-prototyping of software systems. As with
other languages, Lisp provides the glue for using a vat library of canned parts, produced
by members of the user community. In Lisp, procedures are frst-class data, to be passed as
arguments, returned a values, and stored in data structures. This fexibility is valuable, but
most importantly, it provides mechanisms for formalizing, naming, and saving the idioms-the
common patterns of usage that are essential to engineering design. In addition, Lisp programs
can easily manipulate the representations of Lisp programs-a feature that has encouraged the
development of a vast structure of program synthesis and analysis tools, such as cross-referencers.
The Little LISPer is a unique approach to developing the skills underlying creative program­
ming in Lisp. It painlessly packages, with considerable wit , much of the drill and practice that
is necessary to learn the skills of constructing recursive processes and manipulating recursive
data-structures. For the student of Lisp programming, The Li ttle LISPer can perform the same
service that Hanon' s fnger exercises or Czerny' s piano studies perform for the student of piano.
Foreword
Gerald J. Sussman
Cambridge, Massachusetts
ix
ÏrC!BCC
To celebrate the twentieth anniversary of Scheme we revised The Little LISPer a
third time, gave it the more accurate title The Little Schemer, and wrote a sequel :
The Seasoned Schemer.
Programs accept data and produce data. Designing a program requires a thorough understand­
ing of data; a good program refects the shape of the data it deals with. Most collections of data,
and hence most programs, are recursive. Recursion is the act of defning an object or solving a
problem in terms of itself.
The goal of this book is to teach the rader to think rcursively. Our frst task is to decide
which language to use to communicate this concept . There are three obvious choices: a natural
language, such a English; formal mathematics; or a programming language. Natural languages
are ambiguous, imprecise, and sometimes awkwardly verbose. These are all virtues for general
communication, but something of a drawback for communicating concisely as precise a concept
a recursion. The language of mathematics is the opposite of natural language: it can express
powerful formal ideas with only a few symbols. Unfortunately, the language of mathematics
is often cryptic and barely accessible without special training. The marriage of technology
and mathematics presents us with a third, almost ideal choice: a programming language. We
believe that programming languages are the best way to convey the concept of recursion. They
share with mathematics the ability to give a formal meaning to a set of symbols. But unlike
mathematics, programming languages can be directly experienced-you can take the programs in
this book, observe their behavior, modify them, and experience the efect of these modifcations.
Perhaps the best programming language for teaching recursion i s Scheme. Scheme i s
inherently symbolic-the programmer does not have to think about the relationship between the
symbols of his own language and the representations in the computer. Recursion is Scheme' s nat­
ural computational mechanism; the primary programming activity is the creation of (potentially)
recursive defnitions. Scheme implementations are predominantly interactive-the programmer
can immediately participate in and observe the behavior of his programs. And, perhaps most
importantly for our lessons at the end of this book, there is a direct correspondence between
the structure of Scheme programs and the data those programs manipulate.
Although Scheme can be described quite formally, understanding Scheme does not require
a particularly mathematical inclination. In fact , The Little Schemer is based on lecture notes
fom a two-week "quickie" introduction to Scheme for students with no previous programming
experience and an admitted dislike for anything mathematical . Many of these students were
preparing for careers in public afairs. It is our belief that writing prgrams rcursively in Scheme
is essentially simple patter rcognition. Since our only concern is recursive programming, our
treatment is limited to the whys and wherefores of just a few Scheme features: car, cdr, cons,
eq?, null ?, zero? , addl, subl, number?, and, or, quote, lambda, defne, and cond. Indeed, our
language is an idealized Scheme.
The Little Schemer and The Seasoned Schemer will not introduce you to the practical
world of programming, but a mastery of the concepts in these books provides a start toward
understanding the nature of computation.
Preface xi
What You Need to Know to Read This Book
Acknowledgments
We are indebted to many people for their contributions and assistance throughout the devel­
opment of the second and third editions of this book. We thank Bruce Duba, Kent Dybvig,
Chris Haynes, Eugene Kohlbecker, Richard Salter, George Springer, Mitch Wand, and David S.
Wise for countless discussions that infuenced our thinking while conceiving this book. Ghasan
Abbas, Charles Baker, David Boyer, Mike Dunn, Terry Falkenberg, Rob Fiedman, John Gateley,
Mayer Goldberg, Iqbal Khan, Julia Lawall, Jon Mendelsohn, John Nienart , Jefrey D. Perotti,
Ed Robertson, Anne Shpuntof, Erich Smythe, Guy Steele, Todd Stein, and Larry Weisselberg
provided many important comments on the drafts of the book. We especially want to thank Bob
Filman for being such a thorough and uncompromising critic through several readings. Finally
we wish to acknowledge Nancy Garrett, Peg Fletcher, and Bob Filman for contributing to the
design and 'ery.
The fourth and latest edition greatly benefted from Dorai Sitaram' s incredibly clever Scheme
typesetting program SLTE. Kent Dybvig' s Chez Scheme made programming in Scheme a
most pleasant experience. We gratefully acknowledge criticisms and suggestions from Shelaswau
Bushnell, Richard Cobbe, David Combs, Peter Drake, Kent Dybvig, Rob Fiedman, Steve Ganz,
Chris Haynes, Erik Hilsdale, Eugene Kohlbecker, Shriram Krishnamurthi, Julia Lawall, Suzanne
Menzel Collin McCurdy, John Nienart , Jon Rossie, Jonathan Sobel , George Springer, Guy Steele,
John David Stone, Vikram Subramaniam, Mitch Wand, and Melissa Wingard-Phillips.
Do not rush through this book. Read carefully; valuable hints are scattered throughout the
text. Do not read the book in fewer than three sittings. Read systematically. If you do not flly
understand one chapter, you will understand the next one even less. The questions are ordered
by increasing difculty; it will be hard to answer later ones if you cannot solve the earlier ones.
The book is a dialogue between you and us about interesting examples of Scheme programs.
If you can, try the examples while you read. Schemes are readily available. While there are
minor syntactic variations between diferent implementations of Scheme (primarily the spelling of
particular names and the domain of specifc functions) , Scheme is basically the same throughout
the world. To work with Scheme, you will need to defne atom?, subl, and addl. which we
introduced in The Little Schemer:
(define atom?
(labda (x)
(and (not (pair? x)) (not (null? x)))))
To fnd out whether your Scheme has the correct defnition of atom?, try (atom? (quote ()))
and make sure it returns #f. In fact , the material is also suited for modern Lisps such Ü
Common Lisp. To work with Lisp, you will also have to add the function atom?:
xii
(defun atom? (x)
(not (listp x)))
Preface
Moreover, you may need to modify the programs slightly. Typically, the material requires
only a few changes. Suggestions about how to try the programs in the book are provided in
the framenotes. Famenotes preceded by "S: " concern Scheme, those by "L:" concern Common
Lisp.
In chapter 4 we develop basic arithmetic from three operators: add1 , sub1 , and zero ?. Since
Scheme does not provide add1 and sub1 , you must defne them using the built-in primitives for
addition and subtraction. Therefore, to avoid a circularity, our basic arithmetic addition and
subtration must be written using diferent symbols: ÷ and -, respectively.
We do not give any formal defnitions in this book. We believe that you can form your
own defnitions and will thus remember them and understand them better than if we had
written eah one for you. But be sure you know and understand the Laws and Commandments
thoroughly before passing them by. The key to learning Scheme is "pattern recognition. " The
Commandments point out the patterns that you will have already seen. Early in the book,
some concepts are narrowed for simplicity; later, they are expanded and qualifed. You should
also know that , while everything in the book is Scheme, Scheme itself is more general and
incorporates more than we could intelligibly cover in an introductory text . After you have
mastered this book, you can read and understand more advanced and comprehensive books on
Scheme.
We use a few notational conventions throughout the text , primarily changes in typeface
for diferent classes of symbols. Variables and the names of primitive operations are in i talic.
Baic data, including numbers and representations of truth and falsehood, is set in sans serif.
Keywords, i. e. , defne, lambda, cond, else, and, or, and quote, are in boldface. When you
try the programs, you may ignore the typefaces but not the related framenotes. To highlight
this role of typefaces, the programs in framenotes are set in a typewriter face. The typeface
distinctions can be safely ignored until chapter 10, where we treat programs as data.
Finally, Webster defnes "punctuation" as the act of punctuating; specifcally, the act ,
pratice, or system of using standardized marks in writing and printing to separate sentences
or sentence elements or to make the meaning clearer. We have taken this defnition literally
and have abandoned some familiar uses of punctuation in order to make the meaning clearer.
Specifcally, we have dropped the use of punctuation in the left-hand column whenever the item
that precedes such punctuation is a term in our programming language.
Food appears in many of our examples for two reasons. First , food is easier to visualize
than abstract symbols. (This is not a good book to read while dieting. ) We hope the choice of
food will help you understand the examples and concepts we use. Second, we want to provide
you with a little distraction. We know how frustrating the subject matter can be, and a little
You are now ready to start . Good luck! We hope you will enjoy the challenges waiting for
you on the following pages.
Preface
Bon appetit !
Daniel P. Fiedman
Matthias Felleisen
xiii
The Little Schemer
Is it true that this is an atom?
atom1
1 L, S: (quote atom) or 'atom
"L: " and "S:" are described in the preface.
Is it true that this is an atom?
turkey
Is it true that this is an atom?
I4ò2
Is it true that this is an atom?
u
Is it true that this is an atom?
•abc\$
Is it true that this is a list?
(atom)1
1 L, S: (quote (atom) ) or ' ( atom)
Is it true that this is a list?
(atom turkey or)
Toys
Yes,
because atom i s a string of characters
beginning with the letter a .
Yes,
because turkey is a string of characters
beginning with a letter.
Yes,
because I4ò2i s a string of digits.
Yes,
because u i s a string of one character,
which is a letter.
Yes,
because •abc\$ i s a string of characters
beginning with a letter or special character
other than a left ¯( or right ¯)
parenthesis.
Yes,
because (atom) i s an atom enclosed by
parentheses.
Yes,
because it is a collection of atoms enclosed
by parentheses.
3
Is it true that this is a list?
(atom turkey) or
Is it true that this is a list?
( (atom turkey) or)
Is it true that this is an S-expression?
xyz
Is it true that this is an S-expression?
(x y z)
Is it true that this is an S-expression?
((x y) z)
Is it true that this is a list?
(how are you doing so far)
How many S-expressions are in the list
(how are you doi ng so far)
and what are they?
Is it true that this is a list?
( ( ( how) are) ( (you) (doing so) ) far)
How many S-expressions are in the list
( ( ( how) are) ( (you) (doi ng so) ) far)
and what are they?
4
No,
because these are actually two
S-expressions not enclosed by parentheses.
The frst one is a list containing two
atoms, and the second one is an atom.
Yes,
because the two S-expressions are now
enclosed by parentheses.
Yes,
because all atoms ae S-expressions.
Yes,
because it is a list .
Yes,
because all lists are S-expressions.
Yes,
because i t i s a collection of S-expressions
enclosed by parentheses.
Six,
how, are, you, doi ng, so, and far.
Yes,
because i t i s a collection of S-expressions
enclosed by parentheses.
Three,
( ( how) are) , ( (you) (doi ng so) ) , and far.
Chapter 1
Is it true that this is a list?
(
¹
Yes,
because it contains zero S-expressions
enclosed by parentheses. This special
S-expression is called the null (or empty)
list .
Is it true that this is an atom? No,
()
Is it true that this is a list?
(() () () ( ) )
What i s the LÚÍ of l
where l is the argument
(a b c)
What is the LÚÍ of l
where
l is ((a b c) x y z)
What is the LÚÍ of l
where l is hotdog
What is the LÚÍ of l
where l is ()
Toys
because ( ) is j ust a list .
Yes,
a,
because it is a collection of S-expressions
enclosed by parentheses.
because a is the frst atom of this list .
(a b c) ,
because (a b c) i s the frst S-expression of
this non-empty list .
You cannot ak for the LÚÍ of an atom.
You cannot ak for the LÚÍ of the empty
list .
1 L: nil
The Law of Car
The primitive car is defned
only for non-empty lists.
5
What is the car of l
where
l is ( ( ( hotdogs) ) (and) ( pi ckl e) rel i sh)
What is ( car l)
where
l is ( ( ( hotdogs) ) (and) ( pi ckl e) rel i sh)
What is ( car ( car l))
where
l is ( ( ( hotdogs) ) (and) )
What i s the cdr of l
where
l is (a b c)
Note: "cdr" is pronounced "could-er. "
What is the cdr of l
where
l is ((a b c) x y z)
What is the cdr of l
where
l is ( hamburger)
What is ( cdr l)
where
l is ( (x) t r)
What is ( cdr a)
where
a is hotdogs
6
( ( hotdogs) ) ,
"The list of the list of hotdogs. "
( ( hotdogs) ) is the frst S-expression of l .
( ( hotdogs) ) ,
because ( car l) is another way to ak for
"the car of the list l. "
( hotdogs) .
(b c) ,
because (b c) is the list l without ( car l ) .
(x y z) .
( ) .
(t r) ,
because ( cdr l) is just another way to ak
for "the cdr of the list l . "
You cannot ak for the cdr of an atom.
Chapter 1
What is ( cdr l)
where l is ¸)
You cannot ask for the cdr of the null list .
1 L: nÍ1
The Law of Cdr
The primitive cdr is defned only for
non-empty lists. The cdr of any non­
empty list is always another list.
What is ( car ( cdr l ))
where
l is ( ( b) (x y) ((L)))
What is ( cdr ( cdr l ))
where
l is ( ( b) (x y) ((L)))
What is ( cdr ( car l ))
where
l is (a (b (L))d)
What does car take a an argument?
What does cdr take as an argument?
What is the cons of the atom a and the list l
where a is peanut
and
l is ( butter and jel l y)
This can also be written "( cons a l )" .
Read: "cons the atom a onto the list l. "
Toys
(x y) ,
because ((x y) ((L)))is ( cdr l), and (x y) is
the car of ( cdr l ) .
(((L)))ç
because ((x y) ((L))) is ( cdr l ) , and (((L)))
is the cdr of ( cdr l ) .
since ( car l) is an atom, and cdr does not
take an atom as an argument; see The Law
of Cdr.
It takes any non-empty list .
It takes any non-empty list .
( peanut butter and jel l y) ,
because cons adds an atom to the front of
a list .
7
What is the LÛ7b of b and l
where b is ( banana and)
and
l is (peanut butter and jel l y)
What is ¸ LÛ7b b l)
where
b is ( ( hel p) thi s)
and
l is (is very ( ( hard) to l earn) )
What does LÛ7b take a its arguments?
What is ¸ LÛ7b b l)
where
s is (a b (c))
and
l is ( )
What is ¸ LÛ7b b l)
where b is a
and l is ( )
What is ¸ LÛ7b b l)
where
b is ((a b L))
and
l is b
What is ¸ LÛ7b b l)
where b is a
and l is b
8
( (banana and) peanut butter and jel l y) ,
front of a list.
( ( (hel p) thi s) is very ( ( hard) to l earn) ) .
LÛ7b takes two arguments:
the frst one is any S-expression;
the second one is any list .
((a b ¸L)))¿
because () is a list .
(a) .
since the second argument l must be a list .
1 In practice, (cona a /) works for all values a and /, and
(ca (cona a /)) =a
(cd (cona a /)} = /.
Why?
Chapter 1
The Law of Cons
The primitive cons taes two arguments.
The second argument to cons must be a
list. The result is a list.
What is ( cons s ( car l ) )
where s is a
and
l is ((b) c d)
What i s ( cons s ( CJ r l ) )
where s is a
and
l is ((b) c d)
Is i t true that the list l i s the null list
where l is ( )
What is ( null ?1 (quote ( ) ) )
1 L! 0u11
Is ( null? l) true or false
where
l is (a b c)
Toys
¸a b) .
Why?
¸a c d) .
Why?
Yes,
because it i s the list composed of zero
S-expressions.
This question can also be written:
( null ? l ) .
Tue,
because (quote ( ) )
1
is a notation for the
null list .
1 L! Also 0 and • ().
S: Also •().
False,
because l is a non-empty list .
9
Is ( null ? a) true or false
where
a is spaghetti
because you cannot ask null? of Ü atom.
1 In practice, (null? a) is false for everything, except the
empty list.
The Law of Null?
The primitive null ? is de­
fned only for lists.
Is it true or false that b is an atom
where b is Harry
Is ( atom ?
1
b) true or false
where
b is Harry
1
L: (defu atom? (x)
(not (listp x) ) )
S: (define atom?
(labda (x)
(and (not (pair? x)) (not (null? x) ) ) ) )
Is ( atom? b) true or false
where
ö is (Harry had a heap of appl es)
How many arguments does atom? take and
what are they?
10
Tue,
because Harry is a string of characters
beginning with a letter.
Tue,
because ( atom? b) is j ust another way to
ak "Is b is an atom?"
False,
since ö is a list.
It takes one argument. The argument can be
any S-expression.
Chapter 1
Is ( atom? ( car l ) ) true or false
where
l is (Harry had a heap of appl es)
Is ( atom? ( cdr l)) true or false
where
l is (Harry had a heap of apples)
Is ( atom? ( cdr l)) true or false
where
l is (Harry)
Is ( atom? ( car ( cdr l ) ) ) true or false
where
l is (swi ng low sweet cherry oat)
Is ( atom? ( car ( cdr l ) ) ) true or false
where
l is (swi ng (l ow sweet) cherry oat)
Tue or false: a1 and a2 are the same atom
where a1 is Harry
and
a2 is Harry
Is ( eq?1 a1 a2) true or false
where a1 is Harry
and
a2 is Harry
1 L: eq
Is ( eq? a1 a2) true or flse
where a1 is margari ne
and
a2 is butter
Toy
Tue,
because ( car l) is Harry, and Harry is an
atom.
False.
False,
because the list ( ) is not an atom.
Tue,
because ( cdr l) is (l ow sweet cherry oat ) ,
and ( car ( cdr l )) i s l ow, which i s an atom.
False,
since ( cdr l) is ( ( l ow sweet) cherry oat ) , and
( car ( cdr l ) ) is (l ow sweet) , which is a list .
Tue,
because a1 is the atom Harry and a2 is the
atom Harry.
Tue,
because ( eq? a1 a2) is j ust another way to
ak, "Are a1 and a2 the same
non-numeric atom?"
False,
since a1 and a2 are diferent atoms.
1 1
How many arguments does eq? take and
what are they?
Is ( eq? l1 l2) true or false
where l1 is ¸ )
and
l2 is (strawberry)
Is ( eq? n1 n2) true or false
where n1 is 6
and
n2 is 7
It takes two arguments. Both of them must
be non-numeric atoms.
¸ ) and (strawberry) are lists.
1 In pratice, lists m8 be arguments 0Í egT. Tw lists are
egT if they are the same list.
6 and 7 are numbers.
1 In pratice, some numbers may be arguments 0f egT.
The Law of Eq?
The primitive eq? takes two ar­
guments. Each must be a non­
numeric atom.
Tue,
Is ( eq? ( LÚr l) Ú) true or false
where
l is (Mary had a l i ttl e l amb chop)
and
because ( LÚr l) is the atom Mary, and the
argument Ú is also the atom Mary.
Ú is Mary
Is ( eq? ( cdr l) Ú) true or false
where
l is (soured mi l k)
and
Ú is mi l k
12
See The Laws of Eq? and Cdr.
Chapter 1
Is ( eq? ( car l) ( car ( cdr l ) ) ) true or false
where
l is ( beans beans we need jel l y beans)
Tue,
because it compares the frst and second
atoms in the list .
=Now go make yourself a peanut butter and jelly sandwich. ¢
This space reserved for
JÏÏÏY b¯AÏÏbÎ
13
�))�
�m��m�

�� �
���
Tue or false: ( lat ? l)
where
l is (Jack Sprat could eat no chi cken fat)
Tue or false: ( lat ? l)
where
l is ( (Jack) Sprat coul d eat no chi cken fat)
Tue or false: ( lat? l)
where
l is (Jack (Sprat coul d) eat no chi cken fat)
Tue or false: ( lat ? l)
where l is ( )
Tue or false: a lat i s a list of atoms.
Write the function lat ? using some, but not
necessarily all , of the following functions:
car cdr Lonb null ? atom? and eq?
Tue,
because each S-expression in l is an atom.
False,
since ( car l) is a list .
False,
since one of the S-expressions in l is a list.
Tue,
because it does not contain a list .
Tue!
Every lat is a list of atoms!
You were not expected to be able to do this
yet , because you are still missing some
ingredients. Go on to the next question.
Good luck.
Are jOU rested?
Do It, Do It Again, and Again, and Again . . . 15
(defne lat ?1
(lambda ( l )
(cond
( ( null ? l) #t )
( ( atom? ( car l ) ) ( lat ? ( cdr l ) ) )
(else #f ) ) ) )
What i s the value of ( lat ? l )
where
l is the argument ( bacon and eggs)
1 L: (defu 1at? (1}
(cond
( (null 1} t}
( (atom? (ca 1} } (lat? (cdr 1} } }
( t nil} } }
How do we determine the answer #t for the
application
( lat ? l)
What is the frst question asked by ( lat ? l)
What is the meaning of the cond¯line
( (null ? l) #t )
where
l is ( bacon and eggs)
What is the next question?
16
#t .
The application ( lat ? l)
where
l is ( bacon and eggs)
has the value #t -true-because l is a lat .
You were not expected to know this one
either. The answer is determined by
Hint: Write down the defnition of the
function lat ? and refer to it for the next
group of questions.
( null? l)
Note:
( cond . . . ) asks questions;
(lambda . . . ) creates a function; and
(defne . . . ) gives it a name.
( null ? l) asks if the argument l is the null
list. If it is, the value of the application is
true. If it is not , we ask the next question.
In this case, l is not the null list, so we ask
the next question.
( atom? ( car l ) ) .
Chapter 2
What is the meaning of the line
( ( atom? ( car l ) ) ( lat? ( cdr l ) ) )
where
l is (bacon and eggs)
What is the meaning of
( lat? ( cdr l ))
Now what i s the argument l for lat ?
What is the next question?
What is the meaning of the line
( ( null? l) #t )
where
l is now (and eggs)
What is the next question?
What is the meaning of the line
( ( atom? ( car l ) ) ( lat? ( cdr l ) ) )
where
l is (and eggs)
What is the meaning of
( lat ? ( cdr l ) )
Do It, Do It Again, and Again, and Again . . .
( atom? ( car l ) ) asks if the frst S-expression
of the list l is an atom. If ( car l) is an atom,
we want to know if the rest of l is also
composed only of atoms. If ( car l) is not an
atom, we ask the next question. In this case,
( car l) is an atom, so the value of the
function is the value of ( lat ? ( cdr l ) ) .
( lat ? ( cdr l ) ) fnds out i f the rest of the list l
is composed only of atoms, by referring to
the function with a new argument .
Now the argument l is ( cdr l ) , which is
(and eggs) .
( null ? l ) .
( null ? l) asks i f the argument l i s the null
list . If it is, the value of the application is
#t . If it is not , we ask the next question. In
this case, l is not the null list , so we ask the
next question.
( atom? ( car l ) ) .
( atom? ( car l ) ) asks i f ( car l) i s an atom. If
it is an atom, the value of the application is
( lat ? ( cdr l ) ) . If not , we ask the next
question. In this case, ( car l) is an atom, so
we want to fnd out if the rest of the list l is
composed only of atoms.
( lat ? ( cdr l)) fnds out if the rest of l is
composed only of atoms, by referring again
to the function lat ?, but this time, with the
argument ( cdr l ) , which is (eggs) .
17
What is the next question?
What is the meaning of the line
( ( null ? l) #t )
where
l is now (eggs)
What is the next question?
What is the meaning of the line
( ( atom? ( car l ) ) ( lat ? ( cdr l ) ) )
where
l is now (eggs)
What is the meaning of ( lat ? ( cdr l ) )
Now, what i s the argument for lat ?
What is the meaning of the line
( ( null ? l) #t )
where
l is now ( )
Do you remember the question about
( lat ? l)
18
( null ? l ) .
( null ? l) asks i f the argument l i s the null
list . If it is, the value of the application is
#t -true. If it is not, move to the next
question. In this cae, l is not null, so we ask
the next question.
( atom? ( car l ) ) .
( atom? ( car l )) asks i f ( car l) i s an atom. If
it is, the value of the application is
( lat ? ( cdr l ) ) . If ( car l) is not an atom, ak
the next question. In this case, ( car l) is an
atom, so once again we look at ( lat? ( cdr l ) ) .
( lat ? ( cdr l ) ) fnds out i f the rest of the list l
is composed only of atoms, by referring to
the function lat ?, with l becoming the value
of ( cdr l ) .
( ) .
( null ? l) asks if the argument l is the null
list . If it is, the value of the application is
the value of #t . If not , we ak the next
question. In this case, () is the null list. So,
the value of the application ( lat ? l)
where
l is ( bacon and eggs) , is #t -true.
Probably not . The application ( lat? l) has
the value #t if the list l is a list of atoms
where
l is ( bacon and eggs) .
Chapter 2
Can you describe what the function lat ? does
This is the function lat ? again:
(defne lat ?
(lambda ( l )
(cond
( ( null ? l) #t )
( ( atom? ( car l )) ( lat ? ( cdr l ) ) )
(else #f))))
What is the value of ( lat? l)
where
l is now ( bacon (and eggs))
What is the frst question?
What is the meaning of the line
( ( null? l) #t )
where
l is ( bacon (and eggs))
What is the next question?
What is the meaning of the line
( ( atom? ( car l )) ( lat ? ( cdr l ) ) )
where
l is ( bacon (and eggs))
Do It, Do It Again, and Again, and Again ...
Here are our words:
"lat ? looks at each S-expression in a list , in
turn, and asks if each S-expression is an
atom, until it runs out of S-expressions. If
it runs out without encountering a list , the
value is #t . If it fnds a list , the value is
#f-false. "
To see how we could arrive at a value of
"false, " consider the next few questions.
#f,
since the list l contains an S-expression
that is a list .
( null ? l ) .
( null ? l) asks i f l is the null list . If i t i s, the
value is #t . If l is not null, move to the next
question. In this case, it is not null, so we
( atom? ( car l ) ) .
( atom? ( car l )) asks i f ( car l) i s an atom. If
it is, the value is ( lat ? ( cdr l ) ) . If it is not ,
we ask the next question. In this case,
( car l) is an atom, so we want to check if the
rest of the list l is composed only of atoms.
19
What is the meaning of
( lat ? ( cdr l ) )
What i s the meaning of the line
( ( null ? l) #t )
where
l is now ((and eggs) )
What i s the next question?
What is the meaning of the line
( ( atom? ( car l ) ) ( lat ? ( cdr l ) ) )
where
l is now ( (and eggs) )
What i s the next question?
What is the meaning of the question else
Is else true?
else
Why is else the last question?
Why do we not need to ask any more
questions?
What is the meaning of the line
(else #f )
20
( lat ? ( cdr l ) ) checks to see if the rest of the
list l is composed only of atoms, by referring
to lat ? with l replaced by ( cdr l ) .
( null ? l) asks i f l i s the null list. If i t i s null,
the value is #t . If it is not null , we ask the
next question. In this case, l is not null , so
move to the next question.
( atom? ( car l ) ) .
( atom? ( car l )) asks i f ( car l) i s an atom. If
it is, the value is ( lat ? ( cdr l ) ) . If it is not ,
we move to the next question. In this case,
( car l) is not an atom, so we ask the next
question.
else.
else asks if else is true.
Yes, because the question else is always true!
Of course.
Because we do not need to ask any more
questions.
Because a list can be empty, can have an
atom in the frst position, or can have a list
in the frst position.
else asks if else is true. If else is true-a it
always is-then the answer is #f-false.
Chapter 2
What is
) ) )
Can you describe how we determined the
value #f for
( lat? l)
where
l is ( bacon (and eggs) )
Is (oÏ ( null ? ll ) ( atom? l2) ) true or false
where l1 is ¸ )
and
l2 is (d e f g)
Is (oÏ ( null ? l1 ) ( null ? l2) ) true or false
where
l1 is (a b c)
and
l2 is ¸ )
Is ¸LÏ ( null ? l1 } ( null ? l2) ) true or false
where
l1 is (a b c)
and
l2 is (atom)
What does ¸LÏ ) do?
Do It, Do It Agan, and Again, and Again . ..
These are the closing or matching
parentheses of ( cond . . . , (lambda . . . , and
(defne . . . , which appear at the beginning
of a function defnition.
Here is one way to say it :
"( lat? l) looks at each item in its argument
to see if it is an atom. If it runs out of
items before it fnds a list , the value of
( lat ? l) is #t . If it fnds a list , as it did in
the example ( bacon (and eggs) ) , the value
of ( lat ? l) is #f. "
Tue,
because ( null ? l1 ) is true where l1 is ( ) .
Tue,
because ( null ? l2) is true where l2 is ( ) .
False,
because neither ( null? l1 ) nor ( null? l2) is
true where
l1 is (a b c)
and
l2 is (atom) .
(oÏ . . . ) aks two questions, one at a time. If
the frst one is true it stops and answers true.
Otherwise it asks the second question and
answers with whatever the second question
21
Is it true or false that a is a member of lat
where a is tea
and
lat is (coffee tea or mi l k)
Is ( member? a lat) true or false
where a is poached
and
lat is (fri ed eggs and scrambl ed eggs)
This is the function member?
(defne member?
(lambda (a lat)
(LLÏd
( ( null ? lat) #f )
(else ¸LÏ ( eq? ( car lat) a)
( member? a ( cdr lat) ) ) ) ) ) )
What i s the value of ( member? a lat)
where a is meat
and
lat is ( mashed potatoes and meat gravy)
How do we determine the value #t for the
above application?
What is the frst question asked by
( member? a lat)
22
Tue,
because one of the atoms of the lat,
(cofee tea or mi l k)
i s the same as the atom a-tea.
False,
since a is not one of the atoms of the lat.
#t ,
because the atom meat is one of the atoms
of lat ,
( mashed potatoes and meat gravy) .
The value is determined by asking the
questions about ( member? a lat ) .
Hint : Write down the defnition of the
function member? and refer to it while you
work on the next group of questions.
( null ? lat ) .
This i s also the frst question asked by lat ?.
Chapter 2
The First Commandment
(preliminary)
Always ask null ? as the frst question in expressing
any function.
What is the meaning of the line
( ( null ? lat) #f )
where
lat is (mashed potatoes and meat gravy)
What is the next question?
Why is else the next question?
Is else really a question?
What is the meaning of the line
(else ¸LÏ ( eq? ( car lat) a)
( member? a ( cdr lat ) ) ) )
Tue or false:
¸LÏ ( eq? ( car lat) a)
( member? a ( cdr lat) ) )
where a i s meat
and
lat is (mashed potatoes and meat gravy)
Do It, Do It Again, and Again, and Again . . .
( null ? lat) asks if lat is the null list . If it is,
the value is #f, since the atom meat was not
found in lat. If not , we ask the next question.
In this case, it is not null, so we ak the next
question.
else.
Because we do not need to ask any more
questions.
Yes, else is a question whose value is always
true.
Now that we know that lat is not null ?, we
have to fnd out whether the car of lat is the
same atom as a, or whether a is somewhere
in the rest of lat. The answer
¸LÏ ( eq? ( car lat) a)
( member? a ( cdr lat ) ) )
does this.
We will fnd out by looking at each question
in turn.
23
Is ( eq? ( car lat) a) true or false
where a is meat
and
lat is ( mashed potatoes and meat gravy)
What is the second question of ¸LÏ )
Now what are the arguments of member?
What is the next question?
Is ( null ? lat) true or false
where
lat is (potatoes and meat gravy)
What do we do now?
What is the next question?
What is the meaning of
¸LÏ ( eq? ( car lat) a)
( member? a ( cdr lat ) ) )
Is a eq? to the car of lat
24
False,
because meat is not eq? to mashed,
the car of
( mashed potatoes and meat gravy) .
( member? a ( cdr lat ) ) .
This refers t o the function with the
argument lat replaced by ( cdr lat ) .
a i s meat
and
lat is now ( cdr lat) , specifcally
( potatoes and meat gravy) .
( null ? lat ) .
Remember The First Commandment.
#f-false.
else.
¸LÏ ( eq? ( car lat) a)
( member? a ( cdr lat) ) )
fnds out i f a i s eq? to the car of lat or i f a is
a member of the cdr of lat by referring to the
function.
No, because a is meat and the car of lat is
potatoes.
Chapter 2
So what do we do next?
Now, what are the arguments of member?
What is the next question?
What do we do now?
What is the next question?
What is the vlue of
¸LÏ ( eq? ( car lat) a)
( member? a ( cdr lat) ) )
Why?
What do we do now?
We ask ( member? a ( cdr lat ) ) .
a i s meat, and
lat is (and meat gravy) .
( null ? lat ) .
Ask the next question, since ( null ? lat) is
false.
else.
The value of ( member? a ( cdr lat ) ) .
Because ( eq? ( car lat) a) i s false.
Recur-refer to the function with new
arguments.
What are the new arguments? a is meat, and
lat is ( meat gravy) .
What is the next question? ( null ? lat ) .
What do we do now? Since ( null ? lat) is false, ask the next
question.
What is the next question? else.
Do It, Do It Agan, and Again, and Again . . . 25
What is the value of
¸LÏ ( eq? ( car lat) a)
( �e�ber? a ( cdr lat ) ) )
What i s the value of the application
( �e�ber? a lat )
where a is meat
and
lat is ( meat gravy)
What is the value of the application
( �e�ber? a lat)
where a is meat
and
lat is (and meat gravy)
What is the value of the application
( �e�ber? a lat)
where a is meat
and
lat is ( potatoes and meat gravy)
What is the value of the application
¸�e�ber? a lat )
where a is meat
and
lat is ( mashed potatoes and meat gravy)
Just to make sure you have it right , let ' s
quickly run through it again. What is the
value of ( �e�ber? a lat)
where
a is meat
and
lat is ( mashed potatoes and meat gravy)
¸ null ? lat )
26
#t ,
because ( car lat ) , which is meat, and a,
which is meat , are the same atom.
Therefore, ¸LÏ ) answers with #t .
#t ,
because we have found that meat is a
member of
( meat gravy) .
#t ,
because meat is also a member of the lat
(and meat gravy) .
#t ,
because meat is also a member of the lat
( potatoes and meat gravy).
#t ,
because meat is also a member of the lat
( mashed potatoes and meat gravy).
Of course, this is our original lat.
#t.
Hint : Write down the defnition of the
function �e�ber? and its arguments and
refer to them as you go through the next
group of questions.
No. Move to the next line.
Chapter 2
else
Yes.
¸LÏ ( eq? ( car lat) a)
Perhaps.
( member? a ( cdr lat) ) )
( eq? ( car lat) a) No. Ask the next question.
What next?
Recur with a and ( cdr lat)
where a is meat
and
( cdr lat) is ( potatoes and meat gravy) .
( null ? lat) No. Move to the next line.
else Yes, but ( eq? ( car lat) a) is false.
Recur with a and ( cdr lat)
where a is meat
and
( cdr lat) is (and meat gravy) .
( null ? lat) No. Move to the next line.
else Yes, but ( eq? ( car lat) a) is false.
Recur with a and ( cdr lat)
where a is meat
and
( cdr lat) is ( meat gravy) .
( null ? lat) No. Move to the next line.
( eq? ( car lat) a) Yes, the value is #t .
Do It, Do It Again, and Again, and Again . . . 27
¸LÏ ( eqP ( car lat) a)
( memberP a ( cdr lat) ) )
What i s the value of ( memberP a lat)
where a is meat
and
lat is ( meat gravy)
What is the value of ( memberP a lat)
where a is meat
and
lat is (and meat gravy)
What is the value of ( member? a lat)
where a is meat
and
lat is ( potatoes and meat gravy)
What is the value of ( memberP a lat)
where a is meat
and
lat is (mashed potatoes and meat gravy)
What is the value of ( memberP a lat)
where a is l i ver
and
lat is ( bagels and lox)
Let ' s work out why it is #f. What ' s the frst
( null ? lat)
28
#t .
#t .
#t .
#t .
#t .
#f.
(7Ull ? lat) .
No. Move t o the next line.
Chapter 2
else
( null ? lat)
else
( null ? lat)
else
( null ? lat)
What is the value of ( member? a lat)
where a is l i ver
and
lat is ( )
What i s the value of
¸LÏ ( eq? ( car lat) a)
( member? a ( cdr lat) ) )
where
a is l i ver
and
lat is ( l ox)
Do It, Do It Again, and Again, and Again . . .
Yes, but ( eq? ( car lat) a) is false.
Recur with a and ( cdr lat)
where a is l i ver
and
( cdr lat) is (and l ox) .
No. Move to the next line.
Yes, but ( eq? ( car lat) a) is false.
Recur with a and ( cdr lat)
where a is l i ver
and
( cdr lat) is (l ox) .
No. Move to the next line.
Yes, but ( eq? ( car lat) a) is still false.
Recur with a and ( cdr lat)
where a is l i ver
and
( cdr lat) is ¸ ) .
Yes.
#f.
#f.
29
What is the value of ( member? a lat)
where a is l i ver
and
lat is (l ox)
What is the value of
¸LÏ ( eq? ( car lat) a)
( member? a ( cdr lat) ) )
where
a is l i ver
and
lat is (and lox)
What is the value of ( member? a lat)
where a is l i ver
and
lat is (and l ox)
What is the value of
¸LÏ ( eq? ( car lat) a)
( member? a ( cdr lat ) ) )
where
a is l i ver
and
lat is ( bagel s and lox)
What is the value of ( member? a lat)
where a is l i ver
and
lat is ( bagel s and l ox)
30
#f.
#f.
#f.
#f.
#f.
Chapter 2
Do you believe all this? Then you may rest!
This space for doodling
Do It, Do It Agan, and Again, and Again . . . 31
What is ( rmber a lat)
where a is mi nt
and
lat is ( l amb chops and mi nt jelly)
( rmber a lat)
where a is mi nt
and
lat is ( l amb chops and
mi nt fl avored mi nt jel l y)
( rmber a lat)
where a is toast
and
lat is ( bacon l ettuce and tomato)
( rmber a lat)
where a is cup
and
lat is (coffee cup tea cup and hi ck cup)
What does ( rmber a lat) do?
What steps should we use to do this?
And if ( null ? lat) is true?
What do we know if ( null ? lat) is not true?
Is there any other question we should ask
Cons the Magifcent
( l amb chops and jel l y)
"Rember" stands for rmove a member.
( l amb chops and fl avored mi nt jel l y) .
(bacon lettuce and tomato) .
(coffee tea cup and hi ck cup) .
It takes an atom and a lat as i ts arguments,
and makes a new lat with the frst occurrence
of the atom in the old lat removed.
First we will test ( null ? lat)-The First
Commandment .
Return ¸ ) .
We know that there must be at least one
atom in the lat .
No. Either a lat is empty or it contains at
least one atom.
33
What do we do if we know that the lat
contains at least one atom?
How do we ak if a is the same as ( car lat )
What would be t he value of ( rmber a lat ) if
a were the same as ( car lat )
What do we do i f a i s not the same a
( car lat )
How do we remove the frst occurrence of a
in the rest of lat
Is there any other question we should ak?
Now, let ' s write down what we have so far:
(defne rmber
(lambda (a lat)
(cond
( ( null ? lat ) (quote ( )))
(else (Cond
( ( eq? ( car lat) a) ( cdr lat ))
(else ( rmber a
( cdr lat ))))))))
What is the value of ( rember a lat ) where
a is bacon
and
lat is ( bacon l ettuce and tomato)
34
We ak whether a is equal to ( car lat ).
By using
(Cond
(
--
(

( eq? ( car lat ) a).
( cdr lat ).
--
)
-- )) .
We want to keep ( car lat ), but also fnd out
if a is somewhere in the rest of the lat .
( rmber a ( cdr lat )).
No.
( l ettuce and tomato) .
Hint : Write down the function rmber and
its arguments, and refer to them as you go
through the next sequence of questions.
Chapter 3
Now, let ' s see if this function works. What is
the frst question?
What do we do now?
else
What next?
( eq? ( car lat) a)
Is this the correct value?
But did we really use a good example?
What does rmber do?
What do we do now?
What is the value of ¸rmber a lat)
where a is and
and
lat is ( bacon l ettuce and tomato)
Cons the Magnifcent
( null ? lat) .
Move t o the next line and ask the next
question.
Yes.
Yes, so the value is ( cdr lat) . In this case, it
is the list
( l ettuce and tomato) .
Yes, because it is the original list without the
atom bacon.
Who knows? But the proof of the pudding is
in the eating, so let ' s try another example.
It takes an atom and a lat as its arguments,
and makes a new lat with the frst occurrence
of the atom in the old lat removed.
We compare each atom of the lat with the
atom a, and if the comparison fails we build
a list that begins with the atom we j ust
compared.
( bacon l ettuce tomato) .
35
Let us see if our function rmber works.
What is the frst question asked by rmber
What do we do now?
else
( eq? ( car lat) a)
What is the meaning of
(else ( rmber a ( cdr lat ) ) )
( null ? lat)
else
( eq? ( car lat) a)
What is the meaning of
( rmber a ( cdr lat) )
( null ? lat)
else
36
( null ? lat ) .
Move to the next line, and ask the next
question.
Okay, so ask the next question.
No, so move to the next line.
else asks if else is true-a it always is-and
the rest of the line says to recur with a and
( cdr lat) , where
a is and
and
( cdr lat) is (l ettuce and tomato) .
No, so move to the next line.
Sure.
No, so move to the next line.
Recur
where
a is and
and
( cdr lat) is (and tomato) .
No, so move to the next line, and ask the
next question.
Of course.
Chapter 3
( eq? ( car lat) a)
So what is the result?
Is this correct?
What did we do wrong?
How can we keep from losing the atoms
bacon and l ettuce
Yes.
( cdr l at)-(tomato) .
No, since (tomato) is not the list
( bacon l ettuce and tomato)
with j ust a-and-removed.
We dropped and, but we also lost all the
atoms preceding and.
We use Cons the Magnifcent . Remember
cons , from chapter 1?
The Second Commandment
Use cons to build lists.
Let 's see what happens when we use cons
(defne rmber
(lambda (a lat)
(Cond
( ( n·ull? lat) (quote ( ) ) )
(else (cond
(( eq? ( car lat) a) ( cdr lat) )
(else ( cons ( car lat)
( rmber a
( cdr lat) ) ) ) ) ) ) ) )
What i s the value of ( rmber a lat)
where a is and
and
lat is (bacon l ettuce and tomato)
Cons the Magnifcent
( bacon l ettuce tomato) .
Hint : Make a copy of this function with
cons and the arguments a and lat so you
can refer to it for the following questions.
37
What is the frst question?
What do we do now?
else
( eq? ( car lat) a)
What is the meaning of
( cons ( car lat)
( rmber a
( cdr lat ) ) )
where
a is and
and
lat is ( bacon l ettuce and tomato)
What is the meaning of ( rmber a ( cdr lat) )
( null ? lat)
else
( eq? ( car lat) a)
What is the meaning of
( cons ( car lat)
( rmber a
( cdr lat ) ) )
38
( null ? lat ) .
Yes, of course.
No, so move to the next line.
It says to cons the car of lat-bacon-onto
the value of
( rmber a ( cdr lat ) ) .
But since we don' t know the value of
( rmber a ( cdr lat) ) yet , we must fnd it
before we can cons ( car lat) onto it.
This refers to the function with lat replaced
by ( cdr lat)( l ettuce and tomato) .
No, so move to the next line.
No, so move to the next line.
It says to cons the car of lat-l ettuce-onto
the value of
( rmber a ( cdr lat ) ) .
But since we don' t know the value of
( rmber a ( cdr lat) ) yet , we must fnd it
before we can cons ( car lat) onto it .
Chapter 3
What is the meaning of ( rmber a ( cdr lat ) )
( null ? lat)
else
( eq? ( car lat) a)
What is the value of the line
( ( eq? ( car lat) a) ( cdr lat) )
Are we fnished?
We now have a value for
( rmber a ( cdr lat) )
where a i s and
and
( cdr lat) is (and tomato)
This value is (tomato)
What next?
What is the result when we cons l ettuce onto
(tomato)
What does (l ettuce tomato) represent?
Cons the Magifcent
This refers to the function with lat replaced
by ( cdr lat ) , that is, (and tomato) .
No, so ask the next question.
Still .
Yes.
( cdr lat)(tomato ) .
Certainly not ! We know what ( rmber a lat )
is when lat is (and tomato) , but we don' t yet
know what it is when lat is
( l ettuce and tomato)
or
( bacon l ettuce and tomato) .
Recall that we wanted to cons l ettuce onto
the value of ( rmber a ( cdr lat ) )
where
a was and and ( cdr lat ) was (and tomato) .
Now that we have this value, which is
(tomato) , we can cons l ettuce onto it .
( l ettuce tomato) .
It represents the value of
( cons ( car lat )
( rember a
( cdr lat ) ) ) ,
when
lat is ( l ettuce and tomato)
and
( rmber a ( cdr lat ) ) is (tomato) .
39
Are we fnished yet?
We now have a value for ( rmber a ( cdr lat) )
where a i s and
and
( cdr lat) is ( l ettuce and tomato)
This value is ( l ettuce tomato)
This is not the fnal value, so what must we
do again?
What is the result when we cons bacon onto
( l ettuce tomato)
What does ( bacon l ettuce tomato) represent?t
t Lunch?
Are we fnished yet?
Can you put in your own words how we
determined the fnal value
( bacon l ettuce tomato)
40
Not quite. So far we know what
( rmber a lat) is when
lat is ( l ettuce and tomato) ,
but we don' t yet know what i t i s when
lat is ( bacon lettuce and tomato) .
Recall that , at one time, we wanted to cons
bacon onto the value of ( rmber a ( cdr lat) ) ,
where
a wa and
and
( cdr lat) was ( l ettuce and tomato) .
Now that we have this value, which is
( l ettuce tomato) ,
we can cons bacon onto it.
( bacon l ettuce tomato) .
It represents the value of
( cons ( car lat)
( rmber a ( cdr lat ) ) ) ,
when
lat is ( bacon l ettuce and tomato)
and
( rmber a ( cdr lat) ) is ( l ettuce tomato) .
Yes.
In our words:
"The function rmber checked each atom of
the lat , one at a time, to see if it was the
same as the atom and. If the car was not
the same as the atom, we saved it to be
consed to the fnal value later. When
rmber found the atom and, it dropped it,
and consed the previous atoms back onto
the rest of the lat. "
Chapter 3
Can you rewrite rmber so that it refects the
above description?
Do you think this is simpler?
So why don' t we simplify right away?
Let' s see if the new rmber is the same a the
old one. What is the value of the application
( rmber a lat)
where a is and
and
lat is ( bacon l ettuce and tomato)
( null ? lat)
( eq? ( car lat) a)
else
What is the meaning of
( cons ( car lat)
( rmber a ( cdr lat) ) )
Cons the Magifcent
Yes, we can simplify it.
(defÏe rmber
(lambda ( a lat)
(LoÏd
( ( null ? lat) (quote ( ) ) )
( ( eq? ( car lat) a) ( cdr lat ) )
(elÜe ( cons ( car lat)
( rmber a ( cdr lat ) ) ) ) ) ) )
Fnctions like rmber can always be
simplifed in this manner.
Because then a function' s structure does not
coincide with its argument ' s structure.
( bacon l ettuce tomato) .
Hint : Write down the function rmber and
its arguments and refer to them as you go
through the next sequence of questions.
No.
No.
Yes, so the value is
( cons ( car lat)
( rmber a ( cdr lat ) ) ) .
This says t o refer t o the function rmber but
with the argument lat replaced by ( cdr lat) ,
and that after we arrive at a value for
( rmber a ( cdr lat) ) we must cons
( car lat )bacon¨nto it.
41
( null ? lat)
( eq? ( car lat ) a)
else
What is the meaning of
( cons ( car lat)
( rember a ( cdr lat ) ) )
( null ? lat)
( eq? ( car lat) a)
What is the value of the line
( ( eq? ( car lat) a) ( cdr lat) )
Now what?
Now what?
Now that we have completed rmber
try this example: ( rmber a lat)
where a is sauce
and
lat is (soy sauce and tomato sauce)
42
No.
No.
Yes, so the value is
( cons ( car lat)
( rmber a ( cdr lat ) ) ) .
This says we recur using the function
rmber, with the argument lat replaced by
( cdr lat) , and that after we arrive at a value
for ( rmber a ( cdr lat) ) , we must cons
( car lat )-l ettuce-onto it.
No.
Yes.
It is ( cdr lat)-(tomato) .
Now cons ( car lat)-l ettuce-onto (tomato) .
Now cons ( car lat )bacon-onto
( l ettuce tomato) .
( rmber a lat) is (soy and tomato sauce) .
Chapter 3
What is (frsts l)
where
l is ( (appl e peach pumpki n)
(pl um pear cherry)
(grape raisi n pea)
( bean carrot eggpl ant))
What is (frsts l)
where
l is ( ( a b) (c d) (e f) )
What is (frsts l)
where l is ¸ )
What i s (frsts l)
where
l is ( (five pl ums)
(four)
(el even green oranges))
What is (frsts l)
where
l is ( ( (five pl ums) four)
(el even green oranges)
( ( no) more) )
In your own words, what does (frsts l ) do?
Cons the Magnifcent
( appl e pl um grape bean) .
(a c e) .
( ) .
(fi ve four el even) .
( (fi ve pl ums) el even (no) ) .
We tried the following:
"The function frsts takes one argument , a
list, which is either a null list or contains
only non-empty lists. It builds another list
composed of the frst S-expression of each
internal list . "
43
See if you can write the function frsts
Remember the Commandments!
Why
(defne frsts
(lambda ( l )
. . . ) )
Why (cond . . . )
Why ( ( null ? l) . . . )
Why (else
Why (else
Why ( cons
Why (frsts ( cdr l ) )
Why ) ) )
44
This much is easy:
(defne frsts
(lambda ( l )
(cond
( ( null ? l) . . . )
(else ( cons . . . (frsts ( cdr l ) ) ) ) ) ) )
Because we always state the function name,
(lambda, and the argument¸s) of the
function .
Because we need to ak questions about the
actual arguments.
The First Commandment.
Because we only have two questions to ak
about the list l: either it is the null list , or it
contains at least one non-empty list.
See above. And because the last question is
always else.
Because we are building a list-The Second
Commandment .
Because we can only look at one S-expression
at a time. To look at the rest, we must recur.
Because these are the matching parentheses
for ( cond, (lambda, and (defne, and they
always appear at the end of a function
defnition.
Chapter 3
Keeping in mind the defnition of (frsts l)
what is a typical element of the value
of (frsts l)
where
l is ( ( a b) ¸L d) ¸e f) )
What is another typical element?
Consider the function seconds
What would be a typical element of the value
of ( seconds l)
where
l is ¸ ¸a b) ¸L d) ¸e f) )
How do we describe a typical element for
(frsts l)
When we fnd a typical element of (frsts l)
what do we do with it?
a.
L¿ or even e.
b, d, or f.
As the car of an element of l-( car ( car l ) ) .
See chapter 1 .
cons i t onto the recursion-(frsts ( cdr l ) ) .
The Third Commandment
When building a list, describe the frst typical ele­
ment, and then cons it onto the natural recursion.
With The Third Commandment , we can now
fll in more of the function frsts
What does the last line look like now?
Cons the Magifcent
(else ( cons ( car ( car l ) ) (frsts ( cdr l ) ) ) ) .
~~
typical
element
natural
recursion
45
What does (frsts l) do
(defne frsts
(lambda ( l )
(cond
( ( null ? l) . . . )
( else ( cons ( car ( car l ))
(frsts ( cdr l ) ) ) ) ) ) )
where l is ¸ ¸Ü b) ¸L d) ¸e f) )
( null ? l) where l is ¸ ¸Ü b) ¸L d) ¸e f) )
What is the meaning of
( cons ( car ( car l))
(frsts ( cdr l)))
( null ? l) where l is ¸ ¸L d) ¸e f) )
What is the meaning of
( cons ( car ( car l ) )
(frsts ( cdr l ) ) )
( null ? l) where l is ¸ ¸e f) )
What is the meaning of
( cons ( car ( car l ))
(frsts ( cdr l )))
( null ? l)
Now, what is the value of the line
( ( null ? l) . . . )
46
Nothing yet . We are still missing one
important ingredient in our recipe. The frst
line ( ( null ? l) . . . ) needs a value for the case
where l is the null list . We can, however,
proceed without it for now.
No, so move to the next line.
It saves ( car ( car l) ) to cons onto
(frsts ( cdr l) ) . To fnd (frsts ( cdr l) ) , we
refer to the function with the new argument
( cdr l ) .
No, so move t o the next line.
Save ( car ( car l ) ) , and recur with
(frsts ( cdr l ) ) .
No, so move t o t he next line.
Save ( car ( car l)) , and recur with
(frsts ( cdr l ) ) .
Yes.
There is no value; something is missing.
Chapter 3
What do we need to cons atoms onto?
For the purpose of consing, what value can
we give when ( null ? l) is true?
With () as a value, we now have three cons
steps to go back and pick up. We need to:
I. either
1. cons e onto ( )
2. cons c onto the value of 1
3. cons a onto the value of 2
II. or
1 . cons a onto the value of 2
2. cons c onto the value of 3
3. cons e onto ( )
Ill or
cons a onto
the cons of c onto
the cons of e onto
( )
In any case, what is the value of (frsts l)
With which of the three alternatives do you
feel most comfortable?
What is ( insertR new old lat)
where
new is toppi ng
old is fudge
and
lat is (i ce cream wi th fudge for dessert)
( insertR new old lat)
where
new is jal apeno
old is and
and
lat is (tacos tamal es and sal sa)
Cons the Magifcent
A list .
Remember The Law of Cons.
Since the fnal value must be a list , we
cannot use #t or #f. Let ' s try (quote ( ) ) .
(a c e) .
Correct ! Now you should use that one.
(ice cream with fudge toppi ng for dessert ) .
(tacos tamal es and j al apeno sal sa) .
47
( insertR new old lat)
where
new is e
old is d
and
lat is (a b L d f g d h)
I n your own words, what does
( insertR new old lat) do?
See if you can write the frst three lines of
the function insertR
Which argument changes when we recur with
insertR
How many questions can we ak about the
lat?
What do we know if ( null ? lat) is not true?
element?
48
(a b L d e f g d h) .
In our words:
"It takes three arguments: the atoms new
and old, and a lat . The function insertR
builds a lat with new inserted to the right
of the frst occurrence of old. "
(defne insert
(lambda ( new old lat)
(cond . . . ) ) )
lat , because we can only look at one of its
atoms at a time.
Two.
A lat is either the null list or a non-empty
list of atoms.
First , we ak ( null ? lat) . Second, we ak
else, because else is always the last question.
We know that lat ha at leat one element .
First , we ask ( eq? ( car lat) old) . Then we
ak else, because there are no other
interesting cases.
Chapter 3
Now see if you can write the whole function
inserlR
(defne inserlR
(lambda ( new old lat)
(cond
'
(else
(cond
'
'

)
) ) ) ) ) )
What i s the value of the application
( inserl new old lat)
that we just determined
where
new is toppi ng
old is fudge
and
lat is (i ce cream wi th fudge for dessert)
So far this i s the same Ü rmber
What do we do in inserl when
( eq? ( car lat) old) is true?
How is this done?
Now we have
(defne inserlR
(lambda ( new old lat)
(cond
( ( null ? lat) (quote ( ) ) )
(else (cond
( ( eq? ( car lat) old)
( cons new ( cdr lat ) ) )
(else ( cons ( car lat)
Cons the Magifcent
( inserlR new old
( cdr lat ) ) ) ) ) ) ) ) )
Here i s our frst attempt .
(defne inserlR
(lambda ( new old lat)
(cond
( ( null ? lat) (quote ( ) ) )
(else
(cond
( ( eq? ( car lat) old) ( cdr lat) )
(else ( cons ( car lat)
( inserlR new old
( cdr lat) ) ) ) ) ) ) ) )
(i ce cream wi th for dessert ) .
When ( car lat) i s the same Ü old, we want
to insert new to the right .
Let ' s try Lonöing new onto ( cdr lat) .
Yes.
49
So what is ( insertR new old lat) now
where
new is toppi ng
old is fudge
and
lat is (i ce cream wi th fudge for dessert)
Is this the list we wanted?
What still needs to be done?
How can we include old before new
Now let ' s write the rest of the function
insertR
50
(i ce cream with toppi ng for dessert ) .
No, we have only replaced fudge with toppi ng.
Somehow we need to include the atom that is
the same as old before the atom new.
T consing old onto ( cons new ( cdr lat ) ) .
(defne insertR
(lambda ( new old lat)
(cond
( ( null ? lat) (quote ( ) ) )
(else (cond
( ( eq? ( car lat) old)
( cons old
( cons new ( cdr lat ) ) ) )
(else ( cons ( car lat)
( insertR new old
( cdr lat ) ) ) ) ) ) ) ) )
When new i s toppi ng, old i s fudge, and lat is
( i ce cream wi th fudge for dessert ) , the value of
the application, ( insertR new old lat ) , is
(i ce cream wi th fudge toppi ng for dessert) .
If you got this right , have one.
Chapter 3
Now try insertL
Hint: insertL inserts the atom new to the
left of the frst occurrence of the atom old
in lat
Did you think of a diferent way to do it?
Now try subst
Hint: ( subst new old lat) replaces the frst
occurrence of old in the lat with new
For example,
where
new is toppi ng
old is fudge
and
lat is (i ce cream with fudge for dessert)
the value is
(i ce cream with toppi ng for dessert)
Now you have the idea.
Cons the Magifcent
This much is easy, right?
(defne insertL
(lambda ( new old lat)
(cond
( ( null ? lat) (quote ( ) ) )
(else (cond
For example,
( ( eq? ( car lat) old)
( cons new
( cons old ( cdr lat ) ) ) )
(else ( cons ( car lat)
( insertL new old
( cdr lat ) ) ) ) ) ) ) ) )
( ( eq? ( car lat) old)
( cons new ( cons old ( cdr lat ) ) ) )
could have been
( ( eq? ( car lat) old)
( cons new lat) )
since ( cons old ( cdr lat) ) where old i s eq? to
( car lat) is the same as lat.
Obviously,
(defne subst
(lambda ( new old lat)
¸cond
( ( null ? lat) ¸quote ( ) ) )
¸else ¸cond
( ( eq? ( car lat) old)
( cons new ( cdr lat ) ) )
¸else ( cons ( car lat)
( subst new old
( cdr lat ) ) ) ) ) ) ) ) )
This i s the same as one of our incorrect
attempts at insertR.
51
Go cons a piece of cake onto your mouth.
Now try subst2
Hint:
( subst2 new o1 o2 lat)
replaces either the frst occurrence of o1 or
the frst occurrence of o2 by new
For example,
where
new is vani l la
o1 is chocolate
o2 is banana
and
lat is (banana ice cream
with chocolate toppi ng)
the value is
(vani l la ice cream
with chocolate topping)
Did you think of a better way?
(defne subst2
(lambda ( new o1 o2 lat)
(cond
( ( num lat) (quote ( ) ) )
(else (cond
( ( eq� ( car lat) o1 )
( cons new ( cdr lat) ) )
( ( eq� ( car lat) o2)
( cons new ( cdr lat) ) )
(else ( cons ( car lat)
( subst2 new o1 o2
( cdr lat) ) ) ) ) ) ) ) )
Replace the two eq� lines about the ( car lat)
by
( (or ( eq� ( car lat) o1 ) ( eq� ( car lat) o2) )
( cons new ( cdr lat ) ) ) .
If you got the last function, go and repeat the cake-consing.
Do you recall what rmber does?
52
The function rmber looks at each atom of a
lat to see if it is the same a the atom a. If it
is not, rmber saves the atom and proceeds.
When it fnds the frst occurrence of a, it
stops and gives the value ( cdr lat) , or the
rest of the lat , so that the value returned is
the original lat , with only that occurrence of
a removed.
Chapter 3
Write the function multirmber which gives
as its fnal value the lat with all occurrences
of a removed.
(defne multirmber
(lambda ( a lat)
(cond
(

(else
(cond
(

(

)
) ) ) ) ) )
Hint : What do we want as the value when
( eq? ( car lat) a) is true?
Consider the example
where a is cup
and
lat is (cofee cup tea cup and hi ck cup)
Can you see how multirmber works?
( null ? lat)
else
( eq? ( car lat) a)
What is the meaning of
( cons ( car lat)
( multirmber a
( cdr lat) ) )
( null ? lat)
Cons the Magnifcent
¸defne multirmber
(lambda ( a lat)
¸cond
( ( null ? lat) ¸quote ¸ ) ) )
¸else
¸cond
( ( eq? ( car lat) a)
( multirmber a ( cdr lat ) ) )
¸else ( cons ( car lat)
( multirmber a
( cdr lat ) ) ) ) ) ) ) ) )
After the frst occurrence of a, we now
recur with ( multirmber a ( cdr lat ) ) , in
order to remove the other occurrences.
The value of the application is
(cofee tea and hi ck) .
Possibly not , so we will go through the steps
necessary to arrive at the value
(cofee tea and hi ck) .
No, so move to the next line.
Yes.
No, so move to the next line.
Save ( car lat)cofeeto be consed onto
the value of ( multirmber a ( cdr lat) ) later.
Now determine
( multirmber a ( cdr lat ) ) .
No, so move t o the next line.
53
else
( eq? ( car lat) a)
( null ? lat)
else
( eq? ( car lat) a)
What is the meaning of
( cons ( car lat)
( multirmber a
( cdr lat ) ) )
( null ? lat)
else
( eq? ( car lat) a)
( null ? lat)
( eq? ( car lat) a)
What is the meaning of
( cons ( car lat)
54
( multirmber a
( cdr lat ) ) )
Naturally.
Yes, so forget ( car lat ) , and determine
( multirmber a ( cdr lat ) ) .
No, so move to the next line.
Yes!
No, so move to the next line.
Save ( car lat)teato be consed onto the
value of ( multirmber a ( cdr lat) ) later. Now
detennine
( multirmber a ( cdr lat) ) .
No, so move t o the next line.
Okay, move on.
Yes, so forget ( car lat ) , and determine
( multirmber a ( cdr lat ) ) .
No, so move to the next line.
No, so move to the next line.
Save ( car lat)andto be consed onto the
value of ( multirmber a ( cdr lat) ) later. Now
determine
( multirmber a ( cdr lat ) ) .
Chapter 3
( null? lat)
( eq? ( car lat) a)
What is the meaning of
( cons ( car lat)
( multirmber a
( cdr lat) ) )
( null ? lat)
( eq? ( car lat) a)
( null ? lat)
Are we fnished?
What do we do next?
What do we do next?
What do we do next?
What do we do next?
Are we fnished now?
Cons the Magifcent
No, so move to the next line.
No, so move to the next line.
Save ( car lat)hi ckto be consed onto the
value of ( multirember a ( cdr lat) ) later. Now
determine
( multirmber a ( cdr lat ) ) .
No, so move t o the next line.
Yes, so forget ( car lat) , and determine
( multirmber a ( cdr lat ) ) .
Yes, so the value i s ( ) .
No, we still have several conses to pick up.
We cons the most recent ( car lat) we
have-hi ck-onto ( ) .
We cons and onto ¸ hi ck) .
We cons tea onto (and hi ck) .
We cons cofee onto (tea and hi ck) .
Yes.
55
Now write the function multiinsertR
(defne multiinsertR
(lambda ( new old lat)
(cond

(else
(cond
( )
( ) ) ) ) ) )
Is this function defned correctly?
¸defne multiinsertL
(lambda ( new old lat)
¸cond
( ( null ? lat) ¸quote ¸ ) ) )
¸else
¸cond
( ( eq? ( car lat) old)
( cons new
( cons old
( multiinsertL new old
lat ) ) ) )
¸else ( cons ( car lat)
( multiinsertL new old
( cdr lat ) ) ) ) ) ) ) ) )
Was the terminal condition ever reached?
56
¸defne multiinsertR
(lambda ( new old lat)
(cond
( ( null? lat) (quote ( ) ) )
(else
¸cond
( ( eq? ( car lat) old)
( cons ( car lat)
( cons new
( multiinsertR new old
( cdr lat) ) ) ) )
¸else ( cons ( car lat)
( multiinsertR new old
( cdr lat) ) ) ) ) ) ) ) )
It would also be correct to use old i n place
of ( car lat) because we know that
( eq? ( car lat) old) .
Not quite. To fnd out why, go through
( multiinsertL new old lat)
where
new is fried
old is fi sh
and
lat is (chi ps and fish or fi sh and fried) .
No, because we never get pat the frst
occurrence of old.
Chapter 3
Now, try to write the function multiinsertL
aan:
(defne multiinsertL
(lambda ( new old lat)
(cond
'
(else
(cond
( )
( ) ) ) ) ) )
(defne multiinsertL
(lambda ( new old lat)
(cond
( ( null ? lat) (quote ( ) ) )
¸else
¸cond
(( eq? ( car lat) old}
( cons new
( cons old
( multiinsertL new old
( cdr lat) ) ) ) )
¸else ( cons ( car lat)
( multiinsertL new old
( cdr lat) ) ) ) ) ) ) ) )
The Fourth Commandment
(prliminar)
Always change at least one argument while recurring. It
must be changed to be closer to termination. The changing
argument must be tested in the termination condition:
when using cdr, test termination with null?.
Now write the function multisubst
¸defne multisubst
(lambda ( new old lat)
¸cond
'
¸else
¸cond
'
'

)
) ) ) ) ) )
¸defne multisubst
{lambda ( new old lat)
¸cond
( ( null? lat) ¸quote ¸ ) ) )
¸else ¸cond
( ( eq? ( car lat) old}
( cons new
( multisubst new old
( cdr lat) ) ) )
¸else ( cons ( car lat)
( multisubst new old
( cdr lat) ) ) ) ) ) ) ) )
Cons the Magnifcent 57
- 9--�
Is 14 an atom?
Is ( atom? 7) true or false
where
7 is 14
Is -3 a number?
Is 3. 14159 a number?
Are -3 and 3. 14159 numbers?
What is ( add1 1 7)
where 7 is 67
1 L: 1+
Cl11 bda (n}
(+ n 1 } } } }
What i s ( add1 67)
What is ( sub1 1 7)
where 7 is 5
1 L: 1-
S: (define aub1
Cl11 bda (n}
(- n 1 } } } }
Numbers Games
Yes, because all numbers are atoms.
Tue, because 14 is an atom.
Yes,
but we do not consider negative numbers.
Yes,
but we consider only whole numbers.
Yes,
but the only numbers we use are the
nonnegative integers ¸i. e. , 0, 1, 2, 3, 4, . . . ) .
68.
Also 68,
because we don' t need to say "where 7 is
67" when the argument is a number.
4.
59
What is ( sub1 0}
Is ( zer?
1
0} true or false?
¯ L: ZeI0§
Is ( zer? 1492} true or false?
What is ( + 46 12}
Ty to write the function +
Hint : It uses zer? add1 1 and sub1 1
¯ Remember to use our defnitions for addI and aubI.
But didn' t we just violate The First
Commandment?
If zer P is like null?
60
1 ( subl R) , where R i s Û, ha no answer because we consider
only nonnegative numbers. In pratice, this result is -1.
Tue.
Flse.
58.
(defne + 1
(lambda ( n M)
(cond
( ( zero? M) n)
(else ( add1 ( + n ( sub1 M) ) ) ) ) ) )
Wasn' t that easy?
1 L, S: This is like +. Write it a o+ (see prefae) .
Yes, but we can treat zer? like null ? since
zer? asks if a number is empty and null?
asks if a list is empty.
Yes! cons builds lists and add1 builds
numbers.
Chapter 4
What is ( - 14 3)
What is ( - 17 9)
What is ( - 18 25)
Ty to write the function -
Hint: Use sub1
Ca you describe how ( - 7 Û) works?
Is this a tup?
(2 11 3 79 47 6)
Is this a tup?
(8 55 5 555)
Is this a tup?
(1 2 8 appl e 4 3)
Is this a tup?
(3 (7 4) 13 9)
Numbers Games
1 1 .
8.
No answer. There are no negative numbers.
(defne - 1
(lambda ¸7 Û)
(cond
( ( zer? Û) 7)
(else (sub1 ( - 7 (sub1 Û) ) ) ) ) ) )
1 L, S: This i s like -. Write i t a o - (see prefae) .
It takes two numbers Ü arguments, and
reduces the second until it hits zero. It
subtracts one from the result Ü many times
Ü it did to cause the second one to reach
zero.
Yes: tup is short for tuple.
Yes, of course, it is also a list of numbers.
No, it is just a list of atoms.
No, because it is not a list of numbers.
(7 4) is not a number.
61
Is this a tup?
()
where
tup is (3 5 2 8)
where
tup is ( 15 6 7 12 3)
What is the natural way to build numbers
from a list?
When building lists with cons
the value of the terminal condition is ( )
What should be the value of the terminal
condition when building numbers with ÷
What is the natural terminal condition for a
list?
What is the natural terminal condition for a
tup?
When we build a number from a list of
numbers, what should the terminal condition
line look like?
What is the terminal condition line of
62
Yes, it is a list of zero numbers. This special
case is the empty tup.
18.
43.
It builds a number by totaling all the
numbers in its argument.
Use ÷ in place of cons : ÷ builds numbers in
the same way as cons builds lists.
0.
( null ? l ) .
( null ? tup) .
( ( null ? tup) 0) , j ust as ( ( null ? l) (quote ( ) ) )
i s often the terminal condition line for lists.
( ( null? tup) 0) .
Chapter 4
How is a lat defned?
How is a tup defned?
What is used in the natural recursion on a
list?
What is used in the natural recursion on a
tup?
Why?
How many questions do we need to ak
How many questions do we need to ak
How is a number defned?
What is the natural terminal condition for
numbers?
What is the natural recursion on a number?
How many questions do we need to ak
Numbers Games
It is either an empty list , or it contains an
atom, ( car lat) , and a rest, ( cdr lat) , that is
also a lat .
It is either an empty list , or it contains a
number and a rest that is also a tup.
( cdr lat ) .
( cdr tup) .
Because the rest of a non-empty list i s a list
and the rest of a non-empty tup is a tup.
Two.
Two, because it is either empty or it is a
number and a rest, which is again a tup.
It is either zero or it is one added to a rest,
where rest is again a number.
( zero ? 7) .
( sub1 7) ·
Two.
63
The First Commandment
(frst revision)
When recurring on a list of atoms, lat , ask two questions
about it: ( null ? lat ) and else.
it: ( zero ? n) and else.
What does cons do?
What is the terminal condition line of
What is the natural recursion for addtup
What does addtup use to build a number?
Fill in the dots in the following defnition:
(lambda ( tup)
(cond
( ( null ? tup) 0)
(else . . . ) ) ) )
What i s ¸ x 5 3)
What is ¸ x 13 4)
64
It builds lists.
It builds a number by totaling all the
numbers in a tup.
( ( null? tup) 0) .
( addtup ( cdr tup) ) .
It uses +, because + builds numbers, too!
Here is what we flled in:
( + ( car tup) ( addtup ( cdr tup) ) ) .
Notice the similarity between this line, and
the lat line of the function rmber:
( cons ( car lat ) ( rmber a ( cdr lat ) ) ) .
15.
52.
Chapter 4
What does ¸ ? Ñ M) do?
What is the terminal condition line for ?
Since (zerP M) is the terminal condition, M
must eventually be reduced to zero. What
function is used to do this?
It builds up a number by adding Ñ up M
times.
( ( zerP M) 0) , because W ? 0 ~ 0.
sub1 .
The Fourth Commandment
(frst rvision)
Always change at least one argument while recurring. It
must be changed to be closer to termination. The changing
argument must be tested in the termination condition:
when using cdr, test termination with null ? and
when using sub1 , test termination with zero ?.
What is another name for ¸ ? Ñ ( sub1 M) ) in
this case?
Ty to write the function ?
Number Games
It' s the natural recursion for ? .
(defne ?
Î
(lambda ( Ñ M)
(cond
( ( zerP M) 0)
(else ( + Ñ ¦ ? Ñ ( sub1 M) ) ) ) ) ) )
1 L, S: This i s lik •.
65
What is ¸ x 12 3)
( zero ? Û)
What is the meaning of
( + n ¸ x n (sub1 Û) ) )
What are the new arguments of
¸ x n Û)
( zero ? Û)
What is the meaning of
¸ ÷n ¸ x n ( sub1 Û) ) )
What are the new arguments of
¸ x n Û)
( zer? Û)
What is the meaning of
¸ ÷n ¸ x n ( sub1 Û) ) )
What i s the value of the line
( ( zer? M) 0)
Are we fnished yet?
66
36,
but let ' s follow through the function one
time to see how we get this value.
No.
It adds n ¸where n ° 12) to the natural
recursion. If x is correct then
¸ x 12 ( sub1 3))
should be 24.
n is 12, and Û is 2.
No.
It adds n ¸where n ° 12) to ¸ x n ( sub1 Û) ) .
n is 12, and Û is 1 .
No.
It adds n ¸where n ° 12) to ¸ x n ( sub1 Û) ) ·
0, because ( zer? Û) is now true.
No.
Chapter 4
Why not?
What is the value of the original application?
Because we still have three +es to pick up.
Add 12 to 12 to 12 to 0 yielding 36,
Notice that Ñ has been +ed M times.
( x 12 3) = 12 + ( x 12 2)
= 12 + 12 + ( x 12 1)
Argue, using equations, that x is the
conventional multiplication of nonnegative
integers, where n is 12 and Ï is 3.
= 12 + 12 + 12 + ( ? 12 0)
= 12 + 12 + 12 + 0,
Again, why is 0 the value for the terminal
condition line in x
which is as we expected. This technique
works for all recursive functions, not j ust
those that use numbers. You can use this
approach to write functions as well as to
argue their correctness.
Because 0 will not afect +. That is,
n + O = n.
The Fifh Commandment
When building a value with + , always use 0 for the vlue of the
terminating line, for adding 0 does not change the value of an
When building a vlue with x , always use 1 for the vlue of the
terminating line, for multiplying by 1 does not change the vlue
of a multiplication.
When building a vlue with cons, always consider ( ) for the vlue
of the terminating line.
What is ( tup+ tup1 tup2)
where
tup1 is (3 6 9 1 1 4)
and
tup2 is (8 5 2 0 7)
Numbers Games
( 1 1 1 1 11 11 1 1) .
67
What is ( tup+ tup1 tup2)
where
tup1 is (2 3)
and
tup2 is (4 6)
What does ( tup+ tup1 tup2) do?
If you recur on one tup how many questions
When recurring on two tups, how many
Do you mean the questions
(and ( null ' tup1 ) ( nul tup2) )
( nul tup1 )
( nul tup2)
and
else
Can the frst tup be () at the same time as
the second is other than ( )
Does this mean
(and ( nul tup1 ) ( null ' tup2) )
and
else
are the only questions we need to ask?
68
¸69).
It ads the frst number of tup1 to the frst
number of tup2, then it ads the second
number of tup1 to the second number of
tup2, and so on, building a tup of the
answers, for tups of the same length.
It looks at each element of two tups at the
same time, or in other words, it recurs on
two tups.
Two, they are ( null ? tup) and else.
Four: if the frst tup is empty or non-empty,
and if the second tup is empty or non-empty.
Yes.
No, because the tups must have the same
length.
Yes,
because ( null ' tup1 ) is true exactly when
( nul tup2) is true.
Cbapter 4
Write the function tup+
What are the arguments of + in the last line?
What are the arguments of cons in the last
line?
What is ( tup+ tup1 tup2)
where
tup1 is (3 /)
and
tup2 is ¸46)
( null ? tup1 )
( cons
( + ( car tup1 ) ( car tup2) )
( tup+ ( cdr tup1 ) ( cdr tup2) ) )
Why does the natural recursion include the
cdr of both arguments?
( null ? tup1 )
where
tup1 is now (7)
and
tup2 is now (6)
Numbers Games
(defne tup+
(lambda ( tup1 tup2)
(cond
( ( and ( null? tup1 ) ( null ? tup2) )
(quote ( ) ) )
(else
( cons ¸+ ( car tup1 ) ( car tup2) )
( tup+
( cdr tup1 ) ( cdr tup2) ) ) ) ) ) )
( car tup1 ) and ( car tup2) .
( + ( car tup1 ) ( car tup2) ) and
( tup+ ( cdr tup1 ) ( cdr tup2) ) .
(7 13).
But let ' s see how it works.
No.
cons 7 onto the natural recursion:
( tup+ ( cdr tup1 ) ( cdr tup2) ) .
Because the typical element of the fnal value
uses the car of both tups, so now we are
ready to consider the rest of both tups.
No.
69
( cons
( + ( car tup1 ) ( car tup2) )
( tup+ ( cdr tup1 ) ( cdr tup2) ) )
( null ? tup1 )
Then, what must be the value?
What is the value of the application?
What problem arises when we want
( tup+ tup1 tup2)
where
tup1 is (3 7)
and
tup2 is ( 4 6 8 1)
Can we still write tup+ even i f the tups are
not the same length?
What new terminal condition line can we
add to get the correct fnal value?
What is ( tup+ tup1 tup2)
where
tup1 is (3 7 8 1)
and
tup2 is (4 6)
What do we need to include in our function?
What does the second new line look like?
70
cons 13 onto the natural recursion.
Yes.
( ) , because ( null ? tup2) must be true.
(7 13) . That is, the cons of 7 onto the cons
of 13 onto ¸ ) .
No answer, since tup1 will become null
before tup2.
See The First Commandment: We did not
But, we would like the fnal value to be
(7 13 8 1) .
Yes!
( ( null ? tup1 ) tup2) .
No answer, since tup2 will become null
before tup1.
See The First Commandment: We did not
We need to ask two more questions:
( null ? tup1 ) and ( null ? tup2) .
( ( null ? tup2) tup1 ) .
Chapter 4
Here is a defnition of tup+ that works for
any two tups:
(defne tup+
(lambda ( tup1 tup2)
(cond
((and ( null ? tup1 ) ( null ? tup2) )
(quote ( ) ) )
( ( null ? tup1 ) tup2)
( ( null ? tup2) tup1 )
(else
( cons ¸+ ( car tup1 ) ( car tup2) )
( tup+
( cdr tup 1 ) ( cdr tup2) ) ) ) ) ) )
Can you simplify it?
(defne tup+
(lambda ( tup1 tup2)
(cond
( ( null ? tup1 ) tup2)
( ( null ? tup2) tup1 )
¸else
( cons ( + ( car tup1 ) ( car tup2) )
( tup+
( cdr tup1 ) ( cdr tup2) ) ) ) ) ) )
Does the order of the two terminal conditions No.
matter?
Is else the last question?
What is (> 12 133)
What is ( > 120 11)
On how many numbers do we have t o recur?
How do we recur?
When do we recur?
How many questions do we have to ask
Numbers Games
Yes, because either ( null ? tup1 ) or
( null ? tup2) is true if either one of them does
not contain at least one number.
#f-false.
#t -true.
Two, n and Û-
With (sub1 n) and ( sub1 Û) -
When we know neither number is equal to 0.
Three: ( zero ? n), ( zer? M) ¿ and else.
71
Can you write the function > now using
zer? and sub1
Is the way we wrote ¸ > 7 Û) correct?
( zer? 3)
(zer? 3)
What is the meaning of
¸ > (sub1 7) (sub1 Û) )
( zer? 2)
( zer? 2)
What is the meaning of
¸ > ( sub1 7) (sub1 Û) )
(zero ? 1)
(zer? 1)
What i s the meaning of
¸ > ( sub1 7) (sub1 Û) )
72
(defne >
(lambda ( 7 Û)
(CLnd
( ( zer? Û) #t )
( ( zer? 7) #f )
(else ¸ > ( sub1 7) ( sub1 Û) ) ) ) ) )
No, try i t for the case where 7 and Û are the
same number. Let 7 and Û be 3.
No, so move to the next question.
No, so move to the next question.
Recur, but with both arguments reduced by
one.
No, so move to the next question.
No, so move to the next question.
Recur, but with both arguments closer to
zero by one.
No, so move to the next question.
No, so move to the next question.
Recur, but with both arguments reduced by
one.
Chapter 4
(zer? 0) Yes, so the value of ¸ > 7 Û) is #t .
Is this correct? No, because 3 is not greater than 3.
Does the order of the two terminal conditions Think about it.
matter?
Does the order of the two terminal conditions Ty it out !
matter?
Does the order of the two previous answers
matter?
How can we change the function > to take
care of this subtle problem?
What is ( < 4 6)
( < 8 3)
(< 6 6)
Now try to write <
Number Games
Yes. Think frst , then try.
Switch the zero ? lines:
(defne >
(lambda ( 7 Û)
#t .
#f.
#f.
(cond
( ( zero ? 7) #f )
( ( zero ? Û) #t )
(else ¸ > (sub1 7) ( sub1 Û) ) ) ) ) )
(defne <
(lambda ( 7 Û)
(cond
( ( zero ? Û) #f )
( ( zer? 7) #t )
(else ( < (sub1 7) ( sub1 Û) ) ) ) ) )
73
Here is the defnition of =
(defne =
(lambda ¸ H Û)
(cond
( ( zer? Û) ( zero ? H) )
( ( zero ? H) #f )
(else (= ( sub1 H) ( sub1 Û) ) ) ) ) )
Rewrite = using < and >
Does this mean we have two diferent
functions for testing equality of atoms?
¸¡ 1 1 )
¸¡2 3)
¸¡ 5 3)
Now write the function ¡
Hint : See the The First and Fifth
Commandments.
What is a good name for this function?
(defne ???
(lambda ( H Û)
(cond
( ( < H Û) 0)
(else ( add1 ( ??? ( ¯ H Û) Û) ) ) ) ) )
74
(defne =
(lambda ( H Û)
(cond
( ( > H Û) #f )
( ( < H Û) #f )
(else #t ) ) ) )
Yes, they are ° for atoms that are numbers
and eq? for the others.
1 .
8.
125.
(defne ¡'
(lambda ( H Û)
(cond
( ( zero ? Û) 1)
(else ¸ 7 H ¸ ¡H ( sub1 Û) ) ) ) ) ) )
' L, S: This i s like expt.
We have never seen this kind of defnition
before; the natural recursion also looks
strange.
Chapter 4
What does the frst question check?
And what happens in the second line?
So what does the function do?
And what do we call this?
What is ( + 15 4)
How do we get there?
It determines whether the frst argument is
less than the second one.
We recur with a frst argument from which
we subtract the second argument . When the
function returns, we add 1 to the result .
It counts how many times the second
argument fts into the frst one.
Division.
(defne +1
(lambda (n m)
(cond
( ( < n m) O)
(else ( add1 ( + (

n m) m) ) ) ) ) )
¯ L: (defu quotient ( n m}
(values (trucate (/ n m} } } }
S : This is like quotient.
Eay, it is 3.
Eay, too:
( ÷ 15 4) ¯ 1 + ( + 11 4)
1 + ( 1 + ( + 7 4))
1 + ( 1 + ( 1 + ( + 3 4) ) )
¯ 1 + ( 1 + ( 1 + 0) ) .
Wouldn' t a (ham and cheese on rye) be good right now?
Don' t forget the mustard!
Numbers Games 75
What is the value of ( length lat)
where
lat is ¸ hotdogs with mustard sauerkraut
and pi ckl es)
What is ( length lat)
where
lat is ¸ ham and cheese on rye)
Now try to write the function length
What is (pick n lat)
where n is 4
and
lat is ¸ l asagna spaghetti ravi ol i
macaroni meatbal l )
What is (pick 0 lat)
where lat is ¸a)
Ty to write the function pick
What is ¸ ÍÛ pick n lat)
where n is 3
and
lat is ¸ hotdogs wi th hot mustard)
76
6.
5.
(defne length
(lambda ( lat)
(cond
( ( null ? lat) 0)
(else ( add1 ( length ( cdr lat) ) ) ) ) ) )
macaroni .
(defne pick
(lambda ( n lat)
(cond
( ( zer? ( sub1 n) ) ( car lat) )
(else (pick ( sub1 n) ( cdr lat) ) ) ) ) )
¸ hotdogs wi th mustard) .
Chapter 4
Now try to write rmpick
Is ( number?1 a) true or false
where a is tomato
1 L: nuberp
Is ( number? 76) true or false?
Can you write number? which is true if its
argument is a numeric atom and false if it is
anthing else?
Now using number? write the function
no-nums which gives Ü a fnal value a lat
obtained by removing all the numbers from
the lat . For example,
where
lat is (5 pears 6 prunes 9 dates)
the value of ( no-nums lat) is
(pears prunes dates)
Numbers Games
(defne rm pick
(lambda ( n lat)
(cond
False.
Tue.
( ( zero ? ( sub1 n) ) ( cdr lat) )
(else ( cons ( car lat)
( rmpick (sub1 n)
( cdr lat ) ) ) ) ) ) )
No: number?, like add1 , sub1 , zero ?, car,
cdr, cons , null ?, eq?, and atom?, is a
primitive function.
(defne no-nums
(lambda ( lat)
(cond
( ( null ? lat) (quote ( ) ) )
(else (cond
( ( number? ( car lat) )
( no-nums ( cdr lat ) ) )
(else ( cons ( car lat)
( no-nums
( cdr lat ) ) ) ) ) ) ) ) )
77
Now write all-nums which extracts a tup
from a lat using all the numbers in the lat .
Write the function eqan? which is true if its
two arguments ( a1 and a2) are the same
atom. Remember to use ¯ for numbers and
eq? for all other atoms.
Can we assume that all functions written
using eq? can be generalized by replacing eq?
by eqan?
Now write the function occur which counts
the number of times an atom a appears in a
lat
(defne occur
(lambda (a lat)
(cond
78
'
(else
(cond
'
'

)
) ) ) ) ) )
(defne all-nums
(lambda ( lat)
(cond
( ( null ? lat) (quote ( ) ) )
(else
(cond
( ( number? ( car lat) )
( cons ( car lat)
( all-nums ( cdr lat) ) ) )
(else ( all-nums ( cdr lat) ) ) ) ) ) ) )
(defne eqan?
(lambda ( a1 a2)
(cond
((and ( number? a1 ) ( number? a2) )
¸ ¯ a1 a2) )
( (or ( number? a1 ) ( number? a2) )
#f )
(else ( eq? a1 a2) ) ) ) )
Yes, except , of course, for eqan? itself.
(defne occur
(lambda ( a lat)
(cond
( ( null ? lat) 0)
(else
(cond
( ( eq? ( car lat) a)
( add1 ( occur a ( cdr lat) ) ) )
(else ( occur a ( cdr lat ) ) ) ) ) ) ) )
Chapter 4
Write the function one ? where ( one ? n) is #t
if n is 1 and #f ¸i. e. , false) otherwise.
Guess how we can further simplify this
function, making it a one-liner.
Now rewrite the function rmpick that
removes the Î`¯ atom from a lat. For
example,
where
n is 3
and
lat is ( l emon meri ngue sal ty pi e)
the value of ( rm pick n lat) is
(l emon meri ngue pi e)
Numbers Games
(defne one ?
(lambda ( n)
(cond
( ( zer? n) #f )
(else ( zer? ( sub1 n) ) ) ) ) )
or
(defne one ?
(lambda ( n)
(cond
(else ( = n 1) ) ) ) )
By removing the ( cond . . . ) clause:
(defne one ?
(lambda ( n)
( = n 1) ) )
(defne rmpick
(lambda ( n lat)
(cond
( ( one ? n) ( cdr lat) )
(else ( cons ( car lat)
( rmpick ( sub1 n)
( cdr lat ) ) ) ) ) ) )
79
�'
�-·
What is ( rmber* a l}
where a is cup
and
l is ((cofee) cup ((tea) cup)
(and (hick} } cup)
"rmber*" is pronounced "rember-star. "
What is ( rmber* a l}
where a is sauce
and
l is (((tomato sauce))
((bean) sauce)
(and ((flyi ng)) sauce))
Now write rmber*
t
Here is the skeleton:
(defne rmber*
(lambda ( a l}
(cond
'
'
(

¹
¹
) ) ) )
t
" . . . ¯ ma 08 think "oh D gawd."
( lat ? l}
where
l is (( (tomato sauce))
((bean) sauce)
(and ( (flyi ng) ) sauce))
*
Oh My Gawd
*
: It 's Fl of Stas
((cofee) ((tea) ) (and (hick) ) } .
(( (tomato) )
((bean) )
( and ( (flying) ) ) ) .
(defne rmber*
(lambda ( a l}
(cond
( ( nUll ? l} (quote ( ) ) )
( ( atom? ( car l}}
(cond
( ( eq? ( car l} a)
( rmber* a ( cdr l )))
¸else ( cons ( car l)
( rmber* a ( cdr l ) ) } ) ) )
¸else ( cons ( rmber* a ( car l } }
( rmber* a ( cdr l ) ) } ) ) ) )
Using arguments from one of our previous
examples, follow through this to see how it
works. Notice that now we are recurring
down the car of the list , instead of just the
cdr of the list .
#f.
81
Is ( cÛr l) an atom
where
l is ( ( (tomato sauce) )
( ( bean) sauce)
(and ( (flyi ng) ) sauce) )
What is ( insertR * new old l)
where
new is roast
old is chuck
and
l is ( ( ho much (wood) )
coul d
( ( a (wood) chuck) )
( ( (chuck) ) )
( i f ( a) ((wood chuck) ) )
coul d chuck wood)
Now write the function insertR * which
inserts the atom new to the right of old
regardless of where old occurs.
(defÎe insertR *
(lÜbda ( new old l)
(LoÎd
( )
( )
( ) ) ) )
How are insertR * and rmber* similar?
82
No.
( ( how much (wood) )
coul d
((a (wood) chuck roast) )
( ( (chuck roast) ) )
( i f (a) ((wood chuck roast) ) )
coul d chuck roast wood) .
(defÎe insertR *
(lambda ( new old l)
(LoÎd
( ( null ? l) (quote ( ) ) )
( ( atom? ( car l ) )
(LoÎd
(( eq? ( car l) old)
( cons old
¸ cons new
( insertR * new old
( cdr l ) ) ) ) )
(else ¸ cons ¸ car l )
( insertR * new old
( cdr l ) ) ) ) ) )
(else ( cons ( insertR * new old
( car l ))
( insertR * new old
¸ cdr l ) ) ) ) ) ) )
Chapter 5
The First Commandment
(fnal version)
When recurring on a list of atoms, lat , ask two questions
about it: ( null ? lat ) and else.
it: (zero ? n) and else.
When recurring on a list of S-expressions, l , ask three
question about it: ( null ? l ) , ( atom? ( car l ) ) , and else.
How are insertR * and rmber* similar? Each function recurs on the car of its
argument when it fnds out that the
argument ' s car is a list .
How are rmber* and multirmber diferent?
How are insertR * and rmber* similar?
How are all *-functions similar?
Why?
*
Ob My Gawd
*
: It 's Fl of Stars
The function multirmber does not recur
with the car. The function rmber* recurs
with the car as well as with the cdr. It
recurs with the car when it fnds out that
the car is a list .
They both recur with the car, whenever the
car is a list , as well as with the cdr.
They all ask three questions and recur with
the car as well as with the cdr, whenever the
car is a list.
Because all *-functions work on lists that are
either
- empty,
- an atom consed onto a list , or
- a list consed onto a list .
83
The Fourth Commandment
(fnal version)
Always change at least one argument while recurring.
When recurring on a list of atoms, lat , use ( cdr lat ) . When
recurring on a number, n, use ( sub1 n) . And when recur­
ring on a list of S-expressions, l , use ( car l) and ( cdr l) if
neither ( null ? l ) nor ( atom? ( car l ) ) are true.
It must be changed to be closer to termination. The chang­
ing argument must be tested in the termination condition:
when using cdr, test termination with null ? and
when using sub1 , test termination with zero ?.
( occursomething Ú l)
where
5.
Ú i s banana
and
l is ( ( banana)
(spl i t ( ( ( ( banana i ce) ) )
(cream ( banana) )
sherbet) )
( banana)
(banana brandy) )
What is a better name for
occursomething
84
occur*.
Chapter 5
Write occur*
(dÜfne occur*
(lambda (a l)
(cond
<
< -
( --
( subst * new old l)
where
new is orange
old is banana
l is ( ( banana)

)

)
) ) ) )
(spl i t ( ( ( ( banana i ce) ) )
(cream (banana) )
sherbet) )
(banana)
(banana brandy) )
Write subst *
(defne subst *
(lambda ( new old l)
¸cond
< -
<
<

)

)
) ) ) )
*Oh My Gawd*: It 's Fll of Stars
(defne occur*
(lambda (a l)
(cond
( ( null ? l) 0)
( ( atom? ( car l ) )
(cond
( ( eq? ( car l) a)
( add1 ( occur* a ( cdr l ) ) ) )
(else ( occur* a ( cdr l ) ) ) ) )
(else ( + ( occur* a ( car l ) )
( occur* a ( cdr l ) ) ) ) ) ) )
((orange)
(spl i t ( ( ( (orange i ce) ) )
(cream (orange) )
sherbet) )
(orange)
(orange brandy) ) .
¸defne subst*
(lambda ( new old l)
¸cond
( ( null ? l) (quote ( ) ) )
( ( atom? ( car l ) )
¸cond
(( eq? ( car l) old)
( cons new
( subst * new old ( cdr l ) ) ) )
¸else ( cons ( car l)
(subst * new old
( cdr l ) ) ) ) ) )
¸else
( cons ( subst* new old ( car l ) )
( subst* new old ( cdr l ) ) ) ) ) ) )
85
What is ( insertL * new old l)
where
new is peeker
old is chuck
and
l is ( ( how much (wood) )
coul d
( (a (wood) chuck))
( ( (chuck) ) )
( i f ( a) ( (wood chuck) ) )
coul d chuck wood)
Write insertL *
(defne insertL *
(lambda ¸ new old l )
(Lond
(
¯
(

(

¸ member* Ú l)
where Ú is chi ps
and

)

)
) ) ) )
l is ( ( potato) (chi ps ( (wi th) fi sh) (chi ps) ) )
86
( ( how much (wood) )
coul d
((a (wood) peeker chuck))
( ( ( peeker chuck) ) )
( i f (a) ((wood peeker chuck) ) )
coul d peeker chuck wood) .
(defne insertL *
(lambda ¸ new old l)
(Lond
( ¸ null? l) (quote ¸ ) ) )
¸ ¸ atom? ¸ car l) )
(Lond
( ( eq? ¸ car l) old)
¸ cons new
¸ cons old
( insertL * new old
¸ cdr l) ) ) ) )
(else ¸ cons ¸ car l)
( insertL * new old
¸ cdr l) ) ) ) ) )
(else ¸ cons ( insertL * new old
¸ car l) )
( insertL * new old
¸ cdr l) ) ) ) ) ) )
#t g because the atom chi ps appears i n the
list l.
Chapter 5
Write member*
(defne member*
(lambda ¸ Û l)
(cond
(

(

(

¯
)

)
) ) ) )
What i s ( member* a l)
where
Û is chi ps
and
l is ( (potato) (chi ps ( (wi th) fi sh) (chi ps) ) )
Which chi ps did i t fnd?
What is ( lefmost l)
where
l is ( ( potato) (chi ps ( (wi th) fi sh) (chi ps) ) )
What i s ( lefmost l)
where
l is ( ( ( hot) (tuna (and) ) ) cheese)
What is ( lefmost l)
where
l is ( ( ( ( ) four)) 17 (seventeen) )
What i s ( lefmost (quote ( ) ) )
Ca you describe what lefmost does?
*
Oh My Gawd
*
: It 's Fn of Stars
(defne member*
(lambda (a l)
(cond
#t .
( ( null ? l) #f )
( ( atom? ( car l) )
(or ( eq? ( car l) a)
( member* a ( cdr l) ) ) )
(else (or ( member* a ( car l ) )
( member* a ( cdr l ) ) ) ) ) ) )
( ( potato) (chi ps ( (wi th) fi sh) (chi ps) ) ) .
potato.
hot.
Here is our description:
"The function lefmost fnds the leftmost
atom in a non-empty list of S-expressions
that does not contain the empty list. "
87
Is lefmost a *-function?
three possible cases?
Now see if you can write the function
lefmost
(defne lefmost
(lambda ( l )
(LLnd
<
<

)
) ) ) )
Do you remember what ¸LÏ . ) does?
What is
(and ( atom? ( car l ) )
( eq? ( car l) x) )
where
x is pi zza
and
l is (mozzarel l a pi zza)
Why is it false?
88
It works on lists of S-expressions, but it only
recurs on the car.
No, it only needs to ask two questions. We
agreed that lefmost works on non-empty
lists that don' t contain empty lists.
(defne lefmost
(lambda ( l )
(LLnd
( ( atom? ( car l ) ) ( car l ))
(else ( lefmost ( car l ) ) ) ) ) )
¸LÏ . . ) asks questions one at a time until it
fnds one that is true. Then ¸LÏ . ) stops,
making its value true. If it cannot fnd a true
argument , the value of ¸LÏ · · ) is false.
#f.
Since (and . . . ) asks ( atom? ( car l ) ) , which
is true, it then asks ( eq? ( car l) x ) , which is
false; hence it is #f.
Chapter 5
What is
(and ( atom? ( car l ) )
( eq? ( car l) x) )
where
x is pi zza
and
l is ¸ ¸mozzarel l a mushroom) pi zza)
Why is it false?
Give an example for x and l where
(and ( atom? ( car l ) )
( eq? ( car l) x) )
i s true.
Put in your own words what (and . . . ) does.
Tue or false: it is possible that one of the
arguments of (and . . . ) and ¸LÏ . . . ) is not
considered? '
' ( cond . . . ) also ha the property of not considering all of
its arguments. Because of this property, however, neither
(and . . . ) nor (or . . . ) can be def ned a functions in terms
of (cond . . . ) , though both (and . . . ) and (or . . . ) can be
expressed a abbreviations of ( cond . . . ,-expressions:
(and t {) ~ (cond (t {) (else #f))
and
(or t {) ~ (cond , t #t) (else {))
( eqlist ? l1 l2)
where
l1 is ¸strawberry ice cream)
and
l2 is ¸strawberry ice cream)
*Oh My Gawd*: It 's Fll of Stars
#f.
Since (and . . . ) asks ( atom? ( car l ) ) , and
( car l) is not an atom; so it is #f.
Here' s one:
x is pi zza
and
l is ¸ pi zza ¸tastes good) ) .
We put it in our words:
"(and . . . ) asks questions one at a time
until it fnds one whose value is false. Then
(and . . . ) stops with false. If none of the
expressions are false, (and . . . ) is true. "
Tue, because (and . . . ) stops if the frst
argument has the value #f, and ¸LÏ )
stops if the frst argument has the value #t .
#t .
89
( eqlist ? l1 l2)
where
l1 is (strawberry i ce cream)
and
l2 is (strawberry cream i ce)
( eqlist ? l1 l2)
where
l1 is (banana ( (spl i t) ) )
and
l2 is ( ( banana) (spl i t) )
( eqlist ? l 1 l2)
where
l1 is (beef ( (sausage) ) (and (soda) ) )
and
l2 is (beef ( (sal ami ) ) (and (soda) ) )
( eqlist ? l 1 l2)
where
l1 is (beef ( (sausage) ) (and (soda) ) )
and
l2 is ( beef ( (sausage) ) (and (soda) ) )
What i s eqlist ?
How many questions will eqlist ? have to ak
Can you explain why there are nine
questions?
90
#f.
#f.
#f, but almost #t .
#t . That ' s better.
It is a function that determines if two lists
are equal.
Nine.
Here are our words:
"Each argument may be either
- empty,
- an atom consed onto a list , or
- a list consed onto a list .
For example, at the same time a the frst
argument may be the empty list , the
second argument could be the empty list or
have an atom or a list in the LÚÍ position. "
Chapter 5
Write eqlist ? using eqan?
Is it oky to ask ( atom? ( car l2) ) in the
second question?
And why is the third question ( null ? l1 )
Tue or false: if the frst argument is ( )
eqlist ? responds with #t i n only one case.
*Oh My Gawd
*
: It 's Fl of Stars
(defne eqlist ?
(lambda ( l1 l2)
(Lond
( ( and ( null ? l1 ) ( null ? l2) ) #t )
( ( and ( null ? l1 ) ( atom? ( car l2) ) )
#f )
( ( null? l1 ) #f )
( ( and ( atom? ( car l1 ) ) ( null ? l2) )
#f )
( ( and ( atom? ( car l1 ) )
( atom? ( car l2) ) )
(and ( eqan? ( car l1 ) ( car l2) )
( eqlist ? ( cdr l 1 ) ( cdr l2) ) ) )
( ( atom? ( car l1 ) ) #f )
( ( null? l2) #f )
( ( atom? ( car l2) ) #f )
(elÜe
(and ( eqlist ? ( car l1 ) ( car l2) )
( eqlist ? ( cdr l 1 ) ( cdr l2) ) ) ) ) ) )
Yes, because we know that the second list
cannot be empty. Otherwise the frst
question would have been true.
At that point , we know that when the frst
argument is empty, the second argument is
neither the empty list nor a list with Ü atom
as the frst element . If ( null ? l1 ) is true now,
the second argument must be a list whose
frst element is also a list.
Tue.
For ( eqlist ? (quote ( ) ) l2) to be true, l2
must also be the empty list.
91
Does this mean that the questions
(and ( null ? 11 ) ( null ? l2) )
and
¸LÏ ( null ? 11 ) ( null ? 12) )
sufce to determine the answer i n the frst
three cases?
Rewrite eqlist ?
What i s an S-expression?
How many questions does equal? ask to
determine whether two S-expressions are the
same?
Write equal ?
92
Yes. If the frst question is true, eqlist ?
responds with #t ; otherwise, the answer is
#f.
(defne eqlist ?
(lambda ( l1 12)
(LLnd
((and ( null? 11 ) ( null? 12) ) #t )
¸ ¸LÏ ( null ? 11 ) ( null ? 12) ) #f )
((and ( atom? ( car l1 ) )
( atom? ( car 12) ) )
(and ( eqan? ( car 11 ) ( car 12) )
( eqlist ? ( cdr 11 ) ( cdr 12) ) ) )
¸ ¸L1 ( atom? ( car l1 ) )
( atom? ( car 12) ) )
#f )
(else
(and ( eqlist ? ( car 11 ) ( car 12) )
( eqlist? ( cdr 11 ) ( cdr l2) ) )))))
An S-expression i s either an atom or a
(possibly empty) list of S-expressions.
Four. The frst argument may be an atom or
a list of S-expressions at the same time a
the second argument may be an atom or a
list of S-expresssions.
(defne equal ?
(lambda (s1 s2)
(LLnd
((and ( atom? s1 ) ( atom? s2))
( eqan? s1 s2) )
( ( atom? s1 ) #f )
( ( atom? s2) #f )
(else ( eqlist? s1 s2) ) ) ) )
Chapter 5
Why is the second question ( atom? s1 )
And why is the third question ( atom? s2)
Ca we summarize the second question and
the third question as
¸L1 ( atom? s1 ) ( atom? s2) )
Simplify equal ?
Does equal ? ask enough questions?
Now, rewrite eqlist ? using equal?
*Ob My Gawd*: It 's Fll of Stars
If it is true, we know that the frst argument
is an atom and the second argument is a list .
By the time we ask the third question we
know that the frst argument is not an atom.
So all we need to know in order to
distinguish between the two remaining cases
is whether or not the second argument is an
atom. The frst argument must be a list.
Yes, we can!
(defne equal ?
(lambda ( s1 s2)
(LLnd
Yes.
((and ( atom? s1 ) ( atom? s2) )
( eqan? s1 s2) )
¸ ¸L1 ( atom? s1 ) ( atom? s2) )
#f )
(else ( eqlist ? s1 s2) ) ) ) )
The questions cover all four possible cases.
(defne eqlist ?
(lambda ( l 1 l2)
(LLnd
( ( and ( null ? l1 ) ( null ? l2) ) #t )
¸ ¸L1 ( null ? l1 ) ( null? l2) ) #f )
(else
(and ( equal ? ( car l1 ) ( car l2) )
( eqlist ? ( cdr l 1 ) ( cdr l2) ) ) ) ) ) )
93
The Sixth Commandment
Simplif only aer the function is correct.
Here is rmber after we replace lat by a list l
of S-expressions and a by any S-expression.
(defne rmber
(lambda ( s l)
(cond
( ( null ? l) (quote ( ) ) )
( ( atom? ( car l ) )
(cond
( ( equal? ( car l) s) ( cdr l ) )
(else ( cons ( car l)
( rmber s ( cdr l ) ) ) ) ) )
(else (cond
( ( equal ? ( car l) s) ( cdr l ) )
(else ( cons ( car l)
( rmber s
( cdr l ) ) ) ) ) ) ) ) )
Can we simplify it?
And how does that difer?
Is rmber a "star" function now?
Why not ?
Can rmber be further simplifed?
94
Obviously!
(defne rmber
(lambda (s l)
(cond
( ( null ? l) (quote ( ) ) )
(else (cond
( ( equal ? ( car l) s) ( cdr l ))
(else ( cons ( car l)
( rmber s
( cdr l ) ) ) ) ) ) ) ) )
The function rmber now removes the frst
matching S-expression ö in l, instead of the
frst matching atom a in lat.
No.
Because rmber recurs with the cdr of l only.
Yes, the inner ( cond . . . ) asks questions that
the outer ( cond . . . ) could ak!
Chapter 5
Do it !
Does this new defnition look simpler?
And does it work j ust as well?
Simplify insertL*
When functions are correct and
well-designed, we can think about them
easily.
Can all functions that use eq? and = be
generalized by replacing eq' and = by the
function equal ?
*Oh My Gawd
*
: It 's Fll of Stars
(defne rmber
(lÜbda ( s l)
(cond
( ( null ? l) (quote ( ) ) )
( ( equal? ( car l) s) ( cdr l ) )
(else ( cons ( car l)
( rmber s ( cdr l ) ) ) ) ) ) )
Yes, i t does!
Yes, because we knew that all the cases and
all the recursions were right before we
simplifed.
We can' t. Before we can ask ( eq' ( car l) old)
we need to know that ( car l) is an atom.
And that saved us this time from getting it
wrong.
Not quite; this won' t work for eqan?, but will
work for all others. In fact , disregarding the
trivial example of eqan', that is exactly what
we shall assume.
95
(E

d

l@

Is 1 Ü arithmetic expression?
Is 3 an arithmetic expression?
Is 1 + 3 an arithmetic expression?
Is 1 + 3 ? 4 an arithmetic expression?
And, what about 3 t y + 5
What is an arithmetic expression in your
words?
What is (quote a)
What is (quote +)
What does (quote ? ) stand for?
Is ( eqP (quote a) ç)true or false
where ç is a
Is ( eqP x ç) true or false
where x is a
and
çis a
Yes.
Yes, of course.
Yes!
Defnitely.
Yes. Are you almost ready for one?
Yes.
In ours:
"For the purpose of this chapter, an
arithmetic expression is either an atom
(including numbers) , or two arithmetic
expressions combined by +, ? ¿ or t. "
a.
The atom +, not the operation + .
The atom ? ¿ not the operation ? .
Tue.
That ' s the same question again. And the
97
Is ¸ n + 3) an arithmetic expression?
Could we think of ¸ n + 3) Ü an arithmetic
expression?
What would you call ¸ n + 3)
Why is ¸ n + 3) a good representation?
Tue or false: ( numberd? x)
where x is 1
How do you represent 3 + 4 7 5
Tue or false: ( numberd? y)
where y is (3 + (4 t 5) )
Tue or false: ( numberd? Z)
where Z is (2 7 sausage)
What is numberd?
98
Not really, since there are parentheses
around n + 3. Our defnition of arithmetic
expression does not mention parentheses.
Yes, if we keep in mind that the parentheses
are not really there.
We call it a representation for n + 3.
Because
1. ¸ n + 3) is an S-expression.
It can therefore serve Ü an argument for a
function.
2. It structurally resembles n + 3.
Tue.
(3 + (4 ? 5) ) .
Tue.
False, because sausage is not a number.
It is a function that determines whether Ü
representation of an arithmetic expression
contains only numbers besides the +, 7 ¿
and t.
Chapter 6
Now can you write a skeleton for numberd?
What is the frst question?
What is ( eq? ( car ( cdr aex)) (quote +) )
Ca you guess the third one?
And you must know the fourth one.
Should we ak another question about aexp
Why do we ak four, instead of two,
After all , arithmetic expressions like (1 + 3)
are lats.
(defne numberd?
(lambda ( aexp)
(cond
(
¯¯
(

'
(

is a good guess.
( atom? aex ) .

)

)

)
) ) ) )
It i s the second question.
( eq? ( car ( cdr aex) ) (quote 7 ) ) is perfect .
( eq? ( car ( cdr aexp) ) (quote ¡) ) , of course.
No! So we could replace the previous
question by else.
Because we consider (1 + 3) as a
representation of an arithmetic expression in
list form, not a a list itself. And, an
arithmetic expression is either a number, or
two arithmetic expressions combined by +,
7 ¿ or t.
99
Now you can almost write numberd?
Why do we ask ( number? aex) when we
know that aex is an atom?
What do we need to know if the aexp consists
of two arithmetic expressions combined by +
In which position is the frst subexpression?
In which position is the second
subexpression?
So what do we need to ask?
100
Here is our proposal:
(defne numberd?
(lambda ( aex)
(cond
( ( atom? aexp) ( number? ae))
( ( eq? ( car ( cdr aex) ) (quote +) )
. . . )
( ( eq? ( car ( cdr aex)) (quote ?) )
. . . )
( ( eq? ( car ( cdr aex) ) (quote t) )
. . . ) ) ) )
Because we want t o know i f all arithmetic
expressions that are atoms are numbers.
We need to fnd out whether the two
subexpressions are numbered.
It is the car of aex.
It is the car of the cdr of the cdr of aex.
( numberd? ( car aex)) and
( numberd? ( car ( cdr ( cdr aex) ) ) ) .
Both must be true.
(and ( numberd? ( car aex) )
( numberd? ( car ( cdr ( cdr aex) ) ) ) )
Chapter 6
Ty numberd? again.
Since aex wa already understood to be an
arithmetic expression, could we have written
numberd? in a simpler way?
Why can we simplify?
What is ( value u)
where u is 13
( value x)
where
J is (1 + 3)
(defne numberd?
(lambda ( aex)
(cond
( ( atom? aex) ( number? aex) )
( ( eq? ( car ( cdr aexp) ) (quote +) )
(and ( numberd? ( car aexp) )
( numberd?
( car ( cdr ( cdr aex) ) ) ) ) )
( ( eq? ( car ( cdr aex) ) (quote 7 ) )
( and ( numberd? ( car aex) )
( numberd?
( car ( cdr ( cdr aex) ) ) ) ) )
( ( eq? ( car ( cdr aex) ) (quote t) )
(and ( numberd? ( car aex) )
( numberd?
( car ( cdr ( cdr aex) ) ) ) ) ) ) ) )
Yes:
(defne numberd?
(lambda ( aex)
(cond
( ( atom? aex) ( number? aexp) )
(else
(and ( numberd? ( car aex) )
( numberd?
( car ( cdr ( cdr aex) ) ) ) ) ) ) ) )
Because we know we' ve got the function
right .
13.
4.
101
( value y)
where
y is (1 + (3 ¡ 4) )
( value Z)
where Z is cooki e
( value nex) returns what we think is the
natural value of a numbered arithmetic
expression.
82.
We hope.
How many questions does value ak about Four.
nexp
Now, let ' s attempt to write value
What is the natural value of an arithmetic
expression that is a number?
What is the natural value of an arithmetic
expression that consists of two arithmetic
expressions combined by +
Can you think of a way to get the value of
the two subexpressions in ( 1 + (3 7 4) )
102
(defne value
(lambda ( nex)
(cond
( ( atom? nex) . . . )
( ( eq? ( car ( cdr nex)) (quote +) )
. . . )
( ( eq? ( car ( cdr nex)) (quote 7 ) )
. . . )
(else . . . ) ) ) )
It i s j ust that number.
If we had the natural value of the two
subexpressions, we could j ust add up the two
values.
Of course, by applying value to 1 , and
applying value to (3 7 4) .
Chapter 6
And in general? By recurring with value on the
subexpressions.
The Seventh Commandment
Recur on the subparts that are of the same nature:
• On the sublists of a list.
• On the subexpressions of an arithmetic expression.
Give value another try.
Ca you think of a diferent representation of
arithmetic expressions?
Could (3 4 +) represent 3 + 4
Could ¸ + 3 4)
Or (pl us 3 4)
(defne value
(lambda ( nex)
(cond
( ( atom? nex) nex)
( ( eq? ( car ( cdr nexp) ) (quote +))
( + ( value ( car nex) )
( value ( car ( cdr ( cdr nex) ) ) ) ) )
( ( eq? ( car ( cdr nex) ) (quote 7 ) )
( 7 ( value ( car nex ) )
( value ( car ( cdr ( cdr nex) ) ) ) ) )
(else
( ¡ ( value ( car nexp) )
( value
( car ( cdr ( cdr nexp) ) ) ) ) ) ) ) )
There are several of them.
Yes.
Yes.
Yes.
103
Is ( + ¸ 7 3 6) (t 8 2) ) a representation of an
arithmetic expression?
Ty to write the function value for a new
kind of arithmetic expression that is either:
- a number
- a list of the atom + followed by
two arithmetic expressions,
- a list of the atom 7 followed by
two arithmetic expressions, or
- a list of the atom t followed by
two arithmetic expressions.
You guessed it.
Let ' s try an example.
( atom? nex)
where
nexp is ( + 1 3)
( eq? ( car nex) (quote +) )
where
nex is ( + 1 3)
And now recur.
What is ( cdr nex)
where
nex is ( + 1 3)
104
Yes.
(defne vlue
(lambda ( nex)
(cond
( ( atom? nex) nex)
( ( eq? ( car nex) (quote +) )
( + ( value ( cdr nex) )
( value ( cdr ( cdr nex) ) ) ) )
( ( eq? ( car nex) (quote 7) )
( 7 ( value ( cdr nex))
( value ( cdr ( cdr nex) ) ) ) )
(else
(t ( value ( cdr nex ) )
( value ( cdr ( cdr nexp) ) ) ) ) ) ) )
It' s wrong.
(+ 1 3) .
No.
Yes.
Yes.
( 1 3) .
Chapter 6
(1 3) is not our representation of an
arithmetic expression.
How can we get the frst subexpression of a
representation of an arithmetic expression?
Is ( cdr ( cdr nex)) an arithmetic expression
where
nex is ¸ + 1 3)
Again, we were thinking of the list ( + 1 3)
instead of the representation of an arithmetic
expression.
What do we mean if we say the car of the
cdr of nex
Let 's write a function 1st-sub-ex for
arithmetic expressions.
Can we get by without ( cond . . . ) if we
don' t need to ask questions?
No, we violated The Seventh Commandment .
(1 3) is not a subpart that is a representation
of an arithmetic expression! We obviously
recurred on a list . But remember, not all lists
are representations of arithmetic expressions.
We have to recur on subexpressions.
By taking the car of the cdr.
No, the cdr of the cdr is (3) , and (3) is not
an arithmetic expression.
Taking the car of the cdr of the cdr gets us
back on the right track.
The frst subexpression of the representation
of an arithmetic expression.
(defne 1st-sub-ex
(lambda ( aex)
(cond
(else ( car ( cdr aex) ) ) ) ) )
Because the frst question i s also the last
question.
Yes, remember one-liners from chapter 4.
(defne 1st-sub-ex
(lambda ( aex)
( car ( cdr aexp) ) ) )
105
Write 2nd-sub-ex for arithmetic expressions.
Finally, let ' s replace ( car nexp) by
( opertor nex)
Now write value again.
Can we use this value function for the frst
representation of arithmetic expressions in
this chapter?
Do it !
106
(dÜfne 2nd-sub-ex
(lambda ( aex)
( car ( cdr ( cdr aex) ) ) ) )
(defne opertor
(lambda ( aex)
( car aex) ) )
(defne value
(lambda ( nex)
(cond
( ( atom? nex) nex)
( ( eq? ( opertor nex) (quote +) )
( + ( value ( 1 st-sub-ex nex))
( value ( 2nd-sub-ex nex) ) ) )
( ( eq? ( opertor nex) (quote 7 ) )
( 7 ( value ( 1 st-sub-ex nex))
( value ( 2nd-sub- ex nex) ) ) )
(else
( ¡ ( value ( 1st-sub-ex nex) )
( value ( 2nd-sub-ex nex) ) ) ) ) ) )
Yes, by changing 1st-sub-ex ad opertor.
(defne 1st-sub- ex
(lambda ( aex)
( car aexp) ) )
¸defne opertor
(lambda ( aex)
( car ( cdr aex) ) ) )
Chapter 6
Wan' t this eay? Yes, because we used help functions to hide
the representation.
The Eighth Commandment
Use help functions to abstract from representations.
Have we seen representations before?
For what entities have we used
representations?
Numbers are representations?
What else could we have used?
Do you remember how many primitives we
need for numbers?
Let' s try another representation for numbers.
How shall we represent zero now?
How is one represented?
How is two represented?
Yes, we j ust did not tell you that they were
representations.
Tuth-values! Numbers!
Yes. For example 4 stands for the concept
four. We chose that symbol because we are
accustomed to arabic representations.
( () () () ( ) ) would have served j ust as well .
What about ( ( ( ( ( ) ) ) ) ) ? How about ¸ | V)?
Four: number?, zer?, add1 , and sub1 .
() is our choice.
( ( ) ) .
( ( ) ( ) ) .
107
Got it? What ' s three?
Write a function to test for zero.
Can you write a function that is like add1
Is this correct?
What is ( zub1 n) where n is ( )
Rewrite + using this representation.
Ha the defnition of + changed?
108
Three is ( ( ) ( ) ( ) ) .
(defne ser?
(lambda ( n)
( null? n) ) )
(defne edd1
(lambda ( n)
( cons (quote ( ) ) n) ) )
(defne zub1
(lambda ( n)
( cdr n) ) )
Let ' s see.
No answer, but that ' s fne.
- Recall The Law of Cdr.
(defne +
(lambda ( n M)
(cond
( ( ser? M) n)
(else ( edd1 ( + n ( zub1 m) ) ) ) ) ) )
Yes and no. It chaned, but only slightly.
Chapter 6
Recal lat ?
Do you remember what the value of ( lat ? ls)
is where l is (1 2 3)
What is ¸ I2 3) with our new numbers?
What is ( lat? ls) where
ls is
' ' '
) ) ( ( )
'
) ) ( ( )
'
) ( ) ) )
Easy:
(defne lat ?
(lambda (l)
(cond
( ( null ? l) #t )
( ( atom? ( car l ) ) ( lat? ( cdr l ) ) )
(else #f ) ) ) )
#t g of course.
( ( ( ) ) ( ( ) ( ) ) ( ( ) ( ) ( ) ) ) .
It is very false.
109

l � � f� ���
Is this a set?
(appl e peaches appl e pl um)
Tue or false: ( set ? lat)
where
lat is (appl es peaches pears pl ums)
How about ( set ? lat)
where
lat is ( )
Ty t o write set ?
Simplify Iet ?
Does this work for the example
(appl e 3 pear 4 9 appl e 3 4}
Were you surprised to see the function
member? appear in the defnition of set ?
Fiends and Relations
No, since appl e appears more than once.
#t , because no atom appears more than
once.
#t , because no atom appears more than
once.
( defne set ?
(lambda ( lat)
(cond
( ( null ? lat) #t )
(else
(cond
( ( member? ( car lat) ( cdr lat) )
#f )
(else ( set ? ( cdr lat ) ) ) ) ) ) ) )
(defne set ?
(lambda ( lat)
(cond
( ( null ? lat) #t )
( ( member? ( car lat) ( cdr lat) ) #f )
(else ( set? ( cdr lat ) ) ) ) ) )
Yes, since member? i s now written using
You should not be, because we have written
member? already, and now we can use it
whenever we want .
1 1 1
What is ( makeset lat)
where
lat is (appl e peach pear peach
pl um appl e l emon peach)
Ty to write makeset using member?
Are you surprised to see how short this is?
Using the previous defnition, what is the
result of ( makeset lat)
where
lat is (appl e peach pear peach
pl um appl e l emon peach)
Ty to write makeset using multirmber
What is the result of ( makeset lat) using this
second defnition
where
lat is (appl e peach pear peach
pl um appl e l emon peach)
112
(appl e peach pear pl um l emon) .
(defne makeset
(lambda ( lat)
(cond
( ( null? lat) (quote ( ) ) )
( ( member? ( car lat) ( cdr lat) )
( makeset ( cdr lat) ) )
(else ( cons ( car lat)
( makeset ( cdr lat) ) ) ) ) ) )
We hope so. But don' t be afraid: it ' s right.
(pear pl um appl e l emon peach) .
(defne makeset
(lambda ( lat)
(cond
( ( null ? lat) (quote ( ) ) )
(else ( cons ( car lat)
( makeset
( multirmber ( car lat)
( cdr lat ) ) ) ) ) ) ) )
( appl e peach pear pl um l emon) .
Chapter 7
Describe in your own words how the second
defnition of makeset works.
Does the second makeset work for the
example
(appl e 3 pear 4 9 appl e 3 4)
What is ( subset ? set1 set2)
where
set1 is (5 chi cken wi ngs)
and
set2 is (5 hamburgers
2 pi eces fried chi cken and
l i ght duckl i ng wi ngs)
What is ( subset? set1 set2)
where
set1 is (4 pounds of horseradi sh)
and
set2 is (four pounds chi cken and
Write subset?
Fends and Relations
Here are our words:
"The function makeset remembers to cons
the frst atom in the lat onto the result of
the natural recursion, ater removing all
occurrences of the frst atom from the rest
of the lat . "
Yes, since multirember is now written using
#t , because each atom in set1 is also in set2.
#f.
(defne subset ?
(lambda ( set1 set2)
(cond
( ( null ? set1 ) #t )
(else (cond
( ( member? ( car set1 ) set2)
( subset? ( cdr set1 ) set2) )
(else #f ) ) ) ) ) )
1 13
Can you write a shorter version of subset ?
Ty to write subset ? with (and . . . )
What is ( eqset ? set1 set2)
where
set1 is (6 l arge chi ckens with wi ngs)
and
set2 is (6 chi ckens wi th l arge wi ngs)
Write eqset ?
Can you write eqset ? with only one
cond-line?
1 14
(defne subset?
(lambda ( set1 set2)
(cond
( ( null ? set1 ) #t )
( ( member? ( car set1 ) set2}
( subset ? ( cdr set1 ) set2) )
(else #f ) ) ) )
(defne subset ?
(lambda ( set1 set2)
(cond
( ( null ? set1 ) #t )
(else
(and ( member? ( car set1 ) set2)
( subset ? ( cdr set1 ) set2) ) ) ) ) )
#t .
(defne eqset ?
(lambda ( set1 set2)
(cond
( ( subset ? set1 set2)
( subset? set2 set1 ) )
(else #f ) ) ) )
(defne eqset ?
(lambda ( set1 set2)
(cond
(else (and ( subset? set1 set2)
( subset? set2 set1 ) ) ) ) ) )
Chapter 7
Write the one-liner.
What is ( intersect? set1 set2)
where
set1 is (stewed tomatoes and macaroni )
and
set2 is ( macaroni and cheese)
Defne the function intersect?
Write the shorter version.
Ty writing intersect? with ¸LÏ - - . )
Fiends and Relations
(defne eqset ?
(lambda ( set1 set2)
#t ,
(and ( subset ? set1 set2)
( subset ? set2 set1 ) ) ) )
because at least one atom in set1 is in
set2.
(defne intersect?
(lambda ( set1 set2)
(LLnd
( ( null ? set1 ) #f )
(else
(LLnd
( ( member? ( car set1 ) set2) #t )
(else ( intersect ?
( cdr set1 ) set2) ) ) ) ) ) )
(defne intersect?
(lambda ( set1 set2)
(LLnd
( ( null ? set1 ) #f )
( ( member? ( car set1 ) set2) #t )
(else ( intersect? ( cdr set1 ) set2) ) ) ) )
(defne intersect ?
(lambda ( set1 set2)
(LLnd
( ( null ? set1 ) #f )
(else ¦LÏ ( member? ( car set1 ) set2)
( intersect?
( cdr set1 ) set2) ) ) ) ) )
Compare subset? and intersect ?.
115
What is ( intersect set1 set2)
where
set1 is (stewed tomatoes and macaroni )
and
set2 is ( macaroni and cheese)
Now you can write the short version of
intersect
What is ( union set1 set2)
where
set1 is (stewed tomatoes and
macaroni casserol e)
and
set2 is ( macaroni and cheese)
Write union
1 16
(and macaroni ) .
(defne intersect
(lambda ( set1 set2)
(cond
( ( null? set1 ) (quote ( ) ) )
( ( member? ( cr set1 ) set2)
( cons ( car set1 )
( intersect ( cdr set1 ) set2) ) )
(else ( intersect ( cdr set1 ) set2)))))
(stewed tomatoes casserole macaroni
and cheese)
(defne union
(labda ( set1 set2)
(cond
( ( null ? set1 ) set2)
( ( member? ( car set1 ) set2)
( union ( cdr set1 ) set2) )
(else ( cons ( car set1 )
( union ( cdr set1 ) set2))))))
Chapter 7
What is this function?
(defne J
(lambda ( set1 set2)
(cond
( ( null ? set1 ) (quote ( ) ) )
( ( member? ( car set1 ) set2)
( J ( cdr set1 ) set2) )
(else ( cons ( car set1 )
( J ( cdr set1 ) set2) ) ) ) ) )
What i s ( intersect all l-set)
where
l-set is ( ¸a b c) ( c a d e) ( e f g h a b) )
What i s ( intersect all l-set)
where
l-set is ¸( 6 pears and)
(3 peaches and 6 peppers)
(8 pears and 6 pl ums)
(and 6 prunes wi th some appl es) )
Now, using whatever help functions you
need, write intersectall asuming that the list
of sets is non-empty.
Is this a pair?1
¸pear pear)
1 A pair in Scheme (or Lisp) is a diferent but related object.
Fiends and Relations
In our words:
"It is a function that returns all the atoms
in set1 that are not in set2. "
That is, J is the ¸set) diference function.
(a) .
(6 and) .
(defne intersectall
(lambda ( l-set)
(cond
( ( null ? ( cdr l-set ) ) ( car l-set ) )
(else ( intersect ( car l-set)
( intersectall ( cdr l-set ) ) ) ) ) ) )
Yes, because i t i s a list with only two atoms.
117
Is this a pair?
(3 7)
Is this a pair?
((2) ( pai r) )
( a-pair? l)
where
l is (ful l (house))
Defne a-pair?
How can you refer to the frst S-expression of
a pair?
How can you refer to the second S-expression
of a pair?
How can you build a pair with two atoms?
How can you build a pair with two
S-expressions?
Did you notice the diferences between the
1 18
Yes.
Yes, because it is a list with only two
S-expressions.
#t ,
because it is a list with only two
S-expressions.
(defne a-pair?
(lambda (x)
(cond
( ( atÛm? x) #f )
( ( Hull ? x) #f )
( ( Hull ? ( Ldr x) ) #f )
( ( Hull ? ( Ldr ( Ldr x) ) ) #t )
(else #f ) ) ) )
By taking the car of the pair.
By taking the car of the cdr of the pair.
You LÛHb the frst one onto the LÛHb of the
second one onto ( ) o That is,
¸ LÛHb x1 ¸ LÛHb x2 (quote () ) ) ) .
You LÛHb the frst one onto the LÛHb of the
second one onto ( ) . That is,
¸ LÛHb x1 ¸ LÛHb x2 (quote ( ) ) ) ) .
No, there aren' t any.
Chapter 7
(defne frst
(lambda ¸py
(cond
(else ( car p) ) ) ) )
(defne second
(lambda ¸py
(cond
(else ( car ( cdr py y y y y y
(defne build
(lambda ( sl s2)
(cond
¸else ( cons sl
( cons s2 (quote ( ) ) ) ) ) ) ) )
What possible uses do these three functions
have?
Ca you write thir Ü a one-liner?
Is l a rei where
l is (appl es peaches pumpki n pi e)
Is l a rei where
l is ( (appl es peaches)
( pumpki n pi e)
(appl es peaches) )
Is l a rei where
l is ( ( appl es peaches) ( pumpki n pi e) )
Is l a rei where
l is ¸ ¸43) ¸42) (7 6) (6 2) (3 4) )
Fiends and Relations
They are used to make representations of
pairs and to get parts of representations of
pairs. See chapter 6.
They will be used to improve readability,
as you will soon see.
Redefne frst, second, and build as
one-liners.
¸defne thir
(lambda ¸ l )
( car ( cdr ( cdr l ) ) ) ) )
No, since l i s not a list of pairs. We use rei to
stand for relation.
No, since l is not a set of pairs.
Yes.
Yes.
119
Is rl a fun
where
rl is ((4 3) (4 2) (7 6) (6 2) (3 4) )
What i s (fn? rl)
where
rl is ( ( 8 3) (4 2) (7 6) (6 2) (3 4) )
What i s (fn? rl)
where
rl is ( ( d 4) (b 0) (b 9) (e 5) (g 4) )
Write fn? with set ? and frsts
Is fn? a simple one-liner?
How do we represent a fnite function?
What is ( rvrel rl)
where
rl is ( (8 a) ( pumpki n pi e) (got si ck) )
You can now write rvrl
120
No. We use fun to stand for function.
#t , because (frsts rl) is a set
-See chapter 3.
#f, because b is repeated.
(defne fn?
(lambda ( rl)
(set? (frsts rl) ) ) )
It sure is.
For us, a fnite function is a list of pairs in
which no frst element of any pair is the same
Ü any other frst element .
¸ ¸a 8) (pi e pumpki n) (si ck got ) ) .
(defne rvrl
(lambda ( rl)
(cond
( ( null? rl) (quote ( ) ) )
(else ( cons ( build
(second ( car rl) )
(frst ( car rl) ) )
( rvrl ( cdr rl) ) ) ) ) ) )
Chapter 7
Would the following also be correct:
(defne rvrl
(lambda ( rl)
(cond
( ( null ? rl) (quote ( ) ) )
(else ( cons ( cons
( car ( cdr ( car rl) ) )
( cons ( car ( car rl) )
(quote ( ) ) ) )
( rvrl ( cdr rl) ) ) ) ) ) )
Suppose we had the function rvpair that
reversed the two components of a pair like
this:
(defne rvpair
(lambda (pair)
( build (second pair) (frst pair) ) ) )
How would you rewrite rvrl to use this help
function?
Ca you guess why fn is not a fullfun
where
fn is ( (8 3) (4 2) (7 6) (6 2) (3 4) )
Why is #t the value of (fllfn? fn)
where
fn is ( (8 3) (4 8) (7 6) (6 2) (3 4) )
What is (fllfn ' fn)
where
fn is ( (grape raisi n)
( pl um prune)
(stewed prune) )
Fiends and Relations
Yes, but now do you see how representation
No problem, and it is even easier to read:
¸defne revrl
(lambda ( rl)
¦cond
( ( null ? rl) (quote ( ) ) )
¸else ( cons ( rvpair ( car rl) )
( rvrl ( cdr rl) ) ) ) ) ) )
fn i s not a fullfun, since the 2 appears more
than once as a second item of a par.
Because (3 8 6 2 4) is a set .
#f.
121
What is (fullfn? fn)
where
fn is ( (grape rai si n)
( pl um prune)
(stewed grape))
Defne fllfn?
Can you defne seconds
What is another name for fllfn?
Can you think of a second way to write
one-to- one ?
Is ((chocol ate chi p) (doughy cooki e)) a
one-to-one function?
#t , because ( rai si n prune grape) is a set.
(defne fllfn?
(lambda (fn)
(set? (seconds fn) ) ) )
It i s j ust like frsts .
one-to-one ?.
(defne one-to- one ?
(lambda (fn)
(fn? ( rvrl fn) ) ) )
Yes, and you deserve one now!
Go and get one!
122 Chapter 7
Lr UCttCr ÿet , uake ÿour own.
(lÜbda ( )
( bake
(quote (350 degrees) )
(quote ( 12 mi nutes))
( mix
(quote (wal nuts 1 cup))
(quote (chocol ate-chi ps 16 ounces))
( mix
( mix
(quote (flour 2 cups))
(quote (oatmeal 2 cups) )
(quote (sal t .5 teaspoon) )
(quote (baki ng- poder 1 teaspoon) )
(quote (baki ng-soda 1 teaspoon) ) )
( mix
(quote (eggs 2 l arge) )
(quote (vani l la 1 teaspoon))
( cram
(quote (butter 1 cup))
(quote (sugar 2 cups) ) ) ) ) ) ) ) )
123
Remember what we did in rmber and
inserlL at the end of chapter 5?
Ca you write a function rmber-t that
would use either eq? or equal ?
How ca you make rmber remove the frst a
from (b c a)
How can you make rmber remove the frst c
from (b c a)
How can you make rmber-t use equal ?
What is ( rmber-t test ? a l)
where
test? is =1
a is 5
and
l is (6 2 5 3)
' L: (rember-f (fuction •) 5 ' (6 2 5 3) ) ,
but there is more.
What is ( rmber-t test? a l)
where
test? is eq?
a is jel l y
and
l is Uel l y beans are good)
And what i s ( rmber-t test? a l )
where
test ? is equal ?
a is ( pop corn)
and
l is ( l emonade (pop corn) and (cake) )
Lambda the Utimate
We replaced eq? with equal?
No, because we have not yet told you how.
By passing a and (b c a) as arguments to
rmber.
By passing c and (b c a) as arguments to
rmber.
By passing equal? a an argument to
rmber-t.
(6 2 3) .
( beans are good) .
( l emonade and (cake) ) .
125
Try to write rmber-f
How does ( rember-f test ? a l) act
where test? is eq?
And what about ( rmber-f test? a l)
where test? is equal?
Now we have four functions that do almost
the same thing.
126
(defne rmber-f
(lambda ( test ? a l)
(cond
( ( null ? l) (quote ( ) ) )
(else (cond
( ( test ? ( car l) a)
1
( cdr l ))
(else ( cons ( car l)
( rmber-f test ? a
( cdr l ) ) ) ) ) ) ) ) )
This i s good!
1
L: (fucall test? (car 1) a) . Use funcall when
invoking a function argument or a function that ha not
been defued.
(defne rmber-f
(lambda ( test ? a l)
(cond
( ( null ? l) (quote ( ) ) )
( ( test? ( car l) a) ( cdr l ))
(else ( cons ( car l)
( rmber-f test ? a
( cdr l ) ) ) ) ) ) )
( rmber-f test? a l )
where test? i s eq?, acts like rmber.
This is j ust rmber with eq? replaced by
equal ?.
Yes:
rmber with =
rmber with equal ?
rmber with eq?
and
rmber-f.
Chapter 8
And rmber-f ca behave like all the others.
What kind of values can functions return?
Ca you say what (lambda ¸ Ú l) . . . ) is?
Now what is
(lambda ¸ Ú)
(lambda ¸J)
( eq? J Ú) ) )
Is this called "Curry-ing?"
It is not called "Schonfnkel-ing. "
Using (defne . . . ) give the preceding
function a name.
What is ( eq?-c k)
Lambda the Ultimate
Let ' s generate all versions with rmber-f.
Lists and atoms.
Yes,
but you probably did not know that yet .
(lambda ¸ Ú l) . . . ) is a function of two
arguments, Ú and l .
It i s a function that, when passed an
argument Ú¿ returns the function
(lambda ¸J)
( eq? J Ú) )
where Ú i s j ust that argument .
Thank you, Moses Schonfnkel
( 1889-1942) .
Thank you, Hakell B. Curry
( 1900-1982) .
(defne eq?-c1
(lambda ¸ Ú)
(lambda ¸ J)
( eq? J Ú) ) ) )
This i s our choice.
1 L: (defu eq?-c (a)
(fuction
(labda (x)
(eq x a) ) ) )
Its value is a function that takes J a an
argument and tests whether it is eq? to sal ad.
127
So let ' s give it a name using (defne . . . )
I (defne
1
where k i s sal ad
1
Use setq to defne a function that can be fucalled.
What is ( eq? -salad y)
1
1
defued.
And what is ( eq?-salad y)
where y is tuna
Do we need to give a name to eq?-salad
Now rewrite rmber-t as a function of one
argument test? that returns an argument like
rmber with eq? replaced by test ?
128
Okay.
#t .
#
f.
No, we may just as well ak
( ( eq?-c x) y)
1
where
and
y is tuna.
1
L: (fucall (eq?-c x) y) , since (eq?-c x) i s a function
that ha not been defued.
(defne rmber-t
(lambda ( test ?)
(lambda ( a l)
(Cond
( ( null ? l) (quote ( ) ) )
( ( test ? ( car l) a) ( cdr l ) )
(else ( cons ( car l) . . . ) ) ) ) ) )
i s a good start.
Chapter 8
Describe in your own words the result of
( rmber-f test ?)
where
test? is eq?
Give a name to the function returned by
( rmber-f test ?)
where
test? is eq?
What is ( rmber-eq? a l)
where a is tuna
and
l is (tuna sal ad is good)
Did we need to give the name rmber-eq? to
the function ( rmber-f test ?)
where
test? is eq?
Now, complete the line
( cons ( car l) . . . )
in rmber-f so that rmber-f works.
What is ( ( rmber-f eq?) a l)
where Û is tuna
and
Lambda the Ultimate
It is a function that takes two arguments, a
and l. It compares the elements of the list
with a, and the frst one that is eq? to a is
removed.
(defne rmber-eq? ( rmber-f test ?) )
where
test ? is eq?.
No, we could have written
( ( rmber-f test ?) a l)
where
test? is eq?
a is tuna
and
l is (tuna sal ad i s good) .
(defne rmber-f
(lambda ( test ?)
(lambda ( a l)
(cond
( ( null ? l) (quote ( ) ) )
( ( test ? ( car l) a) ( cdr l ) )
(else ( cons ( car l)
( ( rmber-f test ?) a
( cdr l ) ) ) ) ) ) ) )
129
What is ( ( rmber-f eq?) a l)
where a is eq?
and
l is (equal ? eq? eqan? eql i st? eqpai r?) 1
¯ Did you notice the diference between eq? and eq?
Remember that the former is the atom and the latter is the
function.
And now transform insertL to insertL-f the
same way we have transformed rmber into
rmber-f
And, j ust for the exercise, do it to insertR
Are insertR and insertL similar?
Can you write a function insert-g that would
insert either at the left or at the right?
130
(equal ? eqan? eql i st? eqpai r?) .
(defne insertL-f
(lambda ( test ?)
(lambda ( new old l)
(cond
( ( null ? l) (quote ( ) ) )
( ( test ? ( car l) old)
( cons new ( cons old ( cdr l ) ) ) )
(else ( cons ( car l )
( ( insertL-f test ?) new old
( cdr l ) ) ) ) ) ) ) )
(defne insertR-f
(lambda ( test ?)
(lambda ( new old l)
(cond
( ( null? l) (quote ( ) ) )
( ( test ? ( car l) old)
( cons old ( cons new ( cdr l ) ) ) )
(else ( cons ( car |)
( ( insertR-f test ?) new old
( cdr l ) ) ) ) ) ) ) )
Only the middle piece i s a bit diferent .
If you can, get yourself some cofee cake and
relax! Otherwise, don' t give up. You' ll see it
in a minute.
Chapter 8
Which pieces difer?
Put the diference in words!
So how can we get rid of the diference?
Defne a function seqL that
1. takes three arguments, and
2. conses the frst argument
onto the result of consing
the second argument onto
the third argument.
What is:
(defne seqR
(lambda ( new old l)
( cons old ( cons new l ) ) ) )
Do you know why we wrote these functions?
Lambda the Ultimate
The second lines difer from each other. In
insertL it is:
( ( eq? ( car l) old)
( cons new ( cons old ( cdr l ) ) ) ) ,
but in insertR it is:
( ( eq? ( car l) old)
( cons old ( cons new ( cdr l ) ) ) ) .
We say:
"The two functions cons old and new in a
diferent order onto the cdr of the list l . "
You probably guessed it: by passing i n a
function that expresses the appropriate
consing.
(defne seqL
(lambda ( new old l)
( cons new ( cons old l ) ) ) )
A function that
1 . takes three arguments, and
2. conses the second argument
onto the result of consing
the frst argument onto
the third argument.
Because they express what the two difering
lines in insertL and insertR express.
131
Ty to write the function insert-g of one
argument seq
which returns insertL
where seq is seqL
and
which returns insertR
where seq is seqR
Now defne insertL with insert-g
And insertR.
Is there something unusual about these two
defnitions?
Is it necessary to give names to seqL and
seqR
Defne insertL again with insert-g
Do not pass in seqL this time.
132
(defne insert-g
(lambda (seq)
{lambda ( new old l)
(Lond
( ( null ? l) (quote ( ) ) )
( ( eq? ( car l) old)
(seq new old ( cdr l ) ) )
(else ( cons ( car l)
( ( insert-g seq) new old
( cdr l ) ) ) ) ) ) ) )
(defne insertL ( insert-g seqL) )
(defne insertR ( insert-g seqR))
Yes. Earlier we would probably have written
(defne insertL ( insert-g seq))
where
seq is seqL
and
(defne insertR ( insert-g seq))
where
seq is seqR.
But, using "where" is unnecessary when you
pass functions as arguments.
Not really. We could have passed their
(defne insertL
( insert-g
(lambda ( new old l)
( cons new ( cons old l ) ) ) ) )
Chapter 8
Is this better?
Do you remember the defnition of subst
Does this look familiar?
Defne a function like seqL or seqR for subst
And now defne subst using insert-g
And what do you think yyy is
(defne yyy
(lambda ( a l)
( ( insert-g seqrm) #f a l ) ) )
where
¸defne seqrm
(lambda ( new old l)
l ))
Lambda the Ultimate
Yes, because you do not need to remember as
many names. You can
( rmber Jne-name "your-mind" )
where Jne-name is seqL.
Here is one.
(defne subst
(lambda ( new old l)
(cond
( ( null ? l) (quote ( ) ) )
( ( eq? ( car l) old)
( cons new ( cdr l ) ) )
(else ( cons ( car l)
( subst new old ( cdr l ) ) ) ) ) ) )
Yes, i t looks like insertL or insertR. Just the
answer of the second cond-line is diferent .
(defne seqS
(lambda ( new old l)
( cons new l ) ) )
(defne subst ( insert-g seqS))
Surprise! It is our old friend rmber
Hint : Step through the evaluation of
( yyy a l)
where
a is sausage
and
l is ( pizza with sausage and bacon) .
What role does #f play?
133
What you have j ust seen is the power of abstraction.
The Ninth Commandment
Abstract common patterns with a new function.
Have we seen similar functions before?
Do you remember value from chapter 6?
Do you see the similarities?
Can you write the function atom-to-fnction
which:
1 . Takes one argument x and
2. returns the function +
134
if ( eq? x (quote +) )
returns the function 7
if ( eq? x (quote 7) ) and
returns the function t
otherwise?
Yes, we have even seen functions with similar
lines.
(defne value
(lambda ( nex)
(cond
( ( atom? nex) nex)
( ( eq? ( opertor nex)
(quote +) )
( + ( value ( 1st-sub-ex nex))
( value ( 2nd-sub-ex nex) ) ) )
( ( eq? ( opertor nex)
(quote 7 ) )
( 7 ( value ( 1st-sub-ex nex))
( value ( 2nd-sub-exp nexp) ) ) )
(else
( ¡ ( value ( 1st-sub-ex nex))
( value ( 2nd-sub-ex nex) ) ) ) ) ) )
The lat three answers are the same except
for the +, 7 ¿ and t.
(defne atom-to-fnction
(lambda (x)
¸cond
( ( eq? x (quote +) ) + )
( ( eq? x (quote 7 ) ) 7 )
¸else ¡) ) ) )
Chapter 8
What is ( atom-to-fnction ( opertor nexp) )
where
nex is ¸ + 5 3)
Can you use atom-to-fnction to rewrite
value with only two cond-lines?
Is this quite a bit shorter than the frst
version?
Time for an apple?
Here is multirmber again.
(defne multirmber
(lambda ( a lat)
(Lond
( ( null ? lat) (quote ( ) ) )
( ( eq? ( car lat) a)
( multirmber a ( cdr lat) ) )
(else ( cons ( car lat)
( multirmber a
( cdr lat ) ) ) ) ) ) )
Write multirmber-f
What is ( ( multirember-f test?) a lat)
where
test? is eq?
a is tuna
and
Lambda the Ultimate
The function +, not the atom +.
Of course.
(defne value
(lambda ( nex)
(cond
( ( atom? nexp) nex)
(else
( ( atom- to-fnction
( opertor nex) )
( value ( 1 st-sub- ex nex) )
( value ( 2nd-sub- exp nex) ) ) ) ) ) )
Yes, but that ' s oky. We haven' t changed its
meaning.
One a day keeps the doctor away.
No problem.
(defne multirmber-f
(lambda ( test ?)
(lambda ( a lat)
(Lond
( ( null ? lat) (quote ( ) ) )
( ( test ? a ( car lat ) )
( ( multirmber-f test ?) a
( cdr lat ) ) )
(else ( cons ( car lat)
( ( multirmber-f test ?) a
( cdr lat) ) ) ) ) ) ) )
135
Wasn' t that easy?
Defne multirmber- eq? using multirmber-f
Do we really need to tell multirmber-f
Does test ? change as multirmber-f goes
through lat
Can we combine a and test?
How would it do that?
Here is one way to write this function.
I
(defne <q?-tuno
.
( eq?-c k) )
where k i s tuna
Can you think of a diferent way of writing
this function?
Have you ever seen defnitions that contain
atoms?
136
Yes.
(defne multirmber-eq?
( multirmber-f test?) )
where test? i s eq?.
As multirmber-f visits all the elements in
lat , it always looks for tuna.
No, test? always stands for eq?, just a a
always stands for tuna.
Well , test? could be a function of j ust one
argument and could compare that argument
to tuna.
The new test? takes one argument and
compares it to tuna.
Yes, and here i s a diferent way:
(defne eq?-tuna
( eq? -c (quote tuna) ) )
Yes, 0, (quote 7 ) ¿ (quote +) , and many
more.
Chapter 8
Perhaps we should now write multirmberT
which is similar to multirmber-f
Instead of taking testP and returning a
function, multirmberT takes a function like
eq?-tuna and a lat and then does its work.
What is ( multirmberT test ? lat)
where
test? is eq?-tuna
and
Is this easy?
(defne multirmber&co
(lamda ( a lat col)
(cond
( ( null ? lat)
( col ,
quote ( ) ) (quote ( ) ) ) )
( ( eqP ( car lat) a)
( multirmber&co a
( cdr lat)
(lambda ( newlat seen)
( col newlat
( cons ( car lat) seen) ) ) ) )
¸else
( multirmber&co a
( cdr lat)
(lambda ( newlat seen)
( col ( cons ( car lat) newlat)
seen) ) ) ) ) ) )
Lambda the Utimate
This is not really difcult.
(defne multirmberT
(lambda ( test ? lat)
(cond
( ( null ? lat) (quote ( ) ) )
( ( test ? ( car lat) )
( multirmberT test ? ( cdr lat) ) )
(else ( cons ( car lat)
( multirmberT test?
( cdr lat) ) ) ) ) ) )
Now that looks really complicated!
137
Here is something simpler:
(defne a-fiend
(lbda ¦J y)
( null ? y)))
What is the vue of
( multirmber&co a lat col)
where
a is tuna
lat is (strawberries tuna and swrdfish)
and
col is a-fend
So let ' s try a friendlier example. What is the
vlue of ( multirmber&co Ü lat col)
where
a is tuna
lat is ¦ )
and
col is a-fend
And what is ( multirmber&co a lat col)
where
a is tuna
lat is (tuna)
and
col is a-fend
What are the other arguments that
multirmber&co uses for the natural
recursion?
What is the name of the third argument?
Do you know what col stands for?
138
Yes, it is simpler. It is a function that takes
two arguments and aks whether the second
one is the empty list. It ignores its frst
argnment.
This is not simple.
#t , because a-fend is immediately used in
the frst answer on two empty lists, and
a-fend makes sure that its second argument
is empty.
multirmber&co ak
( eqP ( car lat) (quote tuna) )
where
lat is (tuna) .
Then i t recurs on ( ) .
The frst one is clearly tuna. The third
argument is a new function.
col.
The name col is short for "collector. "
A collector is sometimes called a
"continuation. "
Chapter 8
Here is the new collector:
(defne new-fiend
(lambda ( newlat seen)
( col newlat
( cons ( car lat) seen) ) ) )
where
( car lat) is tuna
and
col is a-fiend
Can you write this defnition diferently?
Can we also replace col with a-fiend in such
defnitions because col is to a-fiend what
( car lat) is to tuna
And now?
Which collector is this?
How does a-fiend difer from new-fiend
And what does the old collector do with such
arguments?
What is the value of
( multirmber&co a lat a-fiend)
where a is tuna
and
lat is (and tuna)
Lambda the Utimate
Do you mean the new way where we put t una
into the defnition?
(defne new-fiend
(lambda ( newlat seen)
( col newlat
( cons (quote tuna) seen) ) ) )
where
col is a-fiend.
Yes, we can:
(defne new-fiend
(lambda ( newlat seen)
( a-fiend newlat
( cons (quote tuna) seen) ) ) )
multirember&co fnds out that ( null ? lat ) is
true, which means that it uses the collector
on two empty lists.
It is new-fiend.
new-fiend uses a-fiend on the empty list
and the value of
( cons (quote tuna) (quote ( ) ) ) .
I t answers #f, because its second argument
is (tuna) , which is not the empty list .
This time around multirmber&co recurs
with yet another friend.
(defne latest-fiend
(lambda ( newlat seen)
( a-fiend ( cons (quote and) newlat)
seen) ) )
139
And what is the value of this recursive use of
multirmber&co
What does ( multirmber&co a lat f) do?
Final question: What is the value of
( multirmber&co (quote tuna) ls col)
where
ls is (strawberries tuna and swordfi sh)
and
col is
(defne last-fiend
(lambda ¸J y)
( length x) ) )
Yes!
#f, since ( a-fiend ls1 l2)
where
ls1 is (and)
and
ls2 is (tuna)
is #f.
It looks at every atom of the lat to see
whether it is eq? to a. Those atoms that are
not are collected in one list ls1 ; the others
for which the answer is true are collected in a
second list ls2. Finally, it determines the
value of (! ls1 l2) .
3, because ls contains three things that are
not tuna, and therefore last-fiend is used on
(strawberries and swordfi sh) and (tuna) .
It' s a strange meal , but we have seen foreign
foods before.
The Tenth Commandment
Build functions to collect more than one vlue at a time.
140 Chapter 8
Here is an old friend.
(defne multiinsertL
(lambda ( new old lat)
(cond
( ( null ? lat) (quote ( ) ) )
( ( eq? ( car lat) old)
( cons new
( cons old
( multiinsertL new old
( cdr lat ) ) ) ) )
(else ( cons ( car lat)
( multiinsertL new old
( cdr lat) ) ) ) ) ) )
Do you also remember multiinsertR
Now try multiinsertLR
Hint: multiinsertLR inserts new to the left
of oldL and to the right of oldR in lat if
oldL are oldR are diferent .
The function multiinsertLR&co i s to
multiinsertLR what multirmber&co is to
multirmber
Yes, and what kind of argument is it?
Lambda the Ultimate
No problem.
(defne multiinsertR
(lambda ( new old lat)
(cond
( ( null ? lat) (quote ( ) ) )
( ( eq? ( car lat) old)
( cons old
( cons new
( multiinsertR new old
( cdr lat ) ) ) ) )
(else ( cons ( car lat)
( multiinsertR new old
( cdr lat ) ) ) ) ) ) )
This i s a way of combining the two functions.
(defne multiinsertLR
(lambda ( new oldL oldR lat)
(cond
( ( null ? lat) ¸quote ( ) ) )
( ( eq? ( car lat) oldL)
( cons new
( cons oldL
( multiinsertLR new oldL oldR
( cdr lat ) ) ) ) )
( ( eq? ( car lat) oldR)
( cons oldR
( cons new
( multiinsertLR new oldL oldR
( cdr lat ) ) ) ) )
(else
( cons ( car lat)
( multiinsertLR new oldL oldR
( cdr lat ) ) ) ) ) ) )
Does this mean that multiinsertLR&co takes
one more argument than multiinsertLR?
It is a collector function.
141
When multiinsertLR&co is done, it will use
col on the new lat , on the number of lef
insertions, and the number of rght
insertions. Can you write an outline of
multiinsertLR&co
Why is col used on ¸quote ( ) ) 0 and 0 when
( null ? lat) is true?
So what is the value of
( multiinsertLR&co
¸quote cranberri es)
¸quote fi sh)
¸quote chi ps)
¸quote ¸ ) )
col)
Is it true that multiinsertLR&co will use the
new collector on three arguments when
( car lat) is equal to neither oldL nor oldR
142
Sure, it is j ust like multiinsertLR.
(defne multiinsertLR&co
(lambda ( new oldL oldR lat col)
(cond
( ( null ? lat)
( col ¸quote ( ) ) 0 0) )
( ( eq? ( car lat) oldL)
( multiinsertLR&co new oldL oldR
( cdr lat)
(lambda ( newlat L R)
. . . ) ) )
( ( eq? ( car lat) oldR)
( multiinsertLR&co new oldL oldR
( cdr lat)
(lambda ( newlat L R)
. . . )))
¸else
( multiinsertLR&co new oldL oldR
( cdr lat)
(lambda ( newlat L R)
. . . ) ) ) ) ) )
The empty lat contains neither oldL nor
oldR. And this means that 0 occurrences of
oldL and 0 occurrences of oldR are found and
that multiinsertLR will return () when lat is
empty.
It is the value of ( col (quote ( ) ) 0 0) , which
we cannot determine because we don' t know
what col is.
Yes, the frst is the lat that multiinsertLR
would have produced for ( cdr lat ) , oldL, and
oldR. The second and third are the number
of insertions that occurred to the left and
right of oldL and oldR, respectively.
Chapter 8
Is it true that multiinsertLR&co then uses
the function col on ( cons ( car lat) newlat)
because it copies the list unless an oldL or an
oldR appears?
Why are col' s second and third arguments
j ust L and R
Here is what we have so far. And we have
even thrown in an extra collector:
(defne multiinsertLR&co
(lambda ( new oldL oldR lat col)
(Lond
( ( null ? lat)
( col (quote ( ) ) 0 0) )
( ( eq? ( car lat) oldL)
( multiinsertLR&co new oldL oldR
( cdr lat)
(lambda ( newlat L R)
( col ( cons new
( cons oldL newlat) )
( add1 L) R) ) ) )
( ( eq? ( car lat) oldR)
( multiinsertLR&co new oldL oldR
( cdr lat)
(lambda ( newlat L R)
. . . ) ) )
(elÜe
( multiinsertLR&co new oldL oldR
( cdr lat)
(lambda ( newlat L R)
( col ( cons ( car lat) newlat)
L R) ) ) ) ) ) )
Can you fll i n the dots?
So can you fll in the dots?
Lambda the Ultimate
Yes, it is true, so we know what the new
collector for the last cae is:
(lambda ( newlat L R)
( col ( cons ( car lat) newlat) L R) ) .
If ( car lat) i s neither oldL nor oldR, we do
not need to insert any new elements. So, L
and R are the correct results for both
( cdr lat) and all of lat .
The incomplete collector is similar to the
onto consing oldL onto newlat , it conses oldR
onto the result of consing new onto newlat .
Yes, the fnal collector is
(lambda ( newlat L R)
( col ( cons oldR ( cons new newlat) )
L ( add1 R) ) ) .
143
What is the value of
( multiinsertLR&co new oldL oldR lat col)
where
new is salty
oldL is fi sh
oldR is chi ps
and
lat is (chi ps and fi sh ÕÎ fi sh and chi ps)
Is this healthy?
Do you remember what *-functions are?
Now write the function evens- only* which
removes all odd numbers from a list of nested
lists. Here is even?
(defne even?
(lambda ( n)
¸ ¯ ¸ 7 ¸ ÷ n 2) 2) n) ) )
What i s the value of ( evens- only* l)
where
l is ( (9 1 2 8) 3 10 ( (9 9
) 7 6) 2)
144
It is the value of ( col newlat 2 2)
where
newlat is (chi ps sal ty and sal ty fi sh
ÕÎ sal ty fi sh and chi ps sal ty) .
Looks like lots of salt. Perhaps dessert is
sweeter.
Yes, all *-functions work on lists that are
either
- empty,
- an atom consed onto a list , or
- a list consed onto a list.
Now that we have practiced this way of
writing functions, evens- only* is j ust an
exercise:
(defne evens-only*
(lambda (l )
(cond
( ( null ? l) (quote ( ) ) )
( ( atom? ( car l ) )
(cond
( ( even? ( car l ))
( cons ( car l)
( evens-only* ( cdr l))))
¸else ( evens-only* ( cdr l ) ) ) ) )
(else ( cons ( evens-only* ( car l ))
( evens-only* ( cdr l ) ) ) ) ) ) )
( ( 2 8) 10 ( ( ) 6) 2) .
Chapter 8
What is the sum of the odd numbers in l
where
l is ( (9 1 2 8) 3 10 ( (9 9
) 7 6) 2)
What is the product of the even numbers in l
where
l is ( (9 1 2 8) 3 10 ( (9 9
) 7 6) 2)
Can you write the function evens-only*&co
It builds a nested list of even numbers by
removing the odd ones from its argument
and simultaneously multiplies the even
numbers and sums up the odd numbers that
occur in its argument .
Here is an outline. Can you explain what
( evens-only*&co ( car l) . . . ) accomplishes?
(defne evens-only*&co
(lambda ( l col)
(cond
( ( null ? l)
( col (quote ( ) ) 1 0) )
( ( atom? ( car l ))
fcond
( ( even? ( car l ) )
( evens- only*&co ( cdr l)
(lambda ( newl p s)
( col ( cons ( car l) newl)
( 7 ( car |) p) s) ) ) )
(else ( evens-only*&co ( cdr l)
(lambda ( newl p s)
( col newl
p ( + ( car l) s) ) ) ) ) ) )
(else ( evens-only*&co ( car l )
o . . ) ) ) ) )
What does the function evens-only*&co do
after visiting all the numbers in ( car l)
Lambda the Ultimate
9 + 1 + 3 + 9 + 9 + 7 ° 38.
2 ? 8 ? 10 ? 6 ? 2 ° 1920.
This is full of stars!
It visits every number in the car of l and
collects the list without odd numbers, the
product of the even numbers, and the sum of
the odd numbers.
It uses the collector, which we haven' t
defned yet .
145
And what does the collector do?
Does this mean the unknown collector looks
roughly like this:
(lambda ( al ap as )
( evens-only*&co ( cdr l)
. . . ) )
And when ( evens-only*&co ( cdr l) . . . ) is
done with its job, what happens then?
What does the collector for
( evens-only*&co ( cdr l) . . . )
do?
Does this all make sense now?
What is the value of
( evens-only*&co l the-last-fend)
where
l is ( (9 1 2 8) 3 10 ( (9 9) 7 6) 2) and
the-last-fend is defned as follows:
(defne the-last-fend
(lambda ( newl product sum)
( cons sum
( cons product
newl ) ) ) )
146
It uses evens- only*&co to visit the cdr of l
and to collect the list that is like ( cdr l ) ,
without the odd numbers of course, as well
a the product of the even numbers and the
sum of the odd numbers.
Yes.
The yet-to-be-determined collector is used,
j ust as before.
It conses together the results for the lists in
the car and the cdr and multiplies and adds
the respective products and sums. Then it
passes these values to the old collector:
(lambda ( al ap as)
( evens-only*&co ( cdr l)
(lambda ( dl dp ds )
( col ( cons al dl)
,
7 ap dp)
( + as ds ) ) ) ) ) .
Perfect .
(38 1920 (2 8) 10 ( ( ) 6) 2) .
Chapter 8
Whew! Is your brain twisted up now? Go eat a pretzel ; don' t forget the mustard.
Lambda the Utimate 147
Are you in the mood for cavi ar
What is ( looking a lat)
where a i s caviar
and
lat is (6 2 4 cavi ar 5 7 3)
( looking a lat)
where a is caviar
and
lat is (6 2 grits cavi ar 5 7 3)
Were you expecting something diferent?
Tue enough, but what is the frst number
in the lat?
And what is the sixth element of lat
And what is the seventh element?
So looking clearly can' t fnd cavi ar
Here is looking
(defne looking
(lambda ( a lat)
( keep-looking a (pick 1 lat) lat) ) )
Write keep-looking
( looking a lat)
where a is cavi ar
md
lat is (6 2 4 caviar 5 7 3)
. . . and Agan, and Agan, and Agan, . . .
Then we must go looking for it.
#t ,
cavi ar is obviously in lat .
#f.
Yes, cavi ar is still in lat .
6.
7.
3.
Tue enough,
because the third element is gri ts, which
does not even resemble cavi ar.
We did not expect you to know this.
#t ,
because ( keep-looking a 6 lat) has the same
answer as ( keep-looking a (pick 1 lat) lat) .
149
What is (pick 6 lat)
where
lat is (6 2 4 cavi ar 5 7 3)
So what do we do?
What is (pick 7 lat)
where
lat is (6 2 4 cavi ar 5 7 3)
So what is ( keep-looking a 3 lat)
where a is cavi ar
and
lat is (6 2 4 cavi ar 5 7 3)
Which is?
Write keep-looking
Can you guess what sor stands for?
We call this "unnatural" recursion.
150
7.
( keep-looking a 7 lat)
where a is cavi ar
and
lat is (6 2 4 cavi ar 5 7 3) .
3.
It is the same Ü
( keep-looking a 4 lat ) .
#t .
(defne keep-looking
(lambda ( a sor lat)
(cond
( ( number? sor)
( keep-looking a (pick sor lat) lat) )
(else ( eq? sor a) ) ) ) )
Symbol or number.
It does not recur on Ü part of lat.
It is truly unnatural .
Chapter 9
Does keep-looking appear to get closer to its
goal?
Does it always get closer to its goal?
That is correct. A list may be a tup.
What is ( looking a lat)
where a is cavi ar
and
lat is (7 1 2 caviar 5 6 3)
Yes, it is strange. What happens?
Fnctions like looking are called partial
functions. What do you think the functions
we have seen so far are called?
Can you defne a shorter function that does
not reach its goal for some of its arguments?
For how many of its arguments does eternity
reach its goal?
Is eterity partial?
What is ( shif J)
where
x is ((a b) c)
. . . and Agan, and Again, and Again, . . .
Yes, from all available evidence.
Sometimes the list may contain neither cavi ar
nor gri ts.
Yes, if we start looking in (7 2 4 7 5 6 3), we
will never stop looking.
This is strange!
We keep looking and looking and looking . . .
They are called total .
(defne eterity
(lambda ¸ J)
( eterity J) ) )
None, and this i s the most unnatural
recursion possible.
It is the most partial function.
(a (b c) ) .
151
What is ( shif x)
where
x is ( (a b) ¸L d) )
Defne shif
Describe what shif does.
Now look at this function:
(defne align
(lambda (por)
(cond
( ( atom? por) por)
( ( a-pair? (frst por) )
( align ( shif pora) ) )
(elÜe ( build (frst por)
( align (second por) ) ) ) ) ) )
What does i t have i n common with
keep-looking
Why are we not guaranteed that align makes
progress?
Which commandment does that violate?
152
¦
a ( b
,
L d) ) ) .
This i s trivial ; it' s not even recursive!
(defne shif
(lambda (pair)
( build (frst (frst pair) )
( build (second (frst pair))
( second pair) ) ) ) )
Here are our words:
"The function shif takes a pair whose frst
component is a pair and builds a pair by
shifting the second part of the frst
component into the second component. "
Both functions change their arguments for
their recursive uses but in neither cae is the
change guaranteed to get us closer to the
goal.
In the second cLÏd-line shif creates an
argument for align that is not a part of the
original argument .
The Seventh Commandment .
Chapter 9
Is the new argument at least smaller than
the original one?
Why not?
And?
Can you write a function that counts the
number of atoms in align' s arguments?
Is align a partial function?
Is there something else that changes about
the arguments to align and its recursive uses?
In what way is the frst component simpler?
Doesn' t this mean that length* is the wrong
function for determining the length of the
argument? Can you fnd a better function?
How much more attention should we pay to
the frst component?
. . . and Again, and Again, and Again, . . .
It does not look that way.
The function shif only rearranges the pair it
gets.
Both the result and the argument of shif
have the same number of atoms.
No problem:
(defne length*
(lambda (pora)
(cond
( ( atom? pora) 1)
(else
( + ( length* (frst por) )
( length* (second por) ) ) ) ) ) )
We don' t know yet. There may be arguments
for which it keeps aligning things.
Yes, there is. The frst component of a pair
becomes simpler, though the second
component becomes more complicated.
It is only a part of the original par's frst
component .
A better function should pay more attention
to the frst component.
At least twice as much.
153
Do you mean something like weight*
(defne weight*
(lambda (por)
(cond
( ( atom? por) I)
(else
( + ( 7 ( weight* (frst por) ) 2)
( weight* ( second por) ) ) ) ) ) )
What i s ( weight * x)
where
x is ¸ ¸a b) c)
And what is ( weight* x)
where
x is ¸a (b c) )
Does this mean that the arguments get
simpler?
Is align a partial function?
Here is shuje which is like align but uses
revpair from chapter 7, instead of shif:
(defne shuje
(lambda (por)
(cond
( ( atom? por) por)
( ( a-pair? (frst por) )
( shuje ( revpair por) ) )
( else ( build (frst por)
( shuje ( second por) ) ) ) ) ) )
Does this mean that shuje i s total?
154
That looks right .
7.
5.
Yes, the weight *' s of align' s arguments
become successively smaller.
No, it yields a value for every argument .
The functions shuje and rvpair swap the
components of pairs when the frst
component is a pair.
We don' t know.
Chapter 9
Let 's try it . What is the value of ( shufe x)
where
x is (a (b L) )
( shufe x)
where
x is (a b)
Okay, let ' s try something interesting. What
is the value of ( shufe x)
where
x is ¸ ¸a b) ¸L d) )
And how are we going t o do that?
Doesn' t this mean that we need to know the
value of ( shufe ( rvpair por) )
where
( rvair por) is ¸ ¸a b) ¸L d) )
And?
Is this function total?
(defÏe C
(lambda ( n)
(coÏd
( ( one ? n) 1)
(else
(cond
( ( even? n) ( C (+ n 2) ) )
(else ( C ( add1 ( 7 3 n) ) ) ) ) ) ) ) )
. . . and Agan, and Again, and Again, . . .
¸a ( b L) ) .
¸a b) .
To determine this value, we need to fnd out
what ( shufe ( rvair por) ) is
where
por is ((a b) ¸L d) ) .
We are going t o determine the value of
( shufe por)
where por i s ¸ ¸L d) ¸a b) ) .
Yes, we do.
The function shufe is not total because it
now swaps the components of the pair again,
which means that we start all over.
It doesn' t yield a value for 0, but otherwise
nobody knows. Thank you, Lothar Collatz
( 1910-1990) .
155
What is the value of ¸ A 1 0)
¸ A 1 1)
¸ A 2 2)
Here is the defnition of A
(defne A
(lambda ( n m)
(cond
( ( zero ? n) ( addl m) )
( ( zer? m) ¸ A ( subl n) 1) )
(else ¸ A ( subl n)
¸ A n ( subl m) ) ) ) ) ) )
What does A have i n common with shufe
and looking
Does A always give an answer?
Then what is ¸ A 4 3)
What does that mean?
156
2.
3.
7.
Thank you, Wilhelm Ackermann
( 1853-1946) .
A' s arguments, like shufe' s and looking' s,
do not necessarily decrease for the recursion.
That ' s easy: ¸ A 1 2) needs the value of
¸ A 0 ¸ A 1 1) ) . And that means we need the
value of ¸ A 0 3) .
Yes, i t i s total .
For all practical purposes, there is no answer.
The page that you are reading now will have
decayed long before we could possibly have
calculated the value of ¸ A 4 3) .
And this was scarcely odd, because
They'd eaten every one.
The Walrus and The Carpenter
-Lewis Carroll
Chapter 9
Wouldn' t it be great if we could write a
function that tells us whether some function
returns with a value for every argument ?
Okay, let' s write it .
Then let' s make it simpler. For a warm-up
exercise, let ' s focus on a function that checks
whether some function stops for just the
empty list , the simplest of all arguments.
Here is the beginning of this function:
(defne will-stop?
(lambda U)
. . . ) )
Can you fll i n the dots?
Does will-stop? return a value for all
arguments?
Is will-stop? total then?
Then let' s make up some examples. Here is
the frst one. What is the value of
( will-stop? f)
where
f is length
So?
. . . and Agan, and Agan, and Again, . . .
It sure would. Now that we have seen
functions that never return a value or return
a value so late that it is too late, we should
have some tool like this around.
It sounds complicated. A function can work
for many diferent arguments.
That would simplify it a lot .
What does it do?
That' s the easy part : we said that it either
returns #t or #f, depending on whether the
argument stops when applied to ( ) .
Yes, i t i s. It always returns #t or #f.
We know that ( length l) is 0
where l is ( ) .
Then the value of ( will-stop? length) should
be #t .
157
What is the value of ( will-stop? eterity)
Does this mean the value of
( will-stop ? eterity) is #f
Do we need more examples?
Okay, here is a function that could be an
interesting argument for will-stop?
(defne last-tr
(lambda ¸J)
(and ( will-stop? last-try)
( eterity J) ) ) )
What i s ( will-stop? last-try)
We need to test it on ( )
What is the value of
(and ( will-stop? last-try)
( eternity (quote ( ) ) ) )
There are only two possibilities. Let ' s say
( will-stop? last-try) is #f
So ( last-try (quote ( ) ) ) stopped, right?
But didn' t will-stop? predict j ust the
opposite?
158
( eterity (quote ( ) ) ) doesn' t return a value.
We just saw that.
Yes, it does.
Perhaps we should do one more example.
What does it do?
If we want the value of ( last-tr (quote ( ) ) ) ,
we must determine the value of
(and ( will-stop? last-tr)
( eterity (quote ( ) ) ) ) .
That depends on the value of
( will-stop? last-tr) .
Oky, then (and #f ( eterity (quote ( ) ) ) ) ,
i s #f, since (and #f . . . ) i s always #f.
Yes, it did.
Yes, it did. We said that the value of
( will-stop? last-tr) was #f, which really
means that last-try will not stop.
Chapter 9
So we must have been wrong about
¸ o:||-s|ep?|as|-|rçy
Fine. If ¸ o:||-s|eµ°|as|- |rçy is #t
what is the value of ¸ |as|-|rç (quote ( ) ) )
What i s the value of ¸ e|e¬:/ç (quote ¸ ) ) )
But that means we were wrong again!
What do you think this means?
Is this unique?
Is there any way around this problem?
What is (defne . . . )
. . . and Agan, and Agan, and Again, . . .
That ' s correct . It must return #t , because
o:||-s|ep ?always gives an answer. We said it
was total .
Now we j ust need to determine the value of
(and #t ¸ e|e¬:/ç (quote ( ) ) ) ) ,
which i s the same as the value of
¸ e/e¬:|ç (quote ¸ ) ) ) .
It doesn' t have a value. We know that it
doesn' t stop.
True, since this time we said that
¸ o:||-s/eµ? |as|- |rçy was #t .
Here is our meaning:
"We took a really close look at the two
possible cases. If we can defne o:||-s/ep ?,
then
¸ o:||-s|ep ?|as/- |rçy
must yield either #t or #f. But it
cannot-due to the very defnition of what
o:||-s|eµ?is supposed to do. This must
mean that o:||-s|ep ?cannot be defned. "
Yes, it is. It makes o:||-s|ep ?the frst
function that we can describe precisely but
cannot defne in our language.
No. Thank you,
Alan M. Turing ( 1912-1954)
and
Kurt Godel ( 1906-1978) .
This i s an interesting question. We j ust saw
that (defne . . . ) doesn' t work for o:||-s/ep ?.
159
So what are recursive defnitions?
Is this the function length
(defne length
(lambda ( l )
(cond
( ( null ? l) 0)
(else ( add1 ( length ( cdr l ) ) ) ) ) ) )
What i f we didn' t have (defne . . . )
anymore? Could we still defne length
What does this function do?
(lambda ( l )
(cLnd
( ( null ? l) 0)
(else ( add1 ( eterity ( cdr l ) ) ) ) ) )
What happens when we use i t on a
non-empty list?
What does it mean for this function that
looks like length
Suppose we could name this new function.
What would be a good name?
How would you write a function that
determines the length of lists that contain
one or fewer items?
160
Hold tight, take a deep breath, and plunge
It sure is.
Without (defne . . . ) nothing, and especially
not the body of length, could refer to length.
It determines the length of the emi,tY list
and nothing else.
No answer. If we give eterity an argument,
It j ust won' t give any answer for non-empty
lists.
lengtho
because the function can only determine
the length of the empty list .
Well , we could try the following.
(lambda ( l )
(cond
( ( null ? l) 0)
(else ( add1 ( lengtho ( cdr l ) ) ) ) ) )
Chapter 9
Almost, but (defne . . . ) doesn' t work for
lengtho
And what' s a good name for this function?
Is this the function that would determine the
lenghts of lists that contain two or fewer
items?
(lambda (l )
(cond
( ( null ? l) 0)
(else
((lambda ( l)
(cond
( ( null ? l) 0)
(else
((lambda ( l)
(cond
( ( null? l) 0)
(else
( eterity
( cdr l ) ) ) ) ) )
( cdr l ) ) ) ) ) )
( cdr l ) ) ) ) ) )
Now, what do you think recursion is?
. . . and Again, and Again, and Again, . . .
So, replace length0 by its defnition.
(lambda ( l )
(cond
( ( null ? l) 0)
(else
( (lambda ( l )
(cond
( ( null ? l) 0)
( eterity ( cdr l ) ) ) ) ) )
( cdr l ) ) ) ) ) )
That ' s easy: length9.
Yes, this is length�2 . We j ust replace eterity
with the next version of length.
What do you mean?
161
Well , we have seen how to determine the
length of a list with no items, with no more
than one item, with no more than two items,
and so on. How could we get the function
length back?
How long are the lists that we can make?
But we can' t write an infnite function.
And we still have all these repetitions and
patterns in these functions.
What do these patterns look like?
Let ' s do it !
Do you mean this?
( ( lambda ( length)
(lambda ( l )
(cond
( ( null ? l) 0)
(else ( add1 ( length ( cdr l ) ) ) ) ) ) )
eternity)
162
If we could write an infnite function in the
style of lengtho , length9, length9, . . . , then
we could write length0, which would
determine the length of all lists that we can
make.
Well , a list is either empty, or it contains one
element , or two elements, or three, or four,
. . . , or 1001 , . . .
No, we can' t.
Yes, we do.
All these programs contain a function that
looks like length. Perhaps we should abstract
out this function: see The Ninth
Commandment .
We need a function that looks j ust like length
but starts with (lambda ( length) . . . ) .
Yes, that ' s okay. I t creates length0 .
Chapter 9
Rewrite length9 in the same style.
Do we have to use length to name the
argument?
Close, but there are still repetitions.
Where should we start?
. . . and Agan, and Agan, and Again, . . .
( (lambda (/)
(lambda ( l )
(cond
( ( null ? l) 0)
(else ( add1 (/ ( cdr l ) ) ) ) ) ) )
( ( lambda ( g)
(lambda ( l )
(cond
( ( null ? l) 0)
(else ( add1 ( g ( cdr l ) ) ) ) ) ) )
eterity) )
No, we j ust used f and g. As long as we are
consistent , everything's oky.
( ( lambda ( length)
(lambda ( l )
(cond
( ( null ? l) 0)
(else ( add1 ( length ( cdr l ) ) ) ) ) ) )
( ( lambda ( length)
(lambda ( l )
(cond
( ( null ? l) 0)
(else ( add1 ( length ( cdr l ) ) ) ) ) ) )
( ( lambda ( length)
(lambda ( l )
(cond
( ( null ? l) 0)
(else ( add1 ( length ( cdr l ) ) ) ) ) ) )
eterity) ) )
Tue. Let 's get rid of them.
Name the function that takes length as an
argument and that returns a function that
looks like length.
163
What ' s a good name for this function?
Okay, do this to lengtho
Is this length9
( (lambda ( mk-length)
( mk-length
( mk-length eterity) ) )
(lambda ( length)
(lambda ( l )
(cond
( ( null ? l) 0)
(else ( add1 ( length ( cdr l ) ) ) ) ) ) ) )
Can you write length9 i n this style?
What is recursion like?
164
How about mk-length for "make length" ?
No problem.
( ( lambda ( mk-length)
( mk-length eterity) )
(lambda ( length)
(lambda ( l )
(cond
( ( null ? l) 0)
(else ( add1 ( length ( cdr l ) ) ) ) ) ) ) )
It sure i s. And this i s length9.
( ( lambda ( mk-length)
( mk-length
( mk-length
( mk-length eterity) ) ) )
(lambda ( length)
(lambda ( l )
(cond
( ( null? l) 0)
(else ( add1 ( length ( cdr l ) ) ) ) ) ) ) )
Sure. Here i t is.
( ( lambda ( mk-length)
( mk-length
( mk-length
( mk-length
( mk-length eterity) ) ) ) )
(lambda ( length)
(lambda ( l )
(cond
( ( null? l ) 0)
(else ( add1 ( length ( cdr l ) ) ) ) ) ) ) )
It i s like an infnite tower of applications of
mk-length to an arbitrary function.
Chapter 9
Do we really need an infnite tower?
Could we guess how many we need?
When do we fnd out that we didn' t guess a
large enough number?
What if we could create another application
of mk-length to eterity at this point?
Well , since nobody cares what function we
pass to mk-length we could pass it mk-length
initially.
Then is this still lengtho
( ( lambda ( mk-length)
( mk-length mk-length) )
(lambda ( length)
(lambda ( l )
(cond
( ( null ? l) 0)
( length ( cdr l ) ) ) ) ) ) ) )
Why would we want to do that?
. . . and Agan, and Agan, and Again, . . .
Not really of course. Everytime we use length
we only need a fnite number, but we never
know how many.
Sure, but we may not guess a large enough
number.
When we apply the function eternity that is
pased to the innermost mk-length.
That would only postpone the problem by
one, and besides, how could we do that ?
That ' s the right idea. And then we invoke
mk-length on eterity and the result of this
on the cdr so that we get one more piece of
the tower.
Yes, we could even use mk-length instead of
length.
( (lambda ( mk-length)
( mk-length mk-length) )
(lambda ( mk-length)
(lambda ( l )
(cond
( ( null ? l) 0)
( mk-length ( cdr l ) ) ) ) ) ) ) )
All names are equal , but some names are
more equal than others. 1
1 With apologies t o George Orwell ( 1903- 1950) .
165
True: as long as we use the names
consistently, we are j ust fne.
Now that mk-length is passed to mk-length
can we use the argument to create an
What is the value of
( ( (lambda ( mk-length)
( mk-length mk-length) )
(lambda ( mk-length)
(lambda ( l )
l)
(cond
( ( null ? l) 0)
( ( mk-length eterity)
( cdr l ) ) ) ) ) ) ) )
where
l is ( appl es)
Could we do this more than once?
166
And mk-length is a far more equal name than
length. If we use a name like mk-length, it is
a constant reminder that the frst argument
to mk-length is mk-length.
Yes, when we apply mk-length once, we get
length9
( (lambda ( mk-length)
( mk-length mk-length) )
(lambda ( mk-length)
(lambda ( l )
(cond
( ( null ? l) 0)
( ( mk-length eterity)
( cdr l ) ) ) ) ) ) ) )
This i s a good exercise. Work i t out with
paper and pencil.
Yes, j ust keep passing mk-length to itself,
and we can do this Ü often as we need to!
Chapter 9
What would you call this function?
( (lambda ( mk-length)
( mk-length mk-length) )
(lambda ( mk-length)
(lambda ( l )
(cond
( ( null ? l) 0)
( ( mk-length mk-length)
( cdr l ) ) ) ) ) ) ) )
Ho does i t work?
One problem is left: it no longer contains the
function that looks like length
((lambda ( mk-length)
( mk-length mk-length) )
(lambda ( mk-length)
(lambda ( l )
(cond
( ( null ? l) 0)
.
d1

<
I
<
mk-length mk-length)
( cdr l ) ) ) ) ) ) ) )
Can you fx that?
Why?
. . . and Agan, and Again, and Again, . . .
It is length, of course.
It keeps adding recursive uses by passing
mk-length to itself, j ust Ü it is about to
expire.
We could extract this new application of
mk-length to itself and call it length.
Because it really makes the function length.
167
( ( lambda ( mk-length)
( mk-length mk-length) )
(lambda ( mk-length)
( ( lambda ( length)
(lambda ( l )
(cond
( ( null ? l) 0)
(else ( add1 ( length ( cdr l ) ) ) ) ) ) )
( mk-length mk-length) ) ) )
Yes, this looks j ust fne.
Let ' s see whether it works. Okay.
What is the value of It should be 1.
( ( (lambda ( mk-length)
( mk-length mk-length) )
(lambda ( mk-length)
((lambda ( length)
(lambda ( l )
l)
where
(cond
( ( null ? l) 0)
(else ( add1 ( length ( cdr l ) ) ) ) ) ) )
( mk-length mk-length) ) ) )
l i s (appl es)
First , we need the value of
((lambda ( mk-length)
( mk-length mk-length) )
(lambda ( mk-length)
( (lambda ( length)
(lambda ( l )
That ' s true, because the value of this
expression is the function that we need to
apply to l where
168
(cond
( ( null ? l) 0)
(else ( add1 ( length ( cdr l ) ) ) ) ) ) )
( mk-length mk-length) ) ) )
l i s (appl es)
Chapter 9
So we really need the value of
((lambda ( mk-length)
((lambda ( length)
(lambda ( l )
(cond
( ( null? l) 0)
(else ( add1 ( length ( cdr l ) ) ) ) ) ) )
( mk-length mk-length) ) )
(lambda ( mk-length)
((lambda ( length)
(lambda (l)
(cond
( ( null ? l) 0)
(else ( add1 ( length ( cdr l ) ) ) ) ) ) )
( mk-length mk-length) ) ) )
But then we really need to know the value of
((lambda ( length)
(lambda ( l )
(cond
( ( null ? l) 0)
(else ( add1 ( length ( cdr l ) ) ) ) ) ) )
((lambda ( mk-length)
((lambda ( length)
(lambda ( l)
(cond
( ( null ? l) 0)
(else ( add1 ( length ( cdr l ) ) ) ) ) ) )
( mk-length mk-length) ) )
(lambda ( mk-length)
((lambda ( length)
(lambda (l)
(cond
( ( null? l) 0)
(else ( add1 ( length ( cdr l ) ) ) ) ) ) )
( mk-length mk-length) ) ) ) )
. . . and Again, and Again, and Again, . . .
Tue enough.
Yes, that ' s true, too. Where is the end of
this? Don' t we also need to know the value
of
( (lambda ( length)
(lambda ( l )
(cond
( ( null ? l) 0)
(else ( add1 ( length ( cdr l ) ) ) ) ) ) )
( ( lambda ( length)
(lambda ( l )
(cond
( ( null? l) 0)
(else ( add1 ( length ( cdr l ) ) ) ) ) ) )
( (lambda ( mk-length)
((lambda ( length)
(lambda (l)
(cond
( ( null ? l) 0)
(else ( add1 ( length ( cdr l ) ) ) ) ) ) )
( mk-length mk-length) ) )
(lambda ( mk-length)
( (lambda ( length)
(lambda ( l )
(cond
( ( null? l) 0)
(else ( add1 ( length ( cdr l ) ) ) ) ) ) )
( mk-length mk-length) ) ) ) ) )
169
Yes, there is no end to it . Why?
Is this strange?
But now that we have extracted
( mk-length mk-length)
from the function that makes length
it does not return a function anymore.
Trn the application of mk-length to itself in
our last correct version of length into a
function:
( ( lambda ( mk-length)
( mk-length mk-length) )
(lambda ( mk-length)
(lambda ( l )
(cond
( ( null ? l) 0)
(else
(
(
j'
(
m
-
k
-
-
-
le
-
ng
_
t
_
h
-
m
-
k
-
-l
-
en
-
g
-
th
--
)

(
cdr l ) ) ) ) ) ) ) )
Here i s a diferent way. If f i s a function of
one argument , is (lambda ¸J) (f J) ) a
function of one argument?
If ( mk-length mk-length) returns a function
of one argument , does
(lambda ¸ J)
( ( mk-length mk-length) J) )
return a function of one argument?
170
Because we j ust keep applying mk-length to
itself again and again and again . . .
It is because mk-length used to return a
function when we applied it to an argument.
Indeed, it didn' t matter what we applied it
to.
No it doesn' t. So what do we do?
How?
Yes, it is.
Actually,
(lambda ¸J)
( ( mk-length mk-length) J) )
i s Ü function!
Chapter 9
Okay, let' s do this to the application of
mk-length to itself.
Move out the new function so that we get
length back.
Is it oky to move out the function?
Can we extract the function in the box that
looks lik length and give it a name?
. . . and Agan, and Again, and Again, . . .
( ( lambda ( mk-length)
( mk-length mk-length) )
(lambda ( mk-length)
(lambda ( l )
(cond
( ( null ? l) 0)
(else
("(1 =a-m-b-d;a'(-x )
( ( mk-length mk-length) x) )
( cdr l ) ) ) ) ) ) ) )
( ( lambda ( mk-length)
( mk-length mk-length) )
(lambda ( mk-length)
( (lambda ( length)
(lambda ( l )
(cond
( ( null ? l) 0)
(else
( add1 ( length ( cdr l ) ) ) ) ) ) )
(lambda ( x)
( ( mk-length mk-length) x) ) ) ) )
Yes, we j ust always did the opposite by
replacing a name with its value. Here we
extract a value and give it a name.
Yes, it does not depend on mk-length at all !
171
Is this the right function?
( (lambda ( le)
((lambda ( mk-length)
( mk-length mk-length) )
(lambda ( mk-length)
( le (lambda ( x)
( ( mk-length mk-length) x) ) ) ) ) )
(lambda ( length)
(lambda ( l )
(cond
( ( null? l) 0)
(else ( add1 ( length ( cdr l ) ) ) ) ) ) ) )
Yes.
What did we actually get back? We extracted the original function mk-length.
Let ' s separate the function that makes length That 's easy.
from the function that looks like length
Does this function have a name?
Does (defne . . . ) work again?
Do you now know why Y works?
172
(lambda ( le)
((lambda ( mk-length)
( mk-length mk-length) )
(lambda ( mk-length)
( le (lambda ( x)
( ( mk-length mk-length) x) ) ) ) ) )
Yes, i t i s called the applicative-order Y
combinator.
(defne Y
(lambda ( le)
( ( lambda (/) (/
f
) )
(lambda (/)
( le (lambda ( x) ( (/ f
) x) ) ) ) ) ) )
Sure, now that we know what recursion is.
Read this chapter j ust one more time and
you will.
Chapter 9
What is ( Y Y)
. . . and Again, and Again, and Again, . . .
Who knows, but it works very hard.
Perhaps not after such a mind stretcher.
Stop the World-/ Want to Get Of.
Leslie Bricusse and Anthony Newley
173
:€
�&�W
� J � t C10� �
An entry is a par of lists whose frst list is a
set. Also, the two lists must be of equal
length. Make up some examples for entries.
How can we build an entry from a set of
names and a list of values?
What is ( lookup-in- entr name entr)
where name is entree
and
entr is ( ( appeti zer entree beverage)
(food tastes good) )
What i f name i s dessert
How can we accomplish this?
How many arguments do you think this extra
function should take?
What Is the Vaue of All of This?
Here are our examples:
( (appeti zer entree beverage)
( pate boeuf vi n) )
and
( (appeti zer entree beverage)
( beer beer beer) )
and
( ( beverage dessert)
( (food i s) ( number one wi th us) ) ) .
(defne new-entry build)
Ty to build our examples with this function.
tastes.
In this cae we would like to leave the
decision about what to do with the user of
lookup-in- entry.
the frst list of an entry.
We think it should take one, name. Why?
175
Here is our defnition of lookup-in-entr
(defne lookup-in-entr
(lambda ( name entr entr-f)
( lookup-in-entr-help name
(frst entr)
(second entr)
entr-f) ) )
Finish the function lookup-in-entr-help
(defne lookup-in-entr-help
(lambda ( name names values entr-f)
(cond
(
)
( )
(
) ) ) )
A table ¸also called an environment)i s a list
of entries. Here is one example: the empty
table, represented by ()
Make up some others.
Defne the function extend-table which takes
an entry and a table ¸possibly the empty
one) and creates a new table by putting the
new entry in front of the old table.
What is
( lookup-in-table name table table-f)
where
name is entree
table is ( ( ¸entree dessert)
¸spaghetti spumoni ) )
and
( ¸appetizer entree beverage)
(food tastes good) ) )
table-/ i s ¸lambda ( name) . . . )
176
(defne lookup-in-entr-help
(lambda ( name names values entr-f)
(cond
( ( null? names) ( entr-/ name) )
( ( eq? ( car names) name)
( car values) )
(else ( lookup-in-entry-help name
( cdr names)
( cdr values)
entr-f) ) ) ) )
Here i s another one:
¸ ¸ ¸appeti zer entree beverage)
¸ pate boeuf vi n) )
¸ ¸ beverage dessert)
¸ ¸food i s) ¸number one wi th us) ) ) ) .
¸defne extend-table cons)
It could be either spaghetti or tastes, but
lookup-in-table searches the list of entries in
order. So it is spaghetti .
Chapter 1 0
Write lookup-in-table
Hint: Don't forget to get some help.
Can you describe what the following function
represents:
(lambda ( name)
( lookup-in-table name
( cdr table}
table-f))
In the preface we mentioned that sans serif
typeface would be used to represent atoms.
To this point it ha not mattered.
Henceforth, you must notice whether or not
an atom is in sans serif.
Did you notice that "sans serif" wa not in
sans serif?
Have w chosen a good representation for
expressions?
What kind of functions?
Do you remember value from chapter 6?
What is the value of
( car (quote {a b c) ))
What Is the Vaue of All of This?
(defÎe lookup-in-table
(lambda (name table table-f)
[LoÎd
( ( null ? table) ( table-! name) )
(elÜe ( lookup-in-entry name
( car table}
(lambda ( name)
( lookup-in-table name
( cdr table}
table-f) ) ) ) ) ) )
This function i s the action to take when the
Remember to be very conscious Ûb to whether
or not an atom i s i n sans serif.
We hope so. This is ¯sÛnb serif'
in sans serif.
Yes. They are all S-expressions so they can
be data for functions.
For example, value.
Recal that value is the function that returns
the natural value of expressions.
We don't even know what (quote {Û b c) ) is.
177
What is the value of
¦ cons rp-a
( cons rp-b
( cons rp-c
(quote ( ) ) ) ) )
where
rep-a is a
rp-b is b
and
rp-c is c
Great . And what is the value of
( cons rp-car
( cons ( cons rp-quote
( cons
( cons rep-a
( cons rp-b
¦ cons rep-c
(quote ( ) ) ) ) )
(quote ( ) ) ) )
(quote ( ) ) ) )
where
rp-car is car
rep-quote is quote
rp-a is a
rep-b is b
and
rp-c is c
What is the value of
( car (quote (a b c) ) )
What i s ( value e)
where
e is (car (quote (a b c)) )
What i s ¸ value e)
where
e is (quote (car (quote (a b c) ) ) )
178
It is the same a (a b L) -
It is a representation of the expression:
(car (quote (a b c) ) ) .
a.
a.
(car (quote (a b c) ) ) .
Chapter 10
What is ( value e)
where
What is ( value e)
where e is 6
What is ( value e)
where
e is (quote nothi ng)
What is ( value e)
where
e is nothi ng
What is ( value e)
where
e is ( ( l ambda (nothi ng)
(cons nothi ng (quote ())))
(quote
(from nothi ng comes somethi ng) ) )
What i s ( value e)
where
e is ((l ambda (nothi ng)
(cond
(nothi ng (quote somethi ng))
(else (quote nothi ng) ) ) )
#t )
What is the type of Û
where
e is 6
What is the type of e
where
e is #f
What Is the Vaue of All of This?
7.
6, because numbers are constants.
nothi ng.
nothi ng has no value.
( (from nothi ng comes somethi ng) ) .
somethi ng.
*const.
*const.
179
What is { value e)
where
e is #f
What is the type of e
where e is cons
What is ( value e)
where e i s car
What is the type of e
where
e is {quote nothi ng)
What is the type of e
where
e is nothi ng
What is the type of e
where
e is (l ambda ¸A y) (cons A y))
What is the type of e
where
e is ((l ambda (nothi ng)
(cond
(nothi ng (quote somethi ng))
(else (quote nothi ng) ) ) )
#t )
What is the type of e
where
e is (cond
180
(nothi ng (quote somethi ng))
(else (quote nothi ng)))
#f.
*const.
{pri mi tive car) .
*quote.
*identifer.
*lambda.
*application.
*cond.
Chapte 10
How many types do you think there are?
How do you think we should represent types?
If actions are functions that do "the right
thing" when applied to the appropriate type
of expression, what should value do?
Do you remember atom-to-fnction from
chapter 8?
Below is a function that produces the correct
action (or function)for each possible
S-expression:
(lambda ( e)
(cond
( ( atom? e) ( atom-to-action e) )
(else ( list-to-action e) ) ) ) )
Defne the function atom-to-action
1
1
Ill-formed S-expressions such a (quote a b), () ,
(l ambda (#t) #t), (l ambda (5) 5), ( l ambda (car) car) ,
(l ambda a) , (cond (3 c) (else b) (6 a)), and (1 2) are not
considered here. They can be detected by an appropriate
function to which S-expressions are submitted before they
are pased on to value.
What Is the Value of All of This?
We found six:
*const
*quote
*identifer
*lambda
*cond
and
*application.
We choose functions. We call these functions
"actions. "
You guessed it. It would have to fnd out the
type of expression it was passed and then use
the associated action.
We found. atom-to-fnction useful when we
rewrote value for numbered expresssions.
(defne atom-to-action
(lambda ( e)
(cond
( ( number? e) *const)
( ( eq? e #t ) *const)
( ( eq? e #f ) *const)
( ( eq? e (quote cons) ) *const)
( ( eq? e (quote car) ) *const)
( ( eq? e (quote cdr) ) *const)
( ( eq? e (quote nul l ?) ) *const)
( ( eq? e (quote eq?)) *const)
( ( eq? e (quote atom?)) *const)
( ( eq? e (quote zero?) ) *const)
( ( eq? e (quote add1) ) *const)
( ( eq? e (quote sub1) ) *const)
( ( eq? e (quote number?) ) *const)
(else *identifer))))
181
Now defne the help function list-to-action
we can use it to defne value and meaning
(defne value
(lambda ( e)
(meaning e (quote ( ) ) ) ) )
(defne meaning
(lambda ( e table)
( ( exression-to-action e) e table) ) )
What i s (quote ()) i n the defnition of value
(defne list-to-action
(lambda ( e)
(cond
( ( atom? ( car e) )
(cond
( ( eq? ( car e) (quote quote) )
*quote)
( ( eq? ( car e) (quote l ambda})
*lambda)
( ( eq? ( car e) (quote cond} }
*cond}
(else *application) ) )
(else *application)) ))
It i s the empty table. The function value,
1
together with all the functions it uses, is
called an interpreter.
1
The function value approximates the function eVa1
available in Scheme (and Lisp).
Actions do speak louder than words.
How many arguments should actions take
according to the above?
182
Two, the expression e and a table.
Chapter 1 0
Here is the action for constants.
(defne *const
(lambda ( e table)
(Lond
(( number? e) e)
(( eq? e #t ) #t )
( ( eq? e #f ) #f )
(else ( build (quote pri mi tive) e) ) ) ) )
Is i t correct?
Here is the action for *quote
(defne *quote
(lambda ( e table)
( text-of e)))
Defne the help function text-of
Have we used the table yet?
Why do we need the table?
Given that the table contains the vlues of
identifers, write the action *identifer
Here is initial-table
(defne initial-table
(lambda ( name)
( car (quote ()))))
When i s i t used?
What is the vlue of (l ambda ¦A) A)
What Is the Vaue of All of This?
Yes, for numbers, it j ust returns the
expression, and this is all we have to do for
0, 1 , 2, . . .
For #t , it returns true.
For #f , it returns false.
And all other atoms of constant type
represent primitives.
(defne text-of second)
No, but we will in a moment .
To remember the vlues of identifers.
(defne *identifer
(lambda ( e table)
( lookup-in-table e table initial-table) ) )
Let's hope never. Why?
We don' t know yet, but we know that it
must be the representation of a non-primitive
function.
183
How are non-primitive functions diferent
from primitives?
So when we want to use a non-primitive we
need to remember its formal arguments and
its function body.
And what else do we need to remember?
And how do we represent this?
Here is the action *lambda
(defne *lambda
(lambda ( e table)
( build (quote non-pri mi ti ve)
( cons table ( cdr e) ) ) ) )
What i s ( meaning e table)
where
e is ( l ambda ¸x) (cons x y) )
and
table is (( (y z) ( (8)
9
) ) )
I t i s probably a good idea t o defne some
help functions for getting back the parts in
this three element list (i. e. , the table, the
formal arguments, and the body) . Write
table-of formals- of and body- of
Describe ( cond . . . ) in your own words.
184
We know what primitives do; non-primitives
are defned by their arguments and their
function bodies.
At least. Fortunately this is j ust the cdr of a
l ambda expression.
We will also put the table in, j ust in case we
might need it later.
In a list , of course.
( non-pri mi ti ve
( ( ( (y z) ( (8) 9) ) ) ¸x) (cons x y) ) )
� � �
table formals body
(defne table-of frst)
(defne formals-of second)
(defne body- of thir)
It is a special form that takes any number of
cond-lines. It considers each line in turn. If
the question part on the left is false, it looks
at the rest of the lines. Otherwise it proceeds
to answer the right part . If it sees an
else-line, it treats that cond-line as if its
question part were true.
Chapter 10
Here is the function evcon that does what we
j ust said in words:
(defne evcon
(lambda ( lines table)
(cond
( ( else? ( question- of ( car lines) ) )
( meaning ( answer- of ( car lines ) )
table) )
( ( meaning ( question- of ( car lines ) )
table)
( meaning ( answer-of ( car lines ) )
table) )
(else ( evcon ( cdr lines) table) ) ) ) )
Write else? and the help functions
Didn' t we violate The First Commandment?
Now use the function evcon to write the
*cond action.
Aren' t these help functions useful?
Do you understand *cond now?
How can you become familiar with it?
What Is the Value of All of This?
(defne else ?
(lambda ¸J)
(cond
( ( atom? J) ( eq? J (quote el se) ) )
(else #f ) ) ) )
(defne question-of frst)
Yes, we don' t ask ( null ? lines ) , so one of the
questions in every cond better be true.
¸defne *cond
(lambda ( e table)
( evcon ( cond-lines-of e) table) ) )
¸defne cond-lines-of cdr)
Yes, they make things quite a bit more
Perhaps not .
The best way is to try an example. A good
one is:
( *cond e table)
where
e is (cond (coffee kl atsch) (el se party) )
and
table is (( (cofee) ( #t ) )
( ( kl atsch party) (5 (6) ) ) ) .
185
Have we seen how the table gets used?
But how do the identifers get into the table?
How is an application represented?
How does an application difer from a special
form, like (and . . . ) (or . . . ) or (cond . . . )
Before we can apply a function, do we have
to get the meaning of all of its arguments?
Write a function evlis that takes a list of
( representations of) arguments and a table,
and returns a list composed of the meaning
of each argument .
What else do we need before we can
determine the meaning of an application?
And what then?
Here is *application
(defne *application
(lambda ( e table)
( apply
( meaning (fnction- of e) table)
( evlis ( aruments-of e) table) ) ) )
Is i t correct?
186
Yes, *lambda and *identifer use it .
In the only action we have not defned:
*application.
An application is a list of expressions whose
car position contains an expression whose
value is a function.
An application must always determine the
meaning of all its arguments.
Yes.
(defne evlis
(lambda ( ars table)
(cond
( ( null ? ars) (quote ( ) ) )
(else
( cons ( meaning ( car ars ) table)
( evlis ( cdr ars) table) ) ) ) ) )
We need to fnd out what its fnction-of
means.
Then we apply the meaning of the function
to the meaning of the arguments.
Of course. We j ust have to defne apply,
fnction-of, and aruments-of correctly.
Chapter 1 0
Write fnction-of and aruments-of
How many diferent kinds of functions are
there?
What are the two representations of
functions?
Write primitive ? and non-primitive ?
Now we can write the function apply
What Is the Vaue of All of This?
(defne fnction-of car)
(defne aruments-of cdr)
Two: primitives and non-primitives.
(pri mi tive primitive-name) and
( non- pri mi tive ( table formals body) )
The list ( table formals body) i s called a
closure record.
(defne primitive ?
(lambda ( l )
( eq? (frst l) (quote pri mi tive) ) ) )
(defne non-primitive ?
(lambda ( l )
( eq? (frst l) (quote non-pri mi ti ve) ) ) )
Here i t is:
¸defne apply1
(lambda (fn vals )
(cond
( (primitive ? fn)
( apply-primitive
(second fn) vals) )
( ( non-primitive ? fn)
( apply-closur
(second fn) vals) ) ) ) )
1 I f fn does not evaluate to either a primitive or a
non-primitive a in the expression ( ( l ambda (x) (x 5) ) 3) ,
there i s no answer. The function apply approximates the
function agg1§ available in Scheme ( and Lisp) .
187
This is the defnition of apply-primitive
(defne apply-primitive
(lambda ( name vals )
(cond
(( eq? name 1 )
( cons (frst vals ) (second vals) ) )
( ( eq? name (quote car) )
( car (frst vals) ) )
( ( eq? name (quote cdr) )
( 2 (frst vals) ) )
( ( eq? name (quote nul l ?) )
( null ? (frst vals) ) )
( ( eq? name (quote eq?) )
( 3 (frst vals) 4 ) )
( ( eq? name (quote atom?) )
( 5 (frst vals) ) )
( ( eq? name (quote zero?))
( zer? (frst vals) ) )
( ( eq? name (quote add1) )
( add1 (frst vals) ) )
( ( eq? name (quote sub1) )
(sub1 (frst vals) ) )
( ( eq? name (quote number?) )
( number? (frst vals) ) ) ) ) )
Fill i n the blanks.
Is apply-closur the only function left ?
How could we fnd the result of (f a b)
where
f is ( l ambda (x y) (cons x y) )
a i s 1
and
b is (2)
Why can we do this?
188
1. ¸quote cons)
2. cdr1
3. eq?
4. (second vals)
5. :atom?
¸defne :atom?
(lambda ¸J)
¸cond
( ( atom? J) #t )
( ( null ? J) #f )
( ( eq? ( car J) ¸quote pri mi ti ve) )
#t )
( ( eq? ( car J) ¸quote non-pri mi ti ve) )
#t )
¸else #f ) ) ) )
1 The function apply·primitive could check for applications
of cdr to the empty list or sub1 to 0, etc.
Yes, and apply-closur must extend the table.
That ' s tricky. But we know what to do to
fnd the meaning of
(cons x y)
where
table is ( ( (x y)
(1 (2) ) ) ) .
Here, we don' t need apply-closur.
Chapter 1 0
Can you generalize the last two steps?
Have you followed all this?
This is a complicated function and it
deserves an example.
What will be the new arguments of meaning
What Is the Value of All of This?
Applying a non-primitive function-a
closure-to a list of values is the same as
fnding the meaning of the closure' s body
with its table extended by an entry of the
form
(formals values)
In this entry, formals is the formals of the
closure and values is the result of evlis .
If not , here is the defnition of apply- closur.
(defne apply-closur
(lambda ( closur vals)
( meaning ( body-of closure)
( extend- table
( new-entry
(formals-of closur)
vals )
( table-of closur) ) ) ) )
In the following,
closur is ¸ ¸ ¸ ¸ U V W)
( 1 2 3) )
( (x y z)
(4 5 6) ) )
( x y)
¸LÕHb z x) )
and
vals is ( (a b L) (d Ü f) ) .
The new e for meaning will be ¸LLHb z x) and
the new table for meaning will be
( ( (x y)
((a b L) (d Ü f) ) )
¸ ¸ U V W)
(1 2 3) )
( (x y z)
(4 5 6) ) ) .
189
What is the meaning of ( cons Z J)
where Z is 6
and
J is (a b L)
Let ' s fnd the meaning of all the arguments.
What is
( evlis ars table)
where
ars is ¸Z x)
and
table is ( ( (x y)
( (a b c) (d e f) ) )
¸ ¸ U v W)
( 1 2 3))
( (x y z)
(4 5 6) ) )
What i s the ( meaning e table)
where e is z
What is ( meaning e table)
where e is x
So, what is the result of evlis
What is ( meaning e table)
where e is LÕHb
190
The same Ü
( meaning e table)
where
e is ¸LÕHb z x)
and
table is ( ( (x y)
((a b L) (d e f) ) )
¸ ¸ U v W)
( 1 2 3))
( (x y z)
(4 5 6) ) ) .
I n order t o do this, we must fnd both
( meaning e table)
where
e is z
and
( meaning e table)
where
e is x.
6, by using *identifer.
¸a b L) ¿ by using *identifer.
(6 ¸a b L) ) ¿ because evlis returns a list of the
meanings.
( pri mi ti ve LÕHb) ¿ by using *const.
Chapter 10
We are now ready to ( apply fn vals)
where
fn is ( pri mi ti ve LÕHb)
and
vals is (6 (a b L) )
Which path should we take?
Which cond-line is chosen for
( apply-primitive name vals)
where
name is LÕHb
and
vals is (6 (a b L) )
Are we fnished now?
But what about (defne . . . )
Is (defne . . . ) really not needed?
Does that mean we can run the interpreter
on the interpreter if we do the
transformation with the Y combinator?
What makes value unusual?
Should will-stop? see representations of
expressions?
Does it help?
else
What Is the Vaue of All of This?
The apply-primitive path.
The third:
( ( eq? name (quote LÕHb) )
( cons (frst vals) ( second vals ) ) ) .
Yes, we are exhausted.
It isn' t needed because recursion can be
obtained from the Y combinator.
Yes, but see The Seasoned Schemer.
Yes, but don' t bother.
It sees representations of expressions.
That may help a lot .
No, don' t bother-we can play the same
game again. We would be able to defne a
function like last- try? that will show that we
cannot defne the new and improved
will-stop?.
Yes, it ' s time for a banquet .
191
You' ve reached the intermission. What are your options? You could quickly run out and get the
rest of the show, The Seasoned Schemer, or you could read some of the books that we mention
below. All of these books are classics and some of them are quite old; nevertheless they have
stood the test of time and are all worthy of your notice. Some have nothing whatsoever to do
with mathematics or logic, some have to do with mathematics, but only by way of telling an
interesting story, and still others are j ust worth discovering. There should be no confusion: these
books are not here to prepare you to read the sequel , they are j ust for your entertainment. At
the end of The Seasoned Schemer you can fnd a set of references to Scheme and the reference
to Common Lisp. Do not feel obliged to j ump ahead to the next book. Take some time of and
read some of these books instead. Then, when you have relaxed a bit , perhaps removed some
of the calories that were foisted upon you, go ahead and dive into the sequel . Enjoy!
Abbott, Edwin A. Flatland. Dover Publications, Inc. , New York, 1952. ( Original publica­
tion: Seeley and Co. , Ltd. , London, 1884. )
Carroll, Lewis. The Annotated Alice: Alice 's Adventurs in Wonderland and Through the
Looking Glass. Clarkson M. Potter, Inc. , New York, 1960. Introduction and notes by
Martin Gardner. Original publications under diferent titles: Alice 's Adventurs Under
Grund and Through the Looking Glass and Wat Alice Found Ther, Macmillan and
Company, London 1865 and 1872, respectively.
Halmos, Paul R. Naive Set Theory. Litton Educational Publishers, New York, 1960.
Hein, Piet. Groks. The MIT Press, Cambridge, Masachusetts, 1960.
Hofstadter, Douglas R. Godel, Escher, Bach: an Eteral Golden Bmid. Basic Books, Inc. ,
New York, 1979.
Nagel, Ernest and James R. Newman. Godel 's Prof. New York University Press, New
York, 1958.
P6lya, Gyorgy. How to Solve It. Doubleday and Co. , New York, 1957.
Smullyan, Raymond. To Mock a Mockingbir And Other Logic Puzzles Including an
Amazing Adventur in Combinatory Logic. Alfred A. Knopf, Inc. , New York, 1985.
Suppes, Patrick. Intrduction to Logic. Van Nostrand Co. , Princeton, New Jersey, 1957.
Intermission 193
ÏndCX
*application, 186 eq?-c, 127
*const, 183 eq?-tuna, 136
*identifer, 183 eqan?, 78
*lambda, 184 eqlist ?, 91 , 92, 94
*quote, 183 eqset ?, 1 14, 1 1 5
+, 60, 108 equal ?, 93
¯g 61 eternity, 151
7¿ 65 evcon, 185
+, 75 even?, 144
¡, 74 evens-only*, 144
<, 73 evens-only*fco, 145
¯ç 74 evlis , 186
>, 72, 73 expression-to-action, 181
???, 74 extend-table, 1 76
1st-sub-ex, 105, 106
2nd-sub-ex, 106
frst, 1 1 9
frsts , 43, 44, 46
A, 156 formals-of, 184
a-fiend, 138 fllfn?, 122
a-pair?, 1 18
fn?, 120
align, 152
initial-table, 183
all-nums , 78
insert-g, 132
insertL, 5 1 , 132
apply, 187
insertL *, 86
apply-closure, 189
insertL-f, 130
apply-primitive, 188
insertR, 48-50, 132
aruments-of, 187
insertR *, 82
atom?, 10
insertR-f, 130
atom-to-action, 181
intersect , 116
atom-to-fnction, 134
intersect?, 1 15
intersect all , 1 1 7
body-of, 184
build, 1 19 keep-looking, 150
C, 155 last-fiend, 140
cond-lines-of, 185 last-try, 158
cookies, 123 lat ?, 16, 19, 109
latest-fiend , 139
difernce, 1 1 7
lefmost , 88
edd1 , 108
length, 76, 160
else ?, 185
length*, 1 53
Index 195
list-to- action, 182
looking, 149
lookup-in-entry, 1 76
lookup-in-entry-help, 176
lookup-in-table, 177
makeset , 112
meaning, 182
member* , 87
member?, 22
multiinsertL, 56, 57, 141
multiinsertLR, 141
multiinsertLR&co, 142, 143
multiinsertR, 56, 141
multirmber, 53, 135
multirmber&co, 137
multirmber-eq?, 136
multirmber-f, 135
multirmberT, 137
multisubst, 57
new- entry, 175
new-fiend, 139
no-nums , 77
non-primitive ?, 187
numberd?, 99-101
occur, 78
occur*, 85
one-to- one ?, 122
one ?, 79
opertor, 106
pick, 76
primitive ?, 187
question-of, 185
rmber, 34, 37, 41, 94, 95
rmber*, 81
rmber-eq?, 129
196
rmber-f, 126, 128, 129
rempick, 77, 79
rvpair, 121
rvrl , 120, 121
second, 119
seconds , 122
seqL, 131
seqR, 131
seqrm, 133
seqS, 133
ser?, 108
set ?, 1 1 1
shif, 152
shufe, 154
sub1 , 59
subset ?, 1 13, 114
subst, 51, 133
subst *, 85
subst2, 52
table- of, 184
text-of, 183
the-last-fiend, 146
thir, 1 19
tup+, 69, 71
union, 116
value, 102-104, 106, 134, 135, 182
weight *, 154
will-stop?, 157
JJJ¿ 117
Y, 172
yyy, 133
zub1 , 108
Index

The Ten Commandments

The First Commandment

The Fifth Commandment
When building a value with + ,always use

lat, ask (null? lat) and else. When recurring on a number, n, ask two questions about it: (zero? n) and else. When recurring on a list of S-expressions, l, ask three question about it: (null? l), (atom? ( car l)), and else.
When recurring on a list of atoms, two questions about it:

0 for the value of the terminating line, for adding 0 does not change the value of an

addition. When building a value with x, always use

multiplying by 1 does not change the value of a multiplication. When building a value with consider line.

1 for the value of the terminating line, for

The Second Commandment
Use cons to build lists.

()

cons,

always

for the value of the terminating

The Third Commandment
When building a list, describe the first typical element, and then ral recursion.

cons

The Sixth Commandment
Simplify only after the function is correct.

it onto the natu­

The Fourth Commandment
Always change at least one argument while recurring. When recurring on a list of atoms,

The Seventh Commandment
Recur on the nature:
• •

subparts

that are of the same

(cdrlat). When recurring on a num­ ber, n, use (sub1 n). And when recurring on a list of S-expressions, l, use (carl) and (cdr l) if neither (null? l) nor (atom? (carl)) are
use true. It must be changed to be closer to termina­ tion. The changing argument must be tested in the termination condition: when using and when using

lat,

On the sublists of a list. On the subexpressions of an arithmetic

expression.

The Eighth Commandment
Use help functions to abstract from represen­ tations.

cdr,

test termination with

null? The Ninth Commandment
Abstract common patterns with a new func­ tion.

zero?.

sub1,

test termination with

The Tenth Commandment
Build functions to collect more than one value at a time.

The Five Rules

The Law of Car
The primitive empty lists.

car

is defined only for non­

The Law of Cdr
The primitive

empty lists. The

cdr is defined only for non­ cdrof any non-empty list

is always another list.

The Law of Cons
The primitive

cons

takes two arguments.

The second argument to The result is a list.

cons must be a list.

The Law of Null?
The primitive

null?

is defined only for lists.

The Law of Eq?
The primitive eq'l takes two arguments.

Each must be a non-numeric atom.

The Little Schemer
Fourth Edition

Daniel P. Friedman
Indiana Uni11ersity Bloomington, Indiana

Matthias Felleisen
Rice Uni11ersity Houston, Texas

Drawings by Duane Bibby Foreword by Gerald J. Sussman

The MIT Press Cambridge, Massachusetts London, England

III. recording. ISBN 0-262-56099-2 (pbk: alk.S34F75 005. Title. of: The little LISPer. Matthias. Little LISPer. No part of this book may be reproduced in any form by any electronic or mechanical means (including photocopying. Felleisen. @1989. 1st MIT Press ed. © 1996 Massachusetts Institute of Technology All rights reserved. First MIT Press Edition. Friedman. paper) 1. Rev.13'3-dc20 95-39853 CIP . Daniel P. em. This book was set by the authors and was printed and bound in the United States of America. Includes index. . Daniel P. drawings by Duane Bibby.-4 ed. foreword by Gerald J. or information storage and retrieval) without permission in writing from the publisher. © 1986.73. Friedman and Matthias Felleisen. ed. 3rd ed. The little schemer / Daniel P. 1987. Sussman. II. 1996 QA76. LISP (Computer program language) I. p. 1974 by Scientific Research Associates. Scheme (Computer program language) 2.Original edition published as The Little LISPer . Library of Congress Cataloging-in-Publication Data Friedman.

Helga.To Mary. and our children .

.

Do It. Numbers Games) 58) ((5. . and Again.( (Contents) (Foreword (Preface ix) xi) ((1. What Is the Value of All of T his?) 174) (Intermission 192) (Index 194)) .. Cons the Magnificent) 32) ((4..) 148) ((10.. and Again. Lambda the Ultimate) 124) ((9. Toys) 2) ((2. Friends and Relations) 110) ((8..) 14) ((3. Do It Again. and Again. *Oh My Gawd*: It's Full of Stars) 80) ((6. ... and Again . and Again. Shadows) 96) ((7.

.

and mixing chemicals. Lisp provides the glue for using a vast library of canned parts. One must learn to ritualize the process of developing sensitive material so that one gets consistent results over many years of work. naming. In engineering. much of the drill and practice that is necessary to learn the skills of constructing recursive processes and manipulating recursive data-structures. we must learn to do analysis to support our efforts in synthesis. For the student of Lisp programming. as in other creative arts. In support of this skill we had to learn the use of exposure meters to measure light intensities and the use of exposure time and development time to control the black level and the contrast in the image. The Little L ISPer can perform the same service that Hanon's finger exercises or Czerny's piano studies perform for the student of piano. programming requires a choice of medium. Some photographers choose to use black-and-white 8x10 plates while others choose 35mm slides. it provides mechanisms for formalizing. This is in turn supported by even lower level skills such as loading film . Gerald J . In addition . procedures are first-class data. with considerable wit . Most of the students (including me) came into that course hoping to learn how to be creative-to take pictures like the ones I admired by artists such as Edward Weston. developing and printing. On the first day the teacher patiently explained the long list of technical skills that he was going to teach us during the term. It has developed into a uniquely powerful and flexible family of software development tools. Lisp was initially conceived as a theoretical vehicle for recursion theory and for symbolic algebra. Each has its advantages and disadvantages. and stored in data structures. In 1967 I took an introductory course in photography. The Little LISPer is a unique approach to developing the skills underlying creative program­ ming in Lisp. We reprint it here with the permission of the author. Sussman Cambridge. produced by members of the user community. As with other languages. to be passed as arguments. One cannot build a beautiful and functional bridge without a knowledge of steel and dirt and considerable mathematical technique for using this knowledge to compute the properties of structures. Similarly.Foreword This foreword appeared in the second and third editions of The Little L ISPer. This flexibility is valuable. But what about creative composition? In order to be creative one must first gain control of the medium. providing wrap-around support for the rapid-prototyping of software systems. A key was Ansel Adams' "Zone System" for previsualizing the print values (blackness in the final print) in a photograph and how they derive from the light intensities in the scene. Like photography. In Lisp. The first laboratory session was devoted to finding out that developer feels slippery and that fixer smells awful. Lisp programs can easily manipulate the representations of Lisp programs-a feature that has encouraged the development of a vast structure of program synthesis and analysis tools. such as cross-referencers. and saving the idioms-the common patterns of usage that are essential to engineering design . Massachusetts Foreword ix . It painlessly packages. one cannot build a beautiful computer system without a deep understanding of how to "previsualize" the process generated by the procedures one writes. One can not even begin to think about organizing a great photograph without having the skills to make it happen . but most importantly. returned as values. Lisp is the medium of choice for people who enjoy free style and flexibility.

Preface

To celebrate the twentieth anniversary of Scheme we revised The Little LISPer a third time, gave it the more accurate title The Little Schemer, and wrote a sequel: The Seasoned Schemer. Programs accept data and produce data. Designing a program requires a thorough understand­ ing of data; a good program reflects the shape of the data it deals with. Most collections of data, and hence most programs, are recursive. Recursion is the act of defining an object or solving a problem in terms of itself. The goal of this book is to teach the reader to think recursively. Our first task is to decide which language to use to communicate this concept . There are three obvious choices: a natural language, such as English; formal mathematics; or a programming language. Natural languages are ambiguous, imprecise, and sometimes awkwardly verbose. These are all virtues for general communication, but something of a drawback for communicating concisely as precise a concept as recursion. The language of mathematics is the opposite of natural language: it can express powerful formal ideas with only a few symbols. Unfortunately, the language of mathematics is often cryptic and barely accessible without special training. The marriage of technology and mathematics presents us with a third , almost ideal choice: a programming language. We believe that programming languages are the best way to convey the concept of recursion. They share with mathematics the ability to give a formal meaning to a set of symbols. But unlike mathematics, programming languages can be directly experienced-you can take the programs in this book, observe their behavior, modify them , and experience the effect of these modifications. Perhaps the best programming language for teaching recursion is Scheme. Scheme is inherently symbolic-the programmer does not have to think about the relationship between the symbols of his own language and the representations in the computer. Recursion is Scheme's nat­ ural computational mechanism; the primary programming activity is the creation of (potentially) recursive definitions. Scheme implementations are predominantly interactive-the programmer can immediately participate in and observe the behavior of his programs. And, perhaps most importantly for our lessons at the end of this book, there is a direct correspondence between the structure of Scheme programs and the data those programs manipulate. Although Scheme can be described quite formally, understanding Scheme does not require a particularly mathematical inclination. In fact , The Little Schemer is based on lecture notes from a two-week "quickie" introduction to Scheme for students with no previous programming experience and an admitted dislike for anything mathematical . Many of these students were preparing for careers in public affairs. It is our belief that writing programs recursively in Scheme is essentially simple pattern recognition. Since our only concern is recursive programming, our treatment is limited to the whys and wherefores of just a few Scheme features: car, cdr, cons, eq? , null?, zero? , addl, subl, number?, and, or, quote, lambda, define, and cond. Indeed, our language is an idealized Scheme. The Little Schemer and The Seasoned Schemer will not introduce you to the practical world of programming, but a mastery of the concepts in these books provides a start toward understanding the nature of computation.

Preface

xi

(define at om? (lambda

(x)

(and (not (pair?

x))

(not (null?

x)))))

To find out whether your Scheme has the correct definition of at om?, try (atom? (quote ())) and make sure it returns #f. In fact , the material is also suited for modern Lisps such as Common Lisp. To work with Lisp, you will also have to add the function atom?:
(defun atom? (not (listp

(x) x)))

xii

Preface

Preface

xiii

.

The Little Schemer .

.

which is a letter. Is it true that this is a list? (atom )1 Yes. 1 L. because u is a string of one character. Is it true that this is an atom? tu rkey Yes. 1 L. S: (quote (atom) ) or ' ( atom) Is it true that this is a list? (atom tu rkey or) Yes. because it is a collection of atoms enclosed by parentheses.Is it true that this is an atom? atom1 Yes. because 1492 is a string of digits. because atom is a string of characters beginning with the letter a . because •abc\$ is a string of characters beginning with a letter or special character other than a left " ( " or right " ) " parenthesis. S: (quote atom) or 'atom "L:" and "S:" are described in the preface. because (atom) is an atom enclosed by parentheses. Is it true that this is an atom? 1492 Yes. Toys 3 . Is it true that this is an atom? •abc\$ Yes. Is it true that this is an atom? u Yes. because tu rkey is a string of characters beginning with a letter.

because it is a list . Is it true that this is a list? ( ( ( how) are) ( (you) (doing so) ) far) Yes . because i t i s a collection of S-expressions enclosed by parentheses. you . and the second one is an atom. ( (you ) (doing so) ) . How many S-expressions are in the list (how are you doing so far) and what are they? Six. so. and far. because these are actually two S-expressions not enclosed by parentheses. Is it true that this is a list? (how are you doing so far) Yes . because all atoms are S-expressions. and far. Is it true that this is an S-expression? xyz Yes. how. How many S-expressions are in the list ( ( (how) are) ( (you ) (doing so) ) far) and what are they? Three. ( ( how) are) . Is it true that this is an S-expression? (x y z ) Yes. Is it true that this is a list? ( (atom tu rkey) or) Yes. Is it true that this is an S-expression? ((x y) z ) Yes. 4 Chapter 1 . because all lists are S-expressions. because i t i s a collection of S-expressions enclosed by parentheses. are. doing. The first one is a list containing two atoms. because the two S-expressions are now enclosed by parentheses.Is it true that this is a list? (atom tu rkey) or No.

No answer.Is it true that this is a list? () Yes. No. This special S-expression is called the null (or empty) list . of l car of the empty 1 L: nil The Law of C ar The primitive car is defined only for non-empty lists. because it is a collection of S-expressions enclosed by parentheses. You cannot ask for the No answer. Yes. because (a b c) is the first S-expression of this non-empty list . car of an atom . because it contains zero S-expressions enclosed by parentheses. 1 You cannot as k for the list . because a is the first atom of this list . because () is just a list . Is it true that this is an atom? () Is it true that this is a list? (() () () ( ) ) What i s the car of l where l is the argument (a b c) What is the car of l where l is ((a b c) x y z) What is the c a r of l where l is hotdog What is the where l is () car (a b c). Toys 5 . a.

because (b c) is the list l without ( car l). " ( ( hotdogs)) is the first S-expression of l . What is ( cdr l) where l is ( (x) t r) (t r) ." (b c) . What i s the cdr of l where l is (a b c) Note: "cdr" is pronounced "could-er. What is the cdr of l where l is (hamburger) () . What is the cdr of l where l is ((a b c) x y z) (x y z) ." What is ( cdr a) where a is hotdogs No answer. You cannot ask for the cdr of an atom. because ( cdr l) is just another way to ask for "the cdr of the list l. 6 Chapter 1 . What is ( car l) where l is ( ( ( hotdogs) ) (and) (pickle) relish) ( ( hotdogs) ) ." What is ( car ( car l)) where l is ( ( ( hotdogs)) (and ) ) (hotdogs) .What is the car of l where l is ( ( ( hotdogs)) (and) (pickle) relish) ( ( hotdogs) ) . because ( car l) is another way to ask for "the car of the list l. read as: "The list of the list of hotdogs.

" Toys 7 . 1 L: nil The Law of Cdr The primitive cdr is defined only for non-empty lists. since ( car l) is an atom. N o answer.What is ( cdr l) where l is () No answer. What does car take as an argument? What does cdr take as an argument? What is the cons of the atom a and the list l where a is peanut and l is ( butter and jelly) This can also be written "( cons a l)" . It takes any non-empty list . The cdr of any non­ empty list is always another list. 1 You cannot ask for the cdr of the null list . and cdr does not take an atom as an argument. because ((x y) ((c))) is ( cdr l). and (x y) is the car of ( cdr l ) . (((c))) . ( pea nut butter and jel ly) . What is ( car ( cdr l)) where l is ((b) (x y) ((c))) What is ( cdr ( cdr l)) where l is ((b) (x y) ((c))) What is ( cdr ( car l)) where l is (a (b (c)) d) (x y) . see The Law of Cdr. and (((c))) is the cdr of ( cdr l ) . It takes any non-empty list . because cons adds an atom to the front of a list . Read: "cons the atom a onto the list l. because ((x y) ((c))) is ( cdr l ) .

because cons adds any S-expression to the front of a list. What is ( cons s l) where s is ((a b c)) and l is b No answer. 1 since the second argument l must b e a list .What is the cons of s and l where s is ( banana and) and l is (peanut butter and jelly) ( (ba nana and) peanut butter and jelly) . 1 In practice. What is ( cons where s is a and l is ( ) s l) (a) . What is ( cons s l) where s is (a b (c)) and l is () ((a b ( c))) . What is ( cons where s is a and l is b s l) No answer. Why? 8 Chapter 1 . and (car (cona a /3)) =a (cdr (cona a /3)} = /3. (cona a /3) works for all values a and /3. What does cons take as its arguments? cons takes two arguments: the first one is any S-expression. the second one is any list . because () is a list . What is ( cons s l) where s is ( (help) this) and l is (is very ( ( hard) to learn) ) ( ( (help) this) is very ((hard) to learn ) ) .

The second argument to cons must be a list. Why? Is i t true that the list l i s the null list where l is ( ) Yes. because it is the list composed of zero S-expressions. because (quote ( ) ) 1 is a notation for the null list . The result is a list. This question can also be written: ( null? l). because l is a non-empty list .The Law of Cons The primitive cons takes two arguments. Is ( null? l) true or false where l is (a b c) False. 1 L : null 1 L: Also 0 and •(). Why? What i s ( cons s ( CJJ. S: Also •().r l ) ) where s is a and l is ((b) c d ) (a c d ) . What is ( cons s ( car l)) where s is a and l is ((b) c d ) ( a b) . Toys 9 . What is ( null?1 (quote () )) True.

since s is a list. 1 In practice. except the empty list. The Law of Null? The primitive null ? is de­ fined only for lists. because ( atom ? s ) is just another way to ask "Is s is an atom?" 1 L: (defun atom? ( x ) (not (listp x ) ) ) S: (def ine atom? (lambda (x) (and (not (pair? x)) (not (null? x) ) ) ) ) Is ( atom ? s ) true or false where s is (Harry had a heap of apples) False. because Harry is a string of characters beginning with a letter. How many arguments does atom? take and what are they? It takes one argument. 10 Chapter 1 . Is ( atom ?1 s ) true or false where s is Harry True. (null? a) is false for everything. Is it true or false that where s is Harry s is an atom True.Is ( null ? a) true or false where a is spaghetti No answer/ because you cannot ask null? of an atom. The argument can be any S-expression.

which i s an atom . False. since a1 and a2 are different atoms.Is ( atom ? ( car l)) true or false where l is (Harry had a hea p of a pples) Is ( atom ? ( cdr l)) true or false where l is (Harry had a heap of apples) Is ( atom ? ( cdr l)) true or false where l is (Harry) Is ( atom ? ( car ( cdr l))) true or false where l is (swi ng low sweet cherry oat) Is ( atom ? ( car ( cdr l))) true or false where l is (swing (low sweet) cherry oat) True or false: a1 and a2 are the same atom where a1 is Harry and a2 is Harry Is ( eq?1 a1 a2 ) true or false where a1 is Harry and a2 is Harry True. which is a list . True. False. True. and ( car ( cdr l)) i s low. "Are a1 and a2 the same non-numeric atom?" 1 L: eq Is ( eq? a1 a2) true or false where a1 is margarine and a2 is butter False. because a1 is the atom Harry and a2 is the atom Harry. and ( car ( cdr l)) is (low sweet) . since ( cdr l) is (( low sweet) cherry oat ) . because ( cdr l) is (low sweet cherry oat ) . and Harry is an atom. False. because the list ( ) is not an atom . Toys 11 . True. because ( eq ? a1 a2 ) is just another way to ask. because ( car l) is Harry.

How many arguments does eq ? take and what are they? It takes two arguments. 12 Chapter 1 . Both of them must be non-numeric atoms.1 6 and 7 are numbers. some numbers may be arguments of eq?. eq? if they are the same list. 1 ( ) and (strawberry) are lists. and the argument a is also the atom Mary. because ( car l) is the atom Mary. Is ( eq ? ( cdr l) a ) true or false where l is (sou red milk) and a is m i l k No answer. Is ( eq ? ( car l) a ) true or false where l is (Mary had a little lamb chop) and a is Mary True. The Law of Eq? The primitive eq? takes two ar­ guments. 1 In practice. lists m83' be arguments of eq?. 1 In practice. Each must be a non­ numeric atom. Two lists are Is ( eq ? n1 n2 ) true or false where n1 is 6 and n2 is 7 No answer. See The Laws of Eq? and Cdr. Is ( eq ? l1 l2 ) true or false where l1 is () and l2 is (strawberry) No answer.

==>Now go make yourself a peanut butter and jelly sandwich.Is ( eq? ( car l) ( car ( cdr l))) true or false where l is (beans beans we need jelly beans) True. ¢= This space reserved for JELLY STAINS! 13 . because it compares the first and second atoms in the list .

�))� �m��m �� �t.D��� ��m ��� .

True! Every lat is a list of atoms! Write the function lat ? using some. because each S-expression in l is an atom . Good luck. and Again . but not necessarily all. . and Again. True or false: ( lat ? l) where l is ( (Jack) Sprat cou ld eat no chicken fat) False. True or false: ( lat ? l) where l is () True. of the following functions: car cdr cons null? atom ? and eq ? You were not expected to be able to do this yet . because it does not contain a list . True or false: a lat is a list of atoms. . True or false: ( lat? l) where l is (Jack (Sprat could) eat no chicken fat) False.True or false: ( lat ? l) where l is (Jack Sprat could eat no chicken fat) True. Do It Again. Go on to the next question . 15 . since ( car l) is a list . since one of the S-expressions in l is a list. because you are still missing some ingredients. Are you rested? Do It.

What is the meaning of the cond line ( (null? l) #t ) where l is (bacon and eggs) - ( null? l) asks if the argument l is the null list. . What is the next question? ( atom ? ( car l ) ) . . the value of the application is true. . l is not the null list. In this case. and (define . . If it is not .(define lat ?1 (lambda (l) (cond ( ( null ? l) #t ) ( ( atom ? ( car l)) ( lat ? ( cdr l))) (else #f ) ) ) ) What i s the value of ( lat ? l ) where l is the argument (bacon and eggs) #t . 16 Chapter 2 . 1 L: (defun 1at? (1} (cond ( (null 1} t} ( (atom? (car 1} } (lat? (cdr 1} } } ( t nil} } } How do we determine the answer # t for the application ( lat ? l) You were not expected to know this one either. ) gives it a name. The application ( lat ? l) where l is (bacon and eggs) has the value #t -true-because l is a lat . ) creates a function. (lambda . we ask the next question. ) asks questions. The answer is determined by answering the questions asked by latf Hint: Write down the definition of the function lat ? and refer to it for the next group of questions. so we ask the next question. What is the first question asked by ( lat ? l) ( null? l) Note: (cond . . If it is. .

What is the meaning of the line ( ( atom ? ( car l)) ( lat? ( cdr l))) where l is (and eggs) ( atom ? ( car l)) asks i f ( car l) i s an atom . and Again. If it is. so the value of the function is the value of ( lat ? ( cdr l ) ) . which is (eggs). What is the meaning of ( lat ? ( cdr l)) ( lat ? ( cdr l)) finds out if the rest of l is composed only of atoms. Do It Again. In this case. In this case. we want to know if the rest of l is also composed only of atoms. What is the next question? ( null ? l). so we want to find out if the rest of the list l is composed only of atoms. l is not the null list . we ask the next question. with the argument ( cdr l ) . If ( car l) is an atom . What is the meaning of the line (( null? l) #t ) where l is now (and eggs) ( null ? l) asks if the argument l is the null list . . .What is the meaning of the line (( atom ? ( car l)) ( lat? ( cdr l))) where l is (bacon and eggs) ( atom ? ( car l)) asks if the first S-expression of the list l is an atom . by referring to the function with a new argument . What is the next question? ( atom ? ( car l ) ) . If ( car l) is not an atom. the value of the application is ( lat ? ( cdr l ) ) . we ask the next question . If not . we ask the next question. ( car l) is an atom . ( car l) is an atom . Do It. Now what is the argument l for lat ? Now the argument l is ( cdr l ) . which is (and eggs) . but this time. What is the meaning of ( lat? ( cdr l)) ( lat ? ( cdr l)) finds out i f the rest of the list l is composed only of atoms. the value of the application is #t . by referring again to the function lat ?. In this case. 17 . so we ask the next question. If it is an atom . If it is not . and Again .

ask the next question. move to the next question. If ( car l) is not an atom. What is the meaning of ( lat ? ( cdr l)) ( lat ? ( cdr l)) finds out if the rest of the list l is composed only of atoms. l is not null. D o you remember the question about ( lat ? l) Probably not . If it is. with l becoming the value of ( cdr l ) . is #t -true. ( car l) is an atom . What is the meaning of the line ( ( atom? ( car l)) ( lat ? ( cdr l))) where l is now (eggs) ( atom ? ( car l)) asks i f ( car l) i s an atom. what is the argument for lat ? () . In this case. by referring to the function lat ?. so we ask the next question. So. () is the null list. The application ( lat? l) has the value #t if the list l is a list of atoms where l is (bacon and eggs) . the value of the application ( lat ? l) where l is (bacon and eggs) . What is the meaning of the line ( ( null ? l) #t ) where l is now ( ) ( null ? l) asks if the argument l is the null list . we ask the next question. What is the next question? ( atom ? ( car l ) ) . the value of the application is ( lat ? ( cdr l ) ) . the value of the application is the value of #t . Now. In this case.What is the next question? ( null? l ) . If it is. In this case. If not . the value of the application is #t -true. What is the meaning of the line ( ( null? l) #t ) where l is now (eggs) ( null? l) asks i f the argument l i s the null list . so once again we look at ( lat? ( cdr l)). If it is. 18 Chapter 2 . If it is not.

since the list l contains an S-expression that is a list . and asks if each S-expression is an atom. so we want to check if the rest of the list l is composed only of atoms. Do It. If it runs out without encountering a list . the value is #t . What is the meaning of the line ( ( atom ? ( car l)) ( lat ? ( cdr l))) where l is (bacon (and eggs)) ( atom ? ( car l)) asks i f ( car l) i s a n atom . What is the meaning of the line ( ( null? l) #t ) where l is (bacon (and eggs)) ( null ? l) asks i f l is the null list . the value is ( lat ? ( cdr l ) ) . 19 . What is the first question? ( null ? l ) . What is the next question? ( atom ? ( car l ) ) . move to the next question. If it is not . In this case.. If l is not null. we ask the next question. If it is. it is not null. the value is #t . In this case.Can you describe what the function lat ? does in your own words? Here are our words: "lat ? looks at each S-expression in a list . If it finds a list . and Again ." consider the next few questions. in turn. the value is #f-false. ( car l) is an atom . and Again. Do It Again. so we ask the next question. This is the function lat ? again: (define lat ? (lambda (l) (cond ( ( null? l) #t ) ( ( atom ? ( car l)) ( lat ? ( cdr l))) (else #f)))) What is the value of ( lat? l) where l is now (bacon (and eggs)) #f." To see how we could arrive at a value of "false.. until it runs out of S-expressions . I f i t i s .

If it is null. the value is #t . What is the meaning of the line ( ( null? l) #t ) where l is now ((and eggs) ) ( null? l) asks if l is the null list. What is the next question? else. the value is ( lat ? ( cdr l)). If it is not null. by referring to lat ? with l replaced by ( cdr l). If else is true-as it always is-then the answer is #f-false. can have an atom in the first position. If it is not . ( car l) is not an atom . 20 Chapter 2 . or can have a list in the first position . Because we do not need to ask any more questions. so move to the next question. because the question else is always true! What is the meaning of the question else Is else true? else Why is else the last question? Of course. In this case. Why do we not need to ask any more questions? Because a list can be empty. else asks if else is true.What is the meaning of ( lat ? ( cdr l)) ( lat ? ( cdr l)) checks to see if the rest of the list l is composed only of atoms. In this case. l is not null. so we ask the next question . What is the meaning of the line (else #f ) else asks if else is true. If it is. What i s the next question? ( atom ? ( car l)). we ask the next question. we move to the next question. What is the meaning of the line ( ( atom ? ( car l)) ( lat ? ( cdr l))) where l is now ( (and eggs)) ( atom ? ( car l)) asks if ( car l) is an atom. Yes.

21 .. Do It. " Is ( or ( null? ll ) ( atom ? l2 ) ) true or false where l1 is () and l2 is (d e f g) True.. and Again . If it finds a list . .. . because ( null ? l1 ) is true where l1 is ( ) . If the first one is true it stops and answers true. Can you describe how we determined the value #f for ( lat? l) where l is ( bacon (and eggs)) Here is one way to say it : "( lat? l) looks at each item in its argument to see if it is an atom . What does ( or .What is ))) These are the closing or matching parentheses of ( cond . . the value of ( lat ? l) is #f. (lambda . . one at a time. . which appear at the beginning of a function definition. Do It Again. because ( null ? l2) is true where l2 is ( ) . Otherwise it asks the second question and answers with whatever the second question answers. . . . because neither ( null? l1 ) nor ( null? l2 ) is true where l1 is (a b c) and l2 is (atom ) . as it did in the example ( bacon (and eggs) ) . If it runs out of items before it finds a list . . False. . and (define . ) do? ( or . ) asks two questions. and Again. . the value of ( lat ? l) is #t . . Is ( or ( null? l1 ) ( null? l2 ) ) true or false where l1 is (a b c) and l2 is () Is ( or ( null? l1 } ( null? l2 ) ) true o r false where l1 is (a b c) and l2 is (atom ) True.

(mashed potatoes and meat gravy) . 22 Chapter 2 . What is the first question asked by ( member? a lat) ( null? lat ) . (coffee tea or milk) is the same as the atom a-tea . because one of the atoms of the lat. How do we determine the value #t for the above application? The value is determined by asking the questions about ( member? a lat).Is it true or false that a is a member of lat where a is tea and lat is (coffee tea or m i l k) True. Hint : Write down the definition of the function member? and refer to it while you work on the next group of questions. Is ( member? a lat) true or false where a is poached and lat is (fried eggs and scra m bled eggs) False. because the atom meat is one of the atoms of lat . This is the function member? (define member? (lambda (a lat) ( cond ( ( null ? lat) #f ) (else ( or ( eq? ( car lat) a) ( member? a ( cdr lat) ) ) ) ) ) ) What i s the value of ( member? a lat) where a is m eat and lat is ( mashed potatoes and meat gravy) #t . since a is not one of the atoms of the lat. This i s also the first question asked b y lat ?.

Do It Again. we ask the next question. Is else really a question? Yes. The answer ( or ( eq? ( car lat) a) ( member? a ( cdr lat ) ) ) does this. we have to find out whether the car of lat is the same atom as a. and Again . the value is #f.The First Commandment (preliminary) Always ask null ? as the first question in expressing any function. True or false: ( or ( eq? ( car lat) a) ( member? a ( cdr lat))) where a is meat and lat is (mashed potatoes and meat gravy) We will find out by looking at each question in turn. What is the meaning of the line (else ( or ( eq ? ( car lat) a) ( member? a ( cdr lat)))) Now that we know that lat is not null ?. . it is not null. In this case. and Again. Why is else the next question? Because we do not need to ask any more questions. or whether a is somewhere in the rest of lat. If it is. . since the atom meat was not found in lat. What is the meaning of the line ((null? lat) #f ) where lat is (mashed potatoes and meat gravy) ( null? lat) asks if lat is the null list . What is the next question? else. else is a question whose value is always true. 23 . so we ask the next question. If not . Do It.

What do we do now? Ask the next question. ) ( member? a ( cdr lat ) ) . Is a eq ? to the car of lat No. because a is meat and the car of lat is potatoes.Is ( eq ? ( car lat) a) true or false where a is meat and lat is (mashed potatoes and meat gravy) False. the car of (mashed potatoes and meat gravy) . 24 Chapter 2 . specifically ( potatoes and meat gravy). Remember The First Commandment.. What is the second question of ( or . Now what are the arguments of member? a is meat and lat is now ( cdr lat) . What is the next question? else.. Is ( null ? lat) true or false where lat is (potatoes and meat gravy) #f-false. because meat is not eq ? to mashed . What is the meaning of ( or ( eq ? ( car lat) a) ( member? a ( cdr lat))) ( or ( eq ? ( car lat) a) ( member? a ( cdr lat))) finds out if a is eq ? to the car of lat or if a is a member of the cdr of lat by referring to the function. What is the next question? ( null ? lat ) . This refers t o the function with the argument lat replaced by ( cdr lat).

What are the new arguments? a is meat. and lat is (and meat gravy) . What is the next question? else. Do It. ask the next question. What do we do now? Recur-refer to the function with new arguments. . since ( null ? lat) is false. and Again . Do It Again. and lat is (meat gravy) . what are the arguments of member? a i s meat. 25 . What is the next question? else. . What is the next question? ( null ? lat ) . Why? Because ( eq ? ( car lat) a) i s false. and Again. What is the value of ( or ( eq? ( car lat) a) ( member? a ( cdr lat))) The value of ( member? a ( cdr lat ) ) . What is the next question? ( null ? lat ) . Now.So what do we do next? We ask ( member ? a ( cdr lat ) ) . What d o we do now? Since ( null? lat) is false. What do we do now? Ask the next question .

are the same atom.What is the value of ( or ( eq ? ( car lat) a) ( �e�ber? a ( cdr lat))) #t . #t ... Move to the next line. ( null? lat ) No. #t . What is the value of the application ( �e�ber? a lat ) where a is meat and lat is ( m eat gravy) What is the value of the application ( �e�ber? a lat) where a is meat and lat is (and meat gravy) What is the value of the application ( �e�ber? a lat) where a is meat and lat is ( potatoes and meat gravy) What is the value of the application ( �e�ber? a lat ) where a is meat and lat is ( m ashed potatoes and meat gravy) Just to make sure you have it right . Of course. ( or . because meat is also a member of the lat (and meat gravy) . this is our original lat. #t. which is meat. because meat is also a member of the lat (mashed potatoes and meat gravy). Therefore. 26 Chapter 2 . which is meat . What is the value of ( �e�ber? a lat) where a is meat and lat is (mashed potatoes and meat gravy) #t . because we have found that meat is a member of ( m eat gravy). #t . ) answers with #t . because meat is also a member of the lat ( potatoes and meat gravy). because ( car lat ) . and a. let 's quickly run through it again . Hint : Write down the definition of the function �e�ber? and its arguments and refer to them as you go through the next group of questions.

Recur with a and ( cdr lat) where a is meat and ( cdr lat) is (and meat gravy) . else Yes. ( null? lat) No. Recur with a and ( cdr lat) where a is meat and ( cdr lat) is (meat gravy). ( or ( eq? ( car lat) a) ( member? a ( cdr lat))) ( eq? ( car lat) a) No. ( eq? ( car lat) a) Yes. else Yes. Perhaps. What next? Recur with a and ( cdr lat) where a is meat and ( cdr lat) is ( potatoes and meat gravy) . Ask the next question. Do It Again. the value is #t . Move to the next line. ( null ? lat) No. . but ( eq ? ( car lat) a) is false. Move to the next line. Do It. . 27 . Move to the next line. ( null? lat) No. and Again .else Yes. and Again. but ( eq ? ( car lat) a) is false.

What is the value of ( memberP a lat) where a is meat and lat is (mashed potatoes and meat gravy) #t . ( null? lat) No. What i s the value o f ( memberP a lat) where a is meat and lat is (meat gravy) #t . Move t o the next line. 28 Chapter 2 . What is the value of ( memberP a lat) where a is meat and lat is (and meat gravy) #t . What is the value of ( member? a lat) where a is meat and lat is ( potatoes and meat gravy) #t . What 's the first question member? asks? ( nu ll? lat) . Let 's work out why it is #f. What is the value of ( memberP a lat) where a is liver and lat is ( bagels and lox) #f.( or ( eq P ( car lat) a) ( memberP a ( cdr lat) ) ) #t .

but ( eq ? ( car lat) a) is false. 29 . Yes. ( null? lat) No. . else Yes. Do It. What is the value of ( or ( eq? ( car lat) a) ( member? a ( cdr lat))) where a is liver and lat is ( lox) #f.else Yes. Move to the next line. but ( eq ? ( car lat) a) is still false. ( null? lat) No. Recur with a and ( cdr lat) where a is liver and ( cdr lat) is (lox) . . Recur with a and ( cdr lat) where a is l iver and ( cdr lat) is () . Do It Again. but ( eq ? ( car lat) a) is false. Move to the next line. else Yes. and Again. Recur with a and ( cdr lat) where a is l iver and ( cdr lat) is (and lox) . ( null? lat) What is the value of ( member? a lat) where a is liver and lat is () #f. and Again .

30 Chapter 2 . What is the value of ( or ( eq ? ( car lat) a) ( member? a ( cdr lat))) where a is l iver and lat is (bagels and lox) #f . What is the value of ( or ( eq ? ( car lat) a) ( member? a ( cdr lat))) where a is l iver and lat is (and lox) #f .What is the value of ( member? a lat) where a is liver and lat is (lox) #f. What is the value of ( member? a lat) where a is l iver and lat is (and lox) #f. What is the value of ( member? a lat) where a is l iver and lat is (bagels and lox) #f .

31 . Do It Again. and Again .Do you believe all this? Then you may rest! This space for doodling Do It. . . and Again.

.

Return () . Is there any other question we should ask about the lat? Cons the Magnificent 33 . No.What is ( rember a lat) where a is mint and lat is (lamb chops and mint jelly) (lamb chops and jelly) "Rember" stands for remove a mem ber. What steps should we use to do this? First we will test ( null? lat)-The First Commandment . Either a lat is empty or it contains at least one atom. And if ( null? lat) is true? What do we know if ( null? lat) is not true? We know that there must be at least one atom in the lat . What does ( rember a lat) do? It takes an atom and a lat as its arguments. ( rember a lat) where a is mint and lat is (lamb chops and mint flavored mint jelly) (lamb chops and flavored mint jelly) . ( rember a lat) where a is toast and lat is (bacon lettuce and tomato) (bacon lettuce and tomato). ( rember a lat) where a is cup and lat is (coffee cup tea cup and hick cup) (coffee tea cup and hick cu p). and makes a new lat with the first occurrence of the atom in the old lat removed.

Now.)) . Hint : Write down the function rember and its arguments.) ( --.What do we do if we know that the lat contains at least one atom? How do we ask questions? We ask whether a is equal to ( car lat ). What do we do if a is not the same as ( car lat ) We want to keep ( car lat). let 's write down what we have so far: (define rember (lambda (a lat) (cond ( ( null? lat ) (quote ( ))) (else ( c ond ( ( eq ? ( car lat) a) ( cdr lat)) (else ( rember a ( cdr lat)))))))) What is the value of ( rember a lat ) where a is bacon and lat is ( bacon lettuce and tomato) ( l ettuce and tomato). What would be the value of ( rember a lat) if a were the same as ( car lat) ( cdr lat). No. and refer to them as you go through the next sequence of questions. _ __ How do we ask if a is the same as ( car lat) ( eq ? ( car lat ) a). but also find out if a is somewhere in the rest of the lat . By using ( c ond ( --.--. How d o we remove the first occurrence of a in the rest of lat Is there any other question we should ask? ( rember a ( cdr lat)). 34 Chapter 3 .

What is the value of ( rember a lat) where a is and and lat is (bacon lettuce and tomato) ( bacon lettuce tomato). because it is the original list without the atom bacon. What is the first question? ( null ? lat) . so the value is ( cdr lat) . Cons the Magnificent 35 . let 's see if this function works. But did we really use a good example? Who knows? But the proof of the pudding is in the eating. What next? Ask the next question. ( eq? ( car lat) a) Yes. What do we do now? We compare each atom of the lat with the atom a. it is the list (lettuce and tomato) . What do we do now? Move t o the next line and ask the next question. In this case. What does rember do? It takes an atom and a lat as its arguments.Now. and makes a new lat with the first occurrence of the atom in the old lat removed. so let 's try another example. Is this the correct value? Yes. else Yes. and if the comparison fails we build a list that begins with the atom we just compared.

else Sure. and ask the next question. so ask the next question. What is the first question asked by rember ( null? lat). and ask the next question. else ( eq? ( car lat) a) Okay. ( null ? lat) No. What do we do now? Move to the next line. else Of course. ( eq ? ( car lat) a) What is the meaning of ( rember a ( cdr lat) ) Recur where a is and and ( cdr lat) is (and tomato) . What is the meaning of (else ( rember a ( cdr lat))) else asks if else is true-as it always is-and the rest of the line says to recur with a and ( cdr lat) . No. No. where a is and and ( cdr lat) is (lettuce and tomato) . 36 Chapter 3 .Let us see if our function rember works. so move to the next line. so move to the next line. so move to the next line. ( null ? lat) No. so move to the next line.

since (tomato) is not the list (bacon lettuce and tomato) with j ust a-and-removed. but we also lost all the atoms preceding and. from chapter 1 ? The Second Commandment Use cons to build lists. How can we keep from losing the atoms bacon and lettuce We use Cons the Magnificent . What did we do wrong? We dropped and . Cons the Magnificent 37 . So what is the result? ( cdr l at)-(tomato) . Let 's see what happens when we use cons (define rember (lambda (a lat) ( cond ( ( n·ull? lat) (quote ())) (else (cond (( eq ? ( car lat) a) ( cdr lat) ) (else ( cons ( car lat) ( rember a ( cdr lat) ) ) ) ) ) ) ) ) What i s the value o f ( rember a lat) where a is and and lat is (bacon lettuce and tomato) ( bacon lettuce tomato) . Is this correct? No.( eq? ( car lat) a) Yes. Remember cons . Hint : Make a copy of this function with cons and the arguments a and lat so you can refer to it for the following questions.

so move to the next line. else Yes. ( eq ? ( car lat) a) No. we must find it before we can cons ( car lat) onto it. B u t since we don 't know the value of ( rember a ( cdr lat) ) yet . What is the meaning of ( rember a ( cdr lat) ) This refers to the function with lat replaced by ( cdr lat) ( lettuce and tomato) . of course. ( eq ? ( car lat) a) No. we must find it before we can cons ( car lat) onto it . - ( null ? lat) No. What do we do now? Ask the next question. ask the next question. else Yes. so move to the next line. 38 Chapter 3 . so move to the next line.What is the first question? ( null? lat ) . B u t since we don 't know the value of ( rember a ( cdr lat)) yet . What is the meaning of ( cons ( car lat) ( rember a ( cdr lat))) It says to cons the car of lat-lettuce-onto the value of ( rember a ( cdr lat ) ) . What is the meaning of ( cons ( car lat) ( rember a ( cdr lat))) where a is and and lat is ( bacon lettuce and tomato) It says to cons the car of lat-bacon-onto the value of ( rember a ( cdr lat ) ) .

(and tomato). we can cons lettuce onto it . Certainly not ! We know what ( rember a lat ) is when lat is (and tomato) . ( null? lat) else ( eq? ( car lat) a) What is the value of the line ( ( eq? ( car lat) a) ( cdr lat)) Are we finished? ( cdr lat) - ( tomato ) . Cons the Magnificent 39 . No. when lat is (lettuce and tomato) and ( rember a ( cdr lat )) is (tomato) . Now that we have this value. We now have a value for ( rember a ( cdr lat) ) where a i s and and ( cdr lat) is (and tomato) This value is (tomato) What next? What is the result when we cons lettuce onto (tomato) What does (lettuce tomato) represent? (lettuce tomato). Still. Recall that we wanted to cons lettuce onto the value of ( rember a ( cdr lat )) where a was and and ( cdr lat ) was (and tomato) .What is the meaning of ( rember a ( cdr lat )) This refers to the function with lat replaced by ( cdr lat ) . It represents the value of ( cons ( car lat ) ( rember a ( cdr lat ) ) ) . that is. Yes. but we don't yet know what it is when lat is (lettuce and tomato) or ( bacon lettuce and tomato). which is (tomato) . so ask the next question.

we can cons bacon onto it. Lunch? Are we finished yet? Yes." 40 Chapter 3 . we saved it to be consed to the final value later. which is (lettuce tomato) . Now that we have this value. one at a time. Can you put in your own words how we determined the final value ( bacon lettuce tomato) In our words: "The function rember checked each atom of the lat . and consed the previous atoms back onto the rest of the lat. we wanted to cons bacon onto the value of ( rember a ( cdr lat)). When rember found the atom and. but we don't yet know what it is when lat is (bacon lettuce and tomato) . so what must we do again? Recall that . ( bacon lettuce tomato) . at one time. where a was and and ( cdr lat) was (lettuce and tomato) . What is the result when we cons bacon onto ( l ettuce tomato) What does ( bacon lettuce tomato) represent? t t It represents the value of ( cons ( car lat) ( rember a ( cdr lat))).Are we finished yet? Not quite. it dropped it. So far we know what ( rember a lat) is when lat is (lettuce and tomato). to see if it was the same as the atom and. when lat is ( bacon lettuce and tomato) and ( rember a ( cdr lat)) is (lettuce tomato) . We now have a value for ( rember a ( cdr lat)) where a is and and ( cdr lat) is ( l ettuce and tomato) This value is (lettuce tomato) This is not the final value. If the car was not the same as the atom .

we can simplify it. ( eq? ( car lat) a) No. Hint : Write down the function rember and its arguments and refer to them as you go through the next sequence of questions. so the value is ( cons ( car lat) ( rember a ( cdr lat ) ) ) . What is the meaning of ( cons ( car lat) ( rember a ( cdr lat))) This says t o refer t o the function rember but with the argument lat replaced by ( cdr lat) . - Cons the Magnificent 41 . and that after we arrive at a value for ( rember a ( cdr lat) ) we must cons ( car lat ) bacon-onto it. Let's see if the new rember is the same as the old one. What is the value of the application ( rember a lat) where a is and and lat is (bacon lettuce and tomato) (bacon lettuce tomato) .Can you rewrite rember so that it reflects the above description? Yes. So why don't we simplify right away? Because then a function 's structure does not coincide with its argument 's structure. else Yes. (defi ne rember (lambda ( a lat) ( c o nd ( ( null ? lat) (quote ( ) ) ) ( ( eq ? ( car lat) a) ( cdr lat ) ) (els e ( cons ( car lat) ( rember a ( cdr lat ) ) ) ) ) ) ) Do you think this is simpler? Functions like rember can always be simplified in this manner. ( null? lat) No.

and that after we arrive at a value for ( rember a ( cdr lat) ) . -- baco n-onto Now that we have completed rember try this example: ( rember a lat) where a is sauce and lat is (soy sa uce and tomato sauce) ( rember a lat) is (soy and tomato sauce) . 42 Chapter 3 . Now what? Now cons ( car lat ) (lettuce tomato) . Now what? Now cons ( car lat)-lettuce-onto (tomato) . ( null ? lat) No. so the value is ( cons ( car lat) ( rember a ( cdr lat))). else Yes. ( eq ? ( car lat ) a) No.( null ? lat) No. with the argument lat replaced by ( cdr lat) . we must cons ( car lat )-lettuce-onto it. What is the meaning of ( cons ( car lat) ( rember a ( cdr lat ))) This says we recur using the function rember. What is the value of the line ( ( eq ? ( car lat) a) ( cdr lat) ) It is ( cdr lat)-(tomato) . ( eq ? ( car lat) a) Yes.

What is (firsts l ) where l is ( ( (five plu ms) four) (eleven green oranges) ((no) more)) ( (five plu ms) eleven (no) ) . What is (firsts l ) where l is ((a b) (c d) (e f) ) (a c e ) . which is either a null list or contains only non-empty lists. It builds another list composed of the first S-expression of each internal list . what does (firsts l ) do? We tried the following: "The function firsts takes one argument . (five four eleven) . a list.What is (firsts l ) where l is ( (apple peach pumpkin) (plum pear cherry) (grape raisi n pea ) (bean carrot eggplant)) (apple p l u m gra pe bea n ) . In your own words. What is (firsts l ) where l is () What i s (firsts l ) where l is ( (five plu ms) (four) (eleven green oranges)) ()." Cons the Magnificent 43 .

. Because we can only look at one S-expression at a time. Why ( cons Because we are building a list-The Second Commandment . 44 Chapter 3 . or it contains at least one non-empty list. we must recur. ) Because we need to ask questions about the actual arguments. )) Because we always state the function name. and they always appear at the end of a function definition. . . . and (define. To look at the rest. (lambda. Why (firsts ( cdr l)) Why ) ) ) Because these are the matching parentheses for ( cond. ) (else ( cons . . Why (else Because we only have two questions to ask about the list l: either it is the null list . . And because the last question is always else. ) The First Commandment. Why ( ( null ? l) .See if you can write the function firsts Remember the Commandments! This much is easy: (define firsts (lambda (l) (cond ( ( null? l) . . (firsts ( cdr l))))))) Why (define firsts (lambda (l) . (lambda. . . . and the argument ( s ) of the function . Why (else See above. Why (cond .

d .Keeping in mind the definition of (firsts l) what is a typical element of the value of (firsts l) where l is ((a b) ( c d) ( e f) ) a. and then cons it onto the natural recursion. See chapter 1 . The Third Commandment When building a list. ------._______. o r even e. With The Third Commandment . we can now fill in more of the function firsts What does the last line look like now? (else ( cons ( car ( car l)) (firsts ( cdr l)) ) ) . How do we describe a typical element for (firsts l) As the car of an element of l-( car ( car l ) ) .. When we find a typical element of (firsts l) what do we do with it? cons i t onto the recursion-(firsts ( cdr l ) ) . Consider the function seconds What would be a typical element of the value of ( seconds l) where l is (( a b) ( c d) ( e f) ) b. or f. describe the first typical ele­ ment. typical element natural recursion Cons the Magnificent 45 . What is another typical element? c.

so move to the next line. . ) (else ( cons ( car ( car l)) (firsts ( cdr l ) ) ) ) ) ) ) where l is ( ( a b) ( c d ) ( e f) ) Nothing yet . What is the meaning of ( cons ( car ( car l)) (firsts ( cdr l))) Save ( car ( car l)) . and recur with (firsts ( cdr l)). What is the meaning of ( cons ( car ( car l)) (firsts ( cdr l))) It saves ( car ( car l ) ) to cons onto (firsts ( cdr l ) ) .What does (firsts l) do (define firsts (lambda (l) (cond ( ( null? l) . and recur with (firsts ( cdr l)). 46 Chapter 3 . ( null ? l) where l is ( ( c d) ( e f) ) No. What is the meaning of ( cons ( car ( car l)) (firsts ( cdr l))) Save ( car ( car l ) ) . . however. Now. We can. what is the value of the line ( ( null ? l) . To find (firsts ( cdr l ) ) . so move t o the next line. ( null? l) where l is ( ( e f) ) No . ( null ? l) where l is ( ( a b) ( c d) ( e f) ) No. something is missing. The first line ( ( nu ll ? l ) . We are still missing one important ingredient in our recipe. proceed without it for now. . ) needs a value for the case where l is the null list . . we refer to the function with the new argument ( cdr l ) . ( null ? l) Yes. ) There is no value. . so move to the next line. .

what is the value of (firsts l) With which of the three alternatives do you feel most comfortable? What is ( insertR new old lat) where new is topping old is fudge and lat is (ice cream with fudge for dessert) ( insertR new old lat) where new is jalapeno old is and and lat is (tacos tamales and sa lsa) Correct ! Now you should use that one. or 1 . We need to: I. cons c onto the value of 3 3. (tacos tama les a n d jalapeno sa lsa ) . Cons the Magnificent 47 . cons a onto the value of 2 2. we cannot use #t or #f. Remember The Law of Cons. cons c onto the value of 1 3. For the purpose of cons ing. Since the final value must be a list . either 1. cons e onto () 2. (a c e) . (ice crea m with fudge topping for dessert ) . what value can we give when ( null? l) is true? With () as a value. cons e onto () Ill or cons a onto the cons of c onto the cons of e onto () In any case. Let 's try (quote ( ) ) .What do we need to cons atoms onto? A list . cons a onto the value of 2 II. we now have three cons steps to go back and pick up.

) ) ) Which argument changes when we recur with insertR How many questions can we ask about the lat? lat . Two.( insertR new old lat) where new is e old is d and lat is (a b c d f g d h ) (a b c d e f g d h). because we can only look a t one of its atoms at a time. because else is always the last question. Which questions do we ask? First . The function insertR builds a lat with new inserted to the right of the first occurrence of old . . Second. A lat is either the null list or a non-empty list of atoms. " See if you can write the first three lines of the function insertR (define insertR (lambda ( new old lat) (cond . what does ( insertR new old lat) do? I n our words: "It takes three arguments: the atoms new and old . 48 Chapter 3 . and a lat . What do we know if ( null? lat) is not true? We know that lat has at least one element . I n your own words. we ask ( eq ? ( car lat) old ) . Which questions do we ask about the first element? First . we ask else. Then we ask else. we ask ( null? lat) . because there are no other interesting cases. .

w e want Let 's try con s ing new onto ( cdr lat) .Now see if you can write the whole function inserlR Here i s our first attempt . When ( car lat) i s the same to insert new to the right . Yes. (define inserlR (lambda ( new old lat) (cond < --­ (else (cond < < --­ --__ --- ) )))))) (define inserlR (lambda ( new old lat) (cond ( ( null? lat) (quote ( ) ) ) (else (cond ( ( eq ? ( car lat) old ) ( cdr lat) ) (else ( cons ( car lat) ( inserlR new old ( cdr lat) ) ) ) ) ) ) ) ) What i s the value of the application ( inserlR new old lat) that we just determined where new is topping old is fudge and lat is (ice crea m with fudge for dessert) So far this is the same as rember What do we do in inserlR when ( eq? ( car lat) old ) is true? How is this done? Now we have (ice crea m with for dessert ) . as old. (define inserlR (lambda ( new old lat) (cond ( ( null? lat) (quote ( ) ) ) (else (cond ( ( eq? ( car lat) old ) ( cons new ( cdr lat))) (else ( cons ( car lat) ( inserlR new old ( cdr lat ) ) ) ) ) ) ) ) ) Cons the Magnificent 49 .

we have only replaced fudge with topping. How can we include old before new Now let 's write the rest of the function insertR (define insertR (lambda ( new old lat) (cond ( ( null ? lat) (quote ( ) ) ) (else (cond ( ( eq? ( car lat) old) ( cons old ( cons new ( cdr lat)))) (else ( cons ( car lat) ( insertR new old ( cdr lat))))))))) When new is topping. and lat is (ice crea m with fudge for dessert ) . the value of the application. 50 Chapter 3 . ( insertR new old lat ) . If you got this right . Try consing old onto ( cons new ( cdr lat)). What still needs to be done? Somehow we need to include the atom that is the same as old before the atom new.So what is ( insertR new old lat) now where new is topping old is fudge and lat is (ice crea m with fudge for dessert) (ice crea m with topping for dessert). is (ice cream with fudge topping for dessert). Is this the list we wanted? No. have one. old is fudge.

(define subst (lambda ( new old lat) ( cond ( ( null? lat) ( quote ( ) ) ) ( else ( cond ( ( eq ? ( car lat) old ) ( cons new ( cdr lat ) ) ) ( else ( cons ( car lat) ( subst new old ( cdr lat ) ) ) ) ) ) ) ) ) This i s the same as one of our incorrect attempts at insertR. Now try subst Hint: (subst new old lat) replaces the first occurrence of old in the lat with new For example. right? (define insertL (lambda ( new old lat) (cond ( ( null ? lat) (quote ( ) ) ) (else (cond ( ( eq ? ( car lat) old ) ( cons new ( cons old ( cdr lat ) ) ) ) (else ( cons ( car lat) ( insertL new old ( cdr lat ) ) ) ) ) ) ) ) ) Did you think of a different way to do it? For example. Obviously. ( ( eq ? ( car lat) old ) ( cons new ( cons old ( cdr lat ) ) ) ) could have been ( ( eq ? ( car lat) old ) ( cons new lat) ) since ( cons old ( cdr lat) ) where old i s eq ? to ( car lat) is the same as lat.Now try insertL Hint: insertL inserts the atom new to the left of the first occurrence of the atom old in lat This much is easy. Cons the Magnificent 51 . where new is topping old is fudge and lat is (ice crea m with fudge for dessert) the value is (ice crea m with topping for dessert) Now you have the idea.

or the rest of the lat . If it is not. go and repeat the cake-consing. rember saves the atom and proceeds. so that the value returned is the original lat . Do you recall what rember does? The function rember looks at each atom of a lat to see if it is the same as the atom a. where new is vanilla o1 is chocolate o2 is banana and lat is (banana ice cream with chocolate topping) the value is (vanilla ice cream with chocolate topping) (define subst2 (lambda ( new o1 o2 lat) (cond ( ( num' lat) (quote ())) (else (cond ( ( eq � ( car lat) o1 ) ( cons new ( cdr lat))) ( ( eq � ( car lat) o2) ( cons new ( cdr lat))) (else ( cons ( car lat) ( subst2 new o1 o2 ( cdr lat) )))))))) Did you think of a better way? Replace the two eq � lines about the ( car lat) by ( (or ( eq � ( car lat) o1 ) ( eq� ( car lat) o2 )) ( cons new ( cdr lat))).Go cons a piece of cake onto your mouth. with only that occurrence of a removed. When it finds the first occurrence of a. Now try subst2 Hint: (subst2 new o1 o2 lat) replaces either the first occurrence of o1 or the first occurrence of o2 by new For example. I f you got the last function. it stops and gives the value ( cdr lat) . 52 Chapter 3 .

(define multirember (lambda ( a lat) (cond ( (else (cond ( ( _ __ --___ ___ -- ) )))))) Hint : What d o we want as the value when ( eq? ( car lat) a) is true? Consider the example where a is cup and lat is (coffee cup tea cup and hick cup) Can you see how multirember works? Possibly not . Now determine ( multirember a ( cdr lat ) ) . The value of the application is (coffee tea and hick) . so move to the next line. so move to the next line. so we will go through the steps necessary to arrive at the value (coffee tea and hick) . What is the meaning of ( cons ( car lat) ( multirember a ( cdr lat))) Save ( car lat) coffee to be consed onto the value of ( multirember a ( cdr lat) ) later. w e now recur with ( multirember a ( cdr lat ) ) . - ( null? lat) No. Cons the Magnificen t 53 . ( null ? lat) No. ( define multirember (lambda ( a lat) ( cond ( ( null? lat) ( quote ())) ( else ( cond ( ( eq ? ( car lat) a) ( multirember a ( cdr lat ))) ( else ( cons ( car lat) ( multirember a ( cdr lat ) ) ) ) ) ) ) ) ) After the first occurrence o f a .Write the function multirember which gives as its final value the lat with all occurrences of a removed. else Yes. ( eq? ( car lat) a) No. in order to remove the other occurrences. s o move t o the next line.

and determine ( multirember a ( cdr lat)). so move to the next line. ( eq? ( car lat) a) No. else Yes! No. Now determine ( multirember a ( cdr lat)). ( null ? lat) No. so move t o the next line. so move to the next line. What is the meaning of ( cons ( car lat) ( multirember a ( cdr lat))) Save ( car lat) and to be consed onto the value of ( multirember a ( cdr lat) ) later. so forget ( car lat ) . so forget ( car lat).else Naturally. ( eq ? ( car lat) a) Yes. move on. - ( null ? lat) No. else Okay. so move to the next line. ( null ? lat) No. so move to the next line. - 54 Chapter 3 . and determine ( multirember a ( cdr lat)). ( eq ? ( car lat) a) Yes. ( eq ? ( car lat) a) What is the meaning of ( cons ( car lat) ( multirember a ( cdr lat ) ) ) Save ( car lat) tea to be consed onto the value of ( multirember a ( cdr lat)) later. Now detennine ( multirember a ( cdr lat) ) .

s o move t o the next line. Are we finished? No. What do we do next? What do we do next? Are we finished now? Yes . We cons tea onto (and hick ) . Cons the Magnificent 55 . Now determine ( multirember a ( cdr lat ) ) . What do we do next? We cons and onto ( hick ) . and determine ( multirember a ( cdr lat ) ) . s o the value i s () . ( null? lat) Yes. we still have several conses to pick up. so move to the next line. ( eq? ( car lat) a) Yes . so move to the next line. We cons coffee onto (tea and hick ) . What is the meaning of ( cons ( car lat) ( multirember a ( cdr lat))) Save ( car lat) hick to be consed onto the value of ( multirember a ( cdr lat) ) later. so forget ( car lat) . ( eq? ( car lat) a) No.( null? lat) No. - ( null ? lat) No. What do we do next? We cons the most recent ( car lat) we have-hick--onto () .

Was the terminal condition ever reached? No. because we never get past the first occurrence of old. I s this function defined correctly? ( define multiinsertL (lambda ( new old lat) ( cond ( ( null ? lat) ( quote ())) ( else ( cond ( ( eq ? ( car lat) old ) ( cons new ( cons old ( multiinsertL new old lat)))) ( else ( cons ( car lat) ( multiinsertL new old ( cdr lat ) ) ) ) ) ) ) ) ) Not quite. go through ( multiinsertL new old lat) where new is fried old is fish and lat is (chips and fish or fish and fried) .Now write the function multiinsertR ( define multiinsertR (define multiinsertR (lambda ( new old lat) (cond (else (cond ( ( ( --­ ) )))))) (lambda ( new old lat) (cond ( ( null? lat) (quote ( ))) (else ( cond ( ( eq ? ( car lat) old) ( cons ( car lat) ( cons new ( multiinsertR new old ( cdr lat) ) ) ) ) ( else ( cons ( car lat) ( multiinsertR new old ( cdr lat) )))))))) It would also be correct to use old in place of ( car lat) because we know that ( eq ? ( car lat) old ) . To find out why. 56 Chapter 3 .

test termination with null?.Now. The changing argument must be tested in the termination condition: when using cdr. Now write the function multisubst ( define multisubst (lambda ( new old lat) ( cond ( define multisubst {lambda ( new old lat) ( cond ( ( null? lat) ( quote ())) ( else ( cond ( ( eq? ( car lat) old} ( cons new ( multisubst new old ( cdr lat)))) ( else ( cons ( car lat) ( multisubst new old ( cdr lat))))))))) < --( else ( cond < --­ --. It must be changed to be closer to termination.__ ) ) ) ) ) ) Cons the Magnificent 57 .) < --. try to write the function multiinsertL again: (define multiinsertL (lambda ( new old lat) (cond < --­ (else (cond ( ( ) )))))) (define multiinsertL (lambda ( new old lat) (cond ( ( null? lat) (quote ( ))) ( else ( cond (( eq ? ( car lat) old} ( cons new ( cons old ( multiinsertL new old ( cdr lat))))) ( else ( cons ( car lat) ( multiinsertL new old ( cdr lat))))))))) The Fourt h Commandment (preliminary) Always change at least one argument while recurring.

-9�--� .

True. Are -3 and 3.Is 14 an atom? Is ( atom ? n ) true or false where n is 14 Is -3 a number? Yes. 1 L: 1S: (define aub1 Cl1111bda (n} (. because we don't need to say "where 67" when the argument is a number. 14159 numbers? Yes. but we consider only whole numbers. because all numbers are atoms.n 1 } } } } Numbers Games 59 . because 14 is an atom. 14159 a number? Yes. .e. What is ( a dd 1 1 where n is 67 n ) 1 L: 1+ S: (define add1 Cl1111bda (n} (+ n 1 } } } } What i s ( add1 67 ) Also 68. 1. Is 3. . 2. ) . . 3. 68. but we do not consider negative numbers. 0. . Yes. 4. but the only numbers we use are the nonnegative integers ( i. n is What is (sub1 1 where n is 5 n ) 4.

What is (sub1 0}

( subl n ) , where n i s 0 , h as n o answer because we consider only nonnegative numbers. In practice, this result is -1.

1

Is (zero ?1 0} true or false?

True.

1

L:

zerop

Is (zero ? 1492} true or false?

False.

What is ( + 46 12}

58.

Try to write the function + Hint : It uses zero ? add1 1 and sub1 1

(define + 1 (lambda ( n m ) (cond ((zero ? m ) n) (else ( add1 ( + n (sub1
Wasn't that easy?

m

)))))))

1 Remember to use our definitions for add1 and eub 1 .

1

L, S:

This is like +. Write it as o+ (see preface) .

But didn 't we just violate The First Commandment?

Yes, but we can treat zero ? like null? since zero ? asks if a number is empty and null? asks if a list is empty.

If zero P is like null? is add1 like cons

Yes! cons builds lists and add1 builds numbers.

60

Chapter 4

What is ( - 14 3)

11.

What is ( - 17 9)

8.

What is ( - 18 25)

No answer. There are no negative numbers.

Try to write the function Hint: Use sub1

(define - 1 (lambda ( n m ) (cond ( ( zero ? m ) n ) (else (sub1 ( -

n

(sub1

m

)))))))

1

L, S :

This i s like - . Write i t

as o -

(see preface) .

Can you describe how (-

n m

) works?

I t takes two numbers as arguments, and reduces the second until it hits zero. It subtracts one from the result as many times as it did to cause the second one to reach zero. Yes: tup is short for tuple.

Is this a tup? (2 11 3 79 47 6)

Is this a tup? (8 55 5 555)

Yes, of course, it is also a list of numbers.

Is this a tup? (1 2 8 apple 4 3)

No, it is just a list of atoms.

Is this a tup? (3 (7 4) 13 9)

No, because it is not a list of numbers. (7 4) is not a number.

Numbers Games

61

Is this a tup? ()

Yes, it is a list of zero numbers. This special case is the empty tup.

What is ( addtup tup) where tup is (3 5 2 8)

18.

What is ( addtup tup) where tup is (15 6 7 12 3)

43.

It builds a number by totaling all the numbers in its argument.

What is the natural way to build numbers from a list?

Use + in place of cons : + builds numbers in the same way as cons builds lists.

When building lists with cons the value of the terminal condition is () What should be the value of the terminal condition when building numbers with +

0.

What is the natural terminal condition for a list? What is the natural terminal condition for a tup?

( null? l).

( null? tup).

When we build a number from a list of numbers, what should the terminal condition line look like? What is the terminal condition line of addtup

( ( null ? tup) 0) , just as ( ( null? l) (quote ())) is often the terminal condition line for lists.

( ( null? tup) 0).

62

Chapter 4

How is a lat defined?

It is either an empty list , or it contains an atom , ( car lat) , and a rest, ( cdr lat) , that is also a lat .

How is a tup defined?

It is either an empty list , or it contains a number and a rest that is also a tup.

What is used in the natural recursion on a list?

( cdr lat ) .

What is used in the natural recursion on a tup?

( cdr tup ) .

Why?

Because the rest o f a non-empty list i s a list and the rest of a non-empty tup is a tup.

Two.

Two, because it is either empty or it is a number and a rest, which is again a tup.

How is a number defined?

It is either zero or it is one added to a rest, where rest is again a number.

What is the natural terminal condition for numbers?

(zero ? n ) .

What is the natural recursion on a number?

( sub1

n .

)

Two.

Numbers Games

63

The Fourt h Commandment (first revision) Always change at least one argument while recurring. n up m What is the terminal condition line for x ((zero P m ) 0) . What is another name for ( x this case? n ( sub1 m )) in It's the natural recursion for x.What does ( x n m ) do? It builds up a number by adding times. test termination with zero ?. Since (zero P m ) is the terminal condition. It must be changed to be closer to termination. Try to write the function x (define x 1 (lambda ( n m ) (cond ((zero P m ) 0) (else ( + n ( x n (sub1 m ))))))) 1 L. because n x 0 = 0. What function is used to do this? sub1 . m must eventually be reduced to zero. Numbers Games 65 . The changing argument must be tested in the termination condition: when using cdr. test termination with null? and when using sub1 . S: This i s like • .

but let 's follow through the function one time to see how we get this value. because (zero ? m ) is now true.What is ( x 12 3) 36. (zero ? m ) No. (zero ? m ) No. If x is correct then ( x 12 ( sub1 3)) should be 24. What are the new arguments of ( x n m) n is 12. and m is 1 . What is the meaning of (+ n ( x n ( sub1 m ))) It adds n ( where n = 12) to ( x n (sub1 m )) . What is the meaning of ( + n ( x n (sub1 m ))) It adds n ( where n 12) to the natural recursion. 66 Chapter 4 . (zero ? m ) No. What i s the value of the line ( ( zero ? m ) 0) 0. = What are the new arguments of ( x n m) n is 12. Are we finished yet? No. and m is 2 . What is the meaning of (+ n ( x n ( sub1 m ))) It adds n ( where n = 12) to ( x n (sub1 m )) .

always use 0 for the value of the terminating line. This technique works for all recursive functions. 12 3) Because 0 will not affect + . for multiplying by 1 does not change the value of a multiplication. for adding 0 does not change the value of an addition. where n is 12 and m is 3. You can use this approach to write functions as well as to argue their correctness. Numbers Games 67 . (x Again. not just those that use numbers. Notice that n has been +ed m times.Why not? Because we still have three +es to pick up. When building a value with x . that x is the conventional multiplication of nonnegative integers. using equations. always consider () for the value What is ( tup+ tup1 tup2) where tup1 is (3 6 9 1 1 4 ) and tup2 is (8 5 2 0 7) (11 11 11 11 11). What is the value of the original application? Argue. That is. always use 1 for the value of the terminating line. which is as we expected. n + O = n. = 12 + ( x 12 2) = 12 + 12 + ( x 12 1) = 12 + 12 + 12 + ( X 12 0) = 12 + 12 + 12 + 0. why is 0 the value for the terminal condition line in x The Fifth Commandment When building a value with + . Add 12 to 12 to 12 to 0 yielding 36. When building a value with of the terminating line. cons .

What is (tup+ tup1 tup2 ) where tup1 is (2 3) and tup2 is (4 6) (6 9). and if the second tup is empty or non-empty. because the tups must have the same length. 68 Cbapter 4 . building a tup of the answers. If you recur on one tup how many questions do you have to ask? Two. because ( null'? tup1 ) is true exactly when ( null'? tup2 ) is true. What does ( tup+ tup1 tup2 ) do? It adds the first number of tup1 to the first number of tup2. for tups of the same length. they are ( null? tup) and else. What is unusual about tup+ It looks at each element of two tups at the same time. it recurs on two tups. and so on. No. then it adds the second number of tup1 to the second number of tup2. or in other words. how many questions need to be asked about the tups? Four: if the first tup is empty or non-empty. Does this mean (and ( null'? tup1 ) ( null '? tup2 ) ) and else are the only questions we need to ask? Yes. Do you mean the questions (and ( null'? tup1 ) ( null'? tup2 ) ) ( null'? tup1 ) ( null'? tup2) and else Can the first tup be () at the same time as the second is other than () Yes. When recurring on two tups.

( + ( car tup1 ) ( car tup2 )) and ( tup+ ( cdr tup1 ) ( cdr tup2 ) ) . No. (7 13).Write the function tup+ (define tup+ (lambda ( tup1 tup2) (cond ((and ( null? tup1 ) ( null? tup2)) (quote ())) (else ( cons ( + ( car tup1 ) ( car tup2 )) ( tup+ ( cdr tup1 ) ( cdr tup2 ) ) ) ) ) ) ) What are the arguments of + in the last line? What are the arguments of cons in the last line? ( car tup1 ) and ( car tup2 ) . so now we are ready to consider the rest of both tups. But let 's see how it works. ( null? tup1 ) where tup1 is now (7) and tup2 is now (6) Numbers Games 69 . What is ( tup+ tup1 tup2) where tup1 is (3 7) and tup2 is (4 6) ( null? tup1 ) ( cons ( + ( car tup1 ) ( car tup2 )) ( tup+ ( cdr tup1 ) ( cdr tup2 ))) Why does the natural recursion include the cdr of both arguments? No. Because the typical element of the final value uses the car of both tups. cons 7 onto the natural recursion: ( tup+ ( cdr tup1 ) ( cdr tup2 ) ) .

since tup1 will become null before tup2. the cons of 7 onto the cons of 13 onto ( ) . What does the second new line look like? ( ( null ? tup2 ) tup1 ) . What is the value of the application? (7 13). That is. what must be the value? ( ) . since tup2 will become null before tup1. See The First Commandment: We did not ask all the necessary questions! But.( cons ( + ( car tup1 ) ( car tup2 )) ( tup+ ( cdr tup1 ) ( cdr tup2 ))) cons 13 onto the natural recursion. because ( null ? tup2) must be true. ( null ? tup1 ) Yes. 70 Chapter 4 . What problem arises when we want ( tup+ tup1 tup2 ) where tup1 is (3 7) and tup2 is ( 4 6 8 1 ) Can w e still write tup+ even i f the tups are not the same length? Yes! What new terminal condition line can we add to get the correct final value? Add ( ( null? tup1 ) tup2 ) . See The First Commandment: We did not ask all the necessary questions! What do we need to include in our function? We need to ask two more questions: ( null? tup1 ) and ( null? tup2 ) . Then . we would like the final value to be (7 13 8 1 ) . No answer. What is ( tup+ tup1 tup2) where tup1 is (3 7 8 1 ) and tup2 is (4 6) No answer.

Yes. How many questions do we have to ask about n and m Three: (zero ? n). Numbers Games 71 . because either ( null ? tup1 ) or ( null ? tup2 ) is true if either one of them does not contain at least one number. ) When do we recur? When we know neither number is equal to 0.( car tup1 ) ( car tup2 ) ) ( tup+ ( cdr tup 1 ) ( cdr tup2 ) ) ) ) ) ) ) Can you simplify it? (define tup+ (lambda ( tup1 tup2 ) (cond ( ( null? tup1 ) tup2 ) ( ( null? tup2 ) tup1 ) ( else ( cons ( + ( car tup1 ) ( car tup2 )) ( tup+ ( cdr tup1 ) ( cdr tup2 ) ) ) ) ) ) ) Does the order of the two terminal conditions matter? Is else the last question? No. What is (> 12 133) What is ( > 120 1 1 ) #t -true. n and m. How do we recur? With (sub1 n) and ( sub1 m .Here is a definition of tup+ that works for any two tups: (define tup+ (lambda ( tup1 tup2 ) (cond ((and ( null? tup1 ) ( null? tup2 ) ) (quote ())) ((null? tup1 ) tup2 ) ( ( null? tup2 ) tup1 ) (else ( cons ( -11. ) and else. (zero ? m . #f-false. O n how many numbers d o we have t o recur? Two.

Can you write the function > now using zero ? and sub1 How about (define > (lambda ( n m ) ( cond ((zero ? m ) #t ) ((zero ? n ) #f ) (else ( > (sub1 n ) (sub1 m )))))) Is the way we wrote ( > n m ) correct? No. What is the meaning of ( > (sub1 n ) (sub1 m )) Recur. so move to the next question. No. but with both arguments closer to zero by one. so move to the next question. so move to the next question. so move to the next question. but with both arguments reduced by one. (zero ? 2) No. What is the meaning of ( > (sub1 n ) (sub1 m )) Recur. but with both arguments reduced by one. (zero ? 1 ) No. Let n and m be 3. 72 Chapter 4 . so move to the next question. so move to the next question. try it for the case where n and same number. What is the meaning of ( > ( sub1 n ) (sub1 m )) Recur. m are the ( zero ? 3) (zero ? 3) No. (zero ? 1) No. (zero ? 2) No.

( < 8 3) #f. Switch the zero ? lines: (define > (lambda ( n m ) (cond ( ( zero ? n ) #f ) ((zero ? m ) #t ) (else ( > (sub1 n ) ( sub1 m )))))) What is ( < 4 6) #t . because 3 is not greater than 3. Try it out ! Yes. then try. (define < (lambda ( n m ) (cond ( ( zero ? m ) #f ) ((zero ? n ) #t ) (else ( < (sub1 n ) ( sub1 m )))))) Numbers Games 73 . (< 6 6) Now try to write < #f. Think about it. No.(zero ? 0) Is this correct? Does the order of the two terminal conditions matter? Does the order of the two terminal conditions matter? Does the order of the two previous answers matter? How can we change the function > to take care of this subtle problem? Yes. Think first . so the value of ( > n m ) is #t .

the natural recursion also looks strange. 125 . (define t1 (lambda ( n m ) (cond ((zero ? m ) 1 ) (else ( x n (t n ( sub1 m ))))))) 1 What is a good name for this function? L.Here is the definition of = (define = (lambda ( n m ) (cond ((zero ? m ) (zero ? n )) ((zero ? n ) #f ) (else (= ( sub1 n ) ( sub1 Rewrite = using < and > (define = (lambda ( n m ) (cond ( ( > n m ) #f ) ( ( < n m ) #f ) (else #t ) ) ) ) m )))))) Does this mean we have two different functions for testing equality of atoms? Yes. 8. 1. - n m ) m )))))) 74 Chapter 4 . = (t 1 1 ) (t 2 3) (t 5 3) Now write the function t Hint : See the The First and Fifth Commandments. S: This i s like expt . (define ? ? ? (lambda ( n m ) (cond ( ( < n m ) 0) (else ( add1 ( ? ? ? ( We have never seen this kind of definition before. they are for atoms that are numbers and eq ? for the others.

(define + 1 (lambda (n m) (cond ( ( < n m) O) (else ( add1 ( + ( - n m) m)))))) 1 (values (truncate (/ n m} } } } S : This is like quot ient . When the function returns.What does the first question check? It determines whether the first argument is less than the second one. How do we get there? Easy. it is 3. So what does the function do? It counts how many times the second argument fits into the first one. And what happens in the second line? We recur with a first argument from which we subtract the second argument . we add 1 to the result . = Wouldn't a (ham and cheese on rye) Don't forget the be good right now? m ustard ! 75 Numbers Games . too: ( + 15 4) = 1 + ( + 11 4) 1 + ( 1 + ( + 7 4)) 1 + ( 1 + ( 1 + ( + 3 4) ) ) 1 + ( 1 + ( 1 + 0)). L : (defun quotient ( n m} What is (+ 15 4 ) Easy. And what do we call this? Division.

Now try to write the function length (define length (lambda ( lat) (cond ( ( null ? lat) 0) (else ( add1 ( length ( cdr lat))))))) What is (pick n lat) where n is 4 and lat is ( lasagna spaghetti ravioli macaroni meatball ) macaroni .What is the value of ( length lat) where lat is ( hotdogs with mustard sa uerkraut and pickles ) 6. 76 Chapter 4 . What is (pick 0 lat) where lat is ( a ) N o answer. Try to write the function pick (define pick (lambda ( n lat) (cond ((zero ? ( sub1 n)) ( car lat) ) (else (pick ( sub1 n) ( cdr lat)))))) What is ( rempick n lat) where n is 3 and lat is ( hotdogs with hot mustard ) ( hotdogs with m ustard ) . What is ( length lat) where lat is ( ham and cheese on rye ) 5.

1 L: numberp Is ( number? 76) true or false? True. and atom ?. null?. zero ?.Now try to write rempick (define rempick (lambda ( n lat) (cond ( (zero ? ( sub1 n ) ) ( cdr lat) ) (else ( cons ( car lat) ( rempick (sub1 n) ( cdr lat ) ) ) ) ) ) ) Is ( number?1 a) true or false where a is tomato False. is a primitive function. For example. cdr. where lat is (5 pears 6 prunes 9 dates) the value of ( no-nums lat) is (pears pru nes dates) (define no-nums (lambda ( lat) (cond ( ( null ? lat) (quote ( ) ) ) (else (cond ( ( number? ( car lat) ) ( no-nums ( cdr lat ) ) ) (else ( cons ( car lat) ( no-nums ( cdr lat ) ) ) ) ) ) ) ) ) Numbers Games 77 . eq ?. Now using number? write the function no-nums which gives as a final value a lat obtained by removing all the numbers from the lat . cons . Can you write number? which is true if its argument is a numeric atom and false if it is anthing else? No: number?. like add1 . car. sub1 .

Now write the function occur which counts the number of times an atom a appears in a lat (define occur (lambda (a lat) (cond < --(else (cond (define occur (lambda ( a lat) (cond ( ( null? lat) 0) (else (cond ( ( eq? ( car lat) a) ( add1 ( occur a ( cdr lat)))) (else ( occur a ( cdr lat)))))))) < --­ --.) )))))) < --__ 78 Chapter 4 . for eqan ? itself. of course. Remember to use = for numbers and eq? for all other atoms. (define eqan ? (lambda ( a1 a2 ) (cond ((and ( number? a1 ) ( number? a2 )) ( = a1 a2 ) ) ( (or ( number? a1 ) ( number? a2 )) #f ) (else ( eq? a1 a2 ) ) ) ) ) Can we assume that all functions written using eq ? can be generalized by replacing eq ? by eqan ? Yes. except . (define all-nums (lambda ( lat) (cond ( ( null? lat) (quote ())) (else (cond ( ( number? ( car lat) ) ( cons ( car lat) ( all-nums ( cdr lat) ) ) ) (else ( all-nums ( cdr lat)))))))) Write the function eqan ? which is true if its two arguments ( a1 and a2 ) are the same atom .Now write all-nums which extracts a tup from a lat using all the numbers in the lat .

making it a one-liner. . . . For example.e. By removing the ( cond .Write the function one ? where ( one ? n) is #t if n is 1 and #f ( i. where n is 3 and lat is (lemon meringue salty pie) the value of ( rempick n lat) is (lemon meringue pie) Use the function one ? in your answer. (define rempick (lambda ( n lat) (cond ( ( one ? n) ( cdr lat) ) (else ( cons ( car lat) ( rempick ( sub1 n) ( cdr lat ) ) ) ) ) ) ) Numbers Games 79 . false ) otherwise. ) clause: (define one ? (lambda (n) (= n 1 ) ) ) Now rewrite the function rempick that removes the nth atom from a lat. (define one ? (lambda (n) (cond ((zero ? n) #f ) (else (zero ? ( sub1 n ) ) ) ) ) ) or (define one ? (lambda (n) (cond (else ( = n 1 ) ) ) ) ) Guess how we can further simplify this function.

�' �-· .

. t " ." ((coffee) ((tea)) (and (hick) ) } .What is ( rember* a l} where a is cup and l is ((coffee) cup ((tea) cup) (and (hick}} cup) "rember*" is pronounced "rember-star. Now write rember* t Here is the skeleton: (define rember* (lambda ( a l} (cond ( ( --­ --- --­ ( > --. instead of just the cdr of the list .. *Oh My Gawd *: It 's Full of Stars 81 .> __ --- )))) (define rember* (lambda ( a l} (cond (( n ull? l} (quote ())) ( ( atom ? ( car l}} (cond ( ( eq? ( car l} a) ( rember* a ( cdr l))) ( else ( cons ( car l) ( rember* a ( cdr l))}))) ( else ( cons ( rember* a ( car l}} ( rember* a ( cdr l))})))) Using arguments from one of our previous examples. * " m akes u s think "oh my gawd." ( lat ? l} where l is (((tomato sauce)) ((bean) sauce) (and ((flying)) sauce)) #f. follow through this to see how it works. Notice that now we are recurring down the car of the list . What is ( rember* a l} where a is sauce and l is (((tomato sauce)) ((bean) sauce) (and ((flying)) sauce)) (((tomato)) ((bean)) (and ((flying) ) ) ) .

Is ( car l) an atom where l is ( ( (tomato sa uce) ) ( ( bean) sauce) (and ( (flyi ng) ) sauce) ) No. What is ( insertR * new old l) where new is roast old is chuck and l is ( ( how much (wood ) ) cou ld ((a (wood ) chuck) ) ( ( (chuck) ) ) ( i f ( a ) ((wood chuck) ) ) could chuck wood ) ( (how much (wood) ) cou ld ((a (wood) chuck roast) ) ( ( (chuck roast))) (if (a) ((wood chuck roast))) cou ld chuck roast wood ) . ( define insertR * (lambda ( new old l) ( c o nd ( ( ( ) ) )))) ( defi ne insertR * (lambda ( new old l) ( c o nd ( ( null? l) (quote ())) ( ( atom ? ( car l)) ( c o nd (( eq ? ( car l) old ) ( cons old ( cons new ( insertR * new old ( cdr l))))) (else ( cons ( car l) ( insertR * new old ( cdr l)))))) (else ( cons ( insertR * new old ( car l)) ( insertR * new old ( cdr l))))))) How are insertR * and rember* similar? Each function asks three questions. 82 Chapter 5 . Now write the function insertR * which inserts the atom new to the right of old regardless of where old occurs.

The F irst Commandment (final version) When recurring on a list of atoms. as well as with the cdr. l. ask two questions about it: ( null? lat) and else. n . and else. When recurring on a list of S-expressions. When recurring on a number. ( atom ? ( car l) ) . How are insertR * and rember* similar? How are all *-functions similar? They all ask three questions and recur with the car as well as with the cdr. It recurs with the car when it finds out that the car is a list . They both recur with the car. lat . ask three question about it: ( null? l ) . How are rember* and multirember different? The function multirember does not recur with the car. whenever the car is a list. *Ob My Gawd *: It 's Full of Stars 83 . or . ask two questions about it: (zero ? n) and else.an atom consed onto a list . Why? Because all *-functions work on lists that are either . The function rember* recurs with the car as well as with the cdr. . whenever the car is a list .a list consed onto a list .empty. How are insertR * and rember* similar? Each function recurs on the car of its argument when it finds out that the argument 's car is a list .

The Fourt h Commandment
(final version)
Always change at least one argument while recurring. When recurring on a list of atoms, lat, use ( cdr lat ) . When recurring on a number, n , use (sub1 n) . And when recur­ ring on a list of S-expressions, l , use ( car l) and ( cdr l) if neither ( null ? l) nor ( atom ? ( car l)) are true. It must be changed to be closer to termination. The chang­ ing argument must be tested in the termination condition: when using cdr , test termination with null ? and when using sub1 , test termination with zero ?.

( occursomething a l) where a i s ba nana and l is ((banana) (split ( ( ( ( banana ice) ) ) (crea m ( banana)) sherbet)) ( banana) (bread) (banana brandy) )

5.

What is a better name for occursomething

occur* .

84

Chapter

5

Write occur*

(define occur* (lambda (a l) (cond

< --­ ___ ) < --­ ___ ) ( --- __ ) ) ) )

(define occur* (lambda (a l) (cond ( ( null? l) 0) ( ( atom ? ( car l)) (cond ( ( eq? ( car l) a) ( add1 ( occur * a ( cdr l)))) (else ( occur* a ( cdr l))))) (else ( + ( occur* a ( car l)) ( occur* a ( cdr l ) ) ) ) ) ) )

( subst * new old l) where new is orange old is banana and l is ((banana) (split ((((banana ice) ) ) (cream (banana ) ) sherbet)) (banana) (bread) (banana brandy) )

((orange) (split ( ( ( (orange ice) ) ) (crea m (ora nge)) sherbet)) (orange) (bread) (orange brandy) ) .

Write subst *

( define subst*
(lambda ( new old l) ( cond ( ( null ? l) (quote ( ))) ( ( atom ? ( car l)) ( cond (( eq ? ( car l) old ) ( cons new ( subst * new old ( cdr l)))) ( else ( cons ( car l) (subst * new old ( cdr l)))))) ( else ( cons ( subst* new old ( car l)) ( subst* new old ( cdr l ) ) ) ) ) ) )

(define subst * (lambda ( new old l) ( cond

< --­ ___ ) < --­ ___ ) < --- __ ) ) ) )

*Oh My Gawd*: It 's Full of Stars

85

What is ( insertL * new old l) where new is peeker old is chuck and l is ( (how much (wood ) ) could ( (a (wood ) chuck)) (( (chuck) )) (if (a) ( (wood chuck ) ) ) cou ld chuck wood)

( (how m uch (wood ) ) cou ld ((a (wood) peeker chuck)) ( ( (peeker chuck))) (if (a) ((wood peeker chuck) ) ) cou ld peeker chuck wood) .

Write insertL *

(define insertL * (lambda ( new old l) ( c ond ( ( (
--__ _ ---

---

___ __

) ) ))))

( define insertL * (lambda ( new old l ) ( c ond ( ( null? l ) (quote () ) ) (( atom ? ( car l) ) ( c ond ( ( eq? ( car l) old ) ( cons new ( cons old ( insertL * new old ( cdr l ))))) (else ( cons ( car l ) ( insertL * new old ( cdr l )))))) (else ( cons ( insertL * new old ( car l )) ( insertL * new old ( cdr l )))))))

a l) where a is chi ps and l is ( (potato) (chips ( (with) fish ) (chips) ) )

( member*

#t because the atom chips appears i n the list l.
,

86

Chapter

5

Write member* ( define member* (lambda ( a l) (cond ( ( (

---

___

_ __

--__

---

) ) ))))

( define member* (lambda (a l) (cond ( ( null ? l) #f ) ( ( atom ? ( car l) ) (or ( eq ? ( car l) a) ( member* a ( cdr l) ) ) ) (else (or ( member* a ( car l)) ( member* a ( cdr l ) ) ) ) ) ) )

What i s ( member* a l) where a is chips and l is ( (potato) (chips ( (with) fish) (chips) ) ) Which chips did i t find?

#t .

( ( potato) (chips ( (with) fish ) (chi ps) ) ) . potato.

What is ( leftmost l) where l is ((potato) (chips ( (with) fish) (chips) ) )

What i s ( leftmost l) where l is (((hot) (tuna (and ) ) ) cheese) What is ( leftmost l) where l is (((() four)) 17 (seventeen) ) What i s ( leftmos t (quote ( ) ) )

hot.

C an you describe what leftmost does?

Here is our description : "The function leftmost finds the leftmost atom in a non-empty list of S-expressions that does not contain the empty list."

*Oh My Gawd *: It 's Fun of Stars

87

it only needs to ask two questions. What is (and ( atom ? ( car l)) ( eq ? ( car l) x ) ) where x is pizza and l is (mozzarella pizza) Why is it false? Since (and . No. making its value true. . ) asks questions one at a time until it finds one that is true. but it only recurs on the car.. Then ( or . . . #f. . 88 Chapter 5 . ) does? ( or . which is true. . ) is false. Does leftmost need to ask questions about all three possible cases? Now see if you can write the function leftmost (define leftmost (lambda (l) ( cond (define leftmost (lambda (l) ( co nd ( ( atom ? ( car l)) ( car l)) (else ( leftmost ( car l)))))) < < --­ --__ --- ) )))) D o you remember what ( or . . which is false.Is leftmost a *-function? It works on lists of S-expressions. . the value of ( or . . ) asks ( atom ? ( car l)). hence it is #f. it then asks ( eq ? ( car l) x ) . If it cannot find a true argument . ) stops.. We agreed that leftmost works on non-empty lists that don 't contain empty lists.

" True. If none of the expressions are false. ) stops if the first argument has the value #f. . ) does. . . True or false: it is possible that one of the arguments of (and . Here's one: x is pizza and l is ( pizza ( tastes good ) ) . Put in your own words what (and . . . ) is true. . . . and ( or . -expressions: (and tt {3) (or tt {3) = 1 (cond (tt {3) (else #f)) ) and = (cond ( tt #t) (else {3)) ( eqlist ? l1 l2 ) where l1 is ( strawberry ice crea m ) and l2 is ( strawberry ice cream ) #t . . . ) stops with false. Because of this property. . . . . ) stops if the first argument has the value #t . . (and . however. . ) asks ( atom ? ( car l ) ) . . though both (and . ) can be deft ned as functions in terms of (cond . Since (and . ) also has the property of not considering all of its arguments. so it is #f. . ) nor (or . . . . neither (and . and ( car l) is not an atom . . . . . . ) . Then (and . ) is not considered? 1 ( cond . . .What is (and ( atom ? ( car l)) ( eq? ( car l) x)) where x is pizza and l is (( mozzarella m ushroom ) pizza ) Why is it false? #f. . because (and . ) asks questions one at a time until it finds one whose value is false. ) can be expressed as abbreviations of ( cond . . ) and ( or . . . We put it in our words: "(and . *Oh My Gawd*: It 's Full of Stars 89 . Give an example for x and l where (and ( atom ? ( car l)) ( eq? ( car l) x)) is true. ) and (or .

an atom consed onto a list .empty.( eqlist ? l1 l2 ) where l1 is (strawberry ice crea m ) and l2 is (strawberry crea m ice) ( eqlist ? l1 l2 ) where l1 is (banana ( (split) ) ) and l2 is ( ( banana) (split)) ( eqlist? l1 l2 ) where l1 is (beef ( (sa usage)) (and (soda))) and l2 is (beef ( (sala m i ) ) (and (soda))) ( eqlist ? l1 l2 ) where l1 is (beef ( (sausage)) (and (soda))) and l2 is (beef ( (sa usage) ) (and (sod a ) ) ) What i s eqlist ? #f. It is a function that determines if two lists are equal. For example.a list consed onto a list . That 's better. Nine. or ." 90 Chapter 5 . How many questions will eqlist ? have to ask about its arguments? Can you explain why there are nine questions? Here are our words: "Each argument may be either . the second argument could be the empty list or have an atom or a list in the car position. . #f. but almost #t . #t . #f. at the same time as the first argument may be the empty list .

the second argument is neither the empty list nor a list with an atom as the first element . we know that when the first argument is empty. True or false: if the first argument is () eqlist ? responds with #t in only one case. l2 must also be the empty list. the second argument must be a list whose first element is also a list. True. And why is the third question ( null? l1 ) At that point . *Oh My Gawd *: It 's Full of Stars 91 . If ( null? l1 ) is true now. Otherwise the first question would have been true. For ( eqlist ? (quote ( ) ) l2 ) to be true.Write eqlist ? using eqan ? (define eqlist ? (lambda ( l1 l2 ) ( c ond ((and ( null? l1 ) ( null? l2) ) #t ) ((and ( null? l1 ) ( atom ? ( car l2 ) ) ) #f ) ( ( null? l1 ) #f ) ((and ( atom ? ( car l1 ) ) ( null? l2 ) ) #f ) ((and ( atom ? ( car l1 ) ) ( atom ? ( car l2 ) ) ) (and ( eqan ? ( car l1 ) ( car l2 ) ) ( eqlist ? ( cdr l 1 ) ( cdr l2 ) ) ) ) ( ( atom ? ( car l1 ) ) #f ) ( ( null? l2 ) #f ) ( ( atom ? ( car l2 ) ) #f ) (else (and ( eqlist ? ( car l1 ) ( car l2 ) ) ( eqlist ? ( cdr l 1 ) ( cdr l2 ) ) ) ) ) ) ) Is it okay to ask ( atom ? ( car l2 ) ) in the second question? Yes. because we know that the second list cannot be empty.

Does this mean that the questions (and ( null? 11 ) ( null? l2 ) ) and ( or ( null? 11 ) ( null? 12 )) suffice to determine the answer in the first three cases? Rewrite eqlist ?

Yes. If the first question is true, eqlist ? responds with #t ; otherwise, the answer is #f.

( define eqlist ? (lambda ( l1 12 ) ( cond ((and ( null? 11 ) ( null? 12) ) #t ) (( or ( null? 11 ) ( null? 12)) #f ) ((and ( atom ? ( car l1 )) ( atom ? ( car 12) ) ) (and ( eqan ? ( car 11 ) ( car 12 )) ( eqlist ? ( cdr 11 ) ( cdr 12 )))) (( or ( atom ? ( car l1 ) ) ( atom ? ( car 12 ))) #f ) (else (and ( eqlist ? ( car 11 ) ( car 12) ) ( eqlist? ( cdr 11 ) ( cdr l2 )))))))

What i s an S-expression?

An S-expression is either an atom or a (possibly empty) list of S-expressions. Four. The first argument may be an atom or a list of S-expressions at the same time as the second argument may be an atom or a list of S-expresssions.

How many questions does equal? ask to determine whether two S-expressions are the same?

Write equal?

(define equal? (lambda (s1 s2 ) ( cond ((and ( atom ? s1 ) ( atom ? s2)) ( eqan ? s1 s2 )) ( ( atom ? s1 ) #f ) (( atom? s2 ) #f ) (else ( eqlist? s1 s2 )))))

92

Chapter 5

Why is the second question ( atom ? s1 )

If it is true, we know that the first argument is an atom and the second argument is a list .

And why is the third question ( atom ? s2)

By the time we ask the third question we know that the first argument is not an atom. So all we need to know in order to distinguish between the two remaining cases is whether or not the second argument is an atom . The first argument must be a list.

Can we summarize the second question and the third question as ( or ( atom ? s1 ) ( atom ? s2 ) )

Yes, we can!

Simplify equal?

(define equal ? (lambda (s1 s2 ) ( co nd ((and ( atom ? s1 ) ( atom ? s2 ) ) ( eqan ? s1 s2 ) ) ( ( or ( atom ? s1 ) ( atom ? s2 ) ) #f ) (else ( eqlist ? s1 s2 ) ) ) ) )

Yes. The questions cover all four possible cases.

Now, rewrite eqlist ? using equal?

(define eqlist ? (lambda ( l 1 l2 ) ( co nd ((and ( null? l1 ) ( null? l2 ) ) #t ) ( ( or ( null? l1 ) ( null? l2 ) ) #f ) (else (and ( equal? ( car l1 ) ( car l2 ) ) ( eqlist ? ( cdr l 1 ) ( cdr l2 ) ) ) ) ) ) )

*Ob My Gawd*: It 's Full of Stars

93

T he S ixt h Commandment
Simplify only after the function is correct.

Here is rember after we replace lat by a list l of S-expressions and a by any S-expression.

Obviously!

(define rember (lambda ( s l) (cond ( ( null? l) (quote ( ) ) ) ( ( atom ? ( car l)) (cond ( ( equal? ( car l) s) ( cdr l)) (else ( cons ( car l) ( rember s ( cdr l)))))) (else (cond ( ( equal ? ( car l) s) ( cdr l)) (else ( cons ( car l) ( rember s ( cdr l ) ) ) ) ) ) ) ) )
Can we simplify it?

(define rember (lambda (s l) (cond ( ( null ? l) (quote ())) (else (cond ( ( equal? ( car l) s) ( cdr l)) (else ( cons ( car l) ( rember s ( cdr l)))))))))

And how does that differ?

The function rember now removes the first matching S-expression s in l, instead of the first matching atom a in lat.

Is rember a "star" function now? Why not ?

No.

Because rember recurs with the cdr of l only.

Can rember be further simplified?

Yes, the inner ( cond . . . ) asks questions that the outer ( cond . . . ) could ask!

94

Chapter

5

Do it !

(define rember (lambda ( s l) (cond ( ( null? l) (quote ( ) ) ) ( ( equal? ( car l) s ) ( cdr l)) (else ( cons ( car l) ( rember s ( cdr l ) ) ) ) ) ) )

Does this new definition look simpler?

Yes, i t does!

And does it work just as well?

Yes, because we knew that all the cases and all the recursions were right before we simplified.

Simplify insertL *

We can't. Before we can ask ( eq 'l ( car l) old) we need to know that ( car l) is an atom .

When functions are correct and well-designed, we can think about them easily.

And that saved us this time from getting it wrong.

Can all functions that use eq? and = be generalized by replacing eq 'l and = by the function equal?

Not quite; this won't work for eqan ?, but will work for all others. In fact , disregarding the trivial example of eqan 'l, that is exactly what we shall assume.

*Oh My Gawd *: It 's Full of Stars

95

(])EO �dl�l@) .

What is an arithmetic expression in your words? In ours: "For the purpose of this chapter. Is 1 + 3 an arithmetic expression? Yes! Is 1 +3 x 4 an arithmetic expression? Definitely. Are you almost ready for one? And. What does (quote x ) stand for? The atom x.. what about 3 t y + 5 Yes. an arithmetic expression is either an atom (including numbers). Is cookie an arithmetic expression? Yes. Shadows 97 . or t." What is (quote a) a. not the operation x. That 's the same question again.Is 1 an arithmetic expression? Yes. x . What is (quote +) The atom + . or two arithmetic expressions combined by + . of course. And the answer is still true. not the operation -11. Is ( eqP (quote a ) 11) true or false where 11 is a Is ( eqP x 11) true or false where x is a and 11 is a True. Is 3 an arithmetic expression? Yes .

( n + 3) is an S-expression. What is numbered ? It is a function that determines whether a representation of an arithmetic expression contains only numbers besides the + . 2. Our definition of arithmetic expression does not mention parentheses. How do you represent 3 +4 5 (3 + (4 X 5)). True or false: ( numbered ? y) where y is (3 + (4 t 5)) True. Could we think of ( n + 3) expression? as an arithmetic Yes. because sausage is not a number. 98 Chapter 6 .Is ( n + 3) an arithmetic expression? Not really. True or false: ( numbered ? x ) where x is 1 x True. Because 1. since there are parentheses around n + 3. What would you call ( n + 3) Why is ( n + 3) a good representation? We call it a representation for n + 3. if we keep in mind that the parentheses are not really there. True or false: ( numbered ? z ) where z is (2 x sa usage) False. x . and t. It structurally resembles n + 3. It can therefore serve as an argument for a function.

Shadows 99 . What is ( eq? ( car ( cdr aexp)) (quote + ) ) I t i s the second question. And you must know the fourth one. instead of two. What is the first question? ( atom ? aexp ) . ( eq? ( car ( cdr aexp)) (quote t)). questions about arithmetic expressions? After all. not as a list itself. arithmetic expressions like (1 + 3) are lats. an arithmetic expression is either a number. o f course. or two arithmetic expressions combined by + . And. or t. Because we consider (1 + 3) as a representation of an arithmetic expression in list form . x .Now can you write a skeleton for numbered ? (define numbered ? (lambda ( aexp ) (cond ( ( -- --- ( ( _ __ --- --­ --__ --- ) ) ) )))) is a good guess. C an you guess the third one? ( eq ? ( car ( cdr aexp)) (quote x ) ) is perfect . Why do we ask four. Should we ask another question about aexp No! So we could replace the previous question by else.

Now you can almost write numbered ? Here is our proposal: (define numbered ? (lambda ( aexp) (cond ( ( atom ? aexp) ( number? aexp)) ( ( eq ? ( car ( cdr aexp)) (quote +)) .) ( ( eq? ( car ( cdr aexp)) (quote x )) . . In which position is the first subexpression? It is the car of aexp. . In which position is the second subexpression? It is the car of the cdr of the cdr of aexp. .)))) Why do we ask ( number? aexp) when we know that aexp is an atom? Because we want t o know i f all arithmetic expressions that are atoms are numbers.) ( ( eq? ( car ( cdr aexp)) (quote t) ) . What is the second answer? (and ( numbered ? ( car aexp)) ( numbered ? ( car ( cdr ( cdr aexp) )))) 100 Chapter 6 . . Both must b e true. What do we need to know if the aexp consists of two arithmetic expressions combined by + We need to find out whether the two subexpressions are numbered. So what do we need to ask? ( numbered ? ( car aexp)) and ( numbered ? ( car ( cdr ( cdr aexp) ) ) ) . . .

Shadows 101 . (define numbered ? (lambda ( aexp) (cond ( ( atom ? aexp) ( number? aexp)) ( ( eq ? ( car ( cdr aexp)) (quote + ) ) (and ( numbered ? ( car aexp)) ( numbered ? ( car ( cdr ( cdr aexp ) ) ) ) ) ) ( ( eq? ( car ( cdr aexp)) (quote x )) (and ( numbered ? ( car aexp)) ( numbered ? ( car ( cdr ( cdr aexp ) ) ) ) ) ) ( ( eq ? ( car ( cdr aexp)) (quote t ) ) (and ( numbered ? ( car aexp)) ( numbered ? ( car ( cdr ( cdr aexp) ) ) ) ) ) ) ) ) Since aexp was already understood to be an arithmetic expression. could we have written numbered? in a simpler way? Yes: (define numbered ? (lambda ( aexp) (cond ( ( atom ? aexp) ( number? aexp)) (else (and ( numbered ? ( car aexp)) ( numbered ? ( car ( cdr ( cdr aexp) ) ) ) ) ) ) ) ) Why can we simplify? Because we know we've got the function right . (value x) where x is (1 + 3) 4.Try numbered ? again. What is ( value u) where u is 13 13.

( value nexp) returns what we think is the natural value of a numbered arithmetic expression. .) ( ( eq? ( car ( cdr nexp)) (quote x )) . ) ( ( eq? ( car ( cdr nexp)) (quote +)) . We hope. . . . No answer. . ) ) ) ) What is the natural value of an arithmetic expression that is a number? I t i s just that number. . Now. How many questions does value ask about nexp Four. we could just add up the two values. . 102 Chapter 6 . Can you think of a way to get the value of the two subexpressions in ( 1 + (3 x 4 ) ) Of course. and applying value to (3 x 4) . What is the natural value of an arithmetic expression that consists of two arithmetic expressions combined by + If we had the natural value of the two subexpressions. by applying value to 1 .( value y) where y is (1 + (3 t 4) ) ( value z ) where z is cookie 82.) (else . . let 's attempt to write value (define value (lambda ( nexp) (cond ( ( atom ? nexp) .

Shadows 103 . (define value (lambda ( nexp) (cond ( ( atom ? nexp) nexp) ( ( eq ? ( car ( cdr nexp)) (quote +)) ( + ( value ( car nexp)) ( value ( car ( cdr ( cdr nexp)))))) ( ( eq ? ( car ( cdr nexp)) (quote x ) ) ( x ( value ( car nexp ) ) ( value ( car ( cdr ( cdr nexp)))))) (else ( t ( value ( car nexp)) ( value ( car ( cdr ( cdr nexp) ) ) ) ) )))) Can you think of a different representation of arithmetic expressions? There are several of them. Could (3 4 +) represent 3 + 4 Could ( + 3 4) Yes. Or (plus 3 4 ) Yes.And in general? By recurring with value on the subexpressions. Give value another try. The S event h Commandment Recur on the subparts that are of the same nature: • On the sublists of a list. • On the subexpressions of an arithmetic expression. Yes.

What about (define value (lambda ( nexp) (cond ( ( atom ? nexp) nexp) ( ( eq ? ( car nexp) (quote +)) ( -11. It's wrong.Is (+ ( x 3 6) (t 8 2)) a representation of an arithmetic expression? Yes. 104 Chapter 6 . (+ 1 3) . Let 's try an example.( value ( cdr nexp)) ( value ( cdr ( cdr nexp))))) ( ( eq ? ( car nexp) (quote x )) ( x ( value ( cdr nexp)) ( value ( cdr ( cdr nexp))))) (else (t ( value ( cdr nexp )) ( value ( cdr ( cdr nexp)))))))) You guessed it. ( eq? ( car nexp) (quote + ) ) where nexp is ( + 1 3) Yes.a list of the atom t followed by two arithmetic expressions.a number . And now recur. ( atom ? nexp) where nexp is ( + 1 3 ) No.a list of the atom + followed by two arithmetic expressions. .a list of the atom x followed by two arithmetic expressions. or . Try to write the function value for a new kind of arithmetic expression that is either: . What is ( cdr nexp) where nexp is (+ 1 3) Yes. ( 1 3) .

not all lists are representations of arithmetic expressions. (define 1st-sub-exp (lambda ( aexp) (cond (else ( car ( cdr aexp ) ) ) ) ) ) Why do we ask else Because the first question i s also the last question . ) if we don't need to ask questions? Yes. the cdr of the cdr is (3) . We have to recur on subexpressions. Again. . What do we mean if we say the car of the cdr of nexp The first subexpression of the representation of an arithmetic expression. . But remember. we were thinking of the list ( + 1 3) instead of the representation of an arithmetic expression. remember one-liners from chapter 4. Taking the car of the cdr of the cdr gets us back on the right track. How can we get the first subexpression of a representation of an arithmetic expression? By taking the car of the cdr. (1 3) is not a subpart that is a representation of an arithmetic expression! We obviously recurred on a list . No. we violated The Seventh Commandment .(1 3) is not our representation of an arithmetic expression. Is ( cdr ( cdr nexp)) an arithmetic expression where nexp is ( + 1 3) No. Can we get by without ( cond . and (3) is not an arithmetic expression . (define 1st-sub-exp (lambda ( aexp) ( car ( cdr aexp)))) Shadows 105 . Let 's write a function 1st-sub-exp for arithmetic expressions.

let 's replace ( car nexp) by ( operator nexp) (define operator (lambda ( aexp) ( car aexp))) Now write value again.Write 2nd-sub-exp for arithmetic expressions. by changing 1st-sub-exp and operator. Do it ! (define 1st-sub-exp (lambda ( aexp ) ( car aexp))) ( define operator (lambda ( aexp ) ( car ( cdr aexp)))) 106 Chapter 6 . (define value (lambda ( nexp) (cond ( ( atom ? nexp) nexp) ( ( eq ? ( operator nexp) (quote +)) ( + ( value ( 1 st-sub-exp nexp)) ( value ( 2nd-sub-exp nexp) ))) ( ( eq ? ( operator nexp) (quote x )) ( x ( value ( 1 st-sub-exp nexp)) ( value ( 2nd-sub-exp nexp) ))) (else ( t ( value ( 1st-sub-exp nexp)) ( value ( 2nd-sub-exp nexp) )))))) Can we use this value function for the first representation of arithmetic expressions in this chapter? Yes . (define 2nd-sub-exp (lambda ( aexp) ( car ( cdr ( cdr aexp))))) Finally.

How shall we represent zero now? () is our choice. Shadows 107 . What else could we have used? ( () () () ( ) ) would have served j ust as well. Let's try another representation for numbers. we just did not tell you that they were representations. For what entities have we used representations? Truth-values! Numbers! Numbers are representations? Yes. because we used help functions to hide the representation . How is one represented? (()). We chose that symbol because we are accustomed to arabic representations. How is two represented? (() ()). The Eight h Commandment Use help functions to abstract from representations. Have we seen representations before? Yes. add1 .Wasn't this easy? Yes . For example 4 stands for the concept four. What about ( ( ( ( ( ) ) ) ) ) ? How about (I V)? Do you remember how many primitives we need for numbers? Four: number?. zero ?. and sub1 .

Recall The Law of Cdr. but only slightly. It changed. What is (zub1 n) where n is ( ) No answer.Got it? What 's three? Three is ( () () () ) . 108 Chapter 6 . (define sero ? (lambda (n) ( null? n))) Can you write a function that is like add1 (define edd1 (lambda (n) ( cons (quote ()) n))) What about sub1 (define zub1 (lambda (n) ( cdr n))) Is this correct? Let 's see. (define + (lambda ( n m ) (cond ( ( sero ? m ) n) (else ( edd1 ( + n (zub1 m))))))) Has the definition of + changed? Yes and no. Rewrite + using this representation. Write a function to test for zero. . but that 's fine.

((()) ( ( ) ( ) ) ( ( ) ( ) ( ) ) ) . You must beware of shadows. Shadows 109 . .Recall lat ? Easy: (define lat ? (lambda (l) (cond ( ( null? l) #t ) ( ( atom ? ( car l)) ( lat? ( cdr l))) (else #f ) ) ) ) But why did you ask? Do you remember what the value of ( lat ? ls ) is where ls is (1 2 3) What is ( 1 2 3) with our new numbers? #t of course. What is ( lat? ls ) where ls is ((( )) (() ( )) (() ( )())) Is that bad? It is very false.

� lUhi]� � f!� � - " ��� .

because no atom appears more than once. because no atom appears more than once.Is this a set? (apple peaches apple plum) No. How about ( set ? lat) where lat is ( ) #t . Friends and Relations 111 . Were you surprised to see the function member? appear in the definition of set ? You should not be. True or false: ( set ? lat) where lat is (a pples peaches pears plu ms) #t . and now we can use it whenever we want . Try t o write set ? (define set ? (lambda ( lat) (cond ( ( null? lat) #t ) (else (cond ( ( member? ( car lat) ( cdr lat) ) #f ) (else ( set ? ( cdr lat ) ) ) ) ) ) ) ) Simplify IJet ? (define set ? (lambda ( lat) (cond ( ( null ? lat) #t ) ( ( member? ( car lat) ( cdr lat) ) #f ) (else (set? ( cdr lat ) ) ) ) ) ) Does this work for the example (apple 3 pear 4 9 apple 3 4} Yes. since member? i s now written using equal ? instead of eq ?. since a pple appears more than once. because we have written member? already.

Using the previous definition .What is ( makeset lat) where lat is (apple peach pear peach plum a pple lemon peach) (apple peach pear p l u m lemon ) . Try to write makeset using multirember (define makeset (lambda ( lat) (cond ( ( null ? lat) (quote ())) (else ( cons ( car lat) ( makeset ( multirember ( car lat) ( cdr lat)))))))) What is the result of ( makeset lat) using this second definition where lat is (apple peach pear peach plum apple lemon peach) (apple peach pear plum lemon ) . what is the result of ( makeset lat) where lat is (apple peach pear peach plum apple lemon peach) (pear plum apple lemon peach) . 112 Chapter 7 . Try to write makeset using member ? (define makeset (lambda ( lat) (cond ( ( null? lat) (quote ())) ( ( member? ( car lat) ( cdr lat) ) ( makeset ( cdr lat))) (else ( cons ( car lat) ( makeset ( cdr lat))))))) Are you surprised to see how short this is? We hope so. But don't be afraid: it 's right.

Here are our words: "The function makeset remembers to cons the first atom in the lat onto the result of the natural recursion. since multirember is now written using equal? instead of eq ?.Describe in your own words how the second definition of makeset works. because each atom in set1 is also in set2." Does the second makeset work for the example (apple 3 pear 4 9 apple 3 4) Yes. What is (subset? set1 set2 ) where set1 is (4 pounds of horseradish) and set2 is (four pounds chicken and 5 ou nces horseradish) #f. Write subset? (define subset ? (lambda (set1 set2 ) (cond ( ( null? set1 ) #t ) (else (cond ( ( member? ( car set1 ) set2 ) ( subset? ( cdr set1 ) set2 ) ) (else #f ) ) ) ) ) ) Friends and Relations 1 13 . after removing all occurrences of the first atom from the rest of the lat . What is (subset ? set1 set2 ) where set1 is (5 chicken wings) and set2 is (5 hamburgers 2 pieces fried chicken and light duckling wings) #t .

. ) (define subset ? (lambda (set1 set2 ) (cond ( ( null? set1 ) #t ) (else (and ( member? ( car set1 ) set2 ) ( subset ? ( cdr set1 ) set2)))))) What is ( eqset ? set1 set2 ) where set1 is (6 large chickens with wings) and set2 is (6 chickens with large wings) #t . Write eqset ? (define eqset ? (lambda ( set1 set2 ) (cond ( ( subset ? set1 set2 ) ( subset? set2 set1 ) ) (else #f ) ) ) ) Can you write eqset ? with only one cond-line? (define eqset ? (lambda ( set1 set2 ) (cond (else (and ( subset? set1 set2 ) ( subset? set2 set1 )))))) 1 14 Chapter 7 . .Can you write a shorter version of subset ? (define subset? (lambda ( set1 set2 ) (cond ( ( null ? set1 ) #t ) ( ( member? ( car set1 ) set2 } (subset ? ( cdr set1 ) set2 ) ) (else #f ) ) ) ) Try to write subset ? with (and .

(define eqset ? (lambda ( set1 set2 ) (and ( subset ? set1 set2 ) (subset ? set2 set1 ) ) ) ) What is ( intersect? set1 set2 ) where set1 is (stewed tomatoes and macaroni) and set2 is ( macaroni and cheese) Define the function intersect? #t . . . ) (define intersect ? (lambda ( set1 set2 ) ( cond ( ( null ? set1 ) #f ) (else ( or ( member? ( car set1 ) set2 ) ( intersect? ( cdr set1 ) set2 ) ) ) ) ) ) Compare subset? and intersect ?.Write the one-liner. (define intersect? (lambda ( set1 set2 ) ( cond ( ( null ? set1 ) #f ) (else ( cond ( ( member? ( car set1 ) set2 ) #t ) (else ( intersect ? ( cdr set1 ) set2 ) ) ) ) ) ) ) Write the shorter version. (define intersect? (lambda ( set1 set2 ) ( cond ( ( null? set1 ) #f ) ( ( member? ( car set1 ) set2 ) #t ) (else ( intersect? ( cdr set1 ) set2 ) ) ) ) ) Try writing intersect? with ( or . Friends and Relations 115 . because at least one atom in set1 is in set2 .

set1 and is is set2 Now you can write the short version of intersect (define intersect (lambda (set1 set2 ) (cond (( null? set1 ) (quote ())) (( member? ( car set1 ) set2 ) ( cons ( car set1 ) ( intersect ( cdr set1 ) set2))) (else (intersect ( cdr set1 ) set2))))) What is where ( union set1 set2 ) (stewed tomatoes and macaroni casserole) (macaroni and cheese) (stewed tomatoes casserole macaroni and cheese) set1 and is set2 is Write union (define union (lambda (set1 set2) (cond (( null? set1 ) set2 ) ( ( member? ( car set1 ) set2 ) ( union ( cdr set1 ) set2)) (else ( cons ( car set1 ) ( union ( cdr set1 ) set2)))))) 1 16 Chapter 7 .What is where ( intersect set1 set2 ) (stewed tomatoes and macaroni) (macaroni and cheese) (and macaroni).

Now. because i t i s a list with only two atoms. What i s ( intersectall l-set) where l-set is ( ( a b c) ( c a d e) ( e f g h a b)) (a) .What is this function? (define xxx (lambda ( set1 set2 ) (cond ( ( null? set1 ) (quote ( ) ) ) ( ( member? ( car set1 ) set2 ) ( xxx ( cdr set1 ) set2 ) ) (else ( cons ( car set1 ) ( xxx ( cdr set1 ) set2 ) ) ) ) ) ) In our words: "It is a function that returns all the atoms in set1 that are not in set2. Friends and Relations 117 . What is ( intersectall l-set) where l-set is ( ( 6 pears and) (3 peaches and 6 peppers) (8 pears and 6 plu ms) (and 6 pru nes with some apples) ) (6 a nd ) . (define intersectall (lambda ( l-set) (cond ( ( null? ( cdr l-set)) ( car l-set)) (else ( intersect ( car l-set) ( intersectall ( cdr l-set ) ) ) ) ) ) ) Is this a pair?1 ( pear pear ) Yes. using whatever help functions you need. write intersectall assuming that the list of sets is non-empty. 1 A pair in Scheme (or Lisp) is a different but related object. xxx is the ( set ) difference function." That is.

By taking the car of the cdr of the pair. Define a-pair? (define a-pair? (lambda (x) (cond ( ( ato m ? x) #f ) ( ( n ull ? x) #f ) ( ( null ? ( cdr x)) #f ) ( ( n ull ? ( cdr ( cdr x))) #t ) (else #f ) ) ) ) How can you refer to the first S-expression of a pair? How can you refer to the second S-expression of a pair? By taking the car of the pair. there aren't any. That is. Did you notice the differences between the last two answers? 1 18 Chapter 7 . because it is a list with only two S-expressions.Is this a pair? (3 7) Is this a pair? ((2) ( pair) ) Yes. Yes. How can you build a pair with two atoms? You cons the first one onto the cons of the second one onto ( ) That is. ( cons x1 ( cons x2 (quote ( ) ) ) ) . How can you build a pair with two S-expressions? You cons the first one onto the cons of the second one onto () . ( cons x1 ( cons x2 (quote ( ) ) ) ) . ( a-pair? l ) where l is (fu l l (house)) #t . . because it is a list with only two S-expressions. No.

They will be used to improve readability. We use rei to stand for relation . See chapter 6. Is l a rei where l is ( (a pples peaches) (pumpkin pie) (apples peaches) ) No . and build as one-liners. second .(define first (lambda (p) ( cond (else ( car p))))) (define second (lambda (p) (cond (else ( car ( cdr p ) ) ) ) ) ) (define build (lambda (sl s2 ) (cond ( else ( cons sl ( cons s2 (quote ( ) ) ) ) ) ) ) ) What possible uses d o these three functions have? Can you write third as They are used to make representations of pairs and to get parts of representations of pairs. since l i s not a list of pairs. Is l a rei where l is ((apples peaches) (pum pki n pie) ) Yes. since l is not a set of pairs. Redefine first. a one-liner? ( define third (lambda ( l ) ( car ( cdr ( cdr l))))) Is l a rei where l is (apples peaches pumpkin pie) No. Friends and Relations 119 . Is l a rei where l is ( (4 3) (4 2) (7 6) (6 2) (3 4) ) Yes. as you will soon see.

What i s (fun ? rel) where rel is ((d 4) (b 0) (b 9) (e 5) (g 4) ) #f. What i s (fun ? rel) where rel is ((8 3) (4 2) (7 6) (6 2) (3 4) ) #t . Write fun ? with set ? and firsts (define fun ? (lambda ( rel) (set? (firsts rel) ) ) ) Is fun ? a simple one-liner? How do we represent a finite function? I t sure is. because (firsts rel) is a set -See chapter 3 . We use fun to stand for function. You can now write revrel (define revrel (lambda ( rel) (cond ( ( null? rel) (quote ())) (else ( cons ( build (second ( car rel)) (first ( car rel))) ( revrel ( cdr rel) )))))) 120 Chapter 7 . because b is repeated. a finite function is a list of pairs in which no first element of any pair is the same as any other first element . What is ( revrel rel) where rel is ( (8 a) (pumpkin pie) (got sick)) (( a 8) (pie pumpkin) (sick got ) ) . For us.Is rel a fun where rel is ((4 3) (4 2) (7 6) (6 2) (3 4) ) No.

Friends and Relations 121 . but now do you see how representation aids readability? (define revrel (lambda ( rel) (cond ( ( null? rel) (quote ( ) ) ) (else ( cons ( cons ( car ( cdr ( car rel) ) ) ( cons ( car ( car rel)) (quote ( ) ) ) ) ( revrel ( cdr rel) ) ) ) ) ) ) Suppose we had the function revpair that reversed the two components of a pair like this: No problem . and it is even easier to read: ( define revrel (lambda ( rel) ( cond ( ( null? rel) (quote ())) ( else ( cons ( revpair ( car rel) ) ( revrel ( cdr rel))))))) (define revpair (lambda (pair) ( build (second pair) (first pair)))) How would you rewrite revrel to use this help function? Can you guess why fun is not a fullfun where fun is ( (8 3) (4 2) (7 6) (6 2) (3 4) ) fun is not a fullfun. since the 2 appears more than once as a second item of a pair. What is (fullfun '? fun) where fun is ( (grape raisin) (pl u m prune) (stewed prune) ) #f.Would the following also be correct: Yes. Why is #t the value of (fullfun ? fun) where fun is ( (8 3) (4 8) (7 6) (6 2) (3 4) ) Because (3 8 6 2 4) is a set .

Can you think of a second way to write one-to-one ? (define one-to-one ? (lambda (fun) (fun ? ( revrel fun) ) ) ) Is ((chocolate chip) (doughy cookie)) a one-to-one function? Yes. because ( raisin prune grape) is a set. and you deserve one now! Go and get one! 122 Chapter 7 . Define fullfun ? (define fullfun ? (lambda (fun) (set? (seconds fun) ) ) ) Can you define seconds I t i s just like firsts . What is another name for fullfun ? one-to-one ?.What is (fullfun ? fun) where fun is ( (gra pe raisi n) (plum prune) (stewed grape)) #t .

O r b etter yet .powder 1 teaspoon)) (quote (baking-soda 1 teaspoon) ) ) ( mix (quote (eggs 2 large)) (quote (vanilla 1 teaspoon)) ( cream (quote (butter 1 cup)) (quote (sugar 2 cups)) ) )))))) Friends and Relations 123 .5 teaspoon) ) (quote (ba king. make your own . (define cookies (lambda () ( bake (quote (350 degrees)) (quote ( 12 minutes)) ( mix (quote (walnuts 1 cup)) (quote (chocolate-chips 16 ounces)) ( mix ( mix (quote (flou r 2 cups)) (quote (oatmeal 2 cups) ) (quote (salt .

.

Lambda the Ultimate 125 . (6 2 3) . By passing equal? as an argument to rember-t . because we have not yet told you how. By passing c and (b c a) as arguments to rember. L : (rember-f (function but there is more. 1 • ) 5 ' (6 2 5 3) ) . By passing a and (b c a) as arguments to rember. What is ( rember-t test? a l) where test? is eq? a is jelly and l is Uelly beans are good) And what is ( rember-t test? a l) where test ? is equal? a is (pop corn) and l is (lemonade (pop corn ) and (cake)) ( beans are good ) . ( lemonade a n d (cake) ) .Remember what we did in rember and inserlL at the end of chapter 5? Can you write a function rember-t that would use either eq ? or equal ? How can you make rember remove the first a from (b c a ) How can you make rember remove the first c from (b c a) How can you make rember-t use equal ? instead of eq? What is ( rember-t test ? a l) where test? is = 1 a is 5 and l is (6 2 5 3) We replaced eq ? with equal? No.

Try to write rember-f (define rember-f (lambda ( test ? a l) (cond ( ( null ? l) (quote ())) (else (cond ( ( test ? ( car l) a) 1 ( cdr l)) (else ( cons ( car l) ( rember-f test ? a ( cdr l))))))))) This is good! 1 L : (funcall t e s t ? (car 1 ) a) . 1 26 Chapter 8 . This is j ust rember with eq ? replaced by equal ?. Yes: rember with = rember with equal ? rember with eq ? and rember-f . acts like rember. What about the short version? (define rember-f (lambda ( test ? a l) (cond ( ( null ? l) (quote ( ) ) ) ( ( test? ( car l) a) ( cdr l)) (else ( cons ( car l) ( rember-f test ? a ( cdr l))))))) How does ( rember-f test ? a l) act where test? is eq ? And what about ( rember-f test? a l) where test? is equal? Now we have four functions that do almost the same thing. ( rember-f test? a l) where test? is eq ?. U s e funcall when invoking a function argument or a function that has not been defuned .

Can you say what (lambda ( a l) . ) give the preceding function a name. Lists and atoms. Yes. ." Using (define . It is a function that. returns the function Now what is (lambda ( a ) (lambda ( x ) ( eq? x a ))) (lambda ( x ) ( eq? x a )) where a is just that argument . . Moses Schonfinkel ( 1889--1942 ) . a and l. 1 L: (defun eq?-c (a) (funct ion (lambda (x) (eq x a) ) ) ) What is ( eq?-c k) where k is sa lad Its value is a function that takes x as an argument and tests whether it is eq ? to salad . . Thank you. but you probably did not know that yet . ) is? (lambda ( a l) . . Haskell B. ) is a function of two arguments. (define eq ?-c 1 (lambda ( a ) (lambda ( x ) ( eq? x a )))) This i s our choice. .And rember-f can behave like all the others. Lambda the Ultimate 127 . It is not called "Schonfinkel-ing. when passed an argument a. Curry (1900-1982) . I s this called "Curry-ing?" Thank you. . What kind of values can functions return? What about functions themselves? Let 's generate all versions with rember-f .

complete the line ( cons ( car l) . (define rember-eq ? ( rember-f test ?) ) where test ? is eq ?. and the first one that is eq ? to a is removed. Lambda the Ultimate 129 .Describe in your own words the result of ( rember-f test ?) where test? is eq? Give a name to the function returned by ( rember-f test ?) where test? is eq? It is a function that takes two arguments. Now. (sa lad is good ) . What is ( rember-eq? a l) where a is tuna and l is (tuna salad is good) Did we need to give the name rember-eq? to the function ( rember-f test ?) where test? is eq? No. . It compares the elements of the list with a. we could have written ( ( rember-f test ?) a l) where test? is eq ? a is tuna and l is (tuna salad is good ) . a and l. ) in rember-f so that rember-f works. (define rember-f (lambda ( test ?) (lambda ( a l) (cond ( ( null ? l) (quote ( ) ) ) ( ( test ? ( car l) a) ( cdr l)) (else ( cons ( car l) ( ( rember-f test ?) a ( cdr l ) ) ) ) ) ) ) ) What is ( ( rember-f eq ?) a l) where a is tuna and l is (shri m p salad and tuna salad) (shri m p salad and sa lad ) . .

get yourself some coffee cake and relax! Otherwise.What is ( ( rember-f eq ?) a l) where a is eq? and l is (equal? eq ? eqan? eq list? eqpair? ) 1 1 Did you notice the difference between eq? and eq ? Remember that the former is the atom and the latter is the function. You'll see it in a minute. And now transform insertL to insertL-f the same way we have transformed rember into rember-f (define insertL-f (lambda ( test ?) (lambda ( new old l) (cond ( ( null? l) (quote ())) ( ( test ? ( car l) old) ( cons new ( cons old ( cdr l)))) (else ( cons ( car l) ( ( insertL-f test ?) new old ( cdr l)))))))) And. 130 Chapter 8 . do it to insertR (define insertR-f (lambda ( test ?) (lambda ( new old l) (cond ( ( null? l) (quote ())) ( ( test ? ( car l) old ) ( cons old ( cons new ( cdr l)))) (else ( cons ( car l) ( ( insertR-f test ?) new old ( cdr l)))))))) Are insertR and insertL similar? Can you write a function insert-g that would insert either at the left or at the right? Only the middle piece is a bit different . If you can. (equal? eqan? eqlist? eqpair? ) . j ust for the exercise. don't give up.

Put the difference in words! So how can we get rid of the difference? Define a function seqL that 1. We say: "The two functions cons old and new in a different order onto the cdr of the list l. takes three arguments. but in insertR it is: ( ( eq ? ( car l) old ) ( cons old ( cons new ( cdr l ) ) ) ) . conses the second argument onto the result of cons ing the first argument onto the third argument. Do you know why we wrote these functions? Lambda the Ultimate 131 . What is: (define seqL (lambda ( new old l) ( cons new ( cons old l)))) (define seqR (lambda ( new old l) ( cons old ( cons new l)))) A function that 1 ." You probably guessed it: by passing in a function that expresses the appropriate cons ing. In insertL it is: ( ( eq ? ( car l) old) ( cons new ( cons old ( cdr l))) ) . takes three arguments. cons es the first argument onto the result of cons ing the second argument onto the third argument. and 2. and 2. Because they express what the two differing lines in insertL and insertR express.Which pieces differ? The second lines differ from each other.

Earlier we would probably have written (define insertL ( insert-g seq)) where seq is seqL and (define insertR ( insert-g seq)) where seq is seqR. Not really. (define insertL ( insert-g (lambda ( new old l) ( cons new ( cons old l))))) 132 Chapter 8 . (define insertR ( insert-g seqR)) Is there something unusual about these two definitions? Yes. Is it necessary to give names to seqL and seqR Define insertL again with insert-g Do not pass in seqL this time. We could have passed their definitions instead. using "where" is unnecessary when you pass functions as arguments. But.Try to write the function insert-g of one argument seq which returns insertL where seq is seqL and which returns insertR where seq is seqR (define insert-g (lambda (seq) {lambda ( new old l) ( c ond ( ( null? l) (quote ())) ( ( eq ? ( car l) old) (seq new old ( cdr l))) (else ( cons ( car l) ( ( insert-g seq) new old ( cdr l)))))))) Now define insertL with insert-g (define insertL ( insert-g seqL) ) And insertR .

Just the answer of the second cond-line is different . Here is one. because you do not need to remember as many names. Do you remember the definition of subst (define subst (lambda ( new old l) (cond ( ( null ? l) (quote ( ) ) ) ( ( eq ? ( car l) old ) ( cons new ( cdr l))) (else ( cons ( car l) ( subst new old ( cdr l))))))) Does this look familiar? Yes. What role does # f play? (define yyy (lambda ( a l) ( ( insert-g seqrem) #f a l))) where ( define seqrem (lambda ( new old l) l)) Lambda the Ultimate 133 . You can ( rember June-name "your-mind" ) where June-name is seqL.Is this better? Yes. What do you think about this? Define a function like seqL or seqR for subst (define seqS (lambda ( new old l) ( cons new l))) And now define subst using insert-g (define subst ( insert-g seqS)) And what do you think yyy is Surprise! It is our old friend rember Hint : Step through the evaluation of ( yyy a l) where a is sa usage and l is ( pizza with sa usage and bacon) . i t looks like insertL o r insertR .

What you have j ust seen is the power of abstraction. x . we have even seen functions with similar lines. Have we seen similar functions before? Yes. The Nint h Commandment Abstract common patterns with a new function. and t . Takes one argument x and 2. returns the function + if ( eq? x (quote + ) ) returns the function x if ( eq ? x (quote x )) and returns the function t otherwise? (define atom-to-function (lambda (x) ( cond ( ( eq? x (quote +)) + ) ( ( eq ? x (quote x )) x ) ( else t ) ) ) ) 134 Chapter 8 . Can you write the function atom-to-function which: 1 . Do you remember value from chapter 6? (define value (lambda ( nexp) (cond ( ( atom ? nexp) nexp) ( ( eq ? ( operator nexp) (quote +)) ( + ( value ( 1st-sub-exp nexp)) ( value ( 2nd-sub-exp nexp)))) ( ( eq? ( operator nexp) (quote x ) ) ( x ( value ( 1st-sub-exp nexp)) ( value ( 2nd-sub-exp nexp)))) (else ( t ( value ( 1st-sub-exp nexp)) ( value ( 2nd-sub-exp nexp) )))))) Do you see the similarities? The last three answers are the same except for the + .

No problem . (define value (lambda ( nexp) (cond ( ( atom ? nexp) nexp) (else ( ( atom-to-function ( operator nexp)) ( value ( 1 st-sub-exp nexp)) ( value ( 2nd-sub-exp nexp ) ) ) ) ) ) ) Is this quite a bit shorter than the first version? Time for an apple? Here is multirember again. We haven't changed its meaning. Lambda the Ultimate 135 . Yes.What is ( atom-to-function ( operator nexp)) where nexp is ( + 5 3) Can you use atom-to-function to rewrite value with only two cond-lines? The function + . not the atom + . (define multirember (lambda ( a lat) ( cond ( ( null? lat) (quote ())) ( ( eq? ( car lat) a) ( multirember a ( cdr lat))) (else ( cons ( car lat) ( multirember a ( cdr lat))))))) Write multirember-f (define multirember-f (lambda ( test ?) (lambda ( a lat) ( c ond ( ( null? lat) (quote ( ) ) ) ( ( test ? a ( car lat )) ( ( multirember-f test ?) a ( cdr lat ))) (else ( cons ( car lat) ( ( multirember-f test?) a ( cdr lat) ) ) ) ) ) ) ) What is ( ( multirember-f test?) a lat) where test? is eq? a is tuna and lat is (shri m p sa lad tuna sa lad and tuna) (shri m p sa lad sa lad and ) . but that 's okay. One a day keeps the doctor away. O f course.

Wasn't that easy? Define multirember-eq ? using multirember-f Yes. (quote + ) . 0. (quote more. test? always stands for eq?. and many 136 Chapter 8 . The new test? takes one argument and compares it to tuna. Do we really need to tell multirember-f about tuna Does test ? change as multirember-f goes through lat Can we combine a and test? As multirember-f visits all the elements in lat . No. (define multirember-eq? ( multirember-f test?) ) where test? i s eq ?. and here is a different way: How would it do that? Here is one way to write this function. Well. Yes. just as a always stands for tuna. where k is tuna Can you think of a different way of writing this function? Have you ever seen definitions that contain atoms? Yes. it always looks for tuna. I (define <q ?-tuno ( eq?-c k)) (define eq ?-tuna ( eq?-c (quote tuna ) ) ) . test? could be a function of just one argument and could compare that argument to tuna. x ) .

multiremberT takes a function like eq?-tuna and a lat and then does its work. Now that looks really complicated! (define multirember&co (lambda ( a lat col) (cond ( ( null? lat) ( col ( quote ()) (quote ()))) ( ( eqP ( car lat) a) ( multirember&co a ( cdr lat) (lambda ( newlat seen) ( col newlat ( cons ( car lat) seen))))) ( else ( multirember&co a ( cdr lat) (lambda ( newlat seen) ( col ( cons ( car lat) newlat) seen) ) ) ) ) ) ) Lambda the Ultimate 137 . It 's not bad. This is not really difficult.Perhaps we should now write multiremberT which is similar to multirember-f Instead of taking testP and returning a function. (define multiremberT (lambda ( test ? lat) (cond ( ( null ? lat) (quote ( ) ) ) ( ( test ? ( car lat) ) ( multiremberT test ? ( cdr lat))) (else ( cons ( car lat) ( multiremberT test? ( cdr lat) ) ) ) ) ) ) What is ( multiremberT test ? lat) where test? is eq ?-tuna and lat is (shri m p salad tuna salad and tuna) Is this easy? How about this? (shri m p salad salad a nd ) .

The first one is clearly tuna . What is the value of ( multirember&co a lat col) where a is tuna lat is () and col is a-friend This is not simple. It is a function that takes two arguments and asks whether the second one is the empty list. The name col is short for "collector." A collector is sometimes called a "continuation. it is simpler. What is the value of ( multirember&co a lat col) where a is tuna lat is (strawberries tuna and swordfish) and col is a-friend So let 's try a friendlier example. Then it recurs on ( ) . The third argument is a new function. because a-friend is immediately used in the first answer on two empty lists.Here is something simpler: (define a-friend (lambda ( a: y) ( null? y))) Yes. And what is ( multirember&co a lat col) where a is tuna lat is (tuna) and col is a-friend What are the other arguments that multirember&co uses for the natural recursion? What is the name of the third argument? Do you know what col stands for? multirember&co asks ( eqP ( car lat) (quote tuna )) where lat is (tuna). col." 138 Chapter 8 . It ignores its first argnment. and a-friend makes sure that its second argument is empty. #t .

Which collector is this? How does a-friend differ from new-friend new-friend uses a-friend on the empty list and the value of ( cons (quote tuna) (quote ( ) ) ) . we can : (define new-friend (lambda ( newlat seen ) ( a-friend newlat ( cons (quote tuna) seen)))) And now? multirember&co finds out that ( null ? lat ) is true. which means that it uses the collector on two empty lists. Yes. because its second argument is (tuna ) . I t answers #f. It is new-friend. And what does the old collector do with such arguments? What is the value of ( multirember&co a lat a-friend) where a is tuna and lat is (and tuna) (define latest-friend (lambda ( newlat seen ) ( a-friend ( cons (quote and) newlat) seen))) Lambda the Ultimate 139 . which is not the empty list . This time around multirember&co recurs with yet another friend.Here is the new collector: (define new-friend (lambda ( newlat seen) ( col newlat ( cons ( car lat) seen)))) where ( car lat) is tuna and col is a-friend Can you write this definition differently? Can we also replace col with a-friend in such definitions because col is to a-friend what ( car lat) is to tuna Do you mean the new way where we put tuna into the definition? (define new-friend (lambda ( newlat seen) ( col newlat ( cons (quote tuna) seen)))) where col is a-friend.

but we have seen foreign foods before. 140 Chapter 8 . Finally.And what is the value of this recursive use of multirember&co #f. the others for which the answer is true are collected in a second list ls2 . because ls contains three things that are not tuna. and therefore last-friend is used on (strawberries and swordfish) and (tuna ) . The Tent h Commandment Build functions to collect more than one value at a time. 3. It looks at every atom of the lat to see whether it is eq ? to a. Those atoms that are not are collected in one list ls1 . it determines the value of (! ls1 ls2 ) . since ( a-friend ls1 ls2) where ls1 is (and) and ls2 is (tuna) is #f. What does ( multirember&co a lat f) do? Final question: What is the value of ( multirember&co (quote tuna) ls col) where ls is (strawberries tuna and swordfish) and col is (define last-friend (lambda ( x y) ( length x ) ) ) Yes! It's a strange meal.

(define multiinsertL (lambda ( new old lat) (cond ( ( null? lat) (quote ())) ( ( eq ? ( car lat) old ) ( cons new ( cons old ( multiinsertL new old ( cdr lat))))) (else ( cons ( car lat) ( multiinsertL new old ( cdr lat) ) ) ) ) ) ) D o you also remember multiinsertR Now try multiinsertLR Hint: multiinsertLR inserts new to the left of oldL and to the right of oldR in lat if oldL are oldR are different . (define multiinsertLR (lambda ( new oldL oldR lat) (cond ( ( null? lat) ( quote ( ) ) ) ( ( eq? ( car lat) oldL) ( cons new ( cons oldL ( multiinsertLR new oldL oldR ( cdr lat ) ) ) ) ) ( ( eq ? ( car lat) oldR) ( cons oldR ( cons new ( multiinsertLR new oldL oldR ( cdr lat ) ) ) ) ) (else ( cons ( car lat) ( multiinsertLR new oldL oldR ( cdr lat ) ) ) ) ) ) ) The function multiinsertLR&co i s to multiinsertLR what multirember&co is to multirember Yes. Lambda the Ultimate 141 . No problem.Here is an old friend. and what kind of argument is it? Does this mean that multiinsertLR&co takes one more argument than multiinsertLR? It is a collector function. (define multiinsertR (lambda ( new old lat) (cond ( ( null ? lat) (quote ())) ( ( eq ? ( car lat) old ) ( cons old ( cons new ( multiinsertR new old ( cdr lat ) ) ) ) ) (else ( cons ( car lat) ( multiinsertR new old ( cdr lat ) ) ) ) ) ) ) This i s a way of combining the two functions.

the first is the lat that multiinsertLR would have produced for ( cdr lat ) . So what is the value of ( multiinsertLR &co ( quote cranberries) ( quote fish ) ( quote chi ps) ( quote ()) col) Is it true that multiinsertLR &co will use the new collector on three arguments when ( car lat) is equal to neither oldL nor oldR Yes. . . It is the value of ( col (quote ( )) 0 0) . And this means that 0 occurrences of oldL and 0 occurrences of oldR are found and that multiinsertLR will return () when lat is empty. Can you write an outline of multiinsertLR &co Sure. (define multiinsertLR&co (lambda ( new oldL oldR lat col) (cond ( ( null? lat) ( col ( quote ()) 0 0)) ( ( eq? ( car lat) oldL) ( multiinsertLR&co new oldL oldR ( cdr lat) (lambda ( newlat L R) . and the number of right insertions. . 142 Chapter 8 . ))) ( ( eq ? ( car lat) oldR) ( multiinsertLR&co new oldL oldR ( cdr lat) (lambda ( newlat L R) . and oldR . .When multiinsertLR&co is done. ))) ( else ( multiinsertLR&co new oldL oldR ( cdr lat) (lambda ( newlat L R) . which we cannot determine because we don't know what col is. on the number of left insertions. it will use col on the new lat . respectively. oldL. The second and third are the number of insertions that occurred to the left and right of oldL and oldR . . it is just like multiinsertLR. . )))))) Why is col used on ( quote ( ) ) 0 and 0 when ( null? lat) is true? The empty lat contains neither oldL nor oldR .

it is true. . Lambda the Ultimate 143 . L and R are the correct results for both ( cdr lat) and all of lat .Is it true that multiinsertLR&co then uses the function col on ( cons ( car lat) newlat) because it copies the list unless an oldL or an oldR appears? Why are col's second and third arguments just L and R Yes. Instead of adding one to L. If ( car lat) i s neither oldL nor oldR . it cons es oldR onto the result of cons ing new onto newlat . Here is what we have so far. So. ))) (else ( multiinsertLR&co new oldL oldR ( cdr lat) (lambda ( newlat L R) ( col ( cons ( car lat) newlat) L R))))))) Can you fill in the dots? So can you fill in the dots? Yes. and instead of cons ing new onto cons ing oldL onto newlat . And we have even thrown in an extra collector: (define multiinsertLR&co (lambda ( new oldL oldR lat col) ( cond ( ( null? lat) ( col (quote ( ) ) 0 0)) ( ( eq? ( car lat) oldL) ( multiinsertLR&co new oldL oldR ( cdr lat) (lambda ( newlat L R) ( col ( cons new ( cons oldL newlat)) ( add1 L) R)))) ( ( eq? ( car lat) oldR) ( multiinsertLR&co new oldL oldR ( cdr lat) (lambda ( newlat L R) . we do not need to insert any new elements. it adds one to R. so we know what the new collector for the last case is: (lambda ( newlat L R) ( col ( cons ( car lat) newlat) L R ) ) . . The incomplete collector is similar to the extra collector. the final collector is (lambda ( newlat L R) ( col ( cons oldR ( cons new newlat) ) L ( add1 R ) ) ) .

empty. or . Here is even ? (define even ? (lambda (n) ( = ( x (+ n 2) 2) n))) (define evens-only * (lambda (l) (cond ( ( null ? l) (quote ())) ( ( atom ? ( car l)) (cond ( ( even ? ( car l)) ( cons ( car l) ( evens-only * ( cdr l)))) ( else ( evens-only * ( cdr l))))) (else ( cons ( evens-only * ( car l)) ( evens-only * ( cdr l))))))) What is the value of ( evens-only * l) where l is ( (9 1 2 8) 3 10 ( (9 9 ) 7 6) 2) ((2 8) 10 ( ( ) 6) 2).only * is just an exercise: Do you remember what *-functions are? Now write the function evens-only * which removes all odd numbers from a list of nested lists. evens. Perhaps dessert is sweeter. Now that we have practiced this way of writing functions. Yes. 144 Chapter 8 . . Looks like lots of salt. all *-functions work on lists that are either .an atom consed onto a list .a list consed onto a list.What is the value of ( multiinsertLR&co new oldL oldR lat col) where new is sa lty oldL is fish oldR is chips and lat is (chips and fish or fish and chi ps) Is this healthy? It is the value of ( col newlat 2 2) where newlat is (ch i ps salty and salty fish or sa lty fish and chips salty).

What does the function evens-only *&co do after visiting all the numbers in ( car l) It uses the collector. which we haven't defined yet . Can you explain what ( evens-only *&co ( car l) . . ))))) . and the sum of the odd numbers. . 2 X 8 X 10 X 6 X 2 = 1920. It visits every number in the car of l and collects the list without odd numbers. Lambda the Ultimate 145 . This is full of stars! (define evens-only *&co (lambda ( l col) (cond ( ( null? l) ( col (quote ()) 1 0)) ( ( atom ? ( car l)) fcond ( ( even ? ( car l)) ( evens-only *&co ( cdr l) (lambda ( newl p s ) ( col ( cons ( car l) newl) ( x ( car l) p) s ) ) ) ) (else ( evens-only *&co ( cdr l) (lambda ( newl p s ) ( col newl p ( + ( car l) s ) ) ) ) ) ) ) (else ( evens-only *&co ( car l ) . Here is an outline. ) accomplishes? 9+1+3+9+9+7 = 38.What is the sum of the odd numbers in l where l is ( ( 9 1 2 8) 3 10 ( ( 9 9 ) 7 6) 2) What is the product of the even numbers in l where l is ( ( 9 1 2 8) 3 10 ( ( 9 9 ) 7 6) 2) Can you write the function evens-only *&co It builds a nested list of even numbers by removing the odd ones from its argument and simultaneously multiplies the even numbers and sums up the odd numbers that occur in its argument . . the product of the even numbers.

. Does this mean the unknown collector looks roughly like this: (lambda ( al ap as ) ( evens-only *&co ( cdr l) . ) do? The yet-to-be-determined collector is used.And what does the collector do? It uses evens-only *&co to visit the cdr of l and to collect the list that is like ( cdr l). )) And when ( evens-only *&co ( cdr l) . ) is done with its job. what happens then? What does the collector for ( evens-only *&co ( cdr l) . Then it passes these values to the old collector: (lambda ( al ap as ) ( evens-only *&co ( cdr l) (lambda ( dl dp ds ) ( col ( cons al dl) ( x ap dp) ( + as ds ) ) ) ) ) . (define the-last-friend (lambda ( newl product sum) ( cons sum ( cons product newl)))) 146 Chapter 8 . It conses together the results for the lists in the car and the cdr and multiplies and adds the respective products and sums. Yes. . without the odd numbers of course. Does this all make sense now? What is the value of ( evens-only *&co l the-last-friend) where l is ( ( 9 1 2 8) 3 10 ( ( 9 9 ) 7 6) 2) and the-last-friend is defined as follows: (38 1920 (2 8) 10 (() 6) 2) . . . Perfect . . as well as the product of the even numbers and the sum of the odd numbers. . just as before.

Lambda the Ultimate 147 . don't forget the mustard.Whew! Is your brain twisted up now? Go eat a pretzel.

.

3. and Again. . because the third element is grits. ( looking a lat) where a is caviar and lat is (6 2 grits caviar 5 7 3) Were you expecting something different? True enough. Yes. 6. . . and Again. . Here is looking (define looking (lambda ( a lat) (keep-looking a (pick 1 lat) lat) ) ) Write keep-looking ( looking a lat) where a is caviar md lat is (6 2 4 caviar 5 7 3) #t . which does not even resemble caviar. caviar is obviously in lat .Are you in the mood for caviar What is ( looking a lat) where a is caviar and lat is (6 2 4 caviar 5 7 3) Then we must go looking for it. but what is the first number in the lat? And what is the sixth element of lat And what is the seventh element? So looking clearly can't find caviar #f. 7. We did not expect you to know this. . because (keep-looking a 6 lat) has the same answer as (keep-looking a (pick 1 lat) lat) . #t . True enough . caviar is still in lat . 149 . and Again. .

So what do we do? (keep-looking a 7 lat) where a is caviar and lat is (6 2 4 caviar 5 7 3). It is truly unnatural.What is (pick 6 lat) where lat is (6 2 4 caviar 5 7 3) 7. Which is? #t . So what is (keep-looking a 3 lat) where a is caviar and lat is (6 2 4 caviar 5 7 3) It is the same as (keep-looking a 4 lat). Write keep-looking (define keep-looking (lambda ( a sorn lat) (cond ( ( number? sorn) (keep-looking a (pick sorn lat) lat)) (else ( eq ? sorn a))))) Can you guess what sorn stands for? Symbol or number. What is unusual about keep-looking It does not recur on a part of lat. We call this "unnatural" recursion. What is (pick 7 lat) where lat is (6 2 4 caviar 5 7 3) 3. 150 Chapter 9 .

and Again. What do you think the functions we have seen so far are called? They are called total. Does it always get closer to its goal? Sometimes the list may contain neither caviar nor grits. Yes. . Can you define a shorter function that does not reach its goal for some of its arguments? (define eternity (lambda ( x ) ( eternity x ))) For how many of its arguments does eternity reach its goal? None. . That is correct. . and Again. we will never stop looking. Functions like looking are called partial functions. What is ( looking a lat) where a is caviar and lat is (7 1 2 caviar 5 6 3) This is strange! Yes. . (a (b c) ) . .Does keep-looking appear to get closer to its goal? Yes. Is eternity partial? What is (shift x ) where x is ((a b) c) It is the most partial function. it is strange. 151 . . if we start looking in (7 2 4 7 5 6 3). What happens? We keep looking and looking and looking . . A list may be a tup. and Again. from all available evidence. and this i s the most unnatural recursion possible. .

Define shift This i s trivial. it's not even recursive! (define shift (lambda (pair) ( build (first (first pair)) ( build (second (first pair)) ( second pair))))) Describe what shift does. Why are we not guaranteed that align makes progress? In the second c ond. 152 Chapter 9 ." Now look at this function: (define align (lambda (pora) (cond ( ( atom? pora) pora) ( ( a-pair? (first pora) ) ( align (shift pora))) (els e ( build (first pora) ( align (second pora ) ) ) ) ) ) ) What does i t have i n common with keep-looking Both functions change their arguments for their recursive uses but in neither case is the change guaranteed to get us closer to the goal.What is (shift x ) where x is ( (a b) ( c d ) ) (a ( b (c d ) ) ) . Which commandment does that violate? The Seventh Commandment . Here are our words: "The function shift takes a pair whose first component is a pair and builds a pair by shifting the second part of the first component into the second component.line shift creates an argument for align that is not a part of the original argument .

How much more attention should we pay to the first component? At least twice as much. . . The first component of a pair becomes simpler. and Again. there is. and Again. There may b e arguments for which it keeps aligning things. though the second component becomes more complicated. . and Again. 153 . . Is there something else that changes about the arguments to align and its recursive uses? Yes. In what way is the first component simpler? It is only a part of the original pair's first component .Is the new argument at least smaller than the original one? It does not look that way. Doesn't this mean that length * is the wrong function for determining the length of the argument? Can you find a better function? A better function should pay more attention to the first component. No problem : Can you write a function that counts the number of atoms in align 's arguments? (define length * (lambda (pora) (cond ( ( atom ? pora) 1 ) (else ( + ( length * (first porn) ) ( length * (second porn ) ) ) ) ) ) ) Is align a partial function? We don 't know yet. . Why not? The function shift only rearranges the pair it gets. And? Both the result and the argument of shift have the same number of atoms. .

(define weight* (lambda (pora) (cond ( ( atom ? pora) 1) (else ( + ( x ( weight* (first pora) ) 2) ( weight* (second pora ) ) ) ) ) ) ) What i s ( weight * x ) where x is ( ( a b ) c ) And what is ( weight* x ) where x is ( a (b c ) ) 7. it yields a value for every argument . the weight * 's of align's arguments become successively smaller.Do you mean something like weight* That looks right . Is align a partial function? No. 5. 154 Chapter 9 . instead of shift : The functions shujjle and revpair swap the components of pairs when the first component is a pair. (define shujjle (lambda (pora) (cond ( ( atom ? pora) pora) ( ( a-pair? (first pora)) ( shujjle ( revpair pora))) (else ( build (first pora) ( shujjle ( second pora) ) ) ) ) ) ) Does this mean that shujjle i s total? We don't know. Does this mean that the arguments get simpler? Yes. Here is shujjle which is like align but uses revpair from chapter 7.

What is the value of ( shuffte x) where x is (( a b) ( c d ) ) To determine this value. Lothar Collatz ( 1910-1990) .Let 's try it . let 's try something interesting. Doesn't this mean that we need to know the value of ( shuffte ( revpair pora) ) where (revpair pora) is (( a b) ( c d ) ) Yes. Thank you. And? The function shuffte is not total because it now swaps the components of the pair again. . which means that we start all over. . but otherwise nobody knows. and Again. (shuffte x) where x is (a b) ( a b). Is this function total? (defi ne C (lambda (n) (co nd ( ( one ? n) 1 ) (else (cond (( even ? n) ( C (+ n 2 ) ) ) (else ( C ( add1 ( x 3 n)))) ) ) ) ) ) It doesn 't yield a value for 0. and Again. 155 . we do. we need to find out what ( shuffte ( revpair pora)) is where pora is ((a b) ( c d ) ) . . . Okay. . What is the value of (shuffte x) where x is (a (b c )) ( a ( b c )) . And how are we going t o d o that? We are going t o determine the value of ( shuffte pora) where pora is (( c d ) ( a b) ) . . and Again.

there is no answer. 3. Yes. do not necessarily decrease for the recursion. Does A always give an answer? Then what is ( A 4 3) What does that mean? But answer came there noneAnd this was scarcely odd. The Walrus and The Carpenter -Lewis Carroll 156 Chapter 9 . The page that you are reading now will have decayed long before we could possibly have calculated the value of ( A 4 3). (A 1 1) ( A 2 2) Here is the definition of A (define A (lambda ( n m) (cond ( ( zero ? n) ( addl m)) ( ( zero ? m) ( A ( subl n) 1)) (else ( A (subl n) ( A n ( subl m ) ) ) ) ) ) ) What does A have i n common with shuffie and looking How about an example? A 's arguments. And that means we need the value of ( A 0 3). Wilhelm Ackermann (1853-1946) .What is the value of ( A 1 0) 2. 7. because They 'd eaten every one. it is total. That 's easy: ( A 1 2) needs the value of ( A 0 ( A 1 1 ) ) . like shuffie 's and looking's. For all practical purposes. Thank you .

and Again. That would simplify it a lot . Here is the beginning of this function: What does it do? (define will-stop ? (lambda U ) . and Again. What is the value of ( will-stop ? f ) where f is length We know that ( length l) is 0 where l is ( ) . A function can work for many different arguments. . 157 . .Wouldn't it be great if we could write a function that tells us whether some function returns with a value for every argument ? It sure would. and Again. . It sounds complicated. Then let's make it simpler. . Then let's make up some examples. . . we should have some tool like this around. For a warm-up exercise. Is will-stop ? total then? Yes. . Here is the first one. . Now that we have seen functions that never return a value or return a value so late that it is too late. let 's focus on a function that checks whether some function stops for just the empty list . Okay. depending on whether the argument stops when applied to ( ) . let's write it . I t always returns #t o r #f. i t i s . )) Can you fill in the dots? Does will-stop ? return a value for all arguments? That's the easy part : we said that it either returns #t or #f. the simplest of all arguments. So? Then the value o f ( will-stop ? length) should be #t .

we must determine the value of (and ( will-stop ? last-try) ( eternity (quote ( ) ) ) ) . Okay. 158 Chapter 9 . Do we need more examples? Perhaps we should do one more example. We just saw that. We said that the value of ( will-stop ? last-try) was #f. ) is always #f. since (and #f . here is a function that could be an interesting argument for will-stop ? What does it do? (define last-try (lambda ( x ) (and ( will-stop ? last-try) ( eternity x )))) What i s ( will-stop ? last-try) We need to test it on () If we want the value of ( last-try (quote ())) . . How about another example? What is the value of ( will-stop ? eternity) ( eternity (quote ())) doesn 't return a value. right? Yes. What is the value of (and ( will-stop ? last-try) ( eternity (quote ( ) ) ) ) That depends on the value of ( will-stop ? last-try) . . then (and #f ( eternity (quote ()))). it did. There are only two possibilities. it did. it does. is #f. But didn't will-stop ? predict just the opposite? Yes. Does this mean the value of ( will-stop ? eternity) is #f Yes. which really means that last-try will not stop. Let 's say ( will-stop ? last-try) is #f Okay.Absolutely. So ( last-try (quote ( ) ) ) stopped.

Thank you . It makes will-stop ? the first function that we can describe precisely but cannot define in our language. But it cannot-due to the very definition of what will-stop ? is supposed to do.So we must have been wrong about ( will-stop ? last-try ) That 's correct . We said it was total. and Again. Now we j ust need to determine the value of (and #t ( eternity (quote ( ) ) ) ) . which i s the same as the value of ( eternity (quote ( ) ) ) . . because will-stop ? always gives an answer. . It must return #t . then ( will-stop ? last-try ) must yield either #t or #f. . . This must mean that will-stop ? cannot be defined. . But that means we were wrong again! True. it is. Turing ( 1 9 1 2-1954) and Kurt Godel ( 1906-1978). Is this unique? Is there any way around this problem? No. Fine. We know that it doesn't stop. since this time we said that ( will-stop ? last-try ) was #t . and Again. What is (define . Alan M. ) doesn't work for will-stop ?. What do you think this means? Here is our meaning: "We took a really close look at the two possible cases. If we can define will-stop ?." Yes. If ( will-stop ? last-try ) is #t what is the value of ( last-try (quote ( ) ) ) What i s the value o f ( eternity (quote ( ))) It doesn't have a value. ) This is an interesting question . 159 . . and Again. . . . We j ust saw that (define . .

) nothing. take a deep breath. It just won't give any answer for non-empty lists. we could try the following. .tY list and nothing else. If we give eternity an argument. . (lambda (l) (c o nd ( ( null ? l) 0) (else ( add1 ( eternity ( cdr l)))))) What happens when we use it on a non-empty list? What does it mean for this function that looks like length Suppose we could name this new function. and plunge forward when you're ready. lengtho because the function can only determine the length of the empty list . Is this the function length (define length (lambda (l) (cond ( ( null ? l) 0) (else ( add1 ( length ( cdr l))))))) What i f we didn't have (define . What would be a good name? No answer. . ) anymore? Could we still define length What does this function do? Without (define .So what are recursive definitions? Hold tight. It sure is. It determines the length of the emi. Well. . and especially not the body of length . How would you write a function that determines the length of lists that contain one or fewer items? (lambda (l) (cond ( ( null? l) 0) (else ( add1 ( lengtho ( cdr l)))))) 160 Chapter 9 . could refer to length . it gives no answer.

(lambda (l) (cond ( ( null ? l) 0) (else ( add1 ( (lambda (l) (cond ( ( null? l) 0) (else ( add1 ( eternity ( cdr l ) ) ) ) ) ) ( cdr l)))))) And what's a good name for this function? Is this the function that would determine the lenghts of lists that contain two or fewer items? That 's easy: length9 . Yes. . . (lambda (l) (cond ( ( null? l) 0) (else ( add1 ((lambda ( l) (cond ( ( null? l) 0) (else ( add1 ((lambda ( l) (cond ( ( null? l) 0) (else ( add1 ( eternity ( cdr l)))))) ( cdr l)))))) ( cdr l)))))) Now. . 161 . replace length0 by its definition. and Again. . . . what do you think recursion is? What do you mean? . .Almost. but (define . We just replace eternity with the next version of length . ) doesn't work for lengtho So. and Again. this is length�2 . and Again.

And we still have all these repetitions and patterns in these functions. But we can 't write an infinite function . Do you mean this? Yes. I t creates length0 . we have seen how to determine the length of a list with no items. or it contains one element . What do these patterns look like? All these programs contain a function that looks like length . . . How could we get the function length back? If we could write an infinite function in the style of lengtho . . Yes.Well. . a list is either empty. . ( ( lambda ( length) (lambda ( l ) (cond ( ( null? l) 0) (else ( add1 ( length ( cdr l))))))) eternity) 162 Chapter 9 . which would determine the length of all lists that we can make. . length 9 . . or 100 1 . . and so on. . Perhaps we should abstract out this function : see The Ninth Commandment . or three. that 's okay. . length 9 . . we do. then we could write length00 . How long are the lists that we can make? Well. ) . we can't. with no more than one item . . . Let 's do it ! We need a function that looks just like length but starts with (lambda ( length) . or two elements. or four. No. with no more than two items.

As long as we are consistent . .Rewrite length 9 in the same style. . How about length 9 ( ( lambda ( length ) (lambda (l) (cond ( ( null ? l) 0) (else ( add1 ( length ( cdr l ) ) ) ) ) ) ) ((lambda ( length) (lambda (l) (cond ( ( null ? l) 0) (else ( add1 ( length ( cdr l))))))) ((lambda ( length) (lambda ( l ) (cond ( ( null ? l) 0) (else ( add1 ( length ( cdr l ) ) ) ) ) ) ) eternity) )) Close. everything's okay. True. we just used f and g. Where should we start? Name the function that takes length as an argument and that returns a function that looks like length . and Again. . and Again. but there are still repetitions. and Again. . . 163 . . Let 's get rid of them . ( (lambda (/ ) (lambda (l) (cond ( ( null ? l) 0) (else ( add1 (/ ( cdr l))))))) ((lambda (g) (lambda (l) (cond ( ( null ? l) 0) (else ( add1 (g ( cdr l ) ) ) ) ) ) ) eternity) ) Do we have to use length to name the argument? No.

((lambda ( mk-length) ( mk-length eternity) ) (lambda ( length) (lambda (l) (cond ( ( null? l) 0) (else ( add1 ( length ( cdr l)))))))) Is this length 9 It sure is. 164 Chapter 9 . And this is length9 . ( (lambda ( mk-length) ( mk-length ( mk-length eternity) ) ) (lambda ( length) (lambda (l) (cond ( ( null ? l) 0) (else ( add1 ( length ( cdr l ) ) ) ) ) ) ) ) ((lambda ( mk-length) ( mk-length ( mk-length ( mk-length eternity) ) ) ) (lambda ( length) (lambda (l) (cond ( ( null? l) 0) (else ( add1 ( length ( cdr l)))))))) Can you write length 9 i n this style? Sure. Here it is.What 's a good name for this function? Okay. ((lambda ( mk-length) ( mk-length ( mk-length ( mk-length ( mk-length eternity) ) ) ) ) (lambda ( length) (lambda (l) (cond ( ( null? l) 0) (else ( add1 ( length ( cdr l)))))))) What is recursion like? It is like an infinite tower of applications of mk-length to an arbitrary function. do this to lengtho How about mk-length for "make length" ? No problem.

Do we really need an infinite tower? Not really of course. but we never know how many. and Again. Everytime we use length we only need a finite number. 165 . . . . 1 1 With apologies t o George Orwell ( 1 903. When do we find out that we didn't guess a large enough number? When we apply the function eternity that is passed to the innermost mk-length . how could we do that ? Well. That 's the right idea. Then is this still lengtho ((lambda ( mk-length) ( mk-length mk-length)) (lambda ( length) (lambda (l) (cond ( ( null? l) 0) (else ( add1 ( length ( cdr l)))))))) Yes. we could even use mk-length instead of length . . but we may not guess a large enough number. . What if we could create another application of mk-length to eternity at this point? That would only postpone the problem by one. Could we guess how many we need? Sure. ( (lambda ( mk-length) ( mk-length mk-length )) (lambda ( mk-length) (lambda ( l ) (cond ( ( null? l) 0) (else ( add1 ( mk-length ( cdr l)))) ) ) ) ) Why would we want to do that? All names are equal.1950) . but some names are more equal than others. and Again. And then we invoke mk-length on eternity and the result of this on the cdr so that we get one more piece of the tower. since nobody cares what function we pass to mk-length we could pass it mk-length initially. and Again. . and besides.

True: as long as we use the names consistently. Now that mk-length is passed to mk-length can we use the argument to create an additional recursive use? Yes. and we can do this as often as we need to! 166 Chapter 9 . Could we do this more than once? Yes. If we use a name like mk-length . And mk-length is a far more equal name than length . j ust keep passing mk-length to itself. Work it out with paper and pencil. we are j ust fine. when we apply mk-length once. it is a constant reminder that the first argument to mk-length is mk-length . we get length 9 ( (lambda ( mk-length) ( mk-length mk-length)) (lambda ( mk-length) (lambda (l) (cond ( ( null? l) 0) (else ( add1 ( ( mk-length eternity) ( cdr l)))))))) What is the value of ( ( (lambda ( mk-length ) ( mk-length mk-length)) (lambda ( mk-length) (lambda (l) (cond ( ( null ? l) 0) (else ( add1 ( ( mk-length eternity) ( cdr l )))))))) l) where l is (apples) This is a good exercise.

((lambda ( mk-length) ( mk-length mk-length) ) (lambda ( mk-length) (lambda (l) (cond ((null? l) 0) (else ( add1 ______--. and Again. One problem is left: it no longer contains the function that looks like length We could extract this new application of mk-length to itself and call it length . and Again. . . of course. just as it is about to expire. . I. ( (lambda ( mk-length) ( mk-length mk-length) ) (lambda ( mk-length) (lambda (l) (cond ( ( null? l) 0) (else ( add1 ( ( mk-length mk-length ) ( cdr l )))))))) H ow does i t work? It keeps adding recursive uses by passing mk-length to itself.-mk-length mk-length) I << ( cdr l)))))))) Can you fix that? Why? Because it really makes the function length . . .What would you call this function? It is length . 167 . and Again. .

How about this? Yes. First . this looks just fine. because the value of this expression is the function that we need to apply to l where l is (apples) 168 Chapter 9 . Okay. What is the value of ( ( (lambda ( mk-length) ( mk-length mk-length) ) (lambda ( mk-length) ((lambda ( length) (lambda (l) (cond ( ( null? l) 0) (else ( add1 ( length ( cdr l ) ) ) ) ) ) ) ( mk-length mk-length)))) l) where l is (apples) It should be 1 . we need the value of ((lambda ( mk-length) ( mk-length mk-length)) (lambda ( mk-length) ( (lambda ( length) (lambda (l) (cond ( ( null? l) 0) (else ( add1 ( length ( cdr l))))))) ( mk-length mk-length)))) That 's true. ((lambda ( mk-length) ( mk-length mk-length) ) (lambda ( mk-length) ((lambda ( length) (lambda (l) (cond ( ( null ? l) 0) (else ( add1 ( length ( cdr l))))))) ( mk-length mk-length)))) Let 's see whether it works.

So we really need the value of ((lambda ( mk-length) ((lambda ( length) (lambda (l) (cond ( ( null? l) 0) (else ( add1 ( length ( cdr l))))))) ( mk-length mk-length) ) ) (lambda ( mk-length) ((lambda ( length) (lambda (l) (cond ( ( null? l) 0) (else ( add1 ( length ( cdr l))))))) ( mk-length mk-length)))) But then we really need to know the value of ((lambda ( length) (lambda (l) (cond ( ( null? l) 0) (else ( add1 ( length ( cdr l))))))) ((lambda ( mk-length) ((lambda ( length) (lambda ( l) (cond ( ( null? l) 0) (else ( add1 ( length ( cdr l))))))) ( mk-length mk-length))) (lambda ( mk-length) ((lambda ( length) (lambda (l) (cond ( ( null? l) 0) (else ( add1 ( length ( cdr l))))))) ( mk-length mk-length))))) True enough. Where is the end of this? Don't we also need to know the value of ( (lambda ( length) (lambda (l) (cond ( ( null? l) 0) (else ( add1 ( length ( cdr l))))))) ((lambda ( length) (lambda (l) (cond ( ( null? l) 0) (else ( add1 ( length ( cdr l))))))) ( (lambda ( mk-length) ((lambda ( length) (lambda (l) (cond ( ( null? l) 0) (else ( add1 ( length ( cdr l))))))) ( mk-length mk-length))) (lambda ( mk-length) ( (lambda ( length) (lambda (l) (cond ( ( null? l) 0) (else ( add1 ( length ( cdr l ) ) ) ) ) ) ) ( mk-length mk-length ) ) ) ) ) ) . and Again. and Again. Yes. and Again. . . . that 's true. 169 . too. . .

. there is no end to it . does (lambda ( x ) ( ( mk-length mk-length) x )) return a function of one argument? Actually. No it doesn 't.l ( -k length mk-length ) ( cdr l ) ) ) ) ) ) ) ) Here i s a different way. So what do we do? Turn the application of mk-length to itself in our last correct version of length into a function : How? ((lambda ( mk-length) ( mk-length mk-length) ) (lambda ( mk-length) (lambda (l) (cond ( ( null ? l) 0) (else ( add1 (j'm----__---------. is (lambda ( x ) (f x )) a function of one argument? Yes. (lambda ( x ) ( ( mk-length mk-length) i s a function! x )) 1 70 Chapter 9 . If ( mk-length mk-length) returns a function of one argument . it didn 't matter what we applied it to.Yes. . Indeed. But now that we have extracted ( mk-length mk-length) from the function that makes length it does not return a function anymore. Is this strange? It is because mk-length used to return a function when we applied it to an argument. Why? Because we just keep applying mk-length to itself again and again and again . If f i s a function of one argument . it is.

and Again.--------. . ((lambda ( mk-length) ( mk-length mk-length ) ) (lambda ( mk-length) ( (lambda ( length) (lambda (l) (cond ( ( null? l) 0) (else ( add1 ( length ( cdr l ) ) ) ) ) ) ) (lambda (x) ( ( mk-length mk-length) x ) ) ) ) ) Is it okay to move out the function? Yes. we just always did the opposite by replacing a name with its value.---:. Can we extract the function in the box that looks like length and give it a name? Yes.(-. . ( 1ambd a' x ) ( ( mk-length mk-length) x)) ( cdr l)))))))) Move out the new function so that we get length back. . Here we extract a value and give it a name. it does not depend on mk-length at all! . 171 .Okay. and Again. and Again. .. ((lambda ( mk-length) ( mk-length mk-length) ) (lambda ( mk-length) (lambda (l) (cond ( ( null ? l) 0) (else ( add1 ("'=---. let's do this to the application of mk-length to itself.. .

(define Y (lambda ( le) ((lambda (/ ) (/ f ) ) (lambda (/ ) ( le (lambda ( x ) ( (/ f ) x ))))))) Does (define . 1 72 Chapter 9 . That 's easy.. . Read this chapter just one more time and you will. ( (lambda ( le ) ((lambda ( mk-length) ( mk-length mk-length) ) (lambda ( mk-length) ( le (lambda ( x ) ( ( mk-length mk-length) x )))))) (lambda ( length) (lambda (l) (cond ( ( null? l) 0) (else ( add1 ( length ( cdr l )))))))) What did we actually get back? Let 's separate the function that makes length from the function that looks like length We extracted the original function mk-length... (lambda ( le) ((lambda ( mk-length) ( mk-length mk-length) ) (lambda ( mk-length) ( le (lambda ( x ) ( ( mk-length mk-length) x )))))) Does this function have a name? Yes.----. it is called the applicative-order Y combinator. . now that we know what recursion is. . ) work again? Do you now know why Y works? Sure.Is this the right function? Yes...

and Again. and Again. Does your hat still fit? Perhaps not after such a mind stretcher. 173 . . and Again. . Leslie Bricusse and Anthony Newley . Stop the World-/ Want to Get Off.What is ( Y Y) Who knows. . . but it works very hard. .

:[].(® €:) � &5 � W� � J�SJID � tij Cl1i00� � .

the two lists must be of equal length. Also. Make up some examples for entries. Here are our examples: ( (appetizer entree beverage) ( pate boeuf vi n ) ) and ( (appetizer entree beverage) ( beer beer beer) ) and ( (beverage dessert) ( (food is) ( n u m ber one with us) ) ) . How can we accomplish this? lookup-in-entry takes an additional argument that is invoked when name is not found in the first list of an entry. Why? How many arguments do you think this extra function should take? What Is the Value of All of This ? 1 75 . What is ( lookup-in-entry name entry) where name is entree and entry is ((a ppetizer entree beverage) (food tastes good ) ) What i f name i s dessert In this case we would like to leave the decision about what to do with the user of lookup-in-entry.An entry is a pair of lists whose first list is a set. We think it should take one. tastes. name. How can we build an entry from a set of names and a list of values? (define new-entry build ) Try to build our examples with this function.

. What is ( lookup-in-table name table table-f ) where name is entree table is ( ( ( entree dessert ) ( spaghetti spu moni )) ( ( appetizer entree beverage ) ( food tastes good ) ) ) and table-/ i s ( lambda ( name) . ) ( define extend-table cons) It could be either spaghetti or tastes. So it is spaghetti . Define the function extend-table which takes an entry and a table ( possibly the empty one ) and creates a new table by putting the new entry in front of the old table. Here is one example: the empty table. 176 Chapter 10 . Here is another one: ((( appetizer entree beverage ) ( pate boeuf vin )) (( beverage dessert ) ((food is ) ( n u m ber one with us )))) . but lookup-in-table searches the list of entries in order.Here is our definition of lookup-in-entry (define lookup-in-entry (lambda ( name entry entry-f) ( lookup-in-entry-help name (first entry) (second entry) entry-f))) Finish the function lookup-in-entry-help (define lookup-in-entry-help (lambda ( name names values entry-f) (cond (( null? names) ( entry-/ name) ) ((eq ? ( car names) name) ( car values)) (else ( lookup-in-entry-help name ( cdr names) ( cdr values) entry-f))))) (define lookup-in-entry-help (lambda ( name names values entry-f ) (cond ) ( ( ) )))) ( A table ( also called an environment ) i s a list of entries. . represented by () Make up some others.

as to whether Did you notice that "sans serif" was not in sans serif? We hope so. Henceforth. Have we chosen a good representation for expressions? Yes. is the function that returns Do you remember value from chapter 6? Recall that value the natural value of expressions.Write lookup-in-table Hint : Don't forget to get some help. you must notice whether or not an atom is in sans serif. They are all S-expressions so they can be data for functions . (lambda ( name) (lookup-in-table name ( cdr table} table-f)) In the preface we mentioned that sans serif typeface would be used to represent atoms. This is " a ns s serif' in sans serif. What kind of functions? For example. Remember to be very conscious or not a n atom is i n sans serif. To this point it has not mattered. What is the value of We don 't even know what (quote { a b c) ) is. ( car (quote {a b c) )) What Is the Value of All of This ? 1 77 . value. ( defi ne lookup-in-table (lambda (name table table-f ) ( c o nd (( null? table) ( table-! name)) ( els e ( lookup-in-entry name ( car table} (lambda ( name) (lookup-in-table name ( cdr table} table-f))))))) Can you describe what the following function represents: This function i s the action to take when the name is not found in the first entry.

It is a representation of the expression: (car (quote (a b c))).What is the value of ( cons rep-a ( cons rep-b ( cons rep-c (quote ())))) where rep-a is a rep-b is b and rep-c is c Great . And what is the value of ( cons rep-car ( cons ( cons rep-quote ( cons ( cons rep-a ( cons rep-b ( cons rep-c (quote ())))) (quote ()))) (quote ()))) where rep-car is car rep-quote is quote rep-a is a rep-b is b and rep-c is c What is the value of ( car (quote (a b c) ) ) What i s ( value e ) where e is (car (q uote (a b c))) What is ( value e ) where e is (quote (car (quote (a b c) ))) It is the same as (a b c) . a. (car (q uote (a b c)) ) . a. 178 Chapter 10 .

What is the type of e where e is #f *const. *const. 6. ( (from nothing comes something) ) . nothing has no value. What Is the Value of All of This? 179 . because numbers are constants. What is ( value e) where e is ((lambda (nothing) (cond (nothing (q uote something)) (else (quote nothing)))) #t ) What is the type of where e is 6 e something.What is ( value e) where e is (addl 6 ) What is ( value e) where e is 6 What is ( value e) where e is (quote nothing) What is ( value e) where e is nothing What is ( value e) where e is ((lambda (nothing) (cons nothing (quote ()))) (quote (from nothing comes something) )) 7. nothing.

where e is nothing What is t he type of e where e is (lambda (x y) (cons *lambda. *identifier. *const. 180 Chapter 10 . *cond.What is { value e) where e is # f What is the type of e where e is cons What is ( value e) #f. x y)) What is the type of e where e is ((lambda (nothing) (cond (nothing (quote something)) (else (quote nothing) ))) #t ) What is the type of e where e is (cond (nothing (quote something)) (else (quote nothing))) *application. where e is car What is the type of e where e is {quote nothing) What is the type of e *quote. {primitive car) .

We call these functions "actions. They can be detected by an appropriate function to which S-expressions are submitted before they are passed on to value. (lambda (car) car) . (cond (3 c) (else b) (6 a)). (lambda (#t) #t). atom-to-function useful when we rewrote value for numbered expresssions. We found . How do you think we should represent types? If actions are functions that do "the right thing" when applied to the appropriate type of expression. () . We choose functions." You guessed it. (lambda (5) 5). what should value do? Do you remember atom-to-function from chapter 8? Below is a function that produces the correct action ( or function ) for each possible S-expression: (define expression-to-action (lambda (e) (cond ( ( atom ? e) ( atom-to-action e)) (else (list-to-action e))))) Define the function atom-to-action 1 1 Ill-formed S-expressions such as (quote a b). It would have to find out the type of expression it was passed and then use the associated action . and (1 2) are not considered here.How many types do you think there are? We found six: *const *quote *identifier *lambda *cond and *application. (lambda a). (define atom-to-action (lambda ( e) (cond (( number? e) *const) ( ( eq ? e #t ) *const) ( ( eq ? e #f ) *const) ( ( eq ? e (quote cons )) *const) ( ( eq ? e (quote car )) *const) ( ( eq ? e (quote cd r )) *const) ( ( eq ? e (quote null?)) *const) ( ( eq ? e (quote eq?)) *const) ( ( eq ? e (quote atom?)) *const) ( ( eq ? e (quote zero?) ) *const) ( ( eq ? e (quote add 1 ) ) *const) ( ( eq ? e (quote su b1)) *const) ( ( eq ? e (quote num ber?)) *const) (else *identifier)))) What Is the Value of All of This? 181 .

1 together with all the functions it uses. is called an interpreter. 1 Actions do speak louder than words . the expression e and a table. The function value.Now define the help function list-to-action (define list-to-action (lambda ( e ) (cond (( atom ? ( car e)) (cond ( ( eq ? ( car e) (quote quote )) *quote) ( ( eq ? ( car e) (quote lambda}) *lambda) ( ( eq ? ( car e) (quote cond}} *cond} (else *application))) (else *application)))) Assuming that expression-to-action works. 182 Chapter 10 . we can use it to define value and meaning (define value (lambda ( e ) (meaning e (quote ())))) (define meaning (lambda ( e table) ( ( expression-to-action e) e table))) What is (quote ()) in the definition of value It is the empty table. How many arguments should actions take according to the above? Two. The function value approximates the function eval available in Scheme (and Lisp ) .

(define text-of second) (define *quote (lambda ( e table) ( text-of e))) Define the help function text-of Have we used the table yet? Why do we need the table? Given that the table contains the values of identifiers. For #f . For #t .Here is the action for constants. Why? (define initial-table (lambda ( name) ( car (quote ())))) When is it used? What is the value of (lambda (x ) x ) We don't know yet. but we will in a moment . . . but we know that it must be the representation of a non-primitive function. What Is the Value of All of This ? 183 . (define *identifier (lambda ( e table) (lookup-in-table e table initial-table))) Here is initial-table Let's hope never. To remember the values of identifiers. . write the action *identifier No. it returns true. (define *const (lambda ( e table) ( co nd (( number? e) e) (( eq? e #t ) #t ) ((eq? e #f ) #f ) (else ( build (quote primitive) e))))) Is it correct? Here is the action for *quote Yes. 1 . it just returns the expression. and this is all we have to do for 0. it returns false. for numbers. 2. And all other atoms of constant type represent primitives.

e. . and the body). It is a special form that takes any number of cond-lines. In a list . Otherwise it proceeds to answer the right part . Write table-of formals. It considers each line in turn. At least. the table. And what else do we need to remember? We will also put the table in. non-primitives are defined by their arguments and their function bodies. 184 Chapter 10 . If it sees an else-line. So when we want to use a non-primitive we need to remember its formal arguments and its function body. the formal arguments. .How are non-primitive functions different from primitives? We know what primitives do. just in case we might need it later. If the question part on the left is false. Fortunately this is just the cdr of a lam bda expression. ) in your own words.of and body-of (non-pri mitive ( ( ( (y z) ( (8) 9 ) ) ) table (x) (cons x y ) )) �---v----� � � formals body (define table-of first) (define formals-of second) (define body-of third ) Describe ( cond . And how do we represent this? Here is the action *lambda (define *lambda (lambda ( e table) ( build (quote non-pri m itive) ( cons table ( cdr e ) ) ) ) ) What i s ( meaning e table) where e is (lambda ( x ) (cons x y)) and table is (( (y z) ( (8) 9 ) ) ) I t i s probably a good idea t o define some help functions for getting back the parts in this three element list (i. . it looks at the rest of the lines. it treats that cond-line as if its question part were true. of course.

Do you understand *cond now? How can you become familiar with it? What Is the Value of All of This ? 185 .Here is the function evcon that does what we just said in words: (define evcon (lambda ( lines table ) (cond (( else? ( question-of ( car lines ) ) ) ( meaning ( answer-of ( car lines ) ) table) ) (( meaning ( question-of ( car lines ) ) table) ( meaning ( answer-of ( car lines ) ) table)) (else ( evcon ( cdr lines ) table ) ) ) ) ) Write else? and the help functions question-of and answer-of Didn't we violate The First Commandment? (define else ? (lambda ( x ) (cond ( ( atom ? x ) ( eq ? (else # f ) ) ) ) x (quote else) ) ) (define question-of first) ( define answer-of second) Yes. they make things quite a bit more readable. But you already knew that. A good one is: ( *cond e table) where e is (cond (coffee klatsch) (else party) ) and table is (( (coffee) ( #t ) ) ( ( klatsch party) (5 (6) ) ) ) . we don 't ask ( null ? lines ) . The best way is to try an example. Now use the function evcon to write the *cond action. ( define *cond (lambda ( e table) ( evcon ( cond-lines-of e) table ) ) ) ( define cond-lines-of cdr) Aren't these help functions useful? Yes. so one of the questions in every cond better be true. Perhaps not .

Have we seen how the table gets used? But how do the identifiers get into the table? Yes. Of course. . An application must always determine the meaning of all its arguments. function-of . ) (or . ) or (cond . Then we apply the meaning of the function to the meaning of the arguments. and arguments-of correctly. Yes. .of e) table) ( evlis ( arguments-of e) table)))) Is it correct? 1 86 Chapter 10 . do we have to get the meaning of all of its arguments? Write a function evlis that takes a list of (representations of) arguments and a table. and returns a list composed of the meaning of each argument . (define evlis (lambda ( args table) (cond ( ( null? args ) (quote ())) (else ( cons ( meaning ( car args ) table) ( evlis ( cdr args ) table)))))) What else do we need before we can determine the meaning of an application? And what then? We need to find out what its function-of means. ) Before we can apply a function. . . Here is *application (define *application (lambda ( e table) ( apply ( meaning (function. We just have to define apply. *lambda and *identifier use it . . like (and . An application is a list of expressions whose car position contains an expression whose value is a function . . In the only action we have not defined: *application . How is an application represented? How does an application differ from a special form.

(primitive primitive-name) and (non. there i s n o answer. T h e function apply approximates the function apply available in Scheme ( and Lisp ) .pri mitive ( table formals body) ) The list ( table formals body) i s called a closure record. 1 What Is the Value of All of This ? 187 . Write primitive ? and non-primitive ? (define primitive ? (lambda (l) ( eq ? (first l) (quote pri mitive) ) ) ) (define non-primitive ? (lambda (l) ( eq ? (first l) (quote non-pri m itive) ) ) ) Now we can write the function apply Here i t is: ( define apply 1 (lambda (fun vals ) (cond ( (primitive ? fun) ( apply-primitive (second fun) vals ) ) ( ( non-primitive ? fun) ( apply-closure (second fun) vals ) ) ) ) ) I f fu n does not evaluate to either a primitive o r a non-primitive as in the expression ( ( l a m bda (x) (x 5 ) ) 3).Write function-of and arguments-of (define function-of car) (define arguments-of cdr) How many different kinds of functions are there? What are the two representations of functions? Two: primitives and non-primitives.

5. 2. 188 Chapter 10 . 3. 4. cdr 1 eq ? (second vals ) :atom ? ( quote cons) ( define :atom ? (lambda ( x ) ( cond ( ( atom ? x ) #t ) ( ( null ? x ) #f ) ( ( eq ? ( car x ) ( quote prim itive) ) #t ) ( ( eq ? ( car x ) ( quote non-pri mitive)) #t ) ( else #f ) ) ) ) 1 of The function apply·primitive could check for applications cdr to the empty list or sub1 to 0. But we know what to do to find the meaning of (cons x y) where table is ( ( (x y) (1 (2) ) ) ) . 1. Here. and apply-closure must extend the table.This is the definition of apply-primitive (define apply-primitive (lambda ( name vals ) (cond (( eq ? name 1 ) ( cons (first vals ) (second vals ) ) ) ( ( eq ? name (quote car) ) ( car (first vals ) ) ) ( ( eq? name (quote cd r) ) ( 2 (first vals ) ) ) ( ( eq ? name (quote null?)) ( null ? (first vals ) ) ) ( ( eq ? name (quote eq?)) ( 3 (first vals ) 4 )) ( ( eq ? name (quote atom ?)) (first vals ) ) ) ( 5 ( ( eq ? name (quote zero?)) (zero ? (first vals ) ) ) ( ( eq ? name (quote add 1 ) ) ( add1 (first vals ) ) ) ( ( eq ? name (quote su b1)) (sub1 (first vals ) ) ) ( ( eq ? name (quote n u m ber?)) ( number? (first vals ) ) ) ) ) ) Fill i n the blanks. etc. That 's tricky. Is apply-closure the only function left ? How could we find the result of (f a b) where f is (lambda (x y) (cons x y)) a is 1 and b is (2) Why can we do this? Yes. we don 't need apply-closure .

Can you generalize the last two steps? Applying a non-primitive function-a closure-to a list of values is the same as finding the meaning of the closure's body with its table extended by an entry of the form (formals values ) In this entry. I n the following. The new e for meaning will b e ( cons z x ) and the new table for meaning will be ( ( (x y) ((a b c ) (d e f) ) ) What will be the new arguments of meaning ((u v w ) (1 2 3)) ( (x y z) (4 5 6))). closure is (((( u v w ) ( 1 2 3)) ( (x y z) (4 5 6) ) ) (x y) ( cons z x ) ) and vals is ( (a b c ) (d e f) ) . Have you followed all this? (define apply-closure (lambda ( closure vals ) ( meaning ( body-of closure) ( extend-table ( new-entry (formals-of closure) vals ) ( table-of closure ) ) ) ) ) This is a complicated function and it deserves an example. here is the definition of apply-closure . formals is the formals of the closure and values is the result of evlis . What Is the Value of All of This ? 189 . If not .

by using *identifier. because evlis returns a list of the meanings.What is the meaning of ( cons where z is 6 and x is (a b c ) z x ) The same as ( meaning e table) where e is ( cons z x) and table is ( ( (x y) ((a b c ) (d e f) )) ( ( u v w) ( 1 2 3)) ( (x y z ) (4 5 6) ) ) . we must find both ( meaning e table) where e is z and ( meaning e table) where e is x. cons . what is the result of evlis 6. Let 's find the meaning of all the arguments. What is ( meaning e table) where e is cons ( pri m itive ) by using *const. ( a b c ) . by using *identifier. 190 Chapter 10 . What is ( evlis args table) where args is ( z x) and table is ( ( (x y) ( (a b c) (d e f) ) ) ( ( u v w) ( 1 2 3)) ( (x y z ) (4 5 6) ) ) I n order t o d o this. (6 ( a b c )) . What i s the ( meaning e table) where e is z What is ( meaning e table) where e is x So.

. That may help a lot . . . don't bother-we can play the same game again. but see The Seasoned Schemer.We are now ready to ( apply fun vals ) where fun is ( pri mitive cons ) and vals is (6 (a b c )) Which path should we take? Which cond-line is chosen for ( apply-primitive name vals ) where name is cons and vals is (6 (a b c ) ) Are we finished now? But what about (define . Yes. we are exhausted. else What Is the Value of All of This ? 191 . Yes. The third: ( ( eq? name (quote cons )) ( cons (first vals ) (second vals ) ) ) . Yes. It isn't needed because recursion can be obtained from the Y combinator . Is (define . it 's time for a banquet . We would be able to define a function like last-try ? that will show that we cannot define the new and improved will-stop ?. ) The apply-primitive path. No. but don't bother. ) really not needed? Does that mean we can run the interpreter on the interpreter if we do the transformation with the Y combinator? What makes value unusual? Should will-stop ? see representations of expressions? Does it help? It sees representations of expressions. . Yes.

.