Introduction to List Processing and Functional Programming and Scheme
- Review recursion and lists.
- Chapter 10 (sec. 1-3) of the text (Functional Programming and Scheme)
- The scheme reference manual available with
Introduction to functional programming using scheme (a dialect of lisp).
In functional programming, programs are treated as function (for every input
there is a unique output. In pure functional programming there are no
variables and hence no assignment or side effects. There are function
definitions and function applications and the value of a function only
depends on the values of its arguments (referential transparency).
Furthermore, the value of a function can not depend on the order in which
its arguments are evaluated. A consequence of not having variables is
that there are no loops; repeated operations and other control flow is
obtained through recursion. Finally, functions are treated as first class
values, i.e. functions are viewed as values which
can be computed by other functions and passed as parameters to functions.
The lack of assignment and referential transparancy make the semantics of
functional programs straightforward. There is no state, there is no
concept of memory locations and pointers, only bindings of names to values -
once a name enters the environment its value can never change. Such semantics
is called value semantics. While this seems like a very restricted
type of programming, it can be shown to be turing complete and hence,
theoretically, as powerful as programs with loops and variables.
In practice functional languages, like lisp, scheme, and ML, are not pure;
however, they are conducive to techniques that arise in pure functional
programming and consequently provide the programmer with a different, and at
times very useful, view of programming.
- S-Expressions and Evaluation Rules
Programs and data are the same - in scheme they are lists - the
only difference is how they are interpreted. Given a list,
S-Expression, (E1 E2 ... En), its value is obtained by recursively
evaluating the S-Expressions E1, E2,...,En and applying the value of
E1, which should be a function, with its arguments bound to the values
of E2,...,En. In the base case, atoms, evaluate to themselves. This
evaluation process is called applicative normal order. For example
- (+ 2 (* 3 4)) => (+ 2 12) => 14 since + and * evaluate to the
addition and multiplication functions and the numbers evaluate to
their numeric values.
- (max 1 2 3) => 3, since max is bound to the max function.
- (1 2 3) => error since 1 does not evaluate to a function.
- (list 1 2 3) => (1 2 3), since list is bound to a function that
forms a list from its arguments.
- (list 1 (2 3) 4) => error since recursive (2 3) evaluates to an
error due to the previous problem.
- (list 1 (list 2 3) 4) => (1 (2 3) 4)
- List processing functions
Syntactically lists are enclosed in parentheses with elements separated
by spaces - (a1 ... am). Elements of a list may be lists themselves.
The null list is denoted by (). The null list is detected by the
predicate null? - a predicate is a function that returns true or
false (in scheme denoted by #t and #f respectively). Note that the
null list is considered both a list and an atom and evaluates to
Given a set of atoms, all lists, of
arbitrary order (i.e. nesting) can be built from () and the function
(cons x y), which constructs a list whose first element is x and
remaining elements are the elements of y. Technically y need not be
a list (to be discussed below). The first element of a list is
obtained with the function car and the remaining elements with the
Many other list processing functions are available (see chapter 7 of the
scheme reference manual that is part of the MIT-Scheme distribution):
E.G. append, length and reverse. These functions, and many others,
can be built from the primitive list processing functions car, cdr,
null? and cons using recursion.
- Let L = (1 2 3), this is accomplished with (define L (list 1 2 3)).
the define expression binds the value of the expression given in
the second argument to the name given in the first argument.
- (cons 3 ()) => (3)
- (cons 1 (cons 2 (cons 3 ()))) => (1 2 3)
- (car L) => 1
- (cdr L) => (2 3)
- (null? L) => #f
- (null? ()) => #t
- Lambda expressions
A function in scheme is denoted by (lambda name (formal-parameters) body),
and is called a lambda expression. Lambda expressions evaluate to
a function which can then be applied.
- (define sqr (lambda (x) (* x x))) binds the name sqr
to the function that computes the square of the parameter x.
- Since it is common to define a name to a lambda expression, the
following short hand exists. (define (sqr x) (* x x)).
- Conditional expressions
- The if expression provides a mechanism to return different values
based on the value of a boolean expression. The expression
(if cond val1 val2) returns val1 if cond
evaluates to #t and val2 if cond evaluates
- (if (< 2 3) 0 1) returns 0
- (if (< 3 2) 0 1) returns 1
- The cond expression selectively returns a value based on
a sequence of boolean expressions
(cond ((cond_1) val_1) ... ((cond_n) val_n)) is evaluated
sequentially. First cond_1 is evaluated and if true,
val_1 is returned. If false, then cond_2 is evaluated
and if true cond_2 is returned. This process is continued
until an expression evaluates to true. If none of the expressions
is true, then what is returned is undefined. Note that the keyword
else can be used for a condition and it always evaluates to true.
Since there are no side effects in functional programming, functions
can be thought of as mathematical definitions and the use of recursive
definitions and recursion allows us to define and hence implement
many useful functions. Moreover, without side effects, there can be
no loops since loops require a loop index which is incremented. Hence,
in a pure functional language, all control must be done by recursion.
Note that scheme is not a pure functional language, i.e. there are
variables, however, we will focus on the functional programming style
and use recursion for control.
- The following recursive function computes n factorial.
(define fact (lambda (n) (if (= n 0) 1 (* n (fact (- n 1))))))
- The following recursive function returns a list of n ones.
(define ones (lambda (n) (if (= n 0) () (cons 1 (ones (- n 1))))))
- Higher-order functions
Functions that take functions as parameters and functions that
- sort: sort a list of objects using a specified comparison function
- (sort '(4 3 2 1) <) => (1 2 3 4)
- (sort '("one" "two" "three" "four") string< ?) =>
("four" "one" "three" "two")
- map: map a function over the elements of a list or lists. Given
a function f of t arguments, (map f list_1 ... list_t), where
list_1 = (a_11 ... a_1n), ..., list_t = (a_t1 ... a_tn), produces
the list ((f a_11 ... a_1n) ... (f a_1n ... a_tn)). Examples:
- (define sqr (lambda (n) (* n n)))
- (map sqr '(1 2 3 4)) => (1 4 9 16)
- (map (lambda (n) (* n n)) '(1 2 3 4)) => (1 4 9 16)
- (map list '(1 2 3 4) '(1 4 9 16)) => ((1 1) (2 4) (3 9) (4 16))
- filtering lists - remove/retain the elements of a list satisfying
a given predicate.
- (keep-matching-items '(1 2 3 4 5) odd?) => (1 3 5)
- (delete-matching-items '(1 2 3 4 5) odd?) => (2 4)
- reduce: combines the elements of a list using a given binary
procedure. The combination procedes in a left associative order.
(reduce f initial '(1 2 3 4)) => (f 4 (reduce f initial '(1 2 3)))
=> (f 4 (f 3 (f 2 1))). When the input is empty, initial is returned.
reduce-right is the same as reduce, except a right associative order
The fold-left and fold-right commands are like reduce and reduce-right,
except initial is always used. (fold-left f initial '(1 2 3 4)) =>
(f (fold-left f initial '(1 2 3)) 4) =>
(f (f (f (f ininitial 1) 2) 3) 4). fold-right is the right associative
- (reduce + 0 '(1 2 3 4)) => 10
- (reduce * 1 '(1 2 3 4)) => 24
- (reduce list () '(1 2 3 4)) => (4 (3 (2 1)))
- (reduce-right list () '(1 2 3 4)) => (1 (2 (3 4)))
- (fold-left list () '(1 2 3 4)) => ((((() 1) 2) 3) 4)
- (define (length list)
(fold-left (lambda (sum element) (+ sum 1)) 0 list))
- (define (reverse items)
(fold-left (lambda (x y) (cons y x)) () items))
- function composition
- (define (compose g f) (lambda (x) (g (f x))))
- (define cadr (compose car cdr))
- (define caddr (compose car (compose cdr cdr)))
- (define cadar (compose car (compose cdr car)))
- (cadr '(1 2 3)) => 2
- (caddr '(1 2 3)) => 3
- (cadar '((1 2) 3) => 2
- currying - a function of multiple parameters can be viewed as
a function of a single parameter that returns a function of the
remaining parameters. For example, assume that f is a function of
- (define (curry f b) (lambda (a) (f a b)))
- (define plus1 (curry + 1))
- (plus1 2) => 3
The list processing functions length, concat,
numints, and order are provided.
The length function returns the number of elements in a list and
the concat returns the concatenation of two lists. The function
numints returns the number of integers in an list of arbitrary order.
Since the list may contain elements that are themselves lists, the function
must be called recursively on both the car of the list and the cdr of the list.
Such a function is said to have deep recursion. The functions
length and concat only recurse on the cdr of the list and are
said to have shallow recursion. Note that the built-in functions
eq? and equal? for checking to see of two objects are equal differ
in that eq? uses shallow recursion and equal? uses deep
recursion. The function order computes the order of an object, which
is equal to zero if the object is an atom and 1 plus the maximum order of
its elements if it is a list. The implementation of order uses
reduce and map.
- length.scm - recursive function to
compute the length of a list. E.G. (length '(1 2 3)) => 3.
- concat.scm - recursive function to
compute the concatenation of two lists. E.G.
(concat '(1 2 3) '(4 5 6)) => (1 2 3 4 5 6).
- numints.scm - recursive function to
count the number of integers in an object. E.G.
(numints '(1 (2 3) 4)) => 4.
- order.scm - recursive function to
compute the order (maximum depth) of an object. E.G.
(order '(1 (2 (3)) 4)) => 3 and (order '((()))) => 3.
- intro.scm - text file with scheme
examples and text (from edwin editor)
Created: April 16, 2008 (modified April 19, 2012) by jjohnson AT cs DOT drexel DOT DOT edu
- Install MIT Scheme and run sample programs. You can load programs from
a file using the load function. Use (pwd) to find out which
directory scheme is currently in (this is where it looks for files by
default). You can change directories with the cd function. See
the reference manual for more details.
- Experiment with scheme interpreter (REPL Read-Eval-Print-Loop) - try out
different built-in functions and write some of your own.
- Practice scheme assignment.