Overview

• Why functional programming?
• Historical origins of functional programming
• Functional programming today
• Concepts of functional programming
• A crash course on programming in Scheme

Note: this set of notes covers Chapter 11 Sections 11.1 to 11.2. You are not required to study Sections 11.2.2, 11.2.4, and 11.2.5.

Why Functional Programming?

• Functional programming is a different programming paradigm
• Imperative programming languages are more widely used
• Integrated software development environments for procedural and object oriented programming languages are "industrial strength"
• However, many (commercial) applications exist for functional programming:
• Symbolic data manipulation
• Natural language processing
• Artificial intelligence
• Automatic theorem proving and computer algebra
• Algorithmic optimization of programs written in pure functional languages

Why Functional Programming in This Course?

• A functional language will be used to illustrate a diversity of programming language concepts
• Functional programming languages are
• Compiled and/or interpreted (Section 1.4)
• Have simple syntax (Chapter 2)
• Use garbage collection (Section 3.2.3) for memory management
• Are statically scoped or dynamically scoped (Section 3.3)
• Use higher-order functions and subroutine closures (Section 3.4.1)
• Use first-class function values (Section 3.4.2)
• Depend heavily on polymorphism (Section 3.5)
• Employ recursion (Section 6.6) for repetive execution
• Programs have no side effects and all expressions are referentially transparent (Sections 6.1.2 and 6.3)

Origin of Functional Programming

• Church's thesis:
• All models of computation are equally powerful and can compute any function
• Turing's model of computation: Turing machine
• Reading/writing of values on an infinite tape by a finite state machine
• Church's model of computation: lambda calculus
• This inspired functional programming as a concrete implementation of lambda calculus
• Computability theory
• A program can be viewed as a constructive proof that some mathematical object with a desired property exists
• A function is a mapping from inputs to output objects and computes output objects from appropriate inputs
• For example, the proposition that every pair of nonnegative integers (the inputs) has a greatest common divisor (the output object) has a constructive proof implemented by Euclid's algorithm written as a "function"
 

gcd(a,b) =  ì ï í ï î a  if a = b  gcd(a-b,b)  if a > b  gcd(a,b-a)  if b > a

Functional Programming Today

• Attractive model of computation
• Absence of side effects makes expressions referentially transparent: the value of an expression depends solely on the function return values in it and not on evaluation order and/or values of global variables
• A function can always be counted on to return the same results with the same input parameters
• Dangling and/or uninitialized pointer references do not occur
• Easier to debug and maintain programs
• Significant improvements in theory and practice of functional programming have been made in recent years
• Easier to write functional programs by using their imperative language features which are automatically translated to functional constructs (e.g. loops by recursion)
• Improved efficiency
• Remaining obstacles to functional programming:
• Social: most programmers are trained in imperative programming
• Commercial: not many libraries, not very portable, and no integrated development environments for functional languages

Concepts of Functional Programming

• Functional programming defines the outputs of a program as mathematical function of the inputs with no notion of internal state (no side effects)
• Example pure functional programming languages: Miranda, Haskell, and Sisal
• Non-pure functional programming languages include imperative features with side effects that affect global state (e.g. through destructive assignments to global variables)
• Example: Lisp, Scheme, and ML
• Useful features are found in functional languages that are often missing in imperative languages:
• First-class function values: the ability of functions to return newly constructed functions
• Higher-order functions: functions that take other functions as input parameters or return functions
• Polymorphism: the ability to write functions that operate on more than one type of data
• Aggregate constructs for constructing structured objects: the ability to specify a structured object in-line, e.g. a complete list or record value
• Garbage collection

Lisp

• Lisp (LISt Processing language) was the original functional language
• Lisp and dialects are still the most widely used
• Simple and elegant design of Lisp:
• Homogeneity of programs and data: a Lisp program is a list and can be manipulated in Lisp as a list
• Self-definition: a Lisp interpreter can be written in Lisp
• Interactive: interaction with user through "read-eval-print" loop

A Crash Course on Scheme

• Scheme is a popular Lisp dialect
• Lisp and Scheme adopt Cambridge Polish notation for expressions:
• An expression is an atom, e.g. a number, string, or identifier name
• An expression is a list whose first element is the function name (or operator) followed by the arguments which are expressions:

(function arg1 arg2 arg3 ...)
• The "Read-eval-print" loop provides user interaction: an expression is read, evaluated by evaluating the arguments first and then the function/operator is called after which the result is printed
• Input: 9
• Output: 9
• Input:(+ 3 4)
• Output: 7
• Input:(+ (* 2 3) 1)
• Output: 7
• User can load a program from a file with the load function
• (load "my_scheme_program")
• The file name should use the .scm extension

Note: You can run the Scheme interpreter and try the examples in these notes by executing the scheme command. To exit Scheme, type (exit). You can download an example Scheme program "Eliza". More information on Scheme can be found at http://www.swiss.ai.mit.edu/projects/scheme

Scheme Data Structures

• The only data structures in Lisp and Scheme are atoms and lists
• Atoms are:
• Numbers, e.g. 7
• Strings, e.g. "abc"
• Identifier names (variables), e.g. x
• Boolean values true #t and false #f
• Symbols which are quoted identifiers which will not be evaluated, e.g. 'y
• Input: a
• Output: Error: unbound variable a
• Input: 'a
• Output: a
• Lists:
To distinghuish list data structures from expressions that are written as lists, a quote (') is used to quote the list:
'(elt1 elt2 elt3 ...)
• Input: '(3 4 5)
• Output: (3 4 5)
• Input: '(a 6 (x y) "s")
• Output: (a 6 (x y) "s")
• Input: '(a (+ 3 4))
• Output: (a (+ 3 4))
• Input: '()
• Output: ()
• Note: the empty list () is also identical to false #f in Scheme

Primitive List Operations

• car returns the head (first element) of a list
• Input: (car '(2 3 4))
• Output: 2
• cdr (pronounced "coulder") returns the tail of a list (list without the head)
• Input: (cdr '(2 3 4))
• Output: (3 4)
• cons joins an element and a list to construct a new list
• Input: (cons 2 '(3 4))
• Output: (2 3 4)
• Examples:
• Input: (car '(2))
• Output: 2
• Input: (car '())
• Output: Error
• Input: (cdr '(2 3))
• Output: (3)
• Input: (cdr (cdr '(2 3 4)))
• Output: (4)
• Input: (cdr '(2))
• Output: ()
• Input: (cons 2 '())
• Output: (2)

Type Checking

• The type of an expression is determined only at run-time
• Functions need to check the types of their arguments explicitly
• Type predicate functions:
• (boolean? x)  ; is x a Boolean?
• (char? x)     ; is x a character?
• (string? x)   ; is x a string?
• (symbol? x)   ; is x a symbol?
• (number? x)   ; is x a number?
• (list? x)     ; is x a list?
• (pair? x)     ; is x a non-empty list?
• (null? x)     ; is x an empty list?

If-Then-Else

• Special forms resemble functions but have special evaluation rules
• A conditional expression in Scheme is written using the if special form:

(if condition thenexpr elseexpr)
• Input: (if #t 1 2)
• Output: 1
• Input: (if #f 1 "a")
• Output: "a"
• Input: (if (string? "s") (+ 1 2) 4)
• Output: 3
• Input: (if (> 1 2) "yes" "no")
• Output: "no"
• A more general if-then-else can be written using the cond special form:

(cond listofconditionvaluepairs)
where the condition value pairs is a list of (cond value) pairs and the condition of the last pair can be else to return a default value
• Input: (cond ((< 1 2) 1) ((>= 1 2) 2))
• Output: 1
• Input: (cond ((< 2 1) 1) ((= 2 1) 2) (else 3))
• Output: 3

Testing

• eq? tests whether its two arguments refer to the same object in memory
• Input: (eq? 'a 'a)
• Output: #t
• Input: (eq? '(a b) '(a b))
• Output: () (false: the lists are not stored at the same location in memory!)
• equal? tests whether its arguments have the same structure
• Input: (equal? 'a 'a)
• Output: #t
• Input: (equal? '(a b) '(a b))
• Output: #t
• To test numerical values, use =, <>, >, <, >=, <=, even?, odd?, zero?
• member tests membership of an element in a list and returns the rest of the list that starts with the first occurrence of the element, or returns false
• Input: (member 'y '("s" x 3 y z))
• Output: (y z)
• Input: (member 'y '(x (3 y) z))
• Output: ()

Lambda Abstraction

• A Scheme lambda abstraction is a nameless function specified with the lambda special form:

(lambda formalparameters functionbody)
where the formal parameters are the function inputs and the function body is an expression that is the resulting value of the function
• Examples:
• (lambda (x) (* x x)) ; is a squaring function: x®x²
• (lambda (a b) (sqrt (+ (* a a) (* b b)))) ; is a  function:

(a b)®   _____ Öa2+b2

Lambda Application

• A lambda abstraction is applied by assigning the evaluated actual parameter(s) to the formal parameters and returning the evaluated function body
• The form of a function call in an expression is:

(function arg1 arg2 arg3 ...)
where function can be a lambda abstraction
• Example:
• Input: ((lambda (x) (* x x)) 3)
• Output: 9
• That is, x=3 in (* x x) which evaluates to 9

Defining Global Functions in Scheme

• A function is globally defined using the define special form:

(define name function)
• For example:

(define sqr
  (lambda (x) (* x x))
)
defines function sqr
• Input: (sqr 3)
• Output: 9
• Input: (sqr (sqr 3))
• Output: 81
• (define hypot

  (lambda (a b)
    (sqrt (+ (* a a) (* b b)))
  )
)
defines function hypot
• Input: (hypot 3 4)
• Output: 5

Bindings

• An expression can have local name-value bindings defined with the let special form

(let listofnameandvaluepairs expression)
where name and value pairs is a list of pairs (namevalue) and expression is returned in which each name is replaced with its value in the list
• Input:

(let ((a 3)
      (b 4)
     )
     (hypot a b)
)
• Output: 5
• A name can be bound to a function in let
• Input:

(let ((sqr (lambda (x) (* x x)))
      (y 3)
     )
     (sqr y)
)
• Output: 9

Recursive Bindings

• An expression can have local recursive function bindings defined with the letrec special form

(letrec listofnameandvaluepairs expression)
where name and value pairs is a list of pairs (namevalue) and expression is returned where each name is replaced with its value
• Input:

(letrec ((fact (lambda (n)
                 (if (= n 1)
                   1
                   (* n (fact (- n 1)))
                 )
               )
         )
        )
     (fact 5)
)
• Output: 120
• This allows the local factorial function fact to refer to itself

I/O and Sequencing

• display prints a value
• Input: (display "Hello World!")
• Output: "Hello World!"
• Input: (display (+ 2 3))
• Output: 5
• newline advances to a new line
• Input: (newline)
• read returns a value from standard input
• begin sequences a series of expressions (its value is the value of the last expression)
• Example:

(begin
  (display "Hello World!")
  (newline)
)
• Example:

(let ((x 1)
      (y (read))
      (plus +)
     )
     (begin
       (display (plus x y))
       (newline)
     )
)

Loops

• do takes a list of name-init-update triples, a termination test with final value, and a loop body

(do listoftriples condition body)
• Example:

(do ((i 0 (+ i 1)))
    ((>= i 10) "done")
    (display i)
    (newline)
)
Since everything is an expression in Scheme, a loop must return a value which in this case is the string "done"

Higher-Order Functions

• A function is called a higher-order function (also called a functional form) if it takes a function as an argument or returns a newly constructed function as a result
• Scheme has several built-in higher-order functions, for example:
• apply takes a function and a list and applies the function with the elements of the list as arguments
• Input: (apply '+ '(3 4))
• Output: 7
• Input: (apply (lambda (x) (* x x)) '(3))
• Output: 9
• map takes a function and a list and returns a list after applying the function to each element of the list
• Input: (map odd? '(1 2 3 4))
• Output: (#t () #t ())
• Input: (map (lambda (x) (* x x)) '(1 2 3 4))
• Output: (1 4 9 16)
• Here is a function that applies a function to an argument twice:
• (define twice

  (lambda (f n) (f (f n)))
)
• Input: (twice sqrt 81)
• Output: 3

Non-Pure Constructs: Assignments

• Assignments are considered bad in functional programming because they can change the global state of the program and possibly influence function outcomes
• set! assigns to a variable a new value, for example:
• (define a 0)

...
(set! a 1) ; overwrite a with 1
...
• (let ((a 0))

     (begin
       ...
       (set! a (+ a 1)) ; increment a by 1
       ...
      )
)
• set-car! overwrites the head of a list
• set-cdr! overwrites the tail (rest) of a list

Scheme Examples

• Recursive factorial function:

(define fact
  (lambda (n)
    (if (zero? n) 1 (* n (fact (- n 1))))
  )
)
• Iterative factorial function:

(define iterfact
  (lambda (n)
    (do ((i 1 (+ i 1))
         (f 1 (* f i))
        )
        ((> i n) f)
        ; note: loop body is omitted
   )
  )
)

Example Recursive Functions on Lists

• Sum the elements of a list:

(define sum
  (lambda (lst) 
    (if (null? lst)
      0
      (+ (car lst) (sum (cdr lst))) ; add value of head to sum of rest of list
    )
  )
)
• Input: (sum '(1 2 3))
• Output: 6
• Check if element is in list:

(define in?
  (lambda (elt lst)
    (cond
      ((null? lst) #f) ; if list is empty, return false
      ((= elt (car lst)) #t) ; if element is the head, return true
      (else (in? elt (cdr lst))) ; keep searching rest of list
    )
  )
)
• Input: (in? 2 '(1 2 3))
• Output: #t

Examples of List Functions

• (define fill

  (lambda (num elt)
    (cond
      ((= 0 num) '())
      (else (cons elt (fill (- num 1) elt)))
    )
  )
)
• Input: (fill 3 "a")
• Output: ("a" "a" "a")
• (define between

  (lambda (start end)
    (if (> start end)
      '()
      (cons start (between (+ start 1) end))
    )
  )
)
• Input: (between 1 10)
• Output: (1 2 3 4 5 6 7 8 9 10)
• (define zip

  (lambda (lst1 lst2)
    (cond
      ((null? lst1) '())
      ((null? lst2) '())
      (else (cons (list (car lst1) (car lst2)) (zip (cdr lst1) (cdr lst2))))
    )
  )
)
• Input: (zip '(1 2 3) '(a b c))
• Output: ((1 a) (2 b) (3 c))
• (define take

  (lambda (num lis)
    (cond
      ((= num 0) '())
      (else (cons (car lis) (take (- num 1) (cdr lis))))
    )
  )
)
• Input: (take 3 '(a b c d e f))
• Output: (a b c)

Examples of Higher-Order Functions

• Reduce a list by applying a binary operator to all elements (i.e. elt1 + elt2 + elt3 + ...):

(define reduce
  (lambda (op lst)
    (if (null? (cdr lst))
      (car lst)
      (op (car lst) (reduce op (cdr lst)))
    )
  )
)
• Input: (reduce + '(1 2 3))
• Output: 6
• Filter elements of a list for which a condition (a predicate function) returns true:

(define filter
  (lambda (op lst)
    (cond
      ((null? lst) '())
      ((op (car lst)) (cons (car lst) (filter op (cdr lst))))
      (else (filter op (cdr lst)))
    )
  )
)
• Input: (filter odd? '(1 2 3 4 5))
• Output: (1 3 5)