Administrativa

CS345H: Programming Languages

I Forgot to mention last time: No Textbook
Lecture 2: Lambda Calculus II
I Today thee handouts: L Reference Manual, Written
and Introduction to L
Assignment 1 and Programming Assignment 0.

Thomas Dillig I Piazza course site is set up with discussion forum

I Please use forum instead of email

Thomas Dillig, CS345H: Programming Languages Lecture 2: Lambda Calculus II and Introduction to L 1/27 Thomas Dillig, CS345H: Programming Languages Lecture 2: Lambda Calculus II and Introduction to L 2/27

Recursion Recall: Named Function

I Write function definition as

I I claimed last lecture that λ-calculus was as expressive as any
programming language, e.g. it is Turing-complete fun f with x in e ≡ let f = λx .e in

I Function call is now just application (f e1 ) → (λx .e)e1

I But for Turing completeness, we need to write recursive
functions in λ-calculus
I What about recursion?

Thomas Dillig, CS345H: Programming Languages Lecture 2: Lambda Calculus II and Introduction to L 3/27 Thomas Dillig, CS345H: Programming Languages Lecture 2: Lambda Calculus II and Introduction to L 4/27

Recursion Recursion

I Let us try to define a function that computes the factorial of a

number
I First, expand the function definition:
I Recall recursive factorial definition:
I Factorial of 0 is 1 I fun f with n = ( if n = 0 then 1 else n ∗ (f (n − 1))) in ... →
let f = λn.( if n = 0 then 1 else n ∗ (f (n − 1))) in ...
I Factorial of n is n∗ Factorial of (n − 1)
I Now, consider computing factorial of 3:
I Let’s try to write this in λ-calculus:
I let f = λn.( if n = 0 then 1 else n ∗ (f (n − 1))) in (f 3)
I fun f with n = ( if n = 0 then 1 else n ∗ (f (n − 1))) in ...

I Does this work?

Thomas Dillig, CS345H: Programming Languages Lecture 2: Lambda Calculus II and Introduction to L 5/27 Thomas Dillig, CS345H: Programming Languages Lecture 2: Lambda Calculus II and Introduction to L 6/27

1
Recursion What about Recursion?

I Next, expand the let binding: I On the face of it, λ-calculus does not seem to allow recursion

I Recall: let x = e1 in e2 defined as e2 [e1 /x ] I But this would make λ-calculus very boring; not many
interesting functions can be computed without recursion
I let f = λn.( if n = 0 then 1 else n ∗ (f (n − 1))) in (f 3) →
(λn.( if n = 0 then 1 else n ∗ (f (n − 1))) 3 I Amazing fact: It is possible to encode recursion in λ-calculus

I Left with undefined symbol f I It is just a little bit involved (but very instructive to
understand)
I Conclusion: We cannot encode recursion using named
functions I Any ideas?

Thomas Dillig, CS345H: Programming Languages Lecture 2: Lambda Calculus II and Introduction to L 7/27 Thomas Dillig, CS345H: Programming Languages Lecture 2: Lambda Calculus II and Introduction to L 8/27

Encoding Recursion Encoding Recursion Cont.

I Now, we can use a λ-abstraction to remove n from the

I Recall again the factorial function we would like to compute:
left-hand side: (f n) = ( if n = 0 then 1 else n ∗ (f (n − 1)))
I fun f with n = ( if n = 0 then 1 else n ∗ (f (n − 1)))
I f = λn.( if n = 0 then 1 else n ∗ (f (n − 1)))
I We can view this function definition as an equation:
I Consider defining another function G by moving f to the
I (f n) = ( if n = 0 then 1 else n ∗ (f (n − 1))) right-hand side:

I G = λf .λn.( if n = 0 then 1 else n ∗ (f (n − 1)))

I This states that the value of f n is 1 if n is 0 and
n ∗ (f (n − 1)) otherwise
I To see that this is correct, show that f = G(f ) at home

Thomas Dillig, CS345H: Programming Languages Lecture 2: Lambda Calculus II and Introduction to L 9/27 Thomas Dillig, CS345H: Programming Languages Lecture 2: Lambda Calculus II and Introduction to L 10/27

Fixed Points The Y -combinator

I We can define a fixed-point operator in λ-calculus as follows:

I A fixed point of function h is value v such that v = h(v ) Y = λf .(λx .f (x x ))(λx .f (x x ))

I Intuition: The fixed point of h is applying h until v = h(v ), I This bizarre expression is called Y-combinator
i.e., the base case of the recursion is hit
I Recall property of fixed-point: v = h(v ) for any function h.
I But completely unclear how we can compute fixed-point of h
I Lets confirm that Y has this property:
I An expression that computes a fixed point is called a fixed
I Y h → λf .(λx .f (x x ))(λx .f (x x )) h →
point operator
(λx .h(x x ))(λx .h(x x )) → (h(x x ))[(λx .h(x x ))/x ] →
h(λx .h(x x ))(λx .h(x x )) → h(Yh)

Thomas Dillig, CS345H: Programming Languages Lecture 2: Lambda Calculus II and Introduction to L 11/27 Thomas Dillig, CS345H: Programming Languages Lecture 2: Lambda Calculus II and Introduction to L 12/27

2
Using the Y -combinator Fixed points Summary
I Let’s see how we can use the Y −combinator to compute
factorial:
I We can compute recursive functions in λ-calculus using
I Recall: G = λf .λn.( if n = 0 then 1 else n ∗ (f (n − 1)))
fixed-point operators
I Claim: Factorial of n can be computed as (YG) n
I We have seen the most famous fixed-point operator, the
I Example: Y -combinator

(YG) 2 I However, there are other λ expressions that also compute

→ (G(YG))2 fixed points.
→ (λf .λn( if n = 0 then 1 else n ∗ (f (n − 1)))(YG))2
→ λn( if n = 0 then 1 else n ∗ ((YG) (n − 1)))2 I Remember: Not every recursive function has to terminate, so
→ if 2 = 0 then 1 else 2 ∗ ((YG) (2 − 1)) this means we can write λ terms that will reduce forever
→ if 2 = 0 then 1 else 2 ∗ ((YG) 1)
→ 2 ∗ ((YG) 1)
→ ...

Thomas Dillig, CS345H: Programming Languages Lecture 2: Lambda Calculus II and Introduction to L 13/27 Thomas Dillig, CS345H: Programming Languages Lecture 2: Lambda Calculus II and Introduction to L 14/27

Next: Your course project Course Project Overview

I You will implement a lexer, parser, interpreter for the L

language

I You can find a reference interpreter on the UT Austin

machines to run L programs on (see the L Language handout
for details)

I As the name suggests, L is very similar to λ-calculus, but still

a useful language

I L has a bizarre property that is (almost) unique among

programming languages: At the end of the semester, there
will be many more interpreters for L than L programs

Thomas Dillig, CS345H: Programming Languages Lecture 2: Lambda Calculus II and Introduction to L 15/27 Thomas Dillig, CS345H: Programming Languages Lecture 2: Lambda Calculus II and Introduction to L 16/27

Language Overview Language Overview

I In L, every expression evaluates to a value

I Of course, L has the λ-operator
I The result of running a L program is the value of the program
I Example: (lambda x. x+3 2) will evaluate to ”5”
I Example: let x = 3 in x will evaluate to ”3”
I Note: You must write parenthesis for any applications!
I In addition to integers, L also supports strings
I This means lambda x. x+3 2 is not a valid L program
I Example: let x = "cs312" in x will evaluate to ”cs312”

Thomas Dillig, CS345H: Programming Languages Lecture 2: Lambda Calculus II and Introduction to L 17/27 Thomas Dillig, CS345H: Programming Languages Lecture 2: Lambda Calculus II and Introduction to L 18/27

3
More L Examples Functions in L
I let g =
lambda a.if a>0 then 2*a else 3*a
in I For convenience, L also has built-in function definitions:
let u = 12 in
(g u) I fun compute_grade with percent =
if percent>90 then "A" else
I Value: ”24” if percent>80 then "B" else "F"
in
I L also supports currying:
(compute_grade 73)
I let x = lambda a,b.a+b in I Result: ”F”
let y = (x 2) in
(y 3)

I Value: ”5”

Thomas Dillig, CS345H: Programming Languages Lecture 2: Lambda Calculus II and Introduction to L 19/27 Thomas Dillig, CS345H: Programming Languages Lecture 2: Lambda Calculus II and Introduction to L 20/27

Recursion in L Input/Output in L
I Unlike λ-calculus, L allows you to write ”naturally” recursive
functions
I L has special operators for input/output:
I fun factorial with n =
if n<=0 then 1 else n* (factorial (n-1)) I let _ = print "Please enter an integer: " in
in let i = readInt in
... let _ = print "Please enter a string: " in
let s = readString in
I Can also write ”naturally recursive” anonymous functions: let _ = print "Integer read: " in
let _ = print i in
I let fact = let _ = print "String read: " in
lambda n. if n=0 then 1 else n* (fact (n-1)) let _ = print s in
in 0
(fact 6)

I We will learn later how L allows natural recursion

Thomas Dillig, CS345H: Programming Languages Lecture 2: Lambda Calculus II and Introduction to L 21/27 Thomas Dillig, CS345H: Programming Languages Lecture 2: Lambda Calculus II and Introduction to L 22/27

Lists in L List Examples

I L also supports lists

I let x = 1@2@3@4 in x
I Lists are represented as pairs with a head and tail element
I Value: ”[1, [2, [3, 4]]]”
I This allows very generic data structures, no just linear lists
I let x = 1@2@3@4 in !x
I L has the following list operators:
I isNil: 1 if list is empty, 0 otherwise I Value: ”1”

I e1@e2: Returns a list with e1 as head and e2 as tail I let x = 1@2@3@4 in #x

I !e1:Returns head of e1 if e1 is list, e1 otherwise I Value: ”[2, [3, 4]]”
I #e1:Returns tail of e1 if e1 is list, Nil otherwise

Thomas Dillig, CS345H: Programming Languages Lecture 2: Lambda Calculus II and Introduction to L 23/27 Thomas Dillig, CS345H: Programming Languages Lecture 2: Lambda Calculus II and Introduction to L 24/27

4
More List Examples Run-time errors

I How about computing the length of a list? I There are many run-time errors possible in L programs:

I fun length with l = I Example: let x = "hello" in x+3

if isNil l then 0 else 1 + (length (#l))
in I Result: ”Run-time error: Binop + can only be applied to
(length "A"@"B"@"C") expressions of same type”

I Value: ”3” I The L reference manual lists all possible errors

Thomas Dillig, CS345H: Programming Languages Lecture 2: Lambda Calculus II and Introduction to L 25/27 Thomas Dillig, CS345H: Programming Languages Lecture 2: Lambda Calculus II and Introduction to L 26/27

Course Project Details

I Your first four programming assignments will use L and built
an L interpreter
I Assignment 0: Develop and L program

I Assignment 1: Lexer

I Assignment 2: Parser

I Assignment 3: Interpreter

I Assignment 4: Type Inference

I You will complete these assignments in L and C++

I I posted a quick C++ introduction on the website

I But we will use only tiny subset of C++, easy to pick up

Thomas Dillig, CS345H: Programming Languages Lecture 2: Lambda Calculus II and Introduction to L 27/27