Types, lambda functions and type classes

Function types

• Ordinary data types are for primitive data (like \(Int\) and \(Char\)) and basic data structures (like \([Int]\) and \([Char]\)).

• Algebraic data types are types that combine other types either as records (‘products’), e.g.

    data Pair = Pair Int Double
  

  or as variants (‘sums’), e.g.

    data Bool = False | True
  

• Functions have types containing an arrow, e.g. \(Int \rightarrow String\).
• We now look at function types in more detail.

Lambda expressions

• Lambda expressions (named after the greek letter \(\lambda\)) play a very important role in functional programming in general and Haskell in particular.

Named and anonymous expressions

• You can give a name \(sum\) to an expression \(2+2\):

    sum = 2+2

• But you can also write anonymous expressions — expressions that just appear, but are not given names.

    (-b) + sqrt (b^2 - 4*a*c)

• Without anonymous expressions, writing this would almost be like assembly language:

    e1 = (-b)
    e2 = b^2
    e3 = 4*a
    e4 = e3*c
    e5 = e2-e4
    e6 = e1+e5

Some background

• Sometimes in a mathematics or physics book, there are statements like “the function \(x^2\) is continuous\(\ldots\)”

• This is ok when the context makes it clear what \(x\) is.

• But it can lead to problems. What does \(x*y\) mean?

  • Is it a constant, because both \(x\) and \(y\) have fixed values?

  • Is it a function of \(x\), with a fixed value of \(y\)?

  • Is it a function of \(y\), with a fixed value of \(x\)?

  • Is it a function of both \(x\) and \(y\)?

• In mathematical logic (and computer programming) we need to be precise about this!

• A lambda expression \(\backslash x \rightarrow e\) contains

  • An explicit statement that the formal parameter is \(x\), and

  • the expression \(e\) that defines the value of the function.

Anonymous functions

• A function can be defined and given a name using an equation:

    f :: Int -> Int
    f x = x+1

• Since functions are “first class”, they are ubiquitous, and it’s often useful to denote a function anonymously.

• This is done using lambda expressions.

    \x -> x+1

Pronounced “lambda x arrow x+1”.

There may be any number of arguments:

    \x y z -> 2*x + y*z

Using a lambda expression

Functions are first class: you can use a lambda expression wherever a function is needed. Thus

    f = \x -> x+1

is equivalent to

    f x = x+1

But lambda expressions are most useful when they appear inside larger expressions.

    map (\x -> 2*x + 1) xs

Monomorphic and polymorphic functions

Monomorphic functions

Monomorphic means “having one form”.

    f :: Int -> Char
    f i = "abcdefghijklmnopqrstuvwxyz" !! i

    x :: Int
    x = 3

    f x :: Char
    f x -- > 'd'

Polymorphic functions

Polymorphic means “having many forms”.

    fst :: (a,b) -> a
    fst (x,y) = x

    snd :: (a,b) -> b
    snd (x,y) = y

    fst :: (a,b) -> a
    fst (a,b) = a

    snd :: (a,b) -> b
    snd (a,b) = b

Currying

• Most programming languages allow functions to have any number of arguments.

• But this turns out to be unnecessary: we can restrict all functions to have just one argument, without losing any expressiveness.

• This process is called Currying, in honor of Haskell Curry.

  • The technique makes essential use of higher order functions.

  • It has many advantages, both practical and theoretical.

A function with two arguments

You can write a definition like this, which appears to have two arguments:

    f :: Int -> Int -> Int
    f x y = 2*x + y

But it actually means the following:

    f :: Int -> (Int -> Int)
    f 5 :: Int -> Int

The function takes its arguments one at a time:

    f 3 4 = (f 3) 4

    g :: Int -> Int
    g = f 3
    g 10 -- > (f 3) 10 -- > 2*3 + 10

Grouping: arrow to the right, application left

• The arrow operator takes two types \(a \rightarrow b\), and gives the type of a function with argument type \(a\) and result type \(b\)

• An application \(e_1\; e_2\) applies a function \(e_1\) to an argument \(e_2\)

• Note that for both types and applications, a function has only one argument

• To make the notation work smoothly, arrows group to the right, and application groups to the left.

    f :: a -> b -> c -> d
    f :: a -> (b -> (c -> d))

    f x y z = ((f x) y) z

Type classes and ad-hoc polymorphism

The type of \((+)\)

Note that \(fst\) has the following type, and there is no restriction on what types \(a\) and \(b\) could be.

    fst :: (a,b) -> a

What is the type of \((+)\)? Could it be\(\ldots\)

    (+) :: Int -> Int -> Int
    (+) :: Integer -> Integer -> Integer
    (+) :: Ratio Integer -> Ratio Integer -> Ratio Integer
    (+) :: Double -> Double -> Double

    (+) :: a -> a -> a  -- Wrong! has to be a number

Type classes

Answer: \((+)\) has type \(a \rightarrow a \rightarrow a\) for any type \(a\) that is a member of the type class \(Num\).

    (+) :: Num a => a -> a -> a

• The class \(Num\) is a set of types for which \((+)\) is defined

• It includes \(Int\), \(Integer\), \(Double\), and many more.

• But \(Num\) does not contain types like \(Bool\), \([Char]\), \(Int\rightarrow Double\), and many more.

Two kinds of polymorphism

• Parametric polymorphism.

  • A polymorphic type that can be instantiated to any type.

  • Represented by a type variable. It is conventional to use \(a\), \(b\), \(c\), \(\ldots\)

  • Example: \(length :: [a] \rightarrow Int\) can take the length of a list whose elements could have any type.

• Ad hoc polymorphism.

  • A polymorphic type that can be instantiated to any type chosen from a set, called a “type class”

  • Represented by a type variable that is constrained using the \(\Rightarrow\) notation.

  • Example: \((+) :: Num\, a \Rightarrow a \rightarrow a \rightarrow a\) says that \((+)\) can add values of any type \(a\), provided that \(a\) is an element of the type class \(Num\).

Type inference

• Type checking takes a type declaration and some code, and determines whether the code actually has the type declared.

• Type inference is the analysis of code in order to infer its type.

• Type inference works by

  • Using a set of type inference rules that generate typings based on the program text

  • Combining all the information obtained from the rules to produce the types.

Type inference rules

The type system contains a number of type inference rules, with the form

Context

• Statements about types are written in the form similar to \(\Gamma \vdash e :: \alpha\)

• This means ``if you are given a set \(\Gamma\) of types, then it is proven that \(e\) has type \(\alpha\).

Type of constant

If we know the type \(T\) of a constant \(c\) (for example, we know that \('a' :: Char\)), then this is expressed by saying that there is a given theorem that \(c :: T\). Furthermore, this holds given any context \(\Gamma\).

Type of application

If \(e_1\) is a function with type \(\alpha \rightarrow \beta\), then the application of \(e_1\) to an argument of type \(\alpha\) gives a result of type \(\beta\).

Type of lambda expression

We have a context \(\Gamma\). Suppose that if we’re also given that \(x :: \alpha\), then it can be proven that an expression \(e :: \beta\). Then we can infer that the function \(\lambda x \rightarrow e\) has type \(\alpha \rightarrow \beta\).