Why sum x y is of type (Num a) => a -> a -> a in Haskell?

Question

I\'ve been reading about Haskell and I\'m having a hard time understanding how function definitions are handled in this language.

Let\'s say I\'m defining a su

Answer 1

The Haskell documentation isn't too clear about it, but (Num a) => means that the function works for all cases where a is a Num or derives from it (therefore is a number).

Also see: http://www.cse.unsw.edu.au/~en1000/haskell/inbuilt.html

Answer 2

Num a => means "in the following, a shall refer to a type which is an instance of the typeclass Num" (which is kinda like an interface for number types).

The => operator separates the "typeclass constraints" from the "body" of the type. It's kind of like the where operator for generic constraints in C#. You can read it as a logical implication like "if a is a numeric type then sum can be used with type a -> a -> a".

a -> a -> a means "a function that takes an a and returns a function which takes an a and returns an a". For this to make sense you need to understand that sum x y parses as (sum x) y.

In other words: you first call sum with the argument x. You then get back a new function of type a -> a. You then call that function with the argument y and now you get back a function of type a, where a is the type of x and y and must be an instance of the Num typeclass.

If you want sum to have type Num a => (a,a) -> a, you can define it as sum (x,y) = x+y. In this case you have a function which takes a tuple containing two as and returns an a (where a is again an instance of the Num typeclass).

However the "curry style" (functions returning functions to simulate multiple parameters) is much more often used than the tuple style because it allows you to easily partially apply functions. Example map (sum 5) [1,2,3]. If you had defined sum with a tuple, you'd have to do map (\y -> sum 5 y) [1,2,3].

Answer 3

0. The Haskell => has nothing to do with C#'s =>. In Haskell an anonymous function is created with

\x -> x * x

Also, don't name the function sum because such a function already exists in Prelude. Let's call it plus from now on to avoid confusion.

1. Anyway, the => in Haskell provides a context to the type. For instance:

show :: (Show a) => a -> String

Here, The Show a => means a type must be an instance of the type class Show, which means a must be convertible to a string. Similarly, (Num a) => a -> a -> a means the a type must be an instance of the type class Num, which means a must be like a number. This puts a constraint on a so that show or plus won't accept some unsupported input, e.g. plus "56" "abc". (String is not like a number.)

A type class is similar to C#'s interface, or more specifically, an interface base-type constraint in generics. See the question Explain Type Classes in Haskell for more info.

2. a -> a -> a means a -> (a -> a). Therefore, it is actually a unary function that returns another function.

plus x = \y -> x + y

This makes partial application (currying) very easy. Partial application is used a lot esp. when using higher order functions. For instance we could use

map (plus 4) [1,2,3,4]

to add 4 to every element of the list. In fact we could again use partial application to define:

plusFourToList :: Num a => [a] -> [a]
plusFourToList = map (plus 4)

If a function is written in the form (a,b,c,...)->z by default, we would have to introduce a lot of lambdas:

plusFourToList = \l -> map(\y -> plus(4,y), l) 

Answer 4

This is because

Every function in Haskell takes a single parameter and returns a single value

If a function need to take multiple value, the function would have been a curried function or it have to take a single tuple.

If we add a parentheses, the function signature becomes:

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

In Haskell, the function signature: A -> B means a function the "domain" of the function is A and the "Codomain" of the function is B; or in a programmer's language, the function takes a parameter of type A and returns a value of type B.

Therefore, the function definition sum :: Num -> (Num -> Num) means sum is "a function that takes a parameter of type a and returns a function of type Num -> Num".

In effect, this leads to currying/partial function.

The concept of currying is essential in functional languages like Haskell, because you will want to do things like:

map (sum 5) [1, 2, 3, 5, 3, 1, 3, 4]  -- note: it is usually better to use (+ 5)

In that code, (sum 5) is a function that takes a single parameter, this function (sum 5) will be called for each item in the list, e.g. ((sum 5) 1) returns 6.

If sum had a signature of sum :: (Num, Num) -> Num, then sum would have to receive both of its parameter at the same time because now sum is a "function that receives a tuple (Num, Num) and returns a Num".

Now, the second question, what does Num a => a -> a means? It's basically a shorthand for saying that each time you see a in the signature, replace it with Num or with one of its derived class.

Answer 5

it is a -> a -> a rather than (a, a) -> a because of currying. Fun fact: Currying was (re)invented by Haskell Curry! Basically it means if you supply one one argument you'll get back another function of type a -> a, a partial application of sum.