Manual Currying

Manual Currying

You’ve learned about how Haskell functions work by taking a single argument. One way to write a function that takes multiple arguments is to pass in a tuple of arguments. For example, consider this addition function:

uncurriedAddition nums =
    a = fst nums
    b = snd nums
  in a + b

Haskell’s standard library includes two functions, curry and uncurry, that make it easy for you to convert between functions that take two arguments and functions that take a tuple. The curry function transforms a function like our uncurriedAddition function and turns it into one that takes two separate arguments. For example:

λ addition = curry uncurriedAddition
λ addOne = addition 1
λ addTwo = addition 2
λ addOne 1
λ addOne 2
λ addOne 3
λ addTwo 1
λ addTwo 2
λ addTwo 3

Similarly, the uncurry function takes a regular function with two arguments and converts it into a function that accepts a tuple. For example, using uncurry we could have rewritten uncurredAddition like this:

uncurriedAddition = uncurry (+)

Using what you’ve learned in this chapter, try implementing your own version of curry and uncurry.

Since the standard library already has functions named curry and uncurry, you should select different names for your implementations. After you’ve written your versions, compare the behavior to the standard library implementations to ensure that your versions behave the same way.


Click to reveal

Remember that functions can be passed around as ordinary arguments. For example, imagine that we have a function called addNumbers that adds two numbers:

addNumbers a b = a + b

Next, we could write a function callWithTwoArguments that takes a function, and the two arguments that we should call that function with, and returns the results:

callWithTwoArguments f a b = f a b

Finally, we can pass addNumbers to callWithTwoArguments like any other value. As an example:

addThreeAndFive = callWithTwoArguments addNumbers 3 5

As you’re working on this exercise, remember that you can pass the functions that you want to curry, or uncurry, just like you’d pass around any other argument.

Click to reveal

Remember that you can get the first element of a tuple using the fst function, and you can get the second element of a tuple using the snd function:

λ fst ("hello", "haskell")

λ snd ("hello", "haskell")
Click to reveal

The built-in curry turns a function that takes a tuple into a function that takes two arguments. Let’s look at an example. Imagine that we have a function that adds two numbers from a tuple:

addTuple tuple = fst tuple + snd tuple

We can test this out in ghci and see that it works just like we’d expect.

λ addTuple (1,2)

If we use curry, we can call this like an ordinary function. We can either use curry and pass arguments all in one call:

λ curry addTuple 1 2

Or we can use curry to define a new version of the function that takes two non-tuple arguments:

λ addTwo = curry addTuple
λ addTwo 1 2
λ addTwo 3 4

The uncurry function works the same way, but it converts functions in the other direction. For example:

λ uncurry addTwo (1,2)
λ addTuple' = uncurry addTwo
λ addTuple' (2,3)

You can use these examples as you are testing your own implementation.


Click to reveal

Let’s start by defining our own version of curry called exampleCurry. Our function will need to take three arguments:

  • f is a function that takes in a tuple and returns a value
  • a is a value; it’s the first value in the tuple we’ll pass to f
  • b is a value; it’s the second value in the tuple we’ll pass to f

It’s easiest to write this function as a one-liner:

exampleCurry f a b = f (a,b)

In this code, we’re taking three arguments as input. We’re calling our first argument, the function we want to curry, using a tuple made up of the next two arguments.

If we use this function in ghci we can see it behaves like you’d expect:

λ addTuple tuple = fst tuple + snd tuple
λ exampleCurry addTuple 1 2

Just like with the curry function defined for us in Prelude, we can use exampleCurry to create a new function:

λ addTwo = exampleCurry addTuple
λ addTwo 1 2
λ addTwo 2 3

This might be a bit surprising, since you’re still getting used to Haskell. Remember that Haskell makes it easy for us to do partial application. We could have written addTwo without partial application by taking in the two arguments that we should call the curried function with:

λ addTwo a b = exampleCurry addTuple a b
λ addTwo 1 2
λ addTwo 2 3