Composing functions in Kotlin with extensions and operators

Today I was wondering if I could achieve a more elegant way to compose functions with Kotlin. Let’s imagine we have this:

fun same(val: Int) = val  
fun twice(val: Int) = val * 2  
fun trice(val: Int) = val * 3

fun composed(val: Int) = same(twice(trice(int)))

That gets the job done, but wouldn’t it be better if there was a way to remove all those brackets and get something more similar to a pipeline? Keep in mind: the order of execution will be inverse (trice-twice-same), which is not intuitive.

Turns out there is, if we use two awesome Kotlin features: extensionsand operator overloading. What are those?

Extensions

Extensions let you add extra methods to any type, calling them as if you were declaring a method inside that type. An example could be:

fun Int.double() {  
  return this * 2  
}

2.double() == 4

This basically means: add a method to the type Int with name double, which you can then invoke on any Int. The keyword this is used to get the instance on which you’re invoking the method (the receiver object). How this works in more detail is compiling to a static method of which this is the first parameter.

An interesting thing is that, in Kotlin, a function can be used as a type too! So we might be able to write something like this, as an extension function:

fun (() -> Unit).andHello() {  
  this()  
  println("Hello, world!")  
}

fun greet() {  
  println("Hey there!")  
}

fun greetAndHello() {  
  ::greet.andHello()  
}

() -> Unit is the way you can describe a function type (you’d do the same if it was a parameter). Inside the () you would put the parameters, Unit means it’s a void function (otherwise it would be the return type, for example (Int, Int) -> Int would be a function taking 2 Int parameters and returning an Int.

Here we are referencing a function with the :: operator, and applying the andHello() method to it. Pretty neat!

Operator overloading

Imagine you have an interesting type, and you’d like to use some standard operators with it. Wouldn’t it be great, for example, if you could access the elements in a map simply using square brackets? You can indeed!

val map = mapOf("a" to 1, "b" to 2, "c" to 3)

println(map["a"]) // 1

How does this work? Via operator overloading: you declare a method with the keyword operator in front of it, plus a certain name and signature, and then implement the method as you would do with any other method.

class Map<Key, Value> {  
  operator get(key: Key): Value {  
    // get and return the value from the map here  
  }  
}

You can then use the operator on that type normally. That can come in handy to redeclare operators in a way that makes more sense for your classes (maybe you want to sum twoTime instances together?).

You can find the full list of available operators and respective signatures in the operator overloading documentation page.

So, let’s take a small leap and imagine this is the final result we wanted:

fun composed(val: Int) = (::trice..::twice..::same) (val)

What are we trying to achieve? It’s pretty clear: we want to call, in sequence, in this order, the functions. Notice how the call order is now reversed compared to the first example (and makes it easier to read).

I’m using the .. operator (range) because it reminds me of chaining when we talk about functions. For example, it’s used to call multiple methods in sequence on an object in Dart.

How could we achieve that? All those functions are receiving one parameter and returning something, and we know that we can create an extensions for it.

operator fun <T, R, V> ((T) -> R).rangeTo(other: (R) -> V): ((T) -> V) {  
    return {other(this(it))  
    }}

What is this? Let’s go one step at a time.

First of all, we’re declaring an operator overload. The rangeTo method represents the .. operator.

We’re declaring this extension on any function of type (T) -> R so any function that takes a parameter of type T and returns a type R. Notice that T and R can be the same (for example, Ints).

This extension accepts another function, of type (R) -> V, so it will take the previous function return type and return another type.

Finally, this will generate another function, of type (T) -> V which is what we expect after calling both functions.

With the acquired knowledge, let’s look at this again, does it make more sense?

fun composed(val: Int) = (::trice..::twice..::same) (val)

We’re applying the .. operator to trice, twice, and same, in order. Since the return type is a function, the result of ::trice..::twice can be then chained with ::same. The result of this last operation is again a function, which then we invoke with (val) as an argument.

Kotlin metaprogramming features are extremely intersting, and give you the chance to make your code more expressive, especially when you combine features like extensions and operator overloading. This is only a small example of what you can achieve, so start playing with it if you aren’t already!