Functions as Abstractions

Let's motivate functions by look at structural recursion over the natural numbers, and trying to capture in code the pattern we've used many times.

We have written a lot of structural recursions over the natural numbers. We started with code like

def stackedBoxes(count: Int): Image =
  count match {
    case 0 => Image.empty
    case n => Image.square(20).beside(stackedBoxes(n-1))
  }

and more recently have been writing code like

def polygonPoints(sides: Int, radius: Double): Image = {
  val turn = (1.0 / sides).turns
  def loop(count: Int): Image =
    count match {
      case 0 => Image.empty
      case n =>
        Image
          .circle(5)
          .at(Point(radius, turn * n))
          .on(loop(n - 1))
    }

  loop(sides)
}

The details are different, but the underlying structure of the two examples is the same. In fact that's one of the points we're emphasizing: that code has underlying patterns that apply in many different situations.

When we see this repetition we might wonder if we can somehow capture it in code, so we don't have to write it out again and again. We're going to do this in this section, and in doing so we'll introduce functions.

Let's start by putting the two structural recursions next to each other, and in the case of polygonPoints removing the surrounding code.

def stackedBoxes(count: Int): Image =
  count match {
    case 0 => Image.empty
    case n => aBox.beside(stackedBoxes(n-1))
  }

def loop(count: Int): Image =
  count match {
    case 0 => Image.empty
    case n =>
      Image
        .circle(5)
        .at(Point(radius, turn * n))
        .on(loop(n - 1))
  }

Keeping the common parts of these methods, and replacing those that vary with ???, leaves us with something like

def aMethod(count: Int): Image =
  count match {
    case 0 => Image.empty
    case n => ??? aMethod(count - 1)
  }

This is the code skeleton for structural recursion over the natural numbers.

Now our challenge is to turn this into something we can actually use in Scala. Let's look at the recursive case in the code skeleton

case n => ??? aMethod(count - 1)

This is the only part that has a ???. In the actual methods we have

case n => aBox.beside(stackedBoxes(n-1))

and

case n =>
  Image
    .circle(5)
    .at(Point(radius, turn * n))
    .on(loop(n - 1))

respectively.

To capture this in code we need to allow:

  1. varying how we build the result of the recursion (in one case it's beside, while we use on in the other); and
  2. the Image at the current step to depend on n (as in the case taken from polygonPoints, where we use the value n to determine the location of the circle.)

In other words, the value we build is parameterized by the result of the recursion and the value n. We could express this with a method

def build(n: Int, recursive: Image): Image = ???

and then, with two different implementations of build, we could write both original methods in terms of

def aMethod(count: Int, build: ???): Image =
  count match {
    case 0 => Image.empty
    case n => build(n, aMethod(count - 1, build))
  }

However this won't work; we cannot pass a method as a parameter to a method. The solution, of course, is to use a function.

A function is basically a method, but we can use a function as a first-class value:

Here's how we can solve the problem above using functions. First I'm going to define the method we've been calling aMethod above. I'm going to call it fold, which is the usual name for this kind of method. (This isn't an entirely correct implementation for fold, but we don't have all the tools to do it properly right now. This is something we'll come back to in a later chapter.)

def fold(count: Int, build: (Int, Image) => Image): Image =
  count match {
    case 0 => Image.empty
    case n => build(n, fold(count - 1, build))
  }

This method definition says that the parameter build is a function from two arguments (an Int and an Image) to an Image. To call fold we need to create a function, so let's see an example of that.

val aBox = Image.square(20).fillColor(Color.royalBlue).strokeColor(Color.crimson)

val stack = (count: Int, image: Image) => aBox.beside(image)

We can now call fold with stack.

fold(5, stack)

This produes the output below.

Now that we've seen an example of functions, let's go into the details of defining and using them.

Function Literals

We've just seen an example of a function literal, which was

(count: Int, image: Image) => aBox.above(image)

The general syntax is an extension of this.

Function Literal Syntax

The syntax for declaring a function literal is

(parameter: type, ...) => expression

where

  • the optional parameters are the names given to the function parameters;
  • the types are the types of the function parameters; and
  • the expression determines the result of the function.

The parentheses around the parameters are optional if the function has just a single parameter.

Function Types

To pass functions to methods we need to know how to write down their types (because when we declare a parameter we have to declare its type).

We write a function type like (A, B) => C where A and B are the types of the parameters and C is the result type. The same pattern generalises from functions of no arguments to an arbitrary number of arguments.

Here's an example. We create a method that accepts a function, and that function is from Int to Int. We write this type as Int => Int or (Int) => Int.

def squareF(x: Int, f: Int => Int): Int =
  f(x) * f(x)

We can define a function add42, and pass it to this method

val add42 = (x: Int) => x + 42
squareF(0, add42)
// res3: Int = 1764

We could also pass a function literal

squareF(0, x => x + 42)
// res4: Int = 1764

Note that we didn't have to put the parameter type on the function literal in this case because Scala has enough information to infer the type.

Function Type Declaration Syntax

To declare a function type, write

(A, B, ...) => C

where

  • A, B, ... are the types of the input parameters; and
  • C is the type of the result.

If a function only has one parameter the parentheses may be dropped:

A => B

Functions as Objects

All first class values are objects in Scala, including functions. This means functions can have methods, including some useful means for composition.

val addTen = (a: Int) => a + 10
val double = (a: Int) => a * 2
val combined = addTen.andThen(double) // this composes the two functions
combined(5)
// res5: Int = 30

Calling a function is actually calling the method called apply on the function. Scala allows a shortcut for any object that has a method called apply, where can drop the method name apply and write the call like a function call. This means the following are equivalent.

val halve = (a: Int) => a / 2
halve(4)
// res6: Int = 2
halve.apply(4)
// res7: Int = 2

Converting Methods to Functions

Methods are very similar to functions, so Scala provides a way to convert functions to methods. If we follow a method name with a _ it will be converted to a function.

def times42(x: Int): Int =
  x * 42
val times42Function = times42 _
times42Function(2)
// res8: Int = 84

We can also write a method call but replace all parameters with _ and Scala will convert the method to a function.

val times42Function2 = times42(_)
times42Function2(2)
// res9: Int = 84

Exercises

Exercise: Function Literals

Let's get some practice writing function literals. Write a function literal that:

  • squares it's Int input;
  • has a Color parameter and spins the hue of that Color by 15 degrees; and
  • takes an Image input and creates four copies in a row, where each copy is rotated by 90 degrees relative to the previous image (use the rotate method on Image to achieve this.)

The first function is

(x: Int) => x * x
// res10: Function1[Int, Int] = repl.MdocSession$MdocApp0$$Lambda$13925/0x0000000103aa9040@6ff61491

The second is

(c: Color) => c.spin(15.degrees)
// res11: Function1[Color, HSLA] = repl.MdocSession$MdocApp0$$Lambda$13926/0x0000000103aaa040@753b1556

The third is

(image: Image) => 
  image.beside(image.rotate(90.degrees))
    .beside(image.rotate(180.degrees))
    .beside(image.rotate(270.degrees))
    .beside(image.rotate(360.degrees))
// res12: Function1[Image, Image] = repl.MdocSession$MdocApp0$$Lambda$13927/0x0000000103aaa840@230314d5

Exercise: Function Types

Here's an interesting function we'll do more with in later sections. We don't need to understand what it does right now, though you might want to experiment with it.

val roseFn = (angle: Angle) =>
  Point.cartesian((angle * 7).cos * angle.cos, (angle * 7).cos * angle.sin)

What is the type of the function roseFn defined above? What does this type tell us?

The type is Angle => Point. This means roseFn is a function that takes a single argument of type Angle and returns a value of type Point. In other words, roseFn transforms an Angle to a Point.