Anonymous functions are functions that are defined but not assigned a name.

The following is an anonymous function that takes in two integers and returns the sum.

(x: Int, y: Int) => x + y

The resultant expression can be assigned to a val:

val sum = (x: Int, y: Int) => x + y

Anonymous functions are primarily used as arguments to other functions. For instance, the map function on a collection expects another function as its argument:

// Returns Seq("FOO", "BAR", "QUX")
Seq("Foo", "Bar", "Qux").map((x: String) => x.toUpperCase)

The types of the arguments of the anonymous function can be omitted: the types are inferred automatically:

Seq("Foo", "Bar", "Qux").map((x) => x.toUpperCase)

If there is just one argument, the parentheses around that argument can be omitted:

Seq("Foo", "Bar", "Qux").map(x => x.toUpperCase)

Underscores shorthand

There is an even shorter syntax that doesn't require names for the arguments. The above snippet can be written:

Seq("Foo", "Bar", "Qux").map(_.toUpperCase)

_ represents the anonymous function arguments positionally. With an anonymous function that has multiple parameters, each occurrence of _ will refer to a different argument. For instance, the two following expressions are equivalent:

// Returns "FooBarQux" in both cases
Seq("Foo", "Bar", "Qux").reduce((s1, s2) => s1 + s2)
Seq("Foo", "Bar", "Qux").reduce(_ + _)

When using this shorthand, any argument represented by the positional _ can only be referenced a single time and in the same order.

Anonymous Functions with No Parameters

To create a value for an anonymous function that does not take parameters, leave the parameter list blank:

val sayHello = () => println("hello")

Function composition allows for two functions to operate and be viewed as a single function. Expressed in mathematical terms, given a function f(x) and a function g(x), the function h(x) = f(g(x)).

When a function is compiled, it is compiled to a type related to Function1. Scala provides two methods in the Function1 implementation related to composition: andThen and compose. The compose method fits with the above mathematical definition like so:

val f: B => C = ...
val g: A => B = ...

val h: A => C = f compose g

The andThen (think h(x) = g(f(x))) has a more 'DSL-like' feeling:

val f: A => B = ...
val g: B => C = ...

val h: A => C = f andThen g

A new anonymous function is allocated with that is closed over f and g. This function is bound to the new function h in both cases.

def andThen(g: B => C): A => C = new (A => C){
  def apply(x: A) = g(self(x))
}

If either f or g works via a side-effect, then calling h will cause all side-effects of f and g to happen in the order. The same is true of any mutable state changes.

trait PartialFunction[-A, +B] extends (A => B)

Every single-argument PartialFunction is also a Function1. This is counter-intuitive in a formal mathematical sense, but better fits object oriented design. For this reason Function1 does not have to provide a constant true isDefinedAt method.

To define a partial function (which is also a function), use the following syntax:

{ case i: Int => i + 1 } // or equivalently { case i: Int ⇒ i + 1 }

For further details, take a look at PartialFunctions.