a more readable type lambda trick

Dan Rosen — Sat, 31 Mar 2012 09:43:58 -0700

So you've got some type that takes multiple type parameters, like:

Code:

trait State[S, A]

... and you've got some code that needs to fix one of the type parameters while letting the other(s) vary. We usually use the "type lambda trick" to partially apply type arguments, like this:

Code:

def stateMonad[S] = new Monad[({type L[A] = State[S, A]})#L] { ... }

Sometimes you see this written with actual Greek characters using Unicode, where "L" is written as a lower case lambda, hence the name. Using Greek letters in source code is for jerks, unless you happen to actually be Greek, in which case go for it. However, even with the Latin "L" I don't find this style of code particularly readable, when some type lambda needs to be given a name and referenced more than once. So I propose the following new convention:

Code:

def stateMonad[S] = {
  type `State[S]`[A] = State[S, A]
  new Monad[`State[S]`] { ... }
}

The backticks permit use of the special [] characters which wouldn't otherwise be allowed in an identifier. At any rate, the idea is to mimic the appearance of multiple type parameter lists without yet having them in the Scala language.

Marakana - Scala - Examples Forum

a more readable type lambda trick