CODEWITHSUNDEEP
genericsscalatypesexistential-type
soc
soc

Reputation: 28433

Can all usages of `forSome` be replaced by an equivalent usage of `_`?

For instance, List[T] forSome { type T } is equivalent to List[_], but is this true for every possible usage of forSome or are there cases where forSome cannot be replaced by an equivalent of the second syntax?

Upvotes: 12

Views: 279

Answers (2)

oxbow_lakes
oxbow_lakes

Reputation: 134270

No, not all usages can be thus converted. Something like this (thanks to retronym, below, who should be getting the upvotes on this one)

def foo(xs: Map[T, T] forSome { type T})

The point here is that I can use the existential in more than one place but it is the same existential. Obviously, if you used wildcards, you would be referring to different existentials

Upvotes: 15

retronym
retronym

Reputation: 55028

_ in this context is the "Placeholder Syntax for Existential Types", described in 3.2.10 of the language specification.

The name sounds a lot like another use of _, "Placeholder Syntax for Anonymous Functions". This shorthand can't be used for all functions. For example:

(a: A) => f(g(a))
f(g(_: A)) // WRONG, this means f((a: A) => g(a))

We can construct a similar example for existentials, which can't be represented _:

F[G[A]] forSome { type A }
F[G[_]] // WRONG, this means F[G[A] forSome { type A }]

Side note: existentials are not the type level analog of value level functions. That would be:

({type l[A] = F[G[A]])#l

In fact it's a pity that Scala uses the underscore for existentials, rather than type level functions.

UPDATE

Let me frame @oxbow_lakes' answer in the same way:

(a: A) => h(a, a)
h(_, _) // WRONG, this means (a, b) => f(a, b)

H[A, A] forSome { type A }
H[_, _] // WRONG, this means H[A, B] forSome { type A; type B }

Upvotes: 14

Related Questions