Allow less than all the generic parameters to be specified

[strega-nil]

fn foo<T, U>(u: U) -> T
  where T: From<U>
{
  From::from(u)
}

fn main() {
  println!("{}", foo(0u16): u16); // works
  println!("{}", foo::<u16>(0u16)); // doesn't work
  println!("{}", foo::<u16, _>(0u16)); // works
  println!("{}", foo::<u16, u16>(0u16)); // works
}

I want the second one to work (and to be equivalent to the third). There's no real reason for it not to, and it does work in C++.

[sfackler]

#1196

[strega-nil]

It seems to me that that was decided against due to type defaults? has anything else happened with those?

[est31]

Would implementing support for this be enough?

fn foo<T, U>(u: U) -> T
  where T: From<U>
{
  From::from(u)
}

fn main() {
    println!("{}", foo::<u16, ..>(0u16));
}

This would signal at the call site "more generic params to come" but also allow longer chains of generic parameters to be omitted. (it would allow you to have a function foo<T,U,V,W,X,Y> and still write foo::<u16,..>).

[mark-i-m]

It seems like this would only work if you only need to specify a prefix of the list of types, right? If you only wanted to specify U in your example, you would still be stuck, right?

foo::<_, u16>(0u16)

Perhaps a more flexible (and more dangerous) option would be to make the behavior "hey, compiler, if you get stuck solving for a type, take this as a hint"... I don't know how this would work in general, though...

[KodrAus]

It seems to me that that was decided against due to type defaults? has anything else happened with those?

Haven't type parameter defaults been phased out for functions?

As a motivation, I would like this for cases where I have one 'important' type parameter that may not be elided and a bunch that can be:

fn request<TDocument TIndex, TId>(index: TIndex, id: TId) -> RequestBuilder<TDocument> 
    where TIndex: Into<Index>, 
          TId: Into<Id> 
{ 
}

let req = request::<MyDocument>("some_index", "some_type");

In the above case I don't mind only being able to elide parameters on the right of ones that are explicitly specified.

[strega-nil]

@KodrAus right, that's what I'm talking about. The whole .. thing, I don't like. It's unnecessary.

[Centril]

As we closed #2176 pending more investigation, I'm also closing this in the interest of having fewer open issues. (This does not mean that we do / don't want this feature).