Ability to specify self parameter in static method call

[erickt]

The commit 8693943 removed the ability for us to do:

trait SizeOf {
     fn size_of() -> uint { sys::size_of::<T>() } 
}

fn size_of<T: SizeOf>() -> uint {
    SizeOf::size_of::<T>()
}

@pcwalton has proposed that we add support for expressing the Self type of a static method with:

fn size_of<T: SizeOf>() -> uint {
    SizeOf::<for T>::size_of()
}

but I've heard that we haven't settled on this syntax yet.

[dwrensha]

+1

[SimonSapin]

Is this related to #6894?

[brson]

Nominating

[catamorphism]

1.0, high

[nikomatsakis]

Although I initially proposed this syntax, it sure is ugly. I am still worried we're making a mistake by not supporting 'C++ style'.

[nikomatsakis]

This is also related to the question of whether we want to support multiparameter typeclasses in their full generality.

[SimonSapin]

Shouldn’t this be just T::size_of()?

[pnkfelix]

@SimonSapin not currently; see #6894

[SimonSapin]

@pnkfelix So, we can’t do T::size_of() yet because it requires a bunch of work, but that’s what we want to do eventually. Do I understand this correctly?

[pnkfelix]

@SimonSapin yeah, we want to support T::size_of() eventually, or something very close to it.

[pnkfelix]

I think we want to do this for 1.0. It came up in a recent weekly meeting. Our plan, I think, is to keep supporting the Trait::size_of() invocation form (so its not a backward compatibility risk), but we probably want to add the T::size_of() (with the potential for compile-time error if there is a collision/name-ambiguity) and unambiguous SizeOf::<for T>::size_of() invocation forms for 1.0 as well.

Nominating, presuming that it will get P-high.

[pnkfelix]

Wait, this is already P-high ...

[pnkfelix]

Ah, the description didn't cover the T::size_of() case. I'll check if there's another ticket for that.

[pnkfelix]

The T::size_of() form is in #6894 which was recently nominated.

[nrc]

working on this

[brson]

@nick29581 Do you know what the state of this is? I feel like I haven't heard much about it. Does UFCS solve it?

[nrc]

@brson Yes, UFCS should solve this. I'm not sure who is implementing now.

[aturon]

Note that UFCS is not considered 1.0 blocking, but this issue is.

Nominating: we need to decide whether this functionality blocks 1.0 or not. (Also, this ticket can likely be closed as a dup of #16293).

[pnkfelix]

Would reassign as P-high, not on 1.0 milestone.

But really, just closing as dupe of #16293.