Revert "RFC to require impl MyStruct to be nearby the definition of MyStruct"

[eddyb]

The "Revised UFCS proposal", PR #132, makes the restriction imposed by PR #155 unnecessary:

Whenever a path like ...::a::... resolves to a type (but not a trait), it is rewritten (internally) to <...::a>::....

This applies to all non-trait impls, and because <T>::item has to be handled during type-checking, that removes any need to handle impls (or treat them as modules, for that matter) in resolution, eliminating
all the issues #155 was trying to solve by being overly conservative.

[eddyb]

See also this comment about some of the consequences of UFCS.

[Kimundi]

Sounds good!

[nikomatsakis]

I agree with the sentiment, however there remains some amount of impl work to be done before the arguments in the original RFC no longer apply, no?

[eddyb]

Yes, hopefully I'll get to it in the next couple of weeks (I already figured out the details).

[netvl]

Does this mean that it will be possible to write something like

pub struct SomeType { ... }

mod some_functionality {
    use super::SomeType;

    impl SomeType {
        fn do_something(&self) { ... }
    }

    fn free_func() {
        let s: SomeType = ...;
        s.do_something();
    }
}

or maybe even

pub struct SomeType { ... }

mod some_functionality {
    use super::SomeType;

    impl SomeType {
        fn do_something(&self) { ... }
    }
}

fn free_func() {
    let s: SomeType = ...;
    s.do_something();
}

? If so, it is great! It will help structuring programs immensely.

[eddyb]

@netvl Both work before UFCS if you remove the restriction. I had a PR doing this a while back, it required amending the RFC and I never got back to it.
The difference between those and an impl in the same module is that SomeType::do_something wouldn't work right now (so you only got "extension methods" with that modification of mine).
With UFCS, SomeType::do_something would desugar to <SomeType>::do_something so you couldn't tell them apart.

[brson]

👍 but instead of deleting the original RFC, @eddyb can we have a new RFC that reverses the design? It can be pretty short. I'd prefer to leave the RFC for historical purposes.

[eddyb]

@brson I guess it could be something small explaining how UFCS allows resolution to ignore impls altogether. Will post that once I have an actual implementation.

[brson]

Merged.