Shorten projection types in inlay hints

[WaffleLapkin]

In examples like this, where we have a projection type, which is the only associated item with its name in a trait that is in scope:

trait Trait {
    type Assoc;

    fn f(self) -> Self::Assoc;
}

fn f<T: Trait>(x: T) {
    let y /*here*/ = x.f();
    _ = y;
}

Show T::Assoc instead of <T as Trait>::Assoc.

Rationale: making inlay hints less verbose is generally good, Trait is not that useful there.

---

I'm not sure if the implementation is good and correct, but it seems to be working at least 😅

[lowr]

While I understand the motivation and the logic looks correct, I'm a bit concerned about the performance here; iterating through every associated item of every trait in scope sounds a bit expensive especially for large files. I guess it's okay for typical codebases as it's fairly rare to show projection types, but we might want to assess the effect or add a configuration to opt it out.

Another thing is that it might be somewhat confusing unless the self type of the projection is either placeholder type or ADT. T::Assoc and Vec<i32>::Assoc seem nice, but (T, U)::Assoc, impl Trait + Clone::Assoc, and &dyn Trait + Send::Assoc look foreign to me. Also, <&T as Trait>::Assoc would be displayed as &T::Assoc, which is indistinguishable from &<T as Trait>::Assoc.

[bors]

☔ The latest upstream changes (presumably #13860) made this pull request unmergeable. Please resolve the merge conflicts.

[bors]

☔ The latest upstream changes (presumably #13946) made this pull request unmergeable. Please resolve the merge conflicts.

[Veykril]

I agree with lowr's comment here, this does seem like a performance pitfall and the ambiguity cases mentioned aren't too nice either. So I am inclined to close this for now.

[WaffleLapkin]

Just to be clear: is this an issue of "the implementation is not good enough, needs more work" or "this looks hard/impossible to do in a good way in general"?

[Veykril]

I would say, more work required. I do think having this wouldn't be too bad, but I also don't know if this can be done in a way that is not too performance heavy (inlay hints already take way too long to calculate as is). So unfortunately I can only give a vague answer here :)

[WaffleLapkin]

Okay, I don't have enough time/energy to work on this, but hopefully I'll eventually try to revive this :)