Equality constraints on associated constants

[Centril]

We currently do not support any way to constrain an associated constant as needing to be equal to some constant value. The natural way to achieve this would be to, by analogy with e.g., Foo<0>, also allow Foo<ID = 0>. The same disambiguation rules would apply as for Foo<X> in terms of when a block { ... } is needed, e.g., Foo<ID = { N + 1 }>.

In compiler internal terms, we would redefine ast::AssocTyConstraintKind as:

/// The kinds of an `AssocConstraint`.
#[derive(Clone, RustcEncodable, RustcDecodable, Debug)]
pub enum AssocConstraintKind {
    /// E.g., `A = Bar` in `Foo<A = Bar>`.
    EqualityTy { ty: P<Ty> },
    /// E.g., `A = 0` in `Foo<A = 0>`.
    EqualityConst { ct: AnonConst },
    /// E.g. `A: TraitA + TraitB` in `Foo<A: TraitA + TraitB>`.
    Bound { bounds: GenericBounds },
}

cc @varkor @eddyb

[Centril]

I've assigned myself, as I will be fixing the parser / lowering aspects of this to start with.

[dr-orlovsky]

Any progress on this issue? I see the required changes were made in March, so it seems to be possible to add this feature for now?

[varkor]

I believe only (some of) the parsing aspects were addressed, and nothing more. There is still implementation left before this feature is even ready for a feature gate.

[JulianKnodt]

@varkor (cc @Centril) I can pick up the implementation of this (I actually ran into this issue within 2 hrs of starting a project), unless there's some reason otherwise.

[dtolnay]

This has been implemented in nightly, so I am closing this feature request in favor of the tracking issue at #92827.

Per #92827 (comment), a lang team MCP is eventually going to be required, along with some medium sized doc explaining what is going on and why. @oli-obk will open a hackmd for folks to collaborate in.