Tracking Issue for De-RFC 3307: Remove type ascription

[Manishearth]

This supersedes #23416

This is a tracking issue for the RFC "3307" (rust-lang/rfcs#3307).

About tracking issues

Tracking issues are used to record the overall progress of implementation.
They are also used as hubs connecting to other relevant issues, e.g., bugs or open design questions.
A tracking issue is however not meant for large scale discussion, questions, or bug reports about a feature.
Instead, open a dedicated issue for the specific matter and add the relevant feature gate label.

Steps

• Implement the RFC:
  • Remove type ascription from the parser and diagnostics
  • @rust-lang/compiler: Do y'all have a preference as to how far the removal should go? Unparseable AST nodes could be left in the code, for example (perhaps only instantiable by internal macro)
• Adjust documentation: Remove it from the unstable book
• (No stabilization PR necessary)
• Remove the feature gate.

Unresolved Questions

• Should this be completely removed by the compiler, or left behind in a way that cannot be directly accessed through Rust syntax (or requires using a wrapper macro)?
• [ ]

Implementation history

[Noratrieb]

@rustbot claim

I'll remove all the parsing and diagnostic logic/hacks but leave the AST nodes and everything after it behind for now.

[RalfJung]

Would it make sense to possibly replace type ascription by an unstable macro in the standard library?

macro_rules! type_ascribe {
    ($e:expr, $t:ty) => {{
        let val: $t = $e;
        val
    }};
}

I noticed #104614 added a magic macro that preserves the old semantics, but maybe we can implement this entirely in userland?

[bjorn3]

That macro shortens the lifetime of temporaries created in the expression.

[Skgland]

Would a function eliviate this problems with temporaries?
Or am I missing more temporary stuff?

pub fn type_ascribe<T>(value: T) -> T { value }

#[allow(unused_macros)]
macro_rules! type_ascribe {
    ($e:expr, $t:ty) => {{
        let val: $t = $e;
        val
    }};
}

pub fn main() {
    // macro dosn't work because of temporaries
    // println!("{}", type_ascribe!(String::from("hello world").split_once(' '), _).unwrap().0)
    
    //  identity function with turbo fish works
    println!("{}", type_ascribe::<_>(String::from("hello world").split_once(' ')).unwrap().0)
}

playground

[RalfJung]

That can be wrapped in a macro, of course:

pub fn type_ascribe<T>(value: T) -> T { value }

#[allow(unused_macros)]
macro_rules! type_ascribe {
    ($e:expr, $t:ty) => {type_ascribe::<$t>($e)};
}

pub fn test(x: usize) {
    println!("{}", type_ascribe!(String::from("hello world").split_once(' '), _).unwrap().0);
    let v = type_ascribe!([[0, 1, 2, 3], [4, 5, 6, 7]][x], [u32; 4]);
}

[Noratrieb]

This has taken a lot more time than I thought since lots of diagnostics were relying on it in subtle ways and require restructuring. I don't have enough time to work on this and also the other things I'm working on so I'm dropping my assignment here, feel free to take it. There's a WIP PR #106826 which may or may not be useful.
@estebank maybe you're interested in working on it, given all the pains you've had with type ascription :D
@rustbot release-assignment