Auto completion of `async fn xxxx` for desugared `async fn in trait` form

[hyf0]

Hi. I'm thinking about improving the dx of writing impl DesugaredAsyncTrait for Foo. So far the rust analyzer will only infer auto completion like

However, I found that the compiler will allow the following code compiles. If this is an expected behavior(It's expected, see the following zulipchat link), maybe it's ok to make rust analyzer to support auto completion that starts with async fn foo( the sugar version rather than only desugared version like now) for DesugaredAsyncTrait .

This will surely improve the dx of writing impl DesugaredAsyncTrait for XXX greatly without the need to write impl Future<Output=XXXX> + Send over and over again(Though there are also auto completed by the rust-analyzer :).

use std::future::Future;

trait DesugaredAsyncTrait {
    fn foo(&self) -> impl Future<Output = usize> + Send;
    fn bar(&self) -> impl Future<Output = usize> + Send;
}

struct Foo;

impl DesugaredAsyncTrait for Foo {
    fn foo(&self) -> impl Future<Output = usize> + Send {
        async { 1 }
    }

    //
    async fn bar(&self) -> usize {
        1
    }
}

fn main() {
    let fut = Foo.bar();
    fn _assert_send<T: Send>(_: T) {}
    _assert_send(fut);
}

Related:

• https://rust-lang.zulipchat.com/#narrow/stream/122651-general/topic/Auto.20infered.20.60Send.60.20in.20.60impl.20DesugaredAsync.60.2E.20Expected.3F
• https://x.com/Gankra_/status/1141409682017308672

[hyf0]

There's a question on Quick Fix "Implement missing members" if we support this PR.

On "Implement missing members", should we fill them in sugar version or original form for async fn in trait in desugar form?

[Veykril]

I'd say sugared version, and then if we don't have it yet, have a desugar assist that goes from async to impl Future (I think we already have this though)

[lnicola]

Fixed in #17737