On unresolved names, should the compiler suggest relative paths (`super::…`) instead of absolute ones if possible?

[leonardo-m]

This is wrong code:

use std::iter::once;

mod foo {
    fn bar() {
        let _ = once(1);
    }
}

fn main() {}

Rustc currently (1.45.0-nightly 99cb9cc 2020-05-11) gives:

error[E0425]: cannot find function `once` in this scope
 --> ...\test.rs:5:17
  |
5 |         let _ = once(1);
  |                 ^^^^ not found in this scope
  |
help: consider importing one of these items
  |
4 |     use core::iter::once;
  |
4 |     use std::iter::once;
  |

warning: unused import: `std::iter::once`
 --> ...\test.rs:1:5
  |
1 | use std::iter::once;
  |     ^^^^^^^^^^^^^^^
  |
  = note: `#[warn(unused_imports)]` on by default

error: aborting due to previous error; 1 warning emitted

But I think it could suggest to write super::once(1); instead.

[RalfJung]

Is using parent module imports like that considered good style? I'd rather add a use in the mod, to make each mod self-contained.

[leonardo-m]

If you think this is a suboptimal idea then I'll close this issue down.

[RalfJung]

In no way do I have the power to discard this idea outright.^^ I just provided my input -- before/during an actual implementation of this, the relevant team(s) should be consulted, that is all.

[leonardo-m]

It's OK, thank you. (In the title of this issue I've added a question mark because this is just a moderate suggestion, I am not sure it's worth having. Other my enhancement requests are more "sure" and don't have a question mark).

[RalfJung]

Raising the question certainly makes sense so thanks for the suggestion. :)