Rustc keeps suggesting std::iter::ZipImpl trait who has a "new" method.

[crlf0710]

This is not a real big deal, however rustc keeps suggesting "ZipImpl", which is a implementation detail and rarely is what people want.

Any chance to rename it somehow?

[Mark-Simulacrum]

Could you provide more context? A sample of the code and the error message suggesting ZipImpl would be helpful.

[ExpHP]

struct T;

fn main() {
    T::new();
}

   Compiling playground v0.0.1 (file:///playground)
error[E0599]: no function or associated item named `new` found for type `main::T` in the current scope
 --> src/main.rs:3:5
  |
3 |     T::new();
  |     ^^^^^^
  |
  = help: items from traits can only be used if the trait is implemented and in scope
  = note: the following trait defines an item `new`, perhaps you need to implement it:
          candidate #1: `std::iter::ZipImpl`

[durka]

Maybe the suggestions should filter out #[doc(hidden)] items?

[ExpHP]

oddly enough, ZipImpl isn't just #[doc(hidden)], but it is, in fact, private.

[zackmdavis]

Currently, we just check that at least one of the type and the trait is local, so indeed, this will false-positive for private traits:

rust/src/librustc_typeck/check/method/suggest.rs

Lines 445 to 452 in 565907f

	// we approximate the coherence rules to only suggest
	// traits that are legal to implement by requiring that
	// either the type or trait is local. Multidispatch means
	// this isn't perfect (that is, there are cases when
	// implementing a trait would be legal but is rejected
	// here).
	(type_is_local || info.def_id.is_local())
	&& self.associated_item(info.def_id, item_name, Namespace::Value).is_some()

[estebank]

CC #25358, #26454.