Note if the user wants to use associated items of a trait directly #111324
Conversation
|
(rustbot has picked a reviewer for you, use r? to override) |
rustbot
added
S-waiting-on-review
| diag.note(format!( | ||
| "desired associated item not found in trait `{}` if you want to use such one", | ||
| snippet, | ||
| )); | ||
| } else { | ||
| diag.note(format!( | ||
| "desired associated item not found in this trait if you want to use such one", | ||
| )); |
There was a problem hiding this comment.
You should probably just use a label here, instead of span_to_snippet.
| if in_path && res.len() == 0 { | ||
| if let Ok(snippet) = tcx.sess.source_map().span_to_snippet(self_ty.span) { | ||
| diag.note(format!( | ||
| "desired associated item not found in trait `{}` if you want to use such one", |
There was a problem hiding this comment.
I'm confused by this wording. It also seems redundant given the note that already exists: "no function or associated item named has found for trait object dyn HasNot in the current scope"
There was a problem hiding this comment.
This wording says that there is no direct associated item in this trait, and the dyn HasNot in the existing note causes confusion (even though dyn HasNot also contains associated item of HasNot)
There was a problem hiding this comment.
I'm not sure if I understand what you mean by this, and I still think that this note is not adding any new information over the existing note.
There was a problem hiding this comment.
How about this: "no desired associated item found in this trait, if you do not want to use a bare trait object here".
There was a problem hiding this comment.
The "trait object dyn HasNot" in the existing note will confuse users.
| @@ -3680,6 +3680,19 @@ impl<'o, 'tcx> dyn AstConv<'tcx> + 'o { | |||
| Applicability::MachineApplicable, | |||
| ); | |||
| } | |||
| // note if the user wants to use associated items of a trait directly | |||
| if in_path && res.len() == 0 { | |||
There was a problem hiding this comment.
Why are we checking res here?
There was a problem hiding this comment.
Because I think there is no need to make this note for trait objects with multi traits such as <TraitA + TraitB>::foo()
|
(I'm on mobile) could this use translatable diagnostics? |
I browsed this page about translation, but could not find such directory How do I use translatable diagnostics? Do I need to do anything other than add a new entry in |
| | ^^^ | ||
| | ^^^ no desired associated item found in this trait, if you do not want to use a bare trait object here |
There was a problem hiding this comment.
If this is the only suggestion I don't think it helps much. It's probably better to first do something different for when <dyn A as B>::something and dyn A: B does not hold. Because in that case we are certain that the user didn't want A::something to mean <dyn A as B>::something.
|
This PR is not enough. I think we should improve E0782 instead and open a topic on internals |
WaffleLapkin
removed
the
S-waiting-on-review
Fixes #111312
Furthermore, bare trait objects have been denied after edition 2021. We could not treat traits without
dynas bare trait objects. For such cases, there is no need to use bare trait objects asrecv_ty/self_ty(while usingNonedirectly). This will lead to stop resolving early but it may require a bit more works to resolve compatibility issues.