Removes references to hidden elementes in rustdoc #66715
Conversation
Dylan-DPC-zz
added
the
S-waiting-on-review
| impl<'a> DocFolder for ReferencesStripper<'a> { | ||
| fn fold_item(&mut self, i: Item) -> Option<Item> { | ||
| if let clean::ItemEnum::MethodItem(ref method) = &i.inner { | ||
| if method.parent.map(|did| !self.retained.contains(&did)).unwrap_or(false) { |
There was a problem hiding this comment.
| if method.parent.map(|did| !self.retained.contains(&did)).unwrap_or(false) { | |
| if method.parent.map_or(false, |did| !self.retained.contains(&did)) { |
| impl issue_13698::Foo for Foo {} | ||
| // @!has - '//*[@id="method.foo2"]' 'fn foo2' | ||
| impl issue_13698::FooAux for Foo { | ||
| fn foo2(&self) {} |
There was a problem hiding this comment.
Nit: indent of 2 instead of 4. ;)
| impl Bar for Foo {} | ||
| // @!has - '//*[@id="method.qux"]' 'fn qux' | ||
| impl Bar for Foo { | ||
| fn qux(&self) {} |
There was a problem hiding this comment.
Nit: indent of 2 instead of 4. ;)
|
So far, code looks good to me (you even added tests!). However I wonder: if the child implementation of a trait wants to display a method implementation, how can it be done with this change? Currently, you have to re-use However, I'm open to other opinions on this topic. :) |
Oh, now I'm confused about the scope of the bug I'm supposed to be fixing 🙃 If I understood what you're saying, you are arguing that the current behavior is preferable since it is more flexible? I don't have a strong opinion, I assumed it was undesired because of the (repeated) chatter in #13698 about being unwanted/surprising. If that's the case, I guess this PR does not make much sense, and that in cases like #51147 (comment) what we want is to explicitly mark the implementations with |
|
Well, my point was that you can hide elements but you can't "unhide" them. ;) |
|
#13698 is a WONTFIX |
This is my first PR dealing with
rustc/rustdocinternals, usual warnings apply 😅This is a partial attempt to fix #13698. The approach is pretty blunt: we keep a reference to the defining item in a trait implementation, and we then strip all items with a "parent" marked as
#[doc(hidden)]. It is not finished but I wanted to gather feedback early.TODOs / Questions / Doubts / Alternatives
Method.Cleanimplementation that I have modified is called during thecollect_trait_implspass. I'm not sure whether it would be better to try to tackle hidden elements at that point, instead of waiting tostrip-hidden.parenta good name for defining item? Maybedefinitionwould be better?parentfield? Right now, I'm stuffing it into eachItemEnumvariant, but maybe we could just add it to theItemstruct.