SCIP index symbols for struct/enum/const/static/impl nested in functions do not include function name

[mgsloan]

This is demonstrated in tests included in #18758

The reason I did not fix this in that PR is that hir representation only appears to know what module these definitions are in, and not whether they are contained in a function. I am not familiar enough with rust-analyzer internals to provide such information. The fix would involve a change to the enclosing_definition method or the methods it calls.

[ChayimFriedman2]

The information whether something is inside function (or anything with body) is whether the module is a block module (this exists in hir-def, but adding it to hir should be easy enough). The problem is that block modules usually have no names. Handling it can be done by special-casing block modules to look at the name of the enclosing definitions.

[mgsloan]

@ChayimFriedman2 Thanks for the info, glad this is in the representation. I've attempted to write a method to get the info, but got stuck going from BlockId to the containing FunctionId. Maybe I could find the function name in the syntax tree... But it would be much cleaner to get a FunctionId.

    pub fn function_parent(self, db: &dyn HirDatabase) -> Option<Function> {
        let containing_block = self.id.containing_block()?;
        let node = containing_block.lookup(db).ast_id.file_syntax(db);
    }

Would really appreciate it if on of you rust-analyzer experts would solve this part - feel free to push a commit to this branch adding such a function. I'm happy to make the moniker.rs changes based upon it.

[Veykril]

I believe we don't actually have an association from block to its container, only block to its parent module (which may be a block again).

[ChayimFriedman2]

Indeed; but we do preserve an association between a block and its ast::BlockExpr, and we can use that (ancestors_with_macros()) to find the container.

[mgsloan]

Thanks for the further info @ChayimFriedman2 ! I tried for about 15 minutes to sort out how to write the code for this by referencing this seemingly related code I found by searching for ancestors_with_macros.

I would prefer to not block merging #18758 on this concern. It takes SCIP's symbol information from being wildly out of spec to only having a few known flaws. If we are blocking merge on this concern, then I would really appreciate it if someone that knows rust-analyzer better than me could update enclosing_definition on my branch to return functions when things are nested within them, instead of skipping over the function definitions.

[ChayimFriedman2]

I don't think we should block on that either.