fix: Goto implementation to impls inside blocks #16812
Conversation
rustbot
added
the
S-waiting-on-review
| // Collect nested impls inside function bodies, etc. | ||
| for (_, ns) in module_data.scope.entries() { | ||
| for module in ns.types.iter().map(|t| t.0).chain(ns.values.iter().map(|t| t.0)) { | ||
| let body: Option<DefWithBodyId> = match module { | ||
| ModuleDefId::FunctionId(it) => Some(it.into()), | ||
| ModuleDefId::EnumVariantId(it) => Some(it.into()), | ||
| ModuleDefId::ConstId(it) => Some(it.into()), | ||
| ModuleDefId::StaticId(it) => Some(it.into()), | ||
| _ => None, | ||
| }; | ||
| if let Some(body) = body { | ||
| let body = db.body(body); | ||
| for (_, block_def_map) in body.blocks(db.upcast()) { | ||
| Self::collect_def_map(db, map, &block_def_map, query_depth); | ||
| } | ||
| } | ||
| } | ||
| } |
There was a problem hiding this comment.
The logic is quite verbose, but I couldn't find any good ways that can probe into those blocks 🤔
There was a problem hiding this comment.
We intentionally do not look into all blocks there are, as an IDE we try to be lazy and only analyze the crate structure on a global scale, which allows us to skip out on having to look into every body which would slow us down immensely. So supporting local impls like this isn't really possible.
So what we should do instead to make the first case work is have goto_implementation look into the block defmaps of the ADT/trait definition and iterate those. We can query the relevant impls via the *_impls_in_block queries (check for uses of those queries, that should show you how to use thisI think, if not just ask)
| } | ||
|
|
||
| #[test] | ||
| fn goto_implementation_in_nested_blocks_2() { |
There was a problem hiding this comment.
This test we cannot support due to performance reasons, see the main review message
| } | ||
|
|
||
| #[test] | ||
| fn goto_implementation_in_nested_blocks_1() { |
Veykril
added
S-waiting-on-author
|
I'm seeking some performant implementation; is there any way to find |
|
You can turn block expression ast nodes into blocks via (Note that will only yieldSome for blocks that actually have local items or macro calls in them due to perf optimizations)
Though I tihnk it should be enough to just query the modules of the trait / ADT in question as those should be block modules if they are defined in blocks? |
|
Thank you! I'll look into that point 😄 |
ShoyuVanilla
force-pushed
the
issue-3739
branch
from
c7fde4e
to
536c10b
Compare
ShoyuVanilla
force-pushed
the
issue-3739
branch
from
536c10b
to
967a864
Compare
|
Thanks! |
|
☀️ Test successful - checks-actions |
Fixes #3739