We allow invalid inner attributes to be parsed, e.g. inner attributes that are
not directly after the opening brace of the match block.
Instead we run validation on `MatchArmList` to allow better reporting of errors.
692: [WIP] Correctly parse attributes r=matklad a=DJMcNab
Reference - https://doc.rust-lang.org/reference/attributes.html
This fixes/investigates inner attributes for:
- [x] `impl` blocks
- [x] `extern` blocks
- [x] `fn`s (fixes#689)
- [x] `mod`s (already supported)
- [x] 'block expressions' (the long text just describes all 'blocks' used as statements)
This also investigates/fixes outer attributes for:
- [ ] 'most statements' (see also: #685, https://doc.rust-lang.org/reference/expressions.html#expression-attributes)
- [x] Enum variants, Struct and Union fields (Fixed in #507)
- [ ] 'Match expression arms' (@matklad can you provide a test case which explains what this means?)
- [ ] 'Generic lifetime or type parameters'
- [ ] 'Elements of array expressions, tuple expressions, call expressions, tuple-style struct and enum variant expressions'
- [ ] 'The tail expression of block expressions'
Co-authored-by: DJMcNab <36049421+djmcnab@users.noreply.github.com>
The new rowan uses arena allocator for red nodes which helps
slightly (5% on std completion test) with performance but should help
quite a bit with avoiding memory fragmentation.
630: Fill in DocumentSymbol::detail r=matklad a=hban
Closes: #516
I just pulled type text from the syntax node and "formatted" is bit. VS Code can't really handle multi-line symbol detail (it's will crop it when rendering), so that formatting will just collapse all white-space to singe space. It isn't pretty, but maybe there's a better way.
Issue also mentions "need to be done for `NavigationTarget` to `SymbolInformation`", but `SymbolInformation` doesn't have detail field on it?
Co-authored-by: Hrvoje Ban <hban@users.noreply.github.com>
623: WIP: module id is not def id r=matklad a=matklad
This achieves two things:
* makes module_tree & item_map per crate, not per source_root
* begins the refactoring to remove universal `DefId` in favor of having separate ids for each kind of `Def`. Currently, only modules get a differnt ID though.
Co-authored-by: Aleksey Kladov <aleksey.kladov@gmail.com>
