implements PLN-0136
Intrepid/Prometeu/Studio/pipeline/pr-master Build started...
JaCoCo Coverage #### Project Overview No changes detected, that affect the code coverage. * Line Coverage: 63.75% (19407/30442) * Branch Coverage: 54.17% (7740/14289) * Lines of Code: 30442 * Cyclomatic Complexity: 12365 #### Quality Gates Summary Output truncated.
Test / Build skipped: 15, passed: 685
Intrepid/Prometeu/Studio/pipeline/head This commit looks good

Add PBS quick fixes as frontend-owned repairs transported by the LSP.
Duplicate reserved attributes are deleted at the extra span. Interface
InitAllowed attributes with arguments become [InitAllowed]. The server
matches Diagnostic.code plus range, does not parse messages, and returns
an empty list when the capability or the fix is absent.

Housekeep DSC-0044 with LSN-0073.
This commit is contained in:
2026-09-22 07:43:05 +01:00
parent 6f45f4f205
commit ae891262e1
27 changed files with 1182 additions and 70 deletions
@@ -1,62 +0,0 @@
---
id: AGD-0047
ticket: pbs-lsp-code-actions
title: PBS LSP Code Actions and Quick Fixes
status: open
created: 2026-07-15
resolved:
decision:
tags: [studio, lsp, vscode, compiler-pbs, editor, code-actions, quick-fix]
---
## Pain
Domain owner: `studio/lsp`
PBS diagnostics are visible, but users do not get actionable quick fixes for common compiler/editor problems, making simple repairs slower than necessary.
## Context
The LSP maps diagnostics but does not announce codeActionProvider. Compiler diagnostics already carry codes and ranges, but not every diagnostic has enough structured repair data to generate safe edits.
## Open Questions
- [ ] Which diagnostics should receive quick fixes first?
- [ ] Should code actions be produced from diagnostic codes only, or from structured compiler repair hints?
- [ ] How should imports, reserved attributes, missing methods, and invalid Doc shapes be prioritized?
## Options
### Option A - Diagnostic-code quick fixes
- **Approach:** Map known diagnostic codes to simple edits in the LSP layer.
- **Pro:** Fast for a small set of obvious repairs.
- **Con:** Couples fixes to diagnostic text/ranges and can become brittle.
- **Maintainability:** Moderate if limited to trivial edits.
### Option B - Compiler repair hints
- **Approach:** Extend diagnostics with structured repair metadata and let the LSP map hints to code actions.
- **Pro:** Keeps product logic close to compiler knowledge.
- **Con:** Requires a new diagnostic/repair contract.
- **Maintainability:** Strong for non-trivial fixes and future frontends.
## Tradeoffs
Quick fixes are only useful if safe. Simple code-based fixes can ship first, but imports and generated stubs need stronger compiler context.
## Recommendation
Prefer Option B as the target design, with Option A allowed only for a small bootstrap set of unambiguous repairs.
## Discussion
This should follow diagnostics UX and import assistance for the first high-value fixes.
## Resolution
Ainda em aberto.
## Next Step
Pick the first three diagnostic codes eligible for safe quick fixes.