Hi there,
we would be interested in an improvement of AI Codefix:
In legacy codebases with large number of issues, or when new rules are introduced triggering many findings in a codebase, it is not feasable to apply AI Codefix one by one for each finding.
We experimented with external Coding Agents to fix those in bulks, but the results are much less reliable, then the suggestions made by your AI Codefix.
Instead, a bulk-apply operation within SonarQube is needed, to apply AI Codefixes to a larger number of issues at once, creating a single branch for these, and allowing a single PR Review for all those fixes.
This likely should not auto-include ALL issues, but should allow for more fine granular control, to not overwhelm the reviewers.
Following could be feasable options:
- Allow fixing all issues of a specific issue type/rule violation (one PR per type makes sense, focusing on specific issues, this would also allow to prioritize issue types)
- Bulk select issues in UI, apply fixes for all selected ones, allowing to add further issues from other screens/pages to the same branch
- Option 1) but with an additional MAX COUNT filter (e.g.: Apply fixes for the first 200 findings of a specific type)
- Like 1) but multi-selecting issue types (fix all issues of multiple selected rules/types), optionally together with 3).
- In all options of 1-4) it should be allowed to apply further fixes in additional steps to the same fix-branch, so one can step by step fixes for whatever violations he wants (in bulks)
Could this be something you can consider implementing? This would also help to cleanup maintainability drifts from time to time, which can be easily done by Junior devs or interns using such approach.