Biome fix pass scope—safe versus unsafe edits in automation

Answer

Run only safe fixes in CI and pre-commit hooks. Use the safe-edit flag appropriate for your Biome version (e.g., --write or --apply). Unsafe fixes should never be automated; they must be applied manually on a dedicated branch, followed by a full test suite execution and a manual diff review before merging.

Confirmed Facts

• Safe Fixes: Flags like --write or --apply target edits intended to preserve runtime semantics.
• Unsafe Fixes: Flags such as --unsafe or --apply-unsafe allow edits that may alter the program's behavior.
• No Internal Backups: Biome modifies files in-place without snapshots; version control (Git) is the only mechanism for rollback.
• Diff Volume: Even safe fixes can generate large diffs, such as when reordering imports or applying global formatting changes.

Likely Explanation

The risk of introducing subtle runtime regressions via unsafe fixes outweighs the convenience of automation. Because safe fixes are designed to be semantic-preserving, they are suitable for pre-commit hooks to maintain style consistency. However, because "safe" is a heuristic and "unsafe" is explicit, gating the latter behind a human review ensures that behavioral changes are intentional and verified by the test suite.

Implementation Steps

1. Identify Version: Check your package.json or biome --version to determine the correct flag set.
2. Configure Automation: Add the safe-fix command to your CI pipeline or pre-commit hook:

   # For Biome ≥1.0
   biome check --write
   
   # For older versions
   biome check --apply
   

3. Verify Changes: Use git diff to inspect the scope of automated edits.
4. Validate Runtime: Run the project's full test suite to ensure no regressions were introduced by the safe pass.
5. Handle Unsafe Edits: If unsafe fixes are desired, run them on a separate branch, perform a line-by-line diff review, and re-run all tests before merging.

Missing Diagnostic Detail

Please confirm the specific Biome version used in your project, as the transition from --apply to --write and the naming of unsafe flags vary across major releases.