Back to Plate

Coverage Priority Refresh

docs/plans/2026-03-24-coverage-priority-map.md

53.0.54.8 KB
Original Source

Coverage Priority Refresh

Inputs

  • Coverage source: lcov.info
  • Constraints: exclude /react, ignore coverage vanity, score only real non-React unit-test seams under packages/**/src/**.
  • Scoring signals: seam type, uncovered lines, coverage ratio, recent completed passes, package signal penalty.

Coverage Run

  • Command: bun test --coverage --coverage-reporter=lcov --coverage-dir=.coverage-repo-2026-03-24a --reporter=dots
  • Result: 2643 pass, 0 fail, 493 files, 2.45s

Take

  • Strong take: score >= 6 is the honest next batch.
  • score >= 5 is the wider batch if you want one more sweep before rerunning coverage again.
  • After excluding /react, the real remaining value is concentrated in @udecode/depset and table. Everything after that is smaller leaf work or recently-covered leftovers.

Threshold Counts

  • score >= 8: 1 files
  • score >= 7: 8 files
  • score >= 6: 9 files
  • score >= 5: 18 files
  • score >= 4: 32 files
  • score >= 3: 50 files
  • score >= 2: 78 files
  • score >= 1: 127 files

Packages By Value

  1. udecode/depset — priority 444, top file 10, files >=6: 4, files >=5: 4
  2. table — priority 210, top file 7, files >=6: 1, files >=5: 10
  3. cursor — priority 111, top file 7, files >=6: 1, files >=5: 1
  4. caption — priority 111, top file 7, files >=6: 1, files >=5: 1
  5. callout — priority 111, top file 7, files >=6: 1, files >=5: 1
  6. excalidraw — priority 111, top file 7, files >=6: 1, files >=5: 1

Strict Batch (score >= 6)

Wider Batch (score >= 5)

Notes

  • Recent March 22-24 passes were explicitly penalized so the ranking does not keep resurfacing already-swept packages.
  • /react files were excluded, not merely ranked lower.
  • The TSV outputs contain the full package matrix and the full file matrix for exact triage.