ContextQMD
Back to Plate

Coverage Priority Map Testing Review Non React Post Phase 3

docs/plans/2026-03-25-coverage-priority-map-testing-review-non-react-post-phase-3.md

53.0.57.2 KB
Original Source

Coverage Priority Map Testing Review Non React Post Phase 3

Inputs

  • Coverage source: lcov.info
  • Constraints:
    • temporary exclude /react
    • no browser or e2e coverage work
    • no coverage vanity
    • file-first ranking, not package theater

Coverage Run

  • Command: scoped bun test --coverage --coverage-reporter=lcov --coverage-dir=.coverage-repo-2026-03-25b --reporter=dots over the regenerated non-React file list in .claude/tmp/non_react_test_files.txt
  • Result: Fresh non-React coverage: 2784 pass, 0 fail, 566 files, 2.32s.
  • Important: the old scoped file list was stale. This pass regenerated it to 615 files before rerunning coverage. The earlier list would have lied about remaining gaps.

Suite Health

  • pnpm test:profile -- --top 25: green
  • pnpm test:slowest -- --top 25: green
  • fast suite result: 3022 pass, 0 fail, 614 files, 3.83s
  • average: 0.508ms
  • median: 0.212ms
  • slow-bucket thresholds: 75ms/test, 150ms/file total
  • threshold breaches: none
  • skipped-test debt scan: none worth fixing
  • commented-out spec scan: none worth fixing
  • cross-spec import scan: none

Scoring Rules

  • Scope is packages/**/src/**.
  • Exclude test files, barrels, declaration files, obvious type-only files, /react, and browser-package work under packages/playwright.
  • Penalize wrappers, tiny crumbs, DOM-heavy leftovers, schema dust, and giant serializer sludge.
  • Reward deterministic transforms, queries, parser or serializer seams, plugin contracts, plugin resolution, and bounded runtime helpers.
  • Missing-from-lcov runtime files are treated as uncovered. Pretending they are covered is clown math.
  • package_score is the sum of the top 5 remaining file scores in that package, not every leftover crumb.

Strong Take

Non-React coverage is basically spent.

There are no score >= 6 files left. The entire honest score >= 5 set is just five files, and four of them are one tiny basic-nodes parser-plugin lane. That is not a reason to reopen package theater. It is a reason to finish one last micro-batch if you care, then stop.

Threshold Counts

  • score >= 5: 5 files
  • score >= 4: 25 files
  • score >= 3: 70 files
  • score >= 2: 108 files
  • score >= 1: 210 files

Strict Next Batch

  1. docx: docxListToList.ts — score 5
  2. basic-nodes: BaseCodePlugin.ts — score 5
  3. basic-nodes: BaseStrikethroughPlugin.ts — score 5
  4. basic-nodes: BaseItalicPlugin.ts — score 5
  5. basic-nodes: BaseUnderlinePlugin.ts — score 5

Wider Optional Batch

  1. table: getSelectedCellsBorders.ts — score 4
  2. autoformat: AutoformatPlugin.ts — score 4
  3. core: DebugPlugin.ts — score 4
  4. udecode/depset: get-package-manager.ts — score 4
  5. table: getCellIndices.ts — score 4
  6. basic-nodes: BaseBoldPlugin.ts — score 4
  7. slate: deleteText.ts — score 4

Packages By Value

  1. basic-nodes — package score 24
  2. slate — package score 20
  3. docx — package score 18
  4. table — package score 17
  5. core — package score 16
  6. docx-io — package score 16
  7. list-classic — package score 16
  8. dnd — package score 15
  9. markdown — package score 12
  10. resizable — package score 12
  11. suggestion — package score 10
  12. cursor — package score 9

Best Files By Value

  1. docx — docxListToList.ts — score 5, coverage 78.9%, uncovered 8
  2. basic-nodes — BaseCodePlugin.ts — score 5, coverage 76.9%, uncovered 6
  3. basic-nodes — BaseStrikethroughPlugin.ts — score 5, coverage 83.3%, uncovered 4
  4. basic-nodes — BaseItalicPlugin.ts — score 5, coverage 82.6%, uncovered 4
  5. basic-nodes — BaseUnderlinePlugin.ts — score 5, coverage 82.6%, uncovered 4
  6. table — getSelectedCellsBorders.ts — score 4, coverage 95.1%, uncovered 12
  7. docx-io — settings.ts — score 4, coverage 0.0%, uncovered 11
  8. slate — hasDOMNode.ts — score 4, coverage 18.2%, uncovered 9
  9. autoformat — AutoformatPlugin.ts — score 4, coverage 88.9%, uncovered 8
  10. slate — hasEditableTarget.ts — score 4, coverage 20.0%, uncovered 8
  11. slate — hasSelectableTarget.ts — score 4, coverage 20.0%, uncovered 8
  12. slate — hasTarget.ts — score 4, coverage 20.0%, uncovered 8
  13. core — DebugPlugin.ts — score 4, coverage 90.6%, uncovered 5
  14. udecode/depset — get-package-manager.ts — score 4, coverage 81.5%, uncovered 5
  15. table — getCellIndices.ts — score 4, coverage 81.8%, uncovered 4

Stop Condition

Stop non-React coverage after the phase-4 roadmap below. The remaining misses after that are mostly DOM-only Slate crumbs, schema or type dust, wrapper/plugin residue, and giant low-ROI serializer sludge.

That is where more non-React coverage turns into percentage cosplay. Switch to React or architecture-safety work instead.

Caveats

  • Package totals are noisier than file totals. Trust the file ranking more.
  • basic-nodes looks high because a small parser-mark lane is still partially covered, not because the whole package needs a sweep.
  • slate looks inflated by DOM-editor leftovers. They are real files, but wrong phase for a final non-React cleanup lap.
  • docx-io still overstates itself because schema dust and html-to-docx.ts are uncovered but not high-ROI.

Full Data