ContextQMD
Back to Plate

Coverage Priority Map Testing Review Full Repo

docs/plans/2026-03-24-coverage-priority-map-testing-review-full-repo.md

53.0.57.7 KB
Original Source

Coverage Priority Map Testing Review Full Repo

Inputs

  • Coverage source: lcov.info
  • Constraints:
    • whole repo phase, including React
    • no browser or e2e theater
    • no package sweeps unless a package is still mostly untouched
    • exclude tests, barrels, declarations, snapshots, generated junk, and obvious type-only files
    • score only files still worth real regression detection during breaking changes

Coverage Run

  • Command: bun test --coverage --coverage-reporter=lcov --coverage-dir=.coverage-repo-2026-03-24l --reporter=dots
  • Result: Fresh coverage: 2956 pass, 0 fail, 592 files, 3.56s.

Suite Health

  • pnpm test:profile -- --top 25: clean. Nothing breached the fast-lane budget.
  • pnpm test:slowest -- --top 25: clean. Nothing breached the fast-lane budget.
  • Stale debt scans found no skipped-test junk, commented-out spec junk, or cross-spec imports worth fixing.
  • Audit caveat: the first full coverage run exposed leaked spies in BaseCodeBlockPlugin.spec.ts and withApplyTable.spec.ts. Those are fixed now, so this map is based on a clean run, not poisoned test order.

Scoring Rules

  • Scope is packages/**/src/**.
  • File ranking matters more than package totals.
  • Reward transforms, queries, parser or serializer seams, merge helpers, stateful hooks, and behavior-heavy utils.
  • Reward low runtime coverage and wide uncovered gaps.
  • Penalize wrappers, thin plugin entrypoints, provider or store dust, type-only files, snapshots, generated files, and giant low-ROI slabs.
  • package_score is the sum of the top 5 remaining file scores in that package. It is a rough map, not gospel.

Strong Take

Non-React is done. The next honest phase is React and state-heavy seams.

Do not do another package sweep. The remaining value is split across selection, ai, link, toggle, table, toc, emoji, and a few secondary packages.

The locked next phase is simple:

  1. Execute the full-repo roadmap at 2026-03-24-full-repo-coverage-roadmap.md.
  2. Finish Tier 1 first. That is every remaining score >= 9 file.
  3. Then finish Tier 2. That is every remaining score = 8 file.
  4. Stop and rerun coverage before touching the score-7 ring.

Threshold Counts

  • score = 10: 16 files
  • score >= 9: 40 files
  • score >= 8: 63 files
  • score >= 7: 84 files
  • score >= 6: 131 files
  • score >= 5: 167 files

Packages By Value

Raw package order from the fresh scoring:

  1. ai — package score 50, top files applyAISuggestions.ts:10; submitAIChat.ts:10; callCompletionApi.ts:10; applyTableCellSuggestion.ts:10; triggerCopilotSuggestion.ts:10
  2. selection — package score 50, top files useCursorOverlay.ts:10; useBlockSelectable.ts:10; getSelectionRects.ts:10; getCursorOverlayState.ts:10; getCaretPosition.ts:10
  3. toc — package score 46, top files useTocSideBar.ts:10; useContentController.ts:9; useContentObserver.ts:9; useTocObserver.ts:9; useTocController.ts:9
  4. emoji — package score 45, top files EmojiFloatingLibrary.ts:10; useEmojiPickerState.ts:9; EmojiFloatingGridBuilder.ts:9; EmojiFloatingGrid.ts:9; useEmojiPicker.ts:8
  5. toggle — package score 42, top files getLastEntryEnclosedInToggle.ts:9; isInClosedToggle.ts:9; useToggleToolbarButton.ts:8; useToggleButton.ts:8; findElementIdsHiddenInToggle.ts:8
  6. link — package score 40, top files triggerFloatingLinkInsert.ts:10; triggerFloatingLinkEdit.ts:9; triggerFloatingLink.ts:9; useFloatingLinkInsert.ts:6; submitFloatingLink.ts:6
  7. media — package score 37, top files history.ts:9; usePlaceholderPopover.ts:8; usePlaceholderElement.ts:7; getMediaType.ts:7; insertMedia.ts:6
  8. list — package score 36, top files useListToolbarButton.ts:8; useTodoListElement.ts:8; useTodoListToolbarButton.ts:8; toggleList.ts:6; getSiblingList.ts:6
  9. list-classic — package score 34, top files useTodoListElement.ts:8; useListToolbarButton.ts:8; moveListItemSublistItemsToListItemSublist.ts:6; getHighestEmptyList.ts:6; moveListItemDown.ts:6
  10. table — package score 34, top files useTableMergeState.ts:9; useCellIndices.ts:7; getTableCellBorders.ts:6; deleteRow.ts:6; deleteRowWhenExpanded.ts:6

Actual execution order is a bit stricter than the raw package table:

  1. selection
  2. ai
  3. link
  4. toggle
  5. table, media, combobox
  6. toc, emoji
  7. cursor, math, list, list-classic, callout, excalidraw

Reason: raw package totals overrate some hook clusters. The file queue is the real source of truth.

Best Files By Value

  1. ai — applyAISuggestions.ts — score 10, coverage 0.0%, uncovered 164
  2. selection — useCursorOverlay.ts — score 10, coverage 7.9%, uncovered 116
  3. ai — submitAIChat.ts — score 10, coverage 0.0%, uncovered 97
  4. ai — callCompletionApi.ts — score 10, coverage 0.0%, uncovered 90
  5. selection — useBlockSelectable.ts — score 10, coverage 5.8%, uncovered 81
  6. emoji — EmojiFloatingLibrary.ts — score 10, coverage 0.0%, uncovered 78
  7. selection — getSelectionRects.ts — score 10, coverage 2.7%, uncovered 71
  8. ai — applyTableCellSuggestion.ts — score 10, coverage 0.0%, uncovered 41
  9. ai — triggerCopilotSuggestion.ts — score 10, coverage 0.0%, uncovered 36
  10. link — triggerFloatingLinkInsert.ts — score 10, coverage 0.0%, uncovered 36
  11. ai — acceptAISuggestions.ts — score 10, coverage 0.0%, uncovered 31
  12. ai — rejectAISuggestions.ts — score 10, coverage 0.0%, uncovered 31
  13. ai — resetAIChat.ts — score 10, coverage 0.0%, uncovered 29
  14. selection — getCursorOverlayState.ts — score 10, coverage 11.5%, uncovered 23
  15. selection — getCaretPosition.ts — score 10, coverage 11.1%, uncovered 16

Caveats

  • Raw package totals inflate toc and emoji. Those packages have legit hook clusters, but they are not more important than selection, ai, or link.
  • core, slate, and markdown still show medium leftovers, but they are below the current phase threshold. Do not get distracted.
  • The remaining high-ROI work is mostly stateful React hooks and behavior-heavy utils. That is why the roadmap freezes a file threshold instead of another package march.
  • Trust the file TSV more than the package TSV.

Full Data