Lift the Layered Dependency View's rules into a tested module

[vincentmakes]

Summary

The Layered Dependency View (Dependencies report, a card's dependency section, TurboLens assessments) decided how its cards and lines look and behave inline in components that only a React Flow render can reach, so the nightly scores ~1,100 survivors in LayeredDependencyView.tsx. Those rules move into ldvViewModel.ts with literal, rule-by-rule tests, and the two layouts' triplicated edge assembly collapses onto shared helpers. Nothing a user sees changes. Fourth of the four approved PRs (after #1240, #1241, #1242, all merged); main is merged in, so it carries 2.158.12 above 2.158.11.

Type

• feature
• fix
• refactor
• docs
• chore (CI / build / dependency / housekeeping)
• security

Changes

• New frontend/src/features/reports/ldvViewModel.ts (no React at runtime), taking from the view:
  • card: the type-colour fallback, the tint (12 % / 22 % / 6 % proposed), the lifecycle dot colour, the time-travel accent, the border, the planned/retired badge and its slot, type-icon placement beside the dot, the name's line clamp, handle visibility (mirrored ts-/bt- handles share their base's), the handle's position along its edge, and the 5 px click-versus-drag slop;
  • line: when a line is lit, its colours (a severed line stays red when lit), the 24-character label cut, the label's width, the connector and line stroke widths, and the label placement search (now fed a pointAt(t) function instead of an SVG element);
  • canvas: type counts, "did the type filter empty the view", the background cycle, the export file name, what a card click does in each mode (cardClickAction), whether a line touches the hovered card, and paint order.
• LayeredDependencyView.tsx renders from those answers; markup, sx, i18n keys and the export filter are unchanged.
• ldvLayoutShared.ts gains flowMarkers, relationSevered and routeEdgeData, replacing three copies of the same arrowhead, severed and route-field code in buildLdvFlow and both passes of the aggregate builder. The arrow colours are named once (ARROW_COLOR, SEVERED_ARROW_COLOR, same values).
• ldvEdgeRouting.computeAbsPos is deleted: it had no caller and no test.
• scripts/mutation/floors.toml: a [modules] floor of 91 for ldvViewModel.ts (CI score − 2).

Test Plan

• ldvViewModel.test.ts (new). Every expectation is a literal, never the module's own constant, and test data is built inside each test. It covers:

  • each colour, border and badge rule, including the badge's slot when a NEW badge holds the top;
  • the handle mirror rule both ways (a used mirror does not light its base);
  • the click slop's boundary on both axes and both directions;
  • the label obstacle check's margin on all four sides;
  • label placement: preferred spot, nearest clear sample, the 8 % end skip, the fallback, the exact sample sequence, and the tie going to the earlier sample;
  • every click action, including the fall-through when a mode's handler is missing;
  • paint order with a hovered line, a hovered card and nothing hovered.

• ldvLayoutShared.test.ts:

  • flowMarkers for each direction, bothWays and severed;
  • relationSevered for each change state;
  • routeEdgeData with no centre line, a centre line of 0, and a channel route.

• layeredDependencyLayout.test.ts and ldvAggregate.test.ts: the first tests of each builder's arrowheads. They cover a reverse, an unset-to-retired and a bidirectional line, plus three connector cases: one merging both directions, one carrying a single type's reverse flow, and one whose relations are all severed.

• ldvEdgeRouting.test.ts:

  • the label spread: 0.2 / 0.5 / 0.8 within 80 px of the first label, left to right, with a label exactly 80 px away left out, and unlabelled lines joining no cluster;
  • the nudge off a card (0.2 → 0.32), and its five-step cap;
  • the obstruction clearance as literal numbers (130 when centred, 80 when 50 px off-centre).

• The demo-landscape guard (never lets two connectors leave one node from the same point) now keys on the physical point (ts-N = t-N), so a mirrored handle can no longer hide a shared point. I checked it holds on all six level and centre combinations.

• The suspected spreadOverflow gap (top t and ts bucketed apart) did not reproduce: the aggregate layout puts a box's incoming and outgoing connectors on opposite borders. No fix; the tightened guard above would catch it.

• The view test's React Flow fake is unchanged. Its header leaves edge geometry to the browser suite (jsdom has no getTotalLength), and the edge component's decisions are now unit-tested in the model.

• Locally: tsc -b and eslint on the touched files only. Their one warning (react-refresh on the readableTypeColor re-export) predates this PR. The floors file is validated by backend/tests/core/test_mutation_scripts.py -k floors. As agreed, the tests run in CI.

• Mutation score (CI's Frontend Mutation Tests on 903c02ecc): 88.0 % of 442 changed-line mutants; ldvViewModel.ts 93.8 % (317 of 338), ldvLayoutShared.ts 95.3 % (41 of 43), LayeredDependencyView.tsx 50.8 % (31 of 61, the view's sx and render wiring). Survivors kept in the two modules, none pragma'd: the colour constants (FALLBACK_TYPE_COLOR, the five PHASE_DOT entries, ARROW_COLOR, SEVERED_ARROW_COLOR, the BACKGROUND_CYCLE literals — presentation values the tests read back rather than pin one by one); the pixel arithmetic in DOT_BOX and TYPE_ICON_RIGHT_BESIDE_DOT (layout constants); and three label-search mutants that are equivalent on the sample grid — i < steps → <= adds t = 1, which the end skip drops anyway, t < 0.08 / t > 0.92 → <= / >= on a grid no sample of which lands on either bound, and the preferred-spot early return, since the preferred point is also the first clear sample. The [modules] floor of 91 for ldvViewModel.ts (score − 2) is in floors.toml.

• All CI checks pass (backend lint, backend tests, frontend lint, frontend build, frontend tests, mutation tests)

• Manually tested the affected feature

• Added/updated tests for new or changed behavior

• Mutation gate: my tests kill the mutants on the lines I changed (make mutation-diff). Score: 88.0 % (floor 60 %); survivors listed above, no pragma added.

Checklist

• My changes follow the conventions in CLAUDE.md
• I added permission checks to any new mutating endpoints
• I created an Alembic migration for any schema changes
• I did not introduce hardcoded card types or fields (metamodel is data-driven)
• I used async def for all new route handlers and DB operations
• I did not expose sensitive fields (password hashes, encrypted secrets) in API responses
• I bumped /VERSION and added a CHANGELOG.md entry (if user-facing change)
• I added translations for new UI strings in all 10 locales (if applicable)
• I updated user documentation in docs/ (if UI or feature change)
• Screenshots attached for UI changes (if applicable)

VERSION 2.158.12, one CHANGELOG Changed line. No backend, schema, permission, UI-string or docs change.

🤖 Generated with Claude Code

https://claude.ai/code/session_014eZ98kaESE7YBa84fT1msc

---

Generated by Claude Code

[cloudflare-workers-and-pages]

Deploying turbo-ea-docs with    Cloudflare Pages

Latest commit:	a43ec71
Status:	✅  Deploy successful!
Preview URL:	https://30f62aea.turbo-ea.pages.dev
Branch Preview URL:	https://claude-mutation-lift-ldv.turbo-ea.pages.dev

View logs