Skip to content

Project resumption: lint/security cleanup, CI gate, per-node coverage visualization (P6-1) - #20

Merged
d3mocide merged 6 commits into
devfrom
claude/project-completion-review-pn2rrx
Jul 25, 2026
Merged

Project resumption: lint/security cleanup, CI gate, per-node coverage visualization (P6-1)#20
d3mocide merged 6 commits into
devfrom
claude/project-completion-review-pn2rrx

Conversation

@d3mocide

@d3mocide d3mocide commented Jul 24, 2026

Copy link
Copy Markdown
Owner

Summary

The project had been dormant since v1.16.1 (Feb 2026). This PR picks it back up on four fronts:

  • Lint hygiene: eslint.config.js had no way to recognize JSX component usage (no eslint-plugin-react), so plain no-unused-vars produced ~100 false-positive "unused import" warnings (132 → 35 after the fix). The remaining real issues found underneath (dead variable, unused caught error, stale eslint-disable directives, two intentional-but-unannotated useEffect deps) are fixed/annotated.
  • Security: resolved all production npm audit findings (critical fast-xml-parser via @loaders.gl/xml, high lodash, moderate protocol-buffers-schema) and the dev-tooling ones (vite, vitest, brace-expansion).
  • CI: added .github/workflows/ci.yml — nothing previously ran lint/tests on pushes or PRs (only Docker publish existed).
  • Docs: fixed a literal copy-paste duplication in ROADMAP.md (P5-6/P5-7 repeated twice), wrote the Documentation/pwa-guide.md that README.md linked to but never existed, and dropped the unused itmlogic dependency from rf-engine/requirements.txt (declared, never imported).
  • Feature — ROADMAP P6-1 (Per-Node Coverage Visualization): Multi-Site Analysis previously rendered every selected node's viewshed into one flat cyan composite, making it impossible to tell which node covers which area. The backend now assigns each node a distinct color and alpha-composites per-node layers (overlapping coverage blends naturally); the map markers, popups, and Sites results table are color-matched to the same per-node colors. Also factored out a pixel-projection helper that was duplicated three times in viewshed_proc.py.

Test plan

  • npm run lint — 0 errors (35 pre-existing warnings remain, out of scope for this pass — see below)
  • npm test (vitest) — 7/7 passing
  • npm run build — succeeds
  • pytest rf-engine/tests/ — 9/9 passing (added 2 new tests covering the per-node color/composite logic, which the prior tests never exercised since they only used all-zero grids)
  • Manually verified the new composite-rendering path end-to-end with a two-node scenario (distinct colors assigned, overlap region blends, PNG decodes with multiple visible colors)

Not in scope for this PR

The remaining 35 lint warnings (dead vars in LinkLayer.jsx/MapContainer.jsx/OptimizationLayer.jsx, and the "fast refresh" warnings on context files that mix hook/provider exports) are real but unrelated to what was asked here — flagging as a natural follow-up now that lint output is trustworthy again.

Generated by Claude Code

claude added 6 commits July 24, 2026 17:24
@claude
chore(lint): fix eslint JSX detection, resolve npm audit vulnerabilities
ed7c121 
eslint.config.js had no plugin providing JSX-usage detection, so plain
no-unused-vars couldn't see that a capitalized JSX tag references its
import -- this produced ~100 false-positive warnings across the codebase
that were masking real ones. Add eslint-plugin-react's jsx-uses-vars rule
(only that rule, not its full stylistic/prop-types set).

With the false positives gone, fix what was left: a dead `elevation` var
in CoverageLayerManager, an unused caught SSE-parse error in
useSimulationStore, and annotate two intentional signal-triggered
useEffects (CoverageLayerManager, RFContext) that read the latest render's
closure rather than tracking every dependency.

Also run npm audit fix to close all production vulnerabilities (critical
fast-xml-parser via @loaders.gl/xml, high lodash, moderate
protocol-buffers-schema) plus the dev-tooling ones (vite, vitest,
brace-expansion).

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DwJgxrJr8scmomw9B8Lk9W
@claude
docs: fix roadmap duplication, add pwa guide, drop dead itmlogic dep
ed7be15 
ROADMAP.md had the P5-6/P5-7 sections duplicated verbatim from a
copy-paste error. README.md links to Documentation/pwa-guide.md, which
never existed -- write it from the actual vite-plugin-pwa config.

Drop itmlogic from rf-engine/requirements.txt: it was declared but never
imported (ROADMAP P4-1 is still open, re-add when that's implemented).

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DwJgxrJr8scmomw9B8Lk9W
@claude
ci: add lint/test workflow for frontend and rf-engine
13831c7 
.github/workflows only had docker-publish and cleanup-packages -- nothing
ran npm run lint, npm test, or the rf-engine pytest suite on pushes/PRs.
Add a ci.yml workflow that runs both on dev/main pushes and PRs targeting
them, so regressions don't silently accumulate.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DwJgxrJr8scmomw9B8Lk9W
@claude
feat(site-analysis): per-node coverage visualization (P6-1)
b7d5494 
Multi-Site Analysis previously rendered every selected node's viewshed
into one flat cyan composite, making it impossible to tell which node
covers which area or spot redundant placements visually (the numeric
"unique coverage %" was the only signal).

Backend (rf-engine/core/viewshed_proc.py): assign each selected node an
evenly-spaced distinct hue and render coverage as alpha-composited color
layers instead of a single flat mask -- overlapping nodes blend naturally
where their coverage intersects. Each result now carries a `color` field.
Also factor the pixel-projection math (mapping a node's local viewshed
grid into the shared master-grid coordinates) into one `_project_to_master`
helper, replacing three near-identical inline copies used by the greedy
optimizer, marginal-coverage pass, and composite renderer.

Frontend: color-match the simulation node markers and popups
(OptimizationLayerManager.jsx) and the Sites results table (SitesTab.jsx)
to the same per-node color, so a node's marker, popup, and coverage patch
on the map are visually tied together. Also fixes a pre-existing dead
`res.coverage_points` reference in the marker popup (the backend never
produced that field) by showing unique coverage % instead.

Adds test coverage for the new color assignment and composite rendering,
which the prior tests never exercised (they only used all-zero grids).

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DwJgxrJr8scmomw9B8Lk9W
@claude
docs: add Unreleased changelog entry
b8519fb 
Summarizes the lint/security cleanup, new CI workflow, and P6-1 per-node
coverage visualization from the preceding commits.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DwJgxrJr8scmomw9B8Lk9W
@claude
chore(lint): fix remaining 35 warnings
d3edec2 
Now that JSX usage is correctly detected, clean up what was left:

- Removed genuinely dead code: unused txIcon/rxIcon constants and an
  abandoned fresnelRef in LinkLayer.jsx, an unused syncState helper and
  its backing ref in OptimizationLayer.jsx, and stale proxy-value
  destructuring left over from the P2-4 per-node-config refactor.
- Removed the entire `coverageOverlay` state chain (useLinkTool.js ->
  MapContainer.jsx -> LinkLayerManager.jsx -> LinkLayer.jsx): it was only
  ever set to null, never to a real value, so the ImageOverlay it fed
  never rendered anything.
- Fixed a real gap in HeatmapOverlay.jsx: a per-point opacity was computed
  from score but never applied, the fillOpacity was hardcoded to 0.3
  instead.
- Fixed a dead `res.coverage_points` style leftover (unused args/index
  params) across OptimizationResultsPanel, BatchNodesList,
  ModelComparisonTable, SiteAnalysisResultsPanel.
- Annotated intentional signal-triggered useEffects (Sidebar mobile
  auto-close, DeckGLOverlay mount lifecycle, LinkLayer/OptimizationLayer
  recalc-on-signal) that eslint's exhaustive-deps flagged -- adding the
  suggested deps would have introduced real bugs (e.g. Sidebar's toggle
  button fighting its own auto-close effect).
- Disabled react-refresh/only-export-components at the file level for the
  5 context files (EnvironmentContext, HardwareContext, RFContext,
  RadioContext, UIContext). Splitting each Provider from its hook/constants
  is the "proper" fix but would touch ~25 import sites across the app for
  a dev-only Fast Refresh granularity concern with no effect on
  correctness or production builds -- not proportionate here.

Verified: npm run lint (0 warnings), npm test (7/7), npm run build,
pytest rf-engine/tests (9/9).

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DwJgxrJr8scmomw9B8Lk9W
@d3mocide
d3mocide marked this pull request as ready for review July 25, 2026 04:08
@d3mocide
d3mocide changed the base branch from main to dev July 25, 2026 04:08
@d3mocide
d3mocide merged commit 121987a into dev Jul 25, 2026
6 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

2 participants

@d3mocide @claude