Remove micro-detours from simplified routes, add debug script and regression test

[0hmX]

Motivation

• Eliminate tiny triangular detours and near-duplicate points in simplified PCB traces produced by convertHdRouteToSimplifiedRoute to reduce spurious kinks in output.
• Provide a repro/debug helper to inspect where such micro-detours originate in the pipeline for dataset01 circuit011.
• Add a regression test to prevent reintroduction of the micro-triangle issue.

Description

• Implemented sanitizeLayerPoints and small geometry helpers (distance) along with constants to detect and remove near-duplicate points and micro-detours based on chord length, edge sizes, and perpendicular distance.
• Integrated sanitizeLayerPoints into convertHdRouteToSimplifiedRoute for both layer transitions and the final layer emission so wires are emitted from cleaned point lists.
• Added scripts/debug-dataset01-circuit011-solver3.ts to reproduce and inspect candidate tiny-triangle detours across solver stages.
• Added regression test tests/bugs/dataset01-circuit011-solver3-no-micro-triangle.test.ts and a convenience script entry debug:dataset01:circuit011:solver3 in package.json.

Testing

• Ran the new regression test tests/bugs/dataset01-circuit011-solver3-no-micro-triangle.test.ts via bun test and it passed.
• Executed the project's test suite via bun test to ensure no regressions and the suite passed.

---

Codex Task

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
capacity-node-autorouter	Ready	Preview, Comment	Mar 12, 2026 1:14pm

[tscircuitbot]

🏃 Benchmark This PR

Run benchmarks by commenting on this PR:

/benchmark [solver-name|all] [scenario-limit] --concurrency <n> --effort <n>

Examples:

• /benchmark -> AutoroutingPipelineSolver, all scenarios (default concurrency uses the benchmark runner CPU count)
• /benchmark AutoroutingPipelineSolver -> one solver, all scenarios
• /benchmark all 20 -> all solvers, first 20 scenarios
• /benchmark AutoroutingPipelineSolver 20 --concurrency 8 -> one solver, 20 scenarios, 8 workers
• /benchmark AutoroutingPipelineSolver 20 --effort 2 -> one solver, 20 scenarios, 2x effort

Any PR whose title contains [BENCHMARK TEST] will automatically run the benchmark workflow on PR updates.

[graphite-app]

Logic gap in micro-detour detection. When distAB = 0.005 (tiny) and distBC = 0.025 (medium), the conditions:

• Line 43: Math.min(0.005, 0.025) = 0.005 < 0.01 ✓ (passes)
• Line 44: Math.max(0.005, 0.025) = 0.025 <= 0.03 ✓ (skips removal)

This creates a gap where triangles with one very small edge (< 0.01) and one medium edge (between 0.01 and 0.03) are NOT removed, potentially leaving micro-detours in the output.

// Consider adjusting line 44 to:
if (Math.max(distAB, distBC) < MICRO_DETOUR_MIN_LARGE_EDGE) continue
// Or removing this check entirely if all micro-detours should be caught

This may cause the regression test to pass while still allowing some micro-triangular detours through, as the test only checks chord < 0.08 without the edge size constraints.

Spotted by Graphite



Is this helpful? React 👍 or 👎 to let us know.