webapp-benchmark

name: webapp-benchmark description: > Run WebApp performance benchmarks and record results. Use when the user asks to "benchmark", "measure performance", "run benchmarks", or needs to verify performance impact of changes.

WebApp Benchmark

Measure repository initialization and API performance using the built-in diagnostic tool.

Benchmark Tool

Add ?diag=repo-bench to the URL to run the benchmark automatically.

http://localhost:5173/?diag=repo-bench

The tool runs 12 fixed locations (BENCH_LOCATIONS in src/diagnostics/repo-benchmark.ts) and outputs results to the browser console.

What It Measures

Initialization (from console INFO logs)

API (from [diag:repo-bench] logs)

• getAllStops
• getRouteShapes
• getAllSourceMeta
• getStopsInBounds (12 locations)
• getStopsNearby (12 locations)
• getUpcomingTimetableEntries limit=3 and no-limit
• getRouteTypesForStop
• getFullDayTimetableEntries
• getStopsForRoutes (12 calls)
• Benchmark total

Measurement Protocol

1. Open the page with ?diag=repo-bench.
2. Wait for results to stabilize (discard cold-start runs if unstable).
3. Reload the page and collect results. Repeat at least 5 times.
4. Run 1 may be a JIT warmup outlier. Mark with * but do NOT exclude without increasing sample size.
5. Use median as the representative value.
6. Network conditions (cache, service worker) affect results significantly. Keep conditions consistent across runs.

Report Format

Record results in src/docs/WEBAPP-BENCHMARKS.md. Each report must include:

1. Change description: What was modified and why.
2. Raw data tables: Initialization and API results for all runs.
3. Comparison with previous: Median diff table against the last recorded benchmark.
4. Observations: Root cause analysis of any changes, trade-offs.

Important

• Always use ?diag=repo-bench. Do NOT manually read individual console log lines.
• Always measure both Initialization AND API. One without the other is incomplete.
• Read src/docs/WEBAPP-BENCHMARKS.md before starting. Know the previous baseline.
• Do not discard outliers from a small sample. Increase the sample size instead.
• Note the environment (machine, browser, data size) if different from the recorded baseline.