Skip to content

PERFORMANCE.md

Latest commit

 

History

History
157 lines (102 loc) · 4.12 KB

PERFORMANCE.md

File metadata and controls

157 lines (102 loc) · 4.12 KB
Raw

Performance Optimization

Performance baseline and optimization tracking for the bfra.me/works monorepo.

Targets

Metric Target Status
pnpm validate (local) < 2 minutes 🔄 Measuring
pnpm build (full) < 45 seconds 🔄 Measuring
pnpm build (incremental) < 5 seconds 🔄 Measuring
pnpm test < 30 seconds 🔄 Measuring
pnpm lint < 20 seconds 🔄 Measuring
pnpm analyze < 10 seconds 🔄 Measuring
CI pipeline (total) < 5 minutes 🔄 Measuring

Optimizations Applied

Cache Invalidation Strategy (December 2025)

Problem: Development cache key included **/*.ts which invalidated all caches on any TypeScript file change (tests, docs, fixtures).

Solution: Narrowed cache pattern to packages/*/src/**/*.ts and added dedicated ESLint cache with its own key pattern.

Files Changed:

  • .github/actions/pnpm-install/action.yaml

Expected Impact: 50-70% improvement in cache hit rate

Parallelized Validation Pipeline (December 2025)

Problem: pnpm validate ran all steps sequentially: type-check → build → lint → test → type-coverage

Solution: Run type-check, lint, and test concurrently (they have no dependencies due to vitest's conditions: ['source'] resolver), then build and type-coverage sequentially.

Before:

type-check → build → lint → test → type-coverage
[sequential, ~2-3 minutes]

After:

┌─────────────┬─────────┬──────┐
│ type-check  │  lint   │ test │  (parallel)
└─────────────┴─────────┴──────┘
              ↓
           build
              ↓
        type-coverage

Files Changed:

  • package.json (validate script)

Expected Impact: 30-40% reduction in validate time

CI Job Reordering (December 2025)

Problem: analyze job waited for build to complete unnecessarily.

Solution: Changed analyze to depend on prepare instead of build. Workspace-analyzer scans source files, not build artifacts.

Files Changed:

  • .github/workflows/main.yaml

Expected Impact: 10-15% reduction in CI wall-clock time

Developer Profiling Guide

Build Performance

TIMING=1 pnpm build 2>&1 | grep -E 'packages|total|✓'

Test Performance

pnpm test -- --reporter=verbose

Lint Performance

time pnpm lint

Analyze Performance

time pnpm analyze

Full Validation Timing

time pnpm validate

Type-Check with Diagnostics

pnpm tsc --noEmit --diagnostics

Cache Behavior

pnpm Store Cache

  • Location: pnpm store path
  • Rotation: Monthly (year-month based key)
  • Invalidates on: pnpm-lock.yaml changes

Development Cache

  • Location: .tsup/, tsconfig.tsbuildinfo, .type-coverage/, docs/.astro/
  • Invalidates on: Source code changes (packages/*/src/**/*.ts)
  • Safe to delete: Yes, triggers full rebuild

ESLint Cache

  • Location: .eslintcache
  • Invalidates on: ESLint config or source changes
  • Safe to delete: Yes, next lint run recreates it

Troubleshooting

Slow Builds

  1. Check if ESLint cache exists: ls -la .eslintcache
  2. Check tsup cache: ls -la packages/*/.tsup
  3. Clean and rebuild: pnpm clean && pnpm build

Cache Issues in CI

  1. Check cache hit/miss in GitHub Actions logs
  2. Verify cache key patterns match expected files
  3. Force cache refresh by incrementing version in cache key

Test Performance Degradation

  1. Profile with: pnpm test -- --reporter=verbose
  2. Check for large fixture files or slow setup
  3. Ensure it.concurrent() used for independent tests

Future Improvements

  • Implement task-level build caching (Turbo or manual)
  • Add test sharding across multiple CI runners
  • Incremental workspace-analyzer with change detection
  • Optimize eslint-config type generation caching