Jason Jean
7439d58b22
## Current Behavior No standardized way to measure Nx CLI performance or detect regressions. ## Expected Behavior A `benchmarks/` workspace provides reproducible CLI performance benchmarks using [hyperfine](https://github.com/sharkdp/hyperfine). ### Project Structure 1110 dummy projects in a 3-level fan-out (10 groups → 10 subs → 10 leaves) with implicit dependencies. ### Benchmarks | Benchmark | Command | Outputs | Dependencies | What it measures | | --------------- | --------------------- | ------- | ------------ | ----------------------------------------- | | `version` | `nx --version` | — | — | CLI startup / module loading | | `show-projects` | `nx show projects` | — | — | Project graph query via daemon | | `cat-warm` | `cat lorem.md` × 1110 | No | Flat | Task scheduling + hashing (no output I/O) | | `lint-warm` | `cp lorem.md` × 1110 | Yes | Flat | Cached tasks with output tracking | | `build-warm` | `cp lorem.md` × 1110 | Yes | Topological | Cached tasks with deps (disabled) | ### Usage ```bash # Run all benchmarks pnpm bench # Run a single benchmark pnpm nx bench:cat-warm benchmarks # Set local baseline for comparison pnpm bench -- --set-baseline ``` ### How it works - Each `bench:*` target runs hyperfine with `--setup 'nx reset'` (clean daemon) and `--warmup 1` (warm daemon for measurements) - `goals.json` (committed) defines target times the team agrees on - `baseline.json` (gitignored) captures local per-machine numbers for personal comparison - First run auto-saves a baseline; subsequent runs compare against it - The report shows colored deltas against both goals and baseline - Benchmarks run in CI via `nx affected --targets=bench` on `linux-large` agents --------- Co-authored-by: nx-cloud[bot] <71083854+nx-cloud[bot]@users.noreply.github.com> Co-authored-by: FrozenPandaz <FrozenPandaz@users.noreply.github.com>
7 lines
97 B
Plaintext
7 lines
97 B
Plaintext
benchmarks/packages
|
|
nx-dev/**/jest.config.js
|
|
.next
|
|
_files
|
|
_solution
|
|
nx-dev/tutorial/**/templates
|