Measured Node.js 22.12.0
Benchmarks
node-cmd delegates to Node’s own child-process primitives. These measurements isolate the JavaScript wrapper and then time real child completion, so convenience never hides its cost.
01 / Result
The wrapper stays beneath the work it controls
Common callback, Promise, direct-file, and streaming paths added medians of 0.21–0.22 ns over the matching node:child_process call. The two synchronous result envelopes added 7.56 ns and 10.40 ns. Even the larger value is about 3.8 million times smaller than its matching empty-process completion time.
What this proves. node-cmd’s measured JavaScript dispatch cost is negligible beside process creation on this reference system. It does not claim to make Node or the operating system launch processes faster.
02 / JavaScript dispatch
Nanoseconds before a process exists
The child-process functions were replaced temporarily with counter stubs in an isolated measurement setup. This removes operating-system launch noise and exposes only argument normalization, result-envelope construction, and function dispatch. Bars show median nanoseconds per call; lower is better.
node:child_processnode-cmd
runrunPromiserunSyncrunFilerunFilePromiserunFileSyncrunStream| node-cmd API | Paired median delta | 95% bootstrap CI | Interpretation |
|---|---|---|---|
run | +0.211 ns | +0.203 to +0.220 | Dispatch measured |
runPromise | +0.217 ns | +0.209 to +0.229 | Dispatch measured |
runSync | +7.562 ns | +7.486 to +7.614 | Result normalization measured |
runFile | +0.211 ns | +0.204 to +0.219 | Overload dispatch measured |
runFilePromise | +0.213 ns | +0.204 to +0.224 | Overload dispatch measured |
runFileSync | +10.402 ns | +10.241 to +10.484 | Options + result normalization measured |
runStream | +0.218 ns | +0.204 to +0.227 | Overload dispatch measured |
03 / Real process completion
The wrapper tracks Node through the operating system
Each implementation started the same empty Node.js executable with identical options. The run* scenarios include the platform shell plus that child; runFile* and runStream() start the child directly. The direct and node-cmd calls ran sequentially, with first position alternating to reduce order bias. Bars show median completion time; lower is better.
node:child_processnode-cmd
run · callbackrunPromiserunSyncrunFile · callbackrunFilePromiserunFileSyncrunStream| node-cmd API | Node direct p50 | node-cmd p50 | Paired-delta 95% CI |
|---|---|---|---|
run | 53.040 ms | 52.859 ms | −0.716 to +0.235 ms |
runPromise | 53.712 ms | 53.787 ms | −0.985 to +0.600 ms |
runSync | 54.495 ms | 54.617 ms | −0.853 to +1.536 ms |
runFile | 42.452 ms | 41.932 ms | −1.145 to +0.146 ms |
runFilePromise | 40.393 ms | 40.140 ms | −1.135 to +0.460 ms |
runFileSync | 39.508 ms | 39.485 ms | −0.793 to +0.189 ms |
runStream | 39.121 ms | 38.886 ms | −0.874 to +0.218 ms |
All seven paired-delta confidence intervals include zero, so this run did not resolve a completion-time difference. That is a statement about detection at this sample size—not proof that the implementations are mathematically equivalent.
04 / Method
Reference environment and controls
| Input | Reference value | Why it matters |
|---|---|---|
| Runtime | Node.js 22.12.0 · V8 12.4 | The exact supported runtime floor |
| Source identity | SHA-256 for runtime and both benchmark scripts | Full hashes are included in the raw report |
| Host | Windows x64 · Intel Core Ultra 9 275HX · 24 logical CPUs | Absolute launch time is machine-specific |
| Process suite | 20 warm-ups + 100 balanced pairs per API | Sequential calls and alternating order reduce interference |
| Dispatch suite | 100 calibrated batches per API | At least 100,000 calls per batch; raw batches retained |
| Intervals | 2,000 deterministic bootstrap resamples | Reports uncertainty around the paired median |
| Source parity | CommonJS and ESM identity asserted before timing | Both module systems use the same implementation |
Read the machine, not a universal constant. Absolute process-launch time varies with hardware, operating system, load, security software, and the executable. Compare trends and wrapper scale; reproduce on the environment that matters to you.
05 / Reproduce
Run the dependency-free harness yourself
node ./benchmark/run.js --samples 100 --warmups 20 --output ./benchmark/reference.json
npm run benchmark:chartThe harness uses Node built-ins only. It preserves every process pair and dispatch batch in benchmark-results.json; the README charts are regenerated directly from that file.