@javelin/perf

A runnerless, single-threaded benchmarking library for Node.

Getting Started

npm i @javelin/perf

src/fibonacci.perf.ts

import {fibonacci} from "./fibonacci"
import {perf} from "@javelin/perf"

perf("100 elements", () => fibonacci(100))

> ts-node src/fibonacci.perf.ts

src/fibonacci.perf.ts

 100 elements   106,321 ops/s     =   4.13

Setup

perf("iterator next()", () => {
  let it = fibonacci.iterator({start: 1_000})
  return () => it.next()
})

Options

Run just a single benchmark in a file:

perf.only("name", () => {})

Don't throw an error when a benchmark degrades in performance.

perf("name", () => {}, {
  throwOnFailure: false,
})

Configuration

# Throw an error when performance degrades, halting test execution.
THROW_ON_FAILURE=false
# Write failing test stats to perf results files.
WRITE_FAILURES=false
# Number of times to execute each individual test.
PERF_RUNS=15_000
# Number of samples to ignore from upper and lower extremes when computing perf results.
PERF_SAMPLES_TO_DISCARD_PER_EXTREME=100
# Perf results file extension.
MODULE_RESULTS_EXTENSION=".perf-results"
# Perf file extension.
MODULE_EXTENSION=".perf.ts"