Run benchmark with Jest
npm install jest-benchRun benchmark with Jest. Write benchmark files along with your test files, and benchmark using any existing test environment you're using. This package builds on top of the excellent benchmark package.
Some environments such as jest-electron are only useable with Jest version less than 27, therefore this package takes its version after Jest version for easy installation:
| Jest version | Jest-Bench version |
| ------------ | ------------------ |
| 29.x.x | 29.x.x |
| 28.x.x | 28.x.x |
| 27.x.x | 27.x.x |
| 26.x.x | 26.x.x |
For contributors, branch main works with Jest version 29.
Install
``bash`
npm i -D jest-bench
Create a jest config file just for running benchmarks. You can use names such as jest-bench.config.json.
`javascript`
{
// Jest-bench need its own test environment to function
"testEnvironment": "jest-bench/environment",
"testEnvironmentOptions": {
// still Jest-bench environment will run your environment if you specify it here
"testEnvironment": "jest-environment-node",
"testEnvironmentOptions": {
// specify any option for your environment
}
},
// always include "default" reporter along with Jest-bench reporter
// for error reporting
"reporters": ["default", "jest-bench/reporter"],
// will pick up "*.bench.js" files or files in "__benchmarks__" folder.
"testRegex": "(/__benchmarks__/.*|\\.bench)\\.(ts|tsx|js)$",
...
}
You can optionally define some reporter options for extra control
`javascript`
{
...,
"reporters": ["default", ["jest-bench/reporter", {withOpsPerSecond: true}]],
...
}
Now any files with names that match *.bench.js, or are inside __benchmarks__ folder will be considered benchmark files. More examples:
- jest-jsdom.config.ts
- jest-node.config.ts
`javascript
import { benchmarkSuite } from "jest-bench";
let a;
benchmarkSuite("sample", {
// setup will not run just once, it will run for each loop
setup() {
a = [...Array(10e6).keys()];
},
// same thing with teardown
teardown() {
if (a.length < 10e6) a.unshift(0);
},
["Array.indexOf"]: () => {
a.indexOf(555599);
},
["delete Array[i]"]: () => {
expect(a.length).toEqual(10e6);
delete a[0];
},
["Array.unshift"]: () => {
a.unshift(-1);
},
["Array.push"]: () => {
a.push(1000000);
},
["Async test"]: (deferred) => {
// Call deferred.resolve() at the end of the test.
new Promise((resolve) => setTimeout(resolve, 10)).then(() => deferred.resolve());
},
});
`
To see more examples, check out the test folder. You can now run benchmarks like this:
`bash`
npx jest --projects jest-bench.config.json
Jest-bench saves benchmark results to benchmarks/result.txt in addition to being printed, so you might want to add this folder to .gitignore.
`bash`.gitignore
benchmarks/result.txt
Create and run a new suite. Each suite creates and is associated with a describe block underneath.
- name: string, name of suite.
- description: object, an object with each key represents a single benchmark. Behind the scene, each benchmark runs in a test block. You can also write jest assertions, even though doing so makes little sense as it will affect benchmark results. Special keys include:teardown
- setup: run before each loop of benchmark. Note that this and are evaled together with the benchmark. So once you declare this, any variable defined outside of setup and teardown becomes invisible to the benchmark. If this and teardown are not defined then benchmarks will still be able to see variables in outer scopes.beforeAll
- teardown: run after each loop of benchmark. Note the caveat above.
- setupSuite: run once before all benchmarks. This block, in effect, is the same as a block (and it does call beforeAll underneath). Again you probably don't want to define or initialize variables here if you also include setup or teardown.SuiteOptions
- teardownSuite: run once after all benchmarks have concluded.
- timeoutMsOrOptions:
- number of milliseconds before a benchmark timeouts. Default to 60000,
- or a :delay
- , initCount, maxTime, minSamples, minTime are passed to Benchmark. See the documentation for more info.timeoutSeconds` is the number of seconds before a benchmark timeouts. Default to 60.
-