/kelonio

Performance testing library for Node.js, written in TypeScript

Primary LanguageTypeScriptMIT LicenseMIT

Kelonio

Kelonio is a performance testing library for Node.js, written in TypeScript. Whereas many similar projects are test frameworks in and of themselves, Kelonio is fundamentally a library and therefore aims to integrate with existing test frameworks seamlessly instead of reinventing the wheel. You can use it inside of your existing tests from frameworks such as Jest and Mocha (along with any loaders like ts-jest), and you can use it in the console and scripts as well.

Kelonio also works in the browser (as long as you use a tool like Webpack or Browserify), and it comes with built-in reporters for the following test frameworks without any direct dependency on them:

Usage

Full API documentation: https://mtkennerly.github.io/kelonio/modules

For simple, one-off checks, like in the console or a script, use the measure function:

import { measure } from "kelonio";
import axios from "axios";

measure(() => axios.get("http://www.httpbin.org/get"))
    .then(measurement => console.log(`Mean: ${measurement.mean} ms`));

By default, the check is repeated 100 times, but you can customize this. If you measure a function that returns a promise, Kelonio will automatically measure the time until it's resolved as well. The resulting measurement exposes various stats, like mean time, maximum time, and standard deviation.

For aggregating results from multiple measurements, create a Benchmark and use its record method to store the state:

import { Benchmark, Criteria } from "kelonio";

const benchmark = new Benchmark();
await benchmark.record("RegExp#test", () => /o/.test("Hello World"));
await benchmark.record("String#indexOf", () => "Hello World!".indexOf("o") > -1);

const fastest = benchmark.find(Criteria.Fastest);
console.log(`Fastest: ${fastest?.description} with mean ${fastest?.mean} ms`);
// Fastest: String#indexOf with mean 0.004199049999999999 ms

For aggregating results inside of a test framework, use the default benchmark instance and its record method. Click to expand an example:

Example: Jest

Jest doesn't currently expose a way to get each individual test's name while running, so you have to provide a description to record().

Tests:

import { benchmark } from "kelonio";
import axios from "axios";

describe("An HTTP client", () => {
    it("can send GET requests", async () => {
        await benchmark.record(
            ["HTTP client", "GET"],
            () => axios.get("http://www.httpbin.org/get")
        );
    }, 30_000);

    it("can send POST requests", async () => {
        await benchmark.record(
            ["HTTP client", "POST"],
            () => axios.post("http://www.httpbin.org/post"),
            { iterations: 10, meanUnder: 10 },
        );
    }, 30_000);
});

Output:

FAIL ./index.test.ts (16.576s)
  An HTTP client
    √ can send GET requests (8332ms)
    × can send POST requests (508ms)

  ● An HTTP client › can send POST requests

    Mean time of 49.43073600000001 ms exceeded threshold of 10 ms

Test Suites: 1 failed, 1 total
Tests:       1 failed, 1 passed, 2 total
Snapshots:   0 total
Time:        18.296s

- - - - - - - - - - - - - - - - - Performance - - - - - - - - - - - - - - - - -
HTTP client:
  GET:
    83.25152 ms (+/- 58.77542 ms) from 100 iterations
  POST:
    49.43074 ms (+/- 2.39217 ms) from 10 iterations
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

The first time on each line is the mean duration, and the +/- time is the margin of error at a 95% confidence level.

Example: Mocha

The Mocha reporter can automatically infer the descriptions from the test names, but you're still free to pass additional descriptions to record(), such as if one test performs several different measurements.

Tests:

import { benchmark } from "kelonio";
import axios from "axios";

describe("An HTTP client", () => {
    it("can send GET requests", async function (this: Mocha.Test) {
        this.timeout(30_000);
        await benchmark.record(() => axios.get("http://www.httpbin.org/get"));
    });

    it("can send POST requests", async function (this: Mocha.Test) {
        this.timeout(30_000);
        await benchmark.record(
            () => axios.post("http://www.httpbin.org/post"),
            { iterations: 10, meanUnder: 10 },
        );
    });
});

Output:

  An HTTP client
    √ can send GET requests
    1) can send POST requests


  1 passing (8332ms)
  1 failing

  1) An HTTP client
      can send POST requests:
    Error: Mean time of 49.43073600000001 ms exceeded threshold of 10 ms


- - - - - - - - - - - - - - - - - Performance - - - - - - - - - - - - - - - - -
An HTTP client:
  can send GET requests:
    83.25152 ms (+/- 58.77542 ms) from 100 iterations
  can send POST requests:
    49.43074 ms (+/- 2.39217 ms) from 10 iterations
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

The first time on each line is the mean duration, and the +/- time is the margin of error at a 95% confidence level.

Refer to the examples folder for sample projects that integrate Kelonio with different test frameworks.

Versioning

This project uses Semantic Versioning. Public API:

  • All items that can be imported from "kelonio" and their public attributes.
  • The location of reporter modules:
    • node_modules/kelonio/out/plugin/jestReporter.js.
      • node_modules/kelonio/out/plugin/jestReporterSetup.js.
    • node_modules/kelonio/out/plugin/karmaReporter.js.
      • node_modules/kelonio/out/plugin/karmaReporterSetup.js.
    • node_modules/kelonio/out/plugin/mochaReporter.js.

Comparison with other tools

  • Benchmark:
    • Requires defining tests in its own framework.
    • Doesn't provide a default report format, so you have to write your own reporting in callbacks.
    • Callbacks must be classic function () {} style because they need access to this, which is not accounted for by @types/benchmark.
  • Nanobench:
    • Requires defining tests in its own framework.
    • The CLI can only handle JavaScript code, so in a TypeScript project, you either have to compile the tests in addition to the main source or you have to use ts-node (which appears to degrade the performance results).
    • No typings available for TypeScript.
  • Matcha:
    • Requires defining tests in its own framework.
    • The CLI can only handle JavaScript code, so in a TypeScript project, you either have to compile the tests instead of just the main source or you have to use ts-node (which appears to degrade the performance results).
    • No typings available for TypeScript.
    • Depends on Electron.

Development

Please refer to CONTRIBUTING.md.