• Simplicity: Use the mt test runner with the test function, nothing more.
• Performance: By doing and including less we can run more quick.
• Minimal: Bring your own assertions, snapshots, mt will always be dependency free.

Table of Contents

• Quickstart
• Usage
  • Finding tests
  • Writing tests
  • Running tests
  • Filtering
  • Failing fast
  • Quiet
  • CLI
• Extras
  • TypeScript
  • Coverage
  • JSX, TSX, browsers
• Rationale
• License

Quickstart

1. Install via your tool of choice: pnpm add --dev @sondr3/minitest

2. Add a test script to package.json:

     "scripts": {
       "test": "mt <dir>",
     },

3. Write your tests:

   import { strict as assert } from "assert";
   import { test } from "@sondr3/minitest";
   
   test("it works!", () => {
     assert(true === true, "Phew");
   });

4. Run your tests: pnpm test

   running 1 test
   running 1 test in dir/index.test.js
   test  it works! ... ok
   
   test result: ok. 1 passed; 0 failed; 0 ignored; 0 filtered out; finished in 32ms

Usage

Finding tests

mt will recursively look for files in your current directory:

• named test.{js,mjs},
• or ending in .test.{js,mjs}
• or ending in _test.{js,mjs}

Writing tests

To write tests you need to import the test function from @sondr3/minitest. There are a couple of different styles that can be used to write tests, but we'll get into these later.

import { test } from "@sondr3/minitest";
import { strict as assert } from "node:assert";

test("hello world #1", () => {
  const x = 1 + 2;
  assert.equal(x, 3);
});

test({ name: "hello world #2" }, () => {
  const x = 1 + 2;
  assert.equal(x, 3);
});

You can use any assertion library you want, like Chai for a TDD/BDD like assertions and Sinon.JS to spy and mock functionality. The only requirement is that the assertions throw when they fail.

If you come from a frameworks like Jest you may be surprised to learn that there is no built-in support for nesting tests, or anything like beforeAll or beforeEach in minitest. I can highly recommend this article from Kent C. Dodds about nesting and test hooks. After having used Deno and Rust which do not have such functionality, I don't miss it. I recommend you give it a try!

Async functions

You can also test asynchronous code by turning the test function into a function that returns a promise. Simply add async in front of the function:

import { test } from "@sondr3/minitest";
import { strict as assert } from "node:assert";

test("async hello world", async () => {
  const x = 1 + 2;
  const wait = async () => new Promise((res) => setTimeout(res, 1000));
  await wait();
  assert.equal(x, 3);
});

Running tests

Simply run mt to test all test files matching the pattern in finding tests, or point it to the directory or file you want to run:

# Run all tests in current directory
mt

# Run all tests in the fs directory
mt fs

# Run a single test file
mt fs.test.js

Filtering

There are tree different ways of filtering tests, either by filtering against test names, skipping tests or running a select subset of tests.

Filtering via the CLI

Tests can be filtered out by using the --filter option. This option accepts either a case insensitive string or a Regex pattern as its value.

If you have the following tests:

test("a-test", () => {});
test("test-1", () => {});
test("test-2", () => {});

You can then filter them by filtering all tests that contain the word test:

mt --filter "test"

Or by matching against a Regex pattern:

mt --filter "/test-*\d/"

Ignoring (skipping)

If you want to ignore/skip tests based on some boolean condition or because they are currently incorrect you can use the ignore() method or the ignore field in the options:

test("ignored", () => {}).ignore(); // or .ignore(condition)
test({ name: "other ignored", ignore: process.arch === "x64" }, () => {});
test("final ignored", () => {}, { ignore: true });

Running specific tests

Sometimes you may want to only run a subset of your tests because you are working on a problem where you want to ignore tests not relevant to it. To do this you can mark tests using the only() method or the only field in the options, and only these tests are run. While these tests will run normally and report success and failure normally, the whole test suite will still fail as using only should only be a temporary measure as it disables nearly the whole test suite.

test("other ignored", () => {}).only();
test({ name: "ignored", only: true }, () => {});
test("final ignored", () => {}, { only: true });

Failing fast

If you have a test suite that takes a long time to complete or you simply want to exit after the first error, one can use the --fail-fast option:

# fail immediately
mt --fail-fast

# fail after three failures
mt --fail-fast 3

Quiet

The test output for mt can quickly become verbose once the test suite grows, to help with this one can use the --quiet flag to make it far less verbose:

mt --quiet

CLI

For a complete overview over the available options, use the --help flag:

$ mt --help
minitest v0.1.0
A low-feature and performant test runner inspired by Rust and Deno

USAGE:
        mt <dir> [flags]

OPTIONS:
        -q, --quiet              Quiet output
        -f, --filter=<filter>    Filter tests by name, accepts regex
        -F, --fail-fast=<N>      Fail after N test failures [default: 0]
        -v, --version            Print version
        -h, --help               Print help

Extras

TypeScript

There is no built-in support for natively running TypeScript files, they need to be compiled to JavaScript first. In other words, your build step needs to happen before you run your tests.

Coverage

To generate coverate reporting for your tests, I highly recommend using c8. It generates coverage based on the built-in coverage functionality of NodeJS and will just magically work with your tests. It is also very fast compared to similar functionality in other testing libraries. To generate coverage by default, change your test script to the following after you've added c8 as a dependency:

  "scripts": {
    "test": "c8 mt <dir>",
  },

JSX, TSX, browsers

Like with TypeScript, there is no built-in support for usage with JSX and/or TSX, and no specific functionality for working against browsers. You may be able to make it work, but it is purpose built around testing with Node, so YMMV.

Rationale

Why would you use this over the myriad of other test frameworks, runners and libraries that exist? You probably shouldn't, Jest, Ava, tape, uvu and a whole lot of others are more mature, stable, feature rich and has a wider userbase, but if you want a simple Rust/Deno like test framework, this might just scratch your itch.

I also created this as an exercise in learning how testing works under the hood, and had a lot of fun building it out into a actual useful package. For my needs and packages minitest is just what I want from a test framework: minimal fuzzing, fast and easy and straight forward.

License

MIT.