Split Tests in Jest

Splits test files in Jest and ensures all parallel jobs finish work at a similar time.

Before

| | Job | Time | | --------------------------- | :-: | :--------: | | =========================== | 1 | 46 minutes | | ======== | 2 | 13 minutes | | ============ | 3 | 19 minutes |

After

| | Job | Time | | ------------- | :-: | :--------: | | ============= | 1 | 26 minutes | | ============= | 2 | 26 minutes | | ============= | 3 | 26 minutes |

Setup

Install @split-tests/jest package:

npm install --save-dev @split-tests/jest
yarn add -D @split-tests/jest

In jest.config.js:

module.exports = {
  /* ... */

  // IMPORTANT: set the test sequencer
  testSequencer: require.resolve("@split-tests/jest"),

  // report the run times of the tests, use `jest-unit` reporter
  reporters: [
    "default",
    [
      "jest-junit",
      {
        outputDirectory: "unit",
        addFileAttribute: true,
      },
    ],
  ],

  globals: {
    "split-tests": {
      // collect the times
      junit: "<rootDir>/unit/junit.xml",
    },
  },
};

Usage

Jobs are calculated at run time:

Using environment variables

$ JEST_JOBS_TOTAL=3 JEST_JOBS_INDEX=0 jest

• JEST_JOBS_TOTAL - total number of jobs
• JEST_JOBS_INDEX - index number of the job (starts with 0)

Using cli arguments

$ jest --jobsTotal=3 --jobsIndex=0

• --jobsTotal - total number of jobs
• --jobsIndex - index number of the job (starts with 0)

Custom test reporter

It is possible to read reports not only from a junit file but from any source.

Imagine a this scenario, Jest + jest-junit but saved in CircleCI or in some database.

module.exports = {
  /* your jest config */

  globals: {
    "split-tests": {
      async reader(tests) {
        const reports = await fetchReports(tests.map((t) => t.path));

        return reports;
      },
    },
  },
};

The reader is a function that has to return Array<{path: string; time: number}> (or as a Promise) and its first argument is a list of tests to be executed in Jest.