What is the purpose of `beforeEach` global in Jest?

Jest documentation recommends beforeEach for tests that consume a particular global state for each test, for example, resetting test data in a database before each test is run.

Note the following case:

If beforeEach is inside a describe block, it runs for each test in the describe block.

Using the same example where we have a database with test data, if your tests do not need the test data reset for each test, you can use beforeAll to run some code once, before any tests run.

If you have code that’s common across multiple tests, you can use beforeEach to do some setup before each test runs in order to avoid repetition. In this case, if you have multiple tests that shallow mount Component, you can move the shallow mount to a beforeEach, and the component will be mounted when every test runs. Generally you’ll want to pair this with an afterEach, where you call wrapper.unmount().

describe('tests', () => {
  it('does one thing', () => {
    const wrapper = shallow(<Component />);
    // ...test some things
    wrapper.unmount();
  });

  it('does another thing', () => {
    const wrapper = shallow(<Component />);
    // ...test something else
    wrapper.unmount();
  });

  it('does a third thing', () => {
    const wrapper = shallow(<Component />);
    // ...test a third thing
    wrapper.unmount();
  });
});

becomes:

describe('tests', () => {
  let wrapper;

  beforeEach(() => {
    wrapper = shallow(<Component />);
  });

  afterEach(() => {
    wrapper.unmount();
  });

  it('does something', () => {
    // ...test something
  });

  it('does something else', () => {
    // ...test something else
  });

  it('does another thing', () => {
    // ...test a third something
  });
});

beforeEach is referred to as the “setup” phase, and afterEach as the “teardown” phase.

Are we not deliberately slowing down the test runtime?

No, because you would have to shallow mount the component in each test anyway.

Isn’t it ok to render it once and assign it to wrapper?

Persisting a component (or any statefulness) across multiple tests can cause flaky tests, because (for example) you might get a different result if the tests are run in a different order. Any state (such as a mounted component) should be set up before each test and torn down after each test. This makes your tests completely independent from each other.