Excluding second argument should not throw, but instead mark the test as skipped

[javorosas]

💥 Regression Report

It is a common use case for many teams doing TDD to write the specs first (what the software should do)

it('should sum two numbers');

and implement the behavior later

it('should sum two numbers', () => {
  expect(sum(5, 4)).toBe(9);
});

sometimes these two tasks are even made by two different people.

#5558 introduced throwing an error if the second argument is missing, apparently lacking a proper discussion on the consequences of doing that change.

Last working version

Worked up to version: 22.4.2

Stopped working in version: 23.0.0

To Reproduce

Steps to reproduce the behavior:

Run jest for this test case:

it('should sum two numbers');

Expected behavior

0 tests passed
0 tests failed
1 test skipped

[mattphillips]

Hey @javorosas this issue appears to be the same as #6256

The recommended solution is to add a noop callback:

it.skip('should sum two numbers', () => {});

[rickhanlonii]

Not a regression as it's listed as a breaking change in the changelog, but I'm cool to revert this and add a lint rule to eslint-plugin-jest instead

We should probably still fail tests that provide a function and no description:

test(() => {});

Since we need the description to give to reporters and track some internal state

[tristan-shelton]

This one is driving me nuts. Please do at least add a configuration option for this!

[rickhanlonii]

@tristan-shelton @javorosas we're adding test.todo for this, but if you want to use the previous behavior, here's the config you can use:

// add to setupTestFrameworkScriptFile
// see https://jestjs.io/docs/en/configuration#setuptestframeworkscriptfile-string

const originalTest = test;
test = (description, fn, ...args) => {
  if (!fn) return originalTest(description, () => {});

  return originalTest(description, fn, ...args);
};

test.only = originalTest.only;
test.skip = originalTest.skip;

cc @arcdev1

[tristan-shelton]

Thanks for the response @rickhanlonii . While that code will technically work (barring more unexpected api changes) users shouldn't have to resort to such hacks for the most basic expected behaviors of a test runner. The original pull request should never have been accepted, especially not as a default. This is frankly unacceptable.

I've since moved over to Mocha which provides this behavior by default, runs much faster and has far fewer bugs. I would recommend that anyone having this issue do the same until such time as Jest becomes usable again.

[elyobo]

The change annoyed me at first, but just providing a noop function is fine once I got in to the habit of it; I'm going to have to have that boilerplate when I write the test anyway, so it's not really even any extra work, it's just doing it at the time I write the stub instead of the test.

As for shifting to mocha for speed... well if that works for you, then good, but I've gone the from mocha to jest for the largely the same reason :D

[tristan-shelton]

@elyobo Last I checked a no-op function counts as a passing test, not a pending one. If you're writing a spec across multiple test files it's easy to lose track of unfinished tests that way.

[elyobo]

I should have been more clear; what you describe is the case for test but not for test.skip, which is what I'm using.

describe(' foo', () => {                                                       
  test.skip('does bar', () => {})                                             
})                                                                              

I stub things out like that so that they're correctly reported as skipped, the function that I'll need to write is there already, and the .skip (which did take a few extra keypresses) makes it easy to find the unimplemented tests by searching for .skip.

A small behavioural change for me, but no real pain. The old behaviour was slightly less work, and the change was to address a problem that I've never encountered myself, but now I'm used to it it's fine and if it helps some others then good.