Better setup for tests

I used to put the code for preparing each test in before or setup block, and clean it up in after or teardown. But what if the tests don't share the same setup? Easy, I can write something like this:

test('A', () => {
	// preparing for test A
	...
});

test('B', () => {
	// preparing for test B
	...
});

But what if test B depends on test A:

test('A', () => {
	// preparing for test A
	...
});

test('B', () => {
	// Also depend on "preparing for test A"
	// preparing for test B
	// but if A fails or skipped, the setup of B is not complete
	...
});

Normally this should work if tests are run synchronously. However, A lot of times I only want to run one test with grep. Or, if test A fails, test B also fails but it shouldn't. I was tempted to do something like this:

beforeEach(() => {
	// testId indicates this setup is run immediately before which test
	if (testId === 'A') {
		// setup for test A
		...
	}

	if (testId === 'B') {
		// including the setup for test A
		// setup for test B
		...
	}
});

You might say it is very easy to find a smarter way. This is true for use-cases when your setup doesn't include things like setting up a whole mock git history.

$ npm install --save-dev better-than-before

With this module, you could simplify your setup using beautiful declarative DSL syntax:

import { setups,  preparing } from 'better-than-before';

setups([
	() => {
		// setup for all tests
	},
	() => {
		// setup for all tests except for tests only need the first setup
	},
	...
]);

test('A', () => {
	preparing(1); // I only need the first setup
});

test('B', () => {
	preparing(2); // I need both first and second setup
});

...

The most straightforward way of using this module is to put this in the test block. But test frameworks like mocha also time how long a test runs and you might easily hit timeout. I could hack into the framework but I decided not to. Make sure you increase your timeout and ignore the performance warnings. If you use this module in a performance framework make sure it's not run in the test block.

MIT © Steve Mao