msw does not play well with jests describe blocks

Environment

Name	Version
msw	latest
node	14.6.0
OS	Win10

Request handlers

https://github.com/ljosberinn/personal-react-boilerplate/blob/master/src/client/i18n.test.ts#L21

Actual request

https://github.com/ljosberinn/personal-react-boilerplate/blob/master/src/client/i18n.ts#L209

Current behavior

Trying out this eslint rule which is considered best practice, at least by that plugin, I moved all jest hooks including setupServer into describe blocks.

However, it appears msw just stops working then. Any fetch request will not be intercepted.

If this is a known limitation, it’d be great if the docs mention it; I know the documented usage is outside, but that might also just be because describe isn’t used in the docs at all.

Expected behavior

Continues working when setting up within describe.

Screenshots

I initially wrote the issue yesterday but accidentally didn’t submit so forgive me here if I add screenshots later.

[kettanaito]

In different test cases, I append the same route, with different responses to test different API outcomes. From what you just said this is technically discouraged too?

No, what you are doing is perfectly fine. As @marcosvega91 has mentioned, just make sure to reset the handlers between the tests with server.resetHandlers(), so the changes you make per-test do not leak into irrelevant test suites.

For posterity, I’m going to include the solution that worked for you.

Solution

• Declare a single instance of server (ideally, in jest’s setupFilesAfterEnv configuration option).
• Prepend necessary request handlers per test via server.use() API.
• Don’t forget to reset the runtime state of request handlers with server.resetHandlers(), usually in the afterEach jest hook.

[kettanaito]

Hey, @brianzhou13. Thanks for checking in before creating an issue.

Yes, that statement still holds today, and the solution above is something I’d recommend for MSW users. We may look into this as a part of #474 in the future. I’ve also answered you in the discussion (#821) regarding how to approach the combination of the global MSW setup while still having the per-suite control over your API mocks.