E2E Mocking

When testing things, sometimes you need to fake things out. Imagine you sell plush Koalas, and you want to test the checkout flow. If you were to actually make a request to your payment processor, you'd be charged every time you ran your tests. While it would be fun to have hundreds of Koalas in your house, you'd probably go broke. πŸ˜…
It's a good idea to fake out certain things when testing. Especially third parties. And this isn't just a good idea for testing, it's also a good idea for development as well.
Personally, I prefer to be able to do all my work completely offline, so I mock out all third-party requests during development as well as in testing. This helps a great deal with my productivity.
That said...
Whenever you make a fake version of something during testing, you're losing confidence!
Make certain there's a legitimate benefit for that cost of confidence. Making a fake version of something in testing is called "mocking" and is often necessary, but goes against the principle of ensuring your tests resemble the way the software is used.
Read more about this in The Merits of Mocking.
Another thing you want to watch out for, though, is to make sure you don't impact your source code to make it easier to mock. This is why it's great to have a tool capable of unobtrusively intercepting HTTP requests made by your app. And that tool is called MSW (short for "Mock Service Worker").
MSW originated as a utility for mocking browser requests but is more often used in Node as a testing utility. In Remix apps, it's pretty much only used on the server during testing and development because, typically, most (if not all) external network requests are made on the server.
πŸ“œ You'll find the docs for server setup here. Here's a quick example, modified from the docs:
import { rest, HttpResponse } from 'msw'
import { setupServer } from 'msw/node'
import closeWithGrace from 'close-with-grace'

const { json } = HttpResponse

const server = setupServer(
	// Describe the requests to mock.
	http.get('/book/:bookId', () => {
		return json({
			title: 'Lord of the Rings',
			author: 'J. R. R. Tolkien',
		})
	}),
)

server.listen({ onUnhandledRequest: 'warn' })

if (!process.env.VITEST_POOL_ID) {
	console.info('πŸ”Ά Mock server installed')
}
closeWithGrace(() => {
	server.close()
})
In general, it's best to have the default HTTP mocks handle the happy path, though I do recommend you try to make them as similar to the real deal as reasonable. I often use faker to generate fake data for what's returned from the APIs and even do basic validation on the requests as well (make sure the Authorization header is present and the request body follows the schema, etc).

With Playwright

One of the tricky things we run into when running tests against our app in Playwright is that we can't easily communicate between the playwright server and our app server because they are running in different processes. The best way I've found to communicate is through the file system. However, this is a bit limited, which is one of the reasons I typically focus E2E tests on happy path user flows and edge cases are handled by integration tests with vitest.
When your tests and source code run in the same process, you can take advantage of MSW's server.use function to alter the default handlers, which allows you to test situations where the HTTP request fails or returns an error.
MSW is currently working on a way to improve process-to-process communication for this use case and it's really pretty cool. It will allow us to have much more sophisticated HTTP mocks (using server.use, which we'll do later in the workshop with vitest).

Even with that, writing the emails to the file system has some benefits (like being able to manually inspect the emails that were sent).