Open corneliusroemer opened 1 week ago
IIRC, the reason @fengelniederhammer and Tobias set these tests up really on is that there's no good Astro testing framework. So these were always intended to be mostly website tests. Not covering the whole system (at least not the whole system as it is now).
Maybe the name E2E was sort of ok at the start (even then though the mocking means it's not E2E). As soon as loculus comprised ingest, preprocessing, silo preprocessing, they should have been renamed to something more appropriate.
Calling them E2E has the implications that people misunderstand what is actually tested by them (and what isn't), e.g. Anya thought they tested at least prepro and silo prepro, and didn't mock so much
I think that we need to rework the tests majorly and that there are a few considerations:
So these were always intended to be mostly website tests.
That's not entirely true. As we wrote the tests at the time, it served two purposes:
Also at the time, there was no real preprocessing pipeline, so we had to mock one.
I agree that it might be good to separate the two purposes. AFAIK playwright can also mock backends, which could be used for testing the website. But also for end-to-end tests, we could still use playwright.
The fact we call the playwright website tests E2E is misleading: so much is mocked that it's really just an integration test to check.