fix: Integration tests for async sdk method by breno-costa · Pull Request #4201 · feast-dev/feast

What this PR does / why we need it:

Follow up on #4172 After checking integration tests, I understand this one is relevant to the async method. Let me know if you understand more integration tests are needed.

Which issue(s) this PR fixes:

Fixes

I wonder it there's a way to do this without introducing another test. We could eventually use @pytest.mark.parametrize("use_async_client", [True, False], ids=lambda v: str(v)) on existing tests, but not sure how would we be able to right now limit async tests to only redis (without also limiting sync ones).

I wonder it there's a way to do this without introducing another test. We could eventually use @pytest.mark.parametrize("use_async_client", [True, False], ids=lambda v: str(v)) on existing tests, but not sure how would we be able to right now limit async tests to only redis (without also limiting sync ones).

I'm not sure whether it's possible to configure tests in this way. I refactored a little bit to reuse same setup and assertions for both sync and async sdk methods. This may help. wdyt?

thanks, this looks a lot better.

@breno-costa Can you take a look here? you can ignore test_historical_features fails, but others seem legit.

@breno-costa Can you take a look here? you can ignore test_historical_features fails, but others seem legit.

Fixed the issue.
Would it be possible to run integration tests locally?

@breno-costa thanks. There's a test-python-integration-local command in Makefile, It's used to run integration tests except for the ones that depend on AWS/GCP accounts.