Hey all! I have integration tests that spin up cloud infra, then run tests using the newly created resources. These typically take 10+ minutes. What's the recommended way to approach these kinds of tests with pants? It seems pants works much better with unit tests, interested to hear best practices for integration testing.

I'm not sure how "good" a practice this is, but I have a process that uses both pants tags and pytest marks. All my integration tests are marked with "integration", and I have a separate

python_tests

targets that specifically pull in

integration_test.py

and is tagged "integration"

python_tests(
	name="integration-tests",
	sources=["integration_test.py"],
	tags=["integration"],
)

I can then do fun things like have pytest automatically skip integration tests. Or you can have pants run (or skip) integration tests, either at the pants level (

pants --tag="integration" test ::

) or at the pytest level

pants test :: -- -m "integration"

. You probably only need one of those, but I like the flexibility. There are a few advantages one way or the other. Skipping with pytest is annoying because you can end up with collection errors if all tests are deselected and you need to add a fake empty test.

Tagging looks excellent! I'm wondering if we could maybe use the bash script targets to setup tests first, also like your approach!