pytest fixture with parametrization from another fixture

[ws1088]

I am facing exactly the same situation as the poster of: https://stackoverflow.com/questions/54105134/pytest-fixture-with-parametrization-from-another-fixture

And I also search a lot of places and couldn't find a good answer. I wonder if anyone has come across the same problem and has found a good solution?

Thanks!

[portyu9]

Hi @ws1088

The fundamental limitation is timing:

• parametrization creates test items during collection;
• fixtures execute later during test setup.

So a fixture cannot return a list at runtime and then ask pytest to retroactively create one independent test item for every returned value.

If the objects can be discovered during collection, move that discovery into pytest_generate_tests and parametrize descriptors/IDs there. The actual connection/object creation can still happen later through an indirect fixture.

If the objects genuinely cannot be known until the live fixture is running, keep a runtime loop instead.

On current pytest, the built-in subtests fixture is useful here:

def test_server_objects(server_objects, subtests):
    for obj in server_objects:
        with subtests.test(id=obj.id):
            assert obj.is_valid()

That gives each runtime object a separately reported subtest instead of aborting the whole loop on the first failed assertion.

The tradeoff is that these are subtests, not normal collection-time parametrized node IDs, so they do not have exactly the same individual selection semantics as @pytest.mark.parametrize.

Please mark this answer as accepted if it helped, thank you.