Support for parameterized tests is present in other ports of the approval test tool (see for example https://stackoverflow.com/questions/49035010/approvaltests-with-testcase-nunit-use-the-same-approval-file how it's done in C#). We should have this in the python version too. It's just a question of working out how to make it pythonic and then adding it.

The recent addition of ScenarioNamer seems to make parameterized pytest tests possible. I'm satisfied with closing this issue, but I'll leave that up to you in case you've got further work you'd like to do.

You totally saved me with this issue!
All it took after reading it is:

@pytest.fixture
def verify(request):
    def verify(data, reporter=None):
        verify_with_namer(data, get_scenario_namer(request.node.name), reporter)

    return verify

Unfortunately, this causes duplication in the test name. To fix it, and still match the names printed by pytest, we need to modify ScenarioNamer a bit.

That looks neat! How about putting this code in approvaltests/core/scenario_namer.py? Do you think you could also add a test to tests/test_scenarios.py? I'd like to better understand how this code works and what changes you need in ScenarioNamer.

Emily

PR is up. I added a comment on the pytest_verify fixture as I was unsure where to place it.

Hello there!
Supporting parametrized tests would be great feature. Any updates for this issue?

Any idea when this might be supported? Or is the official solution to use namers?

Honestly I thought it was already fixed. I know there's been quite a bit of development work done on python approvals lately. @isidore can you comment?

Through the use of 'namers' we can make this work, but I wasn't sure if that was the expected way of supporting parametrization or if the pytest integration would do it automagically.

I see some nice examples from @isidore and @SwamyDev from the last few months. Thanks both.

Would it be correct that this is the "official solution"? ( I like it)

AFAICT there is nothing currently that would do this naming automagically. Given some of the more complex cases (below) that makes sense - seems non-trivial. But, I wonder if it would be useful to throw a warning or error for parameterized tests (pytests) that don't provider a namer - I found it surprising on my first use of approvaltests with a parameterized test that only a single pair of received/approved files were generated.

Some more complex cases:

1. multiple parameters, combinatorial:
   • from https://docs.pytest.org/en/6.2.x/parametrize.html#pytest-mark-parametrize

     @pytest.mark.parametrize("x", [0, 1])
     @pytest.mark.parametrize("y", [2, 3])
     def test_foo(x, y):
         pass
     

   • which will run the test with the arguments set to x=0/y=2, x=1/y=2, x=0/y=3, and x=1/y=3
2. Arguments are not constrained to strings - making autonaming challenging
   • I have some fixtures that generate whole data structures that are passed in as parameterized arguments.

Autonaming ought to be trivial with pytest since pytest generates unique names for you in the test node's 'nodeid' field. They're not always pretty names, so a fallback to using a namer would probably be important for some users. Nevertheless, everything we need to do autonaming is already available (as far as I know...I don't know if any situations where pytest won't generate a unique ID, and I don't know how it would operate if it couldn't).

See for example how this is done in godkjenn's pytest plugin.

@abingham - oh that's great. Thanks for the knowledge. Autonaming would be great, but manual works for me for now.

I'd be happy to review/test autonaming if that'd would be helpful