I need to inspect the list of collected tests (specifically, module attributes) before executing any of them.

[ericmburgess]

I've written a fixture that downloads the files from S3 that are necessary to run the test. These files are large and there are lot of them, so we don't want to download everything each time, even if only a few tests are selected, and we don't want them in our git repository, either. It does this by using the value of getattr(request.module, "TEST_ID", None) to know where on S3 to get the needed files.

Everything is working perfectly, but I'm having an unexpected problem. My S3 access requires MFA, and the test suite takes so long to run that I'm prompted multiple times for an MFA code. Obviously I'm not sitting there watching it run, so this causes big delays in getting through the tests.

What I need to do is download the data for all of (and only) the selected tests before executing any of them.

• I see there's a pytest_collection_modifyitems hook, but it's unclear to me whether it's able to access the containing module the way my fixture does via request.
• If there is a way to run all the instances of a given fixture after collection and before execution, that would do the trick, too.
• Maybe there's another way around my problem?

[arthuRHD]

Hi, i'm not sure a hook is necessary here.

you can just create a session fixture which runs by itself; since it's a session fixture it already knows which tests are collected.

import runpy

def get_module_from_path(module_path):
    # Run the module and get the module dictionary
    module_dict = runpy.run_path(module_path)

    return module_dict.get("__main__")

@pytest.fixture(scope="session", autouse=True)
def download_files_session(request):
    test_modules = [get_module_from_path(item.reportinfo()[0]) for item in request.items]

    test_ids = [getattr(test_module, "TEST_ID", None) for test_module in test_modules]

    download_files_for_tests(tests_ids) # The part where you download S3 files idk

If it's not what you need, i'm sorry I didn't understand your problem

[arthuRHD]

its absolutely wrong to use runpy in those places, a module scope fixture for the download before running is all that would be needed without any of those hacks

Yep, I don't know why I needed to centralized into one fixture. Can't you elaborate on why it's "absolutely wrong" to use runpy in those places? It will be insightful, thanks.

[RonnyPfannschmidt]

Running a already imported modules in a new namespace using the tool too run cli modules just to extract data is simply ludicrous when one can just directly use the correctly imported already existing modules

[RonnyPfannschmidt]

Like just item.getparent(pytest.Module).obj

(Please validate, typed from memory on my mobile)

[arthuRHD]

My bad, I only saw item.reportinfo() in the documentation

[ericmburgess]

@arthuRHD if you'd like to edit your reply according to the feedback of @RonnyPfannschmidt, then I'll mark it as the answer. Thank you both!

item.getparent(pytest.Module).obj is in fact the correct syntax to get the module reference.

[RonnyPfannschmidt]

there is no builtin way to manage preloading of something that should be a module scoped session

i would recommend a plugin object that implement the pytest_modifyitems hook and the sessionstart hook

either eagerly download or use a concurrency mechanism (however thats possibly more risky

the pytest cache object can be used to cache the downloads if they are used immutable