45 only run the unit tests which touch changed code in the CI

[max-models]

Solves the following issue(s):

Closes #45

Core changes:

None

Model-specific changes:

None

Documentation changes:

None

[max-models]

It seems like the -k flag isn't working well with --testmon. When running the unit tests twice it still runs all of them unless I remove the -k flag. One option would be to categorize the tests in separate subdirectories instead and use --ignore=<PATH> to specify whether to run unit tests or model tests. I think it would be nice to organize the tests in src/struphy/tests with some subdirectories

[max-models]

Seems like testmon might be problematic for MPI tests https://github.com/struphy-hub/struphy/actions/runs/18819880612/job/53694176303

[max-models]

I find it a bit odd to import a plotting function from a testing file, should it not be the opposite way around?

[spossann]

Yes i agree, please correct.

[spossann]

It seems like the -k flag isn't working well with --testmon. When running the unit tests twice it still runs all of them unless I remove the -k flag. One option would be to categorize the tests in separate subdirectories instead and use --ignore=<PATH> to specify whether to run unit tests or model tests. I think it would be nice to organize the tests in src/struphy/tests with some subdirectories

We can remove the -k flag, move test_models.py and all test_verif_.py to a separate folder and use --ignore. I am against collecting all tests in one struphy/tests/ folder if not really necessary; it is much clearer when the test are next to the code they test.

[spossann]

Seems like testmon might be problematic for MPI tests https://github.com/struphy-hub/struphy/actions/runs/18819880612/job/53694176303

If this problem persists, we can try to shorten the mpi unit tests (using pytest.mark.mpi), for example by replacing the long convergence tests by simpler non-parameter-scan tests.

[max-models]

Seems like testmon might be problematic for MPI tests https://github.com/struphy-hub/struphy/actions/runs/18819880612/job/53694176303

If this problem persists, we can try to shorten the mpi unit tests (using pytest.mark.mpi), for example by replacing the long convergence tests by simpler non-parameter-scan tests.

This could be an option, but I think I can resolve it by only using --testmon to collect the tests which needs to be executed and then only running those, so we would do it in two steps: 1) filter test, 2) execute filtered tests. Now we do both of these in one go, and that seems not to work with --testmon.

[max-models]

Moved to #96