cases_generator doesn't work with parametrize_with_cases

[saroad2]

As of the latest version, cases_data is deprecated and the user is instructed to replace it with parametrize_with_cases.
I tried to replace it as instructed, and got a weird error. A little digging showed me that the problem is that parametrize_with_cases doesn't play nice with cases_generator.

Here is a simple example:

Assume I have the following code:

from pytest_cases import cases_generator, cases_data, THIS_MODULE

@cases_generator("i={i}", i=range(1, 10))
def case_i(i):
    return i + 1


@cases_data(module=THIS_MODULE)
def test_me(case_data):
    j = case_data.get()
    assert j == 0

If I want to replace cases_data with parametrize_with_cases, it should look like this:

from pytest_cases import cases_generator, parametrize_with_cases, THIS_MODULE

@cases_generator("i={i}", i=range(1, 10))
def case_i(i):
    return i + 1


@parametrize_with_cases(argnames="j", cases=THIS_MODULE)
def test_me(j):
    assert j == 0

Even though those tests should fail, the error I'm getting is:

test setup failed
file C:\Users\saroa\repos\eddington\fitting_tests\test_fitting_algorithm.py, line 17: source code not available
file <makefun-gen-8>, line 1: source code not available
file <makefun-gen-6>, line 1: source code not available
E       recursive dependency involving fixture 'i' detected
>       available fixtures: cache, capfd, capfdbinary, caplog, capsys, capsysbinary, class_mocker, clear_fix, doctest_namespace, i, mocker, module_mocker, monkeypatch, package_mocker, pytestconfig, record_property, record_testsuite_property, record_xml_attribute, recwarn, session_mocker, test_me_j, tmp_path, tmp_path_factory, tmpdir, tmpdir_factory
>       use 'pytest --fixtures [testpath]' for help on them.

Is cases_generator deprecated too? If it is, what should I use instead? If it doesn't, it's a bug :)

[smarie]

I @saroad2 thanks for spotting this ! Indeed case_generator is deprecated, but it should work through retrocompatibility.
I had a look now and was able to reproduce the bug, so I'll fix this asap. I'll also provide an example of the new style.

[saroad2]

Thank you mate for the quick response!

I think it doesn't warn that cases_generator is deprecated as it is for cases_data. Maybe you should add this warning.

So my next question is: what is the alternative for cases_generator?
My use case is this: I use cases_generator in order to read cases from json files in a specific directory. Each time I add a json file, it reads it automatically and adds it as a test. You can take a look at it here.

What should I use instead of cases_generator? I must admit that this is a very good ability and I don't really sure why it is deprecated.

[smarie]

Yes, I'll add the deprecation warning. So the conclusion is that @case_generator works as long as it is used with @cases_data (old API continues to work), but if you change one you have to change the other.

This is how you do the same with the new API :

from pytest_cases import parametrize, parametrize_with_cases

@parametrize(i=range(3), idgen="i=={i}")
def case_i(i):
    return i + 1

@parametrize_with_cases(argnames="j", cases='.')
def test_me(j):
    assert j > 0

As you can see the difference is that there is no specific decorator for cases any more: cases can be parametrized with the same decorators than normal pytest test functions. In the example above I use the enhanced @parametrized from pytest-cases so that you can benefit from the alternate **kwargs parametrization style, but you can use the one from pytest if you wish: below an example where part of the cases are parametrized with @parametrize and the other part with pytest.mark.parametrize:

import pytest
from pytest_cases import parametrize, parametrize_with_cases

@parametrize(i=range(2), idgen="i=={i}")
def case_i(i):
    return i + 1

@pytest.mark.parametrize('i', range(2), ids="i=={}".format)
def case_k(i):
    return i + 1

@parametrize_with_cases(argnames="j", cases='.')
def test_me(j):
    assert j > 0

def test_synthesis(module_results_dct):
    assert list(module_results_dct) == [
        'test_me[i-i==0]',
        'test_me[i-i==1]',
        'test_me[k-i==0]',
        'test_me[k-i==1]',
    ]

Let me know if this works for you

EDIT: oh, also you can see above that there is a more concise way to indicate "this module", since now relative module names can be used ; simply use '.' . (but THIS_MODULE is not deprecated and will continue to work).

[smarie]

(closed automatically from commit message for 2.1.3 release but of course feel free to reopen if the above does not work for you)