Extend _pytest.python._idval to return __name__ of functions as well

[st--]

I want to parametrize over factory functions, which I believe is a valid use-case. At the moment, the autogenerated id for the test name is then based on the argument name, plus counter to disambiguate (e.g. feature_factory0, feature_factory1, etc.). With this PR, the default behavior for functions will be the same as for classes: return the __name__. I think this is a more sensible default behaviour.

Semi-pseudocode example:

def factory_IP(T, Kernel): return do_something_with(T, Kernel)
def factory_FF(T, Kernel): return do_something_else(T, Kernel)

pytest.mark.parametrize('feature_factory,kernel_class', [
    (factory_IP, Kernel1),
    (factory_IP, Kernel2),
    (factory_FF, Kernel1),
    (factory_FF, Kernel3),
])
def test_my_model(feature_factory, kernel_class):
    """test code"""

As Kernel1, Kernel2, Kernel3 are classes, currently, the tests would be named as follows:

test_my_model[feature_factory0-Kernel1]
test_my_model[feature_factory1-Kernel2]
test_my_model[feature_factory2-Kernel1]
test_my_model[feature_factory3-Kernel3]

With this PR, the test names would be:

test_my_model[factory_IP-Kernel1]
test_my_model[factory_IP-Kernel2]
test_my_model[factory_FF-Kernel1]
test_my_model[factory_FF-Kernel3]

This allows you to immediately figure out which factory function is being used, and doesn't hide the fact that "feature_factory0" and "feature_factory1" are the same thing.

I hope you'll be happy to merge this :-)

PS: Pytest is great, thanks!!

[coveralls]

Coverage remained the same at 92.697% when pulling 5085aa2 on st--:master into 88ed1ab on pytest-dev:master.

[coveralls]

Coverage remained the same at 92.697% when pulling 5085aa2 on st--:master into 88ed1ab on pytest-dev:master.

[The-Compiler]

This should probably go to the features branch.

[nicoddemus]

Thanks @st-- for the PR!

Agree with @The-Compiler, changed the target to features. Also @st-- can you please add a test for it to avoid future regressions? Thanks!

[coveralls]

Coverage remained the same at 92.697% when pulling fdd4abb on st--:master into 88ed1ab on pytest-dev:features.

[nicoddemus]

Please add a test for it to avoid future regressions.

[st--]

Hadn't written a test because most of the other cases of _idval didn't have any, either. Happy to add one - I guess in testing/python/metafunc.py ? Should I add one for the isclass() case as well while I'm at it?

[nicoddemus]

Happy to add one - I guess in testing/python/metafunc.py ? Should I add one for the isclass() case as well while I'm at it?

Both suggestions sound good, thanks!

[coveralls]

Coverage remained the same at 92.697% when pulling 1fe2e2c on st--:master into 88ed1ab on pytest-dev:features.

[coveralls]

Coverage remained the same at 92.697% when pulling 652936f on st--:master into 88ed1ab on pytest-dev:features.

[st--]

Test added, travis passed :)

[nicoddemus]

Great, thank you!