pip-compile fails if setup.py contains the if __name__ == __main__

[ssbarnea]

pip-compile will display an tracedump if setup.py contains the code such:

if __name__ == "__main__":
    ...

That may be problematic as this construct is quite popular: https://github.com/search?q=%22if+__name__%22+filename%3Asetup.py+path%3A%2F&type=Code&ref=advsearch&l=&l=

[atugushev]

Couldn't reproduce it locally:

# setup.py

from setuptools import setup

if __name__ == "__main__":
    print('main')

setup(name='foo', version='0.1', install_requires=['six'])

$ pip-compile setup.py
#
# This file is autogenerated by pip-compile
# To update, run:
#
#    pip-compile setup.py
#
six==1.15.0
    # via foo (setup.py)

[jdufresne]

IIUC, I believe the scenario would be more like:

# setup.py

from setuptools import setup

if __name__ == "__main__":
    setup(name='foo', version='0.1', install_requires=['six'])

Which makes sense that pip-tools fails. This construct tells Python to only run setuptools if setup.py is executed as the main script, but pip-tools doesn't do that. pip-tools imports setup.py at runtime from within a different main. So, IMO, the if __name__ == "__main__": should be dropped from user code.

Perhaps it would make sense to document this gotcha, but I don't think a code change in pip-tools is required.

[ssbarnea]

Considering that the if __name__ == "__main__" idiom was used and advertised extensively, likely even used by more than 50% of setup.py ever published, it would be highly desirable to find a trick to make it work, if even possible.

if not possible, pip-tools should fail with a clear message that includes a hint that explains that removing the construct should make it work.

[webknjaz]

Agreed, asking the users to dump the best practice sounds wrong. As for the trick, it should be possible to do the same sort of eval setuptools do.

Also, there's https://github.com/jaraco/jaraco.packaging that may have some useful recipes.

[atugushev]

Fixed in #1311. Thanks for the issue!