Jean-Michel Pichavant wrote:
<div class="moz-text-flowed" style="font-family: -moz-fixed">Scott David Daniels wrote:
Jean-Michel Pichavant wrote:
Hello world,

I had recently a very nasty bug in my python application. The context is quite complex, but in the end the problem can be resume as follow:

2 files in the same directory :

lib.py:
 >import foo
 >foo.Foo.BOOM='lib'

foo.py:
 >class Foo:
 >    BOOM = 'Foooo'
 >
 >if __name__=='__main__':
 >    import lib # I'm expecting BOOM to be set to 'lib'
 >    print Foo.BOOM

I was expecting 'lib' as output, but I got 'Fooo'. I don't really understand what python mechanism I'm messing with but I have the feeling I've misunderstood a very basic concept about class, namespace or whatever import notion.


I guess there is 2 different objects for the same class Foo. How I do I make both Foo objects the same object ?

OK, here is one solution (from which you may infer the problem):

lib.py:
    import __main__
    __main__.Foo.BOOM = 'lib'

foo.py:
    class Foo:
        BOOM = 'Foooo'

    if __name__ == '__main__':
        import lib # I'm expecting BOOM to be set to 'lib'
        print(Foo.BOOM)

Here is another solution:

lib.py:
    import foo
    foo.Foo.BOOM = 'lib'

foo.py:
    class Foo:
        BOOM = 'Foooo'

    if __name__ == '__main__':
        import sys
        sys.modules['foo'] = sys.modules['__main__']
        import lib # I'm expecting BOOM to be set to 'lib'
        print(Foo.BOOM)

Here is a demo of what is actually going wrong:

foo.py:
    class Foo:
        inside = __name__

    import foo

    if __name__ == '__main__':
        print(Foo is foo.Foo)
        print(Foo.inside, foo.Foo.inside)

And here is a fix
foo.py:
    if __name__ == '__main__':
        import sys
        sys.modules['foo'] = sys.modules['__main__']

    class Foo:
        inside = __name__

    import foo

    if __name__ == '__main__':
        print(Foo is foo.Foo)
        print(Foo.inside, foo.Foo.inside)


--Scott David Daniels
scott.dani...@acm.org

Thanks for the explanation. I'll have to give it a second thought, I'm still missing something but I'll figure it out.

Jean-Michel

</div>

In general, two or more modules with mutual imports can cause problems. If modulea.py imports moduleb, and moduleb.py imports modulea, you can end up with difficulties.

But even more gross difficulties come from importing the module that started the execution, as it's got another name, "__main__" If you import it again with its traditional name, these two do not get combined, and you end up with multiple instances of things you thought were safe.

Scott has given you a solution for the second problem. But the first problem is more general, and should perhaps be the one you try first.

The simplest answer in many cases is to factor the script into two separate files. One will have "library" functions and classes, like the Foo in your example. And the other will have the logic for parsing command line arguments and suchlike.

The general problem isn't unique to Python. You can also end up with problems in C, where one header includes a second one, which re-includes the first. The workarounds for it are well known, but most of them also contain subtle difficulties.

If it's possible, factor out mutual dependencies into a separate module.

--
http://mail.python.org/mailman/listinfo/python-list

Reply via email to