Adding Python 3 support and fixed Python 2.6 bug.

[bworrell]

Hello,

I ran across some issues when using weakrefmethod on different versions of Python and took a shot at fixing them.

When running unit2 on Python 2.6, I got the following error:

======================================================================
FAIL: test_equality (test_weakmethod.WeakMethodTestCase)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "C:\dev\weakrefmethod\test_weakmethod.py", line 122, in test_equality
    _eq(a, b)
  File "C:\dev\weakrefmethod\test_weakmethod.py", line 111, in _eq
    self.assertFalse(a != b)
AssertionError: True is not false

----------------------------------------------------------------------

I think this is due to a weakref.ref behavior on Python 2.6. It seems the __ne__ method returns a NotImplementedType object, which evaluates to True when used in a boolean expression.

Example:

>>> import weakref
>>> def foo():
...     pass
...
>>>
>>> r1 = weakref.ref(foo)
>>> r2 = weakref.ref(foo)
>>>
>>> r1 == r2
True
>>>
>>> r1 != r2
False
>>>
>>> r1.__ne__(r2)
NotImplemented
>>>
>>> result = r1.__ne__(r2)
>>> bool(result)
True

On Python 3, the weakrefmethod/__init__.py imports fail because of the relative import syntax. I updated it to use explicit relative imports which works on Python 2.6+. After altering the imports and running the tests, weakrefmethod appears to work on Python 3.2 and 3.3.

I also added the tox.ini file that I used for testing across different Python versions.

Thanks for considering this pull request and thank you for providing weakrefmethod to the Python community!

[sjlongland]

Just noticed this package fails on Python 3.3: https://travis-ci.org/mriehl/fysom/jobs/104516422
I'll have to check and see if this pull request fixes it.

[bworrell]

@sjlongland, the PR should fix the issue you are seeing in Travis. The problem is that weakrefmethod uses implicit relative imports, which are not supported in in Python 3.

[twang817]

I'll update pypi a bit later tonight.

[sjlongland]

Excellent news, I'll be giving it a try shortly to see how it all goes. Many thanks. :-)

[bworrell]

Awesome! Thanks @twang817! Hope it works for you @sjlongland :)

[sjlongland]

For the record, I was waiting for the new package to arrive on pypi. In the meantime I decided to add a .travis.yml file (since I don't have Python 3.2/3.3 handy) and give it a shot there.

https://travis-ci.org/vrtsystems/weakrefmethod/jobs/105810978 ← Seems Python 3.2 isn't happy.

[sjlongland]

Okay, seems the .travis.yml file I grabbed (I nicked it from another repository) tries to use latest versions of modules like pip, and the latest pip does not like Python 3.2.

Still, it's worth knowing about 3.2 support because Debian Wheezy ships with it, and we have a lot of kit out in the field with Debian Wheezy.

https://travis-ci.org/vrtsystems/weakrefmethod shows all passing.

Once in pypi, I should be able to trigger a re-test of fysom and get mriehl/fysom#29 squared away.

[sjlongland]

Hi,

Silly question, what's the status of this package these days on pypi? Seems at this present time, any package tested on Python 3.0-3.3 inclusive fails still due to this issue when the package depends on this module.

I haven't been able to get mriehl/fysom#23 resolved as it presently still breaks on Python 3.3 due to incompatibilities in weakrefmethod-1.0.2.

I don't mean to sound impatient, but it has been a while now.

Regards,
Stuart Longland