Python3 support - #51

Hi! Thank you for the patch.

I'm not sure how to best take advantage of it, though, because I can't abandon support for Python 2, and I don't want to support two codebases.

Is it possible to make a single codebase that works in both Python 2 and Python 3, perhaps by using the six library? http://packages.python.org/six/

Regards,

Zooko

Uhm... I don't think six is a good idea. Can you, perhaps, just push it as a separate branch, so people know it is there? Or I can perhaps take over the support for python 3 variant until such time comes that no one needs python 2 anymore :)

I think a major portion of the library
will need to be refactored to make it
work with Python 3.

Using absolute imports may be useful.

Cheers.

I'm willing to look into a Python 2 & 3 port. What's the minimum version of Python 2 which must be supported? The job gets substantially easier with only Python 2.6.

I've been writing one here at
https://github.com/gorakhargosh/pyoauth that
should work on Python 2 and 3 starting at
Python 2.5. However, due to the lack of time
these days, I've been unable to work on it
lately.

Have a look if you're interested.

Cheers, =)

@gorakhargosh : I think it makes sense to focus on python-oauth2, because many people are already using it (going by PyPI download counts), and they'll be reluctant to port their code to another OAuth library in order to use Python 3.

So what would need doing to support Python 2 & 3 from one codebase? The options are roughly:

• Have Python 2 source, and run 2to3 during setup for Python 3. Requires less initial effort, but testing is slowed down by the conversion step. Instructions here.
• Use six or an equivalent abstraction layer to abstract away the differences. This mostly shouldn't be too hard, but catching some exceptions will become ugly, unless we can move the minimum Python version up to 2.6, which supports the new except Exception as e syntax. It looks like there's about a dozen cases of that, so it's not too bad if they have to be ugly. Abstracting u('Unicode strings) and b('byte strings') is also not very beautiful.

@jaitaiwan can you take a look at this? It's related to #158. @hades thanks for the PR! 🎉

@joestump added to the list. I have a feeling that my PR doesn't use the futures lib... but I don't think any others do either. Will continue the chat in #158.

This one fails to work in Pyk2 so I think we may drop it https://travis-ci.org/CentricWebEstate/python-oauth2/jobs/75874253

Thoughts @rickhanlonii @joestump?

Yeah, there are other pk3 wips that are farther along at this point. Thanks for the contribution @hades!