Converted to run on Python 3 - all tests passing

[wbond]

These changes got everything working in Python 3.3. I tried testing it with 2.7, but my 2.7 install is acting up an I could not run nose.

Thanks for all of the tweaks and fixes you've made!

[ikirudennis]

First and foremost: thanks for not only using, but improving my fork.

I haven't started making the switch to Python 3 yet, so it hadn't even crossed my mind to start making this compatible. What you've done here is great, and luckily it only breaks two things:

• You removed a str.encode here. It's needed in order for it to pass this test in 2.7: u'"Übermensch":http://de/wikipedia.org/wiki/Übermensch' Looking up the documentation, it appears that it's redundant in Python 3. Would you rather it be put back for old times' sake (despite it being redundant) or find a way to branch there and work "natively" dependent on the version?
• Your edits to the tests remove the unicode prefix from strings. Was this a conscious choice? It breaks the tests in 2.7.

Again, thank you so much for your efforts.

[wbond]

Yeah, I thought that encode() would likely break something, but with python 3 you can't concatenate bytes and strings (bytes is the output of encode), so lots of tests break with all of those encode() calls. I changed the ur"" strings in the tests to just r"" since ur"" is not valid in Python 3 and r"" is the same thing.

The "right way" to get this all working is probably take a step back and do a revamp of the url code with encodings to make sure all strings are consistently encoded so they can be concatenated. Then fixing the tests may just be as simple as having an if/else branch for the python 2 and the python 3 versions of the strings.