Plans to support Python3?

[rw]

Are there any plans to support Python3?

[dlecocq]

Not that I'm aware of. Though of course, we welcome pull requests :-)

[codinguncut]

I have started to port it to python 3, and would be happy to PR, but am still having some issues with cython "str" vs. "unicode"...
It may be necessary to retrain the sklearn model re: sklearn versions, but that should be OK if given training data, code and instructions are complete.
Another option may be cpy2py, but I haven't quite gotten it to work yet...

[matt-peters]

Excellent! How far along are you? You may want to coordinate with @bdewilde as has also given a python 3 port some thought. Also, there is an open PR #41 that will likely include some API changes and other large changes that may cause some conflicts with your branch.

[rw]

I think if it were easier to regenerate the model then you could have the model serialized for both 2 and 3. Then, you could just switch based on the version (I think :-) ).

[nehalecky]

Hey all, jumping in the py3 compatibility discussion for dragnet with @bdewilde on his PR #41 (which includes progress towards python 3 compatibility). Thought it best if we collaborate on this effort and break apart distinct efforts for refactoring—a py 3 porting roadmap—that could be defined by project leads and key contributors so we can parallelize the effort. Can we work towards this?

A few questions:

• are unit tests up to date for supporting the refactor?
• @bdewilde and others have mentioned significant effort regarding the blocks.pyx cython module, can we get a little more color on the effort there?
• @rw and @codinguncut seem to also have ideas regarding the porting effort, was there any collaboration regarding such with @bdewilde about this. Clearly would be great to have a common vision towards any API changes and general designs. ;)

Thanks all, excited about joining the effort here. 🎆

[matt-peters]

I'm 100% supportive of an effort to port to Python 3. Unfortunately I won't be able to help much other then code reviews and merging PRs. For a roadmap, I'd suggest we first get #41 merged, then someone should take a look at blocks.pyx and figure out the major hurdles. I leave it to interested parties (@nehalecky, @bdewilde, @rw, @codinguncut) to self organize around these.

Regarding some context on blocks.pyx -- this is the major bit of code that interacts with the string parsing in lxml at a low level (via lxml's cython API). Unfortunately this bit of code is pretty ugly and has some performance hacks that obfuscate some of the logic. Fortunately it does have pretty good test coverage now and with luck some of the dirtier bits that handle string encoding won't be necessary in the Python 3 version.

[nehalecky]

Hey All, we finally got up our PR (#52) for the completed port to python 3 (lot's of heavy lifting in streamlining string parsing by @pakelley). Look forward to your feedback there!

[pakelley]

Should this be closed now that #52 has been merged?

[b4hand]

Yup.