Hi. I think it is preferable if we stick to a single project, which gets used as a runtime dependency, and since six is already used in many projects, I'd continue to use it everywhere. The 2to3 script or versions thereof shouldn't introduce a new runtime dependency, as you only need to check in the code produced by these scripts, so you can use those just fine. I've also used the zope.fixers project, which extends 2to3 to apply some zope.interface specific conversions. I did run 2to3 on the Zope2 project and the result wasn't all too helpful. There were many instances were the conversion produced invalid results or changed behavior in undesirable ways, so I ended up manually fixing and looking at each change anyways. One thing I do find helpful is running a Python 3 pep8/flake8 over the codebase, as it points out most of the places where you have invalid Python 3 syntax. Just my 2 cents, Hanno On Tue, Oct 4, 2016, at 11:29, Martin Häcker wrote:
Hi python3-porters,
At the Ressurection-Sprint in Halle, I learned that most of python 3 porting in zope is done using six <http://pythonhosted.org/six/>.
I've stuck to it during the work there, as it seemed to be the standard that other zope packages have adopted.
But since futurize <http://python-future.org> offers a much more comprehensible implementation of the 2to3 Script and seems to be better maintained - I'd like to use it instead. Especially because it promises to make porting to py3 while maintaining py2 compatibility easier and faster. Any good reasons (besides consistency) not to use it? Is there some sort of consensus on this topic that has been reached before?
Best Regards, Martin Häcker