Sorry for the late reply, was a bit busy lately, but I still would like to share my thoughts on this.
For Apache Airflow we're dropping support for Python 2 in the next major release. We're now supporting Python 3.5+. Mostly because: - Easier to maintain and test, and less if/else constructions for the different Python versions. Also, not having to test against Python 2.x reduces the build matrix. - Python 3 has support for typing. From Python 3.5 you can include provisional type hints. An excellent presentation by Guido himself: https://www.youtube.com/watch?v=2wDvzy6Hgxg. From Python 3.5 it is still provisional, but it is a really good idea. From Airflow we've noticed that using mypy is catching bugs early: - This will put less stress on the (boring part of the) reviewing process since a lot of this stuff is checked automatically. - For new developers, it is easier to read the code because of the annotations. - Can be used as an input for generated documentation (or check if it still in sync with the docstrings) - Easier to extend the code since you know what kind of types you can expect, and your IDE will also pick up the hinting. - Python 2.x will be EOL end this year I have a strong preference to migrate everything to Python 3. Cheers, Fokko Op wo 7 aug. 2019 om 12:14 schreef Weichen Xu <weichen...@databricks.com>: > All right we could support both Python 2 and Python 3 for spark 3.0. > > On Wed, Aug 7, 2019 at 6:10 PM Hyukjin Kwon <gurwls...@gmail.com> wrote: > >> We didn't drop Python 2 yet although it's deprecated. So I think It >> should support both Python 2 and Python 3 at the current status. >> >> 2019년 8월 7일 (수) 오후 6:54, Weichen Xu <weichen...@databricks.com>님이 작성: >> >>> Hi all, >>> >>> I would like to discuss the compatibility for dev scripts. Because we >>> already decided to deprecate python2 in spark 3.0, for development scripts >>> under dev/ , we have two choice: >>> 1) Migration from Python 2 to Python 3 >>> 2) Support both Python 2 and Python 3 >>> >>> I tend to option (2) which is more friendly to maintenance. >>> >>> Regards, >>> Weichen >>> >>
