Drop support for EOL Python 2.6 and 3.3?

[hugovk]

It looks like significant effort is going into maintaining support for EOL Python versions. When they reach EOL that means they're no longer receiving security updates.

https://en.wikipedia.org/wiki/CPython#Version_history

Here's the pip installs for jedi from PyPI for the last month (via pypinfo --percent --pip --markdown jedi pyversion).

python_version	percent	download_count
3.6	44.4%	155,231
3.5	27.5%	96,124
3.4	22.2%	77,788
2.7	5.2%	18,142
3.7	0.4%	1,559
3.3	0.2%	800
2.6	0.0%	30
3.2	0.0%	5

From these numbers, it looks like 2.6 and 3.3 can be safely dropped.

This would allow the code to be modernised and ease the maintenance burden quite a bit.

Thoughts?

[valignatev]

I would say that this is an absolutely beautiful idea. But I'm not that familiar with the Jedi source code, and I can't tell how this will simplify maintenance.

[hugovk]

Some of the most recent commits have been to try and wrangle 2.6 and 3.3 into place (eg. e526cb1 and 6291568).

It would look something like #1017, and some more of https://github.com/adamchainz/flake8-comprehensions/blob/master/README.rst#examples are possible too.

[davidhalter]

I'm definitely dropping Python 2.6 in the release that contains the virtualenv branch. Probably not dropping 3.3 though. That's probably going to take another year or at least a few months.

Thanks for the research!

[hugovk]

Sounds good. Of the two, there's definitely more benefit in dropping 2.6.

[davidhalter]

Closing this one in favor of #1018 and #1019. Thanks for your work!