RFC: Make boltons Python 3.7+ only.

[mahmoud]

Not sure how much explication is necessary here, but it's 2023, what better time to finally jump from 2/3 to just 3?

Off the top of my head:

Pros:

1. Get rid of all the PY2/PY3 conditionals cluttering up the code
2. Generally make contribution easier (see the Py2.7 failure in the initial commit of fix: make copy_function work with keyword-only defaults #336 ). Most potential contributors probably don't have a Python 2 installed.

Cons:

1. Will lose some potential utility for something like ecoutils to detect older Python installations

As for the con, there's always the possibility of using older boltons versions, most of the modules have been pretty stable for a while.

[kdeldycke]

Agreed. And Python 3.7+ support is generous IMHO, given Python 3.7 will stop receiving security updates after 2023-06-27 and will be considered EOL.

[rafalkrupinski]

Why not 3.6 or 3.8?
if dropping support for EOL Python -> 3.8
if dropping support for Python 2 -> 3.6 is often said to be the first good version in the 3.x line.

I'm not sure what supporting any of the versions mean for boltons, just saying from a user perspective.

[mahmoud]

@rafalkrupinski / @kdeldycke agreed, all my use cases are on 3.9+, but I had one old contact reach out about still being on 3.7. A quick check of pypi stats shows that 3.7 gets around 5-7.75% of downloads, which is appx ten times as many downloads as 2.7, 2-3x as many as 3.6, about half as many as 3.8.

Another way of looking at it would be to focus on benefits of upgrades. Is there a particular development we're looking forward to taking advantage of in 3.8 vs 3.7? 3.7 has ordered dictionaries and from __future__ import annotations, which are the two I think I use the most from modern Python.

[kdeldycke]

@mahmoud I am impressed by your data-backed maintainership! 🤗 My proposal was just to go with the flow of Python mainline support. That's it. To me it only depends on the effort you would be ready to pour in into supported older versions.

[Avasam]

#190 would be the biggest reason for 3.7 with from __future__ import annotations
With 3.7 in EOL. 3.8 would make it slightly easier to use generic types.

[mgaitan]

+1 to make it 3.7+ only

[hugovk]

Please see PR #362 to drop 3.6 and earlier.

If it's too early for a PR, feel free to close it and use it as an example of what can be done.

[mahmoud]

The biggest PR in boltons history just landed and it's now released on PyPI as boltons==24.0.0. 🎉 Thanks everyone!