Cleanup after we drop support for Python 3.6

[Pierre-Sassoulas]

When we do drop support because 3.6 has reached its end of life, we'll need to do some cleanup:

• remove the code for Python 3.6 only and PY36_PLUS conditionals
• clean up the minimum required version in the tests and functional tests RC files (this has not been done previously)
• Use postponed evaluation of annotations (with from __future__ import annotations) -> Bump py-version and add python-typing-update to CI #6326
• Use dataclasses to simplify the code where it makes the most sense
• Bump py-version in pylintrc -> Bump py-version and add python-typing-update to CI #6326
• Look at coverage drop in Require python 3.7.2+ #5921 and remove dead code > Remove assign-to-new-keyword #6421

[cdce8p]

With the removal, we could also start using features that have been added in 3.7
The most relevant to us IMO would be:

• Postponed evaluation of annotations (with from __future__ import annotations)
• Dataclasses

https://www.python.org/downloads/release/python-370/
https://docs.python.org/3.7/whatsnew/3.7.html#new-features

[cdce8p]

Just for reference: #4310 (Issue discussing when the removal of 3.6 should happen)

[Pierre-Sassoulas]

Thank for the suggestion I added postponed evaluation to the list. Regarding dataclasses, could you explain the advantages compared to keeping NamedTuple for example ?

[cdce8p]

Dataclasses are essentially just classes. So compared to NamedTuple you can change attributes without creating a new instance of it. Where dataclasses really shine is for classes that just implement some boilerplate code (__init__ to assign all attributes to instance attrs, basic __str__, __repr__, __eg__, ...). Basically, they are meant for classes that mostly store data (hence the name). https://docs.python.org/3/library/dataclasses.html

Good examples where to use it can be found in the PR for the Similarity checker.
https://github.com/PyCQA/pylint/blob/65a20b6c5075e1f1756c1e6033c20b289b6edbbf/pylint/checkers/similar.py#L97-L114

https://github.com/PyCQA/pylint/blob/65a20b6c5075e1f1756c1e6033c20b289b6edbbf/pylint/checkers/similar.py#L163-L189

[Pierre-Sassoulas]

I added dataclasses to the list :) I think it should be done in good intelligence though (ie: not everywhere for refactor's sake)

[cdce8p]

I added dataclasses to the list :) I think it should be done in good intelligence though (ie: not everywhere for refactor's sake)

Yeah, I completely agree. The initial list was more meant to show which new features we can take most advantage of with 3.7 as min requirement not necessarily that we should refactor old code because of it.

[DanielNoord]

@cdce8p Is your typing upgrade tool ready to use for pylint? Or will its default value of 3.8 be problematic for us (making it easier to do everything by hand)?

[cdce8p]

@cdce8p Is your typing upgrade tool ready to use for pylint? Or will its default value of 3.8 be problematic for us (making it easier to do everything by hand)?

I just released a new version 0.4.0 which also added an option for 3.7: --py37-plus. Form a quick test, it looks like it's good to go. It's probably best to add it as a temporary pre-commit hook. At least that's the easiest way I found.

  - repo: https://github.com/cdce8p/python-typing-update
    rev: v0.4.0
    hooks:
      - id: python-typing-update
        stages: [manual]
        additional_dependencies:
          - "black==22.3.0"
        args:
          - --py37-plus
          - --force
          - --keep-updates
          - --black
        files: ^(pylint)/.+\.py$  # might need to be changed

pre-commit run --hook-stage manual python-typing-update --all-files

[DanielNoord]

I'll take a look at this after argparse migration is finished!