📦 Move packaging to PEP 517 in-tree backend

[webknjaz]

This essentially allows the cythonization opt-out to be controlled by the --pure-python PEP 517 config setting that can be passed to the corresponding build frontends via their respective CLIs.

Are there changes in behavior for the user?

Yes, the users, and packagers are now supposed to explicitly pass the --pure-python config setting to their PEP 517 build frontends in use, when they need to make pure-python wheels.

$ python -m pip install . --config-settings=--pure-python=
$ python -m pip install .

$ python -m pip install -e . --config-settings=--pure-python=
$ python -m pip install -e .

$ python -m pip wheel . --no-deps --config-settings=--pure-python=
$ python -m pip wheel . --no-deps

$ python -m build --config-settings=--pure-python=
$ python -m build

$ python -m build --wheel --config-settings=--pure-python=
$ python -m build --wheel

Related issue number

N/A

Checklist

• I think the code is well written
• Unit tests for the changes exist
• Documentation reflects the changes
• Add a new news fragment into the CHANGES folder
  • name it <issue_id>.<type> (e.g. 588.bugfix)
  • if you don't have an issue_id change it to the pr id after creating the PR
  • ensure type is one of the following:
    • .feature: Signifying a new feature.
    • .bugfix: Signifying a bug fix.
    • .doc: Signifying a documentation improvement.
    • .removal: Signifying a deprecation or removal of public API.
    • .misc: A ticket has been closed, but it is not of interest to users.
  • Make sure to use full sentences with correct case and punctuation, for example: Fix issue with non-ascii contents in doctest text files.

[webknjaz]

NOTE: for some reason, pure-python wheels get an .so file too. Needs investigation.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (b81a868) 99.74% compared to head (70b26b6) 99.74%.

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #893   +/-   ##
=======================================
  Coverage   99.74%   99.74%           
=======================================
  Files           4        4           
  Lines         772      772           
  Branches      219      219           
=======================================
  Hits          770      770           
  Misses          2        2           

Flag	Coverage Δ
unit	99.61% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[webknjaz]

@mjpieters this still needs work, but you're welcome to poke around. Also, I'm not going to fix linting errors right now since the formatters are too invasive and make incorrect changes sometimes.

[mjpieters]

Also, I'm not going to fix linting errors right now since the formatters are too invasive and make incorrect changes sometimes.

That's fine for now. I'd be interested in examples of incorrect changes, if you have any?

[webknjaz]

Also, I'm not going to fix linting errors right now since the formatters are too invasive and make incorrect changes sometimes.

That's fine for now. I'd be interested in examples of incorrect changes, if you have any?

@mjpieters like moving noqa who-knows-where, so they don't work:

-from ._backend import (  # noqa: WPS436  # Re-exporting PEP 517 hooks
+from ._backend import (  # Re-exporting PEP 517 hooks

[webknjaz]

Status update: I made the packaging work. Will have to polish stuff before undrafting the PR. But in general, I'm pretty happy with the interface here.

[webknjaz]

So while working on this, I realized that introducing a --build-c-extensions inverses the default behavior of building C-extensions, making it an opt-in rather than an opt-out.
With that in mind, I think I'll replace it by --pure-python and preserve the opt-out way.