fix: The file structure of stubs resembles the "package" path.

[Masara]

Closes #81

Summary of Changes

Changed how the file structure for stubs is build. Now we follow the "package" path for the structure.

[github-actions]

🦙 MegaLinter status: ✅ SUCCESS

Descriptor	Linter	Files	Fixed	Errors	Elapsed time
✅ PYTHON	black	7	0	0	1.26s
✅ PYTHON	mypy	7		0	5.6s
✅ PYTHON	ruff	7	0	0	0.04s
✅ REPOSITORY	git_diff	yes		no	0.02s

See detailed report in MegaLinter reports
Set VALIDATE_ALL_CODEBASE: true in mega-linter.yml to validate all sources, not only the diff

MegaLinter is graciously provided by

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 99.53%. Comparing base (660ac0d) to head (5543683).
Report is 37 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #106   +/-   ##
=======================================
  Coverage   99.53%   99.53%           
=======================================
  Files          22       23    +1     
  Lines        2341     2342    +1     
=======================================
+ Hits         2330     2331    +1     
  Misses         11       11           

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

[lars-reimann]

Needs automated tests. I'd suggest changing generate_stubs to return "virtual files", consisting of a Path and a string for the content.

Then handle the file I/O in _run_api_command (which should be renamed BTW).

Tests can work with the result of generate_stubs instead of having to read files from disk again.

---

The leading underscore should also be removed from the created file names.

[Masara]

I refactored a bit here and there like you recommended and there is one test left that calls the create_stub_files function (since it also is needed for codecov). Also, I had no idea how to automate the test_file_creation test, since I'd have to read all __init__.py files of the package and compare package paths etc., which, in the end, would become way too complicated. Or is there maybe a better way, which I can't think of?

[lars-reimann]

There's still an extra folder being created:

Might be related to #117.

[Masara]

For me this seems to work correctly, did you delete the old out directory before generating new stubs? Also regarding #117, in my image.sdsstub I have the package info package safeds.data.image.containers, which also seems to be correct...

[lars-reimann]

For me this seems to work correctly, did you delete the old out directory before generating new stubs? Also regarding #117, in my image.sdsstub I have the package info package safeds.data.image.containers, which also seems to be correct...

Yes, and I've checked on two devices, once with Python 3.12 and once with Python 3.11. Using the latest main of the library in both cases.

[lars-reimann]

Using the latest main of the library in both cases.

That might be the issue: The imports in __init__ files are now in TYPE_CHECKING blocks. We've had to change that to fix the horrendous import times of the library. Maybe that requires a change?

Edit: Nevermind, I'm still getting the bugged behavior at Safe-DS/Library@e856cd5.

[Masara]

I did find and fix a small bug which wouldn't let me parse the latest main, but I don't think this bug was related to your issue. Since I can't recreate this I'd say we look at it together next week.

[lars-reimann]

@Masara No change in behavior, unfortunately. There's still an extra folder.

[lars-reimann]

It works if an absolute path is passed as -s argument (#125):

safe-ds-stubgen -o "./out" -nc -s "E:/Repositories/safe-ds/Library/src" --docstyle NUMPYDOC -v

[lars-reimann]

🎉 This PR is included in version 0.3.0 🎉

The release is available on:

• v0.3.0
• GitHub release

Your semantic-release bot 📦🚀