Drop usage of pkg_resources

[edgarrmondragon]

Type of changes

• Bug fix

Checklist

• I've run the latest black with default args on new code.
• I've updated CHANGELOG.md and CONTRIBUTORS.md where appropriate.
• I've added tests for new code.
• I accept that @PyFilesystem/maintainers may be pedantic in the code review.

Description

Drop pkg_resources in favor of the importlib.metadata API.

Closes #356

[edgarrmondragon]

@althonos the tests that fail seem to do so (rather randomly?) due to a DeprecationWarning in the Python 3.10 importlib.metadata API. There's a few options I can think of:

• Kludgy solution that supports both the deprecated and new APIs
• Require importlib_metadata==2.0.0 for all Python versions (even the newer ones) and use the legacy API
• Drop support for Python 2.7 (EOL Jan 2021) and Python 3.5 (EOL Sept 2020) and require importlib_metadata>=3.6.0;<4.9.0 for Python 3.6 and 3.7.

[matrss]

Is there any progress on this? The CI logs are not available anymore

[edgarrmondragon]

@matrss tbh the project feels abandoned 😢

[lurch]

I occasionally review some of the PRs in this project, but I don't know anything about the code being modified here, so I won't be reviewing this one.

Drop support for Python 2.7

Yeah, see also #317

[anthrotype]

tbh the project feels abandoned 😢

oh.. @willmcgugan is it so?

[althonos]

@anthrotype : Well, technically @willmcgugan is not maintaining this anymore (see #422), so I would be the main maintainer, but with my PhD going on I don't have a lot of time to allocate to this, and I've been fine with the "if it's not broken don't fix it" for now. But we definitely need to sit down on a version 3 at some point because these sort of issues are piling up.

[anthrotype]

willmcgugan is not maintaining this anymore

oh, I somehow missed that, thanks for the clarification/reminder.

"if it's not broken don't fix it"

definitely makes sense. Good luck with the PhD!

[AbdealiLoKo]

@althonos would you like some help maintaining it ?
I use pyfilesystem actively and am happy to help maintain.
I was looking to fix the warnings from pyfilesystem related to pkg_resources and came across this PR

[eli-schwartz]

@althonos the tests that fail seem to do so (rather randomly?) due to a DeprecationWarning in the Python 3.10 importlib.metadata API. There's a few options I can think of:

I'm probably missing something since the CI logs are long gone...

Why would this fail at all? A DeprecationWarning should not cause CI failures since the pytest config doesn't appear to be catching it as an error. You just get:

2654 passed, 26 skipped, 1703 warnings in 38.20s

[eli-schwartz]

Either way I think the obviously correct approach is to handle a fallback to the legacy API if needed, since that means you don't have to block this PR on resolving the python2 question.

[anthrotype]

hello! what it would take to get this PR to the finish line?

[edgarrmondragon]

The flaky 3.10 tests pass locally 🤔

[eli-schwartz]

That doesn't look like a flake to me -- it's running on python 3.7, not 3.10. And, as I said above:

Either way I think the obviously correct approach is to handle a fallback to the legacy API if needed, since that means you don't have to block this PR on resolving the python2 question.

But currently you're exclusively using the legacy API, which is broken.

[edgarrmondragon]

I'm closing the PR since I won't be able to finish this any time soon :)

[mxmlnkn]

I've been fine with the "if it's not broken don't fix it" for now.

Unfortunately, with Python 3.12, it is broken now. Installing PyFilesystem2 seems to still work, but trying to import it, I get:

Traceback (most recent call last):
  [...]
  File "squashfs-root/opt/python3.12/lib/python3.12/site-packages/fs/__init__.py", line 4, in <module>
    __import__("pkg_resources").declare_namespace(__name__)  # type: ignore
    ^^^^^^^^^^^^^^^^^^^^^^^^^^^
ModuleNotFoundError: No module named 'pkg_resources'

[eli-schwartz]

Submitted an alternative PR at #589 / #590