Python 3.12 compat

[auburus]

Hi,

I took a stab at addressing #430, let me describe a little bit the changes because it ended up being a larger PR than what I intended.

The PR is split in 3 commits:

1. Breaking change that made migrating to python's 3.12 pathlib much easier.
2. Make code work and all unit tests pass under python3.12, but break when <= 3.11
3. Make code work in 3.8 - 3.11, and fix tox/pre-commit when running under python3.12.

Why did I have to make a breaking change

First, I tried just upgrading to python 3.12 and working around the changes done in pathlib.PurePath, but I was having to copy/paste and slightly tweak a lot of methods because of the choice of using the first part of the path as the "root".

An example of what I mean:

# This should be true
ArtifactoryPath('http://b.com/artifactory/repo/path/file.txt').is_relative_to('http://b.com/artifactory') == False

This is because the second path doesn't have a root, but IMO this is a relative path.

So, to fix that I've made the first commit, which is changing in python3.11 and before the way the root is handled. This is how the behavior changes, and that may be a BREAKING change, so I am not sure if that is acceptable or not.

# Current behavior
_ArtifactoryFlavour().splitroot('http://b.com/artifactory/repo/folder/file.txt') == ('http://b.com/artifactory', '/repo/', 'folder/file.txt')

# New behavior
_ArtifactoryFlavour().splitroot('http://b.com/artifactory/repo/folder/file.txt') == ('http://b.com/artifactory', '/', 'repo/folder/file.txt')

I think this may not be so bad, since most of the interface is internal. If a code is using ArtifactoryPath.root this will change, if it is using ArtifactoryPath.repo it is backwards compatible.

Final comments

After that change, the migration was much easier. Commit 2 is making the code work in python3.12, and commit 3 is making it work at the same time in python3.12 AND in older python versions.

There are some small changes in precommit files that I had to do to make sure they would work when tox is invoked under python3.12, but I can drop those if needed, or open a separate PR.

Last thing: All unit tests pass, but I couldn't get a trial license to integration test those changes, so I didn't run that, sorry.

Let me know if I can do anything to help the review!

[allburov]

May be insted of defining many "if" across code, we can do something like inheretance (or even duplicating classes

_ArtifactoryPath(...):
   ... # define <3.11 and comment why it's this way

_ArtifactoryPath311(_ArtifactoryPath):
   ... # define <3.11 and comment why it's this way

BaseClass = _ArtifactoryPath if IS_PYTHON_3_12_OR_NEWER else _ArtifactoryPath311
ArtifactoryPath(BaseClass)

This way when 3.11 ends it's lifecycle (31 Oct 2027) - we can just remove the whole _ArtifactoryPath311 and keep using new version

So the idea is to move all 3.11 related code to a separate class and add new things to 3.12 class (but we can name it as regular class, 312 not required in it's name, because it'll be only class soon)

[auburus]

Hi,

I think I've addressed all changes requested by @allburov (and that simplified the diff quite a bit, thank you for that).

Please let me know if there is anything else I can do, happy to help. Otherwise, just waiting to see if @beliaev-maksim can provide his input in the "breaking changes" conversation.

Thank you for your time!

[allburov]

That's a good work and research, thank you! 🫶
Let's wait for some feedback from the community, and merge if it's good one!

I personally don't use the project anymore so don't have anything to test :(

[allburov]

Let's set a deadline - 21 January 2025. If no critical bugs found before that date - we'll merge #463 and make a release 🚀

[offa]

I can provide some data using a minimal example:

Python	Release	Master¹
3.11	✅	✅
3.12	❌	✅
3.13	❌	❌²

¹ 46728f6 which includes this PR
² Different error

Python 3.12 is fixed 👍. Python 3.13 still fails, but with a different error as reported in #430. I'm going to file a new issue for 3.13 compatibility.

Update: #470

[allburov]

@offa thank you for the feedback!