Stop using distutils

[jwodder]

Distutils was officially deprecated in Python 3.10 and will be removed from the standard library completely in 3.12. This PR thus eliminates almost all of the uses of distutils from the code.

Note that I had to leave in some uses of LooseVersion because I wasn't sure whether the stricter packaging.version.Version would accomodate all of its uses. If we're only using LooseVersion for versions of the form N(.N)*, then Version will work out fine, but anything more exotic is not guaranteed to work. If we need support for more exotic versions, we may have to roll our own version class.

[yarikoptic]

FWIW -- LooseVersion was used in not used in core support/versions.py which I am moving to datalad-crawler.
#6115 datalad/datalad-crawler#111 . So the main "use" of LooseVersion would be datalad/support/external_versions.py for versions comparison etc for external libraries/utilities.

LooseVersion was convenient due to its forgiveness and us rarely need "exact" comparison, and largely to capture and then print those versions (e.g. in datalad wtf). packaging.version.Version I guess would strictly follow Python's versioning scheme, and not even semver and blow up on any odd looking version string. For external python modules we can assume that they follow Python versioning and there this Version should be good. But for external tools, we might want either annotate explicitly either it is following Python versioning or semver or ... and then have adapter classes for those types so we could still store them to be displayed and compare whenever needed?

[mih]

This looks all sane to me. Thanks! I left a few comment which are all relatively minor.

The tests seem to suffer from a str vs bytes problem, seems solvable.

[mih]

Test failure seems legitimately caused by these changes.

[jwodder]

@mih I've pushed a fix.

[codecov]

Codecov Report

Merging #6113 (6234d5b) into maint (ece1cbc) will decrease coverage by 38.88%.
The diff coverage is 37.33%.

❗ Current head 6234d5b differs from pull request most recent head b27bd95. Consider uploading reports for the commit b27bd95 to get more accurate results

@@             Coverage Diff             @@
##            maint    #6113       +/-   ##
===========================================
- Coverage   90.23%   51.34%   -38.89%     
===========================================
  Files         312      312               
  Lines       42219    42242       +23     
===========================================
- Hits        38095    21691    -16404     
- Misses       4124    20551    +16427     

Impacted Files	Coverage Δ
datalad/cmdline/tests/test_main.py	0.00% <0.00%> (-84.84%)	⬇️
datalad/core/distributed/tests/test_clone.py	0.00% <0.00%> (-97.44%)	⬇️
datalad/core/local/create.py	87.50% <ø> (-10.30%)	⬇️
datalad/core/local/tests/test_create.py	0.00% <ø> (-100.00%)	⬇️
datalad/customremotes/tests/test_archives.py	0.00% <0.00%> (-89.41%)	⬇️
datalad/customremotes/tests/test_base.py	0.00% <0.00%> (-97.90%)	⬇️
datalad/customremotes/tests/test_datalad.py	0.00% <0.00%> (-97.88%)	⬇️
datalad/distributed/ora_remote.py	21.23% <0.00%> (-10.05%)	⬇️
datalad/distributed/tests/test_ora_http.py	27.53% <0.00%> (-72.47%)	⬇️
datalad/distribution/create_sibling.py	15.31% <0.00%> (-50.56%)	⬇️
... and 282 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update ece1cbc...b27bd95. Read the comment docs.

[mih]

OK, this seems ready. I have kicked off the errored test again, but I expect no failure. The only thing holding me back from merging is the draft state of this PR. @jwodder is this intentionally not meant to be merged?

[jwodder]

@mih It should be merged at some point. The only thing this PR is currently lacking is an alternative to the use of distutils.version.LooseVersion, though it could be merged as-is right now and a separate issue created for that.

[mih]

@jwodder I would prefer if we get this done here, and leave the LooseVersion issue for later. Right now it seems there is no strict requirement to keep it in -core, but maybe I am missing some case.

[jwodder]

@mih

I would prefer if we get this done here, and leave the LooseVersion issue for later.

Then feel free to merge.

Right now it seems there is no strict requirement to keep it in -core, but maybe I am missing some case.

I don't know what you're referring to here.

[mih]

Thanks!

I was referring to keeping LooseVersion or a replacement in this package (vs an extension, for example).