BUG: GroupBy.quantile implicitly sorts index.levels

[Charlie-XIAO]

• Closes BUG: SeriesGroupby with sort=False behaves inconsistently with .quantiles(...).unstack() #53009
• Tests added and passed
• All code checks passed
• Added an entry in the latest doc/source/whatsnew/v2.1.0.rst file

[rhshadrach]

Thanks for the PR! Lookng good - some small requests.

[Charlie-XIAO]

Hi @rhshadrach, thanks for your review! I've updated the changelog and the tests as you suggested. However, I'm not sure if index.levels is being compared. I tried using

qind = pd.MultiIndex(
    [["A", "B"], [0.2, 0.8]],
    [[1, 1, 0, 0], [0, 1, 0, 1]],
    names=["cat1", None],
)

and it still passed the test. Is there any recommended way of fixing this?

[rhshadrach]

Thanks for identifying this @Charlie-XIAO - I opened #53095. For this PR, I think it'd be good to keep the assert_series_equal, and assert the index levels as you had it.

[Charlie-XIAO]

Thanks @rhshadrach! I've retrieved the index levels check.

[Charlie-XIAO]

Thanks for your review @rhshadrach! I've pushed the changes that avoid performance regression when idx is not MultiIndex as you suggested. Please let me know if similar changes should apply to the case when idx is MultiIndex.

[rhshadrach]

Please let me know if similar changes should apply to the case when idx is MultiIndex.

Yes! It looks to me like these two blocks can now be largely combined.

[rhshadrach]

lgtm

[mroeschke]

Thanks @Charlie-XIAO