REGR: Series access with Index of tuples/frozenset

[rhshadrach]

• closes BUG: Cannot select values in series using tuples when indexing with tuples in pandas 1.1.0 #35534
• closes BUG: Cannot index frozenset elements from a pd.Series or pd.DataFrame #35747
• tests added / passed
• passes black pandas
• passes git diff upstream/master -u -- "*.py" | flake8 --diff
• whatsnew entry

[jreback]

instead of doing the if first, can you do it in the keyerror itself e.g.

try:
    result = self._get_value(key)
    return result
except KeyError:
    if isinstance(self.index, MultiIndex):
        return self._get_values_tuple(key)

of course maybe we should actually do this inside _get_value

[rhshadrach]

Since _get_value is called from 4 places, I'll look into doing this for 1.2 in a followup.

[rhshadrach]

Thanks @jreback, I reworked the logic. This patch now retains much more of the original behavior.

When one tries to find a tuple key in a Series with an Index, originally you got the message Can only tuple-index with a MultiIndex. As this is no longer true, I've changed this to key of type tuple not found and not a MultiIndex.

[jreback]

@simonjayhawkins this is ok to merge on green if you haven't tagged (otherwise can move to 1.1.3)

[simonjayhawkins]

@simonjayhawkins this is ok to merge on green if you haven't tagged (otherwise can move to 1.1.3)

yep

[simonjayhawkins]

restarting ci. known failure

=========================== short test summary info ===========================
FAILED pandas/tests/plotting/test_groupby.py::TestDataFrameGroupByPlots::test_series_groupby_plotting_nominally_works
= 1 failed, 74726 passed, 1786 skipped, 1130 xfailed, 6 xpassed, 67 warnings in 1183.92s (0:19:43) =

[simonjayhawkins]

same test failed. may need further investigation

[jreback]

ok let's defer this one

[jbrockmendel]

LGTM whatsnew-placement notwithstanding

[rhshadrach]

Failure is different this time.

=========================== short test summary info ===========================
FAILED pandas/tests/io/test_parquet.py::TestParquetPyArrow::test_s3_roundtrip_for_dir[partition_col0]
= 1 failed, 74674 passed, 1905 skipped, 1108 xfailed, 6 xpassed, 284 warnings in 1178.42s (0:19:38) =

[simonjayhawkins]

Failure is different this time.

I think this one is affecting all PRs

[rhshadrach]

@jreback Same unrelated parquet failure; the groupby plots failure no longer occurs.

[rhshadrach]

Thanks for the review @jbrockmendel, changes made and checks pass.

[jreback]

thanks @rhshadrach for the followon's can open an issue if you want so we can track.

[simonjayhawkins]

@meeseeksdev backport 1.1.x