BUG: Series.groupby fails with InvalidIndexError on time series with a tuple-named grouper.

[ghost]

• I have checked that this issue has not already been reported.

• I have confirmed this bug exists on the latest version of pandas.

• (optional) I have confirmed this bug exists on the master branch of pandas.

---

Code Sample, a copy-pastable example

import pandas as pd

s = pd.Series(index=[pd.Timestamp(2021,7,26)], name=('A', 1))
s.groupby(s==s)  # raises InvalidIndexError

Problem description

There are several things necessary to trigger this:

• The index has to consist of pd.Timestamps. If you change the example to s = pd.Series(index=[0,1], name=('A', 1)) everything is fine.
• The grouper, i.e. groupby's by argument , has to be a Series named with a tuple (like you get if you select a MultiIndex-column from a DataFrame). If you change the example to s.groupby((s==s).rename('foo')) everything is fine. Same goes if you use a pure list or a function.
• The object to group must be a Series. If you change the example to pd.DataFrame(s).groupby(s==s) everything is fine.
• You need pandas>=1.1.3. If you downgrade to an older version - guess what - everything is fine.

Why this is a problem: The name of a grouper should not matter and it is just not to understand why a time series behaves differently than any other kind of Series.

Expected Output

Behavior should be as for non-time-series.

Output of pd.show_versions()

INSTALLED VERSIONS

commit : db08276
python : 3.8.3.final.0
python-bits : 64
OS : Windows
OS-release : 10
Version : 10.0.19041
machine : AMD64
processor : Intel64 Family 6 Model 158 Stepping 13, GenuineIntel
byteorder : little
LC_ALL : None
LANG : None
LOCALE : de_DE.cp1252
pandas : 1.1.3
numpy : 1.20.3
pytz : 2021.1
dateutil : 2.8.2
pip : 21.1.3
setuptools : 52.0.0.post20210125
Cython : None
pytest : 6.2.4
hypothesis : None
sphinx : None
blosc : None
feather : None
xlsxwriter : None
lxml.etree : None
html5lib : None
pymysql : None
psycopg2 : None
jinja2 : None
IPython : None
pandas_datareader: None
bs4 : None
bottleneck : 1.3.2
fsspec : None
fastparquet : None
gcsfs : None
matplotlib : None
numexpr : 2.7.3
odfpy : None
openpyxl : None
pandas_gbq : None
pyarrow : None
pytables : None
pyxlsb : None
s3fs : None
scipy : None
sqlalchemy : None
tables : None
tabulate : None
xarray : None
xlrd : None
xlwt : None
numba : None

[ghost]

There are simple solutions like to catch InvalidIndexError in
pandas.core.groupby.grouper.get_grouper().is_in_obj() or
pandas.core.series.Series.__getitem__() (in the line if is_hashable())
which restores the pre-1.1.3 behavior. But to me this seems to just hide the problem. Why is the name of the grouper(!) of interest at all? I somehow question the implementation of def is_in_obj(gpr).

[simonjayhawkins]

Thanks @suoniq for the report.

• You need pandas>=1.1.3. If you downgrade to an older version - guess what - everything is fine.

first bad commit: [2000334] Backport PR #36147: REGR: Series access with Index of tuples/frozenset (#36332) cc @rhshadrach

similar issue #37755

[rhshadrach]

Why is the name of the grouper(!) of interest at all?

This is to determine whether the grouper is of the form e.g. df.groupby('a') where a is a column of the DataFrame.

There are simple solutions like to catch InvalidIndexError

This seems like the right approach to me - the method is attempting to access the grouper name from the frame, and on any failure return False. It even seems to me this would be an appropriate place to just catch Exception. Would you be interested in putting up a PR to fix?

[rhshadrach]

@simonjayhawkins - I think this should be on the 1.3.2 milestone, is that right?

[ghost]

Would you be interested in putting up a PR to fix?

Sure, count me in. Never contributed here before. So I have to dig into your tests first. But I'll go for it.

[simonjayhawkins]

@simonjayhawkins - I think this should be on the 1.3.2 milestone, is that right?

I don't tend to put older regressions on the current backport milestone as it's more of a tracker than a project task list. If we get a PR to fix, can then assess whether we can/ want to backport. Regressions and bug fixes are allowable in a patch release.