GroupBy any/all Fails with Duplicate Column Names

[WillAyd]

Code Sample, a copy-pastable example if possible

>>> df = pd.DataFrame([[True, True, True]], columns=['key', 'a', 'a'])
>>> df.groupby('key').any()

ValueError: Buffer has wrong number of dimensions (expected 1, got 2)

Problem description

This throws because of the below line:

pandas/pandas/core/groupby/groupby.py

Line 2024 in 0801b8c

mask = isnull(obj.values).view(np.uint8)

Cython is expecting a one dimensional array of masks, but self._iterate_slices() which contains the above statement will provide a multi-dimensional array. Therefore, when the Cython function gets called with mask it is a multi-dimensional object and causes the ValueError.

I'm not sure if this is the expected behavior of self._iterate_slices() or not. If so, there may be implications in other parts of the module.

Output of pd.show_versions()

INSTALLED VERSIONS

commit: 0801b8c
python: 3.6.4.final.0
python-bits: 64
OS: Darwin
OS-release: 17.6.0
machine: x86_64
processor: i386
byteorder: little
LC_ALL: None
LANG: en_US.UTF-8
LOCALE: en_US.UTF-8

pandas: 0.24.0.dev0+184.g0801b8c90
pytest: 3.4.1
pip: 10.0.1
setuptools: 38.5.1
Cython: 0.27.3
numpy: 1.14.1
scipy: 1.0.0
pyarrow: 0.8.0
xarray: 0.10.0
IPython: 6.2.1
sphinx: 1.7.0
patsy: 0.5.0
dateutil: 2.6.1
pytz: 2018.3
blosc: None
bottleneck: 1.2.1
tables: 3.4.2
numexpr: 2.6.4
feather: 0.4.0
matplotlib: 2.1.2
openpyxl: 2.5.0
xlrd: 1.1.0
xlwt: 1.3.0
xlsxwriter: 1.0.2
lxml: 4.1.1
bs4: 4.6.0
html5lib: 1.0.1
sqlalchemy: 1.2.5
pymysql: 0.8.0
psycopg2: 2.7.4 (dt dec pq3 ext lo64)
jinja2: 2.10
s3fs: 0.1.3
fastparquet: 0.1.4
pandas_gbq: 0.4.1
pandas_datareader: None
gcsfs: None

[WillAyd]

Just for a little more clarity on where I think this originates, here's the GroupBy iteration behavior below:

In [19]: df = pd.DataFrame([[True, True, True]], columns=['key', 'a', 'a'])
In [20]: grp = df.groupby('key')
In [21]: grp_iter = grp._iterate_slices()

In [22]: next(grp_iter)
Out[22]: 
('a',       a     a
 0  True  True)

In [23]: next(grp_iter)
Out[23]: 
('a',       a     a
 0  True  True)

In [24]: next(grp_iter)
StopIteration

The duplicated names causes the slices to be multi-dimensional and repetitive. Not sure if this causes problems elsewhere in the code base but is at least responsible for the issue in the original post

[mroeschke]

I think the correct result is returned on master. Could use a test.

In [206]: >>> df = pd.DataFrame([[True, True, True]], columns=['key', 'a', 'a'])
     ...: >>> df.groupby('key').any()
Out[206]:
         a
key
True  True

In [207]: df
Out[207]:
    key     a     a
0  True  True  True

In [208]: pd.__version__
Out[208]: '0.26.0.dev0+682.g08ab156eb'

[WillAyd]

Hmm well still not technically correct since it is dropping one of the columns, though interesting that this no longer raises. Fix is in #29124 - I'll see what I can find on why this no longer raises though