DataFrame Quantile Broken with Datetime Data

[WillAyd]

This works fine:

In [17]: pd.Series(pd.date_range('1/1/18', periods=5)).quantile()                                                                          
Out[17]: Timestamp('2018-01-03 00:00:00')

But the equivalent method with a DataFrame raises:

In [18]: pd.DataFrame(pd.date_range('1/1/18', periods=5)).quantile()                                                                       
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-18-68ffc067f6f0> in <module>
----> 1 pd.DataFrame(pd.date_range('1/1/18', periods=5)).quantile()

~/clones/pandas/pandas/core/frame.py in quantile(self, q, axis, numeric_only, interpolation)
   7569                                      axis=1,
   7570                                      interpolation=interpolation,
-> 7571                                      transposed=is_transposed)
   7572 
   7573         if result.ndim == 2:

~/clones/pandas/pandas/core/internals/managers.py in quantile(self, **kwargs)
    500 
    501     def quantile(self, **kwargs):
--> 502         return self.reduction('quantile', **kwargs)
    503 
    504     def setitem(self, **kwargs):

~/clones/pandas/pandas/core/internals/managers.py in reduction(self, f, axis, consolidate, transposed, **kwargs)
    473 
    474         # single block
--> 475         values = _concat._concat_compat([b.values for b in blocks])
    476 
    477         # compute the orderings of our original data

~/clones/pandas/pandas/core/dtypes/concat.py in _concat_compat(to_concat, axis)
    172                 to_concat = [x.astype('object') for x in to_concat]
    173 
--> 174     return np.concatenate(to_concat, axis=axis)
    175 
    176 

ValueError: need at least one array to concatenate

Expected Output

The DataFrame should return just as Series does

Output of pd.show_versions()

INSTALLED VERSIONS

commit: 07b0c00
python: 3.6.7.final.0
python-bits: 64
OS: Darwin
OS-release: 18.2.0
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+1150.g07b0c003d.dirty
pytest: 4.0.0
pip: 18.1
setuptools: 40.6.2
Cython: 0.29
numpy: 1.15.4
scipy: 1.1.0
pyarrow: 0.11.1
xarray: 0.11.0
IPython: 7.1.1
sphinx: 1.8.2
patsy: 0.5.1
dateutil: 2.7.5
pytz: 2018.7
blosc: None
bottleneck: 1.2.1
tables: 3.4.4
numexpr: 2.6.8
feather: None
matplotlib: 3.0.1
openpyxl: 2.5.9
xlrd: 1.1.0
xlwt: 1.2.0
xlsxwriter: 1.1.2
lxml: 4.2.5
bs4: 4.6.3
html5lib: 1.0.1
sqlalchemy: 1.2.14
pymysql: 0.9.2
psycopg2: None
jinja2: 2.10
s3fs: 0.1.6
fastparquet: 0.1.6
pandas_gbq: None
pandas_datareader: None
gcsfs: 0.2.0

[qwhelan]

It appears that DataFrame.quantile() has a default of numeric_only=True, which Series.quantile() does not have:

In [5]: pd.DataFrame(pd.date_range('1/1/18', periods=5)).quantile(numeric_only=False)
Out[5]: 0   2018-01-03
        Name: 0.5, dtype: datetime64[ns]

Looks like this is a duplicate of #18608 - I'd be in favor of resolving by deprecating this flag.