Pandas 0.18.1 df.groupby().count() throws "ValueError: Buffer has wrong number of dimensions" when one of the counted columns has dtype datetime64

[jfries]

I've confirmed the error does not occur in Pandas 0.16 on a similar machine.

Code Sample, a copy-pastable example if possible

import pandas as pd
df = pd.DataFrame({'x': ['a', 'a', 'b'],
                   'y': [pd.Timestamp('2016-05-07 20:09:25+00:00'), pd.Timestamp('2016-05-07 20:09:29+00:00'), pd.Timestamp('2016-05-07 20:09:29+00:00')]})
df.groupby('x').count()

Observed output

---

ValueError                                Traceback (most recent call last)
<ipython-input-5-3119045de5b1> in <module>()
      2 df = pd.DataFrame({'x': ['a', 'a', 'b'],
      3                    'y': [pd.Timestamp('2016-05-07 20:09:25+00:00'), pd.Timestamp('2016-05-07 20:09:29+00:00'), pd.Timestamp('2016-05-07 20:09:29+00:00')]})
----> 4 print df.groupby('x').count()

/usr/local/lib/python2.7/dist-packages/pandas/core/groupby.pyc in count(self)
   3754         blk = map(make_block, map(counter, val), loc)
   3755 
-> 3756         return self._wrap_agged_blocks(data.items, list(blk))
   3757 
   3758 

pandas/lib.pyx in pandas.lib.count_level_2d (pandas/lib.c:23068)()

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

Expected Output

x  y
a  2
b  1

output of pd.show_versions()

INSTALLED VERSIONS
------------------
commit: None
python: 2.7.6.final.0
python-bits: 64
OS: Linux
OS-release: 3.13.0-32-generic
machine: x86_64
processor: x86_64
byteorder: little
LC_ALL: None
LANG: en_US.UTF-8

pandas: 0.18.1
nose: 1.3.1
pip: 8.1.2
setuptools: 21.2.1
Cython: None
numpy: 1.11.0
scipy: 0.16.0
statsmodels: 0.5.0
xarray: None
IPython: 4.2.0
sphinx: None
patsy: 0.2.1
dateutil: 2.5.3
pytz: 2016.4
blosc: None
bottleneck: 0.8.0
tables: 3.1.1
numexpr: 2.2.2
matplotlib: 1.5.1
openpyxl: 1.7.0
xlrd: 0.9.2
xlwt: 0.7.5
xlsxwriter: 0.5.8
lxml: None
bs4: 4.2.1
html5lib: 0.999
httplib2: None
apiclient: None
sqlalchemy: 0.9.7
pymysql: None
psycopg2: 2.5.3 (dt dec pq3 ext)
jinja2: 2.8
boto: 2.36.0
pandas_datareader: None

[jreback]

Yeah this is an issue, I thought it was a duplicate, but can't seem to find an open one.

its a bit to fix, but pull-requests welcome.

You can do this.

In [4]: df.groupby('x').y.count()
Out[4]: 
x
a    2
b    1
Name: y, dtype: int64

[jreback]

as an FYI, pls paste code as markdown. I edited the above.

[jfries]

sorry about the lack of markdown, will use going forward.
Agreed that your workaround is correct, but I'm pretty sure that a lot of existing pandas code uses the idiom in my example, as it's common in a lot of tutorials & example code.

[jreback]

@jfries sure, and that's why its marked as a bug.

[alan-wong]

This seems related to why the following no longer works in 0.18.1:

df = pd.DataFrame({'a':list('abssbab')}) df.groupby('a').count()

this also used to work

[jorisvandenbossche]

@alan-wong I don't think that is related. In your example, there are no columns left to count the values, so I think it is correct that it does not work. But you are right the behaviour did change. Previously it returned an empty dataframe, now it does give an error.
And the empty dataframe seems more correct, @alan-wong do you want to open an new issue about that?

[alan-wong]

@jorisvandenbossche I get an empty dataframe on version 0.18.1 not sure what version I was running when I posted this answer: http://stackoverflow.com/questions/22391433/count-the-frequency-that-a-value-occurs-in-a-dataframe-column/22391554#22391554 but it used to work it does raise an error if you pass as_index=False arg to the groupby call

[jorisvandenbossche]

@alan-wong Ah, indeed, on 0.18.1 it does work correctly, but on master not anymore. So that indeed looks like a bug. Do you want to open a new issue?

[alan-wong]

@jorisvandenbossche but is it correct behaviour that it's empty? Are you saying that in older versions it shouldn't have worked? I'll post an issue