BUG: not possible to 'cumsum' Timedelta with named aggregation

[yohplala]

• 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
import numpy as np
import random

# Dataset
ts = pd.DatetimeIndex([pd.Timestamp('2021/01/01 00:30'),
                       pd.Timestamp('2021/01/01 00:45'),
                       pd.Timestamp('2021/01/01 02:00'),
                       pd.Timestamp('2021/01/01 03:50'),
                       pd.Timestamp('2021/01/01 05:00')])
length = len(ts)
random.seed(1)
value = random.sample(range(1, length+1), length)
df = pd.DataFrame({'value': value, 'ts': ts})
df['td'] = df['ts'] - df['ts'].shift(1, fill_value=ts[0]-pd.Timedelta('1h'))
df['amount'] = df['value']*10
df.loc[:2,'grps'] = 'a'
df.loc[2:,'grps'] = 'b'

In [16]: df
Out[16]: 
   value                  ts              td  amount grps
0      2 2021-01-01 00:30:00 0 days 01:00:00      20    a
1      1 2021-01-01 00:45:00 0 days 00:15:00      10    a
2      5 2021-01-01 02:00:00 0 days 01:15:00      50    b
3      4 2021-01-01 03:50:00 0 days 01:50:00      40    b
4      3 2021-01-01 05:00:00 0 days 01:10:00      30    b

# cumsum and Timedelta work when used through pd.Series
df['td'].cumsum()

# cumsum and Timedelta do not work when used in groupby with named aggregation
agg_rules = {'td':('td', 'cumsum')}
res = df.groupby('grps').agg(**agg_rules)

Error message that is returned

res = df.groupby('grps').agg(**agg_rules)
Traceback (most recent call last):

  File "<ipython-input-14-c0f4606d675a>", line 2, in <module>
    res = df.groupby('grps').agg(**agg_rules)

  File "/home/pierre/anaconda3/lib/python3.8/site-packages/pandas/core/groupby/generic.py", line 945, in aggregate
    result, how = aggregate(self, func, *args, **kwargs)

  File "/home/pierre/anaconda3/lib/python3.8/site-packages/pandas/core/aggregation.py", line 582, in aggregate
    return agg_dict_like(obj, arg, _axis), True

  File "/home/pierre/anaconda3/lib/python3.8/site-packages/pandas/core/aggregation.py", line 768, in agg_dict_like
    results = {key: obj._gotitem(key, ndim=1).agg(how) for key, how in arg.items()}

  File "/home/pierre/anaconda3/lib/python3.8/site-packages/pandas/core/aggregation.py", line 768, in <dictcomp>
    results = {key: obj._gotitem(key, ndim=1).agg(how) for key, how in arg.items()}

  File "/home/pierre/anaconda3/lib/python3.8/site-packages/pandas/core/groupby/generic.py", line 247, in aggregate
    ret = self._aggregate_multiple_funcs(func)

  File "/home/pierre/anaconda3/lib/python3.8/site-packages/pandas/core/groupby/generic.py", line 315, in _aggregate_multiple_funcs
    results[base.OutputKey(label=name, position=idx)] = obj.aggregate(func)

  File "/home/pierre/anaconda3/lib/python3.8/site-packages/pandas/core/groupby/generic.py", line 241, in aggregate
    return getattr(self, func)(*args, **kwargs)

  File "/home/pierre/anaconda3/lib/python3.8/site-packages/pandas/core/groupby/groupby.py", line 2497, in cumsum
    return self._cython_transform("cumsum", **kwargs)

  File "/home/pierre/anaconda3/lib/python3.8/site-packages/pandas/core/groupby/groupby.py", line 996, in _cython_transform
    raise DataError("No numeric types to aggregate")

DataError: No numeric types to aggregate

Problem description

Cumsum works with Timedelta in some workflows, but for some other workflows (groupby, named aggregation) it does not.

Output of pd.show_versions()

INSTALLED VERSIONS

commit : 2cb9652
python : 3.8.5.final.0
python-bits : 64
OS : Linux
OS-release : 5.8.0-53-generic
Version : #60~20.04.1-Ubuntu SMP Thu May 6 09:52:46 UTC 2021
machine : x86_64
processor : x86_64
byteorder : little
LC_ALL : None
LANG : fr_FR.UTF-8
LOCALE : fr_FR.UTF-8

pandas : 1.2.4
numpy : 1.20.1
pytz : 2021.1
dateutil : 2.8.1
pip : 21.1.1
setuptools : 52.0.0.post20210125
Cython : 0.29.23
pytest : 6.2.3
hypothesis : None
sphinx : 4.0.1
blosc : None
feather : None
xlsxwriter : 1.3.8
lxml.etree : 4.6.3
html5lib : 1.1
pymysql : None
psycopg2 : None
jinja2 : 3.0.0
IPython : 7.22.0
pandas_datareader: None
bs4 : 4.9.3
bottleneck : 1.3.2
fsspec : 0.9.0
fastparquet : 0.6.3
gcsfs : None
matplotlib : 3.3.4
numexpr : 2.7.3
odfpy : None
openpyxl : 3.0.7
pandas_gbq : None
pyarrow : 2.0.0
pyxlsb : None
s3fs : None
scipy : 1.6.2
sqlalchemy : 1.4.15
tables : 3.6.1
tabulate : None
xarray : None
xlrd : 2.0.1
xlwt : 1.3.0
numba : 0.51.2

[phofl]

The error is raised, because cumsum in groupby sets numeric_only=True as default. I think the correct way to call this is:

agg_rules = {'td': 'cumsum'}
res = df.groupby('grps').agg(agg_rules, **{"numeric_only": False})

but the error persists, because the kwargs are not passed through.

[WillAyd]

Looks like this is fixed on main. Could use a test

[seanjedi]

I can take this and work on making a test

[seanjedi]

take

[seanjedi]

@WillAyd Are the results of the two examples above supposed to be the same?
I am trying to create a test utilizing some of the same logic as the sample above, however when I check the results, the results are different
Is this a bug?

[WillAyd]

No they aren't the same - the first is just a normal cumulative whereas the latter is a groupby. So should respect the groupings

[seanjedi]

@WillAyd Added a new test here: #50033
Im sure there will be a lot of things I can improve, as this is my first test with Pandas 😅