BUG: Rolling min_periods not working on groupby object

[justinessert]

• 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

df = pd.DataFrame({
    'segment': 'A',
    'data': range(10)
})

df.rolling(5, center=True, min_periods=1).max()

df.groupby('segment').rolling(5, center=True, min_periods=1).max().reset_index(drop=True)

Problem description

For the DataFrame above, with a single segment 'A', the result of df.rolling(5, center=True, min_periods=1).max() should be identical to that of df.groupby('segment').rolling(5, center=True, min_periods=1).max().reset_index(drop=True). Instead, the latter operation has NaNs in the last two positions of the data column.

Expected Output

Both operations should return the sequence [2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 9.0, 9.0]. Instead, df.groupby('segment').rolling(5, center=True, min_periods=1).max().reset_index(drop=True) returns [2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, NaN, NaN]

Output of pd.show_versions()

INSTALLED VERSIONS

commit : f2ca0a2
python : 3.7.7.final.0
python-bits : 64
OS : Darwin
OS-release : 19.6.0
Version : Darwin Kernel Version 19.6.0: Thu Jun 18 20:49:00 PDT 2020; root:xnu-6153.141.1~1/RELEASE_X86_64
machine : x86_64
processor : i386
byteorder : little
LC_ALL : None
LANG : en_US.UTF-8
LOCALE : en_US.UTF-8

pandas : 1.1.1
numpy : 1.19.1
pytz : 2020.1
dateutil : 2.8.0
pip : 20.1.1
setuptools : 47.3.0.post20200616
Cython : None
pytest : 6.0.0
hypothesis : None
sphinx : 3.1.1
blosc : None
feather : None
xlsxwriter : 1.2.9
lxml.etree : 4.5.1
html5lib : None
pymysql : None
psycopg2 : None
jinja2 : 2.11.2
IPython : 7.15.0
pandas_datareader: None
bs4 : None
bottleneck : None
fsspec : 0.8.0
fastparquet : None
gcsfs : None
matplotlib : 3.2.1
numexpr : None
odfpy : None
openpyxl : None
pandas_gbq : None
pyarrow : 1.0.0
pytables : None
pyxlsb : None
s3fs : 0.4.2
scipy : 1.5.2
sqlalchemy : None
tables : None
tabulate : None
xarray : None
xlrd : 1.2.0
xlwt : None
numba : None

[justinessert]

[Edited]

Additionally, I have found that if there are two segments in the DataFrame, the groupby is not respected, but the NaNs only come in on the last segment.

df = pd.DataFrame({
    'segment': ['A']*10 + ['B']*10,
    'data': range(20)
})
df.groupby('segment').rolling(5, center=True, min_periods=1).max()

Here, the expected result of df.groupby('segment').rolling(5, center=True, min_periods=1).max() is:
for segment 'A' is [2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 9.0, 9.0] but the actual result is [2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0, 11.0]
for segment 'B' is [12.0, 13.0, 14.0, 15.0, 16.0, 17.0, 18.0, 19.0, 19.0, 19.0] but the actual result is [12.0, 13.0, 14.0, 15.0, 16.0, 17.0, 18.0, 19.0, NaN, NaN]

[jreback]

cc @mroeschke

[wfvining]

Seeing what I think is the same problem with the following example.

x = pd.Series([1, 2, 3, 4, 5, 6, 7, 8])
x.groupby(x % 2).rolling(window=3, min_periods=1, center=True).sum()

I expect to see

0  1     6.0                                                                                                               
   3    12.0
   5    18.0 
   7    14.0
1  0     4.0
   2     9.0
   4    15.0
   6    12.0
dtype: float64 

But instead I get

0  1     6.0                                                                                                               
   3    12.0
   5    18.0 
   7     1.0
1  0     4.0
   2     9.0
   4    15.0
   6     NaN
dtype: float64 

If center or min_periods are not specified then I get the expected behavior.

[justinessert]

Issue resolved with PR 36567

[mroeschke]

We'll officially close this issue with your PR in #37035