ZeroDivisionError when groupby on a empty group

[AtomBaf]

Code Sample, a copy-pastable example if possible

# Your code here
import numpy as np
import pandas as pd

df = pd.DataFrame({'A':[0,1,0],'B':[1.,np.nan,2.]})
df.groupby('A').B.rank(pct=True)

Problem description

See code example above. The problem occurs only when pct parameter is True

Regression between 0.22 and 0.23.4.
Before 0.23, the result was a series with NaN values for each group containing only NaN values.
Now it is throwing a ZeroDivisionError

Expected Output

Same output as 0.22

Output of pd.show_versions()

INSTALLED VERSIONS ------------------ commit: None python: 3.5.3.final.0 python-bits: 64 OS: Windows OS-release: 7 machine: AMD64 processor: Intel64 Family 6 Model 45 Stepping 7, GenuineIntel byteorder: little LC_ALL: None LANG: None LOCALE: None.None

pandas: 0.23.4
pytest: None
pip: 18.0
setuptools: 40.2.0
Cython: None
numpy: 1.15.0
scipy: None
pyarrow: None
xarray: None
IPython: None
sphinx: None
patsy: None
dateutil: 2.7.3
pytz: 2018.5
blosc: None
bottleneck: None
tables: None
numexpr: None
feather: None
matplotlib: None
openpyxl: None
xlrd: None
xlwt: None
xlsxwriter: None
lxml: None
bs4: None
html5lib: None
sqlalchemy: None
pymysql: None
psycopg2: None
jinja2: None
s3fs: None
fastparquet: None
pandas_gbq: None
pandas_datareader: None

[gfyoung]

Indeed, that does look like a regression.

cc @jreback

[kokes]

git bisect tells us this regression was introduced in c1068d9, which resolves #15779.

[AtomBaf]

Pardon my interruption, but this is blocking for 0.23 adoption on our side : any schedule on the resolution of this regression ?

[gfyoung]

@AtomBaf : Thank you for the ping! I have marked it for our next minor release.

[jreback]

if a patch is contributed it can be put on 0.23.5 but nothing is currently out there