df.groupby(cols).tshift(0, freq) Drops Index Level When Shift is Zero

[avrixtor]

Code Sample, a copy-pastable example if possible

import pandas as pd

df = pd.DataFrame({"numerical_col": [-1, -2, -3, -4, -5, -6, -7, -8, -9, -10],
                   "date": ["2018-01-01", "2018-01-01", "2018-01-01", "2018-02-01", "2018-02-01",
                               "2018-03-01", "2018-01-01", "2018-01-01", "2018-02-01", "2018-02-01"],
                   "categories": ["a", "a", "a", "a", "a", "a", "b", "b", "b", "b"]})

df["date"] = pd.to_datetime(df["date"])
df.set_index("date").groupby("categories").tshift(0, "M")

Problem description

In short words, tshift after a groupby deletes the groupby columns ("categories" in this case) if the shift used was zero

In [14]: df.set_index("date").groupby("categories").tshift(0, "M")
Out[14]: 
            numerical_col
date                     
2018-01-01             -1
2018-01-01             -2
2018-01-01             -3
2018-02-01             -4
2018-02-01             -5
2018-03-01             -6
2018-01-01             -7
2018-01-01             -8
2018-02-01             -9
2018-02-01            -10

Expected Output

NOTE: This is the output format of any non zero shift (of course with the dates shifted)

In [15]: df.set_index("date").groupby("categories").tshift(0, "M")
Out[15]: 
                       numerical_col
categories date                     
a          2018-01-01             -1
           2018-01-01             -2
           2018-01-01             -3
           2018-02-01             -4
           2018-02-01             -5
           2018-03-01             -6
b          2018-01-01             -7
           2018-01-01             -8
           2018-02-01             -9
           2018-02-01            -10

Output of pd.show_versions()

[paste the output of pd.show_versions() here below this line]
INSTALLED VERSIONS

commit: None
python: 3.6.5.final.0
python-bits: 64
OS: Linux
OS-release: 4.17.3-200.fc28.x86_64
machine: x86_64
processor: x86_64
byteorder: little
LC_ALL: None
LANG: en_US.UTF-8
LOCALE: en_US.UTF-8

pandas: 0.23.3
pytest: None
pip: 9.0.3
setuptools: 39.2.0
Cython: 0.28.4
numpy: 1.14.5
scipy: 1.1.0
pyarrow: None
xarray: None
IPython: 6.4.0
sphinx: None
patsy: None
dateutil: 2.7.3
pytz: 2017.2
blosc: None
bottleneck: None
tables: None
numexpr: None
feather: None
matplotlib: 2.2.2
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

[WillAyd]

Thanks for the report - investigation and PRs are always welcome

[avrixtor]

Now that I think about it... that groupby is not really necessary for this process, I could just shift the date directly. What concerns me is that when tshifting with shift == 0 is not compatible with shift != 0, but maybe that is something intrinsic to tshift

[WillAyd]

See comment in #23918 . The inconsistencies a problem though I think the intended behavior here is the opposite of what is requested, as the request would break alignment with the original caller.

Closing this in favor of the other change