BUG: groupby().groups and .indices always treat as_index as True

[mvashishtha]

Pandas version checks

• I have checked that this issue has not already been reported.

• I have confirmed this bug exists on the latest version of pandas.

• I have confirmed this bug exists on the main branch of pandas.

Reproducible Example

import pandas as pd

df = pd.DataFrame({'col0': [3], 'col1': [4]})

print(df.groupby('col0', as_index=False).groups)
print(df.groupby('col0', as_index=True).groups)
print(df.groupby('col0', as_index=False).indices)
print(df.groupby('col0', as_index=True).indices)

Issue Description

When as_index=False, the keys of groups should be a sequence of consecutive numbers starting from 0. Instead, the keys of the group dictionary are the group keys, as if as_index were True. Same with .indices.

Expected Behavior

When as_index=False, the keys of groups and indices should be a sequence of consecutive numbers starting from 0.

Installed Versions

INSTALLED VERSIONS

commit : a671b5a
python : 3.9.18.final.0
python-bits : 64
OS : Darwin
OS-release : 23.2.0
Version : Darwin Kernel Version 23.2.0: Wed Nov 15 21:55:06 PST 2023; root:xnu-10002.61.3~2/RELEASE_ARM64_T6020
machine : arm64
processor : arm
byteorder : little
LC_ALL : None
LANG : en_US.UTF-8
LOCALE : en_US.UTF-8

pandas : 2.1.4
numpy : 1.26.3
pytz : 2023.3.post1
dateutil : 2.8.2
setuptools : 68.2.2
pip : 23.3.1
Cython : None
pytest : None
hypothesis : None
sphinx : None
blosc : None
feather : None
xlsxwriter : None
lxml.etree : None
html5lib : None
pymysql : None
psycopg2 : None
jinja2 : None
IPython : 8.18.1
pandas_datareader : None
bs4 : None
bottleneck : None
dataframe-api-compat: None
fastparquet : None
fsspec : None
gcsfs : None
matplotlib : None
numba : None
numexpr : None
odfpy : None
openpyxl : None
pandas_gbq : None
pyarrow : None
pyreadstat : None
pyxlsb : None
s3fs : None
scipy : None
sqlalchemy : None
tables : None
tabulate : None
xarray : None
xlrd : None
zstandard : None
tzdata : 2023.4
qtpy : None
pyqt5 : None

[rhshadrach]

Thanks for the report!

When as_index=False, the keys of groups should be a sequence of consecutive numbers starting from 0. Instead, the keys of the group dictionary are the group keys, as if as_index were True. Same with .indices.

I'm not sure I agree. I don't think as_index being True or False changes the groups, but rather just where the groups are placed (index vs column). For example,

print(df.groupby('col0', as_index=False).sum())
   col0  col1
0     3     4

I would still say the group here is 3.

[mvashishtha]

I can see where there is room for debate here. The documented behavior is Dict {group name -> group labels}. If someone does groupby(as_index=False), are the "group names" the new row labels of the results, or the row labels of the dataframe that would result from an aggregation like groupby(as_index=False).sum()? I would say the latter. In the aggregation result, I think the "group names" are the group labels, not the values of the by columns. Note that this principle would hold even in aggregations by expressions that are not part of the original dataframe's columns e.g.

df = pd.DataFrame({'col0': [3, 5], 'col1': [4, 6]})
print(df.groupby(df.col0 % 2, as_index=False).sum())
   col0  col1
0     8    10

In this case with as_index=False, the group key is not part of the original columns, so it's not in the columns or the index.as_index=False effectively eliminates the group key from the result. But df.groupby(df.col0 % 2, as_index=False).groups would keep the group key 1 in the result.

I think it's clearer if the groups that the user sees in groupby().groups always match the row labels of any groupby aggregation. In general, I think all parameters that can affect the sequence of the resulting row labels for aggregations like groupby().sum() should have the same effect on the sequence of keys in groupby().groups. So I stick to my original position here.

[rhshadrach]

are the "group names" the new row labels of the results, or the row labels of the dataframe that would result from an aggregation like groupby(as_index=False).sum()?

I think neither. The DataFrame.groupby documents the by argument as

Used to determine the groups for the groupby. [snip] If a list or ndarray of length equal to the selected axis is passed (see the groupby user guide), the values are used as-is to determine the groups.

These are the groups.

In this case with as_index=False, the group key is not part of the original columns, so it's not in the columns or the index.as_index=False effectively eliminates the group key from the result.

This behavior is deprecated and the groups will be in the columns in pandas 3.0. #49519

[rhshadrach]

Thinking about this a bit more, groups, indices, get_groups, and __iter__ all consistently treat the groups the same way. Changing any one of these I think would mean changing them all, and that's a lot of churn for what I think is little, if any, gain.