SettingWithCopyWarning while iterating on groupby results

[has2k1]

Code Sample

import pandas as pd

df = pd.DataFrame({
    'x': range(4),
    'c': list('aabb')
})

# These operations should have the same consequences.

# 1. No warning
gdfs = [gdf for _, gdf in df.groupby('c')]
for gdf in gdfs:
    gdf['x'] = 1

# 2. SettingWithCopyWarning
gdfs = []
for _, gdf in df.groupby('c'):
    gdf['x'] = 1
    gdfs.append(gdf)

Problem description

Modifying a group dataframe while iterating through the groups triggers a SettingWithCopyWarning.
The dataframe referred to is a result of an internal slicing operation and should not lead to a warning in userspace.

Output of pd.show_versions()

INSTALLED VERSIONS

commit: None
python: 3.6.1.final.0
python-bits: 64
OS: Linux
OS-release: 4.11.1-gentoo
machine: x86_64
processor: Intel
byteorder: little
LC_ALL: en_US.utf8
LANG: en_US.utf8
LOCALE: en_US.UTF-8

pandas: 0.23.0.dev0+81.g78147e9c8.dirty
pytest: 3.3.1
pip: 9.0.1
setuptools: 38.4.0
Cython: 0.27.3
numpy: 1.14.0
scipy: None
pyarrow: None
xarray: None
IPython: 6.2.1
sphinx: None
patsy: None
dateutil: 2.6.1
pytz: 2017.3
blosc: None
bottleneck: None
tables: None
numexpr: 2.6.4
feather: None
matplotlib: None
openpyxl: None
xlrd: None
xlwt: None
xlsxwriter: None
lxml: None
bs4: None
html5lib: 1.0.1
sqlalchemy: None
pymysql: None
psycopg2: None
jinja2: 2.10
s3fs: None
fastparquet: None
pandas_gbq: None
pandas_datareader: None

[jreback]

slightly related to #5758

why do you think these should be able to mutate this object? this is a slice, sure its an implementation detail, but in general you would need to copy this. Mutation like this is plain non-idiomatic as well.

[has2k1]

why do you think these should be able to mutate this object?

Mutation like this is plain non-idiomatic as well.

Okay, here is why.

I have custom split-apply-combine code, some of which nests multiple times. groupby().apply() is/was not an option (functions may have consequences or
to just prevent an explosion of speculative function calls on the first group due to nesting). Where possible I use generators and may reuse the group dataframes. For such a case it is more performant to avoid copying the returned group dataframes.

So other than the oddity about implementation details leaking out, it is nicer to do groupbys if the data is handed back with no strings attached.

[jreback]

@has2k1 you are using pandas non idiomatically
and on your own