BUG: pandas-1.4.0 .loc "shallow" copies do not update

[slackline]

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
from numpy.random import default_rng


rng = default_rng(20220125)
N = 10
original = pd.DataFrame({'a': rng.standard_normal(N),
                         'b': rng.integers(low=0, high=100, size=N),
                         })

print(f'Pandas Version   : {pd.__version__}')
print(f'===========================')
print(f'Original :\n{original}\n')

copy_plain = original.loc[1:2]
copy_shallow = original.loc[1:2].copy(deep=False)
copy_deep = original.loc[1:2].copy(deep=True)

def print_df(df, description):
    """Check the data frame."""
    print(f'===========================')
    print(f'{description}    : \n{df}')


print_df(original.loc[1:2], 'Original slice')
print_df(copy_plain, 'Plain')
print_df(copy_shallow, 'Shallow')
print_df(copy_deep, 'Deep')

print(f'Make a change to original and see if it cascades through')
print(f'original["a"] = original["a"]**2')
original['a'] = original['a']**2

print_df(original.loc[1:2], 'Original slice')
print_df(copy_plain, 'Plain')
print_df(copy_shallow, 'Shallow')
print_df(copy_deep, 'Deep')

Issue Description

When taking a "shallow" slice of a dataframe with .loc subsequent updates should be reflected in the original and vice-versa, this is no longer the case under Pandas 1.4.0 as demonstrated running the reproducible example...

Pandas Version   : 1.4.0
===========================
Original :
          a   b
0  0.778724  58
1  1.529873  40
2 -0.328487  63
3 -0.090816  66
4  0.451966  82
5  1.648603  65
6  1.161088  92
7  0.058186  48
8  0.800209  81
9  1.783765   4

===========================
Original slice    : 
          a   b
1  1.529873  40
2 -0.328487  63
===========================
Plain    : 
          a   b
1  1.529873  40
2 -0.328487  63
===========================
Shallow    : 
          a   b
1  1.529873  40
2 -0.328487  63
===========================
Deep    : 
          a   b
1  1.529873  40
2 -0.328487  63
Make a change to original and see if it cascades through
original["a"] = original["a"]**2
===========================
Original slice    : 
          a   b
1  2.340512  40
2  0.107903  63
===========================
Plain    : 
          a   b
1  1.529873  40
2 -0.328487  63
===========================
Shallow    : 
          a   b
1  1.529873  40
2 -0.328487  63
===========================
Deep    : 
          a   b
1  1.529873  40
2 -0.328487  63

After original["a"] = original["a"]**2 the values at .loc[1:2,"a"]are2.340512and0.107903 and should be the same under a "Plain" copy and "Shallow`" copy as the documentation states only a "deep" copy should take a snapshot of the original and shallow copies share data and index with original..

Expected Behavior

The behaviour under 1.3.5 is as described in the manual, after modifying the original a "plain" and "shallow" copy reflect the modifications...

Pandas Version   : 1.3.5
===========================
Original :
          a   b
0  0.778724  58
1  1.529873  40
2 -0.328487  63
3 -0.090816  66
4  0.451966  82
5  1.648603  65
6  1.161088  92
7  0.058186  48
8  0.800209  81
9  1.783765   4

===========================
Original slice    : 
          a   b
1  1.529873  40
2 -0.328487  63
===========================
Plain    : 
          a   b
1  1.529873  40
2 -0.328487  63
===========================
Shallow    : 
          a   b
1  1.529873  40
2 -0.328487  63
===========================
Deep    : 
          a   b
1  1.529873  40
2 -0.328487  63
Make a change to original and see if it cascades through
original["a"] = original["a"]**2
===========================
Original slice    : 
          a   b
1  2.340512  40
2  0.107903  63
===========================
Plain    : 
          a   b
1  2.340512  40
2  0.107903  63
===========================
Shallow    : 
          a   b
1  2.340512  40
2  0.107903  63
===========================
Deep    : 
          a   b
1  1.529873  40
2 -0.328487  63

Installed Versions

pd.show_versions()

INSTALLED VERSIONS

commit : bb1f651
python : 3.8.12.final.0
python-bits : 64
OS : Linux
OS-release : 4.15.0-123-generic
Version : #126-Ubuntu SMP Wed Oct 21 09:40:11 UTC 2020
machine : x86_64
processor : x86_64
byteorder : little
LC_ALL : None
LANG : en_GB.UTF-8
LOCALE : en_GB.UTF-8

pandas : 1.4.0
numpy : 1.22.1
pytz : 2021.3
dateutil : 2.8.2
pip : 21.3.1
setuptools : 60.5.0
Cython : 0.29.26
pytest : 6.2.5
hypothesis : None
sphinx : 4.4.0
blosc : None
feather : None
xlsxwriter : None
lxml.etree : 4.7.1
html5lib : None
pymysql : None
psycopg2 : 2.9.3
jinja2 : 3.0.3
IPython : 8.0.1
pandas_datareader: None
bs4 : 4.10.0
bottleneck : None
fastparquet : None
fsspec : 2022.01.0
gcsfs : None
matplotlib : 3.5.1
numba : None
numexpr : None
odfpy : None
openpyxl : None
pandas_gbq : None
pyarrow : None
pyreadstat : None
pyxlsb : None
s3fs : None
scipy : 1.7.3
sqlalchemy : 1.4.31
tables : None
tabulate : 0.8.9
xarray : 0.20.2
xlrd : 2.0.1
xlwt : None
zstandard : None

[jbrockmendel]

The .loc views are behaving correctly. The change is that df["a"] = df["a"] * 2 is now creating a new array instead of writing into the existing array. See https://pandas.pydata.org/pandas-docs/stable/whatsnew/v1.4.0.html#bug-fixes Bug in DataFrame.__setitem__() incorrectly writing into an existing column’s array rather than setting a new array when the new dtype and the old dtype match (GH43406)

Setting with df.loc[:, "a"] = ... should be inplace. Trouble is that behavior is inconsistent until #45333.

[slackline]

Cool, thanks for the quick response and my apologies for not linking it to #43406 and other associated issues.

I did search through the bug-fixes of 1.4.0 and issues but was looking for mention of .loc and .copy() behaviour and didn't make the connection to DataFrame.__setitem__(), again my apologies.

Glad to see everything is in hand, thanks.