BUG: Update Values from MultiIndex Dataframe after using loc is not applied correctly

[bbizarro]

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
import numpy as np

df = pd.DataFrame()
df['idx1'] = [0,1,2,3,4]
df['idx2'] = [0,1,2,3,4]
df['abool'] = True
df['a'] = np.arange(5, dtype='int64')
df['b'] = np.arange(5, dtype='float64')
df['c'] = np.arange(5, dtype='int64')
df = df.set_index(['idx1','idx2'])

idx = [(1,1),(3,3)]
df.loc[idx, 'c'] = 0
df.loc[idx, 'c']

df.loc[idx, ['a', 'b']] # This line is the problem

df.loc[idx, 'c'] = 15
df.loc[idx, 'c']

Issue Description

In Pandas 1.4.x the last line is not showing the change in the column c and the values still being 0.

>>> idx1  idx2
1     1       0
3     3       0
Name: c, dtype: int64

When I call the column c as DataFrame I get the correct values

df.loc[idx, ['c']]

>>>         c
idx1 idx2    
1    1     15
3    3     15

Expected Behavior

On Pandas 1.3.5 I get

>>> idx1  idx2
1     1       15
3     3       15
Name: c, dtype: int64

Installed Versions

INSTALLED VERSIONS

commit : 66e3805
python : 3.8.12.final.0
python-bits : 64
OS : Windows
OS-release : 7
Version : 6.1.7601
machine : AMD64
processor : Intel64 Family 6 Model 60 Stepping 3, GenuineIntel
byteorder : little
LC_ALL : None
LANG : None
LOCALE : Spanish_Chile.1252

pandas : 1.3.5
numpy : 1.23.1
pytz : 2022.1
dateutil : 2.8.2
pip : 22.2
setuptools : 59.8.0
Cython : None
pytest : None
hypothesis : None
...
xarray : None
xlrd : 2.0.1
xlwt : None
numba : None

[simonjayhawkins]

@bbizarro Thanks for the report.

In Pandas 1.4.x the last line is not showing the change in the column c and the values still being 0.

Not able to reproduce. The code sample gives the expected result on 1.3.5, 1.4.3 and main for me. Can you double check the code sample.

[BeRT2me]

This is missing a line where you have to try to access df.loc[idx, 'c'] before df.loc[idx, ['a', 'b']], see my above mentioned report for a working example. @simonjayhawkins

[bbizarro]

I updated the code sample with the missing line mentioned by @BeRT2me and I get the problem on 1.4.3

[simonjayhawkins]

In Pandas 1.4.x the last line is not showing the change in the column c and the values still being 0.

first bad commit: [573c063] BUG: frame.loc[2:, 'z'] not setting inplace when multi-block (#44345)

but looks like fixed in commit: [c80f656] BUG: frame[x].loc[y] inconsistent with frame.at[x, y] GH#22372 (#45287)

I'll milestone as 1.4.x for now as we maybe able to backport that fix.