BUG: Setting values on a multiindex df, via loc, does nothing on 1.4.x

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
print("pandas version:", pd.__version__)

index = pd.MultiIndex.from_tuples(
    [("A", "a"), ("A", "b"), ("A", "c"), ("A", "d"), ("B", "a"), ("B", "b")]
)

df = pd.DataFrame(
    [[10, 100], [20, 200], [30, 300], [40, 400], [50, 500], [60, 600]],
    columns=["val1", "val2"],
    index=index,
)

df.loc['A']['val1'] = 1
print(df)

Issue Description

On Pandas 1.4.x (including 1.4.2), the example code does not modify df. The value 1 is not assigned.
It returns:

pandas version: 1.4.2
     val1  val2
A a    10   100
  b    20   200
  c    30   300
  d    40   400
B a    50   500
  b    60   600

(This bug could be linked to #46837)

Expected Behavior

on Pandas 1.3.5, the example returns:

pandas version: 1.3.5
     val1  val2
A a     1   100
  b     1   200
  c     1   300
  d     1   400
B a    50   500
  b    60   600

Installed Versions

INSTALLED VERSIONS

commit : 4bfe3d0
python : 3.8.10.final.0
python-bits : 64
OS : Windows
OS-release : 10
Version : 10.0.19042
machine : AMD64
processor : Intel64 Family 6 Model 142 Stepping 10, GenuineIntel
byteorder : little
LC_ALL : None
LANG : None
LOCALE : English_United Kingdom.1252

pandas : 1.4.2
numpy : 1.22.3
pytz : 2022.1
dateutil : 2.8.2
pip : 22.0.4
setuptools : 56.0.0
Cython : None
pytest : 7.1.1
hypothesis : None
sphinx : 4.5.0
blosc : None
feather : None
xlsxwriter : None
lxml.etree : None
html5lib : None
pymysql : None
psycopg2 : None
jinja2 : 3.1.1
IPython : 8.2.0
pandas_datareader: None
bs4 : 4.11.1
bottleneck : None
brotli : None
fastparquet : None
fsspec : None
gcsfs : None
markupsafe : 2.1.1
matplotlib : 3.5.1
numba : None
numexpr : 2.8.1
odfpy : None
openpyxl : None
pandas_gbq : None
pyarrow : None
pyreadstat : None
pyxlsb : None
s3fs : None
scipy : 1.8.0
snappy : None
sqlalchemy : None
tables : 3.7.0
tabulate : None
xarray : None
xlrd : None
xlwt : None
zstandard : None

[simonjayhawkins]

Thanks @Kerybas for the report.

(This bug could be linked to #46837)

first bad commit: [03dd698] BUG: DataFrame.setitem sometimes operating inplace (#43406)

I think this was considered a bug, even though a long standing behavior.

I've labelled as a regression for now, pending further discussion.

cc @jbrockmendel

Hi @simonjayhawkins,

Which one is considered the buggy behaviour?
what we have in 1.3.5 or what we have in 1.4?

As a user, I would expect the former behaviour, from 1.3.5, where .loc actually modifies values.
And if not, then I would at least expect a warning instead of no change and no feedback.

[phofl]

In general I would recommend to avoid chained indexing and use

df.loc['A', 'val1'] = 1

[simonjayhawkins]

As a workaround, I would suggest using either

df.loc[pd.IndexSlice["A", :], "val1"] = 1

or

df.loc[("A",), "val1"] = 1

to make the code more robust if column names also appear in the levels of the MultiIndex.

---

Which one is considered the buggy behaviour?
what we have in 1.3.5 or what we have in 1.4?

both

As a user, I would expect the former behaviour, from 1.3.5, where .loc actually modifies values.
And if not, then I would at least expect a warning instead of no change and no feedback.

yes. this should raise a warning if we decide not to revert to the 1.3.5 behavior