BUG: Unexpected behavior when a DataFrame has MultiIndex columns and non-unique index

[jonmooser]

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 numpy as np
import pandas as pd


df = pd.DataFrame(
    np.random.random((5, 3)),
    index=[1, 2, 2, 3, 4],
    columns=pd.MultiIndex.from_product([np.arange(3), ["x"]]),
)

print(df)


"""
This works, but is arguably a bug.
It returns a Series of two values, but the contract of the
DataFrame.at, as I understand, is to return a single value
"""
print(df.at[2, 1])
print()

"""
This works and returns a scalar value as expected
"""
print(df.loc[1, (1, "x")])
print()

"""
Fails with 
ValueError: Invalid call for scalar access (getting)!
"""
print(df.at[1, (1, "x")])

Issue Description

When a DataFrame has one axis with non-unique values, and one axis with a multi-index, the DataFrame.at method throws a confusing error.
Even if this is strictly expected, the error should be more clear (And maybe KeyError instead of ValueError)

It feels like a bug because

1. Calling loc with the same parameters works and
2. There is no ambiguity in the values passed
   (there is only one row with index=1)

Traceback (most recent call last):
  File "/Users/jonathan/Dropbox/py/debug_tests/pandas_mi_bug.py", line 33, in <module>
    print(df.at[1, (1, "x")])
          ~~~~~^^^^^^^^^^^^^
  File "/Library/Frameworks/Python.framework/Versions/3.11/lib/python3.11/site-packages/pandas/core/indexing.py", line 2485, in __getitem__
    raise ValueError("Invalid call for scalar access (getting)!")
ValueError: Invalid call for scalar access (getting)!

Expected Behavior

In the example above,

print(df.at[1, (1, "x")])

Should return a single value at that location

If it should throw an Error, make it clear that the duplicates in the index is the problem (concat() does this)
Also note:

• If the index contains unique values, it will work as expected
• It the columns are not a MultiIndex, it will work as expected
  In other words both axes need to have these properties for the bug to appear

Installed Versions

/Library/Frameworks/Python.framework/Versions/3.11/lib/python3.11/site-packages/_distutils_hack/__init__.py:33: UserWarning: Setuptools is replacing distutils. warnings.warn("Setuptools is replacing distutils.")

INSTALLED VERSIONS

commit : ba1cccd
python : 3.11.0.final.0
python-bits : 64
OS : Darwin
OS-release : 21.4.0
Version : Darwin Kernel Version 21.4.0: Fri Mar 18 00:46:32 PDT 2022; root:xnu-8020.101.4~15/RELEASE_ARM64_T6000
machine : arm64
processor : arm
byteorder : little
LC_ALL : None
LANG : en_US.UTF-8
LOCALE : en_US.UTF-8

pandas : 2.1.0
numpy : 1.23.4
pytz : 2022.5
dateutil : 2.8.2
setuptools : 65.5.0
pip : 23.2.1
Cython : None
pytest : 7.2.0
hypothesis : None
sphinx : None
blosc : None
feather : None
xlsxwriter : None
lxml.etree : None
html5lib : None
pymysql : None
psycopg2 : None
jinja2 : 3.1.2
IPython : 8.12.0
pandas_datareader : None
bs4 : 4.12.2
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 : 2023.3
qtpy : None
pyqt5 : None

[Matheus-Ribeiro-ITA]

I faced a similar issue using .at. It works with .loc, but not .at. I'm using pandas 2.2.3.

Unfortunately, I wasn't able to reproduce it on a separate file. When I tried copying the values to test it, it simply worked 😅

This is the code snippet I tried to reproduce the bug, but failed to reproduce it.

import pandas as pd

print(pd.__version__)

data = [
    {
        'date': pd.Timestamp('2024-10-23 23:59:59-0300', tz='Etc/GMT+3'),
        'asset_id': 77,
        'financial_institution_id': 3066.0,
        'dividend_cash_value': 0.0,
    }
]

df = pd.DataFrame(data)


df.set_index(['date', 'asset_id', 'financial_institution_id'], inplace=True)

p_key = (
    pd.Timestamp('2024-10-23 23:59:59-0300', tz='Etc/GMT+3'),
    float(77),
    float(3066.0),
)
if p_key in df.index:
    df.at[p_key, 'dividend_cash_value'] += 100

print(df)

[jonmooser]

Hard to know what is different in your production code, but the code you pasted above won't trigger the bug I reported.
You need to have a DataFrame where either the columns or index has a redundancy. In your case there is only a single column, so you can't have two columns with the same name.