BUG: Inconsistend behaviour when assigning dict to df.loc

[RagBlufThim]

• I have checked that this issue has not already been reported.

• I have confirmed this bug exists on the latest version of pandas.

• (optional) I have confirmed this bug exists on the master branch of pandas.

---

Code Sample, a copy-pastable example

import pandas as pd

dfo = pd.DataFrame({"A": ["abc", "def"], "B": ["ghi", "jkl"]}, dtype="object")
dfs = pd.DataFrame({"A": ["abc", "def"], "B": ["ghi", "jkl"]}, dtype="string")

dfo.loc[0, :] = {"A": "newA", "B": "newB"}
dfs.loc[0, :] = {"A": "newA", "B": "newB"}

print(dfo)
# Output:
#       A     B
# 0  newA  newB
# 1   def   jkl

print(dfs)
# Output:
#      A    B
# 0    A    B
# 1  def  jkl

Problem description

The interpretation of the RHS_dict depends on the dtype of the dataframe assigned to:
Assigning to an object-dataframe: Interpretation as pd.Series(RHS_dict)
Assigning to a string-dataframe: Interpretation as list(RHS_dict)

This is inconsistent and unexpected.

In my script I started with an object-dataframe and later changed to a string-dataframe. This change led to an incorrect result, due to the effect desribed above.

Expected Output

The interpretation of the RHS_dict does NOT depend on the dtype of the dataframe asigned to.
Personally I'd prefer pd.Series(RHS_dict), but if wiser men than me decide that list(RHS_dict) is correct, I won't object, as long as the behaviour is consistent across dtypes.

Output of pd.show_versions()

Details

INSTALLED VERSIONS

commit : 2cb9652
python : 3.9.4.final.0
python-bits : 64
OS : Windows
OS-release : 10
Version : 10.0.18362
machine : AMD64
...

pandas : 1.2.4
numpy : 1.20.2
pytz : 2021.1
dateutil : 2.8.1
...

[asishm-wk]

Seems to be fixed on master.

[phofl]

Yes this works now. But tests would be good.

Dict support inside DataFrame columns is not that great

[JavierRuano]

import pandas as pd

dfo = pd.DataFrame({"A": ["abc", "def"], "B": ["ghi", "jkl"]}, dtype="object")
dfs = pd.DataFrame({"A": ["abc", "def"], "B": ["ghi", "jkl"]}, dtype="string")
dfo=dfo.convert_dtypes(infer_objects=True)
dfo.loc[0, :] = {"A": "newA", "B": "newB"}
dfs.loc[0, :] = {"A": "newA", "B": "newB"}

print(dfo)

Output:

A B

0 A B

1 def jkl

print(dfs)

Output:

A B

0 A B

1 def jkl

[arredond]

take