BUG: DataFrame.update do not preserve string dtype

[ali-cetin-4ss]

• 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 master branch of pandas.

Reproducible Example

df_org = pd.DataFrame({"a": [None, "two"], "b": [1, 2]}, index=[0, 1]).astype({"a": "string", "b": "Int64"})
df_update = pd.DataFrame({"a": ["one"]}, index=[0], dtype="string")

print(df_org, "\n")
print(df_org.dtypes, "\n")

print(df_update, "\n")
print(df_update.dtypes, "\n")

# update with consistent dtype
df_org.update(df_update)

print(df_org, "\n")
print(df_org.dtypes, "\n")

# dtype not conserved - string -> object

Issue Description

String dtype column is casted to object when .update is performed even when the "other" dataframe has consistent dtypes. See code example

Expected Behavior

The dtype should be preserved after the update operation, i.e., string should remain string and not casted to object.

Installed Versions

INSTALLED VERSIONS

commit : 945c9ed
python : 3.8.10.final.0
python-bits : 64
OS : Windows
OS-release : 10
Version : 10.0.19043
machine : AMD64
processor : Intel64 Family 6 Model 142 Stepping 12, GenuineIntel
byteorder : little
LC_ALL : None
LANG : None
LOCALE : Norwegian Bokmål_Norway.1252

pandas : 1.3.4
numpy : 1.20.3
pytz : 2021.3
dateutil : 2.8.2
pip : 21.1.3
setuptools : 58.2.0
Cython : None
pytest : 6.2.5
hypothesis : None
sphinx : 4.0.2
blosc : None
feather : None
xlsxwriter : None
lxml.etree : None
html5lib : None
pymysql : None
psycopg2 : None
jinja2 : 3.0.2
IPython : 7.28.0
pandas_datareader: None
bs4 : 4.9.3
bottleneck : None
fsspec : None
fastparquet : None
gcsfs : None
matplotlib : None
numexpr : None
odfpy : None
openpyxl : None
pandas_gbq : None
pyarrow : 5.0.0
pyxlsb : None
s3fs : None
scipy : 1.7.1
sqlalchemy : None
tables : None
tabulate : None
xarray : None
xlrd : None
xlwt : None
numba : 0.54.0

[CloseChoice]

>>> import pandas as pd
>>> df_org = pd.DataFrame({"a": [None, "two"], "b": [1, 2]}, index=[0, 1]).astype({"a": "string", "b": "Int64"})
>>> df_update = pd.DataFrame({"a": ["one"]}, index=[0], dtype="string")
>>> 
>>> print(df_org, "\n")
      a  b
0  <NA>  1
1   two  2 

>>> print(df_org.dtypes, "\n")
a    string
b     Int64
dtype: object 

>>> 
>>> print(df_update, "\n")
     a
0  one 

>>> print(df_update.dtypes, "\n")
a    string
dtype: object 

>>> 
>>> # update with consistent dtype
>>> df_org.update(df_update)
>>> 
>>> print(df_org, "\n")
     a  b
0  one  1
1  two  2 

>>> print(df_org.dtypes, "\n")
a    object
b     Int64
dtype: object 

Is this a behaviour that was different in previous releases? Is this really unexpected behaviour? I guess there are a couple of situations where we cast to object without necessity.

[ali-cetin-4ss]

For an outsider, this feels like update method is not aware of the "new" built-in native string data type. In practice, it's a bit confusing to unintentionally re-cast dtypes without any apparent reason.

[rpkilby]

This issue seems to affect other new-style dtypes such as Int64. Additionally, I think this is the source of some update() calls generating FutureWarnings about loc/iloc usage.

[rpkilby]

After digging through the issue tracker a bit further, I believe this can be closed as a duplicate of #4094.