REGR: ExtensionArray aggregation on non-numeric types fails

[BryanCutler]

• 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

df = pd.DataFrame({'key': ['a', 'a', 'b', 'b'], 'val': ['a', 'b', 'c', 'd']}, dtype="string")
df.groupby("key").agg({'val': 'first'})

Expected:

    val
key    
a     a
b     c

Actual fails with error:

lib/python3.8/site-packages/pandas/core/groupby/ops.py in _ea_wrap_cython_operation(self, kind, values, how, axis, min_count, **kwargs)
    538             return result
    539 
--> 540         raise NotImplementedError(values.dtype)
    541

NotImplementedError: string

Problem description

Calling groupby on an ExtensionArray and aggregation on a string column fails with the error: NotImplementedError: string. This is because an optimized cython aggregation is attempted and when fails, the error string is improperly checked and does not fallback to a non-cython aggregation.

This was found with the extension arrays from https://github.com/CODAIT/text-extensions-for-pandas/blob/master/text_extensions_for_pandas/io/bert.py#L217 and full error log can be seen here https://github.com/CODAIT/text-extensions-for-pandas/pull/157/checks?check_run_id=1646927381#step:5:1438

Expected Output

Pandas should handle the NotImplementedError and fallback to a non-cython aggregation implementation.

Output of pd.show_versions()

INSTALLED VERSIONS

commit : 3e89b4c
python : 3.8.6.final.0
python-bits : 64
OS : Linux
OS-release : 5.4.0-58-generic
Version : #64-Ubuntu SMP Wed Dec 9 08:16:25 UTC 2020
machine : x86_64
processor : x86_64
byteorder : little
LC_ALL : None
LANG : en_US.UTF-8
LOCALE : en_US.UTF-8

pandas : 1.2.0
numpy : 1.19.1
pytz : 2020.1
dateutil : 2.8.1
pip : 20.2.4
setuptools : 49.6.0.post20201009
Cython : None
pytest : 6.1.1
hypothesis : None
sphinx : None
blosc : None
feather : None
xlsxwriter : None
lxml.etree : None
html5lib : None
pymysql : None
psycopg2 : None
jinja2 : 2.11.2
IPython : 7.18.1
pandas_datareader: None
bs4 : None
bottleneck : None
fsspec : None
fastparquet : None
gcsfs : None
matplotlib : 3.3.2
numexpr : None
odfpy : None
openpyxl : None
pandas_gbq : None
pyarrow : 2.0.0
pyxlsb : None
s3fs : None
scipy : 1.5.2
sqlalchemy : None
tables : None
tabulate : None
xarray : None
xlrd : None
xlwt : None
numba : None

[BryanCutler]

I can work on a patch to fix this

[jorisvandenbossche]

Thanks for the report!
Yes, this is caused by the refactoring in groupby, it was working before correctly, so labeling as regression (and targeting 1.2.1)

[jreback]

pls edit the top include the full example and not just the link

[jorisvandenbossche]

There is a reproducible example now at #38982 (comment)

[simonjayhawkins]

first bad commit: [40ca2b9] REF: implement _ea_wrap_cython_operation (#38162) cc @jbrockmendel