BUG: Rounding on BooleanDtype raises an exception

[JoaquimEsteves]

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


# Using python's native booleans is A-OK
pd.DataFrame([True]).round(0)
pd.DataFrame([True]).round(4)

pd.DataFrame([True]).astype(pd.BooleanDtype()).round(0)

# TypeError: FloatingArray does not support np.float16 dtype.

pd.DataFrame([True]).astype(pd.BooleanDtype()).round(4)

# numpy.core._exceptions._UFuncOutputCastingError: Cannot cast ufunc 'multiply' output from dtype('float64') to dtype('bool') with casting rule 'same_kind'

Issue Description

The type BooleanDtype seems to have a bug when we try to round it.

Internally it looks like the values are stored as a float16 array - this leads to problems when attempting to round those values to 0 as the _maybe_mask_result will attempt to cast the data to a Floating Array. Relevant Line

Note that on an older version of pandas I had lying around (1.4.3) this didn't happen - so it looks to me like a regression.

Expected Behavior

The expected behaviour is that pandas shouldn't crash when attempting to round booleans.

It's noted in a comment in the code that numpy is the one responsible for casting the boolean array into a float16; however the crash only happens with BooleanDtype

Installed Versions

INSTALLED VERSIONS

commit : b2d9ec1
python : 3.10.12.final.0
python-bits : 64
OS : Linux
OS-release : 6.2.0-36-generic
Version : #37~22.04.1-Ubuntu SMP PREEMPT_DYNAMIC Mon Oct 9 15:34:04 UTC 2
machine : x86_64
processor : x86_64
byteorder : little
LC_ALL : None
LANG : en_US.UTF-8
LOCALE : en_US.UTF-8

pandas : 2.2.0.dev0+564.gb2d9ec17c5
numpy : 1.26.2
pytz : 2023.3.post1
dateutil : 2.8.2
setuptools : 59.6.0
pip : 22.0.2
Cython : None
pytest : None
hypothesis : None
sphinx : None
blosc : None
feather : None
xlsxwriter : None
lxml.etree : None
html5lib : None
pymysql : None
psycopg2 : None
jinja2 : None
IPython : None
pandas_datareader : None
bs4 : None
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
python-calamine : 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

[JoaquimEsteves]

After debugging a little more.

It looks like round(0) is crashed by Pandas' self._maybe_mask_result[1]
Whereas round(4) is crashed by numpy's np.round [2]

[1]

pandas/pandas/core/arrays/masked.py

Line 844 in b2d9ec1

return FloatingArray(result, mask, copy=False)

[2]

pandas/pandas/core/arrays/masked.py

Line 387 in b2d9ec1

values = np.round(self._data, decimals=decimals, **kwargs)