BUG: Groupby then sum boolean column gives too large a sum

[agaventa]

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

n = 32

df = pd.DataFrame({'label': [1]*n,'bool_column': [True]*(n)})

print(df.assign(bool_column_dupe=(df.bool_column==True)).groupby("label").bool_column_dupe.sum())

Issue Description

Summing a boolean column after a groupby gives an unreasonably large and incorrect sum.
This is the result from the above code:
1 8128 Name: bool_column_dupe, dtype: int64

The issue does not occur if I instead add an astype(int):
print(df.assign(bool_column_dupe(df.bool_column==True).astype(int)).groupby("label").bool_column_dupe.sum())

The issue does not occur if I don't use the .assign:
print(df.groupby("label").bool_column.sum())

The issue does not occur with n<32.

The issue only occurs with numpy v1.24.0.
It does not occur with numpy v1.23.5.

Expected Behavior

Expected to see:
1 32 Name: bool_column_dupe, dtype: int64

Installed Versions

INSTALLED VERSIONS

commit : 8dab54d
python : 3.9.13.final.0
python-bits : 64
OS : Darwin
OS-release : 22.1.0
Version : Darwin Kernel Version 22.1.0: Sun Oct 9 20:14:54 PDT 2022; root:xnu-8792.41.9~2/RELEASE_X86_64
machine : x86_64
processor : i386
byteorder : little
LC_ALL : None
LANG : None
LOCALE : en_GB.UTF-8
pandas : 1.5.2
numpy : 1.24.0
pytz : 2022.7
dateutil : 2.8.2
setuptools : 60.2.0
pip : 21.3.1
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
brotli : 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
snappy : None
sqlalchemy : None
tables : None
tabulate : None
xarray : None
xlrd : None
xlwt : None
zstandard : None
tzdata : None

[asishm]

this might be a numpy issue

In [77]: a = np.array([True] * 32)

In [78]: b = (a == True)

In [79]: b.view('uint8')
Out[79]:
array([254, 254, 254, 254, 254, 254, 254, 254, 254, 254, 254, 254, 254,
       254, 254, 254, 254, 254, 254, 254, 254, 254, 254, 254, 254, 254,
       254, 254, 254, 254, 254, 254], dtype=uint8)

In [80]: np.__version__
Out[80]: '1.24.0'

[phofl]

This works on main for me, so might need tests.

We made quite a lot of changes on groupby algorithms to preserve dtypes better, so could be related to that

[asishm]

I don't think this works on main

In [3]: pd.__version__, np.__version__
Out[3]: ('2.0.0.dev0+971.gca3e0c875f', '1.24.0')

In [4]: n = 32
   ...: df = pd.DataFrame({'label': 1, 'a': [True] * n})
   ...: df['b'] = df['a'] == True
   ...: df.groupby('label').sum()
Out[4]:
        a     b
label
1      32  8128

[phofl]

Hm good point, did not use numpy 1.24.0
I agree that this looks like a numpy bug

[agaventa]

This is now fixed in numpy 1.24.1

[Daquisu]

take