Unary ~ operator broken for BooleanArray

[birnbera]

This may be related to https://github.com/pandas-dev/pandas/pull/31484.

Unary application of the ~ operator seems to operate bitwise on an underlying integer array rather than the expected negation of a boolean:

>>> s = Series([True, False], dtype="boolean")
>>> ~s
0    -2
1    -1
dtype: object
>>> s = Series([True, False], dtype="bool")
>>> ~s
0    False
1     True
dtype: bool

This behavior prevents the correct handling of boolean indexing:

>>> a = pd.Series(list("abcd"))
>>> s = pd.Series([True, False, True, False], dtype="boolean")
>>> a[s]
0    a
2    c
dtype: object
>>> a[~s]
...
KeyError: "None of [Int64Index([-2, -1, -2, -1], dtype='int64')] are in the [index]"

The problem goes away if both of your series are ExtensionArray types:

>>> a = pd.Series(list('abcd'), dtype="string")
>>> s = pd.Series([True, False, True, False], dtype="boolean")
>>> a[s]
0    a
2    c
dtype: object
>>> a[~s]
1    b
3    d
dtype: string

Edit

I updated to version 1.0.1 and the initial issue goes away (i.e. the unary ~ gives the expected result). However, the dtype reverts to object after assignment:

>>> s = pd.Series([True, False, pd.NA], dtype="boolean") # same behavior whether or not you include an `NA`
>>> s
0     True
1    False
2     <NA>
dtype: boolean
>>> ~s
0    False
1     True
2     <NA>
dtype: boolean
>>> s[3] = True
>>> s
0     True
1    False
2     <NA>
3     True
dtype: object
>>> ~s
0      -2
1      -1
2    <NA>
3      -2
dtype: object

Output of pd.show_versions()

INSTALLED VERSIONS

commit : None
python : 3.8.1.final.0
python-bits : 64
OS : Darwin
OS-release : 18.7.0
machine : x86_64
processor : i386
byteorder : little
LC_ALL : None
LANG : en_US.UTF-8
LOCALE : en_US.UTF-8
pandas : 1.0.0
numpy : 1.18.1
pytz : 2019.3
dateutil : 2.8.1
pip : 20.0.2
setuptools : 45.1.0.post20200119
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 : 7.12.0
pandas_datareader: None
bs4 : None
bottleneck : None
fastparquet : None
gcsfs : None
lxml.etree : None
matplotlib : 3.1.3
numexpr : None
odfpy : None
openpyxl : None
pandas_gbq : None
pyarrow : None
pytables : None
pytest : None
pyxlsb : None
s3fs : None
scipy : 1.4.1
sqlalchemy : 1.3.13
tables : None
tabulate : None
xarray : 0.15.0
xlrd : 1.2.0
xlwt : None
xlsxwriter : None
numba : None

[jorisvandenbossche]

The unary operator is indeed fixed in 1.0.1. For the assignment problem, I created a dedicated issue for that: #32346

Thanks for the report!