API: Series[bool with missing] + bool consistency

[mroeschke]

In [2]: In [1]: import pandas as pd
 ...: ...: import pyarrow as pa
 ...: ...:
 ...: ...: s = pd.Series([True, False, None])
 ...: ...: s + True
Out[2]: 
0 2
1 1
2 NaN
dtype: object

In [3]: In [1]: import pandas as pd
 ...: ...: import pyarrow as pa
 ...: ...:
 ...: ...: s = pd.Series([True, False, None], dtype=pd.BooleanDtype())
 ...: ...: s + True
Out[3]: 
0 True
1 True
2 <NA>
dtype: boolean

In [4]: In [1]: import pandas as pd
 ...: ...: import pyarrow as pa
 ...: ...:
 ...: ...: s = pd.Series([True, False, None], dtype=pd.ArrowDtype(pa.bool_()))
 ...: ...: s + True
...
File ~/pandas/core/arrays/arrow/array.py:1022, in ArrowExtensionArray._evaluate_op_method(self, other, op, arrow_funcs)
 1020 result = pc_func(self._pa_array, other)
 1021 except pa.ArrowNotImplementedError as err:
-> 1022 raise TypeError(self._op_method_error_message(other_original, op)) from err
 1023 return self._from_pyarrow_array(result)

TypeError: operation 'add' not supported for dtype 'bool[pyarrow]' with object of type <class 'bool'>

In addition to consistency, which result is preferred? I could see Out[2] or raising being the sensible behavior

[zhangbowen-coder]

Maybe Out[2] is better, since user may know there is a None in the Series being used and decide to do the plus operation. But in another case, the user may not be aware and need a warning, so my advice is to take Out[2] as the result of the "+" operator and provide another safe operator in the form of a function(e.g. "add") which will raise an exception when there is a None(or Nan) in the Series.