API: ExtensionArray._from_sequence_of_strings assumes true_values and false_values in signature when ExtensionDtype._is_boolean=True

[mroeschke]

ExtensionArray._from_sequence_of_strings has a signature

def _from_sequence_of_strings(
        cls, strings, *, dtype: Dtype | None = None, copy=False
    ):

BooleanArray._from_sequence_of_strings has a signature

def _from_sequence_of_strings(
        cls,
        strings: list[str],
        *,
        dtype: Dtype | None = None,
        copy: bool = False,
        true_values: list[str] | None = None,
        false_values: list[str] | None = None,
    )

This is causing issues toward when trying to develop an ArrowExtensionArray with ArrowDtype that can support multiple underlying dtypes: #46008

pandas/pandas/io/parsers/base_parser.py

Lines 766 to 779 in 612f566

	elif is_extension_array_dtype(cast_type):
	# ensure cast_type is an actual dtype and not a string
	cast_type = pandas_dtype(cast_type)
	array_type = cast_type.construct_array_type()
	try:
	if is_bool_dtype(cast_type):
	return array_type._from_sequence_of_strings(
	values,
	dtype=cast_type,
	true_values=self.true_values,
	false_values=self.false_values,
	)
	else:
	return array_type._from_sequence_of_strings(values, dtype=cast_type)

Possible paths forward:

1. Document this behavior (But this makes parameterized ExtensionDtypes difficult to implement)
2. Generalize & add a new keyword to _from_sequence_of_strings to provide a string conversion e.g.

def _from_sequence_of_strings(
        cls,
        strings: list[str],
        *,
        dtype: Dtype | None = None,
        copy: bool = False,
        string_map: dict[str, Any] | None = None OR conversion_func: Callable[str, Any] | None = None,
    )

[jbrockmendel]

i like the string_map idea, though id prefer to avoid the callable option if that is viable

[jbrockmendel]

I guess the downside of string_map would be that authors/readers might incorrectly think that anything other than true_values/false_values are actually used. Might be better to just add true_values/false_values to the base class signature and document how this is only used deep inside read_csv (and remove any other extraneous uses)