Feature request: support for multiprocessing.Pool "initializer". #231
Description
The multiprocessing.Pool interface provides the ability to pass a custom "initializer" function to each worker: https://docs.python.org/3/library/multiprocessing.html#multiprocessing.pool.Pool.
This is useful for things like suppressing specific warnings, setting up logging, and other "scaffolding" that is occasionally useful (or even required) in applications.
It seems like Pandarallel does not use its own initializer function, so it would be nice if users could provide their own.
Looking over the code, this seems like a relatively unintrusive backward-compatible change that most users won't notice at all, but would benefit the small number of users who do want or need this feature.
Hypothetical usage:
def _suppress_shapely_warning(ignore: bool = True) -> None:
import warnings
warnings.filterwarnings(
'ignore' if ignore else 'default',
message='invalid value encountered in intersects',
category=RuntimeWarning,
module=r'shapely\.predicates',
lineno=758,
append=True,
)
pandarallel.initialize(
nb_workers=10,
progress_bar=False,
initializer=_suppress_shapely_warning,
initargs=(True,),
)