jorisvandenbossche commented on issue #35624: URL: https://github.com/apache/arrow/issues/35624#issuecomment-1561396556
> Oh wow, `fill_null` works quite differently than I expected. Your example of passing it an array and it working _elementwise_ is very surprising to me! `fill_value`'s documentation is a bit misleading, then; it says Yeah, that's indeed certainly not properly documented. I am not sure this was fully intentional (maybe when it was added, mostly the scalar fill_value was considered), but this is essentially an alias for `coalesce`, and it is this function that supports this behaviour. And also for example `fillna` in pandas has the same ability of either filling with a scalar or filling with the corresponding values of a same-length array. Your suggested rewording sounds good! > This issue seems to fracture into three issues: Yes, that's a good summary. > can't make an explicit scalar element of a fixed-size list (which sounds kind of like https://github.com/apache/arrow/issues/18987 perhaps) That's indeed a related issue. It's also related to https://github.com/apache/arrow/issues/21761, for accepting pyarrow values in `pa.array(..)` constructor (and not `pa.scalar(..)`). For this one, there is an open PR, but it will probably depend on how that PR implements the fix whether that would also fix the scalar() version (we don't call `pa.array(..)` inside `pa.scalar(..)`, but directly the lower level `ConvertPySequence`). -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
