Closed
Description
We should refactor out parts of ArrowStringArray
and tests.extension.arrow.arrays.ArrowExtensionArray
that can be re-used for more general arrow-based EAs.
A couple of questions that came up in my first pass at this (cc @jorisvandenbossche):
- Should the general case allow backing by either pa.Array or pa.ChunkedArray, or just pa.ChunkedArray?
- ArrowStringArray.copy relies on immutability. Will that be the case in general? If not, how do we check for it?
- Is the more verbose
nbytes
in ArrowExtensionArray more robust/correct than the one in ArrowStringArray? - Are the slightly-different
_concat_same_type
methods equivalent? Usable for the general case?