Is your feature request related to a problem or challenge?
array_sort is fast for arrays of primitive types. For arrays of non-primitive types, Arrow's sort kernel does sort_by_indices and then take; we do this for every row and then concat the results. It would be faster to accumulate all the (batch-wise) indices and then do a single take to construct the result set.
Describe the solution you'd like
No response
Describe alternatives you've considered
No response
Additional context
No response