pitrou commented on pull request #11019: URL: https://github.com/apache/arrow/pull/11019#issuecomment-914425223
> Note that both algorithms uses NonStablePartitioner to separete Nulls and both are unstable. Oh, you're right! I had forgotten that `partition_nth_indices` was unstable, unlike `sort_indices`. > Clearly, PartitionNthToIndices_nth_element is faster (5-6x) but PartitionNthToIndices kernel is not enough to implement TopK/BottomK because the first K elements needs to be sorted. But if K is small, sorting them should be cheap, right? I'm not sure what the cutoff would be... (sqrt(N) perhaps?) > IMO this PR should first expose a select_k_unstable and look in a follow up PR the best implementation for select_k_stable Agreed! -- 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]
