pitrou commented on pull request #11019:
URL: https://github.com/apache/arrow/pull/11019#issuecomment-914350801
Le 07/09/2021 à 15:46, David Li a écrit :
> Hmm, why not, but sort_indices and partition_nth_indices are
> guaranteed to be stable, so it would be better if topK/bottomK
> wasn't inconsistent.
>
> Since the unstable version is quite a bit faster so far here, maybe we
> should expose a |select_k_unstable|?
If "quite a bit" means 2x or 3x rather than 20%, then that sounds good
to me indeed.
> (Was the reason why we don't have
> an unstable sort because it's not useful and/or wasn't meaningfully
faster?)
I'm not sure we have actually tried to benchmark an unstable sort. But
being stable is usually a desirable guarantee when the sort columns
don't span all columns. Of course, that depends on the use case :-)
--
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]
