westonpace commented on PR #15142:
URL: https://github.com/apache/arrow/pull/15142#issuecomment-1368316255

   > Just a question, is there a sort option which gives precedence to the 
index of the row and decide which comes first, when we have a tie like this?
   
   That's called a "stable sort". The underlying sort kernel (SortIndices) 
**is** stable.  However, if the plan is run in parallel, then there is no 
guarantee the batches will accumulate in the same order.  So even if the sort 
kernel is stable the sort node is not.
   
   Once we add proper ordering we can add a stable option to the sort node 
which resequences the data before sorting so that the sort node can remain 
stable.
   
   However, now that I write this, I realize it might be best to only apply my 
change when testing the parallel case, and to use the old comparison in the 
non-parallel 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]

Reply via email to