Geethapranay1 commented on issue #4131:
URL: 
https://github.com/apache/datafusion-comet/issues/4131#issuecomment-4417990035

   i don’t think this is only a Spark vs DF hashing issue.
   For unordered FIRST/LAST, the intermediate state is just [value, is_set] 
once the partial hash aggregate emits that state, the original row order is 
lost. After that, PartialMerge becomes dependent on hash-table emission order, 
so changing the hasher alone would not make it correct
   
   I added a native-side guard to reject FIRST/LAST in PartialMerge before 
MergeAsPartial, and a unit test for that planner path. This keeps the existing 
fallback and makes the protection explicit on the native side as well 
   can i open pr for this?
   
   A full native fix would need order-preserving merge semantics or richer 
intermediate state.


-- 
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]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to