zhli1142015 commented on code in PR #11197:
URL:
https://github.com/apache/incubator-gluten/pull/11197#discussion_r2568098838
##########
backends-velox/src/main/scala/org/apache/gluten/execution/ColumnarCollectLimitExec.scala:
##########
@@ -28,6 +29,76 @@ case class ColumnarCollectLimitExec(
offset: Int = 0
) extends ColumnarCollectLimitBaseExec(limit, child, offset) {
+ /**
+ * Override doExecuteColumnar to handle zero-column schema specially. Velox
doesn't support
Review Comment:
I think this is doable, but it requires some special handling in both the
shuffle writer and shuffle reader. Since there are no columns, they won’t read
or write column buffers or row data as in the normal logic — the only piece of
information that needs to be transferred is the row count.
My understanding is that this is the only place where empty-schema shuffle
is used. Please let me know if there are other places that need to invoke
empty-schema shuffle. If there are many use cases, then I think it would be
worthwhile to make shuffle support empty-schema input directly.
--
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]
