zhztheplayer commented on code in PR #12215:
URL: https://github.com/apache/gluten/pull/12215#discussion_r3630393697
##########
backends-velox/src-delta/main/scala/org/apache/gluten/component/VeloxDeltaComponent.scala:
##########
@@ -40,9 +40,22 @@ class VeloxDeltaComponent extends Component {
// PreprocessTableWithDVsStrategy injects the skip-row column and filter
during physical
// planning, DeltaPostTransformRules.nativeDeletionVectorRule strips them
when the scan
// offloads, and DeltaScanTransformer materializes the per-file DV
payloads for Velox.
+ //
+ // For native DELETE/UPDATE/MERGE, the DML target row-index scan is
deliberately kept on Spark
+ // until native row-index execution is proven; tag those scans here so the
post-transform rules
+ // can keep the small subtree off the native path.
+ legacy.injectPreTransform(_ =>
DeltaDeletionVectorDmlUtils.tagDmlRowIndexScans)
legacy.injectTransform {
c =>
- val offload = Seq(OffloadDeltaScan(), OffloadDeltaProject(),
OffloadDeltaFilter())
+ val offload = Seq(
+ OffloadDeltaScan(
+ enableNativeDeltaWriteKey =
VeloxDeltaConfig.ENABLE_NATIVE_WRITE.key,
Review Comment:
Is there any particular reason we should check
`VeloxDeltaConfig.ENABLE_NATIVE_WRITE.key` here?
--
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]
