LakshSingla commented on code in PR #16854:
URL: https://github.com/apache/druid/pull/16854#discussion_r1715508754
##########
extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/querykit/WindowOperatorQueryKit.java:
##########
@@ -105,8 +108,20 @@ public QueryDefinition makeQueryDefinition(
final int firstStageNumber = Math.max(minStageNumber,
queryDefBuilder.getNextStageNumber());
final WindowOperatorQuery queryToRun = (WindowOperatorQuery)
originalQuery.withDataSource(dataSourcePlan.getNewDataSource());
- final int maxRowsMaterialized;
+ // Get segment granularity from query context, and create ShuffleSpec and
RowSignature to be used for the final window stage.
+ final Granularity segmentGranularity =
QueryKitUtils.getSegmentGranularityFromContext(jsonMapper,
queryToRun.getContext());
+ final ClusterBy finalWindowClusterBy =
QueryKitUtils.clusterByWithSegmentGranularity(ClusterBy.none(),
segmentGranularity);
+ ShuffleSpec finalWindowStageShuffleSpec =
resultShuffleSpecFactory.build(finalWindowClusterBy, false);
+ if (Objects.equals(segmentGranularity, Granularities.ALL)) {
+ finalWindowStageShuffleSpec = MixShuffleSpec.instance();
+ }
Review Comment:
This seems wrong. We should still be using the `resultShuffleSpecFactory` to
create the final shuffle spec. If the windowing stage is an input to the
segment generator, then the window processor would create one large partition,
while we expect the segments to be sized correctly. Am I missing the reason for
this additional check? And if so, can you please add a comment stating why this
is required.
--
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]
