sohami commented on a change in pull request #1504: DRILL-6792: Find the right
probe side fragment wrapper & fix DrillBuf…
URL: https://github.com/apache/drill/pull/1504#discussion_r226731921
##########
File path:
exec/java-exec/src/main/java/org/apache/drill/exec/planner/physical/visitor/RuntimeFilterVisitor.java
##########
@@ -134,8 +142,15 @@ private RuntimeFilterDef
generateRuntimeFilter(HashJoinPrel hashJoinPrel) {
List<BloomFilterDef> bloomFilterDefs = new ArrayList<>();
//find the possible left scan node of the left join key
- GroupScan groupScan = null;
+ ScanPrel probeSideScanPrel = null;
RelNode left = hashJoinPrel.getLeft();
+ RelNode right = hashJoinPrel.getRight();
+ ExchangePrel exchangePrel = findRightExchangePrel(right);
+ if (exchangePrel == null) {
+ //Does not support the single fragment mode ,that is the right build side
+ //can only be BroadcastExchangePrel or HashToRandomExchangePrel
+ return null;
+ }
List<String> leftFields = left.getRowType().getFieldNames();
List<Integer> leftKeys = hashJoinPrel.getLeftKeys();
RelMetadataQuery metadataQuery = left.getCluster().getMetadataQuery();
Review comment:
For line 160 I wanted to understand why are we trying to get ScanPrel node
for each left key ? Won't all the left keys be coming from same ScanPrel
considering we are not handling blocking operator scenario ? Basically I am
trying to understand when there can be 2 source ScanPrel for set of left keys
to a HashJoinPrel.
Also even if that is the case where we can have multiple source scan prel's
then on line 188 we are only using last known ScanPrel node
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
