ilooner commented on a change in pull request #1606: DRILL-6845: Semi-Hash-Join
to skip incoming build duplicates, automatically stop skipping if too few
URL: https://github.com/apache/drill/pull/1606#discussion_r248151688
##########
File path:
exec/java-exec/src/main/java/org/apache/drill/exec/physical/impl/join/HashJoinBatch.java
##########
@@ -959,14 +964,40 @@ public IterOutcome executeBuildPhase() throws
SchemaChangeException {
numPartitions,
RECORDS_PER_BATCH,
RECORDS_PER_BATCH,
- maxBatchSize,
- maxBatchSize,
+ maxBatchRowCount,
+ maxBatchRowCount,
batchMemoryManager.getOutputBatchSize(),
HashTable.DEFAULT_LOAD_FACTOR);
if (spilledState.isFirstCycle() && doMemoryCalculation) {
// Do auto tuning
- buildCalc = partitionNumTuning(maxBatchSize, buildCalc);
+ buildCalc = partitionNumTuning(maxBatchRowCount, buildCalc);
+ }
+ if ( semiSkipDuplicates ) {
+ // in case of a Semi Join skippinging duplicates, use a "spill
control" calc
+ // (may revert back to the buildCalc if the code decides to stop
skipping)
+ currentCalc = new HashJoinSpillControlImpl(allocator,
RECORDS_PER_BATCH,
+ (int)
context.getOptions().getOption(ExecConstants.HASHJOIN_MIN_BATCHES_IN_AVAILABLE_MEMORY_VALIDATOR),
+ batchMemoryManager, context);
+
+ // calculates the max number of partitions possible
+ if ( spilledState.isFirstCycle() && doMemoryCalculation ) {
+ currentCalc.initialize(spilledState.isFirstCycle(), true, // TODO
Fix after growing hash values bug fixed
+ buildBatch,
+ probeBatch,
+ buildJoinColumns,
+ probeSideIsEmpty.booleanValue(),
+ allocator.getLimit(),
+ numPartitions,
+ RECORDS_PER_BATCH,
+ RECORDS_PER_BATCH,
+ maxBatchRowCount,
+ maxBatchRowCount,
+ batchMemoryManager.getOutputBatchSize(),
+ HashTable.DEFAULT_LOAD_FACTOR);
+
+ numPartitions = currentCalc.getNumPartitions();
Review comment:
Make this the `numPartitions = min(numPartitions,
currentCalc.getNumPartitions())`
----------------------------------------------------------------
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
