[
https://issues.apache.org/jira/browse/DRILL-5457?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16027215#comment-16027215
]
ASF GitHub Bot commented on DRILL-5457:
---------------------------------------
Github user paul-rogers commented on a diff in the pull request:
https://github.com/apache/drill/pull/822#discussion_r118811633
--- Diff:
exec/java-exec/src/main/java/org/apache/drill/exec/physical/impl/aggregate/HashAggBatch.java
---
@@ -136,15 +136,21 @@ public IterOutcome innerNext() {
return IterOutcome.NONE;
}
- if (aggregator.buildComplete() && !aggregator.allFlushed()) {
- // aggregation is complete and not all records have been output yet
- return aggregator.outputCurrentBatch();
+ // if aggregation is complete and not all records have been output yet
+ if (aggregator.buildComplete() ||
+ // or: 1st phase need to return (not fully grouped) partial output
due to memory pressure
+ aggregator.earlyOutput()) {
+ // then output the next batch downstream
+ IterOutcome out = aggregator.outputCurrentBatch();
--- End diff --
Since `HashAggregator` is not an operator executor (AKA record batch), it
does not have to follow the iterator protocol and use the `IterOutcome` enum.
Instead, you can define your own. You won't need the `OK_NEW_SCHEMA`,
`OUT_OF_MEMORY`, `FAIL` or `NOT_YET` values. All you seem to need is `OK`,
`NONE` and `RESTART`.
This approach will avoid the need to change the `IterOutcome` enum and
export your states to all of the Drill iterator protocol.
Did something similar in Sort for the iterator class that returns either
in-memory or merged spilled batches.
> Support Spill to Disk for the Hash Aggregate Operator
> -----------------------------------------------------
>
> Key: DRILL-5457
> URL: https://issues.apache.org/jira/browse/DRILL-5457
> Project: Apache Drill
> Issue Type: Improvement
> Components: Execution - Relational Operators
> Affects Versions: 1.10.0
> Reporter: Boaz Ben-Zvi
> Assignee: Boaz Ben-Zvi
> Fix For: 1.11.0
>
>
> Support gradual spilling memory to disk as the available memory gets too
> small to allow in memory work for the Hash Aggregate Operator.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
