[
https://issues.apache.org/jira/browse/FLINK-2871?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15108313#comment-15108313
]
ASF GitHub Bot commented on FLINK-2871:
---------------------------------------
Github user ChengXiangLi commented on the pull request:
https://github.com/apache/flink/pull/1469#issuecomment-173151704
I did simple regression test based on `HashVsSortMiniBenchmark`, the result
looks like:
Test | Before | After
------ | ------ | --------
testBuildFirst | 6.63s | 6.65s
testBuildSecond | 3.7s | 3.8s
The inner join performance is not influenced by this PR, which fit into my
expectation. There is a flag called `buildsideOuterJoin` in `MutableHashTable`,
all the extra effort only happens while `buildSideOuterJoin` is true.
> Add OuterJoin strategy with HashTable on outer side
> ---------------------------------------------------
>
> Key: FLINK-2871
> URL: https://issues.apache.org/jira/browse/FLINK-2871
> Project: Flink
> Issue Type: New Feature
> Components: Local Runtime, Optimizer
> Affects Versions: 0.10.0
> Reporter: Fabian Hueske
> Assignee: Chengxiang Li
> Priority: Minor
>
> Outer joins are currently supported with two local execution strategies:
> - sort-merge join
> - hash join where the hash table is built on the inner side. Hence, this
> strategy is only supported for left and right outer joins.
> In order to support hash-tables on the outer side, we need a special hash
> table implementation that gives access to all records which have not been
> accessed during the probe phase.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
