You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@hive.apache.org by "Szehon Ho (JIRA)" <ji...@apache.org> on 2014/12/20 01:05:15 UTC
[jira] [Updated] (HIVE-8640) Support hints of SMBJoin [Spark
Branch]
[ https://issues.apache.org/jira/browse/HIVE-8640?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Szehon Ho updated HIVE-8640:
----------------------------
Attachment: HIVE-8640.1-spark.patch
This patch refactors the hinted bucketJoin/SMBJoin to be on the same traversal as the other auto-join optimizers. See RB for details.
One note, there is a dependent issue HIVE-9117 that should be committed first to trunk and merged to spark. But including that change here for reference and pre-commit tests.
About tests, re-enabled all the smb_mapjoin.* tests.
While re-alphabetizing the list, I realized that some tests like 'bucket_map_join_spark.*' and 'join_empty' were missing proper comma deliminters, and windowing.q name is corrupt. So these tests might not have been picked up by test framework. Fixing these.
> Support hints of SMBJoin [Spark Branch]
> ---------------------------------------
>
> Key: HIVE-8640
> URL: https://issues.apache.org/jira/browse/HIVE-8640
> Project: Hive
> Issue Type: Sub-task
> Components: Spark
> Reporter: Szehon Ho
> Assignee: Szehon Ho
> Attachments: HIVE-8640.1-spark.patch
>
>
> HIVE-8202 supports conversion of join to SMB Join automatically, which relies on the configuration property: "hive.auto.convert.sortmerge.join.bigtable.selection.policy".
> This task is to support conversion based on map-hints, instead of this policy. As hints are deprecated, this would not be the primary policy in line with MapReduce, but can be available as a backup to achieve feature parity.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)