[
https://issues.apache.org/jira/browse/DRILL-467?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jinfeng Ni updated DRILL-467:
-----------------------------
Attachment: DRILL-467.1.patch
This patch will support multiple "and" join conditions in merge join operator.
> Merge join should support multipe "and" join conditions
> --------------------------------------------------------
>
> Key: DRILL-467
> URL: https://issues.apache.org/jira/browse/DRILL-467
> Project: Apache Drill
> Issue Type: Bug
> Reporter: Jinfeng Ni
> Assignee: Jinfeng Ni
> Attachments: DRILL-467.1.patch
>
>
> The current MergeJoin operator only supports one single join condition. For
> instance, for the following query:
> "select l.l_partkey, l.l_suppkey, ps.ps_partkey, ps.ps_suppkey "
> + " from cp.`tpch/lineitem.parquet` l join "
> + " cp.`tpch/partsupp.parquet` ps"
> + " on l.l_partkey = ps.ps_partkey and "
> + " l.l_suppkey = ps.ps_suppkey"
> The join conditions are specified by an "and" compound predicates. If we run
> the above query on Drill today, we will see two behaviors:
> 1) if assertion is enabled ( -ea option is specified for java ), then Drill
> will throw assertion error, complaining it can support only single join
> condition.
> 2) if assertion is not enabled, the query will run through. However, only the
> first join condition l.l_partkey = ps.ps_partkey is checked. Therefore, the
> result is not correct.
> We should enhance the current MergeJoin operator to support multiple "and"
> join conditions.
> Some TPCH queries also require to support multiple join conditions.
>
--
This message was sent by Atlassian JIRA
(v6.2#6252)
