[
https://issues.apache.org/jira/browse/HIVE-15619?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15901422#comment-15901422
]
Jesus Camacho Rodriguez commented on HIVE-15619:
------------------------------------------------
[~nishantbangarwa], I took a quick look at the patch.
Instead of using Hive default rules, can we create specific Druid ones that
match the DruidQuery at the bottom, e.g., Project-Filter-DruidQuery, as we do
for instance in DruidRules.PROJECT_SORT rule? That will make the optimization
stage more efficient, since we will get only a few matches and bound the search
space.
This might need changes in Calcite itself, e.g., in the constructor of the
_ProjectFilterTransposeRule.java_, but we are on time to include them in 1.12,
and we will upgrade Hive to that version shortly after it is released.
> Column pruner should handle DruidQuery
> --------------------------------------
>
> Key: HIVE-15619
> URL: https://issues.apache.org/jira/browse/HIVE-15619
> Project: Hive
> Issue Type: Sub-task
> Components: Druid integration
> Affects Versions: 2.2.0
> Reporter: Jesus Camacho Rodriguez
> Assignee: Nishant Bangarwa
> Attachments: HIVE-15619.patch
>
>
> Even when we cannot push any operator into Druid, we might be able to prune
> some of the columns that are read from the Druid sources.
> One solution would be to extend the ColumnPruner so it can push the needed
> columns into DruidQuery.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
