[
https://issues.apache.org/jira/browse/HIVE-15046?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15609822#comment-15609822
]
Ashutosh Chauhan commented on HIVE-15046:
-----------------------------------------
In DruidGroupByQueryRecordReader, I think we need more robust handling, lets
say I have 3 dimensions d1, d2, d3. Its possible that d2,d3 is missing in some
segments and query is on d2,d3. Can current code handle that? What about d2 is
missing in segment and query is on d2,d3. Lastly, d2,d3 is missing in segment
and query is on d3.
Other changes look good.
> Multiple fixes for Druid handler
> --------------------------------
>
> Key: HIVE-15046
> URL: https://issues.apache.org/jira/browse/HIVE-15046
> Project: Hive
> Issue Type: Bug
> Components: Druid integration
> Affects Versions: 2.2.0
> Reporter: Jesus Camacho Rodriguez
> Assignee: Jesus Camacho Rodriguez
> Attachments: HIVE-15046.01.patch, HIVE-15046.patch
>
>
> - Druid query type not recognized after Calcite upgrade; introduced by
> HIVE-13316.
> - Fix handling of NULL values for GroupBy queries.
> - Fix handling of dimension/metrics names, as those names in Druid are case
> sensitive.
> - Select Druid query to effectively return no rows when the result is empty
> (previously returning a single row).
> - When it is split, each of the parts of a Select query might return more
> results than threshold; set threshold to max integer in query so we do not
> face this problem.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
