[
https://issues.apache.org/jira/browse/HIVE-16358?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15974257#comment-15974257
]
Jesus Camacho Rodriguez commented on HIVE-16358:
------------------------------------------------
I think it would depend on whether there is a tangible performance difference
between GroupBy (with OBy + Limit) and TopN, since in that case, user might
choose to have approximate results to get better performance.
About HIVE-16121, I think for the time being we should at least create the flag
to enable/disable support for decimal type in Druid, since using Druid's float
type might lead to loss of precision wrt decimal, however it is still
interesting if user is fine with that.
> Push TOP N to Druid queries
> ---------------------------
>
> Key: HIVE-16358
> URL: https://issues.apache.org/jira/browse/HIVE-16358
> Project: Hive
> Issue Type: Improvement
> Components: Query Planning, Query Processor
> Reporter: Remus Rusanu
> Assignee: Remus Rusanu
>
> After HIVE-15708 promotes Calcite to 1.12, {{TOP N}} queries over Druid
> tables will be implemented by GroupBy on Druid and then (sort and) Limit on
> Hive. The relevant changes on Calcite side are related to CALCITE-1578 and
> steam from fundamental semantic differences between Druid concept of TOP (ie.
> top within a GBy) vs. SQL TOP/LIMIT (ie. global).
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
