[
https://issues.apache.org/jira/browse/HIVE-11676?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Sergey Shelukhin updated HIVE-11676:
------------------------------------
Attachment: HIVE-11676.patch
Implements the API in a simplest possible way, SARG evaluation is done in
metastore but still with full deserialization. Phase 2 (separate JIRA) will
include storing only necessary data in separate CF in HBase to avoid
deserialization and pave the way for phase 3 where we can push this to HBase
filter. Given that filters are right now min-max only we can even get rid of
ORC dependency and use standard HBase filters, but it will make pushdown
difficult for future improvement, need to talk to Gopal and Owen/Prasanth
before doing that.
This is a patch on top of umpteen other patches that are not committed, in
particular HIVE-11553 and HIVE-11705, maybe others. Patch management is
becoming PITA.
[~alangates] can you review? :) This particular patch is not super urgent.
> implement metastore API to do file footer PPD
> ---------------------------------------------
>
> Key: HIVE-11676
> URL: https://issues.apache.org/jira/browse/HIVE-11676
> Project: Hive
> Issue Type: Bug
> Reporter: Sergey Shelukhin
> Attachments: HIVE-11676.patch
>
>
> Need to pass on the expression/sarg, extract column stats from footer (at
> write time?) and then apply one to the other. I may file a separate JIRA for
> ORC changes cause that is usually PITA
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
