[
https://issues.apache.org/jira/browse/HIVE-15723?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Remus Rusanu updated HIVE-15723:
--------------------------------
Attachment: HIVE-15723.02.patch
02.patch adds a conf setting hive.cbo.show.warnings (default false) to control
the logging of misisng columns and a test case. Missing stats warnings looks
like this:
{noformat}
No Stats for default@missing_stats_t1, Columns: value
No Stats for default@missing_stats_t2, Columns: key
No Stats for default@missing_stats_t3, Columns: value
{noformat}
Which is exactly the same output as the missing stats logged (to hive.log) in
RelOpHiveTable.updateColStats, as is actually the very same logMsg.
> Hive should report a warning about missing table/column statistics to user.
> ---------------------------------------------------------------------------
>
> Key: HIVE-15723
> URL: https://issues.apache.org/jira/browse/HIVE-15723
> Project: Hive
> Issue Type: Bug
> Components: Query Processor
> Reporter: Remus Rusanu
> Assignee: Remus Rusanu
> Priority: Minor
> Fix For: 2.2.0
>
> Attachments: HIVE-15723.01.patch, HIVE-15723.02.patch
>
>
> Many Hive performance issues are due to missing statistics. Either all, table
> or column statistics are missing. Potentially a new partition has been added
> and customer forgot to gather stats for that partition.
> A simple warning about a table or column missing statistics can be very
> helpful and makes hive more user friendly. Hive already has this information,
> its a matter of printing it out.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
