hadoop-hive-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Joydeep Sen Sarma (JIRA)" <j...@apache.org>
Subject [jira] Created: (HIVE-218) predicate on partitioning column is ignored in many places
Date Wed, 07 Jan 2009 20:04:44 GMT
predicate on partitioning column is ignored in many places
----------------------------------------------------------

                 Key: HIVE-218
                 URL: https://issues.apache.org/jira/browse/HIVE-218
             Project: Hadoop Hive
          Issue Type: Bug
          Components: Query Processor
            Reporter: Joydeep Sen Sarma
            Priority: Blocker


We tried two queries yesterday that bought up several problems:

1. predicate on partitioning column within a join clause was ignored:

FROM (FROM xxx a SELECT a.xx, a.yy, a.ds WHERE a.ds=2009-01-05 UNION ALL FROM yyy SELECT b.xx,
b.yy, b.ds WHERE b.ds=2009-01-05 UNION ALL FROM zzz c SELECT c.xx, c.yy, c.ds WHERE c.ds=2009-01-05)
d JOIN aaa e ON (d.xx=e.xx AND e.ds=2009-01-05) INSERT OVERWRITE TABLE ...

the plan tried to scan *all* partitions!

2. predicate on partitioning clause inside insert clause was ignored (we took the previous
query and moved the partition filter to the insert statement)

FROM (FROM xxx a SELECT a.xx, a.yy, a.ds WHERE a.ds=2009-01-05 UNION ALL FROM yyy SELECT b.xx,
b.yy, b.ds WHERE b.ds=2009-01-05 UNION ALL FROM zzz c SELECT c.xx, c.yy, c.ds WHERE c.ds=2009-01-05)
d JOIN aaa e ON (d.xx=e.xx ) INSERT OVERWRITE TABLE ... WHERE e.ds=2009-01-05; 

the plan again tried to scan *all* partitions

the really bad thing is that we were able to detect this problem only because of metastore
inconsistencies - otherwise - we would have merrily scanned all the data. This is really critical
to get fixed - because this means that we may actually be scanning tons of unnecessary data
in production.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


Mime
View raw message