hadoop-hive-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "He Yongqiang (JIRA)" <j...@apache.org>
Subject [jira] Commented: (HIVE-1266) partitioning pruing should be more intelligent
Date Tue, 23 Mar 2010 18:34:27 GMT

    [ https://issues.apache.org/jira/browse/HIVE-1266?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12848837#action_12848837
] 

He Yongqiang commented on HIVE-1266:
------------------------------------

Discussed with Namit, and the idea here is to generate a probe pruner expr to see if the partition
predicate has effect or not.
The method to generate the probe pruner expr is to 1) set the partition predicate to be always
false, 2) set non-partition predicate always be true. 

For example, 
"ds='2009' or key>20" will "false or true", and the result will be true. So the partition
predicate has no effect.


> partitioning pruing should be more intelligent
> ----------------------------------------------
>
>                 Key: HIVE-1266
>                 URL: https://issues.apache.org/jira/browse/HIVE-1266
>             Project: Hadoop Hive
>          Issue Type: Bug
>          Components: Query Processor
>            Reporter: Namit Jain
>            Assignee: He Yongqiang
>         Attachments: hive-1266.1.patch
>
>
> Copying the mail from Adam:
> A badly written query:
> select .. from T where partitioning_colmn = 'p1' AND c1 = 100 or c2 = 200
> ...is a command I just foolishly ran: I should have put the disjunction in parentheses.
> But the command actually touched every partition of T without a warning. Is that a bug?
When we force people to state a partition predicate, it seems we are just looking for it to
be referenced in the WHERE or ON clause, when maybe we should be looking at conjuncts?

-- 
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