[jira] Commented: (HIVE-704) join order should be based on table sizes

["Zheng Shao (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/HIVE-704?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12737365#action_12737365 ] 

Zheng Shao commented on HIVE-704:
---------------------------------

I think we can do this in two steps:

1. Allow users to give a hint on which table is the "large" table that needs to have the largest tag;
2. Let hive automatically figure out the "large" table.

1 will help us solve the following problem:
A usual case for join is like this: "fact_table JOIN dim_table_1 JOIN dim_table_2 JOIN dim_table_3". Currently the only way to do that in Hive is to use sub queries for each join (since the "large" table needs to be on the right).
With 1 users don't need to rewrite the query using sub queries, but just need to give a hint.


> join order should be based on table sizes
> -----------------------------------------
>
>                 Key: HIVE-704
>                 URL: https://issues.apache.org/jira/browse/HIVE-704
>             Project: Hadoop Hive
>          Issue Type: Improvement
>          Components: Query Processor
>            Reporter: Namit Jain
>
> Currently, hive uses the same order as specified by the user in the query.
> It would be great if that order can be based on the size of tables - largest table should be outermost table.

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