[jira] [Commented] (HIVE-5831) filter input files for bucketed tables

["Shanzhong Zhu (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/HIVE-5831?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14060321#comment-14060321 ] 

Shanzhong Zhu commented on HIVE-5831:
-------------------------------------

We are very interested in this feature since it can improve the query performance for bucketed table when filtering bucket column in where clause.

However, in our scenario, in a partitioned table, those old partitions are NOT bucketized, while those new partitions are bucketizied. This is because out table was created without buckets at the beginning, and a few months later, we added bucketing (ALTER TABLE table_name CLUSTERED BY (col_name, col_name, ...) INTO num_buckets BUCKETS) but didn't reprocess the old partitions.

In this case, seems this patch cannot work with the old partitions with no buckets. Any insight into this?

Thanks,

> filter input files for bucketed tables
> --------------------------------------
>
>                 Key: HIVE-5831
>                 URL: https://issues.apache.org/jira/browse/HIVE-5831
>             Project: Hive
>          Issue Type: Improvement
>          Components: Query Processor
>            Reporter: Rui Li
>         Attachments: hive-5831.patch
>
>
> When the users query a bucketed table and use the bucketed column in the predicate, only the buckets that satisfy the predicate need to be scanned, thus improving the performance.
> Given a table test:
> CREATE TABLE test (x INT, y STRING) CLUSTERED BY ( x ) INTO 10 BUCKETS;
> The following query only has to scan bucket 5:
> SELECT * FROM test WHERE x=5;



--
This message was sent by Atlassian JIRA
(v6.2#6252)