[HIVE-2284] bucketized map join should allow join key as a superset of bucketized columns

• Type: Improvement
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 0.8.0
• Hadoop Flags: Reviewed
• Assignee: Ning Zhang
• Reporter: Ning Zhang
• Created: 15/Jul/11 08:06
• Updated: 16/Dec/11 23:55
• Resolved: 16/Jul/11 06:42

Description

Currently bucketized mapjoin only allow the join keys being exactly the same as bucketized columns. This is too restrictive and is missing some optimization opportunities.

If tables S and T are both bucketized on column A with the same # of buckets, and the query is something like:

<code>
SELECT /*+ MAPJOIN (S) */ ...
FROM S join T
ON (S.A = T.A AND S.B = T.B)
<code>

We should allow bucketized mapjoin since it's straightforward that bucket 1 from S join with bucket 2 from T on such join condition must be empty.