[
https://issues.apache.org/jira/browse/HIVE-12866?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15291492#comment-15291492
]
Ashutosh Chauhan commented on HIVE-12866:
-----------------------------------------
As a pre-requisite of this _bucket_number_ column introduced by SPDO should be
created as of type ExprNodeColumnDesc. Currently, its incorrectly declared as
ExprNodeConstantDesc.
> Allow ReduceSinkDeDuplication to kick in when there are constant keys
> ----------------------------------------------------------------------
>
> Key: HIVE-12866
> URL: https://issues.apache.org/jira/browse/HIVE-12866
> Project: Hive
> Issue Type: Bug
> Reporter: Hari Sankar Sivarama Subramaniyan
>
> Currently in ReduceSinkDeDuplication.checkExprs() we have the following check
> which prevents the optimization from kicking in as soon as we hit constant
> key. We should lift this restriction as much as possible.
> {code}
> private Integer checkExprs(List<ExprNodeDesc> ckeys, List<ExprNodeDesc>
> pkeys,
> ReduceSinkOperator cRS, ReduceSinkOperator pRS) throws
> SemanticException {
> // If ckeys or pkeys have constant node expressions avoid the merge.
> for (ExprNodeDesc ck : ckeys) {
> if (ck instanceof ExprNodeConstantDesc) {
> return null;
> }
> }
> for (ExprNodeDesc pk : pkeys) {
> if (pk instanceof ExprNodeConstantDesc) {
> return null;
> }
> }
> {code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
