[
https://issues.apache.org/jira/browse/HIVE-750?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12751110#action_12751110
]
Prasad Chakka commented on HIVE-750:
------------------------------------
predicate pushdown was optimization before the new pruner came along so this is
not a bug but a lack of feature :)
patch looks good but since a new child filter operator is created every time
there is a non-deterministic predicate in an expression, there will be two
identical filter operators if there are two non-deterministic are encountered
while walking down the tree.
a better option would be
1) merge the children's pushdown preds with the current filter op's preds
2) run the ppd algorithm to extract pushdown predicates
3) if step 2 does not return an empty set then skip to step 6 (new)
4) if the current op's child is a filter operator then skip to step 6 (new)
5) create a new filter op with the child's pushdown preds (new)
6) pass the results of step2 to current op's parent
so the above algorithm would create a new filter-op whenever a predicate can't
be pushed down. this can help in case multiple sub-queries where some of the
predicates can be pushed down one join but not all the joins in the query.
> new partitionpruner does not work with test mode
> ------------------------------------------------
>
> Key: HIVE-750
> URL: https://issues.apache.org/jira/browse/HIVE-750
> Project: Hadoop Hive
> Issue Type: Bug
> Components: Query Processor
> Reporter: Namit Jain
> Assignee: Namit Jain
> Priority: Critical
> Attachments: hive.750.1.patch
>
>
> set hive.test.mode=true;
> the new partition pruner does not work
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.