[
https://issues.apache.org/jira/browse/HIVE-16775?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16031339#comment-16031339
]
Ashutosh Chauhan commented on HIVE-16775:
-----------------------------------------
While I agree throwing exception and failing query is bad, but we will miss a
huge optimization opportunity if we don't push always false filter past
aggregate. Imagine doing billion row Gby and throwing that immediately away
since there is a false filter immediately after. We ought to do better.
There are 2 options as I see: a) Override Calcite rule and don't generate
values clause in rule. Perhaps, replace with select operator.
b) We already have values clause in our grammar, so on AST we can generate
values token with empty set. Then during genPlan() we can
genTablePlan(DUMMY_TABLE, qb) for values.
> Skip HiveFilterAggregateTransposeRule when filter is always false
> -----------------------------------------------------------------
>
> Key: HIVE-16775
> URL: https://issues.apache.org/jira/browse/HIVE-16775
> Project: Hive
> Issue Type: Sub-task
> Reporter: Pengcheng Xiong
> Assignee: Pengcheng Xiong
> Attachments: HIVE-16775.01.patch
>
>
> query4.q,query74.q
> {code}
> [7e490527-156a-48c7-aa87-8c80093cdfa8 main] ql.Driver: FAILED:
> NullPointerException null
> java.lang.NullPointerException
> at
> org.apache.hadoop.hive.ql.optimizer.calcite.translator.ASTConverter$QBVisitor.visit(ASTConverter.java:457)
> at org.apache.calcite.rel.RelVisitor.go(RelVisitor.java:61)
> at
> org.apache.hadoop.hive.ql.optimizer.calcite.translator.ASTConverter.convert(ASTConverter.java:110)
> at
> org.apache.hadoop.hive.ql.optimizer.calcite.translator.ASTConverter.convertSource(ASTConverter.java:393)
> at
> org.apache.hadoop.hive.ql.optimizer.calcite.translator.ASTConverter.convert(ASTConverter.java:115)
> {code}
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
