Krisztian Kasa created HIVE-28572:
-------------------------------------
Summary: Support Distribute by and Cluster by clauses in CBO
Key: HIVE-28572
URL: https://issues.apache.org/jira/browse/HIVE-28572
Project: Hive
Issue Type: Improvement
Security Level: Public (Viewable by anyone)
Components: CBO
Reporter: Krisztian Kasa
Assignee: Krisztian Kasa
If a query has {{distribute by}} or {{cluster by}} clause CBO is turned off and
only non-CBO optimizations are applied to the query plan.
One impact of not using CBO is that implicit type conversions are not added.
Example:
{code:java}
create table t1 (a string, b int);
insert into t1 values ('2014-03-14 10:10:12', 10);
select * from t1 where a between date_add('2014-03-14', -1) and '2014-03-14'
distribute by a;
{code}
{code:java}
TableScan
alias: t1
filterExpr: a BETWEEN DATE'2014-03-13' AND '2014-03-14'
(type: boolean)
{code}
vs
{code:java}
select * from t1 where a between date_add('2014-03-14', -1) and '2014-03-14'
{code}
{code:java}
TableScan
alias: t1
filterExpr: CAST( a AS DATE) BETWEEN DATE'2014-03-13' AND
DATE'2014-03-14' (type: boolean)
{code}
Moreover, if vectorization is turned off the results of the above queries are
different which leads to data corruption.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
