Zoltan Haindrich created HIVE-21010:
---------------------------------------
Summary: ReduceExpressionRule may oversimplify filter conditions
containing nulls
Key: HIVE-21010
URL: https://issues.apache.org/jira/browse/HIVE-21010
Project: Hive
Issue Type: Bug
Reporter: Zoltan Haindrich
Assignee: Zoltan Haindrich
ReduceExpressionsRule is invoking
[simplifications|https://github.com/apache/calcite/blob/efec74deb80da1708fd42bcc2b57289840869346/core/src/main/java/org/apache/calcite/rel/rules/ReduceExpressionsRule.java#L549]
- by using
[ExprSimplifier|https://github.com/apache/calcite/blob/efec74deb80da1708fd42bcc2b57289840869346/core/src/main/java/org/apache/calcite/rex/RexUtil.java#L2611]
which unfortunately doesn't switch unknownAs mode (because it's a visitor).
Can be reproduced by adding the following test case to RelOptRulesTests
{code}
@Test public void testIncorrectlyRemovedCondition() {
HepProgram program = new HepProgramBuilder()
.addRuleInstance(ReduceExpressionsRule.FILTER_INSTANCE)
.build();
String sql =
// "select * from emp where ( (empno=1 and mgr=1) or (empno=null
and mgr=1) ) is null";
"select * from emp where ( (empno=null and mgr=1) ) is null";
checkPlanning(program, sql);
}
{code}
Plan should retain the condition; right now it incorrectly simplified to false.
{code}
<TestCase name="testIncorrectlyRemovedCondition">
<Resource name="sql">
<![CDATA[select * from emp where ( (empno=null and mgr=1) ) is
null]]>
</Resource>
<Resource name="planBefore">
<![CDATA[
LogicalProject(EMPNO=[$0], ENAME=[$1], JOB=[$2], MGR=[$3], HIREDATE=[$4],
SAL=[$5], COMM=[$6], DEPTNO=[$7], SLACKER=[$8])
LogicalFilter(condition=[IS NULL(AND(=($0, null), =($3, 1)))])
LogicalTableScan(table=[[CATALOG, SALES, EMP]])
]]>
</Resource>
<Resource name="planAfter">
<![CDATA[
LogicalProject(EMPNO=[$0], ENAME=[$1], JOB=[$2], MGR=[$3], HIREDATE=[$4],
SAL=[$5], COMM=[$6], DEPTNO=[$7], SLACKER=[$8])
LogicalValues(tuples=[[]])
]]>
</Resource>
</TestCase>
{code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
