[
https://issues.apache.org/jira/browse/CALCITE-6887?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Yu Xu updated CALCITE-6887:
---------------------------
Description:
Currently IN operator in ReduceExpressionsRule would not distinct values, so
need optimize it with distinct values in IN.
for example *in (1,1,2,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,3,1)* transform to *in
(1,2,3)* is better, but currently would be *in
(1,1,2,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,3,1)* without distinct.
Like case follow:
*test case:*
@Test void testReduceExpressionsWithIn()
{ final String sql = "select deptno, sal " + "from emp " + "where deptno in
(1,1,2,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,3,1) ";
sql(sql).withRule(CoreRules.PROJECT_REDUCE_EXPRESSIONS) .check(); }
*before plan:*
LogicalProject(DEPTNO=[$7], SAL=[$5])
LogicalFilter(condition=[IN($7,
{ LogicalValues(tuples=[[
{ 1 }, \{ 1 }, \{ 2 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 },
\{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 3 },
\{ 1 }]])
})])
LogicalTableScan(table=[[CATALOG, SALES, EMP]])
*after plan:*
LogicalProject(DEPTNO=[$7], SAL=[$5])
LogicalFilter(condition=[IN($7,
{ LogicalValues(tuples=[[
{ 1 }
, \{ 1 }, \{ 2 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1
}, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 3 }, \{ 1
}]])
})])
LogicalTableScan(table=[[CATALOG, SALES, EMP]])
was:
Currently IN operator in ReduceExpressionsRule would not distinct values, so
need optimize it with distinct values in IN.
for example *in (1,1,2,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,3,1)* transform to *in
(1,2,3)* is better, but currently would not
*test case:*
@Test void testReduceExpressionsWithIn()
{ final String sql = "select deptno, sal " + "from emp " + "where deptno in
(1,1,2,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,3,1) ";
sql(sql).withRule(CoreRules.PROJECT_REDUCE_EXPRESSIONS) .check(); }
*before plan:*
LogicalProject(DEPTNO=[$7], SAL=[$5])
LogicalFilter(condition=[IN($7,
{ LogicalValues(tuples=[[
{ 1 }, \{ 1 }, \{ 2 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 },
\{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 3 },
\{ 1 }]])
})])
LogicalTableScan(table=[[CATALOG, SALES, EMP]])
*after plan:*
LogicalProject(DEPTNO=[$7], SAL=[$5])
LogicalFilter(condition=[IN($7,
{ LogicalValues(tuples=[[
{ 1 }, \{ 1 }, \{ 2 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 },
\{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 3 },
\{ 1 }]])
})])
LogicalTableScan(table=[[CATALOG, SALES, EMP]])
> In should distinct values with ReduceExpressionRule
> ---------------------------------------------------
>
> Key: CALCITE-6887
> URL: https://issues.apache.org/jira/browse/CALCITE-6887
> Project: Calcite
> Issue Type: Bug
> Components: core
> Affects Versions: 1.38.0
> Reporter: Yu Xu
> Assignee: Yu Xu
> Priority: Major
> Labels: pull-request-available
> Fix For: 1.40.0
>
>
> Currently IN operator in ReduceExpressionsRule would not distinct values, so
> need optimize it with distinct values in IN.
> for example *in (1,1,2,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,3,1)* transform to
> *in (1,2,3)* is better, but currently would be *in
> (1,1,2,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,3,1)* without distinct.
> Like case follow:
> *test case:*
> @Test void testReduceExpressionsWithIn()
> { final String sql = "select deptno, sal " + "from emp " + "where deptno in
> (1,1,2,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,3,1) ";
> sql(sql).withRule(CoreRules.PROJECT_REDUCE_EXPRESSIONS) .check(); }
>
> *before plan:*
> LogicalProject(DEPTNO=[$7], SAL=[$5])
> LogicalFilter(condition=[IN($7,
> { LogicalValues(tuples=[[
> { 1 }, \{ 1 }, \{ 2 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1
> }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{
> 3 }, \{ 1 }]])
> })])
> LogicalTableScan(table=[[CATALOG, SALES, EMP]])
>
> *after plan:*
> LogicalProject(DEPTNO=[$7], SAL=[$5])
> LogicalFilter(condition=[IN($7,
> { LogicalValues(tuples=[[
> { 1 }
> , \{ 1 }, \{ 2 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{
> 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 1 }, \{ 3 },
> \{ 1 }]])
> })])
> LogicalTableScan(table=[[CATALOG, SALES, EMP]])
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
