[
https://issues.apache.org/jira/browse/CALCITE-4958?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
hujiahua updated CALCITE-4958:
------------------------------
Description:
When we using IN-list predicate in where clause and setting the
DEFAULT_IN_SUB_QUERY_THRESHOLD less than the IN-list elements size, IN-list
predicate will converted to Join. And also using dynamic parameters in a VALUES
clause, I found each dynamic parameter converted to a LogicalValues. Too many
dynamic parameters will lead to poor performance
First I set SqlToRelConverter.DEFAULT_IN_SUB_QUERY_THRESHOLD = 2
And When I use dynamic parameters in query like this:
{noformat}
select * from DEPTS where NAME in ( ?, ?, ?)
{noformat}
The IN-list will convert to union three project.
{noformat}
LogicalProject(DEPTNO=[$0], NAME=[$1])
LogicalJoin(condition=[=($1, $2)], joinType=[inner])
LogicalTableScan(table=[[TEST, DEPTS]])
LogicalAggregate(group=[{0}])
LogicalUnion(all=[true])
LogicalProject(EXPR$0=[?0])
LogicalValues(tuples=[[{ 0 }]])
LogicalProject(EXPR$0=[?1])
LogicalValues(tuples=[[{ 0 }]])
LogicalProject(EXPR$0=[?2])
LogicalValues(tuples=[[{ 0 }]])
{noformat}
But if I not use dynamic parameters in query like this:
{noformat}
select * from DEPTS where NAME in ( 'a', 'b', 'c')
{noformat}
The IN-list will a LogicalValues, this is what I wanted.
{noformat}
LogicalProject(DEPTNO=[$0], NAME=[$1])
LogicalJoin(condition=[=($1, $2)], joinType=[inner])
LogicalTableScan(table=[[TEST, DEPTS]])
LogicalAggregate(group=[{0}])
LogicalValues(tuples=[[{ 'a' }, { 'b' }, { 'c' }]])
{noformat}
Here is my test: // I set DEFAULT_IN_SUB_QUERY_THRESHOLD = 2
{code:java}
final String sql = "select * from \"TEST\".\"DEPTS\" where \"NAME\" in (
?, ?, ?)";
final PreparedStatement statement2 =
calciteConnection.prepareStatement(sql);
statement2.setString(1, "Sales");
statement2.setString(2, "Sales2");
statement2.setString(3, "Sales3");
final ResultSet resultSet1 = statement2.executeQuery();
{code}
was:
When using dynamic parameters in a VALUES clause (e.g. "x IN (?, ?, ?)") ,
First I set SqlToRelConverter.DEFAULT_IN_SUB_QUERY_THRESHOLD = 2
And When I use dynamic parameters in query like this:
{noformat}
select * from DEPTS where NAME in ( ?, ?, ?)
{noformat}
The IN-list will convert to union three project.
{noformat}
LogicalProject(DEPTNO=[$0], NAME=[$1])
LogicalJoin(condition=[=($1, $2)], joinType=[inner])
LogicalTableScan(table=[[TEST, DEPTS]])
LogicalAggregate(group=[{0}])
LogicalUnion(all=[true])
LogicalProject(EXPR$0=[?0])
LogicalValues(tuples=[[{ 0 }]])
LogicalProject(EXPR$0=[?1])
LogicalValues(tuples=[[{ 0 }]])
LogicalProject(EXPR$0=[?2])
LogicalValues(tuples=[[{ 0 }]])
{noformat}
But if I not use dynamic parameters in query like this:
{noformat}
select * from DEPTS where NAME in ( 'a', 'b', 'c')
{noformat}
The IN-list will a LogicalValues, this is what I wanted.
{noformat}
LogicalProject(DEPTNO=[$0], NAME=[$1])
LogicalJoin(condition=[=($1, $2)], joinType=[inner])
LogicalTableScan(table=[[TEST, DEPTS]])
LogicalAggregate(group=[{0}])
LogicalValues(tuples=[[{ 'a' }, { 'b' }, { 'c' }]])
{noformat}
Here is my test: // I set DEFAULT_IN_SUB_QUERY_THRESHOLD = 2
{code:java}
final String sql = "select * from \"TEST\".\"DEPTS\" where \"NAME\" in (
?, ?, ?)";
final PreparedStatement statement2 =
calciteConnection.prepareStatement(sql);
statement2.setString(1, "Sales");
statement2.setString(2, "Sales2");
statement2.setString(3, "Sales3");
final ResultSet resultSet1 = statement2.executeQuery();
{code}
> Poor performance execute plan when use dynamic parameters in IN-list clause
> ---------------------------------------------------------------------------
>
> Key: CALCITE-4958
> URL: https://issues.apache.org/jira/browse/CALCITE-4958
> Project: Calcite
> Issue Type: Bug
> Reporter: hujiahua
> Priority: Minor
>
> When we using IN-list predicate in where clause and setting the
> DEFAULT_IN_SUB_QUERY_THRESHOLD less than the IN-list elements size, IN-list
> predicate will converted to Join. And also using dynamic parameters in a
> VALUES clause, I found each dynamic parameter converted to a LogicalValues.
> Too many dynamic parameters will lead to poor performance
>
> First I set SqlToRelConverter.DEFAULT_IN_SUB_QUERY_THRESHOLD = 2
> And When I use dynamic parameters in query like this:
> {noformat}
> select * from DEPTS where NAME in ( ?, ?, ?)
> {noformat}
> The IN-list will convert to union three project.
> {noformat}
> LogicalProject(DEPTNO=[$0], NAME=[$1])
> LogicalJoin(condition=[=($1, $2)], joinType=[inner])
> LogicalTableScan(table=[[TEST, DEPTS]])
> LogicalAggregate(group=[{0}])
> LogicalUnion(all=[true])
> LogicalProject(EXPR$0=[?0])
> LogicalValues(tuples=[[{ 0 }]])
> LogicalProject(EXPR$0=[?1])
> LogicalValues(tuples=[[{ 0 }]])
> LogicalProject(EXPR$0=[?2])
> LogicalValues(tuples=[[{ 0 }]])
> {noformat}
> But if I not use dynamic parameters in query like this:
> {noformat}
> select * from DEPTS where NAME in ( 'a', 'b', 'c')
> {noformat}
> The IN-list will a LogicalValues, this is what I wanted.
> {noformat}
> LogicalProject(DEPTNO=[$0], NAME=[$1])
> LogicalJoin(condition=[=($1, $2)], joinType=[inner])
> LogicalTableScan(table=[[TEST, DEPTS]])
> LogicalAggregate(group=[{0}])
> LogicalValues(tuples=[[{ 'a' }, { 'b' }, { 'c' }]])
> {noformat}
> Here is my test: // I set DEFAULT_IN_SUB_QUERY_THRESHOLD = 2
> {code:java}
> final String sql = "select * from \"TEST\".\"DEPTS\" where \"NAME\" in
> ( ?, ?, ?)";
> final PreparedStatement statement2 =
> calciteConnection.prepareStatement(sql);
> statement2.setString(1, "Sales");
> statement2.setString(2, "Sales2");
> statement2.setString(3, "Sales3");
> final ResultSet resultSet1 = statement2.executeQuery();
> {code}
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
