[
https://issues.apache.org/jira/browse/CALCITE-6193?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17804947#comment-17804947
]
Mou Wu edited comment on CALCITE-6193 at 1/10/24 2:32 AM:
----------------------------------------------------------
[~julianhyde] Current test case is already a minimal test case, the traits of
this test case is:
# At least three common sub-expressions in query.
# The first one can be replaced by child-node of mv, but can not be replaced
by the mv itself.
# The second one and the third one can be rewritten by mv, but can not be
replaced by mv completely, some compensations are needed.
So the third one may be skipped.
was (Author: JIRAUSER290047):
[~julianhyde] Current test case is already a minimal test case, the traits of
this test case is:
# At least three common sub-expressions in query.
# The first one can be replaced by child-node of mv, but can not be replaced
by the mv itself.
# The second one and the third one can be rewritten by mv, but can not be
replaced by mv completely.
So the third one may be skipped.
> SubstitutionVisitor stop trying incorrect subtree because of wrong parent
> relationship when replace query
> ---------------------------------------------------------------------------------------------------------
>
> Key: CALCITE-6193
> URL: https://issues.apache.org/jira/browse/CALCITE-6193
> Project: Calcite
> Issue Type: Bug
> Reporter: Mou Wu
> Assignee: Mou Wu
> Priority: Major
> Labels: pull-request-available
> Fix For: 1.37.0
>
>
> {code:java}
> @Test void testStopTryIncorrectSubtree() {
> final String mv = ""
> + "select \"empid\", \"deptno\"\n"
> + "from \"emps\"\n"
> + "where \"salary\" > 1000\n"
> + "group by \"empid\", \"deptno\"";
> final String query = ""
> + "select t1.\"deptno\"\n"
> + "from (\n"
> + "select \"deptno\"\n"
> + "from \"emps\"\n"
> + "where \"salary\" > 1000\n"
> + "union all\n"
> + "select \"deptno\"\n"
> + "from \"emps\"\n"
> + "where \"salary\" > 1000\n"
> + "group by \"deptno\"\n"
> + ") as t1 inner join (\n"
> + "select \"deptno\"\n"
> + "from \"emps\"\n"
> + "where \"salary\" > 1000\n"
> + "group by \"deptno\"\n"
> + ") as t2 on t1.\"deptno\" = t2.\"deptno\"\n";
> sql(mv, query)
> .checkingThatResultContains(""
> + "LogicalCalc(expr#0..1=[{inputs}], deptno=[$t0])\n"
> + " LogicalJoin(condition=[=($0, $1)], joinType=[inner])\n"
> + " LogicalUnion(all=[true])\n"
> + " LogicalCalc(expr#0..4=[{inputs}], expr#5=[1000],
> expr#6=[>($t3, $t5)], deptno=[$t1], $condition=[$t6])\n"
> + " LogicalTableScan(table=[[hr, emps]])\n"
> + " LogicalAggregate(group=[{1}])\n"
> + " EnumerableTableScan(table=[[hr, MV0]])\n"
> + " LogicalAggregate(group=[{1}])\n"
> + " EnumerableTableScan(table=[[hr, MV0]])"
> ).ok();
> }{code}
> The test case above will fail because the second mv0 not be matched.
> The root cause is that SubstitutionVisitor replace child nodes withÂ
> targetDescendant node itself, not a deep-copy replica, so they may share the
> same node and the same parent node, thus the incorrect parent relationship
> may occur, it will make stopTrying be wrong.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
