[
https://issues.apache.org/jira/browse/CALCITE-5669?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17715312#comment-17715312
]
Julian Hyde commented on CALCITE-5669:
--------------------------------------
Generally I think it's worthwhile to expand the "correlating expression" as
much as possible. (We may already do this; I haven't checked.) In this case,
don't correlate on "{{e.empno}}" but instead correlate on "{{e.empno = null}}".
Then that is an expression that can be more easily constant-reduced.
(One counterpoint to that: in another case (CALCITE-5644) I am dealing
correlated sub-query that has "{{t.a is not distinct from u.b}}", which is
expanded to "{{t.a is null and u.b is null or t.a = t.b}}", and somehow "{{t.a
is null}}" ends up being computed as a separate boolean column in the left
input of the {{Correlate}} operator. Thus the correlation condition is split
between two operators, and it is very hard to perform further simplification.
The answer in this case is probably to keep {{is not distinct from}} intact,
but even so, it emphasizes the importance of choosing the right grain for
correlation variables.)
But even with the best intentions we may end up with trivial {{Correlate}}
operators. Eliminating variables that are not used, and converting
{{Correlate}} operators with no variables into {{Join}} operators, seem to be
simple and efficient rules. So, my vote would be to do that solution.
> Remove trivial correlates from the query plan
> ---------------------------------------------
>
> Key: CALCITE-5669
> URL: https://issues.apache.org/jira/browse/CALCITE-5669
> Project: Calcite
> Issue Type: Improvement
> Components: core
> Reporter: Stamatis Zampetakis
> Assignee: Stamatis Zampetakis
> Priority: Major
> Labels: pull-request-available
>
> Consider the following query correlated query.
> {code:sql}
> select * from emp as e where exists (select 1 from dept as d where e.empno =
> null)
> {code}
> The query basically returns an empty result because {{e.empno = null}} is
> always false.
> The plan for the query after applying the sub-query remove rule is shown
> below:
> {noformat}
> LogicalProject(EMPNO=[$0], ENAME=[$1], JOB=[$2], MGR=[$3], HIREDATE=[$4],
> SAL=[$5], COMM=[$6], DEPTNO=[$7], SLACKER=[$8])
> LogicalProject(EMPNO=[$0], ENAME=[$1], JOB=[$2], MGR=[$3], HIREDATE=[$4],
> SAL=[$5], COMM=[$6], DEPTNO=[$7], SLACKER=[$8])
> LogicalCorrelate(correlation=[$cor0], joinType=[inner],
> requiredColumns=[{0}])
> LogicalTableScan(table=[[CATALOG, SALES, EMP]])
> LogicalAggregate(group=[{0}])
> LogicalProject(i=[true])
> LogicalFilter(condition=[=($cor0.EMPNO, null)])
> LogicalTableScan(table=[[CATALOG, SALES, DEPT]])
> {noformat}
> After applying the reduce expressions rule the filter with the correlated
> condition will become false and the resulting plan would be the following.
> {noformat}
> LogicalProject(EMPNO=[$0], ENAME=[$1], JOB=[$2], MGR=[$3], HIREDATE=[$4],
> SAL=[$5], COMM=[$6], DEPTNO=[$7], SLACKER=[$8])
> LogicalProject(EMPNO=[$0], ENAME=[$1], JOB=[$2], MGR=[$3], HIREDATE=[$4],
> SAL=[$5], COMM=[$6], DEPTNO=[$7], SLACKER=[$8])
> LogicalCorrelate(correlation=[$cor0], joinType=[inner],
> requiredColumns=[{0}])
> LogicalTableScan(table=[[CATALOG, SALES, EMP]])
> LogicalAggregate(group=[{0}])
> LogicalProject(i=[true])
> LogicalValues(tuples=[[]])
> {noformat}
> Observe that now we have a {{LogicalCorrelate}} but there is no real
> correlation in the plan since the correlation variable on the right side
> disappeared. Depending on how rules are applied and which rules are used
> similar "trivial" correlates may appear.
> The goal of this ticket is to provide the means to get rid of them.
> One option would be to add a new rule (e.g., {{CorrelateToJoinRule}}) which
> detects that a correlate does not have correlations in the right side and
> turn the correlation to a join; then we could employ other existing rules
> (such as PruneEmptyRules) for joins and remove the newly created join
> altogether.
> Another option, would be to introduce new pruning rule(s) for correlate
> (similar to those for joins) that will remove the correlate when its input is
> an empty values expression.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
