[
https://issues.apache.org/jira/browse/CALCITE-7237?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18031266#comment-18031266
]
Ian Bertolacci commented on CALCITE-7237:
-----------------------------------------
We failed the assertion this with the following:
{code:java}
ImmutableOurSpecialTopSecretRule.Config.builder().build()
.withOperandSupplier(root ->
root.operand(Join.class)
.predicate(join -> ...)
.unorderedInputs(
operand1 -> operand1.operand(RelNode.class).predicate(node ->
...).anyInputs(),
operand2 -> operand1.operand(Aggregate.class).predicate(aggr ->
...).anyInputs()
)
)
{code}
And the trace was:
{code:java}
Caused by: java.lang.AssertionError
at
org.apache.calcite.plan.RelOptRuleOperand.<init>(RelOptRuleOperand.java:121)
at
org.apache.calcite.plan.RelRule$OperandDetailBuilderImpl.done(RelRule.java:287)
at
org.apache.calcite.plan.RelRule$OperandDetailBuilderImpl.unorderedInputs(RelRule.java:324)
at
OurSpecialTopSecretRule$Config.lambda$static$5(OurSpecialTopSecretRule.java:98)
{code}
The idea is that we should match a join with these operands, which can appear
in either order.
> RelOptRuleOperand Assertion error with UNORDERED child policy when more than
> 1 child
> ------------------------------------------------------------------------------------
>
> Key: CALCITE-7237
> URL: https://issues.apache.org/jira/browse/CALCITE-7237
> Project: Calcite
> Issue Type: Bug
> Reporter: Ian Bertolacci
> Priority: Major
>
> [In RelOptRuleOperand constructor, there is an assertion that if the policy
> is UNORDERED, there must be exactly one
> child.|https://github.com/apache/calcite/blob/main/core/src/main/java/org/apache/calcite/plan/RelOptRuleOperand.java#L124-L125]
> However, there are many paths to this condition which claim (both literally
> in comments, or through accepting multiple operands) that there *can* be
> multiple children.
> -
> [RelRule.OperandDetailBuilder.unorderedInputs|https://calcite.apache.org/javadocAggregate/org/apache/calcite/plan/RelRule.OperandDetailBuilder.html#unorderedInputs(org.apache.calcite.plan.RelRule.OperandTransform...)]
> -
> [RelRule.OperandDetailBuilderImpl.unorderedInputs|https://github.com/apache/calcite/blob/92a1028d65efc3005eb22c3def97adefd9e8f2fc/core/src/main/java/org/apache/calcite/plan/RelRule.java#L319-L324]
> -
> [HepPlanner|https://github.com/apache/calcite/blob/main/core/src/main/java/org/apache/calcite/plan/hep/HepPlanner.java#L616]
> It seems like theres a contradiction here.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
