[
https://issues.apache.org/jira/browse/CALCITE-2801?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16753318#comment-16753318
]
Hequn Cheng commented on CALCITE-2801:
--------------------------------------
Hi [~julianhyde] , thanks a lot for the tips. It seems that we can only use
RelBuilder.convert to convert to a new rowtype with the same FieldCount. While,
in order to fix this problem, we have to add a Project with a new rowtype with
different FieldCount. (Probably we should use RelOptUtil.createProject() ?)
I have created a pull request. Would be great if you can take a look. Thank you.
> Check input type in AggregateUnionAggregateRule when remove the bottom
> Aggregate
> --------------------------------------------------------------------------------
>
> Key: CALCITE-2801
> URL: https://issues.apache.org/jira/browse/CALCITE-2801
> Project: Calcite
> Issue Type: Bug
> Components: core
> Reporter: Hequn Cheng
> Assignee: Hequn Cheng
> Priority: Major
>
> In {{AggregateUnionAggregateRule}}, we may remove the bottom Aggregate and
> use it's input as the new input of Union directly. However, the bottom
> Aggregate and it's input may not share the same row type. As Union requires
> all inputs with a same row type, once we remove the bottom Aggregate, an
> exception will be thrown.
> The exception can be reproduced by the following test:
> {code:java}
> @Test public void testPullAggregateThroughUnion() {
> HepProgram program = new HepProgramBuilder()
> .addRuleInstance(AggregateProjectMergeRule.INSTANCE)
> .addRuleInstance(AggregateUnionAggregateRule.INSTANCE)
> .build();
> final String sql = "select job, deptno from"
> + " (select job, deptno from emp as e1"
> + " group by job, deptno"
> + " union all"
> + " select job, deptno from emp as e2"
> + " group by job, deptno)"
> + " group by job, deptno";
> sql(sql).with(program).check();
> }
> {code}
> {code:java}
> java.lang.IllegalArgumentException: Cannot compute compatible row type for
> arguments to set op: RecordType(VARCHAR(10) JOB, INTEGER DEPTNO),
> RecordType(INTEGER EMPNO, VARCHAR(20) ENAME, VARCHAR(10) JOB, INTEGER MGR,
> TIMESTAMP(0) HIREDATE, INTEGER SAL, INTEGER COMM, INTEGER DEPTNO, BOOLEAN
> SLACKER)
> at org.apache.calcite.rel.core.SetOp.deriveRowType(SetOp.java:111)
> at
> org.apache.calcite.rel.AbstractRelNode.getRowType(AbstractRelNode.java:222)
> at
> org.apache.calcite.tools.RelBuilder$Frame.<init>(RelBuilder.java:2378)
> at
> org.apache.calcite.tools.RelBuilder$Frame.<init>(RelBuilder.java:2363)
> at org.apache.calcite.tools.RelBuilder.push(RelBuilder.java:260)
> at org.apache.calcite.tools.RelBuilder.setOp(RelBuilder.java:1559)
> at org.apache.calcite.tools.RelBuilder.union(RelBuilder.java:1579)
> at org.apache.calcite.tools.RelBuilder.union(RelBuilder.java:1569)
> at
> org.apache.calcite.rel.rules.AggregateUnionAggregateRule.onMatch(AggregateUnionAggregateRule.java:130)
> at
> org.apache.calcite.plan.AbstractRelOptPlanner.fireRule(AbstractRelOptPlanner.java:319)
> at org.apache.calcite.plan.hep.HepPlanner.applyRule(HepPlanner.java:559)
> at
> org.apache.calcite.plan.hep.HepPlanner.applyRules(HepPlanner.java:418)
> at
> org.apache.calcite.plan.hep.HepPlanner.executeInstruction(HepPlanner.java:255)
> at
> org.apache.calcite.plan.hep.HepInstruction$RuleInstance.execute(HepInstruction.java:127)
> at
> org.apache.calcite.plan.hep.HepPlanner.executeProgram(HepPlanner.java:214)
> at
> org.apache.calcite.plan.hep.HepPlanner.findBestExp(HepPlanner.java:201)
> at
> org.apache.calcite.test.RelOptTestBase.checkPlanning(RelOptTestBase.java:206)
> at
> org.apache.calcite.test.RelOptTestBase$Sql.check(RelOptTestBase.java:334)
> at
> org.apache.calcite.test.RelOptTestBase$Sql.check(RelOptTestBase.java:317)
> at org.apache.calcite.test.RelOptRulesTest.testPullAggregateThroughUnion
> {code}
> I see two option to solve the problem:
> - Check the input type in {{AggregateUnionAggregateRule}} and exit the rule
> directly.
> - Add a Project with a same row type to the bottom Aggregate.
> What do you guys think? Thanks a lot.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
