[ https://issues.apache.org/jira/browse/FLINK-19668?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17216424#comment-17216424 ]
Danny Chen commented on FLINK-19668: ------------------------------------ Another case is `org.apache.flink.table.planner.plan.batch.sql.agg.WindowAggregateTest.testReturnTypeInferenceForWindowAgg` of blink planner. > Fix the plan regression by Calcite upgrade to 1.26 > -------------------------------------------------- > > Key: FLINK-19668 > URL: https://issues.apache.org/jira/browse/FLINK-19668 > Project: Flink > Issue Type: Bug > Components: Table SQL / Planner > Affects Versions: 1.11.3 > Reporter: Danny Chen > Priority: Major > > See org.apache.flink.table.plan.TimeIndicatorConversionTest.testUnion, > {code:java} > @Test > def testUnion(): Unit = { > val util = streamTestUtil() > val t = util.addTable[(Long, Long, Int)]("MyTable", 'rowtime.rowtime, > 'long, 'int) > val result = t.unionAll(t).select('rowtime) > val expected = unaryNode( > "DataStreamCalc", > binaryNode( > "DataStreamUnion", > streamTableNode(t), > streamTableNode(t), > term("all", "true"), > term("union all", "rowtime, long, int") > ), > term("select", "rowtime") > ) > util.verifyTable(result, expected) > } > {code} > The regression is because the new plan has same cost with the old, thus, the > old was not picked as the best. -- This message was sent by Atlassian Jira (v8.3.4#803005)