Github user fhueske commented on a diff in the pull request:
https://github.com/apache/flink/pull/5241#discussion_r160259534
--- Diff:
flink-libraries/flink-table/src/test/scala/org/apache/flink/table/runtime/batch/sql/SetOperatorsITCase.scala
---
@@ -278,7 +278,14 @@ class SetOperatorsITCase(
TestBaseUtils.compareResultAsText(results.asJava, expected)
}
+ /**
+ * This test will checks IN for NULLs based on whether COUNT (*) and
COUNT (a) are equal. Due to
+ *
[[org.apache.flink.table.plan.rules.dataSet.DataSetAggregateWithNullValuesRule]]
will
+ * union a NULL row in to input DataSet for non-groupBy agg. That
caused COUNT (*) and COUNT(a)
+ * are not equal. So this test case ignored before FLINK-8355 be fixed.
--- End diff --
This PR introduces a regression. Previously working queries will compute
incorrect results when we merge it.
IMO, we should only merge it if all existing tests pass. The problem is
that FLINK-8355 cannot be easily fixed. We need to add a NULL tuple, because
Flink's DataSet API won't compute any result if the input is empty. Not sure
how this can be resolved.
---
