Github user fhueske commented on a diff in the pull request:
https://github.com/apache/flink/pull/3715#discussion_r125456339
--- Diff:
flink-libraries/flink-table/src/main/scala/org/apache/flink/table/runtime/join/WindowJoinUtil.scala
---
@@ -162,24 +135,26 @@ object JoinUtil {
(l._1 ++ r._1, l._2 ++ r._2)
}
case c: RexCall =>
- val timeIndicators = analyzeSingleConditionTerm(c, leftFieldCount,
inputType)
+ val timeIndicators = extractTimeIndicatorAccesses(c,
leftFieldCount, inputType)
timeIndicators match {
case Seq() =>
(Seq(), Seq(c))
case Seq(v1, v2) =>
if (v1._1 != v2._1) {
- throw new TableException("The time indicators for each input
should be the same.")
+ throw new TableException(
+ "Both time attributes in a join condition must be of the
same type.")
}
if (v1._2 == v2._2) {
- throw new TableException("Both input's time indicators is
needed.")
+ throw new TableException("Time-based join conditions " +
+ "must reference the time attribute of both input tables.")
}
(Seq((v1._1, v1._2, c)), Seq())
--- End diff --
I think the way the `boolean` field is derived is a bit dangerous. We
assume here that the `RexCall` is a comparison where the first operand is the
left term and the second operand is the right term but it could also be any
other call (with an arbitrary number of operands) that returns a boolean.
Although we later check that the `RexCall` is a comparison, I think we should
do this here.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
