srowen commented on a change in pull request #24890: [SPARK-28074][SS] Log warn
message on possible correctness issue for multiple stateful operations in
single query
URL: https://github.com/apache/spark/pull/24890#discussion_r329153425
##########
File path:
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/analysis/UnsupportedOperationChecker.scala
##########
@@ -41,8 +42,57 @@ object UnsupportedOperationChecker {
}
}
- def checkForStreaming(plan: LogicalPlan, outputMode: OutputMode): Unit = {
+ def checkStreamingQueryGlobalWatermarkLimit(
+ plan: LogicalPlan,
+ outputMode: OutputMode,
+ failWhenDetected: Boolean): Unit = {
+ def isStatefulOperationPossiblyEmitLateRows(p: LogicalPlan): Boolean = p
match {
+ case s: Aggregate
+ if s.isStreaming && outputMode == InternalOutputModes.Append => true
+ case Join(left, right, joinType, _, _)
+ if left.isStreaming && right.isStreaming && joinType != Inner => true
+ case f: FlatMapGroupsWithState
+ if f.isStreaming && f.outputMode == OutputMode.Append() => true
+ case _ => false
+ }
+
+ def isStatefulOperation(p: LogicalPlan): Boolean = p match {
+ case s: Aggregate if s.isStreaming => true
+ case _ @ Join(left, right, _, _, _) if left.isStreaming &&
right.isStreaming => true
+ case f: FlatMapGroupsWithState if f.isStreaming => true
+ case d: Deduplicate if d.isStreaming => true
+ case _ => false
+ }
+
+ var loggedWarnMessage = false
+ plan.foreach { subPlan =>
Review comment:
Because you stop when the first problem is detected, how about just a while
loop, so you don't have to keep examining the whole plan? maybe it's OK as it
only saves time in case of a problem, which is rare.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
