HeartSaVioR commented on code in PR #38503:
URL: https://github.com/apache/spark/pull/38503#discussion_r1021160270
##########
sql/catalyst/src/test/scala/org/apache/spark/sql/catalyst/analysis/UnsupportedOperationsSuite.scala:
##########
@@ -188,17 +194,26 @@ class UnsupportedOperationsSuite extends SparkFunSuite
with SQLHelper {
expectedMsgs = Seq("Complete"))
// FlatMapGroupsWithState(Update) in streaming with aggregation
- for (outputMode <- Seq(Append, Update, Complete)) {
+ for (outputMode <- Seq(Update, Complete)) {
assertNotSupportedInStreamingPlan(
"flatMapGroupsWithState - flatMapGroupsWithState(Update) on streaming
relation " +
s"with aggregation in $outputMode mode",
TestFlatMapGroupsWithState(
null, att, att, Seq(att), Seq(att), att, null, Update,
isMapGroupsWithState = false, null,
Aggregate(Seq(attributeWithWatermark), aggExprs("c"), streamRelation)),
outputMode = outputMode,
- expectedMsgs = Seq("flatMapGroupsWithState in update mode", "with
aggregation"))
+ expectedMsgs = Seq("Multiple stateful operators", "Update", "Complete"))
}
+ assertNotSupportedInStreamingPlan(
+ "flatMapGroupsWithState - flatMapGroupsWithState(Update) on streaming
relation " +
+ s"with aggregation in Append mode",
+ TestFlatMapGroupsWithState(
+ null, att, att, Seq(att), Seq(att), att, null, Update,
isMapGroupsWithState = false, null,
Review Comment:
This is one of the reason I want to see the allowed operations be more
restrictive rather than saying op A - op B is supported with some modes, op B -
op A is supported with some modes, etc.
The major reason I blocked multiple stateful operators at all was that we
figured out stream-stream outer join followed by another stream-stream outer
join is not working, and I realized that is not only the case. Similar risks
are everywhere and it's really hard to think hard on all matrix.
That said, unfortunately, the case we do not fail in unsupported operation
checker does not always mean we guarantee the case to work properly. We
probably have checked in some cases, but I don't believe it's backed by theory
(otherwise we should have known the limit of global watermark much earlier).
--
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.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]