[GitHub] [spark] alex-balikov commented on a diff in pull request #38405: [SPARK-40925][SQL][SS] Fix stateful operator late record filtering

[GitBox <gi...@apache.org>]

alex-balikov commented on code in PR #38405:
URL: https://github.com/apache/spark/pull/38405#discussion_r1007234930


##########
sql/catalyst/src/main/scala/org/apache/spark/sql/internal/SQLConf.scala:
##########
@@ -1941,6 +1941,22 @@ object SQLConf {
       .booleanConf
       .createWithDefault(true)
 
+  val STATEFUL_OPERATOR_ALLOW_MULTIPLE =
+    buildConf("spark.sql.streaming.statefulOperator.allowMultiple")
+      .internal()
+      .doc("When true, multiple stateful operators are allowed to be present in a streaming " +
+        "pipeline. The support for multiple stateful operators introduces a minor (semantically " +
+        "correct) change in respect to late record filtering - late records are detected and " +
+        "filtered in respect to the watermark from the previous microbatch instead of the " +
+        "current one. This is a behavior change for Spark streaming pipelines and we allow " +

Review Comment:
   Currently it is the same watermark passed to all operators. The issue is if anyone has nit tests which check exactly what records are filtered with carefully constructed batches and Trigger.Once - such tests can detect the change in behavior and fail.



-- 
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: reviews-unsubscribe@spark.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org


---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org