Hi, I’m testing a Spark Runner Streaming Pipeline that combines stateful operations with GroupByKey operation (using org.apache.beam.runners.spark.stateful.SparkGroupAlsoByWindowViaWindowSet).
During testing, I’ve observed that GroupByKey occasionally emits empty iterators. I’m wondering if this is intended behavior? I’ve modified the code to prevent empty iterator emissions, and it seems to work well in my tests. If this isn’t intended behavior, would it be appropriate for me to submit a PR with my modifications? Thanks. Best regards.
