HeartSaVioR edited a comment on pull request #35521: URL: https://github.com/apache/spark/pull/35521#issuecomment-1039922686
I agree about the problem description, but I'd like to see more thoughtful solution. Suppose we have perfect watermark with no delay allowance, and there is no event being out of order, then streaming dedup will do nothing on deduplication because effectively it will register the row in the state and evict it immediately. This will happen if you use watermark but use it like the semantic of "processing time". Personally, for this case, applying TTL against state row would be more sense to me. If we don't want to enforce watermark for the functionality, then we will end up with wall/processing time for TTL which may fall into indeterministic result, but setting TTL as huge interval like 2 hours would be acceptable tolerating such behavior. If we want to enforce watermark to let TTL work (TTL working with event time column), we may even produce deterministic result except late events. Reference: https://nightlies.apache.org/flink/flink-docs-master/docs/dev/table/sql/queries/select-distinct/ https://nightlies.apache.org/flink/flink-docs-master/docs/dev/table/config/#table-exec-state-ttl Note that Flink only supports processing time (wall time) semantic for state TTL, if I understand correctly. -- 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]
