Thank you Congxian and Fabian. @Fabian: could you please give a bit more details? My understanding is: to pass the context itself and an OutputTag to the KeyedStateFunction parameter of KeyedBroadcastProcessFunction#Context.applyToKeyedState(), and from within that KeyedStateFunction.process() send out the side output. Am I understand your idea correctly?
BTW, I have another question regarding KeyedBroadcastProcessFunction best practice: I am having two streams: Data and Toggle. The stream Toggle is just a keyed boolean stream, being used to filter data from the stream Data. And I am implementing that filter using a simple RichCoFlatMapFunction. Now that I want to export the list of keys which are currently toggled on. Should I (1) have one additional KeyedBroadcastProcessFunction operator (which has Toggle and BroadCast as the input streams), or (2) replace that RichCoFlatMapFunction with a new KeyedBroadcastProcessFunction, which has both functionalities: filter and export? Doing this would require unioning Toggle and Data into one single keyed stream. Thanks and best regards, Averell -- Sent from: http://apache-flink-user-mailing-list-archive.2336050.n4.nabble.com/