You could load the historical data as flink state and then look up the state with the key derived from input record. That should serve like a join in relational world.
You may also want to think about keeping the writes and querying isolated. Especially if your windows are going to be long (eg cash transactions for last 6 months in your example) and you need your data to be persistent long term, having a durable store outside of Flink will really help. Flink state feature is really nice but I wouldn’t view it as a long term durable storage like a no-sql store or a relational db like oracle. Thanks Ankit From: Tim Stearn <tim.ste...@sas.com> Date: Friday, June 23, 2017 at 3:59 PM To: "user@flink.apache.org" <user@flink.apache.org> Subject: Appending Windowed Aggregates to Events Hello All, I’m *very* new to Flink. I read through the documentation and played with some sample code, but I’m struggling to get started with my requirements. We want to use Flink to maintain windowed aggregates as part of a transaction monitoring application. These would use sliding window definitions. An example would be: “Total amount for CASH transactions in the last 5 days”. Here’s what I need my Flink application to do: 1. Prepare for transaction processing by reading historical aggregates and building windows 2. For each new transaction: a. Update the windowed aggregate with the new transaction data b. Find the window that matches the incoming time stamp and add the aggregate value to the transaction c. Send enhanced transaction (original fields + aggregates from matching window) to downstream processor via RabbitMQ or Kafka sink For every transaction coming in, I want one (and only one) output that contains the original transaction fields plus the aggregates. I see how to do the code to create the window assigner and the code that incrementally maintains the aggregates. I’m not sure how I could join this back to the original transaction record, appending the aggregate values from the window that matches the transaction date stamp. This seems like a join of some kind to me, but I don’t know how to implement in in Flink. I’m hoping someone could reply with some simple code (or even pseudo code) to get me started on the “join” part of the above data flow. Please let me know if I need to clarify. Thanks, Tim Stearn
