Data-driven windowing defines an aggregation through record values and behavioral conditions rather than fixed time boundaries. A sensor aggregation can start or accept readings when a predicate matches, stop when another predicate rejects a reading, and emit after 10 seconds without a closing or additional high reading. section 10.3
A custom Kafka Streams Processor API implementation typically combines:
- Predicates that decide when to start, include, or stop an aggregation. section 10.3
Sensor readings with unreliable connections
Aggregate qualifying sensor readings and update the aggregation’s start time, end time, readings, and average. If no closing or additional high reading arrives for 10 seconds, clean out and forward the unfinished aggregation. section 10.3
