How to emit only the final result of a windowed aggregation in Kafka Streams

A windowed aggregation in Kafka Streams is a stateful operation used to perform a "clustering" or "grouping" of values with the same key over a specified time window. A windowed aggregation will generally return intermediate results while a window is still open, but sometimes you may only care about the final result for a given window. In this tutorial, we take an input topic of sensor pressure readings and output only the final count of readings per sensor for each time window.

The key to emitting only final results is in the call to KTable.suppress in the following snippet that builds the topology:

 builder.stream(INPUT_TOPIC, consumedPressure)
                .selectKey((key, value) -> value.id())
                .groupByKey(groupedPressure)
                .windowedBy(windows)
                .count()
                .suppress(Suppressed.untilWindowCloses(unbounded()))
                .toStream()
                .to(OUTPUT_TOPIC, producedCount);

Let's review this snippet line by line.

         .selectKey((key, value) -> value.id())

Aggregations must group records by key. Since the stream source topic doesn't define any, the code has a selectKey operation specifying that the ID field in the value should be used as the key.

        .groupByKey(groupedPressure)

Since you've changed the key, under the covers Kafka Streams performs a repartition immediately before it performs the grouping.
Repartitioning is simply producing records to an internal topic and consuming them back into the application. By producing the records the updated keys land on the correct partition. Additionally, since the key-value types have changed you need to provide updated Serde objects, via the Grouped configuration object to Kafka Streams for the (de)serialization process for the repartitioning.

        TimeWindows windows = TimeWindows
            .ofSizeAndGrace(Duration.ofSeconds(10), Duration.ofSeconds(20))
            .advanceBy(Duration.ofSeconds(10));

        .windowedBy(windows)

This creates a windowed record stream with 10 second tumbling windows that allow a 20-second grace period during which late events may arrive.

        .count()

Here we count the number of records per grouped key and window using the count convenience method.

        .suppress(Suppressed.untilWindowCloses(unbounded()))

The count operator returns a KTable on which we can suppress updates until the window closes (after the 20-second grace period).

        .toStream()
        .to(OUTPUT_TOPIC, producedCount);

Aggregations in Kafka Streams return a KTable instance, so it's converted to a KStream and emitted to the output topic.