Kafka Streams:如何获取 SessionWindow 的第一条和最后一条记录

Question

默认情况下，.windowedBy(SessionWindows.with(Duration.ofSeconds(60))为每个传入记录返回一条记录。

结合.count()和.filter()，可以轻松检索第一条记录。

用 .suppress(Suppressed.untilWindowCloses(unbounded()))检索上一条记录也很容易。

所以。。。我做了两次处理，如你所看到的适应字数示例：

final KStream<String, String> streamsBranches = builder.<String,String>stream("streams-plaintext-input");
streamsBranches
  .flatMapValues(value -> Arrays.asList(value.toLowerCase(Locale.getDefault()).split("\W+")))
  .groupBy((key, value) -> ""+value)
  .windowedBy(SessionWindows.with(Duration.ofSeconds(60)).grace(Duration.ofSeconds(2)))
  .count(Materialized.with(Serdes.String(), Serdes.Long()))
  .toStream()
  .map((wk, v) -> new KeyValue<>(wk.key(), v == null ? -1l : v))
  .filter((wk, v) -> v == 1)
  .to("streams-wordcount-output", Produced.with(Serdes.String(), Serdes.Long()));
streamsBranches
  .flatMapValues(value -> Arrays.asList(value.toLowerCase(Locale.getDefault()).split("\W+")))
  .groupBy((key, value) -> ""+value)
  .windowedBy(SessionWindows.with(Duration.ofSeconds(60)).grace(Duration.ofSeconds(2)))
  .count(Materialized.with(Serdes.String(), Serdes.Long()))
  .suppress(Suppressed.untilWindowCloses(unbounded()))
  .toStream()
  .map((wk, v) -> new KeyValue<>(wk.key(), v))
  .filter((wk, v) -> v != null)
  .to("streams-wordcount-output", Produced.with(Serdes.String(), Serdes.Long()));

但我想知道是否有一种更简单、更美丽的方法来做同样的事情。

Answer 1

我认为您应该使用SessionWindowedKStream::aggregate(...)并根据您的逻辑在聚合器中累积结果(第一个和最后一个值(

示例代码可能如下所示：

streamsBranches.groupByKey()
        .windowedBy(SessionWindows.with(Duration.ofSeconds(60)).grace(Duration.ofSeconds(2)))
        .aggregate(
                AggClass::new,
                (key, value, oldAgg) -> oldAgg.update(value),
                (key, agg1, agg2) -> agg1.merge(agg2),
                Materialized.with(Serdes.String(), new AggClassSerdes())
        ).suppress(Suppressed.untilWindowCloses(unbounded()))
        .toStream().map((wk, v) -> new KeyValue<>(wk.key(), v))
.to("streams-wordcount-output", Produced.with(Serdes.String(), new AggClassSerdes()));

其中AggClass是累加器，AggClassSerdes是该累加器的Serdes

public class AggClass {
    private String first;
    private String last;
    public AggClass() {}
    public AggClass(String first, String last) {
        this.first = first;
        this.last = last;
    }
    public AggClass update(String value) {
        if (first == null)
            first = value;
        last = value;
        return this;
    }
    public AggClass merge(AggClass other) {
        if (this.first == null)
            return other;
        else return new AggClass(this.first, other.last);
    }
}