Apache Flink using coGroup to achieve left-outer join

Question

I've been trying to join two streams using CoGroupFunction in Flink.

I've two streams; which are;

S1

val m = env
.addSource(new FlinkKafkaConsumer010[String]("topic-1", schema, props))
.map(gson.fromJson(_, classOf[Master]))
.assignAscendingTimestamps(_.time)

S2

val d = env
.addSource(new FlinkKafkaConsumer010[String]("topic-2", schema, props))
.map(gson.fromJson(_, classOf[Detail]))
.assignAscendingTimestamps(_.time)

And my coGroup implementation is;

class MasterDetailOuterJoin extends CoGroupFunction[Master, Detail, 
(Master, Option[Detail])] {

  override def coGroup(
      leftElements : java.lang.Iterable[Master],
      rightElements: java.lang.Iterable[Detail],
      out: Collector[(Master, Option[Detail]) ]): Unit = {

    for (leftElem <- leftElements) {
      var isMatch = false
      println(leftElem.orderNo)
      for (rightElem <- rightElements) {
        println(rightElem.orderNo)
        out.collect((leftElem, Some(rightElem)))
        isMatch = true
      }
      if (!isMatch) {
        out.collect((leftElem, None))
      }
    }
  }
}

And i run it with;

m.coGroup(d)
    .where(_.orderNo)
    .equalTo(_.orderNo)
    .window(TumblingEventTimeWindows.of(Time.of(5, TimeUnit.SECONDS)))
    .apply(new MasterDetailOuterJoin)
    .map(gson.toJson(_, classOf[(Master, Option[Detail])]))
    .print

But nothing is printed even there is a match in master and detail! I monitor the kafka streams with console consumer and they are working fine btw.

If i do it with a inner join instead i get the result.

 m.keyBy(_.orderNo)
    .connect(d.keyBy(_.orderNo))
    .flatMap(new MasterDetailInnerJoin) //RichCoFlatMapFunction
    .map(gson.toJson(_, classOf[(Master, Detail)]))
    .print

Answer 1

It turned out that, what i was missing was;

• env.setStreamTimeCharacteristic(TimeCharacteristic.EventTime)
• and assign a timestamp and watermark extractor to each stream