Flink outputs to Kafka, proper way to use KeyedSerializationSchema - apache-kafka

Question

I have a simple Flink wordcount, which reads from Kafka topic, and outputs its result to another Kafka topic,
DataStream<String> input = env.addSource(new FlinkKafkaConsumer<>(inputTopic, new SimpleStringSchema(), props));
DataStream<Tuple2<String, Long>> counts = ......;
counts.addSink(new FlinkKafkaProducer<>(outputTopic, new WordCountSerializer(), props));
//counts.print();
env.execute("foobar");
The problem is, I see nothing in the output topic via kafka-console-consumer.sh command line.
To address the issue, I try to print the result, and it works fine, I can see the correct workcount result in the log file.
So the guess is that something wrong in WordCountSerializer, which is like,
class WordCountSerializer implements KeyedSerializationSchema<Tuple2<String, Long>>, java.io.Serializable {
public byte[] serializeKey(Tuple2<String, Long> element) {
return new StringSerializer().serialize(null, element.getField(0));
}
public byte[] serializeValue(Tuple2<String, Long> element) {
return new LongSerializer().serialize(null, element.getField(1));
}
public String getTargetTopic(Tuple2<String, Long> element) {
return null;
}
}
After changing the serializeValue to
public byte[] serializeValue(Tuple2<String, Long> element) {
return new StringSerializer().serialize(null, element.getField(1).toString());
}
I can see the count been output to Kafka (the word part of Tuples is still missing), like
1
3
...
My questions are,
I've seen several examples on the internet using the WordCountSerializer mentioned aboved, but it doesn't work for me, am I doing anything wrong here?
After changing the serializeValue method as above, it partially works, but what I actually want is something below, what is the proper way to achieve this?
foo,1
bar,3
...