复杂对象KStream GlobalKTable左联接

Question

我对Kafka的溪流很陌生。我想执行以下KStream-GlobalKTable纯基于DSL的左联接操作，而不使用map操作。

{
    "b_obj": {
        "b_value": "xyz",
        "c_list": [
            {
                "d_obj": {
                    "d_id1": "value1",
                    "d_id2": "value2",
                    "d_value": "some value"
                },
                "c_value": "jkl"
            },
            {
                "d_obj": {
                    "d_id1": "value3",
                    "d_id2": "value4",
                    "d_value": "some value 2"
                },
                "c_value": "pqr"
            }
        ]
    },
    "a_value": "abcd"
}

和另一个输入主题e.topic,它是 ,其中value：

{
    "e_id1": "value1",
    "e_id2": "value2",
    "e_value": "some value"
}

我要执行左联接操作.topic是一个流，主数据.topic是一个全局表，以实现结果值为

{
    "b_obj": {
        "b_value": "xyz",
        "c_list": [
            {
                "d_obj": {
                    "d_id1": "value1",
                    "d_id2": "value2",
                    "d_value": "some value"
                },
                "e_obj": {
                    "e_id1": "value1",
                    "e_id2": "value2",
                    "e_value": "some value a"
                },
                "c_value": "jkl"
            },
            {
                "d_obj": {
                    "d_id1": "value3",
                    "d_id2": "value4",
                    "d_value": "some value 2"
                },
                "e_obj": {
                    "e_id1": "value3",
                    "e_id2": "value4",
                    "e_value": "some value b"
                },
                "c_value": "pqr"
            }
        ]
    },
    "a_value": "abcd"
}

连接条件为a.b.c[i].d.d_id1==e.e_id1&&a.b.c[i].d.d_id2==e.e_id2

代码：

public class ComplexBeanStream {

    public static void main(String[] args) {

        Serde<A> aSerde = new JsonSerde<>(A.class);

        Serde<E> eSerde = new JsonSerde<>(E.class);

        final Properties streamsConfiguration = new Properties();
        streamsConfiguration.put(StreamsConfig.APPLICATION_ID_CONFIG, "complex-bean-app");
        streamsConfiguration.put(StreamsConfig.CLIENT_ID_CONFIG, "complex-bean-client");
        streamsConfiguration.put(StreamsConfig.BOOTSTRAP_SERVERS_CONFIG, "localhost:9092");

        StreamsBuilder builder = new StreamsBuilder();

        final GlobalKTable<String, E> eGlobalTable =
                builder.globalTable(
                        "e.topic",
                        Materialized.<String, E, KeyValueStore<Bytes, byte[]>>
                                as("E-STORE")
                                .withKeySerde(Serdes.String())
                                .withValueSerde(eSerde)
                );

        final KStream<String, A> aStream =
                builder.stream(
                        "a.topic",
                        Consumed.with(Serdes.String(), aSerde));

        // perform left-join here

        Topology topology = builder.build();

        System.out.println("\n\nComplexBeanStream Topology: \n" + topology.describe());

        final KafkaStreams streams = new KafkaStreams(topology, streamsConfiguration);

        streams.cleanUp();

        streams.start();

        Runtime.getRuntime().addShutdownHook(new Thread(streams::close));
    }

}

class A {

    private B b_obj;

    private String a_value;
}

class B {

    private List<C> c_list;

    private String b_value;
}

class C {

    private D d_obj;

    private E e_obj;

    private String c_value;
}

class D {

    private String d_id1;

    private String d_id2;

    private String d_value;
}

class E {

    private String e_id1;

    private String e_id2;

    private String e_value;
}

Answer 1

在您的情况下，如果连接查找是基于值而不是键，则不可能在没有映射的情况下进行连接。

Kafka流只能基于两边的相同键来连接。这意味着，您应该为连接的两侧映射并选择新的键（重新键控），以实现a.b.c[i].d.d_id1==e.e_id1&&a.b.c[i].d.d_id2==e.e_id2。

在这种情况下，一方可以使用[a.b.c[i].d.d_id1,a.b.c[i].d.d_id2]作为其密钥，另一方则使用[e.e_id1,e.e_id2]。如果有匹配项，则可以将这些值连接到一个新对象。您可能应该在重新键控之前对c_list进行平面映射。

阅读Kafaka的加入也很有帮助。