如何在准備就緒時使用反應式 Flux/Mono 將消息推送到上游而不是在間隔中輪詢狀態？

Question

嘗試在消息可用/准備好並在刷新后關閉連接時將消息推送到上游，而不是使用 spring 反應通量間隔輪詢消息。

@GetMapping(value = "/getValue/{randomId}", produces = MediaType.TEXT_EVENT_STREAM_VALUE)
public Flux<String> statusCheck(@PathVariable("randomId") @NonNull String randomId) {

return Flux.<String>interval(Duration.ofSeconds(3))
                .map(status -> {
                    if (getSomething(randomId).
                            equalsIgnoreCase("value"))
                        return "value";
                    return "ping";
                }).take(Duration.ofSeconds(60)).timeout(Duration.ofSeconds(60));
    }

Kafka 偵聽器在獲取時更新地圖中的 randomId 值，getSomething 方法檢查地圖中間隔的 randomId 值。 因此，我想在偵聽器接收到消息時將消息推送到客戶端，而不是檢查間隔並將數據存儲在地圖中。

Answer 1

這聽起來像是一個Flux.create()請求：

return Flux.<String>create(emitter -> {
     if (getSomething(randomId).equalsIgnoreCase("value")) {
          sink.next("value");
     }
     else {
          sink.next("ping");
     }
  });

---

/**
 * Programmatically create a {@link Flux} with the capability of emitting multiple
 * elements in a synchronous or asynchronous manner through the {@link FluxSink} API.
 * This includes emitting elements from multiple threads.
 * <p>
 * <img class="marble" src="doc-files/marbles/createForFlux.svg" alt="">
 * <p>
 * This Flux factory is useful if one wants to adapt some other multi-valued async API
 * and not worry about cancellation and backpressure (which is handled by buffering
 * all signals if the downstream can't keep up).
 * <p>
 * For example:
 *
 * <pre><code>
 * Flux.&lt;String&gt;create(emitter -&gt; {
 *
 *     ActionListener al = e -&gt; {
 *         emitter.next(textField.getText());
 *     };
 *     // without cleanup support:
 *
 *     button.addActionListener(al);
 *
 *     // with cleanup support:
 *
 *     button.addActionListener(al);
 *     emitter.onDispose(() -> {
 *         button.removeListener(al);
 *     });
 * });
 * </code></pre>
 *
 * @reactor.discard The {@link FluxSink} exposed by this operator buffers in case of
 * overflow. The buffer is discarded when the main sequence is cancelled.
 *
 * @param <T> The type of values in the sequence
 * @param emitter Consume the {@link FluxSink} provided per-subscriber by Reactor to generate signals.
 * @return a {@link Flux}
 * @see #push(Consumer)
 */
public static <T> Flux<T> create(Consumer<? super FluxSink<T>> emitter) {

Answer 2

我基於此 stackoverflow Spring 5 Web Reactive - Hot Publishing - How to use EmitterProcessor 將 MessageListener 橋接到事件流答案構建了解決方案，使用 EmitterProcessor 在消息可用時對其進行熱發布。

這是示例代碼

@GetMapping(value = "/getValue/{randomId}", produces = MediaType.TEXT_EVENT_STREAM_VALUE)
public Flux<String> statusCheck(@PathVariable("randomId") @NonNull String randomId) {
    EmitterProcessor<String> emitterProcessor = EmitterProcessor.create();
    Flux<String> autoConnect = emitterProcessor.publish().autoConnect();
    FluxSink<String> sink = emitterProcessor.sink();
    //storing randomId and processor sink details
    randomIdMap.putIfAbsent(randomId, emitterProcessor);
    /** This will return ping status to notify client as 
    connection is alive until the randomId message received. **/
    sendPingStatus(sink, randomId);
}

下面的方法顯示了如何在消息到達 kafka 消費者並關閉通量連接時將消息推送到客戶端。

@KafkaListener(topics = "some-subscription-id",
        containerFactory = "kafkaListenerContainerFactory")
public void pushMessage(SomeMessage message, Acknowledgment acknowledgment) {
    EmitterProcessor emitter = randomIdMap.get("randomId");
    if (emitter != null ) {
        emitter.onNext(message);
        emitter.onComplete();
        randomIdMap.remove("randomId");
        acknowledgment.acknowledge();
    }
}