Spring Boot中通过RabbitTemplate主动pull(get)消息的例子

import java.util.Properties;
import java.util.function.Consumer;

import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import org.springframework.amqp.core.AmqpAdmin;
import org.springframework.amqp.core.Message;
import org.springframework.amqp.core.MessageProperties;
import org.springframework.amqp.rabbit.core.ChannelCallback;
import org.springframework.amqp.rabbit.core.RabbitAdmin;
import org.springframework.amqp.rabbit.core.RabbitTemplate;
import org.springframework.amqp.rabbit.support.DefaultMessagePropertiesConverter;
import org.springframework.amqp.rabbit.support.MessagePropertiesConverter;
import org.springframework.amqp.support.converter.MessageConverter;
import org.springframework.beans.factory.annotation.Autowired;
import org.springframework.stereotype.Service;

import com.rabbitmq.client.Channel;
import com.rabbitmq.client.DefaultConsumer;
import com.rabbitmq.client.GetResponse;

@Service
public class RabbitAdminServices {

    private static final Logger logger = LoggerFactory.getLogger(RabbitAdminServices.class);

    @Autowired
    AmqpAdmin rabbitAdmin;

    @Autowired
    RabbitTemplate rabbitTemplate;

    @Autowired
    MessageConverter messageConverter;

    private volatile MessagePropertiesConverter messagePropertiesConverter = new DefaultMessagePropertiesConverter();

    public int getCount(String queueName) {

        Properties properties = rabbitAdmin.getQueueProperties(queueName);
        return (Integer)properties.get(RabbitAdmin.QUEUE_MESSAGE_COUNT);
    }

    public <T> void processQueue(String queueName, Integer count, Class<T> clazz, Consumer<T> consumer) {

        int reprocessCount = getCount(queueName);
        int requestCount = reprocessCount;
        if(count != null) {
            requestCount = count;
        }
        for(int i = 0; i < reprocessCount && i < requestCount; i++) {
            rabbitTemplate.execute(new ChannelCallback<T>() {

                @Override
                public T doInRabbit(Channel channel) throws Exception {
                    GetResponse response = channel.basicGet(queueName, false);
                    T result = null;
                    try {
                        MessageProperties messageProps = messagePropertiesConverter.toMessageProperties(response.getProps(), response.getEnvelope(), "UTF-8");
                        if(response.getMessageCount() >= 0) {
                            messageProps.setMessageCount(response.getMessageCount());
                        }
                        Message message = new Message(response.getBody(), messageProps);
                        result = (T)messageConverter.fromMessage(message);
                        consumer.accept(result);
                        channel.basicAck(response.getEnvelope().getDeliveryTag(), false);
                    }
                    catch(Exception e) {
                        channel.basicNack(response.getEnvelope().getDeliveryTag(), false, true);
                    }
                    return result;
                }
            });

        }
    }
}

实现RabbitMQ的消费者有两种模式，推模式（Push）和拉模式（Pull）。

实现推模式推荐的方式是继承 DefaultConsumer 基类，也可以使用Spring AMQP的 SimpleMessageListenerContainer 。

推模式是最常用的，但是有些情况下推模式并不适用的，比如说：

• 由于某些限制，消费者在某个条件成立时才能消费消息

• 需要批量拉取消息进行处理

实现拉模式

RabbitMQ的Channel提供了 basicGet 方法用于拉取消息。

1.  
   /**
2.  
   * Retrieve a message from a queue using {@link com.rabbitmq.client.AMQP.Basic.Get}
3.  
   * @see com.rabbitmq.client.AMQP.Basic.Get
4.  
   * @see com.rabbitmq.client.AMQP.Basic.GetOk
5.  
   * @see com.rabbitmq.client.AMQP.Basic.GetEmpty
6.  
   * @param queue the name of the queue
7.  
   * @param autoAck true if the server should consider messages
8.  
   * acknowledged once delivered; false if the server should expect
9.  
   * explicit acknowledgements
10.  
    * @return a {@link GetResponse} containing the retrieved message data
11.  
    * @throws java.io.IOException if an error is encountered
12.  
    */
13.  
    GetResponse basicGet(String queue, boolean autoAck) throws IOException;

basicGet 返回 GetResponse 类。

1.  
   public class GetResponse {
2.  
   private final Envelope envelope;
3.  
   private final BasicProperties props;
4.  
   private final byte[] body;
5.  
   private final int messageCount;
6.  
    
7.  
   // ...

public class GetResponse { private final Envelope envelope; private final BasicProperties props; private final byte[] body; private final int messageCount; // ...

rabbitmq-client版本4.0.3

使用 basicGet 拉取消息需要注意：

1.  
   basicGet
2.  
   DefaultConsumer

示例代码：

1.  
   private void consume(Channel channel) throws IOException, InterruptedException {
2.  
   while (true) {
3.  
   if (!isConditionSatisfied()) {
4.  
   TimeUnit.MILLISECONDS.sleep(1);
5.  
   continue;
6.  
   }
7.  
   GetResponse response = channel.basicGet(CAOSH_TEST_QUEUE, false);
8.  
   if (response == null) {
9.  
   TimeUnit.MILLISECONDS.sleep(1);
10.  
    continue;
11.  
    }
12.  
    String data = new String(response.getBody());
13.  
    logger.info("Get message <= {}", data);
14.  
    channel.basicAck(response.getEnvelope().getDeliveryTag(), false);
15.  
    }
16.  
    }

批量拉取消息

RabbitMQ支持客户端批量拉取消息，客户端可以连续调用 basicGet 方法拉取多条消息，处理完成之后一次性ACK。需要注意：

1.  
   basicGet
2.  
   basicAck

示例代码：

1.  
   String bridgeQueueName = extractorProperties.getBridgeQueueName();
2.  
   int batchSize = extractorProperties.getBatchSize();
3.  
   List<GetResponse> responseList = Lists.newArrayListWithCapacity(batchSize);
4.  
   long tag = 0;
5.  
   while (responseList.size() < batchSize) {
6.  
   GetResponse getResponse = channel.basicGet(bridgeQueueName, false);
7.  
   if (getResponse == null) {
8.  
   break;
9.  
   }
10.  
    responseList.add(getResponse);
11.  
    tag = getResponse.getEnvelope().getDeliveryTag();
12.  
    }
13.  
    if (responseList.isEmpty()) {
14.  
    TimeUnit.MILLISECONDS.sleep(1);
15.  
    } else {
16.  
    logger.info("Get <{}> responses this batch", responseList.size());
17.  
    // handle messages
18.  
    channel.basicAck(tag, true);
19.  
    }

关于QueueingConsumer

QueueingConsumer 在客户端本地使用 BlockingQueue 缓冲消息，其nextDelivery方法也可以用于实现拉模式（其本质上是 BlockingQueue.take ），但是 QueueingConsumer 现在已经标记为Deprecated。