异步转同步
业务需求
有些接口查询反馈结果是异步返回的,无法立刻获取查询结果。
- 正常处理逻辑
触发异步操作,然后传递一个唯一标识。
等到异步结果返回,根据传入的唯一标识,匹配此次结果。
- 如何转换为同步
正常的应用场景很多,但是有时候不想做数据存储,只是想简单获取调用结果。
即想达到同步操作的结果,怎么办呢?
思路
-
发起异步操作
-
在异步结果返回之前,一直等待(可以设置超时)
-
结果返回之后,异步操作结果统一返回
常见的实现方式
(1)循环等待
(2)CountDownLatch
(3)Spring EventListener
(4)wait & notify
(5)使用条件锁
(6)Future
(7)使用 CyclicBarrier
循环等待
- LoopQuery.java
使用 query()
,将异步的操作 remoteCallback()
执行完成后,同步返回。
public class LoopQuery implements Async {
private String result;
private static final Logger LOGGER = LogManager.getLogger(LoopQuery.class.getName());
@Override
public String query(String key) {
startQuery(key);
new Thread(new Runnable() {
@Override
public void run() {
remoteCallback(key);
}
}).start();
final String queryResult = endQuery();
LOGGER.info("查询结果: {}", queryResult);
return queryResult;
}
/**
* 开始查询
* @param key 查询条件
*/
private void startQuery(final String key) {
LOGGER.info("执行查询: {}", key);
}
/**
* 远程的回调是等待是随机的
*
* @param key 查询条件
*/
private void remoteCallback(final String key) {
try {
TimeUnit.SECONDS.sleep(5);
} catch (InterruptedException e) {
e.printStackTrace();
}
this.result = key + "-result";
LOGGER.info("remoteCallback set result: {}", result);
}
/**
* 结束查询
* @return 返回结果
*/
private String endQuery() {
while (true) {
if (null == result) {
try {
TimeUnit.MILLISECONDS.sleep(100);
} catch (InterruptedException e) {
e.printStackTrace();
}
} else {
return result;
}
}
}
}
- main()
public static void main(String[] args) {
new LoopQuery().query("12345");
}
- 测试结果
18:14:16.491 [main] INFO com.github.houbb.thread.learn.aysnc.loop.LoopQuery - 执行查询: 12345
18:14:21.498 [Thread-1] INFO com.github.houbb.thread.learn.aysnc.loop.LoopQuery - remoteCallback set result: 12345-result
18:14:21.548 [main] INFO com.github.houbb.thread.learn.aysnc.loop.LoopQuery - 查询结果: 12345-result
CountDownLatch
- AsyncQuery.java
使用 CountDownLatch
类达到同步的效果。
import org.apache.logging.log4j.LogManager;
import org.apache.logging.log4j.Logger;
import java.util.concurrent.CountDownLatch;
import java.util.concurrent.TimeUnit;
public class AsyncQuery {
private static final Logger LOGGER = LogManager.getLogger(AsyncQuery.class.getName());
/**
* 结果
*/
private String result;
/**
* 异步转同步查询
* @param key
*/
public void asyncQuery(final String key) {
final CountDownLatch latch = new CountDownLatch(1);
this.startQuery(key);
new Thread(new Runnable() {
@Override
public void run() {
LOGGER.info("远程回调线程开始");
remoteCallback(key, latch);
LOGGER.info("远程回调线程结束");
}
}).start();
try {
latch.await();
} catch (InterruptedException e) {
e.printStackTrace();
}
this.endQuery();
}
private void startQuery(final String key) {
LOGGER.info("执行查询: {}", key);
}
/**
* 远程的回调是等待是随机的
* @param key
*/
private void remoteCallback(final String key, CountDownLatch latch) {
try {
TimeUnit.SECONDS.sleep(5);
} catch (InterruptedException e) {
e.printStackTrace();
}
this.result = key + "-result";
latch.countDown();
}
private void endQuery() {
LOGGER.info("查询结果: {}", result);
}
}
- main()
public static void main(String[] args) {
AsyncQuery asyncQuery = new AsyncQuery();
final String key = "123456";
asyncQuery.asyncQuery(key);
}
- 日志
18:19:12.714 [main] INFO com.github.houbb.thread.learn.aysnc.countdownlatch.AsyncQuery - 执行查询: 123456
18:19:12.716 [Thread-1] INFO com.github.houbb.thread.learn.aysnc.countdownlatch.AsyncQuery - 远程回调线程开始
18:19:17.720 [main] INFO com.github.houbb.thread.learn.aysnc.countdownlatch.AsyncQuery - 查询结果: 123456-result
18:19:17.720 [Thread-1] INFO com.github.houbb.thread.learn.aysnc.countdownlatch.AsyncQuery - 远程回调线程结束
Spring EventListener
使用观察者模式也可以。(对方案一的优化)
此处结合 spring 进行使用。
- BookingCreatedEvent.java
定义一个传输属性的对象。
public class BookingCreatedEvent extends ApplicationEvent {
private static final long serialVersionUID = -1387078212317348344L;
private String info;
public BookingCreatedEvent(Object source) {
super(source);
}
public BookingCreatedEvent(Object source, String info) {
super(source);
this.info = info;
}
public String getInfo() {
return info;
}
}
- BookingService.java
说明:当 this.context.publishEvent(bookingCreatedEvent);
触发时,
会被 @EventListener
指定的方法监听到。
import org.springframework.beans.factory.annotation.Autowired;
import org.springframework.context.ApplicationContext;
import org.springframework.context.event.EventListener;
import org.springframework.stereotype.Service;
import java.util.concurrent.TimeUnit;
@Service
public class BookingService {
@Autowired
private ApplicationContext context;
private volatile BookingCreatedEvent bookingCreatedEvent;
/**
* 异步转同步查询
* @param info
* @return
*/
public String asyncQuery(final String info) {
query(info);
new Thread(new Runnable() {
@Override
public void run() {
remoteCallback(info);
}
}).start();
while(bookingCreatedEvent == null) {
//.. 空循环
// 短暂等待。
try {
TimeUnit.MILLISECONDS.sleep(1);
} catch (InterruptedException e) {
//...
}
//2. 使用两个单独的 event...
}
final String result = bookingCreatedEvent.getInfo();
bookingCreatedEvent = null;
return result;
}
@EventListener
public void onApplicationEvent(BookingCreatedEvent bookingCreatedEvent) {
System.out.println("监听到远程的信息: " + bookingCreatedEvent.getInfo());
this.bookingCreatedEvent = bookingCreatedEvent;
System.out.println("监听到远程消息后: " + this.bookingCreatedEvent.getInfo());
}
/**
* 执行查询
* @param info
*/
public void query(final String info) {
System.out.println("开始查询: " + info);
}
/**
* 远程回调
* @param info
*/
public void remoteCallback(final String info) {
System.out.println("远程回调开始: " + info);
try {
TimeUnit.SECONDS.sleep(2);
} catch (InterruptedException e) {
e.printStackTrace();
}
// 重发结果事件
String result = info + "-result";
BookingCreatedEvent bookingCreatedEvent = new BookingCreatedEvent(this, result);
//触发event
this.context.publishEvent(bookingCreatedEvent);
}
}
- 测试方法
@RunWith(SpringJUnit4ClassRunner.class)
@ContextConfiguration(classes = SpringConfig.class)
public class BookServiceTest {
@Autowired
private BookingService bookingService;
@Test
public void asyncQueryTest() {
bookingService.asyncQuery("1234");
}
}
- 日志
2018-08-10 18:27:05.958 INFO [main] com.github.houbb.spring.lean.core.ioc.event.BookingService:84 - 开始查询:1234
2018-08-10 18:27:05.959 INFO [Thread-2] com.github.houbb.spring.lean.core.ioc.event.BookingService:93 - 远程回调开始:1234
接收到信息: 1234-result
2018-08-10 18:27:07.964 INFO [Thread-2] com.github.houbb.spring.lean.core.ioc.event.BookingService:73 - 监听到远程的信息: 1234-result
2018-08-10 18:27:07.964 INFO [Thread-2] com.github.houbb.spring.lean.core.ioc.event.BookingService:75 - 监听到远程消息后: 1234-result
2018-08-10 18:27:07.964 INFO [Thread-2] com.github.houbb.spring.lean.core.ioc.event.BookingService:106 - 已经触发event
2018-08-10 18:27:07.964 INFO [main] com.github.houbb.spring.lean.core.ioc.event.BookingService:67 - 查询结果: 1234-result
2018-08-10 18:27:07.968 INFO [Thread-1] org.springframework.context.support.GenericApplicationContext:993 - Closing org.springframework.context.support.GenericApplicationContext@5cee5251: startup date [Fri Aug 10 18:27:05 CST 2018]; root of context hierarchy
超时和空循环
空循环
空循环会导致 cpu 飙升
while(true) {
}
- 解决方式
while(true) {
// 小睡即可
TimeUnit.sleep(1);
}
超时编写
不可能一直等待反馈,可以设置超时时间。
/**
* 循环等待直到获取结果
* @param key key
* @param timeoutInSeconds 超时时间
* @param <T> 泛型
* @return 结果。如果超时则抛出异常
*/
public <T> T loopWaitForValue(final String key, long timeoutInSeconds) {
long startTime = System.nanoTime();
long deadline = startTime + TimeUnit.SECONDS.toNanos(timeoutInSeconds);
//1. 如果没有新回调,或者 key 对应元素不存在。则一直循环
while(ObjectUtil.isNull(map.get(key))) {
try {
TimeUnit.MILLISECONDS.sleep(5);
} catch (InterruptedException e) {
LOGGER.warn("Loop meet InterruptedException, just ignore it.", e);
}
// 超时判断
long currentTime = System.nanoTime();
if(currentTime >= deadline) {
throw new BussinessException(ErrorCode.READ_TIME_OUT);
}
}
final T target = (T) map.get(key);
LOGGER.debug("loopWaitForValue get value:{} for key:{}", JSON.toJSON(target), key);
//2. 获取到元素之后,需要移除掉对应的值
map.remove(key);
return target;
}