Talk about seven ways to realize asynchronous programming

Recently, many friends left me messages , Can you summarize asynchronous programming , Today, let's talk about this topic briefly .

Early systems were synchronous , Easy to understand , Let's take an example

Synchronous programming

When a user creates an e-commerce transaction order , The business logic process to go through is still very long , Each step takes a certain amount of time , So the whole RT It will be longer .

therefore , Smart people began to think about whether they could separate some non core businesses from the main process , Hence the Asynchronous programming embryonic form .

Asynchronous programming is a means to make programs run concurrently . It allows multiple events to occur simultaneously , When a program calls a method that needs to run for a long time , It does not block the current execution process , The program can continue to run .

The core idea ： Optimize performance with multithreading , Turn serial operation into parallel operation . Programs designed in asynchronous mode can significantly reduce thread waiting , Thus in high throughput scenarios , Greatly improve the overall performance of the system , Significantly reduces latency .

Next , Let's talk about the programming implementation of asynchrony

One 、 Threads Thread

Direct inheritance Thread class Is the easiest way to create asynchronous threads .

First , establish Thread Subclass , Ordinary class or anonymous inner class ; Then create subclass instances ; Finally through start() Method to start the thread .

public class AsyncThread extends Thread{
    @Override
    public void run() {
        System.out.println(" Current thread name :" + this.getName() + ",  Execution thread name :" + Thread.currentThread().getName() + "-hello");
    }
}

public static void main(String[] args) {

  //  Simulate business processes 
  // .......
  
    //  Create asynchronous threads  
    AsyncThread asyncThread = new AsyncThread();

    //  Start asynchronous thread 
    asyncThread.start();
}

Of course, if you create one every time Thread Threads , Frequent creation 、 The destruction , Waste system resources . We can use thread pool

@Bean(name = "executorService")
public ExecutorService downloadExecutorService() {
    return new ThreadPoolExecutor(20, 40, 60, TimeUnit.SECONDS, new ArrayBlockingQueue<>(2000),
            new ThreadFactoryBuilder().setNameFormat("defaultExecutorService-%d").build(),
            (r, executor) -> log.error("defaultExecutor pool is full! "));
}

Encapsulate business logic into Runnable or Callable in , Leave it to Thread pool To execute

Two 、Future

Although the above method achieves multi-threaded parallel processing , But some businesses need more than just process execution , Also get the execution results .

Java from 1.5 Version start , Provides Callable and Future, You can get the task execution result after the task execution .

Of course, it also provides other functions , Such as ： Cancel the task 、 Check whether the task is completed

Future Class is located java.util.concurrent It's a bag , Interface definition ：

public interface Future<V> {
    boolean cancel(boolean mayInterruptIfRunning);
    boolean isCancelled();
    boolean isDone();
    V get() throws InterruptedException, ExecutionException;
    V get(long timeout, TimeUnit unit)
        throws InterruptedException, ExecutionException, TimeoutException;
}

Methods described ：

• cancel()： Cancel the task , If you cancel the task successfully, return true, Returns if the cancel task fails false

• isCancelled()： Indicates whether the task was cancelled successfully , If the mission is cancelled before it is normally completed , Then return to true

• isDone()： Indicates whether the task has been completed , If completed , return true

• get()： Get execution results , This method will block , Will wait until the task is completed to return

• get(long timeout, TimeUnit unit)： Used to get execution results , If within the specified time , We haven't got the result yet , Go straight back null

Code example ：

public class CallableAndFuture {

    public static ExecutorService executorService = new ThreadPoolExecutor(4, 40,
            0L, TimeUnit.MILLISECONDS,
            new LinkedBlockingQueue<Runnable>(1024), new ThreadFactoryBuilder()
            .setNameFormat("demo-pool-%d").build(), new ThreadPoolExecutor.AbortPolicy());


    static class MyCallable implements Callable<String> {
        @Override
        public String call() throws Exception {
            return " Asynchronous processing ,Callable  Return results ";
        }
    }

    public static void main(String[] args) {
        Future<String> future = executorService.submit(new MyCallable());
        try {
            System.out.println(future.get());
        } catch (Exception e) {
            // nodo
        } finally {
            executorService.shutdown();
        }
    }
}

Future Represents the result of an asynchronous task that may not have been completed , adopt get Method to get the execution result , This method blocks until the task returns a result .

3、 ... and 、FutureTask

FutureTask Realized RunnableFuture Interface , be RunnableFuture Interface inherited Runnable Interface and Future Interface , So you can FutureTask Object is submitted as a task to ThreadPoolExecutor To carry out , It can also be directly Thread perform ; Because of the realization of Future Interface , So it can also be used to get the results of the task .

FutureTask Constructors ：

public FutureTask(Callable<V> callable)
public FutureTask(Runnable runnable, V result)

FutureTask Often used to encapsulate Callable and Runnable, It can be submitted to the thread pool as a task for execution . Except as a separate class , Some functional functions are also provided for us to create custom task Class uses .

FutureTask Thread safety by CAS To guarantee .

ExecutorService executor = Executors.newCachedThreadPool();
// FutureTask packing callbale Mission , Then hand it over to the thread pool for execution 
FutureTask<Integer> futureTask = new FutureTask<>(() -> {
    System.out.println(" The child thread starts to calculate ：");
    Integer sum = 0;
    for (int i = 1; i <= 100; i++)
        sum += i;
    return sum;
});

//  Thread pool execution task ,  The running results are in  futureTask  Inside the object 
executor.submit(futureTask);

try {
    System.out.println("task The sum of the running result calculation is ：" + futureTask.get());
} catch (Exception e) {
    e.printStackTrace();
}
executor.shutdown();

Callable and Future The difference between ：Callable Used to produce results ,Future Used to get results

If it is multiple free serial tasks 、 Or parallel combination , It involves synchronous blocking between multiple threads to obtain results ,Future The code implementation will be cumbersome , We need to deal with each intersection manually , It's easy to make a mistake .

Four 、 Asynchronous framework CompletableFuture

Future Class passing get() Method block and wait to get the running result of asynchronous execution , Poor performance .

JDK1.8 in ,Java Provides CompletableFuture class , It is based on asynchronous functional programming . Relatively blocking waiting for the return result ,CompletableFuture The calculation results can be processed by callback , It realizes asynchronous non blocking , Better performance .

advantage ：

• At the end of an asynchronous task , Methods that call back an object automatically

• When an asynchronous task fails , Methods that call back an object automatically

• After the main thread sets the callback , No longer concerned with the execution of asynchronous tasks

Examples of making tea ：

( Excerpt from ： Geek time 《Java Concurrent programming practice 》)

// Mission 1： Wash the kettle -> The boiling water 
CompletableFuture<Void> f1 =
        CompletableFuture.runAsync(() -> {
            System.out.println("T1: Wash the kettle ...");
            sleep(1, TimeUnit.SECONDS);

            System.out.println("T1: The boiling water ...");
            sleep(15, TimeUnit.SECONDS);
        });

// Mission 2： Wash the teapot -> Wash tea cups -> Take the tea 
CompletableFuture<String> f2 =
        CompletableFuture.supplyAsync(() -> {
            System.out.println("T2: Wash the teapot ...");
            sleep(1, TimeUnit.SECONDS);

            System.out.println("T2: Wash tea cups ...");
            sleep(2, TimeUnit.SECONDS);

            System.out.println("T2: Take the tea ...");
            sleep(1, TimeUnit.SECONDS);
            return " Longjing ";
        });

// Mission 3： Mission 1 And tasks 2 When it's done ： Make tea 
CompletableFuture<String> f3 =
        f1.thenCombine(f2, (__, tf) -> {
            System.out.println("T1: Get the tea :" + tf);
            System.out.println("T1: Make tea ...");
            return " Tea :" + tf;
        });

// Wait for the task 3 Execution results 
System.out.println(f3.join());

}

CompletableFuture Provides a very rich API, There are about 50 Kind of processing serial , parallel , Combination and methods of handling errors .

More about an article written before moving , Get it done CompletableFuture, What's the difference between concurrent asynchronous programming and serial programming ？

5、 ... and 、 SpringBoot annotation @Async

In addition to hard coded asynchronous programming processing ,SpringBoot The framework also provides Annotation type Solution , With Method body As boundary , The code logic inside the method body is executed asynchronously .

First , Use @EnableAsync Enable asynchronous annotation

@SpringBootApplication
@EnableAsync
public class StartApplication {

    public static void main(String[] args) {
        SpringApplication.run(StartApplication.class, args);
    }
}

Custom thread pool ：

@Configuration
@Slf4j
public class ThreadPoolConfiguration {

    @Bean(name = "defaultThreadPoolExecutor", destroyMethod = "shutdown")
    public ThreadPoolExecutor systemCheckPoolExecutorService() {

        return new ThreadPoolExecutor(3, 10, 60, TimeUnit.SECONDS,
                new LinkedBlockingQueue<Runnable>(10000),
                new ThreadFactoryBuilder().setNameFormat("default-executor-%d").build(),
                (r, executor) -> log.error("system pool is full! "));
    }
}

Add annotations to the asynchronous processing method @Async , When the execute Method Invocation time , Through the customized thread pool defaultThreadPoolExecutor Asynchronous execution  execute Method

@Service
public class AsyncServiceImpl implements AsyncService {

    @Async("defaultThreadPoolExecutor")
    public Boolean execute(Integer num) {
        System.out.println(" Threads ：" + Thread.currentThread().getName() + " ,  Mission ：" + num);
        return true;
    }

}

use @Async How to annotate and mark , Called asynchronous method . stay spring boot Use in application @Async It's simple ：

• Annotate the asynchronous method class or startup class @EnableAsync

• Add to the method that needs to be called asynchronously @Async

• What is used @Async The class object of the annotation method should be Spring container-managed bean object ;

6、 ... and 、Spring ApplicationEvent event

Event mechanism is often used in some large projects ,Spring It provides a set of interfaces for event mechanism , It meets the decoupling of the architecture in principle .

ApplicationContext adopt ApplicationEvent Classes and ApplicationListener Interface for event handling . If it will achieve ApplicationListener Interface bean Inject into context , Then use it every time ApplicationContext Release ApplicationEvent when , Will inform the bean. Essentially , It's standard Observer design pattern .

ApplicationEvent By Spring All of the Event Class base class

First , Customize business event subclasses , Inherited from ApplicationEvent, Inject business model parameter classes through generics . amount to MQ The body of the message .

public class OrderEvent extends AbstractGenericEvent<OrderModel> {
    public OrderEvent(OrderModel source) {
        super(source);
    }
}

then , Write event listeners .ApplicationListener The interface is made by Spring Provide the interface that the event subscriber must implement , We need to define a subclass , Inherit ApplicationListener. amount to MQ The consumer side of

@Component
public class OrderEventListener implements ApplicationListener<OrderEvent> {
    @Override
    public void onApplicationEvent(OrderEvent event) {

        System.out.println("【OrderEventListener】 Monitor processing ！" + JSON.toJSONString(event.getSource()));

    }
}

Last , Release events , Tell all listeners related to an event . amount to MQ The production side of .

OrderModel orderModel = new OrderModel();
orderModel.setOrderId((long) i);
orderModel.setBuyerName("Tom-" + i);
orderModel.setSellerName("judy-" + i);
orderModel.setAmount(100L);
//  Release Spring Event notification 
SpringUtils.getApplicationContext().publishEvent(new OrderEvent(orderModel));

Add a meal ：

[ The consumer end ] Threads ：http-nio-8090-exec-1, Consumer events  {"amount":100.0,"buyerName":"Tom-1","orderId":1,"sellerName":"judy-1"}
[ Production end ] Threads ：http-nio-8090-exec-1, Release events  1
[ The consumer end ] Threads ：http-nio-8090-exec-1, Consumer events  {"amount":100.0,"buyerName":"Tom-2","orderId":2,"sellerName":"judy-2"}
[ Production end ] Threads ：http-nio-8090-exec-1, Release events  2
[ The consumer end ] Threads ：http-nio-8090-exec-1, Consumer events  {"amount":100.0,"buyerName":"Tom-3","orderId":3,"sellerName":"judy-3"}
[ Production end ] Threads ：http-nio-8090-exec-1, Release events  3

There is a running demo Results of operation , We found that both the production side and the consumption side , The same thread is used http-nio-8090-exec-1,Spring The event mechanism of the framework defaults to synchronous blocking . Just understand the code specification , It has better expansibility , But the bottom layer still adopts synchronous calling mode .

So here comes the question , If you want to implement asynchronous calls , How to deal with it ？

We need to create one manually SimpleApplicationEventMulticaster, And set up TaskExecutor, At this time, all consumption events are executed by asynchronous threads .

@Component
public class SpringConfiguration {

    @Bean
    public SimpleApplicationEventMulticaster applicationEventMulticaster(@Qualifier("defaultThreadPoolExecutor") ThreadPoolExecutor defaultThreadPoolExecutor) {
        SimpleApplicationEventMulticaster simpleApplicationEventMulticaster = new SimpleApplicationEventMulticaster();
        simpleApplicationEventMulticaster.setTaskExecutor(defaultThreadPoolExecutor);
        return simpleApplicationEventMulticaster;
    }

}

Let's take a look at the operation results after the transformation ：

[ Production end ] Threads ：http-nio-8090-exec-1, Release events  1
[ Production end ] Threads ：http-nio-8090-exec-1, Release events  2
[ Production end ] Threads ：http-nio-8090-exec-1, Release events  3
[ The consumer end ] Threads ：default-executor-1, Consumer events  {"amount":100.0,"buyerName":"Tom-2","orderId":2,"sellerName":"judy-2"}
[ The consumer end ] Threads ：default-executor-2, Consumer events  {"amount":100.0,"buyerName":"Tom-1","orderId":1,"sellerName":"judy-1"}
[ The consumer end ] Threads ：default-executor-0, Consumer events  {"amount":100.0,"buyerName":"Tom-3","orderId":3,"sellerName":"judy-3"}

SimpleApplicationEventMulticaster This is instantiated by ourselves Bean How about the default loading order of the system ？ Will there be a conflict ？

Look up the Spring Source code , The processing logic is AbstractApplicationContext#initApplicationEventMulticaster In the method , adopt beanFactory Find out whether there are customized Bean, without , The container itself new One SimpleApplicationEventMulticaster Object into the container .

Code address ：https://github.com/aalansehaiyang/wx-project

7、 ... and 、 Message queue

Asynchronous architecture is a typical architecture pattern in Internet system , Corresponding to synchronous architecture . Message queuing is inherently such an asynchronous architecture , With ultra-high throughput and ultra-low delay .

The main roles of message queue asynchronous architecture include message producers 、 Message queues and message consumers .

The message producer is the main application , The producer encapsulates the call request as a message and sends it to the message queue .

The function of message queue is to buffer messages , Waiting for consumers to spend . According to the consumption mode, it can be divided into Point to point mode and Publish subscribe mode Two kinds of .

Message consumer , Used to pull from the message queue 、 News consumption , Complete business logic processing .

Of course, there are many message queue frameworks on the market , Common are RabbitMQ、Kafka、RocketMQ、ActiveMQ and Pulsar etc.

Different message queues have slightly different functional characteristics , But the overall architecture is similar , It's not going to unfold here .

We just need to remember one key point , With the help of message queue, this middleware can efficiently realize asynchronous programming .

 Previous recommendation 

Ali ' ,SpringBoot Automated Deployment artifact ！

Java in for and foreach Which performance is high ？

How to write elegantly Controller Layer code ？