生产者消费者模型
生产者和消费者问题是线程模型中老生常谈的问题,也是面试中经常遇到的问题。今天我们就盘它!!!
一、生产者消费者模型
何为生产者消费者模型呢?
生产者消费者模型是由两类线程构成:
- 生产者线程:“生产”产品,并把产品放到一个缓冲区里;
- 消费者线程:“消费”产品。
如图所示:
小贴士:
- 生产者持续生产,直到缓冲区满,满时阻塞;缓冲区不满后,继续生产;
- 如果缓冲区里的产品从无到有,生产者就得通知一下消费者,告诉它可以来消费了;
- 消费者持续消费,直到缓冲区空,空时阻塞;缓冲区不空后,继续消费;
- 如果缓冲区里的产品从满到不满,消费者也得去通知下生产者,说你可以来生产了。
那我们自己如何实现这个生产者消费者模型呢?
方式一:使用 wait() / notify() 方式
方式二:使用 BlockingQueue 阻塞队列方式
二、wait()/notify() 方式
wait()
和 notify()
都是 Java 中的 Object
类自带的方法,可以用来实现线程间的通信
wait()
方法是用来让当前线程等待,直到有别的线程调用 notify()
将它唤醒,或者我们可以设定一个时间让它自动苏醒。
notify()
方法只能通知一个线程,如果多个线程在等待,那就唤醒任意一个。
notifyAll()
方法是可以唤醒所有等待线程,然后加入同步队列。
具体代码如下:
// 生产者
public class Producer implements Runnable {
private volatile boolean isRunning = true;
private final Vector sharedQueue; // 内存缓冲区
private final int SIZE; // 缓冲区大小
private static AtomicInteger count = new AtomicInteger(); // 总数,原子操作
private static final int SLEEPTIME = 1000;
public Producer(Vector sharedQueue, int SIZE) {
this.sharedQueue = sharedQueue;
this.SIZE = SIZE;
}
@Override
public void run() {
int data;
Random r = new Random();
System.out.println("start producer id = " + Thread.currentThread().getId());
try {
while (isRunning) {
// 模拟延迟
Thread.sleep(r.nextInt(SLEEPTIME));
// 当队列满时阻塞等待
while (sharedQueue.size() == SIZE) {
synchronized (sharedQueue) {
System.out.println("Queue is full, producer " + Thread.currentThread().getId()
+ " is waiting, size:" + sharedQueue.size());
sharedQueue.wait();
}
}
// 队列不满时持续创造新元素
synchronized (sharedQueue) {
data = count.incrementAndGet(); // 构造任务数据
sharedQueue.add(data);
System.out.println("producer create data:" + data + ", size:" + sharedQueue.size());
sharedQueue.notifyAll();
}
}
} catch (InterruptedException e) {
e.printStackTrace();
Thread.currentThread().interrupted();
}
}
public void stop() {
isRunning = false;
}
}
// 消费者
public class Consumer implements Runnable {
private final Vector sharedQueue; // 内存缓冲区
private final int SIZE; // 缓冲区大小
private static final int SLEEPTIME = 1000;
public Consumer(Vector sharedQueue, int SIZE) {
this.sharedQueue = sharedQueue;
this.SIZE = SIZE;
}
@Override
public void run() {
Random r = new Random();
System.out.println("start consumer id = " + Thread.currentThread().getId());
try {
while (true) {
// 模拟延迟
Thread.sleep(r.nextInt(SLEEPTIME));
// 当队列空时阻塞等待
while (sharedQueue.isEmpty()) {
synchronized (sharedQueue) {
System.out.println("Queue is empty, consumer " + Thread.currentThread().getId()
+ " is waiting, size:" + sharedQueue.size());
sharedQueue.wait();
}
}
// 队列不空时持续消费元素
synchronized (sharedQueue) {
System.out.println("consumer consume data:" + sharedQueue.remove(0) + ", size:" + sharedQueue.size());
sharedQueue.notifyAll();
}
}
} catch (InterruptedException e) {
e.printStackTrace();
Thread.currentThread().interrupt();
}
}
}
// 主线程代码
public static void main(String args[]) throws InterruptedException {
// 1.构建内存缓冲区
Vector sharedQueue = new Vector();
int size = 4;
// 2.建立线程池和线程
ExecutorService service = Executors.newCachedThreadPool();
Producer prodThread1 = new Producer(sharedQueue, size);
Producer prodThread2 = new Producer(sharedQueue, size);
Producer prodThread3 = new Producer(sharedQueue, size);
Consumer consThread1 = new Consumer(sharedQueue, size);
Consumer consThread2 = new Consumer(sharedQueue, size);
Consumer consThread3 = new Consumer(sharedQueue, size);
service.execute(prodThread1);
service.execute(prodThread2);
service.execute(prodThread3);
service.execute(consThread1);
service.execute(consThread2);
service.execute(consThread3);
// 3.睡一会儿然后尝试停止生产者
Thread.sleep(10 * 1000);
prodThread1.stop();
prodThread2.stop();
prodThread3.stop();
// 4.再睡一会儿关闭线程池
Thread.sleep(3000);
service.shutdown();
}
// 执行结果
start producer id = 13
start consumer id = 16
start consumer id = 17
start producer id = 12
start consumer id = 15
start producer id = 14
producer create data:1, size:1
producer create data:2, size:2
consumer consume data:1, size:1
我们解释一下上面的代码吧
- 创造数据:生产者-消费者解决的问题就是数据在多线程间的共享,所以我们首要关心的问题就应该是数据,我们这里采用的是使用一个
AtomicInteger
类来为我们创造数据,使用它的好处是该类是一个保证原子操作的类,我们使用其中的incrementAndGet()
方法不仅能够保证线程安全,还可以达到一个计数的效果,所以是一个既简单又实用的选择,当然也可以使用其他的数据来代替,这里注意的是要保证该类在内存中只存在一份,使用static
修饰; - 内存缓冲区:要保证在多线程环境下内存缓冲区的安全,所以我们考虑使用简单的
Vector
类来作为我们的内存缓冲区,并且使用final
修饰保证内存缓冲区的唯一,然后的话我们需要判断队列是否满,需要手动添加一个标识缓冲区大小的变量SIZE
,注意也是final
修饰; - 模拟延迟:这里主要模拟的是一个网络延迟,我们首先定义了一个
SLEEPTIME
的延迟范围,注意使用的是static final
修饰,然后使用Random()
类的nextInt()
方法来随机选取一个该范围内的值来模拟网络环境中的延迟; - 停止方法:首先需要知道在
Thread
类中有一个弃用的stop()
方法,我们自己增加一个标志位isRunning
来完成我们自己的stop()
功能,需要注意的是使用volatile
来修饰,保证该标志位的可见性; - 错误处理:当捕获到错误时,我们应该使用
Thread
类中的interrupted()
方法来终止当前的进程; - 消息提示:我们主要是要在控制台输出该生产者的信息,包括当前队列的状态,大小,当前线程的生产者信息等,注意的是信息格式的统一;
三、 BlockingQueue 阻塞队列方式
大家还是直接看代码吧:
public class Producer implements Runnable {
private volatile boolean isRunning = true;
private BlockingQueue<Integer> queue; // 内存缓冲区
private static AtomicInteger count = new AtomicInteger(); // 总数,原子操作
private static final int SLEEPTIME = 1000;
public Producer(BlockingQueue<Integer> queue) {
this.queue = queue;
}
@Override
public void run() {
int data;
Random r = new Random();
System.out.println("start producer id = " + Thread.currentThread().getId());
try {
while (isRunning) {
// 模拟延迟
Thread.sleep(r.nextInt(SLEEPTIME));
// 往阻塞队列中添加数据
data = count.incrementAndGet(); // 构造任务数据
System.out.println("producer " + Thread.currentThread().getId() + " create data:" + data
+ ", size:" + queue.size());
if (!queue.offer(data, 2, TimeUnit.SECONDS)) {
System.err.println("failed to put data:" + data);
}
}
} catch (InterruptedException e) {
e.printStackTrace();
Thread.currentThread().interrupted();
}
}
public void stop() {
isRunning = false;
}
}
public class Consumer implements Runnable {
private BlockingQueue<Integer> queue; // 内存缓冲区
private static final int SLEEPTIME = 1000;
public Consumer(BlockingQueue<Integer> queue) {
this.queue = queue;
}
@Override
public void run() {
int data;
Random r = new Random();
System.out.println("start consumer id = " + Thread.currentThread().getId());
try {
while (true) {
// 模拟延迟
Thread.sleep(r.nextInt(SLEEPTIME));
// 从阻塞队列中获取数据
if (!queue.isEmpty()) {
data = queue.take();
System.out.println("consumer " + Thread.currentThread().getId() + " consume data:" + data
+ ", size:" + queue.size());
} else {
System.out.println("Queue is empty, consumer " + Thread.currentThread().getId()
+ " is waiting, size:" + queue.size());
}
}
} catch (InterruptedException e) {
e.printStackTrace();
Thread.currentThread().interrupt();
}
}
}
public static void main(String args[]) throws InterruptedException {
// 1.构建内存缓冲区
BlockingQueue<Integer> queue = new LinkedBlockingDeque<>();
// 2.建立线程池和线程
ExecutorService service = Executors.newCachedThreadPool();
Producer prodThread1 = new Producer(queue);
Producer prodThread2 = new Producer(queue);
Producer prodThread3 = new Producer(queue);
Consumer consThread1 = new Consumer(queue);
Consumer consThread2 = new Consumer(queue);
Consumer consThread3 = new Consumer(queue);
service.execute(prodThread1);
service.execute(prodThread2);
service.execute(prodThread3);
service.execute(consThread1);
service.execute(consThread2);
service.execute(consThread3);
// 3.睡一会儿然后尝试停止生产者
Thread.sleep(10 * 1000);
prodThread1.stop();
prodThread2.stop();
prodThread3.stop();
// 4.再睡一会儿关闭线程池
Thread.sleep(3000);
service.shutdown();
}
// 运行结果
start producer id = 13
start consumer id = 15
start consumer id = 16
start producer id = 14
start producer id = 12
start consumer id = 17
Queue is empty, consumer 16 is waiting, size:0
Queue is empty, consumer 16 is waiting, size:0
Queue is empty, consumer 16 is waiting, size:0
Queue is empty, consumer 16 is waiting, size:0
producer 13 create data:1, size:0
consumer 15 consume data:1, size:0
producer 12 create data:2, size:0
好了,手撕阻塞队列就讲到这里了!!!我们下期见
引用:
https://www.cnblogs.com/wmyskxz/p/9538177.html#_label0
https://segmentfault.com/a/1190000024444906
https://segmentfault.com/a/1190000018169793?utm_source=sf-similar-article