服务器之家:专注于服务器技术及软件下载分享
分类导航

PHP教程|ASP.NET教程|Java教程|ASP教程|编程技术|正则表达式|C/C++|IOS|C#|Swift|Android|JavaScript|易语言|

服务器之家 - 编程语言 - Java教程 - RxJava的消息发送和线程切换实现原理

RxJava的消息发送和线程切换实现原理

2021-06-10 14:16Jack921 Java教程

这篇文章主要介绍了RxJava的消息发送和线程切换实现原理,小编觉得挺不错的,现在分享给大家,也给大家做个参考。一起跟随小编过来看看吧

rxjava是一个在java虚拟机上的响应式扩展,通过使用可观察的序列将异步和基于事件的程序组合起来的一个库。

它扩展了观察者模式来支持数据/事件序列,并且添加了操作符,这些操作符允许你声明性地组合序列,同时抽象出要关注的问题:比如低级线程、同步、线程安全和并发数据结构等。

rxjava相信大家都非常了解吧,今天分享一下rxjava的消息发送和线程源码的分析。最后并分享一个相关demo,让大家更加熟悉我们天天都在用的框架。

消息订阅发送

首先让我们看看消息订阅发送最基本的代码组成:

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
observable observable = observable.create(new observableonsubscribe<string>() {
     @override
     public void subscribe(observableemitter<string> emitter) throws exception {
       emitter.onnext("jack1");
       emitter.onnext("jack2");
       emitter.onnext("jack3");
       emitter.oncomplete();
     }
   });
 
   observer<string> observer = new observer<string>() {
     @override
     public void onsubscribe(disposable d) {
       log.d(tag, "onsubscribe");
     }
 
     @override
     public void onnext(string s) {
       log.d(tag, "onnext : " + s);
     }
 
     @override
     public void onerror(throwable e) {
       log.d(tag, "onerror : " + e.tostring());
     }
 
     @override
     public void oncomplete() {
       log.d(tag, "oncomplete");
     }
   };
 
   observable.subscribe(observer);

代码很简单,observable为被观察者,observer为观察者,然后通过observable.subscribe(observer),把观察者和被观察者关联起来。被观察者发送消息(emitter.onnext("内容")),观察者就可以在onnext()方法里回调出来。

我们先来看observable,创建是用observable.create()方法进行创建,源码如下:

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
public static <t> observable<t> create(observableonsubscribe<t> source) {
  objecthelper.requirenonnull(source, "source is null");
  return rxjavaplugins.onassembly(new observablecreate<t>(source));
}
 
public static <t> t requirenonnull(t object, string message) {
  if (object == null) {
     throw new nullpointerexception(message);
  }
  return object;
 }
 
public static <t> observable<t> onassembly(@nonnull observable<t> source) {
  function<? super observable, ? extends observable> f = onobservableassembly;
  if (f != null) {
     return apply(f, source);
  }
  return source;
}

可以看出,create()方法里最主要的还是创建用observableonsubscribe传入创建了一个observablecreate对象并且保存而已。

?
1
2
3
4
5
6
7
8
public final class observablecreate<t> extends observable<t> {
  final observableonsubscribe<t> source;
 
  public observablecreate(observableonsubscribe<t> source) {
    this.source = source;
  }
 
}

接着是创建observer,这比较简单只是单纯创建一个接口对象而已

?
1
2
3
4
5
6
7
8
9
public interface observer<t> {
  void onsubscribe(@nonnull disposable d);
 
  void onnext(@nonnull t t);
 
  void onerror(@nonnull throwable e);
  
  void oncomplete();
}

订阅发送消息

observable.subscribe(observer)的subscribe方法如下:

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
public final void subscribe(observer<? super t> observer) {
  objecthelper.requirenonnull(observer, "observer is null");
  try {
    observer = rxjavaplugins.onsubscribe(this, observer);
    objecthelper.requirenonnull(observer, "plugin returned null observer");
    subscribeactual(observer);
  } catch (nullpointerexception e) { // nopmd
    throw e;
  } catch (throwable e) {
    exceptions.throwiffatal(e);
    rxjavaplugins.onerror(e);
    nullpointerexception npe = new nullpointerexception("actually not, but can't throw other exceptions due to rs");
    npe.initcause(e);
    throw npe;
  }
}
 
//objecthelper.requirenonnull()方法
public static <t> t requirenonnull(t object, string message) {
  if (object == null) {
     throw new nullpointerexception(message);
  }
  return object;
}
 
//rxjavaplugins.onsubscribe()方法
public static <t> observer<? super t> onsubscribe(@nonnull observable<t> source, @nonnull observer<? super t> observer) {
  bifunction<? super observable, ? super observer, ? extends observer> f = onobservablesubscribe;
  if (f != null) {
    return apply(f, source, observer);
  }
  return observer;
}

从上面源码可以看出requirenonnull()只是做非空判断而已,而rxjavaplugins.onsubscribe()也只是返回最终的观察者而已。所以关键代码是抽象方法subscribeactual(observer);那么subscribeactual对应哪个代码段呢?

还记得observable.create()创建的observablecreate类吗,这就是subscribeactual()具体实现类,源码如下:

?
1
2
3
4
5
6
7
8
9
10
protected void subscribeactual(observer<? super t> observer) {
  createemitter<t> parent = new createemitter<t>(observer);
  observer.onsubscribe(parent);
  try {
    source.subscribe(parent);
  } catch (throwable ex) {
    exceptions.throwiffatal(ex);
    parent.onerror(ex);
  }
}

从上面的代码可以看出,首先创建了一个createemitter对象并传入observer,然后回到observer的onsubscribe()方法,而source就是我们之前创建observablecreate传入的observableonsubscribe对象。

?
1
2
3
4
class createemitter<t> extends atomicreference<disposable>
  implements observableemitter<t>, disposable {
 
 }

而createemitter又继承observableemitter接口,又回调observableonsubscribe的subscribe方法,对应着我们的:

?
1
2
3
4
5
6
7
8
9
observable observable = observable.create(new observableonsubscribe<string>() {
   @override
   public void subscribe(observableemitter<string> emitter) throws exception {
      emitter.onnext("jack1");
      emitter.onnext("jack2");
      emitter.onnext("jack3");
      emitter.oncomplete();
   }
});

当它发送消息既调用emitter.onnext()方法时,既调用了createemitter的onnext()方法:

?
1
2
3
4
5
6
7
8
9
public void onnext(t t) {
  if (t == null) {
    onerror(new nullpointerexception("onnext called with null. null values are generally not allowed in 2.x operators and sources."));
    return;
  }
  if (!isdisposed()) {
    observer.onnext(t);
  }
}

可以看到最终又回调了观察者的onnext()方法,把被观察者的数据传输给了观察者。有人会问

isdisposed()是什么意思,是判断要不要终止传递的,我们看emitter.oncomplete()源码:

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
public void oncomplete() {
  if (!isdisposed()) {
    try {
      observer.oncomplete();
    } finally {
      dispose();
    }
  }
}
 
public static boolean dispose(atomicreference<disposable> field) {
    disposable current = field.get();
    disposable d = disposed;
    if (current != d) {
      current = field.getandset(d);
      if (current != d) {
        if (current != null) {
          current.dispose();
        }
        return true;
      }
    }
    return false;
 }
 
public static boolean isdisposed(disposable d) {
    return d == disposed;
}

dispose()方法是终止消息传递,也就付了个disposed常量,而isdisposed()方法就是判断这个常量而已。这就是整个消息订阅发送的过程,用的是观察者模式。

线程切换

在上面模板代码的基础上,线程切换只是改变了如下代码:

?
1
2
3
observable.subscribeon(schedulers.io())
     .observeon(androidschedulers.mainthread())
     .subscribe(observer);

下面我们对线程切换的源码进行一下分析,分为两部分:subscribeon()和observeon()

subscribeon()

首先是subscribeon()源码如下:

?
1
2
3
4
public final observable<t> subscribeon(scheduler scheduler) {
  objecthelper.requirenonnull(scheduler, "scheduler is null");
  return rxjavaplugins.onassembly(new observablesubscribeon<t>(this, scheduler));
}

我们传进去了一个scheduler类,scheduler是一个调度类,能够延时或周期性地去执行一个任务。

scheduler有如下类型:

 

类型 使用方式 含义 使用场景
ioscheduler schedulers.io() io操作线程 读写sd卡文件,查询数据库,访问网络等io密集型操作
newthreadscheduler schedulers.newthread() 创建新线程 耗时操作等
singlescheduler schedulers.single() 单例线程 只需一个单例线程时
computationscheduler schedulers.computation() cpu计算操作线程 图片压缩取样、xml,json解析等cpu密集型计算
trampolinescheduler schedulers.trampoline() 当前线程 需要在当前线程立即执行任务时
handlerscheduler androidschedulers.mainthread() android主线程 更新ui等

 

接着就没什么了,只是返回一个observablesubscribeon对象而已。

observeon()

首先看源码如下:

?
1
2
3
4
5
6
7
8
9
public final observable<t> observeon(scheduler scheduler) {
  return observeon(scheduler, false, buffersize());
}
 
public final observable<t> observeon(scheduler scheduler, boolean delayerror, int buffersize) {
  objecthelper.requirenonnull(scheduler, "scheduler is null");
  objecthelper.verifypositive(buffersize, "buffersize");
  return rxjavaplugins.onassembly(new observableobserveon<t>(this, scheduler, delayerror, buffersize));
}

这里也是没什么,只是最终返回一个observableobserveon对象而已。

接着还是像原来那样调用subscribe()方法进行订阅,看起来好像整体变化不大,就是封装了一些对象而已,不过着恰恰是rxjava源码的精华,当他再次调用subscribeactual()方法时,已经不是之前的observablecreate()里subscribeactual方法了,而是最先调用observableobserveon的subscribeactual()方法,对应源码如下:

?
1
2
3
4
5
6
7
8
protected void subscribeactual(observer<? super t> observer) {
  if (scheduler instanceof trampolinescheduler) {
    source.subscribe(observer);
  } else {
    scheduler.worker w = scheduler.createworker();
    source.subscribe(new observeonobserver<t>(observer, w, delayerror, buffersize));
  }
}

在这里有两点要讲,一点是observeonobserver是执行观察者的线程,后面还会详解,然后就是source.subscribe,这个source.subscribe调的是observablesubscribeon的subscribe方法,而subscribe方法因为继承的也是observable,是observable里的方法,所以和上面的observablecreate一样的方法,所以会调用observablesubscribeon里的subscribeactual()方法,对应的代码如下:

?
1
2
3
4
5
public void subscribeactual(final observer<? super t> s) {
  final subscribeonobserver<t> parent = new subscribeonobserver<t>(s);
  s.onsubscribe(parent);
  parent.setdisposable(scheduler.scheduledirect(new subscribetask(parent)));
}

上面代码中,首先把observeonobserver返回给来的用subscribeonobserver“包装”起来,然后在回调observer的onsubscribe(),就是对应模板代码的onsubscribe()方法。

接着看subscribetask类的源码:

?
1
2
3
4
5
6
7
8
9
10
final class subscribetask implements runnable {
  private final subscribeonobserver<t> parent;
  subscribetask(subscribeonobserver<t> parent) {
    this.parent = parent;
  }
  @override
  public void run() {
    source.subscribe(parent);
  }
}

其中的source.subscribe(parent),就是我们执行子线程的回调方法,对应我们模板代码里的被观察者的subscribe()方法。它放在run()方法里,并且继承runnable,说明这个类主要是线程运行。接着看scheduler.scheduledirect()方法对应的源码如下:

?
1
2
3
4
5
6
7
8
9
10
11
public disposable scheduledirect(@nonnull runnable run) {
  return scheduledirect(run, 0l, timeunit.nanoseconds);
}
 
public disposable scheduledirect(@nonnull runnable run, long delay, @nonnull timeunit unit) {
  final worker w = createworker();
  final runnable decoratedrun = rxjavaplugins.onschedule(run);
  disposetask task = new disposetask(decoratedrun, w);
  w.schedule(task, delay, unit);
  return task;
}

在这里,createworker()也是一个抽象方法,调用的是我们的调度类对应的schedulers类里面的方法,这里是ioscheduler类,

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
public final class ioscheduler extends scheduler{
 
  final atomicreference<cachedworkerpool> pool;
 
  //省略....
 
  public worker createworker() {
    return new eventloopworker(pool.get());
  }
 
  static final class eventloopworker extends scheduler.worker {
    private final compositedisposable tasks;
    private final cachedworkerpool pool;
    private final threadworker threadworker;
 
    final atomicboolean once = new atomicboolean();
 
    eventloopworker(cachedworkerpool pool) {
      this.pool = pool;
      this.tasks = new compositedisposable();
      this.threadworker = pool.get();
    }
 
    //省略....
 
    @nonnull
    @override
    public disposable schedule(@nonnull runnable action, long delaytime, @nonnull timeunit unit) {
      if (tasks.isdisposed()) {
        // don't schedule, we are unsubscribed
        return emptydisposable.instance;
      }
      return threadworker.scheduleactual(action, delaytime, unit, tasks);
    }
  }
 
}
 
 static final class cachedworkerpool implements runnable {
 
  //省略....
 
  threadworker get() {
    if (allworkers.isdisposed()) {
      return shutdown_thread_worker;
    }
    while (!expiringworkerqueue.isempty()) {
      threadworker threadworker = expiringworkerqueue.poll();
      if (threadworker != null) {
        return threadworker;
      }
    }
 
    threadworker w = new threadworker(threadfactory);
    allworkers.add(w);
    return w;
   }
   //省略....
}

这就是ioscheduler的createworker()的方法,其实最主要的意思就是获取线程池,以便于生成子线程,让subscribetask()可以运行。然后直接调用 w.schedule(task, delay, unit)方法让它在线程池里执行。上面中那threadworker的源码如下:

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
static final class threadworker extends newthreadworker {
  private long expirationtime;
  threadworker(threadfactory threadfactory) {
    super(threadfactory);
    this.expirationtime = 0l;
  }
 
  //省略代码....
 }
 
public class newthreadworker extends scheduler.worker implements disposable {
  private final scheduledexecutorservice executor;
 
  public newthreadworker(threadfactory threadfactory) {
    executor = schedulerpoolfactory.create(threadfactory);
  }
 
  public scheduledrunnable scheduleactual(final runnable run, long delaytime, @nonnull timeunit unit, @nullable disposablecontainer parent) {
    runnable decoratedrun = rxjavaplugins.onschedule(run);
 
    scheduledrunnable sr = new scheduledrunnable(decoratedrun, parent);
 
    if (parent != null) {
      if (!parent.add(sr)) {
        return sr;
      }
    }
 
    future<?> f;
    try {
      if (delaytime <= 0) {
        f = executor.submit((callable<object>)sr);
      } else {
        f = executor.schedule((callable<object>)sr, delaytime, unit);
      }
      sr.setfuture(f);
    } catch (rejectedexecutionexception ex) {
      if (parent != null) {
        parent.remove(sr);
      }
      rxjavaplugins.onerror(ex);
    }
 
    return sr;
  }
}

可以看到,这就调了原始的javaapi来进行线程池操作。

然后最后一环在子线程调用source.subscribe(parent)方法,然后回调刚开始创建的observablecreate的subscribeactual(),既:

?
1
2
3
4
5
6
7
8
9
10
protected void subscribeactual(observer<? super t> observer) {
    createemitter<t> parent = new createemitter<t>(observer);
    observer.onsubscribe(parent);
    try {
      source.subscribe(parent);
    } catch (throwable ex) {
      exceptions.throwiffatal(ex);
      parent.onerror(ex);
    }
}

进行消息的订阅绑定。

当我们在调用 emitter.onnext(内容)时,是在io线程里的,那回调的onnext()又是什么时候切换的?那就是前面为了整个流程流畅性没讲的在observeon()里的observeonobserver是执行观察者的线程的过程。

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
class observeonobserver<t> extends basicintqueuedisposable<t>
  implements observer<t>, runnable {
 
    //省略代码....
 
    observeonobserver(observer<? super t> actual, scheduler.worker worker, boolean delayerror, int buffersize) {
      this.actual = actual;
      this.worker = worker;
      this.delayerror = delayerror;
      this.buffersize = buffersize;
    }
 
    @override
    public void onsubscribe(disposable s) {
      if (disposablehelper.validate(this.s, s)) {
        this.s = s;
        if (s instanceof queuedisposable) {
          @suppresswarnings("unchecked")
          queuedisposable<t> qd = (queuedisposable<t>) s;
          int m = qd.requestfusion(queuedisposable.any | queuedisposable.boundary);
          if (m == queuedisposable.sync) {
            sourcemode = m;
            queue = qd;
            done = true;
            actual.onsubscribe(this);
            schedule();
            return;
          }
          if (m == queuedisposable.async) {
            sourcemode = m;
            queue = qd;
            actual.onsubscribe(this);
            return;
          }
        }
        queue = new spsclinkedarrayqueue<t>(buffersize);
        actual.onsubscribe(this);
      }
    }
 
    @override
    public void onnext(t t) {
      if (done) {
        return;
      }
      if (sourcemode != queuedisposable.async) {
        queue.offer(t);
      }
      schedule();
    
 
    void schedule() {
      if (getandincrement() == 0) {
        worker.schedule(this);
      }
    }
    //省略代码....
  }

当调用emitter.onnext(内容)方法,会调用上面的onnext()方法,然后在这个方法里会把数据压入一个队列,然后执行worker.schedule(this)方法,work是什么呢,还记得androidschedulers.mainthread()吗,这个对应这个handlerscheduler这个类,所以createworker()对应着:

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
private static final class mainholder {
    static final scheduler default = new handlerscheduler(new handler(looper.getmainlooper()));
}
 
 
public worker createworker() {
  return new handlerworker(handler);
}
 
private static final class handlerworker extends worker {
    private final handler handler;
    private volatile boolean disposed;
 
    handlerworker(handler handler) {
      this.handler = handler;
    }
 
    @override
    public disposable schedule(runnable run, long delay, timeunit unit) {
      if (run == null) throw new nullpointerexception("run == null");
      if (unit == null) throw new nullpointerexception("unit == null");
      if (disposed) {
        return disposables.disposed();
      }
      run = rxjavaplugins.onschedule(run);
      scheduledrunnable scheduled = new scheduledrunnable(handler, run);
      message message = message.obtain(handler, scheduled);
      message.obj = this; // used as token for batch disposal of this worker's runnables.
      handler.sendmessagedelayed(message, unit.tomillis(delay));
      if (disposed) {
        handler.removecallbacks(scheduled);
        return disposables.disposed();
      }
      return scheduled;
    }
}

在next()方法里,运用android自带的handler消息机制,通过把方法包裹在message里,同通过handler.sendmessagedelayed()发送消息,就会在ui线程里回调next()方法,从而实现从子线程切换到android主线程的操作。我们在主线程拿到数据就可以进行各种在主线程的操作了。

总结一下:

RxJava的消息发送和线程切换实现原理

observablecreate 一> observablesubscribeon 一> observableobserveon为初始化顺序

当调用observable.subscribe(observer)时的执行顺序
observableobserveon 一> observablesubscribeon 一> observablecreate

当发送消息的执行顺序
observablecreate 一> observablesubscribeon 一> observableobserveon

以上就是消息订阅和线程切换的源码的所有讲解了。

为了让你们理解更清楚,我仿照rxjava写了大概的消息订阅和线程切换的最基本代码和基本功能,以帮助你们理解

https://github.com/jack921/rxjava2demo

以上就是本文的全部内容,希望对大家的学习有所帮助,也希望大家多多支持服务器之家。

原文链接:https://www.jianshu.com/p/264b68fd96fa

延伸 · 阅读

精彩推荐