框架设计|自己撸一个RxJava 可好？（上）

开发者技术前线

发布于 2020-11-23 15:21:43

3590

发布于 2020-11-23 15:21:43

文章被收录于专栏：开发者技术前线

RxJava 拥有繁多的 API 和复杂的逻辑链，学习复杂的知识，一般从整体再到具体，为了学习 RxJava 的原理，参考其源码，自己动手实现一个简化的 RxJava，旨在理解调用链。

阅读本文，建议先下载代码 LittleRx，毕竟在IDE里阅读代码比在网页上要清晰得多，也可以看下打印的日志

RxJava最主要的4个类：Observable、OnSubscribe、Operator、Subscriber

1、创建一个Observable，然后订阅

Observable
        .create(new OnSubscribe<Integer>() {           
           @Override
            public void call(Subscriber<? super Integer> subscriber) {
                subscriber.onNext(1);
            }
        })
        .subscribe(new Subscriber<Integer>() {            
            @Override
            public void onNext(Integer integer) {
                System.out.println(integer);
            }
        });public class Observable<T> {    
           private OnSubscribe<T> onSubscribe;   
           private Observable(OnSubscribe<T> onSubscribe) {        
           this.onSubscribe = onSubscribe;
    }    
    public final void subscribe(Subscriber<? super T> subscriber) {
         onSubscribe.call(subscriber);
    }   
     public static <T> Observable<T> create(OnSubscribe<T> onSubscribe) {        
       return new Observable<>(onSubscribe);
    }
}

这里可以看出

subscribe(subscriber)-->onSubscribe.call(subscriber)，

所以没有订阅动作就不会触发 OnSubscribe.call()。

2、map 和 lift

Observable
        .create(new OnSubscribe<Integer>() {
            @Override            
            public void call(Subscriber<? super Integer> subscriber) {
                subscriber.onNext(1);
            }
        })
        .map(new Func1<Integer, String>() {
            @Override            
            public String call(Integer integer) {                
                return "map" + integer;
            }
        })
        .subscribe(new Subscriber<String>() {
            @Override            
            public void onNext(String s) {
                System.out.println(s);
            }
        });

非链式的写法

OnSubscribe<Integer> onSubscribe = new OnSubscribe<>();
Observable<Integer> observable = Observable.create(onSubscribe);
Func1<Integer, String> func = new Func1<>();
Observable<String> observable2 = observable.map(func);
Subscriber<String> subscriber = new Subscriber<>();
observable2.subscribe(subscriber);

create() 跟之前一样，那么 map() 做了什么

public final <R> Observable<R> map(Func1<? super T, ? extends R> func) {    
     return lift(new OperatorMap<T, R>(func));
}
public final <R> Observable<R> lift(final Operator<? extends R, ? super T> operator) {    
    return new Observable<R>(new OnSubscribeLift<T, R>(onSubscribe, operator));
}

lift() 根据上一个 observable 的 onSubscribe 创建一个新的 OnSubscribeLift 返回一个新的 observable2，上面我们说过 subscribe(subscriber)-->onSubscribe.call(subscriber)，所以我们接着看 OnSubscribeLift

public class OnSubscribeLift<T, R> implements OnSubscribe<R> {    
   final OnSubscribe<T> parent;    
   final Operator<? extends R, ? super T> operator;

    public OnSubscribeLift(OnSubscribe<T> parent, Operator<? extends R, ? super T> operator) {        
      this.parent = parent;        
      this.operator = operator;
    }    
    // 先不用关心具体实现，下面讲到再看
      @Override
      public void call(Subscriber<? super R> r) {
           Subscriber<? super T> st = operator.call(r); 
          // 这个 operator 就是 OperatorMap
          parent.call(st); 
        // parent 就是第一个 observable 的 onSubscribe
    }
}

再看下 OperatorMap

public final class OperatorMap<T, R> implements Operator<R, T> {    
    final Func1<? super T, ? extends R> transformer;    
    public OperatorMap(Func1<? super T, ? extends R> transformer) {        t
         his.transformer = transformer;
    }    
    // 先不用关心具体实现，下面讲到再看
    @Override
    public Subscriber<? super T> call(final Subscriber<? super R> o) {        
         return new MapSubscriber<T, R>(o, transformer);
    }    
    private class MapSubscriber<T, R> extends Subscriber<T> {        
    private Subscriber<? super R> actual;        
    private Func1<? super T, ? extends R> transformer;        
    public MapSubscriber(Subscriber<? super R> o, Func1<? super T, ? extends R> transformer) {            
            this.actual = o;            
            this.transformer = transformer;
        }        
        // 先不用关心具体实现，下面讲到再看
        @Override
        public void onNext(T t) {
            R r = transformer.call(t);
            actual.onNext(r);
        }
    }
}

我们把 map() 和 lift() 都去掉，使用最基本的类来实现

OnSubscribe<Integer> onSubscribe = new OnSubscribe<>();
Observable<Integer> observable = new Observable<>(onSubscribe);
Func1<Integer, String> func = new Func1<>();

OperatorMap<Integer, String> operatorMap = new OperatorMap<>(func);

OnSubscribeLift<Integer, String> onSubscribe2 = new OnSubscribeLift<>(onSubscribe, operatorMap);

Observable<String> observable2 = new Observable<>(onSubscribe2);

Subscriber<String> subscriber = new Subscriber<>();
observable2.subscribe(subscriber);

到这里，清楚了如何把第一个 Observable<Integer> 转成 Observable<String>，包括 OnSubscribe<Integer> onSubscribe 和 OnSubscribeLift<Integer, String> onSubscribe2 的关系

那么最终的 subscribe() 如何调用到第一个 observable.call(Subscriber<Integer>) 里面的 Subscriber<Integer>.onNext(Integer) 又如何调用到最终的订阅者 subscriber<String>().onNext(String)

1) observable2.subscribe(subscriber) --> 2) onSubscribe2.call(subscriber) 即 OnSubscribeLift.call(subscriber) --> 3) Subscriber<Integer> st = operatorMap.call(subscriber) 即 4) Subscriber<Integer> st = new MapSubscriber<Integer, String>(subscriber, func) 5) parent.call(st) 即 onSubscribe.call(st) --> 6) st.onNext(1) 即 MapSubscriber.onNext(1) --> 7) String string = func.call(1) 8) subscriber.onNext(string)

至此 Observable.create().map().subscribe() 的调用链就分析完了很多操作符本质都是 lift()，以此类推，lift() 2次

3、subscribeOn

Scheduler 内部比较繁杂，我们简化下，把 subscribeOn(Scheduler) 简化成 subscribeOnIO()

Observable
        .create(new OnSubscribe<Integer>() {
            @Override            
            public void call(Subscriber<? super Integer> subscriber) {
                subscriber.onNext(1);
            }
        })
        .subscribeOnIO()
        .subscribe(new Subscriber<Integer>() {
            @Override            
            public void onNext(Integer integer) {
                System.out.println(integer);
            }
        });

如何实现 subscribeOnIO() 让第一个 observable 的 onSubscribe 运行在子线程

public final Observable<T> subscribeOnIO() {    
    return create(new OnSubscribeOnIO<T>(this));
}
public final class OnSubscribeOnIO<T> implements OnSubscribe<T> {    
    private static ExecutorService threadPool = Executors.newCachedThreadPool();    
    final Observable<T> source;    
    public OnSubscribeOnIO(Observable<T> source) {        
         this.source = source;
    }    
    @Override
    public void call(final Subscriber<? super T> subscriber) {
        Runnable runnable = new Runnable() {            
             @Override
            public void run() {
                source.subscribe(subscriber);
                 // --> onSubscribe.call(subscriber) --> subscriber.onNext(1)
            }
        };
        threadPool.submit(runnable);
    }
}

从上面看出 subscribeOnIO() 新建了一个线程并加入 CachedThreadPool，在子线程里订阅上一个 Observable，后续的调用都在这个线程里完成

再考虑下复杂点的，加入 map()

Observable
        .create(new OnSubscribe<Integer>() {
            @Override            
            public void call(Subscriber<? super Integer> subscriber) {
                System.out.println(Thread.currentThread());
                subscriber.onNext(1);
            }
        })
        .map(new Func1<Integer, String>() {
            @Override            
            public String call(Integer integer) {
                System.out.println(Thread.currentThread());                
                return "map" + integer;
            }
        })
        .subscribeOnIO()
        .subscribe(new Subscriber<String>() {
            @Override            
            public void onNext(String s) {
                System.out.println(Thread.currentThread());
                System.out.println(s);
            }
        });

非链式的写法

OnSubscribe<Integer> onSubscribe = new OnSubscribe<>();
Observable<Integer> observable = new Observable<>(onSubscribe);
Func1<Integer, String> func = new Func1<>();
OperatorMap<Integer, String> operatorMap = new OperatorMap<>(func);
OnSubscribeLift<Integer, String> onSubscribe2 = new OnSubscribeLift<>(onSubscribe, operatorMap);
Observable<String> observable2 = new Observable<>(onSubscribe2);

OnSubscribeOnIO<String> onSubscribe3 = new OnSubscribeOnIO(observable2);
Observable<String> observable3 = new Observable<>(onSubscribe3);

Subscriber<String> subscriber = new Subscriber<>();
observable3.subscribe(subscriber);

1) observable3.subscribe(subscriber) --> 2) onSubscribe3.call(subscriber) 即 OnSubscribeOnIO.call(subscriber) --> 3) 子线程 new Runnable(){} --> observable2.subscribe(subscriber) 4) onSubscribe2.call(subscriber) 即 OnSubscribeLift.call(subscriber) --> 5) Subscriber<Integer> st = operatorMap.call(subscriber) 即 6) Subscriber<Integer> st = new MapSubscriber<Integer, String>(subscriber, func) 7) parent.call(st) 即 onSubscribe.call(st) --> 8) st.onNext(1) 即 MapSubscriber.onNext(1) --> 9) String string = func.call(1) 10) subscriber.onNext(string)

那要是把 map() 与 subscribeOnIO() 换下位置呢

OnSubscribe<Integer> onSubscribe = new OnSubscribe<>();
Observable<Integer> observable = new Observable<>(onSubscribe);

OnSubscribeOnIO<Integer> onSubscribe2 = new OnSubscribeOnIO(observable);
Observable<Integer> observable2 = new Observable<>(onSubscribe2);

Func1<Integer, String> func = new Func1<>();
OperatorMap<Integer, String> operatorMap = new OperatorMap<>(func);
OnSubscribeLift<Integer, String> onSubscribe3 = new OnSubscribeLift<>(onSubscrib2, operatorMap);
Observable<String> observable3 = new Observable<>(onSubscribe3);
Subscriber<String> subscriber = new Subscriber<>();
observable3.subscribe(subscriber);

1) observable3.subscribe(subscriber) --> 2) onSubscribe3.call(subscriber) 即 OnSubscribeLift.call(subscriber) --> 3) Subscriber<Integer> st = operatorMap.call(subscriber) 即 4) Subscriber<Integer> st = new MapSubscriber<Integer, String>(subscriber, func) 5) parent.call(st) 即 onSubscribe2.call(st) 即 OnSubscribeOnIO.call(st) --> 6) 子线程 new Runnable(){} --> observable.subscribe(st) --> 7) onSubscribe.call(st) --> 7) st.onNext(1) 即 MapSubscriber.onNext(1) --> 8) String string = func.call(1) 9) subscriber.onNext(string)

看得出来，不管 subscribeOnIO() 在哪，第一个 onSubscribe.call() 总是运行在子线程

4、observeOn

先看下 demo 最终写法

Handler handler = new Handler();

Observable
        .create(new OnSubscribe<Integer>() {
            @Override            
            public void call(Subscriber<? super Integer> subscriber) {
                subscriber.onNext(1);
            }
        })
        .observeOn(handler)
        .map(new Func1<Integer, String>() {
            @Override            
            public String call(Integer integer) {                
               return "map" + integer;
            }
        })
        .subscribeOnIO()
        .subscribe(new Subscriber<String>() {
            @Override            
            public void onNext(String s) {
                System.out.println(s);
            }
        });

handler.loop(); 
//队列没有消息时会挂起当前线程，直到收到新的消息

同样我们也自己实现一个简单的可以切换回主线程的 observeOn(Handler)

public class Observable<T> {
    ...    public final Observable<T> observeOn(Handler handler) {        return lift(new OperatorObserveOn<T>(handler));
    }
}

OperatorObserveOn

public final class OperatorObserveOn<T> implements Operator<T, T> {    
   private Handler handler;    
   public OperatorObserveOn(Handler handler) {        
       this.handler = handler;
    }    
    @Override
    public Subscriber<? super T> call(final Subscriber<? super T> subscriber) {
        Subscriber<T> s = new Subscriber<T>() {       
         @Override
         public void onNext(final T t) {                
            handler.post(new Runnable() {                    
         @Override
         public void run() {
             subscriber.onNext(t);
                    }
                });
            }
        };        
        return s;
    }
}

自定义Handler

public class Handler {    
    private ArrayBlockingQueue<Runnable> queue = new ArrayBlockingQueue<Runnable>(10);    
    public void loop() {        
    for (; ; ) {
            Runnable runnable;            
            try {
                runnable = queue.take();
                // 没有数据则一直阻塞，直到有数据自动唤醒
            } catch (InterruptedException e) {                         
                return;
            }            
            if (runnable == null) {                
                 return;
            }
            runnable.run();
        }
    }    
     public void post(Runnable runnable) {       
      try {           
         queue.put(runnable);// 没有空间则一直阻塞，直到有空间
        } catch (InterruptedException e) {                       
           return;
        }
    }
}

非链式写法

OnSubscribe<Integer> onSubscribe = new OnSubscribe<>();
Observable<Integer> observable = new Observable<>(onSubscribe);

OperatorObserveOn<Integer> operatorObserveOn = new OperatorObserveOn(handler);
OnSubscribeLift<Integer, String> onSubscribe2 = new OnSubscribeLift<>(onSubscribe, operatorObserveOn);

Func1<Integer, String> func = new Func1<>();
OperatorMap<Integer, String> operatorMap = new OperatorMap<>(func);
OnSubscribeLift<Integer, String> onSubscribe3 = new OnSubscribeLift<>(onSubscribe2, operatorMap);
Observable<String> observable2 = new Observable<>(onSubscribe3);
OnSubscribeOnIO<String> onSubscribe4 = new OnSubscribeOnIO(observable2);
Observable<String> observable3 = new Observable<>(onSubscribe4);
Subscriber<String> subscriber = new Subscriber<>();
observable3.subscribe(subscriber);

1) observable3.subscribe(subscriber) --> 2) onSubscribe4.call(subscriber) 即 OnSubscribeOnIO.call(subscriber) --> 3) 子线程 new Runnable(){} --> observable2.subscribe(subscriber) 4) onSubscribe3.call(subscriber) 即 OnSubscribeLift.call(subscriber) --> 5) Subscriber<Integer> st = operatorMap.call(subscriber) 即 6) Subscriber<Integer> st = new MapSubscriber<Integer, String>(subscriber, func) 7) parent.call(st) 即 onSubscribe2.call(st) 即 OnSubscribeLift.call(st)--> 8) Subscriber<Integer> st2 = operatorObserveOn.call(st) --> 9) parent.call(st2) 即 onSubscribe.call(st2) --> 8) st2.onNext(1) --> // onNext()里面切换到Handler所在线程 9) st.onNext(1) --> 9) String string = func.call(1) 10) subscriber.onNext(string)