RxJava 复刻简版之四，线程切换

补充知识点：Rx 即 Reactive Extensions，因此 RxJava 是为 java 扩展的响应式编程库，使数据可以异步地运用

1、线程类型

1. Schedulers.io：常用，读写文件/数据库/网络请求
2. Schedulers.newThread：耗时操作
3. Schedulers.computation：CUP密集计算，图片压缩/数据格式解析
4. Schedulers.trampoline：当前线程
5. AndroidSchedulers.mainThread：主线程，更新UI

2、例子

        Single.just(777)
                .subscribeOn(Schedulers.io())
                .observeOn(AndroidSchedulers.mainThread())
                .subscribe(getSingleObserver(tag));

3、流程

1、just(777) ，声明数据处理器

class SingleJust<T>(private val value: T) : Single<T>() {
    override fun subscribeActual(observer: SingleObserver<in T>) {
        observer.onSubscribe(Disposable.disposed())
        observer.onSuccess(value)
    }
}

2、subscribeOn、observeOn，声明处理的观察者

class SingleSubscribeOn<T>(
    private val source: SingleSource<out T>,
    private val scheduler: Scheduler
) : Single<T>() {
    override fun subscribeActual(observer: SingleObserver<in T>) {
        val parent = SubscribeOnObserver(observer, source)
        observer.onSubscribe(parent)
        val f = scheduler.scheduleDirect(parent)
        parent.task.replace(f)
    }

    internal class SubscribeOnObserver<T>(
        private val downstream: SingleObserver<in T>,
        private val source: SingleSource<out T>
    ) :
        AtomicReference<Disposable>(), SingleObserver<T>, Disposable,
        Runnable {
        val task = SequentialDisposable()

        override fun onSubscribe(d: Disposable) {
            DisposableHelper.setOnce(this, d)
        }

        override fun onSuccess(t: T) {
            downstream.onSuccess(t)
        }

        override fun onError(e: Throwable) {
            downstream.onError(e)
        }

        override fun dispose() {
            DisposableHelper.dispose(this)
            task.dispose()
        }

        override fun isDisposed() = DisposableHelper.isDisposed(get())

        override fun run() {
            source.subscribe(this)
        }
    }
}

class SingleObserveOn<T>(
    private val source: SingleSource<out T>,
    private val scheduler: Scheduler
) : Single<T>() {
    override fun subscribeActual(observer: SingleObserver<in T>) {
        source.subscribe(ObserveOnSingleObserver(observer, scheduler))
    }

    internal class ObserveOnSingleObserver<T>(
        private val downstream: SingleObserver<in T>,
        private val scheduler: Scheduler
    ) :
        AtomicReference<Disposable>(), SingleObserver<T>, Disposable,
        Runnable {
        var value: T? = null
        var error: Throwable? = null

        override fun onSubscribe(d: Disposable) {
            if (DisposableHelper.setOnce(this, d)) {
                downstream.onSubscribe(this)
            }
        }

        override fun onSuccess(t: T) {
            this.value = t
            val d = scheduler.scheduleDirect(this)
            DisposableHelper.replace(this, d)
        }

        override fun onError(e: Throwable) {
            error = e
            val d = scheduler.scheduleDirect(this)
            DisposableHelper.replace(this, d)
        }

        override fun run() {
            val ex = error
            if (ex != null) {
                downstream.onError(ex)
            } else {
                downstream.onSuccess(value!!)
            }
        }

        override fun dispose() {
            DisposableHelper.dispose(this)
        }

        override fun isDisposed() = DisposableHelper.isDisposed(get())
    }
}

4、分析

1、执行subscribe，老规矩，通通回调到 subscribeActual 的具体实现
未完待续