RxSwift中就四个内容
可观察序列-Observable
观察者-Observer
调度者-Scheduler
销毁者-Dispose

CurrentThreadScheduler类表示当前线程调度者Scheduler

public class CurrentThreadScheduler : ImmediateSchedulerType {
    typealias ScheduleQueue = RxMutableBox<Queue<ScheduledItemType>>
    /// The singleton instance of the current thread scheduler.
    public static let instance = CurrentThreadScheduler()

    static var queue : ScheduleQueue? {
        get {
            return Thread.getThreadLocalStorageValueForKey(CurrentThreadSchedulerQueueKey.instance)
        }
        set {
            Thread.setThreadLocalStorageValue(newValue, forKey: CurrentThreadSchedulerQueueKey.instance)
        }
    }
    /// Gets a value that indicates whether the caller must call a `schedule` method.
    public static fileprivate(set) var isScheduleRequired: Bool {
        get {
            return pthread_getspecific(CurrentThreadScheduler.isScheduleRequiredKey) == nil
        }
        set(isScheduleRequired) {
            if pthread_setspecific(CurrentThreadScheduler.isScheduleRequiredKey, isScheduleRequired ? nil : scheduleInProgressSentinel) != 0 {
                rxFatalError("pthread_setspecific failed")
            }
        }
    }

    public func schedule<StateType>(_ state: StateType, action: @escaping (StateType) -> Disposable) -> Disposable {
      ...
     }
}

外界获取判断当前队列的是否被关联isScheduleRequired，利用对 queue的set,get方法的观察，绑定我们的当前队列与静态字符串
线程extension

extension Thread {
    static func setThreadLocalStorageValue<T: AnyObject>(_ value: T?, forKey key: NSCopying) {
        let currentThread = Thread.current
        let threadDictionary = currentThread.threadDictionary

        if let newValue = value {
            threadDictionary[key] = newValue
        }
        else {
            threadDictionary[key] = nil
        }
    }

    static func getThreadLocalStorageValueForKey<T>(_ key: NSCopying) -> T? {
        let currentThread = Thread.current
        let threadDictionary = currentThread.threadDictionary
        
        return threadDictionary[key] as? T
    }
}

MainScheduler：表示主线程。如果我们需要执行一些和 UI 相关的任务，就需要切换到该 Scheduler 运行,这里绑定了主队列DispatchQueue.main

public final class MainScheduler : SerialDispatchQueueScheduler {
    private let _mainQueue: DispatchQueue
    let numberEnqueued = AtomicInt(0)

    public init() {
        self._mainQueue = DispatchQueue.main
        super.init(serialQueue: self._mainQueue)
    }
    public static let instance = MainScheduler()
}

同是这里还有继承了SerialDispatchQueueScheduler就是串行调度者。

public class SerialDispatchQueueScheduler : SchedulerType {
    let configuration: DispatchQueueConfiguration
    init(serialQueue: DispatchQueue, leeway:) {
        self.configuration = DispatchQueueConfiguration(queue: leeway:)
    }

public convenience init(internalSerialQueueName: serialQueueConfiguration: leeway: ) {
        let queue = DispatchQueue(label: internalSerialQueueName, attributes: [])
        serialQueueConfiguration?(queue)
        self.init(serialQueue: queue, leeway: leeway)
    }
 }

从这里也可以看出就是接收串行队列，如果没有，自己内部创建一个串行队列

public class ConcurrentDispatchQueueScheduler: SchedulerType {
    public typealias TimeInterval = Foundation.TimeInterval
    public typealias Time = Date
    
    public var now : Date {
        return Date()
    }

    let configuration: DispatchQueueConfiguration
  
    public init(queue: leeway: ) {
        self.configuration = DispatchQueueConfiguration(queue: leeway:)
    }
    
    public convenience init(qos: leeway: ) {
        self.init(queue: DispatchQueue(
            label: "rxswift.queue.\(qos)",
            qos: qos,
            attributes: [DispatchQueue.Attributes.concurrent],
            target: nil),
            leeway: leeway
        )
    }
}

OperationQueueScheduler：封装了 NSOperationQueue, 下面代码非常清晰了，典型的操作队列和操作优先级

public class OperationQueueScheduler: ImmediateSchedulerType {
    public let operationQueue: OperationQueue
    public let queuePriority: Operation.QueuePriority
    public init(operationQueue: queuePriority: ) {
        self.operationQueue = operationQueue
        self.queuePriority = queuePriority
    }
}

调度执行

func schedule<StateType>(_ state: action: ) -> Disposable {
    return self.scheduleInternal(state, action: action)
}

func scheduleInternal<StateType>(_ state:  action: ) -> Disposable {
    return self.configuration.schedule(state, action: action)
}

func scheduleRelative<StateType>(_ state: dueTime: action: ) -> Disposable {
    return self.configuration.scheduleRelative(state, dueTime: action:)
}

func schedulePeriodic<StateType>(state: startAfter:period: action: ) -> Disposable {
    return self.configuration.schedulePeriodic(state, startAfter: period: action:)
}

从上面核心方法：schedule 可以非常轻松看出都是我们的 self.configuration具体施行者

func schedule<StateType>(_ state: StateType, action: ) -> Disposable {
    let cancel = SingleAssignmentDisposable()
    self.queue.async {
        if cancel.isDisposed { return }
        cancel.setDisposable(action(state))
    }
    return cancel
}

调度器（Schedulers）是 RxSwift 实现多线程的核心模块，它主要用于控制任务在哪个线程或队列运行

observeOn&subscribeOn

点击按钮测试

DispatchQueue.global().async {
    self.actionBtn.rx.tap
        .subscribe(onNext: { () in
            print("点击了按钮 --- \(Thread.current)")
        })
        .disposed(by: self.bag)
}

调度主线程判断

public func controlEvent(_ controlEvents: UIControl.Event) -> ControlEvent<()> {
    let source: Observable<Void> = Observable.create { [weak control = self.base] observer in
            MainScheduler.ensureRunningOnMainThread()
        }
    return ControlEvent(events: source)
}

线程切换

public init<Ev: ObservableType>(events: Ev) where Ev.Element == Element {
    self._events = events.subscribeOn(ConcurrentMainScheduler.instance)
}

public func subscribe<Observer: ObserverType>(_ observer: Observer) -> Disposable {
    return self._events.subscribe(observer)
}

OK 很明显我们的 ControlEvent 的序列 subscribe 是调用了一个函数就是：subscribeOn,其中ConcurrentMainScheduler.instance 内部封装了 主队列

public func subscribeOn(_ scheduler: ImmediateSchedulerType)
    -> Observable<Element> {
    return SubscribeOn(source: self, scheduler: scheduler)
}

看到返回值的类型就知道，原来的序列是被subscribeOn进行处理了，封装了中间层：SubscribeOn 的序列

final private class SubscribeOn<Ob: ObservableType>: Producer<Ob.Element> {
    let source: Ob
    let scheduler: ImmediateSchedulerType
    
    init(source: Ob, scheduler: ImmediateSchedulerType) {
        self.source = source
        self.scheduler = scheduler
    }
    
    override func run(_ observer: cancel:) -> (sink:subscription:) {
        let sink = SubscribeOnSink(parent: self, observer: observer, cancel: cancel)
        let subscription = sink.run()
        return (sink: sink, subscription: subscription)
    }
}

看到 SubscribeOn 的继承关系（Producer）
序列订阅的时候，会创建一个observer的观察者
经过Producer 流回SubscribeOn的run
在经过 SubscribeOnSink.run 到观察者的回调（或者内部源序列的订阅，传sink作为观察者回调，后面的流程只是重复走了一次）
由观察者的发送响应，回到 sink 的 on
由 sink的属性观察者（也就是中间封装保存的）响应event事件
最后调用外界的subscribe的闭包

调度源码：

func run() -> Disposable {
    let disposeEverything = SerialDisposable()
    let cancelSchedule = SingleAssignmentDisposable()
    
    disposeEverything.disposable = cancelSchedule
    
    let disposeSchedule = self.parent.scheduler.schedule(()) {

        let subscription = self.parent.source.subscribe(self)
        disposeEverything.disposable = ScheduledDisposable(scheduler: disposable:)
        return Disposables.create()
    }
    cancelSchedule.setDisposable(disposeSchedule)
    return disposeEverything
}

}
这里就有一个非常重要的方法：self.parent.scheduler.schedule()调用self.scheduleInternal(state, action: action)

func schedule<StateType>(_ state: action: ) -> Disposable {
    let cancel = SingleAssignmentDisposable()
    self.queue.async {
        if cancel.isDisposed {
            return
        }
        cancel.setDisposable(action(state))
    }
    return cancel
}

其实这里的action就是一个 schduler 调用时候的闭包，就会执行：let subscription = self.parent.source.subscribe(self), 源序列的subscribe,必然会来到Producer

override func subscribe<Observer: ObserverType>(_ observer: Observer) -> Disposable where Observer.Element == Element {
    if !CurrentThreadScheduler.isScheduleRequired {
        // The returned disposable needs to release all references once it was disposed.
        let disposer = SinkDisposer()
        let sinkAndSubscription = self.run(observer, cancel: disposer)
        disposer.setSinkAndSubscription(sink: sinkAndSubscription.sink, subscription: sinkAndSubscription.subscription)

        return disposer
    }
    else {
        return CurrentThreadScheduler.instance.schedule(()) { _ in
            let disposer = SinkDisposer()
            let sinkAndSubscription = self.run(observer, cancel: disposer)
            disposer.setSinkAndSubscription(sink: sinkAndSubscription.sink, subscription: sinkAndSubscription.subscription)

            return disposer
        }
    }
}

这里会根据当前的调度环境来判断

public func schedule<StateType>(_ state: action: ) -> Disposable {

    if CurrentThreadScheduler.isScheduleRequired {
      // 已经标记，就置false
        CurrentThreadScheduler.isScheduleRequired = false
     // 外界闭包调用执行
        let disposable = action(state)
      // 延迟销毁 
        defer {
            CurrentThreadScheduler.isScheduleRequired = true
            CurrentThreadScheduler.queue = nil
        }
      ...
        return disposable
    }
     ...
    return scheduledItem
}

如果你当前调度环境不变，那就没有问题,如果我这里调度的是子线程，那么就完全不一样，针对当前队列，还有线程安全都是需要处理的

public func scheduleRecursive<State>(_ state: action: ) -> Disposable {
    // 递归调度者
    let recursiveScheduler = RecursiveImmediateScheduler(action: scheduler:)
    // 调度状态执行
    recursiveScheduler.schedule(state)
    return Disposables.create(with: recursiveScheduler.dispose)
}

递归调用

func schedule(_ state: State) {
    var scheduleState: ScheduleState = .initial
    let d = self._scheduler.schedule(state) { state -> Disposable in     
        // 这里因为在递归环境，加了一把锁递归锁，保障安全   
        let action = self._lock.calculateLocked { () -> Action? in
                 return self._action
        }
        
        if let action = action {
            action(state, self.schedule)
        }
        
        return Disposables.create()
    }
...
}

这里因为在递归环境，加了一把锁递归锁，保障安全,通过保护，获取action执行,也就是外界传给递归调度者的闭包任务,RxSwift 的数组调度出来是有顺序的，因为在递归调度，已经加锁了，保障线程资源安全
执行完源序列的响应，会把任务保存进队列

public func schedule<StateType>(_ state: StateType, action: ) -> Disposable {
    // 上面的流程就省略了
    let existingQueue = CurrentThreadScheduler.queue

    let queue: RxMutableBox<Queue<ScheduledItemType>>
    if let existingQueue = existingQueue {
        queue = existingQueue
    }
    else {
        queue = RxMutableBox(Queue<ScheduledItemType>(capacity: 1))
        CurrentThreadScheduler.queue = queue
    }

    let scheduledItem = ScheduledItem(action: action, state: state)
    queue.value.enqueue(scheduledItem)

    return scheduledItem
}

把任务和状态封装成了ScheduledItem,面向对象，更容易传输&执行,把这个事务queue.value.enqueue(scheduledItem),排队进队列

public func schedule<StateType>(_ state: StateType, action: @escaping (StateType) -> Disposable) -> Disposable {
    if CurrentThreadScheduler.isScheduleRequired {

        CurrentThreadScheduler.isScheduleRequired = false
        let disposable = action(state)  

         // 判断当前队列情况，是否存在
        guard let queue = CurrentThreadScheduler.queue else {
            return disposable
        }
        // 从队列去除任务
        while let latest = queue.value.dequeue() {
            if latest.isDisposed {
                continue
            }
            latest.invoke()
        }

        return disposable
    }
   ...
}

流程任务执行action(state) 完毕之后，又会执行下面的流程
判断当前队列情况，是否存在,从队列去除任务 ： queue.value.dequeue()
latest.invoke()

func invoke() {
     self._disposable.setDisposable(self._action(self._state))
}

就是原来响应回来时候保存的 action执行,只不过加了销毁的机制,这个时候我们的流程就会由原来的 源序列 流进 ObserveOnSink,保障了在 ObserveOnSink 的调度环境是有序的进队的：self._queue.enqueue(event)
执行self._scheduler.scheduleRecursive((), action: self.run)

override func onCore(_ event: Event<Element>) {
    let shouldStart = self._lock.calculateLocked { () -> Bool in
        self._queue.enqueue(event)
    }
    if shouldStart {
        self._scheduleDisposable.disposable = 
        self._scheduler.scheduleRecursive((), action: self.run)
    }
}

这里的手法是非常重要的：毕竟并发队列很可能存在

func run(_ state: (), _ recurse: (()) -> Void) {
    // 加锁获取观察者，很队列任务
    let (nextEvent, observer) = self._lock.calculateLocked { 
        if !self._queue.isEmpty {
            return (self._queue.dequeue(), self._observer)
        }
    }
   
    // 观察者发送响应
    if let nextEvent = nextEvent, !self._cancel.isDisposed {
        observer.on(nextEvent)
        if nextEvent.isStopEvent {
            self.dispose()
        }
    }
}

加锁获取观察者，很队列任务 : (self._queue.dequeue(), self._observer)
观察者发送响应: observer.on(nextEvent)

总结：

整个流程是比较复杂
源序列包装
内部序列创建
调度环境&观察者传递准备
源序列订阅 - 根据调度环境调度 - 流程流到观察者就是我们中间内部序列的Sink
Sink 调度执行 响应发给观察者
由观察者响应 订阅事件event

就是两层序列订阅响应，我的第二层的 sink 就是源序列的观察者

--有人觉得简单到令人发指，却令有些人发脱，献给一个姓Co的人。