引言
在c#中,异步的async和await原理,以及运行机制,可以说是老生常谈,经常在各个群里看到有在讨论这个的,而且网上看到的也只是对异步状态机的一些讲解,甚至很多人说异步状态机的时候,他们说的是在运行时去构建状态机对线程状态进行调度,实际上异步状态机是属于编译期间,通过生成dll,然后我们使用反编译工具查看,是可以看到IL构建了异步状态机,并且在运行时添加了两个特性,其中比较重要的是AsyncStateMachine特性这个特性接受的是一个type类型的参数,即指定用的是哪一个异步状态机。所以在写多线程的时候,前面第一篇主要写线程方面的一些具体的使用,以及实现自定义的一些操作,接下来的这篇可能会注重原理方面的讲解,以及结合一些代码实现自定义状态机。
Part 1
在c#中,有的关键字的使用实际上是由对应的类去进行封装的,那例如Lock关键字,是基于Monitor的Enter和Exit两个方法进行封装的,那对应的async和await关键字也是有对应的类或者结构体或者接口去进行封装的,上篇文章中,我们写了自定义的await,可以看到实际上await关键字的限制就是必须继承ICriticalNotifyCompletion, INotifyCompletion这两个接口,然后必须实现它接口的方法,这里有个缺陷就是,await自定义是必须有实现GetResult的方法的这个方法,但是实现那两个接口是没有这个方法的,所以GetResult方法必须是自己手动去实现,返回值的话可以根据自己的情况去写,可以是泛型T 也可以是void类型,然后需要实现一个拓展方法,拓展方法返回类型是你自定义的await,拓展方法是你需要使用await关键字的具体类型;那对应的async的关键字,也是有一个结构体进行封装的AsyncTaskMethodBuilder这个结构体是一个泛型,也有一个不是泛型的,这个可以对标你的自定义await 如果你的await是有返回值的是泛型的,那这个builder也必须是泛型,对标你的返回值类型。
public CustomAwaiter(Func<int, int, string> obj) { Obj = obj; } private bool bIsFinesh; private Timer Timer { get; set; } public bool IsCompleted { get { return bIsFinesh; } } private SpinLock SpinLock = new SpinLock(); private string Result { get; set; } public Func<int, int, string> Obj { get; } public void OnCompleted(Action continuation) { Timer = new Timer(s => { var action = s as Action; var bIsEnter = false; SpinLock.TryEnter(ref bIsEnter); if (bIsEnter) { Result = Obj.Invoke(5, 10); SpinLock.Exit(false); } Thread.Sleep(5000); action?.Invoke(); bIsFinesh = true; }, continuation, 0, int.MaxValue); } public void UnsafeOnCompleted(Action continuation) { Timer = new Timer(s => { var action = s as Action; var bIsEnter = false; SpinLock.TryEnter(ref bIsEnter); if (bIsEnter) { Result = Obj.Invoke(5, 10); SpinLock.Exit(false); } action?.Invoke(); bIsFinesh = true; }, continuation, 5000, int.MaxValue); } public string GetResult() { return Result; }
public static CustomAwaiter GetAwaiter(this Func<int, int, string> obj) { return new CustomAwaiter(obj); }
Part 2
在第一部分中,我们找到了async 和await对应的结构体以及接口,那我们接下来看看实际上的异步的运行方式,下面这一段代码相信大家看起来很熟悉,感觉似曾相识,实际上异步方法加上async和await关键字的时候生成的IL代码转为c#代码基本上就是这个样子的。
GetResult方法,去调用异步状态机
可以看到我们在这里定义了一个方法GetResult,这里面去执行一个异步状态机,这里可以看看自定义状态机的代码,实现了IAsyncStateMachine这个接口,重写了MoveNext的方法和SetStateMachine的两个方法,这里着重讲解MoveNext方法,在c#异步中,都是使用MoveNext方法来进行调度,通过定义的State来判断执行那一步,结合第一段代码片段,可以看到我们刚开始的时候设置的状态是-1,然后调用了Builder的Start方法,这个方法需要传入一个状态机的参数,所以我们传入我们自定义的状态机,
public static Task<string> GetResult() { CustomAsyncStateMechine customAsyncStateMechine = new CustomAsyncStateMechine(); customAsyncStateMechine.builder = AsyncTaskMethodBuilder<string>.Create(); customAsyncStateMechine.State = -1; customAsyncStateMechine.builder.Start(ref customAsyncStateMechine); return customAsyncStateMechine.builder.Task; }
需要执行的异步方法
public async static Task<string> Tests() { return await Task.Run(() => { return "hELLO"; }); }
自定义异步状态机
public class CustomAsyncStateMechine : IAsyncStateMachine { public AsyncTaskMethodBuilder<string> builder; public TaskAwaiter<string> awaiter; public int State; public void MoveNext() { TaskAwaiter<string> taskAwaiter=default; int num = State; CustomAsyncStateMechine state; string Result = string.Empty; switch (num) { case -1: taskAwaiter = Program.Tests().GetAwaiter(); if (!taskAwaiter.IsCompleted) { num = State = 0; awaiter = taskAwaiter; state = this; builder.AwaitUnsafeOnCompleted(ref taskAwaiter, ref state); return; } break; case 0: taskAwaiter = awaiter; awaiter = default(TaskAwaiter<string>); num = State = -1; break; } Result = taskAwaiter.GetResult(); builder.SetResult(Result); } public void SetStateMachine(IAsyncStateMachine stateMachine) { } }
Start方法
可以在下面的代码段看到Start的方法代码,在我们调用了这个方法之后会构建一个用于切换线程上下文的对象,然后调用线程上下文的方法去进行一些操作,这里看一下,这个方法调用了状态机的MoveNext方法,这是第一次执行MoveNext的方法,可以看到我们第一次执行MoveNext方法的时候我们去获取了一下Tests的GetAwaiter,获取的时候实际上这个Tests方法已经执行了,然后我们去判断是否完成,如果没有完成,我们需要去进行下一步操作,在全局变量定义一个Awaiter,需要将Tests的Awaiter保存起来,然后切换State的状态推进到下一步,然后我们调用了Builder的AwaitUnsafeOnCompleted这个方法,
public void Start(ref TStateMachine stateMachine) where TStateMachine : IAsyncStateMachine { // See comment on AsyncMethodBuilderCore.Start // AsyncMethodBuilderCore.Start(ref stateMachine); if (stateMachine == null) throw new ArgumentNullException("stateMachine"); Contract.EndContractBlock(); // Run the MoveNext method within a copy-on-write ExecutionContext scope. // This allows us to undo any ExecutionContext changes made in MoveNext, // so that they won't "leak" out of the first await. ExecutionContextSwitcher ecs = default(ExecutionContextSwitcher); RuntimeHelpers.PrepareConstrainedRegions(); try { ExecutionContext.EstablishCopyOnWriteScope(ref ecs); stateMachine.MoveNext(); } finally { ecs.Undo(); } }
AwaitUnsafeOnCompleted方法
可以看到这个方法内部有调用了一个GetCompletionAction方法
public void AwaitUnsafeOnCompleted( ref TAwaiter awaiter, ref TStateMachine stateMachine) where TAwaiter : ICriticalNotifyCompletion where TStateMachine : IAsyncStateMachine { try { AsyncMethodBuilderCore.MoveNextRunner runnerToInitialize = null; var continuation = m_coreState.GetCompletionAction(AsyncCausalityTracer.LoggingOn ? this.Task : null, ref runnerToInitialize); Contract.Assert(continuation != null, "GetCompletionAction should always return a valid action."); // If this is our first await, such that we've not yet boxed the state machine, do so now. if (m_coreState.m_stateMachine == null) { // Force the Task to be initialized prior to the first suspending await so // that the original stack-based builder has a reference to the right Task. var builtTask = this.Task; // Box the state machine, then tell the boxed instance to call back into its own builder, // so we can cache the boxed reference. Contract.Assert(!Object.ReferenceEquals((object)stateMachine, (object)stateMachine), "Expected an unboxed state machine reference"); m_coreState.PostBoxInitialization(stateMachine, runnerToInitialize, builtTask); } awaiter.UnsafeOnCompleted(continuation); } catch (Exception e) { AsyncMethodBuilderCore.ThrowAsync(e, targetContext: null); } }
GetCompletionAction方法
这里我们着重看一下runner.run方法可以看到Run方法里面不管是怎么去进行操作,最后都是要去执行MoveNext方法,接下来看一下上面的AwaitUnsafeOnCompleted方法,还记得上一篇文章中,我卖了一个关子,询问大家OnCompleted和UnsafeOnCompleted方法里面的Action是哪一个方法,现在已经很明了了,这个Action执行的是状态机的MoveNext方法,它是在Task完成之后,去执行OnCompleted和UnSafeOnCompleted方法的,这里为了方便大家理解,需要结合上一篇文章中自定义任务调度TaskScheduler去给大家演示,最好是希望阅读文章后去下载最新的代码进行调试就会很明白,
internal Action GetCompletionAction(Task taskForTracing, ref MoveNextRunner runnerToInitialize) { Contract.Assert(m_defaultContextAction == null || m_stateMachine != null, "Expected non-null m_stateMachine on non-null m_defaultContextAction"); // Alert a listening debugger that we can't make forward progress unless it slips threads. // If we don't do this, and a method that uses "await foo;" is invoked through funceval, // we could end up hooking up a callback to push forward the async method's state machine, // the debugger would then abort the funceval after it takes too long, and then continuing // execution could result in another callback being hooked up. At that point we have // multiple callbacks registered to push the state machine, which could result in bad behavior. Debugger.NotifyOfCrossThreadDependency(); // The builder needs to flow ExecutionContext, so capture it. var capturedContext = ExecutionContext.FastCapture(); // ok to use FastCapture as we haven't made any permission demands/asserts // If the ExecutionContext is the default context, try to use a cached delegate, creating one if necessary. Action action; MoveNextRunner runner; if (capturedContext != null && capturedContext.IsPreAllocatedDefault) { // Get the cached delegate, and if it's non-null, return it. action = m_defaultContextAction; if (action != null) { Contract.Assert(m_stateMachine != null, "If the delegate was set, the state machine should have been as well."); return action; } // There wasn't a cached delegate, so create one and cache it. // The delegate won't be usable until we set the MoveNextRunner's target state machine. runner = new MoveNextRunner(capturedContext, m_stateMachine); action = new Action(runner.Run); if (taskForTracing != null) { m_defaultContextAction = action = OutputAsyncCausalityEvents(taskForTracing, action); } else { m_defaultContextAction = action; } } // Otherwise, create an Action that flows this context. The context may be null. // The delegate won't be usable until we set the MoveNextRunner's target state machine. else { runner = new MoveNextRunner(capturedContext, m_stateMachine); action = new Action(runner.Run); if (taskForTracing != null) { action = OutputAsyncCausalityEvents(taskForTracing, action); } // NOTE: If capturedContext is null, we could create the Action to point directly // to m_stateMachine.MoveNext. However, that follows a much more expensive // delegate creation path. } if (m_stateMachine == null) runnerToInitialize = runner; return action; }
internal void Run() { Contract.Assert(m_stateMachine != null, "The state machine must have been set before calling Run."); if (m_context != null) { try { // Get the callback, lazily initializing it as necessary ContextCallback callback = s_invokeMoveNext; if (callback == null) { s_invokeMoveNext = callback = InvokeMoveNext; } // Use the context and callback to invoke m_stateMachine.MoveNext. ExecutionContext.Run(m_context, callback, m_stateMachine, preserveSyncCtx: true); } finally { m_context.Dispose(); } } else { m_stateMachine.MoveNext(); } } ///Cached delegate used with ExecutionContext.Run. [SecurityCritical] private static ContextCallback s_invokeMoveNext; // lazily-initialized due to SecurityCritical attribution /// Invokes the MoveNext method on the supplied IAsyncStateMachine. /// The IAsyncStateMachine machine instance. [SecurityCritical] // necessary for ContextCallback in CoreCLR private static void InvokeMoveNext(object stateMachine) { ((IAsyncStateMachine)stateMachine).MoveNext(); }
CustomScheduler 和CustomAwaiter 以及自定义状态机的结合使用,
foreach (var item in Enumerable.Range(0, 1)) { await Task.Run(async () => { var i = item; var ts = new Func<int, int, string>((s, b) => { return Guid.NewGuid().ToString(); }); //var t= await ts; var tash = new TaskCustomScheduler(); var factory = new TaskFactory(tash); await factory.StartNew(async () => { var state = new CustomAsyncStateMechines(); state.State = -1; state.awaiter = ts.GetAwaiter(); state.builder = AsyncTaskMethodBuilder<string>.Create(); state.builder.Start(ref state); var result = await state.builder.Task; Console.WriteLine(result); }); }); }
在上一篇文章中,我们讲解了自定义调度的几个比较重要的方法,我们在使用factory去进行指定了调度器之后,调用了StartNew方法,去执行一段代码,这里的是,实际上在StartNew执行之前,他会先咋自定义任务调度里面添加Task,他会走到QueueTask将Task添加到自己定义的任务池里面去,然后再去RunWork,去通过ThreadPool去执行Task,TryExecuteTask是抽象类提供且内部实现的一个方法,是去执行Task,然后Task执行结束后,我们把它从任务调度池里面移除,那Task结束之后,就会走到自定义Awaiter里面UnSafeOnCompleted方法里面去,然后在这里面再去写执行完成的回调,将状态机向前推进,然后在Movenext方法里面,我们在去获取awaiter的结果,这里就是刚开始所说的就是自定义Awaiter需要自己写的GetResult方法,然后获取到结果之后,我们需要将结果赋值到Task中,就需要调用builder的SetResult方法,实际上对于Task的异常处理也是有SetException方法去进行设置异常的,就需要在MoveNext方法中添加Try Catch 然后捕获之后去迪奥用SetException方法设置异常,这就是async和await异步执行的相关过程,对于内部更深层次的,我目前也是一知半解,但是大体意思都是知道。
public class TaskCustomScheduler : TaskScheduler { private SpinLock SpinLock = new SpinLock(); public TaskCustomScheduler() { } private ConcurrentQueueTasks = new ConcurrentQueue (); protected override IEnumerable GetScheduledTasks() { return Tasks.ToList(); } protected override void QueueTask(Task task) { Tasks.Enqueue(task); RunWork(); } protected override bool TryDequeue(Task task) { return Tasks.TryDequeue(out task); } protected override bool TryExecuteTaskInline(Task task, bool taskWasPreviouslyQueued) { return TryExecuteTask(task); } private void RunWork() { ThreadPool.UnsafeQueueUserWorkItem(_ => { try { foreach (var item in Tasks) { var task = item; var isEnter = false; SpinLock.TryEnter(ref isEnter); TryExecuteTask(task); if (isEnter) { Tasks.TryDequeue(out task); SpinLock.Exit(false); } } } finally { } }, null); } }
Part 2
c#中,实际上所有的Task都是基于ThreadPoolScheduler去进行运行的,这个类开发者是没有办法去new的,但是在TaskScheduler中有一个属性Default实际上它返回的就是这个类,然后Task的时候都是运行在这个类上面,由这个类去进行调度,至于有的人说异步多线程,有的时候异步是多线程有的时候不是多线程,在这里,可以肯定的是async和await的异步是多线程的,但是对于一些类提供的Begin开头的异步,这种的 ,我的观点是,不是多线程的,如果我说的不对的话,希望各位大佬能够进行指正,代码的话,我会放在Gitee里面去,家里的网络上不去Github。抱歉,
总结
多线程方面的文章就讲解到这里,后续可能会出一些,winform方面自绘或者Net Core自定义配置结合Options进行的自定义,敬请各位大佬进行关注,如果对文章或者代码有不懂的地方,可以看自己所在的群里有没有叫四川观察的,那基本上就是我了,或者加QQ群6406277,找我也可以,在这里,谢谢大家的支持,以后会多发表开发方面的知识,大家一起学习,一起进步。