异步编程的几种方式，你知道几种？

{"type":"doc","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"近期尝试在搬砖专用语言 Java 上实现异步，起因和过程就不再详述了，总而言之，心中一万头草泥马奔过。但这个过程也没有白白浪费，趁机回顾了一下各种异步编程的实现。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"这篇文章会涉及到回调、Promise、反应式、async/await、用户态线程等异步编程的实现方案。如果你熟悉它们中的一两种，那应该也能很快理解其他几个。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":1},"content":[{"type":"text","text":"为什么需要异步？","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"操作系统可以看作是个虚拟机（VM），进程生活在操作系统创造的虚拟世界里。进程不用知道到底有多少 core 多少内存，只要进程不要索取的太过分，操作系统就假装有无限多的资源可用。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"基于这个思想，线程（Thread）的个数并不受硬件限制：你的程序可以只有一个线程、也可以有成百上千个。操作系统会默默做好调度，让诸多线程共享有限的 CPU 时间片。这个调度的过程对线程是完全透明 的。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"那么，操作系统是怎样做到在线程无感知的情况下调度呢？答案是上下文切换（Context Switch） ，简单来说，操作系统利用软中断机制，把程序从任意位置打断，然后保存当前所有寄存器——包括最重要的指令寄存器 PC 和栈顶指针 SP，还有一些线程控制信息（TCB），整个过程会产生数个微秒的 overhead。","attrs":{}}]},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/1f/1f0cb4d2b782c9ff8eae6309feb45f2f.webp","alt":"图片","title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"然而作为一位合格的程序员，你一定也听说过，线程是昂贵的：","attrs":{}}]},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"线程的上下文切换有不少的代价，占用宝贵的 CPU 时间；","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"每个线程都会占用一些（至少 1 页）内存。","attrs":{}}]}]}],"attrs":{}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"这两个原因驱使我们尽可能避免创建太多的线程 ，而异步编程的目的就是消除 IO wait 阻塞——绝大多数时候，这是我们创建一堆线程、甚至引入线程池的罪魁祸首。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":1},"content":[{"type":"text","text":"Continuation","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"回调函数知道的人很多，但了解 Continuation 的人不多。Continuation 有时被晦涩地翻译成“计算续体”，咱们还是直接用单词好了。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"把一个计算过程在中间打断，剩下的部分用一个对象表示，这就是 Continuation 。操作系统暂停一个线程时保存的那些现场数据，也可以看作一个 Continuation。有了它，我们就能在这个点接着刚刚的断点继续执行。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"打断一个计算过程听起来很厉害吧！实际上它每时每刻都在发生——假设函数 ","attrs":{}},{"type":"codeinline","content":[{"type":"text","text":"f()","attrs":{}}],"attrs":{}},{"type":"text","text":" 中间调用了 ","attrs":{}},{"type":"codeinline","content":[{"type":"text","text":"g()","attrs":{}}],"attrs":{}},{"type":"text","text":"，那 ","attrs":{}},{"type":"codeinline","content":[{"type":"text","text":"g()","attrs":{}}],"attrs":{}},{"type":"text","text":" 运行完成时，要返回到 ","attrs":{}},{"type":"codeinline","content":[{"type":"text","text":"f()","attrs":{}}],"attrs":{}},{"type":"text","text":" 刚刚调用 ","attrs":{}},{"type":"codeinline","content":[{"type":"text","text":"g()","attrs":{}}],"attrs":{}},{"type":"text","text":" 的地方接着执行。这个过程再自然不过了，以至于所有编程语言（汇编除外）都把它掩藏起来，让你在编程中感觉不到调用栈的存在。","attrs":{}}]},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/03/03e900c4fcaa1dbc8038780bf4c1208b.webp","alt":"图片","title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"操作系统用昂贵的软中断机制实现了栈的保存和恢复。那有没有别的方式实现 Continuation 呢？最朴素的想法就是，把所有用得到的信息包成一个函数对象，在调用 ","attrs":{}},{"type":"codeinline","content":[{"type":"text","text":"g()","attrs":{}}],"attrs":{}},{"type":"text","text":" 的时候一起传进去，并约定：一旦 ","attrs":{}},{"type":"codeinline","content":[{"type":"text","text":"g()","attrs":{}}],"attrs":{}},{"type":"text","text":" 完成，就拿着结果去调用这个 Continuation。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"这种编程模式被称为 Continuation-passing style（CPS）：","attrs":{}}]},{"type":"numberedlist","attrs":{"start":null,"normalizeStart":1},"content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":1,"align":null,"origin":null},"content":[{"type":"text","text":"把调用者 ","attrs":{}},{"type":"codeinline","content":[{"type":"text","text":"f()","attrs":{}}],"attrs":{}},{"type":"text","text":" 还未执行的部分包成一个函数对象 ","attrs":{}},{"type":"codeinline","content":[{"type":"text","text":"cont","attrs":{}}],"attrs":{}},{"type":"text","text":"，一同传给被调用者 ","attrs":{}},{"type":"codeinline","content":[{"type":"text","text":"g()","attrs":{}}],"attrs":{}},{"type":"text","text":"；","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":2,"align":null,"origin":null},"content":[{"type":"text","text":"正常运行 ","attrs":{}},{"type":"codeinline","content":[{"type":"text","text":"g()","attrs":{}}],"attrs":{}},{"type":"text","text":" 函数体；","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":3,"align":null,"origin":null},"content":[{"type":"codeinline","content":[{"type":"text","text":"g()","attrs":{}}],"attrs":{}},{"type":"text","text":" 完成后，连同它的结果一起回调 ","attrs":{}},{"type":"codeinline","content":[{"type":"text","text":"cont","attrs":{}}],"attrs":{}},{"type":"text","text":"，从而继续执行 ","attrs":{}},{"type":"codeinline","content":[{"type":"text","text":"f()","attrs":{}}],"attrs":{}},{"type":"text","text":" 里剩余的代码。","attrs":{}}]}]}]},{"type":"paragraph","attrs":{"indent":0,"number":4,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"再拿 Wikipedia 上的定义巩固一下：","attrs":{}}]},{"type":"blockquote","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"A function written in continuation-passing style takes an extra argument: an explicit \"continuation\", i.e. a function of one argument. When the CPS function has computed its result value, it \"returns\" it by calling the continuation function with this value as the argument.","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"CPS 风格的函数带一个额外的参数：一个显式的 Continuation，具体来说就是个仅有一个参数的函数。当 CPS 函数计算完返回值时，它“返回”的方式就是拿着返回值调用那个 Continuation。","attrs":{}}]}],"attrs":{}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"你应该已经发现了，这也就是回调函数，我只是换了个名字而已。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":1},"content":[{"type":"text","text":"异步的朴素实现：Callback","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"光有回调函数其实并没有卵用。对于纯粹的计算工作，Call Stack 就很好，为何要费时费力用回调来做 Continuation 呢？你说的对，但仅限于没有 IO 的情况。我们知道 IO 通常要比 CPU 慢上好几个数量级，在 BIO 中，线程发起 IO 之后只能暂停，然后等待 IO 完成再由操作系统唤醒。","attrs":{}}]},{"type":"codeblock","attrs":{"lang":null},"content":[{"type":"text","text":"var input = recv_from_socket()  // Block at syscall recv()\n        var result = calculator.calculate(input)\n        send_to_socket(result) // Block at syscall send()\n","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"而异步 IO 中，进程发起 IO 操作时也会一并输入回调（也就是 Continuation），这大大解放了生产力——现场无需等待，可以立即返回去做其他事情。一旦 IO 成功后，AIO 的 Event Loop 会调用刚刚设置的回调函数，把剩下的工作完成。这种模式有时也被称为 Fire and Forget。","attrs":{}}]},{"type":"codeblock","attrs":{"lang":null},"content":[{"type":"text","text":"recv_from_socket((input) -> {\n        var result = calculator.calculate(input)\n        send_to_socket(result) // ignore result\n        })\n","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"就这么简单，通过我们自己实现的 Continuation，线程不再受 IO 阻塞，可以自由自在地跑满 CPU。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":1},"content":[{"type":"text","text":"一颗语法糖：Promise","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"回调函数哪里都好，就是不大好用，以及太丑了。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"第一个问题是可读性大大下降，由于我们绕开操作系统自制 Continuation，所有函数调用都要传入一个 lambda 表达式，你的代码看起来就像要起飞一样，缩进止不住地往右挪（the \"Callback Hell\"）。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"第二个问题是各种细节处理起来很麻烦，比如，考虑下异常处理，看来传一个 Continuation 还不够，最好再传个异常处理的 callback。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Promise 是对异步调用结果的一个封装 ，在 Java 中它叫作 CompletableFuture (JDK8) 或者 ListenableFuture (Guava)。Promise 有两层含义：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"第一层含义是：我现在还不是真正的结果，但是承诺以后会拿到这个结果 。这很容易理解，异步的任务迟早会完成，调用者如果比较蠢萌，他也可以用 ","attrs":{}},{"type":"codeinline","content":[{"type":"text","text":"Promise.get()","attrs":{}}],"attrs":{}},{"type":"text","text":" 强行要拿到结果，顺便阻塞了当前线程，异步变成了同步。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"第二层含义是：如果你（调用者）有什么吩咐，就告诉我好了 。这就有趣了，换句话说，回调函数不再是传给 ","attrs":{}},{"type":"codeinline","content":[{"type":"text","text":"g()","attrs":{}}],"attrs":{}},{"type":"text","text":"，而是 ","attrs":{}},{"type":"codeinline","content":[{"type":"text","text":"g()","attrs":{}}],"attrs":{}},{"type":"text","text":" 返回的 Promise，比如之前那段代码，我们用 Promise 来书写，看起来顺眼了不少。","attrs":{}}]},{"type":"codeblock","attrs":{"lang":null},"content":[{"type":"text","text":"var promise_input = recv_from_socket()\n        promise_input.then((input) -> {\n        var result = calculator.calculate(input)\n        send_to_socket(result) // ignore result\n        })\n","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Promise 改善了 Callback 的可读性，也让异常处理稍稍优雅了些，但终究是颗语法糖。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":1},"content":[{"type":"text","text":"反应式编程","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"反应式（Reactive）最早源于函数式编程中的一种模式，随着微软发起 ReactiveX 项目并一步步壮大，被移植到各种语言和平台上。Reactive 最初在 GUI 编程中有广泛的应用，由于异步调用的高性能，很快也在服务器后端领域遍地开花。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Reactive 可以看作是对 Promise 的极大增强，相比 Promise，反应式引入了流（Flow）的概念。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"ReactiveX 中的事件流从一个 Observable 对象流出，这个对象可以是一个按钮，也可以是 Restful API，总之，它能被外界触发。与 Promise 不同的是，事件可能被触发多次，所以处理代码也会被多次调用。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"一旦允许调用多次，从数据流动的角度看，事实上模型已经是 Push 而非 Pull 。那么问题来了，如果调用频率非常高，以至于我们处理速度跟不上了怎么办？所以 RX 框架又引入了 Backpressure 机制来进行流控，最简单的流控方式就是：一旦 buffer 满，就丢弃掉之后的事件。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"ReactiveX 框架的另一个优点是内置了很多好用的算子，比如：","attrs":{}},{"type":"codeinline","content":[{"type":"text","text":"merge","attrs":{}}],"attrs":{}},{"type":"text","text":"（Flow 合并），","attrs":{}},{"type":"codeinline","content":[{"type":"text","text":"debounce","attrs":{}}],"attrs":{}},{"type":"text","text":"（开关除颤）等等，方便了业务开发。下面是一个 RxJava 的例子：","attrs":{}}]},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/98/98e761b6892e63aae472b4f331350534.gif","alt":"图片","title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"heading","attrs":{"align":null,"level":1},"content":[{"type":"text","text":"CPS 变换：Coroutine 与 async/await","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"无论是反应式还是 Promise，说到底仍然没有摆脱手工构造 Continuation：开发者要把业务逻辑写成回调函数。对于线性的逻辑基本可以应付自如，但是如果逻辑复杂一点呢？（比如，考虑下包含循环的情况）","attrs":{}}]},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/d1/d16ec549d9ee8e25b6a8b6b76e3c74e1.webp","alt":"图片","title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"有些语言例如 C#，JavaScript 和 Python 提供了 ","attrs":{}},{"type":"codeinline","content":[{"type":"text","text":"async/await","attrs":{}}],"attrs":{}},{"type":"text","text":" 关键字。与 Reactive 一样，这同样出自微软 C# 语言。在这些语言中，你会感到前所未有的爽感：异步编程终于摆脱了回调函数！唯一要做的只是在异步函数调用时加上 ","attrs":{}},{"type":"codeinline","content":[{"type":"text","text":"await","attrs":{}}],"attrs":{}},{"type":"text","text":"，编译器就会自动把它转化为协程（Coroutine），而非昂贵的线程。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"魔法的背后是 CPS 变换，CPS 变换把普通函数转换成一个 CPS 的函数，即 Continuation 也能作为一个调用参数。函数不仅能从头运行，还能根据 Continuation 的指示继续某个点（比如调用 IO 的地方）运行。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"例子可以参见我的下一篇文章。由于代码太长，就不贴在这儿了。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"可以看到，函数已经不再是一个函数了，而是变成一个状态机 。每次 call 它、或者它 call 其他异步函数时，状态机都会做一些计算和状态轮转。说好的 Continuation 在哪呢？就是对象自己（","attrs":{}},{"type":"codeinline","content":[{"type":"text","text":"this","attrs":{}}],"attrs":{}},{"type":"text","text":"）啊。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"CPS 变换实现非常复杂，尤其是考虑到 try-catch 之后。但是没关系，复杂性都在编译器里，用户只要学两个关键词即可。这个特性非常优雅，比 Java 那个废柴的 ","attrs":{}},{"type":"codeinline","content":[{"type":"text","text":"CompletableFuture","attrs":{}}],"attrs":{}},{"type":"text","text":"不知道高到哪去了。（更新：也没有那么废柴啦）","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"blockquote","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"JVM 上也有一个实现：electronicarts/ea-async，原理和 C# 的 async/await 类似，在编译期修改 Bytecode 实现 CPS 变换。","attrs":{}}]}],"attrs":{}},{"type":"heading","attrs":{"align":null,"level":1},"content":[{"type":"text","text":"终极方案：用户态线程","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"有了 ","attrs":{}},{"type":"codeinline","content":[{"type":"text","text":"async/await","attrs":{}}],"attrs":{}},{"type":"text","text":"，代码已经简洁很多了，基本上和同步代码无异。是否有可能让异步代码和同步代码完全一样呢？听起来就像免费午餐，但是的确可以做到！","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"用户态线程的代表是 Golang。JVM 上也有些实现，比如 Quasar，不过因为 JDBC、Spring 这些周边生态（它们占据了大部分 IO 操作）的缺失基本没有什么用。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"用户态线程是把操作系统提供的线程机制完全抛弃 ，换句话说，不去用这个 VM 的虚拟化机制。比如硬件有 8 个核心，那就创建 8 个系统线程，然后把 N 个用户线程调度到这 8 个系统线程上跑。N 个用户线程的调度在用户进程里实现，由于一切都在进程内部，切换代价要远远小于操作系统 Context Switch。","attrs":{}}]},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/9c/9cc5aaa0453cb16757e617d740e51052.webp","alt":"图片","title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"另一方面，所有可能阻塞系统级线程的事情，例如 ","attrs":{}},{"type":"codeinline","content":[{"type":"text","text":"sleep()","attrs":{}}],"attrs":{}},{"type":"text","text":"、","attrs":{}},{"type":"codeinline","content":[{"type":"text","text":"recv()","attrs":{}}],"attrs":{}},{"type":"text","text":" 等，用户态线程一定不能碰，否则它一旦阻塞住也就带着那 8 个系统线程中的一个阻塞了。Go Runtime 接管了所有这样的系统调用，并用一个统一的 Event loop 来轮询和分发。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"另外，由于用户态线程很轻量，我们完全没必要再用线程池，如果需要开线程就直接创建。比如 Java 中的 WebServer 几乎一定有个线程池，而 Go 可以给每个请求开辟一个 goroutine 去处理。并发编程从未如此美好！","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":1},"content":[{"type":"text","text":"总结","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"以上方案中，Promise、Reactive 本质上还是回调函数，只是框架的存在一定程度上降低了开发者的心智负担。而 ","attrs":{}},{"type":"codeinline","content":[{"type":"text","text":"async/await","attrs":{}}],"attrs":{}},{"type":"text","text":" 和用户态线程的解决方案要优雅和彻底的多，前者通过编译期的 CPS 变换帮用户创造出 CPS 式的函数调用；后者则绕开操作系统、重新实现一套线程机制，一切调度工作由 Runtime 接管。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"不知道是不是因为历史包袱太重，Java 语言本身提供的异步编程支持弱得可怜，即便是 CompletableFuture 还是在 Java 8 才引入，其后果就是很多库都没有异步的支持。虽然 Quasar 在没有语言级支持的情况下引入了 CPS 变换，但是由于缺少周边生态的支持，实际很难用在项目中。","attrs":{}}]}]}