Promise是如何实现异步编程的?

Promise标准

不能免俗地贴个Promise标准链接Promises/A+。ES6的Promise有很多方法,包括Promise.all()/Promise.resolve()/Promise.reject()等,但其实这些都是Promises/A+规范之外的,Promises/A+规范只定义了一个Promise.then()方法,这是Promise的核心。

基本结构

new Promise((resolve, reject) => {
  let a = 0;
  if (a > 1) {
    resolve(a);
  } else {
    reject(a);
  }
}).then(res => {
  console.log(res);
}, err => {
  console.log(err);
})

Promise接收一个函数作为参数,我们称之为executor,该函数有两个参数resolve和reject,这两个参数也都是函数,并且,它们定义在Promise内部。

那么我们定义一个class并定义一个_isFunction方法,用来校验构造函数的参数必须是函数。再定义resolve和reject这两个方法。

class MyPromise{
  constructor(executor){
    if(!this._isFunction(executor)){
      throw new Error(`Promise resolver ${executor} is not a function`);
    }
  }
  _isFunction(val){
    return Object.prototype.toString.call(val) === '[object Function]';
  }
  _resolve(){
    
  }
  _reject(){
    
  }
}

Promise状态、resolve、reject

Promise有三种状态,分别是pending(等待中)、fulfilled(成功)、rejected(失败)。状态改变只能从pending => fulfilled,或者pending => rejected。

resolve的作用,就是将Promise的状态从pending改为fulfilled,它接收一个参数作为Promise执行成功的值,这个值会传给then的第一个回调函数。reject的作用是将Promise的状态从pending改为rejected,它也接收一个参数作为Promise执行失败的值,这个值会传给then的第二个回调函数。

那么我们定义好状态_status、_resolve、_reject,再定义两个数组_handleFulfilled、_handleRejected,分别存放then的成功和失败回调集合。当用户调用resolve或reject方法后,开始异步调用_handleFulfilled或_handleRejected数组中的回调。

class MyPromise {
  constructor(executor) {
    if (!this._isFunction(executor)) {
      throw new Error(`${executor} is not a function`);
    }
    this._status = "pending";
    this._value = undefined;
    this._handleFulfilled = [];
    this._handleRejected = [];
    // 很多文章在这里给executor加了try catch,实际上原生Promise的executor中的错误并没有捕获
    executor(this._resolve.bind(this), this._reject.bind(this));
  }
  _isFunction(val) {
    return Object.prototype.toString.call(val) === "[object Function]";
  }
  _resolve(value) {
    if(this._status === 'pending'){
      this._status = "fulfilled";
      this._value = value;
      let cb;
      // 异步按顺序调用并清空回调
      setTimeout(() => {
        while(cb = this._handleFulfilled.shift()){
          cb(value);
        }
      }, 0)
    }
  }
  _reject(value) {
    if(this._status === 'pending'){
      this._status = "rejected";
      this._value = value;
      let cb;
      // 异步按顺序调用并清空回调
      setTimeout(() => {
        while ((cb = this._handleRejected.shift())) {
          cb(value);
        }
      }, 0);
    }
  }
}

Promise.then

Promise.then定义了两个回调onFulfilled和onRejected

promise.then(onFulfilled, onRejected)

它们分别在Promise执行成功/失败时执行,它们都是可选的,Promises/A+规范规定,如果onFulfilled或onRejected不是函数,将被忽略,Promise会继续执行下一个then的回调。比如下面的例子会输出1,.then(2)则被忽略了。

new Promise((resolve, reject) => {
  resolve(1);
})
  .then(2)
  .then((res) => {
    console.log(res);
  });

then可以链式调用,是因为每个then都会返回一个新的Promise。then执行onFulfilled还是onRejected,取决于Promise的状态,如果Promise状态为pending,只会将onFulfilled和onRejected分别push到_handleFulfilled和_handleRejected数组;如果状态为fulfilled,会执行对应的onFulfilled;如果状态是rejected,执行对应的onRejected;

那么then方法的基本结构如下

then(onFulfilled, onRejected) {
    const self = this;
    const { _value, _status } = this;

  // 如果onFulfilled、onRejected不是函数,强制改为函数,并且该函数直接返回接收到的参数,传后面的then的回调函数
  onFulfilled = self._isFunction(onFulfilled) ? onFulfilled : (v) => v;
  onRejected = self._isFunction(onRejected) ? onRejected : (v) => v;

  return new MyPromise((resolve, reject) => {
    switch (_status) {
      case "pending":
        self._handleFulfilled.push(onFulfilled);
        self._handleRejected.push(onRejected);
        break;
      case "fulfilled":
        onFulfilled(_value);
        // todo
        break;
      case "rejected":
        onRejected(_value);
        // todo
        break;
      default:
        throw new Error('Promise resolver Unverified status');
        break;
    }
  });
}

在then链式调用的情况下,如果前一个then返回的是一个新Promise,后一个then的回调必须等这个新Promise的状态改变后才会执行。举例,下面的代码输出1之后,等待3秒才会输出2:

new Promise(resolve => {
  resolve()
}).then(() => {
  return new Promise(resolve => {
    console.log(1);
    setTimeout(() => {
      resolve()
    }, 3000)
  })
}).then(() => {
  console.log(2);
})

因此要对then的回调函数的返回值做个判断,如果返回值不是Promise,利用resolve直接返回这个值;如果返回值是Promise,就要等这个Promise状态变化之后再返回,而Promise状态变化之后一定会调用then的回调函数,利用这个特性,将resolve、reject作为then的回调函数即可。

then(onFulfilled, onRejected) {
    const self = this;
    const { _value, _status } = this;

    // 如果onFulfilled、onRejected不是函数,强制改为函数,并且该函数直接返回接收到的参数,传后面的then的回调函数
    onFulfilled = self._isFunction(onFulfilled) ? onFulfilled : (v) => v;
    onRejected = self._isFunction(onRejected) ? onRejected : (v) => v;

    return new MyPromise((resolve, reject) => {
      const fulfilled = (value) => {
        const res = onFulfilled(value);
        if (res instanceof MyPromise) {
          res.then(resolve, reject);
        } else {
          resolve(res);
        }
      };
      const rejected = (value) => {
        const res = onRejected(value);
        if (res instanceof MyPromise) {
          // 这里是重点
          res.then(resolve, reject);
        } else {
          reject(res);
        }
      };
      switch (_status) {
        case "pending":
          self._handleFulfilled.push(fulfilled);
          self._handleRejected.push(rejected);
          break;
        case "fulfilled":
          fulfilled(_value);
          break;
        case "rejected":
          rejected(_value);
          break;
        default:
          throw new Error('Promise resolver Unverified status');
          break;
      }
    });
  }

完整代码

class MyPromise {
  constructor(executor) {
    if (!this._isFunction(executor)) {
      throw new Error(`${executor} is not a function`);
    }
    this._status = "pending";
    this._value = undefined;
    this._handleFulfilled = [];
    this._handleRejected = [];
    // 很多文章在这里给executor加了try catch,实际上原生Promise的executor中的错误并没有捕获
    executor(this._resolve.bind(this), this._reject.bind(this));
  }
  _isFunction(val) {
    return Object.prototype.toString.call(val) === "[object Function]";
  }
  _resolve(value) {
    if (this._status === "pending") {
      this._status = "fulfilled";
      this._value = value;
      let cb;
      // 异步按顺序调用并清空回调
      setTimeout(() => {
        while ((cb = this._handleFulfilled.shift())) {
          cb(value);
        }
      }, 0);
    }
  }
  _reject(value) {
    if (this._status === "pending") {
      this._status = "rejected";
      this._value = value;
      let cb;
      // 异步按顺序调用并清空回调
      setTimeout(() => {
        while ((cb = this._handleRejected.shift())) {
          cb(value);
        }
      }, 0);
    }
  }
  then(onFulfilled, onRejected) {
    const self = this;
    const { _value, _status } = this;

    // 如果onFulfilled、onRejected不是函数,强制改为函数,并且该函数直接返回接收到的参数,传后面的then的回调函数
    onFulfilled = self._isFunction(onFulfilled) ? onFulfilled : (v) => v;
    onRejected = self._isFunction(onRejected) ? onRejected : (v) => v;

    return new MyPromise((resolve, reject) => {
      const fulfilled = (value) => {
        const res = onFulfilled(value);
        if (res instanceof MyPromise) {
          res.then(resolve, reject);
        } else {
          resolve(res);
        }
      };
      const rejected = (value) => {
        const res = onRejected(value);
        if (res instanceof MyPromise) {
          // 这里是重点
          res.then(resolve, reject);
        } else {
          reject(res);
        }
      };
      switch (_status) {
        case "pending":
          self._handleFulfilled.push(fulfilled);
          self._handleRejected.push(rejected);
          break;
        case "fulfilled":
          fulfilled(_value);
          break;
        case "rejected":
          rejected(_value);
          break;
        default:
          throw new Error('Promise resolver Unverified status');
          break;
      }
    });
  }
}

测试一下,先输出1,3秒后输出2,说明MyPromise的基本功能没问题了。

new MyPromise((resolve) => {
  console.log(1);
  setTimeout(() => {
    resolve(2);
  }, 3000)
}).then(res => {
  console.log(res);
})

最后,总结一下,Promise是如何实现异步编程的?

Promise接收一个函数为参数,传入了两个内部的方法resolve和reject,然后用then注册回调函数,手动调用resolve或reject就可以依次执行then的回调,并且给回调函数传值。如果then返回的也是Promise,同样的,手动调用resolve或reject后,才会继续往下执行。

其实本质上还是回调函数,只不过写法变了。

本文GitHub链接:Promise是如何实现异步编程的?