Generator is a very powerful syntax, but it is not widely used. This article mainly introduces how to use Generator in JavaScript. Friends who need it can refer to
Generator is a very powerful syntax, but its use is not widely used (see the survey on twitter below! ). Why is this so? Compared with async/await, its use is more complex and debugging is not easy (most cases are back to the past). Even though we can obtain a similar experience in a very simple way, people generally prefer async/await .
However, Generator allows us to iterate through our own code via the yield keyword! This is a super powerful syntax and we can actually manipulate the execution! Starting with the less obvious cancellation operations, let's start with synchronization operations.
I created a code repository for the features mentioned in the article - github.com/Bloomca/obs…
Batch (or Schedule)
Executing the Generator function will return a traverser object, which means that we can traverse synchronously through it. Why do we want to do this? The reason may be to implement batch processing. Imagine we need to download 1000 items and display them row by row in a table (don't ask me why, assuming we don't use a framework). While there's nothing wrong with showing them off right away, sometimes that might not be the best solution — maybe your MacBook Pro can handle it with ease, but the average person's computer can't (let alone a phone). So, this means we need to delay execution somehow.
Please note that this example is about performance optimization. There is no need to do this before you encounter this problem - premature optimization is the root of all evil!
// 最初的同步实现版本 function renderItems(items) { for (item of items) { renderItem(item); } } // 函数将由我们的执行器遍历执行 // 实际上,我们可以用相同的同步方式来执行它! function* renderItems(items) { // 我使用 for..of 遍历方法来避免新函数的产生 for (item of items) { yield renderItem(item); } }
There is no difference, right? ? Well, the difference here is that now we can run this function differently without changing the source code. Actually, as I mentioned before, there is no need to wait, we can do it synchronously. So, let’s tweak our code. How about adding a delay of 4 ms (one heartbeat in the JavaScript VM) after each yield? We have 1000 items and rendering will take 4 seconds - not bad, assuming I want to render in 2 seconds, the easy way to think of it is to render 2 at a time. Suddenly the solution using Promises becomes more complicated - we have to pass another parameter: the number of items to render each time. Through our executor, we still need to pass this parameter, but the benefit is that it has absolutely no effect on our renderItems method.
function runWithBatch(chunk, fn, ...args) { const gen = fn(...args); let num = 0; return new Promise((resolve, promiseReject) => { callNextStep(); function callNextStep(res) { let result; try { result = gen.next(res); } catch (e) { return reject(e); } next(result); } function next({ done, value }) { if (done) { return resolve(value); } // every chunk we sleep for a tick if (num++ % chunk === 0) { return sleep(4).then(proceed); } else { return proceed(); } function proceed() { return callNextStep(value); } } }); } // 第一个参数 —— 每批处理多少个项目 const items = [...]; batchRunner(2, function*() { for (item of items) { yield renderItem(item); } });
As you can see, we can easily change the number of items per batch, regardless of the executor, and return to normal synchronous execution - all without affecting our renderItems method.
Cancel
Let’s consider the traditional function - Cancel. I have talked about it in detail in my article Promises cancellation in general (Translation: How to cancel your Promise? ). So I'll use some of this code:
function runWithCancel(fn, ...args) { const gen = fn(...args); let cancelled, cancel; const promise = new Promise((resolve, promiseReject) => { // define cancel function to return it from our fn // 定义 cancel 方法,并返回它 cancel = () => { cancelled = true; reject({ reason: 'cancelled' }); }; onFulfilled(); function onFulfilled(res) { if (!cancelled) { let result; try { result = gen.next(res); } catch (e) { return reject(e); } next(result); return null; } } function onRejected(err) { var result; try { result = gen.throw(err); } catch (e) { return reject(e); } next(result); } function next({ done, value }) { if (done) { return resolve(value); } // 假设我们总是接收 Promise,所以不需要检查类型 return value.then(onFulfilled, onRejected); } }); return { promise, cancel }; }
The best part here is that we can cancel all requests that haven't had a chance to execute yet (we can also pass an object parameter like AbortController to our executor, so it can even Cancel the current request!), and we have not modified a single line of code in our business logic.
Pause/Resume
Another special need may be the pause/resume functionality. Why do you want this feature? Imagine we're rendering 1000 rows of data, and it's very slow, and we want to give the user the ability to pause/resume rendering so they can stop all the background work and read what has been downloaded. let's start!
// 实现渲染的方法还是一样的 function* renderItems() { for (item of items) { yield renderItem(item); } } function runWithPause(genFn, ...args) { let pausePromiseResolve = null; let pausePromise; const gen = genFn(...args); const promise = new Promise((resolve, reject) => { onFulfilledWithPromise(); function onFulfilledWithPromise(res) { if (pausePromise) { pausePromise.then(() => onFulfilled(res)); } else { onFulfilled(res); } } function onFulfilled(res) { let result; try { result = gen.next(res); } catch (e) { return reject(e); } next(result); return null; } function onRejected(err) { var result; try { result = gen.throw(err); } catch (e) { return reject(e); } next(result); } function next({ done, value }) { if (done) { return resolve(value); } // 假设我们总是接收 Promise,所以不需要检查类型 return value.then(onFulfilledWithPromise, onRejected); } }); return { pause: () => { pausePromise = new Promise(resolve => { pausePromiseResolve = resolve; }); }, resume: () => { pausePromiseResolve(); pausePromise = null; }, promise }; }
Calling this executor can return us an object with pause/resume functions, all of which can be easily obtained, or use our previous business code! So, if you have a lot of "heavy" request chains that take a long time, and you want to provide pause/resume functionality to your users, feel free to implement this executor in your code.
Error handling
我们有个神秘的 onRejected 调用,这是我们这部分谈论的主题。如果我们使用正常的 async/await 或 Promise 链式写法,我们将通过 try/catch 语句来进行错误处理,如果不添加大量的逻辑代码就很难进行错误处理。通常情况下,如果我们需要以某种方式处理错误(比如重试),我们只是在 Promise 内部进行处理,这将会回调自己,可能再次回到同样的点。而且,这还不是一个通用的解决方案 —— 可悲的是,在这里甚至 Generator 也不能帮助我们。我们发现了 Generator 的局限 —— 虽然我们可以控制执行流程,但不能移动 Generator 函数的主体;所以我们不能后退一步,重新执行我们的命令。一个可行的解决方案是使用command pattern, 它告诉了我们 yield 结果的数据结构 —— 应该是我们需要执行此命令需要的所有信息,这样我们就可以再次执行它了。所以,我们的方法需要改为:
function* renderItems() { for (item of items) { // 我们需要将所有东西传递出去: // 方法, 内容, 参数 yield [renderItem, null, item]; } }
正如你所看到的,这使得我们不清楚发生了什么 —— 所以,也许最好是写一些 wrapWithRetry 方法,它会检查 catch 代码块中的错误类型并再次尝试。但是我们仍然可以做一些不影响我们功能的事情。例如,我们可以增加一个关于忽略错误的策略 —— 在 async/await 中我们不得不使用 try/catch 包装每个调用,或者添加空的 .catch(() => {}) 部分。有了 Generator,我们可以写一个执行器,忽略所有的错误。
function runWithIgnore(fn, ...args) { const gen = fn(...args); return new Promise((resolve, promiseReject) => { onFulfilled(); function onFulfilled(res) { proceed({ data: res }); } // 这些是 yield 返回的错误 // 我们想忽略它们 // 所以我们像往常一样做,但不去传递出错误 function onRejected(error) { proceed({ error }); } function proceed(data) { let result; try { result = gen.next(data); } catch (e) { // 这些错误是同步错误(比如 TypeError 等) return reject(e); } // 为了区分错误和正常的结果 // 我们用它来执行 next(result); } function next({ done, value }) { if (done) { return resolve(value); } // 假设我们总是接收 Promise,所以不需要检查类型 return value.then(onFulfilled, onRejected); } }); }
关于 async/await
Async/await 是现在的首选语法(甚至 co 也谈到了它 ),这也是未来。但是,Generator 也在 ECMAScript 标准内,这意味着为了使用它们,除了写几个工具函数,你不需要任何东西。我试图向你们展示一些不那么简单的例子,这些实例的价值取决于你的看法。请记住,没有那么多人熟悉 Generator,并且如果在整个代码库中只有一个地方使用它们,那么使用 Promise 可能会更容易一些 —— 但是另一方面,通过 Generator 某些问题可以被优雅和简洁的处理。
上面是我整理给大家的,希望今后会对大家有帮助。
相关文章:
The above is the detailed content of How to use Generator methods in JavaScript. For more information, please follow other related articles on the PHP Chinese website!