A brief analysis of the Change Detection mechanism in Angular

What is Change Detection?

In the application development process, state represents the data that needs to be displayed on the application. When state changes, a mechanism is often needed to detect the changed state and update the corresponding interface accordingly. This mechanism is called the Change Detection mechanism. [Related tutorial recommendations: "angular tutorial"]

In WEB development, updating the application interface is actually modifying the DOM tree. Since DOM operations are expensive, an inefficient Change Detection will cause poor application performance. Therefore, the efficiency of the framework in implementing the Change Detection mechanism largely determines its performance.

How Change Detection is implemented

Angular can detect when component data changes and then automatically re-render the view to reflect that change. But how can it do this after a low-level event like a button click?

Through Zone, Angular can automatically trigger the Change Detection mechanism.

What is Zone? In short, Zone is an execution context (execution context), which can be understood as an execution environment. Different from common browser execution environments, all asynchronous tasks executed in this link are called Tasks. Zone provides a bunch of hooks for these Tasks, allowing developers to easily "monitor" the environment. All asynchronous tasks.

Digression: Since Angular strongly advocates the use of observable objects (Observable), if you develop an application completely based on Observable, you can replace Zone to implement the function of tracking the call stack, and improve the performance. It's slightly better than using Zone.

  // Angular 在 v5.0.0-beta.8 起可以通过配置不使用 Zone 
  import { platformBrowser } from '@angular/platform-browser';
  platformBrowser().bootstrapModuleFactory(AppModuleNgFactory, { ngZone: 'noop' });

Override the browser default mechanism

Angular will rewrite the browser low-level API at startup, such as addEventListener, It is a browser function used to register all browser events, including click handling. Angular will replace addEventListener with the new equivalent of this:

// this is the new version of addEventListener                                    
function addEventListener(eventName, callback) { 
    // call the real addEventListener                
    callRealAddEventListener(eventName, function() { 
        //first call the original callback              
        callback(...);
        // and then run Angular-specific functionality
        var changed = angular.runChangeDetection();
        if (changed) {
            angular.reRenderUIPart();
        }
    });
}

The new addEventListener adds more functionality to any event handler: not only registration is called callback, and Angular has a chance to run change detection and update the UI.

Support browser asynchronous API

The following common browser mechanisms have been patched to support change detection:

• All browser events (single click, mouseover, key press, etc.)
• setTimeout()andsetInterval()
• Ajax HTTP request

In fact, Zone.js patches many other browser APIs to transparently trigger Angular change detection, such as Websockets.

One limitation of this mechanism is that if for some reason Zone.js does not support the asynchronous browser API, change detection will not be triggered. This is the case for IndexedDB callbacks, for example.

How does the default change detection mechanism work?

Every Angular component has an associated change detector, which is created when the application starts. For example:

@Component({
    selector: 'todo-item',
    template: `<span class="todo noselect" 
       (click)="onToggle()">{{todo.owner.firstname}} - {{todo.description}}
       - completed: {{todo.completed}}</span>`
})
export class TodoItem {
    @Input()
    todo:Todo;

    @Output()
    toggle = new EventEmitter<Object>();

    onToggle() {
        this.toggle.emit(this.todo);
    }
}

This component will receive a Todo object as input and emit an event when the todo state is toggled.

export class Todo {
    constructor(public id: number, 
        public description: string, 
        public completed: boolean, 
        public owner: Owner) {
    }
}

We can see that Todo has an attribute owner, which itself is an object with two properties: firstname and lastname.

What does a change detector look like?

We can actually see what the change detector looks like at runtime! To see it, just add some code in the Todo class to trigger a breakpoint when a certain property is accessed.

When the breakpoint is hit, we can walk through the stack trace and view the change detection:

This method may seem strange at first, all variables Weird naming. But digging deeper, we noticed that it's doing something very simple: for each expression used in the template, it compares the current value of the property used in the expression to the previous value of that property.

If the property values ​​​​before and after are different, isChanged will be set to true, that’s it! Pretty much, it compares the values ​​using a method called looseNotIdentical().

What about the nested object owner?

We can see in the change detector code that the owner properties of the nested object are also being checked for differences. But only the firstname attribute is compared, not the lastname attribute. This is because lastname is not used in component template! Likewise, the top-level id attribute of Todo is not compared for the same reason.

With this, we can say with confidence:

默认情况下，Angular Change Detection 通过检查模板表达式的值是否已更改来工作。

我们还可以得出结论：

默认情况下，Angular 不做深度对象比较来检测变化，它只考虑模板使用的属性

为什么默认情况下更改检测会这样工作？

Angular 的主要目标之一是更加透明和易于使用，因此框架用户不必费尽心思调试框架并了解内部机制即可有效地使用它。

如果 Angular 默认更改检测机制基于组件输入的参考比较而不是默认机制，那会是什么情况？即使是像 TODO 应用程序这样简单的东西也很难构建：开发人员必须非常小心地创建一个新的 Todo，而不是简单地更新属性。

OnPush 变化检测策略

如果你觉得默认模式影响了性能，我们也可以自定义 Angular 更改检测。将组件更改检测策略更新为OnPush：

@Component({
    selector: 'todo-list',
    changeDetection: ChangeDetectionStrategy.OnPush,
    template: ...
})
export class TodoList {
    ...
}

现在让我们在应用程序中添加几个按钮：一个是通过直接改变列表的第一项来切换列表的第一项，另一个是向整个列表添加一个 Todo。代码如下所示：

@Component({
    selector: 'app',
    template: `<div>
                    <todo-list [todos]="todos"></todo-list>
               </div>
               <button (click)="toggleFirst()">Toggle First Item</button>
               <button (click)="addTodo()">Add Todo to List</button>`
})
export class App {
    todos:Array = initialData;

    constructor() {
    }

    toggleFirst() {
        this.todos[0].completed = ! this.todos[0].completed;
    }

    addTodo() {
        let newTodos = this.todos.slice(0);
        newTodos.push( new Todo(1, "TODO 4", 
            false, new Owner("John", "Doe")));
        this.todos = newTodos;
    }
}

现在让我们看看这两个新按钮的行为：

• 第一个按钮“切换第一项”不起作用！这是因为该toggleFirst()方法直接改变了列表中的一个元素。
  TodoList无法检测到这一点，因为它的输入参考todos没有改变
• 第二个按钮确实有效！请注意，该方法addTodo()创建了 todo 列表的副本，然后将项目添加到副本中，最后将 todos 成员变量替换为复制的列表。这会触发更改检测，因为组件检测到其输入中的参考更改：它收到了一个新列表！
• 在第二个按钮中，直接改变 todos 列表是行不通的！我们真的需要一个新的清单。

OnPush只是通过引用比较输入吗？

情况并非如此。当使用 OnPush 检测器时，框架将在 OnPush 组件的任何输入属性更改、触发事件或 Observable 触发事件时检查

尽管允许更好的性能，但OnPush如果与可变对象一起使用，则使用会带来很高的复杂性成本。它可能会引入难以推理和重现的错误。但是有一种方法可以使使用OnPush可行。

使用 Immutable.js 简化 Angular 应用程序的构建

如果我们只使用不可变对象和不可变列表来构建我们的应用程序，则可以OnPush透明地在任何地方使用，而不会遇到更改检测错误的风险。这是因为对于不可变对象，修改数据的唯一方法是创建一个新的不可变对象并替换之前的对象。使用不可变对象，我们可以保证：

• 新的不可变对象将始终触发OnPush更改检测
• 我们不会因为忘记创建对象的新副本而意外创建错误，因为修改数据的唯一方法是创建新对象

实现不可变的一个不错的选择是使用Immutable.js库。该库为构建应用程序提供了不可变原语，例如不可变对象（映射）和不可变列表。

避免变更检测循环：生产与开发模式

Angular 更改检测的重要属性之一是，与 AngularJs 不同，它强制执行单向数据流：当我们的控制器类上的数据更新时，更改检测运行并更新视图。

如何在 Angular 中触发变更检测循环？

一种方法是如果我们使用生命周期回调。例如，在TodoList组件中，我们可以触发对另一个组件的回调来更改其中一个绑定：

ngAfterViewChecked() {
    if (this.callback && this.clicked) {
        console.log("changing status ...");
        this.callback(Math.random());
    }
}

控制台中将显示一条错误消息：

EXCEPTION: Expression '{{message}} in App@3:20' has changed after it was checked

仅当我们在开发模式下运行 Angular 时才会抛出此错误消息。如果我们启用生产模式会发生什么？ 在生产模式下，错误不会被抛出，问题也不会被发现。

在开发阶段始终使用开发模式会更好，因为这样可以避免问题。这种保证是以 Angular 总是运行两次变更检测为代价的，第二次检测这种情况。在生产模式下，变更检测只运行一次。

打开/关闭变化检测，并手动触发它

在某些特殊情况下，我们确实想要关闭更改检测。想象一下这样一种情况，大量数据通过 websocket 从后端到达。我们可能只想每 5 秒更新一次 UI 的某个部分。为此，我们首先将更改检测器注入到组件中：

constructor(private ref: ChangeDetectorRef) {
    ref.detach();
    setInterval(() => {
      this.ref.detectChanges();
    }, 5000);
  }

正如我们所看到的，我们只是分离了变化检测器，这有效地关闭了变化检测。然后我们只需每 5 秒通过调用手动触发它detectChanges()。

现在让我们快速总结一下我们需要了解的关于 Angular 变更检测的所有内容：它是什么，它是如何工作的以及可用的主要变更检测类型是什么。

概括

Angular 更改检测是一个内置的框架功能，可确保组件数据与其 HTML 模板视图之间的自动同步。

更改检测的工作原理是检测常见的浏览器事件，如鼠标点击、HTTP 请求和其他类型的事件，并确定每个组件的视图是否需要更新。

变更检测有两种类型：

• 默认更改检测：Angular 通过比较事件发生前后的所有模板表达式值来决定是否需要更新视图，用于组件树的所有组件
• OnPush 更改检测：这通过检测是否已通过组件输入或使用异步管道订阅的 Observable 将某些新数据显式推送到组件中来工作

Angular默认更改检测机制实际上与 AngularJs 非常相似：它比较浏览器事件之前和之后模板表达式的值，以查看是否有更改。它对所有组件都这样做。但也有一些重要的区别：

一方面，没有变化检测循环，也没有 AngularJs 中命名的摘要循环。这允许仅通过查看其模板和控制器来推理每个组件。

另一个区别是，由于变化检测器的构建方式，检测组件变化的机制要快得多。

最后，与 AngularJs 不同的是，变化检测机制是可定制的。

更多编程相关知识，请访问：编程教学！！

The above is the detailed content of A brief analysis of the Change Detection mechanism in Angular. For more information, please follow other related articles on the PHP Chinese website!