[转载]AngularJS and scope.$apply

http://jimhoskins.com/2012/12/17/angularjs-and-apply.html

http://www.cnblogs.com/zhrj000/p/3383898.html

If you’ve written a non-trivial amount of code in AngularJS, you may have come across the $scope.$apply() method. On the surface, it may seem like just a method you call to get your bindings to update. But why does it exist? And when do you need to use it?

To really understand when to use $apply, it’s good to know exactly why we need to use it, so let’s dive in!

JavaScript is Turn Based

The JavaScript code we write doesn’t all run in one go, instead it executes in turns. Each of these turns runs uninterupted from start to finish, and when a turn is running, nothing else happens in our browser. No other JavaScript code runs, and our web page interface is completely frozen. This is why poorly coded JavaScript can freeze a web page.

Instead, whenever there is a task that takes some amount of time, such as an Ajax request, waiting for a click event, or setting a timeout, we set up a callback function and finish our current turn. Later, when the Ajax request completes, a click is detected, or the timer completes, a new JavaScript turn is created and the callback is run to completion.

Let’s look at an example JavaScript file:

var button = document.getElementById('clickMe');

function buttonClicked () {
  alert('the button was clicked');
}

button.addEventListener('click', buttonClicked);

function timerComplete () {
  alert('timer complete');
}

setTimeout(timerComplete, 2000);

When the JavaScript code is loaded, that is a single turn. It finds a button, adds a click listener, and sets a timeout. Then the turn is complete, and the browser will update the web page if necessary, and begin accepting user input.

If the browser detects a click on #clickMe, it creates a new turn, which executes the buttonClicked function. When that function returns, that turn is complete.

After 2000 milliseconds, the browser creates a new turn which calls timerComplete.

Our JavaScript code is run in turns, and in between the turns is when the page is repainted, and input is accepted.

How do we update bindings?

So Angular lets us bind parts of our interface to data in our JavaScript code, but how does it know when data changes, and the page needs updating?

There are a few solutions. The code needs to know when a value has changed. Right now there is no way for our code to be directly notified of changes on an object ¹. Instead there are two main strategies.

One strategy is to use special objects, where data is set via methods, not property assignments. Then changes can then be noted, and the page can be updated. This has the downside in that we must extend some special object. Also, for assigning, we must use a more verbose form obj.set('key', 'value') instead of obj.key = 'value'. Frameworks like EmberJS and KnockoutJS use this strategy.

AngularJS takes a different approach: allow any value to be used as a binding target. Then at the end of any JavaScript code turn, check to see if the value has changed. This may seem inneficient at first, but there are some clever strategies to reduce the performance hit. The big benefit is we can use normal objects and update our data however we want, and the changes will be noticed and reflected in our bindings.

For this strategy to work, we need to know when data has possibly changed, and this is where $scope.$apply comes into play.

$apply and $digest

That step that checks to see if any binding values have changed actually has a method, $scope.$digest(). That’s actually where the magic happens, but we almost never call it directly, instead we use $scope.$apply() which will call $scope.$digest() for you.

$scope.$apply() takes a function or an Angular expression string, and executes it, then calls $scope.$digest() to update any bindings or watchers.

So, when do you need to call $apply()? Very rarely, actually. AngularJS actually calls almost all of your code within an $apply call. Events like ng-click, controller initialization, $http callbacks are all wrapped in $scope.$apply(). So you don’t need to call it yourself, in fact you can’t. Calling $apply inside $apply will throw an error.

You do need to use it if you are going to run code in a new turn. And only if that turn isn’t being created from a method in the AngularJS library. Inside that new turn, you should wrap your code in $scope.$apply(). Here is an example. We are using setTimeout, which will execute a function in a new turn after a delay. Since Angular doesn’t know about that new turn, the update will not be reflected.

But, if we wrap the code for that turn in $scope.$apply(), the change will be noticed, and the page is updated.

As a convenience, AngularJS provides $timeout, which is like setTimeout, but automatically wraps your code in $apply by default. Use that, not this

If you write any code that uses Ajax without $http, or listens for events without using Angular’s ng-* listeners, or sets a timeout without $timeout, you should wrap your code in $scope.$apply

$scope.$apply() vs $scope.$apply(fn)

Sometimes I see examples where data is updated, and then $scope.$apply() is called with no arguments. This achieves the desired result, but misses some opportunities.

If your code isn’t wrapped in a function passed to $apply, and it throws an error, that error is thrown outside of AngularJS, which means any error handling being used in your application is going to miss it. $apply not only runs your code, but it runs it in a try/catch so your error is always caught, and the $digest call is in a finally clause, meaning it will run regardless of an error being thrown. That’s pretty nice.

Hopefully now you understand what $apply is and when to use it. If you only use what AngularJS provides you, you shouldn’t need to use it often. But if you begin writing directives where you are observing DOM elements directly, it is going to become necessary.

---

Object.observe has been proposed for ES5, but is only experimentally implemented now