Improve this DocCreating Custom Directives

This document explains when you'd want to create your own directives in your AngularJS app, and how to implement them.

What are Directives?

At a high level, directives are markers on a DOM element (such as an attribute, element name, comment or CSS class) that tell AngularJS's HTML compiler ($compile) to attach a specified behavior to that DOM element or even transform the DOM element and its children.

Angular comes with a set of these directives built-in, like ngBind, ngModel, and ngClass. Much like you create controllers and services, you can create your own directives for Angular to use. When Angular bootstraps your application, the HTML compiler traverses the DOM matching directives against the DOM elements.

Matching Directives

Before we can write a directive, we need to know how Angular's HTML compiler determines when to use a given directive.

In the following example, we say that the <input> element matches the ngModel directive.

<input ng-model="foo">

The following also matches ngModel:

<input data-ng-model="foo">

Normalization

Angular normalizes an element's tag and attribute name to determine which elements match which directives. We typically refer to directives by their case-sensitive camelCase normalized name (e.g. ngModel). However, since HTML is case-insensitive, we refer to directives in the DOM by lower-case forms, typically using dash-delimited attributes on DOM elements (e.g. ng-model).

The normalization process is as follows:

1. Strip x- and data- from the front of the element/attributes.
2. Convert the :, -, or _-delimited name to camelCase.

For example, the following forms are all equivalent and match the ngBind directive:

Directive types

$compile can match directives based on element names, attributes, class names, as well as comments.

All of the Angular-provided directives match attribute name, tag name, comments, or class name. The following demonstrates the various ways a directive (myDir in this case) can be referenced from within a template:

<my-dir></my-dir>
<span my-dir="exp"></span>
<!-- directive: my-dir exp -->
<span class="my-dir: exp;"></span>

Text and attribute bindings

During the compilation process the compiler matches text and attributes using the $interpolate service to see if they contain embedded expressions. These expressions are registered as watches and will update as part of normal digest cycle. An example of interpolation is shown below:

<a ng-href="img/{{username}}.jpg">Hello {{username}}!</a>

ngAttr attribute bindings

Web browsers are sometimes picky about what values they consider valid for attributes.

For example, considering this template:

<svg>
  <circle cx="{{cx}}"></circle>
</svg>

We would expect Angular to be able to bind to this, but when we check the console we see something like Error: Invalid value for attribute cx="{{cx}}". Because of the SVG DOM API's restrictions, you cannot simply write cx="{{cx}}".

With ng-attr-cx you can work around this problem.

If an attribute with a binding is prefixed with the ngAttr prefix (denormalized as ng-attr-) then during the binding it will be applied to the corresponding unprefixed attribute. This allows you to bind to attributes that would otherwise be eagerly processed by browsers (e.g. an SVG element's circle[cx] attributes). When using ngAttr, the allOrNothing flag of $interpolate is used, so if any expression in the interpolated string results in undefined, the attribute is removed and not added to the element.

For example, we could fix the example above by instead writing:

<svg>
  <circle ng-attr-cx="{{cx}}"></circle>
</svg>

If one wants to modify a camelcased attribute (SVG elements have valid camelcased attributes), such as viewBox on the svg element, one can use underscores to denote that the attribute to bind to is naturally camelcased.

For example, to bind to viewBox, we can write:

<svg ng-attr-view_box="{{viewBox}}">
</svg>

Creating Directives

First let's talk about the API for registering directives. Much like controllers, directives are registered on modules. To register a directive, you use the module.directive API. module.directive takes the normalized directive name followed by a factory function. This factory function should return an object with the different options to tell $compile how the directive should behave when matched.

The factory function is invoked only once when the compiler matches the directive for the first time. You can perform any initialization work here. The function is invoked using $injector.invoke which makes it injectable just like a controller.

We'll go over a few common examples of directives, then dive deep into the different options and compilation process.

For the following examples, we'll use the prefix my (e.g. myCustomer).

Template-expanding directive

Let's say you have a chunk of your template that represents a customer's information. This template is repeated many times in your code. When you change it in one place, you have to change it in several others. This is a good opportunity to use a directive to simplify your template.

Let's create a directive that simply replaces its contents with a static template:

Notice that we have bindings in this directive. After $compile compiles and links <div my-customer></div>, it will try to match directives on the element's children. This means you can compose directives of other directives. We'll see how to do that in an example below.

In the example above we in-lined the value of the template option, but this will become annoying as the size of your template grows.

If you are familiar with ngInclude, templateUrl works just like it. Here's the same example using templateUrl instead:

templateUrl can also be a function which returns the URL of an HTML template to be loaded and used for the directive. Angular will call the templateUrl function with two parameters: the element that the directive was called on, and an attr object associated with that element.

The restrict option is typically set to:

• 'A' - only matches attribute name
• 'E' - only matches element name
• 'C' - only matches class name

These restrictions can all be combined as needed:

• 'AEC' - matches either attribute or element or class name

Let's change our directive to use restrict: 'E':

For more on the restrict property, see the API docs.

Using an element for the myCustomer directive is clearly the right choice because you're not decorating an element with some "customer" behavior; you're defining the core behavior of the element as a customer component.

Isolating the Scope of a Directive

Our myCustomer directive above is great, but it has a fatal flaw. We can only use it once within a given scope.

In its current implementation, we'd need to create a different controller each time in order to re-use such a directive:

This is clearly not a great solution.

What we want to be able to do is separate the scope inside a directive from the scope outside, and then map the outer scope to a directive's inner scope. We can do this by creating what we call an isolate scope. To do this, we can use a directive's scope option:

Looking at index.html, the first <my-customer> element binds the info attribute to naomi, which we have exposed on our controller's scope. The second binds info to igor.

Let's take a closer look at the scope option:

//...
scope: {
  customerInfo: '=info'
},
//...

The scope option is an object that contains a property for each isolate scope binding. In this case it has just one property:

• Its name (customerInfo) corresponds to the directive's isolate scope property customerInfo.
• Its value (=info) tells $compile to bind to the info attribute.

For cases where the attribute name is the same as the value you want to bind to inside the directive's scope, you can use this shorthand syntax:

...
scope: {
  // same as '=customer'
  customer: '='
},
...

Besides making it possible to bind different data to the scope inside a directive, using an isolated scope has another effect.

We can show this by adding another property, vojta, to our scope and trying to access it from within our directive's template:

Notice that {{vojta.name}} and {{vojta.address}} are empty, meaning they are undefined. Although we defined vojta in the controller, it's not available within the directive.

As the name suggests, the isolate scope of the directive isolates everything except models that you've explicitly added to the scope: {} hash object. This is helpful when building reusable components because it prevents a component from changing your model state except for the models that you explicitly pass in.

Creating a Directive that Manipulates the DOM

In this example we will build a directive that displays the current time. Once a second, it updates the DOM to reflect the current time.

Directives that want to modify the DOM typically use the link option. link takes a function with the following signature, function link(scope, element, attrs) { ... } where:

• scope is an Angular scope object.
• element is the jqLite-wrapped element that this directive matches.
• attrs is a hash object with key-value pairs of normalized attribute names and their corresponding attribute values.

In our link function, we want to update the displayed time once a second, or whenever a user changes the time formatting string that our directive binds to. We will use the $interval service to call a handler on a regular basis. This is easier than using $timeout but also works better with end-to-end testing, where we want to ensure that all $timeouts have completed before completing the test. We also want to remove the $interval if the directive is deleted so we don't introduce a memory leak.

There are a couple of things to note here. Just like the module.controller API, the function argument in module.directive is dependency injected. Because of this, we can use $interval and dateFilter inside our directive's link function.

We register an event element.on('$destroy', ...). What fires this $destroy event?

There are a few special events that AngularJS emits. When a DOM node that has been compiled with Angular's compiler is destroyed, it emits a $destroy event. Similarly, when an AngularJS scope is destroyed, it broadcasts a $destroy event to listening scopes.

By listening to this event, you can remove event listeners that might cause memory leaks. Listeners registered to scopes and elements are automatically cleaned up when they are destroyed, but if you registered a listener on a service, or registered a listener on a DOM node that isn't being deleted, you'll have to clean it up yourself or you risk introducing a memory leak.

Creating a Directive that Wraps Other Elements

We've seen that you can pass in models to a directive using the isolate scope, but sometimes it's desirable to be able to pass in an entire template rather than a string or an object. Let's say that we want to create a "dialog box" component. The dialog box should be able to wrap any arbitrary content.

To do this, we need to use the transclude option.

What does this transclude option do, exactly? transclude makes the contents of a directive with this option have access to the scope outside of the directive rather than inside.

To illustrate this, see the example below. Notice that we've added a link function in script.js that redefines name as Jeff. What do you think the {{name}} binding will resolve to now?

Ordinarily, we would expect that {{name}} would be Jeff. However, we see in this example that the {{name}} binding is still Tobias.

The transclude option changes the way scopes are nested. It makes it so that the contents of a transcluded directive have whatever scope is outside the directive, rather than whatever scope is on the inside. In doing so, it gives the contents access to the outside scope.

Note that if the directive did not create its own scope, then scope in scope.name = 'Jeff'; would reference the outside scope and we would see Jeff in the output.

This behavior makes sense for a directive that wraps some content, because otherwise you'd have to pass in each model you wanted to use separately. If you have to pass in each model that you want to use, then you can't really have arbitrary contents, can you?

Next, we want to add buttons to this dialog box, and allow someone using the directive to bind their own behavior to it.

We want to run the function we pass by invoking it from the directive's scope, but have it run in the context of the scope where it's registered.

We saw earlier how to use =attr in the scope option, but in the above example, we're using &attr instead. The & binding allows a directive to trigger evaluation of an expression in the context of the original scope, at a specific time. Any legal expression is allowed, including an expression which contains a function call. Because of this, & bindings are ideal for binding callback functions to directive behaviors.

When the user clicks the x in the dialog, the directive's close function is called, thanks to ng-click. This call to close on the isolated scope actually evaluates the expression hideDialog(message) in the context of the original scope, thus running Controller's hideDialog function.

Often it's desirable to pass data from the isolate scope via an expression to the parent scope, this can be done by passing a map of local variable names and values into the expression wrapper fn. For example, the hideDialog function takes a message to display when the dialog is hidden. This is specified in the directive by calling close({message: 'closing for now'}). Then the local variable message will be available within the on-close expression.

Creating a Directive that Adds Event Listeners

Previously, we used the link function to create a directive that manipulated its DOM elements. Building upon that example, let's make a directive that reacts to events on its elements.

For instance, what if we wanted to create a directive that lets a user drag an element?

Creating Directives that Communicate

You can compose any directives by using them within templates.

Sometimes, you want a component that's built from a combination of directives.

Imagine you want to have a container with tabs in which the contents of the container correspond to which tab is active.

The myPane directive has a require option with value ^myTabs. When a directive uses this option, $compile will throw an error unless the specified controller is found. The ^ prefix means that this directive searches for the controller on its parents (without the ^ prefix, the directive would look for the controller on just its own element).

So where does this myTabs controller come from? Directives can specify controllers using the unsurprisingly named controller option. As you can see, the myTabs directive uses this option. Just like ngController, this option attaches a controller to the template of the directive.

If it is necessary to reference the controller or any functions bound to the controller's scope in the template, you can use the option controllerAs to specify the name of the controller as an alias. The directive needs to define a scope for this configuration to be used. This is particularly useful in the case when the directive is used as a component.

Looking back at myPane's definition, notice the last argument in its link function: tabsCtrl. When a directive requires a controller, it receives that controller as the fourth argument of its link function. Taking advantage of this, myPane can call the addPane function of myTabs.

If multiple controllers are required, the require option of the directive can take an array argument. The corresponding parameter being sent to the link function will also be an array.

angular.module('docsTabsExample', [])
.directive('myPane', function() {
  return {
    require: ['^myTabs', '^ngModel'],
    restrict: 'E',
    transclude: true,
    scope: {
      title: '@'
    },
    link: function(scope, element, attrs, controllers) {
      var tabsCtrl = controllers[0],
          modelCtrl = controllers[1];

      tabsCtrl.addPane(scope);
    },
    templateUrl: 'my-pane.html'
  };
});

Savvy readers may be wondering what the difference is between link and controller. The basic difference is that controller can expose an API, and link functions can interact with controllers using require.

Summary

Here we've seen the main use cases for directives. Each of these samples acts as a good starting point for creating your own directives.

You might also be interested in an in-depth explanation of the compilation process that's available in the compiler guide.

The $compile API page has a comprehensive list of directive options for reference.