This guide looks at how Angular manipulates the DOM with structural directives and how you can write your own structural directives to do the same thing.
Table of contents
- What are structural directives?
- NgIf case study
- The asterisk (*) prefix
- Inside NgFor
- Inside the NgSwitch directives
- Prefer the (*) prefix
- The <template> element
- Group sibling elements with <ng-container>
- Write a structural directive
What are structural directives?
Structural directives are responsible for HTML layout. They shape or reshape the DOM's structure, typically by adding, removing, or manipulating elements.
As with other directives, you apply a structural directive to a host element. The directive then does whatever it's supposed to do with that host element and its descendents.
Structural directives are easy to recognize. An asterisk (*) precedes the directive attribute name as in this example.
No brackets. No parentheses. Just
*ngIf set to a string.
You'll learn in this guide that the asterisk (*) is a convenience notation
and the string is a microsyntax rather than the usual
Angular desugars this notation into a marked-up
<template> that surrounds the
host element and its descendents.
Each structural directive does something different with that template.
Three of the common, built-in structural directives—NgIf, NgFor, and NgSwitch...—are described in the Template Syntax guide and seen in samples throughout the Angular documentation. Here's an example of them in a template:
This guide won't repeat how to use them. But it does explain how they work and how to write your own structural directive.
Throughout this guide, you'll see a directive spelled in both UpperCamelCase and lowerCamelCase.
Already you've seen
There's a reason.
NgIf refers to the directive class;
ngIf refers to the directive's attribute name.
A directive class is spelled in UpperCamelCase (
A directive's attribute name is spelled in lowerCamelCase (
The guide refers to the directive class when talking about its properties and what the directive does.
The guide refers to the attribute name when describing how
you apply the directive to an element in the HTML template.
There are two other kinds of Angular directives, described extensively elsewhere: (1) components and (2) attribute directives.
A component manages a region of HTML in the manner of a native HTML element. Technically it's a directive with a template.
An attribute directive changes the appearance or behavior
of an element, component, or another directive.
For example, the built-in
changes several element styles at the same time.
You can apply many attribute directives to one host element. You can only apply one structural directive to a host element.
NgIf case study
NgIf is the simplest structural directive and the easiest to understand.
It takes a boolean expression and makes an entire chunk of the DOM appear or disappear.
ngIf directive doesn't hide elements with CSS. It adds and removes them physically from the DOM.
Confirm that fact using browser developer tools to inspect the DOM.
The top paragraph is in the DOM. The bottom, disused paragraph is not; in its place is a comment about "template bindings" (more about that later).
When the condition is false,
NgIf removes its host element from the DOM,
detaches it from DOM events (the attachments that it made),
detaches the component from Angular change detection, and destroys it.
The component and DOM nodes can be garbage-collected and free up memory.
Why remove rather than hide?
A directive could hide the unwanted paragraph instead by setting its
display style to
While invisible, the element remains in the DOM.
The difference between hiding and removing doesn't matter for a simple paragraph. It does matter when the host element is attached to a resource intensive component. Such a component's behavior continues even when hidden. The component stays attached to its DOM element. It keeps listening to events. Angular keeps checking for changes that could affect data bindings. Whatever the component was doing, it keeps doing.
Although invisible, the component—and all of its descendant components—tie up resources. The performance and memory burden can be substantial, responsiveness can degrade, and the user sees nothing.
On the positive side, showing the element again is quick. The component's previous state is preserved and ready to display. The component doesn't re-initialize—an operation that could be expensive. So hiding and showing is sometimes the right thing to do.
But in the absence of a compelling reason to keep them around,
your preference should be to remove DOM elements that the user can't see
and recover the unused resources with a structural directive like
These same considerations apply to every structural directive, whether built-in or custom. Before applying a structural directive, you might want to pause for a moment to consider the consequences of adding and removing elements and of creating and destroying components.
The asterisk (*) prefix
Surely you noticed the asterisk (*) prefix to the directive name and wondered why it is necessary and what it does.
*ngIf displaying the hero's name if
The asterisk is "syntactic sugar" for something a bit more complicated.
Internally, Angular desugars it in two stages.
First, it translates the
*ngIf="..." into a template attribute,
template="ngIf ...", like this.
Then it translates the template attribute into a template element, wrapped around the host element, like this.
*ngIfdirective moved to the
<template>element where it became a property binding,
- The rest of the
<div>, including its class attribute, moved inside the
None of these forms are actually rendered. Only the finished product ends up in the DOM.
Angular consumed the
<template> content during its actual rendering and
<template> with a diagnostic comment.
NgSwitch... directives follow the same pattern.
Angular transforms the
*ngFor in similar fashion from asterisk (*) syntax through
template attribute to template element.
Here's a full-featured application of
NgFor, written all three ways:
This is manifestly more complicated than
ngIf and rightly so.
NgFor directive has more features, both required and optional, than the
NgIf shown in this guide.
NgFor needs a looping variable (
let hero) and a list (
You enable these features in the string assigned to
ngFor, which you write in Angular's microsyntax.
Everything outside the
ngFor string stays with the host element
<div>) as it moves inside the
In this example, the
[ngClass]="odd" stays on the
The Angular microsyntax lets you configure a directive in a compact, friendly string.
The microsyntax parser translates that string into attributes on the
letkeyword declares a template input variable that you reference within the template. The input variables in this example are
odd. The parser translates
let i, and
let oddinto variables named,
The microsyntax parser takes
trackby, title-cases them (
TrackBy), and prefixes them with the directive's attribute name (
ngFor), yielding the names
ngForTrackBy. Those are the names of two
NgForinput properties . That's how the directive learns that the list is
heroesand the track-by function is
NgFordirective loops through the list, it sets and resets properties of its own context object. These properties include
oddand a special property named
let-oddvariables were defined as
let odd=odd. Angular sets them to the current value of the context's
The context property for
let-herowasn't specified. It's intended source is implicit. Angular sets
let-heroto the value of the context's
NgForhas initialized with the hero for the current iteration.
The API guide describes additional
NgFordirective properties and context properties.
These microsyntax mechanisms are available to you when you write your own structural directives.
Studying the source code for
NgFor is a great way to learn more.
Template input variable
A template input variable is a variable whose value you can reference within a single instance of the template.
There are several such variables in this example:
All are preceded by the keyword
A template input variable is not the same as a template reference variable, neither semantically nor syntactically.
You declare a template input variable using the
let keyword (
The variable's scope is limited to a single instance of the repeated template.
You can use the same variable name again in the definition of other structural directives.
You declare a template reference variable by prefixing the variable name with
A reference variable refers to its attached element, component or directive.
It can be accessed anywhere in the entire template.
Template input and reference variable names have their own namespaces. The
let hero is never the same
variable as the
hero declared as
One structural directive per host element
Someday you'll want to repeat a block of HTML but only when a particular condition is true.
You'll try to put both an
*ngFor and an
*ngIf on the same host element.
Angular won't let you. You may apply only one structural directive to an element.
The reason is simplicity. Structural directives can do complex things with the host element and its descendents.
When two directives lay claim to the same host element, which one takes precedence?
Which should go first, the
NgIf or the
NgFor? Can the
NgIf cancel the effect of the
If so (and it seems like it should be so), how should Angular generalize the ability to cancel for other structural directives?
There are no easy answers to these questions. Prohibiting multiple structural directives makes them moot.
There's an easy solution for this use case: put the
*ngIf on a container element that wraps the
One or both elements can be an
ng-container so you don't have to introduce extra levels of HTML.
Inside NgSwitch directives
The Angular NgSwitch is actually a set of cooperating directives:
Here's an example.
The switch value assigned to
hero.emotion) determines which
(if any) of the switch cases are displayed.
NgSwitch itself is not a structural directive.
It's an attribute directive that controls the behavior of the other two switch directives.
That's why you write
NgSwitchDefault are structural directives.
You attach them to elements using the asterisk (*) prefix notation.
NgSwitchCase displays its host element when its value matches the switch value.
NgSwitchDefault displays its host element when no sibling
NgSwitchCase matches the switch value.
The element to which you apply a directive is its host element.
<happy-hero> is the host element for the happy
<unknown-hero> is the host element for the
As with other structural directives, the
can be desugared into the template attribute form.
That, in turn, can be desugared into the
<template> element form.
Prefer the asterisk (*) syntax.
The asterisk (*) syntax is more clear than the other desugared forms. Use <ng-container> when there's no single element to host the directive.
While there's rarely a good reason to apply a structural directive in template attribute or element form,
it's still important to know that Angular creates a
<template> and to understand how it works.
You'll refer to the
<template> when you write your own structural directive.
The HTML 5 <template>
is a formula for rendering HTML.
It is never displayed directly.
In fact, before rendering the view, Angular replaces the
<template> and its contents with a comment.
If there is no structural directive and you merely wrap some elements in a
those elements disappear.
That's the fate of the middle "Hip!" in the phrase "Hip! Hip! Hooray!".
Angular erases the middle "Hip!", leaving the cheer a bit less enthusiastic.
A structural directive puts a
<template> to work
as you'll see when you write your own structural directive.
Group sibling elements with <ng-container>
There's often a root element that can and should host the structural directive.
The list element (
<li>) is a typical host element of an
When there isn't a host element, you can usually wrap the content in a native HTML container element,
such as a
<div>, and attach the directive to that wrapper.
Introducing another container element—typically a
group the elements under a single root is usually harmless.
Usually ... but not always.
The grouping element may break the template appearance because CSS styles neither expect nor accommodate the new layout. For example, suppose you have the following paragraph layout.
You also have a CSS style rule that happens to apply to a
<span> within a
The constructed paragraph renders strangely.
p span style, intended for use elsewhere, was inadvertently applied here.
Another problem: some HTML elements require all immediate children to be of a specific type.
For example, the
<select> element requires
You can't wrap the options in a conditional
<div> or a
When you try this,
the drop down is empty.
The browser won't display an
<option> within a
<ng-container> to the rescue
<ng-container> is a grouping element that doesn't interfere with styles or layout
because Angular doesn't put it in the DOM.
Here's the conditional paragraph again, this time using
It renders properly.
Now conditionally exclude a select
The drop down works properly.
<ng-container> is a syntax element recognized by the Angular parser.
It's not a directive, component, class, or interface.
when you intend to conditionally execute all of them as a single block.
<ng-container> satisfies a similar need in Angular templates.
Write a structural directive
In this section, you write an
UnlessDirective structural directive
that does the opposite of
NgIf displays the template content when the condition is
UnlessDirective displays the content when the condition is false.
Creating a directive is similar to creating a component.
Directivedecorator (instead of the
ViewContainerRefsymbols; you'll need them for any structural directive.
Apply the decorator to the directive class.
Set the CSS attribute selector that identifies the directive when applied to an element in a template.
Here's how you might begin:
The directive's selector is typically the directive's attribute name in square brackets,
The brackets define a CSS
The directive attribute name should be spelled in lowerCamelCase and begin with a prefix.
ng. That prefix belongs to Angular.
Pick something short that fits you or your company.
In this example, the prefix is
The directive class name ends in
Directive per the style guide.
Angular's own directives do not.
TemplateRef and ViewContainerRef
A simple structural directive like this one creates an
from the Angular-generated
<template> and inserts that view in a
adjacent to the directive's original
<p> host element.
You'll acquire the
<template> contents with a
and access the view container through a
You inject both in the directive constructor as private variables of the class.
The myUnless property
The directive consumer expects to bind a true/false condition to
That means the directive needs a
myUnless property, decorated with
@Input in the Template Syntax guide.
Angular sets the
myUnless property whenever the value of the condition changes.
myUnless property does work, it needs a setter.
If the condition is falsy and the view hasn't been created previously, tell the view container to create the embedded view from the template.
If the condition is truthy and the view is currently displayed, clear the container which also destroys the view.
Nobody reads the
myUnless property so it doesn't need a getter.
The completed directive code looks like this:
Add this directive to the
declarations array of the AppModule.
Then create some HTML to try it.
condition is falsy, the top (A) paragraph appears and the bottom (B) paragraph disappears.
condition is truthy, the top (A) paragraph is removed and the bottom (B) paragraph appears.
You can both try and download the source code for this guide in the
Here is the source from the
- that structural directives manipulate HTML layout.
- to use
<ng-container>as a grouping element when there is no suitable host element.
- that the Angular desugars asterisk (*) syntax into a
- how that works for the
- about the microsyntax that expands into a
- to write a custom structural directive,