Heads Up! It's all about the V1 Spec.
This document is more yours than it is mine. It makes me happy that it has been able to help people. To do better I moved this document from the original gist to this repo so multiple people can work together and improve it.
If you like what you find here, please create issues with ideas as to what more can we add to this repository. Like examples, images, graphical representations for the terminologies, etc. via issues. Let's make everyone love the platform :).
Chinese - https://github.com/Tencent/omi/blob/master/tutorial/shadow-dom-in-depth.cn.md - Thanks @eyea.
- Examples. I am putting examples that'll help everyone understand it better, step by step. Check them out.
Let me know if a "Everything you need to know about Custom Elements" document like this one would help you. If so, I'll put one up 👨💻.
- Chrome : Works
- Firefox : Works
- Opera : Works
- Safari : Works but few things are buggy.
- Edge : Under Consideration.
Comprehensive browser support info can be found here: https://caniuse.com/#feat=shadowdomv1.
In a nutshell, Shadow DOM enables local scoping for HTML & CSS.
Shadow DOM fixes CSS and DOM. It introduces scoped styles to the web platform. Without tools or naming conventions, you can bundle CSS with markup, hide implementation details, and author self-contained components in vanilla JavaScript. - https://developers.google.com/web/fundamentals/web-components/shadowdom
It's like its own little world which hardly affects or gets affected by the outside world.
It's what you write as a component author to abstract away the implementation details of your component. It can also decide what to do with the user-provided light DOM.
- DOM : What we get over the wire (or wireless :|) is a string of text. To render something on the screen, the browsers have to parse that string of text and convert it into a data model so it can understand things better. It also preserves the hierarchy from the original string by putting those parsed objects in a tree structure.
We need to do that to make the machines understand our documents better. This tree like data model of our document is called Document Object Model.
- Component Author : The person who creates a component and defines how it works. Generally the person who writes a lot of shadow DOM code. Example - Browser vendors create the input
element.
- Component User : Well, they use components built by authors. They can pass in light DOM and set attributes and properties on your component. They can even extend the internals of a component if they want. Example - we, the users who use the input
element.
- Shadow Root: It's what gets attached to an element to give that element its shadow DOM. Technically it's a non-element node, a special kind of DocumentFragment.
<custom-picture>
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ ISOLATION
#shadow-root
...
____________________________________________ DOCUMENT FRAGMENT
<!--LIGHT DOM-->
</custom-picture>
Throughout the document, I have put shadow root inside those weird ASCII boundaries. This will put more emphasis on thinking how they are actually document fragments that have a wall around them.
- Shadow Host: The element to which the shadow root gets attached. A host can access its shadow root via a property on itself: .shadowRoot
.
- Shadow Tree : All the elements that go into the Shadow Root, which is scoped from outside world, is called Shadow Tree.
The elements in a shadow tree are not descendants of the shadow host in general (including for the purposes of Selectors like the descendant combinator) - Spec
- Light DOM: - The set of DOM elements we can sandwich between the opening and closing tags. - The DOM that lives outside shadow DOM. - The DOM, the user of your element writes. - The DOM that is the actual children of your element.
<custom-picture>
^^^^^^^^^^^^^^^^^^^^^^^^^^^
#shadow-root
___________________________
<!--Light DOM-->
<img src="https://path.to/a-kitten.png">
<cite>A Nice Kitten!</cite>
<!--Light DOM ends-->
</custom-picture>
- DocumentFragment:
The DocumentFragment interface represents a minimal document object that has no parent. It is used as a lightweight version of Document to store a segment of a document structure comprised of nodes just like a standard document. The key difference is that because the document fragment isn't part of the actual DOM's structure, changes made to the fragment don't affect the document, cause reflow, or incur any performance impact that can occur when changes are made. - MDN
<div class="dom"></div>
let el = document.querySelector('.dom');
el.attachShadow({ mode: 'open' });
// Just like prototype & constructor bi-directional references, we have...
el.shadowRoot; // the shadow root.
el.shadowRoot.host; // the element itself.
// Put something in shadow DOM
el.shadowRoot.innerHTML = 'Hi I am shadowed!';
// Like any other normal DOM operation.
let hello = document.createElement('span');
hello.textContent = 'Hi I am shadowed but wrapped in span';
el.shadowRoot.appendChild(hello);
Yes! But here are the differences from MDN.
ParentNode.append()
allows you to also append DOMString object, whereasNode.appendChild()
only accepts Node objects.ParentNode.append()
has no return value, whereasNode.appendChild()
returns the appended Node object.ParentNode.append()
can append several nodes and strings, whereasNode.appendChild()
can only append one node.
Well, it doesn't work. Because the browser already hosts its own shadow DOM for those elements. Bunch of red colored english alphabets will be thrown at console's face. 😰
- Shadow DOM cannot be removed once created; it can only be replaced with a new one.
- If you are creating a custom element, you should be creating the shadowRoot in its constructor. It can also be probably called in
connectedCallback()
but I am not sure if that introduces performance problems or any other problems. 🤷♂️ - To see how browsers implement shadow DOM for elements like
input
ortextarea
, Go toDevTools > Settings > Elements > [x] Show user agent shadow DOM
.
You saw the {mode: "open"}
in the attachShadow()
method right? Yeah! That's it. What open mode does is that it provides a way for us to reach into the shadow DOM to access the element's contents. It also lets us access the host element from within the shadow DOM.
This is done by the two implicit properties created when we call attachShadow()
in open
mode.
- The element gets a property called
shadowRoot
which points to the shadow DOM being attached to it. - The
shadowRoot
gets a property calledhost
pointing to the element itself.
// From the "How to create shadow DOM" example
el.attachShadow({ mode: 'open' });
// Just like prototype & constructor bi-directional references, we have...
el.shadowRoot; // the shadow root.
el.shadowRoot.host; // the el itself.
Pass {mode: "closed"}
to attachShadow()
to create a closed shadow DOM. It makes the shadow DOM inaccessible from JS.
el.shadowRoot; // null
Almost always use open
mode shadow DOMs because they make it possible for both the component author and user to change things how they want.
Remember we did el.shadowRoot
stuff up there? Yeah! That won't work with closed
mode. The element doesn't get any reference to its shadow DOM, which is a problem when you want to access the shadow DOM for manipulation.
class CustomPicture extends HTMLElement {
constructor() {
this.attachShadow({ mode: 'open' }); // this.shadowRoot exists. Add or remove stuff in there using this ref.
this.attachShadow({ mode: 'closed' }); // this.shadowRoot returns null. Bummer!
}
}
// You could always do the following in your constructor.
// but it totally defies the purpose of using closed mode.
this._shadowRoot = this.attachShadow({ mode: 'closed' });
Also, closed mode isn't a security mechanism. It just gives a fake sense of security. Nobody can stop someone from modifying how Element.prototype.attachShadow()
works.
- They are scoped.
- They don't leak out.
- They can have simple names.
- They are cool, be like them. 😎
<custom-picture>
#shadow-root
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
<style>
/*Applies only to spans inside shadow DOM. Doesn't leak out.*/
span {
color: red;
}
</style>
<span>Hello!</span>
__________________________________________________________________
</custom-picture>
Yeeeeaaaah.. but not in all browsers. 😕
<custom-picture>
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
#shadow-root
<!--All styles coming from custom-picture.css will be scoped inside this shadow root-->
<link rel="stylesheet" href="custom-picture.css">
<span>Hello!</span>
_____________________________________________________
Yeah. In some ways. Only the properties that are inherited will make their way through the shadow DOM boundary. Examples:
- color
- background
- font-family, etc.
The *
selector also affects things because *
means all elements and that includes the element to which you are attaching the shadow root to (the host element). Things which get applied to the host and can be inherited will pass the shadow DOM boundary to apply to inner elements.
- :host: targets the host element. BUT!
/* winner */
custom-picture {
background: red;
}
/* loser */
#shadow-root
<style>
:host {
background: green;
}
</style>
- :host(
<selector>
): Does the component host match the selector ? Basically, allows us to target different states of the same host. Examples:
:host([disabled]) {
...;
}
:host(:focus) {
...;
}
:host(:focus) span {
/*change all spans inside the element when the host has focus*/
}
The :host()
functional pseudo class can only take <compound-selector>
but not a combinator like space or ">" or "+" etc.
That means you can only choose the attributes and other aspects of the host element.
Nope! Only affects how we do things in JS.
Based on how shadow root or in general DocumentFragment works, user-agent styles (global) shouldn't have applied to all the elements inside a shadow root. So how do they work?
From the spec...
Window
s gain a private slot[[defaultElementStylesMap]]
which is a map of local names to stylesheets. This makes it possible to write elements inside shadow root and still get the default browser styles applied to them.
The purpose of having styles inside a shadow DOM is to just have default styles and provide hooks via CSS custom properties so component users can make changes to those default styles via, CSS custom properties (a.k.a CSS variables).
<business-card>
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
#shadow-root
<h1 class="card-title">Hardcoded Title - </h1>
---------------------------------------------------
</business-card>
/*Inside shadow DOM*/
.card-title {
color: var(--card-title-color, #000);
}
/*Component users can then override this color as*/
business-card {
--card-title-color: magenta;
}
- A shadow host is outside of the shadow tree it hosts, and so would ordinarily be untargettable by any selectors evaluated in the context of the shadow tree (as selectors are limited to a single tree), but it is sometimes useful to be able to style it from inside the shadow tree context.
- To work around that problem, the shadow host is treated as replacing the shadow root node.
- When considered within its own shadow trees, the shadow host is featureless. Only the
:host
and:host()
pseudo classes are allowed to match it.
Events work towards maintaining the encapsulation provided by the shadow DOM. Essentially, if an event occurs somewhere in the shadow DOM, to outside world it'll look as if that event has triggered from the host element itself and not a specific part of the shadow DOM. This is called re-targeting of the event.
Inside the shadow DOM, however, the events aren't retargeted and we can find out which specific element an event was associated with.
Say our flattened DOM tree looks like this:
<body>
<custom-picture>
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
#shadow-root
<button> Hello </button>
---------------------------------
</custom-picture>
</body>
On click of the button, to body, or anywhere outside custom-picture, the event.target
will point to the <custom-picture>
itself.
If the shadow tree is open, calling
event.composedPath()
will return an array of nodes that the event traveled through.
Inside <custom-picture>
however, the event target will be the button which was really clicked.
Most events bubble out of the shadow DOM boundary, and when they do they get re-targered. Some events aren't allowed to pass that boundary. Precisely these:
- abort
- error
- select
- change
- load
- reset
- resize
- scroll
- selectstart
Slots are a pretty big thing in Shadow DOM.
Slots are placeholders inside your component that users can fill with their own markup. - https://developers.google.com/web/fundamentals/web-components/shadowdom#slots
When creating custom components, we want to be able to provide only the necessary markup that goes into a particular component and use/group/style that as we want to as component authors.
The DOM that a component user provides is called light DOM and slots are the way we arrange, style, and group those elements.
There are two aspects of a slot:
- Light DOM Elements : They say which slot they wanna go into.
<custom-picture>
<!--Light DOM <img> saying it should be put into the "profile-picture" slot-->
<img src="assets/user.svg" slot="profile-picture">
</custom-picture>
- The Actual Slot : The
<slot>
element, residing somewhere in shadow DOM with a name for itself to be found by light DOM.
<custom-picture>
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
#shadow-root
<slot name="profile-picture">
<!--The <img> from the light DOM gets rendered here!-->
</slot>
_________________________________________
Elements are allowed to "cross" the shadow DOM boundary when a <slot> invites them in. These elements are called distributed nodes. Conceptually, distributed nodes can seem a bit bizarre. Slots don't physically move DOM; they render it at another location inside the shadow DOM. - https://developers.google.com/web/fundamentals/web-components/shadowdom#slots
You'll see nothing rendered. Here's why:
- A host element that has shadow DOM, only renders stuff that goes inside its shadow DOM.
- In order to get light DOM elements rendered, they need to be part of the shadow DOM.
- The way we make them part of the shadow DOM is by putting them in slots.
- In our example above, there's no element in light DOM that wants to go into a slot named
profile-picture
. - Since there's no one, the
<img>
from light DOM is not rendered. - Takeaway: named slots accommodate only those light DOM elements which specify they want to go into that particular slot.
This will require us to have a general purpose slot in our shadow DOM. A slot without a name. Example -
<custom-picture>
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
#shadow-root
<!--General purpose slot, render every element from light DOM that doesn't mention a slot name, here.-->
<slot>
<!--The <img> from the light DOM gets rendered here!-->
</slot>
_________________________________________
<!--Light DOM-->
<img src="assets/user.svg">
</custom-picture>
Woah! 😲 Actually, we can have duplicate unnamed or named slots but they are essentially ignored, since light DOM elements will go into the first slot they match.
<custom-picture>
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
#shadow-root
<slot>
<!--The <img> from the light DOM gets rendered here! Winner!-->
</slot>
<slot>
<!--Doesn't come here!-->
</slot>
_________________________________________
<!--Light DOM-->
<img src="assets/user.svg">
</custom-picture>
Nothing will be rendered unless there's fallback content provided by the slot itself. Providing fallback content is easy:
#shadow-root
<slot name="nobody-comes-here">
<h1> I'll show up when no slot content is provided!</h1>
</slot>
<style>
/*And that fallback can be styled from within the shadow DOM just like we do styles*/
slot[name="nobody-comes-here"] h1 {
color: #bada55;
}
</style>
Light DOM elements that go into a slot are called slotted elements. As mentioned above, these are also called distributed elements which cross the shadow DOM boundary.
These slotted elements can be styled using the ::slotted()
functional pseudo element. The syntax is as follows:
::slotted(<compound-selector >) {
/* styles */
}
Example:
<custom-picture>
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
#shadow-root
<slot>
<!--The <img> from the light DOM gets rendered here!-->
</slot>
<style>
/* find the slotted image and set their width and height */
::slotted(img) {
width: 256px;
height: 256px;
}
</style>
_________________________________________
<!--Light DOM-->
<img src="assets/user.svg">
</custom-picture>
Here's how the spec formally defines it:
The ::slotted() pseudo-element represents the elements assigned, after flattening, to a slot. This pseudo-element only exists on slots.
Flattened trees are here.
An important thing to remember is that only direct children of the host element can be assigned to a slot. For example:
<custom-picture>
<div class="picture-wrapper">
<!--This won't work! Slots can't pick descendants out of the host element's light DOM tree and put them in.-->
<img src="assets/user.svg" slot="assign-me" />
</div>
</custom-picture>
I, however, don't know why that's not possible and what the reasons are behind it, so I created this bug.
It may sound like we don't need to think about this scenario but it's often required.
<parent-element>
<!--parent-element uses child-element in its shadow DOM and we want this span to render inside that child-element's shadow DOM-->
<span slot="parent-slot">Finally</span>
</parent-element>
Here's what our custom elements look like:
class ParentElement extends HTMLElement {
constructor() {
super();
this.attachShadow({ mode: 'open' });
this.shadowRoot.innerHTML = `
<!--We specify a slot property on the slot itself. Which specifies where it goes in the child-element's shadow DOM-->
<child-element>
<slot name="parent-slot" slot="child-slot"></slot>
</child-element>
`;
}
}
class ChildElement extends HTMLElement {
constructor() {
super();
this.attachShadow({ mode: 'open' });
this.shadowRoot.innerHTML = `
<slot name="child-slot">
<!--The span with the thext "Finally" gets rendered here!-->
</slot>
`;
}
}
window.customElements.define('parent-element', ParentElement);
window.customElements.define('child-element', ChildElement);
- To find the elements that went into a slot -
slot.assignedNodes()
; - To find out which slot an light DOM element is assigned to -
element.assignedSlot
;