Giter Club home page Giter Club logo

skatejs's Introduction

Skate

NPM version Build Status Size Join the chat at https://gitter.im/skatejs/skatejs Commitizen friendly Semantic Release Follow @skate_js on Twitter

Sauce Test Status

Skate is a library built on top of the W3C web component specs that enables you to write functional and performant web components with a very small footprint.

  • Functional rendering pipeline backed by Google's Incremental DOM.
  • Inherently cross-framework compatible. For example, it works seamlessly with - and complements - React and other frameworks.
  • Skate itself is only 4k min+gz.
  • It's very fast.
  • It works with multiple versions of itself on the page.

HTML

<x-hello name="Bob"></x-hello>

JavaScript

skate.define('x-hello', {
  props: {
    name: { attribute: true }
  },
  render (elem) {
    return skate.h('div', `Hello, ${elem.name}`);
  }
});

Result

<x-hello name="Bob">Hello, Bob!</x-hello>

Whenever you change the name property - or attribute - the component will re-render, only changing the part of the DOM that requires updating.

Documentation

Installing

There's a couple ways to consume Skate.

NPM

npm install skatejs

Skate exports a UMD definition so you can:

import * as skate from 'skatejs';
const skate = require('skatejs');
require(['skatejs'], function (skate) {});

There's three files in dist/. Each has a UMD definition and a corresponding sourcemap file:

  1. index.js - This is the main entry point in the package.json without dependencies.
  2. index-with-deps.js - Unminified with dependencies.
  3. index-with-deps.min.js - Minified with dependencies.

Script Tag

<script src="https://unpkg.com/skatejs/dist/index-with-deps.min.js"></script>

Since Skate exports a UMD definition, you can then access it via the global:

const skate = window.skate;

Dependencies

Skate doesn't require you provide any external dependencies, but recommends you provide some web component polyfills depending on what browsers you require support for. Skate requires both Custom Elements and Shadow DOM v1.

To get up and running quickly with our recommended configuration, we've created a single package called skatejs-web-components where all you have to do is load it before Skate.

npm install skatejs skatejs-web-components

And then you can import it:

import 'skatejs-web-components';
import { define, vdom } from 'skatejs';

Or you can use script tags:

<script src="https://unpkg.com/skatejs-web-components/dist/index.min.js"></script>
<script src="https://unpkg.com/skatejs/dist/index-with-deps.min.js"></script>

If you want finer grained control about which polyfills you use, you'll have to BYO Custom Element and Shadow DOM polyfills.

Browser Support

Skate supports all major browsers and some older versions of IE. IE support depends on which polyfills you include:

Resources

Questions?

If you have any questions about Skate you can use one of these:

Terminology

Let's define some terms used in these docs:

  • polyfill, polyfilled, polyfill-land - no native web component support.
  • upgrade, upgraded, upgrading - when an element is initialised as a custom element.

Examples

The following are some simple examples to get you started with Skate.

Counter

The following is a simple counter that increments the count for every second that passes.

const sym = Symbol();

skate.define('x-counter', {
  props: {
    // By declaring the property an attribute, we can now pass an initial value
    // for the count as part of the HTML.
    count: skate.prop.number({ attribute: true })
  },
  attached(elem) {
    // We use a symbol so we don't pollute the element's namespace.
    elem[sym] = setInterval(() => ++elem.count, 1000);
  },
  detached(elem) {
    // If we didn't clean up after ourselves, we'd continue to render
    // unnecessarily.
    clearInterval(elem[sym]);
  },
  render(elem) {
    // By separating the strings (and not using template literals or string
    // concatenation) it ensures the strings are diffed indepenedently. If
    // you select "Count" with your mouse, it will not deselect whenr endered.
    return skate.h('div', 'Count ', elem.count);
  }
});

To use this, all you'd need to do in your HTML somewhere is:

<x-counter count="1"></x-counter>

API

Each API point is accessible on the main skate object:

const { define, vdom } = skate;

Or can be imported by name in ES2015:

import { define, vdom } from 'skatejs';

define(name, definition)

The name is a string that is the tag name of the custom element that you are creating. It must be a "valid custom element name" as specified in the spec. For example, my-component, is a valid custom element name.

The definition argument is a class / constructor or object literal of component definition to use for your custom element. The recommended way is to use ES2015 classes:

import { Component, define } from 'skatejs';

const MyComponent = define('my-component', class extends Component {});

A simpler way, especially for ES5 users, would be to just pass an object literal.

import { define } from 'skatejs';

const MyComponent = define('my-component', {});

Using this method will automatically extend the base Component for you.

This is exactly the same thing as doing:

import { Component, define } from 'skatejs';

const MyComponent = define('my-component', Component.extend({}));

You can also use Component.extend() to eliminate the boilerplate of extending base classes in ES5:

import { Component, define } from 'skatejs';

const MyComponent1 = define('my-component-1', {});
const MyComponent2 = define('my-component-2', MyComponent1.extend({});

Whichever method you use, define() will return you a constructor you can use to create a new instance of your element:

const myElement = new MyComponent();

prototype

The element's prototype. This is the first thing that happens in the element's lifecycle.

skate.define('my-component', {
  prototype: {
    get someProperty () {},
    set someProperty () {},
    someMethod () {},
  }
});

props

Custom properties that should be defined on the element. These are set up after the created lifecycle callback is called.

skate.define('my-component', {
  props: { ...props }
});

Custom properties, when set, queue a render(). This happens after a setTimeout() so that you only trigger a single render for a series of property sets.

The custom property definition accepts the following options.

attribute

Whether or not to link the property to an attribute. This can be either a Boolean or String.

  • If it's false, it's not linked to an attribute. This is the default.
  • If it's true, the property name is dash-cased and used as the attribute name it should be linked to.
  • If it's a String, the value is used as the attribute name it should be linked to.

Attributes are set from props on your element once it is inserted into the DOM.

skate.define('my-component', {
  props: {
    myProp: {
      attribute: true
    }
  }
});

When you declare a linked attribute, it automatically adds this attribute to the list of observedAttributes.

coerce

A function that coerces the incoming property value and returns the coerced value. This value is used as the internal value of the property.

skate.define('my-component', {
  props: {
    myProp: {
      coerce (value) {
        return value;
      }
    }
  }
});

The parameters passed to the function are:

  • value - the value that should be coerced
default

Specifies the default value of the property. If the property is ever set to null or undefined, instead of being empty, the default value will be used instead.

skate.define('my-component', {
  props: {
    myProp: {
      default: 'default value'
    }
  }
});

You may also specify a function that returns the default value. This is useful if you are doing calculations or need to return a reference:

skate.define('my-component', {
  props: {
    myProp: {
      default (elem, data) {
        return [];
      }
    }
  }
});

The parameters passed to the function are:

  • elem - the component element
  • data - an object containing information about the property
    • name - the property name

If specified, default will set the property value but will not be reflected back to the attribute, if linked. If at any point the property is set back to the default, the attribute will be removed, if linked.

deserialize

A function that converts the linked attribute value to the linked property value.

skate.define('my-component', {
  props: {
    myProp: {
      deserialize (value) {
        return value.split(',');
      }
    }
  }
});

The parameters passed to the function are:

  • value - the property value that needs to be coerced to the attribute value.
get

A function that is used to return the value of the property. If this is not specified, the internal property value is returned.

skate.define('my-component', {
  props: {
    myProp: {
      get (elem, data) {
        return `prefix_${data.internalValue}`;
      }
    }
  }
});

The parameters passed to the function are:

  • elem - the component element
  • data - an object containing information about the property
    • name - the property name
    • internalValue - the current internal value of the property
initial

The initial value the property should have. This is different from default in the sense that it is only ever invoked once to set the initial value. If this is not specified, then default is used in its place.

skate.define('my-component', {
  props: {
    myProp: {
      initial: 'initial value'
    }
  }
});

It can also be a function that returns the initial value:

skate.define('my-component', {
  props: {
    myProp: {
      initial (elem, data) {
        return 'initial value';
      }
    }
  }
});

The parameters passed to the function are:

  • elem - the component element
  • data - an object containing information about the property
    • name - the property name

Unlike default, initial will be refleted back to the attribute if linked.

serialize

A function that converts the linked property value to the linked attribute value.

skate.define('my-component', {
  props: {
    myProp: {
      serialize (value) {
        return value.join(',');
      }
    }
  }
});

The parameters passed to the function are:

  • value - the attribute value that needs to be coerced to the property value.
set

A function that is called whenever the property is set. This is also called when the property is first initialised.

skate.define('my-component', {
  props: {
    myProp: {
      set (elem, data) {
        // do something
      }
    }
  }
});

The parameters passed to the function are:

  • elem - the component element
  • data - an object containing information about the property
    • name - the property name
    • newValue - the new property value
    • oldValue - the old property value.

When the property is initialised, oldValue will always be undefined and newValue will correspond to the initial value. If the property is set to null or undefined, the value is normalised to be undefined for consistency.

An important thing to note is that native property setters are not invoked if you use the delete keyword. For that reason, Skate property setters are also not able to be invoked, so keep this in mind when using your components.

created

Function that is called when the element is created. This gets called during element construction. We don't use constructor here because Skate does a lot of automation in it and thus offers this as a way to hook into that part of the lifecycle. It is the first lifecycle callback that is called and is called after the prototype is set up.

skate.define('my-component', {
  created (elem) {}
});

The only argument passed to created is component element. In this case that is <my-component>.

updated

Called before render() after props are updated. If it returns falsy, render() is not called. If it returns truthy, render() is called.

skate.define('x-component', {
  updated(elem, prevProps) {
    // The previous props will not be defined if it is the initial render.
    if (!prevProps) {
      return true;
    }

    // The previous props will always contain all of the keys.
    for (let name in prevProps) {
      if (prevProps[name] !== elem[name]) {
        return true;
      }
    }
  }
});

The default implementation does what is described in the example above:

  • If it is the initial update, always call render()
  • If any of the properties have changed according to a strict equality comparison, always call render()
  • In any other scenario, don't render

This generally covers 99% of the use-cases and vastly improves performance over just returning true by default. A good rule of thumb is to always reassign your props. For example, if you have a component that has a string prop and an array prop:

const Elem = skate.define('x-component', {
  props: {
    str: skate.prop.string(),
    arr: skate.prop.array(),
  },
  render() {

  },
});

const elem = new Elem();

// Re-renders:
elem.str = 'updated';

// Will not re-render:
elem.arr.push('something');

// Will re-render:
elem.arr = elem.arr.concat('something');

It is not called if the element is not in the document for the same reasons as render().

If you set properties within updated(), they will not cause it to be called more than once.

The code you write in here is performance critical.

Other use-cases

Generally you'll probably supply a render() function for most of your components. If you require special checks for your props, you can override it:

skate.define('my-component', class extends skate.Component {
  static updated(elem, prev) {
    // You can reuse the original check if you want as part of your new check.
    // You could also call it directly if not extending: skate.Component().
    return super.updated(elem, prev) && myCustomCheck(elem, prev);
  }
});

If you don't have a render() function, sometimes it's still useful to respond to property updates:

skate.define('my-component', {
  props: {
    name: skate.prop.string()
  },
  updated(elem, prev) {
    if (prev.name !== next.name) {
      skate.emit(elem, 'name-changed', { detail: prev });
    }
  }
});

render

Function that is called to render the element. This is called when the element is first created and on subsequent prop updates if the property updated() callback returns true.

skate.define('my-component', {
  render(elem) {
    return skate.h('p', `My name is ${elem.tagName}.`);
  }
});

You may also return an array which negates the need to put a wrapper around several elements:

skate.define('my-component', {
  render(elem) {
    return [
      skate.h('paragraph 1'),
      skate.h('paragraph 2'),
    ];
  }
});

The above isn't restricted to the vdom.builder() API, either; it works with all forms of declaring your virtual DOM.

It is not called if the element is not in the document. It will be called just before attached so that it renders as early as possible, but only if necessary.

Updating props from within render(), while discouraged, will not trigger another render.

Returning the result of your vdom calls is only required when you're using vdom.builder() or h. It is not required when using the (now deprecated) vdom.element() and vdom.text() calls, or if you're using Incremental DOM directly.

rendered

Called after the component has rendered (i.e. called render()). If you need to do any DOM manipulation that can't be done in refs, you can do it here. This is not called if updated() prevents rendering, or render() is not defined.

attached

Function that is called after the element has been inserted to the document. This corresponds to the native attachedCallback. This can be called several times, for example, if you were to remove the element and re-insert it.

skate.define('my-component', {
  attached (elem) {}
});

The only argument passed to attached is component element. In this case that is <my-component>.

detached

Function that is called after the element has been removed from the document. This corresponds to the native detachedCallback. This can be called several times, for example, if you were to remove the element, re-attach it and the remove it again.

skate.define('my-component', {
  detached (elem) {}
});

The only argument passed to detached is component element. In this case that is <my-component>.

attributeChanged

Function that is called whenever an attribute is added, updated or removed. This corresponds to the native attributeChangedCallback. Generally, you'll probably end up using props that have linked attributes instead of this callback, but there are still use cases where this could come in handy.

skate.define('my-component', {
  attributeChanged (elem, data) {
    if (data.oldValue === undefined) {
      // created
    } else if (data.newValue === undefined) {
      // removed
    } else {
      // updated
    }
  }
});

The arguments passed to the attributeChanged() callback differ from the native attributeChangedCallback() to provide consistency and predictability with the rest of the Skate API:

  • elem is the component element
  • data is an object containing attribute name, newValue and oldValue. If newValue and oldValue are empty, the values are undefined.

observedAttributes

This behaves exactly like described in the spec.

For example, the following:

skate.define('my-component', {
  observedAttributes: ['some-attribute'],
  attributeChanged () {}
});

Could similarly be written as:

skate.define('my-component', {
  props: {
    someAttribute: { attribute: true }
  }
});

The differences being that as a result of defining it as a property, it is now linked to an attribute and will cause a re-render. If you only want to observe attribute changes within the attributeChanged() callback, then that is totally valid.

Observing attributes directly using observedAttributes doesn't make sense without specifying an attributeChanged() callback.

emit (elem, eventName, eventOptions = {})

Emits an Event on elem that composed, bubbles and is cancelable by default, or a 'CustomEvent' in older browsers such as IE. This is useful for use in components that are children of a parent component and need to communicate changes to the parent.

skate.define('x-tabs', {
  render(elem) {
    return skate.h('x-tab', { onSelect: () => {} });
  }
});

skate.define('x-tab', {
  render(elem) {
    return skate.h('a', { onClick: () => emit(elem, 'select') });
  }
});

It's preferable not to reach up the DOM hierarchy because that couples your logic to a specific DOM structure that the child has no control over. To decouple this so that your child can be used anywhere, simply trigger an event.

The return value of emit() is the same as dispatchEvent().

Preventing Bubbling or Canceling

If you don't want the event to bubble, or you don't want it to be cancelable, then you can specify those options in the eventOptions argument.

skate.emit(elem, 'event', {
  composed: false,
  bubbles: false,
  cancelable: false
});

Passing Data

You can pass data when emitting the event with the detail option in the eventOptions argument.

skate.emit(elem, 'event', {
  detail: {
    data: 'my-data'
  }
});

link (elem, propSpec)

The link() function returns a function that you can bind as an event listener. The handler will take the event and propagate the changes back to the host element. This essentially allows for 2-way data-binding, but is safer as the propagation of the user input value back to the component element will trigger a re-render, ensuring all dependent UI is up to date.

skate.define('my-input', function () {
  props: {
    value: { attribute: true }
  },
  render (elem) {
    return skate.h('input', { onChange: skate.link(elem), type: 'text' });
  }
});

By default the propSpec defaults to e.currentTarget.getAttribute('name') or "value" which is why it wasn't specified in the example above. In the example above, it would set value on the component. If you were to give your input a name, it would use the name from the event currentTarget as the name that should be set. For example if you changed your input to read:

skate.h('input', { name: 'someValue', onChange: skate.link(elem), type: 'text' });

Then instead of setting value on the component, it would set someValue.

You may explicitly set the property you would like to set by specifying a second argument to link():

skate.link(elem, 'someValue')

The above link would set someValue on the component.

You can also use dot-notation to reach into objects. If you do this, the top-most object will trigger a re-render of the component.

skate.link(elem, 'obj.someValue')

In the above example, the obj property would trigger an update even though only the someValue sub-property was changed. This is so you don't have to worry about re-rendering.

You can even take this a step further and specify a sub-object to modify using the name of the currentTarget (or value, of course) if propSpec ends with a .. For example:

skate.h('input', { name: 'someValue', onChange: skate.link(elem, 'obj.'), type: 'text' });

The above example would set obj.someValue because the name of the input was someValue. This doesn't look much different from the previous example, but this allows you to create a single link handler for use with multiple inputs:

const linkage = skate.link(elem, 'obj.');
skate.h('input', { name: 'someValue1', onChange: linkage, type: 'text' });
skate.h('input', { name: 'someValue2', onChange: linkage, type: 'checkbox' });
skate.h('input', { name: 'someValue3', onChange: linkage, type: 'radio' });
skate.h('select', { name: 'someValue4', onChange: linkage },
  skate.h('option', { value: 2 }, 'Option 2');
  skate.h('option', { value: 1 }, 'Option 1');
);

The above linkage would set:

  • obj.someValue1
  • obj.someValue2
  • obj.someValue3
  • obj.someValue4

prop

Skate has some built-in property definitions to help you with defining consistent property behaviour within your components. All built-in properties are functions that return a property definition.

skate.prop.boolean();

You are able to pass options to these properties to override built-in behaviour, or to define extra options that would normally be supported by a Skate property definition.

You can easily define a new property by calling skate.prop.create() as a function and passing it the default options for the property. All built-in properties are created using this method.

const myNewProp = skate.prop.create({ ... });
myNewProp({ ... });

Built-in properties are accessed on the skate.prop namespace:

skate.prop.boolean({
  coerce () {
    // coerce it differently than the default way
  },
  set () {
    // do something when set
  }
});

Generally built-in properties will only return a definition containing coerce, deserialize and serialize options. They may also define a deafult, such as with the boolean property.

*Empty values are defined as null or undefined. All empty values, if the property accepts them, are normalised to undefined.

Properties are only linked to attributes if the attribute option is set. Each built-in property, if possible, will supply a deserialize and serialize option but will not be linked by default.

array

The array property ensures that any value passed to the property is an array. Serialisation and deserialisation happen using JSON.stringify() and JSON.parse(), respectively, if the property is linked to an attribute.

boolean

The boolean property allows you to define a property that should always have a boolean value to mirror the behaviour of boolean properties / attributes in HTML. By default it is false. If an empty value is passed, then the value is false. If a boolean property is linked to an attribute, the attribute will have no value and its presence indicates whether or not it is true (present) or false (absent).

number

Ensures the value is a Number and is correctly linked to an attribute. Numeric string values such as '10' will be converted to a Number. Non-numeric string values will be converted to undefined. The value will default to 0 if an empty value is passed.

string

Ensures the value is always a String and is correctly linked to an attribute. Empty values are not coerced to strings.

props (elem[, props])

The props function is a getter or setter depending on if you specify the second argument. If you do not provide props, then the current state of the component is returned. If you pass props, then the current state of the component is set. When you set state, the component will re-render synchronously only if it needs to be re-rendered.

Component state is derived from the declared properties. It will only ever return properties that are defined in the props object. However, when you set state, whatever state you specify will be set even if they're not declared in props.

import { define, props } from 'skatejs';

const Elem = define('my-element', {
  props: {
    prop1: null
  }
});
const elem = new Elem();

// Set any property you want.
props(elem, {
  prop1: 'value 1',
  prop2: 'value 2'
});

// Only returns props you've defined on your component.
// { prop1: 'value 1' }
props(elem);

ready (element, callback)

The skate.ready() function allows you to define a callback that is fired when the specified element is has been upgraded. This is useful when you want to ensure an element has been upgraded before doing anything with it. For more information regarding why an element may not be upgraded right away, read the following section.

Background

If you put your component definitions before your components in the DOM loading component-a before component-b:

<script src="component-a.js"></script>
<script src="component-b.js"></script>
<component-a>
  <component-b></component-b>
</component-a>

The initialisation order will be:

  1. component-a
  2. component-b

If you flip that around so that component-b is loaded before component-a, the order is the same. This is because the browser will initialise elements with their corresponding definitions as it descends the DOM tree.

However, if you put your component definitions at the bottom of the page, it gets really hairy. For example:

<component-a>
  <component-b></component-b>
</component-a>
<script src="component-a.js"></script>
<script src="component-b.js"></script>

In this example, we are loading component-a before component-b and the same order will apply. However, if you flip that around so that component-b is loaded before component-a, then component-b will be initialised first. This is because when a definition is registered via window.customElements.define(), it will look for elements to upgrade immediately.

symbols

Symbols allow you to access object information which would be otherwise inaccessible.

name

The name symbol can be used to retrieve the tag name of the component from the constructor. This will be the tag name the component was registered with. If the component has been re-registered with a unique name (see Multiple Component Names and Hot Module Reloading (a.k.a. Webpack HMR)) then this will be the unique name.

const MyComponent1 = skate.define('my-component', {});

// my-component
console.log(MyComponent1[skate.symbols.name]);

// If re-registering in HMR...
const MyComponent2 = skate.define('my-component', {});

// my-component-1
console.log(MyComponent2[skate.symbols.name]);

shadowRoot

When a component renders for the first time, it creates a new shadow root - if it can - and stores this shadow root on the element using this symbol. If a shadow root cannot be created, this returns the element itself.

skate.define('my-component', {
  render() {
    return skate.h('p', 'test');
  },
  ready(elem) {
    // #shadow-root
    //   <p>test</p>
    elem[skate.symbols.shadowRoot];
  }
});

h

The h export is the result of a call to vdom.builder() that allows you to write Hyperscript. This also adds first-class JSX support so you can just set the JSX pragma to h and carry on building stuff.

Hyperscript

You can use the h export to write Hyperscript:

skate.define('my-component', {
  render() {
    return skate.h('p', { style: { fontWeight: 'bold' } }, 'Hello!');
  }
});

Currently id / class selectors are not supported, but will be soon.

JSX

The h export also allows you to write JSX out of the box. All you have to do is include the standard babel transform.

Setting pragma

It's preferred that you set the JSX pragma to h (or skate.h if you're using globals), if possible, so that you don't confuse anyone by using the default React.createElement() in a non-React app.

// .babelrc
{
  "plugins": [
    ["transform-react-jsx", {
      "pragma": "h"
    }]
  ]
}

// my-component.js
import { define, h } from 'skatejs';
define('my-component', {
  render() {
    return <p>Hello!</p>;
  }
});
Default React.createElement()

If you don't have control over the pragma, you can get around it by defining the React.createElement() interface with the h export.

import { define, h } from 'skatejs';
const React = { createElement: h };
define('my-component', {
  render() {
    return <p>Hello!</p>;
  }
});
Other ways to use JSX

You can also enable JSX support in a couple of other ways, though they require slightly more work:

// Incremental DOM needs to be available globally, so you'll have to do:
IncrementalDOM = skate.vdom;

// For the plugin you need to configure the `prefix` option to
// point to Skate's `vdom` or you'll need to do something like this
// so the functions are in scope.
const { elementOpen, elementOpenStart, elementVoid } = skate.vdom;

skate.define('my-element', {
  props: {
    title: skate.prop.string()
  },
  render (elem) {
    return (
      <div>
        <h1>{elem.title}</h1>
        <slot name="description" />
        <article>
          <slot />
        </article>
      </div>
    );
  }
});

vdom

Skate includes several helpers for creating virtual elements with Incremental DOM.

vdom.builder ()

Calling vdom.builder() without any arguments returns a function that you can call in render() to create elements. This is how the h export is created.

const h = vdom.builder();
define('my-component', {
  render() {
    return h('div', { id: 'test', }, h('p', 'test'));
  }
});

vdom.builder (...elements)

When vdom.builder() is called with arguments, it returns an array of functions that create elements corresponding to the arguments that you've passed in. This makes creating a DSL very simple:

const [ div, p ] = skate.vdom.builder('div', 'p');
define('my-component', {
  render() {
    return div({ id: 'mydiv' }, p('test'));
  }
});

[DEPRECATED] vdom.element (elementName, attributesOrChildren, ...children)

This has been deprecated in favour of using the vdom.builder() API, or the h export directly.

The elementName argument is the name of the element you want to create. This can be a string or a function. If it's a function, it is treated as a component constructor or function helper.

The attributesOrChildren argument is either an object, a function that will render the children for this element or a string if you only want to render a text node as the children.

The rest of the arguemnts are functions that render out children.

skate.vdom.element('select', { name: 'my-select' }, function () {
  skate.vdom.element('option', { value: 'myval' }, 'My Value');
});

[DEPRECATED] vdom.text (text)

This has been deprecated in favour of just passing text to an element.

The text() function is exported directly from Incremental DOM and you could use that if you wanted to instead of specifying text as a string to a parent node:

skate.vdom.element('option', { name: 'my-select' }, function () {
  skate.vdom.element('option', { value: 'myval' }, function () {
    skate.vdom.text('My Value');
  });
});

This is very useful if you need to render text with other elements as siblings, or do complex conditional rendering. It's also useful when your custom element may only need to render text nodes to its shadow root.

Incremental DOM

Skate uses Incremental DOM underneath the hood because not only is it performant and memory-efficient, it also acts as a backend to any templating language that can compile down to it. It's less limiting as a transpile target than other virtual DOM implementations so you can use languages other than JSX with it.

We wrap Incremental DOM to add functionality on top of it that we feel is essential to productively building web components:

  • We set properties instead of attributes wherever possible
  • You can pass component constructors and stateless functions as element names directly to the Incremental DOM functions, just like you can in JSX
  • We handle special properties such as class, key, ref, skip and statics
Component constructor

If you pass a component constructor instead of an string as the element name, the name of the component will be used. This means that instead of using hard-coded custom element names, you can import your constructor and pass that instead:

const MyElement = skate.define('my-element');

// Renders <my-element />
skate.h(MyElement);

This is provided at the Incremental DOM level of Skate, so you can also do:

skate.vdom.elementOpen(MyElement);

This is very helpful in JSX:

<MyElement />

However, since this is provided in the Incremental DOM functions that Skate exports, it means that you can do this in any templating language that supports it.

Function helper

Function helpers are passed in the same way as a component constructor but are handled differently. They provide a way for you to write pure, stateless, functions that will render virtual elements in place of the element that you've passed the function to. Essentially they're stateless, private web components.

const MyElement = () => skate.h('div', 'Hello, World!');

// Renders <div>Hello, World!</div>
skate.h(MyElement);

You can customise the output using properties:

const MyElement = props => skate.h('div', `Hello, ${props.name}!`);

// Renders <div>Hello, Bob!</div>
skate.h(MyElement, { name: 'Bob' });

Or you could use children:

const MyElement = (props, chren) => skate.h('div', 'Hello, ', chren, '!');

// Renders <div>Hello, Mary!</div>
skate.h(MyElement, 'Mary');

As with the component constructor, you can also use this in JSX or any other templating language that supports passing functions as tag names:

const MyElement = (props, chren) => <div>Hello, {chren}!</div>;

// Renders <div>Hello, Mary!</div>
<MyElement>Mary</MyElement>
Special Attributes

Skate adds some opinionated behaviour to Incremental DOM.

class

The recommended way to specify a list of classes on an element is by simply specifying the class attribute as you'd normally do in HTML. It's not necessary to specify className, though you can if you really want to.

key

This gives the virtual element a key that Incremental DOM uses to keep track of it for more efficient patches when dealing with arrays of items.

skate.h('ul',
  skate.h('li', { key: 0 });
  skate.h('li', { key: 1 });
);
on*

Any attribute beginning with on followed by an uppercase character or dash, will be bound to the event matching the part found after on.

const onClick = console.log;
skate.h('button', { onClick });
skate.h('button', { 'on-click': onClick });

Additionally, events that exist as properties on DOM elements can also be used:

skate.h('button', { onclick: onClick });

if you need to bind listeners directly to your host element, you should do this in one of your lifecycle callbacks:

skate.define('x-element', {
  created(elem) {
    elem.addEventListener('change', elem.handleChange);
  },
  prototype: {
    handleChange(e) {
      // `this` is the element.
      // The event is passed as the only argument.
    }
  }
});
ref

A callback that is called when the attribute is set on the corresponding element. The only argument is the element that ref is bound to.

const ref = button => button.addEventListener('click', console.log);
skate.h('button', { ref });

Refs are only called on the element when the value of ref changes. This means they get called on the initial set, and subsequent sets if the reference to the value changes.

For example, if you define a function outside of render(), it will only be called when the element is rendered for the first time:

const ref = console.log;
skate.define('my-element', {
  render() {
    return skate.h('div', { ref });
  }
});

However, if you define the ref function within render(), it will be a new reference every time, and thus be called every time:

skate.define('my-element', {
  render() {
    const ref = console.log;
    return skate.h('div', { ref });
  }
});

It's important to understand that this only gets called on the element when either the ref is set up or changed, not when the element is removed from the tree. This is because we discourage saving the value of ref. If you need it to only be called when the element is set up, put the ref outside of render(). If you absolutely need to save the value, try using a WeakMap.

skip

This is helpful when integrating with 3rd-party libraries that may mutate the DOM.

skate.h('div', { ref: e => (e.innerHTML = '<p>oh no you didn\'t</p>'), skip: true });
statics

This is an array that tells Incremental DOM which attributes should be considered static.

skate.h('div', { statics: ['attr1', 'prop2'] });
Boolean Attributes

If you specify false as any attribute value, the attribute will not be added, it will simply be ignored.

Component Lifecycle

The component lifecycle consists of several paths in the following order starting from when the element is first created.

  1. props are defined and set to initial values
  2. created is invoked
  3. attached is invoked when added to the document (or if already in the document)
  4. updated is always invoked before render() when properties have changed
  5. render is invoked to render an HTML structure to the component if it is not prevented by updated()
  6. rendered is always invoked after render(), if it is not prevented by updated()
  7. detached is invoked when removed from the document
  8. attributeChanged is invoked whenever an attribute is changed

Customised built-in elements

The spec mentions this as a way to extend built-in elements. Currently, how this is exposed to the user is still under contention. Skate doesn't need do anything to support this underneath the hood, but be aware of this when building components.

VS other libraries

We hold no ill thoughts against other libraries and we are just trying to articulate why one would choose Skate over another. If any information here is inaccurate, please feel free to raise an issue to discuss how we can make it accurate.

VS WebComponentsJS

webcomponentsjs

WebComponentsJS is a suite of polyfills. If there is native browser support, they're not needed. Skate is a superset of these specifications, thus is a value-add on top of them, if they're needed.

VS Polymer

Polymer uses webcomponentsjs and adds an abstraction on top of it. In their high-level design, Skate and Polymer are very similar in that they're built on top of emerging standards. However, fundamentally, Skate and Polymer are very different.

  • Skate uses a functional programming model for rendering in which you can use any templating language you want that compiles down to Incremental DOM. It calls render() when something changes and then tells Incremental DOM to diff and patch what's different between the two states. With Polymer, you use their custom template syntax that creates links between properties and mutations happen to the DOM directly.
  • Skate only has a single option for its usage, making it simpler to grok what you're getting. Polymer has three different builds, most of which Skate is smaller than. The following comparisons are using non-gzipped, minified versions.
    • polymer-micro.html 17k vs 11k
    • polymer-mini.html 54k vs 11k
    • polymer.html 124k vs 11k
  • Polymer uses HTML Imports to build their codebase. This can be obtuse if you're used to using JavaScript module formats, especially since HTML Imports are currently very contentious and Google are the only ones who are pushing for it.
  • Skate supports JSPM, NPM and more. Polymer currently only supports Bower.

VS X-Tags

Skate is very close to X-Tags in terms of API shape, however, it is very different in the way it is applied and shares a lot of the same differences with X-Tags as it does with Polymer.

  • Skate uses a functional programming model for rendering in which you can use any templating language you want that compiles down to Incremental DOM. X-Tags is not very opinionated about rendering or templating. You define a string of HTML and it will use that as the component's content.
  • Skate's property API is thorough and extensible. We provide implementations or commonly used property patterns and give you an API to easily write your own reusable properties.
  • Skate is about the same size and scales well for building large, complex user interfaces.
  • There's no mutating your component's DOM from property accessors which can become unwieldy.

VS React

React has definitely had an influence on Skate. That said, they're completely different beasts, only sharing a functional rendering pipeline and some aspects of the API.

  • React is 10x the size of Skate.
  • In the performance tests you can see a Skate component is several times faster than a similarly written React component.
  • Skate is written on top of W3C standards. The React authors have been very vocal about this. However, the response to that issue is incorrect. Web Components by nature are declarative: it's just HTML. Web Components also completely solve the integration problems between libraries and frameworks due to the nature of how Custom Elements and Shadow DOM work: Custom Elements provide a declarative API, Shadow DOM hides the implementation details. When integrating with frameworks, you're just writing HTML. In terms of the problems with imperative APIs, it's not the fault of Web Components that force a user to call a method, it's the fault of the design of the Web Component. There's nothing stopping a Web Component from being completely declarative, especially if it's written in Skate. More information about web component design.
  • We have plans to support server-side rendering.

Preventing FOUC

An element may not be initialised right away if your definitions are loaded after the document is parsed. In native custom elements, you can use the :defined pseudo-class to select all elements that have been upgraded, thus allowing you to do :not(:defined) to invert that. Since that only works in native, Skate adds a defined attribute so that you have a cross-browser way of dealing with FOUC and jank.

my-element {
  opacity: 1;
  transition: opacity .3s ease;
}
my-element:not([defined]) {
  opacity: 0;
}

Designing Web Components

A web component's public API should be available both imperatively (via JavaScript) and declaratively (via HTML). You should be able to do everything in one, that you can do in the other within reason.

Imperative

You should always try and make the constructor available whether it's exported from an ES2015 module or a global:

export default skate.define('my-component', {});

Declarative

By declaring a Skate component, you are automatically making your element available to be used as HTML. For example, if you were to create a custom element for a video player:

skate.define('x-video', {});

You could now just write:

<x-video></x-video>

Instead of providing just imperative methods - such as play() for the native <video> element - you should try to provide attributes that offer the same functionality. For example, if you had a player component, you could offer a playing boolean attribute, so that it starts playing when it's put on the page.

<x-video playing></x-video>

To pause / stop the player, you remove the attribute.

<x-video></x-video>

If you're using something like React or Skate to render this component, you don't have to write any imperative code to remove that attribute as the virtual DOM implementations will do that for you.

The nice part about thinking this way is that you get both a declarative and imperative API for free. You can think about this in simpler terms by designing your API around attributes rather than methods.

Naming Collisions

You may write a component that you change in a backward incompatible way. In order for your users to upgrade, they'd have to do so all at once instead of incrementally if you haven't given the new one a different name. You could rename your component so it can co-exist with the old one, or you can create a function that allows your users to define a name for your component:

export default function (name) {
  return skate.define(name, {
    render (elem) {
      return skate.h('div', `This element has been called: ${elem.tagName}.`);
    }
  });
}

If you define the same component more than once, Skate will choose a unique name for subsequent registrations after the first. This generally is something you'd want to avoid, but it is very helpful during development. For more information see the HMR docs.

Compatible with multiple versions of itself

Skate is designed so that you can have multiple versions of it on the same page. This is so that if you have several components, your upgrades and releases aren't coupled. If you have a UI library based on Skate and those consuming your library also have Skate, your versions aren't coupled.

Properties and Attributes

Properties and attributes should represent as much of your public API as possible as this will ensure that no matter which way your component is created, its API remains as consistent as the constraints of HTML will allow. You can do this by ensuring your properties have corresponding attributes:

skate.define('my-component', {
  props: {
    // Links the `name` property to the `name` attribute.
    name: { attribute: true }
  }
});

Sometimes this may not be viable, for example when passing complex data types to attributes. In this scenario, you can try and serialize / deserialize to / from attributes. For example, if you wanted to take a comma-separated list in an attribute and have the property take an array, but still have them linked, you could do something like:

skate.define('my-component', {
  props: {
    values: {
      attribute: true,
      deserialize (val) {
        return val.split(',');
      },
      serialize (val) {
        return val.join(',');
      }
    }
  }
});

Private Members

Skate doesn't have any opinions on how you store or use private methods and properties on your elements. Classically one would normally use scoped functions or underscores to indicate privacy:

function scoped(elem) {}

skate.define('x-element', {
  created(elem) {
    scoped(elem);
    elem._privateButNotReally();
  },
  prototype: {
    _privateButNotReally() {}
  }
});

However, if you're using ES2015 you can use symbols. Using this pattern, your members are completely private and only available if you have access to the symbol:

const sym = Symbol();

skate.define('x-element', {
  created(elem) {
    elem[sym]();
  },
  prototype: {
    [sym]() {}
  }
});

Private Data

A slightly different use-case than using private members would be storing private data. As with members, you can use scoped variables or underscores. However, scoped variables generally aren't specific to an element instance and underscores are only a privacy guideline; anyone can still access the data.

The best way to do this depends on your needs. Generally a WeakMap is a good choice as it will hold weak references to the key:

const map = new WeakMap();

skate.define('x-element', {
  created(elem) {
    map.set(elem, 'some data');
  },
  render(elem) {
    // Renders: "<div>some data</div>"
    return skate.h('div', map.get(elem));
  }
});

You can also use symbols on your element just like we did above with standard methods and properties, if that suits your workflow better.

React Integration

There is a React integration library that allows you to write web components - written with any true web component library - and convert them to react components using a single function. Once converted, it can be used in React just like you would use a normal React component.

Multiple Component Names and Hot Module Reloading (a.k.a. Webpack HMR)

Skate is designed to work with hot-module reloading out of the box:

  • It will always use the canonical name on the initial registration
  • Subsequent registrations will register using the canonical name with a number suffix to identify how many times it's been registered

Skate cannot refresh the component definition as there is no way to reregister a component using the web component APIs.

While this makes the name non-deterministic, you can still get the name from the constructor if you need to using the name symbol.

Form Behaviour and the Shadow DOM

Submission

When you encapsulate a form, button or both inside a shadow root, forms will not be submitted when the submit button is clicked. This is because the shadow boundary prevents each shadow root from communicating with each other. Fortunately this isn't very difficult to wire up.

Let's say we have a custom form and custom button (to reproduce, only one of them would need to be contained in a shadow root):

<x-form>
  <x-button type="submit">Submit</x-button>
</x-form>

The definitions look like the following:

skate.define('x-form', {
  render() {
    return (
      <form>
        <slot />
      </form>
    );
  }
});

skate.define('x-button', {
  render() {
    return (
      <button>
        <slot />
      </button>
    );
  }
});

To wire this up we listen for clicks coming from something that has a type of "submit". You can also check for type, but for the sake of simplicity, we'll just check for type:

function onclick (e) {
  if (e.target.getAttribute('type') === 'submit') {
    // do something submitty
  }
}

Now all you need to do is put that on the <form> inside of <x-form>:

<form { onclick }>

You cane take this a step further and emit a submit event on the form and call submit() on it if the event wasn't canceled:

function onclick (e) {
  if (e.target.getAttribute('type') === 'submit') {
    if (skate.emit(e.currentTarget, 'submit')) {
      e.currentTarget.submit();
    }
  }
}

The full example looks like:

function onclick (e) {
  if (e.target.getAttribute('type') === 'submit') {
    if (skate.emit(e.currentTarget, 'submit')) {
      e.currentTarget.submit();
    }
  }
}

skate.define('x-form', {
  render() {
    return (
      <form>
        <slot />
      </form>
    );
  }
});

skate.define('x-button', {
  render() {
    return (
      <button>
        <slot />
      </button>
    );
  }
});

Form Data

The idea that built-in form elements don't publish their form-data when inside a shadow root is being discussed.

In order to handle this, your custom form would need to gather all the form data associated with it and submit it along with its request.

Stateless Components

If you write a component that manages its props internally, this is called a "smart component"; very much the same as in React. There's many ways you can write components and separate them out in to smart / dumb components. You can then compose the dumb one in the smart one, reusing code while separating concerns.

However, there is one way where you can write a smart component and it can be made "dumb" as part of its API by emitting an event that allows any listeners to optionally prevent it from updating, or to simply update it with new props that it should render with.

skate.define('x-component', {
  updated(elem, prev) {
    // Notify any listeners that the component updated. At this point the
    // listener can update the component's props without fear that this will
    // cause recursion - because it's prevented internally - and it will
    // proceed past this point with the updated props.
    const canRender = skate.emit(elem, 'updated', { detail: prev });

    // This can be custom, or just reuse the default implementation. Since we
    // emitted the event and listeners had a chance to update the component,
    // this will get called with the updated state.
    return canRender && skate.Component.updated(elem, prev);
  }
});

The previous example emits an event that bubbles and is cancelable. If it is canceled, then the component does not render. If the listening component updates the component's props in response to the event, the component will render with the updated props if it passes the default updated() check.

Styling Components

In order to style your components, you should assume Shadow DOM encapsulation. The best-practice here is to simply put styles into a <style> block:

skate.define('x-component', {
  render() {
    return [
      skate.h('style', '.my-class { display: block; }'),
      skate.h('div', { class: 'my-class' }),
    ];
  }
});

If you want to ensure your styles are encapsulated even if using a polyfill, use CSS Modules, they are absolutely amazing!

Using ES2015 classes

When using ES2015 classes, there are slight differences to how you'd specify things when using an object literal. For example, lifecycle callbacks become static:

skate.define('x-component', class extends skate.Component {
  static created(elem) {}
  static attached(elem) {}
  static detached(elem) {}
});

This also means properties like props must be specified as a getter:

skate.define('x-component', class extends skate.Component {
  static get props () {
    return {
      myProp: { attribute: true }
    };
  }
});

Or a class property:

skate.define('x-component', class extends skate.Component {
  static props = {
    myProp: { attribute: true }
  }
});

Adding instance methods to the element prototype is done by simply specifying them as non-static items:

skate.define('x-component', class extends skate.Component {
  myProp1 = 'some value'
  get myProp2 () { return 'another value'; }
  myMethod() {}
});

skatejs's People

Contributors

adevnadia avatar antitoxic avatar bengummer avatar bitdeli-chef avatar bqx avatar bradleyayers avatar dmi3y avatar gitter-badger avatar greenkeeperio-bot avatar jasonberry avatar joscha avatar jpnelson avatar lukebatchelor avatar matiasatlassian avatar nicvenegas avatar npmcdn-to-unpkg-bot avatar nvenegas avatar ruudud avatar stevelacy avatar stevenschobert avatar treshugart avatar wangersnmash avatar zenorocha avatar zephor avatar zpetch avatar zsakare avatar zzarcon avatar

Watchers

avatar avatar

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.