/lighterhtml

The hyperHTML strength & experience without its complexity 🎉

Primary LanguageJavaScriptISC LicenseISC

lighterhtml

Social Media Photo by Kristine Weilert on Unsplash

Build Status WebReflection status License: ISC Greenkeeper badge Blazing Fast

The hyperHTML strength & experience without its complexity 🎉

V4 Breaking Changes

I am afraid this major was necessary due recent bugs/discoveries that made me rethink some practice and patch.

  • the recently introduced data helper could conflict with some node such as <object>, hence it has been replaced by the .dataset utility. Since element.dataset = object is an invalid operation, the sugar to simplify data- attributes is now never ambiguous and future-proof: <element .dataset=${...} /> it is.
  • all cross browsers normalizations and features detection to make the template literal unique has been removed, as these were causing more problems than these were supposed to solve. If you are targeting IE 11 or older browsers, be sure you use Babel 7 to transpile your production code. If you are using TypeScript, be sure you use Babel 7 to transpile your code, as TS has always been broken with transpiled template literals (and classes, and ...).
  • the good old domdiff that served me well, and it still does, has been replaced by its little udomdiff brother, allowing lighterhtml to weight 1K less than before, still keeping lightning fast performance.

Because of these breaking changes, all libraries around lighterhtml will gradually bump major version too, pointing at this paragraph of the README.

V3 Declarative data and aria attributes.

Since the introduction of .setter=${value} made special cases such as data=${...} and props=${...} redundant, as it's always possible to simply attach any kind of data via .data=${...} or .props=${...}, version 3 enhances the declarative power of the template to HTML translation.

// the aria special case
html`<div aria=${{labelledBy: 'id', role: 'button'}} />`;
//=> <div aria-labelledby="id" role="button"></div>

// the *deprecated* dataset special case
html`<div data=${{key: 'value', otherKey: 'otherValue'}} />`;
//=> <div data-key="value" data-other-key="otherValue"></div>

// the *new* dataset special case
html`<div .dataset=${{key: 'value', otherKey: 'otherValue'}} />`;
//=> <div data-key="value" data-other-key="otherValue"></div>

This means the previous data=${...} behavior should be substituted with .data=${...} and it's now possible to better reflect declarative intents in nodes, simplifying both data-* attributes and aria-* ones.

Please note using data-name=${value}, as well as aria-name=${value} is still handled like any other regular attribute, hence it will work as expected, actually faster when the values don't change frequently, as both aria and data special cases simply loop through the object keys and assign their values to node's attributes.

V2.1 Introducing A New Listener Feature

Until version 2.1, there was no way to define different options for any listener. The el.addEventListener(type, listener, false) was the only kind of operation possible behind the scene.

In v2.1 though, whenever a second option is needed, it is now possible to pass an Array instead, in the form [listener, {once: true}], as example, or [listener, true] to use capture instead of bubbling, and so on and so forth.

Bear in mind, specially for the once case, if the listener is different per each update, like onclick=${[() => stuff(), {once: true}]}, it will be set each time that update happens, so that in this case is better to use always the same listener, either via outer scope callback, or via reference, using useRef and the handleEvent pattern, as example.

If you never needed to add a different second option, there is nothing you should do, everything will work exactly as it did before.

V2 Breaking Changes & Improvements

Breaking

  • dropped the ambiguous ability to produce nodes when no render(...) is invoked. When needed, which is the minority of the cases, you need to explicitly use html.node or svg.node, instead of just html or svg. For every other cases, use render(where, what).
  • the render(where, what) does not need a callback anymore. You can now render(node, html`<p>content</p>`) right away. If a callback is provided, that will still be invoked.
  • removed useHook as it's unnecessary since you can use useRef through html.for(...) or svg.for(...) within any useRef provided by your library of choice (i.e. dom-augmentor)
  • the recently introduced inner.html/svg has been removed, as completely unnatural and error prone (just use html anywhere, it'll work).

Improvements

  • a fundamental core-logic implementation that was trashing any node after one or more collections has been refactored and fixed. The current logic create a stack per each array found down the rendering road, isolating those DOM updates per stack. This means that performance have been improved, and GC operations reduced.
  • html and svg template literals tags, now offer both .for(ref[, id]) and .node, to either retain the same content (keyed render) or create fresh new nodes out of the box as one-off operation (via .node).
  • slightly reduced code size, which is always nice to have, after a refactoring

V1 Changes + New Feature

Removed transform export and made default domtagger customizable via custom export.

import { custom } from 'lighterhtml';

const { html, render } = custom({

  // the domtagger attributes handler
  attribute: callback => (node, name, original) => {
    // return a function that will handle the attribute value
    // the function will receive just the new value
    if (name === 'double')
      return value => {
        node[name] = value + value;
      };
    // the received callback is usable as return fallback
    return callback(node, name, original);
  },

  // the domtagger any-content handler
  any: callback => (node, childNodes) => {
    // return a function that will handle handle all special cases
    // the function will receive just the new *hole* value
    if (node.nodeName === 'CUSTOM') {
      return value => {
        node.appendChild(value);
      };
    }
    // the received callback is usable as return fallback
    return callback(node, childNodes);
  },

  // the domtagger text for text only cases
  text: callback => (node) => {
    // return a function that will handle handle text content cases
    // the function will receive just the new text value
    if (node.nodeName === 'WRAP') {
      return value => {
        node.textContent = `(${value})`;
      };
    }
    // the received callback is usable as return fallback
    return callback(node);
  },

  // optionally you can use the special transform handler too
  // in this case, and in V1, the callback is just the String one
  transform: callback => markup => callback(markup),

  // same goes for convert, with the callback being the one
  // originally used to "convert" the template from Array to HTML
  // see: https://github.com/WebReflection/domtagger/issues/17#issuecomment-526151473
  convert: callback => template => callback(template)
});

faster than hyperHTML

Even if sharing 90% of hyperHTML's code, this utility removed all the not strictly necessary parts from it, including:

  • no Component, no define and/or intents, no connect or disconnect, and no promises (possibly in later on), everything these days can be easily handled by hooks, as example using the dom-augmentor utility
  • html content is never implicit, since all you have to do is to write html before any template when you need it. However, the {html: string} is still accepted for extreme cases.

Removing these parts made the output smaller in size (less than 6K) but it also simplified some underlying logic.

Accordingly, lighterhtml delivers raw domdiff and domtagger performance in an optimized way.

If you don't believe it, check the DBMonster benchmark 😉

simpler than lit-html

In lit-html, the html function tag is worthless, if used without its render.

In lighterhtml though, the html.node or svg.node tag, can be used in the wild to create any, one-off, real DOM, as shown in this pen.

// lighterhtml: import the `html` tag and use it right away
import {html} from '//unpkg.com/lighterhtml?module';

// a one off, safe, runtime list 👍
const list = ['some', '<b>nasty</b>', 'list'];
document.body.appendChild(html.node`
  <ul>${list.map(text => html.node`
    <li>${text}</li>
  `)}
  </ul>
`);

Strawberry on top, when the html or svg tag is used through lighterhtml render, it automatically creates all the keyed performance you'd expect from hyperHTML wires, without needing to manually address any reference: pain point? defeated! 🍾

Available as lighterhtml-plus too!

If you are looking for Custom Elements like events out of the box, lighterhtml-plus is your next stop 👍

How to import lighterhtml/-plus

Following, the usual multi import pattern behind every project of mine:

  • via global lighterhtml CDN utility: <script src="https://unpkg.com/lighterhtml"></script>, and const {render, html, svg} = lighterhtml
  • via ESM CDN module: import {render, html, svg} from 'https://unpkg.com/lighterhtml?module'
  • via ESM bundler: import {render, html, svg} from 'lighterhtml'
  • via CJS module: const {render, html, svg} = require('lighterhtml')

What's the API ? What's in the export ?

The module exports the following:

  • html tag function to create any sort of HTML content when used within a render call. It carries two extra utilities, html.for(ref[, id]), to hard-reference a specific node, and html.node to create one-off dom nodes in the wild without using the render.
  • svg tag function to create any sort of SVG content when used within a render call. It carries two extra utilities, svg.for(ref[, id]), to hard-reference a specific node, and svg.node to create one-off dom nodes in the wild without using the render.
  • render(node, fn|Hole) to pollute a node with whatever is returned from the fn parameters, including html or svg tagged layout
  • Hole class for 3rd parts (internal use)

You can test live a hook example in this Code Pen.

What's different from hyperHTML ?

  • the wired content is not strongly referenced as it is for hyperHTML.wire(ref[, type:id]) unless you explicitly ask for it via html.for(ref[, id]) or svg.for(ref[, id]), where in both cases, the id doesn't need any colon to be unique. This creates content hard wired whenever it's needed.
  • the ref=${object} attribute works same as React, you pass an object via const obj = useRef(null) and you'll have obj.current on any effect. If a callback is passed instead, the callback will receive the node right away, same way React ref does.
  • if the attribute name is aria, as in aria=${object}, aria attributes are applied to the node, including the role one.
  • deprecated: if the attribute name is data, as in data=${object}, the node.dataset gets populated with all values.
  • if the attribute name is .dataset, as in .dataset=${object}, the node.dataset gets populated with all values.
  • intents, hence define, are not implemented. Most tasks can be achieved via hooks.
  • promises are not in neither. You can update asynchronously anything via hooks or via custom element forced updates.
  • the onconnected and ondisconnected special events are available only in lighterhtml-plus. These might come back in the future but right now dom-augmentor replaces these via useEffect(callback, []). Please note the empty array as second argument.
  • an array of functions will be called automatically, like functions are already called when found in the wild
  • the Component can be easily replaced with hooks or automatic keyed renders
  • if a listener is an Array such as [listener, {once: true}], the second entry of the array will be used as option.
const {render, html} = lighterhtml;

// all it takes to have components with lighterhtml
const Comp = name => html`<p>Hello ${name}!</p>`;

// for demo purpose, check in console keyed updates
// meaning you won't see a single change per second
setInterval(
  greetings,
  1000,
  [
    'Arianna',
    'Luca',
    'Isa'
  ]
);

function greetings(users) {
  render(document.body, html`${users.map(Comp)}`);
}

Documentation

Excluding the already mentioned removed parts, everything else within the template literal works as described in hyperHTML documentation.

A basic example

Live on Code Pen.

import {render, html} from '//unpkg.com/lighterhtml?module';

document.body.appendChild(
  // as unkeyed one-off content, right away 🎉
  html.node`<strong>any</strong> one-off content!<div/>`
);

// as automatically rendered wired content 🤯
todo(document.body.lastChild);
function todo(node, items = []) {
  render(node, html`
  <ul>${
  items.map((what, i) => html`
    <li data-i=${i} onclick=${remove}> ${what} </li>`)
  }</ul>
  <button onclick=${add}> add </button>`);
  function add() {
    items.push(prompt('do'));
    todo(node, items);
  }
  function remove(e) {
    items.splice(e.currentTarget.dataset.i, 1);
    todo(node, items);
  }
}

What about Custom Elements ?

You got 'em, just bind render arguments once and update the element content whenever you feel like.

Compatible with the node itself, or its shadow root, either opened or closed.

const {render, html} = lighterhtml;

customElements.define('my-ce', class extends HTMLElement {
  constructor() {
    super();
    this.state = {yup: 0, nope: 0};
    this.render = render.bind(
      null,
      // used as target node
      // it could either be the node itself
      // or its shadow root, even a closed one
      this.attachShadow({mode: 'closed'}),
      // the update callback
      this.render.bind(this)
    );
    // first render
    this.render();
  }
  render() {
    const {yup, nope} = this.state;
    return html`
    Isn't this <strong>awesome</strong>?
    <hr>
    <button data-key=yup onclick=${this}>yup ${yup}</button>
    <button data-key=nope onclick=${this}>nope ${nope}</button>`;
  }
  handleEvent(event) {
    this[`on${event.type}`](event);
  }
  onclick(event) {
    event.preventDefault();
    const {key} = event.currentTarget.dataset;
    this.state[key]++;
    this.render();
  }
});

Should I ditch hyperHTML ?

Born at the beginning of 2017, hyperHTML matured so much that no crucial bugs have appeared for a very long time.

It has also been used in production to deliver HyperHTMLElement components to ~100M users, or to show W3C specifications, so that in case of bugs, hyperHTML will most likely be on the fast lane for bug fixes, and lighterhtml will eventually follow, whenever it's needed.

On top of this, most modules used in lighterhtml are also part of hyperHTML core, and the ./tagger.js file is mostly a copy and paste of the hyperHTML ./objects/Update.js one.

However, as tech and software evolve, I wanted to see if squashing together everything I know about template literals, thanks to hyperHTML development, and everything I've recently learned about hooks, could've been merged together to deliver the easiest way ever to declare any non-virtual DOM view on the Web.

And this is what lighterhtml is about, an attempt to simplify to the extreme the .bind(...) and .wire(...) concept of hyperHTML, through a package that requires pretty much zero knowledge about those internals.

lighterhtml is also relatively new, so that some disabled functionality might come back, or slightly change, but if you like the idea, and you have tested it works for your project, feel free to ditch hyperHTML in favor of lighterhtml, so that you can help maturing this project too.

Should I use micro html instead ?

µhtml is a great way to start playing around with most lighterhtml features. As it's simply a subset, you can eventually switch to lighterhtml later on, whenever you miss, or need, some extra feature.

For a complete comparison of features and libraries around my repositories, please have a look at this gist.