threejs-modern-app

Boilerplate and utils for a fullscreen three.js app

                                                                                                                                assets thanks to Poliigon and Blender

It is inspired from mattdesl's threejs-app, but it was rewritten and simplified using ES6 syntax rather than node, making it easier to read and well commented, so it can be easily customized to fit your needs.

DEMO

Example of production scale project

Yeoman Generator (CLI that generates a blank project without the example code)

Features

• All the three.js boilerplate code is tucked away in a file, the exported WebGLApp is easily configurable from the outside, for example you can enable postprocessing, orbit controls, FPS stats, Detect GPU, a controls-gui and use the save screenshot functionality. It also has built-in support for cannon-es. [Read more]
• A scalable three.js component structure where each component is a class which extends THREE.Group, so you can add any object to it. The class also has update, resize, and pointer hooks. [Read more]
• An asset manager which handles the preloading of .gltf models, images, audios, videos and can be easily extended to support other files. It also automatically uploads a texture to the GPU, loads cube env maps or parses equirectangular projection images. [Read more]
• global window.DEBUG flag which is true when the url contains ?debug as a query parameter. So you can enable debug mode both locally and in production. [Read more]
• glslify to import shaders from node_modules. [Read more]
• GPU tiering info using detect-gpu [Read more]
• Modern and customizable development tools such as webpack, babel, eslint, prettier and browserslist.
• Beautiful console output:

NOTE: brew is required for the build command output

Usage

Once you installed the dependencies running yarn, these are the available commands:

• yarn start starts a server locally
• yarn build builds the project for production, ready to be deployed from the build/ folder

All the build tools logic is in the package.json and webpack.config.js.

WebGLApp

import WebGLApp from './lib/WebGLApp'

const webgl = new WebGLApp({ ...options })

The WebGLApp class contains all the code needed for three.js to run a scene, it is always the same so it makes sense to hide it in a standalone file and don't think about it.

You can see an example configuration here:

https://github.com/marcofugaro/threejs-modern-app/blob/bd303c968c0b0ef56a40046e054371ba4b8c7e54/src/index.js#L14-L45

You can pass the class the options you would pass to the THREE.WebGLRenderer, and also some more options:

The webgl instance will contain all the three.js elements such as webgl.scene, webgl.renderer, webgl.camera or webgl.canvas. It also exposes some useful properties and methods:

webgl.isDragging

Wether or not the user is currently dragging. It is true between the onPointerDown and onPointerUp events.

webgl.cursor

Set this property to change the cursor style of the canvas. For example you can use it to display the pointer cursor on some objects:

onPointerMove(event, { x, y }) {
  // raycast and get the intersecting mesh
  const intersectingMesh = getIntersectingMesh([x, y], this, this.webgl)

  if (intersectingMesh) {
    this.webgl.cursor = 'pointer'
  } else {
    this.webgl.cursor = null
  }
}

webgl.saveScreenshot({ ...options })

Save a screenshot of the application as a png.

webgl.onUpdate((dt, time) => {})

Subscribe to the update requestAnimationFrame without having to create a component. If needed you can later unsubscribe the function with webgl.offUpdate(function).

webgl.onPointerDown((event, { x, y }) => {})

Subscribe a function to the pointerdown event on the canvas without having to create a component. If needed you can later unsubscribe the function with webgl.offPointerDown(function).

webgl.onPointerMove((event, { x, y }) => {})

Subscribe a function to the pointermove event on the canvas without having to create a component. If needed you can later unsubscribe the function with webgl.offPointerMove(function).

webgl.onPointerUp((event, { x, y }) => {})

Subscribe a function to the pointerup event on the canvas without having to create a component. If needed you can later unsubscribe the function with webgl.offPointerUp(function).

Component structure

Rather than writing all of your three.js app in one file instruction after instruction, you can split your app into thhree.js components". This makes it easier to manage the app as it grows. Here is a basic component:

https://github.com/marcofugaro/threejs-modern-app/blob/master/src/scene/Box.js

A three.js component is a class which extends THREE.Group (an alias for THREE.Object3D) and subsequently inherits its properties and methods, such as this.add(someMesh) or this.position or this.rotation. Here is a full list.

After having instantiated the class, you can add it directly to the scene.

// attach it to the scene so you can access it in other components
webgl.scene.birds = new Birds(webgl, { count: 1000 })
webgl.scene.add(webgl.scene.birds)

And in the component, you can use the options like this.

export default class Birds extends THREE.Group {
  constructor(webgl, options = {}) {
    super(options)
    // these can be used also in other methods
    this.webgl = webgl
    this.options = options

    // destructure and default values like you do in React
    const { count = 10 } = this.options

    // ...

The class supports some hooks, which get called once the element is in the scene:

update(dt, time) {}

Called each frame of the animation loop of the application. Gets called by the main requestAnimationFrame.

resize({ width, height, pixelRatio }) {}

Called each time the window has been resized.

onPointerDown(event, { x, y }) {}

Called on the pointerdown event on the canvas.

onPointerMove(event, { x, y }) {}

Called on the pointermove event on the canvas.

onPointerUp(event, { x, y }) {}

Called on the pointerup event on the canvas.

Functional Components

If you don't need any of the previous methods, you can use functional components, which are just plain functions with the objective of making code easier to navigate in.

export function addLights(webgl) {
  const directionalLight = new THREE.DirectionalLight(0xffffff, 0.6)
  directionalLight.position.copy(position)
  webgl.scene.add(directionalLight)

  const ambientLight = new THREE.AmbientLight(0xffffff, 0.5)
  webgl.scene.add(ambientLight)
}

// ...

addLights(webgl)

Asset Manager

The Asset Manager handles the preloading of all the assets needed to run the scene, you use it like this:

https://github.com/marcofugaro/threejs-modern-app/blob/bd303c968c0b0ef56a40046e054371ba4b8c7e54/src/scene/Suzanne.js#L15-L46

https://github.com/marcofugaro/threejs-modern-app/blob/bd303c968c0b0ef56a40046e054371ba4b8c7e54/src/index.js#L56

https://github.com/marcofugaro/threejs-modern-app/blob/bd303c968c0b0ef56a40046e054371ba4b8c7e54/src/scene/Suzanne.js#L54

In detail, first you queue the asset you want to preload in the component where you will use it

import assets from '../lib/AssetManager'

const key = assets.queue({
  url: 'assets/model.gltf',
  type: 'gltf',
})

Then you import the component in the index.js so that code gets executed

import Component from './scene/Component'

And then you start the queued assets loading promise, always in the index.js

assets.load({ renderer: webgl.renderer }).then(() => {
  // assets loaded! we can show the canvas
})

After that, you init the component and use the asset in the component like this

const modelGltf = assets.get(key)

These are all the exposed methods:

assets.queue({ url, type, ...others })

Queue an asset to be downloaded later with assets.load().

Returns a key that later you can use with assets.get().

assets.queueStandardMaterial(maps, options)

Utility to queue multiple maps belonging to the same PBR material. They can later be passed directly to the MeshStandardMaterial.

For example, here is how you load a brick PBR texture:

const bricksKeys = assets.queueStandardMaterial(
  {
    map: `assets/bricks/albedo.jpg`,
    roughnessMap: `assets/bricks/roughness.jpg`,
    metalnessMap: `assets/bricks/metallic.jpg`,
    normalMap: `assets/bricks/normal.jpg`,
    displacementMap: `assets/bricks/height.jpg`,
    aoMap: `assets/bricks/ambientocclusion.jpg`,
  },
  {
    repeat: new THREE.Vector2().setScalar(0.5),
    wrapS: THREE.RepeatWrapping,
    wrapT: THREE.RepeatWrapping,
  }
)

As you can see, you can pass as a second argument any property you want to apply to all textures.

If you're using gamma, the textures with color data will be automatically gamma encoded.

Returns a keys object that later you can use with assets.getStandardMaterial().

assets.load({ renderer })

Load all the assets previously queued.

assets.loadSingle({ url, type, renderer, ...others })

Load a single asset without having to pass through the queue. Useful if you want to lazy-load some assets after the application has started. Usually the assets that are not needed immediately.

Returns a key that later you can use with assets.get().

assets.addProgressListener((progress) => {})

Pass a function that gets called each time an assets finishes downloading. The argument progress goes from 0 to 1, with 1 being every asset queued has been downloaded.

assets.get(key)

Retrieve an asset previously loaded with assets.queue() or assets.loadSingle().

assets.getStandardMaterial(keys)

Retrieve an asset previously queued with assets.queueStandardMaterial().

It returns an object of the loaded textures that can be fed directly into MeshStandardMaterial like this:

const textures = assets.getStandardMaterial(keys)
const material = new THREE.MeshStandardMaterial({ ...textures })

Debug mode

Often you want to show the fps count or debug helpers such as the SpotLightHelper only when you're developing or debugging.

A really manageable way is to have a global window.DEBUG constant which is true only if you append ?debug to your url, for example http://localhost:8080/?debug or even in production like https://example.com/?debug.

This is done here in just one line:

window.DEBUG = window.location.search.includes('debug')

You could also add more global constants by just using more query-string parameters, like this ?debug&fps.

glslify

glslify lets you import shader code directly from node_modules.

For example, if you run through glslify a string you're using in three's onBeforeCompile, you can import glsl-noise like this:

import glsl from 'glslify'

// ...

shader.vertexShader = glsl`
  uniform float time;
  uniform float speed;
  uniform float frequency;
  uniform float amplitude;

  #pragma glslify: noise = require('glsl-noise/simplex/3d')

  // the function which defines the displacement
  float displace(vec3 point) {
    return noise(vec3(point.xy * frequency, time * speed)) * amplitude;
  }

  // ...
`

💡 BONUS TIP: you can have glsl syntax highlighting for inline glsl strings in VSCode with the extension glsl-literal.

glslify is applied also to files with the .frag, .vert or .glsl extensions. They are imported as plain strings:

// pass.vert
varying vec2 vUv;

void main() {
  vUv = uv;
  gl_Position = projectionMatrix * modelViewMatrix * vec4(position, 1.0);
}

// index.js
import passVertexShader from '../shaders/pass.vert'

// ...

const material = new THREE.ShaderMaterial({
  // it's a string
  vertexShader: passVert,

  // ...
})

For a list of shaders you can import via glslify check out stack.gl packages list.

GPU Info

Sometimes it might be useful to enable expensive application configuration only on higher-end devices.

This can be done by detecting the user's GPU and checking in which tier it belongs to based on its benchmark score.

This is done thanks to detect-gpu, more detailed info about these mechanics in its README.

For example, here is how to enable shadows only on high-tier devices:

if (webgl.gpu.tier > 1) {
  webgl.renderer.shadowMap.enabled = true

  // soft shadows
  webgl.renderer.shadowMap.type = THREE.PCFSoftShadowMap
}

Here is what the exposed webgl.gpu object contains:

⚠️ WARNING: webgl.gpu is set asyncronously since the benchmark data needs to be fetched. You might want to wait for the exposed promise webgl.loadGPUTier.

More info on this approach also in this great talk by luruke

Hot reload

TODO

(find some use cases, maybe only the hot shader reload?)

Take a look at: https://github.com/mattdesl/canvas-sketch/blob/1cefbcdf2c5302e74a6a84ff803ddbb377e473f5/docs/hot-reloading.md https://github.com/mattdesl/shader-reload