Table of Contents

# threejsfundamentals.org

Fix, Fork, Contribute

# Three.js Primitives

This article is one in a series of articles about three.js. The first article was about fundamentals. If you haven't read that yet you might want to start there.

Three.js has a large number of primitives. Primitives are generally 3D shapes that are generated at runtime with a bunch of parameters.

It's common to use primitives for things like a sphere for a globe or a bunch of boxes to draw a 3D graph. It's especially common to use primitives to experiment and get started with 3D. For the majority of 3D apps it's more common to have an artist make 3D models in a 3D modeling program like Blender or Maya or Cinema 4D. Later in this series we'll cover making and loading data from several 3D modeling programs. For now let's go over some of the available primitives.

Many of the primitives below have defaults for some or all of their parameters so you can use more or less depending on your needs.

A Box
A flat circle
A Cone
A Cylinder
A dodecahedron (12 sides)
An extruded 2d shape with optional bevelling. Here we are extruding a heart shape. Note this is the basis for `TextGeometry`.
An icosahedron (20 sides)
A shape generated by spinning a line. Examples would be: lamps, bowling pins, candles, candle holders, wine glasses, drinking glasses, etc... You provide the 2d silhouette as series of points and then tell three.js how many subdivisions to make as it spins the silhouette around an axis.
An Octahedron (8 sides)
A surface generated by providing a function that takes a 2D point from a grid and returns the corresponding 3d point.
A 2D plane
Takes a set of triangles centered around a point and projects them onto a sphere
A 2D disc with a hole in the center
A 2D outline that gets triangulated
A sphere
A tetrahedron (4 sides)
3D text generated from a 3D font and a string
A torus (donut)
A torus knot
A circle traced down a path
A helper object that takes another geometry as input and generates edges only if the angle between faces is greater than some threshold. For example if you look at the box at the top it shows a line going through each face showing every triangle that makes the box. Using an `EdgesGeometry` instead the middle lines are removed. Adjust the thresholdAngle below and you'll see the edges below that threshold disappear.
Generates geometry that contains one line segment (2 points) per edge in the given geometry. Without this you'd often be missing edges or get extra edges since WebGL generally requires 2 points per line segment. For example if all you had was a single triangle there would only be 3 points. If you tried to draw it using a material with `wireframe: true` you would only get a single line. Passing that triangle geometry to a `WireframeGeometry` will generate a new geometry that has 3 lines segments using 6 points..

We'll go over creating custom geometry in another article. For now let's make an example creating each type of primitive. We'll start with the examples from the previous article.

Near the top let's set a background color

```const scene = new THREE.Scene();
+scene.background = new THREE.Color(0xAAAAAA);
```

This tells three.js to clear to lightish gray.

The camera needs to change position so that we can see all the objects.

```-const fov = 75;
+const fov = 40;
const aspect = 2;  // the canvas default
const near = 0.1;
-const far = 5;
+const far = 1000;
const camera = new THREE.PerspectiveCamera(fov, aspect, near, far);
-camera.position.z = 2;
+camera.position.z = 120;
```

Let's add a function, `addObject`, that takes an x, y position and an `Object3D` and adds the object to the scene.

```const objects = [];
const spread = 15;

function addObject(x, y, obj) {
obj.position.x = x * spread;
obj.position.y = y * spread;

scene.add(obj);
objects.push(obj);
}
```

Let's also make a function to create a random colored material. We'll use a feature of `Color` that lets you set a color based on hue, saturation, and luminance.

`hue` goes from 0 to 1 around the color wheel with red at 0, green at .33 and blue at .66. `saturation` goes from 0 to 1 with 0 having no color and 1 being most saturated. `luminance` goes from 0 to 1 with 0 being black, 1 being white and 0.5 being the maximum amount of color. In other words as `luminance` goes from 0.0 to 0.5 the color will go from black to `hue`. From 0.5 to 1.0 the color will go from `hue` to white.

```function createMaterial() {
const material = new THREE.MeshPhongMaterial({
side: THREE.DoubleSide,
});

const hue = Math.random();
const saturation = 1;
const luminance = .5;
material.color.setHSL(hue, saturation, luminance);

return material;
}
```

We also passed `side: THREE.DoubleSide` to the material. This tells three to draw both sides of the triangles that make up a shape. For a solid shape like a sphere or a cube there's usually no reason to draw the back sides of triangles as they all face inside the shape. In our case though we are drawing a few things like the `PlaneGeometry` and the `ShapeGeometry` which are 2 dimensional and so have no inside. Without setting `side: THREE.DoubleSide` they would disappear when looking at their back sides.

I should note that it's faster to draw when not setting `side: THREE.DoubleSide` so ideally we'd set it only on the materials that really need it but in this case we are not drawing too much so there isn't much reason to worry about it.

Let's make a function, `addSolidGeometry`, that we pass a geometry and it creates a random colored material via `createMaterial` and adds it to the scene via `addObject`.

```function addSolidGeometry(x, y, geometry) {
const mesh = new THREE.Mesh(geometry, createMaterial());
addObject(x, y, mesh);
}
```

Now we can use this for the majority of the primitives we create. For example creating a box

```{
const width = 8;
const height = 8;
const depth = 8;
addSolidGeometry(-2, -2, new THREE.BoxGeometry(width, height, depth));
}
```

If you look in the code below you'll see a similar section for each type of geometry.

Here's the result:

There are a couple of notable exceptions to the pattern above. The biggest is probably the `TextGeometry`. It needs to load 3D font data before it can generate a mesh for the text. That data loads asynchronously so we need to wait for it to load before trying to create the geometry. By promisifiying font loading we can make it mush easier. We create a `FontLoader` and then a function `loadFont` that returns a promise that on resolve will give us the font. We then create an `async` function called `doit` and load the font using `await`. And finally create the geometry and call `addObject` to add it the scene.

```{
const loader = new THREE.FontLoader();
// promisify font loading
function loadFont(url) {
return new Promise((resolve, reject) => {
loader.load(url, resolve, undefined, reject);
});
}

async function doit() {
const font = await loadFont('resources/threejs/fonts/helvetiker_regular.typeface.json');  /* threejsfundamentals: url */
const geometry = new THREE.TextGeometry('three.js', {
font: font,
size: 3.0,
height: .2,
curveSegments: 12,
bevelEnabled: true,
bevelThickness: 0.15,
bevelSize: .3,
bevelSegments: 5,
});
const mesh = new THREE.Mesh(geometry, createMaterial());
geometry.computeBoundingBox();
geometry.boundingBox.getCenter(mesh.position).multiplyScalar(-1);

const parent = new THREE.Object3D();
parent.add(mesh);

addObject(-1, -1, parent);
}
doit();
}
```

There's one other difference. We want to spin the text around its center but by default three.js creates the text such that its center of rotation is on the left edge. To work around this we can ask three.js to compute the bounding box of the geometry. We can then call the `getCenter` method of the bounding box and pass it our mesh's position object. `getCenter` copies the center of the box into the position. It also returns the position object so we can call `multiplyScalar(-1)` to position the entire object such that its center of rotation is at the center of the object.

If we then just called `addSolidGeometry` like with previous examples it would set the position again which is no good. So, in this case we create an `Object3D` which is the standard node for the three.js scene graph. `Mesh` is inherited from `Object3D` as well. We'll cover how the scene graph works in another article. For now it's enough to know that like DOM nodes, children are drawn relative to their parent. By making an `Object3D` and making our mesh a child of that we can position the `Object3D` wherever we want and still keep the center offset we set earlier.

If we didn't do this the text would spin off center.

Notice the one on the left is not spinning around its center whereas the one on the right is.

The other exceptions are the 2 line based examples for `EdgesGeometry` and `WireframeGeometry`. Instead of calling `addSolidGeometry` they call `addLineGeometry` which looks like this

```function addLineGeometry(x, y, geometry) {
const material = new THREE.LineBasicMaterial({color: 0x000000});
const mesh = new THREE.LineSegments(geometry, material);
addObject(x, y, mesh);
}
```

It creates a black `LineBasicMaterial` and then creates a `LineSegments` object which is a wrapper for `Mesh` that helps three know you're rendering line segments (2 points per segment).

Each of the primitives has several parameters you can pass on creation and it's best to look in the documentation for all of them rather than repeat them here. You can also click the links above next to each shape to take you directly to the docs for that shape.

There is one other pair of classes that doesn't really fit the patterns above. Those are the `PointsMaterial` and the `Points` class. `Points` is like `LineSegments` above in that it takes a a `BufferGeometry` but draws points at each vertex instead of lines. To use it you also need to pass it a `PointsMaterial` which take a `size` for how large to make the points.

```const radius = 7;
const widthSegments = 12;
const heightSegments = 8;
const geometry = new THREE.SphereGeometry(radius, widthSegments, heightSegments);
const material = new THREE.PointsMaterial({
color: 'red',
size: 0.2,     // in world units
});
const points = new THREE.Points(geometry, material);
scene.add(points);
```

You can turn off `sizeAttenuation` by setting it to false if you want the points to be the same size regardless of their distance from the camera.

```const material = new THREE.PointsMaterial({
color: 'red',
+    sizeAttenuation: false,
+    size: 3,       // in pixels
-    size: 0.2,     // in world units
});
...
```

One other thing that's important to cover is that almost all shapes have various settings for how much to subdivide them. A good example might be the sphere geometries. Spheres take parameters for how many divisions to make around and how many top to bottom. For example

The first sphere has 5 segments around and 3 high which is 15 segments or 30 triangles. The second sphere has 24 segments by 10. That's 240 segments or 480 triangles. The last one has 50 by 50 which is 2500 segments or 5000 triangles.

It's up to you to decide how many subdivisions you need. It might look like you need a high number of segments but remove the lines and the flat shading and we get this

It's now not so clear that the one on the right with 5000 triangles is entirely better than the one in the middle with only 480. If you're only drawing a few spheres, like say a single globe for a map of the earth, then a single 10000 triangle sphere is not a bad choice. If on the other hand you're trying to draw 1000 spheres then 1000 spheres times 10000 triangles each is 10 million triangles. To animate smoothly you need the browser to draw at 60 frames per second so you'd be asking the browser to draw 600 million triangles per second. That's a lot of computing.

Sometimes it's easy to choose. For example you can also choose to subdivide a plane.

The plane on the left is 2 triangles. The plane on the right is 200 triangles. Unlike the sphere there is really no trade off in quality for most use cases of a plane. You'd most likely only subdivide a plane if you expected to want to modify or warp it in some way. A box is similar.

So, choose whatever is appropriate for your situation. The less subdivisions you choose the more likely things will run smoothly and the less memory they'll take. You'll have to decide for yourself what the correct tradeoff is for your particular situation.

If none of the shapes above fit your use case you can load geometry for example from a .obj file or a .gltf file. You can also create your own custom BufferGeometry.

Next up let's go over how three's scene graph works and how to use it.

Use <pre><code>code goes here</code></pre> for code blocks
comments powered by Disqus