React for Data Visualization
Student Login
  • Introduction

Dancing tree fractal with React

That damn Pythagorean fractal from last month wouldn't leave me alone, so I fixed it. Timing may or may not coincide with a commenter giving a solution to the wonky triangle problem.


Here’s the code on Github. For the story and explanation, keep reading. :)

It took somebody doing the math instead of me to kick my arse into gear. Here's Vinicius Ribeiro schooling me on high school trigonometry:

You are not applying the Law of Sines correctly. The variable 'd' is the diameter of the triangle's circumcircle, not its perimeter.

Tinkering with your code on github, I've managed to accomplish what you were trying to do by doing the following calculations:

const currentH = 0.2 * w,
nextLeft = Math.sqrt(currentH * currentH + 0.7 * w * 0.7 * w),
nextRight = Math.sqrt(currentH * currentH + 0.3 * w * 0.3 * w),
A = Math.deg(Math.atan(currentH / (0.3 * w))),
B = Math.deg(Math.atan(currentH / (0.7 * w)));

The height of the inner triangle is a fraction of the current 'w'. By doing that, we can infer nextLeft and nextRight using the Pythagorean theorem. The angles can then be calculated using the inverse tangent (atan) and the triangle height.

Hope this helps!

Help it did! Thanks, Vinicius.

How you too can build a dancing tree fractal

Equipped with basic trigonometry, you need 3 ingredients to build a dancing tree:

  • a recursive <Pythagoras> component
  • a mousemove listener
  • a memoized next-step-props calculation function

We'll use the <Pythagoras> component from November, add a D3 mouse listener, and put Vinicus's math with some tweaks into a memoized function. We need D3 because its mouse listeners automatically calculate mouse position relative to SVG coordinates, and memoization helps us keep our code faster.

The improved <Pythagoras> component takes a few more arguments than before, and it uses a function to calculate future props. Like this:

const Pythagoras = ({ w,x, y, heightFactor, lean, left, right, lvl, maxlvl }) => {
if (lvl >= maxlvl || w < 1) {
return null;
const { nextRight, nextLeft, A, B } = memoizedCalc({
w: w,
heightFactor: heightFactor,
lean: lean
let rotate = '';
if (left) {
rotate = `rotate(${-A} 0 ${w})`;
}else if (right) {
rotate = `rotate(${B} ${w} ${w})`;
return (
<g transform={`translate(${x} ${y})="" ${rotate}`}="">
<rect width={w} height={w} x={0} y={0} style="{{fill:" interpolateviridis(lvl="" maxlvl)}}="">
<pythagoras w={nextLeft} x={0} y={-nextLeft} lvl={lvl+1} maxlvl={maxlvl} heightfactor={heightFactor} lean={lean} left="">
<pythagoras w={nextRight} x={w-nextRight} y={-nextRight} lvl={lvl+1} maxlvl={maxlvl} heightfactor={heightFactor} lean={lean} right="">

We break recursion whenever we try to draw an invisible square or have reached too deep into the tree. Then we:

  • use memoizedCalc to do the mathematics
  • define different rotate() transforms for the left and right branches
  • and return an SVG <rect> for the current rectangle, and two <Pythagoras> elements for each branch.

Most of this code deals with passing arguments onwards to children. It’s not the most elegant approach, but it works. The rest is about positioning branches so corners match up.


The maths

I don't really understand this math, but I sort of know where it's coming from. It's the sine law applied correctly. You know, the part I failed at miserably last time ?

const memoizedCalc = (function () {
const memo = {};
const key = ({ w, heightFactor, lean }) => [w, heightFactor, lean].join("-");
return (args) => {
const memoKey = key(args);
if (memo[memoKey]) {
return memo[memoKey];
} else {
const { w, heightFactor, lean } = args;
const trigH = heightFactor * w;
const result = {
nextRight: Math.sqrt(trigH ** 2 + (w * (0.5 + lean)) ** 2),
nextLeft: Math.sqrt(trigH ** 2 + (w * (0.5 - lean)) ** 2),
A: Math.deg(Math.atan(trigH / ((0.5 - lean) * w))),
B: Math.deg(Math.atan(trigH / ((0.5 + lean) * w))),
memo[memoKey] = result;
return result;

We added to Vinicius's maths a dynamic heightFactor and lean adjustment. We'll control those with mouse movement.

To improve performance, maybe, our memoizedCalc function has an internal data store that maintains a hash of every argument tuple and its result. This lets us avoid computation and read from memory instead.

At 11 levels of depth, memoizedCalc gets called 2,048 times and only returns 11 different results. You can't find a better candidate for memoization.

Of course, a benchmark would be great here. Maybe sqrt, atan, and ** aren't that slow, and our real bottleneck is redrawing all those nodes on every mouse move. Hint: it totally is.

Now that I spell it out… what the hell was I thinking? I'm impressed it works as well as it does.

The mouse listener

Inside App.js, we add a mouse event listener. We use D3's because it gives us the SVG-relative position calculation out of the box. With React’s, we'd have to do the hard work ourselves.

// App.js
state = {
currentMax: 0,
baseW: 80,
heightFactor: 0,
lean: 0
componentDidMount() {
.on("mousemove", this.onMouseMove.bind(this));
onMouseMove(event) {
const [x, y] = d3mouse(this.refs.svg),
scaleFactor = scaleLinear().domain([this.svg.height, 0])
.range([0, .8]),
scaleLean = scaleLinear().domain([0, this.svg.width/2, this.svg.width])
.range([.5, 0, -.5]);
heightFactor: scaleFactor(y),
lean: scaleLean(x)
// ...
render() {
// ...
<svg ref="svg"> //...
<pythagoras w={this.state.baseW} h={this.state.baseW} heightfactor={this.state.heightFactor} lean={this.state.lean} x={this.svg.width/2-40} y={this.svg.height-this.state.baseW} lvl={0} maxlvl="{this.state.currentMax}/">

A couple of things happen here:

  • we set initial lean and heightFactor to 0
  • in componentDidMount, we use and .on to add a mouse listener
  • we define an onMouseMove method as the listener
  • we render the first <Pythagoras> using values from state

The lean parameter tells us which way the tree is leaning and by how much; the heightFactor tells us how high those triangles should be. We control both with the mouse position.

That happens in onMouseMove:

onMouseMove(event) {
const [x, y] = d3mouse(this.refs.svg),
scaleFactor = scaleLinear().domain([this.svg.height, 0])
.range([0, .8]),
scaleLean = scaleLinear().domain([0, this.svg.width/2, this.svg.width])
.range([.5, 0, -.5]);
heightFactor: scaleFactor(y),
lean: scaleLean(x)

d3mouse – which is an imported mouse function from d3-selection – gives us cursor position relative to the SVG element. Two linear scales give us scaleFactor and scalelean values, which we put into component state.

If you're not used to D3 scales, this reads as:

  • map vertical coordinates between height and 0 evenly to somewhere between 0 and .8
  • map horizontal coordinates between 0 and width/2 evenly to somewhere between .5 and 0, and coordinates between width/2 and width to 0 and -.5

When we feed a change to this.setState, it triggers a re-render of the entire tree, our memoizedCalc function returns new values, and the final result is a dancing tree.


Beautious. ?

PS: last time, I mentioned that recursion stops working when you make a React build optimized for production. That doesn't happen. I don't know what was wrong with the specific case where I saw that behavior. ¯\(ツ)

About the Author

Hi, I’m Swizec Teller. I help coders become software engineers.

Story time 👇

React+D3 started as a bet in April 2015. A friend wanted to learn React and challenged me to publish a book. A month later React+D3 launched with 79 pages of hard earned knowledge.

In April 2016 it became React+D3 ES6. 117 pages and growing beyond a single big project it was a huge success. I kept going, started live streaming, and publishing videos on YouTube.

In 2017, after 10 months of work, React + D3v4 became the best book I'd ever written. At 249 pages, many examples, and code to play with it was designed like a step-by-step course. But I felt something was missing.

So in late 2018 I rebuilt the entire thing as React for Data Visualization — a proper video course. Designed for busy people with real lives like you. Over 8 hours of video material, split into chunks no longer than 5 minutes, a bunch of new chapters, and techniques I discovered along the way.

React for Data Visualization is the best way to learn how to build scalable dataviz components your whole team can understand.

Some of my work has been featured in 👇

Created bySwizecwith ❤️