Skip to content


Repository files navigation


This is a JavaScript Quadtree implementation based on the Java Methods described on by Steven Lambert:

Many games require the use of collision detection algorithms to determine when two objects have collided, but these algorithms are often expensive operations and can greatly slow down a game. One way to speed things up is to reduce the number of checks that have to be made. Two objects that are at opposite ends of the screen can not possibly collide, so there is no need to check for a collision between them. This is where a quadtree comes into play.

This implementation can store and retrieve rectangles in a recursive 2D Quadtree. Every Quadtree node can hold a maximum number of objects before it splits into four subnodes. Objects are only stored on leaf nodes (the lowest level). If an object overlaps into multiple leaf nodes, a reference to the object is stored in each node.

Only 639 Bytes! (Compressed + Gzipped)

checkout "2.0": quadtree-ts with multiple primitives, Typescript and more.



Install this module via npm and import or require it:

npm i -D @timohausmann/quadtree-js
import Quadtree from '@timohausmann/quadtree-js';
const Quadtree = require('@timohausmann/quadtree-js');

Alternatively, download the source and include it the old-fashioned way:

<script src="quadtree.min.js"></script>

Or use an awesome CDN like jsdelivr or unpkg:

<script src=""></script>
<script src=""></script>

How to use

Create a new Quadtree (with default values for max_objects (10) and max_levels (4)).

var myTree = new Quadtree({
    x: 0,
    y: 0,
    width: 400,
    height: 300

MAX_OBJECTS defines how many objects a node can hold before it splits and MAX_LEVELS defines the deepest level subnode.

If you want to specify max_objects and max_levels on your own, you can pass them as a 2nd and 3rd argument. I recommend using low values for max_levels because each level will quadruple the possible amount of nodes. Using lower values for max_levels increases performance but may return more candidates. Finetuning these values depends on your 2D space, the amount and size of the objects and your retrieving areas.

var myTree = new Quadtree({
    x: 0,
    y: 0,
    width: 800,
    height: 600
}, 15, 6);

Insert an element in the Quadtree

    x: 100,
    y: 100,
    width: 100,
    height: 100

Retrieve elements from nodes that intersect with the given bounds

var elements = myTree.retrieve({
    x: 150,
    y: 150,
    width: 100,
    height: 100

Clear the Quadtree


Check out the examples for more information.


Type definitions are included. Inserted objects need to conform to the Quadtree.Rect interface.

import Quadtree, { Rect } from '@timohausmann/quadtree-js';

 * interface Rect {
 *     x: number
 *     y: number
 *     width: number
 *     height: number
 * }

interface Player extends Rect {
    name: string;
    health: number;

const hero:Player = {
    name: 'Shiffman',
    health: 100,
    x: 100,
    y: 100,
    width: 32,
    height: 32


Update single objects

This was often requested and is now supported in quadtree-ts. This might be handy when most of the objects in your Quadtree are static.

Browser Support

This library is supported in all modern browsers and runtimes. 2023 Update: now using ES6 (new Set()) which breaks IE9 compatibility. For IE9 support, use 1.2.5.

Development scripts

  • npm run build to minify the source



  • Performance improvement of retrieve() (was O(n^2), now O(n)) (thanks to xixileng) Pushing the retrieval on a tree with 1.000.000 objects from ~160ms to ~5ms (MacBook M1 Pro).
  • New example: test-retrieve
  • Local copy of quadtree.min.js for the docs folder so it's always up to date


Typescript Definition File Bugfix (thanks to pietrovismara)


Added definition files for Typescript support

JSDoc Fixes


Using for examples (docs), CDN URLs


Removed grunt dev dependency, now using uglify-js to minifiy


Allow float boundaries for Quads

Simplified getIndex function


This implementation now stores objects exclusively on leaf nodes and thus differs from the tutorial it's based on. Objects, that overlap into multiple subnodes are now referenced in each matching subnode instead of their parent node. This drastically reduces the collision candidates. Prior to 1.2.0, overlapping objects were stored in parent nodes.


Support for npm and module.exports