I ran across a great article today by Christer Bystrom that talks of the difference between Quadtree's and Spatial Hashing at implementation and subsequently alludes at the performance implications. I've made a more appropriate demo (in my opinion,) than the confusing code in the spatial hashing implementation on the page to make it more simple to see and understand. I thought that it was worth taking a look at. In my code example I do not show his spatial mapping library. You'll have to use this link to gain access to that. I merely show how I implemented the library. In order to run my demo, you need just reference the easel library which can be included with import {easel} from 'ion-cloud'; after npm install ion-cloud --save to your project.

The highlighted boxes indicate those which have just collided or were forced to overlap due to compression of space allowance (ie, no room for the square to go.) The Spatial Hashing algorithm was created by Christer Bystrom.

This whole project can be viewed on Github here.

const rd0 = uint=> Math.random()*uint;
const rd = (lint,uint)=> Math.random()*(uint-lint)+lint;
const ri = (lint,uint)=> Math.floor(Math.random()*(uint-lint))+lint;

 * This class given an accuracy when created will hold objects and create
 * hashes for objects that are given to it if they don't exist. Hashes that
 * already exist will have those new objects applied to them, indicating that
 * those objects are sharing the same area (overlapping)
class SpatialHash{
    this.accuracy = accuracy || 5;
    this.hash = {};
    this.list = [];

  // perhaps makeKeys would be more accurate, as it makes the
  // keys and then returns them for the object that's passed.
  // it makes the keys based upon the x&ylocation of the object
  // and creates them out to the width and height
    let shift = this.accuracy, //name it appropriately
        sx = obj.x >> shift,
        sy = obj.y >> shift,
        ex = (obj.x + obj.w) >> shift,
        ey = (obj.y + obj.h) >> shift,
        x, y, keys = [];

      } //end for
    } //end for
    return keys;

  // Remove the hashed tree and list
    for(let key in this.hash){
      if(this.hash[key].length === 0){
        delete this.hash[key];
        this.hash[key].length = 0;
      } //end if
    } //end for
    this.list.length = 0;

  // Insert an object into the tree by first adding it to the list, and then
  // computing(and creating) the hash keys to that object if it doesn't exist
  // already
  insert(obj, rect){
    for(let keys = this.getKeys(rect||obj),key,i=0;i<keys.length;i++){
      key = keys[i];
        this.hash[key] = [obj];
      } //end if
    } //end for

  // Retrieve the objects that share the same hash as the specified object.
  // It's important to note that it will also contain the object you're looking
  // for if that object was placed in the spatial hash already
  retrieve(obj, rect){
    let ret = [], keys, i, key;

    if(!obj && !rect) return this.list;
    keys = this.getKeys(rect || obj);
      key = keys[i];
        ret = ret.concat(this.hash[key]);
      } //end if
    } //end for
    return ret;

 * The entities in this demonstration are just rectangles with random width
 * and heights, random velocities on two axis and hold ids and the last objects
 * they interacted with.
class Entity{
    this.viewport = viewport;
    this.id = id;
    this.last = null;
    this.x = rd0(viewport.w);
    this.y = rd0(viewport.h);
    this.w = ri(10,20);
    this.h = ri(10,20);
    this.vx = rd(-1,1);
    this.vy = rd(-1,1);
    this.check = false;

  // Each entity knows not the locations of other entities, and
  // moves blindly. The Mapper tells it when it has collided with
  // another entity and in what direction it should flee to
    this.check = false; // When we move, it's assumed we no longer overlap

 * Mapper is a generic class that holds a collection of entities and a tree
 * for which to relate them to. It initializes the collection, then starts
 * the main loop which tells the entities to move, then calculates their hash
 * and after performing any bouncing interactions, it will draw it and then
 * request a new animation frame from the browser and loop continuously
export class Mapper{
    this.total = 500; //total number of entities we'll draw
    this.collection = []; //holds all of the entities
    this.accuracy = 3; //lower = more accurate = slower
    this.tree = new SpatialHash(this.accuracy);
    this.viewport = viewport;
    this.ctx = ctx;

  // Start the demonstration off by populating a collection of
  // entities up to the total threshold. After everything is populated
  // it calls the main loop function
    for(let i=0;i<this.total;i++){
      this.collection.push(new Entity(i,this.viewport));
    } //end for

  // Separated the drawing portion from the main loop so it was
  // easier to see. We grab the tree from the base node and
  // iterate through all of the objects, highlighting them if they
  // are overlapping another
    let collection = this.tree.retrieve(); //get drawable nodes

        this.ctx.strokeStyle = 'rgba(255,0,0,0.9)'; //overlapping
        this.ctx.strokeStyle = 'rgba(255,255,255,0.2)'; //solitary
      } //end if
      this.ctx.strokeRect(obj.x, obj.y, obj.w, obj.h);

  // The main loop function begins by clearing the scene and then
  // draws each entity before it checks their collisions and moves
  // them coorespondingly.
    let overlapping; //holds all overlapping entities

    this.tree.clear(); //clear the tree and rebuild it
      current.move(); //move the current entity
      this.tree.insert(current); //add the entity back into tree
      overlapping = this.tree.retrieve({
        x: current.x,
        y: current.y,
        w: current.w,
        h: current.h
        let c=current,t=target, //shorten variable names
            hold; //used for swapping velocity variables

          t.check=true;c.check=true; //highlight entities
          if(c.last===t.id||t.last===c.id) return;
          if(c.vx>0&&t.vx>0||t.vx<0&&c.vx<0){ //horizontal collision
          }else if(c.vy>0&&t.vy>0||c.vy<0&&t.vy<0){ //vertical
          }else{ //oblique collision
          } //end if
          c.last=t.id;t.last=c.id; //prevent them from bouncing on overlap
        } //end if