Google Maps - Grid based clustering

2
votes

on this page https://developers.google.com/maps/articles/toomanymarkers

it says how this is the easiest clustering posible, but I can't find any client side solution for this. Does anyone know for free open source, client side solution of grid based clustering for google maps markers ? i don't want MarkerClusterer.

I didn't find it, so i went and start coding my own solution. The first obvious request is that it has to be fast for up to at least 1000 markers loaded on map. Every zoom in out, move and soo on, will repeat procedure/function of clustering. Grid blocks are customizable, default value is 50*50 pixels.

Image below is showing what should the result look like.

I'm having problem with algoritem, now all I can think of to achive this is to itterate through all markers for each grid and ask if it is inside bounderies and then increase array of grid ++. This obviusly can't be done that way. There should be some hash or something that just one itteration through all markers would be needed and I would know in which grid specific marker belongs to.

I hope someone will give me a hint for algoritem I'm after. Thank you !enter image description here

2
google-mapsgoogle-maps-api-3google-maps-markers
The thing with client side clustering is that it defeats the purpose of clustering. With client side clustering you still need to transfer information about each and every marker to the client, (to the browser), and the client has to work hard to do all the number crunching. Unless you have a good reason for doing the clustering on the client, I would suggest that server side clustering is way more efficient.If you do the clustering on the server and then transfer only the minimum amount of data possible to the client, it should load fast on any browser, even on IE. :-) - Marcelo
I want to make plugin, just like MarkerClusterer is. [link]google-maps-utility-library-v3.googlecode.com/svn/trunk/… MarkerClusterer is all client side and works super fast for 1000 markers. The idea must be the same, just go once through markers array. JS file with 5000 markers defined is arround 100kb in size, which is just not that big to be a problem. - displayName
1000 markers is "small stuff". :-) see a sever side solution with 40,000 markers here: maps.forum.nu/server_side_clusterer and here: maps.forum.nu/server_side_clusterer/index2.php - Marcelo
this is nice. but all i want is client side grid clusterer for 1000 markers. simple ... for me hard to do, without some smart algoritem advice. - displayName

2 Answers

2
votes

Seems like geohash is the answer to your problem, you even allude to that solution in your question. It's explained well here:

http://blog.trifork.com/2013/08/01/server-side-clustering-of-geo-points-on-a-map-using-elasticsearch/

I am actually exploring developing a similar clustering algo for Google Maps API v3. I'll post more if I actually get anywhere with it.

0
votes

You can use a geohash or a space filling curve to reduce the dimension and reorder and hash a marker. On the server side you can use a point and a spatial index data stucture but you can also compute the geohash manually. Look for xkcd internet map comic and the hilbert curve to see how it works.