Browse Feature Updated
Thu, 16/07/2009 - 22:08 — matt
I've updated the browse map. Instead of loading thousands of markers individually, they are now loaded to the map as clusters and display like a heat map. The UI controls have also been updated to the Google default. Oh and the maps a bit bigger to.
I've added some screenshots to show the difference.
Comments
North Sea
I've always wondered: why is there a post box in the middle of the North Sea ?
is it an Oil Platform ? There's no other information with it.
Browsing / Searching / Contributing
This heat map is a step in the right direction, but I can't help feeling that you've got the wrong end of the stick: you have three separate pathways: browse, search, and contribute.
For each new postbox that I have to contribute, I have to:
1. Search to see what postboxen exist in a given area, to see if any of the ones that I have are already present.
2. (For those that I have that are new), start from the whole-UK map again and zoom / click / pan several times to get to the right location and then add a new marker.
(1) allows me to specify an approximate area and radius. Then, for step 2, I have to start from the whole-UK map again, clicking / panning / zooming to get to where I was in step 1.
I can't add multiple postboxen at (2) because I have to go back to (1) again to have a look at those already existing.
Why can't I specify an approximate location for the "contribute" phase ? This would cut down enormously on the terribly-long-winded panning and zooming.
It would be a good idea if the browse / search / contribute were all part of the same whole:
Start from a map of the UK
Zoom / Pan to a particular location and/or enter its approx location and radius (with the radius automatically translating to a map scale)
When you're at a particular location, all the extant postboxes in that area should be shewn (whether as cluster points or as individuals) (map as a turnoffable layer).
If wanted, the persun could add a contribution in situ without having to keep starting from the beginning again, and with the added advantage of seeing all the existing postboxen.
It also looks weird when, having specified an approximate location and radius, you zoom / pan away, and see no additional postboxes, simply because the map view no longer corresponds to the given search radius.
You don't expect this to happen with other mapping features: you don't expect e.g. bus stops to no longer be shewn (or only a subset of them shewn) just because you've moved away from the original area.
I'm not trying to be critical: I just see the way it's all implemented at the moment as a bit of a wasted opportunity. The only thing that I can think is that the computation of each PB's (rectangular coords) position is a very expensive process, and not something that can be done on the fly.
New browse function unusably slow
I've tried this new browse function but it is very slow, taking over 20 minutes to load a cluster map. This is more apparent when I'm trying to zoom in. I'm using firefox and eventually FF either locks up with 100% CPU time, or gives a warning about a script taking too long.
Are you precomputing the clusters or are you building them on the fly ? Are you rebuilding the clusters *for the whole database* even when only a portion of it is visible at the current map scale ?
I have a 3.6M mobile broadband connexion and data traffic speed doesn't seem to be the problem.
It works fine in my tests
It works fine in my tests only taking a few seconds to load the markers/clusters, but i've taken browse offline for the time being to investigate further.
Cluster Maps
To be perfectly honest, Postboxr has *always* been slow on all my computers; it's not just the heat map. It's much slower (unusuably so in many cases) than any other web sites that I frequent. Even when the markers were drawn individually, it took *minutes*, admittedly not "20 minutes", like it takes now.
If you remember, I had problems before when it wouldn't update when you added the "Business Box for Franked Mail" icon.
I'm starting to wonder whether t-Mobile is putting some form of weird transparent caching or proxying in. If I had the ability to bypass the proxy ( I don't ), I could test further.
I've tried it on my work's computer and those in the library (both of which use IE). It's a lot faster there, but only by the ratio you'd expect due to an increase in broadband speed: it still takes "minutes" e.g. to zoom in on a place.
Richard [in SG1]
Cluster Maps
OK. I've just tried it on IE and FF.
Doing a search on FF worked the first time round. Then the second time, CPU usage shot up to 100% and I had to kill it with TaskManager. When I managed to get things stable again, FF started up with the "do you want to try and restore the previous session" message. This happens all the time. Trying it with IE gave similar results: "not responding", had to kill with TaskManager.
I'm not getting this with any other web sites (eBay, google, bbc, twitter, whatever).
Is there some unusual technology that you're using ? I find it very bizarre that a simple web page can cause the computer to lock-up in such a way.