In The Pursuit Of Shiny Web Stats
Being a very visual person, I naturally wanted a nice way of visualising the traffic lastgraph gets (especially as I nurse it through the initial round of bugs, most of which have gone). The ever-wonderful gltail always helps, but the same author has gone one further and come up with gltrail:
It draws paths between pages visited, and looks pretty nice when visualising the lastgraph logs. It's also incredibly tempting to put it on a large screen in my room, the only problem being that I lack a large screen. And enough room.
In a related note, it would be nice if there were some sort of AJAXy, real-time, web-based statistics like gltail; anyone know of any? Someone must have put that <canvas> element to good use by now...
comments
once upon a time I tried to implement that spring-graph algorithm apparently used by gltrail in javascript (visualized by SVG as there was no canvas yet) -- it was a complete mess, and a very slow one, too. If you use a naive algorithm, you'll end up with O(N²), which in practise stands for: "10 nodes, 50 edges, 15fps"... Problem is, that implementing any optimisations is nearly impossible thanks to javascripts overhead when handling objects (maybe people with knowledge of the inner workings might get it to run faster...).
So one would end up with something flashy, like digg-labs demonstrated...
Maybe hacking it together in processing.js is the answer? It's processing implemented in javascript using canvas, which sounds very interesting. I haven't done anything with it yet, though.
See: http://ejohn.org/blog/processingjs/
Yes, processing.js was one thought that occurred to me, but I can't help but think it still adds some overhead that might not be liked - I should probably run some tests at some point.