Element point map for html5 canvas element, need algorithm

Question

I'm currently working on a pure html 5 canvas implementation of the "flying tag cloud sphere", which many of you have undoubtedly seen as a flash object in some pages.

The tags are drawn fine, and the performance is satisfactory, but there's one thing in the canvas element that's kind of breaking this idea: you can't identify the objects that you've drawn on a canvas, as it's just a simple flat "image"..

What I have to do in this case is catch the click event, and try to "guess" which element was clicked. So I have to have some kind of matrix, which stores a link to a tag object for each pixel on the canvas, AND I have to update this matrix on every redraw. Now this sounds incredibly inefficient, and before I even start trying to implement this, I want to ask the community - is there some "well known" algorithm that would help me in this case? Or maybe I'm just missing something, and the answer is right behind the corner? :)

Answer 1

This is called the point location problem, and it's one of the basic topics in computational geometry. There are a lot of methods you could use that would be much faster than the approach you're thinking of, but the details depend on what exactly you want to accomplish.

For example, each text string is contained in a bounding box. Do you just want to test whether the user clicked somewhere in that box? Then simply store the minimum and maximum coordinates of each rendered string, and test the point against each bounding box to see if it's contained in that range. If you have a large number of points to test, you can build any number of data structures to speed this up (e.g. R-trees), but for a single point the overhead of constructing such a structure probably isn't worthwhile.

If you care about whether the point actually falls within the opaque area of the stroked characters, the problem is slightly trickier. One solution would be to use the bounding box approach to first eliminate most of the possibilities, and then render the remaining strings one at a time to an offscreen buffer, checking each time to see if the target point has been touched.