Re: events on a lot of triangles: [simplification of a previous post]

[Murray Cumming <murrayc murrayc com>]

On Mon, 2009-02-09 at 13:13 +0200, Dov Grobgeld wrote:
> The problem with any general canvas solution is that it can only make
> limited assumptions about the scale and subdivision of the data. E.g.
> if you
> know that all your line segments are of certain max size, then you can
> use
> that knowledge to do a more efficient lookup. Look e.g. at Google
> Earth that
> is presenting different info depending on the zoom-in. I assume that
> it is
> achieving that through a heavy preprocessing of the data that is
> cached on
> the server. This is difficult to do with a general purpose canvas
> object.
> That's why I turned to a callback model in my widget
> gtk_image_viewer , that
> is much more flexble, though you have to do more work on your own.

This canvas comparison already mentions the idea of "non-scaled items"
which is maybe a sub-set of what you mean, though no canvases seem to
implement it yet:
http://live.gnome.org/ProjectRidley/CanvasOverview

It seems like something that could be done, though many canvas authors
don't seem to like the idea of having any (even optional) higher-level
concepts in their canvases, even where a lower-level implementation
would make the higher-level concepts easier to implement.

-- 
murrayc murrayc com
www.murrayc.com
www.openismus.com