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

[Behdad Esfahbod <behdad behdad org>]

Dov Grobgeld wrote:
> Here's another idea that I use in my program giv. Whenever I am painting
> an object on screen, I am painting the same object in an off-screen
> image with the same drawing commands but with "color" that is a
> combination of R,G,B that encode the index of the object. The important
> catch is that you have to turn off anti-aliasing for this to work.
> During motion events I then extract the corresponding pixel from the
> off-screen image, decode the R,G, and B to get the index of the object,
> which points to a vector of properties that I show.

I always had this idea of using cairo for hit detection in this way.  Didn't
know you have already implemented it Dov.  Great!

Cheers,
behdad


> Regards,
> Dov
> 
> 2009/2/7 Fabio Mariotti <mariotti fabio gmail com
> <mailto:mariotti fabio gmail com>>
> 
>     Ok I have 2 partial answers to myself:
> 
> 
>         I need to draw triangles (and circles) on the screen (any drawable)
>         and have active events on each object I draw: namely mouse_over,
>         click.. etc..
>         [they can overlap]
> 
>         Is there any alternative to gnome-canvas?
>         In particular gnome-canvas was working nicely up to 20000 objects
>         but a bit heavy with 40000. (I would guess that an equivalent
>         canvas solution will suffer the same problem..)
> 
>         Is there a simple way to "recode" a simple z-buffer using
>         existing tools?
>         With simple I mean that I do not have to recalculate exposed pixels.
> 
> 
>     1) There is no simple coding or otherwise I would be doing a canvas!
> 
>     2) for my little application I have a possible solution:
>         I set up a z-buffer as simple integers vector. I store 2 doubles
>         at least for each figure so an integer shouldn't harm the memory.
>         Due to the type of application, on an click at x,y, I can guess
>         a limited number of triangles which might be present in the area.
>         I can redraw the area in the style of a canvas and propagate
>         the event. Grab it. destroy the canvas-like object. and redraw
>         in normal style.
>         I am not sure if I can actually do it... more over I'll still need
>         the canvas part.
>         The advantage is that (of course I do not have to code again
>         the pixel stuff) I can change shape of my
>         objects at will (ok twice) and moreover I can simply use the event
>         stuff which exists already.
>         The overload on the users side shouldn't e that big They will stare
>         at the screen anyway.. ;).
>         On which widget can I draw a canvas at arbitrary coordinates?
>         I'll find it out ;)
>         [
>         I got the idea from an old post...
>         http://mail.gnome.org/archives/gtk-list/2006-September/msg00275.html
>         Where I discussed the problem in different terms.
>         I didn't actually try it. The first reason was that on win
>     machines it would
>         have made no difference.
>         ]
> 
>        A note: would it be a possibility to have a gdk widget named
>     shape(for example)
>       which simply draw itself and gets events?
>       It is just the same problem of shaped windows and the like.
> 
>       sorry again for the long post..
>       Faio
> 
> 
>     _______________________________________________
>     gtk-list mailing list
>     gtk-list gnome org <mailto:gtk-list gnome org>
>     http://mail.gnome.org/mailman/listinfo/gtk-list
> 
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> gtk-list mailing list
> gtk-list gnome org
> http://mail.gnome.org/mailman/listinfo/gtk-list