Re: [AD] new GUI focus selection algorithm

[Elias Pschernig <elias@xxxxxxxxxx>]

On Mon, 2003-11-17 at 15:07, Eric Botcazou wrote:
> > But this computation is not going to be done more often than the user
> > presses a key!
> 
> Sure, but times the number of objects on the screen.
> 
> > That said, we can avoid square roots by comparing the _squares_ of the
> > (Euclidean) distances. It will be faster and not change the behavior.
> 
> I'm a big fan of the man who said: "If you can avoid a multiplication, 
> then... avoid it." :-)
> 
> > If our circles look like squares, then A and B have the same distance to
> > C in this example:
> >
> >   A B
> >
> >
> >       C
> >
> > So moving up from C could equally well move to A or B (or always to A,
> > if it ends one pixel below B).
> 
> Well, it depends upon the orientation of the squares. Certainly we won't take 
> squares (or boxes) that are along the axes. See Shawn's original solution.
> 
> > That could make sense, although it will be tedious to implement because
> > we want to swap the weights when moving horizontally vs vertically.
> 
> I can assure you that I've seen more tedious things :-)
> 

Circles that look like squares, compile time constants which change at
run-time.. I'd say just use the collision approach which is plain and
simple, and clearly defined :)

-- 
Elias Pschernig <elias@xxxxxxxxxx>