Re: Vertical space for imaginary, real collisions

[Joe Neeman]

On Thu, 2006-10-26 at 21:03 +0200, Han-Wen Nienhuys wrote:
> Joe Neeman escreveu:
> > On Thu, 2006-10-26 at 11:06 -0700, Graham Percival wrote:
> >> Marcus Macauley wrote:
> >>
> >>> 1. Vertical interlocking of staff objects should be allowed by default, 
> >>> with an option to disable;
> >> Due to technical issues, I doubt that this will happen by default in the 
> >> next few years.  I agree that it would be nice, but it's not an easy 
> >> task to do.
> > 
> > Is it really so difficult? It's just a skyline algorithm. The current
> > lily implementation is probably too slow to deal with whole staves but
> > an O(n log n) solution is not so hard to write.
> 
> If you want to do it generically, you'd have to write a general 
> skylining algorithm for stencils, and store all bounding boxes in 
> hierarchical structures. I wonder whether that's feasible.

Why not just a Grob::skyline function that
1) if 'elements or 'element is empty, get the bounding box of the
stencil
2) otherwise merge the skylines of child grobs

Obviously (1) would be more accurate if we had a skylining algorithm for
stencils, but I don't know if that's so important in this particular
case. On the other hand, a skylining algorithm for stencils would
probably have several other uses...