Re: skyline vertical spacing

[Werner LEMBERG]

> I don't know if this helps, but I made an algorithm to approximate
> bezier curves with straight lines (and I have a prototype in
> Haskell).

Hmm.  Such skylines don't need to be too detailed -- for example, in
case of a sharp (#) it's sufficient IMHO to handle the top and bottom
as a straight line.

> It seems that the output of potrace is entirely straight lines and
> bezier curves, so I've been using the .eps files generated by
> potrace for testing.

Well, the plan is eventually to move away from potrace but using
mf2pt1.  Since the number of lilypond glyphs is limited it shouldn't
be too great a problem to add skylines manually.

I've already `converted' all symbols except the one for ancient
notation -- Jürgen, do you have time to work on improving the glyph
shapes of the ancient notation so that mf2pt1 produces sensible
results?

> Apart from removing the hard-coding of accidental skylines, I
> thought it could be useful for putting spanners above and below the
> staff (Graham mentioned something like this before).

Definitely.

> The basic idea is that we define an order in which dynamics,
> hairpins, textscripts, etc. are placed. Then we
> 1) generate a skyline of the staff
> 2) place the next spanner so there are no collisions
> 3) merge the skyline of the spanner
> 4) goto 1

It would be even useful to make lilypond automatically rotate hairpins
in case the possible start and positions differ much.

However, the list of bugs and buglets related to non-collision
problems is quite long.  I would like to see you work on such items
also, if you don't mind :-)


    Werner