Re: NoteColumn shift in one voice affects dot position in another voice

[Nick Payne]

On 26/12/09 20:46, David Kastrup wrote:
> Nick Payne<address@hidden>  writes:
>
>    
> > See below. When I shift the first note in voiceTwo to avoid a
> > collision with the note in voiceFour, the dot on the rest that begins
> > voiceOne gets separated from the rest and pretty much disappears under
> > the following accidental. It looks to me as though Lilypond is trying
> > to keep the dots for the rest in voiceOne and the A in voiceFour
> > vertically aligned, regardless of whether the rest and the A are
> > vertically aligned. You can see this quite clearly by setting the
> > hshift value to something larger such as #2.
> >      
> I guess that this sort of thing needs some accounting and/or linear
> programming in order to arrive at a good solution.  You go down the note
> column and put all "natural" dot columns in one list.  Then you assign
> the best columns overall.  Opening a new column in the total is
> expensive (but less so if the vertical distance to the last column is
> large), displacing a dot from its note costs corresponding to the
> displacement, least total cost is taken.  Something like that.
>    
Well, I can fix the output by doing:

\version "2.13.9"

\relative c'' {
    \time 3/4
    \new Staff
<<
    \new Voice { \voiceOne \once \override NoteColumn #'force-hshift = 
#0.1
        \once \override Dots #'extra-offset = #'(-0.9 . 0) r8. <fis 
d>16 e2 }
    \new Voice { \voiceTwo \once \override NoteColumn #'force-hshift = 
#-0.4 e,4 cis'8 e, <a cis> e }
    \new Voice { \voiceFour \once \override NoteColumn #'force-hshift = 
#0.1 a,2. }
>>
}

but that's a rather nasty hack, especially when it has to be repeated 
quite a number of times through a score. I've also found that the 
problem only happens with a dotted rest. If I replace the dotted rest in 
voiceOne with a dotted note, then the dot doesn't get divorced from the 
note in the way that it does from the rest.

Nick