Re: Issue 415 in lilypond: Hidden accidental of tied note still takes space

[Joe Neeman]

On Tue, 2009-03-24 at 15:35 +0000, address@hidden wrote:
> Comment #7 on issue 415 by address@hidden: Hidden accidental of tied  
> note still takes space
> http://code.google.com/p/lilypond/issues/detail?id=415
> 
> I did some looking into this recently to see if I could fix it. What I  
> found is that
> we have a circular dependency situation: the line-break code needs to know  
> how much
> space the accidentals are taking up to know when to insert line breaks, but  
> tied
> accidentals will be printed only if the measure is the first measure in the  
> line,
> making accidental spacing dependent on line breaks.
> 
> When I realized that, I decided that fixing it required more knowledge of  
> the
> LilyPond code than I possess. A possible solution I came up with (but do  
> not know how
> to implement) would be to have the accidentals take up space when the line  
> breaks are
> calculated, but then re-calculate the spacing within each line after  
> deciding where
> the line breaks are, removing tied accidentals where appropriate.

It should be possible to solve this without doing the line-breaking
twice. Have a look at how we handle input/regression/accidental-tie.ly.
First of all, you need to understand that every breakable column (the
non-musical columns that contain barlines, etc, not notes) has 3 copies.
Different copies are used depending on whether the column in question
goes at the beginning, the middle or the end of a line.

The accidental enters the spacing calculation in staff-spacing.cc:189.
You'll see in paper-column.cc:158 that the "conditional skyline" is
added to the column on the right (ie. the column with the note and the
accidentals). The "normal" skyline that is retrieved in
paper-column.cc:152 has no accidentals in it; the "conditional" skyline
may or may not contain the accidentals depending on the "left" column.
If the left column is the breakable paper column that begins a line, all
the accidentals are included in the skyline; the logic that accomplishes
this is accidental-placement.cc:85, which is called from
separation-item.cc:103.

This way, the spacing is done using only the accidentals that are
relevant and it is why bug 415 is fixed. The problem is that the
position of the relevant accidentals is still affected by the position
of the irrelevant accidentals and that is why bug 612 is still open.

One solution might be the following, which involves modifying only
accidental-placement: make 2 copies of every accidental, one for the
start-of-a-line case and one for the middle-of-a-line case (so the
start-of-a-line accidentals would be, in the terminology of
Accidental_placement::split_accidentals, the break_reminder accidentals
and the real_accidentals while the middle-of-a-line accidentals would
have copies of only the real_accidentals). Run the accidental layout
algorithm (currently in Accidental_placement::calc_positioning_done, but
you'll want to move it to a different function and make
calc_positioning_done call it twice) twice, once for each set of
accidentals. Then modify Accidental_placement::get_relevant_accidentals
to return one of the sets of accidentals.

Joe