[Lilypond-auto] Issue 2878 in lilypond: [musicxml2ly] wrong durations in chorded notes with time-modifications

[lilypond]

Status: Accepted
Owner: ----
Labels: Type-Enhancement

New issue 2878 by address@hidden: [musicxml2ly] wrong durations in  
chorded notes with time-modifications
http://code.google.com/p/lilypond/issues/detail?id=2878

Reported by pls:
http://lists.gnu.org/archive/html/bug-lilypond/2012-09/msg00136.html

Hey all,

Sibelius 7.0.0 exports wrong duration values for chorded notes in tuplets.   
The
exported value in this example
(​ 
https://www.dropbox.com/s/vrvdabwym9tezd0/time-modification_and_chorded_notes.xml)
 is 256 instead of 170 for a quarter note in a triplet of eighths
(<divisions>256</divisions>).  <durations>170</durations> is not an optimal
value for a quarter note in a triplet of eighths as there is a rounding
difference (256*2/3=170.6666...).  The same holds for an eighth note in a
triplet of eighths (128*2/3=85.3333... instead of 86 used here).   
musicxml2ly
does not like rounding errors.  The resulting LilyPond-code contains wild
scaling durations and a superfluous spacer rest:
\times 2/3 { <c'' a'>4*255/256 s512*43 r8*129/128 } r4 r2 }

This is bad because we have a bar check problem here!

Sometimes music applications export suboptimal/false values for divisions  
and
durations.
A solution: musicxml2ly should recalculate all durations on the basis of a
divisions value of 6720.   This figure is divisible without remainder by  
e.g.
2, 3, 4, 5, 6, 7, 8, 16, 32, 64 and therefore can provide for note  
durations up
to 256th.  A quarter note (6720) in a triplet of eighths would have a value  
of
4480 (6720*2/3).  An eighth note (6720/2=3360) in a triplet of eighths would
have a value of 2240 (3360*2/3).  (half note: 6720*2 = 13440.)

hth
patrick