Re: Issue 4205: Improve part combiner's rest analysis (issue 174610043 by address@hidden)

[Dan Eble]

On Nov 25, 2014, at 04:21 , Urs Liska <address@hidden> wrote:
> Is the following assumption correct?
> 
> At the beginning of m.2 the partcombiner treats the crotchet and the full 
> measure rests as two voices.

Yes.  The part combiner directs those rests into voices “one” and “two”.

> At the second crotched, when \one begins to play notes this is considered 
> "solo" because \two doesn't play at that moment.

Yes.

> When I explicitly instantiate a "solo" voice in the \score block this will be 
> somehow merged with the voice implicitly created by the partcombiner.

Yes.  You could do the same with voices “one”, “two”, and “shared”.

> OK. It seems this may be a way to fix all issues with the output but as you 
> say it's not pretty. Actually I'd say it's inacceptably ugly. In my concrete 
> score this would mean I'd have to write such a dummy voice for the 800 
> measure piece, for all partcombined instruments.

In a work of that scale, I agree.

For my own work, which is mostly vocal and mostly short, I have modified the 
part combiner never to create solo sections.  When one part rests, both parts 
are engraved; when both parts rest, they are combined into one.  It sounds like 
this is probably not what you need, but if it would help you, I could give you 
a patch.  I do not know when I will be able to contribute it to Lilypond 
because I am trying to find a more general solution to part combiner 
limitations.
— 
Dan