Re: automatic accidental style voice: too many written accidentals

[David Kastrup]

Frédéric Bron <address@hidden> writes:

>> Since the \key command still works at the Staff level (both
>> technically in LilyPond and musically, since there's no notation
>> available to specify separate key signatures for separate voices
>> within a stave), I clearly see it as a bug if the Staff key isn't
>> seen by each Voice.
>
> What if, at the point a voice is instanciated, we automatically set
> its key signature to that of its parent staff context. That would
> solve this issue.

In general, that is a mistake: accidentals are typeset per staff.  Now
this is about an accidental-style 'voice for which we have

     As a result, accidentals from one voice do not get canceled in
     other voices, which is often an unwanted result: in the following
     example, it is hard to determine whether the second `a' should be
     played natural or sharp.  The `voice' option should therefore be
     used only if the voices are to be read solely by individual
     musicians.  If the staff is to be used by one musician (e.g., a
     conductor or in a piano score) then `modern' or
     `modern-cautionary' should be used instead.

Under this premise, I think it makes sense to inherit the key signature.
Now if we have _within_ a voice << { xxx } \\ { yyy } >>, what does this
mean?  Since the purpose of "voice" accidental style is to write stuff
that several different people read, this corresponds to splitting a
tenor voice (for example).  This means that the split voices should not
just inherit the key signature, but also already preexisting accidentals
of the "father voice" since obviously the same persons are singing tenor
1 and tenor 2 that have been singing tenor previously.  Also, any
accidental changes within xxx and yyy need to be consolidated into the
following rejoined voice: if there are different accidentals in xxx and
yyy for a note, a recurrence of the note after the split needs a
cautionary accidental.

-- 
David Kastrup