Re: instrumentSwitch and addInstrumentDefinition use

[Kieren MacMillan]

Hi David,

> You want to write music as if it is for a single voice

A single *player*…

> You are essentially writing polyphonic material (albeit, one voice is playing 
> rests at any one time.)

That’s a valid, but non-intuitive, way of looking at it… I like your 
outside-the-box thinking! =)

> do you expect the MIDI regions to work for these parts?

I never use MIDI myself… but I would certainly expect (demand!) that any proper 
instrument-switching mechanism in Lilypond would Do The Right Thing™ viz-a-viz 
MIDI.

> you will need different MIDI channels, different midi instruments.

Note that one of the parameters of \addInstrumentDefinition is 
#’midiInstrument, i.e., the built-in mechanism already handles that.

> Where does it end, by trying to pretend one staff can satisfy the needs of 
> more than one instrument?

… as engravers have for hundreds of years!?  ;)

> I think that it is reasonable to ask that different voices be encapsulated in 
> separate variables.

As discussed in this thread, there are many ways of doing this.
Yours is yet another — and particularly clever — but it is no simpler (and in 
fact more complex in many ways) than others.

> why not just model it logically from the beginning?

Now *that* I agree with 100%.

> you compose without thought for what instrument is going to be played, then 
> pick and choose in after the fact.

No… I compose/arrange knowing *exactly* which instrument is going to be played, 
and that two or more of them will be played by a single player. (This is a 
standard technique in just about every musical genre: for example, in band and 
orchestral music, flute and clarinet players always have multiple instruments 
at the ready; in the music theatre pit, essentially every player except the 
strings play multiple instruments; in pop and jazz, the reeds often double, and 
the guitar almost always does; etc.)

> how do you choose what notes to write, if you are unsure what instrument will 
> play them?

As I said, you’ve misunderstood the process: I’m *always* sure ahead of time 
(i.e., at the moment of composition and/or arrangement) exactly which 
instrument will play the notes.

> As a multi-instrumentalist, I've seen way too many arrangements where the 
> composer has not put in enough thought into providing suitable rest to 
> actually switch instruments (most recently in das trunke lied). Or, not 
> indicating at the top of the part all the instruments that will be needed.  
> Or, omitting clefs or key signatures when there is a change, etc.

You obviously haven’t seen any of my music.  ;)

> When an instrumentalist has to switch, especially to an instrument in a 
> different transposition, they will necessarily have to think about what the 
> new key is.

Well, they’ll have to *note* what the new key is… “think” might be stretching 
it.

> at the minimum, the composer should be required to also think for a few 
> seconds and supply the current key signature, as well.

The *engraver* (not *composer*) should be required to do that… which is why 
Lilypond should do it automagically.

> a design that doesn't require any considerations leads to people never making 
> those considerations.  

That’s a logical fallacy.

> As such, I suggest an approach by modeling separate, well-defined variables 
> for each instrument, then use partcombine for the part.

I think it’s definitely worth considering, despite the [much] greater effort on 
the part of the engraver. Note that for the *composer* and/or *arranger*, there 
is no extra effort in (between) any of these methods: the music is always (at 
least for proper discussion's sake) already created and ready for engraving.

> I realize this does not fit your desired workflow, because you would have to:
>     Create separate variables for each distinct instrument

This is a little extra effort, yes… Though I have to do this from time to time 
anyway.

>     Fill up the parts with spacer rests besides the sections each instrument 
> is being played

Actually, \pushToTag would solve that.

>     Supply instrument change marks and key signatures at the point of 
> switching

That’s the part that should be done automagically by Lilypond.

> Besides these workflow issues, what things are not suitable with this 
> approach using partcombine and/or staff groups to model a multi-instrument 
> part?

Staff groups wouldn’t work: if there is a very fast switch, it would force a 
certain format (i.e., where the breaks lie) onto the score. This would 
unnecessarily mix content with presentation, which we should avoid (at least, 
we should avoid the design *forcing* the user to mix C&P).

Partcombine, on the other hand, is a very interesting option. I’ve been wanting 
to sponsor a Grand Unified Partcombine Project, Yay! (GUPPY) to fix the 
[myriad] problems with the current implementation — maybe this is the time to 
kill two birds with one stone! Because your suggestion doesn’t [fully] use 
\partcombine as originally intended/designed/coded (i.e., there will never be 
simultaneous notes), I’m not sure if any of its current idiosyncrasies will 
become stumbling blocks, or whether it’s ready to go as is; it will require 
investigation and testing.

In any case, thank you so much for a very eye-opening suggestion!
I look forward to playing around with your idea(s) and reporting back.

Best,
Kieren.
_______________________

Kieren MacMillan, composer
www:  <http://www.kierenmacmillan.info>
email:  address@hidden