Re: Chromatic clashes

[Jean Abou Samra]

Hi,


Le 17/04/2021 à 14:32, Lukas-Fabian Moser a écrit :
> After posting my engraver yesterday, I noticed some unforeseen 
> behaviour that gave me an idea for a possible addition to your 
> excellent tutorial on https://extending-lilypond.readthedocs.io/ 
> (which I unfortunately only today had time to skim through). Namely:
>
> The engraver I posted yesterday collects notes across all voices or 
> staves, for instance in situations like
>
> <<
>   \new Staff { c' }
>   \new Staff { cis' }
> >>
>
> While this might actually be desirable (at least I tell my students 
> never ever to produce
>
> [image]
>
> since nobody knows if the clash is intended or not), I was surprised 
> that it did so. That's how I found out that if I wrap the engraver in 
> a (lambda (ctx) .... ) construct, then it seems to work independently 
> for each voice context, even if it's \consist'ed to all Voices or 
> Staves in the \layout {} block.
>
> Maybe that would be an interesting addition to your engraver tutorial?

Well, blow me down! I never realized that you could use an engraver 
alist directly instead of a function returning the alist, with the 
effect you observed.

Thanks! I'll add some words about this.

As a side note, all engravers defined by LilyPond are like the ones made 
with (lambda (context) ...) (the majority is defined in C++ classes). I 
don't know of a use case where the form without the lambda is necessary 
-- I guess you'd rather consist your engraver in Score, so process-music 
and stop-translation-timestep are run just once for example.

> And while we're at it, here's another question: It turns out it's 
> possible to add note-event properties like force-accidental inside the 
> process-music hook. But when I tried to change the _pitch_ of a note 
> at this stage, there was no effect. I _can_ change the pitch of a 
> note-event in the note-head listener, but that's too early in my use 
> case (since I need to know all simultaneous pitches).
> So, in which parts of an engraver do I still have the opportunity to 
> change pitches?
That's the problem with engravers. They are a modular system, which 
implies strong constraints in order to stay insensitive to the order of 
engravers as well as the order of events. What is happening here is that 
the Note_heads_engraver is reading the pitch in process-music before 
your engraver has had a chance to change it.

The best you can do is set the pitches on all events every time you get 
one. For an example in a very similar situation, take a look at the code 
here:

https://lists.gnu.org/archive/html/lilypond-user/2021-04/msg00025.html

This stores all note events with their original pitches, and naturalizes 
them; when a new key arrives, they would be all re-naturalized using the 
original pitches. ("Would" because I couldn't actually find a case where 
the key comes after the notes.) This is not very pretty, but I don't 
know a better solution.

And if you were implementing something in LilyPond directly (not as a 
user extension), you'd probably go change the Note_heads_engraver itself.

Cheers,
Jean