Re: Auto-beaming infrastructure redo

[Hans Aberg]

On 3 Jul 2010, at 12:37, Carl Sorensen wrote:

> > > > If I throw in tuplets in this AST, setting them to 1:1, I get:
> > > >         +
> > > >         |
> > > >    (1:1, [(2, 1/8), (3, 1/8), +, (3, 1/8)])
> > > >                         |
> > > >                (1:1, [(2, 1/8), ((3, 1/8), (2, 1/8)])
> > > >
> > > > That is, the tuplet p:q just appears as a pair (p, q) attached to  
> > > > the
> > > > list.
> > >
> > > But we don't need to introduce the tuplets in the LilyPond  
> > > structure,
> > > because the p:q for tuplets is already handled by LilyPond, as far
> > > as I can
> > > tell.  At least for right now, I think we can use a single
> > > fundamental unit.
> >
> > I think you may need to have the tuplets as a part of this structure
> > in order to compute the correct beaming. Then the current flexible
> > approach of LilyPond where they can be broken up in pieces might be a
> > complication.
>
> Can you give me a specific example of beaming where the tuplets  
> would be
> necessary?

If one has say a triplet, then I think LilyPond admits breaking it up  
into to separate parts with some non-triplet notes between - LilyPond  
just checks that the timing is correct.

When one wants to beam correctly, a complete triplet group, as above,  
is also a specification that the first not should have the stem top  
the right only and the last one to the left only.

So I suspect that for proper beaming, one has to first make complete  
tuplet groups. LilyPond might need some new syntax to ensure that this  
information is provided or structure enforced.

  Hans