Add support for multi-note tremolo

Add support for multi-note tremolo

-) Remove all checks for exactly one or two notes as arguments to a
   tremolo repeat
-) Create proper beaming for multi-note tremolo (so far, the
   chord-tremolo-engraver always assumed exactly two beamed notes)
-) Extend the make-repeat function to scale the notes properly also
   in the case of multi-note tremolos

[Neil Puttock, 2010-07-09 20:19:11]

Hi Reinhold,

This will be a great enhancement if you can sort out the forbidden breaks
between tremolos.

Cheers,
Neil

http://codereview.appspot.com/1786041/diff/1/2
File input/regression/repeat-tremolo-three-notes.ly (right):

http://codereview.appspot.com/1786041/diff/1/2#newcode4
input/regression/repeat-tremolo-three-notes.ly:4: texidoc = "A tremolo can have
more then two notes."
than

http://codereview.appspot.com/1786041/diff/1/2#newcode9
input/regression/repeat-tremolo-three-notes.ly:9: \new Staff \relative c' {
indent

http://codereview.appspot.com/1786041/diff/1/3
File lily/chord-tremolo-engraver.cc (right):

http://codereview.appspot.com/1786041/diff/1/3#newcode60
lily/chord-tremolo-engraver.cc:60: Grob *previous_grob_;
Since it's always a stem, why not call it previous_stem_ or last_stem_?

http://codereview.appspot.com/1786041/diff/1/3#newcode87
lily/chord-tremolo-engraver.cc:87: announce_end_grob (beam_,
previous_grob_->self_scm ());
This is too late to be acknowledged by the Spanner_break_forbid_engraver before
the next timestep, which means breaks are always forbidden between consecutive
tremolos.

http://codereview.appspot.com/1786041/diff/1/3#newcode130
lily/chord-tremolo-engraver.cc:130: Stem::set_beaming (previous_grob_, flags,
RIGHT);
mixing tabs and spaces

[Reinhold, 2010-07-10 09:51:33]

http://codereview.appspot.com/1786041/diff/1/3
File lily/chord-tremolo-engraver.cc (right):

http://codereview.appspot.com/1786041/diff/1/3#newcode87
lily/chord-tremolo-engraver.cc:87: announce_end_grob (beam_,
previous_grob_->self_scm ());
On 2010/07/09 20:19:11, Neil Puttock wrote:
> This is too late to be acknowledged by the Spanner_break_forbid_engraver
before
> the next timestep, which means breaks are always forbidden between consecutive
> tremolos.

Hmm, but that's the first event where I know that the tremolo has ended. Before
the stopping spanner event, there is no way to determine whether there will be
still some notes or not...

(BTW, this also needs to be wrapped inside a if (previous_grob_) { ... } to
prevent crashed on empty tremolo spanners)

http://codereview.appspot.com/1786041/diff/1/3#newcode128
lily/chord-tremolo-engraver.cc:128: {
Would it be a problem to use announce_end_grob here like before, even though
there might be another note coming up?
I.e. what side-effects does calling announce_end_grob have?

[Neil Puttock, 2010-07-11 17:02:42]

On 2010/07/10 09:51:33, Reinhold wrote:

> Hmm, but that's the first event where I know that the tremolo has ended.
Before
> the stopping spanner event, there is no way to determine whether there will be
> still some notes or not...

It's an unavoidable consequence of the iteration process, since the
TremoloRepeatedMusic triggers the creation of the start/stop events; unlike the
events generated by manual beaming, they're always running behind.  Naturally
this doesn't matter if we're restricted to two-note tremolos since we know once
a second stem's been acknowledged that the beam should end.
 
> Would it be a problem to use announce_end_grob here like before, even though
> there might be another note coming up?
> I.e. what side-effects does calling announce_end_grob have?

AFAICT, it's only used by the Spanner_break_forbid_engraver, so will only affect
beams in a stem tremolo.  But it will have the unfortunate side effect of making
every tremolo longer than two notes breakable (ignoring 'breakable = ##f) after
the second stem.

Cheers,
Neil