Prevents script tie collisions

Prevents script tie collisions

[Keith, 2012-09-04 05:36:53]

I can't get it to work for the script at the start of the tie.
\relative c''' { r2. c4~-> | c-> r2. }

http://codereview.appspot.com/6500058/diff/1/input/regression/script-tie.ly
File input/regression/script-tie.ly (right):

http://codereview.appspot.com/6500058/diff/1/input/regression/script-tie.ly#n...
input/regression/script-tie.ly:10: r2. c4 ~ |
r2. c4-> ~

http://codereview.appspot.com/6500058/diff/1/lily/script-engraver.cc
File lily/script-engraver.cc (right):

http://codereview.appspot.com/6500058/diff/1/lily/script-engraver.cc#newcode54
lily/script-engraver.cc:54: set<Grob *> ties_;
Ties only affect the note at their beginning and at their end, so I do not see
the need to remember them.   Can't you just add them to the support of a script
immediately upon acknowledging their beginnings and ends?

http://codereview.appspot.com/6500058/diff/1/lily/script-engraver.cc#newcode226
lily/script-engraver.cc:226: ties_.erase (ties_to_remove_[i]);
It looks, to me, like you erase any tie ending on this note /before/ you tell
any scripts on this not about the tie.
However, it works at this end of the tie so I must be misunderstanding
something.

[mike7, 2012-09-04 08:14:46]

On 4 sept. 2012, at 07:36, k-ohara5a5a@oco.net wrote:

> I can't get it to work for the script at the start of the tie.
> \relative c''' { r2. c4~-> | c-> r2. }
> 

The reason is because ties are announced at a timestep (maybe several) after
their creation.  This makes it very difficult to account for scripts created on
the first tied note.

What would be best is for the tie engraver to announce ties straightaway and
then to call Grob::suicide if they wind up not being used instead of announcing
them later.  It leads to a proliferation of more grobs but as they're suicided
before the preprocessing of graphical objects they don't eat up computation
time.  This is how flags are created in the Stem_engraver - they don't know if
they'll have a beam until the Auto_beam_engraver finishes off a beam, so we
create more than necessary and kill off the unused ones later.

> 
> http://codereview.appspot.com/6500058/diff/1/input/regression/script-tie.ly
> File input/regression/script-tie.ly (right):
> 
>
http://codereview.appspot.com/6500058/diff/1/input/regression/script-tie.ly#n...
> input/regression/script-tie.ly:10: r2. c4 ~ |
> r2. c4-> ~
> 
> http://codereview.appspot.com/6500058/diff/1/lily/script-engraver.cc
> File lily/script-engraver.cc (right):
> 
>
http://codereview.appspot.com/6500058/diff/1/lily/script-engraver.cc#newcode54
> lily/script-engraver.cc:54: set<Grob *> ties_;
> Ties only affect the note at their beginning and at their end, so I do
> not see the need to remember them.   Can't you just add them to the
> support of a script immediately upon acknowledging their beginnings and
> ends?
> 

Good call.

>
http://codereview.appspot.com/6500058/diff/1/lily/script-engraver.cc#newcode226
> lily/script-engraver.cc:226: ties_.erase (ties_to_remove_[i]);
> It looks, to me, like you erase any tie ending on this note /before/ you
> tell any scripts on this not about the tie.
> However, it works at this end of the tie so I must be misunderstanding
> something.

See above - it's a problem of the moment of announcement versus the actual
timestep where the grob's creation happens.

> 
> http://codereview.appspot.com/6500058/
> 
> _______________________________________________
> lilypond-devel mailing list
> lilypond-devel@gnu.org
> https://lists.gnu.org/mailman/listinfo/lilypond-devel

[Keith, 2012-09-05 06:35:29]

On 2012/09/04 08:14:46, mike7 wrote:
> On 4 sept. 2012, at 07:36, mailto:k-ohara5a5a@oco.net wrote:
> 
> > I can't get it to work for the script at the start of the tie.
> > \relative c''' { r2. c4~-> | c-> r2. }
> > 
> 
> The reason is because ties are announced at a timestep (maybe 
> several) after their creation.  This makes it very difficult to 
> account for scripts created on the first tied note.

Then just forget about it.  When was the last time you saw an accent on the
/second/ note of a tied pair?  or a staccato!?

Or, do it as you say here:

> What would be best is for the tie engraver to announce ties 
> straightaway and then to call Grob::suicide if they wind up 
> not being used instead of announcing them later.

[Keith, 2012-09-06 04:36:09]

Works in realistic cases, and uses less code.
What's not to love?