Re: grob-object information

[Marc Hohl]

Am 11.09.2012 21:54, schrieb address@hidden:
> On 11 sept. 2012, at 22:29, Marc Hohl <address@hidden> wrote:
>
> > > Are you certain that every vertical axis group will always contain the same 
> > > number of bar lines?  If not, it's possible that the matrix you're talking 
> > > about may not have complete rows, in which case it's difficult to know how to 
> > > break it.
> > I am not 100% sure – there could be corner cases where a staff line stops 
> > before the
> > volta bracket is closed, but this could lead to an error when compiling the .ly 
> > file.
> What about pieces with different simultaneous time signatures?  They'd have 
> different numbers of bars for a given section.
Hey, good catch!
> > > One way to check it is to look at the left or right of the spanned_rank_interval of the 
> > > grobs (the values will be the same as you're dealing with items which only ever 
> > > "span" one column).  Once it goes backwards, you know that you've skipped to 
> > > the next row.
> > What kind of values does spanned_rank_interval return?
> The column number of a grob's left and right bound. Columns are linearly 
> ordered from the beginning of a score to the end, so you can use it to know 
> when a new staff is beginning in your array.  For bar lines, the grob only 
> spans one column so the left and right values of the returned will be the same 
> - you can use either one.
Ok, so I try to wrap a scheme caller around spanned_rank_interval and see
how far I can get with that.
> In C++, if I had a vector <Grob *> foo that I wanted to sort into vector <vector 
> <Grob *> > bar, if I understand you correctly, I'd do:
>
>
> int current_column = INT_MAX;
>
> for (vsize i = 0; i < foo.size (); i++)
>   {
>     int column = foo[i]->spanned_rank_interval ()[LEFT];
>     if (column <= current_column)
>       bar.push_back (vector<Grob *> ());
>     bar.back ().push_back (foo[i]);
>     current_column = column;
>   }
>
> The same thing is doable in Scheme - I just write pseudo-code faster in C++.
Yup, I think that's more or less what I want to achieve.

Just a side information: In lily/volta-bracket.cc the bar linformation
is obtained by

extract_grob_set (me, "bars", bars);
Grob *endbar = bars.size () ? bars.back () . 0 ;
SCM glyph = endbar ? endbar->get_property ("glyph_name") : SCM_EOL;

The scheme equivalent is

(let* ((bar-array (ly:grob-object me 'bars))
(bar-array-length (ly:grob-array-length bar-array))
       (endbar (ly:grob-arraybar-array (1- bar-array-length)))
       (glyph (ly:grob-property endbar 'glyph-name))

modulo some checks whether the array is not empty etc.
So the last element in the array is the one used for the
volta bracket, isn't it?

I found out that the last element in bar-array can be a SpanBar grob,
which leads to problems when I want to extract some other properties
that are only defined for BarLine grobs and not for SpanBar grobs.

I don't know whether it is safe just to check whether the last
element is a SpanBar; if yes, take the predecessor, if no, it
is a BarLine grob and erverything is fine.

Hopefully I explained it in a way that's understandable;
I want to know if I can achieve what I need *without* kind of
"parsing" the array.
On the other hand, when I need more detailed layout informations
about the "staffs and stuff", there is probably no way to bypass
this ordering routine.

Regards,

Marc