Re: Give slurs skylines in outside-staff-priority calculations. (issue 5504055)

[Janek Warchoł]

2011/12/28 address@hidden <address@hidden>:
>
> On Dec 25, 2011, at 10:50 PM, Janek Warchoł wrote:
>> My (quick) question is: would be easy to apply your solution to other
>> objects?  I know that lyrics and dynamics really need improvement in
>> this area too - see this example: http://www.sendspace.com/file/dm4ua2
>> (source is attached to pdf).
>
> Vertical skylines are useful for any object whose height changes
> appreciably over its horizontal span.  Slurs (sometimes) fall into this 
> category.
> Accidentals fall into this category as well and have been dealt with similarly
> (I don't know who wrote the code, but there is code in accidental.cc
> to get a boxed approximation of an accidental's form).

Indeed; i've seen that code.  However, it could use some improvement,
as can be seen in
http://code.google.com/p/lilypond/issues/detail?id=2141 .

> I am not sure that dynamics and lyrics have subtleties in their shape
> that could be better approximated by the use of boxes (perhaps dynamics
> more than lyrics as they are slanted, but I'm dubious of this).

There might be a better solution, but they could be done better even
with subdivided boxes, as can be seen by examining the example i
linked to.

> To fix the problem, it'll be necessary to cook up a few minimal examples
> showing each problem (ideally with lines that can be commented out to trigger 
> the problem).

Done!  Examples are attached to issue 2148
http://code.google.com/p/lilypond/issues/detail?id=2148
By the way, i'm attaching examples in pdf format rather than png,
because this gives me unlimited resolution (and i can attach Lily
source code inside pdf).  Do you like it or are pngs better?

> Once you have this, we'll be able to find the cause rather quickly.
> I suspect that the issue arises from too-generous pure height approximations
> and/or oddities in inter-staff vertical spacing, so boxes are likely not the 
> solution.

Um, i think the cause is simpler than that :)


2011/12/28 Werner LEMBERG <address@hidden>:
>
>> I am not sure that dynamics and lyrics have subtleties in their
>> shape that could be better approximated by the use of boxes (perhaps
>> dynamics more than lyrics as they are slanted, but I'm dubious of
>> this).
>
> You might also have a look at issue #1097.  Should be almost trivial
> for you :-)

It would be great to fix this too!
You know, one thing comes to my mind: should we really use boxes for
all sorts of collision calculations?  I think it would give much
better results it we calculated real distance between objects (i.e.
between black areas).  We know the outlines of font glyphs, the real
shape of slur...  Could we implement a global "metric" for measuring
distances between objects?  My friend Łukasz (Luke) thinks this should
be the way to go.


W dniu 28 grudnia 2011 13:01 użytkownik Carl Sorensen
<address@hidden> napisał:
> Looking at Janek's examples, there are places where a hairpin could go closer 
> to the notes
> if it didn't have a rectangular skyline the size of the large end. There are 
> also places
> that the lyrics could go closer if the extended used a skyline that had its 
> tiny height
> instead of the character box height.

Exactly.


2011/12/28 Reinhold Kainhofer <address@hidden>:
> Another situation is that center-aligned barnumbers (right-aligned
> barnumers simply look ugly for >100 bars) are shifted upwards by the
> rectangular box of the treble clef, even though no collision would occur
> at all.

Thank you for a nice example, i've used it in issue 2148.

cheers,
Janek