Re: Approximates cross-staff slurs in VerticalAxisGroup vertical-skylines. (issue 6498077)

[address@hidden]

On 5 sept. 2012, at 00:33, address@hidden wrote:

> 
> On 4 sept. 2012, at 17:45, address@hidden wrote:
> 
>> On 2012/09/02 06:25:58, MikeSol wrote:
>>> On 2012/09/01 23:58:37, Keith wrote:
>>>> I might have a test case for you at
>>>> http://www.mutopiaproject.org/cgibin/piece-info.cgi?id=1776
>>>> 
>>>> It seems you copy each slur into a "slur-stub", and from those keep
>> only the
>>>> ones with "cross-staff" set. Then when figuring system skylines you
>> insert all
>>>> Grobs with the slur-stub-interface into the skylines for each staff.
>>>> 
>>>> Why not insert the original Slur into the skylines, if it has
>> "cross-staff"
>>>> instead of the SlurStub.  For that matter, why not insert all Grobs
>> marked
>>>> "cross-staff"?
>>>> 
>>>> all the cross-staff
>> 
>>> It's not a copy of the original slur because it is using pure heights
>> and
>>> offsets.
>> 
>> I realize I'm a bit late to this party, but why do you need a separate
>> grob instead of just a Slur::get_maybe_pure_vertical_skylines function?
>> 
>> 
>> http://codereview.appspot.com/6498077/
> 
> Welcome to the party!
> 
> We're on the same page - that's exactly what I've been working on over the 
> past day-ish in my spare time.  I don't have a clean result yet as it'd 
> require some trickery with finding minimum translations of VerticalAxisGroups 
> by hand (if the stubs belong to different vertical axis groups this is done 
> automatically when grob offsets are calculated).  I'm almost there though - 
> I'll post a patch once this is done.
> 
> We need to come up with a term for this that's not `pure' as `pure' means in 
> LilyPond jargon "anything that does not result in a set_object, set_property, 
> suicide, the calling of the SpacingSpanner or the calling of a 
> VerticalAlignment."  This would result in the calling of the spacing spanner 
> were it called further upstream, so we need a different thing to call it.
> 
> Cheers,
> MS

After further reflection, the idea of stashing the callback in a single skyline 
doesn't work :-(

If they are part of a single VerticalAxisGroup, they will force tons of space 
between the VerticalAxisGroup to which the skyline belongs and whatever axis 
group falls immediately above or below.  You'll notice that I set the bottom of 
the top skyline and the top of the bottom skyline to empty so that this doesn't 
happen.  If you remove this in Slur::extremal_stub_vertical_skylines, you'll 
see huge vertical shifts in the regtest and other scores.

I know this is supposed to be a comment as per Graham's suggestion but out of 
force of habit I wrote it in an e-mail.  Will write up as a comment.

Cheers,
MS