Allow independent adjustment of minimum length for spanner siblings

Allow independent adjustment of minimum length for spanner siblings

The property minimum-length affects both unbroken and broken spanners, making
independent adjustment impossible.  This patch adds a property,
minimum-length-after-break, which, if set, behaves similarly to minimum-length,
but only affects siblings starting a line.  If minimum-length-after-break is
unset, minimum-length still controls all lengths.

A regtest demonstrates usage of the new property with a number of spanners.

[david.nalesnik, 2015-02-05 21:55:45]

Please review.

[thomasmorley651, 2015-02-05 22:31:34]

Can't review the C++ part.
Some nitpicks in the regtest.

Will it be possible to tweak the first part of the Tie etc at line-end as well?
With a combi of setting 'minimum-length-after-break' and 'minimum-length?
If yes, I'd show an example in the regtest. At least it should be
explained/demonstrated in the docs.

https://codereview.appspot.com/201140043/diff/1/input/regression/minimum-leng...
File input/regression/minimum-length-after-break.ly (right):

https://codereview.appspot.com/201140043/diff/1/input/regression/minimum-leng...
input/regression/minimum-length-after-break.ly:15: \override
Tie.minimum-length-after-break = 20
I'd use \once \override ...
Though, that's only me

https://codereview.appspot.com/201140043/diff/1/input/regression/minimum-leng...
input/regression/minimum-length-after-break.ly:16: a1~
I assume it works for chords as well. 
I'd add at least one example with chords and Tie or Glissando

[david.nalesnik, 2015-02-05 22:51:06]

https://codereview.appspot.com/201140043/diff/1/input/regression/minimum-leng...
File input/regression/minimum-length-after-break.ly (right):

https://codereview.appspot.com/201140043/diff/1/input/regression/minimum-leng...
input/regression/minimum-length-after-break.ly:15: \override
Tie.minimum-length-after-break = 20
On 2015/02/05 22:31:33, thomasmorley651 wrote:
> I'd use \once \override ...
> Though, that's only me

I'm happy to change that.  Would you use \once with the other overrides in the
file?

https://codereview.appspot.com/201140043/diff/1/input/regression/minimum-leng...
input/regression/minimum-length-after-break.ly:16: a1~
On 2015/02/05 22:31:34, thomasmorley651 wrote:
> I assume it works for chords as well. 
> I'd add at least one example with chords and Tie or Glissando

Instead of the one regtest, I could have two.

The first would show a tied chord, and I would show how you can use
'minimum-length and 'minimum-length-after-break in various combinations.

The second would be this regtest, and I could take out the tie example so
there's no overlap with the other test.

What do you think?

[thomasmorley651, 2015-02-05 23:34:22]

https://codereview.appspot.com/201140043/diff/1/input/regression/minimum-leng...
File input/regression/minimum-length-after-break.ly (right):

https://codereview.appspot.com/201140043/diff/1/input/regression/minimum-leng...
input/regression/minimum-length-after-break.ly:15: \override
Tie.minimum-length-after-break = 20
On 2015/02/05 22:51:06, david.nalesnik wrote:
> On 2015/02/05 22:31:33, thomasmorley651 wrote:
> > I'd use \once \override ...
> > Though, that's only me
> 
> I'm happy to change that.  Would you use \once with the other overrides in the
> file?

I'd always use \once. But again, it's only me, what do others think?

https://codereview.appspot.com/201140043/diff/1/input/regression/minimum-leng...
input/regression/minimum-length-after-break.ly:16: a1~
On 2015/02/05 22:51:06, david.nalesnik wrote:
> On 2015/02/05 22:31:34, thomasmorley651 wrote:
> > I assume it works for chords as well. 
> > I'd add at least one example with chords and Tie or Glissando
> 
> Instead of the one regtest, I could have two.
> 
> The first would show a tied chord, and I would show how you can use
> 'minimum-length and 'minimum-length-after-break in various combinations.
> 
> The second would be this regtest, and I could take out the tie example so
> there's no overlap with the other test.
> 
> What do you think?

I can't see an advantage in having two regtests. I'd prefer to extend this one.

[lemzwerg, 2015-02-06 09:03:41]

LGTM.

[david.nalesnik, 2015-02-06 19:17:27]

https://codereview.appspot.com/201140043/diff/1/input/regression/minimum-leng...
File input/regression/minimum-length-after-break.ly (right):

https://codereview.appspot.com/201140043/diff/1/input/regression/minimum-leng...
input/regression/minimum-length-after-break.ly:15: \override
Tie.minimum-length-after-break = 20
On 2015/02/05 23:34:22, thomasmorley651 wrote:
> On 2015/02/05 22:51:06, david.nalesnik wrote:
> > On 2015/02/05 22:31:33, thomasmorley651 wrote:
> > > I'd use \once \override ...
> > > Though, that's only me
> > 
> > I'm happy to change that.  Would you use \once with the other overrides in
the
> > file?
> 
> I'd always use \once. But again, it's only me, what do others think?

I went ahead and changed to \once.  The object being overridden changes, and the
Scheme music representation isn't any shorter, but \once might prevent problems
should the test be extended.

(Done.)

https://codereview.appspot.com/201140043/diff/1/input/regression/minimum-leng...
input/regression/minimum-length-after-break.ly:16: a1~
On 2015/02/05 23:34:22, thomasmorley651 wrote:
> On 2015/02/05 22:51:06, david.nalesnik wrote:
> > On 2015/02/05 22:31:34, thomasmorley651 wrote:
> > > I assume it works for chords as well. 
> > > I'd add at least one example with chords and Tie or Glissando
> > 
> > Instead of the one regtest, I could have two.
> > 
> > The first would show a tied chord, and I would show how you can use
> > 'minimum-length and 'minimum-length-after-break in various combinations.
> > 
> > The second would be this regtest, and I could take out the tie example so
> > there's no overlap with the other test.
> > 
> > What do you think?
> 
> I can't see an advantage in having two regtests. I'd prefer to extend this
one.

After discovering an issue (corrected in patch set #2), I'm leaning towards a
regtest specifically aimed at showing how minimum-distance and
minimum-distance-after-break relate.  They ought to produce the same results in
the spanner piece after the break.  The C++ code uses these values in a way that
is not easy to understand, and I think there should be a check for future
changes to the springs-and-rods callback.

[david.nalesnik, 2015-02-06 20:08:21]

https://codereview.appspot.com/201140043/diff/20001/lily/spanner.cc
File lily/spanner.cc (right):

https://codereview.appspot.com/201140043/diff/20001/lily/spanner.cc#newcode404
lily/spanner.cc:404: Rod r;
Perhaps we should call this rod something else to underscore that it's not the
same rod as the one declared above.  Sure, it's fine because the last r was
declared in a nested scope and "'r' is for rod!", but this is unnecessarily
confusing (even after moving the declaration from where it was originally, which
was really obscure).

[lemzwerg, 2015-02-06 21:47:11]

https://codereview.appspot.com/201140043/diff/20001/lily/spanner.cc
File lily/spanner.cc (right):

https://codereview.appspot.com/201140043/diff/20001/lily/spanner.cc#newcode394
lily/spanner.cc:394: /*
Minor nit: Please use spaces, not tabs.

[david.nalesnik, 2015-02-08 21:51:38]

https://codereview.appspot.com/201140043/diff/20001/lily/spanner.cc
File lily/spanner.cc (right):

https://codereview.appspot.com/201140043/diff/20001/lily/spanner.cc#newcode394
lily/spanner.cc:394: /*
On 2015/02/06 21:47:11, lemzwerg wrote:
> Minor nit: Please use spaces, not tabs.

Done.

[thomasmorley651, 2015-02-09 23:37:56]

LGTM