Issue 4516: Make \offset handle unpure/pure containers

Issue 4516: Make \offset handle unpure/pure containers

I have no clue whether this does what is expected from it.  Not really
understanding \offset, I cannot come up with reasonable test cases.
The linked example from the mailing list does not appear to do
anything.  As opposed to before, it does not print warnings, however.

[david.nalesnik, 2015-07-22 15:20:18]

Thanks for doing this.  The restructuring is a big improvement.

https://codereview.appspot.com/250590043/diff/1/scm/music-functions.scm
File scm/music-functions.scm (right):

https://codereview.appspot.com/250590043/diff/1/scm/music-functions.scm#newco...
scm/music-functions.scm:2331: (offset-multiple-types vals (car offsets)))))
Yup--missing parenthesis.

https://codereview.appspot.com/250590043/diff/1/scm/music-functions.scm#newco...
scm/music-functions.scm:2337: (ly:unpure-pure-container
(ly:make-unpure-pure-container

https://codereview.appspot.com/250590043/diff/1/scm/music-functions.scm#newco...
scm/music-functions.scm:2342: vals))))
(extra paren)

[david.nalesnik, 2015-07-22 15:56:18]

Here's two snippets that show \offset working with unpure-pure containers, one
with the override form, the second using the tweak form.

{
 c'4\f\<
 c'4\!
 \offset Y-offset -2 DynamicLineSpanner
 c'4\f\<
 c'4\!
 c'2
 \offset length 2 Stem
 c'2
}

{
 c'4\f\<
 c'4\!
 c'4\f-\offset DynamicLineSpanner.Y-offset -3 \<
 c'4\!
 c'2
 \offset Stem.length 2
 c'2
}

[david.nalesnik, 2015-07-22 16:20:48]

Yet compare the output of

{
  \offset Y-offset #-5 DynamicLineSpanner
  \repeat unfold 200 {
    c'2\f\< c'2\!
  }
}

with

{
  \override DynamicLineSpanner.Y-offset = #-15
  \repeat unfold 200 {
    c'2\f\< c'2\!
  }
}

Shouldn't the offset to the result of calling the pure function result in better
vertical spacing?

[dak, 2015-07-22 16:59:14]

On 2015/07/22 16:20:48, david.nalesnik wrote:
> Yet compare the output of
> 
> {
>   \offset Y-offset #-5 DynamicLineSpanner
>   \repeat unfold 200 {
>     c'2\f\< c'2\!
>   }
> }
> 
> with
> 
> {
>   \override DynamicLineSpanner.Y-offset = #-15
>   \repeat unfold 200 {
>     c'2\f\< c'2\!
>   }
> }
> 
> Shouldn't the offset to the result of calling the pure function result in
better
> vertical spacing?

Oh, this doesn't work yet: looks nice but I did not think it through correctly. 
And I'm still puzzling this one out: creating an unpure-pure-container that will
contain closures recognizing the unpure-pure-container's identity in the
property chain is actually murderously non-trivial.  Since once you created the
unpure-pure-container (and only then is its identity established), you can no
longer change its contents.

I still got a scheme going that should work for this.  But wrapping the basic
mechanism I'm going to use in code that is reasonably straightforward is quite a
challenge.

[dak, 2015-07-23 07:18:07]

On 2015/07/22 16:59:14, dak wrote:

> Oh, this doesn't work yet: looks nice but I did not think it through
correctly.

To wit: this needs to produce a value for the override at override time but will
only be able to decide about the kind of value to offset at runtime.  That means
it _always_ needs to return the most generic kind of value, namely an
unpure-pure-container.

[dak, 2015-07-23 08:40:09]

This was a lot harder than it appears.

[david.nalesnik, 2015-07-23 15:25:41]

LGTM.

Thanks again!