Don't treat context modification identifiers special

Don't treat context modification identifiers special

This requires using \with before context modifications like
\RemoveEmptyStaves in order to make the syntax get along
with fewer special cases and exceptions.


Also contains commits:

Run scripts/auxiliar/update-with-convert-ly.sh


convert-ly rule to put \with before context mods


\with #*unspecified* should be equivalent to \with { }

[dak, 2016-07-01 16:06:31]

Fix bad test

[thomasmorley651, 2016-07-02 08:50:10]

As always, I can't review parser-code.
Though, after discussion here
http://lilypond.1069038.n5.nabble.com/tiny-function-td192058.html
I vote for it.

A nitpick:

https://codereview.appspot.com/299460043/diff/1/input/regression/context-mod-...
File input/regression/context-mod-with.ly (right):

https://codereview.appspot.com/299460043/diff/1/input/regression/context-mod-...
input/regression/context-mod-with.ly:68: % Mods can be inserted instead of a
\with block (i.e. \with is not required)
This comment is wrong now.

[dak, 2016-07-02 09:09:11]

https://codereview.appspot.com/299460043/diff/1/input/regression/context-mod-...
File input/regression/context-mod-with.ly (right):

https://codereview.appspot.com/299460043/diff/1/input/regression/context-mod-...
input/regression/context-mod-with.ly:68: % Mods can be inserted instead of a
\with block (i.e. \with is not required)
On 2016/07/02 08:50:10, thomasmorley651 wrote:
> This comment is wrong now.

Shiver me timbers, well-spotted.

Will fix.  So one of the very few occurences of this construct in our code base
actually was chosen only for demonstration purposes.

[dak, 2016-07-02 09:27:13]

Fix comments in context-mod-with regtest

[Dan Eble, 2016-07-02 14:19:31]

LGTM, though I rarely work with yacc.  Requiring \with is a good idea.

I don't appreciate the value of \with #*unspecified*. Does it parallel a feature
that was already used in context mod lists?

I glossed over the convert-ly change.

[dak, 2016-07-02 14:52:57]

On 2016/07/02 14:19:31, Dan Eble wrote:
> LGTM, though I rarely work with yacc.  Requiring \with is a good idea.
> 
> I don't appreciate the value of \with #*unspecified*. Does it parallel a
feature
> that was already used in context mod lists?

If there is need for discussion, I can move this change into a separate issue. 
It is indeed almost entirely orthogonal to this issue here.  To be honest, it
has been sitting in a branch of its own since, uh, September 2013.  The
motivation to finally add it to a submission was that we have a number of
situations where

#(if condition expression)

works and is silently ignored when the whole of it evaluates to *unspecified*
(which it will when the condition is false).  For example, you could previously
have written

\with { ... #(if ... #{ \with { ... } #}) ... }

and it would have done nothing if the condition evaluated to false.  However, if
you now add redundancy as

\with { ... \with #(if ...) ... }

it would stop working.  Since the motivation for adding \with to context
modification expressions has now become a bit stronger, I had somewhat more
motivation to extend the "unspecified means nothing" pattern here.

> I glossed over the convert-ly change.

It's a bit fishy since it first tries to find context modification definitions
obeying a certain pattern before trying to replace all \with-less context
modifications to explicit context starts.  Obviously, there is a whole lot of
ways of defining context modifications, and definition and use need not happen
in the same file.

And, uh, this does not catch context modifications added to \chords, \lyrics,
\addlyrics etc.  Uh.  I think I need to extend that rule, come to think of it. 
I'll do that.  Which should give some time to see whether people would rather
have the \with #*unspecified* change forked out to facilitate separate
discussion and/or removed altogether.

[dak, 2016-07-02 15:26:30]

Add other context starters to convert-ly rule

[Dan Eble, 2016-07-02 15:26:43]

On 2016/07/02 14:52:57, dak wrote:
> On 2016/07/02 14:19:31, Dan Eble wrote:
> > I don't appreciate the value of \with #*unspecified*. Does it parallel a
> feature
> > that was already used in context mod lists?
> 
> If there is need for discussion, I can move this change into a separate issue.

> It is indeed almost entirely orthogonal to this issue here.

No, there's no problem.  I was using "appreciate" in the sense of understanding
rather than gratitude.  Thanks for the explanation.