Re: Discuss signature for new function \annotate

[Urs Liska]

Am 06.06.2013 13:35, schrieb David Kastrup:

Urs Liska <address@hidden> writes:

Am 06.06.2013 12:22, schrieb David Kastrup:

If your first argument is something like Script, you might want to make
it of type symbol? instead.  I actually have a syntax sanitizing patch
in limbo where unquoted words are not accepted as strings.  It is not
going to make it into 2.18, but I don't promise I won't push it at some
later point of time.

What type is the 'grob' argument of \shape?

<URL:http://lilypond.org/doc/v2.17/Documentation/notation/available-music-functions#index-shape-1>


 shape [music] - offsets (list) item (symbol list or music)

    Offset control-points of item by offsets. The argument is a list of
    number pairs or list of such lists. Each element of a pair
    represents an offset to one of the coordinates of a
    control-point. If item is a string, the result is \once\override for
    the specified grob type. If item is a music _expression_, the result
    is the same music _expression_ with an appropriate tweak applied.

The documentation string states "If item is a string", but that's
actually inaccurate and would warrant fixing.

And: with \shape the order of arguments was reversed for syntactical
reasons (that I didn't really understand).

When commands with arguments are used as tweaks on a music _expression_,
multiple such commands combine quite awfully when the music _expression_
is not consistently the last argument.

I would also like to \annotate e.g. individual notes from chords
etc. So what would I have to take care of?

Putting an "item" of type symbol-list-or-music? last seems like a
reasonably good idea.

Hm, I think I see what you mean,
but if I write

annotate = 

#(define-void-function (parser location type properties annotation item)

   (string? list? string? symbol-list-or-music?))

\relative g' {

  \annotate 

    "critremark" 

    #'()

    "Tenuto added as in Vc. 2"

    Script

  g1-- 

}

I get an error "Zu viel Vorgriff" (too much read-ahead?).
If I replace 'Script' with '{ c }' for example it compiles without message.
???

It is probably worth considering to make properties of type context-mod?
(potentially optional), then you can write the argument as

\with {
    voice = "vc1"
    source = "Ms. 2"
    author = "Urs Liska"
    date = "2013-06-06"
}

That looks very good. How would the values then be accessed? Are they
simple variables inside the function? Or does that also create an
alist internally

Obviously I'm missing that extra bit outside your quote.

annotate = 

#(define-void-function (parser location type properties annotation item)

   (string? context-mod? string? symbol-list-or-music?))

\relative g' {

  \annotate "critremark" 

      \with {

         voice = "vc1"

         source = "Ms. 2"

         author = "Urs Liska"

         date = "2013-06-06"

       }

    "Tenuto added as in Vc. 2"

    { c }

  g1-- 

}

gives me loads of errors about not being in lyricmode ...

Lists.

address@hidden:/usr/local/tmp/lilypond$ lilypond scheme-sandbox
GNU LilyPond 2.17.20
Processing `/usr/local/share/lilypond/2.17.20/ly/scheme-sandbox.ly'
Parsing...
guile> (ly:get-context-mods #{ \with { voice = "vc1" source = "Ms. 2" } #})
((assign voice "vc1") (assign source "Ms. 2"))
guile> 

Of course, you might get unsets and overrides and reverts slipped in
which you would want to weed out again, but list? is also prone to
permitting a lot that you don't want to see.

Sure. I'd have to iterate over the alist and ignore all keys I can't
handle. But how would that be with a context-mod?

guile> (ly:get-context-mods #{ \with { \unset voice \override Beam.stencil = ##f} #})
((unset voice) (push Beam #f stencil))

Thanks for these

Urs