Hi Paolo,
Am 12.01.21 um 17:01 schrieb Paolo Prete:
Hello,
I just experienced (on Linux) that this compiles with <= 2.20
--------------------------------
foo = #(define-music-function (parser location col staff mus)((color?
red) string? ly:music?)
#{ c' d' #})
{
\foo "bar" {}
}
--------------------------------
But it doesn't compile with 2.22.0
What's wrong?
Last May, Valentin (in 1ea236291f96dbe5) added the ability to use
CSS-style color names (and hex color codes, and four-number RGBA colors
with alpha channel) everywhere where previously one had to use RGB-style
colors. (I only found out about this _great_ addition by researching
your problem, so I'm actually quite happy right now.)
For this, the color? predicate had to be weakened in order to cover more
possible data types: Where previously, we had
(define-public (color? x)
(and (list? x)
(= 3 (length x))
(every number? x)
(every (lambda (y) (<= 0 y 1)) x)))
in scm/output-lib.scm, it now reads
(define-public (color? x)
(or
(string? x)
(and (list? x)
(let ((l (length x)))
(or (= 3 l)
(= 4 l)))
(every number? x)
(every (lambda (y) (<= 0 y 1)) x))))
To wit, while a color? used to be a list of exactly three numbers in
[0,1], now it is _either_ a string _or_ a list of three or four numbers
in the same interval.
So, in your function definition, "bar" is now a valid color?, hence is
taken as "col", and consequently LilyPond tries to read {} as a string?.
I see two solutions for your problem:
a) If you only ever need rgb colors, you could define and old-style
color? predicate and use this in your function definition.
b) You could re-order the parameters to your function in such a way that
the next predicate after the optional color? is not a sub-predicate of
color?. One example would be ly:music?, so the following works (I omit
"parser"/"location"):
\version "2.22"
foo = #(define-music-function (col mus staff) ((color? red) ly:music?
string?)
#{ c' d' #})
{
\foo {} "bar"
}
Best
Lukas
