Re: TupletBracket.shorten-pair with strange output

[Thomas Morley]

2017-04-09 22:38 GMT+02:00 David Nalesnik <address@hidden>:
> Harm,
>
> On Sun, Apr 9, 2017 at 1:53 PM, Thomas Morley <address@hidden> wrote:
>> The following gives strange output, see attached.
>>
>> \version "2.19.58"
>>
>> {
>>     \override TupletBracket.shorten-pair = #'(10 . 0)
>>     \tuplet 1/1 {
>>         c'1 c'1
>>     }
>> }
>>
>> I would have expected the whole bracket to be (much) smaller, instead
>> only the part of the bracket left from TupletNumber is affected.
>>
>
> This behavior is ugly but expected.  You're requesting that the left
> end of the bracket be moved 10 staff-spaces to the right and the right
> end be moved 0 ss.
>
> The way to shorten the bracket would be something like:
>
>  \override TupletBracket.shorten-pair = #'(2 . 2)
>

>
> HTH,
> David

Hi David,

here a better test-code for the topic:

\new Staff {
  $@(map
      (lambda (i)
        #{
        \override TupletBracket.shorten-pair = #(cons i 0)
        \tuplet 1/1 { c'1 c'1 }
        #})
      (iota 7))
}

> Maybe there should be a cap on the amount that a bracket can be
> shortened to prevent the inversion in your example?  The bracket
> either won't shorten any more or that part of it disappears?

I'd second this.

But maybe it's an documentation-issue as well.

shorten-pair description is:
"The lengths to shorten on both sides
a hairpin or text-spanner such as a pedal bracket.  Positive values
shorten the hairpin or text-spanner, while negative values lengthen
it."
Probably we should add/document it's usage is suitable for fine-tuning
not for heavy changes. For that pupose one could use X-positions.

Here a test-code for X-position:

\new Staff {
  $@(map
      (lambda (i)
        #{
        \override TupletBracket.after-line-breaking =
        #(lambda (grob)
          (let ((x-pos (ly:grob-property grob 'X-positions)))
            (ly:grob-set-property! grob 'X-positions
              (cons (+ (car x-pos) i) (cdr x-pos)))))
        \tuplet 1/1 { c'1 c'1 }
        #})
      (iota 7))
}

If you change it to use (iota 8) an assertion failure occurs:

lilypond: /home/harm/lilypond-git/flower/include/interval.hh:227: T
Interval_t<T>::center() const [with T = double]: Assertion `!is_empty
()' failed.
Aborted (core dumped)

Though, it's beyond my depth to fix it in a manner a more useful
error-message is thrown.

Thanks,
  Harm