Re: convert-ly for note/rest-markup

[David Kastrup]

Michael Käppler <xmichael-k@web.de> writes:

> Am 03.02.2021 um 09:05 schrieb Michael Käppler:
>> Hi Harm,
>>
>> Am 02.02.2021 um 22:52 schrieb Thomas Morley:
>>> Hi,
>>>
>>> with e35b7959dfe the note-markup-command was changed to need a
>>> duration-argument, before it was a string.
>>> With, MR 627 "Revisit rest-markup-commands"
>>> https://gitlab.com/lilypond/lilypond/-/merge_requests/627
>>> I attempt to do similar for the rest-markup-command.
>>>
>>> This change would need a convert-rule, imho.
>>> Alas, this is beyond my capabilities.
>>>
>>> Any volunteer?
>>>
>>> Btw, e35b7959dfe contains a convert-rule for the note-markup, though
>>> it does not seem to work.
>> I had a quick look on this, the problem is not in e35b7959dfe,
>> the regex 'matchfullmarkup' (see convertrules.py:3471) which is used in
>> said commit assumes that \markup is followed by at least one set of
>> braces.
>>
>> You can verify this by converting
>>
>> %%%%%%%%%%%%%%
>> \version "2.18.2"
>> \markup { \note #"4." #UP }
>>
>> %%%%%%%%%%%%%%
>>
>> which works as expected.
> Unfortunately, I was wrong. The problem is a more fundamental one.
> 'matchfullmarkup' tries to match the full markup, but nothing more than
> that.
> In other words, it has to make assumptions where the \markup environment
> ends.
> In a case like
>
> \markup { ... } this is simple, because it can simply "wait" for the
> closing brace.
>
> However, when \markup is followed by a single markup command or a
> sequence of markup commands, like
>
> \markup \center-align \fontsize #2 "foo"
>
> this does not work. Instead it assumes that the \markup environment ends
> with a string or another set of balanced braces.
> For markup commands that do not take a markup as last argument, this
> will not work.
> The problem is not limited to \note, it would affect commands like
> \beam, \draw-circle etc. as well, if syntax conversions
> would have to be added to 'convert-ly'.
>
> AFAICT the parser solves this problem by checking the call signature of
> the markup command to
> know how many arguments should be there. (and of which type)
> Since \markup commands are user-definable, I don't have any idea how to
> solve this in a general way
> within convert-ly.
>
> Cc'ing David, maybe you have some inspiration for us?

Short of special-patterning the known exceptions...  a bit of a
nightmare.

-- 
David Kastrup