On 2021-02-11 2:54 pm, Michael Käppler wrote:
Am 11.02.2021 um 23:39 schrieb Aaron Hill:
On 2021-02-11 2:04 pm, David Kastrup wrote:
Short of special-patterning the known exceptions... a bit of a
nightmare.
Does the pattern matching need to be so particular about context?
Consider a primitive approach:
re.sub(r'\\note\s*#"([^"]+)"', r"\\note { \g<1> }", ...)
This matches the `\note #"..."` pattern, converting to `\note { ...
}`.
I agree that this should work for most cases, but consider that it
can't
be taken for
granted that '\note' refers to the markup function:
%%%%%%%%%%%%%%%%
note = { c4 }
\markup \note #"4." #UP
\note #"4." #UP
%%%%%%%%%%%%%%%%
I don't want to say it makes any sense, but it is valid input...
Maybe "valid" is not the best litmus test. The second line is music
followed by a string and a number at the top level. If I did not
actually test this with 2.22.0 and see it work, I would have told you
that would fail to compile. Seems the parser consumes the
Scheme-escaped string and number with no complaints, but there appears
to be no purpose to such a construct.
I apologize if I am being flippant; but permit me to point out that in
your example, if we apply the primitive conversion rule, the result is
still valid input:
%%%%
note = { c4 }
\markup \note { 4. } #UP
\note { 4. } #UP
%%%%
Only difference is an additional staff and note appears, so the useless
string before actually does something.
At the end of the day, I would argue convert-ly should not attempt to
cater to the exotic. `\note #"..."` is distinctive enough a pattern
that broadly it will not be misinterpreted. Certainly, the pattern
could be expanded to include the second argument, which makes false
matches less likely. Additionally, you could go as far as validating
the string contains a suitable duration. But there is always a risk of
overdesigning the regular expression, wasting engineering and execution
resources for no realized gain.
-- Aaron Hill
