Hi,
this question is related to my last merge request
https://gitlab.com/lilypond/lilypond/-/merge_requests/1611. To my surprise,
there seem to be strong objections against it, despite being an improvement
in all tests. I wanted to know more about penalties in
slur-configuration.cc in general. The idea for tweaking the penalty in the
case of the presence of dots is not mine at all. In the function
score_edges, amongst others, we find that the penalty (named "demerit"
here) for a specific case involving stems is tweaked so (line 464 of
current master):
Real demerit = factor * dy;
if (state.extremes_[d].stem_
&& state.extremes_[d].stem_dir_ == state.dir_
// TODO - Stem::get_beaming() should be precomputed.
&& !Stem::get_beaming (state.extremes_[d].stem_, -d))
* demerit /= 5;*
To me this signals that the programmer is aware that there isn't a magical
formula to imitate hand engraving and particular cases will need a
different penalty. Is there something I'm missing? Is the penalty reduced
here in this particular case to fulfill a higher purpose or is that a case
of "last resort fix"? This kind of "tweaking" is all over the code, so it's
difficult not to interpret that as something that is a natural part of the
way penalties are handled. For any information in this regard that helps me
understand the code better I'm thankful.
Regards,
Martín.
--
www.martinrinconbotero.com
