>> Attached. In the end the necessary modifications were surprisingly
>> minor. Please comment, there is certainly room for improvements.
>
> Thanks. This is a bit what I feared: in this state, there is
> potential for things going wrong with other after-line-breaking
> callbacks. Those are the kind of thing that can be buggy in obscure
> circumstances. I'm a little nervous on having my "off-the-top" /
> "first thing that seemed to work" solution in the code base :-)
I think your code is very clear and easy to understand. My extensions
aren't complicated either. It also works nicely and flawless in
combination with the new `VaticanaLyrics` context. So please be more
specific about your fears. You told us that you are working on the
breaking algorithm since a few months, so I guess you are the best
person to voice objections and give counterexamples.
Kieren, can you test this, too?
We should probably continue the discussion on lilypond-devel, since it
seems to get bloody technical. :-)
> OTOH, it's *probably* not too difficult to refactor this to use a
> separate 'original-text property and a 'text property computed by a
> callback, with X-offset computed only after-line-breaking, X-extent
> left empty and springs-and-rods doing the magic that would otherwise
> be done by Separation_item::set_distance, at the condition of
> letting Grob::extent skip the X-offset calculation in case the
> extent is found to be empty (which could be a good idea regardless
> actually).
This is way too complicated for me, sorry.
> Or something like that. Too bad that we don't have pure widths (we
> could if I finally managed to finish a big patch that has been
> pending for about a year), but it should be possible to do without I
> think. I think.
Well, I fear I can't help you with the coding if you think this is the
best route to go. But as usual: The better is the enemy of the
good :-)
> But right now, I'm a bit swamped due to a somewhat important exam
> next Thursday.
I wish you success!
> If you guys are ready to clean up this thing, that would be a relief
> for me given the amount of other contributions that I am trying to
> get around to.
Define 'clean up'. AFAICS, what you have outlined above is not a
cleanup but a complete reimplementation based on a different
algorithm...
Werner
