|
From: | Boris Shingarov |
Subject: | Re: Vertical spacing regression !? |
Date: | Sun, 27 Jun 2010 06:56:31 -0400 |
User-agent: | Mozilla/5.0 (X11; U; Linux x86_64; en-US; rv:1.9.1.9) Gecko/20100423 Thunderbird/3.0.4 |
Hi Arno,Here are some design remarks about what I understood from my work on the breaker. Hopefully a sum of a few [dozen?] of mails like this will together constitute something to transform into a chapter in the CG...
The problem is that page-breaking does not come *after* page-breaking. Well, conceptually, it sort of does. But the optimization of breaking is global over the product space of all line breaking possibilities by all page breaking possibilities. It is done in the hope to get more optimal breaks. The downside is that we are forced to use "height estimations" instead of real heights of things; otherwise we would need to do page layout for every possible break configuration, which would be prohibitively expensive.Thanks, I've reverted the patch in the meantime. However, the information that you see in annotate-spacing is actually computed after line-breaking,ok, so why can't it then be taken into account for page-breaking?
This was discussed on this list only a few weeks ago. I think we are on our way to get rid of global page*line breaking. Height estimation is just too inaccurate, so you either have music running off the page at the bottom, or (no less ugly) empty space. We did fix a bunch of height estimation issues, but I do not know how many more are left. This one we are discussing right now, is another example of problem caused by height estimation.
Since it more than doubles the computation time (often by far) there must be something redundant, no?No, because the height used in page breaking, is not the actual height; it is just an estimate (often wrong). This is what Joe means when he says, "...and they aren't actually correct to begin with".
Boris
[Prev in Thread] | Current Thread | [Next in Thread] |