Re: setting the number of pages for a score

[Joe Neeman]

Han-Wen Nienhuys wrote:

> Joe Neeman wrote:
>
> > I'm just worried that the algorithm will have to deal with scores 
> > that have almost no viable breaks. Either we allow breaks at 
> > practically every rest (with huge penalties) or we don't have enough 
> > break points. The first solution will have very poor performance, and 
> > the second solution is also obviously not good. But maybe there is a 
> > better way to ensure good performance.
>
> I don't understand why this is a problem. If there are no viable 
> breaks, every configuration is as bad, and the algorithm should tend 
> to the one which uses spacing that is as natural as possible, since 
> that one should be the best among the configurations with no specially 
> good page breaks.

It's a bit hard to say without implementing it, but I think it would try 
and use as few breaks as possible (making the music very cramped) 
because it gets a big penalty for every break. The penalties for the 
poor line breaking would get lost compared to the page breaking 
penalties. Ok, so this depends on how big the penalties are, but if they 
are too small then lily might give bad page turns.

Or consider the possibility that I typeset something where I don't care 
about page turns at all. There might be one or 2 decent turns, but not 
enough for good breaking. In trying to ensure good page turns, lilypond 
would give me uneven line breaking, blank pages and all sorts of 
unnecessary ugly stuff.

So first of all, there should be a way to tell the page-breaker to 
completely ignore page turns. But also, the above situation would give 
poor default output. I haven't really thought this through, but could 
the default be to autodetect whether page turns should matter, with an 
override by setting a \paper block variable?