/*
This file is part of LilyPond, the GNU music typesetter.
- Copyright (C) 2006--2010 Joe Neeman <joeneeman@gmail.com>
+ Copyright (C) 2006--2011 Joe Neeman <joeneeman@gmail.com>
LilyPond is free software: you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
discarded after a call to set_current_breakpoints, since that Line_division
refers to a subset of chunks which might be different from the current
subset of chunks under consideration.
+
+ HOW TO WRITE A PAGE BREAKING ALGORITHM
+ All page breakers supported by this class work more-or-less in the same way.
+ First, they request a particular number of systems by saying
+ set_current_breakpoints (0, last_break_position (), system_count)
+ (never mind what the first two arguments do, I'll get to them later).
+ Alternatively, you can do
+ set_to_ideal_line_configuration (0, last_break_position ()),
+ and the number of systems will be automatically chosen according to what
+ the line breaker wants.
+
+ If there are multiple scores, there will be many different ways to achieve
+ a certain number of lines. You can see how many alternatives are available
+ with current_configuration_count (). For every i from 0 to
+ current_configuration_count ()-1, you can see the line division of the
+ corresponding configuration with current_configuration (i), or you can try
+ out various page configurations with one of the space_systems_xxx or
+ pack_systems_xxx functions. The first argument to each of these functions
+ is the configuration index.
+
+ When you're done trying out configurations and you've picked the one
+ you want, do
+ break_into_pieces (0, last_break_position (), line_division_that_you_want);
+ return make_pages (systems_per_page, systems ());
+ where systems_per_page is a vector of numbers telling how many systems are
+ on each page. You can get your systems_per_page vector by looking inside
+ the Page_spacing_results that are returned by space_systems_xxx or
+ pack_systems_xxx.
+
+ A note on performance: set_current_breakpoints is EXPONENTIALLY SLOW unless
+ you constrain it by giving it a lower or an upper bound on the configurations
+ it looks for. Optimal_page_breaking, for example, works by trying
+ out a bunch of configurations, increasing the system count by one, trying
+ again and so on. Each time we increase the system count, we assume that the
+ best new configurations are going to be elementwise larger than the
+ best configuration for the previous system count (in other words, we're going
+ to get a new configuration just by adding an extra line to sone score
+ and leaving the rest the same). Therefore, we pass the best previous line
+ division as an lower bound to set_current_breakpoints.
+
+ Now you should be in a position to understand Optimal_page_breaking::solve.
+ Go ahead and read that before finding out, in the next paragraph,
+ what the first two arguments to set_current_breakpoints do.
+
+ "BREAKS"
+ Sometimes, it's useful to run this whole page-breaking machinery on a subset
+ of the book. To do this, you can mark certain "breaks" in the book (a poor
+ choice of name, perhaps, since a "break" here is different from a page break)
+ and you can run page breaking between any two breaks. You mark your breaks
+ by providing a Break_predicate (and, if you want, a Prob_break_predicate)
+ to Page_breaking's constructor. You then choose a subset of your book
+ by passing the starting and ending breaks to set_current_breakpoints. You
+ can see an example of this in Page_turn_page_breaking, where there is a break
+ everywhere that a page turn is allowed.
*/
#include "page-breaking.hh"
ret = scm_cons (page, ret);
--page_num;
}
+
+ // By reversing the table, we ensure that duplicated labels (eg. those
+ // straddling a page turn) will appear in the table with their last
+ // occurence first.
+ label_page_table = scm_reverse_x (label_page_table, SCM_EOL);
book_->top_paper ()->set_variable (ly_symbol2scm ("label-page-table"), label_page_table);
return ret;
}
return ret;
}
+// Returns the minimum number of _non-title_ lines.
vsize
Page_breaking::min_system_count (vsize start, vsize end)
{
return ret;
}
+// Returns the maximum number of _non-title_ lines.
vsize
Page_breaking::max_system_count (vsize start, vsize end)
{
return ret;
}
+// The numbers returned by this function represent either
+// the maximum or minimum number of _non-title_ lines
+// per chunk.
Page_breaking::Line_division
Page_breaking::system_count_bounds (vector<Break_position> const &chunks,
bool min)
assert (chunks.size () >= 2);
Line_division ret;
- ret.resize (chunks.size () - 1, 1);
+ ret.resize (chunks.size () - 1, 0);
for (vsize i = 0; i + 1 < chunks.size (); i++)
{
div.push_back (line_breaking_[sys].best_solution (start, end).size ());
}
else
- div.push_back (1);
+ div.push_back (0);
system_count_ += div.back ();
}
}
else
{
- assert (div[i] == 1);
+ assert (div[i] == 0);
uncompressed_line_details_.push_back (system_specs_[sys].prob_
? Line_details (system_specs_[sys].prob_, book_->paper_)
: Line_details ());
int real_min = max ((int) min_sys[my_index], (int) system_count - others_max);
int real_max = min ((int) max_sys[my_index], (int) system_count - others_min);
- if (real_min > real_max || real_min <= 0)
+ if (real_min > real_max || real_min < 0)
{
/* this should never happen within a recursive call. If it happens
at all, it means that we were called with an unsolvable problem
projects / lilypond.git / blobdiff