lyx_mirror/src/ParagraphMetrics.cpp
André Pönitz f1cba8ff64 more latin1..utf8 schanges. all of src/* should be utf8 now
git-svn-id: svn://svn.lyx.org/lyx/lyx-devel/trunk@27425 a592a061-630c-0410-9148-cb99ea01b6c8
2008-11-14 15:58:50 +00:00

282 lines
6.2 KiB
C++

/**
* \file Paragraph.cpp
* This file is part of LyX, the document processor.
* Licence details can be found in the file COPYING.
*
* \author Asger Alstrup
* \author Lars Gullik Bjønnes
* \author Jean-Marc Lasgouttes
* \author Angus Leeming
* \author John Levon
* \author André Pönitz
* \author Dekel Tsur
* \author Jürgen Vigna
*
* Full author contact details are available in file CREDITS.
*/
#include <config.h>
#include "ParagraphMetrics.h"
#include "Buffer.h"
#include "BufferParams.h"
#include "BufferView.h"
#include "Counters.h"
#include "Encoding.h"
#include "Language.h"
#include "LaTeXFeatures.h"
#include "Layout.h"
#include "Font.h"
#include "LyXRC.h"
#include "Row.h"
#include "OutputParams.h"
#include "paragraph_funcs.h"
#include "sgml.h"
#include "TextClass.h"
#include "TexRow.h"
#include "VSpace.h"
#include "frontends/FontMetrics.h"
#include "insets/InsetBibitem.h"
#include "insets/InsetOptArg.h"
#include "support/lassert.h"
#include "support/convert.h"
#include "support/debug.h"
#include "support/gettext.h"
#include "support/lstrings.h"
#include "support/textutils.h"
#include <boost/bind.hpp>
#include <boost/crc.hpp>
#include <algorithm>
#include <list>
#include <stack>
#include <sstream>
using namespace std;
using namespace lyx::support;
namespace lyx {
ParagraphMetrics::ParagraphMetrics(Paragraph const & par) :
position_(-1), par_(&par)
{}
ParagraphMetrics & ParagraphMetrics::operator=(
ParagraphMetrics const & pm)
{
rows_ = pm.rows_;
dim_ = pm.dim_;
par_ = pm.par_;
position_ = pm.position_;
return *this;
}
void ParagraphMetrics::reset(Paragraph const & par)
{
par_ = &par;
dim_ = Dimension();
//position_ = -1;
}
size_t ParagraphMetrics::computeRowSignature(Row const & row,
BufferParams const & bparams) const
{
boost::crc_32_type crc;
for (pos_type i = row.pos(); i < row.endpos(); ++i) {
char_type const b[] = { par_->getChar(i) };
crc.process_bytes(b, sizeof(char_type));
if (bparams.trackChanges) {
Change change = par_->lookupChange(i);
char_type const b[] = { change.type };
// 1 byte is enough to encode Change::Type
crc.process_bytes(b, 1);
}
}
Dimension const & d = row.dimension();
char_type const b[] = { row.sel_beg, row.sel_end,
row.begin_margin_sel, row.end_margin_sel, d.wid, d.asc, d.des};
// Each of the variable to process is 4 bytes: 4x7 = 28
crc.process_bytes(b, 28);
return crc.checksum();
}
void ParagraphMetrics::setPosition(int position)
{
position_ = position;
}
Dimension const & ParagraphMetrics::insetDimension(Inset const * inset) const
{
InsetDims::const_iterator it = inset_dims_.find(inset);
if (it != inset_dims_.end())
return it->second;
static Dimension dummy;
return dummy;
}
void ParagraphMetrics::setInsetDimension(Inset const * inset,
Dimension const & dim)
{
inset_dims_[inset] = dim;
}
Row & ParagraphMetrics::getRow(pos_type pos, bool boundary)
{
LASSERT(!rows().empty(), /**/);
// If boundary is set we should return the row on which
// the character before is inside.
if (pos > 0 && boundary)
--pos;
RowList::iterator rit = rows_.end();
RowList::iterator const begin = rows_.begin();
for (--rit; rit != begin && rit->pos() > pos; --rit)
;
return *rit;
}
Row const & ParagraphMetrics::getRow(pos_type pos, bool boundary) const
{
LASSERT(!rows().empty(), /**/);
// If boundary is set we should return the row on which
// the character before is inside.
if (pos > 0 && boundary)
--pos;
RowList::const_iterator rit = rows_.end();
RowList::const_iterator const begin = rows_.begin();
for (--rit; rit != begin && rit->pos() > pos; --rit)
;
return *rit;
}
size_t ParagraphMetrics::pos2row(pos_type pos) const
{
LASSERT(!rows().empty(), /**/);
RowList::const_iterator rit = rows_.end();
RowList::const_iterator const begin = rows_.begin();
for (--rit; rit != begin && rit->pos() > pos; --rit)
;
return rit - begin;
}
void ParagraphMetrics::dump() const
{
lyxerr << "Paragraph::dump: rows.size(): " << rows_.size() << endl;
for (size_t i = 0; i != rows_.size(); ++i) {
lyxerr << " row " << i << ": ";
rows_[i].dump();
}
}
int ParagraphMetrics::rightMargin(BufferView const & bv) const
{
BufferParams const & params = bv.buffer().params();
DocumentClass const & tclass = params.documentClass();
frontend::FontMetrics const & fm = theFontMetrics(params.getFont());
int const r_margin =
bv.rightMargin()
+ fm.signedWidth(tclass.rightmargin())
+ fm.signedWidth(par_->layout().rightmargin)
* 4 / (par_->getDepth() + 4);
return r_margin;
}
int ParagraphMetrics::singleWidth(pos_type pos, Font const & font) const
{
// The most special cases are handled first.
if (Inset const * inset = par_->getInset(pos))
return insetDimension(inset).wid;
char_type c = par_->getChar(pos);
if (c == '\t')
return 4 * theFontMetrics(font).width(' ');
if (!isPrintable(c))
return theFontMetrics(font).width(c);
Language const * language = font.language();
if (language->rightToLeft()) {
if (language->lang() == "arabic_arabtex" ||
language->lang() == "arabic_arabi" ||
language->lang() == "farsi") {
if (Encodings::isArabicComposeChar(c))
return 0;
c = par_->transformChar(c, pos);
} else if (language->lang() == "hebrew" &&
Encodings::isHebrewComposeChar(c)) {
return 0;
}
}
return theFontMetrics(font).width(c);
}
bool ParagraphMetrics::hfillExpansion(Row const & row, pos_type pos) const
{
if (!par_->isHfill(pos))
return false;
LASSERT(pos >= row.pos() && pos < row.endpos(), /**/);
// expand at the end of a row only if there is another hfill on the same row
if (pos == row.endpos() - 1) {
for (pos_type i = row.pos(); i < pos; i++) {
if (par_->isHfill(i))
return true;
}
return false;
}
// expand at the beginning of a row only if it is the first row of a paragraph
if (pos == row.pos()) {
return pos == 0;
}
// do not expand in some labels
if (par_->layout().margintype != MARGIN_MANUAL && pos < par_->beginOfBody())
return false;
// if there is anything between the first char of the row and
// the specified position that is neither a newline nor an hfill,
// the hfill will be expanded, otherwise it won't
for (pos_type i = row.pos(); i < pos; i++) {
if (!par_->isNewline(i) && !par_->isHfill(i))
return true;
}
return false;
}
} // namespace lyx