lyx_mirror/src/output_plaintext.cpp
Richard Heck ecef54500d Introduce max_length parameter for plaintext() output routines,
so we can write a limited amount when using this for TOC and
tooltip output.

This should solve the problem with slowness that Kornel noticed,
which was caused by our trying to write an entire plaintext
bibliography every time we updated the TOC. We did that because
he had a bibliography inside a branch, and we use plaintext for
creating the tooltip that goes with the branch list.

Other related bugs were fixed along the way. E.g., it turns out
that, if someone had an InsetInclude inside a branch, then we would
have been writing a *plaintext file* for that inset every time we
updated the TOC. I wonder if some of the other reports of slowness
we have received might be due to this kind of issue?
2013-03-08 15:12:58 -05:00

250 lines
5.5 KiB
C++

/**
* \file output_plaintext.cpp
* This file is part of LyX, the document processor.
* Licence details can be found in the file COPYING.
*
* \author Lars Gullik Bjønnes
*
* Full author contact details are available in file CREDITS.
*/
#include <config.h>
#include "output_plaintext.h"
#include "Buffer.h"
#include "BufferParams.h"
#include "Layout.h"
#include "output.h"
#include "OutputParams.h"
#include "Paragraph.h"
#include "ParagraphList.h"
#include "ParagraphParameters.h"
#include "support/debug.h"
#include "support/gettext.h"
#include "support/lstrings.h"
using namespace std;
using namespace lyx::support;
namespace lyx {
void writePlaintextFile(Buffer const & buf, FileName const & fname,
OutputParams const & runparams)
{
ofdocstream ofs;
if (!openFileWrite(ofs, fname))
return;
// make sure we are ready to export
buf.updateBuffer();
buf.updateMacroInstances(OutputUpdate);
buf.makeCitationLabels();
writePlaintextFile(buf, ofs, runparams);
}
void writePlaintextFile(Buffer const & buf, odocstream & os,
OutputParams const & runparams)
{
bool ref_printed = false;
ParagraphList const & par = buf.paragraphs();
ParagraphList::const_iterator beg = par.begin();
ParagraphList::const_iterator end = par.end();
ParagraphList::const_iterator it = beg;
for (; it != end; ++it) {
writePlaintextParagraph(buf, *it, os, runparams, ref_printed);
os << "\n";
if (runparams.linelen > 0)
os << "\n";
}
}
static pair<int, docstring> addDepth(int depth, int ldepth)
{
int d = depth * 2;
if (ldepth > depth)
d += (ldepth - depth) * 2;
return make_pair(d, docstring(d, ' '));
}
void writePlaintextParagraph(Buffer const & buf,
Paragraph const & par,
odocstream & ods,
OutputParams const & runparams,
bool & ref_printed, size_t max_length)
{
int ltype = 0;
depth_type ltype_depth = 0;
depth_type depth = par.params().depth();
// First write the layout
string const tmp = to_utf8(par.layout().name());
if (compare_ascii_no_case(tmp, "itemize") == 0) {
ltype = 1;
ltype_depth = depth + 1;
} else if (compare_ascii_no_case(tmp, "enumerate") == 0) {
ltype = 2;
ltype_depth = depth + 1;
} else if (contains(ascii_lowercase(tmp), "ection")) {
ltype = 3;
ltype_depth = depth + 1;
} else if (contains(ascii_lowercase(tmp), "aragraph")) {
ltype = 4;
ltype_depth = depth + 1;
} else if (compare_ascii_no_case(tmp, "description") == 0) {
ltype = 5;
ltype_depth = depth + 1;
} else if (compare_ascii_no_case(tmp, "abstract") == 0) {
ltype = 6;
ltype_depth = 0;
} else if (compare_ascii_no_case(tmp, "bibliography") == 0) {
ltype = 7;
ltype_depth = 0;
} else {
ltype = 0;
ltype_depth = 0;
}
/* the labelwidthstring used in lists */
/* noindent ? */
/* what about the alignment */
// runparams.linelen == 0 is special and means we don't have paragraph breaks
string::size_type currlinelen = 0;
odocstringstream os;
os << docstring(depth * 2, ' ');
currlinelen += depth * 2;
//--
// we should probably change to the paragraph language in the
// support/gettext.here (if possible) so that strings are output in
// the correct language! (20012712 Jug)
//--
switch (ltype) {
case 0: // Standard
case 4: // (Sub)Paragraph
case 5: // Description
break;
case 6: // Abstract
if (runparams.linelen > 0) {
os << buf.B_("Abstract") << "\n\n";
currlinelen = 0;
} else {
docstring const abst = buf.B_("Abstract: ");
os << abst;
currlinelen += abst.length();
}
break;
case 7: // Bibliography
if (!ref_printed) {
if (runparams.linelen > 0) {
os << buf.B_("References") << "\n\n";
currlinelen = 0;
} else {
docstring const refs = buf.B_("References: ");
os << refs;
currlinelen += refs.length();
}
ref_printed = true;
}
break;
default: {
docstring const label = par.params().labelString();
if (!label.empty()) {
os << label << ' ';
currlinelen += label.length() + 1;
}
break;
}
}
if (currlinelen == 0) {
pair<int, docstring> p = addDepth(depth, ltype_depth);
os << p.second;
currlinelen += p.first;
}
docstring word;
for (pos_type i = 0; i < par.size(); ++i) {
// deleted characters don't make much sense in plain text output
if (par.isDeleted(i))
continue;
if (os.str().size() > max_length)
break;
char_type c = par.getUChar(buf.params(), i);
if (par.isInset(i) || c == ' ') {
if (runparams.linelen > 0 &&
currlinelen + word.length() > runparams.linelen) {
os << '\n';
pair<int, docstring> p = addDepth(depth, ltype_depth);
os << p.second;
currlinelen = p.first;
}
os << word;
currlinelen += word.length();
word.erase();
}
if (par.isInset(i)) {
OutputParams rp = runparams;
rp.depth = par.params().depth();
int len = par.getInset(i)->plaintext(os, rp, max_length);
if (len >= Inset::PLAINTEXT_NEWLINE)
currlinelen = len - Inset::PLAINTEXT_NEWLINE;
else
currlinelen += len;
continue;
}
switch (c) {
case ' ':
os << ' ';
currlinelen++;
break;
case '\0':
LYXERR(Debug::INFO, "writePlaintextFile: NUL char in structure.");
break;
default:
word += c;
break;
}
}
// currlinelen may be greater than runparams.linelen!
// => check whether word is empty and do nothing in this case
if (!word.empty()) {
if (runparams.linelen > 0 &&
currlinelen + word.length() > runparams.linelen) {
os << '\n';
pair<int, docstring> p = addDepth(depth, ltype_depth);
os << p.second;
currlinelen = p.first;
}
os << word;
}
ods << os.str();
}
} // namespace lyx