]> git.lyx.org Git - lyx.git/blob - src/Paragraph.cpp
56a19927b8e08cb74c31fee282e7a1d95e147e19
[lyx.git] / src / Paragraph.cpp
1 /**
2  * \file Paragraph.cpp
3  * This file is part of LyX, the document processor.
4  * Licence details can be found in the file COPYING.
5  *
6  * \author Asger Alstrup
7  * \author Lars Gullik Bjønnes
8  * \author Jean-Marc Lasgouttes
9  * \author Angus Leeming
10  * \author John Levon
11  * \author André Pönitz
12  * \author Dekel Tsur
13  * \author Jürgen Vigna
14  *
15  * Full author contact details are available in file CREDITS.
16  */
17
18 #include <config.h>
19
20 #include "Paragraph.h"
21
22 #include "Buffer.h"
23 #include "BufferParams.h"
24 #include "Changes.h"
25 #include "Counters.h"
26 #include "Encoding.h"
27 #include "debug.h"
28 #include "gettext.h"
29 #include "InsetList.h"
30 #include "Language.h"
31 #include "LaTeXFeatures.h"
32 #include "Color.h"
33 #include "Layout.h"
34 #include "Length.h"
35 #include "Font.h"
36 #include "FontList.h"
37 #include "LyXRC.h"
38 #include "Messages.h"
39 #include "OutputParams.h"
40 #include "output_latex.h"
41 #include "paragraph_funcs.h"
42 #include "ParagraphParameters.h"
43 #include "sgml.h"
44 #include "TexRow.h"
45 #include "VSpace.h"
46
47 #include "frontends/alert.h"
48 #include "frontends/FontMetrics.h"
49
50 #include "insets/InsetBibitem.h"
51 #include "insets/InsetLabel.h"
52 #include "insets/InsetOptArg.h"
53
54 #include "support/lstrings.h"
55 #include "support/textutils.h"
56 #include "support/convert.h"
57 #include "support/unicode.h"
58
59 #include <sstream>
60 #include <vector>
61
62 using std::endl;
63 using std::string;
64 using std::ostream;
65
66 namespace lyx {
67
68 using support::contains;
69 using support::lowercase;
70 using support::prefixIs;
71 using support::suffixIs;
72 using support::rsplit;
73 using support::uppercase;
74
75 namespace {
76 /// Inset identifier (above 0x10ffff, for ucs-4)
77 char_type const META_INSET = 0x200001;
78 };
79
80 /////////////////////////////////////////////////////////////////////
81 //
82 // Paragraph::Private
83 //
84 /////////////////////////////////////////////////////////////////////
85
86 class Paragraph::Private
87 {
88 public:
89         ///
90         Private(Paragraph * owner);
91         /// "Copy constructor"
92         Private(Private const &, Paragraph * owner);
93
94         ///
95         void insertChar(pos_type pos, char_type c, Change const & change);
96
97         /// Output the surrogate pair formed by \p c and \p next to \p os.
98         /// \return the number of characters written.
99         int latexSurrogatePair(odocstream & os, char_type c, char_type next,
100                                Encoding const &);
101
102         /// Output a space in appropriate formatting (or a surrogate pair
103         /// if the next character is a combining character).
104         /// \return whether a surrogate pair was output.
105         bool simpleTeXBlanks(Encoding const &,
106                              odocstream &, TexRow & texrow,
107                              pos_type i,
108                              unsigned int & column,
109                              Font const & font,
110                              Layout const & style);
111
112         /// Output consecutive unicode chars, belonging to the same script as
113         /// specified by the latex macro \p ltx, to \p os starting from \p c.
114         /// \return the number of characters written.
115         int writeScriptChars(odocstream & os, char_type c, docstring const & ltx,
116                            Change &, Encoding const &, pos_type &);
117
118         /// This could go to ParagraphParameters if we want to.
119         int startTeXParParams(BufferParams const &, odocstream &, TexRow &,
120                               bool) const;
121
122         /// This could go to ParagraphParameters if we want to.
123         int endTeXParParams(BufferParams const &, odocstream &, TexRow &,
124                             bool) const;
125
126         ///
127         void latexInset(Buffer const &, BufferParams const &,
128                                    odocstream &,
129                                    TexRow & texrow, OutputParams &,
130                                    Font & running_font,
131                                    Font & basefont,
132                                    Font const & outerfont,
133                                    bool & open_font,
134                                    Change & running_change,
135                                    Layout const & style,
136                                    pos_type & i,
137                                    unsigned int & column);
138
139         ///
140         void latexSpecialChar(
141                                    odocstream & os,
142                                    OutputParams & runparams,
143                                    Font & running_font,
144                                    Change & running_change,
145                                    Layout const & style,
146                                    pos_type & i,
147                                    unsigned int & column);
148
149         ///
150         bool latexSpecialT1(
151                 char_type const c,
152                 odocstream & os,
153                 pos_type & i,
154                 unsigned int & column);
155         ///
156         bool latexSpecialTypewriter(
157                 char_type const c,
158                 odocstream & os,
159                 pos_type & i,
160                 unsigned int & column);
161         ///
162         bool latexSpecialPhrase(
163                 odocstream & os,
164                 pos_type & i,
165                 unsigned int & column);
166
167         ///
168         void validate(LaTeXFeatures & features,
169                       Layout const & layout) const;
170
171         /// Checks if the paragraph contains only text and no inset or font change.
172         bool onlyText(Buffer const & buf, Font const & outerfont,
173                       pos_type initial) const;
174
175         /// match a string against a particular point in the paragraph
176         bool isTextAt(std::string const & str, pos_type pos) const;
177         
178         /// Which Paragraph owns us?
179         Paragraph * owner_;
180
181         /// In which Inset?
182         Inset * inset_owner_;
183
184         ///
185         FontList fontlist_;
186
187         ///
188         unsigned int id_;
189         ///
190         static unsigned int paragraph_id;
191         ///
192         ParagraphParameters params_;
193
194         /// for recording and looking up changes
195         Changes changes_;
196
197         ///
198         InsetList insetlist_;
199
200         ///
201         LayoutPtr layout_;
202
203         /// end of label
204         pos_type begin_of_body_;
205
206         typedef docstring TextContainer;
207         ///
208         TextContainer text_;
209 };
210
211
212
213
214 using std::endl;
215 using std::upper_bound;
216 using std::lower_bound;
217 using std::string;
218
219
220 // Initialization of the counter for the paragraph id's,
221 unsigned int Paragraph::Private::paragraph_id = 0;
222
223 namespace {
224
225 struct special_phrase {
226         string phrase;
227         docstring macro;
228         bool builtin;
229 };
230
231 special_phrase const special_phrases[] = {
232         { "LyX", from_ascii("\\protect\\LyX{}"), false },
233         { "TeX", from_ascii("\\protect\\TeX{}"), true },
234         { "LaTeX2e", from_ascii("\\protect\\LaTeXe{}"), true },
235         { "LaTeX", from_ascii("\\protect\\LaTeX{}"), true },
236 };
237
238 size_t const phrases_nr = sizeof(special_phrases)/sizeof(special_phrase);
239
240 } // namespace anon
241
242
243 Paragraph::Private::Private(Paragraph * owner)
244         : owner_(owner), inset_owner_(0), begin_of_body_(0)
245 {
246         id_ = paragraph_id++;
247         text_.reserve(100);
248 }
249
250
251 Paragraph::Private::Private(Private const & p, Paragraph * owner)
252         : owner_(owner), inset_owner_(p.inset_owner_), fontlist_(p.fontlist_), 
253           params_(p.params_), changes_(p.changes_), insetlist_(p.insetlist_),
254           layout_(p.layout_), begin_of_body_(p.begin_of_body_), text_(p.text_)
255 {
256         id_ = paragraph_id++;
257 }
258
259
260 bool Paragraph::isChanged(pos_type start, pos_type end) const
261 {
262         BOOST_ASSERT(start >= 0 && start <= size());
263         BOOST_ASSERT(end > start && end <= size() + 1);
264
265         return d->changes_.isChanged(start, end);
266 }
267
268
269 bool Paragraph::isMergedOnEndOfParDeletion(bool trackChanges) const {
270         // keep the logic here in sync with the logic of eraseChars()
271
272         if (!trackChanges) {
273                 return true;
274         }
275
276         Change change = d->changes_.lookup(size());
277
278         return change.type == Change::INSERTED && change.author == 0;
279 }
280
281
282 void Paragraph::setChange(Change const & change)
283 {
284         // beware of the imaginary end-of-par character!
285         d->changes_.set(change, 0, size() + 1);
286
287         /*
288          * Propagate the change recursively - but not in case of DELETED!
289          *
290          * Imagine that your co-author makes changes in an existing inset. He
291          * sends your document to you and you come to the conclusion that the
292          * inset should go completely. If you erase it, LyX must not delete all
293          * text within the inset. Otherwise, the change tracked insertions of
294          * your co-author get lost and there is no way to restore them later.
295          *
296          * Conclusion: An inset's content should remain untouched if you delete it
297          */
298
299         if (change.type != Change::DELETED) {
300                 for (pos_type pos = 0; pos < size(); ++pos) {
301                         if (isInset(pos))
302                                 getInset(pos)->setChange(change);
303                 }
304         }
305 }
306
307
308 void Paragraph::setChange(pos_type pos, Change const & change)
309 {
310         BOOST_ASSERT(pos >= 0 && pos <= size());
311
312         d->changes_.set(change, pos);
313
314         // see comment in setChange(Change const &) above
315
316         if (change.type != Change::DELETED &&
317             pos < size() && isInset(pos)) {
318                 getInset(pos)->setChange(change);
319         }
320 }
321
322
323 Change const & Paragraph::lookupChange(pos_type pos) const
324 {
325         BOOST_ASSERT(pos >= 0 && pos <= size());
326
327         return d->changes_.lookup(pos);
328 }
329
330
331 void Paragraph::acceptChanges(BufferParams const & bparams, pos_type start,
332                 pos_type end)
333 {
334         BOOST_ASSERT(start >= 0 && start <= size());
335         BOOST_ASSERT(end > start && end <= size() + 1);
336
337         for (pos_type pos = start; pos < end; ++pos) {
338                 switch (lookupChange(pos).type) {
339                         case Change::UNCHANGED:
340                                 // accept changes in nested inset
341                                 if (pos < size() && isInset(pos))
342                                         getInset(pos)->acceptChanges(bparams);
343
344                                 break;
345
346                         case Change::INSERTED:
347                                 d->changes_.set(Change(Change::UNCHANGED), pos);
348                                 // also accept changes in nested inset
349                                 if (pos < size() && isInset(pos)) {
350                                         getInset(pos)->acceptChanges(bparams);
351                                 }
352                                 break;
353
354                         case Change::DELETED:
355                                 // Suppress access to non-existent
356                                 // "end-of-paragraph char"
357                                 if (pos < size()) {
358                                         eraseChar(pos, false);
359                                         --end;
360                                         --pos;
361                                 }
362                                 break;
363                 }
364
365         }
366 }
367
368
369 void Paragraph::rejectChanges(BufferParams const & bparams,
370                 pos_type start, pos_type end)
371 {
372         BOOST_ASSERT(start >= 0 && start <= size());
373         BOOST_ASSERT(end > start && end <= size() + 1);
374
375         for (pos_type pos = start; pos < end; ++pos) {
376                 switch (lookupChange(pos).type) {
377                         case Change::UNCHANGED:
378                                 // reject changes in nested inset
379                                 if (pos < size() && isInset(pos)) {
380                                         getInset(pos)->rejectChanges(bparams);
381                                 }
382                                 break;
383
384                         case Change::INSERTED:
385                                 // Suppress access to non-existent
386                                 // "end-of-paragraph char"
387                                 if (pos < size()) {
388                                         eraseChar(pos, false);
389                                         --end;
390                                         --pos;
391                                 }
392                                 break;
393
394                         case Change::DELETED:
395                                 d->changes_.set(Change(Change::UNCHANGED), pos);
396
397                                 // Do NOT reject changes within a deleted inset!
398                                 // There may be insertions of a co-author inside of it!
399
400                                 break;
401                 }
402         }
403 }
404
405
406 void Paragraph::Private::insertChar(pos_type pos, char_type c,
407                 Change const & change)
408 {
409         BOOST_ASSERT(pos >= 0 && pos <= int(text_.size()));
410
411         // track change
412         changes_.insert(change, pos);
413
414         // This is actually very common when parsing buffers (and
415         // maybe inserting ascii text)
416         if (pos == text_.size()) {
417                 // when appending characters, no need to update tables
418                 text_.push_back(c);
419                 return;
420         }
421
422         text_.insert(text_.begin() + pos, c);
423
424         // Update the font table.
425         fontlist_.increasePosAfterPos(pos);
426
427         // Update the insets
428         insetlist_.increasePosAfterPos(pos);
429 }
430
431
432 void Paragraph::insertInset(pos_type pos, Inset * inset,
433                                    Change const & change)
434 {
435         BOOST_ASSERT(inset);
436         BOOST_ASSERT(pos >= 0 && pos <= size());
437
438         d->insertChar(pos, META_INSET, change);
439         BOOST_ASSERT(d->text_[pos] == META_INSET);
440
441         // Add a new entry in the insetlist_.
442         d->insetlist_.insert(inset, pos);
443 }
444
445
446 bool Paragraph::eraseChar(pos_type pos, bool trackChanges)
447 {
448         BOOST_ASSERT(pos >= 0 && pos <= size());
449
450         // keep the logic here in sync with the logic of isMergedOnEndOfParDeletion()
451
452         if (trackChanges) {
453                 Change change = d->changes_.lookup(pos);
454
455                 // set the character to DELETED if
456                 //  a) it was previously unchanged or
457                 //  b) it was inserted by a co-author
458
459                 if (change.type == Change::UNCHANGED ||
460                     (change.type == Change::INSERTED && change.author != 0)) {
461                         setChange(pos, Change(Change::DELETED));
462                         return false;
463                 }
464
465                 if (change.type == Change::DELETED)
466                         return false;
467         }
468
469         // Don't physically access the imaginary end-of-paragraph character.
470         // eraseChar() can only mark it as DELETED. A physical deletion of
471         // end-of-par must be handled externally.
472         if (pos == size()) {
473                 return false;
474         }
475
476         // track change
477         d->changes_.erase(pos);
478
479         // if it is an inset, delete the inset entry
480         if (d->text_[pos] == META_INSET)
481                 d->insetlist_.erase(pos);
482
483         d->text_.erase(d->text_.begin() + pos);
484
485         // Update the fontlist_
486         d->fontlist_.erase(pos);
487
488         // Update the insetlist_
489         d->insetlist_.decreasePosAfterPos(pos);
490
491         return true;
492 }
493
494
495 int Paragraph::eraseChars(pos_type start, pos_type end, bool trackChanges)
496 {
497         BOOST_ASSERT(start >= 0 && start <= size());
498         BOOST_ASSERT(end >= start && end <= size() + 1);
499
500         pos_type i = start;
501         for (pos_type count = end - start; count; --count) {
502                 if (!eraseChar(i, trackChanges))
503                         ++i;
504         }
505         return end - i;
506 }
507
508
509 int Paragraph::Private::latexSurrogatePair(odocstream & os, char_type c,
510                 char_type next, Encoding const & encoding)
511 {
512         // Writing next here may circumvent a possible font change between
513         // c and next. Since next is only output if it forms a surrogate pair
514         // with c we can ignore this:
515         // A font change inside a surrogate pair does not make sense and is
516         // hopefully impossible to input.
517         // FIXME: change tracking
518         // Is this correct WRT change tracking?
519         docstring const latex1 = encoding.latexChar(next);
520         docstring const latex2 = encoding.latexChar(c);
521         os << latex1 << '{' << latex2 << '}';
522         return latex1.length() + latex2.length() + 2;
523 }
524
525
526 bool Paragraph::Private::simpleTeXBlanks(Encoding const & encoding,
527                                        odocstream & os, TexRow & texrow,
528                                        pos_type i,
529                                        unsigned int & column,
530                                        Font const & font,
531                                        Layout const & style)
532 {
533         if (style.pass_thru)
534                 return false;
535
536         if (i + 1 < int(text_.size())) {
537                 char_type next = text_[i + 1];
538                 if (Encodings::isCombiningChar(next)) {
539                         // This space has an accent, so we must always output it.
540                         column += latexSurrogatePair(os, ' ', next, encoding) - 1;
541                         return true;
542                 }
543         }
544
545         if (lyxrc.plaintext_linelen > 0
546             && column > lyxrc.plaintext_linelen
547             && i
548             && text_[i - 1] != ' '
549             && (i + 1 < int(text_.size()))
550             // same in FreeSpacing mode
551             && !owner_->isFreeSpacing()
552             // In typewriter mode, we want to avoid
553             // ! . ? : at the end of a line
554             && !(font.family() == Font::TYPEWRITER_FAMILY
555                  && (text_[i - 1] == '.'
556                      || text_[i - 1] == '?'
557                      || text_[i - 1] == ':'
558                      || text_[i - 1] == '!'))) {
559                 os << '\n';
560                 texrow.newline();
561                 texrow.start(owner_->id(), i + 1);
562                 column = 0;
563         } else if (style.free_spacing) {
564                 os << '~';
565         } else {
566                 os << ' ';
567         }
568         return false;
569 }
570
571
572 int Paragraph::Private::writeScriptChars(odocstream & os,
573                                          char_type c,
574                                          docstring const & ltx,
575                                          Change & runningChange,
576                                          Encoding const & encoding,
577                                          pos_type & i)
578 {
579         // We only arrive here when a proper language for character c has not
580         // been specified (i.e., it could not be translated in the current
581         // latex encoding) and it belongs to a known script.
582         // Parameter ltx contains the latex translation of c as specified in
583         // the unicodesymbols file and is something like "\textXXX{<spec>}".
584         // The latex macro name "textXXX" specifies the script to which c
585         // belongs and we use it in order to check whether characters from the
586         // same script immediately follow, such that we can collect them in a
587         // single "\textXXX" macro. So, we have to retain "\textXXX{<spec>"
588         // for the first char but only "<spec>" for all subsequent chars.
589         docstring::size_type const brace1 = ltx.find_first_of(from_ascii("{"));
590         docstring::size_type const brace2 = ltx.find_last_of(from_ascii("}"));
591         string script = to_ascii(ltx.substr(1, brace1 - 1));
592         int length = ltx.substr(0, brace2).length();
593         os << ltx.substr(0, brace2);
594         int size = text_.size();
595         while (i + 1 < size) {
596                 char_type const next = text_[i + 1];
597                 // Stop here if next character belongs to another script
598                 // or there is a change in change tracking status.
599                 if (!Encodings::isKnownScriptChar(next, script) ||
600                     runningChange != owner_->lookupChange(i + 1))
601                         break;
602                 Font prev_font;
603                 bool found = false;
604                 FontList::const_iterator cit = fontlist_.begin();
605                 FontList::const_iterator end = fontlist_.end();
606                 for (; cit != end; ++cit) {
607                         if (cit->pos() >= i && !found) {
608                                 prev_font = cit->font();
609                                 found = true;
610                         }
611                         if (cit->pos() >= i + 1)
612                                 break;
613                 }
614                 // Stop here if there is a font attribute or encoding change.
615                 if (found && cit != end && prev_font != cit->font())
616                         break;
617                 docstring const latex = encoding.latexChar(next);
618                 docstring::size_type const b1 =
619                                         latex.find_first_of(from_ascii("{"));
620                 docstring::size_type const b2 =
621                                         latex.find_last_of(from_ascii("}"));
622                 int const len = b2 - b1 - 1;
623                 os << latex.substr(b1 + 1, len);
624                 length += len;
625                 ++i;
626         }
627         os << '}';
628         ++length;
629         return length;
630 }
631
632
633 bool Paragraph::Private::isTextAt(string const & str, pos_type pos) const
634 {
635         pos_type const len = str.length();
636
637         // is the paragraph large enough?
638         if (pos + len > int(text_.size()))
639                 return false;
640
641         // does the wanted text start at point?
642         for (string::size_type i = 0; i < str.length(); ++i) {
643                 // Caution: direct comparison of characters works only
644                 // because str is pure ASCII.
645                 if (str[i] != text_[pos + i])
646                         return false;
647         }
648
649         return fontlist_.hasChangeInRange(pos, len);
650 }
651
652
653 void Paragraph::Private::latexInset(Buffer const & buf,
654                                              BufferParams const & bparams,
655                                              odocstream & os,
656                                              TexRow & texrow,
657                                              OutputParams & runparams,
658                                              Font & running_font,
659                                              Font & basefont,
660                                              Font const & outerfont,
661                                              bool & open_font,
662                                              Change & running_change,
663                                              Layout const & style,
664                                              pos_type & i,
665                                              unsigned int & column)
666 {
667         Inset * inset = owner_->getInset(i);
668         BOOST_ASSERT(inset);
669
670         if (style.pass_thru) {
671                 inset->plaintext(buf, os, runparams);
672                 return;
673         }
674
675         // FIXME: move this to InsetNewline::latex
676         if (inset->lyxCode() == NEWLINE_CODE) {
677                 // newlines are handled differently here than
678                 // the default in simpleTeXSpecialChars().
679                 if (!style.newline_allowed) {
680                         os << '\n';
681                 } else {
682                         if (open_font) {
683                                 column += running_font.latexWriteEndChanges(
684                                         os, bparams, runparams,
685                                         basefont, basefont);
686                                 open_font = false;
687                         }
688
689                         if (running_font.family() == Font::TYPEWRITER_FAMILY)
690                                 os << '~';
691
692                         basefont = owner_->getLayoutFont(bparams, outerfont);
693                         running_font = basefont;
694
695                         if (runparams.moving_arg)
696                                 os << "\\protect ";
697
698                         os << "\\\\\n";
699                 }
700                 texrow.newline();
701                 texrow.start(owner_->id(), i + 1);
702                 column = 0;
703         }
704
705         if (owner_->lookupChange(i).type == Change::DELETED) {
706                 if( ++runparams.inDeletedInset == 1)
707                         runparams.changeOfDeletedInset = owner_->lookupChange(i);
708         }
709
710         if (inset->canTrackChanges()) {
711                 column += Changes::latexMarkChange(os, bparams, running_change,
712                         Change(Change::UNCHANGED));
713                 running_change = Change(Change::UNCHANGED);
714         }
715
716         bool close = false;
717         odocstream::pos_type const len = os.tellp();
718
719         if ((inset->lyxCode() == GRAPHICS_CODE
720              || inset->lyxCode() == MATH_CODE
721              || inset->lyxCode() == HYPERLINK_CODE)
722             && running_font.isRightToLeft()) {
723                 if (running_font.language()->lang() == "farsi")
724                         os << "\\beginL{}";
725                 else
726                         os << "\\L{";
727                 close = true;
728         }
729
730         // FIXME: Bug: we can have an empty font change here!
731         // if there has just been a font change, we are going to close it
732         // right now, which means stupid latex code like \textsf{}. AFAIK,
733         // this does not harm dvi output. A minor bug, thus (JMarc)
734
735         // Some insets cannot be inside a font change command.
736         // However, even such insets *can* be placed in \L or \R
737         // or their equivalents (for RTL language switches), so we don't
738         // close the language in those cases.
739         // ArabTeX, though, cannot handle this special behavior, it seems.
740         bool arabtex = basefont.language()->lang() == "arabic_arabtex"
741                 || running_font.language()->lang() == "arabic_arabtex";
742         if (open_font && inset->noFontChange()) {
743                 bool closeLanguage = arabtex
744                         || basefont.isRightToLeft() == running_font.isRightToLeft();
745                 unsigned int count = running_font.latexWriteEndChanges(os,
746                         bparams, runparams, basefont, basefont, closeLanguage);
747                 column += count;
748                 // if any font properties were closed, update the running_font, 
749                 // making sure, however, to leave the language as it was
750                 if (count > 0) {
751                         // FIXME: probably a better way to keep track of the old 
752                         // language, than copying the entire font?
753                         Font const copy_font(running_font);
754                         basefont = owner_->getLayoutFont(bparams, outerfont);
755                         running_font = basefont;
756                         if (!closeLanguage)
757                                 running_font.setLanguage(copy_font.language());
758                         // leave font open if language is still open
759                         open_font = (running_font.language() == basefont.language());
760                         if (closeLanguage)
761                                 runparams.local_font = &basefont;
762                 }
763         }
764
765         int tmp = inset->latex(buf, os, runparams);
766
767         if (close) {
768         if (running_font.language()->lang() == "farsi")
769                         os << "\\endL{}";
770                 else
771                         os << '}';
772         }
773
774         if (tmp) {
775                 for (int j = 0; j < tmp; ++j)
776                         texrow.newline();
777
778                 texrow.start(owner_->id(), i + 1);
779                 column = 0;
780         } else {
781                 column += os.tellp() - len;
782         }
783
784         if (owner_->lookupChange(i).type == Change::DELETED)
785                 --runparams.inDeletedInset;
786 }
787
788
789 void Paragraph::Private::latexSpecialChar(
790                                              odocstream & os,
791                                              OutputParams & runparams,
792                                              Font & running_font,
793                                              Change & running_change,
794                                              Layout const & style,
795                                              pos_type & i,
796                                              unsigned int & column)
797 {
798         char_type const c = text_[i];
799
800         if (style.pass_thru) {
801                 if (c != '\0')
802                         // FIXME UNICODE: This can fail if c cannot
803                         // be encoded in the current encoding.
804                         os.put(c);
805                 return;
806         }
807
808         if (runparams.verbatim) {
809                 os.put(c);
810                 return;
811         }
812
813         if (lyxrc.fontenc == "T1" && latexSpecialT1(c, os, i, column))
814                 return;
815
816         if (running_font.family() == Font::TYPEWRITER_FAMILY
817                 && latexSpecialTypewriter(c, os, i, column))
818                 return;
819
820         // Otherwise, we use what LaTeX provides us.
821         switch (c) {
822         case '\\':
823                 os << "\\textbackslash{}";
824                 column += 15;
825                 break;
826         case '<':
827                 os << "\\textless{}";
828                 column += 10;
829                 break;
830         case '>':
831                 os << "\\textgreater{}";
832                 column += 13;
833                 break;
834         case '|':
835                 os << "\\textbar{}";
836                 column += 9;
837                 break;
838         case '-':
839                 os << '-';
840                 break;
841         case '\"':
842                 os << "\\char`\\\"{}";
843                 column += 9;
844                 break;
845
846         case '$': case '&':
847         case '%': case '#': case '{':
848         case '}': case '_':
849                 os << '\\';
850                 os.put(c);
851                 column += 1;
852                 break;
853
854         case '~':
855                 os << "\\textasciitilde{}";
856                 column += 16;
857                 break;
858
859         case '^':
860                 os << "\\textasciicircum{}";
861                 column += 17;
862                 break;
863
864         case '*': case '[':
865                 // avoid being mistaken for optional arguments
866                 os << '{';
867                 os.put(c);
868                 os << '}';
869                 column += 2;
870                 break;
871
872         case ' ':
873                 // Blanks are printed before font switching.
874                 // Sure? I am not! (try nice-latex)
875                 // I am sure it's correct. LyX might be smarter
876                 // in the future, but for now, nothing wrong is
877                 // written. (Asger)
878                 break;
879
880         default:
881
882                 // LyX, LaTeX etc.
883                 if (latexSpecialPhrase(os, i, column))
884                         return;
885
886                 if (c == '\0')
887                         return;
888
889                 Encoding const & encoding = *(runparams.encoding);
890                 if (i + 1 < int(text_.size())) {
891                         char_type next = text_[i + 1];
892                         if (Encodings::isCombiningChar(next)) {
893                                 column += latexSurrogatePair(os, c, next, encoding) - 1;
894                                 ++i;
895                                 break;
896                         }
897                 }
898                 string script;
899                 docstring const latex = encoding.latexChar(c);
900                 if (Encodings::isKnownScriptChar(c, script)
901                     && prefixIs(latex, from_ascii("\\" + script)))
902                         column += writeScriptChars(os, c, latex,
903                                         running_change, encoding, i) - 1;
904                 else if (latex.length() > 1 && latex[latex.length() - 1] != '}') {
905                         // Prevent eating of a following
906                         // space or command corruption by
907                         // following characters
908                         column += latex.length() + 1;
909                         os << latex << "{}";
910                 } else {
911                         column += latex.length() - 1;
912                         os << latex;
913                 }
914                 break;
915         }
916 }
917
918
919 bool Paragraph::Private::latexSpecialT1(char_type const c, odocstream & os,
920         pos_type & i, unsigned int & column)
921 {
922         switch (c) {
923         case '>':
924         case '<':
925                 os.put(c);
926                 // In T1 encoding, these characters exist
927                 // but we should avoid ligatures
928                 if (i + 1 > int(text_.size()) || text_[i + 1] != c)
929                         return true;
930                 os << "\\,{}";
931                 column += 3;
932                 // Alternative code:
933                 //os << "\\textcompwordmark{}";
934                 //column += 19;
935                 return true;
936         case '|':
937                 os.put(c);
938                 return true;
939         default:
940                 return false;
941         }
942 }
943
944
945 bool Paragraph::Private::latexSpecialTypewriter(char_type const c, odocstream & os,
946         pos_type & i, unsigned int & column)
947 {
948         switch (c) {
949         case '-':
950                 if (i + 1 < int(text_.size()) && text_[i + 1] == '-') {
951                         // "--" in Typewriter mode -> "-{}-"
952                         os << "-{}";
953                         column += 2;
954                 } else
955                         os << '-';
956                 return true;
957
958         // I assume this is hack treating typewriter as verbatim
959         // FIXME UNICODE: This can fail if c cannot be encoded
960         // in the current encoding.
961
962         case '\0':
963                 return true;
964
965         // Those characters are not directly supported.
966         case '\\':
967         case '\"':
968         case '$': case '&':
969         case '%': case '#': case '{':
970         case '}': case '_':
971         case '~':
972         case '^':
973         case '*': case '[':
974         case ' ':
975                 return false;
976
977         default:
978                 // With Typewriter font, these characters exist.
979                 os.put(c);
980                 return true;
981         }
982 }
983
984
985 bool Paragraph::Private::latexSpecialPhrase(odocstream & os, pos_type & i,
986         unsigned int & column)
987 {
988         // FIXME: if we have "LaTeX" with a font
989         // change in the middle (before the 'T', then
990         // the "TeX" part is still special cased.
991         // Really we should only operate this on
992         // "words" for some definition of word
993
994         for (size_t pnr = 0; pnr < phrases_nr; ++pnr) {
995                 if (!isTextAt(special_phrases[pnr].phrase, i))
996                         continue;
997                 os << special_phrases[pnr].macro;
998                 i += special_phrases[pnr].phrase.length() - 1;
999                 column += special_phrases[pnr].macro.length() - 1;
1000                 return true;
1001         }
1002         return false;
1003 }
1004
1005
1006 void Paragraph::Private::validate(LaTeXFeatures & features,
1007                                 Layout const & layout) const
1008 {
1009         // check the params.
1010         if (!params_.spacing().isDefault())
1011                 features.require("setspace");
1012
1013         // then the layouts
1014         features.useLayout(layout.name());
1015
1016         // then the fonts
1017         fontlist_.validate(features);
1018
1019         // then the indentation
1020         if (!params_.leftIndent().zero())
1021                 features.require("ParagraphLeftIndent");
1022
1023         // then the insets
1024         InsetList::const_iterator icit = insetlist_.begin();
1025         InsetList::const_iterator iend = insetlist_.end();
1026         for (; icit != iend; ++icit) {
1027                 if (icit->inset) {
1028                         icit->inset->validate(features);
1029                         if (layout.needprotect &&
1030                             icit->inset->lyxCode() == FOOT_CODE)
1031                                 features.require("NeedLyXFootnoteCode");
1032                 }
1033         }
1034
1035         // then the contents
1036         for (pos_type i = 0; i < int(text_.size()) ; ++i) {
1037                 for (size_t pnr = 0; pnr < phrases_nr; ++pnr) {
1038                         if (!special_phrases[pnr].builtin
1039                             && isTextAt(special_phrases[pnr].phrase, i)) {
1040                                 features.require(special_phrases[pnr].phrase);
1041                                 break;
1042                         }
1043                 }
1044                 Encodings::validate(text_[i], features);
1045         }
1046 }
1047
1048 /////////////////////////////////////////////////////////////////////
1049 //
1050 // Paragraph
1051 //
1052 /////////////////////////////////////////////////////////////////////
1053
1054 Paragraph::Paragraph()
1055         : d(new Paragraph::Private(this))
1056 {
1057         itemdepth = 0;
1058         d->params_.clear();
1059 }
1060
1061
1062 Paragraph::Paragraph(Paragraph const & par)
1063         : itemdepth(par.itemdepth),
1064         d(new Paragraph::Private(*par.d, this))
1065 {
1066 }
1067
1068
1069 Paragraph & Paragraph::operator=(Paragraph const & par)
1070 {
1071         // needed as we will destroy the private part before copying it
1072         if (&par != this) {
1073                 itemdepth = par.itemdepth;
1074
1075                 delete d;
1076                 d = new Private(*par.d, this);
1077         }
1078         return *this;
1079 }
1080
1081
1082 Paragraph::~Paragraph()
1083 {
1084         delete d;
1085 }
1086
1087
1088 void Paragraph::write(Buffer const & buf, ostream & os,
1089                           BufferParams const & bparams,
1090                           depth_type & dth) const
1091 {
1092         // The beginning or end of a deeper (i.e. nested) area?
1093         if (dth != params().depth()) {
1094                 if (params().depth() > dth) {
1095                         while (params().depth() > dth) {
1096                                 os << "\n\\begin_deeper";
1097                                 ++dth;
1098                         }
1099                 } else {
1100                         while (params().depth() < dth) {
1101                                 os << "\n\\end_deeper";
1102                                 --dth;
1103                         }
1104                 }
1105         }
1106
1107         // First write the layout
1108         os << "\n\\begin_layout " << to_utf8(layout()->name()) << '\n';
1109
1110         params().write(os);
1111
1112         Font font1(Font::ALL_INHERIT, bparams.language);
1113
1114         Change running_change = Change(Change::UNCHANGED);
1115
1116         int column = 0;
1117         for (pos_type i = 0; i <= size(); ++i) {
1118
1119                 Change change = lookupChange(i);
1120                 Changes::lyxMarkChange(os, column, running_change, change);
1121                 running_change = change;
1122
1123                 if (i == size())
1124                         break;
1125
1126                 // Write font changes
1127                 Font font2 = getFontSettings(bparams, i);
1128                 if (font2 != font1) {
1129                         font2.lyxWriteChanges(font1, os);
1130                         column = 0;
1131                         font1 = font2;
1132                 }
1133
1134                 char_type const c = d->text_[i];
1135                 switch (c) {
1136                 case META_INSET:
1137                 {
1138                         Inset const * inset = getInset(i);
1139                         if (inset)
1140                                 if (inset->directWrite()) {
1141                                         // international char, let it write
1142                                         // code directly so it's shorter in
1143                                         // the file
1144                                         inset->write(buf, os);
1145                                 } else {
1146                                         if (i)
1147                                                 os << '\n';
1148                                         os << "\\begin_inset ";
1149                                         inset->write(buf, os);
1150                                         os << "\n\\end_inset\n\n";
1151                                         column = 0;
1152                                 }
1153                 }
1154                 break;
1155                 case '\\':
1156                         os << "\n\\backslash\n";
1157                         column = 0;
1158                         break;
1159                 case '.':
1160                         if (i + 1 < size() && d->text_[i + 1] == ' ') {
1161                                 os << ".\n";
1162                                 column = 0;
1163                         } else
1164                                 os << '.';
1165                         break;
1166                 default:
1167                         if ((column > 70 && c == ' ')
1168                             || column > 79) {
1169                                 os << '\n';
1170                                 column = 0;
1171                         }
1172                         // this check is to amend a bug. LyX sometimes
1173                         // inserts '\0' this could cause problems.
1174                         if (c != '\0') {
1175                                 std::vector<char> tmp = ucs4_to_utf8(c);
1176                                 tmp.push_back('\0');
1177                                 os << &tmp[0];
1178                         } else
1179                                 lyxerr << "ERROR (Paragraph::writeFile):"
1180                                         " NULL char in structure." << endl;
1181                         ++column;
1182                         break;
1183                 }
1184         }
1185
1186         os << "\n\\end_layout\n";
1187 }
1188
1189
1190 void Paragraph::validate(LaTeXFeatures & features) const
1191 {
1192         d->validate(features, *layout());
1193 }
1194
1195
1196 void Paragraph::insert(pos_type start, docstring const & str,
1197                        Font const & font, Change const & change)
1198 {
1199         for (size_t i = 0, n = str.size(); i != n ; ++i)
1200                 insertChar(start + i, str[i], font, change);
1201 }
1202
1203
1204 void Paragraph::appendChar(char_type c, Font const & font,
1205                 Change const & change)
1206 {
1207         // track change
1208         d->changes_.insert(change, d->text_.size());
1209         // when appending characters, no need to update tables
1210         d->text_.push_back(c);
1211         setFont(d->text_.size() - 1, font);
1212 }
1213
1214
1215 void Paragraph::appendString(docstring const & s, Font const & font,
1216                 Change const & change)
1217 {
1218         size_t end = s.size();
1219         size_t oldsize = d->text_.size();
1220         size_t newsize = oldsize + end;
1221         size_t capacity = d->text_.capacity();
1222         if (newsize >= capacity)
1223                 d->text_.reserve(std::max(capacity + 100, newsize));
1224
1225         // when appending characters, no need to update tables
1226         d->text_.append(s);
1227
1228         // FIXME: Optimize this!
1229         for (pos_type i = 0; i != end; ++i) {
1230                 // track change
1231                 d->changes_.insert(change, i);
1232         }
1233         d->fontlist_.setRange(oldsize, newsize, font);
1234 }
1235
1236
1237 void Paragraph::insertChar(pos_type pos, char_type c,
1238                            bool trackChanges)
1239 {
1240         d->insertChar(pos, c, Change(trackChanges ?
1241                            Change::INSERTED : Change::UNCHANGED));
1242 }
1243
1244
1245 void Paragraph::insertChar(pos_type pos, char_type c,
1246                            Font const & font, bool trackChanges)
1247 {
1248         d->insertChar(pos, c, Change(trackChanges ?
1249                            Change::INSERTED : Change::UNCHANGED));
1250         setFont(pos, font);
1251 }
1252
1253
1254 void Paragraph::insertChar(pos_type pos, char_type c,
1255                            Font const & font, Change const & change)
1256 {
1257         d->insertChar(pos, c, change);
1258         setFont(pos, font);
1259 }
1260
1261
1262 void Paragraph::insertInset(pos_type pos, Inset * inset,
1263                             Font const & font, Change const & change)
1264 {
1265         insertInset(pos, inset, change);
1266         // Set the font/language of the inset...
1267         setFont(pos, font);
1268 }
1269
1270
1271 bool Paragraph::insetAllowed(InsetCode code)
1272 {
1273         return !d->inset_owner_ || d->inset_owner_->insetAllowed(code);
1274 }
1275
1276
1277 // Gets uninstantiated font setting at position.
1278 Font const Paragraph::getFontSettings(BufferParams const & bparams,
1279                                          pos_type pos) const
1280 {
1281         if (pos > size()) {
1282                 lyxerr << " pos: " << pos << " size: " << size() << endl;
1283                 BOOST_ASSERT(pos <= size());
1284         }
1285
1286         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
1287         if (cit != d->fontlist_.end())
1288                 return cit->font();
1289
1290         if (pos == size() && !empty())
1291                 return getFontSettings(bparams, pos - 1);
1292
1293         return Font(Font::ALL_INHERIT, getParLanguage(bparams));
1294 }
1295
1296
1297 FontSpan Paragraph::fontSpan(pos_type pos) const
1298 {
1299         BOOST_ASSERT(pos <= size());
1300         pos_type start = 0;
1301
1302         FontList::const_iterator cit = d->fontlist_.begin();
1303         FontList::const_iterator end = d->fontlist_.end();
1304         for (; cit != end; ++cit) {
1305                 if (cit->pos() >= pos) {
1306                         if (pos >= beginOfBody())
1307                                 return FontSpan(std::max(start, beginOfBody()),
1308                                                 cit->pos());
1309                         else
1310                                 return FontSpan(start,
1311                                                 std::min(beginOfBody() - 1,
1312                                                          cit->pos()));
1313                 }
1314                 start = cit->pos() + 1;
1315         }
1316
1317         // This should not happen, but if so, we take no chances.
1318         //lyxerr << "Paragraph::getEndPosOfFontSpan: This should not happen!"
1319         //      << endl;
1320         return FontSpan(pos, pos);
1321 }
1322
1323
1324 // Gets uninstantiated font setting at position 0
1325 Font const Paragraph::getFirstFontSettings(BufferParams const & bparams) const
1326 {
1327         if (!empty() && !d->fontlist_.empty())
1328                 return d->fontlist_.begin()->font();
1329
1330         return Font(Font::ALL_INHERIT, bparams.language);
1331 }
1332
1333
1334 // Gets the fully instantiated font at a given position in a paragraph
1335 // This is basically the same function as Text::GetFont() in text2.cpp.
1336 // The difference is that this one is used for generating the LaTeX file,
1337 // and thus cosmetic "improvements" are disallowed: This has to deliver
1338 // the true picture of the buffer. (Asger)
1339 Font const Paragraph::getFont(BufferParams const & bparams, pos_type pos,
1340                                  Font const & outerfont) const
1341 {
1342         BOOST_ASSERT(pos >= 0);
1343
1344         Font font = getFontSettings(bparams, pos);
1345
1346         pos_type const body_pos = beginOfBody();
1347         if (pos < body_pos)
1348                 font.realize(d->layout_->labelfont);
1349         else
1350                 font.realize(d->layout_->font);
1351
1352         font.realize(outerfont);
1353         font.realize(bparams.getFont());
1354
1355         return font;
1356 }
1357
1358
1359 Font const Paragraph::getLabelFont
1360         (BufferParams const & bparams, Font const & outerfont) const
1361 {
1362         Font tmpfont = layout()->labelfont;
1363         tmpfont.setLanguage(getParLanguage(bparams));
1364         tmpfont.realize(outerfont);
1365         tmpfont.realize(bparams.getFont());
1366         return tmpfont;
1367 }
1368
1369
1370 Font const Paragraph::getLayoutFont
1371         (BufferParams const & bparams, Font const & outerfont) const
1372 {
1373         Font tmpfont = layout()->font;
1374         tmpfont.setLanguage(getParLanguage(bparams));
1375         tmpfont.realize(outerfont);
1376         tmpfont.realize(bparams.getFont());
1377         return tmpfont;
1378 }
1379
1380
1381 /// Returns the height of the highest font in range
1382 Font_size Paragraph::highestFontInRange
1383         (pos_type startpos, pos_type endpos, Font_size def_size) const
1384 {
1385         return d->fontlist_.highestInRange(startpos, endpos, def_size);
1386 }
1387
1388
1389 char_type
1390 Paragraph::getUChar(BufferParams const & bparams, pos_type pos) const
1391 {
1392         char_type c = d->text_[pos];
1393         if (!lyxrc.rtl_support)
1394                 return c;
1395
1396         char_type uc = c;
1397         switch (c) {
1398         case '(':
1399                 uc = ')';
1400                 break;
1401         case ')':
1402                 uc = '(';
1403                 break;
1404         case '[':
1405                 uc = ']';
1406                 break;
1407         case ']':
1408                 uc = '[';
1409                 break;
1410         case '{':
1411                 uc = '}';
1412                 break;
1413         case '}':
1414                 uc = '{';
1415                 break;
1416         case '<':
1417                 uc = '>';
1418                 break;
1419         case '>':
1420                 uc = '<';
1421                 break;
1422         }
1423         if (uc != c && getFontSettings(bparams, pos).isRightToLeft())
1424                 return uc;
1425         else
1426                 return c;
1427 }
1428
1429
1430 void Paragraph::setFont(pos_type pos, Font const & font)
1431 {
1432         BOOST_ASSERT(pos <= size());
1433
1434         // First, reduce font against layout/label font
1435         // Update: The setCharFont() routine in text2.cpp already
1436         // reduces font, so we don't need to do that here. (Asger)
1437         
1438         d->fontlist_.set(pos, font);
1439 }
1440
1441
1442 void Paragraph::makeSameLayout(Paragraph const & par)
1443 {
1444         layout(par.layout());
1445         // move to pimpl?
1446         d->params_ = par.params();
1447 }
1448
1449
1450 bool Paragraph::stripLeadingSpaces(bool trackChanges)
1451 {
1452         if (isFreeSpacing())
1453                 return false;
1454
1455         int pos = 0;
1456         int count = 0;
1457
1458         while (pos < size() && (isNewline(pos) || isLineSeparator(pos))) {
1459                 if (eraseChar(pos, trackChanges))
1460                         ++count;
1461                 else
1462                         ++pos;
1463         }
1464
1465         return count > 0 || pos > 0;
1466 }
1467
1468
1469 bool Paragraph::hasSameLayout(Paragraph const & par) const
1470 {
1471         return par.layout() == layout() && d->params_.sameLayout(par.params());
1472 }
1473
1474
1475 depth_type Paragraph::getDepth() const
1476 {
1477         return params().depth();
1478 }
1479
1480
1481 depth_type Paragraph::getMaxDepthAfter() const
1482 {
1483         if (layout()->isEnvironment())
1484                 return params().depth() + 1;
1485         else
1486                 return params().depth();
1487 }
1488
1489
1490 char Paragraph::getAlign() const
1491 {
1492         if (params().align() == LYX_ALIGN_LAYOUT)
1493                 return layout()->align;
1494         else
1495                 return params().align();
1496 }
1497
1498
1499 docstring const & Paragraph::getLabelstring() const
1500 {
1501         return params().labelString();
1502 }
1503
1504
1505 // the next two functions are for the manual labels
1506 docstring const Paragraph::getLabelWidthString() const
1507 {
1508         if (layout()->margintype == MARGIN_MANUAL)
1509                 return params().labelWidthString();
1510         else
1511                 return _("Senseless with this layout!");
1512 }
1513
1514
1515 void Paragraph::setLabelWidthString(docstring const & s)
1516 {
1517         params().labelWidthString(s);
1518 }
1519
1520
1521 docstring const Paragraph::translateIfPossible(docstring const & s,
1522                 BufferParams const & bparams) const
1523 {
1524         if (!support::isAscii(s) || s.empty()) {
1525                 // This must be a user defined layout. We cannot translate
1526                 // this, since gettext accepts only ascii keys.
1527                 return s;
1528         }
1529         // Probably standard layout, try to translate
1530         Messages & m = getMessages(getParLanguage(bparams)->code());
1531         return m.get(to_ascii(s));
1532 }
1533
1534
1535 docstring Paragraph::expandLabel(LayoutPtr const & layout,
1536                 BufferParams const & bparams, bool process_appendix) const
1537 {
1538         TextClass const & tclass = bparams.getTextClass();
1539
1540         docstring fmt;
1541         if (process_appendix && params().appendix())
1542                 fmt = translateIfPossible(layout->labelstring_appendix(),
1543                         bparams);
1544         else
1545                 fmt = translateIfPossible(layout->labelstring(), bparams);
1546
1547         if (fmt.empty() && layout->labeltype == LABEL_COUNTER 
1548             && !layout->counter.empty())
1549                 fmt = "\\the" + layout->counter;
1550
1551         // handle 'inherited level parts' in 'fmt',
1552         // i.e. the stuff between '@' in   '@Section@.\arabic{subsection}'
1553         size_t const i = fmt.find('@', 0);
1554         if (i != docstring::npos) {
1555                 size_t const j = fmt.find('@', i + 1);
1556                 if (j != docstring::npos) {
1557                         docstring parent(fmt, i + 1, j - i - 1);
1558                         docstring label = from_ascii("??");
1559                         if (tclass.hasLayout(parent))
1560                                 docstring label = expandLabel(tclass[parent], bparams,
1561                                                       process_appendix);
1562                         fmt = docstring(fmt, 0, i) + label 
1563                                 + docstring(fmt, j + 1, docstring::npos);
1564                 }
1565         }
1566
1567         return tclass.counters().counterLabel(fmt);
1568 }
1569
1570
1571 void Paragraph::applyLayout(LayoutPtr const & new_layout)
1572 {
1573         layout(new_layout);
1574         LyXAlignment const oldAlign = params().align();
1575         
1576         if (!(oldAlign & layout()->alignpossible)) {
1577                 frontend::Alert::warning(_("Alignment not permitted"), 
1578                         _("The new layout does not permit the alignment previously used.\nSetting to default."));
1579                 params().align(LYX_ALIGN_LAYOUT);
1580         }
1581 }
1582
1583
1584 pos_type Paragraph::beginOfBody() const
1585 {
1586         return d->begin_of_body_;
1587 }
1588
1589
1590 void Paragraph::setBeginOfBody()
1591 {
1592         if (layout()->labeltype != LABEL_MANUAL) {
1593                 d->begin_of_body_ = 0;
1594                 return;
1595         }
1596
1597         // Unroll the first two cycles of the loop
1598         // and remember the previous character to
1599         // remove unnecessary getChar() calls
1600         pos_type i = 0;
1601         pos_type end = size();
1602         if (i < end && !isNewline(i)) {
1603                 ++i;
1604                 char_type previous_char = 0;
1605                 char_type temp = 0;
1606                 if (i < end) {
1607                         previous_char = d->text_[i];
1608                         if (!isNewline(i)) {
1609                                 ++i;
1610                                 while (i < end && previous_char != ' ') {
1611                                         temp = d->text_[i];
1612                                         if (isNewline(i))
1613                                                 break;
1614                                         ++i;
1615                                         previous_char = temp;
1616                                 }
1617                         }
1618                 }
1619         }
1620
1621         d->begin_of_body_ = i;
1622 }
1623
1624
1625 bool Paragraph::forceDefaultParagraphs() const
1626 {
1627         return inInset() && inInset()->forceDefaultParagraphs(0);
1628 }
1629
1630
1631 namespace {
1632
1633 // paragraphs inside floats need different alignment tags to avoid
1634 // unwanted space
1635
1636 bool noTrivlistCentering(InsetCode code)
1637 {
1638         return code == FLOAT_CODE || code == WRAP_CODE;
1639 }
1640
1641
1642 string correction(string const & orig)
1643 {
1644         if (orig == "flushleft")
1645                 return "raggedright";
1646         if (orig == "flushright")
1647                 return "raggedleft";
1648         if (orig == "center")
1649                 return "centering";
1650         return orig;
1651 }
1652
1653
1654 string const corrected_env(string const & suffix, string const & env,
1655         InsetCode code)
1656 {
1657         string output = suffix + "{";
1658         if (noTrivlistCentering(code))
1659                 output += correction(env);
1660         else
1661                 output += env;
1662         output += "}";
1663         if (suffix == "\\begin")
1664                 output += "\n";
1665         return output;
1666 }
1667
1668
1669 void adjust_row_column(string const & str, TexRow & texrow, int & column)
1670 {
1671         if (!contains(str, "\n"))
1672                 column += str.size();
1673         else {
1674                 string tmp;
1675                 texrow.newline();
1676                 column = rsplit(str, tmp, '\n').size();
1677         }
1678 }
1679
1680 } // namespace anon
1681
1682
1683 int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
1684                                  odocstream & os, TexRow & texrow,
1685                                  bool moving_arg) const
1686 {
1687         int column = 0;
1688
1689         if (params_.noindent()) {
1690                 os << "\\noindent ";
1691                 column += 10;
1692         }
1693         
1694         LyXAlignment const curAlign = params_.align();
1695
1696         if (curAlign == layout_->align)
1697                 return column;
1698
1699         switch (curAlign) {
1700         case LYX_ALIGN_NONE:
1701         case LYX_ALIGN_BLOCK:
1702         case LYX_ALIGN_LAYOUT:
1703         case LYX_ALIGN_SPECIAL:
1704                 break;
1705         case LYX_ALIGN_LEFT:
1706         case LYX_ALIGN_RIGHT:
1707         case LYX_ALIGN_CENTER:
1708                 if (moving_arg) {
1709                         os << "\\protect";
1710                         column += 8;
1711                 }
1712                 break;
1713         }
1714
1715         switch (curAlign) {
1716         case LYX_ALIGN_NONE:
1717         case LYX_ALIGN_BLOCK:
1718         case LYX_ALIGN_LAYOUT:
1719         case LYX_ALIGN_SPECIAL:
1720                 break;
1721         case LYX_ALIGN_LEFT: {
1722                 string output;
1723                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
1724                         output = corrected_env("\\begin", "flushleft", owner_->ownerCode());
1725                 else
1726                         output = corrected_env("\\begin", "flushright", owner_->ownerCode());
1727                 os << from_ascii(output);
1728                 adjust_row_column(output, texrow, column);
1729                 break;
1730         } case LYX_ALIGN_RIGHT: {
1731                 string output;
1732                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
1733                         output = corrected_env("\\begin", "flushright", owner_->ownerCode());
1734                 else
1735                         output = corrected_env("\\begin", "flushleft", owner_->ownerCode());
1736                 os << from_ascii(output);
1737                 adjust_row_column(output, texrow, column);
1738                 break;
1739         } case LYX_ALIGN_CENTER: {
1740                 string output;
1741                 output = corrected_env("\\begin", "center", owner_->ownerCode());
1742                 os << from_ascii(output);
1743                 adjust_row_column(output, texrow, column);
1744                 break;
1745         }
1746         }
1747
1748         return column;
1749 }
1750
1751
1752 int Paragraph::Private::endTeXParParams(BufferParams const & bparams,
1753                                odocstream & os, TexRow & texrow,
1754                                bool moving_arg) const
1755 {
1756         int column = 0;
1757
1758         switch (params_.align()) {
1759         case LYX_ALIGN_NONE:
1760         case LYX_ALIGN_BLOCK:
1761         case LYX_ALIGN_LAYOUT:
1762         case LYX_ALIGN_SPECIAL:
1763                 break;
1764         case LYX_ALIGN_LEFT:
1765         case LYX_ALIGN_RIGHT:
1766         case LYX_ALIGN_CENTER:
1767                 if (moving_arg) {
1768                         os << "\\protect";
1769                         column = 8;
1770                 }
1771                 break;
1772         }
1773
1774         switch (params_.align()) {
1775         case LYX_ALIGN_NONE:
1776         case LYX_ALIGN_BLOCK:
1777         case LYX_ALIGN_LAYOUT:
1778         case LYX_ALIGN_SPECIAL:
1779                 break;
1780         case LYX_ALIGN_LEFT: {
1781                 string output;
1782                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
1783                         output = corrected_env("\n\\par\\end", "flushleft", owner_->ownerCode());
1784                 else
1785                         output = corrected_env("\n\\par\\end", "flushright", owner_->ownerCode());
1786                 os << from_ascii(output);
1787                 adjust_row_column(output, texrow, column);
1788                 break;
1789         } case LYX_ALIGN_RIGHT: {
1790                 string output;
1791                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
1792                         output = corrected_env("\n\\par\\end", "flushright", owner_->ownerCode());
1793                 else
1794                         output = corrected_env("\n\\par\\end", "flushleft", owner_->ownerCode());
1795                 os << from_ascii(output);
1796                 adjust_row_column(output, texrow, column);
1797                 break;
1798         } case LYX_ALIGN_CENTER: {
1799                 string output;
1800                 output = corrected_env("\n\\par\\end", "center", owner_->ownerCode());
1801                 os << from_ascii(output);
1802                 adjust_row_column(output, texrow, column);
1803                 break;
1804         }
1805         }
1806
1807         return column;
1808 }
1809
1810
1811 // This one spits out the text of the paragraph
1812 bool Paragraph::latex(Buffer const & buf,
1813                                 BufferParams const & bparams,
1814                                 Font const & outerfont,
1815                                 odocstream & os, TexRow & texrow,
1816                                 OutputParams const & runparams) const
1817 {
1818         LYXERR(Debug::LATEX) << "SimpleTeXOnePar...     " << this << endl;
1819
1820         bool return_value = false;
1821
1822         LayoutPtr style;
1823
1824         // well we have to check if we are in an inset with unlimited
1825         // length (all in one row) if that is true then we don't allow
1826         // any special options in the paragraph and also we don't allow
1827         // any environment other than the default layout of the text class
1828         // to be valid!
1829         bool asdefault = forceDefaultParagraphs();
1830
1831         if (asdefault) {
1832                 style = bparams.getTextClass().defaultLayout();
1833         } else {
1834                 style = layout();
1835         }
1836
1837         // Current base font for all inherited font changes, without any
1838         // change caused by an individual character, except for the language:
1839         // It is set to the language of the first character.
1840         // As long as we are in the label, this font is the base font of the
1841         // label. Before the first body character it is set to the base font
1842         // of the body.
1843         Font basefont;
1844
1845         // Maybe we have to create a optional argument.
1846         pos_type body_pos = beginOfBody();
1847         unsigned int column = 0;
1848
1849         if (body_pos > 0) {
1850                 // the optional argument is kept in curly brackets in
1851                 // case it contains a ']'
1852                 os << "[{";
1853                 column += 2;
1854                 basefont = getLabelFont(bparams, outerfont);
1855         } else {
1856                 basefont = getLayoutFont(bparams, outerfont);
1857         }
1858
1859         // Which font is currently active?
1860         Font running_font(basefont);
1861         // Do we have an open font change?
1862         bool open_font = false;
1863
1864         Change runningChange = Change(Change::UNCHANGED);
1865
1866         texrow.start(id(), 0);
1867
1868         // if the paragraph is empty, the loop will not be entered at all
1869         if (empty()) {
1870                 if (style->isCommand()) {
1871                         os << '{';
1872                         ++column;
1873                 }
1874                 if (!asdefault)
1875                         column += d->startTeXParParams(bparams, os, texrow,
1876                                                     runparams.moving_arg);
1877         }
1878
1879         for (pos_type i = 0; i < size(); ++i) {
1880                 // First char in paragraph or after label?
1881                 if (i == body_pos) {
1882                         if (body_pos > 0) {
1883                                 if (open_font) {
1884                                         column += running_font.latexWriteEndChanges(
1885                                                 os, bparams, runparams,
1886                                                 basefont, basefont);
1887                                         open_font = false;
1888                                 }
1889                                 basefont = getLayoutFont(bparams, outerfont);
1890                                 running_font = basefont;
1891
1892                                 column += Changes::latexMarkChange(os, bparams,
1893                                                 runningChange, Change(Change::UNCHANGED));
1894                                 runningChange = Change(Change::UNCHANGED);
1895
1896                                 os << "}] ";
1897                                 column +=3;
1898                         }
1899                         if (style->isCommand()) {
1900                                 os << '{';
1901                                 ++column;
1902                         }
1903
1904                         if (!asdefault)
1905                                 column += d->startTeXParParams(bparams, os,
1906                                                             texrow,
1907                                                             runparams.moving_arg);
1908                 }
1909
1910                 Change const & change = runparams.inDeletedInset ? runparams.changeOfDeletedInset
1911                                                                  : lookupChange(i);
1912
1913                 if (bparams.outputChanges && runningChange != change) {
1914                         if (open_font) {
1915                                 column += running_font.latexWriteEndChanges(
1916                                                 os, bparams, runparams, basefont, basefont);
1917                                 open_font = false;
1918                         }
1919                         basefont = getLayoutFont(bparams, outerfont);
1920                         running_font = basefont;
1921
1922                         column += Changes::latexMarkChange(os, bparams, runningChange, change);
1923                         runningChange = change;
1924                 }
1925
1926                 // do not output text which is marked deleted
1927                 // if change tracking output is disabled
1928                 if (!bparams.outputChanges && change.type == Change::DELETED) {
1929                         continue;
1930                 }
1931
1932                 ++column;
1933
1934                 // Fully instantiated font
1935                 Font const font = getFont(bparams, i, outerfont);
1936
1937                 Font const last_font = running_font;
1938
1939                 // Do we need to close the previous font?
1940                 if (open_font &&
1941                     (font != running_font ||
1942                      font.language() != running_font.language()))
1943                 {
1944                         column += running_font.latexWriteEndChanges(
1945                                         os, bparams, runparams, basefont,
1946                                         (i == body_pos-1) ? basefont : font);
1947                         running_font = basefont;
1948                         open_font = false;
1949                 }
1950
1951                 // Switch file encoding if necessary (and allowed)
1952                 if (!runparams.verbatim && 
1953                     runparams.encoding->package() == Encoding::inputenc &&
1954                     font.language()->encoding()->package() == Encoding::inputenc) {
1955                         std::pair<bool, int> const enc_switch = switchEncoding(os, bparams,
1956                                         runparams.moving_arg, *(runparams.encoding),
1957                                         *(font.language()->encoding()));
1958                         if (enc_switch.first) {
1959                                 column += enc_switch.second;
1960                                 runparams.encoding = font.language()->encoding();
1961                         }
1962                 }
1963
1964                 char_type const c = d->text_[i];
1965
1966                 // Do we need to change font?
1967                 if ((font != running_font ||
1968                      font.language() != running_font.language()) &&
1969                         i != body_pos - 1)
1970                 {
1971                         odocstringstream ods;
1972                         column += font.latexWriteStartChanges(ods, bparams,
1973                                                               runparams, basefont,
1974                                                               last_font);
1975                         running_font = font;
1976                         open_font = true;
1977                         docstring fontchange = ods.str();
1978                         // check if the fontchange ends with a trailing blank
1979                         // (like "\small " (see bug 3382)
1980                         if (suffixIs(fontchange, ' ') && c == ' ')
1981                                 os << fontchange.substr(0, fontchange.size() - 1) 
1982                                    << from_ascii("{}");
1983                         else
1984                                 os << fontchange;
1985                 }
1986
1987                 if (c == ' ') {
1988                         // FIXME: integrate this case in latexSpecialChar
1989                         // Do not print the separation of the optional argument
1990                         // if style->pass_thru is false. This works because
1991                         // latexSpecialChar ignores spaces if
1992                         // style->pass_thru is false.
1993                         if (i != body_pos - 1) {
1994                                 if (d->simpleTeXBlanks(
1995                                                 *(runparams.encoding), os, texrow,
1996                                                 i, column, font, *style)) {
1997                                         // A surrogate pair was output. We
1998                                         // must not call latexSpecialChar
1999                                         // in this iteration, since it would output
2000                                         // the combining character again.
2001                                         ++i;
2002                                         continue;
2003                                 }
2004                         }
2005                 }
2006
2007                 OutputParams rp = runparams;
2008                 rp.free_spacing = style->free_spacing;
2009                 rp.local_font = &font;
2010                 rp.intitle = style->intitle;
2011
2012                 // Two major modes:  LaTeX or plain
2013                 // Handle here those cases common to both modes
2014                 // and then split to handle the two modes separately.
2015                 if (c == META_INSET)
2016                         d->latexInset(buf, bparams, os,
2017                                         texrow, rp, running_font,
2018                                         basefont, outerfont, open_font,
2019                                         runningChange, *style, i, column);
2020                 else
2021                         d->latexSpecialChar(os, rp, running_font, runningChange,
2022                                 *style, i, column);
2023
2024                 // Set the encoding to that returned from simpleTeXSpecialChars (see
2025                 // comment for encoding member in OutputParams.h)
2026                 runparams.encoding = rp.encoding;
2027         }
2028
2029         // If we have an open font definition, we have to close it
2030         if (open_font) {
2031 #ifdef FIXED_LANGUAGE_END_DETECTION
2032                 if (next_) {
2033                         running_font
2034                                 .latexWriteEndChanges(os, bparams, runparams,
2035                                         basefont,
2036                                         next_->getFont(bparams, 0, outerfont));
2037                 } else {
2038                         running_font.latexWriteEndChanges(os, bparams,
2039                                         runparams, basefont, basefont);
2040                 }
2041 #else
2042 //FIXME: For now we ALWAYS have to close the foreign font settings if they are
2043 //FIXME: there as we start another \selectlanguage with the next paragraph if
2044 //FIXME: we are in need of this. This should be fixed sometime (Jug)
2045                 running_font.latexWriteEndChanges(os, bparams, runparams,
2046                                 basefont, basefont);
2047 #endif
2048         }
2049
2050         column += Changes::latexMarkChange(os, bparams, runningChange, Change(Change::UNCHANGED));
2051
2052         // Needed if there is an optional argument but no contents.
2053         if (body_pos > 0 && body_pos == size()) {
2054                 os << "}]~";
2055                 return_value = false;
2056         }
2057
2058         if (!asdefault) {
2059                 column += d->endTeXParParams(bparams, os, texrow,
2060                                           runparams.moving_arg);
2061         }
2062
2063         LYXERR(Debug::LATEX) << "SimpleTeXOnePar...done " << this << endl;
2064         return return_value;
2065 }
2066
2067
2068 bool Paragraph::emptyTag() const
2069 {
2070         for (pos_type i = 0; i < size(); ++i) {
2071                 if (isInset(i)) {
2072                         Inset const * inset = getInset(i);
2073                         InsetCode lyx_code = inset->lyxCode();
2074                         if (lyx_code != TOC_CODE &&
2075                             lyx_code != INCLUDE_CODE &&
2076                             lyx_code != GRAPHICS_CODE &&
2077                             lyx_code != ERT_CODE &&
2078                             lyx_code != LISTINGS_CODE &&
2079                             lyx_code != FLOAT_CODE &&
2080                             lyx_code != TABULAR_CODE) {
2081                                 return false;
2082                         }
2083                 } else {
2084                         char_type c = d->text_[i];
2085                         if (c != ' ' && c != '\t')
2086                                 return false;
2087                 }
2088         }
2089         return true;
2090 }
2091
2092
2093 string Paragraph::getID(Buffer const & buf, OutputParams const & runparams) const
2094 {
2095         for (pos_type i = 0; i < size(); ++i) {
2096                 if (isInset(i)) {
2097                         Inset const * inset = getInset(i);
2098                         InsetCode lyx_code = inset->lyxCode();
2099                         if (lyx_code == LABEL_CODE) {
2100                                 InsetLabel const * const il = static_cast<InsetLabel const *>(inset);
2101                                 docstring const & id = il->getParam("name");
2102                                 return "id='" + to_utf8(sgml::cleanID(buf, runparams, id)) + "'";
2103                         }
2104                 }
2105
2106         }
2107         return string();
2108 }
2109
2110
2111 pos_type Paragraph::getFirstWord(Buffer const & buf, odocstream & os, OutputParams const & runparams) const
2112 {
2113         pos_type i;
2114         for (i = 0; i < size(); ++i) {
2115                 if (isInset(i)) {
2116                         Inset const * inset = getInset(i);
2117                         inset->docbook(buf, os, runparams);
2118                 } else {
2119                         char_type c = d->text_[i];
2120                         if (c == ' ')
2121                                 break;
2122                         os << sgml::escapeChar(c);
2123                 }
2124         }
2125         return i;
2126 }
2127
2128
2129 bool Paragraph::Private::onlyText(Buffer const & buf, Font const & outerfont, pos_type initial) const
2130 {
2131         Font font_old;
2132         pos_type size = text_.size();
2133         for (pos_type i = initial; i < size; ++i) {
2134                 Font font = owner_->getFont(buf.params(), i, outerfont);
2135                 if (text_[i] == META_INSET)
2136                         return false;
2137                 if (i != initial && font != font_old)
2138                         return false;
2139                 font_old = font;
2140         }
2141
2142         return true;
2143 }
2144
2145
2146 void Paragraph::simpleDocBookOnePar(Buffer const & buf,
2147                                     odocstream & os,
2148                                     OutputParams const & runparams,
2149                                     Font const & outerfont,
2150                                     pos_type initial) const
2151 {
2152         bool emph_flag = false;
2153
2154         LayoutPtr const & style = layout();
2155         Font font_old =
2156                 style->labeltype == LABEL_MANUAL ? style->labelfont : style->font;
2157
2158         if (style->pass_thru && !d->onlyText(buf, outerfont, initial))
2159                 os << "]]>";
2160
2161         // parsing main loop
2162         for (pos_type i = initial; i < size(); ++i) {
2163                 Font font = getFont(buf.params(), i, outerfont);
2164
2165                 // handle <emphasis> tag
2166                 if (font_old.emph() != font.emph()) {
2167                         if (font.emph() == Font::ON) {
2168                                 os << "<emphasis>";
2169                                 emph_flag = true;
2170                         } else if (i != initial) {
2171                                 os << "</emphasis>";
2172                                 emph_flag = false;
2173                         }
2174                 }
2175
2176                 if (isInset(i)) {
2177                         Inset const * inset = getInset(i);
2178                         inset->docbook(buf, os, runparams);
2179                 } else {
2180                         char_type c = d->text_[i];
2181
2182                         if (style->pass_thru)
2183                                 os.put(c);
2184                         else
2185                                 os << sgml::escapeChar(c);
2186                 }
2187                 font_old = font;
2188         }
2189
2190         if (emph_flag) {
2191                 os << "</emphasis>";
2192         }
2193
2194         if (style->free_spacing)
2195                 os << '\n';
2196         if (style->pass_thru && !d->onlyText(buf, outerfont, initial))
2197                 os << "<![CDATA[";
2198 }
2199
2200
2201 bool Paragraph::isHfill(pos_type pos) const
2202 {
2203         return isInset(pos)
2204                 && getInset(pos)->lyxCode() == HFILL_CODE;
2205 }
2206
2207
2208 bool Paragraph::isNewline(pos_type pos) const
2209 {
2210         return isInset(pos)
2211                 && getInset(pos)->lyxCode() == NEWLINE_CODE;
2212 }
2213
2214
2215 bool Paragraph::isLineSeparator(pos_type pos) const
2216 {
2217         char_type const c = d->text_[pos];
2218         return isLineSeparatorChar(c)
2219                 || (c == META_INSET && getInset(pos) &&
2220                 getInset(pos)->isLineSeparator());
2221 }
2222
2223
2224 /// Used by the spellchecker
2225 bool Paragraph::isLetter(pos_type pos) const
2226 {
2227         if (isInset(pos))
2228                 return getInset(pos)->isLetter();
2229         else {
2230                 char_type const c = d->text_[pos];
2231                 return isLetterChar(c) || isDigit(c);
2232         }
2233 }
2234
2235
2236 Language const *
2237 Paragraph::getParLanguage(BufferParams const & bparams) const
2238 {
2239         if (!empty())
2240                 return getFirstFontSettings(bparams).language();
2241         // FIXME: we should check the prev par as well (Lgb)
2242         return bparams.language;
2243 }
2244
2245
2246 bool Paragraph::isRTL(BufferParams const & bparams) const
2247 {
2248         return lyxrc.rtl_support
2249                 && getParLanguage(bparams)->rightToLeft()
2250                 && ownerCode() != ERT_CODE
2251                 && ownerCode() != LISTINGS_CODE;
2252 }
2253
2254
2255 void Paragraph::changeLanguage(BufferParams const & bparams,
2256                                Language const * from, Language const * to)
2257 {
2258         // change language including dummy font change at the end
2259         for (pos_type i = 0; i <= size(); ++i) {
2260                 Font font = getFontSettings(bparams, i);
2261                 if (font.language() == from) {
2262                         font.setLanguage(to);
2263                         setFont(i, font);
2264                 }
2265         }
2266 }
2267
2268
2269 bool Paragraph::isMultiLingual(BufferParams const & bparams) const
2270 {
2271         Language const * doc_language = bparams.language;
2272         FontList::const_iterator cit = d->fontlist_.begin();
2273         FontList::const_iterator end = d->fontlist_.end();
2274
2275         for (; cit != end; ++cit)
2276                 if (cit->font().language() != ignore_language &&
2277                     cit->font().language() != latex_language &&
2278                     cit->font().language() != doc_language)
2279                         return true;
2280         return false;
2281 }
2282
2283
2284 // Convert the paragraph to a string.
2285 // Used for building the table of contents
2286 docstring const Paragraph::asString(Buffer const & buffer, bool label) const
2287 {
2288         return asString(buffer, 0, size(), label);
2289 }
2290
2291
2292 docstring const Paragraph::asString(Buffer const & buffer,
2293                                  pos_type beg, pos_type end, bool label) const
2294 {
2295
2296         odocstringstream os;
2297
2298         if (beg == 0 && label && !params().labelString().empty())
2299                 os << params().labelString() << ' ';
2300
2301         for (pos_type i = beg; i < end; ++i) {
2302                 char_type const c = d->text_[i];
2303                 if (isPrintable(c))
2304                         os.put(c);
2305                 else if (c == META_INSET)
2306                         getInset(i)->textString(buffer, os);
2307         }
2308
2309         return os.str();
2310 }
2311
2312
2313 void Paragraph::setInsetOwner(Inset * inset)
2314 {
2315         d->inset_owner_ = inset;
2316 }
2317
2318
2319 int Paragraph::id() const
2320 {
2321         return d->id_;
2322 }
2323
2324
2325 LayoutPtr const & Paragraph::layout() const
2326 {
2327         return d->layout_;
2328 }
2329
2330
2331 void Paragraph::layout(LayoutPtr const & new_layout)
2332 {
2333         d->layout_ = new_layout;
2334 }
2335
2336
2337 Inset * Paragraph::inInset() const
2338 {
2339         return d->inset_owner_;
2340 }
2341
2342
2343 InsetCode Paragraph::ownerCode() const
2344 {
2345         return d->inset_owner_ ?
2346                 d->inset_owner_->lyxCode() : NO_CODE;
2347 }
2348
2349
2350 ParagraphParameters & Paragraph::params()
2351 {
2352         return d->params_;
2353 }
2354
2355
2356 ParagraphParameters const & Paragraph::params() const
2357 {
2358         return d->params_;
2359 }
2360
2361
2362 bool Paragraph::isFreeSpacing() const
2363 {
2364         if (layout()->free_spacing)
2365                 return true;
2366
2367         // for now we just need this, later should we need this in some
2368         // other way we can always add a function to Inset too.
2369         return ownerCode() == ERT_CODE || ownerCode() == LISTINGS_CODE;
2370 }
2371
2372
2373 bool Paragraph::allowEmpty() const
2374 {
2375         if (layout()->keepempty)
2376                 return true;
2377         return ownerCode() == ERT_CODE || ownerCode() == LISTINGS_CODE;
2378 }
2379
2380
2381 char_type Paragraph::transformChar(char_type c, pos_type pos) const
2382 {
2383         if (!Encodings::is_arabic(c))
2384                 return c;
2385
2386         char_type prev_char = ' ';
2387         char_type next_char = ' ';
2388
2389         for (pos_type i = pos - 1; i >= 0; --i) {
2390                 char_type const par_char = d->text_[i];
2391                 if (!Encodings::isComposeChar_arabic(par_char)) {
2392                         prev_char = par_char;
2393                         break;
2394                 }
2395         }
2396
2397         for (pos_type i = pos + 1, end = size(); i < end; ++i) {
2398                 char_type const par_char = d->text_[i];
2399                 if (!Encodings::isComposeChar_arabic(par_char)) {
2400                         next_char = par_char;
2401                         break;
2402                 }
2403         }
2404
2405         if (Encodings::is_arabic(next_char)) {
2406                 if (Encodings::is_arabic(prev_char) &&
2407                         !Encodings::is_arabic_special(prev_char))
2408                         return Encodings::transformChar(c, Encodings::FORM_MEDIAL);
2409                 else
2410                         return Encodings::transformChar(c, Encodings::FORM_INITIAL);
2411         } else {
2412                 if (Encodings::is_arabic(prev_char) &&
2413                         !Encodings::is_arabic_special(prev_char))
2414                         return Encodings::transformChar(c, Encodings::FORM_FINAL);
2415                 else
2416                         return Encodings::transformChar(c, Encodings::FORM_ISOLATED);
2417         }
2418 }
2419
2420
2421 int Paragraph::checkBiblio(bool track_changes)
2422 {
2423         //FIXME From JS:
2424         //This is getting more and more a mess. ...We really should clean
2425         //up this bibitem issue for 1.6. See also bug 2743.
2426
2427         // Add bibitem insets if necessary
2428         if (layout()->labeltype != LABEL_BIBLIO)
2429                 return 0;
2430
2431         bool hasbibitem = !d->insetlist_.empty()
2432                 // Insist on it being in pos 0
2433                 && d->text_[0] == META_INSET
2434                 && d->insetlist_.begin()->inset->lyxCode() == BIBITEM_CODE;
2435
2436         docstring oldkey;
2437         docstring oldlabel;
2438
2439         // remove a bibitem in pos != 0
2440         // restore it later in pos 0 if necessary
2441         // (e.g. if a user inserts contents _before_ the item)
2442         // we're assuming there's only one of these, which there
2443         // should be.
2444         int erasedInsetPosition = -1;
2445         InsetList::iterator it = d->insetlist_.begin();
2446         InsetList::iterator end = d->insetlist_.end();
2447         for (; it != end; ++it)
2448                 if (it->inset->lyxCode() == BIBITEM_CODE
2449                     && it->pos > 0) {
2450                         InsetBibitem * olditem = static_cast<InsetBibitem *>(it->inset);
2451                         oldkey = olditem->getParam("key");
2452                         oldlabel = olditem->getParam("label");
2453                         erasedInsetPosition = it->pos;
2454                         eraseChar(erasedInsetPosition, track_changes);
2455                         break;
2456         }
2457
2458         //There was an InsetBibitem at the beginning, and we didn't
2459         //have to erase one.
2460         if (hasbibitem && erasedInsetPosition < 0)
2461                         return 0;
2462
2463         //There was an InsetBibitem at the beginning and we did have to
2464         //erase one. So we give its properties to the beginning inset.
2465         if (hasbibitem) {
2466                 InsetBibitem * inset =
2467                         static_cast<InsetBibitem *>(d->insetlist_.begin()->inset);
2468                 if (!oldkey.empty())
2469                         inset->setParam("key", oldkey);
2470                 inset->setParam("label", oldlabel);
2471                 return -erasedInsetPosition;
2472         }
2473
2474         //There was no inset at the beginning, so we need to create one with
2475         //the key and label of the one we erased.
2476         InsetBibitem * inset(new InsetBibitem(InsetCommandParams(BIBITEM_CODE)));
2477         // restore values of previously deleted item in this par.
2478         if (!oldkey.empty())
2479                 inset->setParam("key", oldkey);
2480         inset->setParam("label", oldlabel);
2481         insertInset(0, static_cast<Inset *>(inset),
2482                     Change(track_changes ? Change::INSERTED : Change::UNCHANGED));
2483
2484         return 1;
2485 }
2486
2487
2488 void Paragraph::checkAuthors(AuthorList const & authorList)
2489 {
2490         d->changes_.checkAuthors(authorList);
2491 }
2492
2493
2494 bool Paragraph::isUnchanged(pos_type pos) const
2495 {
2496         return lookupChange(pos).type == Change::UNCHANGED;
2497 }
2498
2499
2500 bool Paragraph::isInserted(pos_type pos) const
2501 {
2502         return lookupChange(pos).type == Change::INSERTED;
2503 }
2504
2505
2506 bool Paragraph::isDeleted(pos_type pos) const
2507 {
2508         return lookupChange(pos).type == Change::DELETED;
2509 }
2510
2511
2512 InsetList const & Paragraph::insetList() const
2513 {
2514         return d->insetlist_;
2515 }
2516
2517
2518 Inset * Paragraph::releaseInset(pos_type pos)
2519 {
2520         Inset * inset = d->insetlist_.release(pos);
2521         /// does not honour change tracking!
2522         eraseChar(pos, false);
2523         return inset;
2524 }
2525
2526
2527 Inset * Paragraph::getInset(pos_type pos)
2528 {
2529         return d->insetlist_.get(pos);
2530 }
2531
2532
2533 Inset const * Paragraph::getInset(pos_type pos) const
2534 {
2535         return d->insetlist_.get(pos);
2536 }
2537
2538
2539 void Paragraph::changeCase(BufferParams const & bparams, pos_type pos,
2540                 pos_type right, TextCase action)
2541 {
2542         // process sequences of modified characters; in change
2543         // tracking mode, this approach results in much better
2544         // usability than changing case on a char-by-char basis
2545         docstring changes;
2546
2547         bool const trackChanges = bparams.trackChanges;
2548
2549         bool capitalize = true;
2550
2551         for (; pos < right; ++pos) {
2552                 char_type oldChar = d->text_[pos];
2553                 char_type newChar = oldChar;
2554
2555                 // ignore insets and don't play with deleted text!
2556                 if (oldChar != META_INSET && !isDeleted(pos)) {
2557                         switch (action) {
2558                                 case text_lowercase:
2559                                         newChar = lowercase(oldChar);
2560                                         break;
2561                                 case text_capitalization:
2562                                         if (capitalize) {
2563                                                 newChar = uppercase(oldChar);
2564                                                 capitalize = false;
2565                                         }
2566                                         break;
2567                                 case text_uppercase:
2568                                         newChar = uppercase(oldChar);
2569                                         break;
2570                         }
2571                 }
2572
2573                 if (!isLetter(pos) || isDeleted(pos)) {
2574                         // permit capitalization again
2575                         capitalize = true;
2576                 }
2577
2578                 if (oldChar != newChar)
2579                         changes += newChar;
2580
2581                 if (oldChar == newChar || pos == right - 1) {
2582                         if (oldChar != newChar) {
2583                                 // step behind the changing area
2584                                 pos++;
2585                         }
2586                         int erasePos = pos - changes.size();
2587                         for (size_t i = 0; i < changes.size(); i++) {
2588                                 insertChar(pos, changes[i],
2589                                         getFontSettings(bparams,
2590                                         erasePos),
2591                                         trackChanges);
2592                                 if (!eraseChar(erasePos, trackChanges)) {
2593                                         ++erasePos;
2594                                         ++pos; // advance
2595                                         ++right; // expand selection
2596                                 }
2597                         }
2598                         changes.clear();
2599                 }
2600         }
2601 }
2602
2603
2604 bool Paragraph::find(docstring const & str, bool cs, bool mw,
2605                 pos_type pos, bool del) const
2606 {
2607         int const strsize = str.length();
2608         int i = 0;
2609         pos_type const parsize = d->text_.size();
2610         for (i = 0; pos + i < parsize; ++i) {
2611                 if (i >= strsize)
2612                         break;
2613                 if (cs && str[i] != d->text_[pos + i])
2614                         break;
2615                 if (!cs && uppercase(str[i]) != uppercase(d->text_[pos + i]))
2616                         break;
2617                 if (!del && isDeleted(pos + i))
2618                         break;
2619         }
2620
2621         if (i != strsize)
2622                 return false;
2623
2624         // if necessary, check whether string matches word
2625         if (mw) {
2626                 if (pos > 0 && isLetter(pos - 1))
2627                         return false;
2628                 if (pos + strsize < parsize
2629                         && isLetter(pos + strsize))
2630                         return false;
2631         }
2632
2633         return true;
2634 }
2635
2636
2637 char_type Paragraph::getChar(pos_type pos) const
2638 {
2639         return d->text_[pos];
2640 }
2641
2642
2643 pos_type Paragraph::size() const
2644 {
2645         return d->text_.size();
2646 }
2647
2648
2649 bool Paragraph::empty() const
2650 {
2651         return d->text_.empty();
2652 }
2653
2654
2655 bool Paragraph::isInset(pos_type pos) const
2656 {
2657         return d->text_[pos] == META_INSET;
2658 }
2659
2660
2661 bool Paragraph::isSeparator(pos_type pos) const
2662 {
2663         //FIXME: Are we sure this can be the only separator?
2664         return d->text_[pos] == ' ';
2665 }
2666
2667
2668 } // namespace lyx