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 "BaseClassList.h"
  23 #include "Buffer.h"
  24 #include "BufferParams.h"
  25 #include "Changes.h"
  26 #include "Counters.h"
  27 #include "Encoding.h"
  28 #include "InsetList.h"
  29 #include "Language.h"
  30 #include "LaTeXFeatures.h"
  31 #include "Layout.h"
  32 #include "Length.h"
  33 #include "Font.h"
  34 #include "FontList.h"
  35 #include "LyXRC.h"
  36 #include "OutputParams.h"
  37 #include "output_latex.h"
  38 #include "paragraph_funcs.h"
  39 #include "ParagraphParameters.h"
  40 #include "sgml.h"
  41 #include "TextClass.h"
  42 #include "TexRow.h"
  43 #include "Text.h"
  44 #include "VSpace.h"
  45 #include "WordList.h"
  46
  47 #include "frontends/alert.h"
  48
  49 #include "insets/InsetBibitem.h"
  50 #include "insets/InsetLabel.h"
  51
  52 #include "support/convert.h"
  53 #include "support/debug.h"
  54 #include "support/gettext.h"
  55 #include "support/lstrings.h"
  56 #include "support/Messages.h"
  57 #include "support/textutils.h"
  58
  59 #include <sstream>
  60 #include <vector>
  61
  62 using namespace std;
  63 using namespace lyx::support;
  64
  65 namespace lyx {
  66
  67 namespace {
  68 /// Inset identifier (above 0x10ffff, for ucs-4)
  69 char_type const META_INSET = 0x200001;
  70 };
  71
  72 /////////////////////////////////////////////////////////////////////
  73 //
  74 // Paragraph::Private
  75 //
  76 /////////////////////////////////////////////////////////////////////
  77
  78 class Paragraph::Private
  79 {
  80 public:
  81         ///
  82         Private(Paragraph * owner, Layout const & layout);
  83         /// "Copy constructor"
  84         Private(Private const &, Paragraph * owner);
  85
  86         ///
  87         void insertChar(pos_type pos, char_type c, Change const & change);
  88
  89         /// Output the surrogate pair formed by \p c and \p next to \p os.
  90         /// \return the number of characters written.
  91         int latexSurrogatePair(odocstream & os, char_type c, char_type next,
  92                                Encoding const &);
  93
  94         /// Output a space in appropriate formatting (or a surrogate pair
  95         /// if the next character is a combining character).
  96         /// \return whether a surrogate pair was output.
  97         bool simpleTeXBlanks(OutputParams const &,
  98                              odocstream &, TexRow & texrow,
  99                              pos_type i,
 100                              unsigned int & column,
 101                              Font const & font,
 102                              Layout const & style);
 103
 104         /// Output consecutive unicode chars, belonging to the same script as
 105         /// specified by the latex macro \p ltx, to \p os starting from \p i.
 106         /// \return the number of characters written.
 107         int writeScriptChars(odocstream & os, docstring const & ltx,
 108                            Change &, Encoding const &, pos_type & i);
 109
 110         /// This could go to ParagraphParameters if we want to.
 111         int startTeXParParams(BufferParams const &, odocstream &, TexRow &,
 112                               bool) const;
 113
 114         /// This could go to ParagraphParameters if we want to.
 115         int endTeXParParams(BufferParams const &, odocstream &, TexRow &,
 116                             bool) const;
 117
 118         ///
 119         void latexInset(BufferParams const &,
 120                                    odocstream &,
 121                                    TexRow & texrow, OutputParams &,
 122                                    Font & running_font,
 123                                    Font & basefont,
 124                                    Font const & outerfont,
 125                                    bool & open_font,
 126                                    Change & running_change,
 127                                    Layout const & style,
 128                                    pos_type & i,
 129                                    unsigned int & column);
 130
 131         ///
 132         void latexSpecialChar(
 133                                    odocstream & os,
 134                                    OutputParams & runparams,
 135                                    Font & running_font,
 136                                    Change & running_change,
 137                                    Layout const & style,
 138                                    pos_type & i,
 139                                    unsigned int & column);
 140
 141         ///
 142         bool latexSpecialT1(
 143                 char_type const c,
 144                 odocstream & os,
 145                 pos_type & i,
 146                 unsigned int & column);
 147         ///
 148         bool latexSpecialTypewriter(
 149                 char_type const c,
 150                 odocstream & os,
 151                 pos_type & i,
 152                 unsigned int & column);
 153         ///
 154         bool latexSpecialPhrase(
 155                 odocstream & os,
 156                 pos_type & i,
 157                 unsigned int & column,
 158                 OutputParams & runparams);
 159
 160         ///
 161         void validate(LaTeXFeatures & features,
 162                       Layout const * layout) const;
 163
 164         /// Checks if the paragraph contains only text and no inset or font change.
 165         bool onlyText(Buffer const & buf, Font const & outerfont,
 166                       pos_type initial) const;
 167
 168         /// match a string against a particular point in the paragraph
 169         bool isTextAt(string const & str, pos_type pos) const;
 170
 171         /// Which Paragraph owns us?
 172         Paragraph * owner_;
 173
 174         /// In which Inset?
 175         Inset * inset_owner_;
 176
 177         ///
 178         FontList fontlist_;
 179
 180         ///
 181         unsigned int id_;
 182         ///
 183         static unsigned int paragraph_id;
 184         ///
 185         ParagraphParameters params_;
 186
 187         /// for recording and looking up changes
 188         Changes changes_;
 189
 190         ///
 191         InsetList insetlist_;
 192
 193         /// end of label
 194         pos_type begin_of_body_;
 195
 196         typedef docstring TextContainer;
 197         ///
 198         TextContainer text_;
 199
 200         typedef std::set<docstring> Words;
 201         ///
 202         Words words_;
 203         ///
 204         Layout const * layout_;
 205 };
 206
 207
 208 // Initialization of the counter for the paragraph id's,
 209 unsigned int Paragraph::Private::paragraph_id = 0;
 210
 211 namespace {
 212
 213 struct special_phrase {
 214         string phrase;
 215         docstring macro;
 216         bool builtin;
 217 };
 218
 219 special_phrase const special_phrases[] = {
 220         { "LyX", from_ascii("\\LyX{}"), false },
 221         { "TeX", from_ascii("\\TeX{}"), true },
 222         { "LaTeX2e", from_ascii("\\LaTeXe{}"), true },
 223         { "LaTeX", from_ascii("\\LaTeX{}"), true },
 224 };
 225
 226 size_t const phrases_nr = sizeof(special_phrases)/sizeof(special_phrase);
 227
 228 } // namespace anon
 229
 230
 231 Paragraph::Private::Private(Paragraph * owner, Layout const & layout)
 232         : owner_(owner), inset_owner_(0), begin_of_body_(0), layout_(&layout)
 233 {
 234         id_ = paragraph_id++;
 235         text_.reserve(100);
 236 }
 237
 238
 239 Paragraph::Private::Private(Private const & p, Paragraph * owner)
 240         : owner_(owner), inset_owner_(p.inset_owner_), fontlist_(p.fontlist_),
 241           params_(p.params_), changes_(p.changes_), insetlist_(p.insetlist_),
 242           begin_of_body_(p.begin_of_body_), text_(p.text_), words_(p.words_),
 243           layout_(p.layout_)
 244 {
 245         id_ = paragraph_id++;
 246 }
 247
 248
 249 bool Paragraph::isChanged(pos_type start, pos_type end) const
 250 {
 251         BOOST_ASSERT(start >= 0 && start <= size());
 252         BOOST_ASSERT(end > start && end <= size() + 1);
 253
 254         return d->changes_.isChanged(start, end);
 255 }
 256
 257
 258 bool Paragraph::isMergedOnEndOfParDeletion(bool trackChanges) const
 259 {
 260         // keep the logic here in sync with the logic of eraseChars()
 261         if (!trackChanges)
 262                 return true;
 263
 264         Change const change = d->changes_.lookup(size());
 265         return change.type == Change::INSERTED && change.author == 0;
 266 }
 267
 268
 269 void Paragraph::setChange(Change const & change)
 270 {
 271         // beware of the imaginary end-of-par character!
 272         d->changes_.set(change, 0, size() + 1);
 273
 274         /*
 275          * Propagate the change recursively - but not in case of DELETED!
 276          *
 277          * Imagine that your co-author makes changes in an existing inset. He
 278          * sends your document to you and you come to the conclusion that the
 279          * inset should go completely. If you erase it, LyX must not delete all
 280          * text within the inset. Otherwise, the change tracked insertions of
 281          * your co-author get lost and there is no way to restore them later.
 282          *
 283          * Conclusion: An inset's content should remain untouched if you delete it
 284          */
 285
 286         if (change.type != Change::DELETED) {
 287                 for (pos_type pos = 0; pos < size(); ++pos) {
 288                         if (Inset * inset = getInset(pos))
 289                                 inset->setChange(change);
 290                 }
 291         }
 292 }
 293
 294
 295 void Paragraph::setChange(pos_type pos, Change const & change)
 296 {
 297         BOOST_ASSERT(pos >= 0 && pos <= size());
 298         d->changes_.set(change, pos);
 299
 300         // see comment in setChange(Change const &) above
 301         if (change.type != Change::DELETED && pos < size())
 302                         if (Inset * inset = getInset(pos))
 303                                 inset->setChange(change);
 304 }
 305
 306
 307 Change const & Paragraph::lookupChange(pos_type pos) const
 308 {
 309         BOOST_ASSERT(pos >= 0 && pos <= size());
 310         return d->changes_.lookup(pos);
 311 }
 312
 313
 314 void Paragraph::acceptChanges(BufferParams const & bparams, pos_type start,
 315                 pos_type end)
 316 {
 317         BOOST_ASSERT(start >= 0 && start <= size());
 318         BOOST_ASSERT(end > start && end <= size() + 1);
 319
 320         for (pos_type pos = start; pos < end; ++pos) {
 321                 switch (lookupChange(pos).type) {
 322                         case Change::UNCHANGED:
 323                                 // accept changes in nested inset
 324                                 if (Inset * inset = getInset(pos))
 325                                         inset->acceptChanges(bparams);
 326                                 break;
 327
 328                         case Change::INSERTED:
 329                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 330                                 // also accept changes in nested inset
 331                                 if (Inset * inset = getInset(pos))
 332                                         inset->acceptChanges(bparams);
 333                                 break;
 334
 335                         case Change::DELETED:
 336                                 // Suppress access to non-existent
 337                                 // "end-of-paragraph char"
 338                                 if (pos < size()) {
 339                                         eraseChar(pos, false);
 340                                         --end;
 341                                         --pos;
 342                                 }
 343                                 break;
 344                 }
 345
 346         }
 347 }
 348
 349
 350 void Paragraph::rejectChanges(BufferParams const & bparams,
 351                 pos_type start, pos_type end)
 352 {
 353         BOOST_ASSERT(start >= 0 && start <= size());
 354         BOOST_ASSERT(end > start && end <= size() + 1);
 355
 356         for (pos_type pos = start; pos < end; ++pos) {
 357                 switch (lookupChange(pos).type) {
 358                         case Change::UNCHANGED:
 359                                 // reject changes in nested inset
 360                                 if (Inset * inset = getInset(pos))
 361                                                 inset->rejectChanges(bparams);
 362                                 break;
 363
 364                         case Change::INSERTED:
 365                                 // Suppress access to non-existent
 366                                 // "end-of-paragraph char"
 367                                 if (pos < size()) {
 368                                         eraseChar(pos, false);
 369                                         --end;
 370                                         --pos;
 371                                 }
 372                                 break;
 373
 374                         case Change::DELETED:
 375                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 376
 377                                 // Do NOT reject changes within a deleted inset!
 378                                 // There may be insertions of a co-author inside of it!
 379
 380                                 break;
 381                 }
 382         }
 383 }
 384
 385
 386 void Paragraph::Private::insertChar(pos_type pos, char_type c,
 387                 Change const & change)
 388 {
 389         BOOST_ASSERT(pos >= 0 && pos <= int(text_.size()));
 390
 391         // track change
 392         changes_.insert(change, pos);
 393
 394         // This is actually very common when parsing buffers (and
 395         // maybe inserting ascii text)
 396         if (pos == pos_type(text_.size())) {
 397                 // when appending characters, no need to update tables
 398                 text_.push_back(c);
 399                 return;
 400         }
 401
 402         text_.insert(text_.begin() + pos, c);
 403
 404         // Update the font table.
 405         fontlist_.increasePosAfterPos(pos);
 406
 407         // Update the insets
 408         insetlist_.increasePosAfterPos(pos);
 409 }
 410
 411
 412 void Paragraph::insertInset(pos_type pos, Inset * inset,
 413                                    Change const & change)
 414 {
 415         BOOST_ASSERT(inset);
 416         BOOST_ASSERT(pos >= 0 && pos <= size());
 417
 418         d->insertChar(pos, META_INSET, change);
 419         BOOST_ASSERT(d->text_[pos] == META_INSET);
 420
 421         // Add a new entry in the insetlist_.
 422         d->insetlist_.insert(inset, pos);
 423 }
 424
 425
 426 bool Paragraph::eraseChar(pos_type pos, bool trackChanges)
 427 {
 428         BOOST_ASSERT(pos >= 0 && pos <= size());
 429
 430         // keep the logic here in sync with the logic of isMergedOnEndOfParDeletion()
 431
 432         if (trackChanges) {
 433                 Change change = d->changes_.lookup(pos);
 434
 435                 // set the character to DELETED if
 436                 //  a) it was previously unchanged or
 437                 //  b) it was inserted by a co-author
 438
 439                 if (change.type == Change::UNCHANGED ||
 440                     (change.type == Change::INSERTED && change.author != 0)) {
 441                         setChange(pos, Change(Change::DELETED));
 442                         return false;
 443                 }
 444
 445                 if (change.type == Change::DELETED)
 446                         return false;
 447         }
 448
 449         // Don't physically access the imaginary end-of-paragraph character.
 450         // eraseChar() can only mark it as DELETED. A physical deletion of
 451         // end-of-par must be handled externally.
 452         if (pos == size()) {
 453                 return false;
 454         }
 455
 456         // track change
 457         d->changes_.erase(pos);
 458
 459         // if it is an inset, delete the inset entry
 460         if (d->text_[pos] == META_INSET)
 461                 d->insetlist_.erase(pos);
 462
 463         d->text_.erase(d->text_.begin() + pos);
 464
 465         // Update the fontlist_
 466         d->fontlist_.erase(pos);
 467
 468         // Update the insetlist_
 469         d->insetlist_.decreasePosAfterPos(pos);
 470
 471         return true;
 472 }
 473
 474
 475 int Paragraph::eraseChars(pos_type start, pos_type end, bool trackChanges)
 476 {
 477         BOOST_ASSERT(start >= 0 && start <= size());
 478         BOOST_ASSERT(end >= start && end <= size() + 1);
 479
 480         pos_type i = start;
 481         for (pos_type count = end - start; count; --count) {
 482                 if (!eraseChar(i, trackChanges))
 483                         ++i;
 484         }
 485         return end - i;
 486 }
 487
 488
 489 int Paragraph::Private::latexSurrogatePair(odocstream & os, char_type c,
 490                 char_type next, Encoding const & encoding)
 491 {
 492         // Writing next here may circumvent a possible font change between
 493         // c and next. Since next is only output if it forms a surrogate pair
 494         // with c we can ignore this:
 495         // A font change inside a surrogate pair does not make sense and is
 496         // hopefully impossible to input.
 497         // FIXME: change tracking
 498         // Is this correct WRT change tracking?
 499         docstring const latex1 = encoding.latexChar(next);
 500         docstring const latex2 = encoding.latexChar(c);
 501         os << latex1 << '{' << latex2 << '}';
 502         return latex1.length() + latex2.length() + 2;
 503 }
 504
 505
 506 bool Paragraph::Private::simpleTeXBlanks(OutputParams const & runparams,
 507                                        odocstream & os, TexRow & texrow,
 508                                        pos_type i,
 509                                        unsigned int & column,
 510                                        Font const & font,
 511                                        Layout const & style)
 512 {
 513         if (style.pass_thru || runparams.verbatim)
 514                 return false;
 515
 516         if (i + 1 < int(text_.size())) {
 517                 char_type next = text_[i + 1];
 518                 if (Encodings::isCombiningChar(next)) {
 519                         Encoding const & encoding = *(runparams.encoding);
 520                         // This space has an accent, so we must always output it.
 521                         column += latexSurrogatePair(os, ' ', next, encoding) - 1;
 522                         return true;
 523                 }
 524         }
 525
 526         if (lyxrc.plaintext_linelen > 0
 527             && column > lyxrc.plaintext_linelen
 528             && i
 529             && text_[i - 1] != ' '
 530             && (i + 1 < int(text_.size()))
 531             // same in FreeSpacing mode
 532             && !owner_->isFreeSpacing()
 533             // In typewriter mode, we want to avoid
 534             // ! . ? : at the end of a line
 535             && !(font.fontInfo().family() == TYPEWRITER_FAMILY
 536                  && (text_[i - 1] == '.'
 537                      || text_[i - 1] == '?'
 538                      || text_[i - 1] == ':'
 539                      || text_[i - 1] == '!'))) {
 540                 os << '\n';
 541                 texrow.newline();
 542                 texrow.start(owner_->id(), i + 1);
 543                 column = 0;
 544         } else if (style.free_spacing) {
 545                 os << '~';
 546         } else {
 547                 os << ' ';
 548         }
 549         return false;
 550 }
 551
 552
 553 int Paragraph::Private::writeScriptChars(odocstream & os,
 554                                          docstring const & ltx,
 555                                          Change & runningChange,
 556                                          Encoding const & encoding,
 557                                          pos_type & i)
 558 {
 559         // FIXME: modifying i here is not very nice...
 560
 561         // We only arrive here when a proper language for character text_[i] has
 562         // not been specified (i.e., it could not be translated in the current
 563         // latex encoding) and it belongs to a known script.
 564         // Parameter ltx contains the latex translation of text_[i] as specified in
 565         // the unicodesymbols file and is something like "\textXXX{<spec>}".
 566         // The latex macro name "textXXX" specifies the script to which text_[i]
 567         // belongs and we use it in order to check whether characters from the
 568         // same script immediately follow, such that we can collect them in a
 569         // single "\textXXX" macro. So, we have to retain "\textXXX{<spec>"
 570         // for the first char but only "<spec>" for all subsequent chars.
 571         docstring::size_type const brace1 = ltx.find_first_of(from_ascii("{"));
 572         docstring::size_type const brace2 = ltx.find_last_of(from_ascii("}"));
 573         string script = to_ascii(ltx.substr(1, brace1 - 1));
 574         int length = ltx.substr(0, brace2).length();
 575         os << ltx.substr(0, brace2);
 576         int size = text_.size();
 577         while (i + 1 < size) {
 578                 char_type const next = text_[i + 1];
 579                 // Stop here if next character belongs to another script
 580                 // or there is a change in change tracking status.
 581                 if (!Encodings::isKnownScriptChar(next, script) ||
 582                     runningChange != owner_->lookupChange(i + 1))
 583                         break;
 584                 Font prev_font;
 585                 bool found = false;
 586                 FontList::const_iterator cit = fontlist_.begin();
 587                 FontList::const_iterator end = fontlist_.end();
 588                 for (; cit != end; ++cit) {
 589                         if (cit->pos() >= i && !found) {
 590                                 prev_font = cit->font();
 591                                 found = true;
 592                         }
 593                         if (cit->pos() >= i + 1)
 594                                 break;
 595                 }
 596                 // Stop here if there is a font attribute or encoding change.
 597                 if (found && cit != end && prev_font != cit->font())
 598                         break;
 599                 docstring const latex = encoding.latexChar(next);
 600                 docstring::size_type const b1 =
 601                                         latex.find_first_of(from_ascii("{"));
 602                 docstring::size_type const b2 =
 603                                         latex.find_last_of(from_ascii("}"));
 604                 int const len = b2 - b1 - 1;
 605                 os << latex.substr(b1 + 1, len);
 606                 length += len;
 607                 ++i;
 608         }
 609         os << '}';
 610         ++length;
 611         return length;
 612 }
 613
 614
 615 bool Paragraph::Private::isTextAt(string const & str, pos_type pos) const
 616 {
 617         pos_type const len = str.length();
 618
 619         // is the paragraph large enough?
 620         if (pos + len > int(text_.size()))
 621                 return false;
 622
 623         // does the wanted text start at point?
 624         for (string::size_type i = 0; i < str.length(); ++i) {
 625                 // Caution: direct comparison of characters works only
 626                 // because str is pure ASCII.
 627                 if (str[i] != text_[pos + i])
 628                         return false;
 629         }
 630
 631         return fontlist_.hasChangeInRange(pos, len);
 632 }
 633
 634
 635 void Paragraph::Private::latexInset(
 636                                              BufferParams const & bparams,
 637                                              odocstream & os,
 638                                              TexRow & texrow,
 639                                              OutputParams & runparams,
 640                                              Font & running_font,
 641                                              Font & basefont,
 642                                              Font const & outerfont,
 643                                              bool & open_font,
 644                                              Change & running_change,
 645                                              Layout const & style,
 646                                              pos_type & i,
 647                                              unsigned int & column)
 648 {
 649         Inset * inset = owner_->getInset(i);
 650         BOOST_ASSERT(inset);
 651
 652         if (style.pass_thru) {
 653                 inset->plaintext(os, runparams);
 654                 return;
 655         }
 656
 657         // FIXME: move this to InsetNewline::latex
 658         if (inset->lyxCode() == NEWLINE_CODE) {
 659                 // newlines are handled differently here than
 660                 // the default in simpleTeXSpecialChars().
 661                 if (!style.newline_allowed) {
 662                         os << '\n';
 663                 } else {
 664                         if (open_font) {
 665                                 column += running_font.latexWriteEndChanges(
 666                                         os, bparams, runparams,
 667                                         basefont, basefont);
 668                                 open_font = false;
 669                         }
 670
 671                         if (running_font.fontInfo().family() == TYPEWRITER_FAMILY)
 672                                 os << '~';
 673
 674                         basefont = owner_->getLayoutFont(bparams, outerfont);
 675                         running_font = basefont;
 676
 677                         if (runparams.moving_arg)
 678                                 os << "\\protect ";
 679
 680                 }
 681                 texrow.newline();
 682                 texrow.start(owner_->id(), i + 1);
 683                 column = 0;
 684         }
 685
 686         if (owner_->lookupChange(i).type == Change::DELETED) {
 687                 if( ++runparams.inDeletedInset == 1)
 688                         runparams.changeOfDeletedInset = owner_->lookupChange(i);
 689         }
 690
 691         if (inset->canTrackChanges()) {
 692                 column += Changes::latexMarkChange(os, bparams, running_change,
 693                         Change(Change::UNCHANGED));
 694                 running_change = Change(Change::UNCHANGED);
 695         }
 696
 697         bool close = false;
 698         odocstream::pos_type const len = os.tellp();
 699
 700         if (inset->forceLTR()
 701             && running_font.isRightToLeft()
 702                 // ERT is an exception, it should be output with no decorations at all
 703                 && inset->lyxCode() != ERT_CODE) {
 704                 if (running_font.language()->lang() == "farsi")
 705                         os << "\\beginL{}";
 706                 else
 707                         os << "\\L{";
 708                 close = true;
 709         }
 710
 711         // FIXME: Bug: we can have an empty font change here!
 712         // if there has just been a font change, we are going to close it
 713         // right now, which means stupid latex code like \textsf{}. AFAIK,
 714         // this does not harm dvi output. A minor bug, thus (JMarc)
 715
 716         // Some insets cannot be inside a font change command.
 717         // However, even such insets *can* be placed in \L or \R
 718         // or their equivalents (for RTL language switches), so we don't
 719         // close the language in those cases.
 720         // ArabTeX, though, cannot handle this special behavior, it seems.
 721         bool arabtex = basefont.language()->lang() == "arabic_arabtex"
 722                 || running_font.language()->lang() == "arabic_arabtex";
 723         if (open_font && inset->noFontChange()) {
 724                 bool closeLanguage = arabtex
 725                         || basefont.isRightToLeft() == running_font.isRightToLeft();
 726                 unsigned int count = running_font.latexWriteEndChanges(os,
 727                         bparams, runparams, basefont, basefont, closeLanguage);
 728                 column += count;
 729                 // if any font properties were closed, update the running_font,
 730                 // making sure, however, to leave the language as it was
 731                 if (count > 0) {
 732                         // FIXME: probably a better way to keep track of the old
 733                         // language, than copying the entire font?
 734                         Font const copy_font(running_font);
 735                         basefont = owner_->getLayoutFont(bparams, outerfont);
 736                         running_font = basefont;
 737                         if (!closeLanguage)
 738                                 running_font.setLanguage(copy_font.language());
 739                         // leave font open if language is still open
 740                         open_font = (running_font.language() == basefont.language());
 741                         if (closeLanguage)
 742                                 runparams.local_font = &basefont;
 743                 }
 744         }
 745
 746         int tmp = inset->latex(os, runparams);
 747
 748         if (close) {
 749         if (running_font.language()->lang() == "farsi")
 750                         os << "\\endL{}";
 751                 else
 752                         os << '}';
 753         }
 754
 755         if (tmp) {
 756                 for (int j = 0; j < tmp; ++j)
 757                         texrow.newline();
 758
 759                 texrow.start(owner_->id(), i + 1);
 760                 column = 0;
 761         } else {
 762                 column += os.tellp() - len;
 763         }
 764
 765         if (owner_->lookupChange(i).type == Change::DELETED)
 766                 --runparams.inDeletedInset;
 767 }
 768
 769
 770 void Paragraph::Private::latexSpecialChar(
 771                                              odocstream & os,
 772                                              OutputParams & runparams,
 773                                              Font & running_font,
 774                                              Change & running_change,
 775                                              Layout const & style,
 776                                              pos_type & i,
 777                                              unsigned int & column)
 778 {
 779         char_type const c = text_[i];
 780
 781         if (style.pass_thru) {
 782                 if (c != '\0')
 783                         // FIXME UNICODE: This can fail if c cannot
 784                         // be encoded in the current encoding.
 785                         os.put(c);
 786                 return;
 787         }
 788
 789         if (runparams.verbatim) {
 790                 os.put(c);
 791                 return;
 792         }
 793
 794         if (lyxrc.fontenc == "T1" && latexSpecialT1(c, os, i, column))
 795                 return;
 796
 797         if (running_font.fontInfo().family() == TYPEWRITER_FAMILY
 798                 && latexSpecialTypewriter(c, os, i, column))
 799                 return;
 800
 801         // Otherwise, we use what LaTeX provides us.
 802         switch (c) {
 803         case '\\':
 804                 os << "\\textbackslash{}";
 805                 column += 15;
 806                 break;
 807         case '<':
 808                 os << "\\textless{}";
 809                 column += 10;
 810                 break;
 811         case '>':
 812                 os << "\\textgreater{}";
 813                 column += 13;
 814                 break;
 815         case '|':
 816                 os << "\\textbar{}";
 817                 column += 9;
 818                 break;
 819         case '-':
 820                 os << '-';
 821                 break;
 822         case '\"':
 823                 os << "\\char`\\\"{}";
 824                 column += 9;
 825                 break;
 826
 827         case '$': case '&':
 828         case '%': case '#': case '{':
 829         case '}': case '_':
 830                 os << '\\';
 831                 os.put(c);
 832                 column += 1;
 833                 break;
 834
 835         case '~':
 836                 os << "\\textasciitilde{}";
 837                 column += 16;
 838                 break;
 839
 840         case '^':
 841                 os << "\\textasciicircum{}";
 842                 column += 17;
 843                 break;
 844
 845         case '*': case '[':
 846                 // avoid being mistaken for optional arguments
 847                 os << '{';
 848                 os.put(c);
 849                 os << '}';
 850                 column += 2;
 851                 break;
 852
 853         case ' ':
 854                 // Blanks are printed before font switching.
 855                 // Sure? I am not! (try nice-latex)
 856                 // I am sure it's correct. LyX might be smarter
 857                 // in the future, but for now, nothing wrong is
 858                 // written. (Asger)
 859                 break;
 860
 861         default:
 862
 863                 // LyX, LaTeX etc.
 864                 if (latexSpecialPhrase(os, i, column, runparams))
 865                         return;
 866
 867                 if (c == '\0')
 868                         return;
 869
 870                 Encoding const & encoding = *(runparams.encoding);
 871                 if (i + 1 < int(text_.size())) {
 872                         char_type next = text_[i + 1];
 873                         if (Encodings::isCombiningChar(next)) {
 874                                 column += latexSurrogatePair(os, c, next, encoding) - 1;
 875                                 ++i;
 876                                 break;
 877                         }
 878                 }
 879                 string script;
 880                 docstring const latex = encoding.latexChar(c);
 881                 if (Encodings::isKnownScriptChar(c, script)
 882                     && prefixIs(latex, from_ascii("\\" + script)))
 883                         column += writeScriptChars(os, latex,
 884                                         running_change, encoding, i) - 1;
 885                 else if (latex.length() > 1 && latex[latex.length() - 1] != '}') {
 886                         // Prevent eating of a following
 887                         // space or command corruption by
 888                         // following characters
 889                         column += latex.length() + 1;
 890                         os << latex << "{}";
 891                 } else {
 892                         column += latex.length() - 1;
 893                         os << latex;
 894                 }
 895                 break;
 896         }
 897 }
 898
 899
 900 bool Paragraph::Private::latexSpecialT1(char_type const c, odocstream & os,
 901         pos_type & i, unsigned int & column)
 902 {
 903         switch (c) {
 904         case '>':
 905         case '<':
 906                 os.put(c);
 907                 // In T1 encoding, these characters exist
 908                 // but we should avoid ligatures
 909                 if (i + 1 >= int(text_.size()) || text_[i + 1] != c)
 910                         return true;
 911                 os << "\\,{}";
 912                 column += 3;
 913                 // Alternative code:
 914                 //os << "\\textcompwordmark{}";
 915                 //column += 19;
 916                 return true;
 917         case '|':
 918                 os.put(c);
 919                 return true;
 920         default:
 921                 return false;
 922         }
 923 }
 924
 925
 926 bool Paragraph::Private::latexSpecialTypewriter(char_type const c, odocstream & os,
 927         pos_type & i, unsigned int & column)
 928 {
 929         switch (c) {
 930         case '-':
 931                 if (i + 1 < int(text_.size()) && text_[i + 1] == '-') {
 932                         // "--" in Typewriter mode -> "-{}-"
 933                         os << "-{}";
 934                         column += 2;
 935                 } else
 936                         os << '-';
 937                 return true;
 938
 939         // I assume this is hack treating typewriter as verbatim
 940         // FIXME UNICODE: This can fail if c cannot be encoded
 941         // in the current encoding.
 942
 943         case '\0':
 944                 return true;
 945
 946         // Those characters are not directly supported.
 947         case '\\':
 948         case '\"':
 949         case '$': case '&':
 950         case '%': case '#': case '{':
 951         case '}': case '_':
 952         case '~':
 953         case '^':
 954         case '*': case '[':
 955         case ' ':
 956                 return false;
 957
 958         default:
 959                 // With Typewriter font, these characters exist.
 960                 os.put(c);
 961                 return true;
 962         }
 963 }
 964
 965
 966 bool Paragraph::Private::latexSpecialPhrase(odocstream & os, pos_type & i,
 967         unsigned int & column, OutputParams & runparams)
 968 {
 969         // FIXME: if we have "LaTeX" with a font
 970         // change in the middle (before the 'T', then
 971         // the "TeX" part is still special cased.
 972         // Really we should only operate this on
 973         // "words" for some definition of word
 974
 975         for (size_t pnr = 0; pnr < phrases_nr; ++pnr) {
 976                 if (!isTextAt(special_phrases[pnr].phrase, i))
 977                         continue;
 978                 if (runparams.moving_arg)
 979                         os << "\\protect";
 980                 os << special_phrases[pnr].macro;
 981                 i += special_phrases[pnr].phrase.length() - 1;
 982                 column += special_phrases[pnr].macro.length() - 1;
 983                 return true;
 984         }
 985         return false;
 986 }
 987
 988
 989 void Paragraph::Private::validate(LaTeXFeatures & features,
 990                                 Layout const * layout) const
 991 {
 992         // check the params.
 993         if (!params_.spacing().isDefault())
 994                 features.require("setspace");
 995
 996         // then the layouts
 997         features.useLayout(layout->name());
 998
 999         // then the fonts
1000         fontlist_.validate(features);
1001
1002         // then the indentation
1003         if (!params_.leftIndent().zero())
1004                 features.require("ParagraphLeftIndent");
1005
1006         // then the insets
1007         InsetList::const_iterator icit = insetlist_.begin();
1008         InsetList::const_iterator iend = insetlist_.end();
1009         for (; icit != iend; ++icit) {
1010                 if (icit->inset) {
1011                         icit->inset->validate(features);
1012                         if (layout->needprotect &&
1013                             icit->inset->lyxCode() == FOOT_CODE)
1014                                 features.require("NeedLyXFootnoteCode");
1015                 }
1016         }
1017
1018         // then the contents
1019         for (pos_type i = 0; i < int(text_.size()) ; ++i) {
1020                 for (size_t pnr = 0; pnr < phrases_nr; ++pnr) {
1021                         if (!special_phrases[pnr].builtin
1022                             && isTextAt(special_phrases[pnr].phrase, i)) {
1023                                 features.require(special_phrases[pnr].phrase);
1024                                 break;
1025                         }
1026                 }
1027                 Encodings::validate(text_[i], features);
1028         }
1029 }
1030
1031 /////////////////////////////////////////////////////////////////////
1032 //
1033 // Paragraph
1034 //
1035 /////////////////////////////////////////////////////////////////////
1036
1037 namespace {
1038         Layout const emptyParagraphLayout;
1039 }
1040
1041 Paragraph::Paragraph()
1042         : d(new Paragraph::Private(this, emptyParagraphLayout))
1043 {
1044         itemdepth = 0;
1045         d->params_.clear();
1046 }
1047
1048
1049 Paragraph::Paragraph(Paragraph const & par)
1050         : itemdepth(par.itemdepth),
1051         d(new Paragraph::Private(*par.d, this))
1052 {
1053         registerWords();
1054 }
1055
1056
1057 Paragraph & Paragraph::operator=(Paragraph const & par)
1058 {
1059         // needed as we will destroy the private part before copying it
1060         if (&par != this) {
1061                 itemdepth = par.itemdepth;
1062
1063                 deregisterWords();
1064                 delete d;
1065                 d = new Private(*par.d, this);
1066                 registerWords();
1067         }
1068         return *this;
1069 }
1070
1071
1072 Paragraph::~Paragraph()
1073 {
1074         deregisterWords();
1075         delete d;
1076 }
1077
1078
1079 void Paragraph::write(ostream & os, BufferParams const & bparams,
1080         depth_type & dth) const
1081 {
1082         // The beginning or end of a deeper (i.e. nested) area?
1083         if (dth != d->params_.depth()) {
1084                 if (d->params_.depth() > dth) {
1085                         while (d->params_.depth() > dth) {
1086                                 os << "\n\\begin_deeper";
1087                                 ++dth;
1088                         }
1089                 } else {
1090                         while (d->params_.depth() < dth) {
1091                                 os << "\n\\end_deeper";
1092                                 --dth;
1093                         }
1094                 }
1095         }
1096
1097         // First write the layout
1098         os << "\n\\begin_layout " << to_utf8(d->layout_->name()) << '\n';
1099
1100         d->params_.write(os);
1101
1102         Font font1(inherit_font, bparams.language);
1103
1104         Change running_change = Change(Change::UNCHANGED);
1105
1106         int column = 0;
1107         for (pos_type i = 0; i <= size(); ++i) {
1108
1109                 Change change = lookupChange(i);
1110                 Changes::lyxMarkChange(os, column, running_change, change);
1111                 running_change = change;
1112
1113                 if (i == size())
1114                         break;
1115
1116                 // Write font changes
1117                 Font font2 = getFontSettings(bparams, i);
1118                 if (font2 != font1) {
1119                         font2.lyxWriteChanges(font1, os);
1120                         column = 0;
1121                         font1 = font2;
1122                 }
1123
1124                 char_type const c = d->text_[i];
1125                 switch (c) {
1126                 case META_INSET:
1127                         if (Inset const * inset = getInset(i)) {
1128                                 if (inset->directWrite()) {
1129                                         // international char, let it write
1130                                         // code directly so it's shorter in
1131                                         // the file
1132                                         inset->write(os);
1133                                 } else {
1134                                         if (i)
1135                                                 os << '\n';
1136                                         os << "\\begin_inset ";
1137                                         inset->write(os);
1138                                         os << "\n\\end_inset\n\n";
1139                                         column = 0;
1140                                 }
1141                         }
1142                         break;
1143                 case '\\':
1144                         os << "\n\\backslash\n";
1145                         column = 0;
1146                         break;
1147                 case '.':
1148                         if (i + 1 < size() && d->text_[i + 1] == ' ') {
1149                                 os << ".\n";
1150                                 column = 0;
1151                         } else
1152                                 os << '.';
1153                         break;
1154                 default:
1155                         if ((column > 70 && c == ' ')
1156                             || column > 79) {
1157                                 os << '\n';
1158                                 column = 0;
1159                         }
1160                         // this check is to amend a bug. LyX sometimes
1161                         // inserts '\0' this could cause problems.
1162                         if (c != '\0')
1163                                 os << to_utf8(docstring(1, c));
1164                         else
1165                                 lyxerr << "ERROR (Paragraph::writeFile):"
1166                                         " NULL char in structure." << endl;
1167                         ++column;
1168                         break;
1169                 }
1170         }
1171
1172         os << "\n\\end_layout\n";
1173 }
1174
1175
1176 void Paragraph::validate(LaTeXFeatures & features) const
1177 {
1178         d->validate(features, d->layout_);
1179 }
1180
1181
1182 void Paragraph::insert(pos_type start, docstring const & str,
1183                        Font const & font, Change const & change)
1184 {
1185         for (size_t i = 0, n = str.size(); i != n ; ++i)
1186                 insertChar(start + i, str[i], font, change);
1187 }
1188
1189
1190 void Paragraph::appendChar(char_type c, Font const & font,
1191                 Change const & change)
1192 {
1193         // track change
1194         d->changes_.insert(change, d->text_.size());
1195         // when appending characters, no need to update tables
1196         d->text_.push_back(c);
1197         setFont(d->text_.size() - 1, font);
1198 }
1199
1200
1201 void Paragraph::appendString(docstring const & s, Font const & font,
1202                 Change const & change)
1203 {
1204         pos_type end = s.size();
1205         size_t oldsize = d->text_.size();
1206         size_t newsize = oldsize + end;
1207         size_t capacity = d->text_.capacity();
1208         if (newsize >= capacity)
1209                 d->text_.reserve(max(capacity + 100, newsize));
1210
1211         // when appending characters, no need to update tables
1212         d->text_.append(s);
1213
1214         // FIXME: Optimize this!
1215         for (pos_type i = 0; i != end; ++i) {
1216                 // track change
1217                 d->changes_.insert(change, i);
1218         }
1219         d->fontlist_.set(oldsize, font);
1220         d->fontlist_.set(newsize - 1, font);
1221 }
1222
1223
1224 void Paragraph::insertChar(pos_type pos, char_type c,
1225                            bool trackChanges)
1226 {
1227         d->insertChar(pos, c, Change(trackChanges ?
1228                            Change::INSERTED : Change::UNCHANGED));
1229 }
1230
1231
1232 void Paragraph::insertChar(pos_type pos, char_type c,
1233                            Font const & font, bool trackChanges)
1234 {
1235         d->insertChar(pos, c, Change(trackChanges ?
1236                            Change::INSERTED : Change::UNCHANGED));
1237         setFont(pos, font);
1238 }
1239
1240
1241 void Paragraph::insertChar(pos_type pos, char_type c,
1242                            Font const & font, Change const & change)
1243 {
1244         d->insertChar(pos, c, change);
1245         setFont(pos, font);
1246 }
1247
1248
1249 void Paragraph::insertInset(pos_type pos, Inset * inset,
1250                             Font const & font, Change const & change)
1251 {
1252         insertInset(pos, inset, change);
1253         // Set the font/language of the inset...
1254         setFont(pos, font);
1255 }
1256
1257
1258 bool Paragraph::insetAllowed(InsetCode code)
1259 {
1260         return !d->inset_owner_ || d->inset_owner_->insetAllowed(code);
1261 }
1262
1263
1264 void Paragraph::resetFonts(Font const & font)
1265 {
1266         d->fontlist_.clear();
1267         d->fontlist_.set(0, font);
1268         d->fontlist_.set(d->text_.size() - 1, font);
1269 }
1270
1271 // Gets uninstantiated font setting at position.
1272 Font const Paragraph::getFontSettings(BufferParams const & bparams,
1273                                          pos_type pos) const
1274 {
1275         if (pos > size()) {
1276                 lyxerr << " pos: " << pos << " size: " << size() << endl;
1277                 BOOST_ASSERT(pos <= size());
1278         }
1279
1280         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
1281         if (cit != d->fontlist_.end())
1282                 return cit->font();
1283
1284         if (pos == size() && !empty())
1285                 return getFontSettings(bparams, pos - 1);
1286
1287         return Font(inherit_font, getParLanguage(bparams));
1288 }
1289
1290
1291 FontSpan Paragraph::fontSpan(pos_type pos) const
1292 {
1293         BOOST_ASSERT(pos <= size());
1294         pos_type start = 0;
1295
1296         FontList::const_iterator cit = d->fontlist_.begin();
1297         FontList::const_iterator end = d->fontlist_.end();
1298         for (; cit != end; ++cit) {
1299                 if (cit->pos() >= pos) {
1300                         if (pos >= beginOfBody())
1301                                 return FontSpan(max(start, beginOfBody()),
1302                                                 cit->pos());
1303                         else
1304                                 return FontSpan(start,
1305                                                 min(beginOfBody() - 1,
1306                                                          cit->pos()));
1307                 }
1308                 start = cit->pos() + 1;
1309         }
1310
1311         // This should not happen, but if so, we take no chances.
1312         //lyxerr << "Paragraph::getEndPosOfFontSpan: This should not happen!"
1313         //      << endl;
1314         return FontSpan(pos, pos);
1315 }
1316
1317
1318 // Gets uninstantiated font setting at position 0
1319 Font const Paragraph::getFirstFontSettings(BufferParams const & bparams) const
1320 {
1321         if (!empty() && !d->fontlist_.empty())
1322                 return d->fontlist_.begin()->font();
1323
1324         return Font(inherit_font, bparams.language);
1325 }
1326
1327
1328 // Gets the fully instantiated font at a given position in a paragraph
1329 // This is basically the same function as Text::GetFont() in text2.cpp.
1330 // The difference is that this one is used for generating the LaTeX file,
1331 // and thus cosmetic "improvements" are disallowed: This has to deliver
1332 // the true picture of the buffer. (Asger)
1333 Font const Paragraph::getFont(BufferParams const & bparams, pos_type pos,
1334                                  Font const & outerfont) const
1335 {
1336         BOOST_ASSERT(pos >= 0);
1337
1338         Font font = getFontSettings(bparams, pos);
1339
1340         pos_type const body_pos = beginOfBody();
1341         if (pos < body_pos)
1342                 font.fontInfo().realize(d->layout_->labelfont);
1343         else
1344                 font.fontInfo().realize(d->layout_->font);
1345
1346         font.fontInfo().realize(outerfont.fontInfo());
1347         font.fontInfo().realize(bparams.getFont().fontInfo());
1348
1349         return font;
1350 }
1351
1352
1353 Font const Paragraph::getLabelFont
1354         (BufferParams const & bparams, Font const & outerfont) const
1355 {
1356         FontInfo tmpfont = d->layout_->labelfont;
1357         tmpfont.realize(outerfont.fontInfo());
1358         tmpfont.realize(bparams.getFont().fontInfo());
1359         return Font(tmpfont, getParLanguage(bparams));
1360 }
1361
1362
1363 Font const Paragraph::getLayoutFont
1364         (BufferParams const & bparams, Font const & outerfont) const
1365 {
1366         FontInfo tmpfont = d->layout_->font;
1367         tmpfont.realize(outerfont.fontInfo());
1368         tmpfont.realize(bparams.getFont().fontInfo());
1369         return Font(tmpfont, getParLanguage(bparams));
1370 }
1371
1372
1373 /// Returns the height of the highest font in range
1374 FontSize Paragraph::highestFontInRange
1375         (pos_type startpos, pos_type endpos, FontSize def_size) const
1376 {
1377         return d->fontlist_.highestInRange(startpos, endpos, def_size);
1378 }
1379
1380
1381 char_type
1382 Paragraph::getUChar(BufferParams const & bparams, pos_type pos) const
1383 {
1384         char_type c = d->text_[pos];
1385         if (!lyxrc.rtl_support)
1386                 return c;
1387
1388         char_type uc = c;
1389         switch (c) {
1390         case '(':
1391                 uc = ')';
1392                 break;
1393         case ')':
1394                 uc = '(';
1395                 break;
1396         case '[':
1397                 uc = ']';
1398                 break;
1399         case ']':
1400                 uc = '[';
1401                 break;
1402         case '{':
1403                 uc = '}';
1404                 break;
1405         case '}':
1406                 uc = '{';
1407                 break;
1408         case '<':
1409                 uc = '>';
1410                 break;
1411         case '>':
1412                 uc = '<';
1413                 break;
1414         }
1415         if (uc != c && getFontSettings(bparams, pos).isRightToLeft())
1416                 return uc;
1417         else
1418                 return c;
1419 }
1420
1421
1422 void Paragraph::setFont(pos_type pos, Font const & font)
1423 {
1424         BOOST_ASSERT(pos <= size());
1425
1426         // First, reduce font against layout/label font
1427         // Update: The setCharFont() routine in text2.cpp already
1428         // reduces font, so we don't need to do that here. (Asger)
1429
1430         d->fontlist_.set(pos, font);
1431 }
1432
1433
1434 void Paragraph::makeSameLayout(Paragraph const & par)
1435 {
1436         d->layout_ = par.d->layout_;
1437         d->params_ = par.d->params_;
1438 }
1439
1440
1441 bool Paragraph::stripLeadingSpaces(bool trackChanges)
1442 {
1443         if (isFreeSpacing())
1444                 return false;
1445
1446         int pos = 0;
1447         int count = 0;
1448
1449         while (pos < size() && (isNewline(pos) || isLineSeparator(pos))) {
1450                 if (eraseChar(pos, trackChanges))
1451                         ++count;
1452                 else
1453                         ++pos;
1454         }
1455
1456         return count > 0 || pos > 0;
1457 }
1458
1459
1460 bool Paragraph::hasSameLayout(Paragraph const & par) const
1461 {
1462         return par.d->layout_ == d->layout_
1463                 && d->params_.sameLayout(par.d->params_);
1464 }
1465
1466
1467 depth_type Paragraph::getDepth() const
1468 {
1469         return d->params_.depth();
1470 }
1471
1472
1473 depth_type Paragraph::getMaxDepthAfter() const
1474 {
1475         if (d->layout_->isEnvironment())
1476                 return d->params_.depth() + 1;
1477         else
1478                 return d->params_.depth();
1479 }
1480
1481
1482 char Paragraph::getAlign() const
1483 {
1484         if (d->params_.align() == LYX_ALIGN_LAYOUT)
1485                 return d->layout_->align;
1486         else
1487                 return d->params_.align();
1488 }
1489
1490
1491 docstring const & Paragraph::labelString() const
1492 {
1493         return d->params_.labelString();
1494 }
1495
1496
1497 // the next two functions are for the manual labels
1498 docstring const Paragraph::getLabelWidthString() const
1499 {
1500         if (d->layout_->margintype == MARGIN_MANUAL)
1501                 return d->params_.labelWidthString();
1502         else
1503                 return _("Senseless with this layout!");
1504 }
1505
1506
1507 void Paragraph::setLabelWidthString(docstring const & s)
1508 {
1509         d->params_.labelWidthString(s);
1510 }
1511
1512
1513 docstring const Paragraph::translateIfPossible(docstring const & s,
1514                 BufferParams const & bparams) const
1515 {
1516         if (!isAscii(s) || s.empty()) {
1517                 // This must be a user defined layout. We cannot translate
1518                 // this, since gettext accepts only ascii keys.
1519                 return s;
1520         }
1521         // Probably standard layout, try to translate
1522         Messages & m = getMessages(getParLanguage(bparams)->code());
1523         return m.get(to_ascii(s));
1524 }
1525
1526
1527 docstring Paragraph::expandLabel(Layout const & layout,
1528                 BufferParams const & bparams, bool process_appendix) const
1529 {
1530         DocumentClass const & tclass = bparams.documentClass();
1531
1532         docstring fmt;
1533         if (process_appendix && d->params_.appendix())
1534                 fmt = translateIfPossible(layout.labelstring_appendix(),
1535                         bparams);
1536         else
1537                 fmt = translateIfPossible(layout.labelstring(), bparams);
1538
1539         if (fmt.empty() && layout.labeltype == LABEL_COUNTER
1540             && !layout.counter.empty())
1541                 fmt = "\\the" + layout.counter;
1542
1543         // handle 'inherited level parts' in 'fmt',
1544         // i.e. the stuff between '@' in   '@Section@.\arabic{subsection}'
1545         size_t const i = fmt.find('@', 0);
1546         if (i != docstring::npos) {
1547                 size_t const j = fmt.find('@', i + 1);
1548                 if (j != docstring::npos) {
1549                         docstring parent(fmt, i + 1, j - i - 1);
1550                         docstring label = from_ascii("??");
1551                         if (tclass.hasLayout(parent))
1552                                 docstring label = expandLabel(tclass[parent], bparams,
1553                                                       process_appendix);
1554                         fmt = docstring(fmt, 0, i) + label
1555                                 + docstring(fmt, j + 1, docstring::npos);
1556                 }
1557         }
1558
1559         return tclass.counters().counterLabel(fmt);
1560 }
1561
1562
1563 void Paragraph::applyLayout(Layout const & new_layout)
1564 {
1565         d->layout_ = &new_layout;
1566         LyXAlignment const oldAlign = d->params_.align();
1567
1568         if (!(oldAlign & d->layout_->alignpossible)) {
1569                 frontend::Alert::warning(_("Alignment not permitted"),
1570                         _("The new layout does not permit the alignment previously used.\nSetting to default."));
1571                 d->params_.align(LYX_ALIGN_LAYOUT);
1572         }
1573 }
1574
1575
1576 pos_type Paragraph::beginOfBody() const
1577 {
1578         return d->begin_of_body_;
1579 }
1580
1581
1582 void Paragraph::setBeginOfBody()
1583 {
1584         if (d->layout_->labeltype != LABEL_MANUAL) {
1585                 d->begin_of_body_ = 0;
1586                 return;
1587         }
1588
1589         // Unroll the first two cycles of the loop
1590         // and remember the previous character to
1591         // remove unnecessary getChar() calls
1592         pos_type i = 0;
1593         pos_type end = size();
1594         if (i < end && !isNewline(i)) {
1595                 ++i;
1596                 char_type previous_char = 0;
1597                 char_type temp = 0;
1598                 if (i < end) {
1599                         previous_char = d->text_[i];
1600                         if (!isNewline(i)) {
1601                                 ++i;
1602                                 while (i < end && previous_char != ' ') {
1603                                         temp = d->text_[i];
1604                                         if (isNewline(i))
1605                                                 break;
1606                                         ++i;
1607                                         previous_char = temp;
1608                                 }
1609                         }
1610                 }
1611         }
1612
1613         d->begin_of_body_ = i;
1614 }
1615
1616
1617 bool Paragraph::forceEmptyLayout() const
1618 {
1619         return inInset() && inInset()->forceEmptyLayout();
1620 }
1621
1622
1623 bool Paragraph::allowParagraphCustomization() const
1624 {
1625         return inInset() && inInset()->allowParagraphCustomization(0);
1626 }
1627
1628
1629 bool Paragraph::useEmptyLayout() const
1630 {
1631         return inInset() && inInset()->useEmptyLayout();
1632 }
1633
1634
1635 namespace {
1636
1637 // paragraphs inside floats need different alignment tags to avoid
1638 // unwanted space
1639
1640 bool noTrivlistCentering(InsetCode code)
1641 {
1642         return code == FLOAT_CODE || code == WRAP_CODE;
1643 }
1644
1645
1646 string correction(string const & orig)
1647 {
1648         if (orig == "flushleft")
1649                 return "raggedright";
1650         if (orig == "flushright")
1651                 return "raggedleft";
1652         if (orig == "center")
1653                 return "centering";
1654         return orig;
1655 }
1656
1657
1658 string const corrected_env(string const & suffix, string const & env,
1659         InsetCode code)
1660 {
1661         string output = suffix + "{";
1662         if (noTrivlistCentering(code))
1663                 output += correction(env);
1664         else
1665                 output += env;
1666         output += "}";
1667         if (suffix == "\\begin")
1668                 output += "\n";
1669         return output;
1670 }
1671
1672
1673 void adjust_row_column(string const & str, TexRow & texrow, int & column)
1674 {
1675         if (!contains(str, "\n"))
1676                 column += str.size();
1677         else {
1678                 string tmp;
1679                 texrow.newline();
1680                 column = rsplit(str, tmp, '\n').size();
1681         }
1682 }
1683
1684 } // namespace anon
1685
1686
1687 int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
1688                                  odocstream & os, TexRow & texrow,
1689                                  bool moving_arg) const
1690 {
1691         int column = 0;
1692
1693         if (params_.noindent()) {
1694                 os << "\\noindent ";
1695                 column += 10;
1696         }
1697
1698         LyXAlignment const curAlign = params_.align();
1699
1700         if (curAlign == layout_->align)
1701                 return column;
1702
1703         switch (curAlign) {
1704         case LYX_ALIGN_NONE:
1705         case LYX_ALIGN_BLOCK:
1706         case LYX_ALIGN_LAYOUT:
1707         case LYX_ALIGN_SPECIAL:
1708                 break;
1709         case LYX_ALIGN_LEFT:
1710         case LYX_ALIGN_RIGHT:
1711         case LYX_ALIGN_CENTER:
1712                 if (moving_arg) {
1713                         os << "\\protect";
1714                         column += 8;
1715                 }
1716                 break;
1717         }
1718
1719         switch (curAlign) {
1720         case LYX_ALIGN_NONE:
1721         case LYX_ALIGN_BLOCK:
1722         case LYX_ALIGN_LAYOUT:
1723         case LYX_ALIGN_SPECIAL:
1724                 break;
1725         case LYX_ALIGN_LEFT: {
1726                 string output;
1727                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
1728                         output = corrected_env("\\begin", "flushleft", owner_->ownerCode());
1729                 else
1730                         output = corrected_env("\\begin", "flushright", owner_->ownerCode());
1731                 os << from_ascii(output);
1732                 adjust_row_column(output, texrow, column);
1733                 break;
1734         } case LYX_ALIGN_RIGHT: {
1735                 string output;
1736                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
1737                         output = corrected_env("\\begin", "flushright", owner_->ownerCode());
1738                 else
1739                         output = corrected_env("\\begin", "flushleft", owner_->ownerCode());
1740                 os << from_ascii(output);
1741                 adjust_row_column(output, texrow, column);
1742                 break;
1743         } case LYX_ALIGN_CENTER: {
1744                 string output;
1745                 output = corrected_env("\\begin", "center", owner_->ownerCode());
1746                 os << from_ascii(output);
1747                 adjust_row_column(output, texrow, column);
1748                 break;
1749         }
1750         }
1751
1752         return column;
1753 }
1754
1755
1756 int Paragraph::Private::endTeXParParams(BufferParams const & bparams,
1757                                odocstream & os, TexRow & texrow,
1758                                bool moving_arg) const
1759 {
1760         int column = 0;
1761
1762         switch (params_.align()) {
1763         case LYX_ALIGN_NONE:
1764         case LYX_ALIGN_BLOCK:
1765         case LYX_ALIGN_LAYOUT:
1766         case LYX_ALIGN_SPECIAL:
1767                 break;
1768         case LYX_ALIGN_LEFT:
1769         case LYX_ALIGN_RIGHT:
1770         case LYX_ALIGN_CENTER:
1771                 if (moving_arg) {
1772                         os << "\\protect";
1773                         column = 8;
1774                 }
1775                 break;
1776         }
1777
1778         switch (params_.align()) {
1779         case LYX_ALIGN_NONE:
1780         case LYX_ALIGN_BLOCK:
1781         case LYX_ALIGN_LAYOUT:
1782         case LYX_ALIGN_SPECIAL:
1783                 break;
1784         case LYX_ALIGN_LEFT: {
1785                 string output;
1786                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
1787                         output = corrected_env("\n\\par\\end", "flushleft", owner_->ownerCode());
1788                 else
1789                         output = corrected_env("\n\\par\\end", "flushright", owner_->ownerCode());
1790                 os << from_ascii(output);
1791                 adjust_row_column(output, texrow, column);
1792                 break;
1793         } case LYX_ALIGN_RIGHT: {
1794                 string output;
1795                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
1796                         output = corrected_env("\n\\par\\end", "flushright", owner_->ownerCode());
1797                 else
1798                         output = corrected_env("\n\\par\\end", "flushleft", owner_->ownerCode());
1799                 os << from_ascii(output);
1800                 adjust_row_column(output, texrow, column);
1801                 break;
1802         } case LYX_ALIGN_CENTER: {
1803                 string output;
1804                 output = corrected_env("\n\\par\\end", "center", owner_->ownerCode());
1805                 os << from_ascii(output);
1806                 adjust_row_column(output, texrow, column);
1807                 break;
1808         }
1809         }
1810
1811         return column;
1812 }
1813
1814
1815 // This one spits out the text of the paragraph
1816 bool Paragraph::latex(BufferParams const & bparams,
1817                                 Font const & outerfont,
1818                                 odocstream & os, TexRow & texrow,
1819                                 OutputParams const & runparams) const
1820 {
1821         LYXERR(Debug::LATEX, "SimpleTeXOnePar...     " << this);
1822
1823         bool return_value = false;
1824
1825         bool asdefault = forceEmptyLayout();
1826
1827         Layout const & style = asdefault ?
1828                 bparams.documentClass().emptyLayout() :
1829                 *d->layout_;
1830
1831         // Current base font for all inherited font changes, without any
1832         // change caused by an individual character, except for the language:
1833         // It is set to the language of the first character.
1834         // As long as we are in the label, this font is the base font of the
1835         // label. Before the first body character it is set to the base font
1836         // of the body.
1837         Font basefont;
1838
1839         // Maybe we have to create a optional argument.
1840         pos_type body_pos = beginOfBody();
1841         unsigned int column = 0;
1842
1843         if (body_pos > 0) {
1844                 // the optional argument is kept in curly brackets in
1845                 // case it contains a ']'
1846                 os << "[{";
1847                 column += 2;
1848                 basefont = getLabelFont(bparams, outerfont);
1849         } else {
1850                 basefont = getLayoutFont(bparams, outerfont);
1851         }
1852
1853         // Which font is currently active?
1854         Font running_font(basefont);
1855         // Do we have an open font change?
1856         bool open_font = false;
1857
1858         Change runningChange = Change(Change::UNCHANGED);
1859
1860         texrow.start(id(), 0);
1861
1862         // if the paragraph is empty, the loop will not be entered at all
1863         if (empty()) {
1864                 if (style.isCommand()) {
1865                         os << '{';
1866                         ++column;
1867                 }
1868                 if (!asdefault)
1869                         column += d->startTeXParParams(bparams, os, texrow,
1870                                                     runparams.moving_arg);
1871         }
1872
1873         for (pos_type i = 0; i < size(); ++i) {
1874                 // First char in paragraph or after label?
1875                 if (i == body_pos) {
1876                         if (body_pos > 0) {
1877                                 if (open_font) {
1878                                         column += running_font.latexWriteEndChanges(
1879                                                 os, bparams, runparams,
1880                                                 basefont, basefont);
1881                                         open_font = false;
1882                                 }
1883                                 basefont = getLayoutFont(bparams, outerfont);
1884                                 running_font = basefont;
1885
1886                                 column += Changes::latexMarkChange(os, bparams,
1887                                                 runningChange, Change(Change::UNCHANGED));
1888                                 runningChange = Change(Change::UNCHANGED);
1889
1890                                 os << "}] ";
1891                                 column +=3;
1892                         }
1893                         if (style.isCommand()) {
1894                                 os << '{';
1895                                 ++column;
1896                         }
1897
1898                         if (!asdefault)
1899                                 column += d->startTeXParParams(bparams, os,
1900                                                             texrow,
1901                                                             runparams.moving_arg);
1902                 }
1903
1904                 Change const & change = runparams.inDeletedInset ? runparams.changeOfDeletedInset
1905                                                                  : lookupChange(i);
1906
1907                 if (bparams.outputChanges && runningChange != change) {
1908                         if (open_font) {
1909                                 column += running_font.latexWriteEndChanges(
1910                                                 os, bparams, runparams, basefont, basefont);
1911                                 open_font = false;
1912                         }
1913                         basefont = getLayoutFont(bparams, outerfont);
1914                         running_font = basefont;
1915
1916                         column += Changes::latexMarkChange(os, bparams, runningChange, change);
1917                         runningChange = change;
1918                 }
1919
1920                 // do not output text which is marked deleted
1921                 // if change tracking output is disabled
1922                 if (!bparams.outputChanges && change.type == Change::DELETED) {
1923                         continue;
1924                 }
1925
1926                 ++column;
1927
1928                 // Fully instantiated font
1929                 Font const font = getFont(bparams, i, outerfont);
1930
1931                 Font const last_font = running_font;
1932
1933                 // Do we need to close the previous font?
1934                 if (open_font &&
1935                     (font != running_font ||
1936                      font.language() != running_font.language()))
1937                 {
1938                         column += running_font.latexWriteEndChanges(
1939                                         os, bparams, runparams, basefont,
1940                                         (i == body_pos-1) ? basefont : font);
1941                         running_font = basefont;
1942                         open_font = false;
1943                 }
1944
1945                 // close babel's font environment before opening CJK.
1946                 if (!running_font.language()->babel().empty() &&
1947                     font.language()->encoding()->package() == Encoding::CJK) {
1948                                 string end_tag = subst(lyxrc.language_command_end,
1949                                                         "$$lang",
1950                                                         running_font.language()->babel());
1951                                 os << from_ascii(end_tag);
1952                                 column += end_tag.length();
1953                 }
1954
1955                 // Switch file encoding if necessary (and allowed)
1956                 if (!runparams.verbatim &&
1957                     runparams.encoding->package() == Encoding::none &&
1958                     font.language()->encoding()->package() == Encoding::none) {
1959                         pair<bool, int> const enc_switch = switchEncoding(os, bparams,
1960                                         runparams, *(font.language()->encoding()));
1961                         if (enc_switch.first) {
1962                                 column += enc_switch.second;
1963                                 runparams.encoding = font.language()->encoding();
1964                         }
1965                 }
1966
1967                 char_type const c = d->text_[i];
1968
1969                 // Do we need to change font?
1970                 if ((font != running_font ||
1971                      font.language() != running_font.language()) &&
1972                         i != body_pos - 1)
1973                 {
1974                         odocstringstream ods;
1975                         column += font.latexWriteStartChanges(ods, bparams,
1976                                                               runparams, basefont,
1977                                                               last_font);
1978                         running_font = font;
1979                         open_font = true;
1980                         docstring fontchange = ods.str();
1981                         // check if the fontchange ends with a trailing blank
1982                         // (like "\small " (see bug 3382)
1983                         if (suffixIs(fontchange, ' ') && c == ' ')
1984                                 os << fontchange.substr(0, fontchange.size() - 1)
1985                                    << from_ascii("{}");
1986                         else
1987                                 os << fontchange;
1988                 }
1989
1990                 if (c == ' ') {
1991                         // FIXME: integrate this case in latexSpecialChar
1992                         // Do not print the separation of the optional argument
1993                         // if style.pass_thru is false. This works because
1994                         // latexSpecialChar ignores spaces if
1995                         // style.pass_thru is false.
1996                         if (i != body_pos - 1) {
1997                                 if (d->simpleTeXBlanks(
1998                                                 runparams, os, texrow,
1999                                                 i, column, font, style)) {
2000                                         // A surrogate pair was output. We
2001                                         // must not call latexSpecialChar
2002                                         // in this iteration, since it would output
2003                                         // the combining character again.
2004                                         ++i;
2005                                         continue;
2006                                 }
2007                         }
2008                 }
2009
2010                 OutputParams rp = runparams;
2011                 rp.free_spacing = style.free_spacing;
2012                 rp.local_font = &font;
2013                 rp.intitle = style.intitle;
2014
2015                 // Two major modes:  LaTeX or plain
2016                 // Handle here those cases common to both modes
2017                 // and then split to handle the two modes separately.
2018                 if (c == META_INSET)
2019                         d->latexInset(bparams, os,
2020                                         texrow, rp, running_font,
2021                                         basefont, outerfont, open_font,
2022                                         runningChange, style, i, column);
2023                 else {
2024                         try {
2025                                 d->latexSpecialChar(os, rp, running_font, runningChange,
2026                                         style, i, column);
2027                         } catch (EncodingException & e) {
2028                                 if (runparams.dryrun) {
2029                                         os << "<" << _("LyX Warning: ")
2030                                            << _("uncodable character") << " '";
2031                                         os.put(c);
2032                                         os << "'>";
2033                                 } else {
2034                                         // add location information and throw again.
2035                                         e.par_id = id();
2036                                         e.pos = i;
2037                                         throw(e);
2038                                 }
2039                         }
2040                 }
2041
2042                 // Set the encoding to that returned from simpleTeXSpecialChars (see
2043                 // comment for encoding member in OutputParams.h)
2044                 runparams.encoding = rp.encoding;
2045         }
2046
2047         // If we have an open font definition, we have to close it
2048         if (open_font) {
2049 #ifdef FIXED_LANGUAGE_END_DETECTION
2050                 if (next_) {
2051                         running_font
2052                                 .latexWriteEndChanges(os, bparams, runparams,
2053                                         basefont,
2054                                         next_->getFont(bparams, 0, outerfont));
2055                 } else {
2056                         running_font.latexWriteEndChanges(os, bparams,
2057                                         runparams, basefont, basefont);
2058                 }
2059 #else
2060 //FIXME: For now we ALWAYS have to close the foreign font settings if they are
2061 //FIXME: there as we start another \selectlanguage with the next paragraph if
2062 //FIXME: we are in need of this. This should be fixed sometime (Jug)
2063                 running_font.latexWriteEndChanges(os, bparams, runparams,
2064                                 basefont, basefont);
2065 #endif
2066         }
2067
2068         column += Changes::latexMarkChange(os, bparams, runningChange, Change(Change::UNCHANGED));
2069
2070         // Needed if there is an optional argument but no contents.
2071         if (body_pos > 0 && body_pos == size()) {
2072                 os << "}]~";
2073                 return_value = false;
2074         }
2075
2076         if (!asdefault) {
2077                 column += d->endTeXParParams(bparams, os, texrow,
2078                                           runparams.moving_arg);
2079         }
2080
2081         LYXERR(Debug::LATEX, "SimpleTeXOnePar...done " << this);
2082         return return_value;
2083 }
2084
2085
2086 bool Paragraph::emptyTag() const
2087 {
2088         for (pos_type i = 0; i < size(); ++i) {
2089                 if (Inset const * inset = getInset(i)) {
2090                         InsetCode lyx_code = inset->lyxCode();
2091                         if (lyx_code != TOC_CODE &&
2092                             lyx_code != INCLUDE_CODE &&
2093                             lyx_code != GRAPHICS_CODE &&
2094                             lyx_code != ERT_CODE &&
2095                             lyx_code != LISTINGS_CODE &&
2096                             lyx_code != FLOAT_CODE &&
2097                             lyx_code != TABULAR_CODE) {
2098                                 return false;
2099                         }
2100                 } else {
2101                         char_type c = d->text_[i];
2102                         if (c != ' ' && c != '\t')
2103                                 return false;
2104                 }
2105         }
2106         return true;
2107 }
2108
2109
2110 string Paragraph::getID(Buffer const & buf, OutputParams const & runparams)
2111         const
2112 {
2113         for (pos_type i = 0; i < size(); ++i) {
2114                 if (Inset const * inset = getInset(i)) {
2115                         InsetCode lyx_code = inset->lyxCode();
2116                         if (lyx_code == LABEL_CODE) {
2117                                 InsetLabel const * const il = static_cast<InsetLabel const *>(inset);
2118                                 docstring const & id = il->getParam("name");
2119                                 return "id='" + to_utf8(sgml::cleanID(buf, runparams, id)) + "'";
2120                         }
2121                 }
2122         }
2123         return string();
2124 }
2125
2126
2127 pos_type Paragraph::firstWord(odocstream & os, OutputParams const & runparams)
2128         const
2129 {
2130         pos_type i;
2131         for (i = 0; i < size(); ++i) {
2132                 if (Inset const * inset = getInset(i)) {
2133                         inset->docbook(os, runparams);
2134                 } else {
2135                         char_type c = d->text_[i];
2136                         if (c == ' ')
2137                                 break;
2138                         os << sgml::escapeChar(c);
2139                 }
2140         }
2141         return i;
2142 }
2143
2144
2145 bool Paragraph::Private::onlyText(Buffer const & buf, Font const & outerfont, pos_type initial) const
2146 {
2147         Font font_old;
2148         pos_type size = text_.size();
2149         for (pos_type i = initial; i < size; ++i) {
2150                 Font font = owner_->getFont(buf.params(), i, outerfont);
2151                 if (text_[i] == META_INSET)
2152                         return false;
2153                 if (i != initial && font != font_old)
2154                         return false;
2155                 font_old = font;
2156         }
2157
2158         return true;
2159 }
2160
2161
2162 void Paragraph::simpleDocBookOnePar(Buffer const & buf,
2163                                     odocstream & os,
2164                                     OutputParams const & runparams,
2165                                     Font const & outerfont,
2166                                     pos_type initial) const
2167 {
2168         bool emph_flag = false;
2169
2170         Layout const & style = *d->layout_;
2171         FontInfo font_old =
2172                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2173
2174         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2175                 os << "]]>";
2176
2177         // parsing main loop
2178         for (pos_type i = initial; i < size(); ++i) {
2179                 Font font = getFont(buf.params(), i, outerfont);
2180
2181                 // handle <emphasis> tag
2182                 if (font_old.emph() != font.fontInfo().emph()) {
2183                         if (font.fontInfo().emph() == FONT_ON) {
2184                                 os << "<emphasis>";
2185                                 emph_flag = true;
2186                         } else if (i != initial) {
2187                                 os << "</emphasis>";
2188                                 emph_flag = false;
2189                         }
2190                 }
2191
2192                 if (Inset const * inset = getInset(i)) {
2193                         inset->docbook(os, runparams);
2194                 } else {
2195                         char_type c = d->text_[i];
2196
2197                         if (style.pass_thru)
2198                                 os.put(c);
2199                         else
2200                                 os << sgml::escapeChar(c);
2201                 }
2202                 font_old = font.fontInfo();
2203         }
2204
2205         if (emph_flag) {
2206                 os << "</emphasis>";
2207         }
2208
2209         if (style.free_spacing)
2210                 os << '\n';
2211         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2212                 os << "<![CDATA[";
2213 }
2214
2215
2216 bool Paragraph::isHfill(pos_type pos) const
2217 {
2218         Inset const * inset = getInset(pos);
2219         return inset && inset->lyxCode() == HFILL_CODE;
2220 }
2221
2222
2223 bool Paragraph::isNewline(pos_type pos) const
2224 {
2225         Inset const * inset = getInset(pos);
2226         return inset && inset->lyxCode() == NEWLINE_CODE;
2227 }
2228
2229
2230 bool Paragraph::isLineSeparator(pos_type pos) const
2231 {
2232         char_type const c = d->text_[pos];
2233         if (isLineSeparatorChar(c))
2234                 return true;
2235         Inset const * inset = getInset(pos);
2236         return inset && inset->isLineSeparator();
2237 }
2238
2239
2240 /// Used by the spellchecker
2241 bool Paragraph::isLetter(pos_type pos) const
2242 {
2243         if (Inset const * inset = getInset(pos))
2244                 return inset->isLetter();
2245         char_type const c = d->text_[pos];
2246         return isLetterChar(c) || isDigit(c);
2247 }
2248
2249
2250 Language const *
2251 Paragraph::getParLanguage(BufferParams const & bparams) const
2252 {
2253         if (!empty())
2254                 return getFirstFontSettings(bparams).language();
2255         // FIXME: we should check the prev par as well (Lgb)
2256         return bparams.language;
2257 }
2258
2259
2260 bool Paragraph::isRTL(BufferParams const & bparams) const
2261 {
2262         return lyxrc.rtl_support
2263                 && getParLanguage(bparams)->rightToLeft()
2264                 && ownerCode() != ERT_CODE
2265                 && ownerCode() != LISTINGS_CODE;
2266 }
2267
2268
2269 void Paragraph::changeLanguage(BufferParams const & bparams,
2270                                Language const * from, Language const * to)
2271 {
2272         // change language including dummy font change at the end
2273         for (pos_type i = 0; i <= size(); ++i) {
2274                 Font font = getFontSettings(bparams, i);
2275                 if (font.language() == from) {
2276                         font.setLanguage(to);
2277                         setFont(i, font);
2278                 }
2279         }
2280 }
2281
2282
2283 bool Paragraph::isMultiLingual(BufferParams const & bparams) const
2284 {
2285         Language const * doc_language = bparams.language;
2286         FontList::const_iterator cit = d->fontlist_.begin();
2287         FontList::const_iterator end = d->fontlist_.end();
2288
2289         for (; cit != end; ++cit)
2290                 if (cit->font().language() != ignore_language &&
2291                     cit->font().language() != latex_language &&
2292                     cit->font().language() != doc_language)
2293                         return true;
2294         return false;
2295 }
2296
2297
2298 docstring Paragraph::asString(bool label) const
2299 {
2300         return asString(0, size(), label);
2301 }
2302
2303
2304 docstring Paragraph::asString(pos_type beg, pos_type end, bool label) const
2305 {
2306
2307         odocstringstream os;
2308
2309         if (beg == 0 && label && !d->params_.labelString().empty())
2310                 os << d->params_.labelString() << ' ';
2311
2312         for (pos_type i = beg; i < end; ++i) {
2313                 char_type const c = d->text_[i];
2314                 if (isPrintable(c))
2315                         os.put(c);
2316                 else if (c == META_INSET)
2317                         getInset(i)->textString(os);
2318         }
2319
2320         return os.str();
2321 }
2322
2323
2324 void Paragraph::setInsetOwner(Inset * inset)
2325 {
2326         d->inset_owner_ = inset;
2327 }
2328
2329
2330 int Paragraph::id() const
2331 {
2332         return d->id_;
2333 }
2334
2335
2336 Layout const & Paragraph::layout() const
2337 {
2338         return *d->layout_;
2339 }
2340
2341
2342 void Paragraph::setLayout(Layout const & layout)
2343 {
2344         d->layout_ = &layout;
2345 }
2346
2347
2348 void Paragraph::setEmptyOrDefaultLayout(DocumentClass const & tclass)
2349 {
2350         if (useEmptyLayout())
2351                 setLayout(tclass.emptyLayout());
2352         else
2353                 setLayout(tclass.defaultLayout());
2354 }
2355
2356
2357 Inset * Paragraph::inInset() const
2358 {
2359         return d->inset_owner_;
2360 }
2361
2362
2363 InsetCode Paragraph::ownerCode() const
2364 {
2365         return d->inset_owner_ ? d->inset_owner_->lyxCode() : NO_CODE;
2366 }
2367
2368
2369 ParagraphParameters & Paragraph::params()
2370 {
2371         return d->params_;
2372 }
2373
2374
2375 ParagraphParameters const & Paragraph::params() const
2376 {
2377         return d->params_;
2378 }
2379
2380
2381 bool Paragraph::isFreeSpacing() const
2382 {
2383         if (d->layout_->free_spacing)
2384                 return true;
2385         return d->inset_owner_ && d->inset_owner_->isFreeSpacing();
2386 }
2387
2388
2389 bool Paragraph::allowEmpty() const
2390 {
2391         if (d->layout_->keepempty)
2392                 return true;
2393         return d->inset_owner_ && d->inset_owner_->allowEmpty();
2394 }
2395
2396
2397 char_type Paragraph::transformChar(char_type c, pos_type pos) const
2398 {
2399         if (!Encodings::is_arabic(c))
2400                 return c;
2401
2402         char_type prev_char = ' ';
2403         char_type next_char = ' ';
2404
2405         for (pos_type i = pos - 1; i >= 0; --i) {
2406                 char_type const par_char = d->text_[i];
2407                 if (!Encodings::isComposeChar_arabic(par_char)) {
2408                         prev_char = par_char;
2409                         break;
2410                 }
2411         }
2412
2413         for (pos_type i = pos + 1, end = size(); i < end; ++i) {
2414                 char_type const par_char = d->text_[i];
2415                 if (!Encodings::isComposeChar_arabic(par_char)) {
2416                         next_char = par_char;
2417                         break;
2418                 }
2419         }
2420
2421         if (Encodings::is_arabic(next_char)) {
2422                 if (Encodings::is_arabic(prev_char) &&
2423                         !Encodings::is_arabic_special(prev_char))
2424                         return Encodings::transformChar(c, Encodings::FORM_MEDIAL);
2425                 else
2426                         return Encodings::transformChar(c, Encodings::FORM_INITIAL);
2427         } else {
2428                 if (Encodings::is_arabic(prev_char) &&
2429                         !Encodings::is_arabic_special(prev_char))
2430                         return Encodings::transformChar(c, Encodings::FORM_FINAL);
2431                 else
2432                         return Encodings::transformChar(c, Encodings::FORM_ISOLATED);
2433         }
2434 }
2435
2436
2437 int Paragraph::checkBiblio(Buffer const & buffer)
2438 {
2439         // FIXME From JS:
2440         // This is getting more and more a mess. ...We really should clean
2441         // up this bibitem issue for 1.6. See also bug 2743.
2442
2443         // Add bibitem insets if necessary
2444         if (d->layout_->labeltype != LABEL_BIBLIO)
2445                 return 0;
2446
2447         bool hasbibitem = !d->insetlist_.empty()
2448                 // Insist on it being in pos 0
2449                 && d->text_[0] == META_INSET
2450                 && d->insetlist_.begin()->inset->lyxCode() == BIBITEM_CODE;
2451
2452         bool track_changes = buffer.params().trackChanges;
2453
2454         docstring oldkey;
2455         docstring oldlabel;
2456
2457         // remove a bibitem in pos != 0
2458         // restore it later in pos 0 if necessary
2459         // (e.g. if a user inserts contents _before_ the item)
2460         // we're assuming there's only one of these, which there
2461         // should be.
2462         int erasedInsetPosition = -1;
2463         InsetList::iterator it = d->insetlist_.begin();
2464         InsetList::iterator end = d->insetlist_.end();
2465         for (; it != end; ++it)
2466                 if (it->inset->lyxCode() == BIBITEM_CODE
2467                     && it->pos > 0) {
2468                         InsetBibitem * olditem = static_cast<InsetBibitem *>(it->inset);
2469                         oldkey = olditem->getParam("key");
2470                         oldlabel = olditem->getParam("label");
2471                         erasedInsetPosition = it->pos;
2472                         eraseChar(erasedInsetPosition, track_changes);
2473                         break;
2474         }
2475
2476         // There was an InsetBibitem at the beginning, and we didn't
2477         // have to erase one.
2478         if (hasbibitem && erasedInsetPosition < 0)
2479                         return 0;
2480
2481         // There was an InsetBibitem at the beginning and we did have to
2482         // erase one. So we give its properties to the beginning inset.
2483         if (hasbibitem) {
2484                 InsetBibitem * inset =
2485                         static_cast<InsetBibitem *>(d->insetlist_.begin()->inset);
2486                 if (!oldkey.empty())
2487                         inset->setParam("key", oldkey);
2488                 inset->setParam("label", oldlabel);
2489                 return -erasedInsetPosition;
2490         }
2491
2492         // There was no inset at the beginning, so we need to create one with
2493         // the key and label of the one we erased.
2494         InsetBibitem * inset = new InsetBibitem(InsetCommandParams(BIBITEM_CODE));
2495         inset->setBuffer(const_cast<Buffer &>(buffer));
2496         // restore values of previously deleted item in this par.
2497         if (!oldkey.empty())
2498                 inset->setParam("key", oldkey);
2499         inset->setParam("label", oldlabel);
2500         insertInset(0, static_cast<Inset *>(inset),
2501                     Change(track_changes ? Change::INSERTED : Change::UNCHANGED));
2502
2503         return 1;
2504 }
2505
2506
2507 void Paragraph::checkAuthors(AuthorList const & authorList)
2508 {
2509         d->changes_.checkAuthors(authorList);
2510 }
2511
2512
2513 bool Paragraph::isUnchanged(pos_type pos) const
2514 {
2515         return lookupChange(pos).type == Change::UNCHANGED;
2516 }
2517
2518
2519 bool Paragraph::isInserted(pos_type pos) const
2520 {
2521         return lookupChange(pos).type == Change::INSERTED;
2522 }
2523
2524
2525 bool Paragraph::isDeleted(pos_type pos) const
2526 {
2527         return lookupChange(pos).type == Change::DELETED;
2528 }
2529
2530
2531 InsetList const & Paragraph::insetList() const
2532 {
2533         return d->insetlist_;
2534 }
2535
2536
2537 Inset * Paragraph::releaseInset(pos_type pos)
2538 {
2539         Inset * inset = d->insetlist_.release(pos);
2540         /// does not honour change tracking!
2541         eraseChar(pos, false);
2542         return inset;
2543 }
2544
2545
2546 Inset * Paragraph::getInset(pos_type pos)
2547 {
2548         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
2549                  ? d->insetlist_.get(pos) : 0;
2550 }
2551
2552
2553 Inset const * Paragraph::getInset(pos_type pos) const
2554 {
2555         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
2556                  ? d->insetlist_.get(pos) : 0;
2557 }
2558
2559
2560 void Paragraph::changeCase(BufferParams const & bparams, pos_type pos,
2561                 pos_type & right, TextCase action)
2562 {
2563         // process sequences of modified characters; in change
2564         // tracking mode, this approach results in much better
2565         // usability than changing case on a char-by-char basis
2566         docstring changes;
2567
2568         bool const trackChanges = bparams.trackChanges;
2569
2570         bool capitalize = true;
2571
2572         for (; pos < right; ++pos) {
2573                 char_type oldChar = d->text_[pos];
2574                 char_type newChar = oldChar;
2575
2576                 // ignore insets and don't play with deleted text!
2577                 if (oldChar != META_INSET && !isDeleted(pos)) {
2578                         switch (action) {
2579                                 case text_lowercase:
2580                                         newChar = lowercase(oldChar);
2581                                         break;
2582                                 case text_capitalization:
2583                                         if (capitalize) {
2584                                                 newChar = uppercase(oldChar);
2585                                                 capitalize = false;
2586                                         }
2587                                         break;
2588                                 case text_uppercase:
2589                                         newChar = uppercase(oldChar);
2590                                         break;
2591                         }
2592                 }
2593
2594                 if (!isLetter(pos) || isDeleted(pos)) {
2595                         // permit capitalization again
2596                         capitalize = true;
2597                 }
2598
2599                 if (oldChar != newChar)
2600                         changes += newChar;
2601
2602                 if (oldChar == newChar || pos == right - 1) {
2603                         if (oldChar != newChar) {
2604                                 // step behind the changing area
2605                                 pos++;
2606                         }
2607                         int erasePos = pos - changes.size();
2608                         for (size_t i = 0; i < changes.size(); i++) {
2609                                 insertChar(pos, changes[i],
2610                                         getFontSettings(bparams,
2611                                         erasePos),
2612                                         trackChanges);
2613                                 if (!eraseChar(erasePos, trackChanges)) {
2614                                         ++erasePos;
2615                                         ++pos; // advance
2616                                         ++right; // expand selection
2617                                 }
2618                         }
2619                         changes.clear();
2620                 }
2621         }
2622 }
2623
2624
2625 bool Paragraph::find(docstring const & str, bool cs, bool mw,
2626                 pos_type pos, bool del) const
2627 {
2628         int const strsize = str.length();
2629         int i = 0;
2630         pos_type const parsize = d->text_.size();
2631         for (i = 0; pos + i < parsize; ++i) {
2632                 if (i >= strsize)
2633                         break;
2634                 if (cs && str[i] != d->text_[pos + i])
2635                         break;
2636                 if (!cs && uppercase(str[i]) != uppercase(d->text_[pos + i]))
2637                         break;
2638                 if (!del && isDeleted(pos + i))
2639                         break;
2640         }
2641
2642         if (i != strsize)
2643                 return false;
2644
2645         // if necessary, check whether string matches word
2646         if (mw) {
2647                 if (pos > 0 && isLetter(pos - 1))
2648                         return false;
2649                 if (pos + strsize < parsize
2650                         && isLetter(pos + strsize))
2651                         return false;
2652         }
2653
2654         return true;
2655 }
2656
2657
2658 char_type Paragraph::getChar(pos_type pos) const
2659 {
2660         return d->text_[pos];
2661 }
2662
2663
2664 pos_type Paragraph::size() const
2665 {
2666         return d->text_.size();
2667 }
2668
2669
2670 bool Paragraph::empty() const
2671 {
2672         return d->text_.empty();
2673 }
2674
2675
2676 bool Paragraph::isInset(pos_type pos) const
2677 {
2678         return d->text_[pos] == META_INSET;
2679 }
2680
2681
2682 bool Paragraph::isSeparator(pos_type pos) const
2683 {
2684         //FIXME: Are we sure this can be the only separator?
2685         return d->text_[pos] == ' ';
2686 }
2687
2688
2689 void Paragraph::deregisterWords()
2690 {
2691         Private::Words::const_iterator it;
2692         WordList & wl = theWordList();
2693         for (it = d->words_.begin(); it != d->words_.end(); ++it)
2694                 wl.remove(*it);
2695         d->words_.clear();
2696 }
2697
2698
2699 void Paragraph::collectWords(CursorSlice const & sl)
2700 {
2701         // find new words
2702         bool inword = false;
2703
2704         //lyxerr << "Words: ";
2705         pos_type n = size();
2706         for (pos_type pos = 0; pos != n; ++pos) {
2707                 if (isDeleted(pos))
2708                         continue;
2709
2710                 if (!isLetter(pos)) {
2711                         inword = false;
2712                         continue;
2713                 }
2714
2715                 if (inword)
2716                         continue;
2717
2718                 inword = true;
2719                 CursorSlice from = sl;
2720                 CursorSlice to = sl;
2721                 from.pos() = pos;
2722                 to.pos() = pos;
2723                 from.text()->getWord(from, to, WHOLE_WORD);
2724                 if (to.pos() - from.pos() < 6)
2725                         continue;
2726                 docstring word = asString(from.pos(), to.pos(), false);
2727                 d->words_.insert(word);
2728                 //lyxerr << word << " ";
2729         }
2730         //lyxerr << std::endl;
2731 }
2732
2733
2734 void Paragraph::registerWords()
2735 {
2736         Private::Words::const_iterator it;
2737         WordList & wl = theWordList();
2738         for (it = d->words_.begin(); it != d->words_.end(); ++it)
2739                 wl.insert(*it);
2740 }
2741
2742
2743 void Paragraph::updateWords(CursorSlice const & sl)
2744 {
2745         BOOST_ASSERT(&sl.paragraph() == this);
2746         deregisterWords();
2747         collectWords(sl);
2748         registerWords();
2749 }
2750
2751 } // namespace lyx