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 Richard Heck (XHTML output)
   9  * \author Jean-Marc Lasgouttes
  10  * \author Angus Leeming
  11  * \author John Levon
  12  * \author André Pönitz
  13  * \author Dekel Tsur
  14  * \author Jürgen Vigna
  15  *
  16  * Full author contact details are available in file CREDITS.
  17  */
  18
  19 #include <config.h>
  20
  21 #include "Paragraph.h"
  22
  23 #include "LayoutFile.h"
  24 #include "Buffer.h"
  25 #include "BufferParams.h"
  26 #include "Changes.h"
  27 #include "Counters.h"
  28 #include "Encoding.h"
  29 #include "InsetList.h"
  30 #include "Language.h"
  31 #include "LaTeXFeatures.h"
  32 #include "Layout.h"
  33 #include "Length.h"
  34 #include "Font.h"
  35 #include "FontList.h"
  36 #include "LyXRC.h"
  37 #include "OutputParams.h"
  38 #include "output_latex.h"
  39 #include "output_xhtml.h"
  40 #include "ParagraphParameters.h"
  41 #include "SpellChecker.h"
  42 #include "sgml.h"
  43 #include "TextClass.h"
  44 #include "TexRow.h"
  45 #include "Text.h"
  46 #include "VSpace.h"
  47 #include "WordLangTuple.h"
  48 #include "WordList.h"
  49
  50 #include "frontends/alert.h"
  51
  52 #include "insets/InsetBibitem.h"
  53 #include "insets/InsetLabel.h"
  54
  55 #include "support/debug.h"
  56 #include "support/docstring_list.h"
  57 #include "support/ExceptionMessage.h"
  58 #include "support/gettext.h"
  59 #include "support/lassert.h"
  60 #include "support/lstrings.h"
  61 #include "support/textutils.h"
  62
  63 #include <sstream>
  64 #include <vector>
  65
  66 using namespace std;
  67 using namespace lyx::support;
  68
  69 namespace lyx {
  70
  71 namespace {
  72 /// Inset identifier (above 0x10ffff, for ucs-4)
  73 char_type const META_INSET = 0x200001;
  74 };
  75
  76 /////////////////////////////////////////////////////////////////////
  77 //
  78 // Paragraph::Private
  79 //
  80 /////////////////////////////////////////////////////////////////////
  81
  82 class Paragraph::Private
  83 {
  84 public:
  85         ///
  86         Private(Paragraph * owner, Layout const & layout);
  87         /// "Copy constructor"
  88         Private(Private const &, Paragraph * owner);
  89         /// Copy constructor from \p beg  to \p end
  90         Private(Private const &, Paragraph * owner, pos_type beg, pos_type end);
  91
  92         ///
  93         void insertChar(pos_type pos, char_type c, Change const & change);
  94
  95         /// Output the surrogate pair formed by \p c and \p next to \p os.
  96         /// \return the number of characters written.
  97         int latexSurrogatePair(odocstream & os, char_type c, char_type next,
  98                                OutputParams const &);
  99
 100         /// Output a space in appropriate formatting (or a surrogate pair
 101         /// if the next character is a combining character).
 102         /// \return whether a surrogate pair was output.
 103         bool simpleTeXBlanks(OutputParams const &,
 104                              odocstream &, TexRow & texrow,
 105                              pos_type i,
 106                              unsigned int & column,
 107                              Font const & font,
 108                              Layout const & style);
 109
 110         /// Output consecutive unicode chars, belonging to the same script as
 111         /// specified by the latex macro \p ltx, to \p os starting from \p i.
 112         /// \return the number of characters written.
 113         int writeScriptChars(odocstream & os, docstring const & ltx,
 114                            Change const &, Encoding const &, pos_type & i);
 115
 116         /// This could go to ParagraphParameters if we want to.
 117         int startTeXParParams(BufferParams const &, odocstream &, TexRow &,
 118                               OutputParams const &) const;
 119
 120         /// This could go to ParagraphParameters if we want to.
 121         int endTeXParParams(BufferParams const &, odocstream &, TexRow &,
 122                             OutputParams const &) const;
 123
 124         ///
 125         void latexInset(BufferParams const &,
 126                                    odocstream &,
 127                                    TexRow & texrow, OutputParams &,
 128                                    Font & running_font,
 129                                    Font & basefont,
 130                                    Font const & outerfont,
 131                                    bool & open_font,
 132                                    Change & running_change,
 133                                    Layout const & style,
 134                                    pos_type & i,
 135                                    unsigned int & column);
 136
 137         ///
 138         void latexSpecialChar(
 139                                    odocstream & os,
 140                                    OutputParams const & runparams,
 141                                    Font const & running_font,
 142                                    Change const & running_change,
 143                                    Layout const & style,
 144                                    pos_type & i,
 145                                    unsigned int & column);
 146
 147         ///
 148         bool latexSpecialT1(
 149                 char_type const c,
 150                 odocstream & os,
 151                 pos_type i,
 152                 unsigned int & column);
 153         ///
 154         bool latexSpecialTypewriter(
 155                 char_type const c,
 156                 odocstream & os,
 157                 pos_type i,
 158                 unsigned int & column);
 159         ///
 160         bool latexSpecialPhrase(
 161                 odocstream & os,
 162                 pos_type & i,
 163                 unsigned int & column,
 164                 OutputParams const & runparams);
 165
 166         ///
 167         void validate(LaTeXFeatures & features) const;
 168
 169         /// Checks if the paragraph contains only text and no inset or font change.
 170         bool onlyText(Buffer const & buf, Font const & outerfont,
 171                       pos_type initial) const;
 172
 173         /// match a string against a particular point in the paragraph
 174         bool isTextAt(string const & str, pos_type pos) const;
 175
 176
 177         InsetCode ownerCode() const
 178         {
 179                 return inset_owner_ ? inset_owner_->lyxCode() : NO_CODE;
 180         }
 181
 182         /// Which Paragraph owns us?
 183         Paragraph * owner_;
 184
 185         /// In which Inset?
 186         Inset const * inset_owner_;
 187
 188         ///
 189         FontList fontlist_;
 190
 191         ///
 192         int id_;
 193
 194         ///
 195         ParagraphParameters params_;
 196
 197         /// for recording and looking up changes
 198         Changes changes_;
 199
 200         ///
 201         InsetList insetlist_;
 202
 203         /// end of label
 204         pos_type begin_of_body_;
 205
 206         typedef docstring TextContainer;
 207         ///
 208         TextContainer text_;
 209
 210         typedef set<docstring> Words;
 211         typedef map<Language, Words> LangWordsMap;
 212         ///
 213         LangWordsMap words_;
 214         ///
 215         Layout const * layout_;
 216 };
 217
 218
 219 namespace {
 220
 221 struct special_phrase {
 222         string phrase;
 223         docstring macro;
 224         bool builtin;
 225 };
 226
 227 special_phrase const special_phrases[] = {
 228         { "LyX", from_ascii("\\LyX{}"), false },
 229         { "TeX", from_ascii("\\TeX{}"), true },
 230         { "LaTeX2e", from_ascii("\\LaTeXe{}"), true },
 231         { "LaTeX", from_ascii("\\LaTeX{}"), true },
 232 };
 233
 234 size_t const phrases_nr = sizeof(special_phrases)/sizeof(special_phrase);
 235
 236 } // namespace anon
 237
 238
 239 Paragraph::Private::Private(Paragraph * owner, Layout const & layout)
 240         : owner_(owner), inset_owner_(0), id_(-1), begin_of_body_(0), layout_(&layout)
 241 {
 242         text_.reserve(100);
 243 }
 244
 245
 246 // Initialization of the counter for the paragraph id's,
 247 //
 248 // FIXME: There should be a more intelligent way to generate and use the
 249 // paragraph ids per buffer instead a global static counter for all InsetText
 250 // in the running program.
 251 static int paragraph_id = -1;
 252
 253 Paragraph::Private::Private(Private const & p, Paragraph * owner)
 254         : owner_(owner), inset_owner_(p.inset_owner_), fontlist_(p.fontlist_),
 255           params_(p.params_), changes_(p.changes_), insetlist_(p.insetlist_),
 256           begin_of_body_(p.begin_of_body_), text_(p.text_), words_(p.words_),
 257           layout_(p.layout_)
 258 {
 259         id_ = ++paragraph_id;
 260 }
 261
 262
 263 Paragraph::Private::Private(Private const & p, Paragraph * owner,
 264         pos_type beg, pos_type end)
 265         : owner_(owner), inset_owner_(p.inset_owner_),
 266           params_(p.params_), changes_(p.changes_),
 267           insetlist_(p.insetlist_, beg, end),
 268           begin_of_body_(p.begin_of_body_), words_(p.words_),
 269           layout_(p.layout_)
 270 {
 271         id_ = ++paragraph_id;
 272         if (beg >= pos_type(p.text_.size()))
 273                 return;
 274         text_ = p.text_.substr(beg, end - beg);
 275
 276         FontList::const_iterator fcit = fontlist_.begin();
 277         FontList::const_iterator fend = fontlist_.end();
 278         for (; fcit != fend; ++fcit) {
 279                 if (fcit->pos() < beg)
 280                         continue;
 281                 if (fcit->pos() >= end) {
 282                         // Add last entry in the fontlist_.
 283                         fontlist_.set(text_.size() - 1, fcit->font());
 284                         break;
 285                 }
 286                 // Add a new entry in the fontlist_.
 287                 fontlist_.set(fcit->pos() - beg, fcit->font());
 288         }
 289 }
 290
 291
 292 void Paragraph::addChangesToToc(DocIterator const & cdit,
 293         Buffer const & buf) const
 294 {
 295         d->changes_.addToToc(cdit, buf);
 296 }
 297
 298
 299 bool Paragraph::isDeleted(pos_type start, pos_type end) const
 300 {
 301         LASSERT(start >= 0 && start <= size(), /**/);
 302         LASSERT(end > start && end <= size() + 1, /**/);
 303
 304         return d->changes_.isDeleted(start, end);
 305 }
 306
 307
 308 bool Paragraph::isChanged(pos_type start, pos_type end) const
 309 {
 310         LASSERT(start >= 0 && start <= size(), /**/);
 311         LASSERT(end > start && end <= size() + 1, /**/);
 312
 313         return d->changes_.isChanged(start, end);
 314 }
 315
 316
 317 bool Paragraph::isMergedOnEndOfParDeletion(bool trackChanges) const
 318 {
 319         // keep the logic here in sync with the logic of eraseChars()
 320         if (!trackChanges)
 321                 return true;
 322
 323         Change const change = d->changes_.lookup(size());
 324         return change.inserted() && change.currentAuthor();
 325 }
 326
 327
 328 void Paragraph::setChange(Change const & change)
 329 {
 330         // beware of the imaginary end-of-par character!
 331         d->changes_.set(change, 0, size() + 1);
 332
 333         /*
 334          * Propagate the change recursively - but not in case of DELETED!
 335          *
 336          * Imagine that your co-author makes changes in an existing inset. He
 337          * sends your document to you and you come to the conclusion that the
 338          * inset should go completely. If you erase it, LyX must not delete all
 339          * text within the inset. Otherwise, the change tracked insertions of
 340          * your co-author get lost and there is no way to restore them later.
 341          *
 342          * Conclusion: An inset's content should remain untouched if you delete it
 343          */
 344
 345         if (!change.deleted()) {
 346                 for (pos_type pos = 0; pos < size(); ++pos) {
 347                         if (Inset * inset = getInset(pos))
 348                                 inset->setChange(change);
 349                 }
 350         }
 351 }
 352
 353
 354 void Paragraph::setChange(pos_type pos, Change const & change)
 355 {
 356         LASSERT(pos >= 0 && pos <= size(), /**/);
 357         d->changes_.set(change, pos);
 358
 359         // see comment in setChange(Change const &) above
 360         if (!change.deleted() && pos < size())
 361                         if (Inset * inset = getInset(pos))
 362                                 inset->setChange(change);
 363 }
 364
 365
 366 Change const & Paragraph::lookupChange(pos_type pos) const
 367 {
 368         LASSERT(pos >= 0 && pos <= size(), /**/);
 369         return d->changes_.lookup(pos);
 370 }
 371
 372
 373 void Paragraph::acceptChanges(pos_type start, pos_type end)
 374 {
 375         LASSERT(start >= 0 && start <= size(), /**/);
 376         LASSERT(end > start && end <= size() + 1, /**/);
 377
 378         for (pos_type pos = start; pos < end; ++pos) {
 379                 switch (lookupChange(pos).type) {
 380                         case Change::UNCHANGED:
 381                                 // accept changes in nested inset
 382                                 if (Inset * inset = getInset(pos))
 383                                         inset->acceptChanges();
 384                                 break;
 385
 386                         case Change::INSERTED:
 387                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 388                                 // also accept changes in nested inset
 389                                 if (Inset * inset = getInset(pos))
 390                                         inset->acceptChanges();
 391                                 break;
 392
 393                         case Change::DELETED:
 394                                 // Suppress access to non-existent
 395                                 // "end-of-paragraph char"
 396                                 if (pos < size()) {
 397                                         eraseChar(pos, false);
 398                                         --end;
 399                                         --pos;
 400                                 }
 401                                 break;
 402                 }
 403
 404         }
 405 }
 406
 407
 408 void Paragraph::rejectChanges(pos_type start, pos_type end)
 409 {
 410         LASSERT(start >= 0 && start <= size(), /**/);
 411         LASSERT(end > start && end <= size() + 1, /**/);
 412
 413         for (pos_type pos = start; pos < end; ++pos) {
 414                 switch (lookupChange(pos).type) {
 415                         case Change::UNCHANGED:
 416                                 // reject changes in nested inset
 417                                 if (Inset * inset = getInset(pos))
 418                                                 inset->rejectChanges();
 419                                 break;
 420
 421                         case Change::INSERTED:
 422                                 // Suppress access to non-existent
 423                                 // "end-of-paragraph char"
 424                                 if (pos < size()) {
 425                                         eraseChar(pos, false);
 426                                         --end;
 427                                         --pos;
 428                                 }
 429                                 break;
 430
 431                         case Change::DELETED:
 432                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 433
 434                                 // Do NOT reject changes within a deleted inset!
 435                                 // There may be insertions of a co-author inside of it!
 436
 437                                 break;
 438                 }
 439         }
 440 }
 441
 442
 443 void Paragraph::Private::insertChar(pos_type pos, char_type c,
 444                 Change const & change)
 445 {
 446         LASSERT(pos >= 0 && pos <= int(text_.size()), /**/);
 447
 448         // track change
 449         changes_.insert(change, pos);
 450
 451         // This is actually very common when parsing buffers (and
 452         // maybe inserting ascii text)
 453         if (pos == pos_type(text_.size())) {
 454                 // when appending characters, no need to update tables
 455                 text_.push_back(c);
 456                 return;
 457         }
 458
 459         text_.insert(text_.begin() + pos, c);
 460
 461         // Update the font table.
 462         fontlist_.increasePosAfterPos(pos);
 463
 464         // Update the insets
 465         insetlist_.increasePosAfterPos(pos);
 466 }
 467
 468
 469 bool Paragraph::insertInset(pos_type pos, Inset * inset,
 470                                    Change const & change)
 471 {
 472         LASSERT(inset, /**/);
 473         LASSERT(pos >= 0 && pos <= size(), /**/);
 474
 475         // Paragraph::insertInset() can be used in cut/copy/paste operation where
 476         // d->inset_owner_ is not set yet.
 477         if (d->inset_owner_ && !d->inset_owner_->insetAllowed(inset->lyxCode()))
 478                 return false;
 479
 480         d->insertChar(pos, META_INSET, change);
 481         LASSERT(d->text_[pos] == META_INSET, /**/);
 482
 483         // Add a new entry in the insetlist_.
 484         d->insetlist_.insert(inset, pos);
 485         return true;
 486 }
 487
 488
 489 bool Paragraph::eraseChar(pos_type pos, bool trackChanges)
 490 {
 491         LASSERT(pos >= 0 && pos <= size(), return false);
 492
 493         // keep the logic here in sync with the logic of isMergedOnEndOfParDeletion()
 494
 495         if (trackChanges) {
 496                 Change change = d->changes_.lookup(pos);
 497
 498                 // set the character to DELETED if
 499                 //  a) it was previously unchanged or
 500                 //  b) it was inserted by a co-author
 501
 502                 if (!change.changed() ||
 503                       (change.inserted() && !change.currentAuthor())) {
 504                         setChange(pos, Change(Change::DELETED));
 505                         return false;
 506                 }
 507
 508                 if (change.deleted())
 509                         return false;
 510         }
 511
 512         // Don't physically access the imaginary end-of-paragraph character.
 513         // eraseChar() can only mark it as DELETED. A physical deletion of
 514         // end-of-par must be handled externally.
 515         if (pos == size()) {
 516                 return false;
 517         }
 518
 519         // track change
 520         d->changes_.erase(pos);
 521
 522         // if it is an inset, delete the inset entry
 523         if (d->text_[pos] == META_INSET)
 524                 d->insetlist_.erase(pos);
 525
 526         d->text_.erase(d->text_.begin() + pos);
 527
 528         // Update the fontlist_
 529         d->fontlist_.erase(pos);
 530
 531         // Update the insetlist_
 532         d->insetlist_.decreasePosAfterPos(pos);
 533
 534         return true;
 535 }
 536
 537
 538 int Paragraph::eraseChars(pos_type start, pos_type end, bool trackChanges)
 539 {
 540         LASSERT(start >= 0 && start <= size(), /**/);
 541         LASSERT(end >= start && end <= size() + 1, /**/);
 542
 543         pos_type i = start;
 544         for (pos_type count = end - start; count; --count) {
 545                 if (!eraseChar(i, trackChanges))
 546                         ++i;
 547         }
 548         return end - i;
 549 }
 550
 551
 552 int Paragraph::Private::latexSurrogatePair(odocstream & os, char_type c,
 553                 char_type next, OutputParams const & runparams)
 554 {
 555         // Writing next here may circumvent a possible font change between
 556         // c and next. Since next is only output if it forms a surrogate pair
 557         // with c we can ignore this:
 558         // A font change inside a surrogate pair does not make sense and is
 559         // hopefully impossible to input.
 560         // FIXME: change tracking
 561         // Is this correct WRT change tracking?
 562         Encoding const & encoding = *(runparams.encoding);
 563         docstring const latex1 = encoding.latexChar(next);
 564         docstring const latex2 = encoding.latexChar(c);
 565         if (docstring(1, next) == latex1) {
 566                 // the encoding supports the combination
 567                 os << latex2 << latex1;
 568                 return latex1.length() + latex2.length();
 569         } else if (runparams.local_font &&
 570                    runparams.local_font->language()->lang() == "polutonikogreek") {
 571                 // polutonikogreek only works without the brackets
 572                 os << latex1 << latex2;
 573                 return latex1.length() + latex2.length();
 574         } else
 575                 os << latex1 << '{' << latex2 << '}';
 576         return latex1.length() + latex2.length() + 2;
 577 }
 578
 579
 580 bool Paragraph::Private::simpleTeXBlanks(OutputParams const & runparams,
 581                                        odocstream & os, TexRow & texrow,
 582                                        pos_type i,
 583                                        unsigned int & column,
 584                                        Font const & font,
 585                                        Layout const & style)
 586 {
 587         if (style.pass_thru || runparams.verbatim)
 588                 return false;
 589
 590         if (i + 1 < int(text_.size())) {
 591                 char_type next = text_[i + 1];
 592                 if (Encodings::isCombiningChar(next)) {
 593                         // This space has an accent, so we must always output it.
 594                         column += latexSurrogatePair(os, ' ', next, runparams) - 1;
 595                         return true;
 596                 }
 597         }
 598
 599         if (runparams.linelen > 0
 600             && column > runparams.linelen
 601             && i
 602             && text_[i - 1] != ' '
 603             && (i + 1 < int(text_.size()))
 604             // same in FreeSpacing mode
 605             && !owner_->isFreeSpacing()
 606             // In typewriter mode, we want to avoid
 607             // ! . ? : at the end of a line
 608             && !(font.fontInfo().family() == TYPEWRITER_FAMILY
 609                  && (text_[i - 1] == '.'
 610                      || text_[i - 1] == '?'
 611                      || text_[i - 1] == ':'
 612                      || text_[i - 1] == '!'))) {
 613                 os << '\n';
 614                 texrow.newline();
 615                 texrow.start(owner_->id(), i + 1);
 616                 column = 0;
 617         } else if (style.free_spacing) {
 618                 os << '~';
 619         } else {
 620                 os << ' ';
 621         }
 622         return false;
 623 }
 624
 625
 626 int Paragraph::Private::writeScriptChars(odocstream & os,
 627                                          docstring const & ltx,
 628                                          Change const & runningChange,
 629                                          Encoding const & encoding,
 630                                          pos_type & i)
 631 {
 632         // FIXME: modifying i here is not very nice...
 633
 634         // We only arrive here when a proper language for character text_[i] has
 635         // not been specified (i.e., it could not be translated in the current
 636         // latex encoding) or its latex translation has been forced, and it
 637         // belongs to a known script.
 638         // Parameter ltx contains the latex translation of text_[i] as specified
 639         // in the unicodesymbols file and is something like "\textXXX{<spec>}".
 640         // The latex macro name "textXXX" specifies the script to which text_[i]
 641         // belongs and we use it in order to check whether characters from the
 642         // same script immediately follow, such that we can collect them in a
 643         // single "\textXXX" macro. So, we have to retain "\textXXX{<spec>"
 644         // for the first char but only "<spec>" for all subsequent chars.
 645         docstring::size_type const brace1 = ltx.find_first_of(from_ascii("{"));
 646         docstring::size_type const brace2 = ltx.find_last_of(from_ascii("}"));
 647         string script = to_ascii(ltx.substr(1, brace1 - 1));
 648         int pos = 0;
 649         int length = brace2;
 650         bool closing_brace = true;
 651         if (script == "textgreek" && encoding.latexName() == "iso-8859-7") {
 652                 // Correct encoding is being used, so we can avoid \textgreek.
 653                 pos = brace1 + 1;
 654                 length -= pos;
 655                 closing_brace = false;
 656         }
 657         os << ltx.substr(pos, length);
 658         int size = text_.size();
 659         while (i + 1 < size) {
 660                 char_type const next = text_[i + 1];
 661                 // Stop here if next character belongs to another script
 662                 // or there is a change in change tracking status.
 663                 if (!Encodings::isKnownScriptChar(next, script) ||
 664                     runningChange != owner_->lookupChange(i + 1))
 665                         break;
 666                 Font prev_font;
 667                 bool found = false;
 668                 FontList::const_iterator cit = fontlist_.begin();
 669                 FontList::const_iterator end = fontlist_.end();
 670                 for (; cit != end; ++cit) {
 671                         if (cit->pos() >= i && !found) {
 672                                 prev_font = cit->font();
 673                                 found = true;
 674                         }
 675                         if (cit->pos() >= i + 1)
 676                                 break;
 677                 }
 678                 // Stop here if there is a font attribute or encoding change.
 679                 if (found && cit != end && prev_font != cit->font())
 680                         break;
 681                 docstring const latex = encoding.latexChar(next);
 682                 docstring::size_type const b1 =
 683                                         latex.find_first_of(from_ascii("{"));
 684                 docstring::size_type const b2 =
 685                                         latex.find_last_of(from_ascii("}"));
 686                 int const len = b2 - b1 - 1;
 687                 os << latex.substr(b1 + 1, len);
 688                 length += len;
 689                 ++i;
 690         }
 691         if (closing_brace) {
 692                 os << '}';
 693                 ++length;
 694         }
 695         return length;
 696 }
 697
 698
 699 bool Paragraph::Private::isTextAt(string const & str, pos_type pos) const
 700 {
 701         pos_type const len = str.length();
 702
 703         // is the paragraph large enough?
 704         if (pos + len > int(text_.size()))
 705                 return false;
 706
 707         // does the wanted text start at point?
 708         for (string::size_type i = 0; i < str.length(); ++i) {
 709                 // Caution: direct comparison of characters works only
 710                 // because str is pure ASCII.
 711                 if (str[i] != text_[pos + i])
 712                         return false;
 713         }
 714
 715         return fontlist_.hasChangeInRange(pos, len);
 716 }
 717
 718
 719 void Paragraph::Private::latexInset(
 720                                              BufferParams const & bparams,
 721                                              odocstream & os,
 722                                              TexRow & texrow,
 723                                              OutputParams & runparams,
 724                                              Font & running_font,
 725                                              Font & basefont,
 726                                              Font const & outerfont,
 727                                              bool & open_font,
 728                                              Change & running_change,
 729                                              Layout const & style,
 730                                              pos_type & i,
 731                                              unsigned int & column)
 732 {
 733         Inset * inset = owner_->getInset(i);
 734         LASSERT(inset, /**/);
 735
 736         if (style.pass_thru) {
 737                 inset->plaintext(os, runparams);
 738                 return;
 739         }
 740
 741         // FIXME: move this to InsetNewline::latex
 742         if (inset->lyxCode() == NEWLINE_CODE) {
 743                 // newlines are handled differently here than
 744                 // the default in simpleTeXSpecialChars().
 745                 if (!style.newline_allowed) {
 746                         os << '\n';
 747                 } else {
 748                         if (open_font) {
 749                                 column += running_font.latexWriteEndChanges(
 750                                         os, bparams, runparams,
 751                                         basefont, basefont);
 752                                 open_font = false;
 753                         }
 754
 755                         if (running_font.fontInfo().family() == TYPEWRITER_FAMILY)
 756                                 os << '~';
 757
 758                         basefont = owner_->getLayoutFont(bparams, outerfont);
 759                         running_font = basefont;
 760
 761                         if (runparams.moving_arg)
 762                                 os << "\\protect ";
 763
 764                 }
 765                 texrow.newline();
 766                 texrow.start(owner_->id(), i + 1);
 767                 column = 0;
 768         }
 769
 770         if (owner_->isDeleted(i)) {
 771                 if( ++runparams.inDeletedInset == 1)
 772                         runparams.changeOfDeletedInset = owner_->lookupChange(i);
 773         }
 774
 775         if (inset->canTrackChanges()) {
 776                 column += Changes::latexMarkChange(os, bparams, running_change,
 777                         Change(Change::UNCHANGED), runparams);
 778                 running_change = Change(Change::UNCHANGED);
 779         }
 780
 781         bool close = false;
 782         odocstream::pos_type const len = os.tellp();
 783
 784         if (inset->forceLTR()
 785             && running_font.isRightToLeft()
 786             // ERT is an exception, it should be output with no
 787             // decorations at all
 788             && inset->lyxCode() != ERT_CODE) {
 789                 if (running_font.language()->lang() == "farsi")
 790                         os << "\\beginL{}";
 791                 else
 792                         os << "\\L{";
 793                 close = true;
 794         }
 795
 796         // FIXME: Bug: we can have an empty font change here!
 797         // if there has just been a font change, we are going to close it
 798         // right now, which means stupid latex code like \textsf{}. AFAIK,
 799         // this does not harm dvi output. A minor bug, thus (JMarc)
 800
 801         // Some insets cannot be inside a font change command.
 802         // However, even such insets *can* be placed in \L or \R
 803         // or their equivalents (for RTL language switches), so we don't
 804         // close the language in those cases.
 805         // ArabTeX, though, cannot handle this special behavior, it seems.
 806         bool arabtex = basefont.language()->lang() == "arabic_arabtex"
 807                 || running_font.language()->lang() == "arabic_arabtex";
 808         if (open_font && inset->noFontChange()) {
 809                 bool closeLanguage = arabtex
 810                         || basefont.isRightToLeft() == running_font.isRightToLeft();
 811                 unsigned int count = running_font.latexWriteEndChanges(os,
 812                         bparams, runparams, basefont, basefont, closeLanguage);
 813                 column += count;
 814                 // if any font properties were closed, update the running_font,
 815                 // making sure, however, to leave the language as it was
 816                 if (count > 0) {
 817                         // FIXME: probably a better way to keep track of the old
 818                         // language, than copying the entire font?
 819                         Font const copy_font(running_font);
 820                         basefont = owner_->getLayoutFont(bparams, outerfont);
 821                         running_font = basefont;
 822                         if (!closeLanguage)
 823                                 running_font.setLanguage(copy_font.language());
 824                         // leave font open if language is still open
 825                         open_font = (running_font.language() == basefont.language());
 826                         if (closeLanguage)
 827                                 runparams.local_font = &basefont;
 828                 }
 829         }
 830
 831         int tmp;
 832
 833         try {
 834                 tmp = inset->latex(os, runparams);
 835         } catch (EncodingException & e) {
 836                 // add location information and throw again.
 837                 e.par_id = id_;
 838                 e.pos = i;
 839                 throw(e);
 840         }
 841
 842         if (close) {
 843                 if (running_font.language()->lang() == "farsi")
 844                                 os << "\\endL{}";
 845                         else
 846                                 os << '}';
 847         }
 848
 849         if (tmp) {
 850                 texrow.newlines(tmp);
 851                 texrow.start(owner_->id(), i + 1);
 852                 column = 0;
 853         } else {
 854                 column += os.tellp() - len;
 855         }
 856
 857         if (owner_->isDeleted(i))
 858                 --runparams.inDeletedInset;
 859 }
 860
 861
 862 void Paragraph::Private::latexSpecialChar(
 863                                              odocstream & os,
 864                                              OutputParams const & runparams,
 865                                              Font const & running_font,
 866                                              Change const & running_change,
 867                                              Layout const & style,
 868                                              pos_type & i,
 869                                              unsigned int & column)
 870 {
 871         char_type const c = text_[i];
 872
 873         if (style.pass_thru) {
 874                 if (c != '\0')
 875                         // FIXME UNICODE: This can fail if c cannot
 876                         // be encoded in the current encoding.
 877                         os.put(c);
 878                 return;
 879         }
 880
 881         if (runparams.verbatim) {
 882                 // FIXME UNICODE: This can fail if c cannot
 883                 // be encoded in the current encoding.
 884                 os.put(c);
 885                 return;
 886         }
 887
 888         // If T1 font encoding is used, use the special
 889         // characters it provides.
 890         // NOTE: some languages reset the font encoding
 891         // internally
 892         if (!running_font.language()->internalFontEncoding()
 893             && lyxrc.fontenc == "T1" && latexSpecialT1(c, os, i, column))
 894                 return;
 895
 896         // \tt font needs special treatment
 897         if (running_font.fontInfo().family() == TYPEWRITER_FAMILY
 898                 && latexSpecialTypewriter(c, os, i, column))
 899                 return;
 900
 901         // Otherwise, we use what LaTeX provides us.
 902         switch (c) {
 903         case '\\':
 904                 os << "\\textbackslash{}";
 905                 column += 15;
 906                 break;
 907         case '<':
 908                 os << "\\textless{}";
 909                 column += 10;
 910                 break;
 911         case '>':
 912                 os << "\\textgreater{}";
 913                 column += 13;
 914                 break;
 915         case '|':
 916                 os << "\\textbar{}";
 917                 column += 9;
 918                 break;
 919         case '-':
 920                 os << '-';
 921                 break;
 922         case '\"':
 923                 os << "\\char`\\\"{}";
 924                 column += 9;
 925                 break;
 926
 927         case '$': case '&':
 928         case '%': case '#': case '{':
 929         case '}': case '_':
 930                 os << '\\';
 931                 os.put(c);
 932                 column += 1;
 933                 break;
 934
 935         case '~':
 936                 os << "\\textasciitilde{}";
 937                 column += 16;
 938                 break;
 939
 940         case '^':
 941                 os << "\\textasciicircum{}";
 942                 column += 17;
 943                 break;
 944
 945         case '*':
 946         case '[':
 947         case ']':
 948                 // avoid being mistaken for optional arguments
 949                 os << '{';
 950                 os.put(c);
 951                 os << '}';
 952                 column += 2;
 953                 break;
 954
 955         case ' ':
 956                 // Blanks are printed before font switching.
 957                 // Sure? I am not! (try nice-latex)
 958                 // I am sure it's correct. LyX might be smarter
 959                 // in the future, but for now, nothing wrong is
 960                 // written. (Asger)
 961                 break;
 962
 963         default:
 964                 // LyX, LaTeX etc.
 965                 if (latexSpecialPhrase(os, i, column, runparams))
 966                         return;
 967
 968                 if (c == '\0')
 969                         return;
 970
 971                 Encoding const & encoding = *(runparams.encoding);
 972                 if (i + 1 < int(text_.size())) {
 973                         char_type next = text_[i + 1];
 974                         if (Encodings::isCombiningChar(next)) {
 975                                 column += latexSurrogatePair(os, c, next, runparams) - 1;
 976                                 ++i;
 977                                 break;
 978                         }
 979                 }
 980                 string script;
 981                 docstring const latex = encoding.latexChar(c);
 982                 if (Encodings::isKnownScriptChar(c, script)
 983                     && prefixIs(latex, from_ascii("\\" + script)))
 984                         column += writeScriptChars(os, latex,
 985                                         running_change, encoding, i) - 1;
 986                 else if (latex.length() > 1 && latex[latex.length() - 1] != '}') {
 987                         // Prevent eating of a following
 988                         // space or command corruption by
 989                         // following characters
 990                         column += latex.length() + 1;
 991                         os << latex << "{}";
 992                 } else {
 993                         column += latex.length() - 1;
 994                         os << latex;
 995                 }
 996                 break;
 997         }
 998 }
 999
1000
1001 bool Paragraph::Private::latexSpecialT1(char_type const c, odocstream & os,
1002         pos_type i, unsigned int & column)
1003 {
1004         switch (c) {
1005         case '>':
1006         case '<':
1007                 os.put(c);
1008                 // In T1 encoding, these characters exist
1009                 // but we should avoid ligatures
1010                 if (i + 1 >= int(text_.size()) || text_[i + 1] != c)
1011                         return true;
1012                 os << "\\textcompwordmark{}";
1013                 column += 19;
1014                 return true;
1015         case '|':
1016                 os.put(c);
1017                 return true;
1018         case '\"':
1019                 // soul.sty breaks with \char`\"
1020                 os << "\\textquotedbl{}";
1021                 column += 14;
1022                 return true;
1023         default:
1024                 return false;
1025         }
1026 }
1027
1028
1029 bool Paragraph::Private::latexSpecialTypewriter(char_type const c, odocstream & os,
1030         pos_type i, unsigned int & column)
1031 {
1032         switch (c) {
1033         case '-':
1034                 // within \ttfamily, "--" is merged to "-" (no endash)
1035                 // so we avoid this rather irritating ligature
1036                 if (i + 1 < int(text_.size()) && text_[i + 1] == '-') {
1037                         os << "-{}";
1038                         column += 2;
1039                 } else
1040                         os << '-';
1041                 return true;
1042
1043         // everything else has to be checked separately
1044         // (depending on the encoding)
1045         default:
1046                 return false;
1047         }
1048 }
1049
1050
1051 bool Paragraph::Private::latexSpecialPhrase(odocstream & os, pos_type & i,
1052         unsigned int & column, OutputParams const & runparams)
1053 {
1054         // FIXME: if we have "LaTeX" with a font
1055         // change in the middle (before the 'T', then
1056         // the "TeX" part is still special cased.
1057         // Really we should only operate this on
1058         // "words" for some definition of word
1059
1060         for (size_t pnr = 0; pnr < phrases_nr; ++pnr) {
1061                 if (!isTextAt(special_phrases[pnr].phrase, i))
1062                         continue;
1063                 if (runparams.moving_arg)
1064                         os << "\\protect";
1065                 os << special_phrases[pnr].macro;
1066                 i += special_phrases[pnr].phrase.length() - 1;
1067                 column += special_phrases[pnr].macro.length() - 1;
1068                 return true;
1069         }
1070         return false;
1071 }
1072
1073
1074 void Paragraph::Private::validate(LaTeXFeatures & features) const
1075 {
1076         if (layout_->inpreamble && inset_owner_) {
1077                 Buffer const & buf = inset_owner_->buffer();
1078                 BufferParams const & bp = buf.params();
1079                 Font f;
1080                 TexRow tr;
1081                 odocstringstream ods;
1082                 // we have to provide all the optional arguments here, even though
1083                 // the last one is the only one we care about.
1084                 owner_->latex(bp, f, ods, tr, features.runparams(), 0, -1, true);
1085                 docstring const d = ods.str();
1086                 if (!d.empty()) {
1087                         // this will have "{" at the beginning, but not at the end
1088                         string const content = to_utf8(d);
1089                         string const cmd = layout_->latexname();
1090                         features.addPreambleSnippet("\\" + cmd + content + "}");
1091                 }
1092         }
1093
1094         if (features.runparams().flavor == OutputParams::HTML
1095             && layout_->htmltitle()) {
1096                 features.setHTMLTitle(owner_->asString(AS_STR_INSETS));
1097         }
1098
1099         // check the params.
1100         if (!params_.spacing().isDefault())
1101                 features.require("setspace");
1102
1103         // then the layouts
1104         features.useLayout(layout_->name());
1105
1106         // then the fonts
1107         fontlist_.validate(features);
1108
1109         // then the indentation
1110         if (!params_.leftIndent().zero())
1111                 features.require("ParagraphLeftIndent");
1112
1113         // then the insets
1114         InsetList::const_iterator icit = insetlist_.begin();
1115         InsetList::const_iterator iend = insetlist_.end();
1116         for (; icit != iend; ++icit) {
1117                 if (icit->inset) {
1118                         icit->inset->validate(features);
1119                         if (layout_->needprotect &&
1120                             icit->inset->lyxCode() == FOOT_CODE)
1121                                 features.require("NeedLyXFootnoteCode");
1122                 }
1123         }
1124
1125         // then the contents
1126         for (pos_type i = 0; i < int(text_.size()) ; ++i) {
1127                 for (size_t pnr = 0; pnr < phrases_nr; ++pnr) {
1128                         if (!special_phrases[pnr].builtin
1129                             && isTextAt(special_phrases[pnr].phrase, i)) {
1130                                 features.require(special_phrases[pnr].phrase);
1131                                 break;
1132                         }
1133                 }
1134                 Encodings::validate(text_[i], features);
1135         }
1136 }
1137
1138 /////////////////////////////////////////////////////////////////////
1139 //
1140 // Paragraph
1141 //
1142 /////////////////////////////////////////////////////////////////////
1143
1144 namespace {
1145         Layout const emptyParagraphLayout;
1146 }
1147
1148 Paragraph::Paragraph()
1149         : d(new Paragraph::Private(this, emptyParagraphLayout))
1150 {
1151         itemdepth = 0;
1152         d->params_.clear();
1153 }
1154
1155
1156 Paragraph::Paragraph(Paragraph const & par)
1157         : itemdepth(par.itemdepth),
1158         d(new Paragraph::Private(*par.d, this))
1159 {
1160         registerWords();
1161 }
1162
1163
1164 Paragraph::Paragraph(Paragraph const & par, pos_type beg, pos_type end)
1165         : itemdepth(par.itemdepth),
1166         d(new Paragraph::Private(*par.d, this, beg, end))
1167 {
1168         registerWords();
1169 }
1170
1171
1172 Paragraph & Paragraph::operator=(Paragraph const & par)
1173 {
1174         // needed as we will destroy the private part before copying it
1175         if (&par != this) {
1176                 itemdepth = par.itemdepth;
1177
1178                 deregisterWords();
1179                 delete d;
1180                 d = new Private(*par.d, this);
1181                 registerWords();
1182         }
1183         return *this;
1184 }
1185
1186
1187 Paragraph::~Paragraph()
1188 {
1189         deregisterWords();
1190         delete d;
1191 }
1192
1193
1194 namespace {
1195
1196 // this shall be called just before every "os << ..." action.
1197 void flushString(ostream & os, docstring & s)
1198 {
1199         os << to_utf8(s);
1200         s.erase();
1201 }
1202
1203 }
1204
1205
1206 void Paragraph::write(ostream & os, BufferParams const & bparams,
1207         depth_type & dth) const
1208 {
1209         // The beginning or end of a deeper (i.e. nested) area?
1210         if (dth != d->params_.depth()) {
1211                 if (d->params_.depth() > dth) {
1212                         while (d->params_.depth() > dth) {
1213                                 os << "\n\\begin_deeper";
1214                                 ++dth;
1215                         }
1216                 } else {
1217                         while (d->params_.depth() < dth) {
1218                                 os << "\n\\end_deeper";
1219                                 --dth;
1220                         }
1221                 }
1222         }
1223
1224         // First write the layout
1225         os << "\n\\begin_layout " << to_utf8(d->layout_->name()) << '\n';
1226
1227         d->params_.write(os);
1228
1229         Font font1(inherit_font, bparams.language);
1230
1231         Change running_change = Change(Change::UNCHANGED);
1232
1233         // this string is used as a buffer to avoid repetitive calls
1234         // to to_utf8(), which turn out to be expensive (JMarc)
1235         docstring write_buffer;
1236
1237         int column = 0;
1238         for (pos_type i = 0; i <= size(); ++i) {
1239
1240                 Change const change = lookupChange(i);
1241                 if (change != running_change)
1242                         flushString(os, write_buffer);
1243                 Changes::lyxMarkChange(os, bparams, column, running_change, change);
1244                 running_change = change;
1245
1246                 if (i == size())
1247                         break;
1248
1249                 // Write font changes (ignore spelling markers)
1250                 Font font2 = getFontSettings(bparams, i);
1251                 font2.setMisspelled(false);
1252                 if (font2 != font1) {
1253                         flushString(os, write_buffer);
1254                         font2.lyxWriteChanges(font1, os);
1255                         column = 0;
1256                         font1 = font2;
1257                 }
1258
1259                 char_type const c = d->text_[i];
1260                 switch (c) {
1261                 case META_INSET:
1262                         if (Inset const * inset = getInset(i)) {
1263                                 flushString(os, write_buffer);
1264                                 if (inset->directWrite()) {
1265                                         // international char, let it write
1266                                         // code directly so it's shorter in
1267                                         // the file
1268                                         inset->write(os);
1269                                 } else {
1270                                         if (i)
1271                                                 os << '\n';
1272                                         os << "\\begin_inset ";
1273                                         inset->write(os);
1274                                         os << "\n\\end_inset\n\n";
1275                                         column = 0;
1276                                 }
1277                         }
1278                         break;
1279                 case '\\':
1280                         flushString(os, write_buffer);
1281                         os << "\n\\backslash\n";
1282                         column = 0;
1283                         break;
1284                 case '.':
1285                         flushString(os, write_buffer);
1286                         if (i + 1 < size() && d->text_[i + 1] == ' ') {
1287                                 os << ".\n";
1288                                 column = 0;
1289                         } else
1290                                 os << '.';
1291                         break;
1292                 default:
1293                         if ((column > 70 && c == ' ')
1294                             || column > 79) {
1295                                 flushString(os, write_buffer);
1296                                 os << '\n';
1297                                 column = 0;
1298                         }
1299                         // this check is to amend a bug. LyX sometimes
1300                         // inserts '\0' this could cause problems.
1301                         if (c != '\0')
1302                                 write_buffer.push_back(c);
1303                         else
1304                                 LYXERR0("NUL char in structure.");
1305                         ++column;
1306                         break;
1307                 }
1308         }
1309
1310         flushString(os, write_buffer);
1311         os << "\n\\end_layout\n";
1312 }
1313
1314
1315 void Paragraph::validate(LaTeXFeatures & features) const
1316 {
1317         d->validate(features);
1318 }
1319
1320
1321 void Paragraph::insert(pos_type start, docstring const & str,
1322                        Font const & font, Change const & change)
1323 {
1324         for (size_t i = 0, n = str.size(); i != n ; ++i)
1325                 insertChar(start + i, str[i], font, change);
1326 }
1327
1328
1329 void Paragraph::appendChar(char_type c, Font const & font,
1330                 Change const & change)
1331 {
1332         // track change
1333         d->changes_.insert(change, d->text_.size());
1334         // when appending characters, no need to update tables
1335         d->text_.push_back(c);
1336         setFont(d->text_.size() - 1, font);
1337 }
1338
1339
1340 void Paragraph::appendString(docstring const & s, Font const & font,
1341                 Change const & change)
1342 {
1343         pos_type end = s.size();
1344         size_t oldsize = d->text_.size();
1345         size_t newsize = oldsize + end;
1346         size_t capacity = d->text_.capacity();
1347         if (newsize >= capacity)
1348                 d->text_.reserve(max(capacity + 100, newsize));
1349
1350         // when appending characters, no need to update tables
1351         d->text_.append(s);
1352
1353         // FIXME: Optimize this!
1354         for (size_t i = oldsize; i != newsize; ++i) {
1355                 // track change
1356                 d->changes_.insert(change, i);
1357         }
1358         d->fontlist_.set(oldsize, font);
1359         d->fontlist_.set(newsize - 1, font);
1360 }
1361
1362
1363 void Paragraph::insertChar(pos_type pos, char_type c,
1364                            bool trackChanges)
1365 {
1366         d->insertChar(pos, c, Change(trackChanges ?
1367                            Change::INSERTED : Change::UNCHANGED));
1368 }
1369
1370
1371 void Paragraph::insertChar(pos_type pos, char_type c,
1372                            Font const & font, bool trackChanges)
1373 {
1374         d->insertChar(pos, c, Change(trackChanges ?
1375                            Change::INSERTED : Change::UNCHANGED));
1376         setFont(pos, font);
1377 }
1378
1379
1380 void Paragraph::insertChar(pos_type pos, char_type c,
1381                            Font const & font, Change const & change)
1382 {
1383         d->insertChar(pos, c, change);
1384         setFont(pos, font);
1385 }
1386
1387
1388 bool Paragraph::insertInset(pos_type pos, Inset * inset,
1389                             Font const & font, Change const & change)
1390 {
1391         bool const success = insertInset(pos, inset, change);
1392         // Set the font/language of the inset...
1393         setFont(pos, font);
1394         return success;
1395 }
1396
1397
1398 void Paragraph::resetFonts(Font const & font)
1399 {
1400         d->fontlist_.clear();
1401         d->fontlist_.set(0, font);
1402         d->fontlist_.set(d->text_.size() - 1, font);
1403 }
1404
1405 // Gets uninstantiated font setting at position.
1406 Font const & Paragraph::getFontSettings(BufferParams const & bparams,
1407                                          pos_type pos) const
1408 {
1409         if (pos > size()) {
1410                 LYXERR0("pos: " << pos << " size: " << size());
1411                 LASSERT(pos <= size(), /**/);
1412         }
1413
1414         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
1415         if (cit != d->fontlist_.end())
1416                 return cit->font();
1417
1418         if (pos == size() && !empty())
1419                 return getFontSettings(bparams, pos - 1);
1420
1421         // Optimisation: avoid a full font instantiation if there is no
1422         // language change from previous call.
1423         static Font previous_font;
1424         static Language const * previous_lang = 0;
1425         Language const * lang = getParLanguage(bparams);
1426         if (lang != previous_lang) {
1427                 previous_lang = lang;
1428                 previous_font = Font(inherit_font, lang);
1429         }
1430         return previous_font;
1431 }
1432
1433
1434 FontSpan Paragraph::fontSpan(pos_type pos) const
1435 {
1436         LASSERT(pos <= size(), /**/);
1437         pos_type start = 0;
1438
1439         FontList::const_iterator cit = d->fontlist_.begin();
1440         FontList::const_iterator end = d->fontlist_.end();
1441         for (; cit != end; ++cit) {
1442                 if (cit->pos() >= pos) {
1443                         if (pos >= beginOfBody())
1444                                 return FontSpan(max(start, beginOfBody()),
1445                                                 cit->pos());
1446                         else
1447                                 return FontSpan(start,
1448                                                 min(beginOfBody() - 1,
1449                                                          cit->pos()));
1450                 }
1451                 start = cit->pos() + 1;
1452         }
1453
1454         // This should not happen, but if so, we take no chances.
1455         // LYXERR0("Paragraph::getEndPosOfFontSpan: This should not happen!");
1456         return FontSpan(pos, pos);
1457 }
1458
1459
1460 // Gets uninstantiated font setting at position 0
1461 Font const & Paragraph::getFirstFontSettings(BufferParams const & bparams) const
1462 {
1463         if (!empty() && !d->fontlist_.empty())
1464                 return d->fontlist_.begin()->font();
1465
1466         // Optimisation: avoid a full font instantiation if there is no
1467         // language change from previous call.
1468         static Font previous_font;
1469         static Language const * previous_lang = 0;
1470         if (bparams.language != previous_lang) {
1471                 previous_lang = bparams.language;
1472                 previous_font = Font(inherit_font, bparams.language);
1473         }
1474
1475         return previous_font;
1476 }
1477
1478
1479 // Gets the fully instantiated font at a given position in a paragraph
1480 // This is basically the same function as Text::GetFont() in text2.cpp.
1481 // The difference is that this one is used for generating the LaTeX file,
1482 // and thus cosmetic "improvements" are disallowed: This has to deliver
1483 // the true picture of the buffer. (Asger)
1484 Font const Paragraph::getFont(BufferParams const & bparams, pos_type pos,
1485                                  Font const & outerfont) const
1486 {
1487         LASSERT(pos >= 0, /**/);
1488
1489         Font font = getFontSettings(bparams, pos);
1490
1491         pos_type const body_pos = beginOfBody();
1492         FontInfo & fi = font.fontInfo();
1493         if (pos < body_pos)
1494                 fi.realize(d->layout_->labelfont);
1495         else
1496                 fi.realize(d->layout_->font);
1497
1498         fi.realize(outerfont.fontInfo());
1499         fi.realize(bparams.getFont().fontInfo());
1500
1501         return font;
1502 }
1503
1504
1505 Font const Paragraph::getLabelFont
1506         (BufferParams const & bparams, Font const & outerfont) const
1507 {
1508         FontInfo tmpfont = d->layout_->labelfont;
1509         tmpfont.realize(outerfont.fontInfo());
1510         tmpfont.realize(bparams.getFont().fontInfo());
1511         return Font(tmpfont, getParLanguage(bparams));
1512 }
1513
1514
1515 Font const Paragraph::getLayoutFont
1516         (BufferParams const & bparams, Font const & outerfont) const
1517 {
1518         FontInfo tmpfont = d->layout_->font;
1519         tmpfont.realize(outerfont.fontInfo());
1520         tmpfont.realize(bparams.getFont().fontInfo());
1521         return Font(tmpfont, getParLanguage(bparams));
1522 }
1523
1524
1525 /// Returns the height of the highest font in range
1526 FontSize Paragraph::highestFontInRange
1527         (pos_type startpos, pos_type endpos, FontSize def_size) const
1528 {
1529         return d->fontlist_.highestInRange(startpos, endpos, def_size);
1530 }
1531
1532
1533 char_type Paragraph::getUChar(BufferParams const & bparams, pos_type pos) const
1534 {
1535         char_type c = d->text_[pos];
1536         if (!lyxrc.rtl_support)
1537                 return c;
1538
1539         char_type uc = c;
1540         switch (c) {
1541         case '(':
1542                 uc = ')';
1543                 break;
1544         case ')':
1545                 uc = '(';
1546                 break;
1547         case '[':
1548                 uc = ']';
1549                 break;
1550         case ']':
1551                 uc = '[';
1552                 break;
1553         case '{':
1554                 uc = '}';
1555                 break;
1556         case '}':
1557                 uc = '{';
1558                 break;
1559         case '<':
1560                 uc = '>';
1561                 break;
1562         case '>':
1563                 uc = '<';
1564                 break;
1565         }
1566         if (uc != c && getFontSettings(bparams, pos).isRightToLeft())
1567                 return uc;
1568         return c;
1569 }
1570
1571
1572 void Paragraph::setFont(pos_type pos, Font const & font)
1573 {
1574         LASSERT(pos <= size(), /**/);
1575
1576         // First, reduce font against layout/label font
1577         // Update: The setCharFont() routine in text2.cpp already
1578         // reduces font, so we don't need to do that here. (Asger)
1579
1580         d->fontlist_.set(pos, font);
1581 }
1582
1583
1584 void Paragraph::makeSameLayout(Paragraph const & par)
1585 {
1586         d->layout_ = par.d->layout_;
1587         d->params_ = par.d->params_;
1588 }
1589
1590
1591 bool Paragraph::stripLeadingSpaces(bool trackChanges)
1592 {
1593         if (isFreeSpacing())
1594                 return false;
1595
1596         int pos = 0;
1597         int count = 0;
1598
1599         while (pos < size() && (isNewline(pos) || isLineSeparator(pos))) {
1600                 if (eraseChar(pos, trackChanges))
1601                         ++count;
1602                 else
1603                         ++pos;
1604         }
1605
1606         return count > 0 || pos > 0;
1607 }
1608
1609
1610 bool Paragraph::hasSameLayout(Paragraph const & par) const
1611 {
1612         return par.d->layout_ == d->layout_
1613                 && d->params_.sameLayout(par.d->params_);
1614 }
1615
1616
1617 depth_type Paragraph::getDepth() const
1618 {
1619         return d->params_.depth();
1620 }
1621
1622
1623 depth_type Paragraph::getMaxDepthAfter() const
1624 {
1625         if (d->layout_->isEnvironment())
1626                 return d->params_.depth() + 1;
1627         else
1628                 return d->params_.depth();
1629 }
1630
1631
1632 char Paragraph::getAlign() const
1633 {
1634         if (d->params_.align() == LYX_ALIGN_LAYOUT)
1635                 return d->layout_->align;
1636         else
1637                 return d->params_.align();
1638 }
1639
1640
1641 docstring const & Paragraph::labelString() const
1642 {
1643         return d->params_.labelString();
1644 }
1645
1646
1647 // the next two functions are for the manual labels
1648 docstring const Paragraph::getLabelWidthString() const
1649 {
1650         if (d->layout_->margintype == MARGIN_MANUAL
1651             || d->layout_->latextype == LATEX_BIB_ENVIRONMENT)
1652                 return d->params_.labelWidthString();
1653         else
1654                 return _("Senseless with this layout!");
1655 }
1656
1657
1658 void Paragraph::setLabelWidthString(docstring const & s)
1659 {
1660         d->params_.labelWidthString(s);
1661 }
1662
1663
1664 docstring Paragraph::expandLabel(Layout const & layout,
1665                 BufferParams const & bparams) const
1666 {
1667         return expandParagraphLabel(layout, bparams, true);
1668 }
1669
1670
1671 docstring Paragraph::expandDocBookLabel(Layout const & layout,
1672                 BufferParams const & bparams) const
1673 {
1674         return expandParagraphLabel(layout, bparams, false);
1675 }
1676
1677
1678 docstring Paragraph::expandParagraphLabel(Layout const & layout,
1679                 BufferParams const & bparams, bool process_appendix) const
1680 {
1681         DocumentClass const & tclass = bparams.documentClass();
1682         string const & lang = getParLanguage(bparams)->code();
1683         bool const in_appendix = process_appendix && d->params_.appendix();
1684         docstring fmt = translateIfPossible(layout.labelstring(in_appendix), lang);
1685
1686         if (fmt.empty() && layout.labeltype == LABEL_COUNTER
1687             && !layout.counter.empty())
1688                 return tclass.counters().theCounter(layout.counter, lang);
1689
1690         // handle 'inherited level parts' in 'fmt',
1691         // i.e. the stuff between '@' in   '@Section@.\arabic{subsection}'
1692         size_t const i = fmt.find('@', 0);
1693         if (i != docstring::npos) {
1694                 size_t const j = fmt.find('@', i + 1);
1695                 if (j != docstring::npos) {
1696                         docstring parent(fmt, i + 1, j - i - 1);
1697                         docstring label = from_ascii("??");
1698                         if (tclass.hasLayout(parent))
1699                                 docstring label = expandParagraphLabel(tclass[parent], bparams,
1700                                                       process_appendix);
1701                         fmt = docstring(fmt, 0, i) + label
1702                                 + docstring(fmt, j + 1, docstring::npos);
1703                 }
1704         }
1705
1706         return tclass.counters().counterLabel(fmt, lang);
1707 }
1708
1709
1710 void Paragraph::applyLayout(Layout const & new_layout)
1711 {
1712         d->layout_ = &new_layout;
1713         LyXAlignment const oldAlign = d->params_.align();
1714
1715         if (!(oldAlign & d->layout_->alignpossible)) {
1716                 frontend::Alert::warning(_("Alignment not permitted"),
1717                         _("The new layout does not permit the alignment previously used.\nSetting to default."));
1718                 d->params_.align(LYX_ALIGN_LAYOUT);
1719         }
1720 }
1721
1722
1723 pos_type Paragraph::beginOfBody() const
1724 {
1725         return d->begin_of_body_;
1726 }
1727
1728
1729 void Paragraph::setBeginOfBody()
1730 {
1731         if (d->layout_->labeltype != LABEL_MANUAL) {
1732                 d->begin_of_body_ = 0;
1733                 return;
1734         }
1735
1736         // Unroll the first two cycles of the loop
1737         // and remember the previous character to
1738         // remove unnecessary getChar() calls
1739         pos_type i = 0;
1740         pos_type end = size();
1741         if (i < end && !isNewline(i)) {
1742                 ++i;
1743                 char_type previous_char = 0;
1744                 char_type temp = 0;
1745                 if (i < end) {
1746                         previous_char = d->text_[i];
1747                         if (!isNewline(i)) {
1748                                 ++i;
1749                                 while (i < end && previous_char != ' ') {
1750                                         temp = d->text_[i];
1751                                         if (isNewline(i))
1752                                                 break;
1753                                         ++i;
1754                                         previous_char = temp;
1755                                 }
1756                         }
1757                 }
1758         }
1759
1760         d->begin_of_body_ = i;
1761 }
1762
1763
1764 bool Paragraph::allowParagraphCustomization() const
1765 {
1766         return inInset().allowParagraphCustomization();
1767 }
1768
1769
1770 bool Paragraph::usePlainLayout() const
1771 {
1772         return inInset().usePlainLayout();
1773 }
1774
1775
1776 namespace {
1777
1778 // paragraphs inside floats need different alignment tags to avoid
1779 // unwanted space
1780
1781 bool noTrivlistCentering(InsetCode code)
1782 {
1783         return code == FLOAT_CODE
1784                || code == WRAP_CODE
1785                || code == CELL_CODE;
1786 }
1787
1788
1789 string correction(string const & orig)
1790 {
1791         if (orig == "flushleft")
1792                 return "raggedright";
1793         if (orig == "flushright")
1794                 return "raggedleft";
1795         if (orig == "center")
1796                 return "centering";
1797         return orig;
1798 }
1799
1800
1801 string const corrected_env(string const & suffix, string const & env,
1802         InsetCode code, bool const lastpar)
1803 {
1804         string output = suffix + "{";
1805         if (noTrivlistCentering(code)) {
1806                 if (lastpar) {
1807                         // the last paragraph in non-trivlist-aligned
1808                         // context is special (to avoid unwanted whitespace)
1809                         if (suffix == "\\begin")
1810                                 return "\\" + correction(env) + "{}";
1811                         return string();
1812                 }
1813                 output += correction(env);
1814         } else
1815                 output += env;
1816         output += "}";
1817         if (suffix == "\\begin")
1818                 output += "\n";
1819         return output;
1820 }
1821
1822
1823 void adjust_row_column(string const & str, TexRow & texrow, int & column)
1824 {
1825         if (!contains(str, "\n"))
1826                 column += str.size();
1827         else {
1828                 string tmp;
1829                 texrow.newline();
1830                 column = rsplit(str, tmp, '\n').size();
1831         }
1832 }
1833
1834 } // namespace anon
1835
1836
1837 int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
1838                                  odocstream & os, TexRow & texrow,
1839                                  OutputParams const & runparams) const
1840 {
1841         int column = 0;
1842
1843         if (params_.noindent()) {
1844                 os << "\\noindent ";
1845                 column += 10;
1846         }
1847
1848         LyXAlignment const curAlign = params_.align();
1849
1850         if (curAlign == layout_->align)
1851                 return column;
1852
1853         switch (curAlign) {
1854         case LYX_ALIGN_NONE:
1855         case LYX_ALIGN_BLOCK:
1856         case LYX_ALIGN_LAYOUT:
1857         case LYX_ALIGN_SPECIAL:
1858         case LYX_ALIGN_DECIMAL:
1859                 break;
1860         case LYX_ALIGN_LEFT:
1861         case LYX_ALIGN_RIGHT:
1862         case LYX_ALIGN_CENTER:
1863                 if (runparams.moving_arg) {
1864                         os << "\\protect";
1865                         column += 8;
1866                 }
1867                 break;
1868         }
1869
1870         string const begin_tag = "\\begin";
1871         InsetCode code = ownerCode();
1872         bool const lastpar = runparams.isLastPar;
1873
1874         switch (curAlign) {
1875         case LYX_ALIGN_NONE:
1876         case LYX_ALIGN_BLOCK:
1877         case LYX_ALIGN_LAYOUT:
1878         case LYX_ALIGN_SPECIAL:
1879         case LYX_ALIGN_DECIMAL:
1880                 break;
1881         case LYX_ALIGN_LEFT: {
1882                 string output;
1883                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
1884                         output = corrected_env(begin_tag, "flushleft", code, lastpar);
1885                 else
1886                         output = corrected_env(begin_tag, "flushright", code, lastpar);
1887                 os << from_ascii(output);
1888                 adjust_row_column(output, texrow, column);
1889                 break;
1890         } case LYX_ALIGN_RIGHT: {
1891                 string output;
1892                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
1893                         output = corrected_env(begin_tag, "flushright", code, lastpar);
1894                 else
1895                         output = corrected_env(begin_tag, "flushleft", code, lastpar);
1896                 os << from_ascii(output);
1897                 adjust_row_column(output, texrow, column);
1898                 break;
1899         } case LYX_ALIGN_CENTER: {
1900                 string output;
1901                 output = corrected_env(begin_tag, "center", code, lastpar);
1902                 os << from_ascii(output);
1903                 adjust_row_column(output, texrow, column);
1904                 break;
1905         }
1906         }
1907
1908         return column;
1909 }
1910
1911
1912 int Paragraph::Private::endTeXParParams(BufferParams const & bparams,
1913                                odocstream & os, TexRow & texrow,
1914                                OutputParams const & runparams) const
1915 {
1916         int column = 0;
1917
1918         LyXAlignment const curAlign = params_.align();
1919
1920         if (curAlign == layout_->align)
1921                 return column;
1922
1923         switch (curAlign) {
1924         case LYX_ALIGN_NONE:
1925         case LYX_ALIGN_BLOCK:
1926         case LYX_ALIGN_LAYOUT:
1927         case LYX_ALIGN_SPECIAL:
1928         case LYX_ALIGN_DECIMAL:
1929                 break;
1930         case LYX_ALIGN_LEFT:
1931         case LYX_ALIGN_RIGHT:
1932         case LYX_ALIGN_CENTER:
1933                 if (runparams.moving_arg) {
1934                         os << "\\protect";
1935                         column = 8;
1936                 }
1937                 break;
1938         }
1939
1940         string const end_tag = "\n\\par\\end";
1941         InsetCode code = ownerCode();
1942         bool const lastpar = runparams.isLastPar;
1943
1944         switch (curAlign) {
1945         case LYX_ALIGN_NONE:
1946         case LYX_ALIGN_BLOCK:
1947         case LYX_ALIGN_LAYOUT:
1948         case LYX_ALIGN_SPECIAL:
1949         case LYX_ALIGN_DECIMAL:
1950                 break;
1951         case LYX_ALIGN_LEFT: {
1952                 string output;
1953                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
1954                         output = corrected_env(end_tag, "flushleft", code, lastpar);
1955                 else
1956                         output = corrected_env(end_tag, "flushright", code, lastpar);
1957                 os << from_ascii(output);
1958                 adjust_row_column(output, texrow, column);
1959                 break;
1960         } case LYX_ALIGN_RIGHT: {
1961                 string output;
1962                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
1963                         output = corrected_env(end_tag, "flushright", code, lastpar);
1964                 else
1965                         output = corrected_env(end_tag, "flushleft", code, lastpar);
1966                 os << from_ascii(output);
1967                 adjust_row_column(output, texrow, column);
1968                 break;
1969         } case LYX_ALIGN_CENTER: {
1970                 string output;
1971                 output = corrected_env(end_tag, "center", code, lastpar);
1972                 os << from_ascii(output);
1973                 adjust_row_column(output, texrow, column);
1974                 break;
1975         }
1976         }
1977
1978         return column;
1979 }
1980
1981
1982 // This one spits out the text of the paragraph
1983 void Paragraph::latex(BufferParams const & bparams,
1984         Font const & outerfont,
1985         odocstream & os, TexRow & texrow,
1986         OutputParams const & runparams,
1987         int start_pos, int end_pos, bool force) const
1988 {
1989         LYXERR(Debug::LATEX, "Paragraph::latex...     " << this);
1990
1991         // FIXME This check should not be needed. Perhaps issue an
1992         // error if it triggers.
1993         Layout const & style = inInset().forcePlainLayout() ?
1994                 bparams.documentClass().plainLayout() : *d->layout_;
1995
1996         if (!force && style.inpreamble)
1997                 return;
1998
1999         bool const allowcust = allowParagraphCustomization();
2000
2001         // Current base font for all inherited font changes, without any
2002         // change caused by an individual character, except for the language:
2003         // It is set to the language of the first character.
2004         // As long as we are in the label, this font is the base font of the
2005         // label. Before the first body character it is set to the base font
2006         // of the body.
2007         Font basefont;
2008
2009         // Maybe we have to create a optional argument.
2010         pos_type body_pos = beginOfBody();
2011         unsigned int column = 0;
2012
2013         if (body_pos > 0) {
2014                 // the optional argument is kept in curly brackets in
2015                 // case it contains a ']'
2016                 os << "[{";
2017                 column += 2;
2018                 basefont = getLabelFont(bparams, outerfont);
2019         } else {
2020                 basefont = getLayoutFont(bparams, outerfont);
2021         }
2022
2023         // Which font is currently active?
2024         Font running_font(basefont);
2025         // Do we have an open font change?
2026         bool open_font = false;
2027
2028         Change runningChange = Change(Change::UNCHANGED);
2029
2030         Encoding const * const prev_encoding = runparams.encoding;
2031
2032         texrow.start(id(), 0);
2033
2034         // if the paragraph is empty, the loop will not be entered at all
2035         if (empty()) {
2036                 if (style.isCommand()) {
2037                         os << '{';
2038                         ++column;
2039                 }
2040                 if (allowcust)
2041                         column += d->startTeXParParams(bparams, os, texrow,
2042                                                     runparams);
2043         }
2044
2045         for (pos_type i = 0; i < size(); ++i) {
2046                 // First char in paragraph or after label?
2047                 if (i == body_pos) {
2048                         if (body_pos > 0) {
2049                                 if (open_font) {
2050                                         column += running_font.latexWriteEndChanges(
2051                                                 os, bparams, runparams,
2052                                                 basefont, basefont);
2053                                         open_font = false;
2054                                 }
2055                                 basefont = getLayoutFont(bparams, outerfont);
2056                                 running_font = basefont;
2057
2058                                 column += Changes::latexMarkChange(os, bparams,
2059                                                 runningChange, Change(Change::UNCHANGED),
2060                                                 runparams);
2061                                 runningChange = Change(Change::UNCHANGED);
2062
2063                                 os << "}] ";
2064                                 column +=3;
2065                         }
2066                         if (style.isCommand()) {
2067                                 os << '{';
2068                                 ++column;
2069                         }
2070
2071                         if (allowcust)
2072                                 column += d->startTeXParParams(bparams, os,
2073                                                             texrow,
2074                                                             runparams);
2075                 }
2076
2077                 Change const & change = runparams.inDeletedInset ? runparams.changeOfDeletedInset
2078                                                                  : lookupChange(i);
2079
2080                 if (bparams.outputChanges && runningChange != change) {
2081                         if (open_font) {
2082                                 column += running_font.latexWriteEndChanges(
2083                                                 os, bparams, runparams, basefont, basefont);
2084                                 open_font = false;
2085                         }
2086                         basefont = getLayoutFont(bparams, outerfont);
2087                         running_font = basefont;
2088
2089                         column += Changes::latexMarkChange(os, bparams, runningChange,
2090                                                            change, runparams);
2091                         runningChange = change;
2092                 }
2093
2094                 // do not output text which is marked deleted
2095                 // if change tracking output is disabled
2096                 if (!bparams.outputChanges && change.deleted()) {
2097                         continue;
2098                 }
2099
2100                 ++column;
2101
2102                 // Fully instantiated font
2103                 Font const font = getFont(bparams, i, outerfont);
2104
2105                 Font const last_font = running_font;
2106
2107                 // Do we need to close the previous font?
2108                 if (open_font &&
2109                     (font != running_font ||
2110                      font.language() != running_font.language()))
2111                 {
2112                         column += running_font.latexWriteEndChanges(
2113                                         os, bparams, runparams, basefont,
2114                                         (i == body_pos-1) ? basefont : font);
2115                         running_font = basefont;
2116                         open_font = false;
2117                 }
2118
2119                 // close babel's font environment before opening CJK.
2120                 if (!running_font.language()->babel().empty() &&
2121                     font.language()->encoding()->package() == Encoding::CJK) {
2122                                 string end_tag = subst(lyxrc.language_command_end,
2123                                                         "$$lang",
2124                                                         running_font.language()->babel());
2125                                 os << from_ascii(end_tag);
2126                                 column += end_tag.length();
2127                 }
2128
2129                 // Switch file encoding if necessary (and allowed)
2130                 if (!runparams.verbatim &&
2131                     runparams.encoding->package() != Encoding::none &&
2132                     font.language()->encoding()->package() != Encoding::none) {
2133                         pair<bool, int> const enc_switch = switchEncoding(os, bparams,
2134                                         runparams, *(font.language()->encoding()));
2135                         if (enc_switch.first) {
2136                                 column += enc_switch.second;
2137                                 runparams.encoding = font.language()->encoding();
2138                         }
2139                 }
2140
2141                 char_type const c = d->text_[i];
2142
2143                 // Do we need to change font?
2144                 if ((font != running_font ||
2145                      font.language() != running_font.language()) &&
2146                         i != body_pos - 1)
2147                 {
2148                         odocstringstream ods;
2149                         column += font.latexWriteStartChanges(ods, bparams,
2150                                                               runparams, basefont,
2151                                                               last_font);
2152                         running_font = font;
2153                         open_font = true;
2154                         docstring fontchange = ods.str();
2155                         // check whether the fontchange ends with a \\textcolor
2156                         // modifier and the text starts with a space (bug 4473)
2157                         docstring const last_modifier = rsplit(fontchange, '\\');
2158                         if (prefixIs(last_modifier, from_ascii("textcolor")) && c == ' ')
2159                                 os << fontchange << from_ascii("{}");
2160                         // check if the fontchange ends with a trailing blank
2161                         // (like "\small " (see bug 3382)
2162                         else if (suffixIs(fontchange, ' ') && c == ' ')
2163                                 os << fontchange.substr(0, fontchange.size() - 1)
2164                                    << from_ascii("{}");
2165                         else
2166                                 os << fontchange;
2167                 }
2168
2169                 // FIXME: think about end_pos implementation...
2170                 if (c == ' ' && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2171                         // FIXME: integrate this case in latexSpecialChar
2172                         // Do not print the separation of the optional argument
2173                         // if style.pass_thru is false. This works because
2174                         // latexSpecialChar ignores spaces if
2175                         // style.pass_thru is false.
2176                         if (i != body_pos - 1) {
2177                                 if (d->simpleTeXBlanks(
2178                                                 runparams, os, texrow,
2179                                                 i, column, font, style)) {
2180                                         // A surrogate pair was output. We
2181                                         // must not call latexSpecialChar
2182                                         // in this iteration, since it would output
2183                                         // the combining character again.
2184                                         ++i;
2185                                         continue;
2186                                 }
2187                         }
2188                 }
2189
2190                 OutputParams rp = runparams;
2191                 rp.free_spacing = style.free_spacing;
2192                 rp.local_font = &font;
2193                 rp.intitle = style.intitle;
2194
2195                 // Two major modes:  LaTeX or plain
2196                 // Handle here those cases common to both modes
2197                 // and then split to handle the two modes separately.
2198                 if (c == META_INSET) {
2199                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2200                                 d->latexInset(bparams, os,
2201                                                 texrow, rp, running_font,
2202                                                 basefont, outerfont, open_font,
2203                                                 runningChange, style, i, column);
2204                         }
2205                 } else {
2206                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2207                                 try {
2208                                         d->latexSpecialChar(os, rp, running_font, runningChange,
2209                                                 style, i, column);
2210                                 } catch (EncodingException & e) {
2211                                 if (runparams.dryrun) {
2212                                         os << "<" << _("LyX Warning: ")
2213                                            << _("uncodable character") << " '";
2214                                         os.put(c);
2215                                         os << "'>";
2216                                 } else {
2217                                         // add location information and throw again.
2218                                         e.par_id = id();
2219                                         e.pos = i;
2220                                         throw(e);
2221                                 }
2222                         }
2223                 }
2224                 }
2225
2226                 // Set the encoding to that returned from latexSpecialChar (see
2227                 // comment for encoding member in OutputParams.h)
2228                 runparams.encoding = rp.encoding;
2229         }
2230
2231         // If we have an open font definition, we have to close it
2232         if (open_font) {
2233 #ifdef FIXED_LANGUAGE_END_DETECTION
2234                 if (next_) {
2235                         running_font.latexWriteEndChanges(os, bparams,
2236                                         runparams, basefont,
2237                                         next_->getFont(bparams, 0, outerfont));
2238                 } else {
2239                         running_font.latexWriteEndChanges(os, bparams,
2240                                         runparams, basefont, basefont);
2241                 }
2242 #else
2243 //FIXME: For now we ALWAYS have to close the foreign font settings if they are
2244 //FIXME: there as we start another \selectlanguage with the next paragraph if
2245 //FIXME: we are in need of this. This should be fixed sometime (Jug)
2246                 running_font.latexWriteEndChanges(os, bparams, runparams,
2247                                 basefont, basefont);
2248 #endif
2249         }
2250
2251         column += Changes::latexMarkChange(os, bparams, runningChange,
2252                                            Change(Change::UNCHANGED), runparams);
2253
2254         // Needed if there is an optional argument but no contents.
2255         if (body_pos > 0 && body_pos == size()) {
2256                 os << "}]~";
2257         }
2258
2259         if (allowcust && d->endTeXParParams(bparams, os, texrow, runparams)
2260             && runparams.encoding != prev_encoding) {
2261                 runparams.encoding = prev_encoding;
2262                 if (!bparams.useXetex)
2263                         os << setEncoding(prev_encoding->iconvName());
2264         }
2265
2266         LYXERR(Debug::LATEX, "Paragraph::latex... done " << this);
2267 }
2268
2269
2270 bool Paragraph::emptyTag() const
2271 {
2272         for (pos_type i = 0; i < size(); ++i) {
2273                 if (Inset const * inset = getInset(i)) {
2274                         InsetCode lyx_code = inset->lyxCode();
2275                         // FIXME testing like that is wrong. What is
2276                         // the intent?
2277                         if (lyx_code != TOC_CODE &&
2278                             lyx_code != INCLUDE_CODE &&
2279                             lyx_code != GRAPHICS_CODE &&
2280                             lyx_code != ERT_CODE &&
2281                             lyx_code != LISTINGS_CODE &&
2282                             lyx_code != FLOAT_CODE &&
2283                             lyx_code != TABULAR_CODE) {
2284                                 return false;
2285                         }
2286                 } else {
2287                         char_type c = d->text_[i];
2288                         if (c != ' ' && c != '\t')
2289                                 return false;
2290                 }
2291         }
2292         return true;
2293 }
2294
2295
2296 string Paragraph::getID(Buffer const & buf, OutputParams const & runparams)
2297         const
2298 {
2299         for (pos_type i = 0; i < size(); ++i) {
2300                 if (Inset const * inset = getInset(i)) {
2301                         InsetCode lyx_code = inset->lyxCode();
2302                         if (lyx_code == LABEL_CODE) {
2303                                 InsetLabel const * const il = static_cast<InsetLabel const *>(inset);
2304                                 docstring const & id = il->getParam("name");
2305                                 return "id='" + to_utf8(sgml::cleanID(buf, runparams, id)) + "'";
2306                         }
2307                 }
2308         }
2309         return string();
2310 }
2311
2312
2313 pos_type Paragraph::firstWordDocBook(odocstream & os, OutputParams const & runparams)
2314         const
2315 {
2316         pos_type i;
2317         for (i = 0; i < size(); ++i) {
2318                 if (Inset const * inset = getInset(i)) {
2319                         inset->docbook(os, runparams);
2320                 } else {
2321                         char_type c = d->text_[i];
2322                         if (c == ' ')
2323                                 break;
2324                         os << sgml::escapeChar(c);
2325                 }
2326         }
2327         return i;
2328 }
2329
2330
2331 pos_type Paragraph::firstWordLyXHTML(XHTMLStream & xs, OutputParams const & runparams)
2332         const
2333 {
2334         pos_type i;
2335         for (i = 0; i < size(); ++i) {
2336                 if (Inset const * inset = getInset(i)) {
2337                         inset->xhtml(xs, runparams);
2338                 } else {
2339                         char_type c = d->text_[i];
2340                         if (c == ' ')
2341                                 break;
2342                         xs << c;
2343                 }
2344         }
2345         return i;
2346 }
2347
2348
2349 bool Paragraph::Private::onlyText(Buffer const & buf, Font const & outerfont, pos_type initial) const
2350 {
2351         Font font_old;
2352         pos_type size = text_.size();
2353         for (pos_type i = initial; i < size; ++i) {
2354                 Font font = owner_->getFont(buf.params(), i, outerfont);
2355                 if (text_[i] == META_INSET)
2356                         return false;
2357                 if (i != initial && font != font_old)
2358                         return false;
2359                 font_old = font;
2360         }
2361
2362         return true;
2363 }
2364
2365
2366 void Paragraph::simpleDocBookOnePar(Buffer const & buf,
2367                                     odocstream & os,
2368                                     OutputParams const & runparams,
2369                                     Font const & outerfont,
2370                                     pos_type initial) const
2371 {
2372         bool emph_flag = false;
2373
2374         Layout const & style = *d->layout_;
2375         FontInfo font_old =
2376                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2377
2378         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2379                 os << "]]>";
2380
2381         // parsing main loop
2382         for (pos_type i = initial; i < size(); ++i) {
2383                 Font font = getFont(buf.params(), i, outerfont);
2384
2385                 // handle <emphasis> tag
2386                 if (font_old.emph() != font.fontInfo().emph()) {
2387                         if (font.fontInfo().emph() == FONT_ON) {
2388                                 os << "<emphasis>";
2389                                 emph_flag = true;
2390                         } else if (i != initial) {
2391                                 os << "</emphasis>";
2392                                 emph_flag = false;
2393                         }
2394                 }
2395
2396                 if (Inset const * inset = getInset(i)) {
2397                         inset->docbook(os, runparams);
2398                 } else {
2399                         char_type c = d->text_[i];
2400
2401                         if (style.pass_thru)
2402                                 os.put(c);
2403                         else
2404                                 os << sgml::escapeChar(c);
2405                 }
2406                 font_old = font.fontInfo();
2407         }
2408
2409         if (emph_flag) {
2410                 os << "</emphasis>";
2411         }
2412
2413         if (style.free_spacing)
2414                 os << '\n';
2415         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2416                 os << "<![CDATA[";
2417 }
2418
2419
2420 docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
2421                                     XHTMLStream & xs,
2422                                     OutputParams const & runparams,
2423                                     Font const & outerfont,
2424                                     pos_type initial) const
2425 {
2426         docstring retval;
2427
2428         bool emph_flag = false;
2429         bool bold_flag = false;
2430         string closing_tag;
2431
2432         Layout const & style = *d->layout_;
2433
2434         if (!runparams.for_toc && runparams.html_make_pars) {
2435                 // generate a magic label for this paragraph
2436                 string const attr = "id='" + magicLabel() + "'";
2437                 xs << html::CompTag("a", attr);
2438         }
2439
2440         FontInfo font_old =
2441                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2442
2443         // parsing main loop
2444         for (pos_type i = initial; i < size(); ++i) {
2445                 // let's not show deleted material in the output
2446                 if (isDeleted(i))
2447                         continue;
2448
2449                 Font font = getFont(buf.params(), i, outerfont);
2450
2451                 // emphasis
2452                 if (font_old.emph() != font.fontInfo().emph()) {
2453                         if (font.fontInfo().emph() == FONT_ON) {
2454                                 xs << html::StartTag("em");
2455                                 emph_flag = true;
2456                         } else if (emph_flag && i != initial) {
2457                                 xs << html::EndTag("em");
2458                                 emph_flag = false;
2459                         }
2460                 }
2461                 // bold
2462                 if (font_old.series() != font.fontInfo().series()) {
2463                         if (font.fontInfo().series() == BOLD_SERIES) {
2464                                 xs << html::StartTag("strong");
2465                                 bold_flag = true;
2466                         } else if (bold_flag && i != initial) {
2467                                 xs << html::EndTag("strong");
2468                                 bold_flag = false;
2469                         }
2470                 }
2471                 // FIXME XHTML
2472                 // Other such tags? What about the other text ranges?
2473
2474                 Inset const * inset = getInset(i);
2475                 if (inset) {
2476                         InsetCommand const * ic = inset->asInsetCommand();
2477                         InsetLayout const & il = inset->getLayout();
2478                         InsetMath const * im = inset->asInsetMath();
2479                         if (!runparams.for_toc
2480                             || im || il.isInToc() || (ic && ic->isInToc())) {
2481                                 OutputParams np = runparams;
2482                                 if (!il.htmlisblock())
2483                                         np.html_in_par = true;
2484                                 retval += inset->xhtml(xs, np);
2485                         }
2486                 } else {
2487                         char_type c = d->text_[i];
2488
2489                         if (style.pass_thru)
2490                                 xs << c;
2491                         else if (c == '-') {
2492                                 docstring str;
2493                                 int j = i + 1;
2494                                 if (j < size() && d->text_[j] == '-') {
2495                                         j += 1;
2496                                         if (j < size() && d->text_[j] == '-') {
2497                                                 str += from_ascii("&mdash;");
2498                                                 i += 2;
2499                                         } else {
2500                                                 str += from_ascii("&ndash;");
2501                                                 i += 1;
2502                                         }
2503                                 }
2504                                 else
2505                                         str += c;
2506                                 // We don't want to escape the entities. Note that
2507                                 // it is safe to do this, since str can otherwise
2508                                 // only be "-". E.g., it can't be "<".
2509                                 xs << XHTMLStream::NextRaw() << str;
2510                         } else
2511                                 xs << c;
2512                 }
2513                 font_old = font.fontInfo();
2514         }
2515
2516         xs.closeFontTags();
2517         return retval;
2518 }
2519
2520
2521 bool Paragraph::isHfill(pos_type pos) const
2522 {
2523         Inset const * inset = getInset(pos);
2524         return inset && (inset->lyxCode() == SPACE_CODE &&
2525                          inset->isStretchableSpace());
2526 }
2527
2528
2529 bool Paragraph::isNewline(pos_type pos) const
2530 {
2531         Inset const * inset = getInset(pos);
2532         return inset && inset->lyxCode() == NEWLINE_CODE;
2533 }
2534
2535
2536 bool Paragraph::isLineSeparator(pos_type pos) const
2537 {
2538         char_type const c = d->text_[pos];
2539         if (isLineSeparatorChar(c))
2540                 return true;
2541         Inset const * inset = getInset(pos);
2542         return inset && inset->isLineSeparator();
2543 }
2544
2545
2546 bool Paragraph::isWordSeparator(pos_type pos) const
2547 {
2548         if (Inset const * inset = getInset(pos))
2549                 return !inset->isLetter();
2550         char_type const c = d->text_[pos];
2551         // We want to pass the ' and escape chars to the spellchecker
2552         static docstring const quote = from_utf8(lyxrc.spellchecker_esc_chars + '\'');
2553         return (!isLetterChar(c) && !isDigit(c) && !contains(quote, c))
2554                 || pos == size();
2555 }
2556
2557
2558 bool Paragraph::isChar(pos_type pos) const
2559 {
2560         if (Inset const * inset = getInset(pos))
2561                 return inset->isChar();
2562         char_type const c = d->text_[pos];
2563         return !isLetterChar(c) && !isDigit(c) && !lyx::isSpace(c);
2564 }
2565
2566
2567 bool Paragraph::isSpace(pos_type pos) const
2568 {
2569         if (Inset const * inset = getInset(pos))
2570                 return inset->isSpace();
2571         char_type const c = d->text_[pos];
2572         return lyx::isSpace(c);
2573 }
2574
2575
2576 Language const *
2577 Paragraph::getParLanguage(BufferParams const & bparams) const
2578 {
2579         if (!empty())
2580                 return getFirstFontSettings(bparams).language();
2581         // FIXME: we should check the prev par as well (Lgb)
2582         return bparams.language;
2583 }
2584
2585
2586 bool Paragraph::isRTL(BufferParams const & bparams) const
2587 {
2588         return lyxrc.rtl_support
2589                 && getParLanguage(bparams)->rightToLeft()
2590                 && !inInset().getLayout().forceLTR();
2591 }
2592
2593
2594 void Paragraph::changeLanguage(BufferParams const & bparams,
2595                                Language const * from, Language const * to)
2596 {
2597         // change language including dummy font change at the end
2598         for (pos_type i = 0; i <= size(); ++i) {
2599                 Font font = getFontSettings(bparams, i);
2600                 if (font.language() == from) {
2601                         font.setLanguage(to);
2602                         setFont(i, font);
2603                 }
2604         }
2605 }
2606
2607
2608 bool Paragraph::isMultiLingual(BufferParams const & bparams) const
2609 {
2610         Language const * doc_language = bparams.language;
2611         FontList::const_iterator cit = d->fontlist_.begin();
2612         FontList::const_iterator end = d->fontlist_.end();
2613
2614         for (; cit != end; ++cit)
2615                 if (cit->font().language() != ignore_language &&
2616                     cit->font().language() != latex_language &&
2617                     cit->font().language() != doc_language)
2618                         return true;
2619         return false;
2620 }
2621
2622
2623 void Paragraph::getLanguages(std::set<Language const *> & languages) const
2624 {
2625         FontList::const_iterator cit = d->fontlist_.begin();
2626         FontList::const_iterator end = d->fontlist_.end();
2627
2628         for (; cit != end; ++cit) {
2629                 Language const * lang = cit->font().language();
2630                 if (lang != ignore_language &&
2631                     lang != latex_language)
2632                         languages.insert(lang);
2633         }
2634 }
2635
2636
2637 docstring Paragraph::asString(int options) const
2638 {
2639         return asString(0, size(), options);
2640 }
2641
2642
2643 docstring Paragraph::asString(pos_type beg, pos_type end, int options) const
2644 {
2645         odocstringstream os;
2646
2647         if (beg == 0
2648             && options & AS_STR_LABEL
2649             && !d->params_.labelString().empty())
2650                 os << d->params_.labelString() << ' ';
2651
2652         for (pos_type i = beg; i < end; ++i) {
2653                 char_type const c = d->text_[i];
2654                 if (isPrintable(c) || c == '\t'
2655                     || (c == '\n' && (options & AS_STR_NEWLINES)))
2656                         os.put(c);
2657                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
2658                         getInset(i)->tocString(os);
2659                         if (getInset(i)->asInsetMath())
2660                                 os << " ";
2661                 }
2662         }
2663
2664         return os.str();
2665 }
2666
2667
2668 docstring Paragraph::stringify(pos_type beg, pos_type end, int options, OutputParams & runparams) const
2669 {
2670         odocstringstream os;
2671
2672         if (beg == 0
2673                 && options & AS_STR_LABEL
2674                 && !d->params_.labelString().empty())
2675                 os << d->params_.labelString() << ' ';
2676
2677         for (pos_type i = beg; i < end; ++i) {
2678                 char_type const c = d->text_[i];
2679                 if (isPrintable(c) || c == '\t'
2680                     || (c == '\n' && (options & AS_STR_NEWLINES)))
2681                         os.put(c);
2682                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
2683                         getInset(i)->plaintext(os, runparams);
2684                 }
2685         }
2686
2687         return os.str();
2688 }
2689
2690
2691 void Paragraph::setInsetOwner(Inset const * inset)
2692 {
2693         d->inset_owner_ = inset;
2694 }
2695
2696
2697 int Paragraph::id() const
2698 {
2699         return d->id_;
2700 }
2701
2702
2703 void Paragraph::setId(int id)
2704 {
2705         d->id_ = id;
2706 }
2707
2708
2709 Layout const & Paragraph::layout() const
2710 {
2711         return *d->layout_;
2712 }
2713
2714
2715 void Paragraph::setLayout(Layout const & layout)
2716 {
2717         d->layout_ = &layout;
2718 }
2719
2720
2721 void Paragraph::setDefaultLayout(DocumentClass const & tc)
2722 {
2723         setLayout(tc.defaultLayout());
2724 }
2725
2726
2727 void Paragraph::setPlainLayout(DocumentClass const & tc)
2728 {
2729         setLayout(tc.plainLayout());
2730 }
2731
2732
2733 void Paragraph::setPlainOrDefaultLayout(DocumentClass const & tclass)
2734 {
2735         if (usePlainLayout())
2736                 setPlainLayout(tclass);
2737         else
2738                 setDefaultLayout(tclass);
2739 }
2740
2741
2742 Inset const & Paragraph::inInset() const
2743 {
2744         LASSERT(d->inset_owner_, throw ExceptionMessage(BufferException,
2745                 _("Memory problem"), _("Paragraph not properly initialized")));
2746         return *d->inset_owner_;
2747 }
2748
2749
2750 ParagraphParameters & Paragraph::params()
2751 {
2752         return d->params_;
2753 }
2754
2755
2756 ParagraphParameters const & Paragraph::params() const
2757 {
2758         return d->params_;
2759 }
2760
2761
2762 bool Paragraph::isFreeSpacing() const
2763 {
2764         if (d->layout_->free_spacing)
2765                 return true;
2766         return d->inset_owner_ && d->inset_owner_->isFreeSpacing();
2767 }
2768
2769
2770 bool Paragraph::allowEmpty() const
2771 {
2772         if (d->layout_->keepempty)
2773                 return true;
2774         return d->inset_owner_ && d->inset_owner_->allowEmpty();
2775 }
2776
2777
2778 char_type Paragraph::transformChar(char_type c, pos_type pos) const
2779 {
2780         if (!Encodings::isArabicChar(c))
2781                 return c;
2782
2783         char_type prev_char = ' ';
2784         char_type next_char = ' ';
2785
2786         for (pos_type i = pos - 1; i >= 0; --i) {
2787                 char_type const par_char = d->text_[i];
2788                 if (!Encodings::isArabicComposeChar(par_char)) {
2789                         prev_char = par_char;
2790                         break;
2791                 }
2792         }
2793
2794         for (pos_type i = pos + 1, end = size(); i < end; ++i) {
2795                 char_type const par_char = d->text_[i];
2796                 if (!Encodings::isArabicComposeChar(par_char)) {
2797                         next_char = par_char;
2798                         break;
2799                 }
2800         }
2801
2802         if (Encodings::isArabicChar(next_char)) {
2803                 if (Encodings::isArabicChar(prev_char) &&
2804                         !Encodings::isArabicSpecialChar(prev_char))
2805                         return Encodings::transformChar(c, Encodings::FORM_MEDIAL);
2806                 else
2807                         return Encodings::transformChar(c, Encodings::FORM_INITIAL);
2808         } else {
2809                 if (Encodings::isArabicChar(prev_char) &&
2810                         !Encodings::isArabicSpecialChar(prev_char))
2811                         return Encodings::transformChar(c, Encodings::FORM_FINAL);
2812                 else
2813                         return Encodings::transformChar(c, Encodings::FORM_ISOLATED);
2814         }
2815 }
2816
2817
2818 int Paragraph::checkBiblio(Buffer const & buffer)
2819 {
2820         // FIXME From JS:
2821         // This is getting more and more a mess. ...We really should clean
2822         // up this bibitem issue for 1.6. See also bug 2743.
2823
2824         // Add bibitem insets if necessary
2825         if (d->layout_->labeltype != LABEL_BIBLIO)
2826                 return 0;
2827
2828         bool hasbibitem = !d->insetlist_.empty()
2829                 // Insist on it being in pos 0
2830                 && d->text_[0] == META_INSET
2831                 && d->insetlist_.begin()->inset->lyxCode() == BIBITEM_CODE;
2832
2833         bool track_changes = buffer.params().trackChanges;
2834
2835         docstring oldkey;
2836         docstring oldlabel;
2837
2838         // remove a bibitem in pos != 0
2839         // restore it later in pos 0 if necessary
2840         // (e.g. if a user inserts contents _before_ the item)
2841         // we're assuming there's only one of these, which there
2842         // should be.
2843         int erasedInsetPosition = -1;
2844         InsetList::iterator it = d->insetlist_.begin();
2845         InsetList::iterator end = d->insetlist_.end();
2846         for (; it != end; ++it)
2847                 if (it->inset->lyxCode() == BIBITEM_CODE
2848                     && it->pos > 0) {
2849                         InsetBibitem * olditem = static_cast<InsetBibitem *>(it->inset);
2850                         oldkey = olditem->getParam("key");
2851                         oldlabel = olditem->getParam("label");
2852                         erasedInsetPosition = it->pos;
2853                         eraseChar(erasedInsetPosition, track_changes);
2854                         break;
2855         }
2856
2857         // There was an InsetBibitem at the beginning, and we didn't
2858         // have to erase one.
2859         if (hasbibitem && erasedInsetPosition < 0)
2860                         return 0;
2861
2862         // There was an InsetBibitem at the beginning and we did have to
2863         // erase one. So we give its properties to the beginning inset.
2864         if (hasbibitem) {
2865                 InsetBibitem * inset =
2866                         static_cast<InsetBibitem *>(d->insetlist_.begin()->inset);
2867                 if (!oldkey.empty())
2868                         inset->setParam("key", oldkey);
2869                 inset->setParam("label", oldlabel);
2870                 return -erasedInsetPosition;
2871         }
2872
2873         // There was no inset at the beginning, so we need to create one with
2874         // the key and label of the one we erased.
2875         InsetBibitem * inset =
2876                 new InsetBibitem(const_cast<Buffer *>(&buffer), InsetCommandParams(BIBITEM_CODE));
2877         // restore values of previously deleted item in this par.
2878         if (!oldkey.empty())
2879                 inset->setParam("key", oldkey);
2880         inset->setParam("label", oldlabel);
2881         insertInset(0, static_cast<Inset *>(inset),
2882                     Change(track_changes ? Change::INSERTED : Change::UNCHANGED));
2883
2884         return 1;
2885 }
2886
2887
2888 void Paragraph::checkAuthors(AuthorList const & authorList)
2889 {
2890         d->changes_.checkAuthors(authorList);
2891 }
2892
2893
2894 bool Paragraph::isChanged(pos_type pos) const
2895 {
2896         return lookupChange(pos).changed();
2897 }
2898
2899
2900 bool Paragraph::isInserted(pos_type pos) const
2901 {
2902         return lookupChange(pos).inserted();
2903 }
2904
2905
2906 bool Paragraph::isDeleted(pos_type pos) const
2907 {
2908         return lookupChange(pos).deleted();
2909 }
2910
2911
2912 InsetList const & Paragraph::insetList() const
2913 {
2914         return d->insetlist_;
2915 }
2916
2917
2918 void Paragraph::setBuffer(Buffer & b)
2919 {
2920         d->insetlist_.setBuffer(b);
2921 }
2922
2923
2924 Inset * Paragraph::releaseInset(pos_type pos)
2925 {
2926         Inset * inset = d->insetlist_.release(pos);
2927         /// does not honour change tracking!
2928         eraseChar(pos, false);
2929         return inset;
2930 }
2931
2932
2933 Inset * Paragraph::getInset(pos_type pos)
2934 {
2935         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
2936                  ? d->insetlist_.get(pos) : 0;
2937 }
2938
2939
2940 Inset const * Paragraph::getInset(pos_type pos) const
2941 {
2942         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
2943                  ? d->insetlist_.get(pos) : 0;
2944 }
2945
2946
2947 void Paragraph::changeCase(BufferParams const & bparams, pos_type pos,
2948                 pos_type & right, TextCase action)
2949 {
2950         // process sequences of modified characters; in change
2951         // tracking mode, this approach results in much better
2952         // usability than changing case on a char-by-char basis
2953         docstring changes;
2954
2955         bool const trackChanges = bparams.trackChanges;
2956
2957         bool capitalize = true;
2958
2959         for (; pos < right; ++pos) {
2960                 char_type oldChar = d->text_[pos];
2961                 char_type newChar = oldChar;
2962
2963                 // ignore insets and don't play with deleted text!
2964                 if (oldChar != META_INSET && !isDeleted(pos)) {
2965                         switch (action) {
2966                                 case text_lowercase:
2967                                         newChar = lowercase(oldChar);
2968                                         break;
2969                                 case text_capitalization:
2970                                         if (capitalize) {
2971                                                 newChar = uppercase(oldChar);
2972                                                 capitalize = false;
2973                                         }
2974                                         break;
2975                                 case text_uppercase:
2976                                         newChar = uppercase(oldChar);
2977                                         break;
2978                         }
2979                 }
2980
2981                 if (isWordSeparator(pos) || isDeleted(pos)) {
2982                         // permit capitalization again
2983                         capitalize = true;
2984                 }
2985
2986                 if (oldChar != newChar) {
2987                         changes += newChar;
2988                         if (pos != right - 1)
2989                                 continue;
2990                         // step behind the changing area
2991                         pos++;
2992                 }
2993
2994                 int erasePos = pos - changes.size();
2995                 for (size_t i = 0; i < changes.size(); i++) {
2996                         insertChar(pos, changes[i],
2997                                    getFontSettings(bparams,
2998                                                    erasePos),
2999                                    trackChanges);
3000                         if (!eraseChar(erasePos, trackChanges)) {
3001                                 ++erasePos;
3002                                 ++pos; // advance
3003                                 ++right; // expand selection
3004                         }
3005                 }
3006                 changes.clear();
3007         }
3008 }
3009
3010
3011 bool Paragraph::find(docstring const & str, bool cs, bool mw,
3012                 pos_type pos, bool del) const
3013 {
3014         int const strsize = str.length();
3015         int i = 0;
3016         pos_type const parsize = d->text_.size();
3017         for (i = 0; pos + i < parsize; ++i) {
3018                 if (i >= strsize)
3019                         break;
3020                 if (cs && str[i] != d->text_[pos + i])
3021                         break;
3022                 if (!cs && uppercase(str[i]) != uppercase(d->text_[pos + i]))
3023                         break;
3024                 if (!del && isDeleted(pos + i))
3025                         break;
3026         }
3027
3028         if (i != strsize)
3029                 return false;
3030
3031         // if necessary, check whether string matches word
3032         if (mw) {
3033                 if (pos > 0 && !isWordSeparator(pos - 1))
3034                         return false;
3035                 if (pos + strsize < parsize
3036                         && !isWordSeparator(pos + strsize))
3037                         return false;
3038         }
3039
3040         return true;
3041 }
3042
3043
3044 char_type Paragraph::getChar(pos_type pos) const
3045 {
3046         return d->text_[pos];
3047 }
3048
3049
3050 pos_type Paragraph::size() const
3051 {
3052         return d->text_.size();
3053 }
3054
3055
3056 bool Paragraph::empty() const
3057 {
3058         return d->text_.empty();
3059 }
3060
3061
3062 bool Paragraph::isInset(pos_type pos) const
3063 {
3064         return d->text_[pos] == META_INSET;
3065 }
3066
3067
3068 bool Paragraph::isSeparator(pos_type pos) const
3069 {
3070         //FIXME: Are we sure this can be the only separator?
3071         return d->text_[pos] == ' ';
3072 }
3073
3074
3075 void Paragraph::deregisterWords()
3076 {
3077         Private::LangWordsMap::const_iterator itl = d->words_.begin();
3078         Private::LangWordsMap::const_iterator ite = d->words_.end();
3079         for (; itl != ite; ++itl) {
3080                 WordList * wl = theWordList(itl->first);
3081                 Private::Words::const_iterator it = (itl->second).begin();
3082                 Private::Words::const_iterator et = (itl->second).end();
3083                 for (; it != et; ++it)
3084                         wl->remove(*it);
3085         }
3086         d->words_.clear();
3087 }
3088
3089
3090 void Paragraph::locateWord(pos_type & from, pos_type & to,
3091         word_location const loc) const
3092 {
3093         switch (loc) {
3094         case WHOLE_WORD_STRICT:
3095                 if (from == 0 || from == size()
3096                     || isWordSeparator(from)
3097                     || isWordSeparator(from - 1)) {
3098                         to = from;
3099                         return;
3100                 }
3101                 // no break here, we go to the next
3102
3103         case WHOLE_WORD:
3104                 // If we are already at the beginning of a word, do nothing
3105                 if (!from || isWordSeparator(from - 1))
3106                         break;
3107                 // no break here, we go to the next
3108
3109         case PREVIOUS_WORD:
3110                 // always move the cursor to the beginning of previous word
3111                 while (from && !isWordSeparator(from - 1))
3112                         --from;
3113                 break;
3114         case NEXT_WORD:
3115                 LYXERR0("Paragraph::locateWord: NEXT_WORD not implemented yet");
3116                 break;
3117         case PARTIAL_WORD:
3118                 // no need to move the 'from' cursor
3119                 break;
3120         }
3121         to = from;
3122         while (to < size() && !isWordSeparator(to))
3123                 ++to;
3124 }
3125
3126
3127 void Paragraph::collectWords()
3128 {
3129         // This is the value that needs to be exposed in the preferences
3130         // to resolve bug #6760.
3131         static int minlength = 6;
3132         pos_type n = size();
3133         for (pos_type pos = 0; pos < n; ++pos) {
3134                 if (isWordSeparator(pos))
3135                         continue;
3136                 pos_type from = pos;
3137                 locateWord(from, pos, WHOLE_WORD);
3138                 if (pos - from >= minlength) {
3139                         docstring word = asString(from, pos, AS_STR_NONE);
3140                         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
3141                         if (cit == d->fontlist_.end())
3142                                 return;
3143                         Language const * lang = cit->font().language();
3144                         d->words_[*lang].insert(word);
3145                 }
3146         }
3147 }
3148
3149
3150 void Paragraph::registerWords()
3151 {
3152         Private::LangWordsMap::const_iterator itl = d->words_.begin();
3153         Private::LangWordsMap::const_iterator ite = d->words_.end();
3154         for (; itl != ite; ++itl) {
3155                 WordList * wl = theWordList(itl->first);
3156                 Private::Words::const_iterator it = (itl->second).begin();
3157                 Private::Words::const_iterator et = (itl->second).end();
3158                 for (; it != et; ++it)
3159                         wl->insert(*it);
3160         }
3161 }
3162
3163
3164 void Paragraph::updateWords()
3165 {
3166         deregisterWords();
3167         collectWords();
3168         registerWords();
3169 }
3170
3171
3172 bool Paragraph::spellCheck(pos_type & from, pos_type & to, WordLangTuple & wl,
3173         docstring_list & suggestions, bool do_suggestion) const
3174 {
3175         SpellChecker * speller = theSpellChecker();
3176         if (!speller)
3177                 return false;
3178
3179         if (!d->layout_->spellcheck || !inInset().allowSpellCheck())
3180                 return false;
3181
3182         locateWord(from, to, WHOLE_WORD);
3183         if (from == to || from >= pos_type(d->text_.size()))
3184                 return false;
3185
3186         docstring word = asString(from, to, AS_STR_INSETS);
3187         // Ignore words with digits
3188         // FIXME: make this customizable
3189         // (note that hunspell ignores words with digits by default)
3190         bool const ignored = hasDigit(word);
3191         Language * lang = const_cast<Language *>(getFontSettings(
3192                     d->inset_owner_->buffer().params(), from).language());
3193         if (lang == d->inset_owner_->buffer().params().language
3194             && !lyxrc.spellchecker_alt_lang.empty()) {
3195                 string lang_code;
3196                 string const lang_variety =
3197                         split(lyxrc.spellchecker_alt_lang, lang_code, '-');
3198                 lang->setCode(lang_code);
3199                 lang->setVariety(lang_variety);
3200         }
3201         wl = WordLangTuple(word, lang);
3202         SpellChecker::Result res = ignored ?
3203                 SpellChecker::OK : speller->check(wl);
3204 #if 0
3205 // FIXME: the code below makes aspell abort if a word in an unknown
3206 //        language is checked.
3207         // Just ignore any error that the spellchecker reports.
3208         // FIXME: we should through out an exception and catch it in the GUI to
3209         // display the error.
3210         if (!speller->error().empty())
3211                 return false;
3212 #endif
3213
3214         bool const misspelled = res != SpellChecker::OK
3215                 && res != SpellChecker::IGNORED_WORD;
3216
3217         if (lyxrc.spellcheck_continuously)
3218                 d->fontlist_.setMisspelled(from, to, misspelled);
3219
3220         if (misspelled && do_suggestion)
3221                 speller->suggest(wl, suggestions);
3222         else
3223                 suggestions.clear();
3224
3225         return misspelled;
3226 }
3227
3228
3229 bool Paragraph::isMisspelled(pos_type pos) const
3230 {
3231         pos_type from = pos;
3232         pos_type to = pos;
3233         WordLangTuple wl;
3234         docstring_list suggestions;
3235         return spellCheck(from, to, wl, suggestions, false);
3236 }
3237
3238
3239 string Paragraph::magicLabel() const
3240 {
3241         stringstream ss;
3242         ss << "magicparlabel-" << id();
3243         return ss.str();
3244 }
3245
3246
3247 } // namespace lyx