src/Paragraph.cpp

   1 /**
   2  * \file Paragraph.cpp
   3  * This file is part of LyX, the document processor.
   4  * Licence details can be found in the file COPYING.
   5  *
   6  * \author Asger Alstrup
   7  * \author Lars Gullik Bjønnes
   8  * \author Jean-Marc Lasgouttes
   9  * \author Angus Leeming
  10  * \author John Levon
  11  * \author André Pönitz
  12  * \author Dekel Tsur
  13  * \author Jürgen Vigna
  14  *
  15  * Full author contact details are available in file CREDITS.
  16  */
  17
  18 #include <config.h>
  19
  20 #include "Paragraph.h"
  21
  22 #include "LayoutFile.h"
  23 #include "Buffer.h"
  24 #include "BufferParams.h"
  25 #include "Changes.h"
  26 #include "Counters.h"
  27 #include "Encoding.h"
  28 #include "InsetList.h"
  29 #include "Language.h"
  30 #include "LaTeXFeatures.h"
  31 #include "Layout.h"
  32 #include "Length.h"
  33 #include "Font.h"
  34 #include "FontList.h"
  35 #include "LyXRC.h"
  36 #include "OutputParams.h"
  37 #include "output_latex.h"
  38 #include "output_xhtml.h"
  39 #include "paragraph_funcs.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/lassert.h"
  56 #include "support/debug.h"
  57 #include "support/ExceptionMessage.h"
  58 #include "support/gettext.h"
  59 #include "support/lstrings.h"
  60 #include "support/Messages.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                                Encoding 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,
 168                       Layout const & layout) const;
 169
 170         /// Checks if the paragraph contains only text and no inset or font change.
 171         bool onlyText(Buffer const & buf, Font const & outerfont,
 172                       pos_type initial) const;
 173
 174         /// match a string against a particular point in the paragraph
 175         bool isTextAt(string const & str, pos_type pos) const;
 176
 177         /// Which Paragraph owns us?
 178         Paragraph * owner_;
 179
 180         /// In which Inset?
 181         Inset const * inset_owner_;
 182
 183         ///
 184         FontList fontlist_;
 185
 186         ///
 187         unsigned int id_;
 188         ///
 189         static unsigned int paragraph_id;
 190         ///
 191         ParagraphParameters params_;
 192
 193         /// for recording and looking up changes
 194         Changes changes_;
 195
 196         ///
 197         InsetList insetlist_;
 198
 199         /// end of label
 200         pos_type begin_of_body_;
 201
 202         typedef docstring TextContainer;
 203         ///
 204         TextContainer text_;
 205
 206         typedef std::set<docstring> Words;
 207         ///
 208         Words words_;
 209         ///
 210         Layout const * layout_;
 211 };
 212
 213
 214 // Initialization of the counter for the paragraph id's,
 215 unsigned int Paragraph::Private::paragraph_id = 0;
 216
 217 namespace {
 218
 219 struct special_phrase {
 220         string phrase;
 221         docstring macro;
 222         bool builtin;
 223 };
 224
 225 special_phrase const special_phrases[] = {
 226         { "LyX", from_ascii("\\LyX{}"), false },
 227         { "TeX", from_ascii("\\TeX{}"), true },
 228         { "LaTeX2e", from_ascii("\\LaTeXe{}"), true },
 229         { "LaTeX", from_ascii("\\LaTeX{}"), true },
 230 };
 231
 232 size_t const phrases_nr = sizeof(special_phrases)/sizeof(special_phrase);
 233
 234 } // namespace anon
 235
 236
 237 Paragraph::Private::Private(Paragraph * owner, Layout const & layout)
 238         : owner_(owner), inset_owner_(0), begin_of_body_(0), layout_(&layout)
 239 {
 240         id_ = paragraph_id++;
 241         text_.reserve(100);
 242 }
 243
 244
 245 Paragraph::Private::Private(Private const & p, Paragraph * owner)
 246         : owner_(owner), inset_owner_(p.inset_owner_), fontlist_(p.fontlist_),
 247           params_(p.params_), changes_(p.changes_), insetlist_(p.insetlist_),
 248           begin_of_body_(p.begin_of_body_), text_(p.text_), words_(p.words_),
 249           layout_(p.layout_)
 250 {
 251         id_ = paragraph_id++;
 252 }
 253
 254
 255 Paragraph::Private::Private(Private const & p, Paragraph * owner,
 256         pos_type beg, pos_type end)
 257         : owner_(owner), inset_owner_(p.inset_owner_),
 258           params_(p.params_), changes_(p.changes_),
 259           insetlist_(p.insetlist_, beg, end),
 260           begin_of_body_(p.begin_of_body_), words_(p.words_),
 261           layout_(p.layout_)
 262 {
 263         id_ = paragraph_id++;
 264         if (beg >= pos_type(p.text_.size()))
 265                 return;
 266         text_ = p.text_.substr(beg, end - beg);
 267
 268         FontList::const_iterator fcit = fontlist_.begin();
 269         FontList::const_iterator fend = fontlist_.end();
 270         for (; fcit != fend; ++fcit) {
 271                 if (fcit->pos() < beg)
 272                         continue;
 273                 if (fcit->pos() >= end) {
 274                         // Add last entry in the fontlist_.
 275                         fontlist_.set(text_.size() - 1, fcit->font());
 276                         break;
 277                 }
 278                 // Add a new entry in the fontlist_.
 279                 fontlist_.set(fcit->pos() - beg, fcit->font());
 280         }
 281 }
 282
 283
 284 void Paragraph::addChangesToToc(DocIterator const & cdit,
 285         Buffer const & buf) const
 286 {
 287         d->changes_.addToToc(cdit, buf);
 288 }
 289
 290
 291 bool Paragraph::isFullyDeleted(pos_type start, pos_type end) const
 292 {
 293         LASSERT(start >= 0 && start <= size(), /**/);
 294         LASSERT(end > start && end <= size() + 1, /**/);
 295
 296         return d->changes_.isFullyDeleted(start, end);
 297 }
 298
 299
 300 bool Paragraph::isChanged(pos_type start, pos_type end) const
 301 {
 302         LASSERT(start >= 0 && start <= size(), /**/);
 303         LASSERT(end > start && end <= size() + 1, /**/);
 304
 305         return d->changes_.isChanged(start, end);
 306 }
 307
 308
 309 bool Paragraph::isMergedOnEndOfParDeletion(bool trackChanges) const
 310 {
 311         // keep the logic here in sync with the logic of eraseChars()
 312         if (!trackChanges)
 313                 return true;
 314
 315         Change const change = d->changes_.lookup(size());
 316         return change.type == Change::INSERTED && change.author == 0;
 317 }
 318
 319
 320 void Paragraph::setChange(Change const & change)
 321 {
 322         // beware of the imaginary end-of-par character!
 323         d->changes_.set(change, 0, size() + 1);
 324
 325         /*
 326          * Propagate the change recursively - but not in case of DELETED!
 327          *
 328          * Imagine that your co-author makes changes in an existing inset. He
 329          * sends your document to you and you come to the conclusion that the
 330          * inset should go completely. If you erase it, LyX must not delete all
 331          * text within the inset. Otherwise, the change tracked insertions of
 332          * your co-author get lost and there is no way to restore them later.
 333          *
 334          * Conclusion: An inset's content should remain untouched if you delete it
 335          */
 336
 337         if (change.type != Change::DELETED) {
 338                 for (pos_type pos = 0; pos < size(); ++pos) {
 339                         if (Inset * inset = getInset(pos))
 340                                 inset->setChange(change);
 341                 }
 342         }
 343 }
 344
 345
 346 void Paragraph::setChange(pos_type pos, Change const & change)
 347 {
 348         LASSERT(pos >= 0 && pos <= size(), /**/);
 349         d->changes_.set(change, pos);
 350
 351         // see comment in setChange(Change const &) above
 352         if (change.type != Change::DELETED && pos < size())
 353                         if (Inset * inset = getInset(pos))
 354                                 inset->setChange(change);
 355 }
 356
 357
 358 Change const & Paragraph::lookupChange(pos_type pos) const
 359 {
 360         LASSERT(pos >= 0 && pos <= size(), /**/);
 361         return d->changes_.lookup(pos);
 362 }
 363
 364
 365 void Paragraph::acceptChanges(BufferParams const & bparams, pos_type start,
 366                 pos_type end)
 367 {
 368         LASSERT(start >= 0 && start <= size(), /**/);
 369         LASSERT(end > start && end <= size() + 1, /**/);
 370
 371         for (pos_type pos = start; pos < end; ++pos) {
 372                 switch (lookupChange(pos).type) {
 373                         case Change::UNCHANGED:
 374                                 // accept changes in nested inset
 375                                 if (Inset * inset = getInset(pos))
 376                                         inset->acceptChanges(bparams);
 377                                 break;
 378
 379                         case Change::INSERTED:
 380                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 381                                 // also accept changes in nested inset
 382                                 if (Inset * inset = getInset(pos))
 383                                         inset->acceptChanges(bparams);
 384                                 break;
 385
 386                         case Change::DELETED:
 387                                 // Suppress access to non-existent
 388                                 // "end-of-paragraph char"
 389                                 if (pos < size()) {
 390                                         eraseChar(pos, false);
 391                                         --end;
 392                                         --pos;
 393                                 }
 394                                 break;
 395                 }
 396
 397         }
 398 }
 399
 400
 401 void Paragraph::rejectChanges(BufferParams const & bparams,
 402                 pos_type start, pos_type end)
 403 {
 404         LASSERT(start >= 0 && start <= size(), /**/);
 405         LASSERT(end > start && end <= size() + 1, /**/);
 406
 407         for (pos_type pos = start; pos < end; ++pos) {
 408                 switch (lookupChange(pos).type) {
 409                         case Change::UNCHANGED:
 410                                 // reject changes in nested inset
 411                                 if (Inset * inset = getInset(pos))
 412                                                 inset->rejectChanges(bparams);
 413                                 break;
 414
 415                         case Change::INSERTED:
 416                                 // Suppress access to non-existent
 417                                 // "end-of-paragraph char"
 418                                 if (pos < size()) {
 419                                         eraseChar(pos, false);
 420                                         --end;
 421                                         --pos;
 422                                 }
 423                                 break;
 424
 425                         case Change::DELETED:
 426                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 427
 428                                 // Do NOT reject changes within a deleted inset!
 429                                 // There may be insertions of a co-author inside of it!
 430
 431                                 break;
 432                 }
 433         }
 434 }
 435
 436
 437 void Paragraph::Private::insertChar(pos_type pos, char_type c,
 438                 Change const & change)
 439 {
 440         LASSERT(pos >= 0 && pos <= int(text_.size()), /**/);
 441
 442         // track change
 443         changes_.insert(change, pos);
 444
 445         // This is actually very common when parsing buffers (and
 446         // maybe inserting ascii text)
 447         if (pos == pos_type(text_.size())) {
 448                 // when appending characters, no need to update tables
 449                 text_.push_back(c);
 450                 return;
 451         }
 452
 453         text_.insert(text_.begin() + pos, c);
 454
 455         // Update the font table.
 456         fontlist_.increasePosAfterPos(pos);
 457
 458         // Update the insets
 459         insetlist_.increasePosAfterPos(pos);
 460 }
 461
 462
 463 bool Paragraph::insertInset(pos_type pos, Inset * inset,
 464                                    Change const & change)
 465 {
 466         LASSERT(inset, /**/);
 467         LASSERT(pos >= 0 && pos <= size(), /**/);
 468
 469         // Paragraph::insertInset() can be used in cut/copy/paste operation where
 470         // d->inset_owner_ is not set yet.
 471         if (d->inset_owner_ && !d->inset_owner_->insetAllowed(inset->lyxCode()))
 472                 return false;
 473
 474         d->insertChar(pos, META_INSET, change);
 475         LASSERT(d->text_[pos] == META_INSET, /**/);
 476
 477         // Add a new entry in the insetlist_.
 478         d->insetlist_.insert(inset, pos);
 479         return true;
 480 }
 481
 482
 483 bool Paragraph::eraseChar(pos_type pos, bool trackChanges)
 484 {
 485         LASSERT(pos >= 0 && pos <= size(), return false);
 486
 487         // keep the logic here in sync with the logic of isMergedOnEndOfParDeletion()
 488
 489         if (trackChanges) {
 490                 Change change = d->changes_.lookup(pos);
 491
 492                 // set the character to DELETED if
 493                 //  a) it was previously unchanged or
 494                 //  b) it was inserted by a co-author
 495
 496                 if (change.type == Change::UNCHANGED ||
 497                     (change.type == Change::INSERTED && change.author != 0)) {
 498                         setChange(pos, Change(Change::DELETED));
 499                         return false;
 500                 }
 501
 502                 if (change.type == Change::DELETED)
 503                         return false;
 504         }
 505
 506         // Don't physically access the imaginary end-of-paragraph character.
 507         // eraseChar() can only mark it as DELETED. A physical deletion of
 508         // end-of-par must be handled externally.
 509         if (pos == size()) {
 510                 return false;
 511         }
 512
 513         // track change
 514         d->changes_.erase(pos);
 515
 516         // if it is an inset, delete the inset entry
 517         if (d->text_[pos] == META_INSET)
 518                 d->insetlist_.erase(pos);
 519
 520         d->text_.erase(d->text_.begin() + pos);
 521
 522         // Update the fontlist_
 523         d->fontlist_.erase(pos);
 524
 525         // Update the insetlist_
 526         d->insetlist_.decreasePosAfterPos(pos);
 527
 528         return true;
 529 }
 530
 531
 532 int Paragraph::eraseChars(pos_type start, pos_type end, bool trackChanges)
 533 {
 534         LASSERT(start >= 0 && start <= size(), /**/);
 535         LASSERT(end >= start && end <= size() + 1, /**/);
 536
 537         pos_type i = start;
 538         for (pos_type count = end - start; count; --count) {
 539                 if (!eraseChar(i, trackChanges))
 540                         ++i;
 541         }
 542         return end - i;
 543 }
 544
 545
 546 int Paragraph::Private::latexSurrogatePair(odocstream & os, char_type c,
 547                 char_type next, Encoding const & encoding)
 548 {
 549         // Writing next here may circumvent a possible font change between
 550         // c and next. Since next is only output if it forms a surrogate pair
 551         // with c we can ignore this:
 552         // A font change inside a surrogate pair does not make sense and is
 553         // hopefully impossible to input.
 554         // FIXME: change tracking
 555         // Is this correct WRT change tracking?
 556         docstring const latex1 = encoding.latexChar(next);
 557         docstring const latex2 = encoding.latexChar(c);
 558         if (docstring(1, next) == latex1) {
 559                 // the encoding supports the combination
 560                 os << latex2 << latex1;
 561                 return latex1.length() + latex2.length();
 562         } else
 563                 os << latex1 << '{' << latex2 << '}';
 564         return latex1.length() + latex2.length() + 2;
 565 }
 566
 567
 568 bool Paragraph::Private::simpleTeXBlanks(OutputParams const & runparams,
 569                                        odocstream & os, TexRow & texrow,
 570                                        pos_type i,
 571                                        unsigned int & column,
 572                                        Font const & font,
 573                                        Layout const & style)
 574 {
 575         if (style.pass_thru || runparams.verbatim)
 576                 return false;
 577
 578         if (i + 1 < int(text_.size())) {
 579                 char_type next = text_[i + 1];
 580                 if (Encodings::isCombiningChar(next)) {
 581                         Encoding const & encoding = *(runparams.encoding);
 582                         // This space has an accent, so we must always output it.
 583                         column += latexSurrogatePair(os, ' ', next, encoding) - 1;
 584                         return true;
 585                 }
 586         }
 587
 588         if (lyxrc.plaintext_linelen > 0
 589             && column > lyxrc.plaintext_linelen
 590             && i
 591             && text_[i - 1] != ' '
 592             && (i + 1 < int(text_.size()))
 593             // same in FreeSpacing mode
 594             && !owner_->isFreeSpacing()
 595             // In typewriter mode, we want to avoid
 596             // ! . ? : at the end of a line
 597             && !(font.fontInfo().family() == TYPEWRITER_FAMILY
 598                  && (text_[i - 1] == '.'
 599                      || text_[i - 1] == '?'
 600                      || text_[i - 1] == ':'
 601                      || text_[i - 1] == '!'))) {
 602                 os << '\n';
 603                 texrow.newline();
 604                 texrow.start(owner_->id(), i + 1);
 605                 column = 0;
 606         } else if (style.free_spacing) {
 607                 os << '~';
 608         } else {
 609                 os << ' ';
 610         }
 611         return false;
 612 }
 613
 614
 615 int Paragraph::Private::writeScriptChars(odocstream & os,
 616                                          docstring const & ltx,
 617                                          Change const & runningChange,
 618                                          Encoding const & encoding,
 619                                          pos_type & i)
 620 {
 621         // FIXME: modifying i here is not very nice...
 622
 623         // We only arrive here when a proper language for character text_[i] has
 624         // not been specified (i.e., it could not be translated in the current
 625         // latex encoding) or its latex translation has been forced, and it
 626         // belongs to a known script.
 627         // Parameter ltx contains the latex translation of text_[i] as specified
 628         // in the unicodesymbols file and is something like "\textXXX{<spec>}".
 629         // The latex macro name "textXXX" specifies the script to which text_[i]
 630         // belongs and we use it in order to check whether characters from the
 631         // same script immediately follow, such that we can collect them in a
 632         // single "\textXXX" macro. So, we have to retain "\textXXX{<spec>"
 633         // for the first char but only "<spec>" for all subsequent chars.
 634         docstring::size_type const brace1 = ltx.find_first_of(from_ascii("{"));
 635         docstring::size_type const brace2 = ltx.find_last_of(from_ascii("}"));
 636         string script = to_ascii(ltx.substr(1, brace1 - 1));
 637         int pos = 0;
 638         int length = brace2;
 639         bool closing_brace = true;
 640         if (script == "textgreek" && encoding.latexName() == "iso-8859-7") {
 641                 // Correct encoding is being used, so we can avoid \textgreek.
 642                 pos = brace1 + 1;
 643                 length -= pos;
 644                 closing_brace = false;
 645         }
 646         os << ltx.substr(pos, length);
 647         int size = text_.size();
 648         while (i + 1 < size) {
 649                 char_type const next = text_[i + 1];
 650                 // Stop here if next character belongs to another script
 651                 // or there is a change in change tracking status.
 652                 if (!Encodings::isKnownScriptChar(next, script) ||
 653                     runningChange != owner_->lookupChange(i + 1))
 654                         break;
 655                 Font prev_font;
 656                 bool found = false;
 657                 FontList::const_iterator cit = fontlist_.begin();
 658                 FontList::const_iterator end = fontlist_.end();
 659                 for (; cit != end; ++cit) {
 660                         if (cit->pos() >= i && !found) {
 661                                 prev_font = cit->font();
 662                                 found = true;
 663                         }
 664                         if (cit->pos() >= i + 1)
 665                                 break;
 666                 }
 667                 // Stop here if there is a font attribute or encoding change.
 668                 if (found && cit != end && prev_font != cit->font())
 669                         break;
 670                 docstring const latex = encoding.latexChar(next);
 671                 docstring::size_type const b1 =
 672                                         latex.find_first_of(from_ascii("{"));
 673                 docstring::size_type const b2 =
 674                                         latex.find_last_of(from_ascii("}"));
 675                 int const len = b2 - b1 - 1;
 676                 os << latex.substr(b1 + 1, len);
 677                 length += len;
 678                 ++i;
 679         }
 680         if (closing_brace) {
 681                 os << '}';
 682                 ++length;
 683         }
 684         return length;
 685 }
 686
 687
 688 bool Paragraph::Private::isTextAt(string const & str, pos_type pos) const
 689 {
 690         pos_type const len = str.length();
 691
 692         // is the paragraph large enough?
 693         if (pos + len > int(text_.size()))
 694                 return false;
 695
 696         // does the wanted text start at point?
 697         for (string::size_type i = 0; i < str.length(); ++i) {
 698                 // Caution: direct comparison of characters works only
 699                 // because str is pure ASCII.
 700                 if (str[i] != text_[pos + i])
 701                         return false;
 702         }
 703
 704         return fontlist_.hasChangeInRange(pos, len);
 705 }
 706
 707
 708 void Paragraph::Private::latexInset(
 709                                              BufferParams const & bparams,
 710                                              odocstream & os,
 711                                              TexRow & texrow,
 712                                              OutputParams & runparams,
 713                                              Font & running_font,
 714                                              Font & basefont,
 715                                              Font const & outerfont,
 716                                              bool & open_font,
 717                                              Change & running_change,
 718                                              Layout const & style,
 719                                              pos_type & i,
 720                                              unsigned int & column)
 721 {
 722         Inset * inset = owner_->getInset(i);
 723         LASSERT(inset, /**/);
 724
 725         if (style.pass_thru) {
 726                 inset->plaintext(os, runparams);
 727                 return;
 728         }
 729
 730         // FIXME: move this to InsetNewline::latex
 731         if (inset->lyxCode() == NEWLINE_CODE) {
 732                 // newlines are handled differently here than
 733                 // the default in simpleTeXSpecialChars().
 734                 if (!style.newline_allowed) {
 735                         os << '\n';
 736                 } else {
 737                         if (open_font) {
 738                                 column += running_font.latexWriteEndChanges(
 739                                         os, bparams, runparams,
 740                                         basefont, basefont);
 741                                 open_font = false;
 742                         }
 743
 744                         if (running_font.fontInfo().family() == TYPEWRITER_FAMILY)
 745                                 os << '~';
 746
 747                         basefont = owner_->getLayoutFont(bparams, outerfont);
 748                         running_font = basefont;
 749
 750                         if (runparams.moving_arg)
 751                                 os << "\\protect ";
 752
 753                 }
 754                 texrow.newline();
 755                 texrow.start(owner_->id(), i + 1);
 756                 column = 0;
 757         }
 758
 759         if (owner_->lookupChange(i).type == Change::DELETED) {
 760                 if( ++runparams.inDeletedInset == 1)
 761                         runparams.changeOfDeletedInset = owner_->lookupChange(i);
 762         }
 763
 764         if (inset->canTrackChanges()) {
 765                 column += Changes::latexMarkChange(os, bparams, running_change,
 766                         Change(Change::UNCHANGED));
 767                 running_change = Change(Change::UNCHANGED);
 768         }
 769
 770         bool close = false;
 771         odocstream::pos_type const len = os.tellp();
 772
 773         if (inset->forceLTR()
 774             && running_font.isRightToLeft()
 775                 // ERT is an exception, it should be output with no decorations at all
 776                 && inset->lyxCode() != ERT_CODE) {
 777                 if (running_font.language()->lang() == "farsi")
 778                         os << "\\beginL{}";
 779                 else
 780                         os << "\\L{";
 781                 close = true;
 782         }
 783
 784         // FIXME: Bug: we can have an empty font change here!
 785         // if there has just been a font change, we are going to close it
 786         // right now, which means stupid latex code like \textsf{}. AFAIK,
 787         // this does not harm dvi output. A minor bug, thus (JMarc)
 788
 789         // Some insets cannot be inside a font change command.
 790         // However, even such insets *can* be placed in \L or \R
 791         // or their equivalents (for RTL language switches), so we don't
 792         // close the language in those cases.
 793         // ArabTeX, though, cannot handle this special behavior, it seems.
 794         bool arabtex = basefont.language()->lang() == "arabic_arabtex"
 795                 || running_font.language()->lang() == "arabic_arabtex";
 796         if (open_font && inset->noFontChange()) {
 797                 bool closeLanguage = arabtex
 798                         || basefont.isRightToLeft() == running_font.isRightToLeft();
 799                 unsigned int count = running_font.latexWriteEndChanges(os,
 800                         bparams, runparams, basefont, basefont, closeLanguage);
 801                 column += count;
 802                 // if any font properties were closed, update the running_font,
 803                 // making sure, however, to leave the language as it was
 804                 if (count > 0) {
 805                         // FIXME: probably a better way to keep track of the old
 806                         // language, than copying the entire font?
 807                         Font const copy_font(running_font);
 808                         basefont = owner_->getLayoutFont(bparams, outerfont);
 809                         running_font = basefont;
 810                         if (!closeLanguage)
 811                                 running_font.setLanguage(copy_font.language());
 812                         // leave font open if language is still open
 813                         open_font = (running_font.language() == basefont.language());
 814                         if (closeLanguage)
 815                                 runparams.local_font = &basefont;
 816                 }
 817         }
 818
 819         int tmp;
 820
 821         try {
 822                 tmp = inset->latex(os, runparams);
 823         } catch (EncodingException & e) {
 824                 // add location information and throw again.
 825                 e.par_id = id_;
 826                 e.pos = i;
 827                 throw(e);
 828         }
 829
 830         if (close) {
 831                 if (running_font.language()->lang() == "farsi")
 832                                 os << "\\endL{}";
 833                         else
 834                                 os << '}';
 835         }
 836
 837         if (tmp) {
 838                 for (int j = 0; j < tmp; ++j)
 839                         texrow.newline();
 840
 841                 texrow.start(owner_->id(), i + 1);
 842                 column = 0;
 843         } else {
 844                 column += os.tellp() - len;
 845         }
 846
 847         if (owner_->lookupChange(i).type == Change::DELETED)
 848                 --runparams.inDeletedInset;
 849 }
 850
 851
 852 void Paragraph::Private::latexSpecialChar(
 853                                              odocstream & os,
 854                                              OutputParams const & runparams,
 855                                              Font const & running_font,
 856                                              Change const & running_change,
 857                                              Layout const & style,
 858                                              pos_type & i,
 859                                              unsigned int & column)
 860 {
 861         char_type const c = text_[i];
 862
 863         if (style.pass_thru) {
 864                 if (c != '\0')
 865                         // FIXME UNICODE: This can fail if c cannot
 866                         // be encoded in the current encoding.
 867                         os.put(c);
 868                 return;
 869         }
 870
 871         if (runparams.verbatim) {
 872                 // FIXME UNICODE: This can fail if c cannot
 873                 // be encoded in the current encoding.
 874                 os.put(c);
 875                 return;
 876         }
 877
 878         // If T1 font encoding is used, use the special
 879         // characters it provides.
 880         // NOTE: some languages reset the font encoding
 881         // internally
 882         if (!running_font.language()->internalFontEncoding()
 883             && lyxrc.fontenc == "T1" && latexSpecialT1(c, os, i, column))
 884                 return;
 885
 886         // \tt font needs special treatment
 887         if (running_font.fontInfo().family() == TYPEWRITER_FAMILY
 888                 && latexSpecialTypewriter(c, os, i, column))
 889                 return;
 890
 891         // Otherwise, we use what LaTeX provides us.
 892         switch (c) {
 893         case '\\':
 894                 os << "\\textbackslash{}";
 895                 column += 15;
 896                 break;
 897         case '<':
 898                 os << "\\textless{}";
 899                 column += 10;
 900                 break;
 901         case '>':
 902                 os << "\\textgreater{}";
 903                 column += 13;
 904                 break;
 905         case '|':
 906                 os << "\\textbar{}";
 907                 column += 9;
 908                 break;
 909         case '-':
 910                 os << '-';
 911                 break;
 912         case '\"':
 913                 os << "\\char`\\\"{}";
 914                 column += 9;
 915                 break;
 916
 917         case '$': case '&':
 918         case '%': case '#': case '{':
 919         case '}': case '_':
 920                 os << '\\';
 921                 os.put(c);
 922                 column += 1;
 923                 break;
 924
 925         case '~':
 926                 os << "\\textasciitilde{}";
 927                 column += 16;
 928                 break;
 929
 930         case '^':
 931                 os << "\\textasciicircum{}";
 932                 column += 17;
 933                 break;
 934
 935         case '*':
 936         case '[':
 937         case ']':
 938                 // avoid being mistaken for optional arguments
 939                 os << '{';
 940                 os.put(c);
 941                 os << '}';
 942                 column += 2;
 943                 break;
 944
 945         case ' ':
 946                 // Blanks are printed before font switching.
 947                 // Sure? I am not! (try nice-latex)
 948                 // I am sure it's correct. LyX might be smarter
 949                 // in the future, but for now, nothing wrong is
 950                 // written. (Asger)
 951                 break;
 952
 953         default:
 954                 // LyX, LaTeX etc.
 955                 if (latexSpecialPhrase(os, i, column, runparams))
 956                         return;
 957
 958                 if (c == '\0')
 959                         return;
 960
 961                 Encoding const & encoding = *(runparams.encoding);
 962                 if (i + 1 < int(text_.size())) {
 963                         char_type next = text_[i + 1];
 964                         if (Encodings::isCombiningChar(next)) {
 965                                 column += latexSurrogatePair(os, c, next, encoding) - 1;
 966                                 ++i;
 967                                 break;
 968                         }
 969                 }
 970                 string script;
 971                 docstring const latex = encoding.latexChar(c);
 972                 if (Encodings::isKnownScriptChar(c, script)
 973                     && prefixIs(latex, from_ascii("\\" + script)))
 974                         column += writeScriptChars(os, latex,
 975                                         running_change, encoding, i) - 1;
 976                 else if (latex.length() > 1 && latex[latex.length() - 1] != '}') {
 977                         // Prevent eating of a following
 978                         // space or command corruption by
 979                         // following characters
 980                         column += latex.length() + 1;
 981                         os << latex << "{}";
 982                 } else {
 983                         column += latex.length() - 1;
 984                         os << latex;
 985                 }
 986                 break;
 987         }
 988 }
 989
 990
 991 bool Paragraph::Private::latexSpecialT1(char_type const c, odocstream & os,
 992         pos_type i, unsigned int & column)
 993 {
 994         switch (c) {
 995         case '>':
 996         case '<':
 997                 os.put(c);
 998                 // In T1 encoding, these characters exist
 999                 // but we should avoid ligatures
1000                 if (i + 1 >= int(text_.size()) || text_[i + 1] != c)
1001                         return true;
1002                 os << "\\textcompwordmark{}";
1003                 column += 19;
1004                 return true;
1005         case '|':
1006                 os.put(c);
1007                 return true;
1008         case '\"':
1009                 // soul.sty breaks with \char`\"
1010                 os << "\\textquotedbl{}";
1011                 column += 14;
1012                 return true;
1013         default:
1014                 return false;
1015         }
1016 }
1017
1018
1019 bool Paragraph::Private::latexSpecialTypewriter(char_type const c, odocstream & os,
1020         pos_type i, unsigned int & column)
1021 {
1022         switch (c) {
1023         case '-':
1024                 // within \ttfamily, "--" is merged to "-" (no endash)
1025                 // so we avoid this rather irritating ligature
1026                 if (i + 1 < int(text_.size()) && text_[i + 1] == '-') {
1027                         os << "-{}";
1028                         column += 2;
1029                 } else
1030                         os << '-';
1031                 return true;
1032
1033         // everything else has to be checked separately
1034         // (depending on the encoding)
1035         default:
1036                 return false;
1037         }
1038 }
1039
1040
1041 bool Paragraph::Private::latexSpecialPhrase(odocstream & os, pos_type & i,
1042         unsigned int & column, OutputParams const & runparams)
1043 {
1044         // FIXME: if we have "LaTeX" with a font
1045         // change in the middle (before the 'T', then
1046         // the "TeX" part is still special cased.
1047         // Really we should only operate this on
1048         // "words" for some definition of word
1049
1050         for (size_t pnr = 0; pnr < phrases_nr; ++pnr) {
1051                 if (!isTextAt(special_phrases[pnr].phrase, i))
1052                         continue;
1053                 if (runparams.moving_arg)
1054                         os << "\\protect";
1055                 os << special_phrases[pnr].macro;
1056                 i += special_phrases[pnr].phrase.length() - 1;
1057                 column += special_phrases[pnr].macro.length() - 1;
1058                 return true;
1059         }
1060         return false;
1061 }
1062
1063
1064 void Paragraph::Private::validate(LaTeXFeatures & features,
1065                                 Layout const & layout) const
1066 {
1067         // check the params.
1068         if (!params_.spacing().isDefault())
1069                 features.require("setspace");
1070
1071         // then the layouts
1072         features.useLayout(layout.name());
1073
1074         // then the fonts
1075         fontlist_.validate(features);
1076
1077         // then the indentation
1078         if (!params_.leftIndent().zero())
1079                 features.require("ParagraphLeftIndent");
1080
1081         // then the insets
1082         InsetList::const_iterator icit = insetlist_.begin();
1083         InsetList::const_iterator iend = insetlist_.end();
1084         for (; icit != iend; ++icit) {
1085                 if (icit->inset) {
1086                         icit->inset->validate(features);
1087                         if (layout.needprotect &&
1088                             icit->inset->lyxCode() == FOOT_CODE)
1089                                 features.require("NeedLyXFootnoteCode");
1090                 }
1091         }
1092
1093         // then the contents
1094         for (pos_type i = 0; i < int(text_.size()) ; ++i) {
1095                 for (size_t pnr = 0; pnr < phrases_nr; ++pnr) {
1096                         if (!special_phrases[pnr].builtin
1097                             && isTextAt(special_phrases[pnr].phrase, i)) {
1098                                 features.require(special_phrases[pnr].phrase);
1099                                 break;
1100                         }
1101                 }
1102                 Encodings::validate(text_[i], features);
1103         }
1104 }
1105
1106 /////////////////////////////////////////////////////////////////////
1107 //
1108 // Paragraph
1109 //
1110 /////////////////////////////////////////////////////////////////////
1111
1112 namespace {
1113         Layout const emptyParagraphLayout;
1114 }
1115
1116 Paragraph::Paragraph()
1117         : d(new Paragraph::Private(this, emptyParagraphLayout))
1118 {
1119         itemdepth = 0;
1120         d->params_.clear();
1121 }
1122
1123
1124 Paragraph::Paragraph(Paragraph const & par)
1125         : itemdepth(par.itemdepth),
1126         d(new Paragraph::Private(*par.d, this))
1127 {
1128         registerWords();
1129 }
1130
1131
1132 Paragraph::Paragraph(Paragraph const & par, pos_type beg, pos_type end)
1133         : itemdepth(par.itemdepth),
1134         d(new Paragraph::Private(*par.d, this, beg, end))
1135 {
1136         registerWords();
1137 }
1138
1139
1140 Paragraph & Paragraph::operator=(Paragraph const & par)
1141 {
1142         // needed as we will destroy the private part before copying it
1143         if (&par != this) {
1144                 itemdepth = par.itemdepth;
1145
1146                 deregisterWords();
1147                 delete d;
1148                 d = new Private(*par.d, this);
1149                 registerWords();
1150         }
1151         return *this;
1152 }
1153
1154
1155 Paragraph::~Paragraph()
1156 {
1157         deregisterWords();
1158         delete d;
1159 }
1160
1161
1162 void Paragraph::write(ostream & os, BufferParams const & bparams,
1163         depth_type & dth) const
1164 {
1165         // The beginning or end of a deeper (i.e. nested) area?
1166         if (dth != d->params_.depth()) {
1167                 if (d->params_.depth() > dth) {
1168                         while (d->params_.depth() > dth) {
1169                                 os << "\n\\begin_deeper";
1170                                 ++dth;
1171                         }
1172                 } else {
1173                         while (d->params_.depth() < dth) {
1174                                 os << "\n\\end_deeper";
1175                                 --dth;
1176                         }
1177                 }
1178         }
1179
1180         // First write the layout
1181         os << "\n\\begin_layout " << to_utf8(d->layout_->name()) << '\n';
1182
1183         d->params_.write(os);
1184
1185         Font font1(inherit_font, bparams.language);
1186
1187         Change running_change = Change(Change::UNCHANGED);
1188
1189         int column = 0;
1190         for (pos_type i = 0; i <= size(); ++i) {
1191
1192                 Change const change = lookupChange(i);
1193                 Changes::lyxMarkChange(os, column, running_change, change);
1194                 running_change = change;
1195
1196                 if (i == size())
1197                         break;
1198
1199                 // Write font changes (ignore spelling markers)
1200                 Font font2 = getFontSettings(bparams, i);
1201                 font2.setMisspelled(false);
1202                 if (font2 != font1) {
1203                         font2.lyxWriteChanges(font1, os);
1204                         column = 0;
1205                         font1 = font2;
1206                 }
1207
1208                 char_type const c = d->text_[i];
1209                 switch (c) {
1210                 case META_INSET:
1211                         if (Inset const * inset = getInset(i)) {
1212                                 if (inset->directWrite()) {
1213                                         // international char, let it write
1214                                         // code directly so it's shorter in
1215                                         // the file
1216                                         inset->write(os);
1217                                 } else {
1218                                         if (i)
1219                                                 os << '\n';
1220                                         os << "\\begin_inset ";
1221                                         inset->write(os);
1222                                         os << "\n\\end_inset\n\n";
1223                                         column = 0;
1224                                 }
1225                         }
1226                         break;
1227                 case '\\':
1228                         os << "\n\\backslash\n";
1229                         column = 0;
1230                         break;
1231                 case '.':
1232                         if (i + 1 < size() && d->text_[i + 1] == ' ') {
1233                                 os << ".\n";
1234                                 column = 0;
1235                         } else
1236                                 os << '.';
1237                         break;
1238                 default:
1239                         if ((column > 70 && c == ' ')
1240                             || column > 79) {
1241                                 os << '\n';
1242                                 column = 0;
1243                         }
1244                         // this check is to amend a bug. LyX sometimes
1245                         // inserts '\0' this could cause problems.
1246                         if (c != '\0')
1247                                 os << to_utf8(docstring(1, c));
1248                         else
1249                                 LYXERR0("NUL char in structure.");
1250                         ++column;
1251                         break;
1252                 }
1253         }
1254
1255         os << "\n\\end_layout\n";
1256 }
1257
1258
1259 void Paragraph::validate(LaTeXFeatures & features) const
1260 {
1261         d->validate(features, *d->layout_);
1262 }
1263
1264
1265 void Paragraph::insert(pos_type start, docstring const & str,
1266                        Font const & font, Change const & change)
1267 {
1268         for (size_t i = 0, n = str.size(); i != n ; ++i)
1269                 insertChar(start + i, str[i], font, change);
1270 }
1271
1272
1273 void Paragraph::appendChar(char_type c, Font const & font,
1274                 Change const & change)
1275 {
1276         // track change
1277         d->changes_.insert(change, d->text_.size());
1278         // when appending characters, no need to update tables
1279         d->text_.push_back(c);
1280         setFont(d->text_.size() - 1, font);
1281 }
1282
1283
1284 void Paragraph::appendString(docstring const & s, Font const & font,
1285                 Change const & change)
1286 {
1287         pos_type end = s.size();
1288         size_t oldsize = d->text_.size();
1289         size_t newsize = oldsize + end;
1290         size_t capacity = d->text_.capacity();
1291         if (newsize >= capacity)
1292                 d->text_.reserve(max(capacity + 100, newsize));
1293
1294         // when appending characters, no need to update tables
1295         d->text_.append(s);
1296
1297         // FIXME: Optimize this!
1298         for (size_t i = oldsize; i != newsize; ++i) {
1299                 // track change
1300                 d->changes_.insert(change, i);
1301         }
1302         d->fontlist_.set(oldsize, font);
1303         d->fontlist_.set(newsize - 1, font);
1304 }
1305
1306
1307 void Paragraph::insertChar(pos_type pos, char_type c,
1308                            bool trackChanges)
1309 {
1310         d->insertChar(pos, c, Change(trackChanges ?
1311                            Change::INSERTED : Change::UNCHANGED));
1312 }
1313
1314
1315 void Paragraph::insertChar(pos_type pos, char_type c,
1316                            Font const & font, bool trackChanges)
1317 {
1318         d->insertChar(pos, c, Change(trackChanges ?
1319                            Change::INSERTED : Change::UNCHANGED));
1320         setFont(pos, font);
1321 }
1322
1323
1324 void Paragraph::insertChar(pos_type pos, char_type c,
1325                            Font const & font, Change const & change)
1326 {
1327         d->insertChar(pos, c, change);
1328         setFont(pos, font);
1329 }
1330
1331
1332 bool Paragraph::insertInset(pos_type pos, Inset * inset,
1333                             Font const & font, Change const & change)
1334 {
1335         bool const success = insertInset(pos, inset, change);
1336         // Set the font/language of the inset...
1337         setFont(pos, font);
1338         return success;
1339 }
1340
1341
1342 void Paragraph::resetFonts(Font const & font)
1343 {
1344         d->fontlist_.clear();
1345         d->fontlist_.set(0, font);
1346         d->fontlist_.set(d->text_.size() - 1, font);
1347 }
1348
1349 // Gets uninstantiated font setting at position.
1350 Font const & Paragraph::getFontSettings(BufferParams const & bparams,
1351                                          pos_type pos) const
1352 {
1353         if (pos > size()) {
1354                 LYXERR0("pos: " << pos << " size: " << size());
1355                 LASSERT(pos <= size(), /**/);
1356         }
1357
1358         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
1359         if (cit != d->fontlist_.end())
1360                 return cit->font();
1361
1362         if (pos == size() && !empty())
1363                 return getFontSettings(bparams, pos - 1);
1364
1365         // Optimisation: avoid a full font instantiation if there is no
1366         // language change from previous call.
1367         static Font previous_font;
1368         static Language const * previous_lang = 0;
1369         Language const * lang = getParLanguage(bparams);
1370         if (lang != previous_lang) {
1371                 previous_lang = lang;
1372                 previous_font = Font(inherit_font, lang);
1373         }
1374         return previous_font;
1375 }
1376
1377
1378 FontSpan Paragraph::fontSpan(pos_type pos) const
1379 {
1380         LASSERT(pos <= size(), /**/);
1381         pos_type start = 0;
1382
1383         FontList::const_iterator cit = d->fontlist_.begin();
1384         FontList::const_iterator end = d->fontlist_.end();
1385         for (; cit != end; ++cit) {
1386                 if (cit->pos() >= pos) {
1387                         if (pos >= beginOfBody())
1388                                 return FontSpan(max(start, beginOfBody()),
1389                                                 cit->pos());
1390                         else
1391                                 return FontSpan(start,
1392                                                 min(beginOfBody() - 1,
1393                                                          cit->pos()));
1394                 }
1395                 start = cit->pos() + 1;
1396         }
1397
1398         // This should not happen, but if so, we take no chances.
1399         // LYXERR0("Paragraph::getEndPosOfFontSpan: This should not happen!");
1400         return FontSpan(pos, pos);
1401 }
1402
1403
1404 // Gets uninstantiated font setting at position 0
1405 Font const & Paragraph::getFirstFontSettings(BufferParams const & bparams) const
1406 {
1407         if (!empty() && !d->fontlist_.empty())
1408                 return d->fontlist_.begin()->font();
1409
1410         // Optimisation: avoid a full font instantiation if there is no
1411         // language change from previous call.
1412         static Font previous_font;
1413         static Language const * previous_lang = 0;
1414         if (bparams.language != previous_lang) {
1415                 previous_lang = bparams.language;
1416                 previous_font = Font(inherit_font, bparams.language);
1417         }
1418
1419         return previous_font;
1420 }
1421
1422
1423 // Gets the fully instantiated font at a given position in a paragraph
1424 // This is basically the same function as Text::GetFont() in text2.cpp.
1425 // The difference is that this one is used for generating the LaTeX file,
1426 // and thus cosmetic "improvements" are disallowed: This has to deliver
1427 // the true picture of the buffer. (Asger)
1428 Font const Paragraph::getFont(BufferParams const & bparams, pos_type pos,
1429                                  Font const & outerfont) const
1430 {
1431         LASSERT(pos >= 0, /**/);
1432
1433         Font font = getFontSettings(bparams, pos);
1434
1435         pos_type const body_pos = beginOfBody();
1436         FontInfo & fi = font.fontInfo();
1437         if (pos < body_pos)
1438                 fi.realize(d->layout_->labelfont);
1439         else
1440                 fi.realize(d->layout_->font);
1441
1442         fi.realize(outerfont.fontInfo());
1443         fi.realize(bparams.getFont().fontInfo());
1444
1445         return font;
1446 }
1447
1448
1449 Font const Paragraph::getLabelFont
1450         (BufferParams const & bparams, Font const & outerfont) const
1451 {
1452         FontInfo tmpfont = d->layout_->labelfont;
1453         tmpfont.realize(outerfont.fontInfo());
1454         tmpfont.realize(bparams.getFont().fontInfo());
1455         return Font(tmpfont, getParLanguage(bparams));
1456 }
1457
1458
1459 Font const Paragraph::getLayoutFont
1460         (BufferParams const & bparams, Font const & outerfont) const
1461 {
1462         FontInfo tmpfont = d->layout_->font;
1463         tmpfont.realize(outerfont.fontInfo());
1464         tmpfont.realize(bparams.getFont().fontInfo());
1465         return Font(tmpfont, getParLanguage(bparams));
1466 }
1467
1468
1469 /// Returns the height of the highest font in range
1470 FontSize Paragraph::highestFontInRange
1471         (pos_type startpos, pos_type endpos, FontSize def_size) const
1472 {
1473         return d->fontlist_.highestInRange(startpos, endpos, def_size);
1474 }
1475
1476
1477 char_type Paragraph::getUChar(BufferParams const & bparams, pos_type pos) const
1478 {
1479         char_type c = d->text_[pos];
1480         if (!lyxrc.rtl_support)
1481                 return c;
1482
1483         char_type uc = c;
1484         switch (c) {
1485         case '(':
1486                 uc = ')';
1487                 break;
1488         case ')':
1489                 uc = '(';
1490                 break;
1491         case '[':
1492                 uc = ']';
1493                 break;
1494         case ']':
1495                 uc = '[';
1496                 break;
1497         case '{':
1498                 uc = '}';
1499                 break;
1500         case '}':
1501                 uc = '{';
1502                 break;
1503         case '<':
1504                 uc = '>';
1505                 break;
1506         case '>':
1507                 uc = '<';
1508                 break;
1509         }
1510         if (uc != c && getFontSettings(bparams, pos).isRightToLeft())
1511                 return uc;
1512         return c;
1513 }
1514
1515
1516 void Paragraph::setFont(pos_type pos, Font const & font)
1517 {
1518         LASSERT(pos <= size(), /**/);
1519
1520         // First, reduce font against layout/label font
1521         // Update: The setCharFont() routine in text2.cpp already
1522         // reduces font, so we don't need to do that here. (Asger)
1523
1524         d->fontlist_.set(pos, font);
1525 }
1526
1527
1528 void Paragraph::makeSameLayout(Paragraph const & par)
1529 {
1530         d->layout_ = par.d->layout_;
1531         d->params_ = par.d->params_;
1532 }
1533
1534
1535 bool Paragraph::stripLeadingSpaces(bool trackChanges)
1536 {
1537         if (isFreeSpacing())
1538                 return false;
1539
1540         int pos = 0;
1541         int count = 0;
1542
1543         while (pos < size() && (isNewline(pos) || isLineSeparator(pos))) {
1544                 if (eraseChar(pos, trackChanges))
1545                         ++count;
1546                 else
1547                         ++pos;
1548         }
1549
1550         return count > 0 || pos > 0;
1551 }
1552
1553
1554 bool Paragraph::hasSameLayout(Paragraph const & par) const
1555 {
1556         return par.d->layout_ == d->layout_
1557                 && d->params_.sameLayout(par.d->params_);
1558 }
1559
1560
1561 depth_type Paragraph::getDepth() const
1562 {
1563         return d->params_.depth();
1564 }
1565
1566
1567 depth_type Paragraph::getMaxDepthAfter() const
1568 {
1569         if (d->layout_->isEnvironment())
1570                 return d->params_.depth() + 1;
1571         else
1572                 return d->params_.depth();
1573 }
1574
1575
1576 char Paragraph::getAlign() const
1577 {
1578         if (d->params_.align() == LYX_ALIGN_LAYOUT)
1579                 return d->layout_->align;
1580         else
1581                 return d->params_.align();
1582 }
1583
1584
1585 docstring const & Paragraph::labelString() const
1586 {
1587         return d->params_.labelString();
1588 }
1589
1590
1591 // the next two functions are for the manual labels
1592 docstring const Paragraph::getLabelWidthString() const
1593 {
1594         if (d->layout_->margintype == MARGIN_MANUAL
1595             || d->layout_->latextype == LATEX_BIB_ENVIRONMENT)
1596                 return d->params_.labelWidthString();
1597         else
1598                 return _("Senseless with this layout!");
1599 }
1600
1601
1602 void Paragraph::setLabelWidthString(docstring const & s)
1603 {
1604         d->params_.labelWidthString(s);
1605 }
1606
1607
1608 docstring const Paragraph::translateIfPossible(docstring const & s,
1609                 BufferParams const & bparams) const
1610 {
1611         if (!isAscii(s) || s.empty()) {
1612                 // This must be a user defined layout. We cannot translate
1613                 // this, since gettext accepts only ascii keys.
1614                 return s;
1615         }
1616         // Probably standard layout, try to translate
1617         Messages & m = getMessages(getParLanguage(bparams)->code());
1618         return m.get(to_ascii(s));
1619 }
1620
1621
1622 docstring Paragraph::expandLabel(Layout const & layout,
1623                 BufferParams const & bparams, bool process_appendix) const
1624 {
1625         DocumentClass const & tclass = bparams.documentClass();
1626
1627         docstring fmt;
1628         if (process_appendix && d->params_.appendix())
1629                 fmt = translateIfPossible(layout.labelstring_appendix(),
1630                         bparams);
1631         else
1632                 fmt = translateIfPossible(layout.labelstring(), bparams);
1633
1634         if (fmt.empty() && layout.labeltype == LABEL_COUNTER
1635             && !layout.counter.empty())
1636                 return tclass.counters().theCounter(layout.counter);
1637
1638         // handle 'inherited level parts' in 'fmt',
1639         // i.e. the stuff between '@' in   '@Section@.\arabic{subsection}'
1640         size_t const i = fmt.find('@', 0);
1641         if (i != docstring::npos) {
1642                 size_t const j = fmt.find('@', i + 1);
1643                 if (j != docstring::npos) {
1644                         docstring parent(fmt, i + 1, j - i - 1);
1645                         docstring label = from_ascii("??");
1646                         if (tclass.hasLayout(parent))
1647                                 docstring label = expandLabel(tclass[parent], bparams,
1648                                                       process_appendix);
1649                         fmt = docstring(fmt, 0, i) + label
1650                                 + docstring(fmt, j + 1, docstring::npos);
1651                 }
1652         }
1653
1654         return tclass.counters().counterLabel(fmt);
1655 }
1656
1657
1658 void Paragraph::applyLayout(Layout const & new_layout)
1659 {
1660         d->layout_ = &new_layout;
1661         LyXAlignment const oldAlign = d->params_.align();
1662
1663         if (!(oldAlign & d->layout_->alignpossible)) {
1664                 frontend::Alert::warning(_("Alignment not permitted"),
1665                         _("The new layout does not permit the alignment previously used.\nSetting to default."));
1666                 d->params_.align(LYX_ALIGN_LAYOUT);
1667         }
1668 }
1669
1670
1671 pos_type Paragraph::beginOfBody() const
1672 {
1673         return d->begin_of_body_;
1674 }
1675
1676
1677 void Paragraph::setBeginOfBody()
1678 {
1679         if (d->layout_->labeltype != LABEL_MANUAL) {
1680                 d->begin_of_body_ = 0;
1681                 return;
1682         }
1683
1684         // Unroll the first two cycles of the loop
1685         // and remember the previous character to
1686         // remove unnecessary getChar() calls
1687         pos_type i = 0;
1688         pos_type end = size();
1689         if (i < end && !isNewline(i)) {
1690                 ++i;
1691                 char_type previous_char = 0;
1692                 char_type temp = 0;
1693                 if (i < end) {
1694                         previous_char = d->text_[i];
1695                         if (!isNewline(i)) {
1696                                 ++i;
1697                                 while (i < end && previous_char != ' ') {
1698                                         temp = d->text_[i];
1699                                         if (isNewline(i))
1700                                                 break;
1701                                         ++i;
1702                                         previous_char = temp;
1703                                 }
1704                         }
1705                 }
1706         }
1707
1708         d->begin_of_body_ = i;
1709 }
1710
1711
1712 bool Paragraph::forcePlainLayout() const
1713 {
1714         return inInset().forcePlainLayout();
1715 }
1716
1717
1718 bool Paragraph::allowParagraphCustomization() const
1719 {
1720         return inInset().allowParagraphCustomization();
1721 }
1722
1723
1724 bool Paragraph::usePlainLayout() const
1725 {
1726         return inInset().usePlainLayout();
1727 }
1728
1729
1730 namespace {
1731
1732 // paragraphs inside floats need different alignment tags to avoid
1733 // unwanted space
1734
1735 bool noTrivlistCentering(InsetCode code)
1736 {
1737         return code == FLOAT_CODE
1738                || code == WRAP_CODE
1739                || code == CELL_CODE;
1740 }
1741
1742
1743 string correction(string const & orig)
1744 {
1745         if (orig == "flushleft")
1746                 return "raggedright";
1747         if (orig == "flushright")
1748                 return "raggedleft";
1749         if (orig == "center")
1750                 return "centering";
1751         return orig;
1752 }
1753
1754
1755 string const corrected_env(string const & suffix, string const & env,
1756         InsetCode code, bool const lastpar)
1757 {
1758         string output = suffix + "{";
1759         if (noTrivlistCentering(code)) {
1760                 if (lastpar) {
1761                         // the last paragraph in non-trivlist-aligned
1762                         // context is special (to avoid unwanted whitespace)
1763                         if (suffix == "\\begin")
1764                                 return "\\" + correction(env) + "{}";
1765                         return string();
1766                 }
1767                 output += correction(env);
1768         } else
1769                 output += env;
1770         output += "}";
1771         if (suffix == "\\begin")
1772                 output += "\n";
1773         return output;
1774 }
1775
1776
1777 void adjust_row_column(string const & str, TexRow & texrow, int & column)
1778 {
1779         if (!contains(str, "\n"))
1780                 column += str.size();
1781         else {
1782                 string tmp;
1783                 texrow.newline();
1784                 column = rsplit(str, tmp, '\n').size();
1785         }
1786 }
1787
1788 } // namespace anon
1789
1790
1791 int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
1792                                  odocstream & os, TexRow & texrow,
1793                                  OutputParams const & runparams) const
1794 {
1795         int column = 0;
1796
1797         if (params_.noindent()) {
1798                 os << "\\noindent ";
1799                 column += 10;
1800         }
1801
1802         LyXAlignment const curAlign = params_.align();
1803
1804         if (curAlign == layout_->align)
1805                 return column;
1806
1807         switch (curAlign) {
1808         case LYX_ALIGN_NONE:
1809         case LYX_ALIGN_BLOCK:
1810         case LYX_ALIGN_LAYOUT:
1811         case LYX_ALIGN_SPECIAL:
1812                 break;
1813         case LYX_ALIGN_LEFT:
1814         case LYX_ALIGN_RIGHT:
1815         case LYX_ALIGN_CENTER:
1816                 if (runparams.moving_arg) {
1817                         os << "\\protect";
1818                         column += 8;
1819                 }
1820                 break;
1821         }
1822
1823         string const begin_tag = "\\begin";
1824         InsetCode code = owner_->ownerCode();
1825         bool const lastpar = runparams.isLastPar;
1826
1827         switch (curAlign) {
1828         case LYX_ALIGN_NONE:
1829         case LYX_ALIGN_BLOCK:
1830         case LYX_ALIGN_LAYOUT:
1831         case LYX_ALIGN_SPECIAL:
1832                 break;
1833         case LYX_ALIGN_LEFT: {
1834                 string output;
1835                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
1836                         output = corrected_env(begin_tag, "flushleft", code, lastpar);
1837                 else
1838                         output = corrected_env(begin_tag, "flushright", code, lastpar);
1839                 os << from_ascii(output);
1840                 adjust_row_column(output, texrow, column);
1841                 break;
1842         } case LYX_ALIGN_RIGHT: {
1843                 string output;
1844                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
1845                         output = corrected_env(begin_tag, "flushright", code, lastpar);
1846                 else
1847                         output = corrected_env(begin_tag, "flushleft", code, lastpar);
1848                 os << from_ascii(output);
1849                 adjust_row_column(output, texrow, column);
1850                 break;
1851         } case LYX_ALIGN_CENTER: {
1852                 string output;
1853                 output = corrected_env(begin_tag, "center", code, lastpar);
1854                 os << from_ascii(output);
1855                 adjust_row_column(output, texrow, column);
1856                 break;
1857         }
1858         }
1859
1860         return column;
1861 }
1862
1863
1864 int Paragraph::Private::endTeXParParams(BufferParams const & bparams,
1865                                odocstream & os, TexRow & texrow,
1866                                OutputParams const & runparams) const
1867 {
1868         int column = 0;
1869
1870         LyXAlignment const curAlign = params_.align();
1871
1872         if (curAlign == layout_->align)
1873                 return column;
1874
1875         switch (curAlign) {
1876         case LYX_ALIGN_NONE:
1877         case LYX_ALIGN_BLOCK:
1878         case LYX_ALIGN_LAYOUT:
1879         case LYX_ALIGN_SPECIAL:
1880                 break;
1881         case LYX_ALIGN_LEFT:
1882         case LYX_ALIGN_RIGHT:
1883         case LYX_ALIGN_CENTER:
1884                 if (runparams.moving_arg) {
1885                         os << "\\protect";
1886                         column = 8;
1887                 }
1888                 break;
1889         }
1890
1891         string const end_tag = "\n\\par\\end";
1892         InsetCode code = owner_->ownerCode();
1893         bool const lastpar = runparams.isLastPar;
1894
1895         switch (curAlign) {
1896         case LYX_ALIGN_NONE:
1897         case LYX_ALIGN_BLOCK:
1898         case LYX_ALIGN_LAYOUT:
1899         case LYX_ALIGN_SPECIAL:
1900                 break;
1901         case LYX_ALIGN_LEFT: {
1902                 string output;
1903                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
1904                         output = corrected_env(end_tag, "flushleft", code, lastpar);
1905                 else
1906                         output = corrected_env(end_tag, "flushright", code, lastpar);
1907                 os << from_ascii(output);
1908                 adjust_row_column(output, texrow, column);
1909                 break;
1910         } case LYX_ALIGN_RIGHT: {
1911                 string output;
1912                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
1913                         output = corrected_env(end_tag, "flushright", code, lastpar);
1914                 else
1915                         output = corrected_env(end_tag, "flushleft", code, lastpar);
1916                 os << from_ascii(output);
1917                 adjust_row_column(output, texrow, column);
1918                 break;
1919         } case LYX_ALIGN_CENTER: {
1920                 string output;
1921                 output = corrected_env(end_tag, "center", code, lastpar);
1922                 os << from_ascii(output);
1923                 adjust_row_column(output, texrow, column);
1924                 break;
1925         }
1926         }
1927
1928         return column;
1929 }
1930
1931
1932 // This one spits out the text of the paragraph
1933 bool Paragraph::latex(BufferParams const & bparams,
1934         Font const & outerfont,
1935         odocstream & os, TexRow & texrow,
1936         OutputParams const & runparams,
1937         int start_pos, int end_pos) const
1938 {
1939         LYXERR(Debug::LATEX, "Paragraph::latex...     " << this);
1940
1941         bool return_value = false;
1942
1943         bool const allowcust = allowParagraphCustomization();
1944
1945         // FIXME This check should not be needed. Perhaps issue an
1946         // error if it triggers.
1947         Layout const & style = forcePlainLayout() ?
1948                 bparams.documentClass().plainLayout() : *d->layout_;
1949
1950         // Current base font for all inherited font changes, without any
1951         // change caused by an individual character, except for the language:
1952         // It is set to the language of the first character.
1953         // As long as we are in the label, this font is the base font of the
1954         // label. Before the first body character it is set to the base font
1955         // of the body.
1956         Font basefont;
1957
1958         // Maybe we have to create a optional argument.
1959         pos_type body_pos = beginOfBody();
1960         unsigned int column = 0;
1961
1962         if (body_pos > 0) {
1963                 os << '[';
1964                 column += 1;
1965                 basefont = getLabelFont(bparams, outerfont);
1966         } else {
1967                 basefont = getLayoutFont(bparams, outerfont);
1968         }
1969
1970         // Which font is currently active?
1971         Font running_font(basefont);
1972         // Do we have an open font change?
1973         bool open_font = false;
1974
1975         Change runningChange = Change(Change::UNCHANGED);
1976
1977         texrow.start(id(), 0);
1978
1979         // if the paragraph is empty, the loop will not be entered at all
1980         if (empty()) {
1981                 if (style.isCommand()) {
1982                         os << '{';
1983                         ++column;
1984                 }
1985                 if (allowcust)
1986                         column += d->startTeXParParams(bparams, os, texrow,
1987                                                     runparams);
1988         }
1989
1990         for (pos_type i = 0; i < size(); ++i) {
1991                 // First char in paragraph or after label?
1992                 if (i == body_pos) {
1993                         if (body_pos > 0) {
1994                                 if (open_font) {
1995                                         column += running_font.latexWriteEndChanges(
1996                                                 os, bparams, runparams,
1997                                                 basefont, basefont);
1998                                         open_font = false;
1999                                 }
2000                                 basefont = getLayoutFont(bparams, outerfont);
2001                                 running_font = basefont;
2002
2003                                 column += Changes::latexMarkChange(os, bparams,
2004                                                 runningChange, Change(Change::UNCHANGED));
2005                                 runningChange = Change(Change::UNCHANGED);
2006
2007                                 os << "] ";
2008                                 column +=2;
2009                         }
2010                         if (style.isCommand()) {
2011                                 os << '{';
2012                                 ++column;
2013                         }
2014
2015                         if (allowcust)
2016                                 column += d->startTeXParParams(bparams, os,
2017                                                             texrow,
2018                                                             runparams);
2019                 }
2020
2021                 Change const & change = runparams.inDeletedInset ? runparams.changeOfDeletedInset
2022                                                                  : lookupChange(i);
2023
2024                 if (bparams.outputChanges && runningChange != change) {
2025                         if (open_font) {
2026                                 column += running_font.latexWriteEndChanges(
2027                                                 os, bparams, runparams, basefont, basefont);
2028                                 open_font = false;
2029                         }
2030                         basefont = getLayoutFont(bparams, outerfont);
2031                         running_font = basefont;
2032
2033                         column += Changes::latexMarkChange(os, bparams, runningChange, change);
2034                         runningChange = change;
2035                 }
2036
2037                 // do not output text which is marked deleted
2038                 // if change tracking output is disabled
2039                 if (!bparams.outputChanges && change.type == Change::DELETED) {
2040                         continue;
2041                 }
2042
2043                 ++column;
2044
2045                 // Fully instantiated font
2046                 Font const font = getFont(bparams, i, outerfont);
2047
2048                 Font const last_font = running_font;
2049
2050                 // Do we need to close the previous font?
2051                 if (open_font &&
2052                     (font != running_font ||
2053                      font.language() != running_font.language()))
2054                 {
2055                         column += running_font.latexWriteEndChanges(
2056                                         os, bparams, runparams, basefont,
2057                                         (i == body_pos-1) ? basefont : font);
2058                         running_font = basefont;
2059                         open_font = false;
2060                 }
2061
2062                 // close babel's font environment before opening CJK.
2063                 if (!running_font.language()->babel().empty() &&
2064                     font.language()->encoding()->package() == Encoding::CJK) {
2065                                 string end_tag = subst(lyxrc.language_command_end,
2066                                                         "$$lang",
2067                                                         running_font.language()->babel());
2068                                 os << from_ascii(end_tag);
2069                                 column += end_tag.length();
2070                 }
2071
2072                 // Switch file encoding if necessary (and allowed)
2073                 if (!runparams.verbatim &&
2074                     runparams.encoding->package() != Encoding::none &&
2075                     font.language()->encoding()->package() != Encoding::none) {
2076                         pair<bool, int> const enc_switch = switchEncoding(os, bparams,
2077                                         runparams, *(font.language()->encoding()));
2078                         if (enc_switch.first) {
2079                                 column += enc_switch.second;
2080                                 runparams.encoding = font.language()->encoding();
2081                         }
2082                 }
2083
2084                 char_type const c = d->text_[i];
2085
2086                 // Do we need to change font?
2087                 if ((font != running_font ||
2088                      font.language() != running_font.language()) &&
2089                         i != body_pos - 1)
2090                 {
2091                         odocstringstream ods;
2092                         column += font.latexWriteStartChanges(ods, bparams,
2093                                                               runparams, basefont,
2094                                                               last_font);
2095                         running_font = font;
2096                         open_font = true;
2097                         docstring fontchange = ods.str();
2098                         // check if the fontchange ends with a trailing blank
2099                         // (like "\small " (see bug 3382)
2100                         if (suffixIs(fontchange, ' ') && c == ' ')
2101                                 os << fontchange.substr(0, fontchange.size() - 1)
2102                                    << from_ascii("{}");
2103                         else
2104                                 os << fontchange;
2105                 }
2106
2107                 // FIXME: think about end_pos implementation...
2108                 if (c == ' ' && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2109                         // FIXME: integrate this case in latexSpecialChar
2110                         // Do not print the separation of the optional argument
2111                         // if style.pass_thru is false. This works because
2112                         // latexSpecialChar ignores spaces if
2113                         // style.pass_thru is false.
2114                         if (i != body_pos - 1) {
2115                                 if (d->simpleTeXBlanks(
2116                                                 runparams, os, texrow,
2117                                                 i, column, font, style)) {
2118                                         // A surrogate pair was output. We
2119                                         // must not call latexSpecialChar
2120                                         // in this iteration, since it would output
2121                                         // the combining character again.
2122                                         ++i;
2123                                         continue;
2124                                 }
2125                         }
2126                 }
2127
2128                 OutputParams rp = runparams;
2129                 rp.free_spacing = style.free_spacing;
2130                 rp.local_font = &font;
2131                 rp.intitle = style.intitle;
2132
2133                 // Two major modes:  LaTeX or plain
2134                 // Handle here those cases common to both modes
2135                 // and then split to handle the two modes separately.
2136                 if (c == META_INSET) {
2137                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2138                                 d->latexInset(bparams, os,
2139                                                 texrow, rp, running_font,
2140                                                 basefont, outerfont, open_font,
2141                                                 runningChange, style, i, column);
2142                         }
2143                 } else {
2144                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2145                                 try {
2146                                         d->latexSpecialChar(os, rp, running_font, runningChange,
2147                                                 style, i, column);
2148                                 } catch (EncodingException & e) {
2149                                 if (runparams.dryrun) {
2150                                         os << "<" << _("LyX Warning: ")
2151                                            << _("uncodable character") << " '";
2152                                         os.put(c);
2153                                         os << "'>";
2154                                 } else {
2155                                         // add location information and throw again.
2156                                         e.par_id = id();
2157                                         e.pos = i;
2158                                         throw(e);
2159                                 }
2160                         }
2161                 }
2162                 }
2163
2164                 // Set the encoding to that returned from latexSpecialChar (see
2165                 // comment for encoding member in OutputParams.h)
2166                 runparams.encoding = rp.encoding;
2167         }
2168
2169         // If we have an open font definition, we have to close it
2170         if (open_font) {
2171 #ifdef FIXED_LANGUAGE_END_DETECTION
2172                 if (next_) {
2173                         running_font
2174                                 .latexWriteEndChanges(os, bparams, runparams,
2175                                         basefont,
2176                                         next_->getFont(bparams, 0, outerfont));
2177                 } else {
2178                         running_font.latexWriteEndChanges(os, bparams,
2179                                         runparams, basefont, basefont);
2180                 }
2181 #else
2182 //FIXME: For now we ALWAYS have to close the foreign font settings if they are
2183 //FIXME: there as we start another \selectlanguage with the next paragraph if
2184 //FIXME: we are in need of this. This should be fixed sometime (Jug)
2185                 running_font.latexWriteEndChanges(os, bparams, runparams,
2186                                 basefont, basefont);
2187 #endif
2188         }
2189
2190         column += Changes::latexMarkChange(os, bparams, runningChange, Change(Change::UNCHANGED));
2191
2192         // Needed if there is an optional argument but no contents.
2193         if (body_pos > 0 && body_pos == size()) {
2194                 os << "]~";
2195                 return_value = false;
2196         }
2197
2198         if (allowcust) {
2199                 column += d->endTeXParParams(bparams, os, texrow,
2200                                           runparams);
2201         }
2202
2203         LYXERR(Debug::LATEX, "Paragraph::latex... done " << this);
2204         return return_value;
2205 }
2206
2207
2208 bool Paragraph::emptyTag() const
2209 {
2210         for (pos_type i = 0; i < size(); ++i) {
2211                 if (Inset const * inset = getInset(i)) {
2212                         InsetCode lyx_code = inset->lyxCode();
2213                         if (lyx_code != TOC_CODE &&
2214                             lyx_code != INCLUDE_CODE &&
2215                             lyx_code != GRAPHICS_CODE &&
2216                             lyx_code != ERT_CODE &&
2217                             lyx_code != LISTINGS_CODE &&
2218                             lyx_code != FLOAT_CODE &&
2219                             lyx_code != TABULAR_CODE) {
2220                                 return false;
2221                         }
2222                 } else {
2223                         char_type c = d->text_[i];
2224                         if (c != ' ' && c != '\t')
2225                                 return false;
2226                 }
2227         }
2228         return true;
2229 }
2230
2231
2232 string Paragraph::getID(Buffer const & buf, OutputParams const & runparams)
2233         const
2234 {
2235         for (pos_type i = 0; i < size(); ++i) {
2236                 if (Inset const * inset = getInset(i)) {
2237                         InsetCode lyx_code = inset->lyxCode();
2238                         if (lyx_code == LABEL_CODE) {
2239                                 InsetLabel const * const il = static_cast<InsetLabel const *>(inset);
2240                                 docstring const & id = il->getParam("name");
2241                                 return "id='" + to_utf8(sgml::cleanID(buf, runparams, id)) + "'";
2242                         }
2243                 }
2244         }
2245         return string();
2246 }
2247
2248
2249 pos_type Paragraph::firstWordDocBook(odocstream & os, OutputParams const & runparams)
2250         const
2251 {
2252         pos_type i;
2253         for (i = 0; i < size(); ++i) {
2254                 if (Inset const * inset = getInset(i)) {
2255                         inset->docbook(os, runparams);
2256                 } else {
2257                         char_type c = d->text_[i];
2258                         if (c == ' ')
2259                                 break;
2260                         os << sgml::escapeChar(c);
2261                 }
2262         }
2263         return i;
2264 }
2265
2266
2267 pos_type Paragraph::firstWordLyXHTML(odocstream & os, OutputParams const & runparams)
2268         const
2269 {
2270         pos_type i;
2271         for (i = 0; i < size(); ++i) {
2272                 if (Inset const * inset = getInset(i)) {
2273                         inset->xhtml(os, runparams);
2274                 } else {
2275                         char_type c = d->text_[i];
2276                         if (c == ' ')
2277                                 break;
2278                         os << html::escapeChar(c);
2279                 }
2280         }
2281         return i;
2282 }
2283
2284
2285 bool Paragraph::Private::onlyText(Buffer const & buf, Font const & outerfont, pos_type initial) const
2286 {
2287         Font font_old;
2288         pos_type size = text_.size();
2289         for (pos_type i = initial; i < size; ++i) {
2290                 Font font = owner_->getFont(buf.params(), i, outerfont);
2291                 if (text_[i] == META_INSET)
2292                         return false;
2293                 if (i != initial && font != font_old)
2294                         return false;
2295                 font_old = font;
2296         }
2297
2298         return true;
2299 }
2300
2301
2302 void Paragraph::simpleDocBookOnePar(Buffer const & buf,
2303                                     odocstream & os,
2304                                     OutputParams const & runparams,
2305                                     Font const & outerfont,
2306                                     pos_type initial) const
2307 {
2308         bool emph_flag = false;
2309
2310         Layout const & style = *d->layout_;
2311         FontInfo font_old =
2312                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2313
2314         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2315                 os << "]]>";
2316
2317         // parsing main loop
2318         for (pos_type i = initial; i < size(); ++i) {
2319                 Font font = getFont(buf.params(), i, outerfont);
2320
2321                 // handle <emphasis> tag
2322                 if (font_old.emph() != font.fontInfo().emph()) {
2323                         if (font.fontInfo().emph() == FONT_ON) {
2324                                 os << "<emphasis>";
2325                                 emph_flag = true;
2326                         } else if (i != initial) {
2327                                 os << "</emphasis>";
2328                                 emph_flag = false;
2329                         }
2330                 }
2331
2332                 if (Inset const * inset = getInset(i)) {
2333                         inset->docbook(os, runparams);
2334                 } else {
2335                         char_type c = d->text_[i];
2336
2337                         if (style.pass_thru)
2338                                 os.put(c);
2339                         else
2340                                 os << sgml::escapeChar(c);
2341                 }
2342                 font_old = font.fontInfo();
2343         }
2344
2345         if (emph_flag) {
2346                 os << "</emphasis>";
2347         }
2348
2349         if (style.free_spacing)
2350                 os << '\n';
2351         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2352                 os << "<![CDATA[";
2353 }
2354
2355
2356 docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
2357                                     odocstream & os,
2358                                     OutputParams const & runparams,
2359                                     Font const & outerfont,
2360                                     pos_type initial) const
2361 {
2362         docstring retval;
2363
2364         // FIXME We really need to manage the tag nesting here.
2365         // Probably in the same sort of way as in output_xhtml.
2366         bool emph_flag = false;
2367         bool bold_flag = false;
2368         std::string closing_tag;
2369
2370         Layout const & style = *d->layout_;
2371         FontInfo font_old =
2372                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2373
2374         //if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2375         //      os << "]]>";
2376
2377         // parsing main loop
2378         for (pos_type i = initial; i < size(); ++i) {
2379                 Font font = getFont(buf.params(), i, outerfont);
2380
2381                 // emphasis
2382                 if (font_old.emph() != font.fontInfo().emph()) {
2383                         if (font.fontInfo().emph() == FONT_ON) {
2384                                 os << "<em>";
2385                                 emph_flag = true;
2386                         } else if (emph_flag && i != initial) {
2387                                 os << "</em>";
2388                                 emph_flag = false;
2389                         }
2390                 }
2391                 // bold
2392                 if (font_old.series() != font.fontInfo().series()) {
2393                         if (font.fontInfo().series() == BOLD_SERIES) {
2394                                 os << "<strong>";
2395                                 bold_flag = true;
2396                         } else if (bold_flag && i != initial) {
2397                                 os << "</strong>";
2398                                 bold_flag = false;
2399                         }
2400                 }
2401                 // FIXME Other such tags?
2402
2403                 if (Inset const * inset = getInset(i)) {
2404                         retval += inset->xhtml(os, runparams);
2405                 } else {
2406                         char_type c = d->text_[i];
2407
2408                         if (style.pass_thru)
2409                                 os.put(c);
2410                         else if (c == '-') {
2411                                 docstring str;
2412                                 int j = i + 1;
2413                                 if (j < size() && d->text_[j] == '-') {
2414                                         j += 1;
2415                                         if (j < size() && d->text_[j] == '-') {
2416                                                 str += from_ascii("&mdash;");
2417                                                 i += 2;
2418                                         } else {
2419                                                 str += from_ascii("&ndash;");
2420                                                 i += 1;
2421                                         }
2422                                 }
2423                                 else
2424                                         str += c;
2425                                 os << str;
2426                         } else
2427                                 os << html::escapeChar(c);
2428                 }
2429                 font_old = font.fontInfo();
2430         }
2431
2432         // FIXME This could be out of order. See above.
2433         if (emph_flag)
2434                 os << "</em>";
2435         if (bold_flag)
2436                 os << "</strong>";
2437
2438         return retval;
2439 }
2440
2441
2442 bool Paragraph::isHfill(pos_type pos) const
2443 {
2444         Inset const * inset = getInset(pos);
2445         return inset && (inset->lyxCode() == SPACE_CODE &&
2446                          inset->isStretchableSpace());
2447 }
2448
2449
2450 bool Paragraph::isNewline(pos_type pos) const
2451 {
2452         Inset const * inset = getInset(pos);
2453         return inset && inset->lyxCode() == NEWLINE_CODE;
2454 }
2455
2456
2457 bool Paragraph::isLineSeparator(pos_type pos) const
2458 {
2459         char_type const c = d->text_[pos];
2460         if (isLineSeparatorChar(c))
2461                 return true;
2462         Inset const * inset = getInset(pos);
2463         return inset && inset->isLineSeparator();
2464 }
2465
2466
2467 bool Paragraph::isWordSeparator(pos_type pos) const
2468 {
2469         if (Inset const * inset = getInset(pos))
2470                 return !inset->isLetter();
2471         char_type const c = d->text_[pos];
2472     // We want to pass the ' and escape chars to the spellchecker
2473         static docstring const quote = from_utf8(lyxrc.spellchecker_esc_chars + '\'');
2474         return (!isLetterChar(c) && !isDigit(c) && !contains(quote, c))
2475                 || pos == size();
2476 }
2477
2478
2479 bool Paragraph::isChar(pos_type pos) const
2480 {
2481         if (Inset const * inset = getInset(pos))
2482                 return inset->isChar();
2483         char_type const c = d->text_[pos];
2484         return !isLetterChar(c) && !isDigit(c) && !lyx::isSpace(c);
2485 }
2486
2487
2488 bool Paragraph::isSpace(pos_type pos) const
2489 {
2490         if (Inset const * inset = getInset(pos))
2491                 return inset->isSpace();
2492         char_type const c = d->text_[pos];
2493         return lyx::isSpace(c);
2494 }
2495
2496
2497 Language const *
2498 Paragraph::getParLanguage(BufferParams const & bparams) const
2499 {
2500         if (!empty())
2501                 return getFirstFontSettings(bparams).language();
2502         // FIXME: we should check the prev par as well (Lgb)
2503         return bparams.language;
2504 }
2505
2506
2507 bool Paragraph::isRTL(BufferParams const & bparams) const
2508 {
2509         return lyxrc.rtl_support
2510                 && getParLanguage(bparams)->rightToLeft()
2511                 && ownerCode() != ERT_CODE
2512                 && ownerCode() != LISTINGS_CODE;
2513 }
2514
2515
2516 void Paragraph::changeLanguage(BufferParams const & bparams,
2517                                Language const * from, Language const * to)
2518 {
2519         // change language including dummy font change at the end
2520         for (pos_type i = 0; i <= size(); ++i) {
2521                 Font font = getFontSettings(bparams, i);
2522                 if (font.language() == from) {
2523                         font.setLanguage(to);
2524                         setFont(i, font);
2525                 }
2526         }
2527 }
2528
2529
2530 bool Paragraph::isMultiLingual(BufferParams const & bparams) const
2531 {
2532         Language const * doc_language = bparams.language;
2533         FontList::const_iterator cit = d->fontlist_.begin();
2534         FontList::const_iterator end = d->fontlist_.end();
2535
2536         for (; cit != end; ++cit)
2537                 if (cit->font().language() != ignore_language &&
2538                     cit->font().language() != latex_language &&
2539                     cit->font().language() != doc_language)
2540                         return true;
2541         return false;
2542 }
2543
2544
2545 docstring Paragraph::asString(int options) const
2546 {
2547         return asString(0, size(), options);
2548 }
2549
2550
2551 docstring Paragraph::asString(pos_type beg, pos_type end, int options) const
2552 {
2553         odocstringstream os;
2554
2555         if (beg == 0
2556                 && options & AS_STR_LABEL
2557                 && !d->params_.labelString().empty())
2558                 os << d->params_.labelString() << ' ';
2559
2560         for (pos_type i = beg; i < end; ++i) {
2561                 char_type const c = d->text_[i];
2562                 if (isPrintable(c) || c == '\t'
2563                     || (c == '\n' && options & AS_STR_NEWLINES))
2564                         os.put(c);
2565                 else if (c == META_INSET && options & AS_STR_INSETS)
2566                         getInset(i)->tocString(os);
2567         }
2568
2569         return os.str();
2570 }
2571
2572
2573 docstring Paragraph::stringify(pos_type beg, pos_type end, int options, OutputParams & runparams) const
2574 {
2575         odocstringstream os;
2576
2577         if (beg == 0
2578                 && options & AS_STR_LABEL
2579                 && !d->params_.labelString().empty())
2580                 os << d->params_.labelString() << ' ';
2581
2582         for (pos_type i = beg; i < end; ++i) {
2583                 char_type const c = d->text_[i];
2584                 if (isPrintable(c) || c == '\t'
2585                     || (c == '\n' && options & AS_STR_NEWLINES))
2586                         os.put(c);
2587                 else if (c == META_INSET && options & AS_STR_INSETS) {
2588                         getInset(i)->plaintext(os, runparams);
2589                 }
2590         }
2591
2592         return os.str();
2593 }
2594
2595
2596 void Paragraph::setInsetOwner(Inset const * inset)
2597 {
2598         d->inset_owner_ = inset;
2599 }
2600
2601
2602 int Paragraph::id() const
2603 {
2604         return d->id_;
2605 }
2606
2607
2608 Layout const & Paragraph::layout() const
2609 {
2610         return *d->layout_;
2611 }
2612
2613
2614 void Paragraph::setLayout(Layout const & layout)
2615 {
2616         d->layout_ = &layout;
2617 }
2618
2619
2620 void Paragraph::setDefaultLayout(DocumentClass const & tc)
2621 {
2622         setLayout(tc.defaultLayout());
2623 }
2624
2625
2626 void Paragraph::setPlainLayout(DocumentClass const & tc)
2627 {
2628         setLayout(tc.plainLayout());
2629 }
2630
2631
2632 void Paragraph::setPlainOrDefaultLayout(DocumentClass const & tclass)
2633 {
2634         if (usePlainLayout())
2635                 setPlainLayout(tclass);
2636         else
2637                 setDefaultLayout(tclass);
2638 }
2639
2640
2641 Inset const & Paragraph::inInset() const
2642 {
2643         LASSERT(d->inset_owner_, throw ExceptionMessage(BufferException,
2644                 _("Memory problem"), _("Paragraph not properly initialized")));
2645         return *d->inset_owner_;
2646 }
2647
2648
2649 InsetCode Paragraph::ownerCode() const
2650 {
2651         return d->inset_owner_ ? d->inset_owner_->lyxCode() : NO_CODE;
2652 }
2653
2654
2655 ParagraphParameters & Paragraph::params()
2656 {
2657         return d->params_;
2658 }
2659
2660
2661 ParagraphParameters const & Paragraph::params() const
2662 {
2663         return d->params_;
2664 }
2665
2666
2667 bool Paragraph::isFreeSpacing() const
2668 {
2669         if (d->layout_->free_spacing)
2670                 return true;
2671         return d->inset_owner_ && d->inset_owner_->isFreeSpacing();
2672 }
2673
2674
2675 bool Paragraph::allowEmpty() const
2676 {
2677         if (d->layout_->keepempty)
2678                 return true;
2679         return d->inset_owner_ && d->inset_owner_->allowEmpty();
2680 }
2681
2682
2683 char_type Paragraph::transformChar(char_type c, pos_type pos) const
2684 {
2685         if (!Encodings::isArabicChar(c))
2686                 return c;
2687
2688         char_type prev_char = ' ';
2689         char_type next_char = ' ';
2690
2691         for (pos_type i = pos - 1; i >= 0; --i) {
2692                 char_type const par_char = d->text_[i];
2693                 if (!Encodings::isArabicComposeChar(par_char)) {
2694                         prev_char = par_char;
2695                         break;
2696                 }
2697         }
2698
2699         for (pos_type i = pos + 1, end = size(); i < end; ++i) {
2700                 char_type const par_char = d->text_[i];
2701                 if (!Encodings::isArabicComposeChar(par_char)) {
2702                         next_char = par_char;
2703                         break;
2704                 }
2705         }
2706
2707         if (Encodings::isArabicChar(next_char)) {
2708                 if (Encodings::isArabicChar(prev_char) &&
2709                         !Encodings::isArabicSpecialChar(prev_char))
2710                         return Encodings::transformChar(c, Encodings::FORM_MEDIAL);
2711                 else
2712                         return Encodings::transformChar(c, Encodings::FORM_INITIAL);
2713         } else {
2714                 if (Encodings::isArabicChar(prev_char) &&
2715                         !Encodings::isArabicSpecialChar(prev_char))
2716                         return Encodings::transformChar(c, Encodings::FORM_FINAL);
2717                 else
2718                         return Encodings::transformChar(c, Encodings::FORM_ISOLATED);
2719         }
2720 }
2721
2722
2723 int Paragraph::checkBiblio(Buffer const & buffer)
2724 {
2725         // FIXME From JS:
2726         // This is getting more and more a mess. ...We really should clean
2727         // up this bibitem issue for 1.6. See also bug 2743.
2728
2729         // Add bibitem insets if necessary
2730         if (d->layout_->labeltype != LABEL_BIBLIO)
2731                 return 0;
2732
2733         bool hasbibitem = !d->insetlist_.empty()
2734                 // Insist on it being in pos 0
2735                 && d->text_[0] == META_INSET
2736                 && d->insetlist_.begin()->inset->lyxCode() == BIBITEM_CODE;
2737
2738         bool track_changes = buffer.params().trackChanges;
2739
2740         docstring oldkey;
2741         docstring oldlabel;
2742
2743         // remove a bibitem in pos != 0
2744         // restore it later in pos 0 if necessary
2745         // (e.g. if a user inserts contents _before_ the item)
2746         // we're assuming there's only one of these, which there
2747         // should be.
2748         int erasedInsetPosition = -1;
2749         InsetList::iterator it = d->insetlist_.begin();
2750         InsetList::iterator end = d->insetlist_.end();
2751         for (; it != end; ++it)
2752                 if (it->inset->lyxCode() == BIBITEM_CODE
2753                     && it->pos > 0) {
2754                         InsetBibitem * olditem = static_cast<InsetBibitem *>(it->inset);
2755                         oldkey = olditem->getParam("key");
2756                         oldlabel = olditem->getParam("label");
2757                         erasedInsetPosition = it->pos;
2758                         eraseChar(erasedInsetPosition, track_changes);
2759                         break;
2760         }
2761
2762         // There was an InsetBibitem at the beginning, and we didn't
2763         // have to erase one.
2764         if (hasbibitem && erasedInsetPosition < 0)
2765                         return 0;
2766
2767         // There was an InsetBibitem at the beginning and we did have to
2768         // erase one. So we give its properties to the beginning inset.
2769         if (hasbibitem) {
2770                 InsetBibitem * inset =
2771                         static_cast<InsetBibitem *>(d->insetlist_.begin()->inset);
2772                 if (!oldkey.empty())
2773                         inset->setParam("key", oldkey);
2774                 inset->setParam("label", oldlabel);
2775                 return -erasedInsetPosition;
2776         }
2777
2778         // There was no inset at the beginning, so we need to create one with
2779         // the key and label of the one we erased.
2780         InsetBibitem * inset =
2781                 new InsetBibitem(buffer, InsetCommandParams(BIBITEM_CODE));
2782         // restore values of previously deleted item in this par.
2783         if (!oldkey.empty())
2784                 inset->setParam("key", oldkey);
2785         inset->setParam("label", oldlabel);
2786         insertInset(0, static_cast<Inset *>(inset),
2787                     Change(track_changes ? Change::INSERTED : Change::UNCHANGED));
2788
2789         return 1;
2790 }
2791
2792
2793 void Paragraph::checkAuthors(AuthorList const & authorList)
2794 {
2795         d->changes_.checkAuthors(authorList);
2796 }
2797
2798
2799 bool Paragraph::isUnchanged(pos_type pos) const
2800 {
2801         return lookupChange(pos).type == Change::UNCHANGED;
2802 }
2803
2804
2805 bool Paragraph::isInserted(pos_type pos) const
2806 {
2807         return lookupChange(pos).type == Change::INSERTED;
2808 }
2809
2810
2811 bool Paragraph::isDeleted(pos_type pos) const
2812 {
2813         return lookupChange(pos).type == Change::DELETED;
2814 }
2815
2816
2817 InsetList const & Paragraph::insetList() const
2818 {
2819         return d->insetlist_;
2820 }
2821
2822
2823 void Paragraph::setBuffer(Buffer & b)
2824 {
2825         d->insetlist_.setBuffer(b);
2826 }
2827
2828
2829 Inset * Paragraph::releaseInset(pos_type pos)
2830 {
2831         Inset * inset = d->insetlist_.release(pos);
2832         /// does not honour change tracking!
2833         eraseChar(pos, false);
2834         return inset;
2835 }
2836
2837
2838 Inset * Paragraph::getInset(pos_type pos)
2839 {
2840         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
2841                  ? d->insetlist_.get(pos) : 0;
2842 }
2843
2844
2845 Inset const * Paragraph::getInset(pos_type pos) const
2846 {
2847         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
2848                  ? d->insetlist_.get(pos) : 0;
2849 }
2850
2851
2852 void Paragraph::changeCase(BufferParams const & bparams, pos_type pos,
2853                 pos_type & right, TextCase action)
2854 {
2855         // process sequences of modified characters; in change
2856         // tracking mode, this approach results in much better
2857         // usability than changing case on a char-by-char basis
2858         docstring changes;
2859
2860         bool const trackChanges = bparams.trackChanges;
2861
2862         bool capitalize = true;
2863
2864         for (; pos < right; ++pos) {
2865                 char_type oldChar = d->text_[pos];
2866                 char_type newChar = oldChar;
2867
2868                 // ignore insets and don't play with deleted text!
2869                 if (oldChar != META_INSET && !isDeleted(pos)) {
2870                         switch (action) {
2871                                 case text_lowercase:
2872                                         newChar = lowercase(oldChar);
2873                                         break;
2874                                 case text_capitalization:
2875                                         if (capitalize) {
2876                                                 newChar = uppercase(oldChar);
2877                                                 capitalize = false;
2878                                         }
2879                                         break;
2880                                 case text_uppercase:
2881                                         newChar = uppercase(oldChar);
2882                                         break;
2883                         }
2884                 }
2885
2886                 if (isWordSeparator(pos) || isDeleted(pos)) {
2887                         // permit capitalization again
2888                         capitalize = true;
2889                 }
2890
2891                 if (oldChar != newChar)
2892                         changes += newChar;
2893
2894                 if (oldChar == newChar || pos == right - 1) {
2895                         if (oldChar != newChar) {
2896                                 // step behind the changing area
2897                                 pos++;
2898                         }
2899                         int erasePos = pos - changes.size();
2900                         for (size_t i = 0; i < changes.size(); i++) {
2901                                 insertChar(pos, changes[i],
2902                                         getFontSettings(bparams,
2903                                         erasePos),
2904                                         trackChanges);
2905                                 if (!eraseChar(erasePos, trackChanges)) {
2906                                         ++erasePos;
2907                                         ++pos; // advance
2908                                         ++right; // expand selection
2909                                 }
2910                         }
2911                         changes.clear();
2912                 }
2913         }
2914 }
2915
2916
2917 bool Paragraph::find(docstring const & str, bool cs, bool mw,
2918                 pos_type pos, bool del) const
2919 {
2920         int const strsize = str.length();
2921         int i = 0;
2922         pos_type const parsize = d->text_.size();
2923         for (i = 0; pos + i < parsize; ++i) {
2924                 if (i >= strsize)
2925                         break;
2926                 if (cs && str[i] != d->text_[pos + i])
2927                         break;
2928                 if (!cs && uppercase(str[i]) != uppercase(d->text_[pos + i]))
2929                         break;
2930                 if (!del && isDeleted(pos + i))
2931                         break;
2932         }
2933
2934         if (i != strsize)
2935                 return false;
2936
2937         // if necessary, check whether string matches word
2938         if (mw) {
2939                 if (pos > 0 && !isWordSeparator(pos - 1))
2940                         return false;
2941                 if (pos + strsize < parsize
2942                         && !isWordSeparator(pos + strsize))
2943                         return false;
2944         }
2945
2946         return true;
2947 }
2948
2949
2950 char_type Paragraph::getChar(pos_type pos) const
2951 {
2952         return d->text_[pos];
2953 }
2954
2955
2956 pos_type Paragraph::size() const
2957 {
2958         return d->text_.size();
2959 }
2960
2961
2962 bool Paragraph::empty() const
2963 {
2964         return d->text_.empty();
2965 }
2966
2967
2968 bool Paragraph::isInset(pos_type pos) const
2969 {
2970         return d->text_[pos] == META_INSET;
2971 }
2972
2973
2974 bool Paragraph::isSeparator(pos_type pos) const
2975 {
2976         //FIXME: Are we sure this can be the only separator?
2977         return d->text_[pos] == ' ';
2978 }
2979
2980
2981 void Paragraph::deregisterWords()
2982 {
2983         Private::Words::const_iterator it;
2984         WordList & wl = theWordList();
2985         for (it = d->words_.begin(); it != d->words_.end(); ++it)
2986                 wl.remove(*it);
2987         d->words_.clear();
2988 }
2989
2990
2991 void Paragraph::locateWord(pos_type & from, pos_type & to,
2992         word_location const loc) const
2993 {
2994         switch (loc) {
2995         case WHOLE_WORD_STRICT:
2996                 if (from == 0 || from == size()
2997                     || isWordSeparator(from)
2998                     || isWordSeparator(from - 1)) {
2999                         to = from;
3000                         return;
3001                 }
3002                 // no break here, we go to the next
3003
3004         case WHOLE_WORD:
3005                 // If we are already at the beginning of a word, do nothing
3006                 if (!from || isWordSeparator(from - 1))
3007                         break;
3008                 // no break here, we go to the next
3009
3010         case PREVIOUS_WORD:
3011                 // always move the cursor to the beginning of previous word
3012                 while (from && !isWordSeparator(from - 1))
3013                         --from;
3014                 break;
3015         case NEXT_WORD:
3016                 LYXERR0("Paragraph::locateWord: NEXT_WORD not implemented yet");
3017                 break;
3018         case PARTIAL_WORD:
3019                 // no need to move the 'from' cursor
3020                 break;
3021         }
3022         to = from;
3023         while (to < size() && !isWordSeparator(to))
3024                 ++to;
3025 }
3026
3027
3028 void Paragraph::collectWords()
3029 {
3030         SpellChecker * speller = theSpellChecker();
3031
3032         //lyxerr << "Words: ";
3033         pos_type n = size();
3034         for (pos_type pos = 0; pos < n; ++pos) {
3035                 if (isWordSeparator(pos))
3036                         continue;
3037                 pos_type from = pos;
3038                 locateWord(from, pos, WHOLE_WORD);
3039                 if (!lyxrc.spellcheck_continuously && pos - from < 6)
3040                         continue;
3041
3042                 docstring word = asString(from, pos, false);
3043                 if (pos - from >= 6)
3044                         d->words_.insert(word);
3045
3046                 if (!lyxrc.spellcheck_continuously || !speller)
3047                         continue;
3048
3049                 string const lang_code = lyxrc.spellchecker_alt_lang.empty()
3050                         ? getFontSettings(d->inset_owner_->buffer().params(), from).language()->code()
3051                         : lyxrc.spellchecker_alt_lang;
3052                 WordLangTuple wl(word, lang_code);
3053                 SpellChecker::Result res = speller->check(wl);
3054                 // ... just ignore any error that the spellchecker reports.
3055                 if (!speller->error().empty())
3056                         continue;
3057                 bool const misspelled = res != SpellChecker::OK
3058                         && res != SpellChecker::IGNORED_WORD;
3059                 d->fontlist_.setMisspelled(from, pos, misspelled);
3060
3061                 //lyxerr << word << " ";
3062         }
3063         //lyxerr << std::endl;
3064 }
3065
3066
3067 void Paragraph::registerWords()
3068 {
3069         Private::Words::const_iterator it;
3070         WordList & wl = theWordList();
3071         for (it = d->words_.begin(); it != d->words_.end(); ++it)
3072                 wl.insert(*it);
3073 }
3074
3075
3076 void Paragraph::updateWords()
3077 {
3078         deregisterWords();
3079         collectWords();
3080         registerWords();
3081 }
3082
3083
3084 bool Paragraph::isMisspelled(pos_type pos) const
3085 {
3086         SpellChecker * speller = theSpellChecker();
3087         pos_type from = pos;
3088         pos_type to = pos;
3089         locateWord(from, to, WHOLE_WORD);
3090         docstring word = asString(from, to, false);
3091         if (!speller)
3092                 return false;
3093
3094         string const lang_code = lyxrc.spellchecker_alt_lang.empty()
3095                 ? getFontSettings(d->inset_owner_->buffer().params(), from).language()->code()
3096                 : lyxrc.spellchecker_alt_lang;
3097         WordLangTuple wl(word, lang_code);
3098         SpellChecker::Result res = speller->check(wl);
3099         // ... just ignore any error that the spellchecker reports.
3100         if (!speller->error().empty())
3101                 return false;
3102
3103         bool const misspelled = res != SpellChecker::OK
3104                 && res != SpellChecker::IGNORED_WORD;
3105         if (lyxrc.spellcheck_continuously)
3106                 d->fontlist_.setMisspelled(from, pos, misspelled);
3107         return misspelled;
3108 }
3109
3110
3111 } // namespace lyx