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/debug.h"
  56 #include "support/docstring_list.h"
  57 #include "support/ExceptionMessage.h"
  58 #include "support/gettext.h"
  59 #include "support/lassert.h"
  60 #include "support/lstrings.h"
  61 #include "support/textutils.h"
  62
  63 #include <sstream>
  64 #include <vector>
  65
  66 using namespace std;
  67 using namespace lyx::support;
  68
  69 namespace lyx {
  70
  71 namespace {
  72 /// Inset identifier (above 0x10ffff, for ucs-4)
  73 char_type const META_INSET = 0x200001;
  74 };
  75
  76 /////////////////////////////////////////////////////////////////////
  77 //
  78 // Paragraph::Private
  79 //
  80 /////////////////////////////////////////////////////////////////////
  81
  82 class Paragraph::Private
  83 {
  84 public:
  85         ///
  86         Private(Paragraph * owner, Layout const & layout);
  87         /// "Copy constructor"
  88         Private(Private const &, Paragraph * owner);
  89         /// Copy constructor from \p beg  to \p end
  90         Private(Private const &, Paragraph * owner, pos_type beg, pos_type end);
  91
  92         ///
  93         void insertChar(pos_type pos, char_type c, Change const & change);
  94
  95         /// Output the surrogate pair formed by \p c and \p next to \p os.
  96         /// \return the number of characters written.
  97         int latexSurrogatePair(odocstream & os, char_type c, char_type next,
  98                                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 Paragraph::expandLabel(Layout const & layout,
1609                 BufferParams const & bparams, bool process_appendix) const
1610 {
1611         DocumentClass const & tclass = bparams.documentClass();
1612         string const & lang = getParLanguage(bparams)->code();
1613         bool const in_appendix = process_appendix && d->params_.appendix();
1614         docstring fmt = translateIfPossible(layout.labelstring(in_appendix), lang);
1615
1616         if (fmt.empty() && layout.labeltype == LABEL_COUNTER
1617             && !layout.counter.empty())
1618                 return tclass.counters().theCounter(layout.counter, lang);
1619
1620         // handle 'inherited level parts' in 'fmt',
1621         // i.e. the stuff between '@' in   '@Section@.\arabic{subsection}'
1622         size_t const i = fmt.find('@', 0);
1623         if (i != docstring::npos) {
1624                 size_t const j = fmt.find('@', i + 1);
1625                 if (j != docstring::npos) {
1626                         docstring parent(fmt, i + 1, j - i - 1);
1627                         docstring label = from_ascii("??");
1628                         if (tclass.hasLayout(parent))
1629                                 docstring label = expandLabel(tclass[parent], bparams,
1630                                                       process_appendix);
1631                         fmt = docstring(fmt, 0, i) + label
1632                                 + docstring(fmt, j + 1, docstring::npos);
1633                 }
1634         }
1635
1636         return tclass.counters().counterLabel(fmt, lang);
1637 }
1638
1639
1640 void Paragraph::applyLayout(Layout const & new_layout)
1641 {
1642         d->layout_ = &new_layout;
1643         LyXAlignment const oldAlign = d->params_.align();
1644
1645         if (!(oldAlign & d->layout_->alignpossible)) {
1646                 frontend::Alert::warning(_("Alignment not permitted"),
1647                         _("The new layout does not permit the alignment previously used.\nSetting to default."));
1648                 d->params_.align(LYX_ALIGN_LAYOUT);
1649         }
1650 }
1651
1652
1653 pos_type Paragraph::beginOfBody() const
1654 {
1655         return d->begin_of_body_;
1656 }
1657
1658
1659 void Paragraph::setBeginOfBody()
1660 {
1661         if (d->layout_->labeltype != LABEL_MANUAL) {
1662                 d->begin_of_body_ = 0;
1663                 return;
1664         }
1665
1666         // Unroll the first two cycles of the loop
1667         // and remember the previous character to
1668         // remove unnecessary getChar() calls
1669         pos_type i = 0;
1670         pos_type end = size();
1671         if (i < end && !isNewline(i)) {
1672                 ++i;
1673                 char_type previous_char = 0;
1674                 char_type temp = 0;
1675                 if (i < end) {
1676                         previous_char = d->text_[i];
1677                         if (!isNewline(i)) {
1678                                 ++i;
1679                                 while (i < end && previous_char != ' ') {
1680                                         temp = d->text_[i];
1681                                         if (isNewline(i))
1682                                                 break;
1683                                         ++i;
1684                                         previous_char = temp;
1685                                 }
1686                         }
1687                 }
1688         }
1689
1690         d->begin_of_body_ = i;
1691 }
1692
1693
1694 bool Paragraph::forcePlainLayout() const
1695 {
1696         return inInset().forcePlainLayout();
1697 }
1698
1699
1700 bool Paragraph::allowParagraphCustomization() const
1701 {
1702         return inInset().allowParagraphCustomization();
1703 }
1704
1705
1706 bool Paragraph::usePlainLayout() const
1707 {
1708         return inInset().usePlainLayout();
1709 }
1710
1711
1712 namespace {
1713
1714 // paragraphs inside floats need different alignment tags to avoid
1715 // unwanted space
1716
1717 bool noTrivlistCentering(InsetCode code)
1718 {
1719         return code == FLOAT_CODE
1720                || code == WRAP_CODE
1721                || code == CELL_CODE;
1722 }
1723
1724
1725 string correction(string const & orig)
1726 {
1727         if (orig == "flushleft")
1728                 return "raggedright";
1729         if (orig == "flushright")
1730                 return "raggedleft";
1731         if (orig == "center")
1732                 return "centering";
1733         return orig;
1734 }
1735
1736
1737 string const corrected_env(string const & suffix, string const & env,
1738         InsetCode code, bool const lastpar)
1739 {
1740         string output = suffix + "{";
1741         if (noTrivlistCentering(code)) {
1742                 if (lastpar) {
1743                         // the last paragraph in non-trivlist-aligned
1744                         // context is special (to avoid unwanted whitespace)
1745                         if (suffix == "\\begin")
1746                                 return "\\" + correction(env) + "{}";
1747                         return string();
1748                 }
1749                 output += correction(env);
1750         } else
1751                 output += env;
1752         output += "}";
1753         if (suffix == "\\begin")
1754                 output += "\n";
1755         return output;
1756 }
1757
1758
1759 void adjust_row_column(string const & str, TexRow & texrow, int & column)
1760 {
1761         if (!contains(str, "\n"))
1762                 column += str.size();
1763         else {
1764                 string tmp;
1765                 texrow.newline();
1766                 column = rsplit(str, tmp, '\n').size();
1767         }
1768 }
1769
1770 } // namespace anon
1771
1772
1773 int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
1774                                  odocstream & os, TexRow & texrow,
1775                                  OutputParams const & runparams) const
1776 {
1777         int column = 0;
1778
1779         if (params_.noindent()) {
1780                 os << "\\noindent ";
1781                 column += 10;
1782         }
1783
1784         LyXAlignment const curAlign = params_.align();
1785
1786         if (curAlign == layout_->align)
1787                 return column;
1788
1789         switch (curAlign) {
1790         case LYX_ALIGN_NONE:
1791         case LYX_ALIGN_BLOCK:
1792         case LYX_ALIGN_LAYOUT:
1793         case LYX_ALIGN_SPECIAL:
1794                 break;
1795         case LYX_ALIGN_LEFT:
1796         case LYX_ALIGN_RIGHT:
1797         case LYX_ALIGN_CENTER:
1798                 if (runparams.moving_arg) {
1799                         os << "\\protect";
1800                         column += 8;
1801                 }
1802                 break;
1803         }
1804
1805         string const begin_tag = "\\begin";
1806         InsetCode code = owner_->ownerCode();
1807         bool const lastpar = runparams.isLastPar;
1808
1809         switch (curAlign) {
1810         case LYX_ALIGN_NONE:
1811         case LYX_ALIGN_BLOCK:
1812         case LYX_ALIGN_LAYOUT:
1813         case LYX_ALIGN_SPECIAL:
1814                 break;
1815         case LYX_ALIGN_LEFT: {
1816                 string output;
1817                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
1818                         output = corrected_env(begin_tag, "flushleft", code, lastpar);
1819                 else
1820                         output = corrected_env(begin_tag, "flushright", code, lastpar);
1821                 os << from_ascii(output);
1822                 adjust_row_column(output, texrow, column);
1823                 break;
1824         } case LYX_ALIGN_RIGHT: {
1825                 string output;
1826                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
1827                         output = corrected_env(begin_tag, "flushright", code, lastpar);
1828                 else
1829                         output = corrected_env(begin_tag, "flushleft", code, lastpar);
1830                 os << from_ascii(output);
1831                 adjust_row_column(output, texrow, column);
1832                 break;
1833         } case LYX_ALIGN_CENTER: {
1834                 string output;
1835                 output = corrected_env(begin_tag, "center", code, lastpar);
1836                 os << from_ascii(output);
1837                 adjust_row_column(output, texrow, column);
1838                 break;
1839         }
1840         }
1841
1842         return column;
1843 }
1844
1845
1846 int Paragraph::Private::endTeXParParams(BufferParams const & bparams,
1847                                odocstream & os, TexRow & texrow,
1848                                OutputParams const & runparams) const
1849 {
1850         int column = 0;
1851
1852         LyXAlignment const curAlign = params_.align();
1853
1854         if (curAlign == layout_->align)
1855                 return column;
1856
1857         switch (curAlign) {
1858         case LYX_ALIGN_NONE:
1859         case LYX_ALIGN_BLOCK:
1860         case LYX_ALIGN_LAYOUT:
1861         case LYX_ALIGN_SPECIAL:
1862                 break;
1863         case LYX_ALIGN_LEFT:
1864         case LYX_ALIGN_RIGHT:
1865         case LYX_ALIGN_CENTER:
1866                 if (runparams.moving_arg) {
1867                         os << "\\protect";
1868                         column = 8;
1869                 }
1870                 break;
1871         }
1872
1873         string const end_tag = "\n\\par\\end";
1874         InsetCode code = owner_->ownerCode();
1875         bool const lastpar = runparams.isLastPar;
1876
1877         switch (curAlign) {
1878         case LYX_ALIGN_NONE:
1879         case LYX_ALIGN_BLOCK:
1880         case LYX_ALIGN_LAYOUT:
1881         case LYX_ALIGN_SPECIAL:
1882                 break;
1883         case LYX_ALIGN_LEFT: {
1884                 string output;
1885                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
1886                         output = corrected_env(end_tag, "flushleft", code, lastpar);
1887                 else
1888                         output = corrected_env(end_tag, "flushright", code, lastpar);
1889                 os << from_ascii(output);
1890                 adjust_row_column(output, texrow, column);
1891                 break;
1892         } case LYX_ALIGN_RIGHT: {
1893                 string output;
1894                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
1895                         output = corrected_env(end_tag, "flushright", code, lastpar);
1896                 else
1897                         output = corrected_env(end_tag, "flushleft", code, lastpar);
1898                 os << from_ascii(output);
1899                 adjust_row_column(output, texrow, column);
1900                 break;
1901         } case LYX_ALIGN_CENTER: {
1902                 string output;
1903                 output = corrected_env(end_tag, "center", code, lastpar);
1904                 os << from_ascii(output);
1905                 adjust_row_column(output, texrow, column);
1906                 break;
1907         }
1908         }
1909
1910         return column;
1911 }
1912
1913
1914 // This one spits out the text of the paragraph
1915 bool Paragraph::latex(BufferParams const & bparams,
1916         Font const & outerfont,
1917         odocstream & os, TexRow & texrow,
1918         OutputParams const & runparams,
1919         int start_pos, int end_pos) const
1920 {
1921         LYXERR(Debug::LATEX, "Paragraph::latex...     " << this);
1922
1923         bool return_value = false;
1924
1925         bool const allowcust = allowParagraphCustomization();
1926
1927         // FIXME This check should not be needed. Perhaps issue an
1928         // error if it triggers.
1929         Layout const & style = forcePlainLayout() ?
1930                 bparams.documentClass().plainLayout() : *d->layout_;
1931
1932         // Current base font for all inherited font changes, without any
1933         // change caused by an individual character, except for the language:
1934         // It is set to the language of the first character.
1935         // As long as we are in the label, this font is the base font of the
1936         // label. Before the first body character it is set to the base font
1937         // of the body.
1938         Font basefont;
1939
1940         // Maybe we have to create a optional argument.
1941         pos_type body_pos = beginOfBody();
1942         unsigned int column = 0;
1943
1944         if (body_pos > 0) {
1945                 os << '[';
1946                 column += 1;
1947                 basefont = getLabelFont(bparams, outerfont);
1948         } else {
1949                 basefont = getLayoutFont(bparams, outerfont);
1950         }
1951
1952         // Which font is currently active?
1953         Font running_font(basefont);
1954         // Do we have an open font change?
1955         bool open_font = false;
1956
1957         Change runningChange = Change(Change::UNCHANGED);
1958
1959         texrow.start(id(), 0);
1960
1961         // if the paragraph is empty, the loop will not be entered at all
1962         if (empty()) {
1963                 if (style.isCommand()) {
1964                         os << '{';
1965                         ++column;
1966                 }
1967                 if (allowcust)
1968                         column += d->startTeXParParams(bparams, os, texrow,
1969                                                     runparams);
1970         }
1971
1972         for (pos_type i = 0; i < size(); ++i) {
1973                 // First char in paragraph or after label?
1974                 if (i == body_pos) {
1975                         if (body_pos > 0) {
1976                                 if (open_font) {
1977                                         column += running_font.latexWriteEndChanges(
1978                                                 os, bparams, runparams,
1979                                                 basefont, basefont);
1980                                         open_font = false;
1981                                 }
1982                                 basefont = getLayoutFont(bparams, outerfont);
1983                                 running_font = basefont;
1984
1985                                 column += Changes::latexMarkChange(os, bparams,
1986                                                 runningChange, Change(Change::UNCHANGED));
1987                                 runningChange = Change(Change::UNCHANGED);
1988
1989                                 os << "] ";
1990                                 column +=2;
1991                         }
1992                         if (style.isCommand()) {
1993                                 os << '{';
1994                                 ++column;
1995                         }
1996
1997                         if (allowcust)
1998                                 column += d->startTeXParParams(bparams, os,
1999                                                             texrow,
2000                                                             runparams);
2001                 }
2002
2003                 Change const & change = runparams.inDeletedInset ? runparams.changeOfDeletedInset
2004                                                                  : lookupChange(i);
2005
2006                 if (bparams.outputChanges && runningChange != change) {
2007                         if (open_font) {
2008                                 column += running_font.latexWriteEndChanges(
2009                                                 os, bparams, runparams, basefont, basefont);
2010                                 open_font = false;
2011                         }
2012                         basefont = getLayoutFont(bparams, outerfont);
2013                         running_font = basefont;
2014
2015                         column += Changes::latexMarkChange(os, bparams, runningChange, change);
2016                         runningChange = change;
2017                 }
2018
2019                 // do not output text which is marked deleted
2020                 // if change tracking output is disabled
2021                 if (!bparams.outputChanges && change.type == Change::DELETED) {
2022                         continue;
2023                 }
2024
2025                 ++column;
2026
2027                 // Fully instantiated font
2028                 Font const font = getFont(bparams, i, outerfont);
2029
2030                 Font const last_font = running_font;
2031
2032                 // Do we need to close the previous font?
2033                 if (open_font &&
2034                     (font != running_font ||
2035                      font.language() != running_font.language()))
2036                 {
2037                         column += running_font.latexWriteEndChanges(
2038                                         os, bparams, runparams, basefont,
2039                                         (i == body_pos-1) ? basefont : font);
2040                         running_font = basefont;
2041                         open_font = false;
2042                 }
2043
2044                 // close babel's font environment before opening CJK.
2045                 if (!running_font.language()->babel().empty() &&
2046                     font.language()->encoding()->package() == Encoding::CJK) {
2047                                 string end_tag = subst(lyxrc.language_command_end,
2048                                                         "$$lang",
2049                                                         running_font.language()->babel());
2050                                 os << from_ascii(end_tag);
2051                                 column += end_tag.length();
2052                 }
2053
2054                 // Switch file encoding if necessary (and allowed)
2055                 if (!runparams.verbatim &&
2056                     runparams.encoding->package() != Encoding::none &&
2057                     font.language()->encoding()->package() != Encoding::none) {
2058                         pair<bool, int> const enc_switch = switchEncoding(os, bparams,
2059                                         runparams, *(font.language()->encoding()));
2060                         if (enc_switch.first) {
2061                                 column += enc_switch.second;
2062                                 runparams.encoding = font.language()->encoding();
2063                         }
2064                 }
2065
2066                 char_type const c = d->text_[i];
2067
2068                 // Do we need to change font?
2069                 if ((font != running_font ||
2070                      font.language() != running_font.language()) &&
2071                         i != body_pos - 1)
2072                 {
2073                         odocstringstream ods;
2074                         column += font.latexWriteStartChanges(ods, bparams,
2075                                                               runparams, basefont,
2076                                                               last_font);
2077                         running_font = font;
2078                         open_font = true;
2079                         docstring fontchange = ods.str();
2080                         // check if the fontchange ends with a trailing blank
2081                         // (like "\small " (see bug 3382)
2082                         if (suffixIs(fontchange, ' ') && c == ' ')
2083                                 os << fontchange.substr(0, fontchange.size() - 1)
2084                                    << from_ascii("{}");
2085                         else
2086                                 os << fontchange;
2087                 }
2088
2089                 // FIXME: think about end_pos implementation...
2090                 if (c == ' ' && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2091                         // FIXME: integrate this case in latexSpecialChar
2092                         // Do not print the separation of the optional argument
2093                         // if style.pass_thru is false. This works because
2094                         // latexSpecialChar ignores spaces if
2095                         // style.pass_thru is false.
2096                         if (i != body_pos - 1) {
2097                                 if (d->simpleTeXBlanks(
2098                                                 runparams, os, texrow,
2099                                                 i, column, font, style)) {
2100                                         // A surrogate pair was output. We
2101                                         // must not call latexSpecialChar
2102                                         // in this iteration, since it would output
2103                                         // the combining character again.
2104                                         ++i;
2105                                         continue;
2106                                 }
2107                         }
2108                 }
2109
2110                 OutputParams rp = runparams;
2111                 rp.free_spacing = style.free_spacing;
2112                 rp.local_font = &font;
2113                 rp.intitle = style.intitle;
2114
2115                 // Two major modes:  LaTeX or plain
2116                 // Handle here those cases common to both modes
2117                 // and then split to handle the two modes separately.
2118                 if (c == META_INSET) {
2119                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2120                                 d->latexInset(bparams, os,
2121                                                 texrow, rp, running_font,
2122                                                 basefont, outerfont, open_font,
2123                                                 runningChange, style, i, column);
2124                         }
2125                 } else {
2126                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2127                                 try {
2128                                         d->latexSpecialChar(os, rp, running_font, runningChange,
2129                                                 style, i, column);
2130                                 } catch (EncodingException & e) {
2131                                 if (runparams.dryrun) {
2132                                         os << "<" << _("LyX Warning: ")
2133                                            << _("uncodable character") << " '";
2134                                         os.put(c);
2135                                         os << "'>";
2136                                 } else {
2137                                         // add location information and throw again.
2138                                         e.par_id = id();
2139                                         e.pos = i;
2140                                         throw(e);
2141                                 }
2142                         }
2143                 }
2144                 }
2145
2146                 // Set the encoding to that returned from latexSpecialChar (see
2147                 // comment for encoding member in OutputParams.h)
2148                 runparams.encoding = rp.encoding;
2149         }
2150
2151         // If we have an open font definition, we have to close it
2152         if (open_font) {
2153 #ifdef FIXED_LANGUAGE_END_DETECTION
2154                 if (next_) {
2155                         running_font
2156                                 .latexWriteEndChanges(os, bparams, runparams,
2157                                         basefont,
2158                                         next_->getFont(bparams, 0, outerfont));
2159                 } else {
2160                         running_font.latexWriteEndChanges(os, bparams,
2161                                         runparams, basefont, basefont);
2162                 }
2163 #else
2164 //FIXME: For now we ALWAYS have to close the foreign font settings if they are
2165 //FIXME: there as we start another \selectlanguage with the next paragraph if
2166 //FIXME: we are in need of this. This should be fixed sometime (Jug)
2167                 running_font.latexWriteEndChanges(os, bparams, runparams,
2168                                 basefont, basefont);
2169 #endif
2170         }
2171
2172         column += Changes::latexMarkChange(os, bparams, runningChange, Change(Change::UNCHANGED));
2173
2174         // Needed if there is an optional argument but no contents.
2175         if (body_pos > 0 && body_pos == size()) {
2176                 os << "]~";
2177                 return_value = false;
2178         }
2179
2180         if (allowcust) {
2181                 column += d->endTeXParParams(bparams, os, texrow,
2182                                           runparams);
2183         }
2184
2185         LYXERR(Debug::LATEX, "Paragraph::latex... done " << this);
2186         return return_value;
2187 }
2188
2189
2190 bool Paragraph::emptyTag() const
2191 {
2192         for (pos_type i = 0; i < size(); ++i) {
2193                 if (Inset const * inset = getInset(i)) {
2194                         InsetCode lyx_code = inset->lyxCode();
2195                         if (lyx_code != TOC_CODE &&
2196                             lyx_code != INCLUDE_CODE &&
2197                             lyx_code != GRAPHICS_CODE &&
2198                             lyx_code != ERT_CODE &&
2199                             lyx_code != LISTINGS_CODE &&
2200                             lyx_code != FLOAT_CODE &&
2201                             lyx_code != TABULAR_CODE) {
2202                                 return false;
2203                         }
2204                 } else {
2205                         char_type c = d->text_[i];
2206                         if (c != ' ' && c != '\t')
2207                                 return false;
2208                 }
2209         }
2210         return true;
2211 }
2212
2213
2214 string Paragraph::getID(Buffer const & buf, OutputParams const & runparams)
2215         const
2216 {
2217         for (pos_type i = 0; i < size(); ++i) {
2218                 if (Inset const * inset = getInset(i)) {
2219                         InsetCode lyx_code = inset->lyxCode();
2220                         if (lyx_code == LABEL_CODE) {
2221                                 InsetLabel const * const il = static_cast<InsetLabel const *>(inset);
2222                                 docstring const & id = il->getParam("name");
2223                                 return "id='" + to_utf8(sgml::cleanID(buf, runparams, id)) + "'";
2224                         }
2225                 }
2226         }
2227         return string();
2228 }
2229
2230
2231 pos_type Paragraph::firstWordDocBook(odocstream & os, OutputParams const & runparams)
2232         const
2233 {
2234         pos_type i;
2235         for (i = 0; i < size(); ++i) {
2236                 if (Inset const * inset = getInset(i)) {
2237                         inset->docbook(os, runparams);
2238                 } else {
2239                         char_type c = d->text_[i];
2240                         if (c == ' ')
2241                                 break;
2242                         os << sgml::escapeChar(c);
2243                 }
2244         }
2245         return i;
2246 }
2247
2248
2249 pos_type Paragraph::firstWordLyXHTML(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->xhtml(os, runparams);
2256                 } else {
2257                         char_type c = d->text_[i];
2258                         if (c == ' ')
2259                                 break;
2260                         os << html::escapeChar(c);
2261                 }
2262         }
2263         return i;
2264 }
2265
2266
2267 bool Paragraph::Private::onlyText(Buffer const & buf, Font const & outerfont, pos_type initial) const
2268 {
2269         Font font_old;
2270         pos_type size = text_.size();
2271         for (pos_type i = initial; i < size; ++i) {
2272                 Font font = owner_->getFont(buf.params(), i, outerfont);
2273                 if (text_[i] == META_INSET)
2274                         return false;
2275                 if (i != initial && font != font_old)
2276                         return false;
2277                 font_old = font;
2278         }
2279
2280         return true;
2281 }
2282
2283
2284 void Paragraph::simpleDocBookOnePar(Buffer const & buf,
2285                                     odocstream & os,
2286                                     OutputParams const & runparams,
2287                                     Font const & outerfont,
2288                                     pos_type initial) const
2289 {
2290         bool emph_flag = false;
2291
2292         Layout const & style = *d->layout_;
2293         FontInfo font_old =
2294                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2295
2296         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2297                 os << "]]>";
2298
2299         // parsing main loop
2300         for (pos_type i = initial; i < size(); ++i) {
2301                 Font font = getFont(buf.params(), i, outerfont);
2302
2303                 // handle <emphasis> tag
2304                 if (font_old.emph() != font.fontInfo().emph()) {
2305                         if (font.fontInfo().emph() == FONT_ON) {
2306                                 os << "<emphasis>";
2307                                 emph_flag = true;
2308                         } else if (i != initial) {
2309                                 os << "</emphasis>";
2310                                 emph_flag = false;
2311                         }
2312                 }
2313
2314                 if (Inset const * inset = getInset(i)) {
2315                         inset->docbook(os, runparams);
2316                 } else {
2317                         char_type c = d->text_[i];
2318
2319                         if (style.pass_thru)
2320                                 os.put(c);
2321                         else
2322                                 os << sgml::escapeChar(c);
2323                 }
2324                 font_old = font.fontInfo();
2325         }
2326
2327         if (emph_flag) {
2328                 os << "</emphasis>";
2329         }
2330
2331         if (style.free_spacing)
2332                 os << '\n';
2333         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2334                 os << "<![CDATA[";
2335 }
2336
2337
2338 docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
2339                                     odocstream & os,
2340                                     OutputParams const & runparams,
2341                                     Font const & outerfont,
2342                                     pos_type initial) const
2343 {
2344         docstring retval;
2345
2346         // FIXME We really need to manage the tag nesting here.
2347         // Probably in the same sort of way as in output_xhtml.
2348         bool emph_flag = false;
2349         bool bold_flag = false;
2350         std::string closing_tag;
2351
2352         Layout const & style = *d->layout_;
2353         FontInfo font_old =
2354                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2355
2356         //if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2357         //      os << "]]>";
2358
2359         // parsing main loop
2360         for (pos_type i = initial; i < size(); ++i) {
2361                 Font font = getFont(buf.params(), i, outerfont);
2362
2363                 // emphasis
2364                 if (font_old.emph() != font.fontInfo().emph()) {
2365                         if (font.fontInfo().emph() == FONT_ON) {
2366                                 os << "<em>";
2367                                 emph_flag = true;
2368                         } else if (emph_flag && i != initial) {
2369                                 os << "</em>";
2370                                 emph_flag = false;
2371                         }
2372                 }
2373                 // bold
2374                 if (font_old.series() != font.fontInfo().series()) {
2375                         if (font.fontInfo().series() == BOLD_SERIES) {
2376                                 os << "<strong>";
2377                                 bold_flag = true;
2378                         } else if (bold_flag && i != initial) {
2379                                 os << "</strong>";
2380                                 bold_flag = false;
2381                         }
2382                 }
2383                 // FIXME Other such tags?
2384
2385                 if (Inset const * inset = getInset(i)) {
2386                         retval += inset->xhtml(os, runparams);
2387                 } else {
2388                         char_type c = d->text_[i];
2389
2390                         if (style.pass_thru)
2391                                 os.put(c);
2392                         else if (c == '-') {
2393                                 docstring str;
2394                                 int j = i + 1;
2395                                 if (j < size() && d->text_[j] == '-') {
2396                                         j += 1;
2397                                         if (j < size() && d->text_[j] == '-') {
2398                                                 str += from_ascii("&mdash;");
2399                                                 i += 2;
2400                                         } else {
2401                                                 str += from_ascii("&ndash;");
2402                                                 i += 1;
2403                                         }
2404                                 }
2405                                 else
2406                                         str += c;
2407                                 os << str;
2408                         } else
2409                                 os << html::escapeChar(c);
2410                 }
2411                 font_old = font.fontInfo();
2412         }
2413
2414         // FIXME This could be out of order. See above.
2415         if (emph_flag)
2416                 os << "</em>";
2417         if (bold_flag)
2418                 os << "</strong>";
2419
2420         return retval;
2421 }
2422
2423
2424 bool Paragraph::isHfill(pos_type pos) const
2425 {
2426         Inset const * inset = getInset(pos);
2427         return inset && (inset->lyxCode() == SPACE_CODE &&
2428                          inset->isStretchableSpace());
2429 }
2430
2431
2432 bool Paragraph::isNewline(pos_type pos) const
2433 {
2434         Inset const * inset = getInset(pos);
2435         return inset && inset->lyxCode() == NEWLINE_CODE;
2436 }
2437
2438
2439 bool Paragraph::isLineSeparator(pos_type pos) const
2440 {
2441         char_type const c = d->text_[pos];
2442         if (isLineSeparatorChar(c))
2443                 return true;
2444         Inset const * inset = getInset(pos);
2445         return inset && inset->isLineSeparator();
2446 }
2447
2448
2449 bool Paragraph::isWordSeparator(pos_type pos) const
2450 {
2451         if (Inset const * inset = getInset(pos))
2452                 return !inset->isLetter();
2453         char_type const c = d->text_[pos];
2454         // We want to pass the ' and escape chars to the spellchecker
2455         static docstring const quote = from_utf8(lyxrc.spellchecker_esc_chars + '\'');
2456         return (!isLetterChar(c) && !isDigit(c) && !contains(quote, c))
2457                 || pos == size();
2458 }
2459
2460
2461 bool Paragraph::isChar(pos_type pos) const
2462 {
2463         if (Inset const * inset = getInset(pos))
2464                 return inset->isChar();
2465         char_type const c = d->text_[pos];
2466         return !isLetterChar(c) && !isDigit(c) && !lyx::isSpace(c);
2467 }
2468
2469
2470 bool Paragraph::isSpace(pos_type pos) const
2471 {
2472         if (Inset const * inset = getInset(pos))
2473                 return inset->isSpace();
2474         char_type const c = d->text_[pos];
2475         return lyx::isSpace(c);
2476 }
2477
2478
2479 Language const *
2480 Paragraph::getParLanguage(BufferParams const & bparams) const
2481 {
2482         if (!empty())
2483                 return getFirstFontSettings(bparams).language();
2484         // FIXME: we should check the prev par as well (Lgb)
2485         return bparams.language;
2486 }
2487
2488
2489 bool Paragraph::isRTL(BufferParams const & bparams) const
2490 {
2491         return lyxrc.rtl_support
2492                 && getParLanguage(bparams)->rightToLeft()
2493                 && ownerCode() != ERT_CODE
2494                 && ownerCode() != LISTINGS_CODE;
2495 }
2496
2497
2498 void Paragraph::changeLanguage(BufferParams const & bparams,
2499                                Language const * from, Language const * to)
2500 {
2501         // change language including dummy font change at the end
2502         for (pos_type i = 0; i <= size(); ++i) {
2503                 Font font = getFontSettings(bparams, i);
2504                 if (font.language() == from) {
2505                         font.setLanguage(to);
2506                         setFont(i, font);
2507                 }
2508         }
2509 }
2510
2511
2512 bool Paragraph::isMultiLingual(BufferParams const & bparams) const
2513 {
2514         Language const * doc_language = bparams.language;
2515         FontList::const_iterator cit = d->fontlist_.begin();
2516         FontList::const_iterator end = d->fontlist_.end();
2517
2518         for (; cit != end; ++cit)
2519                 if (cit->font().language() != ignore_language &&
2520                     cit->font().language() != latex_language &&
2521                     cit->font().language() != doc_language)
2522                         return true;
2523         return false;
2524 }
2525
2526
2527 docstring Paragraph::asString(int options) const
2528 {
2529         return asString(0, size(), options);
2530 }
2531
2532
2533 docstring Paragraph::asString(pos_type beg, pos_type end, int options) const
2534 {
2535         odocstringstream os;
2536
2537         if (beg == 0
2538                 && options & AS_STR_LABEL
2539                 && !d->params_.labelString().empty())
2540                 os << d->params_.labelString() << ' ';
2541
2542         for (pos_type i = beg; i < end; ++i) {
2543                 char_type const c = d->text_[i];
2544                 if (isPrintable(c) || c == '\t'
2545                     || (c == '\n' && options & AS_STR_NEWLINES))
2546                         os.put(c);
2547                 else if (c == META_INSET && options & AS_STR_INSETS)
2548                         getInset(i)->tocString(os);
2549         }
2550
2551         return os.str();
2552 }
2553
2554
2555 docstring Paragraph::stringify(pos_type beg, pos_type end, int options, OutputParams & runparams) const
2556 {
2557         odocstringstream os;
2558
2559         if (beg == 0
2560                 && options & AS_STR_LABEL
2561                 && !d->params_.labelString().empty())
2562                 os << d->params_.labelString() << ' ';
2563
2564         for (pos_type i = beg; i < end; ++i) {
2565                 char_type const c = d->text_[i];
2566                 if (isPrintable(c) || c == '\t'
2567                     || (c == '\n' && options & AS_STR_NEWLINES))
2568                         os.put(c);
2569                 else if (c == META_INSET && options & AS_STR_INSETS) {
2570                         getInset(i)->plaintext(os, runparams);
2571                 }
2572         }
2573
2574         return os.str();
2575 }
2576
2577
2578 void Paragraph::setInsetOwner(Inset const * inset)
2579 {
2580         d->inset_owner_ = inset;
2581 }
2582
2583
2584 int Paragraph::id() const
2585 {
2586         return d->id_;
2587 }
2588
2589
2590 Layout const & Paragraph::layout() const
2591 {
2592         return *d->layout_;
2593 }
2594
2595
2596 void Paragraph::setLayout(Layout const & layout)
2597 {
2598         d->layout_ = &layout;
2599 }
2600
2601
2602 void Paragraph::setDefaultLayout(DocumentClass const & tc)
2603 {
2604         setLayout(tc.defaultLayout());
2605 }
2606
2607
2608 void Paragraph::setPlainLayout(DocumentClass const & tc)
2609 {
2610         setLayout(tc.plainLayout());
2611 }
2612
2613
2614 void Paragraph::setPlainOrDefaultLayout(DocumentClass const & tclass)
2615 {
2616         if (usePlainLayout())
2617                 setPlainLayout(tclass);
2618         else
2619                 setDefaultLayout(tclass);
2620 }
2621
2622
2623 Inset const & Paragraph::inInset() const
2624 {
2625         LASSERT(d->inset_owner_, throw ExceptionMessage(BufferException,
2626                 _("Memory problem"), _("Paragraph not properly initialized")));
2627         return *d->inset_owner_;
2628 }
2629
2630
2631 InsetCode Paragraph::ownerCode() const
2632 {
2633         return d->inset_owner_ ? d->inset_owner_->lyxCode() : NO_CODE;
2634 }
2635
2636
2637 ParagraphParameters & Paragraph::params()
2638 {
2639         return d->params_;
2640 }
2641
2642
2643 ParagraphParameters const & Paragraph::params() const
2644 {
2645         return d->params_;
2646 }
2647
2648
2649 bool Paragraph::isFreeSpacing() const
2650 {
2651         if (d->layout_->free_spacing)
2652                 return true;
2653         return d->inset_owner_ && d->inset_owner_->isFreeSpacing();
2654 }
2655
2656
2657 bool Paragraph::allowEmpty() const
2658 {
2659         if (d->layout_->keepempty)
2660                 return true;
2661         return d->inset_owner_ && d->inset_owner_->allowEmpty();
2662 }
2663
2664
2665 char_type Paragraph::transformChar(char_type c, pos_type pos) const
2666 {
2667         if (!Encodings::isArabicChar(c))
2668                 return c;
2669
2670         char_type prev_char = ' ';
2671         char_type next_char = ' ';
2672
2673         for (pos_type i = pos - 1; i >= 0; --i) {
2674                 char_type const par_char = d->text_[i];
2675                 if (!Encodings::isArabicComposeChar(par_char)) {
2676                         prev_char = par_char;
2677                         break;
2678                 }
2679         }
2680
2681         for (pos_type i = pos + 1, end = size(); i < end; ++i) {
2682                 char_type const par_char = d->text_[i];
2683                 if (!Encodings::isArabicComposeChar(par_char)) {
2684                         next_char = par_char;
2685                         break;
2686                 }
2687         }
2688
2689         if (Encodings::isArabicChar(next_char)) {
2690                 if (Encodings::isArabicChar(prev_char) &&
2691                         !Encodings::isArabicSpecialChar(prev_char))
2692                         return Encodings::transformChar(c, Encodings::FORM_MEDIAL);
2693                 else
2694                         return Encodings::transformChar(c, Encodings::FORM_INITIAL);
2695         } else {
2696                 if (Encodings::isArabicChar(prev_char) &&
2697                         !Encodings::isArabicSpecialChar(prev_char))
2698                         return Encodings::transformChar(c, Encodings::FORM_FINAL);
2699                 else
2700                         return Encodings::transformChar(c, Encodings::FORM_ISOLATED);
2701         }
2702 }
2703
2704
2705 int Paragraph::checkBiblio(Buffer const & buffer)
2706 {
2707         // FIXME From JS:
2708         // This is getting more and more a mess. ...We really should clean
2709         // up this bibitem issue for 1.6. See also bug 2743.
2710
2711         // Add bibitem insets if necessary
2712         if (d->layout_->labeltype != LABEL_BIBLIO)
2713                 return 0;
2714
2715         bool hasbibitem = !d->insetlist_.empty()
2716                 // Insist on it being in pos 0
2717                 && d->text_[0] == META_INSET
2718                 && d->insetlist_.begin()->inset->lyxCode() == BIBITEM_CODE;
2719
2720         bool track_changes = buffer.params().trackChanges;
2721
2722         docstring oldkey;
2723         docstring oldlabel;
2724
2725         // remove a bibitem in pos != 0
2726         // restore it later in pos 0 if necessary
2727         // (e.g. if a user inserts contents _before_ the item)
2728         // we're assuming there's only one of these, which there
2729         // should be.
2730         int erasedInsetPosition = -1;
2731         InsetList::iterator it = d->insetlist_.begin();
2732         InsetList::iterator end = d->insetlist_.end();
2733         for (; it != end; ++it)
2734                 if (it->inset->lyxCode() == BIBITEM_CODE
2735                     && it->pos > 0) {
2736                         InsetBibitem * olditem = static_cast<InsetBibitem *>(it->inset);
2737                         oldkey = olditem->getParam("key");
2738                         oldlabel = olditem->getParam("label");
2739                         erasedInsetPosition = it->pos;
2740                         eraseChar(erasedInsetPosition, track_changes);
2741                         break;
2742         }
2743
2744         // There was an InsetBibitem at the beginning, and we didn't
2745         // have to erase one.
2746         if (hasbibitem && erasedInsetPosition < 0)
2747                         return 0;
2748
2749         // There was an InsetBibitem at the beginning and we did have to
2750         // erase one. So we give its properties to the beginning inset.
2751         if (hasbibitem) {
2752                 InsetBibitem * inset =
2753                         static_cast<InsetBibitem *>(d->insetlist_.begin()->inset);
2754                 if (!oldkey.empty())
2755                         inset->setParam("key", oldkey);
2756                 inset->setParam("label", oldlabel);
2757                 return -erasedInsetPosition;
2758         }
2759
2760         // There was no inset at the beginning, so we need to create one with
2761         // the key and label of the one we erased.
2762         InsetBibitem * inset =
2763                 new InsetBibitem(buffer, InsetCommandParams(BIBITEM_CODE));
2764         // restore values of previously deleted item in this par.
2765         if (!oldkey.empty())
2766                 inset->setParam("key", oldkey);
2767         inset->setParam("label", oldlabel);
2768         insertInset(0, static_cast<Inset *>(inset),
2769                     Change(track_changes ? Change::INSERTED : Change::UNCHANGED));
2770
2771         return 1;
2772 }
2773
2774
2775 void Paragraph::checkAuthors(AuthorList const & authorList)
2776 {
2777         d->changes_.checkAuthors(authorList);
2778 }
2779
2780
2781 bool Paragraph::isUnchanged(pos_type pos) const
2782 {
2783         return lookupChange(pos).type == Change::UNCHANGED;
2784 }
2785
2786
2787 bool Paragraph::isInserted(pos_type pos) const
2788 {
2789         return lookupChange(pos).type == Change::INSERTED;
2790 }
2791
2792
2793 bool Paragraph::isDeleted(pos_type pos) const
2794 {
2795         return lookupChange(pos).type == Change::DELETED;
2796 }
2797
2798
2799 InsetList const & Paragraph::insetList() const
2800 {
2801         return d->insetlist_;
2802 }
2803
2804
2805 void Paragraph::setBuffer(Buffer & b)
2806 {
2807         d->insetlist_.setBuffer(b);
2808 }
2809
2810
2811 Inset * Paragraph::releaseInset(pos_type pos)
2812 {
2813         Inset * inset = d->insetlist_.release(pos);
2814         /// does not honour change tracking!
2815         eraseChar(pos, false);
2816         return inset;
2817 }
2818
2819
2820 Inset * Paragraph::getInset(pos_type pos)
2821 {
2822         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
2823                  ? d->insetlist_.get(pos) : 0;
2824 }
2825
2826
2827 Inset const * Paragraph::getInset(pos_type pos) const
2828 {
2829         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
2830                  ? d->insetlist_.get(pos) : 0;
2831 }
2832
2833
2834 void Paragraph::changeCase(BufferParams const & bparams, pos_type pos,
2835                 pos_type & right, TextCase action)
2836 {
2837         // process sequences of modified characters; in change
2838         // tracking mode, this approach results in much better
2839         // usability than changing case on a char-by-char basis
2840         docstring changes;
2841
2842         bool const trackChanges = bparams.trackChanges;
2843
2844         bool capitalize = true;
2845
2846         for (; pos < right; ++pos) {
2847                 char_type oldChar = d->text_[pos];
2848                 char_type newChar = oldChar;
2849
2850                 // ignore insets and don't play with deleted text!
2851                 if (oldChar != META_INSET && !isDeleted(pos)) {
2852                         switch (action) {
2853                                 case text_lowercase:
2854                                         newChar = lowercase(oldChar);
2855                                         break;
2856                                 case text_capitalization:
2857                                         if (capitalize) {
2858                                                 newChar = uppercase(oldChar);
2859                                                 capitalize = false;
2860                                         }
2861                                         break;
2862                                 case text_uppercase:
2863                                         newChar = uppercase(oldChar);
2864                                         break;
2865                         }
2866                 }
2867
2868                 if (isWordSeparator(pos) || isDeleted(pos)) {
2869                         // permit capitalization again
2870                         capitalize = true;
2871                 }
2872
2873                 if (oldChar != newChar) {
2874                         changes += newChar;
2875                         if (pos != right - 1)
2876                                 continue;
2877                         // step behind the changing area
2878                         pos++;
2879                 }
2880
2881                 int erasePos = pos - changes.size();
2882                 for (size_t i = 0; i < changes.size(); i++) {
2883                         insertChar(pos, changes[i],
2884                                    getFontSettings(bparams,
2885                                                    erasePos),
2886                                    trackChanges);
2887                         if (!eraseChar(erasePos, trackChanges)) {
2888                                 ++erasePos;
2889                                 ++pos; // advance
2890                                 ++right; // expand selection
2891                         }
2892                 }
2893                 changes.clear();
2894         }
2895 }
2896
2897
2898 bool Paragraph::find(docstring const & str, bool cs, bool mw,
2899                 pos_type pos, bool del) const
2900 {
2901         int const strsize = str.length();
2902         int i = 0;
2903         pos_type const parsize = d->text_.size();
2904         for (i = 0; pos + i < parsize; ++i) {
2905                 if (i >= strsize)
2906                         break;
2907                 if (cs && str[i] != d->text_[pos + i])
2908                         break;
2909                 if (!cs && uppercase(str[i]) != uppercase(d->text_[pos + i]))
2910                         break;
2911                 if (!del && isDeleted(pos + i))
2912                         break;
2913         }
2914
2915         if (i != strsize)
2916                 return false;
2917
2918         // if necessary, check whether string matches word
2919         if (mw) {
2920                 if (pos > 0 && !isWordSeparator(pos - 1))
2921                         return false;
2922                 if (pos + strsize < parsize
2923                         && !isWordSeparator(pos + strsize))
2924                         return false;
2925         }
2926
2927         return true;
2928 }
2929
2930
2931 char_type Paragraph::getChar(pos_type pos) const
2932 {
2933         return d->text_[pos];
2934 }
2935
2936
2937 pos_type Paragraph::size() const
2938 {
2939         return d->text_.size();
2940 }
2941
2942
2943 bool Paragraph::empty() const
2944 {
2945         return d->text_.empty();
2946 }
2947
2948
2949 bool Paragraph::isInset(pos_type pos) const
2950 {
2951         return d->text_[pos] == META_INSET;
2952 }
2953
2954
2955 bool Paragraph::isSeparator(pos_type pos) const
2956 {
2957         //FIXME: Are we sure this can be the only separator?
2958         return d->text_[pos] == ' ';
2959 }
2960
2961
2962 void Paragraph::deregisterWords()
2963 {
2964         Private::Words::const_iterator it;
2965         WordList & wl = theWordList();
2966         for (it = d->words_.begin(); it != d->words_.end(); ++it)
2967                 wl.remove(*it);
2968         d->words_.clear();
2969 }
2970
2971
2972 void Paragraph::locateWord(pos_type & from, pos_type & to,
2973         word_location const loc) const
2974 {
2975         switch (loc) {
2976         case WHOLE_WORD_STRICT:
2977                 if (from == 0 || from == size()
2978                     || isWordSeparator(from)
2979                     || isWordSeparator(from - 1)) {
2980                         to = from;
2981                         return;
2982                 }
2983                 // no break here, we go to the next
2984
2985         case WHOLE_WORD:
2986                 // If we are already at the beginning of a word, do nothing
2987                 if (!from || isWordSeparator(from - 1))
2988                         break;
2989                 // no break here, we go to the next
2990
2991         case PREVIOUS_WORD:
2992                 // always move the cursor to the beginning of previous word
2993                 while (from && !isWordSeparator(from - 1))
2994                         --from;
2995                 break;
2996         case NEXT_WORD:
2997                 LYXERR0("Paragraph::locateWord: NEXT_WORD not implemented yet");
2998                 break;
2999         case PARTIAL_WORD:
3000                 // no need to move the 'from' cursor
3001                 break;
3002         }
3003         to = from;
3004         while (to < size() && !isWordSeparator(to))
3005                 ++to;
3006 }
3007
3008
3009 void Paragraph::collectWords()
3010 {
3011         pos_type n = size();
3012         WordLangTuple wl;
3013         docstring_list suggestions;
3014         for (pos_type pos = 0; pos < n; ++pos) {
3015                 if (isWordSeparator(pos))
3016                         continue;
3017                 pos_type from = pos;
3018                 locateWord(from, pos, WHOLE_WORD);
3019                 if (pos - from >= 6) {
3020                         docstring word = asString(from, pos, AS_STR_NONE);
3021                         d->words_.insert(word);
3022                 }
3023                 if (lyxrc.spellcheck_continuously
3024                     && spellCheck(from, pos, wl, suggestions)) {
3025                         for (size_t i = 0; i != suggestions.size(); ++i)
3026                                 d->words_.insert(suggestions[i]);
3027                 }
3028         }
3029 }
3030
3031
3032 void Paragraph::registerWords()
3033 {
3034         Private::Words::const_iterator it;
3035         WordList & wl = theWordList();
3036         for (it = d->words_.begin(); it != d->words_.end(); ++it)
3037                 wl.insert(*it);
3038 }
3039
3040
3041 void Paragraph::updateWords()
3042 {
3043         deregisterWords();
3044         collectWords();
3045         registerWords();
3046 }
3047
3048
3049 bool Paragraph::spellCheck(pos_type & from, pos_type & to, WordLangTuple & wl,
3050         docstring_list & suggestions) const
3051 {
3052         SpellChecker * speller = theSpellChecker();
3053         if (!speller)
3054                 return false;
3055
3056         locateWord(from, to, WHOLE_WORD);
3057         if (from == to || from >= pos_type(d->text_.size()))
3058                 return false;
3059
3060         docstring word = asString(from, to, AS_STR_INSETS);
3061         string const lang_code = lyxrc.spellchecker_alt_lang.empty()
3062                 ? getFontSettings(d->inset_owner_->buffer().params(), from).language()->code()
3063                 : lyxrc.spellchecker_alt_lang;
3064         wl = WordLangTuple(word, lang_code);
3065         SpellChecker::Result res = speller->check(wl);
3066         // Just ignore any error that the spellchecker reports.
3067         // FIXME: we should through out an exception and catch it in the GUI to
3068         // display the error.
3069         if (!speller->error().empty())
3070                 return false;
3071
3072         bool const misspelled = res != SpellChecker::OK
3073                 && res != SpellChecker::IGNORED_WORD;
3074
3075         if (lyxrc.spellcheck_continuously)
3076                 d->fontlist_.setMisspelled(from, to, misspelled);
3077
3078         if (misspelled) {
3079                 while (!(word = speller->nextMiss()).empty())
3080                         suggestions.push_back(word);
3081         }
3082         return misspelled;
3083 }
3084
3085
3086 bool Paragraph::isMisspelled(pos_type pos) const
3087 {
3088         pos_type from = pos;
3089         pos_type to = pos;
3090         WordLangTuple wl;
3091         docstring_list suggestions;
3092         return spellCheck(from, to, wl, suggestions);
3093 }
3094
3095
3096 } // namespace lyx