src/Paragraph.cpp

   1 /**
   2  * \file Paragraph.cpp
   3  * This file is part of LyX, the document processor.
   4  * Licence details can be found in the file COPYING.
   5  *
   6  * \author Asger Alstrup
   7  * \author Lars Gullik Bjønnes
   8  * \author Richard Heck (XHTML output)
   9  * \author Jean-Marc Lasgouttes
  10  * \author Angus Leeming
  11  * \author John Levon
  12  * \author André Pönitz
  13  * \author Dekel Tsur
  14  * \author Jürgen Vigna
  15  *
  16  * Full author contact details are available in file CREDITS.
  17  */
  18
  19 #include <config.h>
  20
  21 #include "Paragraph.h"
  22
  23 #include "LayoutFile.h"
  24 #include "Buffer.h"
  25 #include "BufferParams.h"
  26 #include "Changes.h"
  27 #include "Counters.h"
  28 #include "Encoding.h"
  29 #include "InsetList.h"
  30 #include "Language.h"
  31 #include "LaTeXFeatures.h"
  32 #include "Layout.h"
  33 #include "Length.h"
  34 #include "Font.h"
  35 #include "FontList.h"
  36 #include "LyXRC.h"
  37 #include "OutputParams.h"
  38 #include "output_latex.h"
  39 #include "output_xhtml.h"
  40 #include "ParagraphParameters.h"
  41 #include "SpellChecker.h"
  42 #include "sgml.h"
  43 #include "TextClass.h"
  44 #include "TexRow.h"
  45 #include "Text.h"
  46 #include "VSpace.h"
  47 #include "WordLangTuple.h"
  48 #include "WordList.h"
  49
  50 #include "frontends/alert.h"
  51
  52 #include "insets/InsetBibitem.h"
  53 #include "insets/InsetLabel.h"
  54
  55 #include "support/debug.h"
  56 #include "support/docstring_list.h"
  57 #include "support/ExceptionMessage.h"
  58 #include "support/gettext.h"
  59 #include "support/lassert.h"
  60 #include "support/lstrings.h"
  61 #include "support/textutils.h"
  62
  63 #include <sstream>
  64 #include <vector>
  65
  66 using namespace std;
  67 using namespace lyx::support;
  68
  69 namespace lyx {
  70
  71 namespace {
  72 /// Inset identifier (above 0x10ffff, for ucs-4)
  73 char_type const META_INSET = 0x200001;
  74 };
  75
  76 /////////////////////////////////////////////////////////////////////
  77 //
  78 // Paragraph::Private
  79 //
  80 /////////////////////////////////////////////////////////////////////
  81
  82 class Paragraph::Private
  83 {
  84 public:
  85         ///
  86         Private(Paragraph * owner, Layout const & layout);
  87         /// "Copy constructor"
  88         Private(Private const &, Paragraph * owner);
  89         /// Copy constructor from \p beg  to \p end
  90         Private(Private const &, Paragraph * owner, pos_type beg, pos_type end);
  91
  92         ///
  93         void insertChar(pos_type pos, char_type c, Change const & change);
  94
  95         /// Output the surrogate pair formed by \p c and \p next to \p os.
  96         /// \return the number of characters written.
  97         int latexSurrogatePair(odocstream & os, char_type c, char_type next,
  98                                OutputParams const &);
  99
 100         /// Output a space in appropriate formatting (or a surrogate pair
 101         /// if the next character is a combining character).
 102         /// \return whether a surrogate pair was output.
 103         bool simpleTeXBlanks(OutputParams const &,
 104                              odocstream &, TexRow & texrow,
 105                              pos_type i,
 106                              unsigned int & column,
 107                              Font const & font,
 108                              Layout const & style);
 109
 110         /// Output consecutive unicode chars, belonging to the same script as
 111         /// specified by the latex macro \p ltx, to \p os starting from \p i.
 112         /// \return the number of characters written.
 113         int writeScriptChars(odocstream & os, docstring const & ltx,
 114                            Change const &, Encoding const &, pos_type & i);
 115
 116         /// This could go to ParagraphParameters if we want to.
 117         int startTeXParParams(BufferParams const &, odocstream &, TexRow &,
 118                               OutputParams const &) const;
 119
 120         /// This could go to ParagraphParameters if we want to.
 121         int endTeXParParams(BufferParams const &, odocstream &, TexRow &,
 122                             OutputParams const &) const;
 123
 124         ///
 125         void latexInset(BufferParams const &,
 126                                    odocstream &,
 127                                    TexRow & texrow, OutputParams &,
 128                                    Font & running_font,
 129                                    Font & basefont,
 130                                    Font const & outerfont,
 131                                    bool & open_font,
 132                                    Change & running_change,
 133                                    Layout const & style,
 134                                    pos_type & i,
 135                                    unsigned int & column);
 136
 137         ///
 138         void latexSpecialChar(
 139                                    odocstream & os,
 140                                    OutputParams const & runparams,
 141                                    Font const & running_font,
 142                                    Change const & running_change,
 143                                    Layout const & style,
 144                                    pos_type & i,
 145                                    unsigned int & column);
 146
 147         ///
 148         bool latexSpecialT1(
 149                 char_type const c,
 150                 odocstream & os,
 151                 pos_type i,
 152                 unsigned int & column);
 153         ///
 154         bool latexSpecialTypewriter(
 155                 char_type const c,
 156                 odocstream & os,
 157                 pos_type i,
 158                 unsigned int & column);
 159         ///
 160         bool latexSpecialPhrase(
 161                 odocstream & os,
 162                 pos_type & i,
 163                 unsigned int & column,
 164                 OutputParams const & runparams);
 165
 166         ///
 167         void validate(LaTeXFeatures & features) const;
 168
 169         /// Checks if the paragraph contains only text and no inset or font change.
 170         bool onlyText(Buffer const & buf, Font const & outerfont,
 171                       pos_type initial) const;
 172
 173         /// match a string against a particular point in the paragraph
 174         bool isTextAt(string const & str, pos_type pos) const;
 175
 176
 177         InsetCode ownerCode() const
 178         {
 179                 return inset_owner_ ? inset_owner_->lyxCode() : NO_CODE;
 180         }
 181
 182         /// Which Paragraph owns us?
 183         Paragraph * owner_;
 184
 185         /// In which Inset?
 186         Inset const * inset_owner_;
 187
 188         ///
 189         FontList fontlist_;
 190
 191         ///
 192         int id_;
 193
 194         ///
 195         ParagraphParameters params_;
 196
 197         /// for recording and looking up changes
 198         Changes changes_;
 199
 200         ///
 201         InsetList insetlist_;
 202
 203         /// end of label
 204         pos_type begin_of_body_;
 205
 206         typedef docstring TextContainer;
 207         ///
 208         TextContainer text_;
 209
 210         typedef set<docstring> Words;
 211         ///
 212         map<Language, Words> words_;
 213         ///
 214         Layout const * layout_;
 215 };
 216
 217
 218 namespace {
 219
 220 struct special_phrase {
 221         string phrase;
 222         docstring macro;
 223         bool builtin;
 224 };
 225
 226 special_phrase const special_phrases[] = {
 227         { "LyX", from_ascii("\\LyX{}"), false },
 228         { "TeX", from_ascii("\\TeX{}"), true },
 229         { "LaTeX2e", from_ascii("\\LaTeXe{}"), true },
 230         { "LaTeX", from_ascii("\\LaTeX{}"), true },
 231 };
 232
 233 size_t const phrases_nr = sizeof(special_phrases)/sizeof(special_phrase);
 234
 235 } // namespace anon
 236
 237
 238 Paragraph::Private::Private(Paragraph * owner, Layout const & layout)
 239         : owner_(owner), inset_owner_(0), id_(-1), begin_of_body_(0), layout_(&layout)
 240 {
 241         text_.reserve(100);
 242 }
 243
 244
 245 // Initialization of the counter for the paragraph id's,
 246 //
 247 // FIXME: There should be a more intelligent way to generate and use the
 248 // paragraph ids per buffer instead a global static counter for all InsetText
 249 // in the running program.
 250 static int paragraph_id = -1;
 251
 252 Paragraph::Private::Private(Private const & p, Paragraph * owner)
 253         : owner_(owner), inset_owner_(p.inset_owner_), fontlist_(p.fontlist_),
 254           params_(p.params_), changes_(p.changes_), insetlist_(p.insetlist_),
 255           begin_of_body_(p.begin_of_body_), text_(p.text_), words_(p.words_),
 256           layout_(p.layout_)
 257 {
 258         id_ = ++paragraph_id;
 259 }
 260
 261
 262 Paragraph::Private::Private(Private const & p, Paragraph * owner,
 263         pos_type beg, pos_type end)
 264         : owner_(owner), inset_owner_(p.inset_owner_),
 265           params_(p.params_), changes_(p.changes_),
 266           insetlist_(p.insetlist_, beg, end),
 267           begin_of_body_(p.begin_of_body_), words_(p.words_),
 268           layout_(p.layout_)
 269 {
 270         id_ = ++paragraph_id;
 271         if (beg >= pos_type(p.text_.size()))
 272                 return;
 273         text_ = p.text_.substr(beg, end - beg);
 274
 275         FontList::const_iterator fcit = fontlist_.begin();
 276         FontList::const_iterator fend = fontlist_.end();
 277         for (; fcit != fend; ++fcit) {
 278                 if (fcit->pos() < beg)
 279                         continue;
 280                 if (fcit->pos() >= end) {
 281                         // Add last entry in the fontlist_.
 282                         fontlist_.set(text_.size() - 1, fcit->font());
 283                         break;
 284                 }
 285                 // Add a new entry in the fontlist_.
 286                 fontlist_.set(fcit->pos() - beg, fcit->font());
 287         }
 288 }
 289
 290
 291 void Paragraph::addChangesToToc(DocIterator const & cdit,
 292         Buffer const & buf) const
 293 {
 294         d->changes_.addToToc(cdit, buf);
 295 }
 296
 297
 298 bool Paragraph::isDeleted(pos_type start, pos_type end) const
 299 {
 300         LASSERT(start >= 0 && start <= size(), /**/);
 301         LASSERT(end > start && end <= size() + 1, /**/);
 302
 303         return d->changes_.isDeleted(start, end);
 304 }
 305
 306
 307 bool Paragraph::isChanged(pos_type start, pos_type end) const
 308 {
 309         LASSERT(start >= 0 && start <= size(), /**/);
 310         LASSERT(end > start && end <= size() + 1, /**/);
 311
 312         return d->changes_.isChanged(start, end);
 313 }
 314
 315
 316 bool Paragraph::isMergedOnEndOfParDeletion(bool trackChanges) const
 317 {
 318         // keep the logic here in sync with the logic of eraseChars()
 319         if (!trackChanges)
 320                 return true;
 321
 322         Change const change = d->changes_.lookup(size());
 323         return change.inserted() && change.currentAuthor();
 324 }
 325
 326
 327 void Paragraph::setChange(Change const & change)
 328 {
 329         // beware of the imaginary end-of-par character!
 330         d->changes_.set(change, 0, size() + 1);
 331
 332         /*
 333          * Propagate the change recursively - but not in case of DELETED!
 334          *
 335          * Imagine that your co-author makes changes in an existing inset. He
 336          * sends your document to you and you come to the conclusion that the
 337          * inset should go completely. If you erase it, LyX must not delete all
 338          * text within the inset. Otherwise, the change tracked insertions of
 339          * your co-author get lost and there is no way to restore them later.
 340          *
 341          * Conclusion: An inset's content should remain untouched if you delete it
 342          */
 343
 344         if (!change.deleted()) {
 345                 for (pos_type pos = 0; pos < size(); ++pos) {
 346                         if (Inset * inset = getInset(pos))
 347                                 inset->setChange(change);
 348                 }
 349         }
 350 }
 351
 352
 353 void Paragraph::setChange(pos_type pos, Change const & change)
 354 {
 355         LASSERT(pos >= 0 && pos <= size(), /**/);
 356         d->changes_.set(change, pos);
 357
 358         // see comment in setChange(Change const &) above
 359         if (!change.deleted() && pos < size())
 360                         if (Inset * inset = getInset(pos))
 361                                 inset->setChange(change);
 362 }
 363
 364
 365 Change const & Paragraph::lookupChange(pos_type pos) const
 366 {
 367         LASSERT(pos >= 0 && pos <= size(), /**/);
 368         return d->changes_.lookup(pos);
 369 }
 370
 371
 372 void Paragraph::acceptChanges(pos_type start, pos_type end)
 373 {
 374         LASSERT(start >= 0 && start <= size(), /**/);
 375         LASSERT(end > start && end <= size() + 1, /**/);
 376
 377         for (pos_type pos = start; pos < end; ++pos) {
 378                 switch (lookupChange(pos).type) {
 379                         case Change::UNCHANGED:
 380                                 // accept changes in nested inset
 381                                 if (Inset * inset = getInset(pos))
 382                                         inset->acceptChanges();
 383                                 break;
 384
 385                         case Change::INSERTED:
 386                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 387                                 // also accept changes in nested inset
 388                                 if (Inset * inset = getInset(pos))
 389                                         inset->acceptChanges();
 390                                 break;
 391
 392                         case Change::DELETED:
 393                                 // Suppress access to non-existent
 394                                 // "end-of-paragraph char"
 395                                 if (pos < size()) {
 396                                         eraseChar(pos, false);
 397                                         --end;
 398                                         --pos;
 399                                 }
 400                                 break;
 401                 }
 402
 403         }
 404 }
 405
 406
 407 void Paragraph::rejectChanges(pos_type start, pos_type end)
 408 {
 409         LASSERT(start >= 0 && start <= size(), /**/);
 410         LASSERT(end > start && end <= size() + 1, /**/);
 411
 412         for (pos_type pos = start; pos < end; ++pos) {
 413                 switch (lookupChange(pos).type) {
 414                         case Change::UNCHANGED:
 415                                 // reject changes in nested inset
 416                                 if (Inset * inset = getInset(pos))
 417                                                 inset->rejectChanges();
 418                                 break;
 419
 420                         case Change::INSERTED:
 421                                 // Suppress access to non-existent
 422                                 // "end-of-paragraph char"
 423                                 if (pos < size()) {
 424                                         eraseChar(pos, false);
 425                                         --end;
 426                                         --pos;
 427                                 }
 428                                 break;
 429
 430                         case Change::DELETED:
 431                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 432
 433                                 // Do NOT reject changes within a deleted inset!
 434                                 // There may be insertions of a co-author inside of it!
 435
 436                                 break;
 437                 }
 438         }
 439 }
 440
 441
 442 void Paragraph::Private::insertChar(pos_type pos, char_type c,
 443                 Change const & change)
 444 {
 445         LASSERT(pos >= 0 && pos <= int(text_.size()), /**/);
 446
 447         // track change
 448         changes_.insert(change, pos);
 449
 450         // This is actually very common when parsing buffers (and
 451         // maybe inserting ascii text)
 452         if (pos == pos_type(text_.size())) {
 453                 // when appending characters, no need to update tables
 454                 text_.push_back(c);
 455                 return;
 456         }
 457
 458         text_.insert(text_.begin() + pos, c);
 459
 460         // Update the font table.
 461         fontlist_.increasePosAfterPos(pos);
 462
 463         // Update the insets
 464         insetlist_.increasePosAfterPos(pos);
 465 }
 466
 467
 468 bool Paragraph::insertInset(pos_type pos, Inset * inset,
 469                                    Change const & change)
 470 {
 471         LASSERT(inset, /**/);
 472         LASSERT(pos >= 0 && pos <= size(), /**/);
 473
 474         // Paragraph::insertInset() can be used in cut/copy/paste operation where
 475         // d->inset_owner_ is not set yet.
 476         if (d->inset_owner_ && !d->inset_owner_->insetAllowed(inset->lyxCode()))
 477                 return false;
 478
 479         d->insertChar(pos, META_INSET, change);
 480         LASSERT(d->text_[pos] == META_INSET, /**/);
 481
 482         // Add a new entry in the insetlist_.
 483         d->insetlist_.insert(inset, pos);
 484         return true;
 485 }
 486
 487
 488 bool Paragraph::eraseChar(pos_type pos, bool trackChanges)
 489 {
 490         LASSERT(pos >= 0 && pos <= size(), return false);
 491
 492         // keep the logic here in sync with the logic of isMergedOnEndOfParDeletion()
 493
 494         if (trackChanges) {
 495                 Change change = d->changes_.lookup(pos);
 496
 497                 // set the character to DELETED if
 498                 //  a) it was previously unchanged or
 499                 //  b) it was inserted by a co-author
 500
 501                 if (!change.changed() ||
 502                       (change.inserted() && !change.currentAuthor())) {
 503                         setChange(pos, Change(Change::DELETED));
 504                         return false;
 505                 }
 506
 507                 if (change.deleted())
 508                         return false;
 509         }
 510
 511         // Don't physically access the imaginary end-of-paragraph character.
 512         // eraseChar() can only mark it as DELETED. A physical deletion of
 513         // end-of-par must be handled externally.
 514         if (pos == size()) {
 515                 return false;
 516         }
 517
 518         // track change
 519         d->changes_.erase(pos);
 520
 521         // if it is an inset, delete the inset entry
 522         if (d->text_[pos] == META_INSET)
 523                 d->insetlist_.erase(pos);
 524
 525         d->text_.erase(d->text_.begin() + pos);
 526
 527         // Update the fontlist_
 528         d->fontlist_.erase(pos);
 529
 530         // Update the insetlist_
 531         d->insetlist_.decreasePosAfterPos(pos);
 532
 533         return true;
 534 }
 535
 536
 537 int Paragraph::eraseChars(pos_type start, pos_type end, bool trackChanges)
 538 {
 539         LASSERT(start >= 0 && start <= size(), /**/);
 540         LASSERT(end >= start && end <= size() + 1, /**/);
 541
 542         pos_type i = start;
 543         for (pos_type count = end - start; count; --count) {
 544                 if (!eraseChar(i, trackChanges))
 545                         ++i;
 546         }
 547         return end - i;
 548 }
 549
 550
 551 int Paragraph::Private::latexSurrogatePair(odocstream & os, char_type c,
 552                 char_type next, OutputParams const & runparams)
 553 {
 554         // Writing next here may circumvent a possible font change between
 555         // c and next. Since next is only output if it forms a surrogate pair
 556         // with c we can ignore this:
 557         // A font change inside a surrogate pair does not make sense and is
 558         // hopefully impossible to input.
 559         // FIXME: change tracking
 560         // Is this correct WRT change tracking?
 561         Encoding const & encoding = *(runparams.encoding);
 562         docstring const latex1 = encoding.latexChar(next);
 563         docstring const latex2 = encoding.latexChar(c);
 564         if (docstring(1, next) == latex1) {
 565                 // the encoding supports the combination
 566                 os << latex2 << latex1;
 567                 return latex1.length() + latex2.length();
 568         } else if (runparams.local_font &&
 569                    runparams.local_font->language()->lang() == "polutonikogreek") {
 570                 // polutonikogreek only works without the brackets
 571                 os << latex1 << latex2;
 572                 return latex1.length() + latex2.length();
 573         } else
 574                 os << latex1 << '{' << latex2 << '}';
 575         return latex1.length() + latex2.length() + 2;
 576 }
 577
 578
 579 bool Paragraph::Private::simpleTeXBlanks(OutputParams const & runparams,
 580                                        odocstream & os, TexRow & texrow,
 581                                        pos_type i,
 582                                        unsigned int & column,
 583                                        Font const & font,
 584                                        Layout const & style)
 585 {
 586         if (style.pass_thru || runparams.verbatim)
 587                 return false;
 588
 589         if (i + 1 < int(text_.size())) {
 590                 char_type next = text_[i + 1];
 591                 if (Encodings::isCombiningChar(next)) {
 592                         // This space has an accent, so we must always output it.
 593                         column += latexSurrogatePair(os, ' ', next, runparams) - 1;
 594                         return true;
 595                 }
 596         }
 597
 598         if (runparams.linelen > 0
 599             && column > runparams.linelen
 600             && i
 601             && text_[i - 1] != ' '
 602             && (i + 1 < int(text_.size()))
 603             // same in FreeSpacing mode
 604             && !owner_->isFreeSpacing()
 605             // In typewriter mode, we want to avoid
 606             // ! . ? : at the end of a line
 607             && !(font.fontInfo().family() == TYPEWRITER_FAMILY
 608                  && (text_[i - 1] == '.'
 609                      || text_[i - 1] == '?'
 610                      || text_[i - 1] == ':'
 611                      || text_[i - 1] == '!'))) {
 612                 os << '\n';
 613                 texrow.newline();
 614                 texrow.start(owner_->id(), i + 1);
 615                 column = 0;
 616         } else if (style.free_spacing) {
 617                 os << '~';
 618         } else {
 619                 os << ' ';
 620         }
 621         return false;
 622 }
 623
 624
 625 int Paragraph::Private::writeScriptChars(odocstream & os,
 626                                          docstring const & ltx,
 627                                          Change const & runningChange,
 628                                          Encoding const & encoding,
 629                                          pos_type & i)
 630 {
 631         // FIXME: modifying i here is not very nice...
 632
 633         // We only arrive here when a proper language for character text_[i] has
 634         // not been specified (i.e., it could not be translated in the current
 635         // latex encoding) or its latex translation has been forced, and it
 636         // belongs to a known script.
 637         // Parameter ltx contains the latex translation of text_[i] as specified
 638         // in the unicodesymbols file and is something like "\textXXX{<spec>}".
 639         // The latex macro name "textXXX" specifies the script to which text_[i]
 640         // belongs and we use it in order to check whether characters from the
 641         // same script immediately follow, such that we can collect them in a
 642         // single "\textXXX" macro. So, we have to retain "\textXXX{<spec>"
 643         // for the first char but only "<spec>" for all subsequent chars.
 644         docstring::size_type const brace1 = ltx.find_first_of(from_ascii("{"));
 645         docstring::size_type const brace2 = ltx.find_last_of(from_ascii("}"));
 646         string script = to_ascii(ltx.substr(1, brace1 - 1));
 647         int pos = 0;
 648         int length = brace2;
 649         bool closing_brace = true;
 650         if (script == "textgreek" && encoding.latexName() == "iso-8859-7") {
 651                 // Correct encoding is being used, so we can avoid \textgreek.
 652                 pos = brace1 + 1;
 653                 length -= pos;
 654                 closing_brace = false;
 655         }
 656         os << ltx.substr(pos, length);
 657         int size = text_.size();
 658         while (i + 1 < size) {
 659                 char_type const next = text_[i + 1];
 660                 // Stop here if next character belongs to another script
 661                 // or there is a change in change tracking status.
 662                 if (!Encodings::isKnownScriptChar(next, script) ||
 663                     runningChange != owner_->lookupChange(i + 1))
 664                         break;
 665                 Font prev_font;
 666                 bool found = false;
 667                 FontList::const_iterator cit = fontlist_.begin();
 668                 FontList::const_iterator end = fontlist_.end();
 669                 for (; cit != end; ++cit) {
 670                         if (cit->pos() >= i && !found) {
 671                                 prev_font = cit->font();
 672                                 found = true;
 673                         }
 674                         if (cit->pos() >= i + 1)
 675                                 break;
 676                 }
 677                 // Stop here if there is a font attribute or encoding change.
 678                 if (found && cit != end && prev_font != cit->font())
 679                         break;
 680                 docstring const latex = encoding.latexChar(next);
 681                 docstring::size_type const b1 =
 682                                         latex.find_first_of(from_ascii("{"));
 683                 docstring::size_type const b2 =
 684                                         latex.find_last_of(from_ascii("}"));
 685                 int const len = b2 - b1 - 1;
 686                 os << latex.substr(b1 + 1, len);
 687                 length += len;
 688                 ++i;
 689         }
 690         if (closing_brace) {
 691                 os << '}';
 692                 ++length;
 693         }
 694         return length;
 695 }
 696
 697
 698 bool Paragraph::Private::isTextAt(string const & str, pos_type pos) const
 699 {
 700         pos_type const len = str.length();
 701
 702         // is the paragraph large enough?
 703         if (pos + len > int(text_.size()))
 704                 return false;
 705
 706         // does the wanted text start at point?
 707         for (string::size_type i = 0; i < str.length(); ++i) {
 708                 // Caution: direct comparison of characters works only
 709                 // because str is pure ASCII.
 710                 if (str[i] != text_[pos + i])
 711                         return false;
 712         }
 713
 714         return fontlist_.hasChangeInRange(pos, len);
 715 }
 716
 717
 718 void Paragraph::Private::latexInset(
 719                                              BufferParams const & bparams,
 720                                              odocstream & os,
 721                                              TexRow & texrow,
 722                                              OutputParams & runparams,
 723                                              Font & running_font,
 724                                              Font & basefont,
 725                                              Font const & outerfont,
 726                                              bool & open_font,
 727                                              Change & running_change,
 728                                              Layout const & style,
 729                                              pos_type & i,
 730                                              unsigned int & column)
 731 {
 732         Inset * inset = owner_->getInset(i);
 733         LASSERT(inset, /**/);
 734
 735         if (style.pass_thru) {
 736                 inset->plaintext(os, runparams);
 737                 return;
 738         }
 739
 740         // FIXME: move this to InsetNewline::latex
 741         if (inset->lyxCode() == NEWLINE_CODE) {
 742                 // newlines are handled differently here than
 743                 // the default in simpleTeXSpecialChars().
 744                 if (!style.newline_allowed) {
 745                         os << '\n';
 746                 } else {
 747                         if (open_font) {
 748                                 column += running_font.latexWriteEndChanges(
 749                                         os, bparams, runparams,
 750                                         basefont, basefont);
 751                                 open_font = false;
 752                         }
 753
 754                         if (running_font.fontInfo().family() == TYPEWRITER_FAMILY)
 755                                 os << '~';
 756
 757                         basefont = owner_->getLayoutFont(bparams, outerfont);
 758                         running_font = basefont;
 759
 760                         if (runparams.moving_arg)
 761                                 os << "\\protect ";
 762
 763                 }
 764                 texrow.newline();
 765                 texrow.start(owner_->id(), i + 1);
 766                 column = 0;
 767         }
 768
 769         if (owner_->isDeleted(i)) {
 770                 if( ++runparams.inDeletedInset == 1)
 771                         runparams.changeOfDeletedInset = owner_->lookupChange(i);
 772         }
 773
 774         if (inset->canTrackChanges()) {
 775                 column += Changes::latexMarkChange(os, bparams, running_change,
 776                         Change(Change::UNCHANGED), runparams);
 777                 running_change = Change(Change::UNCHANGED);
 778         }
 779
 780         bool close = false;
 781         odocstream::pos_type const len = os.tellp();
 782
 783         if (inset->forceLTR()
 784             && running_font.isRightToLeft()
 785             // ERT is an exception, it should be output with no
 786             // decorations at all
 787             && inset->lyxCode() != ERT_CODE) {
 788                 if (running_font.language()->lang() == "farsi")
 789                         os << "\\beginL{}";
 790                 else
 791                         os << "\\L{";
 792                 close = true;
 793         }
 794
 795         // FIXME: Bug: we can have an empty font change here!
 796         // if there has just been a font change, we are going to close it
 797         // right now, which means stupid latex code like \textsf{}. AFAIK,
 798         // this does not harm dvi output. A minor bug, thus (JMarc)
 799
 800         // Some insets cannot be inside a font change command.
 801         // However, even such insets *can* be placed in \L or \R
 802         // or their equivalents (for RTL language switches), so we don't
 803         // close the language in those cases.
 804         // ArabTeX, though, cannot handle this special behavior, it seems.
 805         bool arabtex = basefont.language()->lang() == "arabic_arabtex"
 806                 || running_font.language()->lang() == "arabic_arabtex";
 807         if (open_font && inset->noFontChange()) {
 808                 bool closeLanguage = arabtex
 809                         || basefont.isRightToLeft() == running_font.isRightToLeft();
 810                 unsigned int count = running_font.latexWriteEndChanges(os,
 811                         bparams, runparams, basefont, basefont, closeLanguage);
 812                 column += count;
 813                 // if any font properties were closed, update the running_font,
 814                 // making sure, however, to leave the language as it was
 815                 if (count > 0) {
 816                         // FIXME: probably a better way to keep track of the old
 817                         // language, than copying the entire font?
 818                         Font const copy_font(running_font);
 819                         basefont = owner_->getLayoutFont(bparams, outerfont);
 820                         running_font = basefont;
 821                         if (!closeLanguage)
 822                                 running_font.setLanguage(copy_font.language());
 823                         // leave font open if language is still open
 824                         open_font = (running_font.language() == basefont.language());
 825                         if (closeLanguage)
 826                                 runparams.local_font = &basefont;
 827                 }
 828         }
 829
 830         int tmp;
 831
 832         try {
 833                 tmp = inset->latex(os, runparams);
 834         } catch (EncodingException & e) {
 835                 // add location information and throw again.
 836                 e.par_id = id_;
 837                 e.pos = i;
 838                 throw(e);
 839         }
 840
 841         if (close) {
 842                 if (running_font.language()->lang() == "farsi")
 843                                 os << "\\endL{}";
 844                         else
 845                                 os << '}';
 846         }
 847
 848         if (tmp) {
 849                 for (int j = 0; j < tmp; ++j)
 850                         texrow.newline();
 851
 852                 texrow.start(owner_->id(), i + 1);
 853                 column = 0;
 854         } else {
 855                 column += os.tellp() - len;
 856         }
 857
 858         if (owner_->isDeleted(i))
 859                 --runparams.inDeletedInset;
 860 }
 861
 862
 863 void Paragraph::Private::latexSpecialChar(
 864                                              odocstream & os,
 865                                              OutputParams const & runparams,
 866                                              Font const & running_font,
 867                                              Change const & running_change,
 868                                              Layout const & style,
 869                                              pos_type & i,
 870                                              unsigned int & column)
 871 {
 872         char_type const c = text_[i];
 873
 874         if (style.pass_thru) {
 875                 if (c != '\0')
 876                         // FIXME UNICODE: This can fail if c cannot
 877                         // be encoded in the current encoding.
 878                         os.put(c);
 879                 return;
 880         }
 881
 882         if (runparams.verbatim) {
 883                 // FIXME UNICODE: This can fail if c cannot
 884                 // be encoded in the current encoding.
 885                 os.put(c);
 886                 return;
 887         }
 888
 889         // If T1 font encoding is used, use the special
 890         // characters it provides.
 891         // NOTE: some languages reset the font encoding
 892         // internally
 893         if (!running_font.language()->internalFontEncoding()
 894             && lyxrc.fontenc == "T1" && latexSpecialT1(c, os, i, column))
 895                 return;
 896
 897         // \tt font needs special treatment
 898         if (running_font.fontInfo().family() == TYPEWRITER_FAMILY
 899                 && latexSpecialTypewriter(c, os, i, column))
 900                 return;
 901
 902         // Otherwise, we use what LaTeX provides us.
 903         switch (c) {
 904         case '\\':
 905                 os << "\\textbackslash{}";
 906                 column += 15;
 907                 break;
 908         case '<':
 909                 os << "\\textless{}";
 910                 column += 10;
 911                 break;
 912         case '>':
 913                 os << "\\textgreater{}";
 914                 column += 13;
 915                 break;
 916         case '|':
 917                 os << "\\textbar{}";
 918                 column += 9;
 919                 break;
 920         case '-':
 921                 os << '-';
 922                 break;
 923         case '\"':
 924                 os << "\\char`\\\"{}";
 925                 column += 9;
 926                 break;
 927
 928         case '$': case '&':
 929         case '%': case '#': case '{':
 930         case '}': case '_':
 931                 os << '\\';
 932                 os.put(c);
 933                 column += 1;
 934                 break;
 935
 936         case '~':
 937                 os << "\\textasciitilde{}";
 938                 column += 16;
 939                 break;
 940
 941         case '^':
 942                 os << "\\textasciicircum{}";
 943                 column += 17;
 944                 break;
 945
 946         case '*':
 947         case '[':
 948         case ']':
 949                 // avoid being mistaken for optional arguments
 950                 os << '{';
 951                 os.put(c);
 952                 os << '}';
 953                 column += 2;
 954                 break;
 955
 956         case ' ':
 957                 // Blanks are printed before font switching.
 958                 // Sure? I am not! (try nice-latex)
 959                 // I am sure it's correct. LyX might be smarter
 960                 // in the future, but for now, nothing wrong is
 961                 // written. (Asger)
 962                 break;
 963
 964         default:
 965                 // LyX, LaTeX etc.
 966                 if (latexSpecialPhrase(os, i, column, runparams))
 967                         return;
 968
 969                 if (c == '\0')
 970                         return;
 971
 972                 Encoding const & encoding = *(runparams.encoding);
 973                 if (i + 1 < int(text_.size())) {
 974                         char_type next = text_[i + 1];
 975                         if (Encodings::isCombiningChar(next)) {
 976                                 column += latexSurrogatePair(os, c, next, runparams) - 1;
 977                                 ++i;
 978                                 break;
 979                         }
 980                 }
 981                 string script;
 982                 docstring const latex = encoding.latexChar(c);
 983                 if (Encodings::isKnownScriptChar(c, script)
 984                     && prefixIs(latex, from_ascii("\\" + script)))
 985                         column += writeScriptChars(os, latex,
 986                                         running_change, encoding, i) - 1;
 987                 else if (latex.length() > 1 && latex[latex.length() - 1] != '}') {
 988                         // Prevent eating of a following
 989                         // space or command corruption by
 990                         // following characters
 991                         column += latex.length() + 1;
 992                         os << latex << "{}";
 993                 } else {
 994                         column += latex.length() - 1;
 995                         os << latex;
 996                 }
 997                 break;
 998         }
 999 }
1000
1001
1002 bool Paragraph::Private::latexSpecialT1(char_type const c, odocstream & os,
1003         pos_type i, unsigned int & column)
1004 {
1005         switch (c) {
1006         case '>':
1007         case '<':
1008                 os.put(c);
1009                 // In T1 encoding, these characters exist
1010                 // but we should avoid ligatures
1011                 if (i + 1 >= int(text_.size()) || text_[i + 1] != c)
1012                         return true;
1013                 os << "\\textcompwordmark{}";
1014                 column += 19;
1015                 return true;
1016         case '|':
1017                 os.put(c);
1018                 return true;
1019         case '\"':
1020                 // soul.sty breaks with \char`\"
1021                 os << "\\textquotedbl{}";
1022                 column += 14;
1023                 return true;
1024         default:
1025                 return false;
1026         }
1027 }
1028
1029
1030 bool Paragraph::Private::latexSpecialTypewriter(char_type const c, odocstream & os,
1031         pos_type i, unsigned int & column)
1032 {
1033         switch (c) {
1034         case '-':
1035                 // within \ttfamily, "--" is merged to "-" (no endash)
1036                 // so we avoid this rather irritating ligature
1037                 if (i + 1 < int(text_.size()) && text_[i + 1] == '-') {
1038                         os << "-{}";
1039                         column += 2;
1040                 } else
1041                         os << '-';
1042                 return true;
1043
1044         // everything else has to be checked separately
1045         // (depending on the encoding)
1046         default:
1047                 return false;
1048         }
1049 }
1050
1051
1052 bool Paragraph::Private::latexSpecialPhrase(odocstream & os, pos_type & i,
1053         unsigned int & column, OutputParams const & runparams)
1054 {
1055         // FIXME: if we have "LaTeX" with a font
1056         // change in the middle (before the 'T', then
1057         // the "TeX" part is still special cased.
1058         // Really we should only operate this on
1059         // "words" for some definition of word
1060
1061         for (size_t pnr = 0; pnr < phrases_nr; ++pnr) {
1062                 if (!isTextAt(special_phrases[pnr].phrase, i))
1063                         continue;
1064                 if (runparams.moving_arg)
1065                         os << "\\protect";
1066                 os << special_phrases[pnr].macro;
1067                 i += special_phrases[pnr].phrase.length() - 1;
1068                 column += special_phrases[pnr].macro.length() - 1;
1069                 return true;
1070         }
1071         return false;
1072 }
1073
1074
1075 void Paragraph::Private::validate(LaTeXFeatures & features) const
1076 {
1077         if (layout_->inpreamble && inset_owner_) {
1078                 Buffer const & buf = inset_owner_->buffer();
1079                 BufferParams const & bp = buf.params();
1080                 Font f;
1081                 TexRow tr;
1082                 odocstringstream ods;
1083                 owner_->latex(bp, f, ods, tr, features.runparams());
1084                 docstring d = ods.str();
1085                 if (!d.empty())
1086                         features.addPreambleSnippet(to_utf8(d));
1087         }
1088
1089         if (features.runparams().flavor == OutputParams::HTML
1090             && layout_->htmltitle()) {
1091                 features.setHTMLTitle(owner_->asString(AS_STR_INSETS));
1092         }
1093
1094         // check the params.
1095         if (!params_.spacing().isDefault())
1096                 features.require("setspace");
1097
1098         // then the layouts
1099         features.useLayout(layout_->name());
1100
1101         // then the fonts
1102         fontlist_.validate(features);
1103
1104         // then the indentation
1105         if (!params_.leftIndent().zero())
1106                 features.require("ParagraphLeftIndent");
1107
1108         // then the insets
1109         InsetList::const_iterator icit = insetlist_.begin();
1110         InsetList::const_iterator iend = insetlist_.end();
1111         for (; icit != iend; ++icit) {
1112                 if (icit->inset) {
1113                         icit->inset->validate(features);
1114                         if (layout_->needprotect &&
1115                             icit->inset->lyxCode() == FOOT_CODE)
1116                                 features.require("NeedLyXFootnoteCode");
1117                 }
1118         }
1119
1120         // then the contents
1121         for (pos_type i = 0; i < int(text_.size()) ; ++i) {
1122                 for (size_t pnr = 0; pnr < phrases_nr; ++pnr) {
1123                         if (!special_phrases[pnr].builtin
1124                             && isTextAt(special_phrases[pnr].phrase, i)) {
1125                                 features.require(special_phrases[pnr].phrase);
1126                                 break;
1127                         }
1128                 }
1129                 Encodings::validate(text_[i], features);
1130         }
1131 }
1132
1133 /////////////////////////////////////////////////////////////////////
1134 //
1135 // Paragraph
1136 //
1137 /////////////////////////////////////////////////////////////////////
1138
1139 namespace {
1140         Layout const emptyParagraphLayout;
1141 }
1142
1143 Paragraph::Paragraph()
1144         : d(new Paragraph::Private(this, emptyParagraphLayout))
1145 {
1146         itemdepth = 0;
1147         d->params_.clear();
1148 }
1149
1150
1151 Paragraph::Paragraph(Paragraph const & par)
1152         : itemdepth(par.itemdepth),
1153         d(new Paragraph::Private(*par.d, this))
1154 {
1155         registerWords();
1156 }
1157
1158
1159 Paragraph::Paragraph(Paragraph const & par, pos_type beg, pos_type end)
1160         : itemdepth(par.itemdepth),
1161         d(new Paragraph::Private(*par.d, this, beg, end))
1162 {
1163         registerWords();
1164 }
1165
1166
1167 Paragraph & Paragraph::operator=(Paragraph const & par)
1168 {
1169         // needed as we will destroy the private part before copying it
1170         if (&par != this) {
1171                 itemdepth = par.itemdepth;
1172
1173                 deregisterWords();
1174                 delete d;
1175                 d = new Private(*par.d, this);
1176                 registerWords();
1177         }
1178         return *this;
1179 }
1180
1181
1182 Paragraph::~Paragraph()
1183 {
1184         deregisterWords();
1185         delete d;
1186 }
1187
1188
1189 namespace {
1190
1191 // this shall be called just before every "os << ..." action.
1192 void flushString(ostream & os, docstring & s)
1193 {
1194         os << to_utf8(s);
1195         s.erase();
1196 }
1197
1198 }
1199
1200
1201 void Paragraph::write(ostream & os, BufferParams const & bparams,
1202         depth_type & dth) const
1203 {
1204         // The beginning or end of a deeper (i.e. nested) area?
1205         if (dth != d->params_.depth()) {
1206                 if (d->params_.depth() > dth) {
1207                         while (d->params_.depth() > dth) {
1208                                 os << "\n\\begin_deeper";
1209                                 ++dth;
1210                         }
1211                 } else {
1212                         while (d->params_.depth() < dth) {
1213                                 os << "\n\\end_deeper";
1214                                 --dth;
1215                         }
1216                 }
1217         }
1218
1219         // First write the layout
1220         os << "\n\\begin_layout " << to_utf8(d->layout_->name()) << '\n';
1221
1222         d->params_.write(os);
1223
1224         Font font1(inherit_font, bparams.language);
1225
1226         Change running_change = Change(Change::UNCHANGED);
1227
1228         // this string is used as a buffer to avoid repetitive calls
1229         // to to_utf8(), which turn out to be expensive (JMarc)
1230         docstring write_buffer;
1231
1232         int column = 0;
1233         for (pos_type i = 0; i <= size(); ++i) {
1234
1235                 Change const change = lookupChange(i);
1236                 if (change != running_change)
1237                         flushString(os, write_buffer);
1238                 Changes::lyxMarkChange(os, bparams, column, running_change, change);
1239                 running_change = change;
1240
1241                 if (i == size())
1242                         break;
1243
1244                 // Write font changes (ignore spelling markers)
1245                 Font font2 = getFontSettings(bparams, i);
1246                 font2.setMisspelled(false);
1247                 if (font2 != font1) {
1248                         flushString(os, write_buffer);
1249                         font2.lyxWriteChanges(font1, os);
1250                         column = 0;
1251                         font1 = font2;
1252                 }
1253
1254                 char_type const c = d->text_[i];
1255                 switch (c) {
1256                 case META_INSET:
1257                         if (Inset const * inset = getInset(i)) {
1258                                 flushString(os, write_buffer);
1259                                 if (inset->directWrite()) {
1260                                         // international char, let it write
1261                                         // code directly so it's shorter in
1262                                         // the file
1263                                         inset->write(os);
1264                                 } else {
1265                                         if (i)
1266                                                 os << '\n';
1267                                         os << "\\begin_inset ";
1268                                         inset->write(os);
1269                                         os << "\n\\end_inset\n\n";
1270                                         column = 0;
1271                                 }
1272                         }
1273                         break;
1274                 case '\\':
1275                         flushString(os, write_buffer);
1276                         os << "\n\\backslash\n";
1277                         column = 0;
1278                         break;
1279                 case '.':
1280                         flushString(os, write_buffer);
1281                         if (i + 1 < size() && d->text_[i + 1] == ' ') {
1282                                 os << ".\n";
1283                                 column = 0;
1284                         } else
1285                                 os << '.';
1286                         break;
1287                 default:
1288                         if ((column > 70 && c == ' ')
1289                             || column > 79) {
1290                                 flushString(os, write_buffer);
1291                                 os << '\n';
1292                                 column = 0;
1293                         }
1294                         // this check is to amend a bug. LyX sometimes
1295                         // inserts '\0' this could cause problems.
1296                         if (c != '\0')
1297                                 write_buffer.push_back(c);
1298                         else
1299                                 LYXERR0("NUL char in structure.");
1300                         ++column;
1301                         break;
1302                 }
1303         }
1304
1305         flushString(os, write_buffer);
1306         os << "\n\\end_layout\n";
1307 }
1308
1309
1310 void Paragraph::validate(LaTeXFeatures & features) const
1311 {
1312         d->validate(features);
1313 }
1314
1315
1316 void Paragraph::insert(pos_type start, docstring const & str,
1317                        Font const & font, Change const & change)
1318 {
1319         for (size_t i = 0, n = str.size(); i != n ; ++i)
1320                 insertChar(start + i, str[i], font, change);
1321 }
1322
1323
1324 void Paragraph::appendChar(char_type c, Font const & font,
1325                 Change const & change)
1326 {
1327         // track change
1328         d->changes_.insert(change, d->text_.size());
1329         // when appending characters, no need to update tables
1330         d->text_.push_back(c);
1331         setFont(d->text_.size() - 1, font);
1332 }
1333
1334
1335 void Paragraph::appendString(docstring const & s, Font const & font,
1336                 Change const & change)
1337 {
1338         pos_type end = s.size();
1339         size_t oldsize = d->text_.size();
1340         size_t newsize = oldsize + end;
1341         size_t capacity = d->text_.capacity();
1342         if (newsize >= capacity)
1343                 d->text_.reserve(max(capacity + 100, newsize));
1344
1345         // when appending characters, no need to update tables
1346         d->text_.append(s);
1347
1348         // FIXME: Optimize this!
1349         for (size_t i = oldsize; i != newsize; ++i) {
1350                 // track change
1351                 d->changes_.insert(change, i);
1352         }
1353         d->fontlist_.set(oldsize, font);
1354         d->fontlist_.set(newsize - 1, font);
1355 }
1356
1357
1358 void Paragraph::insertChar(pos_type pos, char_type c,
1359                            bool trackChanges)
1360 {
1361         d->insertChar(pos, c, Change(trackChanges ?
1362                            Change::INSERTED : Change::UNCHANGED));
1363 }
1364
1365
1366 void Paragraph::insertChar(pos_type pos, char_type c,
1367                            Font const & font, bool trackChanges)
1368 {
1369         d->insertChar(pos, c, Change(trackChanges ?
1370                            Change::INSERTED : Change::UNCHANGED));
1371         setFont(pos, font);
1372 }
1373
1374
1375 void Paragraph::insertChar(pos_type pos, char_type c,
1376                            Font const & font, Change const & change)
1377 {
1378         d->insertChar(pos, c, change);
1379         setFont(pos, font);
1380 }
1381
1382
1383 bool Paragraph::insertInset(pos_type pos, Inset * inset,
1384                             Font const & font, Change const & change)
1385 {
1386         bool const success = insertInset(pos, inset, change);
1387         // Set the font/language of the inset...
1388         setFont(pos, font);
1389         return success;
1390 }
1391
1392
1393 void Paragraph::resetFonts(Font const & font)
1394 {
1395         d->fontlist_.clear();
1396         d->fontlist_.set(0, font);
1397         d->fontlist_.set(d->text_.size() - 1, font);
1398 }
1399
1400 // Gets uninstantiated font setting at position.
1401 Font const & Paragraph::getFontSettings(BufferParams const & bparams,
1402                                          pos_type pos) const
1403 {
1404         if (pos > size()) {
1405                 LYXERR0("pos: " << pos << " size: " << size());
1406                 LASSERT(pos <= size(), /**/);
1407         }
1408
1409         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
1410         if (cit != d->fontlist_.end())
1411                 return cit->font();
1412
1413         if (pos == size() && !empty())
1414                 return getFontSettings(bparams, pos - 1);
1415
1416         // Optimisation: avoid a full font instantiation if there is no
1417         // language change from previous call.
1418         static Font previous_font;
1419         static Language const * previous_lang = 0;
1420         Language const * lang = getParLanguage(bparams);
1421         if (lang != previous_lang) {
1422                 previous_lang = lang;
1423                 previous_font = Font(inherit_font, lang);
1424         }
1425         return previous_font;
1426 }
1427
1428
1429 FontSpan Paragraph::fontSpan(pos_type pos) const
1430 {
1431         LASSERT(pos <= size(), /**/);
1432         pos_type start = 0;
1433
1434         FontList::const_iterator cit = d->fontlist_.begin();
1435         FontList::const_iterator end = d->fontlist_.end();
1436         for (; cit != end; ++cit) {
1437                 if (cit->pos() >= pos) {
1438                         if (pos >= beginOfBody())
1439                                 return FontSpan(max(start, beginOfBody()),
1440                                                 cit->pos());
1441                         else
1442                                 return FontSpan(start,
1443                                                 min(beginOfBody() - 1,
1444                                                          cit->pos()));
1445                 }
1446                 start = cit->pos() + 1;
1447         }
1448
1449         // This should not happen, but if so, we take no chances.
1450         // LYXERR0("Paragraph::getEndPosOfFontSpan: This should not happen!");
1451         return FontSpan(pos, pos);
1452 }
1453
1454
1455 // Gets uninstantiated font setting at position 0
1456 Font const & Paragraph::getFirstFontSettings(BufferParams const & bparams) const
1457 {
1458         if (!empty() && !d->fontlist_.empty())
1459                 return d->fontlist_.begin()->font();
1460
1461         // Optimisation: avoid a full font instantiation if there is no
1462         // language change from previous call.
1463         static Font previous_font;
1464         static Language const * previous_lang = 0;
1465         if (bparams.language != previous_lang) {
1466                 previous_lang = bparams.language;
1467                 previous_font = Font(inherit_font, bparams.language);
1468         }
1469
1470         return previous_font;
1471 }
1472
1473
1474 // Gets the fully instantiated font at a given position in a paragraph
1475 // This is basically the same function as Text::GetFont() in text2.cpp.
1476 // The difference is that this one is used for generating the LaTeX file,
1477 // and thus cosmetic "improvements" are disallowed: This has to deliver
1478 // the true picture of the buffer. (Asger)
1479 Font const Paragraph::getFont(BufferParams const & bparams, pos_type pos,
1480                                  Font const & outerfont) const
1481 {
1482         LASSERT(pos >= 0, /**/);
1483
1484         Font font = getFontSettings(bparams, pos);
1485
1486         pos_type const body_pos = beginOfBody();
1487         FontInfo & fi = font.fontInfo();
1488         if (pos < body_pos)
1489                 fi.realize(d->layout_->labelfont);
1490         else
1491                 fi.realize(d->layout_->font);
1492
1493         fi.realize(outerfont.fontInfo());
1494         fi.realize(bparams.getFont().fontInfo());
1495
1496         return font;
1497 }
1498
1499
1500 Font const Paragraph::getLabelFont
1501         (BufferParams const & bparams, Font const & outerfont) const
1502 {
1503         FontInfo tmpfont = d->layout_->labelfont;
1504         tmpfont.realize(outerfont.fontInfo());
1505         tmpfont.realize(bparams.getFont().fontInfo());
1506         return Font(tmpfont, getParLanguage(bparams));
1507 }
1508
1509
1510 Font const Paragraph::getLayoutFont
1511         (BufferParams const & bparams, Font const & outerfont) const
1512 {
1513         FontInfo tmpfont = d->layout_->font;
1514         tmpfont.realize(outerfont.fontInfo());
1515         tmpfont.realize(bparams.getFont().fontInfo());
1516         return Font(tmpfont, getParLanguage(bparams));
1517 }
1518
1519
1520 /// Returns the height of the highest font in range
1521 FontSize Paragraph::highestFontInRange
1522         (pos_type startpos, pos_type endpos, FontSize def_size) const
1523 {
1524         return d->fontlist_.highestInRange(startpos, endpos, def_size);
1525 }
1526
1527
1528 char_type Paragraph::getUChar(BufferParams const & bparams, pos_type pos) const
1529 {
1530         char_type c = d->text_[pos];
1531         if (!lyxrc.rtl_support)
1532                 return c;
1533
1534         char_type uc = c;
1535         switch (c) {
1536         case '(':
1537                 uc = ')';
1538                 break;
1539         case ')':
1540                 uc = '(';
1541                 break;
1542         case '[':
1543                 uc = ']';
1544                 break;
1545         case ']':
1546                 uc = '[';
1547                 break;
1548         case '{':
1549                 uc = '}';
1550                 break;
1551         case '}':
1552                 uc = '{';
1553                 break;
1554         case '<':
1555                 uc = '>';
1556                 break;
1557         case '>':
1558                 uc = '<';
1559                 break;
1560         }
1561         if (uc != c && getFontSettings(bparams, pos).isRightToLeft())
1562                 return uc;
1563         return c;
1564 }
1565
1566
1567 void Paragraph::setFont(pos_type pos, Font const & font)
1568 {
1569         LASSERT(pos <= size(), /**/);
1570
1571         // First, reduce font against layout/label font
1572         // Update: The setCharFont() routine in text2.cpp already
1573         // reduces font, so we don't need to do that here. (Asger)
1574
1575         d->fontlist_.set(pos, font);
1576 }
1577
1578
1579 void Paragraph::makeSameLayout(Paragraph const & par)
1580 {
1581         d->layout_ = par.d->layout_;
1582         d->params_ = par.d->params_;
1583 }
1584
1585
1586 bool Paragraph::stripLeadingSpaces(bool trackChanges)
1587 {
1588         if (isFreeSpacing())
1589                 return false;
1590
1591         int pos = 0;
1592         int count = 0;
1593
1594         while (pos < size() && (isNewline(pos) || isLineSeparator(pos))) {
1595                 if (eraseChar(pos, trackChanges))
1596                         ++count;
1597                 else
1598                         ++pos;
1599         }
1600
1601         return count > 0 || pos > 0;
1602 }
1603
1604
1605 bool Paragraph::hasSameLayout(Paragraph const & par) const
1606 {
1607         return par.d->layout_ == d->layout_
1608                 && d->params_.sameLayout(par.d->params_);
1609 }
1610
1611
1612 depth_type Paragraph::getDepth() const
1613 {
1614         return d->params_.depth();
1615 }
1616
1617
1618 depth_type Paragraph::getMaxDepthAfter() const
1619 {
1620         if (d->layout_->isEnvironment())
1621                 return d->params_.depth() + 1;
1622         else
1623                 return d->params_.depth();
1624 }
1625
1626
1627 char Paragraph::getAlign() const
1628 {
1629         if (d->params_.align() == LYX_ALIGN_LAYOUT)
1630                 return d->layout_->align;
1631         else
1632                 return d->params_.align();
1633 }
1634
1635
1636 docstring const & Paragraph::labelString() const
1637 {
1638         return d->params_.labelString();
1639 }
1640
1641
1642 // the next two functions are for the manual labels
1643 docstring const Paragraph::getLabelWidthString() const
1644 {
1645         if (d->layout_->margintype == MARGIN_MANUAL
1646             || d->layout_->latextype == LATEX_BIB_ENVIRONMENT)
1647                 return d->params_.labelWidthString();
1648         else
1649                 return _("Senseless with this layout!");
1650 }
1651
1652
1653 void Paragraph::setLabelWidthString(docstring const & s)
1654 {
1655         d->params_.labelWidthString(s);
1656 }
1657
1658
1659 docstring Paragraph::expandLabel(Layout const & layout,
1660                 BufferParams const & bparams) const
1661 {
1662         return expandParagraphLabel(layout, bparams, true);
1663 }
1664
1665
1666 docstring Paragraph::expandDocBookLabel(Layout const & layout,
1667                 BufferParams const & bparams) const
1668 {
1669         return expandParagraphLabel(layout, bparams, false);
1670 }
1671
1672
1673 docstring Paragraph::expandParagraphLabel(Layout const & layout,
1674                 BufferParams const & bparams, bool process_appendix) const
1675 {
1676         DocumentClass const & tclass = bparams.documentClass();
1677         string const & lang = getParLanguage(bparams)->code();
1678         bool const in_appendix = process_appendix && d->params_.appendix();
1679         docstring fmt = translateIfPossible(layout.labelstring(in_appendix), lang);
1680
1681         if (fmt.empty() && layout.labeltype == LABEL_COUNTER
1682             && !layout.counter.empty())
1683                 return tclass.counters().theCounter(layout.counter, lang);
1684
1685         // handle 'inherited level parts' in 'fmt',
1686         // i.e. the stuff between '@' in   '@Section@.\arabic{subsection}'
1687         size_t const i = fmt.find('@', 0);
1688         if (i != docstring::npos) {
1689                 size_t const j = fmt.find('@', i + 1);
1690                 if (j != docstring::npos) {
1691                         docstring parent(fmt, i + 1, j - i - 1);
1692                         docstring label = from_ascii("??");
1693                         if (tclass.hasLayout(parent))
1694                                 docstring label = expandParagraphLabel(tclass[parent], bparams,
1695                                                       process_appendix);
1696                         fmt = docstring(fmt, 0, i) + label
1697                                 + docstring(fmt, j + 1, docstring::npos);
1698                 }
1699         }
1700
1701         return tclass.counters().counterLabel(fmt, lang);
1702 }
1703
1704
1705 void Paragraph::applyLayout(Layout const & new_layout)
1706 {
1707         d->layout_ = &new_layout;
1708         LyXAlignment const oldAlign = d->params_.align();
1709
1710         if (!(oldAlign & d->layout_->alignpossible)) {
1711                 frontend::Alert::warning(_("Alignment not permitted"),
1712                         _("The new layout does not permit the alignment previously used.\nSetting to default."));
1713                 d->params_.align(LYX_ALIGN_LAYOUT);
1714         }
1715 }
1716
1717
1718 pos_type Paragraph::beginOfBody() const
1719 {
1720         return d->begin_of_body_;
1721 }
1722
1723
1724 void Paragraph::setBeginOfBody()
1725 {
1726         if (d->layout_->labeltype != LABEL_MANUAL) {
1727                 d->begin_of_body_ = 0;
1728                 return;
1729         }
1730
1731         // Unroll the first two cycles of the loop
1732         // and remember the previous character to
1733         // remove unnecessary getChar() calls
1734         pos_type i = 0;
1735         pos_type end = size();
1736         if (i < end && !isNewline(i)) {
1737                 ++i;
1738                 char_type previous_char = 0;
1739                 char_type temp = 0;
1740                 if (i < end) {
1741                         previous_char = d->text_[i];
1742                         if (!isNewline(i)) {
1743                                 ++i;
1744                                 while (i < end && previous_char != ' ') {
1745                                         temp = d->text_[i];
1746                                         if (isNewline(i))
1747                                                 break;
1748                                         ++i;
1749                                         previous_char = temp;
1750                                 }
1751                         }
1752                 }
1753         }
1754
1755         d->begin_of_body_ = i;
1756 }
1757
1758
1759 bool Paragraph::allowParagraphCustomization() const
1760 {
1761         return inInset().allowParagraphCustomization();
1762 }
1763
1764
1765 bool Paragraph::usePlainLayout() const
1766 {
1767         return inInset().usePlainLayout();
1768 }
1769
1770
1771 namespace {
1772
1773 // paragraphs inside floats need different alignment tags to avoid
1774 // unwanted space
1775
1776 bool noTrivlistCentering(InsetCode code)
1777 {
1778         return code == FLOAT_CODE
1779                || code == WRAP_CODE
1780                || code == CELL_CODE;
1781 }
1782
1783
1784 string correction(string const & orig)
1785 {
1786         if (orig == "flushleft")
1787                 return "raggedright";
1788         if (orig == "flushright")
1789                 return "raggedleft";
1790         if (orig == "center")
1791                 return "centering";
1792         return orig;
1793 }
1794
1795
1796 string const corrected_env(string const & suffix, string const & env,
1797         InsetCode code, bool const lastpar)
1798 {
1799         string output = suffix + "{";
1800         if (noTrivlistCentering(code)) {
1801                 if (lastpar) {
1802                         // the last paragraph in non-trivlist-aligned
1803                         // context is special (to avoid unwanted whitespace)
1804                         if (suffix == "\\begin")
1805                                 return "\\" + correction(env) + "{}";
1806                         return string();
1807                 }
1808                 output += correction(env);
1809         } else
1810                 output += env;
1811         output += "}";
1812         if (suffix == "\\begin")
1813                 output += "\n";
1814         return output;
1815 }
1816
1817
1818 void adjust_row_column(string const & str, TexRow & texrow, int & column)
1819 {
1820         if (!contains(str, "\n"))
1821                 column += str.size();
1822         else {
1823                 string tmp;
1824                 texrow.newline();
1825                 column = rsplit(str, tmp, '\n').size();
1826         }
1827 }
1828
1829 } // namespace anon
1830
1831
1832 int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
1833                                  odocstream & os, TexRow & texrow,
1834                                  OutputParams const & runparams) const
1835 {
1836         int column = 0;
1837
1838         if (params_.noindent()) {
1839                 os << "\\noindent ";
1840                 column += 10;
1841         }
1842
1843         LyXAlignment const curAlign = params_.align();
1844
1845         if (curAlign == layout_->align)
1846                 return column;
1847
1848         switch (curAlign) {
1849         case LYX_ALIGN_NONE:
1850         case LYX_ALIGN_BLOCK:
1851         case LYX_ALIGN_LAYOUT:
1852         case LYX_ALIGN_SPECIAL:
1853                 break;
1854         case LYX_ALIGN_LEFT:
1855         case LYX_ALIGN_RIGHT:
1856         case LYX_ALIGN_CENTER:
1857                 if (runparams.moving_arg) {
1858                         os << "\\protect";
1859                         column += 8;
1860                 }
1861                 break;
1862         }
1863
1864         string const begin_tag = "\\begin";
1865         InsetCode code = ownerCode();
1866         bool const lastpar = runparams.isLastPar;
1867
1868         switch (curAlign) {
1869         case LYX_ALIGN_NONE:
1870         case LYX_ALIGN_BLOCK:
1871         case LYX_ALIGN_LAYOUT:
1872         case LYX_ALIGN_SPECIAL:
1873                 break;
1874         case LYX_ALIGN_LEFT: {
1875                 string output;
1876                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
1877                         output = corrected_env(begin_tag, "flushleft", code, lastpar);
1878                 else
1879                         output = corrected_env(begin_tag, "flushright", code, lastpar);
1880                 os << from_ascii(output);
1881                 adjust_row_column(output, texrow, column);
1882                 break;
1883         } case LYX_ALIGN_RIGHT: {
1884                 string output;
1885                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
1886                         output = corrected_env(begin_tag, "flushright", code, lastpar);
1887                 else
1888                         output = corrected_env(begin_tag, "flushleft", code, lastpar);
1889                 os << from_ascii(output);
1890                 adjust_row_column(output, texrow, column);
1891                 break;
1892         } case LYX_ALIGN_CENTER: {
1893                 string output;
1894                 output = corrected_env(begin_tag, "center", code, lastpar);
1895                 os << from_ascii(output);
1896                 adjust_row_column(output, texrow, column);
1897                 break;
1898         }
1899         }
1900
1901         return column;
1902 }
1903
1904
1905 int Paragraph::Private::endTeXParParams(BufferParams const & bparams,
1906                                odocstream & os, TexRow & texrow,
1907                                OutputParams const & runparams) const
1908 {
1909         int column = 0;
1910
1911         LyXAlignment const curAlign = params_.align();
1912
1913         if (curAlign == layout_->align)
1914                 return column;
1915
1916         switch (curAlign) {
1917         case LYX_ALIGN_NONE:
1918         case LYX_ALIGN_BLOCK:
1919         case LYX_ALIGN_LAYOUT:
1920         case LYX_ALIGN_SPECIAL:
1921                 break;
1922         case LYX_ALIGN_LEFT:
1923         case LYX_ALIGN_RIGHT:
1924         case LYX_ALIGN_CENTER:
1925                 if (runparams.moving_arg) {
1926                         os << "\\protect";
1927                         column = 8;
1928                 }
1929                 break;
1930         }
1931
1932         string const end_tag = "\n\\par\\end";
1933         InsetCode code = ownerCode();
1934         bool const lastpar = runparams.isLastPar;
1935
1936         switch (curAlign) {
1937         case LYX_ALIGN_NONE:
1938         case LYX_ALIGN_BLOCK:
1939         case LYX_ALIGN_LAYOUT:
1940         case LYX_ALIGN_SPECIAL:
1941                 break;
1942         case LYX_ALIGN_LEFT: {
1943                 string output;
1944                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
1945                         output = corrected_env(end_tag, "flushleft", code, lastpar);
1946                 else
1947                         output = corrected_env(end_tag, "flushright", code, lastpar);
1948                 os << from_ascii(output);
1949                 adjust_row_column(output, texrow, column);
1950                 break;
1951         } case LYX_ALIGN_RIGHT: {
1952                 string output;
1953                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
1954                         output = corrected_env(end_tag, "flushright", code, lastpar);
1955                 else
1956                         output = corrected_env(end_tag, "flushleft", code, lastpar);
1957                 os << from_ascii(output);
1958                 adjust_row_column(output, texrow, column);
1959                 break;
1960         } case LYX_ALIGN_CENTER: {
1961                 string output;
1962                 output = corrected_env(end_tag, "center", code, lastpar);
1963                 os << from_ascii(output);
1964                 adjust_row_column(output, texrow, column);
1965                 break;
1966         }
1967         }
1968
1969         return column;
1970 }
1971
1972
1973 // This one spits out the text of the paragraph
1974 bool Paragraph::latex(BufferParams const & bparams,
1975         Font const & outerfont,
1976         odocstream & os, TexRow & texrow,
1977         OutputParams const & runparams,
1978         int start_pos, int end_pos) const
1979 {
1980         LYXERR(Debug::LATEX, "Paragraph::latex...     " << this);
1981
1982         if (layout().inpreamble)
1983                 return true;
1984
1985         bool return_value = false;
1986
1987         bool const allowcust = allowParagraphCustomization();
1988
1989         // FIXME This check should not be needed. Perhaps issue an
1990         // error if it triggers.
1991         Layout const & style = inInset().forcePlainLayout() ?
1992                 bparams.documentClass().plainLayout() : *d->layout_;
1993
1994         // Current base font for all inherited font changes, without any
1995         // change caused by an individual character, except for the language:
1996         // It is set to the language of the first character.
1997         // As long as we are in the label, this font is the base font of the
1998         // label. Before the first body character it is set to the base font
1999         // of the body.
2000         Font basefont;
2001
2002         // Maybe we have to create a optional argument.
2003         pos_type body_pos = beginOfBody();
2004         unsigned int column = 0;
2005
2006         if (body_pos > 0) {
2007                 os << '[';
2008                 column += 1;
2009                 basefont = getLabelFont(bparams, outerfont);
2010         } else {
2011                 basefont = getLayoutFont(bparams, outerfont);
2012         }
2013
2014         // Which font is currently active?
2015         Font running_font(basefont);
2016         // Do we have an open font change?
2017         bool open_font = false;
2018
2019         Change runningChange = Change(Change::UNCHANGED);
2020
2021         Encoding const * const prev_encoding = runparams.encoding;
2022
2023         texrow.start(id(), 0);
2024
2025         // if the paragraph is empty, the loop will not be entered at all
2026         if (empty()) {
2027                 if (style.isCommand()) {
2028                         os << '{';
2029                         ++column;
2030                 }
2031                 if (allowcust)
2032                         column += d->startTeXParParams(bparams, os, texrow,
2033                                                     runparams);
2034         }
2035
2036         for (pos_type i = 0; i < size(); ++i) {
2037                 // First char in paragraph or after label?
2038                 if (i == body_pos) {
2039                         if (body_pos > 0) {
2040                                 if (open_font) {
2041                                         column += running_font.latexWriteEndChanges(
2042                                                 os, bparams, runparams,
2043                                                 basefont, basefont);
2044                                         open_font = false;
2045                                 }
2046                                 basefont = getLayoutFont(bparams, outerfont);
2047                                 running_font = basefont;
2048
2049                                 column += Changes::latexMarkChange(os, bparams,
2050                                                 runningChange, Change(Change::UNCHANGED),
2051                                                 runparams);
2052                                 runningChange = Change(Change::UNCHANGED);
2053
2054                                 os << "] ";
2055                                 column +=2;
2056                         }
2057                         if (style.isCommand()) {
2058                                 os << '{';
2059                                 ++column;
2060                         }
2061
2062                         if (allowcust)
2063                                 column += d->startTeXParParams(bparams, os,
2064                                                             texrow,
2065                                                             runparams);
2066                 }
2067
2068                 Change const & change = runparams.inDeletedInset ? runparams.changeOfDeletedInset
2069                                                                  : lookupChange(i);
2070
2071                 if (bparams.outputChanges && runningChange != change) {
2072                         if (open_font) {
2073                                 column += running_font.latexWriteEndChanges(
2074                                                 os, bparams, runparams, basefont, basefont);
2075                                 open_font = false;
2076                         }
2077                         basefont = getLayoutFont(bparams, outerfont);
2078                         running_font = basefont;
2079
2080                         column += Changes::latexMarkChange(os, bparams, runningChange,
2081                                                            change, runparams);
2082                         runningChange = change;
2083                 }
2084
2085                 // do not output text which is marked deleted
2086                 // if change tracking output is disabled
2087                 if (!bparams.outputChanges && change.deleted()) {
2088                         continue;
2089                 }
2090
2091                 ++column;
2092
2093                 // Fully instantiated font
2094                 Font const font = getFont(bparams, i, outerfont);
2095
2096                 Font const last_font = running_font;
2097
2098                 // Do we need to close the previous font?
2099                 if (open_font &&
2100                     (font != running_font ||
2101                      font.language() != running_font.language()))
2102                 {
2103                         column += running_font.latexWriteEndChanges(
2104                                         os, bparams, runparams, basefont,
2105                                         (i == body_pos-1) ? basefont : font);
2106                         running_font = basefont;
2107                         open_font = false;
2108                 }
2109
2110                 // close babel's font environment before opening CJK.
2111                 if (!running_font.language()->babel().empty() &&
2112                     font.language()->encoding()->package() == Encoding::CJK) {
2113                                 string end_tag = subst(lyxrc.language_command_end,
2114                                                         "$$lang",
2115                                                         running_font.language()->babel());
2116                                 os << from_ascii(end_tag);
2117                                 column += end_tag.length();
2118                 }
2119
2120                 // Switch file encoding if necessary (and allowed)
2121                 if (!runparams.verbatim &&
2122                     runparams.encoding->package() != Encoding::none &&
2123                     font.language()->encoding()->package() != Encoding::none) {
2124                         pair<bool, int> const enc_switch = switchEncoding(os, bparams,
2125                                         runparams, *(font.language()->encoding()));
2126                         if (enc_switch.first) {
2127                                 column += enc_switch.second;
2128                                 runparams.encoding = font.language()->encoding();
2129                         }
2130                 }
2131
2132                 char_type const c = d->text_[i];
2133
2134                 // Do we need to change font?
2135                 if ((font != running_font ||
2136                      font.language() != running_font.language()) &&
2137                         i != body_pos - 1)
2138                 {
2139                         odocstringstream ods;
2140                         column += font.latexWriteStartChanges(ods, bparams,
2141                                                               runparams, basefont,
2142                                                               last_font);
2143                         running_font = font;
2144                         open_font = true;
2145                         docstring fontchange = ods.str();
2146                         // check whether the fontchange ends with a \\textcolor
2147                         // modifier and the text starts with a space (bug 4473)
2148                         docstring const last_modifier = rsplit(fontchange, '\\');
2149                         if (prefixIs(last_modifier, from_ascii("textcolor")) && c == ' ')
2150                                 os << fontchange << from_ascii("{}");
2151                         // check if the fontchange ends with a trailing blank
2152                         // (like "\small " (see bug 3382)
2153                         else if (suffixIs(fontchange, ' ') && c == ' ')
2154                                 os << fontchange.substr(0, fontchange.size() - 1)
2155                                    << from_ascii("{}");
2156                         else
2157                                 os << fontchange;
2158                 }
2159
2160                 // FIXME: think about end_pos implementation...
2161                 if (c == ' ' && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2162                         // FIXME: integrate this case in latexSpecialChar
2163                         // Do not print the separation of the optional argument
2164                         // if style.pass_thru is false. This works because
2165                         // latexSpecialChar ignores spaces if
2166                         // style.pass_thru is false.
2167                         if (i != body_pos - 1) {
2168                                 if (d->simpleTeXBlanks(
2169                                                 runparams, os, texrow,
2170                                                 i, column, font, style)) {
2171                                         // A surrogate pair was output. We
2172                                         // must not call latexSpecialChar
2173                                         // in this iteration, since it would output
2174                                         // the combining character again.
2175                                         ++i;
2176                                         continue;
2177                                 }
2178                         }
2179                 }
2180
2181                 OutputParams rp = runparams;
2182                 rp.free_spacing = style.free_spacing;
2183                 rp.local_font = &font;
2184                 rp.intitle = style.intitle;
2185
2186                 // Two major modes:  LaTeX or plain
2187                 // Handle here those cases common to both modes
2188                 // and then split to handle the two modes separately.
2189                 if (c == META_INSET) {
2190                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2191                                 d->latexInset(bparams, os,
2192                                                 texrow, rp, running_font,
2193                                                 basefont, outerfont, open_font,
2194                                                 runningChange, style, i, column);
2195                         }
2196                 } else {
2197                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2198                                 try {
2199                                         d->latexSpecialChar(os, rp, running_font, runningChange,
2200                                                 style, i, column);
2201                                 } catch (EncodingException & e) {
2202                                 if (runparams.dryrun) {
2203                                         os << "<" << _("LyX Warning: ")
2204                                            << _("uncodable character") << " '";
2205                                         os.put(c);
2206                                         os << "'>";
2207                                 } else {
2208                                         // add location information and throw again.
2209                                         e.par_id = id();
2210                                         e.pos = i;
2211                                         throw(e);
2212                                 }
2213                         }
2214                 }
2215                 }
2216
2217                 // Set the encoding to that returned from latexSpecialChar (see
2218                 // comment for encoding member in OutputParams.h)
2219                 runparams.encoding = rp.encoding;
2220         }
2221
2222         // If we have an open font definition, we have to close it
2223         if (open_font) {
2224 #ifdef FIXED_LANGUAGE_END_DETECTION
2225                 if (next_) {
2226                         running_font
2227                                 .latexWriteEndChanges(os, bparams, runparams,
2228                                         basefont,
2229                                         next_->getFont(bparams, 0, outerfont));
2230                 } else {
2231                         running_font.latexWriteEndChanges(os, bparams,
2232                                         runparams, basefont, basefont);
2233                 }
2234 #else
2235 //FIXME: For now we ALWAYS have to close the foreign font settings if they are
2236 //FIXME: there as we start another \selectlanguage with the next paragraph if
2237 //FIXME: we are in need of this. This should be fixed sometime (Jug)
2238                 running_font.latexWriteEndChanges(os, bparams, runparams,
2239                                 basefont, basefont);
2240 #endif
2241         }
2242
2243         column += Changes::latexMarkChange(os, bparams, runningChange,
2244                                            Change(Change::UNCHANGED), runparams);
2245
2246         // Needed if there is an optional argument but no contents.
2247         if (body_pos > 0 && body_pos == size()) {
2248                 os << "]~";
2249                 return_value = false;
2250         }
2251
2252         if (allowcust && d->endTeXParParams(bparams, os, texrow, runparams)
2253             && runparams.encoding != prev_encoding) {
2254                 runparams.encoding = prev_encoding;
2255                 if (!bparams.useXetex)
2256                         os << setEncoding(prev_encoding->iconvName());
2257         }
2258
2259         LYXERR(Debug::LATEX, "Paragraph::latex... done " << this);
2260         return return_value;
2261 }
2262
2263
2264 bool Paragraph::emptyTag() const
2265 {
2266         for (pos_type i = 0; i < size(); ++i) {
2267                 if (Inset const * inset = getInset(i)) {
2268                         InsetCode lyx_code = inset->lyxCode();
2269                         // FIXME testing like that is wrong. What is
2270                         // the intent?
2271                         if (lyx_code != TOC_CODE &&
2272                             lyx_code != INCLUDE_CODE &&
2273                             lyx_code != GRAPHICS_CODE &&
2274                             lyx_code != ERT_CODE &&
2275                             lyx_code != LISTINGS_CODE &&
2276                             lyx_code != FLOAT_CODE &&
2277                             lyx_code != TABULAR_CODE) {
2278                                 return false;
2279                         }
2280                 } else {
2281                         char_type c = d->text_[i];
2282                         if (c != ' ' && c != '\t')
2283                                 return false;
2284                 }
2285         }
2286         return true;
2287 }
2288
2289
2290 string Paragraph::getID(Buffer const & buf, OutputParams const & runparams)
2291         const
2292 {
2293         for (pos_type i = 0; i < size(); ++i) {
2294                 if (Inset const * inset = getInset(i)) {
2295                         InsetCode lyx_code = inset->lyxCode();
2296                         if (lyx_code == LABEL_CODE) {
2297                                 InsetLabel const * const il = static_cast<InsetLabel const *>(inset);
2298                                 docstring const & id = il->getParam("name");
2299                                 return "id='" + to_utf8(sgml::cleanID(buf, runparams, id)) + "'";
2300                         }
2301                 }
2302         }
2303         return string();
2304 }
2305
2306
2307 pos_type Paragraph::firstWordDocBook(odocstream & os, OutputParams const & runparams)
2308         const
2309 {
2310         pos_type i;
2311         for (i = 0; i < size(); ++i) {
2312                 if (Inset const * inset = getInset(i)) {
2313                         inset->docbook(os, runparams);
2314                 } else {
2315                         char_type c = d->text_[i];
2316                         if (c == ' ')
2317                                 break;
2318                         os << sgml::escapeChar(c);
2319                 }
2320         }
2321         return i;
2322 }
2323
2324
2325 pos_type Paragraph::firstWordLyXHTML(XHTMLStream & xs, OutputParams const & runparams)
2326         const
2327 {
2328         pos_type i;
2329         for (i = 0; i < size(); ++i) {
2330                 if (Inset const * inset = getInset(i)) {
2331                         inset->xhtml(xs, runparams);
2332                 } else {
2333                         char_type c = d->text_[i];
2334                         if (c == ' ')
2335                                 break;
2336                         xs << c;
2337                 }
2338         }
2339         return i;
2340 }
2341
2342
2343 bool Paragraph::Private::onlyText(Buffer const & buf, Font const & outerfont, pos_type initial) const
2344 {
2345         Font font_old;
2346         pos_type size = text_.size();
2347         for (pos_type i = initial; i < size; ++i) {
2348                 Font font = owner_->getFont(buf.params(), i, outerfont);
2349                 if (text_[i] == META_INSET)
2350                         return false;
2351                 if (i != initial && font != font_old)
2352                         return false;
2353                 font_old = font;
2354         }
2355
2356         return true;
2357 }
2358
2359
2360 void Paragraph::simpleDocBookOnePar(Buffer const & buf,
2361                                     odocstream & os,
2362                                     OutputParams const & runparams,
2363                                     Font const & outerfont,
2364                                     pos_type initial) const
2365 {
2366         bool emph_flag = false;
2367
2368         Layout const & style = *d->layout_;
2369         FontInfo font_old =
2370                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2371
2372         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2373                 os << "]]>";
2374
2375         // parsing main loop
2376         for (pos_type i = initial; i < size(); ++i) {
2377                 Font font = getFont(buf.params(), i, outerfont);
2378
2379                 // handle <emphasis> tag
2380                 if (font_old.emph() != font.fontInfo().emph()) {
2381                         if (font.fontInfo().emph() == FONT_ON) {
2382                                 os << "<emphasis>";
2383                                 emph_flag = true;
2384                         } else if (i != initial) {
2385                                 os << "</emphasis>";
2386                                 emph_flag = false;
2387                         }
2388                 }
2389
2390                 if (Inset const * inset = getInset(i)) {
2391                         inset->docbook(os, runparams);
2392                 } else {
2393                         char_type c = d->text_[i];
2394
2395                         if (style.pass_thru)
2396                                 os.put(c);
2397                         else
2398                                 os << sgml::escapeChar(c);
2399                 }
2400                 font_old = font.fontInfo();
2401         }
2402
2403         if (emph_flag) {
2404                 os << "</emphasis>";
2405         }
2406
2407         if (style.free_spacing)
2408                 os << '\n';
2409         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2410                 os << "<![CDATA[";
2411 }
2412
2413
2414 docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
2415                                     XHTMLStream & xs,
2416                                     OutputParams const & runparams,
2417                                     Font const & outerfont,
2418                                     pos_type initial) const
2419 {
2420         docstring retval;
2421
2422         bool emph_flag = false;
2423         bool bold_flag = false;
2424         string closing_tag;
2425
2426         Layout const & style = *d->layout_;
2427
2428         if (!runparams.for_toc && runparams.html_make_pars) {
2429                 // generate a magic label for this paragraph
2430                 string const attr = "id='" + magicLabel() + "'";
2431                 xs << html::CompTag("a", attr);
2432         }
2433
2434         FontInfo font_old =
2435                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2436
2437         // parsing main loop
2438         for (pos_type i = initial; i < size(); ++i) {
2439                 // let's not show deleted material in the output
2440                 if (isDeleted(i))
2441                         continue;
2442
2443                 Font font = getFont(buf.params(), i, outerfont);
2444
2445                 // emphasis
2446                 if (font_old.emph() != font.fontInfo().emph()) {
2447                         if (font.fontInfo().emph() == FONT_ON) {
2448                                 xs << html::StartTag("em");
2449                                 emph_flag = true;
2450                         } else if (emph_flag && i != initial) {
2451                                 xs << html::EndTag("em");
2452                                 emph_flag = false;
2453                         }
2454                 }
2455                 // bold
2456                 if (font_old.series() != font.fontInfo().series()) {
2457                         if (font.fontInfo().series() == BOLD_SERIES) {
2458                                 xs << html::StartTag("strong");
2459                                 bold_flag = true;
2460                         } else if (bold_flag && i != initial) {
2461                                 xs << html::EndTag("strong");
2462                                 bold_flag = false;
2463                         }
2464                 }
2465                 // FIXME XHTML
2466                 // Other such tags? What about the other text ranges?
2467
2468                 Inset const * inset = getInset(i);
2469                 if (inset) {
2470                         InsetCommand const * ic = inset->asInsetCommand();
2471                         InsetLayout const & il = inset->getLayout();
2472                         InsetMath const * im = inset->asInsetMath();
2473                         if (!runparams.for_toc
2474                             || im || il.isInToc() || (ic && ic->isInToc())) {
2475                                 OutputParams np = runparams;
2476                                 if (!il.htmlisblock())
2477                                         np.html_in_par = true;
2478                                 retval += inset->xhtml(xs, np);
2479                         }
2480                 } else {
2481                         char_type c = d->text_[i];
2482
2483                         if (style.pass_thru)
2484                                 xs << c;
2485                         else if (c == '-') {
2486                                 docstring str;
2487                                 int j = i + 1;
2488                                 if (j < size() && d->text_[j] == '-') {
2489                                         j += 1;
2490                                         if (j < size() && d->text_[j] == '-') {
2491                                                 str += from_ascii("&mdash;");
2492                                                 i += 2;
2493                                         } else {
2494                                                 str += from_ascii("&ndash;");
2495                                                 i += 1;
2496                                         }
2497                                 }
2498                                 else
2499                                         str += c;
2500                                 // We don't want to escape the entities. Note that
2501                                 // it is safe to do this, since str can otherwise
2502                                 // only be "-". E.g., it can't be "<".
2503                                 xs << XHTMLStream::NextRaw() << str;
2504                         } else
2505                                 xs << c;
2506                 }
2507                 font_old = font.fontInfo();
2508         }
2509
2510         xs.closeFontTags();
2511         return retval;
2512 }
2513
2514
2515 bool Paragraph::isHfill(pos_type pos) const
2516 {
2517         Inset const * inset = getInset(pos);
2518         return inset && (inset->lyxCode() == SPACE_CODE &&
2519                          inset->isStretchableSpace());
2520 }
2521
2522
2523 bool Paragraph::isNewline(pos_type pos) const
2524 {
2525         Inset const * inset = getInset(pos);
2526         return inset && inset->lyxCode() == NEWLINE_CODE;
2527 }
2528
2529
2530 bool Paragraph::isLineSeparator(pos_type pos) const
2531 {
2532         char_type const c = d->text_[pos];
2533         if (isLineSeparatorChar(c))
2534                 return true;
2535         Inset const * inset = getInset(pos);
2536         return inset && inset->isLineSeparator();
2537 }
2538
2539
2540 bool Paragraph::isWordSeparator(pos_type pos) const
2541 {
2542         if (Inset const * inset = getInset(pos))
2543                 return !inset->isLetter();
2544         char_type const c = d->text_[pos];
2545         // We want to pass the ' and escape chars to the spellchecker
2546         static docstring const quote = from_utf8(lyxrc.spellchecker_esc_chars + '\'');
2547         return (!isLetterChar(c) && !isDigit(c) && !contains(quote, c))
2548                 || pos == size();
2549 }
2550
2551
2552 bool Paragraph::isChar(pos_type pos) const
2553 {
2554         if (Inset const * inset = getInset(pos))
2555                 return inset->isChar();
2556         char_type const c = d->text_[pos];
2557         return !isLetterChar(c) && !isDigit(c) && !lyx::isSpace(c);
2558 }
2559
2560
2561 bool Paragraph::isSpace(pos_type pos) const
2562 {
2563         if (Inset const * inset = getInset(pos))
2564                 return inset->isSpace();
2565         char_type const c = d->text_[pos];
2566         return lyx::isSpace(c);
2567 }
2568
2569
2570 Language const *
2571 Paragraph::getParLanguage(BufferParams const & bparams) const
2572 {
2573         if (!empty())
2574                 return getFirstFontSettings(bparams).language();
2575         // FIXME: we should check the prev par as well (Lgb)
2576         return bparams.language;
2577 }
2578
2579
2580 bool Paragraph::isRTL(BufferParams const & bparams) const
2581 {
2582         return lyxrc.rtl_support
2583                 && getParLanguage(bparams)->rightToLeft()
2584                 && !inInset().getLayout().forceLTR();
2585 }
2586
2587
2588 void Paragraph::changeLanguage(BufferParams const & bparams,
2589                                Language const * from, Language const * to)
2590 {
2591         // change language including dummy font change at the end
2592         for (pos_type i = 0; i <= size(); ++i) {
2593                 Font font = getFontSettings(bparams, i);
2594                 if (font.language() == from) {
2595                         font.setLanguage(to);
2596                         setFont(i, font);
2597                 }
2598         }
2599 }
2600
2601
2602 bool Paragraph::isMultiLingual(BufferParams const & bparams) const
2603 {
2604         Language const * doc_language = bparams.language;
2605         FontList::const_iterator cit = d->fontlist_.begin();
2606         FontList::const_iterator end = d->fontlist_.end();
2607
2608         for (; cit != end; ++cit)
2609                 if (cit->font().language() != ignore_language &&
2610                     cit->font().language() != latex_language &&
2611                     cit->font().language() != doc_language)
2612                         return true;
2613         return false;
2614 }
2615
2616
2617 docstring Paragraph::asString(int options) const
2618 {
2619         return asString(0, size(), options);
2620 }
2621
2622
2623 docstring Paragraph::asString(pos_type beg, pos_type end, int options) const
2624 {
2625         odocstringstream os;
2626
2627         if (beg == 0
2628                 && options & AS_STR_LABEL
2629                 && !d->params_.labelString().empty())
2630                 os << d->params_.labelString() << ' ';
2631
2632         for (pos_type i = beg; i < end; ++i) {
2633                 char_type const c = d->text_[i];
2634                 if (isPrintable(c) || c == '\t'
2635                     || (c == '\n' && (options & AS_STR_NEWLINES)))
2636                         os.put(c);
2637                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
2638                         getInset(i)->tocString(os);
2639                         if (getInset(i)->asInsetMath())
2640                                 os << " ";
2641                 }
2642         }
2643
2644         return os.str();
2645 }
2646
2647
2648 docstring Paragraph::stringify(pos_type beg, pos_type end, int options, OutputParams & runparams) const
2649 {
2650         odocstringstream os;
2651
2652         if (beg == 0
2653                 && options & AS_STR_LABEL
2654                 && !d->params_.labelString().empty())
2655                 os << d->params_.labelString() << ' ';
2656
2657         for (pos_type i = beg; i < end; ++i) {
2658                 char_type const c = d->text_[i];
2659                 if (isPrintable(c) || c == '\t'
2660                     || (c == '\n' && (options & AS_STR_NEWLINES)))
2661                         os.put(c);
2662                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
2663                         getInset(i)->plaintext(os, runparams);
2664                 }
2665         }
2666
2667         return os.str();
2668 }
2669
2670
2671 void Paragraph::setInsetOwner(Inset const * inset)
2672 {
2673         d->inset_owner_ = inset;
2674 }
2675
2676
2677 int Paragraph::id() const
2678 {
2679         return d->id_;
2680 }
2681
2682
2683 void Paragraph::setId(int id)
2684 {
2685         d->id_ = id;
2686 }
2687
2688
2689 Layout const & Paragraph::layout() const
2690 {
2691         return *d->layout_;
2692 }
2693
2694
2695 void Paragraph::setLayout(Layout const & layout)
2696 {
2697         d->layout_ = &layout;
2698 }
2699
2700
2701 void Paragraph::setDefaultLayout(DocumentClass const & tc)
2702 {
2703         setLayout(tc.defaultLayout());
2704 }
2705
2706
2707 void Paragraph::setPlainLayout(DocumentClass const & tc)
2708 {
2709         setLayout(tc.plainLayout());
2710 }
2711
2712
2713 void Paragraph::setPlainOrDefaultLayout(DocumentClass const & tclass)
2714 {
2715         if (usePlainLayout())
2716                 setPlainLayout(tclass);
2717         else
2718                 setDefaultLayout(tclass);
2719 }
2720
2721
2722 Inset const & Paragraph::inInset() const
2723 {
2724         LASSERT(d->inset_owner_, throw ExceptionMessage(BufferException,
2725                 _("Memory problem"), _("Paragraph not properly initialized")));
2726         return *d->inset_owner_;
2727 }
2728
2729
2730 ParagraphParameters & Paragraph::params()
2731 {
2732         return d->params_;
2733 }
2734
2735
2736 ParagraphParameters const & Paragraph::params() const
2737 {
2738         return d->params_;
2739 }
2740
2741
2742 bool Paragraph::isFreeSpacing() const
2743 {
2744         if (d->layout_->free_spacing)
2745                 return true;
2746         return d->inset_owner_ && d->inset_owner_->isFreeSpacing();
2747 }
2748
2749
2750 bool Paragraph::allowEmpty() const
2751 {
2752         if (d->layout_->keepempty)
2753                 return true;
2754         return d->inset_owner_ && d->inset_owner_->allowEmpty();
2755 }
2756
2757
2758 char_type Paragraph::transformChar(char_type c, pos_type pos) const
2759 {
2760         if (!Encodings::isArabicChar(c))
2761                 return c;
2762
2763         char_type prev_char = ' ';
2764         char_type next_char = ' ';
2765
2766         for (pos_type i = pos - 1; i >= 0; --i) {
2767                 char_type const par_char = d->text_[i];
2768                 if (!Encodings::isArabicComposeChar(par_char)) {
2769                         prev_char = par_char;
2770                         break;
2771                 }
2772         }
2773
2774         for (pos_type i = pos + 1, end = size(); i < end; ++i) {
2775                 char_type const par_char = d->text_[i];
2776                 if (!Encodings::isArabicComposeChar(par_char)) {
2777                         next_char = par_char;
2778                         break;
2779                 }
2780         }
2781
2782         if (Encodings::isArabicChar(next_char)) {
2783                 if (Encodings::isArabicChar(prev_char) &&
2784                         !Encodings::isArabicSpecialChar(prev_char))
2785                         return Encodings::transformChar(c, Encodings::FORM_MEDIAL);
2786                 else
2787                         return Encodings::transformChar(c, Encodings::FORM_INITIAL);
2788         } else {
2789                 if (Encodings::isArabicChar(prev_char) &&
2790                         !Encodings::isArabicSpecialChar(prev_char))
2791                         return Encodings::transformChar(c, Encodings::FORM_FINAL);
2792                 else
2793                         return Encodings::transformChar(c, Encodings::FORM_ISOLATED);
2794         }
2795 }
2796
2797
2798 int Paragraph::checkBiblio(Buffer const & buffer)
2799 {
2800         // FIXME From JS:
2801         // This is getting more and more a mess. ...We really should clean
2802         // up this bibitem issue for 1.6. See also bug 2743.
2803
2804         // Add bibitem insets if necessary
2805         if (d->layout_->labeltype != LABEL_BIBLIO)
2806                 return 0;
2807
2808         bool hasbibitem = !d->insetlist_.empty()
2809                 // Insist on it being in pos 0
2810                 && d->text_[0] == META_INSET
2811                 && d->insetlist_.begin()->inset->lyxCode() == BIBITEM_CODE;
2812
2813         bool track_changes = buffer.params().trackChanges;
2814
2815         docstring oldkey;
2816         docstring oldlabel;
2817
2818         // remove a bibitem in pos != 0
2819         // restore it later in pos 0 if necessary
2820         // (e.g. if a user inserts contents _before_ the item)
2821         // we're assuming there's only one of these, which there
2822         // should be.
2823         int erasedInsetPosition = -1;
2824         InsetList::iterator it = d->insetlist_.begin();
2825         InsetList::iterator end = d->insetlist_.end();
2826         for (; it != end; ++it)
2827                 if (it->inset->lyxCode() == BIBITEM_CODE
2828                     && it->pos > 0) {
2829                         InsetBibitem * olditem = static_cast<InsetBibitem *>(it->inset);
2830                         oldkey = olditem->getParam("key");
2831                         oldlabel = olditem->getParam("label");
2832                         erasedInsetPosition = it->pos;
2833                         eraseChar(erasedInsetPosition, track_changes);
2834                         break;
2835         }
2836
2837         // There was an InsetBibitem at the beginning, and we didn't
2838         // have to erase one.
2839         if (hasbibitem && erasedInsetPosition < 0)
2840                         return 0;
2841
2842         // There was an InsetBibitem at the beginning and we did have to
2843         // erase one. So we give its properties to the beginning inset.
2844         if (hasbibitem) {
2845                 InsetBibitem * inset =
2846                         static_cast<InsetBibitem *>(d->insetlist_.begin()->inset);
2847                 if (!oldkey.empty())
2848                         inset->setParam("key", oldkey);
2849                 inset->setParam("label", oldlabel);
2850                 return -erasedInsetPosition;
2851         }
2852
2853         // There was no inset at the beginning, so we need to create one with
2854         // the key and label of the one we erased.
2855         InsetBibitem * inset =
2856                 new InsetBibitem(const_cast<Buffer *>(&buffer), InsetCommandParams(BIBITEM_CODE));
2857         // restore values of previously deleted item in this par.
2858         if (!oldkey.empty())
2859                 inset->setParam("key", oldkey);
2860         inset->setParam("label", oldlabel);
2861         insertInset(0, static_cast<Inset *>(inset),
2862                     Change(track_changes ? Change::INSERTED : Change::UNCHANGED));
2863
2864         return 1;
2865 }
2866
2867
2868 void Paragraph::checkAuthors(AuthorList const & authorList)
2869 {
2870         d->changes_.checkAuthors(authorList);
2871 }
2872
2873
2874 bool Paragraph::isChanged(pos_type pos) const
2875 {
2876         return lookupChange(pos).changed();
2877 }
2878
2879
2880 bool Paragraph::isInserted(pos_type pos) const
2881 {
2882         return lookupChange(pos).inserted();
2883 }
2884
2885
2886 bool Paragraph::isDeleted(pos_type pos) const
2887 {
2888         return lookupChange(pos).deleted();
2889 }
2890
2891
2892 InsetList const & Paragraph::insetList() const
2893 {
2894         return d->insetlist_;
2895 }
2896
2897
2898 void Paragraph::setBuffer(Buffer & b)
2899 {
2900         d->insetlist_.setBuffer(b);
2901 }
2902
2903
2904 Inset * Paragraph::releaseInset(pos_type pos)
2905 {
2906         Inset * inset = d->insetlist_.release(pos);
2907         /// does not honour change tracking!
2908         eraseChar(pos, false);
2909         return inset;
2910 }
2911
2912
2913 Inset * Paragraph::getInset(pos_type pos)
2914 {
2915         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
2916                  ? d->insetlist_.get(pos) : 0;
2917 }
2918
2919
2920 Inset const * Paragraph::getInset(pos_type pos) const
2921 {
2922         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
2923                  ? d->insetlist_.get(pos) : 0;
2924 }
2925
2926
2927 void Paragraph::changeCase(BufferParams const & bparams, pos_type pos,
2928                 pos_type & right, TextCase action)
2929 {
2930         // process sequences of modified characters; in change
2931         // tracking mode, this approach results in much better
2932         // usability than changing case on a char-by-char basis
2933         docstring changes;
2934
2935         bool const trackChanges = bparams.trackChanges;
2936
2937         bool capitalize = true;
2938
2939         for (; pos < right; ++pos) {
2940                 char_type oldChar = d->text_[pos];
2941                 char_type newChar = oldChar;
2942
2943                 // ignore insets and don't play with deleted text!
2944                 if (oldChar != META_INSET && !isDeleted(pos)) {
2945                         switch (action) {
2946                                 case text_lowercase:
2947                                         newChar = lowercase(oldChar);
2948                                         break;
2949                                 case text_capitalization:
2950                                         if (capitalize) {
2951                                                 newChar = uppercase(oldChar);
2952                                                 capitalize = false;
2953                                         }
2954                                         break;
2955                                 case text_uppercase:
2956                                         newChar = uppercase(oldChar);
2957                                         break;
2958                         }
2959                 }
2960
2961                 if (isWordSeparator(pos) || isDeleted(pos)) {
2962                         // permit capitalization again
2963                         capitalize = true;
2964                 }
2965
2966                 if (oldChar != newChar) {
2967                         changes += newChar;
2968                         if (pos != right - 1)
2969                                 continue;
2970                         // step behind the changing area
2971                         pos++;
2972                 }
2973
2974                 int erasePos = pos - changes.size();
2975                 for (size_t i = 0; i < changes.size(); i++) {
2976                         insertChar(pos, changes[i],
2977                                    getFontSettings(bparams,
2978                                                    erasePos),
2979                                    trackChanges);
2980                         if (!eraseChar(erasePos, trackChanges)) {
2981                                 ++erasePos;
2982                                 ++pos; // advance
2983                                 ++right; // expand selection
2984                         }
2985                 }
2986                 changes.clear();
2987         }
2988 }
2989
2990
2991 bool Paragraph::find(docstring const & str, bool cs, bool mw,
2992                 pos_type pos, bool del) const
2993 {
2994         int const strsize = str.length();
2995         int i = 0;
2996         pos_type const parsize = d->text_.size();
2997         for (i = 0; pos + i < parsize; ++i) {
2998                 if (i >= strsize)
2999                         break;
3000                 if (cs && str[i] != d->text_[pos + i])
3001                         break;
3002                 if (!cs && uppercase(str[i]) != uppercase(d->text_[pos + i]))
3003                         break;
3004                 if (!del && isDeleted(pos + i))
3005                         break;
3006         }
3007
3008         if (i != strsize)
3009                 return false;
3010
3011         // if necessary, check whether string matches word
3012         if (mw) {
3013                 if (pos > 0 && !isWordSeparator(pos - 1))
3014                         return false;
3015                 if (pos + strsize < parsize
3016                         && !isWordSeparator(pos + strsize))
3017                         return false;
3018         }
3019
3020         return true;
3021 }
3022
3023
3024 char_type Paragraph::getChar(pos_type pos) const
3025 {
3026         return d->text_[pos];
3027 }
3028
3029
3030 pos_type Paragraph::size() const
3031 {
3032         return d->text_.size();
3033 }
3034
3035
3036 bool Paragraph::empty() const
3037 {
3038         return d->text_.empty();
3039 }
3040
3041
3042 bool Paragraph::isInset(pos_type pos) const
3043 {
3044         return d->text_[pos] == META_INSET;
3045 }
3046
3047
3048 bool Paragraph::isSeparator(pos_type pos) const
3049 {
3050         //FIXME: Are we sure this can be the only separator?
3051         return d->text_[pos] == ' ';
3052 }
3053
3054
3055 void Paragraph::deregisterWords()
3056 {
3057         map<Language, Private::Words>::const_iterator itl;
3058         Private::Words::const_iterator it;
3059         for (itl = d->words_.begin(); itl != d->words_.end(); ++itl) {
3060                 WordList * wl = theWordList(itl->first);
3061                 for (it = (itl->second).begin(); it != (itl->second).end(); ++it)
3062                         wl->remove(*it);
3063         }
3064         d->words_.clear();
3065 }
3066
3067
3068 void Paragraph::locateWord(pos_type & from, pos_type & to,
3069         word_location const loc) const
3070 {
3071         switch (loc) {
3072         case WHOLE_WORD_STRICT:
3073                 if (from == 0 || from == size()
3074                     || isWordSeparator(from)
3075                     || isWordSeparator(from - 1)) {
3076                         to = from;
3077                         return;
3078                 }
3079                 // no break here, we go to the next
3080
3081         case WHOLE_WORD:
3082                 // If we are already at the beginning of a word, do nothing
3083                 if (!from || isWordSeparator(from - 1))
3084                         break;
3085                 // no break here, we go to the next
3086
3087         case PREVIOUS_WORD:
3088                 // always move the cursor to the beginning of previous word
3089                 while (from && !isWordSeparator(from - 1))
3090                         --from;
3091                 break;
3092         case NEXT_WORD:
3093                 LYXERR0("Paragraph::locateWord: NEXT_WORD not implemented yet");
3094                 break;
3095         case PARTIAL_WORD:
3096                 // no need to move the 'from' cursor
3097                 break;
3098         }
3099         to = from;
3100         while (to < size() && !isWordSeparator(to))
3101                 ++to;
3102 }
3103
3104
3105 void Paragraph::collectWords()
3106 {
3107         pos_type n = size();
3108         for (pos_type pos = 0; pos < n; ++pos) {
3109                 if (isWordSeparator(pos))
3110                         continue;
3111                 pos_type from = pos;
3112                 locateWord(from, pos, WHOLE_WORD);
3113                 if (pos - from >= 6) {
3114                         docstring word = asString(from, pos, AS_STR_NONE);
3115                         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
3116                         if (cit == d->fontlist_.end())
3117                                 return;
3118                         Language const * lang = cit->font().language();
3119                         d->words_[*lang].insert(word);
3120                 }
3121         }
3122 }
3123
3124
3125 void Paragraph::registerWords()
3126 {
3127         map<Language, Private::Words>::const_iterator itl;
3128         Private::Words::const_iterator it;
3129         for (itl = d->words_.begin(); itl != d->words_.end(); ++itl) {
3130                 WordList * wl = theWordList(itl->first);
3131                 for (it = (itl->second).begin(); it != (itl->second).end(); ++it)
3132                         wl->insert(*it);
3133         }
3134 }
3135
3136
3137 void Paragraph::updateWords()
3138 {
3139         deregisterWords();
3140         collectWords();
3141         registerWords();
3142 }
3143
3144
3145 bool Paragraph::spellCheck(pos_type & from, pos_type & to, WordLangTuple & wl,
3146         docstring_list & suggestions, bool do_suggestion) const
3147 {
3148         SpellChecker * speller = theSpellChecker();
3149         if (!speller)
3150                 return false;
3151
3152         locateWord(from, to, WHOLE_WORD);
3153         if (from == to || from >= pos_type(d->text_.size()))
3154                 return false;
3155
3156         docstring word = asString(from, to, AS_STR_INSETS);
3157         string lang_code;
3158         string lang_variety;
3159         if (!lyxrc.spellchecker_alt_lang.empty())
3160                 lang_variety = split(lyxrc.spellchecker_alt_lang, lang_code, '-');
3161         else {
3162               lang_code = getFontSettings(
3163                     d->inset_owner_->buffer().params(), from).language()->code();
3164               lang_variety = getFontSettings(
3165                     d->inset_owner_->buffer().params(), from).language()->variety();
3166         }
3167         wl = WordLangTuple(word, lang_code, lang_variety);
3168         SpellChecker::Result res = speller->check(wl);
3169         // Just ignore any error that the spellchecker reports.
3170         // FIXME: we should through out an exception and catch it in the GUI to
3171         // display the error.
3172         if (!speller->error().empty())
3173                 return false;
3174
3175         bool const misspelled = res != SpellChecker::OK
3176                 && res != SpellChecker::IGNORED_WORD;
3177
3178         if (lyxrc.spellcheck_continuously)
3179                 d->fontlist_.setMisspelled(from, to, misspelled);
3180
3181         if (misspelled && do_suggestion)
3182                 speller->suggest(wl, suggestions);
3183         else
3184                 suggestions.clear();
3185
3186         return misspelled;
3187 }
3188
3189
3190 bool Paragraph::isMisspelled(pos_type pos) const
3191 {
3192         pos_type from = pos;
3193         pos_type to = pos;
3194         WordLangTuple wl;
3195         docstring_list suggestions;
3196         return spellCheck(from, to, wl, suggestions, false);
3197 }
3198
3199
3200 string Paragraph::magicLabel() const
3201 {
3202         stringstream ss;
3203         ss << "magicparlabel-" << id();
3204         return ss.str();
3205 }
3206
3207
3208 } // namespace lyx