src/Paragraph.cpp

   1 /**
   2  * \file Paragraph.cpp
   3  * This file is part of LyX, the document processor.
   4  * Licence details can be found in the file COPYING.
   5  *
   6  * \author Asger Alstrup
   7  * \author Lars Gullik Bjønnes
   8  * \author Richard Heck (XHTML output)
   9  * \author Jean-Marc Lasgouttes
  10  * \author Angus Leeming
  11  * \author John Levon
  12  * \author André Pönitz
  13  * \author Dekel Tsur
  14  * \author Jürgen Vigna
  15  *
  16  * Full author contact details are available in file CREDITS.
  17  */
  18
  19 #include <config.h>
  20
  21 #include "Paragraph.h"
  22
  23 #include "LayoutFile.h"
  24 #include "Buffer.h"
  25 #include "BufferParams.h"
  26 #include "Changes.h"
  27 #include "Counters.h"
  28 #include "Encoding.h"
  29 #include "InsetList.h"
  30 #include "Language.h"
  31 #include "LaTeXFeatures.h"
  32 #include "Layout.h"
  33 #include "Length.h"
  34 #include "Font.h"
  35 #include "FontList.h"
  36 #include "LyXRC.h"
  37 #include "OutputParams.h"
  38 #include "output_latex.h"
  39 #include "output_xhtml.h"
  40 #include "ParagraphParameters.h"
  41 #include "SpellChecker.h"
  42 #include "sgml.h"
  43 #include "TextClass.h"
  44 #include "TexRow.h"
  45 #include "Text.h"
  46 #include "VSpace.h"
  47 #include "WordLangTuple.h"
  48 #include "WordList.h"
  49
  50 #include "frontends/alert.h"
  51
  52 #include "insets/InsetBibitem.h"
  53 #include "insets/InsetLabel.h"
  54
  55 #include "support/debug.h"
  56 #include "support/docstring_list.h"
  57 #include "support/ExceptionMessage.h"
  58 #include "support/gettext.h"
  59 #include "support/lassert.h"
  60 #include "support/lstrings.h"
  61 #include "support/textutils.h"
  62
  63 #include <sstream>
  64 #include <vector>
  65
  66 using namespace std;
  67 using namespace lyx::support;
  68
  69 namespace lyx {
  70
  71 namespace {
  72 /// Inset identifier (above 0x10ffff, for ucs-4)
  73 char_type const META_INSET = 0x200001;
  74 };
  75
  76 /////////////////////////////////////////////////////////////////////
  77 //
  78 // Paragraph::Private
  79 //
  80 /////////////////////////////////////////////////////////////////////
  81
  82 class Paragraph::Private
  83 {
  84 public:
  85         ///
  86         Private(Paragraph * owner, Layout const & layout);
  87         /// "Copy constructor"
  88         Private(Private const &, Paragraph * owner);
  89         /// Copy constructor from \p beg  to \p end
  90         Private(Private const &, Paragraph * owner, pos_type beg, pos_type end);
  91
  92         ///
  93         void insertChar(pos_type pos, char_type c, Change const & change);
  94
  95         /// Output the surrogate pair formed by \p c and \p next to \p os.
  96         /// \return the number of characters written.
  97         int latexSurrogatePair(odocstream & os, char_type c, char_type next,
  98                                OutputParams const &);
  99
 100         /// Output a space in appropriate formatting (or a surrogate pair
 101         /// if the next character is a combining character).
 102         /// \return whether a surrogate pair was output.
 103         bool simpleTeXBlanks(OutputParams const &,
 104                              odocstream &, TexRow & texrow,
 105                              pos_type i,
 106                              unsigned int & column,
 107                              Font const & font,
 108                              Layout const & style);
 109
 110         /// Output consecutive unicode chars, belonging to the same script as
 111         /// specified by the latex macro \p ltx, to \p os starting from \p i.
 112         /// \return the number of characters written.
 113         int writeScriptChars(odocstream & os, docstring const & ltx,
 114                            Change const &, Encoding const &, pos_type & i);
 115
 116         /// This could go to ParagraphParameters if we want to.
 117         int startTeXParParams(BufferParams const &, odocstream &, TexRow &,
 118                               OutputParams const &) const;
 119
 120         /// This could go to ParagraphParameters if we want to.
 121         int endTeXParParams(BufferParams const &, odocstream &, TexRow &,
 122                             OutputParams const &) const;
 123
 124         ///
 125         void latexInset(BufferParams const &,
 126                                    odocstream &,
 127                                    TexRow & texrow, OutputParams &,
 128                                    Font & running_font,
 129                                    Font & basefont,
 130                                    Font const & outerfont,
 131                                    bool & open_font,
 132                                    Change & running_change,
 133                                    Layout const & style,
 134                                    pos_type & i,
 135                                    unsigned int & column);
 136
 137         ///
 138         void latexSpecialChar(
 139                                    odocstream & os,
 140                                    OutputParams const & runparams,
 141                                    Font const & running_font,
 142                                    Change const & running_change,
 143                                    Layout const & style,
 144                                    pos_type & i,
 145                                    unsigned int & column);
 146
 147         ///
 148         bool latexSpecialT1(
 149                 char_type const c,
 150                 odocstream & os,
 151                 pos_type i,
 152                 unsigned int & column);
 153         ///
 154         bool latexSpecialTypewriter(
 155                 char_type const c,
 156                 odocstream & os,
 157                 pos_type i,
 158                 unsigned int & column);
 159         ///
 160         bool latexSpecialPhrase(
 161                 odocstream & os,
 162                 pos_type & i,
 163                 unsigned int & column,
 164                 OutputParams const & runparams);
 165
 166         ///
 167         void validate(LaTeXFeatures & features) const;
 168
 169         /// Checks if the paragraph contains only text and no inset or font change.
 170         bool onlyText(Buffer const & buf, Font const & outerfont,
 171                       pos_type initial) const;
 172
 173         /// match a string against a particular point in the paragraph
 174         bool isTextAt(string const & str, pos_type pos) const;
 175
 176
 177         InsetCode ownerCode() const
 178         {
 179                 return inset_owner_ ? inset_owner_->lyxCode() : NO_CODE;
 180         }
 181
 182         /// Which Paragraph owns us?
 183         Paragraph * owner_;
 184
 185         /// In which Inset?
 186         Inset const * inset_owner_;
 187
 188         ///
 189         FontList fontlist_;
 190
 191         ///
 192         int id_;
 193
 194         ///
 195         ParagraphParameters params_;
 196
 197         /// for recording and looking up changes
 198         Changes changes_;
 199
 200         ///
 201         InsetList insetlist_;
 202
 203         /// end of label
 204         pos_type begin_of_body_;
 205
 206         typedef docstring TextContainer;
 207         ///
 208         TextContainer text_;
 209
 210         typedef set<docstring> Words;
 211         typedef map<Language, Words> LangWordsMap;
 212         ///
 213         LangWordsMap words_;
 214         ///
 215         Layout const * layout_;
 216 };
 217
 218
 219 namespace {
 220
 221 struct special_phrase {
 222         string phrase;
 223         docstring macro;
 224         bool builtin;
 225 };
 226
 227 special_phrase const special_phrases[] = {
 228         { "LyX", from_ascii("\\LyX{}"), false },
 229         { "TeX", from_ascii("\\TeX{}"), true },
 230         { "LaTeX2e", from_ascii("\\LaTeXe{}"), true },
 231         { "LaTeX", from_ascii("\\LaTeX{}"), true },
 232 };
 233
 234 size_t const phrases_nr = sizeof(special_phrases)/sizeof(special_phrase);
 235
 236 } // namespace anon
 237
 238
 239 Paragraph::Private::Private(Paragraph * owner, Layout const & layout)
 240         : owner_(owner), inset_owner_(0), id_(-1), begin_of_body_(0), layout_(&layout)
 241 {
 242         text_.reserve(100);
 243 }
 244
 245
 246 // Initialization of the counter for the paragraph id's,
 247 //
 248 // FIXME: There should be a more intelligent way to generate and use the
 249 // paragraph ids per buffer instead a global static counter for all InsetText
 250 // in the running program.
 251 static int paragraph_id = -1;
 252
 253 Paragraph::Private::Private(Private const & p, Paragraph * owner)
 254         : owner_(owner), inset_owner_(p.inset_owner_), fontlist_(p.fontlist_),
 255           params_(p.params_), changes_(p.changes_), insetlist_(p.insetlist_),
 256           begin_of_body_(p.begin_of_body_), text_(p.text_), words_(p.words_),
 257           layout_(p.layout_)
 258 {
 259         id_ = ++paragraph_id;
 260 }
 261
 262
 263 Paragraph::Private::Private(Private const & p, Paragraph * owner,
 264         pos_type beg, pos_type end)
 265         : owner_(owner), inset_owner_(p.inset_owner_),
 266           params_(p.params_), changes_(p.changes_),
 267           insetlist_(p.insetlist_, beg, end),
 268           begin_of_body_(p.begin_of_body_), words_(p.words_),
 269           layout_(p.layout_)
 270 {
 271         id_ = ++paragraph_id;
 272         if (beg >= pos_type(p.text_.size()))
 273                 return;
 274         text_ = p.text_.substr(beg, end - beg);
 275
 276         FontList::const_iterator fcit = fontlist_.begin();
 277         FontList::const_iterator fend = fontlist_.end();
 278         for (; fcit != fend; ++fcit) {
 279                 if (fcit->pos() < beg)
 280                         continue;
 281                 if (fcit->pos() >= end) {
 282                         // Add last entry in the fontlist_.
 283                         fontlist_.set(text_.size() - 1, fcit->font());
 284                         break;
 285                 }
 286                 // Add a new entry in the fontlist_.
 287                 fontlist_.set(fcit->pos() - beg, fcit->font());
 288         }
 289 }
 290
 291
 292 void Paragraph::addChangesToToc(DocIterator const & cdit,
 293         Buffer const & buf) const
 294 {
 295         d->changes_.addToToc(cdit, buf);
 296 }
 297
 298
 299 bool Paragraph::isDeleted(pos_type start, pos_type end) const
 300 {
 301         LASSERT(start >= 0 && start <= size(), /**/);
 302         LASSERT(end > start && end <= size() + 1, /**/);
 303
 304         return d->changes_.isDeleted(start, end);
 305 }
 306
 307
 308 bool Paragraph::isChanged(pos_type start, pos_type end) const
 309 {
 310         LASSERT(start >= 0 && start <= size(), /**/);
 311         LASSERT(end > start && end <= size() + 1, /**/);
 312
 313         return d->changes_.isChanged(start, end);
 314 }
 315
 316
 317 bool Paragraph::isMergedOnEndOfParDeletion(bool trackChanges) const
 318 {
 319         // keep the logic here in sync with the logic of eraseChars()
 320         if (!trackChanges)
 321                 return true;
 322
 323         Change const change = d->changes_.lookup(size());
 324         return change.inserted() && change.currentAuthor();
 325 }
 326
 327
 328 void Paragraph::setChange(Change const & change)
 329 {
 330         // beware of the imaginary end-of-par character!
 331         d->changes_.set(change, 0, size() + 1);
 332
 333         /*
 334          * Propagate the change recursively - but not in case of DELETED!
 335          *
 336          * Imagine that your co-author makes changes in an existing inset. He
 337          * sends your document to you and you come to the conclusion that the
 338          * inset should go completely. If you erase it, LyX must not delete all
 339          * text within the inset. Otherwise, the change tracked insertions of
 340          * your co-author get lost and there is no way to restore them later.
 341          *
 342          * Conclusion: An inset's content should remain untouched if you delete it
 343          */
 344
 345         if (!change.deleted()) {
 346                 for (pos_type pos = 0; pos < size(); ++pos) {
 347                         if (Inset * inset = getInset(pos))
 348                                 inset->setChange(change);
 349                 }
 350         }
 351 }
 352
 353
 354 void Paragraph::setChange(pos_type pos, Change const & change)
 355 {
 356         LASSERT(pos >= 0 && pos <= size(), /**/);
 357         d->changes_.set(change, pos);
 358
 359         // see comment in setChange(Change const &) above
 360         if (!change.deleted() && pos < size())
 361                         if (Inset * inset = getInset(pos))
 362                                 inset->setChange(change);
 363 }
 364
 365
 366 Change const & Paragraph::lookupChange(pos_type pos) const
 367 {
 368         LASSERT(pos >= 0 && pos <= size(), /**/);
 369         return d->changes_.lookup(pos);
 370 }
 371
 372
 373 void Paragraph::acceptChanges(pos_type start, pos_type end)
 374 {
 375         LASSERT(start >= 0 && start <= size(), /**/);
 376         LASSERT(end > start && end <= size() + 1, /**/);
 377
 378         for (pos_type pos = start; pos < end; ++pos) {
 379                 switch (lookupChange(pos).type) {
 380                         case Change::UNCHANGED:
 381                                 // accept changes in nested inset
 382                                 if (Inset * inset = getInset(pos))
 383                                         inset->acceptChanges();
 384                                 break;
 385
 386                         case Change::INSERTED:
 387                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 388                                 // also accept changes in nested inset
 389                                 if (Inset * inset = getInset(pos))
 390                                         inset->acceptChanges();
 391                                 break;
 392
 393                         case Change::DELETED:
 394                                 // Suppress access to non-existent
 395                                 // "end-of-paragraph char"
 396                                 if (pos < size()) {
 397                                         eraseChar(pos, false);
 398                                         --end;
 399                                         --pos;
 400                                 }
 401                                 break;
 402                 }
 403
 404         }
 405 }
 406
 407
 408 void Paragraph::rejectChanges(pos_type start, pos_type end)
 409 {
 410         LASSERT(start >= 0 && start <= size(), /**/);
 411         LASSERT(end > start && end <= size() + 1, /**/);
 412
 413         for (pos_type pos = start; pos < end; ++pos) {
 414                 switch (lookupChange(pos).type) {
 415                         case Change::UNCHANGED:
 416                                 // reject changes in nested inset
 417                                 if (Inset * inset = getInset(pos))
 418                                                 inset->rejectChanges();
 419                                 break;
 420
 421                         case Change::INSERTED:
 422                                 // Suppress access to non-existent
 423                                 // "end-of-paragraph char"
 424                                 if (pos < size()) {
 425                                         eraseChar(pos, false);
 426                                         --end;
 427                                         --pos;
 428                                 }
 429                                 break;
 430
 431                         case Change::DELETED:
 432                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 433
 434                                 // Do NOT reject changes within a deleted inset!
 435                                 // There may be insertions of a co-author inside of it!
 436
 437                                 break;
 438                 }
 439         }
 440 }
 441
 442
 443 void Paragraph::Private::insertChar(pos_type pos, char_type c,
 444                 Change const & change)
 445 {
 446         LASSERT(pos >= 0 && pos <= int(text_.size()), /**/);
 447
 448         // track change
 449         changes_.insert(change, pos);
 450
 451         // This is actually very common when parsing buffers (and
 452         // maybe inserting ascii text)
 453         if (pos == pos_type(text_.size())) {
 454                 // when appending characters, no need to update tables
 455                 text_.push_back(c);
 456                 return;
 457         }
 458
 459         text_.insert(text_.begin() + pos, c);
 460
 461         // Update the font table.
 462         fontlist_.increasePosAfterPos(pos);
 463
 464         // Update the insets
 465         insetlist_.increasePosAfterPos(pos);
 466 }
 467
 468
 469 bool Paragraph::insertInset(pos_type pos, Inset * inset,
 470                                    Change const & change)
 471 {
 472         LASSERT(inset, /**/);
 473         LASSERT(pos >= 0 && pos <= size(), /**/);
 474
 475         // Paragraph::insertInset() can be used in cut/copy/paste operation where
 476         // d->inset_owner_ is not set yet.
 477         if (d->inset_owner_ && !d->inset_owner_->insetAllowed(inset->lyxCode()))
 478                 return false;
 479
 480         d->insertChar(pos, META_INSET, change);
 481         LASSERT(d->text_[pos] == META_INSET, /**/);
 482
 483         // Add a new entry in the insetlist_.
 484         d->insetlist_.insert(inset, pos);
 485         return true;
 486 }
 487
 488
 489 bool Paragraph::eraseChar(pos_type pos, bool trackChanges)
 490 {
 491         LASSERT(pos >= 0 && pos <= size(), return false);
 492
 493         // keep the logic here in sync with the logic of isMergedOnEndOfParDeletion()
 494
 495         if (trackChanges) {
 496                 Change change = d->changes_.lookup(pos);
 497
 498                 // set the character to DELETED if
 499                 //  a) it was previously unchanged or
 500                 //  b) it was inserted by a co-author
 501
 502                 if (!change.changed() ||
 503                       (change.inserted() && !change.currentAuthor())) {
 504                         setChange(pos, Change(Change::DELETED));
 505                         return false;
 506                 }
 507
 508                 if (change.deleted())
 509                         return false;
 510         }
 511
 512         // Don't physically access the imaginary end-of-paragraph character.
 513         // eraseChar() can only mark it as DELETED. A physical deletion of
 514         // end-of-par must be handled externally.
 515         if (pos == size()) {
 516                 return false;
 517         }
 518
 519         // track change
 520         d->changes_.erase(pos);
 521
 522         // if it is an inset, delete the inset entry
 523         if (d->text_[pos] == META_INSET)
 524                 d->insetlist_.erase(pos);
 525
 526         d->text_.erase(d->text_.begin() + pos);
 527
 528         // Update the fontlist_
 529         d->fontlist_.erase(pos);
 530
 531         // Update the insetlist_
 532         d->insetlist_.decreasePosAfterPos(pos);
 533
 534         return true;
 535 }
 536
 537
 538 int Paragraph::eraseChars(pos_type start, pos_type end, bool trackChanges)
 539 {
 540         LASSERT(start >= 0 && start <= size(), /**/);
 541         LASSERT(end >= start && end <= size() + 1, /**/);
 542
 543         pos_type i = start;
 544         for (pos_type count = end - start; count; --count) {
 545                 if (!eraseChar(i, trackChanges))
 546                         ++i;
 547         }
 548         return end - i;
 549 }
 550
 551
 552 int Paragraph::Private::latexSurrogatePair(odocstream & os, char_type c,
 553                 char_type next, OutputParams const & runparams)
 554 {
 555         // Writing next here may circumvent a possible font change between
 556         // c and next. Since next is only output if it forms a surrogate pair
 557         // with c we can ignore this:
 558         // A font change inside a surrogate pair does not make sense and is
 559         // hopefully impossible to input.
 560         // FIXME: change tracking
 561         // Is this correct WRT change tracking?
 562         Encoding const & encoding = *(runparams.encoding);
 563         docstring const latex1 = encoding.latexChar(next);
 564         docstring const latex2 = encoding.latexChar(c);
 565         if (docstring(1, next) == latex1) {
 566                 // the encoding supports the combination
 567                 os << latex2 << latex1;
 568                 return latex1.length() + latex2.length();
 569         } else if (runparams.local_font &&
 570                    runparams.local_font->language()->lang() == "polutonikogreek") {
 571                 // polutonikogreek only works without the brackets
 572                 os << latex1 << latex2;
 573                 return latex1.length() + latex2.length();
 574         } else
 575                 os << latex1 << '{' << latex2 << '}';
 576         return latex1.length() + latex2.length() + 2;
 577 }
 578
 579
 580 bool Paragraph::Private::simpleTeXBlanks(OutputParams const & runparams,
 581                                        odocstream & os, TexRow & texrow,
 582                                        pos_type i,
 583                                        unsigned int & column,
 584                                        Font const & font,
 585                                        Layout const & style)
 586 {
 587         if (style.pass_thru || runparams.pass_thru)
 588                 return false;
 589
 590         if (i + 1 < int(text_.size())) {
 591                 char_type next = text_[i + 1];
 592                 if (Encodings::isCombiningChar(next)) {
 593                         // This space has an accent, so we must always output it.
 594                         column += latexSurrogatePair(os, ' ', next, runparams) - 1;
 595                         return true;
 596                 }
 597         }
 598
 599         if (runparams.linelen > 0
 600             && column > runparams.linelen
 601             && i
 602             && text_[i - 1] != ' '
 603             && (i + 1 < int(text_.size()))
 604             // same in FreeSpacing mode
 605             && !owner_->isFreeSpacing()
 606             // In typewriter mode, we want to avoid
 607             // ! . ? : at the end of a line
 608             && !(font.fontInfo().family() == TYPEWRITER_FAMILY
 609                  && (text_[i - 1] == '.'
 610                      || text_[i - 1] == '?'
 611                      || text_[i - 1] == ':'
 612                      || text_[i - 1] == '!'))) {
 613                 os << '\n';
 614                 texrow.newline();
 615                 texrow.start(owner_->id(), i + 1);
 616                 column = 0;
 617         } else if (style.free_spacing) {
 618                 os << '~';
 619         } else {
 620                 os << ' ';
 621         }
 622         return false;
 623 }
 624
 625
 626 int Paragraph::Private::writeScriptChars(odocstream & os,
 627                                          docstring const & ltx,
 628                                          Change const & runningChange,
 629                                          Encoding const & encoding,
 630                                          pos_type & i)
 631 {
 632         // FIXME: modifying i here is not very nice...
 633
 634         // We only arrive here when a proper language for character text_[i] has
 635         // not been specified (i.e., it could not be translated in the current
 636         // latex encoding) or its latex translation has been forced, and it
 637         // belongs to a known script.
 638         // Parameter ltx contains the latex translation of text_[i] as specified
 639         // in the unicodesymbols file and is something like "\textXXX{<spec>}".
 640         // The latex macro name "textXXX" specifies the script to which text_[i]
 641         // belongs and we use it in order to check whether characters from the
 642         // same script immediately follow, such that we can collect them in a
 643         // single "\textXXX" macro. So, we have to retain "\textXXX{<spec>"
 644         // for the first char but only "<spec>" for all subsequent chars.
 645         docstring::size_type const brace1 = ltx.find_first_of(from_ascii("{"));
 646         docstring::size_type const brace2 = ltx.find_last_of(from_ascii("}"));
 647         string script = to_ascii(ltx.substr(1, brace1 - 1));
 648         int pos = 0;
 649         int length = brace2;
 650         bool closing_brace = true;
 651         if (script == "textgreek" && encoding.latexName() == "iso-8859-7") {
 652                 // Correct encoding is being used, so we can avoid \textgreek.
 653                 pos = brace1 + 1;
 654                 length -= pos;
 655                 closing_brace = false;
 656         }
 657         os << ltx.substr(pos, length);
 658         int size = text_.size();
 659         while (i + 1 < size) {
 660                 char_type const next = text_[i + 1];
 661                 // Stop here if next character belongs to another script
 662                 // or there is a change in change tracking status.
 663                 if (!Encodings::isKnownScriptChar(next, script) ||
 664                     runningChange != owner_->lookupChange(i + 1))
 665                         break;
 666                 Font prev_font;
 667                 bool found = false;
 668                 FontList::const_iterator cit = fontlist_.begin();
 669                 FontList::const_iterator end = fontlist_.end();
 670                 for (; cit != end; ++cit) {
 671                         if (cit->pos() >= i && !found) {
 672                                 prev_font = cit->font();
 673                                 found = true;
 674                         }
 675                         if (cit->pos() >= i + 1)
 676                                 break;
 677                 }
 678                 // Stop here if there is a font attribute or encoding change.
 679                 if (found && cit != end && prev_font != cit->font())
 680                         break;
 681                 docstring const latex = encoding.latexChar(next);
 682                 docstring::size_type const b1 =
 683                                         latex.find_first_of(from_ascii("{"));
 684                 docstring::size_type const b2 =
 685                                         latex.find_last_of(from_ascii("}"));
 686                 int const len = b2 - b1 - 1;
 687                 os << latex.substr(b1 + 1, len);
 688                 length += len;
 689                 ++i;
 690         }
 691         if (closing_brace) {
 692                 os << '}';
 693                 ++length;
 694         }
 695         return length;
 696 }
 697
 698
 699 bool Paragraph::Private::isTextAt(string const & str, pos_type pos) const
 700 {
 701         pos_type const len = str.length();
 702
 703         // is the paragraph large enough?
 704         if (pos + len > int(text_.size()))
 705                 return false;
 706
 707         // does the wanted text start at point?
 708         for (string::size_type i = 0; i < str.length(); ++i) {
 709                 // Caution: direct comparison of characters works only
 710                 // because str is pure ASCII.
 711                 if (str[i] != text_[pos + i])
 712                         return false;
 713         }
 714
 715         return fontlist_.hasChangeInRange(pos, len);
 716 }
 717
 718
 719 void Paragraph::Private::latexInset(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                 texrow.newlines(tmp);
 850                 texrow.start(owner_->id(), i + 1);
 851                 column = 0;
 852         } else {
 853                 column += os.tellp() - len;
 854         }
 855
 856         if (owner_->isDeleted(i))
 857                 --runparams.inDeletedInset;
 858 }
 859
 860
 861 void Paragraph::Private::latexSpecialChar(
 862                                              odocstream & os,
 863                                              OutputParams const & runparams,
 864                                              Font const & running_font,
 865                                              Change const & running_change,
 866                                              Layout const & style,
 867                                              pos_type & i,
 868                                              unsigned int & column)
 869 {
 870         char_type const c = text_[i];
 871
 872         if (style.pass_thru || runparams.pass_thru) {
 873                 if (c != '\0')
 874                         // FIXME UNICODE: This can fail if c cannot
 875                         // be encoded in the current encoding.
 876                         os.put(c);
 877                 return;
 878         }
 879
 880         // If T1 font encoding is used, use the special
 881         // characters it provides.
 882         // NOTE: some languages reset the font encoding
 883         // internally
 884         if (!running_font.language()->internalFontEncoding()
 885             && lyxrc.fontenc == "T1" && latexSpecialT1(c, os, i, column))
 886                 return;
 887
 888         // \tt font needs special treatment
 889         if (running_font.fontInfo().family() == TYPEWRITER_FAMILY
 890                 && latexSpecialTypewriter(c, os, i, column))
 891                 return;
 892
 893         // Otherwise, we use what LaTeX provides us.
 894         switch (c) {
 895         case '\\':
 896                 os << "\\textbackslash{}";
 897                 column += 15;
 898                 break;
 899         case '<':
 900                 os << "\\textless{}";
 901                 column += 10;
 902                 break;
 903         case '>':
 904                 os << "\\textgreater{}";
 905                 column += 13;
 906                 break;
 907         case '|':
 908                 os << "\\textbar{}";
 909                 column += 9;
 910                 break;
 911         case '-':
 912                 os << '-';
 913                 break;
 914         case '\"':
 915                 os << "\\char`\\\"{}";
 916                 column += 9;
 917                 break;
 918
 919         case '$': case '&':
 920         case '%': case '#': case '{':
 921         case '}': case '_':
 922                 os << '\\';
 923                 os.put(c);
 924                 column += 1;
 925                 break;
 926
 927         case '~':
 928                 os << "\\textasciitilde{}";
 929                 column += 16;
 930                 break;
 931
 932         case '^':
 933                 os << "\\textasciicircum{}";
 934                 column += 17;
 935                 break;
 936
 937         case '*':
 938         case '[':
 939         case ']':
 940                 // avoid being mistaken for optional arguments
 941                 os << '{';
 942                 os.put(c);
 943                 os << '}';
 944                 column += 2;
 945                 break;
 946
 947         case ' ':
 948                 // Blanks are printed before font switching.
 949                 // Sure? I am not! (try nice-latex)
 950                 // I am sure it's correct. LyX might be smarter
 951                 // in the future, but for now, nothing wrong is
 952                 // written. (Asger)
 953                 break;
 954
 955         default:
 956                 // LyX, LaTeX etc.
 957                 if (latexSpecialPhrase(os, i, column, runparams))
 958                         return;
 959
 960                 if (c == '\0')
 961                         return;
 962
 963                 Encoding const & encoding = *(runparams.encoding);
 964                 if (i + 1 < int(text_.size())) {
 965                         char_type next = text_[i + 1];
 966                         if (Encodings::isCombiningChar(next)) {
 967                                 column += latexSurrogatePair(os, c, next, runparams) - 1;
 968                                 ++i;
 969                                 break;
 970                         }
 971                 }
 972                 string script;
 973                 docstring const latex = encoding.latexChar(c);
 974                 if (Encodings::isKnownScriptChar(c, script)
 975                     && prefixIs(latex, from_ascii("\\" + script)))
 976                         column += writeScriptChars(os, latex,
 977                                         running_change, encoding, i) - 1;
 978                 else if (latex.length() > 1 && latex[latex.length() - 1] != '}') {
 979                         // Prevent eating of a following
 980                         // space or command corruption by
 981                         // following characters
 982                         column += latex.length() + 1;
 983                         os << latex << "{}";
 984                 } else {
 985                         column += latex.length() - 1;
 986                         os << latex;
 987                 }
 988                 break;
 989         }
 990 }
 991
 992
 993 bool Paragraph::Private::latexSpecialT1(char_type const c, odocstream & os,
 994         pos_type i, unsigned int & column)
 995 {
 996         switch (c) {
 997         case '>':
 998         case '<':
 999                 os.put(c);
1000                 // In T1 encoding, these characters exist
1001                 // but we should avoid ligatures
1002                 if (i + 1 >= int(text_.size()) || text_[i + 1] != c)
1003                         return true;
1004                 os << "\\textcompwordmark{}";
1005                 column += 19;
1006                 return true;
1007         case '|':
1008                 os.put(c);
1009                 return true;
1010         case '\"':
1011                 // soul.sty breaks with \char`\"
1012                 os << "\\textquotedbl{}";
1013                 column += 14;
1014                 return true;
1015         default:
1016                 return false;
1017         }
1018 }
1019
1020
1021 bool Paragraph::Private::latexSpecialTypewriter(char_type const c, odocstream & os,
1022         pos_type i, unsigned int & column)
1023 {
1024         switch (c) {
1025         case '-':
1026                 // within \ttfamily, "--" is merged to "-" (no endash)
1027                 // so we avoid this rather irritating ligature
1028                 if (i + 1 < int(text_.size()) && text_[i + 1] == '-') {
1029                         os << "-{}";
1030                         column += 2;
1031                 } else
1032                         os << '-';
1033                 return true;
1034
1035         // everything else has to be checked separately
1036         // (depending on the encoding)
1037         default:
1038                 return false;
1039         }
1040 }
1041
1042
1043 bool Paragraph::Private::latexSpecialPhrase(odocstream & os, pos_type & i,
1044         unsigned int & column, OutputParams const & runparams)
1045 {
1046         // FIXME: if we have "LaTeX" with a font
1047         // change in the middle (before the 'T', then
1048         // the "TeX" part is still special cased.
1049         // Really we should only operate this on
1050         // "words" for some definition of word
1051
1052         for (size_t pnr = 0; pnr < phrases_nr; ++pnr) {
1053                 if (!isTextAt(special_phrases[pnr].phrase, i))
1054                         continue;
1055                 if (runparams.moving_arg)
1056                         os << "\\protect";
1057                 os << special_phrases[pnr].macro;
1058                 i += special_phrases[pnr].phrase.length() - 1;
1059                 column += special_phrases[pnr].macro.length() - 1;
1060                 return true;
1061         }
1062         return false;
1063 }
1064
1065
1066 void Paragraph::Private::validate(LaTeXFeatures & features) const
1067 {
1068         if (layout_->inpreamble && inset_owner_) {
1069                 Buffer const & buf = inset_owner_->buffer();
1070                 BufferParams const & bp = buf.params();
1071                 Font f;
1072                 TexRow tr;
1073                 odocstringstream ods;
1074                 // we have to provide all the optional arguments here, even though
1075                 // the last one is the only one we care about.
1076                 owner_->latex(bp, f, ods, tr, features.runparams(), 0, -1, true);
1077                 docstring const d = ods.str();
1078                 if (!d.empty()) {
1079                         // this will have "{" at the beginning, but not at the end
1080                         string const content = to_utf8(d);
1081                         string const cmd = layout_->latexname();
1082                         features.addPreambleSnippet("\\" + cmd + content + "}");
1083                 }
1084         }
1085
1086         if (features.runparams().flavor == OutputParams::HTML
1087             && layout_->htmltitle()) {
1088                 features.setHTMLTitle(owner_->asString(AS_STR_INSETS));
1089         }
1090
1091         // check the params.
1092         if (!params_.spacing().isDefault())
1093                 features.require("setspace");
1094
1095         // then the layouts
1096         features.useLayout(layout_->name());
1097
1098         // then the fonts
1099         fontlist_.validate(features);
1100
1101         // then the indentation
1102         if (!params_.leftIndent().zero())
1103                 features.require("ParagraphLeftIndent");
1104
1105         // then the insets
1106         InsetList::const_iterator icit = insetlist_.begin();
1107         InsetList::const_iterator iend = insetlist_.end();
1108         for (; icit != iend; ++icit) {
1109                 if (icit->inset) {
1110                         icit->inset->validate(features);
1111                         if (layout_->needprotect &&
1112                             icit->inset->lyxCode() == FOOT_CODE)
1113                                 features.require("NeedLyXFootnoteCode");
1114                 }
1115         }
1116
1117         // then the contents
1118         for (pos_type i = 0; i < int(text_.size()) ; ++i) {
1119                 for (size_t pnr = 0; pnr < phrases_nr; ++pnr) {
1120                         if (!special_phrases[pnr].builtin
1121                             && isTextAt(special_phrases[pnr].phrase, i)) {
1122                                 features.require(special_phrases[pnr].phrase);
1123                                 break;
1124                         }
1125                 }
1126                 Encodings::validate(text_[i], features);
1127         }
1128 }
1129
1130 /////////////////////////////////////////////////////////////////////
1131 //
1132 // Paragraph
1133 //
1134 /////////////////////////////////////////////////////////////////////
1135
1136 namespace {
1137         Layout const emptyParagraphLayout;
1138 }
1139
1140 Paragraph::Paragraph()
1141         : d(new Paragraph::Private(this, emptyParagraphLayout))
1142 {
1143         itemdepth = 0;
1144         d->params_.clear();
1145 }
1146
1147
1148 Paragraph::Paragraph(Paragraph const & par)
1149         : itemdepth(par.itemdepth),
1150         d(new Paragraph::Private(*par.d, this))
1151 {
1152         registerWords();
1153 }
1154
1155
1156 Paragraph::Paragraph(Paragraph const & par, pos_type beg, pos_type end)
1157         : itemdepth(par.itemdepth),
1158         d(new Paragraph::Private(*par.d, this, beg, end))
1159 {
1160         registerWords();
1161 }
1162
1163
1164 Paragraph & Paragraph::operator=(Paragraph const & par)
1165 {
1166         // needed as we will destroy the private part before copying it
1167         if (&par != this) {
1168                 itemdepth = par.itemdepth;
1169
1170                 deregisterWords();
1171                 delete d;
1172                 d = new Private(*par.d, this);
1173                 registerWords();
1174         }
1175         return *this;
1176 }
1177
1178
1179 Paragraph::~Paragraph()
1180 {
1181         deregisterWords();
1182         delete d;
1183 }
1184
1185
1186 namespace {
1187
1188 // this shall be called just before every "os << ..." action.
1189 void flushString(ostream & os, docstring & s)
1190 {
1191         os << to_utf8(s);
1192         s.erase();
1193 }
1194
1195 }
1196
1197
1198 void Paragraph::write(ostream & os, BufferParams const & bparams,
1199         depth_type & dth) const
1200 {
1201         // The beginning or end of a deeper (i.e. nested) area?
1202         if (dth != d->params_.depth()) {
1203                 if (d->params_.depth() > dth) {
1204                         while (d->params_.depth() > dth) {
1205                                 os << "\n\\begin_deeper";
1206                                 ++dth;
1207                         }
1208                 } else {
1209                         while (d->params_.depth() < dth) {
1210                                 os << "\n\\end_deeper";
1211                                 --dth;
1212                         }
1213                 }
1214         }
1215
1216         // First write the layout
1217         os << "\n\\begin_layout " << to_utf8(d->layout_->name()) << '\n';
1218
1219         d->params_.write(os);
1220
1221         Font font1(inherit_font, bparams.language);
1222
1223         Change running_change = Change(Change::UNCHANGED);
1224
1225         // this string is used as a buffer to avoid repetitive calls
1226         // to to_utf8(), which turn out to be expensive (JMarc)
1227         docstring write_buffer;
1228
1229         int column = 0;
1230         for (pos_type i = 0; i <= size(); ++i) {
1231
1232                 Change const change = lookupChange(i);
1233                 if (change != running_change)
1234                         flushString(os, write_buffer);
1235                 Changes::lyxMarkChange(os, bparams, column, running_change, change);
1236                 running_change = change;
1237
1238                 if (i == size())
1239                         break;
1240
1241                 // Write font changes (ignore spelling markers)
1242                 Font font2 = getFontSettings(bparams, i);
1243                 font2.setMisspelled(false);
1244                 if (font2 != font1) {
1245                         flushString(os, write_buffer);
1246                         font2.lyxWriteChanges(font1, os);
1247                         column = 0;
1248                         font1 = font2;
1249                 }
1250
1251                 char_type const c = d->text_[i];
1252                 switch (c) {
1253                 case META_INSET:
1254                         if (Inset const * inset = getInset(i)) {
1255                                 flushString(os, write_buffer);
1256                                 if (inset->directWrite()) {
1257                                         // international char, let it write
1258                                         // code directly so it's shorter in
1259                                         // the file
1260                                         inset->write(os);
1261                                 } else {
1262                                         if (i)
1263                                                 os << '\n';
1264                                         os << "\\begin_inset ";
1265                                         inset->write(os);
1266                                         os << "\n\\end_inset\n\n";
1267                                         column = 0;
1268                                 }
1269                         }
1270                         break;
1271                 case '\\':
1272                         flushString(os, write_buffer);
1273                         os << "\n\\backslash\n";
1274                         column = 0;
1275                         break;
1276                 case '.':
1277                         flushString(os, write_buffer);
1278                         if (i + 1 < size() && d->text_[i + 1] == ' ') {
1279                                 os << ".\n";
1280                                 column = 0;
1281                         } else
1282                                 os << '.';
1283                         break;
1284                 default:
1285                         if ((column > 70 && c == ' ')
1286                             || column > 79) {
1287                                 flushString(os, write_buffer);
1288                                 os << '\n';
1289                                 column = 0;
1290                         }
1291                         // this check is to amend a bug. LyX sometimes
1292                         // inserts '\0' this could cause problems.
1293                         if (c != '\0')
1294                                 write_buffer.push_back(c);
1295                         else
1296                                 LYXERR0("NUL char in structure.");
1297                         ++column;
1298                         break;
1299                 }
1300         }
1301
1302         flushString(os, write_buffer);
1303         os << "\n\\end_layout\n";
1304 }
1305
1306
1307 void Paragraph::validate(LaTeXFeatures & features) const
1308 {
1309         d->validate(features);
1310 }
1311
1312
1313 void Paragraph::insert(pos_type start, docstring const & str,
1314                        Font const & font, Change const & change)
1315 {
1316         for (size_t i = 0, n = str.size(); i != n ; ++i)
1317                 insertChar(start + i, str[i], font, change);
1318 }
1319
1320
1321 void Paragraph::appendChar(char_type c, Font const & font,
1322                 Change const & change)
1323 {
1324         // track change
1325         d->changes_.insert(change, d->text_.size());
1326         // when appending characters, no need to update tables
1327         d->text_.push_back(c);
1328         setFont(d->text_.size() - 1, font);
1329 }
1330
1331
1332 void Paragraph::appendString(docstring const & s, Font const & font,
1333                 Change const & change)
1334 {
1335         pos_type end = s.size();
1336         size_t oldsize = d->text_.size();
1337         size_t newsize = oldsize + end;
1338         size_t capacity = d->text_.capacity();
1339         if (newsize >= capacity)
1340                 d->text_.reserve(max(capacity + 100, newsize));
1341
1342         // when appending characters, no need to update tables
1343         d->text_.append(s);
1344
1345         // FIXME: Optimize this!
1346         for (size_t i = oldsize; i != newsize; ++i) {
1347                 // track change
1348                 d->changes_.insert(change, i);
1349         }
1350         d->fontlist_.set(oldsize, font);
1351         d->fontlist_.set(newsize - 1, font);
1352 }
1353
1354
1355 void Paragraph::insertChar(pos_type pos, char_type c,
1356                            bool trackChanges)
1357 {
1358         d->insertChar(pos, c, Change(trackChanges ?
1359                            Change::INSERTED : Change::UNCHANGED));
1360 }
1361
1362
1363 void Paragraph::insertChar(pos_type pos, char_type c,
1364                            Font const & font, bool trackChanges)
1365 {
1366         d->insertChar(pos, c, Change(trackChanges ?
1367                            Change::INSERTED : Change::UNCHANGED));
1368         setFont(pos, font);
1369 }
1370
1371
1372 void Paragraph::insertChar(pos_type pos, char_type c,
1373                            Font const & font, Change const & change)
1374 {
1375         d->insertChar(pos, c, change);
1376         setFont(pos, font);
1377 }
1378
1379
1380 bool Paragraph::insertInset(pos_type pos, Inset * inset,
1381                             Font const & font, Change const & change)
1382 {
1383         bool const success = insertInset(pos, inset, change);
1384         // Set the font/language of the inset...
1385         setFont(pos, font);
1386         return success;
1387 }
1388
1389
1390 void Paragraph::resetFonts(Font const & font)
1391 {
1392         d->fontlist_.clear();
1393         d->fontlist_.set(0, font);
1394         d->fontlist_.set(d->text_.size() - 1, font);
1395 }
1396
1397 // Gets uninstantiated font setting at position.
1398 Font const & Paragraph::getFontSettings(BufferParams const & bparams,
1399                                          pos_type pos) const
1400 {
1401         if (pos > size()) {
1402                 LYXERR0("pos: " << pos << " size: " << size());
1403                 LASSERT(pos <= size(), /**/);
1404         }
1405
1406         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
1407         if (cit != d->fontlist_.end())
1408                 return cit->font();
1409
1410         if (pos == size() && !empty())
1411                 return getFontSettings(bparams, pos - 1);
1412
1413         // Optimisation: avoid a full font instantiation if there is no
1414         // language change from previous call.
1415         static Font previous_font;
1416         static Language const * previous_lang = 0;
1417         Language const * lang = getParLanguage(bparams);
1418         if (lang != previous_lang) {
1419                 previous_lang = lang;
1420                 previous_font = Font(inherit_font, lang);
1421         }
1422         return previous_font;
1423 }
1424
1425
1426 FontSpan Paragraph::fontSpan(pos_type pos) const
1427 {
1428         LASSERT(pos <= size(), /**/);
1429         pos_type start = 0;
1430
1431         FontList::const_iterator cit = d->fontlist_.begin();
1432         FontList::const_iterator end = d->fontlist_.end();
1433         for (; cit != end; ++cit) {
1434                 if (cit->pos() >= pos) {
1435                         if (pos >= beginOfBody())
1436                                 return FontSpan(max(start, beginOfBody()),
1437                                                 cit->pos());
1438                         else
1439                                 return FontSpan(start,
1440                                                 min(beginOfBody() - 1,
1441                                                          cit->pos()));
1442                 }
1443                 start = cit->pos() + 1;
1444         }
1445
1446         // This should not happen, but if so, we take no chances.
1447         // LYXERR0("Paragraph::getEndPosOfFontSpan: This should not happen!");
1448         return FontSpan(pos, pos);
1449 }
1450
1451
1452 // Gets uninstantiated font setting at position 0
1453 Font const & Paragraph::getFirstFontSettings(BufferParams const & bparams) const
1454 {
1455         if (!empty() && !d->fontlist_.empty())
1456                 return d->fontlist_.begin()->font();
1457
1458         // Optimisation: avoid a full font instantiation if there is no
1459         // language change from previous call.
1460         static Font previous_font;
1461         static Language const * previous_lang = 0;
1462         if (bparams.language != previous_lang) {
1463                 previous_lang = bparams.language;
1464                 previous_font = Font(inherit_font, bparams.language);
1465         }
1466
1467         return previous_font;
1468 }
1469
1470
1471 // Gets the fully instantiated font at a given position in a paragraph
1472 // This is basically the same function as Text::GetFont() in text2.cpp.
1473 // The difference is that this one is used for generating the LaTeX file,
1474 // and thus cosmetic "improvements" are disallowed: This has to deliver
1475 // the true picture of the buffer. (Asger)
1476 Font const Paragraph::getFont(BufferParams const & bparams, pos_type pos,
1477                                  Font const & outerfont) const
1478 {
1479         LASSERT(pos >= 0, /**/);
1480
1481         Font font = getFontSettings(bparams, pos);
1482
1483         pos_type const body_pos = beginOfBody();
1484         FontInfo & fi = font.fontInfo();
1485         if (pos < body_pos)
1486                 fi.realize(d->layout_->labelfont);
1487         else
1488                 fi.realize(d->layout_->font);
1489
1490         fi.realize(outerfont.fontInfo());
1491         fi.realize(bparams.getFont().fontInfo());
1492
1493         return font;
1494 }
1495
1496
1497 Font const Paragraph::getLabelFont
1498         (BufferParams const & bparams, Font const & outerfont) const
1499 {
1500         FontInfo tmpfont = d->layout_->labelfont;
1501         tmpfont.realize(outerfont.fontInfo());
1502         tmpfont.realize(bparams.getFont().fontInfo());
1503         return Font(tmpfont, getParLanguage(bparams));
1504 }
1505
1506
1507 Font const Paragraph::getLayoutFont
1508         (BufferParams const & bparams, Font const & outerfont) const
1509 {
1510         FontInfo tmpfont = d->layout_->font;
1511         tmpfont.realize(outerfont.fontInfo());
1512         tmpfont.realize(bparams.getFont().fontInfo());
1513         return Font(tmpfont, getParLanguage(bparams));
1514 }
1515
1516
1517 /// Returns the height of the highest font in range
1518 FontSize Paragraph::highestFontInRange
1519         (pos_type startpos, pos_type endpos, FontSize def_size) const
1520 {
1521         return d->fontlist_.highestInRange(startpos, endpos, def_size);
1522 }
1523
1524
1525 char_type Paragraph::getUChar(BufferParams const & bparams, pos_type pos) const
1526 {
1527         char_type c = d->text_[pos];
1528         if (!lyxrc.rtl_support)
1529                 return c;
1530
1531         char_type uc = c;
1532         switch (c) {
1533         case '(':
1534                 uc = ')';
1535                 break;
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         }
1558         if (uc != c && getFontSettings(bparams, pos).isRightToLeft())
1559                 return uc;
1560         return c;
1561 }
1562
1563
1564 void Paragraph::setFont(pos_type pos, Font const & font)
1565 {
1566         LASSERT(pos <= size(), /**/);
1567
1568         // First, reduce font against layout/label font
1569         // Update: The setCharFont() routine in text2.cpp already
1570         // reduces font, so we don't need to do that here. (Asger)
1571
1572         d->fontlist_.set(pos, font);
1573 }
1574
1575
1576 void Paragraph::makeSameLayout(Paragraph const & par)
1577 {
1578         d->layout_ = par.d->layout_;
1579         d->params_ = par.d->params_;
1580 }
1581
1582
1583 bool Paragraph::stripLeadingSpaces(bool trackChanges)
1584 {
1585         if (isFreeSpacing())
1586                 return false;
1587
1588         int pos = 0;
1589         int count = 0;
1590
1591         while (pos < size() && (isNewline(pos) || isLineSeparator(pos))) {
1592                 if (eraseChar(pos, trackChanges))
1593                         ++count;
1594                 else
1595                         ++pos;
1596         }
1597
1598         return count > 0 || pos > 0;
1599 }
1600
1601
1602 bool Paragraph::hasSameLayout(Paragraph const & par) const
1603 {
1604         return par.d->layout_ == d->layout_
1605                 && d->params_.sameLayout(par.d->params_);
1606 }
1607
1608
1609 depth_type Paragraph::getDepth() const
1610 {
1611         return d->params_.depth();
1612 }
1613
1614
1615 depth_type Paragraph::getMaxDepthAfter() const
1616 {
1617         if (d->layout_->isEnvironment())
1618                 return d->params_.depth() + 1;
1619         else
1620                 return d->params_.depth();
1621 }
1622
1623
1624 char Paragraph::getAlign() const
1625 {
1626         if (d->params_.align() == LYX_ALIGN_LAYOUT)
1627                 return d->layout_->align;
1628         else
1629                 return d->params_.align();
1630 }
1631
1632
1633 docstring const & Paragraph::labelString() const
1634 {
1635         return d->params_.labelString();
1636 }
1637
1638
1639 // the next two functions are for the manual labels
1640 docstring const Paragraph::getLabelWidthString() const
1641 {
1642         if (d->layout_->margintype == MARGIN_MANUAL
1643             || d->layout_->latextype == LATEX_BIB_ENVIRONMENT)
1644                 return d->params_.labelWidthString();
1645         else
1646                 return _("Senseless with this layout!");
1647 }
1648
1649
1650 void Paragraph::setLabelWidthString(docstring const & s)
1651 {
1652         d->params_.labelWidthString(s);
1653 }
1654
1655
1656 docstring Paragraph::expandLabel(Layout const & layout,
1657                 BufferParams const & bparams) const
1658 {
1659         return expandParagraphLabel(layout, bparams, true);
1660 }
1661
1662
1663 docstring Paragraph::expandDocBookLabel(Layout const & layout,
1664                 BufferParams const & bparams) const
1665 {
1666         return expandParagraphLabel(layout, bparams, false);
1667 }
1668
1669
1670 docstring Paragraph::expandParagraphLabel(Layout const & layout,
1671                 BufferParams const & bparams, bool process_appendix) const
1672 {
1673         DocumentClass const & tclass = bparams.documentClass();
1674         string const & lang = getParLanguage(bparams)->code();
1675         bool const in_appendix = process_appendix && d->params_.appendix();
1676         docstring fmt = translateIfPossible(layout.labelstring(in_appendix), lang);
1677
1678         if (fmt.empty() && layout.labeltype == LABEL_COUNTER
1679             && !layout.counter.empty())
1680                 return tclass.counters().theCounter(layout.counter, lang);
1681
1682         // handle 'inherited level parts' in 'fmt',
1683         // i.e. the stuff between '@' in   '@Section@.\arabic{subsection}'
1684         size_t const i = fmt.find('@', 0);
1685         if (i != docstring::npos) {
1686                 size_t const j = fmt.find('@', i + 1);
1687                 if (j != docstring::npos) {
1688                         docstring parent(fmt, i + 1, j - i - 1);
1689                         docstring label = from_ascii("??");
1690                         if (tclass.hasLayout(parent))
1691                                 docstring label = expandParagraphLabel(tclass[parent], bparams,
1692                                                       process_appendix);
1693                         fmt = docstring(fmt, 0, i) + label
1694                                 + docstring(fmt, j + 1, docstring::npos);
1695                 }
1696         }
1697
1698         return tclass.counters().counterLabel(fmt, lang);
1699 }
1700
1701
1702 void Paragraph::applyLayout(Layout const & new_layout)
1703 {
1704         d->layout_ = &new_layout;
1705         LyXAlignment const oldAlign = d->params_.align();
1706
1707         if (!(oldAlign & d->layout_->alignpossible)) {
1708                 frontend::Alert::warning(_("Alignment not permitted"),
1709                         _("The new layout does not permit the alignment previously used.\nSetting to default."));
1710                 d->params_.align(LYX_ALIGN_LAYOUT);
1711         }
1712 }
1713
1714
1715 pos_type Paragraph::beginOfBody() const
1716 {
1717         return d->begin_of_body_;
1718 }
1719
1720
1721 void Paragraph::setBeginOfBody()
1722 {
1723         if (d->layout_->labeltype != LABEL_MANUAL) {
1724                 d->begin_of_body_ = 0;
1725                 return;
1726         }
1727
1728         // Unroll the first two cycles of the loop
1729         // and remember the previous character to
1730         // remove unnecessary getChar() calls
1731         pos_type i = 0;
1732         pos_type end = size();
1733         if (i < end && !isNewline(i)) {
1734                 ++i;
1735                 char_type previous_char = 0;
1736                 char_type temp = 0;
1737                 if (i < end) {
1738                         previous_char = d->text_[i];
1739                         if (!isNewline(i)) {
1740                                 ++i;
1741                                 while (i < end && previous_char != ' ') {
1742                                         temp = d->text_[i];
1743                                         if (isNewline(i))
1744                                                 break;
1745                                         ++i;
1746                                         previous_char = temp;
1747                                 }
1748                         }
1749                 }
1750         }
1751
1752         d->begin_of_body_ = i;
1753 }
1754
1755
1756 bool Paragraph::allowParagraphCustomization() const
1757 {
1758         return inInset().allowParagraphCustomization();
1759 }
1760
1761
1762 bool Paragraph::usePlainLayout() const
1763 {
1764         return inInset().usePlainLayout();
1765 }
1766
1767
1768 namespace {
1769
1770 // paragraphs inside floats need different alignment tags to avoid
1771 // unwanted space
1772
1773 bool noTrivlistCentering(InsetCode code)
1774 {
1775         return code == FLOAT_CODE
1776                || code == WRAP_CODE
1777                || code == CELL_CODE;
1778 }
1779
1780
1781 string correction(string const & orig)
1782 {
1783         if (orig == "flushleft")
1784                 return "raggedright";
1785         if (orig == "flushright")
1786                 return "raggedleft";
1787         if (orig == "center")
1788                 return "centering";
1789         return orig;
1790 }
1791
1792
1793 string const corrected_env(string const & suffix, string const & env,
1794         InsetCode code, bool const lastpar)
1795 {
1796         string output = suffix + "{";
1797         if (noTrivlistCentering(code)) {
1798                 if (lastpar) {
1799                         // the last paragraph in non-trivlist-aligned
1800                         // context is special (to avoid unwanted whitespace)
1801                         if (suffix == "\\begin")
1802                                 return "\\" + correction(env) + "{}";
1803                         return string();
1804                 }
1805                 output += correction(env);
1806         } else
1807                 output += env;
1808         output += "}";
1809         if (suffix == "\\begin")
1810                 output += "\n";
1811         return output;
1812 }
1813
1814
1815 void adjust_row_column(string const & str, TexRow & texrow, int & column)
1816 {
1817         if (!contains(str, "\n"))
1818                 column += str.size();
1819         else {
1820                 string tmp;
1821                 texrow.newline();
1822                 column = rsplit(str, tmp, '\n').size();
1823         }
1824 }
1825
1826 } // namespace anon
1827
1828
1829 int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
1830                                  odocstream & os, TexRow & texrow,
1831                                  OutputParams const & runparams) const
1832 {
1833         int column = 0;
1834
1835         if (params_.noindent()) {
1836                 os << "\\noindent ";
1837                 column += 10;
1838         }
1839
1840         LyXAlignment const curAlign = params_.align();
1841
1842         if (curAlign == layout_->align)
1843                 return column;
1844
1845         switch (curAlign) {
1846         case LYX_ALIGN_NONE:
1847         case LYX_ALIGN_BLOCK:
1848         case LYX_ALIGN_LAYOUT:
1849         case LYX_ALIGN_SPECIAL:
1850         case LYX_ALIGN_DECIMAL:
1851                 break;
1852         case LYX_ALIGN_LEFT:
1853         case LYX_ALIGN_RIGHT:
1854         case LYX_ALIGN_CENTER:
1855                 if (runparams.moving_arg) {
1856                         os << "\\protect";
1857                         column += 8;
1858                 }
1859                 break;
1860         }
1861
1862         string const begin_tag = "\\begin";
1863         InsetCode code = ownerCode();
1864         bool const lastpar = runparams.isLastPar;
1865
1866         switch (curAlign) {
1867         case LYX_ALIGN_NONE:
1868         case LYX_ALIGN_BLOCK:
1869         case LYX_ALIGN_LAYOUT:
1870         case LYX_ALIGN_SPECIAL:
1871         case LYX_ALIGN_DECIMAL:
1872                 break;
1873         case LYX_ALIGN_LEFT: {
1874                 string output;
1875                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
1876                         output = corrected_env(begin_tag, "flushleft", code, lastpar);
1877                 else
1878                         output = corrected_env(begin_tag, "flushright", code, lastpar);
1879                 os << from_ascii(output);
1880                 adjust_row_column(output, texrow, column);
1881                 break;
1882         } case LYX_ALIGN_RIGHT: {
1883                 string output;
1884                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
1885                         output = corrected_env(begin_tag, "flushright", code, lastpar);
1886                 else
1887                         output = corrected_env(begin_tag, "flushleft", code, lastpar);
1888                 os << from_ascii(output);
1889                 adjust_row_column(output, texrow, column);
1890                 break;
1891         } case LYX_ALIGN_CENTER: {
1892                 string output;
1893                 output = corrected_env(begin_tag, "center", code, lastpar);
1894                 os << from_ascii(output);
1895                 adjust_row_column(output, texrow, column);
1896                 break;
1897         }
1898         }
1899
1900         return column;
1901 }
1902
1903
1904 int Paragraph::Private::endTeXParParams(BufferParams const & bparams,
1905                                odocstream & os, TexRow & texrow,
1906                                OutputParams const & runparams) const
1907 {
1908         int column = 0;
1909
1910         LyXAlignment const curAlign = params_.align();
1911
1912         if (curAlign == layout_->align)
1913                 return column;
1914
1915         switch (curAlign) {
1916         case LYX_ALIGN_NONE:
1917         case LYX_ALIGN_BLOCK:
1918         case LYX_ALIGN_LAYOUT:
1919         case LYX_ALIGN_SPECIAL:
1920         case LYX_ALIGN_DECIMAL:
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         case LYX_ALIGN_DECIMAL:
1942                 break;
1943         case LYX_ALIGN_LEFT: {
1944                 string output;
1945                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
1946                         output = corrected_env(end_tag, "flushleft", code, lastpar);
1947                 else
1948                         output = corrected_env(end_tag, "flushright", code, lastpar);
1949                 os << from_ascii(output);
1950                 adjust_row_column(output, texrow, column);
1951                 break;
1952         } case LYX_ALIGN_RIGHT: {
1953                 string output;
1954                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
1955                         output = corrected_env(end_tag, "flushright", code, lastpar);
1956                 else
1957                         output = corrected_env(end_tag, "flushleft", code, lastpar);
1958                 os << from_ascii(output);
1959                 adjust_row_column(output, texrow, column);
1960                 break;
1961         } case LYX_ALIGN_CENTER: {
1962                 string output;
1963                 output = corrected_env(end_tag, "center", code, lastpar);
1964                 os << from_ascii(output);
1965                 adjust_row_column(output, texrow, column);
1966                 break;
1967         }
1968         }
1969
1970         return column;
1971 }
1972
1973
1974 // This one spits out the text of the paragraph
1975 void Paragraph::latex(BufferParams const & bparams,
1976         Font const & outerfont,
1977         odocstream & os, TexRow & texrow,
1978         OutputParams const & runparams,
1979         int start_pos, int end_pos, bool force) const
1980 {
1981         LYXERR(Debug::LATEX, "Paragraph::latex...     " << this);
1982
1983         // FIXME This check should not be needed. Perhaps issue an
1984         // error if it triggers.
1985         Layout const & style = inInset().forcePlainLayout() ?
1986                 bparams.documentClass().plainLayout() : *d->layout_;
1987
1988         if (!force && style.inpreamble)
1989                 return;
1990
1991         bool const allowcust = allowParagraphCustomization();
1992
1993         // Current base font for all inherited font changes, without any
1994         // change caused by an individual character, except for the language:
1995         // It is set to the language of the first character.
1996         // As long as we are in the label, this font is the base font of the
1997         // label. Before the first body character it is set to the base font
1998         // of the body.
1999         Font basefont;
2000
2001         // Maybe we have to create a optional argument.
2002         pos_type body_pos = beginOfBody();
2003         unsigned int column = 0;
2004
2005         if (body_pos > 0) {
2006                 // the optional argument is kept in curly brackets in
2007                 // case it contains a ']'
2008                 os << "[{";
2009                 column += 2;
2010                 basefont = getLabelFont(bparams, outerfont);
2011         } else {
2012                 basefont = getLayoutFont(bparams, outerfont);
2013         }
2014
2015         // Which font is currently active?
2016         Font running_font(basefont);
2017         // Do we have an open font change?
2018         bool open_font = false;
2019
2020         Change runningChange = Change(Change::UNCHANGED);
2021
2022         Encoding const * const prev_encoding = runparams.encoding;
2023
2024         texrow.start(id(), 0);
2025
2026         // if the paragraph is empty, the loop will not be entered at all
2027         if (empty()) {
2028                 if (style.isCommand()) {
2029                         os << '{';
2030                         ++column;
2031                 }
2032                 if (allowcust)
2033                         column += d->startTeXParParams(bparams, os, texrow,
2034                                                     runparams);
2035         }
2036
2037         for (pos_type i = 0; i < size(); ++i) {
2038                 // First char in paragraph or after label?
2039                 if (i == body_pos) {
2040                         if (body_pos > 0) {
2041                                 if (open_font) {
2042                                         column += running_font.latexWriteEndChanges(
2043                                                 os, bparams, runparams,
2044                                                 basefont, basefont);
2045                                         open_font = false;
2046                                 }
2047                                 basefont = getLayoutFont(bparams, outerfont);
2048                                 running_font = basefont;
2049
2050                                 column += Changes::latexMarkChange(os, bparams,
2051                                                 runningChange, Change(Change::UNCHANGED),
2052                                                 runparams);
2053                                 runningChange = Change(Change::UNCHANGED);
2054
2055                                 os << "}] ";
2056                                 column +=3;
2057                         }
2058                         if (style.isCommand()) {
2059                                 os << '{';
2060                                 ++column;
2061                         }
2062
2063                         if (allowcust)
2064                                 column += d->startTeXParParams(bparams, os,
2065                                                             texrow,
2066                                                             runparams);
2067                 }
2068
2069                 Change const & change = runparams.inDeletedInset ? runparams.changeOfDeletedInset
2070                                                                  : lookupChange(i);
2071
2072                 if (bparams.outputChanges && runningChange != change) {
2073                         if (open_font) {
2074                                 column += running_font.latexWriteEndChanges(
2075                                                 os, bparams, runparams, basefont, basefont);
2076                                 open_font = false;
2077                         }
2078                         basefont = getLayoutFont(bparams, outerfont);
2079                         running_font = basefont;
2080
2081                         column += Changes::latexMarkChange(os, bparams, runningChange,
2082                                                            change, runparams);
2083                         runningChange = change;
2084                 }
2085
2086                 // do not output text which is marked deleted
2087                 // if change tracking output is disabled
2088                 if (!bparams.outputChanges && change.deleted()) {
2089                         continue;
2090                 }
2091
2092                 ++column;
2093
2094                 // Fully instantiated font
2095                 Font const font = getFont(bparams, i, outerfont);
2096
2097                 Font const last_font = running_font;
2098
2099                 // Do we need to close the previous font?
2100                 if (open_font &&
2101                     (font != running_font ||
2102                      font.language() != running_font.language()))
2103                 {
2104                         column += running_font.latexWriteEndChanges(
2105                                         os, bparams, runparams, basefont,
2106                                         (i == body_pos-1) ? basefont : font);
2107                         running_font = basefont;
2108                         open_font = false;
2109                 }
2110
2111                 // close babel's font environment before opening CJK.
2112                 if (!running_font.language()->babel().empty() &&
2113                     font.language()->encoding()->package() == Encoding::CJK) {
2114                                 string end_tag = subst(lyxrc.language_command_end,
2115                                                         "$$lang",
2116                                                         running_font.language()->babel());
2117                                 os << from_ascii(end_tag);
2118                                 column += end_tag.length();
2119                 }
2120
2121                 // Switch file encoding if necessary (and allowed)
2122                 if (!runparams.pass_thru && !style.pass_thru &&
2123                     runparams.encoding->package() != Encoding::none &&
2124                     font.language()->encoding()->package() != Encoding::none) {
2125                         pair<bool, int> const enc_switch = switchEncoding(os, bparams,
2126                                         runparams, *(font.language()->encoding()));
2127                         if (enc_switch.first) {
2128                                 column += enc_switch.second;
2129                                 runparams.encoding = font.language()->encoding();
2130                         }
2131                 }
2132
2133                 char_type const c = d->text_[i];
2134
2135                 // Do we need to change font?
2136                 if ((font != running_font ||
2137                      font.language() != running_font.language()) &&
2138                         i != body_pos - 1)
2139                 {
2140                         odocstringstream ods;
2141                         column += font.latexWriteStartChanges(ods, bparams,
2142                                                               runparams, basefont,
2143                                                               last_font);
2144                         running_font = font;
2145                         open_font = true;
2146                         docstring fontchange = ods.str();
2147                         // check whether the fontchange ends with a \\textcolor
2148                         // modifier and the text starts with a space (bug 4473)
2149                         docstring const last_modifier = rsplit(fontchange, '\\');
2150                         if (prefixIs(last_modifier, from_ascii("textcolor")) && c == ' ')
2151                                 os << fontchange << from_ascii("{}");
2152                         // check if the fontchange ends with a trailing blank
2153                         // (like "\small " (see bug 3382)
2154                         else if (suffixIs(fontchange, ' ') && c == ' ')
2155                                 os << fontchange.substr(0, fontchange.size() - 1)
2156                                    << from_ascii("{}");
2157                         else
2158                                 os << fontchange;
2159                 }
2160
2161                 // FIXME: think about end_pos implementation...
2162                 if (c == ' ' && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2163                         // FIXME: integrate this case in latexSpecialChar
2164                         // Do not print the separation of the optional argument
2165                         // if style.pass_thru is false. This works because
2166                         // latexSpecialChar ignores spaces if
2167                         // style.pass_thru is false.
2168                         if (i != body_pos - 1) {
2169                                 if (d->simpleTeXBlanks(
2170                                                 runparams, os, texrow,
2171                                                 i, column, font, style)) {
2172                                         // A surrogate pair was output. We
2173                                         // must not call latexSpecialChar
2174                                         // in this iteration, since it would output
2175                                         // the combining character again.
2176                                         ++i;
2177                                         continue;
2178                                 }
2179                         }
2180                 }
2181
2182                 OutputParams rp = runparams;
2183                 rp.free_spacing = style.free_spacing;
2184                 rp.local_font = &font;
2185                 rp.intitle = style.intitle;
2186
2187                 // Two major modes:  LaTeX or plain
2188                 // Handle here those cases common to both modes
2189                 // and then split to handle the two modes separately.
2190                 if (c == META_INSET) {
2191                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2192                                 d->latexInset(bparams, os,
2193                                                 texrow, rp, running_font,
2194                                                 basefont, outerfont, open_font,
2195                                                 runningChange, style, i, column);
2196                         }
2197                 } else {
2198                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2199                                 try {
2200                                         d->latexSpecialChar(os, rp, running_font, runningChange,
2201                                                 style, i, column);
2202                                 } catch (EncodingException & e) {
2203                                 if (runparams.dryrun) {
2204                                         os << "<" << _("LyX Warning: ")
2205                                            << _("uncodable character") << " '";
2206                                         os.put(c);
2207                                         os << "'>";
2208                                 } else {
2209                                         // add location information and throw again.
2210                                         e.par_id = id();
2211                                         e.pos = i;
2212                                         throw(e);
2213                                 }
2214                         }
2215                 }
2216                 }
2217
2218                 // Set the encoding to that returned from latexSpecialChar (see
2219                 // comment for encoding member in OutputParams.h)
2220                 runparams.encoding = rp.encoding;
2221         }
2222
2223         // If we have an open font definition, we have to close it
2224         if (open_font) {
2225 #ifdef FIXED_LANGUAGE_END_DETECTION
2226                 if (next_) {
2227                         running_font.latexWriteEndChanges(os, bparams,
2228                                         runparams, 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         }
2250
2251         if (allowcust && d->endTeXParParams(bparams, os, texrow, runparams)
2252             && runparams.encoding != prev_encoding) {
2253                 runparams.encoding = prev_encoding;
2254                 if (!bparams.useXetex)
2255                         os << setEncoding(prev_encoding->iconvName());
2256         }
2257
2258         LYXERR(Debug::LATEX, "Paragraph::latex... done " << this);
2259 }
2260
2261
2262 bool Paragraph::emptyTag() const
2263 {
2264         for (pos_type i = 0; i < size(); ++i) {
2265                 if (Inset const * inset = getInset(i)) {
2266                         InsetCode lyx_code = inset->lyxCode();
2267                         // FIXME testing like that is wrong. What is
2268                         // the intent?
2269                         if (lyx_code != TOC_CODE &&
2270                             lyx_code != INCLUDE_CODE &&
2271                             lyx_code != GRAPHICS_CODE &&
2272                             lyx_code != ERT_CODE &&
2273                             lyx_code != LISTINGS_CODE &&
2274                             lyx_code != FLOAT_CODE &&
2275                             lyx_code != TABULAR_CODE) {
2276                                 return false;
2277                         }
2278                 } else {
2279                         char_type c = d->text_[i];
2280                         if (c != ' ' && c != '\t')
2281                                 return false;
2282                 }
2283         }
2284         return true;
2285 }
2286
2287
2288 string Paragraph::getID(Buffer const & buf, OutputParams const & runparams)
2289         const
2290 {
2291         for (pos_type i = 0; i < size(); ++i) {
2292                 if (Inset const * inset = getInset(i)) {
2293                         InsetCode lyx_code = inset->lyxCode();
2294                         if (lyx_code == LABEL_CODE) {
2295                                 InsetLabel const * const il = static_cast<InsetLabel const *>(inset);
2296                                 docstring const & id = il->getParam("name");
2297                                 return "id='" + to_utf8(sgml::cleanID(buf, runparams, id)) + "'";
2298                         }
2299                 }
2300         }
2301         return string();
2302 }
2303
2304
2305 pos_type Paragraph::firstWordDocBook(odocstream & os, OutputParams const & runparams)
2306         const
2307 {
2308         pos_type i;
2309         for (i = 0; i < size(); ++i) {
2310                 if (Inset const * inset = getInset(i)) {
2311                         inset->docbook(os, runparams);
2312                 } else {
2313                         char_type c = d->text_[i];
2314                         if (c == ' ')
2315                                 break;
2316                         os << sgml::escapeChar(c);
2317                 }
2318         }
2319         return i;
2320 }
2321
2322
2323 pos_type Paragraph::firstWordLyXHTML(XHTMLStream & xs, OutputParams const & runparams)
2324         const
2325 {
2326         pos_type i;
2327         for (i = 0; i < size(); ++i) {
2328                 if (Inset const * inset = getInset(i)) {
2329                         inset->xhtml(xs, runparams);
2330                 } else {
2331                         char_type c = d->text_[i];
2332                         if (c == ' ')
2333                                 break;
2334                         xs << c;
2335                 }
2336         }
2337         return i;
2338 }
2339
2340
2341 bool Paragraph::Private::onlyText(Buffer const & buf, Font const & outerfont, pos_type initial) const
2342 {
2343         Font font_old;
2344         pos_type size = text_.size();
2345         for (pos_type i = initial; i < size; ++i) {
2346                 Font font = owner_->getFont(buf.params(), i, outerfont);
2347                 if (text_[i] == META_INSET)
2348                         return false;
2349                 if (i != initial && font != font_old)
2350                         return false;
2351                 font_old = font;
2352         }
2353
2354         return true;
2355 }
2356
2357
2358 void Paragraph::simpleDocBookOnePar(Buffer const & buf,
2359                                     odocstream & os,
2360                                     OutputParams const & runparams,
2361                                     Font const & outerfont,
2362                                     pos_type initial) const
2363 {
2364         bool emph_flag = false;
2365
2366         Layout const & style = *d->layout_;
2367         FontInfo font_old =
2368                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2369
2370         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2371                 os << "]]>";
2372
2373         // parsing main loop
2374         for (pos_type i = initial; i < size(); ++i) {
2375                 Font font = getFont(buf.params(), i, outerfont);
2376
2377                 // handle <emphasis> tag
2378                 if (font_old.emph() != font.fontInfo().emph()) {
2379                         if (font.fontInfo().emph() == FONT_ON) {
2380                                 os << "<emphasis>";
2381                                 emph_flag = true;
2382                         } else if (i != initial) {
2383                                 os << "</emphasis>";
2384                                 emph_flag = false;
2385                         }
2386                 }
2387
2388                 if (Inset const * inset = getInset(i)) {
2389                         inset->docbook(os, runparams);
2390                 } else {
2391                         char_type c = d->text_[i];
2392
2393                         if (style.pass_thru)
2394                                 os.put(c);
2395                         else
2396                                 os << sgml::escapeChar(c);
2397                 }
2398                 font_old = font.fontInfo();
2399         }
2400
2401         if (emph_flag) {
2402                 os << "</emphasis>";
2403         }
2404
2405         if (style.free_spacing)
2406                 os << '\n';
2407         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2408                 os << "<![CDATA[";
2409 }
2410
2411
2412 docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
2413                                     XHTMLStream & xs,
2414                                     OutputParams const & runparams,
2415                                     Font const & outerfont,
2416                                     pos_type initial) const
2417 {
2418         docstring retval;
2419
2420         bool emph_flag = false;
2421         bool bold_flag = false;
2422         string closing_tag;
2423
2424         Layout const & style = *d->layout_;
2425
2426         if (!runparams.for_toc && runparams.html_make_pars) {
2427                 // generate a magic label for this paragraph
2428                 string const attr = "id='" + magicLabel() + "'";
2429                 xs << html::CompTag("a", attr);
2430         }
2431
2432         FontInfo font_old =
2433                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2434
2435         // parsing main loop
2436         for (pos_type i = initial; i < size(); ++i) {
2437                 // let's not show deleted material in the output
2438                 if (isDeleted(i))
2439                         continue;
2440
2441                 Font font = getFont(buf.params(), i, outerfont);
2442
2443                 // emphasis
2444                 if (font_old.emph() != font.fontInfo().emph()) {
2445                         if (font.fontInfo().emph() == FONT_ON) {
2446                                 xs << html::StartTag("em");
2447                                 emph_flag = true;
2448                         } else if (emph_flag && i != initial) {
2449                                 xs << html::EndTag("em");
2450                                 emph_flag = false;
2451                         }
2452                 }
2453                 // bold
2454                 if (font_old.series() != font.fontInfo().series()) {
2455                         if (font.fontInfo().series() == BOLD_SERIES) {
2456                                 xs << html::StartTag("strong");
2457                                 bold_flag = true;
2458                         } else if (bold_flag && i != initial) {
2459                                 xs << html::EndTag("strong");
2460                                 bold_flag = false;
2461                         }
2462                 }
2463                 // FIXME XHTML
2464                 // Other such tags? What about the other text ranges?
2465
2466                 Inset const * inset = getInset(i);
2467                 if (inset) {
2468                         InsetCommand const * ic = inset->asInsetCommand();
2469                         InsetLayout const & il = inset->getLayout();
2470                         InsetMath const * im = inset->asInsetMath();
2471                         if (!runparams.for_toc
2472                             || im || il.isInToc() || (ic && ic->isInToc())) {
2473                                 OutputParams np = runparams;
2474                                 if (!il.htmlisblock())
2475                                         np.html_in_par = true;
2476                                 retval += inset->xhtml(xs, np);
2477                         }
2478                 } else {
2479                         char_type c = d->text_[i];
2480
2481                         if (style.pass_thru)
2482                                 xs << c;
2483                         else if (c == '-') {
2484                                 docstring str;
2485                                 int j = i + 1;
2486                                 if (j < size() && d->text_[j] == '-') {
2487                                         j += 1;
2488                                         if (j < size() && d->text_[j] == '-') {
2489                                                 str += from_ascii("&mdash;");
2490                                                 i += 2;
2491                                         } else {
2492                                                 str += from_ascii("&ndash;");
2493                                                 i += 1;
2494                                         }
2495                                 }
2496                                 else
2497                                         str += c;
2498                                 // We don't want to escape the entities. Note that
2499                                 // it is safe to do this, since str can otherwise
2500                                 // only be "-". E.g., it can't be "<".
2501                                 xs << XHTMLStream::NextRaw() << str;
2502                         } else
2503                                 xs << c;
2504                 }
2505                 font_old = font.fontInfo();
2506         }
2507
2508         xs.closeFontTags();
2509         return retval;
2510 }
2511
2512
2513 bool Paragraph::isHfill(pos_type pos) const
2514 {
2515         Inset const * inset = getInset(pos);
2516         return inset && (inset->lyxCode() == SPACE_CODE &&
2517                          inset->isStretchableSpace());
2518 }
2519
2520
2521 bool Paragraph::isNewline(pos_type pos) const
2522 {
2523         Inset const * inset = getInset(pos);
2524         return inset && inset->lyxCode() == NEWLINE_CODE;
2525 }
2526
2527
2528 bool Paragraph::isLineSeparator(pos_type pos) const
2529 {
2530         char_type const c = d->text_[pos];
2531         if (isLineSeparatorChar(c))
2532                 return true;
2533         Inset const * inset = getInset(pos);
2534         return inset && inset->isLineSeparator();
2535 }
2536
2537
2538 bool Paragraph::isWordSeparator(pos_type pos) const
2539 {
2540         if (Inset const * inset = getInset(pos))
2541                 return !inset->isLetter();
2542         char_type const c = d->text_[pos];
2543         // We want to pass the ' and escape chars to the spellchecker
2544         static docstring const quote = from_utf8(lyxrc.spellchecker_esc_chars + '\'');
2545         return (!isLetterChar(c) && !isDigit(c) && !contains(quote, c))
2546                 || pos == size();
2547 }
2548
2549
2550 bool Paragraph::isChar(pos_type pos) const
2551 {
2552         if (Inset const * inset = getInset(pos))
2553                 return inset->isChar();
2554         char_type const c = d->text_[pos];
2555         return !isLetterChar(c) && !isDigit(c) && !lyx::isSpace(c);
2556 }
2557
2558
2559 bool Paragraph::isSpace(pos_type pos) const
2560 {
2561         if (Inset const * inset = getInset(pos))
2562                 return inset->isSpace();
2563         char_type const c = d->text_[pos];
2564         return lyx::isSpace(c);
2565 }
2566
2567
2568 Language const *
2569 Paragraph::getParLanguage(BufferParams const & bparams) const
2570 {
2571         if (!empty())
2572                 return getFirstFontSettings(bparams).language();
2573         // FIXME: we should check the prev par as well (Lgb)
2574         return bparams.language;
2575 }
2576
2577
2578 bool Paragraph::isRTL(BufferParams const & bparams) const
2579 {
2580         return lyxrc.rtl_support
2581                 && getParLanguage(bparams)->rightToLeft()
2582                 && !inInset().getLayout().forceLTR();
2583 }
2584
2585
2586 void Paragraph::changeLanguage(BufferParams const & bparams,
2587                                Language const * from, Language const * to)
2588 {
2589         // change language including dummy font change at the end
2590         for (pos_type i = 0; i <= size(); ++i) {
2591                 Font font = getFontSettings(bparams, i);
2592                 if (font.language() == from) {
2593                         font.setLanguage(to);
2594                         setFont(i, font);
2595                 }
2596         }
2597 }
2598
2599
2600 bool Paragraph::isMultiLingual(BufferParams const & bparams) const
2601 {
2602         Language const * doc_language = bparams.language;
2603         FontList::const_iterator cit = d->fontlist_.begin();
2604         FontList::const_iterator end = d->fontlist_.end();
2605
2606         for (; cit != end; ++cit)
2607                 if (cit->font().language() != ignore_language &&
2608                     cit->font().language() != latex_language &&
2609                     cit->font().language() != doc_language)
2610                         return true;
2611         return false;
2612 }
2613
2614
2615 void Paragraph::getLanguages(std::set<Language const *> & languages) const
2616 {
2617         FontList::const_iterator cit = d->fontlist_.begin();
2618         FontList::const_iterator end = d->fontlist_.end();
2619
2620         for (; cit != end; ++cit) {
2621                 Language const * lang = cit->font().language();
2622                 if (lang != ignore_language &&
2623                     lang != latex_language)
2624                         languages.insert(lang);
2625         }
2626 }
2627
2628
2629 docstring Paragraph::asString(int options) const
2630 {
2631         return asString(0, size(), options);
2632 }
2633
2634
2635 docstring Paragraph::asString(pos_type beg, pos_type end, int options) const
2636 {
2637         odocstringstream os;
2638
2639         if (beg == 0
2640             && options & AS_STR_LABEL
2641             && !d->params_.labelString().empty())
2642                 os << d->params_.labelString() << ' ';
2643
2644         for (pos_type i = beg; i < end; ++i) {
2645                 char_type const c = d->text_[i];
2646                 if (isPrintable(c) || c == '\t'
2647                     || (c == '\n' && (options & AS_STR_NEWLINES)))
2648                         os.put(c);
2649                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
2650                         getInset(i)->tocString(os);
2651                         if (getInset(i)->asInsetMath())
2652                                 os << " ";
2653                 }
2654         }
2655
2656         return os.str();
2657 }
2658
2659
2660 docstring Paragraph::stringify(pos_type beg, pos_type end, int options, OutputParams & runparams) const
2661 {
2662         odocstringstream os;
2663
2664         if (beg == 0
2665                 && options & AS_STR_LABEL
2666                 && !d->params_.labelString().empty())
2667                 os << d->params_.labelString() << ' ';
2668
2669         for (pos_type i = beg; i < end; ++i) {
2670                 char_type const c = d->text_[i];
2671                 if (isPrintable(c) || c == '\t'
2672                     || (c == '\n' && (options & AS_STR_NEWLINES)))
2673                         os.put(c);
2674                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
2675                         getInset(i)->plaintext(os, runparams);
2676                 }
2677         }
2678
2679         return os.str();
2680 }
2681
2682
2683 void Paragraph::setInsetOwner(Inset const * inset)
2684 {
2685         d->inset_owner_ = inset;
2686 }
2687
2688
2689 int Paragraph::id() const
2690 {
2691         return d->id_;
2692 }
2693
2694
2695 void Paragraph::setId(int id)
2696 {
2697         d->id_ = id;
2698 }
2699
2700
2701 Layout const & Paragraph::layout() const
2702 {
2703         return *d->layout_;
2704 }
2705
2706
2707 void Paragraph::setLayout(Layout const & layout)
2708 {
2709         d->layout_ = &layout;
2710 }
2711
2712
2713 void Paragraph::setDefaultLayout(DocumentClass const & tc)
2714 {
2715         setLayout(tc.defaultLayout());
2716 }
2717
2718
2719 void Paragraph::setPlainLayout(DocumentClass const & tc)
2720 {
2721         setLayout(tc.plainLayout());
2722 }
2723
2724
2725 void Paragraph::setPlainOrDefaultLayout(DocumentClass const & tclass)
2726 {
2727         if (usePlainLayout())
2728                 setPlainLayout(tclass);
2729         else
2730                 setDefaultLayout(tclass);
2731 }
2732
2733
2734 Inset const & Paragraph::inInset() const
2735 {
2736         LASSERT(d->inset_owner_, throw ExceptionMessage(BufferException,
2737                 _("Memory problem"), _("Paragraph not properly initialized")));
2738         return *d->inset_owner_;
2739 }
2740
2741
2742 ParagraphParameters & Paragraph::params()
2743 {
2744         return d->params_;
2745 }
2746
2747
2748 ParagraphParameters const & Paragraph::params() const
2749 {
2750         return d->params_;
2751 }
2752
2753
2754 bool Paragraph::isFreeSpacing() const
2755 {
2756         if (d->layout_->free_spacing)
2757                 return true;
2758         return d->inset_owner_ && d->inset_owner_->isFreeSpacing();
2759 }
2760
2761
2762 bool Paragraph::allowEmpty() const
2763 {
2764         if (d->layout_->keepempty)
2765                 return true;
2766         return d->inset_owner_ && d->inset_owner_->allowEmpty();
2767 }
2768
2769
2770 char_type Paragraph::transformChar(char_type c, pos_type pos) const
2771 {
2772         if (!Encodings::isArabicChar(c))
2773                 return c;
2774
2775         char_type prev_char = ' ';
2776         char_type next_char = ' ';
2777
2778         for (pos_type i = pos - 1; i >= 0; --i) {
2779                 char_type const par_char = d->text_[i];
2780                 if (!Encodings::isArabicComposeChar(par_char)) {
2781                         prev_char = par_char;
2782                         break;
2783                 }
2784         }
2785
2786         for (pos_type i = pos + 1, end = size(); i < end; ++i) {
2787                 char_type const par_char = d->text_[i];
2788                 if (!Encodings::isArabicComposeChar(par_char)) {
2789                         next_char = par_char;
2790                         break;
2791                 }
2792         }
2793
2794         if (Encodings::isArabicChar(next_char)) {
2795                 if (Encodings::isArabicChar(prev_char) &&
2796                         !Encodings::isArabicSpecialChar(prev_char))
2797                         return Encodings::transformChar(c, Encodings::FORM_MEDIAL);
2798                 else
2799                         return Encodings::transformChar(c, Encodings::FORM_INITIAL);
2800         } else {
2801                 if (Encodings::isArabicChar(prev_char) &&
2802                         !Encodings::isArabicSpecialChar(prev_char))
2803                         return Encodings::transformChar(c, Encodings::FORM_FINAL);
2804                 else
2805                         return Encodings::transformChar(c, Encodings::FORM_ISOLATED);
2806         }
2807 }
2808
2809
2810 int Paragraph::checkBiblio(Buffer const & buffer)
2811 {
2812         // FIXME From JS:
2813         // This is getting more and more a mess. ...We really should clean
2814         // up this bibitem issue for 1.6. See also bug 2743.
2815
2816         // Add bibitem insets if necessary
2817         if (d->layout_->labeltype != LABEL_BIBLIO)
2818                 return 0;
2819
2820         bool hasbibitem = !d->insetlist_.empty()
2821                 // Insist on it being in pos 0
2822                 && d->text_[0] == META_INSET
2823                 && d->insetlist_.begin()->inset->lyxCode() == BIBITEM_CODE;
2824
2825         bool track_changes = buffer.params().trackChanges;
2826
2827         docstring oldkey;
2828         docstring oldlabel;
2829
2830         // remove a bibitem in pos != 0
2831         // restore it later in pos 0 if necessary
2832         // (e.g. if a user inserts contents _before_ the item)
2833         // we're assuming there's only one of these, which there
2834         // should be.
2835         int erasedInsetPosition = -1;
2836         InsetList::iterator it = d->insetlist_.begin();
2837         InsetList::iterator end = d->insetlist_.end();
2838         for (; it != end; ++it)
2839                 if (it->inset->lyxCode() == BIBITEM_CODE
2840                     && it->pos > 0) {
2841                         InsetBibitem * olditem = static_cast<InsetBibitem *>(it->inset);
2842                         oldkey = olditem->getParam("key");
2843                         oldlabel = olditem->getParam("label");
2844                         erasedInsetPosition = it->pos;
2845                         eraseChar(erasedInsetPosition, track_changes);
2846                         break;
2847         }
2848
2849         // There was an InsetBibitem at the beginning, and we didn't
2850         // have to erase one.
2851         if (hasbibitem && erasedInsetPosition < 0)
2852                         return 0;
2853
2854         // There was an InsetBibitem at the beginning and we did have to
2855         // erase one. So we give its properties to the beginning inset.
2856         if (hasbibitem) {
2857                 InsetBibitem * inset =
2858                         static_cast<InsetBibitem *>(d->insetlist_.begin()->inset);
2859                 if (!oldkey.empty())
2860                         inset->setParam("key", oldkey);
2861                 inset->setParam("label", oldlabel);
2862                 return -erasedInsetPosition;
2863         }
2864
2865         // There was no inset at the beginning, so we need to create one with
2866         // the key and label of the one we erased.
2867         InsetBibitem * inset =
2868                 new InsetBibitem(const_cast<Buffer *>(&buffer), InsetCommandParams(BIBITEM_CODE));
2869         // restore values of previously deleted item in this par.
2870         if (!oldkey.empty())
2871                 inset->setParam("key", oldkey);
2872         inset->setParam("label", oldlabel);
2873         insertInset(0, static_cast<Inset *>(inset),
2874                     Change(track_changes ? Change::INSERTED : Change::UNCHANGED));
2875
2876         return 1;
2877 }
2878
2879
2880 void Paragraph::checkAuthors(AuthorList const & authorList)
2881 {
2882         d->changes_.checkAuthors(authorList);
2883 }
2884
2885
2886 bool Paragraph::isChanged(pos_type pos) const
2887 {
2888         return lookupChange(pos).changed();
2889 }
2890
2891
2892 bool Paragraph::isInserted(pos_type pos) const
2893 {
2894         return lookupChange(pos).inserted();
2895 }
2896
2897
2898 bool Paragraph::isDeleted(pos_type pos) const
2899 {
2900         return lookupChange(pos).deleted();
2901 }
2902
2903
2904 InsetList const & Paragraph::insetList() const
2905 {
2906         return d->insetlist_;
2907 }
2908
2909
2910 void Paragraph::setBuffer(Buffer & b)
2911 {
2912         d->insetlist_.setBuffer(b);
2913 }
2914
2915
2916 Inset * Paragraph::releaseInset(pos_type pos)
2917 {
2918         Inset * inset = d->insetlist_.release(pos);
2919         /// does not honour change tracking!
2920         eraseChar(pos, false);
2921         return inset;
2922 }
2923
2924
2925 Inset * Paragraph::getInset(pos_type pos)
2926 {
2927         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
2928                  ? d->insetlist_.get(pos) : 0;
2929 }
2930
2931
2932 Inset const * Paragraph::getInset(pos_type pos) const
2933 {
2934         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
2935                  ? d->insetlist_.get(pos) : 0;
2936 }
2937
2938
2939 void Paragraph::changeCase(BufferParams const & bparams, pos_type pos,
2940                 pos_type & right, TextCase action)
2941 {
2942         // process sequences of modified characters; in change
2943         // tracking mode, this approach results in much better
2944         // usability than changing case on a char-by-char basis
2945         docstring changes;
2946
2947         bool const trackChanges = bparams.trackChanges;
2948
2949         bool capitalize = true;
2950
2951         for (; pos < right; ++pos) {
2952                 char_type oldChar = d->text_[pos];
2953                 char_type newChar = oldChar;
2954
2955                 // ignore insets and don't play with deleted text!
2956                 if (oldChar != META_INSET && !isDeleted(pos)) {
2957                         switch (action) {
2958                                 case text_lowercase:
2959                                         newChar = lowercase(oldChar);
2960                                         break;
2961                                 case text_capitalization:
2962                                         if (capitalize) {
2963                                                 newChar = uppercase(oldChar);
2964                                                 capitalize = false;
2965                                         }
2966                                         break;
2967                                 case text_uppercase:
2968                                         newChar = uppercase(oldChar);
2969                                         break;
2970                         }
2971                 }
2972
2973                 if (isWordSeparator(pos) || isDeleted(pos)) {
2974                         // permit capitalization again
2975                         capitalize = true;
2976                 }
2977
2978                 if (oldChar != newChar) {
2979                         changes += newChar;
2980                         if (pos != right - 1)
2981                                 continue;
2982                         // step behind the changing area
2983                         pos++;
2984                 }
2985
2986                 int erasePos = pos - changes.size();
2987                 for (size_t i = 0; i < changes.size(); i++) {
2988                         insertChar(pos, changes[i],
2989                                    getFontSettings(bparams,
2990                                                    erasePos),
2991                                    trackChanges);
2992                         if (!eraseChar(erasePos, trackChanges)) {
2993                                 ++erasePos;
2994                                 ++pos; // advance
2995                                 ++right; // expand selection
2996                         }
2997                 }
2998                 changes.clear();
2999         }
3000 }
3001
3002
3003 bool Paragraph::find(docstring const & str, bool cs, bool mw,
3004                 pos_type pos, bool del) const
3005 {
3006         int const strsize = str.length();
3007         int i = 0;
3008         pos_type const parsize = d->text_.size();
3009         for (i = 0; pos + i < parsize; ++i) {
3010                 if (i >= strsize)
3011                         break;
3012                 if (cs && str[i] != d->text_[pos + i])
3013                         break;
3014                 if (!cs && uppercase(str[i]) != uppercase(d->text_[pos + i]))
3015                         break;
3016                 if (!del && isDeleted(pos + i))
3017                         break;
3018         }
3019
3020         if (i != strsize)
3021                 return false;
3022
3023         // if necessary, check whether string matches word
3024         if (mw) {
3025                 if (pos > 0 && !isWordSeparator(pos - 1))
3026                         return false;
3027                 if (pos + strsize < parsize
3028                         && !isWordSeparator(pos + strsize))
3029                         return false;
3030         }
3031
3032         return true;
3033 }
3034
3035
3036 char_type Paragraph::getChar(pos_type pos) const
3037 {
3038         return d->text_[pos];
3039 }
3040
3041
3042 pos_type Paragraph::size() const
3043 {
3044         return d->text_.size();
3045 }
3046
3047
3048 bool Paragraph::empty() const
3049 {
3050         return d->text_.empty();
3051 }
3052
3053
3054 bool Paragraph::isInset(pos_type pos) const
3055 {
3056         return d->text_[pos] == META_INSET;
3057 }
3058
3059
3060 bool Paragraph::isSeparator(pos_type pos) const
3061 {
3062         //FIXME: Are we sure this can be the only separator?
3063         return d->text_[pos] == ' ';
3064 }
3065
3066
3067 void Paragraph::deregisterWords()
3068 {
3069         Private::LangWordsMap::const_iterator itl = d->words_.begin();
3070         Private::LangWordsMap::const_iterator ite = d->words_.end();
3071         for (; itl != ite; ++itl) {
3072                 WordList * wl = theWordList(itl->first);
3073                 Private::Words::const_iterator it = (itl->second).begin();
3074                 Private::Words::const_iterator et = (itl->second).end();
3075                 for (; it != et; ++it)
3076                         wl->remove(*it);
3077         }
3078         d->words_.clear();
3079 }
3080
3081
3082 void Paragraph::locateWord(pos_type & from, pos_type & to,
3083         word_location const loc) const
3084 {
3085         switch (loc) {
3086         case WHOLE_WORD_STRICT:
3087                 if (from == 0 || from == size()
3088                     || isWordSeparator(from)
3089                     || isWordSeparator(from - 1)) {
3090                         to = from;
3091                         return;
3092                 }
3093                 // no break here, we go to the next
3094
3095         case WHOLE_WORD:
3096                 // If we are already at the beginning of a word, do nothing
3097                 if (!from || isWordSeparator(from - 1))
3098                         break;
3099                 // no break here, we go to the next
3100
3101         case PREVIOUS_WORD:
3102                 // always move the cursor to the beginning of previous word
3103                 while (from && !isWordSeparator(from - 1))
3104                         --from;
3105                 break;
3106         case NEXT_WORD:
3107                 LYXERR0("Paragraph::locateWord: NEXT_WORD not implemented yet");
3108                 break;
3109         case PARTIAL_WORD:
3110                 // no need to move the 'from' cursor
3111                 break;
3112         }
3113         to = from;
3114         while (to < size() && !isWordSeparator(to))
3115                 ++to;
3116 }
3117
3118
3119 void Paragraph::collectWords()
3120 {
3121         // This is the value that needs to be exposed in the preferences
3122         // to resolve bug #6760.
3123         static int minlength = 6;
3124         pos_type n = size();
3125         for (pos_type pos = 0; pos < n; ++pos) {
3126                 if (isWordSeparator(pos))
3127                         continue;
3128                 pos_type from = pos;
3129                 locateWord(from, pos, WHOLE_WORD);
3130                 if (pos - from >= minlength) {
3131                         docstring word = asString(from, pos, AS_STR_NONE);
3132                         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
3133                         if (cit == d->fontlist_.end())
3134                                 return;
3135                         Language const * lang = cit->font().language();
3136                         d->words_[*lang].insert(word);
3137                 }
3138         }
3139 }
3140
3141
3142 void Paragraph::registerWords()
3143 {
3144         Private::LangWordsMap::const_iterator itl = d->words_.begin();
3145         Private::LangWordsMap::const_iterator ite = d->words_.end();
3146         for (; itl != ite; ++itl) {
3147                 WordList * wl = theWordList(itl->first);
3148                 Private::Words::const_iterator it = (itl->second).begin();
3149                 Private::Words::const_iterator et = (itl->second).end();
3150                 for (; it != et; ++it)
3151                         wl->insert(*it);
3152         }
3153 }
3154
3155
3156 void Paragraph::updateWords()
3157 {
3158         deregisterWords();
3159         collectWords();
3160         registerWords();
3161 }
3162
3163
3164 SpellChecker::Result Paragraph::spellCheck(pos_type & from, pos_type & to, WordLangTuple & wl,
3165         docstring_list & suggestions, bool do_suggestion) const
3166 {
3167         SpellChecker * speller = theSpellChecker();
3168         if (!speller)
3169                 return SpellChecker::WORD_OK;
3170
3171         if (!d->layout_->spellcheck || !inInset().allowSpellCheck())
3172                 return SpellChecker::WORD_OK;
3173
3174         locateWord(from, to, WHOLE_WORD);
3175         if (from == to || from >= pos_type(d->text_.size()))
3176                 return SpellChecker::WORD_OK;
3177
3178         docstring word = asString(from, to, AS_STR_INSETS);
3179         // Ignore words with digits
3180         // FIXME: make this customizable
3181         // (note that hunspell ignores words with digits by default)
3182         bool const ignored = hasDigit(word);
3183         Language * lang = const_cast<Language *>(getFontSettings(
3184                     d->inset_owner_->buffer().params(), from).language());
3185         if (lang == d->inset_owner_->buffer().params().language
3186             && !lyxrc.spellchecker_alt_lang.empty()) {
3187                 string lang_code;
3188                 string const lang_variety =
3189                         split(lyxrc.spellchecker_alt_lang, lang_code, '-');
3190                 lang->setCode(lang_code);
3191                 lang->setVariety(lang_variety);
3192         }
3193         wl = WordLangTuple(word, lang);
3194         SpellChecker::Result res = ignored ?
3195                 SpellChecker::WORD_OK : speller->check(wl);
3196
3197         bool const misspelled_ = SpellChecker::misspelled(res) ;
3198
3199         if (lyxrc.spellcheck_continuously)
3200                 d->fontlist_.setMisspelled(from, to, misspelled_);
3201
3202         if (misspelled_ && do_suggestion)
3203                 speller->suggest(wl, suggestions);
3204         else
3205                 suggestions.clear();
3206
3207         return res;
3208 }
3209
3210
3211 bool Paragraph::isMisspelled(pos_type pos) const
3212 {
3213         pos_type from = pos;
3214         pos_type to = pos;
3215         WordLangTuple wl;
3216         docstring_list suggestions;
3217         SpellChecker::Result res = spellCheck(from, to, wl, suggestions, false);
3218         return SpellChecker::misspelled(res) ;
3219 }
3220
3221
3222 string Paragraph::magicLabel() const
3223 {
3224         stringstream ss;
3225         ss << "magicparlabel-" << id();
3226         return ss.str();
3227 }
3228
3229
3230 } // namespace lyx