src/Paragraph.cpp

   1 /**
   2  * \file Paragraph.cpp
   3  * This file is part of LyX, the document processor.
   4  * Licence details can be found in the file COPYING.
   5  *
   6  * \author Asger Alstrup
   7  * \author Lars Gullik Bjønnes
   8  * \author Jean-Marc Lasgouttes
   9  * \author Angus Leeming
  10  * \author John Levon
  11  * \author André Pönitz
  12  * \author Dekel Tsur
  13  * \author Jürgen Vigna
  14  *
  15  * Full author contact details are available in file CREDITS.
  16  */
  17
  18 #include <config.h>
  19
  20 #include "Paragraph.h"
  21
  22 #include "Buffer.h"
  23 #include "BufferParams.h"
  24 #include "Changes.h"
  25 #include "Counters.h"
  26 #include "Encoding.h"
  27 #include "debug.h"
  28 #include "gettext.h"
  29 #include "InsetList.h"
  30 #include "Language.h"
  31 #include "LaTeXFeatures.h"
  32 #include "Color.h"
  33 #include "Layout.h"
  34 #include "Length.h"
  35 #include "Font.h"
  36 #include "FontList.h"
  37 #include "LyXRC.h"
  38 #include "Messages.h"
  39 #include "OutputParams.h"
  40 #include "output_latex.h"
  41 #include "paragraph_funcs.h"
  42 #include "ParagraphParameters.h"
  43 #include "sgml.h"
  44 #include "TexRow.h"
  45 #include "VSpace.h"
  46
  47 #include "frontends/alert.h"
  48 #include "frontends/FontMetrics.h"
  49
  50 #include "insets/InsetBibitem.h"
  51 #include "insets/InsetLabel.h"
  52 #include "insets/InsetOptArg.h"
  53
  54 #include "support/lstrings.h"
  55 #include "support/textutils.h"
  56 #include "support/convert.h"
  57 #include "support/unicode.h"
  58
  59 #include <sstream>
  60 #include <vector>
  61
  62 using std::endl;
  63 using std::string;
  64 using std::ostream;
  65
  66 namespace lyx {
  67
  68 using support::contains;
  69 using support::lowercase;
  70 using support::prefixIs;
  71 using support::suffixIs;
  72 using support::rsplit;
  73 using support::rtrim;
  74 using support::uppercase;
  75
  76 namespace {
  77 /// Inset identifier (above 0x10ffff, for ucs-4)
  78 char_type const META_INSET = 0x200001;
  79 };
  80
  81 /////////////////////////////////////////////////////////////////////
  82 //
  83 // Paragraph::Private
  84 //
  85 /////////////////////////////////////////////////////////////////////
  86
  87 class Paragraph::Private
  88 {
  89 public:
  90         ///
  91         Private(Paragraph * owner);
  92         /// "Copy constructor"
  93         Private(Private const &, Paragraph * owner);
  94
  95         ///
  96         void insertChar(pos_type pos, char_type c, Change const & change);
  97
  98         /// Output the surrogate pair formed by \p c and \p next to \p os.
  99         /// \return the number of characters written.
 100         int latexSurrogatePair(odocstream & os, char_type c, char_type next,
 101                                Encoding const &);
 102
 103         /// Output a space in appropriate formatting (or a surrogate pair
 104         /// if the next character is a combining character).
 105         /// \return whether a surrogate pair was output.
 106         bool simpleTeXBlanks(Encoding const &,
 107                              odocstream &, TexRow & texrow,
 108                              pos_type i,
 109                              unsigned int & column,
 110                              Font const & font,
 111                              Layout const & style);
 112
 113         /// Output consecutive known unicode chars, belonging to the same
 114         /// language as specified by \p preamble, to \p os starting from \p c.
 115         /// \return the number of characters written.
 116         int knownLangChars(odocstream & os, char_type c, string & preamble,
 117                            Change &, Encoding const &, pos_type &);
 118
 119         /// This could go to ParagraphParameters if we want to.
 120         int startTeXParParams(BufferParams const &, odocstream &, TexRow &,
 121                               bool) const;
 122
 123         /// This could go to ParagraphParameters if we want to.
 124         int endTeXParParams(BufferParams const &, odocstream &, TexRow &,
 125                             bool) const;
 126
 127         ///
 128         void latexInset(Buffer const &, BufferParams const &,
 129                                    odocstream &,
 130                                    TexRow & texrow, OutputParams &,
 131                                    Font & running_font,
 132                                    Font & basefont,
 133                                    Font const & outerfont,
 134                                    bool & open_font,
 135                                    Change & running_change,
 136                                    Layout const & style,
 137                                    pos_type & i,
 138                                    unsigned int & column);
 139
 140         ///
 141         void latexSpecialChar(
 142                                    odocstream & os,
 143                                    OutputParams & runparams,
 144                                    Font & running_font,
 145                                    Change & running_change,
 146                                    Layout const & style,
 147                                    pos_type & i,
 148                                    unsigned int & column);
 149
 150         ///
 151         bool latexSpecialT1(
 152                 char_type const c,
 153                 odocstream & os,
 154                 pos_type & i,
 155                 unsigned int & column);
 156         ///
 157         bool latexSpecialTypewriter(
 158                 char_type const c,
 159                 odocstream & os,
 160                 pos_type & i,
 161                 unsigned int & column);
 162         ///
 163         bool latexSpecialPhrase(
 164                 odocstream & os,
 165                 pos_type & i,
 166                 unsigned int & column);
 167
 168         ///
 169         void validate(LaTeXFeatures & features,
 170                       Layout const & layout) const;
 171
 172         ///
 173         pos_type size() const { return owner_->size(); }
 174
 175         /// match a string against a particular point in the paragraph
 176         bool isTextAt(std::string const & str, pos_type pos) const;
 177
 178         /// Which Paragraph owns us?
 179         Paragraph * owner_;
 180
 181         /// In which Inset?
 182         Inset * inset_owner_;
 183
 184         ///
 185         FontList fontlist_;
 186
 187         ///
 188         unsigned int id_;
 189         ///
 190         static unsigned int paragraph_id;
 191         ///
 192         ParagraphParameters params_;
 193
 194         /// for recording and looking up changes
 195         Changes changes_;
 196
 197         ///
 198         InsetList insetlist_;
 199
 200         ///
 201         LayoutPtr layout_;
 202
 203         /// end of label
 204         pos_type begin_of_body_;
 205
 206         typedef std::vector<char_type> TextContainer;
 207         ///
 208         TextContainer text_;
 209 };
 210
 211
 212
 213
 214 using std::endl;
 215 using std::upper_bound;
 216 using std::lower_bound;
 217 using std::string;
 218
 219
 220 // Initialization of the counter for the paragraph id's,
 221 unsigned int Paragraph::Private::paragraph_id = 0;
 222
 223 namespace {
 224
 225 struct special_phrase {
 226         string phrase;
 227         docstring macro;
 228         bool builtin;
 229 };
 230
 231 special_phrase const special_phrases[] = {
 232         { "LyX", from_ascii("\\protect\\LyX{}"), false },
 233         { "TeX", from_ascii("\\protect\\TeX{}"), true },
 234         { "LaTeX2e", from_ascii("\\protect\\LaTeXe{}"), true },
 235         { "LaTeX", from_ascii("\\protect\\LaTeX{}"), true },
 236 };
 237
 238 size_t const phrases_nr = sizeof(special_phrases)/sizeof(special_phrase);
 239
 240 } // namespace anon
 241
 242
 243 Paragraph::Private::Private(Paragraph * owner)
 244         : owner_(owner), inset_owner_(0), begin_of_body_(0)
 245 {
 246         id_ = paragraph_id++;
 247         text_.reserve(100);
 248 }
 249
 250
 251 Paragraph::Private::Private(Private const & p, Paragraph * owner)
 252         : owner_(owner), inset_owner_(p.inset_owner_), fontlist_(p.fontlist_),
 253           params_(p.params_), changes_(p.changes_), insetlist_(p.insetlist_),
 254           layout_(p.layout_), begin_of_body_(p.begin_of_body_), text_(p.text_)
 255 {
 256         id_ = paragraph_id++;
 257 }
 258
 259
 260 bool Paragraph::isChanged(pos_type start, pos_type end) const
 261 {
 262         BOOST_ASSERT(start >= 0 && start <= size());
 263         BOOST_ASSERT(end > start && end <= size() + 1);
 264
 265         return d->changes_.isChanged(start, end);
 266 }
 267
 268
 269 bool Paragraph::isMergedOnEndOfParDeletion(bool trackChanges) const {
 270         // keep the logic here in sync with the logic of eraseChars()
 271
 272         if (!trackChanges) {
 273                 return true;
 274         }
 275
 276         Change change = d->changes_.lookup(size());
 277
 278         return change.type == Change::INSERTED && change.author == 0;
 279 }
 280
 281
 282 void Paragraph::setChange(Change const & change)
 283 {
 284         // beware of the imaginary end-of-par character!
 285         d->changes_.set(change, 0, size() + 1);
 286
 287         /*
 288          * Propagate the change recursively - but not in case of DELETED!
 289          *
 290          * Imagine that your co-author makes changes in an existing inset. He
 291          * sends your document to you and you come to the conclusion that the
 292          * inset should go completely. If you erase it, LyX must not delete all
 293          * text within the inset. Otherwise, the change tracked insertions of
 294          * your co-author get lost and there is no way to restore them later.
 295          *
 296          * Conclusion: An inset's content should remain untouched if you delete it
 297          */
 298
 299         if (change.type != Change::DELETED) {
 300                 for (pos_type pos = 0; pos < size(); ++pos) {
 301                         if (isInset(pos))
 302                                 getInset(pos)->setChange(change);
 303                 }
 304         }
 305 }
 306
 307
 308 void Paragraph::setChange(pos_type pos, Change const & change)
 309 {
 310         BOOST_ASSERT(pos >= 0 && pos <= size());
 311
 312         d->changes_.set(change, pos);
 313
 314         // see comment in setChange(Change const &) above
 315
 316         if (change.type != Change::DELETED &&
 317             pos < size() && isInset(pos)) {
 318                 getInset(pos)->setChange(change);
 319         }
 320 }
 321
 322
 323 Change const & Paragraph::lookupChange(pos_type pos) const
 324 {
 325         BOOST_ASSERT(pos >= 0 && pos <= size());
 326
 327         return d->changes_.lookup(pos);
 328 }
 329
 330
 331 void Paragraph::acceptChanges(BufferParams const & bparams, pos_type start,
 332                 pos_type end)
 333 {
 334         BOOST_ASSERT(start >= 0 && start <= size());
 335         BOOST_ASSERT(end > start && end <= size() + 1);
 336
 337         for (pos_type pos = start; pos < end; ++pos) {
 338                 switch (lookupChange(pos).type) {
 339                         case Change::UNCHANGED:
 340                                 // accept changes in nested inset
 341                                 if (pos < size() && isInset(pos))
 342                                         getInset(pos)->acceptChanges(bparams);
 343
 344                                 break;
 345
 346                         case Change::INSERTED:
 347                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 348                                 // also accept changes in nested inset
 349                                 if (pos < size() && isInset(pos)) {
 350                                         getInset(pos)->acceptChanges(bparams);
 351                                 }
 352                                 break;
 353
 354                         case Change::DELETED:
 355                                 // Suppress access to non-existent
 356                                 // "end-of-paragraph char"
 357                                 if (pos < size()) {
 358                                         eraseChar(pos, false);
 359                                         --end;
 360                                         --pos;
 361                                 }
 362                                 break;
 363                 }
 364
 365         }
 366 }
 367
 368
 369 void Paragraph::rejectChanges(BufferParams const & bparams,
 370                 pos_type start, pos_type end)
 371 {
 372         BOOST_ASSERT(start >= 0 && start <= size());
 373         BOOST_ASSERT(end > start && end <= size() + 1);
 374
 375         for (pos_type pos = start; pos < end; ++pos) {
 376                 switch (lookupChange(pos).type) {
 377                         case Change::UNCHANGED:
 378                                 // reject changes in nested inset
 379                                 if (pos < size() && isInset(pos)) {
 380                                         getInset(pos)->rejectChanges(bparams);
 381                                 }
 382                                 break;
 383
 384                         case Change::INSERTED:
 385                                 // Suppress access to non-existent
 386                                 // "end-of-paragraph char"
 387                                 if (pos < size()) {
 388                                         eraseChar(pos, false);
 389                                         --end;
 390                                         --pos;
 391                                 }
 392                                 break;
 393
 394                         case Change::DELETED:
 395                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 396
 397                                 // Do NOT reject changes within a deleted inset!
 398                                 // There may be insertions of a co-author inside of it!
 399
 400                                 break;
 401                 }
 402         }
 403 }
 404
 405
 406 void Paragraph::Private::insertChar(pos_type pos, char_type c,
 407                 Change const & change)
 408 {
 409         BOOST_ASSERT(pos >= 0 && pos <= size());
 410
 411         // track change
 412         changes_.insert(change, pos);
 413
 414         // This is actually very common when parsing buffers (and
 415         // maybe inserting ascii text)
 416         if (pos == size()) {
 417                 // when appending characters, no need to update tables
 418                 text_.push_back(c);
 419                 return;
 420         }
 421
 422         text_.insert(text_.begin() + pos, c);
 423
 424         // Update the font table.
 425         fontlist_.increasePosAfterPos(pos);
 426
 427         // Update the insets
 428         insetlist_.increasePosAfterPos(pos);
 429 }
 430
 431
 432 void Paragraph::insertInset(pos_type pos, Inset * inset,
 433                                    Change const & change)
 434 {
 435         BOOST_ASSERT(inset);
 436         BOOST_ASSERT(pos >= 0 && pos <= size());
 437
 438         d->insertChar(pos, META_INSET, change);
 439         BOOST_ASSERT(d->text_[pos] == META_INSET);
 440
 441         // Add a new entry in the insetlist_.
 442         d->insetlist_.insert(inset, pos);
 443 }
 444
 445
 446 bool Paragraph::eraseChar(pos_type pos, bool trackChanges)
 447 {
 448         BOOST_ASSERT(pos >= 0 && pos <= size());
 449
 450         // keep the logic here in sync with the logic of isMergedOnEndOfParDeletion()
 451
 452         if (trackChanges) {
 453                 Change change = d->changes_.lookup(pos);
 454
 455                 // set the character to DELETED if
 456                 //  a) it was previously unchanged or
 457                 //  b) it was inserted by a co-author
 458
 459                 if (change.type == Change::UNCHANGED ||
 460                     (change.type == Change::INSERTED && change.author != 0)) {
 461                         setChange(pos, Change(Change::DELETED));
 462                         return false;
 463                 }
 464
 465                 if (change.type == Change::DELETED)
 466                         return false;
 467         }
 468
 469         // Don't physically access the imaginary end-of-paragraph character.
 470         // eraseChar() can only mark it as DELETED. A physical deletion of
 471         // end-of-par must be handled externally.
 472         if (pos == size()) {
 473                 return false;
 474         }
 475
 476         // track change
 477         d->changes_.erase(pos);
 478
 479         // if it is an inset, delete the inset entry
 480         if (d->text_[pos] == META_INSET)
 481                 d->insetlist_.erase(pos);
 482
 483         d->text_.erase(d->text_.begin() + pos);
 484
 485         // Update the fontlist_
 486         d->fontlist_.erase(pos);
 487
 488         // Update the insetlist_
 489         d->insetlist_.decreasePosAfterPos(pos);
 490
 491         return true;
 492 }
 493
 494
 495 int Paragraph::eraseChars(pos_type start, pos_type end, bool trackChanges)
 496 {
 497         BOOST_ASSERT(start >= 0 && start <= size());
 498         BOOST_ASSERT(end >= start && end <= size() + 1);
 499
 500         pos_type i = start;
 501         for (pos_type count = end - start; count; --count) {
 502                 if (!eraseChar(i, trackChanges))
 503                         ++i;
 504         }
 505         return end - i;
 506 }
 507
 508
 509 int Paragraph::Private::latexSurrogatePair(odocstream & os, char_type c,
 510                 char_type next, Encoding const & encoding)
 511 {
 512         // Writing next here may circumvent a possible font change between
 513         // c and next. Since next is only output if it forms a surrogate pair
 514         // with c we can ignore this:
 515         // A font change inside a surrogate pair does not make sense and is
 516         // hopefully impossible to input.
 517         // FIXME: change tracking
 518         // Is this correct WRT change tracking?
 519         docstring const latex1 = encoding.latexChar(next);
 520         docstring const latex2 = encoding.latexChar(c);
 521         os << latex1 << '{' << latex2 << '}';
 522         return latex1.length() + latex2.length() + 2;
 523 }
 524
 525
 526 bool Paragraph::Private::simpleTeXBlanks(Encoding const & encoding,
 527                                        odocstream & os, TexRow & texrow,
 528                                        pos_type i,
 529                                        unsigned int & column,
 530                                        Font const & font,
 531                                        Layout const & style)
 532 {
 533         if (style.pass_thru)
 534                 return false;
 535
 536         if (i + 1 < size()) {
 537                 char_type next = text_[i + 1];
 538                 if (Encodings::isCombiningChar(next)) {
 539                         // This space has an accent, so we must always output it.
 540                         column += latexSurrogatePair(os, ' ', next, encoding) - 1;
 541                         return true;
 542                 }
 543         }
 544
 545         if (lyxrc.plaintext_linelen > 0
 546             && column > lyxrc.plaintext_linelen
 547             && i
 548             && text_[i - 1] != ' '
 549             && (i + 1 < size())
 550             // same in FreeSpacing mode
 551             && !owner_->isFreeSpacing()
 552             // In typewriter mode, we want to avoid
 553             // ! . ? : at the end of a line
 554             && !(font.family() == Font::TYPEWRITER_FAMILY
 555                  && (text_[i - 1] == '.'
 556                      || text_[i - 1] == '?'
 557                      || text_[i - 1] == ':'
 558                      || text_[i - 1] == '!'))) {
 559                 os << '\n';
 560                 texrow.newline();
 561                 texrow.start(owner_->id(), i + 1);
 562                 column = 0;
 563         } else if (style.free_spacing) {
 564                 os << '~';
 565         } else {
 566                 os << ' ';
 567         }
 568         return false;
 569 }
 570
 571
 572 int Paragraph::Private::knownLangChars(odocstream & os,
 573                                      char_type c,
 574                                      string & preamble,
 575                                      Change & runningChange,
 576                                      Encoding const & encoding,
 577                                      pos_type & i)
 578 {
 579         // When the character is marked by the proper language, we simply
 580         // get its code point in some encoding, otherwise we get the
 581         // translation specified in the unicodesymbols file, which is
 582         // something like "\textLANG{<spec>}". So, we have to retain
 583         // "\textLANG{<spec>" for the first char but only "<spec>" for
 584         // all subsequent chars.
 585         docstring const latex1 = rtrim(encoding.latexChar(c), "}");
 586         int length = latex1.length();
 587         os << latex1;
 588         while (i + 1 < size()) {
 589                 char_type next = text_[i + 1];
 590                 // Stop here if next character belongs to another
 591                 // language or there is a change tracking status.
 592                 if (!Encodings::isKnownLangChar(next, preamble) ||
 593                     runningChange != owner_->lookupChange(i + 1))
 594                         break;
 595                 Font prev_font;
 596                 bool found = false;
 597                 FontList::const_iterator cit = fontlist_.begin();
 598                 FontList::const_iterator end = fontlist_.end();
 599                 for (; cit != end; ++cit) {
 600                         if (cit->pos() >= i && !found) {
 601                                 prev_font = cit->font();
 602                                 found = true;
 603                         }
 604                         if (cit->pos() >= i + 1)
 605                                 break;
 606                 }
 607                 // Stop here if there is a font attribute change.
 608                 if (found && cit != end && prev_font != cit->font())
 609                         break;
 610                 docstring const latex = rtrim(encoding.latexChar(next), "}");
 611                 docstring::size_type const j =
 612                                         latex.find_first_of(from_ascii("{"));
 613                 if (j == docstring::npos) {
 614                         os << latex;
 615                         length += latex.length();
 616                 } else {
 617                         os << latex.substr(j + 1);
 618                         length += latex.substr(j + 1).length();
 619                 }
 620                 ++i;
 621         }
 622         // When the proper language is set, we are simply passed a code
 623         // point, so we should not try to close the \textLANG command.
 624         if (prefixIs(latex1, from_ascii("\\" + preamble))) {
 625                 os << '}';
 626                 ++length;
 627         }
 628         return length;
 629 }
 630
 631
 632 bool Paragraph::Private::isTextAt(string const & str, pos_type pos) const
 633 {
 634         pos_type const len = str.length();
 635
 636         // is the paragraph large enough?
 637         if (pos + len > size())
 638                 return false;
 639
 640         // does the wanted text start at point?
 641         for (string::size_type i = 0; i < str.length(); ++i) {
 642                 // Caution: direct comparison of characters works only
 643                 // because str is pure ASCII.
 644                 if (str[i] != text_[pos + i])
 645                         return false;
 646         }
 647
 648         return fontlist_.hasChangeInRange(pos, len);
 649 }
 650
 651
 652 void Paragraph::Private::latexInset(Buffer const & buf,
 653                                              BufferParams const & bparams,
 654                                              odocstream & os,
 655                                              TexRow & texrow,
 656                                              OutputParams & runparams,
 657                                              Font & running_font,
 658                                              Font & basefont,
 659                                              Font const & outerfont,
 660                                              bool & open_font,
 661                                              Change & running_change,
 662                                              Layout const & style,
 663                                              pos_type & i,
 664                                              unsigned int & column)
 665 {
 666         Inset * inset = owner_->getInset(i);
 667         BOOST_ASSERT(inset);
 668
 669         if (style.pass_thru) {
 670                 inset->plaintext(buf, os, runparams);
 671                 return;
 672         }
 673
 674         // FIXME: move this to InsetNewline::latex
 675         if (inset->lyxCode() == NEWLINE_CODE) {
 676                 // newlines are handled differently here than
 677                 // the default in simpleTeXSpecialChars().
 678                 if (!style.newline_allowed) {
 679                         os << '\n';
 680                 } else {
 681                         if (open_font) {
 682                                 column += running_font.latexWriteEndChanges(
 683                                         os, bparams, runparams,
 684                                         basefont, basefont);
 685                                 open_font = false;
 686                         }
 687
 688                         if (running_font.family() == Font::TYPEWRITER_FAMILY)
 689                                 os << '~';
 690
 691                         basefont = owner_->getLayoutFont(bparams, outerfont);
 692                         running_font = basefont;
 693
 694                         if (runparams.moving_arg)
 695                                 os << "\\protect ";
 696
 697                         os << "\\\\\n";
 698                 }
 699                 texrow.newline();
 700                 texrow.start(owner_->id(), i + 1);
 701                 column = 0;
 702         }
 703
 704         if (owner_->lookupChange(i).type == Change::DELETED) {
 705                 if( ++runparams.inDeletedInset == 1)
 706                         runparams.changeOfDeletedInset = owner_->lookupChange(i);
 707         }
 708
 709         if (inset->canTrackChanges()) {
 710                 column += Changes::latexMarkChange(os, bparams, running_change,
 711                         Change(Change::UNCHANGED));
 712                 running_change = Change(Change::UNCHANGED);
 713         }
 714
 715         bool close = false;
 716         odocstream::pos_type const len = os.tellp();
 717
 718         if ((inset->lyxCode() == GRAPHICS_CODE
 719              || inset->lyxCode() == MATH_CODE
 720              || inset->lyxCode() == HYPERLINK_CODE)
 721             && running_font.isRightToLeft()) {
 722                 if (running_font.language()->lang() == "farsi")
 723                         os << "\\beginL{}";
 724                 else
 725                         os << "\\L{";
 726                 close = true;
 727         }
 728
 729         // FIXME: Bug: we can have an empty font change here!
 730         // if there has just been a font change, we are going to close it
 731         // right now, which means stupid latex code like \textsf{}. AFAIK,
 732         // this does not harm dvi output. A minor bug, thus (JMarc)
 733
 734         // Some insets cannot be inside a font change command.
 735         // However, even such insets *can* be placed in \L or \R
 736         // or their equivalents (for RTL language switches), so we don't
 737         // close the language in those cases.
 738         // ArabTeX, though, cannot handle this special behavior, it seems.
 739         bool arabtex = basefont.language()->lang() == "arabic_arabtex"
 740                 || running_font.language()->lang() == "arabic_arabtex";
 741         if (open_font && inset->noFontChange()) {
 742                 bool closeLanguage = arabtex
 743                         || basefont.isRightToLeft() == running_font.isRightToLeft();
 744                 unsigned int count = running_font.latexWriteEndChanges(os,
 745                         bparams, runparams, basefont, basefont, closeLanguage);
 746                 column += count;
 747                 // if any font properties were closed, update the running_font,
 748                 // making sure, however, to leave the language as it was
 749                 if (count > 0) {
 750                         // FIXME: probably a better way to keep track of the old
 751                         // language, than copying the entire font?
 752                         Font const copy_font(running_font);
 753                         basefont = owner_->getLayoutFont(bparams, outerfont);
 754                         running_font = basefont;
 755                         if (!closeLanguage)
 756                                 running_font.setLanguage(copy_font.language());
 757                         // leave font open if language is still open
 758                         open_font = (running_font.language() == basefont.language());
 759                         if (closeLanguage)
 760                                 runparams.local_font = &basefont;
 761                 }
 762         }
 763
 764         int tmp = inset->latex(buf, os, runparams);
 765
 766         if (close) {
 767         if (running_font.language()->lang() == "farsi")
 768                         os << "\\endL{}";
 769                 else
 770                         os << '}';
 771         }
 772
 773         if (tmp) {
 774                 for (int j = 0; j < tmp; ++j)
 775                         texrow.newline();
 776
 777                 texrow.start(owner_->id(), i + 1);
 778                 column = 0;
 779         } else {
 780                 column += os.tellp() - len;
 781         }
 782
 783         if (owner_->lookupChange(i).type == Change::DELETED)
 784                 --runparams.inDeletedInset;
 785 }
 786
 787
 788 void Paragraph::Private::latexSpecialChar(
 789                                              odocstream & os,
 790                                              OutputParams & runparams,
 791                                              Font & running_font,
 792                                              Change & running_change,
 793                                              Layout const & style,
 794                                              pos_type & i,
 795                                              unsigned int & column)
 796 {
 797         char_type const c = text_[i];
 798
 799         if (style.pass_thru) {
 800                 if (c != '\0')
 801                         // FIXME UNICODE: This can fail if c cannot
 802                         // be encoded in the current encoding.
 803                         os.put(c);
 804                 return;
 805         }
 806
 807         if (runparams.verbatim) {
 808                 os.put(c);
 809                 return;
 810         }
 811
 812         if (lyxrc.fontenc == "T1" && latexSpecialT1(c, os, i, column))
 813                 return;
 814
 815         if (running_font.family() == Font::TYPEWRITER_FAMILY
 816                 && latexSpecialTypewriter(c, os, i, column))
 817                 return;
 818
 819         // Otherwise, we use what LaTeX provides us.
 820         switch (c) {
 821         case '\\':
 822                 os << "\\textbackslash{}";
 823                 column += 15;
 824                 break;
 825         case '<':
 826                 os << "\\textless{}";
 827                 column += 10;
 828                 break;
 829         case '>':
 830                 os << "\\textgreater{}";
 831                 column += 13;
 832                 break;
 833         case '|':
 834                 os << "\\textbar{}";
 835                 column += 9;
 836                 break;
 837         case '-':
 838                 os << '-';
 839                 break;
 840         case '\"':
 841                 os << "\\char`\\\"{}";
 842                 column += 9;
 843                 break;
 844
 845         case '$': case '&':
 846         case '%': case '#': case '{':
 847         case '}': case '_':
 848                 os << '\\';
 849                 os.put(c);
 850                 column += 1;
 851                 break;
 852
 853         case '~':
 854                 os << "\\textasciitilde{}";
 855                 column += 16;
 856                 break;
 857
 858         case '^':
 859                 os << "\\textasciicircum{}";
 860                 column += 17;
 861                 break;
 862
 863         case '*': case '[':
 864                 // avoid being mistaken for optional arguments
 865                 os << '{';
 866                 os.put(c);
 867                 os << '}';
 868                 column += 2;
 869                 break;
 870
 871         case ' ':
 872                 // Blanks are printed before font switching.
 873                 // Sure? I am not! (try nice-latex)
 874                 // I am sure it's correct. LyX might be smarter
 875                 // in the future, but for now, nothing wrong is
 876                 // written. (Asger)
 877                 break;
 878
 879         default:
 880
 881                 // LyX, LaTeX etc.
 882                 if (latexSpecialPhrase(os, i, column))
 883                         return;
 884
 885                 if (c == '\0')
 886                         return;
 887
 888                 Encoding const & encoding = *(runparams.encoding);
 889                 if (i + 1 < size()) {
 890                         char_type next = text_[i + 1];
 891                         if (Encodings::isCombiningChar(next)) {
 892                                 column += latexSurrogatePair(os, c, next, encoding) - 1;
 893                                 ++i;
 894                                 break;
 895                         }
 896                 }
 897                 string preamble;
 898                 if (Encodings::isKnownLangChar(c, preamble)) {
 899                         column += knownLangChars(os, c, preamble, running_change,
 900                                 encoding, i) - 1;
 901                         break;
 902                 }
 903                 docstring const latex = encoding.latexChar(c);
 904                 if (latex.length() > 1 && latex[latex.length() - 1] != '}') {
 905                         // Prevent eating of a following
 906                         // space or command corruption by
 907                         // following characters
 908                         column += latex.length() + 1;
 909                         os << latex << "{}";
 910                 } else {
 911                         column += latex.length() - 1;
 912                         os << latex;
 913                 }
 914                 break;
 915         }
 916 }
 917
 918
 919 bool Paragraph::Private::latexSpecialT1(char_type const c, odocstream & os,
 920         pos_type & i, unsigned int & column)
 921 {
 922         switch (c) {
 923         case '>':
 924         case '<':
 925                 os.put(c);
 926                 // In T1 encoding, these characters exist
 927                 // but we should avoid ligatures
 928                 if (i + 1 > size() || text_[i + 1] != c)
 929                         return true;
 930                 os << "\\,{}";
 931                 column += 3;
 932                 // Alternative code:
 933                 //os << "\\textcompwordmark{}";
 934                 //column += 19;
 935                 return true;
 936         case '|':
 937                 os.put(c);
 938                 return true;
 939         default:
 940                 return false;
 941         }
 942 }
 943
 944
 945 bool Paragraph::Private::latexSpecialTypewriter(char_type const c, odocstream & os,
 946         pos_type & i, unsigned int & column)
 947 {
 948         switch (c) {
 949         case '-':
 950                 if (i + 1 < size() && text_[i + 1] == '-') {
 951                         // "--" in Typewriter mode -> "-{}-"
 952                         os << "-{}";
 953                         column += 2;
 954                 } else
 955                         os << '-';
 956                 return true;
 957
 958         // I assume this is hack treating typewriter as verbatim
 959         // FIXME UNICODE: This can fail if c cannot be encoded
 960         // in the current encoding.
 961
 962         case '\0':
 963                 return true;
 964
 965         // Those characters are not directly supported.
 966         case '\\':
 967         case '\"':
 968         case '$': case '&':
 969         case '%': case '#': case '{':
 970         case '}': case '_':
 971         case '~':
 972         case '^':
 973         case '*': case '[':
 974         case ' ':
 975                 return false;
 976
 977         default:
 978                 // With Typewriter font, these characters exist.
 979                 os.put(c);
 980                 return true;
 981         }
 982 }
 983
 984
 985 bool Paragraph::Private::latexSpecialPhrase(odocstream & os, pos_type & i,
 986         unsigned int & column)
 987 {
 988         // FIXME: if we have "LaTeX" with a font
 989         // change in the middle (before the 'T', then
 990         // the "TeX" part is still special cased.
 991         // Really we should only operate this on
 992         // "words" for some definition of word
 993
 994         for (size_t pnr = 0; pnr < phrases_nr; ++pnr) {
 995                 if (!isTextAt(special_phrases[pnr].phrase, i))
 996                         continue;
 997                 os << special_phrases[pnr].macro;
 998                 i += special_phrases[pnr].phrase.length() - 1;
 999                 column += special_phrases[pnr].macro.length() - 1;
1000                 return true;
1001         }
1002         return false;
1003 }
1004
1005
1006 void Paragraph::Private::validate(LaTeXFeatures & features,
1007                                 Layout const & layout) const
1008 {
1009         // check the params.
1010         if (!params_.spacing().isDefault())
1011                 features.require("setspace");
1012
1013         // then the layouts
1014         features.useLayout(layout.name());
1015
1016         // then the fonts
1017         fontlist_.validate(features);
1018
1019         // then the indentation
1020         if (!params_.leftIndent().zero())
1021                 features.require("ParagraphLeftIndent");
1022
1023         // then the insets
1024         InsetList::const_iterator icit = insetlist_.begin();
1025         InsetList::const_iterator iend = insetlist_.end();
1026         for (; icit != iend; ++icit) {
1027                 if (icit->inset) {
1028                         icit->inset->validate(features);
1029                         if (layout.needprotect &&
1030                             icit->inset->lyxCode() == FOOT_CODE)
1031                                 features.require("NeedLyXFootnoteCode");
1032                 }
1033         }
1034
1035         // then the contents
1036         for (pos_type i = 0; i < size() ; ++i) {
1037                 for (size_t pnr = 0; pnr < phrases_nr; ++pnr) {
1038                         if (!special_phrases[pnr].builtin
1039                             && isTextAt(special_phrases[pnr].phrase, i)) {
1040                                 features.require(special_phrases[pnr].phrase);
1041                                 break;
1042                         }
1043                 }
1044                 Encodings::validate(text_[i], features);
1045         }
1046 }
1047
1048 /////////////////////////////////////////////////////////////////////
1049 //
1050 // Paragraph
1051 //
1052 /////////////////////////////////////////////////////////////////////
1053
1054 Paragraph::Paragraph()
1055         : d(new Paragraph::Private(this))
1056 {
1057         itemdepth = 0;
1058         d->params_.clear();
1059 }
1060
1061
1062 Paragraph::Paragraph(Paragraph const & par)
1063         : itemdepth(par.itemdepth),
1064         d(new Paragraph::Private(*par.d, this))
1065 {
1066 }
1067
1068
1069 Paragraph & Paragraph::operator=(Paragraph const & par)
1070 {
1071         // needed as we will destroy the private part before copying it
1072         if (&par != this) {
1073                 itemdepth = par.itemdepth;
1074
1075                 delete d;
1076                 d = new Private(*par.d, this);
1077         }
1078         return *this;
1079 }
1080
1081
1082 Paragraph::~Paragraph()
1083 {
1084         delete d;
1085 }
1086
1087
1088 void Paragraph::write(Buffer const & buf, ostream & os,
1089                           BufferParams const & bparams,
1090                           depth_type & dth) const
1091 {
1092         // The beginning or end of a deeper (i.e. nested) area?
1093         if (dth != params().depth()) {
1094                 if (params().depth() > dth) {
1095                         while (params().depth() > dth) {
1096                                 os << "\n\\begin_deeper";
1097                                 ++dth;
1098                         }
1099                 } else {
1100                         while (params().depth() < dth) {
1101                                 os << "\n\\end_deeper";
1102                                 --dth;
1103                         }
1104                 }
1105         }
1106
1107         // First write the layout
1108         os << "\n\\begin_layout " << to_utf8(layout()->name()) << '\n';
1109
1110         params().write(os);
1111
1112         Font font1(Font::ALL_INHERIT, bparams.language);
1113
1114         Change running_change = Change(Change::UNCHANGED);
1115
1116         int column = 0;
1117         for (pos_type i = 0; i <= size(); ++i) {
1118
1119                 Change change = lookupChange(i);
1120                 Changes::lyxMarkChange(os, column, running_change, change);
1121                 running_change = change;
1122
1123                 if (i == size())
1124                         break;
1125
1126                 // Write font changes
1127                 Font font2 = getFontSettings(bparams, i);
1128                 if (font2 != font1) {
1129                         font2.lyxWriteChanges(font1, os);
1130                         column = 0;
1131                         font1 = font2;
1132                 }
1133
1134                 char_type const c = d->text_[i];
1135                 switch (c) {
1136                 case META_INSET:
1137                 {
1138                         Inset const * inset = getInset(i);
1139                         if (inset)
1140                                 if (inset->directWrite()) {
1141                                         // international char, let it write
1142                                         // code directly so it's shorter in
1143                                         // the file
1144                                         inset->write(buf, os);
1145                                 } else {
1146                                         if (i)
1147                                                 os << '\n';
1148                                         os << "\\begin_inset ";
1149                                         inset->write(buf, os);
1150                                         os << "\n\\end_inset\n\n";
1151                                         column = 0;
1152                                 }
1153                 }
1154                 break;
1155                 case '\\':
1156                         os << "\n\\backslash\n";
1157                         column = 0;
1158                         break;
1159                 case '.':
1160                         if (i + 1 < size() && d->text_[i + 1] == ' ') {
1161                                 os << ".\n";
1162                                 column = 0;
1163                         } else
1164                                 os << '.';
1165                         break;
1166                 default:
1167                         if ((column > 70 && c == ' ')
1168                             || column > 79) {
1169                                 os << '\n';
1170                                 column = 0;
1171                         }
1172                         // this check is to amend a bug. LyX sometimes
1173                         // inserts '\0' this could cause problems.
1174                         if (c != '\0') {
1175                                 std::vector<char> tmp = ucs4_to_utf8(c);
1176                                 tmp.push_back('\0');
1177                                 os << &tmp[0];
1178                         } else
1179                                 lyxerr << "ERROR (Paragraph::writeFile):"
1180                                         " NULL char in structure." << endl;
1181                         ++column;
1182                         break;
1183                 }
1184         }
1185
1186         os << "\n\\end_layout\n";
1187 }
1188
1189
1190 void Paragraph::validate(LaTeXFeatures & features) const
1191 {
1192         d->validate(features, *layout());
1193 }
1194
1195
1196 void Paragraph::insert(pos_type start, docstring const & str,
1197                        Font const & font, Change const & change)
1198 {
1199         for (size_t i = 0, n = str.size(); i != n ; ++i)
1200                 insertChar(start + i, str[i], font, change);
1201 }
1202
1203
1204 void Paragraph::appendChar(char_type c, Font const & font,
1205                 Change const & change)
1206 {
1207         // track change
1208         d->changes_.insert(change, d->text_.size());
1209         // when appending characters, no need to update tables
1210         d->text_.push_back(c);
1211         setFont(d->text_.size() - 1, font);
1212 }
1213
1214
1215 void Paragraph::appendString(docstring const & s, Font const & font,
1216                 Change const & change)
1217 {
1218         size_t end = s.size();
1219         size_t oldsize = d->text_.size();
1220         size_t newsize = oldsize + end;
1221         size_t capacity = d->text_.capacity();
1222         if (newsize >= capacity)
1223                 d->text_.reserve(std::max(capacity + 100, newsize));
1224
1225         // FIXME: Optimize this!
1226         for (pos_type i = 0; i != end; ++i) {
1227                 // track change
1228                 d->changes_.insert(change, i);
1229                 // when appending characters, no need to update tables
1230                 d->text_.push_back(s[i]);
1231         }
1232         d->fontlist_.setRange(oldsize, newsize, font);
1233 }
1234
1235
1236 void Paragraph::insertChar(pos_type pos, char_type c,
1237                            bool trackChanges)
1238 {
1239         d->insertChar(pos, c, Change(trackChanges ?
1240                            Change::INSERTED : Change::UNCHANGED));
1241 }
1242
1243
1244 void Paragraph::insertChar(pos_type pos, char_type c,
1245                            Font const & font, bool trackChanges)
1246 {
1247         d->insertChar(pos, c, Change(trackChanges ?
1248                            Change::INSERTED : Change::UNCHANGED));
1249         setFont(pos, font);
1250 }
1251
1252
1253 void Paragraph::insertChar(pos_type pos, char_type c,
1254                            Font const & font, Change const & change)
1255 {
1256         d->insertChar(pos, c, change);
1257         setFont(pos, font);
1258 }
1259
1260
1261 void Paragraph::insertInset(pos_type pos, Inset * inset,
1262                             Font const & font, Change const & change)
1263 {
1264         insertInset(pos, inset, change);
1265         // Set the font/language of the inset...
1266         setFont(pos, font);
1267 }
1268
1269
1270 bool Paragraph::insetAllowed(InsetCode code)
1271 {
1272         return !d->inset_owner_ || d->inset_owner_->insetAllowed(code);
1273 }
1274
1275
1276 // Gets uninstantiated font setting at position.
1277 Font const Paragraph::getFontSettings(BufferParams const & bparams,
1278                                          pos_type pos) const
1279 {
1280         if (pos > size()) {
1281                 lyxerr << " pos: " << pos << " size: " << size() << endl;
1282                 BOOST_ASSERT(pos <= size());
1283         }
1284
1285         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
1286         if (cit != d->fontlist_.end())
1287                 return cit->font();
1288
1289         if (pos == size() && !empty())
1290                 return getFontSettings(bparams, pos - 1);
1291
1292         return Font(Font::ALL_INHERIT, getParLanguage(bparams));
1293 }
1294
1295
1296 FontSpan Paragraph::fontSpan(pos_type pos) const
1297 {
1298         BOOST_ASSERT(pos <= size());
1299         pos_type start = 0;
1300
1301         FontList::const_iterator cit = d->fontlist_.begin();
1302         FontList::const_iterator end = d->fontlist_.end();
1303         for (; cit != end; ++cit) {
1304                 if (cit->pos() >= pos) {
1305                         if (pos >= beginOfBody())
1306                                 return FontSpan(std::max(start, beginOfBody()),
1307                                                 cit->pos());
1308                         else
1309                                 return FontSpan(start,
1310                                                 std::min(beginOfBody() - 1,
1311                                                          cit->pos()));
1312                 }
1313                 start = cit->pos() + 1;
1314         }
1315
1316         // This should not happen, but if so, we take no chances.
1317         //lyxerr << "Paragraph::getEndPosOfFontSpan: This should not happen!"
1318         //      << endl;
1319         return FontSpan(pos, pos);
1320 }
1321
1322
1323 // Gets uninstantiated font setting at position 0
1324 Font const Paragraph::getFirstFontSettings(BufferParams const & bparams) const
1325 {
1326         if (!empty() && !d->fontlist_.empty())
1327                 return d->fontlist_.begin()->font();
1328
1329         return Font(Font::ALL_INHERIT, bparams.language);
1330 }
1331
1332
1333 // Gets the fully instantiated font at a given position in a paragraph
1334 // This is basically the same function as Text::GetFont() in text2.cpp.
1335 // The difference is that this one is used for generating the LaTeX file,
1336 // and thus cosmetic "improvements" are disallowed: This has to deliver
1337 // the true picture of the buffer. (Asger)
1338 Font const Paragraph::getFont(BufferParams const & bparams, pos_type pos,
1339                                  Font const & outerfont) const
1340 {
1341         BOOST_ASSERT(pos >= 0);
1342
1343         Font font = getFontSettings(bparams, pos);
1344
1345         pos_type const body_pos = beginOfBody();
1346         if (pos < body_pos)
1347                 font.realize(d->layout_->labelfont);
1348         else
1349                 font.realize(d->layout_->font);
1350
1351         font.realize(outerfont);
1352         font.realize(bparams.getFont());
1353
1354         return font;
1355 }
1356
1357
1358 Font const Paragraph::getLabelFont
1359         (BufferParams const & bparams, Font const & outerfont) const
1360 {
1361         Font tmpfont = layout()->labelfont;
1362         tmpfont.setLanguage(getParLanguage(bparams));
1363         tmpfont.realize(outerfont);
1364         tmpfont.realize(bparams.getFont());
1365         return tmpfont;
1366 }
1367
1368
1369 Font const Paragraph::getLayoutFont
1370         (BufferParams const & bparams, Font const & outerfont) const
1371 {
1372         Font tmpfont = layout()->font;
1373         tmpfont.setLanguage(getParLanguage(bparams));
1374         tmpfont.realize(outerfont);
1375         tmpfont.realize(bparams.getFont());
1376         return tmpfont;
1377 }
1378
1379
1380 /// Returns the height of the highest font in range
1381 Font_size Paragraph::highestFontInRange
1382         (pos_type startpos, pos_type endpos, Font_size def_size) const
1383 {
1384         return d->fontlist_.highestInRange(startpos, endpos, def_size);
1385 }
1386
1387
1388 char_type
1389 Paragraph::getUChar(BufferParams const & bparams, pos_type pos) const
1390 {
1391         char_type c = d->text_[pos];
1392         if (!lyxrc.rtl_support)
1393                 return c;
1394
1395         char_type uc = c;
1396         switch (c) {
1397         case '(':
1398                 uc = ')';
1399                 break;
1400         case ')':
1401                 uc = '(';
1402                 break;
1403         case '[':
1404                 uc = ']';
1405                 break;
1406         case ']':
1407                 uc = '[';
1408                 break;
1409         case '{':
1410                 uc = '}';
1411                 break;
1412         case '}':
1413                 uc = '{';
1414                 break;
1415         case '<':
1416                 uc = '>';
1417                 break;
1418         case '>':
1419                 uc = '<';
1420                 break;
1421         }
1422         if (uc != c && getFontSettings(bparams, pos).isRightToLeft())
1423                 return uc;
1424         else
1425                 return c;
1426 }
1427
1428
1429 void Paragraph::setFont(pos_type pos, Font const & font)
1430 {
1431         BOOST_ASSERT(pos <= size());
1432
1433         // First, reduce font against layout/label font
1434         // Update: The setCharFont() routine in text2.cpp already
1435         // reduces font, so we don't need to do that here. (Asger)
1436
1437         d->fontlist_.set(pos, font);
1438 }
1439
1440
1441 void Paragraph::makeSameLayout(Paragraph const & par)
1442 {
1443         layout(par.layout());
1444         // move to pimpl?
1445         d->params_ = par.params();
1446 }
1447
1448
1449 bool Paragraph::stripLeadingSpaces(bool trackChanges)
1450 {
1451         if (isFreeSpacing())
1452                 return false;
1453
1454         int pos = 0;
1455         int count = 0;
1456
1457         while (pos < size() && (isNewline(pos) || isLineSeparator(pos))) {
1458                 if (eraseChar(pos, trackChanges))
1459                         ++count;
1460                 else
1461                         ++pos;
1462         }
1463
1464         return count > 0 || pos > 0;
1465 }
1466
1467
1468 bool Paragraph::hasSameLayout(Paragraph const & par) const
1469 {
1470         return par.layout() == layout() && d->params_.sameLayout(par.params());
1471 }
1472
1473
1474 depth_type Paragraph::getDepth() const
1475 {
1476         return params().depth();
1477 }
1478
1479
1480 depth_type Paragraph::getMaxDepthAfter() const
1481 {
1482         if (layout()->isEnvironment())
1483                 return params().depth() + 1;
1484         else
1485                 return params().depth();
1486 }
1487
1488
1489 char Paragraph::getAlign() const
1490 {
1491         if (params().align() == LYX_ALIGN_LAYOUT)
1492                 return layout()->align;
1493         else
1494                 return params().align();
1495 }
1496
1497
1498 docstring const & Paragraph::getLabelstring() const
1499 {
1500         return params().labelString();
1501 }
1502
1503
1504 // the next two functions are for the manual labels
1505 docstring const Paragraph::getLabelWidthString() const
1506 {
1507         if (layout()->margintype == MARGIN_MANUAL)
1508                 return params().labelWidthString();
1509         else
1510                 return _("Senseless with this layout!");
1511 }
1512
1513
1514 void Paragraph::setLabelWidthString(docstring const & s)
1515 {
1516         params().labelWidthString(s);
1517 }
1518
1519
1520 docstring const Paragraph::translateIfPossible(docstring const & s,
1521                 BufferParams const & bparams) const
1522 {
1523         if (!support::isAscii(s) || s.empty()) {
1524                 // This must be a user defined layout. We cannot translate
1525                 // this, since gettext accepts only ascii keys.
1526                 return s;
1527         }
1528         // Probably standard layout, try to translate
1529         Messages & m = getMessages(getParLanguage(bparams)->code());
1530         return m.get(to_ascii(s));
1531 }
1532
1533
1534 docstring Paragraph::expandLabel(LayoutPtr const & layout,
1535                 BufferParams const & bparams, bool process_appendix) const
1536 {
1537         TextClass const & tclass = bparams.getTextClass();
1538
1539         docstring fmt;
1540         if (process_appendix && params().appendix())
1541                 fmt = translateIfPossible(layout->labelstring_appendix(),
1542                         bparams);
1543         else
1544                 fmt = translateIfPossible(layout->labelstring(), bparams);
1545
1546         if (fmt.empty() && layout->labeltype == LABEL_COUNTER
1547             && !layout->counter.empty())
1548                 fmt = "\\the" + layout->counter;
1549
1550         // handle 'inherited level parts' in 'fmt',
1551         // i.e. the stuff between '@' in   '@Section@.\arabic{subsection}'
1552         size_t const i = fmt.find('@', 0);
1553         if (i != docstring::npos) {
1554                 size_t const j = fmt.find('@', i + 1);
1555                 if (j != docstring::npos) {
1556                         docstring parent(fmt, i + 1, j - i - 1);
1557                         docstring label = from_ascii("??");
1558                         if (tclass.hasLayout(parent))
1559                                 docstring label = expandLabel(tclass[parent], bparams,
1560                                                       process_appendix);
1561                         fmt = docstring(fmt, 0, i) + label
1562                                 + docstring(fmt, j + 1, docstring::npos);
1563                 }
1564         }
1565
1566         return tclass.counters().counterLabel(fmt);
1567 }
1568
1569
1570 void Paragraph::applyLayout(LayoutPtr const & new_layout)
1571 {
1572         layout(new_layout);
1573         LyXAlignment const oldAlign = params().align();
1574
1575         if (!(oldAlign & layout()->alignpossible)) {
1576                 frontend::Alert::warning(_("Alignment not permitted"),
1577                         _("The new layout does not permit the alignment previously used.\nSetting to default."));
1578                 params().align(LYX_ALIGN_LAYOUT);
1579         }
1580 }
1581
1582
1583 pos_type Paragraph::beginOfBody() const
1584 {
1585         return d->begin_of_body_;
1586 }
1587
1588
1589 void Paragraph::setBeginOfBody()
1590 {
1591         if (layout()->labeltype != LABEL_MANUAL) {
1592                 d->begin_of_body_ = 0;
1593                 return;
1594         }
1595
1596         // Unroll the first two cycles of the loop
1597         // and remember the previous character to
1598         // remove unnecessary getChar() calls
1599         pos_type i = 0;
1600         pos_type end = size();
1601         if (i < end && !isNewline(i)) {
1602                 ++i;
1603                 char_type previous_char = 0;
1604                 char_type temp = 0;
1605                 if (i < end) {
1606                         previous_char = d->text_[i];
1607                         if (!isNewline(i)) {
1608                                 ++i;
1609                                 while (i < end && previous_char != ' ') {
1610                                         temp = d->text_[i];
1611                                         if (isNewline(i))
1612                                                 break;
1613                                         ++i;
1614                                         previous_char = temp;
1615                                 }
1616                         }
1617                 }
1618         }
1619
1620         d->begin_of_body_ = i;
1621 }
1622
1623
1624 InsetBibitem * Paragraph::bibitem() const
1625 {
1626         if (!d->insetlist_.empty()) {
1627                 Inset * inset = d->insetlist_.begin()->inset;
1628                 if (inset->lyxCode() == BIBITEM_CODE)
1629                         return static_cast<InsetBibitem *>(inset);
1630         }
1631         return 0;
1632 }
1633
1634
1635 bool Paragraph::forceDefaultParagraphs() const
1636 {
1637         return inInset() && inInset()->forceDefaultParagraphs(0);
1638 }
1639
1640
1641 namespace {
1642
1643 // paragraphs inside floats need different alignment tags to avoid
1644 // unwanted space
1645
1646 bool noTrivlistCentering(InsetCode code)
1647 {
1648         return code == FLOAT_CODE || code == WRAP_CODE;
1649 }
1650
1651
1652 string correction(string const & orig)
1653 {
1654         if (orig == "flushleft")
1655                 return "raggedright";
1656         if (orig == "flushright")
1657                 return "raggedleft";
1658         if (orig == "center")
1659                 return "centering";
1660         return orig;
1661 }
1662
1663
1664 string const corrected_env(string const & suffix, string const & env,
1665         InsetCode code)
1666 {
1667         string output = suffix + "{";
1668         if (noTrivlistCentering(code))
1669                 output += correction(env);
1670         else
1671                 output += env;
1672         output += "}";
1673         if (suffix == "\\begin")
1674                 output += "\n";
1675         return output;
1676 }
1677
1678
1679 void adjust_row_column(string const & str, TexRow & texrow, int & column)
1680 {
1681         if (!contains(str, "\n"))
1682                 column += str.size();
1683         else {
1684                 string tmp;
1685                 texrow.newline();
1686                 column = rsplit(str, tmp, '\n').size();
1687         }
1688 }
1689
1690 } // namespace anon
1691
1692
1693 int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
1694                                  odocstream & os, TexRow & texrow,
1695                                  bool moving_arg) const
1696 {
1697         int column = 0;
1698
1699         if (params_.noindent()) {
1700                 os << "\\noindent ";
1701                 column += 10;
1702         }
1703
1704         LyXAlignment const curAlign = params_.align();
1705
1706         if (curAlign == layout_->align)
1707                 return column;
1708
1709         switch (curAlign) {
1710         case LYX_ALIGN_NONE:
1711         case LYX_ALIGN_BLOCK:
1712         case LYX_ALIGN_LAYOUT:
1713         case LYX_ALIGN_SPECIAL:
1714                 break;
1715         case LYX_ALIGN_LEFT:
1716         case LYX_ALIGN_RIGHT:
1717         case LYX_ALIGN_CENTER:
1718                 if (moving_arg) {
1719                         os << "\\protect";
1720                         column += 8;
1721                 }
1722                 break;
1723         }
1724
1725         switch (curAlign) {
1726         case LYX_ALIGN_NONE:
1727         case LYX_ALIGN_BLOCK:
1728         case LYX_ALIGN_LAYOUT:
1729         case LYX_ALIGN_SPECIAL:
1730                 break;
1731         case LYX_ALIGN_LEFT: {
1732                 string output;
1733                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
1734                         output = corrected_env("\\begin", "flushleft", owner_->ownerCode());
1735                 else
1736                         output = corrected_env("\\begin", "flushright", owner_->ownerCode());
1737                 os << from_ascii(output);
1738                 adjust_row_column(output, texrow, column);
1739                 break;
1740         } case LYX_ALIGN_RIGHT: {
1741                 string output;
1742                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
1743                         output = corrected_env("\\begin", "flushright", owner_->ownerCode());
1744                 else
1745                         output = corrected_env("\\begin", "flushleft", owner_->ownerCode());
1746                 os << from_ascii(output);
1747                 adjust_row_column(output, texrow, column);
1748                 break;
1749         } case LYX_ALIGN_CENTER: {
1750                 string output;
1751                 output = corrected_env("\\begin", "center", owner_->ownerCode());
1752                 os << from_ascii(output);
1753                 adjust_row_column(output, texrow, column);
1754                 break;
1755         }
1756         }
1757
1758         return column;
1759 }
1760
1761
1762 int Paragraph::Private::endTeXParParams(BufferParams const & bparams,
1763                                odocstream & os, TexRow & texrow,
1764                                bool moving_arg) const
1765 {
1766         int column = 0;
1767
1768         switch (params_.align()) {
1769         case LYX_ALIGN_NONE:
1770         case LYX_ALIGN_BLOCK:
1771         case LYX_ALIGN_LAYOUT:
1772         case LYX_ALIGN_SPECIAL:
1773                 break;
1774         case LYX_ALIGN_LEFT:
1775         case LYX_ALIGN_RIGHT:
1776         case LYX_ALIGN_CENTER:
1777                 if (moving_arg) {
1778                         os << "\\protect";
1779                         column = 8;
1780                 }
1781                 break;
1782         }
1783
1784         switch (params_.align()) {
1785         case LYX_ALIGN_NONE:
1786         case LYX_ALIGN_BLOCK:
1787         case LYX_ALIGN_LAYOUT:
1788         case LYX_ALIGN_SPECIAL:
1789                 break;
1790         case LYX_ALIGN_LEFT: {
1791                 string output;
1792                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
1793                         output = corrected_env("\n\\par\\end", "flushleft", owner_->ownerCode());
1794                 else
1795                         output = corrected_env("\n\\par\\end", "flushright", owner_->ownerCode());
1796                 os << from_ascii(output);
1797                 adjust_row_column(output, texrow, column);
1798                 break;
1799         } case LYX_ALIGN_RIGHT: {
1800                 string output;
1801                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
1802                         output = corrected_env("\n\\par\\end", "flushright", owner_->ownerCode());
1803                 else
1804                         output = corrected_env("\n\\par\\end", "flushleft", owner_->ownerCode());
1805                 os << from_ascii(output);
1806                 adjust_row_column(output, texrow, column);
1807                 break;
1808         } case LYX_ALIGN_CENTER: {
1809                 string output;
1810                 output = corrected_env("\n\\par\\end", "center", owner_->ownerCode());
1811                 os << from_ascii(output);
1812                 adjust_row_column(output, texrow, column);
1813                 break;
1814         }
1815         }
1816
1817         return column;
1818 }
1819
1820
1821 // This one spits out the text of the paragraph
1822 bool Paragraph::latex(Buffer const & buf,
1823                                 BufferParams const & bparams,
1824                                 Font const & outerfont,
1825                                 odocstream & os, TexRow & texrow,
1826                                 OutputParams const & runparams) const
1827 {
1828         LYXERR(Debug::LATEX) << "SimpleTeXOnePar...     " << this << endl;
1829
1830         bool return_value = false;
1831
1832         LayoutPtr style;
1833
1834         // well we have to check if we are in an inset with unlimited
1835         // length (all in one row) if that is true then we don't allow
1836         // any special options in the paragraph and also we don't allow
1837         // any environment other than the default layout of the text class
1838         // to be valid!
1839         bool asdefault = forceDefaultParagraphs();
1840
1841         if (asdefault) {
1842                 style = bparams.getTextClass().defaultLayout();
1843         } else {
1844                 style = layout();
1845         }
1846
1847         // Current base font for all inherited font changes, without any
1848         // change caused by an individual character, except for the language:
1849         // It is set to the language of the first character.
1850         // As long as we are in the label, this font is the base font of the
1851         // label. Before the first body character it is set to the base font
1852         // of the body.
1853         Font basefont;
1854
1855         // Maybe we have to create a optional argument.
1856         pos_type body_pos = beginOfBody();
1857         unsigned int column = 0;
1858
1859         if (body_pos > 0) {
1860                 // the optional argument is kept in curly brackets in
1861                 // case it contains a ']'
1862                 os << "[{";
1863                 column += 2;
1864                 basefont = getLabelFont(bparams, outerfont);
1865         } else {
1866                 basefont = getLayoutFont(bparams, outerfont);
1867         }
1868
1869         // Which font is currently active?
1870         Font running_font(basefont);
1871         // Do we have an open font change?
1872         bool open_font = false;
1873
1874         Change runningChange = Change(Change::UNCHANGED);
1875
1876         texrow.start(id(), 0);
1877
1878         // if the paragraph is empty, the loop will not be entered at all
1879         if (empty()) {
1880                 if (style->isCommand()) {
1881                         os << '{';
1882                         ++column;
1883                 }
1884                 if (!asdefault)
1885                         column += d->startTeXParParams(bparams, os, texrow,
1886                                                     runparams.moving_arg);
1887         }
1888
1889         for (pos_type i = 0; i < size(); ++i) {
1890                 // First char in paragraph or after label?
1891                 if (i == body_pos) {
1892                         if (body_pos > 0) {
1893                                 if (open_font) {
1894                                         column += running_font.latexWriteEndChanges(
1895                                                 os, bparams, runparams,
1896                                                 basefont, basefont);
1897                                         open_font = false;
1898                                 }
1899                                 basefont = getLayoutFont(bparams, outerfont);
1900                                 running_font = basefont;
1901
1902                                 column += Changes::latexMarkChange(os, bparams,
1903                                                 runningChange, Change(Change::UNCHANGED));
1904                                 runningChange = Change(Change::UNCHANGED);
1905
1906                                 os << "}] ";
1907                                 column +=3;
1908                         }
1909                         if (style->isCommand()) {
1910                                 os << '{';
1911                                 ++column;
1912                         }
1913
1914                         if (!asdefault)
1915                                 column += d->startTeXParParams(bparams, os,
1916                                                             texrow,
1917                                                             runparams.moving_arg);
1918                 }
1919
1920                 Change const & change = runparams.inDeletedInset ? runparams.changeOfDeletedInset
1921                                                                  : lookupChange(i);
1922
1923                 if (bparams.outputChanges && runningChange != change) {
1924                         if (open_font) {
1925                                 column += running_font.latexWriteEndChanges(
1926                                                 os, bparams, runparams, basefont, basefont);
1927                                 open_font = false;
1928                         }
1929                         basefont = getLayoutFont(bparams, outerfont);
1930                         running_font = basefont;
1931
1932                         column += Changes::latexMarkChange(os, bparams, runningChange, change);
1933                         runningChange = change;
1934                 }
1935
1936                 // do not output text which is marked deleted
1937                 // if change tracking output is disabled
1938                 if (!bparams.outputChanges && change.type == Change::DELETED) {
1939                         continue;
1940                 }
1941
1942                 ++column;
1943
1944                 // Fully instantiated font
1945                 Font const font = getFont(bparams, i, outerfont);
1946
1947                 Font const last_font = running_font;
1948
1949                 // Do we need to close the previous font?
1950                 if (open_font &&
1951                     (font != running_font ||
1952                      font.language() != running_font.language()))
1953                 {
1954                         column += running_font.latexWriteEndChanges(
1955                                         os, bparams, runparams, basefont,
1956                                         (i == body_pos-1) ? basefont : font);
1957                         running_font = basefont;
1958                         open_font = false;
1959                 }
1960
1961                 // Switch file encoding if necessary (and allowed)
1962                 if (!runparams.verbatim &&
1963                     runparams.encoding->package() == Encoding::inputenc &&
1964                     font.language()->encoding()->package() == Encoding::inputenc) {
1965                         std::pair<bool, int> const enc_switch = switchEncoding(os, bparams,
1966                                         runparams.moving_arg, *(runparams.encoding),
1967                                         *(font.language()->encoding()));
1968                         if (enc_switch.first) {
1969                                 column += enc_switch.second;
1970                                 runparams.encoding = font.language()->encoding();
1971                         }
1972                 }
1973
1974                 char_type const c = d->text_[i];
1975
1976                 // Do we need to change font?
1977                 if ((font != running_font ||
1978                      font.language() != running_font.language()) &&
1979                         i != body_pos - 1)
1980                 {
1981                         odocstringstream ods;
1982                         column += font.latexWriteStartChanges(ods, bparams,
1983                                                               runparams, basefont,
1984                                                               last_font);
1985                         running_font = font;
1986                         open_font = true;
1987                         docstring fontchange = ods.str();
1988                         // check if the fontchange ends with a trailing blank
1989                         // (like "\small " (see bug 3382)
1990                         if (suffixIs(fontchange, ' ') && c == ' ')
1991                                 os << fontchange.substr(0, fontchange.size() - 1)
1992                                    << from_ascii("{}");
1993                         else
1994                                 os << fontchange;
1995                 }
1996
1997                 if (c == ' ') {
1998                         // FIXME: integrate this case in latexSpecialChar
1999                         // Do not print the separation of the optional argument
2000                         // if style->pass_thru is false. This works because
2001                         // latexSpecialChar ignores spaces if
2002                         // style->pass_thru is false.
2003                         if (i != body_pos - 1) {
2004                                 if (d->simpleTeXBlanks(
2005                                                 *(runparams.encoding), os, texrow,
2006                                                 i, column, font, *style)) {
2007                                         // A surrogate pair was output. We
2008                                         // must not call latexSpecialChar
2009                                         // in this iteration, since it would output
2010                                         // the combining character again.
2011                                         ++i;
2012                                         continue;
2013                                 }
2014                         }
2015                 }
2016
2017                 OutputParams rp = runparams;
2018                 rp.free_spacing = style->free_spacing;
2019                 rp.local_font = &font;
2020                 rp.intitle = style->intitle;
2021
2022                 // Two major modes:  LaTeX or plain
2023                 // Handle here those cases common to both modes
2024                 // and then split to handle the two modes separately.
2025                 if (c == META_INSET)
2026                         d->latexInset(buf, bparams, os,
2027                                         texrow, rp, running_font,
2028                                         basefont, outerfont, open_font,
2029                                         runningChange, *style, i, column);
2030                 else
2031                         d->latexSpecialChar(os, rp, running_font, runningChange,
2032                                 *style, i, column);
2033
2034                 // Set the encoding to that returned from simpleTeXSpecialChars (see
2035                 // comment for encoding member in OutputParams.h)
2036                 runparams.encoding = rp.encoding;
2037         }
2038
2039         // If we have an open font definition, we have to close it
2040         if (open_font) {
2041 #ifdef FIXED_LANGUAGE_END_DETECTION
2042                 if (next_) {
2043                         running_font
2044                                 .latexWriteEndChanges(os, bparams, runparams,
2045                                         basefont,
2046                                         next_->getFont(bparams, 0, outerfont));
2047                 } else {
2048                         running_font.latexWriteEndChanges(os, bparams,
2049                                         runparams, basefont, basefont);
2050                 }
2051 #else
2052 //FIXME: For now we ALWAYS have to close the foreign font settings if they are
2053 //FIXME: there as we start another \selectlanguage with the next paragraph if
2054 //FIXME: we are in need of this. This should be fixed sometime (Jug)
2055                 running_font.latexWriteEndChanges(os, bparams, runparams,
2056                                 basefont, basefont);
2057 #endif
2058         }
2059
2060         column += Changes::latexMarkChange(os, bparams, runningChange, Change(Change::UNCHANGED));
2061
2062         // Needed if there is an optional argument but no contents.
2063         if (body_pos > 0 && body_pos == size()) {
2064                 os << "}]~";
2065                 return_value = false;
2066         }
2067
2068         if (!asdefault) {
2069                 column += d->endTeXParParams(bparams, os, texrow,
2070                                           runparams.moving_arg);
2071         }
2072
2073         LYXERR(Debug::LATEX) << "SimpleTeXOnePar...done " << this << endl;
2074         return return_value;
2075 }
2076
2077
2078 namespace {
2079
2080 enum PAR_TAG {
2081         PAR_NONE=0,
2082         TT = 1,
2083         SF = 2,
2084         BF = 4,
2085         IT = 8,
2086         SL = 16,
2087         EM = 32
2088 };
2089
2090
2091 string tag_name(PAR_TAG const & pt) {
2092         switch (pt) {
2093         case PAR_NONE: return "!-- --";
2094         case TT: return "tt";
2095         case SF: return "sf";
2096         case BF: return "bf";
2097         case IT: return "it";
2098         case SL: return "sl";
2099         case EM: return "em";
2100         }
2101         return "";
2102 }
2103
2104
2105 inline
2106 void operator|=(PAR_TAG & p1, PAR_TAG const & p2)
2107 {
2108         p1 = static_cast<PAR_TAG>(p1 | p2);
2109 }
2110
2111
2112 inline
2113 void reset(PAR_TAG & p1, PAR_TAG const & p2)
2114 {
2115         p1 = static_cast<PAR_TAG>(p1 & ~p2);
2116 }
2117
2118 } // anon
2119
2120
2121 bool Paragraph::emptyTag() const
2122 {
2123         for (pos_type i = 0; i < size(); ++i) {
2124                 if (isInset(i)) {
2125                         Inset const * inset = getInset(i);
2126                         InsetCode lyx_code = inset->lyxCode();
2127                         if (lyx_code != TOC_CODE &&
2128                             lyx_code != INCLUDE_CODE &&
2129                             lyx_code != GRAPHICS_CODE &&
2130                             lyx_code != ERT_CODE &&
2131                             lyx_code != LISTINGS_CODE &&
2132                             lyx_code != FLOAT_CODE &&
2133                             lyx_code != TABULAR_CODE) {
2134                                 return false;
2135                         }
2136                 } else {
2137                         char_type c = d->text_[i];
2138                         if (c != ' ' && c != '\t')
2139                                 return false;
2140                 }
2141         }
2142         return true;
2143 }
2144
2145
2146 string Paragraph::getID(Buffer const & buf, OutputParams const & runparams) const
2147 {
2148         for (pos_type i = 0; i < size(); ++i) {
2149                 if (isInset(i)) {
2150                         Inset const * inset = getInset(i);
2151                         InsetCode lyx_code = inset->lyxCode();
2152                         if (lyx_code == LABEL_CODE) {
2153                                 InsetLabel const * const il = static_cast<InsetLabel const *>(inset);
2154                                 docstring const & id = il->getParam("name");
2155                                 return "id='" + to_utf8(sgml::cleanID(buf, runparams, id)) + "'";
2156                         }
2157                 }
2158
2159         }
2160         return string();
2161 }
2162
2163
2164 pos_type Paragraph::getFirstWord(Buffer const & buf, odocstream & os, OutputParams const & runparams) const
2165 {
2166         pos_type i;
2167         for (i = 0; i < size(); ++i) {
2168                 if (isInset(i)) {
2169                         Inset const * inset = getInset(i);
2170                         inset->docbook(buf, os, runparams);
2171                 } else {
2172                         char_type c = d->text_[i];
2173                         if (c == ' ')
2174                                 break;
2175                         os << sgml::escapeChar(c);
2176                 }
2177         }
2178         return i;
2179 }
2180
2181
2182 bool Paragraph::onlyText(Buffer const & buf, Font const & outerfont, pos_type initial) const
2183 {
2184         Font font_old;
2185
2186         for (pos_type i = initial; i < size(); ++i) {
2187                 Font font = getFont(buf.params(), i, outerfont);
2188                 if (isInset(i))
2189                         return false;
2190                 if (i != initial && font != font_old)
2191                         return false;
2192                 font_old = font;
2193         }
2194
2195         return true;
2196 }
2197
2198
2199 void Paragraph::simpleDocBookOnePar(Buffer const & buf,
2200                                     odocstream & os,
2201                                     OutputParams const & runparams,
2202                                     Font const & outerfont,
2203                                     pos_type initial) const
2204 {
2205         bool emph_flag = false;
2206
2207         LayoutPtr const & style = layout();
2208         Font font_old =
2209                 style->labeltype == LABEL_MANUAL ? style->labelfont : style->font;
2210
2211         if (style->pass_thru && !onlyText(buf, outerfont, initial))
2212                 os << "]]>";
2213
2214         // parsing main loop
2215         for (pos_type i = initial; i < size(); ++i) {
2216                 Font font = getFont(buf.params(), i, outerfont);
2217
2218                 // handle <emphasis> tag
2219                 if (font_old.emph() != font.emph()) {
2220                         if (font.emph() == Font::ON) {
2221                                 os << "<emphasis>";
2222                                 emph_flag = true;
2223                         } else if (i != initial) {
2224                                 os << "</emphasis>";
2225                                 emph_flag = false;
2226                         }
2227                 }
2228
2229                 if (isInset(i)) {
2230                         Inset const * inset = getInset(i);
2231                         inset->docbook(buf, os, runparams);
2232                 } else {
2233                         char_type c = d->text_[i];
2234
2235                         if (style->pass_thru)
2236                                 os.put(c);
2237                         else
2238                                 os << sgml::escapeChar(c);
2239                 }
2240                 font_old = font;
2241         }
2242
2243         if (emph_flag) {
2244                 os << "</emphasis>";
2245         }
2246
2247         if (style->free_spacing)
2248                 os << '\n';
2249         if (style->pass_thru && !onlyText(buf, outerfont, initial))
2250                 os << "<![CDATA[";
2251 }
2252
2253
2254 bool Paragraph::isHfill(pos_type pos) const
2255 {
2256         return isInset(pos)
2257                 && getInset(pos)->lyxCode() == HFILL_CODE;
2258 }
2259
2260
2261 bool Paragraph::isNewline(pos_type pos) const
2262 {
2263         return isInset(pos)
2264                 && getInset(pos)->lyxCode() == NEWLINE_CODE;
2265 }
2266
2267
2268 bool Paragraph::isLineSeparator(pos_type pos) const
2269 {
2270         char_type const c = d->text_[pos];
2271         return isLineSeparatorChar(c)
2272                 || (c == META_INSET && getInset(pos) &&
2273                 getInset(pos)->isLineSeparator());
2274 }
2275
2276
2277 /// Used by the spellchecker
2278 bool Paragraph::isLetter(pos_type pos) const
2279 {
2280         if (isInset(pos))
2281                 return getInset(pos)->isLetter();
2282         else {
2283                 char_type const c = d->text_[pos];
2284                 return isLetterChar(c) || isDigit(c);
2285         }
2286 }
2287
2288
2289 Language const *
2290 Paragraph::getParLanguage(BufferParams const & bparams) const
2291 {
2292         if (!empty())
2293                 return getFirstFontSettings(bparams).language();
2294         // FIXME: we should check the prev par as well (Lgb)
2295         return bparams.language;
2296 }
2297
2298
2299 bool Paragraph::isRTL(BufferParams const & bparams) const
2300 {
2301         return lyxrc.rtl_support
2302                 && getParLanguage(bparams)->rightToLeft()
2303                 && ownerCode() != ERT_CODE
2304                 && ownerCode() != LISTINGS_CODE;
2305 }
2306
2307
2308 void Paragraph::changeLanguage(BufferParams const & bparams,
2309                                Language const * from, Language const * to)
2310 {
2311         // change language including dummy font change at the end
2312         for (pos_type i = 0; i <= size(); ++i) {
2313                 Font font = getFontSettings(bparams, i);
2314                 if (font.language() == from) {
2315                         font.setLanguage(to);
2316                         setFont(i, font);
2317                 }
2318         }
2319 }
2320
2321
2322 bool Paragraph::isMultiLingual(BufferParams const & bparams) const
2323 {
2324         Language const * doc_language = bparams.language;
2325         FontList::const_iterator cit = d->fontlist_.begin();
2326         FontList::const_iterator end = d->fontlist_.end();
2327
2328         for (; cit != end; ++cit)
2329                 if (cit->font().language() != ignore_language &&
2330                     cit->font().language() != latex_language &&
2331                     cit->font().language() != doc_language)
2332                         return true;
2333         return false;
2334 }
2335
2336
2337 // Convert the paragraph to a string.
2338 // Used for building the table of contents
2339 docstring const Paragraph::asString(Buffer const & buffer, bool label) const
2340 {
2341         return asString(buffer, 0, size(), label);
2342 }
2343
2344
2345 docstring const Paragraph::asString(Buffer const & buffer,
2346                                  pos_type beg, pos_type end, bool label) const
2347 {
2348
2349         odocstringstream os;
2350
2351         if (beg == 0 && label && !params().labelString().empty())
2352                 os << params().labelString() << ' ';
2353
2354         for (pos_type i = beg; i < end; ++i) {
2355                 char_type const c = d->text_[i];
2356                 if (isPrintable(c))
2357                         os.put(c);
2358                 else if (c == META_INSET)
2359                         getInset(i)->textString(buffer, os);
2360         }
2361
2362         return os.str();
2363 }
2364
2365
2366 void Paragraph::setInsetOwner(Inset * inset)
2367 {
2368         d->inset_owner_ = inset;
2369 }
2370
2371
2372 int Paragraph::id() const
2373 {
2374         return d->id_;
2375 }
2376
2377
2378 LayoutPtr const & Paragraph::layout() const
2379 {
2380         return d->layout_;
2381 }
2382
2383
2384 void Paragraph::layout(LayoutPtr const & new_layout)
2385 {
2386         d->layout_ = new_layout;
2387 }
2388
2389
2390 Inset * Paragraph::inInset() const
2391 {
2392         return d->inset_owner_;
2393 }
2394
2395
2396 InsetCode Paragraph::ownerCode() const
2397 {
2398         return d->inset_owner_ ?
2399                 d->inset_owner_->lyxCode() : NO_CODE;
2400 }
2401
2402
2403 ParagraphParameters & Paragraph::params()
2404 {
2405         return d->params_;
2406 }
2407
2408
2409 ParagraphParameters const & Paragraph::params() const
2410 {
2411         return d->params_;
2412 }
2413
2414
2415 bool Paragraph::isFreeSpacing() const
2416 {
2417         if (layout()->free_spacing)
2418                 return true;
2419
2420         // for now we just need this, later should we need this in some
2421         // other way we can always add a function to Inset too.
2422         return ownerCode() == ERT_CODE || ownerCode() == LISTINGS_CODE;
2423 }
2424
2425
2426 bool Paragraph::allowEmpty() const
2427 {
2428         if (layout()->keepempty)
2429                 return true;
2430         return ownerCode() == ERT_CODE || ownerCode() == LISTINGS_CODE;
2431 }
2432
2433
2434 char_type Paragraph::transformChar(char_type c, pos_type pos) const
2435 {
2436         if (!Encodings::is_arabic(c))
2437                 return c;
2438
2439         char_type prev_char = ' ';
2440         char_type next_char = ' ';
2441
2442         for (pos_type i = pos - 1; i >= 0; --i) {
2443                 char_type const par_char = d->text_[i];
2444                 if (!Encodings::isComposeChar_arabic(par_char)) {
2445                         prev_char = par_char;
2446                         break;
2447                 }
2448         }
2449
2450         for (pos_type i = pos + 1, end = size(); i < end; ++i) {
2451                 char_type const par_char = d->text_[i];
2452                 if (!Encodings::isComposeChar_arabic(par_char)) {
2453                         next_char = par_char;
2454                         break;
2455                 }
2456         }
2457
2458         if (Encodings::is_arabic(next_char)) {
2459                 if (Encodings::is_arabic(prev_char) &&
2460                         !Encodings::is_arabic_special(prev_char))
2461                         return Encodings::transformChar(c, Encodings::FORM_MEDIAL);
2462                 else
2463                         return Encodings::transformChar(c, Encodings::FORM_INITIAL);
2464         } else {
2465                 if (Encodings::is_arabic(prev_char) &&
2466                         !Encodings::is_arabic_special(prev_char))
2467                         return Encodings::transformChar(c, Encodings::FORM_FINAL);
2468                 else
2469                         return Encodings::transformChar(c, Encodings::FORM_ISOLATED);
2470         }
2471 }
2472
2473
2474 int Paragraph::checkBiblio(bool track_changes)
2475 {
2476         //FIXME From JS:
2477         //This is getting more and more a mess. ...We really should clean
2478         //up this bibitem issue for 1.6. See also bug 2743.
2479
2480         // Add bibitem insets if necessary
2481         if (layout()->labeltype != LABEL_BIBLIO)
2482                 return 0;
2483
2484         bool hasbibitem = !d->insetlist_.empty()
2485                 // Insist on it being in pos 0
2486                 && d->text_[0] == META_INSET
2487                 && d->insetlist_.begin()->inset->lyxCode() == BIBITEM_CODE;
2488
2489         docstring oldkey;
2490         docstring oldlabel;
2491
2492         // remove a bibitem in pos != 0
2493         // restore it later in pos 0 if necessary
2494         // (e.g. if a user inserts contents _before_ the item)
2495         // we're assuming there's only one of these, which there
2496         // should be.
2497         int erasedInsetPosition = -1;
2498         InsetList::iterator it = d->insetlist_.begin();
2499         InsetList::iterator end = d->insetlist_.end();
2500         for (; it != end; ++it)
2501                 if (it->inset->lyxCode() == BIBITEM_CODE
2502                     && it->pos > 0) {
2503                         InsetBibitem * olditem = static_cast<InsetBibitem *>(it->inset);
2504                         oldkey = olditem->getParam("key");
2505                         oldlabel = olditem->getParam("label");
2506                         erasedInsetPosition = it->pos;
2507                         eraseChar(erasedInsetPosition, track_changes);
2508                         break;
2509         }
2510
2511         //There was an InsetBibitem at the beginning, and we didn't
2512         //have to erase one.
2513         if (hasbibitem && erasedInsetPosition < 0)
2514                         return 0;
2515
2516         //There was an InsetBibitem at the beginning and we did have to
2517         //erase one. So we give its properties to the beginning inset.
2518         if (hasbibitem) {
2519                 InsetBibitem * inset =
2520                         static_cast<InsetBibitem *>(d->insetlist_.begin()->inset);
2521                 if (!oldkey.empty())
2522                         inset->setParam("key", oldkey);
2523                 inset->setParam("label", oldlabel);
2524                 return -erasedInsetPosition;
2525         }
2526
2527         //There was no inset at the beginning, so we need to create one with
2528         //the key and label of the one we erased.
2529         InsetBibitem * inset(new InsetBibitem(InsetCommandParams(BIBITEM_CODE)));
2530         // restore values of previously deleted item in this par.
2531         if (!oldkey.empty())
2532                 inset->setParam("key", oldkey);
2533         inset->setParam("label", oldlabel);
2534         insertInset(0, static_cast<Inset *>(inset),
2535                     Change(track_changes ? Change::INSERTED : Change::UNCHANGED));
2536
2537         return 1;
2538 }
2539
2540
2541 void Paragraph::checkAuthors(AuthorList const & authorList)
2542 {
2543         d->changes_.checkAuthors(authorList);
2544 }
2545
2546
2547 bool Paragraph::isUnchanged(pos_type pos) const
2548 {
2549         return lookupChange(pos).type == Change::UNCHANGED;
2550 }
2551
2552
2553 bool Paragraph::isInserted(pos_type pos) const
2554 {
2555         return lookupChange(pos).type == Change::INSERTED;
2556 }
2557
2558
2559 bool Paragraph::isDeleted(pos_type pos) const
2560 {
2561         return lookupChange(pos).type == Change::DELETED;
2562 }
2563
2564
2565 InsetList const & Paragraph::insetList() const
2566 {
2567         return d->insetlist_;
2568 }
2569
2570
2571 Inset * Paragraph::releaseInset(pos_type pos)
2572 {
2573         Inset * inset = d->insetlist_.release(pos);
2574         /// does not honour change tracking!
2575         eraseChar(pos, false);
2576         return inset;
2577 }
2578
2579
2580 Inset * Paragraph::getInset(pos_type pos)
2581 {
2582         return d->insetlist_.get(pos);
2583 }
2584
2585
2586 Inset const * Paragraph::getInset(pos_type pos) const
2587 {
2588         return d->insetlist_.get(pos);
2589 }
2590
2591
2592 int Paragraph::numberOfOptArgs() const
2593 {
2594         int num = 0;
2595         InsetList::const_iterator it = insetList().begin();
2596         InsetList::const_iterator end = insetList().end();
2597         for (; it != end ; ++it) {
2598                 if (it->inset->lyxCode() == OPTARG_CODE)
2599                         ++num;
2600         }
2601         return num;
2602 }
2603
2604
2605 void Paragraph::changeCase(BufferParams const & bparams, pos_type pos,
2606                 pos_type right, TextCase action)
2607 {
2608         // process sequences of modified characters; in change
2609         // tracking mode, this approach results in much better
2610         // usability than changing case on a char-by-char basis
2611         docstring changes;
2612
2613         bool const trackChanges = bparams.trackChanges;
2614
2615         bool capitalize = true;
2616
2617         for (; pos < right; ++pos) {
2618                 char_type oldChar = d->text_[pos];
2619                 char_type newChar = oldChar;
2620
2621                 // ignore insets and don't play with deleted text!
2622                 if (isInset(pos) && !isDeleted(pos)) {
2623                         switch (action) {
2624                                 case text_lowercase:
2625                                         newChar = lowercase(oldChar);
2626                                         break;
2627                                 case text_capitalization:
2628                                         if (capitalize) {
2629                                                 newChar = uppercase(oldChar);
2630                                                 capitalize = false;
2631                                         }
2632                                         break;
2633                                 case text_uppercase:
2634                                         newChar = uppercase(oldChar);
2635                                         break;
2636                         }
2637                 }
2638
2639                 if (!isLetter(pos) || isDeleted(pos)) {
2640                         // permit capitalization again
2641                         capitalize = true;
2642                 }
2643
2644                 if (oldChar != newChar)
2645                         changes += newChar;
2646
2647                 if (oldChar == newChar || pos == right - 1) {
2648                         if (oldChar != newChar) {
2649                                 // step behind the changing area
2650                                 pos++;
2651                         }
2652                         int erasePos = pos - changes.size();
2653                         for (size_t i = 0; i < changes.size(); i++) {
2654                                 insertChar(pos, changes[i],
2655                                         getFontSettings(bparams,
2656                                         erasePos),
2657                                         trackChanges);
2658                                 if (!eraseChar(erasePos, trackChanges)) {
2659                                         ++erasePos;
2660                                         ++pos; // advance
2661                                         ++right; // expand selection
2662                                 }
2663                         }
2664                         changes.clear();
2665                 }
2666         }
2667 }
2668
2669
2670 bool Paragraph::find(docstring const & str, bool cs, bool mw,
2671                 pos_type pos, bool del) const
2672 {
2673         int const strsize = str.length();
2674         int i = 0;
2675         pos_type const parsize = d->text_.size();
2676         for (i = 0; pos + i < parsize; ++i) {
2677                 if (i >= strsize)
2678                         break;
2679                 if (cs && str[i] != d->text_[pos + i])
2680                         break;
2681                 if (!cs && uppercase(str[i]) != uppercase(d->text_[pos + i]))
2682                         break;
2683                 if (!del && isDeleted(pos + i))
2684                         break;
2685         }
2686
2687         if (i != strsize)
2688                 return false;
2689
2690         // if necessary, check whether string matches word
2691         if (mw) {
2692                 if (pos > 0 && isLetter(pos - 1))
2693                         return false;
2694                 if (pos + strsize < parsize
2695                         && isLetter(pos + strsize))
2696                         return false;
2697         }
2698
2699         return true;
2700 }
2701
2702
2703 char_type Paragraph::getChar(pos_type pos) const
2704 {
2705         return d->text_[pos];
2706 }
2707
2708
2709 pos_type Paragraph::size() const
2710 {
2711         return d->text_.size();
2712 }
2713
2714
2715 bool Paragraph::empty() const
2716 {
2717         return d->text_.empty();
2718 }
2719
2720
2721 bool Paragraph::isInset(pos_type pos) const
2722 {
2723         return d->text_[pos] == META_INSET;
2724 }
2725
2726
2727 bool Paragraph::isSeparator(pos_type pos) const
2728 {
2729         //FIXME: Are we sure this can be the only separator?
2730         return d->text_[pos] == ' ';
2731 }
2732
2733
2734 } // namespace lyx