src/Paragraph.cpp

   1 /**
   2  * \file Paragraph.cpp
   3  * This file is part of LyX, the document processor.
   4  * Licence details can be found in the file COPYING.
   5  *
   6  * \author Asger Alstrup
   7  * \author Lars Gullik Bjønnes
   8  * \author Jean-Marc Lasgouttes
   9  * \author Angus Leeming
  10  * \author John Levon
  11  * \author André Pönitz
  12  * \author Dekel Tsur
  13  * \author Jürgen Vigna
  14  *
  15  * Full author contact details are available in file CREDITS.
  16  */
  17
  18 #include <config.h>
  19
  20 #include "Paragraph.h"
  21
  22 #include "LayoutFile.h"
  23 #include "Buffer.h"
  24 #include "BufferParams.h"
  25 #include "Changes.h"
  26 #include "Counters.h"
  27 #include "Encoding.h"
  28 #include "InsetList.h"
  29 #include "Language.h"
  30 #include "LaTeXFeatures.h"
  31 #include "Layout.h"
  32 #include "Length.h"
  33 #include "Font.h"
  34 #include "FontList.h"
  35 #include "LyXRC.h"
  36 #include "OutputParams.h"
  37 #include "output_latex.h"
  38 #include "paragraph_funcs.h"
  39 #include "ParagraphParameters.h"
  40 #include "sgml.h"
  41 #include "TextClass.h"
  42 #include "TexRow.h"
  43 #include "Text.h"
  44 #include "VSpace.h"
  45 #include "WordList.h"
  46
  47 #include "frontends/alert.h"
  48
  49 #include "insets/InsetBibitem.h"
  50 #include "insets/InsetLabel.h"
  51
  52 #include "support/lassert.h"
  53 #include "support/convert.h"
  54 #include "support/debug.h"
  55 #include "support/gettext.h"
  56 #include "support/lstrings.h"
  57 #include "support/Messages.h"
  58 #include "support/textutils.h"
  59
  60 #include <sstream>
  61 #include <vector>
  62
  63 using namespace std;
  64 using namespace lyx::support;
  65
  66 namespace lyx {
  67
  68 namespace {
  69 /// Inset identifier (above 0x10ffff, for ucs-4)
  70 char_type const META_INSET = 0x200001;
  71 };
  72
  73 /////////////////////////////////////////////////////////////////////
  74 //
  75 // Paragraph::Private
  76 //
  77 /////////////////////////////////////////////////////////////////////
  78
  79 class Paragraph::Private
  80 {
  81 public:
  82         ///
  83         Private(Paragraph * owner, Layout const & layout);
  84         /// "Copy constructor"
  85         Private(Private const &, Paragraph * owner);
  86
  87         ///
  88         void insertChar(pos_type pos, char_type c, Change const & change);
  89
  90         /// Output the surrogate pair formed by \p c and \p next to \p os.
  91         /// \return the number of characters written.
  92         int latexSurrogatePair(odocstream & os, char_type c, char_type next,
  93                                Encoding const &);
  94
  95         /// Output a space in appropriate formatting (or a surrogate pair
  96         /// if the next character is a combining character).
  97         /// \return whether a surrogate pair was output.
  98         bool simpleTeXBlanks(OutputParams const &,
  99                              odocstream &, TexRow & texrow,
 100                              pos_type i,
 101                              unsigned int & column,
 102                              Font const & font,
 103                              Layout const & style);
 104
 105         /// Output consecutive unicode chars, belonging to the same script as
 106         /// specified by the latex macro \p ltx, to \p os starting from \p i.
 107         /// \return the number of characters written.
 108         int writeScriptChars(odocstream & os, docstring const & ltx,
 109                            Change &, Encoding const &, pos_type & i);
 110
 111         /// This could go to ParagraphParameters if we want to.
 112         int startTeXParParams(BufferParams const &, odocstream &, TexRow &,
 113                               bool) const;
 114
 115         /// This could go to ParagraphParameters if we want to.
 116         int endTeXParParams(BufferParams const &, odocstream &, TexRow &,
 117                             bool) const;
 118
 119         ///
 120         void latexInset(BufferParams const &,
 121                                    odocstream &,
 122                                    TexRow & texrow, OutputParams &,
 123                                    Font & running_font,
 124                                    Font & basefont,
 125                                    Font const & outerfont,
 126                                    bool & open_font,
 127                                    Change & running_change,
 128                                    Layout const & style,
 129                                    pos_type & i,
 130                                    unsigned int & column);
 131
 132         ///
 133         void latexSpecialChar(
 134                                    odocstream & os,
 135                                    OutputParams & runparams,
 136                                    Font & running_font,
 137                                    Change & running_change,
 138                                    Layout const & style,
 139                                    pos_type & i,
 140                                    unsigned int & column);
 141
 142         ///
 143         bool latexSpecialT1(
 144                 char_type const c,
 145                 odocstream & os,
 146                 pos_type & i,
 147                 unsigned int & column);
 148         ///
 149         bool latexSpecialTypewriter(
 150                 char_type const c,
 151                 odocstream & os,
 152                 pos_type & i,
 153                 unsigned int & column);
 154         ///
 155         bool latexSpecialPhrase(
 156                 odocstream & os,
 157                 pos_type & i,
 158                 unsigned int & column,
 159                 OutputParams & runparams);
 160
 161         ///
 162         void validate(LaTeXFeatures & features,
 163                       Layout const & layout) const;
 164
 165         /// Checks if the paragraph contains only text and no inset or font change.
 166         bool onlyText(Buffer const & buf, Font const & outerfont,
 167                       pos_type initial) const;
 168
 169         /// match a string against a particular point in the paragraph
 170         bool isTextAt(string const & str, pos_type pos) const;
 171
 172         /// Which Paragraph owns us?
 173         Paragraph * owner_;
 174
 175         /// In which Inset?
 176         Inset * inset_owner_;
 177
 178         ///
 179         FontList fontlist_;
 180
 181         ///
 182         unsigned int id_;
 183         ///
 184         static unsigned int paragraph_id;
 185         ///
 186         ParagraphParameters params_;
 187
 188         /// for recording and looking up changes
 189         Changes changes_;
 190
 191         ///
 192         InsetList insetlist_;
 193
 194         /// end of label
 195         pos_type begin_of_body_;
 196
 197         typedef docstring TextContainer;
 198         ///
 199         TextContainer text_;
 200
 201         typedef std::set<docstring> Words;
 202         ///
 203         Words words_;
 204         ///
 205         Layout const * layout_;
 206 };
 207
 208
 209 // Initialization of the counter for the paragraph id's,
 210 unsigned int Paragraph::Private::paragraph_id = 0;
 211
 212 namespace {
 213
 214 struct special_phrase {
 215         string phrase;
 216         docstring macro;
 217         bool builtin;
 218 };
 219
 220 special_phrase const special_phrases[] = {
 221         { "LyX", from_ascii("\\LyX{}"), false },
 222         { "TeX", from_ascii("\\TeX{}"), true },
 223         { "LaTeX2e", from_ascii("\\LaTeXe{}"), true },
 224         { "LaTeX", from_ascii("\\LaTeX{}"), true },
 225 };
 226
 227 size_t const phrases_nr = sizeof(special_phrases)/sizeof(special_phrase);
 228
 229 } // namespace anon
 230
 231
 232 Paragraph::Private::Private(Paragraph * owner, Layout const & layout)
 233         : owner_(owner), inset_owner_(0), begin_of_body_(0), layout_(&layout)
 234 {
 235         id_ = paragraph_id++;
 236         text_.reserve(100);
 237 }
 238
 239
 240 Paragraph::Private::Private(Private const & p, Paragraph * owner)
 241         : owner_(owner), inset_owner_(p.inset_owner_), fontlist_(p.fontlist_),
 242           params_(p.params_), changes_(p.changes_), insetlist_(p.insetlist_),
 243           begin_of_body_(p.begin_of_body_), text_(p.text_), words_(p.words_),
 244           layout_(p.layout_)
 245 {
 246         id_ = paragraph_id++;
 247 }
 248
 249
 250 bool Paragraph::isChanged(pos_type start, pos_type end) const
 251 {
 252         LASSERT(start >= 0 && start <= size(), /**/);
 253         LASSERT(end > start && end <= size() + 1, /**/);
 254
 255         return d->changes_.isChanged(start, end);
 256 }
 257
 258
 259 bool Paragraph::isMergedOnEndOfParDeletion(bool trackChanges) const
 260 {
 261         // keep the logic here in sync with the logic of eraseChars()
 262         if (!trackChanges)
 263                 return true;
 264
 265         Change const change = d->changes_.lookup(size());
 266         return change.type == Change::INSERTED && change.author == 0;
 267 }
 268
 269
 270 void Paragraph::setChange(Change const & change)
 271 {
 272         // beware of the imaginary end-of-par character!
 273         d->changes_.set(change, 0, size() + 1);
 274
 275         /*
 276          * Propagate the change recursively - but not in case of DELETED!
 277          *
 278          * Imagine that your co-author makes changes in an existing inset. He
 279          * sends your document to you and you come to the conclusion that the
 280          * inset should go completely. If you erase it, LyX must not delete all
 281          * text within the inset. Otherwise, the change tracked insertions of
 282          * your co-author get lost and there is no way to restore them later.
 283          *
 284          * Conclusion: An inset's content should remain untouched if you delete it
 285          */
 286
 287         if (change.type != Change::DELETED) {
 288                 for (pos_type pos = 0; pos < size(); ++pos) {
 289                         if (Inset * inset = getInset(pos))
 290                                 inset->setChange(change);
 291                 }
 292         }
 293 }
 294
 295
 296 void Paragraph::setChange(pos_type pos, Change const & change)
 297 {
 298         LASSERT(pos >= 0 && pos <= size(), /**/);
 299         d->changes_.set(change, pos);
 300
 301         // see comment in setChange(Change const &) above
 302         if (change.type != Change::DELETED && pos < size())
 303                         if (Inset * inset = getInset(pos))
 304                                 inset->setChange(change);
 305 }
 306
 307
 308 Change const & Paragraph::lookupChange(pos_type pos) const
 309 {
 310         LASSERT(pos >= 0 && pos <= size(), /**/);
 311         return d->changes_.lookup(pos);
 312 }
 313
 314
 315 void Paragraph::acceptChanges(BufferParams const & bparams, pos_type start,
 316                 pos_type end)
 317 {
 318         LASSERT(start >= 0 && start <= size(), /**/);
 319         LASSERT(end > start && end <= size() + 1, /**/);
 320
 321         for (pos_type pos = start; pos < end; ++pos) {
 322                 switch (lookupChange(pos).type) {
 323                         case Change::UNCHANGED:
 324                                 // accept changes in nested inset
 325                                 if (Inset * inset = getInset(pos))
 326                                         inset->acceptChanges(bparams);
 327                                 break;
 328
 329                         case Change::INSERTED:
 330                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 331                                 // also accept changes in nested inset
 332                                 if (Inset * inset = getInset(pos))
 333                                         inset->acceptChanges(bparams);
 334                                 break;
 335
 336                         case Change::DELETED:
 337                                 // Suppress access to non-existent
 338                                 // "end-of-paragraph char"
 339                                 if (pos < size()) {
 340                                         eraseChar(pos, false);
 341                                         --end;
 342                                         --pos;
 343                                 }
 344                                 break;
 345                 }
 346
 347         }
 348 }
 349
 350
 351 void Paragraph::rejectChanges(BufferParams const & bparams,
 352                 pos_type start, pos_type end)
 353 {
 354         LASSERT(start >= 0 && start <= size(), /**/);
 355         LASSERT(end > start && end <= size() + 1, /**/);
 356
 357         for (pos_type pos = start; pos < end; ++pos) {
 358                 switch (lookupChange(pos).type) {
 359                         case Change::UNCHANGED:
 360                                 // reject changes in nested inset
 361                                 if (Inset * inset = getInset(pos))
 362                                                 inset->rejectChanges(bparams);
 363                                 break;
 364
 365                         case Change::INSERTED:
 366                                 // Suppress access to non-existent
 367                                 // "end-of-paragraph char"
 368                                 if (pos < size()) {
 369                                         eraseChar(pos, false);
 370                                         --end;
 371                                         --pos;
 372                                 }
 373                                 break;
 374
 375                         case Change::DELETED:
 376                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 377
 378                                 // Do NOT reject changes within a deleted inset!
 379                                 // There may be insertions of a co-author inside of it!
 380
 381                                 break;
 382                 }
 383         }
 384 }
 385
 386
 387 void Paragraph::Private::insertChar(pos_type pos, char_type c,
 388                 Change const & change)
 389 {
 390         LASSERT(pos >= 0 && pos <= int(text_.size()), /**/);
 391
 392         // track change
 393         changes_.insert(change, pos);
 394
 395         // This is actually very common when parsing buffers (and
 396         // maybe inserting ascii text)
 397         if (pos == pos_type(text_.size())) {
 398                 // when appending characters, no need to update tables
 399                 text_.push_back(c);
 400                 return;
 401         }
 402
 403         text_.insert(text_.begin() + pos, c);
 404
 405         // Update the font table.
 406         fontlist_.increasePosAfterPos(pos);
 407
 408         // Update the insets
 409         insetlist_.increasePosAfterPos(pos);
 410 }
 411
 412
 413 void Paragraph::insertInset(pos_type pos, Inset * inset,
 414                                    Change const & change)
 415 {
 416         LASSERT(inset, /**/);
 417         LASSERT(pos >= 0 && pos <= size(), /**/);
 418
 419         d->insertChar(pos, META_INSET, change);
 420         LASSERT(d->text_[pos] == META_INSET, /**/);
 421
 422         // Add a new entry in the insetlist_.
 423         d->insetlist_.insert(inset, pos);
 424 }
 425
 426
 427 bool Paragraph::eraseChar(pos_type pos, bool trackChanges)
 428 {
 429         LASSERT(pos >= 0 && pos <= size(), /**/);
 430
 431         // keep the logic here in sync with the logic of isMergedOnEndOfParDeletion()
 432
 433         if (trackChanges) {
 434                 Change change = d->changes_.lookup(pos);
 435
 436                 // set the character to DELETED if
 437                 //  a) it was previously unchanged or
 438                 //  b) it was inserted by a co-author
 439
 440                 if (change.type == Change::UNCHANGED ||
 441                     (change.type == Change::INSERTED && change.author != 0)) {
 442                         setChange(pos, Change(Change::DELETED));
 443                         return false;
 444                 }
 445
 446                 if (change.type == Change::DELETED)
 447                         return false;
 448         }
 449
 450         // Don't physically access the imaginary end-of-paragraph character.
 451         // eraseChar() can only mark it as DELETED. A physical deletion of
 452         // end-of-par must be handled externally.
 453         if (pos == size()) {
 454                 return false;
 455         }
 456
 457         // track change
 458         d->changes_.erase(pos);
 459
 460         // if it is an inset, delete the inset entry
 461         if (d->text_[pos] == META_INSET)
 462                 d->insetlist_.erase(pos);
 463
 464         d->text_.erase(d->text_.begin() + pos);
 465
 466         // Update the fontlist_
 467         d->fontlist_.erase(pos);
 468
 469         // Update the insetlist_
 470         d->insetlist_.decreasePosAfterPos(pos);
 471
 472         return true;
 473 }
 474
 475
 476 int Paragraph::eraseChars(pos_type start, pos_type end, bool trackChanges)
 477 {
 478         LASSERT(start >= 0 && start <= size(), /**/);
 479         LASSERT(end >= start && end <= size() + 1, /**/);
 480
 481         pos_type i = start;
 482         for (pos_type count = end - start; count; --count) {
 483                 if (!eraseChar(i, trackChanges))
 484                         ++i;
 485         }
 486         return end - i;
 487 }
 488
 489
 490 int Paragraph::Private::latexSurrogatePair(odocstream & os, char_type c,
 491                 char_type next, Encoding const & encoding)
 492 {
 493         // Writing next here may circumvent a possible font change between
 494         // c and next. Since next is only output if it forms a surrogate pair
 495         // with c we can ignore this:
 496         // A font change inside a surrogate pair does not make sense and is
 497         // hopefully impossible to input.
 498         // FIXME: change tracking
 499         // Is this correct WRT change tracking?
 500         docstring const latex1 = encoding.latexChar(next);
 501         docstring const latex2 = encoding.latexChar(c);
 502         os << latex1 << '{' << latex2 << '}';
 503         return latex1.length() + latex2.length() + 2;
 504 }
 505
 506
 507 bool Paragraph::Private::simpleTeXBlanks(OutputParams const & runparams,
 508                                        odocstream & os, TexRow & texrow,
 509                                        pos_type i,
 510                                        unsigned int & column,
 511                                        Font const & font,
 512                                        Layout const & style)
 513 {
 514         if (style.pass_thru || runparams.verbatim)
 515                 return false;
 516
 517         if (i + 1 < int(text_.size())) {
 518                 char_type next = text_[i + 1];
 519                 if (Encodings::isCombiningChar(next)) {
 520                         Encoding const & encoding = *(runparams.encoding);
 521                         // This space has an accent, so we must always output it.
 522                         column += latexSurrogatePair(os, ' ', next, encoding) - 1;
 523                         return true;
 524                 }
 525         }
 526
 527         if (lyxrc.plaintext_linelen > 0
 528             && column > lyxrc.plaintext_linelen
 529             && i
 530             && text_[i - 1] != ' '
 531             && (i + 1 < int(text_.size()))
 532             // same in FreeSpacing mode
 533             && !owner_->isFreeSpacing()
 534             // In typewriter mode, we want to avoid
 535             // ! . ? : at the end of a line
 536             && !(font.fontInfo().family() == TYPEWRITER_FAMILY
 537                  && (text_[i - 1] == '.'
 538                      || text_[i - 1] == '?'
 539                      || text_[i - 1] == ':'
 540                      || text_[i - 1] == '!'))) {
 541                 os << '\n';
 542                 texrow.newline();
 543                 texrow.start(owner_->id(), i + 1);
 544                 column = 0;
 545         } else if (style.free_spacing) {
 546                 os << '~';
 547         } else {
 548                 os << ' ';
 549         }
 550         return false;
 551 }
 552
 553
 554 int Paragraph::Private::writeScriptChars(odocstream & os,
 555                                          docstring const & ltx,
 556                                          Change & runningChange,
 557                                          Encoding const & encoding,
 558                                          pos_type & i)
 559 {
 560         // FIXME: modifying i here is not very nice...
 561
 562         // We only arrive here when a proper language for character text_[i] has
 563         // not been specified (i.e., it could not be translated in the current
 564         // latex encoding) and it belongs to a known script.
 565         // Parameter ltx contains the latex translation of text_[i] as specified in
 566         // the unicodesymbols file and is something like "\textXXX{<spec>}".
 567         // The latex macro name "textXXX" specifies the script to which text_[i]
 568         // belongs and we use it in order to check whether characters from the
 569         // same script immediately follow, such that we can collect them in a
 570         // single "\textXXX" macro. So, we have to retain "\textXXX{<spec>"
 571         // for the first char but only "<spec>" for all subsequent chars.
 572         docstring::size_type const brace1 = ltx.find_first_of(from_ascii("{"));
 573         docstring::size_type const brace2 = ltx.find_last_of(from_ascii("}"));
 574         string script = to_ascii(ltx.substr(1, brace1 - 1));
 575         int length = ltx.substr(0, brace2).length();
 576         os << ltx.substr(0, brace2);
 577         int size = text_.size();
 578         while (i + 1 < size) {
 579                 char_type const next = text_[i + 1];
 580                 // Stop here if next character belongs to another script
 581                 // or there is a change in change tracking status.
 582                 if (!Encodings::isKnownScriptChar(next, script) ||
 583                     runningChange != owner_->lookupChange(i + 1))
 584                         break;
 585                 Font prev_font;
 586                 bool found = false;
 587                 FontList::const_iterator cit = fontlist_.begin();
 588                 FontList::const_iterator end = fontlist_.end();
 589                 for (; cit != end; ++cit) {
 590                         if (cit->pos() >= i && !found) {
 591                                 prev_font = cit->font();
 592                                 found = true;
 593                         }
 594                         if (cit->pos() >= i + 1)
 595                                 break;
 596                 }
 597                 // Stop here if there is a font attribute or encoding change.
 598                 if (found && cit != end && prev_font != cit->font())
 599                         break;
 600                 docstring const latex = encoding.latexChar(next);
 601                 docstring::size_type const b1 =
 602                                         latex.find_first_of(from_ascii("{"));
 603                 docstring::size_type const b2 =
 604                                         latex.find_last_of(from_ascii("}"));
 605                 int const len = b2 - b1 - 1;
 606                 os << latex.substr(b1 + 1, len);
 607                 length += len;
 608                 ++i;
 609         }
 610         os << '}';
 611         ++length;
 612         return length;
 613 }
 614
 615
 616 bool Paragraph::Private::isTextAt(string const & str, pos_type pos) const
 617 {
 618         pos_type const len = str.length();
 619
 620         // is the paragraph large enough?
 621         if (pos + len > int(text_.size()))
 622                 return false;
 623
 624         // does the wanted text start at point?
 625         for (string::size_type i = 0; i < str.length(); ++i) {
 626                 // Caution: direct comparison of characters works only
 627                 // because str is pure ASCII.
 628                 if (str[i] != text_[pos + i])
 629                         return false;
 630         }
 631
 632         return fontlist_.hasChangeInRange(pos, len);
 633 }
 634
 635
 636 void Paragraph::Private::latexInset(
 637                                              BufferParams const & bparams,
 638                                              odocstream & os,
 639                                              TexRow & texrow,
 640                                              OutputParams & runparams,
 641                                              Font & running_font,
 642                                              Font & basefont,
 643                                              Font const & outerfont,
 644                                              bool & open_font,
 645                                              Change & running_change,
 646                                              Layout const & style,
 647                                              pos_type & i,
 648                                              unsigned int & column)
 649 {
 650         Inset * inset = owner_->getInset(i);
 651         LASSERT(inset, /**/);
 652
 653         if (style.pass_thru) {
 654                 inset->plaintext(os, runparams);
 655                 return;
 656         }
 657
 658         // FIXME: move this to InsetNewline::latex
 659         if (inset->lyxCode() == NEWLINE_CODE) {
 660                 // newlines are handled differently here than
 661                 // the default in simpleTeXSpecialChars().
 662                 if (!style.newline_allowed) {
 663                         os << '\n';
 664                 } else {
 665                         if (open_font) {
 666                                 column += running_font.latexWriteEndChanges(
 667                                         os, bparams, runparams,
 668                                         basefont, basefont);
 669                                 open_font = false;
 670                         }
 671
 672                         if (running_font.fontInfo().family() == TYPEWRITER_FAMILY)
 673                                 os << '~';
 674
 675                         basefont = owner_->getLayoutFont(bparams, outerfont);
 676                         running_font = basefont;
 677
 678                         if (runparams.moving_arg)
 679                                 os << "\\protect ";
 680
 681                 }
 682                 texrow.newline();
 683                 texrow.start(owner_->id(), i + 1);
 684                 column = 0;
 685         }
 686
 687         if (owner_->lookupChange(i).type == Change::DELETED) {
 688                 if( ++runparams.inDeletedInset == 1)
 689                         runparams.changeOfDeletedInset = owner_->lookupChange(i);
 690         }
 691
 692         if (inset->canTrackChanges()) {
 693                 column += Changes::latexMarkChange(os, bparams, running_change,
 694                         Change(Change::UNCHANGED));
 695                 running_change = Change(Change::UNCHANGED);
 696         }
 697
 698         bool close = false;
 699         odocstream::pos_type const len = os.tellp();
 700
 701         if (inset->forceLTR()
 702             && running_font.isRightToLeft()
 703                 // ERT is an exception, it should be output with no decorations at all
 704                 && inset->lyxCode() != ERT_CODE) {
 705                 if (running_font.language()->lang() == "farsi")
 706                         os << "\\beginL{}";
 707                 else
 708                         os << "\\L{";
 709                 close = true;
 710         }
 711
 712         // FIXME: Bug: we can have an empty font change here!
 713         // if there has just been a font change, we are going to close it
 714         // right now, which means stupid latex code like \textsf{}. AFAIK,
 715         // this does not harm dvi output. A minor bug, thus (JMarc)
 716
 717         // Some insets cannot be inside a font change command.
 718         // However, even such insets *can* be placed in \L or \R
 719         // or their equivalents (for RTL language switches), so we don't
 720         // close the language in those cases.
 721         // ArabTeX, though, cannot handle this special behavior, it seems.
 722         bool arabtex = basefont.language()->lang() == "arabic_arabtex"
 723                 || running_font.language()->lang() == "arabic_arabtex";
 724         if (open_font && inset->noFontChange()) {
 725                 bool closeLanguage = arabtex
 726                         || basefont.isRightToLeft() == running_font.isRightToLeft();
 727                 unsigned int count = running_font.latexWriteEndChanges(os,
 728                         bparams, runparams, basefont, basefont, closeLanguage);
 729                 column += count;
 730                 // if any font properties were closed, update the running_font,
 731                 // making sure, however, to leave the language as it was
 732                 if (count > 0) {
 733                         // FIXME: probably a better way to keep track of the old
 734                         // language, than copying the entire font?
 735                         Font const copy_font(running_font);
 736                         basefont = owner_->getLayoutFont(bparams, outerfont);
 737                         running_font = basefont;
 738                         if (!closeLanguage)
 739                                 running_font.setLanguage(copy_font.language());
 740                         // leave font open if language is still open
 741                         open_font = (running_font.language() == basefont.language());
 742                         if (closeLanguage)
 743                                 runparams.local_font = &basefont;
 744                 }
 745         }
 746
 747         int tmp = inset->latex(os, runparams);
 748
 749         if (close) {
 750         if (running_font.language()->lang() == "farsi")
 751                         os << "\\endL{}";
 752                 else
 753                         os << '}';
 754         }
 755
 756         if (tmp) {
 757                 for (int j = 0; j < tmp; ++j)
 758                         texrow.newline();
 759
 760                 texrow.start(owner_->id(), i + 1);
 761                 column = 0;
 762         } else {
 763                 column += os.tellp() - len;
 764         }
 765
 766         if (owner_->lookupChange(i).type == Change::DELETED)
 767                 --runparams.inDeletedInset;
 768 }
 769
 770
 771 void Paragraph::Private::latexSpecialChar(
 772                                              odocstream & os,
 773                                              OutputParams & runparams,
 774                                              Font & running_font,
 775                                              Change & running_change,
 776                                              Layout const & style,
 777                                              pos_type & i,
 778                                              unsigned int & column)
 779 {
 780         char_type const c = text_[i];
 781
 782         if (style.pass_thru) {
 783                 if (c != '\0')
 784                         // FIXME UNICODE: This can fail if c cannot
 785                         // be encoded in the current encoding.
 786                         os.put(c);
 787                 return;
 788         }
 789
 790         if (runparams.verbatim) {
 791                 os.put(c);
 792                 return;
 793         }
 794
 795         if (lyxrc.fontenc == "T1" && latexSpecialT1(c, os, i, column))
 796                 return;
 797
 798         if (running_font.fontInfo().family() == TYPEWRITER_FAMILY
 799                 && latexSpecialTypewriter(c, os, i, column))
 800                 return;
 801
 802         // Otherwise, we use what LaTeX provides us.
 803         switch (c) {
 804         case '\\':
 805                 os << "\\textbackslash{}";
 806                 column += 15;
 807                 break;
 808         case '<':
 809                 os << "\\textless{}";
 810                 column += 10;
 811                 break;
 812         case '>':
 813                 os << "\\textgreater{}";
 814                 column += 13;
 815                 break;
 816         case '|':
 817                 os << "\\textbar{}";
 818                 column += 9;
 819                 break;
 820         case '-':
 821                 os << '-';
 822                 break;
 823         case '\"':
 824                 os << "\\char`\\\"{}";
 825                 column += 9;
 826                 break;
 827
 828         case '$': case '&':
 829         case '%': case '#': case '{':
 830         case '}': case '_':
 831                 os << '\\';
 832                 os.put(c);
 833                 column += 1;
 834                 break;
 835
 836         case '~':
 837                 os << "\\textasciitilde{}";
 838                 column += 16;
 839                 break;
 840
 841         case '^':
 842                 os << "\\textasciicircum{}";
 843                 column += 17;
 844                 break;
 845
 846         case '*': case '[':
 847                 // avoid being mistaken for optional arguments
 848                 os << '{';
 849                 os.put(c);
 850                 os << '}';
 851                 column += 2;
 852                 break;
 853
 854         case ' ':
 855                 // Blanks are printed before font switching.
 856                 // Sure? I am not! (try nice-latex)
 857                 // I am sure it's correct. LyX might be smarter
 858                 // in the future, but for now, nothing wrong is
 859                 // written. (Asger)
 860                 break;
 861
 862         default:
 863
 864                 // LyX, LaTeX etc.
 865                 if (latexSpecialPhrase(os, i, column, runparams))
 866                         return;
 867
 868                 if (c == '\0')
 869                         return;
 870
 871                 Encoding const & encoding = *(runparams.encoding);
 872                 if (i + 1 < int(text_.size())) {
 873                         char_type next = text_[i + 1];
 874                         if (Encodings::isCombiningChar(next)) {
 875                                 column += latexSurrogatePair(os, c, next, encoding) - 1;
 876                                 ++i;
 877                                 break;
 878                         }
 879                 }
 880                 string script;
 881                 docstring const latex = encoding.latexChar(c);
 882                 if (Encodings::isKnownScriptChar(c, script)
 883                     && prefixIs(latex, from_ascii("\\" + script)))
 884                         column += writeScriptChars(os, latex,
 885                                         running_change, encoding, i) - 1;
 886                 else if (latex.length() > 1 && latex[latex.length() - 1] != '}') {
 887                         // Prevent eating of a following
 888                         // space or command corruption by
 889                         // following characters
 890                         column += latex.length() + 1;
 891                         os << latex << "{}";
 892                 } else {
 893                         column += latex.length() - 1;
 894                         os << latex;
 895                 }
 896                 break;
 897         }
 898 }
 899
 900
 901 bool Paragraph::Private::latexSpecialT1(char_type const c, odocstream & os,
 902         pos_type & i, unsigned int & column)
 903 {
 904         switch (c) {
 905         case '>':
 906         case '<':
 907                 os.put(c);
 908                 // In T1 encoding, these characters exist
 909                 // but we should avoid ligatures
 910                 if (i + 1 >= int(text_.size()) || text_[i + 1] != c)
 911                         return true;
 912                 os << "\\,{}";
 913                 column += 3;
 914                 // Alternative code:
 915                 //os << "\\textcompwordmark{}";
 916                 //column += 19;
 917                 return true;
 918         case '|':
 919                 os.put(c);
 920                 return true;
 921         default:
 922                 return false;
 923         }
 924 }
 925
 926
 927 bool Paragraph::Private::latexSpecialTypewriter(char_type const c, odocstream & os,
 928         pos_type & i, unsigned int & column)
 929 {
 930         switch (c) {
 931         case '-':
 932                 if (i + 1 < int(text_.size()) && text_[i + 1] == '-') {
 933                         // "--" in Typewriter mode -> "-{}-"
 934                         os << "-{}";
 935                         column += 2;
 936                 } else
 937                         os << '-';
 938                 return true;
 939
 940         // I assume this is hack treating typewriter as verbatim
 941         // FIXME UNICODE: This can fail if c cannot be encoded
 942         // in the current encoding.
 943
 944         case '\0':
 945                 return true;
 946
 947         // Those characters are not directly supported.
 948         case '\\':
 949         case '\"':
 950         case '$': case '&':
 951         case '%': case '#': case '{':
 952         case '}': case '_':
 953         case '~':
 954         case '^':
 955         case '*': case '[':
 956         case ' ':
 957                 return false;
 958
 959         default:
 960                 // With Typewriter font, these characters exist.
 961                 os.put(c);
 962                 return true;
 963         }
 964 }
 965
 966
 967 bool Paragraph::Private::latexSpecialPhrase(odocstream & os, pos_type & i,
 968         unsigned int & column, OutputParams & runparams)
 969 {
 970         // FIXME: if we have "LaTeX" with a font
 971         // change in the middle (before the 'T', then
 972         // the "TeX" part is still special cased.
 973         // Really we should only operate this on
 974         // "words" for some definition of word
 975
 976         for (size_t pnr = 0; pnr < phrases_nr; ++pnr) {
 977                 if (!isTextAt(special_phrases[pnr].phrase, i))
 978                         continue;
 979                 if (runparams.moving_arg)
 980                         os << "\\protect";
 981                 os << special_phrases[pnr].macro;
 982                 i += special_phrases[pnr].phrase.length() - 1;
 983                 column += special_phrases[pnr].macro.length() - 1;
 984                 return true;
 985         }
 986         return false;
 987 }
 988
 989
 990 void Paragraph::Private::validate(LaTeXFeatures & features,
 991                                 Layout const & layout) const
 992 {
 993         // check the params.
 994         if (!params_.spacing().isDefault())
 995                 features.require("setspace");
 996
 997         // then the layouts
 998         features.useLayout(layout.name());
 999
1000         // then the fonts
1001         fontlist_.validate(features);
1002
1003         // then the indentation
1004         if (!params_.leftIndent().zero())
1005                 features.require("ParagraphLeftIndent");
1006
1007         // then the insets
1008         InsetList::const_iterator icit = insetlist_.begin();
1009         InsetList::const_iterator iend = insetlist_.end();
1010         for (; icit != iend; ++icit) {
1011                 if (icit->inset) {
1012                         icit->inset->validate(features);
1013                         if (layout.needprotect &&
1014                             icit->inset->lyxCode() == FOOT_CODE)
1015                                 features.require("NeedLyXFootnoteCode");
1016                 }
1017         }
1018
1019         // then the contents
1020         for (pos_type i = 0; i < int(text_.size()) ; ++i) {
1021                 for (size_t pnr = 0; pnr < phrases_nr; ++pnr) {
1022                         if (!special_phrases[pnr].builtin
1023                             && isTextAt(special_phrases[pnr].phrase, i)) {
1024                                 features.require(special_phrases[pnr].phrase);
1025                                 break;
1026                         }
1027                 }
1028                 Encodings::validate(text_[i], features);
1029         }
1030 }
1031
1032 /////////////////////////////////////////////////////////////////////
1033 //
1034 // Paragraph
1035 //
1036 /////////////////////////////////////////////////////////////////////
1037
1038 namespace {
1039         Layout const emptyParagraphLayout;
1040 }
1041
1042 Paragraph::Paragraph()
1043         : d(new Paragraph::Private(this, emptyParagraphLayout))
1044 {
1045         itemdepth = 0;
1046         d->params_.clear();
1047 }
1048
1049
1050 Paragraph::Paragraph(Paragraph const & par)
1051         : itemdepth(par.itemdepth),
1052         d(new Paragraph::Private(*par.d, this))
1053 {
1054         registerWords();
1055 }
1056
1057
1058 Paragraph & Paragraph::operator=(Paragraph const & par)
1059 {
1060         // needed as we will destroy the private part before copying it
1061         if (&par != this) {
1062                 itemdepth = par.itemdepth;
1063
1064                 deregisterWords();
1065                 delete d;
1066                 d = new Private(*par.d, this);
1067                 registerWords();
1068         }
1069         return *this;
1070 }
1071
1072
1073 Paragraph::~Paragraph()
1074 {
1075         deregisterWords();
1076         delete d;
1077 }
1078
1079
1080 void Paragraph::write(ostream & os, BufferParams const & bparams,
1081         depth_type & dth) const
1082 {
1083         // The beginning or end of a deeper (i.e. nested) area?
1084         if (dth != d->params_.depth()) {
1085                 if (d->params_.depth() > dth) {
1086                         while (d->params_.depth() > dth) {
1087                                 os << "\n\\begin_deeper";
1088                                 ++dth;
1089                         }
1090                 } else {
1091                         while (d->params_.depth() < dth) {
1092                                 os << "\n\\end_deeper";
1093                                 --dth;
1094                         }
1095                 }
1096         }
1097
1098         // First write the layout
1099         os << "\n\\begin_layout " << to_utf8(d->layout_->name()) << '\n';
1100
1101         d->params_.write(os);
1102
1103         Font font1(inherit_font, bparams.language);
1104
1105         Change running_change = Change(Change::UNCHANGED);
1106
1107         int column = 0;
1108         for (pos_type i = 0; i <= size(); ++i) {
1109
1110                 Change change = lookupChange(i);
1111                 Changes::lyxMarkChange(os, column, running_change, change);
1112                 running_change = change;
1113
1114                 if (i == size())
1115                         break;
1116
1117                 // Write font changes
1118                 Font font2 = getFontSettings(bparams, i);
1119                 if (font2 != font1) {
1120                         font2.lyxWriteChanges(font1, os);
1121                         column = 0;
1122                         font1 = font2;
1123                 }
1124
1125                 char_type const c = d->text_[i];
1126                 switch (c) {
1127                 case META_INSET:
1128                         if (Inset const * inset = getInset(i)) {
1129                                 if (inset->directWrite()) {
1130                                         // international char, let it write
1131                                         // code directly so it's shorter in
1132                                         // the file
1133                                         inset->write(os);
1134                                 } else {
1135                                         if (i)
1136                                                 os << '\n';
1137                                         os << "\\begin_inset ";
1138                                         inset->write(os);
1139                                         os << "\n\\end_inset\n\n";
1140                                         column = 0;
1141                                 }
1142                         }
1143                         break;
1144                 case '\\':
1145                         os << "\n\\backslash\n";
1146                         column = 0;
1147                         break;
1148                 case '.':
1149                         if (i + 1 < size() && d->text_[i + 1] == ' ') {
1150                                 os << ".\n";
1151                                 column = 0;
1152                         } else
1153                                 os << '.';
1154                         break;
1155                 default:
1156                         if ((column > 70 && c == ' ')
1157                             || column > 79) {
1158                                 os << '\n';
1159                                 column = 0;
1160                         }
1161                         // this check is to amend a bug. LyX sometimes
1162                         // inserts '\0' this could cause problems.
1163                         if (c != '\0')
1164                                 os << to_utf8(docstring(1, c));
1165                         else
1166                                 LYXERR0("NUL char in structure.");
1167                         ++column;
1168                         break;
1169                 }
1170         }
1171
1172         os << "\n\\end_layout\n";
1173 }
1174
1175
1176 void Paragraph::validate(LaTeXFeatures & features) const
1177 {
1178         d->validate(features, *d->layout_);
1179 }
1180
1181
1182 void Paragraph::insert(pos_type start, docstring const & str,
1183                        Font const & font, Change const & change)
1184 {
1185         for (size_t i = 0, n = str.size(); i != n ; ++i)
1186                 insertChar(start + i, str[i], font, change);
1187 }
1188
1189
1190 void Paragraph::appendChar(char_type c, Font const & font,
1191                 Change const & change)
1192 {
1193         // track change
1194         d->changes_.insert(change, d->text_.size());
1195         // when appending characters, no need to update tables
1196         d->text_.push_back(c);
1197         setFont(d->text_.size() - 1, font);
1198 }
1199
1200
1201 void Paragraph::appendString(docstring const & s, Font const & font,
1202                 Change const & change)
1203 {
1204         pos_type end = s.size();
1205         size_t oldsize = d->text_.size();
1206         size_t newsize = oldsize + end;
1207         size_t capacity = d->text_.capacity();
1208         if (newsize >= capacity)
1209                 d->text_.reserve(max(capacity + 100, newsize));
1210
1211         // when appending characters, no need to update tables
1212         d->text_.append(s);
1213
1214         // FIXME: Optimize this!
1215         for (pos_type i = oldsize; i != newsize; ++i) {
1216                 // track change
1217                 d->changes_.insert(change, i);
1218         }
1219         d->fontlist_.set(oldsize, font);
1220         d->fontlist_.set(newsize - 1, font);
1221 }
1222
1223
1224 void Paragraph::insertChar(pos_type pos, char_type c,
1225                            bool trackChanges)
1226 {
1227         d->insertChar(pos, c, Change(trackChanges ?
1228                            Change::INSERTED : Change::UNCHANGED));
1229 }
1230
1231
1232 void Paragraph::insertChar(pos_type pos, char_type c,
1233                            Font const & font, bool trackChanges)
1234 {
1235         d->insertChar(pos, c, Change(trackChanges ?
1236                            Change::INSERTED : Change::UNCHANGED));
1237         setFont(pos, font);
1238 }
1239
1240
1241 void Paragraph::insertChar(pos_type pos, char_type c,
1242                            Font const & font, Change const & change)
1243 {
1244         d->insertChar(pos, c, change);
1245         setFont(pos, font);
1246 }
1247
1248
1249 void Paragraph::insertInset(pos_type pos, Inset * inset,
1250                             Font const & font, Change const & change)
1251 {
1252         insertInset(pos, inset, change);
1253         // Set the font/language of the inset...
1254         setFont(pos, font);
1255 }
1256
1257
1258 bool Paragraph::insetAllowed(InsetCode code)
1259 {
1260         return !d->inset_owner_ || d->inset_owner_->insetAllowed(code);
1261 }
1262
1263
1264 void Paragraph::resetFonts(Font const & font)
1265 {
1266         d->fontlist_.clear();
1267         d->fontlist_.set(0, font);
1268         d->fontlist_.set(d->text_.size() - 1, font);
1269 }
1270
1271 // Gets uninstantiated font setting at position.
1272 Font const Paragraph::getFontSettings(BufferParams const & bparams,
1273                                          pos_type pos) const
1274 {
1275         if (pos > size()) {
1276                 LYXERR0("pos: " << pos << " size: " << size());
1277                 LASSERT(pos <= size(), /**/);
1278         }
1279
1280         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
1281         if (cit != d->fontlist_.end())
1282                 return cit->font();
1283
1284         if (pos == size() && !empty())
1285                 return getFontSettings(bparams, pos - 1);
1286
1287         return Font(inherit_font, getParLanguage(bparams));
1288 }
1289
1290
1291 FontSpan Paragraph::fontSpan(pos_type pos) const
1292 {
1293         LASSERT(pos <= size(), /**/);
1294         pos_type start = 0;
1295
1296         FontList::const_iterator cit = d->fontlist_.begin();
1297         FontList::const_iterator end = d->fontlist_.end();
1298         for (; cit != end; ++cit) {
1299                 if (cit->pos() >= pos) {
1300                         if (pos >= beginOfBody())
1301                                 return FontSpan(max(start, beginOfBody()),
1302                                                 cit->pos());
1303                         else
1304                                 return FontSpan(start,
1305                                                 min(beginOfBody() - 1,
1306                                                          cit->pos()));
1307                 }
1308                 start = cit->pos() + 1;
1309         }
1310
1311         // This should not happen, but if so, we take no chances.
1312         // LYXERR0("Paragraph::getEndPosOfFontSpan: This should not happen!");
1313         return FontSpan(pos, pos);
1314 }
1315
1316
1317 // Gets uninstantiated font setting at position 0
1318 Font const Paragraph::getFirstFontSettings(BufferParams const & bparams) const
1319 {
1320         if (!empty() && !d->fontlist_.empty())
1321                 return d->fontlist_.begin()->font();
1322
1323         return Font(inherit_font, bparams.language);
1324 }
1325
1326
1327 // Gets the fully instantiated font at a given position in a paragraph
1328 // This is basically the same function as Text::GetFont() in text2.cpp.
1329 // The difference is that this one is used for generating the LaTeX file,
1330 // and thus cosmetic "improvements" are disallowed: This has to deliver
1331 // the true picture of the buffer. (Asger)
1332 Font const Paragraph::getFont(BufferParams const & bparams, pos_type pos,
1333                                  Font const & outerfont) const
1334 {
1335         LASSERT(pos >= 0, /**/);
1336
1337         Font font = getFontSettings(bparams, pos);
1338
1339         pos_type const body_pos = beginOfBody();
1340         if (pos < body_pos)
1341                 font.fontInfo().realize(d->layout_->labelfont);
1342         else
1343                 font.fontInfo().realize(d->layout_->font);
1344
1345         font.fontInfo().realize(outerfont.fontInfo());
1346         font.fontInfo().realize(bparams.getFont().fontInfo());
1347
1348         return font;
1349 }
1350
1351
1352 Font const Paragraph::getLabelFont
1353         (BufferParams const & bparams, Font const & outerfont) const
1354 {
1355         FontInfo tmpfont = d->layout_->labelfont;
1356         tmpfont.realize(outerfont.fontInfo());
1357         tmpfont.realize(bparams.getFont().fontInfo());
1358         return Font(tmpfont, getParLanguage(bparams));
1359 }
1360
1361
1362 Font const Paragraph::getLayoutFont
1363         (BufferParams const & bparams, Font const & outerfont) const
1364 {
1365         FontInfo tmpfont = d->layout_->font;
1366         tmpfont.realize(outerfont.fontInfo());
1367         tmpfont.realize(bparams.getFont().fontInfo());
1368         return Font(tmpfont, getParLanguage(bparams));
1369 }
1370
1371
1372 /// Returns the height of the highest font in range
1373 FontSize Paragraph::highestFontInRange
1374         (pos_type startpos, pos_type endpos, FontSize def_size) const
1375 {
1376         return d->fontlist_.highestInRange(startpos, endpos, def_size);
1377 }
1378
1379
1380 char_type Paragraph::getUChar(BufferParams const & bparams, pos_type pos) const
1381 {
1382         char_type c = d->text_[pos];
1383         if (!lyxrc.rtl_support)
1384                 return c;
1385
1386         char_type uc = c;
1387         switch (c) {
1388         case '(':
1389                 uc = ')';
1390                 break;
1391         case ')':
1392                 uc = '(';
1393                 break;
1394         case '[':
1395                 uc = ']';
1396                 break;
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         }
1413         if (uc != c && getFontSettings(bparams, pos).isRightToLeft())
1414                 return uc;
1415         return c;
1416 }
1417
1418
1419 void Paragraph::setFont(pos_type pos, Font const & font)
1420 {
1421         LASSERT(pos <= size(), /**/);
1422
1423         // First, reduce font against layout/label font
1424         // Update: The setCharFont() routine in text2.cpp already
1425         // reduces font, so we don't need to do that here. (Asger)
1426
1427         d->fontlist_.set(pos, font);
1428 }
1429
1430
1431 void Paragraph::makeSameLayout(Paragraph const & par)
1432 {
1433         d->layout_ = par.d->layout_;
1434         d->params_ = par.d->params_;
1435 }
1436
1437
1438 bool Paragraph::stripLeadingSpaces(bool trackChanges)
1439 {
1440         if (isFreeSpacing())
1441                 return false;
1442
1443         int pos = 0;
1444         int count = 0;
1445
1446         while (pos < size() && (isNewline(pos) || isLineSeparator(pos))) {
1447                 if (eraseChar(pos, trackChanges))
1448                         ++count;
1449                 else
1450                         ++pos;
1451         }
1452
1453         return count > 0 || pos > 0;
1454 }
1455
1456
1457 bool Paragraph::hasSameLayout(Paragraph const & par) const
1458 {
1459         return par.d->layout_ == d->layout_
1460                 && d->params_.sameLayout(par.d->params_);
1461 }
1462
1463
1464 depth_type Paragraph::getDepth() const
1465 {
1466         return d->params_.depth();
1467 }
1468
1469
1470 depth_type Paragraph::getMaxDepthAfter() const
1471 {
1472         if (d->layout_->isEnvironment())
1473                 return d->params_.depth() + 1;
1474         else
1475                 return d->params_.depth();
1476 }
1477
1478
1479 char Paragraph::getAlign() const
1480 {
1481         if (d->params_.align() == LYX_ALIGN_LAYOUT)
1482                 return d->layout_->align;
1483         else
1484                 return d->params_.align();
1485 }
1486
1487
1488 docstring const & Paragraph::labelString() const
1489 {
1490         return d->params_.labelString();
1491 }
1492
1493
1494 // the next two functions are for the manual labels
1495 docstring const Paragraph::getLabelWidthString() const
1496 {
1497         if (d->layout_->margintype == MARGIN_MANUAL)
1498                 return d->params_.labelWidthString();
1499         else
1500                 return _("Senseless with this layout!");
1501 }
1502
1503
1504 void Paragraph::setLabelWidthString(docstring const & s)
1505 {
1506         d->params_.labelWidthString(s);
1507 }
1508
1509
1510 docstring const Paragraph::translateIfPossible(docstring const & s,
1511                 BufferParams const & bparams) const
1512 {
1513         if (!isAscii(s) || s.empty()) {
1514                 // This must be a user defined layout. We cannot translate
1515                 // this, since gettext accepts only ascii keys.
1516                 return s;
1517         }
1518         // Probably standard layout, try to translate
1519         Messages & m = getMessages(getParLanguage(bparams)->code());
1520         return m.get(to_ascii(s));
1521 }
1522
1523
1524 docstring Paragraph::expandLabel(Layout const & layout,
1525                 BufferParams const & bparams, bool process_appendix) const
1526 {
1527         DocumentClass const & tclass = bparams.documentClass();
1528
1529         docstring fmt;
1530         if (process_appendix && d->params_.appendix())
1531                 fmt = translateIfPossible(layout.labelstring_appendix(),
1532                         bparams);
1533         else
1534                 fmt = translateIfPossible(layout.labelstring(), bparams);
1535
1536         if (fmt.empty() && layout.labeltype == LABEL_COUNTER
1537             && !layout.counter.empty())
1538                 fmt = "\\the" + layout.counter;
1539
1540         // handle 'inherited level parts' in 'fmt',
1541         // i.e. the stuff between '@' in   '@Section@.\arabic{subsection}'
1542         size_t const i = fmt.find('@', 0);
1543         if (i != docstring::npos) {
1544                 size_t const j = fmt.find('@', i + 1);
1545                 if (j != docstring::npos) {
1546                         docstring parent(fmt, i + 1, j - i - 1);
1547                         docstring label = from_ascii("??");
1548                         if (tclass.hasLayout(parent))
1549                                 docstring label = expandLabel(tclass[parent], bparams,
1550                                                       process_appendix);
1551                         fmt = docstring(fmt, 0, i) + label
1552                                 + docstring(fmt, j + 1, docstring::npos);
1553                 }
1554         }
1555
1556         return tclass.counters().counterLabel(fmt);
1557 }
1558
1559
1560 void Paragraph::applyLayout(Layout const & new_layout)
1561 {
1562         d->layout_ = &new_layout;
1563         LyXAlignment const oldAlign = d->params_.align();
1564
1565         if (!(oldAlign & d->layout_->alignpossible)) {
1566                 frontend::Alert::warning(_("Alignment not permitted"),
1567                         _("The new layout does not permit the alignment previously used.\nSetting to default."));
1568                 d->params_.align(LYX_ALIGN_LAYOUT);
1569         }
1570 }
1571
1572
1573 pos_type Paragraph::beginOfBody() const
1574 {
1575         return d->begin_of_body_;
1576 }
1577
1578
1579 void Paragraph::setBeginOfBody()
1580 {
1581         if (d->layout_->labeltype != LABEL_MANUAL) {
1582                 d->begin_of_body_ = 0;
1583                 return;
1584         }
1585
1586         // Unroll the first two cycles of the loop
1587         // and remember the previous character to
1588         // remove unnecessary getChar() calls
1589         pos_type i = 0;
1590         pos_type end = size();
1591         if (i < end && !isNewline(i)) {
1592                 ++i;
1593                 char_type previous_char = 0;
1594                 char_type temp = 0;
1595                 if (i < end) {
1596                         previous_char = d->text_[i];
1597                         if (!isNewline(i)) {
1598                                 ++i;
1599                                 while (i < end && previous_char != ' ') {
1600                                         temp = d->text_[i];
1601                                         if (isNewline(i))
1602                                                 break;
1603                                         ++i;
1604                                         previous_char = temp;
1605                                 }
1606                         }
1607                 }
1608         }
1609
1610         d->begin_of_body_ = i;
1611 }
1612
1613
1614 bool Paragraph::forceEmptyLayout() const
1615 {
1616         Inset const * const inset = inInset();
1617         if (!inset)
1618                 return true;
1619         return inset->forceEmptyLayout();
1620 }
1621
1622
1623 bool Paragraph::allowParagraphCustomization() const
1624 {
1625         Inset const * const inset = inInset();
1626         if (!inset)
1627                 return true;
1628         return inset->allowParagraphCustomization();
1629 }
1630
1631
1632 bool Paragraph::useEmptyLayout() const
1633 {
1634         Inset const * const inset = inInset();
1635         if (!inset)
1636                 return false;
1637         return inset->useEmptyLayout();
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(BufferParams const & bparams,
1823                                 Font const & outerfont,
1824                                 odocstream & os, TexRow & texrow,
1825                                 OutputParams const & runparams) const
1826 {
1827         LYXERR(Debug::LATEX, "SimpleTeXOnePar...     " << this);
1828
1829         bool return_value = false;
1830
1831         bool asdefault = forceEmptyLayout();
1832
1833         Layout const & style = asdefault ?
1834                 bparams.documentClass().emptyLayout() :
1835                 *d->layout_;
1836
1837         // Current base font for all inherited font changes, without any
1838         // change caused by an individual character, except for the language:
1839         // It is set to the language of the first character.
1840         // As long as we are in the label, this font is the base font of the
1841         // label. Before the first body character it is set to the base font
1842         // of the body.
1843         Font basefont;
1844
1845         // Maybe we have to create a optional argument.
1846         pos_type body_pos = beginOfBody();
1847         unsigned int column = 0;
1848
1849         if (body_pos > 0) {
1850                 // the optional argument is kept in curly brackets in
1851                 // case it contains a ']'
1852                 os << "[{";
1853                 column += 2;
1854                 basefont = getLabelFont(bparams, outerfont);
1855         } else {
1856                 basefont = getLayoutFont(bparams, outerfont);
1857         }
1858
1859         // Which font is currently active?
1860         Font running_font(basefont);
1861         // Do we have an open font change?
1862         bool open_font = false;
1863
1864         Change runningChange = Change(Change::UNCHANGED);
1865
1866         texrow.start(id(), 0);
1867
1868         // if the paragraph is empty, the loop will not be entered at all
1869         if (empty()) {
1870                 if (style.isCommand()) {
1871                         os << '{';
1872                         ++column;
1873                 }
1874                 if (!asdefault)
1875                         column += d->startTeXParParams(bparams, os, texrow,
1876                                                     runparams.moving_arg);
1877         }
1878
1879         for (pos_type i = 0; i < size(); ++i) {
1880                 // First char in paragraph or after label?
1881                 if (i == body_pos) {
1882                         if (body_pos > 0) {
1883                                 if (open_font) {
1884                                         column += running_font.latexWriteEndChanges(
1885                                                 os, bparams, runparams,
1886                                                 basefont, basefont);
1887                                         open_font = false;
1888                                 }
1889                                 basefont = getLayoutFont(bparams, outerfont);
1890                                 running_font = basefont;
1891
1892                                 column += Changes::latexMarkChange(os, bparams,
1893                                                 runningChange, Change(Change::UNCHANGED));
1894                                 runningChange = Change(Change::UNCHANGED);
1895
1896                                 os << "}] ";
1897                                 column +=3;
1898                         }
1899                         if (style.isCommand()) {
1900                                 os << '{';
1901                                 ++column;
1902                         }
1903
1904                         if (!asdefault)
1905                                 column += d->startTeXParParams(bparams, os,
1906                                                             texrow,
1907                                                             runparams.moving_arg);
1908                 }
1909
1910                 Change const & change = runparams.inDeletedInset ? runparams.changeOfDeletedInset
1911                                                                  : lookupChange(i);
1912
1913                 if (bparams.outputChanges && runningChange != change) {
1914                         if (open_font) {
1915                                 column += running_font.latexWriteEndChanges(
1916                                                 os, bparams, runparams, basefont, basefont);
1917                                 open_font = false;
1918                         }
1919                         basefont = getLayoutFont(bparams, outerfont);
1920                         running_font = basefont;
1921
1922                         column += Changes::latexMarkChange(os, bparams, runningChange, change);
1923                         runningChange = change;
1924                 }
1925
1926                 // do not output text which is marked deleted
1927                 // if change tracking output is disabled
1928                 if (!bparams.outputChanges && change.type == Change::DELETED) {
1929                         continue;
1930                 }
1931
1932                 ++column;
1933
1934                 // Fully instantiated font
1935                 Font const font = getFont(bparams, i, outerfont);
1936
1937                 Font const last_font = running_font;
1938
1939                 // Do we need to close the previous font?
1940                 if (open_font &&
1941                     (font != running_font ||
1942                      font.language() != running_font.language()))
1943                 {
1944                         column += running_font.latexWriteEndChanges(
1945                                         os, bparams, runparams, basefont,
1946                                         (i == body_pos-1) ? basefont : font);
1947                         running_font = basefont;
1948                         open_font = false;
1949                 }
1950
1951                 // close babel's font environment before opening CJK.
1952                 if (!running_font.language()->babel().empty() &&
1953                     font.language()->encoding()->package() == Encoding::CJK) {
1954                                 string end_tag = subst(lyxrc.language_command_end,
1955                                                         "$$lang",
1956                                                         running_font.language()->babel());
1957                                 os << from_ascii(end_tag);
1958                                 column += end_tag.length();
1959                 }
1960
1961                 // Switch file encoding if necessary (and allowed)
1962                 if (!runparams.verbatim &&
1963                     runparams.encoding->package() == Encoding::none &&
1964                     font.language()->encoding()->package() == Encoding::none) {
1965                         pair<bool, int> const enc_switch = switchEncoding(os, bparams,
1966                                         runparams, *(font.language()->encoding()));
1967                         if (enc_switch.first) {
1968                                 column += enc_switch.second;
1969                                 runparams.encoding = font.language()->encoding();
1970                         }
1971                 }
1972
1973                 char_type const c = d->text_[i];
1974
1975                 // Do we need to change font?
1976                 if ((font != running_font ||
1977                      font.language() != running_font.language()) &&
1978                         i != body_pos - 1)
1979                 {
1980                         odocstringstream ods;
1981                         column += font.latexWriteStartChanges(ods, bparams,
1982                                                               runparams, basefont,
1983                                                               last_font);
1984                         running_font = font;
1985                         open_font = true;
1986                         docstring fontchange = ods.str();
1987                         // check if the fontchange ends with a trailing blank
1988                         // (like "\small " (see bug 3382)
1989                         if (suffixIs(fontchange, ' ') && c == ' ')
1990                                 os << fontchange.substr(0, fontchange.size() - 1)
1991                                    << from_ascii("{}");
1992                         else
1993                                 os << fontchange;
1994                 }
1995
1996                 if (c == ' ') {
1997                         // FIXME: integrate this case in latexSpecialChar
1998                         // Do not print the separation of the optional argument
1999                         // if style.pass_thru is false. This works because
2000                         // latexSpecialChar ignores spaces if
2001                         // style.pass_thru is false.
2002                         if (i != body_pos - 1) {
2003                                 if (d->simpleTeXBlanks(
2004                                                 runparams, os, texrow,
2005                                                 i, column, font, style)) {
2006                                         // A surrogate pair was output. We
2007                                         // must not call latexSpecialChar
2008                                         // in this iteration, since it would output
2009                                         // the combining character again.
2010                                         ++i;
2011                                         continue;
2012                                 }
2013                         }
2014                 }
2015
2016                 OutputParams rp = runparams;
2017                 rp.free_spacing = style.free_spacing;
2018                 rp.local_font = &font;
2019                 rp.intitle = style.intitle;
2020
2021                 // Two major modes:  LaTeX or plain
2022                 // Handle here those cases common to both modes
2023                 // and then split to handle the two modes separately.
2024                 if (c == META_INSET)
2025                         d->latexInset(bparams, os,
2026                                         texrow, rp, running_font,
2027                                         basefont, outerfont, open_font,
2028                                         runningChange, style, i, column);
2029                 else {
2030                         try {
2031                                 d->latexSpecialChar(os, rp, running_font, runningChange,
2032                                         style, i, column);
2033                         } catch (EncodingException & e) {
2034                                 if (runparams.dryrun) {
2035                                         os << "<" << _("LyX Warning: ")
2036                                            << _("uncodable character") << " '";
2037                                         os.put(c);
2038                                         os << "'>";
2039                                 } else {
2040                                         // add location information and throw again.
2041                                         e.par_id = id();
2042                                         e.pos = i;
2043                                         throw(e);
2044                                 }
2045                         }
2046                 }
2047
2048                 // Set the encoding to that returned from simpleTeXSpecialChars (see
2049                 // comment for encoding member in OutputParams.h)
2050                 runparams.encoding = rp.encoding;
2051         }
2052
2053         // If we have an open font definition, we have to close it
2054         if (open_font) {
2055 #ifdef FIXED_LANGUAGE_END_DETECTION
2056                 if (next_) {
2057                         running_font
2058                                 .latexWriteEndChanges(os, bparams, runparams,
2059                                         basefont,
2060                                         next_->getFont(bparams, 0, outerfont));
2061                 } else {
2062                         running_font.latexWriteEndChanges(os, bparams,
2063                                         runparams, basefont, basefont);
2064                 }
2065 #else
2066 //FIXME: For now we ALWAYS have to close the foreign font settings if they are
2067 //FIXME: there as we start another \selectlanguage with the next paragraph if
2068 //FIXME: we are in need of this. This should be fixed sometime (Jug)
2069                 running_font.latexWriteEndChanges(os, bparams, runparams,
2070                                 basefont, basefont);
2071 #endif
2072         }
2073
2074         column += Changes::latexMarkChange(os, bparams, runningChange, Change(Change::UNCHANGED));
2075
2076         // Needed if there is an optional argument but no contents.
2077         if (body_pos > 0 && body_pos == size()) {
2078                 os << "}]~";
2079                 return_value = false;
2080         }
2081
2082         if (!asdefault) {
2083                 column += d->endTeXParParams(bparams, os, texrow,
2084                                           runparams.moving_arg);
2085         }
2086
2087         LYXERR(Debug::LATEX, "SimpleTeXOnePar...done " << this);
2088         return return_value;
2089 }
2090
2091
2092 bool Paragraph::emptyTag() const
2093 {
2094         for (pos_type i = 0; i < size(); ++i) {
2095                 if (Inset const * inset = getInset(i)) {
2096                         InsetCode lyx_code = inset->lyxCode();
2097                         if (lyx_code != TOC_CODE &&
2098                             lyx_code != INCLUDE_CODE &&
2099                             lyx_code != GRAPHICS_CODE &&
2100                             lyx_code != ERT_CODE &&
2101                             lyx_code != LISTINGS_CODE &&
2102                             lyx_code != FLOAT_CODE &&
2103                             lyx_code != TABULAR_CODE) {
2104                                 return false;
2105                         }
2106                 } else {
2107                         char_type c = d->text_[i];
2108                         if (c != ' ' && c != '\t')
2109                                 return false;
2110                 }
2111         }
2112         return true;
2113 }
2114
2115
2116 string Paragraph::getID(Buffer const & buf, OutputParams const & runparams)
2117         const
2118 {
2119         for (pos_type i = 0; i < size(); ++i) {
2120                 if (Inset const * inset = getInset(i)) {
2121                         InsetCode lyx_code = inset->lyxCode();
2122                         if (lyx_code == LABEL_CODE) {
2123                                 InsetLabel const * const il = static_cast<InsetLabel const *>(inset);
2124                                 docstring const & id = il->getParam("name");
2125                                 return "id='" + to_utf8(sgml::cleanID(buf, runparams, id)) + "'";
2126                         }
2127                 }
2128         }
2129         return string();
2130 }
2131
2132
2133 pos_type Paragraph::firstWord(odocstream & os, OutputParams const & runparams)
2134         const
2135 {
2136         pos_type i;
2137         for (i = 0; i < size(); ++i) {
2138                 if (Inset const * inset = getInset(i)) {
2139                         inset->docbook(os, runparams);
2140                 } else {
2141                         char_type c = d->text_[i];
2142                         if (c == ' ')
2143                                 break;
2144                         os << sgml::escapeChar(c);
2145                 }
2146         }
2147         return i;
2148 }
2149
2150
2151 bool Paragraph::Private::onlyText(Buffer const & buf, Font const & outerfont, pos_type initial) const
2152 {
2153         Font font_old;
2154         pos_type size = text_.size();
2155         for (pos_type i = initial; i < size; ++i) {
2156                 Font font = owner_->getFont(buf.params(), i, outerfont);
2157                 if (text_[i] == META_INSET)
2158                         return false;
2159                 if (i != initial && font != font_old)
2160                         return false;
2161                 font_old = font;
2162         }
2163
2164         return true;
2165 }
2166
2167
2168 void Paragraph::simpleDocBookOnePar(Buffer const & buf,
2169                                     odocstream & os,
2170                                     OutputParams const & runparams,
2171                                     Font const & outerfont,
2172                                     pos_type initial) const
2173 {
2174         bool emph_flag = false;
2175
2176         Layout const & style = *d->layout_;
2177         FontInfo font_old =
2178                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2179
2180         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2181                 os << "]]>";
2182
2183         // parsing main loop
2184         for (pos_type i = initial; i < size(); ++i) {
2185                 Font font = getFont(buf.params(), i, outerfont);
2186
2187                 // handle <emphasis> tag
2188                 if (font_old.emph() != font.fontInfo().emph()) {
2189                         if (font.fontInfo().emph() == FONT_ON) {
2190                                 os << "<emphasis>";
2191                                 emph_flag = true;
2192                         } else if (i != initial) {
2193                                 os << "</emphasis>";
2194                                 emph_flag = false;
2195                         }
2196                 }
2197
2198                 if (Inset const * inset = getInset(i)) {
2199                         inset->docbook(os, runparams);
2200                 } else {
2201                         char_type c = d->text_[i];
2202
2203                         if (style.pass_thru)
2204                                 os.put(c);
2205                         else
2206                                 os << sgml::escapeChar(c);
2207                 }
2208                 font_old = font.fontInfo();
2209         }
2210
2211         if (emph_flag) {
2212                 os << "</emphasis>";
2213         }
2214
2215         if (style.free_spacing)
2216                 os << '\n';
2217         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2218                 os << "<![CDATA[";
2219 }
2220
2221
2222 bool Paragraph::isHfill(pos_type pos) const
2223 {
2224         Inset const * inset = getInset(pos);
2225         return inset && (inset->lyxCode() == SPACE_CODE &&
2226                          inset->isStretchableSpace());
2227 }
2228
2229
2230 bool Paragraph::isNewline(pos_type pos) const
2231 {
2232         Inset const * inset = getInset(pos);
2233         return inset && inset->lyxCode() == NEWLINE_CODE;
2234 }
2235
2236
2237 bool Paragraph::isLineSeparator(pos_type pos) const
2238 {
2239         char_type const c = d->text_[pos];
2240         if (isLineSeparatorChar(c))
2241                 return true;
2242         Inset const * inset = getInset(pos);
2243         return inset && inset->isLineSeparator();
2244 }
2245
2246
2247 /// Used by the spellchecker
2248 bool Paragraph::isLetter(pos_type pos) const
2249 {
2250         if (Inset const * inset = getInset(pos))
2251                 return inset->isLetter();
2252         char_type const c = d->text_[pos];
2253         return isLetterChar(c) || isDigit(c);
2254 }
2255
2256
2257 Language const *
2258 Paragraph::getParLanguage(BufferParams const & bparams) const
2259 {
2260         if (!empty())
2261                 return getFirstFontSettings(bparams).language();
2262         // FIXME: we should check the prev par as well (Lgb)
2263         return bparams.language;
2264 }
2265
2266
2267 bool Paragraph::isRTL(BufferParams const & bparams) const
2268 {
2269         return lyxrc.rtl_support
2270                 && getParLanguage(bparams)->rightToLeft()
2271                 && ownerCode() != ERT_CODE
2272                 && ownerCode() != LISTINGS_CODE;
2273 }
2274
2275
2276 void Paragraph::changeLanguage(BufferParams const & bparams,
2277                                Language const * from, Language const * to)
2278 {
2279         // change language including dummy font change at the end
2280         for (pos_type i = 0; i <= size(); ++i) {
2281                 Font font = getFontSettings(bparams, i);
2282                 if (font.language() == from) {
2283                         font.setLanguage(to);
2284                         setFont(i, font);
2285                 }
2286         }
2287 }
2288
2289
2290 bool Paragraph::isMultiLingual(BufferParams const & bparams) const
2291 {
2292         Language const * doc_language = bparams.language;
2293         FontList::const_iterator cit = d->fontlist_.begin();
2294         FontList::const_iterator end = d->fontlist_.end();
2295
2296         for (; cit != end; ++cit)
2297                 if (cit->font().language() != ignore_language &&
2298                     cit->font().language() != latex_language &&
2299                     cit->font().language() != doc_language)
2300                         return true;
2301         return false;
2302 }
2303
2304
2305 docstring Paragraph::asString(int options) const
2306 {
2307         return asString(0, size(), options);
2308 }
2309
2310
2311 docstring Paragraph::asString(pos_type beg, pos_type end, int options) const
2312 {
2313         odocstringstream os;
2314
2315         if (beg == 0
2316                 && options & AS_STR_LABEL
2317                 && !d->params_.labelString().empty())
2318                 os << d->params_.labelString() << ' ';
2319
2320         for (pos_type i = beg; i < end; ++i) {
2321                 char_type const c = d->text_[i];
2322                 if (isPrintable(c))
2323                         os.put(c);
2324                 else if (c == META_INSET && options & AS_STR_INSETS)
2325                         getInset(i)->textString(os);
2326         }
2327
2328         return os.str();
2329 }
2330
2331
2332 void Paragraph::setInsetOwner(Inset * inset)
2333 {
2334         d->inset_owner_ = inset;
2335 }
2336
2337
2338 int Paragraph::id() const
2339 {
2340         return d->id_;
2341 }
2342
2343
2344 Layout const & Paragraph::layout() const
2345 {
2346         return *d->layout_;
2347 }
2348
2349
2350 void Paragraph::setLayout(Layout const & layout)
2351 {
2352         d->layout_ = &layout;
2353 }
2354
2355
2356 void Paragraph::setEmptyOrDefaultLayout(DocumentClass const & tclass)
2357 {
2358         if (useEmptyLayout())
2359                 setLayout(tclass.emptyLayout());
2360         else
2361                 setLayout(tclass.defaultLayout());
2362 }
2363
2364
2365 Inset * Paragraph::inInset() const
2366 {
2367         return d->inset_owner_;
2368 }
2369
2370
2371 InsetCode Paragraph::ownerCode() const
2372 {
2373         return d->inset_owner_ ? d->inset_owner_->lyxCode() : NO_CODE;
2374 }
2375
2376
2377 ParagraphParameters & Paragraph::params()
2378 {
2379         return d->params_;
2380 }
2381
2382
2383 ParagraphParameters const & Paragraph::params() const
2384 {
2385         return d->params_;
2386 }
2387
2388
2389 bool Paragraph::isFreeSpacing() const
2390 {
2391         if (d->layout_->free_spacing)
2392                 return true;
2393         return d->inset_owner_ && d->inset_owner_->isFreeSpacing();
2394 }
2395
2396
2397 bool Paragraph::allowEmpty() const
2398 {
2399         if (d->layout_->keepempty)
2400                 return true;
2401         return d->inset_owner_ && d->inset_owner_->allowEmpty();
2402 }
2403
2404
2405 char_type Paragraph::transformChar(char_type c, pos_type pos) const
2406 {
2407         if (!Encodings::isArabicChar(c))
2408                 return c;
2409
2410         char_type prev_char = ' ';
2411         char_type next_char = ' ';
2412
2413         for (pos_type i = pos - 1; i >= 0; --i) {
2414                 char_type const par_char = d->text_[i];
2415                 if (!Encodings::isArabicComposeChar(par_char)) {
2416                         prev_char = par_char;
2417                         break;
2418                 }
2419         }
2420
2421         for (pos_type i = pos + 1, end = size(); i < end; ++i) {
2422                 char_type const par_char = d->text_[i];
2423                 if (!Encodings::isArabicComposeChar(par_char)) {
2424                         next_char = par_char;
2425                         break;
2426                 }
2427         }
2428
2429         if (Encodings::isArabicChar(next_char)) {
2430                 if (Encodings::isArabicChar(prev_char) &&
2431                         !Encodings::isArabicSpecialChar(prev_char))
2432                         return Encodings::transformChar(c, Encodings::FORM_MEDIAL);
2433                 else
2434                         return Encodings::transformChar(c, Encodings::FORM_INITIAL);
2435         } else {
2436                 if (Encodings::isArabicChar(prev_char) &&
2437                         !Encodings::isArabicSpecialChar(prev_char))
2438                         return Encodings::transformChar(c, Encodings::FORM_FINAL);
2439                 else
2440                         return Encodings::transformChar(c, Encodings::FORM_ISOLATED);
2441         }
2442 }
2443
2444
2445 int Paragraph::checkBiblio(Buffer const & buffer)
2446 {
2447         // FIXME From JS:
2448         // This is getting more and more a mess. ...We really should clean
2449         // up this bibitem issue for 1.6. See also bug 2743.
2450
2451         // Add bibitem insets if necessary
2452         if (d->layout_->labeltype != LABEL_BIBLIO)
2453                 return 0;
2454
2455         bool hasbibitem = !d->insetlist_.empty()
2456                 // Insist on it being in pos 0
2457                 && d->text_[0] == META_INSET
2458                 && d->insetlist_.begin()->inset->lyxCode() == BIBITEM_CODE;
2459
2460         bool track_changes = buffer.params().trackChanges;
2461
2462         docstring oldkey;
2463         docstring oldlabel;
2464
2465         // remove a bibitem in pos != 0
2466         // restore it later in pos 0 if necessary
2467         // (e.g. if a user inserts contents _before_ the item)
2468         // we're assuming there's only one of these, which there
2469         // should be.
2470         int erasedInsetPosition = -1;
2471         InsetList::iterator it = d->insetlist_.begin();
2472         InsetList::iterator end = d->insetlist_.end();
2473         for (; it != end; ++it)
2474                 if (it->inset->lyxCode() == BIBITEM_CODE
2475                     && it->pos > 0) {
2476                         InsetBibitem * olditem = static_cast<InsetBibitem *>(it->inset);
2477                         oldkey = olditem->getParam("key");
2478                         oldlabel = olditem->getParam("label");
2479                         erasedInsetPosition = it->pos;
2480                         eraseChar(erasedInsetPosition, track_changes);
2481                         break;
2482         }
2483
2484         // There was an InsetBibitem at the beginning, and we didn't
2485         // have to erase one.
2486         if (hasbibitem && erasedInsetPosition < 0)
2487                         return 0;
2488
2489         // There was an InsetBibitem at the beginning and we did have to
2490         // erase one. So we give its properties to the beginning inset.
2491         if (hasbibitem) {
2492                 InsetBibitem * inset =
2493                         static_cast<InsetBibitem *>(d->insetlist_.begin()->inset);
2494                 if (!oldkey.empty())
2495                         inset->setParam("key", oldkey);
2496                 inset->setParam("label", oldlabel);
2497                 return -erasedInsetPosition;
2498         }
2499
2500         // There was no inset at the beginning, so we need to create one with
2501         // the key and label of the one we erased.
2502         InsetBibitem * inset = new InsetBibitem(InsetCommandParams(BIBITEM_CODE));
2503         inset->setBuffer(const_cast<Buffer &>(buffer));
2504         // restore values of previously deleted item in this par.
2505         if (!oldkey.empty())
2506                 inset->setParam("key", oldkey);
2507         inset->setParam("label", oldlabel);
2508         insertInset(0, static_cast<Inset *>(inset),
2509                     Change(track_changes ? Change::INSERTED : Change::UNCHANGED));
2510
2511         return 1;
2512 }
2513
2514
2515 void Paragraph::checkAuthors(AuthorList const & authorList)
2516 {
2517         d->changes_.checkAuthors(authorList);
2518 }
2519
2520
2521 bool Paragraph::isUnchanged(pos_type pos) const
2522 {
2523         return lookupChange(pos).type == Change::UNCHANGED;
2524 }
2525
2526
2527 bool Paragraph::isInserted(pos_type pos) const
2528 {
2529         return lookupChange(pos).type == Change::INSERTED;
2530 }
2531
2532
2533 bool Paragraph::isDeleted(pos_type pos) const
2534 {
2535         return lookupChange(pos).type == Change::DELETED;
2536 }
2537
2538
2539 InsetList const & Paragraph::insetList() const
2540 {
2541         return d->insetlist_;
2542 }
2543
2544
2545 Inset * Paragraph::releaseInset(pos_type pos)
2546 {
2547         Inset * inset = d->insetlist_.release(pos);
2548         /// does not honour change tracking!
2549         eraseChar(pos, false);
2550         return inset;
2551 }
2552
2553
2554 Inset * Paragraph::getInset(pos_type pos)
2555 {
2556         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
2557                  ? d->insetlist_.get(pos) : 0;
2558 }
2559
2560
2561 Inset const * Paragraph::getInset(pos_type pos) const
2562 {
2563         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
2564                  ? d->insetlist_.get(pos) : 0;
2565 }
2566
2567
2568 void Paragraph::changeCase(BufferParams const & bparams, pos_type pos,
2569                 pos_type & right, TextCase action)
2570 {
2571         // process sequences of modified characters; in change
2572         // tracking mode, this approach results in much better
2573         // usability than changing case on a char-by-char basis
2574         docstring changes;
2575
2576         bool const trackChanges = bparams.trackChanges;
2577
2578         bool capitalize = true;
2579
2580         for (; pos < right; ++pos) {
2581                 char_type oldChar = d->text_[pos];
2582                 char_type newChar = oldChar;
2583
2584                 // ignore insets and don't play with deleted text!
2585                 if (oldChar != META_INSET && !isDeleted(pos)) {
2586                         switch (action) {
2587                                 case text_lowercase:
2588                                         newChar = lowercase(oldChar);
2589                                         break;
2590                                 case text_capitalization:
2591                                         if (capitalize) {
2592                                                 newChar = uppercase(oldChar);
2593                                                 capitalize = false;
2594                                         }
2595                                         break;
2596                                 case text_uppercase:
2597                                         newChar = uppercase(oldChar);
2598                                         break;
2599                         }
2600                 }
2601
2602                 if (!isLetter(pos) || isDeleted(pos)) {
2603                         // permit capitalization again
2604                         capitalize = true;
2605                 }
2606
2607                 if (oldChar != newChar)
2608                         changes += newChar;
2609
2610                 if (oldChar == newChar || pos == right - 1) {
2611                         if (oldChar != newChar) {
2612                                 // step behind the changing area
2613                                 pos++;
2614                         }
2615                         int erasePos = pos - changes.size();
2616                         for (size_t i = 0; i < changes.size(); i++) {
2617                                 insertChar(pos, changes[i],
2618                                         getFontSettings(bparams,
2619                                         erasePos),
2620                                         trackChanges);
2621                                 if (!eraseChar(erasePos, trackChanges)) {
2622                                         ++erasePos;
2623                                         ++pos; // advance
2624                                         ++right; // expand selection
2625                                 }
2626                         }
2627                         changes.clear();
2628                 }
2629         }
2630 }
2631
2632
2633 bool Paragraph::find(docstring const & str, bool cs, bool mw,
2634                 pos_type pos, bool del) const
2635 {
2636         int const strsize = str.length();
2637         int i = 0;
2638         pos_type const parsize = d->text_.size();
2639         for (i = 0; pos + i < parsize; ++i) {
2640                 if (i >= strsize)
2641                         break;
2642                 if (cs && str[i] != d->text_[pos + i])
2643                         break;
2644                 if (!cs && uppercase(str[i]) != uppercase(d->text_[pos + i]))
2645                         break;
2646                 if (!del && isDeleted(pos + i))
2647                         break;
2648         }
2649
2650         if (i != strsize)
2651                 return false;
2652
2653         // if necessary, check whether string matches word
2654         if (mw) {
2655                 if (pos > 0 && isLetter(pos - 1))
2656                         return false;
2657                 if (pos + strsize < parsize
2658                         && isLetter(pos + strsize))
2659                         return false;
2660         }
2661
2662         return true;
2663 }
2664
2665
2666 char_type Paragraph::getChar(pos_type pos) const
2667 {
2668         return d->text_[pos];
2669 }
2670
2671
2672 pos_type Paragraph::size() const
2673 {
2674         return d->text_.size();
2675 }
2676
2677
2678 bool Paragraph::empty() const
2679 {
2680         return d->text_.empty();
2681 }
2682
2683
2684 bool Paragraph::isInset(pos_type pos) const
2685 {
2686         return d->text_[pos] == META_INSET;
2687 }
2688
2689
2690 bool Paragraph::isSeparator(pos_type pos) const
2691 {
2692         //FIXME: Are we sure this can be the only separator?
2693         return d->text_[pos] == ' ';
2694 }
2695
2696
2697 void Paragraph::deregisterWords()
2698 {
2699         Private::Words::const_iterator it;
2700         WordList & wl = theWordList();
2701         for (it = d->words_.begin(); it != d->words_.end(); ++it)
2702                 wl.remove(*it);
2703         d->words_.clear();
2704 }
2705
2706
2707 void Paragraph::collectWords(CursorSlice const & sl)
2708 {
2709         // find new words
2710         bool inword = false;
2711
2712         //lyxerr << "Words: ";
2713         pos_type n = size();
2714         for (pos_type pos = 0; pos != n; ++pos) {
2715                 if (isDeleted(pos))
2716                         continue;
2717
2718                 if (!isLetter(pos)) {
2719                         inword = false;
2720                         continue;
2721                 }
2722
2723                 if (inword)
2724                         continue;
2725
2726                 inword = true;
2727                 CursorSlice from = sl;
2728                 CursorSlice to = sl;
2729                 from.pos() = pos;
2730                 to.pos() = pos;
2731                 from.text()->getWord(from, to, WHOLE_WORD);
2732                 if (to.pos() - from.pos() < 6)
2733                         continue;
2734                 docstring word = asString(from.pos(), to.pos(), false);
2735                 d->words_.insert(word);
2736                 //lyxerr << word << " ";
2737         }
2738         //lyxerr << std::endl;
2739 }
2740
2741
2742 void Paragraph::registerWords()
2743 {
2744         Private::Words::const_iterator it;
2745         WordList & wl = theWordList();
2746         for (it = d->words_.begin(); it != d->words_.end(); ++it)
2747                 wl.insert(*it);
2748 }
2749
2750
2751 void Paragraph::updateWords(CursorSlice const & sl)
2752 {
2753         LASSERT(&sl.paragraph() == this, /**/);
2754         deregisterWords();
2755         collectWords(sl);
2756         registerWords();
2757 }
2758
2759 } // namespace lyx