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