src/Paragraph.cpp

   1 /**
   2  * \file Paragraph.cpp
   3  * This file is part of LyX, the document processor.
   4  * Licence details can be found in the file COPYING.
   5  *
   6  * \author Asger Alstrup
   7  * \author Lars Gullik Bjønnes
   8  * \author Richard Heck (XHTML output)
   9  * \author Jean-Marc Lasgouttes
  10  * \author Angus Leeming
  11  * \author John Levon
  12  * \author André Pönitz
  13  * \author Dekel Tsur
  14  * \author Jürgen Vigna
  15  *
  16  * Full author contact details are available in file CREDITS.
  17  */
  18
  19 #include <config.h>
  20
  21 #include "Paragraph.h"
  22
  23 #include "LayoutFile.h"
  24 #include "Buffer.h"
  25 #include "BufferParams.h"
  26 #include "Changes.h"
  27 #include "Counters.h"
  28 #include "BufferEncodings.h"
  29 #include "InsetList.h"
  30 #include "Language.h"
  31 #include "LaTeXFeatures.h"
  32 #include "Layout.h"
  33 #include "Length.h"
  34 #include "Font.h"
  35 #include "FontList.h"
  36 #include "LyXRC.h"
  37 #include "OutputParams.h"
  38 #include "output_latex.h"
  39 #include "output_xhtml.h"
  40 #include "ParagraphParameters.h"
  41 #include "SpellChecker.h"
  42 #include "sgml.h"
  43 #include "TextClass.h"
  44 #include "TexRow.h"
  45 #include "Text.h"
  46 #include "WordLangTuple.h"
  47 #include "WordList.h"
  48
  49 #include "frontends/alert.h"
  50
  51 #include "insets/InsetBibitem.h"
  52 #include "insets/InsetLabel.h"
  53 #include "insets/InsetSpecialChar.h"
  54
  55 #include "support/debug.h"
  56 #include "support/docstring_list.h"
  57 #include "support/ExceptionMessage.h"
  58 #include "support/gettext.h"
  59 #include "support/lassert.h"
  60 #include "support/lstrings.h"
  61 #include "support/textutils.h"
  62
  63 #include <sstream>
  64 #include <vector>
  65
  66 using namespace std;
  67 using namespace lyx::support;
  68
  69 namespace lyx {
  70
  71 namespace {
  72
  73 /// Inset identifier (above 0x10ffff, for ucs-4)
  74 char_type const META_INSET = 0x200001;
  75
  76 }
  77
  78
  79 /////////////////////////////////////////////////////////////////////
  80 //
  81 // SpellResultRange
  82 //
  83 /////////////////////////////////////////////////////////////////////
  84
  85 class SpellResultRange {
  86 public:
  87         SpellResultRange(FontSpan range, SpellChecker::Result result)
  88         : range_(range), result_(result)
  89         {}
  90         ///
  91         FontSpan const & range() const { return range_; }
  92         ///
  93         void range(FontSpan const & r) { range_ = r; }
  94         ///
  95         SpellChecker::Result result() const { return result_; }
  96         ///
  97         void result(SpellChecker::Result r) { result_ = r; }
  98         ///
  99         bool contains(pos_type pos) const { return range_.contains(pos); }
 100         ///
 101         bool covered(FontSpan const & r) const
 102         {
 103                 // 1. first of new range inside current range or
 104                 // 2. last of new range inside current range or
 105                 // 3. first of current range inside new range or
 106                 // 4. last of current range inside new range
 107                 //FIXME: is this the same as !range_.intersect(r).empty() ?
 108                 return range_.contains(r.first) || range_.contains(r.last) ||
 109                         r.contains(range_.first) || r.contains(range_.last);
 110         }
 111         ///
 112         void shift(pos_type pos, int offset)
 113         {
 114                 if (range_.first > pos) {
 115                         range_.first += offset;
 116                         range_.last += offset;
 117                 } else if (range_.last >= pos) {
 118                         range_.last += offset;
 119                 }
 120         }
 121 private:
 122         FontSpan range_ ;
 123         SpellChecker::Result result_ ;
 124 };
 125
 126
 127 /////////////////////////////////////////////////////////////////////
 128 //
 129 // SpellCheckerState
 130 //
 131 /////////////////////////////////////////////////////////////////////
 132
 133 class SpellCheckerState {
 134 public:
 135         SpellCheckerState()
 136         {
 137                 needs_refresh_ = true;
 138                 current_change_number_ = 0;
 139         }
 140
 141         void setRange(FontSpan const & fp, SpellChecker::Result state)
 142         {
 143                 Ranges result;
 144                 RangesIterator et = ranges_.end();
 145                 RangesIterator it = ranges_.begin();
 146                 for (; it != et; ++it) {
 147                         if (!it->covered(fp))
 148                                 result.push_back(SpellResultRange(it->range(), it->result()));
 149                         else if (state == SpellChecker::WORD_OK) {
 150                                 // trim or split the current misspelled range
 151                                 // store misspelled ranges only
 152                                 FontSpan range = it->range();
 153                                 if (fp.first > range.first) {
 154                                         // misspelled area in front of WORD_OK
 155                                         range.last = fp.first - 1;
 156                                         result.push_back(SpellResultRange(range, it->result()));
 157                                         range = it->range();
 158                                 }
 159                                 if (fp.last < range.last) {
 160                                         // misspelled area after WORD_OK range
 161                                         range.first = fp.last + 1;
 162                                         result.push_back(SpellResultRange(range, it->result()));
 163                                 }
 164                         }
 165                 }
 166                 ranges_ = result;
 167                 if (state != SpellChecker::WORD_OK)
 168                         ranges_.push_back(SpellResultRange(fp, state));
 169         }
 170
 171         void increasePosAfterPos(pos_type pos)
 172         {
 173                 correctRangesAfterPos(pos, 1);
 174                 needsRefresh(pos);
 175         }
 176
 177         void decreasePosAfterPos(pos_type pos)
 178         {
 179                 correctRangesAfterPos(pos, -1);
 180                 needsRefresh(pos);
 181         }
 182
 183         void refreshLast(pos_type pos)
 184         {
 185                 if (pos < refresh_.last)
 186                         refresh_.last = pos;
 187         }
 188
 189         SpellChecker::Result getState(pos_type pos) const
 190         {
 191                 SpellChecker::Result result = SpellChecker::WORD_OK;
 192                 RangesIterator et = ranges_.end();
 193                 RangesIterator it = ranges_.begin();
 194                 for (; it != et; ++it) {
 195                         if(it->contains(pos)) {
 196                                 return it->result();
 197                         }
 198                 }
 199                 return result;
 200         }
 201
 202         FontSpan const & getRange(pos_type pos) const
 203         {
 204                 /// empty span to indicate mismatch
 205                 static FontSpan empty_;
 206                 RangesIterator et = ranges_.end();
 207                 RangesIterator it = ranges_.begin();
 208                 for (; it != et; ++it) {
 209                         if(it->contains(pos)) {
 210                                 return it->range();
 211                         }
 212                 }
 213                 return empty_;
 214         }
 215
 216         bool needsRefresh() const
 217         {
 218                 return needs_refresh_;
 219         }
 220
 221         SpellChecker::ChangeNumber currentChangeNumber() const
 222         {
 223                 return current_change_number_;
 224         }
 225
 226         void refreshRange(pos_type & first, pos_type & last) const
 227         {
 228                 first = refresh_.first;
 229                 last = refresh_.last;
 230         }
 231
 232         void needsRefresh(pos_type pos)
 233         {
 234                 if (needs_refresh_ && pos != -1) {
 235                         if (pos < refresh_.first)
 236                                 refresh_.first = pos;
 237                         if (pos > refresh_.last)
 238                                 refresh_.last = pos;
 239                 } else if (pos != -1) {
 240                         // init request check for neighbour positions too
 241                         refresh_.first = pos > 0 ? pos - 1 : 0;
 242                         // no need for special end of paragraph check
 243                         refresh_.last = pos + 1;
 244                 }
 245                 needs_refresh_ = pos != -1;
 246         }
 247
 248         void needsCompleteRefresh(SpellChecker::ChangeNumber change_number)
 249         {
 250                 needs_refresh_ = true;
 251                 refresh_.first = 0;
 252                 refresh_.last = -1;
 253                 current_change_number_ = change_number;
 254         }
 255 private:
 256         typedef vector<SpellResultRange> Ranges;
 257         typedef Ranges::const_iterator RangesIterator;
 258         Ranges ranges_;
 259         /// the area of the paragraph with pending spell check
 260         FontSpan refresh_;
 261         bool needs_refresh_;
 262         /// spell state cache version number
 263         SpellChecker::ChangeNumber current_change_number_;
 264
 265
 266         void correctRangesAfterPos(pos_type pos, int offset)
 267         {
 268                 RangesIterator et = ranges_.end();
 269                 Ranges::iterator it = ranges_.begin();
 270                 for (; it != et; ++it) {
 271                         it->shift(pos, offset);
 272                 }
 273         }
 274
 275 };
 276
 277 /////////////////////////////////////////////////////////////////////
 278 //
 279 // Paragraph::Private
 280 //
 281 /////////////////////////////////////////////////////////////////////
 282
 283 class Paragraph::Private
 284 {
 285         // Enforce our own "copy" constructor by declaring the standard one and
 286         // the assignment operator private without implementing them.
 287         Private(Private const &);
 288         Private & operator=(Private const &);
 289 public:
 290         ///
 291         Private(Paragraph * owner, Layout const & layout);
 292         /// "Copy constructor"
 293         Private(Private const &, Paragraph * owner);
 294         /// Copy constructor from \p beg  to \p end
 295         Private(Private const &, Paragraph * owner, pos_type beg, pos_type end);
 296
 297         ///
 298         void insertChar(pos_type pos, char_type c, Change const & change);
 299
 300         /// Output the surrogate pair formed by \p c and \p next to \p os.
 301         /// \return the number of characters written.
 302         int latexSurrogatePair(otexstream & os, char_type c, char_type next,
 303                                OutputParams const &);
 304
 305         /// Output a space in appropriate formatting (or a surrogate pair
 306         /// if the next character is a combining character).
 307         /// \return whether a surrogate pair was output.
 308         bool simpleTeXBlanks(OutputParams const &,
 309                              otexstream &,
 310                              pos_type i,
 311                              unsigned int & column,
 312                              Font const & font,
 313                              Layout const & style);
 314
 315         /// Output consecutive unicode chars, belonging to the same script as
 316         /// specified by the latex macro \p ltx, to \p os starting from \p i.
 317         /// \return the number of characters written.
 318         int writeScriptChars(otexstream & os, docstring const & ltx,
 319                            Change const &, Encoding const &, pos_type & i);
 320
 321         /// This could go to ParagraphParameters if we want to.
 322         int startTeXParParams(BufferParams const &, otexstream &,
 323                               OutputParams const &) const;
 324
 325         /// This could go to ParagraphParameters if we want to.
 326         bool endTeXParParams(BufferParams const &, otexstream &,
 327                              OutputParams const &) const;
 328
 329         ///
 330         void latexInset(BufferParams const &,
 331                                    otexstream &,
 332                                    OutputParams &,
 333                                    Font & running_font,
 334                                    Font & basefont,
 335                                    Font const & outerfont,
 336                                    bool & open_font,
 337                                    Change & running_change,
 338                                    Layout const & style,
 339                                    pos_type & i,
 340                                    unsigned int & column);
 341
 342         ///
 343         void latexSpecialChar(
 344                                    otexstream & os,
 345                                    BufferParams const & bparams,
 346                                    OutputParams const & runparams,
 347                                    Font const & running_font,
 348                                    Change const & running_change,
 349                                    Layout const & style,
 350                                    pos_type & i,
 351                                    pos_type end_pos,
 352                                    unsigned int & column);
 353
 354         ///
 355         bool latexSpecialT1(
 356                 char_type const c,
 357                 otexstream & os,
 358                 pos_type i,
 359                 unsigned int & column);
 360         ///
 361         bool latexSpecialT3(
 362                 char_type const c,
 363                 otexstream & os,
 364                 pos_type i,
 365                 unsigned int & column);
 366
 367         ///
 368         void validate(LaTeXFeatures & features) const;
 369
 370         /// Checks if the paragraph contains only text and no inset or font change.
 371         bool onlyText(Buffer const & buf, Font const & outerfont,
 372                       pos_type initial) const;
 373
 374         /// a vector of speller skip positions
 375         typedef vector<FontSpan> SkipPositions;
 376         typedef SkipPositions::const_iterator SkipPositionsIterator;
 377
 378         void appendSkipPosition(SkipPositions & skips, pos_type const pos) const;
 379
 380         Language * getSpellLanguage(pos_type const from) const;
 381
 382         Language * locateSpellRange(pos_type & from, pos_type & to,
 383                                     SkipPositions & skips) const;
 384
 385         bool hasSpellerChange() const
 386         {
 387                 SpellChecker::ChangeNumber speller_change_number = 0;
 388                 if (theSpellChecker())
 389                         speller_change_number = theSpellChecker()->changeNumber();
 390                 return speller_change_number > speller_state_.currentChangeNumber();
 391         }
 392
 393         bool ignoreWord(docstring const & word) const ;
 394
 395         void setMisspelled(pos_type from, pos_type to, SpellChecker::Result state)
 396         {
 397                 pos_type textsize = owner_->size();
 398                 // check for sane arguments
 399                 if (to <= from || from >= textsize)
 400                         return;
 401                 FontSpan fp = FontSpan(from, to - 1);
 402                 speller_state_.setRange(fp, state);
 403         }
 404
 405         void requestSpellCheck(pos_type pos)
 406         {
 407                 if (pos == -1)
 408                         speller_state_.needsCompleteRefresh(speller_state_.currentChangeNumber());
 409                 else
 410                         speller_state_.needsRefresh(pos);
 411         }
 412
 413         void readySpellCheck()
 414         {
 415                 speller_state_.needsRefresh(-1);
 416         }
 417
 418         bool needsSpellCheck() const
 419         {
 420                 return speller_state_.needsRefresh();
 421         }
 422
 423         void rangeOfSpellCheck(pos_type & first, pos_type & last) const
 424         {
 425                 speller_state_.refreshRange(first, last);
 426                 if (last == -1) {
 427                         last = owner_->size();
 428                         return;
 429                 }
 430                 pos_type endpos = last;
 431                 owner_->locateWord(first, endpos, WHOLE_WORD);
 432                 if (endpos < last) {
 433                         endpos = last;
 434                         owner_->locateWord(last, endpos, WHOLE_WORD);
 435                 }
 436                 last = endpos;
 437         }
 438
 439         int countSkips(SkipPositionsIterator & it, SkipPositionsIterator const et,
 440                             int & start) const
 441         {
 442                 int numskips = 0;
 443                 while (it != et && it->first < start) {
 444                         int skip = it->last - it->first + 1;
 445                         start += skip;
 446                         numskips += skip;
 447                         ++it;
 448                 }
 449                 return numskips;
 450         }
 451
 452         void markMisspelledWords(pos_type const & first, pos_type const & last,
 453                                                          SpellChecker::Result result,
 454                                                          docstring const & word,
 455                                                          SkipPositions const & skips);
 456
 457         InsetCode ownerCode() const
 458         {
 459                 return inset_owner_ ? inset_owner_->lyxCode() : NO_CODE;
 460         }
 461
 462         /// Which Paragraph owns us?
 463         Paragraph * owner_;
 464
 465         /// In which Inset?
 466         Inset const * inset_owner_;
 467
 468         ///
 469         FontList fontlist_;
 470
 471         ///
 472         int id_;
 473
 474         ///
 475         ParagraphParameters params_;
 476
 477         /// for recording and looking up changes
 478         Changes changes_;
 479
 480         ///
 481         InsetList insetlist_;
 482
 483         /// end of label
 484         pos_type begin_of_body_;
 485
 486         typedef docstring TextContainer;
 487         ///
 488         TextContainer text_;
 489
 490         typedef set<docstring> Words;
 491         typedef map<string, Words> LangWordsMap;
 492         ///
 493         LangWordsMap words_;
 494         ///
 495         Layout const * layout_;
 496         ///
 497         SpellCheckerState speller_state_;
 498 };
 499
 500
 501 Paragraph::Private::Private(Paragraph * owner, Layout const & layout)
 502         : owner_(owner), inset_owner_(0), id_(-1), begin_of_body_(0), layout_(&layout)
 503 {
 504         text_.reserve(100);
 505 }
 506
 507
 508 // Initialization of the counter for the paragraph id's,
 509 //
 510 // FIXME: There should be a more intelligent way to generate and use the
 511 // paragraph ids per buffer instead a global static counter for all InsetText
 512 // in the running program.
 513 // However, this per-session id is used in LFUN_PARAGRAPH_GOTO to
 514 // switch to a different buffer, as used in the outliner for instance.
 515 static int paragraph_id = -1;
 516
 517 Paragraph::Private::Private(Private const & p, Paragraph * owner)
 518         : owner_(owner), inset_owner_(p.inset_owner_), fontlist_(p.fontlist_),
 519           params_(p.params_), changes_(p.changes_), insetlist_(p.insetlist_),
 520           begin_of_body_(p.begin_of_body_), text_(p.text_), words_(p.words_),
 521           layout_(p.layout_)
 522 {
 523         id_ = ++paragraph_id;
 524         requestSpellCheck(p.text_.size());
 525 }
 526
 527
 528 Paragraph::Private::Private(Private const & p, Paragraph * owner,
 529         pos_type beg, pos_type end)
 530         : owner_(owner), inset_owner_(p.inset_owner_),
 531           params_(p.params_), changes_(p.changes_),
 532           insetlist_(p.insetlist_, beg, end),
 533           begin_of_body_(p.begin_of_body_), words_(p.words_),
 534           layout_(p.layout_)
 535 {
 536         id_ = ++paragraph_id;
 537         if (beg >= pos_type(p.text_.size()))
 538                 return;
 539         text_ = p.text_.substr(beg, end - beg);
 540
 541         FontList::const_iterator fcit = fontlist_.begin();
 542         FontList::const_iterator fend = fontlist_.end();
 543         for (; fcit != fend; ++fcit) {
 544                 if (fcit->pos() < beg)
 545                         continue;
 546                 if (fcit->pos() >= end) {
 547                         // Add last entry in the fontlist_.
 548                         fontlist_.set(text_.size() - 1, fcit->font());
 549                         break;
 550                 }
 551                 // Add a new entry in the fontlist_.
 552                 fontlist_.set(fcit->pos() - beg, fcit->font());
 553         }
 554         requestSpellCheck(p.text_.size());
 555 }
 556
 557
 558 void Paragraph::addChangesToToc(DocIterator const & cdit,
 559         Buffer const & buf, bool output_active) const
 560 {
 561         d->changes_.addToToc(cdit, buf, output_active);
 562 }
 563
 564
 565 bool Paragraph::isDeleted(pos_type start, pos_type end) const
 566 {
 567         LASSERT(start >= 0 && start <= size(), return false);
 568         LASSERT(end > start && end <= size() + 1, return false);
 569
 570         return d->changes_.isDeleted(start, end);
 571 }
 572
 573
 574 bool Paragraph::isChanged(pos_type start, pos_type end) const
 575 {
 576         LASSERT(start >= 0 && start <= size(), return false);
 577         LASSERT(end > start && end <= size() + 1, return false);
 578
 579         return d->changes_.isChanged(start, end);
 580 }
 581
 582
 583 bool Paragraph::isMergedOnEndOfParDeletion(bool trackChanges) const
 584 {
 585         // keep the logic here in sync with the logic of eraseChars()
 586         if (!trackChanges)
 587                 return true;
 588
 589         Change const & change = d->changes_.lookup(size());
 590         return change.inserted() && change.currentAuthor();
 591 }
 592
 593
 594 void Paragraph::setChange(Change const & change)
 595 {
 596         // beware of the imaginary end-of-par character!
 597         d->changes_.set(change, 0, size() + 1);
 598
 599         /*
 600          * Propagate the change recursively - but not in case of DELETED!
 601          *
 602          * Imagine that your co-author makes changes in an existing inset. He
 603          * sends your document to you and you come to the conclusion that the
 604          * inset should go completely. If you erase it, LyX must not delete all
 605          * text within the inset. Otherwise, the change tracked insertions of
 606          * your co-author get lost and there is no way to restore them later.
 607          *
 608          * Conclusion: An inset's content should remain untouched if you delete it
 609          */
 610
 611         if (!change.deleted()) {
 612                 for (pos_type pos = 0; pos < size(); ++pos) {
 613                         if (Inset * inset = getInset(pos))
 614                                 inset->setChange(change);
 615                 }
 616         }
 617 }
 618
 619
 620 void Paragraph::setChange(pos_type pos, Change const & change)
 621 {
 622         LASSERT(pos >= 0 && pos <= size(), return);
 623         d->changes_.set(change, pos);
 624
 625         // see comment in setChange(Change const &) above
 626         if (!change.deleted() && pos < size())
 627                         if (Inset * inset = getInset(pos))
 628                                 inset->setChange(change);
 629 }
 630
 631
 632 Change const & Paragraph::lookupChange(pos_type pos) const
 633 {
 634         LBUFERR(pos >= 0 && pos <= size());
 635         return d->changes_.lookup(pos);
 636 }
 637
 638
 639 void Paragraph::acceptChanges(pos_type start, pos_type end)
 640 {
 641         LASSERT(start >= 0 && start <= size(), return);
 642         LASSERT(end > start && end <= size() + 1, return);
 643
 644         for (pos_type pos = start; pos < end; ++pos) {
 645                 switch (lookupChange(pos).type) {
 646                         case Change::UNCHANGED:
 647                                 // accept changes in nested inset
 648                                 if (Inset * inset = getInset(pos))
 649                                         inset->acceptChanges();
 650                                 break;
 651
 652                         case Change::INSERTED:
 653                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 654                                 // also accept changes in nested inset
 655                                 if (Inset * inset = getInset(pos))
 656                                         inset->acceptChanges();
 657                                 break;
 658
 659                         case Change::DELETED:
 660                                 // Suppress access to non-existent
 661                                 // "end-of-paragraph char"
 662                                 if (pos < size()) {
 663                                         eraseChar(pos, false);
 664                                         --end;
 665                                         --pos;
 666                                 }
 667                                 break;
 668                 }
 669
 670         }
 671 }
 672
 673
 674 void Paragraph::rejectChanges(pos_type start, pos_type end)
 675 {
 676         LASSERT(start >= 0 && start <= size(), return);
 677         LASSERT(end > start && end <= size() + 1, return);
 678
 679         for (pos_type pos = start; pos < end; ++pos) {
 680                 switch (lookupChange(pos).type) {
 681                         case Change::UNCHANGED:
 682                                 // reject changes in nested inset
 683                                 if (Inset * inset = getInset(pos))
 684                                                 inset->rejectChanges();
 685                                 break;
 686
 687                         case Change::INSERTED:
 688                                 // Suppress access to non-existent
 689                                 // "end-of-paragraph char"
 690                                 if (pos < size()) {
 691                                         eraseChar(pos, false);
 692                                         --end;
 693                                         --pos;
 694                                 }
 695                                 break;
 696
 697                         case Change::DELETED:
 698                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 699
 700                                 // Do NOT reject changes within a deleted inset!
 701                                 // There may be insertions of a co-author inside of it!
 702
 703                                 break;
 704                 }
 705         }
 706 }
 707
 708
 709 void Paragraph::Private::insertChar(pos_type pos, char_type c,
 710                 Change const & change)
 711 {
 712         LASSERT(pos >= 0 && pos <= int(text_.size()), return);
 713
 714         // track change
 715         changes_.insert(change, pos);
 716
 717         // This is actually very common when parsing buffers (and
 718         // maybe inserting ascii text)
 719         if (pos == pos_type(text_.size())) {
 720                 // when appending characters, no need to update tables
 721                 text_.push_back(c);
 722                 // but we want spell checking
 723                 requestSpellCheck(pos);
 724                 return;
 725         }
 726
 727         text_.insert(text_.begin() + pos, c);
 728
 729         // Update the font table.
 730         fontlist_.increasePosAfterPos(pos);
 731
 732         // Update the insets
 733         insetlist_.increasePosAfterPos(pos);
 734
 735         // Update list of misspelled positions
 736         speller_state_.increasePosAfterPos(pos);
 737 }
 738
 739
 740 bool Paragraph::insertInset(pos_type pos, Inset * inset,
 741                                    Font const & font, Change const & change)
 742 {
 743         LASSERT(inset, return false);
 744         LASSERT(pos >= 0 && pos <= size(), return false);
 745
 746         // Paragraph::insertInset() can be used in cut/copy/paste operation where
 747         // d->inset_owner_ is not set yet.
 748         if (d->inset_owner_ && !d->inset_owner_->insetAllowed(inset->lyxCode()))
 749                 return false;
 750
 751         d->insertChar(pos, META_INSET, change);
 752         LASSERT(d->text_[pos] == META_INSET, return false);
 753
 754         // Add a new entry in the insetlist_.
 755         d->insetlist_.insert(inset, pos);
 756
 757         // Some insets require run of spell checker
 758         requestSpellCheck(pos);
 759         setFont(pos, font);
 760         return true;
 761 }
 762
 763
 764 bool Paragraph::eraseChar(pos_type pos, bool trackChanges)
 765 {
 766         LASSERT(pos >= 0 && pos <= size(), return false);
 767
 768         // keep the logic here in sync with the logic of isMergedOnEndOfParDeletion()
 769
 770         if (trackChanges) {
 771                 Change change = d->changes_.lookup(pos);
 772
 773                 // set the character to DELETED if
 774                 //  a) it was previously unchanged or
 775                 //  b) it was inserted by a co-author
 776
 777                 if (!change.changed() ||
 778                       (change.inserted() && !change.currentAuthor())) {
 779                         setChange(pos, Change(Change::DELETED));
 780                         // request run of spell checker
 781                         requestSpellCheck(pos);
 782                         return false;
 783                 }
 784
 785                 if (change.deleted())
 786                         return false;
 787         }
 788
 789         // Don't physically access the imaginary end-of-paragraph character.
 790         // eraseChar() can only mark it as DELETED. A physical deletion of
 791         // end-of-par must be handled externally.
 792         if (pos == size()) {
 793                 return false;
 794         }
 795
 796         // track change
 797         d->changes_.erase(pos);
 798
 799         // if it is an inset, delete the inset entry
 800         if (d->text_[pos] == META_INSET)
 801                 d->insetlist_.erase(pos);
 802
 803         d->text_.erase(d->text_.begin() + pos);
 804
 805         // Update the fontlist_
 806         d->fontlist_.erase(pos);
 807
 808         // Update the insetlist_
 809         d->insetlist_.decreasePosAfterPos(pos);
 810
 811         // Update list of misspelled positions
 812         d->speller_state_.decreasePosAfterPos(pos);
 813         d->speller_state_.refreshLast(size());
 814
 815         return true;
 816 }
 817
 818
 819 int Paragraph::eraseChars(pos_type start, pos_type end, bool trackChanges)
 820 {
 821         LASSERT(start >= 0 && start <= size(), return 0);
 822         LASSERT(end >= start && end <= size() + 1, return 0);
 823
 824         pos_type i = start;
 825         for (pos_type count = end - start; count; --count) {
 826                 if (!eraseChar(i, trackChanges))
 827                         ++i;
 828         }
 829         return end - i;
 830 }
 831
 832
 833 int Paragraph::Private::latexSurrogatePair(otexstream & os, char_type c,
 834                 char_type next, OutputParams const & runparams)
 835 {
 836         // Writing next here may circumvent a possible font change between
 837         // c and next. Since next is only output if it forms a surrogate pair
 838         // with c we can ignore this:
 839         // A font change inside a surrogate pair does not make sense and is
 840         // hopefully impossible to input.
 841         // FIXME: change tracking
 842         // Is this correct WRT change tracking?
 843         Encoding const & encoding = *(runparams.encoding);
 844         docstring latex1 = encoding.latexChar(next).first;
 845         if (runparams.inIPA) {
 846                 string const tipashortcut = Encodings::TIPAShortcut(next);
 847                 if (!tipashortcut.empty()) {
 848                         latex1 = from_ascii(tipashortcut);
 849                 }
 850         }
 851         docstring const latex2 = encoding.latexChar(c).first;
 852         if (docstring(1, next) == latex1) {
 853                 // the encoding supports the combination
 854                 os << latex2 << latex1;
 855                 return latex1.length() + latex2.length();
 856         } else if (runparams.local_font &&
 857                    runparams.local_font->language()->lang() == "polutonikogreek") {
 858                 // polutonikogreek only works without the brackets
 859                 os << latex1 << latex2;
 860                 return latex1.length() + latex2.length();
 861         } else
 862                 os << latex1 << '{' << latex2 << '}';
 863         return latex1.length() + latex2.length() + 2;
 864 }
 865
 866
 867 bool Paragraph::Private::simpleTeXBlanks(OutputParams const & runparams,
 868                                        otexstream & os,
 869                                        pos_type i,
 870                                        unsigned int & column,
 871                                        Font const & font,
 872                                        Layout const & style)
 873 {
 874         if (style.pass_thru || runparams.pass_thru)
 875                 return false;
 876
 877         if (i + 1 < int(text_.size())) {
 878                 char_type next = text_[i + 1];
 879                 if (Encodings::isCombiningChar(next)) {
 880                         // This space has an accent, so we must always output it.
 881                         column += latexSurrogatePair(os, ' ', next, runparams) - 1;
 882                         return true;
 883                 }
 884         }
 885
 886         if (runparams.linelen > 0
 887             && column > runparams.linelen
 888             && i
 889             && text_[i - 1] != ' '
 890             && (i + 1 < int(text_.size()))
 891             // same in FreeSpacing mode
 892             && !owner_->isFreeSpacing()
 893             // In typewriter mode, we want to avoid
 894             // ! . ? : at the end of a line
 895             && !(font.fontInfo().family() == TYPEWRITER_FAMILY
 896                  && (text_[i - 1] == '.'
 897                      || text_[i - 1] == '?'
 898                      || text_[i - 1] == ':'
 899                      || text_[i - 1] == '!'))) {
 900                 os << '\n';
 901                 os.texrow().start(owner_->id(), i + 1);
 902                 column = 0;
 903         } else if (style.free_spacing) {
 904                 os << '~';
 905         } else {
 906                 os << ' ';
 907         }
 908         return false;
 909 }
 910
 911
 912 int Paragraph::Private::writeScriptChars(otexstream & os,
 913                                          docstring const & ltx,
 914                                          Change const & runningChange,
 915                                          Encoding const & encoding,
 916                                          pos_type & i)
 917 {
 918         // FIXME: modifying i here is not very nice...
 919
 920         // We only arrive here when character text_[i] could not be translated
 921         // into the current latex encoding (or its latex translation has been forced,)
 922         // and it belongs to a known script.
 923         // TODO: We need \textcyr and \textgreek wrappers also for characters
 924         //       that can be encoded in the "LaTeX encoding" but not in the
 925         //       current *font encoding*.
 926         //       (See #9681 for details and test)
 927         // Parameter ltx contains the latex translation of text_[i] as specified
 928         // in the unicodesymbols file and is something like "\textXXX{<spec>}".
 929         // The latex macro name "textXXX" specifies the script to which text_[i]
 930         // belongs and we use it in order to check whether characters from the
 931         // same script immediately follow, such that we can collect them in a
 932         // single "\textXXX" macro. So, we have to retain "\textXXX{<spec>"
 933         // for the first char but only "<spec>" for all subsequent chars.
 934         docstring::size_type const brace1 = ltx.find_first_of(from_ascii("{"));
 935         docstring::size_type const brace2 = ltx.find_last_of(from_ascii("}"));
 936         string script = to_ascii(ltx.substr(1, brace1 - 1));
 937         int pos = 0;
 938         int length = brace2;
 939         bool closing_brace = true;
 940         if (script == "textgreek" && encoding.latexName() == "iso-8859-7") {
 941                 // Correct encoding is being used, so we can avoid \textgreek.
 942                 // TODO: wrong test: we need to check the *font encoding*
 943                 //       (i.e. the active language and its FontEncoding tag)
 944                 //       instead of the LaTeX *input encoding*!
 945                 //       See #9637 for details and test-cases.
 946                 pos = brace1 + 1;
 947                 length -= pos;
 948                 closing_brace = false;
 949         }
 950         os << ltx.substr(pos, length);
 951         int size = text_.size();
 952         while (i + 1 < size) {
 953                 char_type const next = text_[i + 1];
 954                 // Stop here if next character belongs to another script
 955                 // or there is a change in change tracking status.
 956                 if (!Encodings::isKnownScriptChar(next, script) ||
 957                     runningChange != owner_->lookupChange(i + 1))
 958                         break;
 959                 Font prev_font;
 960                 bool found = false;
 961                 FontList::const_iterator cit = fontlist_.begin();
 962                 FontList::const_iterator end = fontlist_.end();
 963                 for (; cit != end; ++cit) {
 964                         if (cit->pos() >= i && !found) {
 965                                 prev_font = cit->font();
 966                                 found = true;
 967                         }
 968                         if (cit->pos() >= i + 1)
 969                                 break;
 970                 }
 971                 // Stop here if there is a font attribute or encoding change.
 972                 if (found && cit != end && prev_font != cit->font())
 973                         break;
 974                 docstring const latex = encoding.latexChar(next).first;
 975                 docstring::size_type const b1 =
 976                                         latex.find_first_of(from_ascii("{"));
 977                 docstring::size_type const b2 =
 978                                         latex.find_last_of(from_ascii("}"));
 979                 int const len = b2 - b1 - 1;
 980                 os << latex.substr(b1 + 1, len);
 981                 length += len;
 982                 ++i;
 983         }
 984         if (closing_brace) {
 985                 os << '}';
 986                 ++length;
 987         }
 988         return length;
 989 }
 990
 991
 992 void Paragraph::Private::latexInset(BufferParams const & bparams,
 993                                     otexstream & os,
 994                                     OutputParams & runparams,
 995                                     Font & running_font,
 996                                     Font & basefont,
 997                                     Font const & outerfont,
 998                                     bool & open_font,
 999                                     Change & running_change,
1000                                     Layout const & style,
1001                                     pos_type & i,
1002                                     unsigned int & column)
1003 {
1004         Inset * inset = owner_->getInset(i);
1005         LBUFERR(inset);
1006
1007         if (style.pass_thru) {
1008                 odocstringstream ods;
1009                 inset->plaintext(ods, runparams);
1010                 os << ods.str();
1011                 return;
1012         }
1013
1014         // FIXME: move this to InsetNewline::latex
1015         if (inset->lyxCode() == NEWLINE_CODE || inset->lyxCode() == SEPARATOR_CODE) {
1016                 // newlines are handled differently here than
1017                 // the default in simpleTeXSpecialChars().
1018                 if (!style.newline_allowed) {
1019                         os << '\n';
1020                 } else {
1021                         if (open_font) {
1022                                 column += running_font.latexWriteEndChanges(
1023                                         os, bparams, runparams,
1024                                         basefont, basefont);
1025                                 open_font = false;
1026                         }
1027
1028                         if (running_font.fontInfo().family() == TYPEWRITER_FAMILY)
1029                                 os << '~';
1030
1031                         basefont = owner_->getLayoutFont(bparams, outerfont);
1032                         running_font = basefont;
1033
1034                         if (runparams.moving_arg)
1035                                 os << "\\protect ";
1036
1037                 }
1038                 os.texrow().start(owner_->id(), i + 1);
1039                 column = 0;
1040         }
1041
1042         if (owner_->isDeleted(i)) {
1043                 if( ++runparams.inDeletedInset == 1)
1044                         runparams.changeOfDeletedInset = owner_->lookupChange(i);
1045         }
1046
1047         if (inset->canTrackChanges()) {
1048                 column += Changes::latexMarkChange(os, bparams, running_change,
1049                         Change(Change::UNCHANGED), runparams);
1050                 running_change = Change(Change::UNCHANGED);
1051         }
1052
1053         bool close = false;
1054         odocstream::pos_type const len = os.os().tellp();
1055
1056         if (inset->forceLTR()
1057             && !runparams.use_polyglossia
1058             && running_font.isRightToLeft()
1059             // ERT is an exception, it should be output with no
1060             // decorations at all
1061             && inset->lyxCode() != ERT_CODE) {
1062                 if (running_font.language()->lang() == "farsi")
1063                         os << "\\beginL{}";
1064                 else
1065                         os << "\\L{";
1066                 close = true;
1067         }
1068
1069         // FIXME: Bug: we can have an empty font change here!
1070         // if there has just been a font change, we are going to close it
1071         // right now, which means stupid latex code like \textsf{}. AFAIK,
1072         // this does not harm dvi output. A minor bug, thus (JMarc)
1073
1074         // Some insets cannot be inside a font change command.
1075         // However, even such insets *can* be placed in \L or \R
1076         // or their equivalents (for RTL language switches), so we don't
1077         // close the language in those cases.
1078         // ArabTeX, though, cannot handle this special behavior, it seems.
1079         bool arabtex = basefont.language()->lang() == "arabic_arabtex"
1080                 || running_font.language()->lang() == "arabic_arabtex";
1081         if (open_font && !inset->inheritFont()) {
1082                 bool closeLanguage = arabtex
1083                         || basefont.isRightToLeft() == running_font.isRightToLeft();
1084                 unsigned int count = running_font.latexWriteEndChanges(os,
1085                         bparams, runparams, basefont, basefont, closeLanguage);
1086                 column += count;
1087                 // if any font properties were closed, update the running_font,
1088                 // making sure, however, to leave the language as it was
1089                 if (count > 0) {
1090                         // FIXME: probably a better way to keep track of the old
1091                         // language, than copying the entire font?
1092                         Font const copy_font(running_font);
1093                         basefont = owner_->getLayoutFont(bparams, outerfont);
1094                         running_font = basefont;
1095                         if (!closeLanguage)
1096                                 running_font.setLanguage(copy_font.language());
1097                         // leave font open if language is still open
1098                         open_font = (running_font.language() == basefont.language());
1099                         if (closeLanguage)
1100                                 runparams.local_font = &basefont;
1101                 }
1102         }
1103
1104         int prev_rows = os.texrow().rows();
1105
1106         try {
1107                 runparams.lastid = id_;
1108                 runparams.lastpos = i;
1109                 inset->latex(os, runparams);
1110         } catch (EncodingException & e) {
1111                 // add location information and throw again.
1112                 e.par_id = id_;
1113                 e.pos = i;
1114                 throw(e);
1115         }
1116
1117         if (close) {
1118                 if (running_font.language()->lang() == "farsi")
1119                                 os << "\\endL{}";
1120                         else
1121                                 os << '}';
1122         }
1123
1124         if (os.texrow().rows() > prev_rows) {
1125                 os.texrow().start(owner_->id(), i + 1);
1126                 column = 0;
1127         } else {
1128                 column += (unsigned int)(os.os().tellp() - len);
1129         }
1130
1131         if (owner_->isDeleted(i))
1132                 --runparams.inDeletedInset;
1133 }
1134
1135
1136 void Paragraph::Private::latexSpecialChar(otexstream & os,
1137                                           BufferParams const & bparams,
1138                                           OutputParams const & runparams,
1139                                           Font const & running_font,
1140                                           Change const & running_change,
1141                                           Layout const & style,
1142                                           pos_type & i,
1143                                           pos_type end_pos,
1144                                           unsigned int & column)
1145 {
1146         // With polyglossia, brackets and stuff need not be reversed
1147         // in RTL scripts (see bug #8251)
1148         char_type const c = (runparams.use_polyglossia) ?
1149                 owner_->getUChar(bparams, i) : text_[i];
1150
1151         if (style.pass_thru || runparams.pass_thru
1152             || contains(style.pass_thru_chars, c)
1153             || contains(runparams.pass_thru_chars, c)) {
1154                 if (c != '\0') {
1155                         Encoding const * const enc = runparams.encoding;
1156                         if (enc && !enc->encodable(c))
1157                                 throw EncodingException(c);
1158                         os.put(c);
1159                 }
1160                 return;
1161         }
1162
1163         // TIPA uses its own T3 encoding
1164         if (runparams.inIPA && latexSpecialT3(c, os, i, column))
1165                 return;
1166         // If T1 font encoding is used, use the special
1167         // characters it provides.
1168         // NOTE: Some languages reset the font encoding internally to a
1169         //       non-standard font encoding. If we are using such a language,
1170         //       we do not output special T1 chars.
1171         if (!runparams.inIPA && !running_font.language()->internalFontEncoding()
1172             && bparams.font_encoding() == "T1" && latexSpecialT1(c, os, i, column))
1173                 return;
1174
1175         // Otherwise, we use what LaTeX provides us.
1176         switch (c) {
1177         case '\\':
1178                 os << "\\textbackslash{}";
1179                 column += 15;
1180                 break;
1181         case '<':
1182                 os << "\\textless{}";
1183                 column += 10;
1184                 break;
1185         case '>':
1186                 os << "\\textgreater{}";
1187                 column += 13;
1188                 break;
1189         case '|':
1190                 os << "\\textbar{}";
1191                 column += 9;
1192                 break;
1193         case '-':
1194                 os << '-';
1195                 if (i + 1 < end_pos && text_[i+1] == '-') {
1196                         // Prevent "--" becoming an endash and "---" becoming
1197                         // an emdash.
1198                         // Within \ttfamily, "--" is merged to "-" (no endash)
1199                         // so we avoid this rather irritating ligature as well
1200                         os << "{}";
1201                         column += 2;
1202                 }
1203                 break;
1204         case '\"':
1205                 os << "\\char`\\\"{}";
1206                 column += 9;
1207                 break;
1208
1209         case '$': case '&':
1210         case '%': case '#': case '{':
1211         case '}': case '_':
1212                 os << '\\';
1213                 os.put(c);
1214                 column += 1;
1215                 break;
1216
1217         case '~':
1218                 os << "\\textasciitilde{}";
1219                 column += 16;
1220                 break;
1221
1222         case '^':
1223                 os << "\\textasciicircum{}";
1224                 column += 17;
1225                 break;
1226
1227         case '*':
1228         case '[':
1229         case ']':
1230                 // avoid being mistaken for optional arguments
1231                 os << '{';
1232                 os.put(c);
1233                 os << '}';
1234                 column += 2;
1235                 break;
1236
1237         case ' ':
1238                 // Blanks are printed before font switching.
1239                 // Sure? I am not! (try nice-latex)
1240                 // I am sure it's correct. LyX might be smarter
1241                 // in the future, but for now, nothing wrong is
1242                 // written. (Asger)
1243                 break;
1244
1245         default:
1246                 if (c == '\0')
1247                         return;
1248
1249                 Encoding const & encoding = *(runparams.encoding);
1250                 char_type next = '\0';
1251                 if (i + 1 < int(text_.size())) {
1252                         next = text_[i + 1];
1253                         if (Encodings::isCombiningChar(next)) {
1254                                 column += latexSurrogatePair(os, c, next, runparams) - 1;
1255                                 ++i;
1256                                 break;
1257                         }
1258                 }
1259                 string script;
1260                 pair<docstring, bool> latex = encoding.latexChar(c);
1261                 docstring nextlatex;
1262                 bool nexttipas = false;
1263                 string nexttipashortcut;
1264                 if (next != '\0' && next != META_INSET && encoding.encodable(next)) {
1265                         nextlatex = encoding.latexChar(next).first;
1266                         if (runparams.inIPA) {
1267                                 nexttipashortcut = Encodings::TIPAShortcut(next);
1268                                 nexttipas = !nexttipashortcut.empty();
1269                         }
1270                 }
1271                 bool tipas = false;
1272                 if (runparams.inIPA) {
1273                         string const tipashortcut = Encodings::TIPAShortcut(c);
1274                         if (!tipashortcut.empty()) {
1275                                 latex.first = from_ascii(tipashortcut);
1276                                 latex.second = false;
1277                                 tipas = true;
1278                         }
1279                 }
1280                 if (Encodings::isKnownScriptChar(c, script)
1281                     && prefixIs(latex.first, from_ascii("\\" + script)))
1282                         column += writeScriptChars(os, latex.first,
1283                                         running_change, encoding, i) - 1;
1284                 else if (latex.second
1285                          && ((!prefixIs(nextlatex, '\\')
1286                                && !prefixIs(nextlatex, '{')
1287                                && !prefixIs(nextlatex, '}'))
1288                              || (nexttipas
1289                                  && !prefixIs(from_ascii(nexttipashortcut), '\\')))
1290                          && !tipas) {
1291                         // Prevent eating of a following
1292                         // space or command corruption by
1293                         // following characters
1294                         if (next == ' ' || next == '\0') {
1295                                 column += latex.first.length() + 1;
1296                                 os << latex.first << "{}";
1297                         } else {
1298                                 column += latex.first.length();
1299                                 os << latex.first << " ";
1300                         }
1301                 } else {
1302                         column += latex.first.length() - 1;
1303                         os << latex.first;
1304                 }
1305                 break;
1306         }
1307 }
1308
1309
1310 bool Paragraph::Private::latexSpecialT1(char_type const c, otexstream & os,
1311         pos_type i, unsigned int & column)
1312 {
1313         switch (c) {
1314         case '>':
1315         case '<':
1316                 os.put(c);
1317                 // In T1 encoding, these characters exist
1318                 // but we should avoid ligatures
1319                 if (i + 1 >= int(text_.size()) || text_[i + 1] != c)
1320                         return true;
1321                 os << "\\textcompwordmark{}";
1322                 column += 19;
1323                 return true;
1324         case '|':
1325                 os.put(c);
1326                 return true;
1327         case '\"':
1328                 // soul.sty breaks with \char`\"
1329                 os << "\\textquotedbl{}";
1330                 column += 14;
1331                 return true;
1332         default:
1333                 return false;
1334         }
1335 }
1336
1337
1338 bool Paragraph::Private::latexSpecialT3(char_type const c, otexstream & os,
1339         pos_type /*i*/, unsigned int & column)
1340 {
1341         switch (c) {
1342         case '*':
1343         case '[':
1344         case ']':
1345         case '\"':
1346                 os.put(c);
1347                 return true;
1348         case '|':
1349                 os << "\\textvertline{}";
1350                 column += 14;
1351                 return true;
1352         default:
1353                 return false;
1354         }
1355 }
1356
1357
1358 void Paragraph::Private::validate(LaTeXFeatures & features) const
1359 {
1360         if (layout_->inpreamble && inset_owner_) {
1361                 bool const is_command = layout_->latextype == LATEX_COMMAND;
1362                 Buffer const & buf = inset_owner_->buffer();
1363                 BufferParams const & bp = features.runparams().is_child
1364                         ? buf.masterParams() : buf.params();
1365                 Font f;
1366                 TexRow texrow;
1367                 // Using a string stream here circumvents the encoding
1368                 // switching machinery of odocstream. Therefore the
1369                 // output is wrong if this paragraph contains content
1370                 // that needs to switch encoding.
1371                 odocstringstream ods;
1372                 otexstream os(ods, texrow);
1373                 if (is_command) {
1374                         os << '\\' << from_ascii(layout_->latexname());
1375                         // we have to provide all the optional arguments here, even though
1376                         // the last one is the only one we care about.
1377                         // Separate handling of optional argument inset.
1378                         if (!layout_->latexargs().empty()) {
1379                                 OutputParams rp = features.runparams();
1380                                 rp.local_font = &owner_->getFirstFontSettings(bp);
1381                                 latexArgInsets(*owner_, os, rp, layout_->latexargs());
1382                         }
1383                         os << from_ascii(layout_->latexparam());
1384                 }
1385                 docstring::size_type const length = ods.str().length();
1386                 // this will output "{" at the beginning, but not at the end
1387                 owner_->latex(bp, f, os, features.runparams(), 0, -1, true);
1388                 if (ods.str().length() > length) {
1389                         if (is_command) {
1390                                 ods << '}';
1391                                 if (!layout_->postcommandargs().empty()) {
1392                                         OutputParams rp = features.runparams();
1393                                         rp.local_font = &owner_->getFirstFontSettings(bp);
1394                                         latexArgInsets(*owner_, os, rp, layout_->postcommandargs(), "post:");
1395                                 }
1396                         }
1397                         string const snippet = to_utf8(ods.str());
1398                         features.addPreambleSnippet(snippet);
1399                 }
1400         }
1401
1402         if (features.runparams().flavor == OutputParams::HTML
1403             && layout_->htmltitle()) {
1404                 features.setHTMLTitle(owner_->asString(AS_STR_INSETS | AS_STR_SKIPDELETE));
1405         }
1406
1407         // check the params.
1408         if (!params_.spacing().isDefault())
1409                 features.require("setspace");
1410
1411         // then the layouts
1412         features.useLayout(layout_->name());
1413
1414         // then the fonts
1415         fontlist_.validate(features);
1416
1417         // then the indentation
1418         if (!params_.leftIndent().zero())
1419                 features.require("ParagraphLeftIndent");
1420
1421         // then the insets
1422         InsetList::const_iterator icit = insetlist_.begin();
1423         InsetList::const_iterator iend = insetlist_.end();
1424         for (; icit != iend; ++icit) {
1425                 if (icit->inset) {
1426                         icit->inset->validate(features);
1427                         if (layout_->needprotect &&
1428                             icit->inset->lyxCode() == FOOT_CODE)
1429                                 features.require("NeedLyXFootnoteCode");
1430                 }
1431         }
1432
1433         // then the contents
1434         for (pos_type i = 0; i < int(text_.size()) ; ++i) {
1435                 BufferEncodings::validate(text_[i], features);
1436         }
1437 }
1438
1439 /////////////////////////////////////////////////////////////////////
1440 //
1441 // Paragraph
1442 //
1443 /////////////////////////////////////////////////////////////////////
1444
1445 namespace {
1446         Layout const emptyParagraphLayout;
1447 }
1448
1449 Paragraph::Paragraph()
1450         : d(new Paragraph::Private(this, emptyParagraphLayout))
1451 {
1452         itemdepth = 0;
1453         d->params_.clear();
1454 }
1455
1456
1457 Paragraph::Paragraph(Paragraph const & par)
1458         : itemdepth(par.itemdepth),
1459         d(new Paragraph::Private(*par.d, this))
1460 {
1461         registerWords();
1462 }
1463
1464
1465 Paragraph::Paragraph(Paragraph const & par, pos_type beg, pos_type end)
1466         : itemdepth(par.itemdepth),
1467         d(new Paragraph::Private(*par.d, this, beg, end))
1468 {
1469         registerWords();
1470 }
1471
1472
1473 Paragraph & Paragraph::operator=(Paragraph const & par)
1474 {
1475         // needed as we will destroy the private part before copying it
1476         if (&par != this) {
1477                 itemdepth = par.itemdepth;
1478
1479                 deregisterWords();
1480                 delete d;
1481                 d = new Private(*par.d, this);
1482                 registerWords();
1483         }
1484         return *this;
1485 }
1486
1487
1488 Paragraph::~Paragraph()
1489 {
1490         deregisterWords();
1491         delete d;
1492 }
1493
1494
1495 namespace {
1496
1497 // this shall be called just before every "os << ..." action.
1498 void flushString(ostream & os, docstring & s)
1499 {
1500         os << to_utf8(s);
1501         s.erase();
1502 }
1503
1504 }
1505
1506
1507 void Paragraph::write(ostream & os, BufferParams const & bparams,
1508         depth_type & dth) const
1509 {
1510         // The beginning or end of a deeper (i.e. nested) area?
1511         if (dth != d->params_.depth()) {
1512                 if (d->params_.depth() > dth) {
1513                         while (d->params_.depth() > dth) {
1514                                 os << "\n\\begin_deeper";
1515                                 ++dth;
1516                         }
1517                 } else {
1518                         while (d->params_.depth() < dth) {
1519                                 os << "\n\\end_deeper";
1520                                 --dth;
1521                         }
1522                 }
1523         }
1524
1525         // First write the layout
1526         os << "\n\\begin_layout " << to_utf8(d->layout_->name()) << '\n';
1527
1528         d->params_.write(os);
1529
1530         Font font1(inherit_font, bparams.language);
1531
1532         Change running_change = Change(Change::UNCHANGED);
1533
1534         // this string is used as a buffer to avoid repetitive calls
1535         // to to_utf8(), which turn out to be expensive (JMarc)
1536         docstring write_buffer;
1537
1538         int column = 0;
1539         for (pos_type i = 0; i <= size(); ++i) {
1540
1541                 Change const & change = lookupChange(i);
1542                 if (change != running_change)
1543                         flushString(os, write_buffer);
1544                 Changes::lyxMarkChange(os, bparams, column, running_change, change);
1545                 running_change = change;
1546
1547                 if (i == size())
1548                         break;
1549
1550                 // Write font changes
1551                 Font font2 = getFontSettings(bparams, i);
1552                 if (font2 != font1) {
1553                         flushString(os, write_buffer);
1554                         font2.lyxWriteChanges(font1, os);
1555                         column = 0;
1556                         font1 = font2;
1557                 }
1558
1559                 char_type const c = d->text_[i];
1560                 switch (c) {
1561                 case META_INSET:
1562                         if (Inset const * inset = getInset(i)) {
1563                                 flushString(os, write_buffer);
1564                                 if (inset->directWrite()) {
1565                                         // international char, let it write
1566                                         // code directly so it's shorter in
1567                                         // the file
1568                                         inset->write(os);
1569                                 } else {
1570                                         if (i)
1571                                                 os << '\n';
1572                                         os << "\\begin_inset ";
1573                                         inset->write(os);
1574                                         os << "\n\\end_inset\n\n";
1575                                         column = 0;
1576                                 }
1577                                 // FIXME This can be removed again once the mystery
1578                                 // crash has been resolved.
1579                                 os << flush;
1580                         }
1581                         break;
1582                 case '\\':
1583                         flushString(os, write_buffer);
1584                         os << "\n\\backslash\n";
1585                         column = 0;
1586                         break;
1587                 case '.':
1588                         flushString(os, write_buffer);
1589                         if (i + 1 < size() && d->text_[i + 1] == ' ') {
1590                                 os << ".\n";
1591                                 column = 0;
1592                         } else
1593                                 os << '.';
1594                         break;
1595                 default:
1596                         if ((column > 70 && c == ' ')
1597                             || column > 79) {
1598                                 flushString(os, write_buffer);
1599                                 os << '\n';
1600                                 column = 0;
1601                         }
1602                         // this check is to amend a bug. LyX sometimes
1603                         // inserts '\0' this could cause problems.
1604                         if (c != '\0')
1605                                 write_buffer.push_back(c);
1606                         else
1607                                 LYXERR0("NUL char in structure.");
1608                         ++column;
1609                         break;
1610                 }
1611         }
1612
1613         flushString(os, write_buffer);
1614         os << "\n\\end_layout\n";
1615         // FIXME This can be removed again once the mystery
1616         // crash has been resolved.
1617         os << flush;
1618 }
1619
1620
1621 void Paragraph::validate(LaTeXFeatures & features) const
1622 {
1623         d->validate(features);
1624 }
1625
1626
1627 void Paragraph::insert(pos_type start, docstring const & str,
1628                        Font const & font, Change const & change)
1629 {
1630         for (size_t i = 0, n = str.size(); i != n ; ++i)
1631                 insertChar(start + i, str[i], font, change);
1632 }
1633
1634
1635 void Paragraph::appendChar(char_type c, Font const & font,
1636                 Change const & change)
1637 {
1638         // track change
1639         d->changes_.insert(change, d->text_.size());
1640         // when appending characters, no need to update tables
1641         d->text_.push_back(c);
1642         setFont(d->text_.size() - 1, font);
1643         d->requestSpellCheck(d->text_.size() - 1);
1644 }
1645
1646
1647 void Paragraph::appendString(docstring const & s, Font const & font,
1648                 Change const & change)
1649 {
1650         pos_type end = s.size();
1651         size_t oldsize = d->text_.size();
1652         size_t newsize = oldsize + end;
1653         size_t capacity = d->text_.capacity();
1654         if (newsize >= capacity)
1655                 d->text_.reserve(max(capacity + 100, newsize));
1656
1657         // when appending characters, no need to update tables
1658         d->text_.append(s);
1659
1660         // FIXME: Optimize this!
1661         for (size_t i = oldsize; i != newsize; ++i) {
1662                 // track change
1663                 d->changes_.insert(change, i);
1664                 d->requestSpellCheck(i);
1665         }
1666         d->fontlist_.set(oldsize, font);
1667         d->fontlist_.set(newsize - 1, font);
1668 }
1669
1670
1671 void Paragraph::insertChar(pos_type pos, char_type c,
1672                            bool trackChanges)
1673 {
1674         d->insertChar(pos, c, Change(trackChanges ?
1675                            Change::INSERTED : Change::UNCHANGED));
1676 }
1677
1678
1679 void Paragraph::insertChar(pos_type pos, char_type c,
1680                            Font const & font, bool trackChanges)
1681 {
1682         d->insertChar(pos, c, Change(trackChanges ?
1683                            Change::INSERTED : Change::UNCHANGED));
1684         setFont(pos, font);
1685 }
1686
1687
1688 void Paragraph::insertChar(pos_type pos, char_type c,
1689                            Font const & font, Change const & change)
1690 {
1691         d->insertChar(pos, c, change);
1692         setFont(pos, font);
1693 }
1694
1695
1696 void Paragraph::resetFonts(Font const & font)
1697 {
1698         d->fontlist_.clear();
1699         d->fontlist_.set(0, font);
1700         d->fontlist_.set(d->text_.size() - 1, font);
1701 }
1702
1703 // Gets uninstantiated font setting at position.
1704 Font const & Paragraph::getFontSettings(BufferParams const & bparams,
1705                                          pos_type pos) const
1706 {
1707         if (pos > size()) {
1708                 LYXERR0("pos: " << pos << " size: " << size());
1709                 LBUFERR(false);
1710         }
1711
1712         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
1713         if (cit != d->fontlist_.end())
1714                 return cit->font();
1715
1716         if (pos == size() && !empty())
1717                 return getFontSettings(bparams, pos - 1);
1718
1719         // Optimisation: avoid a full font instantiation if there is no
1720         // language change from previous call.
1721         static Font previous_font;
1722         static Language const * previous_lang = 0;
1723         Language const * lang = getParLanguage(bparams);
1724         if (lang != previous_lang) {
1725                 previous_lang = lang;
1726                 previous_font = Font(inherit_font, lang);
1727         }
1728         return previous_font;
1729 }
1730
1731
1732 FontSpan Paragraph::fontSpan(pos_type pos) const
1733 {
1734         LBUFERR(pos <= size());
1735
1736         if (pos == size())
1737                 return FontSpan(pos, pos);
1738
1739         pos_type start = 0;
1740         FontList::const_iterator cit = d->fontlist_.begin();
1741         FontList::const_iterator end = d->fontlist_.end();
1742         for (; cit != end; ++cit) {
1743                 if (cit->pos() >= pos) {
1744                         if (pos >= beginOfBody())
1745                                 return FontSpan(max(start, beginOfBody()),
1746                                                 cit->pos());
1747                         else
1748                                 return FontSpan(start,
1749                                                 min(beginOfBody() - 1,
1750                                                          cit->pos()));
1751                 }
1752                 start = cit->pos() + 1;
1753         }
1754
1755         // This should not happen, but if so, we take no chances.
1756         LYXERR0("Paragraph::fontSpan: position not found in fontinfo table!");
1757         LASSERT(false, return FontSpan(pos, pos));
1758 }
1759
1760
1761 // Gets uninstantiated font setting at position 0
1762 Font const & Paragraph::getFirstFontSettings(BufferParams const & bparams) const
1763 {
1764         if (!empty() && !d->fontlist_.empty())
1765                 return d->fontlist_.begin()->font();
1766
1767         // Optimisation: avoid a full font instantiation if there is no
1768         // language change from previous call.
1769         static Font previous_font;
1770         static Language const * previous_lang = 0;
1771         if (bparams.language != previous_lang) {
1772                 previous_lang = bparams.language;
1773                 previous_font = Font(inherit_font, bparams.language);
1774         }
1775
1776         return previous_font;
1777 }
1778
1779
1780 // Gets the fully instantiated font at a given position in a paragraph
1781 // This is basically the same function as Text::GetFont() in text2.cpp.
1782 // The difference is that this one is used for generating the LaTeX file,
1783 // and thus cosmetic "improvements" are disallowed: This has to deliver
1784 // the true picture of the buffer. (Asger)
1785 Font const Paragraph::getFont(BufferParams const & bparams, pos_type pos,
1786                                  Font const & outerfont) const
1787 {
1788         LBUFERR(pos >= 0);
1789
1790         Font font = getFontSettings(bparams, pos);
1791
1792         pos_type const body_pos = beginOfBody();
1793         FontInfo & fi = font.fontInfo();
1794         if (pos < body_pos)
1795                 fi.realize(d->layout_->labelfont);
1796         else
1797                 fi.realize(d->layout_->font);
1798
1799         fi.realize(outerfont.fontInfo());
1800         fi.realize(bparams.getFont().fontInfo());
1801
1802         return font;
1803 }
1804
1805
1806 Font const Paragraph::getLabelFont
1807         (BufferParams const & bparams, Font const & outerfont) const
1808 {
1809         FontInfo tmpfont = d->layout_->labelfont;
1810         tmpfont.realize(outerfont.fontInfo());
1811         tmpfont.realize(bparams.getFont().fontInfo());
1812         return Font(tmpfont, getParLanguage(bparams));
1813 }
1814
1815
1816 Font const Paragraph::getLayoutFont
1817         (BufferParams const & bparams, Font const & outerfont) const
1818 {
1819         FontInfo tmpfont = d->layout_->font;
1820         tmpfont.realize(outerfont.fontInfo());
1821         tmpfont.realize(bparams.getFont().fontInfo());
1822         return Font(tmpfont, getParLanguage(bparams));
1823 }
1824
1825
1826 /// Returns the height of the highest font in range
1827 FontSize Paragraph::highestFontInRange
1828         (pos_type startpos, pos_type endpos, FontSize def_size) const
1829 {
1830         return d->fontlist_.highestInRange(startpos, endpos, def_size);
1831 }
1832
1833
1834 char_type Paragraph::getUChar(BufferParams const & bparams, pos_type pos) const
1835 {
1836         char_type c = d->text_[pos];
1837         if (!getFontSettings(bparams, pos).isRightToLeft())
1838                 return c;
1839
1840         // FIXME: The arabic special casing is due to the difference of arabic
1841         // round brackets input introduced in r18599. Check if this should be
1842         // unified with Hebrew or at least if all bracket types should be
1843         // handled the same (file format change in either case).
1844         string const & lang = getFontSettings(bparams, pos).language()->lang();
1845         bool const arabic = lang == "arabic_arabtex" || lang == "arabic_arabi"
1846                 || lang == "farsi";
1847         char_type uc = c;
1848         switch (c) {
1849         case '(':
1850                 uc = arabic ? c : ')';
1851                 break;
1852         case ')':
1853                 uc = arabic ? c : '(';
1854                 break;
1855         case '[':
1856                 uc = ']';
1857                 break;
1858         case ']':
1859                 uc = '[';
1860                 break;
1861         case '{':
1862                 uc = '}';
1863                 break;
1864         case '}':
1865                 uc = '{';
1866                 break;
1867         case '<':
1868                 uc = '>';
1869                 break;
1870         case '>':
1871                 uc = '<';
1872                 break;
1873         }
1874
1875         return uc;
1876 }
1877
1878
1879 void Paragraph::setFont(pos_type pos, Font const & font)
1880 {
1881         LASSERT(pos <= size(), return);
1882
1883         // First, reduce font against layout/label font
1884         // Update: The setCharFont() routine in text2.cpp already
1885         // reduces font, so we don't need to do that here. (Asger)
1886
1887         d->fontlist_.set(pos, font);
1888 }
1889
1890
1891 void Paragraph::makeSameLayout(Paragraph const & par)
1892 {
1893         d->layout_ = par.d->layout_;
1894         d->params_ = par.d->params_;
1895 }
1896
1897
1898 bool Paragraph::stripLeadingSpaces(bool trackChanges)
1899 {
1900         if (isFreeSpacing())
1901                 return false;
1902
1903         int pos = 0;
1904         int count = 0;
1905
1906         while (pos < size() && (isNewline(pos) || isLineSeparator(pos))) {
1907                 if (eraseChar(pos, trackChanges))
1908                         ++count;
1909                 else
1910                         ++pos;
1911         }
1912
1913         return count > 0 || pos > 0;
1914 }
1915
1916
1917 bool Paragraph::hasSameLayout(Paragraph const & par) const
1918 {
1919         return par.d->layout_ == d->layout_
1920                 && d->params_.sameLayout(par.d->params_);
1921 }
1922
1923
1924 depth_type Paragraph::getDepth() const
1925 {
1926         return d->params_.depth();
1927 }
1928
1929
1930 depth_type Paragraph::getMaxDepthAfter() const
1931 {
1932         if (d->layout_->isEnvironment())
1933                 return d->params_.depth() + 1;
1934         else
1935                 return d->params_.depth();
1936 }
1937
1938
1939 char Paragraph::getAlign() const
1940 {
1941         if (d->params_.align() == LYX_ALIGN_LAYOUT)
1942                 return d->layout_->align;
1943         else
1944                 return d->params_.align();
1945 }
1946
1947
1948 docstring const & Paragraph::labelString() const
1949 {
1950         return d->params_.labelString();
1951 }
1952
1953
1954 // the next two functions are for the manual labels
1955 docstring const Paragraph::getLabelWidthString() const
1956 {
1957         if (d->layout_->margintype == MARGIN_MANUAL
1958             || d->layout_->latextype == LATEX_BIB_ENVIRONMENT)
1959                 return d->params_.labelWidthString();
1960         else
1961                 return _("Senseless with this layout!");
1962 }
1963
1964
1965 void Paragraph::setLabelWidthString(docstring const & s)
1966 {
1967         d->params_.labelWidthString(s);
1968 }
1969
1970
1971 docstring Paragraph::expandLabel(Layout const & layout,
1972                 BufferParams const & bparams) const
1973 {
1974         return expandParagraphLabel(layout, bparams, true);
1975 }
1976
1977
1978 docstring Paragraph::expandDocBookLabel(Layout const & layout,
1979                 BufferParams const & bparams) const
1980 {
1981         return expandParagraphLabel(layout, bparams, false);
1982 }
1983
1984
1985 docstring Paragraph::expandParagraphLabel(Layout const & layout,
1986                 BufferParams const & bparams, bool process_appendix) const
1987 {
1988         DocumentClass const & tclass = bparams.documentClass();
1989         string const & lang = getParLanguage(bparams)->code();
1990         bool const in_appendix = process_appendix && d->params_.appendix();
1991         docstring fmt = translateIfPossible(layout.labelstring(in_appendix), lang);
1992
1993         if (fmt.empty() && !layout.counter.empty())
1994                 return tclass.counters().theCounter(layout.counter, lang);
1995
1996         // handle 'inherited level parts' in 'fmt',
1997         // i.e. the stuff between '@' in   '@Section@.\arabic{subsection}'
1998         size_t const i = fmt.find('@', 0);
1999         if (i != docstring::npos) {
2000                 size_t const j = fmt.find('@', i + 1);
2001                 if (j != docstring::npos) {
2002                         docstring parent(fmt, i + 1, j - i - 1);
2003                         docstring label = from_ascii("??");
2004                         if (tclass.hasLayout(parent))
2005                                 label = expandParagraphLabel(tclass[parent], bparams,
2006                                                       process_appendix);
2007                         fmt = docstring(fmt, 0, i) + label
2008                                 + docstring(fmt, j + 1, docstring::npos);
2009                 }
2010         }
2011
2012         return tclass.counters().counterLabel(fmt, lang);
2013 }
2014
2015
2016 void Paragraph::applyLayout(Layout const & new_layout)
2017 {
2018         d->layout_ = &new_layout;
2019         LyXAlignment const oldAlign = d->params_.align();
2020
2021         if (!(oldAlign & d->layout_->alignpossible)) {
2022                 frontend::Alert::warning(_("Alignment not permitted"),
2023                         _("The new layout does not permit the alignment previously used.\nSetting to default."));
2024                 d->params_.align(LYX_ALIGN_LAYOUT);
2025         }
2026 }
2027
2028
2029 pos_type Paragraph::beginOfBody() const
2030 {
2031         return d->begin_of_body_;
2032 }
2033
2034
2035 void Paragraph::setBeginOfBody()
2036 {
2037         if (d->layout_->labeltype != LABEL_MANUAL) {
2038                 d->begin_of_body_ = 0;
2039                 return;
2040         }
2041
2042         // Unroll the first two cycles of the loop
2043         // and remember the previous character to
2044         // remove unnecessary getChar() calls
2045         pos_type i = 0;
2046         pos_type end = size();
2047         if (i < end && !(isNewline(i) || isEnvSeparator(i))) {
2048                 ++i;
2049                 if (i < end) {
2050                         char_type previous_char = d->text_[i];
2051                         if (!(isNewline(i) || isEnvSeparator(i))) {
2052                                 ++i;
2053                                 while (i < end && previous_char != ' ') {
2054                                         char_type temp = d->text_[i];
2055                                         if (isNewline(i) || isEnvSeparator(i))
2056                                                 break;
2057                                         ++i;
2058                                         previous_char = temp;
2059                                 }
2060                         }
2061                 }
2062         }
2063
2064         d->begin_of_body_ = i;
2065 }
2066
2067
2068 bool Paragraph::allowParagraphCustomization() const
2069 {
2070         return inInset().allowParagraphCustomization();
2071 }
2072
2073
2074 bool Paragraph::usePlainLayout() const
2075 {
2076         return inInset().usePlainLayout();
2077 }
2078
2079
2080 bool Paragraph::isPassThru() const
2081 {
2082         return inInset().isPassThru() || d->layout_->pass_thru;
2083 }
2084
2085 namespace {
2086
2087 // paragraphs inside floats need different alignment tags to avoid
2088 // unwanted space
2089
2090 bool noTrivlistCentering(InsetCode code)
2091 {
2092         return code == FLOAT_CODE
2093                || code == WRAP_CODE
2094                || code == CELL_CODE;
2095 }
2096
2097
2098 string correction(string const & orig)
2099 {
2100         if (orig == "flushleft")
2101                 return "raggedright";
2102         if (orig == "flushright")
2103                 return "raggedleft";
2104         if (orig == "center")
2105                 return "centering";
2106         return orig;
2107 }
2108
2109
2110 bool corrected_env(otexstream & os, string const & suffix, string const & env,
2111         InsetCode code, bool const lastpar, int & col)
2112 {
2113         string macro = suffix + "{";
2114         if (noTrivlistCentering(code)) {
2115                 if (lastpar) {
2116                         // the last paragraph in non-trivlist-aligned
2117                         // context is special (to avoid unwanted whitespace)
2118                         if (suffix == "\\begin") {
2119                                 macro = "\\" + correction(env) + "{}";
2120                                 os << from_ascii(macro);
2121                                 col += macro.size();
2122                                 return true;
2123                         }
2124                         return false;
2125                 }
2126                 macro += correction(env);
2127         } else
2128                 macro += env;
2129         macro += "}";
2130         if (suffix == "\\par\\end") {
2131                 os << breakln;
2132                 col = 0;
2133         }
2134         os << from_ascii(macro);
2135         col += macro.size();
2136         if (suffix == "\\begin") {
2137                 os << breakln;
2138                 col = 0;
2139         }
2140         return true;
2141 }
2142
2143 } // namespace anon
2144
2145
2146 int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
2147                         otexstream & os, OutputParams const & runparams) const
2148 {
2149         int column = 0;
2150
2151         bool canindent =
2152                 (bparams.paragraph_separation == BufferParams::ParagraphIndentSeparation) ?
2153                         (layout_->toggle_indent != ITOGGLE_NEVER) :
2154                         (layout_->toggle_indent == ITOGGLE_ALWAYS);
2155
2156         if (canindent && params_.noindent() && !layout_->pass_thru) {
2157                 os << "\\noindent ";
2158                 column += 10;
2159         }
2160
2161         LyXAlignment const curAlign = params_.align();
2162
2163         if (curAlign == layout_->align)
2164                 return column;
2165
2166         switch (curAlign) {
2167         case LYX_ALIGN_NONE:
2168         case LYX_ALIGN_BLOCK:
2169         case LYX_ALIGN_LAYOUT:
2170         case LYX_ALIGN_SPECIAL:
2171         case LYX_ALIGN_DECIMAL:
2172                 break;
2173         case LYX_ALIGN_LEFT:
2174         case LYX_ALIGN_RIGHT:
2175         case LYX_ALIGN_CENTER:
2176                 if (runparams.moving_arg) {
2177                         os << "\\protect";
2178                         column += 8;
2179                 }
2180                 break;
2181         }
2182
2183         string const begin_tag = "\\begin";
2184         InsetCode code = ownerCode();
2185         bool const lastpar = runparams.isLastPar;
2186
2187         switch (curAlign) {
2188         case LYX_ALIGN_NONE:
2189         case LYX_ALIGN_BLOCK:
2190         case LYX_ALIGN_LAYOUT:
2191         case LYX_ALIGN_SPECIAL:
2192         case LYX_ALIGN_DECIMAL:
2193                 break;
2194         case LYX_ALIGN_LEFT: {
2195                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2196                         corrected_env(os, begin_tag, "flushleft", code, lastpar, column);
2197                 else
2198                         corrected_env(os, begin_tag, "flushright", code, lastpar, column);
2199                 break;
2200         } case LYX_ALIGN_RIGHT: {
2201                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2202                         corrected_env(os, begin_tag, "flushright", code, lastpar, column);
2203                 else
2204                         corrected_env(os, begin_tag, "flushleft", code, lastpar, column);
2205                 break;
2206         } case LYX_ALIGN_CENTER: {
2207                 corrected_env(os, begin_tag, "center", code, lastpar, column);
2208                 break;
2209         }
2210         }
2211
2212         return column;
2213 }
2214
2215
2216 bool Paragraph::Private::endTeXParParams(BufferParams const & bparams,
2217                         otexstream & os, OutputParams const & runparams) const
2218 {
2219         LyXAlignment const curAlign = params_.align();
2220
2221         if (curAlign == layout_->align)
2222                 return false;
2223
2224         switch (curAlign) {
2225         case LYX_ALIGN_NONE:
2226         case LYX_ALIGN_BLOCK:
2227         case LYX_ALIGN_LAYOUT:
2228         case LYX_ALIGN_SPECIAL:
2229         case LYX_ALIGN_DECIMAL:
2230                 break;
2231         case LYX_ALIGN_LEFT:
2232         case LYX_ALIGN_RIGHT:
2233         case LYX_ALIGN_CENTER:
2234                 if (runparams.moving_arg)
2235                         os << "\\protect";
2236                 break;
2237         }
2238
2239         bool output = false;
2240         int col = 0;
2241         string const end_tag = "\\par\\end";
2242         InsetCode code = ownerCode();
2243         bool const lastpar = runparams.isLastPar;
2244
2245         switch (curAlign) {
2246         case LYX_ALIGN_NONE:
2247         case LYX_ALIGN_BLOCK:
2248         case LYX_ALIGN_LAYOUT:
2249         case LYX_ALIGN_SPECIAL:
2250         case LYX_ALIGN_DECIMAL:
2251                 break;
2252         case LYX_ALIGN_LEFT: {
2253                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2254                         output = corrected_env(os, end_tag, "flushleft", code, lastpar, col);
2255                 else
2256                         output = corrected_env(os, end_tag, "flushright", code, lastpar, col);
2257                 break;
2258         } case LYX_ALIGN_RIGHT: {
2259                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2260                         output = corrected_env(os, end_tag, "flushright", code, lastpar, col);
2261                 else
2262                         output = corrected_env(os, end_tag, "flushleft", code, lastpar, col);
2263                 break;
2264         } case LYX_ALIGN_CENTER: {
2265                 corrected_env(os, end_tag, "center", code, lastpar, col);
2266                 break;
2267         }
2268         }
2269
2270         return output || lastpar;
2271 }
2272
2273
2274 // This one spits out the text of the paragraph
2275 void Paragraph::latex(BufferParams const & bparams,
2276         Font const & outerfont,
2277         otexstream & os,
2278         OutputParams const & runparams,
2279         int start_pos, int end_pos, bool force) const
2280 {
2281         LYXERR(Debug::LATEX, "Paragraph::latex...     " << this);
2282
2283         // FIXME This check should not be needed. Perhaps issue an
2284         // error if it triggers.
2285         Layout const & style = inInset().forcePlainLayout() ?
2286                 bparams.documentClass().plainLayout() : *d->layout_;
2287
2288         if (!force && style.inpreamble)
2289                 return;
2290
2291         bool const allowcust = allowParagraphCustomization();
2292
2293         // Current base font for all inherited font changes, without any
2294         // change caused by an individual character, except for the language:
2295         // It is set to the language of the first character.
2296         // As long as we are in the label, this font is the base font of the
2297         // label. Before the first body character it is set to the base font
2298         // of the body.
2299         Font basefont;
2300
2301         // Maybe we have to create a optional argument.
2302         pos_type body_pos = beginOfBody();
2303         unsigned int column = 0;
2304
2305         if (body_pos > 0) {
2306                 // the optional argument is kept in curly brackets in
2307                 // case it contains a ']'
2308                 // This is not strictly needed, but if this is changed it
2309                 // would be a file format change, and tex2lyx would need
2310                 // to be adjusted, since it unconditionally removes the
2311                 // braces when it parses \item.
2312                 os << "[{";
2313                 column += 2;
2314                 basefont = getLabelFont(bparams, outerfont);
2315         } else {
2316                 basefont = getLayoutFont(bparams, outerfont);
2317         }
2318
2319         // Which font is currently active?
2320         Font running_font(basefont);
2321         // Do we have an open font change?
2322         bool open_font = false;
2323
2324         Change runningChange = Change(Change::UNCHANGED);
2325
2326         Encoding const * const prev_encoding = runparams.encoding;
2327
2328         os.texrow().start(id(), 0);
2329
2330         // if the paragraph is empty, the loop will not be entered at all
2331         if (empty()) {
2332                 if (style.isCommand()) {
2333                         os << '{';
2334                         ++column;
2335                 }
2336                 if (!style.leftdelim().empty()) {
2337                         os << style.leftdelim();
2338                         column += style.leftdelim().size();
2339                 }
2340                 if (allowcust)
2341                         column += d->startTeXParParams(bparams, os, runparams);
2342         }
2343
2344         for (pos_type i = 0; i < size(); ++i) {
2345                 // First char in paragraph or after label?
2346                 if (i == body_pos) {
2347                         if (body_pos > 0) {
2348                                 if (open_font) {
2349                                         column += running_font.latexWriteEndChanges(
2350                                                 os, bparams, runparams,
2351                                                 basefont, basefont);
2352                                         open_font = false;
2353                                 }
2354                                 basefont = getLayoutFont(bparams, outerfont);
2355                                 running_font = basefont;
2356
2357                                 column += Changes::latexMarkChange(os, bparams,
2358                                                 runningChange, Change(Change::UNCHANGED),
2359                                                 runparams);
2360                                 runningChange = Change(Change::UNCHANGED);
2361
2362                                 os << "}] ";
2363                                 column +=3;
2364                         }
2365                         if (style.isCommand()) {
2366                                 os << '{';
2367                                 ++column;
2368                         }
2369
2370                         if (!style.leftdelim().empty()) {
2371                                 os << style.leftdelim();
2372                                 column += style.leftdelim().size();
2373                         }
2374
2375                         if (allowcust)
2376                                 column += d->startTeXParParams(bparams, os,
2377                                                             runparams);
2378                 }
2379
2380                 Change const & change = runparams.inDeletedInset
2381                         ? runparams.changeOfDeletedInset : lookupChange(i);
2382
2383                 if (bparams.output_changes && runningChange != change) {
2384                         if (open_font) {
2385                                 column += running_font.latexWriteEndChanges(
2386                                                 os, bparams, runparams, basefont, basefont);
2387                                 open_font = false;
2388                         }
2389                         basefont = getLayoutFont(bparams, outerfont);
2390                         running_font = basefont;
2391
2392                         column += Changes::latexMarkChange(os, bparams, runningChange,
2393                                                            change, runparams);
2394                         runningChange = change;
2395                 }
2396
2397                 // do not output text which is marked deleted
2398                 // if change tracking output is disabled
2399                 if (!bparams.output_changes && change.deleted()) {
2400                         continue;
2401                 }
2402
2403                 ++column;
2404
2405                 // Fully instantiated font
2406                 Font const font = getFont(bparams, i, outerfont);
2407
2408                 Font const last_font = running_font;
2409
2410                 // Do we need to close the previous font?
2411                 if (open_font &&
2412                     (font != running_font ||
2413                      font.language() != running_font.language()))
2414                 {
2415                         column += running_font.latexWriteEndChanges(
2416                                         os, bparams, runparams, basefont,
2417                                         (i == body_pos-1) ? basefont : font);
2418                         running_font = basefont;
2419                         open_font = false;
2420                 }
2421
2422                 string const running_lang = runparams.use_polyglossia ?
2423                         running_font.language()->polyglossia() : running_font.language()->babel();
2424                 // close babel's font environment before opening CJK.
2425                 string const lang_end_command = runparams.use_polyglossia ?
2426                         "\\end{$$lang}" : lyxrc.language_command_end;
2427                 if (!running_lang.empty() &&
2428                     font.language()->encoding()->package() == Encoding::CJK) {
2429                                 string end_tag = subst(lang_end_command,
2430                                                         "$$lang",
2431                                                         running_lang);
2432                                 os << from_ascii(end_tag);
2433                                 column += end_tag.length();
2434                 }
2435
2436                 // Switch file encoding if necessary (and allowed)
2437                 if (!runparams.pass_thru && !style.pass_thru &&
2438                     runparams.encoding->package() != Encoding::none &&
2439                     font.language()->encoding()->package() != Encoding::none) {
2440                         pair<bool, int> const enc_switch =
2441                                 switchEncoding(os.os(), bparams, runparams,
2442                                         *(font.language()->encoding()));
2443                         if (enc_switch.first) {
2444                                 column += enc_switch.second;
2445                                 runparams.encoding = font.language()->encoding();
2446                         }
2447                 }
2448
2449                 char_type const c = d->text_[i];
2450
2451                 // Do we need to change font?
2452                 if ((font != running_font ||
2453                      font.language() != running_font.language()) &&
2454                         i != body_pos - 1)
2455                 {
2456                         odocstringstream ods;
2457                         column += font.latexWriteStartChanges(ods, bparams,
2458                                                               runparams, basefont,
2459                                                               last_font);
2460                         running_font = font;
2461                         open_font = true;
2462                         docstring fontchange = ods.str();
2463                         // check whether the fontchange ends with a \\textcolor
2464                         // modifier and the text starts with a space (bug 4473)
2465                         docstring const last_modifier = rsplit(fontchange, '\\');
2466                         if (prefixIs(last_modifier, from_ascii("textcolor")) && c == ' ')
2467                                 os << fontchange << from_ascii("{}");
2468                         // check if the fontchange ends with a trailing blank
2469                         // (like "\small " (see bug 3382)
2470                         else if (suffixIs(fontchange, ' ') && c == ' ')
2471                                 os << fontchange.substr(0, fontchange.size() - 1)
2472                                    << from_ascii("{}");
2473                         else
2474                                 os << fontchange;
2475                 }
2476
2477                 // FIXME: think about end_pos implementation...
2478                 if (c == ' ' && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2479                         // FIXME: integrate this case in latexSpecialChar
2480                         // Do not print the separation of the optional argument
2481                         // if style.pass_thru is false. This works because
2482                         // latexSpecialChar ignores spaces if
2483                         // style.pass_thru is false.
2484                         if (i != body_pos - 1) {
2485                                 if (d->simpleTeXBlanks(runparams, os,
2486                                                 i, column, font, style)) {
2487                                         // A surrogate pair was output. We
2488                                         // must not call latexSpecialChar
2489                                         // in this iteration, since it would output
2490                                         // the combining character again.
2491                                         ++i;
2492                                         continue;
2493                                 }
2494                         }
2495                 }
2496
2497                 OutputParams rp = runparams;
2498                 rp.free_spacing = style.free_spacing;
2499                 rp.local_font = &font;
2500                 rp.intitle = style.intitle;
2501
2502                 // Two major modes:  LaTeX or plain
2503                 // Handle here those cases common to both modes
2504                 // and then split to handle the two modes separately.
2505                 if (c == META_INSET) {
2506                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2507                                 d->latexInset(bparams, os, rp, running_font,
2508                                                 basefont, outerfont, open_font,
2509                                                 runningChange, style, i, column);
2510                         }
2511                 } else {
2512                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2513                                 try {
2514                                         d->latexSpecialChar(os, bparams, rp, running_font, runningChange,
2515                                                             style, i, end_pos, column);
2516                                 } catch (EncodingException & e) {
2517                                 if (runparams.dryrun) {
2518                                         os << "<" << _("LyX Warning: ")
2519                                            << _("uncodable character") << " '";
2520                                         os.put(c);
2521                                         os << "'>";
2522                                 } else {
2523                                         // add location information and throw again.
2524                                         e.par_id = id();
2525                                         e.pos = i;
2526                                         throw(e);
2527                                 }
2528                         }
2529                 }
2530                 }
2531
2532                 // Set the encoding to that returned from latexSpecialChar (see
2533                 // comment for encoding member in OutputParams.h)
2534                 runparams.encoding = rp.encoding;
2535         }
2536
2537         // If we have an open font definition, we have to close it
2538         if (open_font) {
2539 #ifdef FIXED_LANGUAGE_END_DETECTION
2540                 if (next_) {
2541                         running_font.latexWriteEndChanges(os, bparams,
2542                                         runparams, basefont,
2543                                         next_->getFont(bparams, 0, outerfont));
2544                 } else {
2545                         running_font.latexWriteEndChanges(os, bparams,
2546                                         runparams, basefont, basefont);
2547                 }
2548 #else
2549 //FIXME: For now we ALWAYS have to close the foreign font settings if they are
2550 //FIXME: there as we start another \selectlanguage with the next paragraph if
2551 //FIXME: we are in need of this. This should be fixed sometime (Jug)
2552                 running_font.latexWriteEndChanges(os, bparams, runparams,
2553                                 basefont, basefont);
2554 #endif
2555         }
2556
2557         column += Changes::latexMarkChange(os, bparams, runningChange,
2558                                            Change(Change::UNCHANGED), runparams);
2559
2560         // Needed if there is an optional argument but no contents.
2561         if (body_pos > 0 && body_pos == size()) {
2562                 os << "}]~";
2563         }
2564
2565         if (!style.rightdelim().empty()) {
2566                 os << style.rightdelim();
2567                 column += style.rightdelim().size();
2568         }
2569
2570         if (allowcust && d->endTeXParParams(bparams, os, runparams)
2571             && runparams.encoding != prev_encoding) {
2572                 runparams.encoding = prev_encoding;
2573                 os << setEncoding(prev_encoding->iconvName());
2574         }
2575
2576         LYXERR(Debug::LATEX, "Paragraph::latex... done " << this);
2577 }
2578
2579
2580 bool Paragraph::emptyTag() const
2581 {
2582         for (pos_type i = 0; i < size(); ++i) {
2583                 if (Inset const * inset = getInset(i)) {
2584                         InsetCode lyx_code = inset->lyxCode();
2585                         // FIXME testing like that is wrong. What is
2586                         // the intent?
2587                         if (lyx_code != TOC_CODE &&
2588                             lyx_code != INCLUDE_CODE &&
2589                             lyx_code != GRAPHICS_CODE &&
2590                             lyx_code != ERT_CODE &&
2591                             lyx_code != LISTINGS_CODE &&
2592                             lyx_code != FLOAT_CODE &&
2593                             lyx_code != TABULAR_CODE) {
2594                                 return false;
2595                         }
2596                 } else {
2597                         char_type c = d->text_[i];
2598                         if (c != ' ' && c != '\t')
2599                                 return false;
2600                 }
2601         }
2602         return true;
2603 }
2604
2605
2606 string Paragraph::getID(Buffer const & buf, OutputParams const & runparams)
2607         const
2608 {
2609         for (pos_type i = 0; i < size(); ++i) {
2610                 if (Inset const * inset = getInset(i)) {
2611                         InsetCode lyx_code = inset->lyxCode();
2612                         if (lyx_code == LABEL_CODE) {
2613                                 InsetLabel const * const il = static_cast<InsetLabel const *>(inset);
2614                                 docstring const & id = il->getParam("name");
2615                                 return "id='" + to_utf8(sgml::cleanID(buf, runparams, id)) + "'";
2616                         }
2617                 }
2618         }
2619         return string();
2620 }
2621
2622
2623 pos_type Paragraph::firstWordDocBook(odocstream & os, OutputParams const & runparams)
2624         const
2625 {
2626         pos_type i;
2627         for (i = 0; i < size(); ++i) {
2628                 if (Inset const * inset = getInset(i)) {
2629                         inset->docbook(os, runparams);
2630                 } else {
2631                         char_type c = d->text_[i];
2632                         if (c == ' ')
2633                                 break;
2634                         os << sgml::escapeChar(c);
2635                 }
2636         }
2637         return i;
2638 }
2639
2640
2641 pos_type Paragraph::firstWordLyXHTML(XHTMLStream & xs, OutputParams const & runparams)
2642         const
2643 {
2644         pos_type i;
2645         for (i = 0; i < size(); ++i) {
2646                 if (Inset const * inset = getInset(i)) {
2647                         inset->xhtml(xs, runparams);
2648                 } else {
2649                         char_type c = d->text_[i];
2650                         if (c == ' ')
2651                                 break;
2652                         xs << c;
2653                 }
2654         }
2655         return i;
2656 }
2657
2658
2659 bool Paragraph::Private::onlyText(Buffer const & buf, Font const & outerfont, pos_type initial) const
2660 {
2661         Font font_old;
2662         pos_type size = text_.size();
2663         for (pos_type i = initial; i < size; ++i) {
2664                 Font font = owner_->getFont(buf.params(), i, outerfont);
2665                 if (text_[i] == META_INSET)
2666                         return false;
2667                 if (i != initial && font != font_old)
2668                         return false;
2669                 font_old = font;
2670         }
2671
2672         return true;
2673 }
2674
2675
2676 void Paragraph::simpleDocBookOnePar(Buffer const & buf,
2677                                     odocstream & os,
2678                                     OutputParams const & runparams,
2679                                     Font const & outerfont,
2680                                     pos_type initial) const
2681 {
2682         bool emph_flag = false;
2683
2684         Layout const & style = *d->layout_;
2685         FontInfo font_old =
2686                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2687
2688         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2689                 os << "]]>";
2690
2691         // parsing main loop
2692         for (pos_type i = initial; i < size(); ++i) {
2693                 Font font = getFont(buf.params(), i, outerfont);
2694
2695                 // handle <emphasis> tag
2696                 if (font_old.emph() != font.fontInfo().emph()) {
2697                         if (font.fontInfo().emph() == FONT_ON) {
2698                                 os << "<emphasis>";
2699                                 emph_flag = true;
2700                         } else if (i != initial) {
2701                                 os << "</emphasis>";
2702                                 emph_flag = false;
2703                         }
2704                 }
2705
2706                 if (Inset const * inset = getInset(i)) {
2707                         inset->docbook(os, runparams);
2708                 } else {
2709                         char_type c = d->text_[i];
2710
2711                         if (style.pass_thru)
2712                                 os.put(c);
2713                         else
2714                                 os << sgml::escapeChar(c);
2715                 }
2716                 font_old = font.fontInfo();
2717         }
2718
2719         if (emph_flag) {
2720                 os << "</emphasis>";
2721         }
2722
2723         if (style.free_spacing)
2724                 os << '\n';
2725         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2726                 os << "<![CDATA[";
2727 }
2728
2729
2730 namespace {
2731 void doFontSwitch(vector<html::FontTag> & tagsToOpen,
2732                   vector<html::EndFontTag> & tagsToClose,
2733                   bool & flag, FontState curstate, html::FontTypes type)
2734 {
2735         if (curstate == FONT_ON) {
2736                 tagsToOpen.push_back(html::FontTag(type));
2737                 flag = true;
2738         } else if (flag) {
2739                 tagsToClose.push_back(html::EndFontTag(type));
2740                 flag = false;
2741         }
2742 }
2743 }
2744
2745
2746 docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
2747                                     XHTMLStream & xs,
2748                                     OutputParams const & runparams,
2749                                     Font const & outerfont,
2750                                     pos_type initial) const
2751 {
2752         docstring retval;
2753
2754         // track whether we have opened these tags
2755         bool emph_flag = false;
2756         bool bold_flag = false;
2757         bool noun_flag = false;
2758         bool ubar_flag = false;
2759         bool dbar_flag = false;
2760         bool sout_flag = false;
2761         bool wave_flag = false;
2762         // shape tags
2763         bool shap_flag = false;
2764         // family tags
2765         bool faml_flag = false;
2766         // size tags
2767         bool size_flag = false;
2768
2769         Layout const & style = *d->layout_;
2770
2771         xs.startParagraph(allowEmpty());
2772
2773         FontInfo font_old =
2774                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2775
2776         FontShape  curr_fs   = INHERIT_SHAPE;
2777         FontFamily curr_fam  = INHERIT_FAMILY;
2778         FontSize   curr_size = FONT_SIZE_INHERIT;
2779
2780         string const default_family =
2781                 buf.masterBuffer()->params().fonts_default_family;
2782
2783         vector<html::FontTag> tagsToOpen;
2784         vector<html::EndFontTag> tagsToClose;
2785
2786         // parsing main loop
2787         for (pos_type i = initial; i < size(); ++i) {
2788                 // let's not show deleted material in the output
2789                 if (isDeleted(i))
2790                         continue;
2791
2792                 Font const font = getFont(buf.masterBuffer()->params(), i, outerfont);
2793
2794                 // emphasis
2795                 FontState curstate = font.fontInfo().emph();
2796                 if (font_old.emph() != curstate)
2797                         doFontSwitch(tagsToOpen, tagsToClose, emph_flag, curstate, html::FT_EMPH);
2798
2799                 // noun
2800                 curstate = font.fontInfo().noun();
2801                 if (font_old.noun() != curstate)
2802                         doFontSwitch(tagsToOpen, tagsToClose, noun_flag, curstate, html::FT_NOUN);
2803
2804                 // underbar
2805                 curstate = font.fontInfo().underbar();
2806                 if (font_old.underbar() != curstate)
2807                         doFontSwitch(tagsToOpen, tagsToClose, ubar_flag, curstate, html::FT_UBAR);
2808
2809                 // strikeout
2810                 curstate = font.fontInfo().strikeout();
2811                 if (font_old.strikeout() != curstate)
2812                         doFontSwitch(tagsToOpen, tagsToClose, sout_flag, curstate, html::FT_SOUT);
2813
2814                 // double underbar
2815                 curstate = font.fontInfo().uuline();
2816                 if (font_old.uuline() != curstate)
2817                         doFontSwitch(tagsToOpen, tagsToClose, dbar_flag, curstate, html::FT_DBAR);
2818
2819                 // wavy line
2820                 curstate = font.fontInfo().uwave();
2821                 if (font_old.uwave() != curstate)
2822                         doFontSwitch(tagsToOpen, tagsToClose, wave_flag, curstate, html::FT_WAVE);
2823
2824                 // bold
2825                 // a little hackish, but allows us to reuse what we have.
2826                 curstate = (font.fontInfo().series() == BOLD_SERIES ? FONT_ON : FONT_OFF);
2827                 if (font_old.series() != font.fontInfo().series())
2828                         doFontSwitch(tagsToOpen, tagsToClose, bold_flag, curstate, html::FT_BOLD);
2829
2830                 // Font shape
2831                 curr_fs = font.fontInfo().shape();
2832                 FontShape old_fs = font_old.shape();
2833                 if (old_fs != curr_fs) {
2834                         if (shap_flag) {
2835                                 switch (old_fs) {
2836                                 case ITALIC_SHAPE:
2837                                         tagsToClose.push_back(html::EndFontTag(html::FT_ITALIC));
2838                                         break;
2839                                 case SLANTED_SHAPE:
2840                                         tagsToClose.push_back(html::EndFontTag(html::FT_SLANTED));
2841                                         break;
2842                                 case SMALLCAPS_SHAPE:
2843                                         tagsToClose.push_back(html::EndFontTag(html::FT_SMALLCAPS));
2844                                         break;
2845                                 case UP_SHAPE:
2846                                 case INHERIT_SHAPE:
2847                                         break;
2848                                 default:
2849                                         // the other tags are for internal use
2850                                         LATTEST(false);
2851                                         break;
2852                                 }
2853                                 shap_flag = false;
2854                         }
2855                         switch (curr_fs) {
2856                         case ITALIC_SHAPE:
2857                                 tagsToOpen.push_back(html::FontTag(html::FT_ITALIC));
2858                                 shap_flag = true;
2859                                 break;
2860                         case SLANTED_SHAPE:
2861                                 tagsToOpen.push_back(html::FontTag(html::FT_SLANTED));
2862                                 shap_flag = true;
2863                                 break;
2864                         case SMALLCAPS_SHAPE:
2865                                 tagsToOpen.push_back(html::FontTag(html::FT_SMALLCAPS));
2866                                 shap_flag = true;
2867                                 break;
2868                         case UP_SHAPE:
2869                         case INHERIT_SHAPE:
2870                                 break;
2871                         default:
2872                                 // the other tags are for internal use
2873                                 LATTEST(false);
2874                                 break;
2875                         }
2876                 }
2877
2878                 // Font family
2879                 curr_fam = font.fontInfo().family();
2880                 FontFamily old_fam = font_old.family();
2881                 if (old_fam != curr_fam) {
2882                         if (faml_flag) {
2883                                 switch (old_fam) {
2884                                 case ROMAN_FAMILY:
2885                                         tagsToClose.push_back(html::EndFontTag(html::FT_ROMAN));
2886                                         break;
2887                                 case SANS_FAMILY:
2888                                         tagsToClose.push_back(html::EndFontTag(html::FT_SANS));
2889                                         break;
2890                                 case TYPEWRITER_FAMILY:
2891                                         tagsToClose.push_back(html::EndFontTag(html::FT_TYPE));
2892                                         break;
2893                                 case INHERIT_FAMILY:
2894                                         break;
2895                                 default:
2896                                         // the other tags are for internal use
2897                                         LATTEST(false);
2898                                         break;
2899                                 }
2900                                 faml_flag = false;
2901                         }
2902                         switch (curr_fam) {
2903                         case ROMAN_FAMILY:
2904                                 // we will treat a "default" font family as roman, since we have
2905                                 // no other idea what to do.
2906                                 if (default_family != "rmdefault" && default_family != "default") {
2907                                         tagsToOpen.push_back(html::FontTag(html::FT_ROMAN));
2908                                         faml_flag = true;
2909                                 }
2910                                 break;
2911                         case SANS_FAMILY:
2912                                 if (default_family != "sfdefault") {
2913                                         tagsToOpen.push_back(html::FontTag(html::FT_SANS));
2914                                         faml_flag = true;
2915                                 }
2916                                 break;
2917                         case TYPEWRITER_FAMILY:
2918                                 if (default_family != "ttdefault") {
2919                                         tagsToOpen.push_back(html::FontTag(html::FT_TYPE));
2920                                         faml_flag = true;
2921                                 }
2922                                 break;
2923                         case INHERIT_FAMILY:
2924                                 break;
2925                         default:
2926                                 // the other tags are for internal use
2927                                 LATTEST(false);
2928                                 break;
2929                         }
2930                 }
2931
2932                 // Font size
2933                 curr_size = font.fontInfo().size();
2934                 FontSize old_size = font_old.size();
2935                 if (old_size != curr_size) {
2936                         if (size_flag) {
2937                                 switch (old_size) {
2938                                 case FONT_SIZE_TINY:
2939                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_TINY));
2940                                         break;
2941                                 case FONT_SIZE_SCRIPT:
2942                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_SCRIPT));
2943                                         break;
2944                                 case FONT_SIZE_FOOTNOTE:
2945                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_FOOTNOTE));
2946                                         break;
2947                                 case FONT_SIZE_SMALL:
2948                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_SMALL));
2949                                         break;
2950                                 case FONT_SIZE_LARGE:
2951                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGE));
2952                                         break;
2953                                 case FONT_SIZE_LARGER:
2954                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGER));
2955                                         break;
2956                                 case FONT_SIZE_LARGEST:
2957                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGEST));
2958                                         break;
2959                                 case FONT_SIZE_HUGE:
2960                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_HUGE));
2961                                         break;
2962                                 case FONT_SIZE_HUGER:
2963                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_HUGER));
2964                                         break;
2965                                 case FONT_SIZE_INCREASE:
2966                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_INCREASE));
2967                                         break;
2968                                 case FONT_SIZE_DECREASE:
2969                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_DECREASE));
2970                                         break;
2971                                 case FONT_SIZE_INHERIT:
2972                                 case FONT_SIZE_NORMAL:
2973                                         break;
2974                                 default:
2975                                         // the other tags are for internal use
2976                                         LATTEST(false);
2977                                         break;
2978                                 }
2979                                 size_flag = false;
2980                         }
2981                         switch (curr_size) {
2982                         case FONT_SIZE_TINY:
2983                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_TINY));
2984                                 size_flag = true;
2985                                 break;
2986                         case FONT_SIZE_SCRIPT:
2987                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_SCRIPT));
2988                                 size_flag = true;
2989                                 break;
2990                         case FONT_SIZE_FOOTNOTE:
2991                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_FOOTNOTE));
2992                                 size_flag = true;
2993                                 break;
2994                         case FONT_SIZE_SMALL:
2995                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_SMALL));
2996                                 size_flag = true;
2997                                 break;
2998                         case FONT_SIZE_LARGE:
2999                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGE));
3000                                 size_flag = true;
3001                                 break;
3002                         case FONT_SIZE_LARGER:
3003                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGER));
3004                                 size_flag = true;
3005                                 break;
3006                         case FONT_SIZE_LARGEST:
3007                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGEST));
3008                                 size_flag = true;
3009                                 break;
3010                         case FONT_SIZE_HUGE:
3011                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_HUGE));
3012                                 size_flag = true;
3013                                 break;
3014                         case FONT_SIZE_HUGER:
3015                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_HUGER));
3016                                 size_flag = true;
3017                                 break;
3018                         case FONT_SIZE_INCREASE:
3019                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_INCREASE));
3020                                 size_flag = true;
3021                                 break;
3022                         case FONT_SIZE_DECREASE:
3023                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_DECREASE));
3024                                 size_flag = true;
3025                                 break;
3026                         case FONT_SIZE_NORMAL:
3027                         case FONT_SIZE_INHERIT:
3028                                 break;
3029                         default:
3030                                 // the other tags are for internal use
3031                                 LATTEST(false);
3032                                 break;
3033                         }
3034                 }
3035
3036                 // FIXME XHTML
3037                 // Other such tags? What about the other text ranges?
3038
3039                 vector<html::EndFontTag>::const_iterator cit = tagsToClose.begin();
3040                 vector<html::EndFontTag>::const_iterator cen = tagsToClose.end();
3041                 for (; cit != cen; ++cit)
3042                         xs << *cit;
3043
3044                 vector<html::FontTag>::const_iterator sit = tagsToOpen.begin();
3045                 vector<html::FontTag>::const_iterator sen = tagsToOpen.end();
3046                 for (; sit != sen; ++sit)
3047                         xs << *sit;
3048
3049                 tagsToClose.clear();
3050                 tagsToOpen.clear();
3051
3052                 Inset const * inset = getInset(i);
3053                 if (inset) {
3054                         if (!runparams.for_toc || inset->isInToc()) {
3055                                 OutputParams np = runparams;
3056                                 np.local_font = &font;
3057                                 if (!inset->getLayout().htmlisblock())
3058                                         np.html_in_par = true;
3059                                 retval += inset->xhtml(xs, np);
3060                         }
3061                 } else {
3062                         char_type c = getUChar(buf.masterBuffer()->params(), i);
3063                         xs << c;
3064                 }
3065                 font_old = font.fontInfo();
3066         }
3067
3068         xs.closeFontTags();
3069         xs.endParagraph();
3070         return retval;
3071 }
3072
3073
3074 bool Paragraph::isHfill(pos_type pos) const
3075 {
3076         Inset const * inset = getInset(pos);
3077         return inset && inset->isHfill();
3078 }
3079
3080
3081 bool Paragraph::isNewline(pos_type pos) const
3082 {
3083         Inset const * inset = getInset(pos);
3084         return inset && inset->lyxCode() == NEWLINE_CODE;
3085 }
3086
3087
3088 bool Paragraph::isEnvSeparator(pos_type pos) const
3089 {
3090         Inset const * inset = getInset(pos);
3091         return inset && inset->lyxCode() == SEPARATOR_CODE;
3092 }
3093
3094
3095 bool Paragraph::isLineSeparator(pos_type pos) const
3096 {
3097         char_type const c = d->text_[pos];
3098         if (isLineSeparatorChar(c))
3099                 return true;
3100         Inset const * inset = getInset(pos);
3101         return inset && inset->isLineSeparator();
3102 }
3103
3104
3105 bool Paragraph::isWordSeparator(pos_type pos) const
3106 {
3107         if (pos == size())
3108                 return true;
3109         if (Inset const * inset = getInset(pos))
3110                 return !inset->isLetter();
3111         // if we have a hard hyphen (no en- or emdash) or apostrophe
3112         // we pass this to the spell checker
3113         // FIXME: this method is subject to change, visit
3114         // https://bugzilla.mozilla.org/show_bug.cgi?id=355178
3115         // to get an impression how complex this is.
3116         if (isHardHyphenOrApostrophe(pos))
3117                 return false;
3118         char_type const c = d->text_[pos];
3119         // We want to pass the escape chars to the spellchecker
3120         docstring const escape_chars = from_utf8(lyxrc.spellchecker_esc_chars);
3121         return !isLetterChar(c) && !isDigitASCII(c) && !contains(escape_chars, c);
3122 }
3123
3124
3125 bool Paragraph::isHardHyphenOrApostrophe(pos_type pos) const
3126 {
3127         pos_type const psize = size();
3128         if (pos >= psize)
3129                 return false;
3130         char_type const c = d->text_[pos];
3131         if (c != '-' && c != '\'')
3132                 return false;
3133         int nextpos = pos + 1;
3134         int prevpos = pos > 0 ? pos - 1 : 0;
3135         if ((nextpos == psize || isSpace(nextpos))
3136                 && (pos == 0 || isSpace(prevpos)))
3137                 return false;
3138         return true;
3139 }
3140
3141
3142 FontSpan const & Paragraph::getSpellRange(pos_type pos) const
3143 {
3144         return d->speller_state_.getRange(pos);
3145 }
3146
3147
3148 bool Paragraph::isChar(pos_type pos) const
3149 {
3150         if (Inset const * inset = getInset(pos))
3151                 return inset->isChar();
3152         char_type const c = d->text_[pos];
3153         return !isLetterChar(c) && !isDigitASCII(c) && !lyx::isSpace(c);
3154 }
3155
3156
3157 bool Paragraph::isSpace(pos_type pos) const
3158 {
3159         if (Inset const * inset = getInset(pos))
3160                 return inset->isSpace();
3161         char_type const c = d->text_[pos];
3162         return lyx::isSpace(c);
3163 }
3164
3165
3166 Language const *
3167 Paragraph::getParLanguage(BufferParams const & bparams) const
3168 {
3169         if (!empty())
3170                 return getFirstFontSettings(bparams).language();
3171         // FIXME: we should check the prev par as well (Lgb)
3172         return bparams.language;
3173 }
3174
3175
3176 bool Paragraph::isRTL(BufferParams const & bparams) const
3177 {
3178         return getParLanguage(bparams)->rightToLeft()
3179                 && !inInset().getLayout().forceLTR();
3180 }
3181
3182
3183 void Paragraph::changeLanguage(BufferParams const & bparams,
3184                                Language const * from, Language const * to)
3185 {
3186         // change language including dummy font change at the end
3187         for (pos_type i = 0; i <= size(); ++i) {
3188                 Font font = getFontSettings(bparams, i);
3189                 if (font.language() == from) {
3190                         font.setLanguage(to);
3191                         setFont(i, font);
3192                         d->requestSpellCheck(i);
3193                 }
3194         }
3195 }
3196
3197
3198 bool Paragraph::isMultiLingual(BufferParams const & bparams) const
3199 {
3200         Language const * doc_language = bparams.language;
3201         FontList::const_iterator cit = d->fontlist_.begin();
3202         FontList::const_iterator end = d->fontlist_.end();
3203
3204         for (; cit != end; ++cit)
3205                 if (cit->font().language() != ignore_language &&
3206                     cit->font().language() != latex_language &&
3207                     cit->font().language() != doc_language)
3208                         return true;
3209         return false;
3210 }
3211
3212
3213 void Paragraph::getLanguages(std::set<Language const *> & languages) const
3214 {
3215         FontList::const_iterator cit = d->fontlist_.begin();
3216         FontList::const_iterator end = d->fontlist_.end();
3217
3218         for (; cit != end; ++cit) {
3219                 Language const * lang = cit->font().language();
3220                 if (lang != ignore_language &&
3221                     lang != latex_language)
3222                         languages.insert(lang);
3223         }
3224 }
3225
3226
3227 docstring Paragraph::asString(int options) const
3228 {
3229         return asString(0, size(), options);
3230 }
3231
3232
3233 docstring Paragraph::asString(pos_type beg, pos_type end, int options, const OutputParams *runparams) const
3234 {
3235         odocstringstream os;
3236
3237         if (beg == 0
3238             && options & AS_STR_LABEL
3239             && !d->params_.labelString().empty())
3240                 os << d->params_.labelString() << ' ';
3241
3242         for (pos_type i = beg; i < end; ++i) {
3243                 if ((options & AS_STR_SKIPDELETE) && isDeleted(i))
3244                         continue;
3245                 char_type const c = d->text_[i];
3246                 if (isPrintable(c) || c == '\t'
3247                     || (c == '\n' && (options & AS_STR_NEWLINES)))
3248                         os.put(c);
3249                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
3250                         if (c == META_INSET && (options & AS_STR_PLAINTEXT)) {
3251                                 LASSERT(runparams != 0, return docstring());
3252                                 getInset(i)->plaintext(os, *runparams);
3253                         } else {
3254                                 getInset(i)->toString(os);
3255                         }
3256                 }
3257         }
3258
3259         return os.str();
3260 }
3261
3262
3263 void Paragraph::forOutliner(docstring & os, size_t const maxlen,
3264                                                         bool const shorten) const
3265 {
3266         size_t tmplen = shorten ? maxlen + 1 : maxlen;
3267         if (!d->params_.labelString().empty())
3268                 os += d->params_.labelString() + ' ';
3269         for (pos_type i = 0; i < size() && os.length() < tmplen; ++i) {
3270                 if (isDeleted(i))
3271                         continue;
3272                 char_type const c = d->text_[i];
3273                 if (isPrintable(c))
3274                         os += c;
3275                 else if (c == META_INSET)
3276                         getInset(i)->forOutliner(os, tmplen, false);
3277         }
3278         if (shorten)
3279                 Text::shortenForOutliner(os, maxlen);
3280 }
3281
3282
3283 void Paragraph::setInsetOwner(Inset const * inset)
3284 {
3285         d->inset_owner_ = inset;
3286 }
3287
3288
3289 int Paragraph::id() const
3290 {
3291         return d->id_;
3292 }
3293
3294
3295 void Paragraph::setId(int id)
3296 {
3297         d->id_ = id;
3298 }
3299
3300
3301 Layout const & Paragraph::layout() const
3302 {
3303         return *d->layout_;
3304 }
3305
3306
3307 void Paragraph::setLayout(Layout const & layout)
3308 {
3309         d->layout_ = &layout;
3310 }
3311
3312
3313 void Paragraph::setDefaultLayout(DocumentClass const & tc)
3314 {
3315         setLayout(tc.defaultLayout());
3316 }
3317
3318
3319 void Paragraph::setPlainLayout(DocumentClass const & tc)
3320 {
3321         setLayout(tc.plainLayout());
3322 }
3323
3324
3325 void Paragraph::setPlainOrDefaultLayout(DocumentClass const & tclass)
3326 {
3327         if (usePlainLayout())
3328                 setPlainLayout(tclass);
3329         else
3330                 setDefaultLayout(tclass);
3331 }
3332
3333
3334 Inset const & Paragraph::inInset() const
3335 {
3336         LBUFERR(d->inset_owner_);
3337         return *d->inset_owner_;
3338 }
3339
3340
3341 ParagraphParameters & Paragraph::params()
3342 {
3343         return d->params_;
3344 }
3345
3346
3347 ParagraphParameters const & Paragraph::params() const
3348 {
3349         return d->params_;
3350 }
3351
3352
3353 bool Paragraph::isFreeSpacing() const
3354 {
3355         if (d->layout_->free_spacing)
3356                 return true;
3357         return d->inset_owner_ && d->inset_owner_->isFreeSpacing();
3358 }
3359
3360
3361 bool Paragraph::allowEmpty() const
3362 {
3363         if (d->layout_->keepempty)
3364                 return true;
3365         return d->inset_owner_ && d->inset_owner_->allowEmpty();
3366 }
3367
3368
3369 bool Paragraph::brokenBiblio() const
3370 {
3371         // there is a problem if there is no bibitem at position 0 or
3372         // if there is another bibitem in the paragraph.
3373         return d->layout_->labeltype == LABEL_BIBLIO
3374                 && (d->insetlist_.find(BIBITEM_CODE) != 0
3375                     || d->insetlist_.find(BIBITEM_CODE, 1) > 0);
3376 }
3377
3378
3379 int Paragraph::fixBiblio(Buffer const & buffer)
3380 {
3381         // FIXME: What about the case where paragraph is not BIBLIO
3382         // but there is an InsetBibitem?
3383         // FIXME: when there was already an inset at 0, the return value is 1,
3384         // which does not tell whether another inset has been remove; the
3385         // cursor cannot be correctly updated.
3386
3387         if (d->layout_->labeltype != LABEL_BIBLIO)
3388                 return 0;
3389
3390         bool const track_changes = buffer.params().track_changes;
3391         int bibitem_pos = d->insetlist_.find(BIBITEM_CODE);
3392         bool const hasbibitem0 = bibitem_pos == 0;
3393
3394         if (hasbibitem0) {
3395                 bibitem_pos = d->insetlist_.find(BIBITEM_CODE, 1);
3396                 // There was an InsetBibitem at pos 0, and no other one => OK
3397                 if (bibitem_pos == -1)
3398                         return 0;
3399                 // there is a bibitem at the 0 position, but since
3400                 // there is a second one, we copy the second on the
3401                 // first. We're assuming there are at most two of
3402                 // these, which there should be.
3403                 // FIXME: why does it make sense to do that rather
3404                 // than keep the first? (JMarc)
3405                 Inset * inset = releaseInset(bibitem_pos);
3406                 d->insetlist_.begin()->inset = inset;
3407                 return -bibitem_pos;
3408         }
3409
3410         // We need to create an inset at the beginning
3411         Inset * inset = 0;
3412         if (bibitem_pos > 0) {
3413                 // there was one somewhere in the paragraph, let's move it
3414                 inset = d->insetlist_.release(bibitem_pos);
3415                 eraseChar(bibitem_pos, track_changes);
3416         } else
3417                 // make a fresh one
3418                 inset = new InsetBibitem(const_cast<Buffer *>(&buffer),
3419                                          InsetCommandParams(BIBITEM_CODE));
3420
3421         Font font(inherit_font, buffer.params().language);
3422         insertInset(0, inset, font, Change(track_changes ? Change::INSERTED
3423                                                    : Change::UNCHANGED));
3424
3425         return 1;
3426 }
3427
3428
3429 void Paragraph::checkAuthors(AuthorList const & authorList)
3430 {
3431         d->changes_.checkAuthors(authorList);
3432 }
3433
3434
3435 bool Paragraph::isChanged(pos_type pos) const
3436 {
3437         return lookupChange(pos).changed();
3438 }
3439
3440
3441 bool Paragraph::isInserted(pos_type pos) const
3442 {
3443         return lookupChange(pos).inserted();
3444 }
3445
3446
3447 bool Paragraph::isDeleted(pos_type pos) const
3448 {
3449         return lookupChange(pos).deleted();
3450 }
3451
3452
3453 InsetList const & Paragraph::insetList() const
3454 {
3455         return d->insetlist_;
3456 }
3457
3458
3459 void Paragraph::setBuffer(Buffer & b)
3460 {
3461         d->insetlist_.setBuffer(b);
3462 }
3463
3464
3465 Inset * Paragraph::releaseInset(pos_type pos)
3466 {
3467         Inset * inset = d->insetlist_.release(pos);
3468         /// does not honour change tracking!
3469         eraseChar(pos, false);
3470         return inset;
3471 }
3472
3473
3474 Inset * Paragraph::getInset(pos_type pos)
3475 {
3476         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3477                  ? d->insetlist_.get(pos) : 0;
3478 }
3479
3480
3481 Inset const * Paragraph::getInset(pos_type pos) const
3482 {
3483         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3484                  ? d->insetlist_.get(pos) : 0;
3485 }
3486
3487
3488 void Paragraph::changeCase(BufferParams const & bparams, pos_type pos,
3489                 pos_type & right, TextCase action)
3490 {
3491         // process sequences of modified characters; in change
3492         // tracking mode, this approach results in much better
3493         // usability than changing case on a char-by-char basis
3494         // We also need to track the current font, since font
3495         // changes within sequences can occur.
3496         vector<pair<char_type, Font> > changes;
3497
3498         bool const trackChanges = bparams.track_changes;
3499
3500         bool capitalize = true;
3501
3502         for (; pos < right; ++pos) {
3503                 char_type oldChar = d->text_[pos];
3504                 char_type newChar = oldChar;
3505
3506                 // ignore insets and don't play with deleted text!
3507                 if (oldChar != META_INSET && !isDeleted(pos)) {
3508                         switch (action) {
3509                                 case text_lowercase:
3510                                         newChar = lowercase(oldChar);
3511                                         break;
3512                                 case text_capitalization:
3513                                         if (capitalize) {
3514                                                 newChar = uppercase(oldChar);
3515                                                 capitalize = false;
3516                                         }
3517                                         break;
3518                                 case text_uppercase:
3519                                         newChar = uppercase(oldChar);
3520                                         break;
3521                         }
3522                 }
3523
3524                 if (isWordSeparator(pos) || isDeleted(pos)) {
3525                         // permit capitalization again
3526                         capitalize = true;
3527                 }
3528
3529                 if (oldChar != newChar) {
3530                         changes.push_back(make_pair(newChar, getFontSettings(bparams, pos)));
3531                         if (pos != right - 1)
3532                                 continue;
3533                         // step behind the changing area
3534                         pos++;
3535                 }
3536
3537                 int erasePos = pos - changes.size();
3538                 for (size_t i = 0; i < changes.size(); i++) {
3539                         insertChar(pos, changes[i].first,
3540                                    changes[i].second,
3541                                    trackChanges);
3542                         if (!eraseChar(erasePos, trackChanges)) {
3543                                 ++erasePos;
3544                                 ++pos; // advance
3545                                 ++right; // expand selection
3546                         }
3547                 }
3548                 changes.clear();
3549         }
3550 }
3551
3552
3553 int Paragraph::find(docstring const & str, bool cs, bool mw,
3554                 pos_type start_pos, bool del) const
3555 {
3556         pos_type pos = start_pos;
3557         int const strsize = str.length();
3558         int i = 0;
3559         pos_type const parsize = d->text_.size();
3560         for (i = 0; i < strsize && pos < parsize; ++i, ++pos) {
3561                 // Ignore "invisible" letters such as ligature breaks
3562                 // and hyphenation chars while searching
3563                 while (pos < parsize - 1 && isInset(pos)) {
3564                         odocstringstream os;
3565                         getInset(pos)->toString(os);
3566                         if (!getInset(pos)->isLetter() || !os.str().empty())
3567                                 break;
3568                         pos++;
3569                 }
3570                 if (cs && str[i] != d->text_[pos])
3571                         break;
3572                 if (!cs && uppercase(str[i]) != uppercase(d->text_[pos]))
3573                         break;
3574                 if (!del && isDeleted(pos))
3575                         break;
3576         }
3577
3578         if (i != strsize)
3579                 return 0;
3580
3581         // if necessary, check whether string matches word
3582         if (mw) {
3583                 if (start_pos > 0 && !isWordSeparator(start_pos - 1))
3584                         return 0;
3585                 if (pos < parsize
3586                         && !isWordSeparator(pos))
3587                         return 0;
3588         }
3589
3590         return pos - start_pos;
3591 }
3592
3593
3594 char_type Paragraph::getChar(pos_type pos) const
3595 {
3596         return d->text_[pos];
3597 }
3598
3599
3600 pos_type Paragraph::size() const
3601 {
3602         return d->text_.size();
3603 }
3604
3605
3606 bool Paragraph::empty() const
3607 {
3608         return d->text_.empty();
3609 }
3610
3611
3612 bool Paragraph::isInset(pos_type pos) const
3613 {
3614         return d->text_[pos] == META_INSET;
3615 }
3616
3617
3618 bool Paragraph::isSeparator(pos_type pos) const
3619 {
3620         //FIXME: Are we sure this can be the only separator?
3621         return d->text_[pos] == ' ';
3622 }
3623
3624
3625 void Paragraph::deregisterWords()
3626 {
3627         Private::LangWordsMap::const_iterator itl = d->words_.begin();
3628         Private::LangWordsMap::const_iterator ite = d->words_.end();
3629         for (; itl != ite; ++itl) {
3630                 WordList * wl = theWordList(itl->first);
3631                 Private::Words::const_iterator it = (itl->second).begin();
3632                 Private::Words::const_iterator et = (itl->second).end();
3633                 for (; it != et; ++it)
3634                         wl->remove(*it);
3635         }
3636         d->words_.clear();
3637 }
3638
3639
3640 void Paragraph::locateWord(pos_type & from, pos_type & to,
3641         word_location const loc) const
3642 {
3643         switch (loc) {
3644         case WHOLE_WORD_STRICT:
3645                 if (from == 0 || from == size()
3646                     || isWordSeparator(from)
3647                     || isWordSeparator(from - 1)) {
3648                         to = from;
3649                         return;
3650                 }
3651                 // no break here, we go to the next
3652
3653         case WHOLE_WORD:
3654                 // If we are already at the beginning of a word, do nothing
3655                 if (!from || isWordSeparator(from - 1))
3656                         break;
3657                 // no break here, we go to the next
3658
3659         case PREVIOUS_WORD:
3660                 // always move the cursor to the beginning of previous word
3661                 while (from && !isWordSeparator(from - 1))
3662                         --from;
3663                 break;
3664         case NEXT_WORD:
3665                 LYXERR0("Paragraph::locateWord: NEXT_WORD not implemented yet");
3666                 break;
3667         case PARTIAL_WORD:
3668                 // no need to move the 'from' cursor
3669                 break;
3670         }
3671         to = from;
3672         while (to < size() && !isWordSeparator(to))
3673                 ++to;
3674 }
3675
3676
3677 void Paragraph::collectWords()
3678 {
3679         for (pos_type pos = 0; pos < size(); ++pos) {
3680                 if (isWordSeparator(pos))
3681                         continue;
3682                 pos_type from = pos;
3683                 locateWord(from, pos, WHOLE_WORD);
3684                 // Work around MSVC warning: The statement
3685                 // if (pos < from + lyxrc.completion_minlength)
3686                 // triggers a signed vs. unsigned warning.
3687                 // I don't know why this happens, it could be a MSVC bug, or
3688                 // related to LLP64 (windows) vs. LP64 (unix) programming
3689                 // model, or the C++ standard might be ambigous in the section
3690                 // defining the "usual arithmetic conversions". However, using
3691                 // a temporary variable is safe and works on all compilers.
3692                 pos_type const endpos = from + lyxrc.completion_minlength;
3693                 if (pos < endpos)
3694                         continue;
3695                 FontList::const_iterator cit = d->fontlist_.fontIterator(from);
3696                 if (cit == d->fontlist_.end())
3697                         return;
3698                 Language const * lang = cit->font().language();
3699                 docstring const word = asString(from, pos, AS_STR_NONE);
3700                 d->words_[lang->lang()].insert(word);
3701         }
3702 }
3703
3704
3705 void Paragraph::registerWords()
3706 {
3707         Private::LangWordsMap::const_iterator itl = d->words_.begin();
3708         Private::LangWordsMap::const_iterator ite = d->words_.end();
3709         for (; itl != ite; ++itl) {
3710                 WordList * wl = theWordList(itl->first);
3711                 Private::Words::const_iterator it = (itl->second).begin();
3712                 Private::Words::const_iterator et = (itl->second).end();
3713                 for (; it != et; ++it)
3714                         wl->insert(*it);
3715         }
3716 }
3717
3718
3719 void Paragraph::updateWords()
3720 {
3721         deregisterWords();
3722         collectWords();
3723         registerWords();
3724 }
3725
3726
3727 void Paragraph::Private::appendSkipPosition(SkipPositions & skips, pos_type const pos) const
3728 {
3729         SkipPositionsIterator begin = skips.begin();
3730         SkipPositions::iterator end = skips.end();
3731         if (pos > 0 && begin < end) {
3732                 --end;
3733                 if (end->last == pos - 1) {
3734                         end->last = pos;
3735                         return;
3736                 }
3737         }
3738         skips.insert(end, FontSpan(pos, pos));
3739 }
3740
3741
3742 Language * Paragraph::Private::locateSpellRange(
3743         pos_type & from, pos_type & to,
3744         SkipPositions & skips) const
3745 {
3746         // skip leading white space
3747         while (from < to && owner_->isWordSeparator(from))
3748                 ++from;
3749         // don't check empty range
3750         if (from >= to)
3751                 return 0;
3752         // get current language
3753         Language * lang = getSpellLanguage(from);
3754         pos_type last = from;
3755         bool samelang = true;
3756         bool sameinset = true;
3757         while (last < to && samelang && sameinset) {
3758                 // hop to end of word
3759                 while (last < to && !owner_->isWordSeparator(last)) {
3760                         if (owner_->getInset(last)) {
3761                                 appendSkipPosition(skips, last);
3762                         } else if (owner_->isDeleted(last)) {
3763                                 appendSkipPosition(skips, last);
3764                         }
3765                         ++last;
3766                 }
3767                 // hop to next word while checking for insets
3768                 while (sameinset && last < to && owner_->isWordSeparator(last)) {
3769                         if (Inset const * inset = owner_->getInset(last))
3770                                 sameinset = inset->isChar() && inset->isLetter();
3771                         if (sameinset && owner_->isDeleted(last)) {
3772                                 appendSkipPosition(skips, last);
3773                         }
3774                         if (sameinset)
3775                                 last++;
3776                 }
3777                 if (sameinset && last < to) {
3778                         // now check for language change
3779                         samelang = lang == getSpellLanguage(last);
3780                 }
3781         }
3782         // if language change detected backstep is needed
3783         if (!samelang)
3784                 --last;
3785         to = last;
3786         return lang;
3787 }
3788
3789
3790 Language * Paragraph::Private::getSpellLanguage(pos_type const from) const
3791 {
3792         Language * lang =
3793                 const_cast<Language *>(owner_->getFontSettings(
3794                         inset_owner_->buffer().params(), from).language());
3795         if (lang == inset_owner_->buffer().params().language
3796                 && !lyxrc.spellchecker_alt_lang.empty()) {
3797                 string lang_code;
3798                 string const lang_variety =
3799                         split(lyxrc.spellchecker_alt_lang, lang_code, '-');
3800                 lang->setCode(lang_code);
3801                 lang->setVariety(lang_variety);
3802         }
3803         return lang;
3804 }
3805
3806
3807 void Paragraph::requestSpellCheck(pos_type pos)
3808 {
3809         d->requestSpellCheck(pos);
3810 }
3811
3812
3813 bool Paragraph::needsSpellCheck() const
3814 {
3815         SpellChecker::ChangeNumber speller_change_number = 0;
3816         if (theSpellChecker())
3817                 speller_change_number = theSpellChecker()->changeNumber();
3818         if (speller_change_number > d->speller_state_.currentChangeNumber()) {
3819                 d->speller_state_.needsCompleteRefresh(speller_change_number);
3820         }
3821         return d->needsSpellCheck();
3822 }
3823
3824
3825 bool Paragraph::Private::ignoreWord(docstring const & word) const
3826 {
3827         // Ignore words with digits
3828         // FIXME: make this customizable
3829         // (note that some checkers ignore words with digits by default)
3830         docstring::const_iterator cit = word.begin();
3831         docstring::const_iterator const end = word.end();
3832         for (; cit != end; ++cit) {
3833                 if (isNumber((*cit)))
3834                         return true;
3835         }
3836         return false;
3837 }
3838
3839
3840 SpellChecker::Result Paragraph::spellCheck(pos_type & from, pos_type & to,
3841         WordLangTuple & wl, docstring_list & suggestions,
3842         bool do_suggestion, bool check_learned) const
3843 {
3844         SpellChecker::Result result = SpellChecker::WORD_OK;
3845         SpellChecker * speller = theSpellChecker();
3846         if (!speller)
3847                 return result;
3848
3849         if (!d->layout_->spellcheck || !inInset().allowSpellCheck())
3850                 return result;
3851
3852         locateWord(from, to, WHOLE_WORD);
3853         if (from == to || from >= size())
3854                 return result;
3855
3856         docstring word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
3857         Language * lang = d->getSpellLanguage(from);
3858
3859         wl = WordLangTuple(word, lang);
3860
3861         if (word.empty())
3862                 return result;
3863
3864         if (needsSpellCheck() || check_learned) {
3865                 pos_type end = to;
3866                 if (!d->ignoreWord(word)) {
3867                         bool const trailing_dot = to < size() && d->text_[to] == '.';
3868                         result = speller->check(wl);
3869                         if (SpellChecker::misspelled(result) && trailing_dot) {
3870                                 wl = WordLangTuple(word.append(from_ascii(".")), lang);
3871                                 result = speller->check(wl);
3872                                 if (!SpellChecker::misspelled(result)) {
3873                                         LYXERR(Debug::GUI, "misspelled word is correct with dot: \"" <<
3874                                            word << "\" [" <<
3875                                            from << ".." << to << "]");
3876                                 } else {
3877                                         // spell check with dot appended failed too
3878                                         // restore original word/lang value
3879                                         word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
3880                                         wl = WordLangTuple(word, lang);
3881                                 }
3882                         }
3883                 }
3884                 if (!SpellChecker::misspelled(result)) {
3885                         // area up to the begin of the next word is not misspelled
3886                         while (end < size() && isWordSeparator(end))
3887                                 ++end;
3888                 }
3889                 d->setMisspelled(from, end, result);
3890         } else {
3891                 result = d->speller_state_.getState(from);
3892         }
3893
3894         if (do_suggestion)
3895                 suggestions.clear();
3896
3897         if (SpellChecker::misspelled(result)) {
3898                 LYXERR(Debug::GUI, "misspelled word: \"" <<
3899                            word << "\" [" <<
3900                            from << ".." << to << "]");
3901                 if (do_suggestion)
3902                         speller->suggest(wl, suggestions);
3903         }
3904         return result;
3905 }
3906
3907
3908 void Paragraph::Private::markMisspelledWords(
3909         pos_type const & first, pos_type const & last,
3910         SpellChecker::Result result,
3911         docstring const & word,
3912         SkipPositions const & skips)
3913 {
3914         if (!SpellChecker::misspelled(result)) {
3915                 setMisspelled(first, last, SpellChecker::WORD_OK);
3916                 return;
3917         }
3918         int snext = first;
3919         SpellChecker * speller = theSpellChecker();
3920         // locate and enumerate the error positions
3921         int nerrors = speller->numMisspelledWords();
3922         int numskipped = 0;
3923         SkipPositionsIterator it = skips.begin();
3924         SkipPositionsIterator et = skips.end();
3925         for (int index = 0; index < nerrors; ++index) {
3926                 int wstart;
3927                 int wlen = 0;
3928                 speller->misspelledWord(index, wstart, wlen);
3929                 /// should not happen if speller supports range checks
3930                 if (!wlen) continue;
3931                 docstring const misspelled = word.substr(wstart, wlen);
3932                 wstart += first + numskipped;
3933                 if (snext < wstart) {
3934                         /// mark the range of correct spelling
3935                         numskipped += countSkips(it, et, wstart);
3936                         setMisspelled(snext,
3937                                 wstart - 1, SpellChecker::WORD_OK);
3938                 }
3939                 snext = wstart + wlen;
3940                 numskipped += countSkips(it, et, snext);
3941                 /// mark the range of misspelling
3942                 setMisspelled(wstart, snext, result);
3943                 LYXERR(Debug::GUI, "misspelled word: \"" <<
3944                            misspelled << "\" [" <<
3945                            wstart << ".." << (snext-1) << "]");
3946                 ++snext;
3947         }
3948         if (snext <= last) {
3949                 /// mark the range of correct spelling at end
3950                 setMisspelled(snext, last, SpellChecker::WORD_OK);
3951         }
3952 }
3953
3954
3955 void Paragraph::spellCheck() const
3956 {
3957         SpellChecker * speller = theSpellChecker();
3958         if (!speller || empty() ||!needsSpellCheck())
3959                 return;
3960         pos_type start;
3961         pos_type endpos;
3962         d->rangeOfSpellCheck(start, endpos);
3963         if (speller->canCheckParagraph()) {
3964                 // loop until we leave the range
3965                 for (pos_type first = start; first < endpos; ) {
3966                         pos_type last = endpos;
3967                         Private::SkipPositions skips;
3968                         Language * lang = d->locateSpellRange(first, last, skips);
3969                         if (first >= endpos)
3970                                 break;
3971                         // start the spell checker on the unit of meaning
3972                         docstring word = asString(first, last, AS_STR_INSETS + AS_STR_SKIPDELETE);
3973                         WordLangTuple wl = WordLangTuple(word, lang);
3974                         SpellChecker::Result result = word.size() ?
3975                                 speller->check(wl) : SpellChecker::WORD_OK;
3976                         d->markMisspelledWords(first, last, result, word, skips);
3977                         first = ++last;
3978                 }
3979         } else {
3980                 static docstring_list suggestions;
3981                 pos_type to = endpos;
3982                 while (start < endpos) {
3983                         WordLangTuple wl;
3984                         spellCheck(start, to, wl, suggestions, false);
3985                         start = to + 1;
3986                 }
3987         }
3988         d->readySpellCheck();
3989 }
3990
3991
3992 bool Paragraph::isMisspelled(pos_type pos, bool check_boundary) const
3993 {
3994         bool result = SpellChecker::misspelled(d->speller_state_.getState(pos));
3995         if (result || pos <= 0 || pos > size())
3996                 return result;
3997         if (check_boundary && (pos == size() || isWordSeparator(pos)))
3998                 result = SpellChecker::misspelled(d->speller_state_.getState(pos - 1));
3999         return result;
4000 }
4001
4002
4003 string Paragraph::magicLabel() const
4004 {
4005         stringstream ss;
4006         ss << "magicparlabel-" << id();
4007         return ss.str();
4008 }
4009
4010
4011 } // namespace lyx