src/lyxfind.cpp

   1 /**
   2  * \file lyxfind.cpp
   3  * This file is part of LyX, the document processor.
   4  * License details can be found in the file COPYING.
   5  *
   6  * \author Lars Gullik Bjønnes
   7  * \author John Levon
   8  * \author Jürgen Vigna
   9  * \author Alfredo Braunstein
  10  * \author Tommaso Cucinotta
  11  * \author Kornel Benko
  12  *
  13  * Full author contact details are available in file CREDITS.
  14  */
  15
  16 #include <config.h>
  17 #include <iterator>
  18
  19 #include "lyxfind.h"
  20
  21 #include "Buffer.h"
  22 #include "BufferList.h"
  23 #include "BufferParams.h"
  24 #include "BufferView.h"
  25 #include "Changes.h"
  26 #include "Cursor.h"
  27 #include "CutAndPaste.h"
  28 #include "FuncRequest.h"
  29 #include "LyX.h"
  30 #include "output_latex.h"
  31 #include "OutputParams.h"
  32 #include "Paragraph.h"
  33 #include "Text.h"
  34 #include "Encoding.h"
  35 #include "Language.h"
  36
  37 #include "frontends/Application.h"
  38 #include "frontends/alert.h"
  39
  40 #include "mathed/InsetMath.h"
  41 #include "mathed/InsetMathHull.h"
  42 #include "mathed/MathData.h"
  43 #include "mathed/MathStream.h"
  44 #include "mathed/MathSupport.h"
  45
  46 #include "support/debug.h"
  47 #include "support/docstream.h"
  48 #include "support/FileName.h"
  49 #include "support/gettext.h"
  50 #include "support/lassert.h"
  51 #include "support/lstrings.h"
  52 #include "support/textutils.h"
  53
  54 #include <unordered_map>
  55 #include <regex>
  56
  57 //#define ResultsDebug
  58 #define USE_QT_FOR_SEARCH
  59 #if defined(USE_QT_FOR_SEARCH)
  60         #include <QtCore>       // sets QT_VERSION
  61         #if (QT_VERSION >= 0x050000)
  62                 #include <QRegularExpression>
  63                 #define QTSEARCH 1
  64         #else
  65                 #define QTSEARCH 0
  66         #endif
  67 #else
  68         #define QTSEARCH 0
  69 #endif
  70
  71 using namespace std;
  72 using namespace lyx::support;
  73
  74 namespace lyx {
  75
  76 typedef unordered_map<string, string> AccentsMap;
  77 typedef unordered_map<string,string>::const_iterator AccentsIterator;
  78 static AccentsMap accents = unordered_map<string, string>();
  79
  80 // Helper class for deciding what should be ignored
  81 class IgnoreFormats {
  82  public:
  83         ///
  84         IgnoreFormats() = default;
  85         ///
  86         bool getFamily() const { return ignoreFamily_; }
  87         ///
  88         bool getSeries() const { return ignoreSeries_; }
  89         ///
  90         bool getShape() const { return ignoreShape_; }
  91         ///
  92         bool getSize() const { return ignoreSize_; }
  93         ///
  94         bool getUnderline() const { return ignoreUnderline_; }
  95         ///
  96         bool getMarkUp() const { return ignoreMarkUp_; }
  97         ///
  98         bool getStrikeOut() const { return ignoreStrikeOut_; }
  99         ///
 100         bool getSectioning() const { return ignoreSectioning_; }
 101         ///
 102         bool getFrontMatter() const { return ignoreFrontMatter_; }
 103         ///
 104         bool getColor() const { return ignoreColor_; }
 105         ///
 106         bool getLanguage() const { return ignoreLanguage_; }
 107         ///
 108         bool getDeleted() const { return ignoreDeleted_; }
 109         ///
 110         void setIgnoreDeleted(bool value);
 111         ///
 112         bool getNonContent() const { return searchNonContent_; }
 113         ///
 114         void setIgnoreFormat(string const & type, bool value, bool fromUser = true);
 115
 116 private:
 117         ///
 118         bool ignoreFamily_ = false;
 119         ///
 120         bool ignoreSeries_ = false;
 121         ///
 122         bool ignoreShape_ = false;
 123         ///
 124         bool ignoreSize_ = true;
 125         ///
 126         bool ignoreUnderline_ = false;
 127         ///
 128         bool ignoreMarkUp_ = false;
 129         ///
 130         bool ignoreStrikeOut_ = false;
 131         ///
 132         bool ignoreSectioning_ = false;
 133         ///
 134         bool ignoreFrontMatter_ = false;
 135         ///
 136         bool ignoreColor_ = false;
 137         ///
 138         bool ignoreLanguage_ = false;
 139         bool userSelectedIgnoreLanguage_ = false;
 140         ///
 141         bool ignoreDeleted_ = true;
 142         ///
 143         bool searchNonContent_ = true;
 144 };
 145
 146 void IgnoreFormats::setIgnoreFormat(string const & type, bool value, bool fromUser)
 147 {
 148         if (type == "color") {
 149                 ignoreColor_ = value;
 150         }
 151         else if (type == "language") {
 152                 if (fromUser) {
 153                         userSelectedIgnoreLanguage_ = value;
 154                         ignoreLanguage_ = value;
 155                 }
 156                 else
 157                         ignoreLanguage_ = (value || userSelectedIgnoreLanguage_);
 158         }
 159         else if (type == "sectioning") {
 160                 ignoreSectioning_ = value;
 161                 ignoreFrontMatter_ = value;
 162         }
 163         else if (type == "font") {
 164                 ignoreSeries_ = value;
 165                 ignoreShape_ = value;
 166                 ignoreFamily_ = value;
 167         }
 168         else if (type == "series") {
 169                 ignoreSeries_ = value;
 170         }
 171         else if (type == "shape") {
 172                 ignoreShape_ = value;
 173         }
 174         else if (type == "size") {
 175                 ignoreSize_ = value;
 176         }
 177         else if (type == "family") {
 178                 ignoreFamily_ = value;
 179         }
 180         else if (type == "markup") {
 181                 ignoreMarkUp_ = value;
 182         }
 183         else if (type == "underline") {
 184                 ignoreUnderline_ = value;
 185         }
 186         else if (type == "strike") {
 187                 ignoreStrikeOut_ = value;
 188         }
 189         else if (type == "deleted") {
 190                 ignoreDeleted_ = value;
 191         }
 192         else if (type == "non-output-content") {
 193                 searchNonContent_ = !value;
 194         }
 195 }
 196
 197 // The global variable that can be changed from outside
 198 IgnoreFormats ignoreFormats;
 199
 200
 201 void setIgnoreFormat(string const & type, bool value, bool fromUser)
 202 {
 203         ignoreFormats.setIgnoreFormat(type, value, fromUser);
 204 }
 205
 206
 207 namespace {
 208
 209 bool parse_bool(docstring & howto, bool const defvalue = false)
 210 {
 211         if (howto.empty())
 212                 return defvalue;
 213         docstring var;
 214         howto = split(howto, var, ' ');
 215         return var == "1";
 216 }
 217
 218
 219 class MatchString
 220 {
 221 public:
 222         MatchString(docstring const & s, bool cs, bool mw)
 223                 : str(s), case_sens(cs), whole_words(mw)
 224         {}
 225
 226         // returns true if the specified string is at the specified position
 227         // del specifies whether deleted strings in ct mode will be considered
 228         int operator()(Paragraph const & par, pos_type pos, bool del = true) const
 229         {
 230                 return par.find(str, case_sens, whole_words, pos, del);
 231         }
 232
 233 private:
 234         // search string
 235         docstring str;
 236         // case sensitive
 237         bool case_sens;
 238         // match whole words only
 239         bool whole_words;
 240 };
 241
 242
 243 int findForward(DocIterator & cur, DocIterator const endcur,
 244                 MatchString const & match,
 245                 bool find_del = true, bool onlysel = false)
 246 {
 247         for (; cur; cur.forwardChar()) {
 248                 if (onlysel && endcur.pit() == cur.pit()
 249                     && endcur.idx() == cur.idx() && endcur.pos() < cur.pos())
 250                         break;
 251                 if (cur.inTexted()) {
 252                         int len = match(cur.paragraph(), cur.pos(), find_del);
 253                         if (len > 0)
 254                                 return len;
 255                 }
 256         }
 257         return 0;
 258 }
 259
 260
 261 int findBackwards(DocIterator & cur, DocIterator const endcur,
 262                   MatchString const & match,
 263                   bool find_del = true, bool onlysel = false)
 264 {
 265         while (cur) {
 266                 cur.backwardChar();
 267                 if (onlysel && endcur.pit() == cur.pit()
 268                     && endcur.idx() == cur.idx() && endcur.pos() > cur.pos())
 269                         break;
 270                 if (cur.inTexted()) {
 271                         int len = match(cur.paragraph(), cur.pos(), find_del);
 272                         if (len > 0)
 273                                 return len;
 274                 }
 275         }
 276         return 0;
 277 }
 278
 279
 280 bool searchAllowed(docstring const & str)
 281 {
 282         if (str.empty()) {
 283                 frontend::Alert::error(_("Search error"), _("Search string is empty"));
 284                 return false;
 285         }
 286         return true;
 287 }
 288
 289 } // namespace
 290
 291
 292 bool findOne(BufferView * bv, docstring const & searchstr,
 293              bool case_sens, bool whole, bool forward,
 294              bool find_del, bool check_wrap, bool const auto_wrap,
 295              bool instant, bool onlysel)
 296 {
 297         // Clean up previous selections with empty searchstr on instant
 298         if (searchstr.empty() && instant) {
 299                 if (bv->cursor().selection()) {
 300                         bv->setCursor(bv->cursor().selectionBegin());
 301                         bv->clearSelection();
 302                 }
 303                 return true;
 304         }
 305
 306         if (!searchAllowed(searchstr))
 307                 return false;
 308
 309         DocIterator const endcur = forward ? bv->cursor().selectionEnd() : bv->cursor().selectionBegin();
 310
 311         if (onlysel && bv->cursor().selection()) {
 312                 docstring const matchstring = bv->cursor().selectionAsString(false);
 313                 docstring const lcmatchsting = support::lowercase(matchstring);
 314                 if (matchstring == searchstr || (!case_sens && lcmatchsting == lowercase(searchstr))) {
 315                         docstring q = _("The search string matches the selection, and search is limited to selection.\n"
 316                                         "Continue search outside?");
 317                         int search_answer = frontend::Alert::prompt(_("Search outside selection?"),
 318                                 q, 0, 1, _("&Yes"), _("&No"));
 319                         if (search_answer == 0) {
 320                                 bv->clearSelection();
 321                                 if (findOne(bv, searchstr, case_sens, whole, forward,
 322                                             find_del, check_wrap, auto_wrap, false, false))
 323                                         return true;
 324                         }
 325                         return false;
 326                 }
 327         }
 328
 329         DocIterator cur = forward
 330                 ? ((instant || onlysel) ? bv->cursor().selectionBegin() : bv->cursor().selectionEnd())
 331                 : ((instant || onlysel) ? bv->cursor().selectionEnd() : bv->cursor().selectionBegin());
 332
 333         MatchString const match(searchstr, case_sens, whole);
 334
 335         int match_len = forward
 336                 ? findForward(cur, endcur, match, find_del, onlysel)
 337                 : findBackwards(cur, endcur, match, find_del, onlysel);
 338
 339         if (match_len > 0)
 340                 bv->putSelectionAt(cur, match_len, !forward);
 341         else if (onlysel) {
 342                 docstring q = _("The search string was not found within the selection.\n"
 343                                 "Continue search outside?");
 344                 int search_answer = frontend::Alert::prompt(_("Search outside selection?"),
 345                         q, 0, 1, _("&Yes"), _("&No"));
 346                 if (search_answer == 0) {
 347                         bv->clearSelection();
 348                         if (findOne(bv, searchstr, case_sens, whole, forward,
 349                                     find_del, check_wrap, auto_wrap, false, false))
 350                                 return true;
 351                 }
 352                 return false;
 353         }
 354         else if (check_wrap) {
 355                 DocIterator cur_orig(bv->cursor());
 356                 bool wrap = auto_wrap;
 357                 if (!auto_wrap) {
 358                         docstring q;
 359                         if (forward)
 360                                 q = _("End of file reached while searching forward.\n"
 361                                   "Continue searching from the beginning?");
 362                         else
 363                                 q = _("Beginning of file reached while searching backward.\n"
 364                                   "Continue searching from the end?");
 365                         int wrap_answer = frontend::Alert::prompt(_("Wrap search?"),
 366                                 q, 0, 1, _("&Yes"), _("&No"));
 367                         wrap = wrap_answer == 0;
 368                 }
 369                 if (wrap) {
 370                         if (forward) {
 371                                 bv->cursor().clear();
 372                                 bv->cursor().push_back(CursorSlice(bv->buffer().inset()));
 373                         } else {
 374                                 bv->cursor().setCursor(doc_iterator_end(&bv->buffer()));
 375                                 bv->cursor().backwardPos();
 376                         }
 377                         if (auto_wrap) {
 378                                 docstring const msg = forward
 379                                   ? _("Search reached end of document, continuing from beginning.")
 380                                   : _("Search reached beginning of document, continuing from end.");
 381                                 bv->message(msg);
 382                         }
 383                         bv->clearSelection();
 384                         if (findOne(bv, searchstr, case_sens, whole, forward,
 385                                     find_del, false, false, false, false))
 386                                 return true;
 387                 }
 388                 bv->cursor().setCursor(cur_orig);
 389                 return false;
 390         }
 391
 392         return match_len > 0;
 393 }
 394
 395
 396 namespace {
 397
 398 int replaceAll(BufferView * bv,
 399                docstring const & searchstr, docstring const & replacestr,
 400                bool case_sens, bool whole, bool onlysel)
 401 {
 402         Buffer & buf = bv->buffer();
 403
 404         if (!searchAllowed(searchstr) || buf.isReadonly())
 405                 return 0;
 406
 407         DocIterator startcur = bv->cursor().selectionBegin();
 408         DocIterator endcur = bv->cursor().selectionEnd();
 409         bool const had_selection = bv->cursor().selection();
 410
 411         MatchString const match(searchstr, case_sens, whole);
 412         int num = 0;
 413
 414         int const rsize = replacestr.size();
 415         int const ssize = searchstr.size();
 416
 417         Cursor cur(*bv);
 418         cur.setCursor(doc_iterator_begin(&buf));
 419         int match_len = findForward(cur, endcur, match, false, onlysel);
 420         while (match_len > 0) {
 421                 // Backup current cursor position and font.
 422                 pos_type const pos = cur.pos();
 423                 Font const font = cur.paragraph().getFontSettings(buf.params(), pos);
 424                 cur.recordUndo();
 425                 int ct_deleted_text = ssize -
 426                         cur.paragraph().eraseChars(pos, pos + match_len,
 427                                                    buf.params().track_changes);
 428                 cur.paragraph().insert(pos, replacestr, font,
 429                                        Change(buf.params().track_changes
 430                                               ? Change::INSERTED
 431                                               : Change::UNCHANGED));
 432                 for (int i = 0; i < rsize + ct_deleted_text
 433                      && cur.pos() < cur.lastpos(); ++i)
 434                         cur.forwardPos();
 435                 if (onlysel && cur.pit() == endcur.pit() && cur.idx() == endcur.idx()) {
 436                         // Adjust end of selection for replace-all in selection
 437                         if (rsize > ssize) {
 438                                 int const offset = rsize - ssize;
 439                                 for (int i = 0; i < offset + ct_deleted_text
 440                                      && endcur.pos() < endcur.lastpos(); ++i)
 441                                         endcur.forwardPos();
 442                         } else {
 443                                 int const offset = ssize - rsize;
 444                                 for (int i = 0; i < offset && endcur.pos() > 0; ++i)
 445                                         endcur.backwardPos();
 446                                 for (int i = 0; i < ct_deleted_text
 447                                      && endcur.pos() < endcur.lastpos(); ++i)
 448                                         endcur.forwardPos();
 449                         }
 450                 }
 451                 ++num;
 452                 match_len = findForward(cur, endcur, match, false, onlysel);
 453         }
 454
 455         bv->putSelectionAt(doc_iterator_begin(&buf), 0, false);
 456
 457         startcur.fixIfBroken();
 458         bv->setCursor(startcur);
 459
 460         // Reset selection, accounting for changes in selection
 461         if (had_selection) {
 462                 endcur.fixIfBroken();
 463                 bv->cursor().resetAnchor();
 464                 bv->setCursorSelectionTo(endcur);
 465         }
 466
 467         return num;
 468 }
 469
 470
 471 // the idea here is that we are going to replace the string that
 472 // is selected IF it is the search string.
 473 // if there is a selection, but it is not the search string, then
 474 // we basically ignore it. (FIXME We ought to replace only within
 475 // the selection.)
 476 // if there is no selection, then:
 477 //  (i) if some search string has been provided, then we find it.
 478 //      (think of how the dialog works when you hit "replace" the
 479 //      first time.)
 480 // (ii) if no search string has been provided, then we treat the
 481 //      word the cursor is in as the search string. (why? i have no
 482 //      idea.) but this only works in text?
 483 //
 484 // returns the number of replacements made (one, if any) and
 485 // whether anything at all was done.
 486 pair<bool, int> replaceOne(BufferView * bv, docstring searchstr,
 487                            docstring const & replacestr, bool case_sens,
 488                            bool whole, bool forward, bool findnext, bool wrap,
 489                            bool onlysel)
 490 {
 491         Cursor & cur = bv->cursor();
 492         if (!cur.selection() || onlysel) {
 493                 // no selection, non-empty search string: find it
 494                 if (!searchstr.empty()) {
 495                         bool const found = findOne(bv, searchstr, case_sens, whole,
 496                                                    forward, true, findnext, wrap, false, onlysel);
 497                         return make_pair(found, 0);
 498                 }
 499                 // empty search string
 500                 if (!cur.inTexted())
 501                         // bail in math
 502                         return make_pair(false, 0);
 503                 // select current word and treat it as the search string.
 504                 // This causes a minor bug as undo will restore this selection,
 505                 // which the user did not create (#8986).
 506                 cur.innerText()->selectWord(cur, WHOLE_WORD);
 507                 searchstr = cur.selectionAsString(false, true);
 508         }
 509
 510         // if we still don't have a search string, report the error
 511         // and abort.
 512         if (!searchAllowed(searchstr))
 513                 return make_pair(false, 0);
 514
 515         bool have_selection = cur.selection();
 516         docstring const selected = cur.selectionAsString(false, true);
 517         bool match =
 518                 case_sens
 519                 ? searchstr == selected
 520                 : compare_no_case(searchstr, selected) == 0;
 521
 522         // no selection or current selection is not search word:
 523         // just find the search word
 524         if (!have_selection || !match) {
 525                 bool const found = findOne(bv, searchstr, case_sens, whole, forward,
 526                                            true, findnext, wrap, false, onlysel);
 527                 return make_pair(found, 0);
 528         }
 529
 530         // we're now actually ready to replace. if the buffer is
 531         // read-only, we can't, though.
 532         if (bv->buffer().isReadonly())
 533                 return make_pair(false, 0);
 534
 535         cap::replaceSelectionWithString(cur, replacestr);
 536         if (forward) {
 537                 cur.pos() += replacestr.length();
 538                 LASSERT(cur.pos() <= cur.lastpos(),
 539                         cur.pos() = cur.lastpos());
 540         }
 541         if (findnext)
 542                 findOne(bv, searchstr, case_sens, whole,
 543                         forward, false, findnext, wrap, false, onlysel);
 544
 545         return make_pair(true, 1);
 546 }
 547
 548 } // namespace
 549
 550
 551 docstring const find2string(docstring const & search,
 552                             bool casesensitive, bool matchword,
 553                             bool forward, bool wrap, bool instant,
 554                             bool onlysel)
 555 {
 556         odocstringstream ss;
 557         ss << search << '\n'
 558            << int(casesensitive) << ' '
 559            << int(matchword) << ' '
 560            << int(forward) << ' '
 561            << int(wrap) << ' '
 562            << int(instant) << ' '
 563            << int(onlysel);
 564         return ss.str();
 565 }
 566
 567
 568 docstring const replace2string(docstring const & replace,
 569                                docstring const & search,
 570                                bool casesensitive, bool matchword,
 571                                bool all, bool forward, bool findnext,
 572                                bool wrap, bool onlysel)
 573 {
 574         odocstringstream ss;
 575         ss << replace << '\n'
 576            << search << '\n'
 577            << int(casesensitive) << ' '
 578            << int(matchword) << ' '
 579            << int(all) << ' '
 580            << int(forward) << ' '
 581            << int(findnext) << ' '
 582            << int(wrap) << ' '
 583            << int(onlysel);
 584         return ss.str();
 585 }
 586
 587
 588 docstring const string2find(docstring const & argument,
 589                               bool &casesensitive,
 590                               bool &matchword,
 591                               bool &forward,
 592                               bool &wrap,
 593                               bool &instant,
 594                               bool &onlysel)
 595 {
 596         // data is of the form
 597         // "<search>
 598         //  <casesensitive> <matchword> <forward> <wrap> <onlysel>"
 599         docstring search;
 600         docstring howto = split(argument, search, '\n');
 601
 602         casesensitive = parse_bool(howto);
 603         matchword     = parse_bool(howto);
 604         forward       = parse_bool(howto, true);
 605         wrap          = parse_bool(howto);
 606         instant       = parse_bool(howto);
 607         onlysel       = parse_bool(howto);
 608
 609         return search;
 610 }
 611
 612
 613 bool lyxfind(BufferView * bv, FuncRequest const & ev)
 614 {
 615         if (!bv || ev.action() != LFUN_WORD_FIND)
 616                 return false;
 617
 618         //lyxerr << "find called, cmd: " << ev << endl;
 619         bool casesensitive;
 620         bool matchword;
 621         bool forward;
 622         bool wrap;
 623         bool instant;
 624         bool onlysel;
 625
 626         docstring search = string2find(ev.argument(), casesensitive,
 627                                        matchword, forward, wrap, instant, onlysel);
 628
 629         return findOne(bv, search, casesensitive, matchword, forward,
 630                        false, true, wrap, instant, onlysel);
 631 }
 632
 633
 634 bool lyxreplace(BufferView * bv, FuncRequest const & ev)
 635 {
 636         if (!bv || ev.action() != LFUN_WORD_REPLACE)
 637                 return false;
 638
 639         // data is of the form
 640         // "<search>
 641         //  <replace>
 642         //  <casesensitive> <matchword> <all> <forward> <findnext> <wrap> <onlysel>"
 643         docstring search;
 644         docstring rplc;
 645         docstring howto = split(ev.argument(), rplc, '\n');
 646         howto = split(howto, search, '\n');
 647
 648         bool casesensitive = parse_bool(howto);
 649         bool matchword     = parse_bool(howto);
 650         bool all           = parse_bool(howto);
 651         bool forward       = parse_bool(howto, true);
 652         bool findnext      = parse_bool(howto, true);
 653         bool wrap          = parse_bool(howto);
 654         bool onlysel       = parse_bool(howto);
 655
 656         if (!bv->cursor().selection())
 657                 // only selection only makes sense with selection
 658                 onlysel = false;
 659
 660         bool update = false;
 661
 662         int replace_count = 0;
 663         if (all) {
 664                 replace_count = replaceAll(bv, search, rplc, casesensitive,
 665                                            matchword, onlysel);
 666                 update = replace_count > 0;
 667         } else {
 668                 pair<bool, int> rv =
 669                         replaceOne(bv, search, rplc, casesensitive, matchword,
 670                                    forward, findnext, wrap, onlysel);
 671                 update = rv.first;
 672                 replace_count = rv.second;
 673         }
 674
 675         Buffer const & buf = bv->buffer();
 676         if (!update) {
 677                 // emit message signal.
 678                 if (onlysel)
 679                         buf.message(_("String not found in selection."));
 680                 else
 681                         buf.message(_("String not found."));
 682         } else {
 683                 if (replace_count == 0) {
 684                         buf.message(_("String found."));
 685                 } else if (replace_count == 1) {
 686                         buf.message(_("String has been replaced."));
 687                 } else {
 688                         docstring const str = onlysel
 689                                         ? bformat(_("%1$d strings have been replaced in the selection."), replace_count)
 690                                         : bformat(_("%1$d strings have been replaced."), replace_count);
 691                         buf.message(str);
 692                 }
 693         }
 694         return update;
 695 }
 696
 697
 698 bool findNextChange(BufferView * bv, Cursor & cur, bool const check_wrap)
 699 {
 700         for (; cur; cur.forwardPos())
 701                 if (cur.inTexted() && cur.paragraph().isChanged(cur.pos()))
 702                         return true;
 703
 704         if (check_wrap) {
 705                 DocIterator cur_orig(bv->cursor());
 706                 docstring q = _("End of file reached while searching forward.\n"
 707                           "Continue searching from the beginning?");
 708                 int wrap_answer = frontend::Alert::prompt(_("Wrap search?"),
 709                         q, 0, 1, _("&Yes"), _("&No"));
 710                 if (wrap_answer == 0) {
 711                         bv->cursor().clear();
 712                         bv->cursor().push_back(CursorSlice(bv->buffer().inset()));
 713                         bv->clearSelection();
 714                         cur.setCursor(bv->cursor().selectionBegin());
 715                         if (findNextChange(bv, cur, false))
 716                                 return true;
 717                 }
 718                 bv->cursor().setCursor(cur_orig);
 719         }
 720
 721         return false;
 722 }
 723
 724
 725 bool findPreviousChange(BufferView * bv, Cursor & cur, bool const check_wrap)
 726 {
 727         for (cur.backwardPos(); cur; cur.backwardPos()) {
 728                 if (cur.inTexted() && cur.paragraph().isChanged(cur.pos()))
 729                         return true;
 730         }
 731
 732         if (check_wrap) {
 733                 DocIterator cur_orig(bv->cursor());
 734                 docstring q = _("Beginning of file reached while searching backward.\n"
 735                           "Continue searching from the end?");
 736                 int wrap_answer = frontend::Alert::prompt(_("Wrap search?"),
 737                         q, 0, 1, _("&Yes"), _("&No"));
 738                 if (wrap_answer == 0) {
 739                         bv->cursor().setCursor(doc_iterator_end(&bv->buffer()));
 740                         bv->cursor().backwardPos();
 741                         bv->clearSelection();
 742                         cur.setCursor(bv->cursor().selectionBegin());
 743                         if (findPreviousChange(bv, cur, false))
 744                                 return true;
 745                 }
 746                 bv->cursor().setCursor(cur_orig);
 747         }
 748
 749         return false;
 750 }
 751
 752
 753 bool selectChange(Cursor & cur, bool forward)
 754 {
 755         if (!cur.inTexted() || !cur.paragraph().isChanged(cur.pos()))
 756                 return false;
 757         Change ch = cur.paragraph().lookupChange(cur.pos());
 758
 759         CursorSlice tip1 = cur.top();
 760         for (; tip1.pit() < tip1.lastpit() || tip1.pos() < tip1.lastpos(); tip1.forwardPos()) {
 761                 Change ch2 = tip1.paragraph().lookupChange(tip1.pos());
 762                 if (!ch2.isSimilarTo(ch))
 763                         break;
 764         }
 765         CursorSlice tip2 = cur.top();
 766         for (; tip2.pit() > 0 || tip2.pos() > 0;) {
 767                 tip2.backwardPos();
 768                 Change ch2 = tip2.paragraph().lookupChange(tip2.pos());
 769                 if (!ch2.isSimilarTo(ch)) {
 770                         // take a step forward to correctly set the selection
 771                         tip2.forwardPos();
 772                         break;
 773                 }
 774         }
 775         if (forward)
 776                 swap(tip1, tip2);
 777         cur.top() = tip1;
 778         cur.bv().mouseSetCursor(cur, false);
 779         cur.top() = tip2;
 780         cur.bv().mouseSetCursor(cur, true);
 781         return true;
 782 }
 783
 784
 785 namespace {
 786
 787
 788 bool findChange(BufferView * bv, bool forward)
 789 {
 790         Cursor cur(*bv);
 791         cur.setCursor(forward ? bv->cursor().selectionEnd()
 792                       : bv->cursor().selectionBegin());
 793         forward ? findNextChange(bv, cur, true) : findPreviousChange(bv, cur, true);
 794         return selectChange(cur, forward);
 795 }
 796
 797 } // namespace
 798
 799 bool findNextChange(BufferView * bv)
 800 {
 801         return findChange(bv, true);
 802 }
 803
 804
 805 bool findPreviousChange(BufferView * bv)
 806 {
 807         return findChange(bv, false);
 808 }
 809
 810
 811
 812 namespace {
 813
 814 typedef vector<pair<string, string> > Escapes;
 815
 816 static string getRegexSpaceCount(int count)
 817 {
 818         if (count > 0) {
 819                 if (count > 1)
 820                         return "\\s{" + std::to_string(count) + "}";
 821                 else
 822                         return "\\s";
 823         }
 824         return "";
 825 }
 826
 827 string string2regex(string in)
 828 {
 829         static std::regex specialChars { R"([-[\]{}()*+?.,\^$|#\$\\])" };
 830         string tempx = std::regex_replace(in, specialChars,  R"(\$&)" );
 831         // Special handling for ' '
 832         string temp("");
 833         int blanks = 0;
 834         for (unsigned i = 0; i < tempx.size(); i++) {
 835                 if (tempx[i] == ' ' || tempx[i] == '~' ) {
 836                         // normal blanks
 837                         blanks++;
 838                 }
 839                 else if ((tempx[i] == '\302' && tempx[i+1] == '\240')
 840                         || (tempx[i] == '\342' && tempx[i+1] == '\200')) {
 841                         // protected space
 842                         // thin space
 843                         blanks++;
 844                         i++;
 845                 }
 846                 else {
 847                         if (blanks > 0) {
 848                                 temp += getRegexSpaceCount(blanks);
 849                         }
 850                         temp += tempx[i];
 851                         blanks = 0;
 852                 }
 853         }
 854         if (blanks > 0) {
 855                 temp += getRegexSpaceCount(blanks);
 856         }
 857
 858         string temp2("");
 859         size_t lastpos = 0;
 860         size_t fl_pos = 0;
 861         int offset = 1;
 862         while (fl_pos < temp.size()) {
 863                 fl_pos = temp.find("\\\\foreignlanguage", lastpos + offset);
 864                 if (fl_pos == string::npos)
 865                         break;
 866                 offset = 16;
 867                 temp2 += temp.substr(lastpos, fl_pos - lastpos);
 868                 temp2 += "\\n";
 869                 lastpos = fl_pos;
 870         }
 871         if (lastpos == 0)
 872                 return(temp);
 873         if (lastpos < temp.size()) {
 874                 temp2 += temp.substr(lastpos, temp.size() - lastpos);
 875         }
 876         return temp2;
 877 }
 878
 879 static void buildAccentsMap();
 880
 881 string correctRegex(string t, bool withformat)
 882 {
 883         /* Convert \backslash => \
 884          * and \{, \}, \[, \] => {, }, [, ]
 885          */
 886         string s("");
 887         static std::regex wordre("(\\\\)*(\\\\(( |[A-Za-z]+|[\\{\\}%])( |\\{\\})?|[\\[\\]\\{\\}]))");
 888         static std::regex protectedSpace { R"(~)" };
 889         size_t lastpos = 0;
 890         smatch sub;
 891         bool backslashed = false;
 892         if (accents.empty())
 893                 buildAccentsMap();
 894
 895         //LYXERR0("correctRegex input '" << t << "'");
 896         int skip = 0;
 897         for (sregex_iterator it(t.begin(), t.end(), wordre), end; it != end; ++it) {
 898                 sub = *it;
 899                 string replace;
 900                 if ((sub.position(2) - sub.position(0)) % 2 == 1) {
 901                         continue;
 902                 }
 903                 else {
 904                         if (sub.str(4) == "backslash") {
 905                                 replace = "\\";
 906                                 {
 907                                         // transforms '\backslash \{' into '\{'
 908                                         string next = t.substr(sub.position(2) + sub.str(2).length(), 2);
 909                                         if ((next == "\\{") || (next == "\\}") || (next == "\\ ")) {
 910                                                 replace = "";
 911                                                 backslashed = true;
 912                                         }
 913                                         else if (withformat && next[0] == '$') {
 914                                                 replace = accents["lyxdollar"];
 915                                                 skip = 1;       // Skip following '$'
 916                                         }
 917                                 }
 918                         }
 919                         else if (sub.str(4) == "mathcircumflex")
 920                                 replace = "^";
 921                         else if (backslashed) {
 922                                 backslashed = false;
 923                                 if (withformat) {
 924                                         if (sub.str(3) == "{")
 925                                                 replace = accents["braceleft"];
 926                                         else if (sub.str(3) == "}")
 927                                                 replace = accents["braceright"];
 928                                         else if (sub.str(3) == " ")
 929                                                 replace = "\\ ";
 930                                         else {
 931                                                 // else part should not exist
 932                                                 LASSERT(0, /**/);
 933                                         }
 934                                 }
 935                                 else {
 936                                         if (sub.str(3) == "{")
 937                                                 replace = "\\{";
 938                                         else if (sub.str(3) == "}")
 939                                                 replace = "\\}";
 940                                         else if (sub.str(3) == " ")
 941                                                 replace = "\\ ";
 942                                         else {
 943                                                 // else part should not exist
 944                                                 LASSERT(0, /**/);
 945                                         }
 946                                 }
 947                         }
 948                         else if (sub.str(4) == "{") // transforms '\{' into '{'
 949                                 replace = "{";
 950                         else if (sub.str(4) == "}")
 951                                 replace = "}";
 952                         else if (sub.str(4) == "%")
 953                                 replace = "%";
 954                         else if (sub.str(4) == " ")
 955                                 replace = " ";
 956                         else {
 957                                 AccentsIterator it_ac = accents.find(sub.str(4));
 958                                 if (it_ac == accents.end()) {
 959                                         replace = sub.str(2);
 960                                 }
 961                                 else {
 962                                         replace = it_ac->second;
 963                                 }
 964                         }
 965                 }
 966                 if (lastpos < (size_t) sub.position(2))
 967                         s += std::regex_replace(t.substr(lastpos, sub.position(2) - lastpos), protectedSpace, R"( )");
 968                 s += replace;
 969                 lastpos = sub.position(2) + sub.length(2) + skip;
 970                 skip = 0;
 971         }
 972         if (lastpos == 0)
 973                 s = std::regex_replace(t, protectedSpace, R"( )");
 974         else if (lastpos < t.length())
 975                 s += std::regex_replace(t.substr(lastpos, t.length() - lastpos), protectedSpace, R"( )");
 976         // Handle quotes in regex
 977         // substitute all '„', '“', '»', '«' with '"'
 978         // and all '‚', '‘', '›', '‹' with "\'"
 979         static std::regex plainquotes { R"(„|“|»|«)" };
 980         static std::regex innerquotes { R"(‚|‘|›|‹)" };
 981         t = std::regex_replace(s, plainquotes, R"(")");
 982         s = std::regex_replace(t, innerquotes, R"(')");
 983         //LYXERR0("correctRegex output '" << s << "'");
 984         return s;
 985 }
 986
 987 /// Within \regexp{} apply get_lyx_unescapes() only (i.e., preserve regexp semantics of the string),
 988 /// while outside apply get_lyx_unescapes()+get_regexp_escapes().
 989 /// If match_latex is true, then apply regexp_latex_escapes() to \regexp{} contents as well.
 990 string escape_for_regex(string s, bool withformat)
 991 {
 992         size_t lastpos = 0;
 993         string result = "";
 994         while (lastpos < s.size()) {
 995                 size_t regex_pos = s.find("\\regexp{", lastpos);
 996                 if (regex_pos == string::npos) {
 997                         regex_pos = s.size();
 998                 }
 999                 if (regex_pos > lastpos) {
1000                         result += string2regex(s.substr(lastpos, regex_pos-lastpos));
1001                         lastpos = regex_pos;
1002                         if (lastpos == s.size())
1003                                 break;
1004                 }
1005                 size_t end_pos = s.find("\\endregexp", regex_pos + 8);
1006                 result += correctRegex(s.substr(regex_pos + 8, end_pos -(regex_pos + 8)), withformat);
1007                 lastpos = end_pos + 13;
1008         }
1009         return result;
1010 }
1011
1012
1013 /// Wrapper for lyx::regex_replace with simpler interface
1014 bool regex_replace(string const & s, string & t, string const & searchstr,
1015                    string const & replacestr)
1016 {
1017         regex e(searchstr, regex_constants::ECMAScript);
1018         ostringstream oss;
1019         ostream_iterator<char, char> it(oss);
1020         regex_replace(it, s.begin(), s.end(), e, replacestr);
1021         // tolerate t and s be references to the same variable
1022         bool rv = (s != oss.str());
1023         t = oss.str();
1024         return rv;
1025 }
1026
1027 class MatchResult {
1028 public:
1029         enum range {
1030                 newIsTooFar,
1031                 newIsBetter,
1032                 newIsInvalid
1033         };
1034         int match_len;
1035         int match_prefix;
1036         int match2end;
1037         int pos;
1038         int leadsize;
1039         int pos_len;
1040         int searched_size;
1041         vector <string> result = vector <string>();
1042         MatchResult(int len = 0): match_len(len),match_prefix(0),match2end(0), pos(0),leadsize(0),pos_len(-1),searched_size(0) {}
1043 };
1044
1045 static MatchResult::range interpretMatch(MatchResult &oldres, MatchResult &newres)
1046 {
1047         if (newres.match2end < oldres.match2end)
1048                 return MatchResult::newIsTooFar;
1049         if (newres.match_len < oldres.match_len)
1050                 return MatchResult::newIsTooFar;
1051
1052         if (newres.match_len == oldres.match_len) {
1053                 if (newres.match2end == oldres.match2end)
1054                         return MatchResult::newIsBetter;
1055         }
1056         return MatchResult::newIsInvalid;
1057 }
1058
1059 /** The class performing a match between a position in the document and the FindAdvOptions.
1060  **/
1061
1062 class MatchStringAdv {
1063 public:
1064         MatchStringAdv(lyx::Buffer & buf, FindAndReplaceOptions & opt);
1065
1066         /** Tests if text starting at the supplied position matches with the one provided to the MatchStringAdv
1067          ** constructor as opt.search, under the opt.* options settings.
1068          **
1069          ** @param at_begin
1070          **     If set to MatchStringAdv::MatchFromStart,
1071          **       then match is searched only against beginning of text starting at cur.
1072          **     Otherwise the match is searched anywhere in text starting at cur.
1073          **
1074          ** @return
1075          ** The length of the matching text, or zero if no match was found.
1076          **/
1077         enum matchType {
1078                 MatchAnyPlace,
1079                 MatchFromStart
1080         };
1081         string matchTypeAsString(matchType const x) const { return (x == MatchFromStart ? "MatchFromStart" : "MatchAnyPlace"); }
1082         MatchResult operator()(DocIterator const & cur, int len, matchType at_begin) const;
1083 #if QTSEARCH
1084         bool regexIsValid;
1085         string regexError;
1086 #endif
1087
1088 public:
1089         /// buffer
1090         lyx::Buffer * p_buf;
1091         /// first buffer on which search was started
1092         lyx::Buffer * const p_first_buf;
1093         /// options
1094         FindAndReplaceOptions const & opt;
1095
1096 private:
1097         /// Auxiliary find method (does not account for opt.matchword)
1098         MatchResult findAux(DocIterator const & cur, int len, matchType at_begin) const;
1099         void CreateRegexp(FindAndReplaceOptions const & opt, string regexp_str, string regexp2_str, string par_as_string = "");
1100
1101         /** Normalize a stringified or latexified LyX paragraph.
1102          **
1103          ** Normalize means:
1104          ** <ul>
1105          **   <li>if search is not casesensitive, then lowercase the string;
1106          **   <li>remove any newline at begin or end of the string;
1107          **   <li>replace any newline in the middle of the string with a simple space;
1108          **   <li>remove stale empty styles and environments, like \emph{} and \textbf{}.
1109          ** </ul>
1110          **
1111          ** @todo Normalization should also expand macros, if the corresponding
1112          ** search option was checked.
1113          **/
1114         string normalize(docstring const & s, bool ignore_fomat) const;
1115         // normalized string to search
1116         string par_as_string;
1117         // regular expression to use for searching
1118         // regexp2 is same as regexp, but prefixed with a ".*?"
1119 #if QTSEARCH
1120         QRegularExpression regexp;
1121         QRegularExpression regexp2;
1122 #else
1123         regex regexp;
1124         regex regexp2;
1125 #endif
1126         // leading format material as string
1127         string lead_as_string;
1128         // par_as_string after removal of lead_as_string
1129         string par_as_string_nolead;
1130         // unmatched open braces in the search string/regexp
1131         int open_braces;
1132         // number of (.*?) subexpressions added at end of search regexp for closing
1133         // environments, math mode, styles, etc...
1134         int close_wildcards;
1135 public:
1136         // Are we searching with regular expressions ?
1137         bool use_regexp = false;
1138         static int valid_matches;
1139         static vector <string> matches;
1140         void FillResults(MatchResult &found_mr);
1141 };
1142
1143 int MatchStringAdv::valid_matches = 0;
1144 vector <string> MatchStringAdv::matches = vector <string>(10);
1145
1146 void MatchStringAdv::FillResults(MatchResult &found_mr)
1147 {
1148         if (found_mr.match_len > 0) {
1149                 valid_matches = found_mr.result.size();
1150                 for (size_t i = 0; i < found_mr.result.size(); i++)
1151                         matches[i] = found_mr.result[i];
1152         } else
1153                 valid_matches = 0;
1154 }
1155
1156 static void setFindParams(OutputParams &runparams)
1157 {
1158         runparams.flavor = Flavor::XeTeX;
1159         //runparams.use_polyglossia = true;
1160         runparams.linelen = 10000; //lyxrc.plaintext_linelen;
1161         // No side effect of file copying and image conversion
1162         runparams.dryrun = true;
1163 }
1164
1165 static docstring buffer_to_latex(Buffer & buffer)
1166 {
1167         //OutputParams runparams(&buffer.params().encoding());
1168         OutputParams runparams(encodings.fromLyXName("utf8"));
1169         odocstringstream ods;
1170         otexstream os(ods);
1171         runparams.nice = true;
1172         setFindParams(runparams);
1173         if (ignoreFormats.getDeleted())
1174                 runparams.find_set_feature(OutputParams::SearchWithoutDeleted);
1175         else
1176                 runparams.find_set_feature(OutputParams::SearchWithDeleted);
1177         if (ignoreFormats.getNonContent()) {
1178                 runparams.find_add_feature(OutputParams::SearchNonOutput);
1179         }
1180         pit_type const endpit = buffer.paragraphs().size();
1181         for (pit_type pit = 0; pit != endpit; ++pit) {
1182                 TeXOnePar(buffer, buffer.text(), pit, os, runparams, string(), -1, -1, true);
1183                 LYXERR(Debug::FINDVERBOSE, "searchString up to here: " << ods.str());
1184         }
1185         return ods.str();
1186 }
1187
1188 static string latexNamesToUtf8(docstring strIn, bool withformat)
1189 {
1190         string addtmp = to_utf8(strIn);
1191         static regex const rmAcc("(\\\\)*("
1192                                          "\\\\([A-Za-z]+\\{.\\})"       // e.g. "ddot{A}" == sub.str(3)
1193                                         "|\\\\([A-Za-z]+)( |\\{\\})?"   // e.g. "LyX", "LyX{}", "LyX " == sub.str(4)
1194                                         ")"
1195                                 );
1196         size_t lastpos = 0;
1197         smatch sub;
1198         string replace;
1199         string add("");
1200         if (accents.empty())
1201                 buildAccentsMap();
1202         for (sregex_iterator it_add(addtmp.begin(), addtmp.end(), rmAcc), end; it_add != end; ++it_add) {
1203                 sub = *it_add;
1204                 if ((sub.position(2) - sub.position(0)) % 2 == 1) {
1205                         continue;
1206                 }
1207                 else {
1208                         string key;
1209                         if (sub.length(3) > 0)
1210                                 key = sub.str(3);
1211                         else
1212                                 key = sub.str(4);
1213                         AccentsIterator it_ac = accents.find(key);
1214                         if (it_ac == accents.end()) {
1215                                 replace = sub.str(2);
1216                         }
1217                         else {
1218                                 replace = it_ac->second;
1219                         }
1220                 }
1221                 if (lastpos < (size_t) sub.position(2))
1222                         add += addtmp.substr(lastpos, sub.position(2) - lastpos);
1223                 add += replace;
1224                 lastpos = sub.position(2) + sub.length(2);
1225         }
1226         if (lastpos == 0)
1227                 add = addtmp;
1228         else if (addtmp.length() > lastpos)
1229                 add += addtmp.substr(lastpos, addtmp.length() - lastpos);
1230         if (!withformat) {
1231                 static std::regex repltilde { R"(~)" };
1232                 add = std::regex_replace(add, repltilde, accents["lyxtilde"]);
1233         }
1234         LYXERR(Debug::FINDVERBOSE, "Adding to search string: '"
1235                         << add << "'");
1236         return add;
1237 }
1238
1239 static docstring stringifySearchBuffer(Buffer & buffer, FindAndReplaceOptions const & opt)
1240 {
1241         docstring str;
1242         if (!opt.ignoreformat) {
1243                 str = buffer_to_latex(buffer);
1244         } else {
1245                 // OutputParams runparams(&buffer.params().encoding());
1246                 OutputParams runparams(encodings.fromLyXName("utf8"));
1247                 runparams.nice = true;
1248                 setFindParams(runparams);
1249                 int option = AS_STR_INSETS |AS_STR_PLAINTEXT;
1250                 if (ignoreFormats.getDeleted()) {
1251                         option |= AS_STR_SKIPDELETE;
1252                         runparams.find_set_feature(OutputParams::SearchWithoutDeleted);
1253                 }
1254                 else {
1255                         runparams.find_set_feature(OutputParams::SearchWithDeleted);
1256                 }
1257                 if (ignoreFormats.getNonContent()) {
1258                         runparams.find_add_feature(OutputParams::SearchNonOutput);
1259                 }
1260                 string t("");
1261                 for (pos_type pit = pos_type(0); pit < (pos_type)buffer.paragraphs().size(); ++pit) {
1262                         Paragraph const & par = buffer.paragraphs().at(pit);
1263                         string add = latexNamesToUtf8(par.asString(pos_type(0), par.size(),
1264                                                                 option,
1265                                                                 &runparams), !opt.ignoreformat);
1266                         LYXERR(Debug::FINDVERBOSE, "Adding to search string: '"
1267                                 << add << "'");
1268                         t += add;
1269                 }
1270                 // Even in ignore-format we have to remove "\text{}, \lyxmathsym{}" parts
1271                 while (regex_replace(t, t, "\\\\(text|lyxmathsym|ensuremath)\\{([^\\}]*)\\}", "$2"));
1272                 str = from_utf8(t);
1273         }
1274         return str;
1275 }
1276
1277
1278 /// Return separation pos between the leading material and the rest
1279 static size_t identifyLeading(string const & s)
1280 {
1281         string t = s;
1282         // @TODO Support \item[text]
1283         // Kornel: Added textsl, textsf, textit, texttt and noun
1284         // + allow to search for colored text too
1285         while (regex_replace(t, t, "^\\\\(("
1286                              "(author|title|subtitle|subject|publishers|dedication|uppertitleback|lowertitleback|extratitle|"
1287                              "lyxaddress|lyxrightaddress|"
1288                              "footnotesize|tiny|scriptsize|small|large|Large|LARGE|huge|Huge|"
1289                              "emph|noun|minisec|text(bf|md|sl|sf|it|tt))|"
1290                              "((textcolor|foreignlanguage|latexenvironment)\\{[a-z]+\\*?\\})|"
1291                              "(u|uu)line|(s|x)out|uwave)|((sub)?(((sub)?section)|paragraph)|part|chapter)\\*?)\\{", "")
1292                || regex_replace(t, t, "^\\$", "")
1293                || regex_replace(t, t, "^\\\\\\[", "")
1294                || regex_replace(t, t, "^ ?\\\\item\\{[a-z]+\\}", "")
1295                || regex_replace(t, t, "^\\\\begin\\{[a-zA-Z_]*\\*?\\}", ""))
1296                ;
1297         LYXERR(Debug::FINDVERBOSE, "  after removing leading $, \\[ , \\emph{, \\textbf{, etc.: '" << t << "'");
1298         return s.size() - t.size();
1299 }
1300
1301 /*
1302  * Given a latexified string, retrieve some handled features
1303  * The features of the regex will later be compared with the features
1304  * of the searched text. If the regex features are not a
1305  * subset of the analized, then, in not format ignoring search
1306  * we can early stop the search in the relevant inset.
1307  */
1308 typedef map<string, bool> Features;
1309
1310 static Features identifyFeatures(string const & s)
1311 {
1312         static regex const feature("\\\\(([a-zA-Z]+(\\{([a-z]+\\*?)\\}|\\*)?))\\{");
1313         static regex const valid("^("
1314                 "("
1315                         "(footnotesize|tiny|scriptsize|small|large|Large|LARGE|huge|Huge|"
1316                                 "emph|noun|text(bf|md|sl|sf|it|tt)|"
1317                                 "(textcolor|foreignlanguage|item|listitem|latexenvironment)\\{[a-z]+\\*?\\})|"
1318                         "(u|uu)line|(s|x)out|uwave|"
1319                         "(sub|extra)?title|author|subject|publishers|dedication|(upper|lower)titleback|lyx(right)?address)|"
1320                 "((sub)?(((sub)?section)|paragraph)|part|chapter|lyxslide)\\*?)$");
1321         smatch sub;
1322         bool displ = true;
1323         Features info;
1324
1325         for (sregex_iterator it(s.begin(), s.end(), feature), end; it != end; ++it) {
1326                 sub = *it;
1327                 if (displ) {
1328                         if (sub.str(1).compare("regexp") == 0) {
1329                                 displ = false;
1330                                 continue;
1331                         }
1332                         string token = sub.str(1);
1333                         smatch sub2;
1334                         if (regex_match(token, sub2, valid)) {
1335                                 info[token] = true;
1336                         }
1337                         else {
1338                                 // ignore
1339                         }
1340                 }
1341                 else {
1342                         if (sub.str(1).compare("endregexp") == 0) {
1343                                 displ = true;
1344                                 continue;
1345                         }
1346                 }
1347         }
1348         return info;
1349 }
1350
1351 /*
1352  * defines values features of a key "\\[a-z]+{"
1353  */
1354 class KeyInfo {
1355 public:
1356         enum KeyType {
1357                 /* Char type with content discarded
1358                  * like \hspace{1cm} */
1359                 noContent,
1360                 /* Char, like \backslash */
1361                 isChar,
1362                 /* replace starting backslash with '#' */
1363                 isText,
1364                 /* \part, \section*, ... */
1365                 isSectioning,
1366                 /* title, author etc */
1367                 isTitle,
1368                 /* \foreignlanguage{ngerman}, ... */
1369                 isMain,
1370                 /* inside \code{}
1371                  * to discard language in content */
1372                 noMain,
1373                 isRegex,
1374                 /* \begin{eqnarray}...\end{eqnarray}, ... $...$ */
1375                 isMath,
1376                 /* fonts, colors, markups, ... */
1377                 isStandard,
1378                 /* footnotesize, ... large, ...
1379                  * Ignore all of them */
1380                 isSize,
1381                 invalid,
1382                 /* inputencoding, ...
1383                  * Discard also content, because they do not help in search */
1384                 doRemove,
1385                 /* twocolumns, ...
1386                  * like remove, but also all arguments */
1387                 removeWithArg,
1388                 /* item, listitem */
1389                 isList,
1390                 /* tex, latex, ... like isChar */
1391                 isIgnored,
1392                 /* like \lettrine[lines=5]{}{} */
1393                 cleanToStart,
1394                 // like isStandard, but always remove head
1395                 headRemove,
1396                 /* End of arguments marker for lettrine,
1397                  * so that they can be ignored */
1398                 endArguments
1399         };
1400         KeyInfo() = default;
1401         KeyInfo(KeyType type, int parcount, bool disable)
1402                 : keytype(type),
1403                   parenthesiscount(parcount),
1404                   disabled(disable) {}
1405         KeyType keytype = invalid;
1406         string head;
1407         int _tokensize = -1;
1408         int _tokenstart = -1;
1409         int _dataStart = -1;
1410         int _dataEnd = -1;
1411         int parenthesiscount = 1;
1412         bool disabled = false;
1413         bool used = false; /* by pattern */
1414 };
1415
1416 class Border {
1417 public:
1418         Border(int l=0, int u=0) : low(l), upper(u) {}
1419         int low;
1420         int upper;
1421 };
1422
1423 #define MAXOPENED 30
1424 class Intervall {
1425         bool isPatternString_;
1426 public:
1427         explicit Intervall(bool isPattern, string const & p)
1428                 : isPatternString_(isPattern), par(p), ignoreidx(-1),
1429                   actualdeptindex(0), hasTitle(false), langcount(0)
1430         {
1431                 depts[0] = 0;
1432                 closes[0] = 0;
1433         }
1434
1435         string par;
1436         int ignoreidx;
1437         static vector<Border> borders;
1438         static vector<int> depts;
1439         static vector<int> closes;
1440         int actualdeptindex;
1441         int previousNotIgnored(int) const;
1442         int nextNotIgnored(int) const;
1443         void handleOpenP(int i);
1444         void handleCloseP(int i, bool closingAllowed);
1445         void resetOpenedP(int openPos);
1446         void addIntervall(int upper);
1447         void addIntervall(int low, int upper); /* if explicit */
1448         void removeAccents();
1449         void setForDefaultLang(KeyInfo const & defLang) const;
1450         int findclosing(int start, int end, char up, char down, int repeat);
1451         void removeInvalidClosings(void);
1452         void handleParentheses(int lastpos, bool closingAllowed);
1453         bool hasTitle;
1454         // Number of disabled language specs up
1455         // to current position in actual interval
1456         int langcount;
1457         int isOpeningPar(int pos) const;
1458         string titleValue;
1459         void output(ostringstream &os, int lastpos);
1460         // string show(int lastpos);
1461 };
1462
1463 vector<Border> Intervall::borders = vector<Border>(30);
1464 vector<int> Intervall::depts = vector<int>(30);
1465 vector<int> Intervall::closes = vector<int>(30);
1466
1467 int Intervall::isOpeningPar(int pos) const
1468 {
1469         if ((pos < 0) || (size_t(pos) >= par.size()))
1470                 return 0;
1471         if (par[pos] != '{')
1472                 return 0;
1473         if (size_t(pos) + 2 >= par.size())
1474                 return 1;
1475         if (par[pos+2] != '}')
1476                 return 1;
1477         if (par[pos+1] == '[' || par[pos+1] == ']')
1478                 return 3;
1479         return 1;
1480 }
1481
1482 void Intervall::setForDefaultLang(KeyInfo const & defLang) const
1483 {
1484         // Enable the use of first token again
1485         if (ignoreidx >= 0) {
1486                 int value = defLang._tokenstart + defLang._tokensize;
1487                 int borderidx = 0;
1488                 if (hasTitle)
1489                         borderidx = 1;
1490                 if (value > 0) {
1491                         if (borders[borderidx].low < value)
1492                                 borders[borderidx].low = value;
1493                         if (borders[borderidx].upper < value)
1494                                 borders[borderidx].upper = value;
1495                 }
1496         }
1497 }
1498
1499 #if 0
1500 // Not needed, because dpts and closes are now dynamically expanded
1501 static void checkDepthIndex(int val)
1502 {
1503         static int maxdepthidx = MAXOPENED-2;
1504         static int lastmaxdepth = 0;
1505         if (val > lastmaxdepth) {
1506                 LYXERR(Debug::INFO, "Depth reached " << val);
1507                 lastmaxdepth = val;
1508         }
1509         if (val > maxdepthidx) {
1510                 maxdepthidx = val;
1511                 LYXERR(Debug::INFO, "maxdepthidx now " << val);
1512         }
1513 }
1514 #endif
1515
1516 #if 0
1517 // Not needed, because borders are now dynamically expanded
1518 static void checkIgnoreIdx(int val)
1519 {
1520         static int lastmaxignore = -1;
1521         if ((lastmaxignore < val) && (size_t(val+1) >= borders.size())) {
1522                 LYXERR(Debug::INFO, "IgnoreIdx reached " << val);
1523                 lastmaxignore = val;
1524         }
1525 }
1526 #endif
1527
1528 /*
1529  * Expand the region of ignored parts of the input latex string
1530  * The region is only relevant in output()
1531  */
1532 void Intervall::addIntervall(int low, int upper)
1533 {
1534         int idx;
1535         if (low == upper) return;
1536         for (idx = ignoreidx+1; idx > 0; --idx) {
1537                 if (low > borders[idx-1].upper) {
1538                         break;
1539                 }
1540         }
1541         Border br(low, upper);
1542         if (idx > ignoreidx) {
1543                 if (borders.size() <= size_t(idx)) {
1544                         borders.push_back(br);
1545                 }
1546                 else {
1547                         borders[idx] = br;
1548                 }
1549                 ignoreidx = idx;
1550                 // checkIgnoreIdx(ignoreidx);
1551                 return;
1552         }
1553         else {
1554                 // Expand only if one of the new bound is inside the interwall
1555                 // We know here that br.low > borders[idx-1].upper
1556                 if (br.upper < borders[idx].low) {
1557                         // We have to insert at this pos
1558                         if (size_t(ignoreidx+1) >= borders.size()) {
1559                                 borders.push_back(borders[ignoreidx]);
1560                         }
1561                         else {
1562                                 borders[ignoreidx+1] = borders[ignoreidx];
1563                         }
1564                         for (int i = ignoreidx; i > idx; --i) {
1565                                 borders[i] = borders[i-1];
1566                         }
1567                         borders[idx] = br;
1568                         ignoreidx += 1;
1569                         // checkIgnoreIdx(ignoreidx);
1570                         return;
1571                 }
1572                 // Here we know, that we are overlapping
1573                 if (br.low > borders[idx].low)
1574                         br.low = borders[idx].low;
1575                 // check what has to be concatenated
1576                 int count = 0;
1577                 for (int i = idx; i <= ignoreidx; i++) {
1578                         if (br.upper >= borders[i].low) {
1579                                 count++;
1580                                 if (br.upper < borders[i].upper)
1581                                         br.upper = borders[i].upper;
1582                         }
1583                         else {
1584                                 break;
1585                         }
1586                 }
1587                 // count should be >= 1 here
1588                 borders[idx] = br;
1589                 if (count > 1) {
1590                         for (int i = idx + count; i <= ignoreidx; i++) {
1591                                 borders[i-count+1] = borders[i];
1592                         }
1593                         ignoreidx -= count - 1;
1594                         return;
1595                 }
1596         }
1597 }
1598
1599 static void buildaccent(string n, string param, string values)
1600 {
1601         stringstream s(n);
1602         string name;
1603         const char delim = '|';
1604         while (getline(s, name, delim)) {
1605                 size_t start = 0;
1606                 for (char c : param) {
1607                         string key = name + "{" + c + "}";
1608                         // get the corresponding utf8-value
1609                         if ((values[start] & 0xc0) != 0xc0) {
1610                                 // should not happen, utf8 encoding starts at least with 11xxxxxx
1611                                 // but value for '\dot{i}' is 'i', which is ascii
1612                                 if ((values[start] & 0x80) == 0) {
1613                                         // is ascii
1614                                         accents[key] = values.substr(start, 1);
1615                                         // LYXERR(Debug::INFO, "" << key << "=" << accents[key]);
1616                                 }
1617                                 start++;
1618                                 continue;
1619                         }
1620                         for (int j = 1; ;j++) {
1621                                 if (start + j >= values.size()) {
1622                                         accents[key] = values.substr(start, j);
1623                                         start = values.size() - 1;
1624                                         break;
1625                                 }
1626                                 else if ((values[start+j] & 0xc0) != 0x80) {
1627                                         // This is the first byte of following utf8 char
1628                                         accents[key] = values.substr(start, j);
1629                                         start += j;
1630                                         // LYXERR(Debug::INFO, "" << key << "=" << accents[key]);
1631                                         break;
1632                                 }
1633                         }
1634                 }
1635         }
1636 }
1637
1638 // Helper function
1639 static string getutf8(unsigned uchar)
1640 {
1641 #define maxc 5
1642         string ret = string();
1643         char c[maxc] = {0};
1644         if (uchar <= 0x7f) {
1645                 c[maxc-1] = uchar & 0x7f;
1646         }
1647         else {
1648                 unsigned char rest = 0x40;
1649                 unsigned char first = 0x80;
1650                 int start = maxc-1;
1651                 for (int i = start; i >=0; --i) {
1652                         if (uchar < rest) {
1653                                 c[i] = first + uchar;
1654                                 break;
1655                         }
1656                         c[i] = 0x80 | (uchar &  0x3f);
1657                         uchar >>= 6;
1658                         rest >>= 1;
1659                         first >>= 1;
1660                         first |= 0x80;
1661                 }
1662         }
1663         for (int i = 0; i < maxc; i++) {
1664                 if (c[i] == 0) continue;
1665                 ret += c[i];
1666         }
1667         return(ret);
1668 }
1669
1670 static void addAccents(string latex_in, string unicode_out)
1671 {
1672         latex_in = latex_in.substr(1);
1673         AccentsIterator it_ac = accents.find(latex_in);
1674         if (it_ac == accents.end()) {
1675                 accents[latex_in] = unicode_out;
1676         }
1677         else {
1678                 LYXERR0("Key " << latex_in  << " already set");
1679         }
1680 }
1681
1682 void static fillMissingUnicodesymbols()
1683 {
1684         addAccents("\\pounds", getutf8(0x00a3));
1685         addAccents("\\textsterling", getutf8(0x00a3));
1686         addAccents("\\textyen", getutf8(0x00a5));
1687         addAccents("\\yen", getutf8(0x00a5));
1688         addAccents("\\textsection", getutf8(0x00a7));
1689         addAccents("\\mathsection", getutf8(0x00a7));
1690         addAccents("\\textcopyright", getutf8(0x00a9));
1691         addAccents("\\copyright", getutf8(0x00a9));
1692         addAccents("\\textlnot", getutf8(0x00ac));
1693         addAccents("\\neg", getutf8(0x00ac));
1694         addAccents("\\textregistered", getutf8(0x00ae));
1695         addAccents("\\circledR", getutf8(0x00ae));
1696         addAccents("\\textpm", getutf8(0x00b1));
1697         addAccents("\\pm", getutf8(0x00b1));
1698         addAccents("\\textparagraph", getutf8(0x00b6));
1699         addAccents("\\mathparagraph", getutf8(0x00b6));
1700         addAccents("\\textperiodcentered", getutf8(0x00b7));
1701         addAccents("\\texttimes", getutf8(0x00d7));
1702         addAccents("\\times", getutf8(0x00d7));
1703         addAccents("\\O", getutf8(0x00d8));
1704         addAccents("\\dh", getutf8(0x00f0));
1705         addAccents("\\eth", getutf8(0x00f0));
1706         addAccents("\\textdiv", getutf8(0x00f7));
1707         addAccents("\\div", getutf8(0x00f7));
1708         addAccents("\\o", getutf8(0x00f8));
1709         addAccents("\\textcrlambda", getutf8(0x019b));
1710         addAccents("\\j", getutf8(0x0237));
1711         addAccents("\\textrevepsilon", getutf8(0x025c));
1712         addAccents("\\textbaru", getutf8(0x0289));
1713         addAccents("\\textquoteleft", getutf8(0x02bb));
1714         addAccents("\\textGamma", getutf8(0x0393));
1715         addAccents("\\Gamma", getutf8(0x0393));
1716         addAccents("\\textDelta", getutf8(0x0394));
1717         addAccents("\\Delta", getutf8(0x0394));
1718         addAccents("\\textTheta", getutf8(0x0398));
1719         addAccents("\\Theta", getutf8(0x0398));
1720         addAccents("\\textLambda", getutf8(0x039b));
1721         addAccents("\\Lambda", getutf8(0x039b));
1722         addAccents("\\textXi", getutf8(0x039e));
1723         addAccents("\\Xi", getutf8(0x039e));
1724         addAccents("\\textPi", getutf8(0x03a0));
1725         addAccents("\\Pi", getutf8(0x03a0));
1726         addAccents("\\textSigma", getutf8(0x03a3));
1727         addAccents("\\Sigma", getutf8(0x03a3));
1728         addAccents("\\textUpsilon", getutf8(0x03a5));
1729         addAccents("\\Upsilon", getutf8(0x03a5));
1730         addAccents("\\textPhi", getutf8(0x03a6));
1731         addAccents("\\Phi", getutf8(0x03a6));
1732         addAccents("\\textPsi", getutf8(0x03a8));
1733         addAccents("\\Psi", getutf8(0x03a8));
1734         addAccents("\\textOmega", getutf8(0x03a9));
1735         addAccents("\\Omega", getutf8(0x03a9));
1736         addAccents("\\textalpha", getutf8(0x03b1));
1737         addAccents("\\alpha", getutf8(0x03b1));
1738         addAccents("\\textbeta", getutf8(0x03b2));
1739         addAccents("\\beta", getutf8(0x03b2));
1740         addAccents("\\textgamma", getutf8(0x03b3));
1741         addAccents("\\gamma", getutf8(0x03b3));
1742         addAccents("\\textdelta", getutf8(0x03b4));
1743         addAccents("\\delta", getutf8(0x03b4));
1744         addAccents("\\textepsilon", getutf8(0x03b5));
1745         addAccents("\\varepsilon", getutf8(0x03b5));
1746         addAccents("\\textzeta", getutf8(0x03b6));
1747         addAccents("\\zeta", getutf8(0x03b6));
1748         addAccents("\\texteta", getutf8(0x03b7));
1749         addAccents("\\eta", getutf8(0x03b7));
1750         addAccents("\\texttheta", getutf8(0x03b8));
1751         addAccents("\\theta", getutf8(0x03b8));
1752         addAccents("\\textiota", getutf8(0x03b9));
1753         addAccents("\\iota", getutf8(0x03b9));
1754         addAccents("\\textkappa", getutf8(0x03ba));
1755         addAccents("\\kappa", getutf8(0x03ba));
1756         addAccents("\\textlambda", getutf8(0x03bb));
1757         addAccents("\\lambda", getutf8(0x03bb));
1758         addAccents("\\textmu", getutf8(0x03bc));
1759         addAccents("\\mu", getutf8(0x03bc));
1760         addAccents("\\textnu", getutf8(0x03bd));
1761         addAccents("\\nu", getutf8(0x03bd));
1762         addAccents("\\textxi", getutf8(0x03be));
1763         addAccents("\\xi", getutf8(0x03be));
1764         addAccents("\\textpi", getutf8(0x03c0));
1765         addAccents("\\pi", getutf8(0x03c0));
1766         addAccents("\\textrho", getutf8(0x03c1));
1767         addAccents("\\rho", getutf8(0x03c1));
1768         addAccents("\\textfinalsigma", getutf8(0x03c2));
1769         addAccents("\\varsigma", getutf8(0x03c2));
1770         addAccents("\\textsigma", getutf8(0x03c3));
1771         addAccents("\\sigma", getutf8(0x03c3));
1772         addAccents("\\texttau", getutf8(0x03c4));
1773         addAccents("\\tau", getutf8(0x03c4));
1774         addAccents("\\textupsilon", getutf8(0x03c5));
1775         addAccents("\\upsilon", getutf8(0x03c5));
1776         addAccents("\\textphi", getutf8(0x03c6));
1777         addAccents("\\varphi", getutf8(0x03c6));
1778         addAccents("\\textchi", getutf8(0x03c7));
1779         addAccents("\\chi", getutf8(0x03c7));
1780         addAccents("\\textpsi", getutf8(0x03c8));
1781         addAccents("\\psi", getutf8(0x03c8));
1782         addAccents("\\textomega", getutf8(0x03c9));
1783         addAccents("\\omega", getutf8(0x03c9));
1784         addAccents("\\textdigamma", getutf8(0x03dd));
1785         addAccents("\\digamma", getutf8(0x03dd));
1786         addAccents("\\hebalef", getutf8(0x05d0));
1787         addAccents("\\aleph", getutf8(0x05d0));
1788         addAccents("\\hebbet", getutf8(0x05d1));
1789         addAccents("\\beth", getutf8(0x05d1));
1790         addAccents("\\hebgimel", getutf8(0x05d2));
1791         addAccents("\\gimel", getutf8(0x05d2));
1792         addAccents("\\hebdalet", getutf8(0x05d3));
1793         addAccents("\\daleth", getutf8(0x05d3));
1794         addAccents("\\hebhe", getutf8(0x05d4));
1795         addAccents("\\hebvav", getutf8(0x05d5));
1796         addAccents("\\hebzayin", getutf8(0x05d6));
1797         addAccents("\\hebhet", getutf8(0x05d7));
1798         addAccents("\\hebtet", getutf8(0x05d8));
1799         addAccents("\\hebyod", getutf8(0x05d9));
1800         addAccents("\\hebfinalkaf", getutf8(0x05da));
1801         addAccents("\\hebkaf", getutf8(0x05db));
1802         addAccents("\\heblamed", getutf8(0x05dc));
1803         addAccents("\\hebfinalmem", getutf8(0x05dd));
1804         addAccents("\\hebmem", getutf8(0x05de));
1805         addAccents("\\hebfinalnun", getutf8(0x05df));
1806         addAccents("\\hebnun", getutf8(0x05e0));
1807         addAccents("\\hebsamekh", getutf8(0x05e1));
1808         addAccents("\\hebayin", getutf8(0x05e2));
1809         addAccents("\\hebfinalpe", getutf8(0x05e3));
1810         addAccents("\\hebpe", getutf8(0x05e4));
1811         addAccents("\\hebfinaltsadi", getutf8(0x05e5));
1812         addAccents("\\hebtsadi", getutf8(0x05e6));
1813         addAccents("\\hebqof", getutf8(0x05e7));
1814         addAccents("\\hebresh", getutf8(0x05e8));
1815         addAccents("\\hebshin", getutf8(0x05e9));
1816         addAccents("\\hebtav", getutf8(0x05ea));
1817
1818         // Thai characters
1819         addAccents("\\thaiKoKai", getutf8(0x0e01));
1820         addAccents("\\thaiKhoKhai", getutf8(0x0e02));
1821         addAccents("\\thaiKhoKhuat", getutf8(0x0e03));
1822         addAccents("\\thaiKhoKhwai", getutf8(0x0e04));
1823         addAccents("\\thaiKhoKhon", getutf8(0x0e05));
1824         addAccents("\\thaiKhoRakhang", getutf8(0x0e06));
1825         addAccents("\\thaiNgoNgu", getutf8(0x0e07));
1826         addAccents("\\thaiChoChan", getutf8(0x0e08));
1827         addAccents("\\thaiChoChing", getutf8(0x0e09));
1828         addAccents("\\thaiChoChang", getutf8(0x0e0a));
1829         addAccents("\\thaiSoSo", getutf8(0x0e0b));
1830         addAccents("\\thaiChoChoe", getutf8(0x0e0c));
1831         addAccents("\\thaiYoYing", getutf8(0x0e0d));
1832         addAccents("\\thaiDoChada", getutf8(0x0e0e));
1833         addAccents("\\thaiToPatak", getutf8(0x0e0f));
1834         addAccents("\\thaiThoThan", getutf8(0x0e10));
1835         addAccents("\\thaiThoNangmontho", getutf8(0x0e11));
1836         addAccents("\\thaiThoPhuthao", getutf8(0x0e12));
1837         addAccents("\\thaiNoNen", getutf8(0x0e13));
1838         addAccents("\\thaiDoDek", getutf8(0x0e14));
1839         addAccents("\\thaiToTao", getutf8(0x0e15));
1840         addAccents("\\thaiThoThung", getutf8(0x0e16));
1841         addAccents("\\thaiThoThahan", getutf8(0x0e17));
1842         addAccents("\\thaiThoThong", getutf8(0x0e18));
1843         addAccents("\\thaiNoNu", getutf8(0x0e19));
1844         addAccents("\\thaiBoBaimai", getutf8(0x0e1a));
1845         addAccents("\\thaiPoPla", getutf8(0x0e1b));
1846         addAccents("\\thaiPhoPhung", getutf8(0x0e1c));
1847         addAccents("\\thaiFoFa", getutf8(0x0e1d));
1848         addAccents("\\thaiPhoPhan", getutf8(0x0e1e));
1849         addAccents("\\thaiFoFan", getutf8(0x0e1f));
1850         addAccents("\\thaiPhoSamphao", getutf8(0x0e20));
1851         addAccents("\\thaiMoMa", getutf8(0x0e21));
1852         addAccents("\\thaiYoYak", getutf8(0x0e22));
1853         addAccents("\\thaiRoRua", getutf8(0x0e23));
1854         addAccents("\\thaiRu", getutf8(0x0e24));
1855         addAccents("\\thaiLoLing", getutf8(0x0e25));
1856         addAccents("\\thaiLu", getutf8(0x0e26));
1857         addAccents("\\thaiWoWaen", getutf8(0x0e27));
1858         addAccents("\\thaiSoSala", getutf8(0x0e28));
1859         addAccents("\\thaiSoRusi", getutf8(0x0e29));
1860         addAccents("\\thaiSoSua", getutf8(0x0e2a));
1861         addAccents("\\thaiHoHip", getutf8(0x0e2b));
1862         addAccents("\\thaiLoChula", getutf8(0x0e2c));
1863         addAccents("\\thaiOAng", getutf8(0x0e2d));
1864         addAccents("\\thaiHoNokhuk", getutf8(0x0e2e));
1865         addAccents("\\thaiPaiyannoi", getutf8(0x0e2f));
1866         addAccents("\\thaiSaraA", getutf8(0x0e30));
1867         addAccents("\\thaiMaiHanakat", getutf8(0x0e31));
1868         addAccents("\\thaiSaraAa", getutf8(0x0e32));
1869         addAccents("\\thaiSaraAm", getutf8(0x0e33));
1870         addAccents("\\thaiSaraI", getutf8(0x0e34));
1871         addAccents("\\thaiSaraIi", getutf8(0x0e35));
1872         addAccents("\\thaiSaraUe", getutf8(0x0e36));
1873         addAccents("\\thaiSaraUee", getutf8(0x0e37));
1874         addAccents("\\thaiSaraU", getutf8(0x0e38));
1875         addAccents("\\thaiSaraUu", getutf8(0x0e39));
1876         addAccents("\\thaiPhinthu", getutf8(0x0e3a));
1877         addAccents("\\thaiSaraE", getutf8(0x0e40));
1878         addAccents("\\thaiSaraAe", getutf8(0x0e41));
1879         addAccents("\\thaiSaraO", getutf8(0x0e42));
1880         addAccents("\\thaiSaraAiMaimuan", getutf8(0x0e43));
1881         addAccents("\\thaiSaraAiMaimalai", getutf8(0x0e44));
1882         addAccents("\\thaiLakkhangyao", getutf8(0x0e45));
1883         addAccents("\\thaiMaiyamok", getutf8(0x0e46));
1884         addAccents("\\thaiMaitaikhu", getutf8(0x0e47));
1885         addAccents("\\thaiMaiEk", getutf8(0x0e48));
1886         addAccents("\\thaiMaiTho", getutf8(0x0e49));
1887         addAccents("\\thaiMaiTri", getutf8(0x0e4a));
1888         addAccents("\\thaiMaiChattawa", getutf8(0x0e4b));
1889         addAccents("\\thaiThanthakhat", getutf8(0x0e4c));
1890         addAccents("\\thaiNikhahit", getutf8(0x0e4d));
1891         addAccents("\\thaiYamakkan", getutf8(0x0e4e));
1892         addAccents("\\thaiFongman", getutf8(0x0e4f));
1893         addAccents("\\thaizero", getutf8(0x0e50));
1894         addAccents("\\thaione", getutf8(0x0e51));
1895         addAccents("\\thaitwo", getutf8(0x0e52));
1896         addAccents("\\thaithree", getutf8(0x0e53));
1897         addAccents("\\thaifour", getutf8(0x0e54));
1898         addAccents("\\thaifive", getutf8(0x0e55));
1899         addAccents("\\thaisix", getutf8(0x0e56));
1900         addAccents("\\thaiseven", getutf8(0x0e57));
1901         addAccents("\\thaieight", getutf8(0x0e58));
1902         addAccents("\\thainine", getutf8(0x0e59));
1903         addAccents("\\thaiAngkhankhu", getutf8(0x0e5a));
1904         addAccents("\\thaiKhomut", getutf8(0x0e5b));
1905         addAccents("\\dag", getutf8(0x2020));
1906         addAccents("\\dagger", getutf8(0x2020));
1907         addAccents("\\textdagger", getutf8(0x2020));
1908         addAccents("\\ddag", getutf8(0x2021));
1909         addAccents("\\ddagger", getutf8(0x2021));
1910         addAccents("\\textdaggerdbl", getutf8(0x2021));
1911         addAccents("\\textbullet", getutf8(0x2022));
1912         addAccents("\\bullet", getutf8(0x2022));
1913         addAccents("\\dots", getutf8(0x2026));
1914         addAccents("\\ldots", getutf8(0x2026));
1915         addAccents("\\textellipsis", getutf8(0x2026));
1916         addAccents("\\textasciiacute", getutf8(0x2032));
1917         addAccents("\\prime", getutf8(0x2032));
1918         addAccents("\\textacutedbl", getutf8(0x2033));
1919         addAccents("\\dprime", getutf8(0x2033));
1920         addAccents("\\textasciigrave", getutf8(0x2035));
1921         addAccents("\\backprime", getutf8(0x2035));
1922         addAccents("\\textsubcircum{ }", getutf8(0x2038));
1923         addAccents("\\caretinsert", getutf8(0x2038));
1924         addAccents("\\textasteriskcentered", getutf8(0x204e));
1925         addAccents("\\ast", getutf8(0x204e));
1926         addAccents("\\textmho", getutf8(0x2127));
1927         addAccents("\\mho", getutf8(0x2127));
1928         addAccents("\\textleftarrow", getutf8(0x2190));
1929         addAccents("\\leftarrow", getutf8(0x2190));
1930         addAccents("\\textuparrow", getutf8(0x2191));
1931         addAccents("\\uparrow", getutf8(0x2191));
1932         addAccents("\\textrightarrow", getutf8(0x2192));
1933         addAccents("\\rightarrow", getutf8(0x2192));
1934         addAccents("\\textdownarrow", getutf8(0x2193));
1935         addAccents("\\downarrow", getutf8(0x2193));
1936         addAccents("\\textglobrise", getutf8(0x2197));
1937         addAccents("\\nearrow", getutf8(0x2197));
1938         addAccents("\\textglobfall", getutf8(0x2198));
1939         addAccents("\\searrow", getutf8(0x2198));
1940         addAccents("\\textsurd", getutf8(0x221a));
1941         addAccents("\\surd", getutf8(0x221a));
1942         addAccents("\\textbigcircle", getutf8(0x25ef));
1943         addAccents("\\bigcirc", getutf8(0x25ef));
1944         addAccents("\\FiveStar", getutf8(0x2605));
1945         addAccents("\\bigstar", getutf8(0x2605));
1946         addAccents("\\FiveStarOpen", getutf8(0x2606));
1947         addAccents("\\bigwhitestar", getutf8(0x2606));
1948         addAccents("\\Checkmark", getutf8(0x2713));
1949         addAccents("\\checkmark", getutf8(0x2713));
1950         addAccents("\\CrossMaltese", getutf8(0x2720));
1951         addAccents("\\maltese", getutf8(0x2720));
1952         addAccents("\\textlangle", getutf8(0x27e8));
1953         addAccents("\\langle", getutf8(0x27e8));
1954         addAccents("\\textrangle", getutf8(0x27e9));
1955         addAccents("\\rangle", getutf8(0x27e9));
1956 }
1957
1958 static void buildAccentsMap()
1959 {
1960         accents["imath"] = "ı";
1961         accents["i"] = "ı";
1962         accents["jmath"] = "ȷ";
1963         accents["cdot"] = "·";
1964         accents["textasciicircum"] = "^";
1965         accents["mathcircumflex"] = "^";
1966         accents["guillemotright"] = "»";
1967         accents["guillemotleft"] = "«";
1968         accents["hairspace"]     = getutf8(0xf0000);    // select from free unicode plane 15
1969         accents["thinspace"]     = getutf8(0xf0002);    // and used _only_ by findadv
1970         accents["negthinspace"]  = getutf8(0xf0003);    // to omit backslashed latex macros
1971         accents["medspace"]      = getutf8(0xf0004);    // See https://en.wikipedia.org/wiki/Private_Use_Areas
1972         accents["negmedspace"]   = getutf8(0xf0005);
1973         accents["thickspace"]    = getutf8(0xf0006);
1974         accents["negthickspace"] = getutf8(0xf0007);
1975         accents["lyx"]           = getutf8(0xf0010);    // Used logos
1976         accents["LyX"]           = getutf8(0xf0010);
1977         accents["tex"]           = getutf8(0xf0011);
1978         accents["TeX"]           = getutf8(0xf0011);
1979         accents["latex"]         = getutf8(0xf0012);
1980         accents["LaTeX"]         = getutf8(0xf0012);
1981         accents["latexe"]        = getutf8(0xf0013);
1982         accents["LaTeXe"]        = getutf8(0xf0013);
1983         accents["lyxarrow"]      = getutf8(0xf0020);
1984         accents["braceleft"]     = getutf8(0xf0030);
1985         accents["braceright"]    = getutf8(0xf0031);
1986         accents["lyxtilde"]      = getutf8(0xf0032);
1987         accents["sim"]           = getutf8(0xf0032);
1988         accents["lyxdollar"]     = getutf8(0xf0033);
1989         accents["backslash lyx"]           = getutf8(0xf0010);  // Used logos inserted with starting \backslash
1990         accents["backslash LyX"]           = getutf8(0xf0010);
1991         accents["backslash tex"]           = getutf8(0xf0011);
1992         accents["backslash TeX"]           = getutf8(0xf0011);
1993         accents["backslash latex"]         = getutf8(0xf0012);
1994         accents["backslash LaTeX"]         = getutf8(0xf0012);
1995         accents["backslash latexe"]        = getutf8(0xf0013);
1996         accents["backslash LaTeXe"]        = getutf8(0xf0013);
1997         accents["backslash lyxarrow"]      = getutf8(0xf0020);
1998         accents["ddot{\\imath}"] = "ï";
1999         buildaccent("ddot", "aAeEhHiIoOtuUwWxXyY",
2000                     "äÄëËḧḦïÏöÖẗüÜẅẄẍẌÿŸ");   // umlaut
2001         buildaccent("dot|.", "aAbBcCdDeEfFGghHIimMnNoOpPrRsStTwWxXyYzZ",
2002                     "ȧȦḃḂċĊḋḊėĖḟḞĠġḣḢİİṁṀṅṄȯȮṗṖṙṘṡṠṫṪẇẆẋẊẏẎżŻ");      // dot{i} can only happen if ignoring case, but there is no lowercase of 'İ'
2003         accents["acute{\\imath}"] = "í";
2004         buildaccent("acute", "aAcCeEgGkKlLmMoOnNpPrRsSuUwWyYzZiI",
2005                     "áÁćĆéÉǵǴḱḰĺĹḿḾóÓńŃṕṔŕŔśŚúÚẃẂýÝźŹíÍ");
2006         buildaccent("dacute|H|h", "oOuU", "őŐűŰ");  // double acute
2007         buildaccent("mathring|r", "aAuUwy",
2008                     "åÅůŮẘẙ");  // ring
2009         accents["check{\\imath}"] = "ǐ";
2010         accents["check{\\jmath}"] = "ǰ";
2011         buildaccent("check|v", "cCdDaAeEiIoOuUgGkKhHlLnNrRsSTtzZ",
2012                     "čČďĎǎǍěĚǐǏǒǑǔǓǧǦǩǨȟȞľĽňŇřŘšŠŤťžŽ");        // caron
2013         accents["hat{\\imath}"] = "î";
2014         accents["hat{\\jmath}"] = "ĵ";
2015         buildaccent("hat|^", "aAcCeEgGhHiIjJoOsSuUwWyYzZ",
2016                     "âÂĉĈêÊĝĜĥĤîÎĵĴôÔŝŜûÛŵŴŷŶẑẐ");  // circ
2017         accents["bar{\\imath}"] = "ī";
2018         buildaccent("bar|=", "aAeEiIoOuUyY",
2019                     "āĀēĒīĪōŌūŪȳȲ");        // macron
2020         accents["tilde{\\imath}"] = "ĩ";
2021         buildaccent("tilde", "aAeEiInNoOuUvVyY",
2022                     "ãÃẽẼĩĨñÑõÕũŨṽṼỹỸ");  // tilde
2023         accents["breve{\\imath}"] = "ĭ";
2024         buildaccent("breve|u", "aAeEgGiIoOuU",
2025                     "ăĂĕĔğĞĭĬŏŎŭŬ");        // breve
2026         accents["grave{\\imath}"] = "ì";
2027         buildaccent("grave|`", "aAeEiIoOuUnNwWyY",
2028                     "àÀèÈìÌòÒùÙǹǸẁẀỳỲ");    // grave
2029         buildaccent("subdot|d", "BbDdHhKkLlMmNnRrSsTtVvWwZzAaEeIiOoUuYy",
2030                     "ḄḅḌḍḤḥḲḳḶḷṂṃṆṇṚṛṢṣṬṭṾṿẈẉẒẓẠạẸẹỊịỌọỤụỴỵ");      // dot below
2031         buildaccent("ogonek|k", "AaEeIiUuOo",
2032                     "ĄąĘęĮįŲųǪǫ");    // ogonek
2033         buildaccent("cedilla|c", "CcGgKkLlNnRrSsTtEeDdHh",
2034                     "ÇçĢģĶķĻļŅņŖŗŞşŢţȨȩḐḑḨḩ");        // cedilla
2035         buildaccent("subring|textsubring", "Aa",
2036                     "Ḁḁ");  // subring
2037         buildaccent("subhat|textsubcircum", "DdEeLlNnTtUu",
2038                     "ḒḓḘḙḼḽṊṋṰṱṶṷ");    // subcircum
2039         buildaccent("subtilde|textsubtilde", "EeIiUu",
2040                     "ḚḛḬḭṴṵ");      // subtilde
2041         accents["dgrave{\\imath}"] = "ȉ";
2042         accents["textdoublegrave{\\i}"] = "ȉ";
2043         buildaccent("dgrave|textdoublegrave", "AaEeIiOoRrUu",
2044                     "ȀȁȄȅȈȉȌȍȐȑȔȕ"); // double grave
2045         accents["rcap{\\imath}"] = "ȋ";
2046         accents["textroundcap{\\i}"] = "ȋ";
2047         buildaccent("rcap|textroundcap", "AaEeIiOoRrUu",
2048                     "ȂȃȆȇȊȋȎȏȒȓȖȗ"); // inverted breve
2049         buildaccent("slashed", "oO",
2050                     "øØ"); // slashed
2051         fillMissingUnicodesymbols(); // Add some still not handled entries contained in 'unicodesynbols'
2052         // LYXERR0("Number of accents " << accents.size());
2053 }
2054
2055 /*
2056  * Created accents in math or regexp environment
2057  * are macros, but we need the utf8 equivalent
2058  */
2059 void Intervall::removeAccents()
2060 {
2061         if (accents.empty())
2062                 buildAccentsMap();
2063         static regex const accre("\\\\("
2064                                  "([\\S]|[A-Za-z]+)\\{[^\\\\\\{\\}]+\\}"
2065                                  "|([\\S]|[A-Za-z]+)\\{\\\\[ij](math)?\\}"
2066                                  "|("
2067                                  "(backslash ([lL]y[xX]|[tT]e[xX]|[lL]a[tT]e[xX]e?|lyxarrow))"
2068                                  "|[A-Za-z]+"
2069                                  ")"
2070                                  "(?![a-zA-Z]))");
2071         smatch sub;
2072         for (sregex_iterator itacc(par.begin(), par.end(), accre), end; itacc != end; ++itacc) {
2073                 sub = *itacc;
2074                 string key = sub.str(1);
2075                 AccentsIterator it_ac = accents.find(key);
2076                 if (it_ac != accents.end()) {
2077                         string val = it_ac->second;
2078                         size_t pos = sub.position(size_t(0));
2079                         for (size_t i = 0; i < val.size(); i++) {
2080                                 par[pos+i] = val[i];
2081                         }
2082                         // Remove possibly following space too
2083                         if (par[pos+sub.str(0).size()] == ' ')
2084                                 addIntervall(pos+val.size(), pos + sub.str(0).size()+1);
2085                         else
2086                                 addIntervall(pos+val.size(), pos + sub.str(0).size());
2087                         for (size_t i = pos+val.size(); i < pos + sub.str(0).size(); i++) {
2088                                 // remove traces of any remaining chars
2089                                 par[i] = ' ';
2090                         }
2091                 }
2092                 else {
2093                         LYXERR(Debug::INFO, "Not added accent for \"" << key << "\"");
2094                 }
2095         }
2096 }
2097
2098 void Intervall::handleOpenP(int i)
2099 {
2100         actualdeptindex++;
2101         if ((size_t) actualdeptindex >= depts.size()) {
2102                 depts.resize(actualdeptindex + 30);
2103                 closes.resize(actualdeptindex + 30);
2104         }
2105         depts[actualdeptindex] = i+1;
2106         closes[actualdeptindex] = -1;
2107         // checkDepthIndex(actualdeptindex);
2108 }
2109
2110 void Intervall::handleCloseP(int i, bool closingAllowed)
2111 {
2112         if (actualdeptindex <= 0) {
2113                 if (! closingAllowed)
2114                         LYXERR(Debug::FINDVERBOSE, "Bad closing parenthesis in latex");  /* should not happen, but the latex input may be wrong */
2115                 // if we are at the very end
2116                 addIntervall(i, i+1);
2117         }
2118         else {
2119                 closes[actualdeptindex] = i+1;
2120                 actualdeptindex--;
2121         }
2122 }
2123
2124 void Intervall::resetOpenedP(int openPos)
2125 {
2126         // Used as initializer for foreignlanguage entry
2127         actualdeptindex = 1;
2128         depts[1] = openPos+1;
2129         closes[1] = -1;
2130 }
2131
2132 int Intervall::previousNotIgnored(int start) const
2133 {
2134         int idx = 0;                          /* int intervalls */
2135         for (idx = ignoreidx; idx >= 0; --idx) {
2136                 if (start > borders[idx].upper)
2137                         return start;
2138                 if (start >= borders[idx].low)
2139                         start = borders[idx].low-1;
2140         }
2141         return start;
2142 }
2143
2144 int Intervall::nextNotIgnored(int start) const
2145 {
2146         int idx = 0;                          /* int intervalls */
2147         for (idx = 0; idx <= ignoreidx; idx++) {
2148                 if (start < borders[idx].low)
2149                         return start;
2150                 if (start < borders[idx].upper)
2151                         start = borders[idx].upper;
2152         }
2153         return start;
2154 }
2155
2156 typedef unordered_map<string, KeyInfo> KeysMap;
2157 typedef unordered_map<string, KeyInfo>::const_iterator KeysIterator;
2158 typedef vector< KeyInfo> Entries;
2159 static KeysMap keys = unordered_map<string, KeyInfo>();
2160
2161 class LatexInfo {
2162 private:
2163         int entidx_;
2164         Entries entries_;
2165         Intervall interval_;
2166         void buildKeys(bool);
2167         void buildEntries(bool);
2168         void makeKey(const string &, KeyInfo, bool isPatternString);
2169         void processRegion(int start, int region_end); /*  remove {} parts */
2170         void removeHead(KeyInfo const &, int count=0);
2171
2172 public:
2173         LatexInfo(string const & par, bool isPatternString)
2174                 : entidx_(-1), interval_(isPatternString, par)
2175         {
2176                 buildKeys(isPatternString);
2177                 entries_ = vector<KeyInfo>();
2178                 buildEntries(isPatternString);
2179         }
2180         int getFirstKey() {
2181                 entidx_ = 0;
2182                 if (entries_.empty()) {
2183                         return -1;
2184                 }
2185                 if (entries_[0].keytype == KeyInfo::isTitle) {
2186                         interval_.hasTitle = true;
2187                         if (! entries_[0].disabled) {
2188                                 interval_.titleValue = entries_[0].head;
2189                         }
2190                         else {
2191                                 interval_.titleValue = "";
2192                         }
2193                         removeHead(entries_[0]);
2194                         if (entries_.size() > 1)
2195                                 return 1;
2196                         else
2197                                 return -1;
2198                 }
2199                 return 0;
2200         }
2201         int getNextKey() {
2202                 entidx_++;
2203                 if (int(entries_.size()) > entidx_) {
2204                         return entidx_;
2205                 }
2206                 else {
2207                         return -1;
2208                 }
2209         }
2210         bool setNextKey(int idx) {
2211                 if ((idx == entidx_) && (entidx_ >= 0)) {
2212                         entidx_--;
2213                         return true;
2214                 }
2215                 else
2216                         return false;
2217         }
2218         int find(int start, KeyInfo::KeyType keytype) const {
2219                 if (start < 0)
2220                         return -1;
2221                 int tmpIdx = start;
2222                 while (tmpIdx < int(entries_.size())) {
2223                         if (entries_[tmpIdx].keytype == keytype)
2224                                 return tmpIdx;
2225                         tmpIdx++;
2226                 }
2227                 return -1;
2228         }
2229         int process(ostringstream & os, KeyInfo const & actual);
2230         int dispatch(ostringstream & os, int previousStart, KeyInfo & actual);
2231         // string show(int lastpos) { return interval.show(lastpos);}
2232         int nextNotIgnored(int start) { return interval_.nextNotIgnored(start);}
2233         KeyInfo &getKeyInfo(int keyinfo) {
2234                 static KeyInfo invalidInfo = KeyInfo();
2235                 if ((keyinfo < 0) || ( keyinfo >= int(entries_.size())))
2236                         return invalidInfo;
2237                 else
2238                         return entries_[keyinfo];
2239         }
2240         void setForDefaultLang(KeyInfo const & defLang) {interval_.setForDefaultLang(defLang);}
2241         void addIntervall(int low, int up) { interval_.addIntervall(low, up); }
2242 };
2243
2244
2245 int Intervall::findclosing(int start, int end, char up = '{', char down = '}', int repeat = 1)
2246 {
2247         int skip = 0;
2248         int depth = 0;
2249         for (int i = start; i < end; i += 1 + skip) {
2250                 char c;
2251                 c = par[i];
2252                 skip = 0;
2253                 if (c == '\\') skip = 1;
2254                 else if (c == up) {
2255                         depth++;
2256                 }
2257                 else if (c == down) {
2258                         if (depth == 0) {
2259                                 repeat--;
2260                                 if ((repeat <= 0) || (par[i+1] != up))
2261                                         return i;
2262                         }
2263                         --depth;
2264                 }
2265         }
2266         return end;
2267 }
2268
2269 void Intervall::removeInvalidClosings(void)
2270 {
2271         // this can happen, if there are deleted parts
2272         int skip = 0;
2273         int depth = 0;
2274         for (unsigned i = 0; i < par.size(); i += 1 + skip) {
2275                 char c = par[i];
2276                 skip = 0;
2277                 if (c == '\\') skip = 1;
2278                 else if (c == '{')
2279                         depth++;
2280                 else if (c == '}') {
2281                         if (depth == 0) {
2282                                 addIntervall(i, i+1);
2283                                 LYXERR(Debug::FINDVERBOSE, "removed invalid closing '}' at " << i);
2284                         }
2285                         else
2286                                 --depth;
2287                 }
2288         }
2289 }
2290 class MathInfo {
2291         class MathEntry {
2292         public:
2293                 string wait;
2294                 size_t mathEnd;
2295                 size_t mathpostfixsize;
2296                 size_t mathStart;
2297                 size_t mathprefixsize;
2298                 size_t mathSize;
2299         };
2300         size_t actualIdx_;
2301         vector<MathEntry> entries_;
2302 public:
2303         MathInfo() {
2304                 actualIdx_ = 0;
2305         }
2306         void insert(string const & wait, size_t start, size_t prefixsize, size_t end, size_t postfixsize) {
2307                 MathEntry m = MathEntry();
2308                 m.wait = wait;
2309                 m.mathStart = start;
2310                 m.mathprefixsize = prefixsize;
2311                 m.mathEnd = end + postfixsize;
2312                 m.mathpostfixsize = postfixsize;
2313                 m.mathSize = m.mathEnd - m.mathStart;
2314                 entries_.push_back(m);
2315         }
2316         bool empty() const { return entries_.empty(); }
2317         size_t getEndPos() const {
2318                 if (entries_.empty() || (actualIdx_ >= entries_.size())) {
2319                         return 0;
2320                 }
2321                 return entries_[actualIdx_].mathEnd;
2322         }
2323         size_t getStartPos() const {
2324                 if (entries_.empty() || (actualIdx_ >= entries_.size())) {
2325                         return 100000;                    /*  definitely enough? */
2326                 }
2327                 return entries_[actualIdx_].mathStart;
2328         }
2329         size_t getPrefixSize() const {
2330                 if (entries_.empty() || (actualIdx_ >= entries_.size())) {
2331                         return 0;
2332                 }
2333                 return entries_[actualIdx_].mathprefixsize;
2334         }
2335         size_t getPostfixSize() const {
2336                 if (entries_.empty() || (actualIdx_ >= entries_.size())) {
2337                         return 0;
2338                 }
2339                 return entries_[actualIdx_].mathpostfixsize;
2340         }
2341         size_t getFirstPos() {
2342                 actualIdx_ = 0;
2343                 return getStartPos();
2344         }
2345         size_t getSize() const {
2346                 if (entries_.empty() || (actualIdx_ >= entries_.size())) {
2347                         return size_t(0);
2348                 }
2349                 return entries_[actualIdx_].mathSize;
2350         }
2351         void incrEntry() { actualIdx_++; }
2352 };
2353
2354 void LatexInfo::buildEntries(bool isPatternString)
2355 {
2356         static regex const rmath("(\\\\)*(\\$|\\\\\\[|\\\\\\]|\\\\(begin|end)\\{((eqnarray|equation|flalign|gather|multline|align|x?x?alignat)\\*?\\})(\\{[0-9]+\\})?)");
2357         static regex const rkeys("(\\\\)*(\\$|\\\\\\[|\\\\\\]|\\\\((([a-zA-Z]+\\*?)(\\{([a-z]+\\*?)\\}|=[0-9]+[a-z]+)?)))");
2358         static bool disableLanguageOverride = false;
2359         smatch sub, submath;
2360         bool evaluatingRegexp = false;
2361         MathInfo mi;
2362         bool evaluatingMath = false;
2363         bool evaluatingCode = false;
2364         size_t codeEnd = 0;
2365         bool evaluatingOptional = false;
2366         size_t optionalEnd = 0;
2367         int codeStart = -1;
2368         KeyInfo found;
2369         bool math_end_waiting = false;
2370         size_t math_pos = 10000;
2371         size_t math_prefix_size = 1;
2372         string math_end;
2373         static vector<string> usedText = vector<string>();
2374         static bool removeMathHull = false;
2375
2376         interval_.removeAccents();
2377         interval_.removeInvalidClosings();
2378
2379         for (sregex_iterator itmath(interval_.par.begin(), interval_.par.end(), rmath), end; itmath != end; ++itmath) {
2380                 submath = *itmath;
2381                 if ((submath.position(2) - submath.position(0)) %2 == 1) {
2382                         // prefixed by odd count of '\\'
2383                         continue;
2384                 }
2385                 if (math_end_waiting) {
2386                         size_t pos = submath.position(size_t(2));
2387                         if ((math_end == "$") &&
2388                                         (submath.str(2) == "$")) {
2389                                 mi.insert("$", math_pos, 1, pos, 1);
2390                                 math_end_waiting = false;
2391                         }
2392                         else if ((math_end == "\\]") &&
2393                                  (submath.str(2) == "\\]")) {
2394                                 mi.insert("\\]", math_pos, 2, pos, 2);
2395                                 math_end_waiting = false;
2396                         }
2397                         else if ((submath.str(3).compare("end") == 0) &&
2398                                  (submath.str(5).compare(math_end) == 0)) {
2399                                 mi.insert(math_end, math_pos, math_prefix_size, pos, submath.str(2).length());
2400                                 math_end_waiting = false;
2401                         }
2402                         else
2403                                 continue;
2404                 }
2405                 else {
2406                         if (submath.str(3).compare("begin") == 0) {
2407                                 math_end_waiting = true;
2408                                 math_end = submath.str(5);
2409                                 math_pos = submath.position(size_t(2));
2410                                 math_prefix_size = submath.str(2).length();
2411                         }
2412                         else if (submath.str(2).compare("\\[") == 0) {
2413                                 math_end_waiting = true;
2414                                 math_end = "\\]";
2415                                 math_pos = submath.position(size_t(2));
2416                         }
2417                         else if (submath.str(2) == "$") {
2418                                 size_t pos = submath.position(size_t(2));
2419                                 math_end_waiting = true;
2420                                 math_end = "$";
2421                                 math_pos = pos;
2422                         }
2423                 }
2424         }
2425         // Ignore language if there is math somewhere in pattern-string
2426         if (isPatternString) {
2427                 for (auto s: usedText) {
2428                         // Remove entries created in previous search runs
2429                         keys.erase(s);
2430                 }
2431                 usedText = vector<string>();
2432                 if (! mi.empty()) {
2433                         // Disable language
2434                         keys["foreignlanguage"].disabled = true;
2435                         disableLanguageOverride = true;
2436                         removeMathHull = false;
2437                 }
2438                 else {
2439                         removeMathHull = true;  // used later if not isPatternString
2440                         disableLanguageOverride = false;
2441                 }
2442         }
2443         else {
2444                 if (disableLanguageOverride) {
2445                         keys["foreignlanguage"].disabled = true;
2446                 }
2447         }
2448         math_pos = mi.getFirstPos();
2449         for (sregex_iterator it(interval_.par.begin(), interval_.par.end(), rkeys), end; it != end; ++it) {
2450                 sub = *it;
2451                 if ((sub.position(2) - sub.position(0)) %2 == 1) {
2452                         // prefixed by odd count of '\\'
2453                         continue;
2454                 }
2455                 string key = sub.str(5);
2456                 if (key == "") {
2457                         if (sub.str(2)[0] == '\\')
2458                                 key = sub.str(2)[1];
2459                         else {
2460                                 key = sub.str(2);
2461                         }
2462                 }
2463                 KeysIterator it_key = keys.find(key);
2464                 if (it_key != keys.end()) {
2465                         if (it_key->second.keytype == KeyInfo::headRemove) {
2466                                 KeyInfo found1 = it_key->second;
2467                                 found1.disabled = true;
2468                                 found1.head = "\\" + key + "{";
2469                                 found1._tokenstart = sub.position(size_t(2));
2470                                 found1._tokensize = found1.head.length();
2471                                 found1._dataStart = found1._tokenstart + found1.head.length();
2472                                 int endpos = interval_.findclosing(found1._dataStart, interval_.par.length(), '{', '}', 1);
2473                                 found1._dataEnd = endpos;
2474                                 removeHead(found1);
2475                                 continue;
2476                         }
2477                 }
2478                 if (evaluatingRegexp) {
2479                         if (sub.str(3).compare("endregexp") == 0) {
2480                                 evaluatingRegexp = false;
2481                                 // found._tokenstart already set
2482                                 found._dataEnd = sub.position(size_t(2)) + 13;
2483                                 found._dataStart = found._dataEnd;
2484                                 found._tokensize = found._dataEnd - found._tokenstart;
2485                                 found.parenthesiscount = 0;
2486                                 found.head = interval_.par.substr(found._tokenstart, found._tokensize);
2487                         }
2488                         else {
2489                                 continue;
2490                         }
2491                 }
2492                 else {
2493                         if (evaluatingMath) {
2494                                 if (size_t(sub.position(size_t(2))) < mi.getEndPos())
2495                                         continue;
2496                                 evaluatingMath = false;
2497                                 mi.incrEntry();
2498                                 math_pos = mi.getStartPos();
2499                         }
2500                         if (it_key == keys.end()) {
2501                                 found = KeyInfo(KeyInfo::isStandard, 0, true);
2502                                 LYXERR(Debug::INFO, "Undefined key " << key << " ==> will be used as text");
2503                                 found = KeyInfo(KeyInfo::isText, 0, false);
2504                                 if (isPatternString) {
2505                                         found.keytype = KeyInfo::isChar;
2506                                         found.disabled = false;
2507                                         found.used = true;
2508                                 }
2509                                 keys[key] = found;
2510                                 usedText.push_back(key);
2511                         }
2512                         else
2513                                 found = keys[key];
2514                         if (key.compare("regexp") == 0) {
2515                                 evaluatingRegexp = true;
2516                                 found._tokenstart = sub.position(size_t(2));
2517                                 found._tokensize = 0;
2518                                 continue;
2519                         }
2520                 }
2521                 // Handle the other params of key
2522                 if (found.keytype == KeyInfo::isIgnored)
2523                         continue;
2524                 else if (found.keytype == KeyInfo::isMath) {
2525                         if (size_t(sub.position(size_t(2))) == math_pos) {
2526                                 found = keys[key];
2527                                 found._tokenstart = sub.position(size_t(2));
2528                                 found._tokensize = mi.getSize();
2529                                 found._dataEnd = found._tokenstart + found._tokensize;
2530                                 found._dataStart = found._dataEnd;
2531                                 found.parenthesiscount = 0;
2532                                 found.head = interval_.par.substr(found._tokenstart, found._tokensize);
2533                                 if (removeMathHull) {
2534                                         interval_.addIntervall(found._tokenstart, found._tokenstart + mi.getPrefixSize());
2535                                         interval_.addIntervall(found._dataEnd - mi.getPostfixSize(), found._dataEnd);
2536                                 }
2537                                 else {
2538                                         // Treate all math constructs as simple math
2539                                         interval_.par[found._tokenstart] = '$';
2540                                         interval_.par[found._dataEnd - mi.getPostfixSize()] = '$';
2541                                         interval_.addIntervall(found._tokenstart + 1, found._tokenstart + mi.getPrefixSize());
2542                                         interval_.addIntervall(found._dataEnd - mi.getPostfixSize() + 1, found._dataEnd);
2543                                 }
2544                                 evaluatingMath = true;
2545                         }
2546                         else {
2547                                 // begin|end of unknown env, discard
2548                                 // First handle tables
2549                                 // longtable|tabular
2550                                 bool discardComment;
2551                                 found = keys[key];
2552                                 found.keytype = KeyInfo::doRemove;
2553                                 if ((sub.str(7).compare("longtable") == 0) ||
2554                                                 (sub.str(7).compare("tabular") == 0)) {
2555                                         discardComment = true;        /* '%' */
2556                                 }
2557                                 else {
2558                                         discardComment = false;
2559                                         static regex const removeArgs("^(multicols|multipar|sectionbox|subsectionbox|tcolorbox)$");
2560                                         smatch sub2;
2561                                         string token = sub.str(7);
2562                                         if (regex_match(token, sub2, removeArgs)) {
2563                                                 found.keytype = KeyInfo::removeWithArg;
2564                                         }
2565                                 }
2566                                 // discard spaces before pos(2)
2567                                 int pos = sub.position(size_t(2));
2568                                 int count;
2569                                 for (count = 0; pos - count > 0; count++) {
2570                                         char c = interval_.par[pos-count-1];
2571                                         if (discardComment) {
2572                                                 if ((c != ' ') && (c != '%'))
2573                                                         break;
2574                                         }
2575                                         else if (c != ' ')
2576                                                 break;
2577                                 }
2578                                 found._tokenstart = pos - count;
2579                                 if (sub.str(3).compare(0, 5, "begin") == 0) {
2580                                         size_t pos1 = pos + sub.str(2).length();
2581                                         if (sub.str(7).compare("cjk") == 0) {
2582                                                 pos1 = interval_.findclosing(pos1+1, interval_.par.length()) + 1;
2583                                                 if ((interval_.par[pos1] == '{') && (interval_.par[pos1+1] == '}'))
2584                                                         pos1 += 2;
2585                                                 found.keytype = KeyInfo::isMain;
2586                                                 found._dataStart = pos1;
2587                                                 found._dataEnd = interval_.par.length();
2588                                                 found.disabled = keys["foreignlanguage"].disabled;
2589                                                 found.used = keys["foreignlanguage"].used;
2590                                                 found._tokensize = pos1 - found._tokenstart;
2591                                                 found.head = interval_.par.substr(found._tokenstart, found._tokensize);
2592                                         }
2593                                         else {
2594                                                 // Swallow possible optional params
2595                                                 while (interval_.par[pos1] == '[') {
2596                                                         pos1 = interval_.findclosing(pos1+1, interval_.par.length(), '[', ']')+1;
2597                                                 }
2598                                                 // Swallow also the eventual parameter
2599                                                 if (interval_.par[pos1] == '{') {
2600                                                         found._dataEnd = interval_.findclosing(pos1+1, interval_.par.length()) + 1;
2601                                                 }
2602                                                 else {
2603                                                         found._dataEnd = pos1;
2604                                                 }
2605                                                 found._dataStart = found._dataEnd;
2606                                                 found._tokensize = count + found._dataEnd - pos;
2607                                                 found.parenthesiscount = 0;
2608                                                 found.head = interval_.par.substr(found._tokenstart, found._tokensize);
2609                                                 found.disabled = true;
2610                                         }
2611                                 }
2612                                 else {
2613                                         // Handle "\end{...}"
2614                                         found._dataStart = pos + sub.str(2).length();
2615                                         found._dataEnd = found._dataStart;
2616                                         found._tokensize = count + found._dataEnd - pos;
2617                                         found.parenthesiscount = 0;
2618                                         found.head = interval_.par.substr(found._tokenstart, found._tokensize);
2619                                         found.disabled = true;
2620                                 }
2621                         }
2622                 }
2623                 else if (found.keytype != KeyInfo::isRegex) {
2624                         found._tokenstart = sub.position(size_t(2));
2625                         if (found.parenthesiscount == 0) {
2626                                 // Probably to be discarded
2627                                 size_t following_pos = sub.position(size_t(2)) + sub.str(5).length() + 1;
2628                                 char following = interval_.par[following_pos];
2629                                 if (following == ' ')
2630                                         found.head = "\\" + sub.str(5) + " ";
2631                                 else if (following == '=') {
2632                                         // like \uldepth=1000pt
2633                                         found.head = sub.str(2);
2634                                 }
2635                                 else
2636                                         found.head = "\\" + key;
2637                                 found._tokensize = found.head.length();
2638                                 found._dataEnd = found._tokenstart + found._tokensize;
2639                                 found._dataStart = found._dataEnd;
2640                         }
2641                         else {
2642                                 int params = found._tokenstart + key.length() + 1;
2643                                 if (evaluatingOptional) {
2644                                         if (size_t(found._tokenstart) > optionalEnd) {
2645                                                 evaluatingOptional = false;
2646                                         }
2647                                         else {
2648                                                 found.disabled = true;
2649                                         }
2650                                 }
2651                                 int optend = params;
2652                                 while (interval_.par[optend] == '[') {
2653                                         // discard optional parameters
2654                                         optend = interval_.findclosing(optend+1, interval_.par.length(), '[', ']') + 1;
2655                                 }
2656                                 if (optend > params) {
2657                                         key += interval_.par.substr(params, optend-params);
2658                                         evaluatingOptional = true;
2659                                         optionalEnd = optend;
2660                                         if (found.keytype == KeyInfo::isSectioning) {
2661                                                 // Remove optional values (but still keep in header)
2662                                                 interval_.addIntervall(params, optend);
2663                                         }
2664                                 }
2665                                 string token = sub.str(7);
2666                                 int closings;
2667                                 if (interval_.par[optend] != '{') {
2668                                         closings = 0;
2669                                         found.parenthesiscount = 0;
2670                                         found.head = "\\" + key;
2671                                 }
2672                                 else
2673                                         closings = found.parenthesiscount;
2674                                 if (found.parenthesiscount == 1) {
2675                                         found.head = "\\" + key + "{";
2676                                 }
2677                                 else if (found.parenthesiscount > 1) {
2678                                         if (token != "") {
2679                                                 found.head = sub.str(2) + "{";
2680                                                 closings = found.parenthesiscount - 1;
2681                                         }
2682                                         else {
2683                                                 found.head = "\\" + key + "{";
2684                                         }
2685                                 }
2686                                 found._tokensize = found.head.length();
2687                                 found._dataStart = found._tokenstart + found.head.length();
2688                                 if (found.keytype == KeyInfo::doRemove) {
2689                                         if (closings > 0) {
2690                                                 size_t endpar = 2 + interval_.findclosing(found._dataStart, interval_.par.length(), '{', '}', closings);
2691                                                 if (endpar >= interval_.par.length())
2692                                                         found._dataStart = interval_.par.length();
2693                                                 else
2694                                                         found._dataStart = endpar;
2695                                                 found._tokensize = found._dataStart - found._tokenstart;
2696                                         }
2697                                         else {
2698                                                 found._dataStart = found._tokenstart + found._tokensize;
2699                                         }
2700                                         closings = 0;
2701                                 }
2702                                 if (interval_.par.substr(found._dataStart, 15).compare("\\endarguments{}") == 0) {
2703                                         found._dataStart += 15;
2704                                 }
2705                                 size_t endpos;
2706                                 if (closings < 1)
2707                                         endpos = found._dataStart - 1;
2708                                 else
2709                                         endpos = interval_.findclosing(found._dataStart, interval_.par.length(), '{', '}', closings);
2710                                 if (found.keytype == KeyInfo::isList) {
2711                                         // Check if it really is list env
2712                                         static regex const listre("^([a-z]+)$");
2713                                         smatch sub2;
2714                                         if (!regex_match(token, sub2, listre)) {
2715                                                 // Change the key of this entry. It is not in a list/item environment
2716                                                 found.keytype = KeyInfo::endArguments;
2717                                         }
2718                                 }
2719                                 if (found.keytype == KeyInfo::noMain) {
2720                                         evaluatingCode = true;
2721                                         codeEnd = endpos;
2722                                         codeStart = found._dataStart;
2723                                 }
2724                                 else if (evaluatingCode) {
2725                                         if (size_t(found._dataStart) > codeEnd)
2726                                                 evaluatingCode = false;
2727                                         else if (found.keytype == KeyInfo::isMain) {
2728                                                 // Disable this key, treate it as standard
2729                                                 found.keytype = KeyInfo::isStandard;
2730                                                 found.disabled = true;
2731                                                 if ((codeEnd +1 >= interval_.par.length()) &&
2732                                                                 (found._tokenstart == codeStart)) {
2733                                                         // trickery, because the code inset starts
2734                                                         // with \selectlanguage ...
2735                                                         codeEnd = endpos;
2736                                                         if (entries_.size() > 1) {
2737                                                                 entries_[entries_.size()-1]._dataEnd = codeEnd;
2738                                                         }
2739                                                 }
2740                                         }
2741                                 }
2742                                 if ((endpos == interval_.par.length()) &&
2743                                                 (found.keytype == KeyInfo::doRemove)) {
2744                                         // Missing closing => error in latex-input?
2745                                         // therefore do not delete remaining data
2746                                         found._dataStart -= 1;
2747                                         found._dataEnd = found._dataStart;
2748                                 }
2749                                 else
2750                                         found._dataEnd = endpos;
2751                         }
2752                         if (isPatternString) {
2753                                 keys[key].used = true;
2754                         }
2755                 }
2756                 entries_.push_back(found);
2757         }
2758 }
2759
2760 void LatexInfo::makeKey(const string &keysstring, KeyInfo keyI, bool isPatternString)
2761 {
2762         stringstream s(keysstring);
2763         string key;
2764         const char delim = '|';
2765         while (getline(s, key, delim)) {
2766                 KeyInfo keyII(keyI);
2767                 if (isPatternString) {
2768                         keyII.used = false;
2769                 }
2770                 else if ( !keys[key].used)
2771                         keyII.disabled = true;
2772                 keys[key] = keyII;
2773         }
2774 }
2775
2776 void LatexInfo::buildKeys(bool isPatternString)
2777 {
2778
2779         static bool keysBuilt = false;
2780         if (keysBuilt && !isPatternString) return;
2781
2782         // Keys to ignore in any case
2783         makeKey("text|lyxmathsym|ensuremath", KeyInfo(KeyInfo::headRemove, 1, true), true);
2784         makeKey("nonumber|notag", KeyInfo(KeyInfo::headRemove, 0, true), true);
2785         // Known standard keys with 1 parameter.
2786         // Split is done, if not at start of region
2787         makeKey("textsf|textss|texttt", KeyInfo(KeyInfo::isStandard, 1, ignoreFormats.getFamily()), isPatternString);
2788         makeKey("textbf",               KeyInfo(KeyInfo::isStandard, 1, ignoreFormats.getSeries()), isPatternString);
2789         makeKey("textit|textsc|textsl", KeyInfo(KeyInfo::isStandard, 1, ignoreFormats.getShape()), isPatternString);
2790         makeKey("uuline|uline|uwave",   KeyInfo(KeyInfo::isStandard, 1, ignoreFormats.getUnderline()), isPatternString);
2791         makeKey("emph|noun",            KeyInfo(KeyInfo::isStandard, 1, ignoreFormats.getMarkUp()), isPatternString);
2792         makeKey("sout|xout",            KeyInfo(KeyInfo::isStandard, 1, ignoreFormats.getStrikeOut()), isPatternString);
2793
2794         makeKey("section|subsection|subsubsection|paragraph|subparagraph|minisec",
2795                 KeyInfo(KeyInfo::isSectioning, 1, ignoreFormats.getSectioning()), isPatternString);
2796         makeKey("section*|subsection*|subsubsection*|paragraph*",
2797                 KeyInfo(KeyInfo::isSectioning, 1, ignoreFormats.getSectioning()), isPatternString);
2798         makeKey("part|part*|chapter|chapter*", KeyInfo(KeyInfo::isSectioning, 1, ignoreFormats.getSectioning()), isPatternString);
2799         makeKey("title|subtitle|author|subject|publishers|dedication|uppertitleback|lowertitleback|extratitle|lyxaddress|lyxrightaddress", KeyInfo(KeyInfo::isTitle, 1, ignoreFormats.getFrontMatter()), isPatternString);
2800         // Regex
2801         makeKey("regexp", KeyInfo(KeyInfo::isRegex, 1, false), isPatternString);
2802
2803         // Split is done, if not at start of region
2804         makeKey("textcolor", KeyInfo(KeyInfo::isStandard, 2, ignoreFormats.getColor()), isPatternString);
2805         makeKey("latexenvironment", KeyInfo(KeyInfo::isStandard, 2, false), isPatternString);
2806
2807         // Split is done always.
2808         makeKey("foreignlanguage", KeyInfo(KeyInfo::isMain, 2, ignoreFormats.getLanguage()), isPatternString);
2809
2810         // Known charaters
2811         // No split
2812         makeKey("backslash|textbackslash|slash",  KeyInfo(KeyInfo::isChar, 0, false), isPatternString);
2813         makeKey("textasciicircum|textasciitilde", KeyInfo(KeyInfo::isChar, 0, false), isPatternString);
2814         makeKey("textasciiacute|texemdash",       KeyInfo(KeyInfo::isChar, 0, false), isPatternString);
2815         makeKey("dots|ldots",                     KeyInfo(KeyInfo::isChar, 0, false), isPatternString);
2816         // Spaces
2817         makeKey("quad|qquad|hfill|dotfill",               KeyInfo(KeyInfo::isChar, 0, false), isPatternString);
2818         makeKey("textvisiblespace|nobreakspace",          KeyInfo(KeyInfo::isChar, 0, false), isPatternString);
2819         makeKey("negthickspace|negmedspace|negthinspace", KeyInfo(KeyInfo::isChar, 0, false), isPatternString);
2820         makeKey("thickspace|medspace|thinspace",          KeyInfo(KeyInfo::isChar, 0, false), isPatternString);
2821         // Skip
2822         // makeKey("enskip|smallskip|medskip|bigskip|vfill", KeyInfo(KeyInfo::isChar, 0, false), isPatternString);
2823         // Custom space/skip, remove the content (== length value)
2824         makeKey("vspace|vspace*|hspace|hspace*|mspace", KeyInfo(KeyInfo::noContent, 1, false), isPatternString);
2825         // Found in fr/UserGuide.lyx
2826         makeKey("og|fg", KeyInfo(KeyInfo::isChar, 0, false), isPatternString);
2827         // quotes
2828         makeKey("textquotedbl|quotesinglbase|lyxarrow", KeyInfo(KeyInfo::isChar, 0, false), isPatternString);
2829         makeKey("textquotedblleft|textquotedblright", KeyInfo(KeyInfo::isChar, 0, false), isPatternString);
2830         // Known macros to remove (including their parameter)
2831         // No split
2832         makeKey("input|inputencoding|label|ref|index|bibitem", KeyInfo(KeyInfo::doRemove, 1, false), isPatternString);
2833         makeKey("addtocounter|setlength",                 KeyInfo(KeyInfo::noContent, 2, true), isPatternString);
2834         // handle like standard keys with 1 parameter.
2835         makeKey("url|href|vref|thanks", KeyInfo(KeyInfo::isStandard, 1, false), isPatternString);
2836
2837         if (ignoreFormats.getDeleted()) {
2838                 // Ignore deleted text
2839                 makeKey("lyxdeleted", KeyInfo(KeyInfo::doRemove, 3, false), isPatternString);
2840         }
2841         else {
2842                 // but preserve added text
2843                 makeKey("lyxdeleted", KeyInfo(KeyInfo::doRemove, 2, false), isPatternString);
2844         }
2845         makeKey("lyxadded", KeyInfo(KeyInfo::doRemove, 2, false), isPatternString);
2846
2847         // Macros to remove, but let the parameter survive
2848         // No split
2849         makeKey("menuitem|textmd|textrm", KeyInfo(KeyInfo::isStandard, 1, true), isPatternString);
2850
2851         // Remove language spec from content of these insets
2852         makeKey("code", KeyInfo(KeyInfo::noMain, 1, false), isPatternString);
2853
2854         // Same effect as previous, parameter will survive (because there is no one anyway)
2855         // No split
2856         makeKey("noindent|textcompwordmark|maketitle", KeyInfo(KeyInfo::isStandard, 0, true), isPatternString);
2857         // Remove table decorations
2858         makeKey("hline|tabularnewline|toprule|bottomrule|midrule", KeyInfo(KeyInfo::doRemove, 0, true), isPatternString);
2859         // Discard shape-header.
2860         // For footnote or shortcut too, because of lang settings
2861         // and wrong handling if used 'KeyInfo::noMain'
2862         makeKey("circlepar|diamondpar|heartpar|nutpar",  KeyInfo(KeyInfo::isStandard, 1, true), isPatternString);
2863         makeKey("trianglerightpar|hexagonpar|starpar",   KeyInfo(KeyInfo::isStandard, 1, true), isPatternString);
2864         makeKey("triangleuppar|triangledownpar|droppar", KeyInfo(KeyInfo::isStandard, 1, true), isPatternString);
2865         makeKey("triangleleftpar|shapepar|dropuppar",    KeyInfo(KeyInfo::isStandard, 1, true), isPatternString);
2866         makeKey("hphantom|vphantom|note|footnote|shortcut|include|includegraphics",     KeyInfo(KeyInfo::isStandard, 1, true), isPatternString);
2867         makeKey("textgreek|textcyrillic", KeyInfo(KeyInfo::isStandard, 1, true), false);
2868         makeKey("parbox", KeyInfo(KeyInfo::doRemove, 1, true), isPatternString);
2869         // like ('tiny{}' or '\tiny ' ... )
2870         makeKey("footnotesize|tiny|scriptsize|small|large|Large|LARGE|huge|Huge", KeyInfo(KeyInfo::isSize, 0, ignoreFormats.getSize()), isPatternString);
2871
2872         // Survives, like known character
2873         // makeKey("lyx|LyX|latex|LaTeX|latexe|LaTeXe|tex|TeX", KeyInfo(KeyInfo::isChar, 0, false), isPatternString);
2874         makeKey("tableofcontents", KeyInfo(KeyInfo::isChar, 0, false), isPatternString);
2875         makeKey("item|listitem", KeyInfo(KeyInfo::isList, 1, false), isPatternString);
2876
2877         makeKey("begin|end", KeyInfo(KeyInfo::isMath, 1, false), isPatternString);
2878         makeKey("[|]", KeyInfo(KeyInfo::isMath, 1, false), isPatternString);
2879         makeKey("$", KeyInfo(KeyInfo::isMath, 1, false), isPatternString);
2880
2881         makeKey("par|uldepth|ULdepth|protect|nobreakdash|medskip|relax", KeyInfo(KeyInfo::isStandard, 0, true), isPatternString);
2882         // Remove RTL/LTR marker
2883         makeKey("l|r|textlr|textfr|textar|beginl|endl", KeyInfo(KeyInfo::isStandard, 0, true), isPatternString);
2884         makeKey("lettrine", KeyInfo(KeyInfo::cleanToStart, 0, true), isPatternString);
2885         makeKey("lyxslide", KeyInfo(KeyInfo::isSectioning, 1, true), isPatternString);
2886         makeKey("endarguments", KeyInfo(KeyInfo::endArguments, 0, true), isPatternString);
2887         makeKey("twocolumn", KeyInfo(KeyInfo::removeWithArg, 2, true), isPatternString);
2888         makeKey("tnotetext|ead|fntext|cortext|address", KeyInfo(KeyInfo::removeWithArg, 0, true), isPatternString);
2889         makeKey("lyxend", KeyInfo(KeyInfo::isStandard, 0, true), isPatternString);
2890         if (isPatternString) {
2891                 // Allow the first searched string to rebuild the keys too
2892                 keysBuilt = false;
2893         }
2894         else {
2895                 // no need to rebuild again
2896                 keysBuilt = true;
2897         }
2898 }
2899
2900 /*
2901  * Keep the list of actual opened parentheses actual
2902  * (e.g. depth == 4 means there are 4 '{' not processed yet)
2903  */
2904 void Intervall::handleParentheses(int lastpos, bool closingAllowed)
2905 {
2906         int skip = 0;
2907         for (int i = depts[actualdeptindex]; i < lastpos; i+= 1 + skip) {
2908                 char c;
2909                 c = par[i];
2910                 skip = 0;
2911                 if (c == '\\') skip = 1;
2912                 else if (c == '{') {
2913                         handleOpenP(i);
2914                 }
2915                 else if (c == '}') {
2916                         handleCloseP(i, closingAllowed);
2917                 }
2918         }
2919 }
2920
2921 #if (0)
2922 string Intervall::show(int lastpos)
2923 {
2924         int idx = 0;                          /* int intervalls */
2925         string s;
2926         int i = 0;
2927         if ((unsigned) lastpos > par.size())
2928                 lastpos = par.size();
2929         for (idx = 0; idx <= ignoreidx; idx++) {
2930                 while (i < lastpos) {
2931                         int printsize;
2932                         if (i <= borders[idx].low) {
2933                                 if (borders[idx].low > lastpos)
2934                                         printsize = lastpos - i;
2935                                 else
2936                                         printsize = borders[idx].low - i;
2937                                 s += par.substr(i, printsize);
2938                                 i += printsize;
2939                                 if (i >= borders[idx].low)
2940                                         i = borders[idx].upper;
2941                         }
2942                         else {
2943                                 i = borders[idx].upper;
2944                                 break;
2945                         }
2946                 }
2947         }
2948         if (lastpos > i) {
2949                 s += par.substr(i, lastpos-i);
2950         }
2951         return s;
2952 }
2953 #endif
2954
2955 void Intervall::output(ostringstream &os, int lastpos)
2956 {
2957         // get number of chars to output
2958         int idx = 0;                          /* int intervalls */
2959         int i = 0;
2960         int printed = 0;
2961         string startTitle = titleValue;
2962         for (idx = 0; idx <= ignoreidx; idx++) {
2963                 if (i < lastpos) {
2964                         if (i <= borders[idx].low) {
2965                                 int printsize;
2966                                 if (borders[idx].low > lastpos)
2967                                         printsize = lastpos - i;
2968                                 else
2969                                         printsize = borders[idx].low - i;
2970                                 if (printsize > 0) {
2971                                         os << startTitle << par.substr(i, printsize);
2972                                         i += printsize;
2973                                         printed += printsize;
2974                                         startTitle = "";
2975                                 }
2976                                 handleParentheses(i, false);
2977                                 if (i >= borders[idx].low)
2978                                         i = borders[idx].upper;
2979                         }
2980                         else {
2981                                 i = borders[idx].upper;
2982                         }
2983                 }
2984                 else
2985                         break;
2986         }
2987         if (lastpos > i) {
2988                 os << startTitle << par.substr(i, lastpos-i);
2989                 printed += lastpos-i;
2990         }
2991         handleParentheses(lastpos, false);
2992         int startindex;
2993         if (keys["foreignlanguage"].disabled)
2994                 startindex = actualdeptindex-langcount;
2995         else
2996                 startindex = actualdeptindex;
2997         for (int i = startindex; i > 0; --i) {
2998                 os << "}";
2999         }
3000         if (hasTitle && (printed > 0))
3001                 os << "}";
3002         if (! isPatternString_)
3003                 os << "\n";
3004         handleParentheses(lastpos, true); /* extra closings '}' allowed here */
3005 }
3006
3007 void LatexInfo::processRegion(int start, int region_end)
3008 {
3009         while (start < region_end) {          /* Let {[} and {]} survive */
3010                 int cnt = interval_.isOpeningPar(start);
3011                 if (cnt == 1) {
3012                         // Closing is allowed past the region
3013                         int closing = interval_.findclosing(start+1, interval_.par.length());
3014                         interval_.addIntervall(start, start+1);
3015                         interval_.addIntervall(closing, closing+1);
3016                 }
3017                 else if (cnt == 3)
3018                         start += 2;
3019                 start = interval_.nextNotIgnored(start+1);
3020         }
3021 }
3022
3023 void LatexInfo::removeHead(KeyInfo const & actual, int count)
3024 {
3025         if (actual.parenthesiscount == 0) {
3026                 // "{\tiny{} ...}" ==> "{{} ...}"
3027                 interval_.addIntervall(actual._tokenstart-count, actual._tokenstart + actual._tokensize);
3028         }
3029         else {
3030                 // Remove header hull, that is "\url{abcd}" ==> "abcd"
3031                 interval_.addIntervall(actual._tokenstart - count, actual._dataStart);
3032                 interval_.addIntervall(actual._dataEnd, actual._dataEnd+1);
3033         }
3034 }
3035
3036 int LatexInfo::dispatch(ostringstream &os, int previousStart, KeyInfo &actual)
3037 {
3038         int nextKeyIdx = 0;
3039         switch (actual.keytype)
3040         {
3041         case KeyInfo::isTitle: {
3042                 removeHead(actual);
3043                 nextKeyIdx = getNextKey();
3044                 break;
3045         }
3046         case KeyInfo::cleanToStart: {
3047                 actual._dataEnd = actual._dataStart;
3048                 nextKeyIdx = getNextKey();
3049                 // Search for end of arguments
3050                 int tmpIdx = find(nextKeyIdx, KeyInfo::endArguments);
3051                 if (tmpIdx > 0) {
3052                         for (int i = nextKeyIdx; i <= tmpIdx; i++) {
3053                                 entries_[i].disabled = true;
3054                         }
3055                         actual._dataEnd = entries_[tmpIdx]._dataEnd;
3056                 }
3057                 while (interval_.par[actual._dataEnd] == ' ')
3058                         actual._dataEnd++;
3059                 interval_.addIntervall(0, actual._dataEnd+1);
3060                 interval_.actualdeptindex = 0;
3061                 interval_.depts[0] = actual._dataEnd+1;
3062                 interval_.closes[0] = -1;
3063                 break;
3064         }
3065         case KeyInfo::isText:
3066                 interval_.par[actual._tokenstart] = '#';
3067                 //interval_.addIntervall(actual._tokenstart, actual._tokenstart+1);
3068                 nextKeyIdx = getNextKey();
3069                 break;
3070         case KeyInfo::noContent: {          /* char like "\hspace{2cm}" */
3071                 if (actual.disabled)
3072                         interval_.addIntervall(actual._tokenstart, actual._dataEnd);
3073                 else
3074                         interval_.addIntervall(actual._dataStart, actual._dataEnd);
3075         }
3076                 // fall through
3077         case KeyInfo::isChar: {
3078                 nextKeyIdx = getNextKey();
3079                 break;
3080         }
3081         case KeyInfo::isSize: {
3082                 if (actual.disabled || (interval_.par[actual._dataStart] != '{') || (interval_.par[actual._dataStart-1] == ' ')) {
3083                         if (actual.parenthesiscount == 0)
3084                                 interval_.addIntervall(actual._tokenstart, actual._dataEnd);
3085                         else {
3086                                 interval_.addIntervall(actual._tokenstart, actual._dataEnd+1);
3087                         }
3088                         nextKeyIdx = getNextKey();
3089                 } else {
3090                         // Here _dataStart points to '{', so correct it
3091                         actual._dataStart += 1;
3092                         actual._tokensize += 1;
3093                         actual.parenthesiscount = 1;
3094                         if (interval_.par[actual._dataStart] == '}') {
3095                                 // Determine the end if used like '{\tiny{}...}'
3096                                 actual._dataEnd = interval_.findclosing(actual._dataStart+1, interval_.par.length()) + 1;
3097                                 interval_.addIntervall(actual._dataStart, actual._dataStart+1);
3098                         }
3099                         else {
3100                                 // Determine the end if used like '\tiny{...}'
3101                                 actual._dataEnd = interval_.findclosing(actual._dataStart, interval_.par.length()) + 1;
3102                         }
3103                         // Split on this key if not at start
3104                         int start = interval_.nextNotIgnored(previousStart);
3105                         if (start < actual._tokenstart) {
3106                                 interval_.output(os, actual._tokenstart);
3107                                 interval_.addIntervall(start, actual._tokenstart);
3108                         }
3109                         // discard entry if at end of actual
3110                         nextKeyIdx = process(os, actual);
3111                 }
3112                 break;
3113         }
3114         case KeyInfo::endArguments: {
3115                 // Remove trailing '{}' too
3116                 actual._dataStart += 1;
3117                 actual._dataEnd += 1;
3118                 interval_.addIntervall(actual._tokenstart, actual._dataEnd+1);
3119                 nextKeyIdx = getNextKey();
3120                 break;
3121         }
3122         case KeyInfo::noMain:
3123                 // fall through
3124         case KeyInfo::isStandard: {
3125                 if (actual.disabled) {
3126                         removeHead(actual);
3127                         processRegion(actual._dataStart, actual._dataStart+1);
3128                         nextKeyIdx = getNextKey();
3129                 } else {
3130                         // Split on this key if not at datastart of calling entry
3131                         int start = interval_.nextNotIgnored(previousStart);
3132                         if (start < actual._tokenstart) {
3133                                 interval_.output(os, actual._tokenstart);
3134                                 interval_.addIntervall(start, actual._tokenstart);
3135                         }
3136                         // discard entry if at end of actual
3137                         nextKeyIdx = process(os, actual);
3138                 }
3139                 break;
3140         }
3141         case KeyInfo::removeWithArg: {
3142                 nextKeyIdx = getNextKey();
3143                 // Search for end of arguments
3144                 int tmpIdx = find(nextKeyIdx, KeyInfo::endArguments);
3145                 if (tmpIdx > 0) {
3146                         for (int i = nextKeyIdx; i <= tmpIdx; i++) {
3147                                 entries_[i].disabled = true;
3148                         }
3149                         actual._dataEnd = entries_[tmpIdx]._dataEnd;
3150                 }
3151                 interval_.addIntervall(actual._tokenstart, actual._dataEnd+1);
3152                 break;
3153         }
3154         case KeyInfo::doRemove: {
3155                 // Remove the key with all parameters and following spaces
3156                 size_t pos;
3157                 size_t start;
3158                 if (interval_.par[actual._dataEnd-1] == ' ' || interval_.par[actual._dataEnd-1] == '}')
3159                         start = actual._dataEnd;
3160                 else
3161                         start = actual._dataEnd+1;
3162                 for (pos = start; pos < interval_.par.length(); pos++) {
3163                         if ((interval_.par[pos] != ' ') && (interval_.par[pos] != '%'))
3164                                 break;
3165                 }
3166                 // Remove also enclosing parentheses [] and {}
3167                 int numpars = 0;
3168                 int spaces = 0;
3169                 while (actual._tokenstart > numpars) {
3170                         if (pos+numpars >= interval_.par.size())
3171                                 break;
3172                         else if (interval_.par[pos+numpars] == ']' && interval_.par[actual._tokenstart-numpars-1] == '[')
3173                                 numpars++;
3174                         else if (interval_.par[pos+numpars] == '}' && interval_.par[actual._tokenstart-numpars-1] == '{')
3175                                 numpars++;
3176                         else
3177                                 break;
3178                 }
3179                 if (numpars > 0) {
3180                         if (interval_.par[pos+numpars] == ' ')
3181                                 spaces++;
3182                 }
3183
3184                 interval_.addIntervall(actual._tokenstart-numpars, pos+numpars+spaces);
3185                 nextKeyIdx = getNextKey();
3186                 break;
3187         }
3188         case KeyInfo::isList: {
3189                 // Discard space before _tokenstart
3190                 int count;
3191                 for (count = 0; count < actual._tokenstart; count++) {
3192                         if (interval_.par[actual._tokenstart-count-1] != ' ')
3193                                 break;
3194                 }
3195                 nextKeyIdx = getNextKey();
3196                 int tmpIdx = find(nextKeyIdx, KeyInfo::endArguments);
3197                 if (tmpIdx > 0) {
3198                         // Special case: \item is not a list, but a command (like in Style Author_Biography in maa-monthly.layout)
3199                         // with arguments
3200                         // How else can we catch this one?
3201                         for (int i = nextKeyIdx; i <= tmpIdx; i++) {
3202                                 entries_[i].disabled = true;
3203                         }
3204                         actual._dataEnd = entries_[tmpIdx]._dataEnd;
3205                 }
3206                 else if (nextKeyIdx > 0) {
3207                         // Ignore any lang entries inside data region
3208                         for (int i = nextKeyIdx; i < int(entries_.size()) && entries_[i]._tokenstart < actual._dataEnd; i++) {
3209                                 if (entries_[i].keytype == KeyInfo::isMain)
3210                                         entries_[i].disabled = true;
3211                         }
3212                 }
3213                 if (actual.disabled) {
3214                         interval_.addIntervall(actual._tokenstart-count, actual._dataEnd+1);
3215                 }
3216                 else {
3217                         interval_.addIntervall(actual._tokenstart-count, actual._tokenstart);
3218                 }
3219                 if (interval_.par[actual._dataEnd+1] == '[') {
3220                         int posdown = interval_.findclosing(actual._dataEnd+2, interval_.par.length(), '[', ']');
3221                         if ((interval_.par[actual._dataEnd+2] == '{') &&
3222                                         (interval_.par[posdown-1] == '}')) {
3223                                 interval_.addIntervall(actual._dataEnd+1,actual._dataEnd+3);
3224                                 interval_.addIntervall(posdown-1, posdown+1);
3225                         }
3226                         else {
3227                                 interval_.addIntervall(actual._dataEnd+1, actual._dataEnd+2);
3228                                 interval_.addIntervall(posdown, posdown+1);
3229                         }
3230                         int blk = interval_.nextNotIgnored(actual._dataEnd+1);
3231                         if (blk > posdown) {
3232                                 // Discard at most 1 space after empty item
3233                                 int count;
3234                                 for (count = 0; count < 1; count++) {
3235                                         if (interval_.par[blk+count] != ' ')
3236                                                 break;
3237                                 }
3238                                 if (count > 0)
3239                                         interval_.addIntervall(blk, blk+count);
3240                         }
3241                 }
3242                 break;
3243         }
3244         case KeyInfo::isSectioning: {
3245                 // Discard spaces before _tokenstart
3246                 int count;
3247                 int val = actual._tokenstart;
3248                 for (count = 0; count < actual._tokenstart;) {
3249                         val = interval_.previousNotIgnored(val-1);
3250                         if (val < 0 || interval_.par[val] != ' ')
3251                                 break;
3252                         else {
3253                                 count = actual._tokenstart - val;
3254                         }
3255                 }
3256                 if (actual.disabled) {
3257                         removeHead(actual, count);
3258                         nextKeyIdx = getNextKey();
3259                 } else {
3260                         interval_.addIntervall(actual._tokenstart-count, actual._tokenstart);
3261                         nextKeyIdx = process(os, actual);
3262                 }
3263                 break;
3264         }
3265         case KeyInfo::isMath: {
3266                 // Same as regex, use the content unchanged
3267                 nextKeyIdx = getNextKey();
3268                 break;
3269         }
3270         case KeyInfo::isRegex: {
3271                 // DO NOT SPLIT ON REGEX
3272                 // Do not disable
3273                 nextKeyIdx = getNextKey();
3274                 break;
3275         }
3276         case KeyInfo::isIgnored: {
3277                 // Treat like a character for now
3278                 nextKeyIdx = getNextKey();
3279                 break;
3280         }
3281         case KeyInfo::isMain: {
3282                 if (interval_.par.substr(actual._dataStart, 2) == "% ")
3283                         interval_.addIntervall(actual._dataStart, actual._dataStart+2);
3284                 if (actual._tokenstart > 0) {
3285                         int prev = interval_.previousNotIgnored(actual._tokenstart - 1);
3286                         if ((prev >= 0) && interval_.par[prev] == '%')
3287                                 interval_.addIntervall(prev, prev+1);
3288                 }
3289                 if (actual.disabled) {
3290                         removeHead(actual);
3291                         interval_.langcount++;
3292                         if ((interval_.par.substr(actual._dataStart, 3) == " \\[") ||
3293                                         (interval_.par.substr(actual._dataStart, 8) == " \\begin{")) {
3294                                 // Discard also the space before math-equation
3295                                 interval_.addIntervall(actual._dataStart, actual._dataStart+1);
3296                         }
3297                         nextKeyIdx = getNextKey();
3298                         // interval.resetOpenedP(actual._dataStart-1);
3299                 }
3300                 else {
3301                         if (actual._tokenstart < 26) {
3302                                 // for the first (and maybe dummy) language
3303                                 interval_.setForDefaultLang(actual);
3304                         }
3305                         interval_.resetOpenedP(actual._dataStart-1);
3306                 }
3307                 break;
3308         }
3309         case KeyInfo::invalid:
3310         case KeyInfo::headRemove:
3311                 // These two cases cannot happen, already handled
3312                 // fall through
3313         default: {
3314                 // LYXERR(Debug::INFO, "Unhandled keytype");
3315                 nextKeyIdx = getNextKey();
3316                 break;
3317         }
3318         }
3319         return nextKeyIdx;
3320 }
3321
3322 int LatexInfo::process(ostringstream & os, KeyInfo const & actual )
3323 {
3324         int end = interval_.nextNotIgnored(actual._dataEnd);
3325         int oldStart = actual._dataStart;
3326         int nextKeyIdx = getNextKey();
3327         while (true) {
3328                 if ((nextKeyIdx < 0) ||
3329                                 (entries_[nextKeyIdx]._tokenstart >= actual._dataEnd) ||
3330                                 (entries_[nextKeyIdx].keytype == KeyInfo::invalid)) {
3331                         if (oldStart <= end) {
3332                                 processRegion(oldStart, end);
3333                                 oldStart = end+1;
3334                         }
3335                         break;
3336                 }
3337                 KeyInfo &nextKey = getKeyInfo(nextKeyIdx);
3338
3339                 if ((nextKey.keytype == KeyInfo::isMain) && !nextKey.disabled) {
3340                         (void) dispatch(os, actual._dataStart, nextKey);
3341                         end = nextKey._tokenstart;
3342                         break;
3343                 }
3344                 processRegion(oldStart, nextKey._tokenstart);
3345                 nextKeyIdx = dispatch(os, actual._dataStart, nextKey);
3346
3347                 oldStart = nextKey._dataEnd+1;
3348         }
3349         // now nextKey is either invalid or is outside of actual._dataEnd
3350         // output the remaining and discard myself
3351         if (oldStart <= end) {
3352                 processRegion(oldStart, end);
3353         }
3354         if (interval_.par.size() > (size_t) end && interval_.par[end] == '}') {
3355                 end += 1;
3356                 // This is the normal case.
3357                 // But if using the firstlanguage, the closing may be missing
3358         }
3359         // get minimum of 'end' and  'actual._dataEnd' in case that the nextKey.keytype was 'KeyInfo::isMain'
3360         int output_end;
3361         if (actual._dataEnd < end)
3362                 output_end = interval_.nextNotIgnored(actual._dataEnd);
3363         else if (interval_.par.size() > (size_t) end)
3364                 output_end = interval_.nextNotIgnored(end);
3365         else
3366                 output_end = interval_.par.size();
3367         if ((actual.keytype == KeyInfo::isMain) && actual.disabled) {
3368                 interval_.addIntervall(actual._tokenstart, actual._tokenstart+actual._tokensize);
3369         }
3370         // Remove possible empty data
3371         int dstart = interval_.nextNotIgnored(actual._dataStart);
3372         while (interval_.isOpeningPar(dstart) == 1) {
3373                 interval_.addIntervall(dstart, dstart+1);
3374                 int dend = interval_.findclosing(dstart+1, output_end);
3375                 interval_.addIntervall(dend, dend+1);
3376                 dstart = interval_.nextNotIgnored(dstart+1);
3377         }
3378         if (dstart < output_end)
3379                 interval_.output(os, output_end);
3380         if (nextKeyIdx < 0)
3381                 interval_.addIntervall(0, end);
3382         else
3383                 interval_.addIntervall(actual._tokenstart, end);
3384         return nextKeyIdx;
3385 }
3386
3387 string splitOnKnownMacros(string par, bool isPatternString)
3388 {
3389         ostringstream os;
3390         LatexInfo li(par, isPatternString);
3391         // LYXERR(Debug::INFO, "Berfore split: " << par);
3392         KeyInfo DummyKey = KeyInfo(KeyInfo::KeyType::isMain, 2, true);
3393         DummyKey.head = "";
3394         DummyKey._tokensize = 0;
3395         DummyKey._dataStart = 0;
3396         DummyKey._dataEnd = par.length();
3397         DummyKey.disabled = true;
3398         int firstkeyIdx = li.getFirstKey();
3399         string s;
3400         if (firstkeyIdx >= 0) {
3401                 KeyInfo firstKey = li.getKeyInfo(firstkeyIdx);
3402                 DummyKey._tokenstart = firstKey._tokenstart;
3403                 int nextkeyIdx;
3404                 if ((firstKey.keytype != KeyInfo::isMain) || firstKey.disabled) {
3405                         // Use dummy firstKey
3406                         firstKey = DummyKey;
3407                         (void) li.setNextKey(firstkeyIdx);
3408                 }
3409                 else {
3410                         if (par.substr(firstKey._dataStart, 2) == "% ")
3411                                 li.addIntervall(firstKey._dataStart, firstKey._dataStart+2);
3412                 }
3413                 nextkeyIdx = li.process(os, firstKey);
3414                 while (nextkeyIdx >= 0) {
3415                         // Check for a possible gap between the last
3416                         // entry and this one
3417                         int datastart = li.nextNotIgnored(firstKey._dataStart);
3418                         KeyInfo &nextKey = li.getKeyInfo(nextkeyIdx);
3419                         if ((nextKey._tokenstart > datastart)) {
3420                                 // Handle the gap
3421                                 firstKey._dataStart = datastart;
3422                                 firstKey._dataEnd = par.length();
3423                                 (void) li.setNextKey(nextkeyIdx);
3424                                 // Fake the last opened parenthesis
3425                                 li.setForDefaultLang(firstKey);
3426                                 nextkeyIdx = li.process(os, firstKey);
3427                         }
3428                         else {
3429                                 if (nextKey.keytype != KeyInfo::isMain) {
3430                                         firstKey._dataStart = datastart;
3431                                         firstKey._dataEnd = nextKey._dataEnd+1;
3432                                         (void) li.setNextKey(nextkeyIdx);
3433                                         li.setForDefaultLang(firstKey);
3434                                         nextkeyIdx = li.process(os, firstKey);
3435                                 }
3436                                 else {
3437                                         nextkeyIdx = li.process(os, nextKey);
3438                                 }
3439                         }
3440                 }
3441                 // Handle the remaining
3442                 firstKey._dataStart = li.nextNotIgnored(firstKey._dataStart);
3443                 firstKey._dataEnd = par.length();
3444                 // Check if ! empty
3445                 if ((firstKey._dataStart < firstKey._dataEnd) &&
3446                                 (par[firstKey._dataStart] != '}')) {
3447                         li.setForDefaultLang(firstKey);
3448                         (void) li.process(os, firstKey);
3449                 }
3450                 s = os.str();
3451                 // return string definitelly impossible to match, but should be known
3452         }
3453         else
3454                 s = par;                            /* no known macros found */
3455         // LYXERR(Debug::INFO, "After split: " << s);
3456         return s;
3457 }
3458
3459 /*
3460  * Try to unify the language specs in the latexified text.
3461  * Resulting modified string is set to "", if
3462  * the searched tex does not contain all the features in the search pattern
3463  */
3464 static string correctlanguagesetting(string par, bool isPatternString, bool withformat, lyx::Buffer *pbuf = nullptr)
3465 {
3466         static Features regex_f;
3467         static int missed = 0;
3468         static bool regex_with_format = false;
3469
3470         int parlen = par.length();
3471
3472         while ((parlen > 0) && (par[parlen-1] == '\n')) {
3473                 parlen--;
3474         }
3475 #if 0
3476         if (isPatternString && (parlen > 0) && (par[parlen-1] == '~')) {
3477                 // Happens to be there in case of description or labeling environment
3478                 parlen--;
3479         }
3480 #endif
3481         string result;
3482         if (withformat) {
3483                 // Split the latex input into pieces which
3484                 // can be digested by our search engine
3485                 LYXERR(Debug::FINDVERBOSE, "input: \"" << par << "\"");
3486                 if (isPatternString && (pbuf != nullptr)) { // Check if we should disable/enable test for language
3487                         // We check for polyglossia, because in runparams.flavor we use Flavor::XeTeX
3488                         string doclang = pbuf->params().language->polyglossia();
3489                         static regex langre("\\\\(foreignlanguage)\\{([^\\}]+)\\}");
3490                         smatch sub;
3491                         bool toIgnoreLang = true;
3492                         for (sregex_iterator it(par.begin(), par.end(), langre), end; it != end; ++it) {
3493                                 sub = *it;
3494                                 if (sub.str(2) != doclang) {
3495                                         toIgnoreLang = false;
3496                                         break;
3497                                 }
3498                         }
3499                         setIgnoreFormat("language", toIgnoreLang, false);
3500
3501                 }
3502                 result = splitOnKnownMacros(par.substr(0,parlen), isPatternString);
3503                 LYXERR(Debug::FINDVERBOSE, "After splitOnKnownMacros:\n\"" << result << "\"");
3504         }
3505         else
3506                 result = par.substr(0, parlen);
3507         if (isPatternString) {
3508                 missed = 0;
3509                 if (withformat) {
3510                         regex_f = identifyFeatures(result);
3511                         string features = "";
3512                         for (auto it = regex_f.cbegin(); it != regex_f.cend(); ++it) {
3513                                 string a = it->first;
3514                                 regex_with_format = true;
3515                                 features += " " + a;
3516                                 // LYXERR(Debug::INFO, "Identified regex format:" << a);
3517                         }
3518                         LYXERR(Debug::FINDVERBOSE, "Identified Features" << features);
3519
3520                 }
3521         } else if (regex_with_format) {
3522                 Features info = identifyFeatures(result);
3523                 for (auto it = regex_f.cbegin(); it != regex_f.cend(); ++it) {
3524                         string a = it->first;
3525                         bool b = it->second;
3526                         if (b && ! info[a]) {
3527                                 missed++;
3528                                 LYXERR(Debug::FINDVERBOSE, "Missed(" << missed << " " << a <<", srclen = " << parlen );
3529                                 return "";
3530                         }
3531                 }
3532
3533         }
3534         else {
3535                 // LYXERR(Debug::INFO, "No regex formats");
3536         }
3537         return result;
3538 }
3539
3540
3541 // Remove trailing closure of math, macros and environments, so to catch parts of them.
3542 static void identifyClosing(string & t, bool ignoreformat)
3543 {
3544         do {
3545                 LYXERR(Debug::FINDVERBOSE, "identifyClosing(): t now is '" << t << "'");
3546                 if (regex_replace(t, t, "(.*[^\\\\])\\$$", "$1"))
3547                         continue;
3548                 if (regex_replace(t, t, "(.*[^\\\\])\\\\\\]$", "$1"))
3549                         continue;
3550                 if (regex_replace(t, t, "(.*[^\\\\])\\\\end\\{[a-zA-Z_]+\\*?\\}$", "$1"))
3551                         continue;
3552                 if (! ignoreformat) {
3553                         if (regex_replace(t, t, "(.*[^\\\\])\\}$", "$1"))
3554                                 continue;
3555                 }
3556                 break;
3557         } while (true);
3558         return;
3559 }
3560
3561 static int num_replaced = 0;
3562 static bool previous_single_replace = true;
3563
3564 void MatchStringAdv::CreateRegexp(FindAndReplaceOptions const & opt, string regexp_str, string regexp2_str, string par_as_string)
3565 {
3566 #if QTSEARCH
3567         if (regexp_str.empty() || regexp2_str.empty()) {
3568                 regexIsValid = false;
3569                 regexError = "Invalid empty regex";
3570                 return;
3571         }
3572         // Handle \w properly
3573         QRegularExpression::PatternOptions popts = QRegularExpression::UseUnicodePropertiesOption | QRegularExpression::MultilineOption;
3574         if (! opt.casesensitive) {
3575                 popts |= QRegularExpression::CaseInsensitiveOption;
3576         }
3577         regexp = QRegularExpression(QString::fromStdString(regexp_str), popts);
3578         regexp2 = QRegularExpression(QString::fromStdString(regexp2_str), popts);
3579         regexError = "";
3580         if (regexp.isValid() && regexp2.isValid()) {
3581                 regexIsValid = true;
3582                 // Check '{', '}' pairs inside the regex
3583                 int balanced = 0;
3584                 int skip = 1;
3585                 for (unsigned i = 0; i < par_as_string.size(); i+= skip) {
3586                         char c = par_as_string[i];
3587                         if (c == '\\') {
3588                                 skip = 2;
3589                                 continue;
3590                         }
3591                         if (c == '{')
3592                                 balanced++;
3593                         else if (c == '}') {
3594                                 balanced--;
3595                                 if (balanced < 0)
3596                                         break;
3597                         }
3598                         skip = 1;
3599                 }
3600                 if (balanced != 0) {
3601                         regexIsValid = false;
3602                         regexError = "Unbalanced curly brackets in regexp \"" + regexp_str + "\"";
3603                 }
3604         }
3605         else {
3606                 regexIsValid = false;
3607                 if (!regexp.isValid())
3608                         regexError += "Invalid regexp \"" + regexp_str + "\", error = " + regexp.errorString().toStdString();
3609                 else
3610                         regexError += "Invalid regexp2 \"" + regexp2_str + "\", error = " + regexp2.errorString().toStdString();
3611         }
3612 #else
3613         (void)par_as_string;
3614         if (opt.casesensitive) {
3615                 regexp = regex(regexp_str);
3616                 regexp2 = regex(regexp2_str);
3617         }
3618         else {
3619                 regexp = regex(regexp_str, std::regex_constants::icase);
3620                 regexp2 = regex(regexp2_str, std::regex_constants::icase);
3621         }
3622 #endif
3623 }
3624
3625 static void modifyRegexForMatchWord(string &t)
3626 {
3627         string s("");
3628         regex wordre("(\\\\)*((\\.|\\\\b))");
3629         size_t lastpos = 0;
3630         smatch sub;
3631         for (sregex_iterator it(t.begin(), t.end(), wordre), end; it != end; ++it) {
3632                 sub = *it;
3633                 if ((sub.position(2) - sub.position(0)) % 2 == 1) {
3634                         continue;
3635                 }
3636                 else if (sub.str(2) == "\\\\b")
3637                         return;
3638                 if (lastpos < (size_t) sub.position(2))
3639                         s += t.substr(lastpos, sub.position(2) - lastpos);
3640                 s += "\\S";
3641                 lastpos = sub.position(2) + sub.length(2);
3642         }
3643         if (lastpos == 0) {
3644                 s = "\\b" + t + "\\b";
3645                 t = s;
3646                 return;
3647         }
3648         else if (lastpos < t.length())
3649                 s += t.substr(lastpos, t.length() - lastpos);
3650         t = "\\b" + s + "\\b";
3651 }
3652
3653 MatchStringAdv::MatchStringAdv(lyx::Buffer & buf, FindAndReplaceOptions & opt)
3654         : p_buf(&buf), p_first_buf(&buf), opt(opt)
3655 {
3656         Buffer & find_buf = *theBufferList().getBuffer(FileName(to_utf8(opt.find_buf_name)), true);
3657         docstring const & ds = stringifySearchBuffer(find_buf, opt);
3658         if (ds.empty() ) {
3659                 CreateRegexp(opt, "", "", "");
3660                 return;
3661         }
3662         use_regexp = ds.find(from_utf8("\\regexp{")) != std::string::npos;
3663         if (opt.replace_all && previous_single_replace) {
3664                 previous_single_replace = false;
3665                 num_replaced = 0;
3666         }
3667         else if (!opt.replace_all) {
3668                 num_replaced = 0;       // count number of replaced strings
3669                 previous_single_replace = true;
3670         }
3671         // When using regexp, braces are hacked already by escape_for_regex()
3672         par_as_string = normalize(ds, opt.ignoreformat);
3673         open_braces = 0;
3674         close_wildcards = 0;
3675
3676         size_t lead_size = 0;
3677         // correct the language settings
3678         par_as_string = correctlanguagesetting(par_as_string, true, !opt.ignoreformat, &buf);
3679         if (par_as_string.empty()) {
3680                 CreateRegexp(opt, "", "", "");
3681                 return;
3682         }
3683         opt.matchAtStart = false;
3684         if (!use_regexp) {
3685                 identifyClosing(par_as_string, opt.ignoreformat); // Removes math closings ($, ], ...) at end of string
3686                 if (opt.ignoreformat) {
3687                         lead_size = 0;
3688                 }
3689                 else {
3690                         lead_size = identifyLeading(par_as_string);
3691                 }
3692                 lead_as_string = par_as_string.substr(0, lead_size);
3693                 string lead_as_regex_string = string2regex(lead_as_string);
3694                 par_as_string_nolead = par_as_string.substr(lead_size, par_as_string.size() - lead_size);
3695                 string par_as_regex_string_nolead = string2regex(par_as_string_nolead);
3696                 /* Handle whole words too in this case
3697                 */
3698                 if (opt.matchword) {
3699                         par_as_regex_string_nolead = "\\b" + par_as_regex_string_nolead + "\\b";
3700                         opt.matchword = false;
3701                 }
3702                 string regexp_str = "(" + lead_as_regex_string + ")()" + par_as_regex_string_nolead;
3703                 string regexp2_str = "(" + lead_as_regex_string + ")(.*?)" + par_as_regex_string_nolead;
3704                 CreateRegexp(opt, regexp_str, regexp2_str);
3705                 use_regexp = true;
3706                 LYXERR(Debug::FINDVERBOSE|Debug::FIND, "Setting regexp to : '" << regexp_str << "'");
3707                 LYXERR(Debug::FINDVERBOSE|Debug::FIND, "Setting regexp2 to: '" << regexp2_str << "'");
3708                 return;
3709         }
3710
3711         if (!opt.ignoreformat) {
3712                 lead_size = identifyLeading(par_as_string);
3713                 LYXERR(Debug::FINDVERBOSE, "Lead_size: " << lead_size);
3714                 lead_as_string = par_as_string.substr(0, lead_size);
3715                 par_as_string_nolead = par_as_string.substr(lead_size, par_as_string.size() - lead_size);
3716         }
3717
3718         // Here we are using regexp
3719         LASSERT(use_regexp, /**/);
3720         {
3721                 string lead_as_regexp;
3722                 if (lead_size > 0) {
3723                         lead_as_regexp = string2regex(par_as_string.substr(0, lead_size));
3724                         (void)regex_replace(par_as_string_nolead, par_as_string_nolead, "\\$$", "");
3725                         (void)regex_replace(par_as_string_nolead, par_as_string_nolead, "}$", "");
3726                         par_as_string = par_as_string_nolead;
3727                         LYXERR(Debug::FINDVERBOSE, "lead_as_regexp is '" << lead_as_regexp << "'");
3728                         LYXERR(Debug::FINDVERBOSE, "par_as_string now is '" << par_as_string << "'");
3729                 }
3730                 // LYXERR(Debug::FINDVERBOSE, "par_as_string before escape_for_regex() is '" << par_as_string << "'");
3731                 par_as_string = escape_for_regex(par_as_string, !opt.ignoreformat);
3732                 // Insert (.*?) before trailing closure of math, macros and environments, so to catch parts of them.
3733                 // LYXERR(Debug::FINDVERBOSE, "par_as_string now is '" << par_as_string << "'");
3734                 ++close_wildcards;
3735                 size_t lng = par_as_string.size();
3736                 if (!opt.ignoreformat) {
3737                         // Remove extra '\}' at end if not part of \{\.\}
3738                         while(lng > 2) {
3739                                 if (par_as_string.substr(lng-2, 2).compare("\\}") == 0) {
3740                                         if (lng >= 6) {
3741                                                 if (par_as_string.substr(lng-6,3).compare("\\{\\") == 0)
3742                                                         break;
3743                                         }
3744                                         lng -= 2;
3745                                         open_braces++;
3746                                 }
3747                                 else
3748                                         break;
3749                         }
3750                         if (lng < par_as_string.size())
3751                                 par_as_string.resize(lng);
3752                 }
3753                 LYXERR(Debug::FINDVERBOSE, "par_as_string after correctRegex is '" << par_as_string << "'");
3754                 if ((lng > 0) && (par_as_string[0] == '^')) {
3755                         par_as_string = par_as_string.substr(1);
3756                         --lng;
3757                         opt.matchAtStart = true;
3758                 }
3759                 // LYXERR(Debug::FINDVERBOSE, "par_as_string now is '" << par_as_string << "'");
3760                 // LYXERR(Debug::FINDVERBOSE, "Open braces: " << open_braces);
3761                 // LYXERR(Debug::FINDVERBOSE, "Replaced text (to be used as regex): " << par_as_string);
3762
3763                 // If entered regexp must match at begin of searched string buffer
3764                 // Kornel: Added parentheses to use $1 for size of the leading string
3765                 string regexp_str;
3766                 string regexp2_str;
3767                 {
3768                         // TODO: Adapt '\[12345678]' in par_as_string to acount for the first '()
3769                         // Unfortunately is '\1', '\2', etc not working for strings with extra format
3770                         // so the convert has no effect in that case
3771                         for (int i = 7; i > 0; --i) {
3772                                 string orig = "\\\\" + std::to_string(i);
3773                                 string dest = "\\" + std::to_string(i+2);
3774                                 while (regex_replace(par_as_string, par_as_string, orig, dest));
3775                         }
3776                         if (opt.matchword) {
3777                                 modifyRegexForMatchWord(par_as_string);
3778                                 // opt.matchword = false;
3779                         }
3780                         regexp_str = "(" + lead_as_regexp + ")()" + par_as_string;
3781                         regexp2_str = "(" + lead_as_regexp + ")(.*?)" + par_as_string;
3782                 }
3783                 LYXERR(Debug::FINDVERBOSE|Debug::FIND, "Setting regexp to : '" << regexp_str << "'");
3784                 LYXERR(Debug::FINDVERBOSE|Debug::FIND, "Setting regexp2 to: '" << regexp2_str << "'");
3785                 CreateRegexp(opt, regexp_str, regexp2_str, par_as_string);
3786         }
3787 }
3788
3789 MatchResult MatchStringAdv::findAux(DocIterator const & cur, int len, MatchStringAdv::matchType at_begin) const
3790 {
3791         MatchResult mres;
3792
3793         mres.searched_size = len;
3794
3795         docstring docstr = stringifyFromForSearch(opt, cur, len);
3796         string str;
3797         str = normalize(docstr, opt.ignoreformat);
3798         if (!opt.ignoreformat) {
3799                 str = correctlanguagesetting(str, false, !opt.ignoreformat);
3800                 // remove closing '}' and '\n' to allow for use of '$' in regex
3801                 size_t lng = str.size();
3802                 while ((lng > 1) && ((str[lng -1] == '}') || (str[lng -1] == '\n')))
3803                         lng--;
3804                 if (lng != str.size()) {
3805                         str = str.substr(0, lng);
3806                 }
3807                 // Replace occurences of '~' to ' '
3808                 static std::regex specialChars { R"(~)" };
3809                 str = std::regex_replace(str, specialChars,  R"( )" );
3810         }
3811         if (str.empty()) {
3812                 mres.match_len = -1;
3813                 return mres;
3814         }
3815         LYXERR(Debug::FINDVERBOSE|Debug::FIND, "After normalization: Matching against:\n'" << str << "'");
3816
3817         LASSERT(use_regexp, /**/);
3818         {
3819                 // use_regexp always true
3820                 LYXERR(Debug::FINDVERBOSE, "Searching in regexp mode: at_begin=" << matchTypeAsString(at_begin));
3821 #if QTSEARCH
3822                 QString qstr = QString::fromStdString(str);
3823                 QRegularExpression const *p_regexp;
3824                 QRegularExpression::MatchType flags = QRegularExpression::NormalMatch;
3825                 if (at_begin == MatchStringAdv::MatchFromStart) {
3826                         p_regexp = &regexp;
3827                 } else {
3828                         p_regexp = &regexp2;
3829                 }
3830                 QRegularExpressionMatch match = p_regexp->match(qstr, 0, flags);
3831                 if (!match.hasMatch())
3832                         return mres;
3833 #else
3834                 regex const *p_regexp;
3835                 regex_constants::match_flag_type flags;
3836                 if (at_begin == MatchStringAdv::MatchFromStart) {
3837                         flags = regex_constants::match_continuous;
3838                         p_regexp = &regexp;
3839                 } else {
3840                         flags = regex_constants::match_default;
3841                         p_regexp = &regexp2;
3842                 }
3843                 sregex_iterator re_it(str.begin(), str.end(), *p_regexp, flags);
3844                 if (re_it == sregex_iterator())
3845                         return mres;
3846                 match_results<string::const_iterator> const & m = *re_it;
3847 #endif
3848                 // Whole found string, including the leading
3849                 // std: m[0].second - m[0].first
3850                 // Qt: match.capturedEnd(0) - match.capturedStart(0)
3851                 //
3852                 // Size of the leading string
3853                 // std: m[1].second - m[1].first
3854                 // Qt: match.capturedEnd(1) - match.capturedStart(1)
3855                 int leadingsize = 0;
3856 #if QTSEARCH
3857                 if (match.lastCapturedIndex() > 0) {
3858                         leadingsize = match.capturedEnd(1) - match.capturedStart(1);
3859                 }
3860
3861 #else
3862                 if (m.size() > 2) {
3863                         leadingsize = m[1].second - m[1].first;
3864                 }
3865 #endif
3866 #if QTSEARCH
3867                 mres.match_prefix = match.capturedEnd(2) - match.capturedStart(2);
3868                 mres.match_len = match.capturedEnd(0) - match.capturedEnd(2);
3869                 // because of different number of closing at end of string
3870                 // we have to 'unify' the length of the post-match.
3871                 // Done by ignoring closing parenthesis and linefeeds at string end
3872                 int matchend = match.capturedEnd(0);
3873                 size_t strsize = qstr.size();
3874                 if (!opt.ignoreformat) {
3875                         while (mres.match_len > 1) {
3876                                 QChar c = qstr.at(matchend - 1);
3877                                 if ((c == '\n') || (c == '}') || (c == '{')) {
3878                                         mres.match_len--;
3879                                         matchend--;
3880                                 }
3881                                 else
3882                                         break;
3883                         }
3884                         while (strsize > (size_t) match.capturedEnd(0)) {
3885                                 QChar c = qstr.at(strsize-1);
3886                                 if ((c == '\n') || (c == '}')) {
3887                                         --strsize;
3888                                 }
3889                                 else
3890                                         break;
3891                         }
3892                 }
3893                 // LYXERR0(qstr.toStdString());
3894                 mres.match2end = strsize - matchend;
3895                 mres.pos = match.capturedStart(2);
3896 #else
3897                 mres.match_prefix = m[2].second - m[2].first;
3898                 mres.match_len = m[0].second - m[2].second;
3899                 // ignore closing parenthesis and linefeeds at string end
3900                 size_t strend = m[0].second - m[0].first;
3901                 int matchend = strend;
3902                 size_t strsize = str.size();
3903                 if (!opt.ignoreformat) {
3904                         while (mres.match_len > 1) {
3905                                 char c = str.at(matchend - 1);
3906                                 if ((c == '\n') || (c == '}') || (c == '{')) {
3907                                         mres.match_len--;
3908                                         matchend--;
3909                                 }
3910                                 else
3911                                         break;
3912                         }
3913                         while (strsize > strend) {
3914                                 if ((str.at(strsize-1) == '}') || (str.at(strsize-1) == '\n')) {
3915                                         --strsize;
3916                                 }
3917                                 else
3918                                         break;
3919                         }
3920                 }
3921                 // LYXERR0(str);
3922                 mres.match2end = strsize - matchend;
3923                 mres.pos = m[2].first - m[0].first;;
3924 #endif
3925                 if (mres.match2end < 0)
3926                         mres.match_len = 0;
3927                 mres.leadsize = leadingsize;
3928 #if QTSEARCH
3929                 if (mres.match_len > 0) {
3930                         string a0 = match.captured(0).mid(mres.pos + mres.match_prefix, mres.match_len).toStdString();
3931                         mres.result.push_back(a0);
3932                         for (int i = 3; i <= match.lastCapturedIndex(); i++) {
3933                                 mres.result.push_back(match.captured(i).toStdString());
3934                         }
3935                 }
3936 #else
3937                 if (mres.match_len > 0) {
3938                         string a0 = m[0].str().substr(mres.pos + mres.match_prefix, mres.match_len);
3939                         mres.result.push_back(a0);
3940                         for (size_t i = 3; i < m.size(); i++) {
3941                                 mres.result.push_back(m[i]);
3942                         }
3943                 }
3944 #endif
3945                 return mres;
3946         }
3947 }
3948
3949
3950 MatchResult MatchStringAdv::operator()(DocIterator const & cur, int len, MatchStringAdv::matchType at_begin) const
3951 {
3952         MatchResult mres = findAux(cur, len, at_begin);
3953         LYXERR(Debug::FINDVERBOSE,
3954                "res=" << mres.match_len << ", at_begin=" << matchTypeAsString(at_begin)
3955                << ", matchAtStart=" << opt.matchAtStart
3956                << ", inTexted=" << cur.inTexted());
3957         if (mres.match_len > 0) {
3958                 if (opt.matchAtStart) {
3959                         if (cur.pos() > 0 || mres.match_prefix > 0)
3960                                 mres.match_len = 0;
3961                 }
3962         }
3963         return mres;
3964 }
3965
3966 #if 0
3967 static bool simple_replace(string &t, string from, string to)
3968 {
3969         regex repl("(\\\\)*(" + from + ")");
3970         string s("");
3971         size_t lastpos = 0;
3972         smatch sub;
3973         for (sregex_iterator it(t.begin(), t.end(), repl), end; it != end; ++it) {
3974                 sub = *it;
3975                 if ((sub.position(2) - sub.position(0)) % 2 == 1)
3976                         continue;
3977                 if (lastpos < (size_t) sub.position(2))
3978                         s += t.substr(lastpos, sub.position(2) - lastpos);
3979                 s += to;
3980                 lastpos = sub.position(2) + sub.length(2);
3981         }
3982         if (lastpos == 0)
3983                 return false;
3984         else if (lastpos < t.length())
3985                 s += t.substr(lastpos, t.length() - lastpos);
3986         t = s;
3987         return true;
3988 }
3989 #endif
3990
3991 #if 1
3992 static string convertLF2Space(docstring const &s, bool ignore_format)
3993 {
3994         // Using original docstring to handle '\n'
3995
3996         if (s.size() == 0) return "";
3997         stringstream t;
3998         size_t pos;
3999         size_t start = 0;
4000         size_t end = s.size() - 1;
4001         if (!ignore_format) {
4002                 while (s[start] == '\n' && start <= end)
4003                         start++;
4004                 while (end >= start && s[end] == '\n')
4005                         end--;
4006                 if (start >= end + 1)
4007                         return "";
4008         }
4009         do {
4010                 bool dospace = true;
4011                 int skip = -1;
4012                 pos = s.find('\n', start);
4013                 if (pos >= end) {
4014                         t << lyx::to_utf8(s.substr(start, end + 1 - start));
4015                         break;
4016                 }
4017                 if (!ignore_format) {
4018                         if ((pos > start + 1) &&
4019                              s[pos-1] == '\\' &&
4020                              s[pos-2] == '\\') {
4021                                 skip = 2;
4022                                 if ((pos > start + 2) &&
4023                                     (s[pos+1] == '~' || isSpace(s[pos+1]) ||
4024                                      s[pos-3] == '~' || isSpace(s[pos-3]))) {
4025                                         // discard "\\\\\n", do not replace with space
4026                                         dospace = false;
4027                                 }
4028                         }
4029                         else if (pos > start) {
4030                                 if (s[pos-1] == '%') {
4031                                         skip = 1;
4032                                         while ((pos > start+skip) && (s[pos-1-skip] == '%'))
4033                                                 skip++;
4034                                         if ((pos > start+skip) &&
4035                                             (s[pos+1] == '~' || isSpace(s[pos+1]) ||
4036                                              s[pos-1-skip] == '~' || isSpace(s[pos-1-skip]))) {
4037                                                 // discard '%%%%%\n'
4038                                                 dospace = false;
4039                                         }
4040                                 }
4041                                 else if (!isAlnumASCII(s[pos+1]) || !isAlnumASCII(s[pos-1])) {
4042                                         dospace = false;
4043                                         skip = 0;       // remove the '\n' only
4044                                 }
4045                         }
4046                 }
4047                 else {
4048                         dospace = true;
4049                         skip = 0;
4050                 }
4051                 t << lyx::to_utf8(s.substr(start, pos-skip-start));
4052                 if (dospace)
4053                         t << ' ';
4054                 start = pos+1;
4055         } while (start <= end);
4056         return(t.str());
4057 }
4058
4059 #else
4060 static string convertLF2Space(docstring const & s, bool ignore_format)
4061 {
4062         // Using utf8-converted string to handle '\n'
4063
4064         string t;
4065         t = lyx::to_utf8(s);
4066         // Remove \n at begin
4067         while (!t.empty() && t[0] == '\n')
4068                 t = t.substr(1);
4069         // Remove \n* at end
4070         while (!t.empty() && t[t.size() - 1] == '\n') {
4071                 t = t.substr(0, t.size() - 1);
4072         }
4073         size_t pos;
4074         // Handle all other '\n'
4075         while ((pos = t.find("\n")) != string::npos) {
4076                 if (pos > 1 && t[pos-1] == '\\' && t[pos-2] == '\\' ) {
4077                         // Handle '\\\n'
4078                         if (isPrintableNonspace(t[pos+1]) && ((pos < 3) || isPrintableNonspace(t[pos-3]))) {
4079                                 t.replace(pos-2, 3, " ");
4080                         }
4081                         else {
4082                                 // Already a space there
4083                                 t.replace(pos-2, 3, "");
4084                         }
4085                 }
4086                 else {
4087                         if (!isAlnumASCII(t[pos+1]) || !isAlnumASCII(t[pos-1])) {
4088                                 // '\n' adjacent to non-alpha-numerics, discard
4089                                 t.replace(pos, 1, "");
4090                         }
4091                         else {
4092                                 // Replace all other \n with spaces
4093                                 t.replace(pos, 1, " ");
4094                         }
4095                         if (!ignore_format) {
4096                                 size_t count = 0;
4097                                 while ((pos > count + 1) && (t[pos - 1 -count] == '%')) {
4098                                         count++;
4099                                 }
4100                                 if (count > 0) {
4101                                         t.replace(pos - count, count, "");
4102                                 }
4103                         }
4104                 }
4105         }
4106         return(t);
4107
4108 }
4109 #endif
4110
4111 string MatchStringAdv::normalize(docstring const & s, bool ignore_format) const
4112 {
4113         string t = convertLF2Space(s, ignore_format);
4114
4115         // The following replaces are not appropriate in non-format-search mode
4116         if (!ignore_format) {
4117                 // Remove stale empty \emph{}, \textbf{} and similar blocks from latexify
4118                 // Kornel: Added textsl, textsf, textit, texttt and noun
4119                 // + allow to seach for colored text too
4120                 LYXERR(Debug::FINDVERBOSE, "Removing stale empty macros from: " << t);
4121                 while (regex_replace(t, t, "\\\\(emph|noun|text(bf|sl|sf|it|tt)|(u|uu)line|(s|x)out|uwave)(\\{(\\{\\})?\\})+", ""))
4122                         LYXERR(Debug::FINDVERBOSE, "  further removing stale empty \\emph{}, \\textbf{} macros from: " << t);
4123                 while (regex_replace(t, t, "\\\\((sub)?(((sub)?section)|paragraph)|part)\\*?(\\{(\\{\\})?\\})+", ""))
4124                         LYXERR(Debug::FINDVERBOSE, "  further removing stale empty \\section{}, \\part{}, \\paragraph{} macros from: " << t);
4125                 while (regex_replace(t, t, "\\\\(foreignlanguage|textcolor|item)\\{[a-z]+\\}(\\{(\\{\\})?\\})+", ""));
4126         }
4127         return t;
4128 }
4129
4130 docstring stringifyFromCursor(DocIterator const & cur, int len)
4131 {
4132         LYXERR(Debug::FINDVERBOSE, "Stringifying with len=" << len << " from cursor at pos: " << cur);
4133         if (cur.inTexted()) {
4134                 Paragraph const & par = cur.paragraph();
4135                 // TODO what about searching beyond/across paragraph breaks ?
4136                 // TODO Try adding a AS_STR_INSERTS as last arg
4137                 pos_type end = ( len == -1 || cur.pos() + len > int(par.size()) ) ?
4138                                         int(par.size()) : cur.pos() + len;
4139                 // OutputParams runparams(&cur.buffer()->params().encoding());
4140                 OutputParams runparams(encodings.fromLyXName("utf8"));
4141                 runparams.nice = true;
4142                 setFindParams(runparams);
4143                 int option = AS_STR_INSETS | AS_STR_PLAINTEXT;
4144                 if (ignoreFormats.getDeleted()) {
4145                         option |= AS_STR_SKIPDELETE;
4146                         runparams.find_set_feature(OutputParams::SearchWithoutDeleted);
4147                 }
4148                 else {
4149                         runparams.find_set_feature(OutputParams::SearchWithDeleted);
4150                 }
4151                 if (ignoreFormats.getNonContent()) {
4152                         runparams.find_add_feature(OutputParams::SearchNonOutput);
4153                 }
4154                 LYXERR(Debug::FINDVERBOSE, "Stringifying with cur: "
4155                        << cur << ", from pos: " << cur.pos() << ", end: " << end);
4156                 docstring res = from_utf8(latexNamesToUtf8(par.asString(cur.pos(), end,
4157                                                                         option,
4158                                                                         &runparams), false));
4159                 LYXERR(Debug::FINDVERBOSE|Debug::FIND, "Stringified text from pos(" << cur.pos() << ") len(" << len << "): " << res);
4160                 return res;
4161         } else if (cur.inMathed()) {
4162                 CursorSlice cs = cur.top();
4163                 MathData md = cs.cell();
4164                 MathData::const_iterator it_end =
4165                                 (( len == -1 || cs.pos() + len > int(md.size()))
4166                                  ? md.end()
4167                                  : md.begin() + cs.pos() + len );
4168                 MathData md2;
4169                 for (MathData::const_iterator it = md.begin() + cs.pos(); it != it_end; ++it)
4170                         md2.push_back(*it);
4171                 docstring res = from_utf8(latexNamesToUtf8(asString(md2), false));
4172                 LYXERR(Debug::FINDVERBOSE|Debug::FIND, "Stringified math from pos(" << cur.pos() << ") len(" << len << "): " << res);
4173                 return res;
4174         }
4175         LYXERR(Debug::FINDVERBOSE|Debug::FIND, "Don't know how to stringify from here: " << cur);
4176         return docstring();
4177 }
4178
4179 /** Computes the LaTeX export of buf starting from cur and ending len positions
4180  * after cur, if len is positive, or at the paragraph or innermost inset end
4181  * if len is -1.
4182  */
4183 docstring latexifyFromCursor(DocIterator const & cur, int len)
4184 {
4185         /*
4186         LYXERR(Debug::FINDVERBOSE, "Latexifying with len=" << len << " from cursor at pos: " << cur);
4187         LYXERR(Debug::FINDVERBOSE, "  with cur.lastpost=" << cur.lastpos() << ", cur.lastrow="
4188                << cur.lastrow() << ", cur.lastcol=" << cur.lastcol());
4189         */
4190         Buffer const & buf = *cur.buffer();
4191
4192         odocstringstream ods;
4193         otexstream os(ods);
4194         //OutputParams runparams(&buf.params().encoding());
4195         OutputParams runparams(encodings.fromLyXName("utf8"));
4196         runparams.nice = false;
4197         setFindParams(runparams);
4198         if (ignoreFormats.getDeleted()) {
4199                 runparams.find_set_feature(OutputParams::SearchWithoutDeleted);
4200         }
4201         else {
4202                 runparams.find_set_feature(OutputParams::SearchWithDeleted);
4203         }
4204         if (ignoreFormats.getNonContent()) {
4205                 runparams.find_add_feature(OutputParams::SearchNonOutput);
4206         }
4207
4208         if (cur.inTexted()) {
4209                 // @TODO what about searching beyond/across paragraph breaks ?
4210                 pos_type endpos = cur.paragraph().size();
4211                 if (len != -1 && endpos > cur.pos() + len)
4212                         endpos = cur.pos() + len;
4213                 TeXOnePar(buf, *cur.innerText(), cur.pit(), os, runparams,
4214                           string(), cur.pos(), endpos, true);
4215                 LYXERR(Debug::FINDVERBOSE|Debug::FIND, "Latexified text from pos(" << cur.pos() << ") len(" << len << "): " << ods.str());
4216                 return(ods.str());
4217         } else if (cur.inMathed()) {
4218                 // Retrieve the math environment type, and add '$' or '$[' or others (\begin{equation}) accordingly
4219                 for (int s = cur.depth() - 1; s >= 0; --s) {
4220                         CursorSlice const & cs = cur[s];
4221                         if (cs.asInsetMath() && cs.asInsetMath()->asHullInset()) {
4222                                 TeXMathStream ws(os);
4223                                 cs.asInsetMath()->asHullInset()->header_write(ws);
4224                                 break;
4225                         }
4226                 }
4227
4228                 CursorSlice const & cs = cur.top();
4229                 MathData md = cs.cell();
4230                 MathData::const_iterator it_end =
4231                                 ((len == -1 || cs.pos() + len > int(md.size()))
4232                                  ? md.end()
4233                                  : md.begin() + cs.pos() + len);
4234                 MathData md2;
4235                 for (MathData::const_iterator it = md.begin() + cs.pos();
4236                      it != it_end; ++it)
4237                         md2.push_back(*it);
4238
4239                 ods << asString(md2);
4240                 // Retrieve the math environment type, and add '$' or '$]'
4241                 // or others (\end{equation}) accordingly
4242                 for (int s = cur.depth() - 1; s >= 0; --s) {
4243                         CursorSlice const & cs2 = cur[s];
4244                         InsetMath * inset = cs2.asInsetMath();
4245                         if (inset && inset->asHullInset()) {
4246                                 TeXMathStream ws(os);
4247                                 inset->asHullInset()->footer_write(ws);
4248                                 break;
4249                         }
4250                 }
4251                 LYXERR(Debug::FINDVERBOSE|Debug::FIND, "Latexified math from pos(" << cur.pos() << ") len(" << len << "): " << ods.str());
4252         } else {
4253                 LYXERR(Debug::FINDVERBOSE|Debug::FIND, "Don't know how to stringify from here: " << cur);
4254         }
4255         return ods.str();
4256 }
4257
4258 #if defined(ResultsDebug)
4259 // Debugging output
4260 static void displayMResult(MatchResult &mres, string from, DocIterator & cur)
4261 {
4262         LYXERR0( "from:\t\t\t" << from);
4263         string status;
4264         if (mres.pos_len > 0) {
4265                 // Set in finalize
4266                 status = "FINALSEARCH";
4267         }
4268         else {
4269                 if (mres.match_len > 0) {
4270                         if ((mres.match_prefix == 0) && (mres.pos == mres.leadsize))
4271                                 status = "Good Match";
4272                         else
4273                                 status = "Matched in";
4274                 }
4275                 else
4276                         status = "MissedSearch";
4277         }
4278
4279         LYXERR0( status << "(" << cur.pos() << " ... " << mres.searched_size + cur.pos() << ") cur.lastpos(" << cur.lastpos() << ")");
4280         if ((mres.leadsize > 0) || (mres.match_len > 0) || (mres.match2end > 0))
4281                 LYXERR0( "leadsize(" << mres.leadsize << ") match_len(" << mres.match_len << ") match2end(" << mres.match2end << ")");
4282         if ((mres.pos > 0) || (mres.match_prefix > 0))
4283                 LYXERR0( "pos(" << mres.pos << ") match_prefix(" << mres.match_prefix << ")");
4284         for (size_t i = 0; i < mres.result.size(); i++)
4285                 LYXERR0( "Match " << i << " = \"" << mres.result[i] << "\"");
4286 }
4287 #define displayMres(s, txt, cur) displayMResult(s, txt, cur);
4288 #else
4289 #define displayMres(s, txt, cur)
4290 #endif
4291
4292 /** Finalize an advanced find operation, advancing the cursor to the innermost
4293  ** position that matches, plus computing the length of the matching text to
4294  ** be selected
4295  ** Return the cur.pos() difference between start and end of found match
4296  **/
4297 MatchResult findAdvFinalize(DocIterator & cur, MatchStringAdv const & match, MatchResult const & expected = MatchResult(-1))
4298 {
4299         // Search the foremost position that matches (avoids find of entire math
4300         // inset when match at start of it)
4301         DocIterator old_cur(cur.buffer());
4302         MatchResult mres;
4303         static MatchResult fail = MatchResult();
4304         MatchResult max_match;
4305         // If (prefix_len > 0) means that forwarding 1 position will remove the complete entry
4306         // Happens with e.g. hyperlinks
4307         // either one sees "http://www.bla.bla" or nothing
4308         // so the search for "www" gives prefix_len = 7 (== sizeof("http://")
4309         // and although we search for only 3 chars, we find the whole hyperlink inset
4310         MatchStringAdv::matchType at_begin = (expected.match_prefix == 0) ? MatchStringAdv::MatchFromStart : MatchStringAdv::MatchAnyPlace;
4311         if (!match.opt.forward && match.opt.ignoreformat) {
4312                 if (expected.pos > 0)
4313                         return fail;
4314         }
4315         LASSERT(at_begin == MatchStringAdv::MatchFromStart, /**/);
4316         if (expected.match_len > 0 && at_begin == MatchStringAdv::MatchFromStart) {
4317                 // Search for deepest match
4318                 old_cur = cur;
4319                 max_match = expected;
4320                 do {
4321                         size_t d = cur.depth();
4322                         cur.forwardPos();
4323                         if (!cur)
4324                                 break;
4325                         if (cur.depth() < d)
4326                                 break;
4327                         if (cur.depth() == d)
4328                                 break;
4329                         size_t lastd = d;
4330                         while (cur && cur.depth() > lastd) {
4331                                 lastd = cur.depth();
4332                                 mres = match(cur, -1, at_begin);
4333                                 displayMres(mres, "Checking innermost", cur);
4334                                 if (mres.match_len > 0)
4335                                         break;
4336                                 // maybe deeper?
4337                                 cur.forwardPos();
4338                         }
4339                         if (mres.match_len < expected.match_len)
4340                                 break;
4341                         max_match = mres;
4342                         old_cur = cur;;
4343                 } while(1);
4344                 cur = old_cur;
4345         }
4346         else {
4347                 // (expected.match_len <= 0)
4348                 mres = match(cur, -1, MatchStringAdv::MatchFromStart);      /* match valid only if not searching whole words */
4349                 displayMres(mres, "Start with negative match", cur);
4350                 max_match = mres;
4351         }
4352         // Only now we are really at_begin
4353         if ((max_match.match_len <= 0) ||
4354             (match.opt.restr == FindAndReplaceOptions::R_ONLY_MATHS && !cur.inMathed()))
4355                 return fail;
4356         LYXERR(Debug::FINDVERBOSE, "Ok");
4357
4358         // Compute the match length
4359         int len = 1;
4360         if (cur.pos() + len > cur.lastpos())
4361                 return fail;
4362
4363         LASSERT(match.use_regexp, /**/);
4364         {
4365                 int minl = 1;
4366                 int maxl = cur.lastpos() - cur.pos();
4367                 // Greedy behaviour while matching regexps
4368                 while (maxl > minl) {
4369                         MatchResult mres2;
4370                         mres2 = match(cur, len, at_begin);
4371                         displayMres(mres2, "Finalize loop", cur);
4372                         int actual_match_len = mres2.match_len;
4373                         if (actual_match_len >= max_match.match_len) {
4374                                 // actual_match_len > max_match _can_ happen,
4375                                 // if the search area splits
4376                                 // some following word so that the regex
4377                                 // (e.g. 'r.*r\b' matches 'r' from the middle of the
4378                                 // splitted word)
4379                                 // This means, the len value is too big
4380                                 actual_match_len = max_match.match_len;
4381                                 max_match = mres2;
4382                                 max_match.match_len = actual_match_len;
4383                                 maxl = len;
4384                                 if (maxl - minl < 4)
4385                                         len = (maxl + minl)/2;
4386                                 else
4387                                         len = minl + (maxl - minl + 3)/4;
4388                         }
4389                         else {
4390                                 // (actual_match_len < max_match.match_len)
4391                                 minl = len + 1;
4392                                 len = (maxl + minl)/2;
4393                         }
4394                 }
4395                 len = minl;
4396                 old_cur = cur;
4397                 // Search for real start of matched characters
4398                 while (len > 1) {
4399                         MatchResult actual_match;
4400                         do {
4401                                 cur.forwardPos();
4402                         } while (cur.depth() > old_cur.depth()); /* Skip inner insets */
4403                         if (cur.depth() < old_cur.depth()) {
4404                                 // Outer inset?
4405                                 LYXERR(Debug::INFO, "cur.depth() < old_cur.depth(), this should never happen");
4406                                 break;
4407                         }
4408                         if (cur.pos() != old_cur.pos()) {
4409                                 // OK, forwarded 1 pos in actual inset
4410                                 actual_match = match(cur, len-1, at_begin);
4411                                 if (actual_match.match_len == max_match.match_len) {
4412                                         // Ha, got it! The shorter selection has the same match length
4413                                         len--;
4414                                         old_cur = cur;
4415                                         max_match = actual_match;
4416                                 }
4417                                 else {
4418                                         // OK, the shorter selection matches less chars, revert to previous value
4419                                         cur = old_cur;
4420                                         break;
4421                                 }
4422                         }
4423                         else {
4424                                 LYXERR(Debug::INFO, "cur.pos() == old_cur.pos(), this should never happen");
4425                                 actual_match = match(cur, len, at_begin);
4426                                 if (actual_match.match_len == max_match.match_len) {
4427                                         old_cur = cur;
4428                                         max_match = actual_match;
4429                                 }
4430                         }
4431                 }
4432                 if (len == 0)
4433                         return fail;
4434                 else {
4435                         max_match.pos_len = len;
4436                         displayMres(max_match, "SEARCH RESULT", cur)
4437                                         return max_match;
4438                 }
4439         }
4440 }
4441
4442 /// Finds forward
4443 int findForwardAdv(DocIterator & cur, MatchStringAdv & match)
4444 {
4445         if (!cur)
4446                 return 0;
4447         bool repeat = false;
4448         DocIterator orig_cur;   // to be used if repeat not successful
4449         MatchResult orig_mres;
4450         while (!theApp()->longOperationCancelled() && cur) {
4451                 //(void) findAdvForwardInnermost(cur);
4452                 LYXERR(Debug::FINDVERBOSE, "findForwardAdv() cur: " << cur);
4453                 MatchResult mres = match(cur, -1, MatchStringAdv::MatchAnyPlace);
4454                 string msg = "Starting";
4455                 if (repeat)
4456                         msg = "Repeated";
4457                 displayMres(mres, msg + " findForwardAdv", cur)
4458                                 int match_len = mres.match_len;
4459                 if ((mres.pos > 100000) || (mres.match2end > 100000) || (match_len > 100000)) {
4460                         LYXERR(Debug::INFO, "BIG LENGTHS: " << mres.pos << ", " << match_len << ", " << mres.match2end);
4461                         match_len = 0;
4462                 }
4463                 if (match_len <= 0) {
4464                         // This should exit nested insets, if any, or otherwise undefine the currsor.
4465                         cur.pos() = cur.lastpos();
4466                         LYXERR(Debug::FINDVERBOSE, "Advancing pos: cur=" << cur);
4467                         cur.forwardPos();
4468                 }
4469                 else {  // match_len > 0
4470                         // Try to find the begin of searched string
4471                         int increment;
4472                         int firstInvalid = cur.lastpos() - cur.pos();
4473                         {
4474                                 int incrmatch = (mres.match_prefix + mres.pos - mres.leadsize + 1)*3/4;
4475                                 int incrcur = (firstInvalid + 1 )*3/4;
4476                                 if (incrcur < incrmatch)
4477                                         increment = incrcur;
4478                                 else
4479                                         increment = incrmatch;
4480                                 if (increment < 1)
4481                                         increment = 1;
4482                         }
4483                         LYXERR(Debug::FINDVERBOSE, "Set increment to " << increment);
4484                         while (increment > 0) {
4485                                 DocIterator old_cur = cur;
4486                                 if (cur.pos() + increment >= cur.lastpos()) {
4487                                         increment /= 2;
4488                                         continue;
4489                                 }
4490                                 cur.pos() = cur.pos() + increment;
4491                                 MatchResult mres2 = match(cur, -1, MatchStringAdv::MatchAnyPlace);
4492                                 displayMres(mres2, "findForwardAdv loop", cur)
4493                                                 switch (interpretMatch(mres, mres2)) {
4494                                         case MatchResult::newIsTooFar:
4495                                                 // behind the expected match
4496                                                 firstInvalid = increment;
4497                                                 cur = old_cur;
4498                                                 increment /= 2;
4499                                                 break;
4500                                         case MatchResult::newIsBetter:
4501                                                 // not reached yet, but cur.pos()+increment is bettert
4502                                                 mres = mres2;
4503                                                 firstInvalid -= increment;
4504                                                 if (increment > firstInvalid*3/4)
4505                                                         increment = firstInvalid*3/4;
4506                                                 if ((mres2.pos == mres2.leadsize) && (increment >= mres2.match_prefix)) {
4507                                                         if (increment >= mres2.match_prefix)
4508                                                                 increment = (mres2.match_prefix+1)*3/4;
4509                                                 }
4510                                                 break;
4511                                         default:
4512                                                 // Todo@
4513                                                 // Handle not like MatchResult::newIsTooFar
4514                                                 LYXERR(Debug::FINDVERBOSE, "Probably too far: Increment = " << increment << " match_prefix = " << mres.match_prefix);
4515                                                 firstInvalid--;
4516                                                 increment = increment*3/4;
4517                                                 cur = old_cur;
4518                                                 break;
4519                                 }
4520                         }
4521                         if (mres.match_len > 0) {
4522                                 if (mres.match_prefix + mres.pos - mres.leadsize > 0) {
4523                                         // The match seems to indicate some deeper level
4524                                         repeat = true;
4525                                         orig_cur = cur;
4526                                         orig_mres = mres;
4527                                         cur.forwardPos();
4528                                         continue;
4529                                 }
4530                         }
4531                         else if (repeat) {
4532                                 // should never be reached.
4533                                 cur = orig_cur;
4534                                 mres = orig_mres;
4535                         }
4536                         // LYXERR0("Leaving first loop");
4537                         LYXERR(Debug::FINDVERBOSE, "Finalizing 1");
4538                         MatchResult found_match = findAdvFinalize(cur, match, mres);
4539                         if (found_match.match_len > 0) {
4540                                 match.FillResults(found_match);
4541                                 return found_match.pos_len;
4542                         }
4543                         else {
4544                                 // try next possible match
4545                                 cur.forwardPos();
4546                                 repeat = false;
4547                                 continue;
4548                         }
4549                 }
4550         }
4551         return 0;
4552 }
4553
4554
4555 /// Find the most backward consecutive match within same paragraph while searching backwards.
4556 MatchResult findMostBackwards(DocIterator & cur, MatchStringAdv const & match, MatchResult &expected)
4557 {
4558         DocIterator cur_begin = cur;
4559         cur_begin.pos() = 0;
4560         DocIterator tmp_cur = cur;
4561         MatchResult mr = findAdvFinalize(tmp_cur, match, expected);
4562         Inset & inset = cur.inset();
4563         for (; cur != cur_begin; cur.backwardPos()) {
4564                 LYXERR(Debug::FINDVERBOSE, "findMostBackwards(): cur=" << cur);
4565                 DocIterator new_cur = cur;
4566                 new_cur.backwardPos();
4567                 if (new_cur == cur || &new_cur.inset() != &inset
4568                     || match(new_cur, -1, MatchStringAdv::MatchFromStart).match_len <= 0)
4569                         break;
4570                 MatchResult new_mr = findAdvFinalize(new_cur, match, expected);
4571                 if (new_mr.match_len == mr.match_len)
4572                         break;
4573                 mr = new_mr;
4574         }
4575         LYXERR(Debug::FINDVERBOSE, "findMostBackwards(): exiting with cur=" << cur);
4576         return mr;
4577 }
4578
4579
4580 /// Finds backwards
4581 int findBackwardsAdv(DocIterator & cur, MatchStringAdv & match)
4582 {
4583         if (! cur)
4584                 return 0;
4585         // Backup of original position
4586         DocIterator cur_begin = doc_iterator_begin(cur.buffer());
4587         if (cur == cur_begin)
4588                 return 0;
4589         cur.backwardPos();
4590         DocIterator cur_orig(cur);
4591         bool pit_changed = false;
4592         do {
4593                 cur.pos() = 0;
4594                 MatchResult found_match = match(cur, -1, MatchStringAdv::MatchAnyPlace);
4595
4596                 if (found_match.match_len > 0) {
4597                         if (pit_changed)
4598                                 cur.pos() = cur.lastpos();
4599                         else
4600                                 cur.pos() = cur_orig.pos();
4601                         LYXERR(Debug::FINDVERBOSE, "findBackAdv2: cur: " << cur);
4602                         DocIterator cur_prev_iter;
4603                         do {
4604                                 found_match = match(cur, -1, MatchStringAdv::MatchFromStart);
4605                                 LYXERR(Debug::FINDVERBOSE, "findBackAdv3: found_match="
4606                                        << (found_match.match_len > 0) << ", cur: " << cur);
4607                                 if (found_match.match_len > 0) {
4608                                         MatchResult found_mr = findMostBackwards(cur, match, found_match);
4609                                         if (found_mr.pos_len > 0) {
4610                                                 match.FillResults(found_mr);
4611                                                 return found_mr.pos_len;
4612                                         }
4613                                 }
4614
4615                                 // Stop if begin of document reached
4616                                 if (cur == cur_begin)
4617                                         break;
4618                                 cur_prev_iter = cur;
4619                                 cur.backwardPos();
4620                         } while (true);
4621                 }
4622                 if (cur == cur_begin)
4623                         break;
4624                 if (cur.pit() > 0)
4625                         --cur.pit();
4626                 else
4627                         cur.backwardPos();
4628                 pit_changed = true;
4629         } while (!theApp()->longOperationCancelled());
4630         return 0;
4631 }
4632
4633
4634 } // namespace
4635
4636
4637 docstring stringifyFromForSearch(FindAndReplaceOptions const & opt,
4638                                  DocIterator const & cur, int len)
4639 {
4640         if (cur.pos() < 0 || cur.pos() > cur.lastpos())
4641                 return docstring();
4642         if (!opt.ignoreformat)
4643                 return latexifyFromCursor(cur, len);
4644         else
4645                 return stringifyFromCursor(cur, len);
4646 }
4647
4648
4649 FindAndReplaceOptions::FindAndReplaceOptions(
4650                 docstring const & _find_buf_name, bool _casesensitive,
4651                 bool _matchword, bool _forward, bool _expandmacros, bool _ignoreformat,
4652                 docstring const & _repl_buf_name, bool _keep_case,
4653                 SearchScope _scope, SearchRestriction _restr, bool _replace_all)
4654         : find_buf_name(_find_buf_name), casesensitive(_casesensitive), matchword(_matchword),
4655           forward(_forward), expandmacros(_expandmacros), ignoreformat(_ignoreformat),
4656           repl_buf_name(_repl_buf_name), keep_case(_keep_case), scope(_scope), restr(_restr), replace_all(_replace_all)
4657 {
4658 }
4659
4660
4661 namespace {
4662
4663
4664 /** Check if 'len' letters following cursor are all non-lowercase */
4665 static bool allNonLowercase(Cursor const & cur, int len)
4666 {
4667         pos_type beg_pos = cur.selectionBegin().pos();
4668         pos_type end_pos = cur.selectionBegin().pos() + len;
4669         if (len > cur.lastpos() + 1 - beg_pos) {
4670                 LYXERR(Debug::FINDVERBOSE, "This should not happen, more debug needed");
4671                 len = cur.lastpos() + 1 - beg_pos;
4672                 end_pos = beg_pos + len;
4673         }
4674         for (pos_type pos = beg_pos; pos != end_pos; ++pos)
4675                 if (isLowerCase(cur.paragraph().getChar(pos)))
4676                         return false;
4677         return true;
4678 }
4679
4680
4681 /** Check if first letter is upper case and second one is lower case */
4682 static bool firstUppercase(Cursor const & cur)
4683 {
4684         char_type ch1, ch2;
4685         pos_type pos = cur.selectionBegin().pos();
4686         if (pos >= cur.lastpos() - 1) {
4687                 LYXERR(Debug::FINDVERBOSE, "No upper-case at cur: " << cur);
4688                 return false;
4689         }
4690         ch1 = cur.paragraph().getChar(pos);
4691         ch2 = cur.paragraph().getChar(pos + 1);
4692         bool result = isUpperCase(ch1) && isLowerCase(ch2);
4693         LYXERR(Debug::FINDVERBOSE, "firstUppercase(): "
4694                << "ch1=" << ch1 << "(" << char(ch1) << "), ch2="
4695                << ch2 << "(" << char(ch2) << ")"
4696                << ", result=" << result << ", cur=" << cur);
4697         return result;
4698 }
4699
4700
4701 /** Make first letter of supplied buffer upper-case, and the rest lower-case.
4702  **
4703  ** \fixme What to do with possible further paragraphs in replace buffer ?
4704  **/
4705 static void changeFirstCase(Buffer & buffer, TextCase first_case, TextCase others_case)
4706 {
4707         ParagraphList::iterator pit = buffer.paragraphs().begin();
4708         LASSERT(!pit->empty(), /**/);
4709         pos_type right = pos_type(1);
4710         pit->changeCase(buffer.params(), pos_type(0), right, first_case);
4711         right = pit->size();
4712         pit->changeCase(buffer.params(), pos_type(1), right, others_case);
4713 }
4714 } // namespace
4715
4716 static bool replaceMatches(string &t, int maxmatchnum, vector <string> const & replacements)
4717 {
4718         // Should replace the string "$" + std::to_string(matchnum) with replacement
4719         // if the char '$' is not prefixed with odd number of char '\\'
4720         static regex const rematch("(\\\\)*(\\$\\$([0-9]))");
4721         string s;
4722         size_t lastpos = 0;
4723         smatch sub;
4724         for (sregex_iterator it(t.begin(), t.end(), rematch), end; it != end; ++it) {
4725                 sub = *it;
4726                 if ((sub.position(2) - sub.position(0)) % 2 == 1)
4727                         continue;
4728                 int num = stoi(sub.str(3), nullptr, 10);
4729                 if (num >= maxmatchnum)
4730                         continue;
4731                 if (lastpos < (size_t) sub.position(2))
4732                         s += t.substr(lastpos, sub.position(2) - lastpos);
4733                 s += replacements[num];
4734                 lastpos = sub.position(2) + sub.length(2);
4735         }
4736         if (lastpos == 0)
4737                 return false;
4738         else if (lastpos < t.length())
4739                 s += t.substr(lastpos, t.length() - lastpos);
4740         t = s;
4741         return true;
4742 }
4743
4744 ///
4745 static int findAdvReplace(BufferView * bv, FindAndReplaceOptions const & opt, MatchStringAdv & matchAdv)
4746 {
4747         Cursor & cur = bv->cursor();
4748         if (opt.repl_buf_name.empty()
4749                         || theBufferList().getBuffer(FileName(to_utf8(opt.repl_buf_name)), true) == 0
4750                         || theBufferList().getBuffer(FileName(to_utf8(opt.find_buf_name)), true) == 0)
4751                 return 0;
4752
4753         DocIterator sel_beg = cur.selectionBegin();
4754         DocIterator sel_end = cur.selectionEnd();
4755         if (&sel_beg.inset() != &sel_end.inset()
4756                         || sel_beg.pit() != sel_end.pit()
4757                         || sel_beg.idx() != sel_end.idx())
4758                 return 0;
4759         int sel_len = sel_end.pos() - sel_beg.pos();
4760         LYXERR(Debug::FINDVERBOSE, "sel_beg: " << sel_beg << ", sel_end: " << sel_end
4761                << ", sel_len: " << sel_len << endl);
4762         if (sel_len == 0)
4763                 return 0;
4764         LASSERT(sel_len > 0, return 0);
4765
4766         if (matchAdv(sel_beg, sel_len, MatchStringAdv::MatchFromStart).match_len <= 0)
4767                 return 0;
4768
4769         // Build a copy of the replace buffer, adapted to the KeepCase option
4770         Buffer const & repl_buffer_orig = *theBufferList().getBuffer(FileName(to_utf8(opt.repl_buf_name)), true);
4771         ostringstream oss;
4772         repl_buffer_orig.write(oss);
4773         string lyx = oss.str();
4774         if (matchAdv.valid_matches > 0)
4775                 replaceMatches(lyx, matchAdv.valid_matches, matchAdv.matches);
4776         Buffer repl_buffer(string(), false);
4777         repl_buffer.setInternal(true);
4778         repl_buffer.setUnnamed(true);
4779         LASSERT(repl_buffer.readString(lyx), return 0);
4780         if (opt.keep_case && sel_len >= 2) {
4781                 LYXERR(Debug::FINDVERBOSE, "keep_case true: cur.pos()=" << cur.pos() << ", sel_len=" << sel_len);
4782                 if (cur.inTexted()) {
4783                         if (firstUppercase(cur))
4784                                 changeFirstCase(repl_buffer, text_uppercase, text_lowercase);
4785                         else if (allNonLowercase(cur, sel_len))
4786                                 changeFirstCase(repl_buffer, text_uppercase, text_uppercase);
4787                 }
4788         }
4789         cap::cutSelection(cur, false);
4790         if (cur.inTexted()) {
4791                 repl_buffer.changeLanguage(
4792                                         repl_buffer.language(),
4793                                         cur.getFont().language());
4794                 LYXERR(Debug::FINDVERBOSE, "Replacing by pasteParagraphList()ing repl_buffer");
4795                 LYXERR(Debug::FINDVERBOSE, "Before pasteParagraphList() cur=" << cur << endl);
4796                 cap::pasteParagraphList(cur, repl_buffer.paragraphs(),
4797                                         repl_buffer.params().documentClassPtr(),
4798                                         repl_buffer.params().authors(),
4799                                         bv->buffer().errorList("Paste"));
4800                 LYXERR(Debug::FINDVERBOSE, "After pasteParagraphList() cur=" << cur << endl);
4801                 sel_len = repl_buffer.paragraphs().begin()->size();
4802         } else if (cur.inMathed()) {
4803                 odocstringstream ods;
4804                 otexstream os(ods);
4805                 // OutputParams runparams(&repl_buffer.params().encoding());
4806                 OutputParams runparams(encodings.fromLyXName("utf8"));
4807                 runparams.nice = false;
4808                 setFindParams(runparams);
4809                 TeXOnePar(repl_buffer, repl_buffer.text(), 0, os, runparams, string(), -1, -1, true);
4810                 //repl_buffer.getSourceCode(ods, 0, repl_buffer.paragraphs().size(), false);
4811                 docstring repl_latex = ods.str();
4812                 LYXERR(Debug::FINDVERBOSE, "Latexified replace_buffer: '" << repl_latex << "'");
4813                 string s;
4814                 (void)regex_replace(to_utf8(repl_latex), s, "\\$(.*)\\$", "$1");
4815                 (void)regex_replace(s, s, "\\\\\\[(.*)\\\\\\]", "$1");
4816                 repl_latex = from_utf8(s);
4817                 LYXERR(Debug::FINDVERBOSE, "Replacing by insert()ing latex: '" << repl_latex << "' cur=" << cur << " with depth=" << cur.depth());
4818                 MathData ar(cur.buffer());
4819                 asArray(repl_latex, ar, Parse::NORMAL);
4820                 cur.insert(ar);
4821                 sel_len = ar.size();
4822                 LYXERR(Debug::FINDVERBOSE, "After insert() cur=" << cur << " with depth: " << cur.depth() << " and len: " << sel_len);
4823         }
4824         if (cur.pos() >= sel_len)
4825                 cur.pos() -= sel_len;
4826         else
4827                 cur.pos() = 0;
4828         LYXERR(Debug::FINDVERBOSE, "After pos adj cur=" << cur << " with depth: " << cur.depth() << " and len: " << sel_len);
4829         bv->putSelectionAt(DocIterator(cur), sel_len, !opt.forward);
4830         bv->processUpdateFlags(Update::Force);
4831         return 1;
4832 }
4833
4834 static bool isWordChar(char_type c)
4835 {
4836         return isLetterChar(c) || isDigitASCII(c);
4837 }
4838
4839 /// Perform a FindAdv operation.
4840 bool findAdv(BufferView * bv, FindAndReplaceOptions & opt)
4841 {
4842         DocIterator cur;
4843         int pos_len = 0;
4844
4845         // e.g., when invoking word-findadv from mini-buffer wither with
4846         //       wrong options syntax or before ever opening advanced F&R pane
4847         if (theBufferList().getBuffer(FileName(to_utf8(opt.find_buf_name)), true) == 0)
4848                 return false;
4849
4850         try {
4851                 MatchStringAdv matchAdv(bv->buffer(), opt);
4852 #if QTSEARCH
4853                 if (!matchAdv.regexIsValid) {
4854                         bv->message(lyx::from_utf8(matchAdv.regexError));
4855                         return(false);
4856                 }
4857 #endif
4858                 int length = bv->cursor().selectionEnd().pos() - bv->cursor().selectionBegin().pos();
4859                 if (length > 0)
4860                         bv->putSelectionAt(bv->cursor().selectionBegin(), length, !opt.forward);
4861                 num_replaced += findAdvReplace(bv, opt, matchAdv);
4862                 cur = bv->cursor();
4863                 if (opt.forward) {
4864                         if (opt.matchword && cur.pos() > 0) {  // Skip word-characters if we are in the mid of a word
4865                                 if (cur.inTexted()) {
4866                                         Paragraph const & par = cur.paragraph();
4867                                         if (!par.isWordSeparator(cur.pos() -1, true)) {
4868                                                 class Text *t = cur.text();
4869                                                 CursorSlice to;
4870                                                 CursorSlice from = cur.top();
4871                                                 t->getWord(from, to, WHOLE_WORD);
4872                                                 cur.pos() = to.pos();
4873                                                 cur.pit() = to.pit();
4874                                         }
4875                                 }
4876                                 else if (cur.inMathed()) {
4877                                         // Check if 'cur.pos()-1' and 'cur.pos()' both point to a letter,
4878                                         // I am not sure, we should consider the selection
4879                                         bool sel = bv->cursor().selection();
4880                                         if (!sel && cur.pos() < cur.lastpos()) {
4881                                                 CursorSlice const & cs = cur.top();
4882                                                 MathData md = cs.cell();
4883                                                 int len = -1;
4884                                                 MathData::const_iterator it_end = md.end();
4885                                                 MathData md2;
4886                                                 // Start the check with one character before actual cursor position
4887                                                 for (MathData::const_iterator it = md.begin() + cs.pos() - 1;
4888                                                     it != it_end; ++it)
4889                                                         md2.push_back(*it);
4890                                                 docstring inp = asString(md2);
4891                                                 LYXERR0("Got \"" << inp << "\"");
4892                                                 char_type prev = inp[0];
4893                                                 for (len = 0; (unsigned) len < inp.size() && len + cur.pos() <= cur.lastpos(); len++) {
4894                                                         char_type c = inp[len];
4895                                                         if (isLetterChar(c))
4896                                                                 continue;
4897                                                         if (isDigitASCII(c))
4898                                                                 continue;
4899                                                         break;
4900                                                 }
4901                                                 // len == 0 means previous char was a word separator
4902                                                 // len == 1       search starts with a word separator
4903                                                 // len == 2 ...   we have to skip len -1 chars
4904                                                 if (len > 1)
4905                                                         cur.pos() = cur.pos() + len - 1;
4906                                         }
4907
4908                                 }
4909                                 opt.matchword = false;
4910                         }
4911                         pos_len = findForwardAdv(cur, matchAdv);
4912                 }
4913                 else
4914                         pos_len = findBackwardsAdv(cur, matchAdv);
4915         } catch (exception & ex) {
4916                 bv->message(from_utf8(ex.what()));
4917                 return false;
4918         }
4919
4920         if (pos_len == 0) {
4921                 if (num_replaced > 0) {
4922                         switch (num_replaced)
4923                         {
4924                         case 1:
4925                                 bv->message(_("One match has been replaced."));
4926                                 break;
4927                         case 2:
4928                                 bv->message(_("Two matches have been replaced."));
4929                                 break;
4930                         default:
4931                                 bv->message(bformat(_("%1$d matches have been replaced."), num_replaced));
4932                                 break;
4933                         }
4934                         num_replaced = 0;
4935                 }
4936                 else {
4937                         bv->message(_("Match not found."));
4938                 }
4939                 return false;
4940         }
4941
4942         if (num_replaced > 0)
4943                 bv->message(_("Match has been replaced."));
4944         else
4945                 bv->message(_("Match found."));
4946
4947         if (cur.pos() + pos_len > cur.lastpos()) {
4948                 // Prevent crash in bv->putSelectionAt()
4949                 // Should never happen, maybe LASSERT() here?
4950                 pos_len = cur.lastpos() - cur.pos();
4951         }
4952         LYXERR(Debug::FINDVERBOSE|Debug::FIND, "Putting selection at cur=" << cur << " with len: " << pos_len);
4953         bv->putSelectionAt(cur, pos_len, !opt.forward);
4954
4955         return true;
4956 }
4957
4958
4959 ostringstream & operator<<(ostringstream & os, FindAndReplaceOptions const & opt)
4960 {
4961         os << to_utf8(opt.find_buf_name) << "\nEOSS\n"
4962            << opt.casesensitive << ' '
4963            << opt.matchword << ' '
4964            << opt.forward << ' '
4965            << opt.expandmacros << ' '
4966            << opt.ignoreformat << ' '
4967            << opt.replace_all << ' '
4968            << to_utf8(opt.repl_buf_name) << "\nEOSS\n"
4969            << opt.keep_case << ' '
4970            << int(opt.scope) << ' '
4971            << int(opt.restr);
4972
4973         LYXERR(Debug::FINDVERBOSE, "built: " << os.str());
4974
4975         return os;
4976 }
4977
4978
4979 istringstream & operator>>(istringstream & is, FindAndReplaceOptions & opt)
4980 {
4981         // LYXERR(Debug::FINDVERBOSE, "parsing");
4982         string s;
4983         string line;
4984         getline(is, line);
4985         while (line != "EOSS") {
4986                 if (! s.empty())
4987                         s = s + "\n";
4988                 s = s + line;
4989                 if (is.eof())   // Tolerate malformed request
4990                         break;
4991                 getline(is, line);
4992         }
4993         // LYXERR(Debug::FINDVERBOSE, "file_buf_name: '" << s << "'");
4994         opt.find_buf_name = from_utf8(s);
4995         is >> opt.casesensitive >> opt.matchword >> opt.forward >> opt.expandmacros >> opt.ignoreformat >> opt.replace_all;
4996         is.get();       // Waste space before replace string
4997         s = "";
4998         getline(is, line);
4999         while (line != "EOSS") {
5000                 if (! s.empty())
5001                         s = s + "\n";
5002                 s = s + line;
5003                 if (is.eof())   // Tolerate malformed request
5004                         break;
5005                 getline(is, line);
5006         }
5007         // LYXERR(Debug::FINDVERBOSE, "repl_buf_name: '" << s << "'");
5008         opt.repl_buf_name = from_utf8(s);
5009         is >> opt.keep_case;
5010         int i;
5011         is >> i;
5012         opt.scope = FindAndReplaceOptions::SearchScope(i);
5013         is >> i;
5014         opt.restr = FindAndReplaceOptions::SearchRestriction(i);
5015
5016         /*
5017         LYXERR(Debug::FINDVERBOSE, "parsed: " << opt.casesensitive << ' ' << opt.matchword << ' ' << opt.forward << ' '
5018                << opt.expandmacros << ' ' << opt.ignoreformat << ' ' << opt.keep_case << ' '
5019                << opt.scope << ' ' << opt.restr);
5020         */
5021         return is;
5022 }
5023
5024 } // namespace lyx