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