src/Encoding.cpp

   1 /**
   2  * \file Encoding.cpp
   3  * This file is part of LyX, the document processor.
   4  * Licence details can be found in the file COPYING.
   5  *
   6  * \author Lars Gullik Bjønnes
   7  * \author Jean-Marc Lasgouttes
   8  * \author Dekel Tsur
   9  *
  10  * Full author contact details are available in file CREDITS.
  11  */
  12
  13 #include <config.h>
  14
  15 #include "Encoding.h"
  16
  17 #include "Buffer.h"
  18 #include "BufferList.h"
  19 #include "InsetIterator.h"
  20 #include "LaTeXFeatures.h"
  21 #include "Lexer.h"
  22 #include "LyXRC.h"
  23
  24 #include "support/debug.h"
  25 #include "support/gettext.h"
  26 #include "support/FileName.h"
  27 #include "support/lstrings.h"
  28 #include "support/textutils.h"
  29 #include "support/unicode.h"
  30
  31 #include <boost/cstdint.hpp>
  32
  33 #include <sstream>
  34
  35 using namespace std;
  36 using namespace lyx::support;
  37
  38 namespace lyx {
  39
  40 Encodings encodings;
  41
  42 Encodings::MathCommandSet Encodings::mathcmd;
  43 Encodings::TextCommandSet Encodings::textcmd;
  44 Encodings::MathSymbolSet  Encodings::mathsym;
  45
  46 namespace {
  47
  48 char_type arabic_table[172][4] = {
  49         {0xfe80, 0xfe80, 0xfe80, 0xfe80}, // 0x0621 = hamza
  50         {0xfe81, 0xfe82, 0xfe81, 0xfe82}, // 0x0622 = ligature madda on alef
  51         {0xfe83, 0xfe84, 0xfe83, 0xfe84}, // 0x0623 = ligature hamza on alef
  52         {0xfe85, 0xfe86, 0xfe85, 0xfe86}, // 0x0624 = ligature hamza on waw
  53         {0xfe87, 0xfe88, 0xfe87, 0xfe88}, // 0x0625 = ligature hamza under alef
  54         {0xfe89, 0xfe8a, 0xfe8b, 0xfe8c}, // 0x0626 = ligature hamza on ya
  55         {0xfe8d, 0xfe8e, 0xfe8d, 0xfe8e}, // 0x0627 = alef
  56         {0xfe8f, 0xfe90, 0xfe91, 0xfe92}, // 0x0628 = baa
  57         {0xfe93, 0xfe94, 0xfe93, 0xfe94}, // 0x0629 = taa marbuta
  58         {0xfe95, 0xfe96, 0xfe97, 0xfe98}, // 0x062a = taa
  59         {0xfe99, 0xfe9a, 0xfe9b, 0xfe9c}, // 0x062b = thaa
  60         {0xfe9d, 0xfe9e, 0xfe9f, 0xfea0}, // 0x062c = jeem
  61         {0xfea1, 0xfea2, 0xfea3, 0xfea4}, // 0x062d = haa
  62         {0xfea5, 0xfea6, 0xfea7, 0xfea8}, // 0x062e = khaa
  63         {0xfea9, 0xfeaa, 0xfea9, 0xfeaa}, // 0x062f = dal
  64
  65         {0xfeab, 0xfeac, 0xfeab, 0xfeac}, // 0x0630 = thal
  66         {0xfead, 0xfeae, 0xfead, 0xfeae}, // 0x0631 = ra
  67         {0xfeaf, 0xfeb0, 0xfeaf, 0xfeb0}, // 0x0632 = zain
  68         {0xfeb1, 0xfeb2, 0xfeb3, 0xfeb4}, // 0x0633 = seen
  69         {0xfeb5, 0xfeb6, 0xfeb7, 0xfeb8}, // 0x0634 = sheen
  70         {0xfeb9, 0xfeba, 0xfebb, 0xfebc}, // 0x0635 = sad
  71         {0xfebd, 0xfebe, 0xfebf, 0xfec0}, // 0x0636 = dad
  72         {0xfec1, 0xfec2, 0xfec3, 0xfec4}, // 0x0637 = tah
  73         {0xfec5, 0xfec6, 0xfec7, 0xfec8}, // 0x0638 = zah
  74         {0xfec9, 0xfeca, 0xfecb, 0xfecc}, // 0x0639 = ain
  75         {0xfecd, 0xfece, 0xfecf, 0xfed0}, // 0x063a = ghain
  76         {0, 0, 0, 0}, // 0x063b
  77         {0, 0, 0, 0}, // 0x063c
  78         {0, 0, 0, 0}, // 0x063d
  79         {0, 0, 0, 0}, // 0x063e
  80         {0, 0, 0, 0}, // 0x063f
  81
  82         {0, 0, 0, 0}, // 0x0640
  83         {0xfed1, 0xfed2, 0xfed3, 0xfed4}, // 0x0641 = fa
  84         {0xfed5, 0xfed6, 0xfed7, 0xfed8}, // 0x0642 = qaf
  85         {0xfed9, 0xfeda, 0xfedb, 0xfedc}, // 0x0643 = kaf
  86         {0xfedd, 0xfede, 0xfedf, 0xfee0}, // 0x0644 = lam
  87         {0xfee1, 0xfee2, 0xfee3, 0xfee4}, // 0x0645 = meem
  88         {0xfee5, 0xfee6, 0xfee7, 0xfee8}, // 0x0646 = noon
  89         {0xfee9, 0xfeea, 0xfeeb, 0xfeec}, // 0x0647 = ha
  90         {0xfeed, 0xfeee, 0xfeed, 0xfeee}, // 0x0648 = waw
  91         {0xfeef, 0xfef0, 0xfeef, 0xfef0}, // 0x0649 = alef maksura
  92         {0xfef1, 0xfef2, 0xfef3, 0xfef4}, // 0x064a = ya
  93         {0x065b, 0x065b, 0x065b, 0x065b}, // 0x064b = fathatan
  94         {0x065c, 0x065c, 0x065c, 0x065c}, // 0x064c = dammatan
  95         {0x064d, 0x064d, 0x064d, 0x064d}, // 0x064d = kasratan
  96         {0x064e, 0x064e, 0x064e, 0x064e}, // 0x064e = fatha
  97         {0x064f, 0x064f, 0x064f, 0x064f}, // 0x064f = damma
  98
  99         {0x0650, 0x0650, 0x0650, 0x0650}, // 0x0650 = kasra
 100         {0x0651, 0x0651, 0x0651, 0x0651}, // 0x0651 = shadda
 101         {0x0652, 0x0652, 0x0652, 0x0652}, // 0x0652 = sukun
 102
 103         {0, 0, 0, 0}, // 0x0653
 104         {0, 0, 0, 0}, // 0x0654
 105         {0, 0, 0, 0}, // 0x0655
 106         {0, 0, 0, 0}, // 0x0656
 107         {0, 0, 0, 0}, // 0x0657
 108         {0, 0, 0, 0}, // 0x0658
 109         {0, 0, 0, 0}, // 0x0659
 110         {0, 0, 0, 0}, // 0x065a
 111         {0, 0, 0, 0}, // 0x065b
 112         {0, 0, 0, 0}, // 0x065c
 113         {0, 0, 0, 0}, // 0x065d
 114         {0, 0, 0, 0}, // 0x065e
 115         {0, 0, 0, 0}, // 0x065f
 116         {0, 0, 0, 0}, // 0x0660
 117         {0, 0, 0, 0}, // 0x0661
 118         {0, 0, 0, 0}, // 0x0662
 119         {0, 0, 0, 0}, // 0x0663
 120         {0, 0, 0, 0}, // 0x0664
 121         {0, 0, 0, 0}, // 0x0665
 122         {0, 0, 0, 0}, // 0x0666
 123         {0, 0, 0, 0}, // 0x0667
 124         {0, 0, 0, 0}, // 0x0668
 125         {0, 0, 0, 0}, // 0x0669
 126         {0, 0, 0, 0}, // 0x066a
 127         {0, 0, 0, 0}, // 0x066b
 128         {0, 0, 0, 0}, // 0x066c
 129         {0, 0, 0, 0}, // 0x066d
 130         {0, 0, 0, 0}, // 0x066e
 131         {0, 0, 0, 0}, // 0x066f
 132         {0, 0, 0, 0}, // 0x0670
 133         {0, 0, 0, 0}, // 0x0671
 134         {0, 0, 0, 0}, // 0x0672
 135         {0, 0, 0, 0}, // 0x0673
 136         {0, 0, 0, 0}, // 0x0674
 137         {0, 0, 0, 0}, // 0x0675
 138         {0, 0, 0, 0}, // 0x0676
 139         {0, 0, 0, 0}, // 0x0677
 140         {0, 0, 0, 0}, // 0x0678
 141         {0, 0, 0, 0}, // 0x0679
 142         {0, 0, 0, 0}, // 0x067a
 143         {0, 0, 0, 0}, // 0x067b
 144         {0, 0, 0, 0}, // 0x067c
 145         {0, 0, 0, 0}, // 0x067d
 146         {0xfb56, 0xfb57, 0xfb58, 0xfb59}, // 0x067e = peh
 147         {0, 0, 0, 0}, // 0x067f
 148         {0, 0, 0, 0}, // 0x0680
 149         {0, 0, 0, 0}, // 0x0681
 150         {0, 0, 0, 0}, // 0x0682
 151         {0, 0, 0, 0}, // 0x0683
 152         {0, 0, 0, 0}, // 0x0684
 153         {0, 0, 0, 0}, // 0x0685
 154         {0xfb7a, 0xfb7b, 0xfb7c, 0xfb7d}, // 0x0686 = tcheh
 155         {0, 0, 0, 0}, // 0x0687
 156         {0, 0, 0, 0}, // 0x0688
 157         {0, 0, 0, 0}, // 0x0689
 158         {0, 0, 0, 0}, // 0x068a
 159         {0, 0, 0, 0}, // 0x068b
 160         {0, 0, 0, 0}, // 0x068c
 161         {0, 0, 0, 0}, // 0x068d
 162         {0, 0, 0, 0}, // 0x068e
 163         {0, 0, 0, 0}, // 0x068f
 164         {0, 0, 0, 0}, // 0x0690
 165         {0, 0, 0, 0}, // 0x0691
 166         {0, 0, 0, 0}, // 0x0692
 167         {0, 0, 0, 0}, // 0x0693
 168         {0, 0, 0, 0}, // 0x0694
 169         {0, 0, 0, 0}, // 0x0695
 170         {0, 0, 0, 0}, // 0x0696
 171         {0, 0, 0, 0}, // 0x0697
 172         {0xfb8a, 0xfb8b, 0xfb8a, 0xfb8b}, // 0x0698 = jeh
 173         {0, 0, 0, 0}, // 0x0699
 174         {0, 0, 0, 0}, // 0x069a
 175         {0, 0, 0, 0}, // 0x069b
 176         {0, 0, 0, 0}, // 0x069c
 177         {0, 0, 0, 0}, // 0x069d
 178         {0, 0, 0, 0}, // 0x069e
 179         {0, 0, 0, 0}, // 0x069f
 180         {0, 0, 0, 0}, // 0x06a0
 181         {0, 0, 0, 0}, // 0x06a1
 182         {0, 0, 0, 0}, // 0x06a2
 183         {0, 0, 0, 0}, // 0x06a3
 184         {0, 0, 0, 0}, // 0x06a4
 185         {0, 0, 0, 0}, // 0x06a5
 186         {0, 0, 0, 0}, // 0x06a6
 187         {0, 0, 0, 0}, // 0x06a7
 188         {0, 0, 0, 0}, // 0x06a8
 189         {0xfb8e, 0xfb8f, 0xfb90, 0xfb91}, // 0x06a9 = farsi kaf
 190         {0, 0, 0, 0}, // 0x06aa
 191         {0, 0, 0, 0}, // 0x06ab
 192         {0, 0, 0, 0}, // 0x06ac
 193         {0, 0, 0, 0}, // 0x06ad
 194         {0, 0, 0, 0}, // 0x06ae
 195         {0xfb92, 0xfb93, 0xfb94, 0xfb95}, // 0x06af = gaf
 196         {0, 0, 0, 0}, // 0x06b0
 197         {0, 0, 0, 0}, // 0x06b1
 198         {0, 0, 0, 0}, // 0x06b2
 199         {0, 0, 0, 0}, // 0x06b3
 200         {0, 0, 0, 0}, // 0x06b4
 201         {0, 0, 0, 0}, // 0x06b5
 202         {0, 0, 0, 0}, // 0x06b6
 203         {0, 0, 0, 0}, // 0x06b7
 204         {0, 0, 0, 0}, // 0x06b8
 205         {0, 0, 0, 0}, // 0x06b9
 206         {0, 0, 0, 0}, // 0x06ba
 207         {0, 0, 0, 0}, // 0x06bb
 208         {0, 0, 0, 0}, // 0x06bc
 209         {0, 0, 0, 0}, // 0x06bd
 210         {0, 0, 0, 0}, // 0x06be
 211         {0, 0, 0, 0}, // 0x06bf
 212         {0, 0, 0, 0}, // 0x06c0
 213         {0, 0, 0, 0}, // 0x06c1
 214         {0, 0, 0, 0}, // 0x06c2
 215         {0, 0, 0, 0}, // 0x06c3
 216         {0, 0, 0, 0}, // 0x06c4
 217         {0, 0, 0, 0}, // 0x06c5
 218         {0, 0, 0, 0}, // 0x06c6
 219         {0, 0, 0, 0}, // 0x06c7
 220         {0, 0, 0, 0}, // 0x06c8
 221         {0, 0, 0, 0}, // 0x06c9
 222         {0, 0, 0, 0}, // 0x06ca
 223         {0, 0, 0, 0}, // 0x06cb
 224         {0xfbfc, 0xfbfd, 0xfbfe, 0xfbff} // 0x06cc = farsi yeh
 225 };
 226
 227
 228 char_type const arabic_start = 0x0621;
 229 char_type const arabic_end = 0x06cc;
 230
 231
 232 enum CharInfoFlags {
 233         ///
 234         CharInfoCombining = 1,
 235         ///
 236         CharInfoTextFeature = 2,
 237         ///
 238         CharInfoMathFeature = 4,
 239         ///
 240         CharInfoForce = 8,
 241         ///
 242         CharInfoTextNoTermination = 16,
 243         ///
 244         CharInfoMathNoTermination = 32,
 245         ///
 246         CharInfoForceSelected = 64,
 247 };
 248
 249 /// Information about a single UCS4 character
 250 struct CharInfo {
 251         /// LaTeX command (text mode) for this character
 252         docstring textcommand;
 253         /// LaTeX command (math mode) for this character
 254         docstring mathcommand;
 255         /// Needed LaTeX preamble (or feature) for text mode
 256         string textpreamble;
 257         /// Needed LaTeX preamble (or feature) for math mode
 258         string mathpreamble;
 259         /// Is this a combining character?
 260         bool combining() const { return flags & CharInfoCombining ? true : false; }
 261         /// Is \c textpreamble a feature known by LaTeXFeatures, or a raw LaTeX
 262         /// command?
 263         bool textfeature() const { return flags & CharInfoTextFeature ? true : false; }
 264         /// Is \c mathpreamble a feature known by LaTeXFeatures, or a raw LaTeX
 265         /// command?
 266         bool mathfeature() const { return flags & CharInfoMathFeature ? true : false; }
 267         /// Always force the LaTeX command, even if the encoding contains
 268         /// this character?
 269         bool force() const { return flags & CharInfoForce ? true : false; }
 270         /// Force the LaTeX command for some encodings?
 271         bool forceselected() const { return flags & CharInfoForceSelected ? true : false; }
 272         /// TIPA shortcut
 273         string tipashortcut;
 274         /// \c textcommand needs no termination (such as {} or space).
 275         bool textnotermination() const { return flags & CharInfoTextNoTermination ? true : false; }
 276         /// \c mathcommand needs no termination (such as {} or space).
 277         bool mathnotermination() const { return flags & CharInfoMathNoTermination ? true : false; }
 278         ///
 279         unsigned int flags;
 280 };
 281
 282
 283 typedef map<char_type, CharInfo> CharInfoMap;
 284 CharInfoMap unicodesymbols;
 285
 286 typedef std::set<char_type> CharSet;
 287 typedef std::map<string, CharSet> CharSetMap;
 288 CharSet forced;
 289 CharSetMap forcedselected;
 290
 291 typedef std::set<char_type> MathAlphaSet;
 292 MathAlphaSet mathalpha;
 293
 294
 295 /// The highest code point in UCS4 encoding (1<<20 + 1<<16)
 296 char_type const max_ucs4 = 0x110000;
 297
 298 } // namespace anon
 299
 300
 301 EncodingException::EncodingException(char_type c)
 302         : failed_char(c), par_id(0), pos(0)
 303 {
 304 }
 305
 306
 307 const char * EncodingException::what() const throw()
 308 {
 309         return "Could not find LaTeX command for a character";
 310 }
 311
 312
 313 Encoding::Encoding(string const & n, string const & l, string const & g,
 314                    string const & i, bool f, bool u, Encoding::Package p)
 315         : name_(n), latexName_(l), guiName_(g), iconvName_(i), fixedwidth_(f),
 316           unsafe_(u), forced_(&forcedselected[n]), package_(p)
 317 {
 318         if (n == "ascii") {
 319                 // ASCII can encode 128 code points and nothing else
 320                 start_encodable_ = 128;
 321                 complete_ = true;
 322         } else if (i == "UTF-8") {
 323                 // UTF8 can encode all UCS4 code points
 324                 start_encodable_ = max_ucs4;
 325                 complete_ = true;
 326         } else {
 327                 start_encodable_ = 0;
 328                 complete_ = false;
 329         }
 330 }
 331
 332
 333 void Encoding::init() const
 334 {
 335         if (complete_)
 336                 return;
 337
 338         start_encodable_ = 0;
 339         // temporarily switch off lyxerr, since we will generate iconv errors
 340         lyxerr.disable();
 341         if (fixedwidth_) {
 342                 // We do not need to check all UCS4 code points, it is enough
 343                 // if we check all 256 code points of this encoding.
 344                 for (unsigned short j = 0; j < 256; ++j) {
 345                         char const c = char(j);
 346                         vector<char_type> const ucs4 = eightbit_to_ucs4(&c, 1, iconvName_);
 347                         if (ucs4.size() != 1)
 348                                 continue;
 349                         char_type const uc = ucs4[0];
 350                         CharInfoMap::const_iterator const it = unicodesymbols.find(uc);
 351                         if (it == unicodesymbols.end())
 352                                 encodable_.insert(uc);
 353                         else if (!it->second.force()) {
 354                                 if (forced_->empty() || forced_->find(uc) == forced_->end())
 355                                         encodable_.insert(uc);
 356                         }
 357                 }
 358         } else {
 359                 // We do not know how many code points this encoding has, and
 360                 // they do not have a direct representation as a single byte,
 361                 // therefore we need to check all UCS4 code points.
 362                 // This is expensive!
 363                 for (char_type c = 0; c < max_ucs4; ++c) {
 364                         vector<char> const eightbit = ucs4_to_eightbit(&c, 1, iconvName_);
 365                         if (!eightbit.empty()) {
 366                                 CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 367                                 if (it == unicodesymbols.end())
 368                                         encodable_.insert(c);
 369                                 else if (!it->second.force()) {
 370                                         if (forced_->empty() || forced_->find(c) == forced_->end())
 371                                                 encodable_.insert(c);
 372                                 }
 373                         }
 374                 }
 375         }
 376         lyxerr.enable();
 377         CharSet::iterator it = encodable_.find(start_encodable_);
 378         while (it != encodable_.end()) {
 379                 encodable_.erase(it);
 380                 ++start_encodable_;
 381                 it = encodable_.find(start_encodable_);
 382         }
 383         complete_ = true;
 384 }
 385
 386
 387 bool Encoding::isForced(char_type c) const
 388 {
 389         if (!forced.empty() && forced.find(c) != forced.end())
 390                 return true;
 391         return !forced_->empty() && forced_->find(c) != forced_->end();
 392 }
 393
 394
 395 bool Encoding::encodable(char_type c) const
 396 {
 397         // assure the used encoding is properly initialized
 398         init();
 399
 400         if (iconvName_ == "UTF-8" && package_ == none)
 401                 return true;
 402         if (c < start_encodable_ && !isForced(c))
 403                 return true;
 404         if (encodable_.find(c) != encodable_.end())
 405                 return true;
 406         return false;
 407 }
 408
 409
 410 pair<docstring, bool> Encoding::latexChar(char_type c) const
 411 {
 412         if (encodable(c))
 413                 return make_pair(docstring(1, c), false);
 414
 415         // c cannot (or should not) be encoded in this encoding
 416         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 417         if (it == unicodesymbols.end())
 418                 throw EncodingException(c);
 419         // at least one of mathcommand and textcommand is nonempty
 420         if (it->second.textcommand.empty())
 421                 return make_pair(
 422                         "\\ensuremath{" + it->second.mathcommand + '}', false);
 423         return make_pair(it->second.textcommand, !it->second.textnotermination());
 424 }
 425
 426
 427 pair<docstring, docstring> Encoding::latexString(docstring const input, bool dryrun) const
 428 {
 429         docstring result;
 430         docstring uncodable;
 431         bool terminate = false;
 432         for (size_t n = 0; n < input.size(); ++n) {
 433                 try {
 434                         char_type const c = input[n];
 435                         pair<docstring, bool> latex_char = latexChar(c);
 436                         docstring const latex = latex_char.first;
 437                         if (terminate && !prefixIs(latex, '\\')
 438                             && !prefixIs(latex, '{')
 439                             && !prefixIs(latex, '}')) {
 440                                         // Prevent eating of a following
 441                                         // space or command corruption by
 442                                         // following characters
 443                                         if (latex == " ")
 444                                                 result += "{}";
 445                                         else
 446                                                 result += " ";
 447                                 }
 448                         result += latex;
 449                         terminate = latex_char.second;
 450                 } catch (EncodingException & /* e */) {
 451                         LYXERR0("Uncodable character in latexString!");
 452                         if (dryrun) {
 453                                 result += "<" + _("LyX Warning: ")
 454                                            + _("uncodable character") + " '";
 455                                 result += docstring(1, input[n]);
 456                                 result += "'>";
 457                         } else
 458                                 uncodable += input[n];
 459                 }
 460         }
 461         return make_pair(result, uncodable);
 462 }
 463
 464
 465 vector<char_type> Encoding::symbolsList() const
 466 {
 467         // assure the used encoding is properly initialized
 468         init();
 469
 470         // first all encodable characters
 471         vector<char_type> symbols(encodable_.begin(), encodable_.end());
 472         // add those below start_encodable_
 473         for (char_type c = 0; c < start_encodable_; ++c)
 474                 symbols.push_back(c);
 475         // now the ones from the unicodesymbols file
 476         CharInfoMap::const_iterator const end = unicodesymbols.end();
 477         CharInfoMap::const_iterator it = unicodesymbols.begin();
 478         for (; it != end; ++it)
 479                 symbols.push_back(it->first);
 480         return symbols;
 481 }
 482
 483
 484 bool Encodings::latexMathChar(char_type c, bool mathmode,
 485                         Encoding const * encoding, docstring & command,
 486                         bool & needsTermination)
 487 {
 488         command = empty_docstring();
 489         if (encoding)
 490                 if (encoding->encodable(c))
 491                         command = docstring(1, c);
 492         needsTermination = false;
 493
 494         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 495         if (it == unicodesymbols.end()) {
 496                 if (!encoding || command.empty())
 497                         throw EncodingException(c);
 498                 if (mathmode)
 499                         addMathSym(c);
 500                 return false;
 501         }
 502         // at least one of mathcommand and textcommand is nonempty
 503         bool use_math = (mathmode && !it->second.mathcommand.empty()) ||
 504                         (!mathmode && it->second.textcommand.empty());
 505         if (use_math) {
 506                 command = it->second.mathcommand;
 507                 needsTermination = !it->second.mathnotermination();
 508                 addMathCmd(c);
 509         } else {
 510                 if (!encoding || command.empty()) {
 511                         command = it->second.textcommand;
 512                         needsTermination = !it->second.textnotermination();
 513                         addTextCmd(c);
 514                 }
 515                 if (mathmode)
 516                         addMathSym(c);
 517         }
 518         return use_math;
 519 }
 520
 521
 522 char_type Encodings::fromLaTeXCommand(docstring const & cmd, int cmdtype,
 523                 bool & combining, bool & needsTermination, set<string> * req)
 524 {
 525         CharInfoMap::const_iterator const end = unicodesymbols.end();
 526         CharInfoMap::const_iterator it = unicodesymbols.begin();
 527         for (combining = false; it != end; ++it) {
 528                 docstring const math = it->second.mathcommand;
 529                 docstring const text = it->second.textcommand;
 530                 if ((cmdtype & MATH_CMD) && math == cmd) {
 531                         combining = it->second.combining();
 532                         needsTermination = !it->second.mathnotermination();
 533                         if (req && it->second.mathfeature() &&
 534                             !it->second.mathpreamble.empty())
 535                                 req->insert(it->second.mathpreamble);
 536                         return it->first;
 537                 }
 538                 if ((cmdtype & TEXT_CMD) && text == cmd) {
 539                         combining = it->second.combining();
 540                         needsTermination = !it->second.textnotermination();
 541                         if (req && it->second.textfeature() &&
 542                             !it->second.textpreamble.empty())
 543                                 req->insert(it->second.textpreamble);
 544                         return it->first;
 545                 }
 546         }
 547         needsTermination = false;
 548         return 0;
 549 }
 550
 551
 552 docstring Encodings::fromLaTeXCommand(docstring const & cmd, int cmdtype,
 553                 bool & needsTermination, docstring & rem, set<string> * req)
 554 {
 555         needsTermination = false;
 556         rem = empty_docstring();
 557         bool const mathmode = cmdtype & MATH_CMD;
 558         bool const textmode = cmdtype & TEXT_CMD;
 559         docstring symbols;
 560         size_t i = 0;
 561         size_t const cmdend = cmd.size();
 562         CharInfoMap::const_iterator const uniend = unicodesymbols.end();
 563         for (size_t j = 0; j < cmdend; ++j) {
 564                 // Also get the char after a backslash
 565                 if (j + 1 < cmdend && cmd[j] == '\\')
 566                         ++j;
 567                 // If a macro argument follows, get it, too
 568                 if (j + 1 < cmdend && cmd[j + 1] == '{') {
 569                         size_t k = j + 1;
 570                         int count = 1;
 571                         while (k < cmdend && count && k != docstring::npos) {
 572                                 k = cmd.find_first_of(from_ascii("{}"), k + 1);
 573                                 if (cmd[k] == '{')
 574                                         ++count;
 575                                 else
 576                                         --count;
 577                         }
 578                         if (k != docstring::npos)
 579                                 j = k;
 580                 }
 581                 // Start with this substring and try augmenting it when it is
 582                 // the prefix of some command in the unicodesymbols file
 583                 docstring const subcmd = cmd.substr(i, j - i + 1);
 584
 585                 CharInfoMap::const_iterator it = unicodesymbols.begin();
 586                 size_t unicmd_size = 0;
 587                 char_type c = 0;
 588                 for (; it != uniend; ++it) {
 589                         docstring const math = mathmode ? it->second.mathcommand
 590                                                         : docstring();
 591                         docstring const text = textmode ? it->second.textcommand
 592                                                         : docstring();
 593                         size_t cur_size = max(math.size(), text.size());
 594                         // The current math or text unicode command cannot
 595                         // match, or we already matched a longer one
 596                         if (cur_size < subcmd.size() || cur_size <= unicmd_size)
 597                                 continue;
 598
 599                         docstring tmp = subcmd;
 600                         size_t k = j;
 601                         while (prefixIs(math, tmp) || prefixIs(text, tmp)) {
 602                                 ++k;
 603                                 if (k >= cmdend || cur_size <= tmp.size())
 604                                         break;
 605                                 tmp += cmd[k];
 606                         }
 607                         // No match
 608                         if (k == j)
 609                                 continue;
 610
 611                         // The last added char caused a mismatch, because
 612                         // we didn't exhaust the chars in cmd and didn't
 613                         // exceed the maximum size of the current unicmd
 614                         if (k < cmdend && cur_size > tmp.size())
 615                                 tmp.resize(tmp.size() - 1);
 616
 617                         // If this is an exact match, we found a (longer)
 618                         // matching entry in the unicodesymbols file.
 619                         // If the entry doesn't start with '\', we take note
 620                         // of the match and continue (this is not a ultimate
 621                         // acceptance, as some other entry may match a longer
 622                         // portion of the cmd string). However, if the entry
 623                         // does start with '\', we accept the match only if
 624                         // this is a valid macro, i.e., either it is a single
 625                         // (nonletter) char macro, or nothing else follows,
 626                         // or what follows is a nonletter char, or the last
 627                         // character is a }.
 628                         if ((math == tmp || text == tmp)
 629                             && (tmp[0] != '\\'
 630                                    || (tmp.size() == 2 && !isAlphaASCII(tmp[1]))
 631                                    || k == cmdend
 632                                    || !isAlphaASCII(cmd[k])
 633                                    || tmp[tmp.size() - 1] == '}')
 634                                  ) {
 635                                 c = it->first;
 636                                 j = k - 1;
 637                                 i = j + 1;
 638                                 unicmd_size = cur_size;
 639                                 if (math == tmp)
 640                                         needsTermination = !it->second.mathnotermination();
 641                                 else
 642                                         needsTermination = !it->second.textnotermination();
 643                                 if (req) {
 644                                         if (math == tmp && it->second.mathfeature() &&
 645                                             !it->second.mathpreamble.empty())
 646                                                 req->insert(it->second.mathpreamble);
 647                                         if (text == tmp && it->second.textfeature() &&
 648                                             !it->second.textpreamble.empty())
 649                                                 req->insert(it->second.textpreamble);
 650                                 }
 651                         }
 652                 }
 653                 if (unicmd_size)
 654                         symbols += c;
 655                 else if (j + 1 == cmdend) {
 656                         // No luck. Return what remains
 657                         rem = cmd.substr(i);
 658                         if (needsTermination && !rem.empty()) {
 659                                 if (rem.substr(0, 2) == "{}") {
 660                                         rem = rem.substr(2);
 661                                         needsTermination = false;
 662                                 } else if (rem[0] == ' ') {
 663                                         needsTermination = false;
 664                                         // LaTeX would swallow all spaces
 665                                         rem = ltrim(rem);
 666                                 }
 667                         }
 668                 }
 669         }
 670         return symbols;
 671 }
 672
 673
 674 void Encodings::initUnicodeMath(Buffer const & buffer, bool for_master)
 675 {
 676 #ifdef TEX2LYX
 677         // The code below is not needed in tex2lyx and requires additional stuff
 678         (void)buffer;
 679         (void)for_master;
 680 #else
 681         if (for_master) {
 682                 mathcmd.clear();
 683                 textcmd.clear();
 684                 mathsym.clear();
 685         }
 686
 687         // Check this buffer
 688         Inset & inset = buffer.inset();
 689         InsetIterator it = inset_iterator_begin(inset);
 690         InsetIterator const end = inset_iterator_end(inset);
 691         for (; it != end; ++it)
 692                 it->initUnicodeMath();
 693
 694         if (!for_master)
 695                 return;
 696
 697         // Check children
 698         ListOfBuffers blist = buffer.getDescendents();
 699         ListOfBuffers::const_iterator bit = blist.begin();
 700         ListOfBuffers::const_iterator const bend = blist.end();
 701         for (; bit != bend; ++bit)
 702                 initUnicodeMath(**bit, false);
 703 #endif
 704 }
 705
 706
 707 void Encodings::validate(char_type c, LaTeXFeatures & features, bool for_mathed)
 708 {
 709 #ifdef TEX2LYX
 710         // The code below is not needed in tex2lyx and requires additional stuff
 711         (void)c;
 712         (void)features;
 713         (void)for_mathed;
 714 #else
 715         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 716         if (it != unicodesymbols.end()) {
 717                 // In mathed, c could be used both in textmode and mathmode
 718                 bool const math_mode = for_mathed && isMathCmd(c);
 719                 bool const use_math = math_mode ||
 720                                       (!for_mathed && it->second.textcommand.empty());
 721                 bool const use_text = (for_mathed && isTextCmd(c)) ||
 722                                       (!for_mathed && !it->second.textcommand.empty());
 723                 bool const plain_utf8 = (features.runparams().encoding->name() == "utf8-plain");
 724                 bool const unicode_math = (features.isRequired("unicode-math")
 725                         && features.isAvailable("unicode-math"));
 726                 // with utf8-plain, we only load packages when in mathed (see #7766)
 727                 // and if we do not use unicode-math
 728                 if ((math_mode && !unicode_math)
 729                      || (use_math && !plain_utf8)) {
 730                         if (!it->second.mathpreamble.empty()) {
 731                                 if (it->second.mathfeature()) {
 732                                         string feats = it->second.mathpreamble;
 733                                         while (!feats.empty()) {
 734                                                 string feat;
 735                                                 feats = split(feats, feat, ',');
 736                                                 features.require(feat);
 737                                         }
 738                                 } else
 739                                         features.addPreambleSnippet(it->second.mathpreamble);
 740                         }
 741                 }
 742                 // with utf8-plain, we do not load packages (see #7766)
 743                 if (use_text && !plain_utf8) {
 744                         if (!it->second.textpreamble.empty()) {
 745                                 if (it->second.textfeature()) {
 746                                         string feats = it->second.textpreamble;
 747                                         while (!feats.empty()) {
 748                                                 string feat;
 749                                                 feats = split(feats, feat, ',');
 750                                                 features.require(feat);
 751                                         }
 752                                 } else
 753                                         features.addPreambleSnippet(it->second.textpreamble);
 754                         }
 755                 }
 756         }
 757         if (for_mathed && isMathSym(c)) {
 758                 features.require("amstext");
 759                 features.require("lyxmathsym");
 760         }
 761 #endif
 762 }
 763
 764
 765 bool Encodings::isHebrewComposeChar(char_type c)
 766 {
 767         return c <= 0x05c2 && c >= 0x05b0 && c != 0x05be && c != 0x05c0;
 768 }
 769
 770
 771 // Special Arabic letters are ones that do not get connected from left
 772 // they are hamza, alef_madda, alef_hamza, waw_hamza, alef_hamza_under,
 773 // alef, tah_marbota, dal, thal, rah, zai, wow, alef_maksoura
 774
 775 bool Encodings::isArabicSpecialChar(char_type c)
 776 {
 777         return (c >= 0x0621 && c <= 0x0625) || (c >= 0x0630 && c <= 0x0632)
 778                 || c == 0x0627 || c == 0x0629 || c == 0x062f || c == 0x0648
 779                 || c == 0x0649 || c == 0x0698;
 780 }
 781
 782
 783 bool Encodings::isArabicComposeChar(char_type c)
 784 {
 785         return c >= 0x064b && c <= 0x0652;
 786 }
 787
 788
 789 bool Encodings::isArabicChar(char_type c)
 790 {
 791         return c >= arabic_start && c <= arabic_end
 792                 && arabic_table[c-arabic_start][0];
 793 }
 794
 795
 796 char_type Encodings::transformChar(char_type c, Encodings::LetterForm form)
 797 {
 798         return isArabicChar(c) ? arabic_table[c-arabic_start][form] : c;
 799 }
 800
 801
 802 bool Encodings::isCombiningChar(char_type c)
 803 {
 804         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 805         if (it != unicodesymbols.end())
 806                 return it->second.combining();
 807         return false;
 808 }
 809
 810
 811 string const Encodings::TIPAShortcut(char_type c)
 812 {
 813         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 814         if (it != unicodesymbols.end())
 815                 return it->second.tipashortcut;
 816         return string();
 817 }
 818
 819
 820 bool Encodings::isKnownScriptChar(char_type const c, string & preamble)
 821 {
 822         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 823
 824         if (it == unicodesymbols.end())
 825                 return false;
 826
 827         if (it->second.textpreamble != "textgreek" && it->second.textpreamble != "textcyr")
 828                 return false;
 829
 830         if (preamble.empty()) {
 831                 preamble = it->second.textpreamble;
 832                 return true;
 833         }
 834         return it->second.textpreamble == preamble;
 835 }
 836
 837
 838 bool Encodings::isMathAlpha(char_type c)
 839 {
 840         return mathalpha.count(c);
 841 }
 842
 843
 844 Encoding const *
 845 Encodings::fromLyXName(string const & name, bool allowUnsafe) const
 846 {
 847         EncodingList::const_iterator const it = encodinglist.find(name);
 848         if (!allowUnsafe && it->second.unsafe())
 849                 return 0;
 850         return it != encodinglist.end() ? &it->second : 0;
 851 }
 852
 853
 854 Encoding const *
 855 Encodings::fromLaTeXName(string const & n, bool allowUnsafe) const
 856 {
 857         string name = n;
 858         // FIXME: if we have to test for too many of these synonyms,
 859         // we should instead extend the format of lib/encodings
 860         if (n == "ansinew")
 861                 name = "cp1252";
 862
 863         // We don't use find_if because it makes copies of the pairs in
 864         // the map.
 865         // This linear search is OK since we don't have many encodings.
 866         // Users could even optimize it by putting the encodings they use
 867         // most at the top of lib/encodings.
 868         EncodingList::const_iterator const end = encodinglist.end();
 869         for (EncodingList::const_iterator it = encodinglist.begin(); it != end; ++it)
 870                 if (it->second.latexName() == name) {
 871                         if (!allowUnsafe && it->second.unsafe())
 872                                 return 0;
 873                         return &it->second;
 874                 }
 875         return 0;
 876 }
 877
 878
 879 Encodings::Encodings()
 880 {}
 881
 882
 883 void Encodings::read(FileName const & encfile, FileName const & symbolsfile)
 884 {
 885         // We must read the symbolsfile first, because the Encoding
 886         // constructor depends on it.
 887         Lexer symbolslex;
 888         symbolslex.setFile(symbolsfile);
 889         bool getNextToken = true;
 890         while (symbolslex.isOK()) {
 891                 char_type symbol;
 892                 CharInfo info;
 893                 string flags;
 894
 895                 if (getNextToken) {
 896                         if (!symbolslex.next(true))
 897                                 break;
 898                 } else
 899                         getNextToken = true;
 900
 901                 istringstream is(symbolslex.getString());
 902                 // reading symbol directly does not work if
 903                 // char_type == wchar_t.
 904                 boost::uint32_t tmp;
 905                 if(!(is >> hex >> tmp))
 906                         break;
 907                 symbol = tmp;
 908
 909                 if (!symbolslex.next(true))
 910                         break;
 911                 info.textcommand = symbolslex.getDocString();
 912                 if (!symbolslex.next(true))
 913                         break;
 914                 info.textpreamble = symbolslex.getString();
 915                 if (!symbolslex.next(true))
 916                         break;
 917                 flags = symbolslex.getString();
 918
 919                 info.flags = 0;
 920                 if (suffixIs(info.textcommand, '}'))
 921                         info.flags |= CharInfoTextNoTermination;
 922                 if (suffixIs(info.mathcommand, '}'))
 923                         info.flags |= CharInfoMathNoTermination;
 924                 while (!flags.empty()) {
 925                         string flag;
 926                         flags = split(flags, flag, ',');
 927                         if (flag == "combining") {
 928                                 info.flags |= CharInfoCombining;
 929                         } else if (flag == "force") {
 930                                 info.flags |= CharInfoForce;
 931                                 forced.insert(symbol);
 932                         } else if (prefixIs(flag, "force=")) {
 933                                 vector<string> encodings =
 934                                         getVectorFromString(flag.substr(6));
 935                                 for (size_t i = 0; i < encodings.size(); ++i)
 936                                         forcedselected[encodings[i]].insert(symbol);
 937                                 info.flags |= CharInfoForceSelected;
 938                         } else if (flag == "mathalpha") {
 939                                 mathalpha.insert(symbol);
 940                         } else if (flag == "notermination=text") {
 941                                 info.flags |= CharInfoTextNoTermination;
 942                         } else if (flag == "notermination=math") {
 943                                 info.flags |= CharInfoMathNoTermination;
 944                         } else if (flag == "notermination=both") {
 945                                 info.flags |= CharInfoTextNoTermination;
 946                                 info.flags |= CharInfoMathNoTermination;
 947                         } else if (flag == "notermination=none") {
 948                                 info.flags &= ~CharInfoTextNoTermination;
 949                                 info.flags &= ~CharInfoMathNoTermination;
 950                         } else if (contains(flag, "tipashortcut=")) {
 951                                 info.tipashortcut = split(flag, '=');
 952                         } else {
 953                                 lyxerr << "Ignoring unknown flag `" << flag
 954                                        << "' for symbol `0x"
 955                                        << hex << symbol << dec
 956                                        << "'." << endl;
 957                         }
 958                 }
 959                 // mathcommand and mathpreamble have been added for 1.6.0.
 960                 // make them optional so that old files still work.
 961                 int const lineno = symbolslex.lineNumber();
 962                 bool breakout = false;
 963                 if (symbolslex.next(true)) {
 964                         if (symbolslex.lineNumber() != lineno) {
 965                                 // line in old format without mathcommand and mathpreamble
 966                                 getNextToken = false;
 967                         } else {
 968                                 info.mathcommand = symbolslex.getDocString();
 969                                 if (symbolslex.next(true)) {
 970                                         if (symbolslex.lineNumber() != lineno) {
 971                                                 // line in new format with mathcommand only
 972                                                 getNextToken = false;
 973                                         } else {
 974                                                 // line in new format with mathcommand and mathpreamble
 975                                                 info.mathpreamble = symbolslex.getString();
 976                                         }
 977                                 } else
 978                                         breakout = true;
 979                         }
 980                 } else {
 981                         breakout = true;
 982                 }
 983
 984                 // backward compatibility
 985                 if (info.mathpreamble == "esintoramsmath")
 986                         info.mathpreamble = "esint|amsmath";
 987
 988                 if (!info.textpreamble.empty())
 989                         if (info.textpreamble[0] != '\\')
 990                                 info.flags |= CharInfoTextFeature;
 991                 if (!info.mathpreamble.empty())
 992                         if (info.mathpreamble[0] != '\\')
 993                                 info.flags |= CharInfoMathFeature;
 994
 995                 LYXERR(Debug::INFO, "Read unicode symbol " << symbol << " '"
 996                         << to_utf8(info.textcommand) << "' '" << info.textpreamble
 997                         << " '" << info.textfeature() << ' ' << info.textnotermination()
 998                         << ' ' << to_utf8(info.mathcommand) << "' '" << info.mathpreamble
 999                         << "' " << info.mathfeature() << ' ' << info.mathnotermination()
1000                         << ' ' << info.combining() << ' ' << info.force()
1001                         << ' ' << info.forceselected());
1002
1003                 // we assume that at least one command is nonempty when using unicodesymbols
1004                 if (!info.textcommand.empty() || !info.mathcommand.empty())
1005                         unicodesymbols[symbol] = info;
1006
1007                 if (breakout)
1008                         break;
1009         }
1010
1011         // Now read the encodings
1012         enum {
1013                 et_encoding = 1,
1014                 et_end
1015         };
1016
1017         LexerKeyword encodingtags[] = {
1018                 { "encoding", et_encoding },
1019                 { "end", et_end }
1020         };
1021
1022         Lexer lex(encodingtags);
1023         lex.setFile(encfile);
1024         lex.setContext("Encodings::read");
1025         while (lex.isOK()) {
1026                 switch (lex.lex()) {
1027                 case et_encoding:
1028                 {
1029                         lex.next();
1030                         string const name = lex.getString();
1031                         lex.next();
1032                         string const latexname = lex.getString();
1033                         lex.next();
1034                         string const guiname = lex.getString();
1035                         lex.next();
1036                         string const iconvname = lex.getString();
1037                         lex.next();
1038                         string const width = lex.getString();
1039                         bool fixedwidth = false;
1040                         bool unsafe = false;
1041                         if (width == "fixed")
1042                                 fixedwidth = true;
1043                         else if (width == "variable")
1044                                 fixedwidth = false;
1045                         else if (width == "variableunsafe") {
1046                                 fixedwidth = false;
1047                                 unsafe = true;
1048                         }
1049                         else
1050                                 lex.printError("Unknown width");
1051
1052                         lex.next();
1053                         string const p = lex.getString();
1054                         Encoding::Package package = Encoding::none;
1055                         if (p == "none")
1056                                 package = Encoding::none;
1057                         else if (p == "inputenc")
1058                                 package = Encoding::inputenc;
1059                         else if (p == "CJK")
1060                                 package = Encoding::CJK;
1061                         else if (p == "japanese")
1062                                 package = Encoding::japanese;
1063                         else
1064                                 lex.printError("Unknown package");
1065
1066                         LYXERR(Debug::INFO, "Reading encoding " << name);
1067                         encodinglist[name] = Encoding(name, latexname,
1068                                 guiname, iconvname, fixedwidth, unsafe,
1069                                 package);
1070
1071                         if (lex.lex() != et_end)
1072                                 lex.printError("Missing end");
1073                         break;
1074                 }
1075                 case et_end:
1076                         lex.printError("Misplaced end");
1077                         break;
1078                 case Lexer::LEX_FEOF:
1079                         break;
1080                 default:
1081                         lex.printError("Unknown tag");
1082                         break;
1083                 }
1084         }
1085 }
1086
1087
1088 } // namespace lyx