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