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 "Lexer.h"
  18
  19 #include "support/debug.h"
  20 #include "support/gettext.h"
  21 #include "support/lstrings.h"
  22 #include "support/textutils.h"
  23 #include "support/unicode.h"
  24
  25 #include <boost/cstdint.hpp>
  26
  27 #include <sstream>
  28
  29 using namespace std;
  30 using namespace lyx::support;
  31
  32 namespace lyx {
  33
  34 int const Encoding::any = -1;
  35
  36 Encodings encodings;
  37
  38 Encodings::MathCommandSet Encodings::mathcmd;
  39 Encodings::TextCommandSet Encodings::textcmd;
  40 Encodings::MathSymbolSet  Encodings::mathsym;
  41
  42 namespace {
  43
  44 char_type arabic_table[172][4] = {
  45         {0xfe80, 0xfe80, 0xfe80, 0xfe80}, // 0x0621 = hamza
  46         {0xfe81, 0xfe82, 0xfe81, 0xfe82}, // 0x0622 = ligature madda on alef
  47         {0xfe83, 0xfe84, 0xfe83, 0xfe84}, // 0x0623 = ligature hamza on alef
  48         {0xfe85, 0xfe86, 0xfe85, 0xfe86}, // 0x0624 = ligature hamza on waw
  49         {0xfe87, 0xfe88, 0xfe87, 0xfe88}, // 0x0625 = ligature hamza under alef
  50         {0xfe89, 0xfe8a, 0xfe8b, 0xfe8c}, // 0x0626 = ligature hamza on ya
  51         {0xfe8d, 0xfe8e, 0xfe8d, 0xfe8e}, // 0x0627 = alef
  52         {0xfe8f, 0xfe90, 0xfe91, 0xfe92}, // 0x0628 = baa
  53         {0xfe93, 0xfe94, 0xfe93, 0xfe94}, // 0x0629 = taa marbuta
  54         {0xfe95, 0xfe96, 0xfe97, 0xfe98}, // 0x062a = taa
  55         {0xfe99, 0xfe9a, 0xfe9b, 0xfe9c}, // 0x062b = thaa
  56         {0xfe9d, 0xfe9e, 0xfe9f, 0xfea0}, // 0x062c = jeem
  57         {0xfea1, 0xfea2, 0xfea3, 0xfea4}, // 0x062d = haa
  58         {0xfea5, 0xfea6, 0xfea7, 0xfea8}, // 0x062e = khaa
  59         {0xfea9, 0xfeaa, 0xfea9, 0xfeaa}, // 0x062f = dal
  60
  61         {0xfeab, 0xfeac, 0xfeab, 0xfeac}, // 0x0630 = thal
  62         {0xfead, 0xfeae, 0xfead, 0xfeae}, // 0x0631 = ra
  63         {0xfeaf, 0xfeb0, 0xfeaf, 0xfeb0}, // 0x0632 = zain
  64         {0xfeb1, 0xfeb2, 0xfeb3, 0xfeb4}, // 0x0633 = seen
  65         {0xfeb5, 0xfeb6, 0xfeb7, 0xfeb8}, // 0x0634 = sheen
  66         {0xfeb9, 0xfeba, 0xfebb, 0xfebc}, // 0x0635 = sad
  67         {0xfebd, 0xfebe, 0xfebf, 0xfec0}, // 0x0636 = dad
  68         {0xfec1, 0xfec2, 0xfec3, 0xfec4}, // 0x0637 = tah
  69         {0xfec5, 0xfec6, 0xfec7, 0xfec8}, // 0x0638 = zah
  70         {0xfec9, 0xfeca, 0xfecb, 0xfecc}, // 0x0639 = ain
  71         {0xfecd, 0xfece, 0xfecf, 0xfed0}, // 0x063a = ghain
  72         {0, 0, 0, 0}, // 0x063b
  73         {0, 0, 0, 0}, // 0x063c
  74         {0, 0, 0, 0}, // 0x063d
  75         {0, 0, 0, 0}, // 0x063e
  76         {0, 0, 0, 0}, // 0x063f
  77
  78         {0, 0, 0, 0}, // 0x0640
  79         {0xfed1, 0xfed2, 0xfed3, 0xfed4}, // 0x0641 = fa
  80         {0xfed5, 0xfed6, 0xfed7, 0xfed8}, // 0x0642 = qaf
  81         {0xfed9, 0xfeda, 0xfedb, 0xfedc}, // 0x0643 = kaf
  82         {0xfedd, 0xfede, 0xfedf, 0xfee0}, // 0x0644 = lam
  83         {0xfee1, 0xfee2, 0xfee3, 0xfee4}, // 0x0645 = meem
  84         {0xfee5, 0xfee6, 0xfee7, 0xfee8}, // 0x0646 = noon
  85         {0xfee9, 0xfeea, 0xfeeb, 0xfeec}, // 0x0647 = ha
  86         {0xfeed, 0xfeee, 0xfeed, 0xfeee}, // 0x0648 = waw
  87         {0xfeef, 0xfef0, 0xfeef, 0xfef0}, // 0x0649 = alef maksura
  88         {0xfef1, 0xfef2, 0xfef3, 0xfef4}, // 0x064a = ya
  89         {0x065b, 0x065b, 0x065b, 0x065b}, // 0x064b = fathatan
  90         {0x065c, 0x065c, 0x065c, 0x065c}, // 0x064c = dammatan
  91         {0x064d, 0x064d, 0x064d, 0x064d}, // 0x064d = kasratan
  92         {0x064e, 0x064e, 0x064e, 0x064e}, // 0x064e = fatha
  93         {0x064f, 0x064f, 0x064f, 0x064f}, // 0x064f = damma
  94
  95         {0x0650, 0x0650, 0x0650, 0x0650}, // 0x0650 = kasra
  96         {0x0651, 0x0651, 0x0651, 0x0651}, // 0x0651 = shadda
  97         {0x0652, 0x0652, 0x0652, 0x0652}, // 0x0652 = sukun
  98
  99         {0, 0, 0, 0}, // 0x0653
 100         {0, 0, 0, 0}, // 0x0654
 101         {0, 0, 0, 0}, // 0x0655
 102         {0, 0, 0, 0}, // 0x0656
 103         {0, 0, 0, 0}, // 0x0657
 104         {0, 0, 0, 0}, // 0x0658
 105         {0, 0, 0, 0}, // 0x0659
 106         {0, 0, 0, 0}, // 0x065a
 107         {0, 0, 0, 0}, // 0x065b
 108         {0, 0, 0, 0}, // 0x065c
 109         {0, 0, 0, 0}, // 0x065d
 110         {0, 0, 0, 0}, // 0x065e
 111         {0, 0, 0, 0}, // 0x065f
 112         {0, 0, 0, 0}, // 0x0660
 113         {0, 0, 0, 0}, // 0x0661
 114         {0, 0, 0, 0}, // 0x0662
 115         {0, 0, 0, 0}, // 0x0663
 116         {0, 0, 0, 0}, // 0x0664
 117         {0, 0, 0, 0}, // 0x0665
 118         {0, 0, 0, 0}, // 0x0666
 119         {0, 0, 0, 0}, // 0x0667
 120         {0, 0, 0, 0}, // 0x0668
 121         {0, 0, 0, 0}, // 0x0669
 122         {0, 0, 0, 0}, // 0x066a
 123         {0, 0, 0, 0}, // 0x066b
 124         {0, 0, 0, 0}, // 0x066c
 125         {0, 0, 0, 0}, // 0x066d
 126         {0, 0, 0, 0}, // 0x066e
 127         {0, 0, 0, 0}, // 0x066f
 128         {0, 0, 0, 0}, // 0x0670
 129         {0, 0, 0, 0}, // 0x0671
 130         {0, 0, 0, 0}, // 0x0672
 131         {0, 0, 0, 0}, // 0x0673
 132         {0, 0, 0, 0}, // 0x0674
 133         {0, 0, 0, 0}, // 0x0675
 134         {0, 0, 0, 0}, // 0x0676
 135         {0, 0, 0, 0}, // 0x0677
 136         {0, 0, 0, 0}, // 0x0678
 137         {0, 0, 0, 0}, // 0x0679
 138         {0, 0, 0, 0}, // 0x067a
 139         {0, 0, 0, 0}, // 0x067b
 140         {0, 0, 0, 0}, // 0x067c
 141         {0, 0, 0, 0}, // 0x067d
 142         {0xfb56, 0xfb57, 0xfb58, 0xfb59}, // 0x067e = peh
 143         {0, 0, 0, 0}, // 0x067f
 144         {0, 0, 0, 0}, // 0x0680
 145         {0, 0, 0, 0}, // 0x0681
 146         {0, 0, 0, 0}, // 0x0682
 147         {0, 0, 0, 0}, // 0x0683
 148         {0, 0, 0, 0}, // 0x0684
 149         {0, 0, 0, 0}, // 0x0685
 150         {0xfb7a, 0xfb7b, 0xfb7c, 0xfb7d}, // 0x0686 = tcheh
 151         {0, 0, 0, 0}, // 0x0687
 152         {0, 0, 0, 0}, // 0x0688
 153         {0, 0, 0, 0}, // 0x0689
 154         {0, 0, 0, 0}, // 0x068a
 155         {0, 0, 0, 0}, // 0x068b
 156         {0, 0, 0, 0}, // 0x068c
 157         {0, 0, 0, 0}, // 0x068d
 158         {0, 0, 0, 0}, // 0x068e
 159         {0, 0, 0, 0}, // 0x068f
 160         {0, 0, 0, 0}, // 0x0690
 161         {0, 0, 0, 0}, // 0x0691
 162         {0, 0, 0, 0}, // 0x0692
 163         {0, 0, 0, 0}, // 0x0693
 164         {0, 0, 0, 0}, // 0x0694
 165         {0, 0, 0, 0}, // 0x0695
 166         {0, 0, 0, 0}, // 0x0696
 167         {0, 0, 0, 0}, // 0x0697
 168         {0xfb8a, 0xfb8b, 0xfb8a, 0xfb8b}, // 0x0698 = jeh
 169         {0, 0, 0, 0}, // 0x0699
 170         {0, 0, 0, 0}, // 0x069a
 171         {0, 0, 0, 0}, // 0x069b
 172         {0, 0, 0, 0}, // 0x069c
 173         {0, 0, 0, 0}, // 0x069d
 174         {0, 0, 0, 0}, // 0x069e
 175         {0, 0, 0, 0}, // 0x069f
 176         {0, 0, 0, 0}, // 0x06a0
 177         {0, 0, 0, 0}, // 0x06a1
 178         {0, 0, 0, 0}, // 0x06a2
 179         {0, 0, 0, 0}, // 0x06a3
 180         {0, 0, 0, 0}, // 0x06a4
 181         {0, 0, 0, 0}, // 0x06a5
 182         {0, 0, 0, 0}, // 0x06a6
 183         {0, 0, 0, 0}, // 0x06a7
 184         {0, 0, 0, 0}, // 0x06a8
 185         {0xfb8e, 0xfb8f, 0xfb90, 0xfb91}, // 0x06a9 = farsi kaf
 186         {0, 0, 0, 0}, // 0x06aa
 187         {0, 0, 0, 0}, // 0x06ab
 188         {0, 0, 0, 0}, // 0x06ac
 189         {0, 0, 0, 0}, // 0x06ad
 190         {0, 0, 0, 0}, // 0x06ae
 191         {0xfb92, 0xfb93, 0xfb94, 0xfb95}, // 0x06af = gaf
 192         {0, 0, 0, 0}, // 0x06b0
 193         {0, 0, 0, 0}, // 0x06b1
 194         {0, 0, 0, 0}, // 0x06b2
 195         {0, 0, 0, 0}, // 0x06b3
 196         {0, 0, 0, 0}, // 0x06b4
 197         {0, 0, 0, 0}, // 0x06b5
 198         {0, 0, 0, 0}, // 0x06b6
 199         {0, 0, 0, 0}, // 0x06b7
 200         {0, 0, 0, 0}, // 0x06b8
 201         {0, 0, 0, 0}, // 0x06b9
 202         {0, 0, 0, 0}, // 0x06ba
 203         {0, 0, 0, 0}, // 0x06bb
 204         {0, 0, 0, 0}, // 0x06bc
 205         {0, 0, 0, 0}, // 0x06bd
 206         {0, 0, 0, 0}, // 0x06be
 207         {0, 0, 0, 0}, // 0x06bf
 208         {0, 0, 0, 0}, // 0x06c0
 209         {0, 0, 0, 0}, // 0x06c1
 210         {0, 0, 0, 0}, // 0x06c2
 211         {0, 0, 0, 0}, // 0x06c3
 212         {0, 0, 0, 0}, // 0x06c4
 213         {0, 0, 0, 0}, // 0x06c5
 214         {0, 0, 0, 0}, // 0x06c6
 215         {0, 0, 0, 0}, // 0x06c7
 216         {0, 0, 0, 0}, // 0x06c8
 217         {0, 0, 0, 0}, // 0x06c9
 218         {0, 0, 0, 0}, // 0x06ca
 219         {0, 0, 0, 0}, // 0x06cb
 220         {0xfbfc, 0xfbfd, 0xfbfe, 0xfbff} // 0x06cc = farsi yeh
 221 };
 222
 223
 224 char_type const arabic_start = 0x0621;
 225 char_type const arabic_end = 0x06cc;
 226
 227
 228 typedef map<char_type, CharInfo> CharInfoMap;
 229 CharInfoMap unicodesymbols;
 230
 231 typedef set<char_type> CharSet;
 232 typedef map<string, CharSet> CharSetMap;
 233 CharSet forced;
 234 CharSetMap forcedselected;
 235
 236 typedef set<char_type> MathAlphaSet;
 237 MathAlphaSet mathalpha;
 238
 239
 240 /// The highest code point in UCS4 encoding (1<<20 + 1<<16)
 241 char_type const max_ucs4 = 0x110000;
 242
 243 } // namespace anon
 244
 245
 246 EncodingException::EncodingException(char_type c)
 247         : failed_char(c), par_id(0), pos(0)
 248 {
 249 }
 250
 251
 252 const char * EncodingException::what() const throw()
 253 {
 254         return "Could not find LaTeX command for a character";
 255 }
 256
 257
 258 CharInfo::CharInfo(
 259         docstring const textcommand, docstring const mathcommand,
 260         std::string const textpreamble, std::string const mathpreamble,
 261         std::string const tipashortcut, unsigned int flags)
 262         : textcommand_(textcommand), mathcommand_(mathcommand),
 263           textpreamble_(textpreamble), mathpreamble_(mathpreamble),
 264           tipashortcut_(tipashortcut), flags_(flags)
 265 {
 266 }
 267
 268 Encoding::Encoding(string const & n, string const & l, string const & g,
 269                    string const & i, bool f, bool u, Encoding::Package p)
 270         : name_(n), latexName_(l), guiName_(g), iconvName_(i), fixedwidth_(f),
 271           unsafe_(u), forced_(&forcedselected[n]), package_(p)
 272 {
 273         if (n == "ascii") {
 274                 // ASCII can encode 128 code points and nothing else
 275                 start_encodable_ = 128;
 276                 complete_ = true;
 277         } else if (i == "UTF-8") {
 278                 // UTF8 can encode all UCS4 code points
 279                 start_encodable_ = max_ucs4;
 280                 complete_ = true;
 281         } else {
 282                 start_encodable_ = 0;
 283                 complete_ = false;
 284         }
 285 }
 286
 287
 288 void Encoding::init() const
 289 {
 290         if (complete_)
 291                 return;
 292
 293         start_encodable_ = 0;
 294         // temporarily switch off lyxerr, since we will generate iconv errors
 295         lyxerr.disable();
 296         if (fixedwidth_) {
 297                 // We do not need to check all UCS4 code points, it is enough
 298                 // if we check all 256 code points of this encoding.
 299                 for (unsigned short j = 0; j < 256; ++j) {
 300                         char const c = char(j);
 301                         vector<char_type> const ucs4 = eightbit_to_ucs4(&c, 1, iconvName_);
 302                         if (ucs4.size() != 1)
 303                                 continue;
 304                         char_type const uc = ucs4[0];
 305                         CharInfoMap::const_iterator const it = unicodesymbols.find(uc);
 306                         if (it == unicodesymbols.end())
 307                                 encodable_.insert(uc);
 308                         else if (!it->second.force()) {
 309                                 if (forced_->empty() || forced_->find(uc) == forced_->end())
 310                                         encodable_.insert(uc);
 311                         }
 312                 }
 313         } else {
 314                 // We do not know how many code points this encoding has, and
 315                 // they do not have a direct representation as a single byte,
 316                 // therefore we need to check all UCS4 code points.
 317                 // This is expensive!
 318                 for (char_type c = 0; c < max_ucs4; ++c) {
 319                         vector<char> const eightbit = ucs4_to_eightbit(&c, 1, iconvName_);
 320                         if (!eightbit.empty()) {
 321                                 CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 322                                 if (it == unicodesymbols.end())
 323                                         encodable_.insert(c);
 324                                 else if (!it->second.force()) {
 325                                         if (forced_->empty() || forced_->find(c) == forced_->end())
 326                                                 encodable_.insert(c);
 327                                 }
 328                         }
 329                 }
 330         }
 331         lyxerr.enable();
 332         CharSet::iterator it = encodable_.find(start_encodable_);
 333         while (it != encodable_.end()) {
 334                 encodable_.erase(it);
 335                 ++start_encodable_;
 336                 it = encodable_.find(start_encodable_);
 337         }
 338         complete_ = true;
 339 }
 340
 341
 342 bool Encoding::isForced(char_type c) const
 343 {
 344         if (!forced.empty() && forced.find(c) != forced.end())
 345                 return true;
 346         return !forced_->empty() && forced_->find(c) != forced_->end();
 347 }
 348
 349
 350 bool Encoding::encodable(char_type c) const
 351 {
 352         // assure the used encoding is properly initialized
 353         init();
 354
 355         if (iconvName_ == "UTF-8" && package_ == none)
 356                 return true;
 357         if (c < start_encodable_ && !isForced(c))
 358                 return true;
 359         if (encodable_.find(c) != encodable_.end())
 360                 return true;
 361         return false;
 362 }
 363
 364
 365 pair<docstring, bool> Encoding::latexChar(char_type c) const
 366 {
 367         if (encodable(c))
 368                 return make_pair(docstring(1, c), false);
 369
 370         // c cannot (or should not) be encoded in this encoding
 371         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 372         if (it == unicodesymbols.end())
 373                 throw EncodingException(c);
 374         // at least one of mathcommand and textcommand is nonempty
 375         if (it->second.textcommand().empty())
 376                 return make_pair(
 377                         "\\ensuremath{" + it->second.mathcommand() + '}', false);
 378         return make_pair(it->second.textcommand(), !it->second.textnotermination());
 379 }
 380
 381
 382 pair<docstring, docstring> Encoding::latexString(docstring const input, bool dryrun) const
 383 {
 384         docstring result;
 385         docstring uncodable;
 386         bool terminate = false;
 387         for (size_t n = 0; n < input.size(); ++n) {
 388                 try {
 389                         char_type const c = input[n];
 390                         pair<docstring, bool> latex_char = latexChar(c);
 391                         docstring const latex = latex_char.first;
 392                         if (terminate && !prefixIs(latex, '\\')
 393                             && !prefixIs(latex, '{')
 394                             && !prefixIs(latex, '}')) {
 395                                         // Prevent eating of a following
 396                                         // space or command corruption by
 397                                         // following characters
 398                                         if (latex == " ")
 399                                                 result += "{}";
 400                                         else
 401                                                 result += " ";
 402                                 }
 403                         result += latex;
 404                         terminate = latex_char.second;
 405                 } catch (EncodingException & /* e */) {
 406                         LYXERR0("Uncodable character in latexString!");
 407                         if (dryrun) {
 408                                 result += "<" + _("LyX Warning: ")
 409                                            + _("uncodable character") + " '";
 410                                 result += docstring(1, input[n]);
 411                                 result += "'>";
 412                         } else
 413                                 uncodable += input[n];
 414                 }
 415         }
 416         return make_pair(result, uncodable);
 417 }
 418
 419
 420 vector<char_type> Encoding::symbolsList() const
 421 {
 422         // assure the used encoding is properly initialized
 423         init();
 424
 425         // first all encodable characters
 426         vector<char_type> symbols(encodable_.begin(), encodable_.end());
 427         // add those below start_encodable_
 428         for (char_type c = 0; c < start_encodable_; ++c)
 429                 symbols.push_back(c);
 430         // now the ones from the unicodesymbols file
 431         CharInfoMap::const_iterator const end = unicodesymbols.end();
 432         CharInfoMap::const_iterator it = unicodesymbols.begin();
 433         for (; it != end; ++it)
 434                 symbols.push_back(it->first);
 435         return symbols;
 436 }
 437
 438
 439 bool Encodings::latexMathChar(char_type c, bool mathmode,
 440                         Encoding const * encoding, docstring & command,
 441                         bool & needsTermination)
 442 {
 443         command = empty_docstring();
 444         if (encoding)
 445                 if (encoding->encodable(c))
 446                         command = docstring(1, c);
 447         needsTermination = false;
 448
 449         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 450         if (it == unicodesymbols.end()) {
 451                 if (!encoding || command.empty())
 452                         throw EncodingException(c);
 453                 if (mathmode)
 454                         addMathSym(c);
 455                 return false;
 456         }
 457         // at least one of mathcommand and textcommand is nonempty
 458         bool use_math = (mathmode && !it->second.mathcommand().empty()) ||
 459                         (!mathmode && it->second.textcommand().empty());
 460         if (use_math) {
 461                 command = it->second.mathcommand();
 462                 needsTermination = !it->second.mathnotermination();
 463                 addMathCmd(c);
 464         } else {
 465                 if (!encoding || command.empty()) {
 466                         command = it->second.textcommand();
 467                         needsTermination = !it->second.textnotermination();
 468                         addTextCmd(c);
 469                 }
 470                 if (mathmode)
 471                         addMathSym(c);
 472         }
 473         return use_math;
 474 }
 475
 476
 477 char_type Encodings::fromLaTeXCommand(docstring const & cmd, int cmdtype,
 478                 bool & combining, bool & needsTermination, set<string> * req)
 479 {
 480         CharInfoMap::const_iterator const end = unicodesymbols.end();
 481         CharInfoMap::const_iterator it = unicodesymbols.begin();
 482         for (combining = false; it != end; ++it) {
 483                 docstring const math = it->second.mathcommand();
 484                 docstring const text = it->second.textcommand();
 485                 if ((cmdtype & MATH_CMD) && math == cmd) {
 486                         combining = it->second.combining();
 487                         needsTermination = !it->second.mathnotermination();
 488                         if (req && it->second.mathfeature() &&
 489                             !it->second.mathpreamble().empty())
 490                                 req->insert(it->second.mathpreamble());
 491                         return it->first;
 492                 }
 493                 if ((cmdtype & TEXT_CMD) && text == cmd) {
 494                         combining = it->second.combining();
 495                         needsTermination = !it->second.textnotermination();
 496                         if (req && it->second.textfeature() &&
 497                             !it->second.textpreamble().empty())
 498                                 req->insert(it->second.textpreamble());
 499                         return it->first;
 500                 }
 501         }
 502         needsTermination = false;
 503         return 0;
 504 }
 505
 506
 507 docstring Encodings::fromLaTeXCommand(docstring const & cmd, int cmdtype,
 508                 bool & needsTermination, docstring & rem, set<string> * req)
 509 {
 510         needsTermination = false;
 511         rem = empty_docstring();
 512         bool const mathmode = cmdtype & MATH_CMD;
 513         bool const textmode = cmdtype & TEXT_CMD;
 514         docstring symbols;
 515         size_t const cmdend = cmd.size();
 516         size_t prefix = 0;
 517         CharInfoMap::const_iterator const uniend = unicodesymbols.end();
 518         for (size_t i = 0, j = 0; j < cmdend; ++j) {
 519                 // Also get the char after a backslash
 520                 if (j + 1 < cmdend && cmd[j] == '\\') {
 521                         ++j;
 522                         prefix = 1;
 523                         // Detect things like \=*{e} as well
 524                         if (j + 3 < cmdend && cmd[j+1] == '*' &&
 525                             cmd[j+2] == '{') {
 526                                 ++j;
 527                                 prefix = 2;
 528                         }
 529                 }
 530                 // position of the last character before a possible macro
 531                 // argument
 532                 size_t m = j;
 533                 // If a macro argument follows, get it, too
 534                 // Do it here only for single character commands. Other
 535                 // combining commands need this too, but they are handled in
 536                 // the loop below for performance reasons.
 537                 if (j + 1 < cmdend && cmd[j + 1] == '{') {
 538                         size_t k = j + 1;
 539                         int count = 1;
 540                         while (k < cmdend && count) {
 541                                 k = cmd.find_first_of(from_ascii("{}"), k + 1);
 542                                 // braces may not be balanced
 543                                 if (k == docstring::npos)
 544                                         break;
 545                                 if (cmd[k] == '{')
 546                                         ++count;
 547                                 else
 548                                         --count;
 549                         }
 550                         if (k != docstring::npos)
 551                                 j = k;
 552                 } else if (m + 1 < cmdend && isAlphaASCII(cmd[m])) {
 553                         while (m + 2 < cmdend && isAlphaASCII(cmd[m+1]))
 554                                 m++;
 555                 }
 556                 // Start with this substring and try augmenting it when it is
 557                 // the prefix of some command in the unicodesymbols file
 558                 docstring subcmd = cmd.substr(i, j - i + 1);
 559
 560                 CharInfoMap::const_iterator it = unicodesymbols.begin();
 561                 // First part of subcmd which might be a combining character
 562                 docstring combcmd = (m == j) ? docstring() : cmd.substr(i, m - i + 1);
 563                 // The combining character of combcmd if it exists
 564                 CharInfoMap::const_iterator combining = uniend;
 565                 size_t unicmd_size = 0;
 566                 char_type c = 0;
 567                 for (; it != uniend; ++it) {
 568                         docstring const math = mathmode ? it->second.mathcommand()
 569                                                         : docstring();
 570                         docstring const text = textmode ? it->second.textcommand()
 571                                                         : docstring();
 572                         if (!combcmd.empty() && it->second.combining() &&
 573                             (math == combcmd || text == combcmd))
 574                                 combining = it;
 575                         size_t cur_size = max(math.size(), text.size());
 576                         // The current math or text unicode command cannot
 577                         // match, or we already matched a longer one
 578                         if (cur_size < subcmd.size() || cur_size <= unicmd_size)
 579                                 continue;
 580
 581                         docstring tmp = subcmd;
 582                         size_t k = j;
 583                         while (prefixIs(math, tmp) || prefixIs(text, tmp)) {
 584                                 ++k;
 585                                 if (k >= cmdend || cur_size <= tmp.size())
 586                                         break;
 587                                 tmp += cmd[k];
 588                         }
 589                         // No match
 590                         if (k == j)
 591                                 continue;
 592
 593                         // The last added char caused a mismatch, because
 594                         // we didn't exhaust the chars in cmd and didn't
 595                         // exceed the maximum size of the current unicmd
 596                         if (k < cmdend && cur_size > tmp.size())
 597                                 tmp.resize(tmp.size() - 1);
 598
 599                         // If this is an exact match, we found a (longer)
 600                         // matching entry in the unicodesymbols file.
 601                         if (math != tmp && text != tmp)
 602                                 continue;
 603                         // If we found a combining command, we need to append
 604                         // the macro argument if this has not been done above.
 605                         if (tmp == combcmd && combining != uniend &&
 606                             k < cmdend && cmd[k] == '{') {
 607                                 size_t l = k;
 608                                 int count = 1;
 609                                 while (l < cmdend && count) {
 610                                         l = cmd.find_first_of(from_ascii("{}"), l + 1);
 611                                         // braces may not be balanced
 612                                         if (l == docstring::npos)
 613                                                 break;
 614                                         if (cmd[l] == '{')
 615                                                 ++count;
 616                                         else
 617                                                 --count;
 618                                 }
 619                                 if (l != docstring::npos) {
 620                                         j = l;
 621                                         subcmd = cmd.substr(i, j - i + 1);
 622                                 }
 623                         }
 624                         // If the entry doesn't start with '\', we take note
 625                         // of the match and continue (this is not a ultimate
 626                         // acceptance, as some other entry may match a longer
 627                         // portion of the cmd string). However, if the entry
 628                         // does start with '\', we accept the match only if
 629                         // this is a valid macro, i.e., either it is a single
 630                         // (nonletter) char macro, or nothing else follows,
 631                         // or what follows is a nonletter char, or the last
 632                         // character is a }.
 633                         else if (tmp[0] != '\\'
 634                                    || (tmp.size() == prefix + 1 &&
 635                                        !isAlphaASCII(tmp[1]) &&
 636                                        (prefix == 1 || !isAlphaASCII(tmp[2])))
 637                                    || k == cmdend
 638                                    || !isAlphaASCII(cmd[k])
 639                                    || tmp[tmp.size() - 1] == '}'
 640                                  ) {
 641                                 c = it->first;
 642                                 j = k - 1;
 643                                 i = j + 1;
 644                                 unicmd_size = cur_size;
 645                                 if (math == tmp)
 646                                         needsTermination = !it->second.mathnotermination();
 647                                 else
 648                                         needsTermination = !it->second.textnotermination();
 649                                 if (req) {
 650                                         if (math == tmp && it->second.mathfeature() &&
 651                                             !it->second.mathpreamble().empty())
 652                                                 req->insert(it->second.mathpreamble());
 653                                         if (text == tmp && it->second.textfeature() &&
 654                                             !it->second.textpreamble().empty())
 655                                                 req->insert(it->second.textpreamble());
 656                                 }
 657                         }
 658                 }
 659                 if (unicmd_size)
 660                         symbols += c;
 661                 else if (combining != uniend &&
 662                          prefixIs(subcmd, combcmd + '{')) {
 663                         // We know that subcmd starts with combcmd and
 664                         // contains an argument in braces.
 665                         docstring const arg = subcmd.substr(
 666                                 combcmd.length() + 1,
 667                                 subcmd.length() - combcmd.length() - 2);
 668                         // If arg is a single character we can construct a
 669                         // combining sequence.
 670                         char_type a;
 671                         bool argcomb = false;
 672                         if (arg.size() == 1 && isAlnumASCII(arg[0]))
 673                                 a = arg[0];
 674                         else {
 675                                 // Use the version of fromLaTeXCommand() that
 676                                 // parses only one command, since we cannot
 677                                 // use more than one character.
 678                                 bool dummy = false;
 679                                 set<string> r;
 680                                 a = fromLaTeXCommand(arg, cmdtype, argcomb,
 681                                                      dummy, &r);
 682                                 if (a && req && !argcomb)
 683                                         req->insert(r.begin(), r.end());
 684                         }
 685                         if (a && !argcomb) {
 686                                 // In unicode the combining character comes
 687                                 // after its base
 688                                 symbols += a;
 689                                 symbols += combining->first;
 690                                 i = j + 1;
 691                                 unicmd_size = 2;
 692                         }
 693                 }
 694                 if (j + 1 == cmdend && !unicmd_size) {
 695                         // No luck. Return what remains
 696                         rem = cmd.substr(i);
 697                         if (needsTermination && !rem.empty()) {
 698                                 if (rem.substr(0, 2) == "{}") {
 699                                         rem = rem.substr(2);
 700                                         needsTermination = false;
 701                                 } else if (rem[0] == ' ') {
 702                                         needsTermination = false;
 703                                         // LaTeX would swallow all spaces
 704                                         rem = ltrim(rem);
 705                                 }
 706                         }
 707                 }
 708         }
 709         return symbols;
 710 }
 711
 712
 713 bool Encodings::isHebrewComposeChar(char_type c)
 714 {
 715         return c <= 0x05c2 && c >= 0x05b0 && c != 0x05be && c != 0x05c0;
 716 }
 717
 718
 719 // Special Arabic letters are ones that do not get connected from left
 720 // they are hamza, alef_madda, alef_hamza, waw_hamza, alef_hamza_under,
 721 // alef, tah_marbota, dal, thal, rah, zai, wow, alef_maksoura
 722
 723 bool Encodings::isArabicSpecialChar(char_type c)
 724 {
 725         return (c >= 0x0621 && c <= 0x0625) || (c >= 0x0630 && c <= 0x0632)
 726                 || c == 0x0627 || c == 0x0629 || c == 0x062f || c == 0x0648
 727                 || c == 0x0649 || c == 0x0698;
 728 }
 729
 730
 731 bool Encodings::isArabicComposeChar(char_type c)
 732 {
 733         return c >= 0x064b && c <= 0x0652;
 734 }
 735
 736
 737 bool Encodings::isArabicChar(char_type c)
 738 {
 739         return c >= arabic_start && c <= arabic_end
 740                 && arabic_table[c-arabic_start][0];
 741 }
 742
 743
 744 CharInfo const & Encodings::unicodeCharInfo(char_type c)
 745 {
 746         static CharInfo empty;
 747         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 748         return it != unicodesymbols.end() ? it->second : empty;
 749 }
 750
 751
 752 char_type Encodings::transformChar(char_type c, Encodings::LetterForm form)
 753 {
 754         return isArabicChar(c) ? arabic_table[c-arabic_start][form] : c;
 755 }
 756
 757
 758 bool Encodings::isCombiningChar(char_type c)
 759 {
 760         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 761         if (it != unicodesymbols.end())
 762                 return it->second.combining();
 763         return false;
 764 }
 765
 766
 767 string const Encodings::TIPAShortcut(char_type c)
 768 {
 769         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 770         if (it != unicodesymbols.end())
 771                 return it->second.tipashortcut();
 772         return string();
 773 }
 774
 775
 776 bool Encodings::isKnownScriptChar(char_type const c, string & preamble)
 777 {
 778         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 779
 780         if (it == unicodesymbols.end())
 781                 return false;
 782
 783         if (it->second.textpreamble() != "textgreek" && it->second.textpreamble() != "textcyr")
 784                 return false;
 785
 786         if (preamble.empty()) {
 787                 preamble = it->second.textpreamble();
 788                 return true;
 789         }
 790         return it->second.textpreamble() == preamble;
 791 }
 792
 793
 794 bool Encodings::isMathAlpha(char_type c)
 795 {
 796         return mathalpha.count(c);
 797 }
 798
 799
 800 Encoding const *
 801 Encodings::fromLyXName(string const & name, bool allowUnsafe) const
 802 {
 803         EncodingList::const_iterator const it = encodinglist.find(name);
 804         if (!allowUnsafe && it->second.unsafe())
 805                 return 0;
 806         return it != encodinglist.end() ? &it->second : 0;
 807 }
 808
 809
 810 Encoding const *
 811 Encodings::fromLaTeXName(string const & n, int const & p, bool allowUnsafe) const
 812 {
 813         string name = n;
 814         // FIXME: if we have to test for too many of these synonyms,
 815         // we should instead extend the format of lib/encodings
 816         if (n == "ansinew")
 817                 name = "cp1252";
 818
 819         // We don't use find_if because it makes copies of the pairs in
 820         // the map.
 821         // This linear search is OK since we don't have many encodings.
 822         // Users could even optimize it by putting the encodings they use
 823         // most at the top of lib/encodings.
 824         EncodingList::const_iterator const end = encodinglist.end();
 825         for (EncodingList::const_iterator it = encodinglist.begin(); it != end; ++it)
 826                 if ((it->second.latexName() == name) && (it->second.package() & p)
 827                                 && (!it->second.unsafe() || allowUnsafe))
 828                         return &it->second;
 829         return 0;
 830 }
 831
 832
 833 Encoding const *
 834 Encodings::fromIconvName(string const & n, int const & p, bool allowUnsafe) const
 835 {
 836         EncodingList::const_iterator const end = encodinglist.end();
 837         for (EncodingList::const_iterator it = encodinglist.begin(); it != end; ++it)
 838                 if ((it->second.iconvName() == n) && (it->second.package() & p)
 839                                 && (!it->second.unsafe() || allowUnsafe))
 840                         return &it->second;
 841         return 0;
 842 }
 843
 844
 845 Encodings::Encodings()
 846 {}
 847
 848
 849 void Encodings::read(FileName const & encfile, FileName const & symbolsfile)
 850 {
 851         // We must read the symbolsfile first, because the Encoding
 852         // constructor depends on it.
 853         CharSetMap forcednotselected;
 854         Lexer symbolslex;
 855         symbolslex.setFile(symbolsfile);
 856         bool getNextToken = true;
 857         while (symbolslex.isOK()) {
 858                 char_type symbol;
 859
 860                 if (getNextToken) {
 861                         if (!symbolslex.next(true))
 862                                 break;
 863                 } else
 864                         getNextToken = true;
 865
 866                 istringstream is(symbolslex.getString());
 867                 // reading symbol directly does not work if
 868                 // char_type == wchar_t.
 869                 boost::uint32_t tmp;
 870                 if(!(is >> hex >> tmp))
 871                         break;
 872                 symbol = tmp;
 873
 874                 if (!symbolslex.next(true))
 875                         break;
 876                 docstring textcommand = symbolslex.getDocString();
 877                 if (!symbolslex.next(true))
 878                         break;
 879                 string textpreamble = symbolslex.getString();
 880                 if (!symbolslex.next(true))
 881                         break;
 882                 string sflags = symbolslex.getString();
 883
 884                 string tipashortcut;
 885                 int flags = 0;
 886
 887                 if (suffixIs(textcommand, '}'))
 888                         flags |= CharInfoTextNoTermination;
 889                 while (!sflags.empty()) {
 890                         string flag;
 891                         sflags = split(sflags, flag, ',');
 892                         if (flag == "combining") {
 893                                 flags |= CharInfoCombining;
 894                         } else if (flag == "force") {
 895                                 flags |= CharInfoForce;
 896                                 forced.insert(symbol);
 897                         } else if (prefixIs(flag, "force=")) {
 898                                 vector<string> encodings =
 899                                         getVectorFromString(flag.substr(6), ";");
 900                                 for (size_t i = 0; i < encodings.size(); ++i)
 901                                         forcedselected[encodings[i]].insert(symbol);
 902                                 flags |= CharInfoForceSelected;
 903                         } else if (prefixIs(flag, "force!=")) {
 904                                 vector<string> encodings =
 905                                         getVectorFromString(flag.substr(7), ";");
 906                                 for (size_t i = 0; i < encodings.size(); ++i)
 907                                         forcednotselected[encodings[i]].insert(symbol);
 908                                 flags |= CharInfoForceSelected;
 909                         } else if (flag == "mathalpha") {
 910                                 mathalpha.insert(symbol);
 911                         } else if (flag == "notermination=text") {
 912                                 flags |= CharInfoTextNoTermination;
 913                         } else if (flag == "notermination=math") {
 914                                 flags |= CharInfoMathNoTermination;
 915                         } else if (flag == "notermination=both") {
 916                                 flags |= CharInfoTextNoTermination;
 917                                 flags |= CharInfoMathNoTermination;
 918                         } else if (flag == "notermination=none") {
 919                                 flags &= ~CharInfoTextNoTermination;
 920                                 flags &= ~CharInfoMathNoTermination;
 921                         } else if (contains(flag, "tipashortcut=")) {
 922                                 tipashortcut = split(flag, '=');
 923                         } else {
 924                                 lyxerr << "Ignoring unknown flag `" << flag
 925                                        << "' for symbol `0x"
 926                                        << hex << symbol << dec
 927                                        << "'." << endl;
 928                         }
 929                 }
 930                 // mathcommand and mathpreamble have been added for 1.6.0.
 931                 // make them optional so that old files still work.
 932                 int const lineno = symbolslex.lineNumber();
 933                 bool breakout = false;
 934                 docstring mathcommand;
 935                 string mathpreamble;
 936                 if (symbolslex.next(true)) {
 937                         if (symbolslex.lineNumber() != lineno) {
 938                                 // line in old format without mathcommand and mathpreamble
 939                                 getNextToken = false;
 940                         } else {
 941                                 mathcommand = symbolslex.getDocString();
 942                                 if (suffixIs(mathcommand, '}'))
 943                                         flags |= CharInfoMathNoTermination;
 944                                 if (symbolslex.next(true)) {
 945                                         if (symbolslex.lineNumber() != lineno) {
 946                                                 // line in new format with mathcommand only
 947                                                 getNextToken = false;
 948                                         } else {
 949                                                 // line in new format with mathcommand and mathpreamble
 950                                                 mathpreamble = symbolslex.getString();
 951                                         }
 952                                 } else
 953                                         breakout = true;
 954                         }
 955                 } else {
 956                         breakout = true;
 957                 }
 958
 959                 // backward compatibility
 960                 if (mathpreamble == "esintoramsmath")
 961                         mathpreamble = "esint|amsmath";
 962
 963                 if (!textpreamble.empty())
 964                         if (textpreamble[0] != '\\')
 965                                 flags |= CharInfoTextFeature;
 966                 if (!mathpreamble.empty())
 967                         if (mathpreamble[0] != '\\')
 968                                 flags |= CharInfoMathFeature;
 969
 970                 CharInfo info = CharInfo(
 971                         textcommand, mathcommand,
 972                         textpreamble, mathpreamble,
 973                         tipashortcut, flags);
 974                 LYXERR(Debug::INFO, "Read unicode symbol " << symbol << " '"
 975                            << to_utf8(info.textcommand()) << "' '" << info.textpreamble()
 976                            << " '" << info.textfeature() << ' ' << info.textnotermination()
 977                            << ' ' << to_utf8(info.mathcommand()) << "' '" << info.mathpreamble()
 978                            << "' " << info.mathfeature() << ' ' << info.mathnotermination()
 979                            << ' ' << info.combining() << ' ' << info.force()
 980                            << ' ' << info.forceselected());
 981
 982                 // we assume that at least one command is nonempty when using unicodesymbols
 983                 if (info.isUnicodeSymbol()) {
 984                         unicodesymbols[symbol] = info;
 985                 }
 986
 987                 if (breakout)
 988                         break;
 989         }
 990
 991         // Now read the encodings
 992         enum {
 993                 et_encoding = 1,
 994                 et_end
 995         };
 996
 997         LexerKeyword encodingtags[] = {
 998                 { "encoding", et_encoding },
 999                 { "end", et_end }
1000         };
1001
1002         Lexer lex(encodingtags);
1003         lex.setFile(encfile);
1004         lex.setContext("Encodings::read");
1005         while (lex.isOK()) {
1006                 switch (lex.lex()) {
1007                 case et_encoding:
1008                 {
1009                         lex.next();
1010                         string const name = lex.getString();
1011                         lex.next();
1012                         string const latexname = lex.getString();
1013                         lex.next();
1014                         string const guiname = lex.getString();
1015                         lex.next();
1016                         string const iconvname = lex.getString();
1017                         lex.next();
1018                         string const width = lex.getString();
1019                         bool fixedwidth = false;
1020                         bool unsafe = false;
1021                         if (width == "fixed")
1022                                 fixedwidth = true;
1023                         else if (width == "variable")
1024                                 fixedwidth = false;
1025                         else if (width == "variableunsafe") {
1026                                 fixedwidth = false;
1027                                 unsafe = true;
1028                         }
1029                         else
1030                                 lex.printError("Unknown width");
1031
1032                         lex.next();
1033                         string const p = lex.getString();
1034                         Encoding::Package package = Encoding::none;
1035                         if (p == "none")
1036                                 package = Encoding::none;
1037                         else if (p == "inputenc")
1038                                 package = Encoding::inputenc;
1039                         else if (p == "CJK")
1040                                 package = Encoding::CJK;
1041                         else if (p == "japanese")
1042                                 package = Encoding::japanese;
1043                         else
1044                                 lex.printError("Unknown package");
1045
1046                         LYXERR(Debug::INFO, "Reading encoding " << name);
1047                         encodinglist[name] = Encoding(name, latexname,
1048                                 guiname, iconvname, fixedwidth, unsafe,
1049                                 package);
1050
1051                         if (lex.lex() != et_end)
1052                                 lex.printError("Missing end");
1053                         break;
1054                 }
1055                 case et_end:
1056                         lex.printError("Misplaced end");
1057                         break;
1058                 case Lexer::LEX_FEOF:
1059                         break;
1060                 default:
1061                         lex.printError("Unknown tag");
1062                         break;
1063                 }
1064         }
1065
1066         // Move all information from forcednotselected to forcedselected
1067         for (CharSetMap::const_iterator it1 = forcednotselected.begin(); it1 != forcednotselected.end(); ++it1) {
1068                 for (CharSetMap::iterator it2 = forcedselected.begin(); it2 != forcedselected.end(); ++it2) {
1069                         if (it2->first != it1->first)
1070                                 it2->second.insert(it1->second.begin(), it1->second.end());
1071                 }
1072         }
1073
1074 }
1075
1076
1077 } // namespace lyx