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 #include <algorithm>
  29
  30 using namespace std;
  31 using namespace lyx::support;
  32
  33 namespace lyx {
  34
  35 int const Encoding::any = -1;
  36
  37 Encodings encodings;
  38
  39 Encodings::MathCommandSet Encodings::mathcmd;
  40 Encodings::TextCommandSet Encodings::textcmd;
  41 Encodings::MathSymbolSet  Encodings::mathsym;
  42
  43 namespace {
  44
  45 char_type arabic_table[172][4] = {
  46         {0xfe80, 0xfe80, 0xfe80, 0xfe80}, // 0x0621 = hamza
  47         {0xfe81, 0xfe82, 0xfe81, 0xfe82}, // 0x0622 = ligature madda on alef
  48         {0xfe83, 0xfe84, 0xfe83, 0xfe84}, // 0x0623 = ligature hamza on alef
  49         {0xfe85, 0xfe86, 0xfe85, 0xfe86}, // 0x0624 = ligature hamza on waw
  50         {0xfe87, 0xfe88, 0xfe87, 0xfe88}, // 0x0625 = ligature hamza under alef
  51         {0xfe89, 0xfe8a, 0xfe8b, 0xfe8c}, // 0x0626 = ligature hamza on ya
  52         {0xfe8d, 0xfe8e, 0xfe8d, 0xfe8e}, // 0x0627 = alef
  53         {0xfe8f, 0xfe90, 0xfe91, 0xfe92}, // 0x0628 = baa
  54         {0xfe93, 0xfe94, 0xfe93, 0xfe94}, // 0x0629 = taa marbuta
  55         {0xfe95, 0xfe96, 0xfe97, 0xfe98}, // 0x062a = taa
  56         {0xfe99, 0xfe9a, 0xfe9b, 0xfe9c}, // 0x062b = thaa
  57         {0xfe9d, 0xfe9e, 0xfe9f, 0xfea0}, // 0x062c = jeem
  58         {0xfea1, 0xfea2, 0xfea3, 0xfea4}, // 0x062d = haa
  59         {0xfea5, 0xfea6, 0xfea7, 0xfea8}, // 0x062e = khaa
  60         {0xfea9, 0xfeaa, 0xfea9, 0xfeaa}, // 0x062f = dal
  61
  62         {0xfeab, 0xfeac, 0xfeab, 0xfeac}, // 0x0630 = thal
  63         {0xfead, 0xfeae, 0xfead, 0xfeae}, // 0x0631 = ra
  64         {0xfeaf, 0xfeb0, 0xfeaf, 0xfeb0}, // 0x0632 = zain
  65         {0xfeb1, 0xfeb2, 0xfeb3, 0xfeb4}, // 0x0633 = seen
  66         {0xfeb5, 0xfeb6, 0xfeb7, 0xfeb8}, // 0x0634 = sheen
  67         {0xfeb9, 0xfeba, 0xfebb, 0xfebc}, // 0x0635 = sad
  68         {0xfebd, 0xfebe, 0xfebf, 0xfec0}, // 0x0636 = dad
  69         {0xfec1, 0xfec2, 0xfec3, 0xfec4}, // 0x0637 = tah
  70         {0xfec5, 0xfec6, 0xfec7, 0xfec8}, // 0x0638 = zah
  71         {0xfec9, 0xfeca, 0xfecb, 0xfecc}, // 0x0639 = ain
  72         {0xfecd, 0xfece, 0xfecf, 0xfed0}, // 0x063a = ghain
  73         {0, 0, 0, 0}, // 0x063b
  74         {0, 0, 0, 0}, // 0x063c
  75         {0, 0, 0, 0}, // 0x063d
  76         {0, 0, 0, 0}, // 0x063e
  77         {0, 0, 0, 0}, // 0x063f
  78
  79         {0, 0, 0, 0}, // 0x0640
  80         {0xfed1, 0xfed2, 0xfed3, 0xfed4}, // 0x0641 = fa
  81         {0xfed5, 0xfed6, 0xfed7, 0xfed8}, // 0x0642 = qaf
  82         {0xfed9, 0xfeda, 0xfedb, 0xfedc}, // 0x0643 = kaf
  83         {0xfedd, 0xfede, 0xfedf, 0xfee0}, // 0x0644 = lam
  84         {0xfee1, 0xfee2, 0xfee3, 0xfee4}, // 0x0645 = meem
  85         {0xfee5, 0xfee6, 0xfee7, 0xfee8}, // 0x0646 = noon
  86         {0xfee9, 0xfeea, 0xfeeb, 0xfeec}, // 0x0647 = ha
  87         {0xfeed, 0xfeee, 0xfeed, 0xfeee}, // 0x0648 = waw
  88         {0xfeef, 0xfef0, 0xfeef, 0xfef0}, // 0x0649 = alef maksura
  89         {0xfef1, 0xfef2, 0xfef3, 0xfef4}, // 0x064a = ya
  90         {0x065b, 0x065b, 0x065b, 0x065b}, // 0x064b = fathatan
  91         {0x065c, 0x065c, 0x065c, 0x065c}, // 0x064c = dammatan
  92         {0x064d, 0x064d, 0x064d, 0x064d}, // 0x064d = kasratan
  93         {0x064e, 0x064e, 0x064e, 0x064e}, // 0x064e = fatha
  94         {0x064f, 0x064f, 0x064f, 0x064f}, // 0x064f = damma
  95
  96         {0x0650, 0x0650, 0x0650, 0x0650}, // 0x0650 = kasra
  97         {0x0651, 0x0651, 0x0651, 0x0651}, // 0x0651 = shadda
  98         {0x0652, 0x0652, 0x0652, 0x0652}, // 0x0652 = sukun
  99
 100         {0, 0, 0, 0}, // 0x0653
 101         {0, 0, 0, 0}, // 0x0654
 102         {0, 0, 0, 0}, // 0x0655
 103         {0, 0, 0, 0}, // 0x0656
 104         {0, 0, 0, 0}, // 0x0657
 105         {0, 0, 0, 0}, // 0x0658
 106         {0, 0, 0, 0}, // 0x0659
 107         {0, 0, 0, 0}, // 0x065a
 108         {0, 0, 0, 0}, // 0x065b
 109         {0, 0, 0, 0}, // 0x065c
 110         {0, 0, 0, 0}, // 0x065d
 111         {0, 0, 0, 0}, // 0x065e
 112         {0, 0, 0, 0}, // 0x065f
 113         {0, 0, 0, 0}, // 0x0660
 114         {0, 0, 0, 0}, // 0x0661
 115         {0, 0, 0, 0}, // 0x0662
 116         {0, 0, 0, 0}, // 0x0663
 117         {0, 0, 0, 0}, // 0x0664
 118         {0, 0, 0, 0}, // 0x0665
 119         {0, 0, 0, 0}, // 0x0666
 120         {0, 0, 0, 0}, // 0x0667
 121         {0, 0, 0, 0}, // 0x0668
 122         {0, 0, 0, 0}, // 0x0669
 123         {0, 0, 0, 0}, // 0x066a
 124         {0, 0, 0, 0}, // 0x066b
 125         {0, 0, 0, 0}, // 0x066c
 126         {0, 0, 0, 0}, // 0x066d
 127         {0, 0, 0, 0}, // 0x066e
 128         {0, 0, 0, 0}, // 0x066f
 129         {0, 0, 0, 0}, // 0x0670
 130         {0, 0, 0, 0}, // 0x0671
 131         {0, 0, 0, 0}, // 0x0672
 132         {0, 0, 0, 0}, // 0x0673
 133         {0, 0, 0, 0}, // 0x0674
 134         {0, 0, 0, 0}, // 0x0675
 135         {0, 0, 0, 0}, // 0x0676
 136         {0, 0, 0, 0}, // 0x0677
 137         {0, 0, 0, 0}, // 0x0678
 138         {0, 0, 0, 0}, // 0x0679
 139         {0, 0, 0, 0}, // 0x067a
 140         {0, 0, 0, 0}, // 0x067b
 141         {0, 0, 0, 0}, // 0x067c
 142         {0, 0, 0, 0}, // 0x067d
 143         {0xfb56, 0xfb57, 0xfb58, 0xfb59}, // 0x067e = peh
 144         {0, 0, 0, 0}, // 0x067f
 145         {0, 0, 0, 0}, // 0x0680
 146         {0, 0, 0, 0}, // 0x0681
 147         {0, 0, 0, 0}, // 0x0682
 148         {0, 0, 0, 0}, // 0x0683
 149         {0, 0, 0, 0}, // 0x0684
 150         {0, 0, 0, 0}, // 0x0685
 151         {0xfb7a, 0xfb7b, 0xfb7c, 0xfb7d}, // 0x0686 = tcheh
 152         {0, 0, 0, 0}, // 0x0687
 153         {0, 0, 0, 0}, // 0x0688
 154         {0, 0, 0, 0}, // 0x0689
 155         {0, 0, 0, 0}, // 0x068a
 156         {0, 0, 0, 0}, // 0x068b
 157         {0, 0, 0, 0}, // 0x068c
 158         {0, 0, 0, 0}, // 0x068d
 159         {0, 0, 0, 0}, // 0x068e
 160         {0, 0, 0, 0}, // 0x068f
 161         {0, 0, 0, 0}, // 0x0690
 162         {0, 0, 0, 0}, // 0x0691
 163         {0, 0, 0, 0}, // 0x0692
 164         {0, 0, 0, 0}, // 0x0693
 165         {0, 0, 0, 0}, // 0x0694
 166         {0, 0, 0, 0}, // 0x0695
 167         {0, 0, 0, 0}, // 0x0696
 168         {0, 0, 0, 0}, // 0x0697
 169         {0xfb8a, 0xfb8b, 0xfb8a, 0xfb8b}, // 0x0698 = jeh
 170         {0, 0, 0, 0}, // 0x0699
 171         {0, 0, 0, 0}, // 0x069a
 172         {0, 0, 0, 0}, // 0x069b
 173         {0, 0, 0, 0}, // 0x069c
 174         {0, 0, 0, 0}, // 0x069d
 175         {0, 0, 0, 0}, // 0x069e
 176         {0, 0, 0, 0}, // 0x069f
 177         {0, 0, 0, 0}, // 0x06a0
 178         {0, 0, 0, 0}, // 0x06a1
 179         {0, 0, 0, 0}, // 0x06a2
 180         {0, 0, 0, 0}, // 0x06a3
 181         {0, 0, 0, 0}, // 0x06a4
 182         {0, 0, 0, 0}, // 0x06a5
 183         {0, 0, 0, 0}, // 0x06a6
 184         {0, 0, 0, 0}, // 0x06a7
 185         {0, 0, 0, 0}, // 0x06a8
 186         {0xfb8e, 0xfb8f, 0xfb90, 0xfb91}, // 0x06a9 = farsi kaf
 187         {0, 0, 0, 0}, // 0x06aa
 188         {0, 0, 0, 0}, // 0x06ab
 189         {0, 0, 0, 0}, // 0x06ac
 190         {0, 0, 0, 0}, // 0x06ad
 191         {0, 0, 0, 0}, // 0x06ae
 192         {0xfb92, 0xfb93, 0xfb94, 0xfb95}, // 0x06af = gaf
 193         {0, 0, 0, 0}, // 0x06b0
 194         {0, 0, 0, 0}, // 0x06b1
 195         {0, 0, 0, 0}, // 0x06b2
 196         {0, 0, 0, 0}, // 0x06b3
 197         {0, 0, 0, 0}, // 0x06b4
 198         {0, 0, 0, 0}, // 0x06b5
 199         {0, 0, 0, 0}, // 0x06b6
 200         {0, 0, 0, 0}, // 0x06b7
 201         {0, 0, 0, 0}, // 0x06b8
 202         {0, 0, 0, 0}, // 0x06b9
 203         {0, 0, 0, 0}, // 0x06ba
 204         {0, 0, 0, 0}, // 0x06bb
 205         {0, 0, 0, 0}, // 0x06bc
 206         {0, 0, 0, 0}, // 0x06bd
 207         {0, 0, 0, 0}, // 0x06be
 208         {0, 0, 0, 0}, // 0x06bf
 209         {0, 0, 0, 0}, // 0x06c0
 210         {0, 0, 0, 0}, // 0x06c1
 211         {0, 0, 0, 0}, // 0x06c2
 212         {0, 0, 0, 0}, // 0x06c3
 213         {0, 0, 0, 0}, // 0x06c4
 214         {0, 0, 0, 0}, // 0x06c5
 215         {0, 0, 0, 0}, // 0x06c6
 216         {0, 0, 0, 0}, // 0x06c7
 217         {0, 0, 0, 0}, // 0x06c8
 218         {0, 0, 0, 0}, // 0x06c9
 219         {0, 0, 0, 0}, // 0x06ca
 220         {0, 0, 0, 0}, // 0x06cb
 221         {0xfbfc, 0xfbfd, 0xfbfe, 0xfbff} // 0x06cc = farsi yeh
 222 };
 223
 224
 225 char_type const arabic_start = 0x0621;
 226 char_type const arabic_end = 0x06cc;
 227
 228
 229 typedef map<char_type, CharInfo> CharInfoMap;
 230 CharInfoMap unicodesymbols;
 231
 232 typedef set<char_type> CharSet;
 233 typedef map<string, CharSet> CharSetMap;
 234 CharSet forced;
 235 CharSetMap forcedselected;
 236
 237 typedef set<char_type> MathAlphaSet;
 238 MathAlphaSet mathalpha;
 239
 240
 241 /// The highest code point in UCS4 encoding (1<<20 + 1<<16)
 242 char_type const max_ucs4 = 0x110000;
 243
 244 } // namespace anon
 245
 246
 247 EncodingException::EncodingException(char_type c)
 248         : failed_char(c), par_id(0), pos(0)
 249 {
 250 }
 251
 252
 253 const char * EncodingException::what() const throw()
 254 {
 255         return "Could not find LaTeX command for a character";
 256 }
 257
 258
 259 CharInfo::CharInfo(
 260         docstring const & textcommand, docstring const & mathcommand,
 261         std::string const & textpreamble, std::string const & mathpreamble,
 262         std::string const & tipashortcut, unsigned int flags)
 263         : textcommand_(textcommand), mathcommand_(mathcommand),
 264           textpreamble_(textpreamble), mathpreamble_(mathpreamble),
 265           tipashortcut_(tipashortcut), flags_(flags)
 266 {
 267 }
 268
 269 Encoding::Encoding(string const & n, string const & l, string const & g,
 270                    string const & i, bool f, bool u, Encoding::Package p)
 271         : name_(n), latexName_(l), guiName_(g), iconvName_(i), fixedwidth_(f),
 272           unsafe_(u), forced_(&forcedselected[n]), package_(p)
 273 {
 274         if (n == "ascii") {
 275                 // ASCII can encode 128 code points and nothing else
 276                 start_encodable_ = 128;
 277                 complete_ = true;
 278         } else if (i == "UTF-8") {
 279                 // UTF8 can encode all UCS4 code points
 280                 start_encodable_ = max_ucs4;
 281                 complete_ = true;
 282         } else {
 283                 start_encodable_ = 0;
 284                 complete_ = false;
 285         }
 286 }
 287
 288
 289 void Encoding::init() const
 290 {
 291         if (complete_)
 292                 return;
 293
 294         start_encodable_ = 0;
 295         // temporarily switch off lyxerr, since we will generate iconv errors
 296         lyxerr.disable();
 297         if (fixedwidth_) {
 298                 // We do not need to check all UCS4 code points, it is enough
 299                 // if we check all 256 code points of this encoding.
 300                 for (unsigned short j = 0; j < 256; ++j) {
 301                         char const c = char(j);
 302                         vector<char_type> const ucs4 = eightbit_to_ucs4(&c, 1, iconvName_);
 303                         if (ucs4.size() != 1)
 304                                 continue;
 305                         char_type const uc = ucs4[0];
 306                         CharInfoMap::const_iterator const it = unicodesymbols.find(uc);
 307                         if (it == unicodesymbols.end())
 308                                 encodable_.insert(uc);
 309                         else if (!it->second.force()) {
 310                                 if (forced_->empty() || forced_->find(uc) == forced_->end())
 311                                         encodable_.insert(uc);
 312                         }
 313                 }
 314         } else {
 315                 // We do not know how many code points this encoding has, and
 316                 // they do not have a direct representation as a single byte,
 317                 // therefore we need to check all UCS4 code points.
 318                 // This is expensive!
 319                 for (char_type c = 0; c < max_ucs4; ++c) {
 320                         vector<char> const eightbit = ucs4_to_eightbit(&c, 1, iconvName_);
 321                         if (!eightbit.empty()) {
 322                                 CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 323                                 if (it == unicodesymbols.end())
 324                                         encodable_.insert(c);
 325                                 else if (!it->second.force()) {
 326                                         if (forced_->empty() || forced_->find(c) == forced_->end())
 327                                                 encodable_.insert(c);
 328                                 }
 329                         }
 330                 }
 331         }
 332         lyxerr.enable();
 333         CharSet::iterator it = encodable_.find(start_encodable_);
 334         while (it != encodable_.end()) {
 335                 encodable_.erase(it);
 336                 ++start_encodable_;
 337                 it = encodable_.find(start_encodable_);
 338         }
 339         complete_ = true;
 340 }
 341
 342
 343 bool Encoding::isForced(char_type c) const
 344 {
 345         if (!forced.empty() && forced.find(c) != forced.end())
 346                 return true;
 347         return !forced_->empty() && forced_->find(c) != forced_->end();
 348 }
 349
 350
 351 bool Encoding::encodable(char_type c) const
 352 {
 353         // assure the used encoding is properly initialized
 354         init();
 355
 356         if (iconvName_ == "UTF-8" && package_ == none)
 357                 return true;
 358         if (c < start_encodable_ && !isForced(c))
 359                 return true;
 360         if (encodable_.find(c) != encodable_.end())
 361                 return true;
 362         return false;
 363 }
 364
 365
 366 pair<docstring, bool> Encoding::latexChar(char_type c) const
 367 {
 368         if (encodable(c))
 369                 return make_pair(docstring(1, c), false);
 370
 371         // c cannot (or should not) be encoded in this encoding
 372         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 373         if (it == unicodesymbols.end())
 374                 throw EncodingException(c);
 375         // at least one of mathcommand and textcommand is nonempty
 376         if (it->second.textcommand().empty())
 377                 return make_pair(
 378                         "\\ensuremath{" + it->second.mathcommand() + '}', false);
 379         return make_pair(it->second.textcommand(), !it->second.textnotermination());
 380 }
 381
 382
 383 pair<docstring, docstring> Encoding::latexString(docstring const & input, bool dryrun) const
 384 {
 385         docstring result;
 386         docstring uncodable;
 387         bool terminate = false;
 388         for (size_t n = 0; n < input.size(); ++n) {
 389                 try {
 390                         char_type const c = input[n];
 391                         pair<docstring, bool> latex_char = latexChar(c);
 392                         docstring const latex = latex_char.first;
 393                         if (terminate && !prefixIs(latex, '\\')
 394                             && !prefixIs(latex, '{')
 395                             && !prefixIs(latex, '}')) {
 396                                         // Prevent eating of a following
 397                                         // space or command corruption by
 398                                         // following characters
 399                                         if (latex == " ")
 400                                                 result += "{}";
 401                                         else
 402                                                 result += " ";
 403                                 }
 404                         result += latex;
 405                         terminate = latex_char.second;
 406                 } catch (EncodingException & /* e */) {
 407                         LYXERR0("Uncodable character in latexString!");
 408                         if (dryrun) {
 409                                 result += "<" + _("LyX Warning: ")
 410                                            + _("uncodable character") + " '";
 411                                 result += docstring(1, input[n]);
 412                                 result += "'>";
 413                         } else
 414                                 uncodable += input[n];
 415                 }
 416         }
 417         return make_pair(result, uncodable);
 418 }
 419
 420
 421 vector<char_type> Encoding::symbolsList() const
 422 {
 423         // assure the used encoding is properly initialized
 424         init();
 425
 426         // first all encodable characters
 427         vector<char_type> symbols(encodable_.begin(), encodable_.end());
 428         // add those below start_encodable_
 429         for (char_type c = 0; c < start_encodable_; ++c)
 430                 symbols.push_back(c);
 431         // now the ones from the unicodesymbols file
 432         CharInfoMap::const_iterator const end = unicodesymbols.end();
 433         CharInfoMap::const_iterator it = unicodesymbols.begin();
 434         for (; it != end; ++it)
 435                 symbols.push_back(it->first);
 436         return symbols;
 437 }
 438
 439
 440 bool Encodings::latexMathChar(char_type c, bool mathmode,
 441                         Encoding const * encoding, docstring & command,
 442                         bool & needsTermination)
 443 {
 444         command = empty_docstring();
 445         if (encoding)
 446                 if (encoding->encodable(c))
 447                         command = docstring(1, c);
 448         needsTermination = false;
 449
 450         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 451         if (it == unicodesymbols.end()) {
 452                 if (!encoding || command.empty())
 453                         throw EncodingException(c);
 454                 if (mathmode)
 455                         addMathSym(c);
 456                 return false;
 457         }
 458         // at least one of mathcommand and textcommand is nonempty
 459         bool use_math = (mathmode && !it->second.mathcommand().empty()) ||
 460                         (!mathmode && it->second.textcommand().empty());
 461         if (use_math) {
 462                 command = it->second.mathcommand();
 463                 needsTermination = !it->second.mathnotermination();
 464                 addMathCmd(c);
 465         } else {
 466                 if (!encoding || command.empty()) {
 467                         command = it->second.textcommand();
 468                         needsTermination = !it->second.textnotermination();
 469                         addTextCmd(c);
 470                 }
 471                 if (mathmode)
 472                         addMathSym(c);
 473         }
 474         return use_math;
 475 }
 476
 477
 478 char_type Encodings::fromLaTeXCommand(docstring const & cmd, int cmdtype,
 479                 bool & combining, bool & needsTermination, set<string> * req)
 480 {
 481         CharInfoMap::const_iterator const end = unicodesymbols.end();
 482         CharInfoMap::const_iterator it = unicodesymbols.begin();
 483         for (combining = false; it != end; ++it) {
 484                 docstring const math = it->second.mathcommand();
 485                 docstring const text = it->second.textcommand();
 486                 if ((cmdtype & MATH_CMD) && math == cmd) {
 487                         combining = it->second.combining();
 488                         needsTermination = !it->second.mathnotermination();
 489                         if (req && it->second.mathfeature() &&
 490                             !it->second.mathpreamble().empty())
 491                                 req->insert(it->second.mathpreamble());
 492                         return it->first;
 493                 }
 494                 if ((cmdtype & TEXT_CMD) && text == cmd) {
 495                         combining = it->second.combining();
 496                         needsTermination = !it->second.textnotermination();
 497                         if (req && it->second.textfeature() &&
 498                             !it->second.textpreamble().empty())
 499                                 req->insert(it->second.textpreamble());
 500                         return it->first;
 501                 }
 502         }
 503         needsTermination = false;
 504         return 0;
 505 }
 506
 507
 508 docstring Encodings::fromLaTeXCommand(docstring const & cmd, int cmdtype,
 509                 bool & needsTermination, docstring & rem, set<string> * req)
 510 {
 511         needsTermination = false;
 512         rem = empty_docstring();
 513         bool const mathmode = cmdtype & MATH_CMD;
 514         bool const textmode = cmdtype & TEXT_CMD;
 515         docstring symbols;
 516         size_t const cmdend = cmd.size();
 517         size_t prefix = 0;
 518         CharInfoMap::const_iterator const uniend = unicodesymbols.end();
 519         for (size_t i = 0, j = 0; j < cmdend; ++j) {
 520                 // Also get the char after a backslash
 521                 if (j + 1 < cmdend && cmd[j] == '\\') {
 522                         ++j;
 523                         prefix = 1;
 524                         // Detect things like \=*{e} as well
 525                         if (j + 3 < cmdend && cmd[j+1] == '*' &&
 526                             cmd[j+2] == '{') {
 527                                 ++j;
 528                                 prefix = 2;
 529                         }
 530                 }
 531                 // position of the last character before a possible macro
 532                 // argument
 533                 size_t m = j;
 534                 // If a macro argument follows, get it, too
 535                 // Do it here only for single character commands. Other
 536                 // combining commands need this too, but they are handled in
 537                 // the loop below for performance reasons.
 538                 if (j + 1 < cmdend && cmd[j + 1] == '{') {
 539                         size_t k = j + 1;
 540                         int count = 1;
 541                         while (k < cmdend && count) {
 542                                 k = cmd.find_first_of(from_ascii("{}"), k + 1);
 543                                 // braces may not be balanced
 544                                 if (k == docstring::npos)
 545                                         break;
 546                                 if (cmd[k] == '{')
 547                                         ++count;
 548                                 else
 549                                         --count;
 550                         }
 551                         if (k != docstring::npos)
 552                                 j = k;
 553                 } else if (m + 1 < cmdend && isAlphaASCII(cmd[m])) {
 554                         while (m + 2 < cmdend && isAlphaASCII(cmd[m+1]))
 555                                 m++;
 556                 }
 557                 // Start with this substring and try augmenting it when it is
 558                 // the prefix of some command in the unicodesymbols file
 559                 docstring subcmd = cmd.substr(i, j - i + 1);
 560
 561                 CharInfoMap::const_iterator it = unicodesymbols.begin();
 562                 // First part of subcmd which might be a combining character
 563                 docstring combcmd = (m == j) ? docstring() : cmd.substr(i, m - i + 1);
 564                 // The combining character of combcmd if it exists
 565                 CharInfoMap::const_iterator combining = uniend;
 566                 size_t unicmd_size = 0;
 567                 char_type c = 0;
 568                 for (; it != uniend; ++it) {
 569                         docstring const math = mathmode ? it->second.mathcommand()
 570                                                         : docstring();
 571                         docstring const text = textmode ? it->second.textcommand()
 572                                                         : docstring();
 573                         if (!combcmd.empty() && it->second.combining() &&
 574                             (math == combcmd || text == combcmd))
 575                                 combining = it;
 576                         size_t cur_size = max(math.size(), text.size());
 577                         // The current math or text unicode command cannot
 578                         // match, or we already matched a longer one
 579                         if (cur_size < subcmd.size() || cur_size <= unicmd_size)
 580                                 continue;
 581
 582                         docstring tmp = subcmd;
 583                         size_t k = j;
 584                         while (prefixIs(math, tmp) || prefixIs(text, tmp)) {
 585                                 ++k;
 586                                 if (k >= cmdend || cur_size <= tmp.size())
 587                                         break;
 588                                 tmp += cmd[k];
 589                         }
 590                         // No match
 591                         if (k == j)
 592                                 continue;
 593
 594                         // The last added char caused a mismatch, because
 595                         // we didn't exhaust the chars in cmd and didn't
 596                         // exceed the maximum size of the current unicmd
 597                         if (k < cmdend && cur_size > tmp.size())
 598                                 tmp.resize(tmp.size() - 1);
 599
 600                         // If this is an exact match, we found a (longer)
 601                         // matching entry in the unicodesymbols file.
 602                         if (math != tmp && text != tmp)
 603                                 continue;
 604                         // If we found a combining command, we need to append
 605                         // the macro argument if this has not been done above.
 606                         if (tmp == combcmd && combining != uniend &&
 607                             k < cmdend && cmd[k] == '{') {
 608                                 size_t l = k;
 609                                 int count = 1;
 610                                 while (l < cmdend && count) {
 611                                         l = cmd.find_first_of(from_ascii("{}"), l + 1);
 612                                         // braces may not be balanced
 613                                         if (l == docstring::npos)
 614                                                 break;
 615                                         if (cmd[l] == '{')
 616                                                 ++count;
 617                                         else
 618                                                 --count;
 619                                 }
 620                                 if (l != docstring::npos) {
 621                                         j = l;
 622                                         subcmd = cmd.substr(i, j - i + 1);
 623                                 }
 624                         }
 625                         // If the entry doesn't start with '\', we take note
 626                         // of the match and continue (this is not a ultimate
 627                         // acceptance, as some other entry may match a longer
 628                         // portion of the cmd string). However, if the entry
 629                         // does start with '\', we accept the match only if
 630                         // this is a valid macro, i.e., either it is a single
 631                         // (nonletter) char macro, or nothing else follows,
 632                         // or what follows is a nonletter char, or the last
 633                         // character is a }.
 634                         else if (tmp[0] != '\\'
 635                                    || (tmp.size() == prefix + 1 &&
 636                                        !isAlphaASCII(tmp[1]) &&
 637                                        (prefix == 1 || !isAlphaASCII(tmp[2])))
 638                                    || k == cmdend
 639                                    || !isAlphaASCII(cmd[k])
 640                                    || tmp[tmp.size() - 1] == '}'
 641                                  ) {
 642                                 c = it->first;
 643                                 j = k - 1;
 644                                 i = j + 1;
 645                                 unicmd_size = cur_size;
 646                                 if (math == tmp)
 647                                         needsTermination = !it->second.mathnotermination();
 648                                 else
 649                                         needsTermination = !it->second.textnotermination();
 650                                 if (req) {
 651                                         if (math == tmp && it->second.mathfeature() &&
 652                                             !it->second.mathpreamble().empty())
 653                                                 req->insert(it->second.mathpreamble());
 654                                         if (text == tmp && it->second.textfeature() &&
 655                                             !it->second.textpreamble().empty())
 656                                                 req->insert(it->second.textpreamble());
 657                                 }
 658                         }
 659                 }
 660                 if (unicmd_size)
 661                         symbols += c;
 662                 else if (combining != uniend &&
 663                          prefixIs(subcmd, combcmd + '{')) {
 664                         // We know that subcmd starts with combcmd and
 665                         // contains an argument in braces.
 666                         docstring const arg = subcmd.substr(
 667                                 combcmd.length() + 1,
 668                                 subcmd.length() - combcmd.length() - 2);
 669                         // If arg is a single character we can construct a
 670                         // combining sequence.
 671                         char_type a;
 672                         bool argcomb = false;
 673                         if (arg.size() == 1 && isAlnumASCII(arg[0]))
 674                                 a = arg[0];
 675                         else {
 676                                 // Use the version of fromLaTeXCommand() that
 677                                 // parses only one command, since we cannot
 678                                 // use more than one character.
 679                                 bool dummy = false;
 680                                 set<string> r;
 681                                 a = fromLaTeXCommand(arg, cmdtype, argcomb,
 682                                                      dummy, &r);
 683                                 if (a && req && !argcomb)
 684                                         req->insert(r.begin(), r.end());
 685                         }
 686                         if (a && !argcomb) {
 687                                 // In unicode the combining character comes
 688                                 // after its base
 689                                 symbols += a;
 690                                 symbols += combining->first;
 691                                 i = j + 1;
 692                                 unicmd_size = 2;
 693                         }
 694                 }
 695                 if (j + 1 == cmdend && !unicmd_size) {
 696                         // No luck. Return what remains
 697                         rem = cmd.substr(i);
 698                         if (needsTermination && !rem.empty()) {
 699                                 if (rem.substr(0, 2) == "{}") {
 700                                         rem = rem.substr(2);
 701                                         needsTermination = false;
 702                                 } else if (rem[0] == ' ') {
 703                                         needsTermination = false;
 704                                         // LaTeX would swallow all spaces
 705                                         rem = ltrim(rem);
 706                                 }
 707                         }
 708                 }
 709         }
 710         return symbols;
 711 }
 712
 713
 714 bool Encodings::isHebrewComposeChar(char_type c)
 715 {
 716         return c <= 0x05c2 && c >= 0x05b0 && c != 0x05be && c != 0x05c0;
 717 }
 718
 719
 720 // Special Arabic letters are ones that do not get connected from left
 721 // they are hamza, alef_madda, alef_hamza, waw_hamza, alef_hamza_under,
 722 // alef, tah_marbota, dal, thal, rah, zai, wow, alef_maksoura
 723
 724 bool Encodings::isArabicSpecialChar(char_type c)
 725 {
 726         return (c >= 0x0621 && c <= 0x0625) || (c >= 0x0630 && c <= 0x0632)
 727                 || c == 0x0627 || c == 0x0629 || c == 0x062f || c == 0x0648
 728                 || c == 0x0649 || c == 0x0698;
 729 }
 730
 731
 732 bool Encodings::isArabicComposeChar(char_type c)
 733 {
 734         return c >= 0x064b && c <= 0x0652;
 735 }
 736
 737
 738 bool Encodings::isArabicChar(char_type c)
 739 {
 740         return c >= arabic_start && c <= arabic_end
 741                 && arabic_table[c-arabic_start][0];
 742 }
 743
 744
 745 CharInfo const & Encodings::unicodeCharInfo(char_type c)
 746 {
 747         static CharInfo empty;
 748         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 749         return it != unicodesymbols.end() ? it->second : empty;
 750 }
 751
 752
 753 char_type Encodings::transformChar(char_type c, Encodings::LetterForm form)
 754 {
 755         return isArabicChar(c) ? arabic_table[c-arabic_start][form] : c;
 756 }
 757
 758
 759 bool Encodings::isCombiningChar(char_type c)
 760 {
 761         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 762         if (it != unicodesymbols.end())
 763                 return it->second.combining();
 764         return false;
 765 }
 766
 767
 768 string const Encodings::TIPAShortcut(char_type c)
 769 {
 770         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 771         if (it != unicodesymbols.end())
 772                 return it->second.tipashortcut();
 773         return string();
 774 }
 775
 776
 777 bool Encodings::isKnownScriptChar(char_type const c, string & preamble)
 778 {
 779         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 780
 781         if (it == unicodesymbols.end())
 782                 return false;
 783
 784         if (it->second.textpreamble() != "textgreek" && it->second.textpreamble() != "textcyr")
 785                 return false;
 786
 787         if (preamble.empty()) {
 788                 preamble = it->second.textpreamble();
 789                 return true;
 790         }
 791         return it->second.textpreamble() == preamble;
 792 }
 793
 794
 795 bool Encodings::isMathAlpha(char_type c)
 796 {
 797         return mathalpha.count(c);
 798 }
 799
 800
 801 Encoding const *
 802 Encodings::fromLyXName(string const & name, bool allowUnsafe) const
 803 {
 804         EncodingList::const_iterator const it = encodinglist.find(name);
 805         if (!allowUnsafe && it->second.unsafe())
 806                 return 0;
 807         return it != encodinglist.end() ? &it->second : 0;
 808 }
 809
 810
 811 Encoding const *
 812 Encodings::fromLaTeXName(string const & n, int const & p, bool allowUnsafe) const
 813 {
 814         string name = n;
 815         // FIXME: if we have to test for too many of these synonyms,
 816         // we should instead extend the format of lib/encodings
 817         if (n == "ansinew")
 818                 name = "cp1252";
 819
 820         // We don't use find_if because it makes copies of the pairs in
 821         // the map.
 822         // This linear search is OK since we don't have many encodings.
 823         // Users could even optimize it by putting the encodings they use
 824         // most at the top of lib/encodings.
 825         EncodingList::const_iterator const end = encodinglist.end();
 826         for (EncodingList::const_iterator it = encodinglist.begin(); it != end; ++it)
 827                 if ((it->second.latexName() == name) && (it->second.package() & p)
 828                                 && (!it->second.unsafe() || allowUnsafe))
 829                         return &it->second;
 830         return 0;
 831 }
 832
 833
 834 Encoding const *
 835 Encodings::fromIconvName(string const & n, int const & p, bool allowUnsafe) const
 836 {
 837         EncodingList::const_iterator const end = encodinglist.end();
 838         for (EncodingList::const_iterator it = encodinglist.begin(); it != end; ++it)
 839                 if ((it->second.iconvName() == n) && (it->second.package() & p)
 840                                 && (!it->second.unsafe() || allowUnsafe))
 841                         return &it->second;
 842         return 0;
 843 }
 844
 845
 846 Encodings::Encodings()
 847 {}
 848
 849
 850 void Encodings::read(FileName const & encfile, FileName const & symbolsfile)
 851 {
 852         // We must read the symbolsfile first, because the Encoding
 853         // constructor depends on it.
 854         CharSetMap forcednotselected;
 855         Lexer symbolslex;
 856         symbolslex.setFile(symbolsfile);
 857         bool getNextToken = true;
 858         while (symbolslex.isOK()) {
 859                 char_type symbol;
 860
 861                 if (getNextToken) {
 862                         if (!symbolslex.next(true))
 863                                 break;
 864                 } else
 865                         getNextToken = true;
 866
 867                 istringstream is(symbolslex.getString());
 868                 // reading symbol directly does not work if
 869                 // char_type == wchar_t.
 870                 boost::uint32_t tmp;
 871                 if(!(is >> hex >> tmp))
 872                         break;
 873                 symbol = tmp;
 874
 875                 if (!symbolslex.next(true))
 876                         break;
 877                 docstring textcommand = symbolslex.getDocString();
 878                 if (!symbolslex.next(true))
 879                         break;
 880                 string textpreamble = symbolslex.getString();
 881                 if (!symbolslex.next(true))
 882                         break;
 883                 string sflags = symbolslex.getString();
 884
 885                 string tipashortcut;
 886                 int flags = 0;
 887
 888                 if (suffixIs(textcommand, '}'))
 889                         flags |= CharInfoTextNoTermination;
 890                 while (!sflags.empty()) {
 891                         string flag;
 892                         sflags = split(sflags, flag, ',');
 893                         if (flag == "combining") {
 894                                 flags |= CharInfoCombining;
 895                         } else if (flag == "force") {
 896                                 flags |= CharInfoForce;
 897                                 forced.insert(symbol);
 898                         } else if (prefixIs(flag, "force=")) {
 899                                 vector<string> encodings =
 900                                         getVectorFromString(flag.substr(6), ";");
 901                                 for (size_t i = 0; i < encodings.size(); ++i)
 902                                         forcedselected[encodings[i]].insert(symbol);
 903                                 flags |= CharInfoForceSelected;
 904                         } else if (prefixIs(flag, "force!=")) {
 905                                 vector<string> encodings =
 906                                         getVectorFromString(flag.substr(7), ";");
 907                                 for (size_t i = 0; i < encodings.size(); ++i)
 908                                         forcednotselected[encodings[i]].insert(symbol);
 909                                 flags |= CharInfoForceSelected;
 910                         } else if (flag == "mathalpha") {
 911                                 mathalpha.insert(symbol);
 912                         } else if (flag == "notermination=text") {
 913                                 flags |= CharInfoTextNoTermination;
 914                         } else if (flag == "notermination=math") {
 915                                 flags |= CharInfoMathNoTermination;
 916                         } else if (flag == "notermination=both") {
 917                                 flags |= CharInfoTextNoTermination;
 918                                 flags |= CharInfoMathNoTermination;
 919                         } else if (flag == "notermination=none") {
 920                                 flags &= ~CharInfoTextNoTermination;
 921                                 flags &= ~CharInfoMathNoTermination;
 922                         } else if (contains(flag, "tipashortcut=")) {
 923                                 tipashortcut = split(flag, '=');
 924                         } else {
 925                                 lyxerr << "Ignoring unknown flag `" << flag
 926                                        << "' for symbol `0x"
 927                                        << hex << symbol << dec
 928                                        << "'." << endl;
 929                         }
 930                 }
 931                 // mathcommand and mathpreamble have been added for 1.6.0.
 932                 // make them optional so that old files still work.
 933                 int const lineno = symbolslex.lineNumber();
 934                 bool breakout = false;
 935                 docstring mathcommand;
 936                 string mathpreamble;
 937                 if (symbolslex.next(true)) {
 938                         if (symbolslex.lineNumber() != lineno) {
 939                                 // line in old format without mathcommand and mathpreamble
 940                                 getNextToken = false;
 941                         } else {
 942                                 mathcommand = symbolslex.getDocString();
 943                                 if (suffixIs(mathcommand, '}'))
 944                                         flags |= CharInfoMathNoTermination;
 945                                 if (symbolslex.next(true)) {
 946                                         if (symbolslex.lineNumber() != lineno) {
 947                                                 // line in new format with mathcommand only
 948                                                 getNextToken = false;
 949                                         } else {
 950                                                 // line in new format with mathcommand and mathpreamble
 951                                                 mathpreamble = symbolslex.getString();
 952                                         }
 953                                 } else
 954                                         breakout = true;
 955                         }
 956                 } else {
 957                         breakout = true;
 958                 }
 959
 960                 // backward compatibility
 961                 if (mathpreamble == "esintoramsmath")
 962                         mathpreamble = "esint|amsmath";
 963
 964                 if (!textpreamble.empty())
 965                         if (textpreamble[0] != '\\')
 966                                 flags |= CharInfoTextFeature;
 967                 if (!mathpreamble.empty())
 968                         if (mathpreamble[0] != '\\')
 969                                 flags |= CharInfoMathFeature;
 970
 971                 CharInfo info = CharInfo(
 972                         textcommand, mathcommand,
 973                         textpreamble, mathpreamble,
 974                         tipashortcut, flags);
 975                 LYXERR(Debug::INFO, "Read unicode symbol " << symbol << " '"
 976                            << to_utf8(info.textcommand()) << "' '" << info.textpreamble()
 977                            << " '" << info.textfeature() << ' ' << info.textnotermination()
 978                            << ' ' << to_utf8(info.mathcommand()) << "' '" << info.mathpreamble()
 979                            << "' " << info.mathfeature() << ' ' << info.mathnotermination()
 980                            << ' ' << info.combining() << ' ' << info.force()
 981                            << ' ' << info.forceselected());
 982
 983                 // we assume that at least one command is nonempty when using unicodesymbols
 984                 if (info.isUnicodeSymbol()) {
 985                         unicodesymbols[symbol] = info;
 986                 }
 987
 988                 if (breakout)
 989                         break;
 990         }
 991
 992         // Now read the encodings
 993         enum {
 994                 et_encoding = 1,
 995                 et_end
 996         };
 997
 998         LexerKeyword encodingtags[] = {
 999                 { "encoding", et_encoding },
1000                 { "end", et_end }
1001         };
1002
1003         Lexer lex(encodingtags);
1004         lex.setFile(encfile);
1005         lex.setContext("Encodings::read");
1006         while (lex.isOK()) {
1007                 switch (lex.lex()) {
1008                 case et_encoding:
1009                 {
1010                         lex.next();
1011                         string const name = lex.getString();
1012                         lex.next();
1013                         string const latexname = lex.getString();
1014                         lex.next();
1015                         string const guiname = lex.getString();
1016                         lex.next();
1017                         string const iconvname = lex.getString();
1018                         lex.next();
1019                         string const width = lex.getString();
1020                         bool fixedwidth = false;
1021                         bool unsafe = false;
1022                         if (width == "fixed")
1023                                 fixedwidth = true;
1024                         else if (width == "variable")
1025                                 fixedwidth = false;
1026                         else if (width == "variableunsafe") {
1027                                 fixedwidth = false;
1028                                 unsafe = true;
1029                         }
1030                         else
1031                                 lex.printError("Unknown width");
1032
1033                         lex.next();
1034                         string const p = lex.getString();
1035                         Encoding::Package package = Encoding::none;
1036                         if (p == "none")
1037                                 package = Encoding::none;
1038                         else if (p == "inputenc")
1039                                 package = Encoding::inputenc;
1040                         else if (p == "CJK")
1041                                 package = Encoding::CJK;
1042                         else if (p == "japanese")
1043                                 package = Encoding::japanese;
1044                         else
1045                                 lex.printError("Unknown package");
1046
1047                         LYXERR(Debug::INFO, "Reading encoding " << name);
1048                         encodinglist[name] = Encoding(name, latexname,
1049                                 guiname, iconvname, fixedwidth, unsafe,
1050                                 package);
1051
1052                         if (lex.lex() != et_end)
1053                                 lex.printError("Missing end");
1054                         break;
1055                 }
1056                 case et_end:
1057                         lex.printError("Misplaced end");
1058                         break;
1059                 case Lexer::LEX_FEOF:
1060                         break;
1061                 default:
1062                         lex.printError("Unknown tag");
1063                         break;
1064                 }
1065         }
1066
1067         // Move all information from forcednotselected to forcedselected
1068         for (CharSetMap::const_iterator it1 = forcednotselected.begin(); it1 != forcednotselected.end(); ++it1) {
1069                 for (CharSetMap::iterator it2 = forcedselected.begin(); it2 != forcedselected.end(); ++it2) {
1070                         if (it2->first != it1->first)
1071                                 it2->second.insert(it1->second.begin(), it1->second.end());
1072                 }
1073         }
1074
1075 }
1076
1077
1078 } // namespace lyx