src/Encoding.cpp

   1 /**
   2  * \file Encoding.cpp
   3  * This file is part of LyX, the document processor.
   4  * Licence details can be found in the file COPYING.
   5  *
   6  * \author Lars Gullik Bjønnes
   7  * \author Jean-Marc Lasgouttes
   8  * \author Dekel Tsur
   9  *
  10  * Full author contact details are available in file CREDITS.
  11  */
  12
  13 #include <config.h>
  14
  15 #include "Encoding.h"
  16
  17 #include "Buffer.h"
  18 #include "BufferList.h"
  19 #include "InsetIterator.h"
  20 #include "LaTeXFeatures.h"
  21 #include "Lexer.h"
  22 #include "LyXRC.h"
  23
  24 #include "support/debug.h"
  25 #include "support/gettext.h"
  26 #include "support/FileName.h"
  27 #include "support/lstrings.h"
  28 #include "support/textutils.h"
  29 #include "support/unicode.h"
  30
  31 #include <boost/cstdint.hpp>
  32
  33 #include <sstream>
  34
  35 using namespace std;
  36 using namespace lyx::support;
  37
  38 namespace lyx {
  39
  40 int const Encoding::any = -1;
  41
  42 Encodings encodings;
  43
  44 Encodings::MathCommandSet Encodings::mathcmd;
  45 Encodings::TextCommandSet Encodings::textcmd;
  46 Encodings::MathSymbolSet  Encodings::mathsym;
  47
  48 namespace {
  49
  50 char_type arabic_table[172][4] = {
  51         {0xfe80, 0xfe80, 0xfe80, 0xfe80}, // 0x0621 = hamza
  52         {0xfe81, 0xfe82, 0xfe81, 0xfe82}, // 0x0622 = ligature madda on alef
  53         {0xfe83, 0xfe84, 0xfe83, 0xfe84}, // 0x0623 = ligature hamza on alef
  54         {0xfe85, 0xfe86, 0xfe85, 0xfe86}, // 0x0624 = ligature hamza on waw
  55         {0xfe87, 0xfe88, 0xfe87, 0xfe88}, // 0x0625 = ligature hamza under alef
  56         {0xfe89, 0xfe8a, 0xfe8b, 0xfe8c}, // 0x0626 = ligature hamza on ya
  57         {0xfe8d, 0xfe8e, 0xfe8d, 0xfe8e}, // 0x0627 = alef
  58         {0xfe8f, 0xfe90, 0xfe91, 0xfe92}, // 0x0628 = baa
  59         {0xfe93, 0xfe94, 0xfe93, 0xfe94}, // 0x0629 = taa marbuta
  60         {0xfe95, 0xfe96, 0xfe97, 0xfe98}, // 0x062a = taa
  61         {0xfe99, 0xfe9a, 0xfe9b, 0xfe9c}, // 0x062b = thaa
  62         {0xfe9d, 0xfe9e, 0xfe9f, 0xfea0}, // 0x062c = jeem
  63         {0xfea1, 0xfea2, 0xfea3, 0xfea4}, // 0x062d = haa
  64         {0xfea5, 0xfea6, 0xfea7, 0xfea8}, // 0x062e = khaa
  65         {0xfea9, 0xfeaa, 0xfea9, 0xfeaa}, // 0x062f = dal
  66
  67         {0xfeab, 0xfeac, 0xfeab, 0xfeac}, // 0x0630 = thal
  68         {0xfead, 0xfeae, 0xfead, 0xfeae}, // 0x0631 = ra
  69         {0xfeaf, 0xfeb0, 0xfeaf, 0xfeb0}, // 0x0632 = zain
  70         {0xfeb1, 0xfeb2, 0xfeb3, 0xfeb4}, // 0x0633 = seen
  71         {0xfeb5, 0xfeb6, 0xfeb7, 0xfeb8}, // 0x0634 = sheen
  72         {0xfeb9, 0xfeba, 0xfebb, 0xfebc}, // 0x0635 = sad
  73         {0xfebd, 0xfebe, 0xfebf, 0xfec0}, // 0x0636 = dad
  74         {0xfec1, 0xfec2, 0xfec3, 0xfec4}, // 0x0637 = tah
  75         {0xfec5, 0xfec6, 0xfec7, 0xfec8}, // 0x0638 = zah
  76         {0xfec9, 0xfeca, 0xfecb, 0xfecc}, // 0x0639 = ain
  77         {0xfecd, 0xfece, 0xfecf, 0xfed0}, // 0x063a = ghain
  78         {0, 0, 0, 0}, // 0x063b
  79         {0, 0, 0, 0}, // 0x063c
  80         {0, 0, 0, 0}, // 0x063d
  81         {0, 0, 0, 0}, // 0x063e
  82         {0, 0, 0, 0}, // 0x063f
  83
  84         {0, 0, 0, 0}, // 0x0640
  85         {0xfed1, 0xfed2, 0xfed3, 0xfed4}, // 0x0641 = fa
  86         {0xfed5, 0xfed6, 0xfed7, 0xfed8}, // 0x0642 = qaf
  87         {0xfed9, 0xfeda, 0xfedb, 0xfedc}, // 0x0643 = kaf
  88         {0xfedd, 0xfede, 0xfedf, 0xfee0}, // 0x0644 = lam
  89         {0xfee1, 0xfee2, 0xfee3, 0xfee4}, // 0x0645 = meem
  90         {0xfee5, 0xfee6, 0xfee7, 0xfee8}, // 0x0646 = noon
  91         {0xfee9, 0xfeea, 0xfeeb, 0xfeec}, // 0x0647 = ha
  92         {0xfeed, 0xfeee, 0xfeed, 0xfeee}, // 0x0648 = waw
  93         {0xfeef, 0xfef0, 0xfeef, 0xfef0}, // 0x0649 = alef maksura
  94         {0xfef1, 0xfef2, 0xfef3, 0xfef4}, // 0x064a = ya
  95         {0x065b, 0x065b, 0x065b, 0x065b}, // 0x064b = fathatan
  96         {0x065c, 0x065c, 0x065c, 0x065c}, // 0x064c = dammatan
  97         {0x064d, 0x064d, 0x064d, 0x064d}, // 0x064d = kasratan
  98         {0x064e, 0x064e, 0x064e, 0x064e}, // 0x064e = fatha
  99         {0x064f, 0x064f, 0x064f, 0x064f}, // 0x064f = damma
 100
 101         {0x0650, 0x0650, 0x0650, 0x0650}, // 0x0650 = kasra
 102         {0x0651, 0x0651, 0x0651, 0x0651}, // 0x0651 = shadda
 103         {0x0652, 0x0652, 0x0652, 0x0652}, // 0x0652 = sukun
 104
 105         {0, 0, 0, 0}, // 0x0653
 106         {0, 0, 0, 0}, // 0x0654
 107         {0, 0, 0, 0}, // 0x0655
 108         {0, 0, 0, 0}, // 0x0656
 109         {0, 0, 0, 0}, // 0x0657
 110         {0, 0, 0, 0}, // 0x0658
 111         {0, 0, 0, 0}, // 0x0659
 112         {0, 0, 0, 0}, // 0x065a
 113         {0, 0, 0, 0}, // 0x065b
 114         {0, 0, 0, 0}, // 0x065c
 115         {0, 0, 0, 0}, // 0x065d
 116         {0, 0, 0, 0}, // 0x065e
 117         {0, 0, 0, 0}, // 0x065f
 118         {0, 0, 0, 0}, // 0x0660
 119         {0, 0, 0, 0}, // 0x0661
 120         {0, 0, 0, 0}, // 0x0662
 121         {0, 0, 0, 0}, // 0x0663
 122         {0, 0, 0, 0}, // 0x0664
 123         {0, 0, 0, 0}, // 0x0665
 124         {0, 0, 0, 0}, // 0x0666
 125         {0, 0, 0, 0}, // 0x0667
 126         {0, 0, 0, 0}, // 0x0668
 127         {0, 0, 0, 0}, // 0x0669
 128         {0, 0, 0, 0}, // 0x066a
 129         {0, 0, 0, 0}, // 0x066b
 130         {0, 0, 0, 0}, // 0x066c
 131         {0, 0, 0, 0}, // 0x066d
 132         {0, 0, 0, 0}, // 0x066e
 133         {0, 0, 0, 0}, // 0x066f
 134         {0, 0, 0, 0}, // 0x0670
 135         {0, 0, 0, 0}, // 0x0671
 136         {0, 0, 0, 0}, // 0x0672
 137         {0, 0, 0, 0}, // 0x0673
 138         {0, 0, 0, 0}, // 0x0674
 139         {0, 0, 0, 0}, // 0x0675
 140         {0, 0, 0, 0}, // 0x0676
 141         {0, 0, 0, 0}, // 0x0677
 142         {0, 0, 0, 0}, // 0x0678
 143         {0, 0, 0, 0}, // 0x0679
 144         {0, 0, 0, 0}, // 0x067a
 145         {0, 0, 0, 0}, // 0x067b
 146         {0, 0, 0, 0}, // 0x067c
 147         {0, 0, 0, 0}, // 0x067d
 148         {0xfb56, 0xfb57, 0xfb58, 0xfb59}, // 0x067e = peh
 149         {0, 0, 0, 0}, // 0x067f
 150         {0, 0, 0, 0}, // 0x0680
 151         {0, 0, 0, 0}, // 0x0681
 152         {0, 0, 0, 0}, // 0x0682
 153         {0, 0, 0, 0}, // 0x0683
 154         {0, 0, 0, 0}, // 0x0684
 155         {0, 0, 0, 0}, // 0x0685
 156         {0xfb7a, 0xfb7b, 0xfb7c, 0xfb7d}, // 0x0686 = tcheh
 157         {0, 0, 0, 0}, // 0x0687
 158         {0, 0, 0, 0}, // 0x0688
 159         {0, 0, 0, 0}, // 0x0689
 160         {0, 0, 0, 0}, // 0x068a
 161         {0, 0, 0, 0}, // 0x068b
 162         {0, 0, 0, 0}, // 0x068c
 163         {0, 0, 0, 0}, // 0x068d
 164         {0, 0, 0, 0}, // 0x068e
 165         {0, 0, 0, 0}, // 0x068f
 166         {0, 0, 0, 0}, // 0x0690
 167         {0, 0, 0, 0}, // 0x0691
 168         {0, 0, 0, 0}, // 0x0692
 169         {0, 0, 0, 0}, // 0x0693
 170         {0, 0, 0, 0}, // 0x0694
 171         {0, 0, 0, 0}, // 0x0695
 172         {0, 0, 0, 0}, // 0x0696
 173         {0, 0, 0, 0}, // 0x0697
 174         {0xfb8a, 0xfb8b, 0xfb8a, 0xfb8b}, // 0x0698 = jeh
 175         {0, 0, 0, 0}, // 0x0699
 176         {0, 0, 0, 0}, // 0x069a
 177         {0, 0, 0, 0}, // 0x069b
 178         {0, 0, 0, 0}, // 0x069c
 179         {0, 0, 0, 0}, // 0x069d
 180         {0, 0, 0, 0}, // 0x069e
 181         {0, 0, 0, 0}, // 0x069f
 182         {0, 0, 0, 0}, // 0x06a0
 183         {0, 0, 0, 0}, // 0x06a1
 184         {0, 0, 0, 0}, // 0x06a2
 185         {0, 0, 0, 0}, // 0x06a3
 186         {0, 0, 0, 0}, // 0x06a4
 187         {0, 0, 0, 0}, // 0x06a5
 188         {0, 0, 0, 0}, // 0x06a6
 189         {0, 0, 0, 0}, // 0x06a7
 190         {0, 0, 0, 0}, // 0x06a8
 191         {0xfb8e, 0xfb8f, 0xfb90, 0xfb91}, // 0x06a9 = farsi kaf
 192         {0, 0, 0, 0}, // 0x06aa
 193         {0, 0, 0, 0}, // 0x06ab
 194         {0, 0, 0, 0}, // 0x06ac
 195         {0, 0, 0, 0}, // 0x06ad
 196         {0, 0, 0, 0}, // 0x06ae
 197         {0xfb92, 0xfb93, 0xfb94, 0xfb95}, // 0x06af = gaf
 198         {0, 0, 0, 0}, // 0x06b0
 199         {0, 0, 0, 0}, // 0x06b1
 200         {0, 0, 0, 0}, // 0x06b2
 201         {0, 0, 0, 0}, // 0x06b3
 202         {0, 0, 0, 0}, // 0x06b4
 203         {0, 0, 0, 0}, // 0x06b5
 204         {0, 0, 0, 0}, // 0x06b6
 205         {0, 0, 0, 0}, // 0x06b7
 206         {0, 0, 0, 0}, // 0x06b8
 207         {0, 0, 0, 0}, // 0x06b9
 208         {0, 0, 0, 0}, // 0x06ba
 209         {0, 0, 0, 0}, // 0x06bb
 210         {0, 0, 0, 0}, // 0x06bc
 211         {0, 0, 0, 0}, // 0x06bd
 212         {0, 0, 0, 0}, // 0x06be
 213         {0, 0, 0, 0}, // 0x06bf
 214         {0, 0, 0, 0}, // 0x06c0
 215         {0, 0, 0, 0}, // 0x06c1
 216         {0, 0, 0, 0}, // 0x06c2
 217         {0, 0, 0, 0}, // 0x06c3
 218         {0, 0, 0, 0}, // 0x06c4
 219         {0, 0, 0, 0}, // 0x06c5
 220         {0, 0, 0, 0}, // 0x06c6
 221         {0, 0, 0, 0}, // 0x06c7
 222         {0, 0, 0, 0}, // 0x06c8
 223         {0, 0, 0, 0}, // 0x06c9
 224         {0, 0, 0, 0}, // 0x06ca
 225         {0, 0, 0, 0}, // 0x06cb
 226         {0xfbfc, 0xfbfd, 0xfbfe, 0xfbff} // 0x06cc = farsi yeh
 227 };
 228
 229
 230 char_type const arabic_start = 0x0621;
 231 char_type const arabic_end = 0x06cc;
 232
 233
 234 enum CharInfoFlags {
 235         ///
 236         CharInfoCombining = 1,
 237         ///
 238         CharInfoTextFeature = 2,
 239         ///
 240         CharInfoMathFeature = 4,
 241         ///
 242         CharInfoForce = 8,
 243         ///
 244         CharInfoTextNoTermination = 16,
 245         ///
 246         CharInfoMathNoTermination = 32,
 247         ///
 248         CharInfoForceSelected = 64,
 249 };
 250
 251 /// Information about a single UCS4 character
 252 struct CharInfo {
 253         /// LaTeX command (text mode) for this character
 254         docstring textcommand;
 255         /// LaTeX command (math mode) for this character
 256         docstring mathcommand;
 257         /// Needed LaTeX preamble (or feature) for text mode
 258         string textpreamble;
 259         /// Needed LaTeX preamble (or feature) for math mode
 260         string mathpreamble;
 261         /// Is this a combining character?
 262         bool combining() const { return flags & CharInfoCombining ? true : false; }
 263         /// Is \c textpreamble a feature known by LaTeXFeatures, or a raw LaTeX
 264         /// command?
 265         bool textfeature() const { return flags & CharInfoTextFeature ? true : false; }
 266         /// Is \c mathpreamble a feature known by LaTeXFeatures, or a raw LaTeX
 267         /// command?
 268         bool mathfeature() const { return flags & CharInfoMathFeature ? true : false; }
 269         /// Always force the LaTeX command, even if the encoding contains
 270         /// this character?
 271         bool force() const { return flags & CharInfoForce ? true : false; }
 272         /// Force the LaTeX command for some encodings?
 273         bool forceselected() const { return flags & CharInfoForceSelected ? true : false; }
 274         /// TIPA shortcut
 275         string tipashortcut;
 276         /// \c textcommand needs no termination (such as {} or space).
 277         bool textnotermination() const { return flags & CharInfoTextNoTermination ? true : false; }
 278         /// \c mathcommand needs no termination (such as {} or space).
 279         bool mathnotermination() const { return flags & CharInfoMathNoTermination ? true : false; }
 280         ///
 281         unsigned int flags;
 282 };
 283
 284
 285 typedef map<char_type, CharInfo> CharInfoMap;
 286 CharInfoMap unicodesymbols;
 287
 288 typedef std::set<char_type> CharSet;
 289 typedef std::map<string, CharSet> CharSetMap;
 290 CharSet forced;
 291 CharSetMap forcedselected;
 292
 293 typedef std::set<char_type> MathAlphaSet;
 294 MathAlphaSet mathalpha;
 295
 296
 297 /// The highest code point in UCS4 encoding (1<<20 + 1<<16)
 298 char_type const max_ucs4 = 0x110000;
 299
 300 } // namespace anon
 301
 302
 303 EncodingException::EncodingException(char_type c)
 304         : failed_char(c), par_id(0), pos(0)
 305 {
 306 }
 307
 308
 309 const char * EncodingException::what() const throw()
 310 {
 311         return "Could not find LaTeX command for a character";
 312 }
 313
 314
 315 Encoding::Encoding(string const & n, string const & l, string const & g,
 316                    string const & i, bool f, bool u, Encoding::Package p)
 317         : name_(n), latexName_(l), guiName_(g), iconvName_(i), fixedwidth_(f),
 318           unsafe_(u), forced_(&forcedselected[n]), package_(p)
 319 {
 320         if (n == "ascii") {
 321                 // ASCII can encode 128 code points and nothing else
 322                 start_encodable_ = 128;
 323                 complete_ = true;
 324         } else if (i == "UTF-8") {
 325                 // UTF8 can encode all UCS4 code points
 326                 start_encodable_ = max_ucs4;
 327                 complete_ = true;
 328         } else {
 329                 start_encodable_ = 0;
 330                 complete_ = false;
 331         }
 332 }
 333
 334
 335 void Encoding::init() const
 336 {
 337         if (complete_)
 338                 return;
 339
 340         start_encodable_ = 0;
 341         // temporarily switch off lyxerr, since we will generate iconv errors
 342         lyxerr.disable();
 343         if (fixedwidth_) {
 344                 // We do not need to check all UCS4 code points, it is enough
 345                 // if we check all 256 code points of this encoding.
 346                 for (unsigned short j = 0; j < 256; ++j) {
 347                         char const c = char(j);
 348                         vector<char_type> const ucs4 = eightbit_to_ucs4(&c, 1, iconvName_);
 349                         if (ucs4.size() != 1)
 350                                 continue;
 351                         char_type const uc = ucs4[0];
 352                         CharInfoMap::const_iterator const it = unicodesymbols.find(uc);
 353                         if (it == unicodesymbols.end())
 354                                 encodable_.insert(uc);
 355                         else if (!it->second.force()) {
 356                                 if (forced_->empty() || forced_->find(uc) == forced_->end())
 357                                         encodable_.insert(uc);
 358                         }
 359                 }
 360         } else {
 361                 // We do not know how many code points this encoding has, and
 362                 // they do not have a direct representation as a single byte,
 363                 // therefore we need to check all UCS4 code points.
 364                 // This is expensive!
 365                 for (char_type c = 0; c < max_ucs4; ++c) {
 366                         vector<char> const eightbit = ucs4_to_eightbit(&c, 1, iconvName_);
 367                         if (!eightbit.empty()) {
 368                                 CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 369                                 if (it == unicodesymbols.end())
 370                                         encodable_.insert(c);
 371                                 else if (!it->second.force()) {
 372                                         if (forced_->empty() || forced_->find(c) == forced_->end())
 373                                                 encodable_.insert(c);
 374                                 }
 375                         }
 376                 }
 377         }
 378         lyxerr.enable();
 379         CharSet::iterator it = encodable_.find(start_encodable_);
 380         while (it != encodable_.end()) {
 381                 encodable_.erase(it);
 382                 ++start_encodable_;
 383                 it = encodable_.find(start_encodable_);
 384         }
 385         complete_ = true;
 386 }
 387
 388
 389 bool Encoding::isForced(char_type c) const
 390 {
 391         if (!forced.empty() && forced.find(c) != forced.end())
 392                 return true;
 393         return !forced_->empty() && forced_->find(c) != forced_->end();
 394 }
 395
 396
 397 bool Encoding::encodable(char_type c) const
 398 {
 399         // assure the used encoding is properly initialized
 400         init();
 401
 402         if (iconvName_ == "UTF-8" && package_ == none)
 403                 return true;
 404         if (c < start_encodable_ && !isForced(c))
 405                 return true;
 406         if (encodable_.find(c) != encodable_.end())
 407                 return true;
 408         return false;
 409 }
 410
 411
 412 pair<docstring, bool> Encoding::latexChar(char_type c) const
 413 {
 414         if (encodable(c))
 415                 return make_pair(docstring(1, c), false);
 416
 417         // c cannot (or should not) be encoded in this encoding
 418         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 419         if (it == unicodesymbols.end())
 420                 throw EncodingException(c);
 421         // at least one of mathcommand and textcommand is nonempty
 422         if (it->second.textcommand.empty())
 423                 return make_pair(
 424                         "\\ensuremath{" + it->second.mathcommand + '}', false);
 425         return make_pair(it->second.textcommand, !it->second.textnotermination());
 426 }
 427
 428
 429 pair<docstring, docstring> Encoding::latexString(docstring const input, bool dryrun) const
 430 {
 431         docstring result;
 432         docstring uncodable;
 433         bool terminate = false;
 434         for (size_t n = 0; n < input.size(); ++n) {
 435                 try {
 436                         char_type const c = input[n];
 437                         pair<docstring, bool> latex_char = latexChar(c);
 438                         docstring const latex = latex_char.first;
 439                         if (terminate && !prefixIs(latex, '\\')
 440                             && !prefixIs(latex, '{')
 441                             && !prefixIs(latex, '}')) {
 442                                         // Prevent eating of a following
 443                                         // space or command corruption by
 444                                         // following characters
 445                                         if (latex == " ")
 446                                                 result += "{}";
 447                                         else
 448                                                 result += " ";
 449                                 }
 450                         result += latex;
 451                         terminate = latex_char.second;
 452                 } catch (EncodingException & /* e */) {
 453                         LYXERR0("Uncodable character in latexString!");
 454                         if (dryrun) {
 455                                 result += "<" + _("LyX Warning: ")
 456                                            + _("uncodable character") + " '";
 457                                 result += docstring(1, input[n]);
 458                                 result += "'>";
 459                         } else
 460                                 uncodable += input[n];
 461                 }
 462         }
 463         return make_pair(result, uncodable);
 464 }
 465
 466
 467 vector<char_type> Encoding::symbolsList() const
 468 {
 469         // assure the used encoding is properly initialized
 470         init();
 471
 472         // first all encodable characters
 473         vector<char_type> symbols(encodable_.begin(), encodable_.end());
 474         // add those below start_encodable_
 475         for (char_type c = 0; c < start_encodable_; ++c)
 476                 symbols.push_back(c);
 477         // now the ones from the unicodesymbols file
 478         CharInfoMap::const_iterator const end = unicodesymbols.end();
 479         CharInfoMap::const_iterator it = unicodesymbols.begin();
 480         for (; it != end; ++it)
 481                 symbols.push_back(it->first);
 482         return symbols;
 483 }
 484
 485
 486 bool Encodings::latexMathChar(char_type c, bool mathmode,
 487                         Encoding const * encoding, docstring & command,
 488                         bool & needsTermination)
 489 {
 490         command = empty_docstring();
 491         if (encoding)
 492                 if (encoding->encodable(c))
 493                         command = docstring(1, c);
 494         needsTermination = false;
 495
 496         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 497         if (it == unicodesymbols.end()) {
 498                 if (!encoding || command.empty())
 499                         throw EncodingException(c);
 500                 if (mathmode)
 501                         addMathSym(c);
 502                 return false;
 503         }
 504         // at least one of mathcommand and textcommand is nonempty
 505         bool use_math = (mathmode && !it->second.mathcommand.empty()) ||
 506                         (!mathmode && it->second.textcommand.empty());
 507         if (use_math) {
 508                 command = it->second.mathcommand;
 509                 needsTermination = !it->second.mathnotermination();
 510                 addMathCmd(c);
 511         } else {
 512                 if (!encoding || command.empty()) {
 513                         command = it->second.textcommand;
 514                         needsTermination = !it->second.textnotermination();
 515                         addTextCmd(c);
 516                 }
 517                 if (mathmode)
 518                         addMathSym(c);
 519         }
 520         return use_math;
 521 }
 522
 523
 524 char_type Encodings::fromLaTeXCommand(docstring const & cmd, int cmdtype,
 525                 bool & combining, bool & needsTermination, set<string> * req)
 526 {
 527         CharInfoMap::const_iterator const end = unicodesymbols.end();
 528         CharInfoMap::const_iterator it = unicodesymbols.begin();
 529         for (combining = false; it != end; ++it) {
 530                 docstring const math = it->second.mathcommand;
 531                 docstring const text = it->second.textcommand;
 532                 if ((cmdtype & MATH_CMD) && math == cmd) {
 533                         combining = it->second.combining();
 534                         needsTermination = !it->second.mathnotermination();
 535                         if (req && it->second.mathfeature() &&
 536                             !it->second.mathpreamble.empty())
 537                                 req->insert(it->second.mathpreamble);
 538                         return it->first;
 539                 }
 540                 if ((cmdtype & TEXT_CMD) && text == cmd) {
 541                         combining = it->second.combining();
 542                         needsTermination = !it->second.textnotermination();
 543                         if (req && it->second.textfeature() &&
 544                             !it->second.textpreamble.empty())
 545                                 req->insert(it->second.textpreamble);
 546                         return it->first;
 547                 }
 548         }
 549         needsTermination = false;
 550         return 0;
 551 }
 552
 553
 554 docstring Encodings::fromLaTeXCommand(docstring const & cmd, int cmdtype,
 555                 bool & needsTermination, docstring & rem, set<string> * req)
 556 {
 557         needsTermination = false;
 558         rem = empty_docstring();
 559         bool const mathmode = cmdtype & MATH_CMD;
 560         bool const textmode = cmdtype & TEXT_CMD;
 561         docstring symbols;
 562         size_t i = 0;
 563         size_t const cmdend = cmd.size();
 564         CharInfoMap::const_iterator const uniend = unicodesymbols.end();
 565         for (size_t j = 0; j < cmdend; ++j) {
 566                 // Also get the char after a backslash
 567                 if (j + 1 < cmdend && cmd[j] == '\\')
 568                         ++j;
 569                 // If a macro argument follows, get it, too
 570                 if (j + 1 < cmdend && cmd[j + 1] == '{') {
 571                         size_t k = j + 1;
 572                         int count = 1;
 573                         while (k < cmdend && count && k != docstring::npos) {
 574                                 k = cmd.find_first_of(from_ascii("{}"), k + 1);
 575                                 if (cmd[k] == '{')
 576                                         ++count;
 577                                 else
 578                                         --count;
 579                         }
 580                         if (k != docstring::npos)
 581                                 j = k;
 582                 }
 583                 // Start with this substring and try augmenting it when it is
 584                 // the prefix of some command in the unicodesymbols file
 585                 docstring const subcmd = cmd.substr(i, j - i + 1);
 586
 587                 CharInfoMap::const_iterator it = unicodesymbols.begin();
 588                 size_t unicmd_size = 0;
 589                 char_type c = 0;
 590                 for (; it != uniend; ++it) {
 591                         docstring const math = mathmode ? it->second.mathcommand
 592                                                         : docstring();
 593                         docstring const text = textmode ? it->second.textcommand
 594                                                         : docstring();
 595                         size_t cur_size = max(math.size(), text.size());
 596                         // The current math or text unicode command cannot
 597                         // match, or we already matched a longer one
 598                         if (cur_size < subcmd.size() || cur_size <= unicmd_size)
 599                                 continue;
 600
 601                         docstring tmp = subcmd;
 602                         size_t k = j;
 603                         while (prefixIs(math, tmp) || prefixIs(text, tmp)) {
 604                                 ++k;
 605                                 if (k >= cmdend || cur_size <= tmp.size())
 606                                         break;
 607                                 tmp += cmd[k];
 608                         }
 609                         // No match
 610                         if (k == j)
 611                                 continue;
 612
 613                         // The last added char caused a mismatch, because
 614                         // we didn't exhaust the chars in cmd and didn't
 615                         // exceed the maximum size of the current unicmd
 616                         if (k < cmdend && cur_size > tmp.size())
 617                                 tmp.resize(tmp.size() - 1);
 618
 619                         // If this is an exact match, we found a (longer)
 620                         // matching entry in the unicodesymbols file.
 621                         // If the entry doesn't start with '\', we take note
 622                         // of the match and continue (this is not a ultimate
 623                         // acceptance, as some other entry may match a longer
 624                         // portion of the cmd string). However, if the entry
 625                         // does start with '\', we accept the match only if
 626                         // this is a valid macro, i.e., either it is a single
 627                         // (nonletter) char macro, or nothing else follows,
 628                         // or what follows is a nonletter char, or the last
 629                         // character is a }.
 630                         if ((math == tmp || text == tmp)
 631                             && (tmp[0] != '\\'
 632                                    || (tmp.size() == 2 && !isAlphaASCII(tmp[1]))
 633                                    || k == cmdend
 634                                    || !isAlphaASCII(cmd[k])
 635                                    || tmp[tmp.size() - 1] == '}')
 636                                  ) {
 637                                 c = it->first;
 638                                 j = k - 1;
 639                                 i = j + 1;
 640                                 unicmd_size = cur_size;
 641                                 if (math == tmp)
 642                                         needsTermination = !it->second.mathnotermination();
 643                                 else
 644                                         needsTermination = !it->second.textnotermination();
 645                                 if (req) {
 646                                         if (math == tmp && it->second.mathfeature() &&
 647                                             !it->second.mathpreamble.empty())
 648                                                 req->insert(it->second.mathpreamble);
 649                                         if (text == tmp && it->second.textfeature() &&
 650                                             !it->second.textpreamble.empty())
 651                                                 req->insert(it->second.textpreamble);
 652                                 }
 653                         }
 654                 }
 655                 if (unicmd_size)
 656                         symbols += c;
 657                 else if (j + 1 == cmdend) {
 658                         // No luck. Return what remains
 659                         rem = cmd.substr(i);
 660                         if (needsTermination && !rem.empty()) {
 661                                 if (rem.substr(0, 2) == "{}") {
 662                                         rem = rem.substr(2);
 663                                         needsTermination = false;
 664                                 } else if (rem[0] == ' ') {
 665                                         needsTermination = false;
 666                                         // LaTeX would swallow all spaces
 667                                         rem = ltrim(rem);
 668                                 }
 669                         }
 670                 }
 671         }
 672         return symbols;
 673 }
 674
 675
 676 void Encodings::initUnicodeMath(Buffer const & buffer, bool for_master)
 677 {
 678 #ifdef TEX2LYX
 679         // The code below is not needed in tex2lyx and requires additional stuff
 680         (void)buffer;
 681         (void)for_master;
 682 #else
 683         if (for_master) {
 684                 mathcmd.clear();
 685                 textcmd.clear();
 686                 mathsym.clear();
 687         }
 688
 689         // Check this buffer
 690         Inset & inset = buffer.inset();
 691         InsetIterator it = inset_iterator_begin(inset);
 692         InsetIterator const end = inset_iterator_end(inset);
 693         for (; it != end; ++it)
 694                 it->initUnicodeMath();
 695
 696         if (!for_master)
 697                 return;
 698
 699         // Check children
 700         ListOfBuffers blist = buffer.getDescendents();
 701         ListOfBuffers::const_iterator bit = blist.begin();
 702         ListOfBuffers::const_iterator const bend = blist.end();
 703         for (; bit != bend; ++bit)
 704                 initUnicodeMath(**bit, false);
 705 #endif
 706 }
 707
 708
 709 void Encodings::validate(char_type c, LaTeXFeatures & features, bool for_mathed)
 710 {
 711 #ifdef TEX2LYX
 712         // The code below is not needed in tex2lyx and requires additional stuff
 713         (void)c;
 714         (void)features;
 715         (void)for_mathed;
 716 #else
 717         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 718         if (it != unicodesymbols.end()) {
 719                 // In mathed, c could be used both in textmode and mathmode
 720                 bool const math_mode = for_mathed && isMathCmd(c);
 721                 bool const use_math = math_mode ||
 722                                       (!for_mathed && it->second.textcommand.empty());
 723                 bool const use_text = (for_mathed && isTextCmd(c)) ||
 724                                       (!for_mathed && !it->second.textcommand.empty());
 725                 bool const plain_utf8 = (features.runparams().encoding->name() == "utf8-plain");
 726                 bool const unicode_math = (features.isRequired("unicode-math")
 727                         && features.isAvailable("unicode-math"));
 728                 // with utf8-plain, we only load packages when in mathed (see #7766)
 729                 // and if we do not use unicode-math
 730                 if ((math_mode && !unicode_math)
 731                      || (use_math && !plain_utf8)) {
 732                         if (!it->second.mathpreamble.empty()) {
 733                                 if (it->second.mathfeature()) {
 734                                         string feats = it->second.mathpreamble;
 735                                         while (!feats.empty()) {
 736                                                 string feat;
 737                                                 feats = split(feats, feat, ',');
 738                                                 features.require(feat);
 739                                         }
 740                                 } else
 741                                         features.addPreambleSnippet(it->second.mathpreamble);
 742                         }
 743                 }
 744                 // with utf8-plain, we do not load packages (see #7766)
 745                 if (use_text && !plain_utf8) {
 746                         if (!it->second.textpreamble.empty()) {
 747                                 if (it->second.textfeature()) {
 748                                         string feats = it->second.textpreamble;
 749                                         while (!feats.empty()) {
 750                                                 string feat;
 751                                                 feats = split(feats, feat, ',');
 752                                                 features.require(feat);
 753                                         }
 754                                 } else
 755                                         features.addPreambleSnippet(it->second.textpreamble);
 756                         }
 757                 }
 758         }
 759         if (for_mathed && isMathSym(c)) {
 760                 features.require("amstext");
 761                 features.require("lyxmathsym");
 762         }
 763 #endif
 764 }
 765
 766
 767 bool Encodings::isHebrewComposeChar(char_type c)
 768 {
 769         return c <= 0x05c2 && c >= 0x05b0 && c != 0x05be && c != 0x05c0;
 770 }
 771
 772
 773 // Special Arabic letters are ones that do not get connected from left
 774 // they are hamza, alef_madda, alef_hamza, waw_hamza, alef_hamza_under,
 775 // alef, tah_marbota, dal, thal, rah, zai, wow, alef_maksoura
 776
 777 bool Encodings::isArabicSpecialChar(char_type c)
 778 {
 779         return (c >= 0x0621 && c <= 0x0625) || (c >= 0x0630 && c <= 0x0632)
 780                 || c == 0x0627 || c == 0x0629 || c == 0x062f || c == 0x0648
 781                 || c == 0x0649 || c == 0x0698;
 782 }
 783
 784
 785 bool Encodings::isArabicComposeChar(char_type c)
 786 {
 787         return c >= 0x064b && c <= 0x0652;
 788 }
 789
 790
 791 bool Encodings::isArabicChar(char_type c)
 792 {
 793         return c >= arabic_start && c <= arabic_end
 794                 && arabic_table[c-arabic_start][0];
 795 }
 796
 797
 798 char_type Encodings::transformChar(char_type c, Encodings::LetterForm form)
 799 {
 800         return isArabicChar(c) ? arabic_table[c-arabic_start][form] : c;
 801 }
 802
 803
 804 bool Encodings::isCombiningChar(char_type c)
 805 {
 806         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 807         if (it != unicodesymbols.end())
 808                 return it->second.combining();
 809         return false;
 810 }
 811
 812
 813 string const Encodings::TIPAShortcut(char_type c)
 814 {
 815         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 816         if (it != unicodesymbols.end())
 817                 return it->second.tipashortcut;
 818         return string();
 819 }
 820
 821
 822 bool Encodings::isKnownScriptChar(char_type const c, string & preamble)
 823 {
 824         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 825
 826         if (it == unicodesymbols.end())
 827                 return false;
 828
 829         if (it->second.textpreamble != "textgreek" && it->second.textpreamble != "textcyr")
 830                 return false;
 831
 832         if (preamble.empty()) {
 833                 preamble = it->second.textpreamble;
 834                 return true;
 835         }
 836         return it->second.textpreamble == preamble;
 837 }
 838
 839
 840 bool Encodings::isMathAlpha(char_type c)
 841 {
 842         return mathalpha.count(c);
 843 }
 844
 845
 846 Encoding const *
 847 Encodings::fromLyXName(string const & name, bool allowUnsafe) const
 848 {
 849         EncodingList::const_iterator const it = encodinglist.find(name);
 850         if (!allowUnsafe && it->second.unsafe())
 851                 return 0;
 852         return it != encodinglist.end() ? &it->second : 0;
 853 }
 854
 855
 856 Encoding const *
 857 Encodings::fromLaTeXName(string const & n, int const & p, bool allowUnsafe) const
 858 {
 859         string name = n;
 860         // FIXME: if we have to test for too many of these synonyms,
 861         // we should instead extend the format of lib/encodings
 862         if (n == "ansinew")
 863                 name = "cp1252";
 864
 865         // We don't use find_if because it makes copies of the pairs in
 866         // the map.
 867         // This linear search is OK since we don't have many encodings.
 868         // Users could even optimize it by putting the encodings they use
 869         // most at the top of lib/encodings.
 870         EncodingList::const_iterator const end = encodinglist.end();
 871         for (EncodingList::const_iterator it = encodinglist.begin(); it != end; ++it)
 872                 if ((it->second.latexName() == name) && (it->second.package() & p)
 873                                 && (!it->second.unsafe() || allowUnsafe))
 874                         return &it->second;
 875         return 0;
 876 }
 877
 878
 879 Encoding const *
 880 Encodings::fromIconvName(string const & n, int const & p, bool allowUnsafe) const
 881 {
 882         EncodingList::const_iterator const end = encodinglist.end();
 883         for (EncodingList::const_iterator it = encodinglist.begin(); it != end; ++it)
 884                 if ((it->second.iconvName() == n) && (it->second.package() & p)
 885                                 && (!it->second.unsafe() || allowUnsafe))
 886                         return &it->second;
 887         return 0;
 888 }
 889
 890
 891 Encodings::Encodings()
 892 {}
 893
 894
 895 void Encodings::read(FileName const & encfile, FileName const & symbolsfile)
 896 {
 897         // We must read the symbolsfile first, because the Encoding
 898         // constructor depends on it.
 899         Lexer symbolslex;
 900         symbolslex.setFile(symbolsfile);
 901         bool getNextToken = true;
 902         while (symbolslex.isOK()) {
 903                 char_type symbol;
 904                 CharInfo info;
 905                 string flags;
 906
 907                 if (getNextToken) {
 908                         if (!symbolslex.next(true))
 909                                 break;
 910                 } else
 911                         getNextToken = true;
 912
 913                 istringstream is(symbolslex.getString());
 914                 // reading symbol directly does not work if
 915                 // char_type == wchar_t.
 916                 boost::uint32_t tmp;
 917                 if(!(is >> hex >> tmp))
 918                         break;
 919                 symbol = tmp;
 920
 921                 if (!symbolslex.next(true))
 922                         break;
 923                 info.textcommand = symbolslex.getDocString();
 924                 if (!symbolslex.next(true))
 925                         break;
 926                 info.textpreamble = symbolslex.getString();
 927                 if (!symbolslex.next(true))
 928                         break;
 929                 flags = symbolslex.getString();
 930
 931                 info.flags = 0;
 932                 if (suffixIs(info.textcommand, '}'))
 933                         info.flags |= CharInfoTextNoTermination;
 934                 if (suffixIs(info.mathcommand, '}'))
 935                         info.flags |= CharInfoMathNoTermination;
 936                 while (!flags.empty()) {
 937                         string flag;
 938                         flags = split(flags, flag, ',');
 939                         if (flag == "combining") {
 940                                 info.flags |= CharInfoCombining;
 941                         } else if (flag == "force") {
 942                                 info.flags |= CharInfoForce;
 943                                 forced.insert(symbol);
 944                         } else if (prefixIs(flag, "force=")) {
 945                                 vector<string> encodings =
 946                                         getVectorFromString(flag.substr(6));
 947                                 for (size_t i = 0; i < encodings.size(); ++i)
 948                                         forcedselected[encodings[i]].insert(symbol);
 949                                 info.flags |= CharInfoForceSelected;
 950                         } else if (flag == "mathalpha") {
 951                                 mathalpha.insert(symbol);
 952                         } else if (flag == "notermination=text") {
 953                                 info.flags |= CharInfoTextNoTermination;
 954                         } else if (flag == "notermination=math") {
 955                                 info.flags |= CharInfoMathNoTermination;
 956                         } else if (flag == "notermination=both") {
 957                                 info.flags |= CharInfoTextNoTermination;
 958                                 info.flags |= CharInfoMathNoTermination;
 959                         } else if (flag == "notermination=none") {
 960                                 info.flags &= ~CharInfoTextNoTermination;
 961                                 info.flags &= ~CharInfoMathNoTermination;
 962                         } else if (contains(flag, "tipashortcut=")) {
 963                                 info.tipashortcut = split(flag, '=');
 964                         } else {
 965                                 lyxerr << "Ignoring unknown flag `" << flag
 966                                        << "' for symbol `0x"
 967                                        << hex << symbol << dec
 968                                        << "'." << endl;
 969                         }
 970                 }
 971                 // mathcommand and mathpreamble have been added for 1.6.0.
 972                 // make them optional so that old files still work.
 973                 int const lineno = symbolslex.lineNumber();
 974                 bool breakout = false;
 975                 if (symbolslex.next(true)) {
 976                         if (symbolslex.lineNumber() != lineno) {
 977                                 // line in old format without mathcommand and mathpreamble
 978                                 getNextToken = false;
 979                         } else {
 980                                 info.mathcommand = symbolslex.getDocString();
 981                                 if (symbolslex.next(true)) {
 982                                         if (symbolslex.lineNumber() != lineno) {
 983                                                 // line in new format with mathcommand only
 984                                                 getNextToken = false;
 985                                         } else {
 986                                                 // line in new format with mathcommand and mathpreamble
 987                                                 info.mathpreamble = symbolslex.getString();
 988                                         }
 989                                 } else
 990                                         breakout = true;
 991                         }
 992                 } else {
 993                         breakout = true;
 994                 }
 995
 996                 // backward compatibility
 997                 if (info.mathpreamble == "esintoramsmath")
 998                         info.mathpreamble = "esint|amsmath";
 999
1000                 if (!info.textpreamble.empty())
1001                         if (info.textpreamble[0] != '\\')
1002                                 info.flags |= CharInfoTextFeature;
1003                 if (!info.mathpreamble.empty())
1004                         if (info.mathpreamble[0] != '\\')
1005                                 info.flags |= CharInfoMathFeature;
1006
1007                 LYXERR(Debug::INFO, "Read unicode symbol " << symbol << " '"
1008                         << to_utf8(info.textcommand) << "' '" << info.textpreamble
1009                         << " '" << info.textfeature() << ' ' << info.textnotermination()
1010                         << ' ' << to_utf8(info.mathcommand) << "' '" << info.mathpreamble
1011                         << "' " << info.mathfeature() << ' ' << info.mathnotermination()
1012                         << ' ' << info.combining() << ' ' << info.force()
1013                         << ' ' << info.forceselected());
1014
1015                 // we assume that at least one command is nonempty when using unicodesymbols
1016                 if (!info.textcommand.empty() || !info.mathcommand.empty())
1017                         unicodesymbols[symbol] = info;
1018
1019                 if (breakout)
1020                         break;
1021         }
1022
1023         // Now read the encodings
1024         enum {
1025                 et_encoding = 1,
1026                 et_end
1027         };
1028
1029         LexerKeyword encodingtags[] = {
1030                 { "encoding", et_encoding },
1031                 { "end", et_end }
1032         };
1033
1034         Lexer lex(encodingtags);
1035         lex.setFile(encfile);
1036         lex.setContext("Encodings::read");
1037         while (lex.isOK()) {
1038                 switch (lex.lex()) {
1039                 case et_encoding:
1040                 {
1041                         lex.next();
1042                         string const name = lex.getString();
1043                         lex.next();
1044                         string const latexname = lex.getString();
1045                         lex.next();
1046                         string const guiname = lex.getString();
1047                         lex.next();
1048                         string const iconvname = lex.getString();
1049                         lex.next();
1050                         string const width = lex.getString();
1051                         bool fixedwidth = false;
1052                         bool unsafe = false;
1053                         if (width == "fixed")
1054                                 fixedwidth = true;
1055                         else if (width == "variable")
1056                                 fixedwidth = false;
1057                         else if (width == "variableunsafe") {
1058                                 fixedwidth = false;
1059                                 unsafe = true;
1060                         }
1061                         else
1062                                 lex.printError("Unknown width");
1063
1064                         lex.next();
1065                         string const p = lex.getString();
1066                         Encoding::Package package = Encoding::none;
1067                         if (p == "none")
1068                                 package = Encoding::none;
1069                         else if (p == "inputenc")
1070                                 package = Encoding::inputenc;
1071                         else if (p == "CJK")
1072                                 package = Encoding::CJK;
1073                         else if (p == "japanese")
1074                                 package = Encoding::japanese;
1075                         else
1076                                 lex.printError("Unknown package");
1077
1078                         LYXERR(Debug::INFO, "Reading encoding " << name);
1079                         encodinglist[name] = Encoding(name, latexname,
1080                                 guiname, iconvname, fixedwidth, unsafe,
1081                                 package);
1082
1083                         if (lex.lex() != et_end)
1084                                 lex.printError("Missing end");
1085                         break;
1086                 }
1087                 case et_end:
1088                         lex.printError("Misplaced end");
1089                         break;
1090                 case Lexer::LEX_FEOF:
1091                         break;
1092                 default:
1093                         lex.printError("Unknown tag");
1094                         break;
1095                 }
1096         }
1097 }
1098
1099
1100 } // namespace lyx