src/encoding.C

   1 /* This file is part of
   2  * ======================================================
   3  *
   4  *           LyX, The Document Processor
   5  *
   6  *           Copyright 1995 Matthias Ettrich
   7  *           Copyright 1995-2001 The LyX Team.
   8  *
   9  * ====================================================== */
  10
  11 #include <config.h>
  12
  13 #include "encoding.h"
  14 #include "lyxlex.h"
  15 #include "debug.h"
  16
  17 #ifndef CXX_GLOBAL_CSTD
  18 using std::strtol;
  19 #endif
  20
  21 using std::endl;
  22
  23 Encodings encodings;
  24
  25 namespace {
  26
  27 Uchar tab_iso8859_1[256] = {
  28    0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, 0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f,
  29    0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17, 0x18, 0x19, 0x1a, 0x1b, 0x1c, 0x1d, 0x1e, 0x1f,
  30    0x20, 0x21, 0x22, 0x23, 0x24, 0x25, 0x26, 0x27, 0x28, 0x29, 0x2a, 0x2b, 0x2c, 0x2d, 0x2e, 0x2f,
  31    0x30, 0x31, 0x32, 0x33, 0x34, 0x35, 0x36, 0x37, 0x38, 0x39, 0x3a, 0x3b, 0x3c, 0x3d, 0x3e, 0x3f,
  32    0x40, 0x41, 0x42, 0x43, 0x44, 0x45, 0x46, 0x47, 0x48, 0x49, 0x4a, 0x4b, 0x4c, 0x4d, 0x4e, 0x4f,
  33    0x50, 0x51, 0x52, 0x53, 0x54, 0x55, 0x56, 0x57, 0x58, 0x59, 0x5a, 0x5b, 0x5c, 0x5d, 0x5e, 0x5f,
  34    0x60, 0x61, 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0x6a, 0x6b, 0x6c, 0x6d, 0x6e, 0x6f,
  35    0x70, 0x71, 0x72, 0x73, 0x74, 0x75, 0x76, 0x77, 0x78, 0x79, 0x7a, 0x7b, 0x7c, 0x7d, 0x7e, 0x7f,
  36    0x80, 0x81, 0x82, 0x83, 0x84, 0x85, 0x86, 0x87, 0x88, 0x89, 0x8a, 0x8b, 0x8c, 0x8d, 0x8e, 0x8f,
  37    0x90, 0x91, 0x92, 0x93, 0x94, 0x95, 0x96, 0x97, 0x98, 0x99, 0x9a, 0x9b, 0x9c, 0x9d, 0x9e, 0x9f,
  38    0xa0, 0xa1, 0xa2, 0xa3, 0xa4, 0xa5, 0xa6, 0xa7, 0xa8, 0xa9, 0xaa, 0xab, 0xac, 0xad, 0xae, 0xaf,
  39    0xb0, 0xb1, 0xb2, 0xb3, 0xb4, 0xb5, 0xb6, 0xb7, 0xb8, 0xb9, 0xba, 0xbb, 0xbc, 0xbd, 0xbe, 0xbf,
  40    0xc0, 0xc1, 0xc2, 0xc3, 0xc4, 0xc5, 0xc6, 0xc7, 0xc8, 0xc9, 0xca, 0xcb, 0xcc, 0xcd, 0xce, 0xcf,
  41    0xd0, 0xd1, 0xd2, 0xd3, 0xd4, 0xd5, 0xd6, 0xd7, 0xd8, 0xd9, 0xda, 0xdb, 0xdc, 0xdd, 0xde, 0xdf,
  42    0xe0, 0xe1, 0xe2, 0xe3, 0xe4, 0xe5, 0xe6, 0xe7, 0xe8, 0xe9, 0xea, 0xeb, 0xec, 0xed, 0xee, 0xef,
  43    0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff
  44 };
  45
  46 #ifdef USE_UNICODE_FOR_SYMBOLS
  47 Uchar tab_symbol[256] = {
  48    0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, 0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f,
  49    0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17, 0x18, 0x19, 0x1a, 0x1b, 0x1c, 0x1d, 0x1e, 0x1f,
  50
  51    0x0020, 0x0021, 0x2200, 0x0023, 0x2203, 0x0025, 0x0026, 0x220b,
  52    0x0028, 0x0029, 0x002a, 0x002b, 0x002c, 0x002d, 0x002e, 0x002f,
  53
  54    0x30, 0x31, 0x32, 0x33, 0x34, 0x35, 0x36, 0x37, 0x38, 0x39, 0x3a, 0x3b, 0x3c, 0x3d, 0x3e, 0x3f,
  55
  56 //   0x2245, 0x0391, 0x0392, 0x03a7, 0x0394, 0x0395, 0x03a6, 0x0393,
  57 //   0x0397, 0x0399, 0x03d1, 0x039a, 0x039b, 0x039c, 0x039d, 0x039f,
  58
  59 //   0x03a0, 0x0398, 0x03a1, 0x03a3, 0x03a4, 0x03a5, 0x03c2, 0x03a9,
  60 //   0x039e, 0x03a8, 0x0396, 0x005b, 0x2234, 0x005d, 0x22a5, 0xffff,
  61    0x2245, 0x0041, 0x0042, 0x0058, 0x0394, 0x0045, 0x03c2, 0x03a9,
  62    0x0048, 0x0049, 0x03d1, 0x004b, 0x039b, 0x004d, 0x004e, 0x004f,
  63
  64    0x03a0, 0x0398, 0x0050, 0x03a3, 0x0054, 0x0059, 0x03c2, 0x03a9,
  65    0x039e, 0x03a8, 0x005a, 0x005b, 0x2234, 0x005d, 0x22a5, 0xffff,
  66
  67    0xffff, 0x03b1, 0x03b2, 0x03c7, 0x03b4, 0x03b5, 0x03d5, 0x03b3,
  68    0x03b7, 0x03b9, 0x03c6, 0x03ba, 0x03bb, 0x03bc, 0x03bd, 0x03bf,
  69
  70    0x03c0, 0x03b8, 0x03c1, 0x03c3, 0x03c4, 0x03c5, 0x03d6, 0x03c9,
  71    0x03be, 0x03c8, 0x03b6, 0x007b, 0x007c, 0x007d, 0x007e, 0xffff,
  72
  73    0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff,
  74    0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff,
  75
  76    0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff,
  77    0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff,
  78
  79    0xffff, 0x03d2, 0x2032, 0x2264, 0x2215, 0x221e, 0xffff, 0x2663,
  80    0x2666, 0x2665, 0x2660, 0x2194, 0x2190, 0x2191, 0x2192, 0x2193,
  81
  82    0x00b0, 0x00b1, 0x2033, 0x2265, 0x00d7, 0x221d, 0x2202, 0x2022,
  83    0x00f7, 0x2260, 0x2261, 0x2248, 0x22ef, 0xffff, 0xffff, 0x21b5,
  84
  85    0x2135, 0x2111, 0x211c, 0x2118, 0x2297, 0x2295, 0x2205, 0x2229,
  86    0x222a, 0x2283, 0x2287, 0x2284, 0x2282, 0x2286, 0x2208, 0x2209,
  87
  88    0x2220, 0x2207, 0x00ae, 0x00a9, 0x2122, 0x220f, 0x221a, 0x22c5,
  89    0x00ac, 0x2227, 0x2228, 0x21d4, 0x21d0, 0x21d1, 0x21d2, 0x21d3,
  90
  91    0x2662, 0x2329, 0x00ae, 0x00a9, 0x2122, 0x2211, 0xffff, 0xffff,
  92    0xffff, 0x2308, 0xffff, 0x230a, 0xffff, 0xffff, 0xffff, 0xffff,
  93
  94    0xffff, 0x232a, 0x222b, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff,
  95    0xffff, 0x2309, 0xffff, 0x230b, 0xffff, 0xffff, 0xffff, 0xffff
  96 };
  97 #endif
  98
  99
 100 unsigned char arabic_table2[63][4] = {
 101         {0x41, 0x41, 0x41, 0x41}, // 0xc1 = hamza
 102         {0x42, 0xa1, 0x42, 0xa1}, // 0xc2 = ligature madda on alef
 103         {0x43, 0xa2, 0x43, 0xa2}, // 0xc3 = ligature hamza on alef
 104         {0x44, 0xa3, 0x44, 0xa3}, // 0xc4 = ligature hamza on waw
 105         {0x45, 0xa4, 0x45, 0xa4}, // 0xc5 = ligature hamza under alef
 106         {0x46, 0xf9, 0xf8, 0xa0}, // 0xc6 = ligature hamza on ya
 107         {0x47, 0xa5, 0x47, 0xa5}, // 0xc7 = alef
 108         {0x48, 0xae, 0xac, 0xad}, // 0xc8 = baa
 109         {0x49, 0xb1, 0x49, 0xb1}, // 0xc9 = taa marbuta
 110         {0x4a, 0xb4, 0xb2, 0xb3}, // 0xca = taa
 111         {0x4b, 0xb7, 0xb5, 0xb6}, // 0xcb = thaa
 112         {0x4c, 0xba, 0xb8, 0xb9}, // 0xcc = jeem
 113         {0x4d, 0xbd, 0xbb, 0xbc}, // 0xcd = haa
 114         {0x4e, 0xc0, 0xbe, 0xbf}, // 0xce = khaa
 115         {0x4f, 0xa6, 0x4f, 0xa6}, // 0xcf = dal
 116
 117         {0x50, 0xa7, 0x50, 0xa7}, // 0xd0 = thal
 118         {0x51, 0xa8, 0x51, 0xa8}, // 0xd1 = ra
 119         {0x52, 0xa9, 0x52, 0xa9}, // 0xd2 = zain
 120         {0x53, 0xc3, 0xc1, 0xc2}, // 0xd3 = seen
 121         {0x54, 0xc6, 0xc4, 0xc5}, // 0xd4 = sheen
 122         {0x55, 0xc9, 0xc7, 0xc8}, // 0xd5 = sad
 123         {0x56, 0xcc, 0xca, 0xcb}, // 0xd6 = dad
 124         {0x57, 0xcf, 0xcd, 0xce}, // 0xd7 = tah
 125         {0x58, 0xd2, 0xd0, 0xd1}, // 0xd8 = zah
 126         {0x59, 0xd5, 0xd3, 0xd4}, // 0xd9 = ain
 127         {0x5a, 0xd8, 0xd6, 0xd7}, // 0xda = ghain
 128         {0,0,0,0}, // 0xdb
 129         {0,0,0,0}, // 0xdc
 130         {0,0,0,0}, // 0xdd
 131         {0,0,0,0}, // 0xde
 132         {0,0,0,0}, // 0xdf
 133
 134         {0,0,0,0}, // 0xe0
 135         {0x61, 0xdb, 0xd9, 0xda}, // 0xe1 = fa
 136         {0x62, 0xde, 0xdc, 0xdd}, // 0xe2 = qaf
 137         {0x63, 0xe1, 0xdf, 0xe0}, // 0xe3 = kaf
 138         {0x64, 0xe4, 0xe2, 0xe3}, // 0xe4 = lam
 139         {0x65, 0xe7, 0xe5, 0xe6}, // 0xe5 = meem
 140         {0x66, 0xea, 0xe8, 0xe9}, // 0xe6 = noon
 141         {0x67, 0xed, 0xeb, 0xec}, // 0xe7 = ha
 142         {0x68, 0xaa, 0x68, 0xaa}, // 0xe8 = waw
 143         {0x69, 0xab, 0x69, 0xab}, // 0xe9 = alef maksura
 144         {0x6a, 0xf0, 0xee, 0xef}, // 0xea = ya
 145         {0x6b, 0x6b, 0x6b, 0x6b}, // 0xeb = fathatan
 146         {0x6c, 0x6c, 0x6c, 0x6c}, // 0xec = dammatan
 147         {0x6d, 0x6d, 0x6d, 0x6d}, // 0xed = kasratan
 148         {0x6e, 0x6e, 0x6e, 0x6e}, // 0xee = fatha
 149         {0x6f, 0x6f, 0x6f, 0x6f}, // 0xef = damma
 150
 151         {0x70, 0x70, 0x70, 0x70}, // 0xf0 = kasra
 152         {0x71, 0x71, 0x71, 0x71}, // 0xf1 = shadda
 153         {0x72, 0x72, 0x72, 0x72}, // 0xf2 = sukun
 154         {0,0,0,0}, // 0xf3
 155         {0,0,0,0}, // 0xf4
 156         {0,0,0,0}, // 0xf5
 157         {0,0,0,0}, // 0xf6
 158         {0,0,0,0}, // 0xf7
 159         {0,0,0,0}, // 0xf8
 160         {0,0,0,0}, // 0xf9
 161         {0,0,0,0}, // 0xfa
 162         {0,0,0,0}, // 0xfb
 163         {0,0,0,0}, // 0xfc
 164         {0,0,0,0}, // 0xfd
 165         {0,0,0,0}, // 0xfe
 166         {0,0,0,0}, // 0xff
 167 };
 168
 169
 170 unsigned char arabic_table[63][2] = {
 171         {0xc1, 0xc1}, // 0xc1 = hamza
 172         {0xc2, 0xc2}, // 0xc2 = ligature madda on alef
 173         {0xc3, 0xc3}, // 0xc3 = ligature hamza on alef
 174         {0xc4, 0xc4}, // 0xc4 = ligature hamza on waw
 175         {0xc5, 0xc5}, // 0xc5 = ligature hamza under alef
 176         {0xc6, 0xc0}, // 0xc6 = ligature hamza on ya
 177         {0xc7, 0xc7}, // 0xc7 = alef
 178         {0xc8, 0xeb}, // 0xc8 = baa
 179         {0xc9, 0xc9}, // 0xc9 = taa marbuta
 180         {0xca, 0xec}, // 0xca = taa
 181         {0xcb, 0xed}, // 0xcb = thaa
 182         {0xcc, 0xee}, // 0xcc = jeem
 183         {0xcd, 0xef}, // 0xcd = haa
 184         {0xce, 0xf0}, // 0xce = khaa
 185         {0xcf, 0xcf}, // 0xcf = dal
 186
 187         {0xd0, 0xd0}, // 0xd0 = thal
 188         {0xd1, 0xd1}, // 0xd1 = ra
 189         {0xd2, 0xd2}, // 0xd2 = zain
 190         {0xd3, 0xf1}, // 0xd3 = seen
 191         {0xd4, 0xf2}, // 0xd4 = sheen
 192         {0xd5, 0xf3}, // 0xd5 = sad
 193         {0xd6, 0xf4}, // 0xd6 = dad
 194         {0xd7, 0xd7}, // 0xd7 = tah
 195         {0xd8, 0xd8}, // 0xd8 = zah
 196         {0xd9, 0xf5}, // 0xd9 = ain
 197         {0xda, 0xf6}, // 0xda = ghain
 198         {0,0}, // 0xdb
 199         {0,0}, // 0xdc
 200         {0,0}, // 0xdd
 201         {0,0}, // 0xde
 202         {0,0}, // 0xdf
 203
 204         {0,0},  // 0xe0
 205         {0xe1, 0xf7},   // 0xe1 = fa
 206         {0xe2, 0xf8},   // 0xe2 = qaf
 207         {0xe3, 0xf9},   // 0xe3 = kaf
 208         {0xe4, 0xfa},   // 0xe4 = lam
 209         {0xe5, 0xfb},   // 0xe5 = meem
 210         {0xe6, 0xfc},   // 0xe6 = noon
 211         {0xe7, 0xfd},   // 0xe7 = ha
 212         {0xe8, 0xe8},   // 0xe8 = waw
 213         {0xe9, 0xe9},   // 0xe9 = alef maksura
 214         {0xea, 0xfe},   // 0xea = ya
 215         {0xa8, 0xa8},   // 0xeb = fathatan
 216         {0xa9, 0xa9},   // 0xec = dammatan
 217         {0xaa, 0xaa},   // 0xed = kasratan
 218         {0xab, 0xab},   // 0xee = fatha
 219         {0xac, 0xac},   // 0xef = damma
 220
 221         {0xad, 0xad},   // 0xf0 = kasra
 222         {0xae, 0xae},   // 0xf1 = shadda
 223         {0xaf, 0xaf},   // 0xf2 = sukun
 224         {0,0}, // 0xf3
 225         {0,0}, // 0xf4
 226         {0,0}, // 0xf5
 227         {0,0}, // 0xf6
 228         {0,0}, // 0xf7
 229         {0,0}, // 0xf8
 230         {0,0}, // 0xf9
 231         {0,0}, // 0xfa
 232         {0,0}, // 0xfb
 233         {0,0}, // 0xfc
 234         {0,0}, // 0xfd
 235         {0,0}, // 0xfe
 236         {0,0} // 0xff
 237 };
 238
 239
 240 unsigned char const arabic_start = 0xc1;
 241
 242 } // namespace anon
 243
 244
 245 bool Encodings::IsComposeChar_hebrew(unsigned char c)
 246 {
 247         return c <= 0xd2 && c >= 0xc0 &&
 248                 c != 0xce && c != 0xd0;
 249 }
 250
 251
 252 // Special Arabic letters are ones that do not get connected from left
 253 // they are hamza, alef_madda, alef_hamza, waw_hamza, alef_hamza_under,
 254 // alef, tah_marbota, dal, thal, rah, zai, wow, alef_maksoura
 255
 256 bool Encodings::is_arabic_special(unsigned char c)
 257 {
 258         return  (c >= 0xc1 && c <= 0xc5) ||
 259                  c == 0xc7 || c  == 0xc9  ||
 260                  c == 0xcf || c  == 0xe8  ||
 261                 (c >= 0xd0 && c <= 0xd2) ||
 262                  c == 0xe9;
 263 }
 264
 265 bool Encodings::IsComposeChar_arabic(unsigned char c)
 266 {
 267         return c >= 0xeb && c <= 0xf2;
 268 }
 269
 270
 271 bool Encodings::is_arabic(unsigned char c)
 272 {
 273         return c >= arabic_start && arabic_table[c-arabic_start][0];
 274 }
 275
 276
 277 unsigned char Encodings::TransformChar(unsigned char c,
 278                                       Encodings::Letter_Form form)
 279 {
 280         if (!is_arabic(c))
 281                 return c;
 282
 283         if (lyxrc.font_norm_type == LyXRC::ISO_10646_1)
 284                 return arabic_table2[c-arabic_start][form];
 285         else
 286                 return arabic_table[c-arabic_start][form >> 1];
 287 }
 288
 289
 290 Encoding const * Encodings::getEncoding(string const & encoding) const
 291 {
 292         EncodingList::const_iterator it = encodinglist.find(encoding);
 293         if (it != encodinglist.end())
 294                 return &it->second;
 295         else
 296                 return 0;
 297 }
 298
 299 Encodings::Encodings()
 300 {
 301         encodinglist["iso8859-1"] = Encoding("iso8859-1", "latin1", tab_iso8859_1);
 302         symbol_encoding_ =
 303 #ifdef USE_UNICODE_FOR_SYMBOLS
 304                 Encoding("symbol", "", tab_symbol);
 305 #else
 306                 Encoding("symbol", "", tab_iso8859_1);
 307 #endif
 308 }
 309
 310 void Encodings::read(string const & filename)
 311 {
 312         enum Encodingtags {
 313                 et_encoding = 1,
 314                 et_end,
 315                 et_last
 316         };
 317
 318         struct keyword_item encodingtags[et_last - 1] = {
 319                 { "encoding", et_encoding },
 320                 { "end", et_end }
 321         };
 322
 323         LyXLex lex(encodingtags, et_last - 1);
 324         lex.setFile(filename);
 325         while (lex.isOK()) {
 326                 switch (lex.lex()) {
 327                 case et_encoding:
 328                 {
 329                         lex.next();
 330                         string const name = lex.getString();
 331                         lex.next();
 332                         string const latexname = lex.getString();
 333                         lyxerr[Debug::INIT] << "Reading encoding " << name << endl;
 334                         Uchar table[256];
 335                         for (unsigned int i = 0; i < 256; ++i) {
 336                                 lex.next();
 337                                 string const tmp = lex.getString();
 338                                 table[i] = ::strtol(tmp.c_str(), 0 , 16);
 339                         }
 340                         encodinglist[name] = Encoding(name, latexname, table);
 341                         if (lex.lex() != et_end)
 342                                 lex.printError("Encodings::read: "
 343                                                "missing end");
 344                         break;
 345                 }
 346                 case et_end:
 347                         lex.printError("Encodings::read: Misplaced end");
 348                         break;
 349                 case LyXLex::LEX_FEOF:
 350                         break;
 351                 default:
 352                         lex.printError("Encodings::read: "
 353                                        "Unknown tag: `$$Token'");
 354                         break;
 355                 }
 356         }
 357 }