| 1 | /* |
| 2 | * iso2022s.c - support for ISO-2022 subset encodings. |
| 3 | * |
| 4 | * (The `s' suffix on the filename is there to leave `iso2022.c' |
| 5 | * free for the unlikely event that I ever attempt to implement |
| 6 | * _full_ ISO-2022 in this library!) |
| 7 | */ |
| 8 | |
| 9 | #ifndef ENUM_CHARSETS |
| 10 | |
| 11 | #include <stdio.h> |
| 12 | #include <string.h> |
| 13 | #include <assert.h> |
| 14 | |
| 15 | #include "charset.h" |
| 16 | #include "internal.h" |
| 17 | #include "sbcsdat.h" |
| 18 | |
| 19 | #define SO (0x0E) |
| 20 | #define SI (0x0F) |
| 21 | #define ESC (0x1B) |
| 22 | |
| 23 | /* Functional description of a single ISO 2022 escape sequence. */ |
| 24 | struct iso2022_escape { |
| 25 | char const *sequence; |
| 26 | unsigned long andbits, xorbits; |
| 27 | /* |
| 28 | * For output, these variables help us figure out which escape |
| 29 | * sequences we need to get where we want to be. |
| 30 | * |
| 31 | * `container' should be in the range 0-3, but can also be ORed |
| 32 | * with the bit flag RO to indicate that this is not a |
| 33 | * preferred container to use for this charset during output. |
| 34 | */ |
| 35 | int container, subcharset; |
| 36 | }; |
| 37 | #define RO 0x80 |
| 38 | |
| 39 | struct iso2022 { |
| 40 | /* |
| 41 | * List of escape sequences supported in this subset. Must be |
| 42 | * in ASCII order, so that we can narrow down the list as |
| 43 | * necessary. |
| 44 | */ |
| 45 | const struct iso2022_escape *escapes;/* must be sorted in ASCII order! */ |
| 46 | int nescapes; |
| 47 | |
| 48 | /* |
| 49 | * We assign indices from 0 upwards to the sub-charsets of a |
| 50 | * given ISO 2022 subset. nbytes[i] tells us how many bytes per |
| 51 | * character are required by sub-charset i. (It's a string |
| 52 | * mainly because that makes it easier to declare in C syntax |
| 53 | * than an int array.) |
| 54 | */ |
| 55 | char const *nbytes; |
| 56 | |
| 57 | /* |
| 58 | * The characters in this string are indices-plus-one (so that |
| 59 | * NUL can still terminate) of escape sequences in `escapes'. |
| 60 | * These escapes are output in the given sequence to reset the |
| 61 | * encoding state, unless it turns out that a given escape |
| 62 | * would not change the state at all. |
| 63 | */ |
| 64 | char const *reset; |
| 65 | |
| 66 | /* |
| 67 | * Initial value of s1, in case the default container contents |
| 68 | * needs to be something other than charset 0 in all cases. |
| 69 | * (Note that this must have the top bit set!) |
| 70 | */ |
| 71 | unsigned long s1; |
| 72 | |
| 73 | /* |
| 74 | * For output, some ISO 2022 subsets _mandate_ an initial shift |
| 75 | * sequence. If so, here it is so we can output it. (For the |
| 76 | * sake of basic sanity we won't bother to _require_ it on |
| 77 | * input, although it should of course be listed under |
| 78 | * `escapes' above so that we ignore it when present.) |
| 79 | */ |
| 80 | char const *initial_sequence; |
| 81 | |
| 82 | /* |
| 83 | * Is this an 8-bit ISO 2022 subset? |
| 84 | */ |
| 85 | int eightbit; |
| 86 | |
| 87 | /* |
| 88 | * Function calls to do the actual translation. |
| 89 | */ |
| 90 | long int (*to_ucs)(int subcharset, unsigned long bytes); |
| 91 | int (*from_ucs)(long int ucs, int *subcharset, unsigned long *bytes); |
| 92 | }; |
| 93 | |
| 94 | static void read_iso2022s(charset_spec const *charset, long int input_chr, |
| 95 | charset_state *state, |
| 96 | void (*emit)(void *ctx, long int output), |
| 97 | void *emitctx) |
| 98 | { |
| 99 | struct iso2022 const *iso = (struct iso2022 *)charset->data; |
| 100 | |
| 101 | /* |
| 102 | * For reading ISO-2022 subsets, we divide up our state |
| 103 | * variables as follows: |
| 104 | * |
| 105 | * - The top byte of s0 (bits 31:24) indicates, if nonzero, |
| 106 | * that we are part-way through a recognised ISO-2022 escape |
| 107 | * sequence. Five of those bits (31:27) give the index of |
| 108 | * the first member of the escapes list matching what we |
| 109 | * have so far; the remaining three (26:24) give the number |
| 110 | * of characters we have seen so far. |
| 111 | * |
| 112 | * - The top bit of s1 (bit 31) is non-zero at all times, to |
| 113 | * indicate that we have performed any necessary |
| 114 | * initialisation. When we start, we detect a zero s1 and |
| 115 | * respond to it by initialising the default container |
| 116 | * contents. |
| 117 | * |
| 118 | * - The next three bits of s1 (bits 30:28) indicate which |
| 119 | * _container_ is currently selected. This isn't quite as |
| 120 | * simple as it sounds, since we have to preserve memory of |
| 121 | * which of the SI/SO containers we came from when we're |
| 122 | * temporarily in SS2/SS3. Hence, what happens is: |
| 123 | * + bit 28 indicates SI/SO. |
| 124 | * + if we're in an SS2/SS3 container, that's indicated by |
| 125 | * the two bits above that being nonzero and holding |
| 126 | * either 2 or 3. |
| 127 | * + Hence: 0 is SI, 1 is SO, 4 is SS2-from-SI, 5 is |
| 128 | * SS2-from-SO, 6 is SS3-from-SI, 7 is SS3-from-SO. |
| 129 | * + For added fun: in an _8-bit_ ISO 2022 subset, we have |
| 130 | * the further special value 2, which means that we're |
| 131 | * theoretically in SI but the current character being |
| 132 | * accumulated is composed of 8-bit characters and will |
| 133 | * therefore be interpreted as if in SO. |
| 134 | * |
| 135 | * - The next nibble of s1 (27:24) indicates how many bytes |
| 136 | * have been accumulated in the current character. |
| 137 | * |
| 138 | * - The remaining three bytes of s1 are divided into four |
| 139 | * six-bit sections, and each section gives the current |
| 140 | * sub-charset selected in one of the possible containers. |
| 141 | * (Those containers are SI, SO, SS2 and SS3, respectively |
| 142 | * and in order from the bottom of s0 to the top.) |
| 143 | * |
| 144 | * - The bottom 24 bits of s0 give the accumulated character |
| 145 | * data so far. |
| 146 | * |
| 147 | * (Note that this means s1 contains all the parts of the state |
| 148 | * which might need to be operated on by escape sequences. |
| 149 | * Cunning, eh?) |
| 150 | */ |
| 151 | |
| 152 | if (!(state->s1 & 0x80000000)) { |
| 153 | state->s1 = iso->s1; |
| 154 | } |
| 155 | |
| 156 | /* |
| 157 | * So. Firstly, we process escape sequences, if we're in the |
| 158 | * middle of one or if we see a possible introducer (SI, SO, |
| 159 | * ESC). |
| 160 | */ |
| 161 | if ((state->s0 >> 24) || |
| 162 | (input_chr == SO || input_chr == SI || input_chr == ESC)) { |
| 163 | int n = (state->s0 >> 24) & 7, i = (state->s0 >> 27), oi = i, j; |
| 164 | |
| 165 | /* |
| 166 | * If this is the start of an escape sequence, we might be |
| 167 | * in mid-character. If so, clear the character state and |
| 168 | * emit an error token for the incomplete character. |
| 169 | */ |
| 170 | if (state->s1 & 0x0F000000) { |
| 171 | state->s1 &= ~0x0F000000; |
| 172 | state->s0 &= 0xFF000000; |
| 173 | /* |
| 174 | * If we were in the SS2 or SS3 container, we |
| 175 | * automatically exit it. |
| 176 | */ |
| 177 | if (state->s1 & 0x60000000) |
| 178 | state->s1 &= 0x9FFFFFFF; |
| 179 | emit(emitctx, ERROR); |
| 180 | } |
| 181 | |
| 182 | j = i; |
| 183 | while (j < iso->nescapes && |
| 184 | !memcmp(iso->escapes[j].sequence, |
| 185 | iso->escapes[oi].sequence, n)) { |
| 186 | if (iso->escapes[j].sequence[n] < input_chr) |
| 187 | i = ++j; |
| 188 | else |
| 189 | break; |
| 190 | } |
| 191 | if (i >= iso->nescapes || |
| 192 | memcmp(iso->escapes[i].sequence, |
| 193 | iso->escapes[oi].sequence, n) || |
| 194 | iso->escapes[i].sequence[n] != input_chr) { |
| 195 | /* |
| 196 | * This character does not appear in any valid escape |
| 197 | * sequence. Therefore, we must emit all the characters |
| 198 | * we had previously swallowed, plus this one, and |
| 199 | * return to non-escape-sequence state. |
| 200 | */ |
| 201 | for (j = 0; j < n; j++) |
| 202 | emit(emitctx, iso->escapes[oi].sequence[j]); |
| 203 | emit(emitctx, input_chr); |
| 204 | state->s0 = 0; |
| 205 | return; |
| 206 | } |
| 207 | |
| 208 | /* |
| 209 | * Otherwise, we have found an additional character in our |
| 210 | * escape sequence. See if we have reached the _end_ of our |
| 211 | * sequence (and therefore must process the sequence). |
| 212 | */ |
| 213 | n++; |
| 214 | if (!iso->escapes[i].sequence[n]) { |
| 215 | state->s0 = 0; |
| 216 | state->s1 &= iso->escapes[i].andbits; |
| 217 | state->s1 ^= iso->escapes[i].xorbits; |
| 218 | return; |
| 219 | } |
| 220 | |
| 221 | /* |
| 222 | * Failing _that_, we simply update our escape-sequence- |
| 223 | * tracking state. |
| 224 | */ |
| 225 | assert(i < 32 && n < 8); |
| 226 | state->s0 = (i << 27) | (n << 24); |
| 227 | return; |
| 228 | } |
| 229 | |
| 230 | /* |
| 231 | * If this isn't an escape sequence, it must be part of a |
| 232 | * character. One possibility is that it's a control character |
| 233 | * (00-20 or 7F-9F; also in non-8-bit ISO 2022 subsets I'm |
| 234 | * going to treat all top-half characters as controls), in |
| 235 | * which case we output it verbatim. |
| 236 | */ |
| 237 | if (input_chr < 0x21 || |
| 238 | (input_chr > 0x7E && (!iso->eightbit || input_chr < 0xA0))) { |
| 239 | /* |
| 240 | * We might be in mid-multibyte-character. If so, clear the |
| 241 | * character state and emit an error token for the |
| 242 | * incomplete character. |
| 243 | */ |
| 244 | if (state->s1 & 0x0F000000) { |
| 245 | state->s1 &= ~0x0F000000; |
| 246 | state->s0 &= 0xFF000000; |
| 247 | emit(emitctx, ERROR); |
| 248 | /* |
| 249 | * If we were in the SS2 or SS3 container, we |
| 250 | * automatically exit it. |
| 251 | */ |
| 252 | if (state->s1 & 0x60000000) |
| 253 | state->s1 &= 0x9FFFFFFF; |
| 254 | } |
| 255 | |
| 256 | emit(emitctx, input_chr); |
| 257 | return; |
| 258 | } |
| 259 | |
| 260 | /* |
| 261 | * Otherwise, accumulate character data. |
| 262 | */ |
| 263 | { |
| 264 | unsigned long chr; |
| 265 | int chrlen, cont, subcharset, bytes; |
| 266 | |
| 267 | /* |
| 268 | * Verify that we've seen the right kind of character for |
| 269 | * what we're currently doing. This only matters in 8-bit |
| 270 | * subsets. |
| 271 | */ |
| 272 | if (iso->eightbit) { |
| 273 | cont = (state->s1 >> 28) & 7; |
| 274 | /* |
| 275 | * If cont==0, we're entitled to see either GL or GR |
| 276 | * characters. If cont==2, we expect only GR; otherwise |
| 277 | * we expect only GL. |
| 278 | * |
| 279 | * If we see a GR character while cont==0, we set |
| 280 | * cont=2 immediately. |
| 281 | */ |
| 282 | if ((cont == 2 && !(input_chr & 0x80)) || |
| 283 | (cont != 0 && cont != 2 && (input_chr & 0x80))) { |
| 284 | /* |
| 285 | * Clear the previous character; it was prematurely |
| 286 | * terminated by this error. |
| 287 | */ |
| 288 | state->s1 &= ~0x0F000000; |
| 289 | state->s0 &= 0xFF000000; |
| 290 | emit(emitctx, ERROR); |
| 291 | /* |
| 292 | * If we were in the SS2 or SS3 container, we |
| 293 | * automatically exit it. |
| 294 | */ |
| 295 | if (state->s1 & 0x60000000) |
| 296 | state->s1 &= 0x9FFFFFFF; |
| 297 | } |
| 298 | |
| 299 | if (cont == 0 && (input_chr & 0x80)) { |
| 300 | state->s1 |= 0x20000000; |
| 301 | } |
| 302 | } |
| 303 | |
| 304 | /* The current character and its length. */ |
| 305 | chr = ((state->s0 & 0x00FFFFFF) << 8) | (input_chr & 0x7F); |
| 306 | chrlen = ((state->s1 >> 24) & 0xF) + 1; |
| 307 | /* The current sub-charset. */ |
| 308 | cont = (state->s1 >> 28) & 7; |
| 309 | if (cont > 1) cont >>= 1; |
| 310 | subcharset = (state->s1 >> (6*cont)) & 0x3F; |
| 311 | /* The number of bytes-per-character in that sub-charset. */ |
| 312 | bytes = iso->nbytes[subcharset]; |
| 313 | |
| 314 | /* |
| 315 | * If this character is now complete, we convert and emit |
| 316 | * it. Otherwise, we simply update the state and return. |
| 317 | */ |
| 318 | if (chrlen >= bytes) { |
| 319 | emit(emitctx, iso->to_ucs(subcharset, chr)); |
| 320 | chr = chrlen = 0; |
| 321 | /* |
| 322 | * If we were in the SS2 or SS3 container, we |
| 323 | * automatically exit it. |
| 324 | */ |
| 325 | if (state->s1 & 0x60000000) |
| 326 | state->s1 &= 0x9FFFFFFF; |
| 327 | } |
| 328 | state->s0 = (state->s0 & 0xFF000000) | chr; |
| 329 | state->s1 = (state->s1 & 0xF0FFFFFF) | (chrlen << 24); |
| 330 | } |
| 331 | } |
| 332 | |
| 333 | static int write_iso2022s(charset_spec const *charset, long int input_chr, |
| 334 | charset_state *state, |
| 335 | void (*emit)(void *ctx, long int output), |
| 336 | void *emitctx) |
| 337 | { |
| 338 | struct iso2022 const *iso = (struct iso2022 *)charset->data; |
| 339 | int subcharset, len, i, j, cont, topbit = 0; |
| 340 | unsigned long bytes; |
| 341 | |
| 342 | /* |
| 343 | * For output, our s1 state variable contains most of the same |
| 344 | * stuff as it did for input - initial-state indicator bit, |
| 345 | * current container, and current subcharset selected in each |
| 346 | * container. |
| 347 | */ |
| 348 | |
| 349 | /* |
| 350 | * Analyse the character and find out what subcharset it needs |
| 351 | * to go in. |
| 352 | */ |
| 353 | if (input_chr >= 0 && !iso->from_ucs(input_chr, &subcharset, &bytes)) |
| 354 | return FALSE; |
| 355 | |
| 356 | if (!(state->s1 & 0x80000000)) { |
| 357 | state->s1 = iso->s1; |
| 358 | if (iso->initial_sequence) |
| 359 | for (i = 0; iso->initial_sequence[i]; i++) |
| 360 | emit(emitctx, iso->initial_sequence[i]); |
| 361 | } |
| 362 | |
| 363 | if (input_chr == -1) { |
| 364 | unsigned long oldstate; |
| 365 | int k; |
| 366 | |
| 367 | /* |
| 368 | * Special case: reset encoding state. |
| 369 | */ |
| 370 | for (i = 0; iso->reset[i]; i++) { |
| 371 | j = iso->reset[i] - 1; |
| 372 | oldstate = state->s1; |
| 373 | state->s1 &= iso->escapes[j].andbits; |
| 374 | state->s1 ^= iso->escapes[j].xorbits; |
| 375 | if (state->s1 != oldstate) { |
| 376 | /* We must actually emit this sequence. */ |
| 377 | for (k = 0; iso->escapes[j].sequence[k]; k++) |
| 378 | emit(emitctx, iso->escapes[j].sequence[k]); |
| 379 | } |
| 380 | } |
| 381 | |
| 382 | return TRUE; |
| 383 | } |
| 384 | |
| 385 | /* |
| 386 | * Now begins the fun. We now know what subcharset we want. So |
| 387 | * we must find out which container we should select it into, |
| 388 | * select it into it if necessary, select that _container_ if |
| 389 | * necessary, and then output the given bytes. |
| 390 | */ |
| 391 | for (i = 0; i < iso->nescapes; i++) |
| 392 | if (iso->escapes[i].subcharset == subcharset && |
| 393 | !(iso->escapes[i].container & RO)) |
| 394 | break; |
| 395 | assert(i < iso->nescapes); |
| 396 | |
| 397 | /* |
| 398 | * We've found the escape sequence which would select this |
| 399 | * subcharset into a container. However, that subcharset might |
| 400 | * already _be_ selected in that container! Check before we go |
| 401 | * to the effort of emitting the sequence. |
| 402 | */ |
| 403 | cont = iso->escapes[i].container &~ RO; |
| 404 | if (((state->s1 >> (6*cont)) & 0x3F) != (unsigned)subcharset) { |
| 405 | for (j = 0; iso->escapes[i].sequence[j]; j++) |
| 406 | emit(emitctx, iso->escapes[i].sequence[j]); |
| 407 | state->s1 &= iso->escapes[i].andbits; |
| 408 | state->s1 ^= iso->escapes[i].xorbits; |
| 409 | } |
| 410 | |
| 411 | /* |
| 412 | * Now we know what container our subcharset is in, so we want |
| 413 | * to select that container. |
| 414 | */ |
| 415 | if (cont > 1) { |
| 416 | /* SS2 or SS3; just output the sequence and be done. */ |
| 417 | emit(emitctx, ESC); |
| 418 | emit(emitctx, 'L' + cont); /* comes out to 'N' or 'O' */ |
| 419 | } else { |
| 420 | /* |
| 421 | * Emit SI or SO, but only if the current container isn't already |
| 422 | * the right one. |
| 423 | * |
| 424 | * Also, in an 8-bit subset, we need not do this; we'll |
| 425 | * just use 8-bit characters to output SO-container |
| 426 | * characters. |
| 427 | */ |
| 428 | if (iso->eightbit && cont == 1 && ((state->s1 >> 28) & 7) == 0) { |
| 429 | topbit = 0x80; |
| 430 | } else if (((state->s1 >> 28) & 7) != (unsigned)cont) { |
| 431 | emit(emitctx, cont ? SO : SI); |
| 432 | state->s1 = (state->s1 & 0x8FFFFFFF) | (cont << 28); |
| 433 | } |
| 434 | } |
| 435 | |
| 436 | /* |
| 437 | * We're done. Subcharset is selected in container, container |
| 438 | * is selected. All we need now is to write out the bytes. |
| 439 | */ |
| 440 | len = iso->nbytes[subcharset]; |
| 441 | while (len--) |
| 442 | emit(emitctx, ((bytes >> (8*len)) & 0xFF) | topbit); |
| 443 | |
| 444 | return TRUE; |
| 445 | } |
| 446 | |
| 447 | /* |
| 448 | * ISO-2022-JP, defined in RFC 1468. |
| 449 | */ |
| 450 | static long int iso2022jp_to_ucs(int subcharset, unsigned long bytes) |
| 451 | { |
| 452 | switch (subcharset) { |
| 453 | case 0: return bytes; /* one-byte ASCII */ |
| 454 | case 1: /* JIS X 0201 half-width katakana */ |
| 455 | if (bytes >= 0x21 && bytes <= 0x5F) |
| 456 | return bytes + (0xFF61 - 0x21); |
| 457 | else |
| 458 | return ERROR; |
| 459 | /* (no break needed since all control paths have returned) */ |
| 460 | case 2: return jisx0208_to_unicode(((bytes >> 8) & 0xFF) - 0x21, |
| 461 | ((bytes ) & 0xFF) - 0x21); |
| 462 | default: return ERROR; |
| 463 | } |
| 464 | } |
| 465 | static int iso2022jp_from_ucs(long int ucs, int *subcharset, |
| 466 | unsigned long *bytes) |
| 467 | { |
| 468 | int r, c; |
| 469 | if (ucs < 0x80) { |
| 470 | *subcharset = 0; |
| 471 | *bytes = ucs; |
| 472 | return 1; |
| 473 | } else if (ucs >= 0xFF61 && ucs <= 0xFF9F) { |
| 474 | *subcharset = 1; |
| 475 | *bytes = ucs - (0xFF61 - 0x21); |
| 476 | return 1; |
| 477 | } else if (unicode_to_jisx0208(ucs, &r, &c)) { |
| 478 | *subcharset = 2; |
| 479 | *bytes = ((r+0x21) << 8) | (c+0x21); |
| 480 | return 1; |
| 481 | } else { |
| 482 | return 0; |
| 483 | } |
| 484 | } |
| 485 | static const struct iso2022_escape iso2022jp_escapes[] = { |
| 486 | {"\033$@", 0xFFFFFFC0, 0x00000002, -1, -1}, /* we ignore this one */ |
| 487 | {"\033$B", 0xFFFFFFC0, 0x00000002, 0, 2}, |
| 488 | {"\033(B", 0xFFFFFFC0, 0x00000000, 0, 0}, |
| 489 | {"\033(J", 0xFFFFFFC0, 0x00000001, 0, 1}, |
| 490 | }; |
| 491 | static const struct iso2022 iso2022jp = { |
| 492 | iso2022jp_escapes, lenof(iso2022jp_escapes), |
| 493 | "\1\1\2", "\3", 0x80000000, NULL, FALSE, |
| 494 | iso2022jp_to_ucs, iso2022jp_from_ucs |
| 495 | }; |
| 496 | const charset_spec charset_CS_ISO2022_JP = { |
| 497 | CS_ISO2022_JP, read_iso2022s, write_iso2022s, &iso2022jp |
| 498 | }; |
| 499 | |
| 500 | /* |
| 501 | * ISO-2022-KR, defined in RFC 1557. |
| 502 | */ |
| 503 | static long int iso2022kr_to_ucs(int subcharset, unsigned long bytes) |
| 504 | { |
| 505 | switch (subcharset) { |
| 506 | case 0: return bytes; /* one-byte ASCII */ |
| 507 | case 1: return ksx1001_to_unicode(((bytes >> 8) & 0xFF) - 0x21, |
| 508 | ((bytes ) & 0xFF) - 0x21); |
| 509 | default: return ERROR; |
| 510 | } |
| 511 | } |
| 512 | static int iso2022kr_from_ucs(long int ucs, int *subcharset, |
| 513 | unsigned long *bytes) |
| 514 | { |
| 515 | int r, c; |
| 516 | if (ucs < 0x80) { |
| 517 | *subcharset = 0; |
| 518 | *bytes = ucs; |
| 519 | return 1; |
| 520 | } else if (unicode_to_ksx1001(ucs, &r, &c)) { |
| 521 | *subcharset = 1; |
| 522 | *bytes = ((r+0x21) << 8) | (c+0x21); |
| 523 | return 1; |
| 524 | } else { |
| 525 | return 0; |
| 526 | } |
| 527 | } |
| 528 | static const struct iso2022_escape iso2022kr_escapes[] = { |
| 529 | {"\016", 0x8FFFFFFF, 0x10000000, -1, -1}, |
| 530 | {"\017", 0x8FFFFFFF, 0x00000000, 0, 0}, |
| 531 | {"\033$)C", 0xFFFFF03F, 0x00000040, 1, 1}, /* bits[11:6] <- 1 */ |
| 532 | }; |
| 533 | static const struct iso2022 iso2022kr = { |
| 534 | iso2022kr_escapes, lenof(iso2022kr_escapes), |
| 535 | "\1\2", "\2", 0x80000040, "\033$)C", FALSE, |
| 536 | iso2022kr_to_ucs, iso2022kr_from_ucs |
| 537 | }; |
| 538 | const charset_spec charset_CS_ISO2022_KR = { |
| 539 | CS_ISO2022_KR, read_iso2022s, write_iso2022s, &iso2022kr |
| 540 | }; |
| 541 | |
| 542 | /* |
| 543 | * The COMPOUND_TEXT encoding used in X selections. Defined by the |
| 544 | * X consortium. |
| 545 | * |
| 546 | * This encoding has quite a few sub-charsets. The order I assign |
| 547 | * to them here is given in an enum. |
| 548 | */ |
| 549 | enum { |
| 550 | /* This must match the bytes-per-character string given below. */ |
| 551 | CTEXT_ASCII, |
| 552 | CTEXT_JISX0201_LEFT, |
| 553 | CTEXT_JISX0201_RIGHT, |
| 554 | CTEXT_ISO8859_1, |
| 555 | CTEXT_ISO8859_2, |
| 556 | CTEXT_ISO8859_3, |
| 557 | CTEXT_ISO8859_4, |
| 558 | CTEXT_ISO8859_5, |
| 559 | CTEXT_ISO8859_6, |
| 560 | CTEXT_ISO8859_7, |
| 561 | CTEXT_ISO8859_8, |
| 562 | CTEXT_ISO8859_9, |
| 563 | CTEXT_GB2312, |
| 564 | CTEXT_KSC5601, |
| 565 | CTEXT_JISX0208, |
| 566 | CTEXT_JISX0212 |
| 567 | }; |
| 568 | static long int ctext_to_ucs(int subcharset, unsigned long bytes) |
| 569 | { |
| 570 | switch (subcharset) { |
| 571 | case CTEXT_ASCII: return bytes; /* one-byte ASCII */ |
| 572 | case CTEXT_JISX0201_LEFT: /* ASCII with yen and overline */ |
| 573 | return sbcs_to_unicode(&sbcsdata_CS_JISX0201, bytes & 0x7F); |
| 574 | case CTEXT_JISX0201_RIGHT: /* JIS X 0201 half-width katakana */ |
| 575 | return sbcs_to_unicode(&sbcsdata_CS_JISX0201, (bytes & 0x7F) | 0x80); |
| 576 | case CTEXT_ISO8859_1: |
| 577 | return sbcs_to_unicode(&sbcsdata_CS_ISO8859_1, (bytes & 0x7F) | 0x80); |
| 578 | case CTEXT_ISO8859_2: |
| 579 | return sbcs_to_unicode(&sbcsdata_CS_ISO8859_2, (bytes & 0x7F) | 0x80); |
| 580 | case CTEXT_ISO8859_3: |
| 581 | return sbcs_to_unicode(&sbcsdata_CS_ISO8859_3, (bytes & 0x7F) | 0x80); |
| 582 | case CTEXT_ISO8859_4: |
| 583 | return sbcs_to_unicode(&sbcsdata_CS_ISO8859_4, (bytes & 0x7F) | 0x80); |
| 584 | case CTEXT_ISO8859_5: |
| 585 | return sbcs_to_unicode(&sbcsdata_CS_ISO8859_5, (bytes & 0x7F) | 0x80); |
| 586 | case CTEXT_ISO8859_6: |
| 587 | return sbcs_to_unicode(&sbcsdata_CS_ISO8859_6, (bytes & 0x7F) | 0x80); |
| 588 | case CTEXT_ISO8859_7: |
| 589 | return sbcs_to_unicode(&sbcsdata_CS_ISO8859_7, (bytes & 0x7F) | 0x80); |
| 590 | case CTEXT_ISO8859_8: |
| 591 | return sbcs_to_unicode(&sbcsdata_CS_ISO8859_8, (bytes & 0x7F) | 0x80); |
| 592 | case CTEXT_ISO8859_9: |
| 593 | return sbcs_to_unicode(&sbcsdata_CS_ISO8859_9, (bytes & 0x7F) | 0x80); |
| 594 | case CTEXT_GB2312: |
| 595 | return gb2312_to_unicode(((bytes >> 8) & 0xFF) - 0x21, |
| 596 | ((bytes ) & 0xFF) - 0x21); |
| 597 | case CTEXT_KSC5601: |
| 598 | return ksx1001_to_unicode(((bytes >> 8) & 0xFF) - 0x21, |
| 599 | ((bytes ) & 0xFF) - 0x21); |
| 600 | case CTEXT_JISX0208: |
| 601 | return jisx0208_to_unicode(((bytes >> 8) & 0xFF) - 0x21, |
| 602 | ((bytes ) & 0xFF) - 0x21); |
| 603 | case CTEXT_JISX0212: |
| 604 | return jisx0212_to_unicode(((bytes >> 8) & 0xFF) - 0x21, |
| 605 | ((bytes ) & 0xFF) - 0x21); |
| 606 | default: return ERROR; |
| 607 | } |
| 608 | } |
| 609 | static int ctext_from_ucs(long int ucs, int *subcharset, unsigned long *bytes) |
| 610 | { |
| 611 | int r, c; |
| 612 | if (ucs < 0x80) { |
| 613 | *subcharset = CTEXT_ASCII; |
| 614 | *bytes = ucs; |
| 615 | return 1; |
| 616 | } else if ((c = sbcs_from_unicode(&sbcsdata_CS_ISO8859_1, ucs)) != ERROR) { |
| 617 | *subcharset = CTEXT_ISO8859_1; |
| 618 | *bytes = c - 0x80; |
| 619 | return 1; |
| 620 | } else if ((c = sbcs_from_unicode(&sbcsdata_CS_ISO8859_2, ucs)) != ERROR) { |
| 621 | *subcharset = CTEXT_ISO8859_2; |
| 622 | *bytes = c - 0x80; |
| 623 | return 1; |
| 624 | } else if ((c = sbcs_from_unicode(&sbcsdata_CS_ISO8859_3, ucs)) != ERROR) { |
| 625 | *subcharset = CTEXT_ISO8859_3; |
| 626 | *bytes = c - 0x80; |
| 627 | return 1; |
| 628 | } else if ((c = sbcs_from_unicode(&sbcsdata_CS_ISO8859_4, ucs)) != ERROR) { |
| 629 | *subcharset = CTEXT_ISO8859_4; |
| 630 | *bytes = c - 0x80; |
| 631 | return 1; |
| 632 | } else if ((c = sbcs_from_unicode(&sbcsdata_CS_ISO8859_5, ucs)) != ERROR) { |
| 633 | *subcharset = CTEXT_ISO8859_5; |
| 634 | *bytes = c - 0x80; |
| 635 | return 1; |
| 636 | } else if ((c = sbcs_from_unicode(&sbcsdata_CS_ISO8859_6, ucs)) != ERROR) { |
| 637 | *subcharset = CTEXT_ISO8859_6; |
| 638 | *bytes = c - 0x80; |
| 639 | return 1; |
| 640 | } else if ((c = sbcs_from_unicode(&sbcsdata_CS_ISO8859_7, ucs)) != ERROR) { |
| 641 | *subcharset = CTEXT_ISO8859_7; |
| 642 | *bytes = c - 0x80; |
| 643 | return 1; |
| 644 | } else if ((c = sbcs_from_unicode(&sbcsdata_CS_ISO8859_8, ucs)) != ERROR) { |
| 645 | *subcharset = CTEXT_ISO8859_8; |
| 646 | *bytes = c - 0x80; |
| 647 | return 1; |
| 648 | } else if ((c = sbcs_from_unicode(&sbcsdata_CS_ISO8859_9, ucs)) != ERROR) { |
| 649 | *subcharset = CTEXT_ISO8859_9; |
| 650 | *bytes = c - 0x80; |
| 651 | return 1; |
| 652 | } else if ((c = sbcs_from_unicode(&sbcsdata_CS_JISX0201, ucs)) != ERROR) { |
| 653 | if (c < 0x80) { |
| 654 | *subcharset = CTEXT_JISX0201_LEFT; |
| 655 | } else { |
| 656 | *subcharset = CTEXT_JISX0201_RIGHT; |
| 657 | c -= 0x80; |
| 658 | } |
| 659 | *bytes = c; |
| 660 | return 1; |
| 661 | } else if (unicode_to_gb2312(ucs, &r, &c)) { |
| 662 | *subcharset = CTEXT_GB2312; |
| 663 | *bytes = ((r+0x21) << 8) | (c+0x21); |
| 664 | return 1; |
| 665 | } else if (unicode_to_ksx1001(ucs, &r, &c)) { |
| 666 | *subcharset = CTEXT_KSC5601; |
| 667 | *bytes = ((r+0x21) << 8) | (c+0x21); |
| 668 | return 1; |
| 669 | } else if (unicode_to_jisx0208(ucs, &r, &c)) { |
| 670 | *subcharset = CTEXT_JISX0208; |
| 671 | *bytes = ((r+0x21) << 8) | (c+0x21); |
| 672 | return 1; |
| 673 | } else if (unicode_to_jisx0212(ucs, &r, &c)) { |
| 674 | *subcharset = CTEXT_JISX0212; |
| 675 | *bytes = ((r+0x21) << 8) | (c+0x21); |
| 676 | return 1; |
| 677 | } else { |
| 678 | return 0; |
| 679 | } |
| 680 | } |
| 681 | #define SEQ(str,cont,cs) \ |
| 682 | {str,~(63<<(6*(((cont)&~RO)))),(cs)<<(6*(((cont)&~RO))),(cont),(cs)} |
| 683 | /* |
| 684 | * Compound text defines restrictions on which container can take |
| 685 | * which character sets. Things labelled `left half of' can only go |
| 686 | * in GL; things labelled `right half of' can only go in GR; and 96 |
| 687 | * or 96^n character sets only _fit_ in GR. Thus: |
| 688 | * - ASCII can only go in GL since it is the left half of 8859-*. |
| 689 | * - All the 8859 sets can only go in GR. |
| 690 | * - JISX0201 left is GL only; JISX0201 right is GR only. |
| 691 | * - The three multibyte sets (GB2312, JISX0208, KSC5601) can go |
| 692 | * in either; we prefer GR where possible since this leads to a |
| 693 | * more compact EUC-like encoding. |
| 694 | */ |
| 695 | static const struct iso2022_escape ctext_escapes[] = { |
| 696 | SEQ("\033$(A", 0|RO, CTEXT_GB2312), |
| 697 | SEQ("\033$(B", 0|RO, CTEXT_JISX0208), |
| 698 | SEQ("\033$(C", 0|RO, CTEXT_KSC5601), |
| 699 | SEQ("\033$(D", 0|RO, CTEXT_JISX0212), |
| 700 | SEQ("\033$)A", 1, CTEXT_GB2312), |
| 701 | SEQ("\033$)B", 1, CTEXT_JISX0208), |
| 702 | SEQ("\033$)C", 1, CTEXT_KSC5601), |
| 703 | SEQ("\033$)D", 1, CTEXT_JISX0212), |
| 704 | SEQ("\033(B", 0, CTEXT_ASCII), |
| 705 | SEQ("\033(J", 0, CTEXT_JISX0201_LEFT), |
| 706 | SEQ("\033)I", 1, CTEXT_JISX0201_RIGHT), |
| 707 | SEQ("\033-A", 1, CTEXT_ISO8859_1), |
| 708 | SEQ("\033-B", 1, CTEXT_ISO8859_2), |
| 709 | SEQ("\033-C", 1, CTEXT_ISO8859_3), |
| 710 | SEQ("\033-D", 1, CTEXT_ISO8859_4), |
| 711 | SEQ("\033-F", 1, CTEXT_ISO8859_7), |
| 712 | SEQ("\033-G", 1, CTEXT_ISO8859_6), |
| 713 | SEQ("\033-H", 1, CTEXT_ISO8859_8), |
| 714 | SEQ("\033-L", 1, CTEXT_ISO8859_5), |
| 715 | SEQ("\033-M", 1, CTEXT_ISO8859_9), |
| 716 | }; |
| 717 | static const struct iso2022 ctext = { |
| 718 | ctext_escapes, lenof(ctext_escapes), |
| 719 | "\1\1\1\1\1\1\1\1\1\1\1\1\2\2\2\2", /* must match the enum above */ |
| 720 | "", 0x80000000 | (CTEXT_ASCII<<0) | (CTEXT_ISO8859_1<<6), "", TRUE, |
| 721 | ctext_to_ucs, ctext_from_ucs |
| 722 | }; |
| 723 | const charset_spec charset_CS_CTEXT = { |
| 724 | CS_CTEXT, read_iso2022s, write_iso2022s, &ctext |
| 725 | }; |
| 726 | |
| 727 | #else /* ENUM_CHARSETS */ |
| 728 | |
| 729 | ENUM_CHARSET(CS_ISO2022_JP) |
| 730 | ENUM_CHARSET(CS_ISO2022_KR) |
| 731 | ENUM_CHARSET(CS_CTEXT) |
| 732 | |
| 733 | #endif /* ENUM_CHARSETS */ |