| 1 | /*
|
|---|
| 2 | * _codecs_cn.c: Codecs collection for Mainland Chinese encodings
|
|---|
| 3 | *
|
|---|
| 4 | * Written by Hye-Shik Chang <[email protected]>
|
|---|
| 5 | */
|
|---|
| 6 |
|
|---|
| 7 | #include "cjkcodecs.h"
|
|---|
| 8 | #include "mappings_cn.h"
|
|---|
| 9 |
|
|---|
| 10 | /**
|
|---|
| 11 | * hz is predefined as 100 on AIX. So we undefine it to avoid
|
|---|
| 12 | * conflict against hz codec's.
|
|---|
| 13 | */
|
|---|
| 14 | #ifdef _AIX
|
|---|
| 15 | #undef hz
|
|---|
| 16 | #endif
|
|---|
| 17 |
|
|---|
| 18 | /* GBK and GB2312 map differently in few codepoints that are listed below:
|
|---|
| 19 | *
|
|---|
| 20 | * gb2312 gbk
|
|---|
| 21 | * A1A4 U+30FB KATAKANA MIDDLE DOT U+00B7 MIDDLE DOT
|
|---|
| 22 | * A1AA U+2015 HORIZONTAL BAR U+2014 EM DASH
|
|---|
| 23 | * A844 undefined U+2015 HORIZONTAL BAR
|
|---|
| 24 | */
|
|---|
| 25 |
|
|---|
| 26 | #define GBK_DECODE(dc1, dc2, assi) \
|
|---|
| 27 | if ((dc1) == 0xa1 && (dc2) == 0xaa) (assi) = 0x2014; \
|
|---|
| 28 | else if ((dc1) == 0xa8 && (dc2) == 0x44) (assi) = 0x2015; \
|
|---|
| 29 | else if ((dc1) == 0xa1 && (dc2) == 0xa4) (assi) = 0x00b7; \
|
|---|
| 30 | else TRYMAP_DEC(gb2312, assi, dc1 ^ 0x80, dc2 ^ 0x80); \
|
|---|
| 31 | else TRYMAP_DEC(gbkext, assi, dc1, dc2);
|
|---|
| 32 |
|
|---|
| 33 | #define GBK_ENCODE(code, assi) \
|
|---|
| 34 | if ((code) == 0x2014) (assi) = 0xa1aa; \
|
|---|
| 35 | else if ((code) == 0x2015) (assi) = 0xa844; \
|
|---|
| 36 | else if ((code) == 0x00b7) (assi) = 0xa1a4; \
|
|---|
| 37 | else if ((code) != 0x30fb && TRYMAP_ENC_COND(gbcommon, assi, code));
|
|---|
| 38 |
|
|---|
| 39 | /*
|
|---|
| 40 | * GB2312 codec
|
|---|
| 41 | */
|
|---|
| 42 |
|
|---|
| 43 | ENCODER(gb2312)
|
|---|
| 44 | {
|
|---|
| 45 | while (inleft > 0) {
|
|---|
| 46 | Py_UNICODE c = IN1;
|
|---|
| 47 | DBCHAR code;
|
|---|
| 48 |
|
|---|
| 49 | if (c < 0x80) {
|
|---|
| 50 | WRITE1((unsigned char)c)
|
|---|
| 51 | NEXT(1, 1)
|
|---|
| 52 | continue;
|
|---|
| 53 | }
|
|---|
| 54 | UCS4INVALID(c)
|
|---|
| 55 |
|
|---|
| 56 | REQUIRE_OUTBUF(2)
|
|---|
| 57 | TRYMAP_ENC(gbcommon, code, c);
|
|---|
| 58 | else return 1;
|
|---|
| 59 |
|
|---|
| 60 | if (code & 0x8000) /* MSB set: GBK */
|
|---|
| 61 | return 1;
|
|---|
| 62 |
|
|---|
| 63 | OUT1((code >> 8) | 0x80)
|
|---|
| 64 | OUT2((code & 0xFF) | 0x80)
|
|---|
| 65 | NEXT(1, 2)
|
|---|
| 66 | }
|
|---|
| 67 |
|
|---|
| 68 | return 0;
|
|---|
| 69 | }
|
|---|
| 70 |
|
|---|
| 71 | DECODER(gb2312)
|
|---|
| 72 | {
|
|---|
| 73 | while (inleft > 0) {
|
|---|
| 74 | unsigned char c = **inbuf;
|
|---|
| 75 |
|
|---|
| 76 | REQUIRE_OUTBUF(1)
|
|---|
| 77 |
|
|---|
| 78 | if (c < 0x80) {
|
|---|
| 79 | OUT1(c)
|
|---|
| 80 | NEXT(1, 1)
|
|---|
| 81 | continue;
|
|---|
| 82 | }
|
|---|
| 83 |
|
|---|
| 84 | REQUIRE_INBUF(2)
|
|---|
| 85 | TRYMAP_DEC(gb2312, **outbuf, c ^ 0x80, IN2 ^ 0x80) {
|
|---|
| 86 | NEXT(2, 1)
|
|---|
| 87 | }
|
|---|
| 88 | else return 2;
|
|---|
| 89 | }
|
|---|
| 90 |
|
|---|
| 91 | return 0;
|
|---|
| 92 | }
|
|---|
| 93 |
|
|---|
| 94 |
|
|---|
| 95 | /*
|
|---|
| 96 | * GBK codec
|
|---|
| 97 | */
|
|---|
| 98 |
|
|---|
| 99 | ENCODER(gbk)
|
|---|
| 100 | {
|
|---|
| 101 | while (inleft > 0) {
|
|---|
| 102 | Py_UNICODE c = IN1;
|
|---|
| 103 | DBCHAR code;
|
|---|
| 104 |
|
|---|
| 105 | if (c < 0x80) {
|
|---|
| 106 | WRITE1((unsigned char)c)
|
|---|
| 107 | NEXT(1, 1)
|
|---|
| 108 | continue;
|
|---|
| 109 | }
|
|---|
| 110 | UCS4INVALID(c)
|
|---|
| 111 |
|
|---|
| 112 | REQUIRE_OUTBUF(2)
|
|---|
| 113 |
|
|---|
| 114 | GBK_ENCODE(c, code)
|
|---|
| 115 | else return 1;
|
|---|
| 116 |
|
|---|
| 117 | OUT1((code >> 8) | 0x80)
|
|---|
| 118 | if (code & 0x8000)
|
|---|
| 119 | OUT2((code & 0xFF)) /* MSB set: GBK */
|
|---|
| 120 | else
|
|---|
| 121 | OUT2((code & 0xFF) | 0x80) /* MSB unset: GB2312 */
|
|---|
| 122 | NEXT(1, 2)
|
|---|
| 123 | }
|
|---|
| 124 |
|
|---|
| 125 | return 0;
|
|---|
| 126 | }
|
|---|
| 127 |
|
|---|
| 128 | DECODER(gbk)
|
|---|
| 129 | {
|
|---|
| 130 | while (inleft > 0) {
|
|---|
| 131 | unsigned char c = IN1;
|
|---|
| 132 |
|
|---|
| 133 | REQUIRE_OUTBUF(1)
|
|---|
| 134 |
|
|---|
| 135 | if (c < 0x80) {
|
|---|
| 136 | OUT1(c)
|
|---|
| 137 | NEXT(1, 1)
|
|---|
| 138 | continue;
|
|---|
| 139 | }
|
|---|
| 140 |
|
|---|
| 141 | REQUIRE_INBUF(2)
|
|---|
| 142 |
|
|---|
| 143 | GBK_DECODE(c, IN2, **outbuf)
|
|---|
| 144 | else return 2;
|
|---|
| 145 |
|
|---|
| 146 | NEXT(2, 1)
|
|---|
| 147 | }
|
|---|
| 148 |
|
|---|
| 149 | return 0;
|
|---|
| 150 | }
|
|---|
| 151 |
|
|---|
| 152 |
|
|---|
| 153 | /*
|
|---|
| 154 | * GB18030 codec
|
|---|
| 155 | */
|
|---|
| 156 |
|
|---|
| 157 | ENCODER(gb18030)
|
|---|
| 158 | {
|
|---|
| 159 | while (inleft > 0) {
|
|---|
| 160 | ucs4_t c = IN1;
|
|---|
| 161 | DBCHAR code;
|
|---|
| 162 |
|
|---|
| 163 | if (c < 0x80) {
|
|---|
| 164 | WRITE1(c)
|
|---|
| 165 | NEXT(1, 1)
|
|---|
| 166 | continue;
|
|---|
| 167 | }
|
|---|
| 168 |
|
|---|
| 169 | DECODE_SURROGATE(c)
|
|---|
| 170 | if (c > 0x10FFFF)
|
|---|
| 171 | #if Py_UNICODE_SIZE == 2
|
|---|
| 172 | return 2; /* surrogates pair */
|
|---|
| 173 | #else
|
|---|
| 174 | return 1;
|
|---|
| 175 | #endif
|
|---|
| 176 | else if (c >= 0x10000) {
|
|---|
| 177 | ucs4_t tc = c - 0x10000;
|
|---|
| 178 |
|
|---|
| 179 | REQUIRE_OUTBUF(4)
|
|---|
| 180 |
|
|---|
| 181 | OUT4((unsigned char)(tc % 10) + 0x30)
|
|---|
| 182 | tc /= 10;
|
|---|
| 183 | OUT3((unsigned char)(tc % 126) + 0x81)
|
|---|
| 184 | tc /= 126;
|
|---|
| 185 | OUT2((unsigned char)(tc % 10) + 0x30)
|
|---|
| 186 | tc /= 10;
|
|---|
| 187 | OUT1((unsigned char)(tc + 0x90))
|
|---|
| 188 |
|
|---|
| 189 | #if Py_UNICODE_SIZE == 2
|
|---|
| 190 | NEXT(2, 4) /* surrogates pair */
|
|---|
| 191 | #else
|
|---|
| 192 | NEXT(1, 4)
|
|---|
| 193 | #endif
|
|---|
| 194 | continue;
|
|---|
| 195 | }
|
|---|
| 196 |
|
|---|
| 197 | REQUIRE_OUTBUF(2)
|
|---|
| 198 |
|
|---|
| 199 | GBK_ENCODE(c, code)
|
|---|
| 200 | else {
|
|---|
| 201 | const struct _gb18030_to_unibmp_ranges *utrrange;
|
|---|
| 202 |
|
|---|
| 203 | REQUIRE_OUTBUF(4)
|
|---|
| 204 |
|
|---|
| 205 | for (utrrange = gb18030_to_unibmp_ranges;
|
|---|
| 206 | utrrange->first != 0;
|
|---|
| 207 | utrrange++)
|
|---|
| 208 | if (utrrange->first <= c &&
|
|---|
| 209 | c <= utrrange->last) {
|
|---|
| 210 | Py_UNICODE tc;
|
|---|
| 211 |
|
|---|
| 212 | tc = c - utrrange->first +
|
|---|
| 213 | utrrange->base;
|
|---|
| 214 |
|
|---|
| 215 | OUT4((unsigned char)(tc % 10) + 0x30)
|
|---|
| 216 | tc /= 10;
|
|---|
| 217 | OUT3((unsigned char)(tc % 126) + 0x81)
|
|---|
| 218 | tc /= 126;
|
|---|
| 219 | OUT2((unsigned char)(tc % 10) + 0x30)
|
|---|
| 220 | tc /= 10;
|
|---|
| 221 | OUT1((unsigned char)tc + 0x81)
|
|---|
| 222 |
|
|---|
| 223 | NEXT(1, 4)
|
|---|
| 224 | break;
|
|---|
| 225 | }
|
|---|
| 226 |
|
|---|
| 227 | if (utrrange->first == 0)
|
|---|
| 228 | return 1;
|
|---|
| 229 | continue;
|
|---|
| 230 | }
|
|---|
| 231 |
|
|---|
| 232 | OUT1((code >> 8) | 0x80)
|
|---|
| 233 | if (code & 0x8000)
|
|---|
| 234 | OUT2((code & 0xFF)) /* MSB set: GBK or GB18030ext */
|
|---|
| 235 | else
|
|---|
| 236 | OUT2((code & 0xFF) | 0x80) /* MSB unset: GB2312 */
|
|---|
| 237 |
|
|---|
| 238 | NEXT(1, 2)
|
|---|
| 239 | }
|
|---|
| 240 |
|
|---|
| 241 | return 0;
|
|---|
| 242 | }
|
|---|
| 243 |
|
|---|
| 244 | DECODER(gb18030)
|
|---|
| 245 | {
|
|---|
| 246 | while (inleft > 0) {
|
|---|
| 247 | unsigned char c = IN1, c2;
|
|---|
| 248 |
|
|---|
| 249 | REQUIRE_OUTBUF(1)
|
|---|
| 250 |
|
|---|
| 251 | if (c < 0x80) {
|
|---|
| 252 | OUT1(c)
|
|---|
| 253 | NEXT(1, 1)
|
|---|
| 254 | continue;
|
|---|
| 255 | }
|
|---|
| 256 |
|
|---|
| 257 | REQUIRE_INBUF(2)
|
|---|
| 258 |
|
|---|
| 259 | c2 = IN2;
|
|---|
| 260 | if (c2 >= 0x30 && c2 <= 0x39) { /* 4 bytes seq */
|
|---|
| 261 | const struct _gb18030_to_unibmp_ranges *utr;
|
|---|
| 262 | unsigned char c3, c4;
|
|---|
| 263 | ucs4_t lseq;
|
|---|
| 264 |
|
|---|
| 265 | REQUIRE_INBUF(4)
|
|---|
| 266 | c3 = IN3;
|
|---|
| 267 | c4 = IN4;
|
|---|
| 268 | if (c < 0x81 || c3 < 0x81 || c4 < 0x30 || c4 > 0x39)
|
|---|
| 269 | return 4;
|
|---|
| 270 | c -= 0x81; c2 -= 0x30;
|
|---|
| 271 | c3 -= 0x81; c4 -= 0x30;
|
|---|
| 272 |
|
|---|
| 273 | if (c < 4) { /* U+0080 - U+FFFF */
|
|---|
| 274 | lseq = ((ucs4_t)c * 10 + c2) * 1260 +
|
|---|
| 275 | (ucs4_t)c3 * 10 + c4;
|
|---|
| 276 | if (lseq < 39420) {
|
|---|
| 277 | for (utr = gb18030_to_unibmp_ranges;
|
|---|
| 278 | lseq >= (utr + 1)->base;
|
|---|
| 279 | utr++) ;
|
|---|
| 280 | OUT1(utr->first - utr->base + lseq)
|
|---|
| 281 | NEXT(4, 1)
|
|---|
| 282 | continue;
|
|---|
| 283 | }
|
|---|
| 284 | }
|
|---|
| 285 | else if (c >= 15) { /* U+10000 - U+10FFFF */
|
|---|
| 286 | lseq = 0x10000 + (((ucs4_t)c-15) * 10 + c2)
|
|---|
| 287 | * 1260 + (ucs4_t)c3 * 10 + c4;
|
|---|
| 288 | if (lseq <= 0x10FFFF) {
|
|---|
| 289 | WRITEUCS4(lseq);
|
|---|
| 290 | NEXT_IN(4)
|
|---|
| 291 | continue;
|
|---|
| 292 | }
|
|---|
| 293 | }
|
|---|
| 294 | return 4;
|
|---|
| 295 | }
|
|---|
| 296 |
|
|---|
| 297 | GBK_DECODE(c, c2, **outbuf)
|
|---|
| 298 | else TRYMAP_DEC(gb18030ext, **outbuf, c, c2);
|
|---|
| 299 | else return 2;
|
|---|
| 300 |
|
|---|
| 301 | NEXT(2, 1)
|
|---|
| 302 | }
|
|---|
| 303 |
|
|---|
| 304 | return 0;
|
|---|
| 305 | }
|
|---|
| 306 |
|
|---|
| 307 |
|
|---|
| 308 | /*
|
|---|
| 309 | * HZ codec
|
|---|
| 310 | */
|
|---|
| 311 |
|
|---|
| 312 | ENCODER_INIT(hz)
|
|---|
| 313 | {
|
|---|
| 314 | state->i = 0;
|
|---|
| 315 | return 0;
|
|---|
| 316 | }
|
|---|
| 317 |
|
|---|
| 318 | ENCODER_RESET(hz)
|
|---|
| 319 | {
|
|---|
| 320 | if (state->i != 0) {
|
|---|
| 321 | WRITE2('~', '}')
|
|---|
| 322 | state->i = 0;
|
|---|
| 323 | NEXT_OUT(2)
|
|---|
| 324 | }
|
|---|
| 325 | return 0;
|
|---|
| 326 | }
|
|---|
| 327 |
|
|---|
| 328 | ENCODER(hz)
|
|---|
| 329 | {
|
|---|
| 330 | while (inleft > 0) {
|
|---|
| 331 | Py_UNICODE c = IN1;
|
|---|
| 332 | DBCHAR code;
|
|---|
| 333 |
|
|---|
| 334 | if (c < 0x80) {
|
|---|
| 335 | if (state->i == 0) {
|
|---|
| 336 | WRITE1((unsigned char)c)
|
|---|
| 337 | NEXT(1, 1)
|
|---|
| 338 | }
|
|---|
| 339 | else {
|
|---|
| 340 | WRITE3('~', '}', (unsigned char)c)
|
|---|
| 341 | NEXT(1, 3)
|
|---|
| 342 | state->i = 0;
|
|---|
| 343 | }
|
|---|
| 344 | continue;
|
|---|
| 345 | }
|
|---|
| 346 |
|
|---|
| 347 | UCS4INVALID(c)
|
|---|
| 348 |
|
|---|
| 349 | TRYMAP_ENC(gbcommon, code, c);
|
|---|
| 350 | else return 1;
|
|---|
| 351 |
|
|---|
| 352 | if (code & 0x8000) /* MSB set: GBK */
|
|---|
| 353 | return 1;
|
|---|
| 354 |
|
|---|
| 355 | if (state->i == 0) {
|
|---|
| 356 | WRITE4('~', '{', code >> 8, code & 0xff)
|
|---|
| 357 | NEXT(1, 4)
|
|---|
| 358 | state->i = 1;
|
|---|
| 359 | }
|
|---|
| 360 | else {
|
|---|
| 361 | WRITE2(code >> 8, code & 0xff)
|
|---|
| 362 | NEXT(1, 2)
|
|---|
| 363 | }
|
|---|
| 364 | }
|
|---|
| 365 |
|
|---|
| 366 | return 0;
|
|---|
| 367 | }
|
|---|
| 368 |
|
|---|
| 369 | DECODER_INIT(hz)
|
|---|
| 370 | {
|
|---|
| 371 | state->i = 0;
|
|---|
| 372 | return 0;
|
|---|
| 373 | }
|
|---|
| 374 |
|
|---|
| 375 | DECODER_RESET(hz)
|
|---|
| 376 | {
|
|---|
| 377 | state->i = 0;
|
|---|
| 378 | return 0;
|
|---|
| 379 | }
|
|---|
| 380 |
|
|---|
| 381 | DECODER(hz)
|
|---|
| 382 | {
|
|---|
| 383 | while (inleft > 0) {
|
|---|
| 384 | unsigned char c = IN1;
|
|---|
| 385 |
|
|---|
| 386 | if (c == '~') {
|
|---|
| 387 | unsigned char c2 = IN2;
|
|---|
| 388 |
|
|---|
| 389 | REQUIRE_INBUF(2)
|
|---|
| 390 | if (c2 == '~') {
|
|---|
| 391 | WRITE1('~')
|
|---|
| 392 | NEXT(2, 1)
|
|---|
| 393 | continue;
|
|---|
| 394 | }
|
|---|
| 395 | else if (c2 == '{' && state->i == 0)
|
|---|
| 396 | state->i = 1; /* set GB */
|
|---|
| 397 | else if (c2 == '}' && state->i == 1)
|
|---|
| 398 | state->i = 0; /* set ASCII */
|
|---|
| 399 | else if (c2 == '\n')
|
|---|
| 400 | ; /* line-continuation */
|
|---|
| 401 | else
|
|---|
| 402 | return 2;
|
|---|
| 403 | NEXT(2, 0);
|
|---|
| 404 | continue;
|
|---|
| 405 | }
|
|---|
| 406 |
|
|---|
| 407 | if (c & 0x80)
|
|---|
| 408 | return 1;
|
|---|
| 409 |
|
|---|
| 410 | if (state->i == 0) { /* ASCII mode */
|
|---|
| 411 | WRITE1(c)
|
|---|
| 412 | NEXT(1, 1)
|
|---|
| 413 | }
|
|---|
| 414 | else { /* GB mode */
|
|---|
| 415 | REQUIRE_INBUF(2)
|
|---|
| 416 | REQUIRE_OUTBUF(1)
|
|---|
| 417 | TRYMAP_DEC(gb2312, **outbuf, c, IN2) {
|
|---|
| 418 | NEXT(2, 1)
|
|---|
| 419 | }
|
|---|
| 420 | else
|
|---|
| 421 | return 2;
|
|---|
| 422 | }
|
|---|
| 423 | }
|
|---|
| 424 |
|
|---|
| 425 | return 0;
|
|---|
| 426 | }
|
|---|
| 427 |
|
|---|
| 428 |
|
|---|
| 429 | BEGIN_MAPPINGS_LIST
|
|---|
| 430 | MAPPING_DECONLY(gb2312)
|
|---|
| 431 | MAPPING_DECONLY(gbkext)
|
|---|
| 432 | MAPPING_ENCONLY(gbcommon)
|
|---|
| 433 | MAPPING_ENCDEC(gb18030ext)
|
|---|
| 434 | END_MAPPINGS_LIST
|
|---|
| 435 |
|
|---|
| 436 | BEGIN_CODECS_LIST
|
|---|
| 437 | CODEC_STATELESS(gb2312)
|
|---|
| 438 | CODEC_STATELESS(gbk)
|
|---|
| 439 | CODEC_STATELESS(gb18030)
|
|---|
| 440 | CODEC_STATEFUL(hz)
|
|---|
| 441 | END_CODECS_LIST
|
|---|
| 442 |
|
|---|
| 443 | I_AM_A_MODULE_FOR(cn)
|
|---|