src/locale/iconv.c

   1 #include <iconv.h>
   2 #include <errno.h>
   3 #include <wchar.h>
   4 #include <string.h>
   5 #include <stdlib.h>
   6 #include <limits.h>
   7 #include <stdint.h>
   8 #include "locale_impl.h"
   9
  10 #define UTF_32BE    0300
  11 #define UTF_16LE    0301
  12 #define UTF_16BE    0302
  13 #define UTF_32LE    0303
  14 #define UCS2BE      0304
  15 #define UCS2LE      0305
  16 #define WCHAR_T     0306
  17 #define US_ASCII    0307
  18 #define UTF_8       0310
  19 #define EUC_JP      0320
  20 #define SHIFT_JIS   0321
  21 #define GB18030     0330
  22 #define GBK         0331
  23 #define GB2312      0332
  24 #define BIG5        0340
  25 #define EUC_KR      0350
  26
  27 /* Definitions of charmaps. Each charmap consists of:
  28  * 1. Empty-string-terminated list of null-terminated aliases.
  29  * 2. Special type code or number of elided entries.
  30  * 3. Character table (size determined by field 2). */
  31
  32 static const unsigned char charmaps[] =
  33 "utf8\0char\0\0\310"
  34 "wchart\0\0\306"
  35 "ucs2\0ucs2be\0\0\304"
  36 "ucs2le\0\0\305"
  37 "utf16\0utf16be\0\0\302"
  38 "utf16le\0\0\301"
  39 "ucs4\0ucs4be\0utf32\0utf32be\0\0\300"
  40 "ucs4le\0utf32le\0\0\303"
  41 "ascii\0usascii\0iso646\0iso646us\0\0\307"
  42 "eucjp\0\0\320"
  43 "shiftjis\0sjis\0\0\321"
  44 "gb18030\0\0\330"
  45 "gbk\0\0\331"
  46 "gb2312\0\0\332"
  47 "big5\0bigfive\0cp950\0big5hkscs\0\0\340"
  48 "euckr\0ksc5601\0ksx1001\0cp949\0\0\350"
  49 #include "codepages.h"
  50 ;
  51
  52 static const unsigned short legacy_chars[] = {
  53 #include "legacychars.h"
  54 };
  55
  56 static const unsigned short jis0208[84][94] = {
  57 #include "jis0208.h"
  58 };
  59
  60 static const unsigned short gb18030[126][190] = {
  61 #include "gb18030.h"
  62 };
  63
  64 static const unsigned short big5[89][157] = {
  65 #include "big5.h"
  66 };
  67
  68 static const unsigned short hkscs[] = {
  69 #include "hkscs.h"
  70 };
  71
  72 static const unsigned short ksc[93][94] = {
  73 #include "ksc.h"
  74 };
  75
  76 static int fuzzycmp(const unsigned char *a, const unsigned char *b)
  77 {
  78         for (; *a && *b; a++, b++) {
  79                 while (*a && (*a|32U)-'a'>26 && *a-'0'>10U) a++;
  80                 if ((*a|32U) != *b) return 1;
  81         }
  82         return *a != *b;
  83 }
  84
  85 static size_t find_charmap(const void *name)
  86 {
  87         const unsigned char *s;
  88         if (!*(char *)name) name=charmaps; /* "utf8" */
  89         for (s=charmaps; *s; ) {
  90                 if (!fuzzycmp(name, s)) {
  91                         for (; *s; s+=strlen((void *)s)+1);
  92                         return s+1-charmaps;
  93                 }
  94                 s += strlen((void *)s)+1;
  95                 if (!*s) {
  96                         if (s[1] > 0200) s+=2;
  97                         else s+=2+(128U-s[1])/4*5;
  98                 }
  99         }
 100         return -1;
 101 }
 102
 103 static iconv_t combine_to_from(size_t t, size_t f)
 104 {
 105         return (void *)(f<<16 | t);
 106 }
 107
 108 static size_t extract_from(iconv_t cd)
 109 {
 110         return (size_t)cd >> 16;
 111 }
 112
 113 static size_t extract_to(iconv_t cd)
 114 {
 115         return (size_t)cd & 0xffff;
 116 }
 117
 118 iconv_t iconv_open(const char *to, const char *from)
 119 {
 120         size_t f, t;
 121
 122         if ((t = find_charmap(to))==-1
 123          || (f = find_charmap(from))==-1
 124          || (charmaps[t] >= 0320)) {
 125                 errno = EINVAL;
 126                 return (iconv_t)-1;
 127         }
 128
 129         return combine_to_from(t, f);
 130 }
 131
 132 static unsigned get_16(const unsigned char *s, int e)
 133 {
 134         e &= 1;
 135         return s[e]<<8 | s[1-e];
 136 }
 137
 138 static void put_16(unsigned char *s, unsigned c, int e)
 139 {
 140         e &= 1;
 141         s[e] = c>>8;
 142         s[1-e] = c;
 143 }
 144
 145 static unsigned get_32(const unsigned char *s, int e)
 146 {
 147         e &= 3;
 148         return s[e]+0U<<24 | s[e^1]<<16 | s[e^2]<<8 | s[e^3];
 149 }
 150
 151 static void put_32(unsigned char *s, unsigned c, int e)
 152 {
 153         e &= 3;
 154         s[e^0] = c>>24;
 155         s[e^1] = c>>16;
 156         s[e^2] = c>>8;
 157         s[e^3] = c;
 158 }
 159
 160 /* Adapt as needed */
 161 #define mbrtowc_utf8 mbrtowc
 162 #define wctomb_utf8 wctomb
 163
 164 static unsigned legacy_map(const unsigned char *map, unsigned c)
 165 {
 166         unsigned x = c - 128 - map[-1];
 167         x = legacy_chars[ map[x*5/4]>>2*x%8 |
 168                 map[x*5/4+1]<<8-2*x%8 & 1023 ];
 169         return x ? x : c;
 170 }
 171
 172 size_t iconv(iconv_t cd, char **restrict in, size_t *restrict inb, char **restrict out, size_t *restrict outb)
 173 {
 174         size_t x=0;
 175         unsigned to = extract_to(cd);
 176         unsigned from = extract_from(cd);
 177         const unsigned char *map = charmaps+from+1;
 178         const unsigned char *tomap = charmaps+to+1;
 179         mbstate_t st = {0};
 180         wchar_t wc;
 181         unsigned c, d;
 182         size_t k, l;
 183         int err;
 184         unsigned char type = map[-1];
 185         unsigned char totype = tomap[-1];
 186         locale_t *ploc = &CURRENT_LOCALE, loc = *ploc;
 187
 188         if (!in || !*in || !*inb) return 0;
 189
 190         *ploc = UTF8_LOCALE;
 191
 192         for (; *inb; *in+=l, *inb-=l) {
 193                 c = *(unsigned char *)*in;
 194                 l = 1;
 195
 196                 switch (type) {
 197                 case UTF_8:
 198                         if (c < 128) break; // optimization
 199                         l = mbrtowc_utf8(&wc, *in, *inb, &st);
 200                         if (!l) l++;
 201                         else if (l == (size_t)-1) goto ilseq;
 202                         else if (l == (size_t)-2) goto starved;
 203                         c = wc;
 204                         break;
 205                 case US_ASCII:
 206                         if (c >= 128) goto ilseq;
 207                         break;
 208                 case WCHAR_T:
 209                         l = sizeof(wchar_t);
 210                         if (*inb < l) goto starved;
 211                         c = *(wchar_t *)*in;
 212                         if (0) {
 213                 case UTF_32BE:
 214                 case UTF_32LE:
 215                         l = 4;
 216                         if (*inb < 4) goto starved;
 217                         c = get_32((void *)*in, type);
 218                         }
 219                         if (c-0xd800u < 0x800u || c >= 0x110000u) goto ilseq;
 220                         break;
 221                 case UCS2BE:
 222                 case UCS2LE:
 223                 case UTF_16BE:
 224                 case UTF_16LE:
 225                         l = 2;
 226                         if (*inb < 2) goto starved;
 227                         c = get_16((void *)*in, type);
 228                         if ((unsigned)(c-0xdc00) < 0x400) goto ilseq;
 229                         if ((unsigned)(c-0xd800) < 0x400) {
 230                                 if (type-UCS2BE < 2U) goto ilseq;
 231                                 l = 4;
 232                                 if (*inb < 4) goto starved;
 233                                 d = get_16((void *)(*in + 2), type);
 234                                 if ((unsigned)(d-0xdc00) >= 0x400) goto ilseq;
 235                                 c = ((c-0xd7c0)<<10) + (d-0xdc00);
 236                         }
 237                         break;
 238                 case SHIFT_JIS:
 239                         if (c < 128) break;
 240                         if (c-0xa1 <= 0xdf-0xa1) {
 241                                 c += 0xff61-0xa1;
 242                                 break;
 243                         }
 244                         l = 2;
 245                         if (*inb < 2) goto starved;
 246                         d = *((unsigned char *)*in + 1);
 247                         if (c-129 <= 159-129) c -= 129;
 248                         else if (c-224 <= 239-224) c -= 193;
 249                         else goto ilseq;
 250                         c *= 2;
 251                         if (d-64 <= 158-64) {
 252                                 if (d==127) goto ilseq;
 253                                 if (d>127) d--;
 254                                 d -= 64;
 255                         } else if (d-159 <= 252-159) {
 256                                 c++;
 257                                 d -= 159;
 258                         }
 259                         c = jis0208[c][d];
 260                         if (!c) goto ilseq;
 261                         break;
 262                 case EUC_JP:
 263                         if (c < 128) break;
 264                         l = 2;
 265                         if (*inb < 2) goto starved;
 266                         d = *((unsigned char *)*in + 1);
 267                         if (c==0x8e) {
 268                                 c = d;
 269                                 if (c-0xa1 > 0xdf-0xa1) goto ilseq;
 270                                 c += 0xff61 - 0xa1;
 271                                 break;
 272                         }
 273                         c -= 0xa1;
 274                         d -= 0xa1;
 275                         if (c >= 84 || d >= 94) goto ilseq;
 276                         c = jis0208[c][d];
 277                         if (!c) goto ilseq;
 278                         break;
 279                 case GB2312:
 280                         if (c < 128) break;
 281                         if (c < 0xa1) goto ilseq;
 282                 case GBK:
 283                 case GB18030:
 284                         if (c < 128) break;
 285                         c -= 0x81;
 286                         if (c >= 126) goto ilseq;
 287                         l = 2;
 288                         if (*inb < 2) goto starved;
 289                         d = *((unsigned char *)*in + 1);
 290                         if (d < 0xa1 && type == GB2312) goto ilseq;
 291                         if (d-0x40>=191 || d==127) {
 292                                 if (d-'0'>9 || type != GB18030)
 293                                         goto ilseq;
 294                                 l = 4;
 295                                 if (*inb < 4) goto starved;
 296                                 c = (10*c + d-'0') * 1260;
 297                                 d = *((unsigned char *)*in + 2);
 298                                 if (d-0x81>126) goto ilseq;
 299                                 c += 10*(d-0x81);
 300                                 d = *((unsigned char *)*in + 3);
 301                                 if (d-'0'>9) goto ilseq;
 302                                 c += d-'0';
 303                                 c += 128;
 304                                 for (d=0; d<=c; ) {
 305                                         k = 0;
 306                                         for (int i=0; i<126; i++)
 307                                                 for (int j=0; j<190; j++)
 308                                                         if (gb18030[i][j]-d <= c-d)
 309                                                                 k++;
 310                                         d = c+1;
 311                                         c += k;
 312                                 }
 313                                 break;
 314                         }
 315                         d -= 0x40;
 316                         if (d>63) d--;
 317                         c = gb18030[c][d];
 318                         break;
 319                 case BIG5:
 320                         if (c < 128) break;
 321                         l = 2;
 322                         if (*inb < 2) goto starved;
 323                         d = *((unsigned char *)*in + 1);
 324                         if (d-0x40>=0xff-0x40 || d-0x7f<0xa1-0x7f) goto ilseq;
 325                         d -= 0x40;
 326                         if (d > 0x3e) d -= 0x22;
 327                         if (c-0xa1>=0xfa-0xa1) {
 328                                 if (c-0x87>=0xff-0x87) goto ilseq;
 329                                 if (c < 0xa1) c -= 0x87;
 330                                 else c -= 0x87 + (0xfa-0xa1);
 331                                 c = (hkscs[4867+(c*157+d)/16]>>(c*157+d)%16)%2<<17
 332                                         | hkscs[c*157+d];
 333                                 /* A few HKSCS characters map to pairs of UCS
 334                                  * characters. These are mapped to surrogate
 335                                  * range in the hkscs table then hard-coded
 336                                  * here. Ugly, yes. */
 337                                 if (c/256 == 0xdc) {
 338                                         if (totype-0300U > 8) k = 2;
 339                                         else k = "\10\4\4\10\4\4\10\2\4"[totype-0300];
 340                                         if (k > *outb) goto toobig;
 341                                         x += iconv(combine_to_from(to, 0),
 342                                                 &(char *){"\303\212\314\204"
 343                                                 "\303\212\314\214"
 344                                                 "\303\252\314\204"
 345                                                 "\303\252\314\214"
 346                                                 +c%256}, &(size_t){4},
 347                                                 out, outb);
 348                                         continue;
 349                                 }
 350                                 if (!c) goto ilseq;
 351                                 break;
 352                         }
 353                         c -= 0xa1;
 354                         c = big5[c][d]|(c==0x27&&(d==0x3a||d==0x3c||d==0x42))<<17;
 355                         if (!c) goto ilseq;
 356                         break;
 357                 case EUC_KR:
 358                         if (c < 128) break;
 359                         l = 2;
 360                         if (*inb < 2) goto starved;
 361                         d = *((unsigned char *)*in + 1);
 362                         c -= 0xa1;
 363                         d -= 0xa1;
 364                         if (c >= 93 || d >= 94) {
 365                                 c += (0xa1-0x81);
 366                                 d += 0xa1;
 367                                 if (c >= 93 || c>=0xc6-0x81 && d>0x52)
 368                                         goto ilseq;
 369                                 if (d-'A'<26) d = d-'A';
 370                                 else if (d-'a'<26) d = d-'a'+26;
 371                                 else if (d-0x81<0xff-0x81) d = d-0x81+52;
 372                                 else goto ilseq;
 373                                 if (c < 0x20) c = 178*c + d;
 374                                 else c = 178*0x20 + 84*(c-0x20) + d;
 375                                 c += 0xac00;
 376                                 for (d=0xac00; d<=c; ) {
 377                                         k = 0;
 378                                         for (int i=0; i<93; i++)
 379                                                 for (int j=0; j<94; j++)
 380                                                         if (ksc[i][j]-d <= c-d)
 381                                                                 k++;
 382                                         d = c+1;
 383                                         c += k;
 384                                 }
 385                                 break;
 386                         }
 387                         c = ksc[c][d];
 388                         if (!c) goto ilseq;
 389                         break;
 390                 default:
 391                         if (c < 128+type) break;
 392                         c = legacy_map(map, c);
 393                         if (c==1) goto ilseq;
 394                 }
 395
 396                 switch (totype) {
 397                 case WCHAR_T:
 398                         if (*outb < sizeof(wchar_t)) goto toobig;
 399                         *(wchar_t *)*out = c;
 400                         *out += sizeof(wchar_t);
 401                         *outb -= sizeof(wchar_t);
 402                         break;
 403                 case UTF_8:
 404                         if (*outb < 4) {
 405                                 char tmp[4];
 406                                 k = wctomb_utf8(tmp, c);
 407                                 if (*outb < k) goto toobig;
 408                                 memcpy(*out, tmp, k);
 409                         } else k = wctomb_utf8(*out, c);
 410                         *out += k;
 411                         *outb -= k;
 412                         break;
 413                 case US_ASCII:
 414                         if (c > 0x7f) subst: x++, c='*';
 415                 default:
 416                         if (*outb < 1) goto toobig;
 417                         if (c < 128+totype || (c<256 && c==legacy_map(tomap, c))) {
 418                         revout:
 419                                 *(*out)++ = c;
 420                                 *outb -= 1;
 421                                 break;
 422                         }
 423                         d = c;
 424                         for (c=128+totype; c<256; c++) {
 425                                 if (d == legacy_map(tomap, c)) {
 426                                         goto revout;
 427                                 }
 428                         }
 429                         goto subst;
 430                 case UCS2BE:
 431                 case UCS2LE:
 432                 case UTF_16BE:
 433                 case UTF_16LE:
 434                         if (c < 0x10000 || type-UCS2BE < 2U) {
 435                                 if (c >= 0x10000) c = 0xFFFD;
 436                                 if (*outb < 2) goto toobig;
 437                                 put_16((void *)*out, c, totype);
 438                                 *out += 2;
 439                                 *outb -= 2;
 440                                 break;
 441                         }
 442                         if (*outb < 4) goto toobig;
 443                         c -= 0x10000;
 444                         put_16((void *)*out, (c>>10)|0xd800, totype);
 445                         put_16((void *)(*out + 2), (c&0x3ff)|0xdc00, totype);
 446                         *out += 4;
 447                         *outb -= 4;
 448                         break;
 449                 case UTF_32BE:
 450                 case UTF_32LE:
 451                         if (*outb < 4) goto toobig;
 452                         put_32((void *)*out, c, totype);
 453                         *out += 4;
 454                         *outb -= 4;
 455                         break;
 456                 }
 457         }
 458         *ploc = loc;
 459         return x;
 460 ilseq:
 461         err = EILSEQ;
 462         x = -1;
 463         goto end;
 464 toobig:
 465         err = E2BIG;
 466         x = -1;
 467         goto end;
 468 starved:
 469         err = EINVAL;
 470         x = -1;
 471 end:
 472         errno = err;
 473         *ploc = loc;
 474         return x;
 475 }