nsz Git - musl/blob - 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
   9 #define UTF_32BE    0300
  10 #define UTF_16LE    0301
  11 #define UTF_16BE    0302
  12 #define UTF_32LE    0303
  13 #define UCS2BE      0304
  14 #define UCS2LE      0305
  15 #define WCHAR_T     0306
  16 #define US_ASCII    0307
  17 #define UTF_8       0310
  18 #define EUC_JP      0320
  19 #define SHIFT_JIS   0321
  20 #define GB18030     0330
  21 #define GBK         0331
  22 #define GB2312      0332
  23 #define BIG5        0340
  24 #define EUC_KR      0350
  25
  26 /* FIXME: these are not implemented yet
  27  * EUC:   A1-FE A1-FE
  28  * GBK:   81-FE 40-7E,80-FE
  29  * Big5:  A1-FE 40-7E,A1-FE
  30  */
  31
  32 /* Definitions of charmaps. Each charmap consists of:
  33  * 1. Empty-string-terminated list of null-terminated aliases.
  34  * 2. Special type code or number of elided entries.
  35  * 3. Character table (size determined by field 2). */
  36
  37 static const unsigned char charmaps[] =
  38 "utf8\0\0\310"
  39 "wchart\0\0\306"
  40 "ucs2\0ucs2be\0\0\304"
  41 "ucs2le\0\0\305"
  42 "utf16\0utf16be\0\0\302"
  43 "utf16le\0\0\301"
  44 "ucs4\0ucs4be\0utf32\0utf32be\0\0\300"
  45 "ucs4le\0utf32le\0\0\303"
  46 "ascii\0usascii\0iso646\0iso646us\0\0\307"
  47 "eucjp\0\0\320"
  48 "shiftjis\0sjis\0\0\321"
  49 "gb18030\0\0\330"
  50 "gbk\0\0\331"
  51 "gb2312\0\0\332"
  52 "big5\0bigfive\0cp950\0\0\340"
  53 "euckr\0ksc5601\0ksx1001\0cp949\0\0\350"
  54 #include "codepages.h"
  55 ;
  56
  57 static const unsigned short legacy_chars[] = {
  58 #include "legacychars.h"
  59 };
  60
  61 static const unsigned short jis0208[84][94] = {
  62 #include "jis0208.h"
  63 };
  64
  65 static const unsigned short gb18030[126][190] = {
  66 #include "gb18030.h"
  67 };
  68
  69 static const unsigned short big5[89][157] = {
  70 #include "big5.h"
  71 };
  72
  73 static const unsigned short ksc[93][94] = {
  74 #include "ksc.h"
  75 };
  76
  77 static int fuzzycmp(const unsigned char *a, const unsigned char *b)
  78 {
  79         for (; *a && *b; a++, b++) {
  80                 while (*a && (*a|32U)-'a'>26 && *a-'0'>10U) a++;
  81                 if ((*a|32U) != *b) return 1;
  82         }
  83         return *a != *b;
  84 }
  85
  86 static size_t find_charmap(const void *name)
  87 {
  88         const unsigned char *s;
  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 iconv_t iconv_open(const char *to, const char *from)
 104 {
 105         size_t f, t;
 106
 107         if ((t = find_charmap(to))==-1
 108          || (f = find_charmap(from))==-1
 109          || (charmaps[t] >= 0320)) {
 110                 errno = EINVAL;
 111                 return (iconv_t)-1;
 112         }
 113
 114         return (void *)(f<<16 | t);
 115 }
 116
 117 int iconv_close(iconv_t cd)
 118 {
 119         return 0;
 120 }
 121
 122 static unsigned get_16(const unsigned char *s, int e)
 123 {
 124         e &= 1;
 125         return s[e]<<8 | s[1-e];
 126 }
 127
 128 static void put_16(unsigned char *s, unsigned c, int e)
 129 {
 130         e &= 1;
 131         s[e] = c>>8;
 132         s[1-e] = c;
 133 }
 134
 135 static unsigned get_32(const unsigned char *s, int e)
 136 {
 137         e &= 3;
 138         return s[e]+0U<<24 | s[e^1]<<16 | s[e^2]<<8 | s[e^3];
 139 }
 140
 141 static void put_32(unsigned char *s, unsigned c, int e)
 142 {
 143         e &= 3;
 144         s[e^0] = c>>24;
 145         s[e^1] = c>>16;
 146         s[e^2] = c>>8;
 147         s[e^3] = c;
 148 }
 149
 150 /* Adapt as needed */
 151 #define mbrtowc_utf8 mbrtowc
 152 #define wctomb_utf8 wctomb
 153
 154 size_t iconv(iconv_t cd0, char **restrict in, size_t *restrict inb, char **restrict out, size_t *restrict outb)
 155 {
 156         size_t x=0;
 157         unsigned long cd = (unsigned long)cd0;
 158         unsigned to = cd & 0xffff;
 159         unsigned from = cd >> 16;
 160         const unsigned char *map = charmaps+from+1;
 161         const unsigned char *tomap = charmaps+to+1;
 162         mbstate_t st = {0};
 163         wchar_t wc;
 164         unsigned c, d;
 165         size_t k, l;
 166         int err;
 167         unsigned char type = map[-1];
 168         unsigned char totype = tomap[-1];
 169
 170         if (!in || !*in || !*inb) return 0;
 171
 172         for (; *inb; *in+=l, *inb-=l) {
 173                 c = *(unsigned char *)*in;
 174                 l = 1;
 175
 176                 if (c >= 128 || type-UTF_32BE < 7U) switch (type) {
 177                 case UTF_8:
 178                         l = mbrtowc_utf8(&wc, *in, *inb, &st);
 179                         if (!l) l++;
 180                         else if (l == (size_t)-1) goto ilseq;
 181                         else if (l == (size_t)-2) goto starved;
 182                         c = wc;
 183                         break;
 184                 case US_ASCII:
 185                         goto ilseq;
 186                 case WCHAR_T:
 187                         l = sizeof(wchar_t);
 188                         if (*inb < l) goto starved;
 189                         c = *(wchar_t *)*in;
 190                         if (0) {
 191                 case UTF_32BE:
 192                 case UTF_32LE:
 193                         l = 4;
 194                         if (*inb < 4) goto starved;
 195                         c = get_32((void *)*in, type);
 196                         }
 197                         if (c-0xd800u < 0x800u || c >= 0x110000u) goto ilseq;
 198                         break;
 199                 case UCS2BE:
 200                 case UCS2LE:
 201                 case UTF_16BE:
 202                 case UTF_16LE:
 203                         l = 2;
 204                         if (*inb < 2) goto starved;
 205                         c = get_16((void *)*in, type);
 206                         if ((unsigned)(c-0xdc00) < 0x400) goto ilseq;
 207                         if ((unsigned)(c-0xd800) < 0x400) {
 208                                 if (type-UCS2BE < 2U) goto ilseq;
 209                                 l = 4;
 210                                 if (*inb < 4) goto starved;
 211                                 d = get_16((void *)(*in + 2), type);
 212                                 if ((unsigned)(d-0xdc00) >= 0x400) goto ilseq;
 213                                 c = ((c-0xd7c0)<<10) + (d-0xdc00);
 214                         }
 215                         break;
 216                 case SHIFT_JIS:
 217                         if (c-0xa1 <= 0xdf-0xa1) {
 218                                 c += 0xff61-0xa1;
 219                                 break;
 220                         }
 221                         l = 2;
 222                         if (*inb < 2) goto starved;
 223                         d = *((unsigned char *)*in + 1);
 224                         if (c-129 <= 159-129) c -= 129;
 225                         else if (c-224 <= 239-224) c -= 193;
 226                         else goto ilseq;
 227                         c *= 2;
 228                         if (d-64 <= 158-64) {
 229                                 if (d==127) goto ilseq;
 230                                 if (d>127) d--;
 231                                 d -= 64;
 232                         } else if (d-159 <= 252-159) {
 233                                 c++;
 234                                 d -= 159;
 235                         }
 236                         c = jis0208[c][d];
 237                         if (!c) goto ilseq;
 238                         break;
 239                 case EUC_JP:
 240                         l = 2;
 241                         if (*inb < 2) goto starved;
 242                         d = *((unsigned char *)*in + 1);
 243                         if (c==0x8e) {
 244                                 c = d;
 245                                 if (c-0xa1 > 0xdf-0xa1) goto ilseq;
 246                                 c += 0xff61 - 0xa1;
 247                                 break;
 248                         }
 249                         c -= 0xa1;
 250                         d -= 0xa1;
 251                         if (c >= 84 || d >= 94) goto ilseq;
 252                         c = jis0208[c][d];
 253                         if (!c) goto ilseq;
 254                         break;
 255                 case GB2312:
 256                         if (c < 0xa1) goto ilseq;
 257                 case GBK:
 258                 case GB18030:
 259                         c -= 0x81;
 260                         if (c >= 126) goto ilseq;
 261                         l = 2;
 262                         if (*inb < 2) goto starved;
 263                         d = *((unsigned char *)*in + 1);
 264                         if (d < 0xa1 && type == GB2312) goto ilseq;
 265                         if (d-0x40>=191 || d==127) {
 266                                 if (d-'0'>9 || type != GB18030)
 267                                         goto ilseq;
 268                                 l = 4;
 269                                 if (*inb < 4) goto starved;
 270                                 c = (10*c + d-'0') * 1260;
 271                                 d = *((unsigned char *)*in + 2);
 272                                 if (d-0x81>126) goto ilseq;
 273                                 c += 10*(d-0x81);
 274                                 d = *((unsigned char *)*in + 3);
 275                                 if (d-'0'>9) goto ilseq;
 276                                 c += d-'0';
 277                                 c += 128;
 278                                 for (d=0; d<=c; ) {
 279                                         k = 0;
 280                                         for (int i=0; i<126; i++)
 281                                                 for (int j=0; j<190; j++)
 282                                                         if (gb18030[i][j]-d <= c-d)
 283                                                                 k++;
 284                                         d = c+1;
 285                                         c += k;
 286                                 }
 287                                 break;
 288                         }
 289                         d -= 0x40;
 290                         if (d>63) d--;
 291                         c = gb18030[c][d];
 292                         break;
 293                 case BIG5:
 294                         l = 2;
 295                         if (*inb < 2) goto starved;
 296                         d = *((unsigned char *)*in + 1);
 297                         if (c-0xa1>=0xfa-0xa1) goto ilseq;
 298                         c -= 0xa1;
 299                         if (d-0x40>=0xff-0x40 || d-0x7f<0xa1-0x7f) goto ilseq;
 300                         d -= 0x40;
 301                         if (d > 0x3e) d -= 0x22;
 302                         c = big5[c][d];
 303                         if (!c) goto ilseq;
 304                         break;
 305                 case EUC_KR:
 306                         l = 2;
 307                         if (*inb < 2) goto starved;
 308                         d = *((unsigned char *)*in + 1);
 309                         c -= 0xa1;
 310                         d -= 0xa1;
 311                         if (c >= 93 || d >= 94) {
 312                                 c += (0xa1-0x81);
 313                                 d += 0xa1;
 314                                 if (c >= 93 || c>=0xc6-0x81 && d>0x52)
 315                                         goto ilseq;
 316                                 if (d-'A'<26) d = d-'A';
 317                                 else if (d-'a'<26) d = d-'a'+26;
 318                                 else if (d-0x81<0xff-0x81) d = d-0x81+52;
 319                                 else goto ilseq;
 320                                 if (c < 0x20) c = 178*c + d;
 321                                 else c = 178*0x20 + 84*(c-0x20) + d;
 322                                 c += 0xac00;
 323                                 for (d=0xac00; d<=c; ) {
 324                                         k = 0;
 325                                         for (int i=0; i<93; i++)
 326                                                 for (int j=0; j<94; j++)
 327                                                         if (ksc[i][j]-d <= c-d)
 328                                                                 k++;
 329                                         d = c+1;
 330                                         c += k;
 331                                 }
 332                                 break;
 333                         }
 334                         c = ksc[c][d];
 335                         if (!c) goto ilseq;
 336                         break;
 337                 default:
 338                         if (c < 128+type) break;
 339                         c -= 128+type;
 340                         c = legacy_chars[ map[c*5/4]>>2*c%8 |
 341                                 map[c*5/4+1]<<8-2*c%8 & 1023 ];
 342                         if (!c) c = *(unsigned char *)*in;
 343                         if (c==1) goto ilseq;
 344                 }
 345
 346                 switch (totype) {
 347                 case WCHAR_T:
 348                         if (*outb < sizeof(wchar_t)) goto toobig;
 349                         *(wchar_t *)*out = c;
 350                         *out += sizeof(wchar_t);
 351                         *outb -= sizeof(wchar_t);
 352                         break;
 353                 case UTF_8:
 354                         if (*outb < 4) {
 355                                 char tmp[4];
 356                                 k = wctomb_utf8(tmp, c);
 357                                 if (*outb < k) goto toobig;
 358                                 memcpy(*out, tmp, k);
 359                         } else k = wctomb_utf8(*out, c);
 360                         *out += k;
 361                         *outb -= k;
 362                         break;
 363                 case US_ASCII:
 364                         if (c > 0x7f) subst: x++, c='*';
 365                 default:
 366                         if (*outb < 1) goto toobig;
 367                         if (c < 128+totype) {
 368                         revout:
 369                                 *(*out)++ = c;
 370                                 *outb -= 1;
 371                                 break;
 372                         }
 373                         d = c;
 374                         for (c=0; c<128-totype; c++) {
 375                                 if (d == legacy_chars[ tomap[c*5/4]>>2*c%8 |
 376                                         tomap[c*5/4+1]<<8-2*c%8 & 1023 ]) {
 377                                         c += 128;
 378                                         goto revout;
 379                                 }
 380                         }
 381                         goto subst;
 382                 case UCS2BE:
 383                 case UCS2LE:
 384                 case UTF_16BE:
 385                 case UTF_16LE:
 386                         if (c < 0x10000 || type-UCS2BE < 2U) {
 387                                 if (c >= 0x10000) c = 0xFFFD;
 388                                 if (*outb < 2) goto toobig;
 389                                 put_16((void *)*out, c, totype);
 390                                 *out += 2;
 391                                 *outb -= 2;
 392                                 break;
 393                         }
 394                         if (*outb < 4) goto toobig;
 395                         c -= 0x10000;
 396                         put_16((void *)*out, (c>>10)|0xd800, totype);
 397                         put_16((void *)(*out + 2), (c&0x3ff)|0xdc00, totype);
 398                         *out += 4;
 399                         *outb -= 4;
 400                         break;
 401                 case UTF_32BE:
 402                 case UTF_32LE:
 403                         if (*outb < 4) goto toobig;
 404                         put_32((void *)*out, c, totype);
 405                         *out += 4;
 406                         *outb -= 4;
 407                         break;
 408                 }
 409         }
 410         return x;
 411 ilseq:
 412         err = EILSEQ;
 413         x = -1;
 414         goto end;
 415 toobig:
 416         err = E2BIG;
 417         x = -1;
 418         goto end;
 419 starved:
 420         err = EINVAL;
 421         x = -1;
 422 end:
 423         errno = err;
 424         return x;
 425 }