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 EUC_KR      0350
  24
  25 /* FIXME: these are not implemented yet
  26  * EUC:   A1-FE A1-FE
  27  * GBK:   81-FE 40-7E,80-FE
  28  * Big5:  A1-FE 40-7E,A1-FE
  29  */
  30
  31 /* Definitions of charmaps. Each charmap consists of:
  32  * 1. Empty-string-terminated list of null-terminated aliases.
  33  * 2. Special type code or number of elided entries.
  34  * 3. Character table (size determined by field 2). */
  35
  36 static const unsigned char charmaps[] =
  37 "utf8\0\0\310"
  38 "wchart\0\0\306"
  39 "ucs2\0ucs2be\0\0\304"
  40 "ucs2le\0\0\305"
  41 "utf16\0utf16be\0\0\302"
  42 "utf16le\0\0\301"
  43 "ucs4\0ucs4be\0utf32\0utf32be\0\0\300"
  44 "ucs4le\0utf32le\0\0\303"
  45 "ascii\0usascii\0iso646\0iso646us\0\0\307"
  46 "eucjp\0\0\320"
  47 "shiftjis\0sjis\0\0\321"
  48 "gb18030\0\0\330"
  49 "gbk\0\0\331"
  50 "gb2312\0\0\332"
  51 "euckr\0ksc5601\0ksx1001\0cp949\0\0\350"
  52 #include "codepages.h"
  53 ;
  54
  55 static const unsigned short legacy_chars[] = {
  56 #include "legacychars.h"
  57 };
  58
  59 static const unsigned short jis0208[84][94] = {
  60 #include "jis0208.h"
  61 };
  62
  63 static const unsigned short gb18030[126][190] = {
  64 #include "gb18030.h"
  65 };
  66
  67 static const unsigned short ksc[93][94] = {
  68 #include "ksc.h"
  69 };
  70
  71 static int fuzzycmp(const unsigned char *a, const unsigned char *b)
  72 {
  73         for (; *a && *b; a++, b++) {
  74                 while (*a && (*a|32U)-'a'>26 && *a-'0'>10U) a++;
  75                 if ((*a|32U) != *b) return 1;
  76         }
  77         return *a != *b;
  78 }
  79
  80 static size_t find_charmap(const void *name)
  81 {
  82         const unsigned char *s;
  83         for (s=charmaps; *s; ) {
  84                 if (!fuzzycmp(name, s)) {
  85                         for (; *s; s+=strlen((void *)s)+1);
  86                         return s+1-charmaps;
  87                 }
  88                 s += strlen((void *)s)+1;
  89                 if (!*s) {
  90                         if (s[1] > 0200) s+=2;
  91                         else s+=2+(128U-s[1])/4*5;
  92                 }
  93         }
  94         return -1;
  95 }
  96
  97 iconv_t iconv_open(const char *to, const char *from)
  98 {
  99         size_t f, t;
 100
 101         if ((t = find_charmap(to))==-1
 102          || (f = find_charmap(from))==-1
 103          || (charmaps[t] >= 0320)) {
 104                 errno = EINVAL;
 105                 return (iconv_t)-1;
 106         }
 107
 108         return (void *)(f<<16 | t);
 109 }
 110
 111 int iconv_close(iconv_t cd)
 112 {
 113         return 0;
 114 }
 115
 116 static unsigned get_16(const unsigned char *s, int e)
 117 {
 118         e &= 1;
 119         return s[e]<<8 | s[1-e];
 120 }
 121
 122 static void put_16(unsigned char *s, unsigned c, int e)
 123 {
 124         e &= 1;
 125         s[e] = c>>8;
 126         s[1-e] = c;
 127 }
 128
 129 static unsigned get_32(const unsigned char *s, int e)
 130 {
 131         e &= 3;
 132         return s[e]+0U<<24 | s[e^1]<<16 | s[e^2]<<8 | s[e^3];
 133 }
 134
 135 static void put_32(unsigned char *s, unsigned c, int e)
 136 {
 137         e &= 3;
 138         s[e^0] = c>>24;
 139         s[e^1] = c>>16;
 140         s[e^2] = c>>8;
 141         s[e^3] = c;
 142 }
 143
 144 /* Adapt as needed */
 145 #define mbrtowc_utf8 mbrtowc
 146 #define wctomb_utf8 wctomb
 147
 148 size_t iconv(iconv_t cd0, char **restrict in, size_t *restrict inb, char **restrict out, size_t *restrict outb)
 149 {
 150         size_t x=0;
 151         unsigned long cd = (unsigned long)cd0;
 152         unsigned to = cd & 0xffff;
 153         unsigned from = cd >> 16;
 154         const unsigned char *map = charmaps+from+1;
 155         const unsigned char *tomap = charmaps+to+1;
 156         mbstate_t st = {0};
 157         wchar_t wc;
 158         unsigned c, d;
 159         size_t k, l;
 160         int err;
 161         unsigned char type = map[-1];
 162         unsigned char totype = tomap[-1];
 163
 164         if (!in || !*in || !*inb) return 0;
 165
 166         for (; *inb; *in+=l, *inb-=l) {
 167                 c = *(unsigned char *)*in;
 168                 l = 1;
 169
 170                 if (c >= 128 || type-UTF_32BE < 7U) switch (type) {
 171                 case UTF_8:
 172                         l = mbrtowc_utf8(&wc, *in, *inb, &st);
 173                         if (!l) l++;
 174                         else if (l == (size_t)-1) goto ilseq;
 175                         else if (l == (size_t)-2) goto starved;
 176                         c = wc;
 177                         break;
 178                 case US_ASCII:
 179                         goto ilseq;
 180                 case WCHAR_T:
 181                         l = sizeof(wchar_t);
 182                         if (*inb < l) goto starved;
 183                         c = *(wchar_t *)*in;
 184                         if (0) {
 185                 case UTF_32BE:
 186                 case UTF_32LE:
 187                         l = 4;
 188                         if (*inb < 4) goto starved;
 189                         c = get_32((void *)*in, type);
 190                         }
 191                         if (c-0xd800u < 0x800u || c >= 0x110000u) goto ilseq;
 192                         break;
 193                 case UCS2BE:
 194                 case UCS2LE:
 195                 case UTF_16BE:
 196                 case UTF_16LE:
 197                         l = 2;
 198                         if (*inb < 2) goto starved;
 199                         c = get_16((void *)*in, type);
 200                         if ((unsigned)(c-0xdc00) < 0x400) goto ilseq;
 201                         if ((unsigned)(c-0xd800) < 0x400) {
 202                                 if (type-UCS2BE < 2U) goto ilseq;
 203                                 l = 4;
 204                                 if (*inb < 4) goto starved;
 205                                 d = get_16((void *)(*in + 2), type);
 206                                 if ((unsigned)(d-0xdc00) >= 0x400) goto ilseq;
 207                                 c = ((c-0xd7c0)<<10) + (d-0xdc00);
 208                         }
 209                         break;
 210                 case SHIFT_JIS:
 211                         if (c-0xa1 <= 0xdf-0xa1) {
 212                                 c += 0xff61-0xa1;
 213                                 break;
 214                         }
 215                         l = 2;
 216                         if (*inb < 2) goto starved;
 217                         d = *((unsigned char *)*in + 1);
 218                         if (c-129 <= 159-129) c -= 129;
 219                         else if (c-224 <= 239-224) c -= 193;
 220                         else goto ilseq;
 221                         c *= 2;
 222                         if (d-64 <= 158-64) {
 223                                 if (d==127) goto ilseq;
 224                                 if (d>127) d--;
 225                                 d -= 64;
 226                         } else if (d-159 <= 252-159) {
 227                                 c++;
 228                                 d -= 159;
 229                         }
 230                         c = jis0208[c][d];
 231                         if (!c) goto ilseq;
 232                         break;
 233                 case EUC_JP:
 234                         l = 2;
 235                         if (*inb < 2) goto starved;
 236                         d = *((unsigned char *)*in + 1);
 237                         if (c==0x8e) {
 238                                 c = d;
 239                                 if (c-0xa1 > 0xdf-0xa1) goto ilseq;
 240                                 c += 0xff61 - 0xa1;
 241                                 break;
 242                         }
 243                         c -= 0xa1;
 244                         d -= 0xa1;
 245                         if (c >= 84 || d >= 94) goto ilseq;
 246                         c = jis0208[c][d];
 247                         if (!c) goto ilseq;
 248                         break;
 249                 case GB2312:
 250                         if (c < 0xa1) goto ilseq;
 251                 case GBK:
 252                 case GB18030:
 253                         c -= 0x81;
 254                         if (c >= 126) goto ilseq;
 255                         l = 2;
 256                         if (*inb < 2) goto starved;
 257                         d = *((unsigned char *)*in + 1);
 258                         if (d < 0xa1 && type == GB2312) goto ilseq;
 259                         if (d-0x40>=191 || d==127) {
 260                                 if (d-'0'>9 || type != GB18030)
 261                                         goto ilseq;
 262                                 l = 4;
 263                                 if (*inb < 4) goto starved;
 264                                 c = (10*c + d-'0') * 1260;
 265                                 d = *((unsigned char *)*in + 2);
 266                                 if (d-0x81>126) goto ilseq;
 267                                 c += 10*(d-0x81);
 268                                 d = *((unsigned char *)*in + 3);
 269                                 if (d-'0'>9) goto ilseq;
 270                                 c += d-'0';
 271                                 c += 128;
 272                                 for (d=0; d<=c; ) {
 273                                         k = 0;
 274                                         for (int i=0; i<126; i++)
 275                                                 for (int j=0; j<190; j++)
 276                                                         if (gb18030[i][j]-d <= c-d)
 277                                                                 k++;
 278                                         d = c+1;
 279                                         c += k;
 280                                 }
 281                                 break;
 282                         }
 283                         d -= 0x40;
 284                         if (d>63) d--;
 285                         c = gb18030[c][d];
 286                         break;
 287                 case EUC_KR:
 288                         l = 2;
 289                         if (*inb < 2) goto starved;
 290                         d = *((unsigned char *)*in + 1);
 291                         c -= 0xa1;
 292                         d -= 0xa1;
 293                         if (c >= 93 || d >= 94) {
 294                                 c += (0xa1-0x81);
 295                                 d += 0xa1;
 296                                 if (c >= 93 || c>=0xc6-0x81 && d>0x52)
 297                                         goto ilseq;
 298                                 if (d-'A'<26) d = d-'A';
 299                                 else if (d-'a'<26) d = d-'a'+26;
 300                                 else if (d-0x81<0xff-0x81) d = d-0x81+52;
 301                                 else goto ilseq;
 302                                 if (c < 0x20) c = 178*c + d;
 303                                 else c = 178*0x20 + 84*(c-0x20) + d;
 304                                 c += 0xac00;
 305                                 for (d=0xac00; d<=c; ) {
 306                                         k = 0;
 307                                         for (int i=0; i<93; i++)
 308                                                 for (int j=0; j<94; j++)
 309                                                         if (ksc[i][j]-d <= c-d)
 310                                                                 k++;
 311                                         d = c+1;
 312                                         c += k;
 313                                 }
 314                                 break;
 315                         }
 316                         c = ksc[c][d];
 317                         if (!c) goto ilseq;
 318                         break;
 319                 default:
 320                         if (c < 128+type) break;
 321                         c -= 128+type;
 322                         c = legacy_chars[ map[c*5/4]>>2*c%8 |
 323                                 map[c*5/4+1]<<8-2*c%8 & 1023 ];
 324                         if (!c) c = *(unsigned char *)*in;
 325                         if (c==1) goto ilseq;
 326                 }
 327
 328                 switch (totype) {
 329                 case WCHAR_T:
 330                         if (*outb < sizeof(wchar_t)) goto toobig;
 331                         *(wchar_t *)*out = c;
 332                         *out += sizeof(wchar_t);
 333                         *outb -= sizeof(wchar_t);
 334                         break;
 335                 case UTF_8:
 336                         if (*outb < 4) {
 337                                 char tmp[4];
 338                                 k = wctomb_utf8(tmp, c);
 339                                 if (*outb < k) goto toobig;
 340                                 memcpy(*out, tmp, k);
 341                         } else k = wctomb_utf8(*out, c);
 342                         *out += k;
 343                         *outb -= k;
 344                         break;
 345                 case US_ASCII:
 346                         if (c > 0x7f) subst: x++, c='*';
 347                 default:
 348                         if (*outb < 1) goto toobig;
 349                         if (c < 128+totype) {
 350                         revout:
 351                                 *(*out)++ = c;
 352                                 *outb -= 1;
 353                                 break;
 354                         }
 355                         d = c;
 356                         for (c=0; c<128-totype; c++) {
 357                                 if (d == legacy_chars[ tomap[c*5/4]>>2*c%8 |
 358                                         tomap[c*5/4+1]<<8-2*c%8 & 1023 ]) {
 359                                         c += 128;
 360                                         goto revout;
 361                                 }
 362                         }
 363                         goto subst;
 364                 case UCS2BE:
 365                 case UCS2LE:
 366                 case UTF_16BE:
 367                 case UTF_16LE:
 368                         if (c < 0x10000 || type-UCS2BE < 2U) {
 369                                 if (c >= 0x10000) c = 0xFFFD;
 370                                 if (*outb < 2) goto toobig;
 371                                 put_16((void *)*out, c, totype);
 372                                 *out += 2;
 373                                 *outb -= 2;
 374                                 break;
 375                         }
 376                         if (*outb < 4) goto toobig;
 377                         c -= 0x10000;
 378                         put_16((void *)*out, (c>>10)|0xd800, totype);
 379                         put_16((void *)(*out + 2), (c&0x3ff)|0xdc00, totype);
 380                         *out += 4;
 381                         *outb -= 4;
 382                         break;
 383                 case UTF_32BE:
 384                 case UTF_32LE:
 385                         if (*outb < 4) goto toobig;
 386                         put_32((void *)*out, c, totype);
 387                         *out += 4;
 388                         *outb -= 4;
 389                         break;
 390                 }
 391         }
 392         return x;
 393 ilseq:
 394         err = EILSEQ;
 395         x = -1;
 396         goto end;
 397 toobig:
 398         err = E2BIG;
 399         x = -1;
 400         goto end;
 401 starved:
 402         err = EINVAL;
 403         x = -1;
 404 end:
 405         errno = err;
 406         return x;
 407 }