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