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\0char\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\0big5hkscs\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 hkscs[] = {
  74 #include "hkscs.h"
  75 };
  76
  77 static const unsigned short ksc[93][94] = {
  78 #include "ksc.h"
  79 };
  80
  81 static int fuzzycmp(const unsigned char *a, const unsigned char *b)
  82 {
  83         for (; *a && *b; a++, b++) {
  84                 while (*a && (*a|32U)-'a'>26 && *a-'0'>10U) a++;
  85                 if ((*a|32U) != *b) return 1;
  86         }
  87         return *a != *b;
  88 }
  89
  90 static size_t find_charmap(const void *name)
  91 {
  92         const unsigned char *s;
  93         if (!*(char *)name) name=charmaps; /* "utf8" */
  94         for (s=charmaps; *s; ) {
  95                 if (!fuzzycmp(name, s)) {
  96                         for (; *s; s+=strlen((void *)s)+1);
  97                         return s+1-charmaps;
  98                 }
  99                 s += strlen((void *)s)+1;
 100                 if (!*s) {
 101                         if (s[1] > 0200) s+=2;
 102                         else s+=2+(128U-s[1])/4*5;
 103                 }
 104         }
 105         return -1;
 106 }
 107
 108 iconv_t iconv_open(const char *to, const char *from)
 109 {
 110         size_t f, t;
 111
 112         if ((t = find_charmap(to))==-1
 113          || (f = find_charmap(from))==-1
 114          || (charmaps[t] >= 0320)) {
 115                 errno = EINVAL;
 116                 return (iconv_t)-1;
 117         }
 118
 119         return (void *)(f<<16 | t);
 120 }
 121
 122 int iconv_close(iconv_t cd)
 123 {
 124         return 0;
 125 }
 126
 127 static unsigned get_16(const unsigned char *s, int e)
 128 {
 129         e &= 1;
 130         return s[e]<<8 | s[1-e];
 131 }
 132
 133 static void put_16(unsigned char *s, unsigned c, int e)
 134 {
 135         e &= 1;
 136         s[e] = c>>8;
 137         s[1-e] = c;
 138 }
 139
 140 static unsigned get_32(const unsigned char *s, int e)
 141 {
 142         e &= 3;
 143         return s[e]+0U<<24 | s[e^1]<<16 | s[e^2]<<8 | s[e^3];
 144 }
 145
 146 static void put_32(unsigned char *s, unsigned c, int e)
 147 {
 148         e &= 3;
 149         s[e^0] = c>>24;
 150         s[e^1] = c>>16;
 151         s[e^2] = c>>8;
 152         s[e^3] = c;
 153 }
 154
 155 /* Adapt as needed */
 156 #define mbrtowc_utf8 mbrtowc
 157 #define wctomb_utf8 wctomb
 158
 159 size_t iconv(iconv_t cd0, char **restrict in, size_t *restrict inb, char **restrict out, size_t *restrict outb)
 160 {
 161         size_t x=0;
 162         unsigned long cd = (unsigned long)cd0;
 163         unsigned to = cd & 0xffff;
 164         unsigned from = cd >> 16;
 165         const unsigned char *map = charmaps+from+1;
 166         const unsigned char *tomap = charmaps+to+1;
 167         mbstate_t st = {0};
 168         wchar_t wc;
 169         unsigned c, d;
 170         size_t k, l;
 171         int err;
 172         unsigned char type = map[-1];
 173         unsigned char totype = tomap[-1];
 174
 175         if (!in || !*in || !*inb) return 0;
 176
 177         for (; *inb; *in+=l, *inb-=l) {
 178                 c = *(unsigned char *)*in;
 179                 l = 1;
 180
 181                 if (c >= 128 || type-UTF_32BE < 7U) switch (type) {
 182                 case UTF_8:
 183                         l = mbrtowc_utf8(&wc, *in, *inb, &st);
 184                         if (!l) l++;
 185                         else if (l == (size_t)-1) goto ilseq;
 186                         else if (l == (size_t)-2) goto starved;
 187                         c = wc;
 188                         break;
 189                 case US_ASCII:
 190                         goto ilseq;
 191                 case WCHAR_T:
 192                         l = sizeof(wchar_t);
 193                         if (*inb < l) goto starved;
 194                         c = *(wchar_t *)*in;
 195                         if (0) {
 196                 case UTF_32BE:
 197                 case UTF_32LE:
 198                         l = 4;
 199                         if (*inb < 4) goto starved;
 200                         c = get_32((void *)*in, type);
 201                         }
 202                         if (c-0xd800u < 0x800u || c >= 0x110000u) goto ilseq;
 203                         break;
 204                 case UCS2BE:
 205                 case UCS2LE:
 206                 case UTF_16BE:
 207                 case UTF_16LE:
 208                         l = 2;
 209                         if (*inb < 2) goto starved;
 210                         c = get_16((void *)*in, type);
 211                         if ((unsigned)(c-0xdc00) < 0x400) goto ilseq;
 212                         if ((unsigned)(c-0xd800) < 0x400) {
 213                                 if (type-UCS2BE < 2U) goto ilseq;
 214                                 l = 4;
 215                                 if (*inb < 4) goto starved;
 216                                 d = get_16((void *)(*in + 2), type);
 217                                 if ((unsigned)(d-0xdc00) >= 0x400) goto ilseq;
 218                                 c = ((c-0xd7c0)<<10) + (d-0xdc00);
 219                         }
 220                         break;
 221                 case SHIFT_JIS:
 222                         if (c-0xa1 <= 0xdf-0xa1) {
 223                                 c += 0xff61-0xa1;
 224                                 break;
 225                         }
 226                         l = 2;
 227                         if (*inb < 2) goto starved;
 228                         d = *((unsigned char *)*in + 1);
 229                         if (c-129 <= 159-129) c -= 129;
 230                         else if (c-224 <= 239-224) c -= 193;
 231                         else goto ilseq;
 232                         c *= 2;
 233                         if (d-64 <= 158-64) {
 234                                 if (d==127) goto ilseq;
 235                                 if (d>127) d--;
 236                                 d -= 64;
 237                         } else if (d-159 <= 252-159) {
 238                                 c++;
 239                                 d -= 159;
 240                         }
 241                         c = jis0208[c][d];
 242                         if (!c) goto ilseq;
 243                         break;
 244                 case EUC_JP:
 245                         l = 2;
 246                         if (*inb < 2) goto starved;
 247                         d = *((unsigned char *)*in + 1);
 248                         if (c==0x8e) {
 249                                 c = d;
 250                                 if (c-0xa1 > 0xdf-0xa1) goto ilseq;
 251                                 c += 0xff61 - 0xa1;
 252                                 break;
 253                         }
 254                         c -= 0xa1;
 255                         d -= 0xa1;
 256                         if (c >= 84 || d >= 94) goto ilseq;
 257                         c = jis0208[c][d];
 258                         if (!c) goto ilseq;
 259                         break;
 260                 case GB2312:
 261                         if (c < 0xa1) goto ilseq;
 262                 case GBK:
 263                 case GB18030:
 264                         c -= 0x81;
 265                         if (c >= 126) goto ilseq;
 266                         l = 2;
 267                         if (*inb < 2) goto starved;
 268                         d = *((unsigned char *)*in + 1);
 269                         if (d < 0xa1 && type == GB2312) goto ilseq;
 270                         if (d-0x40>=191 || d==127) {
 271                                 if (d-'0'>9 || type != GB18030)
 272                                         goto ilseq;
 273                                 l = 4;
 274                                 if (*inb < 4) goto starved;
 275                                 c = (10*c + d-'0') * 1260;
 276                                 d = *((unsigned char *)*in + 2);
 277                                 if (d-0x81>126) goto ilseq;
 278                                 c += 10*(d-0x81);
 279                                 d = *((unsigned char *)*in + 3);
 280                                 if (d-'0'>9) goto ilseq;
 281                                 c += d-'0';
 282                                 c += 128;
 283                                 for (d=0; d<=c; ) {
 284                                         k = 0;
 285                                         for (int i=0; i<126; i++)
 286                                                 for (int j=0; j<190; j++)
 287                                                         if (gb18030[i][j]-d <= c-d)
 288                                                                 k++;
 289                                         d = c+1;
 290                                         c += k;
 291                                 }
 292                                 break;
 293                         }
 294                         d -= 0x40;
 295                         if (d>63) d--;
 296                         c = gb18030[c][d];
 297                         break;
 298                 case BIG5:
 299                         l = 2;
 300                         if (*inb < 2) goto starved;
 301                         d = *((unsigned char *)*in + 1);
 302                         if (d-0x40>=0xff-0x40 || d-0x7f<0xa1-0x7f) goto ilseq;
 303                         d -= 0x40;
 304                         if (d > 0x3e) d -= 0x22;
 305                         if (c-0xa1>=0xfa-0xa1) {
 306                                 if (c-0x87>=0xff-0x87) goto ilseq;
 307                                 if (c < 0xa1) c -= 0x87;
 308                                 else c -= 0x87 + (0xfa-0xa1);
 309                                 c = (hkscs[4867+(c*157+d)/16]>>(c*157+d)%16)%2<<17
 310                                         | hkscs[c*157+d];
 311                                 /* A few HKSCS characters map to pairs of UCS
 312                                  * characters. These are mapped to surrogate
 313                                  * range in the hkscs table then hard-coded
 314                                  * here. Ugly, yes. */
 315                                 if (c/256 == 0xdc) {
 316                                         if (totype-0300U > 8) k = 2;
 317                                         else k = "\10\4\4\10\4\4\10\2\4"[totype-0300];
 318                                         if (k > *outb) goto toobig;
 319                                         x += iconv((iconv_t)(uintptr_t)to,
 320                                                 &(char *){"\303\212\314\204"
 321                                                 "\303\212\314\214"
 322                                                 "\303\252\314\204"
 323                                                 "\303\252\314\214"
 324                                                 +c%256}, &(size_t){4},
 325                                                 out, outb);
 326                                         continue;
 327                                 }
 328                                 if (!c) goto ilseq;
 329                                 break;
 330                         }
 331                         c -= 0xa1;
 332                         c = big5[c][d]|(c==0x27&&(d==0x3a||d==0x3c||d==0x42))<<17;
 333                         if (!c) goto ilseq;
 334                         break;
 335                 case EUC_KR:
 336                         l = 2;
 337                         if (*inb < 2) goto starved;
 338                         d = *((unsigned char *)*in + 1);
 339                         c -= 0xa1;
 340                         d -= 0xa1;
 341                         if (c >= 93 || d >= 94) {
 342                                 c += (0xa1-0x81);
 343                                 d += 0xa1;
 344                                 if (c >= 93 || c>=0xc6-0x81 && d>0x52)
 345                                         goto ilseq;
 346                                 if (d-'A'<26) d = d-'A';
 347                                 else if (d-'a'<26) d = d-'a'+26;
 348                                 else if (d-0x81<0xff-0x81) d = d-0x81+52;
 349                                 else goto ilseq;
 350                                 if (c < 0x20) c = 178*c + d;
 351                                 else c = 178*0x20 + 84*(c-0x20) + d;
 352                                 c += 0xac00;
 353                                 for (d=0xac00; d<=c; ) {
 354                                         k = 0;
 355                                         for (int i=0; i<93; i++)
 356                                                 for (int j=0; j<94; j++)
 357                                                         if (ksc[i][j]-d <= c-d)
 358                                                                 k++;
 359                                         d = c+1;
 360                                         c += k;
 361                                 }
 362                                 break;
 363                         }
 364                         c = ksc[c][d];
 365                         if (!c) goto ilseq;
 366                         break;
 367                 default:
 368                         if (c < 128+type) break;
 369                         c -= 128+type;
 370                         c = legacy_chars[ map[c*5/4]>>2*c%8 |
 371                                 map[c*5/4+1]<<8-2*c%8 & 1023 ];
 372                         if (!c) c = *(unsigned char *)*in;
 373                         if (c==1) goto ilseq;
 374                 }
 375
 376                 switch (totype) {
 377                 case WCHAR_T:
 378                         if (*outb < sizeof(wchar_t)) goto toobig;
 379                         *(wchar_t *)*out = c;
 380                         *out += sizeof(wchar_t);
 381                         *outb -= sizeof(wchar_t);
 382                         break;
 383                 case UTF_8:
 384                         if (*outb < 4) {
 385                                 char tmp[4];
 386                                 k = wctomb_utf8(tmp, c);
 387                                 if (*outb < k) goto toobig;
 388                                 memcpy(*out, tmp, k);
 389                         } else k = wctomb_utf8(*out, c);
 390                         *out += k;
 391                         *outb -= k;
 392                         break;
 393                 case US_ASCII:
 394                         if (c > 0x7f) subst: x++, c='*';
 395                 default:
 396                         if (*outb < 1) goto toobig;
 397                         if (c < 128+totype) {
 398                         revout:
 399                                 *(*out)++ = c;
 400                                 *outb -= 1;
 401                                 break;
 402                         }
 403                         d = c;
 404                         for (c=0; c<128-totype; c++) {
 405                                 if (d == legacy_chars[ tomap[c*5/4]>>2*c%8 |
 406                                         tomap[c*5/4+1]<<8-2*c%8 & 1023 ]) {
 407                                         c += 128;
 408                                         goto revout;
 409                                 }
 410                         }
 411                         goto subst;
 412                 case UCS2BE:
 413                 case UCS2LE:
 414                 case UTF_16BE:
 415                 case UTF_16LE:
 416                         if (c < 0x10000 || type-UCS2BE < 2U) {
 417                                 if (c >= 0x10000) c = 0xFFFD;
 418                                 if (*outb < 2) goto toobig;
 419                                 put_16((void *)*out, c, totype);
 420                                 *out += 2;
 421                                 *outb -= 2;
 422                                 break;
 423                         }
 424                         if (*outb < 4) goto toobig;
 425                         c -= 0x10000;
 426                         put_16((void *)*out, (c>>10)|0xd800, totype);
 427                         put_16((void *)(*out + 2), (c&0x3ff)|0xdc00, totype);
 428                         *out += 4;
 429                         *outb -= 4;
 430                         break;
 431                 case UTF_32BE:
 432                 case UTF_32LE:
 433                         if (*outb < 4) goto toobig;
 434                         put_32((void *)*out, c, totype);
 435                         *out += 4;
 436                         *outb -= 4;
 437                         break;
 438                 }
 439         }
 440         return x;
 441 ilseq:
 442         err = EILSEQ;
 443         x = -1;
 444         goto end;
 445 toobig:
 446         err = E2BIG;
 447         x = -1;
 448         goto end;
 449 starved:
 450         err = EINVAL;
 451         x = -1;
 452 end:
 453         errno = err;
 454         return x;
 455 }