2107b0555e3346fc7cef3c6f61eeba66259af5ef
[musl] / 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 ISO2022_JP  0322
22 #define GB18030     0330
23 #define GBK         0331
24 #define GB2312      0332
25 #define BIG5        0340
26 #define EUC_KR      0350
27
28 /* Definitions of charmaps. Each charmap consists of:
29  * 1. Empty-string-terminated list of null-terminated aliases.
30  * 2. Special type code or number of elided entries.
31  * 3. Character table (size determined by field 2). */
32
33 static const unsigned char charmaps[] =
34 "utf8\0char\0\0\310"
35 "wchart\0\0\306"
36 "ucs2\0ucs2be\0\0\304"
37 "ucs2le\0\0\305"
38 "utf16\0utf16be\0\0\302"
39 "utf16le\0\0\301"
40 "ucs4\0ucs4be\0utf32\0utf32be\0\0\300"
41 "ucs4le\0utf32le\0\0\303"
42 "ascii\0usascii\0iso646\0iso646us\0\0\307"
43 "eucjp\0\0\320"
44 "shiftjis\0sjis\0\0\321"
45 "iso2022jp\0\0\322"
46 "gb18030\0\0\330"
47 "gbk\0\0\331"
48 "gb2312\0\0\332"
49 "big5\0bigfive\0cp950\0big5hkscs\0\0\340"
50 "euckr\0ksc5601\0ksx1001\0cp949\0\0\350"
51 #include "codepages.h"
52 ;
53
54 static const unsigned short legacy_chars[] = {
55 #include "legacychars.h"
56 };
57
58 static const unsigned short jis0208[84][94] = {
59 #include "jis0208.h"
60 };
61
62 static const unsigned short gb18030[126][190] = {
63 #include "gb18030.h"
64 };
65
66 static const unsigned short big5[89][157] = {
67 #include "big5.h"
68 };
69
70 static const unsigned short hkscs[] = {
71 #include "hkscs.h"
72 };
73
74 static const unsigned short ksc[93][94] = {
75 #include "ksc.h"
76 };
77
78 static int fuzzycmp(const unsigned char *a, const unsigned char *b)
79 {
80         for (; *a && *b; a++, b++) {
81                 while (*a && (*a|32U)-'a'>26 && *a-'0'>10U) a++;
82                 if ((*a|32U) != *b) return 1;
83         }
84         return *a != *b;
85 }
86
87 static size_t find_charmap(const void *name)
88 {
89         const unsigned char *s;
90         if (!*(char *)name) name=charmaps; /* "utf8" */
91         for (s=charmaps; *s; ) {
92                 if (!fuzzycmp(name, s)) {
93                         for (; *s; s+=strlen((void *)s)+1);
94                         return s+1-charmaps;
95                 }
96                 s += strlen((void *)s)+1;
97                 if (!*s) {
98                         if (s[1] > 0200) s+=2;
99                         else s+=2+(128U-s[1])/4*5;
100                 }
101         }
102         return -1;
103 }
104
105 struct stateful_cd {
106         iconv_t base_cd;
107         unsigned state;
108 };
109
110 static iconv_t combine_to_from(size_t t, size_t f)
111 {
112         return (void *)(f<<16 | t<<1 | 1);
113 }
114
115 static size_t extract_from(iconv_t cd)
116 {
117         return (size_t)cd >> 16;
118 }
119
120 static size_t extract_to(iconv_t cd)
121 {
122         return (size_t)cd >> 1 & 0x7fff;
123 }
124
125 iconv_t iconv_open(const char *to, const char *from)
126 {
127         size_t f, t;
128         struct stateful_cd *scd;
129
130         if ((t = find_charmap(to))==-1
131          || (f = find_charmap(from))==-1
132          || (charmaps[t] >= 0320)) {
133                 errno = EINVAL;
134                 return (iconv_t)-1;
135         }
136         iconv_t cd = combine_to_from(t, f);
137
138         switch (charmaps[f]) {
139         case ISO2022_JP:
140                 scd = malloc(sizeof *scd);
141                 if (!scd) return (iconv_t)-1;
142                 scd->base_cd = cd;
143                 scd->state = 0;
144                 cd = (iconv_t)scd;
145         }
146
147         return cd;
148 }
149
150 static unsigned get_16(const unsigned char *s, int e)
151 {
152         e &= 1;
153         return s[e]<<8 | s[1-e];
154 }
155
156 static void put_16(unsigned char *s, unsigned c, int e)
157 {
158         e &= 1;
159         s[e] = c>>8;
160         s[1-e] = c;
161 }
162
163 static unsigned get_32(const unsigned char *s, int e)
164 {
165         e &= 3;
166         return s[e]+0U<<24 | s[e^1]<<16 | s[e^2]<<8 | s[e^3];
167 }
168
169 static void put_32(unsigned char *s, unsigned c, int e)
170 {
171         e &= 3;
172         s[e^0] = c>>24;
173         s[e^1] = c>>16;
174         s[e^2] = c>>8;
175         s[e^3] = c;
176 }
177
178 /* Adapt as needed */
179 #define mbrtowc_utf8 mbrtowc
180 #define wctomb_utf8 wctomb
181
182 static unsigned legacy_map(const unsigned char *map, unsigned c)
183 {
184         unsigned x = c - 128 - map[-1];
185         x = legacy_chars[ map[x*5/4]>>2*x%8 |
186                 map[x*5/4+1]<<8-2*x%8 & 1023 ];
187         return x ? x : c;
188 }
189
190 size_t iconv(iconv_t cd, char **restrict in, size_t *restrict inb, char **restrict out, size_t *restrict outb)
191 {
192         size_t x=0;
193         struct stateful_cd *scd=0;
194         if (!((size_t)cd & 1)) {
195                 scd = (void *)cd;
196                 cd = scd->base_cd;
197         }
198         unsigned to = extract_to(cd);
199         unsigned from = extract_from(cd);
200         const unsigned char *map = charmaps+from+1;
201         const unsigned char *tomap = charmaps+to+1;
202         mbstate_t st = {0};
203         wchar_t wc;
204         unsigned c, d;
205         size_t k, l;
206         int err;
207         unsigned char type = map[-1];
208         unsigned char totype = tomap[-1];
209         locale_t *ploc = &CURRENT_LOCALE, loc = *ploc;
210
211         if (!in || !*in || !*inb) return 0;
212
213         *ploc = UTF8_LOCALE;
214
215         for (; *inb; *in+=l, *inb-=l) {
216                 c = *(unsigned char *)*in;
217                 l = 1;
218
219                 switch (type) {
220                 case UTF_8:
221                         if (c < 128) break;
222                         l = mbrtowc_utf8(&wc, *in, *inb, &st);
223                         if (l == (size_t)-1) goto ilseq;
224                         if (l == (size_t)-2) goto starved;
225                         c = wc;
226                         break;
227                 case US_ASCII:
228                         if (c >= 128) goto ilseq;
229                         break;
230                 case WCHAR_T:
231                         l = sizeof(wchar_t);
232                         if (*inb < l) goto starved;
233                         c = *(wchar_t *)*in;
234                         if (0) {
235                 case UTF_32BE:
236                 case UTF_32LE:
237                         l = 4;
238                         if (*inb < 4) goto starved;
239                         c = get_32((void *)*in, type);
240                         }
241                         if (c-0xd800u < 0x800u || c >= 0x110000u) goto ilseq;
242                         break;
243                 case UCS2BE:
244                 case UCS2LE:
245                 case UTF_16BE:
246                 case UTF_16LE:
247                         l = 2;
248                         if (*inb < 2) goto starved;
249                         c = get_16((void *)*in, type);
250                         if ((unsigned)(c-0xdc00) < 0x400) goto ilseq;
251                         if ((unsigned)(c-0xd800) < 0x400) {
252                                 if (type-UCS2BE < 2U) goto ilseq;
253                                 l = 4;
254                                 if (*inb < 4) goto starved;
255                                 d = get_16((void *)(*in + 2), type);
256                                 if ((unsigned)(d-0xdc00) >= 0x400) goto ilseq;
257                                 c = ((c-0xd7c0)<<10) + (d-0xdc00);
258                         }
259                         break;
260                 case SHIFT_JIS:
261                         if (c < 128) break;
262                         if (c-0xa1 <= 0xdf-0xa1) {
263                                 c += 0xff61-0xa1;
264                                 break;
265                         }
266                         l = 2;
267                         if (*inb < 2) goto starved;
268                         d = *((unsigned char *)*in + 1);
269                         if (c-129 <= 159-129) c -= 129;
270                         else if (c-224 <= 239-224) c -= 193;
271                         else goto ilseq;
272                         c *= 2;
273                         if (d-64 <= 158-64) {
274                                 if (d==127) goto ilseq;
275                                 if (d>127) d--;
276                                 d -= 64;
277                         } else if (d-159 <= 252-159) {
278                                 c++;
279                                 d -= 159;
280                         }
281                         c = jis0208[c][d];
282                         if (!c) goto ilseq;
283                         break;
284                 case EUC_JP:
285                         if (c < 128) break;
286                         l = 2;
287                         if (*inb < 2) goto starved;
288                         d = *((unsigned char *)*in + 1);
289                         if (c==0x8e) {
290                                 c = d;
291                                 if (c-0xa1 > 0xdf-0xa1) goto ilseq;
292                                 c += 0xff61 - 0xa1;
293                                 break;
294                         }
295                         c -= 0xa1;
296                         d -= 0xa1;
297                         if (c >= 84 || d >= 94) goto ilseq;
298                         c = jis0208[c][d];
299                         if (!c) goto ilseq;
300                         break;
301                 case ISO2022_JP:
302                         if (c >= 128) goto ilseq;
303                         if (c == '\033') {
304                                 l = 3;
305                                 if (*inb < 3) goto starved;
306                                 c = *((unsigned char *)*in + 1);
307                                 d = *((unsigned char *)*in + 2);
308                                 if (c != '(' && c != '$') goto ilseq;
309                                 switch (128*(c=='$') + d) {
310                                 case 'B': scd->state=0; continue;
311                                 case 'J': scd->state=1; continue;
312                                 case 'I': scd->state=4; continue;
313                                 case 128+'@': scd->state=2; continue;
314                                 case 128+'B': scd->state=3; continue;
315                                 }
316                                 goto ilseq;
317                         }
318                         switch (scd->state) {
319                         case 1:
320                                 if (c=='\\') c = 0xa5;
321                                 if (c=='~') c = 0x203e;
322                                 break;
323                         case 2:
324                         case 3:
325                                 l = 2;
326                                 if (*inb < 2) goto starved;
327                                 d = *((unsigned char *)*in + 1);
328                                 c -= 0x21;
329                                 d -= 0x21;
330                                 if (c >= 84 || d >= 94) goto ilseq;
331                                 c = jis0208[c][d];
332                                 if (!c) goto ilseq;
333                                 break;
334                         case 4:
335                                 if (c-0x60 < 0x1f) goto ilseq;
336                                 if (c-0x21 < 0x5e) c += 0xff61-0x21;
337                                 break;
338                         }
339                         break;
340                 case GB2312:
341                         if (c < 128) break;
342                         if (c < 0xa1) goto ilseq;
343                 case GBK:
344                 case GB18030:
345                         if (c < 128) break;
346                         c -= 0x81;
347                         if (c >= 126) goto ilseq;
348                         l = 2;
349                         if (*inb < 2) goto starved;
350                         d = *((unsigned char *)*in + 1);
351                         if (d < 0xa1 && type == GB2312) goto ilseq;
352                         if (d-0x40>=191 || d==127) {
353                                 if (d-'0'>9 || type != GB18030)
354                                         goto ilseq;
355                                 l = 4;
356                                 if (*inb < 4) goto starved;
357                                 c = (10*c + d-'0') * 1260;
358                                 d = *((unsigned char *)*in + 2);
359                                 if (d-0x81>126) goto ilseq;
360                                 c += 10*(d-0x81);
361                                 d = *((unsigned char *)*in + 3);
362                                 if (d-'0'>9) goto ilseq;
363                                 c += d-'0';
364                                 c += 128;
365                                 for (d=0; d<=c; ) {
366                                         k = 0;
367                                         for (int i=0; i<126; i++)
368                                                 for (int j=0; j<190; j++)
369                                                         if (gb18030[i][j]-d <= c-d)
370                                                                 k++;
371                                         d = c+1;
372                                         c += k;
373                                 }
374                                 break;
375                         }
376                         d -= 0x40;
377                         if (d>63) d--;
378                         c = gb18030[c][d];
379                         break;
380                 case BIG5:
381                         if (c < 128) break;
382                         l = 2;
383                         if (*inb < 2) goto starved;
384                         d = *((unsigned char *)*in + 1);
385                         if (d-0x40>=0xff-0x40 || d-0x7f<0xa1-0x7f) goto ilseq;
386                         d -= 0x40;
387                         if (d > 0x3e) d -= 0x22;
388                         if (c-0xa1>=0xfa-0xa1) {
389                                 if (c-0x87>=0xff-0x87) goto ilseq;
390                                 if (c < 0xa1) c -= 0x87;
391                                 else c -= 0x87 + (0xfa-0xa1);
392                                 c = (hkscs[4867+(c*157+d)/16]>>(c*157+d)%16)%2<<17
393                                         | hkscs[c*157+d];
394                                 /* A few HKSCS characters map to pairs of UCS
395                                  * characters. These are mapped to surrogate
396                                  * range in the hkscs table then hard-coded
397                                  * here. Ugly, yes. */
398                                 if (c/256 == 0xdc) {
399                                         if (totype-0300U > 8) k = 2;
400                                         else k = "\10\4\4\10\4\4\10\2\4"[totype-0300];
401                                         if (k > *outb) goto toobig;
402                                         x += iconv(combine_to_from(to, 0),
403                                                 &(char *){"\303\212\314\204"
404                                                 "\303\212\314\214"
405                                                 "\303\252\314\204"
406                                                 "\303\252\314\214"
407                                                 +c%256}, &(size_t){4},
408                                                 out, outb);
409                                         continue;
410                                 }
411                                 if (!c) goto ilseq;
412                                 break;
413                         }
414                         c -= 0xa1;
415                         c = big5[c][d]|(c==0x27&&(d==0x3a||d==0x3c||d==0x42))<<17;
416                         if (!c) goto ilseq;
417                         break;
418                 case EUC_KR:
419                         if (c < 128) break;
420                         l = 2;
421                         if (*inb < 2) goto starved;
422                         d = *((unsigned char *)*in + 1);
423                         c -= 0xa1;
424                         d -= 0xa1;
425                         if (c >= 93 || d >= 94) {
426                                 c += (0xa1-0x81);
427                                 d += 0xa1;
428                                 if (c >= 93 || c>=0xc6-0x81 && d>0x52)
429                                         goto ilseq;
430                                 if (d-'A'<26) d = d-'A';
431                                 else if (d-'a'<26) d = d-'a'+26;
432                                 else if (d-0x81<0xff-0x81) d = d-0x81+52;
433                                 else goto ilseq;
434                                 if (c < 0x20) c = 178*c + d;
435                                 else c = 178*0x20 + 84*(c-0x20) + d;
436                                 c += 0xac00;
437                                 for (d=0xac00; d<=c; ) {
438                                         k = 0;
439                                         for (int i=0; i<93; i++)
440                                                 for (int j=0; j<94; j++)
441                                                         if (ksc[i][j]-d <= c-d)
442                                                                 k++;
443                                         d = c+1;
444                                         c += k;
445                                 }
446                                 break;
447                         }
448                         c = ksc[c][d];
449                         if (!c) goto ilseq;
450                         break;
451                 default:
452                         if (c < 128+type) break;
453                         c = legacy_map(map, c);
454                         if (c==1) goto ilseq;
455                 }
456
457                 switch (totype) {
458                 case WCHAR_T:
459                         if (*outb < sizeof(wchar_t)) goto toobig;
460                         *(wchar_t *)*out = c;
461                         *out += sizeof(wchar_t);
462                         *outb -= sizeof(wchar_t);
463                         break;
464                 case UTF_8:
465                         if (*outb < 4) {
466                                 char tmp[4];
467                                 k = wctomb_utf8(tmp, c);
468                                 if (*outb < k) goto toobig;
469                                 memcpy(*out, tmp, k);
470                         } else k = wctomb_utf8(*out, c);
471                         *out += k;
472                         *outb -= k;
473                         break;
474                 case US_ASCII:
475                         if (c > 0x7f) subst: x++, c='*';
476                 default:
477                         if (*outb < 1) goto toobig;
478                         if (c < 128+totype || (c<256 && c==legacy_map(tomap, c))) {
479                         revout:
480                                 *(*out)++ = c;
481                                 *outb -= 1;
482                                 break;
483                         }
484                         d = c;
485                         for (c=128+totype; c<256; c++) {
486                                 if (d == legacy_map(tomap, c)) {
487                                         goto revout;
488                                 }
489                         }
490                         goto subst;
491                 case UCS2BE:
492                 case UCS2LE:
493                 case UTF_16BE:
494                 case UTF_16LE:
495                         if (c < 0x10000 || type-UCS2BE < 2U) {
496                                 if (c >= 0x10000) c = 0xFFFD;
497                                 if (*outb < 2) goto toobig;
498                                 put_16((void *)*out, c, totype);
499                                 *out += 2;
500                                 *outb -= 2;
501                                 break;
502                         }
503                         if (*outb < 4) goto toobig;
504                         c -= 0x10000;
505                         put_16((void *)*out, (c>>10)|0xd800, totype);
506                         put_16((void *)(*out + 2), (c&0x3ff)|0xdc00, totype);
507                         *out += 4;
508                         *outb -= 4;
509                         break;
510                 case UTF_32BE:
511                 case UTF_32LE:
512                         if (*outb < 4) goto toobig;
513                         put_32((void *)*out, c, totype);
514                         *out += 4;
515                         *outb -= 4;
516                         break;
517                 }
518         }
519         *ploc = loc;
520         return x;
521 ilseq:
522         err = EILSEQ;
523         x = -1;
524         goto end;
525 toobig:
526         err = E2BIG;
527         x = -1;
528         goto end;
529 starved:
530         err = EINVAL;
531         x = -1;
532 end:
533         errno = err;
534         *ploc = loc;
535         return x;
536 }