8 #include "locale_impl.h"
23 #define SHIFT_JIS 0321
24 #define ISO2022_JP 0322
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 quads of entries.
34 * 3. Character table (size determined by field 2), consisting
35 * of 5 bytes for every 4 characters, interpreted as 10-bit
36 * indices into the legacy_chars table. */
38 static const unsigned char charmaps[] =
45 "ucs4be\0utf32be\0\0\300"
46 "ucs4le\0utf32le\0\0\303"
47 "ascii\0usascii\0iso646\0iso646us\0\0\307"
52 "shiftjis\0sjis\0\0\321"
57 "big5\0bigfive\0cp950\0big5hkscs\0\0\340"
58 "euckr\0ksc5601\0ksx1001\0cp949\0\0\350"
59 #include "codepages.h"
62 /* Table of characters that appear in legacy 8-bit codepages,
63 * limited to 1024 slots (10 bit indices). The first 256 entries
64 * are elided since those characters are obviously all included. */
65 static const unsigned short legacy_chars[] = {
66 #include "legacychars.h"
69 static const unsigned short jis0208[84][94] = {
73 static const unsigned short gb18030[126][190] = {
77 static const unsigned short big5[89][157] = {
81 static const unsigned short hkscs[] = {
85 static const unsigned short ksc[93][94] = {
89 static const unsigned short rev_jis[] = {
93 static int fuzzycmp(const unsigned char *a, const unsigned char *b)
95 for (; *a && *b; a++, b++) {
96 while (*a && (*a|32U)-'a'>26 && *a-'0'>10U) a++;
97 if ((*a|32U) != *b) return 1;
102 static size_t find_charmap(const void *name)
104 const unsigned char *s;
105 if (!*(char *)name) name=charmaps; /* "utf8" */
106 for (s=charmaps; *s; ) {
107 if (!fuzzycmp(name, s)) {
108 for (; *s; s+=strlen((void *)s)+1);
111 s += strlen((void *)s)+1;
113 if (s[1] > 0200) s+=2;
114 else s+=2+(64U-s[1])*5;
125 static iconv_t combine_to_from(size_t t, size_t f)
127 return (void *)(f<<16 | t<<1 | 1);
130 static size_t extract_from(iconv_t cd)
132 return (size_t)cd >> 16;
135 static size_t extract_to(iconv_t cd)
137 return (size_t)cd >> 1 & 0x7fff;
140 iconv_t iconv_open(const char *to, const char *from)
143 struct stateful_cd *scd;
145 if ((t = find_charmap(to))==-1
146 || (f = find_charmap(from))==-1
147 || (charmaps[t] >= 0330)) {
151 iconv_t cd = combine_to_from(t, f);
153 switch (charmaps[f]) {
158 scd = malloc(sizeof *scd);
159 if (!scd) return (iconv_t)-1;
168 static unsigned get_16(const unsigned char *s, int e)
171 return s[e]<<8 | s[1-e];
174 static void put_16(unsigned char *s, unsigned c, int e)
181 static unsigned get_32(const unsigned char *s, int e)
184 return s[e]+0U<<24 | s[e^1]<<16 | s[e^2]<<8 | s[e^3];
187 static void put_32(unsigned char *s, unsigned c, int e)
196 /* Adapt as needed */
197 #define mbrtowc_utf8 mbrtowc
198 #define wctomb_utf8 wctomb
200 static unsigned legacy_map(const unsigned char *map, unsigned c)
202 if (c < 4*map[-1]) return c;
203 unsigned x = c - 4*map[-1];
204 x = map[x*5/4]>>2*x%8 | map[x*5/4+1]<<8-2*x%8 & 1023;
205 return x < 256 ? x : legacy_chars[x-256];
208 static unsigned uni_to_jis(unsigned c)
210 unsigned nel = sizeof rev_jis / sizeof *rev_jis;
211 unsigned d, j, i, b = 0;
215 d = jis0208[j/256][j%256];
216 if (d==c) return j + 0x2121;
217 else if (nel == 1) return 0;
227 size_t iconv(iconv_t cd, char **restrict in, size_t *restrict inb, char **restrict out, size_t *restrict outb)
230 struct stateful_cd *scd=0;
231 if (!((size_t)cd & 1)) {
235 unsigned to = extract_to(cd);
236 unsigned from = extract_from(cd);
237 const unsigned char *map = charmaps+from+1;
238 const unsigned char *tomap = charmaps+to+1;
244 unsigned char type = map[-1];
245 unsigned char totype = tomap[-1];
246 locale_t *ploc = &CURRENT_LOCALE, loc = *ploc;
248 if (!in || !*in || !*inb) return 0;
252 for (; *inb; *in+=l, *inb-=l) {
253 c = *(unsigned char *)*in;
259 l = mbrtowc_utf8(&wc, *in, *inb, &st);
260 if (l == (size_t)-1) goto ilseq;
261 if (l == (size_t)-2) goto starved;
265 if (c >= 128) goto ilseq;
269 if (*inb < l) goto starved;
275 if (*inb < 4) goto starved;
276 c = get_32((void *)*in, type);
278 if (c-0xd800u < 0x800u || c >= 0x110000u) goto ilseq;
285 if (*inb < 2) goto starved;
286 c = get_16((void *)*in, type);
287 if ((unsigned)(c-0xdc00) < 0x400) goto ilseq;
288 if ((unsigned)(c-0xd800) < 0x400) {
289 if (type-UCS2BE < 2U) goto ilseq;
291 if (*inb < 4) goto starved;
292 d = get_16((void *)(*in + 2), type);
293 if ((unsigned)(d-0xdc00) >= 0x400) goto ilseq;
294 c = ((c-0xd7c0)<<10) + (d-0xdc00);
301 if (*inb < 2) goto starved;
302 c = get_16((void *)*in, 0);
303 scd->state = type==UCS2
304 ? c==0xfffe ? UCS2LE : UCS2BE
305 : c==0xfffe ? UTF_16LE : UTF_16BE;
306 if (c == 0xfffe || c == 0xfeff)
314 if (*inb < 4) goto starved;
315 c = get_32((void *)*in, 0);
316 scd->state = c==0xfffe0000 ? UTF_32LE : UTF_32BE;
317 if (c == 0xfffe0000 || c == 0xfeff)
324 if (c-0xa1 <= 0xdf-0xa1) {
329 if (*inb < 2) goto starved;
330 d = *((unsigned char *)*in + 1);
331 if (c-129 <= 159-129) c -= 129;
332 else if (c-224 <= 239-224) c -= 193;
335 if (d-64 <= 158-64) {
336 if (d==127) goto ilseq;
339 } else if (d-159 <= 252-159) {
349 if (*inb < 2) goto starved;
350 d = *((unsigned char *)*in + 1);
353 if (c-0xa1 > 0xdf-0xa1) goto ilseq;
359 if (c >= 84 || d >= 94) goto ilseq;
364 if (c >= 128) goto ilseq;
367 if (*inb < 3) goto starved;
368 c = *((unsigned char *)*in + 1);
369 d = *((unsigned char *)*in + 2);
370 if (c != '(' && c != '$') goto ilseq;
371 switch (128*(c=='$') + d) {
372 case 'B': scd->state=0; continue;
373 case 'J': scd->state=1; continue;
374 case 'I': scd->state=4; continue;
375 case 128+'@': scd->state=2; continue;
376 case 128+'B': scd->state=3; continue;
380 switch (scd->state) {
382 if (c=='\\') c = 0xa5;
383 if (c=='~') c = 0x203e;
388 if (*inb < 2) goto starved;
389 d = *((unsigned char *)*in + 1);
392 if (c >= 84 || d >= 94) goto ilseq;
397 if (c-0x60 < 0x1f) goto ilseq;
398 if (c-0x21 < 0x5e) c += 0xff61-0x21;
404 if (c < 0xa1) goto ilseq;
409 if (c >= 126) goto ilseq;
411 if (*inb < 2) goto starved;
412 d = *((unsigned char *)*in + 1);
413 if (d < 0xa1 && type == GB2312) goto ilseq;
414 if (d-0x40>=191 || d==127) {
415 if (d-'0'>9 || type != GB18030)
418 if (*inb < 4) goto starved;
419 c = (10*c + d-'0') * 1260;
420 d = *((unsigned char *)*in + 2);
421 if (d-0x81>126) goto ilseq;
423 d = *((unsigned char *)*in + 3);
424 if (d-'0'>9) goto ilseq;
429 for (int i=0; i<126; i++)
430 for (int j=0; j<190; j++)
431 if (gb18030[i][j]-d <= c-d)
445 if (*inb < 2) goto starved;
446 d = *((unsigned char *)*in + 1);
447 if (d-0x40>=0xff-0x40 || d-0x7f<0xa1-0x7f) goto ilseq;
449 if (d > 0x3e) d -= 0x22;
450 if (c-0xa1>=0xfa-0xa1) {
451 if (c-0x87>=0xff-0x87) goto ilseq;
452 if (c < 0xa1) c -= 0x87;
453 else c -= 0x87 + (0xfa-0xa1);
454 c = (hkscs[4867+(c*157+d)/16]>>(c*157+d)%16)%2<<17
456 /* A few HKSCS characters map to pairs of UCS
457 * characters. These are mapped to surrogate
458 * range in the hkscs table then hard-coded
459 * here. Ugly, yes. */
466 size_t tmpx = iconv(combine_to_from(to, find_charmap("utf8")),
467 &(char *){"\303\212\314\204"
471 +c%256}, &(size_t){4},
472 &ptmp, &(size_t){sizeof tmp});
473 size_t tmplen = ptmp - tmp.c;
474 if (tmplen > *outb) goto toobig;
476 memcpy(*out, &tmp, tmplen);
485 c = big5[c][d]|(c==0x27&&(d==0x3a||d==0x3c||d==0x42))<<17;
491 if (*inb < 2) goto starved;
492 d = *((unsigned char *)*in + 1);
495 if (c >= 93 || d >= 94) {
498 if (c >= 93 || c>=0xc6-0x81 && d>0x52)
500 if (d-'A'<26) d = d-'A';
501 else if (d-'a'<26) d = d-'a'+26;
502 else if (d-0x81<0xff-0x81) d = d-0x81+52;
504 if (c < 0x20) c = 178*c + d;
505 else c = 178*0x20 + 84*(c-0x20) + d;
507 for (d=0xac00; d<=c; ) {
509 for (int i=0; i<93; i++)
510 for (int j=0; j<94; j++)
511 if (ksc[i][j]-d <= c-d)
523 c = legacy_map(map, c);
529 if (*outb < sizeof(wchar_t)) goto toobig;
530 *(wchar_t *)*out = c;
531 *out += sizeof(wchar_t);
532 *outb -= sizeof(wchar_t);
537 k = wctomb_utf8(tmp, c);
538 if (*outb < k) goto toobig;
539 memcpy(*out, tmp, k);
540 } else k = wctomb_utf8(*out, c);
545 if (c > 0x7f) subst: x++, c='*';
547 if (*outb < 1) goto toobig;
548 if (c<256 && c==legacy_map(tomap, c)) {
550 if (*outb < 1) goto toobig;
556 for (c=4*totype; c<256; c++) {
557 if (d == legacy_map(tomap, c)) {
563 if (c < 128) goto revout;
574 if (c-0xff61 <= 0xdf-0xa1) {
580 if (*outb < 2) goto toobig;
583 *(*out)++ = (c+1)/2 + (c<95 ? 112 : 176);
584 *(*out)++ = c%2 ? d + 31 + d/96 : d + 126;
588 if (c < 128) goto revout;
589 if (c-0xff61 <= 0xdf-0xa1) {
590 c += 0x0e00 + 0x21 - 0xff61;
595 if (*outb < 2) goto toobig;
596 *(*out)++ = c/256 + 0x80;
597 *(*out)++ = c%256 + 0x80;
601 if (c < 128) goto revout;
602 if (c-0xff61 <= 0xdf-0xa1 || c==0xa5 || c==0x203e) {
603 if (*outb < 7) goto toobig;
609 } else if (c==0x203e) {
614 *(*out)++ = c-0xff61+0x21;
624 if (*outb < 8) goto toobig;
642 if (c < 0x10000 || totype-UCS2BE < 2U) {
643 if (c >= 0x10000) c = 0xFFFD;
644 if (*outb < 2) goto toobig;
645 put_16((void *)*out, c, totype);
650 if (*outb < 4) goto toobig;
652 put_16((void *)*out, (c>>10)|0xd800, totype);
653 put_16((void *)(*out + 2), (c&0x3ff)|0xdc00, totype);
661 if (*outb < 4) goto toobig;
662 put_32((void *)*out, c, totype);