Commit | Line | Data |
a0ed51b3 |
1 | /* utf8.c |
2 | * |
3818b22b |
3 | * Copyright (c) 1998-2000, Larry Wall |
a0ed51b3 |
4 | * |
5 | * You may distribute under the terms of either the GNU General Public |
6 | * License or the Artistic License, as specified in the README file. |
7 | * |
8 | */ |
9 | |
10 | /* |
11 | * 'What a fix!' said Sam. 'That's the one place in all the lands we've ever |
12 | * heard of that we don't want to see any closer; and that's the one place |
13 | * we're trying to get to! And that's just where we can't get, nohow.' |
14 | * |
15 | * 'Well do I understand your speech,' he answered in the same language; |
16 | * 'yet few strangers do so. Why then do you not speak in the Common Tongue, |
17 | * as is the custom in the West, if you wish to be answered?' |
18 | * |
19 | * ...the travellers perceived that the floor was paved with stones of many |
20 | * hues; branching runes and strange devices intertwined beneath their feet. |
21 | */ |
22 | |
23 | #include "EXTERN.h" |
864dbfa3 |
24 | #define PERL_IN_UTF8_C |
a0ed51b3 |
25 | #include "perl.h" |
26 | |
27 | /* Unicode support */ |
28 | |
dfe13c55 |
29 | U8 * |
864dbfa3 |
30 | Perl_uv_to_utf8(pTHX_ U8 *d, UV uv) |
a0ed51b3 |
31 | { |
32 | if (uv < 0x80) { |
33 | *d++ = uv; |
34 | return d; |
35 | } |
36 | if (uv < 0x800) { |
37 | *d++ = (( uv >> 6) | 0xc0); |
38 | *d++ = (( uv & 0x3f) | 0x80); |
39 | return d; |
40 | } |
41 | if (uv < 0x10000) { |
42 | *d++ = (( uv >> 12) | 0xe0); |
43 | *d++ = (((uv >> 6) & 0x3f) | 0x80); |
44 | *d++ = (( uv & 0x3f) | 0x80); |
45 | return d; |
46 | } |
47 | if (uv < 0x200000) { |
48 | *d++ = (( uv >> 18) | 0xf0); |
49 | *d++ = (((uv >> 12) & 0x3f) | 0x80); |
50 | *d++ = (((uv >> 6) & 0x3f) | 0x80); |
51 | *d++ = (( uv & 0x3f) | 0x80); |
52 | return d; |
53 | } |
54 | if (uv < 0x4000000) { |
55 | *d++ = (( uv >> 24) | 0xf8); |
56 | *d++ = (((uv >> 18) & 0x3f) | 0x80); |
57 | *d++ = (((uv >> 12) & 0x3f) | 0x80); |
58 | *d++ = (((uv >> 6) & 0x3f) | 0x80); |
59 | *d++ = (( uv & 0x3f) | 0x80); |
60 | return d; |
61 | } |
62 | if (uv < 0x80000000) { |
63 | *d++ = (( uv >> 30) | 0xfc); |
64 | *d++ = (((uv >> 24) & 0x3f) | 0x80); |
65 | *d++ = (((uv >> 18) & 0x3f) | 0x80); |
66 | *d++ = (((uv >> 12) & 0x3f) | 0x80); |
67 | *d++ = (((uv >> 6) & 0x3f) | 0x80); |
68 | *d++ = (( uv & 0x3f) | 0x80); |
69 | return d; |
70 | } |
6b8eaf93 |
71 | #ifdef HAS_QUAD |
628e1a40 |
72 | if (uv < 0x1000000000LL) |
a0ed51b3 |
73 | #endif |
74 | { |
75 | *d++ = 0xfe; /* Can't match U+FEFF! */ |
76 | *d++ = (((uv >> 30) & 0x3f) | 0x80); |
77 | *d++ = (((uv >> 24) & 0x3f) | 0x80); |
78 | *d++ = (((uv >> 18) & 0x3f) | 0x80); |
79 | *d++ = (((uv >> 12) & 0x3f) | 0x80); |
80 | *d++ = (((uv >> 6) & 0x3f) | 0x80); |
81 | *d++ = (( uv & 0x3f) | 0x80); |
82 | return d; |
83 | } |
6b8eaf93 |
84 | #ifdef HAS_QUAD |
a0ed51b3 |
85 | { |
86 | *d++ = 0xff; /* Can't match U+FFFE! */ |
3c77ea2b |
87 | *d++ = 0x80; /* 6 Reserved bits */ |
88 | *d++ = (((uv >> 60) & 0x0f) | 0x80); /* 2 Reserved bits */ |
89 | *d++ = (((uv >> 54) & 0x3f) | 0x80); |
90 | *d++ = (((uv >> 48) & 0x3f) | 0x80); |
91 | *d++ = (((uv >> 42) & 0x3f) | 0x80); |
a0ed51b3 |
92 | *d++ = (((uv >> 36) & 0x3f) | 0x80); |
93 | *d++ = (((uv >> 30) & 0x3f) | 0x80); |
94 | *d++ = (((uv >> 24) & 0x3f) | 0x80); |
95 | *d++ = (((uv >> 18) & 0x3f) | 0x80); |
96 | *d++ = (((uv >> 12) & 0x3f) | 0x80); |
97 | *d++ = (((uv >> 6) & 0x3f) | 0x80); |
98 | *d++ = (( uv & 0x3f) | 0x80); |
99 | return d; |
100 | } |
101 | #endif |
102 | } |
103 | |
386d01d6 |
104 | /* Tests if some arbitrary number of bytes begins in a valid UTF-8 character. |
105 | * The actual number of bytes in the UTF-8 character will be returned if it |
106 | * is valid, otherwise 0. */ |
107 | int |
108 | Perl_is_utf8_char(pTHX_ U8 *s) |
109 | { |
110 | U8 u = *s; |
111 | int slen, len; |
112 | |
113 | if (!(u & 0x80)) |
114 | return 1; |
115 | |
116 | if (!(u & 0x40)) |
117 | return 0; |
118 | |
119 | if (!(u & 0x20)) { len = 2; } |
120 | else if (!(u & 0x10)) { len = 3; } |
121 | else if (!(u & 0x08)) { len = 4; } |
122 | else if (!(u & 0x04)) { len = 5; } |
123 | else if (!(u & 0x02)) { len = 6; } |
124 | else if (!(u & 0x01)) { len = 7; } |
125 | else { len = 13; } /* whoa! */ |
126 | |
127 | slen = len - 1; |
128 | s++; |
129 | while (slen--) { |
130 | if ((*s & 0xc0) != 0x80) |
131 | return 0; |
132 | s++; |
133 | } |
134 | return len; |
135 | } |
136 | |
6662521e |
137 | /* |
138 | =for apidoc Am|bool_utf8_string|U8 *s|STRLEN len |
139 | |
140 | Returns true if first C<len> bytes of the given string form valid a UTF8 |
141 | string, false otherwise. |
142 | |
143 | =cut |
144 | */ |
145 | |
146 | bool |
147 | Perl_is_utf8_string(pTHX_ U8 *s, STRLEN len) |
148 | { |
149 | U8* x=s; |
150 | U8* send=s+len; |
151 | int c; |
152 | while (x < send) { |
153 | c = is_utf8_char(x); |
154 | x += c; |
155 | if (!c || x > send) |
156 | return 0; |
157 | } |
158 | return 1; |
159 | } |
160 | |
a0ed51b3 |
161 | UV |
864dbfa3 |
162 | Perl_utf8_to_uv(pTHX_ U8* s, I32* retlen) |
a0ed51b3 |
163 | { |
164 | UV uv = *s; |
165 | int len; |
166 | if (!(uv & 0x80)) { |
167 | if (retlen) |
168 | *retlen = 1; |
169 | return *s; |
170 | } |
171 | if (!(uv & 0x40)) { |
0453d815 |
172 | dTHR; |
173 | if (ckWARN_d(WARN_UTF8)) |
174 | Perl_warner(aTHX_ WARN_UTF8, "Malformed UTF-8 character"); |
a0ed51b3 |
175 | if (retlen) |
176 | *retlen = 1; |
177 | return *s; |
178 | } |
179 | |
180 | if (!(uv & 0x20)) { len = 2; uv &= 0x1f; } |
181 | else if (!(uv & 0x10)) { len = 3; uv &= 0x0f; } |
182 | else if (!(uv & 0x08)) { len = 4; uv &= 0x07; } |
183 | else if (!(uv & 0x04)) { len = 5; uv &= 0x03; } |
184 | else if (!(uv & 0x02)) { len = 6; uv &= 0x01; } |
3c77ea2b |
185 | else if (!(uv & 0x01)) { len = 7; uv = 0; } |
186 | else { len = 13; uv = 0; } /* whoa! */ |
a0ed51b3 |
187 | |
188 | if (retlen) |
189 | *retlen = len; |
190 | --len; |
191 | s++; |
192 | while (len--) { |
193 | if ((*s & 0xc0) != 0x80) { |
0453d815 |
194 | dTHR; |
195 | if (ckWARN_d(WARN_UTF8)) |
196 | Perl_warner(aTHX_ WARN_UTF8, "Malformed UTF-8 character"); |
a0ed51b3 |
197 | if (retlen) |
198 | *retlen -= len + 1; |
199 | return 0xfffd; |
200 | } |
201 | else |
202 | uv = (uv << 6) | (*s++ & 0x3f); |
203 | } |
204 | return uv; |
205 | } |
206 | |
207 | /* utf8_distance(a,b) is intended to be a - b in pointer arithmetic */ |
208 | |
209 | I32 |
864dbfa3 |
210 | Perl_utf8_distance(pTHX_ U8 *a, U8 *b) |
a0ed51b3 |
211 | { |
212 | I32 off = 0; |
213 | if (a < b) { |
214 | while (a < b) { |
215 | a += UTF8SKIP(a); |
216 | off--; |
217 | } |
218 | } |
219 | else { |
220 | while (b < a) { |
221 | b += UTF8SKIP(b); |
222 | off++; |
223 | } |
224 | } |
225 | return off; |
226 | } |
227 | |
228 | /* WARNING: do not use the following unless you *know* off is within bounds */ |
229 | |
230 | U8 * |
864dbfa3 |
231 | Perl_utf8_hop(pTHX_ U8 *s, I32 off) |
a0ed51b3 |
232 | { |
233 | if (off >= 0) { |
234 | while (off--) |
235 | s += UTF8SKIP(s); |
236 | } |
237 | else { |
238 | while (off++) { |
239 | s--; |
240 | if (*s & 0x80) { |
241 | while ((*s & 0xc0) == 0x80) |
242 | s--; |
243 | } |
244 | } |
245 | } |
246 | return s; |
247 | } |
248 | |
6940069f |
249 | /* |
497711e7 |
250 | =for apidoc Am|U8 *|utf8_to_bytes|U8 *s|STRLEN len |
6940069f |
251 | |
252 | Converts a string C<s> of length C<len> from UTF8 into ASCII encoding. |
253 | Unlike C<bytes_to_utf8>, this over-writes the original string. |
6662521e |
254 | Returns zero on failure after converting as much as possible. |
6940069f |
255 | |
256 | =cut |
257 | */ |
258 | |
259 | U8 * |
260 | Perl_utf8_to_bytes(pTHX_ U8* s, STRLEN len) |
261 | { |
262 | dTHR; |
263 | U8 *send; |
264 | U8 *d; |
265 | U8 *save; |
266 | |
267 | send = s + len; |
268 | d = save = s; |
269 | while (s < send) { |
270 | if (*s < 0x80) |
271 | *d++ = *s++; |
272 | else { |
273 | I32 ulen; |
274 | UV uv = utf8_to_uv(s, &ulen); |
6662521e |
275 | if (uv > 255) { |
276 | *d = '\0'; |
277 | return 0; |
278 | } |
6940069f |
279 | s += ulen; |
280 | *d++ = (U8)uv; |
281 | } |
282 | } |
283 | *d = '\0'; |
284 | return save; |
285 | } |
286 | |
287 | /* |
6662521e |
288 | =for apidoc Am|U8 *|bytes_to_utf8|U8 *s|STRLEN *len |
6940069f |
289 | |
290 | Converts a string C<s> of length C<len> from ASCII into UTF8 encoding. |
6662521e |
291 | Returns a pointer to the newly-created string, and sets C<len> to |
292 | reflect the new length. |
6940069f |
293 | |
497711e7 |
294 | =cut |
6940069f |
295 | */ |
296 | |
297 | U8* |
6662521e |
298 | Perl_bytes_to_utf8(pTHX_ U8* s, STRLEN *len) |
6940069f |
299 | { |
300 | dTHR; |
301 | U8 *send; |
302 | U8 *d; |
303 | U8 *dst; |
6662521e |
304 | send = s + (*len); |
6940069f |
305 | |
6662521e |
306 | Newz(801, d, (*len) * 2 + 1, U8); |
6940069f |
307 | dst = d; |
308 | |
309 | while (s < send) { |
310 | if (*s < 0x80) |
311 | *d++ = *s++; |
312 | else { |
313 | UV uv = *s++; |
314 | *d++ = (( uv >> 6) | 0xc0); |
315 | *d++ = (( uv & 0x3f) | 0x80); |
316 | } |
317 | } |
318 | *d = '\0'; |
6662521e |
319 | *len = d-dst; |
6940069f |
320 | return dst; |
321 | } |
322 | |
a0ed51b3 |
323 | /* |
dea0fc0b |
324 | * Convert native (big-endian) or reversed (little-endian) UTF-16 to UTF-8. |
a0ed51b3 |
325 | * |
326 | * Destination must be pre-extended to 3/2 source. Do not use in-place. |
327 | * We optimize for native, for obvious reasons. */ |
328 | |
329 | U8* |
dea0fc0b |
330 | Perl_utf16_to_utf8(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen) |
a0ed51b3 |
331 | { |
dea0fc0b |
332 | U8* pend; |
333 | U8* dstart = d; |
334 | |
335 | if (bytelen & 1) |
a7867d0a |
336 | Perl_croak(aTHX_ "panic: utf16_to_utf8: odd bytelen"); |
dea0fc0b |
337 | |
338 | pend = p + bytelen; |
339 | |
a0ed51b3 |
340 | while (p < pend) { |
dea0fc0b |
341 | UV uv = (p[0] << 8) + p[1]; /* UTF-16BE */ |
342 | p += 2; |
a0ed51b3 |
343 | if (uv < 0x80) { |
344 | *d++ = uv; |
345 | continue; |
346 | } |
347 | if (uv < 0x800) { |
348 | *d++ = (( uv >> 6) | 0xc0); |
349 | *d++ = (( uv & 0x3f) | 0x80); |
350 | continue; |
351 | } |
352 | if (uv >= 0xd800 && uv < 0xdbff) { /* surrogates */ |
0453d815 |
353 | dTHR; |
dea0fc0b |
354 | UV low = *p++; |
355 | if (low < 0xdc00 || low >= 0xdfff) |
356 | Perl_croak(aTHX_ "Malformed UTF-16 surrogate"); |
a0ed51b3 |
357 | uv = ((uv - 0xd800) << 10) + (low - 0xdc00) + 0x10000; |
358 | } |
359 | if (uv < 0x10000) { |
360 | *d++ = (( uv >> 12) | 0xe0); |
361 | *d++ = (((uv >> 6) & 0x3f) | 0x80); |
362 | *d++ = (( uv & 0x3f) | 0x80); |
363 | continue; |
364 | } |
365 | else { |
366 | *d++ = (( uv >> 18) | 0xf0); |
367 | *d++ = (((uv >> 12) & 0x3f) | 0x80); |
368 | *d++ = (((uv >> 6) & 0x3f) | 0x80); |
369 | *d++ = (( uv & 0x3f) | 0x80); |
370 | continue; |
371 | } |
372 | } |
dea0fc0b |
373 | *newlen = d - dstart; |
a0ed51b3 |
374 | return d; |
375 | } |
376 | |
377 | /* Note: this one is slightly destructive of the source. */ |
378 | |
379 | U8* |
dea0fc0b |
380 | Perl_utf16_to_utf8_reversed(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen) |
a0ed51b3 |
381 | { |
382 | U8* s = (U8*)p; |
383 | U8* send = s + bytelen; |
384 | while (s < send) { |
385 | U8 tmp = s[0]; |
386 | s[0] = s[1]; |
387 | s[1] = tmp; |
388 | s += 2; |
389 | } |
dea0fc0b |
390 | return utf16_to_utf8(p, d, bytelen, newlen); |
a0ed51b3 |
391 | } |
392 | |
393 | /* for now these are all defined (inefficiently) in terms of the utf8 versions */ |
394 | |
395 | bool |
864dbfa3 |
396 | Perl_is_uni_alnum(pTHX_ U32 c) |
a0ed51b3 |
397 | { |
aa6ffa16 |
398 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
399 | uv_to_utf8(tmpbuf, (UV)c); |
400 | return is_utf8_alnum(tmpbuf); |
401 | } |
402 | |
403 | bool |
b8c5462f |
404 | Perl_is_uni_alnumc(pTHX_ U32 c) |
405 | { |
aa6ffa16 |
406 | U8 tmpbuf[UTF8_MAXLEN]; |
b8c5462f |
407 | uv_to_utf8(tmpbuf, (UV)c); |
408 | return is_utf8_alnumc(tmpbuf); |
409 | } |
410 | |
411 | bool |
864dbfa3 |
412 | Perl_is_uni_idfirst(pTHX_ U32 c) |
a0ed51b3 |
413 | { |
aa6ffa16 |
414 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
415 | uv_to_utf8(tmpbuf, (UV)c); |
416 | return is_utf8_idfirst(tmpbuf); |
417 | } |
418 | |
419 | bool |
864dbfa3 |
420 | Perl_is_uni_alpha(pTHX_ U32 c) |
a0ed51b3 |
421 | { |
aa6ffa16 |
422 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
423 | uv_to_utf8(tmpbuf, (UV)c); |
424 | return is_utf8_alpha(tmpbuf); |
425 | } |
426 | |
427 | bool |
4d61ec05 |
428 | Perl_is_uni_ascii(pTHX_ U32 c) |
429 | { |
aa6ffa16 |
430 | U8 tmpbuf[UTF8_MAXLEN]; |
4d61ec05 |
431 | uv_to_utf8(tmpbuf, (UV)c); |
432 | return is_utf8_ascii(tmpbuf); |
433 | } |
434 | |
435 | bool |
864dbfa3 |
436 | Perl_is_uni_space(pTHX_ U32 c) |
a0ed51b3 |
437 | { |
aa6ffa16 |
438 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
439 | uv_to_utf8(tmpbuf, (UV)c); |
440 | return is_utf8_space(tmpbuf); |
441 | } |
442 | |
443 | bool |
864dbfa3 |
444 | Perl_is_uni_digit(pTHX_ U32 c) |
a0ed51b3 |
445 | { |
aa6ffa16 |
446 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
447 | uv_to_utf8(tmpbuf, (UV)c); |
448 | return is_utf8_digit(tmpbuf); |
449 | } |
450 | |
451 | bool |
864dbfa3 |
452 | Perl_is_uni_upper(pTHX_ U32 c) |
a0ed51b3 |
453 | { |
aa6ffa16 |
454 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
455 | uv_to_utf8(tmpbuf, (UV)c); |
456 | return is_utf8_upper(tmpbuf); |
457 | } |
458 | |
459 | bool |
864dbfa3 |
460 | Perl_is_uni_lower(pTHX_ U32 c) |
a0ed51b3 |
461 | { |
aa6ffa16 |
462 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
463 | uv_to_utf8(tmpbuf, (UV)c); |
464 | return is_utf8_lower(tmpbuf); |
465 | } |
466 | |
467 | bool |
b8c5462f |
468 | Perl_is_uni_cntrl(pTHX_ U32 c) |
469 | { |
aa6ffa16 |
470 | U8 tmpbuf[UTF8_MAXLEN]; |
b8c5462f |
471 | uv_to_utf8(tmpbuf, (UV)c); |
472 | return is_utf8_cntrl(tmpbuf); |
473 | } |
474 | |
475 | bool |
476 | Perl_is_uni_graph(pTHX_ U32 c) |
477 | { |
aa6ffa16 |
478 | U8 tmpbuf[UTF8_MAXLEN]; |
b8c5462f |
479 | uv_to_utf8(tmpbuf, (UV)c); |
480 | return is_utf8_graph(tmpbuf); |
481 | } |
482 | |
483 | bool |
864dbfa3 |
484 | Perl_is_uni_print(pTHX_ U32 c) |
a0ed51b3 |
485 | { |
aa6ffa16 |
486 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
487 | uv_to_utf8(tmpbuf, (UV)c); |
488 | return is_utf8_print(tmpbuf); |
489 | } |
490 | |
b8c5462f |
491 | bool |
f248d071 |
492 | Perl_is_uni_punct(pTHX_ U32 c) |
b8c5462f |
493 | { |
aa6ffa16 |
494 | U8 tmpbuf[UTF8_MAXLEN]; |
b8c5462f |
495 | uv_to_utf8(tmpbuf, (UV)c); |
496 | return is_utf8_punct(tmpbuf); |
497 | } |
498 | |
4d61ec05 |
499 | bool |
500 | Perl_is_uni_xdigit(pTHX_ U32 c) |
501 | { |
aa6ffa16 |
502 | U8 tmpbuf[UTF8_MAXLEN]; |
4d61ec05 |
503 | uv_to_utf8(tmpbuf, (UV)c); |
504 | return is_utf8_xdigit(tmpbuf); |
505 | } |
506 | |
a0ed51b3 |
507 | U32 |
864dbfa3 |
508 | Perl_to_uni_upper(pTHX_ U32 c) |
a0ed51b3 |
509 | { |
aa6ffa16 |
510 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
511 | uv_to_utf8(tmpbuf, (UV)c); |
512 | return to_utf8_upper(tmpbuf); |
513 | } |
514 | |
515 | U32 |
864dbfa3 |
516 | Perl_to_uni_title(pTHX_ U32 c) |
a0ed51b3 |
517 | { |
aa6ffa16 |
518 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
519 | uv_to_utf8(tmpbuf, (UV)c); |
520 | return to_utf8_title(tmpbuf); |
521 | } |
522 | |
523 | U32 |
864dbfa3 |
524 | Perl_to_uni_lower(pTHX_ U32 c) |
a0ed51b3 |
525 | { |
aa6ffa16 |
526 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
527 | uv_to_utf8(tmpbuf, (UV)c); |
528 | return to_utf8_lower(tmpbuf); |
529 | } |
530 | |
531 | /* for now these all assume no locale info available for Unicode > 255 */ |
532 | |
533 | bool |
864dbfa3 |
534 | Perl_is_uni_alnum_lc(pTHX_ U32 c) |
a0ed51b3 |
535 | { |
536 | return is_uni_alnum(c); /* XXX no locale support yet */ |
537 | } |
538 | |
539 | bool |
b8c5462f |
540 | Perl_is_uni_alnumc_lc(pTHX_ U32 c) |
541 | { |
542 | return is_uni_alnumc(c); /* XXX no locale support yet */ |
543 | } |
544 | |
545 | bool |
864dbfa3 |
546 | Perl_is_uni_idfirst_lc(pTHX_ U32 c) |
a0ed51b3 |
547 | { |
548 | return is_uni_idfirst(c); /* XXX no locale support yet */ |
549 | } |
550 | |
551 | bool |
864dbfa3 |
552 | Perl_is_uni_alpha_lc(pTHX_ U32 c) |
a0ed51b3 |
553 | { |
554 | return is_uni_alpha(c); /* XXX no locale support yet */ |
555 | } |
556 | |
557 | bool |
4d61ec05 |
558 | Perl_is_uni_ascii_lc(pTHX_ U32 c) |
559 | { |
560 | return is_uni_ascii(c); /* XXX no locale support yet */ |
561 | } |
562 | |
563 | bool |
864dbfa3 |
564 | Perl_is_uni_space_lc(pTHX_ U32 c) |
a0ed51b3 |
565 | { |
566 | return is_uni_space(c); /* XXX no locale support yet */ |
567 | } |
568 | |
569 | bool |
864dbfa3 |
570 | Perl_is_uni_digit_lc(pTHX_ U32 c) |
a0ed51b3 |
571 | { |
572 | return is_uni_digit(c); /* XXX no locale support yet */ |
573 | } |
574 | |
575 | bool |
864dbfa3 |
576 | Perl_is_uni_upper_lc(pTHX_ U32 c) |
a0ed51b3 |
577 | { |
578 | return is_uni_upper(c); /* XXX no locale support yet */ |
579 | } |
580 | |
581 | bool |
864dbfa3 |
582 | Perl_is_uni_lower_lc(pTHX_ U32 c) |
a0ed51b3 |
583 | { |
584 | return is_uni_lower(c); /* XXX no locale support yet */ |
585 | } |
586 | |
587 | bool |
b8c5462f |
588 | Perl_is_uni_cntrl_lc(pTHX_ U32 c) |
589 | { |
590 | return is_uni_cntrl(c); /* XXX no locale support yet */ |
591 | } |
592 | |
593 | bool |
594 | Perl_is_uni_graph_lc(pTHX_ U32 c) |
595 | { |
596 | return is_uni_graph(c); /* XXX no locale support yet */ |
597 | } |
598 | |
599 | bool |
864dbfa3 |
600 | Perl_is_uni_print_lc(pTHX_ U32 c) |
a0ed51b3 |
601 | { |
602 | return is_uni_print(c); /* XXX no locale support yet */ |
603 | } |
604 | |
b8c5462f |
605 | bool |
606 | Perl_is_uni_punct_lc(pTHX_ U32 c) |
607 | { |
608 | return is_uni_punct(c); /* XXX no locale support yet */ |
609 | } |
610 | |
4d61ec05 |
611 | bool |
612 | Perl_is_uni_xdigit_lc(pTHX_ U32 c) |
613 | { |
614 | return is_uni_xdigit(c); /* XXX no locale support yet */ |
615 | } |
616 | |
a0ed51b3 |
617 | U32 |
864dbfa3 |
618 | Perl_to_uni_upper_lc(pTHX_ U32 c) |
a0ed51b3 |
619 | { |
620 | return to_uni_upper(c); /* XXX no locale support yet */ |
621 | } |
622 | |
623 | U32 |
864dbfa3 |
624 | Perl_to_uni_title_lc(pTHX_ U32 c) |
a0ed51b3 |
625 | { |
626 | return to_uni_title(c); /* XXX no locale support yet */ |
627 | } |
628 | |
629 | U32 |
864dbfa3 |
630 | Perl_to_uni_lower_lc(pTHX_ U32 c) |
a0ed51b3 |
631 | { |
632 | return to_uni_lower(c); /* XXX no locale support yet */ |
633 | } |
634 | |
a0ed51b3 |
635 | bool |
864dbfa3 |
636 | Perl_is_utf8_alnum(pTHX_ U8 *p) |
a0ed51b3 |
637 | { |
386d01d6 |
638 | if (!is_utf8_char(p)) |
639 | return FALSE; |
a0ed51b3 |
640 | if (!PL_utf8_alnum) |
e24b16f9 |
641 | PL_utf8_alnum = swash_init("utf8", "IsAlnum", &PL_sv_undef, 0, 0); |
a0ed51b3 |
642 | return swash_fetch(PL_utf8_alnum, p); |
643 | /* return *p == '_' || is_utf8_alpha(p) || is_utf8_digit(p); */ |
644 | #ifdef SURPRISINGLY_SLOWER /* probably because alpha is usually true */ |
645 | if (!PL_utf8_alnum) |
646 | PL_utf8_alnum = swash_init("utf8", "", |
647 | sv_2mortal(newSVpv("+utf8::IsAlpha\n+utf8::IsDigit\n005F\n",0)), 0, 0); |
648 | return swash_fetch(PL_utf8_alnum, p); |
649 | #endif |
650 | } |
651 | |
652 | bool |
b8c5462f |
653 | Perl_is_utf8_alnumc(pTHX_ U8 *p) |
654 | { |
386d01d6 |
655 | if (!is_utf8_char(p)) |
656 | return FALSE; |
b8c5462f |
657 | if (!PL_utf8_alnum) |
658 | PL_utf8_alnum = swash_init("utf8", "IsAlnumC", &PL_sv_undef, 0, 0); |
659 | return swash_fetch(PL_utf8_alnum, p); |
660 | /* return is_utf8_alpha(p) || is_utf8_digit(p); */ |
661 | #ifdef SURPRISINGLY_SLOWER /* probably because alpha is usually true */ |
662 | if (!PL_utf8_alnum) |
663 | PL_utf8_alnum = swash_init("utf8", "", |
664 | sv_2mortal(newSVpv("+utf8::IsAlpha\n+utf8::IsDigit\n005F\n",0)), 0, 0); |
665 | return swash_fetch(PL_utf8_alnum, p); |
666 | #endif |
667 | } |
668 | |
669 | bool |
864dbfa3 |
670 | Perl_is_utf8_idfirst(pTHX_ U8 *p) |
a0ed51b3 |
671 | { |
672 | return *p == '_' || is_utf8_alpha(p); |
673 | } |
674 | |
675 | bool |
864dbfa3 |
676 | Perl_is_utf8_alpha(pTHX_ U8 *p) |
a0ed51b3 |
677 | { |
386d01d6 |
678 | if (!is_utf8_char(p)) |
679 | return FALSE; |
a0ed51b3 |
680 | if (!PL_utf8_alpha) |
e24b16f9 |
681 | PL_utf8_alpha = swash_init("utf8", "IsAlpha", &PL_sv_undef, 0, 0); |
a0ed51b3 |
682 | return swash_fetch(PL_utf8_alpha, p); |
683 | } |
684 | |
685 | bool |
b8c5462f |
686 | Perl_is_utf8_ascii(pTHX_ U8 *p) |
687 | { |
386d01d6 |
688 | if (!is_utf8_char(p)) |
689 | return FALSE; |
b8c5462f |
690 | if (!PL_utf8_ascii) |
691 | PL_utf8_ascii = swash_init("utf8", "IsAscii", &PL_sv_undef, 0, 0); |
692 | return swash_fetch(PL_utf8_ascii, p); |
693 | } |
694 | |
695 | bool |
864dbfa3 |
696 | Perl_is_utf8_space(pTHX_ U8 *p) |
a0ed51b3 |
697 | { |
386d01d6 |
698 | if (!is_utf8_char(p)) |
699 | return FALSE; |
a0ed51b3 |
700 | if (!PL_utf8_space) |
e24b16f9 |
701 | PL_utf8_space = swash_init("utf8", "IsSpace", &PL_sv_undef, 0, 0); |
a0ed51b3 |
702 | return swash_fetch(PL_utf8_space, p); |
703 | } |
704 | |
705 | bool |
864dbfa3 |
706 | Perl_is_utf8_digit(pTHX_ U8 *p) |
a0ed51b3 |
707 | { |
386d01d6 |
708 | if (!is_utf8_char(p)) |
709 | return FALSE; |
a0ed51b3 |
710 | if (!PL_utf8_digit) |
e24b16f9 |
711 | PL_utf8_digit = swash_init("utf8", "IsDigit", &PL_sv_undef, 0, 0); |
a0ed51b3 |
712 | return swash_fetch(PL_utf8_digit, p); |
713 | } |
714 | |
715 | bool |
864dbfa3 |
716 | Perl_is_utf8_upper(pTHX_ U8 *p) |
a0ed51b3 |
717 | { |
386d01d6 |
718 | if (!is_utf8_char(p)) |
719 | return FALSE; |
a0ed51b3 |
720 | if (!PL_utf8_upper) |
e24b16f9 |
721 | PL_utf8_upper = swash_init("utf8", "IsUpper", &PL_sv_undef, 0, 0); |
a0ed51b3 |
722 | return swash_fetch(PL_utf8_upper, p); |
723 | } |
724 | |
725 | bool |
864dbfa3 |
726 | Perl_is_utf8_lower(pTHX_ U8 *p) |
a0ed51b3 |
727 | { |
386d01d6 |
728 | if (!is_utf8_char(p)) |
729 | return FALSE; |
a0ed51b3 |
730 | if (!PL_utf8_lower) |
e24b16f9 |
731 | PL_utf8_lower = swash_init("utf8", "IsLower", &PL_sv_undef, 0, 0); |
a0ed51b3 |
732 | return swash_fetch(PL_utf8_lower, p); |
733 | } |
734 | |
735 | bool |
b8c5462f |
736 | Perl_is_utf8_cntrl(pTHX_ U8 *p) |
737 | { |
386d01d6 |
738 | if (!is_utf8_char(p)) |
739 | return FALSE; |
b8c5462f |
740 | if (!PL_utf8_cntrl) |
741 | PL_utf8_cntrl = swash_init("utf8", "IsCntrl", &PL_sv_undef, 0, 0); |
742 | return swash_fetch(PL_utf8_cntrl, p); |
743 | } |
744 | |
745 | bool |
746 | Perl_is_utf8_graph(pTHX_ U8 *p) |
747 | { |
386d01d6 |
748 | if (!is_utf8_char(p)) |
749 | return FALSE; |
b8c5462f |
750 | if (!PL_utf8_graph) |
751 | PL_utf8_graph = swash_init("utf8", "IsGraph", &PL_sv_undef, 0, 0); |
752 | return swash_fetch(PL_utf8_graph, p); |
753 | } |
754 | |
755 | bool |
864dbfa3 |
756 | Perl_is_utf8_print(pTHX_ U8 *p) |
a0ed51b3 |
757 | { |
386d01d6 |
758 | if (!is_utf8_char(p)) |
759 | return FALSE; |
a0ed51b3 |
760 | if (!PL_utf8_print) |
e24b16f9 |
761 | PL_utf8_print = swash_init("utf8", "IsPrint", &PL_sv_undef, 0, 0); |
a0ed51b3 |
762 | return swash_fetch(PL_utf8_print, p); |
763 | } |
764 | |
765 | bool |
b8c5462f |
766 | Perl_is_utf8_punct(pTHX_ U8 *p) |
767 | { |
386d01d6 |
768 | if (!is_utf8_char(p)) |
769 | return FALSE; |
b8c5462f |
770 | if (!PL_utf8_punct) |
771 | PL_utf8_punct = swash_init("utf8", "IsPunct", &PL_sv_undef, 0, 0); |
772 | return swash_fetch(PL_utf8_punct, p); |
773 | } |
774 | |
775 | bool |
776 | Perl_is_utf8_xdigit(pTHX_ U8 *p) |
777 | { |
386d01d6 |
778 | if (!is_utf8_char(p)) |
779 | return FALSE; |
b8c5462f |
780 | if (!PL_utf8_xdigit) |
781 | PL_utf8_xdigit = swash_init("utf8", "IsXDigit", &PL_sv_undef, 0, 0); |
782 | return swash_fetch(PL_utf8_xdigit, p); |
783 | } |
784 | |
785 | bool |
864dbfa3 |
786 | Perl_is_utf8_mark(pTHX_ U8 *p) |
a0ed51b3 |
787 | { |
386d01d6 |
788 | if (!is_utf8_char(p)) |
789 | return FALSE; |
a0ed51b3 |
790 | if (!PL_utf8_mark) |
e24b16f9 |
791 | PL_utf8_mark = swash_init("utf8", "IsM", &PL_sv_undef, 0, 0); |
a0ed51b3 |
792 | return swash_fetch(PL_utf8_mark, p); |
793 | } |
794 | |
2104c8d9 |
795 | UV |
864dbfa3 |
796 | Perl_to_utf8_upper(pTHX_ U8 *p) |
a0ed51b3 |
797 | { |
798 | UV uv; |
799 | |
800 | if (!PL_utf8_toupper) |
e24b16f9 |
801 | PL_utf8_toupper = swash_init("utf8", "ToUpper", &PL_sv_undef, 4, 0); |
a0ed51b3 |
802 | uv = swash_fetch(PL_utf8_toupper, p); |
803 | return uv ? uv : utf8_to_uv(p,0); |
804 | } |
805 | |
2104c8d9 |
806 | UV |
864dbfa3 |
807 | Perl_to_utf8_title(pTHX_ U8 *p) |
a0ed51b3 |
808 | { |
809 | UV uv; |
810 | |
811 | if (!PL_utf8_totitle) |
e24b16f9 |
812 | PL_utf8_totitle = swash_init("utf8", "ToTitle", &PL_sv_undef, 4, 0); |
a0ed51b3 |
813 | uv = swash_fetch(PL_utf8_totitle, p); |
814 | return uv ? uv : utf8_to_uv(p,0); |
815 | } |
816 | |
2104c8d9 |
817 | UV |
864dbfa3 |
818 | Perl_to_utf8_lower(pTHX_ U8 *p) |
a0ed51b3 |
819 | { |
820 | UV uv; |
821 | |
822 | if (!PL_utf8_tolower) |
e24b16f9 |
823 | PL_utf8_tolower = swash_init("utf8", "ToLower", &PL_sv_undef, 4, 0); |
a0ed51b3 |
824 | uv = swash_fetch(PL_utf8_tolower, p); |
825 | return uv ? uv : utf8_to_uv(p,0); |
826 | } |
827 | |
828 | /* a "swash" is a swatch hash */ |
829 | |
830 | SV* |
864dbfa3 |
831 | Perl_swash_init(pTHX_ char* pkg, char* name, SV *listsv, I32 minbits, I32 none) |
a0ed51b3 |
832 | { |
833 | SV* retval; |
834 | char tmpbuf[256]; |
835 | dSP; |
ce3b816e |
836 | |
837 | if (!gv_stashpv(pkg, 0)) { /* demand load utf8 */ |
838 | ENTER; |
839 | Perl_load_module(aTHX_ PERL_LOADMOD_NOIMPORT, newSVpv(pkg,0), Nullsv); |
840 | LEAVE; |
841 | } |
842 | SPAGAIN; |
a0ed51b3 |
843 | PUSHSTACKi(PERLSI_MAGIC); |
844 | PUSHMARK(SP); |
845 | EXTEND(SP,5); |
846 | PUSHs(sv_2mortal(newSVpvn(pkg, strlen(pkg)))); |
847 | PUSHs(sv_2mortal(newSVpvn(name, strlen(name)))); |
848 | PUSHs(listsv); |
849 | PUSHs(sv_2mortal(newSViv(minbits))); |
850 | PUSHs(sv_2mortal(newSViv(none))); |
851 | PUTBACK; |
852 | ENTER; |
853 | SAVEI32(PL_hints); |
854 | PL_hints = 0; |
855 | save_re_context(); |
e24b16f9 |
856 | if (PL_curcop == &PL_compiling) /* XXX ought to be handled by lex_start */ |
a0ed51b3 |
857 | strncpy(tmpbuf, PL_tokenbuf, sizeof tmpbuf); |
864dbfa3 |
858 | if (call_method("SWASHNEW", G_SCALAR)) |
e24b16f9 |
859 | retval = newSVsv(*PL_stack_sp--); |
a0ed51b3 |
860 | else |
e24b16f9 |
861 | retval = &PL_sv_undef; |
a0ed51b3 |
862 | LEAVE; |
863 | POPSTACK; |
e24b16f9 |
864 | if (PL_curcop == &PL_compiling) { |
a0ed51b3 |
865 | strncpy(PL_tokenbuf, tmpbuf, sizeof tmpbuf); |
e24b16f9 |
866 | PL_curcop->op_private = PL_hints; |
a0ed51b3 |
867 | } |
868 | if (!SvROK(retval) || SvTYPE(SvRV(retval)) != SVt_PVHV) |
cea2e8a9 |
869 | Perl_croak(aTHX_ "SWASHNEW didn't return an HV ref"); |
a0ed51b3 |
870 | return retval; |
871 | } |
872 | |
873 | UV |
864dbfa3 |
874 | Perl_swash_fetch(pTHX_ SV *sv, U8 *ptr) |
a0ed51b3 |
875 | { |
876 | HV* hv = (HV*)SvRV(sv); |
877 | U32 klen = UTF8SKIP(ptr) - 1; |
878 | U32 off = ptr[klen] & 127; /* NB: 64 bit always 0 when len > 1 */ |
879 | STRLEN slen; |
880 | STRLEN needents = (klen ? 64 : 128); |
dfe13c55 |
881 | U8 *tmps; |
a0ed51b3 |
882 | U32 bit; |
883 | SV *retval; |
884 | |
885 | /* |
886 | * This single-entry cache saves about 1/3 of the utf8 overhead in test |
887 | * suite. (That is, only 7-8% overall over just a hash cache. Still, |
888 | * it's nothing to sniff at.) Pity we usually come through at least |
889 | * two function calls to get here... |
890 | * |
891 | * NB: this code assumes that swatches are never modified, once generated! |
892 | */ |
893 | |
894 | if (hv == PL_last_swash_hv && |
895 | klen == PL_last_swash_klen && |
12ae5dfc |
896 | (!klen || memEQ((char *)ptr,(char *)PL_last_swash_key,klen)) ) |
a0ed51b3 |
897 | { |
898 | tmps = PL_last_swash_tmps; |
899 | slen = PL_last_swash_slen; |
900 | } |
901 | else { |
902 | /* Try our second-level swatch cache, kept in a hash. */ |
dfe13c55 |
903 | SV** svp = hv_fetch(hv, (char*)ptr, klen, FALSE); |
a0ed51b3 |
904 | |
905 | /* If not cached, generate it via utf8::SWASHGET */ |
dfe13c55 |
906 | if (!svp || !SvPOK(*svp) || !(tmps = (U8*)SvPV(*svp, slen))) { |
a0ed51b3 |
907 | dSP; |
908 | ENTER; |
909 | SAVETMPS; |
910 | save_re_context(); |
911 | PUSHSTACKi(PERLSI_MAGIC); |
912 | PUSHMARK(SP); |
913 | EXTEND(SP,3); |
914 | PUSHs((SV*)sv); |
915 | PUSHs(sv_2mortal(newSViv(utf8_to_uv(ptr, 0) & ~(needents - 1)))); |
916 | PUSHs(sv_2mortal(newSViv(needents))); |
917 | PUTBACK; |
864dbfa3 |
918 | if (call_method("SWASHGET", G_SCALAR)) |
e24b16f9 |
919 | retval = newSVsv(*PL_stack_sp--); |
a0ed51b3 |
920 | else |
e24b16f9 |
921 | retval = &PL_sv_undef; |
a0ed51b3 |
922 | POPSTACK; |
923 | FREETMPS; |
924 | LEAVE; |
e24b16f9 |
925 | if (PL_curcop == &PL_compiling) |
926 | PL_curcop->op_private = PL_hints; |
a0ed51b3 |
927 | |
dfe13c55 |
928 | svp = hv_store(hv, (char*)ptr, klen, retval, 0); |
a0ed51b3 |
929 | |
dfe13c55 |
930 | if (!svp || !(tmps = (U8*)SvPV(*svp, slen)) || slen < 8) |
cea2e8a9 |
931 | Perl_croak(aTHX_ "SWASHGET didn't return result of proper length"); |
a0ed51b3 |
932 | } |
933 | |
934 | PL_last_swash_hv = hv; |
935 | PL_last_swash_klen = klen; |
936 | PL_last_swash_tmps = tmps; |
937 | PL_last_swash_slen = slen; |
938 | if (klen) |
939 | Copy(ptr, PL_last_swash_key, klen, U8); |
940 | } |
941 | |
942 | switch ((slen << 3) / needents) { |
943 | case 1: |
944 | bit = 1 << (off & 7); |
945 | off >>= 3; |
946 | return (tmps[off] & bit) != 0; |
947 | case 8: |
948 | return tmps[off]; |
949 | case 16: |
950 | off <<= 1; |
951 | return (tmps[off] << 8) + tmps[off + 1] ; |
952 | case 32: |
953 | off <<= 2; |
954 | return (tmps[off] << 24) + (tmps[off+1] << 16) + (tmps[off+2] << 8) + tmps[off + 3] ; |
955 | } |
cea2e8a9 |
956 | Perl_croak(aTHX_ "panic: swash_fetch"); |
a0ed51b3 |
957 | return 0; |
958 | } |