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 | |
a0ed51b3 |
137 | UV |
864dbfa3 |
138 | Perl_utf8_to_uv(pTHX_ U8* s, I32* retlen) |
a0ed51b3 |
139 | { |
140 | UV uv = *s; |
141 | int len; |
142 | if (!(uv & 0x80)) { |
143 | if (retlen) |
144 | *retlen = 1; |
145 | return *s; |
146 | } |
147 | if (!(uv & 0x40)) { |
0453d815 |
148 | dTHR; |
149 | if (ckWARN_d(WARN_UTF8)) |
150 | Perl_warner(aTHX_ WARN_UTF8, "Malformed UTF-8 character"); |
a0ed51b3 |
151 | if (retlen) |
152 | *retlen = 1; |
153 | return *s; |
154 | } |
155 | |
156 | if (!(uv & 0x20)) { len = 2; uv &= 0x1f; } |
157 | else if (!(uv & 0x10)) { len = 3; uv &= 0x0f; } |
158 | else if (!(uv & 0x08)) { len = 4; uv &= 0x07; } |
159 | else if (!(uv & 0x04)) { len = 5; uv &= 0x03; } |
160 | else if (!(uv & 0x02)) { len = 6; uv &= 0x01; } |
3c77ea2b |
161 | else if (!(uv & 0x01)) { len = 7; uv = 0; } |
162 | else { len = 13; uv = 0; } /* whoa! */ |
a0ed51b3 |
163 | |
164 | if (retlen) |
165 | *retlen = len; |
166 | --len; |
167 | s++; |
168 | while (len--) { |
169 | if ((*s & 0xc0) != 0x80) { |
0453d815 |
170 | dTHR; |
171 | if (ckWARN_d(WARN_UTF8)) |
172 | Perl_warner(aTHX_ WARN_UTF8, "Malformed UTF-8 character"); |
a0ed51b3 |
173 | if (retlen) |
174 | *retlen -= len + 1; |
175 | return 0xfffd; |
176 | } |
177 | else |
178 | uv = (uv << 6) | (*s++ & 0x3f); |
179 | } |
180 | return uv; |
181 | } |
182 | |
183 | /* utf8_distance(a,b) is intended to be a - b in pointer arithmetic */ |
184 | |
185 | I32 |
864dbfa3 |
186 | Perl_utf8_distance(pTHX_ U8 *a, U8 *b) |
a0ed51b3 |
187 | { |
188 | I32 off = 0; |
189 | if (a < b) { |
190 | while (a < b) { |
191 | a += UTF8SKIP(a); |
192 | off--; |
193 | } |
194 | } |
195 | else { |
196 | while (b < a) { |
197 | b += UTF8SKIP(b); |
198 | off++; |
199 | } |
200 | } |
201 | return off; |
202 | } |
203 | |
204 | /* WARNING: do not use the following unless you *know* off is within bounds */ |
205 | |
206 | U8 * |
864dbfa3 |
207 | Perl_utf8_hop(pTHX_ U8 *s, I32 off) |
a0ed51b3 |
208 | { |
209 | if (off >= 0) { |
210 | while (off--) |
211 | s += UTF8SKIP(s); |
212 | } |
213 | else { |
214 | while (off++) { |
215 | s--; |
216 | if (*s & 0x80) { |
217 | while ((*s & 0xc0) == 0x80) |
218 | s--; |
219 | } |
220 | } |
221 | } |
222 | return s; |
223 | } |
224 | |
84393cd9 |
225 | /* |
732bce8c |
226 | =for apidoc Am|U8 *|utf8_to_bytes|U8 *s|STRLEN len |
84393cd9 |
227 | |
228 | Converts a string C<s> of length C<len> from UTF8 into ASCII encoding. |
229 | Unlike C<bytes_to_utf8>, this over-writes the original string. |
1e72252a |
230 | Returns zero on failure after converting as much as possible. |
84393cd9 |
231 | |
232 | =cut |
233 | */ |
234 | |
235 | U8 * |
236 | Perl_utf8_to_bytes(pTHX_ U8* s, STRLEN len) |
237 | { |
238 | dTHR; |
239 | U8 *send; |
240 | U8 *d; |
241 | U8 *save; |
242 | |
243 | send = s + len; |
244 | d = save = s; |
245 | while (s < send) { |
246 | if (*s < 0x80) |
247 | *d++ = *s++; |
248 | else { |
249 | I32 ulen; |
250 | UV uv = utf8_to_uv(s, &ulen); |
1e72252a |
251 | if (uv > 255) { |
252 | *d = '\0'; |
253 | return 0; |
254 | } |
84393cd9 |
255 | s += ulen; |
256 | *d++ = (U8)uv; |
257 | } |
258 | } |
259 | *d = '\0'; |
260 | return save; |
261 | } |
262 | |
263 | /* |
1e72252a |
264 | =for apidoc Am|U8 *|bytes_to_utf8|U8 *s|STRLEN *len |
84393cd9 |
265 | |
266 | Converts a string C<s> of length C<len> from ASCII into UTF8 encoding. |
1e72252a |
267 | Returns a pointer to the newly-created string, and sets C<len> to |
268 | reflect the new length. |
84393cd9 |
269 | |
732bce8c |
270 | =cut |
84393cd9 |
271 | */ |
272 | |
273 | U8* |
1e72252a |
274 | Perl_bytes_to_utf8(pTHX_ U8* s, STRLEN *len) |
84393cd9 |
275 | { |
276 | dTHR; |
277 | U8 *send; |
278 | U8 *d; |
279 | U8 *dst; |
1e72252a |
280 | send = s + (*len); |
84393cd9 |
281 | |
1e72252a |
282 | Newz(801, d, (*len) * 2 + 1, U8); |
84393cd9 |
283 | dst = d; |
284 | |
285 | while (s < send) { |
286 | if (*s < 0x80) |
287 | *d++ = *s++; |
288 | else { |
289 | UV uv = *s++; |
290 | *d++ = (( uv >> 6) | 0xc0); |
291 | *d++ = (( uv & 0x3f) | 0x80); |
292 | } |
293 | } |
294 | *d = '\0'; |
1e72252a |
295 | *len = d-dst; |
84393cd9 |
296 | return dst; |
297 | } |
298 | |
a0ed51b3 |
299 | /* XXX NOTHING CALLS THE FOLLOWING TWO ROUTINES YET!!! */ |
300 | /* |
301 | * Convert native or reversed UTF-16 to UTF-8. |
302 | * |
303 | * Destination must be pre-extended to 3/2 source. Do not use in-place. |
304 | * We optimize for native, for obvious reasons. */ |
305 | |
306 | U8* |
864dbfa3 |
307 | Perl_utf16_to_utf8(pTHX_ U16* p, U8* d, I32 bytelen) |
a0ed51b3 |
308 | { |
309 | U16* pend = p + bytelen / 2; |
310 | while (p < pend) { |
311 | UV uv = *p++; |
312 | if (uv < 0x80) { |
313 | *d++ = uv; |
314 | continue; |
315 | } |
316 | if (uv < 0x800) { |
317 | *d++ = (( uv >> 6) | 0xc0); |
318 | *d++ = (( uv & 0x3f) | 0x80); |
319 | continue; |
320 | } |
321 | if (uv >= 0xd800 && uv < 0xdbff) { /* surrogates */ |
0453d815 |
322 | dTHR; |
a0ed51b3 |
323 | int low = *p++; |
324 | if (low < 0xdc00 || low >= 0xdfff) { |
0453d815 |
325 | if (ckWARN_d(WARN_UTF8)) |
326 | Perl_warner(aTHX_ WARN_UTF8, "Malformed UTF-16 surrogate"); |
a0ed51b3 |
327 | p--; |
328 | uv = 0xfffd; |
329 | } |
330 | uv = ((uv - 0xd800) << 10) + (low - 0xdc00) + 0x10000; |
331 | } |
332 | if (uv < 0x10000) { |
333 | *d++ = (( uv >> 12) | 0xe0); |
334 | *d++ = (((uv >> 6) & 0x3f) | 0x80); |
335 | *d++ = (( uv & 0x3f) | 0x80); |
336 | continue; |
337 | } |
338 | else { |
339 | *d++ = (( uv >> 18) | 0xf0); |
340 | *d++ = (((uv >> 12) & 0x3f) | 0x80); |
341 | *d++ = (((uv >> 6) & 0x3f) | 0x80); |
342 | *d++ = (( uv & 0x3f) | 0x80); |
343 | continue; |
344 | } |
345 | } |
346 | return d; |
347 | } |
348 | |
349 | /* Note: this one is slightly destructive of the source. */ |
350 | |
351 | U8* |
864dbfa3 |
352 | Perl_utf16_to_utf8_reversed(pTHX_ U16* p, U8* d, I32 bytelen) |
a0ed51b3 |
353 | { |
354 | U8* s = (U8*)p; |
355 | U8* send = s + bytelen; |
356 | while (s < send) { |
357 | U8 tmp = s[0]; |
358 | s[0] = s[1]; |
359 | s[1] = tmp; |
360 | s += 2; |
361 | } |
362 | return utf16_to_utf8(p, d, bytelen); |
363 | } |
364 | |
365 | /* for now these are all defined (inefficiently) in terms of the utf8 versions */ |
366 | |
367 | bool |
864dbfa3 |
368 | Perl_is_uni_alnum(pTHX_ U32 c) |
a0ed51b3 |
369 | { |
aa6ffa16 |
370 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
371 | uv_to_utf8(tmpbuf, (UV)c); |
372 | return is_utf8_alnum(tmpbuf); |
373 | } |
374 | |
375 | bool |
b8c5462f |
376 | Perl_is_uni_alnumc(pTHX_ U32 c) |
377 | { |
aa6ffa16 |
378 | U8 tmpbuf[UTF8_MAXLEN]; |
b8c5462f |
379 | uv_to_utf8(tmpbuf, (UV)c); |
380 | return is_utf8_alnumc(tmpbuf); |
381 | } |
382 | |
383 | bool |
864dbfa3 |
384 | Perl_is_uni_idfirst(pTHX_ U32 c) |
a0ed51b3 |
385 | { |
aa6ffa16 |
386 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
387 | uv_to_utf8(tmpbuf, (UV)c); |
388 | return is_utf8_idfirst(tmpbuf); |
389 | } |
390 | |
391 | bool |
864dbfa3 |
392 | Perl_is_uni_alpha(pTHX_ U32 c) |
a0ed51b3 |
393 | { |
aa6ffa16 |
394 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
395 | uv_to_utf8(tmpbuf, (UV)c); |
396 | return is_utf8_alpha(tmpbuf); |
397 | } |
398 | |
399 | bool |
4d61ec05 |
400 | Perl_is_uni_ascii(pTHX_ U32 c) |
401 | { |
aa6ffa16 |
402 | U8 tmpbuf[UTF8_MAXLEN]; |
4d61ec05 |
403 | uv_to_utf8(tmpbuf, (UV)c); |
404 | return is_utf8_ascii(tmpbuf); |
405 | } |
406 | |
407 | bool |
864dbfa3 |
408 | Perl_is_uni_space(pTHX_ U32 c) |
a0ed51b3 |
409 | { |
aa6ffa16 |
410 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
411 | uv_to_utf8(tmpbuf, (UV)c); |
412 | return is_utf8_space(tmpbuf); |
413 | } |
414 | |
415 | bool |
864dbfa3 |
416 | Perl_is_uni_digit(pTHX_ U32 c) |
a0ed51b3 |
417 | { |
aa6ffa16 |
418 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
419 | uv_to_utf8(tmpbuf, (UV)c); |
420 | return is_utf8_digit(tmpbuf); |
421 | } |
422 | |
423 | bool |
864dbfa3 |
424 | Perl_is_uni_upper(pTHX_ U32 c) |
a0ed51b3 |
425 | { |
aa6ffa16 |
426 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
427 | uv_to_utf8(tmpbuf, (UV)c); |
428 | return is_utf8_upper(tmpbuf); |
429 | } |
430 | |
431 | bool |
864dbfa3 |
432 | Perl_is_uni_lower(pTHX_ U32 c) |
a0ed51b3 |
433 | { |
aa6ffa16 |
434 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
435 | uv_to_utf8(tmpbuf, (UV)c); |
436 | return is_utf8_lower(tmpbuf); |
437 | } |
438 | |
439 | bool |
b8c5462f |
440 | Perl_is_uni_cntrl(pTHX_ U32 c) |
441 | { |
aa6ffa16 |
442 | U8 tmpbuf[UTF8_MAXLEN]; |
b8c5462f |
443 | uv_to_utf8(tmpbuf, (UV)c); |
444 | return is_utf8_cntrl(tmpbuf); |
445 | } |
446 | |
447 | bool |
448 | Perl_is_uni_graph(pTHX_ U32 c) |
449 | { |
aa6ffa16 |
450 | U8 tmpbuf[UTF8_MAXLEN]; |
b8c5462f |
451 | uv_to_utf8(tmpbuf, (UV)c); |
452 | return is_utf8_graph(tmpbuf); |
453 | } |
454 | |
455 | bool |
864dbfa3 |
456 | Perl_is_uni_print(pTHX_ U32 c) |
a0ed51b3 |
457 | { |
aa6ffa16 |
458 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
459 | uv_to_utf8(tmpbuf, (UV)c); |
460 | return is_utf8_print(tmpbuf); |
461 | } |
462 | |
b8c5462f |
463 | bool |
f248d071 |
464 | Perl_is_uni_punct(pTHX_ U32 c) |
b8c5462f |
465 | { |
aa6ffa16 |
466 | U8 tmpbuf[UTF8_MAXLEN]; |
b8c5462f |
467 | uv_to_utf8(tmpbuf, (UV)c); |
468 | return is_utf8_punct(tmpbuf); |
469 | } |
470 | |
4d61ec05 |
471 | bool |
472 | Perl_is_uni_xdigit(pTHX_ U32 c) |
473 | { |
aa6ffa16 |
474 | U8 tmpbuf[UTF8_MAXLEN]; |
4d61ec05 |
475 | uv_to_utf8(tmpbuf, (UV)c); |
476 | return is_utf8_xdigit(tmpbuf); |
477 | } |
478 | |
a0ed51b3 |
479 | U32 |
864dbfa3 |
480 | Perl_to_uni_upper(pTHX_ U32 c) |
a0ed51b3 |
481 | { |
aa6ffa16 |
482 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
483 | uv_to_utf8(tmpbuf, (UV)c); |
484 | return to_utf8_upper(tmpbuf); |
485 | } |
486 | |
487 | U32 |
864dbfa3 |
488 | Perl_to_uni_title(pTHX_ U32 c) |
a0ed51b3 |
489 | { |
aa6ffa16 |
490 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
491 | uv_to_utf8(tmpbuf, (UV)c); |
492 | return to_utf8_title(tmpbuf); |
493 | } |
494 | |
495 | U32 |
864dbfa3 |
496 | Perl_to_uni_lower(pTHX_ U32 c) |
a0ed51b3 |
497 | { |
aa6ffa16 |
498 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
499 | uv_to_utf8(tmpbuf, (UV)c); |
500 | return to_utf8_lower(tmpbuf); |
501 | } |
502 | |
503 | /* for now these all assume no locale info available for Unicode > 255 */ |
504 | |
505 | bool |
864dbfa3 |
506 | Perl_is_uni_alnum_lc(pTHX_ U32 c) |
a0ed51b3 |
507 | { |
508 | return is_uni_alnum(c); /* XXX no locale support yet */ |
509 | } |
510 | |
511 | bool |
b8c5462f |
512 | Perl_is_uni_alnumc_lc(pTHX_ U32 c) |
513 | { |
514 | return is_uni_alnumc(c); /* XXX no locale support yet */ |
515 | } |
516 | |
517 | bool |
864dbfa3 |
518 | Perl_is_uni_idfirst_lc(pTHX_ U32 c) |
a0ed51b3 |
519 | { |
520 | return is_uni_idfirst(c); /* XXX no locale support yet */ |
521 | } |
522 | |
523 | bool |
864dbfa3 |
524 | Perl_is_uni_alpha_lc(pTHX_ U32 c) |
a0ed51b3 |
525 | { |
526 | return is_uni_alpha(c); /* XXX no locale support yet */ |
527 | } |
528 | |
529 | bool |
4d61ec05 |
530 | Perl_is_uni_ascii_lc(pTHX_ U32 c) |
531 | { |
532 | return is_uni_ascii(c); /* XXX no locale support yet */ |
533 | } |
534 | |
535 | bool |
864dbfa3 |
536 | Perl_is_uni_space_lc(pTHX_ U32 c) |
a0ed51b3 |
537 | { |
538 | return is_uni_space(c); /* XXX no locale support yet */ |
539 | } |
540 | |
541 | bool |
864dbfa3 |
542 | Perl_is_uni_digit_lc(pTHX_ U32 c) |
a0ed51b3 |
543 | { |
544 | return is_uni_digit(c); /* XXX no locale support yet */ |
545 | } |
546 | |
547 | bool |
864dbfa3 |
548 | Perl_is_uni_upper_lc(pTHX_ U32 c) |
a0ed51b3 |
549 | { |
550 | return is_uni_upper(c); /* XXX no locale support yet */ |
551 | } |
552 | |
553 | bool |
864dbfa3 |
554 | Perl_is_uni_lower_lc(pTHX_ U32 c) |
a0ed51b3 |
555 | { |
556 | return is_uni_lower(c); /* XXX no locale support yet */ |
557 | } |
558 | |
559 | bool |
b8c5462f |
560 | Perl_is_uni_cntrl_lc(pTHX_ U32 c) |
561 | { |
562 | return is_uni_cntrl(c); /* XXX no locale support yet */ |
563 | } |
564 | |
565 | bool |
566 | Perl_is_uni_graph_lc(pTHX_ U32 c) |
567 | { |
568 | return is_uni_graph(c); /* XXX no locale support yet */ |
569 | } |
570 | |
571 | bool |
864dbfa3 |
572 | Perl_is_uni_print_lc(pTHX_ U32 c) |
a0ed51b3 |
573 | { |
574 | return is_uni_print(c); /* XXX no locale support yet */ |
575 | } |
576 | |
b8c5462f |
577 | bool |
578 | Perl_is_uni_punct_lc(pTHX_ U32 c) |
579 | { |
580 | return is_uni_punct(c); /* XXX no locale support yet */ |
581 | } |
582 | |
4d61ec05 |
583 | bool |
584 | Perl_is_uni_xdigit_lc(pTHX_ U32 c) |
585 | { |
586 | return is_uni_xdigit(c); /* XXX no locale support yet */ |
587 | } |
588 | |
a0ed51b3 |
589 | U32 |
864dbfa3 |
590 | Perl_to_uni_upper_lc(pTHX_ U32 c) |
a0ed51b3 |
591 | { |
592 | return to_uni_upper(c); /* XXX no locale support yet */ |
593 | } |
594 | |
595 | U32 |
864dbfa3 |
596 | Perl_to_uni_title_lc(pTHX_ U32 c) |
a0ed51b3 |
597 | { |
598 | return to_uni_title(c); /* XXX no locale support yet */ |
599 | } |
600 | |
601 | U32 |
864dbfa3 |
602 | Perl_to_uni_lower_lc(pTHX_ U32 c) |
a0ed51b3 |
603 | { |
604 | return to_uni_lower(c); /* XXX no locale support yet */ |
605 | } |
606 | |
a0ed51b3 |
607 | bool |
864dbfa3 |
608 | Perl_is_utf8_alnum(pTHX_ U8 *p) |
a0ed51b3 |
609 | { |
386d01d6 |
610 | if (!is_utf8_char(p)) |
611 | return FALSE; |
a0ed51b3 |
612 | if (!PL_utf8_alnum) |
e24b16f9 |
613 | PL_utf8_alnum = swash_init("utf8", "IsAlnum", &PL_sv_undef, 0, 0); |
a0ed51b3 |
614 | return swash_fetch(PL_utf8_alnum, p); |
615 | /* return *p == '_' || is_utf8_alpha(p) || is_utf8_digit(p); */ |
616 | #ifdef SURPRISINGLY_SLOWER /* probably because alpha is usually true */ |
617 | if (!PL_utf8_alnum) |
618 | PL_utf8_alnum = swash_init("utf8", "", |
619 | sv_2mortal(newSVpv("+utf8::IsAlpha\n+utf8::IsDigit\n005F\n",0)), 0, 0); |
620 | return swash_fetch(PL_utf8_alnum, p); |
621 | #endif |
622 | } |
623 | |
624 | bool |
b8c5462f |
625 | Perl_is_utf8_alnumc(pTHX_ U8 *p) |
626 | { |
386d01d6 |
627 | if (!is_utf8_char(p)) |
628 | return FALSE; |
b8c5462f |
629 | if (!PL_utf8_alnum) |
630 | PL_utf8_alnum = swash_init("utf8", "IsAlnumC", &PL_sv_undef, 0, 0); |
631 | return swash_fetch(PL_utf8_alnum, p); |
632 | /* return is_utf8_alpha(p) || is_utf8_digit(p); */ |
633 | #ifdef SURPRISINGLY_SLOWER /* probably because alpha is usually true */ |
634 | if (!PL_utf8_alnum) |
635 | PL_utf8_alnum = swash_init("utf8", "", |
636 | sv_2mortal(newSVpv("+utf8::IsAlpha\n+utf8::IsDigit\n005F\n",0)), 0, 0); |
637 | return swash_fetch(PL_utf8_alnum, p); |
638 | #endif |
639 | } |
640 | |
641 | bool |
864dbfa3 |
642 | Perl_is_utf8_idfirst(pTHX_ U8 *p) |
a0ed51b3 |
643 | { |
644 | return *p == '_' || is_utf8_alpha(p); |
645 | } |
646 | |
647 | bool |
864dbfa3 |
648 | Perl_is_utf8_alpha(pTHX_ U8 *p) |
a0ed51b3 |
649 | { |
386d01d6 |
650 | if (!is_utf8_char(p)) |
651 | return FALSE; |
a0ed51b3 |
652 | if (!PL_utf8_alpha) |
e24b16f9 |
653 | PL_utf8_alpha = swash_init("utf8", "IsAlpha", &PL_sv_undef, 0, 0); |
a0ed51b3 |
654 | return swash_fetch(PL_utf8_alpha, p); |
655 | } |
656 | |
657 | bool |
b8c5462f |
658 | Perl_is_utf8_ascii(pTHX_ U8 *p) |
659 | { |
386d01d6 |
660 | if (!is_utf8_char(p)) |
661 | return FALSE; |
b8c5462f |
662 | if (!PL_utf8_ascii) |
663 | PL_utf8_ascii = swash_init("utf8", "IsAscii", &PL_sv_undef, 0, 0); |
664 | return swash_fetch(PL_utf8_ascii, p); |
665 | } |
666 | |
667 | bool |
864dbfa3 |
668 | Perl_is_utf8_space(pTHX_ U8 *p) |
a0ed51b3 |
669 | { |
386d01d6 |
670 | if (!is_utf8_char(p)) |
671 | return FALSE; |
a0ed51b3 |
672 | if (!PL_utf8_space) |
e24b16f9 |
673 | PL_utf8_space = swash_init("utf8", "IsSpace", &PL_sv_undef, 0, 0); |
a0ed51b3 |
674 | return swash_fetch(PL_utf8_space, p); |
675 | } |
676 | |
677 | bool |
864dbfa3 |
678 | Perl_is_utf8_digit(pTHX_ U8 *p) |
a0ed51b3 |
679 | { |
386d01d6 |
680 | if (!is_utf8_char(p)) |
681 | return FALSE; |
a0ed51b3 |
682 | if (!PL_utf8_digit) |
e24b16f9 |
683 | PL_utf8_digit = swash_init("utf8", "IsDigit", &PL_sv_undef, 0, 0); |
a0ed51b3 |
684 | return swash_fetch(PL_utf8_digit, p); |
685 | } |
686 | |
687 | bool |
864dbfa3 |
688 | Perl_is_utf8_upper(pTHX_ U8 *p) |
a0ed51b3 |
689 | { |
386d01d6 |
690 | if (!is_utf8_char(p)) |
691 | return FALSE; |
a0ed51b3 |
692 | if (!PL_utf8_upper) |
e24b16f9 |
693 | PL_utf8_upper = swash_init("utf8", "IsUpper", &PL_sv_undef, 0, 0); |
a0ed51b3 |
694 | return swash_fetch(PL_utf8_upper, p); |
695 | } |
696 | |
697 | bool |
864dbfa3 |
698 | Perl_is_utf8_lower(pTHX_ U8 *p) |
a0ed51b3 |
699 | { |
386d01d6 |
700 | if (!is_utf8_char(p)) |
701 | return FALSE; |
a0ed51b3 |
702 | if (!PL_utf8_lower) |
e24b16f9 |
703 | PL_utf8_lower = swash_init("utf8", "IsLower", &PL_sv_undef, 0, 0); |
a0ed51b3 |
704 | return swash_fetch(PL_utf8_lower, p); |
705 | } |
706 | |
707 | bool |
b8c5462f |
708 | Perl_is_utf8_cntrl(pTHX_ U8 *p) |
709 | { |
386d01d6 |
710 | if (!is_utf8_char(p)) |
711 | return FALSE; |
b8c5462f |
712 | if (!PL_utf8_cntrl) |
713 | PL_utf8_cntrl = swash_init("utf8", "IsCntrl", &PL_sv_undef, 0, 0); |
714 | return swash_fetch(PL_utf8_cntrl, p); |
715 | } |
716 | |
717 | bool |
718 | Perl_is_utf8_graph(pTHX_ U8 *p) |
719 | { |
386d01d6 |
720 | if (!is_utf8_char(p)) |
721 | return FALSE; |
b8c5462f |
722 | if (!PL_utf8_graph) |
723 | PL_utf8_graph = swash_init("utf8", "IsGraph", &PL_sv_undef, 0, 0); |
724 | return swash_fetch(PL_utf8_graph, p); |
725 | } |
726 | |
727 | bool |
864dbfa3 |
728 | Perl_is_utf8_print(pTHX_ U8 *p) |
a0ed51b3 |
729 | { |
386d01d6 |
730 | if (!is_utf8_char(p)) |
731 | return FALSE; |
a0ed51b3 |
732 | if (!PL_utf8_print) |
e24b16f9 |
733 | PL_utf8_print = swash_init("utf8", "IsPrint", &PL_sv_undef, 0, 0); |
a0ed51b3 |
734 | return swash_fetch(PL_utf8_print, p); |
735 | } |
736 | |
737 | bool |
b8c5462f |
738 | Perl_is_utf8_punct(pTHX_ U8 *p) |
739 | { |
386d01d6 |
740 | if (!is_utf8_char(p)) |
741 | return FALSE; |
b8c5462f |
742 | if (!PL_utf8_punct) |
743 | PL_utf8_punct = swash_init("utf8", "IsPunct", &PL_sv_undef, 0, 0); |
744 | return swash_fetch(PL_utf8_punct, p); |
745 | } |
746 | |
747 | bool |
748 | Perl_is_utf8_xdigit(pTHX_ U8 *p) |
749 | { |
386d01d6 |
750 | if (!is_utf8_char(p)) |
751 | return FALSE; |
b8c5462f |
752 | if (!PL_utf8_xdigit) |
753 | PL_utf8_xdigit = swash_init("utf8", "IsXDigit", &PL_sv_undef, 0, 0); |
754 | return swash_fetch(PL_utf8_xdigit, p); |
755 | } |
756 | |
757 | bool |
864dbfa3 |
758 | Perl_is_utf8_mark(pTHX_ U8 *p) |
a0ed51b3 |
759 | { |
386d01d6 |
760 | if (!is_utf8_char(p)) |
761 | return FALSE; |
a0ed51b3 |
762 | if (!PL_utf8_mark) |
e24b16f9 |
763 | PL_utf8_mark = swash_init("utf8", "IsM", &PL_sv_undef, 0, 0); |
a0ed51b3 |
764 | return swash_fetch(PL_utf8_mark, p); |
765 | } |
766 | |
2104c8d9 |
767 | UV |
864dbfa3 |
768 | Perl_to_utf8_upper(pTHX_ U8 *p) |
a0ed51b3 |
769 | { |
770 | UV uv; |
771 | |
772 | if (!PL_utf8_toupper) |
e24b16f9 |
773 | PL_utf8_toupper = swash_init("utf8", "ToUpper", &PL_sv_undef, 4, 0); |
a0ed51b3 |
774 | uv = swash_fetch(PL_utf8_toupper, p); |
775 | return uv ? uv : utf8_to_uv(p,0); |
776 | } |
777 | |
2104c8d9 |
778 | UV |
864dbfa3 |
779 | Perl_to_utf8_title(pTHX_ U8 *p) |
a0ed51b3 |
780 | { |
781 | UV uv; |
782 | |
783 | if (!PL_utf8_totitle) |
e24b16f9 |
784 | PL_utf8_totitle = swash_init("utf8", "ToTitle", &PL_sv_undef, 4, 0); |
a0ed51b3 |
785 | uv = swash_fetch(PL_utf8_totitle, p); |
786 | return uv ? uv : utf8_to_uv(p,0); |
787 | } |
788 | |
2104c8d9 |
789 | UV |
864dbfa3 |
790 | Perl_to_utf8_lower(pTHX_ U8 *p) |
a0ed51b3 |
791 | { |
792 | UV uv; |
793 | |
794 | if (!PL_utf8_tolower) |
e24b16f9 |
795 | PL_utf8_tolower = swash_init("utf8", "ToLower", &PL_sv_undef, 4, 0); |
a0ed51b3 |
796 | uv = swash_fetch(PL_utf8_tolower, p); |
797 | return uv ? uv : utf8_to_uv(p,0); |
798 | } |
799 | |
800 | /* a "swash" is a swatch hash */ |
801 | |
802 | SV* |
864dbfa3 |
803 | Perl_swash_init(pTHX_ char* pkg, char* name, SV *listsv, I32 minbits, I32 none) |
a0ed51b3 |
804 | { |
805 | SV* retval; |
806 | char tmpbuf[256]; |
807 | dSP; |
ce3b816e |
808 | |
809 | if (!gv_stashpv(pkg, 0)) { /* demand load utf8 */ |
810 | ENTER; |
811 | Perl_load_module(aTHX_ PERL_LOADMOD_NOIMPORT, newSVpv(pkg,0), Nullsv); |
812 | LEAVE; |
813 | } |
814 | SPAGAIN; |
a0ed51b3 |
815 | PUSHSTACKi(PERLSI_MAGIC); |
816 | PUSHMARK(SP); |
817 | EXTEND(SP,5); |
818 | PUSHs(sv_2mortal(newSVpvn(pkg, strlen(pkg)))); |
819 | PUSHs(sv_2mortal(newSVpvn(name, strlen(name)))); |
820 | PUSHs(listsv); |
821 | PUSHs(sv_2mortal(newSViv(minbits))); |
822 | PUSHs(sv_2mortal(newSViv(none))); |
823 | PUTBACK; |
824 | ENTER; |
825 | SAVEI32(PL_hints); |
826 | PL_hints = 0; |
827 | save_re_context(); |
e24b16f9 |
828 | if (PL_curcop == &PL_compiling) /* XXX ought to be handled by lex_start */ |
a0ed51b3 |
829 | strncpy(tmpbuf, PL_tokenbuf, sizeof tmpbuf); |
864dbfa3 |
830 | if (call_method("SWASHNEW", G_SCALAR)) |
e24b16f9 |
831 | retval = newSVsv(*PL_stack_sp--); |
a0ed51b3 |
832 | else |
e24b16f9 |
833 | retval = &PL_sv_undef; |
a0ed51b3 |
834 | LEAVE; |
835 | POPSTACK; |
e24b16f9 |
836 | if (PL_curcop == &PL_compiling) { |
a0ed51b3 |
837 | strncpy(PL_tokenbuf, tmpbuf, sizeof tmpbuf); |
e24b16f9 |
838 | PL_curcop->op_private = PL_hints; |
a0ed51b3 |
839 | } |
840 | if (!SvROK(retval) || SvTYPE(SvRV(retval)) != SVt_PVHV) |
cea2e8a9 |
841 | Perl_croak(aTHX_ "SWASHNEW didn't return an HV ref"); |
a0ed51b3 |
842 | return retval; |
843 | } |
844 | |
845 | UV |
864dbfa3 |
846 | Perl_swash_fetch(pTHX_ SV *sv, U8 *ptr) |
a0ed51b3 |
847 | { |
848 | HV* hv = (HV*)SvRV(sv); |
849 | U32 klen = UTF8SKIP(ptr) - 1; |
850 | U32 off = ptr[klen] & 127; /* NB: 64 bit always 0 when len > 1 */ |
851 | STRLEN slen; |
852 | STRLEN needents = (klen ? 64 : 128); |
dfe13c55 |
853 | U8 *tmps; |
a0ed51b3 |
854 | U32 bit; |
855 | SV *retval; |
856 | |
857 | /* |
858 | * This single-entry cache saves about 1/3 of the utf8 overhead in test |
859 | * suite. (That is, only 7-8% overall over just a hash cache. Still, |
860 | * it's nothing to sniff at.) Pity we usually come through at least |
861 | * two function calls to get here... |
862 | * |
863 | * NB: this code assumes that swatches are never modified, once generated! |
864 | */ |
865 | |
866 | if (hv == PL_last_swash_hv && |
867 | klen == PL_last_swash_klen && |
12ae5dfc |
868 | (!klen || memEQ((char *)ptr,(char *)PL_last_swash_key,klen)) ) |
a0ed51b3 |
869 | { |
870 | tmps = PL_last_swash_tmps; |
871 | slen = PL_last_swash_slen; |
872 | } |
873 | else { |
874 | /* Try our second-level swatch cache, kept in a hash. */ |
dfe13c55 |
875 | SV** svp = hv_fetch(hv, (char*)ptr, klen, FALSE); |
a0ed51b3 |
876 | |
877 | /* If not cached, generate it via utf8::SWASHGET */ |
dfe13c55 |
878 | if (!svp || !SvPOK(*svp) || !(tmps = (U8*)SvPV(*svp, slen))) { |
a0ed51b3 |
879 | dSP; |
880 | ENTER; |
881 | SAVETMPS; |
882 | save_re_context(); |
883 | PUSHSTACKi(PERLSI_MAGIC); |
884 | PUSHMARK(SP); |
885 | EXTEND(SP,3); |
886 | PUSHs((SV*)sv); |
887 | PUSHs(sv_2mortal(newSViv(utf8_to_uv(ptr, 0) & ~(needents - 1)))); |
888 | PUSHs(sv_2mortal(newSViv(needents))); |
889 | PUTBACK; |
864dbfa3 |
890 | if (call_method("SWASHGET", G_SCALAR)) |
e24b16f9 |
891 | retval = newSVsv(*PL_stack_sp--); |
a0ed51b3 |
892 | else |
e24b16f9 |
893 | retval = &PL_sv_undef; |
a0ed51b3 |
894 | POPSTACK; |
895 | FREETMPS; |
896 | LEAVE; |
e24b16f9 |
897 | if (PL_curcop == &PL_compiling) |
898 | PL_curcop->op_private = PL_hints; |
a0ed51b3 |
899 | |
dfe13c55 |
900 | svp = hv_store(hv, (char*)ptr, klen, retval, 0); |
a0ed51b3 |
901 | |
dfe13c55 |
902 | if (!svp || !(tmps = (U8*)SvPV(*svp, slen)) || slen < 8) |
cea2e8a9 |
903 | Perl_croak(aTHX_ "SWASHGET didn't return result of proper length"); |
a0ed51b3 |
904 | } |
905 | |
906 | PL_last_swash_hv = hv; |
907 | PL_last_swash_klen = klen; |
908 | PL_last_swash_tmps = tmps; |
909 | PL_last_swash_slen = slen; |
910 | if (klen) |
911 | Copy(ptr, PL_last_swash_key, klen, U8); |
912 | } |
913 | |
914 | switch ((slen << 3) / needents) { |
915 | case 1: |
916 | bit = 1 << (off & 7); |
917 | off >>= 3; |
918 | return (tmps[off] & bit) != 0; |
919 | case 8: |
920 | return tmps[off]; |
921 | case 16: |
922 | off <<= 1; |
923 | return (tmps[off] << 8) + tmps[off + 1] ; |
924 | case 32: |
925 | off <<= 2; |
926 | return (tmps[off] << 24) + (tmps[off+1] << 16) + (tmps[off+2] << 8) + tmps[off + 3] ; |
927 | } |
cea2e8a9 |
928 | Perl_croak(aTHX_ "panic: swash_fetch"); |
a0ed51b3 |
929 | return 0; |
930 | } |