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 | /* |
324 | * Convert native or reversed UTF-16 to UTF-8. |
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 | |
4ac469dc |
329 | /* There are several problems with utf16_to_utf8(). |
330 | * (1) U16 is not necessarily *exactly* two bytes. |
331 | * (2) Secondly, no check is made for odd length. |
332 | * (3) Thirdly, the "Malformed UTF-16 surrogate" should probably be |
333 | * a hard error (and it should be listed in perldiag). |
334 | * (4) The tests (in comp/t/require.t) are a joke: the UTF16 BOM |
335 | * really ought to be followed by valid UTF16 characters. |
336 | * --Mike Guy */ |
a0ed51b3 |
337 | U8* |
864dbfa3 |
338 | Perl_utf16_to_utf8(pTHX_ U16* p, U8* d, I32 bytelen) |
a0ed51b3 |
339 | { |
340 | U16* pend = p + bytelen / 2; |
341 | while (p < pend) { |
342 | UV uv = *p++; |
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; |
a0ed51b3 |
354 | int low = *p++; |
355 | if (low < 0xdc00 || low >= 0xdfff) { |
0453d815 |
356 | if (ckWARN_d(WARN_UTF8)) |
357 | Perl_warner(aTHX_ WARN_UTF8, "Malformed UTF-16 surrogate"); |
a0ed51b3 |
358 | p--; |
359 | uv = 0xfffd; |
360 | } |
361 | uv = ((uv - 0xd800) << 10) + (low - 0xdc00) + 0x10000; |
362 | } |
363 | if (uv < 0x10000) { |
364 | *d++ = (( uv >> 12) | 0xe0); |
365 | *d++ = (((uv >> 6) & 0x3f) | 0x80); |
366 | *d++ = (( uv & 0x3f) | 0x80); |
367 | continue; |
368 | } |
369 | else { |
370 | *d++ = (( uv >> 18) | 0xf0); |
371 | *d++ = (((uv >> 12) & 0x3f) | 0x80); |
372 | *d++ = (((uv >> 6) & 0x3f) | 0x80); |
373 | *d++ = (( uv & 0x3f) | 0x80); |
374 | continue; |
375 | } |
376 | } |
377 | return d; |
378 | } |
379 | |
380 | /* Note: this one is slightly destructive of the source. */ |
381 | |
382 | U8* |
864dbfa3 |
383 | Perl_utf16_to_utf8_reversed(pTHX_ U16* p, U8* d, I32 bytelen) |
a0ed51b3 |
384 | { |
385 | U8* s = (U8*)p; |
386 | U8* send = s + bytelen; |
387 | while (s < send) { |
388 | U8 tmp = s[0]; |
389 | s[0] = s[1]; |
390 | s[1] = tmp; |
391 | s += 2; |
392 | } |
393 | return utf16_to_utf8(p, d, bytelen); |
394 | } |
395 | |
396 | /* for now these are all defined (inefficiently) in terms of the utf8 versions */ |
397 | |
398 | bool |
864dbfa3 |
399 | Perl_is_uni_alnum(pTHX_ U32 c) |
a0ed51b3 |
400 | { |
aa6ffa16 |
401 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
402 | uv_to_utf8(tmpbuf, (UV)c); |
403 | return is_utf8_alnum(tmpbuf); |
404 | } |
405 | |
406 | bool |
b8c5462f |
407 | Perl_is_uni_alnumc(pTHX_ U32 c) |
408 | { |
aa6ffa16 |
409 | U8 tmpbuf[UTF8_MAXLEN]; |
b8c5462f |
410 | uv_to_utf8(tmpbuf, (UV)c); |
411 | return is_utf8_alnumc(tmpbuf); |
412 | } |
413 | |
414 | bool |
864dbfa3 |
415 | Perl_is_uni_idfirst(pTHX_ U32 c) |
a0ed51b3 |
416 | { |
aa6ffa16 |
417 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
418 | uv_to_utf8(tmpbuf, (UV)c); |
419 | return is_utf8_idfirst(tmpbuf); |
420 | } |
421 | |
422 | bool |
864dbfa3 |
423 | Perl_is_uni_alpha(pTHX_ U32 c) |
a0ed51b3 |
424 | { |
aa6ffa16 |
425 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
426 | uv_to_utf8(tmpbuf, (UV)c); |
427 | return is_utf8_alpha(tmpbuf); |
428 | } |
429 | |
430 | bool |
4d61ec05 |
431 | Perl_is_uni_ascii(pTHX_ U32 c) |
432 | { |
aa6ffa16 |
433 | U8 tmpbuf[UTF8_MAXLEN]; |
4d61ec05 |
434 | uv_to_utf8(tmpbuf, (UV)c); |
435 | return is_utf8_ascii(tmpbuf); |
436 | } |
437 | |
438 | bool |
864dbfa3 |
439 | Perl_is_uni_space(pTHX_ U32 c) |
a0ed51b3 |
440 | { |
aa6ffa16 |
441 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
442 | uv_to_utf8(tmpbuf, (UV)c); |
443 | return is_utf8_space(tmpbuf); |
444 | } |
445 | |
446 | bool |
864dbfa3 |
447 | Perl_is_uni_digit(pTHX_ U32 c) |
a0ed51b3 |
448 | { |
aa6ffa16 |
449 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
450 | uv_to_utf8(tmpbuf, (UV)c); |
451 | return is_utf8_digit(tmpbuf); |
452 | } |
453 | |
454 | bool |
864dbfa3 |
455 | Perl_is_uni_upper(pTHX_ U32 c) |
a0ed51b3 |
456 | { |
aa6ffa16 |
457 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
458 | uv_to_utf8(tmpbuf, (UV)c); |
459 | return is_utf8_upper(tmpbuf); |
460 | } |
461 | |
462 | bool |
864dbfa3 |
463 | Perl_is_uni_lower(pTHX_ U32 c) |
a0ed51b3 |
464 | { |
aa6ffa16 |
465 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
466 | uv_to_utf8(tmpbuf, (UV)c); |
467 | return is_utf8_lower(tmpbuf); |
468 | } |
469 | |
470 | bool |
b8c5462f |
471 | Perl_is_uni_cntrl(pTHX_ U32 c) |
472 | { |
aa6ffa16 |
473 | U8 tmpbuf[UTF8_MAXLEN]; |
b8c5462f |
474 | uv_to_utf8(tmpbuf, (UV)c); |
475 | return is_utf8_cntrl(tmpbuf); |
476 | } |
477 | |
478 | bool |
479 | Perl_is_uni_graph(pTHX_ U32 c) |
480 | { |
aa6ffa16 |
481 | U8 tmpbuf[UTF8_MAXLEN]; |
b8c5462f |
482 | uv_to_utf8(tmpbuf, (UV)c); |
483 | return is_utf8_graph(tmpbuf); |
484 | } |
485 | |
486 | bool |
864dbfa3 |
487 | Perl_is_uni_print(pTHX_ U32 c) |
a0ed51b3 |
488 | { |
aa6ffa16 |
489 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
490 | uv_to_utf8(tmpbuf, (UV)c); |
491 | return is_utf8_print(tmpbuf); |
492 | } |
493 | |
b8c5462f |
494 | bool |
f248d071 |
495 | Perl_is_uni_punct(pTHX_ U32 c) |
b8c5462f |
496 | { |
aa6ffa16 |
497 | U8 tmpbuf[UTF8_MAXLEN]; |
b8c5462f |
498 | uv_to_utf8(tmpbuf, (UV)c); |
499 | return is_utf8_punct(tmpbuf); |
500 | } |
501 | |
4d61ec05 |
502 | bool |
503 | Perl_is_uni_xdigit(pTHX_ U32 c) |
504 | { |
aa6ffa16 |
505 | U8 tmpbuf[UTF8_MAXLEN]; |
4d61ec05 |
506 | uv_to_utf8(tmpbuf, (UV)c); |
507 | return is_utf8_xdigit(tmpbuf); |
508 | } |
509 | |
a0ed51b3 |
510 | U32 |
864dbfa3 |
511 | Perl_to_uni_upper(pTHX_ U32 c) |
a0ed51b3 |
512 | { |
aa6ffa16 |
513 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
514 | uv_to_utf8(tmpbuf, (UV)c); |
515 | return to_utf8_upper(tmpbuf); |
516 | } |
517 | |
518 | U32 |
864dbfa3 |
519 | Perl_to_uni_title(pTHX_ U32 c) |
a0ed51b3 |
520 | { |
aa6ffa16 |
521 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
522 | uv_to_utf8(tmpbuf, (UV)c); |
523 | return to_utf8_title(tmpbuf); |
524 | } |
525 | |
526 | U32 |
864dbfa3 |
527 | Perl_to_uni_lower(pTHX_ U32 c) |
a0ed51b3 |
528 | { |
aa6ffa16 |
529 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
530 | uv_to_utf8(tmpbuf, (UV)c); |
531 | return to_utf8_lower(tmpbuf); |
532 | } |
533 | |
534 | /* for now these all assume no locale info available for Unicode > 255 */ |
535 | |
536 | bool |
864dbfa3 |
537 | Perl_is_uni_alnum_lc(pTHX_ U32 c) |
a0ed51b3 |
538 | { |
539 | return is_uni_alnum(c); /* XXX no locale support yet */ |
540 | } |
541 | |
542 | bool |
b8c5462f |
543 | Perl_is_uni_alnumc_lc(pTHX_ U32 c) |
544 | { |
545 | return is_uni_alnumc(c); /* XXX no locale support yet */ |
546 | } |
547 | |
548 | bool |
864dbfa3 |
549 | Perl_is_uni_idfirst_lc(pTHX_ U32 c) |
a0ed51b3 |
550 | { |
551 | return is_uni_idfirst(c); /* XXX no locale support yet */ |
552 | } |
553 | |
554 | bool |
864dbfa3 |
555 | Perl_is_uni_alpha_lc(pTHX_ U32 c) |
a0ed51b3 |
556 | { |
557 | return is_uni_alpha(c); /* XXX no locale support yet */ |
558 | } |
559 | |
560 | bool |
4d61ec05 |
561 | Perl_is_uni_ascii_lc(pTHX_ U32 c) |
562 | { |
563 | return is_uni_ascii(c); /* XXX no locale support yet */ |
564 | } |
565 | |
566 | bool |
864dbfa3 |
567 | Perl_is_uni_space_lc(pTHX_ U32 c) |
a0ed51b3 |
568 | { |
569 | return is_uni_space(c); /* XXX no locale support yet */ |
570 | } |
571 | |
572 | bool |
864dbfa3 |
573 | Perl_is_uni_digit_lc(pTHX_ U32 c) |
a0ed51b3 |
574 | { |
575 | return is_uni_digit(c); /* XXX no locale support yet */ |
576 | } |
577 | |
578 | bool |
864dbfa3 |
579 | Perl_is_uni_upper_lc(pTHX_ U32 c) |
a0ed51b3 |
580 | { |
581 | return is_uni_upper(c); /* XXX no locale support yet */ |
582 | } |
583 | |
584 | bool |
864dbfa3 |
585 | Perl_is_uni_lower_lc(pTHX_ U32 c) |
a0ed51b3 |
586 | { |
587 | return is_uni_lower(c); /* XXX no locale support yet */ |
588 | } |
589 | |
590 | bool |
b8c5462f |
591 | Perl_is_uni_cntrl_lc(pTHX_ U32 c) |
592 | { |
593 | return is_uni_cntrl(c); /* XXX no locale support yet */ |
594 | } |
595 | |
596 | bool |
597 | Perl_is_uni_graph_lc(pTHX_ U32 c) |
598 | { |
599 | return is_uni_graph(c); /* XXX no locale support yet */ |
600 | } |
601 | |
602 | bool |
864dbfa3 |
603 | Perl_is_uni_print_lc(pTHX_ U32 c) |
a0ed51b3 |
604 | { |
605 | return is_uni_print(c); /* XXX no locale support yet */ |
606 | } |
607 | |
b8c5462f |
608 | bool |
609 | Perl_is_uni_punct_lc(pTHX_ U32 c) |
610 | { |
611 | return is_uni_punct(c); /* XXX no locale support yet */ |
612 | } |
613 | |
4d61ec05 |
614 | bool |
615 | Perl_is_uni_xdigit_lc(pTHX_ U32 c) |
616 | { |
617 | return is_uni_xdigit(c); /* XXX no locale support yet */ |
618 | } |
619 | |
a0ed51b3 |
620 | U32 |
864dbfa3 |
621 | Perl_to_uni_upper_lc(pTHX_ U32 c) |
a0ed51b3 |
622 | { |
623 | return to_uni_upper(c); /* XXX no locale support yet */ |
624 | } |
625 | |
626 | U32 |
864dbfa3 |
627 | Perl_to_uni_title_lc(pTHX_ U32 c) |
a0ed51b3 |
628 | { |
629 | return to_uni_title(c); /* XXX no locale support yet */ |
630 | } |
631 | |
632 | U32 |
864dbfa3 |
633 | Perl_to_uni_lower_lc(pTHX_ U32 c) |
a0ed51b3 |
634 | { |
635 | return to_uni_lower(c); /* XXX no locale support yet */ |
636 | } |
637 | |
a0ed51b3 |
638 | bool |
864dbfa3 |
639 | Perl_is_utf8_alnum(pTHX_ U8 *p) |
a0ed51b3 |
640 | { |
386d01d6 |
641 | if (!is_utf8_char(p)) |
642 | return FALSE; |
a0ed51b3 |
643 | if (!PL_utf8_alnum) |
e24b16f9 |
644 | PL_utf8_alnum = swash_init("utf8", "IsAlnum", &PL_sv_undef, 0, 0); |
a0ed51b3 |
645 | return swash_fetch(PL_utf8_alnum, p); |
646 | /* return *p == '_' || is_utf8_alpha(p) || is_utf8_digit(p); */ |
647 | #ifdef SURPRISINGLY_SLOWER /* probably because alpha is usually true */ |
648 | if (!PL_utf8_alnum) |
649 | PL_utf8_alnum = swash_init("utf8", "", |
650 | sv_2mortal(newSVpv("+utf8::IsAlpha\n+utf8::IsDigit\n005F\n",0)), 0, 0); |
651 | return swash_fetch(PL_utf8_alnum, p); |
652 | #endif |
653 | } |
654 | |
655 | bool |
b8c5462f |
656 | Perl_is_utf8_alnumc(pTHX_ U8 *p) |
657 | { |
386d01d6 |
658 | if (!is_utf8_char(p)) |
659 | return FALSE; |
b8c5462f |
660 | if (!PL_utf8_alnum) |
661 | PL_utf8_alnum = swash_init("utf8", "IsAlnumC", &PL_sv_undef, 0, 0); |
662 | return swash_fetch(PL_utf8_alnum, p); |
663 | /* return is_utf8_alpha(p) || is_utf8_digit(p); */ |
664 | #ifdef SURPRISINGLY_SLOWER /* probably because alpha is usually true */ |
665 | if (!PL_utf8_alnum) |
666 | PL_utf8_alnum = swash_init("utf8", "", |
667 | sv_2mortal(newSVpv("+utf8::IsAlpha\n+utf8::IsDigit\n005F\n",0)), 0, 0); |
668 | return swash_fetch(PL_utf8_alnum, p); |
669 | #endif |
670 | } |
671 | |
672 | bool |
864dbfa3 |
673 | Perl_is_utf8_idfirst(pTHX_ U8 *p) |
a0ed51b3 |
674 | { |
675 | return *p == '_' || is_utf8_alpha(p); |
676 | } |
677 | |
678 | bool |
864dbfa3 |
679 | Perl_is_utf8_alpha(pTHX_ U8 *p) |
a0ed51b3 |
680 | { |
386d01d6 |
681 | if (!is_utf8_char(p)) |
682 | return FALSE; |
a0ed51b3 |
683 | if (!PL_utf8_alpha) |
e24b16f9 |
684 | PL_utf8_alpha = swash_init("utf8", "IsAlpha", &PL_sv_undef, 0, 0); |
a0ed51b3 |
685 | return swash_fetch(PL_utf8_alpha, p); |
686 | } |
687 | |
688 | bool |
b8c5462f |
689 | Perl_is_utf8_ascii(pTHX_ U8 *p) |
690 | { |
386d01d6 |
691 | if (!is_utf8_char(p)) |
692 | return FALSE; |
b8c5462f |
693 | if (!PL_utf8_ascii) |
694 | PL_utf8_ascii = swash_init("utf8", "IsAscii", &PL_sv_undef, 0, 0); |
695 | return swash_fetch(PL_utf8_ascii, p); |
696 | } |
697 | |
698 | bool |
864dbfa3 |
699 | Perl_is_utf8_space(pTHX_ U8 *p) |
a0ed51b3 |
700 | { |
386d01d6 |
701 | if (!is_utf8_char(p)) |
702 | return FALSE; |
a0ed51b3 |
703 | if (!PL_utf8_space) |
e24b16f9 |
704 | PL_utf8_space = swash_init("utf8", "IsSpace", &PL_sv_undef, 0, 0); |
a0ed51b3 |
705 | return swash_fetch(PL_utf8_space, p); |
706 | } |
707 | |
708 | bool |
864dbfa3 |
709 | Perl_is_utf8_digit(pTHX_ U8 *p) |
a0ed51b3 |
710 | { |
386d01d6 |
711 | if (!is_utf8_char(p)) |
712 | return FALSE; |
a0ed51b3 |
713 | if (!PL_utf8_digit) |
e24b16f9 |
714 | PL_utf8_digit = swash_init("utf8", "IsDigit", &PL_sv_undef, 0, 0); |
a0ed51b3 |
715 | return swash_fetch(PL_utf8_digit, p); |
716 | } |
717 | |
718 | bool |
864dbfa3 |
719 | Perl_is_utf8_upper(pTHX_ U8 *p) |
a0ed51b3 |
720 | { |
386d01d6 |
721 | if (!is_utf8_char(p)) |
722 | return FALSE; |
a0ed51b3 |
723 | if (!PL_utf8_upper) |
e24b16f9 |
724 | PL_utf8_upper = swash_init("utf8", "IsUpper", &PL_sv_undef, 0, 0); |
a0ed51b3 |
725 | return swash_fetch(PL_utf8_upper, p); |
726 | } |
727 | |
728 | bool |
864dbfa3 |
729 | Perl_is_utf8_lower(pTHX_ U8 *p) |
a0ed51b3 |
730 | { |
386d01d6 |
731 | if (!is_utf8_char(p)) |
732 | return FALSE; |
a0ed51b3 |
733 | if (!PL_utf8_lower) |
e24b16f9 |
734 | PL_utf8_lower = swash_init("utf8", "IsLower", &PL_sv_undef, 0, 0); |
a0ed51b3 |
735 | return swash_fetch(PL_utf8_lower, p); |
736 | } |
737 | |
738 | bool |
b8c5462f |
739 | Perl_is_utf8_cntrl(pTHX_ U8 *p) |
740 | { |
386d01d6 |
741 | if (!is_utf8_char(p)) |
742 | return FALSE; |
b8c5462f |
743 | if (!PL_utf8_cntrl) |
744 | PL_utf8_cntrl = swash_init("utf8", "IsCntrl", &PL_sv_undef, 0, 0); |
745 | return swash_fetch(PL_utf8_cntrl, p); |
746 | } |
747 | |
748 | bool |
749 | Perl_is_utf8_graph(pTHX_ U8 *p) |
750 | { |
386d01d6 |
751 | if (!is_utf8_char(p)) |
752 | return FALSE; |
b8c5462f |
753 | if (!PL_utf8_graph) |
754 | PL_utf8_graph = swash_init("utf8", "IsGraph", &PL_sv_undef, 0, 0); |
755 | return swash_fetch(PL_utf8_graph, p); |
756 | } |
757 | |
758 | bool |
864dbfa3 |
759 | Perl_is_utf8_print(pTHX_ U8 *p) |
a0ed51b3 |
760 | { |
386d01d6 |
761 | if (!is_utf8_char(p)) |
762 | return FALSE; |
a0ed51b3 |
763 | if (!PL_utf8_print) |
e24b16f9 |
764 | PL_utf8_print = swash_init("utf8", "IsPrint", &PL_sv_undef, 0, 0); |
a0ed51b3 |
765 | return swash_fetch(PL_utf8_print, p); |
766 | } |
767 | |
768 | bool |
b8c5462f |
769 | Perl_is_utf8_punct(pTHX_ U8 *p) |
770 | { |
386d01d6 |
771 | if (!is_utf8_char(p)) |
772 | return FALSE; |
b8c5462f |
773 | if (!PL_utf8_punct) |
774 | PL_utf8_punct = swash_init("utf8", "IsPunct", &PL_sv_undef, 0, 0); |
775 | return swash_fetch(PL_utf8_punct, p); |
776 | } |
777 | |
778 | bool |
779 | Perl_is_utf8_xdigit(pTHX_ U8 *p) |
780 | { |
386d01d6 |
781 | if (!is_utf8_char(p)) |
782 | return FALSE; |
b8c5462f |
783 | if (!PL_utf8_xdigit) |
784 | PL_utf8_xdigit = swash_init("utf8", "IsXDigit", &PL_sv_undef, 0, 0); |
785 | return swash_fetch(PL_utf8_xdigit, p); |
786 | } |
787 | |
788 | bool |
864dbfa3 |
789 | Perl_is_utf8_mark(pTHX_ U8 *p) |
a0ed51b3 |
790 | { |
386d01d6 |
791 | if (!is_utf8_char(p)) |
792 | return FALSE; |
a0ed51b3 |
793 | if (!PL_utf8_mark) |
e24b16f9 |
794 | PL_utf8_mark = swash_init("utf8", "IsM", &PL_sv_undef, 0, 0); |
a0ed51b3 |
795 | return swash_fetch(PL_utf8_mark, p); |
796 | } |
797 | |
2104c8d9 |
798 | UV |
864dbfa3 |
799 | Perl_to_utf8_upper(pTHX_ U8 *p) |
a0ed51b3 |
800 | { |
801 | UV uv; |
802 | |
803 | if (!PL_utf8_toupper) |
e24b16f9 |
804 | PL_utf8_toupper = swash_init("utf8", "ToUpper", &PL_sv_undef, 4, 0); |
a0ed51b3 |
805 | uv = swash_fetch(PL_utf8_toupper, p); |
806 | return uv ? uv : utf8_to_uv(p,0); |
807 | } |
808 | |
2104c8d9 |
809 | UV |
864dbfa3 |
810 | Perl_to_utf8_title(pTHX_ U8 *p) |
a0ed51b3 |
811 | { |
812 | UV uv; |
813 | |
814 | if (!PL_utf8_totitle) |
e24b16f9 |
815 | PL_utf8_totitle = swash_init("utf8", "ToTitle", &PL_sv_undef, 4, 0); |
a0ed51b3 |
816 | uv = swash_fetch(PL_utf8_totitle, p); |
817 | return uv ? uv : utf8_to_uv(p,0); |
818 | } |
819 | |
2104c8d9 |
820 | UV |
864dbfa3 |
821 | Perl_to_utf8_lower(pTHX_ U8 *p) |
a0ed51b3 |
822 | { |
823 | UV uv; |
824 | |
825 | if (!PL_utf8_tolower) |
e24b16f9 |
826 | PL_utf8_tolower = swash_init("utf8", "ToLower", &PL_sv_undef, 4, 0); |
a0ed51b3 |
827 | uv = swash_fetch(PL_utf8_tolower, p); |
828 | return uv ? uv : utf8_to_uv(p,0); |
829 | } |
830 | |
831 | /* a "swash" is a swatch hash */ |
832 | |
833 | SV* |
864dbfa3 |
834 | Perl_swash_init(pTHX_ char* pkg, char* name, SV *listsv, I32 minbits, I32 none) |
a0ed51b3 |
835 | { |
836 | SV* retval; |
837 | char tmpbuf[256]; |
838 | dSP; |
ce3b816e |
839 | |
840 | if (!gv_stashpv(pkg, 0)) { /* demand load utf8 */ |
841 | ENTER; |
842 | Perl_load_module(aTHX_ PERL_LOADMOD_NOIMPORT, newSVpv(pkg,0), Nullsv); |
843 | LEAVE; |
844 | } |
845 | SPAGAIN; |
a0ed51b3 |
846 | PUSHSTACKi(PERLSI_MAGIC); |
847 | PUSHMARK(SP); |
848 | EXTEND(SP,5); |
849 | PUSHs(sv_2mortal(newSVpvn(pkg, strlen(pkg)))); |
850 | PUSHs(sv_2mortal(newSVpvn(name, strlen(name)))); |
851 | PUSHs(listsv); |
852 | PUSHs(sv_2mortal(newSViv(minbits))); |
853 | PUSHs(sv_2mortal(newSViv(none))); |
854 | PUTBACK; |
855 | ENTER; |
856 | SAVEI32(PL_hints); |
857 | PL_hints = 0; |
858 | save_re_context(); |
e24b16f9 |
859 | if (PL_curcop == &PL_compiling) /* XXX ought to be handled by lex_start */ |
a0ed51b3 |
860 | strncpy(tmpbuf, PL_tokenbuf, sizeof tmpbuf); |
864dbfa3 |
861 | if (call_method("SWASHNEW", G_SCALAR)) |
e24b16f9 |
862 | retval = newSVsv(*PL_stack_sp--); |
a0ed51b3 |
863 | else |
e24b16f9 |
864 | retval = &PL_sv_undef; |
a0ed51b3 |
865 | LEAVE; |
866 | POPSTACK; |
e24b16f9 |
867 | if (PL_curcop == &PL_compiling) { |
a0ed51b3 |
868 | strncpy(PL_tokenbuf, tmpbuf, sizeof tmpbuf); |
e24b16f9 |
869 | PL_curcop->op_private = PL_hints; |
a0ed51b3 |
870 | } |
871 | if (!SvROK(retval) || SvTYPE(SvRV(retval)) != SVt_PVHV) |
cea2e8a9 |
872 | Perl_croak(aTHX_ "SWASHNEW didn't return an HV ref"); |
a0ed51b3 |
873 | return retval; |
874 | } |
875 | |
876 | UV |
864dbfa3 |
877 | Perl_swash_fetch(pTHX_ SV *sv, U8 *ptr) |
a0ed51b3 |
878 | { |
879 | HV* hv = (HV*)SvRV(sv); |
880 | U32 klen = UTF8SKIP(ptr) - 1; |
881 | U32 off = ptr[klen] & 127; /* NB: 64 bit always 0 when len > 1 */ |
882 | STRLEN slen; |
883 | STRLEN needents = (klen ? 64 : 128); |
dfe13c55 |
884 | U8 *tmps; |
a0ed51b3 |
885 | U32 bit; |
886 | SV *retval; |
887 | |
888 | /* |
889 | * This single-entry cache saves about 1/3 of the utf8 overhead in test |
890 | * suite. (That is, only 7-8% overall over just a hash cache. Still, |
891 | * it's nothing to sniff at.) Pity we usually come through at least |
892 | * two function calls to get here... |
893 | * |
894 | * NB: this code assumes that swatches are never modified, once generated! |
895 | */ |
896 | |
897 | if (hv == PL_last_swash_hv && |
898 | klen == PL_last_swash_klen && |
12ae5dfc |
899 | (!klen || memEQ((char *)ptr,(char *)PL_last_swash_key,klen)) ) |
a0ed51b3 |
900 | { |
901 | tmps = PL_last_swash_tmps; |
902 | slen = PL_last_swash_slen; |
903 | } |
904 | else { |
905 | /* Try our second-level swatch cache, kept in a hash. */ |
dfe13c55 |
906 | SV** svp = hv_fetch(hv, (char*)ptr, klen, FALSE); |
a0ed51b3 |
907 | |
908 | /* If not cached, generate it via utf8::SWASHGET */ |
dfe13c55 |
909 | if (!svp || !SvPOK(*svp) || !(tmps = (U8*)SvPV(*svp, slen))) { |
a0ed51b3 |
910 | dSP; |
911 | ENTER; |
912 | SAVETMPS; |
913 | save_re_context(); |
914 | PUSHSTACKi(PERLSI_MAGIC); |
915 | PUSHMARK(SP); |
916 | EXTEND(SP,3); |
917 | PUSHs((SV*)sv); |
918 | PUSHs(sv_2mortal(newSViv(utf8_to_uv(ptr, 0) & ~(needents - 1)))); |
919 | PUSHs(sv_2mortal(newSViv(needents))); |
920 | PUTBACK; |
864dbfa3 |
921 | if (call_method("SWASHGET", G_SCALAR)) |
e24b16f9 |
922 | retval = newSVsv(*PL_stack_sp--); |
a0ed51b3 |
923 | else |
e24b16f9 |
924 | retval = &PL_sv_undef; |
a0ed51b3 |
925 | POPSTACK; |
926 | FREETMPS; |
927 | LEAVE; |
e24b16f9 |
928 | if (PL_curcop == &PL_compiling) |
929 | PL_curcop->op_private = PL_hints; |
a0ed51b3 |
930 | |
dfe13c55 |
931 | svp = hv_store(hv, (char*)ptr, klen, retval, 0); |
a0ed51b3 |
932 | |
dfe13c55 |
933 | if (!svp || !(tmps = (U8*)SvPV(*svp, slen)) || slen < 8) |
cea2e8a9 |
934 | Perl_croak(aTHX_ "SWASHGET didn't return result of proper length"); |
a0ed51b3 |
935 | } |
936 | |
937 | PL_last_swash_hv = hv; |
938 | PL_last_swash_klen = klen; |
939 | PL_last_swash_tmps = tmps; |
940 | PL_last_swash_slen = slen; |
941 | if (klen) |
942 | Copy(ptr, PL_last_swash_key, klen, U8); |
943 | } |
944 | |
945 | switch ((slen << 3) / needents) { |
946 | case 1: |
947 | bit = 1 << (off & 7); |
948 | off >>= 3; |
949 | return (tmps[off] & bit) != 0; |
950 | case 8: |
951 | return tmps[off]; |
952 | case 16: |
953 | off <<= 1; |
954 | return (tmps[off] << 8) + tmps[off + 1] ; |
955 | case 32: |
956 | off <<= 2; |
957 | return (tmps[off] << 24) + (tmps[off+1] << 16) + (tmps[off+2] << 8) + tmps[off + 3] ; |
958 | } |
cea2e8a9 |
959 | Perl_croak(aTHX_ "panic: swash_fetch"); |
a0ed51b3 |
960 | return 0; |
961 | } |