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 | |
246fae53 |
207 | /* utf8_distance(a,b) returns the number of UTF8 characters between |
208 | the pointers a and b */ |
a0ed51b3 |
209 | |
210 | I32 |
864dbfa3 |
211 | Perl_utf8_distance(pTHX_ U8 *a, U8 *b) |
a0ed51b3 |
212 | { |
213 | I32 off = 0; |
214 | if (a < b) { |
215 | while (a < b) { |
216 | a += UTF8SKIP(a); |
217 | off--; |
218 | } |
219 | } |
220 | else { |
221 | while (b < a) { |
222 | b += UTF8SKIP(b); |
223 | off++; |
224 | } |
225 | } |
226 | return off; |
227 | } |
228 | |
229 | /* WARNING: do not use the following unless you *know* off is within bounds */ |
230 | |
231 | U8 * |
864dbfa3 |
232 | Perl_utf8_hop(pTHX_ U8 *s, I32 off) |
a0ed51b3 |
233 | { |
234 | if (off >= 0) { |
235 | while (off--) |
236 | s += UTF8SKIP(s); |
237 | } |
238 | else { |
239 | while (off++) { |
240 | s--; |
241 | if (*s & 0x80) { |
242 | while ((*s & 0xc0) == 0x80) |
243 | s--; |
244 | } |
245 | } |
246 | } |
247 | return s; |
248 | } |
249 | |
6940069f |
250 | /* |
246fae53 |
251 | =for apidoc Am|U8 *|utf8_to_bytes|U8 *s|STRLEN *len |
6940069f |
252 | |
246fae53 |
253 | Converts a string C<s> of length C<len> from UTF8 into byte encoding. |
254 | Unlike C<bytes_to_utf8>, this over-writes the original string, and |
255 | updates len to contain the new length. |
256 | Returns zero on failure leaving the string and len unchanged |
6940069f |
257 | |
258 | =cut |
259 | */ |
260 | |
261 | U8 * |
246fae53 |
262 | Perl_utf8_to_bytes(pTHX_ U8* s, STRLEN *len) |
6940069f |
263 | { |
264 | dTHR; |
265 | U8 *send; |
266 | U8 *d; |
267 | U8 *save; |
268 | |
246fae53 |
269 | send = s + *len; |
6940069f |
270 | d = save = s; |
246fae53 |
271 | |
272 | /* ensure valid UTF8 and chars < 256 before updating string */ |
273 | while (s < send) { |
274 | U8 c = *s++; |
275 | if (c >= 0x80 && |
276 | ( (s >= send) || ((*s++ & 0xc0) != 0x80) || ((c & 0xfe) != 0xc2))) |
277 | return 0; |
278 | } |
279 | s = save; |
6940069f |
280 | while (s < send) { |
281 | if (*s < 0x80) |
282 | *d++ = *s++; |
283 | else { |
284 | I32 ulen; |
246fae53 |
285 | *d++ = (U8)utf8_to_uv(s, &ulen); |
6940069f |
286 | s += ulen; |
6940069f |
287 | } |
288 | } |
289 | *d = '\0'; |
246fae53 |
290 | *len = d - save; |
6940069f |
291 | return save; |
292 | } |
293 | |
294 | /* |
6662521e |
295 | =for apidoc Am|U8 *|bytes_to_utf8|U8 *s|STRLEN *len |
6940069f |
296 | |
297 | Converts a string C<s> of length C<len> from ASCII into UTF8 encoding. |
6662521e |
298 | Returns a pointer to the newly-created string, and sets C<len> to |
299 | reflect the new length. |
6940069f |
300 | |
497711e7 |
301 | =cut |
6940069f |
302 | */ |
303 | |
304 | U8* |
6662521e |
305 | Perl_bytes_to_utf8(pTHX_ U8* s, STRLEN *len) |
6940069f |
306 | { |
307 | dTHR; |
308 | U8 *send; |
309 | U8 *d; |
310 | U8 *dst; |
6662521e |
311 | send = s + (*len); |
6940069f |
312 | |
6662521e |
313 | Newz(801, d, (*len) * 2 + 1, U8); |
6940069f |
314 | dst = d; |
315 | |
316 | while (s < send) { |
317 | if (*s < 0x80) |
318 | *d++ = *s++; |
319 | else { |
320 | UV uv = *s++; |
321 | *d++ = (( uv >> 6) | 0xc0); |
322 | *d++ = (( uv & 0x3f) | 0x80); |
323 | } |
324 | } |
325 | *d = '\0'; |
6662521e |
326 | *len = d-dst; |
6940069f |
327 | return dst; |
328 | } |
329 | |
a0ed51b3 |
330 | /* |
dea0fc0b |
331 | * Convert native (big-endian) or reversed (little-endian) UTF-16 to UTF-8. |
a0ed51b3 |
332 | * |
333 | * Destination must be pre-extended to 3/2 source. Do not use in-place. |
334 | * We optimize for native, for obvious reasons. */ |
335 | |
336 | U8* |
dea0fc0b |
337 | Perl_utf16_to_utf8(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen) |
a0ed51b3 |
338 | { |
dea0fc0b |
339 | U8* pend; |
340 | U8* dstart = d; |
341 | |
342 | if (bytelen & 1) |
a7867d0a |
343 | Perl_croak(aTHX_ "panic: utf16_to_utf8: odd bytelen"); |
dea0fc0b |
344 | |
345 | pend = p + bytelen; |
346 | |
a0ed51b3 |
347 | while (p < pend) { |
dea0fc0b |
348 | UV uv = (p[0] << 8) + p[1]; /* UTF-16BE */ |
349 | p += 2; |
a0ed51b3 |
350 | if (uv < 0x80) { |
351 | *d++ = uv; |
352 | continue; |
353 | } |
354 | if (uv < 0x800) { |
355 | *d++ = (( uv >> 6) | 0xc0); |
356 | *d++ = (( uv & 0x3f) | 0x80); |
357 | continue; |
358 | } |
359 | if (uv >= 0xd800 && uv < 0xdbff) { /* surrogates */ |
0453d815 |
360 | dTHR; |
dea0fc0b |
361 | UV low = *p++; |
362 | if (low < 0xdc00 || low >= 0xdfff) |
363 | Perl_croak(aTHX_ "Malformed UTF-16 surrogate"); |
a0ed51b3 |
364 | uv = ((uv - 0xd800) << 10) + (low - 0xdc00) + 0x10000; |
365 | } |
366 | if (uv < 0x10000) { |
367 | *d++ = (( uv >> 12) | 0xe0); |
368 | *d++ = (((uv >> 6) & 0x3f) | 0x80); |
369 | *d++ = (( uv & 0x3f) | 0x80); |
370 | continue; |
371 | } |
372 | else { |
373 | *d++ = (( uv >> 18) | 0xf0); |
374 | *d++ = (((uv >> 12) & 0x3f) | 0x80); |
375 | *d++ = (((uv >> 6) & 0x3f) | 0x80); |
376 | *d++ = (( uv & 0x3f) | 0x80); |
377 | continue; |
378 | } |
379 | } |
dea0fc0b |
380 | *newlen = d - dstart; |
a0ed51b3 |
381 | return d; |
382 | } |
383 | |
384 | /* Note: this one is slightly destructive of the source. */ |
385 | |
386 | U8* |
dea0fc0b |
387 | Perl_utf16_to_utf8_reversed(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen) |
a0ed51b3 |
388 | { |
389 | U8* s = (U8*)p; |
390 | U8* send = s + bytelen; |
391 | while (s < send) { |
392 | U8 tmp = s[0]; |
393 | s[0] = s[1]; |
394 | s[1] = tmp; |
395 | s += 2; |
396 | } |
dea0fc0b |
397 | return utf16_to_utf8(p, d, bytelen, newlen); |
a0ed51b3 |
398 | } |
399 | |
400 | /* for now these are all defined (inefficiently) in terms of the utf8 versions */ |
401 | |
402 | bool |
864dbfa3 |
403 | Perl_is_uni_alnum(pTHX_ U32 c) |
a0ed51b3 |
404 | { |
aa6ffa16 |
405 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
406 | uv_to_utf8(tmpbuf, (UV)c); |
407 | return is_utf8_alnum(tmpbuf); |
408 | } |
409 | |
410 | bool |
b8c5462f |
411 | Perl_is_uni_alnumc(pTHX_ U32 c) |
412 | { |
aa6ffa16 |
413 | U8 tmpbuf[UTF8_MAXLEN]; |
b8c5462f |
414 | uv_to_utf8(tmpbuf, (UV)c); |
415 | return is_utf8_alnumc(tmpbuf); |
416 | } |
417 | |
418 | bool |
864dbfa3 |
419 | Perl_is_uni_idfirst(pTHX_ U32 c) |
a0ed51b3 |
420 | { |
aa6ffa16 |
421 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
422 | uv_to_utf8(tmpbuf, (UV)c); |
423 | return is_utf8_idfirst(tmpbuf); |
424 | } |
425 | |
426 | bool |
864dbfa3 |
427 | Perl_is_uni_alpha(pTHX_ U32 c) |
a0ed51b3 |
428 | { |
aa6ffa16 |
429 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
430 | uv_to_utf8(tmpbuf, (UV)c); |
431 | return is_utf8_alpha(tmpbuf); |
432 | } |
433 | |
434 | bool |
4d61ec05 |
435 | Perl_is_uni_ascii(pTHX_ U32 c) |
436 | { |
aa6ffa16 |
437 | U8 tmpbuf[UTF8_MAXLEN]; |
4d61ec05 |
438 | uv_to_utf8(tmpbuf, (UV)c); |
439 | return is_utf8_ascii(tmpbuf); |
440 | } |
441 | |
442 | bool |
864dbfa3 |
443 | Perl_is_uni_space(pTHX_ U32 c) |
a0ed51b3 |
444 | { |
aa6ffa16 |
445 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
446 | uv_to_utf8(tmpbuf, (UV)c); |
447 | return is_utf8_space(tmpbuf); |
448 | } |
449 | |
450 | bool |
864dbfa3 |
451 | Perl_is_uni_digit(pTHX_ U32 c) |
a0ed51b3 |
452 | { |
aa6ffa16 |
453 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
454 | uv_to_utf8(tmpbuf, (UV)c); |
455 | return is_utf8_digit(tmpbuf); |
456 | } |
457 | |
458 | bool |
864dbfa3 |
459 | Perl_is_uni_upper(pTHX_ U32 c) |
a0ed51b3 |
460 | { |
aa6ffa16 |
461 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
462 | uv_to_utf8(tmpbuf, (UV)c); |
463 | return is_utf8_upper(tmpbuf); |
464 | } |
465 | |
466 | bool |
864dbfa3 |
467 | Perl_is_uni_lower(pTHX_ U32 c) |
a0ed51b3 |
468 | { |
aa6ffa16 |
469 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
470 | uv_to_utf8(tmpbuf, (UV)c); |
471 | return is_utf8_lower(tmpbuf); |
472 | } |
473 | |
474 | bool |
b8c5462f |
475 | Perl_is_uni_cntrl(pTHX_ U32 c) |
476 | { |
aa6ffa16 |
477 | U8 tmpbuf[UTF8_MAXLEN]; |
b8c5462f |
478 | uv_to_utf8(tmpbuf, (UV)c); |
479 | return is_utf8_cntrl(tmpbuf); |
480 | } |
481 | |
482 | bool |
483 | Perl_is_uni_graph(pTHX_ U32 c) |
484 | { |
aa6ffa16 |
485 | U8 tmpbuf[UTF8_MAXLEN]; |
b8c5462f |
486 | uv_to_utf8(tmpbuf, (UV)c); |
487 | return is_utf8_graph(tmpbuf); |
488 | } |
489 | |
490 | bool |
864dbfa3 |
491 | Perl_is_uni_print(pTHX_ U32 c) |
a0ed51b3 |
492 | { |
aa6ffa16 |
493 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
494 | uv_to_utf8(tmpbuf, (UV)c); |
495 | return is_utf8_print(tmpbuf); |
496 | } |
497 | |
b8c5462f |
498 | bool |
f248d071 |
499 | Perl_is_uni_punct(pTHX_ U32 c) |
b8c5462f |
500 | { |
aa6ffa16 |
501 | U8 tmpbuf[UTF8_MAXLEN]; |
b8c5462f |
502 | uv_to_utf8(tmpbuf, (UV)c); |
503 | return is_utf8_punct(tmpbuf); |
504 | } |
505 | |
4d61ec05 |
506 | bool |
507 | Perl_is_uni_xdigit(pTHX_ U32 c) |
508 | { |
aa6ffa16 |
509 | U8 tmpbuf[UTF8_MAXLEN]; |
4d61ec05 |
510 | uv_to_utf8(tmpbuf, (UV)c); |
511 | return is_utf8_xdigit(tmpbuf); |
512 | } |
513 | |
a0ed51b3 |
514 | U32 |
864dbfa3 |
515 | Perl_to_uni_upper(pTHX_ U32 c) |
a0ed51b3 |
516 | { |
aa6ffa16 |
517 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
518 | uv_to_utf8(tmpbuf, (UV)c); |
519 | return to_utf8_upper(tmpbuf); |
520 | } |
521 | |
522 | U32 |
864dbfa3 |
523 | Perl_to_uni_title(pTHX_ U32 c) |
a0ed51b3 |
524 | { |
aa6ffa16 |
525 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
526 | uv_to_utf8(tmpbuf, (UV)c); |
527 | return to_utf8_title(tmpbuf); |
528 | } |
529 | |
530 | U32 |
864dbfa3 |
531 | Perl_to_uni_lower(pTHX_ U32 c) |
a0ed51b3 |
532 | { |
aa6ffa16 |
533 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
534 | uv_to_utf8(tmpbuf, (UV)c); |
535 | return to_utf8_lower(tmpbuf); |
536 | } |
537 | |
538 | /* for now these all assume no locale info available for Unicode > 255 */ |
539 | |
540 | bool |
864dbfa3 |
541 | Perl_is_uni_alnum_lc(pTHX_ U32 c) |
a0ed51b3 |
542 | { |
543 | return is_uni_alnum(c); /* XXX no locale support yet */ |
544 | } |
545 | |
546 | bool |
b8c5462f |
547 | Perl_is_uni_alnumc_lc(pTHX_ U32 c) |
548 | { |
549 | return is_uni_alnumc(c); /* XXX no locale support yet */ |
550 | } |
551 | |
552 | bool |
864dbfa3 |
553 | Perl_is_uni_idfirst_lc(pTHX_ U32 c) |
a0ed51b3 |
554 | { |
555 | return is_uni_idfirst(c); /* XXX no locale support yet */ |
556 | } |
557 | |
558 | bool |
864dbfa3 |
559 | Perl_is_uni_alpha_lc(pTHX_ U32 c) |
a0ed51b3 |
560 | { |
561 | return is_uni_alpha(c); /* XXX no locale support yet */ |
562 | } |
563 | |
564 | bool |
4d61ec05 |
565 | Perl_is_uni_ascii_lc(pTHX_ U32 c) |
566 | { |
567 | return is_uni_ascii(c); /* XXX no locale support yet */ |
568 | } |
569 | |
570 | bool |
864dbfa3 |
571 | Perl_is_uni_space_lc(pTHX_ U32 c) |
a0ed51b3 |
572 | { |
573 | return is_uni_space(c); /* XXX no locale support yet */ |
574 | } |
575 | |
576 | bool |
864dbfa3 |
577 | Perl_is_uni_digit_lc(pTHX_ U32 c) |
a0ed51b3 |
578 | { |
579 | return is_uni_digit(c); /* XXX no locale support yet */ |
580 | } |
581 | |
582 | bool |
864dbfa3 |
583 | Perl_is_uni_upper_lc(pTHX_ U32 c) |
a0ed51b3 |
584 | { |
585 | return is_uni_upper(c); /* XXX no locale support yet */ |
586 | } |
587 | |
588 | bool |
864dbfa3 |
589 | Perl_is_uni_lower_lc(pTHX_ U32 c) |
a0ed51b3 |
590 | { |
591 | return is_uni_lower(c); /* XXX no locale support yet */ |
592 | } |
593 | |
594 | bool |
b8c5462f |
595 | Perl_is_uni_cntrl_lc(pTHX_ U32 c) |
596 | { |
597 | return is_uni_cntrl(c); /* XXX no locale support yet */ |
598 | } |
599 | |
600 | bool |
601 | Perl_is_uni_graph_lc(pTHX_ U32 c) |
602 | { |
603 | return is_uni_graph(c); /* XXX no locale support yet */ |
604 | } |
605 | |
606 | bool |
864dbfa3 |
607 | Perl_is_uni_print_lc(pTHX_ U32 c) |
a0ed51b3 |
608 | { |
609 | return is_uni_print(c); /* XXX no locale support yet */ |
610 | } |
611 | |
b8c5462f |
612 | bool |
613 | Perl_is_uni_punct_lc(pTHX_ U32 c) |
614 | { |
615 | return is_uni_punct(c); /* XXX no locale support yet */ |
616 | } |
617 | |
4d61ec05 |
618 | bool |
619 | Perl_is_uni_xdigit_lc(pTHX_ U32 c) |
620 | { |
621 | return is_uni_xdigit(c); /* XXX no locale support yet */ |
622 | } |
623 | |
a0ed51b3 |
624 | U32 |
864dbfa3 |
625 | Perl_to_uni_upper_lc(pTHX_ U32 c) |
a0ed51b3 |
626 | { |
627 | return to_uni_upper(c); /* XXX no locale support yet */ |
628 | } |
629 | |
630 | U32 |
864dbfa3 |
631 | Perl_to_uni_title_lc(pTHX_ U32 c) |
a0ed51b3 |
632 | { |
633 | return to_uni_title(c); /* XXX no locale support yet */ |
634 | } |
635 | |
636 | U32 |
864dbfa3 |
637 | Perl_to_uni_lower_lc(pTHX_ U32 c) |
a0ed51b3 |
638 | { |
639 | return to_uni_lower(c); /* XXX no locale support yet */ |
640 | } |
641 | |
a0ed51b3 |
642 | bool |
864dbfa3 |
643 | Perl_is_utf8_alnum(pTHX_ U8 *p) |
a0ed51b3 |
644 | { |
386d01d6 |
645 | if (!is_utf8_char(p)) |
646 | return FALSE; |
a0ed51b3 |
647 | if (!PL_utf8_alnum) |
e24b16f9 |
648 | PL_utf8_alnum = swash_init("utf8", "IsAlnum", &PL_sv_undef, 0, 0); |
a0ed51b3 |
649 | return swash_fetch(PL_utf8_alnum, p); |
650 | /* return *p == '_' || is_utf8_alpha(p) || is_utf8_digit(p); */ |
651 | #ifdef SURPRISINGLY_SLOWER /* probably because alpha is usually true */ |
652 | if (!PL_utf8_alnum) |
653 | PL_utf8_alnum = swash_init("utf8", "", |
654 | sv_2mortal(newSVpv("+utf8::IsAlpha\n+utf8::IsDigit\n005F\n",0)), 0, 0); |
655 | return swash_fetch(PL_utf8_alnum, p); |
656 | #endif |
657 | } |
658 | |
659 | bool |
b8c5462f |
660 | Perl_is_utf8_alnumc(pTHX_ U8 *p) |
661 | { |
386d01d6 |
662 | if (!is_utf8_char(p)) |
663 | return FALSE; |
b8c5462f |
664 | if (!PL_utf8_alnum) |
665 | PL_utf8_alnum = swash_init("utf8", "IsAlnumC", &PL_sv_undef, 0, 0); |
666 | return swash_fetch(PL_utf8_alnum, p); |
667 | /* return is_utf8_alpha(p) || is_utf8_digit(p); */ |
668 | #ifdef SURPRISINGLY_SLOWER /* probably because alpha is usually true */ |
669 | if (!PL_utf8_alnum) |
670 | PL_utf8_alnum = swash_init("utf8", "", |
671 | sv_2mortal(newSVpv("+utf8::IsAlpha\n+utf8::IsDigit\n005F\n",0)), 0, 0); |
672 | return swash_fetch(PL_utf8_alnum, p); |
673 | #endif |
674 | } |
675 | |
676 | bool |
864dbfa3 |
677 | Perl_is_utf8_idfirst(pTHX_ U8 *p) |
a0ed51b3 |
678 | { |
679 | return *p == '_' || is_utf8_alpha(p); |
680 | } |
681 | |
682 | bool |
864dbfa3 |
683 | Perl_is_utf8_alpha(pTHX_ U8 *p) |
a0ed51b3 |
684 | { |
386d01d6 |
685 | if (!is_utf8_char(p)) |
686 | return FALSE; |
a0ed51b3 |
687 | if (!PL_utf8_alpha) |
e24b16f9 |
688 | PL_utf8_alpha = swash_init("utf8", "IsAlpha", &PL_sv_undef, 0, 0); |
a0ed51b3 |
689 | return swash_fetch(PL_utf8_alpha, p); |
690 | } |
691 | |
692 | bool |
b8c5462f |
693 | Perl_is_utf8_ascii(pTHX_ U8 *p) |
694 | { |
386d01d6 |
695 | if (!is_utf8_char(p)) |
696 | return FALSE; |
b8c5462f |
697 | if (!PL_utf8_ascii) |
698 | PL_utf8_ascii = swash_init("utf8", "IsAscii", &PL_sv_undef, 0, 0); |
699 | return swash_fetch(PL_utf8_ascii, p); |
700 | } |
701 | |
702 | bool |
864dbfa3 |
703 | Perl_is_utf8_space(pTHX_ U8 *p) |
a0ed51b3 |
704 | { |
386d01d6 |
705 | if (!is_utf8_char(p)) |
706 | return FALSE; |
a0ed51b3 |
707 | if (!PL_utf8_space) |
e24b16f9 |
708 | PL_utf8_space = swash_init("utf8", "IsSpace", &PL_sv_undef, 0, 0); |
a0ed51b3 |
709 | return swash_fetch(PL_utf8_space, p); |
710 | } |
711 | |
712 | bool |
864dbfa3 |
713 | Perl_is_utf8_digit(pTHX_ U8 *p) |
a0ed51b3 |
714 | { |
386d01d6 |
715 | if (!is_utf8_char(p)) |
716 | return FALSE; |
a0ed51b3 |
717 | if (!PL_utf8_digit) |
e24b16f9 |
718 | PL_utf8_digit = swash_init("utf8", "IsDigit", &PL_sv_undef, 0, 0); |
a0ed51b3 |
719 | return swash_fetch(PL_utf8_digit, p); |
720 | } |
721 | |
722 | bool |
864dbfa3 |
723 | Perl_is_utf8_upper(pTHX_ U8 *p) |
a0ed51b3 |
724 | { |
386d01d6 |
725 | if (!is_utf8_char(p)) |
726 | return FALSE; |
a0ed51b3 |
727 | if (!PL_utf8_upper) |
e24b16f9 |
728 | PL_utf8_upper = swash_init("utf8", "IsUpper", &PL_sv_undef, 0, 0); |
a0ed51b3 |
729 | return swash_fetch(PL_utf8_upper, p); |
730 | } |
731 | |
732 | bool |
864dbfa3 |
733 | Perl_is_utf8_lower(pTHX_ U8 *p) |
a0ed51b3 |
734 | { |
386d01d6 |
735 | if (!is_utf8_char(p)) |
736 | return FALSE; |
a0ed51b3 |
737 | if (!PL_utf8_lower) |
e24b16f9 |
738 | PL_utf8_lower = swash_init("utf8", "IsLower", &PL_sv_undef, 0, 0); |
a0ed51b3 |
739 | return swash_fetch(PL_utf8_lower, p); |
740 | } |
741 | |
742 | bool |
b8c5462f |
743 | Perl_is_utf8_cntrl(pTHX_ U8 *p) |
744 | { |
386d01d6 |
745 | if (!is_utf8_char(p)) |
746 | return FALSE; |
b8c5462f |
747 | if (!PL_utf8_cntrl) |
748 | PL_utf8_cntrl = swash_init("utf8", "IsCntrl", &PL_sv_undef, 0, 0); |
749 | return swash_fetch(PL_utf8_cntrl, p); |
750 | } |
751 | |
752 | bool |
753 | Perl_is_utf8_graph(pTHX_ U8 *p) |
754 | { |
386d01d6 |
755 | if (!is_utf8_char(p)) |
756 | return FALSE; |
b8c5462f |
757 | if (!PL_utf8_graph) |
758 | PL_utf8_graph = swash_init("utf8", "IsGraph", &PL_sv_undef, 0, 0); |
759 | return swash_fetch(PL_utf8_graph, p); |
760 | } |
761 | |
762 | bool |
864dbfa3 |
763 | Perl_is_utf8_print(pTHX_ U8 *p) |
a0ed51b3 |
764 | { |
386d01d6 |
765 | if (!is_utf8_char(p)) |
766 | return FALSE; |
a0ed51b3 |
767 | if (!PL_utf8_print) |
e24b16f9 |
768 | PL_utf8_print = swash_init("utf8", "IsPrint", &PL_sv_undef, 0, 0); |
a0ed51b3 |
769 | return swash_fetch(PL_utf8_print, p); |
770 | } |
771 | |
772 | bool |
b8c5462f |
773 | Perl_is_utf8_punct(pTHX_ U8 *p) |
774 | { |
386d01d6 |
775 | if (!is_utf8_char(p)) |
776 | return FALSE; |
b8c5462f |
777 | if (!PL_utf8_punct) |
778 | PL_utf8_punct = swash_init("utf8", "IsPunct", &PL_sv_undef, 0, 0); |
779 | return swash_fetch(PL_utf8_punct, p); |
780 | } |
781 | |
782 | bool |
783 | Perl_is_utf8_xdigit(pTHX_ U8 *p) |
784 | { |
386d01d6 |
785 | if (!is_utf8_char(p)) |
786 | return FALSE; |
b8c5462f |
787 | if (!PL_utf8_xdigit) |
788 | PL_utf8_xdigit = swash_init("utf8", "IsXDigit", &PL_sv_undef, 0, 0); |
789 | return swash_fetch(PL_utf8_xdigit, p); |
790 | } |
791 | |
792 | bool |
864dbfa3 |
793 | Perl_is_utf8_mark(pTHX_ U8 *p) |
a0ed51b3 |
794 | { |
386d01d6 |
795 | if (!is_utf8_char(p)) |
796 | return FALSE; |
a0ed51b3 |
797 | if (!PL_utf8_mark) |
e24b16f9 |
798 | PL_utf8_mark = swash_init("utf8", "IsM", &PL_sv_undef, 0, 0); |
a0ed51b3 |
799 | return swash_fetch(PL_utf8_mark, p); |
800 | } |
801 | |
2104c8d9 |
802 | UV |
864dbfa3 |
803 | Perl_to_utf8_upper(pTHX_ U8 *p) |
a0ed51b3 |
804 | { |
805 | UV uv; |
806 | |
807 | if (!PL_utf8_toupper) |
e24b16f9 |
808 | PL_utf8_toupper = swash_init("utf8", "ToUpper", &PL_sv_undef, 4, 0); |
a0ed51b3 |
809 | uv = swash_fetch(PL_utf8_toupper, p); |
810 | return uv ? uv : utf8_to_uv(p,0); |
811 | } |
812 | |
2104c8d9 |
813 | UV |
864dbfa3 |
814 | Perl_to_utf8_title(pTHX_ U8 *p) |
a0ed51b3 |
815 | { |
816 | UV uv; |
817 | |
818 | if (!PL_utf8_totitle) |
e24b16f9 |
819 | PL_utf8_totitle = swash_init("utf8", "ToTitle", &PL_sv_undef, 4, 0); |
a0ed51b3 |
820 | uv = swash_fetch(PL_utf8_totitle, p); |
821 | return uv ? uv : utf8_to_uv(p,0); |
822 | } |
823 | |
2104c8d9 |
824 | UV |
864dbfa3 |
825 | Perl_to_utf8_lower(pTHX_ U8 *p) |
a0ed51b3 |
826 | { |
827 | UV uv; |
828 | |
829 | if (!PL_utf8_tolower) |
e24b16f9 |
830 | PL_utf8_tolower = swash_init("utf8", "ToLower", &PL_sv_undef, 4, 0); |
a0ed51b3 |
831 | uv = swash_fetch(PL_utf8_tolower, p); |
832 | return uv ? uv : utf8_to_uv(p,0); |
833 | } |
834 | |
835 | /* a "swash" is a swatch hash */ |
836 | |
837 | SV* |
864dbfa3 |
838 | Perl_swash_init(pTHX_ char* pkg, char* name, SV *listsv, I32 minbits, I32 none) |
a0ed51b3 |
839 | { |
840 | SV* retval; |
841 | char tmpbuf[256]; |
842 | dSP; |
ce3b816e |
843 | |
844 | if (!gv_stashpv(pkg, 0)) { /* demand load utf8 */ |
845 | ENTER; |
846 | Perl_load_module(aTHX_ PERL_LOADMOD_NOIMPORT, newSVpv(pkg,0), Nullsv); |
847 | LEAVE; |
848 | } |
849 | SPAGAIN; |
a0ed51b3 |
850 | PUSHSTACKi(PERLSI_MAGIC); |
851 | PUSHMARK(SP); |
852 | EXTEND(SP,5); |
853 | PUSHs(sv_2mortal(newSVpvn(pkg, strlen(pkg)))); |
854 | PUSHs(sv_2mortal(newSVpvn(name, strlen(name)))); |
855 | PUSHs(listsv); |
856 | PUSHs(sv_2mortal(newSViv(minbits))); |
857 | PUSHs(sv_2mortal(newSViv(none))); |
858 | PUTBACK; |
859 | ENTER; |
860 | SAVEI32(PL_hints); |
861 | PL_hints = 0; |
862 | save_re_context(); |
e24b16f9 |
863 | if (PL_curcop == &PL_compiling) /* XXX ought to be handled by lex_start */ |
a0ed51b3 |
864 | strncpy(tmpbuf, PL_tokenbuf, sizeof tmpbuf); |
864dbfa3 |
865 | if (call_method("SWASHNEW", G_SCALAR)) |
e24b16f9 |
866 | retval = newSVsv(*PL_stack_sp--); |
a0ed51b3 |
867 | else |
e24b16f9 |
868 | retval = &PL_sv_undef; |
a0ed51b3 |
869 | LEAVE; |
870 | POPSTACK; |
e24b16f9 |
871 | if (PL_curcop == &PL_compiling) { |
a0ed51b3 |
872 | strncpy(PL_tokenbuf, tmpbuf, sizeof tmpbuf); |
e24b16f9 |
873 | PL_curcop->op_private = PL_hints; |
a0ed51b3 |
874 | } |
875 | if (!SvROK(retval) || SvTYPE(SvRV(retval)) != SVt_PVHV) |
cea2e8a9 |
876 | Perl_croak(aTHX_ "SWASHNEW didn't return an HV ref"); |
a0ed51b3 |
877 | return retval; |
878 | } |
879 | |
880 | UV |
864dbfa3 |
881 | Perl_swash_fetch(pTHX_ SV *sv, U8 *ptr) |
a0ed51b3 |
882 | { |
883 | HV* hv = (HV*)SvRV(sv); |
884 | U32 klen = UTF8SKIP(ptr) - 1; |
885 | U32 off = ptr[klen] & 127; /* NB: 64 bit always 0 when len > 1 */ |
886 | STRLEN slen; |
887 | STRLEN needents = (klen ? 64 : 128); |
dfe13c55 |
888 | U8 *tmps; |
a0ed51b3 |
889 | U32 bit; |
890 | SV *retval; |
891 | |
892 | /* |
893 | * This single-entry cache saves about 1/3 of the utf8 overhead in test |
894 | * suite. (That is, only 7-8% overall over just a hash cache. Still, |
895 | * it's nothing to sniff at.) Pity we usually come through at least |
896 | * two function calls to get here... |
897 | * |
898 | * NB: this code assumes that swatches are never modified, once generated! |
899 | */ |
900 | |
901 | if (hv == PL_last_swash_hv && |
902 | klen == PL_last_swash_klen && |
12ae5dfc |
903 | (!klen || memEQ((char *)ptr,(char *)PL_last_swash_key,klen)) ) |
a0ed51b3 |
904 | { |
905 | tmps = PL_last_swash_tmps; |
906 | slen = PL_last_swash_slen; |
907 | } |
908 | else { |
909 | /* Try our second-level swatch cache, kept in a hash. */ |
dfe13c55 |
910 | SV** svp = hv_fetch(hv, (char*)ptr, klen, FALSE); |
a0ed51b3 |
911 | |
912 | /* If not cached, generate it via utf8::SWASHGET */ |
dfe13c55 |
913 | if (!svp || !SvPOK(*svp) || !(tmps = (U8*)SvPV(*svp, slen))) { |
a0ed51b3 |
914 | dSP; |
915 | ENTER; |
916 | SAVETMPS; |
917 | save_re_context(); |
918 | PUSHSTACKi(PERLSI_MAGIC); |
919 | PUSHMARK(SP); |
920 | EXTEND(SP,3); |
921 | PUSHs((SV*)sv); |
922 | PUSHs(sv_2mortal(newSViv(utf8_to_uv(ptr, 0) & ~(needents - 1)))); |
923 | PUSHs(sv_2mortal(newSViv(needents))); |
924 | PUTBACK; |
864dbfa3 |
925 | if (call_method("SWASHGET", G_SCALAR)) |
e24b16f9 |
926 | retval = newSVsv(*PL_stack_sp--); |
a0ed51b3 |
927 | else |
e24b16f9 |
928 | retval = &PL_sv_undef; |
a0ed51b3 |
929 | POPSTACK; |
930 | FREETMPS; |
931 | LEAVE; |
e24b16f9 |
932 | if (PL_curcop == &PL_compiling) |
933 | PL_curcop->op_private = PL_hints; |
a0ed51b3 |
934 | |
dfe13c55 |
935 | svp = hv_store(hv, (char*)ptr, klen, retval, 0); |
a0ed51b3 |
936 | |
dfe13c55 |
937 | if (!svp || !(tmps = (U8*)SvPV(*svp, slen)) || slen < 8) |
cea2e8a9 |
938 | Perl_croak(aTHX_ "SWASHGET didn't return result of proper length"); |
a0ed51b3 |
939 | } |
940 | |
941 | PL_last_swash_hv = hv; |
942 | PL_last_swash_klen = klen; |
943 | PL_last_swash_tmps = tmps; |
944 | PL_last_swash_slen = slen; |
945 | if (klen) |
946 | Copy(ptr, PL_last_swash_key, klen, U8); |
947 | } |
948 | |
949 | switch ((slen << 3) / needents) { |
950 | case 1: |
951 | bit = 1 << (off & 7); |
952 | off >>= 3; |
953 | return (tmps[off] & bit) != 0; |
954 | case 8: |
955 | return tmps[off]; |
956 | case 16: |
957 | off <<= 1; |
958 | return (tmps[off] << 8) + tmps[off + 1] ; |
959 | case 32: |
960 | off <<= 2; |
961 | return (tmps[off] << 24) + (tmps[off+1] << 16) + (tmps[off+2] << 8) + tmps[off + 3] ; |
962 | } |
cea2e8a9 |
963 | Perl_croak(aTHX_ "panic: swash_fetch"); |
a0ed51b3 |
964 | return 0; |
965 | } |