Commit | Line | Data |
a0ed51b3 |
1 | /* utf8.c |
2 | * |
0e06870b |
3 | * Copyright (c) 1998-2001, 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 * |
0e06870b |
30 | Perl_uv_to_utf8(pTHX_ U8 *d, UV uv) /* the d must be UTF8_MAXLEN+1 deep */ |
a0ed51b3 |
31 | { |
32 | if (uv < 0x80) { |
33 | *d++ = uv; |
0e06870b |
34 | *d = 0; |
a0ed51b3 |
35 | return d; |
36 | } |
37 | if (uv < 0x800) { |
38 | *d++ = (( uv >> 6) | 0xc0); |
39 | *d++ = (( uv & 0x3f) | 0x80); |
0e06870b |
40 | *d = 0; |
a0ed51b3 |
41 | return d; |
42 | } |
43 | if (uv < 0x10000) { |
44 | *d++ = (( uv >> 12) | 0xe0); |
45 | *d++ = (((uv >> 6) & 0x3f) | 0x80); |
46 | *d++ = (( uv & 0x3f) | 0x80); |
0e06870b |
47 | *d = 0; |
a0ed51b3 |
48 | return d; |
49 | } |
50 | if (uv < 0x200000) { |
51 | *d++ = (( uv >> 18) | 0xf0); |
52 | *d++ = (((uv >> 12) & 0x3f) | 0x80); |
53 | *d++ = (((uv >> 6) & 0x3f) | 0x80); |
54 | *d++ = (( uv & 0x3f) | 0x80); |
0e06870b |
55 | *d = 0; |
a0ed51b3 |
56 | return d; |
57 | } |
58 | if (uv < 0x4000000) { |
59 | *d++ = (( uv >> 24) | 0xf8); |
60 | *d++ = (((uv >> 18) & 0x3f) | 0x80); |
61 | *d++ = (((uv >> 12) & 0x3f) | 0x80); |
62 | *d++ = (((uv >> 6) & 0x3f) | 0x80); |
63 | *d++ = (( uv & 0x3f) | 0x80); |
0e06870b |
64 | *d = 0; |
a0ed51b3 |
65 | return d; |
66 | } |
67 | if (uv < 0x80000000) { |
68 | *d++ = (( uv >> 30) | 0xfc); |
69 | *d++ = (((uv >> 24) & 0x3f) | 0x80); |
70 | *d++ = (((uv >> 18) & 0x3f) | 0x80); |
71 | *d++ = (((uv >> 12) & 0x3f) | 0x80); |
72 | *d++ = (((uv >> 6) & 0x3f) | 0x80); |
73 | *d++ = (( uv & 0x3f) | 0x80); |
0e06870b |
74 | *d = 0; |
a0ed51b3 |
75 | return d; |
76 | } |
6b8eaf93 |
77 | #ifdef HAS_QUAD |
0e06870b |
78 | if (uv < UTF8_QUAD_MAX) |
a0ed51b3 |
79 | #endif |
80 | { |
81 | *d++ = 0xfe; /* Can't match U+FEFF! */ |
82 | *d++ = (((uv >> 30) & 0x3f) | 0x80); |
83 | *d++ = (((uv >> 24) & 0x3f) | 0x80); |
84 | *d++ = (((uv >> 18) & 0x3f) | 0x80); |
85 | *d++ = (((uv >> 12) & 0x3f) | 0x80); |
86 | *d++ = (((uv >> 6) & 0x3f) | 0x80); |
87 | *d++ = (( uv & 0x3f) | 0x80); |
0e06870b |
88 | *d = 0; |
a0ed51b3 |
89 | return d; |
90 | } |
6b8eaf93 |
91 | #ifdef HAS_QUAD |
a0ed51b3 |
92 | { |
93 | *d++ = 0xff; /* Can't match U+FFFE! */ |
3c77ea2b |
94 | *d++ = 0x80; /* 6 Reserved bits */ |
95 | *d++ = (((uv >> 60) & 0x0f) | 0x80); /* 2 Reserved bits */ |
96 | *d++ = (((uv >> 54) & 0x3f) | 0x80); |
97 | *d++ = (((uv >> 48) & 0x3f) | 0x80); |
98 | *d++ = (((uv >> 42) & 0x3f) | 0x80); |
a0ed51b3 |
99 | *d++ = (((uv >> 36) & 0x3f) | 0x80); |
100 | *d++ = (((uv >> 30) & 0x3f) | 0x80); |
101 | *d++ = (((uv >> 24) & 0x3f) | 0x80); |
102 | *d++ = (((uv >> 18) & 0x3f) | 0x80); |
103 | *d++ = (((uv >> 12) & 0x3f) | 0x80); |
104 | *d++ = (((uv >> 6) & 0x3f) | 0x80); |
105 | *d++ = (( uv & 0x3f) | 0x80); |
0e06870b |
106 | *d = 0; |
a0ed51b3 |
107 | return d; |
108 | } |
109 | #endif |
110 | } |
111 | |
ee8c7f54 |
112 | /* Tests if some arbitrary number of bytes begins in a valid UTF-8 character. |
113 | * The actual number of bytes in the UTF-8 character will be returned if it |
114 | * is valid, otherwise 0. */ |
0e06870b |
115 | STRLEN |
ee8c7f54 |
116 | Perl_is_utf8_char(pTHX_ U8 *s) |
117 | { |
118 | U8 u = *s; |
0e06870b |
119 | STRLEN slen, len; |
120 | UV uv, ouv; |
ee8c7f54 |
121 | |
0e06870b |
122 | if (UTF8_IS_ASCII(u)) |
ee8c7f54 |
123 | return 1; |
124 | |
0e06870b |
125 | if (!UTF8_IS_START(u)) |
ee8c7f54 |
126 | return 0; |
127 | |
22d4bb9c |
128 | len = UTF8SKIP(s); |
ee8c7f54 |
129 | |
0e06870b |
130 | if (len < 2 || !UTF8_IS_CONTINUATION(s[1])) |
131 | return 0; |
132 | |
ee8c7f54 |
133 | slen = len - 1; |
134 | s++; |
0e06870b |
135 | uv = u; |
136 | ouv = uv; |
ee8c7f54 |
137 | while (slen--) { |
0e06870b |
138 | if (!UTF8_IS_CONTINUATION(*s)) |
ee8c7f54 |
139 | return 0; |
0e06870b |
140 | uv = UTF8_ACCUMULATE(uv, *s); |
141 | if (uv < ouv) |
142 | return 0; |
143 | ouv = uv; |
ee8c7f54 |
144 | s++; |
145 | } |
0e06870b |
146 | |
147 | if (UNISKIP(uv) < len) |
148 | return 0; |
149 | |
ee8c7f54 |
150 | return len; |
151 | } |
152 | |
4b19af01 |
153 | /* |
22d4bb9c |
154 | =for apidoc Am|is_utf8_string|U8 *s|STRLEN len |
4b19af01 |
155 | |
156 | Returns true if first C<len> bytes of the given string form valid a UTF8 |
157 | string, false otherwise. |
158 | |
159 | =cut |
160 | */ |
161 | |
22d4bb9c |
162 | bool |
4b19af01 |
163 | Perl_is_utf8_string(pTHX_ U8 *s, STRLEN len) |
164 | { |
0e06870b |
165 | U8* x = s; |
166 | U8* send; |
167 | STRLEN c; |
168 | |
169 | if (!len) |
170 | len = strlen((char *)s); |
171 | send = s + len; |
172 | |
4b19af01 |
173 | while (x < send) { |
174 | c = is_utf8_char(x); |
0e06870b |
175 | if (!c) |
176 | return FALSE; |
4b19af01 |
177 | x += c; |
4b19af01 |
178 | } |
0e06870b |
179 | if (x != send) |
180 | return FALSE; |
181 | |
182 | return TRUE; |
4b19af01 |
183 | } |
184 | |
22d4bb9c |
185 | /* |
0e06870b |
186 | =for apidoc Am|U8* s|utf8_to_uv|STRLEN curlen|STRLEN *retlen|U32 flags |
22d4bb9c |
187 | |
188 | Returns the character value of the first character in the string C<s> |
0e06870b |
189 | which is assumed to be in UTF8 encoding and no longer than C<curlen>; |
190 | C<retlen> will be set to the length, in bytes, of that character. |
22d4bb9c |
191 | |
192 | If C<s> does not point to a well-formed UTF8 character, the behaviour |
0e06870b |
193 | is dependent on the value of C<flags>: if it contains UTF8_CHECK_ONLY, |
194 | it is assumed that the caller will raise a warning, and this function |
195 | will silently just set C<retlen> to C<-1> and return zero. If the |
196 | C<flags> does not contain UTF8_CHECK_ONLY, warnings about |
197 | malformations will be given, C<retlen> will be set to the expected |
198 | length of the UTF-8 character in bytes, and zero will be returned. |
22d4bb9c |
199 | |
0e06870b |
200 | The C<flags> can also contain various flags to allow deviations from |
201 | the strict UTF-8 encoding (see F<utf8.h>). |
202 | |
203 | =cut */ |
22d4bb9c |
204 | |
a0ed51b3 |
205 | UV |
0e06870b |
206 | Perl_utf8_to_uv(pTHX_ U8* s, STRLEN curlen, STRLEN* retlen, U32 flags) |
207 | { |
208 | UV uv = *s, ouv; |
209 | STRLEN len = 1; |
210 | #ifdef EBCDIC |
211 | bool dowarn = 0; |
212 | #else |
213 | bool dowarn = ckWARN_d(WARN_UTF8); |
214 | #endif |
215 | STRLEN expectlen = 0; |
216 | U32 warning = 0; |
217 | |
218 | /* This list is a superset of the UTF8_ALLOW_XXX. */ |
219 | |
220 | #define UTF8_WARN_EMPTY 1 |
221 | #define UTF8_WARN_CONTINUATION 2 |
222 | #define UTF8_WARN_NON_CONTINUATION 3 |
223 | #define UTF8_WARN_FE_FF 4 |
224 | #define UTF8_WARN_SHORT 5 |
225 | #define UTF8_WARN_OVERFLOW 6 |
226 | #define UTF8_WARN_SURROGATE 7 |
227 | #define UTF8_WARN_BOM 8 |
228 | #define UTF8_WARN_LONG 9 |
229 | #define UTF8_WARN_FFFF 10 |
230 | |
231 | if (curlen == 0 && |
232 | !(flags & UTF8_ALLOW_EMPTY)) { |
233 | warning = UTF8_WARN_EMPTY; |
234 | goto malformed; |
a0ed51b3 |
235 | } |
22d4bb9c |
236 | |
0e06870b |
237 | if (UTF8_IS_ASCII(uv)) { |
a0ed51b3 |
238 | if (retlen) |
239 | *retlen = 1; |
240 | return *s; |
241 | } |
242 | |
0e06870b |
243 | if (UTF8_IS_CONTINUATION(uv) && |
244 | !(flags & UTF8_ALLOW_CONTINUATION)) { |
245 | warning = UTF8_WARN_CONTINUATION; |
246 | goto malformed; |
247 | } |
a0ed51b3 |
248 | |
0e06870b |
249 | if (UTF8_IS_START(uv) && curlen > 1 && !UTF8_IS_CONTINUATION(s[1]) && |
250 | !(flags & UTF8_ALLOW_NON_CONTINUATION)) { |
251 | warning = UTF8_WARN_NON_CONTINUATION; |
252 | goto malformed; |
253 | } |
254 | |
255 | if ((uv == 0xfe || uv == 0xff) && |
256 | !(flags & UTF8_ALLOW_FE_FF)) { |
257 | warning = UTF8_WARN_FE_FF; |
258 | goto malformed; |
259 | } |
260 | |
261 | if (!(uv & 0x20)) { len = 2; uv &= 0x1f; } |
262 | else if (!(uv & 0x10)) { len = 3; uv &= 0x0f; } |
263 | else if (!(uv & 0x08)) { len = 4; uv &= 0x07; } |
264 | else if (!(uv & 0x04)) { len = 5; uv &= 0x03; } |
265 | else if (!(uv & 0x02)) { len = 6; uv &= 0x01; } |
266 | else if (!(uv & 0x01)) { len = 7; uv = 0; } |
267 | else { len = 13; uv = 0; } /* whoa! */ |
268 | |
a0ed51b3 |
269 | if (retlen) |
270 | *retlen = len; |
0e06870b |
271 | |
272 | expectlen = len; |
273 | |
274 | if ((curlen < expectlen) && |
275 | !(flags & UTF8_ALLOW_SHORT)) { |
276 | warning = UTF8_WARN_SHORT; |
277 | goto malformed; |
278 | } |
279 | |
280 | len--; |
a0ed51b3 |
281 | s++; |
0e06870b |
282 | ouv = uv; |
283 | |
a0ed51b3 |
284 | while (len--) { |
0e06870b |
285 | if (!UTF8_IS_CONTINUATION(*s) && |
286 | !(flags & UTF8_ALLOW_NON_CONTINUATION)) { |
287 | s--; |
288 | warning = UTF8_WARN_NON_CONTINUATION; |
289 | goto malformed; |
a0ed51b3 |
290 | } |
291 | else |
0e06870b |
292 | uv = UTF8_ACCUMULATE(uv, *s); |
293 | if (!(uv > ouv)) { |
294 | /* These cannot be allowed. */ |
295 | if (uv == ouv) { |
296 | if (!(flags & UTF8_ALLOW_LONG)) { |
297 | warning = UTF8_WARN_LONG; |
298 | goto malformed; |
299 | } |
300 | } |
301 | else { /* uv < ouv */ |
302 | /* This cannot be allowed. */ |
303 | warning = UTF8_WARN_OVERFLOW; |
304 | goto malformed; |
305 | } |
306 | } |
307 | s++; |
308 | ouv = uv; |
a0ed51b3 |
309 | } |
0e06870b |
310 | |
311 | if (UNICODE_IS_SURROGATE(uv) && |
312 | !(flags & UTF8_ALLOW_SURROGATE)) { |
313 | warning = UTF8_WARN_SURROGATE; |
314 | goto malformed; |
315 | } else if (UNICODE_IS_BYTE_ORDER_MARK(uv) && |
316 | !(flags & UTF8_ALLOW_BOM)) { |
317 | warning = UTF8_WARN_BOM; |
318 | goto malformed; |
319 | } else if ((expectlen > UNISKIP(uv)) && |
320 | !(flags & UTF8_ALLOW_LONG)) { |
321 | warning = UTF8_WARN_LONG; |
322 | goto malformed; |
323 | } else if (UNICODE_IS_ILLEGAL(uv) && |
324 | !(flags & UTF8_ALLOW_FFFF)) { |
325 | warning = UTF8_WARN_FFFF; |
326 | goto malformed; |
327 | } |
328 | |
a0ed51b3 |
329 | return uv; |
0e06870b |
330 | |
331 | malformed: |
332 | |
333 | if (flags & UTF8_CHECK_ONLY) { |
334 | if (retlen) |
335 | *retlen = -1; |
336 | return 0; |
337 | } |
338 | |
339 | if (dowarn) { |
340 | SV* sv = sv_2mortal(newSVpv("Malformed UTF-8 character ", 0)); |
341 | |
342 | switch (warning) { |
343 | case 0: /* Intentionally empty. */ break; |
344 | case UTF8_WARN_EMPTY: |
345 | Perl_sv_catpvf(aTHX_ sv, "(empty string)"); |
346 | break; |
347 | case UTF8_WARN_CONTINUATION: |
348 | Perl_sv_catpvf(aTHX_ sv, "(unexpected continuation byte 0x%02"UVxf")", uv); |
349 | break; |
350 | case UTF8_WARN_NON_CONTINUATION: |
351 | Perl_sv_catpvf(aTHX_ sv, "(unexpected non-continuation byte 0x%02"UVxf" after start byte 0x%02"UVxf")", |
352 | (UV)s[1], uv); |
353 | break; |
354 | case UTF8_WARN_FE_FF: |
355 | Perl_sv_catpvf(aTHX_ sv, "(byte 0x%02"UVxf")", uv); |
356 | break; |
357 | case UTF8_WARN_SHORT: |
358 | Perl_sv_catpvf(aTHX_ sv, "(%d byte%s, need %d)", |
359 | curlen, curlen == 1 ? "" : "s", expectlen); |
360 | break; |
361 | case UTF8_WARN_OVERFLOW: |
362 | Perl_sv_catpvf(aTHX_ sv, "(overflow at 0x%"UVxf", byte 0x%02x)", |
363 | ouv, *s); |
364 | break; |
365 | case UTF8_WARN_SURROGATE: |
366 | Perl_sv_catpvf(aTHX_ sv, "(UTF-16 surrogate 0x%04"UVxf")", uv); |
367 | break; |
368 | case UTF8_WARN_BOM: |
369 | Perl_sv_catpvf(aTHX_ sv, "(byte order mark 0x%04"UVxf")", uv); |
370 | break; |
371 | case UTF8_WARN_LONG: |
372 | Perl_sv_catpvf(aTHX_ sv, "(%d byte%s, need %d)", |
373 | expectlen, expectlen == 1 ? "": "s", UNISKIP(uv)); |
374 | break; |
375 | case UTF8_WARN_FFFF: |
376 | Perl_sv_catpvf(aTHX_ sv, "(character 0x%04"UVxf")", uv); |
377 | break; |
378 | default: |
379 | Perl_sv_catpvf(aTHX_ sv, "(unknown reason)"); |
380 | break; |
381 | } |
382 | |
383 | if (warning) { |
384 | char *s = SvPVX(sv); |
385 | |
386 | if (PL_op) |
387 | Perl_warner(aTHX_ WARN_UTF8, |
388 | "%s in %s", s, PL_op_desc[PL_op->op_type]); |
389 | else |
390 | Perl_warner(aTHX_ WARN_UTF8, "%s", s); |
391 | } |
392 | } |
393 | |
394 | if (retlen) |
395 | *retlen = expectlen ? expectlen : len; |
396 | |
397 | return 0; |
a0ed51b3 |
398 | } |
399 | |
22d4bb9c |
400 | /* |
0e06870b |
401 | =for apidoc Am|U8* s|utf8_to_uv_simple|STRLEN *retlen |
22d4bb9c |
402 | |
403 | Returns the character value of the first character in the string C<s> |
404 | which is assumed to be in UTF8 encoding; C<retlen> will be set to the |
0e06870b |
405 | length, in bytes, of that character. |
22d4bb9c |
406 | |
0e06870b |
407 | If C<s> does not point to a well-formed UTF8 character, zero is |
408 | returned and retlen is set, if possible, to -1. |
22d4bb9c |
409 | |
410 | =cut |
411 | */ |
412 | |
413 | UV |
0e06870b |
414 | Perl_utf8_to_uv_simple(pTHX_ U8* s, STRLEN* retlen) |
22d4bb9c |
415 | { |
0e06870b |
416 | return Perl_utf8_to_uv(aTHX_ s, UTF8_MAXLEN, retlen, 0); |
22d4bb9c |
417 | } |
418 | |
0e06870b |
419 | /* |
420 | =for apidoc Am|STRLEN|utf8_length|U8* s|U8 *e |
421 | |
422 | Return the length of the UTF-8 char encoded string C<s> in characters. |
423 | Stops at C<e> (inclusive). If C<e E<lt> s> or if the scan would end |
424 | up past C<e>, croaks. |
425 | |
426 | =cut |
427 | */ |
428 | |
429 | STRLEN |
430 | Perl_utf8_length(pTHX_ U8* s, U8* e) |
431 | { |
432 | STRLEN len = 0; |
433 | |
434 | /* Note: cannot use UTF8_IS_...() too eagerly here since e.g. |
435 | * the bitops (especially ~) can create illegal UTF-8. |
436 | * In other words: in Perl UTF-8 is not just for Unicode. */ |
437 | |
438 | if (e < s) |
439 | Perl_croak(aTHX_ "panic: utf8_length: unexpected end"); |
440 | while (s < e) { |
441 | U8 t = UTF8SKIP(s); |
442 | |
443 | if (e - s < t) |
444 | Perl_croak(aTHX_ "panic: utf8_length: unaligned end"); |
445 | s += t; |
446 | len++; |
447 | } |
448 | |
449 | return len; |
450 | } |
451 | |
452 | /* |
453 | =for apidoc Am|IV|utf8_distance|U8 *a|U8 *b |
454 | |
455 | Returns the number of UTF8 characters between the UTF-8 pointers C<a> |
456 | and C<b>. |
457 | |
458 | WARNING: use only if you *know* that the pointers point inside the |
459 | same UTF-8 buffer. |
460 | |
461 | =cut */ |
a0ed51b3 |
462 | |
0e06870b |
463 | IV |
864dbfa3 |
464 | Perl_utf8_distance(pTHX_ U8 *a, U8 *b) |
a0ed51b3 |
465 | { |
0e06870b |
466 | IV off = 0; |
467 | |
468 | /* Note: cannot use UTF8_IS_...() too eagerly here since e.g. |
469 | * the bitops (especially ~) can create illegal UTF-8. |
470 | * In other words: in Perl UTF-8 is not just for Unicode. */ |
471 | |
a0ed51b3 |
472 | if (a < b) { |
473 | while (a < b) { |
0e06870b |
474 | U8 c = UTF8SKIP(a); |
475 | |
476 | if (b - a < c) |
477 | Perl_croak(aTHX_ "panic: utf8_distance: unaligned end"); |
478 | a += c; |
a0ed51b3 |
479 | off--; |
480 | } |
481 | } |
482 | else { |
483 | while (b < a) { |
0e06870b |
484 | U8 c = UTF8SKIP(b); |
485 | |
486 | if (a - b < c) |
487 | Perl_croak(aTHX_ "panic: utf8_distance: unaligned end"); |
488 | b += c; |
a0ed51b3 |
489 | off++; |
490 | } |
491 | } |
0e06870b |
492 | |
a0ed51b3 |
493 | return off; |
494 | } |
495 | |
0e06870b |
496 | /* |
497 | =for apidoc Am|U8*|utf8_hop|U8 *s|I32 off |
498 | |
499 | Return the UTF-8 pointer C<s> displaced by C<off> characters, either |
500 | forward or backward. |
501 | |
502 | WARNING: do not use the following unless you *know* C<off> is within |
503 | the UTF-8 data pointed to by C<s> *and* that on entry C<s> is aligned |
504 | on the first byte of character or just after the last byte of a character. |
505 | |
506 | =cut */ |
a0ed51b3 |
507 | |
508 | U8 * |
864dbfa3 |
509 | Perl_utf8_hop(pTHX_ U8 *s, I32 off) |
a0ed51b3 |
510 | { |
0e06870b |
511 | /* Note: cannot use UTF8_IS_...() too eagerly here since e.g |
512 | * the bitops (especially ~) can create illegal UTF-8. |
513 | * In other words: in Perl UTF-8 is not just for Unicode. */ |
514 | |
a0ed51b3 |
515 | if (off >= 0) { |
516 | while (off--) |
517 | s += UTF8SKIP(s); |
518 | } |
519 | else { |
520 | while (off++) { |
521 | s--; |
0e06870b |
522 | while (UTF8_IS_CONTINUATION(*s)) |
523 | s--; |
a0ed51b3 |
524 | } |
525 | } |
526 | return s; |
527 | } |
528 | |
a0ed51b3 |
529 | /* |
22d4bb9c |
530 | =for apidoc Am|U8 *|utf8_to_bytes|U8 *s|STRLEN *len |
4b19af01 |
531 | |
22d4bb9c |
532 | Converts a string C<s> of length C<len> from UTF8 into byte encoding. |
533 | Unlike C<bytes_to_utf8>, this over-writes the original string, and |
534 | updates len to contain the new length. |
535 | Returns zero on failure, setting C<len> to -1. |
4b19af01 |
536 | |
537 | =cut |
538 | */ |
539 | |
540 | U8 * |
22d4bb9c |
541 | Perl_utf8_to_bytes(pTHX_ U8* s, STRLEN *len) |
4b19af01 |
542 | { |
4b19af01 |
543 | U8 *send; |
544 | U8 *d; |
0e06870b |
545 | U8 *save = s; |
22d4bb9c |
546 | |
547 | /* ensure valid UTF8 and chars < 256 before updating string */ |
0e06870b |
548 | for (send = s + *len; s < send; ) { |
549 | U8 c = *s++; |
550 | |
22d4bb9c |
551 | if (c >= 0x80 && |
0e06870b |
552 | ((s >= send) || |
553 | ((*s++ & 0xc0) != 0x80) || ((c & 0xfe) != 0xc2))) { |
554 | *len = -1; |
555 | return 0; |
556 | } |
22d4bb9c |
557 | } |
0e06870b |
558 | |
559 | d = s = save; |
4b19af01 |
560 | while (s < send) { |
0e06870b |
561 | STRLEN ulen; |
562 | *d++ = (U8)utf8_to_uv_simple(s, &ulen); |
563 | s += ulen; |
4b19af01 |
564 | } |
565 | *d = '\0'; |
22d4bb9c |
566 | *len = d - save; |
4b19af01 |
567 | return save; |
568 | } |
569 | |
570 | /* |
571 | =for apidoc Am|U8 *|bytes_to_utf8|U8 *s|STRLEN *len |
572 | |
573 | Converts a string C<s> of length C<len> from ASCII into UTF8 encoding. |
574 | Returns a pointer to the newly-created string, and sets C<len> to |
575 | reflect the new length. |
576 | |
577 | =cut |
578 | */ |
579 | |
580 | U8* |
581 | Perl_bytes_to_utf8(pTHX_ U8* s, STRLEN *len) |
582 | { |
4b19af01 |
583 | U8 *send; |
584 | U8 *d; |
585 | U8 *dst; |
586 | send = s + (*len); |
587 | |
588 | Newz(801, d, (*len) * 2 + 1, U8); |
589 | dst = d; |
590 | |
591 | while (s < send) { |
592 | if (*s < 0x80) |
593 | *d++ = *s++; |
594 | else { |
595 | UV uv = *s++; |
596 | *d++ = (( uv >> 6) | 0xc0); |
597 | *d++ = (( uv & 0x3f) | 0x80); |
598 | } |
599 | } |
600 | *d = '\0'; |
601 | *len = d-dst; |
602 | return dst; |
603 | } |
604 | |
605 | /* |
606 | * Convert native (big-endian) or reversed (little-endian) UTF-16 to UTF-8. |
a0ed51b3 |
607 | * |
608 | * Destination must be pre-extended to 3/2 source. Do not use in-place. |
609 | * We optimize for native, for obvious reasons. */ |
610 | |
611 | U8* |
4b19af01 |
612 | Perl_utf16_to_utf8(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen) |
a0ed51b3 |
613 | { |
4b19af01 |
614 | U8* pend; |
615 | U8* dstart = d; |
616 | |
617 | if (bytelen & 1) |
618 | Perl_croak(aTHX_ "panic: utf16_to_utf8: odd bytelen"); |
619 | |
620 | pend = p + bytelen; |
621 | |
a0ed51b3 |
622 | while (p < pend) { |
4b19af01 |
623 | UV uv = (p[0] << 8) + p[1]; /* UTF-16BE */ |
624 | p += 2; |
a0ed51b3 |
625 | if (uv < 0x80) { |
626 | *d++ = uv; |
627 | continue; |
628 | } |
629 | if (uv < 0x800) { |
630 | *d++ = (( uv >> 6) | 0xc0); |
631 | *d++ = (( uv & 0x3f) | 0x80); |
632 | continue; |
633 | } |
634 | if (uv >= 0xd800 && uv < 0xdbff) { /* surrogates */ |
4b19af01 |
635 | UV low = *p++; |
636 | if (low < 0xdc00 || low >= 0xdfff) |
637 | Perl_croak(aTHX_ "Malformed UTF-16 surrogate"); |
a0ed51b3 |
638 | uv = ((uv - 0xd800) << 10) + (low - 0xdc00) + 0x10000; |
639 | } |
640 | if (uv < 0x10000) { |
641 | *d++ = (( uv >> 12) | 0xe0); |
642 | *d++ = (((uv >> 6) & 0x3f) | 0x80); |
643 | *d++ = (( uv & 0x3f) | 0x80); |
644 | continue; |
645 | } |
646 | else { |
647 | *d++ = (( uv >> 18) | 0xf0); |
648 | *d++ = (((uv >> 12) & 0x3f) | 0x80); |
649 | *d++ = (((uv >> 6) & 0x3f) | 0x80); |
650 | *d++ = (( uv & 0x3f) | 0x80); |
651 | continue; |
652 | } |
653 | } |
4b19af01 |
654 | *newlen = d - dstart; |
a0ed51b3 |
655 | return d; |
656 | } |
657 | |
658 | /* Note: this one is slightly destructive of the source. */ |
659 | |
660 | U8* |
4b19af01 |
661 | Perl_utf16_to_utf8_reversed(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen) |
a0ed51b3 |
662 | { |
663 | U8* s = (U8*)p; |
664 | U8* send = s + bytelen; |
665 | while (s < send) { |
666 | U8 tmp = s[0]; |
667 | s[0] = s[1]; |
668 | s[1] = tmp; |
669 | s += 2; |
670 | } |
4b19af01 |
671 | return utf16_to_utf8(p, d, bytelen, newlen); |
a0ed51b3 |
672 | } |
673 | |
674 | /* for now these are all defined (inefficiently) in terms of the utf8 versions */ |
675 | |
676 | bool |
864dbfa3 |
677 | Perl_is_uni_alnum(pTHX_ U32 c) |
a0ed51b3 |
678 | { |
0e06870b |
679 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
680 | uv_to_utf8(tmpbuf, (UV)c); |
681 | return is_utf8_alnum(tmpbuf); |
682 | } |
683 | |
684 | bool |
b8c5462f |
685 | Perl_is_uni_alnumc(pTHX_ U32 c) |
686 | { |
0e06870b |
687 | U8 tmpbuf[UTF8_MAXLEN+1]; |
b8c5462f |
688 | uv_to_utf8(tmpbuf, (UV)c); |
689 | return is_utf8_alnumc(tmpbuf); |
690 | } |
691 | |
692 | bool |
864dbfa3 |
693 | Perl_is_uni_idfirst(pTHX_ U32 c) |
a0ed51b3 |
694 | { |
0e06870b |
695 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
696 | uv_to_utf8(tmpbuf, (UV)c); |
697 | return is_utf8_idfirst(tmpbuf); |
698 | } |
699 | |
700 | bool |
864dbfa3 |
701 | Perl_is_uni_alpha(pTHX_ U32 c) |
a0ed51b3 |
702 | { |
0e06870b |
703 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
704 | uv_to_utf8(tmpbuf, (UV)c); |
705 | return is_utf8_alpha(tmpbuf); |
706 | } |
707 | |
708 | bool |
4d61ec05 |
709 | Perl_is_uni_ascii(pTHX_ U32 c) |
710 | { |
0e06870b |
711 | U8 tmpbuf[UTF8_MAXLEN+1]; |
4d61ec05 |
712 | uv_to_utf8(tmpbuf, (UV)c); |
713 | return is_utf8_ascii(tmpbuf); |
714 | } |
715 | |
716 | bool |
864dbfa3 |
717 | Perl_is_uni_space(pTHX_ U32 c) |
a0ed51b3 |
718 | { |
0e06870b |
719 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
720 | uv_to_utf8(tmpbuf, (UV)c); |
721 | return is_utf8_space(tmpbuf); |
722 | } |
723 | |
724 | bool |
864dbfa3 |
725 | Perl_is_uni_digit(pTHX_ U32 c) |
a0ed51b3 |
726 | { |
0e06870b |
727 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
728 | uv_to_utf8(tmpbuf, (UV)c); |
729 | return is_utf8_digit(tmpbuf); |
730 | } |
731 | |
732 | bool |
864dbfa3 |
733 | Perl_is_uni_upper(pTHX_ U32 c) |
a0ed51b3 |
734 | { |
0e06870b |
735 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
736 | uv_to_utf8(tmpbuf, (UV)c); |
737 | return is_utf8_upper(tmpbuf); |
738 | } |
739 | |
740 | bool |
864dbfa3 |
741 | Perl_is_uni_lower(pTHX_ U32 c) |
a0ed51b3 |
742 | { |
0e06870b |
743 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
744 | uv_to_utf8(tmpbuf, (UV)c); |
745 | return is_utf8_lower(tmpbuf); |
746 | } |
747 | |
748 | bool |
b8c5462f |
749 | Perl_is_uni_cntrl(pTHX_ U32 c) |
750 | { |
0e06870b |
751 | U8 tmpbuf[UTF8_MAXLEN+1]; |
b8c5462f |
752 | uv_to_utf8(tmpbuf, (UV)c); |
753 | return is_utf8_cntrl(tmpbuf); |
754 | } |
755 | |
756 | bool |
757 | Perl_is_uni_graph(pTHX_ U32 c) |
758 | { |
0e06870b |
759 | U8 tmpbuf[UTF8_MAXLEN+1]; |
b8c5462f |
760 | uv_to_utf8(tmpbuf, (UV)c); |
761 | return is_utf8_graph(tmpbuf); |
762 | } |
763 | |
764 | bool |
864dbfa3 |
765 | Perl_is_uni_print(pTHX_ U32 c) |
a0ed51b3 |
766 | { |
0e06870b |
767 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
768 | uv_to_utf8(tmpbuf, (UV)c); |
769 | return is_utf8_print(tmpbuf); |
770 | } |
771 | |
b8c5462f |
772 | bool |
f248d071 |
773 | Perl_is_uni_punct(pTHX_ U32 c) |
b8c5462f |
774 | { |
0e06870b |
775 | U8 tmpbuf[UTF8_MAXLEN+1]; |
b8c5462f |
776 | uv_to_utf8(tmpbuf, (UV)c); |
777 | return is_utf8_punct(tmpbuf); |
778 | } |
779 | |
4d61ec05 |
780 | bool |
781 | Perl_is_uni_xdigit(pTHX_ U32 c) |
782 | { |
0e06870b |
783 | U8 tmpbuf[UTF8_MAXLEN+1]; |
4d61ec05 |
784 | uv_to_utf8(tmpbuf, (UV)c); |
785 | return is_utf8_xdigit(tmpbuf); |
786 | } |
787 | |
a0ed51b3 |
788 | U32 |
864dbfa3 |
789 | Perl_to_uni_upper(pTHX_ U32 c) |
a0ed51b3 |
790 | { |
0e06870b |
791 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
792 | uv_to_utf8(tmpbuf, (UV)c); |
793 | return to_utf8_upper(tmpbuf); |
794 | } |
795 | |
796 | U32 |
864dbfa3 |
797 | Perl_to_uni_title(pTHX_ U32 c) |
a0ed51b3 |
798 | { |
0e06870b |
799 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
800 | uv_to_utf8(tmpbuf, (UV)c); |
801 | return to_utf8_title(tmpbuf); |
802 | } |
803 | |
804 | U32 |
864dbfa3 |
805 | Perl_to_uni_lower(pTHX_ U32 c) |
a0ed51b3 |
806 | { |
0e06870b |
807 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
808 | uv_to_utf8(tmpbuf, (UV)c); |
809 | return to_utf8_lower(tmpbuf); |
810 | } |
811 | |
812 | /* for now these all assume no locale info available for Unicode > 255 */ |
813 | |
814 | bool |
864dbfa3 |
815 | Perl_is_uni_alnum_lc(pTHX_ U32 c) |
a0ed51b3 |
816 | { |
817 | return is_uni_alnum(c); /* XXX no locale support yet */ |
818 | } |
819 | |
820 | bool |
b8c5462f |
821 | Perl_is_uni_alnumc_lc(pTHX_ U32 c) |
822 | { |
823 | return is_uni_alnumc(c); /* XXX no locale support yet */ |
824 | } |
825 | |
826 | bool |
864dbfa3 |
827 | Perl_is_uni_idfirst_lc(pTHX_ U32 c) |
a0ed51b3 |
828 | { |
829 | return is_uni_idfirst(c); /* XXX no locale support yet */ |
830 | } |
831 | |
832 | bool |
864dbfa3 |
833 | Perl_is_uni_alpha_lc(pTHX_ U32 c) |
a0ed51b3 |
834 | { |
835 | return is_uni_alpha(c); /* XXX no locale support yet */ |
836 | } |
837 | |
838 | bool |
4d61ec05 |
839 | Perl_is_uni_ascii_lc(pTHX_ U32 c) |
840 | { |
841 | return is_uni_ascii(c); /* XXX no locale support yet */ |
842 | } |
843 | |
844 | bool |
864dbfa3 |
845 | Perl_is_uni_space_lc(pTHX_ U32 c) |
a0ed51b3 |
846 | { |
847 | return is_uni_space(c); /* XXX no locale support yet */ |
848 | } |
849 | |
850 | bool |
864dbfa3 |
851 | Perl_is_uni_digit_lc(pTHX_ U32 c) |
a0ed51b3 |
852 | { |
853 | return is_uni_digit(c); /* XXX no locale support yet */ |
854 | } |
855 | |
856 | bool |
864dbfa3 |
857 | Perl_is_uni_upper_lc(pTHX_ U32 c) |
a0ed51b3 |
858 | { |
859 | return is_uni_upper(c); /* XXX no locale support yet */ |
860 | } |
861 | |
862 | bool |
864dbfa3 |
863 | Perl_is_uni_lower_lc(pTHX_ U32 c) |
a0ed51b3 |
864 | { |
865 | return is_uni_lower(c); /* XXX no locale support yet */ |
866 | } |
867 | |
868 | bool |
b8c5462f |
869 | Perl_is_uni_cntrl_lc(pTHX_ U32 c) |
870 | { |
871 | return is_uni_cntrl(c); /* XXX no locale support yet */ |
872 | } |
873 | |
874 | bool |
875 | Perl_is_uni_graph_lc(pTHX_ U32 c) |
876 | { |
877 | return is_uni_graph(c); /* XXX no locale support yet */ |
878 | } |
879 | |
880 | bool |
864dbfa3 |
881 | Perl_is_uni_print_lc(pTHX_ U32 c) |
a0ed51b3 |
882 | { |
883 | return is_uni_print(c); /* XXX no locale support yet */ |
884 | } |
885 | |
b8c5462f |
886 | bool |
887 | Perl_is_uni_punct_lc(pTHX_ U32 c) |
888 | { |
889 | return is_uni_punct(c); /* XXX no locale support yet */ |
890 | } |
891 | |
4d61ec05 |
892 | bool |
893 | Perl_is_uni_xdigit_lc(pTHX_ U32 c) |
894 | { |
895 | return is_uni_xdigit(c); /* XXX no locale support yet */ |
896 | } |
897 | |
a0ed51b3 |
898 | U32 |
864dbfa3 |
899 | Perl_to_uni_upper_lc(pTHX_ U32 c) |
a0ed51b3 |
900 | { |
901 | return to_uni_upper(c); /* XXX no locale support yet */ |
902 | } |
903 | |
904 | U32 |
864dbfa3 |
905 | Perl_to_uni_title_lc(pTHX_ U32 c) |
a0ed51b3 |
906 | { |
907 | return to_uni_title(c); /* XXX no locale support yet */ |
908 | } |
909 | |
910 | U32 |
864dbfa3 |
911 | Perl_to_uni_lower_lc(pTHX_ U32 c) |
a0ed51b3 |
912 | { |
913 | return to_uni_lower(c); /* XXX no locale support yet */ |
914 | } |
915 | |
a0ed51b3 |
916 | bool |
864dbfa3 |
917 | Perl_is_utf8_alnum(pTHX_ U8 *p) |
a0ed51b3 |
918 | { |
ee8c7f54 |
919 | if (!is_utf8_char(p)) |
920 | return FALSE; |
a0ed51b3 |
921 | if (!PL_utf8_alnum) |
22d4bb9c |
922 | /* NOTE: "IsWord", not "IsAlnum", since Alnum is a true |
923 | * descendant of isalnum(3), in other words, it doesn't |
924 | * contain the '_'. --jhi */ |
925 | PL_utf8_alnum = swash_init("utf8", "IsWord", &PL_sv_undef, 0, 0); |
a0ed51b3 |
926 | return swash_fetch(PL_utf8_alnum, p); |
927 | /* return *p == '_' || is_utf8_alpha(p) || is_utf8_digit(p); */ |
928 | #ifdef SURPRISINGLY_SLOWER /* probably because alpha is usually true */ |
929 | if (!PL_utf8_alnum) |
930 | PL_utf8_alnum = swash_init("utf8", "", |
931 | sv_2mortal(newSVpv("+utf8::IsAlpha\n+utf8::IsDigit\n005F\n",0)), 0, 0); |
932 | return swash_fetch(PL_utf8_alnum, p); |
933 | #endif |
934 | } |
935 | |
936 | bool |
b8c5462f |
937 | Perl_is_utf8_alnumc(pTHX_ U8 *p) |
938 | { |
ee8c7f54 |
939 | if (!is_utf8_char(p)) |
940 | return FALSE; |
b8c5462f |
941 | if (!PL_utf8_alnum) |
942 | PL_utf8_alnum = swash_init("utf8", "IsAlnumC", &PL_sv_undef, 0, 0); |
943 | return swash_fetch(PL_utf8_alnum, p); |
944 | /* return is_utf8_alpha(p) || is_utf8_digit(p); */ |
945 | #ifdef SURPRISINGLY_SLOWER /* probably because alpha is usually true */ |
946 | if (!PL_utf8_alnum) |
947 | PL_utf8_alnum = swash_init("utf8", "", |
948 | sv_2mortal(newSVpv("+utf8::IsAlpha\n+utf8::IsDigit\n005F\n",0)), 0, 0); |
949 | return swash_fetch(PL_utf8_alnum, p); |
950 | #endif |
951 | } |
952 | |
953 | bool |
864dbfa3 |
954 | Perl_is_utf8_idfirst(pTHX_ U8 *p) |
a0ed51b3 |
955 | { |
956 | return *p == '_' || is_utf8_alpha(p); |
957 | } |
958 | |
959 | bool |
864dbfa3 |
960 | Perl_is_utf8_alpha(pTHX_ U8 *p) |
a0ed51b3 |
961 | { |
ee8c7f54 |
962 | if (!is_utf8_char(p)) |
963 | return FALSE; |
a0ed51b3 |
964 | if (!PL_utf8_alpha) |
e24b16f9 |
965 | PL_utf8_alpha = swash_init("utf8", "IsAlpha", &PL_sv_undef, 0, 0); |
a0ed51b3 |
966 | return swash_fetch(PL_utf8_alpha, p); |
967 | } |
968 | |
969 | bool |
b8c5462f |
970 | Perl_is_utf8_ascii(pTHX_ U8 *p) |
971 | { |
ee8c7f54 |
972 | if (!is_utf8_char(p)) |
973 | return FALSE; |
b8c5462f |
974 | if (!PL_utf8_ascii) |
975 | PL_utf8_ascii = swash_init("utf8", "IsAscii", &PL_sv_undef, 0, 0); |
976 | return swash_fetch(PL_utf8_ascii, p); |
977 | } |
978 | |
979 | bool |
864dbfa3 |
980 | Perl_is_utf8_space(pTHX_ U8 *p) |
a0ed51b3 |
981 | { |
ee8c7f54 |
982 | if (!is_utf8_char(p)) |
983 | return FALSE; |
a0ed51b3 |
984 | if (!PL_utf8_space) |
0e06870b |
985 | PL_utf8_space = swash_init("utf8", "IsSpacePerl", &PL_sv_undef, 0, 0); |
a0ed51b3 |
986 | return swash_fetch(PL_utf8_space, p); |
987 | } |
988 | |
989 | bool |
864dbfa3 |
990 | Perl_is_utf8_digit(pTHX_ U8 *p) |
a0ed51b3 |
991 | { |
ee8c7f54 |
992 | if (!is_utf8_char(p)) |
993 | return FALSE; |
a0ed51b3 |
994 | if (!PL_utf8_digit) |
e24b16f9 |
995 | PL_utf8_digit = swash_init("utf8", "IsDigit", &PL_sv_undef, 0, 0); |
a0ed51b3 |
996 | return swash_fetch(PL_utf8_digit, p); |
997 | } |
998 | |
999 | bool |
864dbfa3 |
1000 | Perl_is_utf8_upper(pTHX_ U8 *p) |
a0ed51b3 |
1001 | { |
ee8c7f54 |
1002 | if (!is_utf8_char(p)) |
1003 | return FALSE; |
a0ed51b3 |
1004 | if (!PL_utf8_upper) |
e24b16f9 |
1005 | PL_utf8_upper = swash_init("utf8", "IsUpper", &PL_sv_undef, 0, 0); |
a0ed51b3 |
1006 | return swash_fetch(PL_utf8_upper, p); |
1007 | } |
1008 | |
1009 | bool |
864dbfa3 |
1010 | Perl_is_utf8_lower(pTHX_ U8 *p) |
a0ed51b3 |
1011 | { |
ee8c7f54 |
1012 | if (!is_utf8_char(p)) |
1013 | return FALSE; |
a0ed51b3 |
1014 | if (!PL_utf8_lower) |
e24b16f9 |
1015 | PL_utf8_lower = swash_init("utf8", "IsLower", &PL_sv_undef, 0, 0); |
a0ed51b3 |
1016 | return swash_fetch(PL_utf8_lower, p); |
1017 | } |
1018 | |
1019 | bool |
b8c5462f |
1020 | Perl_is_utf8_cntrl(pTHX_ U8 *p) |
1021 | { |
ee8c7f54 |
1022 | if (!is_utf8_char(p)) |
1023 | return FALSE; |
b8c5462f |
1024 | if (!PL_utf8_cntrl) |
1025 | PL_utf8_cntrl = swash_init("utf8", "IsCntrl", &PL_sv_undef, 0, 0); |
1026 | return swash_fetch(PL_utf8_cntrl, p); |
1027 | } |
1028 | |
1029 | bool |
1030 | Perl_is_utf8_graph(pTHX_ U8 *p) |
1031 | { |
ee8c7f54 |
1032 | if (!is_utf8_char(p)) |
1033 | return FALSE; |
b8c5462f |
1034 | if (!PL_utf8_graph) |
1035 | PL_utf8_graph = swash_init("utf8", "IsGraph", &PL_sv_undef, 0, 0); |
1036 | return swash_fetch(PL_utf8_graph, p); |
1037 | } |
1038 | |
1039 | bool |
864dbfa3 |
1040 | Perl_is_utf8_print(pTHX_ U8 *p) |
a0ed51b3 |
1041 | { |
ee8c7f54 |
1042 | if (!is_utf8_char(p)) |
1043 | return FALSE; |
a0ed51b3 |
1044 | if (!PL_utf8_print) |
e24b16f9 |
1045 | PL_utf8_print = swash_init("utf8", "IsPrint", &PL_sv_undef, 0, 0); |
a0ed51b3 |
1046 | return swash_fetch(PL_utf8_print, p); |
1047 | } |
1048 | |
1049 | bool |
b8c5462f |
1050 | Perl_is_utf8_punct(pTHX_ U8 *p) |
1051 | { |
ee8c7f54 |
1052 | if (!is_utf8_char(p)) |
1053 | return FALSE; |
b8c5462f |
1054 | if (!PL_utf8_punct) |
1055 | PL_utf8_punct = swash_init("utf8", "IsPunct", &PL_sv_undef, 0, 0); |
1056 | return swash_fetch(PL_utf8_punct, p); |
1057 | } |
1058 | |
1059 | bool |
1060 | Perl_is_utf8_xdigit(pTHX_ U8 *p) |
1061 | { |
ee8c7f54 |
1062 | if (!is_utf8_char(p)) |
1063 | return FALSE; |
b8c5462f |
1064 | if (!PL_utf8_xdigit) |
1065 | PL_utf8_xdigit = swash_init("utf8", "IsXDigit", &PL_sv_undef, 0, 0); |
1066 | return swash_fetch(PL_utf8_xdigit, p); |
1067 | } |
1068 | |
1069 | bool |
864dbfa3 |
1070 | Perl_is_utf8_mark(pTHX_ U8 *p) |
a0ed51b3 |
1071 | { |
ee8c7f54 |
1072 | if (!is_utf8_char(p)) |
1073 | return FALSE; |
a0ed51b3 |
1074 | if (!PL_utf8_mark) |
e24b16f9 |
1075 | PL_utf8_mark = swash_init("utf8", "IsM", &PL_sv_undef, 0, 0); |
a0ed51b3 |
1076 | return swash_fetch(PL_utf8_mark, p); |
1077 | } |
1078 | |
2104c8d9 |
1079 | UV |
864dbfa3 |
1080 | Perl_to_utf8_upper(pTHX_ U8 *p) |
a0ed51b3 |
1081 | { |
1082 | UV uv; |
1083 | |
1084 | if (!PL_utf8_toupper) |
e24b16f9 |
1085 | PL_utf8_toupper = swash_init("utf8", "ToUpper", &PL_sv_undef, 4, 0); |
a0ed51b3 |
1086 | uv = swash_fetch(PL_utf8_toupper, p); |
0e06870b |
1087 | return uv ? uv : utf8_to_uv(p,UTF8_MAXLEN,0,0); |
a0ed51b3 |
1088 | } |
1089 | |
2104c8d9 |
1090 | UV |
864dbfa3 |
1091 | Perl_to_utf8_title(pTHX_ U8 *p) |
a0ed51b3 |
1092 | { |
1093 | UV uv; |
1094 | |
1095 | if (!PL_utf8_totitle) |
e24b16f9 |
1096 | PL_utf8_totitle = swash_init("utf8", "ToTitle", &PL_sv_undef, 4, 0); |
a0ed51b3 |
1097 | uv = swash_fetch(PL_utf8_totitle, p); |
0e06870b |
1098 | return uv ? uv : utf8_to_uv(p,UTF8_MAXLEN,0,0); |
a0ed51b3 |
1099 | } |
1100 | |
2104c8d9 |
1101 | UV |
864dbfa3 |
1102 | Perl_to_utf8_lower(pTHX_ U8 *p) |
a0ed51b3 |
1103 | { |
1104 | UV uv; |
1105 | |
1106 | if (!PL_utf8_tolower) |
e24b16f9 |
1107 | PL_utf8_tolower = swash_init("utf8", "ToLower", &PL_sv_undef, 4, 0); |
a0ed51b3 |
1108 | uv = swash_fetch(PL_utf8_tolower, p); |
0e06870b |
1109 | return uv ? uv : utf8_to_uv(p,UTF8_MAXLEN,0,0); |
a0ed51b3 |
1110 | } |
1111 | |
1112 | /* a "swash" is a swatch hash */ |
1113 | |
1114 | SV* |
864dbfa3 |
1115 | Perl_swash_init(pTHX_ char* pkg, char* name, SV *listsv, I32 minbits, I32 none) |
a0ed51b3 |
1116 | { |
1117 | SV* retval; |
0e06870b |
1118 | SV* tokenbufsv = sv_2mortal(NEWSV(0,0)); |
22d4bb9c |
1119 | dSP; |
ce3b816e |
1120 | |
1121 | if (!gv_stashpv(pkg, 0)) { /* demand load utf8 */ |
1122 | ENTER; |
1123 | Perl_load_module(aTHX_ PERL_LOADMOD_NOIMPORT, newSVpv(pkg,0), Nullsv); |
1124 | LEAVE; |
1125 | } |
1126 | SPAGAIN; |
a0ed51b3 |
1127 | PUSHSTACKi(PERLSI_MAGIC); |
1128 | PUSHMARK(SP); |
1129 | EXTEND(SP,5); |
1130 | PUSHs(sv_2mortal(newSVpvn(pkg, strlen(pkg)))); |
1131 | PUSHs(sv_2mortal(newSVpvn(name, strlen(name)))); |
1132 | PUSHs(listsv); |
1133 | PUSHs(sv_2mortal(newSViv(minbits))); |
1134 | PUSHs(sv_2mortal(newSViv(none))); |
1135 | PUTBACK; |
1136 | ENTER; |
1137 | SAVEI32(PL_hints); |
1138 | PL_hints = 0; |
1139 | save_re_context(); |
0e06870b |
1140 | if (PL_curcop == &PL_compiling) |
1141 | /* XXX ought to be handled by lex_start */ |
1142 | sv_setpv(tokenbufsv, PL_tokenbuf); |
864dbfa3 |
1143 | if (call_method("SWASHNEW", G_SCALAR)) |
22d4bb9c |
1144 | retval = newSVsv(*PL_stack_sp--); |
a0ed51b3 |
1145 | else |
e24b16f9 |
1146 | retval = &PL_sv_undef; |
a0ed51b3 |
1147 | LEAVE; |
1148 | POPSTACK; |
e24b16f9 |
1149 | if (PL_curcop == &PL_compiling) { |
0e06870b |
1150 | STRLEN len; |
1151 | char* pv = SvPV(tokenbufsv, len); |
1152 | |
1153 | Copy(pv, PL_tokenbuf, len+1, char); |
e24b16f9 |
1154 | PL_curcop->op_private = PL_hints; |
a0ed51b3 |
1155 | } |
1156 | if (!SvROK(retval) || SvTYPE(SvRV(retval)) != SVt_PVHV) |
cea2e8a9 |
1157 | Perl_croak(aTHX_ "SWASHNEW didn't return an HV ref"); |
a0ed51b3 |
1158 | return retval; |
1159 | } |
1160 | |
1161 | UV |
864dbfa3 |
1162 | Perl_swash_fetch(pTHX_ SV *sv, U8 *ptr) |
a0ed51b3 |
1163 | { |
1164 | HV* hv = (HV*)SvRV(sv); |
1165 | U32 klen = UTF8SKIP(ptr) - 1; |
1166 | U32 off = ptr[klen] & 127; /* NB: 64 bit always 0 when len > 1 */ |
1167 | STRLEN slen; |
1168 | STRLEN needents = (klen ? 64 : 128); |
dfe13c55 |
1169 | U8 *tmps; |
a0ed51b3 |
1170 | U32 bit; |
1171 | SV *retval; |
1172 | |
1173 | /* |
1174 | * This single-entry cache saves about 1/3 of the utf8 overhead in test |
1175 | * suite. (That is, only 7-8% overall over just a hash cache. Still, |
1176 | * it's nothing to sniff at.) Pity we usually come through at least |
1177 | * two function calls to get here... |
1178 | * |
1179 | * NB: this code assumes that swatches are never modified, once generated! |
1180 | */ |
1181 | |
1182 | if (hv == PL_last_swash_hv && |
1183 | klen == PL_last_swash_klen && |
4b19af01 |
1184 | (!klen || memEQ((char *)ptr,(char *)PL_last_swash_key,klen)) ) |
a0ed51b3 |
1185 | { |
1186 | tmps = PL_last_swash_tmps; |
1187 | slen = PL_last_swash_slen; |
1188 | } |
1189 | else { |
1190 | /* Try our second-level swatch cache, kept in a hash. */ |
dfe13c55 |
1191 | SV** svp = hv_fetch(hv, (char*)ptr, klen, FALSE); |
a0ed51b3 |
1192 | |
1193 | /* If not cached, generate it via utf8::SWASHGET */ |
dfe13c55 |
1194 | if (!svp || !SvPOK(*svp) || !(tmps = (U8*)SvPV(*svp, slen))) { |
a0ed51b3 |
1195 | dSP; |
1196 | ENTER; |
1197 | SAVETMPS; |
1198 | save_re_context(); |
1199 | PUSHSTACKi(PERLSI_MAGIC); |
1200 | PUSHMARK(SP); |
1201 | EXTEND(SP,3); |
1202 | PUSHs((SV*)sv); |
0e06870b |
1203 | PUSHs(sv_2mortal(newSViv(utf8_to_uv(ptr, UTF8_MAXLEN, 0, 0) & ~(needents - 1)))); |
a0ed51b3 |
1204 | PUSHs(sv_2mortal(newSViv(needents))); |
1205 | PUTBACK; |
864dbfa3 |
1206 | if (call_method("SWASHGET", G_SCALAR)) |
22d4bb9c |
1207 | retval = newSVsv(*PL_stack_sp--); |
a0ed51b3 |
1208 | else |
e24b16f9 |
1209 | retval = &PL_sv_undef; |
a0ed51b3 |
1210 | POPSTACK; |
1211 | FREETMPS; |
1212 | LEAVE; |
e24b16f9 |
1213 | if (PL_curcop == &PL_compiling) |
1214 | PL_curcop->op_private = PL_hints; |
a0ed51b3 |
1215 | |
dfe13c55 |
1216 | svp = hv_store(hv, (char*)ptr, klen, retval, 0); |
a0ed51b3 |
1217 | |
dfe13c55 |
1218 | if (!svp || !(tmps = (U8*)SvPV(*svp, slen)) || slen < 8) |
cea2e8a9 |
1219 | Perl_croak(aTHX_ "SWASHGET didn't return result of proper length"); |
a0ed51b3 |
1220 | } |
1221 | |
1222 | PL_last_swash_hv = hv; |
1223 | PL_last_swash_klen = klen; |
1224 | PL_last_swash_tmps = tmps; |
1225 | PL_last_swash_slen = slen; |
1226 | if (klen) |
1227 | Copy(ptr, PL_last_swash_key, klen, U8); |
1228 | } |
1229 | |
0e06870b |
1230 | switch ((int)((slen << 3) / needents)) { |
a0ed51b3 |
1231 | case 1: |
1232 | bit = 1 << (off & 7); |
1233 | off >>= 3; |
1234 | return (tmps[off] & bit) != 0; |
1235 | case 8: |
1236 | return tmps[off]; |
1237 | case 16: |
1238 | off <<= 1; |
1239 | return (tmps[off] << 8) + tmps[off + 1] ; |
1240 | case 32: |
1241 | off <<= 2; |
1242 | return (tmps[off] << 24) + (tmps[off+1] << 16) + (tmps[off+2] << 8) + tmps[off + 3] ; |
1243 | } |
cea2e8a9 |
1244 | Perl_croak(aTHX_ "panic: swash_fetch"); |
a0ed51b3 |
1245 | return 0; |
1246 | } |