Commit | Line | Data |
a0ed51b3 |
1 | /* utf8.c |
2 | * |
bc89e66f |
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 * |
ad391ad9 |
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; |
ad391ad9 |
34 | *d = 0; |
a0ed51b3 |
35 | return d; |
36 | } |
37 | if (uv < 0x800) { |
38 | *d++ = (( uv >> 6) | 0xc0); |
39 | *d++ = (( uv & 0x3f) | 0x80); |
ad391ad9 |
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); |
ad391ad9 |
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); |
ad391ad9 |
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); |
ad391ad9 |
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); |
ad391ad9 |
74 | *d = 0; |
a0ed51b3 |
75 | return d; |
76 | } |
6b8eaf93 |
77 | #ifdef HAS_QUAD |
d7578b48 |
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); |
ad391ad9 |
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); |
ad391ad9 |
106 | *d = 0; |
a0ed51b3 |
107 | return d; |
108 | } |
109 | #endif |
110 | } |
111 | |
386d01d6 |
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. */ |
067a85ef |
115 | STRLEN |
386d01d6 |
116 | Perl_is_utf8_char(pTHX_ U8 *s) |
117 | { |
118 | U8 u = *s; |
067a85ef |
119 | STRLEN slen, len; |
120 | UV uv, ouv; |
386d01d6 |
121 | |
60006e79 |
122 | if (UTF8_IS_ASCII(u)) |
386d01d6 |
123 | return 1; |
124 | |
60006e79 |
125 | if (!UTF8_IS_START(u)) |
386d01d6 |
126 | return 0; |
127 | |
9f07fdcd |
128 | len = UTF8SKIP(s); |
386d01d6 |
129 | |
60006e79 |
130 | if (len < 2 || !UTF8_IS_CONTINUATION(s[1])) |
067a85ef |
131 | return 0; |
132 | |
386d01d6 |
133 | slen = len - 1; |
134 | s++; |
067a85ef |
135 | uv = u; |
136 | ouv = uv; |
386d01d6 |
137 | while (slen--) { |
60006e79 |
138 | if (!UTF8_IS_CONTINUATION(*s)) |
386d01d6 |
139 | return 0; |
8850bf83 |
140 | uv = UTF8_ACCUMULATE(uv, *s); |
067a85ef |
141 | if (uv < ouv) |
142 | return 0; |
143 | ouv = uv; |
386d01d6 |
144 | s++; |
145 | } |
067a85ef |
146 | |
5bbb0b5a |
147 | if (UNISKIP(uv) < len) |
067a85ef |
148 | return 0; |
149 | |
386d01d6 |
150 | return len; |
151 | } |
152 | |
6662521e |
153 | /* |
b2a2e44b |
154 | =for apidoc Am|is_utf8_string|U8 *s|STRLEN len |
6662521e |
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 | |
8e84507e |
162 | bool |
6662521e |
163 | Perl_is_utf8_string(pTHX_ U8 *s, STRLEN len) |
164 | { |
067a85ef |
165 | U8* x = s; |
1aa99e6b |
166 | U8* send; |
067a85ef |
167 | STRLEN c; |
168 | |
1aa99e6b |
169 | if (!len) |
6cd5fe39 |
170 | len = strlen((char *)s); |
1aa99e6b |
171 | send = s + len; |
172 | |
6662521e |
173 | while (x < send) { |
174 | c = is_utf8_char(x); |
067a85ef |
175 | if (!c) |
176 | return FALSE; |
6662521e |
177 | x += c; |
6662521e |
178 | } |
60006e79 |
179 | if (x != send) |
180 | return FALSE; |
067a85ef |
181 | |
182 | return TRUE; |
6662521e |
183 | } |
184 | |
67e989fb |
185 | /* |
be2c7115 |
186 | =for apidoc Am|U8* s|utf8_to_uv|STRLEN curlen|STRLEN *retlen|U32 flags |
67e989fb |
187 | |
188 | Returns the character value of the first character in the string C<s> |
ba210ebe |
189 | which is assumed to be in UTF8 encoding and no longer than C<curlen>; |
7df053ec |
190 | C<retlen> will be set to the length, in bytes, of that character. |
67e989fb |
191 | |
192 | If C<s> does not point to a well-formed UTF8 character, the behaviour |
dcad2880 |
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 |
28d3d195 |
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. |
199 | |
200 | The C<flags> can also contain various flags to allow deviations from |
201 | the strict UTF-8 encoding (see F<utf8.h>). |
67e989fb |
202 | |
dcad2880 |
203 | =cut */ |
67e989fb |
204 | |
a0ed51b3 |
205 | UV |
dcad2880 |
206 | Perl_utf8_to_uv(pTHX_ U8* s, STRLEN curlen, STRLEN* retlen, U32 flags) |
a0ed51b3 |
207 | { |
ba210ebe |
208 | UV uv = *s, ouv; |
209 | STRLEN len = 1; |
7bf1b6bb |
210 | #ifdef EBCDIC |
211 | bool dowarn = 0; |
212 | #else |
ba210ebe |
213 | bool dowarn = ckWARN_d(WARN_UTF8); |
7bf1b6bb |
214 | #endif |
ba210ebe |
215 | STRLEN expectlen = 0; |
216 | |
0c443dc2 |
217 | if (curlen == 0) { |
218 | if (dowarn) |
219 | Perl_warner(aTHX_ WARN_UTF8, |
220 | "Malformed UTF-8 character (an empty string)"); |
221 | goto malformed; |
222 | } |
223 | |
421a8bf2 |
224 | if (UTF8_IS_ASCII(uv)) { |
a0ed51b3 |
225 | if (retlen) |
226 | *retlen = 1; |
227 | return *s; |
228 | } |
67e989fb |
229 | |
421a8bf2 |
230 | if (UTF8_IS_CONTINUATION(uv) && |
fcc8fcf6 |
231 | !(flags & UTF8_ALLOW_CONTINUATION)) { |
ba210ebe |
232 | if (dowarn) |
233 | Perl_warner(aTHX_ WARN_UTF8, |
efbcad09 |
234 | "Malformed UTF-8 character (unexpected continuation byte 0x%02"UVxf")", |
ba210ebe |
235 | uv); |
236 | goto malformed; |
237 | } |
238 | |
421a8bf2 |
239 | if (UTF8_IS_START(uv) && curlen > 1 && !UTF8_IS_CONTINUATION(s[1]) && |
fcc8fcf6 |
240 | !(flags & UTF8_ALLOW_NON_CONTINUATION)) { |
ba210ebe |
241 | if (dowarn) |
242 | Perl_warner(aTHX_ WARN_UTF8, |
421a8bf2 |
243 | "Malformed UTF-8 character (unexpected non-continuation byte 0x%02"UVxf" after start byte 0x%02"UVxf")", |
9e29e5ff |
244 | (UV)s[1], uv); |
ba210ebe |
245 | goto malformed; |
246 | } |
fcc8fcf6 |
247 | |
248 | if ((uv == 0xfe || uv == 0xff) && |
249 | !(flags & UTF8_ALLOW_FE_FF)) { |
ba210ebe |
250 | if (dowarn) |
251 | Perl_warner(aTHX_ WARN_UTF8, |
efbcad09 |
252 | "Malformed UTF-8 character (byte 0x%02"UVxf")", |
ba210ebe |
253 | uv); |
254 | goto malformed; |
a0ed51b3 |
255 | } |
fcc8fcf6 |
256 | |
ba210ebe |
257 | if (!(uv & 0x20)) { len = 2; uv &= 0x1f; } |
258 | else if (!(uv & 0x10)) { len = 3; uv &= 0x0f; } |
259 | else if (!(uv & 0x08)) { len = 4; uv &= 0x07; } |
260 | else if (!(uv & 0x04)) { len = 5; uv &= 0x03; } |
261 | else if (!(uv & 0x02)) { len = 6; uv &= 0x01; } |
262 | else if (!(uv & 0x01)) { len = 7; uv = 0; } |
3c77ea2b |
263 | else { len = 13; uv = 0; } /* whoa! */ |
fcc8fcf6 |
264 | |
a0ed51b3 |
265 | if (retlen) |
266 | *retlen = len; |
ba210ebe |
267 | |
268 | expectlen = len; |
269 | |
fcc8fcf6 |
270 | if ((curlen < expectlen) && |
271 | !(flags & UTF8_ALLOW_SHORT)) { |
ba210ebe |
272 | if (dowarn) |
273 | Perl_warner(aTHX_ WARN_UTF8, |
274 | "Malformed UTF-8 character (%d byte%s, need %d)", |
0c443dc2 |
275 | curlen, curlen == 1 ? "" : "s", expectlen); |
ba210ebe |
276 | goto malformed; |
277 | } |
278 | |
279 | len--; |
a0ed51b3 |
280 | s++; |
ba210ebe |
281 | ouv = uv; |
282 | |
a0ed51b3 |
283 | while (len--) { |
421a8bf2 |
284 | if (!UTF8_IS_CONTINUATION(*s) && |
285 | !(flags & UTF8_ALLOW_NON_CONTINUATION)) { |
ba210ebe |
286 | if (dowarn) |
287 | Perl_warner(aTHX_ WARN_UTF8, |
421a8bf2 |
288 | "Malformed UTF-8 character (unexpected non-continuation byte 0x%02x)", |
ba210ebe |
289 | *s); |
290 | goto malformed; |
a0ed51b3 |
291 | } |
292 | else |
8850bf83 |
293 | uv = UTF8_ACCUMULATE(uv, *s); |
ba210ebe |
294 | if (uv < ouv) { |
fcc8fcf6 |
295 | /* This cannot be allowed. */ |
ba210ebe |
296 | if (dowarn) |
297 | Perl_warner(aTHX_ WARN_UTF8, |
298 | "Malformed UTF-8 character (overflow at 0x%"UVxf", byte 0x%02x)", |
299 | ouv, *s); |
300 | goto malformed; |
301 | } |
302 | s++; |
303 | ouv = uv; |
304 | } |
305 | |
421a8bf2 |
306 | if (UNICODE_IS_SURROGATE(uv) && |
fcc8fcf6 |
307 | !(flags & UTF8_ALLOW_SURROGATE)) { |
ba210ebe |
308 | if (dowarn) |
309 | Perl_warner(aTHX_ WARN_UTF8, |
310 | "Malformed UTF-8 character (UTF-16 surrogate 0x%04"UVxf")", |
311 | uv); |
312 | goto malformed; |
421a8bf2 |
313 | } else if (UNICODE_IS_BYTE_ORDER_MARK(uv) && |
fcc8fcf6 |
314 | !(flags & UTF8_ALLOW_BOM)) { |
ba210ebe |
315 | if (dowarn) |
316 | Perl_warner(aTHX_ WARN_UTF8, |
317 | "Malformed UTF-8 character (byte order mark 0x%04"UVxf")", |
318 | uv); |
319 | goto malformed; |
fcc8fcf6 |
320 | } else if ((expectlen > UNISKIP(uv)) && |
321 | !(flags & UTF8_ALLOW_LONG)) { |
ba210ebe |
322 | if (dowarn) |
323 | Perl_warner(aTHX_ WARN_UTF8, |
324 | "Malformed UTF-8 character (%d byte%s, need %d)", |
0c443dc2 |
325 | expectlen, expectlen == 1 ? "": "s", UNISKIP(uv)); |
ba210ebe |
326 | goto malformed; |
421a8bf2 |
327 | } else if (UNICODE_IS_ILLEGAL(uv) && |
a9917092 |
328 | !(flags & UTF8_ALLOW_FFFF)) { |
329 | if (dowarn) |
330 | Perl_warner(aTHX_ WARN_UTF8, |
331 | "Malformed UTF-8 character (character 0x%04"UVxf")", |
332 | uv); |
333 | goto malformed; |
a0ed51b3 |
334 | } |
ba210ebe |
335 | |
a0ed51b3 |
336 | return uv; |
ba210ebe |
337 | |
338 | malformed: |
339 | |
fcc8fcf6 |
340 | if (flags & UTF8_CHECK_ONLY) { |
ba210ebe |
341 | if (retlen) |
cc366d4b |
342 | *retlen = -1; |
ba210ebe |
343 | return 0; |
344 | } |
345 | |
346 | if (retlen) |
28d3d195 |
347 | *retlen = expectlen ? expectlen : len; |
ba210ebe |
348 | |
28d3d195 |
349 | return 0; |
a0ed51b3 |
350 | } |
351 | |
8e84507e |
352 | /* |
dcad2880 |
353 | =for apidoc Am|U8* s|utf8_to_uv_simple|STRLEN *retlen |
8e84507e |
354 | |
355 | Returns the character value of the first character in the string C<s> |
356 | which is assumed to be in UTF8 encoding; C<retlen> will be set to the |
7df053ec |
357 | length, in bytes, of that character. |
8e84507e |
358 | |
ba210ebe |
359 | If C<s> does not point to a well-formed UTF8 character, zero is |
360 | returned and retlen is set, if possible, to -1. |
8e84507e |
361 | |
362 | =cut |
363 | */ |
364 | |
365 | UV |
dcad2880 |
366 | Perl_utf8_to_uv_simple(pTHX_ U8* s, STRLEN* retlen) |
8e84507e |
367 | { |
2e4dc9fc |
368 | return Perl_utf8_to_uv(aTHX_ s, UTF8_MAXLEN, retlen, 0); |
8e84507e |
369 | } |
370 | |
b76347f2 |
371 | /* |
b06226ff |
372 | =for apidoc Am|STRLEN|utf8_length|U8* s|U8 *e |
b76347f2 |
373 | |
374 | Return the length of the UTF-8 char encoded string C<s> in characters. |
02eb7b47 |
375 | Stops at C<e> (inclusive). If C<e E<lt> s> or if the scan would end |
376 | up past C<e>, croaks. |
b76347f2 |
377 | |
378 | =cut |
379 | */ |
380 | |
381 | STRLEN |
382 | Perl_utf8_length(pTHX_ U8* s, U8* e) |
383 | { |
384 | STRLEN len = 0; |
385 | |
8850bf83 |
386 | /* Note: cannot use UTF8_IS_...() too eagerly here since e.g. |
387 | * the bitops (especially ~) can create illegal UTF-8. |
388 | * In other words: in Perl UTF-8 is not just for Unicode. */ |
389 | |
b76347f2 |
390 | if (e < s) |
02eb7b47 |
391 | Perl_croak(aTHX_ "panic: utf8_length: unexpected end"); |
b76347f2 |
392 | while (s < e) { |
02eb7b47 |
393 | U8 t = UTF8SKIP(s); |
b76347f2 |
394 | |
395 | if (e - s < t) |
02eb7b47 |
396 | Perl_croak(aTHX_ "panic: utf8_length: unaligned end"); |
b76347f2 |
397 | s += t; |
398 | len++; |
399 | } |
400 | |
401 | return len; |
402 | } |
403 | |
b06226ff |
404 | /* |
405 | =for apidoc Am|IV|utf8_distance|U8 *a|U8 *b |
406 | |
407 | Returns the number of UTF8 characters between the UTF-8 pointers C<a> |
408 | and C<b>. |
409 | |
410 | WARNING: use only if you *know* that the pointers point inside the |
411 | same UTF-8 buffer. |
412 | |
413 | =cut */ |
a0ed51b3 |
414 | |
02eb7b47 |
415 | IV |
864dbfa3 |
416 | Perl_utf8_distance(pTHX_ U8 *a, U8 *b) |
a0ed51b3 |
417 | { |
02eb7b47 |
418 | IV off = 0; |
419 | |
8850bf83 |
420 | /* Note: cannot use UTF8_IS_...() too eagerly here since e.g. |
421 | * the bitops (especially ~) can create illegal UTF-8. |
422 | * In other words: in Perl UTF-8 is not just for Unicode. */ |
423 | |
a0ed51b3 |
424 | if (a < b) { |
425 | while (a < b) { |
02eb7b47 |
426 | U8 c = UTF8SKIP(a); |
427 | |
428 | if (b - a < c) |
429 | Perl_croak(aTHX_ "panic: utf8_distance: unaligned end"); |
430 | a += c; |
a0ed51b3 |
431 | off--; |
432 | } |
433 | } |
434 | else { |
435 | while (b < a) { |
02eb7b47 |
436 | U8 c = UTF8SKIP(b); |
437 | |
438 | if (a - b < c) |
439 | Perl_croak(aTHX_ "panic: utf8_distance: unaligned end"); |
440 | b += c; |
a0ed51b3 |
441 | off++; |
442 | } |
443 | } |
02eb7b47 |
444 | |
a0ed51b3 |
445 | return off; |
446 | } |
447 | |
b06226ff |
448 | /* |
449 | =for apidoc Am|U8*|utf8_hop|U8 *s|I32 off |
450 | |
8850bf83 |
451 | Return the UTF-8 pointer C<s> displaced by C<off> characters, either |
452 | forward or backward. |
b06226ff |
453 | |
454 | WARNING: do not use the following unless you *know* C<off> is within |
8850bf83 |
455 | the UTF-8 data pointed to by C<s> *and* that on entry C<s> is aligned |
456 | on the first byte of character or just after the last byte of a character. |
b06226ff |
457 | |
458 | =cut */ |
a0ed51b3 |
459 | |
460 | U8 * |
864dbfa3 |
461 | Perl_utf8_hop(pTHX_ U8 *s, I32 off) |
a0ed51b3 |
462 | { |
8850bf83 |
463 | /* Note: cannot use UTF8_IS_...() too eagerly here since e.g |
464 | * the bitops (especially ~) can create illegal UTF-8. |
465 | * In other words: in Perl UTF-8 is not just for Unicode. */ |
466 | |
a0ed51b3 |
467 | if (off >= 0) { |
468 | while (off--) |
469 | s += UTF8SKIP(s); |
470 | } |
471 | else { |
472 | while (off++) { |
473 | s--; |
8850bf83 |
474 | while (UTF8_IS_CONTINUATION(*s)) |
475 | s--; |
a0ed51b3 |
476 | } |
477 | } |
478 | return s; |
479 | } |
480 | |
6940069f |
481 | /* |
246fae53 |
482 | =for apidoc Am|U8 *|utf8_to_bytes|U8 *s|STRLEN *len |
6940069f |
483 | |
246fae53 |
484 | Converts a string C<s> of length C<len> from UTF8 into byte encoding. |
485 | Unlike C<bytes_to_utf8>, this over-writes the original string, and |
486 | updates len to contain the new length. |
67e989fb |
487 | Returns zero on failure, setting C<len> to -1. |
6940069f |
488 | |
489 | =cut |
490 | */ |
491 | |
492 | U8 * |
246fae53 |
493 | Perl_utf8_to_bytes(pTHX_ U8* s, STRLEN *len) |
6940069f |
494 | { |
6940069f |
495 | U8 *send; |
496 | U8 *d; |
dcad2880 |
497 | U8 *save = s; |
246fae53 |
498 | |
499 | /* ensure valid UTF8 and chars < 256 before updating string */ |
dcad2880 |
500 | for (send = s + *len; s < send; ) { |
501 | U8 c = *s++; |
502 | |
9f9ab905 |
503 | if (c >= 0x80 && |
dcad2880 |
504 | ((s >= send) || |
505 | ((*s++ & 0xc0) != 0x80) || ((c & 0xfe) != 0xc2))) { |
506 | *len = -1; |
507 | return 0; |
508 | } |
246fae53 |
509 | } |
dcad2880 |
510 | |
511 | d = s = save; |
6940069f |
512 | while (s < send) { |
ed646e6e |
513 | STRLEN ulen; |
514 | *d++ = (U8)utf8_to_uv_simple(s, &ulen); |
515 | s += ulen; |
6940069f |
516 | } |
517 | *d = '\0'; |
246fae53 |
518 | *len = d - save; |
6940069f |
519 | return save; |
520 | } |
521 | |
522 | /* |
6662521e |
523 | =for apidoc Am|U8 *|bytes_to_utf8|U8 *s|STRLEN *len |
6940069f |
524 | |
525 | Converts a string C<s> of length C<len> from ASCII into UTF8 encoding. |
6662521e |
526 | Returns a pointer to the newly-created string, and sets C<len> to |
527 | reflect the new length. |
6940069f |
528 | |
497711e7 |
529 | =cut |
6940069f |
530 | */ |
531 | |
532 | U8* |
6662521e |
533 | Perl_bytes_to_utf8(pTHX_ U8* s, STRLEN *len) |
6940069f |
534 | { |
6940069f |
535 | U8 *send; |
536 | U8 *d; |
537 | U8 *dst; |
6662521e |
538 | send = s + (*len); |
6940069f |
539 | |
6662521e |
540 | Newz(801, d, (*len) * 2 + 1, U8); |
6940069f |
541 | dst = d; |
542 | |
543 | while (s < send) { |
544 | if (*s < 0x80) |
545 | *d++ = *s++; |
546 | else { |
547 | UV uv = *s++; |
548 | *d++ = (( uv >> 6) | 0xc0); |
549 | *d++ = (( uv & 0x3f) | 0x80); |
550 | } |
551 | } |
552 | *d = '\0'; |
6662521e |
553 | *len = d-dst; |
6940069f |
554 | return dst; |
555 | } |
556 | |
a0ed51b3 |
557 | /* |
dea0fc0b |
558 | * Convert native (big-endian) or reversed (little-endian) UTF-16 to UTF-8. |
a0ed51b3 |
559 | * |
560 | * Destination must be pre-extended to 3/2 source. Do not use in-place. |
561 | * We optimize for native, for obvious reasons. */ |
562 | |
563 | U8* |
dea0fc0b |
564 | Perl_utf16_to_utf8(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen) |
a0ed51b3 |
565 | { |
dea0fc0b |
566 | U8* pend; |
567 | U8* dstart = d; |
568 | |
569 | if (bytelen & 1) |
a7867d0a |
570 | Perl_croak(aTHX_ "panic: utf16_to_utf8: odd bytelen"); |
dea0fc0b |
571 | |
572 | pend = p + bytelen; |
573 | |
a0ed51b3 |
574 | while (p < pend) { |
dea0fc0b |
575 | UV uv = (p[0] << 8) + p[1]; /* UTF-16BE */ |
576 | p += 2; |
a0ed51b3 |
577 | if (uv < 0x80) { |
578 | *d++ = uv; |
579 | continue; |
580 | } |
581 | if (uv < 0x800) { |
582 | *d++ = (( uv >> 6) | 0xc0); |
583 | *d++ = (( uv & 0x3f) | 0x80); |
584 | continue; |
585 | } |
586 | if (uv >= 0xd800 && uv < 0xdbff) { /* surrogates */ |
dea0fc0b |
587 | UV low = *p++; |
588 | if (low < 0xdc00 || low >= 0xdfff) |
589 | Perl_croak(aTHX_ "Malformed UTF-16 surrogate"); |
a0ed51b3 |
590 | uv = ((uv - 0xd800) << 10) + (low - 0xdc00) + 0x10000; |
591 | } |
592 | if (uv < 0x10000) { |
593 | *d++ = (( uv >> 12) | 0xe0); |
594 | *d++ = (((uv >> 6) & 0x3f) | 0x80); |
595 | *d++ = (( uv & 0x3f) | 0x80); |
596 | continue; |
597 | } |
598 | else { |
599 | *d++ = (( uv >> 18) | 0xf0); |
600 | *d++ = (((uv >> 12) & 0x3f) | 0x80); |
601 | *d++ = (((uv >> 6) & 0x3f) | 0x80); |
602 | *d++ = (( uv & 0x3f) | 0x80); |
603 | continue; |
604 | } |
605 | } |
dea0fc0b |
606 | *newlen = d - dstart; |
a0ed51b3 |
607 | return d; |
608 | } |
609 | |
610 | /* Note: this one is slightly destructive of the source. */ |
611 | |
612 | U8* |
dea0fc0b |
613 | Perl_utf16_to_utf8_reversed(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen) |
a0ed51b3 |
614 | { |
615 | U8* s = (U8*)p; |
616 | U8* send = s + bytelen; |
617 | while (s < send) { |
618 | U8 tmp = s[0]; |
619 | s[0] = s[1]; |
620 | s[1] = tmp; |
621 | s += 2; |
622 | } |
dea0fc0b |
623 | return utf16_to_utf8(p, d, bytelen, newlen); |
a0ed51b3 |
624 | } |
625 | |
626 | /* for now these are all defined (inefficiently) in terms of the utf8 versions */ |
627 | |
628 | bool |
864dbfa3 |
629 | Perl_is_uni_alnum(pTHX_ U32 c) |
a0ed51b3 |
630 | { |
ad391ad9 |
631 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
632 | uv_to_utf8(tmpbuf, (UV)c); |
633 | return is_utf8_alnum(tmpbuf); |
634 | } |
635 | |
636 | bool |
b8c5462f |
637 | Perl_is_uni_alnumc(pTHX_ U32 c) |
638 | { |
ad391ad9 |
639 | U8 tmpbuf[UTF8_MAXLEN+1]; |
b8c5462f |
640 | uv_to_utf8(tmpbuf, (UV)c); |
641 | return is_utf8_alnumc(tmpbuf); |
642 | } |
643 | |
644 | bool |
864dbfa3 |
645 | Perl_is_uni_idfirst(pTHX_ U32 c) |
a0ed51b3 |
646 | { |
ad391ad9 |
647 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
648 | uv_to_utf8(tmpbuf, (UV)c); |
649 | return is_utf8_idfirst(tmpbuf); |
650 | } |
651 | |
652 | bool |
864dbfa3 |
653 | Perl_is_uni_alpha(pTHX_ U32 c) |
a0ed51b3 |
654 | { |
ad391ad9 |
655 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
656 | uv_to_utf8(tmpbuf, (UV)c); |
657 | return is_utf8_alpha(tmpbuf); |
658 | } |
659 | |
660 | bool |
4d61ec05 |
661 | Perl_is_uni_ascii(pTHX_ U32 c) |
662 | { |
ad391ad9 |
663 | U8 tmpbuf[UTF8_MAXLEN+1]; |
4d61ec05 |
664 | uv_to_utf8(tmpbuf, (UV)c); |
665 | return is_utf8_ascii(tmpbuf); |
666 | } |
667 | |
668 | bool |
864dbfa3 |
669 | Perl_is_uni_space(pTHX_ U32 c) |
a0ed51b3 |
670 | { |
ad391ad9 |
671 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
672 | uv_to_utf8(tmpbuf, (UV)c); |
673 | return is_utf8_space(tmpbuf); |
674 | } |
675 | |
676 | bool |
864dbfa3 |
677 | Perl_is_uni_digit(pTHX_ U32 c) |
a0ed51b3 |
678 | { |
ad391ad9 |
679 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
680 | uv_to_utf8(tmpbuf, (UV)c); |
681 | return is_utf8_digit(tmpbuf); |
682 | } |
683 | |
684 | bool |
864dbfa3 |
685 | Perl_is_uni_upper(pTHX_ U32 c) |
a0ed51b3 |
686 | { |
ad391ad9 |
687 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
688 | uv_to_utf8(tmpbuf, (UV)c); |
689 | return is_utf8_upper(tmpbuf); |
690 | } |
691 | |
692 | bool |
864dbfa3 |
693 | Perl_is_uni_lower(pTHX_ U32 c) |
a0ed51b3 |
694 | { |
ad391ad9 |
695 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
696 | uv_to_utf8(tmpbuf, (UV)c); |
697 | return is_utf8_lower(tmpbuf); |
698 | } |
699 | |
700 | bool |
b8c5462f |
701 | Perl_is_uni_cntrl(pTHX_ U32 c) |
702 | { |
ad391ad9 |
703 | U8 tmpbuf[UTF8_MAXLEN+1]; |
b8c5462f |
704 | uv_to_utf8(tmpbuf, (UV)c); |
705 | return is_utf8_cntrl(tmpbuf); |
706 | } |
707 | |
708 | bool |
709 | Perl_is_uni_graph(pTHX_ U32 c) |
710 | { |
ad391ad9 |
711 | U8 tmpbuf[UTF8_MAXLEN+1]; |
b8c5462f |
712 | uv_to_utf8(tmpbuf, (UV)c); |
713 | return is_utf8_graph(tmpbuf); |
714 | } |
715 | |
716 | bool |
864dbfa3 |
717 | Perl_is_uni_print(pTHX_ U32 c) |
a0ed51b3 |
718 | { |
ad391ad9 |
719 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
720 | uv_to_utf8(tmpbuf, (UV)c); |
721 | return is_utf8_print(tmpbuf); |
722 | } |
723 | |
b8c5462f |
724 | bool |
f248d071 |
725 | Perl_is_uni_punct(pTHX_ U32 c) |
b8c5462f |
726 | { |
ad391ad9 |
727 | U8 tmpbuf[UTF8_MAXLEN+1]; |
b8c5462f |
728 | uv_to_utf8(tmpbuf, (UV)c); |
729 | return is_utf8_punct(tmpbuf); |
730 | } |
731 | |
4d61ec05 |
732 | bool |
733 | Perl_is_uni_xdigit(pTHX_ U32 c) |
734 | { |
ad391ad9 |
735 | U8 tmpbuf[UTF8_MAXLEN+1]; |
4d61ec05 |
736 | uv_to_utf8(tmpbuf, (UV)c); |
737 | return is_utf8_xdigit(tmpbuf); |
738 | } |
739 | |
a0ed51b3 |
740 | U32 |
864dbfa3 |
741 | Perl_to_uni_upper(pTHX_ U32 c) |
a0ed51b3 |
742 | { |
ad391ad9 |
743 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
744 | uv_to_utf8(tmpbuf, (UV)c); |
745 | return to_utf8_upper(tmpbuf); |
746 | } |
747 | |
748 | U32 |
864dbfa3 |
749 | Perl_to_uni_title(pTHX_ U32 c) |
a0ed51b3 |
750 | { |
ad391ad9 |
751 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
752 | uv_to_utf8(tmpbuf, (UV)c); |
753 | return to_utf8_title(tmpbuf); |
754 | } |
755 | |
756 | U32 |
864dbfa3 |
757 | Perl_to_uni_lower(pTHX_ U32 c) |
a0ed51b3 |
758 | { |
ad391ad9 |
759 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
760 | uv_to_utf8(tmpbuf, (UV)c); |
761 | return to_utf8_lower(tmpbuf); |
762 | } |
763 | |
764 | /* for now these all assume no locale info available for Unicode > 255 */ |
765 | |
766 | bool |
864dbfa3 |
767 | Perl_is_uni_alnum_lc(pTHX_ U32 c) |
a0ed51b3 |
768 | { |
769 | return is_uni_alnum(c); /* XXX no locale support yet */ |
770 | } |
771 | |
772 | bool |
b8c5462f |
773 | Perl_is_uni_alnumc_lc(pTHX_ U32 c) |
774 | { |
775 | return is_uni_alnumc(c); /* XXX no locale support yet */ |
776 | } |
777 | |
778 | bool |
864dbfa3 |
779 | Perl_is_uni_idfirst_lc(pTHX_ U32 c) |
a0ed51b3 |
780 | { |
781 | return is_uni_idfirst(c); /* XXX no locale support yet */ |
782 | } |
783 | |
784 | bool |
864dbfa3 |
785 | Perl_is_uni_alpha_lc(pTHX_ U32 c) |
a0ed51b3 |
786 | { |
787 | return is_uni_alpha(c); /* XXX no locale support yet */ |
788 | } |
789 | |
790 | bool |
4d61ec05 |
791 | Perl_is_uni_ascii_lc(pTHX_ U32 c) |
792 | { |
793 | return is_uni_ascii(c); /* XXX no locale support yet */ |
794 | } |
795 | |
796 | bool |
864dbfa3 |
797 | Perl_is_uni_space_lc(pTHX_ U32 c) |
a0ed51b3 |
798 | { |
799 | return is_uni_space(c); /* XXX no locale support yet */ |
800 | } |
801 | |
802 | bool |
864dbfa3 |
803 | Perl_is_uni_digit_lc(pTHX_ U32 c) |
a0ed51b3 |
804 | { |
805 | return is_uni_digit(c); /* XXX no locale support yet */ |
806 | } |
807 | |
808 | bool |
864dbfa3 |
809 | Perl_is_uni_upper_lc(pTHX_ U32 c) |
a0ed51b3 |
810 | { |
811 | return is_uni_upper(c); /* XXX no locale support yet */ |
812 | } |
813 | |
814 | bool |
864dbfa3 |
815 | Perl_is_uni_lower_lc(pTHX_ U32 c) |
a0ed51b3 |
816 | { |
817 | return is_uni_lower(c); /* XXX no locale support yet */ |
818 | } |
819 | |
820 | bool |
b8c5462f |
821 | Perl_is_uni_cntrl_lc(pTHX_ U32 c) |
822 | { |
823 | return is_uni_cntrl(c); /* XXX no locale support yet */ |
824 | } |
825 | |
826 | bool |
827 | Perl_is_uni_graph_lc(pTHX_ U32 c) |
828 | { |
829 | return is_uni_graph(c); /* XXX no locale support yet */ |
830 | } |
831 | |
832 | bool |
864dbfa3 |
833 | Perl_is_uni_print_lc(pTHX_ U32 c) |
a0ed51b3 |
834 | { |
835 | return is_uni_print(c); /* XXX no locale support yet */ |
836 | } |
837 | |
b8c5462f |
838 | bool |
839 | Perl_is_uni_punct_lc(pTHX_ U32 c) |
840 | { |
841 | return is_uni_punct(c); /* XXX no locale support yet */ |
842 | } |
843 | |
4d61ec05 |
844 | bool |
845 | Perl_is_uni_xdigit_lc(pTHX_ U32 c) |
846 | { |
847 | return is_uni_xdigit(c); /* XXX no locale support yet */ |
848 | } |
849 | |
a0ed51b3 |
850 | U32 |
864dbfa3 |
851 | Perl_to_uni_upper_lc(pTHX_ U32 c) |
a0ed51b3 |
852 | { |
853 | return to_uni_upper(c); /* XXX no locale support yet */ |
854 | } |
855 | |
856 | U32 |
864dbfa3 |
857 | Perl_to_uni_title_lc(pTHX_ U32 c) |
a0ed51b3 |
858 | { |
859 | return to_uni_title(c); /* XXX no locale support yet */ |
860 | } |
861 | |
862 | U32 |
864dbfa3 |
863 | Perl_to_uni_lower_lc(pTHX_ U32 c) |
a0ed51b3 |
864 | { |
865 | return to_uni_lower(c); /* XXX no locale support yet */ |
866 | } |
867 | |
a0ed51b3 |
868 | bool |
864dbfa3 |
869 | Perl_is_utf8_alnum(pTHX_ U8 *p) |
a0ed51b3 |
870 | { |
386d01d6 |
871 | if (!is_utf8_char(p)) |
872 | return FALSE; |
a0ed51b3 |
873 | if (!PL_utf8_alnum) |
289d4f09 |
874 | /* NOTE: "IsWord", not "IsAlnum", since Alnum is a true |
875 | * descendant of isalnum(3), in other words, it doesn't |
876 | * contain the '_'. --jhi */ |
877 | PL_utf8_alnum = swash_init("utf8", "IsWord", &PL_sv_undef, 0, 0); |
a0ed51b3 |
878 | return swash_fetch(PL_utf8_alnum, p); |
879 | /* return *p == '_' || is_utf8_alpha(p) || is_utf8_digit(p); */ |
880 | #ifdef SURPRISINGLY_SLOWER /* probably because alpha is usually true */ |
881 | if (!PL_utf8_alnum) |
882 | PL_utf8_alnum = swash_init("utf8", "", |
883 | sv_2mortal(newSVpv("+utf8::IsAlpha\n+utf8::IsDigit\n005F\n",0)), 0, 0); |
884 | return swash_fetch(PL_utf8_alnum, p); |
885 | #endif |
886 | } |
887 | |
888 | bool |
b8c5462f |
889 | Perl_is_utf8_alnumc(pTHX_ U8 *p) |
890 | { |
386d01d6 |
891 | if (!is_utf8_char(p)) |
892 | return FALSE; |
b8c5462f |
893 | if (!PL_utf8_alnum) |
894 | PL_utf8_alnum = swash_init("utf8", "IsAlnumC", &PL_sv_undef, 0, 0); |
895 | return swash_fetch(PL_utf8_alnum, p); |
896 | /* return is_utf8_alpha(p) || is_utf8_digit(p); */ |
897 | #ifdef SURPRISINGLY_SLOWER /* probably because alpha is usually true */ |
898 | if (!PL_utf8_alnum) |
899 | PL_utf8_alnum = swash_init("utf8", "", |
900 | sv_2mortal(newSVpv("+utf8::IsAlpha\n+utf8::IsDigit\n005F\n",0)), 0, 0); |
901 | return swash_fetch(PL_utf8_alnum, p); |
902 | #endif |
903 | } |
904 | |
905 | bool |
864dbfa3 |
906 | Perl_is_utf8_idfirst(pTHX_ U8 *p) |
a0ed51b3 |
907 | { |
908 | return *p == '_' || is_utf8_alpha(p); |
909 | } |
910 | |
911 | bool |
864dbfa3 |
912 | Perl_is_utf8_alpha(pTHX_ U8 *p) |
a0ed51b3 |
913 | { |
386d01d6 |
914 | if (!is_utf8_char(p)) |
915 | return FALSE; |
a0ed51b3 |
916 | if (!PL_utf8_alpha) |
e24b16f9 |
917 | PL_utf8_alpha = swash_init("utf8", "IsAlpha", &PL_sv_undef, 0, 0); |
a0ed51b3 |
918 | return swash_fetch(PL_utf8_alpha, p); |
919 | } |
920 | |
921 | bool |
b8c5462f |
922 | Perl_is_utf8_ascii(pTHX_ U8 *p) |
923 | { |
386d01d6 |
924 | if (!is_utf8_char(p)) |
925 | return FALSE; |
b8c5462f |
926 | if (!PL_utf8_ascii) |
927 | PL_utf8_ascii = swash_init("utf8", "IsAscii", &PL_sv_undef, 0, 0); |
928 | return swash_fetch(PL_utf8_ascii, p); |
929 | } |
930 | |
931 | bool |
864dbfa3 |
932 | Perl_is_utf8_space(pTHX_ U8 *p) |
a0ed51b3 |
933 | { |
386d01d6 |
934 | if (!is_utf8_char(p)) |
935 | return FALSE; |
a0ed51b3 |
936 | if (!PL_utf8_space) |
3bec3564 |
937 | PL_utf8_space = swash_init("utf8", "IsSpacePerl", &PL_sv_undef, 0, 0); |
a0ed51b3 |
938 | return swash_fetch(PL_utf8_space, p); |
939 | } |
940 | |
941 | bool |
864dbfa3 |
942 | Perl_is_utf8_digit(pTHX_ U8 *p) |
a0ed51b3 |
943 | { |
386d01d6 |
944 | if (!is_utf8_char(p)) |
945 | return FALSE; |
a0ed51b3 |
946 | if (!PL_utf8_digit) |
e24b16f9 |
947 | PL_utf8_digit = swash_init("utf8", "IsDigit", &PL_sv_undef, 0, 0); |
a0ed51b3 |
948 | return swash_fetch(PL_utf8_digit, p); |
949 | } |
950 | |
951 | bool |
864dbfa3 |
952 | Perl_is_utf8_upper(pTHX_ U8 *p) |
a0ed51b3 |
953 | { |
386d01d6 |
954 | if (!is_utf8_char(p)) |
955 | return FALSE; |
a0ed51b3 |
956 | if (!PL_utf8_upper) |
e24b16f9 |
957 | PL_utf8_upper = swash_init("utf8", "IsUpper", &PL_sv_undef, 0, 0); |
a0ed51b3 |
958 | return swash_fetch(PL_utf8_upper, p); |
959 | } |
960 | |
961 | bool |
864dbfa3 |
962 | Perl_is_utf8_lower(pTHX_ U8 *p) |
a0ed51b3 |
963 | { |
386d01d6 |
964 | if (!is_utf8_char(p)) |
965 | return FALSE; |
a0ed51b3 |
966 | if (!PL_utf8_lower) |
e24b16f9 |
967 | PL_utf8_lower = swash_init("utf8", "IsLower", &PL_sv_undef, 0, 0); |
a0ed51b3 |
968 | return swash_fetch(PL_utf8_lower, p); |
969 | } |
970 | |
971 | bool |
b8c5462f |
972 | Perl_is_utf8_cntrl(pTHX_ U8 *p) |
973 | { |
386d01d6 |
974 | if (!is_utf8_char(p)) |
975 | return FALSE; |
b8c5462f |
976 | if (!PL_utf8_cntrl) |
977 | PL_utf8_cntrl = swash_init("utf8", "IsCntrl", &PL_sv_undef, 0, 0); |
978 | return swash_fetch(PL_utf8_cntrl, p); |
979 | } |
980 | |
981 | bool |
982 | Perl_is_utf8_graph(pTHX_ U8 *p) |
983 | { |
386d01d6 |
984 | if (!is_utf8_char(p)) |
985 | return FALSE; |
b8c5462f |
986 | if (!PL_utf8_graph) |
987 | PL_utf8_graph = swash_init("utf8", "IsGraph", &PL_sv_undef, 0, 0); |
988 | return swash_fetch(PL_utf8_graph, p); |
989 | } |
990 | |
991 | bool |
864dbfa3 |
992 | Perl_is_utf8_print(pTHX_ U8 *p) |
a0ed51b3 |
993 | { |
386d01d6 |
994 | if (!is_utf8_char(p)) |
995 | return FALSE; |
a0ed51b3 |
996 | if (!PL_utf8_print) |
e24b16f9 |
997 | PL_utf8_print = swash_init("utf8", "IsPrint", &PL_sv_undef, 0, 0); |
a0ed51b3 |
998 | return swash_fetch(PL_utf8_print, p); |
999 | } |
1000 | |
1001 | bool |
b8c5462f |
1002 | Perl_is_utf8_punct(pTHX_ U8 *p) |
1003 | { |
386d01d6 |
1004 | if (!is_utf8_char(p)) |
1005 | return FALSE; |
b8c5462f |
1006 | if (!PL_utf8_punct) |
1007 | PL_utf8_punct = swash_init("utf8", "IsPunct", &PL_sv_undef, 0, 0); |
1008 | return swash_fetch(PL_utf8_punct, p); |
1009 | } |
1010 | |
1011 | bool |
1012 | Perl_is_utf8_xdigit(pTHX_ U8 *p) |
1013 | { |
386d01d6 |
1014 | if (!is_utf8_char(p)) |
1015 | return FALSE; |
b8c5462f |
1016 | if (!PL_utf8_xdigit) |
1017 | PL_utf8_xdigit = swash_init("utf8", "IsXDigit", &PL_sv_undef, 0, 0); |
1018 | return swash_fetch(PL_utf8_xdigit, p); |
1019 | } |
1020 | |
1021 | bool |
864dbfa3 |
1022 | Perl_is_utf8_mark(pTHX_ U8 *p) |
a0ed51b3 |
1023 | { |
386d01d6 |
1024 | if (!is_utf8_char(p)) |
1025 | return FALSE; |
a0ed51b3 |
1026 | if (!PL_utf8_mark) |
e24b16f9 |
1027 | PL_utf8_mark = swash_init("utf8", "IsM", &PL_sv_undef, 0, 0); |
a0ed51b3 |
1028 | return swash_fetch(PL_utf8_mark, p); |
1029 | } |
1030 | |
2104c8d9 |
1031 | UV |
864dbfa3 |
1032 | Perl_to_utf8_upper(pTHX_ U8 *p) |
a0ed51b3 |
1033 | { |
1034 | UV uv; |
1035 | |
1036 | if (!PL_utf8_toupper) |
e24b16f9 |
1037 | PL_utf8_toupper = swash_init("utf8", "ToUpper", &PL_sv_undef, 4, 0); |
a0ed51b3 |
1038 | uv = swash_fetch(PL_utf8_toupper, p); |
756820e3 |
1039 | return uv ? uv : utf8_to_uv(p,UTF8_MAXLEN,0,0); |
a0ed51b3 |
1040 | } |
1041 | |
2104c8d9 |
1042 | UV |
864dbfa3 |
1043 | Perl_to_utf8_title(pTHX_ U8 *p) |
a0ed51b3 |
1044 | { |
1045 | UV uv; |
1046 | |
1047 | if (!PL_utf8_totitle) |
e24b16f9 |
1048 | PL_utf8_totitle = swash_init("utf8", "ToTitle", &PL_sv_undef, 4, 0); |
a0ed51b3 |
1049 | uv = swash_fetch(PL_utf8_totitle, p); |
756820e3 |
1050 | return uv ? uv : utf8_to_uv(p,UTF8_MAXLEN,0,0); |
a0ed51b3 |
1051 | } |
1052 | |
2104c8d9 |
1053 | UV |
864dbfa3 |
1054 | Perl_to_utf8_lower(pTHX_ U8 *p) |
a0ed51b3 |
1055 | { |
1056 | UV uv; |
1057 | |
1058 | if (!PL_utf8_tolower) |
e24b16f9 |
1059 | PL_utf8_tolower = swash_init("utf8", "ToLower", &PL_sv_undef, 4, 0); |
a0ed51b3 |
1060 | uv = swash_fetch(PL_utf8_tolower, p); |
756820e3 |
1061 | return uv ? uv : utf8_to_uv(p,UTF8_MAXLEN,0,0); |
a0ed51b3 |
1062 | } |
1063 | |
1064 | /* a "swash" is a swatch hash */ |
1065 | |
1066 | SV* |
864dbfa3 |
1067 | Perl_swash_init(pTHX_ char* pkg, char* name, SV *listsv, I32 minbits, I32 none) |
a0ed51b3 |
1068 | { |
1069 | SV* retval; |
1070 | char tmpbuf[256]; |
8e84507e |
1071 | dSP; |
ce3b816e |
1072 | |
1073 | if (!gv_stashpv(pkg, 0)) { /* demand load utf8 */ |
1074 | ENTER; |
1075 | Perl_load_module(aTHX_ PERL_LOADMOD_NOIMPORT, newSVpv(pkg,0), Nullsv); |
1076 | LEAVE; |
1077 | } |
1078 | SPAGAIN; |
a0ed51b3 |
1079 | PUSHSTACKi(PERLSI_MAGIC); |
1080 | PUSHMARK(SP); |
1081 | EXTEND(SP,5); |
1082 | PUSHs(sv_2mortal(newSVpvn(pkg, strlen(pkg)))); |
1083 | PUSHs(sv_2mortal(newSVpvn(name, strlen(name)))); |
1084 | PUSHs(listsv); |
1085 | PUSHs(sv_2mortal(newSViv(minbits))); |
1086 | PUSHs(sv_2mortal(newSViv(none))); |
1087 | PUTBACK; |
1088 | ENTER; |
1089 | SAVEI32(PL_hints); |
1090 | PL_hints = 0; |
1091 | save_re_context(); |
e24b16f9 |
1092 | if (PL_curcop == &PL_compiling) /* XXX ought to be handled by lex_start */ |
a0ed51b3 |
1093 | strncpy(tmpbuf, PL_tokenbuf, sizeof tmpbuf); |
864dbfa3 |
1094 | if (call_method("SWASHNEW", G_SCALAR)) |
8e84507e |
1095 | retval = newSVsv(*PL_stack_sp--); |
a0ed51b3 |
1096 | else |
e24b16f9 |
1097 | retval = &PL_sv_undef; |
a0ed51b3 |
1098 | LEAVE; |
1099 | POPSTACK; |
e24b16f9 |
1100 | if (PL_curcop == &PL_compiling) { |
a0ed51b3 |
1101 | strncpy(PL_tokenbuf, tmpbuf, sizeof tmpbuf); |
e24b16f9 |
1102 | PL_curcop->op_private = PL_hints; |
a0ed51b3 |
1103 | } |
1104 | if (!SvROK(retval) || SvTYPE(SvRV(retval)) != SVt_PVHV) |
cea2e8a9 |
1105 | Perl_croak(aTHX_ "SWASHNEW didn't return an HV ref"); |
a0ed51b3 |
1106 | return retval; |
1107 | } |
1108 | |
1109 | UV |
864dbfa3 |
1110 | Perl_swash_fetch(pTHX_ SV *sv, U8 *ptr) |
a0ed51b3 |
1111 | { |
1112 | HV* hv = (HV*)SvRV(sv); |
1113 | U32 klen = UTF8SKIP(ptr) - 1; |
1114 | U32 off = ptr[klen] & 127; /* NB: 64 bit always 0 when len > 1 */ |
1115 | STRLEN slen; |
1116 | STRLEN needents = (klen ? 64 : 128); |
dfe13c55 |
1117 | U8 *tmps; |
a0ed51b3 |
1118 | U32 bit; |
1119 | SV *retval; |
1120 | |
1121 | /* |
1122 | * This single-entry cache saves about 1/3 of the utf8 overhead in test |
1123 | * suite. (That is, only 7-8% overall over just a hash cache. Still, |
1124 | * it's nothing to sniff at.) Pity we usually come through at least |
1125 | * two function calls to get here... |
1126 | * |
1127 | * NB: this code assumes that swatches are never modified, once generated! |
1128 | */ |
1129 | |
1130 | if (hv == PL_last_swash_hv && |
1131 | klen == PL_last_swash_klen && |
12ae5dfc |
1132 | (!klen || memEQ((char *)ptr,(char *)PL_last_swash_key,klen)) ) |
a0ed51b3 |
1133 | { |
1134 | tmps = PL_last_swash_tmps; |
1135 | slen = PL_last_swash_slen; |
1136 | } |
1137 | else { |
1138 | /* Try our second-level swatch cache, kept in a hash. */ |
dfe13c55 |
1139 | SV** svp = hv_fetch(hv, (char*)ptr, klen, FALSE); |
a0ed51b3 |
1140 | |
1141 | /* If not cached, generate it via utf8::SWASHGET */ |
dfe13c55 |
1142 | if (!svp || !SvPOK(*svp) || !(tmps = (U8*)SvPV(*svp, slen))) { |
a0ed51b3 |
1143 | dSP; |
1144 | ENTER; |
1145 | SAVETMPS; |
1146 | save_re_context(); |
1147 | PUSHSTACKi(PERLSI_MAGIC); |
1148 | PUSHMARK(SP); |
1149 | EXTEND(SP,3); |
1150 | PUSHs((SV*)sv); |
756820e3 |
1151 | PUSHs(sv_2mortal(newSViv(utf8_to_uv(ptr, UTF8_MAXLEN, 0, 0) & ~(needents - 1)))); |
a0ed51b3 |
1152 | PUSHs(sv_2mortal(newSViv(needents))); |
1153 | PUTBACK; |
864dbfa3 |
1154 | if (call_method("SWASHGET", G_SCALAR)) |
8e84507e |
1155 | retval = newSVsv(*PL_stack_sp--); |
a0ed51b3 |
1156 | else |
e24b16f9 |
1157 | retval = &PL_sv_undef; |
a0ed51b3 |
1158 | POPSTACK; |
1159 | FREETMPS; |
1160 | LEAVE; |
e24b16f9 |
1161 | if (PL_curcop == &PL_compiling) |
1162 | PL_curcop->op_private = PL_hints; |
a0ed51b3 |
1163 | |
dfe13c55 |
1164 | svp = hv_store(hv, (char*)ptr, klen, retval, 0); |
a0ed51b3 |
1165 | |
dfe13c55 |
1166 | if (!svp || !(tmps = (U8*)SvPV(*svp, slen)) || slen < 8) |
cea2e8a9 |
1167 | Perl_croak(aTHX_ "SWASHGET didn't return result of proper length"); |
a0ed51b3 |
1168 | } |
1169 | |
1170 | PL_last_swash_hv = hv; |
1171 | PL_last_swash_klen = klen; |
1172 | PL_last_swash_tmps = tmps; |
1173 | PL_last_swash_slen = slen; |
1174 | if (klen) |
1175 | Copy(ptr, PL_last_swash_key, klen, U8); |
1176 | } |
1177 | |
9faf8d75 |
1178 | switch ((int)((slen << 3) / needents)) { |
a0ed51b3 |
1179 | case 1: |
1180 | bit = 1 << (off & 7); |
1181 | off >>= 3; |
1182 | return (tmps[off] & bit) != 0; |
1183 | case 8: |
1184 | return tmps[off]; |
1185 | case 16: |
1186 | off <<= 1; |
1187 | return (tmps[off] << 8) + tmps[off + 1] ; |
1188 | case 32: |
1189 | off <<= 2; |
1190 | return (tmps[off] << 24) + (tmps[off+1] << 16) + (tmps[off+2] << 8) + tmps[off + 3] ; |
1191 | } |
cea2e8a9 |
1192 | Perl_croak(aTHX_ "panic: swash_fetch"); |
a0ed51b3 |
1193 | return 0; |
1194 | } |