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 |
d7578b48 |
72 | if (uv < UTF8_QUAD_MAX) |
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. */ |
067a85ef |
107 | STRLEN |
386d01d6 |
108 | Perl_is_utf8_char(pTHX_ U8 *s) |
109 | { |
110 | U8 u = *s; |
067a85ef |
111 | STRLEN slen, len; |
112 | UV uv, ouv; |
386d01d6 |
113 | |
067a85ef |
114 | if (u <= 0x7f) |
386d01d6 |
115 | return 1; |
116 | |
067a85ef |
117 | if (u >= 0x80 && u <= 0xbf) |
386d01d6 |
118 | return 0; |
119 | |
9f07fdcd |
120 | len = UTF8SKIP(s); |
386d01d6 |
121 | |
067a85ef |
122 | if (len < 2 || (u >= 0xc0 && u <= 0xfd && s[1] < 0x80)) |
123 | return 0; |
124 | |
386d01d6 |
125 | slen = len - 1; |
126 | s++; |
067a85ef |
127 | uv = u; |
128 | ouv = uv; |
386d01d6 |
129 | while (slen--) { |
130 | if ((*s & 0xc0) != 0x80) |
131 | return 0; |
067a85ef |
132 | uv = (uv << 6) | (*s & 0x3f); |
133 | if (uv < ouv) |
134 | return 0; |
135 | ouv = uv; |
386d01d6 |
136 | s++; |
137 | } |
067a85ef |
138 | |
5bbb0b5a |
139 | if (UNISKIP(uv) < len) |
067a85ef |
140 | return 0; |
141 | |
386d01d6 |
142 | return len; |
143 | } |
144 | |
6662521e |
145 | /* |
b2a2e44b |
146 | =for apidoc Am|is_utf8_string|U8 *s|STRLEN len |
6662521e |
147 | |
148 | Returns true if first C<len> bytes of the given string form valid a UTF8 |
149 | string, false otherwise. |
150 | |
151 | =cut |
152 | */ |
153 | |
8e84507e |
154 | bool |
6662521e |
155 | Perl_is_utf8_string(pTHX_ U8 *s, STRLEN len) |
156 | { |
067a85ef |
157 | U8* x = s; |
158 | U8* send = s + len; |
159 | STRLEN c; |
160 | |
6662521e |
161 | while (x < send) { |
162 | c = is_utf8_char(x); |
067a85ef |
163 | if (!c) |
164 | return FALSE; |
6662521e |
165 | x += c; |
067a85ef |
166 | if (x > send) |
167 | return FALSE; |
6662521e |
168 | } |
067a85ef |
169 | |
170 | return TRUE; |
6662521e |
171 | } |
172 | |
67e989fb |
173 | /* |
dcad2880 |
174 | =for apidoc Am|U8* s|utf8_to_uv|STRLEN curlen|I32 *retlen|U32 flags |
67e989fb |
175 | |
176 | Returns the character value of the first character in the string C<s> |
ba210ebe |
177 | which is assumed to be in UTF8 encoding and no longer than C<curlen>; |
178 | C<retlen> will be set to the length, in bytes, of that character, |
179 | and the pointer C<s> will be advanced to the end of the character. |
67e989fb |
180 | |
181 | If C<s> does not point to a well-formed UTF8 character, the behaviour |
dcad2880 |
182 | is dependent on the value of C<flags>: if it contains UTF8_CHECK_ONLY, |
183 | it is assumed that the caller will raise a warning, and this function |
184 | will set C<retlen> to C<-1> and return. The C<flags> can also contain |
185 | various flags to allow deviations from the strict UTF-8 encoding. |
67e989fb |
186 | |
dcad2880 |
187 | =cut */ |
67e989fb |
188 | |
a0ed51b3 |
189 | UV |
dcad2880 |
190 | Perl_utf8_to_uv(pTHX_ U8* s, STRLEN curlen, STRLEN* retlen, U32 flags) |
a0ed51b3 |
191 | { |
ba210ebe |
192 | dTHR; |
193 | UV uv = *s, ouv; |
194 | STRLEN len = 1; |
7bf1b6bb |
195 | #ifdef EBCDIC |
196 | bool dowarn = 0; |
197 | #else |
ba210ebe |
198 | bool dowarn = ckWARN_d(WARN_UTF8); |
7bf1b6bb |
199 | #endif |
ba210ebe |
200 | STRLEN expectlen = 0; |
201 | |
0c443dc2 |
202 | if (curlen == 0) { |
203 | if (dowarn) |
204 | Perl_warner(aTHX_ WARN_UTF8, |
205 | "Malformed UTF-8 character (an empty string)"); |
206 | goto malformed; |
207 | } |
208 | |
ba210ebe |
209 | if (uv <= 0x7f) { /* Pure ASCII. */ |
a0ed51b3 |
210 | if (retlen) |
211 | *retlen = 1; |
212 | return *s; |
213 | } |
67e989fb |
214 | |
fcc8fcf6 |
215 | if ((uv >= 0x80 && uv <= 0xbf) && |
216 | !(flags & UTF8_ALLOW_CONTINUATION)) { |
ba210ebe |
217 | if (dowarn) |
218 | Perl_warner(aTHX_ WARN_UTF8, |
efbcad09 |
219 | "Malformed UTF-8 character (unexpected continuation byte 0x%02"UVxf")", |
ba210ebe |
220 | uv); |
221 | goto malformed; |
222 | } |
223 | |
0c443dc2 |
224 | if ((uv >= 0xc0 && uv <= 0xfd && curlen > 1 && s[1] < 0x80) && |
fcc8fcf6 |
225 | !(flags & UTF8_ALLOW_NON_CONTINUATION)) { |
ba210ebe |
226 | if (dowarn) |
227 | Perl_warner(aTHX_ WARN_UTF8, |
efbcad09 |
228 | "Malformed UTF-8 character (unexpected non-continuation byte 0x%02"UVxf" after byte 0x%02"UVxf")", |
9e29e5ff |
229 | (UV)s[1], uv); |
ba210ebe |
230 | goto malformed; |
231 | } |
fcc8fcf6 |
232 | |
233 | if ((uv == 0xfe || uv == 0xff) && |
234 | !(flags & UTF8_ALLOW_FE_FF)) { |
ba210ebe |
235 | if (dowarn) |
236 | Perl_warner(aTHX_ WARN_UTF8, |
efbcad09 |
237 | "Malformed UTF-8 character (byte 0x%02"UVxf")", |
ba210ebe |
238 | uv); |
239 | goto malformed; |
a0ed51b3 |
240 | } |
fcc8fcf6 |
241 | |
ba210ebe |
242 | if (!(uv & 0x20)) { len = 2; uv &= 0x1f; } |
243 | else if (!(uv & 0x10)) { len = 3; uv &= 0x0f; } |
244 | else if (!(uv & 0x08)) { len = 4; uv &= 0x07; } |
245 | else if (!(uv & 0x04)) { len = 5; uv &= 0x03; } |
246 | else if (!(uv & 0x02)) { len = 6; uv &= 0x01; } |
247 | else if (!(uv & 0x01)) { len = 7; uv = 0; } |
3c77ea2b |
248 | else { len = 13; uv = 0; } /* whoa! */ |
fcc8fcf6 |
249 | |
a0ed51b3 |
250 | if (retlen) |
251 | *retlen = len; |
ba210ebe |
252 | |
253 | expectlen = len; |
254 | |
fcc8fcf6 |
255 | if ((curlen < expectlen) && |
256 | !(flags & UTF8_ALLOW_SHORT)) { |
ba210ebe |
257 | if (dowarn) |
258 | Perl_warner(aTHX_ WARN_UTF8, |
259 | "Malformed UTF-8 character (%d byte%s, need %d)", |
0c443dc2 |
260 | curlen, curlen == 1 ? "" : "s", expectlen); |
ba210ebe |
261 | goto malformed; |
262 | } |
263 | |
264 | len--; |
a0ed51b3 |
265 | s++; |
ba210ebe |
266 | ouv = uv; |
267 | |
a0ed51b3 |
268 | while (len--) { |
269 | if ((*s & 0xc0) != 0x80) { |
ba210ebe |
270 | if (dowarn) |
271 | Perl_warner(aTHX_ WARN_UTF8, |
272 | "Malformed UTF-8 character (unexpected continuation byte 0x%02x)", |
273 | *s); |
274 | goto malformed; |
a0ed51b3 |
275 | } |
276 | else |
ba210ebe |
277 | uv = (uv << 6) | (*s & 0x3f); |
278 | if (uv < ouv) { |
fcc8fcf6 |
279 | /* This cannot be allowed. */ |
ba210ebe |
280 | if (dowarn) |
281 | Perl_warner(aTHX_ WARN_UTF8, |
282 | "Malformed UTF-8 character (overflow at 0x%"UVxf", byte 0x%02x)", |
283 | ouv, *s); |
284 | goto malformed; |
285 | } |
286 | s++; |
287 | ouv = uv; |
288 | } |
289 | |
fcc8fcf6 |
290 | if ((uv >= 0xd800 && uv <= 0xdfff) && |
291 | !(flags & UTF8_ALLOW_SURROGATE)) { |
ba210ebe |
292 | if (dowarn) |
293 | Perl_warner(aTHX_ WARN_UTF8, |
294 | "Malformed UTF-8 character (UTF-16 surrogate 0x%04"UVxf")", |
295 | uv); |
296 | goto malformed; |
fcc8fcf6 |
297 | } else if ((uv == 0xfffe) && |
298 | !(flags & UTF8_ALLOW_BOM)) { |
ba210ebe |
299 | if (dowarn) |
300 | Perl_warner(aTHX_ WARN_UTF8, |
301 | "Malformed UTF-8 character (byte order mark 0x%04"UVxf")", |
302 | uv); |
303 | goto malformed; |
fcc8fcf6 |
304 | } else if ((uv == 0xffff) && |
305 | !(flags & UTF8_ALLOW_FFFF)) { |
ba210ebe |
306 | if (dowarn) |
307 | Perl_warner(aTHX_ WARN_UTF8, |
fcc8fcf6 |
308 | "Malformed UTF-8 character (character 0x%04"UVxf")", |
ba210ebe |
309 | uv); |
310 | goto malformed; |
fcc8fcf6 |
311 | } else if ((expectlen > UNISKIP(uv)) && |
312 | !(flags & UTF8_ALLOW_LONG)) { |
ba210ebe |
313 | if (dowarn) |
314 | Perl_warner(aTHX_ WARN_UTF8, |
315 | "Malformed UTF-8 character (%d byte%s, need %d)", |
0c443dc2 |
316 | expectlen, expectlen == 1 ? "": "s", UNISKIP(uv)); |
ba210ebe |
317 | goto malformed; |
a0ed51b3 |
318 | } |
ba210ebe |
319 | |
a0ed51b3 |
320 | return uv; |
ba210ebe |
321 | |
322 | malformed: |
323 | |
fcc8fcf6 |
324 | if (flags & UTF8_CHECK_ONLY) { |
ba210ebe |
325 | if (retlen) |
cc366d4b |
326 | *retlen = -1; |
ba210ebe |
327 | return 0; |
328 | } |
329 | |
330 | if (retlen) |
cc366d4b |
331 | *retlen = expectlen ? expectlen : len; |
ba210ebe |
332 | |
333 | return UNICODE_REPLACEMENT_CHARACTER; |
a0ed51b3 |
334 | } |
335 | |
8e84507e |
336 | /* |
dcad2880 |
337 | =for apidoc Am|U8* s|utf8_to_uv_simple|STRLEN *retlen |
8e84507e |
338 | |
339 | Returns the character value of the first character in the string C<s> |
340 | which is assumed to be in UTF8 encoding; C<retlen> will be set to the |
341 | length, in bytes, of that character, and the pointer C<s> will be |
342 | advanced to the end of the character. |
343 | |
ba210ebe |
344 | If C<s> does not point to a well-formed UTF8 character, zero is |
345 | returned and retlen is set, if possible, to -1. |
8e84507e |
346 | |
347 | =cut |
348 | */ |
349 | |
350 | UV |
dcad2880 |
351 | Perl_utf8_to_uv_simple(pTHX_ U8* s, STRLEN* retlen) |
8e84507e |
352 | { |
2e4dc9fc |
353 | return Perl_utf8_to_uv(aTHX_ s, UTF8_MAXLEN, retlen, 0); |
8e84507e |
354 | } |
355 | |
b76347f2 |
356 | /* |
357 | =for apidoc|utf8_length|U8 *s|U8 *e |
358 | |
359 | Return the length of the UTF-8 char encoded string C<s> in characters. |
02eb7b47 |
360 | Stops at C<e> (inclusive). If C<e E<lt> s> or if the scan would end |
361 | up past C<e>, croaks. |
b76347f2 |
362 | |
363 | =cut |
364 | */ |
365 | |
366 | STRLEN |
367 | Perl_utf8_length(pTHX_ U8* s, U8* e) |
368 | { |
369 | STRLEN len = 0; |
370 | |
371 | if (e < s) |
02eb7b47 |
372 | Perl_croak(aTHX_ "panic: utf8_length: unexpected end"); |
b76347f2 |
373 | while (s < e) { |
02eb7b47 |
374 | U8 t = UTF8SKIP(s); |
b76347f2 |
375 | |
376 | if (e - s < t) |
02eb7b47 |
377 | Perl_croak(aTHX_ "panic: utf8_length: unaligned end"); |
b76347f2 |
378 | s += t; |
379 | len++; |
380 | } |
381 | |
382 | return len; |
383 | } |
384 | |
246fae53 |
385 | /* utf8_distance(a,b) returns the number of UTF8 characters between |
386 | the pointers a and b */ |
a0ed51b3 |
387 | |
02eb7b47 |
388 | IV |
864dbfa3 |
389 | Perl_utf8_distance(pTHX_ U8 *a, U8 *b) |
a0ed51b3 |
390 | { |
02eb7b47 |
391 | IV off = 0; |
392 | |
a0ed51b3 |
393 | if (a < b) { |
394 | while (a < b) { |
02eb7b47 |
395 | U8 c = UTF8SKIP(a); |
396 | |
397 | if (b - a < c) |
398 | Perl_croak(aTHX_ "panic: utf8_distance: unaligned end"); |
399 | a += c; |
a0ed51b3 |
400 | off--; |
401 | } |
402 | } |
403 | else { |
404 | while (b < a) { |
02eb7b47 |
405 | U8 c = UTF8SKIP(b); |
406 | |
407 | if (a - b < c) |
408 | Perl_croak(aTHX_ "panic: utf8_distance: unaligned end"); |
409 | b += c; |
a0ed51b3 |
410 | off++; |
411 | } |
412 | } |
02eb7b47 |
413 | |
a0ed51b3 |
414 | return off; |
415 | } |
416 | |
417 | /* WARNING: do not use the following unless you *know* off is within bounds */ |
418 | |
419 | U8 * |
864dbfa3 |
420 | Perl_utf8_hop(pTHX_ U8 *s, I32 off) |
a0ed51b3 |
421 | { |
422 | if (off >= 0) { |
423 | while (off--) |
424 | s += UTF8SKIP(s); |
425 | } |
426 | else { |
427 | while (off++) { |
428 | s--; |
429 | if (*s & 0x80) { |
430 | while ((*s & 0xc0) == 0x80) |
431 | s--; |
432 | } |
433 | } |
434 | } |
435 | return s; |
436 | } |
437 | |
6940069f |
438 | /* |
246fae53 |
439 | =for apidoc Am|U8 *|utf8_to_bytes|U8 *s|STRLEN *len |
6940069f |
440 | |
246fae53 |
441 | Converts a string C<s> of length C<len> from UTF8 into byte encoding. |
442 | Unlike C<bytes_to_utf8>, this over-writes the original string, and |
443 | updates len to contain the new length. |
67e989fb |
444 | Returns zero on failure, setting C<len> to -1. |
6940069f |
445 | |
446 | =cut |
447 | */ |
448 | |
449 | U8 * |
246fae53 |
450 | Perl_utf8_to_bytes(pTHX_ U8* s, STRLEN *len) |
6940069f |
451 | { |
6940069f |
452 | U8 *send; |
453 | U8 *d; |
dcad2880 |
454 | U8 *save = s; |
246fae53 |
455 | |
456 | /* ensure valid UTF8 and chars < 256 before updating string */ |
dcad2880 |
457 | for (send = s + *len; s < send; ) { |
458 | U8 c = *s++; |
459 | |
9f9ab905 |
460 | if (c >= 0x80 && |
dcad2880 |
461 | ((s >= send) || |
462 | ((*s++ & 0xc0) != 0x80) || ((c & 0xfe) != 0xc2))) { |
463 | *len = -1; |
464 | return 0; |
465 | } |
246fae53 |
466 | } |
dcad2880 |
467 | |
468 | d = s = save; |
6940069f |
469 | while (s < send) { |
dcad2880 |
470 | if (*s < 0x80) { |
471 | *d++ = *s++; |
472 | } |
6940069f |
473 | else { |
ba210ebe |
474 | STRLEN ulen; |
dcad2880 |
475 | *d++ = (U8)utf8_to_uv_simple(s, &ulen); |
6940069f |
476 | s += ulen; |
6940069f |
477 | } |
478 | } |
479 | *d = '\0'; |
246fae53 |
480 | *len = d - save; |
6940069f |
481 | return save; |
482 | } |
483 | |
484 | /* |
6662521e |
485 | =for apidoc Am|U8 *|bytes_to_utf8|U8 *s|STRLEN *len |
6940069f |
486 | |
487 | Converts a string C<s> of length C<len> from ASCII into UTF8 encoding. |
6662521e |
488 | Returns a pointer to the newly-created string, and sets C<len> to |
489 | reflect the new length. |
6940069f |
490 | |
497711e7 |
491 | =cut |
6940069f |
492 | */ |
493 | |
494 | U8* |
6662521e |
495 | Perl_bytes_to_utf8(pTHX_ U8* s, STRLEN *len) |
6940069f |
496 | { |
497 | dTHR; |
498 | U8 *send; |
499 | U8 *d; |
500 | U8 *dst; |
6662521e |
501 | send = s + (*len); |
6940069f |
502 | |
6662521e |
503 | Newz(801, d, (*len) * 2 + 1, U8); |
6940069f |
504 | dst = d; |
505 | |
506 | while (s < send) { |
507 | if (*s < 0x80) |
508 | *d++ = *s++; |
509 | else { |
510 | UV uv = *s++; |
511 | *d++ = (( uv >> 6) | 0xc0); |
512 | *d++ = (( uv & 0x3f) | 0x80); |
513 | } |
514 | } |
515 | *d = '\0'; |
6662521e |
516 | *len = d-dst; |
6940069f |
517 | return dst; |
518 | } |
519 | |
a0ed51b3 |
520 | /* |
dea0fc0b |
521 | * Convert native (big-endian) or reversed (little-endian) UTF-16 to UTF-8. |
a0ed51b3 |
522 | * |
523 | * Destination must be pre-extended to 3/2 source. Do not use in-place. |
524 | * We optimize for native, for obvious reasons. */ |
525 | |
526 | U8* |
dea0fc0b |
527 | Perl_utf16_to_utf8(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen) |
a0ed51b3 |
528 | { |
dea0fc0b |
529 | U8* pend; |
530 | U8* dstart = d; |
531 | |
532 | if (bytelen & 1) |
a7867d0a |
533 | Perl_croak(aTHX_ "panic: utf16_to_utf8: odd bytelen"); |
dea0fc0b |
534 | |
535 | pend = p + bytelen; |
536 | |
a0ed51b3 |
537 | while (p < pend) { |
dea0fc0b |
538 | UV uv = (p[0] << 8) + p[1]; /* UTF-16BE */ |
539 | p += 2; |
a0ed51b3 |
540 | if (uv < 0x80) { |
541 | *d++ = uv; |
542 | continue; |
543 | } |
544 | if (uv < 0x800) { |
545 | *d++ = (( uv >> 6) | 0xc0); |
546 | *d++ = (( uv & 0x3f) | 0x80); |
547 | continue; |
548 | } |
549 | if (uv >= 0xd800 && uv < 0xdbff) { /* surrogates */ |
0453d815 |
550 | dTHR; |
dea0fc0b |
551 | UV low = *p++; |
552 | if (low < 0xdc00 || low >= 0xdfff) |
553 | Perl_croak(aTHX_ "Malformed UTF-16 surrogate"); |
a0ed51b3 |
554 | uv = ((uv - 0xd800) << 10) + (low - 0xdc00) + 0x10000; |
555 | } |
556 | if (uv < 0x10000) { |
557 | *d++ = (( uv >> 12) | 0xe0); |
558 | *d++ = (((uv >> 6) & 0x3f) | 0x80); |
559 | *d++ = (( uv & 0x3f) | 0x80); |
560 | continue; |
561 | } |
562 | else { |
563 | *d++ = (( uv >> 18) | 0xf0); |
564 | *d++ = (((uv >> 12) & 0x3f) | 0x80); |
565 | *d++ = (((uv >> 6) & 0x3f) | 0x80); |
566 | *d++ = (( uv & 0x3f) | 0x80); |
567 | continue; |
568 | } |
569 | } |
dea0fc0b |
570 | *newlen = d - dstart; |
a0ed51b3 |
571 | return d; |
572 | } |
573 | |
574 | /* Note: this one is slightly destructive of the source. */ |
575 | |
576 | U8* |
dea0fc0b |
577 | Perl_utf16_to_utf8_reversed(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen) |
a0ed51b3 |
578 | { |
579 | U8* s = (U8*)p; |
580 | U8* send = s + bytelen; |
581 | while (s < send) { |
582 | U8 tmp = s[0]; |
583 | s[0] = s[1]; |
584 | s[1] = tmp; |
585 | s += 2; |
586 | } |
dea0fc0b |
587 | return utf16_to_utf8(p, d, bytelen, newlen); |
a0ed51b3 |
588 | } |
589 | |
590 | /* for now these are all defined (inefficiently) in terms of the utf8 versions */ |
591 | |
592 | bool |
864dbfa3 |
593 | Perl_is_uni_alnum(pTHX_ U32 c) |
a0ed51b3 |
594 | { |
aa6ffa16 |
595 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
596 | uv_to_utf8(tmpbuf, (UV)c); |
597 | return is_utf8_alnum(tmpbuf); |
598 | } |
599 | |
600 | bool |
b8c5462f |
601 | Perl_is_uni_alnumc(pTHX_ U32 c) |
602 | { |
aa6ffa16 |
603 | U8 tmpbuf[UTF8_MAXLEN]; |
b8c5462f |
604 | uv_to_utf8(tmpbuf, (UV)c); |
605 | return is_utf8_alnumc(tmpbuf); |
606 | } |
607 | |
608 | bool |
864dbfa3 |
609 | Perl_is_uni_idfirst(pTHX_ U32 c) |
a0ed51b3 |
610 | { |
aa6ffa16 |
611 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
612 | uv_to_utf8(tmpbuf, (UV)c); |
613 | return is_utf8_idfirst(tmpbuf); |
614 | } |
615 | |
616 | bool |
864dbfa3 |
617 | Perl_is_uni_alpha(pTHX_ U32 c) |
a0ed51b3 |
618 | { |
aa6ffa16 |
619 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
620 | uv_to_utf8(tmpbuf, (UV)c); |
621 | return is_utf8_alpha(tmpbuf); |
622 | } |
623 | |
624 | bool |
4d61ec05 |
625 | Perl_is_uni_ascii(pTHX_ U32 c) |
626 | { |
aa6ffa16 |
627 | U8 tmpbuf[UTF8_MAXLEN]; |
4d61ec05 |
628 | uv_to_utf8(tmpbuf, (UV)c); |
629 | return is_utf8_ascii(tmpbuf); |
630 | } |
631 | |
632 | bool |
864dbfa3 |
633 | Perl_is_uni_space(pTHX_ U32 c) |
a0ed51b3 |
634 | { |
aa6ffa16 |
635 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
636 | uv_to_utf8(tmpbuf, (UV)c); |
637 | return is_utf8_space(tmpbuf); |
638 | } |
639 | |
640 | bool |
864dbfa3 |
641 | Perl_is_uni_digit(pTHX_ U32 c) |
a0ed51b3 |
642 | { |
aa6ffa16 |
643 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
644 | uv_to_utf8(tmpbuf, (UV)c); |
645 | return is_utf8_digit(tmpbuf); |
646 | } |
647 | |
648 | bool |
864dbfa3 |
649 | Perl_is_uni_upper(pTHX_ U32 c) |
a0ed51b3 |
650 | { |
aa6ffa16 |
651 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
652 | uv_to_utf8(tmpbuf, (UV)c); |
653 | return is_utf8_upper(tmpbuf); |
654 | } |
655 | |
656 | bool |
864dbfa3 |
657 | Perl_is_uni_lower(pTHX_ U32 c) |
a0ed51b3 |
658 | { |
aa6ffa16 |
659 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
660 | uv_to_utf8(tmpbuf, (UV)c); |
661 | return is_utf8_lower(tmpbuf); |
662 | } |
663 | |
664 | bool |
b8c5462f |
665 | Perl_is_uni_cntrl(pTHX_ U32 c) |
666 | { |
aa6ffa16 |
667 | U8 tmpbuf[UTF8_MAXLEN]; |
b8c5462f |
668 | uv_to_utf8(tmpbuf, (UV)c); |
669 | return is_utf8_cntrl(tmpbuf); |
670 | } |
671 | |
672 | bool |
673 | Perl_is_uni_graph(pTHX_ U32 c) |
674 | { |
aa6ffa16 |
675 | U8 tmpbuf[UTF8_MAXLEN]; |
b8c5462f |
676 | uv_to_utf8(tmpbuf, (UV)c); |
677 | return is_utf8_graph(tmpbuf); |
678 | } |
679 | |
680 | bool |
864dbfa3 |
681 | Perl_is_uni_print(pTHX_ U32 c) |
a0ed51b3 |
682 | { |
aa6ffa16 |
683 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
684 | uv_to_utf8(tmpbuf, (UV)c); |
685 | return is_utf8_print(tmpbuf); |
686 | } |
687 | |
b8c5462f |
688 | bool |
f248d071 |
689 | Perl_is_uni_punct(pTHX_ U32 c) |
b8c5462f |
690 | { |
aa6ffa16 |
691 | U8 tmpbuf[UTF8_MAXLEN]; |
b8c5462f |
692 | uv_to_utf8(tmpbuf, (UV)c); |
693 | return is_utf8_punct(tmpbuf); |
694 | } |
695 | |
4d61ec05 |
696 | bool |
697 | Perl_is_uni_xdigit(pTHX_ U32 c) |
698 | { |
aa6ffa16 |
699 | U8 tmpbuf[UTF8_MAXLEN]; |
4d61ec05 |
700 | uv_to_utf8(tmpbuf, (UV)c); |
701 | return is_utf8_xdigit(tmpbuf); |
702 | } |
703 | |
a0ed51b3 |
704 | U32 |
864dbfa3 |
705 | Perl_to_uni_upper(pTHX_ U32 c) |
a0ed51b3 |
706 | { |
aa6ffa16 |
707 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
708 | uv_to_utf8(tmpbuf, (UV)c); |
709 | return to_utf8_upper(tmpbuf); |
710 | } |
711 | |
712 | U32 |
864dbfa3 |
713 | Perl_to_uni_title(pTHX_ U32 c) |
a0ed51b3 |
714 | { |
aa6ffa16 |
715 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
716 | uv_to_utf8(tmpbuf, (UV)c); |
717 | return to_utf8_title(tmpbuf); |
718 | } |
719 | |
720 | U32 |
864dbfa3 |
721 | Perl_to_uni_lower(pTHX_ U32 c) |
a0ed51b3 |
722 | { |
aa6ffa16 |
723 | U8 tmpbuf[UTF8_MAXLEN]; |
a0ed51b3 |
724 | uv_to_utf8(tmpbuf, (UV)c); |
725 | return to_utf8_lower(tmpbuf); |
726 | } |
727 | |
728 | /* for now these all assume no locale info available for Unicode > 255 */ |
729 | |
730 | bool |
864dbfa3 |
731 | Perl_is_uni_alnum_lc(pTHX_ U32 c) |
a0ed51b3 |
732 | { |
733 | return is_uni_alnum(c); /* XXX no locale support yet */ |
734 | } |
735 | |
736 | bool |
b8c5462f |
737 | Perl_is_uni_alnumc_lc(pTHX_ U32 c) |
738 | { |
739 | return is_uni_alnumc(c); /* XXX no locale support yet */ |
740 | } |
741 | |
742 | bool |
864dbfa3 |
743 | Perl_is_uni_idfirst_lc(pTHX_ U32 c) |
a0ed51b3 |
744 | { |
745 | return is_uni_idfirst(c); /* XXX no locale support yet */ |
746 | } |
747 | |
748 | bool |
864dbfa3 |
749 | Perl_is_uni_alpha_lc(pTHX_ U32 c) |
a0ed51b3 |
750 | { |
751 | return is_uni_alpha(c); /* XXX no locale support yet */ |
752 | } |
753 | |
754 | bool |
4d61ec05 |
755 | Perl_is_uni_ascii_lc(pTHX_ U32 c) |
756 | { |
757 | return is_uni_ascii(c); /* XXX no locale support yet */ |
758 | } |
759 | |
760 | bool |
864dbfa3 |
761 | Perl_is_uni_space_lc(pTHX_ U32 c) |
a0ed51b3 |
762 | { |
763 | return is_uni_space(c); /* XXX no locale support yet */ |
764 | } |
765 | |
766 | bool |
864dbfa3 |
767 | Perl_is_uni_digit_lc(pTHX_ U32 c) |
a0ed51b3 |
768 | { |
769 | return is_uni_digit(c); /* XXX no locale support yet */ |
770 | } |
771 | |
772 | bool |
864dbfa3 |
773 | Perl_is_uni_upper_lc(pTHX_ U32 c) |
a0ed51b3 |
774 | { |
775 | return is_uni_upper(c); /* XXX no locale support yet */ |
776 | } |
777 | |
778 | bool |
864dbfa3 |
779 | Perl_is_uni_lower_lc(pTHX_ U32 c) |
a0ed51b3 |
780 | { |
781 | return is_uni_lower(c); /* XXX no locale support yet */ |
782 | } |
783 | |
784 | bool |
b8c5462f |
785 | Perl_is_uni_cntrl_lc(pTHX_ U32 c) |
786 | { |
787 | return is_uni_cntrl(c); /* XXX no locale support yet */ |
788 | } |
789 | |
790 | bool |
791 | Perl_is_uni_graph_lc(pTHX_ U32 c) |
792 | { |
793 | return is_uni_graph(c); /* XXX no locale support yet */ |
794 | } |
795 | |
796 | bool |
864dbfa3 |
797 | Perl_is_uni_print_lc(pTHX_ U32 c) |
a0ed51b3 |
798 | { |
799 | return is_uni_print(c); /* XXX no locale support yet */ |
800 | } |
801 | |
b8c5462f |
802 | bool |
803 | Perl_is_uni_punct_lc(pTHX_ U32 c) |
804 | { |
805 | return is_uni_punct(c); /* XXX no locale support yet */ |
806 | } |
807 | |
4d61ec05 |
808 | bool |
809 | Perl_is_uni_xdigit_lc(pTHX_ U32 c) |
810 | { |
811 | return is_uni_xdigit(c); /* XXX no locale support yet */ |
812 | } |
813 | |
a0ed51b3 |
814 | U32 |
864dbfa3 |
815 | Perl_to_uni_upper_lc(pTHX_ U32 c) |
a0ed51b3 |
816 | { |
817 | return to_uni_upper(c); /* XXX no locale support yet */ |
818 | } |
819 | |
820 | U32 |
864dbfa3 |
821 | Perl_to_uni_title_lc(pTHX_ U32 c) |
a0ed51b3 |
822 | { |
823 | return to_uni_title(c); /* XXX no locale support yet */ |
824 | } |
825 | |
826 | U32 |
864dbfa3 |
827 | Perl_to_uni_lower_lc(pTHX_ U32 c) |
a0ed51b3 |
828 | { |
829 | return to_uni_lower(c); /* XXX no locale support yet */ |
830 | } |
831 | |
a0ed51b3 |
832 | bool |
864dbfa3 |
833 | Perl_is_utf8_alnum(pTHX_ U8 *p) |
a0ed51b3 |
834 | { |
386d01d6 |
835 | if (!is_utf8_char(p)) |
836 | return FALSE; |
a0ed51b3 |
837 | if (!PL_utf8_alnum) |
289d4f09 |
838 | /* NOTE: "IsWord", not "IsAlnum", since Alnum is a true |
839 | * descendant of isalnum(3), in other words, it doesn't |
840 | * contain the '_'. --jhi */ |
841 | PL_utf8_alnum = swash_init("utf8", "IsWord", &PL_sv_undef, 0, 0); |
a0ed51b3 |
842 | return swash_fetch(PL_utf8_alnum, p); |
843 | /* return *p == '_' || is_utf8_alpha(p) || is_utf8_digit(p); */ |
844 | #ifdef SURPRISINGLY_SLOWER /* probably because alpha is usually true */ |
845 | if (!PL_utf8_alnum) |
846 | PL_utf8_alnum = swash_init("utf8", "", |
847 | sv_2mortal(newSVpv("+utf8::IsAlpha\n+utf8::IsDigit\n005F\n",0)), 0, 0); |
848 | return swash_fetch(PL_utf8_alnum, p); |
849 | #endif |
850 | } |
851 | |
852 | bool |
b8c5462f |
853 | Perl_is_utf8_alnumc(pTHX_ U8 *p) |
854 | { |
386d01d6 |
855 | if (!is_utf8_char(p)) |
856 | return FALSE; |
b8c5462f |
857 | if (!PL_utf8_alnum) |
858 | PL_utf8_alnum = swash_init("utf8", "IsAlnumC", &PL_sv_undef, 0, 0); |
859 | return swash_fetch(PL_utf8_alnum, p); |
860 | /* return is_utf8_alpha(p) || is_utf8_digit(p); */ |
861 | #ifdef SURPRISINGLY_SLOWER /* probably because alpha is usually true */ |
862 | if (!PL_utf8_alnum) |
863 | PL_utf8_alnum = swash_init("utf8", "", |
864 | sv_2mortal(newSVpv("+utf8::IsAlpha\n+utf8::IsDigit\n005F\n",0)), 0, 0); |
865 | return swash_fetch(PL_utf8_alnum, p); |
866 | #endif |
867 | } |
868 | |
869 | bool |
864dbfa3 |
870 | Perl_is_utf8_idfirst(pTHX_ U8 *p) |
a0ed51b3 |
871 | { |
872 | return *p == '_' || is_utf8_alpha(p); |
873 | } |
874 | |
875 | bool |
864dbfa3 |
876 | Perl_is_utf8_alpha(pTHX_ U8 *p) |
a0ed51b3 |
877 | { |
386d01d6 |
878 | if (!is_utf8_char(p)) |
879 | return FALSE; |
a0ed51b3 |
880 | if (!PL_utf8_alpha) |
e24b16f9 |
881 | PL_utf8_alpha = swash_init("utf8", "IsAlpha", &PL_sv_undef, 0, 0); |
a0ed51b3 |
882 | return swash_fetch(PL_utf8_alpha, p); |
883 | } |
884 | |
885 | bool |
b8c5462f |
886 | Perl_is_utf8_ascii(pTHX_ U8 *p) |
887 | { |
386d01d6 |
888 | if (!is_utf8_char(p)) |
889 | return FALSE; |
b8c5462f |
890 | if (!PL_utf8_ascii) |
891 | PL_utf8_ascii = swash_init("utf8", "IsAscii", &PL_sv_undef, 0, 0); |
892 | return swash_fetch(PL_utf8_ascii, p); |
893 | } |
894 | |
895 | bool |
864dbfa3 |
896 | Perl_is_utf8_space(pTHX_ U8 *p) |
a0ed51b3 |
897 | { |
386d01d6 |
898 | if (!is_utf8_char(p)) |
899 | return FALSE; |
a0ed51b3 |
900 | if (!PL_utf8_space) |
e24b16f9 |
901 | PL_utf8_space = swash_init("utf8", "IsSpace", &PL_sv_undef, 0, 0); |
a0ed51b3 |
902 | return swash_fetch(PL_utf8_space, p); |
903 | } |
904 | |
905 | bool |
864dbfa3 |
906 | Perl_is_utf8_digit(pTHX_ U8 *p) |
a0ed51b3 |
907 | { |
386d01d6 |
908 | if (!is_utf8_char(p)) |
909 | return FALSE; |
a0ed51b3 |
910 | if (!PL_utf8_digit) |
e24b16f9 |
911 | PL_utf8_digit = swash_init("utf8", "IsDigit", &PL_sv_undef, 0, 0); |
a0ed51b3 |
912 | return swash_fetch(PL_utf8_digit, p); |
913 | } |
914 | |
915 | bool |
864dbfa3 |
916 | Perl_is_utf8_upper(pTHX_ U8 *p) |
a0ed51b3 |
917 | { |
386d01d6 |
918 | if (!is_utf8_char(p)) |
919 | return FALSE; |
a0ed51b3 |
920 | if (!PL_utf8_upper) |
e24b16f9 |
921 | PL_utf8_upper = swash_init("utf8", "IsUpper", &PL_sv_undef, 0, 0); |
a0ed51b3 |
922 | return swash_fetch(PL_utf8_upper, p); |
923 | } |
924 | |
925 | bool |
864dbfa3 |
926 | Perl_is_utf8_lower(pTHX_ U8 *p) |
a0ed51b3 |
927 | { |
386d01d6 |
928 | if (!is_utf8_char(p)) |
929 | return FALSE; |
a0ed51b3 |
930 | if (!PL_utf8_lower) |
e24b16f9 |
931 | PL_utf8_lower = swash_init("utf8", "IsLower", &PL_sv_undef, 0, 0); |
a0ed51b3 |
932 | return swash_fetch(PL_utf8_lower, p); |
933 | } |
934 | |
935 | bool |
b8c5462f |
936 | Perl_is_utf8_cntrl(pTHX_ U8 *p) |
937 | { |
386d01d6 |
938 | if (!is_utf8_char(p)) |
939 | return FALSE; |
b8c5462f |
940 | if (!PL_utf8_cntrl) |
941 | PL_utf8_cntrl = swash_init("utf8", "IsCntrl", &PL_sv_undef, 0, 0); |
942 | return swash_fetch(PL_utf8_cntrl, p); |
943 | } |
944 | |
945 | bool |
946 | Perl_is_utf8_graph(pTHX_ U8 *p) |
947 | { |
386d01d6 |
948 | if (!is_utf8_char(p)) |
949 | return FALSE; |
b8c5462f |
950 | if (!PL_utf8_graph) |
951 | PL_utf8_graph = swash_init("utf8", "IsGraph", &PL_sv_undef, 0, 0); |
952 | return swash_fetch(PL_utf8_graph, p); |
953 | } |
954 | |
955 | bool |
864dbfa3 |
956 | Perl_is_utf8_print(pTHX_ U8 *p) |
a0ed51b3 |
957 | { |
386d01d6 |
958 | if (!is_utf8_char(p)) |
959 | return FALSE; |
a0ed51b3 |
960 | if (!PL_utf8_print) |
e24b16f9 |
961 | PL_utf8_print = swash_init("utf8", "IsPrint", &PL_sv_undef, 0, 0); |
a0ed51b3 |
962 | return swash_fetch(PL_utf8_print, p); |
963 | } |
964 | |
965 | bool |
b8c5462f |
966 | Perl_is_utf8_punct(pTHX_ U8 *p) |
967 | { |
386d01d6 |
968 | if (!is_utf8_char(p)) |
969 | return FALSE; |
b8c5462f |
970 | if (!PL_utf8_punct) |
971 | PL_utf8_punct = swash_init("utf8", "IsPunct", &PL_sv_undef, 0, 0); |
972 | return swash_fetch(PL_utf8_punct, p); |
973 | } |
974 | |
975 | bool |
976 | Perl_is_utf8_xdigit(pTHX_ U8 *p) |
977 | { |
386d01d6 |
978 | if (!is_utf8_char(p)) |
979 | return FALSE; |
b8c5462f |
980 | if (!PL_utf8_xdigit) |
981 | PL_utf8_xdigit = swash_init("utf8", "IsXDigit", &PL_sv_undef, 0, 0); |
982 | return swash_fetch(PL_utf8_xdigit, p); |
983 | } |
984 | |
985 | bool |
864dbfa3 |
986 | Perl_is_utf8_mark(pTHX_ U8 *p) |
a0ed51b3 |
987 | { |
386d01d6 |
988 | if (!is_utf8_char(p)) |
989 | return FALSE; |
a0ed51b3 |
990 | if (!PL_utf8_mark) |
e24b16f9 |
991 | PL_utf8_mark = swash_init("utf8", "IsM", &PL_sv_undef, 0, 0); |
a0ed51b3 |
992 | return swash_fetch(PL_utf8_mark, p); |
993 | } |
994 | |
2104c8d9 |
995 | UV |
864dbfa3 |
996 | Perl_to_utf8_upper(pTHX_ U8 *p) |
a0ed51b3 |
997 | { |
998 | UV uv; |
999 | |
1000 | if (!PL_utf8_toupper) |
e24b16f9 |
1001 | PL_utf8_toupper = swash_init("utf8", "ToUpper", &PL_sv_undef, 4, 0); |
a0ed51b3 |
1002 | uv = swash_fetch(PL_utf8_toupper, p); |
756820e3 |
1003 | return uv ? uv : utf8_to_uv(p,UTF8_MAXLEN,0,0); |
a0ed51b3 |
1004 | } |
1005 | |
2104c8d9 |
1006 | UV |
864dbfa3 |
1007 | Perl_to_utf8_title(pTHX_ U8 *p) |
a0ed51b3 |
1008 | { |
1009 | UV uv; |
1010 | |
1011 | if (!PL_utf8_totitle) |
e24b16f9 |
1012 | PL_utf8_totitle = swash_init("utf8", "ToTitle", &PL_sv_undef, 4, 0); |
a0ed51b3 |
1013 | uv = swash_fetch(PL_utf8_totitle, p); |
756820e3 |
1014 | return uv ? uv : utf8_to_uv(p,UTF8_MAXLEN,0,0); |
a0ed51b3 |
1015 | } |
1016 | |
2104c8d9 |
1017 | UV |
864dbfa3 |
1018 | Perl_to_utf8_lower(pTHX_ U8 *p) |
a0ed51b3 |
1019 | { |
1020 | UV uv; |
1021 | |
1022 | if (!PL_utf8_tolower) |
e24b16f9 |
1023 | PL_utf8_tolower = swash_init("utf8", "ToLower", &PL_sv_undef, 4, 0); |
a0ed51b3 |
1024 | uv = swash_fetch(PL_utf8_tolower, p); |
756820e3 |
1025 | return uv ? uv : utf8_to_uv(p,UTF8_MAXLEN,0,0); |
a0ed51b3 |
1026 | } |
1027 | |
1028 | /* a "swash" is a swatch hash */ |
1029 | |
1030 | SV* |
864dbfa3 |
1031 | Perl_swash_init(pTHX_ char* pkg, char* name, SV *listsv, I32 minbits, I32 none) |
a0ed51b3 |
1032 | { |
1033 | SV* retval; |
1034 | char tmpbuf[256]; |
8e84507e |
1035 | dSP; |
ce3b816e |
1036 | |
1037 | if (!gv_stashpv(pkg, 0)) { /* demand load utf8 */ |
1038 | ENTER; |
1039 | Perl_load_module(aTHX_ PERL_LOADMOD_NOIMPORT, newSVpv(pkg,0), Nullsv); |
1040 | LEAVE; |
1041 | } |
1042 | SPAGAIN; |
a0ed51b3 |
1043 | PUSHSTACKi(PERLSI_MAGIC); |
1044 | PUSHMARK(SP); |
1045 | EXTEND(SP,5); |
1046 | PUSHs(sv_2mortal(newSVpvn(pkg, strlen(pkg)))); |
1047 | PUSHs(sv_2mortal(newSVpvn(name, strlen(name)))); |
1048 | PUSHs(listsv); |
1049 | PUSHs(sv_2mortal(newSViv(minbits))); |
1050 | PUSHs(sv_2mortal(newSViv(none))); |
1051 | PUTBACK; |
1052 | ENTER; |
1053 | SAVEI32(PL_hints); |
1054 | PL_hints = 0; |
1055 | save_re_context(); |
e24b16f9 |
1056 | if (PL_curcop == &PL_compiling) /* XXX ought to be handled by lex_start */ |
a0ed51b3 |
1057 | strncpy(tmpbuf, PL_tokenbuf, sizeof tmpbuf); |
864dbfa3 |
1058 | if (call_method("SWASHNEW", G_SCALAR)) |
8e84507e |
1059 | retval = newSVsv(*PL_stack_sp--); |
a0ed51b3 |
1060 | else |
e24b16f9 |
1061 | retval = &PL_sv_undef; |
a0ed51b3 |
1062 | LEAVE; |
1063 | POPSTACK; |
e24b16f9 |
1064 | if (PL_curcop == &PL_compiling) { |
a0ed51b3 |
1065 | strncpy(PL_tokenbuf, tmpbuf, sizeof tmpbuf); |
e24b16f9 |
1066 | PL_curcop->op_private = PL_hints; |
a0ed51b3 |
1067 | } |
1068 | if (!SvROK(retval) || SvTYPE(SvRV(retval)) != SVt_PVHV) |
cea2e8a9 |
1069 | Perl_croak(aTHX_ "SWASHNEW didn't return an HV ref"); |
a0ed51b3 |
1070 | return retval; |
1071 | } |
1072 | |
1073 | UV |
864dbfa3 |
1074 | Perl_swash_fetch(pTHX_ SV *sv, U8 *ptr) |
a0ed51b3 |
1075 | { |
1076 | HV* hv = (HV*)SvRV(sv); |
1077 | U32 klen = UTF8SKIP(ptr) - 1; |
1078 | U32 off = ptr[klen] & 127; /* NB: 64 bit always 0 when len > 1 */ |
1079 | STRLEN slen; |
1080 | STRLEN needents = (klen ? 64 : 128); |
dfe13c55 |
1081 | U8 *tmps; |
a0ed51b3 |
1082 | U32 bit; |
1083 | SV *retval; |
1084 | |
1085 | /* |
1086 | * This single-entry cache saves about 1/3 of the utf8 overhead in test |
1087 | * suite. (That is, only 7-8% overall over just a hash cache. Still, |
1088 | * it's nothing to sniff at.) Pity we usually come through at least |
1089 | * two function calls to get here... |
1090 | * |
1091 | * NB: this code assumes that swatches are never modified, once generated! |
1092 | */ |
1093 | |
1094 | if (hv == PL_last_swash_hv && |
1095 | klen == PL_last_swash_klen && |
12ae5dfc |
1096 | (!klen || memEQ((char *)ptr,(char *)PL_last_swash_key,klen)) ) |
a0ed51b3 |
1097 | { |
1098 | tmps = PL_last_swash_tmps; |
1099 | slen = PL_last_swash_slen; |
1100 | } |
1101 | else { |
1102 | /* Try our second-level swatch cache, kept in a hash. */ |
dfe13c55 |
1103 | SV** svp = hv_fetch(hv, (char*)ptr, klen, FALSE); |
a0ed51b3 |
1104 | |
1105 | /* If not cached, generate it via utf8::SWASHGET */ |
dfe13c55 |
1106 | if (!svp || !SvPOK(*svp) || !(tmps = (U8*)SvPV(*svp, slen))) { |
a0ed51b3 |
1107 | dSP; |
1108 | ENTER; |
1109 | SAVETMPS; |
1110 | save_re_context(); |
1111 | PUSHSTACKi(PERLSI_MAGIC); |
1112 | PUSHMARK(SP); |
1113 | EXTEND(SP,3); |
1114 | PUSHs((SV*)sv); |
756820e3 |
1115 | PUSHs(sv_2mortal(newSViv(utf8_to_uv(ptr, UTF8_MAXLEN, 0, 0) & ~(needents - 1)))); |
a0ed51b3 |
1116 | PUSHs(sv_2mortal(newSViv(needents))); |
1117 | PUTBACK; |
864dbfa3 |
1118 | if (call_method("SWASHGET", G_SCALAR)) |
8e84507e |
1119 | retval = newSVsv(*PL_stack_sp--); |
a0ed51b3 |
1120 | else |
e24b16f9 |
1121 | retval = &PL_sv_undef; |
a0ed51b3 |
1122 | POPSTACK; |
1123 | FREETMPS; |
1124 | LEAVE; |
e24b16f9 |
1125 | if (PL_curcop == &PL_compiling) |
1126 | PL_curcop->op_private = PL_hints; |
a0ed51b3 |
1127 | |
dfe13c55 |
1128 | svp = hv_store(hv, (char*)ptr, klen, retval, 0); |
a0ed51b3 |
1129 | |
dfe13c55 |
1130 | if (!svp || !(tmps = (U8*)SvPV(*svp, slen)) || slen < 8) |
cea2e8a9 |
1131 | Perl_croak(aTHX_ "SWASHGET didn't return result of proper length"); |
a0ed51b3 |
1132 | } |
1133 | |
1134 | PL_last_swash_hv = hv; |
1135 | PL_last_swash_klen = klen; |
1136 | PL_last_swash_tmps = tmps; |
1137 | PL_last_swash_slen = slen; |
1138 | if (klen) |
1139 | Copy(ptr, PL_last_swash_key, klen, U8); |
1140 | } |
1141 | |
9faf8d75 |
1142 | switch ((int)((slen << 3) / needents)) { |
a0ed51b3 |
1143 | case 1: |
1144 | bit = 1 << (off & 7); |
1145 | off >>= 3; |
1146 | return (tmps[off] & bit) != 0; |
1147 | case 8: |
1148 | return tmps[off]; |
1149 | case 16: |
1150 | off <<= 1; |
1151 | return (tmps[off] << 8) + tmps[off + 1] ; |
1152 | case 32: |
1153 | off <<= 2; |
1154 | return (tmps[off] << 24) + (tmps[off+1] << 16) + (tmps[off+2] << 8) + tmps[off + 3] ; |
1155 | } |
cea2e8a9 |
1156 | Perl_croak(aTHX_ "panic: swash_fetch"); |
a0ed51b3 |
1157 | return 0; |
1158 | } |