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 * |
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 | |
067a85ef |
122 | if (u <= 0x7f) |
386d01d6 |
123 | return 1; |
124 | |
067a85ef |
125 | if (u >= 0x80 && u <= 0xbf) |
386d01d6 |
126 | return 0; |
127 | |
9f07fdcd |
128 | len = UTF8SKIP(s); |
386d01d6 |
129 | |
067a85ef |
130 | if (len < 2 || (u >= 0xc0 && u <= 0xfd && s[1] < 0x80)) |
131 | return 0; |
132 | |
386d01d6 |
133 | slen = len - 1; |
134 | s++; |
067a85ef |
135 | uv = u; |
136 | ouv = uv; |
386d01d6 |
137 | while (slen--) { |
138 | if ((*s & 0xc0) != 0x80) |
139 | return 0; |
067a85ef |
140 | uv = (uv << 6) | (*s & 0x3f); |
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; |
166 | U8* send = s + len; |
167 | STRLEN c; |
168 | |
6662521e |
169 | while (x < send) { |
170 | c = is_utf8_char(x); |
067a85ef |
171 | if (!c) |
172 | return FALSE; |
6662521e |
173 | x += c; |
067a85ef |
174 | if (x > send) |
175 | return FALSE; |
6662521e |
176 | } |
067a85ef |
177 | |
178 | return TRUE; |
6662521e |
179 | } |
180 | |
67e989fb |
181 | /* |
be2c7115 |
182 | =for apidoc Am|U8* s|utf8_to_uv|STRLEN curlen|STRLEN *retlen|U32 flags |
67e989fb |
183 | |
184 | Returns the character value of the first character in the string C<s> |
ba210ebe |
185 | which is assumed to be in UTF8 encoding and no longer than C<curlen>; |
186 | C<retlen> will be set to the length, in bytes, of that character, |
187 | and the pointer C<s> will be advanced to the end of the character. |
67e989fb |
188 | |
189 | If C<s> does not point to a well-formed UTF8 character, the behaviour |
dcad2880 |
190 | is dependent on the value of C<flags>: if it contains UTF8_CHECK_ONLY, |
191 | it is assumed that the caller will raise a warning, and this function |
192 | will set C<retlen> to C<-1> and return. The C<flags> can also contain |
be2c7115 |
193 | various flags to allow deviations from the strict UTF-8 encoding |
194 | (see F<utf8.h>). |
67e989fb |
195 | |
dcad2880 |
196 | =cut */ |
67e989fb |
197 | |
a0ed51b3 |
198 | UV |
dcad2880 |
199 | Perl_utf8_to_uv(pTHX_ U8* s, STRLEN curlen, STRLEN* retlen, U32 flags) |
a0ed51b3 |
200 | { |
ba210ebe |
201 | UV uv = *s, ouv; |
202 | STRLEN len = 1; |
7bf1b6bb |
203 | #ifdef EBCDIC |
204 | bool dowarn = 0; |
205 | #else |
ba210ebe |
206 | bool dowarn = ckWARN_d(WARN_UTF8); |
7bf1b6bb |
207 | #endif |
ba210ebe |
208 | STRLEN expectlen = 0; |
209 | |
0c443dc2 |
210 | if (curlen == 0) { |
211 | if (dowarn) |
212 | Perl_warner(aTHX_ WARN_UTF8, |
213 | "Malformed UTF-8 character (an empty string)"); |
214 | goto malformed; |
215 | } |
216 | |
ba210ebe |
217 | if (uv <= 0x7f) { /* Pure ASCII. */ |
a0ed51b3 |
218 | if (retlen) |
219 | *retlen = 1; |
220 | return *s; |
221 | } |
67e989fb |
222 | |
fcc8fcf6 |
223 | if ((uv >= 0x80 && uv <= 0xbf) && |
224 | !(flags & UTF8_ALLOW_CONTINUATION)) { |
ba210ebe |
225 | if (dowarn) |
226 | Perl_warner(aTHX_ WARN_UTF8, |
efbcad09 |
227 | "Malformed UTF-8 character (unexpected continuation byte 0x%02"UVxf")", |
ba210ebe |
228 | uv); |
229 | goto malformed; |
230 | } |
231 | |
0c443dc2 |
232 | if ((uv >= 0xc0 && uv <= 0xfd && curlen > 1 && s[1] < 0x80) && |
fcc8fcf6 |
233 | !(flags & UTF8_ALLOW_NON_CONTINUATION)) { |
ba210ebe |
234 | if (dowarn) |
235 | Perl_warner(aTHX_ WARN_UTF8, |
efbcad09 |
236 | "Malformed UTF-8 character (unexpected non-continuation byte 0x%02"UVxf" after byte 0x%02"UVxf")", |
9e29e5ff |
237 | (UV)s[1], uv); |
ba210ebe |
238 | goto malformed; |
239 | } |
fcc8fcf6 |
240 | |
241 | if ((uv == 0xfe || uv == 0xff) && |
242 | !(flags & UTF8_ALLOW_FE_FF)) { |
ba210ebe |
243 | if (dowarn) |
244 | Perl_warner(aTHX_ WARN_UTF8, |
efbcad09 |
245 | "Malformed UTF-8 character (byte 0x%02"UVxf")", |
ba210ebe |
246 | uv); |
247 | goto malformed; |
a0ed51b3 |
248 | } |
fcc8fcf6 |
249 | |
ba210ebe |
250 | if (!(uv & 0x20)) { len = 2; uv &= 0x1f; } |
251 | else if (!(uv & 0x10)) { len = 3; uv &= 0x0f; } |
252 | else if (!(uv & 0x08)) { len = 4; uv &= 0x07; } |
253 | else if (!(uv & 0x04)) { len = 5; uv &= 0x03; } |
254 | else if (!(uv & 0x02)) { len = 6; uv &= 0x01; } |
255 | else if (!(uv & 0x01)) { len = 7; uv = 0; } |
3c77ea2b |
256 | else { len = 13; uv = 0; } /* whoa! */ |
fcc8fcf6 |
257 | |
a0ed51b3 |
258 | if (retlen) |
259 | *retlen = len; |
ba210ebe |
260 | |
261 | expectlen = len; |
262 | |
fcc8fcf6 |
263 | if ((curlen < expectlen) && |
264 | !(flags & UTF8_ALLOW_SHORT)) { |
ba210ebe |
265 | if (dowarn) |
266 | Perl_warner(aTHX_ WARN_UTF8, |
267 | "Malformed UTF-8 character (%d byte%s, need %d)", |
0c443dc2 |
268 | curlen, curlen == 1 ? "" : "s", expectlen); |
ba210ebe |
269 | goto malformed; |
270 | } |
271 | |
272 | len--; |
a0ed51b3 |
273 | s++; |
ba210ebe |
274 | ouv = uv; |
275 | |
a0ed51b3 |
276 | while (len--) { |
277 | if ((*s & 0xc0) != 0x80) { |
ba210ebe |
278 | if (dowarn) |
279 | Perl_warner(aTHX_ WARN_UTF8, |
280 | "Malformed UTF-8 character (unexpected continuation byte 0x%02x)", |
281 | *s); |
282 | goto malformed; |
a0ed51b3 |
283 | } |
284 | else |
ba210ebe |
285 | uv = (uv << 6) | (*s & 0x3f); |
286 | if (uv < ouv) { |
fcc8fcf6 |
287 | /* This cannot be allowed. */ |
ba210ebe |
288 | if (dowarn) |
289 | Perl_warner(aTHX_ WARN_UTF8, |
290 | "Malformed UTF-8 character (overflow at 0x%"UVxf", byte 0x%02x)", |
291 | ouv, *s); |
292 | goto malformed; |
293 | } |
294 | s++; |
295 | ouv = uv; |
296 | } |
297 | |
fcc8fcf6 |
298 | if ((uv >= 0xd800 && uv <= 0xdfff) && |
299 | !(flags & UTF8_ALLOW_SURROGATE)) { |
ba210ebe |
300 | if (dowarn) |
301 | Perl_warner(aTHX_ WARN_UTF8, |
302 | "Malformed UTF-8 character (UTF-16 surrogate 0x%04"UVxf")", |
303 | uv); |
304 | goto malformed; |
fcc8fcf6 |
305 | } else if ((uv == 0xfffe) && |
306 | !(flags & UTF8_ALLOW_BOM)) { |
ba210ebe |
307 | if (dowarn) |
308 | Perl_warner(aTHX_ WARN_UTF8, |
309 | "Malformed UTF-8 character (byte order mark 0x%04"UVxf")", |
310 | uv); |
311 | goto malformed; |
fcc8fcf6 |
312 | } else if ((uv == 0xffff) && |
313 | !(flags & UTF8_ALLOW_FFFF)) { |
ba210ebe |
314 | if (dowarn) |
315 | Perl_warner(aTHX_ WARN_UTF8, |
fcc8fcf6 |
316 | "Malformed UTF-8 character (character 0x%04"UVxf")", |
ba210ebe |
317 | uv); |
318 | goto malformed; |
fcc8fcf6 |
319 | } else if ((expectlen > UNISKIP(uv)) && |
320 | !(flags & UTF8_ALLOW_LONG)) { |
ba210ebe |
321 | if (dowarn) |
322 | Perl_warner(aTHX_ WARN_UTF8, |
323 | "Malformed UTF-8 character (%d byte%s, need %d)", |
0c443dc2 |
324 | expectlen, expectlen == 1 ? "": "s", UNISKIP(uv)); |
ba210ebe |
325 | goto malformed; |
a0ed51b3 |
326 | } |
ba210ebe |
327 | |
a0ed51b3 |
328 | return uv; |
ba210ebe |
329 | |
330 | malformed: |
331 | |
fcc8fcf6 |
332 | if (flags & UTF8_CHECK_ONLY) { |
ba210ebe |
333 | if (retlen) |
cc366d4b |
334 | *retlen = -1; |
ba210ebe |
335 | return 0; |
336 | } |
337 | |
338 | if (retlen) |
cc366d4b |
339 | *retlen = expectlen ? expectlen : len; |
ba210ebe |
340 | |
341 | return UNICODE_REPLACEMENT_CHARACTER; |
a0ed51b3 |
342 | } |
343 | |
8e84507e |
344 | /* |
dcad2880 |
345 | =for apidoc Am|U8* s|utf8_to_uv_simple|STRLEN *retlen |
8e84507e |
346 | |
347 | Returns the character value of the first character in the string C<s> |
348 | which is assumed to be in UTF8 encoding; C<retlen> will be set to the |
349 | length, in bytes, of that character, and the pointer C<s> will be |
350 | advanced to the end of the character. |
351 | |
ba210ebe |
352 | If C<s> does not point to a well-formed UTF8 character, zero is |
353 | returned and retlen is set, if possible, to -1. |
8e84507e |
354 | |
355 | =cut |
356 | */ |
357 | |
358 | UV |
dcad2880 |
359 | Perl_utf8_to_uv_simple(pTHX_ U8* s, STRLEN* retlen) |
8e84507e |
360 | { |
2e4dc9fc |
361 | return Perl_utf8_to_uv(aTHX_ s, UTF8_MAXLEN, retlen, 0); |
8e84507e |
362 | } |
363 | |
b76347f2 |
364 | /* |
365 | =for apidoc|utf8_length|U8 *s|U8 *e |
366 | |
367 | Return the length of the UTF-8 char encoded string C<s> in characters. |
02eb7b47 |
368 | Stops at C<e> (inclusive). If C<e E<lt> s> or if the scan would end |
369 | up past C<e>, croaks. |
b76347f2 |
370 | |
371 | =cut |
372 | */ |
373 | |
374 | STRLEN |
375 | Perl_utf8_length(pTHX_ U8* s, U8* e) |
376 | { |
377 | STRLEN len = 0; |
378 | |
379 | if (e < s) |
02eb7b47 |
380 | Perl_croak(aTHX_ "panic: utf8_length: unexpected end"); |
b76347f2 |
381 | while (s < e) { |
02eb7b47 |
382 | U8 t = UTF8SKIP(s); |
b76347f2 |
383 | |
384 | if (e - s < t) |
02eb7b47 |
385 | Perl_croak(aTHX_ "panic: utf8_length: unaligned end"); |
b76347f2 |
386 | s += t; |
387 | len++; |
388 | } |
389 | |
390 | return len; |
391 | } |
392 | |
246fae53 |
393 | /* utf8_distance(a,b) returns the number of UTF8 characters between |
394 | the pointers a and b */ |
a0ed51b3 |
395 | |
02eb7b47 |
396 | IV |
864dbfa3 |
397 | Perl_utf8_distance(pTHX_ U8 *a, U8 *b) |
a0ed51b3 |
398 | { |
02eb7b47 |
399 | IV off = 0; |
400 | |
a0ed51b3 |
401 | if (a < b) { |
402 | while (a < b) { |
02eb7b47 |
403 | U8 c = UTF8SKIP(a); |
404 | |
405 | if (b - a < c) |
406 | Perl_croak(aTHX_ "panic: utf8_distance: unaligned end"); |
407 | a += c; |
a0ed51b3 |
408 | off--; |
409 | } |
410 | } |
411 | else { |
412 | while (b < a) { |
02eb7b47 |
413 | U8 c = UTF8SKIP(b); |
414 | |
415 | if (a - b < c) |
416 | Perl_croak(aTHX_ "panic: utf8_distance: unaligned end"); |
417 | b += c; |
a0ed51b3 |
418 | off++; |
419 | } |
420 | } |
02eb7b47 |
421 | |
a0ed51b3 |
422 | return off; |
423 | } |
424 | |
425 | /* WARNING: do not use the following unless you *know* off is within bounds */ |
426 | |
427 | U8 * |
864dbfa3 |
428 | Perl_utf8_hop(pTHX_ U8 *s, I32 off) |
a0ed51b3 |
429 | { |
430 | if (off >= 0) { |
431 | while (off--) |
432 | s += UTF8SKIP(s); |
433 | } |
434 | else { |
435 | while (off++) { |
436 | s--; |
437 | if (*s & 0x80) { |
438 | while ((*s & 0xc0) == 0x80) |
439 | s--; |
440 | } |
441 | } |
442 | } |
443 | return s; |
444 | } |
445 | |
6940069f |
446 | /* |
246fae53 |
447 | =for apidoc Am|U8 *|utf8_to_bytes|U8 *s|STRLEN *len |
6940069f |
448 | |
246fae53 |
449 | Converts a string C<s> of length C<len> from UTF8 into byte encoding. |
450 | Unlike C<bytes_to_utf8>, this over-writes the original string, and |
451 | updates len to contain the new length. |
67e989fb |
452 | Returns zero on failure, setting C<len> to -1. |
6940069f |
453 | |
454 | =cut |
455 | */ |
456 | |
457 | U8 * |
246fae53 |
458 | Perl_utf8_to_bytes(pTHX_ U8* s, STRLEN *len) |
6940069f |
459 | { |
6940069f |
460 | U8 *send; |
461 | U8 *d; |
dcad2880 |
462 | U8 *save = s; |
246fae53 |
463 | |
464 | /* ensure valid UTF8 and chars < 256 before updating string */ |
dcad2880 |
465 | for (send = s + *len; s < send; ) { |
466 | U8 c = *s++; |
467 | |
9f9ab905 |
468 | if (c >= 0x80 && |
dcad2880 |
469 | ((s >= send) || |
470 | ((*s++ & 0xc0) != 0x80) || ((c & 0xfe) != 0xc2))) { |
471 | *len = -1; |
472 | return 0; |
473 | } |
246fae53 |
474 | } |
dcad2880 |
475 | |
476 | d = s = save; |
6940069f |
477 | while (s < send) { |
dcad2880 |
478 | if (*s < 0x80) { |
479 | *d++ = *s++; |
480 | } |
6940069f |
481 | else { |
ba210ebe |
482 | STRLEN ulen; |
dcad2880 |
483 | *d++ = (U8)utf8_to_uv_simple(s, &ulen); |
6940069f |
484 | s += ulen; |
6940069f |
485 | } |
486 | } |
487 | *d = '\0'; |
246fae53 |
488 | *len = d - save; |
6940069f |
489 | return save; |
490 | } |
491 | |
492 | /* |
6662521e |
493 | =for apidoc Am|U8 *|bytes_to_utf8|U8 *s|STRLEN *len |
6940069f |
494 | |
495 | Converts a string C<s> of length C<len> from ASCII into UTF8 encoding. |
6662521e |
496 | Returns a pointer to the newly-created string, and sets C<len> to |
497 | reflect the new length. |
6940069f |
498 | |
497711e7 |
499 | =cut |
6940069f |
500 | */ |
501 | |
502 | U8* |
6662521e |
503 | Perl_bytes_to_utf8(pTHX_ U8* s, STRLEN *len) |
6940069f |
504 | { |
6940069f |
505 | U8 *send; |
506 | U8 *d; |
507 | U8 *dst; |
6662521e |
508 | send = s + (*len); |
6940069f |
509 | |
6662521e |
510 | Newz(801, d, (*len) * 2 + 1, U8); |
6940069f |
511 | dst = d; |
512 | |
513 | while (s < send) { |
514 | if (*s < 0x80) |
515 | *d++ = *s++; |
516 | else { |
517 | UV uv = *s++; |
518 | *d++ = (( uv >> 6) | 0xc0); |
519 | *d++ = (( uv & 0x3f) | 0x80); |
520 | } |
521 | } |
522 | *d = '\0'; |
6662521e |
523 | *len = d-dst; |
6940069f |
524 | return dst; |
525 | } |
526 | |
a0ed51b3 |
527 | /* |
dea0fc0b |
528 | * Convert native (big-endian) or reversed (little-endian) UTF-16 to UTF-8. |
a0ed51b3 |
529 | * |
530 | * Destination must be pre-extended to 3/2 source. Do not use in-place. |
531 | * We optimize for native, for obvious reasons. */ |
532 | |
533 | U8* |
dea0fc0b |
534 | Perl_utf16_to_utf8(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen) |
a0ed51b3 |
535 | { |
dea0fc0b |
536 | U8* pend; |
537 | U8* dstart = d; |
538 | |
539 | if (bytelen & 1) |
a7867d0a |
540 | Perl_croak(aTHX_ "panic: utf16_to_utf8: odd bytelen"); |
dea0fc0b |
541 | |
542 | pend = p + bytelen; |
543 | |
a0ed51b3 |
544 | while (p < pend) { |
dea0fc0b |
545 | UV uv = (p[0] << 8) + p[1]; /* UTF-16BE */ |
546 | p += 2; |
a0ed51b3 |
547 | if (uv < 0x80) { |
548 | *d++ = uv; |
549 | continue; |
550 | } |
551 | if (uv < 0x800) { |
552 | *d++ = (( uv >> 6) | 0xc0); |
553 | *d++ = (( uv & 0x3f) | 0x80); |
554 | continue; |
555 | } |
556 | if (uv >= 0xd800 && uv < 0xdbff) { /* surrogates */ |
dea0fc0b |
557 | UV low = *p++; |
558 | if (low < 0xdc00 || low >= 0xdfff) |
559 | Perl_croak(aTHX_ "Malformed UTF-16 surrogate"); |
a0ed51b3 |
560 | uv = ((uv - 0xd800) << 10) + (low - 0xdc00) + 0x10000; |
561 | } |
562 | if (uv < 0x10000) { |
563 | *d++ = (( uv >> 12) | 0xe0); |
564 | *d++ = (((uv >> 6) & 0x3f) | 0x80); |
565 | *d++ = (( uv & 0x3f) | 0x80); |
566 | continue; |
567 | } |
568 | else { |
569 | *d++ = (( uv >> 18) | 0xf0); |
570 | *d++ = (((uv >> 12) & 0x3f) | 0x80); |
571 | *d++ = (((uv >> 6) & 0x3f) | 0x80); |
572 | *d++ = (( uv & 0x3f) | 0x80); |
573 | continue; |
574 | } |
575 | } |
dea0fc0b |
576 | *newlen = d - dstart; |
a0ed51b3 |
577 | return d; |
578 | } |
579 | |
580 | /* Note: this one is slightly destructive of the source. */ |
581 | |
582 | U8* |
dea0fc0b |
583 | Perl_utf16_to_utf8_reversed(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen) |
a0ed51b3 |
584 | { |
585 | U8* s = (U8*)p; |
586 | U8* send = s + bytelen; |
587 | while (s < send) { |
588 | U8 tmp = s[0]; |
589 | s[0] = s[1]; |
590 | s[1] = tmp; |
591 | s += 2; |
592 | } |
dea0fc0b |
593 | return utf16_to_utf8(p, d, bytelen, newlen); |
a0ed51b3 |
594 | } |
595 | |
596 | /* for now these are all defined (inefficiently) in terms of the utf8 versions */ |
597 | |
598 | bool |
864dbfa3 |
599 | Perl_is_uni_alnum(pTHX_ U32 c) |
a0ed51b3 |
600 | { |
ad391ad9 |
601 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
602 | uv_to_utf8(tmpbuf, (UV)c); |
603 | return is_utf8_alnum(tmpbuf); |
604 | } |
605 | |
606 | bool |
b8c5462f |
607 | Perl_is_uni_alnumc(pTHX_ U32 c) |
608 | { |
ad391ad9 |
609 | U8 tmpbuf[UTF8_MAXLEN+1]; |
b8c5462f |
610 | uv_to_utf8(tmpbuf, (UV)c); |
611 | return is_utf8_alnumc(tmpbuf); |
612 | } |
613 | |
614 | bool |
864dbfa3 |
615 | Perl_is_uni_idfirst(pTHX_ U32 c) |
a0ed51b3 |
616 | { |
ad391ad9 |
617 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
618 | uv_to_utf8(tmpbuf, (UV)c); |
619 | return is_utf8_idfirst(tmpbuf); |
620 | } |
621 | |
622 | bool |
864dbfa3 |
623 | Perl_is_uni_alpha(pTHX_ U32 c) |
a0ed51b3 |
624 | { |
ad391ad9 |
625 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
626 | uv_to_utf8(tmpbuf, (UV)c); |
627 | return is_utf8_alpha(tmpbuf); |
628 | } |
629 | |
630 | bool |
4d61ec05 |
631 | Perl_is_uni_ascii(pTHX_ U32 c) |
632 | { |
ad391ad9 |
633 | U8 tmpbuf[UTF8_MAXLEN+1]; |
4d61ec05 |
634 | uv_to_utf8(tmpbuf, (UV)c); |
635 | return is_utf8_ascii(tmpbuf); |
636 | } |
637 | |
638 | bool |
864dbfa3 |
639 | Perl_is_uni_space(pTHX_ U32 c) |
a0ed51b3 |
640 | { |
ad391ad9 |
641 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
642 | uv_to_utf8(tmpbuf, (UV)c); |
643 | return is_utf8_space(tmpbuf); |
644 | } |
645 | |
646 | bool |
864dbfa3 |
647 | Perl_is_uni_digit(pTHX_ U32 c) |
a0ed51b3 |
648 | { |
ad391ad9 |
649 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
650 | uv_to_utf8(tmpbuf, (UV)c); |
651 | return is_utf8_digit(tmpbuf); |
652 | } |
653 | |
654 | bool |
864dbfa3 |
655 | Perl_is_uni_upper(pTHX_ U32 c) |
a0ed51b3 |
656 | { |
ad391ad9 |
657 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
658 | uv_to_utf8(tmpbuf, (UV)c); |
659 | return is_utf8_upper(tmpbuf); |
660 | } |
661 | |
662 | bool |
864dbfa3 |
663 | Perl_is_uni_lower(pTHX_ U32 c) |
a0ed51b3 |
664 | { |
ad391ad9 |
665 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
666 | uv_to_utf8(tmpbuf, (UV)c); |
667 | return is_utf8_lower(tmpbuf); |
668 | } |
669 | |
670 | bool |
b8c5462f |
671 | Perl_is_uni_cntrl(pTHX_ U32 c) |
672 | { |
ad391ad9 |
673 | U8 tmpbuf[UTF8_MAXLEN+1]; |
b8c5462f |
674 | uv_to_utf8(tmpbuf, (UV)c); |
675 | return is_utf8_cntrl(tmpbuf); |
676 | } |
677 | |
678 | bool |
679 | Perl_is_uni_graph(pTHX_ U32 c) |
680 | { |
ad391ad9 |
681 | U8 tmpbuf[UTF8_MAXLEN+1]; |
b8c5462f |
682 | uv_to_utf8(tmpbuf, (UV)c); |
683 | return is_utf8_graph(tmpbuf); |
684 | } |
685 | |
686 | bool |
864dbfa3 |
687 | Perl_is_uni_print(pTHX_ U32 c) |
a0ed51b3 |
688 | { |
ad391ad9 |
689 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
690 | uv_to_utf8(tmpbuf, (UV)c); |
691 | return is_utf8_print(tmpbuf); |
692 | } |
693 | |
b8c5462f |
694 | bool |
f248d071 |
695 | Perl_is_uni_punct(pTHX_ U32 c) |
b8c5462f |
696 | { |
ad391ad9 |
697 | U8 tmpbuf[UTF8_MAXLEN+1]; |
b8c5462f |
698 | uv_to_utf8(tmpbuf, (UV)c); |
699 | return is_utf8_punct(tmpbuf); |
700 | } |
701 | |
4d61ec05 |
702 | bool |
703 | Perl_is_uni_xdigit(pTHX_ U32 c) |
704 | { |
ad391ad9 |
705 | U8 tmpbuf[UTF8_MAXLEN+1]; |
4d61ec05 |
706 | uv_to_utf8(tmpbuf, (UV)c); |
707 | return is_utf8_xdigit(tmpbuf); |
708 | } |
709 | |
a0ed51b3 |
710 | U32 |
864dbfa3 |
711 | Perl_to_uni_upper(pTHX_ U32 c) |
a0ed51b3 |
712 | { |
ad391ad9 |
713 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
714 | uv_to_utf8(tmpbuf, (UV)c); |
715 | return to_utf8_upper(tmpbuf); |
716 | } |
717 | |
718 | U32 |
864dbfa3 |
719 | Perl_to_uni_title(pTHX_ U32 c) |
a0ed51b3 |
720 | { |
ad391ad9 |
721 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
722 | uv_to_utf8(tmpbuf, (UV)c); |
723 | return to_utf8_title(tmpbuf); |
724 | } |
725 | |
726 | U32 |
864dbfa3 |
727 | Perl_to_uni_lower(pTHX_ U32 c) |
a0ed51b3 |
728 | { |
ad391ad9 |
729 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
730 | uv_to_utf8(tmpbuf, (UV)c); |
731 | return to_utf8_lower(tmpbuf); |
732 | } |
733 | |
734 | /* for now these all assume no locale info available for Unicode > 255 */ |
735 | |
736 | bool |
864dbfa3 |
737 | Perl_is_uni_alnum_lc(pTHX_ U32 c) |
a0ed51b3 |
738 | { |
739 | return is_uni_alnum(c); /* XXX no locale support yet */ |
740 | } |
741 | |
742 | bool |
b8c5462f |
743 | Perl_is_uni_alnumc_lc(pTHX_ U32 c) |
744 | { |
745 | return is_uni_alnumc(c); /* XXX no locale support yet */ |
746 | } |
747 | |
748 | bool |
864dbfa3 |
749 | Perl_is_uni_idfirst_lc(pTHX_ U32 c) |
a0ed51b3 |
750 | { |
751 | return is_uni_idfirst(c); /* XXX no locale support yet */ |
752 | } |
753 | |
754 | bool |
864dbfa3 |
755 | Perl_is_uni_alpha_lc(pTHX_ U32 c) |
a0ed51b3 |
756 | { |
757 | return is_uni_alpha(c); /* XXX no locale support yet */ |
758 | } |
759 | |
760 | bool |
4d61ec05 |
761 | Perl_is_uni_ascii_lc(pTHX_ U32 c) |
762 | { |
763 | return is_uni_ascii(c); /* XXX no locale support yet */ |
764 | } |
765 | |
766 | bool |
864dbfa3 |
767 | Perl_is_uni_space_lc(pTHX_ U32 c) |
a0ed51b3 |
768 | { |
769 | return is_uni_space(c); /* XXX no locale support yet */ |
770 | } |
771 | |
772 | bool |
864dbfa3 |
773 | Perl_is_uni_digit_lc(pTHX_ U32 c) |
a0ed51b3 |
774 | { |
775 | return is_uni_digit(c); /* XXX no locale support yet */ |
776 | } |
777 | |
778 | bool |
864dbfa3 |
779 | Perl_is_uni_upper_lc(pTHX_ U32 c) |
a0ed51b3 |
780 | { |
781 | return is_uni_upper(c); /* XXX no locale support yet */ |
782 | } |
783 | |
784 | bool |
864dbfa3 |
785 | Perl_is_uni_lower_lc(pTHX_ U32 c) |
a0ed51b3 |
786 | { |
787 | return is_uni_lower(c); /* XXX no locale support yet */ |
788 | } |
789 | |
790 | bool |
b8c5462f |
791 | Perl_is_uni_cntrl_lc(pTHX_ U32 c) |
792 | { |
793 | return is_uni_cntrl(c); /* XXX no locale support yet */ |
794 | } |
795 | |
796 | bool |
797 | Perl_is_uni_graph_lc(pTHX_ U32 c) |
798 | { |
799 | return is_uni_graph(c); /* XXX no locale support yet */ |
800 | } |
801 | |
802 | bool |
864dbfa3 |
803 | Perl_is_uni_print_lc(pTHX_ U32 c) |
a0ed51b3 |
804 | { |
805 | return is_uni_print(c); /* XXX no locale support yet */ |
806 | } |
807 | |
b8c5462f |
808 | bool |
809 | Perl_is_uni_punct_lc(pTHX_ U32 c) |
810 | { |
811 | return is_uni_punct(c); /* XXX no locale support yet */ |
812 | } |
813 | |
4d61ec05 |
814 | bool |
815 | Perl_is_uni_xdigit_lc(pTHX_ U32 c) |
816 | { |
817 | return is_uni_xdigit(c); /* XXX no locale support yet */ |
818 | } |
819 | |
a0ed51b3 |
820 | U32 |
864dbfa3 |
821 | Perl_to_uni_upper_lc(pTHX_ U32 c) |
a0ed51b3 |
822 | { |
823 | return to_uni_upper(c); /* XXX no locale support yet */ |
824 | } |
825 | |
826 | U32 |
864dbfa3 |
827 | Perl_to_uni_title_lc(pTHX_ U32 c) |
a0ed51b3 |
828 | { |
829 | return to_uni_title(c); /* XXX no locale support yet */ |
830 | } |
831 | |
832 | U32 |
864dbfa3 |
833 | Perl_to_uni_lower_lc(pTHX_ U32 c) |
a0ed51b3 |
834 | { |
835 | return to_uni_lower(c); /* XXX no locale support yet */ |
836 | } |
837 | |
a0ed51b3 |
838 | bool |
864dbfa3 |
839 | Perl_is_utf8_alnum(pTHX_ U8 *p) |
a0ed51b3 |
840 | { |
386d01d6 |
841 | if (!is_utf8_char(p)) |
842 | return FALSE; |
a0ed51b3 |
843 | if (!PL_utf8_alnum) |
289d4f09 |
844 | /* NOTE: "IsWord", not "IsAlnum", since Alnum is a true |
845 | * descendant of isalnum(3), in other words, it doesn't |
846 | * contain the '_'. --jhi */ |
847 | PL_utf8_alnum = swash_init("utf8", "IsWord", &PL_sv_undef, 0, 0); |
a0ed51b3 |
848 | return swash_fetch(PL_utf8_alnum, p); |
849 | /* return *p == '_' || is_utf8_alpha(p) || is_utf8_digit(p); */ |
850 | #ifdef SURPRISINGLY_SLOWER /* probably because alpha is usually true */ |
851 | if (!PL_utf8_alnum) |
852 | PL_utf8_alnum = swash_init("utf8", "", |
853 | sv_2mortal(newSVpv("+utf8::IsAlpha\n+utf8::IsDigit\n005F\n",0)), 0, 0); |
854 | return swash_fetch(PL_utf8_alnum, p); |
855 | #endif |
856 | } |
857 | |
858 | bool |
b8c5462f |
859 | Perl_is_utf8_alnumc(pTHX_ U8 *p) |
860 | { |
386d01d6 |
861 | if (!is_utf8_char(p)) |
862 | return FALSE; |
b8c5462f |
863 | if (!PL_utf8_alnum) |
864 | PL_utf8_alnum = swash_init("utf8", "IsAlnumC", &PL_sv_undef, 0, 0); |
865 | return swash_fetch(PL_utf8_alnum, p); |
866 | /* return is_utf8_alpha(p) || is_utf8_digit(p); */ |
867 | #ifdef SURPRISINGLY_SLOWER /* probably because alpha is usually true */ |
868 | if (!PL_utf8_alnum) |
869 | PL_utf8_alnum = swash_init("utf8", "", |
870 | sv_2mortal(newSVpv("+utf8::IsAlpha\n+utf8::IsDigit\n005F\n",0)), 0, 0); |
871 | return swash_fetch(PL_utf8_alnum, p); |
872 | #endif |
873 | } |
874 | |
875 | bool |
864dbfa3 |
876 | Perl_is_utf8_idfirst(pTHX_ U8 *p) |
a0ed51b3 |
877 | { |
878 | return *p == '_' || is_utf8_alpha(p); |
879 | } |
880 | |
881 | bool |
864dbfa3 |
882 | Perl_is_utf8_alpha(pTHX_ U8 *p) |
a0ed51b3 |
883 | { |
386d01d6 |
884 | if (!is_utf8_char(p)) |
885 | return FALSE; |
a0ed51b3 |
886 | if (!PL_utf8_alpha) |
e24b16f9 |
887 | PL_utf8_alpha = swash_init("utf8", "IsAlpha", &PL_sv_undef, 0, 0); |
a0ed51b3 |
888 | return swash_fetch(PL_utf8_alpha, p); |
889 | } |
890 | |
891 | bool |
b8c5462f |
892 | Perl_is_utf8_ascii(pTHX_ U8 *p) |
893 | { |
386d01d6 |
894 | if (!is_utf8_char(p)) |
895 | return FALSE; |
b8c5462f |
896 | if (!PL_utf8_ascii) |
897 | PL_utf8_ascii = swash_init("utf8", "IsAscii", &PL_sv_undef, 0, 0); |
898 | return swash_fetch(PL_utf8_ascii, p); |
899 | } |
900 | |
901 | bool |
864dbfa3 |
902 | Perl_is_utf8_space(pTHX_ U8 *p) |
a0ed51b3 |
903 | { |
386d01d6 |
904 | if (!is_utf8_char(p)) |
905 | return FALSE; |
a0ed51b3 |
906 | if (!PL_utf8_space) |
3bec3564 |
907 | PL_utf8_space = swash_init("utf8", "IsSpacePerl", &PL_sv_undef, 0, 0); |
a0ed51b3 |
908 | return swash_fetch(PL_utf8_space, p); |
909 | } |
910 | |
911 | bool |
864dbfa3 |
912 | Perl_is_utf8_digit(pTHX_ U8 *p) |
a0ed51b3 |
913 | { |
386d01d6 |
914 | if (!is_utf8_char(p)) |
915 | return FALSE; |
a0ed51b3 |
916 | if (!PL_utf8_digit) |
e24b16f9 |
917 | PL_utf8_digit = swash_init("utf8", "IsDigit", &PL_sv_undef, 0, 0); |
a0ed51b3 |
918 | return swash_fetch(PL_utf8_digit, p); |
919 | } |
920 | |
921 | bool |
864dbfa3 |
922 | Perl_is_utf8_upper(pTHX_ U8 *p) |
a0ed51b3 |
923 | { |
386d01d6 |
924 | if (!is_utf8_char(p)) |
925 | return FALSE; |
a0ed51b3 |
926 | if (!PL_utf8_upper) |
e24b16f9 |
927 | PL_utf8_upper = swash_init("utf8", "IsUpper", &PL_sv_undef, 0, 0); |
a0ed51b3 |
928 | return swash_fetch(PL_utf8_upper, p); |
929 | } |
930 | |
931 | bool |
864dbfa3 |
932 | Perl_is_utf8_lower(pTHX_ U8 *p) |
a0ed51b3 |
933 | { |
386d01d6 |
934 | if (!is_utf8_char(p)) |
935 | return FALSE; |
a0ed51b3 |
936 | if (!PL_utf8_lower) |
e24b16f9 |
937 | PL_utf8_lower = swash_init("utf8", "IsLower", &PL_sv_undef, 0, 0); |
a0ed51b3 |
938 | return swash_fetch(PL_utf8_lower, p); |
939 | } |
940 | |
941 | bool |
b8c5462f |
942 | Perl_is_utf8_cntrl(pTHX_ U8 *p) |
943 | { |
386d01d6 |
944 | if (!is_utf8_char(p)) |
945 | return FALSE; |
b8c5462f |
946 | if (!PL_utf8_cntrl) |
947 | PL_utf8_cntrl = swash_init("utf8", "IsCntrl", &PL_sv_undef, 0, 0); |
948 | return swash_fetch(PL_utf8_cntrl, p); |
949 | } |
950 | |
951 | bool |
952 | Perl_is_utf8_graph(pTHX_ U8 *p) |
953 | { |
386d01d6 |
954 | if (!is_utf8_char(p)) |
955 | return FALSE; |
b8c5462f |
956 | if (!PL_utf8_graph) |
957 | PL_utf8_graph = swash_init("utf8", "IsGraph", &PL_sv_undef, 0, 0); |
958 | return swash_fetch(PL_utf8_graph, p); |
959 | } |
960 | |
961 | bool |
864dbfa3 |
962 | Perl_is_utf8_print(pTHX_ U8 *p) |
a0ed51b3 |
963 | { |
386d01d6 |
964 | if (!is_utf8_char(p)) |
965 | return FALSE; |
a0ed51b3 |
966 | if (!PL_utf8_print) |
e24b16f9 |
967 | PL_utf8_print = swash_init("utf8", "IsPrint", &PL_sv_undef, 0, 0); |
a0ed51b3 |
968 | return swash_fetch(PL_utf8_print, p); |
969 | } |
970 | |
971 | bool |
b8c5462f |
972 | Perl_is_utf8_punct(pTHX_ U8 *p) |
973 | { |
386d01d6 |
974 | if (!is_utf8_char(p)) |
975 | return FALSE; |
b8c5462f |
976 | if (!PL_utf8_punct) |
977 | PL_utf8_punct = swash_init("utf8", "IsPunct", &PL_sv_undef, 0, 0); |
978 | return swash_fetch(PL_utf8_punct, p); |
979 | } |
980 | |
981 | bool |
982 | Perl_is_utf8_xdigit(pTHX_ U8 *p) |
983 | { |
386d01d6 |
984 | if (!is_utf8_char(p)) |
985 | return FALSE; |
b8c5462f |
986 | if (!PL_utf8_xdigit) |
987 | PL_utf8_xdigit = swash_init("utf8", "IsXDigit", &PL_sv_undef, 0, 0); |
988 | return swash_fetch(PL_utf8_xdigit, p); |
989 | } |
990 | |
991 | bool |
864dbfa3 |
992 | Perl_is_utf8_mark(pTHX_ U8 *p) |
a0ed51b3 |
993 | { |
386d01d6 |
994 | if (!is_utf8_char(p)) |
995 | return FALSE; |
a0ed51b3 |
996 | if (!PL_utf8_mark) |
e24b16f9 |
997 | PL_utf8_mark = swash_init("utf8", "IsM", &PL_sv_undef, 0, 0); |
a0ed51b3 |
998 | return swash_fetch(PL_utf8_mark, p); |
999 | } |
1000 | |
2104c8d9 |
1001 | UV |
864dbfa3 |
1002 | Perl_to_utf8_upper(pTHX_ U8 *p) |
a0ed51b3 |
1003 | { |
1004 | UV uv; |
1005 | |
1006 | if (!PL_utf8_toupper) |
e24b16f9 |
1007 | PL_utf8_toupper = swash_init("utf8", "ToUpper", &PL_sv_undef, 4, 0); |
a0ed51b3 |
1008 | uv = swash_fetch(PL_utf8_toupper, p); |
756820e3 |
1009 | return uv ? uv : utf8_to_uv(p,UTF8_MAXLEN,0,0); |
a0ed51b3 |
1010 | } |
1011 | |
2104c8d9 |
1012 | UV |
864dbfa3 |
1013 | Perl_to_utf8_title(pTHX_ U8 *p) |
a0ed51b3 |
1014 | { |
1015 | UV uv; |
1016 | |
1017 | if (!PL_utf8_totitle) |
e24b16f9 |
1018 | PL_utf8_totitle = swash_init("utf8", "ToTitle", &PL_sv_undef, 4, 0); |
a0ed51b3 |
1019 | uv = swash_fetch(PL_utf8_totitle, p); |
756820e3 |
1020 | return uv ? uv : utf8_to_uv(p,UTF8_MAXLEN,0,0); |
a0ed51b3 |
1021 | } |
1022 | |
2104c8d9 |
1023 | UV |
864dbfa3 |
1024 | Perl_to_utf8_lower(pTHX_ U8 *p) |
a0ed51b3 |
1025 | { |
1026 | UV uv; |
1027 | |
1028 | if (!PL_utf8_tolower) |
e24b16f9 |
1029 | PL_utf8_tolower = swash_init("utf8", "ToLower", &PL_sv_undef, 4, 0); |
a0ed51b3 |
1030 | uv = swash_fetch(PL_utf8_tolower, p); |
756820e3 |
1031 | return uv ? uv : utf8_to_uv(p,UTF8_MAXLEN,0,0); |
a0ed51b3 |
1032 | } |
1033 | |
1034 | /* a "swash" is a swatch hash */ |
1035 | |
1036 | SV* |
864dbfa3 |
1037 | Perl_swash_init(pTHX_ char* pkg, char* name, SV *listsv, I32 minbits, I32 none) |
a0ed51b3 |
1038 | { |
1039 | SV* retval; |
1040 | char tmpbuf[256]; |
8e84507e |
1041 | dSP; |
ce3b816e |
1042 | |
1043 | if (!gv_stashpv(pkg, 0)) { /* demand load utf8 */ |
1044 | ENTER; |
1045 | Perl_load_module(aTHX_ PERL_LOADMOD_NOIMPORT, newSVpv(pkg,0), Nullsv); |
1046 | LEAVE; |
1047 | } |
1048 | SPAGAIN; |
a0ed51b3 |
1049 | PUSHSTACKi(PERLSI_MAGIC); |
1050 | PUSHMARK(SP); |
1051 | EXTEND(SP,5); |
1052 | PUSHs(sv_2mortal(newSVpvn(pkg, strlen(pkg)))); |
1053 | PUSHs(sv_2mortal(newSVpvn(name, strlen(name)))); |
1054 | PUSHs(listsv); |
1055 | PUSHs(sv_2mortal(newSViv(minbits))); |
1056 | PUSHs(sv_2mortal(newSViv(none))); |
1057 | PUTBACK; |
1058 | ENTER; |
1059 | SAVEI32(PL_hints); |
1060 | PL_hints = 0; |
1061 | save_re_context(); |
e24b16f9 |
1062 | if (PL_curcop == &PL_compiling) /* XXX ought to be handled by lex_start */ |
a0ed51b3 |
1063 | strncpy(tmpbuf, PL_tokenbuf, sizeof tmpbuf); |
864dbfa3 |
1064 | if (call_method("SWASHNEW", G_SCALAR)) |
8e84507e |
1065 | retval = newSVsv(*PL_stack_sp--); |
a0ed51b3 |
1066 | else |
e24b16f9 |
1067 | retval = &PL_sv_undef; |
a0ed51b3 |
1068 | LEAVE; |
1069 | POPSTACK; |
e24b16f9 |
1070 | if (PL_curcop == &PL_compiling) { |
a0ed51b3 |
1071 | strncpy(PL_tokenbuf, tmpbuf, sizeof tmpbuf); |
e24b16f9 |
1072 | PL_curcop->op_private = PL_hints; |
a0ed51b3 |
1073 | } |
1074 | if (!SvROK(retval) || SvTYPE(SvRV(retval)) != SVt_PVHV) |
cea2e8a9 |
1075 | Perl_croak(aTHX_ "SWASHNEW didn't return an HV ref"); |
a0ed51b3 |
1076 | return retval; |
1077 | } |
1078 | |
1079 | UV |
864dbfa3 |
1080 | Perl_swash_fetch(pTHX_ SV *sv, U8 *ptr) |
a0ed51b3 |
1081 | { |
1082 | HV* hv = (HV*)SvRV(sv); |
1083 | U32 klen = UTF8SKIP(ptr) - 1; |
1084 | U32 off = ptr[klen] & 127; /* NB: 64 bit always 0 when len > 1 */ |
1085 | STRLEN slen; |
1086 | STRLEN needents = (klen ? 64 : 128); |
dfe13c55 |
1087 | U8 *tmps; |
a0ed51b3 |
1088 | U32 bit; |
1089 | SV *retval; |
1090 | |
1091 | /* |
1092 | * This single-entry cache saves about 1/3 of the utf8 overhead in test |
1093 | * suite. (That is, only 7-8% overall over just a hash cache. Still, |
1094 | * it's nothing to sniff at.) Pity we usually come through at least |
1095 | * two function calls to get here... |
1096 | * |
1097 | * NB: this code assumes that swatches are never modified, once generated! |
1098 | */ |
1099 | |
1100 | if (hv == PL_last_swash_hv && |
1101 | klen == PL_last_swash_klen && |
12ae5dfc |
1102 | (!klen || memEQ((char *)ptr,(char *)PL_last_swash_key,klen)) ) |
a0ed51b3 |
1103 | { |
1104 | tmps = PL_last_swash_tmps; |
1105 | slen = PL_last_swash_slen; |
1106 | } |
1107 | else { |
1108 | /* Try our second-level swatch cache, kept in a hash. */ |
dfe13c55 |
1109 | SV** svp = hv_fetch(hv, (char*)ptr, klen, FALSE); |
a0ed51b3 |
1110 | |
1111 | /* If not cached, generate it via utf8::SWASHGET */ |
dfe13c55 |
1112 | if (!svp || !SvPOK(*svp) || !(tmps = (U8*)SvPV(*svp, slen))) { |
a0ed51b3 |
1113 | dSP; |
1114 | ENTER; |
1115 | SAVETMPS; |
1116 | save_re_context(); |
1117 | PUSHSTACKi(PERLSI_MAGIC); |
1118 | PUSHMARK(SP); |
1119 | EXTEND(SP,3); |
1120 | PUSHs((SV*)sv); |
756820e3 |
1121 | PUSHs(sv_2mortal(newSViv(utf8_to_uv(ptr, UTF8_MAXLEN, 0, 0) & ~(needents - 1)))); |
a0ed51b3 |
1122 | PUSHs(sv_2mortal(newSViv(needents))); |
1123 | PUTBACK; |
864dbfa3 |
1124 | if (call_method("SWASHGET", G_SCALAR)) |
8e84507e |
1125 | retval = newSVsv(*PL_stack_sp--); |
a0ed51b3 |
1126 | else |
e24b16f9 |
1127 | retval = &PL_sv_undef; |
a0ed51b3 |
1128 | POPSTACK; |
1129 | FREETMPS; |
1130 | LEAVE; |
e24b16f9 |
1131 | if (PL_curcop == &PL_compiling) |
1132 | PL_curcop->op_private = PL_hints; |
a0ed51b3 |
1133 | |
dfe13c55 |
1134 | svp = hv_store(hv, (char*)ptr, klen, retval, 0); |
a0ed51b3 |
1135 | |
dfe13c55 |
1136 | if (!svp || !(tmps = (U8*)SvPV(*svp, slen)) || slen < 8) |
cea2e8a9 |
1137 | Perl_croak(aTHX_ "SWASHGET didn't return result of proper length"); |
a0ed51b3 |
1138 | } |
1139 | |
1140 | PL_last_swash_hv = hv; |
1141 | PL_last_swash_klen = klen; |
1142 | PL_last_swash_tmps = tmps; |
1143 | PL_last_swash_slen = slen; |
1144 | if (klen) |
1145 | Copy(ptr, PL_last_swash_key, klen, U8); |
1146 | } |
1147 | |
9faf8d75 |
1148 | switch ((int)((slen << 3) / needents)) { |
a0ed51b3 |
1149 | case 1: |
1150 | bit = 1 << (off & 7); |
1151 | off >>= 3; |
1152 | return (tmps[off] & bit) != 0; |
1153 | case 8: |
1154 | return tmps[off]; |
1155 | case 16: |
1156 | off <<= 1; |
1157 | return (tmps[off] << 8) + tmps[off + 1] ; |
1158 | case 32: |
1159 | off <<= 2; |
1160 | return (tmps[off] << 24) + (tmps[off+1] << 16) + (tmps[off+2] << 8) + tmps[off + 3] ; |
1161 | } |
cea2e8a9 |
1162 | Perl_croak(aTHX_ "panic: swash_fetch"); |
a0ed51b3 |
1163 | return 0; |
1164 | } |