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 | |
eebe1485 |
29 | /* |
30 | =for apidoc A|U8*|uv_to_utf8|U8 *d|UV uv |
31 | |
32 | Adds the UTF8 representation of the Unicode codepoint C<uv> to the end |
33 | of the string C<d>; C<d> should be have at least C<UTF8_MAXLEN+1> free |
34 | bytes available. The return value is the pointer to the byte after the |
35 | end of the new character. In other words, |
36 | |
37 | d = uv_to_utf8(d, uv); |
38 | |
39 | is the recommended Unicode-aware way of saying |
40 | |
41 | *(d++) = uv; |
42 | |
43 | =cut |
44 | */ |
45 | |
dfe13c55 |
46 | U8 * |
eebe1485 |
47 | Perl_uv_to_utf8(pTHX_ U8 *d, UV uv) |
a0ed51b3 |
48 | { |
3bd709b1 |
49 | if (uv < 0x100) |
50 | uv = NATIVE_TO_ASCII(uv); |
a0ed51b3 |
51 | if (uv < 0x80) { |
52 | *d++ = uv; |
53 | return d; |
54 | } |
55 | if (uv < 0x800) { |
56 | *d++ = (( uv >> 6) | 0xc0); |
57 | *d++ = (( uv & 0x3f) | 0x80); |
58 | return d; |
59 | } |
60 | if (uv < 0x10000) { |
61 | *d++ = (( uv >> 12) | 0xe0); |
62 | *d++ = (((uv >> 6) & 0x3f) | 0x80); |
63 | *d++ = (( uv & 0x3f) | 0x80); |
64 | return d; |
65 | } |
66 | if (uv < 0x200000) { |
67 | *d++ = (( uv >> 18) | 0xf0); |
68 | *d++ = (((uv >> 12) & 0x3f) | 0x80); |
69 | *d++ = (((uv >> 6) & 0x3f) | 0x80); |
70 | *d++ = (( uv & 0x3f) | 0x80); |
71 | return d; |
72 | } |
73 | if (uv < 0x4000000) { |
74 | *d++ = (( uv >> 24) | 0xf8); |
75 | *d++ = (((uv >> 18) & 0x3f) | 0x80); |
76 | *d++ = (((uv >> 12) & 0x3f) | 0x80); |
77 | *d++ = (((uv >> 6) & 0x3f) | 0x80); |
78 | *d++ = (( uv & 0x3f) | 0x80); |
79 | return d; |
80 | } |
81 | if (uv < 0x80000000) { |
82 | *d++ = (( uv >> 30) | 0xfc); |
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); |
88 | return d; |
89 | } |
6b8eaf93 |
90 | #ifdef HAS_QUAD |
d7578b48 |
91 | if (uv < UTF8_QUAD_MAX) |
a0ed51b3 |
92 | #endif |
93 | { |
94 | *d++ = 0xfe; /* Can't match U+FEFF! */ |
95 | *d++ = (((uv >> 30) & 0x3f) | 0x80); |
96 | *d++ = (((uv >> 24) & 0x3f) | 0x80); |
97 | *d++ = (((uv >> 18) & 0x3f) | 0x80); |
98 | *d++ = (((uv >> 12) & 0x3f) | 0x80); |
99 | *d++ = (((uv >> 6) & 0x3f) | 0x80); |
100 | *d++ = (( uv & 0x3f) | 0x80); |
101 | return d; |
102 | } |
6b8eaf93 |
103 | #ifdef HAS_QUAD |
a0ed51b3 |
104 | { |
105 | *d++ = 0xff; /* Can't match U+FFFE! */ |
3c77ea2b |
106 | *d++ = 0x80; /* 6 Reserved bits */ |
107 | *d++ = (((uv >> 60) & 0x0f) | 0x80); /* 2 Reserved bits */ |
108 | *d++ = (((uv >> 54) & 0x3f) | 0x80); |
109 | *d++ = (((uv >> 48) & 0x3f) | 0x80); |
110 | *d++ = (((uv >> 42) & 0x3f) | 0x80); |
a0ed51b3 |
111 | *d++ = (((uv >> 36) & 0x3f) | 0x80); |
112 | *d++ = (((uv >> 30) & 0x3f) | 0x80); |
113 | *d++ = (((uv >> 24) & 0x3f) | 0x80); |
114 | *d++ = (((uv >> 18) & 0x3f) | 0x80); |
115 | *d++ = (((uv >> 12) & 0x3f) | 0x80); |
116 | *d++ = (((uv >> 6) & 0x3f) | 0x80); |
117 | *d++ = (( uv & 0x3f) | 0x80); |
118 | return d; |
119 | } |
120 | #endif |
121 | } |
122 | |
eebe1485 |
123 | /* |
124 | =for apidoc A|STRLEN|is_utf8_char|U8 *s |
125 | |
5da9da9e |
126 | Tests if some arbitrary number of bytes begins in a valid UTF-8 |
127 | character. Note that an ASCII character is a valid UTF-8 character. |
128 | The actual number of bytes in the UTF-8 character will be returned if |
129 | it is valid, otherwise 0. |
eebe1485 |
130 | |
5da9da9e |
131 | =cut */ |
067a85ef |
132 | STRLEN |
386d01d6 |
133 | Perl_is_utf8_char(pTHX_ U8 *s) |
134 | { |
135 | U8 u = *s; |
067a85ef |
136 | STRLEN slen, len; |
137 | UV uv, ouv; |
386d01d6 |
138 | |
60006e79 |
139 | if (UTF8_IS_ASCII(u)) |
386d01d6 |
140 | return 1; |
141 | |
60006e79 |
142 | if (!UTF8_IS_START(u)) |
386d01d6 |
143 | return 0; |
144 | |
9f07fdcd |
145 | len = UTF8SKIP(s); |
386d01d6 |
146 | |
60006e79 |
147 | if (len < 2 || !UTF8_IS_CONTINUATION(s[1])) |
067a85ef |
148 | return 0; |
149 | |
386d01d6 |
150 | slen = len - 1; |
151 | s++; |
067a85ef |
152 | uv = u; |
153 | ouv = uv; |
386d01d6 |
154 | while (slen--) { |
60006e79 |
155 | if (!UTF8_IS_CONTINUATION(*s)) |
386d01d6 |
156 | return 0; |
8850bf83 |
157 | uv = UTF8_ACCUMULATE(uv, *s); |
067a85ef |
158 | if (uv < ouv) |
159 | return 0; |
160 | ouv = uv; |
386d01d6 |
161 | s++; |
162 | } |
067a85ef |
163 | |
5bbb0b5a |
164 | if (UNISKIP(uv) < len) |
067a85ef |
165 | return 0; |
166 | |
386d01d6 |
167 | return len; |
168 | } |
169 | |
6662521e |
170 | /* |
eebe1485 |
171 | =for apidoc A|bool|is_utf8_string|U8 *s|STRLEN len |
6662521e |
172 | |
5da9da9e |
173 | Returns true if first C<len> bytes of the given string form a valid UTF8 |
174 | string, false otherwise. Note that 'a valid UTF8 string' does not mean |
175 | 'a string that contains UTF8' because a valid ASCII string is a valid |
176 | UTF8 string. |
6662521e |
177 | |
178 | =cut |
179 | */ |
180 | |
8e84507e |
181 | bool |
6662521e |
182 | Perl_is_utf8_string(pTHX_ U8 *s, STRLEN len) |
183 | { |
067a85ef |
184 | U8* x = s; |
1aa99e6b |
185 | U8* send; |
067a85ef |
186 | STRLEN c; |
187 | |
1aa99e6b |
188 | if (!len) |
6cd5fe39 |
189 | len = strlen((char *)s); |
1aa99e6b |
190 | send = s + len; |
191 | |
6662521e |
192 | while (x < send) { |
193 | c = is_utf8_char(x); |
067a85ef |
194 | if (!c) |
195 | return FALSE; |
6662521e |
196 | x += c; |
6662521e |
197 | } |
60006e79 |
198 | if (x != send) |
199 | return FALSE; |
067a85ef |
200 | |
201 | return TRUE; |
6662521e |
202 | } |
203 | |
67e989fb |
204 | /* |
65c50fde |
205 | =for apidoc A|UV|utf8_to_uv|U8 *s|STRLEN curlen|STRLEN *retlen|U32 flags |
67e989fb |
206 | |
207 | Returns the character value of the first character in the string C<s> |
ba210ebe |
208 | which is assumed to be in UTF8 encoding and no longer than C<curlen>; |
7df053ec |
209 | C<retlen> will be set to the length, in bytes, of that character. |
67e989fb |
210 | |
211 | If C<s> does not point to a well-formed UTF8 character, the behaviour |
dcad2880 |
212 | is dependent on the value of C<flags>: if it contains UTF8_CHECK_ONLY, |
213 | it is assumed that the caller will raise a warning, and this function |
28d3d195 |
214 | will silently just set C<retlen> to C<-1> and return zero. If the |
215 | C<flags> does not contain UTF8_CHECK_ONLY, warnings about |
216 | malformations will be given, C<retlen> will be set to the expected |
217 | length of the UTF-8 character in bytes, and zero will be returned. |
218 | |
219 | The C<flags> can also contain various flags to allow deviations from |
220 | the strict UTF-8 encoding (see F<utf8.h>). |
67e989fb |
221 | |
dcad2880 |
222 | =cut */ |
67e989fb |
223 | |
a0ed51b3 |
224 | UV |
dcad2880 |
225 | Perl_utf8_to_uv(pTHX_ U8* s, STRLEN curlen, STRLEN* retlen, U32 flags) |
a0ed51b3 |
226 | { |
ba210ebe |
227 | UV uv = *s, ouv; |
228 | STRLEN len = 1; |
7bf1b6bb |
229 | #ifdef EBCDIC |
230 | bool dowarn = 0; |
231 | #else |
ba210ebe |
232 | bool dowarn = ckWARN_d(WARN_UTF8); |
7bf1b6bb |
233 | #endif |
ba210ebe |
234 | STRLEN expectlen = 0; |
a0dbb045 |
235 | U32 warning = 0; |
236 | |
237 | /* This list is a superset of the UTF8_ALLOW_XXX. */ |
238 | |
239 | #define UTF8_WARN_EMPTY 1 |
240 | #define UTF8_WARN_CONTINUATION 2 |
241 | #define UTF8_WARN_NON_CONTINUATION 3 |
242 | #define UTF8_WARN_FE_FF 4 |
243 | #define UTF8_WARN_SHORT 5 |
244 | #define UTF8_WARN_OVERFLOW 6 |
245 | #define UTF8_WARN_SURROGATE 7 |
246 | #define UTF8_WARN_BOM 8 |
247 | #define UTF8_WARN_LONG 9 |
248 | #define UTF8_WARN_FFFF 10 |
249 | |
250 | if (curlen == 0 && |
251 | !(flags & UTF8_ALLOW_EMPTY)) { |
252 | warning = UTF8_WARN_EMPTY; |
0c443dc2 |
253 | goto malformed; |
254 | } |
255 | |
421a8bf2 |
256 | if (UTF8_IS_ASCII(uv)) { |
a0ed51b3 |
257 | if (retlen) |
258 | *retlen = 1; |
3bd709b1 |
259 | return ASCII_TO_NATIVE(*s); |
a0ed51b3 |
260 | } |
67e989fb |
261 | |
421a8bf2 |
262 | if (UTF8_IS_CONTINUATION(uv) && |
fcc8fcf6 |
263 | !(flags & UTF8_ALLOW_CONTINUATION)) { |
a0dbb045 |
264 | warning = UTF8_WARN_CONTINUATION; |
ba210ebe |
265 | goto malformed; |
266 | } |
267 | |
421a8bf2 |
268 | if (UTF8_IS_START(uv) && curlen > 1 && !UTF8_IS_CONTINUATION(s[1]) && |
fcc8fcf6 |
269 | !(flags & UTF8_ALLOW_NON_CONTINUATION)) { |
a0dbb045 |
270 | warning = UTF8_WARN_NON_CONTINUATION; |
ba210ebe |
271 | goto malformed; |
272 | } |
fcc8fcf6 |
273 | |
274 | if ((uv == 0xfe || uv == 0xff) && |
275 | !(flags & UTF8_ALLOW_FE_FF)) { |
a0dbb045 |
276 | warning = UTF8_WARN_FE_FF; |
ba210ebe |
277 | goto malformed; |
a0ed51b3 |
278 | } |
fcc8fcf6 |
279 | |
ba210ebe |
280 | if (!(uv & 0x20)) { len = 2; uv &= 0x1f; } |
281 | else if (!(uv & 0x10)) { len = 3; uv &= 0x0f; } |
282 | else if (!(uv & 0x08)) { len = 4; uv &= 0x07; } |
283 | else if (!(uv & 0x04)) { len = 5; uv &= 0x03; } |
284 | else if (!(uv & 0x02)) { len = 6; uv &= 0x01; } |
285 | else if (!(uv & 0x01)) { len = 7; uv = 0; } |
3c77ea2b |
286 | else { len = 13; uv = 0; } /* whoa! */ |
fcc8fcf6 |
287 | |
a0ed51b3 |
288 | if (retlen) |
289 | *retlen = len; |
ba210ebe |
290 | |
291 | expectlen = len; |
292 | |
fcc8fcf6 |
293 | if ((curlen < expectlen) && |
294 | !(flags & UTF8_ALLOW_SHORT)) { |
a0dbb045 |
295 | warning = UTF8_WARN_SHORT; |
ba210ebe |
296 | goto malformed; |
297 | } |
298 | |
299 | len--; |
a0ed51b3 |
300 | s++; |
ba210ebe |
301 | ouv = uv; |
302 | |
a0ed51b3 |
303 | while (len--) { |
421a8bf2 |
304 | if (!UTF8_IS_CONTINUATION(*s) && |
305 | !(flags & UTF8_ALLOW_NON_CONTINUATION)) { |
a0dbb045 |
306 | s--; |
307 | warning = UTF8_WARN_NON_CONTINUATION; |
ba210ebe |
308 | goto malformed; |
a0ed51b3 |
309 | } |
310 | else |
8850bf83 |
311 | uv = UTF8_ACCUMULATE(uv, *s); |
a0dbb045 |
312 | if (!(uv > ouv)) { |
313 | /* These cannot be allowed. */ |
314 | if (uv == ouv) { |
315 | if (!(flags & UTF8_ALLOW_LONG)) { |
316 | warning = UTF8_WARN_LONG; |
317 | goto malformed; |
318 | } |
319 | } |
320 | else { /* uv < ouv */ |
321 | /* This cannot be allowed. */ |
322 | warning = UTF8_WARN_OVERFLOW; |
323 | goto malformed; |
324 | } |
ba210ebe |
325 | } |
326 | s++; |
327 | ouv = uv; |
328 | } |
329 | |
421a8bf2 |
330 | if (UNICODE_IS_SURROGATE(uv) && |
fcc8fcf6 |
331 | !(flags & UTF8_ALLOW_SURROGATE)) { |
a0dbb045 |
332 | warning = UTF8_WARN_SURROGATE; |
ba210ebe |
333 | goto malformed; |
421a8bf2 |
334 | } else if (UNICODE_IS_BYTE_ORDER_MARK(uv) && |
fcc8fcf6 |
335 | !(flags & UTF8_ALLOW_BOM)) { |
a0dbb045 |
336 | warning = UTF8_WARN_BOM; |
ba210ebe |
337 | goto malformed; |
fcc8fcf6 |
338 | } else if ((expectlen > UNISKIP(uv)) && |
339 | !(flags & UTF8_ALLOW_LONG)) { |
a0dbb045 |
340 | warning = UTF8_WARN_LONG; |
ba210ebe |
341 | goto malformed; |
421a8bf2 |
342 | } else if (UNICODE_IS_ILLEGAL(uv) && |
a9917092 |
343 | !(flags & UTF8_ALLOW_FFFF)) { |
a0dbb045 |
344 | warning = UTF8_WARN_FFFF; |
a9917092 |
345 | goto malformed; |
a0ed51b3 |
346 | } |
ba210ebe |
347 | |
a0ed51b3 |
348 | return uv; |
ba210ebe |
349 | |
350 | malformed: |
351 | |
fcc8fcf6 |
352 | if (flags & UTF8_CHECK_ONLY) { |
ba210ebe |
353 | if (retlen) |
cc366d4b |
354 | *retlen = -1; |
ba210ebe |
355 | return 0; |
356 | } |
357 | |
a0dbb045 |
358 | if (dowarn) { |
359 | SV* sv = sv_2mortal(newSVpv("Malformed UTF-8 character ", 0)); |
360 | |
361 | switch (warning) { |
362 | case 0: /* Intentionally empty. */ break; |
363 | case UTF8_WARN_EMPTY: |
364 | Perl_sv_catpvf(aTHX_ sv, "(empty string)"); |
365 | break; |
366 | case UTF8_WARN_CONTINUATION: |
367 | Perl_sv_catpvf(aTHX_ sv, "(unexpected continuation byte 0x%02"UVxf")", uv); |
368 | break; |
369 | case UTF8_WARN_NON_CONTINUATION: |
370 | Perl_sv_catpvf(aTHX_ sv, "(unexpected non-continuation byte 0x%02"UVxf" after start byte 0x%02"UVxf")", |
371 | (UV)s[1], uv); |
372 | break; |
373 | case UTF8_WARN_FE_FF: |
374 | Perl_sv_catpvf(aTHX_ sv, "(byte 0x%02"UVxf")", uv); |
375 | break; |
376 | case UTF8_WARN_SHORT: |
377 | Perl_sv_catpvf(aTHX_ sv, "(%d byte%s, need %d)", |
378 | curlen, curlen == 1 ? "" : "s", expectlen); |
379 | break; |
380 | case UTF8_WARN_OVERFLOW: |
381 | Perl_sv_catpvf(aTHX_ sv, "(overflow at 0x%"UVxf", byte 0x%02x)", |
382 | ouv, *s); |
383 | break; |
384 | case UTF8_WARN_SURROGATE: |
385 | Perl_sv_catpvf(aTHX_ sv, "(UTF-16 surrogate 0x%04"UVxf")", uv); |
386 | break; |
387 | case UTF8_WARN_BOM: |
388 | Perl_sv_catpvf(aTHX_ sv, "(byte order mark 0x%04"UVxf")", uv); |
389 | break; |
390 | case UTF8_WARN_LONG: |
391 | Perl_sv_catpvf(aTHX_ sv, "(%d byte%s, need %d)", |
392 | expectlen, expectlen == 1 ? "": "s", UNISKIP(uv)); |
393 | break; |
394 | case UTF8_WARN_FFFF: |
395 | Perl_sv_catpvf(aTHX_ sv, "(character 0x%04"UVxf")", uv); |
396 | break; |
397 | default: |
398 | Perl_sv_catpvf(aTHX_ sv, "(unknown reason)"); |
399 | break; |
400 | } |
401 | |
402 | if (warning) { |
403 | char *s = SvPVX(sv); |
404 | |
405 | if (PL_op) |
406 | Perl_warner(aTHX_ WARN_UTF8, |
407 | "%s in %s", s, PL_op_desc[PL_op->op_type]); |
408 | else |
409 | Perl_warner(aTHX_ WARN_UTF8, "%s", s); |
410 | } |
411 | } |
412 | |
ba210ebe |
413 | if (retlen) |
28d3d195 |
414 | *retlen = expectlen ? expectlen : len; |
ba210ebe |
415 | |
28d3d195 |
416 | return 0; |
a0ed51b3 |
417 | } |
418 | |
8e84507e |
419 | /* |
eebe1485 |
420 | =for apidoc A|U8* s|utf8_to_uv_simple|STRLEN *retlen |
8e84507e |
421 | |
422 | Returns the character value of the first character in the string C<s> |
423 | which is assumed to be in UTF8 encoding; C<retlen> will be set to the |
7df053ec |
424 | length, in bytes, of that character. |
8e84507e |
425 | |
ba210ebe |
426 | If C<s> does not point to a well-formed UTF8 character, zero is |
427 | returned and retlen is set, if possible, to -1. |
8e84507e |
428 | |
429 | =cut |
430 | */ |
431 | |
432 | UV |
dcad2880 |
433 | Perl_utf8_to_uv_simple(pTHX_ U8* s, STRLEN* retlen) |
8e84507e |
434 | { |
2e4dc9fc |
435 | return Perl_utf8_to_uv(aTHX_ s, UTF8_MAXLEN, retlen, 0); |
8e84507e |
436 | } |
437 | |
b76347f2 |
438 | /* |
eebe1485 |
439 | =for apidoc A|STRLEN|utf8_length|U8* s|U8 *e |
b76347f2 |
440 | |
441 | Return the length of the UTF-8 char encoded string C<s> in characters. |
02eb7b47 |
442 | Stops at C<e> (inclusive). If C<e E<lt> s> or if the scan would end |
443 | up past C<e>, croaks. |
b76347f2 |
444 | |
445 | =cut |
446 | */ |
447 | |
448 | STRLEN |
449 | Perl_utf8_length(pTHX_ U8* s, U8* e) |
450 | { |
451 | STRLEN len = 0; |
452 | |
8850bf83 |
453 | /* Note: cannot use UTF8_IS_...() too eagerly here since e.g. |
454 | * the bitops (especially ~) can create illegal UTF-8. |
455 | * In other words: in Perl UTF-8 is not just for Unicode. */ |
456 | |
b76347f2 |
457 | if (e < s) |
02eb7b47 |
458 | Perl_croak(aTHX_ "panic: utf8_length: unexpected end"); |
b76347f2 |
459 | while (s < e) { |
02eb7b47 |
460 | U8 t = UTF8SKIP(s); |
b76347f2 |
461 | |
462 | if (e - s < t) |
02eb7b47 |
463 | Perl_croak(aTHX_ "panic: utf8_length: unaligned end"); |
b76347f2 |
464 | s += t; |
465 | len++; |
466 | } |
467 | |
468 | return len; |
469 | } |
470 | |
b06226ff |
471 | /* |
eebe1485 |
472 | =for apidoc A|IV|utf8_distance|U8 *a|U8 *b |
b06226ff |
473 | |
474 | Returns the number of UTF8 characters between the UTF-8 pointers C<a> |
475 | and C<b>. |
476 | |
477 | WARNING: use only if you *know* that the pointers point inside the |
478 | same UTF-8 buffer. |
479 | |
480 | =cut */ |
a0ed51b3 |
481 | |
02eb7b47 |
482 | IV |
864dbfa3 |
483 | Perl_utf8_distance(pTHX_ U8 *a, U8 *b) |
a0ed51b3 |
484 | { |
02eb7b47 |
485 | IV off = 0; |
486 | |
8850bf83 |
487 | /* Note: cannot use UTF8_IS_...() too eagerly here since e.g. |
488 | * the bitops (especially ~) can create illegal UTF-8. |
489 | * In other words: in Perl UTF-8 is not just for Unicode. */ |
490 | |
a0ed51b3 |
491 | if (a < b) { |
492 | while (a < b) { |
02eb7b47 |
493 | U8 c = UTF8SKIP(a); |
494 | |
495 | if (b - a < c) |
496 | Perl_croak(aTHX_ "panic: utf8_distance: unaligned end"); |
497 | a += c; |
a0ed51b3 |
498 | off--; |
499 | } |
500 | } |
501 | else { |
502 | while (b < a) { |
02eb7b47 |
503 | U8 c = UTF8SKIP(b); |
504 | |
505 | if (a - b < c) |
506 | Perl_croak(aTHX_ "panic: utf8_distance: unaligned end"); |
507 | b += c; |
a0ed51b3 |
508 | off++; |
509 | } |
510 | } |
02eb7b47 |
511 | |
a0ed51b3 |
512 | return off; |
513 | } |
514 | |
b06226ff |
515 | /* |
eebe1485 |
516 | =for apidoc A|U8*|utf8_hop|U8 *s|I32 off |
b06226ff |
517 | |
8850bf83 |
518 | Return the UTF-8 pointer C<s> displaced by C<off> characters, either |
519 | forward or backward. |
b06226ff |
520 | |
521 | WARNING: do not use the following unless you *know* C<off> is within |
8850bf83 |
522 | the UTF-8 data pointed to by C<s> *and* that on entry C<s> is aligned |
523 | on the first byte of character or just after the last byte of a character. |
b06226ff |
524 | |
525 | =cut */ |
a0ed51b3 |
526 | |
527 | U8 * |
864dbfa3 |
528 | Perl_utf8_hop(pTHX_ U8 *s, I32 off) |
a0ed51b3 |
529 | { |
8850bf83 |
530 | /* Note: cannot use UTF8_IS_...() too eagerly here since e.g |
531 | * the bitops (especially ~) can create illegal UTF-8. |
532 | * In other words: in Perl UTF-8 is not just for Unicode. */ |
533 | |
a0ed51b3 |
534 | if (off >= 0) { |
535 | while (off--) |
536 | s += UTF8SKIP(s); |
537 | } |
538 | else { |
539 | while (off++) { |
540 | s--; |
8850bf83 |
541 | while (UTF8_IS_CONTINUATION(*s)) |
542 | s--; |
a0ed51b3 |
543 | } |
544 | } |
545 | return s; |
546 | } |
547 | |
6940069f |
548 | /* |
eebe1485 |
549 | =for apidoc A|U8 *|utf8_to_bytes|U8 *s|STRLEN *len |
6940069f |
550 | |
246fae53 |
551 | Converts a string C<s> of length C<len> from UTF8 into byte encoding. |
552 | Unlike C<bytes_to_utf8>, this over-writes the original string, and |
553 | updates len to contain the new length. |
67e989fb |
554 | Returns zero on failure, setting C<len> to -1. |
6940069f |
555 | |
556 | =cut |
557 | */ |
558 | |
559 | U8 * |
246fae53 |
560 | Perl_utf8_to_bytes(pTHX_ U8* s, STRLEN *len) |
6940069f |
561 | { |
6940069f |
562 | U8 *send; |
563 | U8 *d; |
dcad2880 |
564 | U8 *save = s; |
246fae53 |
565 | |
566 | /* ensure valid UTF8 and chars < 256 before updating string */ |
dcad2880 |
567 | for (send = s + *len; s < send; ) { |
568 | U8 c = *s++; |
569 | |
9f9ab905 |
570 | if (c >= 0x80 && |
dcad2880 |
571 | ((s >= send) || |
572 | ((*s++ & 0xc0) != 0x80) || ((c & 0xfe) != 0xc2))) { |
573 | *len = -1; |
574 | return 0; |
575 | } |
246fae53 |
576 | } |
dcad2880 |
577 | |
578 | d = s = save; |
6940069f |
579 | while (s < send) { |
ed646e6e |
580 | STRLEN ulen; |
581 | *d++ = (U8)utf8_to_uv_simple(s, &ulen); |
582 | s += ulen; |
6940069f |
583 | } |
584 | *d = '\0'; |
246fae53 |
585 | *len = d - save; |
6940069f |
586 | return save; |
587 | } |
588 | |
589 | /* |
f9a63242 |
590 | =for apidoc A|U8 *|bytes_from_utf8|U8 *s|STRLEN *len|bool *is_utf8 |
591 | |
592 | Converts a string C<s> of length C<len> from UTF8 into byte encoding. |
593 | Unlike <utf8_to_bytes> but like C<bytes_to_utf8>, returns a pointer to |
ef9edfd0 |
594 | the newly-created string, and updates C<len> to contain the new |
595 | length. Returns the original string if no conversion occurs, C<len> |
596 | is unchanged. Do nothing if C<is_utf8> points to 0. Sets C<is_utf8> to |
597 | 0 if C<s> is converted or contains all 7bit characters. |
f9a63242 |
598 | |
599 | =cut */ |
600 | |
601 | U8 * |
602 | Perl_bytes_from_utf8(pTHX_ U8* s, STRLEN *len, bool *is_utf8) |
603 | { |
604 | U8 *send; |
605 | U8 *d; |
606 | U8 *start = s; |
607 | I32 count = 0; |
608 | |
609 | if (!*is_utf8) |
610 | return start; |
611 | |
ef9edfd0 |
612 | /* ensure valid UTF8 and chars < 256 before converting string */ |
f9a63242 |
613 | for (send = s + *len; s < send;) { |
614 | U8 c = *s++; |
615 | if (!UTF8_IS_ASCII(c)) { |
616 | if (UTF8_IS_CONTINUATION(c) || s >= send || |
df84a23b |
617 | !UTF8_IS_CONTINUATION(*s) || UTF8_IS_DOWNGRADEABLE_START(c)) |
f9a63242 |
618 | return start; |
619 | s++, count++; |
620 | } |
621 | } |
622 | |
623 | *is_utf8 = 0; |
624 | |
625 | if (!count) |
626 | return start; |
627 | |
628 | Newz(801, d, (*len) - count + 1, U8); |
ef9edfd0 |
629 | s = start; start = d; |
f9a63242 |
630 | while (s < send) { |
631 | U8 c = *s++; |
90f44359 |
632 | |
f9a63242 |
633 | if (UTF8_IS_ASCII(c)) |
634 | *d++ = c; |
635 | else |
90f44359 |
636 | *d++ = UTF8_ACCUMULATE(c, *s++); |
f9a63242 |
637 | } |
638 | *d = '\0'; |
639 | *len = d - start; |
640 | return start; |
641 | } |
642 | |
643 | /* |
eebe1485 |
644 | =for apidoc A|U8 *|bytes_to_utf8|U8 *s|STRLEN *len |
6940069f |
645 | |
646 | Converts a string C<s> of length C<len> from ASCII into UTF8 encoding. |
6662521e |
647 | Returns a pointer to the newly-created string, and sets C<len> to |
648 | reflect the new length. |
6940069f |
649 | |
497711e7 |
650 | =cut |
6940069f |
651 | */ |
652 | |
653 | U8* |
6662521e |
654 | Perl_bytes_to_utf8(pTHX_ U8* s, STRLEN *len) |
6940069f |
655 | { |
6940069f |
656 | U8 *send; |
657 | U8 *d; |
658 | U8 *dst; |
6662521e |
659 | send = s + (*len); |
6940069f |
660 | |
6662521e |
661 | Newz(801, d, (*len) * 2 + 1, U8); |
6940069f |
662 | dst = d; |
663 | |
664 | while (s < send) { |
90f44359 |
665 | if (UTF8_IS_ASCII(*s)) |
6940069f |
666 | *d++ = *s++; |
667 | else { |
668 | UV uv = *s++; |
90f44359 |
669 | |
670 | *d++ = UTF8_EIGHT_BIT_HI(uv); |
671 | *d++ = UTF8_EIGHT_BIT_LO(uv); |
6940069f |
672 | } |
673 | } |
674 | *d = '\0'; |
6662521e |
675 | *len = d-dst; |
6940069f |
676 | return dst; |
677 | } |
678 | |
a0ed51b3 |
679 | /* |
dea0fc0b |
680 | * Convert native (big-endian) or reversed (little-endian) UTF-16 to UTF-8. |
a0ed51b3 |
681 | * |
682 | * Destination must be pre-extended to 3/2 source. Do not use in-place. |
683 | * We optimize for native, for obvious reasons. */ |
684 | |
685 | U8* |
dea0fc0b |
686 | Perl_utf16_to_utf8(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen) |
a0ed51b3 |
687 | { |
dea0fc0b |
688 | U8* pend; |
689 | U8* dstart = d; |
690 | |
691 | if (bytelen & 1) |
a7867d0a |
692 | Perl_croak(aTHX_ "panic: utf16_to_utf8: odd bytelen"); |
dea0fc0b |
693 | |
694 | pend = p + bytelen; |
695 | |
a0ed51b3 |
696 | while (p < pend) { |
dea0fc0b |
697 | UV uv = (p[0] << 8) + p[1]; /* UTF-16BE */ |
698 | p += 2; |
a0ed51b3 |
699 | if (uv < 0x80) { |
700 | *d++ = uv; |
701 | continue; |
702 | } |
703 | if (uv < 0x800) { |
704 | *d++ = (( uv >> 6) | 0xc0); |
705 | *d++ = (( uv & 0x3f) | 0x80); |
706 | continue; |
707 | } |
708 | if (uv >= 0xd800 && uv < 0xdbff) { /* surrogates */ |
dea0fc0b |
709 | UV low = *p++; |
710 | if (low < 0xdc00 || low >= 0xdfff) |
711 | Perl_croak(aTHX_ "Malformed UTF-16 surrogate"); |
a0ed51b3 |
712 | uv = ((uv - 0xd800) << 10) + (low - 0xdc00) + 0x10000; |
713 | } |
714 | if (uv < 0x10000) { |
715 | *d++ = (( uv >> 12) | 0xe0); |
716 | *d++ = (((uv >> 6) & 0x3f) | 0x80); |
717 | *d++ = (( uv & 0x3f) | 0x80); |
718 | continue; |
719 | } |
720 | else { |
721 | *d++ = (( uv >> 18) | 0xf0); |
722 | *d++ = (((uv >> 12) & 0x3f) | 0x80); |
723 | *d++ = (((uv >> 6) & 0x3f) | 0x80); |
724 | *d++ = (( uv & 0x3f) | 0x80); |
725 | continue; |
726 | } |
727 | } |
dea0fc0b |
728 | *newlen = d - dstart; |
a0ed51b3 |
729 | return d; |
730 | } |
731 | |
732 | /* Note: this one is slightly destructive of the source. */ |
733 | |
734 | U8* |
dea0fc0b |
735 | Perl_utf16_to_utf8_reversed(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen) |
a0ed51b3 |
736 | { |
737 | U8* s = (U8*)p; |
738 | U8* send = s + bytelen; |
739 | while (s < send) { |
740 | U8 tmp = s[0]; |
741 | s[0] = s[1]; |
742 | s[1] = tmp; |
743 | s += 2; |
744 | } |
dea0fc0b |
745 | return utf16_to_utf8(p, d, bytelen, newlen); |
a0ed51b3 |
746 | } |
747 | |
748 | /* for now these are all defined (inefficiently) in terms of the utf8 versions */ |
749 | |
750 | bool |
864dbfa3 |
751 | Perl_is_uni_alnum(pTHX_ U32 c) |
a0ed51b3 |
752 | { |
ad391ad9 |
753 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
754 | uv_to_utf8(tmpbuf, (UV)c); |
755 | return is_utf8_alnum(tmpbuf); |
756 | } |
757 | |
758 | bool |
b8c5462f |
759 | Perl_is_uni_alnumc(pTHX_ U32 c) |
760 | { |
ad391ad9 |
761 | U8 tmpbuf[UTF8_MAXLEN+1]; |
b8c5462f |
762 | uv_to_utf8(tmpbuf, (UV)c); |
763 | return is_utf8_alnumc(tmpbuf); |
764 | } |
765 | |
766 | bool |
864dbfa3 |
767 | Perl_is_uni_idfirst(pTHX_ U32 c) |
a0ed51b3 |
768 | { |
ad391ad9 |
769 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
770 | uv_to_utf8(tmpbuf, (UV)c); |
771 | return is_utf8_idfirst(tmpbuf); |
772 | } |
773 | |
774 | bool |
864dbfa3 |
775 | Perl_is_uni_alpha(pTHX_ U32 c) |
a0ed51b3 |
776 | { |
ad391ad9 |
777 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
778 | uv_to_utf8(tmpbuf, (UV)c); |
779 | return is_utf8_alpha(tmpbuf); |
780 | } |
781 | |
782 | bool |
4d61ec05 |
783 | Perl_is_uni_ascii(pTHX_ U32 c) |
784 | { |
ad391ad9 |
785 | U8 tmpbuf[UTF8_MAXLEN+1]; |
4d61ec05 |
786 | uv_to_utf8(tmpbuf, (UV)c); |
787 | return is_utf8_ascii(tmpbuf); |
788 | } |
789 | |
790 | bool |
864dbfa3 |
791 | Perl_is_uni_space(pTHX_ U32 c) |
a0ed51b3 |
792 | { |
ad391ad9 |
793 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
794 | uv_to_utf8(tmpbuf, (UV)c); |
795 | return is_utf8_space(tmpbuf); |
796 | } |
797 | |
798 | bool |
864dbfa3 |
799 | Perl_is_uni_digit(pTHX_ U32 c) |
a0ed51b3 |
800 | { |
ad391ad9 |
801 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
802 | uv_to_utf8(tmpbuf, (UV)c); |
803 | return is_utf8_digit(tmpbuf); |
804 | } |
805 | |
806 | bool |
864dbfa3 |
807 | Perl_is_uni_upper(pTHX_ U32 c) |
a0ed51b3 |
808 | { |
ad391ad9 |
809 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
810 | uv_to_utf8(tmpbuf, (UV)c); |
811 | return is_utf8_upper(tmpbuf); |
812 | } |
813 | |
814 | bool |
864dbfa3 |
815 | Perl_is_uni_lower(pTHX_ U32 c) |
a0ed51b3 |
816 | { |
ad391ad9 |
817 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
818 | uv_to_utf8(tmpbuf, (UV)c); |
819 | return is_utf8_lower(tmpbuf); |
820 | } |
821 | |
822 | bool |
b8c5462f |
823 | Perl_is_uni_cntrl(pTHX_ U32 c) |
824 | { |
ad391ad9 |
825 | U8 tmpbuf[UTF8_MAXLEN+1]; |
b8c5462f |
826 | uv_to_utf8(tmpbuf, (UV)c); |
827 | return is_utf8_cntrl(tmpbuf); |
828 | } |
829 | |
830 | bool |
831 | Perl_is_uni_graph(pTHX_ U32 c) |
832 | { |
ad391ad9 |
833 | U8 tmpbuf[UTF8_MAXLEN+1]; |
b8c5462f |
834 | uv_to_utf8(tmpbuf, (UV)c); |
835 | return is_utf8_graph(tmpbuf); |
836 | } |
837 | |
838 | bool |
864dbfa3 |
839 | Perl_is_uni_print(pTHX_ U32 c) |
a0ed51b3 |
840 | { |
ad391ad9 |
841 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
842 | uv_to_utf8(tmpbuf, (UV)c); |
843 | return is_utf8_print(tmpbuf); |
844 | } |
845 | |
b8c5462f |
846 | bool |
f248d071 |
847 | Perl_is_uni_punct(pTHX_ U32 c) |
b8c5462f |
848 | { |
ad391ad9 |
849 | U8 tmpbuf[UTF8_MAXLEN+1]; |
b8c5462f |
850 | uv_to_utf8(tmpbuf, (UV)c); |
851 | return is_utf8_punct(tmpbuf); |
852 | } |
853 | |
4d61ec05 |
854 | bool |
855 | Perl_is_uni_xdigit(pTHX_ U32 c) |
856 | { |
ad391ad9 |
857 | U8 tmpbuf[UTF8_MAXLEN+1]; |
4d61ec05 |
858 | uv_to_utf8(tmpbuf, (UV)c); |
859 | return is_utf8_xdigit(tmpbuf); |
860 | } |
861 | |
a0ed51b3 |
862 | U32 |
864dbfa3 |
863 | Perl_to_uni_upper(pTHX_ U32 c) |
a0ed51b3 |
864 | { |
ad391ad9 |
865 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
866 | uv_to_utf8(tmpbuf, (UV)c); |
867 | return to_utf8_upper(tmpbuf); |
868 | } |
869 | |
870 | U32 |
864dbfa3 |
871 | Perl_to_uni_title(pTHX_ U32 c) |
a0ed51b3 |
872 | { |
ad391ad9 |
873 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
874 | uv_to_utf8(tmpbuf, (UV)c); |
875 | return to_utf8_title(tmpbuf); |
876 | } |
877 | |
878 | U32 |
864dbfa3 |
879 | Perl_to_uni_lower(pTHX_ U32 c) |
a0ed51b3 |
880 | { |
ad391ad9 |
881 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
882 | uv_to_utf8(tmpbuf, (UV)c); |
883 | return to_utf8_lower(tmpbuf); |
884 | } |
885 | |
886 | /* for now these all assume no locale info available for Unicode > 255 */ |
887 | |
888 | bool |
864dbfa3 |
889 | Perl_is_uni_alnum_lc(pTHX_ U32 c) |
a0ed51b3 |
890 | { |
891 | return is_uni_alnum(c); /* XXX no locale support yet */ |
892 | } |
893 | |
894 | bool |
b8c5462f |
895 | Perl_is_uni_alnumc_lc(pTHX_ U32 c) |
896 | { |
897 | return is_uni_alnumc(c); /* XXX no locale support yet */ |
898 | } |
899 | |
900 | bool |
864dbfa3 |
901 | Perl_is_uni_idfirst_lc(pTHX_ U32 c) |
a0ed51b3 |
902 | { |
903 | return is_uni_idfirst(c); /* XXX no locale support yet */ |
904 | } |
905 | |
906 | bool |
864dbfa3 |
907 | Perl_is_uni_alpha_lc(pTHX_ U32 c) |
a0ed51b3 |
908 | { |
909 | return is_uni_alpha(c); /* XXX no locale support yet */ |
910 | } |
911 | |
912 | bool |
4d61ec05 |
913 | Perl_is_uni_ascii_lc(pTHX_ U32 c) |
914 | { |
915 | return is_uni_ascii(c); /* XXX no locale support yet */ |
916 | } |
917 | |
918 | bool |
864dbfa3 |
919 | Perl_is_uni_space_lc(pTHX_ U32 c) |
a0ed51b3 |
920 | { |
921 | return is_uni_space(c); /* XXX no locale support yet */ |
922 | } |
923 | |
924 | bool |
864dbfa3 |
925 | Perl_is_uni_digit_lc(pTHX_ U32 c) |
a0ed51b3 |
926 | { |
927 | return is_uni_digit(c); /* XXX no locale support yet */ |
928 | } |
929 | |
930 | bool |
864dbfa3 |
931 | Perl_is_uni_upper_lc(pTHX_ U32 c) |
a0ed51b3 |
932 | { |
933 | return is_uni_upper(c); /* XXX no locale support yet */ |
934 | } |
935 | |
936 | bool |
864dbfa3 |
937 | Perl_is_uni_lower_lc(pTHX_ U32 c) |
a0ed51b3 |
938 | { |
939 | return is_uni_lower(c); /* XXX no locale support yet */ |
940 | } |
941 | |
942 | bool |
b8c5462f |
943 | Perl_is_uni_cntrl_lc(pTHX_ U32 c) |
944 | { |
945 | return is_uni_cntrl(c); /* XXX no locale support yet */ |
946 | } |
947 | |
948 | bool |
949 | Perl_is_uni_graph_lc(pTHX_ U32 c) |
950 | { |
951 | return is_uni_graph(c); /* XXX no locale support yet */ |
952 | } |
953 | |
954 | bool |
864dbfa3 |
955 | Perl_is_uni_print_lc(pTHX_ U32 c) |
a0ed51b3 |
956 | { |
957 | return is_uni_print(c); /* XXX no locale support yet */ |
958 | } |
959 | |
b8c5462f |
960 | bool |
961 | Perl_is_uni_punct_lc(pTHX_ U32 c) |
962 | { |
963 | return is_uni_punct(c); /* XXX no locale support yet */ |
964 | } |
965 | |
4d61ec05 |
966 | bool |
967 | Perl_is_uni_xdigit_lc(pTHX_ U32 c) |
968 | { |
969 | return is_uni_xdigit(c); /* XXX no locale support yet */ |
970 | } |
971 | |
a0ed51b3 |
972 | U32 |
864dbfa3 |
973 | Perl_to_uni_upper_lc(pTHX_ U32 c) |
a0ed51b3 |
974 | { |
975 | return to_uni_upper(c); /* XXX no locale support yet */ |
976 | } |
977 | |
978 | U32 |
864dbfa3 |
979 | Perl_to_uni_title_lc(pTHX_ U32 c) |
a0ed51b3 |
980 | { |
981 | return to_uni_title(c); /* XXX no locale support yet */ |
982 | } |
983 | |
984 | U32 |
864dbfa3 |
985 | Perl_to_uni_lower_lc(pTHX_ U32 c) |
a0ed51b3 |
986 | { |
987 | return to_uni_lower(c); /* XXX no locale support yet */ |
988 | } |
989 | |
a0ed51b3 |
990 | bool |
864dbfa3 |
991 | Perl_is_utf8_alnum(pTHX_ U8 *p) |
a0ed51b3 |
992 | { |
386d01d6 |
993 | if (!is_utf8_char(p)) |
994 | return FALSE; |
a0ed51b3 |
995 | if (!PL_utf8_alnum) |
289d4f09 |
996 | /* NOTE: "IsWord", not "IsAlnum", since Alnum is a true |
997 | * descendant of isalnum(3), in other words, it doesn't |
998 | * contain the '_'. --jhi */ |
999 | PL_utf8_alnum = swash_init("utf8", "IsWord", &PL_sv_undef, 0, 0); |
a0ed51b3 |
1000 | return swash_fetch(PL_utf8_alnum, p); |
1001 | /* return *p == '_' || is_utf8_alpha(p) || is_utf8_digit(p); */ |
1002 | #ifdef SURPRISINGLY_SLOWER /* probably because alpha is usually true */ |
1003 | if (!PL_utf8_alnum) |
1004 | PL_utf8_alnum = swash_init("utf8", "", |
1005 | sv_2mortal(newSVpv("+utf8::IsAlpha\n+utf8::IsDigit\n005F\n",0)), 0, 0); |
1006 | return swash_fetch(PL_utf8_alnum, p); |
1007 | #endif |
1008 | } |
1009 | |
1010 | bool |
b8c5462f |
1011 | Perl_is_utf8_alnumc(pTHX_ U8 *p) |
1012 | { |
386d01d6 |
1013 | if (!is_utf8_char(p)) |
1014 | return FALSE; |
b8c5462f |
1015 | if (!PL_utf8_alnum) |
1016 | PL_utf8_alnum = swash_init("utf8", "IsAlnumC", &PL_sv_undef, 0, 0); |
1017 | return swash_fetch(PL_utf8_alnum, p); |
1018 | /* return is_utf8_alpha(p) || is_utf8_digit(p); */ |
1019 | #ifdef SURPRISINGLY_SLOWER /* probably because alpha is usually true */ |
1020 | if (!PL_utf8_alnum) |
1021 | PL_utf8_alnum = swash_init("utf8", "", |
1022 | sv_2mortal(newSVpv("+utf8::IsAlpha\n+utf8::IsDigit\n005F\n",0)), 0, 0); |
1023 | return swash_fetch(PL_utf8_alnum, p); |
1024 | #endif |
1025 | } |
1026 | |
1027 | bool |
864dbfa3 |
1028 | Perl_is_utf8_idfirst(pTHX_ U8 *p) |
a0ed51b3 |
1029 | { |
1030 | return *p == '_' || is_utf8_alpha(p); |
1031 | } |
1032 | |
1033 | bool |
864dbfa3 |
1034 | Perl_is_utf8_alpha(pTHX_ U8 *p) |
a0ed51b3 |
1035 | { |
386d01d6 |
1036 | if (!is_utf8_char(p)) |
1037 | return FALSE; |
a0ed51b3 |
1038 | if (!PL_utf8_alpha) |
e24b16f9 |
1039 | PL_utf8_alpha = swash_init("utf8", "IsAlpha", &PL_sv_undef, 0, 0); |
a0ed51b3 |
1040 | return swash_fetch(PL_utf8_alpha, p); |
1041 | } |
1042 | |
1043 | bool |
b8c5462f |
1044 | Perl_is_utf8_ascii(pTHX_ U8 *p) |
1045 | { |
386d01d6 |
1046 | if (!is_utf8_char(p)) |
1047 | return FALSE; |
b8c5462f |
1048 | if (!PL_utf8_ascii) |
1049 | PL_utf8_ascii = swash_init("utf8", "IsAscii", &PL_sv_undef, 0, 0); |
1050 | return swash_fetch(PL_utf8_ascii, p); |
1051 | } |
1052 | |
1053 | bool |
864dbfa3 |
1054 | Perl_is_utf8_space(pTHX_ U8 *p) |
a0ed51b3 |
1055 | { |
386d01d6 |
1056 | if (!is_utf8_char(p)) |
1057 | return FALSE; |
a0ed51b3 |
1058 | if (!PL_utf8_space) |
3bec3564 |
1059 | PL_utf8_space = swash_init("utf8", "IsSpacePerl", &PL_sv_undef, 0, 0); |
a0ed51b3 |
1060 | return swash_fetch(PL_utf8_space, p); |
1061 | } |
1062 | |
1063 | bool |
864dbfa3 |
1064 | Perl_is_utf8_digit(pTHX_ U8 *p) |
a0ed51b3 |
1065 | { |
386d01d6 |
1066 | if (!is_utf8_char(p)) |
1067 | return FALSE; |
a0ed51b3 |
1068 | if (!PL_utf8_digit) |
e24b16f9 |
1069 | PL_utf8_digit = swash_init("utf8", "IsDigit", &PL_sv_undef, 0, 0); |
a0ed51b3 |
1070 | return swash_fetch(PL_utf8_digit, p); |
1071 | } |
1072 | |
1073 | bool |
864dbfa3 |
1074 | Perl_is_utf8_upper(pTHX_ U8 *p) |
a0ed51b3 |
1075 | { |
386d01d6 |
1076 | if (!is_utf8_char(p)) |
1077 | return FALSE; |
a0ed51b3 |
1078 | if (!PL_utf8_upper) |
e24b16f9 |
1079 | PL_utf8_upper = swash_init("utf8", "IsUpper", &PL_sv_undef, 0, 0); |
a0ed51b3 |
1080 | return swash_fetch(PL_utf8_upper, p); |
1081 | } |
1082 | |
1083 | bool |
864dbfa3 |
1084 | Perl_is_utf8_lower(pTHX_ U8 *p) |
a0ed51b3 |
1085 | { |
386d01d6 |
1086 | if (!is_utf8_char(p)) |
1087 | return FALSE; |
a0ed51b3 |
1088 | if (!PL_utf8_lower) |
e24b16f9 |
1089 | PL_utf8_lower = swash_init("utf8", "IsLower", &PL_sv_undef, 0, 0); |
a0ed51b3 |
1090 | return swash_fetch(PL_utf8_lower, p); |
1091 | } |
1092 | |
1093 | bool |
b8c5462f |
1094 | Perl_is_utf8_cntrl(pTHX_ U8 *p) |
1095 | { |
386d01d6 |
1096 | if (!is_utf8_char(p)) |
1097 | return FALSE; |
b8c5462f |
1098 | if (!PL_utf8_cntrl) |
1099 | PL_utf8_cntrl = swash_init("utf8", "IsCntrl", &PL_sv_undef, 0, 0); |
1100 | return swash_fetch(PL_utf8_cntrl, p); |
1101 | } |
1102 | |
1103 | bool |
1104 | Perl_is_utf8_graph(pTHX_ U8 *p) |
1105 | { |
386d01d6 |
1106 | if (!is_utf8_char(p)) |
1107 | return FALSE; |
b8c5462f |
1108 | if (!PL_utf8_graph) |
1109 | PL_utf8_graph = swash_init("utf8", "IsGraph", &PL_sv_undef, 0, 0); |
1110 | return swash_fetch(PL_utf8_graph, p); |
1111 | } |
1112 | |
1113 | bool |
864dbfa3 |
1114 | Perl_is_utf8_print(pTHX_ U8 *p) |
a0ed51b3 |
1115 | { |
386d01d6 |
1116 | if (!is_utf8_char(p)) |
1117 | return FALSE; |
a0ed51b3 |
1118 | if (!PL_utf8_print) |
e24b16f9 |
1119 | PL_utf8_print = swash_init("utf8", "IsPrint", &PL_sv_undef, 0, 0); |
a0ed51b3 |
1120 | return swash_fetch(PL_utf8_print, p); |
1121 | } |
1122 | |
1123 | bool |
b8c5462f |
1124 | Perl_is_utf8_punct(pTHX_ U8 *p) |
1125 | { |
386d01d6 |
1126 | if (!is_utf8_char(p)) |
1127 | return FALSE; |
b8c5462f |
1128 | if (!PL_utf8_punct) |
1129 | PL_utf8_punct = swash_init("utf8", "IsPunct", &PL_sv_undef, 0, 0); |
1130 | return swash_fetch(PL_utf8_punct, p); |
1131 | } |
1132 | |
1133 | bool |
1134 | Perl_is_utf8_xdigit(pTHX_ U8 *p) |
1135 | { |
386d01d6 |
1136 | if (!is_utf8_char(p)) |
1137 | return FALSE; |
b8c5462f |
1138 | if (!PL_utf8_xdigit) |
1139 | PL_utf8_xdigit = swash_init("utf8", "IsXDigit", &PL_sv_undef, 0, 0); |
1140 | return swash_fetch(PL_utf8_xdigit, p); |
1141 | } |
1142 | |
1143 | bool |
864dbfa3 |
1144 | Perl_is_utf8_mark(pTHX_ U8 *p) |
a0ed51b3 |
1145 | { |
386d01d6 |
1146 | if (!is_utf8_char(p)) |
1147 | return FALSE; |
a0ed51b3 |
1148 | if (!PL_utf8_mark) |
e24b16f9 |
1149 | PL_utf8_mark = swash_init("utf8", "IsM", &PL_sv_undef, 0, 0); |
a0ed51b3 |
1150 | return swash_fetch(PL_utf8_mark, p); |
1151 | } |
1152 | |
2104c8d9 |
1153 | UV |
864dbfa3 |
1154 | Perl_to_utf8_upper(pTHX_ U8 *p) |
a0ed51b3 |
1155 | { |
1156 | UV uv; |
1157 | |
1158 | if (!PL_utf8_toupper) |
e24b16f9 |
1159 | PL_utf8_toupper = swash_init("utf8", "ToUpper", &PL_sv_undef, 4, 0); |
a0ed51b3 |
1160 | uv = swash_fetch(PL_utf8_toupper, p); |
756820e3 |
1161 | return uv ? uv : utf8_to_uv(p,UTF8_MAXLEN,0,0); |
a0ed51b3 |
1162 | } |
1163 | |
2104c8d9 |
1164 | UV |
864dbfa3 |
1165 | Perl_to_utf8_title(pTHX_ U8 *p) |
a0ed51b3 |
1166 | { |
1167 | UV uv; |
1168 | |
1169 | if (!PL_utf8_totitle) |
e24b16f9 |
1170 | PL_utf8_totitle = swash_init("utf8", "ToTitle", &PL_sv_undef, 4, 0); |
a0ed51b3 |
1171 | uv = swash_fetch(PL_utf8_totitle, p); |
756820e3 |
1172 | return uv ? uv : utf8_to_uv(p,UTF8_MAXLEN,0,0); |
a0ed51b3 |
1173 | } |
1174 | |
2104c8d9 |
1175 | UV |
864dbfa3 |
1176 | Perl_to_utf8_lower(pTHX_ U8 *p) |
a0ed51b3 |
1177 | { |
1178 | UV uv; |
1179 | |
1180 | if (!PL_utf8_tolower) |
e24b16f9 |
1181 | PL_utf8_tolower = swash_init("utf8", "ToLower", &PL_sv_undef, 4, 0); |
a0ed51b3 |
1182 | uv = swash_fetch(PL_utf8_tolower, p); |
756820e3 |
1183 | return uv ? uv : utf8_to_uv(p,UTF8_MAXLEN,0,0); |
a0ed51b3 |
1184 | } |
1185 | |
1186 | /* a "swash" is a swatch hash */ |
1187 | |
1188 | SV* |
864dbfa3 |
1189 | Perl_swash_init(pTHX_ char* pkg, char* name, SV *listsv, I32 minbits, I32 none) |
a0ed51b3 |
1190 | { |
1191 | SV* retval; |
bf1fed83 |
1192 | SV* tokenbufsv = sv_2mortal(NEWSV(0,0)); |
8e84507e |
1193 | dSP; |
ce3b816e |
1194 | |
1195 | if (!gv_stashpv(pkg, 0)) { /* demand load utf8 */ |
1196 | ENTER; |
1197 | Perl_load_module(aTHX_ PERL_LOADMOD_NOIMPORT, newSVpv(pkg,0), Nullsv); |
1198 | LEAVE; |
1199 | } |
1200 | SPAGAIN; |
a0ed51b3 |
1201 | PUSHSTACKi(PERLSI_MAGIC); |
1202 | PUSHMARK(SP); |
1203 | EXTEND(SP,5); |
1204 | PUSHs(sv_2mortal(newSVpvn(pkg, strlen(pkg)))); |
1205 | PUSHs(sv_2mortal(newSVpvn(name, strlen(name)))); |
1206 | PUSHs(listsv); |
1207 | PUSHs(sv_2mortal(newSViv(minbits))); |
1208 | PUSHs(sv_2mortal(newSViv(none))); |
1209 | PUTBACK; |
1210 | ENTER; |
1211 | SAVEI32(PL_hints); |
1212 | PL_hints = 0; |
1213 | save_re_context(); |
bf1fed83 |
1214 | if (PL_curcop == &PL_compiling) |
1215 | /* XXX ought to be handled by lex_start */ |
1216 | sv_setpv(tokenbufsv, PL_tokenbuf); |
864dbfa3 |
1217 | if (call_method("SWASHNEW", G_SCALAR)) |
8e84507e |
1218 | retval = newSVsv(*PL_stack_sp--); |
a0ed51b3 |
1219 | else |
e24b16f9 |
1220 | retval = &PL_sv_undef; |
a0ed51b3 |
1221 | LEAVE; |
1222 | POPSTACK; |
e24b16f9 |
1223 | if (PL_curcop == &PL_compiling) { |
bf1fed83 |
1224 | STRLEN len; |
1225 | char* pv = SvPV(tokenbufsv, len); |
1226 | |
1227 | Copy(pv, PL_tokenbuf, len+1, char); |
e24b16f9 |
1228 | PL_curcop->op_private = PL_hints; |
a0ed51b3 |
1229 | } |
1230 | if (!SvROK(retval) || SvTYPE(SvRV(retval)) != SVt_PVHV) |
cea2e8a9 |
1231 | Perl_croak(aTHX_ "SWASHNEW didn't return an HV ref"); |
a0ed51b3 |
1232 | return retval; |
1233 | } |
1234 | |
1235 | UV |
864dbfa3 |
1236 | Perl_swash_fetch(pTHX_ SV *sv, U8 *ptr) |
a0ed51b3 |
1237 | { |
1238 | HV* hv = (HV*)SvRV(sv); |
1239 | U32 klen = UTF8SKIP(ptr) - 1; |
1240 | U32 off = ptr[klen] & 127; /* NB: 64 bit always 0 when len > 1 */ |
1241 | STRLEN slen; |
1242 | STRLEN needents = (klen ? 64 : 128); |
dfe13c55 |
1243 | U8 *tmps; |
a0ed51b3 |
1244 | U32 bit; |
1245 | SV *retval; |
1246 | |
1247 | /* |
1248 | * This single-entry cache saves about 1/3 of the utf8 overhead in test |
1249 | * suite. (That is, only 7-8% overall over just a hash cache. Still, |
1250 | * it's nothing to sniff at.) Pity we usually come through at least |
1251 | * two function calls to get here... |
1252 | * |
1253 | * NB: this code assumes that swatches are never modified, once generated! |
1254 | */ |
1255 | |
1256 | if (hv == PL_last_swash_hv && |
1257 | klen == PL_last_swash_klen && |
12ae5dfc |
1258 | (!klen || memEQ((char *)ptr,(char *)PL_last_swash_key,klen)) ) |
a0ed51b3 |
1259 | { |
1260 | tmps = PL_last_swash_tmps; |
1261 | slen = PL_last_swash_slen; |
1262 | } |
1263 | else { |
1264 | /* Try our second-level swatch cache, kept in a hash. */ |
dfe13c55 |
1265 | SV** svp = hv_fetch(hv, (char*)ptr, klen, FALSE); |
a0ed51b3 |
1266 | |
1267 | /* If not cached, generate it via utf8::SWASHGET */ |
dfe13c55 |
1268 | if (!svp || !SvPOK(*svp) || !(tmps = (U8*)SvPV(*svp, slen))) { |
a0ed51b3 |
1269 | dSP; |
1270 | ENTER; |
1271 | SAVETMPS; |
1272 | save_re_context(); |
1273 | PUSHSTACKi(PERLSI_MAGIC); |
1274 | PUSHMARK(SP); |
1275 | EXTEND(SP,3); |
1276 | PUSHs((SV*)sv); |
756820e3 |
1277 | PUSHs(sv_2mortal(newSViv(utf8_to_uv(ptr, UTF8_MAXLEN, 0, 0) & ~(needents - 1)))); |
a0ed51b3 |
1278 | PUSHs(sv_2mortal(newSViv(needents))); |
1279 | PUTBACK; |
864dbfa3 |
1280 | if (call_method("SWASHGET", G_SCALAR)) |
8e84507e |
1281 | retval = newSVsv(*PL_stack_sp--); |
a0ed51b3 |
1282 | else |
e24b16f9 |
1283 | retval = &PL_sv_undef; |
a0ed51b3 |
1284 | POPSTACK; |
1285 | FREETMPS; |
1286 | LEAVE; |
e24b16f9 |
1287 | if (PL_curcop == &PL_compiling) |
1288 | PL_curcop->op_private = PL_hints; |
a0ed51b3 |
1289 | |
dfe13c55 |
1290 | svp = hv_store(hv, (char*)ptr, klen, retval, 0); |
a0ed51b3 |
1291 | |
dfe13c55 |
1292 | if (!svp || !(tmps = (U8*)SvPV(*svp, slen)) || slen < 8) |
cea2e8a9 |
1293 | Perl_croak(aTHX_ "SWASHGET didn't return result of proper length"); |
a0ed51b3 |
1294 | } |
1295 | |
1296 | PL_last_swash_hv = hv; |
1297 | PL_last_swash_klen = klen; |
1298 | PL_last_swash_tmps = tmps; |
1299 | PL_last_swash_slen = slen; |
1300 | if (klen) |
1301 | Copy(ptr, PL_last_swash_key, klen, U8); |
1302 | } |
1303 | |
9faf8d75 |
1304 | switch ((int)((slen << 3) / needents)) { |
a0ed51b3 |
1305 | case 1: |
1306 | bit = 1 << (off & 7); |
1307 | off >>= 3; |
1308 | return (tmps[off] & bit) != 0; |
1309 | case 8: |
1310 | return tmps[off]; |
1311 | case 16: |
1312 | off <<= 1; |
1313 | return (tmps[off] << 8) + tmps[off + 1] ; |
1314 | case 32: |
1315 | off <<= 2; |
1316 | return (tmps[off] << 24) + (tmps[off+1] << 16) + (tmps[off+2] << 8) + tmps[off + 3] ; |
1317 | } |
cea2e8a9 |
1318 | Perl_croak(aTHX_ "panic: swash_fetch"); |
a0ed51b3 |
1319 | return 0; |
1320 | } |