Commit | Line | Data |
---|---|---|
a0ed51b3 LW |
1 | /* utf8.c |
2 | * | |
bc89e66f | 3 | * Copyright (c) 1998-2001, Larry Wall |
a0ed51b3 LW |
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 LW |
25 | #include "perl.h" |
26 | ||
27 | /* Unicode support */ | |
28 | ||
eebe1485 SC |
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 LW |
48 | { |
49 | if (uv < 0x80) { | |
50 | *d++ = uv; | |
51 | return d; | |
52 | } | |
53 | if (uv < 0x800) { | |
54 | *d++ = (( uv >> 6) | 0xc0); | |
55 | *d++ = (( uv & 0x3f) | 0x80); | |
56 | return d; | |
57 | } | |
58 | if (uv < 0x10000) { | |
59 | *d++ = (( uv >> 12) | 0xe0); | |
60 | *d++ = (((uv >> 6) & 0x3f) | 0x80); | |
61 | *d++ = (( uv & 0x3f) | 0x80); | |
62 | return d; | |
63 | } | |
64 | if (uv < 0x200000) { | |
65 | *d++ = (( uv >> 18) | 0xf0); | |
66 | *d++ = (((uv >> 12) & 0x3f) | 0x80); | |
67 | *d++ = (((uv >> 6) & 0x3f) | 0x80); | |
68 | *d++ = (( uv & 0x3f) | 0x80); | |
69 | return d; | |
70 | } | |
71 | if (uv < 0x4000000) { | |
72 | *d++ = (( uv >> 24) | 0xf8); | |
73 | *d++ = (((uv >> 18) & 0x3f) | 0x80); | |
74 | *d++ = (((uv >> 12) & 0x3f) | 0x80); | |
75 | *d++ = (((uv >> 6) & 0x3f) | 0x80); | |
76 | *d++ = (( uv & 0x3f) | 0x80); | |
77 | return d; | |
78 | } | |
79 | if (uv < 0x80000000) { | |
80 | *d++ = (( uv >> 30) | 0xfc); | |
81 | *d++ = (((uv >> 24) & 0x3f) | 0x80); | |
82 | *d++ = (((uv >> 18) & 0x3f) | 0x80); | |
83 | *d++ = (((uv >> 12) & 0x3f) | 0x80); | |
84 | *d++ = (((uv >> 6) & 0x3f) | 0x80); | |
85 | *d++ = (( uv & 0x3f) | 0x80); | |
86 | return d; | |
87 | } | |
6b8eaf93 | 88 | #ifdef HAS_QUAD |
d7578b48 | 89 | if (uv < UTF8_QUAD_MAX) |
a0ed51b3 LW |
90 | #endif |
91 | { | |
92 | *d++ = 0xfe; /* Can't match U+FEFF! */ | |
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 | } | |
6b8eaf93 | 101 | #ifdef HAS_QUAD |
a0ed51b3 LW |
102 | { |
103 | *d++ = 0xff; /* Can't match U+FFFE! */ | |
3c77ea2b GS |
104 | *d++ = 0x80; /* 6 Reserved bits */ |
105 | *d++ = (((uv >> 60) & 0x0f) | 0x80); /* 2 Reserved bits */ | |
106 | *d++ = (((uv >> 54) & 0x3f) | 0x80); | |
107 | *d++ = (((uv >> 48) & 0x3f) | 0x80); | |
108 | *d++ = (((uv >> 42) & 0x3f) | 0x80); | |
a0ed51b3 LW |
109 | *d++ = (((uv >> 36) & 0x3f) | 0x80); |
110 | *d++ = (((uv >> 30) & 0x3f) | 0x80); | |
111 | *d++ = (((uv >> 24) & 0x3f) | 0x80); | |
112 | *d++ = (((uv >> 18) & 0x3f) | 0x80); | |
113 | *d++ = (((uv >> 12) & 0x3f) | 0x80); | |
114 | *d++ = (((uv >> 6) & 0x3f) | 0x80); | |
115 | *d++ = (( uv & 0x3f) | 0x80); | |
116 | return d; | |
117 | } | |
118 | #endif | |
119 | } | |
120 | ||
eebe1485 SC |
121 | /* |
122 | =for apidoc A|STRLEN|is_utf8_char|U8 *s | |
123 | ||
5da9da9e JH |
124 | Tests if some arbitrary number of bytes begins in a valid UTF-8 |
125 | character. Note that an ASCII character is a valid UTF-8 character. | |
126 | The actual number of bytes in the UTF-8 character will be returned if | |
127 | it is valid, otherwise 0. | |
eebe1485 | 128 | |
5da9da9e | 129 | =cut */ |
067a85ef | 130 | STRLEN |
386d01d6 GS |
131 | Perl_is_utf8_char(pTHX_ U8 *s) |
132 | { | |
133 | U8 u = *s; | |
067a85ef A |
134 | STRLEN slen, len; |
135 | UV uv, ouv; | |
386d01d6 | 136 | |
60006e79 | 137 | if (UTF8_IS_ASCII(u)) |
386d01d6 GS |
138 | return 1; |
139 | ||
60006e79 | 140 | if (!UTF8_IS_START(u)) |
386d01d6 GS |
141 | return 0; |
142 | ||
9f07fdcd | 143 | len = UTF8SKIP(s); |
386d01d6 | 144 | |
60006e79 | 145 | if (len < 2 || !UTF8_IS_CONTINUATION(s[1])) |
067a85ef A |
146 | return 0; |
147 | ||
386d01d6 GS |
148 | slen = len - 1; |
149 | s++; | |
067a85ef A |
150 | uv = u; |
151 | ouv = uv; | |
386d01d6 | 152 | while (slen--) { |
60006e79 | 153 | if (!UTF8_IS_CONTINUATION(*s)) |
386d01d6 | 154 | return 0; |
8850bf83 | 155 | uv = UTF8_ACCUMULATE(uv, *s); |
067a85ef A |
156 | if (uv < ouv) |
157 | return 0; | |
158 | ouv = uv; | |
386d01d6 GS |
159 | s++; |
160 | } | |
067a85ef | 161 | |
5bbb0b5a | 162 | if (UNISKIP(uv) < len) |
067a85ef A |
163 | return 0; |
164 | ||
386d01d6 GS |
165 | return len; |
166 | } | |
167 | ||
6662521e | 168 | /* |
eebe1485 | 169 | =for apidoc A|bool|is_utf8_string|U8 *s|STRLEN len |
6662521e | 170 | |
5da9da9e JH |
171 | Returns true if first C<len> bytes of the given string form a valid UTF8 |
172 | string, false otherwise. Note that 'a valid UTF8 string' does not mean | |
173 | 'a string that contains UTF8' because a valid ASCII string is a valid | |
174 | UTF8 string. | |
6662521e GS |
175 | |
176 | =cut | |
177 | */ | |
178 | ||
8e84507e | 179 | bool |
6662521e GS |
180 | Perl_is_utf8_string(pTHX_ U8 *s, STRLEN len) |
181 | { | |
067a85ef | 182 | U8* x = s; |
1aa99e6b | 183 | U8* send; |
067a85ef A |
184 | STRLEN c; |
185 | ||
1aa99e6b | 186 | if (!len) |
6cd5fe39 | 187 | len = strlen((char *)s); |
1aa99e6b IH |
188 | send = s + len; |
189 | ||
6662521e GS |
190 | while (x < send) { |
191 | c = is_utf8_char(x); | |
067a85ef A |
192 | if (!c) |
193 | return FALSE; | |
6662521e | 194 | x += c; |
6662521e | 195 | } |
60006e79 JH |
196 | if (x != send) |
197 | return FALSE; | |
067a85ef A |
198 | |
199 | return TRUE; | |
6662521e GS |
200 | } |
201 | ||
67e989fb | 202 | /* |
eebe1485 | 203 | =for apidoc A|U8* s|utf8_to_uv|STRLEN curlen|STRLEN *retlen|U32 flags |
67e989fb JH |
204 | |
205 | Returns the character value of the first character in the string C<s> | |
ba210ebe | 206 | which is assumed to be in UTF8 encoding and no longer than C<curlen>; |
7df053ec | 207 | C<retlen> will be set to the length, in bytes, of that character. |
67e989fb JH |
208 | |
209 | If C<s> does not point to a well-formed UTF8 character, the behaviour | |
dcad2880 JH |
210 | is dependent on the value of C<flags>: if it contains UTF8_CHECK_ONLY, |
211 | it is assumed that the caller will raise a warning, and this function | |
28d3d195 JH |
212 | will silently just set C<retlen> to C<-1> and return zero. If the |
213 | C<flags> does not contain UTF8_CHECK_ONLY, warnings about | |
214 | malformations will be given, C<retlen> will be set to the expected | |
215 | length of the UTF-8 character in bytes, and zero will be returned. | |
216 | ||
217 | The C<flags> can also contain various flags to allow deviations from | |
218 | the strict UTF-8 encoding (see F<utf8.h>). | |
67e989fb | 219 | |
dcad2880 | 220 | =cut */ |
67e989fb | 221 | |
a0ed51b3 | 222 | UV |
dcad2880 | 223 | Perl_utf8_to_uv(pTHX_ U8* s, STRLEN curlen, STRLEN* retlen, U32 flags) |
a0ed51b3 | 224 | { |
ba210ebe JH |
225 | UV uv = *s, ouv; |
226 | STRLEN len = 1; | |
7bf1b6bb PP |
227 | #ifdef EBCDIC |
228 | bool dowarn = 0; | |
229 | #else | |
ba210ebe | 230 | bool dowarn = ckWARN_d(WARN_UTF8); |
7bf1b6bb | 231 | #endif |
ba210ebe | 232 | STRLEN expectlen = 0; |
a0dbb045 JH |
233 | U32 warning = 0; |
234 | ||
235 | /* This list is a superset of the UTF8_ALLOW_XXX. */ | |
236 | ||
237 | #define UTF8_WARN_EMPTY 1 | |
238 | #define UTF8_WARN_CONTINUATION 2 | |
239 | #define UTF8_WARN_NON_CONTINUATION 3 | |
240 | #define UTF8_WARN_FE_FF 4 | |
241 | #define UTF8_WARN_SHORT 5 | |
242 | #define UTF8_WARN_OVERFLOW 6 | |
243 | #define UTF8_WARN_SURROGATE 7 | |
244 | #define UTF8_WARN_BOM 8 | |
245 | #define UTF8_WARN_LONG 9 | |
246 | #define UTF8_WARN_FFFF 10 | |
247 | ||
248 | if (curlen == 0 && | |
249 | !(flags & UTF8_ALLOW_EMPTY)) { | |
250 | warning = UTF8_WARN_EMPTY; | |
0c443dc2 JH |
251 | goto malformed; |
252 | } | |
253 | ||
421a8bf2 | 254 | if (UTF8_IS_ASCII(uv)) { |
a0ed51b3 LW |
255 | if (retlen) |
256 | *retlen = 1; | |
257 | return *s; | |
258 | } | |
67e989fb | 259 | |
421a8bf2 | 260 | if (UTF8_IS_CONTINUATION(uv) && |
fcc8fcf6 | 261 | !(flags & UTF8_ALLOW_CONTINUATION)) { |
a0dbb045 | 262 | warning = UTF8_WARN_CONTINUATION; |
ba210ebe JH |
263 | goto malformed; |
264 | } | |
265 | ||
421a8bf2 | 266 | if (UTF8_IS_START(uv) && curlen > 1 && !UTF8_IS_CONTINUATION(s[1]) && |
fcc8fcf6 | 267 | !(flags & UTF8_ALLOW_NON_CONTINUATION)) { |
a0dbb045 | 268 | warning = UTF8_WARN_NON_CONTINUATION; |
ba210ebe JH |
269 | goto malformed; |
270 | } | |
fcc8fcf6 JH |
271 | |
272 | if ((uv == 0xfe || uv == 0xff) && | |
273 | !(flags & UTF8_ALLOW_FE_FF)) { | |
a0dbb045 | 274 | warning = UTF8_WARN_FE_FF; |
ba210ebe | 275 | goto malformed; |
a0ed51b3 | 276 | } |
fcc8fcf6 | 277 | |
ba210ebe JH |
278 | if (!(uv & 0x20)) { len = 2; uv &= 0x1f; } |
279 | else if (!(uv & 0x10)) { len = 3; uv &= 0x0f; } | |
280 | else if (!(uv & 0x08)) { len = 4; uv &= 0x07; } | |
281 | else if (!(uv & 0x04)) { len = 5; uv &= 0x03; } | |
282 | else if (!(uv & 0x02)) { len = 6; uv &= 0x01; } | |
283 | else if (!(uv & 0x01)) { len = 7; uv = 0; } | |
3c77ea2b | 284 | else { len = 13; uv = 0; } /* whoa! */ |
fcc8fcf6 | 285 | |
a0ed51b3 LW |
286 | if (retlen) |
287 | *retlen = len; | |
ba210ebe JH |
288 | |
289 | expectlen = len; | |
290 | ||
fcc8fcf6 JH |
291 | if ((curlen < expectlen) && |
292 | !(flags & UTF8_ALLOW_SHORT)) { | |
a0dbb045 | 293 | warning = UTF8_WARN_SHORT; |
ba210ebe JH |
294 | goto malformed; |
295 | } | |
296 | ||
297 | len--; | |
a0ed51b3 | 298 | s++; |
ba210ebe JH |
299 | ouv = uv; |
300 | ||
a0ed51b3 | 301 | while (len--) { |
421a8bf2 JH |
302 | if (!UTF8_IS_CONTINUATION(*s) && |
303 | !(flags & UTF8_ALLOW_NON_CONTINUATION)) { | |
a0dbb045 JH |
304 | s--; |
305 | warning = UTF8_WARN_NON_CONTINUATION; | |
ba210ebe | 306 | goto malformed; |
a0ed51b3 LW |
307 | } |
308 | else | |
8850bf83 | 309 | uv = UTF8_ACCUMULATE(uv, *s); |
a0dbb045 JH |
310 | if (!(uv > ouv)) { |
311 | /* These cannot be allowed. */ | |
312 | if (uv == ouv) { | |
313 | if (!(flags & UTF8_ALLOW_LONG)) { | |
314 | warning = UTF8_WARN_LONG; | |
315 | goto malformed; | |
316 | } | |
317 | } | |
318 | else { /* uv < ouv */ | |
319 | /* This cannot be allowed. */ | |
320 | warning = UTF8_WARN_OVERFLOW; | |
321 | goto malformed; | |
322 | } | |
ba210ebe JH |
323 | } |
324 | s++; | |
325 | ouv = uv; | |
326 | } | |
327 | ||
421a8bf2 | 328 | if (UNICODE_IS_SURROGATE(uv) && |
fcc8fcf6 | 329 | !(flags & UTF8_ALLOW_SURROGATE)) { |
a0dbb045 | 330 | warning = UTF8_WARN_SURROGATE; |
ba210ebe | 331 | goto malformed; |
421a8bf2 | 332 | } else if (UNICODE_IS_BYTE_ORDER_MARK(uv) && |
fcc8fcf6 | 333 | !(flags & UTF8_ALLOW_BOM)) { |
a0dbb045 | 334 | warning = UTF8_WARN_BOM; |
ba210ebe | 335 | goto malformed; |
fcc8fcf6 JH |
336 | } else if ((expectlen > UNISKIP(uv)) && |
337 | !(flags & UTF8_ALLOW_LONG)) { | |
a0dbb045 | 338 | warning = UTF8_WARN_LONG; |
ba210ebe | 339 | goto malformed; |
421a8bf2 | 340 | } else if (UNICODE_IS_ILLEGAL(uv) && |
a9917092 | 341 | !(flags & UTF8_ALLOW_FFFF)) { |
a0dbb045 | 342 | warning = UTF8_WARN_FFFF; |
a9917092 | 343 | goto malformed; |
a0ed51b3 | 344 | } |
ba210ebe | 345 | |
a0ed51b3 | 346 | return uv; |
ba210ebe JH |
347 | |
348 | malformed: | |
349 | ||
fcc8fcf6 | 350 | if (flags & UTF8_CHECK_ONLY) { |
ba210ebe | 351 | if (retlen) |
cc366d4b | 352 | *retlen = -1; |
ba210ebe JH |
353 | return 0; |
354 | } | |
355 | ||
a0dbb045 JH |
356 | if (dowarn) { |
357 | SV* sv = sv_2mortal(newSVpv("Malformed UTF-8 character ", 0)); | |
358 | ||
359 | switch (warning) { | |
360 | case 0: /* Intentionally empty. */ break; | |
361 | case UTF8_WARN_EMPTY: | |
362 | Perl_sv_catpvf(aTHX_ sv, "(empty string)"); | |
363 | break; | |
364 | case UTF8_WARN_CONTINUATION: | |
365 | Perl_sv_catpvf(aTHX_ sv, "(unexpected continuation byte 0x%02"UVxf")", uv); | |
366 | break; | |
367 | case UTF8_WARN_NON_CONTINUATION: | |
368 | Perl_sv_catpvf(aTHX_ sv, "(unexpected non-continuation byte 0x%02"UVxf" after start byte 0x%02"UVxf")", | |
369 | (UV)s[1], uv); | |
370 | break; | |
371 | case UTF8_WARN_FE_FF: | |
372 | Perl_sv_catpvf(aTHX_ sv, "(byte 0x%02"UVxf")", uv); | |
373 | break; | |
374 | case UTF8_WARN_SHORT: | |
375 | Perl_sv_catpvf(aTHX_ sv, "(%d byte%s, need %d)", | |
376 | curlen, curlen == 1 ? "" : "s", expectlen); | |
377 | break; | |
378 | case UTF8_WARN_OVERFLOW: | |
379 | Perl_sv_catpvf(aTHX_ sv, "(overflow at 0x%"UVxf", byte 0x%02x)", | |
380 | ouv, *s); | |
381 | break; | |
382 | case UTF8_WARN_SURROGATE: | |
383 | Perl_sv_catpvf(aTHX_ sv, "(UTF-16 surrogate 0x%04"UVxf")", uv); | |
384 | break; | |
385 | case UTF8_WARN_BOM: | |
386 | Perl_sv_catpvf(aTHX_ sv, "(byte order mark 0x%04"UVxf")", uv); | |
387 | break; | |
388 | case UTF8_WARN_LONG: | |
389 | Perl_sv_catpvf(aTHX_ sv, "(%d byte%s, need %d)", | |
390 | expectlen, expectlen == 1 ? "": "s", UNISKIP(uv)); | |
391 | break; | |
392 | case UTF8_WARN_FFFF: | |
393 | Perl_sv_catpvf(aTHX_ sv, "(character 0x%04"UVxf")", uv); | |
394 | break; | |
395 | default: | |
396 | Perl_sv_catpvf(aTHX_ sv, "(unknown reason)"); | |
397 | break; | |
398 | } | |
399 | ||
400 | if (warning) { | |
401 | char *s = SvPVX(sv); | |
402 | ||
403 | if (PL_op) | |
404 | Perl_warner(aTHX_ WARN_UTF8, | |
405 | "%s in %s", s, PL_op_desc[PL_op->op_type]); | |
406 | else | |
407 | Perl_warner(aTHX_ WARN_UTF8, "%s", s); | |
408 | } | |
409 | } | |
410 | ||
ba210ebe | 411 | if (retlen) |
28d3d195 | 412 | *retlen = expectlen ? expectlen : len; |
ba210ebe | 413 | |
28d3d195 | 414 | return 0; |
a0ed51b3 LW |
415 | } |
416 | ||
8e84507e | 417 | /* |
eebe1485 | 418 | =for apidoc A|U8* s|utf8_to_uv_simple|STRLEN *retlen |
8e84507e NIS |
419 | |
420 | Returns the character value of the first character in the string C<s> | |
421 | which is assumed to be in UTF8 encoding; C<retlen> will be set to the | |
7df053ec | 422 | length, in bytes, of that character. |
8e84507e | 423 | |
ba210ebe JH |
424 | If C<s> does not point to a well-formed UTF8 character, zero is |
425 | returned and retlen is set, if possible, to -1. | |
8e84507e NIS |
426 | |
427 | =cut | |
428 | */ | |
429 | ||
430 | UV | |
dcad2880 | 431 | Perl_utf8_to_uv_simple(pTHX_ U8* s, STRLEN* retlen) |
8e84507e | 432 | { |
2e4dc9fc | 433 | return Perl_utf8_to_uv(aTHX_ s, UTF8_MAXLEN, retlen, 0); |
8e84507e NIS |
434 | } |
435 | ||
b76347f2 | 436 | /* |
eebe1485 | 437 | =for apidoc A|STRLEN|utf8_length|U8* s|U8 *e |
b76347f2 JH |
438 | |
439 | Return the length of the UTF-8 char encoded string C<s> in characters. | |
02eb7b47 JH |
440 | Stops at C<e> (inclusive). If C<e E<lt> s> or if the scan would end |
441 | up past C<e>, croaks. | |
b76347f2 JH |
442 | |
443 | =cut | |
444 | */ | |
445 | ||
446 | STRLEN | |
447 | Perl_utf8_length(pTHX_ U8* s, U8* e) | |
448 | { | |
449 | STRLEN len = 0; | |
450 | ||
8850bf83 JH |
451 | /* Note: cannot use UTF8_IS_...() too eagerly here since e.g. |
452 | * the bitops (especially ~) can create illegal UTF-8. | |
453 | * In other words: in Perl UTF-8 is not just for Unicode. */ | |
454 | ||
b76347f2 | 455 | if (e < s) |
02eb7b47 | 456 | Perl_croak(aTHX_ "panic: utf8_length: unexpected end"); |
b76347f2 | 457 | while (s < e) { |
02eb7b47 | 458 | U8 t = UTF8SKIP(s); |
b76347f2 JH |
459 | |
460 | if (e - s < t) | |
02eb7b47 | 461 | Perl_croak(aTHX_ "panic: utf8_length: unaligned end"); |
b76347f2 JH |
462 | s += t; |
463 | len++; | |
464 | } | |
465 | ||
466 | return len; | |
467 | } | |
468 | ||
b06226ff | 469 | /* |
eebe1485 | 470 | =for apidoc A|IV|utf8_distance|U8 *a|U8 *b |
b06226ff JH |
471 | |
472 | Returns the number of UTF8 characters between the UTF-8 pointers C<a> | |
473 | and C<b>. | |
474 | ||
475 | WARNING: use only if you *know* that the pointers point inside the | |
476 | same UTF-8 buffer. | |
477 | ||
478 | =cut */ | |
a0ed51b3 | 479 | |
02eb7b47 | 480 | IV |
864dbfa3 | 481 | Perl_utf8_distance(pTHX_ U8 *a, U8 *b) |
a0ed51b3 | 482 | { |
02eb7b47 JH |
483 | IV off = 0; |
484 | ||
8850bf83 JH |
485 | /* Note: cannot use UTF8_IS_...() too eagerly here since e.g. |
486 | * the bitops (especially ~) can create illegal UTF-8. | |
487 | * In other words: in Perl UTF-8 is not just for Unicode. */ | |
488 | ||
a0ed51b3 LW |
489 | if (a < b) { |
490 | while (a < b) { | |
02eb7b47 JH |
491 | U8 c = UTF8SKIP(a); |
492 | ||
493 | if (b - a < c) | |
494 | Perl_croak(aTHX_ "panic: utf8_distance: unaligned end"); | |
495 | a += c; | |
a0ed51b3 LW |
496 | off--; |
497 | } | |
498 | } | |
499 | else { | |
500 | while (b < a) { | |
02eb7b47 JH |
501 | U8 c = UTF8SKIP(b); |
502 | ||
503 | if (a - b < c) | |
504 | Perl_croak(aTHX_ "panic: utf8_distance: unaligned end"); | |
505 | b += c; | |
a0ed51b3 LW |
506 | off++; |
507 | } | |
508 | } | |
02eb7b47 | 509 | |
a0ed51b3 LW |
510 | return off; |
511 | } | |
512 | ||
b06226ff | 513 | /* |
eebe1485 | 514 | =for apidoc A|U8*|utf8_hop|U8 *s|I32 off |
b06226ff | 515 | |
8850bf83 JH |
516 | Return the UTF-8 pointer C<s> displaced by C<off> characters, either |
517 | forward or backward. | |
b06226ff JH |
518 | |
519 | WARNING: do not use the following unless you *know* C<off> is within | |
8850bf83 JH |
520 | the UTF-8 data pointed to by C<s> *and* that on entry C<s> is aligned |
521 | on the first byte of character or just after the last byte of a character. | |
b06226ff JH |
522 | |
523 | =cut */ | |
a0ed51b3 LW |
524 | |
525 | U8 * | |
864dbfa3 | 526 | Perl_utf8_hop(pTHX_ U8 *s, I32 off) |
a0ed51b3 | 527 | { |
8850bf83 JH |
528 | /* Note: cannot use UTF8_IS_...() too eagerly here since e.g |
529 | * the bitops (especially ~) can create illegal UTF-8. | |
530 | * In other words: in Perl UTF-8 is not just for Unicode. */ | |
531 | ||
a0ed51b3 LW |
532 | if (off >= 0) { |
533 | while (off--) | |
534 | s += UTF8SKIP(s); | |
535 | } | |
536 | else { | |
537 | while (off++) { | |
538 | s--; | |
8850bf83 JH |
539 | while (UTF8_IS_CONTINUATION(*s)) |
540 | s--; | |
a0ed51b3 LW |
541 | } |
542 | } | |
543 | return s; | |
544 | } | |
545 | ||
6940069f | 546 | /* |
eebe1485 | 547 | =for apidoc A|U8 *|utf8_to_bytes|U8 *s|STRLEN *len |
6940069f | 548 | |
246fae53 MG |
549 | Converts a string C<s> of length C<len> from UTF8 into byte encoding. |
550 | Unlike C<bytes_to_utf8>, this over-writes the original string, and | |
551 | updates len to contain the new length. | |
67e989fb | 552 | Returns zero on failure, setting C<len> to -1. |
6940069f GS |
553 | |
554 | =cut | |
555 | */ | |
556 | ||
557 | U8 * | |
246fae53 | 558 | Perl_utf8_to_bytes(pTHX_ U8* s, STRLEN *len) |
6940069f | 559 | { |
6940069f GS |
560 | U8 *send; |
561 | U8 *d; | |
dcad2880 | 562 | U8 *save = s; |
246fae53 MG |
563 | |
564 | /* ensure valid UTF8 and chars < 256 before updating string */ | |
dcad2880 JH |
565 | for (send = s + *len; s < send; ) { |
566 | U8 c = *s++; | |
567 | ||
9f9ab905 | 568 | if (c >= 0x80 && |
dcad2880 JH |
569 | ((s >= send) || |
570 | ((*s++ & 0xc0) != 0x80) || ((c & 0xfe) != 0xc2))) { | |
571 | *len = -1; | |
572 | return 0; | |
573 | } | |
246fae53 | 574 | } |
dcad2880 JH |
575 | |
576 | d = s = save; | |
6940069f | 577 | while (s < send) { |
ed646e6e SC |
578 | STRLEN ulen; |
579 | *d++ = (U8)utf8_to_uv_simple(s, &ulen); | |
580 | s += ulen; | |
6940069f GS |
581 | } |
582 | *d = '\0'; | |
246fae53 | 583 | *len = d - save; |
6940069f GS |
584 | return save; |
585 | } | |
586 | ||
587 | /* | |
f9a63242 JH |
588 | =for apidoc A|U8 *|bytes_from_utf8|U8 *s|STRLEN *len|bool *is_utf8 |
589 | ||
590 | Converts a string C<s> of length C<len> from UTF8 into byte encoding. | |
591 | Unlike <utf8_to_bytes> but like C<bytes_to_utf8>, returns a pointer to | |
ef9edfd0 JH |
592 | the newly-created string, and updates C<len> to contain the new |
593 | length. Returns the original string if no conversion occurs, C<len> | |
594 | is unchanged. Do nothing if C<is_utf8> points to 0. Sets C<is_utf8> to | |
595 | 0 if C<s> is converted or contains all 7bit characters. | |
f9a63242 JH |
596 | |
597 | =cut */ | |
598 | ||
599 | U8 * | |
600 | Perl_bytes_from_utf8(pTHX_ U8* s, STRLEN *len, bool *is_utf8) | |
601 | { | |
602 | U8 *send; | |
603 | U8 *d; | |
604 | U8 *start = s; | |
605 | I32 count = 0; | |
606 | ||
607 | if (!*is_utf8) | |
608 | return start; | |
609 | ||
ef9edfd0 | 610 | /* ensure valid UTF8 and chars < 256 before converting string */ |
f9a63242 JH |
611 | for (send = s + *len; s < send;) { |
612 | U8 c = *s++; | |
613 | if (!UTF8_IS_ASCII(c)) { | |
614 | if (UTF8_IS_CONTINUATION(c) || s >= send || | |
df84a23b | 615 | !UTF8_IS_CONTINUATION(*s) || UTF8_IS_DOWNGRADEABLE_START(c)) |
f9a63242 JH |
616 | return start; |
617 | s++, count++; | |
618 | } | |
619 | } | |
620 | ||
621 | *is_utf8 = 0; | |
622 | ||
623 | if (!count) | |
624 | return start; | |
625 | ||
626 | Newz(801, d, (*len) - count + 1, U8); | |
ef9edfd0 | 627 | s = start; start = d; |
f9a63242 JH |
628 | while (s < send) { |
629 | U8 c = *s++; | |
90f44359 | 630 | |
f9a63242 JH |
631 | if (UTF8_IS_ASCII(c)) |
632 | *d++ = c; | |
633 | else | |
90f44359 | 634 | *d++ = UTF8_ACCUMULATE(c, *s++); |
f9a63242 JH |
635 | } |
636 | *d = '\0'; | |
637 | *len = d - start; | |
638 | return start; | |
639 | } | |
640 | ||
641 | /* | |
eebe1485 | 642 | =for apidoc A|U8 *|bytes_to_utf8|U8 *s|STRLEN *len |
6940069f GS |
643 | |
644 | Converts a string C<s> of length C<len> from ASCII into UTF8 encoding. | |
6662521e GS |
645 | Returns a pointer to the newly-created string, and sets C<len> to |
646 | reflect the new length. | |
6940069f | 647 | |
497711e7 | 648 | =cut |
6940069f GS |
649 | */ |
650 | ||
651 | U8* | |
6662521e | 652 | Perl_bytes_to_utf8(pTHX_ U8* s, STRLEN *len) |
6940069f | 653 | { |
6940069f GS |
654 | U8 *send; |
655 | U8 *d; | |
656 | U8 *dst; | |
6662521e | 657 | send = s + (*len); |
6940069f | 658 | |
6662521e | 659 | Newz(801, d, (*len) * 2 + 1, U8); |
6940069f GS |
660 | dst = d; |
661 | ||
662 | while (s < send) { | |
90f44359 | 663 | if (UTF8_IS_ASCII(*s)) |
6940069f GS |
664 | *d++ = *s++; |
665 | else { | |
666 | UV uv = *s++; | |
90f44359 JH |
667 | |
668 | *d++ = UTF8_EIGHT_BIT_HI(uv); | |
669 | *d++ = UTF8_EIGHT_BIT_LO(uv); | |
6940069f GS |
670 | } |
671 | } | |
672 | *d = '\0'; | |
6662521e | 673 | *len = d-dst; |
6940069f GS |
674 | return dst; |
675 | } | |
676 | ||
a0ed51b3 | 677 | /* |
dea0fc0b | 678 | * Convert native (big-endian) or reversed (little-endian) UTF-16 to UTF-8. |
a0ed51b3 LW |
679 | * |
680 | * Destination must be pre-extended to 3/2 source. Do not use in-place. | |
681 | * We optimize for native, for obvious reasons. */ | |
682 | ||
683 | U8* | |
dea0fc0b | 684 | Perl_utf16_to_utf8(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen) |
a0ed51b3 | 685 | { |
dea0fc0b JH |
686 | U8* pend; |
687 | U8* dstart = d; | |
688 | ||
689 | if (bytelen & 1) | |
a7867d0a | 690 | Perl_croak(aTHX_ "panic: utf16_to_utf8: odd bytelen"); |
dea0fc0b JH |
691 | |
692 | pend = p + bytelen; | |
693 | ||
a0ed51b3 | 694 | while (p < pend) { |
dea0fc0b JH |
695 | UV uv = (p[0] << 8) + p[1]; /* UTF-16BE */ |
696 | p += 2; | |
a0ed51b3 LW |
697 | if (uv < 0x80) { |
698 | *d++ = uv; | |
699 | continue; | |
700 | } | |
701 | if (uv < 0x800) { | |
702 | *d++ = (( uv >> 6) | 0xc0); | |
703 | *d++ = (( uv & 0x3f) | 0x80); | |
704 | continue; | |
705 | } | |
706 | if (uv >= 0xd800 && uv < 0xdbff) { /* surrogates */ | |
dea0fc0b JH |
707 | UV low = *p++; |
708 | if (low < 0xdc00 || low >= 0xdfff) | |
709 | Perl_croak(aTHX_ "Malformed UTF-16 surrogate"); | |
a0ed51b3 LW |
710 | uv = ((uv - 0xd800) << 10) + (low - 0xdc00) + 0x10000; |
711 | } | |
712 | if (uv < 0x10000) { | |
713 | *d++ = (( uv >> 12) | 0xe0); | |
714 | *d++ = (((uv >> 6) & 0x3f) | 0x80); | |
715 | *d++ = (( uv & 0x3f) | 0x80); | |
716 | continue; | |
717 | } | |
718 | else { | |
719 | *d++ = (( uv >> 18) | 0xf0); | |
720 | *d++ = (((uv >> 12) & 0x3f) | 0x80); | |
721 | *d++ = (((uv >> 6) & 0x3f) | 0x80); | |
722 | *d++ = (( uv & 0x3f) | 0x80); | |
723 | continue; | |
724 | } | |
725 | } | |
dea0fc0b | 726 | *newlen = d - dstart; |
a0ed51b3 LW |
727 | return d; |
728 | } | |
729 | ||
730 | /* Note: this one is slightly destructive of the source. */ | |
731 | ||
732 | U8* | |
dea0fc0b | 733 | Perl_utf16_to_utf8_reversed(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen) |
a0ed51b3 LW |
734 | { |
735 | U8* s = (U8*)p; | |
736 | U8* send = s + bytelen; | |
737 | while (s < send) { | |
738 | U8 tmp = s[0]; | |
739 | s[0] = s[1]; | |
740 | s[1] = tmp; | |
741 | s += 2; | |
742 | } | |
dea0fc0b | 743 | return utf16_to_utf8(p, d, bytelen, newlen); |
a0ed51b3 LW |
744 | } |
745 | ||
746 | /* for now these are all defined (inefficiently) in terms of the utf8 versions */ | |
747 | ||
748 | bool | |
864dbfa3 | 749 | Perl_is_uni_alnum(pTHX_ U32 c) |
a0ed51b3 | 750 | { |
ad391ad9 | 751 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 LW |
752 | uv_to_utf8(tmpbuf, (UV)c); |
753 | return is_utf8_alnum(tmpbuf); | |
754 | } | |
755 | ||
756 | bool | |
b8c5462f JH |
757 | Perl_is_uni_alnumc(pTHX_ U32 c) |
758 | { | |
ad391ad9 | 759 | U8 tmpbuf[UTF8_MAXLEN+1]; |
b8c5462f JH |
760 | uv_to_utf8(tmpbuf, (UV)c); |
761 | return is_utf8_alnumc(tmpbuf); | |
762 | } | |
763 | ||
764 | bool | |
864dbfa3 | 765 | Perl_is_uni_idfirst(pTHX_ U32 c) |
a0ed51b3 | 766 | { |
ad391ad9 | 767 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 LW |
768 | uv_to_utf8(tmpbuf, (UV)c); |
769 | return is_utf8_idfirst(tmpbuf); | |
770 | } | |
771 | ||
772 | bool | |
864dbfa3 | 773 | Perl_is_uni_alpha(pTHX_ U32 c) |
a0ed51b3 | 774 | { |
ad391ad9 | 775 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 LW |
776 | uv_to_utf8(tmpbuf, (UV)c); |
777 | return is_utf8_alpha(tmpbuf); | |
778 | } | |
779 | ||
780 | bool | |
4d61ec05 GS |
781 | Perl_is_uni_ascii(pTHX_ U32 c) |
782 | { | |
ad391ad9 | 783 | U8 tmpbuf[UTF8_MAXLEN+1]; |
4d61ec05 GS |
784 | uv_to_utf8(tmpbuf, (UV)c); |
785 | return is_utf8_ascii(tmpbuf); | |
786 | } | |
787 | ||
788 | bool | |
864dbfa3 | 789 | Perl_is_uni_space(pTHX_ U32 c) |
a0ed51b3 | 790 | { |
ad391ad9 | 791 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 LW |
792 | uv_to_utf8(tmpbuf, (UV)c); |
793 | return is_utf8_space(tmpbuf); | |
794 | } | |
795 | ||
796 | bool | |
864dbfa3 | 797 | Perl_is_uni_digit(pTHX_ U32 c) |
a0ed51b3 | 798 | { |
ad391ad9 | 799 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 LW |
800 | uv_to_utf8(tmpbuf, (UV)c); |
801 | return is_utf8_digit(tmpbuf); | |
802 | } | |
803 | ||
804 | bool | |
864dbfa3 | 805 | Perl_is_uni_upper(pTHX_ U32 c) |
a0ed51b3 | 806 | { |
ad391ad9 | 807 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 LW |
808 | uv_to_utf8(tmpbuf, (UV)c); |
809 | return is_utf8_upper(tmpbuf); | |
810 | } | |
811 | ||
812 | bool | |
864dbfa3 | 813 | Perl_is_uni_lower(pTHX_ U32 c) |
a0ed51b3 | 814 | { |
ad391ad9 | 815 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 LW |
816 | uv_to_utf8(tmpbuf, (UV)c); |
817 | return is_utf8_lower(tmpbuf); | |
818 | } | |
819 | ||
820 | bool | |
b8c5462f JH |
821 | Perl_is_uni_cntrl(pTHX_ U32 c) |
822 | { | |
ad391ad9 | 823 | U8 tmpbuf[UTF8_MAXLEN+1]; |
b8c5462f JH |
824 | uv_to_utf8(tmpbuf, (UV)c); |
825 | return is_utf8_cntrl(tmpbuf); | |
826 | } | |
827 | ||
828 | bool | |
829 | Perl_is_uni_graph(pTHX_ U32 c) | |
830 | { | |
ad391ad9 | 831 | U8 tmpbuf[UTF8_MAXLEN+1]; |
b8c5462f JH |
832 | uv_to_utf8(tmpbuf, (UV)c); |
833 | return is_utf8_graph(tmpbuf); | |
834 | } | |
835 | ||
836 | bool | |
864dbfa3 | 837 | Perl_is_uni_print(pTHX_ U32 c) |
a0ed51b3 | 838 | { |
ad391ad9 | 839 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 LW |
840 | uv_to_utf8(tmpbuf, (UV)c); |
841 | return is_utf8_print(tmpbuf); | |
842 | } | |
843 | ||
b8c5462f | 844 | bool |
f248d071 | 845 | Perl_is_uni_punct(pTHX_ U32 c) |
b8c5462f | 846 | { |
ad391ad9 | 847 | U8 tmpbuf[UTF8_MAXLEN+1]; |
b8c5462f JH |
848 | uv_to_utf8(tmpbuf, (UV)c); |
849 | return is_utf8_punct(tmpbuf); | |
850 | } | |
851 | ||
4d61ec05 GS |
852 | bool |
853 | Perl_is_uni_xdigit(pTHX_ U32 c) | |
854 | { | |
ad391ad9 | 855 | U8 tmpbuf[UTF8_MAXLEN+1]; |
4d61ec05 GS |
856 | uv_to_utf8(tmpbuf, (UV)c); |
857 | return is_utf8_xdigit(tmpbuf); | |
858 | } | |
859 | ||
a0ed51b3 | 860 | U32 |
864dbfa3 | 861 | Perl_to_uni_upper(pTHX_ U32 c) |
a0ed51b3 | 862 | { |
ad391ad9 | 863 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 LW |
864 | uv_to_utf8(tmpbuf, (UV)c); |
865 | return to_utf8_upper(tmpbuf); | |
866 | } | |
867 | ||
868 | U32 | |
864dbfa3 | 869 | Perl_to_uni_title(pTHX_ U32 c) |
a0ed51b3 | 870 | { |
ad391ad9 | 871 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 LW |
872 | uv_to_utf8(tmpbuf, (UV)c); |
873 | return to_utf8_title(tmpbuf); | |
874 | } | |
875 | ||
876 | U32 | |
864dbfa3 | 877 | Perl_to_uni_lower(pTHX_ U32 c) |
a0ed51b3 | 878 | { |
ad391ad9 | 879 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 LW |
880 | uv_to_utf8(tmpbuf, (UV)c); |
881 | return to_utf8_lower(tmpbuf); | |
882 | } | |
883 | ||
884 | /* for now these all assume no locale info available for Unicode > 255 */ | |
885 | ||
886 | bool | |
864dbfa3 | 887 | Perl_is_uni_alnum_lc(pTHX_ U32 c) |
a0ed51b3 LW |
888 | { |
889 | return is_uni_alnum(c); /* XXX no locale support yet */ | |
890 | } | |
891 | ||
892 | bool | |
b8c5462f JH |
893 | Perl_is_uni_alnumc_lc(pTHX_ U32 c) |
894 | { | |
895 | return is_uni_alnumc(c); /* XXX no locale support yet */ | |
896 | } | |
897 | ||
898 | bool | |
864dbfa3 | 899 | Perl_is_uni_idfirst_lc(pTHX_ U32 c) |
a0ed51b3 LW |
900 | { |
901 | return is_uni_idfirst(c); /* XXX no locale support yet */ | |
902 | } | |
903 | ||
904 | bool | |
864dbfa3 | 905 | Perl_is_uni_alpha_lc(pTHX_ U32 c) |
a0ed51b3 LW |
906 | { |
907 | return is_uni_alpha(c); /* XXX no locale support yet */ | |
908 | } | |
909 | ||
910 | bool | |
4d61ec05 GS |
911 | Perl_is_uni_ascii_lc(pTHX_ U32 c) |
912 | { | |
913 | return is_uni_ascii(c); /* XXX no locale support yet */ | |
914 | } | |
915 | ||
916 | bool | |
864dbfa3 | 917 | Perl_is_uni_space_lc(pTHX_ U32 c) |
a0ed51b3 LW |
918 | { |
919 | return is_uni_space(c); /* XXX no locale support yet */ | |
920 | } | |
921 | ||
922 | bool | |
864dbfa3 | 923 | Perl_is_uni_digit_lc(pTHX_ U32 c) |
a0ed51b3 LW |
924 | { |
925 | return is_uni_digit(c); /* XXX no locale support yet */ | |
926 | } | |
927 | ||
928 | bool | |
864dbfa3 | 929 | Perl_is_uni_upper_lc(pTHX_ U32 c) |
a0ed51b3 LW |
930 | { |
931 | return is_uni_upper(c); /* XXX no locale support yet */ | |
932 | } | |
933 | ||
934 | bool | |
864dbfa3 | 935 | Perl_is_uni_lower_lc(pTHX_ U32 c) |
a0ed51b3 LW |
936 | { |
937 | return is_uni_lower(c); /* XXX no locale support yet */ | |
938 | } | |
939 | ||
940 | bool | |
b8c5462f JH |
941 | Perl_is_uni_cntrl_lc(pTHX_ U32 c) |
942 | { | |
943 | return is_uni_cntrl(c); /* XXX no locale support yet */ | |
944 | } | |
945 | ||
946 | bool | |
947 | Perl_is_uni_graph_lc(pTHX_ U32 c) | |
948 | { | |
949 | return is_uni_graph(c); /* XXX no locale support yet */ | |
950 | } | |
951 | ||
952 | bool | |
864dbfa3 | 953 | Perl_is_uni_print_lc(pTHX_ U32 c) |
a0ed51b3 LW |
954 | { |
955 | return is_uni_print(c); /* XXX no locale support yet */ | |
956 | } | |
957 | ||
b8c5462f JH |
958 | bool |
959 | Perl_is_uni_punct_lc(pTHX_ U32 c) | |
960 | { | |
961 | return is_uni_punct(c); /* XXX no locale support yet */ | |
962 | } | |
963 | ||
4d61ec05 GS |
964 | bool |
965 | Perl_is_uni_xdigit_lc(pTHX_ U32 c) | |
966 | { | |
967 | return is_uni_xdigit(c); /* XXX no locale support yet */ | |
968 | } | |
969 | ||
a0ed51b3 | 970 | U32 |
864dbfa3 | 971 | Perl_to_uni_upper_lc(pTHX_ U32 c) |
a0ed51b3 LW |
972 | { |
973 | return to_uni_upper(c); /* XXX no locale support yet */ | |
974 | } | |
975 | ||
976 | U32 | |
864dbfa3 | 977 | Perl_to_uni_title_lc(pTHX_ U32 c) |
a0ed51b3 LW |
978 | { |
979 | return to_uni_title(c); /* XXX no locale support yet */ | |
980 | } | |
981 | ||
982 | U32 | |
864dbfa3 | 983 | Perl_to_uni_lower_lc(pTHX_ U32 c) |
a0ed51b3 LW |
984 | { |
985 | return to_uni_lower(c); /* XXX no locale support yet */ | |
986 | } | |
987 | ||
a0ed51b3 | 988 | bool |
864dbfa3 | 989 | Perl_is_utf8_alnum(pTHX_ U8 *p) |
a0ed51b3 | 990 | { |
386d01d6 GS |
991 | if (!is_utf8_char(p)) |
992 | return FALSE; | |
a0ed51b3 | 993 | if (!PL_utf8_alnum) |
289d4f09 ML |
994 | /* NOTE: "IsWord", not "IsAlnum", since Alnum is a true |
995 | * descendant of isalnum(3), in other words, it doesn't | |
996 | * contain the '_'. --jhi */ | |
997 | PL_utf8_alnum = swash_init("utf8", "IsWord", &PL_sv_undef, 0, 0); | |
a0ed51b3 LW |
998 | return swash_fetch(PL_utf8_alnum, p); |
999 | /* return *p == '_' || is_utf8_alpha(p) || is_utf8_digit(p); */ | |
1000 | #ifdef SURPRISINGLY_SLOWER /* probably because alpha is usually true */ | |
1001 | if (!PL_utf8_alnum) | |
1002 | PL_utf8_alnum = swash_init("utf8", "", | |
1003 | sv_2mortal(newSVpv("+utf8::IsAlpha\n+utf8::IsDigit\n005F\n",0)), 0, 0); | |
1004 | return swash_fetch(PL_utf8_alnum, p); | |
1005 | #endif | |
1006 | } | |
1007 | ||
1008 | bool | |
b8c5462f JH |
1009 | Perl_is_utf8_alnumc(pTHX_ U8 *p) |
1010 | { | |
386d01d6 GS |
1011 | if (!is_utf8_char(p)) |
1012 | return FALSE; | |
b8c5462f JH |
1013 | if (!PL_utf8_alnum) |
1014 | PL_utf8_alnum = swash_init("utf8", "IsAlnumC", &PL_sv_undef, 0, 0); | |
1015 | return swash_fetch(PL_utf8_alnum, p); | |
1016 | /* return is_utf8_alpha(p) || is_utf8_digit(p); */ | |
1017 | #ifdef SURPRISINGLY_SLOWER /* probably because alpha is usually true */ | |
1018 | if (!PL_utf8_alnum) | |
1019 | PL_utf8_alnum = swash_init("utf8", "", | |
1020 | sv_2mortal(newSVpv("+utf8::IsAlpha\n+utf8::IsDigit\n005F\n",0)), 0, 0); | |
1021 | return swash_fetch(PL_utf8_alnum, p); | |
1022 | #endif | |
1023 | } | |
1024 | ||
1025 | bool | |
864dbfa3 | 1026 | Perl_is_utf8_idfirst(pTHX_ U8 *p) |
a0ed51b3 LW |
1027 | { |
1028 | return *p == '_' || is_utf8_alpha(p); | |
1029 | } | |
1030 | ||
1031 | bool | |
864dbfa3 | 1032 | Perl_is_utf8_alpha(pTHX_ U8 *p) |
a0ed51b3 | 1033 | { |
386d01d6 GS |
1034 | if (!is_utf8_char(p)) |
1035 | return FALSE; | |
a0ed51b3 | 1036 | if (!PL_utf8_alpha) |
e24b16f9 | 1037 | PL_utf8_alpha = swash_init("utf8", "IsAlpha", &PL_sv_undef, 0, 0); |
a0ed51b3 LW |
1038 | return swash_fetch(PL_utf8_alpha, p); |
1039 | } | |
1040 | ||
1041 | bool | |
b8c5462f JH |
1042 | Perl_is_utf8_ascii(pTHX_ U8 *p) |
1043 | { | |
386d01d6 GS |
1044 | if (!is_utf8_char(p)) |
1045 | return FALSE; | |
b8c5462f JH |
1046 | if (!PL_utf8_ascii) |
1047 | PL_utf8_ascii = swash_init("utf8", "IsAscii", &PL_sv_undef, 0, 0); | |
1048 | return swash_fetch(PL_utf8_ascii, p); | |
1049 | } | |
1050 | ||
1051 | bool | |
864dbfa3 | 1052 | Perl_is_utf8_space(pTHX_ U8 *p) |
a0ed51b3 | 1053 | { |
386d01d6 GS |
1054 | if (!is_utf8_char(p)) |
1055 | return FALSE; | |
a0ed51b3 | 1056 | if (!PL_utf8_space) |
3bec3564 | 1057 | PL_utf8_space = swash_init("utf8", "IsSpacePerl", &PL_sv_undef, 0, 0); |
a0ed51b3 LW |
1058 | return swash_fetch(PL_utf8_space, p); |
1059 | } | |
1060 | ||
1061 | bool | |
864dbfa3 | 1062 | Perl_is_utf8_digit(pTHX_ U8 *p) |
a0ed51b3 | 1063 | { |
386d01d6 GS |
1064 | if (!is_utf8_char(p)) |
1065 | return FALSE; | |
a0ed51b3 | 1066 | if (!PL_utf8_digit) |
e24b16f9 | 1067 | PL_utf8_digit = swash_init("utf8", "IsDigit", &PL_sv_undef, 0, 0); |
a0ed51b3 LW |
1068 | return swash_fetch(PL_utf8_digit, p); |
1069 | } | |
1070 | ||
1071 | bool | |
864dbfa3 | 1072 | Perl_is_utf8_upper(pTHX_ U8 *p) |
a0ed51b3 | 1073 | { |
386d01d6 GS |
1074 | if (!is_utf8_char(p)) |
1075 | return FALSE; | |
a0ed51b3 | 1076 | if (!PL_utf8_upper) |
e24b16f9 | 1077 | PL_utf8_upper = swash_init("utf8", "IsUpper", &PL_sv_undef, 0, 0); |
a0ed51b3 LW |
1078 | return swash_fetch(PL_utf8_upper, p); |
1079 | } | |
1080 | ||
1081 | bool | |
864dbfa3 | 1082 | Perl_is_utf8_lower(pTHX_ U8 *p) |
a0ed51b3 | 1083 | { |
386d01d6 GS |
1084 | if (!is_utf8_char(p)) |
1085 | return FALSE; | |
a0ed51b3 | 1086 | if (!PL_utf8_lower) |
e24b16f9 | 1087 | PL_utf8_lower = swash_init("utf8", "IsLower", &PL_sv_undef, 0, 0); |
a0ed51b3 LW |
1088 | return swash_fetch(PL_utf8_lower, p); |
1089 | } | |
1090 | ||
1091 | bool | |
b8c5462f JH |
1092 | Perl_is_utf8_cntrl(pTHX_ U8 *p) |
1093 | { | |
386d01d6 GS |
1094 | if (!is_utf8_char(p)) |
1095 | return FALSE; | |
b8c5462f JH |
1096 | if (!PL_utf8_cntrl) |
1097 | PL_utf8_cntrl = swash_init("utf8", "IsCntrl", &PL_sv_undef, 0, 0); | |
1098 | return swash_fetch(PL_utf8_cntrl, p); | |
1099 | } | |
1100 | ||
1101 | bool | |
1102 | Perl_is_utf8_graph(pTHX_ U8 *p) | |
1103 | { | |
386d01d6 GS |
1104 | if (!is_utf8_char(p)) |
1105 | return FALSE; | |
b8c5462f JH |
1106 | if (!PL_utf8_graph) |
1107 | PL_utf8_graph = swash_init("utf8", "IsGraph", &PL_sv_undef, 0, 0); | |
1108 | return swash_fetch(PL_utf8_graph, p); | |
1109 | } | |
1110 | ||
1111 | bool | |
864dbfa3 | 1112 | Perl_is_utf8_print(pTHX_ U8 *p) |
a0ed51b3 | 1113 | { |
386d01d6 GS |
1114 | if (!is_utf8_char(p)) |
1115 | return FALSE; | |
a0ed51b3 | 1116 | if (!PL_utf8_print) |
e24b16f9 | 1117 | PL_utf8_print = swash_init("utf8", "IsPrint", &PL_sv_undef, 0, 0); |
a0ed51b3 LW |
1118 | return swash_fetch(PL_utf8_print, p); |
1119 | } | |
1120 | ||
1121 | bool | |
b8c5462f JH |
1122 | Perl_is_utf8_punct(pTHX_ U8 *p) |
1123 | { | |
386d01d6 GS |
1124 | if (!is_utf8_char(p)) |
1125 | return FALSE; | |
b8c5462f JH |
1126 | if (!PL_utf8_punct) |
1127 | PL_utf8_punct = swash_init("utf8", "IsPunct", &PL_sv_undef, 0, 0); | |
1128 | return swash_fetch(PL_utf8_punct, p); | |
1129 | } | |
1130 | ||
1131 | bool | |
1132 | Perl_is_utf8_xdigit(pTHX_ U8 *p) | |
1133 | { | |
386d01d6 GS |
1134 | if (!is_utf8_char(p)) |
1135 | return FALSE; | |
b8c5462f JH |
1136 | if (!PL_utf8_xdigit) |
1137 | PL_utf8_xdigit = swash_init("utf8", "IsXDigit", &PL_sv_undef, 0, 0); | |
1138 | return swash_fetch(PL_utf8_xdigit, p); | |
1139 | } | |
1140 | ||
1141 | bool | |
864dbfa3 | 1142 | Perl_is_utf8_mark(pTHX_ U8 *p) |
a0ed51b3 | 1143 | { |
386d01d6 GS |
1144 | if (!is_utf8_char(p)) |
1145 | return FALSE; | |
a0ed51b3 | 1146 | if (!PL_utf8_mark) |
e24b16f9 | 1147 | PL_utf8_mark = swash_init("utf8", "IsM", &PL_sv_undef, 0, 0); |
a0ed51b3 LW |
1148 | return swash_fetch(PL_utf8_mark, p); |
1149 | } | |
1150 | ||
2104c8d9 | 1151 | UV |
864dbfa3 | 1152 | Perl_to_utf8_upper(pTHX_ U8 *p) |
a0ed51b3 LW |
1153 | { |
1154 | UV uv; | |
1155 | ||
1156 | if (!PL_utf8_toupper) | |
e24b16f9 | 1157 | PL_utf8_toupper = swash_init("utf8", "ToUpper", &PL_sv_undef, 4, 0); |
a0ed51b3 | 1158 | uv = swash_fetch(PL_utf8_toupper, p); |
756820e3 | 1159 | return uv ? uv : utf8_to_uv(p,UTF8_MAXLEN,0,0); |
a0ed51b3 LW |
1160 | } |
1161 | ||
2104c8d9 | 1162 | UV |
864dbfa3 | 1163 | Perl_to_utf8_title(pTHX_ U8 *p) |
a0ed51b3 LW |
1164 | { |
1165 | UV uv; | |
1166 | ||
1167 | if (!PL_utf8_totitle) | |
e24b16f9 | 1168 | PL_utf8_totitle = swash_init("utf8", "ToTitle", &PL_sv_undef, 4, 0); |
a0ed51b3 | 1169 | uv = swash_fetch(PL_utf8_totitle, p); |
756820e3 | 1170 | return uv ? uv : utf8_to_uv(p,UTF8_MAXLEN,0,0); |
a0ed51b3 LW |
1171 | } |
1172 | ||
2104c8d9 | 1173 | UV |
864dbfa3 | 1174 | Perl_to_utf8_lower(pTHX_ U8 *p) |
a0ed51b3 LW |
1175 | { |
1176 | UV uv; | |
1177 | ||
1178 | if (!PL_utf8_tolower) | |
e24b16f9 | 1179 | PL_utf8_tolower = swash_init("utf8", "ToLower", &PL_sv_undef, 4, 0); |
a0ed51b3 | 1180 | uv = swash_fetch(PL_utf8_tolower, p); |
756820e3 | 1181 | return uv ? uv : utf8_to_uv(p,UTF8_MAXLEN,0,0); |
a0ed51b3 LW |
1182 | } |
1183 | ||
1184 | /* a "swash" is a swatch hash */ | |
1185 | ||
1186 | SV* | |
864dbfa3 | 1187 | Perl_swash_init(pTHX_ char* pkg, char* name, SV *listsv, I32 minbits, I32 none) |
a0ed51b3 LW |
1188 | { |
1189 | SV* retval; | |
bf1fed83 | 1190 | SV* tokenbufsv = sv_2mortal(NEWSV(0,0)); |
8e84507e | 1191 | dSP; |
ce3b816e GS |
1192 | |
1193 | if (!gv_stashpv(pkg, 0)) { /* demand load utf8 */ | |
1194 | ENTER; | |
1195 | Perl_load_module(aTHX_ PERL_LOADMOD_NOIMPORT, newSVpv(pkg,0), Nullsv); | |
1196 | LEAVE; | |
1197 | } | |
1198 | SPAGAIN; | |
a0ed51b3 LW |
1199 | PUSHSTACKi(PERLSI_MAGIC); |
1200 | PUSHMARK(SP); | |
1201 | EXTEND(SP,5); | |
1202 | PUSHs(sv_2mortal(newSVpvn(pkg, strlen(pkg)))); | |
1203 | PUSHs(sv_2mortal(newSVpvn(name, strlen(name)))); | |
1204 | PUSHs(listsv); | |
1205 | PUSHs(sv_2mortal(newSViv(minbits))); | |
1206 | PUSHs(sv_2mortal(newSViv(none))); | |
1207 | PUTBACK; | |
1208 | ENTER; | |
1209 | SAVEI32(PL_hints); | |
1210 | PL_hints = 0; | |
1211 | save_re_context(); | |
bf1fed83 JH |
1212 | if (PL_curcop == &PL_compiling) |
1213 | /* XXX ought to be handled by lex_start */ | |
1214 | sv_setpv(tokenbufsv, PL_tokenbuf); | |
864dbfa3 | 1215 | if (call_method("SWASHNEW", G_SCALAR)) |
8e84507e | 1216 | retval = newSVsv(*PL_stack_sp--); |
a0ed51b3 | 1217 | else |
e24b16f9 | 1218 | retval = &PL_sv_undef; |
a0ed51b3 LW |
1219 | LEAVE; |
1220 | POPSTACK; | |
e24b16f9 | 1221 | if (PL_curcop == &PL_compiling) { |
bf1fed83 JH |
1222 | STRLEN len; |
1223 | char* pv = SvPV(tokenbufsv, len); | |
1224 | ||
1225 | Copy(pv, PL_tokenbuf, len+1, char); | |
e24b16f9 | 1226 | PL_curcop->op_private = PL_hints; |
a0ed51b3 LW |
1227 | } |
1228 | if (!SvROK(retval) || SvTYPE(SvRV(retval)) != SVt_PVHV) | |
cea2e8a9 | 1229 | Perl_croak(aTHX_ "SWASHNEW didn't return an HV ref"); |
a0ed51b3 LW |
1230 | return retval; |
1231 | } | |
1232 | ||
1233 | UV | |
864dbfa3 | 1234 | Perl_swash_fetch(pTHX_ SV *sv, U8 *ptr) |
a0ed51b3 LW |
1235 | { |
1236 | HV* hv = (HV*)SvRV(sv); | |
1237 | U32 klen = UTF8SKIP(ptr) - 1; | |
1238 | U32 off = ptr[klen] & 127; /* NB: 64 bit always 0 when len > 1 */ | |
1239 | STRLEN slen; | |
1240 | STRLEN needents = (klen ? 64 : 128); | |
dfe13c55 | 1241 | U8 *tmps; |
a0ed51b3 LW |
1242 | U32 bit; |
1243 | SV *retval; | |
1244 | ||
1245 | /* | |
1246 | * This single-entry cache saves about 1/3 of the utf8 overhead in test | |
1247 | * suite. (That is, only 7-8% overall over just a hash cache. Still, | |
1248 | * it's nothing to sniff at.) Pity we usually come through at least | |
1249 | * two function calls to get here... | |
1250 | * | |
1251 | * NB: this code assumes that swatches are never modified, once generated! | |
1252 | */ | |
1253 | ||
1254 | if (hv == PL_last_swash_hv && | |
1255 | klen == PL_last_swash_klen && | |
12ae5dfc | 1256 | (!klen || memEQ((char *)ptr,(char *)PL_last_swash_key,klen)) ) |
a0ed51b3 LW |
1257 | { |
1258 | tmps = PL_last_swash_tmps; | |
1259 | slen = PL_last_swash_slen; | |
1260 | } | |
1261 | else { | |
1262 | /* Try our second-level swatch cache, kept in a hash. */ | |
dfe13c55 | 1263 | SV** svp = hv_fetch(hv, (char*)ptr, klen, FALSE); |
a0ed51b3 LW |
1264 | |
1265 | /* If not cached, generate it via utf8::SWASHGET */ | |
dfe13c55 | 1266 | if (!svp || !SvPOK(*svp) || !(tmps = (U8*)SvPV(*svp, slen))) { |
a0ed51b3 LW |
1267 | dSP; |
1268 | ENTER; | |
1269 | SAVETMPS; | |
1270 | save_re_context(); | |
1271 | PUSHSTACKi(PERLSI_MAGIC); | |
1272 | PUSHMARK(SP); | |
1273 | EXTEND(SP,3); | |
1274 | PUSHs((SV*)sv); | |
756820e3 | 1275 | PUSHs(sv_2mortal(newSViv(utf8_to_uv(ptr, UTF8_MAXLEN, 0, 0) & ~(needents - 1)))); |
a0ed51b3 LW |
1276 | PUSHs(sv_2mortal(newSViv(needents))); |
1277 | PUTBACK; | |
864dbfa3 | 1278 | if (call_method("SWASHGET", G_SCALAR)) |
8e84507e | 1279 | retval = newSVsv(*PL_stack_sp--); |
a0ed51b3 | 1280 | else |
e24b16f9 | 1281 | retval = &PL_sv_undef; |
a0ed51b3 LW |
1282 | POPSTACK; |
1283 | FREETMPS; | |
1284 | LEAVE; | |
e24b16f9 GS |
1285 | if (PL_curcop == &PL_compiling) |
1286 | PL_curcop->op_private = PL_hints; | |
a0ed51b3 | 1287 | |
dfe13c55 | 1288 | svp = hv_store(hv, (char*)ptr, klen, retval, 0); |
a0ed51b3 | 1289 | |
dfe13c55 | 1290 | if (!svp || !(tmps = (U8*)SvPV(*svp, slen)) || slen < 8) |
cea2e8a9 | 1291 | Perl_croak(aTHX_ "SWASHGET didn't return result of proper length"); |
a0ed51b3 LW |
1292 | } |
1293 | ||
1294 | PL_last_swash_hv = hv; | |
1295 | PL_last_swash_klen = klen; | |
1296 | PL_last_swash_tmps = tmps; | |
1297 | PL_last_swash_slen = slen; | |
1298 | if (klen) | |
1299 | Copy(ptr, PL_last_swash_key, klen, U8); | |
1300 | } | |
1301 | ||
9faf8d75 | 1302 | switch ((int)((slen << 3) / needents)) { |
a0ed51b3 LW |
1303 | case 1: |
1304 | bit = 1 << (off & 7); | |
1305 | off >>= 3; | |
1306 | return (tmps[off] & bit) != 0; | |
1307 | case 8: | |
1308 | return tmps[off]; | |
1309 | case 16: | |
1310 | off <<= 1; | |
1311 | return (tmps[off] << 8) + tmps[off + 1] ; | |
1312 | case 32: | |
1313 | off <<= 2; | |
1314 | return (tmps[off] << 24) + (tmps[off+1] << 16) + (tmps[off+2] << 8) + tmps[off + 3] ; | |
1315 | } | |
cea2e8a9 | 1316 | Perl_croak(aTHX_ "panic: swash_fetch"); |
a0ed51b3 LW |
1317 | return 0; |
1318 | } |