This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Forgot the latest mktables.PL from #9899.
[perl5.git] / utf8.c
CommitLineData
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 29/*
37607a96 30=for apidoc A|U8 *|uvuni_to_utf8|U8 *d|UV uv
eebe1485
SC
31
32Adds the UTF8 representation of the Unicode codepoint C<uv> to the end
33of the string C<d>; C<d> should be have at least C<UTF8_MAXLEN+1> free
34bytes available. The return value is the pointer to the byte after the
9041c2e3 35end of the new character. In other words,
eebe1485 36
9041c2e3 37 d = uvuni_to_utf8(d, uv);
eebe1485
SC
38
39is the recommended Unicode-aware way of saying
40
41 *(d++) = uv;
42
43=cut
44*/
45
dfe13c55 46U8 *
9041c2e3 47Perl_uvuni_to_utf8(pTHX_ U8 *d, UV uv)
a0ed51b3 48{
c4d5f83a
NIS
49 if (UNI_IS_INVARIANT(uv)) {
50 *d++ = UTF_TO_NATIVE(uv);
a0ed51b3
LW
51 return d;
52 }
1d72bdf6
NIS
53#if defined(EBCDIC) || 1 /* always for testing */
54 else {
55 STRLEN len = UNISKIP(uv);
56 U8 *p = d+len-1;
57 while (p > d) {
58 *p-- = UTF_TO_NATIVE((uv & UTF_CONTINUATION_MASK) | UTF_CONTINUATION_MARK);
59 uv >>= UTF_ACCUMULATION_SHIFT;
60 }
61 *p = UTF_TO_NATIVE((uv & UTF_START_MASK(len)) | UTF_START_MARK(len));
62 return d+len;
63 }
64#else /* Non loop style */
a0ed51b3
LW
65 if (uv < 0x800) {
66 *d++ = (( uv >> 6) | 0xc0);
67 *d++ = (( uv & 0x3f) | 0x80);
68 return d;
69 }
70 if (uv < 0x10000) {
71 *d++ = (( uv >> 12) | 0xe0);
72 *d++ = (((uv >> 6) & 0x3f) | 0x80);
73 *d++ = (( uv & 0x3f) | 0x80);
74 return d;
75 }
76 if (uv < 0x200000) {
77 *d++ = (( uv >> 18) | 0xf0);
78 *d++ = (((uv >> 12) & 0x3f) | 0x80);
79 *d++ = (((uv >> 6) & 0x3f) | 0x80);
80 *d++ = (( uv & 0x3f) | 0x80);
81 return d;
82 }
83 if (uv < 0x4000000) {
84 *d++ = (( uv >> 24) | 0xf8);
85 *d++ = (((uv >> 18) & 0x3f) | 0x80);
86 *d++ = (((uv >> 12) & 0x3f) | 0x80);
87 *d++ = (((uv >> 6) & 0x3f) | 0x80);
88 *d++ = (( uv & 0x3f) | 0x80);
89 return d;
90 }
91 if (uv < 0x80000000) {
92 *d++ = (( uv >> 30) | 0xfc);
93 *d++ = (((uv >> 24) & 0x3f) | 0x80);
94 *d++ = (((uv >> 18) & 0x3f) | 0x80);
95 *d++ = (((uv >> 12) & 0x3f) | 0x80);
96 *d++ = (((uv >> 6) & 0x3f) | 0x80);
97 *d++ = (( uv & 0x3f) | 0x80);
98 return d;
99 }
6b8eaf93 100#ifdef HAS_QUAD
d7578b48 101 if (uv < UTF8_QUAD_MAX)
a0ed51b3
LW
102#endif
103 {
104 *d++ = 0xfe; /* Can't match U+FEFF! */
105 *d++ = (((uv >> 30) & 0x3f) | 0x80);
106 *d++ = (((uv >> 24) & 0x3f) | 0x80);
107 *d++ = (((uv >> 18) & 0x3f) | 0x80);
108 *d++ = (((uv >> 12) & 0x3f) | 0x80);
109 *d++ = (((uv >> 6) & 0x3f) | 0x80);
110 *d++ = (( uv & 0x3f) | 0x80);
111 return d;
112 }
6b8eaf93 113#ifdef HAS_QUAD
a0ed51b3
LW
114 {
115 *d++ = 0xff; /* Can't match U+FFFE! */
3c77ea2b
GS
116 *d++ = 0x80; /* 6 Reserved bits */
117 *d++ = (((uv >> 60) & 0x0f) | 0x80); /* 2 Reserved bits */
118 *d++ = (((uv >> 54) & 0x3f) | 0x80);
119 *d++ = (((uv >> 48) & 0x3f) | 0x80);
120 *d++ = (((uv >> 42) & 0x3f) | 0x80);
a0ed51b3
LW
121 *d++ = (((uv >> 36) & 0x3f) | 0x80);
122 *d++ = (((uv >> 30) & 0x3f) | 0x80);
123 *d++ = (((uv >> 24) & 0x3f) | 0x80);
124 *d++ = (((uv >> 18) & 0x3f) | 0x80);
125 *d++ = (((uv >> 12) & 0x3f) | 0x80);
126 *d++ = (((uv >> 6) & 0x3f) | 0x80);
127 *d++ = (( uv & 0x3f) | 0x80);
128 return d;
129 }
130#endif
1d72bdf6 131#endif /* Loop style */
a0ed51b3
LW
132}
133
9041c2e3
NIS
134
135
136/*
eebe1485
SC
137=for apidoc A|STRLEN|is_utf8_char|U8 *s
138
5da9da9e 139Tests if some arbitrary number of bytes begins in a valid UTF-8
1d72bdf6 140character. Note that an INVARIANT (i.e. ASCII) character is a valid UTF-8 character.
5da9da9e
JH
141The actual number of bytes in the UTF-8 character will be returned if
142it is valid, otherwise 0.
9041c2e3 143
37607a96
PK
144=cut
145*/
067a85ef 146STRLEN
386d01d6
GS
147Perl_is_utf8_char(pTHX_ U8 *s)
148{
149 U8 u = *s;
067a85ef
A
150 STRLEN slen, len;
151 UV uv, ouv;
386d01d6 152
1d72bdf6 153 if (UTF8_IS_INVARIANT(u))
386d01d6
GS
154 return 1;
155
60006e79 156 if (!UTF8_IS_START(u))
386d01d6
GS
157 return 0;
158
9f07fdcd 159 len = UTF8SKIP(s);
386d01d6 160
60006e79 161 if (len < 2 || !UTF8_IS_CONTINUATION(s[1]))
067a85ef
A
162 return 0;
163
386d01d6
GS
164 slen = len - 1;
165 s++;
6eb6869e 166 u &= UTF_START_MASK(len);
1d72bdf6 167 uv = u;
067a85ef 168 ouv = uv;
386d01d6 169 while (slen--) {
60006e79 170 if (!UTF8_IS_CONTINUATION(*s))
386d01d6 171 return 0;
8850bf83 172 uv = UTF8_ACCUMULATE(uv, *s);
209a9bc1 173 if (uv < ouv)
067a85ef
A
174 return 0;
175 ouv = uv;
386d01d6
GS
176 s++;
177 }
067a85ef 178
5bbb0b5a 179 if (UNISKIP(uv) < len)
067a85ef
A
180 return 0;
181
386d01d6
GS
182 return len;
183}
184
6662521e 185/*
eebe1485 186=for apidoc A|bool|is_utf8_string|U8 *s|STRLEN len
6662521e 187
5da9da9e
JH
188Returns true if first C<len> bytes of the given string form a valid UTF8
189string, false otherwise. Note that 'a valid UTF8 string' does not mean
190'a string that contains UTF8' because a valid ASCII string is a valid
191UTF8 string.
6662521e
GS
192
193=cut
194*/
195
8e84507e 196bool
6662521e
GS
197Perl_is_utf8_string(pTHX_ U8 *s, STRLEN len)
198{
067a85ef 199 U8* x = s;
1aa99e6b 200 U8* send;
067a85ef
A
201 STRLEN c;
202
1aa99e6b 203 if (!len)
6cd5fe39 204 len = strlen((char *)s);
1aa99e6b
IH
205 send = s + len;
206
6662521e
GS
207 while (x < send) {
208 c = is_utf8_char(x);
067a85ef
A
209 if (!c)
210 return FALSE;
6662521e 211 x += c;
6662521e 212 }
60006e79
JH
213 if (x != send)
214 return FALSE;
067a85ef
A
215
216 return TRUE;
6662521e
GS
217}
218
67e989fb 219/*
9041c2e3 220=for apidoc A|UV|utf8n_to_uvuni|U8 *s|STRLEN curlen|STRLEN *retlen|U32 flags
67e989fb 221
9041c2e3
NIS
222Bottom level UTF-8 decode routine.
223Returns the unicode code point value of the first character in the string C<s>
ba210ebe 224which is assumed to be in UTF8 encoding and no longer than C<curlen>;
7df053ec 225C<retlen> will be set to the length, in bytes, of that character.
67e989fb
JH
226
227If C<s> does not point to a well-formed UTF8 character, the behaviour
dcad2880
JH
228is dependent on the value of C<flags>: if it contains UTF8_CHECK_ONLY,
229it is assumed that the caller will raise a warning, and this function
28d3d195
JH
230will silently just set C<retlen> to C<-1> and return zero. If the
231C<flags> does not contain UTF8_CHECK_ONLY, warnings about
232malformations will be given, C<retlen> will be set to the expected
233length of the UTF-8 character in bytes, and zero will be returned.
234
235The C<flags> can also contain various flags to allow deviations from
236the strict UTF-8 encoding (see F<utf8.h>).
67e989fb 237
9041c2e3
NIS
238Most code should use utf8_to_uvchr() rather than call this directly.
239
37607a96
PK
240=cut
241*/
67e989fb 242
a0ed51b3 243UV
37607a96 244Perl_utf8n_to_uvuni(pTHX_ U8 *s, STRLEN curlen, STRLEN *retlen, U32 flags)
a0ed51b3 245{
ba210ebe
JH
246 UV uv = *s, ouv;
247 STRLEN len = 1;
248 bool dowarn = ckWARN_d(WARN_UTF8);
249 STRLEN expectlen = 0;
a0dbb045
JH
250 U32 warning = 0;
251
252/* This list is a superset of the UTF8_ALLOW_XXX. */
253
254#define UTF8_WARN_EMPTY 1
255#define UTF8_WARN_CONTINUATION 2
256#define UTF8_WARN_NON_CONTINUATION 3
257#define UTF8_WARN_FE_FF 4
258#define UTF8_WARN_SHORT 5
259#define UTF8_WARN_OVERFLOW 6
260#define UTF8_WARN_SURROGATE 7
261#define UTF8_WARN_BOM 8
262#define UTF8_WARN_LONG 9
263#define UTF8_WARN_FFFF 10
264
265 if (curlen == 0 &&
266 !(flags & UTF8_ALLOW_EMPTY)) {
267 warning = UTF8_WARN_EMPTY;
0c443dc2
JH
268 goto malformed;
269 }
270
1d72bdf6 271 if (UTF8_IS_INVARIANT(uv)) {
a0ed51b3
LW
272 if (retlen)
273 *retlen = 1;
c4d5f83a 274 return (UV) (NATIVE_TO_UTF(*s));
a0ed51b3 275 }
67e989fb 276
421a8bf2 277 if (UTF8_IS_CONTINUATION(uv) &&
fcc8fcf6 278 !(flags & UTF8_ALLOW_CONTINUATION)) {
a0dbb045 279 warning = UTF8_WARN_CONTINUATION;
ba210ebe
JH
280 goto malformed;
281 }
282
421a8bf2 283 if (UTF8_IS_START(uv) && curlen > 1 && !UTF8_IS_CONTINUATION(s[1]) &&
fcc8fcf6 284 !(flags & UTF8_ALLOW_NON_CONTINUATION)) {
a0dbb045 285 warning = UTF8_WARN_NON_CONTINUATION;
ba210ebe
JH
286 goto malformed;
287 }
9041c2e3 288
1d72bdf6 289#ifdef EBCDIC
75383841 290 uv = NATIVE_TO_UTF(uv);
1d72bdf6 291#else
fcc8fcf6
JH
292 if ((uv == 0xfe || uv == 0xff) &&
293 !(flags & UTF8_ALLOW_FE_FF)) {
a0dbb045 294 warning = UTF8_WARN_FE_FF;
ba210ebe 295 goto malformed;
a0ed51b3 296 }
1d72bdf6
NIS
297#endif
298
ba210ebe
JH
299 if (!(uv & 0x20)) { len = 2; uv &= 0x1f; }
300 else if (!(uv & 0x10)) { len = 3; uv &= 0x0f; }
301 else if (!(uv & 0x08)) { len = 4; uv &= 0x07; }
302 else if (!(uv & 0x04)) { len = 5; uv &= 0x03; }
1d72bdf6
NIS
303#ifdef EBCDIC
304 else if (!(uv & 0x02)) { len = 6; uv &= 0x01; }
305 else { len = 7; uv &= 0x01; }
306#else
ba210ebe
JH
307 else if (!(uv & 0x02)) { len = 6; uv &= 0x01; }
308 else if (!(uv & 0x01)) { len = 7; uv = 0; }
1d72bdf6
NIS
309 else { len = 13; uv = 0; } /* whoa! */
310#endif
311
a0ed51b3
LW
312 if (retlen)
313 *retlen = len;
9041c2e3 314
ba210ebe
JH
315 expectlen = len;
316
fcc8fcf6
JH
317 if ((curlen < expectlen) &&
318 !(flags & UTF8_ALLOW_SHORT)) {
a0dbb045 319 warning = UTF8_WARN_SHORT;
ba210ebe
JH
320 goto malformed;
321 }
322
323 len--;
a0ed51b3 324 s++;
ba210ebe
JH
325 ouv = uv;
326
a0ed51b3 327 while (len--) {
421a8bf2
JH
328 if (!UTF8_IS_CONTINUATION(*s) &&
329 !(flags & UTF8_ALLOW_NON_CONTINUATION)) {
a0dbb045
JH
330 s--;
331 warning = UTF8_WARN_NON_CONTINUATION;
ba210ebe 332 goto malformed;
a0ed51b3
LW
333 }
334 else
8850bf83 335 uv = UTF8_ACCUMULATE(uv, *s);
a0dbb045
JH
336 if (!(uv > ouv)) {
337 /* These cannot be allowed. */
338 if (uv == ouv) {
339 if (!(flags & UTF8_ALLOW_LONG)) {
340 warning = UTF8_WARN_LONG;
341 goto malformed;
342 }
343 }
344 else { /* uv < ouv */
345 /* This cannot be allowed. */
346 warning = UTF8_WARN_OVERFLOW;
347 goto malformed;
348 }
ba210ebe
JH
349 }
350 s++;
351 ouv = uv;
352 }
353
421a8bf2 354 if (UNICODE_IS_SURROGATE(uv) &&
fcc8fcf6 355 !(flags & UTF8_ALLOW_SURROGATE)) {
a0dbb045 356 warning = UTF8_WARN_SURROGATE;
ba210ebe 357 goto malformed;
421a8bf2 358 } else if (UNICODE_IS_BYTE_ORDER_MARK(uv) &&
fcc8fcf6 359 !(flags & UTF8_ALLOW_BOM)) {
a0dbb045 360 warning = UTF8_WARN_BOM;
ba210ebe 361 goto malformed;
fcc8fcf6
JH
362 } else if ((expectlen > UNISKIP(uv)) &&
363 !(flags & UTF8_ALLOW_LONG)) {
a0dbb045 364 warning = UTF8_WARN_LONG;
ba210ebe 365 goto malformed;
421a8bf2 366 } else if (UNICODE_IS_ILLEGAL(uv) &&
a9917092 367 !(flags & UTF8_ALLOW_FFFF)) {
a0dbb045 368 warning = UTF8_WARN_FFFF;
a9917092 369 goto malformed;
a0ed51b3 370 }
ba210ebe 371
a0ed51b3 372 return uv;
ba210ebe
JH
373
374malformed:
375
fcc8fcf6 376 if (flags & UTF8_CHECK_ONLY) {
ba210ebe 377 if (retlen)
cc366d4b 378 *retlen = -1;
ba210ebe
JH
379 return 0;
380 }
381
a0dbb045
JH
382 if (dowarn) {
383 SV* sv = sv_2mortal(newSVpv("Malformed UTF-8 character ", 0));
384
385 switch (warning) {
386 case 0: /* Intentionally empty. */ break;
387 case UTF8_WARN_EMPTY:
388 Perl_sv_catpvf(aTHX_ sv, "(empty string)");
389 break;
390 case UTF8_WARN_CONTINUATION:
391 Perl_sv_catpvf(aTHX_ sv, "(unexpected continuation byte 0x%02"UVxf")", uv);
392 break;
393 case UTF8_WARN_NON_CONTINUATION:
394 Perl_sv_catpvf(aTHX_ sv, "(unexpected non-continuation byte 0x%02"UVxf" after start byte 0x%02"UVxf")",
395 (UV)s[1], uv);
396 break;
397 case UTF8_WARN_FE_FF:
398 Perl_sv_catpvf(aTHX_ sv, "(byte 0x%02"UVxf")", uv);
399 break;
400 case UTF8_WARN_SHORT:
401 Perl_sv_catpvf(aTHX_ sv, "(%d byte%s, need %d)",
402 curlen, curlen == 1 ? "" : "s", expectlen);
403 break;
404 case UTF8_WARN_OVERFLOW:
405 Perl_sv_catpvf(aTHX_ sv, "(overflow at 0x%"UVxf", byte 0x%02x)",
406 ouv, *s);
407 break;
408 case UTF8_WARN_SURROGATE:
409 Perl_sv_catpvf(aTHX_ sv, "(UTF-16 surrogate 0x%04"UVxf")", uv);
410 break;
411 case UTF8_WARN_BOM:
412 Perl_sv_catpvf(aTHX_ sv, "(byte order mark 0x%04"UVxf")", uv);
413 break;
414 case UTF8_WARN_LONG:
415 Perl_sv_catpvf(aTHX_ sv, "(%d byte%s, need %d)",
416 expectlen, expectlen == 1 ? "": "s", UNISKIP(uv));
417 break;
418 case UTF8_WARN_FFFF:
419 Perl_sv_catpvf(aTHX_ sv, "(character 0x%04"UVxf")", uv);
420 break;
421 default:
422 Perl_sv_catpvf(aTHX_ sv, "(unknown reason)");
423 break;
424 }
425
426 if (warning) {
427 char *s = SvPVX(sv);
428
429 if (PL_op)
430 Perl_warner(aTHX_ WARN_UTF8,
431 "%s in %s", s, PL_op_desc[PL_op->op_type]);
432 else
433 Perl_warner(aTHX_ WARN_UTF8, "%s", s);
434 }
435 }
436
ba210ebe 437 if (retlen)
28d3d195 438 *retlen = expectlen ? expectlen : len;
ba210ebe 439
28d3d195 440 return 0;
a0ed51b3
LW
441}
442
8e84507e 443/*
37607a96 444=for apidoc A|UV|utf8_to_uvchr|U8 *s|STRLEN *retlen
9041c2e3
NIS
445
446Returns the native character value of the first character in the string C<s>
447which is assumed to be in UTF8 encoding; C<retlen> will be set to the
448length, in bytes, of that character.
449
450If C<s> does not point to a well-formed UTF8 character, zero is
451returned and retlen is set, if possible, to -1.
452
453=cut
454*/
455
456UV
37607a96 457Perl_utf8_to_uvchr(pTHX_ U8 *s, STRLEN *retlen)
9041c2e3
NIS
458{
459 return Perl_utf8n_to_uvchr(aTHX_ s, UTF8_MAXLEN, retlen, 0);
460}
461
462/*
37607a96 463=for apidoc A|UV|utf8_to_uvuni|U8 *s|STRLEN *retlen
9041c2e3
NIS
464
465Returns the Unicode code point of the first character in the string C<s>
466which is assumed to be in UTF8 encoding; C<retlen> will be set to the
467length, in bytes, of that character.
468
469This function should only be used when returned UV is considered
470an index into the Unicode semantic tables (e.g. swashes).
471
ba210ebe
JH
472If C<s> does not point to a well-formed UTF8 character, zero is
473returned and retlen is set, if possible, to -1.
8e84507e
NIS
474
475=cut
476*/
477
478UV
37607a96 479Perl_utf8_to_uvuni(pTHX_ U8 *s, STRLEN *retlen)
8e84507e 480{
9041c2e3
NIS
481 /* Call the low level routine asking for checks */
482 return Perl_utf8n_to_uvuni(aTHX_ s, UTF8_MAXLEN, retlen, 0);
8e84507e
NIS
483}
484
b76347f2 485/*
37607a96 486=for apidoc A|STRLEN|utf8_length|U8 *s|U8 *e
b76347f2
JH
487
488Return the length of the UTF-8 char encoded string C<s> in characters.
02eb7b47
JH
489Stops at C<e> (inclusive). If C<e E<lt> s> or if the scan would end
490up past C<e>, croaks.
b76347f2
JH
491
492=cut
493*/
494
495STRLEN
37607a96 496Perl_utf8_length(pTHX_ U8 *s, U8 *e)
b76347f2
JH
497{
498 STRLEN len = 0;
499
8850bf83
JH
500 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g.
501 * the bitops (especially ~) can create illegal UTF-8.
502 * In other words: in Perl UTF-8 is not just for Unicode. */
503
b76347f2 504 if (e < s)
02eb7b47 505 Perl_croak(aTHX_ "panic: utf8_length: unexpected end");
b76347f2 506 while (s < e) {
02eb7b47 507 U8 t = UTF8SKIP(s);
b76347f2
JH
508
509 if (e - s < t)
c4d5f83a 510 Perl_croak(aTHX_ "panic: utf8_length: s=%p (%02X) e=%p l=%d - unaligned end",s,*s,e,t);
b76347f2
JH
511 s += t;
512 len++;
513 }
514
515 return len;
516}
517
b06226ff 518/*
eebe1485 519=for apidoc A|IV|utf8_distance|U8 *a|U8 *b
b06226ff
JH
520
521Returns the number of UTF8 characters between the UTF-8 pointers C<a>
522and C<b>.
523
524WARNING: use only if you *know* that the pointers point inside the
525same UTF-8 buffer.
526
37607a96
PK
527=cut
528*/
a0ed51b3 529
02eb7b47 530IV
864dbfa3 531Perl_utf8_distance(pTHX_ U8 *a, U8 *b)
a0ed51b3 532{
02eb7b47
JH
533 IV off = 0;
534
8850bf83
JH
535 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g.
536 * the bitops (especially ~) can create illegal UTF-8.
537 * In other words: in Perl UTF-8 is not just for Unicode. */
538
a0ed51b3
LW
539 if (a < b) {
540 while (a < b) {
02eb7b47
JH
541 U8 c = UTF8SKIP(a);
542
543 if (b - a < c)
544 Perl_croak(aTHX_ "panic: utf8_distance: unaligned end");
545 a += c;
a0ed51b3
LW
546 off--;
547 }
548 }
549 else {
550 while (b < a) {
02eb7b47
JH
551 U8 c = UTF8SKIP(b);
552
553 if (a - b < c)
554 Perl_croak(aTHX_ "panic: utf8_distance: unaligned end");
555 b += c;
a0ed51b3
LW
556 off++;
557 }
558 }
02eb7b47 559
a0ed51b3
LW
560 return off;
561}
562
b06226ff 563/*
37607a96 564=for apidoc A|U8 *|utf8_hop|U8 *s|I32 off
b06226ff 565
8850bf83
JH
566Return the UTF-8 pointer C<s> displaced by C<off> characters, either
567forward or backward.
b06226ff
JH
568
569WARNING: do not use the following unless you *know* C<off> is within
8850bf83
JH
570the UTF-8 data pointed to by C<s> *and* that on entry C<s> is aligned
571on the first byte of character or just after the last byte of a character.
b06226ff 572
37607a96
PK
573=cut
574*/
a0ed51b3
LW
575
576U8 *
864dbfa3 577Perl_utf8_hop(pTHX_ U8 *s, I32 off)
a0ed51b3 578{
8850bf83
JH
579 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g
580 * the bitops (especially ~) can create illegal UTF-8.
581 * In other words: in Perl UTF-8 is not just for Unicode. */
582
a0ed51b3
LW
583 if (off >= 0) {
584 while (off--)
585 s += UTF8SKIP(s);
586 }
587 else {
588 while (off++) {
589 s--;
8850bf83
JH
590 while (UTF8_IS_CONTINUATION(*s))
591 s--;
a0ed51b3
LW
592 }
593 }
594 return s;
595}
596
6940069f 597/*
eebe1485 598=for apidoc A|U8 *|utf8_to_bytes|U8 *s|STRLEN *len
6940069f 599
246fae53
MG
600Converts a string C<s> of length C<len> from UTF8 into byte encoding.
601Unlike C<bytes_to_utf8>, this over-writes the original string, and
602updates len to contain the new length.
67e989fb 603Returns zero on failure, setting C<len> to -1.
6940069f
GS
604
605=cut
606*/
607
608U8 *
37607a96 609Perl_utf8_to_bytes(pTHX_ U8 *s, STRLEN *len)
6940069f 610{
6940069f
GS
611 U8 *send;
612 U8 *d;
dcad2880 613 U8 *save = s;
246fae53
MG
614
615 /* ensure valid UTF8 and chars < 256 before updating string */
dcad2880
JH
616 for (send = s + *len; s < send; ) {
617 U8 c = *s++;
618
1d72bdf6
NIS
619 if (!UTF8_IS_INVARIANT(c) &&
620 (!UTF8_IS_DOWNGRADEABLE_START(c) || (s >= send)
621 || !(c = *s++) || !UTF8_IS_CONTINUATION(c))) {
dcad2880
JH
622 *len = -1;
623 return 0;
624 }
246fae53 625 }
dcad2880
JH
626
627 d = s = save;
6940069f 628 while (s < send) {
ed646e6e 629 STRLEN ulen;
9041c2e3 630 *d++ = (U8)utf8_to_uvchr(s, &ulen);
ed646e6e 631 s += ulen;
6940069f
GS
632 }
633 *d = '\0';
246fae53 634 *len = d - save;
6940069f
GS
635 return save;
636}
637
638/*
f9a63242
JH
639=for apidoc A|U8 *|bytes_from_utf8|U8 *s|STRLEN *len|bool *is_utf8
640
641Converts a string C<s> of length C<len> from UTF8 into byte encoding.
642Unlike <utf8_to_bytes> but like C<bytes_to_utf8>, returns a pointer to
ef9edfd0
JH
643the newly-created string, and updates C<len> to contain the new
644length. Returns the original string if no conversion occurs, C<len>
645is unchanged. Do nothing if C<is_utf8> points to 0. Sets C<is_utf8> to
6460 if C<s> is converted or contains all 7bit characters.
f9a63242 647
37607a96
PK
648=cut
649*/
f9a63242
JH
650
651U8 *
37607a96 652Perl_bytes_from_utf8(pTHX_ U8 *s, STRLEN *len, bool *is_utf8)
f9a63242 653{
f9a63242
JH
654 U8 *d;
655 U8 *start = s;
db42d148 656 U8 *send;
f9a63242
JH
657 I32 count = 0;
658
659 if (!*is_utf8)
660 return start;
661
ef9edfd0 662 /* ensure valid UTF8 and chars < 256 before converting string */
f9a63242
JH
663 for (send = s + *len; s < send;) {
664 U8 c = *s++;
1d72bdf6 665 if (!UTF8_IS_INVARIANT(c)) {
db42d148
NIS
666 if (UTF8_IS_DOWNGRADEABLE_START(c) && s < send &&
667 (c = *s++) && UTF8_IS_CONTINUATION(c))
668 count++;
669 else
f9a63242 670 return start;
db42d148 671 }
f9a63242
JH
672 }
673
674 *is_utf8 = 0;
675
f9a63242 676 Newz(801, d, (*len) - count + 1, U8);
ef9edfd0 677 s = start; start = d;
f9a63242
JH
678 while (s < send) {
679 U8 c = *s++;
c4d5f83a
NIS
680 if (!UTF8_IS_INVARIANT(c)) {
681 /* Then it is two-byte encoded */
682 c = UTF8_ACCUMULATE(NATIVE_TO_UTF(c), *s++);
683 c = ASCII_TO_NATIVE(c);
684 }
685 *d++ = c;
f9a63242
JH
686 }
687 *d = '\0';
688 *len = d - start;
689 return start;
690}
691
692/*
eebe1485 693=for apidoc A|U8 *|bytes_to_utf8|U8 *s|STRLEN *len
6940069f
GS
694
695Converts a string C<s> of length C<len> from ASCII into UTF8 encoding.
6662521e
GS
696Returns a pointer to the newly-created string, and sets C<len> to
697reflect the new length.
6940069f 698
497711e7 699=cut
6940069f
GS
700*/
701
702U8*
37607a96 703Perl_bytes_to_utf8(pTHX_ U8 *s, STRLEN *len)
6940069f 704{
6940069f
GS
705 U8 *send;
706 U8 *d;
707 U8 *dst;
6662521e 708 send = s + (*len);
6940069f 709
6662521e 710 Newz(801, d, (*len) * 2 + 1, U8);
6940069f
GS
711 dst = d;
712
713 while (s < send) {
db42d148 714 UV uv = NATIVE_TO_ASCII(*s++);
c4d5f83a
NIS
715 if (UNI_IS_INVARIANT(uv))
716 *d++ = UTF_TO_NATIVE(uv);
6940069f 717 else {
90f44359
JH
718 *d++ = UTF8_EIGHT_BIT_HI(uv);
719 *d++ = UTF8_EIGHT_BIT_LO(uv);
6940069f
GS
720 }
721 }
722 *d = '\0';
6662521e 723 *len = d-dst;
6940069f
GS
724 return dst;
725}
726
a0ed51b3 727/*
dea0fc0b 728 * Convert native (big-endian) or reversed (little-endian) UTF-16 to UTF-8.
a0ed51b3
LW
729 *
730 * Destination must be pre-extended to 3/2 source. Do not use in-place.
731 * We optimize for native, for obvious reasons. */
732
733U8*
dea0fc0b 734Perl_utf16_to_utf8(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen)
a0ed51b3 735{
dea0fc0b
JH
736 U8* pend;
737 U8* dstart = d;
738
739 if (bytelen & 1)
a7867d0a 740 Perl_croak(aTHX_ "panic: utf16_to_utf8: odd bytelen");
dea0fc0b
JH
741
742 pend = p + bytelen;
743
a0ed51b3 744 while (p < pend) {
dea0fc0b
JH
745 UV uv = (p[0] << 8) + p[1]; /* UTF-16BE */
746 p += 2;
a0ed51b3
LW
747 if (uv < 0x80) {
748 *d++ = uv;
749 continue;
750 }
751 if (uv < 0x800) {
752 *d++ = (( uv >> 6) | 0xc0);
753 *d++ = (( uv & 0x3f) | 0x80);
754 continue;
755 }
756 if (uv >= 0xd800 && uv < 0xdbff) { /* surrogates */
dea0fc0b
JH
757 UV low = *p++;
758 if (low < 0xdc00 || low >= 0xdfff)
759 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
a0ed51b3
LW
760 uv = ((uv - 0xd800) << 10) + (low - 0xdc00) + 0x10000;
761 }
762 if (uv < 0x10000) {
763 *d++ = (( uv >> 12) | 0xe0);
764 *d++ = (((uv >> 6) & 0x3f) | 0x80);
765 *d++ = (( uv & 0x3f) | 0x80);
766 continue;
767 }
768 else {
769 *d++ = (( uv >> 18) | 0xf0);
770 *d++ = (((uv >> 12) & 0x3f) | 0x80);
771 *d++ = (((uv >> 6) & 0x3f) | 0x80);
772 *d++ = (( uv & 0x3f) | 0x80);
773 continue;
774 }
775 }
dea0fc0b 776 *newlen = d - dstart;
a0ed51b3
LW
777 return d;
778}
779
780/* Note: this one is slightly destructive of the source. */
781
782U8*
dea0fc0b 783Perl_utf16_to_utf8_reversed(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen)
a0ed51b3
LW
784{
785 U8* s = (U8*)p;
786 U8* send = s + bytelen;
787 while (s < send) {
788 U8 tmp = s[0];
789 s[0] = s[1];
790 s[1] = tmp;
791 s += 2;
792 }
dea0fc0b 793 return utf16_to_utf8(p, d, bytelen, newlen);
a0ed51b3
LW
794}
795
796/* for now these are all defined (inefficiently) in terms of the utf8 versions */
797
798bool
864dbfa3 799Perl_is_uni_alnum(pTHX_ U32 c)
a0ed51b3 800{
ad391ad9 801 U8 tmpbuf[UTF8_MAXLEN+1];
2b9d42f0 802 uvchr_to_utf8(tmpbuf, (UV)c);
a0ed51b3
LW
803 return is_utf8_alnum(tmpbuf);
804}
805
806bool
b8c5462f
JH
807Perl_is_uni_alnumc(pTHX_ U32 c)
808{
ad391ad9 809 U8 tmpbuf[UTF8_MAXLEN+1];
2b9d42f0 810 uvchr_to_utf8(tmpbuf, (UV)c);
b8c5462f
JH
811 return is_utf8_alnumc(tmpbuf);
812}
813
814bool
864dbfa3 815Perl_is_uni_idfirst(pTHX_ U32 c)
a0ed51b3 816{
ad391ad9 817 U8 tmpbuf[UTF8_MAXLEN+1];
2b9d42f0 818 uvchr_to_utf8(tmpbuf, (UV)c);
a0ed51b3
LW
819 return is_utf8_idfirst(tmpbuf);
820}
821
822bool
864dbfa3 823Perl_is_uni_alpha(pTHX_ U32 c)
a0ed51b3 824{
ad391ad9 825 U8 tmpbuf[UTF8_MAXLEN+1];
2b9d42f0 826 uvchr_to_utf8(tmpbuf, (UV)c);
a0ed51b3
LW
827 return is_utf8_alpha(tmpbuf);
828}
829
830bool
4d61ec05
GS
831Perl_is_uni_ascii(pTHX_ U32 c)
832{
ad391ad9 833 U8 tmpbuf[UTF8_MAXLEN+1];
2b9d42f0 834 uvchr_to_utf8(tmpbuf, (UV)c);
4d61ec05
GS
835 return is_utf8_ascii(tmpbuf);
836}
837
838bool
864dbfa3 839Perl_is_uni_space(pTHX_ U32 c)
a0ed51b3 840{
ad391ad9 841 U8 tmpbuf[UTF8_MAXLEN+1];
2b9d42f0 842 uvchr_to_utf8(tmpbuf, (UV)c);
a0ed51b3
LW
843 return is_utf8_space(tmpbuf);
844}
845
846bool
864dbfa3 847Perl_is_uni_digit(pTHX_ U32 c)
a0ed51b3 848{
ad391ad9 849 U8 tmpbuf[UTF8_MAXLEN+1];
2b9d42f0 850 uvchr_to_utf8(tmpbuf, (UV)c);
a0ed51b3
LW
851 return is_utf8_digit(tmpbuf);
852}
853
854bool
864dbfa3 855Perl_is_uni_upper(pTHX_ U32 c)
a0ed51b3 856{
ad391ad9 857 U8 tmpbuf[UTF8_MAXLEN+1];
2b9d42f0 858 uvchr_to_utf8(tmpbuf, (UV)c);
a0ed51b3
LW
859 return is_utf8_upper(tmpbuf);
860}
861
862bool
864dbfa3 863Perl_is_uni_lower(pTHX_ U32 c)
a0ed51b3 864{
ad391ad9 865 U8 tmpbuf[UTF8_MAXLEN+1];
2b9d42f0 866 uvchr_to_utf8(tmpbuf, (UV)c);
a0ed51b3
LW
867 return is_utf8_lower(tmpbuf);
868}
869
870bool
b8c5462f
JH
871Perl_is_uni_cntrl(pTHX_ U32 c)
872{
ad391ad9 873 U8 tmpbuf[UTF8_MAXLEN+1];
2b9d42f0 874 uvchr_to_utf8(tmpbuf, (UV)c);
b8c5462f
JH
875 return is_utf8_cntrl(tmpbuf);
876}
877
878bool
879Perl_is_uni_graph(pTHX_ U32 c)
880{
ad391ad9 881 U8 tmpbuf[UTF8_MAXLEN+1];
2b9d42f0 882 uvchr_to_utf8(tmpbuf, (UV)c);
b8c5462f
JH
883 return is_utf8_graph(tmpbuf);
884}
885
886bool
864dbfa3 887Perl_is_uni_print(pTHX_ U32 c)
a0ed51b3 888{
ad391ad9 889 U8 tmpbuf[UTF8_MAXLEN+1];
2b9d42f0 890 uvchr_to_utf8(tmpbuf, (UV)c);
a0ed51b3
LW
891 return is_utf8_print(tmpbuf);
892}
893
b8c5462f 894bool
f248d071 895Perl_is_uni_punct(pTHX_ U32 c)
b8c5462f 896{
ad391ad9 897 U8 tmpbuf[UTF8_MAXLEN+1];
2b9d42f0 898 uvchr_to_utf8(tmpbuf, (UV)c);
b8c5462f
JH
899 return is_utf8_punct(tmpbuf);
900}
901
4d61ec05
GS
902bool
903Perl_is_uni_xdigit(pTHX_ U32 c)
904{
ad391ad9 905 U8 tmpbuf[UTF8_MAXLEN+1];
2b9d42f0 906 uvchr_to_utf8(tmpbuf, (UV)c);
4d61ec05
GS
907 return is_utf8_xdigit(tmpbuf);
908}
909
a0ed51b3 910U32
864dbfa3 911Perl_to_uni_upper(pTHX_ U32 c)
a0ed51b3 912{
ad391ad9 913 U8 tmpbuf[UTF8_MAXLEN+1];
2b9d42f0 914 uvchr_to_utf8(tmpbuf, (UV)c);
a0ed51b3
LW
915 return to_utf8_upper(tmpbuf);
916}
917
918U32
864dbfa3 919Perl_to_uni_title(pTHX_ U32 c)
a0ed51b3 920{
ad391ad9 921 U8 tmpbuf[UTF8_MAXLEN+1];
2b9d42f0 922 uvchr_to_utf8(tmpbuf, (UV)c);
a0ed51b3
LW
923 return to_utf8_title(tmpbuf);
924}
925
926U32
864dbfa3 927Perl_to_uni_lower(pTHX_ U32 c)
a0ed51b3 928{
ad391ad9 929 U8 tmpbuf[UTF8_MAXLEN+1];
2b9d42f0 930 uvchr_to_utf8(tmpbuf, (UV)c);
a0ed51b3
LW
931 return to_utf8_lower(tmpbuf);
932}
933
934/* for now these all assume no locale info available for Unicode > 255 */
935
936bool
864dbfa3 937Perl_is_uni_alnum_lc(pTHX_ U32 c)
a0ed51b3
LW
938{
939 return is_uni_alnum(c); /* XXX no locale support yet */
940}
941
942bool
b8c5462f
JH
943Perl_is_uni_alnumc_lc(pTHX_ U32 c)
944{
945 return is_uni_alnumc(c); /* XXX no locale support yet */
946}
947
948bool
864dbfa3 949Perl_is_uni_idfirst_lc(pTHX_ U32 c)
a0ed51b3
LW
950{
951 return is_uni_idfirst(c); /* XXX no locale support yet */
952}
953
954bool
864dbfa3 955Perl_is_uni_alpha_lc(pTHX_ U32 c)
a0ed51b3
LW
956{
957 return is_uni_alpha(c); /* XXX no locale support yet */
958}
959
960bool
4d61ec05
GS
961Perl_is_uni_ascii_lc(pTHX_ U32 c)
962{
963 return is_uni_ascii(c); /* XXX no locale support yet */
964}
965
966bool
864dbfa3 967Perl_is_uni_space_lc(pTHX_ U32 c)
a0ed51b3
LW
968{
969 return is_uni_space(c); /* XXX no locale support yet */
970}
971
972bool
864dbfa3 973Perl_is_uni_digit_lc(pTHX_ U32 c)
a0ed51b3
LW
974{
975 return is_uni_digit(c); /* XXX no locale support yet */
976}
977
978bool
864dbfa3 979Perl_is_uni_upper_lc(pTHX_ U32 c)
a0ed51b3
LW
980{
981 return is_uni_upper(c); /* XXX no locale support yet */
982}
983
984bool
864dbfa3 985Perl_is_uni_lower_lc(pTHX_ U32 c)
a0ed51b3
LW
986{
987 return is_uni_lower(c); /* XXX no locale support yet */
988}
989
990bool
b8c5462f
JH
991Perl_is_uni_cntrl_lc(pTHX_ U32 c)
992{
993 return is_uni_cntrl(c); /* XXX no locale support yet */
994}
995
996bool
997Perl_is_uni_graph_lc(pTHX_ U32 c)
998{
999 return is_uni_graph(c); /* XXX no locale support yet */
1000}
1001
1002bool
864dbfa3 1003Perl_is_uni_print_lc(pTHX_ U32 c)
a0ed51b3
LW
1004{
1005 return is_uni_print(c); /* XXX no locale support yet */
1006}
1007
b8c5462f
JH
1008bool
1009Perl_is_uni_punct_lc(pTHX_ U32 c)
1010{
1011 return is_uni_punct(c); /* XXX no locale support yet */
1012}
1013
4d61ec05
GS
1014bool
1015Perl_is_uni_xdigit_lc(pTHX_ U32 c)
1016{
1017 return is_uni_xdigit(c); /* XXX no locale support yet */
1018}
1019
a0ed51b3 1020U32
864dbfa3 1021Perl_to_uni_upper_lc(pTHX_ U32 c)
a0ed51b3
LW
1022{
1023 return to_uni_upper(c); /* XXX no locale support yet */
1024}
1025
1026U32
864dbfa3 1027Perl_to_uni_title_lc(pTHX_ U32 c)
a0ed51b3
LW
1028{
1029 return to_uni_title(c); /* XXX no locale support yet */
1030}
1031
1032U32
864dbfa3 1033Perl_to_uni_lower_lc(pTHX_ U32 c)
a0ed51b3
LW
1034{
1035 return to_uni_lower(c); /* XXX no locale support yet */
1036}
1037
a0ed51b3 1038bool
864dbfa3 1039Perl_is_utf8_alnum(pTHX_ U8 *p)
a0ed51b3 1040{
386d01d6
GS
1041 if (!is_utf8_char(p))
1042 return FALSE;
a0ed51b3 1043 if (!PL_utf8_alnum)
289d4f09
ML
1044 /* NOTE: "IsWord", not "IsAlnum", since Alnum is a true
1045 * descendant of isalnum(3), in other words, it doesn't
1046 * contain the '_'. --jhi */
1047 PL_utf8_alnum = swash_init("utf8", "IsWord", &PL_sv_undef, 0, 0);
a0ed51b3
LW
1048 return swash_fetch(PL_utf8_alnum, p);
1049/* return *p == '_' || is_utf8_alpha(p) || is_utf8_digit(p); */
1050#ifdef SURPRISINGLY_SLOWER /* probably because alpha is usually true */
1051 if (!PL_utf8_alnum)
1052 PL_utf8_alnum = swash_init("utf8", "",
1053 sv_2mortal(newSVpv("+utf8::IsAlpha\n+utf8::IsDigit\n005F\n",0)), 0, 0);
1054 return swash_fetch(PL_utf8_alnum, p);
1055#endif
1056}
1057
1058bool
b8c5462f
JH
1059Perl_is_utf8_alnumc(pTHX_ U8 *p)
1060{
386d01d6
GS
1061 if (!is_utf8_char(p))
1062 return FALSE;
b8c5462f
JH
1063 if (!PL_utf8_alnum)
1064 PL_utf8_alnum = swash_init("utf8", "IsAlnumC", &PL_sv_undef, 0, 0);
1065 return swash_fetch(PL_utf8_alnum, p);
1066/* return is_utf8_alpha(p) || is_utf8_digit(p); */
1067#ifdef SURPRISINGLY_SLOWER /* probably because alpha is usually true */
1068 if (!PL_utf8_alnum)
1069 PL_utf8_alnum = swash_init("utf8", "",
1070 sv_2mortal(newSVpv("+utf8::IsAlpha\n+utf8::IsDigit\n005F\n",0)), 0, 0);
1071 return swash_fetch(PL_utf8_alnum, p);
1072#endif
1073}
1074
1075bool
864dbfa3 1076Perl_is_utf8_idfirst(pTHX_ U8 *p)
a0ed51b3
LW
1077{
1078 return *p == '_' || is_utf8_alpha(p);
1079}
1080
1081bool
864dbfa3 1082Perl_is_utf8_alpha(pTHX_ U8 *p)
a0ed51b3 1083{
386d01d6
GS
1084 if (!is_utf8_char(p))
1085 return FALSE;
a0ed51b3 1086 if (!PL_utf8_alpha)
e24b16f9 1087 PL_utf8_alpha = swash_init("utf8", "IsAlpha", &PL_sv_undef, 0, 0);
a0ed51b3
LW
1088 return swash_fetch(PL_utf8_alpha, p);
1089}
1090
1091bool
b8c5462f
JH
1092Perl_is_utf8_ascii(pTHX_ U8 *p)
1093{
386d01d6
GS
1094 if (!is_utf8_char(p))
1095 return FALSE;
b8c5462f
JH
1096 if (!PL_utf8_ascii)
1097 PL_utf8_ascii = swash_init("utf8", "IsAscii", &PL_sv_undef, 0, 0);
1098 return swash_fetch(PL_utf8_ascii, p);
1099}
1100
1101bool
864dbfa3 1102Perl_is_utf8_space(pTHX_ U8 *p)
a0ed51b3 1103{
386d01d6
GS
1104 if (!is_utf8_char(p))
1105 return FALSE;
a0ed51b3 1106 if (!PL_utf8_space)
3bec3564 1107 PL_utf8_space = swash_init("utf8", "IsSpacePerl", &PL_sv_undef, 0, 0);
a0ed51b3
LW
1108 return swash_fetch(PL_utf8_space, p);
1109}
1110
1111bool
864dbfa3 1112Perl_is_utf8_digit(pTHX_ U8 *p)
a0ed51b3 1113{
386d01d6
GS
1114 if (!is_utf8_char(p))
1115 return FALSE;
a0ed51b3 1116 if (!PL_utf8_digit)
e24b16f9 1117 PL_utf8_digit = swash_init("utf8", "IsDigit", &PL_sv_undef, 0, 0);
a0ed51b3
LW
1118 return swash_fetch(PL_utf8_digit, p);
1119}
1120
1121bool
864dbfa3 1122Perl_is_utf8_upper(pTHX_ U8 *p)
a0ed51b3 1123{
386d01d6
GS
1124 if (!is_utf8_char(p))
1125 return FALSE;
a0ed51b3 1126 if (!PL_utf8_upper)
e24b16f9 1127 PL_utf8_upper = swash_init("utf8", "IsUpper", &PL_sv_undef, 0, 0);
a0ed51b3
LW
1128 return swash_fetch(PL_utf8_upper, p);
1129}
1130
1131bool
864dbfa3 1132Perl_is_utf8_lower(pTHX_ U8 *p)
a0ed51b3 1133{
386d01d6
GS
1134 if (!is_utf8_char(p))
1135 return FALSE;
a0ed51b3 1136 if (!PL_utf8_lower)
e24b16f9 1137 PL_utf8_lower = swash_init("utf8", "IsLower", &PL_sv_undef, 0, 0);
a0ed51b3
LW
1138 return swash_fetch(PL_utf8_lower, p);
1139}
1140
1141bool
b8c5462f
JH
1142Perl_is_utf8_cntrl(pTHX_ U8 *p)
1143{
386d01d6
GS
1144 if (!is_utf8_char(p))
1145 return FALSE;
b8c5462f
JH
1146 if (!PL_utf8_cntrl)
1147 PL_utf8_cntrl = swash_init("utf8", "IsCntrl", &PL_sv_undef, 0, 0);
1148 return swash_fetch(PL_utf8_cntrl, p);
1149}
1150
1151bool
1152Perl_is_utf8_graph(pTHX_ U8 *p)
1153{
386d01d6
GS
1154 if (!is_utf8_char(p))
1155 return FALSE;
b8c5462f
JH
1156 if (!PL_utf8_graph)
1157 PL_utf8_graph = swash_init("utf8", "IsGraph", &PL_sv_undef, 0, 0);
1158 return swash_fetch(PL_utf8_graph, p);
1159}
1160
1161bool
864dbfa3 1162Perl_is_utf8_print(pTHX_ U8 *p)
a0ed51b3 1163{
386d01d6
GS
1164 if (!is_utf8_char(p))
1165 return FALSE;
a0ed51b3 1166 if (!PL_utf8_print)
e24b16f9 1167 PL_utf8_print = swash_init("utf8", "IsPrint", &PL_sv_undef, 0, 0);
a0ed51b3
LW
1168 return swash_fetch(PL_utf8_print, p);
1169}
1170
1171bool
b8c5462f
JH
1172Perl_is_utf8_punct(pTHX_ U8 *p)
1173{
386d01d6
GS
1174 if (!is_utf8_char(p))
1175 return FALSE;
b8c5462f
JH
1176 if (!PL_utf8_punct)
1177 PL_utf8_punct = swash_init("utf8", "IsPunct", &PL_sv_undef, 0, 0);
1178 return swash_fetch(PL_utf8_punct, p);
1179}
1180
1181bool
1182Perl_is_utf8_xdigit(pTHX_ U8 *p)
1183{
386d01d6
GS
1184 if (!is_utf8_char(p))
1185 return FALSE;
b8c5462f
JH
1186 if (!PL_utf8_xdigit)
1187 PL_utf8_xdigit = swash_init("utf8", "IsXDigit", &PL_sv_undef, 0, 0);
1188 return swash_fetch(PL_utf8_xdigit, p);
1189}
1190
1191bool
864dbfa3 1192Perl_is_utf8_mark(pTHX_ U8 *p)
a0ed51b3 1193{
386d01d6
GS
1194 if (!is_utf8_char(p))
1195 return FALSE;
a0ed51b3 1196 if (!PL_utf8_mark)
e24b16f9 1197 PL_utf8_mark = swash_init("utf8", "IsM", &PL_sv_undef, 0, 0);
a0ed51b3
LW
1198 return swash_fetch(PL_utf8_mark, p);
1199}
1200
2104c8d9 1201UV
864dbfa3 1202Perl_to_utf8_upper(pTHX_ U8 *p)
a0ed51b3
LW
1203{
1204 UV uv;
1205
1206 if (!PL_utf8_toupper)
e24b16f9 1207 PL_utf8_toupper = swash_init("utf8", "ToUpper", &PL_sv_undef, 4, 0);
a0ed51b3 1208 uv = swash_fetch(PL_utf8_toupper, p);
9041c2e3 1209 return uv ? UNI_TO_NATIVE(uv) : utf8_to_uvchr(p,0);
a0ed51b3
LW
1210}
1211
2104c8d9 1212UV
864dbfa3 1213Perl_to_utf8_title(pTHX_ U8 *p)
a0ed51b3
LW
1214{
1215 UV uv;
1216
1217 if (!PL_utf8_totitle)
e24b16f9 1218 PL_utf8_totitle = swash_init("utf8", "ToTitle", &PL_sv_undef, 4, 0);
a0ed51b3 1219 uv = swash_fetch(PL_utf8_totitle, p);
9041c2e3 1220 return uv ? UNI_TO_NATIVE(uv) : utf8_to_uvchr(p,0);
a0ed51b3
LW
1221}
1222
2104c8d9 1223UV
864dbfa3 1224Perl_to_utf8_lower(pTHX_ U8 *p)
a0ed51b3
LW
1225{
1226 UV uv;
1227
1228 if (!PL_utf8_tolower)
e24b16f9 1229 PL_utf8_tolower = swash_init("utf8", "ToLower", &PL_sv_undef, 4, 0);
a0ed51b3 1230 uv = swash_fetch(PL_utf8_tolower, p);
9041c2e3 1231 return uv ? UNI_TO_NATIVE(uv) : utf8_to_uvchr(p,0);
a0ed51b3
LW
1232}
1233
1234/* a "swash" is a swatch hash */
1235
1236SV*
864dbfa3 1237Perl_swash_init(pTHX_ char* pkg, char* name, SV *listsv, I32 minbits, I32 none)
a0ed51b3
LW
1238{
1239 SV* retval;
bf1fed83 1240 SV* tokenbufsv = sv_2mortal(NEWSV(0,0));
8e84507e 1241 dSP;
1b026014 1242 HV *stash = gv_stashpvn(pkg, strlen(pkg), FALSE);
ce3b816e 1243
1b026014 1244 if (!gv_fetchmeth(stash, "SWASHNEW", 8, -1)) { /* demand load utf8 */
ce3b816e
GS
1245 ENTER;
1246 Perl_load_module(aTHX_ PERL_LOADMOD_NOIMPORT, newSVpv(pkg,0), Nullsv);
1247 LEAVE;
1248 }
1249 SPAGAIN;
a0ed51b3
LW
1250 PUSHSTACKi(PERLSI_MAGIC);
1251 PUSHMARK(SP);
1252 EXTEND(SP,5);
1253 PUSHs(sv_2mortal(newSVpvn(pkg, strlen(pkg))));
1254 PUSHs(sv_2mortal(newSVpvn(name, strlen(name))));
1255 PUSHs(listsv);
1256 PUSHs(sv_2mortal(newSViv(minbits)));
1257 PUSHs(sv_2mortal(newSViv(none)));
1258 PUTBACK;
1259 ENTER;
1260 SAVEI32(PL_hints);
1261 PL_hints = 0;
1262 save_re_context();
bf1fed83
JH
1263 if (PL_curcop == &PL_compiling)
1264 /* XXX ought to be handled by lex_start */
1265 sv_setpv(tokenbufsv, PL_tokenbuf);
864dbfa3 1266 if (call_method("SWASHNEW", G_SCALAR))
8e84507e 1267 retval = newSVsv(*PL_stack_sp--);
a0ed51b3 1268 else
e24b16f9 1269 retval = &PL_sv_undef;
a0ed51b3
LW
1270 LEAVE;
1271 POPSTACK;
e24b16f9 1272 if (PL_curcop == &PL_compiling) {
bf1fed83
JH
1273 STRLEN len;
1274 char* pv = SvPV(tokenbufsv, len);
1275
1276 Copy(pv, PL_tokenbuf, len+1, char);
e24b16f9 1277 PL_curcop->op_private = PL_hints;
a0ed51b3
LW
1278 }
1279 if (!SvROK(retval) || SvTYPE(SvRV(retval)) != SVt_PVHV)
cea2e8a9 1280 Perl_croak(aTHX_ "SWASHNEW didn't return an HV ref");
a0ed51b3
LW
1281 return retval;
1282}
1283
1284UV
864dbfa3 1285Perl_swash_fetch(pTHX_ SV *sv, U8 *ptr)
a0ed51b3
LW
1286{
1287 HV* hv = (HV*)SvRV(sv);
7d85a32c
JH
1288 /* Given a UTF-X encoded char 0xAA..0xYY,0xZZ
1289 then the "swatch" is a vec() for al the chars which start
1290 with 0xAA..0xYY
1291 So the key in the hash is length of encoded char -1
1292 */
a0ed51b3 1293 U32 klen = UTF8SKIP(ptr) - 1;
7d85a32c 1294 U32 off = ptr[klen];
a0ed51b3 1295 STRLEN slen;
7d85a32c 1296 STRLEN needents;
dfe13c55 1297 U8 *tmps;
a0ed51b3
LW
1298 U32 bit;
1299 SV *retval;
1300
7d85a32c
JH
1301 if (klen == 0)
1302 {
1303 /* If char in invariant then swatch is for all the invariant chars
1304 * In both UTF-8 and UTF8-MOD that happens to be UTF_CONTINUATION_MARK
1305 */
1306 needents = UTF_CONTINUATION_MARK;
1307 off = NATIVE_TO_UTF(ptr[klen]);
1308 }
1309 else
1310 {
1311 /* If char is encoded then swatch is for the prefix */
1312 needents = (1 << UTF_ACCUMULATION_SHIFT);
1313 off = NATIVE_TO_UTF(ptr[klen]) & UTF_CONTINUATION_MASK;
1314 }
1315
a0ed51b3
LW
1316 /*
1317 * This single-entry cache saves about 1/3 of the utf8 overhead in test
1318 * suite. (That is, only 7-8% overall over just a hash cache. Still,
1319 * it's nothing to sniff at.) Pity we usually come through at least
1320 * two function calls to get here...
1321 *
1322 * NB: this code assumes that swatches are never modified, once generated!
1323 */
1324
1325 if (hv == PL_last_swash_hv &&
1326 klen == PL_last_swash_klen &&
12ae5dfc 1327 (!klen || memEQ((char *)ptr,(char *)PL_last_swash_key,klen)) )
a0ed51b3
LW
1328 {
1329 tmps = PL_last_swash_tmps;
1330 slen = PL_last_swash_slen;
1331 }
1332 else {
1333 /* Try our second-level swatch cache, kept in a hash. */
dfe13c55 1334 SV** svp = hv_fetch(hv, (char*)ptr, klen, FALSE);
a0ed51b3
LW
1335
1336 /* If not cached, generate it via utf8::SWASHGET */
dfe13c55 1337 if (!svp || !SvPOK(*svp) || !(tmps = (U8*)SvPV(*svp, slen))) {
a0ed51b3 1338 dSP;
2b9d42f0
NIS
1339 /* We use utf8n_to_uvuni() as we want an index into
1340 Unicode tables, not a native character number.
1341 */
1342 UV code_point = utf8n_to_uvuni(ptr, UTF8_MAXLEN, NULL, 0);
a0ed51b3
LW
1343 ENTER;
1344 SAVETMPS;
1345 save_re_context();
1346 PUSHSTACKi(PERLSI_MAGIC);
1347 PUSHMARK(SP);
1348 EXTEND(SP,3);
1349 PUSHs((SV*)sv);
ffbc6a93
JH
1350 /* On EBCDIC & ~(0xA0-1) isn't a useful thing to do */
1351 PUSHs(sv_2mortal(newSViv((klen) ? (code_point & ~(needents - 1)) : 0)));
a0ed51b3
LW
1352 PUSHs(sv_2mortal(newSViv(needents)));
1353 PUTBACK;
864dbfa3 1354 if (call_method("SWASHGET", G_SCALAR))
8e84507e 1355 retval = newSVsv(*PL_stack_sp--);
a0ed51b3 1356 else
e24b16f9 1357 retval = &PL_sv_undef;
a0ed51b3
LW
1358 POPSTACK;
1359 FREETMPS;
1360 LEAVE;
e24b16f9
GS
1361 if (PL_curcop == &PL_compiling)
1362 PL_curcop->op_private = PL_hints;
a0ed51b3 1363
dfe13c55 1364 svp = hv_store(hv, (char*)ptr, klen, retval, 0);
a0ed51b3 1365
7d85a32c 1366 if (!svp || !(tmps = (U8*)SvPV(*svp, slen)) || (slen << 3) < needents)
cea2e8a9 1367 Perl_croak(aTHX_ "SWASHGET didn't return result of proper length");
a0ed51b3
LW
1368 }
1369
1370 PL_last_swash_hv = hv;
1371 PL_last_swash_klen = klen;
1372 PL_last_swash_tmps = tmps;
1373 PL_last_swash_slen = slen;
1374 if (klen)
1375 Copy(ptr, PL_last_swash_key, klen, U8);
1376 }
1377
9faf8d75 1378 switch ((int)((slen << 3) / needents)) {
a0ed51b3
LW
1379 case 1:
1380 bit = 1 << (off & 7);
1381 off >>= 3;
1382 return (tmps[off] & bit) != 0;
1383 case 8:
1384 return tmps[off];
1385 case 16:
1386 off <<= 1;
1387 return (tmps[off] << 8) + tmps[off + 1] ;
1388 case 32:
1389 off <<= 2;
1390 return (tmps[off] << 24) + (tmps[off+1] << 16) + (tmps[off+2] << 8) + tmps[off + 3] ;
1391 }
cea2e8a9 1392 Perl_croak(aTHX_ "panic: swash_fetch");
a0ed51b3
LW
1393 return 0;
1394}
2b9d42f0
NIS
1395
1396
1397/*
37607a96 1398=for apidoc A|U8 *|uvchr_to_utf8|U8 *d|UV uv
2b9d42f0
NIS
1399
1400Adds the UTF8 representation of the Native codepoint C<uv> to the end
1401of the string C<d>; C<d> should be have at least C<UTF8_MAXLEN+1> free
1402bytes available. The return value is the pointer to the byte after the
1403end of the new character. In other words,
1404
1405 d = uvchr_to_utf8(d, uv);
1406
1407is the recommended wide native character-aware way of saying
1408
1409 *(d++) = uv;
1410
1411=cut
1412*/
1413
1414/* On ASCII machines this is normally a macro but we want a
1415 real function in case XS code wants it
1416*/
1417#undef Perl_uvchr_to_utf8
1418U8 *
1419Perl_uvchr_to_utf8(pTHX_ U8 *d, UV uv)
1420{
1421 return Perl_uvuni_to_utf8(aTHX_ d, NATIVE_TO_UNI(uv));
1422}
1423
1424
1425/*
37607a96 1426=for apidoc A|UV|utf8n_to_uvchr|U8 *s|STRLEN curlen|STRLEN *retlen|U32 flags
2b9d42f0
NIS
1427
1428Returns the native character value of the first character in the string C<s>
1429which is assumed to be in UTF8 encoding; C<retlen> will be set to the
1430length, in bytes, of that character.
1431
1432Allows length and flags to be passed to low level routine.
1433
1434=cut
1435*/
1436/* On ASCII machines this is normally a macro but we want a
1437 real function in case XS code wants it
1438*/
1439#undef Perl_utf8n_to_uvchr
1440UV
37607a96 1441Perl_utf8n_to_uvchr(pTHX_ U8 *s, STRLEN curlen, STRLEN *retlen, U32 flags)
2b9d42f0
NIS
1442{
1443 UV uv = Perl_utf8n_to_uvuni(aTHX_ s, curlen, retlen, flags);
1444 return UNI_TO_NATIVE(uv);
1445}
1446
1447