This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Perl_sv_uni_display() needs to be aware of RX_WRAPPED()
[perl5.git] / utf8.c
CommitLineData
a0ed51b3
LW
1/* utf8.c
2 *
1129b882 3 * Copyright (C) 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008
b94e2f88 4 * by Larry Wall and others
a0ed51b3
LW
5 *
6 * You may distribute under the terms of either the GNU General Public
7 * License or the Artistic License, as specified in the README file.
8 *
9 */
10
11/*
4ac71550
TC
12 * 'What a fix!' said Sam. 'That's the one place in all the lands we've ever
13 * heard of that we don't want to see any closer; and that's the one place
14 * we're trying to get to! And that's just where we can't get, nohow.'
15 *
cdad3b53 16 * [p.603 of _The Lord of the Rings_, IV/I: "The Taming of Sméagol"]
a0ed51b3
LW
17 *
18 * 'Well do I understand your speech,' he answered in the same language;
19 * 'yet few strangers do so. Why then do you not speak in the Common Tongue,
4ac71550 20 * as is the custom in the West, if you wish to be answered?'
cdad3b53 21 * --Gandalf, addressing Théoden's door wardens
4ac71550
TC
22 *
23 * [p.508 of _The Lord of the Rings_, III/vi: "The King of the Golden Hall"]
a0ed51b3
LW
24 *
25 * ...the travellers perceived that the floor was paved with stones of many
26 * hues; branching runes and strange devices intertwined beneath their feet.
4ac71550
TC
27 *
28 * [p.512 of _The Lord of the Rings_, III/vi: "The King of the Golden Hall"]
a0ed51b3
LW
29 */
30
31#include "EXTERN.h"
864dbfa3 32#define PERL_IN_UTF8_C
a0ed51b3 33#include "perl.h"
81e983c1 34#include "inline_invlist.c"
a0ed51b3 35
a0c21aa1 36#ifndef EBCDIC
970ea3cb 37/* Separate prototypes needed because in ASCII systems these are
a0c21aa1
JH
38 * usually macros but they still are compiled as code, too. */
39PERL_CALLCONV UV Perl_utf8n_to_uvchr(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen, U32 flags);
6dd9dce9 40PERL_CALLCONV UV Perl_valid_utf8_to_uvchr(pTHX_ const U8 *s, STRLEN *retlen);
a0c21aa1
JH
41PERL_CALLCONV U8* Perl_uvchr_to_utf8(pTHX_ U8 *d, UV uv);
42#endif
43
27da23d5
JH
44static const char unees[] =
45 "Malformed UTF-8 character (unexpected end of string)";
901b21bf 46
48ef279e 47/*
ccfc67b7 48=head1 Unicode Support
a0ed51b3 49
166f8a29
DM
50This file contains various utility functions for manipulating UTF8-encoded
51strings. For the uninitiated, this is a method of representing arbitrary
61296642 52Unicode characters as a variable number of bytes, in such a way that
56da48f7
DM
53characters in the ASCII range are unmodified, and a zero byte never appears
54within non-zero characters.
166f8a29 55
eaf7a4d2
CS
56=cut
57*/
58
59/*
60=for apidoc is_ascii_string
61
a1433954 62Returns true if the first C<len> bytes of the string C<s> are the same whether
970ea3cb
KW
63or not the string is encoded in UTF-8 (or UTF-EBCDIC on EBCDIC machines). That
64is, if they are invariant. On ASCII-ish machines, only ASCII characters
65fit this definition, hence the function's name.
eaf7a4d2 66
9f7e3d64
MH
67If C<len> is 0, it will be calculated using C<strlen(s)>.
68
a1433954 69See also L</is_utf8_string>(), L</is_utf8_string_loclen>(), and L</is_utf8_string_loc>().
eaf7a4d2
CS
70
71=cut
72*/
73
74bool
668b6d8d 75Perl_is_ascii_string(const U8 *s, STRLEN len)
eaf7a4d2
CS
76{
77 const U8* const send = s + (len ? len : strlen((const char *)s));
78 const U8* x = s;
79
80 PERL_ARGS_ASSERT_IS_ASCII_STRING;
eaf7a4d2
CS
81
82 for (; x < send; ++x) {
83 if (!UTF8_IS_INVARIANT(*x))
84 break;
85 }
86
87 return x == send;
88}
89
90/*
87cea99e 91=for apidoc uvuni_to_utf8_flags
eebe1485 92
537124e4 93Adds the UTF-8 representation of the Unicode code point C<uv> to the end
6ee84de2 94of the string C<d>; C<d> should have at least C<UTF8_MAXBYTES+1> free
eebe1485 95bytes available. The return value is the pointer to the byte after the
9041c2e3 96end of the new character. In other words,
eebe1485 97
b851fbc1
JH
98 d = uvuni_to_utf8_flags(d, uv, flags);
99
100or, in most cases,
101
9041c2e3 102 d = uvuni_to_utf8(d, uv);
eebe1485 103
b851fbc1
JH
104(which is equivalent to)
105
106 d = uvuni_to_utf8_flags(d, uv, 0);
107
949cf498 108This is the recommended Unicode-aware way of saying
eebe1485
SC
109
110 *(d++) = uv;
111
537124e4
KW
112where uv is a code point expressed in Latin-1 or above, not the platform's
113native character set. B<Almost all code should instead use L</uvchr_to_utf8>
114or L</uvchr_to_utf8_flags>>.
115
949cf498
KW
116This function will convert to UTF-8 (and not warn) even code points that aren't
117legal Unicode or are problematic, unless C<flags> contains one or more of the
a1433954
KW
118following flags:
119
949cf498
KW
120If C<uv> is a Unicode surrogate code point and UNICODE_WARN_SURROGATE is set,
121the function will raise a warning, provided UTF8 warnings are enabled. If instead
122UNICODE_DISALLOW_SURROGATE is set, the function will fail and return NULL.
123If both flags are set, the function will both warn and return NULL.
124
125The UNICODE_WARN_NONCHAR and UNICODE_DISALLOW_NONCHAR flags correspondingly
537124e4
KW
126affect how the function handles a Unicode non-character. And likewise, the
127UNICODE_WARN_SUPER and UNICODE_DISALLOW_SUPER flags, affect the handling of
128code points that are
949cf498
KW
129above the Unicode maximum of 0x10FFFF. Code points above 0x7FFF_FFFF (which are
130even less portable) can be warned and/or disallowed even if other above-Unicode
131code points are accepted by the UNICODE_WARN_FE_FF and UNICODE_DISALLOW_FE_FF
132flags.
133
134And finally, the flag UNICODE_WARN_ILLEGAL_INTERCHANGE selects all four of the
135above WARN flags; and UNICODE_DISALLOW_ILLEGAL_INTERCHANGE selects all four
136DISALLOW flags.
137
138
eebe1485
SC
139=cut
140*/
141
dfe13c55 142U8 *
b851fbc1 143Perl_uvuni_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags)
a0ed51b3 144{
7918f24d
NC
145 PERL_ARGS_ASSERT_UVUNI_TO_UTF8_FLAGS;
146
979f77b6
KW
147 /* The first problematic code point is the first surrogate */
148 if (uv >= UNICODE_SURROGATE_FIRST
149 && ckWARN4_d(WARN_UTF8, WARN_SURROGATE, WARN_NON_UNICODE, WARN_NONCHAR))
150 {
949cf498
KW
151 if (UNICODE_IS_SURROGATE(uv)) {
152 if (flags & UNICODE_WARN_SURROGATE) {
8457b38f 153 Perl_ck_warner_d(aTHX_ packWARN(WARN_SURROGATE),
949cf498
KW
154 "UTF-16 surrogate U+%04"UVXf, uv);
155 }
156 if (flags & UNICODE_DISALLOW_SURROGATE) {
157 return NULL;
158 }
159 }
160 else if (UNICODE_IS_SUPER(uv)) {
161 if (flags & UNICODE_WARN_SUPER
162 || (UNICODE_IS_FE_FF(uv) && (flags & UNICODE_WARN_FE_FF)))
163 {
8457b38f 164 Perl_ck_warner_d(aTHX_ packWARN(WARN_NON_UNICODE),
949cf498
KW
165 "Code point 0x%04"UVXf" is not Unicode, may not be portable", uv);
166 }
167 if (flags & UNICODE_DISALLOW_SUPER
168 || (UNICODE_IS_FE_FF(uv) && (flags & UNICODE_DISALLOW_FE_FF)))
169 {
170 return NULL;
171 }
172 }
173 else if (UNICODE_IS_NONCHAR(uv)) {
174 if (flags & UNICODE_WARN_NONCHAR) {
8457b38f 175 Perl_ck_warner_d(aTHX_ packWARN(WARN_NONCHAR),
949cf498
KW
176 "Unicode non-character U+%04"UVXf" is illegal for open interchange",
177 uv);
178 }
179 if (flags & UNICODE_DISALLOW_NONCHAR) {
180 return NULL;
181 }
182 }
507b9800 183 }
c4d5f83a 184 if (UNI_IS_INVARIANT(uv)) {
eb160463 185 *d++ = (U8)UTF_TO_NATIVE(uv);
a0ed51b3
LW
186 return d;
187 }
2d331972 188#if defined(EBCDIC)
1d72bdf6
NIS
189 else {
190 STRLEN len = UNISKIP(uv);
191 U8 *p = d+len-1;
192 while (p > d) {
eb160463 193 *p-- = (U8)UTF_TO_NATIVE((uv & UTF_CONTINUATION_MASK) | UTF_CONTINUATION_MARK);
1d72bdf6
NIS
194 uv >>= UTF_ACCUMULATION_SHIFT;
195 }
eb160463 196 *p = (U8)UTF_TO_NATIVE((uv & UTF_START_MASK(len)) | UTF_START_MARK(len));
1d72bdf6
NIS
197 return d+len;
198 }
199#else /* Non loop style */
a0ed51b3 200 if (uv < 0x800) {
eb160463
GS
201 *d++ = (U8)(( uv >> 6) | 0xc0);
202 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
203 return d;
204 }
205 if (uv < 0x10000) {
eb160463
GS
206 *d++ = (U8)(( uv >> 12) | 0xe0);
207 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
208 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
209 return d;
210 }
211 if (uv < 0x200000) {
eb160463
GS
212 *d++ = (U8)(( uv >> 18) | 0xf0);
213 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
214 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
215 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
216 return d;
217 }
218 if (uv < 0x4000000) {
eb160463
GS
219 *d++ = (U8)(( uv >> 24) | 0xf8);
220 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
221 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
222 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
223 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
224 return d;
225 }
226 if (uv < 0x80000000) {
eb160463
GS
227 *d++ = (U8)(( uv >> 30) | 0xfc);
228 *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80);
229 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
230 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
231 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
232 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
233 return d;
234 }
6b8eaf93 235#ifdef HAS_QUAD
d7578b48 236 if (uv < UTF8_QUAD_MAX)
a0ed51b3
LW
237#endif
238 {
eb160463
GS
239 *d++ = 0xfe; /* Can't match U+FEFF! */
240 *d++ = (U8)(((uv >> 30) & 0x3f) | 0x80);
241 *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80);
242 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
243 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
244 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
245 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
246 return d;
247 }
6b8eaf93 248#ifdef HAS_QUAD
a0ed51b3 249 {
eb160463
GS
250 *d++ = 0xff; /* Can't match U+FFFE! */
251 *d++ = 0x80; /* 6 Reserved bits */
252 *d++ = (U8)(((uv >> 60) & 0x0f) | 0x80); /* 2 Reserved bits */
253 *d++ = (U8)(((uv >> 54) & 0x3f) | 0x80);
254 *d++ = (U8)(((uv >> 48) & 0x3f) | 0x80);
255 *d++ = (U8)(((uv >> 42) & 0x3f) | 0x80);
256 *d++ = (U8)(((uv >> 36) & 0x3f) | 0x80);
257 *d++ = (U8)(((uv >> 30) & 0x3f) | 0x80);
258 *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80);
259 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
260 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
261 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
262 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
263 return d;
264 }
265#endif
537124e4 266#endif /* Non loop style */
a0ed51b3 267}
9041c2e3 268
646ca15d
JH
269/*
270
f7d739d1 271Tests if the first C<len> bytes of string C<s> form a valid UTF-8
646ca15d 272character. Note that an INVARIANT (i.e. ASCII) character is a valid
f7d739d1 273UTF-8 character. The number of bytes in the UTF-8 character
646ca15d
JH
274will be returned if it is valid, otherwise 0.
275
276This is the "slow" version as opposed to the "fast" version which is
277the "unrolled" IS_UTF8_CHAR(). E.g. for t/uni/class.t the speed
278difference is a factor of 2 to 3. For lengths (UTF8SKIP(s)) of four
279or less you should use the IS_UTF8_CHAR(), for lengths of five or more
280you should use the _slow(). In practice this means that the _slow()
281will be used very rarely, since the maximum Unicode code point (as of
282Unicode 4.1) is U+10FFFF, which encodes in UTF-8 to four bytes. Only
537124e4 283the "Perl extended UTF-8" (e.g, the infamous 'v-strings') will encode into
646ca15d
JH
284five bytes or more.
285
286=cut */
7af276bc 287PERL_STATIC_INLINE STRLEN
5f66b61c 288S_is_utf8_char_slow(const U8 *s, const STRLEN len)
646ca15d 289{
cd7e6c88 290 dTHX; /* The function called below requires thread context */
646ca15d 291
cd7e6c88 292 STRLEN actual_len;
646ca15d 293
cd7e6c88 294 PERL_ARGS_ASSERT_IS_UTF8_CHAR_SLOW;
646ca15d 295
cd7e6c88 296 utf8n_to_uvuni(s, len, &actual_len, UTF8_CHECK_ONLY);
646ca15d 297
cd7e6c88 298 return (actual_len == (STRLEN) -1) ? 0 : actual_len;
646ca15d 299}
9041c2e3
NIS
300
301/*
492a624f
KW
302=for apidoc is_utf8_char_buf
303
304Returns the number of bytes that comprise the first UTF-8 encoded character in
305buffer C<buf>. C<buf_end> should point to one position beyond the end of the
306buffer. 0 is returned if C<buf> does not point to a complete, valid UTF-8
307encoded character.
308
309Note that an INVARIANT character (i.e. ASCII on non-EBCDIC
310machines) is a valid UTF-8 character.
311
312=cut */
313
314STRLEN
315Perl_is_utf8_char_buf(const U8 *buf, const U8* buf_end)
316{
317
318 STRLEN len;
319
320 PERL_ARGS_ASSERT_IS_UTF8_CHAR_BUF;
321
322 if (buf_end <= buf) {
323 return 0;
324 }
325
326 len = buf_end - buf;
327 if (len > UTF8SKIP(buf)) {
328 len = UTF8SKIP(buf);
329 }
330
331#ifdef IS_UTF8_CHAR
332 if (IS_UTF8_CHAR_FAST(len))
333 return IS_UTF8_CHAR(buf, len) ? len : 0;
334#endif /* #ifdef IS_UTF8_CHAR */
335 return is_utf8_char_slow(buf, len);
336}
337
338/*
87cea99e 339=for apidoc is_utf8_char
eebe1485 340
76848387
KW
341DEPRECATED!
342
5da9da9e 343Tests if some arbitrary number of bytes begins in a valid UTF-8
2bbc8d55
SP
344character. Note that an INVARIANT (i.e. ASCII on non-EBCDIC machines)
345character is a valid UTF-8 character. The actual number of bytes in the UTF-8
346character will be returned if it is valid, otherwise 0.
9041c2e3 347
76848387 348This function is deprecated due to the possibility that malformed input could
a1433954 349cause reading beyond the end of the input buffer. Use L</is_utf8_char_buf>
76848387 350instead.
e0328548 351
82686b01 352=cut */
76848387 353
067a85ef 354STRLEN
668b6d8d 355Perl_is_utf8_char(const U8 *s)
386d01d6 356{
7918f24d 357 PERL_ARGS_ASSERT_IS_UTF8_CHAR;
492a624f 358
76848387 359 /* Assumes we have enough space, which is why this is deprecated */
492a624f 360 return is_utf8_char_buf(s, s + UTF8SKIP(s));
386d01d6
GS
361}
362
eaf7a4d2 363
6662521e 364/*
87cea99e 365=for apidoc is_utf8_string
6662521e 366
a1433954 367Returns true if the first C<len> bytes of string C<s> form a valid
9f7e3d64 368UTF-8 string, false otherwise. If C<len> is 0, it will be calculated
e0328548
KW
369using C<strlen(s)> (which means if you use this option, that C<s> has to have a
370terminating NUL byte). Note that all characters being ASCII constitute 'a
371valid UTF-8 string'.
6662521e 372
a1433954 373See also L</is_ascii_string>(), L</is_utf8_string_loclen>(), and L</is_utf8_string_loc>().
768c67ee 374
6662521e
GS
375=cut
376*/
377
8e84507e 378bool
668b6d8d 379Perl_is_utf8_string(const U8 *s, STRLEN len)
6662521e 380{
35da51f7 381 const U8* const send = s + (len ? len : strlen((const char *)s));
7fc63493 382 const U8* x = s;
067a85ef 383
7918f24d 384 PERL_ARGS_ASSERT_IS_UTF8_STRING;
1aa99e6b 385
6662521e 386 while (x < send) {
1acdb0da 387 /* Inline the easy bits of is_utf8_char() here for speed... */
e0328548
KW
388 if (UTF8_IS_INVARIANT(*x)) {
389 x++;
390 }
1acdb0da
JH
391 else {
392 /* ... and call is_utf8_char() only if really needed. */
e0328548
KW
393 const STRLEN c = UTF8SKIP(x);
394 const U8* const next_char_ptr = x + c;
395
396 if (next_char_ptr > send) {
397 return FALSE;
398 }
399
768c67ee
JH
400 if (IS_UTF8_CHAR_FAST(c)) {
401 if (!IS_UTF8_CHAR(x, c))
e0328548 402 return FALSE;
3c614e38 403 }
e0328548
KW
404 else if (! is_utf8_char_slow(x, c)) {
405 return FALSE;
406 }
407 x = next_char_ptr;
1acdb0da 408 }
6662521e 409 }
768c67ee 410
067a85ef 411 return TRUE;
6662521e
GS
412}
413
67e989fb 414/*
814fafa7
NC
415Implemented as a macro in utf8.h
416
87cea99e 417=for apidoc is_utf8_string_loc
814fafa7 418
a1433954
KW
419Like L</is_utf8_string> but stores the location of the failure (in the
420case of "utf8ness failure") or the location C<s>+C<len> (in the case of
814fafa7
NC
421"utf8ness success") in the C<ep>.
422
a1433954 423See also L</is_utf8_string_loclen>() and L</is_utf8_string>().
814fafa7 424
87cea99e 425=for apidoc is_utf8_string_loclen
81cd54e3 426
a1433954
KW
427Like L</is_utf8_string>() but stores the location of the failure (in the
428case of "utf8ness failure") or the location C<s>+C<len> (in the case of
768c67ee
JH
429"utf8ness success") in the C<ep>, and the number of UTF-8
430encoded characters in the C<el>.
431
a1433954 432See also L</is_utf8_string_loc>() and L</is_utf8_string>().
81cd54e3
JH
433
434=cut
435*/
436
437bool
668b6d8d 438Perl_is_utf8_string_loclen(const U8 *s, STRLEN len, const U8 **ep, STRLEN *el)
81cd54e3 439{
35da51f7 440 const U8* const send = s + (len ? len : strlen((const char *)s));
7fc63493 441 const U8* x = s;
81cd54e3 442 STRLEN c;
3ebfea28 443 STRLEN outlen = 0;
7918f24d
NC
444
445 PERL_ARGS_ASSERT_IS_UTF8_STRING_LOCLEN;
81cd54e3 446
81cd54e3 447 while (x < send) {
e0328548
KW
448 const U8* next_char_ptr;
449
81cd54e3
JH
450 /* Inline the easy bits of is_utf8_char() here for speed... */
451 if (UTF8_IS_INVARIANT(*x))
e0328548 452 next_char_ptr = x + 1;
81cd54e3 453 else {
768c67ee 454 /* ... and call is_utf8_char() only if really needed. */
768c67ee 455 c = UTF8SKIP(x);
e0328548
KW
456 next_char_ptr = c + x;
457 if (next_char_ptr > send) {
458 goto out;
459 }
768c67ee
JH
460 if (IS_UTF8_CHAR_FAST(c)) {
461 if (!IS_UTF8_CHAR(x, c))
462 c = 0;
463 } else
464 c = is_utf8_char_slow(x, c);
768c67ee
JH
465 if (!c)
466 goto out;
81cd54e3 467 }
e0328548 468 x = next_char_ptr;
3ebfea28 469 outlen++;
81cd54e3 470 }
768c67ee
JH
471
472 out:
3ebfea28
AL
473 if (el)
474 *el = outlen;
475
768c67ee
JH
476 if (ep)
477 *ep = x;
3ebfea28 478 return (x == send);
81cd54e3
JH
479}
480
481/*
768c67ee 482
87cea99e 483=for apidoc utf8n_to_uvuni
67e989fb 484
9041c2e3 485Bottom level UTF-8 decode routine.
746afd53
KW
486Returns the code point value of the first character in the string C<s>,
487which is assumed to be in UTF-8 (or UTF-EBCDIC) encoding, and no longer than
488C<curlen> bytes; C<*retlen> (if C<retlen> isn't NULL) will be set to
489the length, in bytes, of that character.
949cf498
KW
490
491The value of C<flags> determines the behavior when C<s> does not point to a
492well-formed UTF-8 character. If C<flags> is 0, when a malformation is found,
524080c4
KW
493zero is returned and C<*retlen> is set so that (S<C<s> + C<*retlen>>) is the
494next possible position in C<s> that could begin a non-malformed character.
495Also, if UTF-8 warnings haven't been lexically disabled, a warning is raised.
949cf498
KW
496
497Various ALLOW flags can be set in C<flags> to allow (and not warn on)
498individual types of malformations, such as the sequence being overlong (that
499is, when there is a shorter sequence that can express the same code point;
500overlong sequences are expressly forbidden in the UTF-8 standard due to
501potential security issues). Another malformation example is the first byte of
502a character not being a legal first byte. See F<utf8.h> for the list of such
524080c4
KW
503flags. For allowed 0 length strings, this function returns 0; for allowed
504overlong sequences, the computed code point is returned; for all other allowed
505malformations, the Unicode REPLACEMENT CHARACTER is returned, as these have no
506determinable reasonable value.
949cf498
KW
507
508The UTF8_CHECK_ONLY flag overrides the behavior when a non-allowed (by other
509flags) malformation is found. If this flag is set, the routine assumes that
510the caller will raise a warning, and this function will silently just set
d088425d
KW
511C<retlen> to C<-1> (cast to C<STRLEN>) and return zero.
512
513Note that this API requires disambiguation between successful decoding a NUL
514character, and an error return (unless the UTF8_CHECK_ONLY flag is set), as
515in both cases, 0 is returned. To disambiguate, upon a zero return, see if the
516first byte of C<s> is 0 as well. If so, the input was a NUL; if not, the input
517had an error.
949cf498
KW
518
519Certain code points are considered problematic. These are Unicode surrogates,
746afd53 520Unicode non-characters, and code points above the Unicode maximum of 0x10FFFF.
949cf498 521By default these are considered regular code points, but certain situations
5eafe189 522warrant special handling for them. If C<flags> contains
949cf498
KW
523UTF8_DISALLOW_ILLEGAL_INTERCHANGE, all three classes are treated as
524malformations and handled as such. The flags UTF8_DISALLOW_SURROGATE,
525UTF8_DISALLOW_NONCHAR, and UTF8_DISALLOW_SUPER (meaning above the legal Unicode
526maximum) can be set to disallow these categories individually.
527
528The flags UTF8_WARN_ILLEGAL_INTERCHANGE, UTF8_WARN_SURROGATE,
529UTF8_WARN_NONCHAR, and UTF8_WARN_SUPER will cause warning messages to be raised
530for their respective categories, but otherwise the code points are considered
531valid (not malformations). To get a category to both be treated as a
532malformation and raise a warning, specify both the WARN and DISALLOW flags.
533(But note that warnings are not raised if lexically disabled nor if
534UTF8_CHECK_ONLY is also specified.)
535
536Very large code points (above 0x7FFF_FFFF) are considered more problematic than
537the others that are above the Unicode legal maximum. There are several
eb83ed87
KW
538reasons: they requre at least 32 bits to represent them on ASCII platforms, are
539not representable at all on EBCDIC platforms, and the original UTF-8
540specification never went above this number (the current 0x10FFFF limit was
541imposed later). (The smaller ones, those that fit into 32 bits, are
542representable by a UV on ASCII platforms, but not by an IV, which means that
543the number of operations that can be performed on them is quite restricted.)
544The UTF-8 encoding on ASCII platforms for these large code points begins with a
545byte containing 0xFE or 0xFF. The UTF8_DISALLOW_FE_FF flag will cause them to
546be treated as malformations, while allowing smaller above-Unicode code points.
547(Of course UTF8_DISALLOW_SUPER will treat all above-Unicode code points,
548including these, as malformations.) Similarly, UTF8_WARN_FE_FF acts just like
549the other WARN flags, but applies just to these code points.
949cf498
KW
550
551All other code points corresponding to Unicode characters, including private
552use and those yet to be assigned, are never considered malformed and never
553warn.
67e989fb 554
ec5f19d0 555Most code should use L</utf8_to_uvchr_buf>() rather than call this directly.
9041c2e3 556
37607a96
PK
557=cut
558*/
67e989fb 559
a0ed51b3 560UV
7fc63493 561Perl_utf8n_to_uvuni(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen, U32 flags)
a0ed51b3 562{
97aff369 563 dVAR;
d4c19fe8 564 const U8 * const s0 = s;
eb83ed87 565 U8 overflow_byte = '\0'; /* Save byte in case of overflow */
0b8d30e8 566 U8 * send;
eb83ed87
KW
567 UV uv = *s;
568 STRLEN expectlen;
949cf498 569 SV* sv = NULL;
eb83ed87
KW
570 UV outlier_ret = 0; /* return value when input is in error or problematic
571 */
572 UV pack_warn = 0; /* Save result of packWARN() for later */
573 bool unexpected_non_continuation = FALSE;
574 bool overflowed = FALSE;
2f8f112e 575 bool do_overlong_test = TRUE; /* May have to skip this test */
a0dbb045 576
eb83ed87 577 const char* const malformed_text = "Malformed UTF-8 character";
7918f24d 578
eb83ed87 579 PERL_ARGS_ASSERT_UTF8N_TO_UVUNI;
a0dbb045 580
eb83ed87
KW
581 /* The order of malformation tests here is important. We should consume as
582 * few bytes as possible in order to not skip any valid character. This is
583 * required by the Unicode Standard (section 3.9 of Unicode 6.0); see also
584 * http://unicode.org/reports/tr36 for more discussion as to why. For
585 * example, once we've done a UTF8SKIP, we can tell the expected number of
586 * bytes, and could fail right off the bat if the input parameters indicate
587 * that there are too few available. But it could be that just that first
588 * byte is garbled, and the intended character occupies fewer bytes. If we
589 * blindly assumed that the first byte is correct, and skipped based on
590 * that number, we could skip over a valid input character. So instead, we
591 * always examine the sequence byte-by-byte.
592 *
593 * We also should not consume too few bytes, otherwise someone could inject
594 * things. For example, an input could be deliberately designed to
595 * overflow, and if this code bailed out immediately upon discovering that,
596 * returning to the caller *retlen pointing to the very next byte (one
597 * which is actually part of of the overflowing sequence), that could look
598 * legitimate to the caller, which could discard the initial partial
599 * sequence and process the rest, inappropriately */
600
601 /* Zero length strings, if allowed, of necessity are zero */
b5b9af04 602 if (UNLIKELY(curlen == 0)) {
eb83ed87
KW
603 if (retlen) {
604 *retlen = 0;
605 }
a0dbb045 606
eb83ed87
KW
607 if (flags & UTF8_ALLOW_EMPTY) {
608 return 0;
609 }
610 if (! (flags & UTF8_CHECK_ONLY)) {
611 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (empty string)", malformed_text));
612 }
0c443dc2
JH
613 goto malformed;
614 }
615
eb83ed87
KW
616 expectlen = UTF8SKIP(s);
617
618 /* A well-formed UTF-8 character, as the vast majority of calls to this
619 * function will be for, has this expected length. For efficiency, set
620 * things up here to return it. It will be overriden only in those rare
621 * cases where a malformation is found */
622 if (retlen) {
623 *retlen = expectlen;
624 }
625
626 /* An invariant is trivially well-formed */
1d72bdf6 627 if (UTF8_IS_INVARIANT(uv)) {
c4d5f83a 628 return (UV) (NATIVE_TO_UTF(*s));
a0ed51b3 629 }
67e989fb 630
eb83ed87 631 /* A continuation character can't start a valid sequence */
b5b9af04 632 if (UNLIKELY(UTF8_IS_CONTINUATION(uv))) {
eb83ed87
KW
633 if (flags & UTF8_ALLOW_CONTINUATION) {
634 if (retlen) {
635 *retlen = 1;
636 }
637 return UNICODE_REPLACEMENT;
638 }
ba210ebe 639
eb83ed87
KW
640 if (! (flags & UTF8_CHECK_ONLY)) {
641 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (unexpected continuation byte 0x%02x, with no preceding start byte)", malformed_text, *s0));
642 }
643 curlen = 1;
ba210ebe
JH
644 goto malformed;
645 }
9041c2e3 646
1d72bdf6 647#ifdef EBCDIC
75383841 648 uv = NATIVE_TO_UTF(uv);
1d72bdf6
NIS
649#endif
650
eb83ed87
KW
651 /* Here is not a continuation byte, nor an invariant. The only thing left
652 * is a start byte (possibly for an overlong) */
ba210ebe 653
eb83ed87
KW
654 /* Remove the leading bits that indicate the number of bytes in the
655 * character's whole UTF-8 sequence, leaving just the bits that are part of
656 * the value */
657 uv &= UTF_START_MASK(expectlen);
ba210ebe 658
eb83ed87
KW
659 /* Now, loop through the remaining bytes in the character's sequence,
660 * accumulating each into the working value as we go. Be sure to not look
661 * past the end of the input string */
0b8d30e8
KW
662 send = (U8*) s0 + ((expectlen <= curlen) ? expectlen : curlen);
663
eb83ed87 664 for (s = s0 + 1; s < send; s++) {
b5b9af04 665 if (LIKELY(UTF8_IS_CONTINUATION(*s))) {
eb83ed87
KW
666#ifndef EBCDIC /* Can't overflow in EBCDIC */
667 if (uv & UTF_ACCUMULATION_OVERFLOW_MASK) {
668
669 /* The original implementors viewed this malformation as more
670 * serious than the others (though I, khw, don't understand
671 * why, since other malformations also give very very wrong
672 * results), so there is no way to turn off checking for it.
673 * Set a flag, but keep going in the loop, so that we absorb
674 * the rest of the bytes that comprise the character. */
675 overflowed = TRUE;
676 overflow_byte = *s; /* Save for warning message's use */
677 }
678#endif
8850bf83 679 uv = UTF8_ACCUMULATE(uv, *s);
eb83ed87
KW
680 }
681 else {
682 /* Here, found a non-continuation before processing all expected
683 * bytes. This byte begins a new character, so quit, even if
684 * allowing this malformation. */
685 unexpected_non_continuation = TRUE;
686 break;
687 }
688 } /* End of loop through the character's bytes */
689
690 /* Save how many bytes were actually in the character */
691 curlen = s - s0;
692
693 /* The loop above finds two types of malformations: non-continuation and/or
694 * overflow. The non-continuation malformation is really a too-short
695 * malformation, as it means that the current character ended before it was
696 * expected to (being terminated prematurely by the beginning of the next
697 * character, whereas in the too-short malformation there just are too few
698 * bytes available to hold the character. In both cases, the check below
699 * that we have found the expected number of bytes would fail if executed.)
700 * Thus the non-continuation malformation is really unnecessary, being a
701 * subset of the too-short malformation. But there may be existing
702 * applications that are expecting the non-continuation type, so we retain
703 * it, and return it in preference to the too-short malformation. (If this
704 * code were being written from scratch, the two types might be collapsed
705 * into one.) I, khw, am also giving priority to returning the
706 * non-continuation and too-short malformations over overflow when multiple
707 * ones are present. I don't know of any real reason to prefer one over
708 * the other, except that it seems to me that multiple-byte errors trumps
709 * errors from a single byte */
b5b9af04 710 if (UNLIKELY(unexpected_non_continuation)) {
eb83ed87
KW
711 if (!(flags & UTF8_ALLOW_NON_CONTINUATION)) {
712 if (! (flags & UTF8_CHECK_ONLY)) {
713 if (curlen == 1) {
714 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (unexpected non-continuation byte 0x%02x, immediately after start byte 0x%02x)", malformed_text, *s, *s0));
715 }
716 else {
717 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (unexpected non-continuation byte 0x%02x, %d bytes after start byte 0x%02x, expected %d bytes)", malformed_text, *s, (int) curlen, *s0, (int)expectlen));
a0dbb045
JH
718 }
719 }
eb83ed87
KW
720 goto malformed;
721 }
722 uv = UNICODE_REPLACEMENT;
2f8f112e
KW
723
724 /* Skip testing for overlongs, as the REPLACEMENT may not be the same
725 * as what the original expectations were. */
726 do_overlong_test = FALSE;
eb83ed87
KW
727 if (retlen) {
728 *retlen = curlen;
729 }
730 }
b5b9af04 731 else if (UNLIKELY(curlen < expectlen)) {
eb83ed87
KW
732 if (! (flags & UTF8_ALLOW_SHORT)) {
733 if (! (flags & UTF8_CHECK_ONLY)) {
734 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (%d byte%s, need %d, after start byte 0x%02x)", malformed_text, (int)curlen, curlen == 1 ? "" : "s", (int)expectlen, *s0));
a0dbb045 735 }
eb83ed87
KW
736 goto malformed;
737 }
738 uv = UNICODE_REPLACEMENT;
2f8f112e 739 do_overlong_test = FALSE;
eb83ed87
KW
740 if (retlen) {
741 *retlen = curlen;
742 }
743 }
744
745#ifndef EBCDIC /* EBCDIC allows FE, FF, can't overflow */
2f8f112e 746 if ((*s0 & 0xFE) == 0xFE /* matches both FE, FF */
eb83ed87
KW
747 && (flags & (UTF8_WARN_FE_FF|UTF8_DISALLOW_FE_FF)))
748 {
749 /* By adding UTF8_CHECK_ONLY to the test, we avoid unnecessary
750 * generation of the sv, since no warnings are raised under CHECK */
751 if ((flags & (UTF8_WARN_FE_FF|UTF8_CHECK_ONLY)) == UTF8_WARN_FE_FF
752 && ckWARN_d(WARN_UTF8))
753 {
42303544
KW
754 /* This message is deliberately not of the same syntax as the other
755 * messages for malformations, for backwards compatibility in the
756 * unlikely event that code is relying on its precise earlier text
757 */
eb83ed87
KW
758 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s Code point beginning with byte 0x%02X is not Unicode, and not portable", malformed_text, *s0));
759 pack_warn = packWARN(WARN_UTF8);
760 }
761 if (flags & UTF8_DISALLOW_FE_FF) {
762 goto malformed;
ba210ebe 763 }
ba210ebe 764 }
b5b9af04 765 if (UNLIKELY(overflowed)) {
ba210ebe 766
eb83ed87
KW
767 /* If the first byte is FF, it will overflow a 32-bit word. If the
768 * first byte is FE, it will overflow a signed 32-bit word. The
769 * above preserves backward compatibility, since its message was used
770 * in earlier versions of this code in preference to overflow */
771 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (overflow at byte 0x%02x, after start byte 0x%02x)", malformed_text, overflow_byte, *s0));
ba210ebe 772 goto malformed;
eb83ed87
KW
773 }
774#endif
775
2f8f112e
KW
776 if (do_overlong_test
777 && expectlen > (STRLEN)UNISKIP(uv)
778 && ! (flags & UTF8_ALLOW_LONG))
779 {
eb83ed87
KW
780 /* The overlong malformation has lower precedence than the others.
781 * Note that if this malformation is allowed, we return the actual
782 * value, instead of the replacement character. This is because this
783 * value is actually well-defined. */
784 if (! (flags & UTF8_CHECK_ONLY)) {
785 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (%d byte%s, need %d, after start byte 0x%02x)", malformed_text, (int)expectlen, expectlen == 1 ? "": "s", UNISKIP(uv), *s0));
786 }
787 goto malformed;
788 }
789
790 /* Here, the input is considered to be well-formed , but could be a
791 * problematic code point that is not allowed by the input parameters. */
792 if (uv >= UNICODE_SURROGATE_FIRST /* isn't problematic if < this */
793 && (flags & (UTF8_DISALLOW_ILLEGAL_INTERCHANGE
794 |UTF8_WARN_ILLEGAL_INTERCHANGE)))
795 {
949cf498 796 if (UNICODE_IS_SURROGATE(uv)) {
eb83ed87
KW
797 if ((flags & (UTF8_WARN_SURROGATE|UTF8_CHECK_ONLY)) == UTF8_WARN_SURROGATE
798 && ckWARN2_d(WARN_UTF8, WARN_SURROGATE))
799 {
111d382d 800 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "UTF-16 surrogate U+%04"UVXf"", uv));
eb83ed87 801 pack_warn = packWARN2(WARN_UTF8, WARN_SURROGATE);
949cf498
KW
802 }
803 if (flags & UTF8_DISALLOW_SURROGATE) {
804 goto disallowed;
805 }
806 }
949cf498 807 else if ((uv > PERL_UNICODE_MAX)) {
eb83ed87
KW
808 if ((flags & (UTF8_WARN_SUPER|UTF8_CHECK_ONLY)) == UTF8_WARN_SUPER
809 && ckWARN2_d(WARN_UTF8, WARN_NON_UNICODE))
810 {
111d382d 811 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "Code point 0x%04"UVXf" is not Unicode, may not be portable", uv));
eb83ed87 812 pack_warn = packWARN2(WARN_UTF8, WARN_NON_UNICODE);
949cf498
KW
813 }
814 if (flags & UTF8_DISALLOW_SUPER) {
815 goto disallowed;
816 }
817 }
4190d317
KW
818 else if (UNICODE_IS_NONCHAR(uv)) {
819 if ((flags & (UTF8_WARN_NONCHAR|UTF8_CHECK_ONLY)) == UTF8_WARN_NONCHAR
820 && ckWARN2_d(WARN_UTF8, WARN_NONCHAR))
821 {
822 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "Unicode non-character U+%04"UVXf" is illegal for open interchange", uv));
823 pack_warn = packWARN2(WARN_UTF8, WARN_NONCHAR);
824 }
825 if (flags & UTF8_DISALLOW_NONCHAR) {
826 goto disallowed;
827 }
828 }
949cf498 829
eb83ed87
KW
830 if (sv) {
831 outlier_ret = uv;
832 goto do_warn;
833 }
834
949cf498
KW
835 /* Here, this is not considered a malformed character, so drop through
836 * to return it */
a0ed51b3 837 }
ba210ebe 838
a0ed51b3 839 return uv;
ba210ebe 840
eb83ed87
KW
841 /* There are three cases which get to beyond this point. In all 3 cases:
842 * <sv> if not null points to a string to print as a warning.
843 * <curlen> is what <*retlen> should be set to if UTF8_CHECK_ONLY isn't
844 * set.
845 * <outlier_ret> is what return value to use if UTF8_CHECK_ONLY isn't set.
846 * This is done by initializing it to 0, and changing it only
847 * for case 1).
848 * The 3 cases are:
849 * 1) The input is valid but problematic, and to be warned about. The
850 * return value is the resultant code point; <*retlen> is set to
851 * <curlen>, the number of bytes that comprise the code point.
852 * <pack_warn> contains the result of packWARN() for the warning
853 * types. The entry point for this case is the label <do_warn>;
854 * 2) The input is a valid code point but disallowed by the parameters to
855 * this function. The return value is 0. If UTF8_CHECK_ONLY is set,
856 * <*relen> is -1; otherwise it is <curlen>, the number of bytes that
857 * comprise the code point. <pack_warn> contains the result of
858 * packWARN() for the warning types. The entry point for this case is
859 * the label <disallowed>.
860 * 3) The input is malformed. The return value is 0. If UTF8_CHECK_ONLY
861 * is set, <*relen> is -1; otherwise it is <curlen>, the number of
862 * bytes that comprise the malformation. All such malformations are
863 * assumed to be warning type <utf8>. The entry point for this case
864 * is the label <malformed>.
865 */
949cf498 866
ba210ebe
JH
867malformed:
868
eb83ed87
KW
869 if (sv && ckWARN_d(WARN_UTF8)) {
870 pack_warn = packWARN(WARN_UTF8);
871 }
872
873disallowed:
874
fcc8fcf6 875 if (flags & UTF8_CHECK_ONLY) {
ba210ebe 876 if (retlen)
10edeb5d 877 *retlen = ((STRLEN) -1);
ba210ebe
JH
878 return 0;
879 }
880
eb83ed87 881do_warn:
5b311467 882
eb83ed87
KW
883 if (pack_warn) { /* <pack_warn> was initialized to 0, and changed only
884 if warnings are to be raised. */
f555bc63 885 const char * const string = SvPVX_const(sv);
a0dbb045 886
f555bc63
KW
887 if (PL_op)
888 Perl_warner(aTHX_ pack_warn, "%s in %s", string, OP_DESC(PL_op));
889 else
890 Perl_warner(aTHX_ pack_warn, "%s", string);
a0dbb045
JH
891 }
892
eb83ed87
KW
893 if (retlen) {
894 *retlen = curlen;
895 }
ba210ebe 896
eb83ed87 897 return outlier_ret;
a0ed51b3
LW
898}
899
8e84507e 900/*
ec5f19d0
KW
901=for apidoc utf8_to_uvchr_buf
902
903Returns the native code point of the first character in the string C<s> which
904is assumed to be in UTF-8 encoding; C<send> points to 1 beyond the end of C<s>.
524080c4 905C<*retlen> will be set to the length, in bytes, of that character.
ec5f19d0 906
524080c4
KW
907If C<s> does not point to a well-formed UTF-8 character and UTF8 warnings are
908enabled, zero is returned and C<*retlen> is set (if C<retlen> isn't
173db420
KW
909NULL) to -1. If those warnings are off, the computed value, if well-defined
910(or the Unicode REPLACEMENT CHARACTER if not), is silently returned, and
911C<*retlen> is set (if C<retlen> isn't NULL) so that (S<C<s> + C<*retlen>>) is
912the next possible position in C<s> that could begin a non-malformed character.
913See L</utf8n_to_uvuni> for details on when the REPLACEMENT CHARACTER is
914returned.
ec5f19d0
KW
915
916=cut
917*/
918
919
920UV
921Perl_utf8_to_uvchr_buf(pTHX_ const U8 *s, const U8 *send, STRLEN *retlen)
922{
923 PERL_ARGS_ASSERT_UTF8_TO_UVCHR_BUF;
924
925 assert(s < send);
926
927 return utf8n_to_uvchr(s, send - s, retlen,
928 ckWARN_d(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
929}
930
27d6c58a 931/* Like L</utf8_to_uvchr_buf>(), but should only be called when it is known that
3986bb7c 932 * there are no malformations in the input UTF-8 string C<s>. surrogates,
6dd9dce9
KW
933 * non-character code points, and non-Unicode code points are allowed. A macro
934 * in utf8.h is used to normally avoid this function wrapper */
27d6c58a
KW
935
936UV
937Perl_valid_utf8_to_uvchr(pTHX_ const U8 *s, STRLEN *retlen)
938{
3986bb7c
KW
939 const UV uv = valid_utf8_to_uvuni(s, retlen);
940
27d6c58a
KW
941 PERL_ARGS_ASSERT_VALID_UTF8_TO_UVCHR;
942
3986bb7c 943 return UNI_TO_NATIVE(uv);
27d6c58a
KW
944}
945
ec5f19d0 946/*
87cea99e 947=for apidoc utf8_to_uvchr
9041c2e3 948
977c1d31
KW
949DEPRECATED!
950
6ee84de2 951Returns the native code point of the first character in the string C<s>
1e54db1a 952which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
9041c2e3
NIS
953length, in bytes, of that character.
954
4b88fb76 955Some, but not all, UTF-8 malformations are detected, and in fact, some
977c1d31
KW
956malformed input could cause reading beyond the end of the input buffer, which
957is why this function is deprecated. Use L</utf8_to_uvchr_buf> instead.
4b88fb76 958
524080c4
KW
959If C<s> points to one of the detected malformations, and UTF8 warnings are
960enabled, zero is returned and C<*retlen> is set (if C<retlen> isn't
961NULL) to -1. If those warnings are off, the computed value if well-defined (or
962the Unicode REPLACEMENT CHARACTER, if not) is silently returned, and C<*retlen>
963is set (if C<retlen> isn't NULL) so that (S<C<s> + C<*retlen>>) is the
964next possible position in C<s> that could begin a non-malformed character.
965See L</utf8n_to_uvuni> for details on when the REPLACEMENT CHARACTER is returned.
9041c2e3
NIS
966
967=cut
968*/
969
970UV
7fc63493 971Perl_utf8_to_uvchr(pTHX_ const U8 *s, STRLEN *retlen)
9041c2e3 972{
7918f24d
NC
973 PERL_ARGS_ASSERT_UTF8_TO_UVCHR;
974
2ff6c191 975 return utf8_to_uvchr_buf(s, s + UTF8_MAXBYTES, retlen);
9041c2e3
NIS
976}
977
978/*
ec5f19d0
KW
979=for apidoc utf8_to_uvuni_buf
980
981Returns the Unicode code point of the first character in the string C<s> which
982is assumed to be in UTF-8 encoding; C<send> points to 1 beyond the end of C<s>.
983C<retlen> will be set to the length, in bytes, of that character.
984
985This function should only be used when the returned UV is considered
986an index into the Unicode semantic tables (e.g. swashes).
987
524080c4
KW
988If C<s> does not point to a well-formed UTF-8 character and UTF8 warnings are
989enabled, zero is returned and C<*retlen> is set (if C<retlen> isn't
990NULL) to -1. If those warnings are off, the computed value if well-defined (or
991the Unicode REPLACEMENT CHARACTER, if not) is silently returned, and C<*retlen>
992is set (if C<retlen> isn't NULL) so that (S<C<s> + C<*retlen>>) is the
993next possible position in C<s> that could begin a non-malformed character.
994See L</utf8n_to_uvuni> for details on when the REPLACEMENT CHARACTER is returned.
ec5f19d0
KW
995
996=cut
997*/
998
999UV
1000Perl_utf8_to_uvuni_buf(pTHX_ const U8 *s, const U8 *send, STRLEN *retlen)
1001{
1002 PERL_ARGS_ASSERT_UTF8_TO_UVUNI_BUF;
1003
1004 assert(send > s);
1005
1006 /* Call the low level routine asking for checks */
1007 return Perl_utf8n_to_uvuni(aTHX_ s, send -s, retlen,
1008 ckWARN_d(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
1009}
1010
27d6c58a 1011/* Like L</utf8_to_uvuni_buf>(), but should only be called when it is known that
2114036c 1012 * there are no malformations in the input UTF-8 string C<s>. Surrogates,
3986bb7c 1013 * non-character code points, and non-Unicode code points are allowed */
27d6c58a
KW
1014
1015UV
1016Perl_valid_utf8_to_uvuni(pTHX_ const U8 *s, STRLEN *retlen)
1017{
3986bb7c
KW
1018 UV expectlen = UTF8SKIP(s);
1019 const U8* send = s + expectlen;
1020 UV uv = NATIVE_TO_UTF(*s);
1021
27d6c58a
KW
1022 PERL_ARGS_ASSERT_VALID_UTF8_TO_UVUNI;
1023
3986bb7c
KW
1024 if (retlen) {
1025 *retlen = expectlen;
1026 }
1027
1028 /* An invariant is trivially returned */
1029 if (expectlen == 1) {
1030 return uv;
1031 }
1032
1033 /* Remove the leading bits that indicate the number of bytes, leaving just
1034 * the bits that are part of the value */
1035 uv &= UTF_START_MASK(expectlen);
1036
1037 /* Now, loop through the remaining bytes, accumulating each into the
08bc774e
KW
1038 * working total as we go. (I khw tried unrolling the loop for up to 4
1039 * bytes, but there was no performance improvement) */
3986bb7c
KW
1040 for (++s; s < send; s++) {
1041 uv = UTF8_ACCUMULATE(uv, *s);
1042 }
1043
1044 return uv;
27d6c58a
KW
1045}
1046
ec5f19d0 1047/*
87cea99e 1048=for apidoc utf8_to_uvuni
9041c2e3 1049
977c1d31
KW
1050DEPRECATED!
1051
9041c2e3 1052Returns the Unicode code point of the first character in the string C<s>
1e54db1a 1053which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
9041c2e3
NIS
1054length, in bytes, of that character.
1055
39e518fd
KW
1056This function should only be used when the returned UV is considered
1057an index into the Unicode semantic tables (e.g. swashes).
1058
4b88fb76 1059Some, but not all, UTF-8 malformations are detected, and in fact, some
977c1d31
KW
1060malformed input could cause reading beyond the end of the input buffer, which
1061is why this function is deprecated. Use L</utf8_to_uvuni_buf> instead.
9041c2e3 1062
524080c4
KW
1063If C<s> points to one of the detected malformations, and UTF8 warnings are
1064enabled, zero is returned and C<*retlen> is set (if C<retlen> doesn't point to
1065NULL) to -1. If those warnings are off, the computed value if well-defined (or
1066the Unicode REPLACEMENT CHARACTER, if not) is silently returned, and C<*retlen>
1067is set (if C<retlen> isn't NULL) so that (S<C<s> + C<*retlen>>) is the
1068next possible position in C<s> that could begin a non-malformed character.
1069See L</utf8n_to_uvuni> for details on when the REPLACEMENT CHARACTER is returned.
8e84507e
NIS
1070
1071=cut
1072*/
1073
1074UV
7fc63493 1075Perl_utf8_to_uvuni(pTHX_ const U8 *s, STRLEN *retlen)
8e84507e 1076{
7918f24d
NC
1077 PERL_ARGS_ASSERT_UTF8_TO_UVUNI;
1078
4b88fb76 1079 return valid_utf8_to_uvuni(s, retlen);
8e84507e
NIS
1080}
1081
b76347f2 1082/*
87cea99e 1083=for apidoc utf8_length
b76347f2
JH
1084
1085Return the length of the UTF-8 char encoded string C<s> in characters.
02eb7b47
JH
1086Stops at C<e> (inclusive). If C<e E<lt> s> or if the scan would end
1087up past C<e>, croaks.
b76347f2
JH
1088
1089=cut
1090*/
1091
1092STRLEN
35a4481c 1093Perl_utf8_length(pTHX_ const U8 *s, const U8 *e)
b76347f2 1094{
97aff369 1095 dVAR;
b76347f2
JH
1096 STRLEN len = 0;
1097
7918f24d
NC
1098 PERL_ARGS_ASSERT_UTF8_LENGTH;
1099
8850bf83
JH
1100 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g.
1101 * the bitops (especially ~) can create illegal UTF-8.
1102 * In other words: in Perl UTF-8 is not just for Unicode. */
1103
a3b680e6
AL
1104 if (e < s)
1105 goto warn_and_return;
b76347f2 1106 while (s < e) {
4cbf4130 1107 s += UTF8SKIP(s);
8e91ec7f
AV
1108 len++;
1109 }
1110
1111 if (e != s) {
1112 len--;
1113 warn_and_return:
9b387841
NC
1114 if (PL_op)
1115 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8),
1116 "%s in %s", unees, OP_DESC(PL_op));
1117 else
61a12c31 1118 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), "%s", unees);
b76347f2
JH
1119 }
1120
1121 return len;
1122}
1123
b06226ff 1124/*
87cea99e 1125=for apidoc utf8_distance
b06226ff 1126
1e54db1a 1127Returns the number of UTF-8 characters between the UTF-8 pointers C<a>
b06226ff
JH
1128and C<b>.
1129
1130WARNING: use only if you *know* that the pointers point inside the
1131same UTF-8 buffer.
1132
37607a96
PK
1133=cut
1134*/
a0ed51b3 1135
02eb7b47 1136IV
35a4481c 1137Perl_utf8_distance(pTHX_ const U8 *a, const U8 *b)
a0ed51b3 1138{
7918f24d
NC
1139 PERL_ARGS_ASSERT_UTF8_DISTANCE;
1140
bf1665bc 1141 return (a < b) ? -1 * (IV) utf8_length(a, b) : (IV) utf8_length(b, a);
a0ed51b3
LW
1142}
1143
b06226ff 1144/*
87cea99e 1145=for apidoc utf8_hop
b06226ff 1146
8850bf83
JH
1147Return the UTF-8 pointer C<s> displaced by C<off> characters, either
1148forward or backward.
b06226ff
JH
1149
1150WARNING: do not use the following unless you *know* C<off> is within
8850bf83
JH
1151the UTF-8 data pointed to by C<s> *and* that on entry C<s> is aligned
1152on the first byte of character or just after the last byte of a character.
b06226ff 1153
37607a96
PK
1154=cut
1155*/
a0ed51b3
LW
1156
1157U8 *
4373e329 1158Perl_utf8_hop(pTHX_ const U8 *s, I32 off)
a0ed51b3 1159{
7918f24d
NC
1160 PERL_ARGS_ASSERT_UTF8_HOP;
1161
96a5add6 1162 PERL_UNUSED_CONTEXT;
8850bf83
JH
1163 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g
1164 * the bitops (especially ~) can create illegal UTF-8.
1165 * In other words: in Perl UTF-8 is not just for Unicode. */
1166
a0ed51b3
LW
1167 if (off >= 0) {
1168 while (off--)
1169 s += UTF8SKIP(s);
1170 }
1171 else {
1172 while (off++) {
1173 s--;
8850bf83
JH
1174 while (UTF8_IS_CONTINUATION(*s))
1175 s--;
a0ed51b3
LW
1176 }
1177 }
4373e329 1178 return (U8 *)s;
a0ed51b3
LW
1179}
1180
6940069f 1181/*
fed3ba5d
NC
1182=for apidoc bytes_cmp_utf8
1183
a1433954
KW
1184Compares the sequence of characters (stored as octets) in C<b>, C<blen> with the
1185sequence of characters (stored as UTF-8) in C<u>, C<ulen>. Returns 0 if they are
fed3ba5d
NC
1186equal, -1 or -2 if the first string is less than the second string, +1 or +2
1187if the first string is greater than the second string.
1188
1189-1 or +1 is returned if the shorter string was identical to the start of the
1190longer string. -2 or +2 is returned if the was a difference between characters
1191within the strings.
1192
1193=cut
1194*/
1195
1196int
1197Perl_bytes_cmp_utf8(pTHX_ const U8 *b, STRLEN blen, const U8 *u, STRLEN ulen)
1198{
1199 const U8 *const bend = b + blen;
1200 const U8 *const uend = u + ulen;
1201
1202 PERL_ARGS_ASSERT_BYTES_CMP_UTF8;
1203
1204 PERL_UNUSED_CONTEXT;
1205
1206 while (b < bend && u < uend) {
1207 U8 c = *u++;
1208 if (!UTF8_IS_INVARIANT(c)) {
1209 if (UTF8_IS_DOWNGRADEABLE_START(c)) {
1210 if (u < uend) {
1211 U8 c1 = *u++;
1212 if (UTF8_IS_CONTINUATION(c1)) {
356979f4 1213 c = UNI_TO_NATIVE(TWO_BYTE_UTF8_TO_UNI(c, c1));
fed3ba5d
NC
1214 } else {
1215 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8),
1216 "Malformed UTF-8 character "
1217 "(unexpected non-continuation byte 0x%02x"
1218 ", immediately after start byte 0x%02x)"
1219 /* Dear diag.t, it's in the pod. */
1220 "%s%s", c1, c,
1221 PL_op ? " in " : "",
1222 PL_op ? OP_DESC(PL_op) : "");
1223 return -2;
1224 }
1225 } else {
1226 if (PL_op)
1227 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8),
1228 "%s in %s", unees, OP_DESC(PL_op));
1229 else
61a12c31 1230 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), "%s", unees);
fed3ba5d
NC
1231 return -2; /* Really want to return undef :-) */
1232 }
1233 } else {
1234 return -2;
1235 }
1236 }
1237 if (*b != c) {
1238 return *b < c ? -2 : +2;
1239 }
1240 ++b;
1241 }
1242
1243 if (b == bend && u == uend)
1244 return 0;
1245
1246 return b < bend ? +1 : -1;
1247}
1248
1249/*
87cea99e 1250=for apidoc utf8_to_bytes
6940069f 1251
2bbc8d55 1252Converts a string C<s> of length C<len> from UTF-8 into native byte encoding.
a1433954
KW
1253Unlike L</bytes_to_utf8>, this over-writes the original string, and
1254updates C<len> to contain the new length.
67e989fb 1255Returns zero on failure, setting C<len> to -1.
6940069f 1256
a1433954 1257If you need a copy of the string, see L</bytes_from_utf8>.
95be277c 1258
6940069f
GS
1259=cut
1260*/
1261
1262U8 *
37607a96 1263Perl_utf8_to_bytes(pTHX_ U8 *s, STRLEN *len)
6940069f 1264{
d4c19fe8
AL
1265 U8 * const save = s;
1266 U8 * const send = s + *len;
6940069f 1267 U8 *d;
246fae53 1268
7918f24d
NC
1269 PERL_ARGS_ASSERT_UTF8_TO_BYTES;
1270
1e54db1a 1271 /* ensure valid UTF-8 and chars < 256 before updating string */
d4c19fe8 1272 while (s < send) {
dcad2880
JH
1273 U8 c = *s++;
1274
1d72bdf6
NIS
1275 if (!UTF8_IS_INVARIANT(c) &&
1276 (!UTF8_IS_DOWNGRADEABLE_START(c) || (s >= send)
1277 || !(c = *s++) || !UTF8_IS_CONTINUATION(c))) {
10edeb5d 1278 *len = ((STRLEN) -1);
dcad2880
JH
1279 return 0;
1280 }
246fae53 1281 }
dcad2880
JH
1282
1283 d = s = save;
6940069f 1284 while (s < send) {
ed646e6e 1285 STRLEN ulen;
4b88fb76 1286 *d++ = (U8)utf8_to_uvchr_buf(s, send, &ulen);
ed646e6e 1287 s += ulen;
6940069f
GS
1288 }
1289 *d = '\0';
246fae53 1290 *len = d - save;
6940069f
GS
1291 return save;
1292}
1293
1294/*
87cea99e 1295=for apidoc bytes_from_utf8
f9a63242 1296
2bbc8d55 1297Converts a string C<s> of length C<len> from UTF-8 into native byte encoding.
a1433954 1298Unlike L</utf8_to_bytes> but like L</bytes_to_utf8>, returns a pointer to
ef9edfd0
JH
1299the newly-created string, and updates C<len> to contain the new
1300length. Returns the original string if no conversion occurs, C<len>
1301is unchanged. Do nothing if C<is_utf8> points to 0. Sets C<is_utf8> to
2bbc8d55
SP
13020 if C<s> is converted or consisted entirely of characters that are invariant
1303in utf8 (i.e., US-ASCII on non-EBCDIC machines).
f9a63242 1304
37607a96
PK
1305=cut
1306*/
f9a63242
JH
1307
1308U8 *
e1ec3a88 1309Perl_bytes_from_utf8(pTHX_ const U8 *s, STRLEN *len, bool *is_utf8)
f9a63242 1310{
f9a63242 1311 U8 *d;
e1ec3a88
AL
1312 const U8 *start = s;
1313 const U8 *send;
f9a63242
JH
1314 I32 count = 0;
1315
7918f24d
NC
1316 PERL_ARGS_ASSERT_BYTES_FROM_UTF8;
1317
96a5add6 1318 PERL_UNUSED_CONTEXT;
f9a63242 1319 if (!*is_utf8)
73d840c0 1320 return (U8 *)start;
f9a63242 1321
1e54db1a 1322 /* ensure valid UTF-8 and chars < 256 before converting string */
f9a63242 1323 for (send = s + *len; s < send;) {
e1ec3a88 1324 U8 c = *s++;
1d72bdf6 1325 if (!UTF8_IS_INVARIANT(c)) {
db42d148
NIS
1326 if (UTF8_IS_DOWNGRADEABLE_START(c) && s < send &&
1327 (c = *s++) && UTF8_IS_CONTINUATION(c))
1328 count++;
1329 else
73d840c0 1330 return (U8 *)start;
db42d148 1331 }
f9a63242
JH
1332 }
1333
35da51f7 1334 *is_utf8 = FALSE;
f9a63242 1335
212542aa 1336 Newx(d, (*len) - count + 1, U8);
ef9edfd0 1337 s = start; start = d;
f9a63242
JH
1338 while (s < send) {
1339 U8 c = *s++;
c4d5f83a
NIS
1340 if (!UTF8_IS_INVARIANT(c)) {
1341 /* Then it is two-byte encoded */
356979f4 1342 c = UNI_TO_NATIVE(TWO_BYTE_UTF8_TO_UNI(c, *s++));
c4d5f83a
NIS
1343 }
1344 *d++ = c;
f9a63242
JH
1345 }
1346 *d = '\0';
1347 *len = d - start;
73d840c0 1348 return (U8 *)start;
f9a63242
JH
1349}
1350
1351/*
87cea99e 1352=for apidoc bytes_to_utf8
6940069f 1353
ff97e5cf
KW
1354Converts a string C<s> of length C<len> bytes from the native encoding into
1355UTF-8.
6662521e 1356Returns a pointer to the newly-created string, and sets C<len> to
ff97e5cf 1357reflect the new length in bytes.
6940069f 1358
2bbc8d55
SP
1359A NUL character will be written after the end of the string.
1360
1361If you want to convert to UTF-8 from encodings other than
1362the native (Latin1 or EBCDIC),
a1433954 1363see L</sv_recode_to_utf8>().
c9ada85f 1364
497711e7 1365=cut
6940069f
GS
1366*/
1367
c682ebef
FC
1368/* This logic is duplicated in sv_catpvn_flags, so any bug fixes will
1369 likewise need duplication. */
1370
6940069f 1371U8*
35a4481c 1372Perl_bytes_to_utf8(pTHX_ const U8 *s, STRLEN *len)
6940069f 1373{
35a4481c 1374 const U8 * const send = s + (*len);
6940069f
GS
1375 U8 *d;
1376 U8 *dst;
7918f24d
NC
1377
1378 PERL_ARGS_ASSERT_BYTES_TO_UTF8;
96a5add6 1379 PERL_UNUSED_CONTEXT;
6940069f 1380
212542aa 1381 Newx(d, (*len) * 2 + 1, U8);
6940069f
GS
1382 dst = d;
1383
1384 while (s < send) {
35a4481c 1385 const UV uv = NATIVE_TO_ASCII(*s++);
c4d5f83a 1386 if (UNI_IS_INVARIANT(uv))
eb160463 1387 *d++ = (U8)UTF_TO_NATIVE(uv);
6940069f 1388 else {
eb160463
GS
1389 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
1390 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
6940069f
GS
1391 }
1392 }
1393 *d = '\0';
6662521e 1394 *len = d-dst;
6940069f
GS
1395 return dst;
1396}
1397
a0ed51b3 1398/*
dea0fc0b 1399 * Convert native (big-endian) or reversed (little-endian) UTF-16 to UTF-8.
a0ed51b3
LW
1400 *
1401 * Destination must be pre-extended to 3/2 source. Do not use in-place.
1402 * We optimize for native, for obvious reasons. */
1403
1404U8*
dea0fc0b 1405Perl_utf16_to_utf8(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen)
a0ed51b3 1406{
dea0fc0b
JH
1407 U8* pend;
1408 U8* dstart = d;
1409
7918f24d
NC
1410 PERL_ARGS_ASSERT_UTF16_TO_UTF8;
1411
dea0fc0b 1412 if (bytelen & 1)
f5992bc4 1413 Perl_croak(aTHX_ "panic: utf16_to_utf8: odd bytelen %"UVuf, (UV)bytelen);
dea0fc0b
JH
1414
1415 pend = p + bytelen;
1416
a0ed51b3 1417 while (p < pend) {
dea0fc0b
JH
1418 UV uv = (p[0] << 8) + p[1]; /* UTF-16BE */
1419 p += 2;
a0ed51b3 1420 if (uv < 0x80) {
e294cc5d
JH
1421#ifdef EBCDIC
1422 *d++ = UNI_TO_NATIVE(uv);
1423#else
eb160463 1424 *d++ = (U8)uv;
e294cc5d 1425#endif
a0ed51b3
LW
1426 continue;
1427 }
1428 if (uv < 0x800) {
eb160463
GS
1429 *d++ = (U8)(( uv >> 6) | 0xc0);
1430 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
1431 continue;
1432 }
52b9aa85 1433 if (uv >= 0xd800 && uv <= 0xdbff) { /* surrogates */
01ea242b 1434 if (p >= pend) {
dea0fc0b 1435 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
01ea242b
NC
1436 } else {
1437 UV low = (p[0] << 8) + p[1];
1438 p += 2;
52b9aa85 1439 if (low < 0xdc00 || low > 0xdfff)
01ea242b
NC
1440 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
1441 uv = ((uv - 0xd800) << 10) + (low - 0xdc00) + 0x10000;
1442 }
dbde1951
NC
1443 } else if (uv >= 0xdc00 && uv <= 0xdfff) {
1444 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
a0ed51b3
LW
1445 }
1446 if (uv < 0x10000) {
eb160463
GS
1447 *d++ = (U8)(( uv >> 12) | 0xe0);
1448 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
1449 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
1450 continue;
1451 }
1452 else {
eb160463
GS
1453 *d++ = (U8)(( uv >> 18) | 0xf0);
1454 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
1455 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
1456 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
1457 continue;
1458 }
1459 }
dea0fc0b 1460 *newlen = d - dstart;
a0ed51b3
LW
1461 return d;
1462}
1463
1464/* Note: this one is slightly destructive of the source. */
1465
1466U8*
dea0fc0b 1467Perl_utf16_to_utf8_reversed(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen)
a0ed51b3
LW
1468{
1469 U8* s = (U8*)p;
d4c19fe8 1470 U8* const send = s + bytelen;
7918f24d
NC
1471
1472 PERL_ARGS_ASSERT_UTF16_TO_UTF8_REVERSED;
1473
e0ea5e2d
NC
1474 if (bytelen & 1)
1475 Perl_croak(aTHX_ "panic: utf16_to_utf8_reversed: odd bytelen %"UVuf,
1476 (UV)bytelen);
1477
a0ed51b3 1478 while (s < send) {
d4c19fe8 1479 const U8 tmp = s[0];
a0ed51b3
LW
1480 s[0] = s[1];
1481 s[1] = tmp;
1482 s += 2;
1483 }
dea0fc0b 1484 return utf16_to_utf8(p, d, bytelen, newlen);
a0ed51b3
LW
1485}
1486
922e8cb4
KW
1487bool
1488Perl__is_uni_FOO(pTHX_ const U8 classnum, const UV c)
1489{
1490 U8 tmpbuf[UTF8_MAXBYTES+1];
1491 uvchr_to_utf8(tmpbuf, c);
1492 return _is_utf8_FOO(classnum, tmpbuf);
1493}
1494
c3fd2246
KW
1495/* for now these are all defined (inefficiently) in terms of the utf8 versions.
1496 * Note that the macros in handy.h that call these short-circuit calling them
1497 * for Latin-1 range inputs */
a0ed51b3
LW
1498
1499bool
84afefe6 1500Perl_is_uni_alnum(pTHX_ UV c)
a0ed51b3 1501{
89ebb4a3 1502 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1503 uvchr_to_utf8(tmpbuf, c);
922e8cb4 1504 return _is_utf8_FOO(_CC_WORDCHAR, tmpbuf);
a0ed51b3
LW
1505}
1506
1507bool
5092f92a
KW
1508Perl_is_uni_alnumc(pTHX_ UV c)
1509{
1510 U8 tmpbuf[UTF8_MAXBYTES+1];
1511 uvchr_to_utf8(tmpbuf, c);
922e8cb4 1512 return _is_utf8_FOO(_CC_ALPHANUMERIC, tmpbuf);
5092f92a
KW
1513}
1514
61b19385
KW
1515bool /* Internal function so we can deprecate the external one, and call
1516 this one from other deprecated functions in this file */
1517S_is_utf8_idfirst(pTHX_ const U8 *p)
1518{
1519 dVAR;
1520
1521 if (*p == '_')
1522 return TRUE;
1523 /* is_utf8_idstart would be more logical. */
1524 return is_utf8_common(p, &PL_utf8_idstart, "IdStart");
1525}
1526
5092f92a 1527bool
84afefe6 1528Perl_is_uni_idfirst(pTHX_ UV c)
a0ed51b3 1529{
89ebb4a3 1530 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1531 uvchr_to_utf8(tmpbuf, c);
61b19385 1532 return S_is_utf8_idfirst(aTHX_ tmpbuf);
a0ed51b3
LW
1533}
1534
1535bool
eba68aa0
KW
1536Perl__is_uni_perl_idcont(pTHX_ UV c)
1537{
1538 U8 tmpbuf[UTF8_MAXBYTES+1];
1539 uvchr_to_utf8(tmpbuf, c);
1540 return _is_utf8_perl_idcont(tmpbuf);
1541}
1542
1543bool
f91dcd13
KW
1544Perl__is_uni_perl_idstart(pTHX_ UV c)
1545{
1546 U8 tmpbuf[UTF8_MAXBYTES+1];
1547 uvchr_to_utf8(tmpbuf, c);
1548 return _is_utf8_perl_idstart(tmpbuf);
1549}
1550
1551bool
84afefe6 1552Perl_is_uni_alpha(pTHX_ UV c)
a0ed51b3 1553{
89ebb4a3 1554 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1555 uvchr_to_utf8(tmpbuf, c);
922e8cb4 1556 return _is_utf8_FOO(_CC_ALPHA, tmpbuf);
a0ed51b3
LW
1557}
1558
1559bool
84afefe6 1560Perl_is_uni_ascii(pTHX_ UV c)
4d61ec05 1561{
bc39fe24 1562 return isASCII(c);
4d61ec05
GS
1563}
1564
1565bool
bdd8600f
KW
1566Perl_is_uni_blank(pTHX_ UV c)
1567{
2cafb56b 1568 return isBLANK_uni(c);
bdd8600f
KW
1569}
1570
1571bool
84afefe6 1572Perl_is_uni_space(pTHX_ UV c)
a0ed51b3 1573{
add4123a 1574 return isSPACE_uni(c);
a0ed51b3
LW
1575}
1576
1577bool
84afefe6 1578Perl_is_uni_digit(pTHX_ UV c)
a0ed51b3 1579{
89ebb4a3 1580 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1581 uvchr_to_utf8(tmpbuf, c);
922e8cb4 1582 return _is_utf8_FOO(_CC_DIGIT, tmpbuf);
a0ed51b3
LW
1583}
1584
1585bool
84afefe6 1586Perl_is_uni_upper(pTHX_ UV c)
a0ed51b3 1587{
89ebb4a3 1588 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1589 uvchr_to_utf8(tmpbuf, c);
922e8cb4 1590 return _is_utf8_FOO(_CC_UPPER, tmpbuf);
a0ed51b3
LW
1591}
1592
1593bool
84afefe6 1594Perl_is_uni_lower(pTHX_ UV c)
a0ed51b3 1595{
89ebb4a3 1596 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1597 uvchr_to_utf8(tmpbuf, c);
922e8cb4 1598 return _is_utf8_FOO(_CC_LOWER, tmpbuf);
a0ed51b3
LW
1599}
1600
1601bool
84afefe6 1602Perl_is_uni_cntrl(pTHX_ UV c)
b8c5462f 1603{
7b952154 1604 return isCNTRL_L1(c);
b8c5462f
JH
1605}
1606
1607bool
84afefe6 1608Perl_is_uni_graph(pTHX_ UV c)
b8c5462f 1609{
89ebb4a3 1610 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1611 uvchr_to_utf8(tmpbuf, c);
922e8cb4 1612 return _is_utf8_FOO(_CC_GRAPH, tmpbuf);
b8c5462f
JH
1613}
1614
1615bool
84afefe6 1616Perl_is_uni_print(pTHX_ UV c)
a0ed51b3 1617{
89ebb4a3 1618 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1619 uvchr_to_utf8(tmpbuf, c);
922e8cb4 1620 return _is_utf8_FOO(_CC_PRINT, tmpbuf);
a0ed51b3
LW
1621}
1622
b8c5462f 1623bool
84afefe6 1624Perl_is_uni_punct(pTHX_ UV c)
b8c5462f 1625{
89ebb4a3 1626 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1627 uvchr_to_utf8(tmpbuf, c);
922e8cb4 1628 return _is_utf8_FOO(_CC_PUNCT, tmpbuf);
b8c5462f
JH
1629}
1630
4d61ec05 1631bool
84afefe6 1632Perl_is_uni_xdigit(pTHX_ UV c)
4d61ec05 1633{
4ac6419d 1634 return isXDIGIT_uni(c);
4d61ec05
GS
1635}
1636
3a4c58c9
KW
1637UV
1638Perl__to_upper_title_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp, const char S_or_s)
1639{
1640 /* We have the latin1-range values compiled into the core, so just use
1641 * those, converting the result to utf8. The only difference between upper
1642 * and title case in this range is that LATIN_SMALL_LETTER_SHARP_S is
1643 * either "SS" or "Ss". Which one to use is passed into the routine in
1644 * 'S_or_s' to avoid a test */
1645
1646 UV converted = toUPPER_LATIN1_MOD(c);
1647
1648 PERL_ARGS_ASSERT__TO_UPPER_TITLE_LATIN1;
1649
1650 assert(S_or_s == 'S' || S_or_s == 's');
1651
1652 if (UNI_IS_INVARIANT(converted)) { /* No difference between the two for
1653 characters in this range */
1654 *p = (U8) converted;
1655 *lenp = 1;
1656 return converted;
1657 }
1658
1659 /* toUPPER_LATIN1_MOD gives the correct results except for three outliers,
1660 * which it maps to one of them, so as to only have to have one check for
1661 * it in the main case */
1662 if (UNLIKELY(converted == LATIN_SMALL_LETTER_Y_WITH_DIAERESIS)) {
1663 switch (c) {
1664 case LATIN_SMALL_LETTER_Y_WITH_DIAERESIS:
1665 converted = LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS;
1666 break;
1667 case MICRO_SIGN:
1668 converted = GREEK_CAPITAL_LETTER_MU;
1669 break;
1670 case LATIN_SMALL_LETTER_SHARP_S:
1671 *(p)++ = 'S';
1672 *p = S_or_s;
1673 *lenp = 2;
1674 return 'S';
1675 default:
1676 Perl_croak(aTHX_ "panic: to_upper_title_latin1 did not expect '%c' to map to '%c'", c, LATIN_SMALL_LETTER_Y_WITH_DIAERESIS);
118e2215 1677 assert(0); /* NOTREACHED */
3a4c58c9
KW
1678 }
1679 }
1680
1681 *(p)++ = UTF8_TWO_BYTE_HI(converted);
1682 *p = UTF8_TWO_BYTE_LO(converted);
1683 *lenp = 2;
1684
1685 return converted;
1686}
1687
50bda2c3
KW
1688/* Call the function to convert a UTF-8 encoded character to the specified case.
1689 * Note that there may be more than one character in the result.
1690 * INP is a pointer to the first byte of the input character
1691 * OUTP will be set to the first byte of the string of changed characters. It
1692 * needs to have space for UTF8_MAXBYTES_CASE+1 bytes
1693 * LENP will be set to the length in bytes of the string of changed characters
1694 *
1695 * The functions return the ordinal of the first character in the string of OUTP */
f90a9a02
KW
1696#define CALL_UPPER_CASE(INP, OUTP, LENP) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_toupper, "ToUc", "utf8::ToSpecUc")
1697#define CALL_TITLE_CASE(INP, OUTP, LENP) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_totitle, "ToTc", "utf8::ToSpecTc")
1698#define CALL_LOWER_CASE(INP, OUTP, LENP) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_tolower, "ToLc", "utf8::ToSpecLc")
50bda2c3
KW
1699
1700/* This additionally has the input parameter SPECIALS, which if non-zero will
1701 * cause this to use the SPECIALS hash for folding (meaning get full case
1702 * folding); otherwise, when zero, this implies a simple case fold */
f90a9a02 1703#define CALL_FOLD_CASE(INP, OUTP, LENP, SPECIALS) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_tofold, "ToCf", (SPECIALS) ? "utf8::ToSpecCf" : NULL)
c3fd2246 1704
84afefe6
JH
1705UV
1706Perl_to_uni_upper(pTHX_ UV c, U8* p, STRLEN *lenp)
a0ed51b3 1707{
3a4c58c9
KW
1708 dVAR;
1709
a1433954
KW
1710 /* Convert the Unicode character whose ordinal is <c> to its uppercase
1711 * version and store that in UTF-8 in <p> and its length in bytes in <lenp>.
1712 * Note that the <p> needs to be at least UTF8_MAXBYTES_CASE+1 bytes since
c3fd2246
KW
1713 * the changed version may be longer than the original character.
1714 *
1715 * The ordinal of the first character of the changed version is returned
1716 * (but note, as explained above, that there may be more.) */
1717
7918f24d
NC
1718 PERL_ARGS_ASSERT_TO_UNI_UPPER;
1719
3a4c58c9
KW
1720 if (c < 256) {
1721 return _to_upper_title_latin1((U8) c, p, lenp, 'S');
1722 }
1723
0ebc6274 1724 uvchr_to_utf8(p, c);
3a4c58c9 1725 return CALL_UPPER_CASE(p, p, lenp);
a0ed51b3
LW
1726}
1727
84afefe6
JH
1728UV
1729Perl_to_uni_title(pTHX_ UV c, U8* p, STRLEN *lenp)
a0ed51b3 1730{
3a4c58c9
KW
1731 dVAR;
1732
7918f24d
NC
1733 PERL_ARGS_ASSERT_TO_UNI_TITLE;
1734
3a4c58c9
KW
1735 if (c < 256) {
1736 return _to_upper_title_latin1((U8) c, p, lenp, 's');
1737 }
1738
0ebc6274 1739 uvchr_to_utf8(p, c);
3a4c58c9 1740 return CALL_TITLE_CASE(p, p, lenp);
a0ed51b3
LW
1741}
1742
afc16117
KW
1743STATIC U8
1744S_to_lower_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp)
1745{
1746 /* We have the latin1-range values compiled into the core, so just use
1747 * those, converting the result to utf8. Since the result is always just
a1433954 1748 * one character, we allow <p> to be NULL */
afc16117
KW
1749
1750 U8 converted = toLOWER_LATIN1(c);
1751
1752 if (p != NULL) {
1753 if (UNI_IS_INVARIANT(converted)) {
1754 *p = converted;
1755 *lenp = 1;
1756 }
1757 else {
1758 *p = UTF8_TWO_BYTE_HI(converted);
1759 *(p+1) = UTF8_TWO_BYTE_LO(converted);
1760 *lenp = 2;
1761 }
1762 }
1763 return converted;
1764}
1765
84afefe6
JH
1766UV
1767Perl_to_uni_lower(pTHX_ UV c, U8* p, STRLEN *lenp)
a0ed51b3 1768{
968c5e6a
KW
1769 dVAR;
1770
7918f24d
NC
1771 PERL_ARGS_ASSERT_TO_UNI_LOWER;
1772
afc16117
KW
1773 if (c < 256) {
1774 return to_lower_latin1((U8) c, p, lenp);
bca00c02
KW
1775 }
1776
afc16117 1777 uvchr_to_utf8(p, c);
968c5e6a 1778 return CALL_LOWER_CASE(p, p, lenp);
a0ed51b3
LW
1779}
1780
84afefe6 1781UV
f673fad4 1782Perl__to_fold_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp, const bool flags)
a1dde8de 1783{
a1433954 1784 /* Corresponds to to_lower_latin1(), <flags> is TRUE if to use full case
f673fad4
KW
1785 * folding */
1786
a1dde8de
KW
1787 UV converted;
1788
1789 PERL_ARGS_ASSERT__TO_FOLD_LATIN1;
1790
1791 if (c == MICRO_SIGN) {
1792 converted = GREEK_SMALL_LETTER_MU;
1793 }
1794 else if (flags && c == LATIN_SMALL_LETTER_SHARP_S) {
1795 *(p)++ = 's';
1796 *p = 's';
1797 *lenp = 2;
1798 return 's';
1799 }
1800 else { /* In this range the fold of all other characters is their lower
1801 case */
1802 converted = toLOWER_LATIN1(c);
1803 }
1804
1805 if (UNI_IS_INVARIANT(converted)) {
1806 *p = (U8) converted;
1807 *lenp = 1;
1808 }
1809 else {
1810 *(p)++ = UTF8_TWO_BYTE_HI(converted);
1811 *p = UTF8_TWO_BYTE_LO(converted);
1812 *lenp = 2;
1813 }
1814
1815 return converted;
1816}
1817
1818UV
a0270393 1819Perl__to_uni_fold_flags(pTHX_ UV c, U8* p, STRLEN *lenp, const U8 flags)
84afefe6 1820{
4b593389 1821
a0270393
KW
1822 /* Not currently externally documented, and subject to change
1823 * <flags> bits meanings:
1824 * FOLD_FLAGS_FULL iff full folding is to be used;
1825 * FOLD_FLAGS_LOCALE iff in locale
1826 * FOLD_FLAGS_NOMIX_ASCII iff non-ASCII to ASCII folds are prohibited
1827 */
4b593389 1828
36bb2ab6 1829 PERL_ARGS_ASSERT__TO_UNI_FOLD_FLAGS;
7918f24d 1830
a1dde8de 1831 if (c < 256) {
a0270393
KW
1832 UV result = _to_fold_latin1((U8) c, p, lenp,
1833 cBOOL(((flags & FOLD_FLAGS_FULL)
1834 /* If ASCII-safe, don't allow full folding,
1835 * as that could include SHARP S => ss;
1836 * otherwise there is no crossing of
1837 * ascii/non-ascii in the latin1 range */
1838 && ! (flags & FOLD_FLAGS_NOMIX_ASCII))));
1839 /* It is illegal for the fold to cross the 255/256 boundary under
1840 * locale; in this case return the original */
1841 return (result > 256 && flags & FOLD_FLAGS_LOCALE)
1842 ? c
1843 : result;
a1dde8de
KW
1844 }
1845
a0270393
KW
1846 /* If no special needs, just use the macro */
1847 if ( ! (flags & (FOLD_FLAGS_LOCALE|FOLD_FLAGS_NOMIX_ASCII))) {
1848 uvchr_to_utf8(p, c);
1849 return CALL_FOLD_CASE(p, p, lenp, flags & FOLD_FLAGS_FULL);
1850 }
1851 else { /* Otherwise, _to_utf8_fold_flags has the intelligence to deal with
1852 the special flags. */
1853 U8 utf8_c[UTF8_MAXBYTES + 1];
1854 uvchr_to_utf8(utf8_c, c);
1855 return _to_utf8_fold_flags(utf8_c, p, lenp, flags, NULL);
1856 }
84afefe6
JH
1857}
1858
a0ed51b3 1859bool
84afefe6 1860Perl_is_uni_alnum_lc(pTHX_ UV c)
a0ed51b3 1861{
edfb3318
KW
1862 if (c < 256) {
1863 return isALNUM_LC(UNI_TO_NATIVE(c));
1864 }
922e8cb4 1865 return _is_uni_FOO(_CC_WORDCHAR, c);
a0ed51b3
LW
1866}
1867
1868bool
5092f92a
KW
1869Perl_is_uni_alnumc_lc(pTHX_ UV c)
1870{
1871 if (c < 256) {
15861f94 1872 return isALPHANUMERIC_LC(UNI_TO_NATIVE(c));
5092f92a 1873 }
922e8cb4 1874 return _is_uni_FOO(_CC_ALPHANUMERIC, c);
5092f92a
KW
1875}
1876
1877bool
84afefe6 1878Perl_is_uni_idfirst_lc(pTHX_ UV c)
a0ed51b3 1879{
edfb3318
KW
1880 if (c < 256) {
1881 return isIDFIRST_LC(UNI_TO_NATIVE(c));
1882 }
1883 return _is_uni_perl_idstart(c);
a0ed51b3
LW
1884}
1885
1886bool
84afefe6 1887Perl_is_uni_alpha_lc(pTHX_ UV c)
a0ed51b3 1888{
edfb3318
KW
1889 if (c < 256) {
1890 return isALPHA_LC(UNI_TO_NATIVE(c));
1891 }
922e8cb4 1892 return _is_uni_FOO(_CC_ALPHA, c);
a0ed51b3
LW
1893}
1894
1895bool
84afefe6 1896Perl_is_uni_ascii_lc(pTHX_ UV c)
4d61ec05 1897{
edfb3318
KW
1898 if (c < 256) {
1899 return isASCII_LC(UNI_TO_NATIVE(c));
1900 }
1901 return 0;
4d61ec05
GS
1902}
1903
1904bool
bdd8600f
KW
1905Perl_is_uni_blank_lc(pTHX_ UV c)
1906{
edfb3318
KW
1907 if (c < 256) {
1908 return isBLANK_LC(UNI_TO_NATIVE(c));
1909 }
61b19385 1910 return isBLANK_uni(c);
bdd8600f
KW
1911}
1912
1913bool
84afefe6 1914Perl_is_uni_space_lc(pTHX_ UV c)
a0ed51b3 1915{
edfb3318
KW
1916 if (c < 256) {
1917 return isSPACE_LC(UNI_TO_NATIVE(c));
1918 }
61b19385 1919 return isSPACE_uni(c);
a0ed51b3
LW
1920}
1921
1922bool
84afefe6 1923Perl_is_uni_digit_lc(pTHX_ UV c)
a0ed51b3 1924{
edfb3318
KW
1925 if (c < 256) {
1926 return isDIGIT_LC(UNI_TO_NATIVE(c));
1927 }
922e8cb4 1928 return _is_uni_FOO(_CC_DIGIT, c);
a0ed51b3
LW
1929}
1930
1931bool
84afefe6 1932Perl_is_uni_upper_lc(pTHX_ UV c)
a0ed51b3 1933{
edfb3318
KW
1934 if (c < 256) {
1935 return isUPPER_LC(UNI_TO_NATIVE(c));
1936 }
922e8cb4 1937 return _is_uni_FOO(_CC_UPPER, c);
a0ed51b3
LW
1938}
1939
1940bool
84afefe6 1941Perl_is_uni_lower_lc(pTHX_ UV c)
a0ed51b3 1942{
edfb3318
KW
1943 if (c < 256) {
1944 return isLOWER_LC(UNI_TO_NATIVE(c));
1945 }
922e8cb4 1946 return _is_uni_FOO(_CC_LOWER, c);
a0ed51b3
LW
1947}
1948
1949bool
84afefe6 1950Perl_is_uni_cntrl_lc(pTHX_ UV c)
b8c5462f 1951{
edfb3318
KW
1952 if (c < 256) {
1953 return isCNTRL_LC(UNI_TO_NATIVE(c));
1954 }
61b19385 1955 return 0;
b8c5462f
JH
1956}
1957
1958bool
84afefe6 1959Perl_is_uni_graph_lc(pTHX_ UV c)
b8c5462f 1960{
edfb3318
KW
1961 if (c < 256) {
1962 return isGRAPH_LC(UNI_TO_NATIVE(c));
1963 }
922e8cb4 1964 return _is_uni_FOO(_CC_GRAPH, c);
b8c5462f
JH
1965}
1966
1967bool
84afefe6 1968Perl_is_uni_print_lc(pTHX_ UV c)
a0ed51b3 1969{
edfb3318
KW
1970 if (c < 256) {
1971 return isPRINT_LC(UNI_TO_NATIVE(c));
1972 }
922e8cb4 1973 return _is_uni_FOO(_CC_PRINT, c);
a0ed51b3
LW
1974}
1975
b8c5462f 1976bool
84afefe6 1977Perl_is_uni_punct_lc(pTHX_ UV c)
b8c5462f 1978{
edfb3318
KW
1979 if (c < 256) {
1980 return isPUNCT_LC(UNI_TO_NATIVE(c));
1981 }
922e8cb4 1982 return _is_uni_FOO(_CC_PUNCT, c);
b8c5462f
JH
1983}
1984
4d61ec05 1985bool
84afefe6 1986Perl_is_uni_xdigit_lc(pTHX_ UV c)
4d61ec05 1987{
edfb3318
KW
1988 if (c < 256) {
1989 return isXDIGIT_LC(UNI_TO_NATIVE(c));
1990 }
61b19385 1991 return isXDIGIT_uni(c);
4d61ec05
GS
1992}
1993
b7ac61fa
JH
1994U32
1995Perl_to_uni_upper_lc(pTHX_ U32 c)
1996{
ee099d14
JH
1997 /* XXX returns only the first character -- do not use XXX */
1998 /* XXX no locale support yet */
1999 STRLEN len;
89ebb4a3 2000 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
ee099d14 2001 return (U32)to_uni_upper(c, tmpbuf, &len);
b7ac61fa
JH
2002}
2003
2004U32
2005Perl_to_uni_title_lc(pTHX_ U32 c)
2006{
ee099d14
JH
2007 /* XXX returns only the first character XXX -- do not use XXX */
2008 /* XXX no locale support yet */
2009 STRLEN len;
89ebb4a3 2010 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
ee099d14 2011 return (U32)to_uni_title(c, tmpbuf, &len);
b7ac61fa
JH
2012}
2013
2014U32
2015Perl_to_uni_lower_lc(pTHX_ U32 c)
2016{
ee099d14
JH
2017 /* XXX returns only the first character -- do not use XXX */
2018 /* XXX no locale support yet */
2019 STRLEN len;
89ebb4a3 2020 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
ee099d14 2021 return (U32)to_uni_lower(c, tmpbuf, &len);
b7ac61fa
JH
2022}
2023
26483009 2024PERL_STATIC_INLINE bool
5141f98e 2025S_is_utf8_common(pTHX_ const U8 *const p, SV **swash,
bde6a22d
NC
2026 const char *const swashname)
2027{
ea317ccb
KW
2028 /* returns a boolean giving whether or not the UTF8-encoded character that
2029 * starts at <p> is in the swash indicated by <swashname>. <swash>
2030 * contains a pointer to where the swash indicated by <swashname>
2031 * is to be stored; which this routine will do, so that future calls will
2032 * look at <*swash> and only generate a swash if it is not null
2033 *
2034 * Note that it is assumed that the buffer length of <p> is enough to
2035 * contain all the bytes that comprise the character. Thus, <*p> should
2036 * have been checked before this call for mal-formedness enough to assure
2037 * that. */
2038
97aff369 2039 dVAR;
7918f24d
NC
2040
2041 PERL_ARGS_ASSERT_IS_UTF8_COMMON;
2042
492a624f 2043 /* The API should have included a length for the UTF-8 character in <p>,
28123549 2044 * but it doesn't. We therefore assume that p has been validated at least
492a624f
KW
2045 * as far as there being enough bytes available in it to accommodate the
2046 * character without reading beyond the end, and pass that number on to the
2047 * validating routine */
28123549
KW
2048 if (! is_utf8_char_buf(p, p + UTF8SKIP(p))) {
2049 if (ckWARN_d(WARN_UTF8)) {
2050 Perl_warner(aTHX_ packWARN2(WARN_DEPRECATED,WARN_UTF8),
9816f121 2051 "Passing malformed UTF-8 to \"%s\" is deprecated", swashname);
28123549
KW
2052 if (ckWARN(WARN_UTF8)) { /* This will output details as to the
2053 what the malformation is */
2054 utf8_to_uvchr_buf(p, p + UTF8SKIP(p), NULL);
2055 }
2056 }
2057 return FALSE;
2058 }
87367d5f
KW
2059 if (!*swash) {
2060 U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST;
2061 *swash = _core_swash_init("utf8", swashname, &PL_sv_undef, 1, 0, NULL, &flags);
2062 }
28123549 2063
bde6a22d
NC
2064 return swash_fetch(*swash, p, TRUE) != 0;
2065}
2066
2067bool
922e8cb4
KW
2068Perl__is_utf8_FOO(pTHX_ const U8 classnum, const U8 *p)
2069{
2070 dVAR;
2071
2072 PERL_ARGS_ASSERT__IS_UTF8_FOO;
2073
2074 assert(classnum < _FIRST_NON_SWASH_CC);
2075
2076 return is_utf8_common(p, &PL_utf8_swash_ptrs[classnum], swash_property_names[classnum]);
2077}
2078
2079bool
7fc63493 2080Perl_is_utf8_alnum(pTHX_ const U8 *p)
a0ed51b3 2081{
97aff369 2082 dVAR;
7918f24d
NC
2083
2084 PERL_ARGS_ASSERT_IS_UTF8_ALNUM;
2085
671c33bf
NC
2086 /* NOTE: "IsWord", not "IsAlnum", since Alnum is a true
2087 * descendant of isalnum(3), in other words, it doesn't
2088 * contain the '_'. --jhi */
03940dc2 2089 return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_WORDCHAR], "IsWord");
a0ed51b3
LW
2090}
2091
2092bool
5092f92a
KW
2093Perl_is_utf8_alnumc(pTHX_ const U8 *p)
2094{
2095 dVAR;
2096
2097 PERL_ARGS_ASSERT_IS_UTF8_ALNUMC;
2098
03940dc2 2099 return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_ALPHANUMERIC], "IsAlnum");
5092f92a
KW
2100}
2101
2102bool
7fc63493 2103Perl_is_utf8_idfirst(pTHX_ const U8 *p) /* The naming is historical. */
a0ed51b3 2104{
97aff369 2105 dVAR;
7918f24d
NC
2106
2107 PERL_ARGS_ASSERT_IS_UTF8_IDFIRST;
2108
61b19385 2109 return S_is_utf8_idfirst(aTHX_ p);
82686b01
JH
2110}
2111
2112bool
c11ff943
KW
2113Perl_is_utf8_xidfirst(pTHX_ const U8 *p) /* The naming is historical. */
2114{
2115 dVAR;
2116
2117 PERL_ARGS_ASSERT_IS_UTF8_XIDFIRST;
2118
2119 if (*p == '_')
2120 return TRUE;
2121 /* is_utf8_idstart would be more logical. */
2122 return is_utf8_common(p, &PL_utf8_xidstart, "XIdStart");
2123}
2124
2125bool
d65654cb 2126Perl__is_utf8_perl_idstart(pTHX_ const U8 *p)
b6912c02
KW
2127{
2128 dVAR;
2129
d65654cb 2130 PERL_ARGS_ASSERT__IS_UTF8_PERL_IDSTART;
b6912c02
KW
2131
2132 return is_utf8_common(p, &PL_utf8_perl_idstart, "_Perl_IDStart");
2133}
2134
2135bool
eba68aa0
KW
2136Perl__is_utf8_perl_idcont(pTHX_ const U8 *p)
2137{
2138 dVAR;
2139
2140 PERL_ARGS_ASSERT__IS_UTF8_PERL_IDCONT;
2141
2142 return is_utf8_common(p, &PL_utf8_perl_idcont, "_Perl_IDCont");
2143}
2144
2145
2146bool
7fc63493 2147Perl_is_utf8_idcont(pTHX_ const U8 *p)
82686b01 2148{
97aff369 2149 dVAR;
7918f24d
NC
2150
2151 PERL_ARGS_ASSERT_IS_UTF8_IDCONT;
2152
d4c19fe8 2153 return is_utf8_common(p, &PL_utf8_idcont, "IdContinue");
a0ed51b3
LW
2154}
2155
2156bool
c11ff943
KW
2157Perl_is_utf8_xidcont(pTHX_ const U8 *p)
2158{
2159 dVAR;
2160
2161 PERL_ARGS_ASSERT_IS_UTF8_XIDCONT;
2162
c11ff943
KW
2163 return is_utf8_common(p, &PL_utf8_idcont, "XIdContinue");
2164}
2165
2166bool
7fc63493 2167Perl_is_utf8_alpha(pTHX_ const U8 *p)
a0ed51b3 2168{
97aff369 2169 dVAR;
7918f24d
NC
2170
2171 PERL_ARGS_ASSERT_IS_UTF8_ALPHA;
2172
03940dc2 2173 return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_ALPHA], "IsAlpha");
a0ed51b3
LW
2174}
2175
2176bool
7fc63493 2177Perl_is_utf8_ascii(pTHX_ const U8 *p)
b8c5462f 2178{
97aff369 2179 dVAR;
7918f24d
NC
2180
2181 PERL_ARGS_ASSERT_IS_UTF8_ASCII;
2182
bc39fe24
KW
2183 /* ASCII characters are the same whether in utf8 or not. So the macro
2184 * works on both utf8 and non-utf8 representations. */
2185 return isASCII(*p);
b8c5462f
JH
2186}
2187
2188bool
bdd8600f
KW
2189Perl_is_utf8_blank(pTHX_ const U8 *p)
2190{
2191 dVAR;
2192
2193 PERL_ARGS_ASSERT_IS_UTF8_BLANK;
2194
2cafb56b 2195 return isBLANK_utf8(p);
bdd8600f
KW
2196}
2197
2198bool
7fc63493 2199Perl_is_utf8_space(pTHX_ const U8 *p)
a0ed51b3 2200{
97aff369 2201 dVAR;
7918f24d
NC
2202
2203 PERL_ARGS_ASSERT_IS_UTF8_SPACE;
2204
add4123a 2205 return isSPACE_utf8(p);
a0ed51b3
LW
2206}
2207
2208bool
d1eb3177
YO
2209Perl_is_utf8_perl_space(pTHX_ const U8 *p)
2210{
2211 dVAR;
2212
2213 PERL_ARGS_ASSERT_IS_UTF8_PERL_SPACE;
2214
c4428693
KW
2215 /* Only true if is an ASCII space-like character, and ASCII is invariant
2216 * under utf8, so can just use the macro */
2217 return isSPACE_A(*p);
d1eb3177
YO
2218}
2219
2220bool
2221Perl_is_utf8_perl_word(pTHX_ const U8 *p)
2222{
2223 dVAR;
2224
2225 PERL_ARGS_ASSERT_IS_UTF8_PERL_WORD;
2226
c4428693
KW
2227 /* Only true if is an ASCII word character, and ASCII is invariant
2228 * under utf8, so can just use the macro */
2229 return isWORDCHAR_A(*p);
d1eb3177
YO
2230}
2231
2232bool
7fc63493 2233Perl_is_utf8_digit(pTHX_ const U8 *p)
a0ed51b3 2234{
97aff369 2235 dVAR;
7918f24d
NC
2236
2237 PERL_ARGS_ASSERT_IS_UTF8_DIGIT;
2238
03940dc2 2239 return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_DIGIT], "IsDigit");
a0ed51b3
LW
2240}
2241
2242bool
d1eb3177
YO
2243Perl_is_utf8_posix_digit(pTHX_ const U8 *p)
2244{
2245 dVAR;
2246
2247 PERL_ARGS_ASSERT_IS_UTF8_POSIX_DIGIT;
2248
c4428693
KW
2249 /* Only true if is an ASCII digit character, and ASCII is invariant
2250 * under utf8, so can just use the macro */
2251 return isDIGIT_A(*p);
d1eb3177
YO
2252}
2253
2254bool
7fc63493 2255Perl_is_utf8_upper(pTHX_ const U8 *p)
a0ed51b3 2256{
97aff369 2257 dVAR;
7918f24d
NC
2258
2259 PERL_ARGS_ASSERT_IS_UTF8_UPPER;
2260
03940dc2 2261 return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_UPPER], "IsUppercase");
a0ed51b3
LW
2262}
2263
2264bool
7fc63493 2265Perl_is_utf8_lower(pTHX_ const U8 *p)
a0ed51b3 2266{
97aff369 2267 dVAR;
7918f24d
NC
2268
2269 PERL_ARGS_ASSERT_IS_UTF8_LOWER;
2270
03940dc2 2271 return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_LOWER], "IsLowercase");
a0ed51b3
LW
2272}
2273
2274bool
7fc63493 2275Perl_is_utf8_cntrl(pTHX_ const U8 *p)
b8c5462f 2276{
97aff369 2277 dVAR;
7918f24d
NC
2278
2279 PERL_ARGS_ASSERT_IS_UTF8_CNTRL;
2280
a35d759a 2281 return isCNTRL_utf8(p);
b8c5462f
JH
2282}
2283
2284bool
7fc63493 2285Perl_is_utf8_graph(pTHX_ const U8 *p)
b8c5462f 2286{
97aff369 2287 dVAR;
7918f24d
NC
2288
2289 PERL_ARGS_ASSERT_IS_UTF8_GRAPH;
2290
03940dc2 2291 return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_GRAPH], "IsGraph");
b8c5462f
JH
2292}
2293
2294bool
7fc63493 2295Perl_is_utf8_print(pTHX_ const U8 *p)
a0ed51b3 2296{
97aff369 2297 dVAR;
7918f24d
NC
2298
2299 PERL_ARGS_ASSERT_IS_UTF8_PRINT;
2300
03940dc2 2301 return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_PRINT], "IsPrint");
a0ed51b3
LW
2302}
2303
2304bool
7fc63493 2305Perl_is_utf8_punct(pTHX_ const U8 *p)
b8c5462f 2306{
97aff369 2307 dVAR;
7918f24d
NC
2308
2309 PERL_ARGS_ASSERT_IS_UTF8_PUNCT;
2310
03940dc2 2311 return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_PUNCT], "IsPunct");
b8c5462f
JH
2312}
2313
2314bool
7fc63493 2315Perl_is_utf8_xdigit(pTHX_ const U8 *p)
b8c5462f 2316{
97aff369 2317 dVAR;
7918f24d
NC
2318
2319 PERL_ARGS_ASSERT_IS_UTF8_XDIGIT;
2320
4ac6419d 2321 return is_XDIGIT_utf8(p);
b8c5462f
JH
2322}
2323
2324bool
7dbf68d2
KW
2325Perl__is_utf8_mark(pTHX_ const U8 *p)
2326{
2327 dVAR;
2328
2329 PERL_ARGS_ASSERT__IS_UTF8_MARK;
2330
2331 return is_utf8_common(p, &PL_utf8_mark, "IsM");
2332}
2333
2334
2335bool
7fc63493 2336Perl_is_utf8_mark(pTHX_ const U8 *p)
a0ed51b3 2337{
97aff369 2338 dVAR;
7918f24d
NC
2339
2340 PERL_ARGS_ASSERT_IS_UTF8_MARK;
2341
7dbf68d2 2342 return _is_utf8_mark(p);
a0ed51b3
LW
2343}
2344
6b5c0936 2345/*
87cea99e 2346=for apidoc to_utf8_case
6b5c0936 2347
a1433954
KW
2348The C<p> contains the pointer to the UTF-8 string encoding
2349the character that is being converted. This routine assumes that the character
2350at C<p> is well-formed.
6b5c0936 2351
a1433954
KW
2352The C<ustrp> is a pointer to the character buffer to put the
2353conversion result to. The C<lenp> is a pointer to the length
6b5c0936
JH
2354of the result.
2355
a1433954 2356The C<swashp> is a pointer to the swash to use.
6b5c0936 2357
a1433954
KW
2358Both the special and normal mappings are stored in F<lib/unicore/To/Foo.pl>,
2359and loaded by SWASHNEW, using F<lib/utf8_heavy.pl>. The C<special> (usually,
0134edef 2360but not always, a multicharacter mapping), is tried first.
6b5c0936 2361
a1433954 2362The C<special> is a string like "utf8::ToSpecLower", which means the
0134edef
JH
2363hash %utf8::ToSpecLower. The access to the hash is through
2364Perl_to_utf8_case().
6b5c0936 2365
a1433954 2366The C<normal> is a string like "ToLower" which means the swash
0134edef
JH
2367%utf8::ToLower.
2368
2369=cut */
6b5c0936 2370
2104c8d9 2371UV
9a957fbc
AL
2372Perl_to_utf8_case(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp,
2373 SV **swashp, const char *normal, const char *special)
a0ed51b3 2374{
97aff369 2375 dVAR;
89ebb4a3 2376 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
0134edef 2377 STRLEN len = 0;
4b88fb76 2378 const UV uv0 = valid_utf8_to_uvchr(p, NULL);
1feea2c7
JH
2379 /* The NATIVE_TO_UNI() and UNI_TO_NATIVE() mappings
2380 * are necessary in EBCDIC, they are redundant no-ops
2381 * in ASCII-ish platforms, and hopefully optimized away. */
f54cb97a 2382 const UV uv1 = NATIVE_TO_UNI(uv0);
7918f24d
NC
2383
2384 PERL_ARGS_ASSERT_TO_UTF8_CASE;
2385
9ae3ac1a
KW
2386 /* Note that swash_fetch() doesn't output warnings for these because it
2387 * assumes we will */
8457b38f 2388 if (uv1 >= UNICODE_SURROGATE_FIRST) {
9ae3ac1a 2389 if (uv1 <= UNICODE_SURROGATE_LAST) {
8457b38f
KW
2390 if (ckWARN_d(WARN_SURROGATE)) {
2391 const char* desc = (PL_op) ? OP_DESC(PL_op) : normal;
2392 Perl_warner(aTHX_ packWARN(WARN_SURROGATE),
2393 "Operation \"%s\" returns its argument for UTF-16 surrogate U+%04"UVXf"", desc, uv1);
2394 }
9ae3ac1a
KW
2395 }
2396 else if (UNICODE_IS_SUPER(uv1)) {
8457b38f
KW
2397 if (ckWARN_d(WARN_NON_UNICODE)) {
2398 const char* desc = (PL_op) ? OP_DESC(PL_op) : normal;
2399 Perl_warner(aTHX_ packWARN(WARN_NON_UNICODE),
2400 "Operation \"%s\" returns its argument for non-Unicode code point 0x%04"UVXf"", desc, uv1);
2401 }
9ae3ac1a
KW
2402 }
2403
2404 /* Note that non-characters are perfectly legal, so no warning should
2405 * be given */
2406 }
2407
1feea2c7 2408 uvuni_to_utf8(tmpbuf, uv1);
0134edef
JH
2409
2410 if (!*swashp) /* load on-demand */
5ab9d2ef 2411 *swashp = _core_swash_init("utf8", normal, &PL_sv_undef, 4, 0, NULL, NULL);
0134edef 2412
a6f87d8c 2413 if (special) {
0134edef 2414 /* It might be "special" (sometimes, but not always,
2a37f04d 2415 * a multicharacter mapping) */
6673a63c 2416 HV * const hv = get_hv(special, 0);
b08cf34e
JH
2417 SV **svp;
2418
35da51f7 2419 if (hv &&
b08cf34e
JH
2420 (svp = hv_fetch(hv, (const char*)tmpbuf, UNISKIP(uv1), FALSE)) &&
2421 (*svp)) {
cfd0369c 2422 const char *s;
47654450 2423
cfd0369c 2424 s = SvPV_const(*svp, len);
47654450
JH
2425 if (len == 1)
2426 len = uvuni_to_utf8(ustrp, NATIVE_TO_UNI(*(U8*)s)) - ustrp;
2a37f04d 2427 else {
2f9475ad
JH
2428#ifdef EBCDIC
2429 /* If we have EBCDIC we need to remap the characters
2430 * since any characters in the low 256 are Unicode
2431 * code points, not EBCDIC. */
7cda7a3d 2432 U8 *t = (U8*)s, *tend = t + len, *d;
2f9475ad
JH
2433
2434 d = tmpbuf;
b08cf34e 2435 if (SvUTF8(*svp)) {
2f9475ad
JH
2436 STRLEN tlen = 0;
2437
2438 while (t < tend) {
6bd1c396 2439 const UV c = utf8_to_uvchr_buf(t, tend, &tlen);
2f9475ad
JH
2440 if (tlen > 0) {
2441 d = uvchr_to_utf8(d, UNI_TO_NATIVE(c));
2442 t += tlen;
2443 }
2444 else
2445 break;
2446 }
2447 }
2448 else {
36fec512
JH
2449 while (t < tend) {
2450 d = uvchr_to_utf8(d, UNI_TO_NATIVE(*t));
2451 t++;
2452 }
2f9475ad
JH
2453 }
2454 len = d - tmpbuf;
2455 Copy(tmpbuf, ustrp, len, U8);
2456#else
d2dcd0fb 2457 Copy(s, ustrp, len, U8);
2f9475ad 2458#endif
29e98929 2459 }
983ffd37 2460 }
0134edef
JH
2461 }
2462
2463 if (!len && *swashp) {
2114036c 2464 const UV uv2 = swash_fetch(*swashp, tmpbuf, TRUE /* => is utf8 */);
d4c19fe8 2465
0134edef
JH
2466 if (uv2) {
2467 /* It was "normal" (a single character mapping). */
d4c19fe8 2468 const UV uv3 = UNI_TO_NATIVE(uv2);
e9101d72 2469 len = uvchr_to_utf8(ustrp, uv3) - ustrp;
2a37f04d
JH
2470 }
2471 }
1feea2c7 2472
cbe07460
KW
2473 if (len) {
2474 if (lenp) {
2475 *lenp = len;
2476 }
2477 return valid_utf8_to_uvchr(ustrp, 0);
2478 }
2479
2480 /* Here, there was no mapping defined, which means that the code point maps
2481 * to itself. Return the inputs */
bfdf22ec 2482 len = UTF8SKIP(p);
ca9fab46
KW
2483 if (p != ustrp) { /* Don't copy onto itself */
2484 Copy(p, ustrp, len, U8);
2485 }
0134edef 2486
2a37f04d
JH
2487 if (lenp)
2488 *lenp = len;
2489
cbe07460
KW
2490 return uv0;
2491
a0ed51b3
LW
2492}
2493
051a06d4
KW
2494STATIC UV
2495S_check_locale_boundary_crossing(pTHX_ const U8* const p, const UV result, U8* const ustrp, STRLEN *lenp)
2496{
2497 /* This is called when changing the case of a utf8-encoded character above
2498 * the Latin1 range, and the operation is in locale. If the result
2499 * contains a character that crosses the 255/256 boundary, disallow the
2500 * change, and return the original code point. See L<perlfunc/lc> for why;
2501 *
a1433954
KW
2502 * p points to the original string whose case was changed; assumed
2503 * by this routine to be well-formed
051a06d4
KW
2504 * result the code point of the first character in the changed-case string
2505 * ustrp points to the changed-case string (<result> represents its first char)
2506 * lenp points to the length of <ustrp> */
2507
2508 UV original; /* To store the first code point of <p> */
2509
2510 PERL_ARGS_ASSERT_CHECK_LOCALE_BOUNDARY_CROSSING;
2511
2512 assert(! UTF8_IS_INVARIANT(*p) && ! UTF8_IS_DOWNGRADEABLE_START(*p));
2513
2514 /* We know immediately if the first character in the string crosses the
2515 * boundary, so can skip */
2516 if (result > 255) {
2517
2518 /* Look at every character in the result; if any cross the
2519 * boundary, the whole thing is disallowed */
2520 U8* s = ustrp + UTF8SKIP(ustrp);
2521 U8* e = ustrp + *lenp;
2522 while (s < e) {
2523 if (UTF8_IS_INVARIANT(*s) || UTF8_IS_DOWNGRADEABLE_START(*s))
2524 {
2525 goto bad_crossing;
2526 }
2527 s += UTF8SKIP(s);
2528 }
2529
2530 /* Here, no characters crossed, result is ok as-is */
2531 return result;
2532 }
2533
2534bad_crossing:
2535
2536 /* Failed, have to return the original */
4b88fb76 2537 original = valid_utf8_to_uvchr(p, lenp);
051a06d4
KW
2538 Copy(p, ustrp, *lenp, char);
2539 return original;
2540}
2541
d3e79532 2542/*
87cea99e 2543=for apidoc to_utf8_upper
d3e79532 2544
a1433954
KW
2545Convert the UTF-8 encoded character at C<p> to its uppercase version and
2546store that in UTF-8 in C<ustrp> and its length in bytes in C<lenp>. Note
89ebb4a3
JH
2547that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since
2548the uppercase version may be longer than the original character.
d3e79532
JH
2549
2550The first character of the uppercased version is returned
2551(but note, as explained above, that there may be more.)
2552
a1433954
KW
2553The character at C<p> is assumed by this routine to be well-formed.
2554
d3e79532
JH
2555=cut */
2556
051a06d4
KW
2557/* Not currently externally documented, and subject to change:
2558 * <flags> is set iff locale semantics are to be used for code points < 256
2559 * <tainted_ptr> if non-null, *tainted_ptr will be set TRUE iff locale rules
2560 * were used in the calculation; otherwise unchanged. */
2561
2104c8d9 2562UV
051a06d4 2563Perl__to_utf8_upper_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, const bool flags, bool* tainted_ptr)
a0ed51b3 2564{
97aff369 2565 dVAR;
7918f24d 2566
051a06d4
KW
2567 UV result;
2568
2569 PERL_ARGS_ASSERT__TO_UTF8_UPPER_FLAGS;
7918f24d 2570
3a4c58c9 2571 if (UTF8_IS_INVARIANT(*p)) {
051a06d4
KW
2572 if (flags) {
2573 result = toUPPER_LC(*p);
2574 }
2575 else {
81c6c7ce 2576 return _to_upper_title_latin1(*p, ustrp, lenp, 'S');
051a06d4 2577 }
3a4c58c9
KW
2578 }
2579 else if UTF8_IS_DOWNGRADEABLE_START(*p) {
051a06d4
KW
2580 if (flags) {
2581 result = toUPPER_LC(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)));
2582 }
2583 else {
81c6c7ce
KW
2584 return _to_upper_title_latin1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)),
2585 ustrp, lenp, 'S');
051a06d4
KW
2586 }
2587 }
2588 else { /* utf8, ord above 255 */
2589 result = CALL_UPPER_CASE(p, ustrp, lenp);
2590
2591 if (flags) {
2592 result = check_locale_boundary_crossing(p, result, ustrp, lenp);
2593 }
2594 return result;
2595 }
2596
2597 /* Here, used locale rules. Convert back to utf8 */
2598 if (UTF8_IS_INVARIANT(result)) {
2599 *ustrp = (U8) result;
2600 *lenp = 1;
2601 }
2602 else {
2603 *ustrp = UTF8_EIGHT_BIT_HI(result);
2604 *(ustrp + 1) = UTF8_EIGHT_BIT_LO(result);
2605 *lenp = 2;
3a4c58c9
KW
2606 }
2607
051a06d4
KW
2608 if (tainted_ptr) {
2609 *tainted_ptr = TRUE;
2610 }
2611 return result;
983ffd37 2612}
a0ed51b3 2613
d3e79532 2614/*
87cea99e 2615=for apidoc to_utf8_title
d3e79532 2616
a1433954
KW
2617Convert the UTF-8 encoded character at C<p> to its titlecase version and
2618store that in UTF-8 in C<ustrp> and its length in bytes in C<lenp>. Note
2619that the C<ustrp> needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
89ebb4a3 2620titlecase version may be longer than the original character.
d3e79532
JH
2621
2622The first character of the titlecased version is returned
2623(but note, as explained above, that there may be more.)
2624
a1433954
KW
2625The character at C<p> is assumed by this routine to be well-formed.
2626
d3e79532
JH
2627=cut */
2628
051a06d4
KW
2629/* Not currently externally documented, and subject to change:
2630 * <flags> is set iff locale semantics are to be used for code points < 256
2631 * Since titlecase is not defined in POSIX, uppercase is used instead
2632 * for these/
2633 * <tainted_ptr> if non-null, *tainted_ptr will be set TRUE iff locale rules
2634 * were used in the calculation; otherwise unchanged. */
2635
983ffd37 2636UV
051a06d4 2637Perl__to_utf8_title_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, const bool flags, bool* tainted_ptr)
983ffd37 2638{
97aff369 2639 dVAR;
7918f24d 2640
051a06d4
KW
2641 UV result;
2642
2643 PERL_ARGS_ASSERT__TO_UTF8_TITLE_FLAGS;
7918f24d 2644
3a4c58c9 2645 if (UTF8_IS_INVARIANT(*p)) {
051a06d4
KW
2646 if (flags) {
2647 result = toUPPER_LC(*p);
2648 }
2649 else {
81c6c7ce 2650 return _to_upper_title_latin1(*p, ustrp, lenp, 's');
051a06d4 2651 }
3a4c58c9
KW
2652 }
2653 else if UTF8_IS_DOWNGRADEABLE_START(*p) {
051a06d4
KW
2654 if (flags) {
2655 result = toUPPER_LC(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)));
2656 }
2657 else {
81c6c7ce
KW
2658 return _to_upper_title_latin1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)),
2659 ustrp, lenp, 's');
051a06d4
KW
2660 }
2661 }
2662 else { /* utf8, ord above 255 */
2663 result = CALL_TITLE_CASE(p, ustrp, lenp);
2664
2665 if (flags) {
2666 result = check_locale_boundary_crossing(p, result, ustrp, lenp);
2667 }
2668 return result;
2669 }
2670
2671 /* Here, used locale rules. Convert back to utf8 */
2672 if (UTF8_IS_INVARIANT(result)) {
2673 *ustrp = (U8) result;
2674 *lenp = 1;
2675 }
2676 else {
2677 *ustrp = UTF8_EIGHT_BIT_HI(result);
2678 *(ustrp + 1) = UTF8_EIGHT_BIT_LO(result);
2679 *lenp = 2;
3a4c58c9
KW
2680 }
2681
051a06d4
KW
2682 if (tainted_ptr) {
2683 *tainted_ptr = TRUE;
2684 }
2685 return result;
a0ed51b3
LW
2686}
2687
d3e79532 2688/*
87cea99e 2689=for apidoc to_utf8_lower
d3e79532 2690
a1433954
KW
2691Convert the UTF-8 encoded character at C<p> to its lowercase version and
2692store that in UTF-8 in ustrp and its length in bytes in C<lenp>. Note
2693that the C<ustrp> needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
89ebb4a3 2694lowercase version may be longer than the original character.
d3e79532
JH
2695
2696The first character of the lowercased version is returned
2697(but note, as explained above, that there may be more.)
2698
a1433954
KW
2699The character at C<p> is assumed by this routine to be well-formed.
2700
d3e79532
JH
2701=cut */
2702
051a06d4
KW
2703/* Not currently externally documented, and subject to change:
2704 * <flags> is set iff locale semantics are to be used for code points < 256
2705 * <tainted_ptr> if non-null, *tainted_ptr will be set TRUE iff locale rules
2706 * were used in the calculation; otherwise unchanged. */
2707
2104c8d9 2708UV
051a06d4 2709Perl__to_utf8_lower_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, const bool flags, bool* tainted_ptr)
a0ed51b3 2710{
051a06d4
KW
2711 UV result;
2712
97aff369 2713 dVAR;
7918f24d 2714
051a06d4 2715 PERL_ARGS_ASSERT__TO_UTF8_LOWER_FLAGS;
7918f24d 2716
968c5e6a 2717 if (UTF8_IS_INVARIANT(*p)) {
051a06d4
KW
2718 if (flags) {
2719 result = toLOWER_LC(*p);
2720 }
2721 else {
81c6c7ce 2722 return to_lower_latin1(*p, ustrp, lenp);
051a06d4 2723 }
968c5e6a
KW
2724 }
2725 else if UTF8_IS_DOWNGRADEABLE_START(*p) {
051a06d4
KW
2726 if (flags) {
2727 result = toLOWER_LC(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)));
2728 }
2729 else {
81c6c7ce
KW
2730 return to_lower_latin1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)),
2731 ustrp, lenp);
051a06d4 2732 }
968c5e6a 2733 }
051a06d4
KW
2734 else { /* utf8, ord above 255 */
2735 result = CALL_LOWER_CASE(p, ustrp, lenp);
2736
2737 if (flags) {
2738 result = check_locale_boundary_crossing(p, result, ustrp, lenp);
2739 }
968c5e6a 2740
051a06d4
KW
2741 return result;
2742 }
2743
2744 /* Here, used locale rules. Convert back to utf8 */
2745 if (UTF8_IS_INVARIANT(result)) {
2746 *ustrp = (U8) result;
2747 *lenp = 1;
2748 }
2749 else {
2750 *ustrp = UTF8_EIGHT_BIT_HI(result);
2751 *(ustrp + 1) = UTF8_EIGHT_BIT_LO(result);
2752 *lenp = 2;
2753 }
2754
2755 if (tainted_ptr) {
2756 *tainted_ptr = TRUE;
2757 }
2758 return result;
b4e400f9
JH
2759}
2760
d3e79532 2761/*
87cea99e 2762=for apidoc to_utf8_fold
d3e79532 2763
a1433954
KW
2764Convert the UTF-8 encoded character at C<p> to its foldcase version and
2765store that in UTF-8 in C<ustrp> and its length in bytes in C<lenp>. Note
2766that the C<ustrp> needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
d3e79532
JH
2767foldcase version may be longer than the original character (up to
2768three characters).
2769
2770The first character of the foldcased version is returned
2771(but note, as explained above, that there may be more.)
2772
a1433954
KW
2773The character at C<p> is assumed by this routine to be well-formed.
2774
d3e79532
JH
2775=cut */
2776
051a06d4
KW
2777/* Not currently externally documented, and subject to change,
2778 * in <flags>
2779 * bit FOLD_FLAGS_LOCALE is set iff locale semantics are to be used for code
2780 * points < 256. Since foldcase is not defined in
2781 * POSIX, lowercase is used instead
2782 * bit FOLD_FLAGS_FULL is set iff full case folds are to be used;
2783 * otherwise simple folds
a0270393
KW
2784 * bit FOLD_FLAGS_NOMIX_ASCII is set iff folds of non-ASCII to ASCII are
2785 * prohibited
051a06d4
KW
2786 * <tainted_ptr> if non-null, *tainted_ptr will be set TRUE iff locale rules
2787 * were used in the calculation; otherwise unchanged. */
36bb2ab6 2788
b4e400f9 2789UV
051a06d4 2790Perl__to_utf8_fold_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, U8 flags, bool* tainted_ptr)
b4e400f9 2791{
97aff369 2792 dVAR;
7918f24d 2793
051a06d4
KW
2794 UV result;
2795
36bb2ab6 2796 PERL_ARGS_ASSERT__TO_UTF8_FOLD_FLAGS;
7918f24d 2797
a0270393
KW
2798 /* These are mutually exclusive */
2799 assert (! ((flags & FOLD_FLAGS_LOCALE) && (flags & FOLD_FLAGS_NOMIX_ASCII)));
2800
50ba90ff
KW
2801 assert(p != ustrp); /* Otherwise overwrites */
2802
a1dde8de 2803 if (UTF8_IS_INVARIANT(*p)) {
051a06d4
KW
2804 if (flags & FOLD_FLAGS_LOCALE) {
2805 result = toLOWER_LC(*p);
2806 }
2807 else {
81c6c7ce
KW
2808 return _to_fold_latin1(*p, ustrp, lenp,
2809 cBOOL(flags & FOLD_FLAGS_FULL));
051a06d4 2810 }
a1dde8de
KW
2811 }
2812 else if UTF8_IS_DOWNGRADEABLE_START(*p) {
051a06d4
KW
2813 if (flags & FOLD_FLAGS_LOCALE) {
2814 result = toLOWER_LC(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)));
2815 }
2816 else {
81c6c7ce 2817 return _to_fold_latin1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)),
a0270393
KW
2818 ustrp, lenp,
2819 cBOOL((flags & FOLD_FLAGS_FULL
2820 /* If ASCII safe, don't allow full
2821 * folding, as that could include SHARP
2822 * S => ss; otherwise there is no
2823 * crossing of ascii/non-ascii in the
2824 * latin1 range */
2825 && ! (flags & FOLD_FLAGS_NOMIX_ASCII))));
051a06d4 2826 }
a1dde8de 2827 }
051a06d4 2828 else { /* utf8, ord above 255 */
a0270393 2829 result = CALL_FOLD_CASE(p, ustrp, lenp, flags & FOLD_FLAGS_FULL);
a1dde8de 2830
051a06d4 2831 if ((flags & FOLD_FLAGS_LOCALE)) {
a0270393 2832 return check_locale_boundary_crossing(p, result, ustrp, lenp);
051a06d4 2833 }
a0270393
KW
2834 else if (! (flags & FOLD_FLAGS_NOMIX_ASCII)) {
2835 return result;
2836 }
2837 else {
2838 /* This is called when changing the case of a utf8-encoded
2839 * character above the Latin1 range, and the result should not
2840 * contain an ASCII character. */
2841
2842 UV original; /* To store the first code point of <p> */
2843
2844 /* Look at every character in the result; if any cross the
2845 * boundary, the whole thing is disallowed */
2846 U8* s = ustrp;
2847 U8* e = ustrp + *lenp;
2848 while (s < e) {
2849 if (isASCII(*s)) {
2850 /* Crossed, have to return the original */
2851 original = valid_utf8_to_uvchr(p, lenp);
2852 Copy(p, ustrp, *lenp, char);
2853 return original;
2854 }
2855 s += UTF8SKIP(s);
2856 }
051a06d4 2857
a0270393
KW
2858 /* Here, no characters crossed, result is ok as-is */
2859 return result;
2860 }
051a06d4
KW
2861 }
2862
2863 /* Here, used locale rules. Convert back to utf8 */
2864 if (UTF8_IS_INVARIANT(result)) {
2865 *ustrp = (U8) result;
2866 *lenp = 1;
2867 }
2868 else {
2869 *ustrp = UTF8_EIGHT_BIT_HI(result);
2870 *(ustrp + 1) = UTF8_EIGHT_BIT_LO(result);
2871 *lenp = 2;
2872 }
2873
2874 if (tainted_ptr) {
2875 *tainted_ptr = TRUE;
2876 }
2877 return result;
a0ed51b3
LW
2878}
2879
711a919c 2880/* Note:
f90a9a02 2881 * Returns a "swash" which is a hash described in utf8.c:Perl_swash_fetch().
711a919c
ST
2882 * C<pkg> is a pointer to a package name for SWASHNEW, should be "utf8".
2883 * For other parameters, see utf8::SWASHNEW in lib/utf8_heavy.pl.
2884 */
c4a5db0c 2885
a0ed51b3 2886SV*
7fc63493 2887Perl_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv, I32 minbits, I32 none)
a0ed51b3 2888{
c4a5db0c
KW
2889 PERL_ARGS_ASSERT_SWASH_INIT;
2890
2891 /* Returns a copy of a swash initiated by the called function. This is the
2892 * public interface, and returning a copy prevents others from doing
2893 * mischief on the original */
2894
5d3d13d1 2895 return newSVsv(_core_swash_init(pkg, name, listsv, minbits, none, NULL, NULL));
c4a5db0c
KW
2896}
2897
2898SV*
5d3d13d1 2899Perl__core_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv, I32 minbits, I32 none, SV* invlist, U8* const flags_p)
c4a5db0c
KW
2900{
2901 /* Initialize and return a swash, creating it if necessary. It does this
87367d5f
KW
2902 * by calling utf8_heavy.pl in the general case. The returned value may be
2903 * the swash's inversion list instead if the input parameters allow it.
2904 * Which is returned should be immaterial to callers, as the only
923b6d4e
KW
2905 * operations permitted on a swash, swash_fetch(), _get_swash_invlist(),
2906 * and swash_to_invlist() handle both these transparently.
c4a5db0c
KW
2907 *
2908 * This interface should only be used by functions that won't destroy or
2909 * adversely change the swash, as doing so affects all other uses of the
2910 * swash in the program; the general public should use 'Perl_swash_init'
2911 * instead.
2912 *
2913 * pkg is the name of the package that <name> should be in.
2914 * name is the name of the swash to find. Typically it is a Unicode
2915 * property name, including user-defined ones
2916 * listsv is a string to initialize the swash with. It must be of the form
2917 * documented as the subroutine return value in
2918 * L<perlunicode/User-Defined Character Properties>
2919 * minbits is the number of bits required to represent each data element.
2920 * It is '1' for binary properties.
2921 * none I (khw) do not understand this one, but it is used only in tr///.
9a53f6cf 2922 * invlist is an inversion list to initialize the swash with (or NULL)
83199d38
KW
2923 * flags_p if non-NULL is the address of various input and output flag bits
2924 * to the routine, as follows: ('I' means is input to the routine;
2925 * 'O' means output from the routine. Only flags marked O are
2926 * meaningful on return.)
2927 * _CORE_SWASH_INIT_USER_DEFINED_PROPERTY indicates if the swash
2928 * came from a user-defined property. (I O)
5d3d13d1
KW
2929 * _CORE_SWASH_INIT_RETURN_IF_UNDEF indicates that instead of croaking
2930 * when the swash cannot be located, to simply return NULL. (I)
87367d5f
KW
2931 * _CORE_SWASH_INIT_ACCEPT_INVLIST indicates that the caller will accept a
2932 * return of an inversion list instead of a swash hash if this routine
2933 * thinks that would result in faster execution of swash_fetch() later
2934 * on. (I)
9a53f6cf
KW
2935 *
2936 * Thus there are three possible inputs to find the swash: <name>,
2937 * <listsv>, and <invlist>. At least one must be specified. The result
2938 * will be the union of the specified ones, although <listsv>'s various
2939 * actions can intersect, etc. what <name> gives.
2940 *
2941 * <invlist> is only valid for binary properties */
c4a5db0c 2942
27da23d5 2943 dVAR;
c4a5db0c 2944 SV* retval = &PL_sv_undef;
83199d38 2945 HV* swash_hv = NULL;
87367d5f
KW
2946 const int invlist_swash_boundary =
2947 (flags_p && *flags_p & _CORE_SWASH_INIT_ACCEPT_INVLIST)
2948 ? 512 /* Based on some benchmarking, but not extensive, see commit
2949 message */
2950 : -1; /* Never return just an inversion list */
9a53f6cf
KW
2951
2952 assert(listsv != &PL_sv_undef || strNE(name, "") || invlist);
2953 assert(! invlist || minbits == 1);
2954
2955 /* If data was passed in to go out to utf8_heavy to find the swash of, do
2956 * so */
2957 if (listsv != &PL_sv_undef || strNE(name, "")) {
69794297
KW
2958 dSP;
2959 const size_t pkg_len = strlen(pkg);
2960 const size_t name_len = strlen(name);
2961 HV * const stash = gv_stashpvn(pkg, pkg_len, 0);
2962 SV* errsv_save;
2963 GV *method;
2964
2965 PERL_ARGS_ASSERT__CORE_SWASH_INIT;
2966
2967 PUSHSTACKi(PERLSI_MAGIC);
ce3b816e 2968 ENTER;
69794297
KW
2969 SAVEHINTS();
2970 save_re_context();
650f067c
JL
2971 /* We might get here via a subroutine signature which uses a utf8
2972 * parameter name, at which point PL_subname will have been set
2973 * but not yet used. */
2974 save_item(PL_subname);
69794297
KW
2975 if (PL_parser && PL_parser->error_count)
2976 SAVEI8(PL_parser->error_count), PL_parser->error_count = 0;
2977 method = gv_fetchmeth(stash, "SWASHNEW", 8, -1);
2978 if (!method) { /* demand load utf8 */
2979 ENTER;
db2c6cb3
FC
2980 if ((errsv_save = GvSV(PL_errgv))) SAVEFREESV(errsv_save);
2981 GvSV(PL_errgv) = NULL;
69794297
KW
2982 /* It is assumed that callers of this routine are not passing in
2983 * any user derived data. */
2984 /* Need to do this after save_re_context() as it will set
2985 * PL_tainted to 1 while saving $1 etc (see the code after getrx:
2986 * in Perl_magic_get). Even line to create errsv_save can turn on
2987 * PL_tainted. */
284167a5
SM
2988#ifndef NO_TAINT_SUPPORT
2989 SAVEBOOL(TAINT_get);
2990 TAINT_NOT;
2991#endif
69794297
KW
2992 Perl_load_module(aTHX_ PERL_LOADMOD_NOIMPORT, newSVpvn(pkg,pkg_len),
2993 NULL);
eed484f9 2994 {
db2c6cb3
FC
2995 /* Not ERRSV, as there is no need to vivify a scalar we are
2996 about to discard. */
2997 SV * const errsv = GvSV(PL_errgv);
2998 if (!SvTRUE(errsv)) {
2999 GvSV(PL_errgv) = SvREFCNT_inc_simple(errsv_save);
3000 SvREFCNT_dec(errsv);
3001 }
eed484f9 3002 }
69794297
KW
3003 LEAVE;
3004 }
3005 SPAGAIN;
3006 PUSHMARK(SP);
3007 EXTEND(SP,5);
3008 mPUSHp(pkg, pkg_len);
3009 mPUSHp(name, name_len);
3010 PUSHs(listsv);
3011 mPUSHi(minbits);
3012 mPUSHi(none);
3013 PUTBACK;
db2c6cb3
FC
3014 if ((errsv_save = GvSV(PL_errgv))) SAVEFREESV(errsv_save);
3015 GvSV(PL_errgv) = NULL;
69794297
KW
3016 /* If we already have a pointer to the method, no need to use
3017 * call_method() to repeat the lookup. */
c41800a8
KW
3018 if (method
3019 ? call_sv(MUTABLE_SV(method), G_SCALAR)
69794297
KW
3020 : call_sv(newSVpvs_flags("SWASHNEW", SVs_TEMP), G_SCALAR | G_METHOD))
3021 {
3022 retval = *PL_stack_sp--;
3023 SvREFCNT_inc(retval);
3024 }
eed484f9 3025 {
db2c6cb3
FC
3026 /* Not ERRSV. See above. */
3027 SV * const errsv = GvSV(PL_errgv);
3028 if (!SvTRUE(errsv)) {
3029 GvSV(PL_errgv) = SvREFCNT_inc_simple(errsv_save);
3030 SvREFCNT_dec(errsv);
3031 }
eed484f9 3032 }
ce3b816e 3033 LEAVE;
69794297
KW
3034 POPSTACK;
3035 if (IN_PERL_COMPILETIME) {
3036 CopHINTS_set(PL_curcop, PL_hints);
3037 }
3038 if (!SvROK(retval) || SvTYPE(SvRV(retval)) != SVt_PVHV) {
3039 if (SvPOK(retval))
3040
3041 /* If caller wants to handle missing properties, let them */
5d3d13d1 3042 if (flags_p && *flags_p & _CORE_SWASH_INIT_RETURN_IF_UNDEF) {
69794297
KW
3043 return NULL;
3044 }
3045 Perl_croak(aTHX_
3046 "Can't find Unicode property definition \"%"SVf"\"",
3047 SVfARG(retval));
3048 Perl_croak(aTHX_ "SWASHNEW didn't return an HV ref");
3049 }
9a53f6cf 3050 } /* End of calling the module to find the swash */
36eb48b4 3051
83199d38
KW
3052 /* If this operation fetched a swash, and we will need it later, get it */
3053 if (retval != &PL_sv_undef
3054 && (minbits == 1 || (flags_p
3055 && ! (*flags_p
3056 & _CORE_SWASH_INIT_USER_DEFINED_PROPERTY))))
3057 {
3058 swash_hv = MUTABLE_HV(SvRV(retval));
3059
3060 /* If we don't already know that there is a user-defined component to
3061 * this swash, and the user has indicated they wish to know if there is
3062 * one (by passing <flags_p>), find out */
3063 if (flags_p && ! (*flags_p & _CORE_SWASH_INIT_USER_DEFINED_PROPERTY)) {
3064 SV** user_defined = hv_fetchs(swash_hv, "USER_DEFINED", FALSE);
3065 if (user_defined && SvUV(*user_defined)) {
3066 *flags_p |= _CORE_SWASH_INIT_USER_DEFINED_PROPERTY;
3067 }
3068 }
3069 }
3070
36eb48b4
KW
3071 /* Make sure there is an inversion list for binary properties */
3072 if (minbits == 1) {
3073 SV** swash_invlistsvp = NULL;
3074 SV* swash_invlist = NULL;
9a53f6cf 3075 bool invlist_in_swash_is_valid = FALSE;
02c85471
FC
3076 bool swash_invlist_unclaimed = FALSE; /* whether swash_invlist has
3077 an unclaimed reference count */
36eb48b4 3078
9a53f6cf 3079 /* If this operation fetched a swash, get its already existing
83199d38 3080 * inversion list, or create one for it */
36eb48b4 3081
83199d38 3082 if (swash_hv) {
5c9f4bd2 3083 swash_invlistsvp = hv_fetchs(swash_hv, "V", FALSE);
9a53f6cf
KW
3084 if (swash_invlistsvp) {
3085 swash_invlist = *swash_invlistsvp;
3086 invlist_in_swash_is_valid = TRUE;
3087 }
3088 else {
36eb48b4 3089 swash_invlist = _swash_to_invlist(retval);
02c85471 3090 swash_invlist_unclaimed = TRUE;
9a53f6cf
KW
3091 }
3092 }
3093
3094 /* If an inversion list was passed in, have to include it */
3095 if (invlist) {
3096
3097 /* Any fetched swash will by now have an inversion list in it;
3098 * otherwise <swash_invlist> will be NULL, indicating that we
3099 * didn't fetch a swash */
3100 if (swash_invlist) {
3101
3102 /* Add the passed-in inversion list, which invalidates the one
3103 * already stored in the swash */
3104 invlist_in_swash_is_valid = FALSE;
3105 _invlist_union(invlist, swash_invlist, &swash_invlist);
3106 }
3107 else {
3108
87367d5f
KW
3109 /* Here, there is no swash already. Set up a minimal one, if
3110 * we are going to return a swash */
3111 if ((int) _invlist_len(invlist) > invlist_swash_boundary) {
971d486f 3112 swash_hv = newHV();
4aca0fe6 3113 retval = newRV_noinc(MUTABLE_SV(swash_hv));
87367d5f 3114 }
9a53f6cf
KW
3115 swash_invlist = invlist;
3116 }
9a53f6cf
KW
3117 }
3118
3119 /* Here, we have computed the union of all the passed-in data. It may
3120 * be that there was an inversion list in the swash which didn't get
3121 * touched; otherwise save the one computed one */
87367d5f
KW
3122 if (! invlist_in_swash_is_valid
3123 && (int) _invlist_len(swash_invlist) > invlist_swash_boundary)
3124 {
5c9f4bd2 3125 if (! hv_stores(MUTABLE_HV(SvRV(retval)), "V", swash_invlist))
69794297
KW
3126 {
3127 Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed");
3128 }
cc34d8c5
FC
3129 /* We just stole a reference count. */
3130 if (swash_invlist_unclaimed) swash_invlist_unclaimed = FALSE;
3131 else SvREFCNT_inc_simple_void_NN(swash_invlist);
9a53f6cf 3132 }
87367d5f 3133
c41800a8 3134 /* Use the inversion list stand-alone if small enough */
87367d5f
KW
3135 if ((int) _invlist_len(swash_invlist) <= invlist_swash_boundary) {
3136 SvREFCNT_dec(retval);
02c85471
FC
3137 if (!swash_invlist_unclaimed)
3138 SvREFCNT_inc_simple_void_NN(swash_invlist);
3139 retval = newRV_noinc(swash_invlist);
87367d5f 3140 }
36eb48b4
KW
3141 }
3142
a0ed51b3
LW
3143 return retval;
3144}
3145
035d37be
JH
3146
3147/* This API is wrong for special case conversions since we may need to
3148 * return several Unicode characters for a single Unicode character
3149 * (see lib/unicore/SpecCase.txt) The SWASHGET in lib/utf8_heavy.pl is
3150 * the lower-level routine, and it is similarly broken for returning
38684baa
KW
3151 * multiple values. --jhi
3152 * For those, you should use to_utf8_case() instead */
b0e3252e 3153/* Now SWASHGET is recasted into S_swatch_get in this file. */
680c470c
ST
3154
3155/* Note:
3156 * Returns the value of property/mapping C<swash> for the first character
3157 * of the string C<ptr>. If C<do_utf8> is true, the string C<ptr> is
3158 * assumed to be in utf8. If C<do_utf8> is false, the string C<ptr> is
3159 * assumed to be in native 8-bit encoding. Caches the swatch in C<swash>.
af2af982
KW
3160 *
3161 * A "swash" is a hash which contains initially the keys/values set up by
3162 * SWASHNEW. The purpose is to be able to completely represent a Unicode
3163 * property for all possible code points. Things are stored in a compact form
3164 * (see utf8_heavy.pl) so that calculation is required to find the actual
3165 * property value for a given code point. As code points are looked up, new
3166 * key/value pairs are added to the hash, so that the calculation doesn't have
3167 * to ever be re-done. Further, each calculation is done, not just for the
3168 * desired one, but for a whole block of code points adjacent to that one.
3169 * For binary properties on ASCII machines, the block is usually for 64 code
3170 * points, starting with a code point evenly divisible by 64. Thus if the
3171 * property value for code point 257 is requested, the code goes out and
3172 * calculates the property values for all 64 code points between 256 and 319,
3173 * and stores these as a single 64-bit long bit vector, called a "swatch",
3174 * under the key for code point 256. The key is the UTF-8 encoding for code
3175 * point 256, minus the final byte. Thus, if the length of the UTF-8 encoding
3176 * for a code point is 13 bytes, the key will be 12 bytes long. If the value
3177 * for code point 258 is then requested, this code realizes that it would be
3178 * stored under the key for 256, and would find that value and extract the
3179 * relevant bit, offset from 256.
3180 *
3181 * Non-binary properties are stored in as many bits as necessary to represent
3182 * their values (32 currently, though the code is more general than that), not
3183 * as single bits, but the principal is the same: the value for each key is a
3184 * vector that encompasses the property values for all code points whose UTF-8
3185 * representations are represented by the key. That is, for all code points
3186 * whose UTF-8 representations are length N bytes, and the key is the first N-1
3187 * bytes of that.
680c470c 3188 */
a0ed51b3 3189UV
680c470c 3190Perl_swash_fetch(pTHX_ SV *swash, const U8 *ptr, bool do_utf8)
a0ed51b3 3191{
27da23d5 3192 dVAR;
ef8f7699 3193 HV *const hv = MUTABLE_HV(SvRV(swash));
3568d838
JH
3194 U32 klen;
3195 U32 off;
a0ed51b3 3196 STRLEN slen;
7d85a32c 3197 STRLEN needents;
cfd0369c 3198 const U8 *tmps = NULL;
a0ed51b3 3199 U32 bit;
979f2922 3200 SV *swatch;
3568d838 3201 U8 tmputf8[2];
35da51f7 3202 const UV c = NATIVE_TO_ASCII(*ptr);
3568d838 3203
7918f24d
NC
3204 PERL_ARGS_ASSERT_SWASH_FETCH;
3205
87367d5f
KW
3206 /* If it really isn't a hash, it isn't really swash; must be an inversion
3207 * list */
3208 if (SvTYPE(hv) != SVt_PVHV) {
3209 return _invlist_contains_cp((SV*)hv,
3210 (do_utf8)
3211 ? valid_utf8_to_uvchr(ptr, NULL)
3212 : c);
3213 }
3214
dbe7a391 3215 /* Convert to utf8 if not already */
3568d838 3216 if (!do_utf8 && !UNI_IS_INVARIANT(c)) {
979f2922
ST
3217 tmputf8[0] = (U8)UTF8_EIGHT_BIT_HI(c);
3218 tmputf8[1] = (U8)UTF8_EIGHT_BIT_LO(c);
3219 ptr = tmputf8;
3568d838
JH
3220 }
3221 /* Given a UTF-X encoded char 0xAA..0xYY,0xZZ
37e2e78e 3222 * then the "swatch" is a vec() for all the chars which start
3568d838
JH
3223 * with 0xAA..0xYY
3224 * So the key in the hash (klen) is length of encoded char -1
3225 */
3226 klen = UTF8SKIP(ptr) - 1;
3227 off = ptr[klen];
a0ed51b3 3228
979f2922 3229 if (klen == 0) {
37e2e78e 3230 /* If char is invariant then swatch is for all the invariant chars
1e54db1a 3231 * In both UTF-8 and UTF-8-MOD that happens to be UTF_CONTINUATION_MARK
7d85a32c 3232 */
979f2922
ST
3233 needents = UTF_CONTINUATION_MARK;
3234 off = NATIVE_TO_UTF(ptr[klen]);
3235 }
3236 else {
7d85a32c 3237 /* If char is encoded then swatch is for the prefix */
979f2922
ST
3238 needents = (1 << UTF_ACCUMULATION_SHIFT);
3239 off = NATIVE_TO_UTF(ptr[klen]) & UTF_CONTINUATION_MASK;
3240 }
7d85a32c 3241
a0ed51b3
LW
3242 /*
3243 * This single-entry cache saves about 1/3 of the utf8 overhead in test
3244 * suite. (That is, only 7-8% overall over just a hash cache. Still,
3245 * it's nothing to sniff at.) Pity we usually come through at least
3246 * two function calls to get here...
3247 *
3248 * NB: this code assumes that swatches are never modified, once generated!
3249 */
3250
3568d838 3251 if (hv == PL_last_swash_hv &&
a0ed51b3 3252 klen == PL_last_swash_klen &&
27da23d5 3253 (!klen || memEQ((char *)ptr, (char *)PL_last_swash_key, klen)) )
a0ed51b3
LW
3254 {
3255 tmps = PL_last_swash_tmps;
3256 slen = PL_last_swash_slen;
3257 }
3258 else {
3259 /* Try our second-level swatch cache, kept in a hash. */
e1ec3a88 3260 SV** svp = hv_fetch(hv, (const char*)ptr, klen, FALSE);
a0ed51b3 3261
b0e3252e 3262 /* If not cached, generate it via swatch_get */
979f2922
ST
3263 if (!svp || !SvPOK(*svp)
3264 || !(tmps = (const U8*)SvPV_const(*svp, slen))) {
2b9d42f0
NIS
3265 /* We use utf8n_to_uvuni() as we want an index into
3266 Unicode tables, not a native character number.
3267 */
aec46f14 3268 const UV code_point = utf8n_to_uvuni(ptr, UTF8_MAXBYTES, 0,
872c91ae
JH
3269 ckWARN(WARN_UTF8) ?
3270 0 : UTF8_ALLOW_ANY);
b0e3252e 3271 swatch = swatch_get(swash,
979f2922 3272 /* On EBCDIC & ~(0xA0-1) isn't a useful thing to do */
361ee0fe 3273 (klen) ? (code_point & ~((UV)needents - 1)) : 0,
979f2922
ST
3274 needents);
3275
923e4eb5 3276 if (IN_PERL_COMPILETIME)
623e6609 3277 CopHINTS_set(PL_curcop, PL_hints);
a0ed51b3 3278
979f2922 3279 svp = hv_store(hv, (const char *)ptr, klen, swatch, 0);
a0ed51b3 3280
979f2922
ST
3281 if (!svp || !(tmps = (U8*)SvPV(*svp, slen))
3282 || (slen << 3) < needents)
5637ef5b
NC
3283 Perl_croak(aTHX_ "panic: swash_fetch got improper swatch, "
3284 "svp=%p, tmps=%p, slen=%"UVuf", needents=%"UVuf,
3285 svp, tmps, (UV)slen, (UV)needents);
a0ed51b3
LW
3286 }
3287
3288 PL_last_swash_hv = hv;
16d8f38a 3289 assert(klen <= sizeof(PL_last_swash_key));
eac04b2e 3290 PL_last_swash_klen = (U8)klen;
cfd0369c
NC
3291 /* FIXME change interpvar.h? */
3292 PL_last_swash_tmps = (U8 *) tmps;
a0ed51b3
LW
3293 PL_last_swash_slen = slen;
3294 if (klen)
3295 Copy(ptr, PL_last_swash_key, klen, U8);
3296 }
3297
9faf8d75 3298 switch ((int)((slen << 3) / needents)) {
a0ed51b3
LW
3299 case 1:
3300 bit = 1 << (off & 7);
3301 off >>= 3;
3302 return (tmps[off] & bit) != 0;
3303 case 8:
3304 return tmps[off];
3305 case 16:
3306 off <<= 1;
3307 return (tmps[off] << 8) + tmps[off + 1] ;
3308 case 32:
3309 off <<= 2;
3310 return (tmps[off] << 24) + (tmps[off+1] << 16) + (tmps[off+2] << 8) + tmps[off + 3] ;
3311 }
5637ef5b
NC
3312 Perl_croak(aTHX_ "panic: swash_fetch got swatch of unexpected bit width, "
3313 "slen=%"UVuf", needents=%"UVuf, (UV)slen, (UV)needents);
670f1322 3314 NORETURN_FUNCTION_END;
a0ed51b3 3315}
2b9d42f0 3316
319009ee
KW
3317/* Read a single line of the main body of the swash input text. These are of
3318 * the form:
3319 * 0053 0056 0073
3320 * where each number is hex. The first two numbers form the minimum and
3321 * maximum of a range, and the third is the value associated with the range.
3322 * Not all swashes should have a third number
3323 *
3324 * On input: l points to the beginning of the line to be examined; it points
3325 * to somewhere in the string of the whole input text, and is
3326 * terminated by a \n or the null string terminator.
3327 * lend points to the null terminator of that string
3328 * wants_value is non-zero if the swash expects a third number
3329 * typestr is the name of the swash's mapping, like 'ToLower'
3330 * On output: *min, *max, and *val are set to the values read from the line.
3331 * returns a pointer just beyond the line examined. If there was no
3332 * valid min number on the line, returns lend+1
3333 */
3334
3335STATIC U8*
3336S_swash_scan_list_line(pTHX_ U8* l, U8* const lend, UV* min, UV* max, UV* val,
3337 const bool wants_value, const U8* const typestr)
3338{
3339 const int typeto = typestr[0] == 'T' && typestr[1] == 'o';
3340 STRLEN numlen; /* Length of the number */
02470786
KW
3341 I32 flags = PERL_SCAN_SILENT_ILLDIGIT
3342 | PERL_SCAN_DISALLOW_PREFIX
3343 | PERL_SCAN_SILENT_NON_PORTABLE;
319009ee
KW
3344
3345 /* nl points to the next \n in the scan */
3346 U8* const nl = (U8*)memchr(l, '\n', lend - l);
3347
3348 /* Get the first number on the line: the range minimum */
3349 numlen = lend - l;
3350 *min = grok_hex((char *)l, &numlen, &flags, NULL);
3351 if (numlen) /* If found a hex number, position past it */
3352 l += numlen;
3353 else if (nl) { /* Else, go handle next line, if any */
3354 return nl + 1; /* 1 is length of "\n" */
3355 }
3356 else { /* Else, no next line */
3357 return lend + 1; /* to LIST's end at which \n is not found */
3358 }
3359
3360 /* The max range value follows, separated by a BLANK */
3361 if (isBLANK(*l)) {
3362 ++l;
02470786
KW
3363 flags = PERL_SCAN_SILENT_ILLDIGIT
3364 | PERL_SCAN_DISALLOW_PREFIX
3365 | PERL_SCAN_SILENT_NON_PORTABLE;
319009ee
KW
3366 numlen = lend - l;
3367 *max = grok_hex((char *)l, &numlen, &flags, NULL);
3368 if (numlen)
3369 l += numlen;
3370 else /* If no value here, it is a single element range */
3371 *max = *min;
3372
3373 /* Non-binary tables have a third entry: what the first element of the
3374 * range maps to */
3375 if (wants_value) {
3376 if (isBLANK(*l)) {
3377 ++l;
f90a9a02
KW
3378
3379 /* The ToLc, etc table mappings are not in hex, and must be
3380 * corrected by adding the code point to them */
3381 if (typeto) {
3382 char *after_strtol = (char *) lend;
3383 *val = Strtol((char *)l, &after_strtol, 10);
3384 l = (U8 *) after_strtol;
f90a9a02
KW
3385 }
3386 else { /* Other tables are in hex, and are the correct result
3387 without tweaking */
a9d188b3
KW
3388 flags = PERL_SCAN_SILENT_ILLDIGIT
3389 | PERL_SCAN_DISALLOW_PREFIX
3390 | PERL_SCAN_SILENT_NON_PORTABLE;
3391 numlen = lend - l;
3392 *val = grok_hex((char *)l, &numlen, &flags, NULL);
3393 if (numlen)
3394 l += numlen;
3395 else
3396 *val = 0;
f90a9a02 3397 }
319009ee
KW
3398 }
3399 else {
3400 *val = 0;
3401 if (typeto) {
dcbac5bb 3402 /* diag_listed_as: To%s: illegal mapping '%s' */
319009ee
KW
3403 Perl_croak(aTHX_ "%s: illegal mapping '%s'",
3404 typestr, l);
3405 }
3406 }
3407 }
3408 else
3409 *val = 0; /* bits == 1, then any val should be ignored */
3410 }
3411 else { /* Nothing following range min, should be single element with no
3412 mapping expected */
3413 *max = *min;
3414 if (wants_value) {
3415 *val = 0;
3416 if (typeto) {
dcbac5bb 3417 /* diag_listed_as: To%s: illegal mapping '%s' */
319009ee
KW
3418 Perl_croak(aTHX_ "%s: illegal mapping '%s'", typestr, l);
3419 }
3420 }
3421 else
3422 *val = 0; /* bits == 1, then val should be ignored */
3423 }
3424
3425 /* Position to next line if any, or EOF */
3426 if (nl)
3427 l = nl + 1;
3428 else
3429 l = lend;
3430
3431 return l;
3432}
3433
979f2922
ST
3434/* Note:
3435 * Returns a swatch (a bit vector string) for a code point sequence
3436 * that starts from the value C<start> and comprises the number C<span>.
3437 * A C<swash> must be an object created by SWASHNEW (see lib/utf8_heavy.pl).
3438 * Should be used via swash_fetch, which will cache the swatch in C<swash>.
3439 */
3440STATIC SV*
b0e3252e 3441S_swatch_get(pTHX_ SV* swash, UV start, UV span)
979f2922
ST
3442{
3443 SV *swatch;
77f9f126 3444 U8 *l, *lend, *x, *xend, *s, *send;
979f2922 3445 STRLEN lcur, xcur, scur;
ef8f7699 3446 HV *const hv = MUTABLE_HV(SvRV(swash));
5c9f4bd2 3447 SV** const invlistsvp = hv_fetchs(hv, "V", FALSE);
36eb48b4 3448
88d45d28
KW
3449 SV** listsvp = NULL; /* The string containing the main body of the table */
3450 SV** extssvp = NULL;
3451 SV** invert_it_svp = NULL;
3452 U8* typestr = NULL;
786861f5
KW
3453 STRLEN bits;
3454 STRLEN octets; /* if bits == 1, then octets == 0 */
3455 UV none;
3456 UV end = start + span;
972dd592 3457
36eb48b4 3458 if (invlistsvp == NULL) {
786861f5
KW
3459 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
3460 SV** const nonesvp = hv_fetchs(hv, "NONE", FALSE);
3461 SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE);
3462 extssvp = hv_fetchs(hv, "EXTRAS", FALSE);
3463 listsvp = hv_fetchs(hv, "LIST", FALSE);
3464 invert_it_svp = hv_fetchs(hv, "INVERT_IT", FALSE);
3465
3466 bits = SvUV(*bitssvp);
3467 none = SvUV(*nonesvp);
3468 typestr = (U8*)SvPV_nolen(*typesvp);
3469 }
36eb48b4
KW
3470 else {
3471 bits = 1;
3472 none = 0;
3473 }
786861f5 3474 octets = bits >> 3; /* if bits == 1, then octets == 0 */
979f2922 3475
b0e3252e 3476 PERL_ARGS_ASSERT_SWATCH_GET;
7918f24d 3477
979f2922 3478 if (bits != 1 && bits != 8 && bits != 16 && bits != 32) {
b0e3252e 3479 Perl_croak(aTHX_ "panic: swatch_get doesn't expect bits %"UVuf,
660a4616 3480 (UV)bits);
979f2922
ST
3481 }
3482
84ea5ef6
KW
3483 /* If overflowed, use the max possible */
3484 if (end < start) {
3485 end = UV_MAX;
3486 span = end - start;
3487 }
3488
979f2922 3489 /* create and initialize $swatch */
979f2922 3490 scur = octets ? (span * octets) : (span + 7) / 8;
e524fe40
NC
3491 swatch = newSV(scur);
3492 SvPOK_on(swatch);
979f2922
ST
3493 s = (U8*)SvPVX(swatch);
3494 if (octets && none) {
0bd48802 3495 const U8* const e = s + scur;
979f2922
ST
3496 while (s < e) {
3497 if (bits == 8)
3498 *s++ = (U8)(none & 0xff);
3499 else if (bits == 16) {
3500 *s++ = (U8)((none >> 8) & 0xff);
3501 *s++ = (U8)( none & 0xff);
3502 }
3503 else if (bits == 32) {
3504 *s++ = (U8)((none >> 24) & 0xff);
3505 *s++ = (U8)((none >> 16) & 0xff);
3506 *s++ = (U8)((none >> 8) & 0xff);
3507 *s++ = (U8)( none & 0xff);
3508 }
3509 }
3510 *s = '\0';
3511 }
3512 else {
3513 (void)memzero((U8*)s, scur + 1);
3514 }
3515 SvCUR_set(swatch, scur);
3516 s = (U8*)SvPVX(swatch);
3517
36eb48b4
KW
3518 if (invlistsvp) { /* If has an inversion list set up use that */
3519 _invlist_populate_swatch(*invlistsvp, start, end, s);
3520 return swatch;
3521 }
3522
3523 /* read $swash->{LIST} */
979f2922
ST
3524 l = (U8*)SvPV(*listsvp, lcur);
3525 lend = l + lcur;
3526 while (l < lend) {
8ed25d53 3527 UV min, max, val, upper;
319009ee
KW
3528 l = S_swash_scan_list_line(aTHX_ l, lend, &min, &max, &val,
3529 cBOOL(octets), typestr);
3530 if (l > lend) {
979f2922
ST
3531 break;
3532 }
3533
972dd592 3534 /* If looking for something beyond this range, go try the next one */
979f2922
ST
3535 if (max < start)
3536 continue;
3537
8ed25d53
KW
3538 /* <end> is generally 1 beyond where we want to set things, but at the
3539 * platform's infinity, where we can't go any higher, we want to
3540 * include the code point at <end> */
3541 upper = (max < end)
3542 ? max
3543 : (max != UV_MAX || end != UV_MAX)
3544 ? end - 1
3545 : end;
3546
979f2922 3547 if (octets) {
35da51f7 3548 UV key;
979f2922
ST
3549 if (min < start) {
3550 if (!none || val < none) {
3551 val += start - min;
3552 }
3553 min = start;
3554 }
8ed25d53 3555 for (key = min; key <= upper; key++) {
979f2922 3556 STRLEN offset;
979f2922
ST
3557 /* offset must be non-negative (start <= min <= key < end) */
3558 offset = octets * (key - start);
3559 if (bits == 8)
3560 s[offset] = (U8)(val & 0xff);
3561 else if (bits == 16) {
3562 s[offset ] = (U8)((val >> 8) & 0xff);
3563 s[offset + 1] = (U8)( val & 0xff);
3564 }
3565 else if (bits == 32) {
3566 s[offset ] = (U8)((val >> 24) & 0xff);
3567 s[offset + 1] = (U8)((val >> 16) & 0xff);
3568 s[offset + 2] = (U8)((val >> 8) & 0xff);
3569 s[offset + 3] = (U8)( val & 0xff);
3570 }
3571
3572 if (!none || val < none)
3573 ++val;
3574 }
3575 }
711a919c 3576 else { /* bits == 1, then val should be ignored */
35da51f7 3577 UV key;
979f2922
ST
3578 if (min < start)
3579 min = start;
6cb05c12 3580
8ed25d53 3581 for (key = min; key <= upper; key++) {
0bd48802 3582 const STRLEN offset = (STRLEN)(key - start);
979f2922
ST
3583 s[offset >> 3] |= 1 << (offset & 7);
3584 }
3585 }
3586 } /* while */
979f2922 3587
9479a769 3588 /* Invert if the data says it should be. Assumes that bits == 1 */
77f9f126 3589 if (invert_it_svp && SvUV(*invert_it_svp)) {
0bda3001
KW
3590
3591 /* Unicode properties should come with all bits above PERL_UNICODE_MAX
3592 * be 0, and their inversion should also be 0, as we don't succeed any
3593 * Unicode property matches for non-Unicode code points */
3594 if (start <= PERL_UNICODE_MAX) {
3595
3596 /* The code below assumes that we never cross the
3597 * Unicode/above-Unicode boundary in a range, as otherwise we would
3598 * have to figure out where to stop flipping the bits. Since this
3599 * boundary is divisible by a large power of 2, and swatches comes
3600 * in small powers of 2, this should be a valid assumption */
3601 assert(start + span - 1 <= PERL_UNICODE_MAX);
3602
507a8485
KW
3603 send = s + scur;
3604 while (s < send) {
3605 *s = ~(*s);
3606 s++;
3607 }
0bda3001 3608 }
77f9f126
KW
3609 }
3610
d73c39c5
KW
3611 /* read $swash->{EXTRAS}
3612 * This code also copied to swash_to_invlist() below */
979f2922
ST
3613 x = (U8*)SvPV(*extssvp, xcur);
3614 xend = x + xcur;
3615 while (x < xend) {
3616 STRLEN namelen;
3617 U8 *namestr;
3618 SV** othersvp;
3619 HV* otherhv;
3620 STRLEN otherbits;
3621 SV **otherbitssvp, *other;
711a919c 3622 U8 *s, *o, *nl;
979f2922
ST
3623 STRLEN slen, olen;
3624
35da51f7 3625 const U8 opc = *x++;
979f2922
ST
3626 if (opc == '\n')
3627 continue;
3628
3629 nl = (U8*)memchr(x, '\n', xend - x);
3630
3631 if (opc != '-' && opc != '+' && opc != '!' && opc != '&') {
3632 if (nl) {
3633 x = nl + 1; /* 1 is length of "\n" */
3634 continue;
3635 }
3636 else {
3637 x = xend; /* to EXTRAS' end at which \n is not found */
3638 break;
3639 }
3640 }
3641
3642 namestr = x;
3643 if (nl) {
3644 namelen = nl - namestr;
3645 x = nl + 1;
3646 }
3647 else {
3648 namelen = xend - namestr;
3649 x = xend;
3650 }
3651
3652 othersvp = hv_fetch(hv, (char *)namestr, namelen, FALSE);
ef8f7699 3653 otherhv = MUTABLE_HV(SvRV(*othersvp));
017a3ce5 3654 otherbitssvp = hv_fetchs(otherhv, "BITS", FALSE);
979f2922
ST
3655 otherbits = (STRLEN)SvUV(*otherbitssvp);
3656 if (bits < otherbits)
5637ef5b
NC
3657 Perl_croak(aTHX_ "panic: swatch_get found swatch size mismatch, "
3658 "bits=%"UVuf", otherbits=%"UVuf, (UV)bits, (UV)otherbits);
979f2922
ST
3659
3660 /* The "other" swatch must be destroyed after. */
b0e3252e 3661 other = swatch_get(*othersvp, start, span);
979f2922
ST
3662 o = (U8*)SvPV(other, olen);
3663
3664 if (!olen)
b0e3252e 3665 Perl_croak(aTHX_ "panic: swatch_get got improper swatch");
979f2922
ST
3666
3667 s = (U8*)SvPV(swatch, slen);
3668 if (bits == 1 && otherbits == 1) {
3669 if (slen != olen)
5637ef5b
NC
3670 Perl_croak(aTHX_ "panic: swatch_get found swatch length "
3671 "mismatch, slen=%"UVuf", olen=%"UVuf,
3672 (UV)slen, (UV)olen);
979f2922
ST
3673
3674 switch (opc) {
3675 case '+':
3676 while (slen--)
3677 *s++ |= *o++;
3678 break;
3679 case '!':
3680 while (slen--)
3681 *s++ |= ~*o++;
3682 break;
3683 case '-':
3684 while (slen--)
3685 *s++ &= ~*o++;
3686 break;
3687 case '&':
3688 while (slen--)
3689 *s++ &= *o++;
3690 break;
3691 default: