This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
handy.h: Improve isDIGIT_utf8() and isXDIGIT_utf8() macros
[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
6ee84de2
KW
93Adds the UTF-8 representation of the code point C<uv> to the end
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
949cf498
KW
112This function will convert to UTF-8 (and not warn) even code points that aren't
113legal Unicode or are problematic, unless C<flags> contains one or more of the
a1433954
KW
114following flags:
115
949cf498
KW
116If C<uv> is a Unicode surrogate code point and UNICODE_WARN_SURROGATE is set,
117the function will raise a warning, provided UTF8 warnings are enabled. If instead
118UNICODE_DISALLOW_SURROGATE is set, the function will fail and return NULL.
119If both flags are set, the function will both warn and return NULL.
120
121The UNICODE_WARN_NONCHAR and UNICODE_DISALLOW_NONCHAR flags correspondingly
122affect how the function handles a Unicode non-character. And, likewise for the
123UNICODE_WARN_SUPER and UNICODE_DISALLOW_SUPER flags, and code points that are
124above the Unicode maximum of 0x10FFFF. Code points above 0x7FFF_FFFF (which are
125even less portable) can be warned and/or disallowed even if other above-Unicode
126code points are accepted by the UNICODE_WARN_FE_FF and UNICODE_DISALLOW_FE_FF
127flags.
128
129And finally, the flag UNICODE_WARN_ILLEGAL_INTERCHANGE selects all four of the
130above WARN flags; and UNICODE_DISALLOW_ILLEGAL_INTERCHANGE selects all four
131DISALLOW flags.
132
133
eebe1485
SC
134=cut
135*/
136
dfe13c55 137U8 *
b851fbc1 138Perl_uvuni_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags)
a0ed51b3 139{
7918f24d
NC
140 PERL_ARGS_ASSERT_UVUNI_TO_UTF8_FLAGS;
141
979f77b6
KW
142 /* The first problematic code point is the first surrogate */
143 if (uv >= UNICODE_SURROGATE_FIRST
144 && ckWARN4_d(WARN_UTF8, WARN_SURROGATE, WARN_NON_UNICODE, WARN_NONCHAR))
145 {
949cf498
KW
146 if (UNICODE_IS_SURROGATE(uv)) {
147 if (flags & UNICODE_WARN_SURROGATE) {
8457b38f 148 Perl_ck_warner_d(aTHX_ packWARN(WARN_SURROGATE),
949cf498
KW
149 "UTF-16 surrogate U+%04"UVXf, uv);
150 }
151 if (flags & UNICODE_DISALLOW_SURROGATE) {
152 return NULL;
153 }
154 }
155 else if (UNICODE_IS_SUPER(uv)) {
156 if (flags & UNICODE_WARN_SUPER
157 || (UNICODE_IS_FE_FF(uv) && (flags & UNICODE_WARN_FE_FF)))
158 {
8457b38f 159 Perl_ck_warner_d(aTHX_ packWARN(WARN_NON_UNICODE),
949cf498
KW
160 "Code point 0x%04"UVXf" is not Unicode, may not be portable", uv);
161 }
162 if (flags & UNICODE_DISALLOW_SUPER
163 || (UNICODE_IS_FE_FF(uv) && (flags & UNICODE_DISALLOW_FE_FF)))
164 {
165 return NULL;
166 }
167 }
168 else if (UNICODE_IS_NONCHAR(uv)) {
169 if (flags & UNICODE_WARN_NONCHAR) {
8457b38f 170 Perl_ck_warner_d(aTHX_ packWARN(WARN_NONCHAR),
949cf498
KW
171 "Unicode non-character U+%04"UVXf" is illegal for open interchange",
172 uv);
173 }
174 if (flags & UNICODE_DISALLOW_NONCHAR) {
175 return NULL;
176 }
177 }
507b9800 178 }
c4d5f83a 179 if (UNI_IS_INVARIANT(uv)) {
eb160463 180 *d++ = (U8)UTF_TO_NATIVE(uv);
a0ed51b3
LW
181 return d;
182 }
2d331972 183#if defined(EBCDIC)
1d72bdf6
NIS
184 else {
185 STRLEN len = UNISKIP(uv);
186 U8 *p = d+len-1;
187 while (p > d) {
eb160463 188 *p-- = (U8)UTF_TO_NATIVE((uv & UTF_CONTINUATION_MASK) | UTF_CONTINUATION_MARK);
1d72bdf6
NIS
189 uv >>= UTF_ACCUMULATION_SHIFT;
190 }
eb160463 191 *p = (U8)UTF_TO_NATIVE((uv & UTF_START_MASK(len)) | UTF_START_MARK(len));
1d72bdf6
NIS
192 return d+len;
193 }
194#else /* Non loop style */
a0ed51b3 195 if (uv < 0x800) {
eb160463
GS
196 *d++ = (U8)(( uv >> 6) | 0xc0);
197 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
198 return d;
199 }
200 if (uv < 0x10000) {
eb160463
GS
201 *d++ = (U8)(( uv >> 12) | 0xe0);
202 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
203 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
204 return d;
205 }
206 if (uv < 0x200000) {
eb160463
GS
207 *d++ = (U8)(( uv >> 18) | 0xf0);
208 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
209 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
210 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
211 return d;
212 }
213 if (uv < 0x4000000) {
eb160463
GS
214 *d++ = (U8)(( uv >> 24) | 0xf8);
215 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
216 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
217 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
218 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
219 return d;
220 }
221 if (uv < 0x80000000) {
eb160463
GS
222 *d++ = (U8)(( uv >> 30) | 0xfc);
223 *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80);
224 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
225 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
226 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
227 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
228 return d;
229 }
6b8eaf93 230#ifdef HAS_QUAD
d7578b48 231 if (uv < UTF8_QUAD_MAX)
a0ed51b3
LW
232#endif
233 {
eb160463
GS
234 *d++ = 0xfe; /* Can't match U+FEFF! */
235 *d++ = (U8)(((uv >> 30) & 0x3f) | 0x80);
236 *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80);
237 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
238 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
239 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
240 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
241 return d;
242 }
6b8eaf93 243#ifdef HAS_QUAD
a0ed51b3 244 {
eb160463
GS
245 *d++ = 0xff; /* Can't match U+FFFE! */
246 *d++ = 0x80; /* 6 Reserved bits */
247 *d++ = (U8)(((uv >> 60) & 0x0f) | 0x80); /* 2 Reserved bits */
248 *d++ = (U8)(((uv >> 54) & 0x3f) | 0x80);
249 *d++ = (U8)(((uv >> 48) & 0x3f) | 0x80);
250 *d++ = (U8)(((uv >> 42) & 0x3f) | 0x80);
251 *d++ = (U8)(((uv >> 36) & 0x3f) | 0x80);
252 *d++ = (U8)(((uv >> 30) & 0x3f) | 0x80);
253 *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80);
254 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
255 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
256 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
257 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
258 return d;
259 }
260#endif
1d72bdf6 261#endif /* Loop style */
a0ed51b3 262}
9041c2e3 263
646ca15d
JH
264/*
265
f7d739d1 266Tests if the first C<len> bytes of string C<s> form a valid UTF-8
646ca15d 267character. Note that an INVARIANT (i.e. ASCII) character is a valid
f7d739d1 268UTF-8 character. The number of bytes in the UTF-8 character
646ca15d
JH
269will be returned if it is valid, otherwise 0.
270
271This is the "slow" version as opposed to the "fast" version which is
272the "unrolled" IS_UTF8_CHAR(). E.g. for t/uni/class.t the speed
273difference is a factor of 2 to 3. For lengths (UTF8SKIP(s)) of four
274or less you should use the IS_UTF8_CHAR(), for lengths of five or more
275you should use the _slow(). In practice this means that the _slow()
276will be used very rarely, since the maximum Unicode code point (as of
277Unicode 4.1) is U+10FFFF, which encodes in UTF-8 to four bytes. Only
278the "Perl extended UTF-8" (the infamous 'v-strings') will encode into
279five bytes or more.
280
281=cut */
7af276bc 282PERL_STATIC_INLINE STRLEN
5f66b61c 283S_is_utf8_char_slow(const U8 *s, const STRLEN len)
646ca15d 284{
cd7e6c88 285 dTHX; /* The function called below requires thread context */
646ca15d 286
cd7e6c88 287 STRLEN actual_len;
646ca15d 288
cd7e6c88 289 PERL_ARGS_ASSERT_IS_UTF8_CHAR_SLOW;
646ca15d 290
cd7e6c88 291 utf8n_to_uvuni(s, len, &actual_len, UTF8_CHECK_ONLY);
646ca15d 292
cd7e6c88 293 return (actual_len == (STRLEN) -1) ? 0 : actual_len;
646ca15d 294}
9041c2e3
NIS
295
296/*
492a624f
KW
297=for apidoc is_utf8_char_buf
298
299Returns the number of bytes that comprise the first UTF-8 encoded character in
300buffer C<buf>. C<buf_end> should point to one position beyond the end of the
301buffer. 0 is returned if C<buf> does not point to a complete, valid UTF-8
302encoded character.
303
304Note that an INVARIANT character (i.e. ASCII on non-EBCDIC
305machines) is a valid UTF-8 character.
306
307=cut */
308
309STRLEN
310Perl_is_utf8_char_buf(const U8 *buf, const U8* buf_end)
311{
312
313 STRLEN len;
314
315 PERL_ARGS_ASSERT_IS_UTF8_CHAR_BUF;
316
317 if (buf_end <= buf) {
318 return 0;
319 }
320
321 len = buf_end - buf;
322 if (len > UTF8SKIP(buf)) {
323 len = UTF8SKIP(buf);
324 }
325
326#ifdef IS_UTF8_CHAR
327 if (IS_UTF8_CHAR_FAST(len))
328 return IS_UTF8_CHAR(buf, len) ? len : 0;
329#endif /* #ifdef IS_UTF8_CHAR */
330 return is_utf8_char_slow(buf, len);
331}
332
333/*
87cea99e 334=for apidoc is_utf8_char
eebe1485 335
76848387
KW
336DEPRECATED!
337
5da9da9e 338Tests if some arbitrary number of bytes begins in a valid UTF-8
2bbc8d55
SP
339character. Note that an INVARIANT (i.e. ASCII on non-EBCDIC machines)
340character is a valid UTF-8 character. The actual number of bytes in the UTF-8
341character will be returned if it is valid, otherwise 0.
9041c2e3 342
76848387 343This function is deprecated due to the possibility that malformed input could
a1433954 344cause reading beyond the end of the input buffer. Use L</is_utf8_char_buf>
76848387 345instead.
e0328548 346
82686b01 347=cut */
76848387 348
067a85ef 349STRLEN
668b6d8d 350Perl_is_utf8_char(const U8 *s)
386d01d6 351{
7918f24d 352 PERL_ARGS_ASSERT_IS_UTF8_CHAR;
492a624f 353
76848387 354 /* Assumes we have enough space, which is why this is deprecated */
492a624f 355 return is_utf8_char_buf(s, s + UTF8SKIP(s));
386d01d6
GS
356}
357
eaf7a4d2 358
6662521e 359/*
87cea99e 360=for apidoc is_utf8_string
6662521e 361
a1433954 362Returns true if the first C<len> bytes of string C<s> form a valid
9f7e3d64 363UTF-8 string, false otherwise. If C<len> is 0, it will be calculated
e0328548
KW
364using C<strlen(s)> (which means if you use this option, that C<s> has to have a
365terminating NUL byte). Note that all characters being ASCII constitute 'a
366valid UTF-8 string'.
6662521e 367
a1433954 368See also L</is_ascii_string>(), L</is_utf8_string_loclen>(), and L</is_utf8_string_loc>().
768c67ee 369
6662521e
GS
370=cut
371*/
372
8e84507e 373bool
668b6d8d 374Perl_is_utf8_string(const U8 *s, STRLEN len)
6662521e 375{
35da51f7 376 const U8* const send = s + (len ? len : strlen((const char *)s));
7fc63493 377 const U8* x = s;
067a85ef 378
7918f24d 379 PERL_ARGS_ASSERT_IS_UTF8_STRING;
1aa99e6b 380
6662521e 381 while (x < send) {
1acdb0da 382 /* Inline the easy bits of is_utf8_char() here for speed... */
e0328548
KW
383 if (UTF8_IS_INVARIANT(*x)) {
384 x++;
385 }
1acdb0da
JH
386 else {
387 /* ... and call is_utf8_char() only if really needed. */
e0328548
KW
388 const STRLEN c = UTF8SKIP(x);
389 const U8* const next_char_ptr = x + c;
390
391 if (next_char_ptr > send) {
392 return FALSE;
393 }
394
768c67ee
JH
395 if (IS_UTF8_CHAR_FAST(c)) {
396 if (!IS_UTF8_CHAR(x, c))
e0328548 397 return FALSE;
3c614e38 398 }
e0328548
KW
399 else if (! is_utf8_char_slow(x, c)) {
400 return FALSE;
401 }
402 x = next_char_ptr;
1acdb0da 403 }
6662521e 404 }
768c67ee 405
067a85ef 406 return TRUE;
6662521e
GS
407}
408
67e989fb 409/*
814fafa7
NC
410Implemented as a macro in utf8.h
411
87cea99e 412=for apidoc is_utf8_string_loc
814fafa7 413
a1433954
KW
414Like L</is_utf8_string> but stores the location of the failure (in the
415case of "utf8ness failure") or the location C<s>+C<len> (in the case of
814fafa7
NC
416"utf8ness success") in the C<ep>.
417
a1433954 418See also L</is_utf8_string_loclen>() and L</is_utf8_string>().
814fafa7 419
87cea99e 420=for apidoc is_utf8_string_loclen
81cd54e3 421
a1433954
KW
422Like L</is_utf8_string>() but stores the location of the failure (in the
423case of "utf8ness failure") or the location C<s>+C<len> (in the case of
768c67ee
JH
424"utf8ness success") in the C<ep>, and the number of UTF-8
425encoded characters in the C<el>.
426
a1433954 427See also L</is_utf8_string_loc>() and L</is_utf8_string>().
81cd54e3
JH
428
429=cut
430*/
431
432bool
668b6d8d 433Perl_is_utf8_string_loclen(const U8 *s, STRLEN len, const U8 **ep, STRLEN *el)
81cd54e3 434{
35da51f7 435 const U8* const send = s + (len ? len : strlen((const char *)s));
7fc63493 436 const U8* x = s;
81cd54e3 437 STRLEN c;
3ebfea28 438 STRLEN outlen = 0;
7918f24d
NC
439
440 PERL_ARGS_ASSERT_IS_UTF8_STRING_LOCLEN;
81cd54e3 441
81cd54e3 442 while (x < send) {
e0328548
KW
443 const U8* next_char_ptr;
444
81cd54e3
JH
445 /* Inline the easy bits of is_utf8_char() here for speed... */
446 if (UTF8_IS_INVARIANT(*x))
e0328548 447 next_char_ptr = x + 1;
81cd54e3 448 else {
768c67ee 449 /* ... and call is_utf8_char() only if really needed. */
768c67ee 450 c = UTF8SKIP(x);
e0328548
KW
451 next_char_ptr = c + x;
452 if (next_char_ptr > send) {
453 goto out;
454 }
768c67ee
JH
455 if (IS_UTF8_CHAR_FAST(c)) {
456 if (!IS_UTF8_CHAR(x, c))
457 c = 0;
458 } else
459 c = is_utf8_char_slow(x, c);
768c67ee
JH
460 if (!c)
461 goto out;
81cd54e3 462 }
e0328548 463 x = next_char_ptr;
3ebfea28 464 outlen++;
81cd54e3 465 }
768c67ee
JH
466
467 out:
3ebfea28
AL
468 if (el)
469 *el = outlen;
470
768c67ee
JH
471 if (ep)
472 *ep = x;
3ebfea28 473 return (x == send);
81cd54e3
JH
474}
475
476/*
768c67ee 477
87cea99e 478=for apidoc utf8n_to_uvuni
67e989fb 479
9041c2e3 480Bottom level UTF-8 decode routine.
746afd53
KW
481Returns the code point value of the first character in the string C<s>,
482which is assumed to be in UTF-8 (or UTF-EBCDIC) encoding, and no longer than
483C<curlen> bytes; C<*retlen> (if C<retlen> isn't NULL) will be set to
484the length, in bytes, of that character.
949cf498
KW
485
486The value of C<flags> determines the behavior when C<s> does not point to a
487well-formed UTF-8 character. If C<flags> is 0, when a malformation is found,
524080c4
KW
488zero is returned and C<*retlen> is set so that (S<C<s> + C<*retlen>>) is the
489next possible position in C<s> that could begin a non-malformed character.
490Also, if UTF-8 warnings haven't been lexically disabled, a warning is raised.
949cf498
KW
491
492Various ALLOW flags can be set in C<flags> to allow (and not warn on)
493individual types of malformations, such as the sequence being overlong (that
494is, when there is a shorter sequence that can express the same code point;
495overlong sequences are expressly forbidden in the UTF-8 standard due to
496potential security issues). Another malformation example is the first byte of
497a character not being a legal first byte. See F<utf8.h> for the list of such
524080c4
KW
498flags. For allowed 0 length strings, this function returns 0; for allowed
499overlong sequences, the computed code point is returned; for all other allowed
500malformations, the Unicode REPLACEMENT CHARACTER is returned, as these have no
501determinable reasonable value.
949cf498
KW
502
503The UTF8_CHECK_ONLY flag overrides the behavior when a non-allowed (by other
504flags) malformation is found. If this flag is set, the routine assumes that
505the caller will raise a warning, and this function will silently just set
d088425d
KW
506C<retlen> to C<-1> (cast to C<STRLEN>) and return zero.
507
508Note that this API requires disambiguation between successful decoding a NUL
509character, and an error return (unless the UTF8_CHECK_ONLY flag is set), as
510in both cases, 0 is returned. To disambiguate, upon a zero return, see if the
511first byte of C<s> is 0 as well. If so, the input was a NUL; if not, the input
512had an error.
949cf498
KW
513
514Certain code points are considered problematic. These are Unicode surrogates,
746afd53 515Unicode non-characters, and code points above the Unicode maximum of 0x10FFFF.
949cf498 516By default these are considered regular code points, but certain situations
5eafe189 517warrant special handling for them. If C<flags> contains
949cf498
KW
518UTF8_DISALLOW_ILLEGAL_INTERCHANGE, all three classes are treated as
519malformations and handled as such. The flags UTF8_DISALLOW_SURROGATE,
520UTF8_DISALLOW_NONCHAR, and UTF8_DISALLOW_SUPER (meaning above the legal Unicode
521maximum) can be set to disallow these categories individually.
522
523The flags UTF8_WARN_ILLEGAL_INTERCHANGE, UTF8_WARN_SURROGATE,
524UTF8_WARN_NONCHAR, and UTF8_WARN_SUPER will cause warning messages to be raised
525for their respective categories, but otherwise the code points are considered
526valid (not malformations). To get a category to both be treated as a
527malformation and raise a warning, specify both the WARN and DISALLOW flags.
528(But note that warnings are not raised if lexically disabled nor if
529UTF8_CHECK_ONLY is also specified.)
530
531Very large code points (above 0x7FFF_FFFF) are considered more problematic than
532the others that are above the Unicode legal maximum. There are several
eb83ed87
KW
533reasons: they requre at least 32 bits to represent them on ASCII platforms, are
534not representable at all on EBCDIC platforms, and the original UTF-8
535specification never went above this number (the current 0x10FFFF limit was
536imposed later). (The smaller ones, those that fit into 32 bits, are
537representable by a UV on ASCII platforms, but not by an IV, which means that
538the number of operations that can be performed on them is quite restricted.)
539The UTF-8 encoding on ASCII platforms for these large code points begins with a
540byte containing 0xFE or 0xFF. The UTF8_DISALLOW_FE_FF flag will cause them to
541be treated as malformations, while allowing smaller above-Unicode code points.
542(Of course UTF8_DISALLOW_SUPER will treat all above-Unicode code points,
543including these, as malformations.) Similarly, UTF8_WARN_FE_FF acts just like
544the other WARN flags, but applies just to these code points.
949cf498
KW
545
546All other code points corresponding to Unicode characters, including private
547use and those yet to be assigned, are never considered malformed and never
548warn.
67e989fb 549
ec5f19d0 550Most code should use L</utf8_to_uvchr_buf>() rather than call this directly.
9041c2e3 551
37607a96
PK
552=cut
553*/
67e989fb 554
a0ed51b3 555UV
7fc63493 556Perl_utf8n_to_uvuni(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen, U32 flags)
a0ed51b3 557{
97aff369 558 dVAR;
d4c19fe8 559 const U8 * const s0 = s;
eb83ed87 560 U8 overflow_byte = '\0'; /* Save byte in case of overflow */
0b8d30e8 561 U8 * send;
eb83ed87
KW
562 UV uv = *s;
563 STRLEN expectlen;
949cf498 564 SV* sv = NULL;
eb83ed87
KW
565 UV outlier_ret = 0; /* return value when input is in error or problematic
566 */
567 UV pack_warn = 0; /* Save result of packWARN() for later */
568 bool unexpected_non_continuation = FALSE;
569 bool overflowed = FALSE;
2f8f112e 570 bool do_overlong_test = TRUE; /* May have to skip this test */
a0dbb045 571
eb83ed87 572 const char* const malformed_text = "Malformed UTF-8 character";
7918f24d 573
eb83ed87 574 PERL_ARGS_ASSERT_UTF8N_TO_UVUNI;
a0dbb045 575
eb83ed87
KW
576 /* The order of malformation tests here is important. We should consume as
577 * few bytes as possible in order to not skip any valid character. This is
578 * required by the Unicode Standard (section 3.9 of Unicode 6.0); see also
579 * http://unicode.org/reports/tr36 for more discussion as to why. For
580 * example, once we've done a UTF8SKIP, we can tell the expected number of
581 * bytes, and could fail right off the bat if the input parameters indicate
582 * that there are too few available. But it could be that just that first
583 * byte is garbled, and the intended character occupies fewer bytes. If we
584 * blindly assumed that the first byte is correct, and skipped based on
585 * that number, we could skip over a valid input character. So instead, we
586 * always examine the sequence byte-by-byte.
587 *
588 * We also should not consume too few bytes, otherwise someone could inject
589 * things. For example, an input could be deliberately designed to
590 * overflow, and if this code bailed out immediately upon discovering that,
591 * returning to the caller *retlen pointing to the very next byte (one
592 * which is actually part of of the overflowing sequence), that could look
593 * legitimate to the caller, which could discard the initial partial
594 * sequence and process the rest, inappropriately */
595
596 /* Zero length strings, if allowed, of necessity are zero */
b5b9af04 597 if (UNLIKELY(curlen == 0)) {
eb83ed87
KW
598 if (retlen) {
599 *retlen = 0;
600 }
a0dbb045 601
eb83ed87
KW
602 if (flags & UTF8_ALLOW_EMPTY) {
603 return 0;
604 }
605 if (! (flags & UTF8_CHECK_ONLY)) {
606 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (empty string)", malformed_text));
607 }
0c443dc2
JH
608 goto malformed;
609 }
610
eb83ed87
KW
611 expectlen = UTF8SKIP(s);
612
613 /* A well-formed UTF-8 character, as the vast majority of calls to this
614 * function will be for, has this expected length. For efficiency, set
615 * things up here to return it. It will be overriden only in those rare
616 * cases where a malformation is found */
617 if (retlen) {
618 *retlen = expectlen;
619 }
620
621 /* An invariant is trivially well-formed */
1d72bdf6 622 if (UTF8_IS_INVARIANT(uv)) {
c4d5f83a 623 return (UV) (NATIVE_TO_UTF(*s));
a0ed51b3 624 }
67e989fb 625
eb83ed87 626 /* A continuation character can't start a valid sequence */
b5b9af04 627 if (UNLIKELY(UTF8_IS_CONTINUATION(uv))) {
eb83ed87
KW
628 if (flags & UTF8_ALLOW_CONTINUATION) {
629 if (retlen) {
630 *retlen = 1;
631 }
632 return UNICODE_REPLACEMENT;
633 }
ba210ebe 634
eb83ed87
KW
635 if (! (flags & UTF8_CHECK_ONLY)) {
636 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (unexpected continuation byte 0x%02x, with no preceding start byte)", malformed_text, *s0));
637 }
638 curlen = 1;
ba210ebe
JH
639 goto malformed;
640 }
9041c2e3 641
1d72bdf6 642#ifdef EBCDIC
75383841 643 uv = NATIVE_TO_UTF(uv);
1d72bdf6
NIS
644#endif
645
eb83ed87
KW
646 /* Here is not a continuation byte, nor an invariant. The only thing left
647 * is a start byte (possibly for an overlong) */
ba210ebe 648
eb83ed87
KW
649 /* Remove the leading bits that indicate the number of bytes in the
650 * character's whole UTF-8 sequence, leaving just the bits that are part of
651 * the value */
652 uv &= UTF_START_MASK(expectlen);
ba210ebe 653
eb83ed87
KW
654 /* Now, loop through the remaining bytes in the character's sequence,
655 * accumulating each into the working value as we go. Be sure to not look
656 * past the end of the input string */
0b8d30e8
KW
657 send = (U8*) s0 + ((expectlen <= curlen) ? expectlen : curlen);
658
eb83ed87 659 for (s = s0 + 1; s < send; s++) {
b5b9af04 660 if (LIKELY(UTF8_IS_CONTINUATION(*s))) {
eb83ed87
KW
661#ifndef EBCDIC /* Can't overflow in EBCDIC */
662 if (uv & UTF_ACCUMULATION_OVERFLOW_MASK) {
663
664 /* The original implementors viewed this malformation as more
665 * serious than the others (though I, khw, don't understand
666 * why, since other malformations also give very very wrong
667 * results), so there is no way to turn off checking for it.
668 * Set a flag, but keep going in the loop, so that we absorb
669 * the rest of the bytes that comprise the character. */
670 overflowed = TRUE;
671 overflow_byte = *s; /* Save for warning message's use */
672 }
673#endif
8850bf83 674 uv = UTF8_ACCUMULATE(uv, *s);
eb83ed87
KW
675 }
676 else {
677 /* Here, found a non-continuation before processing all expected
678 * bytes. This byte begins a new character, so quit, even if
679 * allowing this malformation. */
680 unexpected_non_continuation = TRUE;
681 break;
682 }
683 } /* End of loop through the character's bytes */
684
685 /* Save how many bytes were actually in the character */
686 curlen = s - s0;
687
688 /* The loop above finds two types of malformations: non-continuation and/or
689 * overflow. The non-continuation malformation is really a too-short
690 * malformation, as it means that the current character ended before it was
691 * expected to (being terminated prematurely by the beginning of the next
692 * character, whereas in the too-short malformation there just are too few
693 * bytes available to hold the character. In both cases, the check below
694 * that we have found the expected number of bytes would fail if executed.)
695 * Thus the non-continuation malformation is really unnecessary, being a
696 * subset of the too-short malformation. But there may be existing
697 * applications that are expecting the non-continuation type, so we retain
698 * it, and return it in preference to the too-short malformation. (If this
699 * code were being written from scratch, the two types might be collapsed
700 * into one.) I, khw, am also giving priority to returning the
701 * non-continuation and too-short malformations over overflow when multiple
702 * ones are present. I don't know of any real reason to prefer one over
703 * the other, except that it seems to me that multiple-byte errors trumps
704 * errors from a single byte */
b5b9af04 705 if (UNLIKELY(unexpected_non_continuation)) {
eb83ed87
KW
706 if (!(flags & UTF8_ALLOW_NON_CONTINUATION)) {
707 if (! (flags & UTF8_CHECK_ONLY)) {
708 if (curlen == 1) {
709 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (unexpected non-continuation byte 0x%02x, immediately after start byte 0x%02x)", malformed_text, *s, *s0));
710 }
711 else {
712 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
713 }
714 }
eb83ed87
KW
715 goto malformed;
716 }
717 uv = UNICODE_REPLACEMENT;
2f8f112e
KW
718
719 /* Skip testing for overlongs, as the REPLACEMENT may not be the same
720 * as what the original expectations were. */
721 do_overlong_test = FALSE;
eb83ed87
KW
722 if (retlen) {
723 *retlen = curlen;
724 }
725 }
b5b9af04 726 else if (UNLIKELY(curlen < expectlen)) {
eb83ed87
KW
727 if (! (flags & UTF8_ALLOW_SHORT)) {
728 if (! (flags & UTF8_CHECK_ONLY)) {
729 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 730 }
eb83ed87
KW
731 goto malformed;
732 }
733 uv = UNICODE_REPLACEMENT;
2f8f112e 734 do_overlong_test = FALSE;
eb83ed87
KW
735 if (retlen) {
736 *retlen = curlen;
737 }
738 }
739
740#ifndef EBCDIC /* EBCDIC allows FE, FF, can't overflow */
2f8f112e 741 if ((*s0 & 0xFE) == 0xFE /* matches both FE, FF */
eb83ed87
KW
742 && (flags & (UTF8_WARN_FE_FF|UTF8_DISALLOW_FE_FF)))
743 {
744 /* By adding UTF8_CHECK_ONLY to the test, we avoid unnecessary
745 * generation of the sv, since no warnings are raised under CHECK */
746 if ((flags & (UTF8_WARN_FE_FF|UTF8_CHECK_ONLY)) == UTF8_WARN_FE_FF
747 && ckWARN_d(WARN_UTF8))
748 {
42303544
KW
749 /* This message is deliberately not of the same syntax as the other
750 * messages for malformations, for backwards compatibility in the
751 * unlikely event that code is relying on its precise earlier text
752 */
eb83ed87
KW
753 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s Code point beginning with byte 0x%02X is not Unicode, and not portable", malformed_text, *s0));
754 pack_warn = packWARN(WARN_UTF8);
755 }
756 if (flags & UTF8_DISALLOW_FE_FF) {
757 goto malformed;
ba210ebe 758 }
ba210ebe 759 }
b5b9af04 760 if (UNLIKELY(overflowed)) {
ba210ebe 761
eb83ed87
KW
762 /* If the first byte is FF, it will overflow a 32-bit word. If the
763 * first byte is FE, it will overflow a signed 32-bit word. The
764 * above preserves backward compatibility, since its message was used
765 * in earlier versions of this code in preference to overflow */
766 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (overflow at byte 0x%02x, after start byte 0x%02x)", malformed_text, overflow_byte, *s0));
ba210ebe 767 goto malformed;
eb83ed87
KW
768 }
769#endif
770
2f8f112e
KW
771 if (do_overlong_test
772 && expectlen > (STRLEN)UNISKIP(uv)
773 && ! (flags & UTF8_ALLOW_LONG))
774 {
eb83ed87
KW
775 /* The overlong malformation has lower precedence than the others.
776 * Note that if this malformation is allowed, we return the actual
777 * value, instead of the replacement character. This is because this
778 * value is actually well-defined. */
779 if (! (flags & UTF8_CHECK_ONLY)) {
780 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));
781 }
782 goto malformed;
783 }
784
785 /* Here, the input is considered to be well-formed , but could be a
786 * problematic code point that is not allowed by the input parameters. */
787 if (uv >= UNICODE_SURROGATE_FIRST /* isn't problematic if < this */
788 && (flags & (UTF8_DISALLOW_ILLEGAL_INTERCHANGE
789 |UTF8_WARN_ILLEGAL_INTERCHANGE)))
790 {
949cf498 791 if (UNICODE_IS_SURROGATE(uv)) {
eb83ed87
KW
792 if ((flags & (UTF8_WARN_SURROGATE|UTF8_CHECK_ONLY)) == UTF8_WARN_SURROGATE
793 && ckWARN2_d(WARN_UTF8, WARN_SURROGATE))
794 {
111d382d 795 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "UTF-16 surrogate U+%04"UVXf"", uv));
eb83ed87 796 pack_warn = packWARN2(WARN_UTF8, WARN_SURROGATE);
949cf498
KW
797 }
798 if (flags & UTF8_DISALLOW_SURROGATE) {
799 goto disallowed;
800 }
801 }
949cf498 802 else if ((uv > PERL_UNICODE_MAX)) {
eb83ed87
KW
803 if ((flags & (UTF8_WARN_SUPER|UTF8_CHECK_ONLY)) == UTF8_WARN_SUPER
804 && ckWARN2_d(WARN_UTF8, WARN_NON_UNICODE))
805 {
111d382d 806 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "Code point 0x%04"UVXf" is not Unicode, may not be portable", uv));
eb83ed87 807 pack_warn = packWARN2(WARN_UTF8, WARN_NON_UNICODE);
949cf498
KW
808 }
809 if (flags & UTF8_DISALLOW_SUPER) {
810 goto disallowed;
811 }
812 }
4190d317
KW
813 else if (UNICODE_IS_NONCHAR(uv)) {
814 if ((flags & (UTF8_WARN_NONCHAR|UTF8_CHECK_ONLY)) == UTF8_WARN_NONCHAR
815 && ckWARN2_d(WARN_UTF8, WARN_NONCHAR))
816 {
817 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "Unicode non-character U+%04"UVXf" is illegal for open interchange", uv));
818 pack_warn = packWARN2(WARN_UTF8, WARN_NONCHAR);
819 }
820 if (flags & UTF8_DISALLOW_NONCHAR) {
821 goto disallowed;
822 }
823 }
949cf498 824
eb83ed87
KW
825 if (sv) {
826 outlier_ret = uv;
827 goto do_warn;
828 }
829
949cf498
KW
830 /* Here, this is not considered a malformed character, so drop through
831 * to return it */
a0ed51b3 832 }
ba210ebe 833
a0ed51b3 834 return uv;
ba210ebe 835
eb83ed87
KW
836 /* There are three cases which get to beyond this point. In all 3 cases:
837 * <sv> if not null points to a string to print as a warning.
838 * <curlen> is what <*retlen> should be set to if UTF8_CHECK_ONLY isn't
839 * set.
840 * <outlier_ret> is what return value to use if UTF8_CHECK_ONLY isn't set.
841 * This is done by initializing it to 0, and changing it only
842 * for case 1).
843 * The 3 cases are:
844 * 1) The input is valid but problematic, and to be warned about. The
845 * return value is the resultant code point; <*retlen> is set to
846 * <curlen>, the number of bytes that comprise the code point.
847 * <pack_warn> contains the result of packWARN() for the warning
848 * types. The entry point for this case is the label <do_warn>;
849 * 2) The input is a valid code point but disallowed by the parameters to
850 * this function. The return value is 0. If UTF8_CHECK_ONLY is set,
851 * <*relen> is -1; otherwise it is <curlen>, the number of bytes that
852 * comprise the code point. <pack_warn> contains the result of
853 * packWARN() for the warning types. The entry point for this case is
854 * the label <disallowed>.
855 * 3) The input is malformed. The return value is 0. If UTF8_CHECK_ONLY
856 * is set, <*relen> is -1; otherwise it is <curlen>, the number of
857 * bytes that comprise the malformation. All such malformations are
858 * assumed to be warning type <utf8>. The entry point for this case
859 * is the label <malformed>.
860 */
949cf498 861
ba210ebe
JH
862malformed:
863
eb83ed87
KW
864 if (sv && ckWARN_d(WARN_UTF8)) {
865 pack_warn = packWARN(WARN_UTF8);
866 }
867
868disallowed:
869
fcc8fcf6 870 if (flags & UTF8_CHECK_ONLY) {
ba210ebe 871 if (retlen)
10edeb5d 872 *retlen = ((STRLEN) -1);
ba210ebe
JH
873 return 0;
874 }
875
eb83ed87 876do_warn:
5b311467 877
eb83ed87
KW
878 if (pack_warn) { /* <pack_warn> was initialized to 0, and changed only
879 if warnings are to be raised. */
f555bc63 880 const char * const string = SvPVX_const(sv);
a0dbb045 881
f555bc63
KW
882 if (PL_op)
883 Perl_warner(aTHX_ pack_warn, "%s in %s", string, OP_DESC(PL_op));
884 else
885 Perl_warner(aTHX_ pack_warn, "%s", string);
a0dbb045
JH
886 }
887
eb83ed87
KW
888 if (retlen) {
889 *retlen = curlen;
890 }
ba210ebe 891
eb83ed87 892 return outlier_ret;
a0ed51b3
LW
893}
894
8e84507e 895/*
ec5f19d0
KW
896=for apidoc utf8_to_uvchr_buf
897
898Returns the native code point of the first character in the string C<s> which
899is assumed to be in UTF-8 encoding; C<send> points to 1 beyond the end of C<s>.
524080c4 900C<*retlen> will be set to the length, in bytes, of that character.
ec5f19d0 901
524080c4
KW
902If C<s> does not point to a well-formed UTF-8 character and UTF8 warnings are
903enabled, zero is returned and C<*retlen> is set (if C<retlen> isn't
904NULL) to -1. If those warnings are off, the computed value if well-defined (or
905the Unicode REPLACEMENT CHARACTER, if not) is silently returned, and C<*retlen>
906is set (if C<retlen> isn't NULL) so that (S<C<s> + C<*retlen>>) is the
907next possible position in C<s> that could begin a non-malformed character.
908See L</utf8n_to_uvuni> for details on when the REPLACEMENT CHARACTER is returned.
ec5f19d0
KW
909
910=cut
911*/
912
913
914UV
915Perl_utf8_to_uvchr_buf(pTHX_ const U8 *s, const U8 *send, STRLEN *retlen)
916{
917 PERL_ARGS_ASSERT_UTF8_TO_UVCHR_BUF;
918
919 assert(s < send);
920
921 return utf8n_to_uvchr(s, send - s, retlen,
922 ckWARN_d(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
923}
924
27d6c58a 925/* Like L</utf8_to_uvchr_buf>(), but should only be called when it is known that
3986bb7c 926 * there are no malformations in the input UTF-8 string C<s>. surrogates,
6dd9dce9
KW
927 * non-character code points, and non-Unicode code points are allowed. A macro
928 * in utf8.h is used to normally avoid this function wrapper */
27d6c58a
KW
929
930UV
931Perl_valid_utf8_to_uvchr(pTHX_ const U8 *s, STRLEN *retlen)
932{
3986bb7c
KW
933 const UV uv = valid_utf8_to_uvuni(s, retlen);
934
27d6c58a
KW
935 PERL_ARGS_ASSERT_VALID_UTF8_TO_UVCHR;
936
3986bb7c 937 return UNI_TO_NATIVE(uv);
27d6c58a
KW
938}
939
ec5f19d0 940/*
87cea99e 941=for apidoc utf8_to_uvchr
9041c2e3 942
977c1d31
KW
943DEPRECATED!
944
6ee84de2 945Returns the native code point of the first character in the string C<s>
1e54db1a 946which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
9041c2e3
NIS
947length, in bytes, of that character.
948
4b88fb76 949Some, but not all, UTF-8 malformations are detected, and in fact, some
977c1d31
KW
950malformed input could cause reading beyond the end of the input buffer, which
951is why this function is deprecated. Use L</utf8_to_uvchr_buf> instead.
4b88fb76 952
524080c4
KW
953If C<s> points to one of the detected malformations, and UTF8 warnings are
954enabled, zero is returned and C<*retlen> is set (if C<retlen> isn't
955NULL) to -1. If those warnings are off, the computed value if well-defined (or
956the Unicode REPLACEMENT CHARACTER, if not) is silently returned, and C<*retlen>
957is set (if C<retlen> isn't NULL) so that (S<C<s> + C<*retlen>>) is the
958next possible position in C<s> that could begin a non-malformed character.
959See L</utf8n_to_uvuni> for details on when the REPLACEMENT CHARACTER is returned.
9041c2e3
NIS
960
961=cut
962*/
963
964UV
7fc63493 965Perl_utf8_to_uvchr(pTHX_ const U8 *s, STRLEN *retlen)
9041c2e3 966{
7918f24d
NC
967 PERL_ARGS_ASSERT_UTF8_TO_UVCHR;
968
2ff6c191 969 return utf8_to_uvchr_buf(s, s + UTF8_MAXBYTES, retlen);
9041c2e3
NIS
970}
971
972/*
ec5f19d0
KW
973=for apidoc utf8_to_uvuni_buf
974
975Returns the Unicode code point of the first character in the string C<s> which
976is assumed to be in UTF-8 encoding; C<send> points to 1 beyond the end of C<s>.
977C<retlen> will be set to the length, in bytes, of that character.
978
979This function should only be used when the returned UV is considered
980an index into the Unicode semantic tables (e.g. swashes).
981
524080c4
KW
982If C<s> does not point to a well-formed UTF-8 character and UTF8 warnings are
983enabled, zero is returned and C<*retlen> is set (if C<retlen> isn't
984NULL) to -1. If those warnings are off, the computed value if well-defined (or
985the Unicode REPLACEMENT CHARACTER, if not) is silently returned, and C<*retlen>
986is set (if C<retlen> isn't NULL) so that (S<C<s> + C<*retlen>>) is the
987next possible position in C<s> that could begin a non-malformed character.
988See L</utf8n_to_uvuni> for details on when the REPLACEMENT CHARACTER is returned.
ec5f19d0
KW
989
990=cut
991*/
992
993UV
994Perl_utf8_to_uvuni_buf(pTHX_ const U8 *s, const U8 *send, STRLEN *retlen)
995{
996 PERL_ARGS_ASSERT_UTF8_TO_UVUNI_BUF;
997
998 assert(send > s);
999
1000 /* Call the low level routine asking for checks */
1001 return Perl_utf8n_to_uvuni(aTHX_ s, send -s, retlen,
1002 ckWARN_d(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
1003}
1004
27d6c58a 1005/* Like L</utf8_to_uvuni_buf>(), but should only be called when it is known that
2114036c 1006 * there are no malformations in the input UTF-8 string C<s>. Surrogates,
3986bb7c 1007 * non-character code points, and non-Unicode code points are allowed */
27d6c58a
KW
1008
1009UV
1010Perl_valid_utf8_to_uvuni(pTHX_ const U8 *s, STRLEN *retlen)
1011{
3986bb7c
KW
1012 UV expectlen = UTF8SKIP(s);
1013 const U8* send = s + expectlen;
1014 UV uv = NATIVE_TO_UTF(*s);
1015
27d6c58a
KW
1016 PERL_ARGS_ASSERT_VALID_UTF8_TO_UVUNI;
1017
3986bb7c
KW
1018 if (retlen) {
1019 *retlen = expectlen;
1020 }
1021
1022 /* An invariant is trivially returned */
1023 if (expectlen == 1) {
1024 return uv;
1025 }
1026
1027 /* Remove the leading bits that indicate the number of bytes, leaving just
1028 * the bits that are part of the value */
1029 uv &= UTF_START_MASK(expectlen);
1030
1031 /* Now, loop through the remaining bytes, accumulating each into the
08bc774e
KW
1032 * working total as we go. (I khw tried unrolling the loop for up to 4
1033 * bytes, but there was no performance improvement) */
3986bb7c
KW
1034 for (++s; s < send; s++) {
1035 uv = UTF8_ACCUMULATE(uv, *s);
1036 }
1037
1038 return uv;
27d6c58a
KW
1039}
1040
ec5f19d0 1041/*
87cea99e 1042=for apidoc utf8_to_uvuni
9041c2e3 1043
977c1d31
KW
1044DEPRECATED!
1045
9041c2e3 1046Returns the Unicode code point of the first character in the string C<s>
1e54db1a 1047which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
9041c2e3
NIS
1048length, in bytes, of that character.
1049
39e518fd
KW
1050This function should only be used when the returned UV is considered
1051an index into the Unicode semantic tables (e.g. swashes).
1052
4b88fb76 1053Some, but not all, UTF-8 malformations are detected, and in fact, some
977c1d31
KW
1054malformed input could cause reading beyond the end of the input buffer, which
1055is why this function is deprecated. Use L</utf8_to_uvuni_buf> instead.
9041c2e3 1056
524080c4
KW
1057If C<s> points to one of the detected malformations, and UTF8 warnings are
1058enabled, zero is returned and C<*retlen> is set (if C<retlen> doesn't point to
1059NULL) to -1. If those warnings are off, the computed value if well-defined (or
1060the Unicode REPLACEMENT CHARACTER, if not) is silently returned, and C<*retlen>
1061is set (if C<retlen> isn't NULL) so that (S<C<s> + C<*retlen>>) is the
1062next possible position in C<s> that could begin a non-malformed character.
1063See L</utf8n_to_uvuni> for details on when the REPLACEMENT CHARACTER is returned.
8e84507e
NIS
1064
1065=cut
1066*/
1067
1068UV
7fc63493 1069Perl_utf8_to_uvuni(pTHX_ const U8 *s, STRLEN *retlen)
8e84507e 1070{
7918f24d
NC
1071 PERL_ARGS_ASSERT_UTF8_TO_UVUNI;
1072
4b88fb76 1073 return valid_utf8_to_uvuni(s, retlen);
8e84507e
NIS
1074}
1075
b76347f2 1076/*
87cea99e 1077=for apidoc utf8_length
b76347f2
JH
1078
1079Return the length of the UTF-8 char encoded string C<s> in characters.
02eb7b47
JH
1080Stops at C<e> (inclusive). If C<e E<lt> s> or if the scan would end
1081up past C<e>, croaks.
b76347f2
JH
1082
1083=cut
1084*/
1085
1086STRLEN
35a4481c 1087Perl_utf8_length(pTHX_ const U8 *s, const U8 *e)
b76347f2 1088{
97aff369 1089 dVAR;
b76347f2
JH
1090 STRLEN len = 0;
1091
7918f24d
NC
1092 PERL_ARGS_ASSERT_UTF8_LENGTH;
1093
8850bf83
JH
1094 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g.
1095 * the bitops (especially ~) can create illegal UTF-8.
1096 * In other words: in Perl UTF-8 is not just for Unicode. */
1097
a3b680e6
AL
1098 if (e < s)
1099 goto warn_and_return;
b76347f2 1100 while (s < e) {
4cbf4130 1101 s += UTF8SKIP(s);
8e91ec7f
AV
1102 len++;
1103 }
1104
1105 if (e != s) {
1106 len--;
1107 warn_and_return:
9b387841
NC
1108 if (PL_op)
1109 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8),
1110 "%s in %s", unees, OP_DESC(PL_op));
1111 else
61a12c31 1112 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), "%s", unees);
b76347f2
JH
1113 }
1114
1115 return len;
1116}
1117
b06226ff 1118/*
87cea99e 1119=for apidoc utf8_distance
b06226ff 1120
1e54db1a 1121Returns the number of UTF-8 characters between the UTF-8 pointers C<a>
b06226ff
JH
1122and C<b>.
1123
1124WARNING: use only if you *know* that the pointers point inside the
1125same UTF-8 buffer.
1126
37607a96
PK
1127=cut
1128*/
a0ed51b3 1129
02eb7b47 1130IV
35a4481c 1131Perl_utf8_distance(pTHX_ const U8 *a, const U8 *b)
a0ed51b3 1132{
7918f24d
NC
1133 PERL_ARGS_ASSERT_UTF8_DISTANCE;
1134
bf1665bc 1135 return (a < b) ? -1 * (IV) utf8_length(a, b) : (IV) utf8_length(b, a);
a0ed51b3
LW
1136}
1137
b06226ff 1138/*
87cea99e 1139=for apidoc utf8_hop
b06226ff 1140
8850bf83
JH
1141Return the UTF-8 pointer C<s> displaced by C<off> characters, either
1142forward or backward.
b06226ff
JH
1143
1144WARNING: do not use the following unless you *know* C<off> is within
8850bf83
JH
1145the UTF-8 data pointed to by C<s> *and* that on entry C<s> is aligned
1146on the first byte of character or just after the last byte of a character.
b06226ff 1147
37607a96
PK
1148=cut
1149*/
a0ed51b3
LW
1150
1151U8 *
4373e329 1152Perl_utf8_hop(pTHX_ const U8 *s, I32 off)
a0ed51b3 1153{
7918f24d
NC
1154 PERL_ARGS_ASSERT_UTF8_HOP;
1155
96a5add6 1156 PERL_UNUSED_CONTEXT;
8850bf83
JH
1157 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g
1158 * the bitops (especially ~) can create illegal UTF-8.
1159 * In other words: in Perl UTF-8 is not just for Unicode. */
1160
a0ed51b3
LW
1161 if (off >= 0) {
1162 while (off--)
1163 s += UTF8SKIP(s);
1164 }
1165 else {
1166 while (off++) {
1167 s--;
8850bf83
JH
1168 while (UTF8_IS_CONTINUATION(*s))
1169 s--;
a0ed51b3
LW
1170 }
1171 }
4373e329 1172 return (U8 *)s;
a0ed51b3
LW
1173}
1174
6940069f 1175/*
fed3ba5d
NC
1176=for apidoc bytes_cmp_utf8
1177
a1433954
KW
1178Compares the sequence of characters (stored as octets) in C<b>, C<blen> with the
1179sequence of characters (stored as UTF-8) in C<u>, C<ulen>. Returns 0 if they are
fed3ba5d
NC
1180equal, -1 or -2 if the first string is less than the second string, +1 or +2
1181if the first string is greater than the second string.
1182
1183-1 or +1 is returned if the shorter string was identical to the start of the
1184longer string. -2 or +2 is returned if the was a difference between characters
1185within the strings.
1186
1187=cut
1188*/
1189
1190int
1191Perl_bytes_cmp_utf8(pTHX_ const U8 *b, STRLEN blen, const U8 *u, STRLEN ulen)
1192{
1193 const U8 *const bend = b + blen;
1194 const U8 *const uend = u + ulen;
1195
1196 PERL_ARGS_ASSERT_BYTES_CMP_UTF8;
1197
1198 PERL_UNUSED_CONTEXT;
1199
1200 while (b < bend && u < uend) {
1201 U8 c = *u++;
1202 if (!UTF8_IS_INVARIANT(c)) {
1203 if (UTF8_IS_DOWNGRADEABLE_START(c)) {
1204 if (u < uend) {
1205 U8 c1 = *u++;
1206 if (UTF8_IS_CONTINUATION(c1)) {
356979f4 1207 c = UNI_TO_NATIVE(TWO_BYTE_UTF8_TO_UNI(c, c1));
fed3ba5d
NC
1208 } else {
1209 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8),
1210 "Malformed UTF-8 character "
1211 "(unexpected non-continuation byte 0x%02x"
1212 ", immediately after start byte 0x%02x)"
1213 /* Dear diag.t, it's in the pod. */
1214 "%s%s", c1, c,
1215 PL_op ? " in " : "",
1216 PL_op ? OP_DESC(PL_op) : "");
1217 return -2;
1218 }
1219 } else {
1220 if (PL_op)
1221 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8),
1222 "%s in %s", unees, OP_DESC(PL_op));
1223 else
61a12c31 1224 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), "%s", unees);
fed3ba5d
NC
1225 return -2; /* Really want to return undef :-) */
1226 }
1227 } else {
1228 return -2;
1229 }
1230 }
1231 if (*b != c) {
1232 return *b < c ? -2 : +2;
1233 }
1234 ++b;
1235 }
1236
1237 if (b == bend && u == uend)
1238 return 0;
1239
1240 return b < bend ? +1 : -1;
1241}
1242
1243/*
87cea99e 1244=for apidoc utf8_to_bytes
6940069f 1245
2bbc8d55 1246Converts a string C<s> of length C<len> from UTF-8 into native byte encoding.
a1433954
KW
1247Unlike L</bytes_to_utf8>, this over-writes the original string, and
1248updates C<len> to contain the new length.
67e989fb 1249Returns zero on failure, setting C<len> to -1.
6940069f 1250
a1433954 1251If you need a copy of the string, see L</bytes_from_utf8>.
95be277c 1252
6940069f
GS
1253=cut
1254*/
1255
1256U8 *
37607a96 1257Perl_utf8_to_bytes(pTHX_ U8 *s, STRLEN *len)
6940069f 1258{
d4c19fe8
AL
1259 U8 * const save = s;
1260 U8 * const send = s + *len;
6940069f 1261 U8 *d;
246fae53 1262
7918f24d
NC
1263 PERL_ARGS_ASSERT_UTF8_TO_BYTES;
1264
1e54db1a 1265 /* ensure valid UTF-8 and chars < 256 before updating string */
d4c19fe8 1266 while (s < send) {
dcad2880
JH
1267 U8 c = *s++;
1268
1d72bdf6
NIS
1269 if (!UTF8_IS_INVARIANT(c) &&
1270 (!UTF8_IS_DOWNGRADEABLE_START(c) || (s >= send)
1271 || !(c = *s++) || !UTF8_IS_CONTINUATION(c))) {
10edeb5d 1272 *len = ((STRLEN) -1);
dcad2880
JH
1273 return 0;
1274 }
246fae53 1275 }
dcad2880
JH
1276
1277 d = s = save;
6940069f 1278 while (s < send) {
ed646e6e 1279 STRLEN ulen;
4b88fb76 1280 *d++ = (U8)utf8_to_uvchr_buf(s, send, &ulen);
ed646e6e 1281 s += ulen;
6940069f
GS
1282 }
1283 *d = '\0';
246fae53 1284 *len = d - save;
6940069f
GS
1285 return save;
1286}
1287
1288/*
87cea99e 1289=for apidoc bytes_from_utf8
f9a63242 1290
2bbc8d55 1291Converts a string C<s> of length C<len> from UTF-8 into native byte encoding.
a1433954 1292Unlike L</utf8_to_bytes> but like L</bytes_to_utf8>, returns a pointer to
ef9edfd0
JH
1293the newly-created string, and updates C<len> to contain the new
1294length. Returns the original string if no conversion occurs, C<len>
1295is unchanged. Do nothing if C<is_utf8> points to 0. Sets C<is_utf8> to
2bbc8d55
SP
12960 if C<s> is converted or consisted entirely of characters that are invariant
1297in utf8 (i.e., US-ASCII on non-EBCDIC machines).
f9a63242 1298
37607a96
PK
1299=cut
1300*/
f9a63242
JH
1301
1302U8 *
e1ec3a88 1303Perl_bytes_from_utf8(pTHX_ const U8 *s, STRLEN *len, bool *is_utf8)
f9a63242 1304{
f9a63242 1305 U8 *d;
e1ec3a88
AL
1306 const U8 *start = s;
1307 const U8 *send;
f9a63242
JH
1308 I32 count = 0;
1309
7918f24d
NC
1310 PERL_ARGS_ASSERT_BYTES_FROM_UTF8;
1311
96a5add6 1312 PERL_UNUSED_CONTEXT;
f9a63242 1313 if (!*is_utf8)
73d840c0 1314 return (U8 *)start;
f9a63242 1315
1e54db1a 1316 /* ensure valid UTF-8 and chars < 256 before converting string */
f9a63242 1317 for (send = s + *len; s < send;) {
e1ec3a88 1318 U8 c = *s++;
1d72bdf6 1319 if (!UTF8_IS_INVARIANT(c)) {
db42d148
NIS
1320 if (UTF8_IS_DOWNGRADEABLE_START(c) && s < send &&
1321 (c = *s++) && UTF8_IS_CONTINUATION(c))
1322 count++;
1323 else
73d840c0 1324 return (U8 *)start;
db42d148 1325 }
f9a63242
JH
1326 }
1327
35da51f7 1328 *is_utf8 = FALSE;
f9a63242 1329
212542aa 1330 Newx(d, (*len) - count + 1, U8);
ef9edfd0 1331 s = start; start = d;
f9a63242
JH
1332 while (s < send) {
1333 U8 c = *s++;
c4d5f83a
NIS
1334 if (!UTF8_IS_INVARIANT(c)) {
1335 /* Then it is two-byte encoded */
356979f4 1336 c = UNI_TO_NATIVE(TWO_BYTE_UTF8_TO_UNI(c, *s++));
c4d5f83a
NIS
1337 }
1338 *d++ = c;
f9a63242
JH
1339 }
1340 *d = '\0';
1341 *len = d - start;
73d840c0 1342 return (U8 *)start;
f9a63242
JH
1343}
1344
1345/*
87cea99e 1346=for apidoc bytes_to_utf8
6940069f 1347
ff97e5cf
KW
1348Converts a string C<s> of length C<len> bytes from the native encoding into
1349UTF-8.
6662521e 1350Returns a pointer to the newly-created string, and sets C<len> to
ff97e5cf 1351reflect the new length in bytes.
6940069f 1352
2bbc8d55
SP
1353A NUL character will be written after the end of the string.
1354
1355If you want to convert to UTF-8 from encodings other than
1356the native (Latin1 or EBCDIC),
a1433954 1357see L</sv_recode_to_utf8>().
c9ada85f 1358
497711e7 1359=cut
6940069f
GS
1360*/
1361
c682ebef
FC
1362/* This logic is duplicated in sv_catpvn_flags, so any bug fixes will
1363 likewise need duplication. */
1364
6940069f 1365U8*
35a4481c 1366Perl_bytes_to_utf8(pTHX_ const U8 *s, STRLEN *len)
6940069f 1367{
35a4481c 1368 const U8 * const send = s + (*len);
6940069f
GS
1369 U8 *d;
1370 U8 *dst;
7918f24d
NC
1371
1372 PERL_ARGS_ASSERT_BYTES_TO_UTF8;
96a5add6 1373 PERL_UNUSED_CONTEXT;
6940069f 1374
212542aa 1375 Newx(d, (*len) * 2 + 1, U8);
6940069f
GS
1376 dst = d;
1377
1378 while (s < send) {
35a4481c 1379 const UV uv = NATIVE_TO_ASCII(*s++);
c4d5f83a 1380 if (UNI_IS_INVARIANT(uv))
eb160463 1381 *d++ = (U8)UTF_TO_NATIVE(uv);
6940069f 1382 else {
eb160463
GS
1383 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
1384 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
6940069f
GS
1385 }
1386 }
1387 *d = '\0';
6662521e 1388 *len = d-dst;
6940069f
GS
1389 return dst;
1390}
1391
a0ed51b3 1392/*
dea0fc0b 1393 * Convert native (big-endian) or reversed (little-endian) UTF-16 to UTF-8.
a0ed51b3
LW
1394 *
1395 * Destination must be pre-extended to 3/2 source. Do not use in-place.
1396 * We optimize for native, for obvious reasons. */
1397
1398U8*
dea0fc0b 1399Perl_utf16_to_utf8(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen)
a0ed51b3 1400{
dea0fc0b
JH
1401 U8* pend;
1402 U8* dstart = d;
1403
7918f24d
NC
1404 PERL_ARGS_ASSERT_UTF16_TO_UTF8;
1405
dea0fc0b 1406 if (bytelen & 1)
f5992bc4 1407 Perl_croak(aTHX_ "panic: utf16_to_utf8: odd bytelen %"UVuf, (UV)bytelen);
dea0fc0b
JH
1408
1409 pend = p + bytelen;
1410
a0ed51b3 1411 while (p < pend) {
dea0fc0b
JH
1412 UV uv = (p[0] << 8) + p[1]; /* UTF-16BE */
1413 p += 2;
a0ed51b3 1414 if (uv < 0x80) {
e294cc5d
JH
1415#ifdef EBCDIC
1416 *d++ = UNI_TO_NATIVE(uv);
1417#else
eb160463 1418 *d++ = (U8)uv;
e294cc5d 1419#endif
a0ed51b3
LW
1420 continue;
1421 }
1422 if (uv < 0x800) {
eb160463
GS
1423 *d++ = (U8)(( uv >> 6) | 0xc0);
1424 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
1425 continue;
1426 }
52b9aa85 1427 if (uv >= 0xd800 && uv <= 0xdbff) { /* surrogates */
01ea242b 1428 if (p >= pend) {
dea0fc0b 1429 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
01ea242b
NC
1430 } else {
1431 UV low = (p[0] << 8) + p[1];
1432 p += 2;
52b9aa85 1433 if (low < 0xdc00 || low > 0xdfff)
01ea242b
NC
1434 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
1435 uv = ((uv - 0xd800) << 10) + (low - 0xdc00) + 0x10000;
1436 }
dbde1951
NC
1437 } else if (uv >= 0xdc00 && uv <= 0xdfff) {
1438 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
a0ed51b3
LW
1439 }
1440 if (uv < 0x10000) {
eb160463
GS
1441 *d++ = (U8)(( uv >> 12) | 0xe0);
1442 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
1443 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
1444 continue;
1445 }
1446 else {
eb160463
GS
1447 *d++ = (U8)(( uv >> 18) | 0xf0);
1448 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
1449 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
1450 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
1451 continue;
1452 }
1453 }
dea0fc0b 1454 *newlen = d - dstart;
a0ed51b3
LW
1455 return d;
1456}
1457
1458/* Note: this one is slightly destructive of the source. */
1459
1460U8*
dea0fc0b 1461Perl_utf16_to_utf8_reversed(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen)
a0ed51b3
LW
1462{
1463 U8* s = (U8*)p;
d4c19fe8 1464 U8* const send = s + bytelen;
7918f24d
NC
1465
1466 PERL_ARGS_ASSERT_UTF16_TO_UTF8_REVERSED;
1467
e0ea5e2d
NC
1468 if (bytelen & 1)
1469 Perl_croak(aTHX_ "panic: utf16_to_utf8_reversed: odd bytelen %"UVuf,
1470 (UV)bytelen);
1471
a0ed51b3 1472 while (s < send) {
d4c19fe8 1473 const U8 tmp = s[0];
a0ed51b3
LW
1474 s[0] = s[1];
1475 s[1] = tmp;
1476 s += 2;
1477 }
dea0fc0b 1478 return utf16_to_utf8(p, d, bytelen, newlen);
a0ed51b3
LW
1479}
1480
c3fd2246
KW
1481/* for now these are all defined (inefficiently) in terms of the utf8 versions.
1482 * Note that the macros in handy.h that call these short-circuit calling them
1483 * for Latin-1 range inputs */
a0ed51b3
LW
1484
1485bool
84afefe6 1486Perl_is_uni_alnum(pTHX_ UV c)
a0ed51b3 1487{
89ebb4a3 1488 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1489 uvchr_to_utf8(tmpbuf, c);
a0ed51b3
LW
1490 return is_utf8_alnum(tmpbuf);
1491}
1492
1493bool
5092f92a
KW
1494Perl_is_uni_alnumc(pTHX_ UV c)
1495{
1496 U8 tmpbuf[UTF8_MAXBYTES+1];
1497 uvchr_to_utf8(tmpbuf, c);
1498 return is_utf8_alnumc(tmpbuf);
1499}
1500
1501bool
84afefe6 1502Perl_is_uni_idfirst(pTHX_ UV c)
a0ed51b3 1503{
89ebb4a3 1504 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1505 uvchr_to_utf8(tmpbuf, c);
a0ed51b3
LW
1506 return is_utf8_idfirst(tmpbuf);
1507}
1508
1509bool
f91dcd13
KW
1510Perl__is_uni_perl_idstart(pTHX_ UV c)
1511{
1512 U8 tmpbuf[UTF8_MAXBYTES+1];
1513 uvchr_to_utf8(tmpbuf, c);
1514 return _is_utf8_perl_idstart(tmpbuf);
1515}
1516
1517bool
84afefe6 1518Perl_is_uni_alpha(pTHX_ UV c)
a0ed51b3 1519{
89ebb4a3 1520 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1521 uvchr_to_utf8(tmpbuf, c);
a0ed51b3
LW
1522 return is_utf8_alpha(tmpbuf);
1523}
1524
1525bool
84afefe6 1526Perl_is_uni_ascii(pTHX_ UV c)
4d61ec05 1527{
bc39fe24 1528 return isASCII(c);
4d61ec05
GS
1529}
1530
1531bool
bdd8600f
KW
1532Perl_is_uni_blank(pTHX_ UV c)
1533{
2cafb56b 1534 return isBLANK_uni(c);
bdd8600f
KW
1535}
1536
1537bool
84afefe6 1538Perl_is_uni_space(pTHX_ UV c)
a0ed51b3 1539{
add4123a 1540 return isSPACE_uni(c);
a0ed51b3
LW
1541}
1542
1543bool
84afefe6 1544Perl_is_uni_digit(pTHX_ UV c)
a0ed51b3 1545{
89ebb4a3 1546 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1547 uvchr_to_utf8(tmpbuf, c);
a0ed51b3
LW
1548 return is_utf8_digit(tmpbuf);
1549}
1550
1551bool
84afefe6 1552Perl_is_uni_upper(pTHX_ UV c)
a0ed51b3 1553{
89ebb4a3 1554 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1555 uvchr_to_utf8(tmpbuf, c);
a0ed51b3
LW
1556 return is_utf8_upper(tmpbuf);
1557}
1558
1559bool
84afefe6 1560Perl_is_uni_lower(pTHX_ UV c)
a0ed51b3 1561{
89ebb4a3 1562 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1563 uvchr_to_utf8(tmpbuf, c);
a0ed51b3
LW
1564 return is_utf8_lower(tmpbuf);
1565}
1566
1567bool
84afefe6 1568Perl_is_uni_cntrl(pTHX_ UV c)
b8c5462f 1569{
7b952154 1570 return isCNTRL_L1(c);
b8c5462f
JH
1571}
1572
1573bool
84afefe6 1574Perl_is_uni_graph(pTHX_ UV c)
b8c5462f 1575{
89ebb4a3 1576 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1577 uvchr_to_utf8(tmpbuf, c);
b8c5462f
JH
1578 return is_utf8_graph(tmpbuf);
1579}
1580
1581bool
84afefe6 1582Perl_is_uni_print(pTHX_ UV c)
a0ed51b3 1583{
89ebb4a3 1584 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1585 uvchr_to_utf8(tmpbuf, c);
a0ed51b3
LW
1586 return is_utf8_print(tmpbuf);
1587}
1588
b8c5462f 1589bool
84afefe6 1590Perl_is_uni_punct(pTHX_ UV c)
b8c5462f 1591{
89ebb4a3 1592 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1593 uvchr_to_utf8(tmpbuf, c);
b8c5462f
JH
1594 return is_utf8_punct(tmpbuf);
1595}
1596
4d61ec05 1597bool
84afefe6 1598Perl_is_uni_xdigit(pTHX_ UV c)
4d61ec05 1599{
4ac6419d 1600 return isXDIGIT_uni(c);
4d61ec05
GS
1601}
1602
3a4c58c9
KW
1603UV
1604Perl__to_upper_title_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp, const char S_or_s)
1605{
1606 /* We have the latin1-range values compiled into the core, so just use
1607 * those, converting the result to utf8. The only difference between upper
1608 * and title case in this range is that LATIN_SMALL_LETTER_SHARP_S is
1609 * either "SS" or "Ss". Which one to use is passed into the routine in
1610 * 'S_or_s' to avoid a test */
1611
1612 UV converted = toUPPER_LATIN1_MOD(c);
1613
1614 PERL_ARGS_ASSERT__TO_UPPER_TITLE_LATIN1;
1615
1616 assert(S_or_s == 'S' || S_or_s == 's');
1617
1618 if (UNI_IS_INVARIANT(converted)) { /* No difference between the two for
1619 characters in this range */
1620 *p = (U8) converted;
1621 *lenp = 1;
1622 return converted;
1623 }
1624
1625 /* toUPPER_LATIN1_MOD gives the correct results except for three outliers,
1626 * which it maps to one of them, so as to only have to have one check for
1627 * it in the main case */
1628 if (UNLIKELY(converted == LATIN_SMALL_LETTER_Y_WITH_DIAERESIS)) {
1629 switch (c) {
1630 case LATIN_SMALL_LETTER_Y_WITH_DIAERESIS:
1631 converted = LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS;
1632 break;
1633 case MICRO_SIGN:
1634 converted = GREEK_CAPITAL_LETTER_MU;
1635 break;
1636 case LATIN_SMALL_LETTER_SHARP_S:
1637 *(p)++ = 'S';
1638 *p = S_or_s;
1639 *lenp = 2;
1640 return 'S';
1641 default:
1642 Perl_croak(aTHX_ "panic: to_upper_title_latin1 did not expect '%c' to map to '%c'", c, LATIN_SMALL_LETTER_Y_WITH_DIAERESIS);
118e2215 1643 assert(0); /* NOTREACHED */
3a4c58c9
KW
1644 }
1645 }
1646
1647 *(p)++ = UTF8_TWO_BYTE_HI(converted);
1648 *p = UTF8_TWO_BYTE_LO(converted);
1649 *lenp = 2;
1650
1651 return converted;
1652}
1653
50bda2c3
KW
1654/* Call the function to convert a UTF-8 encoded character to the specified case.
1655 * Note that there may be more than one character in the result.
1656 * INP is a pointer to the first byte of the input character
1657 * OUTP will be set to the first byte of the string of changed characters. It
1658 * needs to have space for UTF8_MAXBYTES_CASE+1 bytes
1659 * LENP will be set to the length in bytes of the string of changed characters
1660 *
1661 * The functions return the ordinal of the first character in the string of OUTP */
f90a9a02
KW
1662#define CALL_UPPER_CASE(INP, OUTP, LENP) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_toupper, "ToUc", "utf8::ToSpecUc")
1663#define CALL_TITLE_CASE(INP, OUTP, LENP) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_totitle, "ToTc", "utf8::ToSpecTc")
1664#define CALL_LOWER_CASE(INP, OUTP, LENP) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_tolower, "ToLc", "utf8::ToSpecLc")
50bda2c3
KW
1665
1666/* This additionally has the input parameter SPECIALS, which if non-zero will
1667 * cause this to use the SPECIALS hash for folding (meaning get full case
1668 * folding); otherwise, when zero, this implies a simple case fold */
f90a9a02 1669#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 1670
84afefe6
JH
1671UV
1672Perl_to_uni_upper(pTHX_ UV c, U8* p, STRLEN *lenp)
a0ed51b3 1673{
3a4c58c9
KW
1674 dVAR;
1675
a1433954
KW
1676 /* Convert the Unicode character whose ordinal is <c> to its uppercase
1677 * version and store that in UTF-8 in <p> and its length in bytes in <lenp>.
1678 * Note that the <p> needs to be at least UTF8_MAXBYTES_CASE+1 bytes since
c3fd2246
KW
1679 * the changed version may be longer than the original character.
1680 *
1681 * The ordinal of the first character of the changed version is returned
1682 * (but note, as explained above, that there may be more.) */
1683
7918f24d
NC
1684 PERL_ARGS_ASSERT_TO_UNI_UPPER;
1685
3a4c58c9
KW
1686 if (c < 256) {
1687 return _to_upper_title_latin1((U8) c, p, lenp, 'S');
1688 }
1689
0ebc6274 1690 uvchr_to_utf8(p, c);
3a4c58c9 1691 return CALL_UPPER_CASE(p, p, lenp);
a0ed51b3
LW
1692}
1693
84afefe6
JH
1694UV
1695Perl_to_uni_title(pTHX_ UV c, U8* p, STRLEN *lenp)
a0ed51b3 1696{
3a4c58c9
KW
1697 dVAR;
1698
7918f24d
NC
1699 PERL_ARGS_ASSERT_TO_UNI_TITLE;
1700
3a4c58c9
KW
1701 if (c < 256) {
1702 return _to_upper_title_latin1((U8) c, p, lenp, 's');
1703 }
1704
0ebc6274 1705 uvchr_to_utf8(p, c);
3a4c58c9 1706 return CALL_TITLE_CASE(p, p, lenp);
a0ed51b3
LW
1707}
1708
afc16117
KW
1709STATIC U8
1710S_to_lower_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp)
1711{
1712 /* We have the latin1-range values compiled into the core, so just use
1713 * those, converting the result to utf8. Since the result is always just
a1433954 1714 * one character, we allow <p> to be NULL */
afc16117
KW
1715
1716 U8 converted = toLOWER_LATIN1(c);
1717
1718 if (p != NULL) {
1719 if (UNI_IS_INVARIANT(converted)) {
1720 *p = converted;
1721 *lenp = 1;
1722 }
1723 else {
1724 *p = UTF8_TWO_BYTE_HI(converted);
1725 *(p+1) = UTF8_TWO_BYTE_LO(converted);
1726 *lenp = 2;
1727 }
1728 }
1729 return converted;
1730}
1731
84afefe6
JH
1732UV
1733Perl_to_uni_lower(pTHX_ UV c, U8* p, STRLEN *lenp)
a0ed51b3 1734{
968c5e6a
KW
1735 dVAR;
1736
7918f24d
NC
1737 PERL_ARGS_ASSERT_TO_UNI_LOWER;
1738
afc16117
KW
1739 if (c < 256) {
1740 return to_lower_latin1((U8) c, p, lenp);
bca00c02
KW
1741 }
1742
afc16117 1743 uvchr_to_utf8(p, c);
968c5e6a 1744 return CALL_LOWER_CASE(p, p, lenp);
a0ed51b3
LW
1745}
1746
84afefe6 1747UV
f673fad4 1748Perl__to_fold_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp, const bool flags)
a1dde8de 1749{
a1433954 1750 /* Corresponds to to_lower_latin1(), <flags> is TRUE if to use full case
f673fad4
KW
1751 * folding */
1752
a1dde8de
KW
1753 UV converted;
1754
1755 PERL_ARGS_ASSERT__TO_FOLD_LATIN1;
1756
1757 if (c == MICRO_SIGN) {
1758 converted = GREEK_SMALL_LETTER_MU;
1759 }
1760 else if (flags && c == LATIN_SMALL_LETTER_SHARP_S) {
1761 *(p)++ = 's';
1762 *p = 's';
1763 *lenp = 2;
1764 return 's';
1765 }
1766 else { /* In this range the fold of all other characters is their lower
1767 case */
1768 converted = toLOWER_LATIN1(c);
1769 }
1770
1771 if (UNI_IS_INVARIANT(converted)) {
1772 *p = (U8) converted;
1773 *lenp = 1;
1774 }
1775 else {
1776 *(p)++ = UTF8_TWO_BYTE_HI(converted);
1777 *p = UTF8_TWO_BYTE_LO(converted);
1778 *lenp = 2;
1779 }
1780
1781 return converted;
1782}
1783
1784UV
a0270393 1785Perl__to_uni_fold_flags(pTHX_ UV c, U8* p, STRLEN *lenp, const U8 flags)
84afefe6 1786{
4b593389 1787
a0270393
KW
1788 /* Not currently externally documented, and subject to change
1789 * <flags> bits meanings:
1790 * FOLD_FLAGS_FULL iff full folding is to be used;
1791 * FOLD_FLAGS_LOCALE iff in locale
1792 * FOLD_FLAGS_NOMIX_ASCII iff non-ASCII to ASCII folds are prohibited
1793 */
4b593389 1794
36bb2ab6 1795 PERL_ARGS_ASSERT__TO_UNI_FOLD_FLAGS;
7918f24d 1796
a1dde8de 1797 if (c < 256) {
a0270393
KW
1798 UV result = _to_fold_latin1((U8) c, p, lenp,
1799 cBOOL(((flags & FOLD_FLAGS_FULL)
1800 /* If ASCII-safe, don't allow full folding,
1801 * as that could include SHARP S => ss;
1802 * otherwise there is no crossing of
1803 * ascii/non-ascii in the latin1 range */
1804 && ! (flags & FOLD_FLAGS_NOMIX_ASCII))));
1805 /* It is illegal for the fold to cross the 255/256 boundary under
1806 * locale; in this case return the original */
1807 return (result > 256 && flags & FOLD_FLAGS_LOCALE)
1808 ? c
1809 : result;
a1dde8de
KW
1810 }
1811
a0270393
KW
1812 /* If no special needs, just use the macro */
1813 if ( ! (flags & (FOLD_FLAGS_LOCALE|FOLD_FLAGS_NOMIX_ASCII))) {
1814 uvchr_to_utf8(p, c);
1815 return CALL_FOLD_CASE(p, p, lenp, flags & FOLD_FLAGS_FULL);
1816 }
1817 else { /* Otherwise, _to_utf8_fold_flags has the intelligence to deal with
1818 the special flags. */
1819 U8 utf8_c[UTF8_MAXBYTES + 1];
1820 uvchr_to_utf8(utf8_c, c);
1821 return _to_utf8_fold_flags(utf8_c, p, lenp, flags, NULL);
1822 }
84afefe6
JH
1823}
1824
a0ed51b3 1825bool
84afefe6 1826Perl_is_uni_alnum_lc(pTHX_ UV c)
a0ed51b3 1827{
edfb3318
KW
1828 if (c < 256) {
1829 return isALNUM_LC(UNI_TO_NATIVE(c));
1830 }
1831 return is_uni_alnum(c);
a0ed51b3
LW
1832}
1833
1834bool
5092f92a
KW
1835Perl_is_uni_alnumc_lc(pTHX_ UV c)
1836{
1837 if (c < 256) {
1838 return isALNUMC_LC(UNI_TO_NATIVE(c));
1839 }
1840 return is_uni_alnumc(c);
1841}
1842
1843bool
84afefe6 1844Perl_is_uni_idfirst_lc(pTHX_ UV c)
a0ed51b3 1845{
edfb3318
KW
1846 if (c < 256) {
1847 return isIDFIRST_LC(UNI_TO_NATIVE(c));
1848 }
1849 return _is_uni_perl_idstart(c);
a0ed51b3
LW
1850}
1851
1852bool
84afefe6 1853Perl_is_uni_alpha_lc(pTHX_ UV c)
a0ed51b3 1854{
edfb3318
KW
1855 if (c < 256) {
1856 return isALPHA_LC(UNI_TO_NATIVE(c));
1857 }
1858 return is_uni_alpha(c);
a0ed51b3
LW
1859}
1860
1861bool
84afefe6 1862Perl_is_uni_ascii_lc(pTHX_ UV c)
4d61ec05 1863{
edfb3318
KW
1864 if (c < 256) {
1865 return isASCII_LC(UNI_TO_NATIVE(c));
1866 }
1867 return 0;
4d61ec05
GS
1868}
1869
1870bool
bdd8600f
KW
1871Perl_is_uni_blank_lc(pTHX_ UV c)
1872{
edfb3318
KW
1873 if (c < 256) {
1874 return isBLANK_LC(UNI_TO_NATIVE(c));
1875 }
1876 return is_uni_blank(c);
bdd8600f
KW
1877}
1878
1879bool
84afefe6 1880Perl_is_uni_space_lc(pTHX_ UV c)
a0ed51b3 1881{
edfb3318
KW
1882 if (c < 256) {
1883 return isSPACE_LC(UNI_TO_NATIVE(c));
1884 }
1885 return is_uni_space(c);
a0ed51b3
LW
1886}
1887
1888bool
84afefe6 1889Perl_is_uni_digit_lc(pTHX_ UV c)
a0ed51b3 1890{
edfb3318
KW
1891 if (c < 256) {
1892 return isDIGIT_LC(UNI_TO_NATIVE(c));
1893 }
1894 return is_uni_digit(c);
a0ed51b3
LW
1895}
1896
1897bool
84afefe6 1898Perl_is_uni_upper_lc(pTHX_ UV c)
a0ed51b3 1899{
edfb3318
KW
1900 if (c < 256) {
1901 return isUPPER_LC(UNI_TO_NATIVE(c));
1902 }
1903 return is_uni_upper(c);
a0ed51b3
LW
1904}
1905
1906bool
84afefe6 1907Perl_is_uni_lower_lc(pTHX_ UV c)
a0ed51b3 1908{
edfb3318
KW
1909 if (c < 256) {
1910 return isLOWER_LC(UNI_TO_NATIVE(c));
1911 }
1912 return is_uni_lower(c);
a0ed51b3
LW
1913}
1914
1915bool
84afefe6 1916Perl_is_uni_cntrl_lc(pTHX_ UV c)
b8c5462f 1917{
edfb3318
KW
1918 if (c < 256) {
1919 return isCNTRL_LC(UNI_TO_NATIVE(c));
1920 }
1921 return is_uni_cntrl(c);
b8c5462f
JH
1922}
1923
1924bool
84afefe6 1925Perl_is_uni_graph_lc(pTHX_ UV c)
b8c5462f 1926{
edfb3318
KW
1927 if (c < 256) {
1928 return isGRAPH_LC(UNI_TO_NATIVE(c));
1929 }
1930 return is_uni_graph(c);
b8c5462f
JH
1931}
1932
1933bool
84afefe6 1934Perl_is_uni_print_lc(pTHX_ UV c)
a0ed51b3 1935{
edfb3318
KW
1936 if (c < 256) {
1937 return isPRINT_LC(UNI_TO_NATIVE(c));
1938 }
1939 return is_uni_print(c);
a0ed51b3
LW
1940}
1941
b8c5462f 1942bool
84afefe6 1943Perl_is_uni_punct_lc(pTHX_ UV c)
b8c5462f 1944{
edfb3318
KW
1945 if (c < 256) {
1946 return isPUNCT_LC(UNI_TO_NATIVE(c));
1947 }
1948 return is_uni_punct(c);
b8c5462f
JH
1949}
1950
4d61ec05 1951bool
84afefe6 1952Perl_is_uni_xdigit_lc(pTHX_ UV c)
4d61ec05 1953{
edfb3318
KW
1954 if (c < 256) {
1955 return isXDIGIT_LC(UNI_TO_NATIVE(c));
1956 }
1957 return is_uni_xdigit(c);
4d61ec05
GS
1958}
1959
b7ac61fa
JH
1960U32
1961Perl_to_uni_upper_lc(pTHX_ U32 c)
1962{
ee099d14
JH
1963 /* XXX returns only the first character -- do not use XXX */
1964 /* XXX no locale support yet */
1965 STRLEN len;
89ebb4a3 1966 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
ee099d14 1967 return (U32)to_uni_upper(c, tmpbuf, &len);
b7ac61fa
JH
1968}
1969
1970U32
1971Perl_to_uni_title_lc(pTHX_ U32 c)
1972{
ee099d14
JH
1973 /* XXX returns only the first character XXX -- do not use XXX */
1974 /* XXX no locale support yet */
1975 STRLEN len;
89ebb4a3 1976 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
ee099d14 1977 return (U32)to_uni_title(c, tmpbuf, &len);
b7ac61fa
JH
1978}
1979
1980U32
1981Perl_to_uni_lower_lc(pTHX_ U32 c)
1982{
ee099d14
JH
1983 /* XXX returns only the first character -- do not use XXX */
1984 /* XXX no locale support yet */
1985 STRLEN len;
89ebb4a3 1986 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
ee099d14 1987 return (U32)to_uni_lower(c, tmpbuf, &len);
b7ac61fa
JH
1988}
1989
26483009 1990PERL_STATIC_INLINE bool
5141f98e 1991S_is_utf8_common(pTHX_ const U8 *const p, SV **swash,
bde6a22d
NC
1992 const char *const swashname)
1993{
ea317ccb
KW
1994 /* returns a boolean giving whether or not the UTF8-encoded character that
1995 * starts at <p> is in the swash indicated by <swashname>. <swash>
1996 * contains a pointer to where the swash indicated by <swashname>
1997 * is to be stored; which this routine will do, so that future calls will
1998 * look at <*swash> and only generate a swash if it is not null
1999 *
2000 * Note that it is assumed that the buffer length of <p> is enough to
2001 * contain all the bytes that comprise the character. Thus, <*p> should
2002 * have been checked before this call for mal-formedness enough to assure
2003 * that. */
2004
97aff369 2005 dVAR;
7918f24d
NC
2006
2007 PERL_ARGS_ASSERT_IS_UTF8_COMMON;
2008
492a624f
KW
2009 /* The API should have included a length for the UTF-8 character in <p>,
2010 * but it doesn't. We therefor assume that p has been validated at least
2011 * as far as there being enough bytes available in it to accommodate the
2012 * character without reading beyond the end, and pass that number on to the
2013 * validating routine */
2014 if (!is_utf8_char_buf(p, p + UTF8SKIP(p)))
bde6a22d 2015 return FALSE;
87367d5f
KW
2016 if (!*swash) {
2017 U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST;
2018 *swash = _core_swash_init("utf8", swashname, &PL_sv_undef, 1, 0, NULL, &flags);
2019 }
bde6a22d
NC
2020 return swash_fetch(*swash, p, TRUE) != 0;
2021}
2022
2023bool
7fc63493 2024Perl_is_utf8_alnum(pTHX_ const U8 *p)
a0ed51b3 2025{
97aff369 2026 dVAR;
7918f24d
NC
2027
2028 PERL_ARGS_ASSERT_IS_UTF8_ALNUM;
2029
671c33bf
NC
2030 /* NOTE: "IsWord", not "IsAlnum", since Alnum is a true
2031 * descendant of isalnum(3), in other words, it doesn't
2032 * contain the '_'. --jhi */
d4c19fe8 2033 return is_utf8_common(p, &PL_utf8_alnum, "IsWord");
a0ed51b3
LW
2034}
2035
2036bool
5092f92a
KW
2037Perl_is_utf8_alnumc(pTHX_ const U8 *p)
2038{
2039 dVAR;
2040
2041 PERL_ARGS_ASSERT_IS_UTF8_ALNUMC;
2042
2043 return is_utf8_common(p, &PL_utf8_alnumc, "IsAlnum");
2044}
2045
2046bool
7fc63493 2047Perl_is_utf8_idfirst(pTHX_ const U8 *p) /* The naming is historical. */
a0ed51b3 2048{
97aff369 2049 dVAR;
7918f24d
NC
2050
2051 PERL_ARGS_ASSERT_IS_UTF8_IDFIRST;
2052
82686b01
JH
2053 if (*p == '_')
2054 return TRUE;
bde6a22d 2055 /* is_utf8_idstart would be more logical. */
d4c19fe8 2056 return is_utf8_common(p, &PL_utf8_idstart, "IdStart");
82686b01
JH
2057}
2058
2059bool
c11ff943
KW
2060Perl_is_utf8_xidfirst(pTHX_ const U8 *p) /* The naming is historical. */
2061{
2062 dVAR;
2063
2064 PERL_ARGS_ASSERT_IS_UTF8_XIDFIRST;
2065
2066 if (*p == '_')
2067 return TRUE;
2068 /* is_utf8_idstart would be more logical. */
2069 return is_utf8_common(p, &PL_utf8_xidstart, "XIdStart");
2070}
2071
2072bool
d65654cb 2073Perl__is_utf8_perl_idstart(pTHX_ const U8 *p)
b6912c02
KW
2074{
2075 dVAR;
2076
d65654cb 2077 PERL_ARGS_ASSERT__IS_UTF8_PERL_IDSTART;
b6912c02
KW
2078
2079 return is_utf8_common(p, &PL_utf8_perl_idstart, "_Perl_IDStart");
2080}
2081
2082bool
7fc63493 2083Perl_is_utf8_idcont(pTHX_ const U8 *p)
82686b01 2084{
97aff369 2085 dVAR;
7918f24d
NC
2086
2087 PERL_ARGS_ASSERT_IS_UTF8_IDCONT;
2088
d4c19fe8 2089 return is_utf8_common(p, &PL_utf8_idcont, "IdContinue");
a0ed51b3
LW
2090}
2091
2092bool
c11ff943
KW
2093Perl_is_utf8_xidcont(pTHX_ const U8 *p)
2094{
2095 dVAR;
2096
2097 PERL_ARGS_ASSERT_IS_UTF8_XIDCONT;
2098
c11ff943
KW
2099 return is_utf8_common(p, &PL_utf8_idcont, "XIdContinue");
2100}
2101
2102bool
7fc63493 2103Perl_is_utf8_alpha(pTHX_ const U8 *p)
a0ed51b3 2104{
97aff369 2105 dVAR;
7918f24d
NC
2106
2107 PERL_ARGS_ASSERT_IS_UTF8_ALPHA;
2108
d4c19fe8 2109 return is_utf8_common(p, &PL_utf8_alpha, "IsAlpha");
a0ed51b3
LW
2110}
2111
2112bool
7fc63493 2113Perl_is_utf8_ascii(pTHX_ const U8 *p)
b8c5462f 2114{
97aff369 2115 dVAR;
7918f24d
NC
2116
2117 PERL_ARGS_ASSERT_IS_UTF8_ASCII;
2118
bc39fe24
KW
2119 /* ASCII characters are the same whether in utf8 or not. So the macro
2120 * works on both utf8 and non-utf8 representations. */
2121 return isASCII(*p);
b8c5462f
JH
2122}
2123
2124bool
bdd8600f
KW
2125Perl_is_utf8_blank(pTHX_ const U8 *p)
2126{
2127 dVAR;
2128
2129 PERL_ARGS_ASSERT_IS_UTF8_BLANK;
2130
2cafb56b 2131 return isBLANK_utf8(p);
bdd8600f
KW
2132}
2133
2134bool
7fc63493 2135Perl_is_utf8_space(pTHX_ const U8 *p)
a0ed51b3 2136{
97aff369 2137 dVAR;
7918f24d
NC
2138
2139 PERL_ARGS_ASSERT_IS_UTF8_SPACE;
2140
add4123a 2141 return isSPACE_utf8(p);
a0ed51b3
LW
2142}
2143
2144bool
d1eb3177
YO
2145Perl_is_utf8_perl_space(pTHX_ const U8 *p)
2146{
2147 dVAR;
2148
2149 PERL_ARGS_ASSERT_IS_UTF8_PERL_SPACE;
2150
c4428693
KW
2151 /* Only true if is an ASCII space-like character, and ASCII is invariant
2152 * under utf8, so can just use the macro */
2153 return isSPACE_A(*p);
d1eb3177
YO
2154}
2155
2156bool
2157Perl_is_utf8_perl_word(pTHX_ const U8 *p)
2158{
2159 dVAR;
2160
2161 PERL_ARGS_ASSERT_IS_UTF8_PERL_WORD;
2162
c4428693
KW
2163 /* Only true if is an ASCII word character, and ASCII is invariant
2164 * under utf8, so can just use the macro */
2165 return isWORDCHAR_A(*p);
d1eb3177
YO
2166}
2167
2168bool
7fc63493 2169Perl_is_utf8_digit(pTHX_ const U8 *p)
a0ed51b3 2170{
97aff369 2171 dVAR;
7918f24d
NC
2172
2173 PERL_ARGS_ASSERT_IS_UTF8_DIGIT;
2174
d4c19fe8 2175 return is_utf8_common(p, &PL_utf8_digit, "IsDigit");
a0ed51b3
LW
2176}
2177
2178bool
d1eb3177
YO
2179Perl_is_utf8_posix_digit(pTHX_ const U8 *p)
2180{
2181 dVAR;
2182
2183 PERL_ARGS_ASSERT_IS_UTF8_POSIX_DIGIT;
2184
c4428693
KW
2185 /* Only true if is an ASCII digit character, and ASCII is invariant
2186 * under utf8, so can just use the macro */
2187 return isDIGIT_A(*p);
d1eb3177
YO
2188}
2189
2190bool
7fc63493 2191Perl_is_utf8_upper(pTHX_ const U8 *p)
a0ed51b3 2192{
97aff369 2193 dVAR;
7918f24d
NC
2194
2195 PERL_ARGS_ASSERT_IS_UTF8_UPPER;
2196
d4c19fe8 2197 return is_utf8_common(p, &PL_utf8_upper, "IsUppercase");
a0ed51b3
LW
2198}
2199
2200bool
7fc63493 2201Perl_is_utf8_lower(pTHX_ const U8 *p)
a0ed51b3 2202{
97aff369 2203 dVAR;
7918f24d
NC
2204
2205 PERL_ARGS_ASSERT_IS_UTF8_LOWER;
2206
d4c19fe8 2207 return is_utf8_common(p, &PL_utf8_lower, "IsLowercase");
a0ed51b3
LW
2208}
2209
2210bool
7fc63493 2211Perl_is_utf8_cntrl(pTHX_ const U8 *p)
b8c5462f 2212{
97aff369 2213 dVAR;
7918f24d
NC
2214
2215 PERL_ARGS_ASSERT_IS_UTF8_CNTRL;
2216
a35d759a 2217 return isCNTRL_utf8(p);
b8c5462f
JH
2218}
2219
2220bool
7fc63493 2221Perl_is_utf8_graph(pTHX_ const U8 *p)
b8c5462f 2222{
97aff369 2223 dVAR;
7918f24d
NC
2224
2225 PERL_ARGS_ASSERT_IS_UTF8_GRAPH;
2226
d4c19fe8 2227 return is_utf8_common(p, &PL_utf8_graph, "IsGraph");
b8c5462f
JH
2228}
2229
2230bool
7fc63493 2231Perl_is_utf8_print(pTHX_ const U8 *p)
a0ed51b3 2232{
97aff369 2233 dVAR;
7918f24d
NC
2234
2235 PERL_ARGS_ASSERT_IS_UTF8_PRINT;
2236
d4c19fe8 2237 return is_utf8_common(p, &PL_utf8_print, "IsPrint");
a0ed51b3
LW
2238}
2239
2240bool
7fc63493 2241Perl_is_utf8_punct(pTHX_ const U8 *p)
b8c5462f 2242{
97aff369 2243 dVAR;
7918f24d
NC
2244
2245 PERL_ARGS_ASSERT_IS_UTF8_PUNCT;
2246
d4c19fe8 2247 return is_utf8_common(p, &PL_utf8_punct, "IsPunct");
b8c5462f
JH
2248}
2249
2250bool
7fc63493 2251Perl_is_utf8_xdigit(pTHX_ const U8 *p)
b8c5462f 2252{
97aff369 2253 dVAR;
7918f24d
NC
2254
2255 PERL_ARGS_ASSERT_IS_UTF8_XDIGIT;
2256
4ac6419d 2257 return is_XDIGIT_utf8(p);
b8c5462f
JH
2258}
2259
2260bool
7fc63493 2261Perl_is_utf8_mark(pTHX_ const U8 *p)
a0ed51b3 2262{
97aff369 2263 dVAR;
7918f24d
NC
2264
2265 PERL_ARGS_ASSERT_IS_UTF8_MARK;
2266
d4c19fe8 2267 return is_utf8_common(p, &PL_utf8_mark, "IsM");
a0ed51b3
LW
2268}
2269
37e2e78e 2270bool
27d4fc33 2271Perl_is_utf8_X_regular_begin(pTHX_ const U8 *p)
37e2e78e
KW
2272{
2273 dVAR;
2274
27d4fc33 2275 PERL_ARGS_ASSERT_IS_UTF8_X_REGULAR_BEGIN;
37e2e78e 2276
27d4fc33 2277 return is_utf8_common(p, &PL_utf8_X_regular_begin, "_X_Regular_Begin");
37e2e78e
KW
2278}
2279
2280bool
2281Perl_is_utf8_X_extend(pTHX_ const U8 *p)
2282{
2283 dVAR;
2284
2285 PERL_ARGS_ASSERT_IS_UTF8_X_EXTEND;
2286
2287 return is_utf8_common(p, &PL_utf8_X_extend, "_X_Extend");
2288}
2289
6b5c0936 2290/*
87cea99e 2291=for apidoc to_utf8_case
6b5c0936 2292
a1433954
KW
2293The C<p> contains the pointer to the UTF-8 string encoding
2294the character that is being converted. This routine assumes that the character
2295at C<p> is well-formed.
6b5c0936 2296
a1433954
KW
2297The C<ustrp> is a pointer to the character buffer to put the
2298conversion result to. The C<lenp> is a pointer to the length
6b5c0936
JH
2299of the result.
2300
a1433954 2301The C<swashp> is a pointer to the swash to use.
6b5c0936 2302
a1433954
KW
2303Both the special and normal mappings are stored in F<lib/unicore/To/Foo.pl>,
2304and loaded by SWASHNEW, using F<lib/utf8_heavy.pl>. The C<special> (usually,
0134edef 2305but not always, a multicharacter mapping), is tried first.
6b5c0936 2306
a1433954 2307The C<special> is a string like "utf8::ToSpecLower", which means the
0134edef
JH
2308hash %utf8::ToSpecLower. The access to the hash is through
2309Perl_to_utf8_case().
6b5c0936 2310
a1433954 2311The C<normal> is a string like "ToLower" which means the swash
0134edef
JH
2312%utf8::ToLower.
2313
2314=cut */
6b5c0936 2315
2104c8d9 2316UV
9a957fbc
AL
2317Perl_to_utf8_case(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp,
2318 SV **swashp, const char *normal, const char *special)
a0ed51b3 2319{
97aff369 2320 dVAR;
89ebb4a3 2321 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
0134edef 2322 STRLEN len = 0;
4b88fb76 2323 const UV uv0 = valid_utf8_to_uvchr(p, NULL);
1feea2c7
JH
2324 /* The NATIVE_TO_UNI() and UNI_TO_NATIVE() mappings
2325 * are necessary in EBCDIC, they are redundant no-ops
2326 * in ASCII-ish platforms, and hopefully optimized away. */
f54cb97a 2327 const UV uv1 = NATIVE_TO_UNI(uv0);
7918f24d
NC
2328
2329 PERL_ARGS_ASSERT_TO_UTF8_CASE;
2330
9ae3ac1a
KW
2331 /* Note that swash_fetch() doesn't output warnings for these because it
2332 * assumes we will */
8457b38f 2333 if (uv1 >= UNICODE_SURROGATE_FIRST) {
9ae3ac1a 2334 if (uv1 <= UNICODE_SURROGATE_LAST) {
8457b38f
KW
2335 if (ckWARN_d(WARN_SURROGATE)) {
2336 const char* desc = (PL_op) ? OP_DESC(PL_op) : normal;
2337 Perl_warner(aTHX_ packWARN(WARN_SURROGATE),
2338 "Operation \"%s\" returns its argument for UTF-16 surrogate U+%04"UVXf"", desc, uv1);
2339 }
9ae3ac1a
KW
2340 }
2341 else if (UNICODE_IS_SUPER(uv1)) {
8457b38f
KW
2342 if (ckWARN_d(WARN_NON_UNICODE)) {
2343 const char* desc = (PL_op) ? OP_DESC(PL_op) : normal;
2344 Perl_warner(aTHX_ packWARN(WARN_NON_UNICODE),
2345 "Operation \"%s\" returns its argument for non-Unicode code point 0x%04"UVXf"", desc, uv1);
2346 }
9ae3ac1a
KW
2347 }
2348
2349 /* Note that non-characters are perfectly legal, so no warning should
2350 * be given */
2351 }
2352
1feea2c7 2353 uvuni_to_utf8(tmpbuf, uv1);
0134edef
JH
2354
2355 if (!*swashp) /* load on-demand */
5ab9d2ef 2356 *swashp = _core_swash_init("utf8", normal, &PL_sv_undef, 4, 0, NULL, NULL);
0134edef 2357
a6f87d8c 2358 if (special) {
0134edef 2359 /* It might be "special" (sometimes, but not always,
2a37f04d 2360 * a multicharacter mapping) */
6673a63c 2361 HV * const hv = get_hv(special, 0);
b08cf34e
JH
2362 SV **svp;
2363
35da51f7 2364 if (hv &&
b08cf34e
JH
2365 (svp = hv_fetch(hv, (const char*)tmpbuf, UNISKIP(uv1), FALSE)) &&
2366 (*svp)) {
cfd0369c 2367 const char *s;
47654450 2368
cfd0369c 2369 s = SvPV_const(*svp, len);
47654450
JH
2370 if (len == 1)
2371 len = uvuni_to_utf8(ustrp, NATIVE_TO_UNI(*(U8*)s)) - ustrp;
2a37f04d 2372 else {
2f9475ad
JH
2373#ifdef EBCDIC
2374 /* If we have EBCDIC we need to remap the characters
2375 * since any characters in the low 256 are Unicode
2376 * code points, not EBCDIC. */
7cda7a3d 2377 U8 *t = (U8*)s, *tend = t + len, *d;
2f9475ad
JH
2378
2379 d = tmpbuf;
b08cf34e 2380 if (SvUTF8(*svp)) {
2f9475ad
JH
2381 STRLEN tlen = 0;
2382
2383 while (t < tend) {
6bd1c396 2384 const UV c = utf8_to_uvchr_buf(t, tend, &tlen);
2f9475ad
JH
2385 if (tlen > 0) {
2386 d = uvchr_to_utf8(d, UNI_TO_NATIVE(c));
2387 t += tlen;
2388 }
2389 else
2390 break;
2391 }
2392 }
2393 else {
36fec512
JH
2394 while (t < tend) {
2395 d = uvchr_to_utf8(d, UNI_TO_NATIVE(*t));
2396 t++;
2397 }
2f9475ad
JH
2398 }
2399 len = d - tmpbuf;
2400 Copy(tmpbuf, ustrp, len, U8);
2401#else
d2dcd0fb 2402 Copy(s, ustrp, len, U8);
2f9475ad 2403#endif
29e98929 2404 }
983ffd37 2405 }
0134edef
JH
2406 }
2407
2408 if (!len && *swashp) {
2114036c 2409 const UV uv2 = swash_fetch(*swashp, tmpbuf, TRUE /* => is utf8 */);
d4c19fe8 2410
0134edef
JH
2411 if (uv2) {
2412 /* It was "normal" (a single character mapping). */
d4c19fe8 2413 const UV uv3 = UNI_TO_NATIVE(uv2);
e9101d72 2414 len = uvchr_to_utf8(ustrp, uv3) - ustrp;
2a37f04d
JH
2415 }
2416 }
1feea2c7 2417
cbe07460
KW
2418 if (len) {
2419 if (lenp) {
2420 *lenp = len;
2421 }
2422 return valid_utf8_to_uvchr(ustrp, 0);
2423 }
2424
2425 /* Here, there was no mapping defined, which means that the code point maps
2426 * to itself. Return the inputs */
bfdf22ec 2427 len = UTF8SKIP(p);
ca9fab46
KW
2428 if (p != ustrp) { /* Don't copy onto itself */
2429 Copy(p, ustrp, len, U8);
2430 }
0134edef 2431
2a37f04d
JH
2432 if (lenp)
2433 *lenp = len;
2434
cbe07460
KW
2435 return uv0;
2436
a0ed51b3
LW
2437}
2438
051a06d4
KW
2439STATIC UV
2440S_check_locale_boundary_crossing(pTHX_ const U8* const p, const UV result, U8* const ustrp, STRLEN *lenp)
2441{
2442 /* This is called when changing the case of a utf8-encoded character above
2443 * the Latin1 range, and the operation is in locale. If the result
2444 * contains a character that crosses the 255/256 boundary, disallow the
2445 * change, and return the original code point. See L<perlfunc/lc> for why;
2446 *
a1433954
KW
2447 * p points to the original string whose case was changed; assumed
2448 * by this routine to be well-formed
051a06d4
KW
2449 * result the code point of the first character in the changed-case string
2450 * ustrp points to the changed-case string (<result> represents its first char)
2451 * lenp points to the length of <ustrp> */
2452
2453 UV original; /* To store the first code point of <p> */
2454
2455 PERL_ARGS_ASSERT_CHECK_LOCALE_BOUNDARY_CROSSING;
2456
2457 assert(! UTF8_IS_INVARIANT(*p) && ! UTF8_IS_DOWNGRADEABLE_START(*p));
2458
2459 /* We know immediately if the first character in the string crosses the
2460 * boundary, so can skip */
2461 if (result > 255) {
2462
2463 /* Look at every character in the result; if any cross the
2464 * boundary, the whole thing is disallowed */
2465 U8* s = ustrp + UTF8SKIP(ustrp);
2466 U8* e = ustrp + *lenp;
2467 while (s < e) {
2468 if (UTF8_IS_INVARIANT(*s) || UTF8_IS_DOWNGRADEABLE_START(*s))
2469 {
2470 goto bad_crossing;
2471 }
2472 s += UTF8SKIP(s);
2473 }
2474
2475 /* Here, no characters crossed, result is ok as-is */
2476 return result;
2477 }
2478
2479bad_crossing:
2480
2481 /* Failed, have to return the original */
4b88fb76 2482 original = valid_utf8_to_uvchr(p, lenp);
051a06d4
KW
2483 Copy(p, ustrp, *lenp, char);
2484 return original;
2485}
2486
d3e79532 2487/*
87cea99e 2488=for apidoc to_utf8_upper
d3e79532 2489
a1433954
KW
2490Convert the UTF-8 encoded character at C<p> to its uppercase version and
2491store that in UTF-8 in C<ustrp> and its length in bytes in C<lenp>. Note
89ebb4a3
JH
2492that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since
2493the uppercase version may be longer than the original character.
d3e79532
JH
2494
2495The first character of the uppercased version is returned
2496(but note, as explained above, that there may be more.)
2497
a1433954
KW
2498The character at C<p> is assumed by this routine to be well-formed.
2499
d3e79532
JH
2500=cut */
2501
051a06d4
KW
2502/* Not currently externally documented, and subject to change:
2503 * <flags> is set iff locale semantics are to be used for code points < 256
2504 * <tainted_ptr> if non-null, *tainted_ptr will be set TRUE iff locale rules
2505 * were used in the calculation; otherwise unchanged. */
2506
2104c8d9 2507UV
051a06d4 2508Perl__to_utf8_upper_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, const bool flags, bool* tainted_ptr)
a0ed51b3 2509{
97aff369 2510 dVAR;
7918f24d 2511
051a06d4
KW
2512 UV result;
2513
2514 PERL_ARGS_ASSERT__TO_UTF8_UPPER_FLAGS;
7918f24d 2515
3a4c58c9 2516 if (UTF8_IS_INVARIANT(*p)) {
051a06d4
KW
2517 if (flags) {
2518 result = toUPPER_LC(*p);
2519 }
2520 else {
81c6c7ce 2521 return _to_upper_title_latin1(*p, ustrp, lenp, 'S');
051a06d4 2522 }
3a4c58c9
KW
2523 }
2524 else if UTF8_IS_DOWNGRADEABLE_START(*p) {
051a06d4
KW
2525 if (flags) {
2526 result = toUPPER_LC(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)));
2527 }
2528 else {
81c6c7ce
KW
2529 return _to_upper_title_latin1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)),
2530 ustrp, lenp, 'S');
051a06d4
KW
2531 }
2532 }
2533 else { /* utf8, ord above 255 */
2534 result = CALL_UPPER_CASE(p, ustrp, lenp);
2535
2536 if (flags) {
2537 result = check_locale_boundary_crossing(p, result, ustrp, lenp);
2538 }
2539 return result;
2540 }
2541
2542 /* Here, used locale rules. Convert back to utf8 */
2543 if (UTF8_IS_INVARIANT(result)) {
2544 *ustrp = (U8) result;
2545 *lenp = 1;
2546 }
2547 else {
2548 *ustrp = UTF8_EIGHT_BIT_HI(result);
2549 *(ustrp + 1) = UTF8_EIGHT_BIT_LO(result);
2550 *lenp = 2;
3a4c58c9
KW
2551 }
2552
051a06d4
KW
2553 if (tainted_ptr) {
2554 *tainted_ptr = TRUE;
2555 }
2556 return result;
983ffd37 2557}
a0ed51b3 2558
d3e79532 2559/*
87cea99e 2560=for apidoc to_utf8_title
d3e79532 2561
a1433954
KW
2562Convert the UTF-8 encoded character at C<p> to its titlecase version and
2563store that in UTF-8 in C<ustrp> and its length in bytes in C<lenp>. Note
2564that the C<ustrp> needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
89ebb4a3 2565titlecase version may be longer than the original character.
d3e79532
JH
2566
2567The first character of the titlecased version is returned
2568(but note, as explained above, that there may be more.)
2569
a1433954
KW
2570The character at C<p> is assumed by this routine to be well-formed.
2571
d3e79532
JH
2572=cut */
2573
051a06d4
KW
2574/* Not currently externally documented, and subject to change:
2575 * <flags> is set iff locale semantics are to be used for code points < 256
2576 * Since titlecase is not defined in POSIX, uppercase is used instead
2577 * for these/
2578 * <tainted_ptr> if non-null, *tainted_ptr will be set TRUE iff locale rules
2579 * were used in the calculation; otherwise unchanged. */
2580
983ffd37 2581UV
051a06d4 2582Perl__to_utf8_title_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, const bool flags, bool* tainted_ptr)
983ffd37 2583{
97aff369 2584 dVAR;
7918f24d 2585
051a06d4
KW
2586 UV result;
2587
2588 PERL_ARGS_ASSERT__TO_UTF8_TITLE_FLAGS;
7918f24d 2589
3a4c58c9 2590 if (UTF8_IS_INVARIANT(*p)) {
051a06d4
KW
2591 if (flags) {
2592 result = toUPPER_LC(*p);
2593 }
2594 else {
81c6c7ce 2595 return _to_upper_title_latin1(*p, ustrp, lenp, 's');
051a06d4 2596 }
3a4c58c9
KW
2597 }
2598 else if UTF8_IS_DOWNGRADEABLE_START(*p) {
051a06d4
KW
2599 if (flags) {
2600 result = toUPPER_LC(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)));
2601 }
2602 else {
81c6c7ce
KW
2603 return _to_upper_title_latin1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)),
2604 ustrp, lenp, 's');
051a06d4
KW
2605 }
2606 }
2607 else { /* utf8, ord above 255 */
2608 result = CALL_TITLE_CASE(p, ustrp, lenp);
2609
2610 if (flags) {
2611 result = check_locale_boundary_crossing(p, result, ustrp, lenp);
2612 }
2613 return result;
2614 }
2615
2616 /* Here, used locale rules. Convert back to utf8 */
2617 if (UTF8_IS_INVARIANT(result)) {
2618 *ustrp = (U8) result;
2619 *lenp = 1;
2620 }
2621 else {
2622 *ustrp = UTF8_EIGHT_BIT_HI(result);
2623 *(ustrp + 1) = UTF8_EIGHT_BIT_LO(result);
2624 *lenp = 2;
3a4c58c9
KW
2625 }
2626
051a06d4
KW
2627 if (tainted_ptr) {
2628 *tainted_ptr = TRUE;
2629 }
2630 return result;
a0ed51b3
LW
2631}
2632
d3e79532 2633/*
87cea99e 2634=for apidoc to_utf8_lower
d3e79532 2635
a1433954
KW
2636Convert the UTF-8 encoded character at C<p> to its lowercase version and
2637store that in UTF-8 in ustrp and its length in bytes in C<lenp>. Note
2638that the C<ustrp> needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
89ebb4a3 2639lowercase version may be longer than the original character.
d3e79532
JH
2640
2641The first character of the lowercased version is returned
2642(but note, as explained above, that there may be more.)
2643
a1433954
KW
2644The character at C<p> is assumed by this routine to be well-formed.
2645
d3e79532
JH
2646=cut */
2647
051a06d4
KW
2648/* Not currently externally documented, and subject to change:
2649 * <flags> is set iff locale semantics are to be used for code points < 256
2650 * <tainted_ptr> if non-null, *tainted_ptr will be set TRUE iff locale rules
2651 * were used in the calculation; otherwise unchanged. */
2652
2104c8d9 2653UV
051a06d4 2654Perl__to_utf8_lower_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, const bool flags, bool* tainted_ptr)
a0ed51b3 2655{
051a06d4
KW
2656 UV result;
2657
97aff369 2658 dVAR;
7918f24d 2659
051a06d4 2660 PERL_ARGS_ASSERT__TO_UTF8_LOWER_FLAGS;
7918f24d 2661
968c5e6a 2662 if (UTF8_IS_INVARIANT(*p)) {
051a06d4
KW
2663 if (flags) {
2664 result = toLOWER_LC(*p);
2665 }
2666 else {
81c6c7ce 2667 return to_lower_latin1(*p, ustrp, lenp);
051a06d4 2668 }
968c5e6a
KW
2669 }
2670 else if UTF8_IS_DOWNGRADEABLE_START(*p) {
051a06d4
KW
2671 if (flags) {
2672 result = toLOWER_LC(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)));
2673 }
2674 else {
81c6c7ce
KW
2675 return to_lower_latin1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)),
2676 ustrp, lenp);
051a06d4 2677 }
968c5e6a 2678 }
051a06d4
KW
2679 else { /* utf8, ord above 255 */
2680 result = CALL_LOWER_CASE(p, ustrp, lenp);
2681
2682 if (flags) {
2683 result = check_locale_boundary_crossing(p, result, ustrp, lenp);
2684 }
968c5e6a 2685
051a06d4
KW
2686 return result;
2687 }
2688
2689 /* Here, used locale rules. Convert back to utf8 */
2690 if (UTF8_IS_INVARIANT(result)) {
2691 *ustrp = (U8) result;
2692 *lenp = 1;
2693 }
2694 else {
2695 *ustrp = UTF8_EIGHT_BIT_HI(result);
2696 *(ustrp + 1) = UTF8_EIGHT_BIT_LO(result);
2697 *lenp = 2;
2698 }
2699
2700 if (tainted_ptr) {
2701 *tainted_ptr = TRUE;
2702 }
2703 return result;
b4e400f9
JH
2704}
2705
d3e79532 2706/*
87cea99e 2707=for apidoc to_utf8_fold
d3e79532 2708
a1433954
KW
2709Convert the UTF-8 encoded character at C<p> to its foldcase version and
2710store that in UTF-8 in C<ustrp> and its length in bytes in C<lenp>. Note
2711that the C<ustrp> needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
d3e79532
JH
2712foldcase version may be longer than the original character (up to
2713three characters).
2714
2715The first character of the foldcased version is returned
2716(but note, as explained above, that there may be more.)
2717
a1433954
KW
2718The character at C<p> is assumed by this routine to be well-formed.
2719
d3e79532
JH
2720=cut */
2721
051a06d4
KW
2722/* Not currently externally documented, and subject to change,
2723 * in <flags>
2724 * bit FOLD_FLAGS_LOCALE is set iff locale semantics are to be used for code
2725 * points < 256. Since foldcase is not defined in
2726 * POSIX, lowercase is used instead
2727 * bit FOLD_FLAGS_FULL is set iff full case folds are to be used;
2728 * otherwise simple folds
a0270393
KW
2729 * bit FOLD_FLAGS_NOMIX_ASCII is set iff folds of non-ASCII to ASCII are
2730 * prohibited
051a06d4
KW
2731 * <tainted_ptr> if non-null, *tainted_ptr will be set TRUE iff locale rules
2732 * were used in the calculation; otherwise unchanged. */
36bb2ab6 2733
b4e400f9 2734UV
051a06d4 2735Perl__to_utf8_fold_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, U8 flags, bool* tainted_ptr)
b4e400f9 2736{
97aff369 2737 dVAR;
7918f24d 2738
051a06d4
KW
2739 UV result;
2740
36bb2ab6 2741 PERL_ARGS_ASSERT__TO_UTF8_FOLD_FLAGS;
7918f24d 2742
a0270393
KW
2743 /* These are mutually exclusive */
2744 assert (! ((flags & FOLD_FLAGS_LOCALE) && (flags & FOLD_FLAGS_NOMIX_ASCII)));
2745
50ba90ff
KW
2746 assert(p != ustrp); /* Otherwise overwrites */
2747
a1dde8de 2748 if (UTF8_IS_INVARIANT(*p)) {
051a06d4
KW
2749 if (flags & FOLD_FLAGS_LOCALE) {
2750 result = toLOWER_LC(*p);
2751 }
2752 else {
81c6c7ce
KW
2753 return _to_fold_latin1(*p, ustrp, lenp,
2754 cBOOL(flags & FOLD_FLAGS_FULL));
051a06d4 2755 }
a1dde8de
KW
2756 }
2757 else if UTF8_IS_DOWNGRADEABLE_START(*p) {
051a06d4
KW
2758 if (flags & FOLD_FLAGS_LOCALE) {
2759 result = toLOWER_LC(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)));
2760 }
2761 else {
81c6c7ce 2762 return _to_fold_latin1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)),
a0270393
KW
2763 ustrp, lenp,
2764 cBOOL((flags & FOLD_FLAGS_FULL
2765 /* If ASCII safe, don't allow full
2766 * folding, as that could include SHARP
2767 * S => ss; otherwise there is no
2768 * crossing of ascii/non-ascii in the
2769 * latin1 range */
2770 && ! (flags & FOLD_FLAGS_NOMIX_ASCII))));
051a06d4 2771 }
a1dde8de 2772 }
051a06d4 2773 else { /* utf8, ord above 255 */
a0270393 2774 result = CALL_FOLD_CASE(p, ustrp, lenp, flags & FOLD_FLAGS_FULL);
a1dde8de 2775
051a06d4 2776 if ((flags & FOLD_FLAGS_LOCALE)) {
a0270393 2777 return check_locale_boundary_crossing(p, result, ustrp, lenp);
051a06d4 2778 }
a0270393
KW
2779 else if (! (flags & FOLD_FLAGS_NOMIX_ASCII)) {
2780 return result;
2781 }
2782 else {
2783 /* This is called when changing the case of a utf8-encoded
2784 * character above the Latin1 range, and the result should not
2785 * contain an ASCII character. */
2786
2787 UV original; /* To store the first code point of <p> */
2788
2789 /* Look at every character in the result; if any cross the
2790 * boundary, the whole thing is disallowed */
2791 U8* s = ustrp;
2792 U8* e = ustrp + *lenp;
2793 while (s < e) {
2794 if (isASCII(*s)) {
2795 /* Crossed, have to return the original */
2796 original = valid_utf8_to_uvchr(p, lenp);
2797 Copy(p, ustrp, *lenp, char);
2798 return original;
2799 }
2800 s += UTF8SKIP(s);
2801 }
051a06d4 2802
a0270393
KW
2803 /* Here, no characters crossed, result is ok as-is */
2804 return result;
2805 }
051a06d4
KW
2806 }
2807
2808 /* Here, used locale rules. Convert back to utf8 */
2809 if (UTF8_IS_INVARIANT(result)) {
2810 *ustrp = (U8) result;
2811 *lenp = 1;
2812 }
2813 else {
2814 *ustrp = UTF8_EIGHT_BIT_HI(result);
2815 *(ustrp + 1) = UTF8_EIGHT_BIT_LO(result);
2816 *lenp = 2;
2817 }
2818
2819 if (tainted_ptr) {
2820 *tainted_ptr = TRUE;
2821 }
2822 return result;
a0ed51b3
LW
2823}
2824
711a919c 2825/* Note:
f90a9a02 2826 * Returns a "swash" which is a hash described in utf8.c:Perl_swash_fetch().
711a919c
ST
2827 * C<pkg> is a pointer to a package name for SWASHNEW, should be "utf8".
2828 * For other parameters, see utf8::SWASHNEW in lib/utf8_heavy.pl.
2829 */
c4a5db0c 2830
a0ed51b3 2831SV*
7fc63493 2832Perl_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv, I32 minbits, I32 none)
a0ed51b3 2833{
c4a5db0c
KW
2834 PERL_ARGS_ASSERT_SWASH_INIT;
2835
2836 /* Returns a copy of a swash initiated by the called function. This is the
2837 * public interface, and returning a copy prevents others from doing
2838 * mischief on the original */
2839
5d3d13d1 2840 return newSVsv(_core_swash_init(pkg, name, listsv, minbits, none, NULL, NULL));
c4a5db0c
KW
2841}
2842
2843SV*
5d3d13d1 2844Perl__core_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv, I32 minbits, I32 none, SV* invlist, U8* const flags_p)
c4a5db0c
KW
2845{
2846 /* Initialize and return a swash, creating it if necessary. It does this
87367d5f
KW
2847 * by calling utf8_heavy.pl in the general case. The returned value may be
2848 * the swash's inversion list instead if the input parameters allow it.
2849 * Which is returned should be immaterial to callers, as the only
923b6d4e
KW
2850 * operations permitted on a swash, swash_fetch(), _get_swash_invlist(),
2851 * and swash_to_invlist() handle both these transparently.
c4a5db0c
KW
2852 *
2853 * This interface should only be used by functions that won't destroy or
2854 * adversely change the swash, as doing so affects all other uses of the
2855 * swash in the program; the general public should use 'Perl_swash_init'
2856 * instead.
2857 *
2858 * pkg is the name of the package that <name> should be in.
2859 * name is the name of the swash to find. Typically it is a Unicode
2860 * property name, including user-defined ones
2861 * listsv is a string to initialize the swash with. It must be of the form
2862 * documented as the subroutine return value in
2863 * L<perlunicode/User-Defined Character Properties>
2864 * minbits is the number of bits required to represent each data element.
2865 * It is '1' for binary properties.
2866 * none I (khw) do not understand this one, but it is used only in tr///.
9a53f6cf 2867 * invlist is an inversion list to initialize the swash with (or NULL)
83199d38
KW
2868 * flags_p if non-NULL is the address of various input and output flag bits
2869 * to the routine, as follows: ('I' means is input to the routine;
2870 * 'O' means output from the routine. Only flags marked O are
2871 * meaningful on return.)
2872 * _CORE_SWASH_INIT_USER_DEFINED_PROPERTY indicates if the swash
2873 * came from a user-defined property. (I O)
5d3d13d1
KW
2874 * _CORE_SWASH_INIT_RETURN_IF_UNDEF indicates that instead of croaking
2875 * when the swash cannot be located, to simply return NULL. (I)
87367d5f
KW
2876 * _CORE_SWASH_INIT_ACCEPT_INVLIST indicates that the caller will accept a
2877 * return of an inversion list instead of a swash hash if this routine
2878 * thinks that would result in faster execution of swash_fetch() later
2879 * on. (I)
9a53f6cf
KW
2880 *
2881 * Thus there are three possible inputs to find the swash: <name>,
2882 * <listsv>, and <invlist>. At least one must be specified. The result
2883 * will be the union of the specified ones, although <listsv>'s various
2884 * actions can intersect, etc. what <name> gives.
2885 *
2886 * <invlist> is only valid for binary properties */
c4a5db0c 2887
27da23d5 2888 dVAR;
c4a5db0c 2889 SV* retval = &PL_sv_undef;
83199d38 2890 HV* swash_hv = NULL;
87367d5f
KW
2891 const int invlist_swash_boundary =
2892 (flags_p && *flags_p & _CORE_SWASH_INIT_ACCEPT_INVLIST)
2893 ? 512 /* Based on some benchmarking, but not extensive, see commit
2894 message */
2895 : -1; /* Never return just an inversion list */
9a53f6cf
KW
2896
2897 assert(listsv != &PL_sv_undef || strNE(name, "") || invlist);
2898 assert(! invlist || minbits == 1);
2899
2900 /* If data was passed in to go out to utf8_heavy to find the swash of, do
2901 * so */
2902 if (listsv != &PL_sv_undef || strNE(name, "")) {
69794297
KW
2903 dSP;
2904 const size_t pkg_len = strlen(pkg);
2905 const size_t name_len = strlen(name);
2906 HV * const stash = gv_stashpvn(pkg, pkg_len, 0);
2907 SV* errsv_save;
2908 GV *method;
2909
2910 PERL_ARGS_ASSERT__CORE_SWASH_INIT;
2911
2912 PUSHSTACKi(PERLSI_MAGIC);
ce3b816e 2913 ENTER;
69794297
KW
2914 SAVEHINTS();
2915 save_re_context();
650f067c
JL
2916 /* We might get here via a subroutine signature which uses a utf8
2917 * parameter name, at which point PL_subname will have been set
2918 * but not yet used. */
2919 save_item(PL_subname);
69794297
KW
2920 if (PL_parser && PL_parser->error_count)
2921 SAVEI8(PL_parser->error_count), PL_parser->error_count = 0;
2922 method = gv_fetchmeth(stash, "SWASHNEW", 8, -1);
2923 if (!method) { /* demand load utf8 */
2924 ENTER;
db2c6cb3
FC
2925 if ((errsv_save = GvSV(PL_errgv))) SAVEFREESV(errsv_save);
2926 GvSV(PL_errgv) = NULL;
69794297
KW
2927 /* It is assumed that callers of this routine are not passing in
2928 * any user derived data. */
2929 /* Need to do this after save_re_context() as it will set
2930 * PL_tainted to 1 while saving $1 etc (see the code after getrx:
2931 * in Perl_magic_get). Even line to create errsv_save can turn on
2932 * PL_tainted. */
284167a5
SM
2933#ifndef NO_TAINT_SUPPORT
2934 SAVEBOOL(TAINT_get);
2935 TAINT_NOT;
2936#endif
69794297
KW
2937 Perl_load_module(aTHX_ PERL_LOADMOD_NOIMPORT, newSVpvn(pkg,pkg_len),
2938 NULL);
eed484f9 2939 {
db2c6cb3
FC
2940 /* Not ERRSV, as there is no need to vivify a scalar we are
2941 about to discard. */
2942 SV * const errsv = GvSV(PL_errgv);
2943 if (!SvTRUE(errsv)) {
2944 GvSV(PL_errgv) = SvREFCNT_inc_simple(errsv_save);
2945 SvREFCNT_dec(errsv);
2946 }
eed484f9 2947 }
69794297
KW
2948 LEAVE;
2949 }
2950 SPAGAIN;
2951 PUSHMARK(SP);
2952 EXTEND(SP,5);
2953 mPUSHp(pkg, pkg_len);
2954 mPUSHp(name, name_len);
2955 PUSHs(listsv);
2956 mPUSHi(minbits);
2957 mPUSHi(none);
2958 PUTBACK;
db2c6cb3
FC
2959 if ((errsv_save = GvSV(PL_errgv))) SAVEFREESV(errsv_save);
2960 GvSV(PL_errgv) = NULL;
69794297
KW
2961 /* If we already have a pointer to the method, no need to use
2962 * call_method() to repeat the lookup. */
c41800a8
KW
2963 if (method
2964 ? call_sv(MUTABLE_SV(method), G_SCALAR)
69794297
KW
2965 : call_sv(newSVpvs_flags("SWASHNEW", SVs_TEMP), G_SCALAR | G_METHOD))
2966 {
2967 retval = *PL_stack_sp--;
2968 SvREFCNT_inc(retval);
2969 }
eed484f9 2970 {
db2c6cb3
FC
2971 /* Not ERRSV. See above. */
2972 SV * const errsv = GvSV(PL_errgv);
2973 if (!SvTRUE(errsv)) {
2974 GvSV(PL_errgv) = SvREFCNT_inc_simple(errsv_save);
2975 SvREFCNT_dec(errsv);
2976 }
eed484f9 2977 }
ce3b816e 2978 LEAVE;
69794297
KW
2979 POPSTACK;
2980 if (IN_PERL_COMPILETIME) {
2981 CopHINTS_set(PL_curcop, PL_hints);
2982 }
2983 if (!SvROK(retval) || SvTYPE(SvRV(retval)) != SVt_PVHV) {
2984 if (SvPOK(retval))
2985
2986 /* If caller wants to handle missing properties, let them */
5d3d13d1 2987 if (flags_p && *flags_p & _CORE_SWASH_INIT_RETURN_IF_UNDEF) {
69794297
KW
2988 return NULL;
2989 }
2990 Perl_croak(aTHX_
2991 "Can't find Unicode property definition \"%"SVf"\"",
2992 SVfARG(retval));
2993 Perl_croak(aTHX_ "SWASHNEW didn't return an HV ref");
2994 }
9a53f6cf 2995 } /* End of calling the module to find the swash */
36eb48b4 2996
83199d38
KW
2997 /* If this operation fetched a swash, and we will need it later, get it */
2998 if (retval != &PL_sv_undef
2999 && (minbits == 1 || (flags_p
3000 && ! (*flags_p
3001 & _CORE_SWASH_INIT_USER_DEFINED_PROPERTY))))
3002 {
3003 swash_hv = MUTABLE_HV(SvRV(retval));
3004
3005 /* If we don't already know that there is a user-defined component to
3006 * this swash, and the user has indicated they wish to know if there is
3007 * one (by passing <flags_p>), find out */
3008 if (flags_p && ! (*flags_p & _CORE_SWASH_INIT_USER_DEFINED_PROPERTY)) {
3009 SV** user_defined = hv_fetchs(swash_hv, "USER_DEFINED", FALSE);
3010 if (user_defined && SvUV(*user_defined)) {
3011 *flags_p |= _CORE_SWASH_INIT_USER_DEFINED_PROPERTY;
3012 }
3013 }
3014 }
3015
36eb48b4
KW
3016 /* Make sure there is an inversion list for binary properties */
3017 if (minbits == 1) {
3018 SV** swash_invlistsvp = NULL;
3019 SV* swash_invlist = NULL;
9a53f6cf 3020 bool invlist_in_swash_is_valid = FALSE;
02c85471
FC
3021 bool swash_invlist_unclaimed = FALSE; /* whether swash_invlist has
3022 an unclaimed reference count */
36eb48b4 3023
9a53f6cf 3024 /* If this operation fetched a swash, get its already existing
83199d38 3025 * inversion list, or create one for it */
36eb48b4 3026
83199d38 3027 if (swash_hv) {
5c9f4bd2 3028 swash_invlistsvp = hv_fetchs(swash_hv, "V", FALSE);
9a53f6cf
KW
3029 if (swash_invlistsvp) {
3030 swash_invlist = *swash_invlistsvp;
3031 invlist_in_swash_is_valid = TRUE;
3032 }
3033 else {
36eb48b4 3034 swash_invlist = _swash_to_invlist(retval);
02c85471 3035 swash_invlist_unclaimed = TRUE;
9a53f6cf
KW
3036 }
3037 }
3038
3039 /* If an inversion list was passed in, have to include it */
3040 if (invlist) {
3041
3042 /* Any fetched swash will by now have an inversion list in it;
3043 * otherwise <swash_invlist> will be NULL, indicating that we
3044 * didn't fetch a swash */
3045 if (swash_invlist) {
3046
3047 /* Add the passed-in inversion list, which invalidates the one
3048 * already stored in the swash */
3049 invlist_in_swash_is_valid = FALSE;
3050 _invlist_union(invlist, swash_invlist, &swash_invlist);
3051 }
3052 else {
3053
87367d5f
KW
3054 /* Here, there is no swash already. Set up a minimal one, if
3055 * we are going to return a swash */
3056 if ((int) _invlist_len(invlist) > invlist_swash_boundary) {
971d486f 3057 swash_hv = newHV();
4aca0fe6 3058 retval = newRV_noinc(MUTABLE_SV(swash_hv));
87367d5f 3059 }
9a53f6cf
KW
3060 swash_invlist = invlist;
3061 }
9a53f6cf
KW
3062 }
3063
3064 /* Here, we have computed the union of all the passed-in data. It may
3065 * be that there was an inversion list in the swash which didn't get
3066 * touched; otherwise save the one computed one */
87367d5f
KW
3067 if (! invlist_in_swash_is_valid
3068 && (int) _invlist_len(swash_invlist) > invlist_swash_boundary)
3069 {
5c9f4bd2 3070 if (! hv_stores(MUTABLE_HV(SvRV(retval)), "V", swash_invlist))
69794297
KW
3071 {
3072 Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed");
3073 }
cc34d8c5
FC
3074 /* We just stole a reference count. */
3075 if (swash_invlist_unclaimed) swash_invlist_unclaimed = FALSE;
3076 else SvREFCNT_inc_simple_void_NN(swash_invlist);
9a53f6cf 3077 }
87367d5f 3078
c41800a8 3079 /* Use the inversion list stand-alone if small enough */
87367d5f
KW
3080 if ((int) _invlist_len(swash_invlist) <= invlist_swash_boundary) {
3081 SvREFCNT_dec(retval);
02c85471
FC
3082 if (!swash_invlist_unclaimed)
3083 SvREFCNT_inc_simple_void_NN(swash_invlist);
3084 retval = newRV_noinc(swash_invlist);
87367d5f 3085 }
36eb48b4
KW
3086 }
3087
a0ed51b3
LW
3088 return retval;
3089}
3090
035d37be
JH
3091
3092/* This API is wrong for special case conversions since we may need to
3093 * return several Unicode characters for a single Unicode character
3094 * (see lib/unicore/SpecCase.txt) The SWASHGET in lib/utf8_heavy.pl is
3095 * the lower-level routine, and it is similarly broken for returning
38684baa
KW
3096 * multiple values. --jhi
3097 * For those, you should use to_utf8_case() instead */
b0e3252e 3098/* Now SWASHGET is recasted into S_swatch_get in this file. */
680c470c
ST
3099
3100/* Note:
3101 * Returns the value of property/mapping C<swash> for the first character
3102 * of the string C<ptr>. If C<do_utf8> is true, the string C<ptr> is
3103 * assumed to be in utf8. If C<do_utf8> is false, the string C<ptr> is
3104 * assumed to be in native 8-bit encoding. Caches the swatch in C<swash>.
af2af982
KW
3105 *
3106 * A "swash" is a hash which contains initially the keys/values set up by
3107 * SWASHNEW. The purpose is to be able to completely represent a Unicode
3108 * property for all possible code points. Things are stored in a compact form
3109 * (see utf8_heavy.pl) so that calculation is required to find the actual
3110 * property value for a given code point. As code points are looked up, new
3111 * key/value pairs are added to the hash, so that the calculation doesn't have
3112 * to ever be re-done. Further, each calculation is done, not just for the
3113 * desired one, but for a whole block of code points adjacent to that one.
3114 * For binary properties on ASCII machines, the block is usually for 64 code
3115 * points, starting with a code point evenly divisible by 64. Thus if the
3116 * property value for code point 257 is requested, the code goes out and
3117 * calculates the property values for all 64 code points between 256 and 319,
3118 * and stores these as a single 64-bit long bit vector, called a "swatch",
3119 * under the key for code point 256. The key is the UTF-8 encoding for code
3120 * point 256, minus the final byte. Thus, if the length of the UTF-8 encoding
3121 * for a code point is 13 bytes, the key will be 12 bytes long. If the value
3122 * for code point 258 is then requested, this code realizes that it would be
3123 * stored under the key for 256, and would find that value and extract the
3124 * relevant bit, offset from 256.
3125 *
3126 * Non-binary properties are stored in as many bits as necessary to represent
3127 * their values (32 currently, though the code is more general than that), not
3128 * as single bits, but the principal is the same: the value for each key is a
3129 * vector that encompasses the property values for all code points whose UTF-8
3130 * representations are represented by the key. That is, for all code points
3131 * whose UTF-8 representations are length N bytes, and the key is the first N-1
3132 * bytes of that.
680c470c 3133 */
a0ed51b3 3134UV
680c470c 3135Perl_swash_fetch(pTHX_ SV *swash, const U8 *ptr, bool do_utf8)
a0ed51b3 3136{
27da23d5 3137 dVAR;
ef8f7699 3138 HV *const hv = MUTABLE_HV(SvRV(swash));
3568d838
JH
3139 U32 klen;
3140 U32 off;
a0ed51b3 3141 STRLEN slen;
7d85a32c 3142 STRLEN needents;
cfd0369c 3143 const U8 *tmps = NULL;
a0ed51b3 3144 U32 bit;
979f2922 3145 SV *swatch;
3568d838 3146 U8 tmputf8[2];
35da51f7 3147 const UV c = NATIVE_TO_ASCII(*ptr);
3568d838 3148
7918f24d
NC
3149 PERL_ARGS_ASSERT_SWASH_FETCH;
3150
87367d5f
KW
3151 /* If it really isn't a hash, it isn't really swash; must be an inversion
3152 * list */
3153 if (SvTYPE(hv) != SVt_PVHV) {
3154 return _invlist_contains_cp((SV*)hv,
3155 (do_utf8)
3156 ? valid_utf8_to_uvchr(ptr, NULL)
3157 : c);
3158 }
3159
dbe7a391 3160 /* Convert to utf8 if not already */
3568d838 3161 if (!do_utf8 && !UNI_IS_INVARIANT(c)) {
979f2922
ST
3162 tmputf8[0] = (U8)UTF8_EIGHT_BIT_HI(c);
3163 tmputf8[1] = (U8)UTF8_EIGHT_BIT_LO(c);
3164 ptr = tmputf8;
3568d838
JH
3165 }
3166 /* Given a UTF-X encoded char 0xAA..0xYY,0xZZ
37e2e78e 3167 * then the "swatch" is a vec() for all the chars which start
3568d838
JH
3168 * with 0xAA..0xYY
3169 * So the key in the hash (klen) is length of encoded char -1
3170 */
3171 klen = UTF8SKIP(ptr) - 1;
3172 off = ptr[klen];
a0ed51b3 3173
979f2922 3174 if (klen == 0) {
37e2e78e 3175 /* If char is invariant then swatch is for all the invariant chars
1e54db1a 3176 * In both UTF-8 and UTF-8-MOD that happens to be UTF_CONTINUATION_MARK
7d85a32c 3177 */
979f2922
ST
3178 needents = UTF_CONTINUATION_MARK;
3179 off = NATIVE_TO_UTF(ptr[klen]);
3180 }
3181 else {
7d85a32c 3182 /* If char is encoded then swatch is for the prefix */
979f2922
ST
3183 needents = (1 << UTF_ACCUMULATION_SHIFT);
3184 off = NATIVE_TO_UTF(ptr[klen]) & UTF_CONTINUATION_MASK;
3185 }
7d85a32c 3186
a0ed51b3
LW
3187 /*
3188 * This single-entry cache saves about 1/3 of the utf8 overhead in test
3189 * suite. (That is, only 7-8% overall over just a hash cache. Still,
3190 * it's nothing to sniff at.) Pity we usually come through at least
3191 * two function calls to get here...
3192 *
3193 * NB: this code assumes that swatches are never modified, once generated!
3194 */
3195
3568d838 3196 if (hv == PL_last_swash_hv &&
a0ed51b3 3197 klen == PL_last_swash_klen &&
27da23d5 3198 (!klen || memEQ((char *)ptr, (char *)PL_last_swash_key, klen)) )
a0ed51b3
LW
3199 {
3200 tmps = PL_last_swash_tmps;
3201 slen = PL_last_swash_slen;
3202 }
3203 else {
3204 /* Try our second-level swatch cache, kept in a hash. */
e1ec3a88 3205 SV** svp = hv_fetch(hv, (const char*)ptr, klen, FALSE);
a0ed51b3 3206
b0e3252e 3207 /* If not cached, generate it via swatch_get */
979f2922
ST
3208 if (!svp || !SvPOK(*svp)
3209 || !(tmps = (const U8*)SvPV_const(*svp, slen))) {
2b9d42f0
NIS
3210 /* We use utf8n_to_uvuni() as we want an index into
3211 Unicode tables, not a native character number.
3212 */
aec46f14 3213 const UV code_point = utf8n_to_uvuni(ptr, UTF8_MAXBYTES, 0,
872c91ae
JH
3214 ckWARN(WARN_UTF8) ?
3215 0 : UTF8_ALLOW_ANY);
b0e3252e 3216 swatch = swatch_get(swash,
979f2922 3217 /* On EBCDIC & ~(0xA0-1) isn't a useful thing to do */
361ee0fe 3218 (klen) ? (code_point & ~((UV)needents - 1)) : 0,
979f2922
ST
3219 needents);
3220
923e4eb5 3221 if (IN_PERL_COMPILETIME)
623e6609 3222 CopHINTS_set(PL_curcop, PL_hints);
a0ed51b3 3223
979f2922 3224 svp = hv_store(hv, (const char *)ptr, klen, swatch, 0);
a0ed51b3 3225
979f2922
ST
3226 if (!svp || !(tmps = (U8*)SvPV(*svp, slen))
3227 || (slen << 3) < needents)
5637ef5b
NC
3228 Perl_croak(aTHX_ "panic: swash_fetch got improper swatch, "
3229 "svp=%p, tmps=%p, slen=%"UVuf", needents=%"UVuf,
3230 svp, tmps, (UV)slen, (UV)needents);
a0ed51b3
LW
3231 }
3232
3233 PL_last_swash_hv = hv;
16d8f38a 3234 assert(klen <= sizeof(PL_last_swash_key));
eac04b2e 3235 PL_last_swash_klen = (U8)klen;
cfd0369c
NC
3236 /* FIXME change interpvar.h? */
3237 PL_last_swash_tmps = (U8 *) tmps;
a0ed51b3
LW
3238 PL_last_swash_slen = slen;
3239 if (klen)
3240 Copy(ptr, PL_last_swash_key, klen, U8);
3241 }
3242
9faf8d75 3243 switch ((int)((slen << 3) / needents)) {
a0ed51b3
LW
3244 case 1:
3245 bit = 1 << (off & 7);
3246 off >>= 3;
3247 return (tmps[off] & bit) != 0;
3248 case 8:
3249 return tmps[off];
3250 case 16:
3251 off <<= 1;
3252 return (tmps[off] << 8) + tmps[off + 1] ;
3253 case 32:
3254 off <<= 2;
3255 return (tmps[off] << 24) + (tmps[off+1] << 16) + (tmps[off+2] << 8) + tmps[off + 3] ;
3256 }
5637ef5b
NC
3257 Perl_croak(aTHX_ "panic: swash_fetch got swatch of unexpected bit width, "
3258 "slen=%"UVuf", needents=%"UVuf, (UV)slen, (UV)needents);
670f1322 3259 NORETURN_FUNCTION_END;
a0ed51b3 3260}
2b9d42f0 3261
319009ee
KW
3262/* Read a single line of the main body of the swash input text. These are of
3263 * the form:
3264 * 0053 0056 0073
3265 * where each number is hex. The first two numbers form the minimum and
3266 * maximum of a range, and the third is the value associated with the range.
3267 * Not all swashes should have a third number
3268 *
3269 * On input: l points to the beginning of the line to be examined; it points
3270 * to somewhere in the string of the whole input text, and is
3271 * terminated by a \n or the null string terminator.
3272 * lend points to the null terminator of that string
3273 * wants_value is non-zero if the swash expects a third number
3274 * typestr is the name of the swash's mapping, like 'ToLower'
3275 * On output: *min, *max, and *val are set to the values read from the line.
3276 * returns a pointer just beyond the line examined. If there was no
3277 * valid min number on the line, returns lend+1
3278 */
3279
3280STATIC U8*
3281S_swash_scan_list_line(pTHX_ U8* l, U8* const lend, UV* min, UV* max, UV* val,
3282 const bool wants_value, const U8* const typestr)
3283{
3284 const int typeto = typestr[0] == 'T' && typestr[1] == 'o';
3285 STRLEN numlen; /* Length of the number */
02470786
KW
3286 I32 flags = PERL_SCAN_SILENT_ILLDIGIT
3287 | PERL_SCAN_DISALLOW_PREFIX
3288 | PERL_SCAN_SILENT_NON_PORTABLE;
319009ee
KW
3289
3290 /* nl points to the next \n in the scan */
3291 U8* const nl = (U8*)memchr(l, '\n', lend - l);
3292
3293 /* Get the first number on the line: the range minimum */
3294 numlen = lend - l;
3295 *min = grok_hex((char *)l, &numlen, &flags, NULL);
3296 if (numlen) /* If found a hex number, position past it */
3297 l += numlen;
3298 else if (nl) { /* Else, go handle next line, if any */
3299 return nl + 1; /* 1 is length of "\n" */
3300 }
3301 else { /* Else, no next line */
3302 return lend + 1; /* to LIST's end at which \n is not found */
3303 }
3304
3305 /* The max range value follows, separated by a BLANK */
3306 if (isBLANK(*l)) {
3307 ++l;
02470786
KW
3308 flags = PERL_SCAN_SILENT_ILLDIGIT
3309 | PERL_SCAN_DISALLOW_PREFIX
3310 | PERL_SCAN_SILENT_NON_PORTABLE;
319009ee
KW
3311 numlen = lend - l;
3312 *max = grok_hex((char *)l, &numlen, &flags, NULL);
3313 if (numlen)
3314 l += numlen;
3315 else /* If no value here, it is a single element range */
3316 *max = *min;
3317
3318 /* Non-binary tables have a third entry: what the first element of the
3319 * range maps to */
3320 if (wants_value) {
3321 if (isBLANK(*l)) {
3322 ++l;
f90a9a02
KW
3323
3324 /* The ToLc, etc table mappings are not in hex, and must be
3325 * corrected by adding the code point to them */
3326 if (typeto) {
3327 char *after_strtol = (char *) lend;
3328 *val = Strtol((char *)l, &after_strtol, 10);
3329 l = (U8 *) after_strtol;
f90a9a02
KW
3330 }
3331 else { /* Other tables are in hex, and are the correct result
3332 without tweaking */
a9d188b3
KW
3333 flags = PERL_SCAN_SILENT_ILLDIGIT
3334 | PERL_SCAN_DISALLOW_PREFIX
3335 | PERL_SCAN_SILENT_NON_PORTABLE;
3336 numlen = lend - l;
3337 *val = grok_hex((char *)l, &numlen, &flags, NULL);
3338 if (numlen)
3339 l += numlen;
3340 else
3341 *val = 0;
f90a9a02 3342 }
319009ee
KW
3343 }
3344 else {
3345 *val = 0;
3346 if (typeto) {
dcbac5bb 3347 /* diag_listed_as: To%s: illegal mapping '%s' */
319009ee
KW
3348 Perl_croak(aTHX_ "%s: illegal mapping '%s'",
3349 typestr, l);
3350 }
3351 }
3352 }
3353 else
3354 *val = 0; /* bits == 1, then any val should be ignored */
3355 }
3356 else { /* Nothing following range min, should be single element with no
3357 mapping expected */
3358 *max = *min;
3359 if (wants_value) {
3360 *val = 0;
3361 if (typeto) {
dcbac5bb 3362 /* diag_listed_as: To%s: illegal mapping '%s' */
319009ee
KW
3363 Perl_croak(aTHX_ "%s: illegal mapping '%s'", typestr, l);
3364 }
3365 }
3366 else
3367 *val = 0; /* bits == 1, then val should be ignored */
3368 }
3369
3370 /* Position to next line if any, or EOF */
3371 if (nl)
3372 l = nl + 1;
3373 else
3374 l = lend;
3375
3376 return l;
3377}
3378
979f2922
ST
3379/* Note:
3380 * Returns a swatch (a bit vector string) for a code point sequence
3381 * that starts from the value C<start> and comprises the number C<span>.
3382 * A C<swash> must be an object created by SWASHNEW (see lib/utf8_heavy.pl).
3383 * Should be used via swash_fetch, which will cache the swatch in C<swash>.
3384 */
3385STATIC SV*
b0e3252e 3386S_swatch_get(pTHX_ SV* swash, UV start, UV span)
979f2922
ST
3387{
3388 SV *swatch;
77f9f126 3389 U8 *l, *lend, *x, *xend, *s, *send;
979f2922 3390 STRLEN lcur, xcur, scur;
ef8f7699 3391 HV *const hv = MUTABLE_HV(SvRV(swash));
5c9f4bd2 3392 SV** const invlistsvp = hv_fetchs(hv, "V", FALSE);
36eb48b4 3393
88d45d28
KW
3394 SV** listsvp = NULL; /* The string containing the main body of the table */
3395 SV** extssvp = NULL;
3396 SV** invert_it_svp = NULL;
3397 U8* typestr = NULL;
786861f5
KW
3398 STRLEN bits;
3399 STRLEN octets; /* if bits == 1, then octets == 0 */
3400 UV none;
3401 UV end = start + span;
972dd592 3402
36eb48b4 3403 if (invlistsvp == NULL) {
786861f5
KW
3404 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
3405 SV** const nonesvp = hv_fetchs(hv, "NONE", FALSE);
3406 SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE);
3407 extssvp = hv_fetchs(hv, "EXTRAS", FALSE);
3408 listsvp = hv_fetchs(hv, "LIST", FALSE);
3409 invert_it_svp = hv_fetchs(hv, "INVERT_IT", FALSE);
3410
3411 bits = SvUV(*bitssvp);
3412 none = SvUV(*nonesvp);
3413 typestr = (U8*)SvPV_nolen(*typesvp);
3414 }
36eb48b4
KW
3415 else {
3416 bits = 1;
3417 none = 0;
3418 }
786861f5 3419 octets = bits >> 3; /* if bits == 1, then octets == 0 */
979f2922 3420
b0e3252e 3421 PERL_ARGS_ASSERT_SWATCH_GET;
7918f24d 3422
979f2922 3423 if (bits != 1 && bits != 8 && bits != 16 && bits != 32) {
b0e3252e 3424 Perl_croak(aTHX_ "panic: swatch_get doesn't expect bits %"UVuf,
660a4616 3425 (UV)bits);
979f2922
ST
3426 }
3427
84ea5ef6
KW
3428 /* If overflowed, use the max possible */
3429 if (end < start) {
3430 end = UV_MAX;
3431 span = end - start;
3432 }
3433
979f2922 3434 /* create and initialize $swatch */
979f2922 3435 scur = octets ? (span * octets) : (span + 7) / 8;
e524fe40
NC
3436 swatch = newSV(scur);
3437 SvPOK_on(swatch);
979f2922
ST
3438 s = (U8*)SvPVX(swatch);
3439 if (octets && none) {
0bd48802 3440 const U8* const e = s + scur;
979f2922
ST
3441 while (s < e) {
3442 if (bits == 8)
3443 *s++ = (U8)(none & 0xff);
3444 else if (bits == 16) {
3445 *s++ = (U8)((none >> 8) & 0xff);
3446 *s++ = (U8)( none & 0xff);
3447 }
3448 else if (bits == 32) {
3449 *s++ = (U8)((none >> 24) & 0xff);
3450 *s++ = (U8)((none >> 16) & 0xff);
3451 *s++ = (U8)((none >> 8) & 0xff);
3452 *s++ = (U8)( none & 0xff);
3453 }
3454 }
3455 *s = '\0';
3456 }
3457 else {
3458 (void)memzero((U8*)s, scur + 1);
3459 }
3460 SvCUR_set(swatch, scur);
3461 s = (U8*)SvPVX(swatch);
3462
36eb48b4
KW
3463 if (invlistsvp) { /* If has an inversion list set up use that */
3464 _invlist_populate_swatch(*invlistsvp, start, end, s);
3465 return swatch;
3466 }
3467
3468 /* read $swash->{LIST} */
979f2922
ST
3469 l = (U8*)SvPV(*listsvp, lcur);
3470 lend = l + lcur;
3471 while (l < lend) {
8ed25d53 3472 UV min, max, val, upper;
319009ee
KW
3473 l = S_swash_scan_list_line(aTHX_ l, lend, &min, &max, &val,
3474 cBOOL(octets), typestr);
3475 if (l > lend) {
979f2922
ST
3476 break;
3477 }
3478
972dd592 3479 /* If looking for something beyond this range, go try the next one */
979f2922
ST
3480 if (max < start)
3481 continue;
3482
8ed25d53
KW
3483 /* <end> is generally 1 beyond where we want to set things, but at the
3484 * platform's infinity, where we can't go any higher, we want to
3485 * include the code point at <end> */
3486 upper = (max < end)
3487 ? max
3488 : (max != UV_MAX || end != UV_MAX)
3489 ? end - 1
3490 : end;
3491
979f2922 3492 if (octets) {
35da51f7 3493 UV key;
979f2922
ST
3494 if (min < start) {
3495 if (!none || val < none) {
3496 val += start - min;
3497 }
3498 min = start;
3499 }
8ed25d53 3500 for (key = min; key <= upper; key++) {
979f2922 3501 STRLEN offset;
979f2922
ST
3502 /* offset must be non-negative (start <= min <= key < end) */
3503 offset = octets * (key - start);
3504 if (bits == 8)
3505 s[offset] = (U8)(val & 0xff);
3506 else if (bits == 16) {
3507 s[offset ] = (U8)((val >> 8) & 0xff);
3508 s[offset + 1] = (U8)( val & 0xff);
3509 }
3510 else if (bits == 32) {
3511 s[offset ] = (U8)((val >> 24) & 0xff);
3512 s[offset + 1] = (U8)((val >> 16) & 0xff);
3513 s[offset + 2] = (U8)((val >> 8) & 0xff);
3514 s[offset + 3] = (U8)( val & 0xff);
3515 }
3516
3517 if (!none || val < none)
3518 ++val;
3519 }
3520 }
711a919c 3521 else { /* bits == 1, then val should be ignored */
35da51f7 3522 UV key;
979f2922
ST
3523 if (min < start)
3524 min = start;
6cb05c12 3525
8ed25d53 3526 for (key = min; key <= upper; key++) {
0bd48802 3527 const STRLEN offset = (STRLEN)(key - start);
979f2922
ST
3528 s[offset >> 3] |= 1 << (offset & 7);
3529 }
3530 }
3531 } /* while */
979f2922 3532
9479a769 3533 /* Invert if the data says it should be. Assumes that bits == 1 */
77f9f126 3534 if (invert_it_svp && SvUV(*invert_it_svp)) {
0bda3001
KW
3535
3536 /* Unicode properties should come with all bits above PERL_UNICODE_MAX
3537 * be 0, and their inversion should also be 0, as we don't succeed any
3538 * Unicode property matches for non-Unicode code points */
3539 if (start <= PERL_UNICODE_MAX) {
3540
3541 /* The code below assumes that we never cross the
3542 * Unicode/above-Unicode boundary in a range, as otherwise we would
3543 * have to figure out where to stop flipping the bits. Since this
3544 * boundary is divisible by a large power of 2, and swatches comes
3545 * in small powers of 2, this should be a valid assumption */
3546 assert(start + span - 1 <= PERL_UNICODE_MAX);
3547
507a8485
KW
3548 send = s + scur;
3549 while (s < send) {
3550 *s = ~(*s);
3551 s++;
3552 }
0bda3001 3553 }
77f9f126
KW
3554 }
3555
d73c39c5
KW
3556 /* read $swash->{EXTRAS}
3557 * This code also copied to swash_to_invlist() below */
979f2922
ST
3558 x = (U8*)SvPV(*extssvp, xcur);
3559 xend = x + xcur;
3560 while (x < xend) {
3561 STRLEN namelen;
3562 U8 *namestr;
3563 SV** othersvp;
3564 HV* otherhv;
3565 STRLEN otherbits;
3566 SV **otherbitssvp, *other;
711a919c 3567 U8 *s, *o, *nl;
979f2922
ST
3568 STRLEN slen, olen;
3569
35da51f7 3570 const U8 opc = *x++;
979f2922
ST
3571 if (opc == '\n')
3572 continue;
3573
3574 nl = (U8*)memchr(x, '\n', xend - x);
3575
3576 if (opc != '-' && opc != '+' && opc != '!' && opc != '&') {
3577 if (nl) {
3578 x = nl + 1; /* 1 is length of "\n" */
3579 continue;
3580 }
3581 else {
3582 x = xend; /* to EXTRAS' end at which \n is not found */
3583 break;
3584 }
3585 }
3586
3587 namestr = x;
3588 if (nl) {
3589 namelen = nl - namestr;
3590 x = nl + 1;
3591 }
3592 else {
3593 namelen = xend - namestr;
3594 x = xend;
3595 }
3596
3597 othersvp = hv_fetch(hv, (char *)namestr, namelen, FALSE);
ef8f7699 3598 otherhv = MUTABLE_HV(SvRV(*othersvp));
017a3ce5 3599 otherbitssvp = hv_fetchs(otherhv, "BITS", FALSE);
979f2922
ST
3600 otherbits = (STRLEN)SvUV(*otherbitssvp);
3601 if (bits < otherbits)
5637ef5b
NC
3602 Perl_croak(aTHX_ "panic: swatch_get found swatch size mismatch, "
3603 "bits=%"UVuf", otherbits=%"UVuf, (UV)bits, (UV)otherbits);
979f2922
ST
3604
3605 /* The "other" swatch must be destroyed after. */
b0e3252e 3606 other = swatch_get(*othersvp, start, span);
979f2922
ST
3607 o = (U8*)SvPV(other, olen);
3608
3609 if (!olen)
b0e3252e 3610 Perl_croak(aTHX_ "panic: swatch_get got improper swatch");
979f2922
ST
3611
3612 s = (U8*)SvPV(swatch, slen);
3613 if (bits == 1 && otherbits == 1) {
3614 if (slen != olen)
5637ef5b
NC
3615 Perl_croak(aTHX_ "panic: swatch_get found swatch length "
3616 "mismatch, slen=%"UVuf", olen=%"UVuf,
3617 (UV)slen, (UV)olen);
979f2922
ST
3618
3619 switch (opc) {
3620 case '+':
3621 while (slen--)
3622 *s++ |= *o++;
3623 break;
3624 case '!':
3625 while (slen--)
3626 *s++ |= ~*o++;
3627 break;
3628 case '-':
3629 while (slen--)
3630 *s++ &= ~*o++;
3631 break;
3632 case '&':
3633 while (slen--)
3634 *s++ &= *o++;
3635 break;
3636 default:
3637 break;
3638 }
3639 }
711a919c 3640 else {
979f2922
ST
3641 STRLEN otheroctets = otherbits >> 3;
3642 STRLEN offset = 0;
35da51f7 3643 U8* const send = s + slen;
979f2922
ST
3644
3645 while (s < send) {
3646 UV otherval = 0;
3647
3648 if (otherbits == 1) {
3649 otherval = (o[offset >> 3] >> (offset & 7)) & 1;
3650 ++offset;
3651 }
3652 else {
3653 STRLEN vlen = otheroctets;
3654 otherval = *o++;
3655 while (--vlen) {
3656 otherval <<= 8;
3657 otherval |= *o++;
3658 }
3659 }
3660
711a919c 3661 if (opc == '+' && otherval)
6f207bd3 3662 NOOP; /* replace with otherval */
979f2922
ST
3663 else if (opc == '!' && !otherval)
3664 otherval = 1;
3665 else if (opc == '-' && otherval)
3666 otherval = 0;
3667 else if (opc == '&' && !otherval)
3668 otherval = 0;
3669 else {
711a919c 3670 s += octets; /* no replacement */
979f2922
ST
3671 continue;
3672 }
3673
3674 if (bits == 8)
3675 *s++ = (U8)( otherval & 0xff);
3676 else if (bits == 16) {
3677 *s++ = (U8)((otherval >> 8) & 0xff);
3678 *s++ = (U8)( otherval & 0xff);
3679 }
3680 else if (bits == 32) {
3681 *s++ = (U8)((otherval >> 24) & 0xff);
3682 *s++ = (U8)((otherval >> 16) & 0xff);
3683 *s++ = (U8)((otherval >> 8) & 0xff);
3684 *s++ = (U8)( otherval & 0xff);
3685 }
3686 }
3687 }
3688 sv_free(other); /* through with it! */
3689 } /* while */
3690 return swatch;
3691}
3692
064c021d 3693HV*
4c2e1131 3694Perl__swash_inversion_hash(pTHX_ SV* const swash)
064c021d
KW
3695{
3696
79a2a0e8 3697 /* Subject to change or removal. For use only in regcomp.c and regexec.c
5662e334
KW
3698 * Can't be used on a property that is subject to user override, as it
3699 * relies on the value of SPECIALS in the swash which would be set by
3700 * utf8_heavy.pl to the hash in the non-overriden file, and hence is not set
3701 * for overridden properties
064c021d
KW
3702 *
3703 * Returns a hash which is the inversion and closure of a swash mapping.
3704 * For example, consider the input lines:
3705 * 004B 006B
3706 * 004C 006C
3707 * 212A 006B
3708 *
3709 * The returned hash would have two keys, the utf8 for 006B and the utf8 for
3710 * 006C. The value for each key is an array. For 006C, the array would
3711 * have a two elements, the utf8 for itself, and for 004C. For 006B, there
3712 * would be