This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
charnames: fix nit in comment
[perl5.git] / utf8.c
CommitLineData
a0ed51b3
LW
1/* utf8.c
2 *
1129b882 3 * Copyright (C) 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008
b94e2f88 4 * by Larry Wall and others
a0ed51b3
LW
5 *
6 * You may distribute under the terms of either the GNU General Public
7 * License or the Artistic License, as specified in the README file.
8 *
9 */
10
11/*
4ac71550
TC
12 * 'What a fix!' said Sam. 'That's the one place in all the lands we've ever
13 * heard of that we don't want to see any closer; and that's the one place
14 * we're trying to get to! And that's just where we can't get, nohow.'
15 *
cdad3b53 16 * [p.603 of _The Lord of the Rings_, IV/I: "The Taming of Sméagol"]
a0ed51b3
LW
17 *
18 * 'Well do I understand your speech,' he answered in the same language;
19 * 'yet few strangers do so. Why then do you not speak in the Common Tongue,
4ac71550 20 * as is the custom in the West, if you wish to be answered?'
cdad3b53 21 * --Gandalf, addressing Théoden's door wardens
4ac71550
TC
22 *
23 * [p.508 of _The Lord of the Rings_, III/vi: "The King of the Golden Hall"]
a0ed51b3
LW
24 *
25 * ...the travellers perceived that the floor was paved with stones of many
26 * hues; branching runes and strange devices intertwined beneath their feet.
4ac71550
TC
27 *
28 * [p.512 of _The Lord of the Rings_, III/vi: "The King of the Golden Hall"]
a0ed51b3
LW
29 */
30
31#include "EXTERN.h"
864dbfa3 32#define PERL_IN_UTF8_C
a0ed51b3 33#include "perl.h"
81e983c1 34#include "inline_invlist.c"
a0ed51b3 35
a0c21aa1 36#ifndef EBCDIC
970ea3cb 37/* Separate prototypes needed because in ASCII systems these are
a0c21aa1
JH
38 * usually macros but they still are compiled as code, too. */
39PERL_CALLCONV UV Perl_utf8n_to_uvchr(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen, U32 flags);
6dd9dce9 40PERL_CALLCONV UV Perl_valid_utf8_to_uvchr(pTHX_ const U8 *s, STRLEN *retlen);
a0c21aa1
JH
41PERL_CALLCONV U8* Perl_uvchr_to_utf8(pTHX_ U8 *d, UV uv);
42#endif
43
27da23d5
JH
44static const char unees[] =
45 "Malformed UTF-8 character (unexpected end of string)";
901b21bf 46
48ef279e 47/*
ccfc67b7 48=head1 Unicode Support
a0ed51b3 49
166f8a29
DM
50This file contains various utility functions for manipulating UTF8-encoded
51strings. For the uninitiated, this is a method of representing arbitrary
61296642 52Unicode characters as a variable number of bytes, in such a way that
56da48f7
DM
53characters in the ASCII range are unmodified, and a zero byte never appears
54within non-zero characters.
166f8a29 55
eaf7a4d2
CS
56=cut
57*/
58
59/*
60=for apidoc is_ascii_string
61
a1433954 62Returns true if the first C<len> bytes of the string C<s> are the same whether
970ea3cb
KW
63or not the string is encoded in UTF-8 (or UTF-EBCDIC on EBCDIC machines). That
64is, if they are invariant. On ASCII-ish machines, only ASCII characters
65fit this definition, hence the function's name.
eaf7a4d2 66
9f7e3d64
MH
67If C<len> is 0, it will be calculated using C<strlen(s)>.
68
a1433954 69See also L</is_utf8_string>(), L</is_utf8_string_loclen>(), and L</is_utf8_string_loc>().
eaf7a4d2
CS
70
71=cut
72*/
73
74bool
668b6d8d 75Perl_is_ascii_string(const U8 *s, STRLEN len)
eaf7a4d2
CS
76{
77 const U8* const send = s + (len ? len : strlen((const char *)s));
78 const U8* x = s;
79
80 PERL_ARGS_ASSERT_IS_ASCII_STRING;
eaf7a4d2
CS
81
82 for (; x < send; ++x) {
83 if (!UTF8_IS_INVARIANT(*x))
84 break;
85 }
86
87 return x == send;
88}
89
90/*
87cea99e 91=for apidoc uvuni_to_utf8_flags
eebe1485 92
537124e4 93Adds the UTF-8 representation of the Unicode code point C<uv> to the end
6ee84de2 94of the string C<d>; C<d> should have at least C<UTF8_MAXBYTES+1> free
eebe1485 95bytes available. The return value is the pointer to the byte after the
9041c2e3 96end of the new character. In other words,
eebe1485 97
b851fbc1
JH
98 d = uvuni_to_utf8_flags(d, uv, flags);
99
100or, in most cases,
101
9041c2e3 102 d = uvuni_to_utf8(d, uv);
eebe1485 103
b851fbc1
JH
104(which is equivalent to)
105
106 d = uvuni_to_utf8_flags(d, uv, 0);
107
949cf498 108This is the recommended Unicode-aware way of saying
eebe1485
SC
109
110 *(d++) = uv;
111
537124e4
KW
112where uv is a code point expressed in Latin-1 or above, not the platform's
113native character set. B<Almost all code should instead use L</uvchr_to_utf8>
114or L</uvchr_to_utf8_flags>>.
115
949cf498
KW
116This function will convert to UTF-8 (and not warn) even code points that aren't
117legal Unicode or are problematic, unless C<flags> contains one or more of the
a1433954
KW
118following flags:
119
949cf498
KW
120If C<uv> is a Unicode surrogate code point and UNICODE_WARN_SURROGATE is set,
121the function will raise a warning, provided UTF8 warnings are enabled. If instead
122UNICODE_DISALLOW_SURROGATE is set, the function will fail and return NULL.
123If both flags are set, the function will both warn and return NULL.
124
125The UNICODE_WARN_NONCHAR and UNICODE_DISALLOW_NONCHAR flags correspondingly
537124e4
KW
126affect how the function handles a Unicode non-character. And likewise, the
127UNICODE_WARN_SUPER and UNICODE_DISALLOW_SUPER flags, affect the handling of
128code points that are
949cf498
KW
129above the Unicode maximum of 0x10FFFF. Code points above 0x7FFF_FFFF (which are
130even less portable) can be warned and/or disallowed even if other above-Unicode
131code points are accepted by the UNICODE_WARN_FE_FF and UNICODE_DISALLOW_FE_FF
132flags.
133
134And finally, the flag UNICODE_WARN_ILLEGAL_INTERCHANGE selects all four of the
135above WARN flags; and UNICODE_DISALLOW_ILLEGAL_INTERCHANGE selects all four
136DISALLOW flags.
137
138
eebe1485
SC
139=cut
140*/
141
dfe13c55 142U8 *
b851fbc1 143Perl_uvuni_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags)
a0ed51b3 144{
7918f24d
NC
145 PERL_ARGS_ASSERT_UVUNI_TO_UTF8_FLAGS;
146
979f77b6
KW
147 /* The first problematic code point is the first surrogate */
148 if (uv >= UNICODE_SURROGATE_FIRST
149 && ckWARN4_d(WARN_UTF8, WARN_SURROGATE, WARN_NON_UNICODE, WARN_NONCHAR))
150 {
949cf498
KW
151 if (UNICODE_IS_SURROGATE(uv)) {
152 if (flags & UNICODE_WARN_SURROGATE) {
8457b38f 153 Perl_ck_warner_d(aTHX_ packWARN(WARN_SURROGATE),
949cf498
KW
154 "UTF-16 surrogate U+%04"UVXf, uv);
155 }
156 if (flags & UNICODE_DISALLOW_SURROGATE) {
157 return NULL;
158 }
159 }
160 else if (UNICODE_IS_SUPER(uv)) {
161 if (flags & UNICODE_WARN_SUPER
162 || (UNICODE_IS_FE_FF(uv) && (flags & UNICODE_WARN_FE_FF)))
163 {
8457b38f 164 Perl_ck_warner_d(aTHX_ packWARN(WARN_NON_UNICODE),
949cf498
KW
165 "Code point 0x%04"UVXf" is not Unicode, may not be portable", uv);
166 }
167 if (flags & UNICODE_DISALLOW_SUPER
168 || (UNICODE_IS_FE_FF(uv) && (flags & UNICODE_DISALLOW_FE_FF)))
169 {
170 return NULL;
171 }
172 }
173 else if (UNICODE_IS_NONCHAR(uv)) {
174 if (flags & UNICODE_WARN_NONCHAR) {
8457b38f 175 Perl_ck_warner_d(aTHX_ packWARN(WARN_NONCHAR),
949cf498
KW
176 "Unicode non-character U+%04"UVXf" is illegal for open interchange",
177 uv);
178 }
179 if (flags & UNICODE_DISALLOW_NONCHAR) {
180 return NULL;
181 }
182 }
507b9800 183 }
c4d5f83a 184 if (UNI_IS_INVARIANT(uv)) {
eb160463 185 *d++ = (U8)UTF_TO_NATIVE(uv);
a0ed51b3
LW
186 return d;
187 }
2d331972 188#if defined(EBCDIC)
1d72bdf6
NIS
189 else {
190 STRLEN len = UNISKIP(uv);
191 U8 *p = d+len-1;
192 while (p > d) {
eb160463 193 *p-- = (U8)UTF_TO_NATIVE((uv & UTF_CONTINUATION_MASK) | UTF_CONTINUATION_MARK);
1d72bdf6
NIS
194 uv >>= UTF_ACCUMULATION_SHIFT;
195 }
eb160463 196 *p = (U8)UTF_TO_NATIVE((uv & UTF_START_MASK(len)) | UTF_START_MARK(len));
1d72bdf6
NIS
197 return d+len;
198 }
199#else /* Non loop style */
a0ed51b3 200 if (uv < 0x800) {
eb160463
GS
201 *d++ = (U8)(( uv >> 6) | 0xc0);
202 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
203 return d;
204 }
205 if (uv < 0x10000) {
eb160463
GS
206 *d++ = (U8)(( uv >> 12) | 0xe0);
207 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
208 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
209 return d;
210 }
211 if (uv < 0x200000) {
eb160463
GS
212 *d++ = (U8)(( uv >> 18) | 0xf0);
213 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
214 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
215 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
216 return d;
217 }
218 if (uv < 0x4000000) {
eb160463
GS
219 *d++ = (U8)(( uv >> 24) | 0xf8);
220 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
221 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
222 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
223 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
224 return d;
225 }
226 if (uv < 0x80000000) {
eb160463
GS
227 *d++ = (U8)(( uv >> 30) | 0xfc);
228 *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80);
229 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
230 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
231 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
232 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
233 return d;
234 }
6b8eaf93 235#ifdef HAS_QUAD
d7578b48 236 if (uv < UTF8_QUAD_MAX)
a0ed51b3
LW
237#endif
238 {
eb160463
GS
239 *d++ = 0xfe; /* Can't match U+FEFF! */
240 *d++ = (U8)(((uv >> 30) & 0x3f) | 0x80);
241 *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80);
242 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
243 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
244 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
245 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
246 return d;
247 }
6b8eaf93 248#ifdef HAS_QUAD
a0ed51b3 249 {
eb160463
GS
250 *d++ = 0xff; /* Can't match U+FFFE! */
251 *d++ = 0x80; /* 6 Reserved bits */
252 *d++ = (U8)(((uv >> 60) & 0x0f) | 0x80); /* 2 Reserved bits */
253 *d++ = (U8)(((uv >> 54) & 0x3f) | 0x80);
254 *d++ = (U8)(((uv >> 48) & 0x3f) | 0x80);
255 *d++ = (U8)(((uv >> 42) & 0x3f) | 0x80);
256 *d++ = (U8)(((uv >> 36) & 0x3f) | 0x80);
257 *d++ = (U8)(((uv >> 30) & 0x3f) | 0x80);
258 *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80);
259 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
260 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
261 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
262 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
263 return d;
264 }
265#endif
537124e4 266#endif /* Non loop style */
a0ed51b3 267}
9041c2e3 268
646ca15d
JH
269/*
270
f7d739d1 271Tests if the first C<len> bytes of string C<s> form a valid UTF-8
646ca15d 272character. Note that an INVARIANT (i.e. ASCII) character is a valid
f7d739d1 273UTF-8 character. The number of bytes in the UTF-8 character
646ca15d
JH
274will be returned if it is valid, otherwise 0.
275
276This is the "slow" version as opposed to the "fast" version which is
277the "unrolled" IS_UTF8_CHAR(). E.g. for t/uni/class.t the speed
278difference is a factor of 2 to 3. For lengths (UTF8SKIP(s)) of four
279or less you should use the IS_UTF8_CHAR(), for lengths of five or more
280you should use the _slow(). In practice this means that the _slow()
281will be used very rarely, since the maximum Unicode code point (as of
282Unicode 4.1) is U+10FFFF, which encodes in UTF-8 to four bytes. Only
537124e4 283the "Perl extended UTF-8" (e.g, the infamous 'v-strings') will encode into
646ca15d
JH
284five bytes or more.
285
286=cut */
7af276bc 287PERL_STATIC_INLINE STRLEN
5f66b61c 288S_is_utf8_char_slow(const U8 *s, const STRLEN len)
646ca15d 289{
cd7e6c88 290 dTHX; /* The function called below requires thread context */
646ca15d 291
cd7e6c88 292 STRLEN actual_len;
646ca15d 293
cd7e6c88 294 PERL_ARGS_ASSERT_IS_UTF8_CHAR_SLOW;
646ca15d 295
cd7e6c88 296 utf8n_to_uvuni(s, len, &actual_len, UTF8_CHECK_ONLY);
646ca15d 297
cd7e6c88 298 return (actual_len == (STRLEN) -1) ? 0 : actual_len;
646ca15d 299}
9041c2e3
NIS
300
301/*
492a624f
KW
302=for apidoc is_utf8_char_buf
303
304Returns the number of bytes that comprise the first UTF-8 encoded character in
305buffer C<buf>. C<buf_end> should point to one position beyond the end of the
306buffer. 0 is returned if C<buf> does not point to a complete, valid UTF-8
307encoded character.
308
309Note that an INVARIANT character (i.e. ASCII on non-EBCDIC
310machines) is a valid UTF-8 character.
311
312=cut */
313
314STRLEN
315Perl_is_utf8_char_buf(const U8 *buf, const U8* buf_end)
316{
317
318 STRLEN len;
319
320 PERL_ARGS_ASSERT_IS_UTF8_CHAR_BUF;
321
322 if (buf_end <= buf) {
323 return 0;
324 }
325
326 len = buf_end - buf;
327 if (len > UTF8SKIP(buf)) {
328 len = UTF8SKIP(buf);
329 }
330
331#ifdef IS_UTF8_CHAR
332 if (IS_UTF8_CHAR_FAST(len))
333 return IS_UTF8_CHAR(buf, len) ? len : 0;
334#endif /* #ifdef IS_UTF8_CHAR */
335 return is_utf8_char_slow(buf, len);
336}
337
338/*
87cea99e 339=for apidoc is_utf8_char
eebe1485 340
76848387
KW
341DEPRECATED!
342
5da9da9e 343Tests if some arbitrary number of bytes begins in a valid UTF-8
2bbc8d55
SP
344character. Note that an INVARIANT (i.e. ASCII on non-EBCDIC machines)
345character is a valid UTF-8 character. The actual number of bytes in the UTF-8
346character will be returned if it is valid, otherwise 0.
9041c2e3 347
76848387 348This function is deprecated due to the possibility that malformed input could
a1433954 349cause reading beyond the end of the input buffer. Use L</is_utf8_char_buf>
76848387 350instead.
e0328548 351
82686b01 352=cut */
76848387 353
067a85ef 354STRLEN
668b6d8d 355Perl_is_utf8_char(const U8 *s)
386d01d6 356{
7918f24d 357 PERL_ARGS_ASSERT_IS_UTF8_CHAR;
492a624f 358
76848387 359 /* Assumes we have enough space, which is why this is deprecated */
492a624f 360 return is_utf8_char_buf(s, s + UTF8SKIP(s));
386d01d6
GS
361}
362
eaf7a4d2 363
6662521e 364/*
87cea99e 365=for apidoc is_utf8_string
6662521e 366
a1433954 367Returns true if the first C<len> bytes of string C<s> form a valid
9f7e3d64 368UTF-8 string, false otherwise. If C<len> is 0, it will be calculated
e0328548
KW
369using C<strlen(s)> (which means if you use this option, that C<s> has to have a
370terminating NUL byte). Note that all characters being ASCII constitute 'a
371valid UTF-8 string'.
6662521e 372
a1433954 373See also L</is_ascii_string>(), L</is_utf8_string_loclen>(), and L</is_utf8_string_loc>().
768c67ee 374
6662521e
GS
375=cut
376*/
377
8e84507e 378bool
668b6d8d 379Perl_is_utf8_string(const U8 *s, STRLEN len)
6662521e 380{
35da51f7 381 const U8* const send = s + (len ? len : strlen((const char *)s));
7fc63493 382 const U8* x = s;
067a85ef 383
7918f24d 384 PERL_ARGS_ASSERT_IS_UTF8_STRING;
1aa99e6b 385
6662521e 386 while (x < send) {
1acdb0da 387 /* Inline the easy bits of is_utf8_char() here for speed... */
e0328548
KW
388 if (UTF8_IS_INVARIANT(*x)) {
389 x++;
390 }
1acdb0da
JH
391 else {
392 /* ... and call is_utf8_char() only if really needed. */
e0328548
KW
393 const STRLEN c = UTF8SKIP(x);
394 const U8* const next_char_ptr = x + c;
395
396 if (next_char_ptr > send) {
397 return FALSE;
398 }
399
768c67ee
JH
400 if (IS_UTF8_CHAR_FAST(c)) {
401 if (!IS_UTF8_CHAR(x, c))
e0328548 402 return FALSE;
3c614e38 403 }
e0328548
KW
404 else if (! is_utf8_char_slow(x, c)) {
405 return FALSE;
406 }
407 x = next_char_ptr;
1acdb0da 408 }
6662521e 409 }
768c67ee 410
067a85ef 411 return TRUE;
6662521e
GS
412}
413
67e989fb 414/*
814fafa7
NC
415Implemented as a macro in utf8.h
416
87cea99e 417=for apidoc is_utf8_string_loc
814fafa7 418
a1433954
KW
419Like L</is_utf8_string> but stores the location of the failure (in the
420case of "utf8ness failure") or the location C<s>+C<len> (in the case of
814fafa7
NC
421"utf8ness success") in the C<ep>.
422
a1433954 423See also L</is_utf8_string_loclen>() and L</is_utf8_string>().
814fafa7 424
87cea99e 425=for apidoc is_utf8_string_loclen
81cd54e3 426
a1433954
KW
427Like L</is_utf8_string>() but stores the location of the failure (in the
428case of "utf8ness failure") or the location C<s>+C<len> (in the case of
768c67ee
JH
429"utf8ness success") in the C<ep>, and the number of UTF-8
430encoded characters in the C<el>.
431
a1433954 432See also L</is_utf8_string_loc>() and L</is_utf8_string>().
81cd54e3
JH
433
434=cut
435*/
436
437bool
668b6d8d 438Perl_is_utf8_string_loclen(const U8 *s, STRLEN len, const U8 **ep, STRLEN *el)
81cd54e3 439{
35da51f7 440 const U8* const send = s + (len ? len : strlen((const char *)s));
7fc63493 441 const U8* x = s;
81cd54e3 442 STRLEN c;
3ebfea28 443 STRLEN outlen = 0;
7918f24d
NC
444
445 PERL_ARGS_ASSERT_IS_UTF8_STRING_LOCLEN;
81cd54e3 446
81cd54e3 447 while (x < send) {
e0328548
KW
448 const U8* next_char_ptr;
449
81cd54e3
JH
450 /* Inline the easy bits of is_utf8_char() here for speed... */
451 if (UTF8_IS_INVARIANT(*x))
e0328548 452 next_char_ptr = x + 1;
81cd54e3 453 else {
768c67ee 454 /* ... and call is_utf8_char() only if really needed. */
768c67ee 455 c = UTF8SKIP(x);
e0328548
KW
456 next_char_ptr = c + x;
457 if (next_char_ptr > send) {
458 goto out;
459 }
768c67ee
JH
460 if (IS_UTF8_CHAR_FAST(c)) {
461 if (!IS_UTF8_CHAR(x, c))
462 c = 0;
463 } else
464 c = is_utf8_char_slow(x, c);
768c67ee
JH
465 if (!c)
466 goto out;
81cd54e3 467 }
e0328548 468 x = next_char_ptr;
3ebfea28 469 outlen++;
81cd54e3 470 }
768c67ee
JH
471
472 out:
3ebfea28
AL
473 if (el)
474 *el = outlen;
475
768c67ee
JH
476 if (ep)
477 *ep = x;
3ebfea28 478 return (x == send);
81cd54e3
JH
479}
480
481/*
768c67ee 482
87cea99e 483=for apidoc utf8n_to_uvuni
67e989fb 484
9041c2e3 485Bottom level UTF-8 decode routine.
746afd53
KW
486Returns the code point value of the first character in the string C<s>,
487which is assumed to be in UTF-8 (or UTF-EBCDIC) encoding, and no longer than
488C<curlen> bytes; C<*retlen> (if C<retlen> isn't NULL) will be set to
489the length, in bytes, of that character.
949cf498
KW
490
491The value of C<flags> determines the behavior when C<s> does not point to a
492well-formed UTF-8 character. If C<flags> is 0, when a malformation is found,
524080c4
KW
493zero is returned and C<*retlen> is set so that (S<C<s> + C<*retlen>>) is the
494next possible position in C<s> that could begin a non-malformed character.
495Also, if UTF-8 warnings haven't been lexically disabled, a warning is raised.
949cf498
KW
496
497Various ALLOW flags can be set in C<flags> to allow (and not warn on)
498individual types of malformations, such as the sequence being overlong (that
499is, when there is a shorter sequence that can express the same code point;
500overlong sequences are expressly forbidden in the UTF-8 standard due to
501potential security issues). Another malformation example is the first byte of
502a character not being a legal first byte. See F<utf8.h> for the list of such
524080c4
KW
503flags. For allowed 0 length strings, this function returns 0; for allowed
504overlong sequences, the computed code point is returned; for all other allowed
505malformations, the Unicode REPLACEMENT CHARACTER is returned, as these have no
506determinable reasonable value.
949cf498
KW
507
508The UTF8_CHECK_ONLY flag overrides the behavior when a non-allowed (by other
509flags) malformation is found. If this flag is set, the routine assumes that
510the caller will raise a warning, and this function will silently just set
d088425d
KW
511C<retlen> to C<-1> (cast to C<STRLEN>) and return zero.
512
513Note that this API requires disambiguation between successful decoding a NUL
514character, and an error return (unless the UTF8_CHECK_ONLY flag is set), as
515in both cases, 0 is returned. To disambiguate, upon a zero return, see if the
516first byte of C<s> is 0 as well. If so, the input was a NUL; if not, the input
517had an error.
949cf498
KW
518
519Certain code points are considered problematic. These are Unicode surrogates,
746afd53 520Unicode non-characters, and code points above the Unicode maximum of 0x10FFFF.
949cf498 521By default these are considered regular code points, but certain situations
5eafe189 522warrant special handling for them. If C<flags> contains
949cf498
KW
523UTF8_DISALLOW_ILLEGAL_INTERCHANGE, all three classes are treated as
524malformations and handled as such. The flags UTF8_DISALLOW_SURROGATE,
525UTF8_DISALLOW_NONCHAR, and UTF8_DISALLOW_SUPER (meaning above the legal Unicode
526maximum) can be set to disallow these categories individually.
527
528The flags UTF8_WARN_ILLEGAL_INTERCHANGE, UTF8_WARN_SURROGATE,
529UTF8_WARN_NONCHAR, and UTF8_WARN_SUPER will cause warning messages to be raised
530for their respective categories, but otherwise the code points are considered
531valid (not malformations). To get a category to both be treated as a
532malformation and raise a warning, specify both the WARN and DISALLOW flags.
533(But note that warnings are not raised if lexically disabled nor if
534UTF8_CHECK_ONLY is also specified.)
535
536Very large code points (above 0x7FFF_FFFF) are considered more problematic than
537the others that are above the Unicode legal maximum. There are several
eb83ed87
KW
538reasons: they requre at least 32 bits to represent them on ASCII platforms, are
539not representable at all on EBCDIC platforms, and the original UTF-8
540specification never went above this number (the current 0x10FFFF limit was
541imposed later). (The smaller ones, those that fit into 32 bits, are
542representable by a UV on ASCII platforms, but not by an IV, which means that
543the number of operations that can be performed on them is quite restricted.)
544The UTF-8 encoding on ASCII platforms for these large code points begins with a
545byte containing 0xFE or 0xFF. The UTF8_DISALLOW_FE_FF flag will cause them to
546be treated as malformations, while allowing smaller above-Unicode code points.
547(Of course UTF8_DISALLOW_SUPER will treat all above-Unicode code points,
548including these, as malformations.) Similarly, UTF8_WARN_FE_FF acts just like
549the other WARN flags, but applies just to these code points.
949cf498
KW
550
551All other code points corresponding to Unicode characters, including private
552use and those yet to be assigned, are never considered malformed and never
553warn.
67e989fb 554
ec5f19d0 555Most code should use L</utf8_to_uvchr_buf>() rather than call this directly.
9041c2e3 556
37607a96
PK
557=cut
558*/
67e989fb 559
a0ed51b3 560UV
7fc63493 561Perl_utf8n_to_uvuni(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen, U32 flags)
a0ed51b3 562{
97aff369 563 dVAR;
d4c19fe8 564 const U8 * const s0 = s;
eb83ed87 565 U8 overflow_byte = '\0'; /* Save byte in case of overflow */
0b8d30e8 566 U8 * send;
eb83ed87
KW
567 UV uv = *s;
568 STRLEN expectlen;
949cf498 569 SV* sv = NULL;
eb83ed87
KW
570 UV outlier_ret = 0; /* return value when input is in error or problematic
571 */
572 UV pack_warn = 0; /* Save result of packWARN() for later */
573 bool unexpected_non_continuation = FALSE;
574 bool overflowed = FALSE;
2f8f112e 575 bool do_overlong_test = TRUE; /* May have to skip this test */
a0dbb045 576
eb83ed87 577 const char* const malformed_text = "Malformed UTF-8 character";
7918f24d 578
eb83ed87 579 PERL_ARGS_ASSERT_UTF8N_TO_UVUNI;
a0dbb045 580
eb83ed87
KW
581 /* The order of malformation tests here is important. We should consume as
582 * few bytes as possible in order to not skip any valid character. This is
583 * required by the Unicode Standard (section 3.9 of Unicode 6.0); see also
584 * http://unicode.org/reports/tr36 for more discussion as to why. For
585 * example, once we've done a UTF8SKIP, we can tell the expected number of
586 * bytes, and could fail right off the bat if the input parameters indicate
587 * that there are too few available. But it could be that just that first
588 * byte is garbled, and the intended character occupies fewer bytes. If we
589 * blindly assumed that the first byte is correct, and skipped based on
590 * that number, we could skip over a valid input character. So instead, we
591 * always examine the sequence byte-by-byte.
592 *
593 * We also should not consume too few bytes, otherwise someone could inject
594 * things. For example, an input could be deliberately designed to
595 * overflow, and if this code bailed out immediately upon discovering that,
596 * returning to the caller *retlen pointing to the very next byte (one
597 * which is actually part of of the overflowing sequence), that could look
598 * legitimate to the caller, which could discard the initial partial
599 * sequence and process the rest, inappropriately */
600
601 /* Zero length strings, if allowed, of necessity are zero */
b5b9af04 602 if (UNLIKELY(curlen == 0)) {
eb83ed87
KW
603 if (retlen) {
604 *retlen = 0;
605 }
a0dbb045 606
eb83ed87
KW
607 if (flags & UTF8_ALLOW_EMPTY) {
608 return 0;
609 }
610 if (! (flags & UTF8_CHECK_ONLY)) {
611 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (empty string)", malformed_text));
612 }
0c443dc2
JH
613 goto malformed;
614 }
615
eb83ed87
KW
616 expectlen = UTF8SKIP(s);
617
618 /* A well-formed UTF-8 character, as the vast majority of calls to this
619 * function will be for, has this expected length. For efficiency, set
620 * things up here to return it. It will be overriden only in those rare
621 * cases where a malformation is found */
622 if (retlen) {
623 *retlen = expectlen;
624 }
625
626 /* An invariant is trivially well-formed */
1d72bdf6 627 if (UTF8_IS_INVARIANT(uv)) {
c4d5f83a 628 return (UV) (NATIVE_TO_UTF(*s));
a0ed51b3 629 }
67e989fb 630
eb83ed87 631 /* A continuation character can't start a valid sequence */
b5b9af04 632 if (UNLIKELY(UTF8_IS_CONTINUATION(uv))) {
eb83ed87
KW
633 if (flags & UTF8_ALLOW_CONTINUATION) {
634 if (retlen) {
635 *retlen = 1;
636 }
637 return UNICODE_REPLACEMENT;
638 }
ba210ebe 639
eb83ed87
KW
640 if (! (flags & UTF8_CHECK_ONLY)) {
641 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (unexpected continuation byte 0x%02x, with no preceding start byte)", malformed_text, *s0));
642 }
643 curlen = 1;
ba210ebe
JH
644 goto malformed;
645 }
9041c2e3 646
1d72bdf6 647#ifdef EBCDIC
75383841 648 uv = NATIVE_TO_UTF(uv);
1d72bdf6
NIS
649#endif
650
eb83ed87
KW
651 /* Here is not a continuation byte, nor an invariant. The only thing left
652 * is a start byte (possibly for an overlong) */
ba210ebe 653
eb83ed87
KW
654 /* Remove the leading bits that indicate the number of bytes in the
655 * character's whole UTF-8 sequence, leaving just the bits that are part of
656 * the value */
657 uv &= UTF_START_MASK(expectlen);
ba210ebe 658
eb83ed87
KW
659 /* Now, loop through the remaining bytes in the character's sequence,
660 * accumulating each into the working value as we go. Be sure to not look
661 * past the end of the input string */
0b8d30e8
KW
662 send = (U8*) s0 + ((expectlen <= curlen) ? expectlen : curlen);
663
eb83ed87 664 for (s = s0 + 1; s < send; s++) {
b5b9af04 665 if (LIKELY(UTF8_IS_CONTINUATION(*s))) {
eb83ed87
KW
666#ifndef EBCDIC /* Can't overflow in EBCDIC */
667 if (uv & UTF_ACCUMULATION_OVERFLOW_MASK) {
668
669 /* The original implementors viewed this malformation as more
670 * serious than the others (though I, khw, don't understand
671 * why, since other malformations also give very very wrong
672 * results), so there is no way to turn off checking for it.
673 * Set a flag, but keep going in the loop, so that we absorb
674 * the rest of the bytes that comprise the character. */
675 overflowed = TRUE;
676 overflow_byte = *s; /* Save for warning message's use */
677 }
678#endif
8850bf83 679 uv = UTF8_ACCUMULATE(uv, *s);
eb83ed87
KW
680 }
681 else {
682 /* Here, found a non-continuation before processing all expected
683 * bytes. This byte begins a new character, so quit, even if
684 * allowing this malformation. */
685 unexpected_non_continuation = TRUE;
686 break;
687 }
688 } /* End of loop through the character's bytes */
689
690 /* Save how many bytes were actually in the character */
691 curlen = s - s0;
692
693 /* The loop above finds two types of malformations: non-continuation and/or
694 * overflow. The non-continuation malformation is really a too-short
695 * malformation, as it means that the current character ended before it was
696 * expected to (being terminated prematurely by the beginning of the next
697 * character, whereas in the too-short malformation there just are too few
698 * bytes available to hold the character. In both cases, the check below
699 * that we have found the expected number of bytes would fail if executed.)
700 * Thus the non-continuation malformation is really unnecessary, being a
701 * subset of the too-short malformation. But there may be existing
702 * applications that are expecting the non-continuation type, so we retain
703 * it, and return it in preference to the too-short malformation. (If this
704 * code were being written from scratch, the two types might be collapsed
705 * into one.) I, khw, am also giving priority to returning the
706 * non-continuation and too-short malformations over overflow when multiple
707 * ones are present. I don't know of any real reason to prefer one over
708 * the other, except that it seems to me that multiple-byte errors trumps
709 * errors from a single byte */
b5b9af04 710 if (UNLIKELY(unexpected_non_continuation)) {
eb83ed87
KW
711 if (!(flags & UTF8_ALLOW_NON_CONTINUATION)) {
712 if (! (flags & UTF8_CHECK_ONLY)) {
713 if (curlen == 1) {
714 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (unexpected non-continuation byte 0x%02x, immediately after start byte 0x%02x)", malformed_text, *s, *s0));
715 }
716 else {
717 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (unexpected non-continuation byte 0x%02x, %d bytes after start byte 0x%02x, expected %d bytes)", malformed_text, *s, (int) curlen, *s0, (int)expectlen));
a0dbb045
JH
718 }
719 }
eb83ed87
KW
720 goto malformed;
721 }
722 uv = UNICODE_REPLACEMENT;
2f8f112e
KW
723
724 /* Skip testing for overlongs, as the REPLACEMENT may not be the same
725 * as what the original expectations were. */
726 do_overlong_test = FALSE;
eb83ed87
KW
727 if (retlen) {
728 *retlen = curlen;
729 }
730 }
b5b9af04 731 else if (UNLIKELY(curlen < expectlen)) {
eb83ed87
KW
732 if (! (flags & UTF8_ALLOW_SHORT)) {
733 if (! (flags & UTF8_CHECK_ONLY)) {
734 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (%d byte%s, need %d, after start byte 0x%02x)", malformed_text, (int)curlen, curlen == 1 ? "" : "s", (int)expectlen, *s0));
a0dbb045 735 }
eb83ed87
KW
736 goto malformed;
737 }
738 uv = UNICODE_REPLACEMENT;
2f8f112e 739 do_overlong_test = FALSE;
eb83ed87
KW
740 if (retlen) {
741 *retlen = curlen;
742 }
743 }
744
745#ifndef EBCDIC /* EBCDIC allows FE, FF, can't overflow */
2f8f112e 746 if ((*s0 & 0xFE) == 0xFE /* matches both FE, FF */
eb83ed87
KW
747 && (flags & (UTF8_WARN_FE_FF|UTF8_DISALLOW_FE_FF)))
748 {
749 /* By adding UTF8_CHECK_ONLY to the test, we avoid unnecessary
750 * generation of the sv, since no warnings are raised under CHECK */
751 if ((flags & (UTF8_WARN_FE_FF|UTF8_CHECK_ONLY)) == UTF8_WARN_FE_FF
752 && ckWARN_d(WARN_UTF8))
753 {
42303544
KW
754 /* This message is deliberately not of the same syntax as the other
755 * messages for malformations, for backwards compatibility in the
756 * unlikely event that code is relying on its precise earlier text
757 */
eb83ed87
KW
758 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s Code point beginning with byte 0x%02X is not Unicode, and not portable", malformed_text, *s0));
759 pack_warn = packWARN(WARN_UTF8);
760 }
761 if (flags & UTF8_DISALLOW_FE_FF) {
762 goto malformed;
ba210ebe 763 }
ba210ebe 764 }
b5b9af04 765 if (UNLIKELY(overflowed)) {
ba210ebe 766
eb83ed87
KW
767 /* If the first byte is FF, it will overflow a 32-bit word. If the
768 * first byte is FE, it will overflow a signed 32-bit word. The
769 * above preserves backward compatibility, since its message was used
770 * in earlier versions of this code in preference to overflow */
771 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (overflow at byte 0x%02x, after start byte 0x%02x)", malformed_text, overflow_byte, *s0));
ba210ebe 772 goto malformed;
eb83ed87
KW
773 }
774#endif
775
2f8f112e
KW
776 if (do_overlong_test
777 && expectlen > (STRLEN)UNISKIP(uv)
778 && ! (flags & UTF8_ALLOW_LONG))
779 {
eb83ed87
KW
780 /* The overlong malformation has lower precedence than the others.
781 * Note that if this malformation is allowed, we return the actual
782 * value, instead of the replacement character. This is because this
783 * value is actually well-defined. */
784 if (! (flags & UTF8_CHECK_ONLY)) {
785 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (%d byte%s, need %d, after start byte 0x%02x)", malformed_text, (int)expectlen, expectlen == 1 ? "": "s", UNISKIP(uv), *s0));
786 }
787 goto malformed;
788 }
789
790 /* Here, the input is considered to be well-formed , but could be a
791 * problematic code point that is not allowed by the input parameters. */
792 if (uv >= UNICODE_SURROGATE_FIRST /* isn't problematic if < this */
793 && (flags & (UTF8_DISALLOW_ILLEGAL_INTERCHANGE
794 |UTF8_WARN_ILLEGAL_INTERCHANGE)))
795 {
949cf498 796 if (UNICODE_IS_SURROGATE(uv)) {
eb83ed87
KW
797 if ((flags & (UTF8_WARN_SURROGATE|UTF8_CHECK_ONLY)) == UTF8_WARN_SURROGATE
798 && ckWARN2_d(WARN_UTF8, WARN_SURROGATE))
799 {
111d382d 800 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "UTF-16 surrogate U+%04"UVXf"", uv));
eb83ed87 801 pack_warn = packWARN2(WARN_UTF8, WARN_SURROGATE);
949cf498
KW
802 }
803 if (flags & UTF8_DISALLOW_SURROGATE) {
804 goto disallowed;
805 }
806 }
949cf498 807 else if ((uv > PERL_UNICODE_MAX)) {
eb83ed87
KW
808 if ((flags & (UTF8_WARN_SUPER|UTF8_CHECK_ONLY)) == UTF8_WARN_SUPER
809 && ckWARN2_d(WARN_UTF8, WARN_NON_UNICODE))
810 {
111d382d 811 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "Code point 0x%04"UVXf" is not Unicode, may not be portable", uv));
eb83ed87 812 pack_warn = packWARN2(WARN_UTF8, WARN_NON_UNICODE);
949cf498
KW
813 }
814 if (flags & UTF8_DISALLOW_SUPER) {
815 goto disallowed;
816 }
817 }
4190d317
KW
818 else if (UNICODE_IS_NONCHAR(uv)) {
819 if ((flags & (UTF8_WARN_NONCHAR|UTF8_CHECK_ONLY)) == UTF8_WARN_NONCHAR
820 && ckWARN2_d(WARN_UTF8, WARN_NONCHAR))
821 {
822 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "Unicode non-character U+%04"UVXf" is illegal for open interchange", uv));
823 pack_warn = packWARN2(WARN_UTF8, WARN_NONCHAR);
824 }
825 if (flags & UTF8_DISALLOW_NONCHAR) {
826 goto disallowed;
827 }
828 }
949cf498 829
eb83ed87
KW
830 if (sv) {
831 outlier_ret = uv;
832 goto do_warn;
833 }
834
949cf498
KW
835 /* Here, this is not considered a malformed character, so drop through
836 * to return it */
a0ed51b3 837 }
ba210ebe 838
a0ed51b3 839 return uv;
ba210ebe 840
eb83ed87
KW
841 /* There are three cases which get to beyond this point. In all 3 cases:
842 * <sv> if not null points to a string to print as a warning.
843 * <curlen> is what <*retlen> should be set to if UTF8_CHECK_ONLY isn't
844 * set.
845 * <outlier_ret> is what return value to use if UTF8_CHECK_ONLY isn't set.
846 * This is done by initializing it to 0, and changing it only
847 * for case 1).
848 * The 3 cases are:
849 * 1) The input is valid but problematic, and to be warned about. The
850 * return value is the resultant code point; <*retlen> is set to
851 * <curlen>, the number of bytes that comprise the code point.
852 * <pack_warn> contains the result of packWARN() for the warning
853 * types. The entry point for this case is the label <do_warn>;
854 * 2) The input is a valid code point but disallowed by the parameters to
855 * this function. The return value is 0. If UTF8_CHECK_ONLY is set,
856 * <*relen> is -1; otherwise it is <curlen>, the number of bytes that
857 * comprise the code point. <pack_warn> contains the result of
858 * packWARN() for the warning types. The entry point for this case is
859 * the label <disallowed>.
860 * 3) The input is malformed. The return value is 0. If UTF8_CHECK_ONLY
861 * is set, <*relen> is -1; otherwise it is <curlen>, the number of
862 * bytes that comprise the malformation. All such malformations are
863 * assumed to be warning type <utf8>. The entry point for this case
864 * is the label <malformed>.
865 */
949cf498 866
ba210ebe
JH
867malformed:
868
eb83ed87
KW
869 if (sv && ckWARN_d(WARN_UTF8)) {
870 pack_warn = packWARN(WARN_UTF8);
871 }
872
873disallowed:
874
fcc8fcf6 875 if (flags & UTF8_CHECK_ONLY) {
ba210ebe 876 if (retlen)
10edeb5d 877 *retlen = ((STRLEN) -1);
ba210ebe
JH
878 return 0;
879 }
880
eb83ed87 881do_warn:
5b311467 882
eb83ed87
KW
883 if (pack_warn) { /* <pack_warn> was initialized to 0, and changed only
884 if warnings are to be raised. */
f555bc63 885 const char * const string = SvPVX_const(sv);
a0dbb045 886
f555bc63
KW
887 if (PL_op)
888 Perl_warner(aTHX_ pack_warn, "%s in %s", string, OP_DESC(PL_op));
889 else
890 Perl_warner(aTHX_ pack_warn, "%s", string);
a0dbb045
JH
891 }
892
eb83ed87
KW
893 if (retlen) {
894 *retlen = curlen;
895 }
ba210ebe 896
eb83ed87 897 return outlier_ret;
a0ed51b3
LW
898}
899
8e84507e 900/*
ec5f19d0
KW
901=for apidoc utf8_to_uvchr_buf
902
903Returns the native code point of the first character in the string C<s> which
904is assumed to be in UTF-8 encoding; C<send> points to 1 beyond the end of C<s>.
524080c4 905C<*retlen> will be set to the length, in bytes, of that character.
ec5f19d0 906
524080c4
KW
907If C<s> does not point to a well-formed UTF-8 character and UTF8 warnings are
908enabled, zero is returned and C<*retlen> is set (if C<retlen> isn't
173db420
KW
909NULL) to -1. If those warnings are off, the computed value, if well-defined
910(or the Unicode REPLACEMENT CHARACTER if not), is silently returned, and
911C<*retlen> is set (if C<retlen> isn't NULL) so that (S<C<s> + C<*retlen>>) is
912the next possible position in C<s> that could begin a non-malformed character.
913See L</utf8n_to_uvuni> for details on when the REPLACEMENT CHARACTER is
914returned.
ec5f19d0
KW
915
916=cut
917*/
918
919
920UV
921Perl_utf8_to_uvchr_buf(pTHX_ const U8 *s, const U8 *send, STRLEN *retlen)
922{
923 PERL_ARGS_ASSERT_UTF8_TO_UVCHR_BUF;
924
925 assert(s < send);
926
927 return utf8n_to_uvchr(s, send - s, retlen,
928 ckWARN_d(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
929}
930
27d6c58a 931/* Like L</utf8_to_uvchr_buf>(), but should only be called when it is known that
3986bb7c 932 * there are no malformations in the input UTF-8 string C<s>. surrogates,
6dd9dce9
KW
933 * non-character code points, and non-Unicode code points are allowed. A macro
934 * in utf8.h is used to normally avoid this function wrapper */
27d6c58a
KW
935
936UV
937Perl_valid_utf8_to_uvchr(pTHX_ const U8 *s, STRLEN *retlen)
938{
3986bb7c
KW
939 const UV uv = valid_utf8_to_uvuni(s, retlen);
940
27d6c58a
KW
941 PERL_ARGS_ASSERT_VALID_UTF8_TO_UVCHR;
942
3986bb7c 943 return UNI_TO_NATIVE(uv);
27d6c58a
KW
944}
945
ec5f19d0 946/*
87cea99e 947=for apidoc utf8_to_uvchr
9041c2e3 948
977c1d31
KW
949DEPRECATED!
950
6ee84de2 951Returns the native code point of the first character in the string C<s>
1e54db1a 952which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
9041c2e3
NIS
953length, in bytes, of that character.
954
4b88fb76 955Some, but not all, UTF-8 malformations are detected, and in fact, some
977c1d31
KW
956malformed input could cause reading beyond the end of the input buffer, which
957is why this function is deprecated. Use L</utf8_to_uvchr_buf> instead.
4b88fb76 958
524080c4
KW
959If C<s> points to one of the detected malformations, and UTF8 warnings are
960enabled, zero is returned and C<*retlen> is set (if C<retlen> isn't
961NULL) to -1. If those warnings are off, the computed value if well-defined (or
962the Unicode REPLACEMENT CHARACTER, if not) is silently returned, and C<*retlen>
963is set (if C<retlen> isn't NULL) so that (S<C<s> + C<*retlen>>) is the
964next possible position in C<s> that could begin a non-malformed character.
965See L</utf8n_to_uvuni> for details on when the REPLACEMENT CHARACTER is returned.
9041c2e3
NIS
966
967=cut
968*/
969
970UV
7fc63493 971Perl_utf8_to_uvchr(pTHX_ const U8 *s, STRLEN *retlen)
9041c2e3 972{
7918f24d
NC
973 PERL_ARGS_ASSERT_UTF8_TO_UVCHR;
974
2ff6c191 975 return utf8_to_uvchr_buf(s, s + UTF8_MAXBYTES, retlen);
9041c2e3
NIS
976}
977
978/*
ec5f19d0
KW
979=for apidoc utf8_to_uvuni_buf
980
981Returns the Unicode code point of the first character in the string C<s> which
982is assumed to be in UTF-8 encoding; C<send> points to 1 beyond the end of C<s>.
983C<retlen> will be set to the length, in bytes, of that character.
984
985This function should only be used when the returned UV is considered
986an index into the Unicode semantic tables (e.g. swashes).
987
524080c4
KW
988If C<s> does not point to a well-formed UTF-8 character and UTF8 warnings are
989enabled, zero is returned and C<*retlen> is set (if C<retlen> isn't
990NULL) to -1. If those warnings are off, the computed value if well-defined (or
991the Unicode REPLACEMENT CHARACTER, if not) is silently returned, and C<*retlen>
992is set (if C<retlen> isn't NULL) so that (S<C<s> + C<*retlen>>) is the
993next possible position in C<s> that could begin a non-malformed character.
994See L</utf8n_to_uvuni> for details on when the REPLACEMENT CHARACTER is returned.
ec5f19d0
KW
995
996=cut
997*/
998
999UV
1000Perl_utf8_to_uvuni_buf(pTHX_ const U8 *s, const U8 *send, STRLEN *retlen)
1001{
1002 PERL_ARGS_ASSERT_UTF8_TO_UVUNI_BUF;
1003
1004 assert(send > s);
1005
1006 /* Call the low level routine asking for checks */
1007 return Perl_utf8n_to_uvuni(aTHX_ s, send -s, retlen,
1008 ckWARN_d(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
1009}
1010
27d6c58a 1011/* Like L</utf8_to_uvuni_buf>(), but should only be called when it is known that
2114036c 1012 * there are no malformations in the input UTF-8 string C<s>. Surrogates,
3986bb7c 1013 * non-character code points, and non-Unicode code points are allowed */
27d6c58a
KW
1014
1015UV
1016Perl_valid_utf8_to_uvuni(pTHX_ const U8 *s, STRLEN *retlen)
1017{
3986bb7c
KW
1018 UV expectlen = UTF8SKIP(s);
1019 const U8* send = s + expectlen;
1020 UV uv = NATIVE_TO_UTF(*s);
1021
27d6c58a
KW
1022 PERL_ARGS_ASSERT_VALID_UTF8_TO_UVUNI;
1023
3986bb7c
KW
1024 if (retlen) {
1025 *retlen = expectlen;
1026 }
1027
1028 /* An invariant is trivially returned */
1029 if (expectlen == 1) {
1030 return uv;
1031 }
1032
1033 /* Remove the leading bits that indicate the number of bytes, leaving just
1034 * the bits that are part of the value */
1035 uv &= UTF_START_MASK(expectlen);
1036
1037 /* Now, loop through the remaining bytes, accumulating each into the
08bc774e
KW
1038 * working total as we go. (I khw tried unrolling the loop for up to 4
1039 * bytes, but there was no performance improvement) */
3986bb7c
KW
1040 for (++s; s < send; s++) {
1041 uv = UTF8_ACCUMULATE(uv, *s);
1042 }
1043
1044 return uv;
27d6c58a
KW
1045}
1046
ec5f19d0 1047/*
87cea99e 1048=for apidoc utf8_to_uvuni
9041c2e3 1049
977c1d31
KW
1050DEPRECATED!
1051
9041c2e3 1052Returns the Unicode code point of the first character in the string C<s>
1e54db1a 1053which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
9041c2e3
NIS
1054length, in bytes, of that character.
1055
39e518fd
KW
1056This function should only be used when the returned UV is considered
1057an index into the Unicode semantic tables (e.g. swashes).
1058
4b88fb76 1059Some, but not all, UTF-8 malformations are detected, and in fact, some
977c1d31
KW
1060malformed input could cause reading beyond the end of the input buffer, which
1061is why this function is deprecated. Use L</utf8_to_uvuni_buf> instead.
9041c2e3 1062
524080c4
KW
1063If C<s> points to one of the detected malformations, and UTF8 warnings are
1064enabled, zero is returned and C<*retlen> is set (if C<retlen> doesn't point to
1065NULL) to -1. If those warnings are off, the computed value if well-defined (or
1066the Unicode REPLACEMENT CHARACTER, if not) is silently returned, and C<*retlen>
1067is set (if C<retlen> isn't NULL) so that (S<C<s> + C<*retlen>>) is the
1068next possible position in C<s> that could begin a non-malformed character.
1069See L</utf8n_to_uvuni> for details on when the REPLACEMENT CHARACTER is returned.
8e84507e
NIS
1070
1071=cut
1072*/
1073
1074UV
7fc63493 1075Perl_utf8_to_uvuni(pTHX_ const U8 *s, STRLEN *retlen)
8e84507e 1076{
7918f24d
NC
1077 PERL_ARGS_ASSERT_UTF8_TO_UVUNI;
1078
4b88fb76 1079 return valid_utf8_to_uvuni(s, retlen);
8e84507e
NIS
1080}
1081
b76347f2 1082/*
87cea99e 1083=for apidoc utf8_length
b76347f2
JH
1084
1085Return the length of the UTF-8 char encoded string C<s> in characters.
02eb7b47
JH
1086Stops at C<e> (inclusive). If C<e E<lt> s> or if the scan would end
1087up past C<e>, croaks.
b76347f2
JH
1088
1089=cut
1090*/
1091
1092STRLEN
35a4481c 1093Perl_utf8_length(pTHX_ const U8 *s, const U8 *e)
b76347f2 1094{
97aff369 1095 dVAR;
b76347f2
JH
1096 STRLEN len = 0;
1097
7918f24d
NC
1098 PERL_ARGS_ASSERT_UTF8_LENGTH;
1099
8850bf83
JH
1100 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g.
1101 * the bitops (especially ~) can create illegal UTF-8.
1102 * In other words: in Perl UTF-8 is not just for Unicode. */
1103
a3b680e6
AL
1104 if (e < s)
1105 goto warn_and_return;
b76347f2 1106 while (s < e) {
4cbf4130 1107 s += UTF8SKIP(s);
8e91ec7f
AV
1108 len++;
1109 }
1110
1111 if (e != s) {
1112 len--;
1113 warn_and_return:
9b387841
NC
1114 if (PL_op)
1115 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8),
1116 "%s in %s", unees, OP_DESC(PL_op));
1117 else
61a12c31 1118 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), "%s", unees);
b76347f2
JH
1119 }
1120
1121 return len;
1122}
1123
b06226ff 1124/*
87cea99e 1125=for apidoc utf8_distance
b06226ff 1126
1e54db1a 1127Returns the number of UTF-8 characters between the UTF-8 pointers C<a>
b06226ff
JH
1128and C<b>.
1129
1130WARNING: use only if you *know* that the pointers point inside the
1131same UTF-8 buffer.
1132
37607a96
PK
1133=cut
1134*/
a0ed51b3 1135
02eb7b47 1136IV
35a4481c 1137Perl_utf8_distance(pTHX_ const U8 *a, const U8 *b)
a0ed51b3 1138{
7918f24d
NC
1139 PERL_ARGS_ASSERT_UTF8_DISTANCE;
1140
bf1665bc 1141 return (a < b) ? -1 * (IV) utf8_length(a, b) : (IV) utf8_length(b, a);
a0ed51b3
LW
1142}
1143
b06226ff 1144/*
87cea99e 1145=for apidoc utf8_hop
b06226ff 1146
8850bf83
JH
1147Return the UTF-8 pointer C<s> displaced by C<off> characters, either
1148forward or backward.
b06226ff
JH
1149
1150WARNING: do not use the following unless you *know* C<off> is within
8850bf83
JH
1151the UTF-8 data pointed to by C<s> *and* that on entry C<s> is aligned
1152on the first byte of character or just after the last byte of a character.
b06226ff 1153
37607a96
PK
1154=cut
1155*/
a0ed51b3
LW
1156
1157U8 *
4373e329 1158Perl_utf8_hop(pTHX_ const U8 *s, I32 off)
a0ed51b3 1159{
7918f24d
NC
1160 PERL_ARGS_ASSERT_UTF8_HOP;
1161
96a5add6 1162 PERL_UNUSED_CONTEXT;
8850bf83
JH
1163 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g
1164 * the bitops (especially ~) can create illegal UTF-8.
1165 * In other words: in Perl UTF-8 is not just for Unicode. */
1166
a0ed51b3
LW
1167 if (off >= 0) {
1168 while (off--)
1169 s += UTF8SKIP(s);
1170 }
1171 else {
1172 while (off++) {
1173 s--;
8850bf83
JH
1174 while (UTF8_IS_CONTINUATION(*s))
1175 s--;
a0ed51b3
LW
1176 }
1177 }
4373e329 1178 return (U8 *)s;
a0ed51b3
LW
1179}
1180
6940069f 1181/*
fed3ba5d
NC
1182=for apidoc bytes_cmp_utf8
1183
a1433954
KW
1184Compares the sequence of characters (stored as octets) in C<b>, C<blen> with the
1185sequence of characters (stored as UTF-8) in C<u>, C<ulen>. Returns 0 if they are
fed3ba5d
NC
1186equal, -1 or -2 if the first string is less than the second string, +1 or +2
1187if the first string is greater than the second string.
1188
1189-1 or +1 is returned if the shorter string was identical to the start of the
1190longer string. -2 or +2 is returned if the was a difference between characters
1191within the strings.
1192
1193=cut
1194*/
1195
1196int
1197Perl_bytes_cmp_utf8(pTHX_ const U8 *b, STRLEN blen, const U8 *u, STRLEN ulen)
1198{
1199 const U8 *const bend = b + blen;
1200 const U8 *const uend = u + ulen;
1201
1202 PERL_ARGS_ASSERT_BYTES_CMP_UTF8;
1203
1204 PERL_UNUSED_CONTEXT;
1205
1206 while (b < bend && u < uend) {
1207 U8 c = *u++;
1208 if (!UTF8_IS_INVARIANT(c)) {
1209 if (UTF8_IS_DOWNGRADEABLE_START(c)) {
1210 if (u < uend) {
1211 U8 c1 = *u++;
1212 if (UTF8_IS_CONTINUATION(c1)) {
356979f4 1213 c = UNI_TO_NATIVE(TWO_BYTE_UTF8_TO_UNI(c, c1));
fed3ba5d
NC
1214 } else {
1215 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8),
1216 "Malformed UTF-8 character "
1217 "(unexpected non-continuation byte 0x%02x"
1218 ", immediately after start byte 0x%02x)"
1219 /* Dear diag.t, it's in the pod. */
1220 "%s%s", c1, c,
1221 PL_op ? " in " : "",
1222 PL_op ? OP_DESC(PL_op) : "");
1223 return -2;
1224 }
1225 } else {
1226 if (PL_op)
1227 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8),
1228 "%s in %s", unees, OP_DESC(PL_op));
1229 else
61a12c31 1230 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), "%s", unees);
fed3ba5d
NC
1231 return -2; /* Really want to return undef :-) */
1232 }
1233 } else {
1234 return -2;
1235 }
1236 }
1237 if (*b != c) {
1238 return *b < c ? -2 : +2;
1239 }
1240 ++b;
1241 }
1242
1243 if (b == bend && u == uend)
1244 return 0;
1245
1246 return b < bend ? +1 : -1;
1247}
1248
1249/*
87cea99e 1250=for apidoc utf8_to_bytes
6940069f 1251
2bbc8d55 1252Converts a string C<s> of length C<len> from UTF-8 into native byte encoding.
a1433954
KW
1253Unlike L</bytes_to_utf8>, this over-writes the original string, and
1254updates C<len> to contain the new length.
67e989fb 1255Returns zero on failure, setting C<len> to -1.
6940069f 1256
a1433954 1257If you need a copy of the string, see L</bytes_from_utf8>.
95be277c 1258
6940069f
GS
1259=cut
1260*/
1261
1262U8 *
37607a96 1263Perl_utf8_to_bytes(pTHX_ U8 *s, STRLEN *len)
6940069f 1264{
d4c19fe8
AL
1265 U8 * const save = s;
1266 U8 * const send = s + *len;
6940069f 1267 U8 *d;
246fae53 1268
7918f24d
NC
1269 PERL_ARGS_ASSERT_UTF8_TO_BYTES;
1270
1e54db1a 1271 /* ensure valid UTF-8 and chars < 256 before updating string */
d4c19fe8 1272 while (s < send) {
dcad2880
JH
1273 U8 c = *s++;
1274
1d72bdf6
NIS
1275 if (!UTF8_IS_INVARIANT(c) &&
1276 (!UTF8_IS_DOWNGRADEABLE_START(c) || (s >= send)
1277 || !(c = *s++) || !UTF8_IS_CONTINUATION(c))) {
10edeb5d 1278 *len = ((STRLEN) -1);
dcad2880
JH
1279 return 0;
1280 }
246fae53 1281 }
dcad2880
JH
1282
1283 d = s = save;
6940069f 1284 while (s < send) {
ed646e6e 1285 STRLEN ulen;
4b88fb76 1286 *d++ = (U8)utf8_to_uvchr_buf(s, send, &ulen);
ed646e6e 1287 s += ulen;
6940069f
GS
1288 }
1289 *d = '\0';
246fae53 1290 *len = d - save;
6940069f
GS
1291 return save;
1292}
1293
1294/*
87cea99e 1295=for apidoc bytes_from_utf8
f9a63242 1296
2bbc8d55 1297Converts a string C<s> of length C<len> from UTF-8 into native byte encoding.
a1433954 1298Unlike L</utf8_to_bytes> but like L</bytes_to_utf8>, returns a pointer to
ef9edfd0
JH
1299the newly-created string, and updates C<len> to contain the new
1300length. Returns the original string if no conversion occurs, C<len>
1301is unchanged. Do nothing if C<is_utf8> points to 0. Sets C<is_utf8> to
2bbc8d55
SP
13020 if C<s> is converted or consisted entirely of characters that are invariant
1303in utf8 (i.e., US-ASCII on non-EBCDIC machines).
f9a63242 1304
37607a96
PK
1305=cut
1306*/
f9a63242
JH
1307
1308U8 *
e1ec3a88 1309Perl_bytes_from_utf8(pTHX_ const U8 *s, STRLEN *len, bool *is_utf8)
f9a63242 1310{
f9a63242 1311 U8 *d;
e1ec3a88
AL
1312 const U8 *start = s;
1313 const U8 *send;
f9a63242
JH
1314 I32 count = 0;
1315
7918f24d
NC
1316 PERL_ARGS_ASSERT_BYTES_FROM_UTF8;
1317
96a5add6 1318 PERL_UNUSED_CONTEXT;
f9a63242 1319 if (!*is_utf8)
73d840c0 1320 return (U8 *)start;
f9a63242 1321
1e54db1a 1322 /* ensure valid UTF-8 and chars < 256 before converting string */
f9a63242 1323 for (send = s + *len; s < send;) {
e1ec3a88 1324 U8 c = *s++;
1d72bdf6 1325 if (!UTF8_IS_INVARIANT(c)) {
db42d148
NIS
1326 if (UTF8_IS_DOWNGRADEABLE_START(c) && s < send &&
1327 (c = *s++) && UTF8_IS_CONTINUATION(c))
1328 count++;
1329 else
73d840c0 1330 return (U8 *)start;
db42d148 1331 }
f9a63242
JH
1332 }
1333
35da51f7 1334 *is_utf8 = FALSE;
f9a63242 1335
212542aa 1336 Newx(d, (*len) - count + 1, U8);
ef9edfd0 1337 s = start; start = d;
f9a63242
JH
1338 while (s < send) {
1339 U8 c = *s++;
c4d5f83a
NIS
1340 if (!UTF8_IS_INVARIANT(c)) {
1341 /* Then it is two-byte encoded */
356979f4 1342 c = UNI_TO_NATIVE(TWO_BYTE_UTF8_TO_UNI(c, *s++));
c4d5f83a
NIS
1343 }
1344 *d++ = c;
f9a63242
JH
1345 }
1346 *d = '\0';
1347 *len = d - start;
73d840c0 1348 return (U8 *)start;
f9a63242
JH
1349}
1350
1351/*
87cea99e 1352=for apidoc bytes_to_utf8
6940069f 1353
ff97e5cf
KW
1354Converts a string C<s> of length C<len> bytes from the native encoding into
1355UTF-8.
6662521e 1356Returns a pointer to the newly-created string, and sets C<len> to
ff97e5cf 1357reflect the new length in bytes.
6940069f 1358
2bbc8d55
SP
1359A NUL character will be written after the end of the string.
1360
1361If you want to convert to UTF-8 from encodings other than
1362the native (Latin1 or EBCDIC),
a1433954 1363see L</sv_recode_to_utf8>().
c9ada85f 1364
497711e7 1365=cut
6940069f
GS
1366*/
1367
c682ebef
FC
1368/* This logic is duplicated in sv_catpvn_flags, so any bug fixes will
1369 likewise need duplication. */
1370
6940069f 1371U8*
35a4481c 1372Perl_bytes_to_utf8(pTHX_ const U8 *s, STRLEN *len)
6940069f 1373{
35a4481c 1374 const U8 * const send = s + (*len);
6940069f
GS
1375 U8 *d;
1376 U8 *dst;
7918f24d
NC
1377
1378 PERL_ARGS_ASSERT_BYTES_TO_UTF8;
96a5add6 1379 PERL_UNUSED_CONTEXT;
6940069f 1380
212542aa 1381 Newx(d, (*len) * 2 + 1, U8);
6940069f
GS
1382 dst = d;
1383
1384 while (s < send) {
35a4481c 1385 const UV uv = NATIVE_TO_ASCII(*s++);
c4d5f83a 1386 if (UNI_IS_INVARIANT(uv))
eb160463 1387 *d++ = (U8)UTF_TO_NATIVE(uv);
6940069f 1388 else {
eb160463
GS
1389 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
1390 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
6940069f
GS
1391 }
1392 }
1393 *d = '\0';
6662521e 1394 *len = d-dst;
6940069f
GS
1395 return dst;
1396}
1397
a0ed51b3 1398/*
dea0fc0b 1399 * Convert native (big-endian) or reversed (little-endian) UTF-16 to UTF-8.
a0ed51b3
LW
1400 *
1401 * Destination must be pre-extended to 3/2 source. Do not use in-place.
1402 * We optimize for native, for obvious reasons. */
1403
1404U8*
dea0fc0b 1405Perl_utf16_to_utf8(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen)
a0ed51b3 1406{
dea0fc0b
JH
1407 U8* pend;
1408 U8* dstart = d;
1409
7918f24d
NC
1410 PERL_ARGS_ASSERT_UTF16_TO_UTF8;
1411
dea0fc0b 1412 if (bytelen & 1)
f5992bc4 1413 Perl_croak(aTHX_ "panic: utf16_to_utf8: odd bytelen %"UVuf, (UV)bytelen);
dea0fc0b
JH
1414
1415 pend = p + bytelen;
1416
a0ed51b3 1417 while (p < pend) {
dea0fc0b
JH
1418 UV uv = (p[0] << 8) + p[1]; /* UTF-16BE */
1419 p += 2;
a0ed51b3 1420 if (uv < 0x80) {
e294cc5d
JH
1421#ifdef EBCDIC
1422 *d++ = UNI_TO_NATIVE(uv);
1423#else
eb160463 1424 *d++ = (U8)uv;
e294cc5d 1425#endif
a0ed51b3
LW
1426 continue;
1427 }
1428 if (uv < 0x800) {
eb160463
GS
1429 *d++ = (U8)(( uv >> 6) | 0xc0);
1430 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
1431 continue;
1432 }
52b9aa85 1433 if (uv >= 0xd800 && uv <= 0xdbff) { /* surrogates */
01ea242b 1434 if (p >= pend) {
dea0fc0b 1435 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
01ea242b
NC
1436 } else {
1437 UV low = (p[0] << 8) + p[1];
1438 p += 2;
52b9aa85 1439 if (low < 0xdc00 || low > 0xdfff)
01ea242b
NC
1440 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
1441 uv = ((uv - 0xd800) << 10) + (low - 0xdc00) + 0x10000;
1442 }
dbde1951
NC
1443 } else if (uv >= 0xdc00 && uv <= 0xdfff) {
1444 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
a0ed51b3
LW
1445 }
1446 if (uv < 0x10000) {
eb160463
GS
1447 *d++ = (U8)(( uv >> 12) | 0xe0);
1448 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
1449 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
1450 continue;
1451 }
1452 else {
eb160463
GS
1453 *d++ = (U8)(( uv >> 18) | 0xf0);
1454 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
1455 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
1456 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
1457 continue;
1458 }
1459 }
dea0fc0b 1460 *newlen = d - dstart;
a0ed51b3
LW
1461 return d;
1462}
1463
1464/* Note: this one is slightly destructive of the source. */
1465
1466U8*
dea0fc0b 1467Perl_utf16_to_utf8_reversed(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen)
a0ed51b3
LW
1468{
1469 U8* s = (U8*)p;
d4c19fe8 1470 U8* const send = s + bytelen;
7918f24d
NC
1471
1472 PERL_ARGS_ASSERT_UTF16_TO_UTF8_REVERSED;
1473
e0ea5e2d
NC
1474 if (bytelen & 1)
1475 Perl_croak(aTHX_ "panic: utf16_to_utf8_reversed: odd bytelen %"UVuf,
1476 (UV)bytelen);
1477
a0ed51b3 1478 while (s < send) {
d4c19fe8 1479 const U8 tmp = s[0];
a0ed51b3
LW
1480 s[0] = s[1];
1481 s[1] = tmp;
1482 s += 2;
1483 }
dea0fc0b 1484 return utf16_to_utf8(p, d, bytelen, newlen);
a0ed51b3
LW
1485}
1486
922e8cb4
KW
1487bool
1488Perl__is_uni_FOO(pTHX_ const U8 classnum, const UV c)
1489{
1490 U8 tmpbuf[UTF8_MAXBYTES+1];
1491 uvchr_to_utf8(tmpbuf, c);
1492 return _is_utf8_FOO(classnum, tmpbuf);
1493}
1494
c3fd2246
KW
1495/* for now these are all defined (inefficiently) in terms of the utf8 versions.
1496 * Note that the macros in handy.h that call these short-circuit calling them
1497 * for Latin-1 range inputs */
a0ed51b3
LW
1498
1499bool
84afefe6 1500Perl_is_uni_alnum(pTHX_ UV c)
a0ed51b3 1501{
89ebb4a3 1502 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1503 uvchr_to_utf8(tmpbuf, c);
922e8cb4 1504 return _is_utf8_FOO(_CC_WORDCHAR, tmpbuf);
a0ed51b3
LW
1505}
1506
1507bool
5092f92a
KW
1508Perl_is_uni_alnumc(pTHX_ UV c)
1509{
1510 U8 tmpbuf[UTF8_MAXBYTES+1];
1511 uvchr_to_utf8(tmpbuf, c);
922e8cb4 1512 return _is_utf8_FOO(_CC_ALPHANUMERIC, tmpbuf);
5092f92a
KW
1513}
1514
f9ae8fb6
JD
1515/* Internal function so we can deprecate the external one, and call
1516 this one from other deprecated functions in this file */
1517
1518PERL_STATIC_INLINE bool
61b19385
KW
1519S_is_utf8_idfirst(pTHX_ const U8 *p)
1520{
1521 dVAR;
1522
1523 if (*p == '_')
1524 return TRUE;
1525 /* is_utf8_idstart would be more logical. */
1526 return is_utf8_common(p, &PL_utf8_idstart, "IdStart");
1527}
1528
5092f92a 1529bool
84afefe6 1530Perl_is_uni_idfirst(pTHX_ UV c)
a0ed51b3 1531{
89ebb4a3 1532 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1533 uvchr_to_utf8(tmpbuf, c);
61b19385 1534 return S_is_utf8_idfirst(aTHX_ tmpbuf);
a0ed51b3
LW
1535}
1536
1537bool
eba68aa0
KW
1538Perl__is_uni_perl_idcont(pTHX_ UV c)
1539{
1540 U8 tmpbuf[UTF8_MAXBYTES+1];
1541 uvchr_to_utf8(tmpbuf, c);
1542 return _is_utf8_perl_idcont(tmpbuf);
1543}
1544
1545bool
f91dcd13
KW
1546Perl__is_uni_perl_idstart(pTHX_ UV c)
1547{
1548 U8 tmpbuf[UTF8_MAXBYTES+1];
1549 uvchr_to_utf8(tmpbuf, c);
1550 return _is_utf8_perl_idstart(tmpbuf);
1551}
1552
1553bool
84afefe6 1554Perl_is_uni_alpha(pTHX_ UV c)
a0ed51b3 1555{
89ebb4a3 1556 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1557 uvchr_to_utf8(tmpbuf, c);
922e8cb4 1558 return _is_utf8_FOO(_CC_ALPHA, tmpbuf);
a0ed51b3
LW
1559}
1560
1561bool
84afefe6 1562Perl_is_uni_ascii(pTHX_ UV c)
4d61ec05 1563{
bc39fe24 1564 return isASCII(c);
4d61ec05
GS
1565}
1566
1567bool
bdd8600f
KW
1568Perl_is_uni_blank(pTHX_ UV c)
1569{
2cafb56b 1570 return isBLANK_uni(c);
bdd8600f
KW
1571}
1572
1573bool
84afefe6 1574Perl_is_uni_space(pTHX_ UV c)
a0ed51b3 1575{
add4123a 1576 return isSPACE_uni(c);
a0ed51b3
LW
1577}
1578
1579bool
84afefe6 1580Perl_is_uni_digit(pTHX_ UV c)
a0ed51b3 1581{
89ebb4a3 1582 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1583 uvchr_to_utf8(tmpbuf, c);
922e8cb4 1584 return _is_utf8_FOO(_CC_DIGIT, tmpbuf);
a0ed51b3
LW
1585}
1586
1587bool
84afefe6 1588Perl_is_uni_upper(pTHX_ UV c)
a0ed51b3 1589{
89ebb4a3 1590 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1591 uvchr_to_utf8(tmpbuf, c);
922e8cb4 1592 return _is_utf8_FOO(_CC_UPPER, tmpbuf);
a0ed51b3
LW
1593}
1594
1595bool
84afefe6 1596Perl_is_uni_lower(pTHX_ UV c)
a0ed51b3 1597{
89ebb4a3 1598 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1599 uvchr_to_utf8(tmpbuf, c);
922e8cb4 1600 return _is_utf8_FOO(_CC_LOWER, tmpbuf);
a0ed51b3
LW
1601}
1602
1603bool
84afefe6 1604Perl_is_uni_cntrl(pTHX_ UV c)
b8c5462f 1605{
7b952154 1606 return isCNTRL_L1(c);
b8c5462f
JH
1607}
1608
1609bool
84afefe6 1610Perl_is_uni_graph(pTHX_ UV c)
b8c5462f 1611{
89ebb4a3 1612 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1613 uvchr_to_utf8(tmpbuf, c);
922e8cb4 1614 return _is_utf8_FOO(_CC_GRAPH, tmpbuf);
b8c5462f
JH
1615}
1616
1617bool
84afefe6 1618Perl_is_uni_print(pTHX_ UV c)
a0ed51b3 1619{
89ebb4a3 1620 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1621 uvchr_to_utf8(tmpbuf, c);
922e8cb4 1622 return _is_utf8_FOO(_CC_PRINT, tmpbuf);
a0ed51b3
LW
1623}
1624
b8c5462f 1625bool
84afefe6 1626Perl_is_uni_punct(pTHX_ UV c)
b8c5462f 1627{
89ebb4a3 1628 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1629 uvchr_to_utf8(tmpbuf, c);
922e8cb4 1630 return _is_utf8_FOO(_CC_PUNCT, tmpbuf);
b8c5462f
JH
1631}
1632
4d61ec05 1633bool
84afefe6 1634Perl_is_uni_xdigit(pTHX_ UV c)
4d61ec05 1635{
4ac6419d 1636 return isXDIGIT_uni(c);
4d61ec05
GS
1637}
1638
3a4c58c9
KW
1639UV
1640Perl__to_upper_title_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp, const char S_or_s)
1641{
1642 /* We have the latin1-range values compiled into the core, so just use
1643 * those, converting the result to utf8. The only difference between upper
1644 * and title case in this range is that LATIN_SMALL_LETTER_SHARP_S is
1645 * either "SS" or "Ss". Which one to use is passed into the routine in
1646 * 'S_or_s' to avoid a test */
1647
1648 UV converted = toUPPER_LATIN1_MOD(c);
1649
1650 PERL_ARGS_ASSERT__TO_UPPER_TITLE_LATIN1;
1651
1652 assert(S_or_s == 'S' || S_or_s == 's');
1653
1654 if (UNI_IS_INVARIANT(converted)) { /* No difference between the two for
1655 characters in this range */
1656 *p = (U8) converted;
1657 *lenp = 1;
1658 return converted;
1659 }
1660
1661 /* toUPPER_LATIN1_MOD gives the correct results except for three outliers,
1662 * which it maps to one of them, so as to only have to have one check for
1663 * it in the main case */
1664 if (UNLIKELY(converted == LATIN_SMALL_LETTER_Y_WITH_DIAERESIS)) {
1665 switch (c) {
1666 case LATIN_SMALL_LETTER_Y_WITH_DIAERESIS:
1667 converted = LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS;
1668 break;
1669 case MICRO_SIGN:
1670 converted = GREEK_CAPITAL_LETTER_MU;
1671 break;
1672 case LATIN_SMALL_LETTER_SHARP_S:
1673 *(p)++ = 'S';
1674 *p = S_or_s;
1675 *lenp = 2;
1676 return 'S';
1677 default:
1678 Perl_croak(aTHX_ "panic: to_upper_title_latin1 did not expect '%c' to map to '%c'", c, LATIN_SMALL_LETTER_Y_WITH_DIAERESIS);
118e2215 1679 assert(0); /* NOTREACHED */
3a4c58c9
KW
1680 }
1681 }
1682
1683 *(p)++ = UTF8_TWO_BYTE_HI(converted);
1684 *p = UTF8_TWO_BYTE_LO(converted);
1685 *lenp = 2;
1686
1687 return converted;
1688}
1689
50bda2c3
KW
1690/* Call the function to convert a UTF-8 encoded character to the specified case.
1691 * Note that there may be more than one character in the result.
1692 * INP is a pointer to the first byte of the input character
1693 * OUTP will be set to the first byte of the string of changed characters. It
1694 * needs to have space for UTF8_MAXBYTES_CASE+1 bytes
1695 * LENP will be set to the length in bytes of the string of changed characters
1696 *
1697 * The functions return the ordinal of the first character in the string of OUTP */
f90a9a02
KW
1698#define CALL_UPPER_CASE(INP, OUTP, LENP) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_toupper, "ToUc", "utf8::ToSpecUc")
1699#define CALL_TITLE_CASE(INP, OUTP, LENP) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_totitle, "ToTc", "utf8::ToSpecTc")
1700#define CALL_LOWER_CASE(INP, OUTP, LENP) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_tolower, "ToLc", "utf8::ToSpecLc")
50bda2c3
KW
1701
1702/* This additionally has the input parameter SPECIALS, which if non-zero will
1703 * cause this to use the SPECIALS hash for folding (meaning get full case
1704 * folding); otherwise, when zero, this implies a simple case fold */
f90a9a02 1705#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 1706
84afefe6
JH
1707UV
1708Perl_to_uni_upper(pTHX_ UV c, U8* p, STRLEN *lenp)
a0ed51b3 1709{
3a4c58c9
KW
1710 dVAR;
1711
a1433954
KW
1712 /* Convert the Unicode character whose ordinal is <c> to its uppercase
1713 * version and store that in UTF-8 in <p> and its length in bytes in <lenp>.
1714 * Note that the <p> needs to be at least UTF8_MAXBYTES_CASE+1 bytes since
c3fd2246
KW
1715 * the changed version may be longer than the original character.
1716 *
1717 * The ordinal of the first character of the changed version is returned
1718 * (but note, as explained above, that there may be more.) */
1719
7918f24d
NC
1720 PERL_ARGS_ASSERT_TO_UNI_UPPER;
1721
3a4c58c9
KW
1722 if (c < 256) {
1723 return _to_upper_title_latin1((U8) c, p, lenp, 'S');
1724 }
1725
0ebc6274 1726 uvchr_to_utf8(p, c);
3a4c58c9 1727 return CALL_UPPER_CASE(p, p, lenp);
a0ed51b3
LW
1728}
1729
84afefe6
JH
1730UV
1731Perl_to_uni_title(pTHX_ UV c, U8* p, STRLEN *lenp)
a0ed51b3 1732{
3a4c58c9
KW
1733 dVAR;
1734
7918f24d
NC
1735 PERL_ARGS_ASSERT_TO_UNI_TITLE;
1736
3a4c58c9
KW
1737 if (c < 256) {
1738 return _to_upper_title_latin1((U8) c, p, lenp, 's');
1739 }
1740
0ebc6274 1741 uvchr_to_utf8(p, c);
3a4c58c9 1742 return CALL_TITLE_CASE(p, p, lenp);
a0ed51b3
LW
1743}
1744
afc16117
KW
1745STATIC U8
1746S_to_lower_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp)
1747{
1748 /* We have the latin1-range values compiled into the core, so just use
1749 * those, converting the result to utf8. Since the result is always just
a1433954 1750 * one character, we allow <p> to be NULL */
afc16117
KW
1751
1752 U8 converted = toLOWER_LATIN1(c);
1753
1754 if (p != NULL) {
1755 if (UNI_IS_INVARIANT(converted)) {
1756 *p = converted;
1757 *lenp = 1;
1758 }
1759 else {
1760 *p = UTF8_TWO_BYTE_HI(converted);
1761 *(p+1) = UTF8_TWO_BYTE_LO(converted);
1762 *lenp = 2;
1763 }
1764 }
1765 return converted;
1766}
1767
84afefe6
JH
1768UV
1769Perl_to_uni_lower(pTHX_ UV c, U8* p, STRLEN *lenp)
a0ed51b3 1770{
968c5e6a
KW
1771 dVAR;
1772
7918f24d
NC
1773 PERL_ARGS_ASSERT_TO_UNI_LOWER;
1774
afc16117
KW
1775 if (c < 256) {
1776 return to_lower_latin1((U8) c, p, lenp);
bca00c02
KW
1777 }
1778
afc16117 1779 uvchr_to_utf8(p, c);
968c5e6a 1780 return CALL_LOWER_CASE(p, p, lenp);
a0ed51b3
LW
1781}
1782
84afefe6 1783UV
f673fad4 1784Perl__to_fold_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp, const bool flags)
a1dde8de 1785{
a1433954 1786 /* Corresponds to to_lower_latin1(), <flags> is TRUE if to use full case
f673fad4
KW
1787 * folding */
1788
a1dde8de
KW
1789 UV converted;
1790
1791 PERL_ARGS_ASSERT__TO_FOLD_LATIN1;
1792
1793 if (c == MICRO_SIGN) {
1794 converted = GREEK_SMALL_LETTER_MU;
1795 }
1796 else if (flags && c == LATIN_SMALL_LETTER_SHARP_S) {
1797 *(p)++ = 's';
1798 *p = 's';
1799 *lenp = 2;
1800 return 's';
1801 }
1802 else { /* In this range the fold of all other characters is their lower
1803 case */
1804 converted = toLOWER_LATIN1(c);
1805 }
1806
1807 if (UNI_IS_INVARIANT(converted)) {
1808 *p = (U8) converted;
1809 *lenp = 1;
1810 }
1811 else {
1812 *(p)++ = UTF8_TWO_BYTE_HI(converted);
1813 *p = UTF8_TWO_BYTE_LO(converted);
1814 *lenp = 2;
1815 }
1816
1817 return converted;
1818}
1819
1820UV
a0270393 1821Perl__to_uni_fold_flags(pTHX_ UV c, U8* p, STRLEN *lenp, const U8 flags)
84afefe6 1822{
4b593389 1823
a0270393
KW
1824 /* Not currently externally documented, and subject to change
1825 * <flags> bits meanings:
1826 * FOLD_FLAGS_FULL iff full folding is to be used;
1827 * FOLD_FLAGS_LOCALE iff in locale
1828 * FOLD_FLAGS_NOMIX_ASCII iff non-ASCII to ASCII folds are prohibited
1829 */
4b593389 1830
36bb2ab6 1831 PERL_ARGS_ASSERT__TO_UNI_FOLD_FLAGS;
7918f24d 1832
a1dde8de 1833 if (c < 256) {
a0270393
KW
1834 UV result = _to_fold_latin1((U8) c, p, lenp,
1835 cBOOL(((flags & FOLD_FLAGS_FULL)
1836 /* If ASCII-safe, don't allow full folding,
1837 * as that could include SHARP S => ss;
1838 * otherwise there is no crossing of
1839 * ascii/non-ascii in the latin1 range */
1840 && ! (flags & FOLD_FLAGS_NOMIX_ASCII))));
1841 /* It is illegal for the fold to cross the 255/256 boundary under
1842 * locale; in this case return the original */
1843 return (result > 256 && flags & FOLD_FLAGS_LOCALE)
1844 ? c
1845 : result;
a1dde8de
KW
1846 }
1847
a0270393
KW
1848 /* If no special needs, just use the macro */
1849 if ( ! (flags & (FOLD_FLAGS_LOCALE|FOLD_FLAGS_NOMIX_ASCII))) {
1850 uvchr_to_utf8(p, c);
1851 return CALL_FOLD_CASE(p, p, lenp, flags & FOLD_FLAGS_FULL);
1852 }
1853 else { /* Otherwise, _to_utf8_fold_flags has the intelligence to deal with
1854 the special flags. */
1855 U8 utf8_c[UTF8_MAXBYTES + 1];
1856 uvchr_to_utf8(utf8_c, c);
1857 return _to_utf8_fold_flags(utf8_c, p, lenp, flags, NULL);
1858 }
84afefe6
JH
1859}
1860
a0ed51b3 1861bool
84afefe6 1862Perl_is_uni_alnum_lc(pTHX_ UV c)
a0ed51b3 1863{
edfb3318
KW
1864 if (c < 256) {
1865 return isALNUM_LC(UNI_TO_NATIVE(c));
1866 }
922e8cb4 1867 return _is_uni_FOO(_CC_WORDCHAR, c);
a0ed51b3
LW
1868}
1869
1870bool
5092f92a
KW
1871Perl_is_uni_alnumc_lc(pTHX_ UV c)
1872{
1873 if (c < 256) {
15861f94 1874 return isALPHANUMERIC_LC(UNI_TO_NATIVE(c));
5092f92a 1875 }
922e8cb4 1876 return _is_uni_FOO(_CC_ALPHANUMERIC, c);
5092f92a
KW
1877}
1878
1879bool
84afefe6 1880Perl_is_uni_idfirst_lc(pTHX_ UV c)
a0ed51b3 1881{
edfb3318
KW
1882 if (c < 256) {
1883 return isIDFIRST_LC(UNI_TO_NATIVE(c));
1884 }
1885 return _is_uni_perl_idstart(c);
a0ed51b3
LW
1886}
1887
1888bool
84afefe6 1889Perl_is_uni_alpha_lc(pTHX_ UV c)
a0ed51b3 1890{
edfb3318
KW
1891 if (c < 256) {
1892 return isALPHA_LC(UNI_TO_NATIVE(c));
1893 }
922e8cb4 1894 return _is_uni_FOO(_CC_ALPHA, c);
a0ed51b3
LW
1895}
1896
1897bool
84afefe6 1898Perl_is_uni_ascii_lc(pTHX_ UV c)
4d61ec05 1899{
edfb3318
KW
1900 if (c < 256) {
1901 return isASCII_LC(UNI_TO_NATIVE(c));
1902 }
1903 return 0;
4d61ec05
GS
1904}
1905
1906bool
bdd8600f
KW
1907Perl_is_uni_blank_lc(pTHX_ UV c)
1908{
edfb3318
KW
1909 if (c < 256) {
1910 return isBLANK_LC(UNI_TO_NATIVE(c));
1911 }
61b19385 1912 return isBLANK_uni(c);
bdd8600f
KW
1913}
1914
1915bool
84afefe6 1916Perl_is_uni_space_lc(pTHX_ UV c)
a0ed51b3 1917{
edfb3318
KW
1918 if (c < 256) {
1919 return isSPACE_LC(UNI_TO_NATIVE(c));
1920 }
61b19385 1921 return isSPACE_uni(c);
a0ed51b3
LW
1922}
1923
1924bool
84afefe6 1925Perl_is_uni_digit_lc(pTHX_ UV c)
a0ed51b3 1926{
edfb3318
KW
1927 if (c < 256) {
1928 return isDIGIT_LC(UNI_TO_NATIVE(c));
1929 }
922e8cb4 1930 return _is_uni_FOO(_CC_DIGIT, c);
a0ed51b3
LW
1931}
1932
1933bool
84afefe6 1934Perl_is_uni_upper_lc(pTHX_ UV c)
a0ed51b3 1935{
edfb3318
KW
1936 if (c < 256) {
1937 return isUPPER_LC(UNI_TO_NATIVE(c));
1938 }
922e8cb4 1939 return _is_uni_FOO(_CC_UPPER, c);
a0ed51b3
LW
1940}
1941
1942bool
84afefe6 1943Perl_is_uni_lower_lc(pTHX_ UV c)
a0ed51b3 1944{
edfb3318
KW
1945 if (c < 256) {
1946 return isLOWER_LC(UNI_TO_NATIVE(c));
1947 }
922e8cb4 1948 return _is_uni_FOO(_CC_LOWER, c);
a0ed51b3
LW
1949}
1950
1951bool
84afefe6 1952Perl_is_uni_cntrl_lc(pTHX_ UV c)
b8c5462f 1953{
edfb3318
KW
1954 if (c < 256) {
1955 return isCNTRL_LC(UNI_TO_NATIVE(c));
1956 }
61b19385 1957 return 0;
b8c5462f
JH
1958}
1959
1960bool
84afefe6 1961Perl_is_uni_graph_lc(pTHX_ UV c)
b8c5462f 1962{
edfb3318
KW
1963 if (c < 256) {
1964 return isGRAPH_LC(UNI_TO_NATIVE(c));
1965 }
922e8cb4 1966 return _is_uni_FOO(_CC_GRAPH, c);
b8c5462f
JH
1967}
1968
1969bool
84afefe6 1970Perl_is_uni_print_lc(pTHX_ UV c)
a0ed51b3 1971{
edfb3318
KW
1972 if (c < 256) {
1973 return isPRINT_LC(UNI_TO_NATIVE(c));
1974 }
922e8cb4 1975 return _is_uni_FOO(_CC_PRINT, c);
a0ed51b3
LW
1976}
1977
b8c5462f 1978bool
84afefe6 1979Perl_is_uni_punct_lc(pTHX_ UV c)
b8c5462f 1980{
edfb3318
KW
1981 if (c < 256) {
1982 return isPUNCT_LC(UNI_TO_NATIVE(c));
1983 }
922e8cb4 1984 return _is_uni_FOO(_CC_PUNCT, c);
b8c5462f
JH
1985}
1986
4d61ec05 1987bool
84afefe6 1988Perl_is_uni_xdigit_lc(pTHX_ UV c)
4d61ec05 1989{
edfb3318
KW
1990 if (c < 256) {
1991 return isXDIGIT_LC(UNI_TO_NATIVE(c));
1992 }
61b19385 1993 return isXDIGIT_uni(c);
4d61ec05
GS
1994}
1995
b7ac61fa
JH
1996U32
1997Perl_to_uni_upper_lc(pTHX_ U32 c)
1998{
ee099d14
JH
1999 /* XXX returns only the first character -- do not use XXX */
2000 /* XXX no locale support yet */
2001 STRLEN len;
89ebb4a3 2002 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
ee099d14 2003 return (U32)to_uni_upper(c, tmpbuf, &len);
b7ac61fa
JH
2004}
2005
2006U32
2007Perl_to_uni_title_lc(pTHX_ U32 c)
2008{
ee099d14
JH
2009 /* XXX returns only the first character XXX -- do not use XXX */
2010 /* XXX no locale support yet */
2011 STRLEN len;
89ebb4a3 2012 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
ee099d14 2013 return (U32)to_uni_title(c, tmpbuf, &len);
b7ac61fa
JH
2014}
2015
2016U32
2017Perl_to_uni_lower_lc(pTHX_ U32 c)
2018{
ee099d14
JH
2019 /* XXX returns only the first character -- do not use XXX */
2020 /* XXX no locale support yet */
2021 STRLEN len;
89ebb4a3 2022 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
ee099d14 2023 return (U32)to_uni_lower(c, tmpbuf, &len);
b7ac61fa
JH
2024}
2025
26483009 2026PERL_STATIC_INLINE bool
5141f98e 2027S_is_utf8_common(pTHX_ const U8 *const p, SV **swash,
bde6a22d
NC
2028 const char *const swashname)
2029{
ea317ccb
KW
2030 /* returns a boolean giving whether or not the UTF8-encoded character that
2031 * starts at <p> is in the swash indicated by <swashname>. <swash>
2032 * contains a pointer to where the swash indicated by <swashname>
2033 * is to be stored; which this routine will do, so that future calls will
2034 * look at <*swash> and only generate a swash if it is not null
2035 *
2036 * Note that it is assumed that the buffer length of <p> is enough to
2037 * contain all the bytes that comprise the character. Thus, <*p> should
2038 * have been checked before this call for mal-formedness enough to assure
2039 * that. */
2040
97aff369 2041 dVAR;
7918f24d
NC
2042
2043 PERL_ARGS_ASSERT_IS_UTF8_COMMON;
2044
492a624f 2045 /* The API should have included a length for the UTF-8 character in <p>,
28123549 2046 * but it doesn't. We therefore assume that p has been validated at least
492a624f
KW
2047 * as far as there being enough bytes available in it to accommodate the
2048 * character without reading beyond the end, and pass that number on to the
2049 * validating routine */
28123549
KW
2050 if (! is_utf8_char_buf(p, p + UTF8SKIP(p))) {
2051 if (ckWARN_d(WARN_UTF8)) {
2052 Perl_warner(aTHX_ packWARN2(WARN_DEPRECATED,WARN_UTF8),
9816f121 2053 "Passing malformed UTF-8 to \"%s\" is deprecated", swashname);
28123549
KW
2054 if (ckWARN(WARN_UTF8)) { /* This will output details as to the
2055 what the malformation is */
2056 utf8_to_uvchr_buf(p, p + UTF8SKIP(p), NULL);
2057 }
2058 }
2059 return FALSE;
2060 }
87367d5f
KW
2061 if (!*swash) {
2062 U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST;
2063 *swash = _core_swash_init("utf8", swashname, &PL_sv_undef, 1, 0, NULL, &flags);
2064 }
28123549 2065
bde6a22d
NC
2066 return swash_fetch(*swash, p, TRUE) != 0;
2067}
2068
2069bool
922e8cb4
KW
2070Perl__is_utf8_FOO(pTHX_ const U8 classnum, const U8 *p)
2071{
2072 dVAR;
2073
2074 PERL_ARGS_ASSERT__IS_UTF8_FOO;
2075
2076 assert(classnum < _FIRST_NON_SWASH_CC);
2077
2078 return is_utf8_common(p, &PL_utf8_swash_ptrs[classnum], swash_property_names[classnum]);
2079}
2080
2081bool
7fc63493 2082Perl_is_utf8_alnum(pTHX_ const U8 *p)
a0ed51b3 2083{
97aff369 2084 dVAR;
7918f24d
NC
2085
2086 PERL_ARGS_ASSERT_IS_UTF8_ALNUM;
2087
671c33bf
NC
2088 /* NOTE: "IsWord", not "IsAlnum", since Alnum is a true
2089 * descendant of isalnum(3), in other words, it doesn't
2090 * contain the '_'. --jhi */
03940dc2 2091 return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_WORDCHAR], "IsWord");
a0ed51b3
LW
2092}
2093
2094bool
5092f92a
KW
2095Perl_is_utf8_alnumc(pTHX_ const U8 *p)
2096{
2097 dVAR;
2098
2099 PERL_ARGS_ASSERT_IS_UTF8_ALNUMC;
2100
03940dc2 2101 return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_ALPHANUMERIC], "IsAlnum");
5092f92a
KW
2102}
2103
2104bool
7fc63493 2105Perl_is_utf8_idfirst(pTHX_ const U8 *p) /* The naming is historical. */
a0ed51b3 2106{
97aff369 2107 dVAR;
7918f24d
NC
2108
2109 PERL_ARGS_ASSERT_IS_UTF8_IDFIRST;
2110
61b19385 2111 return S_is_utf8_idfirst(aTHX_ p);
82686b01
JH
2112}
2113
2114bool
c11ff943
KW
2115Perl_is_utf8_xidfirst(pTHX_ const U8 *p) /* The naming is historical. */
2116{
2117 dVAR;
2118
2119 PERL_ARGS_ASSERT_IS_UTF8_XIDFIRST;
2120
2121 if (*p == '_')
2122 return TRUE;
2123 /* is_utf8_idstart would be more logical. */
2124 return is_utf8_common(p, &PL_utf8_xidstart, "XIdStart");
2125}
2126
2127bool
d65654cb 2128Perl__is_utf8_perl_idstart(pTHX_ const U8 *p)
b6912c02
KW
2129{
2130 dVAR;
2131
d65654cb 2132 PERL_ARGS_ASSERT__IS_UTF8_PERL_IDSTART;
b6912c02
KW
2133
2134 return is_utf8_common(p, &PL_utf8_perl_idstart, "_Perl_IDStart");
2135}
2136
2137bool
eba68aa0
KW
2138Perl__is_utf8_perl_idcont(pTHX_ const U8 *p)
2139{
2140 dVAR;
2141
2142 PERL_ARGS_ASSERT__IS_UTF8_PERL_IDCONT;
2143
2144 return is_utf8_common(p, &PL_utf8_perl_idcont, "_Perl_IDCont");
2145}
2146
2147
2148bool
7fc63493 2149Perl_is_utf8_idcont(pTHX_ const U8 *p)
82686b01 2150{
97aff369 2151 dVAR;
7918f24d
NC
2152
2153 PERL_ARGS_ASSERT_IS_UTF8_IDCONT;
2154
d4c19fe8 2155 return is_utf8_common(p, &PL_utf8_idcont, "IdContinue");
a0ed51b3
LW
2156}
2157
2158bool
c11ff943
KW
2159Perl_is_utf8_xidcont(pTHX_ const U8 *p)
2160{
2161 dVAR;
2162
2163 PERL_ARGS_ASSERT_IS_UTF8_XIDCONT;
2164
c11ff943
KW
2165 return is_utf8_common(p, &PL_utf8_idcont, "XIdContinue");
2166}
2167
2168bool
7fc63493 2169Perl_is_utf8_alpha(pTHX_ const U8 *p)
a0ed51b3 2170{
97aff369 2171 dVAR;
7918f24d
NC
2172
2173 PERL_ARGS_ASSERT_IS_UTF8_ALPHA;
2174
03940dc2 2175 return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_ALPHA], "IsAlpha");
a0ed51b3
LW
2176}
2177
2178bool
7fc63493 2179Perl_is_utf8_ascii(pTHX_ const U8 *p)
b8c5462f 2180{
97aff369 2181 dVAR;
7918f24d
NC
2182
2183 PERL_ARGS_ASSERT_IS_UTF8_ASCII;
2184
bc39fe24
KW
2185 /* ASCII characters are the same whether in utf8 or not. So the macro
2186 * works on both utf8 and non-utf8 representations. */
2187 return isASCII(*p);
b8c5462f
JH
2188}
2189
2190bool
bdd8600f
KW
2191Perl_is_utf8_blank(pTHX_ const U8 *p)
2192{
2193 dVAR;
2194
2195 PERL_ARGS_ASSERT_IS_UTF8_BLANK;
2196
2cafb56b 2197 return isBLANK_utf8(p);
bdd8600f
KW
2198}
2199
2200bool
7fc63493 2201Perl_is_utf8_space(pTHX_ const U8 *p)
a0ed51b3 2202{
97aff369 2203 dVAR;
7918f24d
NC
2204
2205 PERL_ARGS_ASSERT_IS_UTF8_SPACE;
2206
add4123a 2207 return isSPACE_utf8(p);
a0ed51b3
LW
2208}
2209
2210bool
d1eb3177
YO
2211Perl_is_utf8_perl_space(pTHX_ const U8 *p)
2212{
2213 dVAR;
2214
2215 PERL_ARGS_ASSERT_IS_UTF8_PERL_SPACE;
2216
c4428693
KW
2217 /* Only true if is an ASCII space-like character, and ASCII is invariant
2218 * under utf8, so can just use the macro */
2219 return isSPACE_A(*p);
d1eb3177
YO
2220}
2221
2222bool
2223Perl_is_utf8_perl_word(pTHX_ const U8 *p)
2224{
2225 dVAR;
2226
2227 PERL_ARGS_ASSERT_IS_UTF8_PERL_WORD;
2228
c4428693
KW
2229 /* Only true if is an ASCII word character, and ASCII is invariant
2230 * under utf8, so can just use the macro */
2231 return isWORDCHAR_A(*p);
d1eb3177
YO
2232}
2233
2234bool
7fc63493 2235Perl_is_utf8_digit(pTHX_ const U8 *p)
a0ed51b3 2236{
97aff369 2237 dVAR;
7918f24d
NC
2238
2239 PERL_ARGS_ASSERT_IS_UTF8_DIGIT;
2240
03940dc2 2241 return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_DIGIT], "IsDigit");
a0ed51b3
LW
2242}
2243
2244bool
d1eb3177
YO
2245Perl_is_utf8_posix_digit(pTHX_ const U8 *p)
2246{
2247 dVAR;
2248
2249 PERL_ARGS_ASSERT_IS_UTF8_POSIX_DIGIT;
2250
c4428693
KW
2251 /* Only true if is an ASCII digit character, and ASCII is invariant
2252 * under utf8, so can just use the macro */
2253 return isDIGIT_A(*p);
d1eb3177
YO
2254}
2255
2256bool
7fc63493 2257Perl_is_utf8_upper(pTHX_ const U8 *p)
a0ed51b3 2258{
97aff369 2259 dVAR;
7918f24d
NC
2260
2261 PERL_ARGS_ASSERT_IS_UTF8_UPPER;
2262
03940dc2 2263 return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_UPPER], "IsUppercase");
a0ed51b3
LW
2264}
2265
2266bool
7fc63493 2267Perl_is_utf8_lower(pTHX_ const U8 *p)
a0ed51b3 2268{
97aff369 2269 dVAR;
7918f24d
NC
2270
2271 PERL_ARGS_ASSERT_IS_UTF8_LOWER;
2272
03940dc2 2273 return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_LOWER], "IsLowercase");
a0ed51b3
LW
2274}
2275
2276bool
7fc63493 2277Perl_is_utf8_cntrl(pTHX_ const U8 *p)
b8c5462f 2278{
97aff369 2279 dVAR;
7918f24d
NC
2280
2281 PERL_ARGS_ASSERT_IS_UTF8_CNTRL;
2282
a35d759a 2283 return isCNTRL_utf8(p);
b8c5462f
JH
2284}
2285
2286bool
7fc63493 2287Perl_is_utf8_graph(pTHX_ const U8 *p)
b8c5462f 2288{
97aff369 2289 dVAR;
7918f24d
NC
2290
2291 PERL_ARGS_ASSERT_IS_UTF8_GRAPH;
2292
03940dc2 2293 return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_GRAPH], "IsGraph");
b8c5462f
JH
2294}
2295
2296bool
7fc63493 2297Perl_is_utf8_print(pTHX_ const U8 *p)
a0ed51b3 2298{
97aff369 2299 dVAR;
7918f24d
NC
2300
2301 PERL_ARGS_ASSERT_IS_UTF8_PRINT;
2302
03940dc2 2303 return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_PRINT], "IsPrint");
a0ed51b3
LW
2304}
2305
2306bool
7fc63493 2307Perl_is_utf8_punct(pTHX_ const U8 *p)
b8c5462f 2308{
97aff369 2309 dVAR;
7918f24d
NC
2310
2311 PERL_ARGS_ASSERT_IS_UTF8_PUNCT;
2312
03940dc2 2313 return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_PUNCT], "IsPunct");
b8c5462f
JH
2314}
2315
2316bool
7fc63493 2317Perl_is_utf8_xdigit(pTHX_ const U8 *p)
b8c5462f 2318{
97aff369 2319 dVAR;
7918f24d
NC
2320
2321 PERL_ARGS_ASSERT_IS_UTF8_XDIGIT;
2322
4ac6419d 2323 return is_XDIGIT_utf8(p);
b8c5462f
JH
2324}
2325
2326bool
7dbf68d2
KW
2327Perl__is_utf8_mark(pTHX_ const U8 *p)
2328{
2329 dVAR;
2330
2331 PERL_ARGS_ASSERT__IS_UTF8_MARK;
2332
2333 return is_utf8_common(p, &PL_utf8_mark, "IsM");
2334}
2335
2336
2337bool
7fc63493 2338Perl_is_utf8_mark(pTHX_ const U8 *p)
a0ed51b3 2339{
97aff369 2340 dVAR;
7918f24d
NC
2341
2342 PERL_ARGS_ASSERT_IS_UTF8_MARK;
2343
7dbf68d2 2344 return _is_utf8_mark(p);
a0ed51b3
LW
2345}
2346
6b5c0936 2347/*
87cea99e 2348=for apidoc to_utf8_case
6b5c0936 2349
a1433954
KW
2350The C<p> contains the pointer to the UTF-8 string encoding
2351the character that is being converted. This routine assumes that the character
2352at C<p> is well-formed.
6b5c0936 2353
a1433954
KW
2354The C<ustrp> is a pointer to the character buffer to put the
2355conversion result to. The C<lenp> is a pointer to the length
6b5c0936
JH
2356of the result.
2357
a1433954 2358The C<swashp> is a pointer to the swash to use.
6b5c0936 2359
a1433954
KW
2360Both the special and normal mappings are stored in F<lib/unicore/To/Foo.pl>,
2361and loaded by SWASHNEW, using F<lib/utf8_heavy.pl>. The C<special> (usually,
0134edef 2362but not always, a multicharacter mapping), is tried first.
6b5c0936 2363
a1433954 2364The C<special> is a string like "utf8::ToSpecLower", which means the
0134edef
JH
2365hash %utf8::ToSpecLower. The access to the hash is through
2366Perl_to_utf8_case().
6b5c0936 2367
a1433954 2368The C<normal> is a string like "ToLower" which means the swash
0134edef
JH
2369%utf8::ToLower.
2370
2371=cut */
6b5c0936 2372
2104c8d9 2373UV
9a957fbc
AL
2374Perl_to_utf8_case(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp,
2375 SV **swashp, const char *normal, const char *special)
a0ed51b3 2376{
97aff369 2377 dVAR;
89ebb4a3 2378 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
0134edef 2379 STRLEN len = 0;
4b88fb76 2380 const UV uv0 = valid_utf8_to_uvchr(p, NULL);
1feea2c7
JH
2381 /* The NATIVE_TO_UNI() and UNI_TO_NATIVE() mappings
2382 * are necessary in EBCDIC, they are redundant no-ops
2383 * in ASCII-ish platforms, and hopefully optimized away. */
f54cb97a 2384 const UV uv1 = NATIVE_TO_UNI(uv0);
7918f24d
NC
2385
2386 PERL_ARGS_ASSERT_TO_UTF8_CASE;
2387
9ae3ac1a
KW
2388 /* Note that swash_fetch() doesn't output warnings for these because it
2389 * assumes we will */
8457b38f 2390 if (uv1 >= UNICODE_SURROGATE_FIRST) {
9ae3ac1a 2391 if (uv1 <= UNICODE_SURROGATE_LAST) {
8457b38f
KW
2392 if (ckWARN_d(WARN_SURROGATE)) {
2393 const char* desc = (PL_op) ? OP_DESC(PL_op) : normal;
2394 Perl_warner(aTHX_ packWARN(WARN_SURROGATE),
2395 "Operation \"%s\" returns its argument for UTF-16 surrogate U+%04"UVXf"", desc, uv1);
2396 }
9ae3ac1a
KW
2397 }
2398 else if (UNICODE_IS_SUPER(uv1)) {
8457b38f
KW
2399 if (ckWARN_d(WARN_NON_UNICODE)) {
2400 const char* desc = (PL_op) ? OP_DESC(PL_op) : normal;
2401 Perl_warner(aTHX_ packWARN(WARN_NON_UNICODE),
2402 "Operation \"%s\" returns its argument for non-Unicode code point 0x%04"UVXf"", desc, uv1);
2403 }
9ae3ac1a
KW
2404 }
2405
2406 /* Note that non-characters are perfectly legal, so no warning should
2407 * be given */
2408 }
2409
1feea2c7 2410 uvuni_to_utf8(tmpbuf, uv1);
0134edef
JH
2411
2412 if (!*swashp) /* load on-demand */
5ab9d2ef 2413 *swashp = _core_swash_init("utf8", normal, &PL_sv_undef, 4, 0, NULL, NULL);
0134edef 2414
a6f87d8c 2415 if (special) {
0134edef 2416 /* It might be "special" (sometimes, but not always,
2a37f04d 2417 * a multicharacter mapping) */
6673a63c 2418 HV * const hv = get_hv(special, 0);
b08cf34e
JH
2419 SV **svp;
2420
35da51f7 2421 if (hv &&
b08cf34e
JH
2422 (svp = hv_fetch(hv, (const char*)tmpbuf, UNISKIP(uv1), FALSE)) &&
2423 (*svp)) {
cfd0369c 2424 const char *s;
47654450 2425
cfd0369c 2426 s = SvPV_const(*svp, len);
47654450
JH
2427 if (len == 1)
2428 len = uvuni_to_utf8(ustrp, NATIVE_TO_UNI(*(U8*)s)) - ustrp;
2a37f04d 2429 else {
2f9475ad
JH
2430#ifdef EBCDIC
2431 /* If we have EBCDIC we need to remap the characters
2432 * since any characters in the low 256 are Unicode
2433 * code points, not EBCDIC. */
7cda7a3d 2434 U8 *t = (U8*)s, *tend = t + len, *d;
2f9475ad
JH
2435
2436 d = tmpbuf;
b08cf34e 2437 if (SvUTF8(*svp)) {
2f9475ad
JH
2438 STRLEN tlen = 0;
2439
2440 while (t < tend) {
6bd1c396 2441 const UV c = utf8_to_uvchr_buf(t, tend, &tlen);
2f9475ad
JH
2442 if (tlen > 0) {
2443 d = uvchr_to_utf8(d, UNI_TO_NATIVE(c));
2444 t += tlen;
2445 }
2446 else
2447 break;
2448 }
2449 }
2450 else {
36fec512
JH
2451 while (t < tend) {
2452 d = uvchr_to_utf8(d, UNI_TO_NATIVE(*t));
2453 t++;
2454 }
2f9475ad
JH
2455 }
2456 len = d - tmpbuf;
2457 Copy(tmpbuf, ustrp, len, U8);
2458#else
d2dcd0fb 2459 Copy(s, ustrp, len, U8);
2f9475ad 2460#endif
29e98929 2461 }
983ffd37 2462 }
0134edef
JH
2463 }
2464
2465 if (!len && *swashp) {
2114036c 2466 const UV uv2 = swash_fetch(*swashp, tmpbuf, TRUE /* => is utf8 */);
d4c19fe8 2467
0134edef
JH
2468 if (uv2) {
2469 /* It was "normal" (a single character mapping). */
d4c19fe8 2470 const UV uv3 = UNI_TO_NATIVE(uv2);
e9101d72 2471 len = uvchr_to_utf8(ustrp, uv3) - ustrp;
2a37f04d
JH
2472 }
2473 }
1feea2c7 2474
cbe07460
KW
2475 if (len) {
2476 if (lenp) {
2477 *lenp = len;
2478 }
2479 return valid_utf8_to_uvchr(ustrp, 0);
2480 }
2481
2482 /* Here, there was no mapping defined, which means that the code point maps
2483 * to itself. Return the inputs */
bfdf22ec 2484 len = UTF8SKIP(p);
ca9fab46
KW
2485 if (p != ustrp) { /* Don't copy onto itself */
2486 Copy(p, ustrp, len, U8);
2487 }
0134edef 2488
2a37f04d
JH
2489 if (lenp)
2490 *lenp = len;
2491
cbe07460
KW
2492 return uv0;
2493
a0ed51b3
LW
2494}
2495
051a06d4
KW
2496STATIC UV
2497S_check_locale_boundary_crossing(pTHX_ const U8* const p, const UV result, U8* const ustrp, STRLEN *lenp)
2498{
2499 /* This is called when changing the case of a utf8-encoded character above
2500 * the Latin1 range, and the operation is in locale. If the result
2501 * contains a character that crosses the 255/256 boundary, disallow the
2502 * change, and return the original code point. See L<perlfunc/lc> for why;
2503 *
a1433954
KW
2504 * p points to the original string whose case was changed; assumed
2505 * by this routine to be well-formed
051a06d4
KW
2506 * result the code point of the first character in the changed-case string
2507 * ustrp points to the changed-case string (<result> represents its first char)
2508 * lenp points to the length of <ustrp> */
2509
2510 UV original; /* To store the first code point of <p> */
2511
2512 PERL_ARGS_ASSERT_CHECK_LOCALE_BOUNDARY_CROSSING;
2513
2514 assert(! UTF8_IS_INVARIANT(*p) && ! UTF8_IS_DOWNGRADEABLE_START(*p));
2515
2516 /* We know immediately if the first character in the string crosses the
2517 * boundary, so can skip */
2518 if (result > 255) {
2519
2520 /* Look at every character in the result; if any cross the
2521 * boundary, the whole thing is disallowed */
2522 U8* s = ustrp + UTF8SKIP(ustrp);
2523 U8* e = ustrp + *lenp;
2524 while (s < e) {
2525 if (UTF8_IS_INVARIANT(*s) || UTF8_IS_DOWNGRADEABLE_START(*s))
2526 {
2527 goto bad_crossing;
2528 }
2529 s += UTF8SKIP(s);
2530 }
2531
2532 /* Here, no characters crossed, result is ok as-is */
2533 return result;
2534 }
2535
2536bad_crossing:
2537
2538 /* Failed, have to return the original */
4b88fb76 2539 original = valid_utf8_to_uvchr(p, lenp);
051a06d4
KW
2540 Copy(p, ustrp, *lenp, char);
2541 return original;
2542}
2543
d3e79532 2544/*
87cea99e 2545=for apidoc to_utf8_upper
d3e79532 2546
a1433954
KW
2547Convert the UTF-8 encoded character at C<p> to its uppercase version and
2548store that in UTF-8 in C<ustrp> and its length in bytes in C<lenp>. Note
89ebb4a3
JH
2549that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since
2550the uppercase version may be longer than the original character.
d3e79532
JH
2551
2552The first character of the uppercased version is returned
2553(but note, as explained above, that there may be more.)
2554
a1433954
KW
2555The character at C<p> is assumed by this routine to be well-formed.
2556
d3e79532
JH
2557=cut */
2558
051a06d4
KW
2559/* Not currently externally documented, and subject to change:
2560 * <flags> is set iff locale semantics are to be used for code points < 256
2561 * <tainted_ptr> if non-null, *tainted_ptr will be set TRUE iff locale rules
2562 * were used in the calculation; otherwise unchanged. */
2563
2104c8d9 2564UV
051a06d4 2565Perl__to_utf8_upper_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, const bool flags, bool* tainted_ptr)
a0ed51b3 2566{
97aff369 2567 dVAR;
7918f24d 2568
051a06d4
KW
2569 UV result;
2570
2571 PERL_ARGS_ASSERT__TO_UTF8_UPPER_FLAGS;
7918f24d 2572
3a4c58c9 2573 if (UTF8_IS_INVARIANT(*p)) {
051a06d4
KW
2574 if (flags) {
2575 result = toUPPER_LC(*p);
2576 }
2577 else {
81c6c7ce 2578 return _to_upper_title_latin1(*p, ustrp, lenp, 'S');
051a06d4 2579 }
3a4c58c9
KW
2580 }
2581 else if UTF8_IS_DOWNGRADEABLE_START(*p) {
051a06d4
KW
2582 if (flags) {
2583 result = toUPPER_LC(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)));
2584 }
2585 else {
81c6c7ce
KW
2586 return _to_upper_title_latin1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)),
2587 ustrp, lenp, 'S');
051a06d4
KW
2588 }
2589 }
2590 else { /* utf8, ord above 255 */
2591 result = CALL_UPPER_CASE(p, ustrp, lenp);
2592
2593 if (flags) {
2594 result = check_locale_boundary_crossing(p, result, ustrp, lenp);
2595 }
2596 return result;
2597 }
2598
2599 /* Here, used locale rules. Convert back to utf8 */
2600 if (UTF8_IS_INVARIANT(result)) {
2601 *ustrp = (U8) result;
2602 *lenp = 1;
2603 }
2604 else {
2605 *ustrp = UTF8_EIGHT_BIT_HI(result);
2606 *(ustrp + 1) = UTF8_EIGHT_BIT_LO(result);
2607 *lenp = 2;
3a4c58c9
KW
2608 }
2609
051a06d4
KW
2610 if (tainted_ptr) {
2611 *tainted_ptr = TRUE;
2612 }
2613 return result;
983ffd37 2614}
a0ed51b3 2615
d3e79532 2616/*
87cea99e 2617=for apidoc to_utf8_title
d3e79532 2618
a1433954
KW
2619Convert the UTF-8 encoded character at C<p> to its titlecase version and
2620store that in UTF-8 in C<ustrp> and its length in bytes in C<lenp>. Note
2621that the C<ustrp> needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
89ebb4a3 2622titlecase version may be longer than the original character.
d3e79532
JH
2623
2624The first character of the titlecased version is returned
2625(but note, as explained above, that there may be more.)
2626
a1433954
KW
2627The character at C<p> is assumed by this routine to be well-formed.
2628
d3e79532
JH
2629=cut */
2630
051a06d4
KW
2631/* Not currently externally documented, and subject to change:
2632 * <flags> is set iff locale semantics are to be used for code points < 256
2633 * Since titlecase is not defined in POSIX, uppercase is used instead
2634 * for these/
2635 * <tainted_ptr> if non-null, *tainted_ptr will be set TRUE iff locale rules
2636 * were used in the calculation; otherwise unchanged. */
2637
983ffd37 2638UV
051a06d4 2639Perl__to_utf8_title_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, const bool flags, bool* tainted_ptr)
983ffd37 2640{
97aff369 2641 dVAR;
7918f24d 2642
051a06d4
KW
2643 UV result;
2644
2645 PERL_ARGS_ASSERT__TO_UTF8_TITLE_FLAGS;
7918f24d 2646
3a4c58c9 2647 if (UTF8_IS_INVARIANT(*p)) {
051a06d4
KW
2648 if (flags) {
2649 result = toUPPER_LC(*p);
2650 }
2651 else {
81c6c7ce 2652 return _to_upper_title_latin1(*p, ustrp, lenp, 's');
051a06d4 2653 }
3a4c58c9
KW
2654 }
2655 else if UTF8_IS_DOWNGRADEABLE_START(*p) {
051a06d4
KW
2656 if (flags) {
2657 result = toUPPER_LC(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)));
2658 }
2659 else {
81c6c7ce
KW
2660 return _to_upper_title_latin1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)),
2661 ustrp, lenp, 's');
051a06d4
KW
2662 }
2663 }
2664 else { /* utf8, ord above 255 */
2665 result = CALL_TITLE_CASE(p, ustrp, lenp);
2666
2667 if (flags) {
2668 result = check_locale_boundary_crossing(p, result, ustrp, lenp);
2669 }
2670 return result;
2671 }
2672
2673 /* Here, used locale rules. Convert back to utf8 */
2674 if (UTF8_IS_INVARIANT(result)) {
2675 *ustrp = (U8) result;
2676 *lenp = 1;
2677 }
2678 else {
2679 *ustrp = UTF8_EIGHT_BIT_HI(result);
2680 *(ustrp + 1) = UTF8_EIGHT_BIT_LO(result);
2681 *lenp = 2;
3a4c58c9
KW
2682 }
2683
051a06d4
KW
2684 if (tainted_ptr) {
2685 *tainted_ptr = TRUE;
2686 }
2687 return result;
a0ed51b3
LW
2688}
2689
d3e79532 2690/*
87cea99e 2691=for apidoc to_utf8_lower
d3e79532 2692
a1433954
KW
2693Convert the UTF-8 encoded character at C<p> to its lowercase version and
2694store that in UTF-8 in ustrp and its length in bytes in C<lenp>. Note
2695that the C<ustrp> needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
89ebb4a3 2696lowercase version may be longer than the original character.
d3e79532
JH
2697
2698The first character of the lowercased version is returned
2699(but note, as explained above, that there may be more.)
2700
a1433954
KW
2701The character at C<p> is assumed by this routine to be well-formed.
2702
d3e79532
JH
2703=cut */
2704
051a06d4
KW
2705/* Not currently externally documented, and subject to change:
2706 * <flags> is set iff locale semantics are to be used for code points < 256
2707 * <tainted_ptr> if non-null, *tainted_ptr will be set TRUE iff locale rules
2708 * were used in the calculation; otherwise unchanged. */
2709
2104c8d9 2710UV
051a06d4 2711Perl__to_utf8_lower_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, const bool flags, bool* tainted_ptr)
a0ed51b3 2712{
051a06d4
KW
2713 UV result;
2714
97aff369 2715 dVAR;
7918f24d 2716
051a06d4 2717 PERL_ARGS_ASSERT__TO_UTF8_LOWER_FLAGS;
7918f24d 2718
968c5e6a 2719 if (UTF8_IS_INVARIANT(*p)) {
051a06d4
KW
2720 if (flags) {
2721 result = toLOWER_LC(*p);
2722 }
2723 else {
81c6c7ce 2724 return to_lower_latin1(*p, ustrp, lenp);
051a06d4 2725 }
968c5e6a
KW
2726 }
2727 else if UTF8_IS_DOWNGRADEABLE_START(*p) {
051a06d4
KW
2728 if (flags) {
2729 result = toLOWER_LC(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)));
2730 }
2731 else {
81c6c7ce
KW
2732 return to_lower_latin1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)),
2733 ustrp, lenp);
051a06d4 2734 }
968c5e6a 2735 }
051a06d4
KW
2736 else { /* utf8, ord above 255 */
2737 result = CALL_LOWER_CASE(p, ustrp, lenp);
2738
2739 if (flags) {
2740 result = check_locale_boundary_crossing(p, result, ustrp, lenp);
2741 }
968c5e6a 2742
051a06d4
KW
2743 return result;
2744 }
2745
2746 /* Here, used locale rules. Convert back to utf8 */
2747 if (UTF8_IS_INVARIANT(result)) {
2748 *ustrp = (U8) result;
2749 *lenp = 1;
2750 }
2751 else {
2752 *ustrp = UTF8_EIGHT_BIT_HI(result);
2753 *(ustrp + 1) = UTF8_EIGHT_BIT_LO(result);
2754 *lenp = 2;
2755 }
2756
2757 if (tainted_ptr) {
2758 *tainted_ptr = TRUE;
2759 }
2760 return result;
b4e400f9
JH
2761}
2762
d3e79532 2763/*
87cea99e 2764=for apidoc to_utf8_fold
d3e79532 2765
a1433954
KW
2766Convert the UTF-8 encoded character at C<p> to its foldcase version and
2767store that in UTF-8 in C<ustrp> and its length in bytes in C<lenp>. Note
2768that the C<ustrp> needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
d3e79532
JH
2769foldcase version may be longer than the original character (up to
2770three characters).
2771
2772The first character of the foldcased version is returned
2773(but note, as explained above, that there may be more.)
2774
a1433954
KW
2775The character at C<p> is assumed by this routine to be well-formed.
2776
d3e79532
JH
2777=cut */
2778
051a06d4
KW
2779/* Not currently externally documented, and subject to change,
2780 * in <flags>
2781 * bit FOLD_FLAGS_LOCALE is set iff locale semantics are to be used for code
2782 * points < 256. Since foldcase is not defined in
2783 * POSIX, lowercase is used instead
2784 * bit FOLD_FLAGS_FULL is set iff full case folds are to be used;
2785 * otherwise simple folds
a0270393
KW
2786 * bit FOLD_FLAGS_NOMIX_ASCII is set iff folds of non-ASCII to ASCII are
2787 * prohibited
051a06d4
KW
2788 * <tainted_ptr> if non-null, *tainted_ptr will be set TRUE iff locale rules
2789 * were used in the calculation; otherwise unchanged. */
36bb2ab6 2790
b4e400f9 2791UV
051a06d4 2792Perl__to_utf8_fold_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, U8 flags, bool* tainted_ptr)
b4e400f9 2793{
97aff369 2794 dVAR;
7918f24d 2795
051a06d4
KW
2796 UV result;
2797
36bb2ab6 2798 PERL_ARGS_ASSERT__TO_UTF8_FOLD_FLAGS;
7918f24d 2799
a0270393
KW
2800 /* These are mutually exclusive */
2801 assert (! ((flags & FOLD_FLAGS_LOCALE) && (flags & FOLD_FLAGS_NOMIX_ASCII)));
2802
50ba90ff
KW
2803 assert(p != ustrp); /* Otherwise overwrites */
2804
a1dde8de 2805 if (UTF8_IS_INVARIANT(*p)) {
051a06d4
KW
2806 if (flags & FOLD_FLAGS_LOCALE) {
2807 result = toLOWER_LC(*p);
2808 }
2809 else {
81c6c7ce
KW
2810 return _to_fold_latin1(*p, ustrp, lenp,
2811 cBOOL(flags & FOLD_FLAGS_FULL));
051a06d4 2812 }
a1dde8de
KW
2813 }
2814 else if UTF8_IS_DOWNGRADEABLE_START(*p) {
051a06d4
KW
2815 if (flags & FOLD_FLAGS_LOCALE) {
2816 result = toLOWER_LC(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)));
2817 }
2818 else {
81c6c7ce 2819 return _to_fold_latin1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)),
a0270393
KW
2820 ustrp, lenp,
2821 cBOOL((flags & FOLD_FLAGS_FULL
2822 /* If ASCII safe, don't allow full
2823 * folding, as that could include SHARP
2824 * S => ss; otherwise there is no
2825 * crossing of ascii/non-ascii in the
2826 * latin1 range */
2827 && ! (flags & FOLD_FLAGS_NOMIX_ASCII))));
051a06d4 2828 }
a1dde8de 2829 }
051a06d4 2830 else { /* utf8, ord above 255 */
a0270393 2831 result = CALL_FOLD_CASE(p, ustrp, lenp, flags & FOLD_FLAGS_FULL);
a1dde8de 2832
051a06d4 2833 if ((flags & FOLD_FLAGS_LOCALE)) {
a0270393 2834 return check_locale_boundary_crossing(p, result, ustrp, lenp);
051a06d4 2835 }
a0270393
KW
2836 else if (! (flags & FOLD_FLAGS_NOMIX_ASCII)) {
2837 return result;
2838 }
2839 else {
2840 /* This is called when changing the case of a utf8-encoded
2841 * character above the Latin1 range, and the result should not
2842 * contain an ASCII character. */
2843
2844 UV original; /* To store the first code point of <p> */
2845
2846 /* Look at every character in the result; if any cross the
2847 * boundary, the whole thing is disallowed */
2848 U8* s = ustrp;
2849 U8* e = ustrp + *lenp;
2850 while (s < e) {
2851 if (isASCII(*s)) {
2852 /* Crossed, have to return the original */
2853 original = valid_utf8_to_uvchr(p, lenp);
2854 Copy(p, ustrp, *lenp, char);
2855 return original;
2856 }
2857 s += UTF8SKIP(s);
2858 }
051a06d4 2859
a0270393
KW
2860 /* Here, no characters crossed, result is ok as-is */
2861 return result;
2862 }
051a06d4
KW
2863 }
2864
2865 /* Here, used locale rules. Convert back to utf8 */
2866 if (UTF8_IS_INVARIANT(result)) {
2867 *ustrp = (U8) result;
2868 *lenp = 1;
2869 }
2870 else {
2871 *ustrp = UTF8_EIGHT_BIT_HI(result);
2872 *(ustrp + 1) = UTF8_EIGHT_BIT_LO(result);
2873 *lenp = 2;
2874 }
2875
2876 if (tainted_ptr) {
2877 *tainted_ptr = TRUE;
2878 }
2879 return result;
a0ed51b3
LW
2880}
2881
711a919c 2882/* Note:
f90a9a02 2883 * Returns a "swash" which is a hash described in utf8.c:Perl_swash_fetch().
711a919c
ST
2884 * C<pkg> is a pointer to a package name for SWASHNEW, should be "utf8".
2885 * For other parameters, see utf8::SWASHNEW in lib/utf8_heavy.pl.
2886 */
c4a5db0c 2887
a0ed51b3 2888SV*
7fc63493 2889Perl_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv, I32 minbits, I32 none)
a0ed51b3 2890{
c4a5db0c
KW
2891 PERL_ARGS_ASSERT_SWASH_INIT;
2892
2893 /* Returns a copy of a swash initiated by the called function. This is the
2894 * public interface, and returning a copy prevents others from doing
2895 * mischief on the original */
2896
5d3d13d1 2897 return newSVsv(_core_swash_init(pkg, name, listsv, minbits, none, NULL, NULL));
c4a5db0c
KW
2898}
2899
2900SV*
5d3d13d1 2901Perl__core_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv, I32 minbits, I32 none, SV* invlist, U8* const flags_p)
c4a5db0c
KW
2902{
2903 /* Initialize and return a swash, creating it if necessary. It does this
87367d5f
KW
2904 * by calling utf8_heavy.pl in the general case. The returned value may be
2905 * the swash's inversion list instead if the input parameters allow it.
2906 * Which is returned should be immaterial to callers, as the only
923b6d4e
KW
2907 * operations permitted on a swash, swash_fetch(), _get_swash_invlist(),
2908 * and swash_to_invlist() handle both these transparently.
c4a5db0c
KW
2909 *
2910 * This interface should only be used by functions that won't destroy or
2911 * adversely change the swash, as doing so affects all other uses of the
2912 * swash in the program; the general public should use 'Perl_swash_init'
2913 * instead.
2914 *
2915 * pkg is the name of the package that <name> should be in.
2916 * name is the name of the swash to find. Typically it is a Unicode
2917 * property name, including user-defined ones
2918 * listsv is a string to initialize the swash with. It must be of the form
2919 * documented as the subroutine return value in
2920 * L<perlunicode/User-Defined Character Properties>
2921 * minbits is the number of bits required to represent each data element.
2922 * It is '1' for binary properties.
2923 * none I (khw) do not understand this one, but it is used only in tr///.
9a53f6cf 2924 * invlist is an inversion list to initialize the swash with (or NULL)
83199d38
KW
2925 * flags_p if non-NULL is the address of various input and output flag bits
2926 * to the routine, as follows: ('I' means is input to the routine;
2927 * 'O' means output from the routine. Only flags marked O are
2928 * meaningful on return.)
2929 * _CORE_SWASH_INIT_USER_DEFINED_PROPERTY indicates if the swash
2930 * came from a user-defined property. (I O)
5d3d13d1
KW
2931 * _CORE_SWASH_INIT_RETURN_IF_UNDEF indicates that instead of croaking
2932 * when the swash cannot be located, to simply return NULL. (I)
87367d5f
KW
2933 * _CORE_SWASH_INIT_ACCEPT_INVLIST indicates that the caller will accept a
2934 * return of an inversion list instead of a swash hash if this routine
2935 * thinks that would result in faster execution of swash_fetch() later
2936 * on. (I)
9a53f6cf
KW
2937 *
2938 * Thus there are three possible inputs to find the swash: <name>,
2939 * <listsv>, and <invlist>. At least one must be specified. The result
2940 * will be the union of the specified ones, although <listsv>'s various
2941 * actions can intersect, etc. what <name> gives.
2942 *
2943 * <invlist> is only valid for binary properties */
c4a5db0c 2944
27da23d5 2945 dVAR;
c4a5db0c 2946 SV* retval = &PL_sv_undef;
83199d38 2947 HV* swash_hv = NULL;
87367d5f
KW
2948 const int invlist_swash_boundary =
2949 (flags_p && *flags_p & _CORE_SWASH_INIT_ACCEPT_INVLIST)
2950 ? 512 /* Based on some benchmarking, but not extensive, see commit
2951 message */
2952 : -1; /* Never return just an inversion list */
9a53f6cf
KW
2953
2954 assert(listsv != &PL_sv_undef || strNE(name, "") || invlist);
2955 assert(! invlist || minbits == 1);
2956
2957 /* If data was passed in to go out to utf8_heavy to find the swash of, do
2958 * so */
2959 if (listsv != &PL_sv_undef || strNE(name, "")) {
69794297
KW
2960 dSP;
2961 const size_t pkg_len = strlen(pkg);
2962 const size_t name_len = strlen(name);
2963 HV * const stash = gv_stashpvn(pkg, pkg_len, 0);
2964 SV* errsv_save;
2965 GV *method;
2966
2967 PERL_ARGS_ASSERT__CORE_SWASH_INIT;
2968
2969 PUSHSTACKi(PERLSI_MAGIC);
ce3b816e 2970 ENTER;
69794297
KW
2971 SAVEHINTS();
2972 save_re_context();
650f067c
JL
2973 /* We might get here via a subroutine signature which uses a utf8
2974 * parameter name, at which point PL_subname will have been set
2975 * but not yet used. */
2976 save_item(PL_subname);
69794297
KW
2977 if (PL_parser && PL_parser->error_count)
2978 SAVEI8(PL_parser->error_count), PL_parser->error_count = 0;
2979 method = gv_fetchmeth(stash, "SWASHNEW", 8, -1);
2980 if (!method) { /* demand load utf8 */
2981 ENTER;
db2c6cb3
FC
2982 if ((errsv_save = GvSV(PL_errgv))) SAVEFREESV(errsv_save);
2983 GvSV(PL_errgv) = NULL;
69794297
KW
2984 /* It is assumed that callers of this routine are not passing in
2985 * any user derived data. */
2986 /* Need to do this after save_re_context() as it will set
2987 * PL_tainted to 1 while saving $1 etc (see the code after getrx:
2988 * in Perl_magic_get). Even line to create errsv_save can turn on
2989 * PL_tainted. */
284167a5
SM
2990#ifndef NO_TAINT_SUPPORT
2991 SAVEBOOL(TAINT_get);
2992 TAINT_NOT;
2993#endif
69794297
KW
2994 Perl_load_module(aTHX_ PERL_LOADMOD_NOIMPORT, newSVpvn(pkg,pkg_len),
2995 NULL);
eed484f9 2996 {
db2c6cb3
FC
2997 /* Not ERRSV, as there is no need to vivify a scalar we are
2998 about to discard. */
2999 SV * const errsv = GvSV(PL_errgv);
3000 if (!SvTRUE(errsv)) {
3001 GvSV(PL_errgv) = SvREFCNT_inc_simple(errsv_save);
3002 SvREFCNT_dec(errsv);
3003 }
eed484f9 3004 }
69794297
KW
3005 LEAVE;
3006 }
3007 SPAGAIN;
3008 PUSHMARK(SP);
3009 EXTEND(SP,5);
3010 mPUSHp(pkg, pkg_len);
3011 mPUSHp(name, name_len);
3012 PUSHs(listsv);
3013 mPUSHi(minbits);
3014 mPUSHi(none);
3015 PUTBACK;
db2c6cb3
FC
3016 if ((errsv_save = GvSV(PL_errgv))) SAVEFREESV(errsv_save);
3017 GvSV(PL_errgv) = NULL;
69794297
KW
3018 /* If we already have a pointer to the method, no need to use
3019 * call_method() to repeat the lookup. */
c41800a8
KW
3020 if (method
3021 ? call_sv(MUTABLE_SV(method), G_SCALAR)
69794297
KW
3022 : call_sv(newSVpvs_flags("SWASHNEW", SVs_TEMP), G_SCALAR | G_METHOD))
3023 {
3024 retval = *PL_stack_sp--;
3025 SvREFCNT_inc(retval);
3026 }
eed484f9 3027 {
db2c6cb3
FC
3028 /* Not ERRSV. See above. */
3029 SV * const errsv = GvSV(PL_errgv);
3030 if (!SvTRUE(errsv)) {
3031 GvSV(PL_errgv) = SvREFCNT_inc_simple(errsv_save);
3032 SvREFCNT_dec(errsv);
3033 }
eed484f9 3034 }
ce3b816e 3035 LEAVE;
69794297
KW
3036 POPSTACK;
3037 if (IN_PERL_COMPILETIME) {
3038 CopHINTS_set(PL_curcop, PL_hints);
3039 }
3040 if (!SvROK(retval) || SvTYPE(SvRV(retval)) != SVt_PVHV) {
3041 if (SvPOK(retval))
3042
3043 /* If caller wants to handle missing properties, let them */
5d3d13d1 3044 if (flags_p && *flags_p & _CORE_SWASH_INIT_RETURN_IF_UNDEF) {
69794297
KW
3045 return NULL;
3046 }
3047 Perl_croak(aTHX_
3048 "Can't find Unicode property definition \"%"SVf"\"",
3049 SVfARG(retval));
3050 Perl_croak(aTHX_ "SWASHNEW didn't return an HV ref");
3051 }
9a53f6cf 3052 } /* End of calling the module to find the swash */
36eb48b4 3053
83199d38
KW
3054 /* If this operation fetched a swash, and we will need it later, get it */
3055 if (retval != &PL_sv_undef
3056 && (minbits == 1 || (flags_p
3057 && ! (*flags_p
3058 & _CORE_SWASH_INIT_USER_DEFINED_PROPERTY))))
3059 {
3060 swash_hv = MUTABLE_HV(SvRV(retval));
3061
3062 /* If we don't already know that there is a user-defined component to
3063 * this swash, and the user has indicated they wish to know if there is
3064 * one (by passing <flags_p>), find out */
3065 if (flags_p && ! (*flags_p & _CORE_SWASH_INIT_USER_DEFINED_PROPERTY)) {
3066 SV** user_defined = hv_fetchs(swash_hv, "USER_DEFINED", FALSE);
3067 if (user_defined && SvUV(*user_defined)) {
3068 *flags_p |= _CORE_SWASH_INIT_USER_DEFINED_PROPERTY;
3069 }
3070 }
3071 }
3072
36eb48b4
KW
3073 /* Make sure there is an inversion list for binary properties */
3074 if (minbits == 1) {
3075 SV** swash_invlistsvp = NULL;
3076 SV* swash_invlist = NULL;
9a53f6cf 3077 bool invlist_in_swash_is_valid = FALSE;
02c85471
FC
3078 bool swash_invlist_unclaimed = FALSE; /* whether swash_invlist has
3079 an unclaimed reference count */
36eb48b4 3080
9a53f6cf 3081 /* If this operation fetched a swash, get its already existing
83199d38 3082 * inversion list, or create one for it */
36eb48b4 3083
83199d38 3084 if (swash_hv) {
5c9f4bd2 3085 swash_invlistsvp = hv_fetchs(swash_hv, "V", FALSE);
9a53f6cf
KW
3086 if (swash_invlistsvp) {
3087 swash_invlist = *swash_invlistsvp;
3088 invlist_in_swash_is_valid = TRUE;
3089 }
3090 else {
36eb48b4 3091 swash_invlist = _swash_to_invlist(retval);
02c85471 3092 swash_invlist_unclaimed = TRUE;
9a53f6cf
KW
3093 }
3094 }
3095
3096 /* If an inversion list was passed in, have to include it */
3097 if (invlist) {
3098
3099 /* Any fetched swash will by now have an inversion list in it;
3100 * otherwise <swash_invlist> will be NULL, indicating that we
3101 * didn't fetch a swash */
3102 if (swash_invlist) {
3103
3104 /* Add the passed-in inversion list, which invalidates the one
3105 * already stored in the swash */
3106 invlist_in_swash_is_valid = FALSE;
3107 _invlist_union(invlist, swash_invlist, &swash_invlist);
3108 }
3109 else {
3110
87367d5f
KW
3111 /* Here, there is no swash already. Set up a minimal one, if
3112 * we are going to return a swash */
3113 if ((int) _invlist_len(invlist) > invlist_swash_boundary) {
971d486f 3114 swash_hv = newHV();
4aca0fe6 3115 retval = newRV_noinc(MUTABLE_SV(swash_hv));
87367d5f 3116 }
9a53f6cf
KW
3117 swash_invlist = invlist;
3118 }
9a53f6cf
KW
3119 }
3120
3121 /* Here, we have computed the union of all the passed-in data. It may
3122 * be that there was an inversion list in the swash which didn't get
3123 * touched; otherwise save the one computed one */
87367d5f
KW
3124 if (! invlist_in_swash_is_valid
3125 && (int) _invlist_len(swash_invlist) > invlist_swash_boundary)
3126 {
5c9f4bd2 3127 if (! hv_stores(MUTABLE_HV(SvRV(retval)), "V", swash_invlist))
69794297
KW
3128 {
3129 Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed");
3130 }
cc34d8c5
FC
3131 /* We just stole a reference count. */
3132 if (swash_invlist_unclaimed) swash_invlist_unclaimed = FALSE;
3133 else SvREFCNT_inc_simple_void_NN(swash_invlist);
9a53f6cf 3134 }
87367d5f 3135
c41800a8 3136 /* Use the inversion list stand-alone if small enough */
87367d5f
KW
3137 if ((int) _invlist_len(swash_invlist) <= invlist_swash_boundary) {
3138 SvREFCNT_dec(retval);
02c85471
FC
3139 if (!swash_invlist_unclaimed)
3140 SvREFCNT_inc_simple_void_NN(swash_invlist);
3141 retval = newRV_noinc(swash_invlist);
87367d5f 3142 }
36eb48b4
KW
3143 }
3144
a0ed51b3
LW
3145 return retval;
3146}
3147
035d37be
JH
3148
3149/* This API is wrong for special case conversions since we may need to
3150 * return several Unicode characters for a single Unicode character
3151 * (see lib/unicore/SpecCase.txt) The SWASHGET in lib/utf8_heavy.pl is
3152 * the lower-level routine, and it is similarly broken for returning
38684baa
KW
3153 * multiple values. --jhi
3154 * For those, you should use to_utf8_case() instead */
b0e3252e 3155/* Now SWASHGET is recasted into S_swatch_get in this file. */
680c470c
ST
3156
3157/* Note:
3158 * Returns the value of property/mapping C<swash> for the first character
3159 * of the string C<ptr>. If C<do_utf8> is true, the string C<ptr> is
3160 * assumed to be in utf8. If C<do_utf8> is false, the string C<ptr> is
3161 * assumed to be in native 8-bit encoding. Caches the swatch in C<swash>.
af2af982
KW
3162 *
3163 * A "swash" is a hash which contains initially the keys/values set up by
3164 * SWASHNEW. The purpose is to be able to completely represent a Unicode
3165 * property for all possible code points. Things are stored in a compact form
3166 * (see utf8_heavy.pl) so that calculation is required to find the actual
3167 * property value for a given code point. As code points are looked up, new
3168 * key/value pairs are added to the hash, so that the calculation doesn't have
3169 * to ever be re-done. Further, each calculation is done, not just for the
3170 * desired one, but for a whole block of code points adjacent to that one.
3171 * For binary properties on ASCII machines, the block is usually for 64 code
3172 * points, starting with a code point evenly divisible by 64. Thus if the
3173 * property value for code point 257 is requested, the code goes out and
3174 * calculates the property values for all 64 code points between 256 and 319,
3175 * and stores these as a single 64-bit long bit vector, called a "swatch",
3176 * under the key for code point 256. The key is the UTF-8 encoding for code
3177 * point 256, minus the final byte. Thus, if the length of the UTF-8 encoding
3178 * for a code point is 13 bytes, the key will be 12 bytes long. If the value
3179 * for code point 258 is then requested, this code realizes that it would be
3180 * stored under the key for 256, and would find that value and extract the
3181 * relevant bit, offset from 256.
3182 *
3183 * Non-binary properties are stored in as many bits as necessary to represent
3184 * their values (32 currently, though the code is more general than that), not
3185 * as single bits, but the principal is the same: the value for each key is a
3186 * vector that encompasses the property values for all code points whose UTF-8
3187 * representations are represented by the key. That is, for all code points
3188 * whose UTF-8 representations are length N bytes, and the key is the first N-1
3189 * bytes of that.
680c470c 3190 */
a0ed51b3 3191UV
680c470c 3192Perl_swash_fetch(pTHX_ SV *swash, const U8 *ptr, bool do_utf8)
a0ed51b3 3193{
27da23d5 3194 dVAR;
ef8f7699 3195 HV *const hv = MUTABLE_HV(SvRV(swash));
3568d838
JH
3196 U32 klen;
3197 U32 off;
a0ed51b3 3198 STRLEN slen;
7d85a32c 3199 STRLEN needents;
cfd0369c 3200 const U8 *tmps = NULL;
a0ed51b3 3201 U32 bit;
979f2922 3202 SV *swatch;
3568d838 3203 U8 tmputf8[2];
35da51f7 3204 const UV c = NATIVE_TO_ASCII(*ptr);
3568d838 3205
7918f24d
NC
3206 PERL_ARGS_ASSERT_SWASH_FETCH;
3207
87367d5f
KW
3208 /* If it really isn't a hash, it isn't really swash; must be an inversion
3209 * list */
3210 if (SvTYPE(hv) != SVt_PVHV) {
3211 return _invlist_contains_cp((SV*)hv,
3212 (do_utf8)
3213 ? valid_utf8_to_uvchr(ptr, NULL)
3214 : c);
3215 }
3216
dbe7a391 3217 /* Convert to utf8 if not already */
3568d838 3218 if (!do_utf8 && !UNI_IS_INVARIANT(c)) {
979f2922
ST
3219 tmputf8[0] = (U8)UTF8_EIGHT_BIT_HI(c);
3220 tmputf8[1] = (U8)UTF8_EIGHT_BIT_LO(c);
3221 ptr = tmputf8;
3568d838
JH
3222 }
3223 /* Given a UTF-X encoded char 0xAA..0xYY,0xZZ
37e2e78e 3224 * then the "swatch" is a vec() for all the chars which start
3568d838
JH
3225 * with 0xAA..0xYY
3226 * So the key in the hash (klen) is length of encoded char -1
3227 */
3228 klen = UTF8SKIP(ptr) - 1;
3229 off = ptr[klen];
a0ed51b3 3230
979f2922 3231 if (klen == 0) {
37e2e78e 3232 /* If char is invariant then swatch is for all the invariant chars
1e54db1a 3233 * In both UTF-8 and UTF-8-MOD that happens to be UTF_CONTINUATION_MARK
7d85a32c 3234 */
979f2922
ST
3235 needents = UTF_CONTINUATION_MARK;
3236 off = NATIVE_TO_UTF(ptr[klen]);
3237 }
3238 else {
7d85a32c 3239 /* If char is encoded then swatch is for the prefix */
979f2922
ST
3240 needents = (1 << UTF_ACCUMULATION_SHIFT);
3241 off = NATIVE_TO_UTF(ptr[klen]) & UTF_CONTINUATION_MASK;
3242 }
7d85a32c 3243
a0ed51b3
LW
3244 /*
3245 * This single-entry cache saves about 1/3 of the utf8 overhead in test
3246 * suite. (That is, only 7-8% overall over just a hash cache. Still,
3247 * it's nothing to sniff at.) Pity we usually come through at least
3248 * two function calls to get here...
3249 *
3250 * NB: this code assumes that swatches are never modified, once generated!
3251 */
3252
3568d838 3253 if (hv == PL_last_swash_hv &&
a0ed51b3 3254 klen == PL_last_swash_klen &&
27da23d5 3255 (!klen || memEQ((char *)ptr, (char *)PL_last_swash_key, klen)) )
a0ed51b3
LW
3256 {
3257 tmps = PL_last_swash_tmps;
3258 slen = PL_last_swash_slen;
3259 }
3260 else {
3261 /* Try our second-level swatch cache, kept in a hash. */
e1ec3a88 3262 SV** svp = hv_fetch(hv, (const char*)ptr, klen, FALSE);
a0ed51b3 3263
b0e3252e 3264 /* If not cached, generate it via swatch_get */
979f2922
ST
3265 if (!svp || !SvPOK(*svp)
3266 || !(tmps = (const U8*)SvPV_const(*svp, slen))) {
2b9d42f0
NIS
3267 /* We use utf8n_to_uvuni() as we want an index into
3268 Unicode tables, not a native character number.
3269 */
aec46f14 3270 const UV code_point = utf8n_to_uvuni(ptr, UTF8_MAXBYTES, 0,
872c91ae
JH
3271 ckWARN(WARN_UTF8) ?
3272 0 : UTF8_ALLOW_ANY);
b0e3252e 3273 swatch = swatch_get(swash,
979f2922 3274 /* On EBCDIC & ~(0xA0-1) isn't a useful thing to do */
361ee0fe 3275 (klen) ? (code_point & ~((UV)needents - 1)) : 0,
979f2922
ST
3276 needents);
3277
923e4eb5 3278 if (IN_PERL_COMPILETIME)
623e6609 3279 CopHINTS_set(PL_curcop, PL_hints);
a0ed51b3 3280
979f2922 3281 svp = hv_store(hv, (const char *)ptr, klen, swatch, 0);
a0ed51b3 3282
979f2922
ST
3283 if (!svp || !(tmps = (U8*)SvPV(*svp, slen))
3284 || (slen << 3) < needents)
5637ef5b
NC
3285 Perl_croak(aTHX_ "panic: swash_fetch got improper swatch, "
3286 "svp=%p, tmps=%p, slen=%"UVuf", needents=%"UVuf,
3287 svp, tmps, (UV)slen, (UV)needents);
a0ed51b3
LW
3288 }
3289
3290 PL_last_swash_hv = hv;
16d8f38a 3291 assert(klen <= sizeof(PL_last_swash_key));
eac04b2e 3292 PL_last_swash_klen = (U8)klen;
cfd0369c
NC
3293 /* FIXME change interpvar.h? */
3294 PL_last_swash_tmps = (U8 *) tmps;
a0ed51b3
LW
3295 PL_last_swash_slen = slen;
3296 if (klen)
3297 Copy(ptr, PL_last_swash_key, klen, U8);
3298 }
3299
9faf8d75 3300 switch ((int)((slen << 3) / needents)) {
a0ed51b3
LW
3301 case 1:
3302 bit = 1 << (off & 7);
3303 off >>= 3;
3304 return (tmps[off] & bit) != 0;
3305 case 8:
3306 return tmps[off];
3307 case 16:
3308 off <<= 1;
3309 return (tmps[off] << 8) + tmps[off + 1] ;
3310 case 32:
3311 off <<= 2;
3312 return (tmps[off] << 24) + (tmps[off+1] << 16) + (tmps[off+2] << 8) + tmps[off + 3] ;
3313 }
5637ef5b
NC
3314 Perl_croak(aTHX_ "panic: swash_fetch got swatch of unexpected bit width, "
3315 "slen=%"UVuf", needents=%"UVuf, (UV)slen, (UV)needents);
670f1322 3316 NORETURN_FUNCTION_END;
a0ed51b3 3317}
2b9d42f0 3318
319009ee
KW
3319/* Read a single line of the main body of the swash input text. These are of
3320 * the form:
3321 * 0053 0056 0073
3322 * where each number is hex. The first two numbers form the minimum and
3323 * maximum of a range, and the third is the value associated with the range.
3324 * Not all swashes should have a third number
3325 *
3326 * On input: l points to the beginning of the line to be examined; it points
3327 * to somewhere in the string of the whole input text, and is
3328 * terminated by a \n or the null string terminator.
3329 * lend points to the null terminator of that string
3330 * wants_value is non-zero if the swash expects a third number
3331 * typestr is the name of the swash's mapping, like 'ToLower'
3332 * On output: *min, *max, and *val are set to the values read from the line.
3333 * returns a pointer just beyond the line examined. If there was no
3334 * valid min number on the line, returns lend+1
3335 */
3336
3337STATIC U8*
3338S_swash_scan_list_line(pTHX_ U8* l, U8* const lend, UV* min, UV* max, UV* val,
3339 const bool wants_value, const U8* const typestr)
3340{
3341 const int typeto = typestr[0] == 'T' && typestr[1] == 'o';
3342 STRLEN numlen; /* Length of the number */
02470786
KW
3343 I32 flags = PERL_SCAN_SILENT_ILLDIGIT
3344 | PERL_SCAN_DISALLOW_PREFIX
3345 | PERL_SCAN_SILENT_NON_PORTABLE;
319009ee
KW
3346
3347 /* nl points to the next \n in the scan */
3348 U8* const nl = (U8*)memchr(l, '\n', lend - l);
3349
3350 /* Get the first number on the line: the range minimum */
3351 numlen = lend - l;
3352 *min = grok_hex((char *)l, &numlen, &flags, NULL);
3353 if (numlen) /* If found a hex number, position past it */
3354 l += numlen;
3355 else if (nl) { /* Else, go handle next line, if any */
3356 return nl + 1; /* 1 is length of "\n" */
3357 }
3358 else { /* Else, no next line */
3359 return lend + 1; /* to LIST's end at which \n is not found */
3360 }
3361
3362 /* The max range value follows, separated by a BLANK */
3363 if (isBLANK(*l)) {
3364 ++l;
02470786
KW
3365 flags = PERL_SCAN_SILENT_ILLDIGIT
3366 | PERL_SCAN_DISALLOW_PREFIX
3367 | PERL_SCAN_SILENT_NON_PORTABLE;
319009ee
KW
3368 numlen = lend - l;
3369 *max = grok_hex((char *)l, &numlen, &flags, NULL);
3370 if (numlen)
3371 l += numlen;
3372 else /* If no value here, it is a single element range */
3373 *max = *min;
3374
3375 /* Non-binary tables have a third entry: what the first element of the
3376 * range maps to */
3377 if (wants_value) {
3378 if (isBLANK(*l)) {
3379 ++l;
f90a9a02
KW
3380
3381 /* The ToLc, etc table mappings are not in hex, and must be
3382 * corrected by adding the code point to them */
3383 if (typeto) {
3384 char *after_strtol = (char *) lend;
3385 *val = Strtol((char *)l, &after_strtol, 10);
3386 l = (U8 *) after_strtol;
f90a9a02
KW
3387 }
3388 else { /* Other tables are in hex, and are the correct result
3389 without tweaking */
a9d188b3
KW
3390 flags = PERL_SCAN_SILENT_ILLDIGIT
3391 | PERL_SCAN_DISALLOW_PREFIX
3392 | PERL_SCAN_SILENT_NON_PORTABLE;
3393 numlen = lend - l;
3394 *val = grok_hex((char *)l, &numlen, &flags, NULL);
3395 if (numlen)
3396 l += numlen;
3397 else
3398 *val = 0;
f90a9a02 3399 }
319009ee
KW
3400 }
3401 else {
3402 *val = 0;
3403 if (typeto) {
dcbac5bb 3404 /* diag_listed_as: To%s: illegal mapping '%s' */
319009ee
KW
3405 Perl_croak(aTHX_ "%s: illegal mapping '%s'",
3406 typestr, l);
3407 }
3408 }
3409 }
3410 else
3411 *val = 0; /* bits == 1, then any val should be ignored */
3412 }
3413 else { /* Nothing following range min, should be single element with no
3414 mapping expected */
3415 *max = *min;
3416 if (wants_value) {
3417 *val = 0;
3418 if (typeto) {
dcbac5bb 3419 /* diag_listed_as: To%s: illegal mapping '%s' */
319009ee
KW
3420 Perl_croak(aTHX_ "%s: illegal mapping '%s'", typestr, l);
3421 }
3422 }
3423 else
3424 *val = 0; /* bits == 1, then val should be ignored */
3425 }
3426
3427 /* Position to next line if any, or EOF */
3428 if (nl)
3429 l = nl + 1;
3430 else
3431 l = lend;
3432
3433 return l;
3434}
3435
979f2922
ST
3436/* Note:
3437 * Returns a swatch (a bit vector string) for a code point sequence
3438 * that starts from the value C<start> and comprises the number C<span>.
3439 * A C<swash> must be an object created by SWASHNEW (see lib/utf8_heavy.pl).
3440 * Should be used via swash_fetch, which will cache the swatch in C<swash>.
3441 */
3442STATIC SV*
b0e3252e 3443S_swatch_get(pTHX_ SV* swash, UV start, UV span)
979f2922
ST
3444{
3445 SV *swatch;
77f9f126 3446 U8 *l, *lend, *x, *xend, *s, *send;
979f2922 3447 STRLEN lcur, xcur, scur;
ef8f7699 3448 HV *const hv = MUTABLE_HV(SvRV(swash));
5c9f4bd2 3449 SV** const invlistsvp = hv_fetchs(hv, "V", FALSE);
36eb48b4 3450
88d45d28
KW
3451 SV** listsvp = NULL; /* The string containing the main body of the table */
3452 SV** extssvp = NULL;
3453 SV** invert_it_svp = NULL;
3454 U8* typestr = NULL;
786861f5
KW
3455 STRLEN bits;
3456 STRLEN octets; /* if bits == 1, then octets == 0 */
3457 UV none;
3458 UV end = start + span;
972dd592 3459
36eb48b4 3460 if (invlistsvp == NULL) {
786861f5
KW
3461 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
3462 SV** const nonesvp = hv_fetchs(hv, "NONE", FALSE);
3463 SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE);
3464 extssvp = hv_fetchs(hv, "EXTRAS", FALSE);
3465 listsvp = hv_fetchs(hv, "LIST", FALSE);
3466 invert_it_svp = hv_fetchs(hv, "INVERT_IT", FALSE);
3467
3468 bits = SvUV(*bitssvp);
3469 none = SvUV(*nonesvp);
3470 typestr = (U8*)SvPV_nolen(*typesvp);
3471 }
36eb48b4
KW
3472 else {
3473 bits = 1;
3474 none = 0;
3475 }
786861f5 3476 octets = bits >> 3; /* if bits == 1, then octets == 0 */
979f2922 3477
b0e3252e 3478 PERL_ARGS_ASSERT_SWATCH_GET;
7918f24d 3479
979f2922 3480 if (bits != 1 && bits != 8 && bits != 16 && bits != 32) {
b0e3252e 3481 Perl_croak(aTHX_ "panic: swatch_get doesn't expect bits %"UVuf,
660a4616 3482 (UV)bits);
979f2922
ST
3483 }
3484
84ea5ef6
KW
3485 /* If overflowed, use the max possible */
3486 if (end < start) {
3487 end = UV_MAX;
3488 span = end - start;
3489 }
3490
979f2922 3491 /* create and initialize $swatch */
979f2922 3492 scur = octets ? (span * octets) : (span + 7) / 8;
e524fe40
NC
3493 swatch = newSV(scur);
3494 SvPOK_on(swatch);
979f2922
ST
3495 s = (U8*)SvPVX(swatch);
3496 if (octets && none) {
0bd48802 3497 const U8* const e = s + scur;
979f2922
ST
3498 while (s < e) {
3499 if (bits == 8)
3500 *s++ = (U8)(none & 0xff);
3501 else if (bits == 16) {
3502 *s++ = (U8)((none >> 8) & 0xff);
3503 *s++ = (U8)( none & 0xff);
3504 }
3505 else if (bits == 32) {
3506 *s++ = (U8)((none >> 24) & 0xff);
3507 *s++ = (U8)((none >> 16) & 0xff);
3508 *s++ = (U8)((none >> 8) & 0xff);
3509 *s++ = (U8)( none & 0xff);
3510 }
3511 }
3512 *s = '\0';
3513 }
3514 else {
3515 (void)memzero((U8*)s, scur + 1);
3516 }
3517 SvCUR_set(swatch, scur);
3518 s = (U8*)SvPVX(swatch);
3519
36eb48b4
KW
3520 if (invlistsvp) { /* If has an inversion list set up use that */
3521 _invlist_populate_swatch(*invlistsvp, start, end, s);
3522 return swatch;
3523 }
3524
3525 /* read $swash->{LIST} */
979f2922
ST
3526 l = (U8*)SvPV(*listsvp, lcur);
3527 lend = l + lcur;
3528 while (l < lend) {
8ed25d53 3529 UV min, max, val, upper;
319009ee
KW
3530 l = S_swash_scan_list_line(aTHX_ l, lend, &min, &max, &val,
3531 cBOOL(octets), typestr);
3532 if (l > lend) {
979f2922
ST
3533 break;
3534 }
3535
972dd592 3536 /* If looking for something beyond this range, go try the next one */
979f2922
ST
3537 if (max < start)
3538 continue;
3539
8ed25d53
KW
3540 /* <end> is generally 1 beyond where we want to set things, but at the
3541 * platform's infinity, where we can't go any higher, we want to
3542 * include the code point at <end> */
3543 upper = (max < end)
3544 ? max
3545 : (max != UV_MAX || end != UV_MAX)
3546 ? end - 1
3547 : end;
3548
979f2922 3549 if (octets) {
35da51f7 3550 UV key;
979f2922
ST
3551 if (min < start) {
3552 if (!none || val < none) {
3553 val += start - min;
3554 }
3555 min = start;
3556 }
8ed25d53 3557 for (key = min; key <= upper; key++) {
979f2922 3558 STRLEN offset;
979f2922
ST
3559 /* offset must be non-negative (start <= min <= key < end) */
3560 offset = octets * (key - start);
3561 if (bits == 8)
3562 s[offset] = (U8)(val & 0xff);
3563 else if (bits == 16) {
3564 s[offset ] = (U8)((val >> 8) & 0xff);
3565 s[offset + 1] = (U8)( val & 0xff);
3566 }
3567 else if (bits == 32) {
3568 s[offset ] = (U8)((val >> 24) & 0xff);
3569 s[offset + 1] = (U8)((val >> 16) & 0xff);
3570 s[offset + 2] = (U8)((val >> 8) & 0xff);
3571 s[offset + 3] = (U8)( val & 0xff);
3572 }
3573
3574 if (!none || val < none)
3575 ++val;
3576 }
3577 }
711a919c 3578 else { /* bits == 1, then val should be ignored */
35da51f7 3579 UV key;
979f2922
ST
3580 if (min < start)
3581 min = start;
6cb05c12 3582
8ed25d53 3583 for (key = min; key <= upper; key++) {
0bd48802 3584 const STRLEN offset = (STRLEN)(key - start);
979f2922
ST
3585 s[offset >> 3] |= 1 << (offset & 7);
3586 }
3587 }
3588 } /* while */
979f2922 3589
9479a769 3590 /* Invert if the data says it should be. Assumes that bits == 1 */
77f9f126 3591 if (invert_it_svp && SvUV(*invert_it_svp)) {
0bda3001
KW
3592
3593 /* Unicode properties should come with all bits above PERL_UNICODE_MAX
3594 * be 0, and their inversion should also be 0, as we don't succeed any
3595 * Unicode property matches for non-Unicode code points */
3596 if (start <= PERL_UNICODE_MAX) {
3597
3598 /* The code below assumes that we never cross the
3599 * Unicode/above-Unicode boundary in a range, as otherwise we would
3600 * have to figure out where to stop flipping the bits. Since this
3601 * boundary is divisible by a large power of 2, and swatches comes
3602 * in small powers of 2, this should be a valid assumption */
3603 assert(start + span - 1 <= PERL_UNICODE_MAX);
3604
507a8485
KW
3605 send = s + scur;
3606 while (s < send) {
3607 *s = ~(*s);
3608 s++;
3609 }
0bda3001 3610 }
77f9f126
KW
3611 }
3612
d73c39c5
KW
3613 /* read $swash->{EXTRAS}
3614 * This code also copied to swash_to_invlist() below */
979f2922
ST
3615 x = (U8*)SvPV(*extssvp, xcur);
3616 xend = x + xcur;
3617 while (x < xend) {
3618 STRLEN namelen;
3619 U8 *namestr;
3620 SV** othersvp;
3621 HV* otherhv;
3622 STRLEN otherbits;
3623 SV **otherbitssvp, *other;
711a919c 3624 U8 *s, *o, *nl;
979f2922
ST
3625 STRLEN slen, olen;
3626
35da51f7 3627 const U8 opc = *x++;
979f2922
ST
3628 if (opc == '\n')
3629 continue;
3630
3631 nl = (U8*)memchr(x, '\n', xend - x);
3632
3633 if (opc != '-' && opc != '+' && opc != '!' && opc != '&') {
3634 if (nl) {
3635 x = nl + 1; /* 1 is length of "\n" */
3636 continue;
3637 }
3638 else {
3639 x = xend; /* to EXTRAS' end at which \n is not found */
3640 break;
3641 }
3642 }
3643
3644 namestr = x;
3645 if (nl) {
3646 namelen = nl - namestr;
3647 x = nl + 1;
3648 }
3649 else {
3650 namelen = xend - namestr;
3651 x = xend;
3652 }
3653
3654 othersvp = hv_fetch(hv, (char *)namestr, namelen, FALSE);
ef8f7699 3655 otherhv = MUTABLE_HV(SvRV(*othersvp));
017a3ce5 3656 otherbitssvp = hv_fetchs(otherhv, "BITS", FALSE);
979f2922
ST
3657 otherbits = (STRLEN)SvUV(*otherbitssvp);
3658 if (bits < otherbits)
5637ef5b
NC
3659 Perl_croak(aTHX_ "panic: swatch_get found swatch size mismatch, "
3660 "bits=%"UVuf", otherbits=%"UVuf, (UV)bits, (UV)otherbits);
979f2922
ST
3661
3662 /* The "other" swatch must be destroyed after. */
b0e3252e 3663 other = swatch_get(*othersvp, start, span);
979f2922
ST
3664 o = (U8*)SvPV(other, olen);
3665
3666 if (!olen)
b0e3252e 3667 Perl_croak(aTHX_ "panic: swatch_get got improper swatch");
979f2922
ST
3668
3669 s = (U8*)SvPV(swatch, slen);
3670 if (bits == 1 && otherbits == 1) {
3671 if (slen != olen)
5637ef5b
NC
3672 Perl_croak(aTHX_ "panic: swatch_get found swatch length "
3673 "mismatch, slen=%"UVuf", olen=%"UVuf,
3674 (UV)slen, (UV)olen);
979f2922
ST
3675
3676 switch (opc) {
3677 case '+':
3678 while (slen--)
3679 *s++ |= *o++;
3680 break;
3681 case '!':
3682 while (slen--)
3683 *s++ |= ~*o++;
3684 break;
3685 case '-':
3686 while (slen--)
3687 *s++ &= ~*o++;
3688 break;
3689 case '&':
3690 while (slen--)
3691 *s++ &= *o++;