This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
regexec.c: Combine adjacent 'ifs' with same clause
[perl5.git] / utf8.c
CommitLineData
a0ed51b3
LW
1/* utf8.c
2 *
1129b882 3 * Copyright (C) 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008
b94e2f88 4 * by Larry Wall and others
a0ed51b3
LW
5 *
6 * You may distribute under the terms of either the GNU General Public
7 * License or the Artistic License, as specified in the README file.
8 *
9 */
10
11/*
4ac71550
TC
12 * 'What a fix!' said Sam. 'That's the one place in all the lands we've ever
13 * heard of that we don't want to see any closer; and that's the one place
14 * we're trying to get to! And that's just where we can't get, nohow.'
15 *
cdad3b53 16 * [p.603 of _The Lord of the Rings_, IV/I: "The Taming of Sméagol"]
a0ed51b3
LW
17 *
18 * 'Well do I understand your speech,' he answered in the same language;
19 * 'yet few strangers do so. Why then do you not speak in the Common Tongue,
4ac71550 20 * as is the custom in the West, if you wish to be answered?'
cdad3b53 21 * --Gandalf, addressing Théoden's door wardens
4ac71550
TC
22 *
23 * [p.508 of _The Lord of the Rings_, III/vi: "The King of the Golden Hall"]
a0ed51b3
LW
24 *
25 * ...the travellers perceived that the floor was paved with stones of many
26 * hues; branching runes and strange devices intertwined beneath their feet.
4ac71550
TC
27 *
28 * [p.512 of _The Lord of the Rings_, III/vi: "The King of the Golden Hall"]
a0ed51b3
LW
29 */
30
31#include "EXTERN.h"
864dbfa3 32#define PERL_IN_UTF8_C
a0ed51b3 33#include "perl.h"
81e983c1 34#include "inline_invlist.c"
a0ed51b3 35
a0c21aa1 36#ifndef EBCDIC
970ea3cb 37/* Separate prototypes needed because in ASCII systems these are
a0c21aa1
JH
38 * usually macros but they still are compiled as code, too. */
39PERL_CALLCONV UV Perl_utf8n_to_uvchr(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen, U32 flags);
6dd9dce9 40PERL_CALLCONV UV Perl_valid_utf8_to_uvchr(pTHX_ const U8 *s, STRLEN *retlen);
a0c21aa1
JH
41PERL_CALLCONV U8* Perl_uvchr_to_utf8(pTHX_ U8 *d, UV uv);
42#endif
43
27da23d5
JH
44static const char unees[] =
45 "Malformed UTF-8 character (unexpected end of string)";
901b21bf 46
48ef279e 47/*
ccfc67b7 48=head1 Unicode Support
a0ed51b3 49
166f8a29
DM
50This file contains various utility functions for manipulating UTF8-encoded
51strings. For the uninitiated, this is a method of representing arbitrary
61296642 52Unicode characters as a variable number of bytes, in such a way that
56da48f7
DM
53characters in the ASCII range are unmodified, and a zero byte never appears
54within non-zero characters.
166f8a29 55
eaf7a4d2
CS
56=cut
57*/
58
59/*
60=for apidoc is_ascii_string
61
a1433954 62Returns true if the first C<len> bytes of the string C<s> are the same whether
970ea3cb
KW
63or not the string is encoded in UTF-8 (or UTF-EBCDIC on EBCDIC machines). That
64is, if they are invariant. On ASCII-ish machines, only ASCII characters
65fit this definition, hence the function's name.
eaf7a4d2 66
9f7e3d64
MH
67If C<len> is 0, it will be calculated using C<strlen(s)>.
68
a1433954 69See also L</is_utf8_string>(), L</is_utf8_string_loclen>(), and L</is_utf8_string_loc>().
eaf7a4d2
CS
70
71=cut
72*/
73
74bool
668b6d8d 75Perl_is_ascii_string(const U8 *s, STRLEN len)
eaf7a4d2
CS
76{
77 const U8* const send = s + (len ? len : strlen((const char *)s));
78 const U8* x = s;
79
80 PERL_ARGS_ASSERT_IS_ASCII_STRING;
eaf7a4d2
CS
81
82 for (; x < send; ++x) {
83 if (!UTF8_IS_INVARIANT(*x))
84 break;
85 }
86
87 return x == send;
88}
89
90/*
87cea99e 91=for apidoc uvuni_to_utf8_flags
eebe1485 92
6ee84de2
KW
93Adds the UTF-8 representation of the code point C<uv> to the end
94of the string C<d>; C<d> should have at least C<UTF8_MAXBYTES+1> free
eebe1485 95bytes available. The return value is the pointer to the byte after the
9041c2e3 96end of the new character. In other words,
eebe1485 97
b851fbc1
JH
98 d = uvuni_to_utf8_flags(d, uv, flags);
99
100or, in most cases,
101
9041c2e3 102 d = uvuni_to_utf8(d, uv);
eebe1485 103
b851fbc1
JH
104(which is equivalent to)
105
106 d = uvuni_to_utf8_flags(d, uv, 0);
107
949cf498 108This is the recommended Unicode-aware way of saying
eebe1485
SC
109
110 *(d++) = uv;
111
949cf498
KW
112This function will convert to UTF-8 (and not warn) even code points that aren't
113legal Unicode or are problematic, unless C<flags> contains one or more of the
a1433954
KW
114following flags:
115
949cf498
KW
116If C<uv> is a Unicode surrogate code point and UNICODE_WARN_SURROGATE is set,
117the function will raise a warning, provided UTF8 warnings are enabled. If instead
118UNICODE_DISALLOW_SURROGATE is set, the function will fail and return NULL.
119If both flags are set, the function will both warn and return NULL.
120
121The UNICODE_WARN_NONCHAR and UNICODE_DISALLOW_NONCHAR flags correspondingly
122affect how the function handles a Unicode non-character. And, likewise for the
123UNICODE_WARN_SUPER and UNICODE_DISALLOW_SUPER flags, and code points that are
124above the Unicode maximum of 0x10FFFF. Code points above 0x7FFF_FFFF (which are
125even less portable) can be warned and/or disallowed even if other above-Unicode
126code points are accepted by the UNICODE_WARN_FE_FF and UNICODE_DISALLOW_FE_FF
127flags.
128
129And finally, the flag UNICODE_WARN_ILLEGAL_INTERCHANGE selects all four of the
130above WARN flags; and UNICODE_DISALLOW_ILLEGAL_INTERCHANGE selects all four
131DISALLOW flags.
132
133
eebe1485
SC
134=cut
135*/
136
dfe13c55 137U8 *
b851fbc1 138Perl_uvuni_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags)
a0ed51b3 139{
7918f24d
NC
140 PERL_ARGS_ASSERT_UVUNI_TO_UTF8_FLAGS;
141
979f77b6
KW
142 /* The first problematic code point is the first surrogate */
143 if (uv >= UNICODE_SURROGATE_FIRST
144 && ckWARN4_d(WARN_UTF8, WARN_SURROGATE, WARN_NON_UNICODE, WARN_NONCHAR))
145 {
949cf498
KW
146 if (UNICODE_IS_SURROGATE(uv)) {
147 if (flags & UNICODE_WARN_SURROGATE) {
8457b38f 148 Perl_ck_warner_d(aTHX_ packWARN(WARN_SURROGATE),
949cf498
KW
149 "UTF-16 surrogate U+%04"UVXf, uv);
150 }
151 if (flags & UNICODE_DISALLOW_SURROGATE) {
152 return NULL;
153 }
154 }
155 else if (UNICODE_IS_SUPER(uv)) {
156 if (flags & UNICODE_WARN_SUPER
157 || (UNICODE_IS_FE_FF(uv) && (flags & UNICODE_WARN_FE_FF)))
158 {
8457b38f 159 Perl_ck_warner_d(aTHX_ packWARN(WARN_NON_UNICODE),
949cf498
KW
160 "Code point 0x%04"UVXf" is not Unicode, may not be portable", uv);
161 }
162 if (flags & UNICODE_DISALLOW_SUPER
163 || (UNICODE_IS_FE_FF(uv) && (flags & UNICODE_DISALLOW_FE_FF)))
164 {
165 return NULL;
166 }
167 }
168 else if (UNICODE_IS_NONCHAR(uv)) {
169 if (flags & UNICODE_WARN_NONCHAR) {
8457b38f 170 Perl_ck_warner_d(aTHX_ packWARN(WARN_NONCHAR),
949cf498
KW
171 "Unicode non-character U+%04"UVXf" is illegal for open interchange",
172 uv);
173 }
174 if (flags & UNICODE_DISALLOW_NONCHAR) {
175 return NULL;
176 }
177 }
507b9800 178 }
c4d5f83a 179 if (UNI_IS_INVARIANT(uv)) {
eb160463 180 *d++ = (U8)UTF_TO_NATIVE(uv);
a0ed51b3
LW
181 return d;
182 }
2d331972 183#if defined(EBCDIC)
1d72bdf6
NIS
184 else {
185 STRLEN len = UNISKIP(uv);
186 U8 *p = d+len-1;
187 while (p > d) {
eb160463 188 *p-- = (U8)UTF_TO_NATIVE((uv & UTF_CONTINUATION_MASK) | UTF_CONTINUATION_MARK);
1d72bdf6
NIS
189 uv >>= UTF_ACCUMULATION_SHIFT;
190 }
eb160463 191 *p = (U8)UTF_TO_NATIVE((uv & UTF_START_MASK(len)) | UTF_START_MARK(len));
1d72bdf6
NIS
192 return d+len;
193 }
194#else /* Non loop style */
a0ed51b3 195 if (uv < 0x800) {
eb160463
GS
196 *d++ = (U8)(( uv >> 6) | 0xc0);
197 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
198 return d;
199 }
200 if (uv < 0x10000) {
eb160463
GS
201 *d++ = (U8)(( uv >> 12) | 0xe0);
202 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
203 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
204 return d;
205 }
206 if (uv < 0x200000) {
eb160463
GS
207 *d++ = (U8)(( uv >> 18) | 0xf0);
208 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
209 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
210 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
211 return d;
212 }
213 if (uv < 0x4000000) {
eb160463
GS
214 *d++ = (U8)(( uv >> 24) | 0xf8);
215 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
216 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
217 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
218 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
219 return d;
220 }
221 if (uv < 0x80000000) {
eb160463
GS
222 *d++ = (U8)(( uv >> 30) | 0xfc);
223 *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80);
224 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
225 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
226 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
227 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
228 return d;
229 }
6b8eaf93 230#ifdef HAS_QUAD
d7578b48 231 if (uv < UTF8_QUAD_MAX)
a0ed51b3
LW
232#endif
233 {
eb160463
GS
234 *d++ = 0xfe; /* Can't match U+FEFF! */
235 *d++ = (U8)(((uv >> 30) & 0x3f) | 0x80);
236 *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80);
237 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
238 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
239 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
240 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
241 return d;
242 }
6b8eaf93 243#ifdef HAS_QUAD
a0ed51b3 244 {
eb160463
GS
245 *d++ = 0xff; /* Can't match U+FFFE! */
246 *d++ = 0x80; /* 6 Reserved bits */
247 *d++ = (U8)(((uv >> 60) & 0x0f) | 0x80); /* 2 Reserved bits */
248 *d++ = (U8)(((uv >> 54) & 0x3f) | 0x80);
249 *d++ = (U8)(((uv >> 48) & 0x3f) | 0x80);
250 *d++ = (U8)(((uv >> 42) & 0x3f) | 0x80);
251 *d++ = (U8)(((uv >> 36) & 0x3f) | 0x80);
252 *d++ = (U8)(((uv >> 30) & 0x3f) | 0x80);
253 *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80);
254 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
255 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
256 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
257 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
258 return d;
259 }
260#endif
1d72bdf6 261#endif /* Loop style */
a0ed51b3 262}
9041c2e3 263
646ca15d
JH
264/*
265
f7d739d1 266Tests if the first C<len> bytes of string C<s> form a valid UTF-8
646ca15d 267character. Note that an INVARIANT (i.e. ASCII) character is a valid
f7d739d1 268UTF-8 character. The number of bytes in the UTF-8 character
646ca15d
JH
269will be returned if it is valid, otherwise 0.
270
271This is the "slow" version as opposed to the "fast" version which is
272the "unrolled" IS_UTF8_CHAR(). E.g. for t/uni/class.t the speed
273difference is a factor of 2 to 3. For lengths (UTF8SKIP(s)) of four
274or less you should use the IS_UTF8_CHAR(), for lengths of five or more
275you should use the _slow(). In practice this means that the _slow()
276will be used very rarely, since the maximum Unicode code point (as of
277Unicode 4.1) is U+10FFFF, which encodes in UTF-8 to four bytes. Only
278the "Perl extended UTF-8" (the infamous 'v-strings') will encode into
279five bytes or more.
280
281=cut */
7af276bc 282PERL_STATIC_INLINE STRLEN
5f66b61c 283S_is_utf8_char_slow(const U8 *s, const STRLEN len)
646ca15d 284{
cd7e6c88 285 dTHX; /* The function called below requires thread context */
646ca15d 286
cd7e6c88 287 STRLEN actual_len;
646ca15d 288
cd7e6c88 289 PERL_ARGS_ASSERT_IS_UTF8_CHAR_SLOW;
646ca15d 290
cd7e6c88 291 utf8n_to_uvuni(s, len, &actual_len, UTF8_CHECK_ONLY);
646ca15d 292
cd7e6c88 293 return (actual_len == (STRLEN) -1) ? 0 : actual_len;
646ca15d 294}
9041c2e3
NIS
295
296/*
492a624f
KW
297=for apidoc is_utf8_char_buf
298
299Returns the number of bytes that comprise the first UTF-8 encoded character in
300buffer C<buf>. C<buf_end> should point to one position beyond the end of the
301buffer. 0 is returned if C<buf> does not point to a complete, valid UTF-8
302encoded character.
303
304Note that an INVARIANT character (i.e. ASCII on non-EBCDIC
305machines) is a valid UTF-8 character.
306
307=cut */
308
309STRLEN
310Perl_is_utf8_char_buf(const U8 *buf, const U8* buf_end)
311{
312
313 STRLEN len;
314
315 PERL_ARGS_ASSERT_IS_UTF8_CHAR_BUF;
316
317 if (buf_end <= buf) {
318 return 0;
319 }
320
321 len = buf_end - buf;
322 if (len > UTF8SKIP(buf)) {
323 len = UTF8SKIP(buf);
324 }
325
326#ifdef IS_UTF8_CHAR
327 if (IS_UTF8_CHAR_FAST(len))
328 return IS_UTF8_CHAR(buf, len) ? len : 0;
329#endif /* #ifdef IS_UTF8_CHAR */
330 return is_utf8_char_slow(buf, len);
331}
332
333/*
87cea99e 334=for apidoc is_utf8_char
eebe1485 335
76848387
KW
336DEPRECATED!
337
5da9da9e 338Tests if some arbitrary number of bytes begins in a valid UTF-8
2bbc8d55
SP
339character. Note that an INVARIANT (i.e. ASCII on non-EBCDIC machines)
340character is a valid UTF-8 character. The actual number of bytes in the UTF-8
341character will be returned if it is valid, otherwise 0.
9041c2e3 342
76848387 343This function is deprecated due to the possibility that malformed input could
a1433954 344cause reading beyond the end of the input buffer. Use L</is_utf8_char_buf>
76848387 345instead.
e0328548 346
82686b01 347=cut */
76848387 348
067a85ef 349STRLEN
668b6d8d 350Perl_is_utf8_char(const U8 *s)
386d01d6 351{
7918f24d 352 PERL_ARGS_ASSERT_IS_UTF8_CHAR;
492a624f 353
76848387 354 /* Assumes we have enough space, which is why this is deprecated */
492a624f 355 return is_utf8_char_buf(s, s + UTF8SKIP(s));
386d01d6
GS
356}
357
eaf7a4d2 358
6662521e 359/*
87cea99e 360=for apidoc is_utf8_string
6662521e 361
a1433954 362Returns true if the first C<len> bytes of string C<s> form a valid
9f7e3d64 363UTF-8 string, false otherwise. If C<len> is 0, it will be calculated
e0328548
KW
364using C<strlen(s)> (which means if you use this option, that C<s> has to have a
365terminating NUL byte). Note that all characters being ASCII constitute 'a
366valid UTF-8 string'.
6662521e 367
a1433954 368See also L</is_ascii_string>(), L</is_utf8_string_loclen>(), and L</is_utf8_string_loc>().
768c67ee 369
6662521e
GS
370=cut
371*/
372
8e84507e 373bool
668b6d8d 374Perl_is_utf8_string(const U8 *s, STRLEN len)
6662521e 375{
35da51f7 376 const U8* const send = s + (len ? len : strlen((const char *)s));
7fc63493 377 const U8* x = s;
067a85ef 378
7918f24d 379 PERL_ARGS_ASSERT_IS_UTF8_STRING;
1aa99e6b 380
6662521e 381 while (x < send) {
1acdb0da 382 /* Inline the easy bits of is_utf8_char() here for speed... */
e0328548
KW
383 if (UTF8_IS_INVARIANT(*x)) {
384 x++;
385 }
1acdb0da
JH
386 else {
387 /* ... and call is_utf8_char() only if really needed. */
e0328548
KW
388 const STRLEN c = UTF8SKIP(x);
389 const U8* const next_char_ptr = x + c;
390
391 if (next_char_ptr > send) {
392 return FALSE;
393 }
394
768c67ee
JH
395 if (IS_UTF8_CHAR_FAST(c)) {
396 if (!IS_UTF8_CHAR(x, c))
e0328548 397 return FALSE;
3c614e38 398 }
e0328548
KW
399 else if (! is_utf8_char_slow(x, c)) {
400 return FALSE;
401 }
402 x = next_char_ptr;
1acdb0da 403 }
6662521e 404 }
768c67ee 405
067a85ef 406 return TRUE;
6662521e
GS
407}
408
67e989fb 409/*
814fafa7
NC
410Implemented as a macro in utf8.h
411
87cea99e 412=for apidoc is_utf8_string_loc
814fafa7 413
a1433954
KW
414Like L</is_utf8_string> but stores the location of the failure (in the
415case of "utf8ness failure") or the location C<s>+C<len> (in the case of
814fafa7
NC
416"utf8ness success") in the C<ep>.
417
a1433954 418See also L</is_utf8_string_loclen>() and L</is_utf8_string>().
814fafa7 419
87cea99e 420=for apidoc is_utf8_string_loclen
81cd54e3 421
a1433954
KW
422Like L</is_utf8_string>() but stores the location of the failure (in the
423case of "utf8ness failure") or the location C<s>+C<len> (in the case of
768c67ee
JH
424"utf8ness success") in the C<ep>, and the number of UTF-8
425encoded characters in the C<el>.
426
a1433954 427See also L</is_utf8_string_loc>() and L</is_utf8_string>().
81cd54e3
JH
428
429=cut
430*/
431
432bool
668b6d8d 433Perl_is_utf8_string_loclen(const U8 *s, STRLEN len, const U8 **ep, STRLEN *el)
81cd54e3 434{
35da51f7 435 const U8* const send = s + (len ? len : strlen((const char *)s));
7fc63493 436 const U8* x = s;
81cd54e3 437 STRLEN c;
3ebfea28 438 STRLEN outlen = 0;
7918f24d
NC
439
440 PERL_ARGS_ASSERT_IS_UTF8_STRING_LOCLEN;
81cd54e3 441
81cd54e3 442 while (x < send) {
e0328548
KW
443 const U8* next_char_ptr;
444
81cd54e3
JH
445 /* Inline the easy bits of is_utf8_char() here for speed... */
446 if (UTF8_IS_INVARIANT(*x))
e0328548 447 next_char_ptr = x + 1;
81cd54e3 448 else {
768c67ee 449 /* ... and call is_utf8_char() only if really needed. */
768c67ee 450 c = UTF8SKIP(x);
e0328548
KW
451 next_char_ptr = c + x;
452 if (next_char_ptr > send) {
453 goto out;
454 }
768c67ee
JH
455 if (IS_UTF8_CHAR_FAST(c)) {
456 if (!IS_UTF8_CHAR(x, c))
457 c = 0;
458 } else
459 c = is_utf8_char_slow(x, c);
768c67ee
JH
460 if (!c)
461 goto out;
81cd54e3 462 }
e0328548 463 x = next_char_ptr;
3ebfea28 464 outlen++;
81cd54e3 465 }
768c67ee
JH
466
467 out:
3ebfea28
AL
468 if (el)
469 *el = outlen;
470
768c67ee
JH
471 if (ep)
472 *ep = x;
3ebfea28 473 return (x == send);
81cd54e3
JH
474}
475
476/*
768c67ee 477
87cea99e 478=for apidoc utf8n_to_uvuni
67e989fb 479
9041c2e3 480Bottom level UTF-8 decode routine.
746afd53
KW
481Returns the code point value of the first character in the string C<s>,
482which is assumed to be in UTF-8 (or UTF-EBCDIC) encoding, and no longer than
483C<curlen> bytes; C<*retlen> (if C<retlen> isn't NULL) will be set to
484the length, in bytes, of that character.
949cf498
KW
485
486The value of C<flags> determines the behavior when C<s> does not point to a
487well-formed UTF-8 character. If C<flags> is 0, when a malformation is found,
524080c4
KW
488zero is returned and C<*retlen> is set so that (S<C<s> + C<*retlen>>) is the
489next possible position in C<s> that could begin a non-malformed character.
490Also, if UTF-8 warnings haven't been lexically disabled, a warning is raised.
949cf498
KW
491
492Various ALLOW flags can be set in C<flags> to allow (and not warn on)
493individual types of malformations, such as the sequence being overlong (that
494is, when there is a shorter sequence that can express the same code point;
495overlong sequences are expressly forbidden in the UTF-8 standard due to
496potential security issues). Another malformation example is the first byte of
497a character not being a legal first byte. See F<utf8.h> for the list of such
524080c4
KW
498flags. For allowed 0 length strings, this function returns 0; for allowed
499overlong sequences, the computed code point is returned; for all other allowed
500malformations, the Unicode REPLACEMENT CHARACTER is returned, as these have no
501determinable reasonable value.
949cf498
KW
502
503The UTF8_CHECK_ONLY flag overrides the behavior when a non-allowed (by other
504flags) malformation is found. If this flag is set, the routine assumes that
505the caller will raise a warning, and this function will silently just set
d088425d
KW
506C<retlen> to C<-1> (cast to C<STRLEN>) and return zero.
507
508Note that this API requires disambiguation between successful decoding a NUL
509character, and an error return (unless the UTF8_CHECK_ONLY flag is set), as
510in both cases, 0 is returned. To disambiguate, upon a zero return, see if the
511first byte of C<s> is 0 as well. If so, the input was a NUL; if not, the input
512had an error.
949cf498
KW
513
514Certain code points are considered problematic. These are Unicode surrogates,
746afd53 515Unicode non-characters, and code points above the Unicode maximum of 0x10FFFF.
949cf498 516By default these are considered regular code points, but certain situations
5eafe189 517warrant special handling for them. If C<flags> contains
949cf498
KW
518UTF8_DISALLOW_ILLEGAL_INTERCHANGE, all three classes are treated as
519malformations and handled as such. The flags UTF8_DISALLOW_SURROGATE,
520UTF8_DISALLOW_NONCHAR, and UTF8_DISALLOW_SUPER (meaning above the legal Unicode
521maximum) can be set to disallow these categories individually.
522
523The flags UTF8_WARN_ILLEGAL_INTERCHANGE, UTF8_WARN_SURROGATE,
524UTF8_WARN_NONCHAR, and UTF8_WARN_SUPER will cause warning messages to be raised
525for their respective categories, but otherwise the code points are considered
526valid (not malformations). To get a category to both be treated as a
527malformation and raise a warning, specify both the WARN and DISALLOW flags.
528(But note that warnings are not raised if lexically disabled nor if
529UTF8_CHECK_ONLY is also specified.)
530
531Very large code points (above 0x7FFF_FFFF) are considered more problematic than
532the others that are above the Unicode legal maximum. There are several
eb83ed87
KW
533reasons: they requre at least 32 bits to represent them on ASCII platforms, are
534not representable at all on EBCDIC platforms, and the original UTF-8
535specification never went above this number (the current 0x10FFFF limit was
536imposed later). (The smaller ones, those that fit into 32 bits, are
537representable by a UV on ASCII platforms, but not by an IV, which means that
538the number of operations that can be performed on them is quite restricted.)
539The UTF-8 encoding on ASCII platforms for these large code points begins with a
540byte containing 0xFE or 0xFF. The UTF8_DISALLOW_FE_FF flag will cause them to
541be treated as malformations, while allowing smaller above-Unicode code points.
542(Of course UTF8_DISALLOW_SUPER will treat all above-Unicode code points,
543including these, as malformations.) Similarly, UTF8_WARN_FE_FF acts just like
544the other WARN flags, but applies just to these code points.
949cf498
KW
545
546All other code points corresponding to Unicode characters, including private
547use and those yet to be assigned, are never considered malformed and never
548warn.
67e989fb 549
ec5f19d0 550Most code should use L</utf8_to_uvchr_buf>() rather than call this directly.
9041c2e3 551
37607a96
PK
552=cut
553*/
67e989fb 554
a0ed51b3 555UV
7fc63493 556Perl_utf8n_to_uvuni(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen, U32 flags)
a0ed51b3 557{
97aff369 558 dVAR;
d4c19fe8 559 const U8 * const s0 = s;
eb83ed87 560 U8 overflow_byte = '\0'; /* Save byte in case of overflow */
0b8d30e8 561 U8 * send;
eb83ed87
KW
562 UV uv = *s;
563 STRLEN expectlen;
949cf498 564 SV* sv = NULL;
eb83ed87
KW
565 UV outlier_ret = 0; /* return value when input is in error or problematic
566 */
567 UV pack_warn = 0; /* Save result of packWARN() for later */
568 bool unexpected_non_continuation = FALSE;
569 bool overflowed = FALSE;
2f8f112e 570 bool do_overlong_test = TRUE; /* May have to skip this test */
a0dbb045 571
eb83ed87 572 const char* const malformed_text = "Malformed UTF-8 character";
7918f24d 573
eb83ed87 574 PERL_ARGS_ASSERT_UTF8N_TO_UVUNI;
a0dbb045 575
eb83ed87
KW
576 /* The order of malformation tests here is important. We should consume as
577 * few bytes as possible in order to not skip any valid character. This is
578 * required by the Unicode Standard (section 3.9 of Unicode 6.0); see also
579 * http://unicode.org/reports/tr36 for more discussion as to why. For
580 * example, once we've done a UTF8SKIP, we can tell the expected number of
581 * bytes, and could fail right off the bat if the input parameters indicate
582 * that there are too few available. But it could be that just that first
583 * byte is garbled, and the intended character occupies fewer bytes. If we
584 * blindly assumed that the first byte is correct, and skipped based on
585 * that number, we could skip over a valid input character. So instead, we
586 * always examine the sequence byte-by-byte.
587 *
588 * We also should not consume too few bytes, otherwise someone could inject
589 * things. For example, an input could be deliberately designed to
590 * overflow, and if this code bailed out immediately upon discovering that,
591 * returning to the caller *retlen pointing to the very next byte (one
592 * which is actually part of of the overflowing sequence), that could look
593 * legitimate to the caller, which could discard the initial partial
594 * sequence and process the rest, inappropriately */
595
596 /* Zero length strings, if allowed, of necessity are zero */
b5b9af04 597 if (UNLIKELY(curlen == 0)) {
eb83ed87
KW
598 if (retlen) {
599 *retlen = 0;
600 }
a0dbb045 601
eb83ed87
KW
602 if (flags & UTF8_ALLOW_EMPTY) {
603 return 0;
604 }
605 if (! (flags & UTF8_CHECK_ONLY)) {
606 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (empty string)", malformed_text));
607 }
0c443dc2
JH
608 goto malformed;
609 }
610
eb83ed87
KW
611 expectlen = UTF8SKIP(s);
612
613 /* A well-formed UTF-8 character, as the vast majority of calls to this
614 * function will be for, has this expected length. For efficiency, set
615 * things up here to return it. It will be overriden only in those rare
616 * cases where a malformation is found */
617 if (retlen) {
618 *retlen = expectlen;
619 }
620
621 /* An invariant is trivially well-formed */
1d72bdf6 622 if (UTF8_IS_INVARIANT(uv)) {
c4d5f83a 623 return (UV) (NATIVE_TO_UTF(*s));
a0ed51b3 624 }
67e989fb 625
eb83ed87 626 /* A continuation character can't start a valid sequence */
b5b9af04 627 if (UNLIKELY(UTF8_IS_CONTINUATION(uv))) {
eb83ed87
KW
628 if (flags & UTF8_ALLOW_CONTINUATION) {
629 if (retlen) {
630 *retlen = 1;
631 }
632 return UNICODE_REPLACEMENT;
633 }
ba210ebe 634
eb83ed87
KW
635 if (! (flags & UTF8_CHECK_ONLY)) {
636 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (unexpected continuation byte 0x%02x, with no preceding start byte)", malformed_text, *s0));
637 }
638 curlen = 1;
ba210ebe
JH
639 goto malformed;
640 }
9041c2e3 641
1d72bdf6 642#ifdef EBCDIC
75383841 643 uv = NATIVE_TO_UTF(uv);
1d72bdf6
NIS
644#endif
645
eb83ed87
KW
646 /* Here is not a continuation byte, nor an invariant. The only thing left
647 * is a start byte (possibly for an overlong) */
ba210ebe 648
eb83ed87
KW
649 /* Remove the leading bits that indicate the number of bytes in the
650 * character's whole UTF-8 sequence, leaving just the bits that are part of
651 * the value */
652 uv &= UTF_START_MASK(expectlen);
ba210ebe 653
eb83ed87
KW
654 /* Now, loop through the remaining bytes in the character's sequence,
655 * accumulating each into the working value as we go. Be sure to not look
656 * past the end of the input string */
0b8d30e8
KW
657 send = (U8*) s0 + ((expectlen <= curlen) ? expectlen : curlen);
658
eb83ed87 659 for (s = s0 + 1; s < send; s++) {
b5b9af04 660 if (LIKELY(UTF8_IS_CONTINUATION(*s))) {
eb83ed87
KW
661#ifndef EBCDIC /* Can't overflow in EBCDIC */
662 if (uv & UTF_ACCUMULATION_OVERFLOW_MASK) {
663
664 /* The original implementors viewed this malformation as more
665 * serious than the others (though I, khw, don't understand
666 * why, since other malformations also give very very wrong
667 * results), so there is no way to turn off checking for it.
668 * Set a flag, but keep going in the loop, so that we absorb
669 * the rest of the bytes that comprise the character. */
670 overflowed = TRUE;
671 overflow_byte = *s; /* Save for warning message's use */
672 }
673#endif
8850bf83 674 uv = UTF8_ACCUMULATE(uv, *s);
eb83ed87
KW
675 }
676 else {
677 /* Here, found a non-continuation before processing all expected
678 * bytes. This byte begins a new character, so quit, even if
679 * allowing this malformation. */
680 unexpected_non_continuation = TRUE;
681 break;
682 }
683 } /* End of loop through the character's bytes */
684
685 /* Save how many bytes were actually in the character */
686 curlen = s - s0;
687
688 /* The loop above finds two types of malformations: non-continuation and/or
689 * overflow. The non-continuation malformation is really a too-short
690 * malformation, as it means that the current character ended before it was
691 * expected to (being terminated prematurely by the beginning of the next
692 * character, whereas in the too-short malformation there just are too few
693 * bytes available to hold the character. In both cases, the check below
694 * that we have found the expected number of bytes would fail if executed.)
695 * Thus the non-continuation malformation is really unnecessary, being a
696 * subset of the too-short malformation. But there may be existing
697 * applications that are expecting the non-continuation type, so we retain
698 * it, and return it in preference to the too-short malformation. (If this
699 * code were being written from scratch, the two types might be collapsed
700 * into one.) I, khw, am also giving priority to returning the
701 * non-continuation and too-short malformations over overflow when multiple
702 * ones are present. I don't know of any real reason to prefer one over
703 * the other, except that it seems to me that multiple-byte errors trumps
704 * errors from a single byte */
b5b9af04 705 if (UNLIKELY(unexpected_non_continuation)) {
eb83ed87
KW
706 if (!(flags & UTF8_ALLOW_NON_CONTINUATION)) {
707 if (! (flags & UTF8_CHECK_ONLY)) {
708 if (curlen == 1) {
709 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (unexpected non-continuation byte 0x%02x, immediately after start byte 0x%02x)", malformed_text, *s, *s0));
710 }
711 else {
712 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (unexpected non-continuation byte 0x%02x, %d bytes after start byte 0x%02x, expected %d bytes)", malformed_text, *s, (int) curlen, *s0, (int)expectlen));
a0dbb045
JH
713 }
714 }
eb83ed87
KW
715 goto malformed;
716 }
717 uv = UNICODE_REPLACEMENT;
2f8f112e
KW
718
719 /* Skip testing for overlongs, as the REPLACEMENT may not be the same
720 * as what the original expectations were. */
721 do_overlong_test = FALSE;
eb83ed87
KW
722 if (retlen) {
723 *retlen = curlen;
724 }
725 }
b5b9af04 726 else if (UNLIKELY(curlen < expectlen)) {
eb83ed87
KW
727 if (! (flags & UTF8_ALLOW_SHORT)) {
728 if (! (flags & UTF8_CHECK_ONLY)) {
729 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (%d byte%s, need %d, after start byte 0x%02x)", malformed_text, (int)curlen, curlen == 1 ? "" : "s", (int)expectlen, *s0));
a0dbb045 730 }
eb83ed87
KW
731 goto malformed;
732 }
733 uv = UNICODE_REPLACEMENT;
2f8f112e 734 do_overlong_test = FALSE;
eb83ed87
KW
735 if (retlen) {
736 *retlen = curlen;
737 }
738 }
739
740#ifndef EBCDIC /* EBCDIC allows FE, FF, can't overflow */
2f8f112e 741 if ((*s0 & 0xFE) == 0xFE /* matches both FE, FF */
eb83ed87
KW
742 && (flags & (UTF8_WARN_FE_FF|UTF8_DISALLOW_FE_FF)))
743 {
744 /* By adding UTF8_CHECK_ONLY to the test, we avoid unnecessary
745 * generation of the sv, since no warnings are raised under CHECK */
746 if ((flags & (UTF8_WARN_FE_FF|UTF8_CHECK_ONLY)) == UTF8_WARN_FE_FF
747 && ckWARN_d(WARN_UTF8))
748 {
42303544
KW
749 /* This message is deliberately not of the same syntax as the other
750 * messages for malformations, for backwards compatibility in the
751 * unlikely event that code is relying on its precise earlier text
752 */
eb83ed87
KW
753 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s Code point beginning with byte 0x%02X is not Unicode, and not portable", malformed_text, *s0));
754 pack_warn = packWARN(WARN_UTF8);
755 }
756 if (flags & UTF8_DISALLOW_FE_FF) {
757 goto malformed;
ba210ebe 758 }
ba210ebe 759 }
b5b9af04 760 if (UNLIKELY(overflowed)) {
ba210ebe 761
eb83ed87
KW
762 /* If the first byte is FF, it will overflow a 32-bit word. If the
763 * first byte is FE, it will overflow a signed 32-bit word. The
764 * above preserves backward compatibility, since its message was used
765 * in earlier versions of this code in preference to overflow */
766 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (overflow at byte 0x%02x, after start byte 0x%02x)", malformed_text, overflow_byte, *s0));
ba210ebe 767 goto malformed;
eb83ed87
KW
768 }
769#endif
770
2f8f112e
KW
771 if (do_overlong_test
772 && expectlen > (STRLEN)UNISKIP(uv)
773 && ! (flags & UTF8_ALLOW_LONG))
774 {
eb83ed87
KW
775 /* The overlong malformation has lower precedence than the others.
776 * Note that if this malformation is allowed, we return the actual
777 * value, instead of the replacement character. This is because this
778 * value is actually well-defined. */
779 if (! (flags & UTF8_CHECK_ONLY)) {
780 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (%d byte%s, need %d, after start byte 0x%02x)", malformed_text, (int)expectlen, expectlen == 1 ? "": "s", UNISKIP(uv), *s0));
781 }
782 goto malformed;
783 }
784
785 /* Here, the input is considered to be well-formed , but could be a
786 * problematic code point that is not allowed by the input parameters. */
787 if (uv >= UNICODE_SURROGATE_FIRST /* isn't problematic if < this */
788 && (flags & (UTF8_DISALLOW_ILLEGAL_INTERCHANGE
789 |UTF8_WARN_ILLEGAL_INTERCHANGE)))
790 {
949cf498 791 if (UNICODE_IS_SURROGATE(uv)) {
eb83ed87
KW
792 if ((flags & (UTF8_WARN_SURROGATE|UTF8_CHECK_ONLY)) == UTF8_WARN_SURROGATE
793 && ckWARN2_d(WARN_UTF8, WARN_SURROGATE))
794 {
111d382d 795 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "UTF-16 surrogate U+%04"UVXf"", uv));
eb83ed87 796 pack_warn = packWARN2(WARN_UTF8, WARN_SURROGATE);
949cf498
KW
797 }
798 if (flags & UTF8_DISALLOW_SURROGATE) {
799 goto disallowed;
800 }
801 }
949cf498 802 else if ((uv > PERL_UNICODE_MAX)) {
eb83ed87
KW
803 if ((flags & (UTF8_WARN_SUPER|UTF8_CHECK_ONLY)) == UTF8_WARN_SUPER
804 && ckWARN2_d(WARN_UTF8, WARN_NON_UNICODE))
805 {
111d382d 806 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "Code point 0x%04"UVXf" is not Unicode, may not be portable", uv));
eb83ed87 807 pack_warn = packWARN2(WARN_UTF8, WARN_NON_UNICODE);
949cf498
KW
808 }
809 if (flags & UTF8_DISALLOW_SUPER) {
810 goto disallowed;
811 }
812 }
4190d317
KW
813 else if (UNICODE_IS_NONCHAR(uv)) {
814 if ((flags & (UTF8_WARN_NONCHAR|UTF8_CHECK_ONLY)) == UTF8_WARN_NONCHAR
815 && ckWARN2_d(WARN_UTF8, WARN_NONCHAR))
816 {
817 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "Unicode non-character U+%04"UVXf" is illegal for open interchange", uv));
818 pack_warn = packWARN2(WARN_UTF8, WARN_NONCHAR);
819 }
820 if (flags & UTF8_DISALLOW_NONCHAR) {
821 goto disallowed;
822 }
823 }
949cf498 824
eb83ed87
KW
825 if (sv) {
826 outlier_ret = uv;
827 goto do_warn;
828 }
829
949cf498
KW
830 /* Here, this is not considered a malformed character, so drop through
831 * to return it */
a0ed51b3 832 }
ba210ebe 833
a0ed51b3 834 return uv;
ba210ebe 835
eb83ed87
KW
836 /* There are three cases which get to beyond this point. In all 3 cases:
837 * <sv> if not null points to a string to print as a warning.
838 * <curlen> is what <*retlen> should be set to if UTF8_CHECK_ONLY isn't
839 * set.
840 * <outlier_ret> is what return value to use if UTF8_CHECK_ONLY isn't set.
841 * This is done by initializing it to 0, and changing it only
842 * for case 1).
843 * The 3 cases are:
844 * 1) The input is valid but problematic, and to be warned about. The
845 * return value is the resultant code point; <*retlen> is set to
846 * <curlen>, the number of bytes that comprise the code point.
847 * <pack_warn> contains the result of packWARN() for the warning
848 * types. The entry point for this case is the label <do_warn>;
849 * 2) The input is a valid code point but disallowed by the parameters to
850 * this function. The return value is 0. If UTF8_CHECK_ONLY is set,
851 * <*relen> is -1; otherwise it is <curlen>, the number of bytes that
852 * comprise the code point. <pack_warn> contains the result of
853 * packWARN() for the warning types. The entry point for this case is
854 * the label <disallowed>.
855 * 3) The input is malformed. The return value is 0. If UTF8_CHECK_ONLY
856 * is set, <*relen> is -1; otherwise it is <curlen>, the number of
857 * bytes that comprise the malformation. All such malformations are
858 * assumed to be warning type <utf8>. The entry point for this case
859 * is the label <malformed>.
860 */
949cf498 861
ba210ebe
JH
862malformed:
863
eb83ed87
KW
864 if (sv && ckWARN_d(WARN_UTF8)) {
865 pack_warn = packWARN(WARN_UTF8);
866 }
867
868disallowed:
869
fcc8fcf6 870 if (flags & UTF8_CHECK_ONLY) {
ba210ebe 871 if (retlen)
10edeb5d 872 *retlen = ((STRLEN) -1);
ba210ebe
JH
873 return 0;
874 }
875
eb83ed87 876do_warn:
5b311467 877
eb83ed87
KW
878 if (pack_warn) { /* <pack_warn> was initialized to 0, and changed only
879 if warnings are to be raised. */
f555bc63 880 const char * const string = SvPVX_const(sv);
a0dbb045 881
f555bc63
KW
882 if (PL_op)
883 Perl_warner(aTHX_ pack_warn, "%s in %s", string, OP_DESC(PL_op));
884 else
885 Perl_warner(aTHX_ pack_warn, "%s", string);
a0dbb045
JH
886 }
887
eb83ed87
KW
888 if (retlen) {
889 *retlen = curlen;
890 }
ba210ebe 891
eb83ed87 892 return outlier_ret;
a0ed51b3
LW
893}
894
8e84507e 895/*
ec5f19d0
KW
896=for apidoc utf8_to_uvchr_buf
897
898Returns the native code point of the first character in the string C<s> which
899is assumed to be in UTF-8 encoding; C<send> points to 1 beyond the end of C<s>.
524080c4 900C<*retlen> will be set to the length, in bytes, of that character.
ec5f19d0 901
524080c4
KW
902If C<s> does not point to a well-formed UTF-8 character and UTF8 warnings are
903enabled, zero is returned and C<*retlen> is set (if C<retlen> isn't
173db420
KW
904NULL) to -1. If those warnings are off, the computed value, if well-defined
905(or the Unicode REPLACEMENT CHARACTER if not), is silently returned, and
906C<*retlen> is set (if C<retlen> isn't NULL) so that (S<C<s> + C<*retlen>>) is
907the next possible position in C<s> that could begin a non-malformed character.
908See L</utf8n_to_uvuni> for details on when the REPLACEMENT CHARACTER is
909returned.
ec5f19d0
KW
910
911=cut
912*/
913
914
915UV
916Perl_utf8_to_uvchr_buf(pTHX_ const U8 *s, const U8 *send, STRLEN *retlen)
917{
918 PERL_ARGS_ASSERT_UTF8_TO_UVCHR_BUF;
919
920 assert(s < send);
921
922 return utf8n_to_uvchr(s, send - s, retlen,
923 ckWARN_d(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
924}
925
27d6c58a 926/* Like L</utf8_to_uvchr_buf>(), but should only be called when it is known that
3986bb7c 927 * there are no malformations in the input UTF-8 string C<s>. surrogates,
6dd9dce9
KW
928 * non-character code points, and non-Unicode code points are allowed. A macro
929 * in utf8.h is used to normally avoid this function wrapper */
27d6c58a
KW
930
931UV
932Perl_valid_utf8_to_uvchr(pTHX_ const U8 *s, STRLEN *retlen)
933{
3986bb7c
KW
934 const UV uv = valid_utf8_to_uvuni(s, retlen);
935
27d6c58a
KW
936 PERL_ARGS_ASSERT_VALID_UTF8_TO_UVCHR;
937
3986bb7c 938 return UNI_TO_NATIVE(uv);
27d6c58a
KW
939}
940
ec5f19d0 941/*
87cea99e 942=for apidoc utf8_to_uvchr
9041c2e3 943
977c1d31
KW
944DEPRECATED!
945
6ee84de2 946Returns the native code point of the first character in the string C<s>
1e54db1a 947which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
9041c2e3
NIS
948length, in bytes, of that character.
949
4b88fb76 950Some, but not all, UTF-8 malformations are detected, and in fact, some
977c1d31
KW
951malformed input could cause reading beyond the end of the input buffer, which
952is why this function is deprecated. Use L</utf8_to_uvchr_buf> instead.
4b88fb76 953
524080c4
KW
954If C<s> points to one of the detected malformations, and UTF8 warnings are
955enabled, zero is returned and C<*retlen> is set (if C<retlen> isn't
956NULL) to -1. If those warnings are off, the computed value if well-defined (or
957the Unicode REPLACEMENT CHARACTER, if not) is silently returned, and C<*retlen>
958is set (if C<retlen> isn't NULL) so that (S<C<s> + C<*retlen>>) is the
959next possible position in C<s> that could begin a non-malformed character.
960See L</utf8n_to_uvuni> for details on when the REPLACEMENT CHARACTER is returned.
9041c2e3
NIS
961
962=cut
963*/
964
965UV
7fc63493 966Perl_utf8_to_uvchr(pTHX_ const U8 *s, STRLEN *retlen)
9041c2e3 967{
7918f24d
NC
968 PERL_ARGS_ASSERT_UTF8_TO_UVCHR;
969
2ff6c191 970 return utf8_to_uvchr_buf(s, s + UTF8_MAXBYTES, retlen);
9041c2e3
NIS
971}
972
973/*
ec5f19d0
KW
974=for apidoc utf8_to_uvuni_buf
975
976Returns the Unicode code point of the first character in the string C<s> which
977is assumed to be in UTF-8 encoding; C<send> points to 1 beyond the end of C<s>.
978C<retlen> will be set to the length, in bytes, of that character.
979
980This function should only be used when the returned UV is considered
981an index into the Unicode semantic tables (e.g. swashes).
982
524080c4
KW
983If C<s> does not point to a well-formed UTF-8 character and UTF8 warnings are
984enabled, zero is returned and C<*retlen> is set (if C<retlen> isn't
985NULL) to -1. If those warnings are off, the computed value if well-defined (or
986the Unicode REPLACEMENT CHARACTER, if not) is silently returned, and C<*retlen>
987is set (if C<retlen> isn't NULL) so that (S<C<s> + C<*retlen>>) is the
988next possible position in C<s> that could begin a non-malformed character.
989See L</utf8n_to_uvuni> for details on when the REPLACEMENT CHARACTER is returned.
ec5f19d0
KW
990
991=cut
992*/
993
994UV
995Perl_utf8_to_uvuni_buf(pTHX_ const U8 *s, const U8 *send, STRLEN *retlen)
996{
997 PERL_ARGS_ASSERT_UTF8_TO_UVUNI_BUF;
998
999 assert(send > s);
1000
1001 /* Call the low level routine asking for checks */
1002 return Perl_utf8n_to_uvuni(aTHX_ s, send -s, retlen,
1003 ckWARN_d(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
1004}
1005
27d6c58a 1006/* Like L</utf8_to_uvuni_buf>(), but should only be called when it is known that
2114036c 1007 * there are no malformations in the input UTF-8 string C<s>. Surrogates,
3986bb7c 1008 * non-character code points, and non-Unicode code points are allowed */
27d6c58a
KW
1009
1010UV
1011Perl_valid_utf8_to_uvuni(pTHX_ const U8 *s, STRLEN *retlen)
1012{
3986bb7c
KW
1013 UV expectlen = UTF8SKIP(s);
1014 const U8* send = s + expectlen;
1015 UV uv = NATIVE_TO_UTF(*s);
1016
27d6c58a
KW
1017 PERL_ARGS_ASSERT_VALID_UTF8_TO_UVUNI;
1018
3986bb7c
KW
1019 if (retlen) {
1020 *retlen = expectlen;
1021 }
1022
1023 /* An invariant is trivially returned */
1024 if (expectlen == 1) {
1025 return uv;
1026 }
1027
1028 /* Remove the leading bits that indicate the number of bytes, leaving just
1029 * the bits that are part of the value */
1030 uv &= UTF_START_MASK(expectlen);
1031
1032 /* Now, loop through the remaining bytes, accumulating each into the
08bc774e
KW
1033 * working total as we go. (I khw tried unrolling the loop for up to 4
1034 * bytes, but there was no performance improvement) */
3986bb7c
KW
1035 for (++s; s < send; s++) {
1036 uv = UTF8_ACCUMULATE(uv, *s);
1037 }
1038
1039 return uv;
27d6c58a
KW
1040}
1041
ec5f19d0 1042/*
87cea99e 1043=for apidoc utf8_to_uvuni
9041c2e3 1044
977c1d31
KW
1045DEPRECATED!
1046
9041c2e3 1047Returns the Unicode code point of the first character in the string C<s>
1e54db1a 1048which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
9041c2e3
NIS
1049length, in bytes, of that character.
1050
39e518fd
KW
1051This function should only be used when the returned UV is considered
1052an index into the Unicode semantic tables (e.g. swashes).
1053
4b88fb76 1054Some, but not all, UTF-8 malformations are detected, and in fact, some
977c1d31
KW
1055malformed input could cause reading beyond the end of the input buffer, which
1056is why this function is deprecated. Use L</utf8_to_uvuni_buf> instead.
9041c2e3 1057
524080c4
KW
1058If C<s> points to one of the detected malformations, and UTF8 warnings are
1059enabled, zero is returned and C<*retlen> is set (if C<retlen> doesn't point to
1060NULL) to -1. If those warnings are off, the computed value if well-defined (or
1061the Unicode REPLACEMENT CHARACTER, if not) is silently returned, and C<*retlen>
1062is set (if C<retlen> isn't NULL) so that (S<C<s> + C<*retlen>>) is the
1063next possible position in C<s> that could begin a non-malformed character.
1064See L</utf8n_to_uvuni> for details on when the REPLACEMENT CHARACTER is returned.
8e84507e
NIS
1065
1066=cut
1067*/
1068
1069UV
7fc63493 1070Perl_utf8_to_uvuni(pTHX_ const U8 *s, STRLEN *retlen)
8e84507e 1071{
7918f24d
NC
1072 PERL_ARGS_ASSERT_UTF8_TO_UVUNI;
1073
4b88fb76 1074 return valid_utf8_to_uvuni(s, retlen);
8e84507e
NIS
1075}
1076
b76347f2 1077/*
87cea99e 1078=for apidoc utf8_length
b76347f2
JH
1079
1080Return the length of the UTF-8 char encoded string C<s> in characters.
02eb7b47
JH
1081Stops at C<e> (inclusive). If C<e E<lt> s> or if the scan would end
1082up past C<e>, croaks.
b76347f2
JH
1083
1084=cut
1085*/
1086
1087STRLEN
35a4481c 1088Perl_utf8_length(pTHX_ const U8 *s, const U8 *e)
b76347f2 1089{
97aff369 1090 dVAR;
b76347f2
JH
1091 STRLEN len = 0;
1092
7918f24d
NC
1093 PERL_ARGS_ASSERT_UTF8_LENGTH;
1094
8850bf83
JH
1095 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g.
1096 * the bitops (especially ~) can create illegal UTF-8.
1097 * In other words: in Perl UTF-8 is not just for Unicode. */
1098
a3b680e6
AL
1099 if (e < s)
1100 goto warn_and_return;
b76347f2 1101 while (s < e) {
4cbf4130 1102 s += UTF8SKIP(s);
8e91ec7f
AV
1103 len++;
1104 }
1105
1106 if (e != s) {
1107 len--;
1108 warn_and_return:
9b387841
NC
1109 if (PL_op)
1110 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8),
1111 "%s in %s", unees, OP_DESC(PL_op));
1112 else
61a12c31 1113 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), "%s", unees);
b76347f2
JH
1114 }
1115
1116 return len;
1117}
1118
b06226ff 1119/*
87cea99e 1120=for apidoc utf8_distance
b06226ff 1121
1e54db1a 1122Returns the number of UTF-8 characters between the UTF-8 pointers C<a>
b06226ff
JH
1123and C<b>.
1124
1125WARNING: use only if you *know* that the pointers point inside the
1126same UTF-8 buffer.
1127
37607a96
PK
1128=cut
1129*/
a0ed51b3 1130
02eb7b47 1131IV
35a4481c 1132Perl_utf8_distance(pTHX_ const U8 *a, const U8 *b)
a0ed51b3 1133{
7918f24d
NC
1134 PERL_ARGS_ASSERT_UTF8_DISTANCE;
1135
bf1665bc 1136 return (a < b) ? -1 * (IV) utf8_length(a, b) : (IV) utf8_length(b, a);
a0ed51b3
LW
1137}
1138
b06226ff 1139/*
87cea99e 1140=for apidoc utf8_hop
b06226ff 1141
8850bf83
JH
1142Return the UTF-8 pointer C<s> displaced by C<off> characters, either
1143forward or backward.
b06226ff
JH
1144
1145WARNING: do not use the following unless you *know* C<off> is within
8850bf83
JH
1146the UTF-8 data pointed to by C<s> *and* that on entry C<s> is aligned
1147on the first byte of character or just after the last byte of a character.
b06226ff 1148
37607a96
PK
1149=cut
1150*/
a0ed51b3
LW
1151
1152U8 *
4373e329 1153Perl_utf8_hop(pTHX_ const U8 *s, I32 off)
a0ed51b3 1154{
7918f24d
NC
1155 PERL_ARGS_ASSERT_UTF8_HOP;
1156
96a5add6 1157 PERL_UNUSED_CONTEXT;
8850bf83
JH
1158 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g
1159 * the bitops (especially ~) can create illegal UTF-8.
1160 * In other words: in Perl UTF-8 is not just for Unicode. */
1161
a0ed51b3
LW
1162 if (off >= 0) {
1163 while (off--)
1164 s += UTF8SKIP(s);
1165 }
1166 else {
1167 while (off++) {
1168 s--;
8850bf83
JH
1169 while (UTF8_IS_CONTINUATION(*s))
1170 s--;
a0ed51b3
LW
1171 }
1172 }
4373e329 1173 return (U8 *)s;
a0ed51b3
LW
1174}
1175
6940069f 1176/*
fed3ba5d
NC
1177=for apidoc bytes_cmp_utf8
1178
a1433954
KW
1179Compares the sequence of characters (stored as octets) in C<b>, C<blen> with the
1180sequence of characters (stored as UTF-8) in C<u>, C<ulen>. Returns 0 if they are
fed3ba5d
NC
1181equal, -1 or -2 if the first string is less than the second string, +1 or +2
1182if the first string is greater than the second string.
1183
1184-1 or +1 is returned if the shorter string was identical to the start of the
1185longer string. -2 or +2 is returned if the was a difference between characters
1186within the strings.
1187
1188=cut
1189*/
1190
1191int
1192Perl_bytes_cmp_utf8(pTHX_ const U8 *b, STRLEN blen, const U8 *u, STRLEN ulen)
1193{
1194 const U8 *const bend = b + blen;
1195 const U8 *const uend = u + ulen;
1196
1197 PERL_ARGS_ASSERT_BYTES_CMP_UTF8;
1198
1199 PERL_UNUSED_CONTEXT;
1200
1201 while (b < bend && u < uend) {
1202 U8 c = *u++;
1203 if (!UTF8_IS_INVARIANT(c)) {
1204 if (UTF8_IS_DOWNGRADEABLE_START(c)) {
1205 if (u < uend) {
1206 U8 c1 = *u++;
1207 if (UTF8_IS_CONTINUATION(c1)) {
356979f4 1208 c = UNI_TO_NATIVE(TWO_BYTE_UTF8_TO_UNI(c, c1));
fed3ba5d
NC
1209 } else {
1210 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8),
1211 "Malformed UTF-8 character "
1212 "(unexpected non-continuation byte 0x%02x"
1213 ", immediately after start byte 0x%02x)"
1214 /* Dear diag.t, it's in the pod. */
1215 "%s%s", c1, c,
1216 PL_op ? " in " : "",
1217 PL_op ? OP_DESC(PL_op) : "");
1218 return -2;
1219 }
1220 } else {
1221 if (PL_op)
1222 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8),
1223 "%s in %s", unees, OP_DESC(PL_op));
1224 else
61a12c31 1225 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), "%s", unees);
fed3ba5d
NC
1226 return -2; /* Really want to return undef :-) */
1227 }
1228 } else {
1229 return -2;
1230 }
1231 }
1232 if (*b != c) {
1233 return *b < c ? -2 : +2;
1234 }
1235 ++b;
1236 }
1237
1238 if (b == bend && u == uend)
1239 return 0;
1240
1241 return b < bend ? +1 : -1;
1242}
1243
1244/*
87cea99e 1245=for apidoc utf8_to_bytes
6940069f 1246
2bbc8d55 1247Converts a string C<s> of length C<len> from UTF-8 into native byte encoding.
a1433954
KW
1248Unlike L</bytes_to_utf8>, this over-writes the original string, and
1249updates C<len> to contain the new length.
67e989fb 1250Returns zero on failure, setting C<len> to -1.
6940069f 1251
a1433954 1252If you need a copy of the string, see L</bytes_from_utf8>.
95be277c 1253
6940069f
GS
1254=cut
1255*/
1256
1257U8 *
37607a96 1258Perl_utf8_to_bytes(pTHX_ U8 *s, STRLEN *len)
6940069f 1259{
d4c19fe8
AL
1260 U8 * const save = s;
1261 U8 * const send = s + *len;
6940069f 1262 U8 *d;
246fae53 1263
7918f24d
NC
1264 PERL_ARGS_ASSERT_UTF8_TO_BYTES;
1265
1e54db1a 1266 /* ensure valid UTF-8 and chars < 256 before updating string */
d4c19fe8 1267 while (s < send) {
dcad2880
JH
1268 U8 c = *s++;
1269
1d72bdf6
NIS
1270 if (!UTF8_IS_INVARIANT(c) &&
1271 (!UTF8_IS_DOWNGRADEABLE_START(c) || (s >= send)
1272 || !(c = *s++) || !UTF8_IS_CONTINUATION(c))) {
10edeb5d 1273 *len = ((STRLEN) -1);
dcad2880
JH
1274 return 0;
1275 }
246fae53 1276 }
dcad2880
JH
1277
1278 d = s = save;
6940069f 1279 while (s < send) {
ed646e6e 1280 STRLEN ulen;
4b88fb76 1281 *d++ = (U8)utf8_to_uvchr_buf(s, send, &ulen);
ed646e6e 1282 s += ulen;
6940069f
GS
1283 }
1284 *d = '\0';
246fae53 1285 *len = d - save;
6940069f
GS
1286 return save;
1287}
1288
1289/*
87cea99e 1290=for apidoc bytes_from_utf8
f9a63242 1291
2bbc8d55 1292Converts a string C<s> of length C<len> from UTF-8 into native byte encoding.
a1433954 1293Unlike L</utf8_to_bytes> but like L</bytes_to_utf8>, returns a pointer to
ef9edfd0
JH
1294the newly-created string, and updates C<len> to contain the new
1295length. Returns the original string if no conversion occurs, C<len>
1296is unchanged. Do nothing if C<is_utf8> points to 0. Sets C<is_utf8> to
2bbc8d55
SP
12970 if C<s> is converted or consisted entirely of characters that are invariant
1298in utf8 (i.e., US-ASCII on non-EBCDIC machines).
f9a63242 1299
37607a96
PK
1300=cut
1301*/
f9a63242
JH
1302
1303U8 *
e1ec3a88 1304Perl_bytes_from_utf8(pTHX_ const U8 *s, STRLEN *len, bool *is_utf8)
f9a63242 1305{
f9a63242 1306 U8 *d;
e1ec3a88
AL
1307 const U8 *start = s;
1308 const U8 *send;
f9a63242
JH
1309 I32 count = 0;
1310
7918f24d
NC
1311 PERL_ARGS_ASSERT_BYTES_FROM_UTF8;
1312
96a5add6 1313 PERL_UNUSED_CONTEXT;
f9a63242 1314 if (!*is_utf8)
73d840c0 1315 return (U8 *)start;
f9a63242 1316
1e54db1a 1317 /* ensure valid UTF-8 and chars < 256 before converting string */
f9a63242 1318 for (send = s + *len; s < send;) {
e1ec3a88 1319 U8 c = *s++;
1d72bdf6 1320 if (!UTF8_IS_INVARIANT(c)) {
db42d148
NIS
1321 if (UTF8_IS_DOWNGRADEABLE_START(c) && s < send &&
1322 (c = *s++) && UTF8_IS_CONTINUATION(c))
1323 count++;
1324 else
73d840c0 1325 return (U8 *)start;
db42d148 1326 }
f9a63242
JH
1327 }
1328
35da51f7 1329 *is_utf8 = FALSE;
f9a63242 1330
212542aa 1331 Newx(d, (*len) - count + 1, U8);
ef9edfd0 1332 s = start; start = d;
f9a63242
JH
1333 while (s < send) {
1334 U8 c = *s++;
c4d5f83a
NIS
1335 if (!UTF8_IS_INVARIANT(c)) {
1336 /* Then it is two-byte encoded */
356979f4 1337 c = UNI_TO_NATIVE(TWO_BYTE_UTF8_TO_UNI(c, *s++));
c4d5f83a
NIS
1338 }
1339 *d++ = c;
f9a63242
JH
1340 }
1341 *d = '\0';
1342 *len = d - start;
73d840c0 1343 return (U8 *)start;
f9a63242
JH
1344}
1345
1346/*
87cea99e 1347=for apidoc bytes_to_utf8
6940069f 1348
ff97e5cf
KW
1349Converts a string C<s> of length C<len> bytes from the native encoding into
1350UTF-8.
6662521e 1351Returns a pointer to the newly-created string, and sets C<len> to
ff97e5cf 1352reflect the new length in bytes.
6940069f 1353
2bbc8d55
SP
1354A NUL character will be written after the end of the string.
1355
1356If you want to convert to UTF-8 from encodings other than
1357the native (Latin1 or EBCDIC),
a1433954 1358see L</sv_recode_to_utf8>().
c9ada85f 1359
497711e7 1360=cut
6940069f
GS
1361*/
1362
c682ebef
FC
1363/* This logic is duplicated in sv_catpvn_flags, so any bug fixes will
1364 likewise need duplication. */
1365
6940069f 1366U8*
35a4481c 1367Perl_bytes_to_utf8(pTHX_ const U8 *s, STRLEN *len)
6940069f 1368{
35a4481c 1369 const U8 * const send = s + (*len);
6940069f
GS
1370 U8 *d;
1371 U8 *dst;
7918f24d
NC
1372
1373 PERL_ARGS_ASSERT_BYTES_TO_UTF8;
96a5add6 1374 PERL_UNUSED_CONTEXT;
6940069f 1375
212542aa 1376 Newx(d, (*len) * 2 + 1, U8);
6940069f
GS
1377 dst = d;
1378
1379 while (s < send) {
35a4481c 1380 const UV uv = NATIVE_TO_ASCII(*s++);
c4d5f83a 1381 if (UNI_IS_INVARIANT(uv))
eb160463 1382 *d++ = (U8)UTF_TO_NATIVE(uv);
6940069f 1383 else {
eb160463
GS
1384 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
1385 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
6940069f
GS
1386 }
1387 }
1388 *d = '\0';
6662521e 1389 *len = d-dst;
6940069f
GS
1390 return dst;
1391}
1392
a0ed51b3 1393/*
dea0fc0b 1394 * Convert native (big-endian) or reversed (little-endian) UTF-16 to UTF-8.
a0ed51b3
LW
1395 *
1396 * Destination must be pre-extended to 3/2 source. Do not use in-place.
1397 * We optimize for native, for obvious reasons. */
1398
1399U8*
dea0fc0b 1400Perl_utf16_to_utf8(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen)
a0ed51b3 1401{
dea0fc0b
JH
1402 U8* pend;
1403 U8* dstart = d;
1404
7918f24d
NC
1405 PERL_ARGS_ASSERT_UTF16_TO_UTF8;
1406
dea0fc0b 1407 if (bytelen & 1)
f5992bc4 1408 Perl_croak(aTHX_ "panic: utf16_to_utf8: odd bytelen %"UVuf, (UV)bytelen);
dea0fc0b
JH
1409
1410 pend = p + bytelen;
1411
a0ed51b3 1412 while (p < pend) {
dea0fc0b
JH
1413 UV uv = (p[0] << 8) + p[1]; /* UTF-16BE */
1414 p += 2;
a0ed51b3 1415 if (uv < 0x80) {
e294cc5d
JH
1416#ifdef EBCDIC
1417 *d++ = UNI_TO_NATIVE(uv);
1418#else
eb160463 1419 *d++ = (U8)uv;
e294cc5d 1420#endif
a0ed51b3
LW
1421 continue;
1422 }
1423 if (uv < 0x800) {
eb160463
GS
1424 *d++ = (U8)(( uv >> 6) | 0xc0);
1425 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
1426 continue;
1427 }
52b9aa85 1428 if (uv >= 0xd800 && uv <= 0xdbff) { /* surrogates */
01ea242b 1429 if (p >= pend) {
dea0fc0b 1430 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
01ea242b
NC
1431 } else {
1432 UV low = (p[0] << 8) + p[1];
1433 p += 2;
52b9aa85 1434 if (low < 0xdc00 || low > 0xdfff)
01ea242b
NC
1435 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
1436 uv = ((uv - 0xd800) << 10) + (low - 0xdc00) + 0x10000;
1437 }
dbde1951
NC
1438 } else if (uv >= 0xdc00 && uv <= 0xdfff) {
1439 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
a0ed51b3
LW
1440 }
1441 if (uv < 0x10000) {
eb160463
GS
1442 *d++ = (U8)(( uv >> 12) | 0xe0);
1443 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
1444 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
1445 continue;
1446 }
1447 else {
eb160463
GS
1448 *d++ = (U8)(( uv >> 18) | 0xf0);
1449 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
1450 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
1451 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
1452 continue;
1453 }
1454 }
dea0fc0b 1455 *newlen = d - dstart;
a0ed51b3
LW
1456 return d;
1457}
1458
1459/* Note: this one is slightly destructive of the source. */
1460
1461U8*
dea0fc0b 1462Perl_utf16_to_utf8_reversed(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen)
a0ed51b3
LW
1463{
1464 U8* s = (U8*)p;
d4c19fe8 1465 U8* const send = s + bytelen;
7918f24d
NC
1466
1467 PERL_ARGS_ASSERT_UTF16_TO_UTF8_REVERSED;
1468
e0ea5e2d
NC
1469 if (bytelen & 1)
1470 Perl_croak(aTHX_ "panic: utf16_to_utf8_reversed: odd bytelen %"UVuf,
1471 (UV)bytelen);
1472
a0ed51b3 1473 while (s < send) {
d4c19fe8 1474 const U8 tmp = s[0];
a0ed51b3
LW
1475 s[0] = s[1];
1476 s[1] = tmp;
1477 s += 2;
1478 }
dea0fc0b 1479 return utf16_to_utf8(p, d, bytelen, newlen);
a0ed51b3
LW
1480}
1481
922e8cb4
KW
1482bool
1483Perl__is_uni_FOO(pTHX_ const U8 classnum, const UV c)
1484{
1485 U8 tmpbuf[UTF8_MAXBYTES+1];
1486 uvchr_to_utf8(tmpbuf, c);
1487 return _is_utf8_FOO(classnum, tmpbuf);
1488}
1489
c3fd2246
KW
1490/* for now these are all defined (inefficiently) in terms of the utf8 versions.
1491 * Note that the macros in handy.h that call these short-circuit calling them
1492 * for Latin-1 range inputs */
a0ed51b3
LW
1493
1494bool
84afefe6 1495Perl_is_uni_alnum(pTHX_ UV c)
a0ed51b3 1496{
89ebb4a3 1497 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1498 uvchr_to_utf8(tmpbuf, c);
922e8cb4 1499 return _is_utf8_FOO(_CC_WORDCHAR, tmpbuf);
a0ed51b3
LW
1500}
1501
1502bool
5092f92a
KW
1503Perl_is_uni_alnumc(pTHX_ UV c)
1504{
1505 U8 tmpbuf[UTF8_MAXBYTES+1];
1506 uvchr_to_utf8(tmpbuf, c);
922e8cb4 1507 return _is_utf8_FOO(_CC_ALPHANUMERIC, tmpbuf);
5092f92a
KW
1508}
1509
61b19385
KW
1510bool /* Internal function so we can deprecate the external one, and call
1511 this one from other deprecated functions in this file */
1512S_is_utf8_idfirst(pTHX_ const U8 *p)
1513{
1514 dVAR;
1515
1516 if (*p == '_')
1517 return TRUE;
1518 /* is_utf8_idstart would be more logical. */
1519 return is_utf8_common(p, &PL_utf8_idstart, "IdStart");
1520}
1521
5092f92a 1522bool
84afefe6 1523Perl_is_uni_idfirst(pTHX_ UV c)
a0ed51b3 1524{
89ebb4a3 1525 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1526 uvchr_to_utf8(tmpbuf, c);
61b19385 1527 return S_is_utf8_idfirst(aTHX_ tmpbuf);
a0ed51b3
LW
1528}
1529
1530bool
f91dcd13
KW
1531Perl__is_uni_perl_idstart(pTHX_ UV c)
1532{
1533 U8 tmpbuf[UTF8_MAXBYTES+1];
1534 uvchr_to_utf8(tmpbuf, c);
1535 return _is_utf8_perl_idstart(tmpbuf);
1536}
1537
1538bool
84afefe6 1539Perl_is_uni_alpha(pTHX_ UV c)
a0ed51b3 1540{
89ebb4a3 1541 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1542 uvchr_to_utf8(tmpbuf, c);
922e8cb4 1543 return _is_utf8_FOO(_CC_ALPHA, tmpbuf);
a0ed51b3
LW
1544}
1545
1546bool
84afefe6 1547Perl_is_uni_ascii(pTHX_ UV c)
4d61ec05 1548{
bc39fe24 1549 return isASCII(c);
4d61ec05
GS
1550}
1551
1552bool
bdd8600f
KW
1553Perl_is_uni_blank(pTHX_ UV c)
1554{
2cafb56b 1555 return isBLANK_uni(c);
bdd8600f
KW
1556}
1557
1558bool
84afefe6 1559Perl_is_uni_space(pTHX_ UV c)
a0ed51b3 1560{
add4123a 1561 return isSPACE_uni(c);
a0ed51b3
LW
1562}
1563
1564bool
84afefe6 1565Perl_is_uni_digit(pTHX_ UV c)
a0ed51b3 1566{
89ebb4a3 1567 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1568 uvchr_to_utf8(tmpbuf, c);
922e8cb4 1569 return _is_utf8_FOO(_CC_DIGIT, tmpbuf);
a0ed51b3
LW
1570}
1571
1572bool
84afefe6 1573Perl_is_uni_upper(pTHX_ UV c)
a0ed51b3 1574{
89ebb4a3 1575 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1576 uvchr_to_utf8(tmpbuf, c);
922e8cb4 1577 return _is_utf8_FOO(_CC_UPPER, tmpbuf);
a0ed51b3
LW
1578}
1579
1580bool
84afefe6 1581Perl_is_uni_lower(pTHX_ UV c)
a0ed51b3 1582{
89ebb4a3 1583 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1584 uvchr_to_utf8(tmpbuf, c);
922e8cb4 1585 return _is_utf8_FOO(_CC_LOWER, tmpbuf);
a0ed51b3
LW
1586}
1587
1588bool
84afefe6 1589Perl_is_uni_cntrl(pTHX_ UV c)
b8c5462f 1590{
7b952154 1591 return isCNTRL_L1(c);
b8c5462f
JH
1592}
1593
1594bool
84afefe6 1595Perl_is_uni_graph(pTHX_ UV c)
b8c5462f 1596{
89ebb4a3 1597 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1598 uvchr_to_utf8(tmpbuf, c);
922e8cb4 1599 return _is_utf8_FOO(_CC_GRAPH, tmpbuf);
b8c5462f
JH
1600}
1601
1602bool
84afefe6 1603Perl_is_uni_print(pTHX_ UV c)
a0ed51b3 1604{
89ebb4a3 1605 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1606 uvchr_to_utf8(tmpbuf, c);
922e8cb4 1607 return _is_utf8_FOO(_CC_PRINT, tmpbuf);
a0ed51b3
LW
1608}
1609
b8c5462f 1610bool
84afefe6 1611Perl_is_uni_punct(pTHX_ UV c)
b8c5462f 1612{
89ebb4a3 1613 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1614 uvchr_to_utf8(tmpbuf, c);
922e8cb4 1615 return _is_utf8_FOO(_CC_PUNCT, tmpbuf);
b8c5462f
JH
1616}
1617
4d61ec05 1618bool
84afefe6 1619Perl_is_uni_xdigit(pTHX_ UV c)
4d61ec05 1620{
4ac6419d 1621 return isXDIGIT_uni(c);
4d61ec05
GS
1622}
1623
3a4c58c9
KW
1624UV
1625Perl__to_upper_title_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp, const char S_or_s)
1626{
1627 /* We have the latin1-range values compiled into the core, so just use
1628 * those, converting the result to utf8. The only difference between upper
1629 * and title case in this range is that LATIN_SMALL_LETTER_SHARP_S is
1630 * either "SS" or "Ss". Which one to use is passed into the routine in
1631 * 'S_or_s' to avoid a test */
1632
1633 UV converted = toUPPER_LATIN1_MOD(c);
1634
1635 PERL_ARGS_ASSERT__TO_UPPER_TITLE_LATIN1;
1636
1637 assert(S_or_s == 'S' || S_or_s == 's');
1638
1639 if (UNI_IS_INVARIANT(converted)) { /* No difference between the two for
1640 characters in this range */
1641 *p = (U8) converted;
1642 *lenp = 1;
1643 return converted;
1644 }
1645
1646 /* toUPPER_LATIN1_MOD gives the correct results except for three outliers,
1647 * which it maps to one of them, so as to only have to have one check for
1648 * it in the main case */
1649 if (UNLIKELY(converted == LATIN_SMALL_LETTER_Y_WITH_DIAERESIS)) {
1650 switch (c) {
1651 case LATIN_SMALL_LETTER_Y_WITH_DIAERESIS:
1652 converted = LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS;
1653 break;
1654 case MICRO_SIGN:
1655 converted = GREEK_CAPITAL_LETTER_MU;
1656 break;
1657 case LATIN_SMALL_LETTER_SHARP_S:
1658 *(p)++ = 'S';
1659 *p = S_or_s;
1660 *lenp = 2;
1661 return 'S';
1662 default:
1663 Perl_croak(aTHX_ "panic: to_upper_title_latin1 did not expect '%c' to map to '%c'", c, LATIN_SMALL_LETTER_Y_WITH_DIAERESIS);
118e2215 1664 assert(0); /* NOTREACHED */
3a4c58c9
KW
1665 }
1666 }
1667
1668 *(p)++ = UTF8_TWO_BYTE_HI(converted);
1669 *p = UTF8_TWO_BYTE_LO(converted);
1670 *lenp = 2;
1671
1672 return converted;
1673}
1674
50bda2c3
KW
1675/* Call the function to convert a UTF-8 encoded character to the specified case.
1676 * Note that there may be more than one character in the result.
1677 * INP is a pointer to the first byte of the input character
1678 * OUTP will be set to the first byte of the string of changed characters. It
1679 * needs to have space for UTF8_MAXBYTES_CASE+1 bytes
1680 * LENP will be set to the length in bytes of the string of changed characters
1681 *
1682 * The functions return the ordinal of the first character in the string of OUTP */
f90a9a02
KW
1683#define CALL_UPPER_CASE(INP, OUTP, LENP) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_toupper, "ToUc", "utf8::ToSpecUc")
1684#define CALL_TITLE_CASE(INP, OUTP, LENP) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_totitle, "ToTc", "utf8::ToSpecTc")
1685#define CALL_LOWER_CASE(INP, OUTP, LENP) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_tolower, "ToLc", "utf8::ToSpecLc")
50bda2c3
KW
1686
1687/* This additionally has the input parameter SPECIALS, which if non-zero will
1688 * cause this to use the SPECIALS hash for folding (meaning get full case
1689 * folding); otherwise, when zero, this implies a simple case fold */
f90a9a02 1690#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 1691
84afefe6
JH
1692UV
1693Perl_to_uni_upper(pTHX_ UV c, U8* p, STRLEN *lenp)
a0ed51b3 1694{
3a4c58c9
KW
1695 dVAR;
1696
a1433954
KW
1697 /* Convert the Unicode character whose ordinal is <c> to its uppercase
1698 * version and store that in UTF-8 in <p> and its length in bytes in <lenp>.
1699 * Note that the <p> needs to be at least UTF8_MAXBYTES_CASE+1 bytes since
c3fd2246
KW
1700 * the changed version may be longer than the original character.
1701 *
1702 * The ordinal of the first character of the changed version is returned
1703 * (but note, as explained above, that there may be more.) */
1704
7918f24d
NC
1705 PERL_ARGS_ASSERT_TO_UNI_UPPER;
1706
3a4c58c9
KW
1707 if (c < 256) {
1708 return _to_upper_title_latin1((U8) c, p, lenp, 'S');
1709 }
1710
0ebc6274 1711 uvchr_to_utf8(p, c);
3a4c58c9 1712 return CALL_UPPER_CASE(p, p, lenp);
a0ed51b3
LW
1713}
1714
84afefe6
JH
1715UV
1716Perl_to_uni_title(pTHX_ UV c, U8* p, STRLEN *lenp)
a0ed51b3 1717{
3a4c58c9
KW
1718 dVAR;
1719
7918f24d
NC
1720 PERL_ARGS_ASSERT_TO_UNI_TITLE;
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_TITLE_CASE(p, p, lenp);
a0ed51b3
LW
1728}
1729
afc16117
KW
1730STATIC U8
1731S_to_lower_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp)
1732{
1733 /* We have the latin1-range values compiled into the core, so just use
1734 * those, converting the result to utf8. Since the result is always just
a1433954 1735 * one character, we allow <p> to be NULL */
afc16117
KW
1736
1737 U8 converted = toLOWER_LATIN1(c);
1738
1739 if (p != NULL) {
1740 if (UNI_IS_INVARIANT(converted)) {
1741 *p = converted;
1742 *lenp = 1;
1743 }
1744 else {
1745 *p = UTF8_TWO_BYTE_HI(converted);
1746 *(p+1) = UTF8_TWO_BYTE_LO(converted);
1747 *lenp = 2;
1748 }
1749 }
1750 return converted;
1751}
1752
84afefe6
JH
1753UV
1754Perl_to_uni_lower(pTHX_ UV c, U8* p, STRLEN *lenp)
a0ed51b3 1755{
968c5e6a
KW
1756 dVAR;
1757
7918f24d
NC
1758 PERL_ARGS_ASSERT_TO_UNI_LOWER;
1759
afc16117
KW
1760 if (c < 256) {
1761 return to_lower_latin1((U8) c, p, lenp);
bca00c02
KW
1762 }
1763
afc16117 1764 uvchr_to_utf8(p, c);
968c5e6a 1765 return CALL_LOWER_CASE(p, p, lenp);
a0ed51b3
LW
1766}
1767
84afefe6 1768UV
f673fad4 1769Perl__to_fold_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp, const bool flags)
a1dde8de 1770{
a1433954 1771 /* Corresponds to to_lower_latin1(), <flags> is TRUE if to use full case
f673fad4
KW
1772 * folding */
1773
a1dde8de
KW
1774 UV converted;
1775
1776 PERL_ARGS_ASSERT__TO_FOLD_LATIN1;
1777
1778 if (c == MICRO_SIGN) {
1779 converted = GREEK_SMALL_LETTER_MU;
1780 }
1781 else if (flags && c == LATIN_SMALL_LETTER_SHARP_S) {
1782 *(p)++ = 's';
1783 *p = 's';
1784 *lenp = 2;
1785 return 's';
1786 }
1787 else { /* In this range the fold of all other characters is their lower
1788 case */
1789 converted = toLOWER_LATIN1(c);
1790 }
1791
1792 if (UNI_IS_INVARIANT(converted)) {
1793 *p = (U8) converted;
1794 *lenp = 1;
1795 }
1796 else {
1797 *(p)++ = UTF8_TWO_BYTE_HI(converted);
1798 *p = UTF8_TWO_BYTE_LO(converted);
1799 *lenp = 2;
1800 }
1801
1802 return converted;
1803}
1804
1805UV
a0270393 1806Perl__to_uni_fold_flags(pTHX_ UV c, U8* p, STRLEN *lenp, const U8 flags)
84afefe6 1807{
4b593389 1808
a0270393
KW
1809 /* Not currently externally documented, and subject to change
1810 * <flags> bits meanings:
1811 * FOLD_FLAGS_FULL iff full folding is to be used;
1812 * FOLD_FLAGS_LOCALE iff in locale
1813 * FOLD_FLAGS_NOMIX_ASCII iff non-ASCII to ASCII folds are prohibited
1814 */
4b593389 1815
36bb2ab6 1816 PERL_ARGS_ASSERT__TO_UNI_FOLD_FLAGS;
7918f24d 1817
a1dde8de 1818 if (c < 256) {
a0270393
KW
1819 UV result = _to_fold_latin1((U8) c, p, lenp,
1820 cBOOL(((flags & FOLD_FLAGS_FULL)
1821 /* If ASCII-safe, don't allow full folding,
1822 * as that could include SHARP S => ss;
1823 * otherwise there is no crossing of
1824 * ascii/non-ascii in the latin1 range */
1825 && ! (flags & FOLD_FLAGS_NOMIX_ASCII))));
1826 /* It is illegal for the fold to cross the 255/256 boundary under
1827 * locale; in this case return the original */
1828 return (result > 256 && flags & FOLD_FLAGS_LOCALE)
1829 ? c
1830 : result;
a1dde8de
KW
1831 }
1832
a0270393
KW
1833 /* If no special needs, just use the macro */
1834 if ( ! (flags & (FOLD_FLAGS_LOCALE|FOLD_FLAGS_NOMIX_ASCII))) {
1835 uvchr_to_utf8(p, c);
1836 return CALL_FOLD_CASE(p, p, lenp, flags & FOLD_FLAGS_FULL);
1837 }
1838 else { /* Otherwise, _to_utf8_fold_flags has the intelligence to deal with
1839 the special flags. */
1840 U8 utf8_c[UTF8_MAXBYTES + 1];
1841 uvchr_to_utf8(utf8_c, c);
1842 return _to_utf8_fold_flags(utf8_c, p, lenp, flags, NULL);
1843 }
84afefe6
JH
1844}
1845
a0ed51b3 1846bool
84afefe6 1847Perl_is_uni_alnum_lc(pTHX_ UV c)
a0ed51b3 1848{
edfb3318
KW
1849 if (c < 256) {
1850 return isALNUM_LC(UNI_TO_NATIVE(c));
1851 }
922e8cb4 1852 return _is_uni_FOO(_CC_WORDCHAR, c);
a0ed51b3
LW
1853}
1854
1855bool
5092f92a
KW
1856Perl_is_uni_alnumc_lc(pTHX_ UV c)
1857{
1858 if (c < 256) {
15861f94 1859 return isALPHANUMERIC_LC(UNI_TO_NATIVE(c));
5092f92a 1860 }
922e8cb4 1861 return _is_uni_FOO(_CC_ALPHANUMERIC, c);
5092f92a
KW
1862}
1863
1864bool
84afefe6 1865Perl_is_uni_idfirst_lc(pTHX_ UV c)
a0ed51b3 1866{
edfb3318
KW
1867 if (c < 256) {
1868 return isIDFIRST_LC(UNI_TO_NATIVE(c));
1869 }
1870 return _is_uni_perl_idstart(c);
a0ed51b3
LW
1871}
1872
1873bool
84afefe6 1874Perl_is_uni_alpha_lc(pTHX_ UV c)
a0ed51b3 1875{
edfb3318
KW
1876 if (c < 256) {
1877 return isALPHA_LC(UNI_TO_NATIVE(c));
1878 }
922e8cb4 1879 return _is_uni_FOO(_CC_ALPHA, c);
a0ed51b3
LW
1880}
1881
1882bool
84afefe6 1883Perl_is_uni_ascii_lc(pTHX_ UV c)
4d61ec05 1884{
edfb3318
KW
1885 if (c < 256) {
1886 return isASCII_LC(UNI_TO_NATIVE(c));
1887 }
1888 return 0;
4d61ec05
GS
1889}
1890
1891bool
bdd8600f
KW
1892Perl_is_uni_blank_lc(pTHX_ UV c)
1893{
edfb3318
KW
1894 if (c < 256) {
1895 return isBLANK_LC(UNI_TO_NATIVE(c));
1896 }
61b19385 1897 return isBLANK_uni(c);
bdd8600f
KW
1898}
1899
1900bool
84afefe6 1901Perl_is_uni_space_lc(pTHX_ UV c)
a0ed51b3 1902{
edfb3318
KW
1903 if (c < 256) {
1904 return isSPACE_LC(UNI_TO_NATIVE(c));
1905 }
61b19385 1906 return isSPACE_uni(c);
a0ed51b3
LW
1907}
1908
1909bool
84afefe6 1910Perl_is_uni_digit_lc(pTHX_ UV c)
a0ed51b3 1911{
edfb3318
KW
1912 if (c < 256) {
1913 return isDIGIT_LC(UNI_TO_NATIVE(c));
1914 }
922e8cb4 1915 return _is_uni_FOO(_CC_DIGIT, c);
a0ed51b3
LW
1916}
1917
1918bool
84afefe6 1919Perl_is_uni_upper_lc(pTHX_ UV c)
a0ed51b3 1920{
edfb3318
KW
1921 if (c < 256) {
1922 return isUPPER_LC(UNI_TO_NATIVE(c));
1923 }
922e8cb4 1924 return _is_uni_FOO(_CC_UPPER, c);
a0ed51b3
LW
1925}
1926
1927bool
84afefe6 1928Perl_is_uni_lower_lc(pTHX_ UV c)
a0ed51b3 1929{
edfb3318
KW
1930 if (c < 256) {
1931 return isLOWER_LC(UNI_TO_NATIVE(c));
1932 }
922e8cb4 1933 return _is_uni_FOO(_CC_LOWER, c);
a0ed51b3
LW
1934}
1935
1936bool
84afefe6 1937Perl_is_uni_cntrl_lc(pTHX_ UV c)
b8c5462f 1938{
edfb3318
KW
1939 if (c < 256) {
1940 return isCNTRL_LC(UNI_TO_NATIVE(c));
1941 }
61b19385 1942 return 0;
b8c5462f
JH
1943}
1944
1945bool
84afefe6 1946Perl_is_uni_graph_lc(pTHX_ UV c)
b8c5462f 1947{
edfb3318
KW
1948 if (c < 256) {
1949 return isGRAPH_LC(UNI_TO_NATIVE(c));
1950 }
922e8cb4 1951 return _is_uni_FOO(_CC_GRAPH, c);
b8c5462f
JH
1952}
1953
1954bool
84afefe6 1955Perl_is_uni_print_lc(pTHX_ UV c)
a0ed51b3 1956{
edfb3318
KW
1957 if (c < 256) {
1958 return isPRINT_LC(UNI_TO_NATIVE(c));
1959 }
922e8cb4 1960 return _is_uni_FOO(_CC_PRINT, c);
a0ed51b3
LW
1961}
1962
b8c5462f 1963bool
84afefe6 1964Perl_is_uni_punct_lc(pTHX_ UV c)
b8c5462f 1965{
edfb3318
KW
1966 if (c < 256) {
1967 return isPUNCT_LC(UNI_TO_NATIVE(c));
1968 }
922e8cb4 1969 return _is_uni_FOO(_CC_PUNCT, c);
b8c5462f
JH
1970}
1971
4d61ec05 1972bool
84afefe6 1973Perl_is_uni_xdigit_lc(pTHX_ UV c)
4d61ec05 1974{
edfb3318
KW
1975 if (c < 256) {
1976 return isXDIGIT_LC(UNI_TO_NATIVE(c));
1977 }
61b19385 1978 return isXDIGIT_uni(c);
4d61ec05
GS
1979}
1980
b7ac61fa
JH
1981U32
1982Perl_to_uni_upper_lc(pTHX_ U32 c)
1983{
ee099d14
JH
1984 /* XXX returns only the first character -- do not use XXX */
1985 /* XXX no locale support yet */
1986 STRLEN len;
89ebb4a3 1987 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
ee099d14 1988 return (U32)to_uni_upper(c, tmpbuf, &len);
b7ac61fa
JH
1989}
1990
1991U32
1992Perl_to_uni_title_lc(pTHX_ U32 c)
1993{
ee099d14
JH
1994 /* XXX returns only the first character XXX -- do not use XXX */
1995 /* XXX no locale support yet */
1996 STRLEN len;
89ebb4a3 1997 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
ee099d14 1998 return (U32)to_uni_title(c, tmpbuf, &len);
b7ac61fa
JH
1999}
2000
2001U32
2002Perl_to_uni_lower_lc(pTHX_ U32 c)
2003{
ee099d14
JH
2004 /* XXX returns only the first character -- do not use XXX */
2005 /* XXX no locale support yet */
2006 STRLEN len;
89ebb4a3 2007 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
ee099d14 2008 return (U32)to_uni_lower(c, tmpbuf, &len);
b7ac61fa
JH
2009}
2010
26483009 2011PERL_STATIC_INLINE bool
5141f98e 2012S_is_utf8_common(pTHX_ const U8 *const p, SV **swash,
bde6a22d
NC
2013 const char *const swashname)
2014{
ea317ccb
KW
2015 /* returns a boolean giving whether or not the UTF8-encoded character that
2016 * starts at <p> is in the swash indicated by <swashname>. <swash>
2017 * contains a pointer to where the swash indicated by <swashname>
2018 * is to be stored; which this routine will do, so that future calls will
2019 * look at <*swash> and only generate a swash if it is not null
2020 *
2021 * Note that it is assumed that the buffer length of <p> is enough to
2022 * contain all the bytes that comprise the character. Thus, <*p> should
2023 * have been checked before this call for mal-formedness enough to assure
2024 * that. */
2025
97aff369 2026 dVAR;
7918f24d
NC
2027
2028 PERL_ARGS_ASSERT_IS_UTF8_COMMON;
2029
492a624f
KW
2030 /* The API should have included a length for the UTF-8 character in <p>,
2031 * but it doesn't. We therefor assume that p has been validated at least
2032 * as far as there being enough bytes available in it to accommodate the
2033 * character without reading beyond the end, and pass that number on to the
2034 * validating routine */
2035 if (!is_utf8_char_buf(p, p + UTF8SKIP(p)))
bde6a22d 2036 return FALSE;
87367d5f
KW
2037 if (!*swash) {
2038 U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST;
2039 *swash = _core_swash_init("utf8", swashname, &PL_sv_undef, 1, 0, NULL, &flags);
2040 }
bde6a22d
NC
2041 return swash_fetch(*swash, p, TRUE) != 0;
2042}
2043
2044bool
922e8cb4
KW
2045Perl__is_utf8_FOO(pTHX_ const U8 classnum, const U8 *p)
2046{
2047 dVAR;
2048
2049 PERL_ARGS_ASSERT__IS_UTF8_FOO;
2050
2051 assert(classnum < _FIRST_NON_SWASH_CC);
2052
2053 return is_utf8_common(p, &PL_utf8_swash_ptrs[classnum], swash_property_names[classnum]);
2054}
2055
2056bool
7fc63493 2057Perl_is_utf8_alnum(pTHX_ const U8 *p)
a0ed51b3 2058{
97aff369 2059 dVAR;
7918f24d
NC
2060
2061 PERL_ARGS_ASSERT_IS_UTF8_ALNUM;
2062
671c33bf
NC
2063 /* NOTE: "IsWord", not "IsAlnum", since Alnum is a true
2064 * descendant of isalnum(3), in other words, it doesn't
2065 * contain the '_'. --jhi */
03940dc2 2066 return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_WORDCHAR], "IsWord");
a0ed51b3
LW
2067}
2068
2069bool
5092f92a
KW
2070Perl_is_utf8_alnumc(pTHX_ const U8 *p)
2071{
2072 dVAR;
2073
2074 PERL_ARGS_ASSERT_IS_UTF8_ALNUMC;
2075
03940dc2 2076 return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_ALPHANUMERIC], "IsAlnum");
5092f92a
KW
2077}
2078
2079bool
7fc63493 2080Perl_is_utf8_idfirst(pTHX_ const U8 *p) /* The naming is historical. */
a0ed51b3 2081{
97aff369 2082 dVAR;
7918f24d
NC
2083
2084 PERL_ARGS_ASSERT_IS_UTF8_IDFIRST;
2085
61b19385 2086 return S_is_utf8_idfirst(aTHX_ p);
82686b01
JH
2087}
2088
2089bool
c11ff943
KW
2090Perl_is_utf8_xidfirst(pTHX_ const U8 *p) /* The naming is historical. */
2091{
2092 dVAR;
2093
2094 PERL_ARGS_ASSERT_IS_UTF8_XIDFIRST;
2095
2096 if (*p == '_')
2097 return TRUE;
2098 /* is_utf8_idstart would be more logical. */
2099 return is_utf8_common(p, &PL_utf8_xidstart, "XIdStart");
2100}
2101
2102bool
d65654cb 2103Perl__is_utf8_perl_idstart(pTHX_ const U8 *p)
b6912c02
KW
2104{
2105 dVAR;
2106
d65654cb 2107 PERL_ARGS_ASSERT__IS_UTF8_PERL_IDSTART;
b6912c02
KW
2108
2109 return is_utf8_common(p, &PL_utf8_perl_idstart, "_Perl_IDStart");
2110}
2111
2112bool
7fc63493 2113Perl_is_utf8_idcont(pTHX_ const U8 *p)
82686b01 2114{
97aff369 2115 dVAR;
7918f24d
NC
2116
2117 PERL_ARGS_ASSERT_IS_UTF8_IDCONT;
2118
d4c19fe8 2119 return is_utf8_common(p, &PL_utf8_idcont, "IdContinue");
a0ed51b3
LW
2120}
2121
2122bool
c11ff943
KW
2123Perl_is_utf8_xidcont(pTHX_ const U8 *p)
2124{
2125 dVAR;
2126
2127 PERL_ARGS_ASSERT_IS_UTF8_XIDCONT;
2128
c11ff943
KW
2129 return is_utf8_common(p, &PL_utf8_idcont, "XIdContinue");
2130}
2131
2132bool
7fc63493 2133Perl_is_utf8_alpha(pTHX_ const U8 *p)
a0ed51b3 2134{
97aff369 2135 dVAR;
7918f24d
NC
2136
2137 PERL_ARGS_ASSERT_IS_UTF8_ALPHA;
2138
03940dc2 2139 return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_ALPHA], "IsAlpha");
a0ed51b3
LW
2140}
2141
2142bool
7fc63493 2143Perl_is_utf8_ascii(pTHX_ const U8 *p)
b8c5462f 2144{
97aff369 2145 dVAR;
7918f24d
NC
2146
2147 PERL_ARGS_ASSERT_IS_UTF8_ASCII;
2148
bc39fe24
KW
2149 /* ASCII characters are the same whether in utf8 or not. So the macro
2150 * works on both utf8 and non-utf8 representations. */
2151 return isASCII(*p);
b8c5462f
JH
2152}
2153
2154bool
bdd8600f
KW
2155Perl_is_utf8_blank(pTHX_ const U8 *p)
2156{
2157 dVAR;
2158
2159 PERL_ARGS_ASSERT_IS_UTF8_BLANK;
2160
2cafb56b 2161 return isBLANK_utf8(p);
bdd8600f
KW
2162}
2163
2164bool
7fc63493 2165Perl_is_utf8_space(pTHX_ const U8 *p)
a0ed51b3 2166{
97aff369 2167 dVAR;
7918f24d
NC
2168
2169 PERL_ARGS_ASSERT_IS_UTF8_SPACE;
2170
add4123a 2171 return isSPACE_utf8(p);
a0ed51b3
LW
2172}
2173
2174bool
d1eb3177
YO
2175Perl_is_utf8_perl_space(pTHX_ const U8 *p)
2176{
2177 dVAR;
2178
2179 PERL_ARGS_ASSERT_IS_UTF8_PERL_SPACE;
2180
c4428693
KW
2181 /* Only true if is an ASCII space-like character, and ASCII is invariant
2182 * under utf8, so can just use the macro */
2183 return isSPACE_A(*p);
d1eb3177
YO
2184}
2185
2186bool
2187Perl_is_utf8_perl_word(pTHX_ const U8 *p)
2188{
2189 dVAR;
2190
2191 PERL_ARGS_ASSERT_IS_UTF8_PERL_WORD;
2192
c4428693
KW
2193 /* Only true if is an ASCII word character, and ASCII is invariant
2194 * under utf8, so can just use the macro */
2195 return isWORDCHAR_A(*p);
d1eb3177
YO
2196}
2197
2198bool
7fc63493 2199Perl_is_utf8_digit(pTHX_ const U8 *p)
a0ed51b3 2200{
97aff369 2201 dVAR;
7918f24d
NC
2202
2203 PERL_ARGS_ASSERT_IS_UTF8_DIGIT;
2204
03940dc2 2205 return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_DIGIT], "IsDigit");
a0ed51b3
LW
2206}
2207
2208bool
d1eb3177
YO
2209Perl_is_utf8_posix_digit(pTHX_ const U8 *p)
2210{
2211 dVAR;
2212
2213 PERL_ARGS_ASSERT_IS_UTF8_POSIX_DIGIT;
2214
c4428693
KW
2215 /* Only true if is an ASCII digit character, and ASCII is invariant
2216 * under utf8, so can just use the macro */
2217 return isDIGIT_A(*p);
d1eb3177
YO
2218}
2219
2220bool
7fc63493 2221Perl_is_utf8_upper(pTHX_ const U8 *p)
a0ed51b3 2222{
97aff369 2223 dVAR;
7918f24d
NC
2224
2225 PERL_ARGS_ASSERT_IS_UTF8_UPPER;
2226
03940dc2 2227 return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_UPPER], "IsUppercase");
a0ed51b3
LW
2228}
2229
2230bool
7fc63493 2231Perl_is_utf8_lower(pTHX_ const U8 *p)
a0ed51b3 2232{
97aff369 2233 dVAR;
7918f24d
NC
2234
2235 PERL_ARGS_ASSERT_IS_UTF8_LOWER;
2236
03940dc2 2237 return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_LOWER], "IsLowercase");
a0ed51b3
LW
2238}
2239
2240bool
7fc63493 2241Perl_is_utf8_cntrl(pTHX_ const U8 *p)
b8c5462f 2242{
97aff369 2243 dVAR;
7918f24d
NC
2244
2245 PERL_ARGS_ASSERT_IS_UTF8_CNTRL;
2246
a35d759a 2247 return isCNTRL_utf8(p);
b8c5462f
JH
2248}
2249
2250bool
7fc63493 2251Perl_is_utf8_graph(pTHX_ const U8 *p)
b8c5462f 2252{
97aff369 2253 dVAR;
7918f24d
NC
2254
2255 PERL_ARGS_ASSERT_IS_UTF8_GRAPH;
2256
03940dc2 2257 return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_GRAPH], "IsGraph");
b8c5462f
JH
2258}
2259
2260bool
7fc63493 2261Perl_is_utf8_print(pTHX_ const U8 *p)
a0ed51b3 2262{
97aff369 2263 dVAR;
7918f24d
NC
2264
2265 PERL_ARGS_ASSERT_IS_UTF8_PRINT;
2266
03940dc2 2267 return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_PRINT], "IsPrint");
a0ed51b3
LW
2268}
2269
2270bool
7fc63493 2271Perl_is_utf8_punct(pTHX_ const U8 *p)
b8c5462f 2272{
97aff369 2273 dVAR;
7918f24d
NC
2274
2275 PERL_ARGS_ASSERT_IS_UTF8_PUNCT;
2276
03940dc2 2277 return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_PUNCT], "IsPunct");
b8c5462f
JH
2278}
2279
2280bool
7fc63493 2281Perl_is_utf8_xdigit(pTHX_ const U8 *p)
b8c5462f 2282{
97aff369 2283 dVAR;
7918f24d
NC
2284
2285 PERL_ARGS_ASSERT_IS_UTF8_XDIGIT;
2286
4ac6419d 2287 return is_XDIGIT_utf8(p);
b8c5462f
JH
2288}
2289
2290bool
7fc63493 2291Perl_is_utf8_mark(pTHX_ const U8 *p)
a0ed51b3 2292{
97aff369 2293 dVAR;
7918f24d
NC
2294
2295 PERL_ARGS_ASSERT_IS_UTF8_MARK;
2296
d4c19fe8 2297 return is_utf8_common(p, &PL_utf8_mark, "IsM");
a0ed51b3
LW
2298}
2299
6b5c0936 2300/*
87cea99e 2301=for apidoc to_utf8_case
6b5c0936 2302
a1433954
KW
2303The C<p> contains the pointer to the UTF-8 string encoding
2304the character that is being converted. This routine assumes that the character
2305at C<p> is well-formed.
6b5c0936 2306
a1433954
KW
2307The C<ustrp> is a pointer to the character buffer to put the
2308conversion result to. The C<lenp> is a pointer to the length
6b5c0936
JH
2309of the result.
2310
a1433954 2311The C<swashp> is a pointer to the swash to use.
6b5c0936 2312
a1433954
KW
2313Both the special and normal mappings are stored in F<lib/unicore/To/Foo.pl>,
2314and loaded by SWASHNEW, using F<lib/utf8_heavy.pl>. The C<special> (usually,
0134edef 2315but not always, a multicharacter mapping), is tried first.
6b5c0936 2316
a1433954 2317The C<special> is a string like "utf8::ToSpecLower", which means the
0134edef
JH
2318hash %utf8::ToSpecLower. The access to the hash is through
2319Perl_to_utf8_case().
6b5c0936 2320
a1433954 2321The C<normal> is a string like "ToLower" which means the swash
0134edef
JH
2322%utf8::ToLower.
2323
2324=cut */
6b5c0936 2325
2104c8d9 2326UV
9a957fbc
AL
2327Perl_to_utf8_case(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp,
2328 SV **swashp, const char *normal, const char *special)
a0ed51b3 2329{
97aff369 2330 dVAR;
89ebb4a3 2331 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
0134edef 2332 STRLEN len = 0;
4b88fb76 2333 const UV uv0 = valid_utf8_to_uvchr(p, NULL);
1feea2c7
JH
2334 /* The NATIVE_TO_UNI() and UNI_TO_NATIVE() mappings
2335 * are necessary in EBCDIC, they are redundant no-ops
2336 * in ASCII-ish platforms, and hopefully optimized away. */
f54cb97a 2337 const UV uv1 = NATIVE_TO_UNI(uv0);
7918f24d
NC
2338
2339 PERL_ARGS_ASSERT_TO_UTF8_CASE;
2340
9ae3ac1a
KW
2341 /* Note that swash_fetch() doesn't output warnings for these because it
2342 * assumes we will */
8457b38f 2343 if (uv1 >= UNICODE_SURROGATE_FIRST) {
9ae3ac1a 2344 if (uv1 <= UNICODE_SURROGATE_LAST) {
8457b38f
KW
2345 if (ckWARN_d(WARN_SURROGATE)) {
2346 const char* desc = (PL_op) ? OP_DESC(PL_op) : normal;
2347 Perl_warner(aTHX_ packWARN(WARN_SURROGATE),
2348 "Operation \"%s\" returns its argument for UTF-16 surrogate U+%04"UVXf"", desc, uv1);
2349 }
9ae3ac1a
KW
2350 }
2351 else if (UNICODE_IS_SUPER(uv1)) {
8457b38f
KW
2352 if (ckWARN_d(WARN_NON_UNICODE)) {
2353 const char* desc = (PL_op) ? OP_DESC(PL_op) : normal;
2354 Perl_warner(aTHX_ packWARN(WARN_NON_UNICODE),
2355 "Operation \"%s\" returns its argument for non-Unicode code point 0x%04"UVXf"", desc, uv1);
2356 }
9ae3ac1a
KW
2357 }
2358
2359 /* Note that non-characters are perfectly legal, so no warning should
2360 * be given */
2361 }
2362
1feea2c7 2363 uvuni_to_utf8(tmpbuf, uv1);
0134edef
JH
2364
2365 if (!*swashp) /* load on-demand */
5ab9d2ef 2366 *swashp = _core_swash_init("utf8", normal, &PL_sv_undef, 4, 0, NULL, NULL);
0134edef 2367
a6f87d8c 2368 if (special) {
0134edef 2369 /* It might be "special" (sometimes, but not always,
2a37f04d 2370 * a multicharacter mapping) */
6673a63c 2371 HV * const hv = get_hv(special, 0);
b08cf34e
JH
2372 SV **svp;
2373
35da51f7 2374 if (hv &&
b08cf34e
JH
2375 (svp = hv_fetch(hv, (const char*)tmpbuf, UNISKIP(uv1), FALSE)) &&
2376 (*svp)) {
cfd0369c 2377 const char *s;
47654450 2378
cfd0369c 2379 s = SvPV_const(*svp, len);
47654450
JH
2380 if (len == 1)
2381 len = uvuni_to_utf8(ustrp, NATIVE_TO_UNI(*(U8*)s)) - ustrp;
2a37f04d 2382 else {
2f9475ad
JH
2383#ifdef EBCDIC
2384 /* If we have EBCDIC we need to remap the characters
2385 * since any characters in the low 256 are Unicode
2386 * code points, not EBCDIC. */
7cda7a3d 2387 U8 *t = (U8*)s, *tend = t + len, *d;
2f9475ad
JH
2388
2389 d = tmpbuf;
b08cf34e 2390 if (SvUTF8(*svp)) {
2f9475ad
JH
2391 STRLEN tlen = 0;
2392
2393 while (t < tend) {
6bd1c396 2394 const UV c = utf8_to_uvchr_buf(t, tend, &tlen);
2f9475ad
JH
2395 if (tlen > 0) {
2396 d = uvchr_to_utf8(d, UNI_TO_NATIVE(c));
2397 t += tlen;
2398 }
2399 else
2400 break;
2401 }
2402 }
2403 else {
36fec512
JH
2404 while (t < tend) {
2405 d = uvchr_to_utf8(d, UNI_TO_NATIVE(*t));
2406 t++;
2407 }
2f9475ad
JH
2408 }
2409 len = d - tmpbuf;
2410 Copy(tmpbuf, ustrp, len, U8);
2411#else
d2dcd0fb 2412 Copy(s, ustrp, len, U8);
2f9475ad 2413#endif
29e98929 2414 }
983ffd37 2415 }
0134edef
JH
2416 }
2417
2418 if (!len && *swashp) {
2114036c 2419 const UV uv2 = swash_fetch(*swashp, tmpbuf, TRUE /* => is utf8 */);
d4c19fe8 2420
0134edef
JH
2421 if (uv2) {
2422 /* It was "normal" (a single character mapping). */
d4c19fe8 2423 const UV uv3 = UNI_TO_NATIVE(uv2);
e9101d72 2424 len = uvchr_to_utf8(ustrp, uv3) - ustrp;
2a37f04d
JH
2425 }
2426 }
1feea2c7 2427
cbe07460
KW
2428 if (len) {
2429 if (lenp) {
2430 *lenp = len;
2431 }
2432 return valid_utf8_to_uvchr(ustrp, 0);
2433 }
2434
2435 /* Here, there was no mapping defined, which means that the code point maps
2436 * to itself. Return the inputs */
bfdf22ec 2437 len = UTF8SKIP(p);
ca9fab46
KW
2438 if (p != ustrp) { /* Don't copy onto itself */
2439 Copy(p, ustrp, len, U8);
2440 }
0134edef 2441
2a37f04d
JH
2442 if (lenp)
2443 *lenp = len;
2444
cbe07460
KW
2445 return uv0;
2446
a0ed51b3
LW
2447}
2448
051a06d4
KW
2449STATIC UV
2450S_check_locale_boundary_crossing(pTHX_ const U8* const p, const UV result, U8* const ustrp, STRLEN *lenp)
2451{
2452 /* This is called when changing the case of a utf8-encoded character above
2453 * the Latin1 range, and the operation is in locale. If the result
2454 * contains a character that crosses the 255/256 boundary, disallow the
2455 * change, and return the original code point. See L<perlfunc/lc> for why;
2456 *
a1433954
KW
2457 * p points to the original string whose case was changed; assumed
2458 * by this routine to be well-formed
051a06d4
KW
2459 * result the code point of the first character in the changed-case string
2460 * ustrp points to the changed-case string (<result> represents its first char)
2461 * lenp points to the length of <ustrp> */
2462
2463 UV original; /* To store the first code point of <p> */
2464
2465 PERL_ARGS_ASSERT_CHECK_LOCALE_BOUNDARY_CROSSING;
2466
2467 assert(! UTF8_IS_INVARIANT(*p) && ! UTF8_IS_DOWNGRADEABLE_START(*p));
2468
2469 /* We know immediately if the first character in the string crosses the
2470 * boundary, so can skip */
2471 if (result > 255) {
2472
2473 /* Look at every character in the result; if any cross the
2474 * boundary, the whole thing is disallowed */
2475 U8* s = ustrp + UTF8SKIP(ustrp);
2476 U8* e = ustrp + *lenp;
2477 while (s < e) {
2478 if (UTF8_IS_INVARIANT(*s) || UTF8_IS_DOWNGRADEABLE_START(*s))
2479 {
2480 goto bad_crossing;
2481 }
2482 s += UTF8SKIP(s);
2483 }
2484
2485 /* Here, no characters crossed, result is ok as-is */
2486 return result;
2487 }
2488
2489bad_crossing:
2490
2491 /* Failed, have to return the original */
4b88fb76 2492 original = valid_utf8_to_uvchr(p, lenp);
051a06d4
KW
2493 Copy(p, ustrp, *lenp, char);
2494 return original;
2495}
2496
d3e79532 2497/*
87cea99e 2498=for apidoc to_utf8_upper
d3e79532 2499
a1433954
KW
2500Convert the UTF-8 encoded character at C<p> to its uppercase version and
2501store that in UTF-8 in C<ustrp> and its length in bytes in C<lenp>. Note
89ebb4a3
JH
2502that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since
2503the uppercase version may be longer than the original character.
d3e79532
JH
2504
2505The first character of the uppercased version is returned
2506(but note, as explained above, that there may be more.)
2507
a1433954
KW
2508The character at C<p> is assumed by this routine to be well-formed.
2509
d3e79532
JH
2510=cut */
2511
051a06d4
KW
2512/* Not currently externally documented, and subject to change:
2513 * <flags> is set iff locale semantics are to be used for code points < 256
2514 * <tainted_ptr> if non-null, *tainted_ptr will be set TRUE iff locale rules
2515 * were used in the calculation; otherwise unchanged. */
2516
2104c8d9 2517UV
051a06d4 2518Perl__to_utf8_upper_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, const bool flags, bool* tainted_ptr)
a0ed51b3 2519{
97aff369 2520 dVAR;
7918f24d 2521
051a06d4
KW
2522 UV result;
2523
2524 PERL_ARGS_ASSERT__TO_UTF8_UPPER_FLAGS;
7918f24d 2525
3a4c58c9 2526 if (UTF8_IS_INVARIANT(*p)) {
051a06d4
KW
2527 if (flags) {
2528 result = toUPPER_LC(*p);
2529 }
2530 else {
81c6c7ce 2531 return _to_upper_title_latin1(*p, ustrp, lenp, 'S');
051a06d4 2532 }
3a4c58c9
KW
2533 }
2534 else if UTF8_IS_DOWNGRADEABLE_START(*p) {
051a06d4
KW
2535 if (flags) {
2536 result = toUPPER_LC(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)));
2537 }
2538 else {
81c6c7ce
KW
2539 return _to_upper_title_latin1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)),
2540 ustrp, lenp, 'S');
051a06d4
KW
2541 }
2542 }
2543 else { /* utf8, ord above 255 */
2544 result = CALL_UPPER_CASE(p, ustrp, lenp);
2545
2546 if (flags) {
2547 result = check_locale_boundary_crossing(p, result, ustrp, lenp);
2548 }
2549 return result;
2550 }
2551
2552 /* Here, used locale rules. Convert back to utf8 */
2553 if (UTF8_IS_INVARIANT(result)) {
2554 *ustrp = (U8) result;
2555 *lenp = 1;
2556 }
2557 else {
2558 *ustrp = UTF8_EIGHT_BIT_HI(result);
2559 *(ustrp + 1) = UTF8_EIGHT_BIT_LO(result);
2560 *lenp = 2;
3a4c58c9
KW
2561 }
2562
051a06d4
KW
2563 if (tainted_ptr) {
2564 *tainted_ptr = TRUE;
2565 }
2566 return result;
983ffd37 2567}
a0ed51b3 2568
d3e79532 2569/*
87cea99e 2570=for apidoc to_utf8_title
d3e79532 2571
a1433954
KW
2572Convert the UTF-8 encoded character at C<p> to its titlecase version and
2573store that in UTF-8 in C<ustrp> and its length in bytes in C<lenp>. Note
2574that the C<ustrp> needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
89ebb4a3 2575titlecase version may be longer than the original character.
d3e79532
JH
2576
2577The first character of the titlecased version is returned
2578(but note, as explained above, that there may be more.)
2579
a1433954
KW
2580The character at C<p> is assumed by this routine to be well-formed.
2581
d3e79532
JH
2582=cut */
2583
051a06d4
KW
2584/* Not currently externally documented, and subject to change:
2585 * <flags> is set iff locale semantics are to be used for code points < 256
2586 * Since titlecase is not defined in POSIX, uppercase is used instead
2587 * for these/
2588 * <tainted_ptr> if non-null, *tainted_ptr will be set TRUE iff locale rules
2589 * were used in the calculation; otherwise unchanged. */
2590
983ffd37 2591UV
051a06d4 2592Perl__to_utf8_title_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, const bool flags, bool* tainted_ptr)
983ffd37 2593{
97aff369 2594 dVAR;
7918f24d 2595
051a06d4
KW
2596 UV result;
2597
2598 PERL_ARGS_ASSERT__TO_UTF8_TITLE_FLAGS;
7918f24d 2599
3a4c58c9 2600 if (UTF8_IS_INVARIANT(*p)) {
051a06d4
KW
2601 if (flags) {
2602 result = toUPPER_LC(*p);
2603 }
2604 else {
81c6c7ce 2605 return _to_upper_title_latin1(*p, ustrp, lenp, 's');
051a06d4 2606 }
3a4c58c9
KW
2607 }
2608 else if UTF8_IS_DOWNGRADEABLE_START(*p) {
051a06d4
KW
2609 if (flags) {
2610 result = toUPPER_LC(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)));
2611 }
2612 else {
81c6c7ce
KW
2613 return _to_upper_title_latin1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)),
2614 ustrp, lenp, 's');
051a06d4
KW
2615 }
2616 }
2617 else { /* utf8, ord above 255 */
2618 result = CALL_TITLE_CASE(p, ustrp, lenp);
2619
2620 if (flags) {
2621 result = check_locale_boundary_crossing(p, result, ustrp, lenp);
2622 }
2623 return result;
2624 }
2625
2626 /* Here, used locale rules. Convert back to utf8 */
2627 if (UTF8_IS_INVARIANT(result)) {
2628 *ustrp = (U8) result;
2629 *lenp = 1;
2630 }
2631 else {
2632 *ustrp = UTF8_EIGHT_BIT_HI(result);
2633 *(ustrp + 1) = UTF8_EIGHT_BIT_LO(result);
2634 *lenp = 2;
3a4c58c9
KW
2635 }
2636
051a06d4
KW
2637 if (tainted_ptr) {
2638 *tainted_ptr = TRUE;
2639 }
2640 return result;
a0ed51b3
LW
2641}
2642
d3e79532 2643/*
87cea99e 2644=for apidoc to_utf8_lower
d3e79532 2645
a1433954
KW
2646Convert the UTF-8 encoded character at C<p> to its lowercase version and
2647store that in UTF-8 in ustrp and its length in bytes in C<lenp>. Note
2648that the C<ustrp> needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
89ebb4a3 2649lowercase version may be longer than the original character.
d3e79532
JH
2650
2651The first character of the lowercased version is returned
2652(but note, as explained above, that there may be more.)
2653
a1433954
KW
2654The character at C<p> is assumed by this routine to be well-formed.
2655
d3e79532
JH
2656=cut */
2657
051a06d4
KW
2658/* Not currently externally documented, and subject to change:
2659 * <flags> is set iff locale semantics are to be used for code points < 256
2660 * <tainted_ptr> if non-null, *tainted_ptr will be set TRUE iff locale rules
2661 * were used in the calculation; otherwise unchanged. */
2662
2104c8d9 2663UV
051a06d4 2664Perl__to_utf8_lower_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, const bool flags, bool* tainted_ptr)
a0ed51b3 2665{
051a06d4
KW
2666 UV result;
2667
97aff369 2668 dVAR;
7918f24d 2669
051a06d4 2670 PERL_ARGS_ASSERT__TO_UTF8_LOWER_FLAGS;
7918f24d 2671
968c5e6a 2672 if (UTF8_IS_INVARIANT(*p)) {
051a06d4
KW
2673 if (flags) {
2674 result = toLOWER_LC(*p);
2675 }
2676 else {
81c6c7ce 2677 return to_lower_latin1(*p, ustrp, lenp);
051a06d4 2678 }
968c5e6a
KW
2679 }
2680 else if UTF8_IS_DOWNGRADEABLE_START(*p) {
051a06d4
KW
2681 if (flags) {
2682 result = toLOWER_LC(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)));
2683 }
2684 else {
81c6c7ce
KW
2685 return to_lower_latin1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)),
2686 ustrp, lenp);
051a06d4 2687 }
968c5e6a 2688 }
051a06d4
KW
2689 else { /* utf8, ord above 255 */
2690 result = CALL_LOWER_CASE(p, ustrp, lenp);
2691
2692 if (flags) {
2693 result = check_locale_boundary_crossing(p, result, ustrp, lenp);
2694 }
968c5e6a 2695
051a06d4
KW
2696 return result;
2697 }
2698
2699 /* Here, used locale rules. Convert back to utf8 */
2700 if (UTF8_IS_INVARIANT(result)) {
2701 *ustrp = (U8) result;
2702 *lenp = 1;
2703 }
2704 else {
2705 *ustrp = UTF8_EIGHT_BIT_HI(result);
2706 *(ustrp + 1) = UTF8_EIGHT_BIT_LO(result);
2707 *lenp = 2;
2708 }
2709
2710 if (tainted_ptr) {
2711 *tainted_ptr = TRUE;
2712 }
2713 return result;
b4e400f9
JH
2714}
2715
d3e79532 2716/*
87cea99e 2717=for apidoc to_utf8_fold
d3e79532 2718
a1433954
KW
2719Convert the UTF-8 encoded character at C<p> to its foldcase version and
2720store that in UTF-8 in C<ustrp> and its length in bytes in C<lenp>. Note
2721that the C<ustrp> needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
d3e79532
JH
2722foldcase version may be longer than the original character (up to
2723three characters).
2724
2725The first character of the foldcased version is returned
2726(but note, as explained above, that there may be more.)
2727
a1433954
KW
2728The character at C<p> is assumed by this routine to be well-formed.
2729
d3e79532
JH
2730=cut */
2731
051a06d4
KW
2732/* Not currently externally documented, and subject to change,
2733 * in <flags>
2734 * bit FOLD_FLAGS_LOCALE is set iff locale semantics are to be used for code
2735 * points < 256. Since foldcase is not defined in
2736 * POSIX, lowercase is used instead
2737 * bit FOLD_FLAGS_FULL is set iff full case folds are to be used;
2738 * otherwise simple folds
a0270393
KW
2739 * bit FOLD_FLAGS_NOMIX_ASCII is set iff folds of non-ASCII to ASCII are
2740 * prohibited
051a06d4
KW
2741 * <tainted_ptr> if non-null, *tainted_ptr will be set TRUE iff locale rules
2742 * were used in the calculation; otherwise unchanged. */
36bb2ab6 2743
b4e400f9 2744UV
051a06d4 2745Perl__to_utf8_fold_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, U8 flags, bool* tainted_ptr)
b4e400f9 2746{
97aff369 2747 dVAR;
7918f24d 2748
051a06d4
KW
2749 UV result;
2750
36bb2ab6 2751 PERL_ARGS_ASSERT__TO_UTF8_FOLD_FLAGS;
7918f24d 2752
a0270393
KW
2753 /* These are mutually exclusive */
2754 assert (! ((flags & FOLD_FLAGS_LOCALE) && (flags & FOLD_FLAGS_NOMIX_ASCII)));
2755
50ba90ff
KW
2756 assert(p != ustrp); /* Otherwise overwrites */
2757
a1dde8de 2758 if (UTF8_IS_INVARIANT(*p)) {
051a06d4
KW
2759 if (flags & FOLD_FLAGS_LOCALE) {
2760 result = toLOWER_LC(*p);
2761 }
2762 else {
81c6c7ce
KW
2763 return _to_fold_latin1(*p, ustrp, lenp,
2764 cBOOL(flags & FOLD_FLAGS_FULL));
051a06d4 2765 }
a1dde8de
KW
2766 }
2767 else if UTF8_IS_DOWNGRADEABLE_START(*p) {
051a06d4
KW
2768 if (flags & FOLD_FLAGS_LOCALE) {
2769 result = toLOWER_LC(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)));
2770 }
2771 else {
81c6c7ce 2772 return _to_fold_latin1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)),
a0270393
KW
2773 ustrp, lenp,
2774 cBOOL((flags & FOLD_FLAGS_FULL
2775 /* If ASCII safe, don't allow full
2776 * folding, as that could include SHARP
2777 * S => ss; otherwise there is no
2778 * crossing of ascii/non-ascii in the
2779 * latin1 range */
2780 && ! (flags & FOLD_FLAGS_NOMIX_ASCII))));
051a06d4 2781 }
a1dde8de 2782 }
051a06d4 2783 else { /* utf8, ord above 255 */
a0270393 2784 result = CALL_FOLD_CASE(p, ustrp, lenp, flags & FOLD_FLAGS_FULL);
a1dde8de 2785
051a06d4 2786 if ((flags & FOLD_FLAGS_LOCALE)) {
a0270393 2787 return check_locale_boundary_crossing(p, result, ustrp, lenp);
051a06d4 2788 }
a0270393
KW
2789 else if (! (flags & FOLD_FLAGS_NOMIX_ASCII)) {
2790 return result;
2791 }
2792 else {
2793 /* This is called when changing the case of a utf8-encoded
2794 * character above the Latin1 range, and the result should not
2795 * contain an ASCII character. */
2796
2797 UV original; /* To store the first code point of <p> */
2798
2799 /* Look at every character in the result; if any cross the
2800 * boundary, the whole thing is disallowed */
2801 U8* s = ustrp;
2802 U8* e = ustrp + *lenp;
2803 while (s < e) {
2804 if (isASCII(*s)) {
2805 /* Crossed, have to return the original */
2806 original = valid_utf8_to_uvchr(p, lenp);
2807 Copy(p, ustrp, *lenp, char);
2808 return original;
2809 }
2810 s += UTF8SKIP(s);
2811 }
051a06d4 2812
a0270393
KW
2813 /* Here, no characters crossed, result is ok as-is */
2814 return result;
2815 }
051a06d4
KW
2816 }
2817
2818 /* Here, used locale rules. Convert back to utf8 */
2819 if (UTF8_IS_INVARIANT(result)) {
2820 *ustrp = (U8) result;
2821 *lenp = 1;
2822 }
2823 else {
2824 *ustrp = UTF8_EIGHT_BIT_HI(result);
2825 *(ustrp + 1) = UTF8_EIGHT_BIT_LO(result);
2826 *lenp = 2;
2827 }
2828
2829 if (tainted_ptr) {
2830 *tainted_ptr = TRUE;
2831 }
2832 return result;
a0ed51b3
LW
2833}
2834
711a919c 2835/* Note:
f90a9a02 2836 * Returns a "swash" which is a hash described in utf8.c:Perl_swash_fetch().
711a919c
ST
2837 * C<pkg> is a pointer to a package name for SWASHNEW, should be "utf8".
2838 * For other parameters, see utf8::SWASHNEW in lib/utf8_heavy.pl.
2839 */
c4a5db0c 2840
a0ed51b3 2841SV*
7fc63493 2842Perl_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv, I32 minbits, I32 none)
a0ed51b3 2843{
c4a5db0c
KW
2844 PERL_ARGS_ASSERT_SWASH_INIT;
2845
2846 /* Returns a copy of a swash initiated by the called function. This is the
2847 * public interface, and returning a copy prevents others from doing
2848 * mischief on the original */
2849
5d3d13d1 2850 return newSVsv(_core_swash_init(pkg, name, listsv, minbits, none, NULL, NULL));
c4a5db0c
KW
2851}
2852
2853SV*
5d3d13d1 2854Perl__core_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv, I32 minbits, I32 none, SV* invlist, U8* const flags_p)
c4a5db0c
KW
2855{
2856 /* Initialize and return a swash, creating it if necessary. It does this
87367d5f
KW
2857 * by calling utf8_heavy.pl in the general case. The returned value may be
2858 * the swash's inversion list instead if the input parameters allow it.
2859 * Which is returned should be immaterial to callers, as the only
923b6d4e
KW
2860 * operations permitted on a swash, swash_fetch(), _get_swash_invlist(),
2861 * and swash_to_invlist() handle both these transparently.
c4a5db0c
KW
2862 *
2863 * This interface should only be used by functions that won't destroy or
2864 * adversely change the swash, as doing so affects all other uses of the
2865 * swash in the program; the general public should use 'Perl_swash_init'
2866 * instead.
2867 *
2868 * pkg is the name of the package that <name> should be in.
2869 * name is the name of the swash to find. Typically it is a Unicode
2870 * property name, including user-defined ones
2871 * listsv is a string to initialize the swash with. It must be of the form
2872 * documented as the subroutine return value in
2873 * L<perlunicode/User-Defined Character Properties>
2874 * minbits is the number of bits required to represent each data element.
2875 * It is '1' for binary properties.
2876 * none I (khw) do not understand this one, but it is used only in tr///.
9a53f6cf 2877 * invlist is an inversion list to initialize the swash with (or NULL)
83199d38
KW
2878 * flags_p if non-NULL is the address of various input and output flag bits
2879 * to the routine, as follows: ('I' means is input to the routine;
2880 * 'O' means output from the routine. Only flags marked O are
2881 * meaningful on return.)
2882 * _CORE_SWASH_INIT_USER_DEFINED_PROPERTY indicates if the swash
2883 * came from a user-defined property. (I O)
5d3d13d1
KW
2884 * _CORE_SWASH_INIT_RETURN_IF_UNDEF indicates that instead of croaking
2885 * when the swash cannot be located, to simply return NULL. (I)
87367d5f
KW
2886 * _CORE_SWASH_INIT_ACCEPT_INVLIST indicates that the caller will accept a
2887 * return of an inversion list instead of a swash hash if this routine
2888 * thinks that would result in faster execution of swash_fetch() later
2889 * on. (I)
9a53f6cf
KW
2890 *
2891 * Thus there are three possible inputs to find the swash: <name>,
2892 * <listsv>, and <invlist>. At least one must be specified. The result
2893 * will be the union of the specified ones, although <listsv>'s various
2894 * actions can intersect, etc. what <name> gives.
2895 *
2896 * <invlist> is only valid for binary properties */
c4a5db0c 2897
27da23d5 2898 dVAR;
c4a5db0c 2899 SV* retval = &PL_sv_undef;
83199d38 2900 HV* swash_hv = NULL;
87367d5f
KW
2901 const int invlist_swash_boundary =
2902 (flags_p && *flags_p & _CORE_SWASH_INIT_ACCEPT_INVLIST)
2903 ? 512 /* Based on some benchmarking, but not extensive, see commit
2904 message */
2905 : -1; /* Never return just an inversion list */
9a53f6cf
KW
2906
2907 assert(listsv != &PL_sv_undef || strNE(name, "") || invlist);
2908 assert(! invlist || minbits == 1);
2909
2910 /* If data was passed in to go out to utf8_heavy to find the swash of, do
2911 * so */
2912 if (listsv != &PL_sv_undef || strNE(name, "")) {
69794297
KW
2913 dSP;
2914 const size_t pkg_len = strlen(pkg);
2915 const size_t name_len = strlen(name);
2916 HV * const stash = gv_stashpvn(pkg, pkg_len, 0);
2917 SV* errsv_save;
2918 GV *method;
2919
2920 PERL_ARGS_ASSERT__CORE_SWASH_INIT;
2921
2922 PUSHSTACKi(PERLSI_MAGIC);
ce3b816e 2923 ENTER;
69794297
KW
2924 SAVEHINTS();
2925 save_re_context();
650f067c
JL
2926 /* We might get here via a subroutine signature which uses a utf8
2927 * parameter name, at which point PL_subname will have been set
2928 * but not yet used. */
2929 save_item(PL_subname);
69794297
KW
2930 if (PL_parser && PL_parser->error_count)
2931 SAVEI8(PL_parser->error_count), PL_parser->error_count = 0;
2932 method = gv_fetchmeth(stash, "SWASHNEW", 8, -1);
2933 if (!method) { /* demand load utf8 */
2934 ENTER;
db2c6cb3
FC
2935 if ((errsv_save = GvSV(PL_errgv))) SAVEFREESV(errsv_save);
2936 GvSV(PL_errgv) = NULL;
69794297
KW
2937 /* It is assumed that callers of this routine are not passing in
2938 * any user derived data. */
2939 /* Need to do this after save_re_context() as it will set
2940 * PL_tainted to 1 while saving $1 etc (see the code after getrx:
2941 * in Perl_magic_get). Even line to create errsv_save can turn on
2942 * PL_tainted. */
284167a5
SM
2943#ifndef NO_TAINT_SUPPORT
2944 SAVEBOOL(TAINT_get);
2945 TAINT_NOT;
2946#endif
69794297
KW
2947 Perl_load_module(aTHX_ PERL_LOADMOD_NOIMPORT, newSVpvn(pkg,pkg_len),
2948 NULL);
eed484f9 2949 {
db2c6cb3
FC
2950 /* Not ERRSV, as there is no need to vivify a scalar we are
2951 about to discard. */
2952 SV * const errsv = GvSV(PL_errgv);
2953 if (!SvTRUE(errsv)) {
2954 GvSV(PL_errgv) = SvREFCNT_inc_simple(errsv_save);
2955 SvREFCNT_dec(errsv);
2956 }
eed484f9 2957 }
69794297
KW
2958 LEAVE;
2959 }
2960 SPAGAIN;
2961 PUSHMARK(SP);
2962 EXTEND(SP,5);
2963 mPUSHp(pkg, pkg_len);
2964 mPUSHp(name, name_len);
2965 PUSHs(listsv);
2966 mPUSHi(minbits);
2967 mPUSHi(none);
2968 PUTBACK;
db2c6cb3
FC
2969 if ((errsv_save = GvSV(PL_errgv))) SAVEFREESV(errsv_save);
2970 GvSV(PL_errgv) = NULL;
69794297
KW
2971 /* If we already have a pointer to the method, no need to use
2972 * call_method() to repeat the lookup. */
c41800a8
KW
2973 if (method
2974 ? call_sv(MUTABLE_SV(method), G_SCALAR)
69794297
KW
2975 : call_sv(newSVpvs_flags("SWASHNEW", SVs_TEMP), G_SCALAR | G_METHOD))
2976 {
2977 retval = *PL_stack_sp--;
2978 SvREFCNT_inc(retval);
2979 }
eed484f9 2980 {
db2c6cb3
FC
2981 /* Not ERRSV. See above. */
2982 SV * const errsv = GvSV(PL_errgv);
2983 if (!SvTRUE(errsv)) {
2984 GvSV(PL_errgv) = SvREFCNT_inc_simple(errsv_save);
2985 SvREFCNT_dec(errsv);
2986 }
eed484f9 2987 }
ce3b816e 2988 LEAVE;
69794297
KW
2989 POPSTACK;
2990 if (IN_PERL_COMPILETIME) {
2991 CopHINTS_set(PL_curcop, PL_hints);
2992 }
2993 if (!SvROK(retval) || SvTYPE(SvRV(retval)) != SVt_PVHV) {
2994 if (SvPOK(retval))
2995
2996 /* If caller wants to handle missing properties, let them */
5d3d13d1 2997 if (flags_p && *flags_p & _CORE_SWASH_INIT_RETURN_IF_UNDEF) {
69794297
KW
2998 return NULL;
2999 }
3000 Perl_croak(aTHX_
3001 "Can't find Unicode property definition \"%"SVf"\"",
3002 SVfARG(retval));
3003 Perl_croak(aTHX_ "SWASHNEW didn't return an HV ref");
3004 }
9a53f6cf 3005 } /* End of calling the module to find the swash */
36eb48b4 3006
83199d38
KW
3007 /* If this operation fetched a swash, and we will need it later, get it */
3008 if (retval != &PL_sv_undef
3009 && (minbits == 1 || (flags_p
3010 && ! (*flags_p
3011 & _CORE_SWASH_INIT_USER_DEFINED_PROPERTY))))
3012 {
3013 swash_hv = MUTABLE_HV(SvRV(retval));
3014
3015 /* If we don't already know that there is a user-defined component to
3016 * this swash, and the user has indicated they wish to know if there is
3017 * one (by passing <flags_p>), find out */
3018 if (flags_p && ! (*flags_p & _CORE_SWASH_INIT_USER_DEFINED_PROPERTY)) {
3019 SV** user_defined = hv_fetchs(swash_hv, "USER_DEFINED", FALSE);
3020 if (user_defined && SvUV(*user_defined)) {
3021 *flags_p |= _CORE_SWASH_INIT_USER_DEFINED_PROPERTY;
3022 }
3023 }
3024 }
3025
36eb48b4
KW
3026 /* Make sure there is an inversion list for binary properties */
3027 if (minbits == 1) {
3028 SV** swash_invlistsvp = NULL;
3029 SV* swash_invlist = NULL;
9a53f6cf 3030 bool invlist_in_swash_is_valid = FALSE;
02c85471
FC
3031 bool swash_invlist_unclaimed = FALSE; /* whether swash_invlist has
3032 an unclaimed reference count */
36eb48b4 3033
9a53f6cf 3034 /* If this operation fetched a swash, get its already existing
83199d38 3035 * inversion list, or create one for it */
36eb48b4 3036
83199d38 3037 if (swash_hv) {
5c9f4bd2 3038 swash_invlistsvp = hv_fetchs(swash_hv, "V", FALSE);
9a53f6cf
KW
3039 if (swash_invlistsvp) {
3040 swash_invlist = *swash_invlistsvp;
3041 invlist_in_swash_is_valid = TRUE;
3042 }
3043 else {
36eb48b4 3044 swash_invlist = _swash_to_invlist(retval);
02c85471 3045 swash_invlist_unclaimed = TRUE;
9a53f6cf
KW
3046 }
3047 }
3048
3049 /* If an inversion list was passed in, have to include it */
3050 if (invlist) {
3051
3052 /* Any fetched swash will by now have an inversion list in it;
3053 * otherwise <swash_invlist> will be NULL, indicating that we
3054 * didn't fetch a swash */
3055 if (swash_invlist) {
3056
3057 /* Add the passed-in inversion list, which invalidates the one
3058 * already stored in the swash */
3059 invlist_in_swash_is_valid = FALSE;
3060 _invlist_union(invlist, swash_invlist, &swash_invlist);
3061 }
3062 else {
3063
87367d5f
KW
3064 /* Here, there is no swash already. Set up a minimal one, if
3065 * we are going to return a swash */
3066 if ((int) _invlist_len(invlist) > invlist_swash_boundary) {
971d486f 3067 swash_hv = newHV();
4aca0fe6 3068 retval = newRV_noinc(MUTABLE_SV(swash_hv));
87367d5f 3069 }
9a53f6cf
KW
3070 swash_invlist = invlist;
3071 }
9a53f6cf
KW
3072 }
3073
3074 /* Here, we have computed the union of all the passed-in data. It may
3075 * be that there was an inversion list in the swash which didn't get
3076 * touched; otherwise save the one computed one */
87367d5f
KW
3077 if (! invlist_in_swash_is_valid
3078 && (int) _invlist_len(swash_invlist) > invlist_swash_boundary)
3079 {
5c9f4bd2 3080 if (! hv_stores(MUTABLE_HV(SvRV(retval)), "V", swash_invlist))
69794297
KW
3081 {
3082 Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed");
3083 }
cc34d8c5
FC
3084 /* We just stole a reference count. */
3085 if (swash_invlist_unclaimed) swash_invlist_unclaimed = FALSE;
3086 else SvREFCNT_inc_simple_void_NN(swash_invlist);
9a53f6cf 3087 }
87367d5f 3088
c41800a8 3089 /* Use the inversion list stand-alone if small enough */
87367d5f
KW
3090 if ((int) _invlist_len(swash_invlist) <= invlist_swash_boundary) {
3091 SvREFCNT_dec(retval);
02c85471
FC
3092 if (!swash_invlist_unclaimed)
3093 SvREFCNT_inc_simple_void_NN(swash_invlist);
3094 retval = newRV_noinc(swash_invlist);
87367d5f 3095 }
36eb48b4
KW
3096 }
3097
a0ed51b3
LW
3098 return retval;
3099}
3100
035d37be
JH
3101
3102/* This API is wrong for special case conversions since we may need to
3103 * return several Unicode characters for a single Unicode character
3104 * (see lib/unicore/SpecCase.txt) The SWASHGET in lib/utf8_heavy.pl is
3105 * the lower-level routine, and it is similarly broken for returning
38684baa
KW
3106 * multiple values. --jhi
3107 * For those, you should use to_utf8_case() instead */
b0e3252e 3108/* Now SWASHGET is recasted into S_swatch_get in this file. */
680c470c
ST
3109
3110/* Note:
3111 * Returns the value of property/mapping C<swash> for the first character
3112 * of the string C<ptr>. If C<do_utf8> is true, the string C<ptr> is
3113 * assumed to be in utf8. If C<do_utf8> is false, the string C<ptr> is
3114 * assumed to be in native 8-bit encoding. Caches the swatch in C<swash>.
af2af982
KW
3115 *
3116 * A "swash" is a hash which contains initially the keys/values set up by
3117 * SWASHNEW. The purpose is to be able to completely represent a Unicode
3118 * property for all possible code points. Things are stored in a compact form
3119 * (see utf8_heavy.pl) so that calculation is required to find the actual
3120 * property value for a given code point. As code points are looked up, new
3121 * key/value pairs are added to the hash, so that the calculation doesn't have
3122 * to ever be re-done. Further, each calculation is done, not just for the
3123 * desired one, but for a whole block of code points adjacent to that one.
3124 * For binary properties on ASCII machines, the block is usually for 64 code
3125 * points, starting with a code point evenly divisible by 64. Thus if the
3126 * property value for code point 257 is requested, the code goes out and
3127 * calculates the property values for all 64 code points between 256 and 319,
3128 * and stores these as a single 64-bit long bit vector, called a "swatch",
3129 * under the key for code point 256. The key is the UTF-8 encoding for code
3130 * point 256, minus the final byte. Thus, if the length of the UTF-8 encoding
3131 * for a code point is 13 bytes, the key will be 12 bytes long. If the value
3132 * for code point 258 is then requested, this code realizes that it would be
3133 * stored under the key for 256, and would find that value and extract the
3134 * relevant bit, offset from 256.
3135 *
3136 * Non-binary properties are stored in as many bits as necessary to represent
3137 * their values (32 currently, though the code is more general than that), not
3138 * as single bits, but the principal is the same: the value for each key is a
3139 * vector that encompasses the property values for all code points whose UTF-8
3140 * representations are represented by the key. That is, for all code points
3141 * whose UTF-8 representations are length N bytes, and the key is the first N-1
3142 * bytes of that.
680c470c 3143 */
a0ed51b3 3144UV
680c470c 3145Perl_swash_fetch(pTHX_ SV *swash, const U8 *ptr, bool do_utf8)
a0ed51b3 3146{
27da23d5 3147 dVAR;
ef8f7699 3148 HV *const hv = MUTABLE_HV(SvRV(swash));
3568d838
JH
3149 U32 klen;
3150 U32 off;
a0ed51b3 3151 STRLEN slen;
7d85a32c 3152 STRLEN needents;
cfd0369c 3153 const U8 *tmps = NULL;
a0ed51b3 3154 U32 bit;
979f2922 3155 SV *swatch;
3568d838 3156 U8 tmputf8[2];
35da51f7 3157 const UV c = NATIVE_TO_ASCII(*ptr);
3568d838 3158
7918f24d
NC
3159 PERL_ARGS_ASSERT_SWASH_FETCH;
3160
87367d5f
KW
3161 /* If it really isn't a hash, it isn't really swash; must be an inversion
3162 * list */
3163 if (SvTYPE(hv) != SVt_PVHV) {
3164 return _invlist_contains_cp((SV*)hv,
3165 (do_utf8)
3166 ? valid_utf8_to_uvchr(ptr, NULL)
3167 : c);
3168 }
3169
dbe7a391 3170 /* Convert to utf8 if not already */
3568d838 3171 if (!do_utf8 && !UNI_IS_INVARIANT(c)) {
979f2922
ST
3172 tmputf8[0] = (U8)UTF8_EIGHT_BIT_HI(c);
3173 tmputf8[1] = (U8)UTF8_EIGHT_BIT_LO(c);
3174 ptr = tmputf8;
3568d838
JH
3175 }
3176 /* Given a UTF-X encoded char 0xAA..0xYY,0xZZ
37e2e78e 3177 * then the "swatch" is a vec() for all the chars which start
3568d838
JH
3178 * with 0xAA..0xYY
3179 * So the key in the hash (klen) is length of encoded char -1
3180 */
3181 klen = UTF8SKIP(ptr) - 1;
3182 off = ptr[klen];
a0ed51b3 3183
979f2922 3184 if (klen == 0) {
37e2e78e 3185 /* If char is invariant then swatch is for all the invariant chars
1e54db1a 3186 * In both UTF-8 and UTF-8-MOD that happens to be UTF_CONTINUATION_MARK
7d85a32c 3187 */
979f2922
ST
3188 needents = UTF_CONTINUATION_MARK;
3189 off = NATIVE_TO_UTF(ptr[klen]);
3190 }
3191 else {
7d85a32c 3192 /* If char is encoded then swatch is for the prefix */
979f2922
ST
3193 needents = (1 << UTF_ACCUMULATION_SHIFT);
3194 off = NATIVE_TO_UTF(ptr[klen]) & UTF_CONTINUATION_MASK;
3195 }
7d85a32c 3196
a0ed51b3
LW
3197 /*
3198 * This single-entry cache saves about 1/3 of the utf8 overhead in test
3199 * suite. (That is, only 7-8% overall over just a hash cache. Still,
3200 * it's nothing to sniff at.) Pity we usually come through at least
3201 * two function calls to get here...
3202 *
3203 * NB: this code assumes that swatches are never modified, once generated!
3204 */
3205
3568d838 3206 if (hv == PL_last_swash_hv &&
a0ed51b3 3207 klen == PL_last_swash_klen &&
27da23d5 3208 (!klen || memEQ((char *)ptr, (char *)PL_last_swash_key, klen)) )
a0ed51b3
LW
3209 {
3210 tmps = PL_last_swash_tmps;
3211 slen = PL_last_swash_slen;
3212 }
3213 else {
3214 /* Try our second-level swatch cache, kept in a hash. */
e1ec3a88 3215 SV** svp = hv_fetch(hv, (const char*)ptr, klen, FALSE);
a0ed51b3 3216
b0e3252e 3217 /* If not cached, generate it via swatch_get */
979f2922
ST
3218 if (!svp || !SvPOK(*svp)
3219 || !(tmps = (const U8*)SvPV_const(*svp, slen))) {
2b9d42f0
NIS
3220 /* We use utf8n_to_uvuni() as we want an index into
3221 Unicode tables, not a native character number.
3222 */
aec46f14 3223 const UV code_point = utf8n_to_uvuni(ptr, UTF8_MAXBYTES, 0,
872c91ae
JH
3224 ckWARN(WARN_UTF8) ?
3225 0 : UTF8_ALLOW_ANY);
b0e3252e 3226 swatch = swatch_get(swash,
979f2922 3227 /* On EBCDIC & ~(0xA0-1) isn't a useful thing to do */
361ee0fe 3228 (klen) ? (code_point & ~((UV)needents - 1)) : 0,
979f2922
ST
3229 needents);
3230
923e4eb5 3231 if (IN_PERL_COMPILETIME)
623e6609 3232 CopHINTS_set(PL_curcop, PL_hints);
a0ed51b3 3233
979f2922 3234 svp = hv_store(hv, (const char *)ptr, klen, swatch, 0);
a0ed51b3 3235
979f2922
ST
3236 if (!svp || !(tmps = (U8*)SvPV(*svp, slen))
3237 || (slen << 3) < needents)
5637ef5b
NC
3238 Perl_croak(aTHX_ "panic: swash_fetch got improper swatch, "
3239 "svp=%p, tmps=%p, slen=%"UVuf", needents=%"UVuf,
3240 svp, tmps, (UV)slen, (UV)needents);
a0ed51b3
LW
3241 }
3242
3243 PL_last_swash_hv = hv;
16d8f38a 3244 assert(klen <= sizeof(PL_last_swash_key));
eac04b2e 3245 PL_last_swash_klen = (U8)klen;
cfd0369c
NC
3246 /* FIXME change interpvar.h? */
3247 PL_last_swash_tmps = (U8 *) tmps;
a0ed51b3
LW
3248 PL_last_swash_slen = slen;
3249 if (klen)
3250 Copy(ptr, PL_last_swash_key, klen, U8);
3251 }
3252
9faf8d75 3253 switch ((int)((slen << 3) / needents)) {
a0ed51b3
LW
3254 case 1:
3255 bit = 1 << (off & 7);
3256 off >>= 3;
3257 return (tmps[off] & bit) != 0;
3258 case 8:
3259 return tmps[off];
3260 case 16:
3261 off <<= 1;
3262 return (tmps[off] << 8) + tmps[off + 1] ;
3263 case 32:
3264 off <<= 2;
3265 return (tmps[off] << 24) + (tmps[off+1] << 16) + (tmps[off+2] << 8) + tmps[off + 3] ;
3266 }
5637ef5b
NC
3267 Perl_croak(aTHX_ "panic: swash_fetch got swatch of unexpected bit width, "
3268 "slen=%"UVuf", needents=%"UVuf, (UV)slen, (UV)needents);
670f1322 3269 NORETURN_FUNCTION_END;
a0ed51b3 3270}
2b9d42f0 3271
319009ee
KW
3272/* Read a single line of the main body of the swash input text. These are of
3273 * the form:
3274 * 0053 0056 0073
3275 * where each number is hex. The first two numbers form the minimum and
3276 * maximum of a range, and the third is the value associated with the range.
3277 * Not all swashes should have a third number
3278 *
3279 * On input: l points to the beginning of the line to be examined; it points
3280 * to somewhere in the string of the whole input text, and is
3281 * terminated by a \n or the null string terminator.
3282 * lend points to the null terminator of that string
3283 * wants_value is non-zero if the swash expects a third number
3284 * typestr is the name of the swash's mapping, like 'ToLower'
3285 * On output: *min, *max, and *val are set to the values read from the line.
3286 * returns a pointer just beyond the line examined. If there was no
3287 * valid min number on the line, returns lend+1
3288 */
3289
3290STATIC U8*
3291S_swash_scan_list_line(pTHX_ U8* l, U8* const lend, UV* min, UV* max, UV* val,
3292 const bool wants_value, const U8* const typestr)
3293{
3294 const int typeto = typestr[0] == 'T' && typestr[1] == 'o';
3295 STRLEN numlen; /* Length of the number */
02470786
KW
3296 I32 flags = PERL_SCAN_SILENT_ILLDIGIT
3297 | PERL_SCAN_DISALLOW_PREFIX
3298 | PERL_SCAN_SILENT_NON_PORTABLE;
319009ee
KW
3299
3300 /* nl points to the next \n in the scan */
3301 U8* const nl = (U8*)memchr(l, '\n', lend - l);
3302
3303 /* Get the first number on the line: the range minimum */
3304 numlen = lend - l;
3305 *min = grok_hex((char *)l, &numlen, &flags, NULL);
3306 if (numlen) /* If found a hex number, position past it */
3307 l += numlen;
3308 else if (nl) { /* Else, go handle next line, if any */
3309 return nl + 1; /* 1 is length of "\n" */
3310 }
3311 else { /* Else, no next line */
3312 return lend + 1; /* to LIST's end at which \n is not found */
3313 }
3314
3315 /* The max range value follows, separated by a BLANK */
3316 if (isBLANK(*l)) {
3317 ++l;
02470786
KW
3318 flags = PERL_SCAN_SILENT_ILLDIGIT
3319 | PERL_SCAN_DISALLOW_PREFIX
3320 | PERL_SCAN_SILENT_NON_PORTABLE;
319009ee
KW
3321 numlen = lend - l;
3322 *max = grok_hex((char *)l, &numlen, &flags, NULL);
3323 if (numlen)
3324 l += numlen;
3325 else /* If no value here, it is a single element range */
3326 *max = *min;
3327
3328 /* Non-binary tables have a third entry: what the first element of the
3329 * range maps to */
3330 if (wants_value) {
3331 if (isBLANK(*l)) {
3332 ++l;
f90a9a02
KW
3333
3334 /* The ToLc, etc table mappings are not in hex, and must be
3335 * corrected by adding the code point to them */
3336 if (typeto) {
3337 char *after_strtol = (char *) lend;
3338 *val = Strtol((char *)l, &after_strtol, 10);
3339 l = (U8 *) after_strtol;
f90a9a02
KW
3340 }
3341 else { /* Other tables are in hex, and are the correct result
3342 without tweaking */
a9d188b3
KW
3343 flags = PERL_SCAN_SILENT_ILLDIGIT
3344 | PERL_SCAN_DISALLOW_PREFIX
3345 | PERL_SCAN_SILENT_NON_PORTABLE;
3346 numlen = lend - l;
3347 *val = grok_hex((char *)l, &numlen, &flags, NULL);
3348 if (numlen)
3349 l += numlen;
3350 else
3351 *val = 0;
f90a9a02 3352 }
319009ee
KW
3353 }
3354 else {
3355 *val = 0;
3356 if (typeto) {
dcbac5bb 3357 /* diag_listed_as: To%s: illegal mapping '%s' */
319009ee
KW
3358 Perl_croak(aTHX_ "%s: illegal mapping '%s'",
3359 typestr, l);
3360 }
3361 }
3362 }
3363 else
3364 *val = 0; /* bits == 1, then any val should be ignored */
3365 }
3366 else { /* Nothing following range min, should be single element with no
3367 mapping expected */
3368 *max = *min;
3369 if (wants_value) {
3370 *val = 0;
3371 if (typeto) {
dcbac5bb 3372 /* diag_listed_as: To%s: illegal mapping '%s' */
319009ee
KW
3373 Perl_croak(aTHX_ "%s: illegal mapping '%s'", typestr, l);
3374 }
3375 }
3376 else
3377 *val = 0; /* bits == 1, then val should be ignored */
3378 }
3379
3380 /* Position to next line if any, or EOF */
3381 if (nl)
3382 l = nl + 1;
3383 else
3384 l = lend;
3385
3386 return l;
3387}
3388
979f2922
ST
3389/* Note:
3390 * Returns a swatch (a bit vector string) for a code point sequence
3391 * that starts from the value C<start> and comprises the number C<span>.
3392 * A C<swash> must be an object created by SWASHNEW (see lib/utf8_heavy.pl).
3393 * Should be used via swash_fetch, which will cache the swatch in C<swash>.
3394 */
3395STATIC SV*
b0e3252e 3396S_swatch_get(pTHX_ SV* swash, UV start, UV span)
979f2922
ST
3397{
3398 SV *swatch;
77f9f126 3399 U8 *l, *lend, *x, *xend, *s, *send;
979f2922 3400 STRLEN lcur, xcur, scur;
ef8f7699 3401 HV *const hv = MUTABLE_HV(SvRV(swash));
5c9f4bd2 3402 SV** const invlistsvp = hv_fetchs(hv, "V", FALSE);
36eb48b4 3403
88d45d28
KW
3404 SV** listsvp = NULL; /* The string containing the main body of the table */
3405 SV** extssvp = NULL;
3406 SV** invert_it_svp = NULL;
3407 U8* typestr = NULL;
786861f5
KW
3408 STRLEN bits;
3409 STRLEN octets; /* if bits == 1, then octets == 0 */
3410 UV none;
3411 UV end = start + span;
972dd592 3412
36eb48b4 3413 if (invlistsvp == NULL) {
786861f5
KW
3414 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
3415 SV** const nonesvp = hv_fetchs(hv, "NONE", FALSE);
3416 SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE);
3417 extssvp = hv_fetchs(hv, "EXTRAS", FALSE);
3418 listsvp = hv_fetchs(hv, "LIST", FALSE);
3419 invert_it_svp = hv_fetchs(hv, "INVERT_IT", FALSE);
3420
3421 bits = SvUV(*bitssvp);
3422 none = SvUV(*nonesvp);
3423 typestr = (U8*)SvPV_nolen(*typesvp);
3424 }
36eb48b4
KW
3425 else {
3426 bits = 1;
3427 none = 0;
3428 }
786861f5 3429 octets = bits >> 3; /* if bits == 1, then octets == 0 */
979f2922 3430
b0e3252e 3431 PERL_ARGS_ASSERT_SWATCH_GET;
7918f24d 3432
979f2922 3433 if (bits != 1 && bits != 8 && bits != 16 && bits != 32) {
b0e3252e 3434 Perl_croak(aTHX_ "panic: swatch_get doesn't expect bits %"UVuf,
660a4616 3435 (UV)bits);
979f2922
ST
3436 }
3437
84ea5ef6
KW
3438 /* If overflowed, use the max possible */
3439 if (end < start) {
3440 end = UV_MAX;
3441 span = end - start;
3442 }
3443
979f2922 3444 /* create and initialize $swatch */
979f2922 3445 scur = octets ? (span * octets) : (span + 7) / 8;
e524fe40
NC
3446 swatch = newSV(scur);
3447 SvPOK_on(swatch);
979f2922
ST
3448 s = (U8*)SvPVX(swatch);
3449 if (octets && none) {
0bd48802 3450 const U8* const e = s + scur;
979f2922
ST
3451 while (s < e) {
3452 if (bits == 8)
3453 *s++ = (U8)(none & 0xff);
3454 else if (bits == 16) {
3455 *s++ = (U8)((none >> 8) & 0xff);
3456 *s++ = (U8)( none & 0xff);
3457 }
3458 else if (bits == 32) {
3459 *s++ = (U8)((none >> 24) & 0xff);
3460 *s++ = (U8)((none >> 16) & 0xff);
3461 *s++ = (U8)((none >> 8) & 0xff);
3462 *s++ = (U8)( none & 0xff);
3463 }
3464 }
3465 *s = '\0';
3466 }
3467 else {
3468 (void)memzero((U8*)s, scur + 1);
3469 }
3470 SvCUR_set(swatch, scur);
3471 s = (U8*)SvPVX(swatch);
3472
36eb48b4
KW
3473 if (invlistsvp) { /* If has an inversion list set up use that */
3474 _invlist_populate_swatch(*invlistsvp, start, end, s);
3475 return swatch;
3476 }
3477
3478 /* read $swash->{LIST} */
979f2922
ST
3479 l = (U8*)SvPV(*listsvp, lcur);
3480 lend = l + lcur;
3481 while (l < lend) {
8ed25d53 3482 UV min, max, val, upper;
319009ee
KW
3483 l = S_swash_scan_list_line(aTHX_ l, lend, &min, &max, &val,
3484 cBOOL(octets), typestr);
3485 if (l > lend) {
979f2922
ST
3486 break;
3487 }
3488
972dd592 3489 /* If looking for something beyond this range, go try the next one */
979f2922
ST
3490 if (max < start)
3491 continue;
3492
8ed25d53
KW
3493 /* <end> is generally 1 beyond where we want to set things, but at the
3494 * platform's infinity, where we can't go any higher, we want to
3495 * include the code point at <end> */
3496 upper = (max < end)
3497 ? max
3498 : (max != UV_MAX || end != UV_MAX)
3499 ? end - 1
3500 : end;
3501
979f2922 3502 if (octets) {
35da51f7 3503 UV key;
979f2922
ST
3504 if (min < start) {
3505 if (!none || val < none) {
3506 val += start - min;
3507 }
3508 min = start;
3509 }
8ed25d53 3510 for (key = min; key <= upper; key++) {
979f2922 3511 STRLEN offset;
979f2922
ST
3512 /* offset must be non-negative (start <= min <= key < end) */
3513 offset = octets * (key - start);
3514 if (bits == 8)
3515 s[offset] = (U8)(val & 0xff);
3516 else if (bits == 16) {
3517 s[offset ] = (U8)((val >> 8) & 0xff);
3518 s[offset + 1] = (U8)( val & 0xff);
3519 }
3520 else if (bits == 32) {
3521 s[offset ] = (U8)((val >> 24) & 0xff);
3522 s[offset + 1] = (U8)((val >> 16) & 0xff);
3523 s[offset + 2] = (U8)((val >> 8) & 0xff);
3524 s[offset + 3] = (U8)( val & 0xff);
3525 }
3526
3527 if (!none || val < none)
3528 ++val;
3529 }
3530 }
711a919c 3531 else { /* bits == 1, then val should be ignored */
35da51f7 3532 UV key;
979f2922
ST
3533 if (min < start)
3534 min = start;
6cb05c12 3535
8ed25d53 3536 for (key = min; key <= upper; key++) {
0bd48802 3537 const STRLEN offset = (STRLEN)(key - start);
979f2922
ST
3538 s[offset >> 3] |= 1 << (offset & 7);
3539 }
3540 }
3541 } /* while */
979f2922 3542
9479a769 3543 /* Invert if the data says it should be. Assumes that bits == 1 */
77f9f126 3544 if (invert_it_svp && SvUV(*invert_it_svp)) {
0bda3001
KW
3545
3546 /* Unicode properties should come with all bits above PERL_UNICODE_MAX
3547 * be 0, and their inversion should also be 0, as we don't succeed any
3548 * Unicode property matches for non-Unicode code points */
3549 if (start <= PERL_UNICODE_MAX) {
3550
3551 /* The code below assumes that we never cross the
3552 * Unicode/above-Unicode boundary in a range, as otherwise we would
3553 * have to figure out where to stop flipping the bits. Since this
3554 * boundary is divisible by a large power of 2, and swatches comes
3555 * in small powers of 2, this should be a valid assumption */
3556 assert(start + span - 1 <= PERL_UNICODE_MAX);
3557
507a8485
KW
3558 send = s + scur;
3559 while (s < send) {
3560 *s = ~(*s);
3561 s++;
3562 }
0bda3001 3563 }
77f9f126
KW
3564 }
3565
d73c39c5
KW
3566 /* read $swash->{EXTRAS}
3567 * This code also copied to swash_to_invlist() below */
979f2922
ST
3568 x = (U8*)SvPV(*extssvp, xcur);
3569 xend = x + xcur;
3570 while (x < xend) {
3571 STRLEN namelen;
3572 U8 *namestr;
3573 SV** othersvp;
3574 HV* otherhv;
3575 STRLEN otherbits;
3576 SV **otherbitssvp, *other;
711a919c 3577 U8 *s, *o, *nl;
979f2922
ST
3578 STRLEN slen, olen;
3579
35da51f7 3580 const U8 opc = *x++;
979f2922
ST
3581 if (opc == '\n')
3582 continue;
3583
3584 nl = (U8*)memchr(x, '\n', xend - x);
3585
3586 if (opc != '-' && opc != '+' && opc != '!' && opc != '&') {
3587 if (nl) {
3588 x = nl + 1; /* 1 is length of "\n" */
3589 continue;
3590 }
3591 else {
3592 x = xend; /* to EXTRAS' end at which \n is not found */
3593 break;
3594 }
3595 }
3596
3597 namestr = x;
3598 if (nl) {
3599 namelen = nl - namestr;
3600 x = nl + 1;
3601 }
3602 else {
3603 namelen = xend - namestr;
3604 x = xend;
3605 }
3606
3607 othersvp = hv_fetch(hv, (char *)namestr, namelen, FALSE);
ef8f7699 3608 otherhv = MUTABLE_HV(SvRV(*othersvp));
017a3ce5 3609 otherbitssvp = hv_fetchs(otherhv, "BITS", FALSE);
979f2922
ST
3610 otherbits = (STRLEN)SvUV(*otherbitssvp);
3611 if (bits < otherbits)
5637ef5b
NC
3612 Perl_croak(aTHX_ "panic: swatch_get found swatch size mismatch, "
3613 "bits=%"UVuf", otherbits=%"UVuf, (UV)bits, (UV)otherbits);
979f2922
ST
3614
3615 /* The "other" swatch must be destroyed after. */
b0e3252e 3616 other = swatch_get(*othersvp, start, span);
979f2922
ST
3617 o = (U8*)SvPV(other, olen);
3618
3619 if (!olen)
b0e3252e 3620 Perl_croak(aTHX_ "panic: swatch_get got improper swatch");
979f2922
ST
3621
3622 s = (U8*)SvPV(swatch, slen);
3623 if (bits == 1 && otherbits == 1) {
3624 if (slen != olen)
5637ef5b
NC
3625 Perl_croak(aTHX_ "panic: swatch_get found swatch length "
3626 "mismatch, slen=%"UVuf", olen=%"UVuf,
3627 (UV)slen, (UV)olen);
979f2922
ST
3628
3629 switch (opc) {
3630 case '+':
3631 while (slen--)
3632 *s++ |= *o++;
3633 break;
3634 case '!':
3635 while (slen--)
3636 *s++ |= ~*o++;
3637 break;
3638 case '-':
3639 while (slen--)
3640 *s++ &= ~*o++;
3641 break;
3642 case '&':
3643 while (slen--)
3644 *s++ &= *o++;
3645 break;
3646 default:
3647 break;
3648 }
3649 }
711a919c 3650 else {
979f2922
ST
3651 STRLEN otheroctets = otherbits >> 3;
3652 STRLEN offset = 0;
35da51f7 3653 U8* const send = s + slen;
979f2922
ST
3654
3655 while (s < send) {
3656 UV otherval = 0;
3657
3658 if (otherbits == 1) {
3659 otherval = (o[offset >> 3] >> (offset & 7)) & 1;
3660 ++offset;
3661 }
3662 else {
3663 STRLEN vlen = otheroctets;
3664 otherval = *o++;
3665 while (--vlen) {
3666 otherval <<= 8;
3667 otherval |= *o++;
3668 }
3669 }
3670
711a919c 3671 if (opc == '+' && otherval)
6f207bd3 3672 NOOP; /* replace with otherval */
979f2922
ST
3673 else if (opc == '!' && !otherval)
3674 otherval = 1;
3675 else if (opc == '-' && otherval)
3676 otherval = 0;
3677 else if (opc == '&' && !otherval)
3678 otherval = 0;
3679 else {
711a919c 3680 s += octets; /* no replacement */
979f2922
ST
3681 continue;
3682 }
3683
3684 if (bits == 8)
3685 *s++ = (U8)( otherval & 0xff);
3686 else if (bits == 16) {
3687 *s++ = (U8)((otherval >> 8) & 0xff);
3688 *s++ = (U8)( otherval & 0xff);
3689 }
3690 else if (bits == 32) {
3691 *s++ = (U8)((otherval >> 24) & 0xff);
3692 *s++ = (U8)((otherval >> 16) & 0xff);
3693 *s++ = (U8)((otherval >> 8) & 0xff);
3694 *s++ = (U8)( otherval & 0xff);
3695 }