This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Convert uvuni_to_utf8() to function
[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);
40PERL_CALLCONV U8* Perl_uvchr_to_utf8(pTHX_ U8 *d, UV uv);
41#endif
42
27da23d5
JH
43static const char unees[] =
44 "Malformed UTF-8 character (unexpected end of string)";
901b21bf 45
48ef279e 46/*
ccfc67b7 47=head1 Unicode Support
a0ed51b3 48
166f8a29
DM
49This file contains various utility functions for manipulating UTF8-encoded
50strings. For the uninitiated, this is a method of representing arbitrary
61296642 51Unicode characters as a variable number of bytes, in such a way that
56da48f7
DM
52characters in the ASCII range are unmodified, and a zero byte never appears
53within non-zero characters.
166f8a29 54
eaf7a4d2
CS
55=cut
56*/
57
58/*
59=for apidoc is_ascii_string
60
a1433954 61Returns true if the first C<len> bytes of the string C<s> are the same whether
970ea3cb
KW
62or not the string is encoded in UTF-8 (or UTF-EBCDIC on EBCDIC machines). That
63is, if they are invariant. On ASCII-ish machines, only ASCII characters
64fit this definition, hence the function's name.
eaf7a4d2 65
9f7e3d64
MH
66If C<len> is 0, it will be calculated using C<strlen(s)>.
67
a1433954 68See also L</is_utf8_string>(), L</is_utf8_string_loclen>(), and L</is_utf8_string_loc>().
eaf7a4d2
CS
69
70=cut
71*/
72
73bool
668b6d8d 74Perl_is_ascii_string(const U8 *s, STRLEN len)
eaf7a4d2
CS
75{
76 const U8* const send = s + (len ? len : strlen((const char *)s));
77 const U8* x = s;
78
79 PERL_ARGS_ASSERT_IS_ASCII_STRING;
eaf7a4d2
CS
80
81 for (; x < send; ++x) {
82 if (!UTF8_IS_INVARIANT(*x))
83 break;
84 }
85
86 return x == send;
87}
88
89/*
87cea99e 90=for apidoc uvuni_to_utf8_flags
eebe1485 91
a27992cc
KW
92THIS FUNCTION SHOULD BE USED IN ONLY VERY SPECIALIZED CIRCUMSTANCES.
93
94It adds the UTF-8 representation of the Unicode code point C<uv> to the end
6ee84de2 95of the string C<d>; C<d> should have at least C<UTF8_MAXBYTES+1> free
eebe1485 96bytes available. The return value is the pointer to the byte after the
9041c2e3 97end of the new character. In other words,
eebe1485 98
b851fbc1
JH
99 d = uvuni_to_utf8_flags(d, uv, flags);
100
101or, in most cases,
102
b851fbc1
JH
103 d = uvuni_to_utf8_flags(d, uv, 0);
104
a27992cc 105This is the Unicode-aware way of saying
eebe1485
SC
106
107 *(d++) = uv;
108
537124e4
KW
109where uv is a code point expressed in Latin-1 or above, not the platform's
110native character set. B<Almost all code should instead use L</uvchr_to_utf8>
111or L</uvchr_to_utf8_flags>>.
112
949cf498
KW
113This function will convert to UTF-8 (and not warn) even code points that aren't
114legal Unicode or are problematic, unless C<flags> contains one or more of the
a1433954
KW
115following flags:
116
949cf498
KW
117If C<uv> is a Unicode surrogate code point and UNICODE_WARN_SURROGATE is set,
118the function will raise a warning, provided UTF8 warnings are enabled. If instead
119UNICODE_DISALLOW_SURROGATE is set, the function will fail and return NULL.
120If both flags are set, the function will both warn and return NULL.
121
122The UNICODE_WARN_NONCHAR and UNICODE_DISALLOW_NONCHAR flags correspondingly
537124e4
KW
123affect how the function handles a Unicode non-character. And likewise, the
124UNICODE_WARN_SUPER and UNICODE_DISALLOW_SUPER flags, affect the handling of
125code points that are
949cf498
KW
126above the Unicode maximum of 0x10FFFF. Code points above 0x7FFF_FFFF (which are
127even less portable) can be warned and/or disallowed even if other above-Unicode
128code points are accepted by the UNICODE_WARN_FE_FF and UNICODE_DISALLOW_FE_FF
129flags.
130
131And finally, the flag UNICODE_WARN_ILLEGAL_INTERCHANGE selects all four of the
132above WARN flags; and UNICODE_DISALLOW_ILLEGAL_INTERCHANGE selects all four
133DISALLOW flags.
134
135
eebe1485
SC
136=cut
137*/
138
dfe13c55 139U8 *
b851fbc1 140Perl_uvuni_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags)
a0ed51b3 141{
7918f24d
NC
142 PERL_ARGS_ASSERT_UVUNI_TO_UTF8_FLAGS;
143
979f77b6
KW
144 /* The first problematic code point is the first surrogate */
145 if (uv >= UNICODE_SURROGATE_FIRST
146 && ckWARN4_d(WARN_UTF8, WARN_SURROGATE, WARN_NON_UNICODE, WARN_NONCHAR))
147 {
949cf498
KW
148 if (UNICODE_IS_SURROGATE(uv)) {
149 if (flags & UNICODE_WARN_SURROGATE) {
8457b38f 150 Perl_ck_warner_d(aTHX_ packWARN(WARN_SURROGATE),
949cf498
KW
151 "UTF-16 surrogate U+%04"UVXf, uv);
152 }
153 if (flags & UNICODE_DISALLOW_SURROGATE) {
154 return NULL;
155 }
156 }
157 else if (UNICODE_IS_SUPER(uv)) {
158 if (flags & UNICODE_WARN_SUPER
159 || (UNICODE_IS_FE_FF(uv) && (flags & UNICODE_WARN_FE_FF)))
160 {
8457b38f 161 Perl_ck_warner_d(aTHX_ packWARN(WARN_NON_UNICODE),
949cf498
KW
162 "Code point 0x%04"UVXf" is not Unicode, may not be portable", uv);
163 }
164 if (flags & UNICODE_DISALLOW_SUPER
165 || (UNICODE_IS_FE_FF(uv) && (flags & UNICODE_DISALLOW_FE_FF)))
166 {
167 return NULL;
168 }
169 }
170 else if (UNICODE_IS_NONCHAR(uv)) {
171 if (flags & UNICODE_WARN_NONCHAR) {
8457b38f 172 Perl_ck_warner_d(aTHX_ packWARN(WARN_NONCHAR),
949cf498
KW
173 "Unicode non-character U+%04"UVXf" is illegal for open interchange",
174 uv);
175 }
176 if (flags & UNICODE_DISALLOW_NONCHAR) {
177 return NULL;
178 }
179 }
507b9800 180 }
c4d5f83a 181 if (UNI_IS_INVARIANT(uv)) {
bc3632a8 182 *d++ = (U8) I8_TO_NATIVE_UTF8(uv);
a0ed51b3
LW
183 return d;
184 }
2d331972 185#if defined(EBCDIC)
1d72bdf6
NIS
186 else {
187 STRLEN len = UNISKIP(uv);
188 U8 *p = d+len-1;
189 while (p > d) {
bc3632a8 190 *p-- = (U8) I8_TO_NATIVE_UTF8((uv & UTF_CONTINUATION_MASK) | UTF_CONTINUATION_MARK);
1d72bdf6
NIS
191 uv >>= UTF_ACCUMULATION_SHIFT;
192 }
bc3632a8 193 *p = (U8) I8_TO_NATIVE_UTF8((uv & UTF_START_MASK(len)) | UTF_START_MARK(len));
1d72bdf6
NIS
194 return d+len;
195 }
196#else /* Non loop style */
a0ed51b3 197 if (uv < 0x800) {
eb160463
GS
198 *d++ = (U8)(( uv >> 6) | 0xc0);
199 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
200 return d;
201 }
202 if (uv < 0x10000) {
eb160463
GS
203 *d++ = (U8)(( uv >> 12) | 0xe0);
204 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
205 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
206 return d;
207 }
208 if (uv < 0x200000) {
eb160463
GS
209 *d++ = (U8)(( uv >> 18) | 0xf0);
210 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
211 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
212 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
213 return d;
214 }
215 if (uv < 0x4000000) {
eb160463
GS
216 *d++ = (U8)(( uv >> 24) | 0xf8);
217 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
218 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
219 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
220 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
221 return d;
222 }
223 if (uv < 0x80000000) {
eb160463
GS
224 *d++ = (U8)(( uv >> 30) | 0xfc);
225 *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80);
226 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
227 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
228 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
229 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
230 return d;
231 }
6b8eaf93 232#ifdef HAS_QUAD
d7578b48 233 if (uv < UTF8_QUAD_MAX)
a0ed51b3
LW
234#endif
235 {
eb160463
GS
236 *d++ = 0xfe; /* Can't match U+FEFF! */
237 *d++ = (U8)(((uv >> 30) & 0x3f) | 0x80);
238 *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80);
239 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
240 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
241 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
242 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
243 return d;
244 }
6b8eaf93 245#ifdef HAS_QUAD
a0ed51b3 246 {
eb160463
GS
247 *d++ = 0xff; /* Can't match U+FFFE! */
248 *d++ = 0x80; /* 6 Reserved bits */
249 *d++ = (U8)(((uv >> 60) & 0x0f) | 0x80); /* 2 Reserved bits */
250 *d++ = (U8)(((uv >> 54) & 0x3f) | 0x80);
251 *d++ = (U8)(((uv >> 48) & 0x3f) | 0x80);
252 *d++ = (U8)(((uv >> 42) & 0x3f) | 0x80);
253 *d++ = (U8)(((uv >> 36) & 0x3f) | 0x80);
254 *d++ = (U8)(((uv >> 30) & 0x3f) | 0x80);
255 *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80);
256 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
257 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
258 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
259 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
260 return d;
261 }
262#endif
537124e4 263#endif /* Non loop style */
a0ed51b3 264}
9041c2e3 265
646ca15d
JH
266/*
267
f7d739d1 268Tests if the first C<len> bytes of string C<s> form a valid UTF-8
646ca15d 269character. Note that an INVARIANT (i.e. ASCII) character is a valid
f7d739d1 270UTF-8 character. The number of bytes in the UTF-8 character
646ca15d
JH
271will be returned if it is valid, otherwise 0.
272
273This is the "slow" version as opposed to the "fast" version which is
274the "unrolled" IS_UTF8_CHAR(). E.g. for t/uni/class.t the speed
275difference is a factor of 2 to 3. For lengths (UTF8SKIP(s)) of four
276or less you should use the IS_UTF8_CHAR(), for lengths of five or more
277you should use the _slow(). In practice this means that the _slow()
278will be used very rarely, since the maximum Unicode code point (as of
279Unicode 4.1) is U+10FFFF, which encodes in UTF-8 to four bytes. Only
537124e4 280the "Perl extended UTF-8" (e.g, the infamous 'v-strings') will encode into
646ca15d
JH
281five bytes or more.
282
283=cut */
7af276bc 284PERL_STATIC_INLINE STRLEN
5f66b61c 285S_is_utf8_char_slow(const U8 *s, const STRLEN len)
646ca15d 286{
cd7e6c88 287 dTHX; /* The function called below requires thread context */
646ca15d 288
cd7e6c88 289 STRLEN actual_len;
646ca15d 290
cd7e6c88 291 PERL_ARGS_ASSERT_IS_UTF8_CHAR_SLOW;
646ca15d 292
18712bce 293 utf8n_to_uvchr(s, len, &actual_len, UTF8_CHECK_ONLY);
646ca15d 294
cd7e6c88 295 return (actual_len == (STRLEN) -1) ? 0 : actual_len;
646ca15d 296}
9041c2e3
NIS
297
298/*
492a624f
KW
299=for apidoc is_utf8_char_buf
300
301Returns the number of bytes that comprise the first UTF-8 encoded character in
302buffer C<buf>. C<buf_end> should point to one position beyond the end of the
303buffer. 0 is returned if C<buf> does not point to a complete, valid UTF-8
304encoded character.
305
306Note that an INVARIANT character (i.e. ASCII on non-EBCDIC
307machines) is a valid UTF-8 character.
308
309=cut */
310
311STRLEN
312Perl_is_utf8_char_buf(const U8 *buf, const U8* buf_end)
313{
314
315 STRLEN len;
316
317 PERL_ARGS_ASSERT_IS_UTF8_CHAR_BUF;
318
319 if (buf_end <= buf) {
320 return 0;
321 }
322
323 len = buf_end - buf;
324 if (len > UTF8SKIP(buf)) {
325 len = UTF8SKIP(buf);
326 }
327
328#ifdef IS_UTF8_CHAR
329 if (IS_UTF8_CHAR_FAST(len))
330 return IS_UTF8_CHAR(buf, len) ? len : 0;
331#endif /* #ifdef IS_UTF8_CHAR */
332 return is_utf8_char_slow(buf, len);
333}
334
335/*
87cea99e 336=for apidoc is_utf8_char
eebe1485 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)) {
38bce45e 623 return uv;
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
dcd27b3c
KW
642 /* Here is not a continuation byte, nor an invariant. The only thing left
643 * is a start byte (possibly for an overlong) */
644
1d72bdf6 645#ifdef EBCDIC
bc3632a8 646 uv = NATIVE_UTF8_TO_I8(uv);
1d72bdf6
NIS
647#endif
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{
010ab96b
KW
934 UV expectlen = UTF8SKIP(s);
935 const U8* send = s + expectlen;
936 UV uv = NATIVE_UTF8_TO_I8(*s);
3986bb7c 937
27d6c58a
KW
938 PERL_ARGS_ASSERT_VALID_UTF8_TO_UVCHR;
939
010ab96b
KW
940 if (retlen) {
941 *retlen = expectlen;
942 }
943
944 /* An invariant is trivially returned */
945 if (expectlen == 1) {
946 return LATIN1_TO_NATIVE(uv);
947 }
948
949 /* Remove the leading bits that indicate the number of bytes, leaving just
950 * the bits that are part of the value */
951 uv &= UTF_START_MASK(expectlen);
952
953 /* Now, loop through the remaining bytes, accumulating each into the
954 * working total as we go. (I khw tried unrolling the loop for up to 4
955 * bytes, but there was no performance improvement) */
956 for (++s; s < send; s++) {
957 uv = UTF8_ACCUMULATE(uv, *s);
958 }
959
3986bb7c 960 return UNI_TO_NATIVE(uv);
010ab96b 961
27d6c58a
KW
962}
963
ec5f19d0 964/*
87cea99e 965=for apidoc utf8_to_uvchr
9041c2e3 966
6ee84de2 967Returns the native code point of the first character in the string C<s>
1e54db1a 968which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
9041c2e3
NIS
969length, in bytes, of that character.
970
4b88fb76 971Some, but not all, UTF-8 malformations are detected, and in fact, some
977c1d31
KW
972malformed input could cause reading beyond the end of the input buffer, which
973is why this function is deprecated. Use L</utf8_to_uvchr_buf> instead.
4b88fb76 974
524080c4
KW
975If C<s> points to one of the detected malformations, and UTF8 warnings are
976enabled, zero is returned and C<*retlen> is set (if C<retlen> isn't
977NULL) to -1. If those warnings are off, the computed value if well-defined (or
978the Unicode REPLACEMENT CHARACTER, if not) is silently returned, and C<*retlen>
979is set (if C<retlen> isn't NULL) so that (S<C<s> + C<*retlen>>) is the
980next possible position in C<s> that could begin a non-malformed character.
981See L</utf8n_to_uvuni> for details on when the REPLACEMENT CHARACTER is returned.
9041c2e3
NIS
982
983=cut
984*/
985
986UV
7fc63493 987Perl_utf8_to_uvchr(pTHX_ const U8 *s, STRLEN *retlen)
9041c2e3 988{
7918f24d
NC
989 PERL_ARGS_ASSERT_UTF8_TO_UVCHR;
990
2ff6c191 991 return utf8_to_uvchr_buf(s, s + UTF8_MAXBYTES, retlen);
9041c2e3
NIS
992}
993
994/*
ec5f19d0
KW
995=for apidoc utf8_to_uvuni_buf
996
4f83cdcd
KW
997Only in very rare circumstances should code need to be dealing in the Unicode
998code point. Use L</utf8_to_uvchr_buf> instead.
999
1000Returns the Unicode (not-native) code point of the first character in the
1001string C<s> which
ec5f19d0
KW
1002is assumed to be in UTF-8 encoding; C<send> points to 1 beyond the end of C<s>.
1003C<retlen> will be set to the length, in bytes, of that character.
1004
524080c4
KW
1005If C<s> does not point to a well-formed UTF-8 character and UTF8 warnings are
1006enabled, zero is returned and C<*retlen> is set (if C<retlen> isn't
1007NULL) to -1. If those warnings are off, the computed value if well-defined (or
1008the Unicode REPLACEMENT CHARACTER, if not) is silently returned, and C<*retlen>
1009is set (if C<retlen> isn't NULL) so that (S<C<s> + C<*retlen>>) is the
1010next possible position in C<s> that could begin a non-malformed character.
1011See L</utf8n_to_uvuni> for details on when the REPLACEMENT CHARACTER is returned.
ec5f19d0
KW
1012
1013=cut
1014*/
1015
1016UV
1017Perl_utf8_to_uvuni_buf(pTHX_ const U8 *s, const U8 *send, STRLEN *retlen)
1018{
1019 PERL_ARGS_ASSERT_UTF8_TO_UVUNI_BUF;
1020
1021 assert(send > s);
1022
1023 /* Call the low level routine asking for checks */
1024 return Perl_utf8n_to_uvuni(aTHX_ s, send -s, retlen,
1025 ckWARN_d(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
1026}
1027
5495102a
KW
1028/* DEPRECATED!
1029 * Like L</utf8_to_uvuni_buf>(), but should only be called when it is known that
2114036c 1030 * there are no malformations in the input UTF-8 string C<s>. Surrogates,
3986bb7c 1031 * non-character code points, and non-Unicode code points are allowed */
27d6c58a
KW
1032
1033UV
1034Perl_valid_utf8_to_uvuni(pTHX_ const U8 *s, STRLEN *retlen)
1035{
1036 PERL_ARGS_ASSERT_VALID_UTF8_TO_UVUNI;
1037
010ab96b 1038 return NATIVE_TO_UNI(valid_utf8_to_uvchr(s, retlen));
27d6c58a
KW
1039}
1040
ec5f19d0 1041/*
87cea99e 1042=for apidoc utf8_to_uvuni
9041c2e3
NIS
1043
1044Returns the Unicode code point of the first character in the string C<s>
1e54db1a 1045which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
9041c2e3
NIS
1046length, in bytes, of that character.
1047
4b88fb76 1048Some, but not all, UTF-8 malformations are detected, and in fact, some
977c1d31 1049malformed input could cause reading beyond the end of the input buffer, which
4f83cdcd
KW
1050is one reason why this function is deprecated. The other is that only in
1051extremely limited circumstances should the Unicode versus native code point be
1052of any interest to you. Use L</utf8_to_uvchr_buf> instead.
9041c2e3 1053
524080c4
KW
1054If C<s> points to one of the detected malformations, and UTF8 warnings are
1055enabled, zero is returned and C<*retlen> is set (if C<retlen> doesn't point to
1056NULL) to -1. If those warnings are off, the computed value if well-defined (or
1057the Unicode REPLACEMENT CHARACTER, if not) is silently returned, and C<*retlen>
1058is set (if C<retlen> isn't NULL) so that (S<C<s> + C<*retlen>>) is the
1059next possible position in C<s> that could begin a non-malformed character.
1060See L</utf8n_to_uvuni> for details on when the REPLACEMENT CHARACTER is returned.
8e84507e
NIS
1061
1062=cut
1063*/
1064
1065UV
7fc63493 1066Perl_utf8_to_uvuni(pTHX_ const U8 *s, STRLEN *retlen)
8e84507e 1067{
7918f24d
NC
1068 PERL_ARGS_ASSERT_UTF8_TO_UVUNI;
1069
5495102a 1070 return NATIVE_TO_UNI(valid_utf8_to_uvchr(s, retlen));
8e84507e
NIS
1071}
1072
b76347f2 1073/*
87cea99e 1074=for apidoc utf8_length
b76347f2
JH
1075
1076Return the length of the UTF-8 char encoded string C<s> in characters.
02eb7b47
JH
1077Stops at C<e> (inclusive). If C<e E<lt> s> or if the scan would end
1078up past C<e>, croaks.
b76347f2
JH
1079
1080=cut
1081*/
1082
1083STRLEN
35a4481c 1084Perl_utf8_length(pTHX_ const U8 *s, const U8 *e)
b76347f2 1085{
97aff369 1086 dVAR;
b76347f2
JH
1087 STRLEN len = 0;
1088
7918f24d
NC
1089 PERL_ARGS_ASSERT_UTF8_LENGTH;
1090
8850bf83
JH
1091 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g.
1092 * the bitops (especially ~) can create illegal UTF-8.
1093 * In other words: in Perl UTF-8 is not just for Unicode. */
1094
a3b680e6
AL
1095 if (e < s)
1096 goto warn_and_return;
b76347f2 1097 while (s < e) {
4cbf4130 1098 s += UTF8SKIP(s);
8e91ec7f
AV
1099 len++;
1100 }
1101
1102 if (e != s) {
1103 len--;
1104 warn_and_return:
9b387841
NC
1105 if (PL_op)
1106 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8),
1107 "%s in %s", unees, OP_DESC(PL_op));
1108 else
61a12c31 1109 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), "%s", unees);
b76347f2
JH
1110 }
1111
1112 return len;
1113}
1114
b06226ff 1115/*
87cea99e 1116=for apidoc utf8_distance
b06226ff 1117
1e54db1a 1118Returns the number of UTF-8 characters between the UTF-8 pointers C<a>
b06226ff
JH
1119and C<b>.
1120
1121WARNING: use only if you *know* that the pointers point inside the
1122same UTF-8 buffer.
1123
37607a96
PK
1124=cut
1125*/
a0ed51b3 1126
02eb7b47 1127IV
35a4481c 1128Perl_utf8_distance(pTHX_ const U8 *a, const U8 *b)
a0ed51b3 1129{
7918f24d
NC
1130 PERL_ARGS_ASSERT_UTF8_DISTANCE;
1131
bf1665bc 1132 return (a < b) ? -1 * (IV) utf8_length(a, b) : (IV) utf8_length(b, a);
a0ed51b3
LW
1133}
1134
b06226ff 1135/*
87cea99e 1136=for apidoc utf8_hop
b06226ff 1137
8850bf83
JH
1138Return the UTF-8 pointer C<s> displaced by C<off> characters, either
1139forward or backward.
b06226ff
JH
1140
1141WARNING: do not use the following unless you *know* C<off> is within
8850bf83
JH
1142the UTF-8 data pointed to by C<s> *and* that on entry C<s> is aligned
1143on the first byte of character or just after the last byte of a character.
b06226ff 1144
37607a96
PK
1145=cut
1146*/
a0ed51b3
LW
1147
1148U8 *
4373e329 1149Perl_utf8_hop(pTHX_ const U8 *s, I32 off)
a0ed51b3 1150{
7918f24d
NC
1151 PERL_ARGS_ASSERT_UTF8_HOP;
1152
96a5add6 1153 PERL_UNUSED_CONTEXT;
8850bf83
JH
1154 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g
1155 * the bitops (especially ~) can create illegal UTF-8.
1156 * In other words: in Perl UTF-8 is not just for Unicode. */
1157
a0ed51b3
LW
1158 if (off >= 0) {
1159 while (off--)
1160 s += UTF8SKIP(s);
1161 }
1162 else {
1163 while (off++) {
1164 s--;
8850bf83
JH
1165 while (UTF8_IS_CONTINUATION(*s))
1166 s--;
a0ed51b3
LW
1167 }
1168 }
4373e329 1169 return (U8 *)s;
a0ed51b3
LW
1170}
1171
6940069f 1172/*
fed3ba5d
NC
1173=for apidoc bytes_cmp_utf8
1174
a1433954
KW
1175Compares the sequence of characters (stored as octets) in C<b>, C<blen> with the
1176sequence of characters (stored as UTF-8) in C<u>, C<ulen>. Returns 0 if they are
fed3ba5d
NC
1177equal, -1 or -2 if the first string is less than the second string, +1 or +2
1178if the first string is greater than the second string.
1179
1180-1 or +1 is returned if the shorter string was identical to the start of the
1181longer string. -2 or +2 is returned if the was a difference between characters
1182within the strings.
1183
1184=cut
1185*/
1186
1187int
1188Perl_bytes_cmp_utf8(pTHX_ const U8 *b, STRLEN blen, const U8 *u, STRLEN ulen)
1189{
1190 const U8 *const bend = b + blen;
1191 const U8 *const uend = u + ulen;
1192
1193 PERL_ARGS_ASSERT_BYTES_CMP_UTF8;
1194
1195 PERL_UNUSED_CONTEXT;
1196
1197 while (b < bend && u < uend) {
1198 U8 c = *u++;
1199 if (!UTF8_IS_INVARIANT(c)) {
1200 if (UTF8_IS_DOWNGRADEABLE_START(c)) {
1201 if (u < uend) {
1202 U8 c1 = *u++;
1203 if (UTF8_IS_CONTINUATION(c1)) {
94bb8c36 1204 c = TWO_BYTE_UTF8_TO_NATIVE(c, c1);
fed3ba5d
NC
1205 } else {
1206 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8),
1207 "Malformed UTF-8 character "
1208 "(unexpected non-continuation byte 0x%02x"
1209 ", immediately after start byte 0x%02x)"
1210 /* Dear diag.t, it's in the pod. */
1211 "%s%s", c1, c,
1212 PL_op ? " in " : "",
1213 PL_op ? OP_DESC(PL_op) : "");
1214 return -2;
1215 }
1216 } else {
1217 if (PL_op)
1218 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8),
1219 "%s in %s", unees, OP_DESC(PL_op));
1220 else
61a12c31 1221 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), "%s", unees);
fed3ba5d
NC
1222 return -2; /* Really want to return undef :-) */
1223 }
1224 } else {
1225 return -2;
1226 }
1227 }
1228 if (*b != c) {
1229 return *b < c ? -2 : +2;
1230 }
1231 ++b;
1232 }
1233
1234 if (b == bend && u == uend)
1235 return 0;
1236
1237 return b < bend ? +1 : -1;
1238}
1239
1240/*
87cea99e 1241=for apidoc utf8_to_bytes
6940069f 1242
2bbc8d55 1243Converts a string C<s> of length C<len> from UTF-8 into native byte encoding.
a1433954
KW
1244Unlike L</bytes_to_utf8>, this over-writes the original string, and
1245updates C<len> to contain the new length.
67e989fb 1246Returns zero on failure, setting C<len> to -1.
6940069f 1247
a1433954 1248If you need a copy of the string, see L</bytes_from_utf8>.
95be277c 1249
6940069f
GS
1250=cut
1251*/
1252
1253U8 *
37607a96 1254Perl_utf8_to_bytes(pTHX_ U8 *s, STRLEN *len)
6940069f 1255{
d4c19fe8
AL
1256 U8 * const save = s;
1257 U8 * const send = s + *len;
6940069f 1258 U8 *d;
246fae53 1259
7918f24d
NC
1260 PERL_ARGS_ASSERT_UTF8_TO_BYTES;
1261
1e54db1a 1262 /* ensure valid UTF-8 and chars < 256 before updating string */
d4c19fe8 1263 while (s < send) {
dcad2880
JH
1264 U8 c = *s++;
1265
1d72bdf6
NIS
1266 if (!UTF8_IS_INVARIANT(c) &&
1267 (!UTF8_IS_DOWNGRADEABLE_START(c) || (s >= send)
1268 || !(c = *s++) || !UTF8_IS_CONTINUATION(c))) {
10edeb5d 1269 *len = ((STRLEN) -1);
dcad2880
JH
1270 return 0;
1271 }
246fae53 1272 }
dcad2880
JH
1273
1274 d = s = save;
6940069f 1275 while (s < send) {
ed646e6e 1276 STRLEN ulen;
4b88fb76 1277 *d++ = (U8)utf8_to_uvchr_buf(s, send, &ulen);
ed646e6e 1278 s += ulen;
6940069f
GS
1279 }
1280 *d = '\0';
246fae53 1281 *len = d - save;
6940069f
GS
1282 return save;
1283}
1284
1285/*
87cea99e 1286=for apidoc bytes_from_utf8
f9a63242 1287
2bbc8d55 1288Converts a string C<s> of length C<len> from UTF-8 into native byte encoding.
a1433954 1289Unlike L</utf8_to_bytes> but like L</bytes_to_utf8>, returns a pointer to
ef9edfd0
JH
1290the newly-created string, and updates C<len> to contain the new
1291length. Returns the original string if no conversion occurs, C<len>
1292is unchanged. Do nothing if C<is_utf8> points to 0. Sets C<is_utf8> to
2bbc8d55
SP
12930 if C<s> is converted or consisted entirely of characters that are invariant
1294in utf8 (i.e., US-ASCII on non-EBCDIC machines).
f9a63242 1295
37607a96
PK
1296=cut
1297*/
f9a63242
JH
1298
1299U8 *
e1ec3a88 1300Perl_bytes_from_utf8(pTHX_ const U8 *s, STRLEN *len, bool *is_utf8)
f9a63242 1301{
f9a63242 1302 U8 *d;
e1ec3a88
AL
1303 const U8 *start = s;
1304 const U8 *send;
f9a63242
JH
1305 I32 count = 0;
1306
7918f24d
NC
1307 PERL_ARGS_ASSERT_BYTES_FROM_UTF8;
1308
96a5add6 1309 PERL_UNUSED_CONTEXT;
f9a63242 1310 if (!*is_utf8)
73d840c0 1311 return (U8 *)start;
f9a63242 1312
1e54db1a 1313 /* ensure valid UTF-8 and chars < 256 before converting string */
f9a63242 1314 for (send = s + *len; s < send;) {
e1ec3a88 1315 U8 c = *s++;
1d72bdf6 1316 if (!UTF8_IS_INVARIANT(c)) {
db42d148
NIS
1317 if (UTF8_IS_DOWNGRADEABLE_START(c) && s < send &&
1318 (c = *s++) && UTF8_IS_CONTINUATION(c))
1319 count++;
1320 else
73d840c0 1321 return (U8 *)start;
db42d148 1322 }
f9a63242
JH
1323 }
1324
35da51f7 1325 *is_utf8 = FALSE;
f9a63242 1326
212542aa 1327 Newx(d, (*len) - count + 1, U8);
ef9edfd0 1328 s = start; start = d;
f9a63242
JH
1329 while (s < send) {
1330 U8 c = *s++;
c4d5f83a
NIS
1331 if (!UTF8_IS_INVARIANT(c)) {
1332 /* Then it is two-byte encoded */
94bb8c36 1333 c = TWO_BYTE_UTF8_TO_NATIVE(c, *s++);
c4d5f83a
NIS
1334 }
1335 *d++ = c;
f9a63242
JH
1336 }
1337 *d = '\0';
1338 *len = d - start;
73d840c0 1339 return (U8 *)start;
f9a63242
JH
1340}
1341
1342/*
87cea99e 1343=for apidoc bytes_to_utf8
6940069f 1344
ff97e5cf
KW
1345Converts a string C<s> of length C<len> bytes from the native encoding into
1346UTF-8.
6662521e 1347Returns a pointer to the newly-created string, and sets C<len> to
ff97e5cf 1348reflect the new length in bytes.
6940069f 1349
2bbc8d55
SP
1350A NUL character will be written after the end of the string.
1351
1352If you want to convert to UTF-8 from encodings other than
1353the native (Latin1 or EBCDIC),
a1433954 1354see L</sv_recode_to_utf8>().
c9ada85f 1355
497711e7 1356=cut
6940069f
GS
1357*/
1358
c682ebef
FC
1359/* This logic is duplicated in sv_catpvn_flags, so any bug fixes will
1360 likewise need duplication. */
1361
6940069f 1362U8*
35a4481c 1363Perl_bytes_to_utf8(pTHX_ const U8 *s, STRLEN *len)
6940069f 1364{
35a4481c 1365 const U8 * const send = s + (*len);
6940069f
GS
1366 U8 *d;
1367 U8 *dst;
7918f24d
NC
1368
1369 PERL_ARGS_ASSERT_BYTES_TO_UTF8;
96a5add6 1370 PERL_UNUSED_CONTEXT;
6940069f 1371
212542aa 1372 Newx(d, (*len) * 2 + 1, U8);
6940069f
GS
1373 dst = d;
1374
1375 while (s < send) {
55d09dc8
KW
1376 append_utf8_from_native_byte(*s, &d);
1377 s++;
6940069f
GS
1378 }
1379 *d = '\0';
6662521e 1380 *len = d-dst;
6940069f
GS
1381 return dst;
1382}
1383
a0ed51b3 1384/*
dea0fc0b 1385 * Convert native (big-endian) or reversed (little-endian) UTF-16 to UTF-8.
a0ed51b3
LW
1386 *
1387 * Destination must be pre-extended to 3/2 source. Do not use in-place.
1388 * We optimize for native, for obvious reasons. */
1389
1390U8*
dea0fc0b 1391Perl_utf16_to_utf8(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen)
a0ed51b3 1392{
dea0fc0b
JH
1393 U8* pend;
1394 U8* dstart = d;
1395
7918f24d
NC
1396 PERL_ARGS_ASSERT_UTF16_TO_UTF8;
1397
dea0fc0b 1398 if (bytelen & 1)
f5992bc4 1399 Perl_croak(aTHX_ "panic: utf16_to_utf8: odd bytelen %"UVuf, (UV)bytelen);
dea0fc0b
JH
1400
1401 pend = p + bytelen;
1402
a0ed51b3 1403 while (p < pend) {
dea0fc0b
JH
1404 UV uv = (p[0] << 8) + p[1]; /* UTF-16BE */
1405 p += 2;
a0ed51b3 1406 if (uv < 0x80) {
eb160463 1407 *d++ = (U8)uv;
a0ed51b3
LW
1408 continue;
1409 }
1410 if (uv < 0x800) {
eb160463
GS
1411 *d++ = (U8)(( uv >> 6) | 0xc0);
1412 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
1413 continue;
1414 }
46956fad
KW
1415#define FIRST_HIGH_SURROGATE UNICODE_SURROGATE_FIRST
1416#define LAST_HIGH_SURROGATE 0xDBFF
1417#define FIRST_LOW_SURROGATE 0xDC00
1418#define LAST_LOW_SURROGATE UNICODE_SURROGATE_LAST
1419 if (uv >= FIRST_HIGH_SURROGATE && uv <= LAST_HIGH_SURROGATE) {
01ea242b 1420 if (p >= pend) {
dea0fc0b 1421 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
01ea242b
NC
1422 } else {
1423 UV low = (p[0] << 8) + p[1];
1424 p += 2;
46956fad 1425 if (low < FIRST_LOW_SURROGATE || low > LAST_LOW_SURROGATE)
01ea242b 1426 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
46956fad
KW
1427 uv = ((uv - FIRST_HIGH_SURROGATE) << 10)
1428 + (low - FIRST_LOW_SURROGATE) + 0x10000;
01ea242b 1429 }
46956fad 1430 } else if (uv >= FIRST_LOW_SURROGATE && uv <= LAST_LOW_SURROGATE) {
dbde1951 1431 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
a0ed51b3
LW
1432 }
1433 if (uv < 0x10000) {
eb160463
GS
1434 *d++ = (U8)(( uv >> 12) | 0xe0);
1435 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
1436 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
1437 continue;
1438 }
1439 else {
eb160463
GS
1440 *d++ = (U8)(( uv >> 18) | 0xf0);
1441 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
1442 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
1443 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
1444 continue;
1445 }
1446 }
dea0fc0b 1447 *newlen = d - dstart;
a0ed51b3
LW
1448 return d;
1449}
1450
1451/* Note: this one is slightly destructive of the source. */
1452
1453U8*
dea0fc0b 1454Perl_utf16_to_utf8_reversed(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen)
a0ed51b3
LW
1455{
1456 U8* s = (U8*)p;
d4c19fe8 1457 U8* const send = s + bytelen;
7918f24d
NC
1458
1459 PERL_ARGS_ASSERT_UTF16_TO_UTF8_REVERSED;
1460
e0ea5e2d
NC
1461 if (bytelen & 1)
1462 Perl_croak(aTHX_ "panic: utf16_to_utf8_reversed: odd bytelen %"UVuf,
1463 (UV)bytelen);
1464
a0ed51b3 1465 while (s < send) {
d4c19fe8 1466 const U8 tmp = s[0];
a0ed51b3
LW
1467 s[0] = s[1];
1468 s[1] = tmp;
1469 s += 2;
1470 }
dea0fc0b 1471 return utf16_to_utf8(p, d, bytelen, newlen);
a0ed51b3
LW
1472}
1473
922e8cb4
KW
1474bool
1475Perl__is_uni_FOO(pTHX_ const U8 classnum, const UV c)
1476{
1477 U8 tmpbuf[UTF8_MAXBYTES+1];
1478 uvchr_to_utf8(tmpbuf, c);
1479 return _is_utf8_FOO(classnum, tmpbuf);
1480}
1481
c3fd2246
KW
1482/* for now these are all defined (inefficiently) in terms of the utf8 versions.
1483 * Note that the macros in handy.h that call these short-circuit calling them
1484 * for Latin-1 range inputs */
a0ed51b3
LW
1485
1486bool
84afefe6 1487Perl_is_uni_alnum(pTHX_ UV c)
a0ed51b3 1488{
89ebb4a3 1489 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1490 uvchr_to_utf8(tmpbuf, c);
922e8cb4 1491 return _is_utf8_FOO(_CC_WORDCHAR, tmpbuf);
a0ed51b3
LW
1492}
1493
1494bool
5092f92a
KW
1495Perl_is_uni_alnumc(pTHX_ UV c)
1496{
1497 U8 tmpbuf[UTF8_MAXBYTES+1];
1498 uvchr_to_utf8(tmpbuf, c);
922e8cb4 1499 return _is_utf8_FOO(_CC_ALPHANUMERIC, tmpbuf);
5092f92a
KW
1500}
1501
f9ae8fb6
JD
1502/* Internal function so we can deprecate the external one, and call
1503 this one from other deprecated functions in this file */
1504
1505PERL_STATIC_INLINE bool
61b19385
KW
1506S_is_utf8_idfirst(pTHX_ const U8 *p)
1507{
1508 dVAR;
1509
1510 if (*p == '_')
1511 return TRUE;
1512 /* is_utf8_idstart would be more logical. */
1513 return is_utf8_common(p, &PL_utf8_idstart, "IdStart");
1514}
1515
5092f92a 1516bool
84afefe6 1517Perl_is_uni_idfirst(pTHX_ UV c)
a0ed51b3 1518{
89ebb4a3 1519 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1520 uvchr_to_utf8(tmpbuf, c);
61b19385 1521 return S_is_utf8_idfirst(aTHX_ tmpbuf);
a0ed51b3
LW
1522}
1523
1524bool
eba68aa0
KW
1525Perl__is_uni_perl_idcont(pTHX_ UV c)
1526{
1527 U8 tmpbuf[UTF8_MAXBYTES+1];
1528 uvchr_to_utf8(tmpbuf, c);
1529 return _is_utf8_perl_idcont(tmpbuf);
1530}
1531
1532bool
f91dcd13
KW
1533Perl__is_uni_perl_idstart(pTHX_ UV c)
1534{
1535 U8 tmpbuf[UTF8_MAXBYTES+1];
1536 uvchr_to_utf8(tmpbuf, c);
1537 return _is_utf8_perl_idstart(tmpbuf);
1538}
1539
1540bool
84afefe6 1541Perl_is_uni_alpha(pTHX_ UV c)
a0ed51b3 1542{
89ebb4a3 1543 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1544 uvchr_to_utf8(tmpbuf, c);
922e8cb4 1545 return _is_utf8_FOO(_CC_ALPHA, tmpbuf);
a0ed51b3
LW
1546}
1547
1548bool
84afefe6 1549Perl_is_uni_ascii(pTHX_ UV c)
4d61ec05 1550{
bc39fe24 1551 return isASCII(c);
4d61ec05
GS
1552}
1553
1554bool
bdd8600f
KW
1555Perl_is_uni_blank(pTHX_ UV c)
1556{
2cafb56b 1557 return isBLANK_uni(c);
bdd8600f
KW
1558}
1559
1560bool
84afefe6 1561Perl_is_uni_space(pTHX_ UV c)
a0ed51b3 1562{
add4123a 1563 return isSPACE_uni(c);
a0ed51b3
LW
1564}
1565
1566bool
84afefe6 1567Perl_is_uni_digit(pTHX_ UV c)
a0ed51b3 1568{
89ebb4a3 1569 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1570 uvchr_to_utf8(tmpbuf, c);
922e8cb4 1571 return _is_utf8_FOO(_CC_DIGIT, tmpbuf);
a0ed51b3
LW
1572}
1573
1574bool
84afefe6 1575Perl_is_uni_upper(pTHX_ UV c)
a0ed51b3 1576{
89ebb4a3 1577 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1578 uvchr_to_utf8(tmpbuf, c);
922e8cb4 1579 return _is_utf8_FOO(_CC_UPPER, tmpbuf);
a0ed51b3
LW
1580}
1581
1582bool
84afefe6 1583Perl_is_uni_lower(pTHX_ UV c)
a0ed51b3 1584{
89ebb4a3 1585 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1586 uvchr_to_utf8(tmpbuf, c);
922e8cb4 1587 return _is_utf8_FOO(_CC_LOWER, tmpbuf);
a0ed51b3
LW
1588}
1589
1590bool
84afefe6 1591Perl_is_uni_cntrl(pTHX_ UV c)
b8c5462f 1592{
7b952154 1593 return isCNTRL_L1(c);
b8c5462f
JH
1594}
1595
1596bool
84afefe6 1597Perl_is_uni_graph(pTHX_ UV c)
b8c5462f 1598{
89ebb4a3 1599 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1600 uvchr_to_utf8(tmpbuf, c);
922e8cb4 1601 return _is_utf8_FOO(_CC_GRAPH, tmpbuf);
b8c5462f
JH
1602}
1603
1604bool
84afefe6 1605Perl_is_uni_print(pTHX_ UV c)
a0ed51b3 1606{
89ebb4a3 1607 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1608 uvchr_to_utf8(tmpbuf, c);
922e8cb4 1609 return _is_utf8_FOO(_CC_PRINT, tmpbuf);
a0ed51b3
LW
1610}
1611
b8c5462f 1612bool
84afefe6 1613Perl_is_uni_punct(pTHX_ UV c)
b8c5462f 1614{
89ebb4a3 1615 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1616 uvchr_to_utf8(tmpbuf, c);
922e8cb4 1617 return _is_utf8_FOO(_CC_PUNCT, tmpbuf);
b8c5462f
JH
1618}
1619
4d61ec05 1620bool
84afefe6 1621Perl_is_uni_xdigit(pTHX_ UV c)
4d61ec05 1622{
4ac6419d 1623 return isXDIGIT_uni(c);
4d61ec05
GS
1624}
1625
3a4c58c9
KW
1626UV
1627Perl__to_upper_title_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp, const char S_or_s)
1628{
1629 /* We have the latin1-range values compiled into the core, so just use
1630 * those, converting the result to utf8. The only difference between upper
1631 * and title case in this range is that LATIN_SMALL_LETTER_SHARP_S is
1632 * either "SS" or "Ss". Which one to use is passed into the routine in
1633 * 'S_or_s' to avoid a test */
1634
1635 UV converted = toUPPER_LATIN1_MOD(c);
1636
1637 PERL_ARGS_ASSERT__TO_UPPER_TITLE_LATIN1;
1638
1639 assert(S_or_s == 'S' || S_or_s == 's');
1640
f4cd282c
KW
1641 if (NATIVE_IS_INVARIANT(converted)) { /* No difference between the two for
1642 characters in this range */
3a4c58c9
KW
1643 *p = (U8) converted;
1644 *lenp = 1;
1645 return converted;
1646 }
1647
1648 /* toUPPER_LATIN1_MOD gives the correct results except for three outliers,
1649 * which it maps to one of them, so as to only have to have one check for
1650 * it in the main case */
1651 if (UNLIKELY(converted == LATIN_SMALL_LETTER_Y_WITH_DIAERESIS)) {
1652 switch (c) {
1653 case LATIN_SMALL_LETTER_Y_WITH_DIAERESIS:
1654 converted = LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS;
1655 break;
1656 case MICRO_SIGN:
1657 converted = GREEK_CAPITAL_LETTER_MU;
1658 break;
1659 case LATIN_SMALL_LETTER_SHARP_S:
1660 *(p)++ = 'S';
1661 *p = S_or_s;
1662 *lenp = 2;
1663 return 'S';
1664 default:
1665 Perl_croak(aTHX_ "panic: to_upper_title_latin1 did not expect '%c' to map to '%c'", c, LATIN_SMALL_LETTER_Y_WITH_DIAERESIS);
118e2215 1666 assert(0); /* NOTREACHED */
3a4c58c9
KW
1667 }
1668 }
1669
1670 *(p)++ = UTF8_TWO_BYTE_HI(converted);
1671 *p = UTF8_TWO_BYTE_LO(converted);
1672 *lenp = 2;
1673
1674 return converted;
1675}
1676
50bda2c3
KW
1677/* Call the function to convert a UTF-8 encoded character to the specified case.
1678 * Note that there may be more than one character in the result.
1679 * INP is a pointer to the first byte of the input character
1680 * OUTP will be set to the first byte of the string of changed characters. It
1681 * needs to have space for UTF8_MAXBYTES_CASE+1 bytes
1682 * LENP will be set to the length in bytes of the string of changed characters
1683 *
1684 * The functions return the ordinal of the first character in the string of OUTP */
f90a9a02
KW
1685#define CALL_UPPER_CASE(INP, OUTP, LENP) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_toupper, "ToUc", "utf8::ToSpecUc")
1686#define CALL_TITLE_CASE(INP, OUTP, LENP) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_totitle, "ToTc", "utf8::ToSpecTc")
1687#define CALL_LOWER_CASE(INP, OUTP, LENP) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_tolower, "ToLc", "utf8::ToSpecLc")
50bda2c3
KW
1688
1689/* This additionally has the input parameter SPECIALS, which if non-zero will
1690 * cause this to use the SPECIALS hash for folding (meaning get full case
1691 * folding); otherwise, when zero, this implies a simple case fold */
f90a9a02 1692#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 1693
84afefe6
JH
1694UV
1695Perl_to_uni_upper(pTHX_ UV c, U8* p, STRLEN *lenp)
a0ed51b3 1696{
3a4c58c9
KW
1697 dVAR;
1698
a1433954
KW
1699 /* Convert the Unicode character whose ordinal is <c> to its uppercase
1700 * version and store that in UTF-8 in <p> and its length in bytes in <lenp>.
1701 * Note that the <p> needs to be at least UTF8_MAXBYTES_CASE+1 bytes since
c3fd2246
KW
1702 * the changed version may be longer than the original character.
1703 *
1704 * The ordinal of the first character of the changed version is returned
1705 * (but note, as explained above, that there may be more.) */
1706
7918f24d
NC
1707 PERL_ARGS_ASSERT_TO_UNI_UPPER;
1708
3a4c58c9
KW
1709 if (c < 256) {
1710 return _to_upper_title_latin1((U8) c, p, lenp, 'S');
1711 }
1712
0ebc6274 1713 uvchr_to_utf8(p, c);
3a4c58c9 1714 return CALL_UPPER_CASE(p, p, lenp);
a0ed51b3
LW
1715}
1716
84afefe6
JH
1717UV
1718Perl_to_uni_title(pTHX_ UV c, U8* p, STRLEN *lenp)
a0ed51b3 1719{
3a4c58c9
KW
1720 dVAR;
1721
7918f24d
NC
1722 PERL_ARGS_ASSERT_TO_UNI_TITLE;
1723
3a4c58c9
KW
1724 if (c < 256) {
1725 return _to_upper_title_latin1((U8) c, p, lenp, 's');
1726 }
1727
0ebc6274 1728 uvchr_to_utf8(p, c);
3a4c58c9 1729 return CALL_TITLE_CASE(p, p, lenp);
a0ed51b3
LW
1730}
1731
afc16117
KW
1732STATIC U8
1733S_to_lower_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp)
1734{
1735 /* We have the latin1-range values compiled into the core, so just use
1736 * those, converting the result to utf8. Since the result is always just
a1433954 1737 * one character, we allow <p> to be NULL */
afc16117
KW
1738
1739 U8 converted = toLOWER_LATIN1(c);
1740
1741 if (p != NULL) {
f4cd282c 1742 if (NATIVE_IS_INVARIANT(converted)) {
afc16117
KW
1743 *p = converted;
1744 *lenp = 1;
1745 }
1746 else {
1747 *p = UTF8_TWO_BYTE_HI(converted);
1748 *(p+1) = UTF8_TWO_BYTE_LO(converted);
1749 *lenp = 2;
1750 }
1751 }
1752 return converted;
1753}
1754
84afefe6
JH
1755UV
1756Perl_to_uni_lower(pTHX_ UV c, U8* p, STRLEN *lenp)
a0ed51b3 1757{
968c5e6a
KW
1758 dVAR;
1759
7918f24d
NC
1760 PERL_ARGS_ASSERT_TO_UNI_LOWER;
1761
afc16117
KW
1762 if (c < 256) {
1763 return to_lower_latin1((U8) c, p, lenp);
bca00c02
KW
1764 }
1765
afc16117 1766 uvchr_to_utf8(p, c);
968c5e6a 1767 return CALL_LOWER_CASE(p, p, lenp);
a0ed51b3
LW
1768}
1769
84afefe6 1770UV
51910141 1771Perl__to_fold_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp, const unsigned int flags)
a1dde8de 1772{
51910141 1773 /* Corresponds to to_lower_latin1(); <flags> bits meanings:
1ca267a5 1774 * FOLD_FLAGS_NOMIX_ASCII iff non-ASCII to ASCII folds are prohibited
51910141 1775 * FOLD_FLAGS_FULL iff full folding is to be used;
1ca267a5
KW
1776 *
1777 * Not to be used for locale folds
51910141 1778 */
f673fad4 1779
a1dde8de
KW
1780 UV converted;
1781
1782 PERL_ARGS_ASSERT__TO_FOLD_LATIN1;
1783
1ca267a5
KW
1784 assert (! (flags & FOLD_FLAGS_LOCALE));
1785
a1dde8de
KW
1786 if (c == MICRO_SIGN) {
1787 converted = GREEK_SMALL_LETTER_MU;
1788 }
51910141 1789 else if ((flags & FOLD_FLAGS_FULL) && c == LATIN_SMALL_LETTER_SHARP_S) {
1ca267a5
KW
1790
1791 /* If can't cross 127/128 boundary, can't return "ss"; instead return
1792 * two U+017F characters, as fc("\df") should eq fc("\x{17f}\x{17f}")
1793 * under those circumstances. */
1794 if (flags & FOLD_FLAGS_NOMIX_ASCII) {
1795 *lenp = 2 * sizeof(LATIN_SMALL_LETTER_LONG_S_UTF8) - 2;
1796 Copy(LATIN_SMALL_LETTER_LONG_S_UTF8 LATIN_SMALL_LETTER_LONG_S_UTF8,
1797 p, *lenp, U8);
1798 return LATIN_SMALL_LETTER_LONG_S;
1799 }
1800 else {
4f489194
KW
1801 *(p)++ = 's';
1802 *p = 's';
1803 *lenp = 2;
1804 return 's';
1ca267a5 1805 }
a1dde8de
KW
1806 }
1807 else { /* In this range the fold of all other characters is their lower
1808 case */
1809 converted = toLOWER_LATIN1(c);
1810 }
1811
f4cd282c 1812 if (NATIVE_IS_INVARIANT(converted)) {
a1dde8de
KW
1813 *p = (U8) converted;
1814 *lenp = 1;
1815 }
1816 else {
1817 *(p)++ = UTF8_TWO_BYTE_HI(converted);
1818 *p = UTF8_TWO_BYTE_LO(converted);
1819 *lenp = 2;
1820 }
1821
1822 return converted;
1823}
1824
1825UV
a0270393 1826Perl__to_uni_fold_flags(pTHX_ UV c, U8* p, STRLEN *lenp, const U8 flags)
84afefe6 1827{
4b593389 1828
a0270393
KW
1829 /* Not currently externally documented, and subject to change
1830 * <flags> bits meanings:
1831 * FOLD_FLAGS_FULL iff full folding is to be used;
1832 * FOLD_FLAGS_LOCALE iff in locale
1833 * FOLD_FLAGS_NOMIX_ASCII iff non-ASCII to ASCII folds are prohibited
1834 */
4b593389 1835
36bb2ab6 1836 PERL_ARGS_ASSERT__TO_UNI_FOLD_FLAGS;
7918f24d 1837
a1dde8de 1838 if (c < 256) {
a0270393 1839 UV result = _to_fold_latin1((U8) c, p, lenp,
1ca267a5 1840 flags & (FOLD_FLAGS_FULL | FOLD_FLAGS_NOMIX_ASCII));
a0270393
KW
1841 /* It is illegal for the fold to cross the 255/256 boundary under
1842 * locale; in this case return the original */
1843 return (result > 256 && flags & FOLD_FLAGS_LOCALE)
1844 ? c
1845 : result;
a1dde8de
KW
1846 }
1847
a0270393
KW
1848 /* If no special needs, just use the macro */
1849 if ( ! (flags & (FOLD_FLAGS_LOCALE|FOLD_FLAGS_NOMIX_ASCII))) {
1850 uvchr_to_utf8(p, c);
1851 return CALL_FOLD_CASE(p, p, lenp, flags & FOLD_FLAGS_FULL);
1852 }
1853 else { /* Otherwise, _to_utf8_fold_flags has the intelligence to deal with
1854 the special flags. */
1855 U8 utf8_c[UTF8_MAXBYTES + 1];
1856 uvchr_to_utf8(utf8_c, c);
1857 return _to_utf8_fold_flags(utf8_c, p, lenp, flags, NULL);
1858 }
84afefe6
JH
1859}
1860
a0ed51b3 1861bool
84afefe6 1862Perl_is_uni_alnum_lc(pTHX_ UV c)
a0ed51b3 1863{
edfb3318 1864 if (c < 256) {
f4cd282c 1865 return isALNUM_LC(c);
edfb3318 1866 }
922e8cb4 1867 return _is_uni_FOO(_CC_WORDCHAR, c);
a0ed51b3
LW
1868}
1869
1870bool
5092f92a
KW
1871Perl_is_uni_alnumc_lc(pTHX_ UV c)
1872{
1873 if (c < 256) {
f4cd282c 1874 return isALPHANUMERIC_LC(c);
5092f92a 1875 }
922e8cb4 1876 return _is_uni_FOO(_CC_ALPHANUMERIC, c);
5092f92a
KW
1877}
1878
1879bool
84afefe6 1880Perl_is_uni_idfirst_lc(pTHX_ UV c)
a0ed51b3 1881{
edfb3318 1882 if (c < 256) {
f4cd282c 1883 return isIDFIRST_LC(c);
edfb3318
KW
1884 }
1885 return _is_uni_perl_idstart(c);
a0ed51b3
LW
1886}
1887
1888bool
84afefe6 1889Perl_is_uni_alpha_lc(pTHX_ UV c)
a0ed51b3 1890{
edfb3318 1891 if (c < 256) {
f4cd282c 1892 return isALPHA_LC(c);
edfb3318 1893 }
922e8cb4 1894 return _is_uni_FOO(_CC_ALPHA, c);
a0ed51b3
LW
1895}
1896
1897bool
84afefe6 1898Perl_is_uni_ascii_lc(pTHX_ UV c)
4d61ec05 1899{
edfb3318 1900 if (c < 256) {
f4cd282c 1901 return isASCII_LC(c);
edfb3318
KW
1902 }
1903 return 0;
4d61ec05
GS
1904}
1905
1906bool
bdd8600f
KW
1907Perl_is_uni_blank_lc(pTHX_ UV c)
1908{
edfb3318 1909 if (c < 256) {
f4cd282c 1910 return isBLANK_LC(c);
edfb3318 1911 }
61b19385 1912 return isBLANK_uni(c);
bdd8600f
KW
1913}
1914
1915bool
84afefe6 1916Perl_is_uni_space_lc(pTHX_ UV c)
a0ed51b3 1917{
edfb3318 1918 if (c < 256) {
f4cd282c 1919 return isSPACE_LC(c);
edfb3318 1920 }
61b19385 1921 return isSPACE_uni(c);
a0ed51b3
LW
1922}
1923
1924bool
84afefe6 1925Perl_is_uni_digit_lc(pTHX_ UV c)
a0ed51b3 1926{
edfb3318 1927 if (c < 256) {
f4cd282c 1928 return isDIGIT_LC(c);
edfb3318 1929 }
922e8cb4 1930 return _is_uni_FOO(_CC_DIGIT, c);
a0ed51b3
LW
1931}
1932
1933bool
84afefe6 1934Perl_is_uni_upper_lc(pTHX_ UV c)
a0ed51b3 1935{
edfb3318 1936 if (c < 256) {
f4cd282c 1937 return isUPPER_LC(c);
edfb3318 1938 }
922e8cb4 1939 return _is_uni_FOO(_CC_UPPER, c);
a0ed51b3
LW
1940}
1941
1942bool
84afefe6 1943Perl_is_uni_lower_lc(pTHX_ UV c)
a0ed51b3 1944{
edfb3318 1945 if (c < 256) {
f4cd282c 1946 return isLOWER_LC(c);
edfb3318 1947 }
922e8cb4 1948 return _is_uni_FOO(_CC_LOWER, c);
a0ed51b3
LW
1949}
1950
1951bool
84afefe6 1952Perl_is_uni_cntrl_lc(pTHX_ UV c)
b8c5462f 1953{
edfb3318 1954 if (c < 256) {
f4cd282c 1955 return isCNTRL_LC(c);
edfb3318 1956 }
61b19385 1957 return 0;
b8c5462f
JH
1958}
1959
1960bool
84afefe6 1961Perl_is_uni_graph_lc(pTHX_ UV c)
b8c5462f 1962{
edfb3318 1963 if (c < 256) {
f4cd282c 1964 return isGRAPH_LC(c);
edfb3318 1965 }
922e8cb4 1966 return _is_uni_FOO(_CC_GRAPH, c);
b8c5462f
JH
1967}
1968
1969bool
84afefe6 1970Perl_is_uni_print_lc(pTHX_ UV c)
a0ed51b3 1971{
edfb3318 1972 if (c < 256) {
f4cd282c 1973 return isPRINT_LC(c);
edfb3318 1974 }
922e8cb4 1975 return _is_uni_FOO(_CC_PRINT, c);
a0ed51b3
LW
1976}
1977
b8c5462f 1978bool
84afefe6 1979Perl_is_uni_punct_lc(pTHX_ UV c)
b8c5462f 1980{
edfb3318 1981 if (c < 256) {
f4cd282c 1982 return isPUNCT_LC(c);
edfb3318 1983 }
922e8cb4 1984 return _is_uni_FOO(_CC_PUNCT, c);
b8c5462f
JH
1985}
1986
4d61ec05 1987bool
84afefe6 1988Perl_is_uni_xdigit_lc(pTHX_ UV c)
4d61ec05 1989{
edfb3318 1990 if (c < 256) {
f4cd282c 1991 return isXDIGIT_LC(c);
edfb3318 1992 }
61b19385 1993 return isXDIGIT_uni(c);
4d61ec05
GS
1994}
1995
b7ac61fa
JH
1996U32
1997Perl_to_uni_upper_lc(pTHX_ U32 c)
1998{
ee099d14
JH
1999 /* XXX returns only the first character -- do not use XXX */
2000 /* XXX no locale support yet */
2001 STRLEN len;
89ebb4a3 2002 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
ee099d14 2003 return (U32)to_uni_upper(c, tmpbuf, &len);
b7ac61fa
JH
2004}
2005
2006U32
2007Perl_to_uni_title_lc(pTHX_ U32 c)
2008{
ee099d14
JH
2009 /* XXX returns only the first character XXX -- do not use XXX */
2010 /* XXX no locale support yet */
2011 STRLEN len;
89ebb4a3 2012 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
ee099d14 2013 return (U32)to_uni_title(c, tmpbuf, &len);
b7ac61fa
JH
2014}
2015
2016U32
2017Perl_to_uni_lower_lc(pTHX_ U32 c)
2018{
ee099d14
JH
2019 /* XXX returns only the first character -- do not use XXX */
2020 /* XXX no locale support yet */
2021 STRLEN len;
89ebb4a3 2022 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
ee099d14 2023 return (U32)to_uni_lower(c, tmpbuf, &len);
b7ac61fa
JH
2024}
2025
26483009 2026PERL_STATIC_INLINE bool
5141f98e 2027S_is_utf8_common(pTHX_ const U8 *const p, SV **swash,
bde6a22d
NC
2028 const char *const swashname)
2029{
ea317ccb
KW
2030 /* returns a boolean giving whether or not the UTF8-encoded character that
2031 * starts at <p> is in the swash indicated by <swashname>. <swash>
2032 * contains a pointer to where the swash indicated by <swashname>
2033 * is to be stored; which this routine will do, so that future calls will
2034 * look at <*swash> and only generate a swash if it is not null
2035 *
2036 * Note that it is assumed that the buffer length of <p> is enough to
2037 * contain all the bytes that comprise the character. Thus, <*p> should
2038 * have been checked before this call for mal-formedness enough to assure
2039 * that. */
2040
97aff369 2041 dVAR;
7918f24d
NC
2042
2043 PERL_ARGS_ASSERT_IS_UTF8_COMMON;
2044
492a624f 2045 /* The API should have included a length for the UTF-8 character in <p>,
28123549 2046 * but it doesn't. We therefore assume that p has been validated at least
492a624f
KW
2047 * as far as there being enough bytes available in it to accommodate the
2048 * character without reading beyond the end, and pass that number on to the
2049 * validating routine */
28123549
KW
2050 if (! is_utf8_char_buf(p, p + UTF8SKIP(p))) {
2051 if (ckWARN_d(WARN_UTF8)) {
2052 Perl_warner(aTHX_ packWARN2(WARN_DEPRECATED,WARN_UTF8),
9816f121 2053 "Passing malformed UTF-8 to \"%s\" is deprecated", swashname);
28123549
KW
2054 if (ckWARN(WARN_UTF8)) { /* This will output details as to the
2055 what the malformation is */
2056 utf8_to_uvchr_buf(p, p + UTF8SKIP(p), NULL);
2057 }
2058 }
2059 return FALSE;
2060 }
87367d5f
KW
2061 if (!*swash) {
2062 U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST;
2063 *swash = _core_swash_init("utf8", swashname, &PL_sv_undef, 1, 0, NULL, &flags);
2064 }
28123549 2065
bde6a22d
NC
2066 return swash_fetch(*swash, p, TRUE) != 0;
2067}
2068
2069bool
922e8cb4
KW
2070Perl__is_utf8_FOO(pTHX_ const U8 classnum, const U8 *p)
2071{
2072 dVAR;
2073
2074 PERL_ARGS_ASSERT__IS_UTF8_FOO;
2075
2076 assert(classnum < _FIRST_NON_SWASH_CC);
2077
2078 return is_utf8_common(p, &PL_utf8_swash_ptrs[classnum], swash_property_names[classnum]);
2079}
2080
2081bool
7fc63493 2082Perl_is_utf8_alnum(pTHX_ const U8 *p)
a0ed51b3 2083{
97aff369 2084 dVAR;
7918f24d
NC
2085
2086 PERL_ARGS_ASSERT_IS_UTF8_ALNUM;
2087
671c33bf
NC
2088 /* NOTE: "IsWord", not "IsAlnum", since Alnum is a true
2089 * descendant of isalnum(3), in other words, it doesn't
2090 * contain the '_'. --jhi */
03940dc2 2091 return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_WORDCHAR], "IsWord");
a0ed51b3
LW
2092}
2093
2094bool
5092f92a
KW
2095Perl_is_utf8_alnumc(pTHX_ const U8 *p)
2096{
2097 dVAR;
2098
2099 PERL_ARGS_ASSERT_IS_UTF8_ALNUMC;
2100
03940dc2 2101 return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_ALPHANUMERIC], "IsAlnum");
5092f92a
KW
2102}
2103
2104bool
7fc63493 2105Perl_is_utf8_idfirst(pTHX_ const U8 *p) /* The naming is historical. */
a0ed51b3 2106{
97aff369 2107 dVAR;
7918f24d
NC
2108
2109 PERL_ARGS_ASSERT_IS_UTF8_IDFIRST;
2110
61b19385 2111 return S_is_utf8_idfirst(aTHX_ p);
82686b01
JH
2112}
2113
2114bool
c11ff943
KW
2115Perl_is_utf8_xidfirst(pTHX_ const U8 *p) /* The naming is historical. */
2116{
2117 dVAR;
2118
2119 PERL_ARGS_ASSERT_IS_UTF8_XIDFIRST;
2120
2121 if (*p == '_')
2122 return TRUE;
2123 /* is_utf8_idstart would be more logical. */
2124 return is_utf8_common(p, &PL_utf8_xidstart, "XIdStart");
2125}
2126
2127bool
d65654cb 2128Perl__is_utf8_perl_idstart(pTHX_ const U8 *p)
b6912c02
KW
2129{
2130 dVAR;
2131
d65654cb 2132 PERL_ARGS_ASSERT__IS_UTF8_PERL_IDSTART;
b6912c02
KW
2133
2134 return is_utf8_common(p, &PL_utf8_perl_idstart, "_Perl_IDStart");
2135}
2136
2137bool
eba68aa0
KW
2138Perl__is_utf8_perl_idcont(pTHX_ const U8 *p)
2139{
2140 dVAR;
2141
2142 PERL_ARGS_ASSERT__IS_UTF8_PERL_IDCONT;
2143
2144 return is_utf8_common(p, &PL_utf8_perl_idcont, "_Perl_IDCont");
2145}
2146
2147
2148bool
7fc63493 2149Perl_is_utf8_idcont(pTHX_ const U8 *p)
82686b01 2150{
97aff369 2151 dVAR;
7918f24d
NC
2152
2153 PERL_ARGS_ASSERT_IS_UTF8_IDCONT;
2154
d4c19fe8 2155 return is_utf8_common(p, &PL_utf8_idcont, "IdContinue");
a0ed51b3
LW
2156}
2157
2158bool
c11ff943
KW
2159Perl_is_utf8_xidcont(pTHX_ const U8 *p)
2160{
2161 dVAR;
2162
2163 PERL_ARGS_ASSERT_IS_UTF8_XIDCONT;
2164
c11ff943
KW
2165 return is_utf8_common(p, &PL_utf8_idcont, "XIdContinue");
2166}
2167
2168bool
7fc63493 2169Perl_is_utf8_alpha(pTHX_ const U8 *p)
a0ed51b3 2170{
97aff369 2171 dVAR;
7918f24d
NC
2172
2173 PERL_ARGS_ASSERT_IS_UTF8_ALPHA;
2174
03940dc2 2175 return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_ALPHA], "IsAlpha");
a0ed51b3
LW
2176}
2177
2178bool
7fc63493 2179Perl_is_utf8_ascii(pTHX_ const U8 *p)
b8c5462f 2180{
97aff369 2181 dVAR;
7918f24d
NC
2182
2183 PERL_ARGS_ASSERT_IS_UTF8_ASCII;
2184
bc39fe24
KW
2185 /* ASCII characters are the same whether in utf8 or not. So the macro
2186 * works on both utf8 and non-utf8 representations. */
2187 return isASCII(*p);
b8c5462f
JH
2188}
2189
2190bool
bdd8600f
KW
2191Perl_is_utf8_blank(pTHX_ const U8 *p)
2192{
2193 dVAR;
2194
2195 PERL_ARGS_ASSERT_IS_UTF8_BLANK;
2196
2cafb56b 2197 return isBLANK_utf8(p);
bdd8600f
KW
2198}
2199
2200bool
7fc63493 2201Perl_is_utf8_space(pTHX_ const U8 *p)
a0ed51b3 2202{
97aff369 2203 dVAR;
7918f24d
NC
2204
2205 PERL_ARGS_ASSERT_IS_UTF8_SPACE;
2206
add4123a 2207 return isSPACE_utf8(p);
a0ed51b3
LW
2208}
2209
2210bool
d1eb3177
YO
2211Perl_is_utf8_perl_space(pTHX_ const U8 *p)
2212{
2213 dVAR;
2214
2215 PERL_ARGS_ASSERT_IS_UTF8_PERL_SPACE;
2216
c4428693
KW
2217 /* Only true if is an ASCII space-like character, and ASCII is invariant
2218 * under utf8, so can just use the macro */
2219 return isSPACE_A(*p);
d1eb3177
YO
2220}
2221
2222bool
2223Perl_is_utf8_perl_word(pTHX_ const U8 *p)
2224{
2225 dVAR;
2226
2227 PERL_ARGS_ASSERT_IS_UTF8_PERL_WORD;
2228
c4428693
KW
2229 /* Only true if is an ASCII word character, and ASCII is invariant
2230 * under utf8, so can just use the macro */
2231 return isWORDCHAR_A(*p);
d1eb3177
YO
2232}
2233
2234bool
7fc63493 2235Perl_is_utf8_digit(pTHX_ const U8 *p)
a0ed51b3 2236{
97aff369 2237 dVAR;
7918f24d
NC
2238
2239 PERL_ARGS_ASSERT_IS_UTF8_DIGIT;
2240
03940dc2 2241 return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_DIGIT], "IsDigit");
a0ed51b3
LW
2242}
2243
2244bool
d1eb3177
YO
2245Perl_is_utf8_posix_digit(pTHX_ const U8 *p)
2246{
2247 dVAR;
2248
2249 PERL_ARGS_ASSERT_IS_UTF8_POSIX_DIGIT;
2250
c4428693
KW
2251 /* Only true if is an ASCII digit character, and ASCII is invariant
2252 * under utf8, so can just use the macro */
2253 return isDIGIT_A(*p);
d1eb3177
YO
2254}
2255
2256bool
7fc63493 2257Perl_is_utf8_upper(pTHX_ const U8 *p)
a0ed51b3 2258{
97aff369 2259 dVAR;
7918f24d
NC
2260
2261 PERL_ARGS_ASSERT_IS_UTF8_UPPER;
2262
03940dc2 2263 return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_UPPER], "IsUppercase");
a0ed51b3
LW
2264}
2265
2266bool
7fc63493 2267Perl_is_utf8_lower(pTHX_ const U8 *p)
a0ed51b3 2268{
97aff369 2269 dVAR;
7918f24d
NC
2270
2271 PERL_ARGS_ASSERT_IS_UTF8_LOWER;
2272
03940dc2 2273 return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_LOWER], "IsLowercase");
a0ed51b3
LW
2274}
2275
2276bool
7fc63493 2277Perl_is_utf8_cntrl(pTHX_ const U8 *p)
b8c5462f 2278{
97aff369 2279 dVAR;
7918f24d
NC
2280
2281 PERL_ARGS_ASSERT_IS_UTF8_CNTRL;
2282
a35d759a 2283 return isCNTRL_utf8(p);
b8c5462f
JH
2284}
2285
2286bool
7fc63493 2287Perl_is_utf8_graph(pTHX_ const U8 *p)
b8c5462f 2288{
97aff369 2289 dVAR;
7918f24d
NC
2290
2291 PERL_ARGS_ASSERT_IS_UTF8_GRAPH;
2292
03940dc2 2293 return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_GRAPH], "IsGraph");
b8c5462f
JH
2294}
2295
2296bool
7fc63493 2297Perl_is_utf8_print(pTHX_ const U8 *p)
a0ed51b3 2298{
97aff369 2299 dVAR;
7918f24d
NC
2300
2301 PERL_ARGS_ASSERT_IS_UTF8_PRINT;
2302
03940dc2 2303 return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_PRINT], "IsPrint");
a0ed51b3
LW
2304}
2305
2306bool
7fc63493 2307Perl_is_utf8_punct(pTHX_ const U8 *p)
b8c5462f 2308{
97aff369 2309 dVAR;
7918f24d
NC
2310
2311 PERL_ARGS_ASSERT_IS_UTF8_PUNCT;
2312
03940dc2 2313 return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_PUNCT], "IsPunct");
b8c5462f
JH
2314}
2315
2316bool
7fc63493 2317Perl_is_utf8_xdigit(pTHX_ const U8 *p)
b8c5462f 2318{
97aff369 2319 dVAR;
7918f24d
NC
2320
2321 PERL_ARGS_ASSERT_IS_UTF8_XDIGIT;
2322
4ac6419d 2323 return is_XDIGIT_utf8(p);
b8c5462f
JH
2324}
2325
2326bool
7dbf68d2
KW
2327Perl__is_utf8_mark(pTHX_ const U8 *p)
2328{
2329 dVAR;
2330
2331 PERL_ARGS_ASSERT__IS_UTF8_MARK;
2332
2333 return is_utf8_common(p, &PL_utf8_mark, "IsM");
2334}
2335
2336
2337bool
7fc63493 2338Perl_is_utf8_mark(pTHX_ const U8 *p)
a0ed51b3 2339{
97aff369 2340 dVAR;
7918f24d
NC
2341
2342 PERL_ARGS_ASSERT_IS_UTF8_MARK;
2343
7dbf68d2 2344 return _is_utf8_mark(p);
a0ed51b3
LW
2345}
2346
6b5c0936 2347/*
87cea99e 2348=for apidoc to_utf8_case
6b5c0936 2349
a1433954
KW
2350The C<p> contains the pointer to the UTF-8 string encoding
2351the character that is being converted. This routine assumes that the character
2352at C<p> is well-formed.
6b5c0936 2353
a1433954
KW
2354The C<ustrp> is a pointer to the character buffer to put the
2355conversion result to. The C<lenp> is a pointer to the length
6b5c0936
JH
2356of the result.
2357
a1433954 2358The C<swashp> is a pointer to the swash to use.
6b5c0936 2359
a1433954
KW
2360Both the special and normal mappings are stored in F<lib/unicore/To/Foo.pl>,
2361and loaded by SWASHNEW, using F<lib/utf8_heavy.pl>. The C<special> (usually,
0134edef 2362but not always, a multicharacter mapping), is tried first.
6b5c0936 2363
a1433954 2364The C<special> is a string like "utf8::ToSpecLower", which means the
0134edef
JH
2365hash %utf8::ToSpecLower. The access to the hash is through
2366Perl_to_utf8_case().
6b5c0936 2367
a1433954 2368The C<normal> is a string like "ToLower" which means the swash
0134edef
JH
2369%utf8::ToLower.
2370
2371=cut */
6b5c0936 2372
2104c8d9 2373UV
9a957fbc
AL
2374Perl_to_utf8_case(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp,
2375 SV **swashp, const char *normal, const char *special)
a0ed51b3 2376{
97aff369 2377 dVAR;
0134edef 2378 STRLEN len = 0;
f4cd282c 2379 const UV uv1 = valid_utf8_to_uvchr(p, NULL);
7918f24d
NC
2380
2381 PERL_ARGS_ASSERT_TO_UTF8_CASE;
2382
9ae3ac1a
KW
2383 /* Note that swash_fetch() doesn't output warnings for these because it
2384 * assumes we will */
8457b38f 2385 if (uv1 >= UNICODE_SURROGATE_FIRST) {
9ae3ac1a 2386 if (uv1 <= UNICODE_SURROGATE_LAST) {
8457b38f
KW
2387 if (ckWARN_d(WARN_SURROGATE)) {
2388 const char* desc = (PL_op) ? OP_DESC(PL_op) : normal;
2389 Perl_warner(aTHX_ packWARN(WARN_SURROGATE),
2390 "Operation \"%s\" returns its argument for UTF-16 surrogate U+%04"UVXf"", desc, uv1);
2391 }
9ae3ac1a
KW
2392 }
2393 else if (UNICODE_IS_SUPER(uv1)) {
8457b38f
KW
2394 if (ckWARN_d(WARN_NON_UNICODE)) {
2395 const char* desc = (PL_op) ? OP_DESC(PL_op) : normal;
2396 Perl_warner(aTHX_ packWARN(WARN_NON_UNICODE),
2397 "Operation \"%s\" returns its argument for non-Unicode code point 0x%04"UVXf"", desc, uv1);
2398 }
9ae3ac1a
KW
2399 }
2400
2401 /* Note that non-characters are perfectly legal, so no warning should
2402 * be given */
2403 }
2404
0134edef 2405 if (!*swashp) /* load on-demand */
5ab9d2ef 2406 *swashp = _core_swash_init("utf8", normal, &PL_sv_undef, 4, 0, NULL, NULL);
0134edef 2407
a6f87d8c 2408 if (special) {
0134edef 2409 /* It might be "special" (sometimes, but not always,
2a37f04d 2410 * a multicharacter mapping) */
6673a63c 2411 HV * const hv = get_hv(special, 0);
b08cf34e
JH
2412 SV **svp;
2413
35da51f7 2414 if (hv &&
f4cd282c 2415 (svp = hv_fetch(hv, (const char*)p, UNISKIP(uv1), FALSE)) &&
b08cf34e 2416 (*svp)) {
cfd0369c 2417 const char *s;
47654450 2418
cfd0369c 2419 s = SvPV_const(*svp, len);
47654450 2420 if (len == 1)
f4cd282c 2421 /* EIGHTBIT */
c80e42f3 2422 len = uvchr_to_utf8(ustrp, *(U8*)s) - ustrp;
2a37f04d 2423 else {
d2dcd0fb 2424 Copy(s, ustrp, len, U8);
29e98929 2425 }
983ffd37 2426 }
0134edef
JH
2427 }
2428
2429 if (!len && *swashp) {
f4cd282c 2430 const UV uv2 = swash_fetch(*swashp, p, TRUE /* => is utf8 */);
d4c19fe8 2431
0134edef
JH
2432 if (uv2) {
2433 /* It was "normal" (a single character mapping). */
f4cd282c 2434 len = uvchr_to_utf8(ustrp, uv2) - ustrp;
2a37f04d
JH
2435 }
2436 }
1feea2c7 2437
cbe07460
KW
2438 if (len) {
2439 if (lenp) {
2440 *lenp = len;
2441 }
2442 return valid_utf8_to_uvchr(ustrp, 0);
2443 }
2444
2445 /* Here, there was no mapping defined, which means that the code point maps
2446 * to itself. Return the inputs */
bfdf22ec 2447 len = UTF8SKIP(p);
ca9fab46
KW
2448 if (p != ustrp) { /* Don't copy onto itself */
2449 Copy(p, ustrp, len, U8);
2450 }
0134edef 2451
2a37f04d
JH
2452 if (lenp)
2453 *lenp = len;
2454
f4cd282c 2455 return uv1;
cbe07460 2456
a0ed51b3
LW
2457}
2458
051a06d4
KW
2459STATIC UV
2460S_check_locale_boundary_crossing(pTHX_ const U8* const p, const UV result, U8* const ustrp, STRLEN *lenp)
2461{
2462 /* This is called when changing the case of a utf8-encoded character above
2463 * the Latin1 range, and the operation is in locale. If the result
2464 * contains a character that crosses the 255/256 boundary, disallow the
2465 * change, and return the original code point. See L<perlfunc/lc> for why;
2466 *
a1433954
KW
2467 * p points to the original string whose case was changed; assumed
2468 * by this routine to be well-formed
051a06d4
KW
2469 * result the code point of the first character in the changed-case string
2470 * ustrp points to the changed-case string (<result> represents its first char)
2471 * lenp points to the length of <ustrp> */
2472
2473 UV original; /* To store the first code point of <p> */
2474
2475 PERL_ARGS_ASSERT_CHECK_LOCALE_BOUNDARY_CROSSING;
2476
a4f12ed7 2477 assert(UTF8_IS_ABOVE_LATIN1(*p));
051a06d4
KW
2478
2479 /* We know immediately if the first character in the string crosses the
2480 * boundary, so can skip */
2481 if (result > 255) {
2482
2483 /* Look at every character in the result; if any cross the
2484 * boundary, the whole thing is disallowed */
2485 U8* s = ustrp + UTF8SKIP(ustrp);
2486 U8* e = ustrp + *lenp;
2487 while (s < e) {
a4f12ed7 2488 if (! UTF8_IS_ABOVE_LATIN1(*s)) {
051a06d4
KW
2489 goto bad_crossing;
2490 }
2491 s += UTF8SKIP(s);
2492 }
2493
2494 /* Here, no characters crossed, result is ok as-is */
2495 return result;
2496 }
2497
2498bad_crossing:
2499
2500 /* Failed, have to return the original */
4b88fb76 2501 original = valid_utf8_to_uvchr(p, lenp);
051a06d4
KW
2502 Copy(p, ustrp, *lenp, char);
2503 return original;
2504}
2505
d3e79532 2506/*
87cea99e 2507=for apidoc to_utf8_upper
d3e79532 2508
1f607577 2509Instead use L</toUPPER_utf8>.
a1433954 2510
d3e79532
JH
2511=cut */
2512
051a06d4
KW
2513/* Not currently externally documented, and subject to change:
2514 * <flags> is set iff locale semantics are to be used for code points < 256
2515 * <tainted_ptr> if non-null, *tainted_ptr will be set TRUE iff locale rules
2516 * were used in the calculation; otherwise unchanged. */
2517
2104c8d9 2518UV
051a06d4 2519Perl__to_utf8_upper_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, const bool flags, bool* tainted_ptr)
a0ed51b3 2520{
97aff369 2521 dVAR;
7918f24d 2522
051a06d4
KW
2523 UV result;
2524
2525 PERL_ARGS_ASSERT__TO_UTF8_UPPER_FLAGS;
7918f24d 2526
3a4c58c9 2527 if (UTF8_IS_INVARIANT(*p)) {
051a06d4
KW
2528 if (flags) {
2529 result = toUPPER_LC(*p);
2530 }
2531 else {
81c6c7ce 2532 return _to_upper_title_latin1(*p, ustrp, lenp, 'S');
051a06d4 2533 }
3a4c58c9
KW
2534 }
2535 else if UTF8_IS_DOWNGRADEABLE_START(*p) {
051a06d4 2536 if (flags) {
94bb8c36 2537 result = toUPPER_LC(TWO_BYTE_UTF8_TO_NATIVE(*p, *(p+1)));
051a06d4
KW
2538 }
2539 else {
94bb8c36 2540 return _to_upper_title_latin1(TWO_BYTE_UTF8_TO_NATIVE(*p, *(p+1)),
81c6c7ce 2541 ustrp, lenp, 'S');
051a06d4
KW
2542 }
2543 }
2544 else { /* utf8, ord above 255 */
2545 result = CALL_UPPER_CASE(p, ustrp, lenp);
2546
2547 if (flags) {
2548 result = check_locale_boundary_crossing(p, result, ustrp, lenp);
2549 }
2550 return result;
2551 }
2552
2553 /* Here, used locale rules. Convert back to utf8 */
2554 if (UTF8_IS_INVARIANT(result)) {
2555 *ustrp = (U8) result;
2556 *lenp = 1;
2557 }
2558 else {
2559 *ustrp = UTF8_EIGHT_BIT_HI(result);
2560 *(ustrp + 1) = UTF8_EIGHT_BIT_LO(result);
2561 *lenp = 2;
3a4c58c9
KW
2562 }
2563
051a06d4
KW
2564 if (tainted_ptr) {
2565 *tainted_ptr = TRUE;
2566 }
2567 return result;
983ffd37 2568}
a0ed51b3 2569
d3e79532 2570/*
87cea99e 2571=for apidoc to_utf8_title
d3e79532 2572
1f607577 2573Instead use L</toTITLE_utf8>.
a1433954 2574
d3e79532
JH
2575=cut */
2576
051a06d4
KW
2577/* Not currently externally documented, and subject to change:
2578 * <flags> is set iff locale semantics are to be used for code points < 256
2579 * Since titlecase is not defined in POSIX, uppercase is used instead
2580 * for these/
2581 * <tainted_ptr> if non-null, *tainted_ptr will be set TRUE iff locale rules
2582 * were used in the calculation; otherwise unchanged. */
2583
983ffd37 2584UV
051a06d4 2585Perl__to_utf8_title_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, const bool flags, bool* tainted_ptr)
983ffd37 2586{
97aff369 2587 dVAR;
7918f24d 2588
051a06d4
KW
2589 UV result;
2590
2591 PERL_ARGS_ASSERT__TO_UTF8_TITLE_FLAGS;
7918f24d 2592
3a4c58c9 2593 if (UTF8_IS_INVARIANT(*p)) {
051a06d4
KW
2594 if (flags) {
2595 result = toUPPER_LC(*p);
2596 }
2597 else {
81c6c7ce 2598 return _to_upper_title_latin1(*p, ustrp, lenp, 's');
051a06d4 2599 }
3a4c58c9
KW
2600 }
2601 else if UTF8_IS_DOWNGRADEABLE_START(*p) {
051a06d4 2602 if (flags) {
94bb8c36 2603 result = toUPPER_LC(TWO_BYTE_UTF8_TO_NATIVE(*p, *(p+1)));
051a06d4
KW
2604 }
2605 else {
94bb8c36 2606 return _to_upper_title_latin1(TWO_BYTE_UTF8_TO_NATIVE(*p, *(p+1)),
81c6c7ce 2607 ustrp, lenp, 's');
051a06d4
KW
2608 }
2609 }
2610 else { /* utf8, ord above 255 */
2611 result = CALL_TITLE_CASE(p, ustrp, lenp);
2612
2613 if (flags) {
2614 result = check_locale_boundary_crossing(p, result, ustrp, lenp);
2615 }
2616 return result;
2617 }
2618
2619 /* Here, used locale rules. Convert back to utf8 */
2620 if (UTF8_IS_INVARIANT(result)) {
2621 *ustrp = (U8) result;
2622 *lenp = 1;
2623 }
2624 else {
2625 *ustrp = UTF8_EIGHT_BIT_HI(result);
2626 *(ustrp + 1) = UTF8_EIGHT_BIT_LO(result);
2627 *lenp = 2;
3a4c58c9
KW
2628 }
2629
051a06d4
KW
2630 if (tainted_ptr) {
2631 *tainted_ptr = TRUE;
2632 }
2633 return result;
a0ed51b3
LW
2634}
2635
d3e79532 2636/*
87cea99e 2637=for apidoc to_utf8_lower
d3e79532 2638
1f607577 2639Instead use L</toLOWER_utf8>.
a1433954 2640
d3e79532
JH
2641=cut */
2642
051a06d4
KW
2643/* Not currently externally documented, and subject to change:
2644 * <flags> is set iff locale semantics are to be used for code points < 256
2645 * <tainted_ptr> if non-null, *tainted_ptr will be set TRUE iff locale rules
2646 * were used in the calculation; otherwise unchanged. */
2647
2104c8d9 2648UV
051a06d4 2649Perl__to_utf8_lower_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, const bool flags, bool* tainted_ptr)
a0ed51b3 2650{
051a06d4
KW
2651 UV result;
2652
97aff369 2653 dVAR;
7918f24d 2654
051a06d4 2655 PERL_ARGS_ASSERT__TO_UTF8_LOWER_FLAGS;
7918f24d 2656
968c5e6a 2657 if (UTF8_IS_INVARIANT(*p)) {
051a06d4
KW
2658 if (flags) {
2659 result = toLOWER_LC(*p);
2660 }
2661 else {
81c6c7ce 2662 return to_lower_latin1(*p, ustrp, lenp);
051a06d4 2663 }
968c5e6a
KW
2664 }
2665 else if UTF8_IS_DOWNGRADEABLE_START(*p) {
051a06d4 2666 if (flags) {
94bb8c36 2667 result = toLOWER_LC(TWO_BYTE_UTF8_TO_NATIVE(*p, *(p+1)));
051a06d4
KW
2668 }
2669 else {
94bb8c36 2670 return to_lower_latin1(TWO_BYTE_UTF8_TO_NATIVE(*p, *(p+1)),
81c6c7ce 2671 ustrp, lenp);
051a06d4 2672 }
968c5e6a 2673 }
051a06d4
KW
2674 else { /* utf8, ord above 255 */
2675 result = CALL_LOWER_CASE(p, ustrp, lenp);
2676
2677 if (flags) {
2678 result = check_locale_boundary_crossing(p, result, ustrp, lenp);
2679 }
968c5e6a 2680
051a06d4
KW
2681 return result;
2682 }
2683
2684 /* Here, used locale rules. Convert back to utf8 */
2685 if (UTF8_IS_INVARIANT(result)) {
2686 *ustrp = (U8) result;
2687 *lenp = 1;
2688 }
2689 else {
2690 *ustrp = UTF8_EIGHT_BIT_HI(result);
2691 *(ustrp + 1) = UTF8_EIGHT_BIT_LO(result);
2692 *lenp = 2;
2693 }
2694
2695 if (tainted_ptr) {
2696 *tainted_ptr = TRUE;
2697 }
2698 return result;
b4e400f9
JH
2699}
2700
d3e79532 2701/*
87cea99e 2702=for apidoc to_utf8_fold
d3e79532 2703
1f607577 2704Instead use L</toFOLD_utf8>.
a1433954 2705
d3e79532
JH
2706=cut */
2707
051a06d4
KW
2708/* Not currently externally documented, and subject to change,
2709 * in <flags>
2710 * bit FOLD_FLAGS_LOCALE is set iff locale semantics are to be used for code
2711 * points < 256. Since foldcase is not defined in
2712 * POSIX, lowercase is used instead
2713 * bit FOLD_FLAGS_FULL is set iff full case folds are to be used;
2714 * otherwise simple folds
a0270393
KW
2715 * bit FOLD_FLAGS_NOMIX_ASCII is set iff folds of non-ASCII to ASCII are
2716 * prohibited
051a06d4
KW
2717 * <tainted_ptr> if non-null, *tainted_ptr will be set TRUE iff locale rules
2718 * were used in the calculation; otherwise unchanged. */
36bb2ab6 2719
b4e400f9 2720UV
051a06d4 2721Perl__to_utf8_fold_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, U8 flags, bool* tainted_ptr)
b4e400f9 2722{
97aff369 2723 dVAR;
7918f24d 2724
051a06d4
KW
2725 UV result;
2726
36bb2ab6 2727 PERL_ARGS_ASSERT__TO_UTF8_FOLD_FLAGS;
7918f24d 2728
a0270393
KW
2729 /* These are mutually exclusive */
2730 assert (! ((flags & FOLD_FLAGS_LOCALE) && (flags & FOLD_FLAGS_NOMIX_ASCII)));
2731
50ba90ff
KW
2732 assert(p != ustrp); /* Otherwise overwrites */
2733
a1dde8de 2734 if (UTF8_IS_INVARIANT(*p)) {
051a06d4 2735 if (flags & FOLD_FLAGS_LOCALE) {
d22b930b 2736 result = toFOLD_LC(*p);
051a06d4
KW
2737 }
2738 else {
81c6c7ce 2739 return _to_fold_latin1(*p, ustrp, lenp,
1ca267a5 2740 flags & (FOLD_FLAGS_FULL | FOLD_FLAGS_NOMIX_ASCII));
051a06d4 2741 }
a1dde8de
KW
2742 }
2743 else if UTF8_IS_DOWNGRADEABLE_START(*p) {
051a06d4 2744 if (flags & FOLD_FLAGS_LOCALE) {
94bb8c36 2745 result = toFOLD_LC(TWO_BYTE_UTF8_TO_NATIVE(*p, *(p+1)));
051a06d4
KW
2746 }
2747 else {
94bb8c36 2748 return _to_fold_latin1(TWO_BYTE_UTF8_TO_NATIVE(*p, *(p+1)),
51910141 2749 ustrp, lenp,
1ca267a5 2750 flags & (FOLD_FLAGS_FULL | FOLD_FLAGS_NOMIX_ASCII));
051a06d4 2751 }
a1dde8de 2752 }
051a06d4 2753 else { /* utf8, ord above 255 */
a0270393 2754 result = CALL_FOLD_CASE(p, ustrp, lenp, flags & FOLD_FLAGS_FULL);
a1dde8de 2755
1ca267a5
KW
2756 if (flags & FOLD_FLAGS_LOCALE) {
2757
2758 /* Special case this character, as what normally gets returned
2759 * under locale doesn't work */
2760 if (UTF8SKIP(p) == sizeof(LATIN_CAPITAL_LETTER_SHARP_S_UTF8) - 1
2761 && memEQ((char *) p, LATIN_CAPITAL_LETTER_SHARP_S_UTF8,
2762 sizeof(LATIN_CAPITAL_LETTER_SHARP_S_UTF8) - 1))
2763 {
2764 goto return_long_s;
2765 }
a0270393 2766 return check_locale_boundary_crossing(p, result, ustrp, lenp);
051a06d4 2767 }
a0270393
KW
2768 else if (! (flags & FOLD_FLAGS_NOMIX_ASCII)) {
2769 return result;
2770 }
2771 else {
2772 /* This is called when changing the case of a utf8-encoded
2773 * character above the Latin1 range, and the result should not
2774 * contain an ASCII character. */
2775
2776 UV original; /* To store the first code point of <p> */
2777
2778 /* Look at every character in the result; if any cross the
2779 * boundary, the whole thing is disallowed */
2780 U8* s = ustrp;
2781 U8* e = ustrp + *lenp;
2782 while (s < e) {
2783 if (isASCII(*s)) {
2784 /* Crossed, have to return the original */
2785 original = valid_utf8_to_uvchr(p, lenp);
1ca267a5
KW
2786
2787 /* But in this one instance, there is an alternative we can
2788 * return that is valid */
2789 if (original == LATIN_CAPITAL_LETTER_SHARP_S) {
2790 goto return_long_s;
2791 }
a0270393
KW
2792 Copy(p, ustrp, *lenp, char);
2793 return original;
2794 }
2795 s += UTF8SKIP(s);
2796 }
051a06d4 2797
a0270393
KW
2798 /* Here, no characters crossed, result is ok as-is */
2799 return result;
2800 }
051a06d4
KW
2801 }
2802
2803 /* Here, used locale rules. Convert back to utf8 */
2804 if (UTF8_IS_INVARIANT(result)) {
2805 *ustrp = (U8) result;
2806 *lenp = 1;
2807 }
2808 else {
2809 *ustrp = UTF8_EIGHT_BIT_HI(result);
2810 *(ustrp + 1) = UTF8_EIGHT_BIT_LO(result);
2811 *lenp = 2;
2812 }
2813
2814 if (tainted_ptr) {
2815 *tainted_ptr = TRUE;
2816 }
2817 return result;
1ca267a5
KW
2818
2819 return_long_s:
2820 /* Certain folds to 'ss' are prohibited by the options, but they do allow
2821 * folds to a string of two of these characters. By returning this
2822 * instead, then, e.g.,
2823 * fc("\x{1E9E}") eq fc("\x{17F}\x{17F}")
2824 * works. */
2825
2826 *lenp = 2 * sizeof(LATIN_SMALL_LETTER_LONG_S_UTF8) - 2;
2827 Copy(LATIN_SMALL_LETTER_LONG_S_UTF8 LATIN_SMALL_LETTER_LONG_S_UTF8,
2828 ustrp, *lenp, U8);
2829 return LATIN_SMALL_LETTER_LONG_S;
a0ed51b3
LW
2830}
2831
711a919c 2832/* Note:
f90a9a02 2833 * Returns a "swash" which is a hash described in utf8.c:Perl_swash_fetch().
711a919c
ST
2834 * C<pkg> is a pointer to a package name for SWASHNEW, should be "utf8".
2835 * For other parameters, see utf8::SWASHNEW in lib/utf8_heavy.pl.
2836 */
c4a5db0c 2837
a0ed51b3 2838SV*
7fc63493 2839Perl_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv, I32 minbits, I32 none)
a0ed51b3 2840{
c4a5db0c
KW
2841 PERL_ARGS_ASSERT_SWASH_INIT;
2842
2843 /* Returns a copy of a swash initiated by the called function. This is the
2844 * public interface, and returning a copy prevents others from doing
2845 * mischief on the original */
2846
5d3d13d1 2847 return newSVsv(_core_swash_init(pkg, name, listsv, minbits, none, NULL, NULL));
c4a5db0c
KW
2848}
2849
2850SV*
5d3d13d1 2851Perl__core_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv, I32 minbits, I32 none, SV* invlist, U8* const flags_p)
c4a5db0c
KW
2852{
2853 /* Initialize and return a swash, creating it if necessary. It does this
87367d5f
KW
2854 * by calling utf8_heavy.pl in the general case. The returned value may be
2855 * the swash's inversion list instead if the input parameters allow it.
2856 * Which is returned should be immaterial to callers, as the only
923b6d4e
KW
2857 * operations permitted on a swash, swash_fetch(), _get_swash_invlist(),
2858 * and swash_to_invlist() handle both these transparently.
c4a5db0c
KW
2859 *
2860 * This interface should only be used by functions that won't destroy or
2861 * adversely change the swash, as doing so affects all other uses of the
2862 * swash in the program; the general public should use 'Perl_swash_init'
2863 * instead.
2864 *
2865 * pkg is the name of the package that <name> should be in.
2866 * name is the name of the swash to find. Typically it is a Unicode
2867 * property name, including user-defined ones
2868 * listsv is a string to initialize the swash with. It must be of the form
2869 * documented as the subroutine return value in
2870 * L<perlunicode/User-Defined Character Properties>
2871 * minbits is the number of bits required to represent each data element.
2872 * It is '1' for binary properties.
2873 * none I (khw) do not understand this one, but it is used only in tr///.
9a53f6cf 2874 * invlist is an inversion list to initialize the swash with (or NULL)
83199d38
KW
2875 * flags_p if non-NULL is the address of various input and output flag bits
2876 * to the routine, as follows: ('I' means is input to the routine;
2877 * 'O' means output from the routine. Only flags marked O are
2878 * meaningful on return.)
2879 * _CORE_SWASH_INIT_USER_DEFINED_PROPERTY indicates if the swash
2880 * came from a user-defined property. (I O)
5d3d13d1
KW
2881 * _CORE_SWASH_INIT_RETURN_IF_UNDEF indicates that instead of croaking
2882 * when the swash cannot be located, to simply return NULL. (I)
87367d5f
KW
2883 * _CORE_SWASH_INIT_ACCEPT_INVLIST indicates that the caller will accept a
2884 * return of an inversion list instead of a swash hash if this routine
2885 * thinks that would result in faster execution of swash_fetch() later
2886 * on. (I)
9a53f6cf
KW
2887 *
2888 * Thus there are three possible inputs to find the swash: <name>,
2889 * <listsv>, and <invlist>. At least one must be specified. The result
2890 * will be the union of the specified ones, although <listsv>'s various
2891 * actions can intersect, etc. what <name> gives.
2892 *
2893 * <invlist> is only valid for binary properties */
c4a5db0c 2894
27da23d5 2895 dVAR;
c4a5db0c 2896 SV* retval = &PL_sv_undef;
83199d38 2897 HV* swash_hv = NULL;
87367d5f
KW
2898 const int invlist_swash_boundary =
2899 (flags_p && *flags_p & _CORE_SWASH_INIT_ACCEPT_INVLIST)
2900 ? 512 /* Based on some benchmarking, but not extensive, see commit
2901 message */
2902 : -1; /* Never return just an inversion list */
9a53f6cf
KW
2903
2904 assert(listsv != &PL_sv_undef || strNE(name, "") || invlist);
2905 assert(! invlist || minbits == 1);
2906
2907 /* If data was passed in to go out to utf8_heavy to find the swash of, do
2908 * so */
2909 if (listsv != &PL_sv_undef || strNE(name, "")) {
69794297
KW
2910 dSP;
2911 const size_t pkg_len = strlen(pkg);
2912 const size_t name_len = strlen(name);
2913 HV * const stash = gv_stashpvn(pkg, pkg_len, 0);
2914 SV* errsv_save;
2915 GV *method;
2916
2917 PERL_ARGS_ASSERT__CORE_SWASH_INIT;
2918
2919 PUSHSTACKi(PERLSI_MAGIC);
ce3b816e 2920 ENTER;
69794297
KW
2921 SAVEHINTS();
2922 save_re_context();
650f067c
JL
2923 /* We might get here via a subroutine signature which uses a utf8
2924 * parameter name, at which point PL_subname will have been set
2925 * but not yet used. */
2926 save_item(PL_subname);
69794297
KW
2927 if (PL_parser && PL_parser->error_count)
2928 SAVEI8(PL_parser->error_count), PL_parser->error_count = 0;
2929 method = gv_fetchmeth(stash, "SWASHNEW", 8, -1);
2930 if (!method) { /* demand load utf8 */
2931 ENTER;
db2c6cb3
FC
2932 if ((errsv_save = GvSV(PL_errgv))) SAVEFREESV(errsv_save);
2933 GvSV(PL_errgv) = NULL;
69794297
KW
2934 /* It is assumed that callers of this routine are not passing in
2935 * any user derived data. */
2936 /* Need to do this after save_re_context() as it will set
2937 * PL_tainted to 1 while saving $1 etc (see the code after getrx:
2938 * in Perl_magic_get). Even line to create errsv_save can turn on
2939 * PL_tainted. */
284167a5
SM
2940#ifndef NO_TAINT_SUPPORT
2941 SAVEBOOL(TAINT_get);
2942 TAINT_NOT;
2943#endif
69794297
KW
2944 Perl_load_module(aTHX_ PERL_LOADMOD_NOIMPORT, newSVpvn(pkg,pkg_len),
2945 NULL);
eed484f9 2946 {
db2c6cb3
FC
2947 /* Not ERRSV, as there is no need to vivify a scalar we are
2948 about to discard. */
2949 SV * const errsv = GvSV(PL_errgv);
2950 if (!SvTRUE(errsv)) {
2951 GvSV(PL_errgv) = SvREFCNT_inc_simple(errsv_save);
2952 SvREFCNT_dec(errsv);
2953 }
eed484f9 2954 }
69794297
KW
2955 LEAVE;
2956 }
2957 SPAGAIN;
2958 PUSHMARK(SP);
2959 EXTEND(SP,5);
2960 mPUSHp(pkg, pkg_len);
2961 mPUSHp(name, name_len);
2962 PUSHs(listsv);
2963 mPUSHi(minbits);
2964 mPUSHi(none);
2965 PUTBACK;
db2c6cb3
FC
2966 if ((errsv_save = GvSV(PL_errgv))) SAVEFREESV(errsv_save);
2967 GvSV(PL_errgv) = NULL;
69794297
KW
2968 /* If we already have a pointer to the method, no need to use
2969 * call_method() to repeat the lookup. */
c41800a8
KW
2970 if (method
2971 ? call_sv(MUTABLE_SV(method), G_SCALAR)
69794297
KW
2972 : call_sv(newSVpvs_flags("SWASHNEW", SVs_TEMP), G_SCALAR | G_METHOD))
2973 {
2974 retval = *PL_stack_sp--;
2975 SvREFCNT_inc(retval);
2976 }
eed484f9 2977 {
db2c6cb3
FC
2978 /* Not ERRSV. See above. */
2979 SV * const errsv = GvSV(PL_errgv);
2980 if (!SvTRUE(errsv)) {
2981 GvSV(PL_errgv) = SvREFCNT_inc_simple(errsv_save);
2982 SvREFCNT_dec(errsv);
2983 }
eed484f9 2984 }
ce3b816e 2985 LEAVE;
69794297
KW
2986 POPSTACK;
2987 if (IN_PERL_COMPILETIME) {
2988 CopHINTS_set(PL_curcop, PL_hints);
2989 }
2990 if (!SvROK(retval) || SvTYPE(SvRV(retval)) != SVt_PVHV) {
2991 if (SvPOK(retval))
2992
2993 /* If caller wants to handle missing properties, let them */
5d3d13d1 2994 if (flags_p && *flags_p & _CORE_SWASH_INIT_RETURN_IF_UNDEF) {
69794297
KW
2995 return NULL;
2996 }
2997 Perl_croak(aTHX_
2998 "Can't find Unicode property definition \"%"SVf"\"",
2999 SVfARG(retval));
3000 Perl_croak(aTHX_ "SWASHNEW didn't return an HV ref");
3001 }
9a53f6cf 3002 } /* End of calling the module to find the swash */
36eb48b4 3003
83199d38
KW
3004 /* If this operation fetched a swash, and we will need it later, get it */
3005 if (retval != &PL_sv_undef
3006 && (minbits == 1 || (flags_p
3007 && ! (*flags_p
3008 & _CORE_SWASH_INIT_USER_DEFINED_PROPERTY))))
3009 {
3010 swash_hv = MUTABLE_HV(SvRV(retval));
3011
3012 /* If we don't already know that there is a user-defined component to
3013 * this swash, and the user has indicated they wish to know if there is
3014 * one (by passing <flags_p>), find out */
3015 if (flags_p && ! (*flags_p & _CORE_SWASH_INIT_USER_DEFINED_PROPERTY)) {
3016 SV** user_defined = hv_fetchs(swash_hv, "USER_DEFINED", FALSE);
3017 if (user_defined && SvUV(*user_defined)) {
3018 *flags_p |= _CORE_SWASH_INIT_USER_DEFINED_PROPERTY;
3019 }
3020 }
3021 }
3022
36eb48b4
KW
3023 /* Make sure there is an inversion list for binary properties */
3024 if (minbits == 1) {
3025 SV** swash_invlistsvp = NULL;
3026 SV* swash_invlist = NULL;
9a53f6cf 3027 bool invlist_in_swash_is_valid = FALSE;
02c85471
FC
3028 bool swash_invlist_unclaimed = FALSE; /* whether swash_invlist has
3029 an unclaimed reference count */
36eb48b4 3030
9a53f6cf 3031 /* If this operation fetched a swash, get its already existing
83199d38 3032 * inversion list, or create one for it */
36eb48b4 3033
83199d38 3034 if (swash_hv) {
5c9f4bd2 3035 swash_invlistsvp = hv_fetchs(swash_hv, "V", FALSE);
9a53f6cf
KW
3036 if (swash_invlistsvp) {
3037 swash_invlist = *swash_invlistsvp;
3038 invlist_in_swash_is_valid = TRUE;
3039 }
3040 else {
36eb48b4 3041 swash_invlist = _swash_to_invlist(retval);
02c85471 3042 swash_invlist_unclaimed = TRUE;
9a53f6cf
KW
3043 }
3044 }
3045
3046 /* If an inversion list was passed in, have to include it */
3047 if (invlist) {
3048
3049 /* Any fetched swash will by now have an inversion list in it;
3050 * otherwise <swash_invlist> will be NULL, indicating that we
3051 * didn't fetch a swash */
3052 if (swash_invlist) {
3053
3054 /* Add the passed-in inversion list, which invalidates the one
3055 * already stored in the swash */
3056 invlist_in_swash_is_valid = FALSE;
3057 _invlist_union(invlist, swash_invlist, &swash_invlist);
3058 }
3059 else {
3060
87367d5f
KW
3061 /* Here, there is no swash already. Set up a minimal one, if
3062 * we are going to return a swash */
3063 if ((int) _invlist_len(invlist) > invlist_swash_boundary) {
971d486f 3064 swash_hv = newHV();
4aca0fe6 3065 retval = newRV_noinc(MUTABLE_SV(swash_hv));
87367d5f 3066 }
9a53f6cf
KW
3067 swash_invlist = invlist;
3068 }
9a53f6cf
KW
3069 }
3070
3071 /* Here, we have computed the union of all the passed-in data. It may
3072 * be that there was an inversion list in the swash which didn't get
3073 * touched; otherwise save the one computed one */
87367d5f
KW
3074 if (! invlist_in_swash_is_valid
3075 && (int) _invlist_len(swash_invlist) > invlist_swash_boundary)
3076 {
5c9f4bd2 3077 if (! hv_stores(MUTABLE_HV(SvRV(retval)), "V", swash_invlist))
69794297
KW
3078 {
3079 Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed");
3080 }
cc34d8c5
FC
3081 /* We just stole a reference count. */
3082 if (swash_invlist_unclaimed) swash_invlist_unclaimed = FALSE;
3083 else SvREFCNT_inc_simple_void_NN(swash_invlist);
9a53f6cf 3084 }
87367d5f 3085
c41800a8 3086 /* Use the inversion list stand-alone if small enough */
87367d5f
KW
3087 if ((int) _invlist_len(swash_invlist) <= invlist_swash_boundary) {
3088 SvREFCNT_dec(retval);
02c85471
FC
3089 if (!swash_invlist_unclaimed)
3090 SvREFCNT_inc_simple_void_NN(swash_invlist);
3091 retval = newRV_noinc(swash_invlist);
87367d5f 3092 }
36eb48b4
KW
3093 }
3094
a0ed51b3
LW
3095 return retval;
3096}
3097
035d37be
JH
3098
3099/* This API is wrong for special case conversions since we may need to
3100 * return several Unicode characters for a single Unicode character
3101 * (see lib/unicore/SpecCase.txt) The SWASHGET in lib/utf8_heavy.pl is
3102 * the lower-level routine, and it is similarly broken for returning
38684baa
KW
3103 * multiple values. --jhi
3104 * For those, you should use to_utf8_case() instead */
b0e3252e 3105/* Now SWASHGET is recasted into S_swatch_get in this file. */
680c470c
ST
3106
3107/* Note:
3108 * Returns the value of property/mapping C<swash> for the first character
3109 * of the string C<ptr>. If C<do_utf8> is true, the string C<ptr> is
3110 * assumed to be in utf8. If C<do_utf8> is false, the string C<ptr> is
3111 * assumed to be in native 8-bit encoding. Caches the swatch in C<swash>.
af2af982
KW
3112 *
3113 * A "swash" is a hash which contains initially the keys/values set up by
3114 * SWASHNEW. The purpose is to be able to completely represent a Unicode
3115 * property for all possible code points. Things are stored in a compact form
3116 * (see utf8_heavy.pl) so that calculation is required to find the actual
3117 * property value for a given code point. As code points are looked up, new
3118 * key/value pairs are added to the hash, so that the calculation doesn't have
3119 * to ever be re-done. Further, each calculation is done, not just for the
3120 * desired one, but for a whole block of code points adjacent to that one.
3121 * For binary properties on ASCII machines, the block is usually for 64 code
3122 * points, starting with a code point evenly divisible by 64. Thus if the
3123 * property value for code point 257 is requested, the code goes out and
3124 * calculates the property values for all 64 code points between 256 and 319,
3125 * and stores these as a single 64-bit long bit vector, called a "swatch",
3126 * under the key for code point 256. The key is the UTF-8 encoding for code
3127 * point 256, minus the final byte. Thus, if the length of the UTF-8 encoding
3128 * for a code point is 13 bytes, the key will be 12 bytes long. If the value
3129 * for code point 258 is then requested, this code realizes that it would be
3130 * stored under the key for 256, and would find that value and extract the
3131 * relevant bit, offset from 256.
3132 *
3133 * Non-binary properties are stored in as many bits as necessary to represent
3134 * their values (32 currently, though the code is more general than that), not
3135 * as single bits, but the principal is the same: the value for each key is a
3136 * vector that encompasses the property values for all code points whose UTF-8
3137 * representations are represented by the key. That is, for all code points
3138 * whose UTF-8 representations are length N bytes, and the key is the first N-1
3139 * bytes of that.
680c470c 3140 */
a0ed51b3 3141UV
680c470c 3142Perl_swash_fetch(pTHX_ SV *swash, const U8 *ptr, bool do_utf8)
a0ed51b3 3143{
27da23d5 3144 dVAR;
ef8f7699 3145 HV *const hv = MUTABLE_HV(SvRV(swash));
3568d838
JH
3146 U32 klen;
3147 U32 off;
a0ed51b3 3148 STRLEN slen;
7d85a32c 3149 STRLEN needents;
cfd0369c 3150 const U8 *tmps = NULL;
a0ed51b3 3151 U32 bit;
979f2922 3152 SV *swatch;
3568d838 3153 U8 tmputf8[2];
f4cd282c 3154 const UV c = *ptr;
3568d838 3155
7918f24d
NC
3156 PERL_ARGS_ASSERT_SWASH_FETCH;
3157
87367d5f
KW
3158 /* If it really isn't a hash, it isn't really swash; must be an inversion
3159 * list */
3160 if (SvTYPE(hv) != SVt_PVHV) {
3161 return _invlist_contains_cp((SV*)hv,
3162 (do_utf8)
3163 ? valid_utf8_to_uvchr(ptr, NULL)
3164 : c);
3165 }
3166
dbe7a391 3167 /* Convert to utf8 if not already */
f4cd282c 3168 if (!do_utf8 && !NATIVE_IS_INVARIANT(c)) {
979f2922
ST
3169 tmputf8[0] = (U8)UTF8_EIGHT_BIT_HI(c);
3170 tmputf8[1] = (U8)UTF8_EIGHT_BIT_LO(c);
3171 ptr = tmputf8;
3568d838
JH
3172 }
3173 /* Given a UTF-X encoded char 0xAA..0xYY,0xZZ
37e2e78e 3174 * then the "swatch" is a vec() for all the chars which start
3568d838
JH
3175 * with 0xAA..0xYY
3176 * So the key in the hash (klen) is length of encoded char -1
3177 */
3178 klen = UTF8SKIP(ptr) - 1;
a0ed51b3 3179
979f2922 3180 if (klen == 0) {
37e2e78e 3181 /* If char is invariant then swatch is for all the invariant chars
1e54db1a 3182 * In both UTF-8 and UTF-8-MOD that happens to be UTF_CONTINUATION_MARK
7d85a32c 3183 */
979f2922 3184 needents = UTF_CONTINUATION_MARK;
bc3632a8 3185 off = NATIVE_UTF8_TO_I8(ptr[klen]);
979f2922
ST
3186 }
3187 else {
7d85a32c 3188 /* If char is encoded then swatch is for the prefix */
979f2922 3189 needents = (1 << UTF_ACCUMULATION_SHIFT);
bc3632a8 3190 off = NATIVE_UTF8_TO_I8(ptr[klen]) & UTF_CONTINUATION_MASK;
979f2922 3191 }
7d85a32c 3192
a0ed51b3
LW
3193 /*
3194 * This single-entry cache saves about 1/3 of the utf8 overhead in test
3195 * suite. (That is, only 7-8% overall over just a hash cache. Still,
3196 * it's nothing to sniff at.) Pity we usually come through at least
3197 * two function calls to get here...
3198 *
3199 * NB: this code assumes that swatches are never modified, once generated!
3200 */
3201
3568d838 3202 if (hv == PL_last_swash_hv &&
a0ed51b3 3203 klen == PL_last_swash_klen &&
27da23d5 3204 (!klen || memEQ((char *)ptr, (char *)PL_last_swash_key, klen)) )
a0ed51b3
LW
3205 {
3206 tmps = PL_last_swash_tmps;
3207 slen = PL_last_swash_slen;
3208 }
3209 else {
3210 /* Try our second-level swatch cache, kept in a hash. */
e1ec3a88 3211 SV** svp = hv_fetch(hv, (const char*)ptr, klen, FALSE);
a0ed51b3 3212
b0e3252e 3213 /* If not cached, generate it via swatch_get */
979f2922
ST
3214 if (!svp || !SvPOK(*svp)
3215 || !(tmps = (const U8*)SvPV_const(*svp, slen))) {
c80e42f3 3216 const UV code_point = utf8n_to_uvchr(ptr, UTF8_MAXBYTES, 0,
872c91ae
JH
3217 ckWARN(WARN_UTF8) ?
3218 0 : UTF8_ALLOW_ANY);
b0e3252e 3219 swatch = swatch_get(swash,
979f2922 3220 /* On EBCDIC & ~(0xA0-1) isn't a useful thing to do */
361ee0fe 3221 (klen) ? (code_point & ~((UV)needents - 1)) : 0,
979f2922
ST
3222 needents);
3223
923e4eb5 3224 if (IN_PERL_COMPILETIME)
623e6609 3225 CopHINTS_set(PL_curcop, PL_hints);
a0ed51b3 3226
979f2922 3227 svp = hv_store(hv, (const char *)ptr, klen, swatch, 0);
a0ed51b3 3228
979f2922
ST
3229 if (!svp || !(tmps = (U8*)SvPV(*svp, slen))
3230 || (slen << 3) < needents)
5637ef5b
NC
3231 Perl_croak(aTHX_ "panic: swash_fetch got improper swatch, "
3232 "svp=%p, tmps=%p, slen=%"UVuf", needents=%"UVuf,
3233 svp, tmps, (UV)slen, (UV)needents);
a0ed51b3
LW
3234 }
3235
3236 PL_last_swash_hv = hv;
16d8f38a 3237 assert(klen <= sizeof(PL_last_swash_key));
eac04b2e 3238 PL_last_swash_klen = (U8)klen;
cfd0369c
NC
3239 /* FIXME change interpvar.h? */
3240 PL_last_swash_tmps = (U8 *) tmps;
a0ed51b3
LW
3241 PL_last_swash_slen = slen;
3242 if (klen)
3243 Copy(ptr, PL_last_swash_key, klen, U8);
3244 }
3245
9faf8d75 3246 switch ((int)((slen << 3) / needents)) {
a0ed51b3
LW
3247 case 1:
3248 bit = 1 << (off & 7);
3249 off >>= 3;
3250 return (tmps[off] & bit) != 0;
3251 case 8:
3252 return tmps[off];
3253 case 16:
3254 off <<= 1;
3255 return (tmps[off] << 8) + tmps[off + 1] ;
3256 case 32:
3257 off <<= 2;
3258 return (tmps[off] << 24) + (tmps[off+1] << 16) + (tmps[off+2] << 8) + tmps[off + 3] ;
3259 }
5637ef5b
NC
3260 Perl_croak(aTHX_ "panic: swash_fetch got swatch of unexpected bit width, "
3261 "slen=%"UVuf", needents=%"UVuf, (UV)slen, (UV)needents);
670f1322 3262 NORETURN_FUNCTION_END;
a0ed51b3 3263}
2b9d42f0 3264
319009ee
KW
3265/* Read a single line of the main body of the swash input text. These are of
3266 * the form:
3267 * 0053 0056 0073
3268 * where each number is hex. The first two numbers form the minimum and
3269 * maximum of a range, and the third is the value associated with the range.
3270 * Not all swashes should have a third number
3271 *
3272 * On input: l points to the beginning of the line to be examined; it points
3273 * to somewhere in the string of the whole input text, and is
3274 * terminated by a \n or the null string terminator.
3275 * lend points to the null terminator of that string
3276 * wants_value is non-zero if the swash expects a third number
3277 * typestr is the name of the swash's mapping, like 'ToLower'
3278 * On output: *min, *max, and *val are set to the values read from the line.
3279 * returns a pointer just beyond the line examined. If there was no
3280 * valid min number on the line, returns lend+1
3281 */
3282
3283STATIC U8*
3284S_swash_scan_list_line(pTHX_ U8* l, U8* const lend, UV* min, UV* max, UV* val,
3285 const bool wants_value, const U8* const typestr)
3286{
3287 const int typeto = typestr[0] == 'T' && typestr[1] == 'o';
3288 STRLEN numlen; /* Length of the number */
02470786
KW
3289 I32 flags = PERL_SCAN_SILENT_ILLDIGIT
3290 | PERL_SCAN_DISALLOW_PREFIX
3291 | PERL_SCAN_SILENT_NON_PORTABLE;
319009ee
KW
3292
3293 /* nl points to the next \n in the scan */
3294 U8* const nl = (U8*)memchr(l, '\n', lend - l);
3295
3296 /* Get the first number on the line: the range minimum */
3297 numlen = lend - l;
3298 *min = grok_hex((char *)l, &numlen, &flags, NULL);
3299 if (numlen) /* If found a hex number, position past it */
3300 l += numlen;
3301 else if (nl) { /* Else, go handle next line, if any */
3302 return nl + 1; /* 1 is length of "\n" */
3303 }
3304 else { /* Else, no next line */
3305 return lend + 1; /* to LIST's end at which \n is not found */
3306 }
3307
3308 /* The max range value follows, separated by a BLANK */
3309 if (isBLANK(*l)) {
3310 ++l;
02470786
KW
3311 flags = PERL_SCAN_SILENT_ILLDIGIT
3312 | PERL_SCAN_DISALLOW_PREFIX
3313 | PERL_SCAN_SILENT_NON_PORTABLE;
319009ee
KW
3314 numlen = lend - l;
3315 *max = grok_hex((char *)l, &numlen, &flags, NULL);
3316 if (numlen)
3317 l += numlen;
3318 else /* If no value here, it is a single element range */
3319 *max = *min;
3320
3321 /* Non-binary tables have a third entry: what the first element of the
3322 * range maps to */
3323 if (wants_value) {
3324 if (isBLANK(*l)) {
3325 ++l;
f90a9a02
KW
3326
3327 /* The ToLc, etc table mappings are not in hex, and must be
3328 * corrected by adding the code point to them */
3329 if (typeto) {
3330 char *after_strtol = (char *) lend;
3331 *val = Strtol((char *)l, &after_strtol, 10);
3332 l = (U8 *) after_strtol;
f90a9a02
KW
3333 }
3334 else { /* Other tables are in hex, and are the correct result
3335 without tweaking */
a9d188b3
KW
3336 flags = PERL_SCAN_SILENT_ILLDIGIT
3337 | PERL_SCAN_DISALLOW_PREFIX
3338 | PERL_SCAN_SILENT_NON_PORTABLE;
3339 numlen = lend - l;
3340 *val = grok_hex((char *)l, &numlen, &flags, NULL);
3341 if (numlen)
3342 l += numlen;
3343 else
3344 *val = 0;
f90a9a02 3345 }
319009ee
KW
3346 }
3347 else {
3348 *val = 0;
3349 if (typeto) {
dcbac5bb 3350 /* diag_listed_as: To%s: illegal mapping '%s' */
319009ee
KW
3351 Perl_croak(aTHX_ "%s: illegal mapping '%s'",
3352 typestr, l);
3353 }
3354 }
3355 }
3356 else
3357 *val = 0; /* bits == 1, then any val should be ignored */
3358 }
3359 else { /* Nothing following range min, should be single element with no
3360 mapping expected */
3361 *max = *min;
3362 if (wants_value) {
3363 *val = 0;
3364 if (typeto) {
dcbac5bb 3365 /* diag_listed_as: To%s: illegal mapping '%s' */
319009ee
KW
3366 Perl_croak(aTHX_ "%s: illegal mapping '%s'", typestr, l);
3367 }
3368 }
3369 else
3370 *val = 0; /* bits == 1, then val should be ignored */
3371 }
3372
3373 /* Position to next line if any, or EOF */
3374 if (nl)
3375 l = nl + 1;
3376 else
3377 l = lend;
3378
3379 return l;
3380}
3381
979f2922
ST
3382/* Note:
3383 * Returns a swatch (a bit vector string) for a code point sequence
3384 * that starts from the value C<start> and comprises the number C<span>.
3385 * A C<swash> must be an object created by SWASHNEW (see lib/utf8_heavy.pl).
3386 * Should be used via swash_fetch, which will cache the swatch in C<swash>.
3387 */
3388STATIC SV*
b0e3252e 3389S_swatch_get(pTHX_ SV* swash, UV start, UV span)
979f2922
ST
3390{
3391 SV *swatch;
77f9f126 3392 U8 *l, *lend, *x, *xend, *s, *send;
979f2922 3393 STRLEN lcur, xcur, scur;
ef8f7699 3394 HV *const hv = MUTABLE_HV(SvRV(swash));
5c9f4bd2 3395 SV** const invlistsvp = hv_fetchs(hv, "V", FALSE);
36eb48b4 3396
88d45d28
KW
3397 SV** listsvp = NULL; /* The string containing the main body of the table */
3398 SV** extssvp = NULL;
3399 SV** invert_it_svp = NULL;
3400 U8* typestr = NULL;
786861f5
KW
3401 STRLEN bits;
3402 STRLEN octets; /* if bits == 1, then octets == 0 */
3403 UV none;
3404 UV end = start + span;
972dd592 3405
36eb48b4 3406 if (invlistsvp == NULL) {
786861f5
KW
3407 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
3408 SV** const nonesvp = hv_fetchs(hv, "NONE", FALSE);
3409 SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE);
3410 extssvp = hv_fetchs(hv, "EXTRAS", FALSE);
3411 listsvp = hv_fetchs(hv, "LIST", FALSE);
3412 invert_it_svp = hv_fetchs(hv, "INVERT_IT", FALSE);
3413
3414 bits = SvUV(*bitssvp);
3415 none = SvUV(*nonesvp);
3416 typestr = (U8*)SvPV_nolen(*typesvp);
3417 }
36eb48b4
KW
3418 else {
3419 bits = 1;
3420 none = 0;
3421 }
786861f5 3422 octets = bits >> 3; /* if bits == 1, then octets == 0 */
979f2922 3423
b0e3252e 3424 PERL_ARGS_ASSERT_SWATCH_GET;
7918f24d 3425
979f2922 3426 if (bits != 1 && bits != 8 && bits != 16 && bits != 32) {
b0e3252e 3427 Perl_croak(aTHX_ "panic: swatch_get doesn't expect bits %"UVuf,
660a4616 3428 (UV)bits);
979f2922
ST
3429 }
3430
84ea5ef6
KW
3431 /* If overflowed, use the max possible */
3432 if (end < start) {
3433 end = UV_MAX;
3434 span = end - start;
3435 }
3436
979f2922 3437 /* create and initialize $swatch */
979f2922 3438 scur = octets ? (span * octets) : (span + 7) / 8;
e524fe40
NC
3439 swatch = newSV(scur);
3440 SvPOK_on(swatch);
979f2922
ST
3441 s = (U8*)SvPVX(swatch);
3442 if (octets && none) {
0bd48802 3443 const U8* const e = s + scur;
979f2922
ST
3444 while (s < e) {
3445 if (bits == 8)
3446 *s++ = (U8)(none & 0xff);
3447 else if (bits == 16) {
3448 *s++ = (U8)((none >> 8) & 0xff);
3449 *s++ = (U8)( none & 0xff);
3450 }
3451 else if (bits == 32) {
3452 *s++ = (U8)((none >> 24) & 0xff);
3453 *s++ = (U8)((none >> 16) & 0xff);
3454 *s++ = (U8)((none >> 8) & 0xff);
3455 *s++ = (U8)( none & 0xff);
3456 }
3457 }
3458 *s = '\0';
3459 }
3460 else {
3461 (void)memzero((U8*)s, scur + 1);
3462 }
3463 SvCUR_set(swatch, scur);
3464 s = (U8*)SvPVX(swatch);
3465
36eb48b4
KW
3466 if (invlistsvp) { /* If has an inversion list set up use that */
3467 _invlist_populate_swatch(*invlistsvp, start, end, s);
3468 return swatch;
3469 }
3470
3471 /* read $swash->{LIST} */
979f2922
ST
3472 l = (U8*)SvPV(*listsvp, lcur);
3473 lend = l + lcur;
3474 while (l < lend) {
8ed25d53 3475 UV min, max, val, upper;
319009ee
KW
3476 l = S_swash_scan_list_line(aTHX_ l, lend, &min, &max, &val,
3477 cBOOL(octets), typestr);
3478 if (l > lend) {
979f2922
ST
3479 break;
3480 }
3481
972dd592 3482 /* If looking for something beyond this range, go try the next one */
979f2922
ST
3483 if (max < start)
3484 continue;
3485
8ed25d53
KW
3486 /* <end> is generally 1 beyond where we want to set things, but at the
3487 * platform's infinity, where we can't go any higher, we want to
3488 * include the code point at <end> */
3489 upper = (max < end)
3490 ? max
3491 : (max != UV_MAX || end != UV_MAX)
3492 ? end - 1
3493 : end;
3494
979f2922 3495 if (octets) {
35da51f7 3496 UV key;
979f2922
ST
3497 if (min < start) {
3498 if (!none || val < none) {
3499 val += start - min;
3500 }
3501 min = start;
3502 }
8ed25d53 3503 for (key = min; key <= upper; key++) {
979f2922 3504 STRLEN offset;
979f2922
ST
3505 /* offset must be non-negative (start <= min <= key < end) */
3506 offset = octets * (key - start);
3507 if (bits == 8)
3508 s[offset] = (U8)(val & 0xff);
3509 else if (bits == 16) {
3510 s[offset ] = (U8)((val >> 8) & 0xff);
3511 s[offset + 1] = (U8)( val & 0xff);
3512 }
3513 else if (bits == 32) {
3514 s[offset ] = (U8)((val >> 24) & 0xff);
3515 s[offset + 1] = (U8)((val >> 16) & 0xff);
3516 s[offset + 2] = (U8)((val >> 8) & 0xff);
3517 s[offset + 3] = (U8)( val & 0xff);
3518 }
3519
3520 if (!none || val < none)
3521 ++val;
3522 }
3523 }
711a919c 3524 else { /* bits == 1, then val should be ignored */
35da51f7 3525 UV key;
979f2922
ST
3526 if (min < start)
3527 min = start;
6cb05c12 3528
8ed25d53 3529 for (key = min; key <= upper; key++) {
0bd48802 3530 const STRLEN offset = (STRLEN)(key - start);
979f2922
ST
3531 s[offset >> 3] |= 1 << (offset & 7);
3532 }
3533 }
3534 } /* while */
979f2922 3535
9479a769 3536 /* Invert if the data says it should be. Assumes that bits == 1 */
77f9f126 3537 if (invert_it_svp && SvUV(*invert_it_svp)) {
0bda3001
KW
3538
3539 /* Unicode properties should come with all bits above PERL_UNICODE_MAX
3540 * be 0, and their inversion should also be 0, as we don't succeed any
3541 * Unicode property matches for non-Unicode code points */
3542 if (start <= PERL_UNICODE_MAX) {
3543
3544 /* The code below assumes that we never cross the
3545 * Unicode/above-Unicode boundary in a range, as otherwise we would
3546 * have to figure out where to stop flipping the bits. Since this
3547 * boundary is divisible by a large power of 2, and swatches comes
3548 * in small powers of 2, this should be a valid assumption */
3549 assert(start + span - 1 <= PERL_UNICODE_MAX);
3550
507a8485
KW
3551 send = s + scur;
3552 while (s < send) {
3553 *s = ~(*s);
3554 s++;
3555 }
0bda3001 3556 }
77f9f126
KW
3557 }
3558
d73c39c5
KW
3559 /* read $swash->{EXTRAS}
3560 * This code also copied to swash_to_invlist() below */
979f2922
ST
3561 x = (U8*)SvPV(*extssvp, xcur);
3562 xend = x + xcur;
3563 while (x < xend) {
3564 STRLEN namelen;
3565 U8 *namestr;
3566 SV** othersvp;
3567 HV* otherhv;
3568 STRLEN otherbits;
3569 SV **otherbitssvp, *other;
711a919c 3570 U8 *s, *o, *nl;
979f2922
ST
3571 STRLEN slen, olen;
3572
35da51f7 3573 const U8 opc = *x++;
979f2922
ST
3574 if (opc == '\n')
3575 continue;
3576
3577 nl = (U8*)memchr(x, '\n', xend - x);
3578
3579 if (opc != '-' && opc != '+' && opc != '!' && opc != '&') {
3580 if (nl) {
3581 x = nl + 1; /* 1 is length of "\n" */
3582 continue;
3583 }
3584 else {
3585 x = xend; /* to EXTRAS' end at which \n is not found */
3586 break;
3587 }
3588 }
3589
3590 namestr = x;
3591 if (nl) {
3592 namelen = nl - namestr;
3593 x = nl + 1;
3594 }
3595 else {
3596 namelen = xend - namestr;
3597 x = xend;
3598 }
3599
3600 othersvp = hv_fetch(hv, (char *)namestr, namelen, FALSE);
ef8f7699 3601 otherhv = MUTABLE_HV(SvRV(*othersvp));
017a3ce5 3602 otherbitssvp = hv_fetchs(otherhv, "BITS", FALSE);
979f2922
ST
3603 otherbits = (STRLEN)SvUV(*otherbitssvp);
3604 if (bits < otherbits)
5637ef5b
NC
3605 Perl_croak(aTHX_ "panic: swatch_get found swatch size mismatch, "
3606 "bits=%"UVuf", otherbits=%"UVuf, (UV)bits, (UV)otherbits);
979f2922
ST
3607
3608 /* The "other" swatch must be destroyed after. */
b0e3252e 3609 other = swatch_get(*othersvp, start, span);
979f2922
ST
3610 o = (U8*)SvPV(other, olen);
3611
3612 if (!olen)
b0e3252e 3613 Perl_croak(aTHX_ "panic: swatch_get got improper swatch");
979f2922
ST
3614
3615 s = (U8*)SvPV(swatch, slen);
3616 if (bits == 1 && otherbits == 1) {
3617 if (slen != olen)
5637ef5b
NC
3618 Perl_croak(aTHX_ "panic: swatch_get found swatch length "
3619 "mismatch, slen=%"UVuf", olen=%"UVuf,
3620 (UV)slen, (UV)olen);
979f2922
ST
3621
3622 switch (opc) {
3623 case '+':
3624 while (slen--)
3625 *s++ |= *o++;
3626 break;
3627 case '!':
3628 while (slen--)
3629 *s++ |= ~*o++;
3630 break;
3631 case '-':
3632 while (slen--)
3633 *s++ &= ~*o++;
3634 break;
3635 case '&':
3636 while (slen--)
3637 *s++ &= *o++;
3638 break;
3639 default:
3640 break;
3641 }
3642 }
711a919c 3643 else {
979f2922
ST
3644 STRLEN otheroctets = otherbits >> 3;
3645 STRLEN offset = 0;
35da51f7 3646 U8* const send = s + slen;
979f2922
ST
3647
3648 while (s < send) {
3649 UV otherval = 0;
3650
3651 if (otherbits == 1) {
3652 otherval = (o[offset >> 3] >> (offset & 7)) & 1;
3653 ++offset;
3654 }
3655 else {
3656 STRLEN vlen = otheroctets;
3657 otherval = *o++;
3658 while (--vlen) {
3659 otherval <<= 8;
3660