This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
utf8.c: Add assertion
[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
LW
33#include "perl.h"
34
a0c21aa1 35#ifndef EBCDIC
970ea3cb 36/* Separate prototypes needed because in ASCII systems these are
a0c21aa1
JH
37 * usually macros but they still are compiled as code, too. */
38PERL_CALLCONV UV Perl_utf8n_to_uvchr(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen, U32 flags);
39PERL_CALLCONV U8* Perl_uvchr_to_utf8(pTHX_ U8 *d, UV uv);
40#endif
41
27da23d5
JH
42static const char unees[] =
43 "Malformed UTF-8 character (unexpected end of string)";
901b21bf 44
48ef279e 45/*
ccfc67b7 46=head1 Unicode Support
a0ed51b3 47
166f8a29
DM
48This file contains various utility functions for manipulating UTF8-encoded
49strings. For the uninitiated, this is a method of representing arbitrary
61296642 50Unicode characters as a variable number of bytes, in such a way that
56da48f7
DM
51characters in the ASCII range are unmodified, and a zero byte never appears
52within non-zero characters.
166f8a29 53
eaf7a4d2
CS
54=cut
55*/
56
57/*
58=for apidoc is_ascii_string
59
a1433954 60Returns true if the first C<len> bytes of the string C<s> are the same whether
970ea3cb
KW
61or not the string is encoded in UTF-8 (or UTF-EBCDIC on EBCDIC machines). That
62is, if they are invariant. On ASCII-ish machines, only ASCII characters
63fit this definition, hence the function's name.
eaf7a4d2 64
9f7e3d64
MH
65If C<len> is 0, it will be calculated using C<strlen(s)>.
66
a1433954 67See also L</is_utf8_string>(), L</is_utf8_string_loclen>(), and L</is_utf8_string_loc>().
eaf7a4d2
CS
68
69=cut
70*/
71
72bool
668b6d8d 73Perl_is_ascii_string(const U8 *s, STRLEN len)
eaf7a4d2
CS
74{
75 const U8* const send = s + (len ? len : strlen((const char *)s));
76 const U8* x = s;
77
78 PERL_ARGS_ASSERT_IS_ASCII_STRING;
eaf7a4d2
CS
79
80 for (; x < send; ++x) {
81 if (!UTF8_IS_INVARIANT(*x))
82 break;
83 }
84
85 return x == send;
86}
87
88/*
87cea99e 89=for apidoc uvuni_to_utf8_flags
eebe1485 90
6ee84de2
KW
91Adds the UTF-8 representation of the code point C<uv> to the end
92of the string C<d>; C<d> should have at least C<UTF8_MAXBYTES+1> free
eebe1485 93bytes available. The return value is the pointer to the byte after the
9041c2e3 94end of the new character. In other words,
eebe1485 95
b851fbc1
JH
96 d = uvuni_to_utf8_flags(d, uv, flags);
97
98or, in most cases,
99
9041c2e3 100 d = uvuni_to_utf8(d, uv);
eebe1485 101
b851fbc1
JH
102(which is equivalent to)
103
104 d = uvuni_to_utf8_flags(d, uv, 0);
105
949cf498 106This is the recommended Unicode-aware way of saying
eebe1485
SC
107
108 *(d++) = uv;
109
949cf498
KW
110This function will convert to UTF-8 (and not warn) even code points that aren't
111legal Unicode or are problematic, unless C<flags> contains one or more of the
a1433954
KW
112following flags:
113
949cf498
KW
114If C<uv> is a Unicode surrogate code point and UNICODE_WARN_SURROGATE is set,
115the function will raise a warning, provided UTF8 warnings are enabled. If instead
116UNICODE_DISALLOW_SURROGATE is set, the function will fail and return NULL.
117If both flags are set, the function will both warn and return NULL.
118
119The UNICODE_WARN_NONCHAR and UNICODE_DISALLOW_NONCHAR flags correspondingly
120affect how the function handles a Unicode non-character. And, likewise for the
121UNICODE_WARN_SUPER and UNICODE_DISALLOW_SUPER flags, and code points that are
122above the Unicode maximum of 0x10FFFF. Code points above 0x7FFF_FFFF (which are
123even less portable) can be warned and/or disallowed even if other above-Unicode
124code points are accepted by the UNICODE_WARN_FE_FF and UNICODE_DISALLOW_FE_FF
125flags.
126
127And finally, the flag UNICODE_WARN_ILLEGAL_INTERCHANGE selects all four of the
128above WARN flags; and UNICODE_DISALLOW_ILLEGAL_INTERCHANGE selects all four
129DISALLOW flags.
130
131
eebe1485
SC
132=cut
133*/
134
dfe13c55 135U8 *
b851fbc1 136Perl_uvuni_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags)
a0ed51b3 137{
7918f24d
NC
138 PERL_ARGS_ASSERT_UVUNI_TO_UTF8_FLAGS;
139
979f77b6
KW
140 /* The first problematic code point is the first surrogate */
141 if (uv >= UNICODE_SURROGATE_FIRST
142 && ckWARN4_d(WARN_UTF8, WARN_SURROGATE, WARN_NON_UNICODE, WARN_NONCHAR))
143 {
949cf498
KW
144 if (UNICODE_IS_SURROGATE(uv)) {
145 if (flags & UNICODE_WARN_SURROGATE) {
8457b38f 146 Perl_ck_warner_d(aTHX_ packWARN(WARN_SURROGATE),
949cf498
KW
147 "UTF-16 surrogate U+%04"UVXf, uv);
148 }
149 if (flags & UNICODE_DISALLOW_SURROGATE) {
150 return NULL;
151 }
152 }
153 else if (UNICODE_IS_SUPER(uv)) {
154 if (flags & UNICODE_WARN_SUPER
155 || (UNICODE_IS_FE_FF(uv) && (flags & UNICODE_WARN_FE_FF)))
156 {
8457b38f 157 Perl_ck_warner_d(aTHX_ packWARN(WARN_NON_UNICODE),
949cf498
KW
158 "Code point 0x%04"UVXf" is not Unicode, may not be portable", uv);
159 }
160 if (flags & UNICODE_DISALLOW_SUPER
161 || (UNICODE_IS_FE_FF(uv) && (flags & UNICODE_DISALLOW_FE_FF)))
162 {
163 return NULL;
164 }
165 }
166 else if (UNICODE_IS_NONCHAR(uv)) {
167 if (flags & UNICODE_WARN_NONCHAR) {
8457b38f 168 Perl_ck_warner_d(aTHX_ packWARN(WARN_NONCHAR),
949cf498
KW
169 "Unicode non-character U+%04"UVXf" is illegal for open interchange",
170 uv);
171 }
172 if (flags & UNICODE_DISALLOW_NONCHAR) {
173 return NULL;
174 }
175 }
507b9800 176 }
c4d5f83a 177 if (UNI_IS_INVARIANT(uv)) {
eb160463 178 *d++ = (U8)UTF_TO_NATIVE(uv);
a0ed51b3
LW
179 return d;
180 }
2d331972 181#if defined(EBCDIC)
1d72bdf6
NIS
182 else {
183 STRLEN len = UNISKIP(uv);
184 U8 *p = d+len-1;
185 while (p > d) {
eb160463 186 *p-- = (U8)UTF_TO_NATIVE((uv & UTF_CONTINUATION_MASK) | UTF_CONTINUATION_MARK);
1d72bdf6
NIS
187 uv >>= UTF_ACCUMULATION_SHIFT;
188 }
eb160463 189 *p = (U8)UTF_TO_NATIVE((uv & UTF_START_MASK(len)) | UTF_START_MARK(len));
1d72bdf6
NIS
190 return d+len;
191 }
192#else /* Non loop style */
a0ed51b3 193 if (uv < 0x800) {
eb160463
GS
194 *d++ = (U8)(( uv >> 6) | 0xc0);
195 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
196 return d;
197 }
198 if (uv < 0x10000) {
eb160463
GS
199 *d++ = (U8)(( uv >> 12) | 0xe0);
200 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
201 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
202 return d;
203 }
204 if (uv < 0x200000) {
eb160463
GS
205 *d++ = (U8)(( uv >> 18) | 0xf0);
206 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
207 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
208 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
209 return d;
210 }
211 if (uv < 0x4000000) {
eb160463
GS
212 *d++ = (U8)(( uv >> 24) | 0xf8);
213 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
214 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
215 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
216 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
217 return d;
218 }
219 if (uv < 0x80000000) {
eb160463
GS
220 *d++ = (U8)(( uv >> 30) | 0xfc);
221 *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80);
222 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
223 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
224 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
225 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
226 return d;
227 }
6b8eaf93 228#ifdef HAS_QUAD
d7578b48 229 if (uv < UTF8_QUAD_MAX)
a0ed51b3
LW
230#endif
231 {
eb160463
GS
232 *d++ = 0xfe; /* Can't match U+FEFF! */
233 *d++ = (U8)(((uv >> 30) & 0x3f) | 0x80);
234 *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80);
235 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
236 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
237 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
238 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
239 return d;
240 }
6b8eaf93 241#ifdef HAS_QUAD
a0ed51b3 242 {
eb160463
GS
243 *d++ = 0xff; /* Can't match U+FFFE! */
244 *d++ = 0x80; /* 6 Reserved bits */
245 *d++ = (U8)(((uv >> 60) & 0x0f) | 0x80); /* 2 Reserved bits */
246 *d++ = (U8)(((uv >> 54) & 0x3f) | 0x80);
247 *d++ = (U8)(((uv >> 48) & 0x3f) | 0x80);
248 *d++ = (U8)(((uv >> 42) & 0x3f) | 0x80);
249 *d++ = (U8)(((uv >> 36) & 0x3f) | 0x80);
250 *d++ = (U8)(((uv >> 30) & 0x3f) | 0x80);
251 *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80);
252 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
253 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
254 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
255 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
256 return d;
257 }
258#endif
1d72bdf6 259#endif /* Loop style */
a0ed51b3 260}
9041c2e3 261
646ca15d
JH
262/*
263
f7d739d1 264Tests if the first C<len> bytes of string C<s> form a valid UTF-8
646ca15d 265character. Note that an INVARIANT (i.e. ASCII) character is a valid
f7d739d1 266UTF-8 character. The number of bytes in the UTF-8 character
646ca15d
JH
267will be returned if it is valid, otherwise 0.
268
269This is the "slow" version as opposed to the "fast" version which is
270the "unrolled" IS_UTF8_CHAR(). E.g. for t/uni/class.t the speed
271difference is a factor of 2 to 3. For lengths (UTF8SKIP(s)) of four
272or less you should use the IS_UTF8_CHAR(), for lengths of five or more
273you should use the _slow(). In practice this means that the _slow()
274will be used very rarely, since the maximum Unicode code point (as of
275Unicode 4.1) is U+10FFFF, which encodes in UTF-8 to four bytes. Only
276the "Perl extended UTF-8" (the infamous 'v-strings') will encode into
277five bytes or more.
278
279=cut */
c053b435 280STATIC STRLEN
5f66b61c 281S_is_utf8_char_slow(const U8 *s, const STRLEN len)
646ca15d 282{
cd7e6c88 283 dTHX; /* The function called below requires thread context */
646ca15d 284
cd7e6c88 285 STRLEN actual_len;
646ca15d 286
cd7e6c88 287 PERL_ARGS_ASSERT_IS_UTF8_CHAR_SLOW;
646ca15d 288
cd7e6c88 289 utf8n_to_uvuni(s, len, &actual_len, UTF8_CHECK_ONLY);
646ca15d 290
cd7e6c88 291 return (actual_len == (STRLEN) -1) ? 0 : actual_len;
646ca15d 292}
9041c2e3
NIS
293
294/*
492a624f
KW
295=for apidoc is_utf8_char_buf
296
297Returns the number of bytes that comprise the first UTF-8 encoded character in
298buffer C<buf>. C<buf_end> should point to one position beyond the end of the
299buffer. 0 is returned if C<buf> does not point to a complete, valid UTF-8
300encoded character.
301
302Note that an INVARIANT character (i.e. ASCII on non-EBCDIC
303machines) is a valid UTF-8 character.
304
305=cut */
306
307STRLEN
308Perl_is_utf8_char_buf(const U8 *buf, const U8* buf_end)
309{
310
311 STRLEN len;
312
313 PERL_ARGS_ASSERT_IS_UTF8_CHAR_BUF;
314
315 if (buf_end <= buf) {
316 return 0;
317 }
318
319 len = buf_end - buf;
320 if (len > UTF8SKIP(buf)) {
321 len = UTF8SKIP(buf);
322 }
323
324#ifdef IS_UTF8_CHAR
325 if (IS_UTF8_CHAR_FAST(len))
326 return IS_UTF8_CHAR(buf, len) ? len : 0;
327#endif /* #ifdef IS_UTF8_CHAR */
328 return is_utf8_char_slow(buf, len);
329}
330
331/*
87cea99e 332=for apidoc is_utf8_char
eebe1485 333
76848387
KW
334DEPRECATED!
335
5da9da9e 336Tests if some arbitrary number of bytes begins in a valid UTF-8
2bbc8d55
SP
337character. Note that an INVARIANT (i.e. ASCII on non-EBCDIC machines)
338character is a valid UTF-8 character. The actual number of bytes in the UTF-8
339character will be returned if it is valid, otherwise 0.
9041c2e3 340
76848387 341This function is deprecated due to the possibility that malformed input could
a1433954 342cause reading beyond the end of the input buffer. Use L</is_utf8_char_buf>
76848387 343instead.
e0328548 344
82686b01 345=cut */
76848387 346
067a85ef 347STRLEN
668b6d8d 348Perl_is_utf8_char(const U8 *s)
386d01d6 349{
7918f24d 350 PERL_ARGS_ASSERT_IS_UTF8_CHAR;
492a624f 351
76848387 352 /* Assumes we have enough space, which is why this is deprecated */
492a624f 353 return is_utf8_char_buf(s, s + UTF8SKIP(s));
386d01d6
GS
354}
355
eaf7a4d2 356
6662521e 357/*
87cea99e 358=for apidoc is_utf8_string
6662521e 359
a1433954 360Returns true if the first C<len> bytes of string C<s> form a valid
9f7e3d64 361UTF-8 string, false otherwise. If C<len> is 0, it will be calculated
e0328548
KW
362using C<strlen(s)> (which means if you use this option, that C<s> has to have a
363terminating NUL byte). Note that all characters being ASCII constitute 'a
364valid UTF-8 string'.
6662521e 365
a1433954 366See also L</is_ascii_string>(), L</is_utf8_string_loclen>(), and L</is_utf8_string_loc>().
768c67ee 367
6662521e
GS
368=cut
369*/
370
8e84507e 371bool
668b6d8d 372Perl_is_utf8_string(const U8 *s, STRLEN len)
6662521e 373{
35da51f7 374 const U8* const send = s + (len ? len : strlen((const char *)s));
7fc63493 375 const U8* x = s;
067a85ef 376
7918f24d 377 PERL_ARGS_ASSERT_IS_UTF8_STRING;
1aa99e6b 378
6662521e 379 while (x < send) {
1acdb0da 380 /* Inline the easy bits of is_utf8_char() here for speed... */
e0328548
KW
381 if (UTF8_IS_INVARIANT(*x)) {
382 x++;
383 }
1acdb0da 384 else if (!UTF8_IS_START(*x))
e0328548 385 return FALSE;
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;
768c67ee
JH
448 else if (!UTF8_IS_START(*x))
449 goto out;
81cd54e3 450 else {
768c67ee 451 /* ... and call is_utf8_char() only if really needed. */
768c67ee 452 c = UTF8SKIP(x);
e0328548
KW
453 next_char_ptr = c + x;
454 if (next_char_ptr > send) {
455 goto out;
456 }
768c67ee
JH
457 if (IS_UTF8_CHAR_FAST(c)) {
458 if (!IS_UTF8_CHAR(x, c))
459 c = 0;
460 } else
461 c = is_utf8_char_slow(x, c);
768c67ee
JH
462 if (!c)
463 goto out;
81cd54e3 464 }
e0328548 465 x = next_char_ptr;
3ebfea28 466 outlen++;
81cd54e3 467 }
768c67ee
JH
468
469 out:
3ebfea28
AL
470 if (el)
471 *el = outlen;
472
768c67ee
JH
473 if (ep)
474 *ep = x;
3ebfea28 475 return (x == send);
81cd54e3
JH
476}
477
478/*
768c67ee 479
87cea99e 480=for apidoc utf8n_to_uvuni
67e989fb 481
9041c2e3 482Bottom level UTF-8 decode routine.
746afd53
KW
483Returns the code point value of the first character in the string C<s>,
484which is assumed to be in UTF-8 (or UTF-EBCDIC) encoding, and no longer than
485C<curlen> bytes; C<*retlen> (if C<retlen> isn't NULL) will be set to
486the length, in bytes, of that character.
949cf498
KW
487
488The value of C<flags> determines the behavior when C<s> does not point to a
489well-formed UTF-8 character. If C<flags> is 0, when a malformation is found,
524080c4
KW
490zero is returned and C<*retlen> is set so that (S<C<s> + C<*retlen>>) is the
491next possible position in C<s> that could begin a non-malformed character.
492Also, if UTF-8 warnings haven't been lexically disabled, a warning is raised.
949cf498
KW
493
494Various ALLOW flags can be set in C<flags> to allow (and not warn on)
495individual types of malformations, such as the sequence being overlong (that
496is, when there is a shorter sequence that can express the same code point;
497overlong sequences are expressly forbidden in the UTF-8 standard due to
498potential security issues). Another malformation example is the first byte of
499a character not being a legal first byte. See F<utf8.h> for the list of such
524080c4
KW
500flags. For allowed 0 length strings, this function returns 0; for allowed
501overlong sequences, the computed code point is returned; for all other allowed
502malformations, the Unicode REPLACEMENT CHARACTER is returned, as these have no
503determinable reasonable value.
949cf498
KW
504
505The UTF8_CHECK_ONLY flag overrides the behavior when a non-allowed (by other
506flags) malformation is found. If this flag is set, the routine assumes that
507the caller will raise a warning, and this function will silently just set
508C<retlen> to C<-1> and return zero.
509
510Certain code points are considered problematic. These are Unicode surrogates,
746afd53 511Unicode non-characters, and code points above the Unicode maximum of 0x10FFFF.
949cf498 512By default these are considered regular code points, but certain situations
5eafe189 513warrant special handling for them. If C<flags> contains
949cf498
KW
514UTF8_DISALLOW_ILLEGAL_INTERCHANGE, all three classes are treated as
515malformations and handled as such. The flags UTF8_DISALLOW_SURROGATE,
516UTF8_DISALLOW_NONCHAR, and UTF8_DISALLOW_SUPER (meaning above the legal Unicode
517maximum) can be set to disallow these categories individually.
518
519The flags UTF8_WARN_ILLEGAL_INTERCHANGE, UTF8_WARN_SURROGATE,
520UTF8_WARN_NONCHAR, and UTF8_WARN_SUPER will cause warning messages to be raised
521for their respective categories, but otherwise the code points are considered
522valid (not malformations). To get a category to both be treated as a
523malformation and raise a warning, specify both the WARN and DISALLOW flags.
524(But note that warnings are not raised if lexically disabled nor if
525UTF8_CHECK_ONLY is also specified.)
526
527Very large code points (above 0x7FFF_FFFF) are considered more problematic than
528the others that are above the Unicode legal maximum. There are several
eb83ed87
KW
529reasons: they requre at least 32 bits to represent them on ASCII platforms, are
530not representable at all on EBCDIC platforms, and the original UTF-8
531specification never went above this number (the current 0x10FFFF limit was
532imposed later). (The smaller ones, those that fit into 32 bits, are
533representable by a UV on ASCII platforms, but not by an IV, which means that
534the number of operations that can be performed on them is quite restricted.)
535The UTF-8 encoding on ASCII platforms for these large code points begins with a
536byte containing 0xFE or 0xFF. The UTF8_DISALLOW_FE_FF flag will cause them to
537be treated as malformations, while allowing smaller above-Unicode code points.
538(Of course UTF8_DISALLOW_SUPER will treat all above-Unicode code points,
539including these, as malformations.) Similarly, UTF8_WARN_FE_FF acts just like
540the other WARN flags, but applies just to these code points.
949cf498
KW
541
542All other code points corresponding to Unicode characters, including private
543use and those yet to be assigned, are never considered malformed and never
544warn.
67e989fb 545
ec5f19d0 546Most code should use L</utf8_to_uvchr_buf>() rather than call this directly.
9041c2e3 547
37607a96
PK
548=cut
549*/
67e989fb 550
a0ed51b3 551UV
7fc63493 552Perl_utf8n_to_uvuni(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen, U32 flags)
a0ed51b3 553{
97aff369 554 dVAR;
d4c19fe8 555 const U8 * const s0 = s;
eb83ed87 556 U8 overflow_byte = '\0'; /* Save byte in case of overflow */
0b8d30e8 557 U8 * send;
eb83ed87
KW
558 UV uv = *s;
559 STRLEN expectlen;
949cf498 560 SV* sv = NULL;
eb83ed87
KW
561 UV outlier_ret = 0; /* return value when input is in error or problematic
562 */
563 UV pack_warn = 0; /* Save result of packWARN() for later */
564 bool unexpected_non_continuation = FALSE;
565 bool overflowed = FALSE;
2f8f112e 566 bool do_overlong_test = TRUE; /* May have to skip this test */
a0dbb045 567
eb83ed87 568 const char* const malformed_text = "Malformed UTF-8 character";
7918f24d 569
eb83ed87 570 PERL_ARGS_ASSERT_UTF8N_TO_UVUNI;
a0dbb045 571
eb83ed87
KW
572 /* The order of malformation tests here is important. We should consume as
573 * few bytes as possible in order to not skip any valid character. This is
574 * required by the Unicode Standard (section 3.9 of Unicode 6.0); see also
575 * http://unicode.org/reports/tr36 for more discussion as to why. For
576 * example, once we've done a UTF8SKIP, we can tell the expected number of
577 * bytes, and could fail right off the bat if the input parameters indicate
578 * that there are too few available. But it could be that just that first
579 * byte is garbled, and the intended character occupies fewer bytes. If we
580 * blindly assumed that the first byte is correct, and skipped based on
581 * that number, we could skip over a valid input character. So instead, we
582 * always examine the sequence byte-by-byte.
583 *
584 * We also should not consume too few bytes, otherwise someone could inject
585 * things. For example, an input could be deliberately designed to
586 * overflow, and if this code bailed out immediately upon discovering that,
587 * returning to the caller *retlen pointing to the very next byte (one
588 * which is actually part of of the overflowing sequence), that could look
589 * legitimate to the caller, which could discard the initial partial
590 * sequence and process the rest, inappropriately */
591
592 /* Zero length strings, if allowed, of necessity are zero */
b5b9af04 593 if (UNLIKELY(curlen == 0)) {
eb83ed87
KW
594 if (retlen) {
595 *retlen = 0;
596 }
a0dbb045 597
eb83ed87
KW
598 if (flags & UTF8_ALLOW_EMPTY) {
599 return 0;
600 }
601 if (! (flags & UTF8_CHECK_ONLY)) {
602 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (empty string)", malformed_text));
603 }
0c443dc2
JH
604 goto malformed;
605 }
606
eb83ed87
KW
607 expectlen = UTF8SKIP(s);
608
609 /* A well-formed UTF-8 character, as the vast majority of calls to this
610 * function will be for, has this expected length. For efficiency, set
611 * things up here to return it. It will be overriden only in those rare
612 * cases where a malformation is found */
613 if (retlen) {
614 *retlen = expectlen;
615 }
616
617 /* An invariant is trivially well-formed */
1d72bdf6 618 if (UTF8_IS_INVARIANT(uv)) {
c4d5f83a 619 return (UV) (NATIVE_TO_UTF(*s));
a0ed51b3 620 }
67e989fb 621
eb83ed87 622 /* A continuation character can't start a valid sequence */
b5b9af04 623 if (UNLIKELY(UTF8_IS_CONTINUATION(uv))) {
eb83ed87
KW
624 if (flags & UTF8_ALLOW_CONTINUATION) {
625 if (retlen) {
626 *retlen = 1;
627 }
628 return UNICODE_REPLACEMENT;
629 }
ba210ebe 630
eb83ed87
KW
631 if (! (flags & UTF8_CHECK_ONLY)) {
632 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (unexpected continuation byte 0x%02x, with no preceding start byte)", malformed_text, *s0));
633 }
634 curlen = 1;
ba210ebe
JH
635 goto malformed;
636 }
9041c2e3 637
1d72bdf6 638#ifdef EBCDIC
75383841 639 uv = NATIVE_TO_UTF(uv);
1d72bdf6
NIS
640#endif
641
eb83ed87
KW
642 /* Here is not a continuation byte, nor an invariant. The only thing left
643 * is a start byte (possibly for an overlong) */
ba210ebe 644
eb83ed87
KW
645 /* Remove the leading bits that indicate the number of bytes in the
646 * character's whole UTF-8 sequence, leaving just the bits that are part of
647 * the value */
648 uv &= UTF_START_MASK(expectlen);
ba210ebe 649
eb83ed87
KW
650 /* Now, loop through the remaining bytes in the character's sequence,
651 * accumulating each into the working value as we go. Be sure to not look
652 * past the end of the input string */
0b8d30e8
KW
653 send = (U8*) s0 + ((expectlen <= curlen) ? expectlen : curlen);
654
eb83ed87 655 for (s = s0 + 1; s < send; s++) {
b5b9af04 656 if (LIKELY(UTF8_IS_CONTINUATION(*s))) {
eb83ed87
KW
657#ifndef EBCDIC /* Can't overflow in EBCDIC */
658 if (uv & UTF_ACCUMULATION_OVERFLOW_MASK) {
659
660 /* The original implementors viewed this malformation as more
661 * serious than the others (though I, khw, don't understand
662 * why, since other malformations also give very very wrong
663 * results), so there is no way to turn off checking for it.
664 * Set a flag, but keep going in the loop, so that we absorb
665 * the rest of the bytes that comprise the character. */
666 overflowed = TRUE;
667 overflow_byte = *s; /* Save for warning message's use */
668 }
669#endif
8850bf83 670 uv = UTF8_ACCUMULATE(uv, *s);
eb83ed87
KW
671 }
672 else {
673 /* Here, found a non-continuation before processing all expected
674 * bytes. This byte begins a new character, so quit, even if
675 * allowing this malformation. */
676 unexpected_non_continuation = TRUE;
677 break;
678 }
679 } /* End of loop through the character's bytes */
680
681 /* Save how many bytes were actually in the character */
682 curlen = s - s0;
683
684 /* The loop above finds two types of malformations: non-continuation and/or
685 * overflow. The non-continuation malformation is really a too-short
686 * malformation, as it means that the current character ended before it was
687 * expected to (being terminated prematurely by the beginning of the next
688 * character, whereas in the too-short malformation there just are too few
689 * bytes available to hold the character. In both cases, the check below
690 * that we have found the expected number of bytes would fail if executed.)
691 * Thus the non-continuation malformation is really unnecessary, being a
692 * subset of the too-short malformation. But there may be existing
693 * applications that are expecting the non-continuation type, so we retain
694 * it, and return it in preference to the too-short malformation. (If this
695 * code were being written from scratch, the two types might be collapsed
696 * into one.) I, khw, am also giving priority to returning the
697 * non-continuation and too-short malformations over overflow when multiple
698 * ones are present. I don't know of any real reason to prefer one over
699 * the other, except that it seems to me that multiple-byte errors trumps
700 * errors from a single byte */
b5b9af04 701 if (UNLIKELY(unexpected_non_continuation)) {
eb83ed87
KW
702 if (!(flags & UTF8_ALLOW_NON_CONTINUATION)) {
703 if (! (flags & UTF8_CHECK_ONLY)) {
704 if (curlen == 1) {
705 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (unexpected non-continuation byte 0x%02x, immediately after start byte 0x%02x)", malformed_text, *s, *s0));
706 }
707 else {
708 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
709 }
710 }
eb83ed87
KW
711 goto malformed;
712 }
713 uv = UNICODE_REPLACEMENT;
2f8f112e
KW
714
715 /* Skip testing for overlongs, as the REPLACEMENT may not be the same
716 * as what the original expectations were. */
717 do_overlong_test = FALSE;
eb83ed87
KW
718 if (retlen) {
719 *retlen = curlen;
720 }
721 }
b5b9af04 722 else if (UNLIKELY(curlen < expectlen)) {
eb83ed87
KW
723 if (! (flags & UTF8_ALLOW_SHORT)) {
724 if (! (flags & UTF8_CHECK_ONLY)) {
725 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 726 }
eb83ed87
KW
727 goto malformed;
728 }
729 uv = UNICODE_REPLACEMENT;
2f8f112e 730 do_overlong_test = FALSE;
eb83ed87
KW
731 if (retlen) {
732 *retlen = curlen;
733 }
734 }
735
736#ifndef EBCDIC /* EBCDIC allows FE, FF, can't overflow */
2f8f112e 737 if ((*s0 & 0xFE) == 0xFE /* matches both FE, FF */
eb83ed87
KW
738 && (flags & (UTF8_WARN_FE_FF|UTF8_DISALLOW_FE_FF)))
739 {
740 /* By adding UTF8_CHECK_ONLY to the test, we avoid unnecessary
741 * generation of the sv, since no warnings are raised under CHECK */
742 if ((flags & (UTF8_WARN_FE_FF|UTF8_CHECK_ONLY)) == UTF8_WARN_FE_FF
743 && ckWARN_d(WARN_UTF8))
744 {
42303544
KW
745 /* This message is deliberately not of the same syntax as the other
746 * messages for malformations, for backwards compatibility in the
747 * unlikely event that code is relying on its precise earlier text
748 */
eb83ed87
KW
749 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s Code point beginning with byte 0x%02X is not Unicode, and not portable", malformed_text, *s0));
750 pack_warn = packWARN(WARN_UTF8);
751 }
752 if (flags & UTF8_DISALLOW_FE_FF) {
753 goto malformed;
ba210ebe 754 }
ba210ebe 755 }
b5b9af04 756 if (UNLIKELY(overflowed)) {
ba210ebe 757
eb83ed87
KW
758 /* If the first byte is FF, it will overflow a 32-bit word. If the
759 * first byte is FE, it will overflow a signed 32-bit word. The
760 * above preserves backward compatibility, since its message was used
761 * in earlier versions of this code in preference to overflow */
762 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (overflow at byte 0x%02x, after start byte 0x%02x)", malformed_text, overflow_byte, *s0));
ba210ebe 763 goto malformed;
eb83ed87
KW
764 }
765#endif
766
2f8f112e
KW
767 if (do_overlong_test
768 && expectlen > (STRLEN)UNISKIP(uv)
769 && ! (flags & UTF8_ALLOW_LONG))
770 {
eb83ed87
KW
771 /* The overlong malformation has lower precedence than the others.
772 * Note that if this malformation is allowed, we return the actual
773 * value, instead of the replacement character. This is because this
774 * value is actually well-defined. */
775 if (! (flags & UTF8_CHECK_ONLY)) {
776 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));
777 }
778 goto malformed;
779 }
780
781 /* Here, the input is considered to be well-formed , but could be a
782 * problematic code point that is not allowed by the input parameters. */
783 if (uv >= UNICODE_SURROGATE_FIRST /* isn't problematic if < this */
784 && (flags & (UTF8_DISALLOW_ILLEGAL_INTERCHANGE
785 |UTF8_WARN_ILLEGAL_INTERCHANGE)))
786 {
949cf498 787 if (UNICODE_IS_SURROGATE(uv)) {
eb83ed87
KW
788 if ((flags & (UTF8_WARN_SURROGATE|UTF8_CHECK_ONLY)) == UTF8_WARN_SURROGATE
789 && ckWARN2_d(WARN_UTF8, WARN_SURROGATE))
790 {
111d382d 791 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "UTF-16 surrogate U+%04"UVXf"", uv));
eb83ed87 792 pack_warn = packWARN2(WARN_UTF8, WARN_SURROGATE);
949cf498
KW
793 }
794 if (flags & UTF8_DISALLOW_SURROGATE) {
795 goto disallowed;
796 }
797 }
949cf498 798 else if ((uv > PERL_UNICODE_MAX)) {
eb83ed87
KW
799 if ((flags & (UTF8_WARN_SUPER|UTF8_CHECK_ONLY)) == UTF8_WARN_SUPER
800 && ckWARN2_d(WARN_UTF8, WARN_NON_UNICODE))
801 {
111d382d 802 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "Code point 0x%04"UVXf" is not Unicode, may not be portable", uv));
eb83ed87 803 pack_warn = packWARN2(WARN_UTF8, WARN_NON_UNICODE);
949cf498
KW
804 }
805 if (flags & UTF8_DISALLOW_SUPER) {
806 goto disallowed;
807 }
808 }
4190d317
KW
809 else if (UNICODE_IS_NONCHAR(uv)) {
810 if ((flags & (UTF8_WARN_NONCHAR|UTF8_CHECK_ONLY)) == UTF8_WARN_NONCHAR
811 && ckWARN2_d(WARN_UTF8, WARN_NONCHAR))
812 {
813 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "Unicode non-character U+%04"UVXf" is illegal for open interchange", uv));
814 pack_warn = packWARN2(WARN_UTF8, WARN_NONCHAR);
815 }
816 if (flags & UTF8_DISALLOW_NONCHAR) {
817 goto disallowed;
818 }
819 }
949cf498 820
eb83ed87
KW
821 if (sv) {
822 outlier_ret = uv;
823 goto do_warn;
824 }
825
949cf498
KW
826 /* Here, this is not considered a malformed character, so drop through
827 * to return it */
a0ed51b3 828 }
ba210ebe 829
a0ed51b3 830 return uv;
ba210ebe 831
eb83ed87
KW
832 /* There are three cases which get to beyond this point. In all 3 cases:
833 * <sv> if not null points to a string to print as a warning.
834 * <curlen> is what <*retlen> should be set to if UTF8_CHECK_ONLY isn't
835 * set.
836 * <outlier_ret> is what return value to use if UTF8_CHECK_ONLY isn't set.
837 * This is done by initializing it to 0, and changing it only
838 * for case 1).
839 * The 3 cases are:
840 * 1) The input is valid but problematic, and to be warned about. The
841 * return value is the resultant code point; <*retlen> is set to
842 * <curlen>, the number of bytes that comprise the code point.
843 * <pack_warn> contains the result of packWARN() for the warning
844 * types. The entry point for this case is the label <do_warn>;
845 * 2) The input is a valid code point but disallowed by the parameters to
846 * this function. The return value is 0. If UTF8_CHECK_ONLY is set,
847 * <*relen> is -1; otherwise it is <curlen>, the number of bytes that
848 * comprise the code point. <pack_warn> contains the result of
849 * packWARN() for the warning types. The entry point for this case is
850 * the label <disallowed>.
851 * 3) The input is malformed. The return value is 0. If UTF8_CHECK_ONLY
852 * is set, <*relen> is -1; otherwise it is <curlen>, the number of
853 * bytes that comprise the malformation. All such malformations are
854 * assumed to be warning type <utf8>. The entry point for this case
855 * is the label <malformed>.
856 */
949cf498 857
ba210ebe
JH
858malformed:
859
eb83ed87
KW
860 if (sv && ckWARN_d(WARN_UTF8)) {
861 pack_warn = packWARN(WARN_UTF8);
862 }
863
864disallowed:
865
fcc8fcf6 866 if (flags & UTF8_CHECK_ONLY) {
ba210ebe 867 if (retlen)
10edeb5d 868 *retlen = ((STRLEN) -1);
ba210ebe
JH
869 return 0;
870 }
871
eb83ed87 872do_warn:
5b311467 873
eb83ed87
KW
874 if (pack_warn) { /* <pack_warn> was initialized to 0, and changed only
875 if warnings are to be raised. */
f555bc63 876 const char * const string = SvPVX_const(sv);
a0dbb045 877
f555bc63
KW
878 if (PL_op)
879 Perl_warner(aTHX_ pack_warn, "%s in %s", string, OP_DESC(PL_op));
880 else
881 Perl_warner(aTHX_ pack_warn, "%s", string);
a0dbb045
JH
882 }
883
eb83ed87
KW
884 if (retlen) {
885 *retlen = curlen;
886 }
ba210ebe 887
eb83ed87 888 return outlier_ret;
a0ed51b3
LW
889}
890
8e84507e 891/*
ec5f19d0
KW
892=for apidoc utf8_to_uvchr_buf
893
894Returns the native code point of the first character in the string C<s> which
895is assumed to be in UTF-8 encoding; C<send> points to 1 beyond the end of C<s>.
524080c4 896C<*retlen> will be set to the length, in bytes, of that character.
ec5f19d0 897
524080c4
KW
898If C<s> does not point to a well-formed UTF-8 character and UTF8 warnings are
899enabled, zero is returned and C<*retlen> is set (if C<retlen> isn't
900NULL) to -1. If those warnings are off, the computed value if well-defined (or
901the Unicode REPLACEMENT CHARACTER, if not) is silently returned, and C<*retlen>
902is set (if C<retlen> isn't NULL) so that (S<C<s> + C<*retlen>>) is the
903next possible position in C<s> that could begin a non-malformed character.
904See L</utf8n_to_uvuni> for details on when the REPLACEMENT CHARACTER is returned.
ec5f19d0
KW
905
906=cut
907*/
908
909
910UV
911Perl_utf8_to_uvchr_buf(pTHX_ const U8 *s, const U8 *send, STRLEN *retlen)
912{
913 PERL_ARGS_ASSERT_UTF8_TO_UVCHR_BUF;
914
915 assert(s < send);
916
917 return utf8n_to_uvchr(s, send - s, retlen,
918 ckWARN_d(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
919}
920
27d6c58a 921/* Like L</utf8_to_uvchr_buf>(), but should only be called when it is known that
3986bb7c
KW
922 * there are no malformations in the input UTF-8 string C<s>. surrogates,
923 * non-character code points, and non-Unicode code points are allowed */
27d6c58a
KW
924
925UV
926Perl_valid_utf8_to_uvchr(pTHX_ const U8 *s, STRLEN *retlen)
927{
3986bb7c
KW
928 const UV uv = valid_utf8_to_uvuni(s, retlen);
929
27d6c58a
KW
930 PERL_ARGS_ASSERT_VALID_UTF8_TO_UVCHR;
931
3986bb7c 932 return UNI_TO_NATIVE(uv);
27d6c58a
KW
933}
934
ec5f19d0 935/*
87cea99e 936=for apidoc utf8_to_uvchr
9041c2e3 937
977c1d31
KW
938DEPRECATED!
939
6ee84de2 940Returns the native code point of the first character in the string C<s>
1e54db1a 941which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
9041c2e3
NIS
942length, in bytes, of that character.
943
4b88fb76 944Some, but not all, UTF-8 malformations are detected, and in fact, some
977c1d31
KW
945malformed input could cause reading beyond the end of the input buffer, which
946is why this function is deprecated. Use L</utf8_to_uvchr_buf> instead.
4b88fb76 947
524080c4
KW
948If C<s> points to one of the detected malformations, and UTF8 warnings are
949enabled, zero is returned and C<*retlen> is set (if C<retlen> isn't
950NULL) to -1. If those warnings are off, the computed value if well-defined (or
951the Unicode REPLACEMENT CHARACTER, if not) is silently returned, and C<*retlen>
952is set (if C<retlen> isn't NULL) so that (S<C<s> + C<*retlen>>) is the
953next possible position in C<s> that could begin a non-malformed character.
954See L</utf8n_to_uvuni> for details on when the REPLACEMENT CHARACTER is returned.
9041c2e3
NIS
955
956=cut
957*/
958
959UV
7fc63493 960Perl_utf8_to_uvchr(pTHX_ const U8 *s, STRLEN *retlen)
9041c2e3 961{
7918f24d
NC
962 PERL_ARGS_ASSERT_UTF8_TO_UVCHR;
963
2ff6c191 964 return utf8_to_uvchr_buf(s, s + UTF8_MAXBYTES, retlen);
9041c2e3
NIS
965}
966
967/*
ec5f19d0
KW
968=for apidoc utf8_to_uvuni_buf
969
970Returns the Unicode code point of the first character in the string C<s> which
971is assumed to be in UTF-8 encoding; C<send> points to 1 beyond the end of C<s>.
972C<retlen> will be set to the length, in bytes, of that character.
973
974This function should only be used when the returned UV is considered
975an index into the Unicode semantic tables (e.g. swashes).
976
524080c4
KW
977If C<s> does not point to a well-formed UTF-8 character and UTF8 warnings are
978enabled, zero is returned and C<*retlen> is set (if C<retlen> isn't
979NULL) to -1. If those warnings are off, the computed value if well-defined (or
980the Unicode REPLACEMENT CHARACTER, if not) is silently returned, and C<*retlen>
981is set (if C<retlen> isn't NULL) so that (S<C<s> + C<*retlen>>) is the
982next possible position in C<s> that could begin a non-malformed character.
983See L</utf8n_to_uvuni> for details on when the REPLACEMENT CHARACTER is returned.
ec5f19d0
KW
984
985=cut
986*/
987
988UV
989Perl_utf8_to_uvuni_buf(pTHX_ const U8 *s, const U8 *send, STRLEN *retlen)
990{
991 PERL_ARGS_ASSERT_UTF8_TO_UVUNI_BUF;
992
993 assert(send > s);
994
995 /* Call the low level routine asking for checks */
996 return Perl_utf8n_to_uvuni(aTHX_ s, send -s, retlen,
997 ckWARN_d(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
998}
999
27d6c58a 1000/* Like L</utf8_to_uvuni_buf>(), but should only be called when it is known that
3986bb7c
KW
1001 * there are no malformations in the input UTF-8 string C<s>. surrogates,
1002 * non-character code points, and non-Unicode code points are allowed */
27d6c58a
KW
1003
1004UV
1005Perl_valid_utf8_to_uvuni(pTHX_ const U8 *s, STRLEN *retlen)
1006{
3986bb7c
KW
1007 UV expectlen = UTF8SKIP(s);
1008 const U8* send = s + expectlen;
1009 UV uv = NATIVE_TO_UTF(*s);
1010
27d6c58a
KW
1011 PERL_ARGS_ASSERT_VALID_UTF8_TO_UVUNI;
1012
3986bb7c
KW
1013 if (retlen) {
1014 *retlen = expectlen;
1015 }
1016
1017 /* An invariant is trivially returned */
1018 if (expectlen == 1) {
1019 return uv;
1020 }
1021
1022 /* Remove the leading bits that indicate the number of bytes, leaving just
1023 * the bits that are part of the value */
1024 uv &= UTF_START_MASK(expectlen);
1025
1026 /* Now, loop through the remaining bytes, accumulating each into the
1027 * working total as we go */
1028 for (++s; s < send; s++) {
1029 uv = UTF8_ACCUMULATE(uv, *s);
1030 }
1031
1032 return uv;
27d6c58a
KW
1033}
1034
ec5f19d0 1035/*
87cea99e 1036=for apidoc utf8_to_uvuni
9041c2e3 1037
977c1d31
KW
1038DEPRECATED!
1039
9041c2e3 1040Returns the Unicode code point of the first character in the string C<s>
1e54db1a 1041which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
9041c2e3
NIS
1042length, in bytes, of that character.
1043
39e518fd
KW
1044This function should only be used when the returned UV is considered
1045an index into the Unicode semantic tables (e.g. swashes).
1046
4b88fb76 1047Some, but not all, UTF-8 malformations are detected, and in fact, some
977c1d31
KW
1048malformed input could cause reading beyond the end of the input buffer, which
1049is why this function is deprecated. Use L</utf8_to_uvuni_buf> instead.
9041c2e3 1050
524080c4
KW
1051If C<s> points to one of the detected malformations, and UTF8 warnings are
1052enabled, zero is returned and C<*retlen> is set (if C<retlen> doesn't point to
1053NULL) to -1. If those warnings are off, the computed value if well-defined (or
1054the Unicode REPLACEMENT CHARACTER, if not) is silently returned, and C<*retlen>
1055is set (if C<retlen> isn't NULL) so that (S<C<s> + C<*retlen>>) is the
1056next possible position in C<s> that could begin a non-malformed character.
1057See L</utf8n_to_uvuni> for details on when the REPLACEMENT CHARACTER is returned.
8e84507e
NIS
1058
1059=cut
1060*/
1061
1062UV
7fc63493 1063Perl_utf8_to_uvuni(pTHX_ const U8 *s, STRLEN *retlen)
8e84507e 1064{
7918f24d
NC
1065 PERL_ARGS_ASSERT_UTF8_TO_UVUNI;
1066
4b88fb76 1067 return valid_utf8_to_uvuni(s, retlen);
8e84507e
NIS
1068}
1069
b76347f2 1070/*
87cea99e 1071=for apidoc utf8_length
b76347f2
JH
1072
1073Return the length of the UTF-8 char encoded string C<s> in characters.
02eb7b47
JH
1074Stops at C<e> (inclusive). If C<e E<lt> s> or if the scan would end
1075up past C<e>, croaks.
b76347f2
JH
1076
1077=cut
1078*/
1079
1080STRLEN
35a4481c 1081Perl_utf8_length(pTHX_ const U8 *s, const U8 *e)
b76347f2 1082{
97aff369 1083 dVAR;
b76347f2
JH
1084 STRLEN len = 0;
1085
7918f24d
NC
1086 PERL_ARGS_ASSERT_UTF8_LENGTH;
1087
8850bf83
JH
1088 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g.
1089 * the bitops (especially ~) can create illegal UTF-8.
1090 * In other words: in Perl UTF-8 is not just for Unicode. */
1091
a3b680e6
AL
1092 if (e < s)
1093 goto warn_and_return;
b76347f2 1094 while (s < e) {
8e91ec7f
AV
1095 if (!UTF8_IS_INVARIANT(*s))
1096 s += UTF8SKIP(s);
1097 else
1098 s++;
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)) {
356979f4 1204 c = UNI_TO_NATIVE(TWO_BYTE_UTF8_TO_UNI(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 */
356979f4 1333 c = UNI_TO_NATIVE(TWO_BYTE_UTF8_TO_UNI(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) {
35a4481c 1376 const UV uv = NATIVE_TO_ASCII(*s++);
c4d5f83a 1377 if (UNI_IS_INVARIANT(uv))
eb160463 1378 *d++ = (U8)UTF_TO_NATIVE(uv);
6940069f 1379 else {
eb160463
GS
1380 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
1381 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
6940069f
GS
1382 }
1383 }
1384 *d = '\0';
6662521e 1385 *len = d-dst;
6940069f
GS
1386 return dst;
1387}
1388
a0ed51b3 1389/*
dea0fc0b 1390 * Convert native (big-endian) or reversed (little-endian) UTF-16 to UTF-8.
a0ed51b3
LW
1391 *
1392 * Destination must be pre-extended to 3/2 source. Do not use in-place.
1393 * We optimize for native, for obvious reasons. */
1394
1395U8*
dea0fc0b 1396Perl_utf16_to_utf8(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen)
a0ed51b3 1397{
dea0fc0b
JH
1398 U8* pend;
1399 U8* dstart = d;
1400
7918f24d
NC
1401 PERL_ARGS_ASSERT_UTF16_TO_UTF8;
1402
dea0fc0b 1403 if (bytelen & 1)
f5992bc4 1404 Perl_croak(aTHX_ "panic: utf16_to_utf8: odd bytelen %"UVuf, (UV)bytelen);
dea0fc0b
JH
1405
1406 pend = p + bytelen;
1407
a0ed51b3 1408 while (p < pend) {
dea0fc0b
JH
1409 UV uv = (p[0] << 8) + p[1]; /* UTF-16BE */
1410 p += 2;
a0ed51b3 1411 if (uv < 0x80) {
e294cc5d
JH
1412#ifdef EBCDIC
1413 *d++ = UNI_TO_NATIVE(uv);
1414#else
eb160463 1415 *d++ = (U8)uv;
e294cc5d 1416#endif
a0ed51b3
LW
1417 continue;
1418 }
1419 if (uv < 0x800) {
eb160463
GS
1420 *d++ = (U8)(( uv >> 6) | 0xc0);
1421 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
1422 continue;
1423 }
52b9aa85 1424 if (uv >= 0xd800 && uv <= 0xdbff) { /* surrogates */
01ea242b 1425 if (p >= pend) {
dea0fc0b 1426 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
01ea242b
NC
1427 } else {
1428 UV low = (p[0] << 8) + p[1];
1429 p += 2;
52b9aa85 1430 if (low < 0xdc00 || low > 0xdfff)
01ea242b
NC
1431 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
1432 uv = ((uv - 0xd800) << 10) + (low - 0xdc00) + 0x10000;
1433 }
dbde1951
NC
1434 } else if (uv >= 0xdc00 && uv <= 0xdfff) {
1435 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
a0ed51b3
LW
1436 }
1437 if (uv < 0x10000) {
eb160463
GS
1438 *d++ = (U8)(( uv >> 12) | 0xe0);
1439 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
1440 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
1441 continue;
1442 }
1443 else {
eb160463
GS
1444 *d++ = (U8)(( uv >> 18) | 0xf0);
1445 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
1446 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
1447 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
1448 continue;
1449 }
1450 }
dea0fc0b 1451 *newlen = d - dstart;
a0ed51b3
LW
1452 return d;
1453}
1454
1455/* Note: this one is slightly destructive of the source. */
1456
1457U8*
dea0fc0b 1458Perl_utf16_to_utf8_reversed(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen)
a0ed51b3
LW
1459{
1460 U8* s = (U8*)p;
d4c19fe8 1461 U8* const send = s + bytelen;
7918f24d
NC
1462
1463 PERL_ARGS_ASSERT_UTF16_TO_UTF8_REVERSED;
1464
e0ea5e2d
NC
1465 if (bytelen & 1)
1466 Perl_croak(aTHX_ "panic: utf16_to_utf8_reversed: odd bytelen %"UVuf,
1467 (UV)bytelen);
1468
a0ed51b3 1469 while (s < send) {
d4c19fe8 1470 const U8 tmp = s[0];
a0ed51b3
LW
1471 s[0] = s[1];
1472 s[1] = tmp;
1473 s += 2;
1474 }
dea0fc0b 1475 return utf16_to_utf8(p, d, bytelen, newlen);
a0ed51b3
LW
1476}
1477
c3fd2246
KW
1478/* for now these are all defined (inefficiently) in terms of the utf8 versions.
1479 * Note that the macros in handy.h that call these short-circuit calling them
1480 * for Latin-1 range inputs */
a0ed51b3
LW
1481
1482bool
84afefe6 1483Perl_is_uni_alnum(pTHX_ UV c)
a0ed51b3 1484{
89ebb4a3 1485 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1486 uvchr_to_utf8(tmpbuf, c);
a0ed51b3
LW
1487 return is_utf8_alnum(tmpbuf);
1488}
1489
1490bool
84afefe6 1491Perl_is_uni_idfirst(pTHX_ UV c)
a0ed51b3 1492{
89ebb4a3 1493 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1494 uvchr_to_utf8(tmpbuf, c);
a0ed51b3
LW
1495 return is_utf8_idfirst(tmpbuf);
1496}
1497
1498bool
84afefe6 1499Perl_is_uni_alpha(pTHX_ UV c)
a0ed51b3 1500{
89ebb4a3 1501 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1502 uvchr_to_utf8(tmpbuf, c);
a0ed51b3
LW
1503 return is_utf8_alpha(tmpbuf);
1504}
1505
1506bool
84afefe6 1507Perl_is_uni_ascii(pTHX_ UV c)
4d61ec05 1508{
bc39fe24 1509 return isASCII(c);
4d61ec05
GS
1510}
1511
1512bool
84afefe6 1513Perl_is_uni_space(pTHX_ UV c)
a0ed51b3 1514{
89ebb4a3 1515 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1516 uvchr_to_utf8(tmpbuf, c);
a0ed51b3
LW
1517 return is_utf8_space(tmpbuf);
1518}
1519
1520bool
84afefe6 1521Perl_is_uni_digit(pTHX_ UV c)
a0ed51b3 1522{
89ebb4a3 1523 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1524 uvchr_to_utf8(tmpbuf, c);
a0ed51b3
LW
1525 return is_utf8_digit(tmpbuf);
1526}
1527
1528bool
84afefe6 1529Perl_is_uni_upper(pTHX_ UV c)
a0ed51b3 1530{
89ebb4a3 1531 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1532 uvchr_to_utf8(tmpbuf, c);
a0ed51b3
LW
1533 return is_utf8_upper(tmpbuf);
1534}
1535
1536bool
84afefe6 1537Perl_is_uni_lower(pTHX_ UV c)
a0ed51b3 1538{
89ebb4a3 1539 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1540 uvchr_to_utf8(tmpbuf, c);
a0ed51b3
LW
1541 return is_utf8_lower(tmpbuf);
1542}
1543
1544bool
84afefe6 1545Perl_is_uni_cntrl(pTHX_ UV c)
b8c5462f 1546{
7b952154 1547 return isCNTRL_L1(c);
b8c5462f
JH
1548}
1549
1550bool
84afefe6 1551Perl_is_uni_graph(pTHX_ UV c)
b8c5462f 1552{
89ebb4a3 1553 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1554 uvchr_to_utf8(tmpbuf, c);
b8c5462f
JH
1555 return is_utf8_graph(tmpbuf);
1556}
1557
1558bool
84afefe6 1559Perl_is_uni_print(pTHX_ UV c)
a0ed51b3 1560{
89ebb4a3 1561 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1562 uvchr_to_utf8(tmpbuf, c);
a0ed51b3
LW
1563 return is_utf8_print(tmpbuf);
1564}
1565
b8c5462f 1566bool
84afefe6 1567Perl_is_uni_punct(pTHX_ UV c)
b8c5462f 1568{
89ebb4a3 1569 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1570 uvchr_to_utf8(tmpbuf, c);
b8c5462f
JH
1571 return is_utf8_punct(tmpbuf);
1572}
1573
4d61ec05 1574bool
84afefe6 1575Perl_is_uni_xdigit(pTHX_ UV c)
4d61ec05 1576{
89ebb4a3 1577 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
230880c1 1578 uvchr_to_utf8(tmpbuf, c);
4d61ec05
GS
1579 return is_utf8_xdigit(tmpbuf);
1580}
1581
3a4c58c9
KW
1582UV
1583Perl__to_upper_title_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp, const char S_or_s)
1584{
1585 /* We have the latin1-range values compiled into the core, so just use
1586 * those, converting the result to utf8. The only difference between upper
1587 * and title case in this range is that LATIN_SMALL_LETTER_SHARP_S is
1588 * either "SS" or "Ss". Which one to use is passed into the routine in
1589 * 'S_or_s' to avoid a test */
1590
1591 UV converted = toUPPER_LATIN1_MOD(c);
1592
1593 PERL_ARGS_ASSERT__TO_UPPER_TITLE_LATIN1;
1594
1595 assert(S_or_s == 'S' || S_or_s == 's');
1596
1597 if (UNI_IS_INVARIANT(converted)) { /* No difference between the two for
1598 characters in this range */
1599 *p = (U8) converted;
1600 *lenp = 1;
1601 return converted;
1602 }
1603
1604 /* toUPPER_LATIN1_MOD gives the correct results except for three outliers,
1605 * which it maps to one of them, so as to only have to have one check for
1606 * it in the main case */
1607 if (UNLIKELY(converted == LATIN_SMALL_LETTER_Y_WITH_DIAERESIS)) {
1608 switch (c) {
1609 case LATIN_SMALL_LETTER_Y_WITH_DIAERESIS:
1610 converted = LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS;
1611 break;
1612 case MICRO_SIGN:
1613 converted = GREEK_CAPITAL_LETTER_MU;
1614 break;
1615 case LATIN_SMALL_LETTER_SHARP_S:
1616 *(p)++ = 'S';
1617 *p = S_or_s;
1618 *lenp = 2;
1619 return 'S';
1620 default:
1621 Perl_croak(aTHX_ "panic: to_upper_title_latin1 did not expect '%c' to map to '%c'", c, LATIN_SMALL_LETTER_Y_WITH_DIAERESIS);
1622 /* NOTREACHED */
1623 }
1624 }
1625
1626 *(p)++ = UTF8_TWO_BYTE_HI(converted);
1627 *p = UTF8_TWO_BYTE_LO(converted);
1628 *lenp = 2;
1629
1630 return converted;
1631}
1632
50bda2c3
KW
1633/* Call the function to convert a UTF-8 encoded character to the specified case.
1634 * Note that there may be more than one character in the result.
1635 * INP is a pointer to the first byte of the input character
1636 * OUTP will be set to the first byte of the string of changed characters. It
1637 * needs to have space for UTF8_MAXBYTES_CASE+1 bytes
1638 * LENP will be set to the length in bytes of the string of changed characters
1639 *
1640 * The functions return the ordinal of the first character in the string of OUTP */
f90a9a02
KW
1641#define CALL_UPPER_CASE(INP, OUTP, LENP) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_toupper, "ToUc", "utf8::ToSpecUc")
1642#define CALL_TITLE_CASE(INP, OUTP, LENP) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_totitle, "ToTc", "utf8::ToSpecTc")
1643#define CALL_LOWER_CASE(INP, OUTP, LENP) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_tolower, "ToLc", "utf8::ToSpecLc")
50bda2c3
KW
1644
1645/* This additionally has the input parameter SPECIALS, which if non-zero will
1646 * cause this to use the SPECIALS hash for folding (meaning get full case
1647 * folding); otherwise, when zero, this implies a simple case fold */
f90a9a02 1648#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 1649
84afefe6
JH
1650UV
1651Perl_to_uni_upper(pTHX_ UV c, U8* p, STRLEN *lenp)
a0ed51b3 1652{
3a4c58c9
KW
1653 dVAR;
1654
a1433954
KW
1655 /* Convert the Unicode character whose ordinal is <c> to its uppercase
1656 * version and store that in UTF-8 in <p> and its length in bytes in <lenp>.
1657 * Note that the <p> needs to be at least UTF8_MAXBYTES_CASE+1 bytes since
c3fd2246
KW
1658 * the changed version may be longer than the original character.
1659 *
1660 * The ordinal of the first character of the changed version is returned
1661 * (but note, as explained above, that there may be more.) */
1662
7918f24d
NC
1663 PERL_ARGS_ASSERT_TO_UNI_UPPER;
1664
3a4c58c9
KW
1665 if (c < 256) {
1666 return _to_upper_title_latin1((U8) c, p, lenp, 'S');
1667 }
1668
0ebc6274 1669 uvchr_to_utf8(p, c);
3a4c58c9 1670 return CALL_UPPER_CASE(p, p, lenp);
a0ed51b3
LW
1671}
1672
84afefe6
JH
1673UV
1674Perl_to_uni_title(pTHX_ UV c, U8* p, STRLEN *lenp)
a0ed51b3 1675{
3a4c58c9
KW
1676 dVAR;
1677
7918f24d
NC
1678 PERL_ARGS_ASSERT_TO_UNI_TITLE;
1679
3a4c58c9
KW
1680 if (c < 256) {
1681 return _to_upper_title_latin1((U8) c, p, lenp, 's');
1682 }
1683
0ebc6274 1684 uvchr_to_utf8(p, c);
3a4c58c9 1685 return CALL_TITLE_CASE(p, p, lenp);
a0ed51b3
LW
1686}
1687
afc16117
KW
1688STATIC U8
1689S_to_lower_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp)
1690{
1691 /* We have the latin1-range values compiled into the core, so just use
1692 * those, converting the result to utf8. Since the result is always just
a1433954 1693 * one character, we allow <p> to be NULL */
afc16117
KW
1694
1695 U8 converted = toLOWER_LATIN1(c);
1696
1697 if (p != NULL) {
1698 if (UNI_IS_INVARIANT(converted)) {
1699 *p = converted;
1700 *lenp = 1;
1701 }
1702 else {
1703 *p = UTF8_TWO_BYTE_HI(converted);
1704 *(p+1) = UTF8_TWO_BYTE_LO(converted);
1705 *lenp = 2;
1706 }
1707 }
1708 return converted;
1709}
1710
84afefe6
JH
1711UV
1712Perl_to_uni_lower(pTHX_ UV c, U8* p, STRLEN *lenp)
a0ed51b3 1713{
968c5e6a
KW
1714 dVAR;
1715
7918f24d
NC
1716 PERL_ARGS_ASSERT_TO_UNI_LOWER;
1717
afc16117
KW
1718 if (c < 256) {
1719 return to_lower_latin1((U8) c, p, lenp);
bca00c02
KW
1720 }
1721
afc16117 1722 uvchr_to_utf8(p, c);
968c5e6a 1723 return CALL_LOWER_CASE(p, p, lenp);
a0ed51b3
LW
1724}
1725
84afefe6 1726UV
f673fad4 1727Perl__to_fold_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp, const bool flags)
a1dde8de 1728{
a1433954 1729 /* Corresponds to to_lower_latin1(), <flags> is TRUE if to use full case
f673fad4
KW
1730 * folding */
1731
a1dde8de
KW
1732 UV converted;
1733
1734 PERL_ARGS_ASSERT__TO_FOLD_LATIN1;
1735
1736 if (c == MICRO_SIGN) {
1737 converted = GREEK_SMALL_LETTER_MU;
1738 }
1739 else if (flags && c == LATIN_SMALL_LETTER_SHARP_S) {
1740 *(p)++ = 's';
1741 *p = 's';
1742 *lenp = 2;
1743 return 's';
1744 }
1745 else { /* In this range the fold of all other characters is their lower
1746 case */
1747 converted = toLOWER_LATIN1(c);
1748 }
1749
1750 if (UNI_IS_INVARIANT(converted)) {
1751 *p = (U8) converted;
1752 *lenp = 1;
1753 }
1754 else {
1755 *(p)++ = UTF8_TWO_BYTE_HI(converted);
1756 *p = UTF8_TWO_BYTE_LO(converted);
1757 *lenp = 2;
1758 }
1759
1760 return converted;
1761}
1762
1763UV
f673fad4 1764Perl__to_uni_fold_flags(pTHX_ UV c, U8* p, STRLEN *lenp, const bool flags)
84afefe6 1765{
4b593389
KW
1766
1767 /* Not currently externally documented, and subject to change, <flags> is
f673fad4 1768 * TRUE iff full folding is to be used */
4b593389 1769
36bb2ab6 1770 PERL_ARGS_ASSERT__TO_UNI_FOLD_FLAGS;
7918f24d 1771
a1dde8de
KW
1772 if (c < 256) {
1773 return _to_fold_latin1((U8) c, p, lenp, flags);
1774 }
1775
0ebc6274 1776 uvchr_to_utf8(p, c);
a1dde8de 1777 return CALL_FOLD_CASE(p, p, lenp, flags);
84afefe6
JH
1778}
1779
ea317ccb
KW
1780/* for now these all assume no locale info available for Unicode > 255; and
1781 * the corresponding macros in handy.h (like isALNUM_LC_uvchr) should have been
1782 * called instead, so that these don't get called for < 255 */
a0ed51b3
LW
1783
1784bool
84afefe6 1785Perl_is_uni_alnum_lc(pTHX_ UV c)
a0ed51b3
LW
1786{
1787 return is_uni_alnum(c); /* XXX no locale support yet */
1788}
1789
1790bool
84afefe6 1791Perl_is_uni_idfirst_lc(pTHX_ UV c)
a0ed51b3
LW
1792{
1793 return is_uni_idfirst(c); /* XXX no locale support yet */
1794}
1795
1796bool
84afefe6 1797Perl_is_uni_alpha_lc(pTHX_ UV c)
a0ed51b3
LW
1798{
1799 return is_uni_alpha(c); /* XXX no locale support yet */
1800}
1801
1802bool
84afefe6 1803Perl_is_uni_ascii_lc(pTHX_ UV c)
4d61ec05
GS
1804{
1805 return is_uni_ascii(c); /* XXX no locale support yet */
1806}
1807
1808bool
84afefe6 1809Perl_is_uni_space_lc(pTHX_ UV c)
a0ed51b3
LW
1810{
1811 return is_uni_space(c); /* XXX no locale support yet */
1812}
1813
1814bool
84afefe6 1815Perl_is_uni_digit_lc(pTHX_ UV c)
a0ed51b3
LW
1816{
1817 return is_uni_digit(c); /* XXX no locale support yet */
1818}
1819
1820bool
84afefe6 1821Perl_is_uni_upper_lc(pTHX_ UV c)
a0ed51b3
LW
1822{
1823 return is_uni_upper(c); /* XXX no locale support yet */
1824}
1825
1826bool
84afefe6 1827Perl_is_uni_lower_lc(pTHX_ UV c)
a0ed51b3
LW
1828{
1829 return is_uni_lower(c); /* XXX no locale support yet */
1830}
1831
1832bool
84afefe6 1833Perl_is_uni_cntrl_lc(pTHX_ UV c)
b8c5462f
JH
1834{
1835 return is_uni_cntrl(c); /* XXX no locale support yet */
1836}
1837
1838bool
84afefe6 1839Perl_is_uni_graph_lc(pTHX_ UV c)
b8c5462f
JH
1840{
1841 return is_uni_graph(c); /* XXX no locale support yet */
1842}
1843
1844bool
84afefe6 1845Perl_is_uni_print_lc(pTHX_ UV c)
a0ed51b3
LW
1846{
1847 return is_uni_print(c); /* XXX no locale support yet */
1848}
1849
b8c5462f 1850bool
84afefe6 1851Perl_is_uni_punct_lc(pTHX_ UV c)
b8c5462f
JH
1852{
1853 return is_uni_punct(c); /* XXX no locale support yet */
1854}
1855
4d61ec05 1856bool
84afefe6 1857Perl_is_uni_xdigit_lc(pTHX_ UV c)
4d61ec05
GS
1858{
1859 return is_uni_xdigit(c); /* XXX no locale support yet */
1860}
1861
b7ac61fa
JH
1862U32
1863Perl_to_uni_upper_lc(pTHX_ U32 c)
1864{
ee099d14
JH
1865 /* XXX returns only the first character -- do not use XXX */
1866 /* XXX no locale support yet */
1867 STRLEN len;
89ebb4a3 1868 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
ee099d14 1869 return (U32)to_uni_upper(c, tmpbuf, &len);
b7ac61fa
JH
1870}
1871
1872U32
1873Perl_to_uni_title_lc(pTHX_ U32 c)
1874{
ee099d14
JH
1875 /* XXX returns only the first character XXX -- do not use XXX */
1876 /* XXX no locale support yet */
1877 STRLEN len;
89ebb4a3 1878 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
ee099d14 1879 return (U32)to_uni_title(c, tmpbuf, &len);
b7ac61fa
JH
1880}
1881
1882U32
1883Perl_to_uni_lower_lc(pTHX_ U32 c)
1884{
ee099d14
JH
1885 /* XXX returns only the first character -- do not use XXX */
1886 /* XXX no locale support yet */
1887 STRLEN len;
89ebb4a3 1888 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
ee099d14 1889 return (U32)to_uni_lower(c, tmpbuf, &len);
b7ac61fa
JH
1890}
1891
7452cf6a 1892static bool
5141f98e 1893S_is_utf8_common(pTHX_ const U8 *const p, SV **swash,
bde6a22d
NC
1894 const char *const swashname)
1895{
ea317ccb
KW
1896 /* returns a boolean giving whether or not the UTF8-encoded character that
1897 * starts at <p> is in the swash indicated by <swashname>. <swash>
1898 * contains a pointer to where the swash indicated by <swashname>
1899 * is to be stored; which this routine will do, so that future calls will
1900 * look at <*swash> and only generate a swash if it is not null
1901 *
1902 * Note that it is assumed that the buffer length of <p> is enough to
1903 * contain all the bytes that comprise the character. Thus, <*p> should
1904 * have been checked before this call for mal-formedness enough to assure
1905 * that. */
1906
97aff369 1907 dVAR;
7918f24d
NC
1908
1909 PERL_ARGS_ASSERT_IS_UTF8_COMMON;
1910
492a624f
KW
1911 /* The API should have included a length for the UTF-8 character in <p>,
1912 * but it doesn't. We therefor assume that p has been validated at least
1913 * as far as there being enough bytes available in it to accommodate the
1914 * character without reading beyond the end, and pass that number on to the
1915 * validating routine */
1916 if (!is_utf8_char_buf(p, p + UTF8SKIP(p)))
bde6a22d
NC
1917 return FALSE;
1918 if (!*swash)
711a919c 1919 *swash = swash_init("utf8", swashname, &PL_sv_undef, 1, 0);
bde6a22d
NC
1920 return swash_fetch(*swash, p, TRUE) != 0;
1921}
1922
1923bool
7fc63493 1924Perl_is_utf8_alnum(pTHX_ const U8 *p)
a0ed51b3 1925{
97aff369 1926 dVAR;
7918f24d
NC
1927
1928 PERL_ARGS_ASSERT_IS_UTF8_ALNUM;
1929
671c33bf
NC
1930 /* NOTE: "IsWord", not "IsAlnum", since Alnum is a true
1931 * descendant of isalnum(3), in other words, it doesn't
1932 * contain the '_'. --jhi */
d4c19fe8 1933 return is_utf8_common(p, &PL_utf8_alnum, "IsWord");
a0ed51b3
LW
1934}
1935
1936bool
7fc63493 1937Perl_is_utf8_idfirst(pTHX_ const U8 *p) /* The naming is historical. */
a0ed51b3 1938{
97aff369 1939 dVAR;
7918f24d
NC
1940
1941 PERL_ARGS_ASSERT_IS_UTF8_IDFIRST;
1942
82686b01
JH
1943 if (*p == '_')
1944 return TRUE;
bde6a22d 1945 /* is_utf8_idstart would be more logical. */
d4c19fe8 1946 return is_utf8_common(p, &PL_utf8_idstart, "IdStart");
82686b01
JH
1947}
1948
1949bool
c11ff943
KW
1950Perl_is_utf8_xidfirst(pTHX_ const U8 *p) /* The naming is historical. */
1951{
1952 dVAR;
1953
1954 PERL_ARGS_ASSERT_IS_UTF8_XIDFIRST;
1955
1956 if (*p == '_')
1957 return TRUE;
1958 /* is_utf8_idstart would be more logical. */
1959 return is_utf8_common(p, &PL_utf8_xidstart, "XIdStart");
1960}
1961
1962bool
b6912c02
KW
1963Perl__is_utf8__perl_idstart(pTHX_ const U8 *p)
1964{
1965 dVAR;
1966
1967 PERL_ARGS_ASSERT__IS_UTF8__PERL_IDSTART;
1968
1969 return is_utf8_common(p, &PL_utf8_perl_idstart, "_Perl_IDStart");
1970}
1971
1972bool
7fc63493 1973Perl_is_utf8_idcont(pTHX_ const U8 *p)
82686b01 1974{
97aff369 1975 dVAR;
7918f24d
NC
1976
1977 PERL_ARGS_ASSERT_IS_UTF8_IDCONT;
1978
d4c19fe8 1979 return is_utf8_common(p, &PL_utf8_idcont, "IdContinue");
a0ed51b3
LW
1980}
1981
1982bool
c11ff943
KW
1983Perl_is_utf8_xidcont(pTHX_ const U8 *p)
1984{
1985 dVAR;
1986
1987 PERL_ARGS_ASSERT_IS_UTF8_XIDCONT;
1988
c11ff943
KW
1989 return is_utf8_common(p, &PL_utf8_idcont, "XIdContinue");
1990}
1991
1992bool
7fc63493 1993Perl_is_utf8_alpha(pTHX_ const U8 *p)
a0ed51b3 1994{
97aff369 1995 dVAR;
7918f24d
NC
1996
1997 PERL_ARGS_ASSERT_IS_UTF8_ALPHA;
1998
d4c19fe8 1999 return is_utf8_common(p, &PL_utf8_alpha, "IsAlpha");
a0ed51b3
LW
2000}
2001
2002bool
7fc63493 2003Perl_is_utf8_ascii(pTHX_ const U8 *p)
b8c5462f 2004{
97aff369 2005 dVAR;
7918f24d
NC
2006
2007 PERL_ARGS_ASSERT_IS_UTF8_ASCII;
2008
bc39fe24
KW
2009 /* ASCII characters are the same whether in utf8 or not. So the macro
2010 * works on both utf8 and non-utf8 representations. */
2011 return isASCII(*p);
b8c5462f
JH
2012}
2013
2014bool
7fc63493 2015Perl_is_utf8_space(pTHX_ const U8 *p)
a0ed51b3 2016{
97aff369 2017 dVAR;
7918f24d
NC
2018
2019 PERL_ARGS_ASSERT_IS_UTF8_SPACE;
2020
a34094a9 2021 return is_utf8_common(p, &PL_utf8_space, "IsXPerlSpace");
a0ed51b3
LW
2022}
2023
2024bool
d1eb3177
YO
2025Perl_is_utf8_perl_space(pTHX_ const U8 *p)
2026{
2027 dVAR;
2028
2029 PERL_ARGS_ASSERT_IS_UTF8_PERL_SPACE;
2030
c4428693
KW
2031 /* Only true if is an ASCII space-like character, and ASCII is invariant
2032 * under utf8, so can just use the macro */
2033 return isSPACE_A(*p);
d1eb3177
YO
2034}
2035
2036bool
2037Perl_is_utf8_perl_word(pTHX_ const U8 *p)
2038{
2039 dVAR;
2040
2041 PERL_ARGS_ASSERT_IS_UTF8_PERL_WORD;
2042
c4428693
KW
2043 /* Only true if is an ASCII word character, and ASCII is invariant
2044 * under utf8, so can just use the macro */
2045 return isWORDCHAR_A(*p);
d1eb3177
YO
2046}
2047
2048bool
7fc63493 2049Perl_is_utf8_digit(pTHX_ const U8 *p)
a0ed51b3 2050{
97aff369 2051 dVAR;
7918f24d
NC
2052
2053 PERL_ARGS_ASSERT_IS_UTF8_DIGIT;
2054
d4c19fe8 2055 return is_utf8_common(p, &PL_utf8_digit, "IsDigit");
a0ed51b3
LW
2056}
2057
2058bool
d1eb3177
YO
2059Perl_is_utf8_posix_digit(pTHX_ const U8 *p)
2060{
2061 dVAR;
2062
2063 PERL_ARGS_ASSERT_IS_UTF8_POSIX_DIGIT;
2064
c4428693
KW
2065 /* Only true if is an ASCII digit character, and ASCII is invariant
2066 * under utf8, so can just use the macro */
2067 return isDIGIT_A(*p);
d1eb3177
YO
2068}
2069
2070bool
7fc63493 2071Perl_is_utf8_upper(pTHX_ const U8 *p)
a0ed51b3 2072{
97aff369 2073 dVAR;
7918f24d
NC
2074
2075 PERL_ARGS_ASSERT_IS_UTF8_UPPER;
2076
d4c19fe8 2077 return is_utf8_common(p, &PL_utf8_upper, "IsUppercase");
a0ed51b3
LW
2078}
2079
2080bool
7fc63493 2081Perl_is_utf8_lower(pTHX_ const U8 *p)
a0ed51b3 2082{
97aff369 2083 dVAR;
7918f24d
NC
2084
2085 PERL_ARGS_ASSERT_IS_UTF8_LOWER;
2086
d4c19fe8 2087 return is_utf8_common(p, &PL_utf8_lower, "IsLowercase");
a0ed51b3
LW
2088}
2089
2090bool
7fc63493 2091Perl_is_utf8_cntrl(pTHX_ const U8 *p)
b8c5462f 2092{
97aff369 2093 dVAR;
7918f24d
NC
2094
2095 PERL_ARGS_ASSERT_IS_UTF8_CNTRL;
2096
7b952154
KW
2097 if (isASCII(*p)) {
2098 return isCNTRL_A(*p);
2099 }
2100
2101 /* All controls are in Latin1 */
2102 if (! UTF8_IS_DOWNGRADEABLE_START(*p)) {
2103 return 0;
2104 }
2105 return isCNTRL_L1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)));
b8c5462f
JH
2106}
2107
2108bool
7fc63493 2109Perl_is_utf8_graph(pTHX_ const U8 *p)
b8c5462f 2110{
97aff369 2111 dVAR;
7918f24d
NC
2112
2113 PERL_ARGS_ASSERT_IS_UTF8_GRAPH;
2114
d4c19fe8 2115 return is_utf8_common(p, &PL_utf8_graph, "IsGraph");
b8c5462f
JH
2116}
2117
2118bool
7fc63493 2119Perl_is_utf8_print(pTHX_ const U8 *p)
a0ed51b3 2120{
97aff369 2121 dVAR;
7918f24d
NC
2122
2123 PERL_ARGS_ASSERT_IS_UTF8_PRINT;
2124
d4c19fe8 2125 return is_utf8_common(p, &PL_utf8_print, "IsPrint");
a0ed51b3
LW
2126}
2127
2128bool
7fc63493 2129Perl_is_utf8_punct(pTHX_ const U8 *p)
b8c5462f 2130{
97aff369 2131 dVAR;
7918f24d
NC
2132
2133 PERL_ARGS_ASSERT_IS_UTF8_PUNCT;
2134
d4c19fe8 2135 return is_utf8_common(p, &PL_utf8_punct, "IsPunct");
b8c5462f
JH
2136}
2137
2138bool
7fc63493 2139Perl_is_utf8_xdigit(pTHX_ const U8 *p)
b8c5462f 2140{
97aff369 2141 dVAR;
7918f24d
NC
2142
2143 PERL_ARGS_ASSERT_IS_UTF8_XDIGIT;
2144
d1eb3177 2145 return is_utf8_common(p, &PL_utf8_xdigit, "IsXDigit");
b8c5462f
JH
2146}
2147
2148bool
7fc63493 2149Perl_is_utf8_mark(pTHX_ const U8 *p)
a0ed51b3 2150{
97aff369 2151 dVAR;
7918f24d
NC
2152
2153 PERL_ARGS_ASSERT_IS_UTF8_MARK;
2154
d4c19fe8 2155 return is_utf8_common(p, &PL_utf8_mark, "IsM");
a0ed51b3
LW
2156}
2157
37e2e78e
KW
2158bool
2159Perl_is_utf8_X_begin(pTHX_ const U8 *p)
2160{
2161 dVAR;
2162
2163 PERL_ARGS_ASSERT_IS_UTF8_X_BEGIN;
2164
2165 return is_utf8_common(p, &PL_utf8_X_begin, "_X_Begin");
2166}
2167
2168bool
2169Perl_is_utf8_X_extend(pTHX_ const U8 *p)
2170{
2171 dVAR;
2172
2173 PERL_ARGS_ASSERT_IS_UTF8_X_EXTEND;
2174
2175 return is_utf8_common(p, &PL_utf8_X_extend, "_X_Extend");
2176}
2177
2178bool
2179Perl_is_utf8_X_prepend(pTHX_ const U8 *p)
2180{
2181 dVAR;
2182
2183 PERL_ARGS_ASSERT_IS_UTF8_X_PREPEND;
2184
2185 return is_utf8_common(p, &PL_utf8_X_prepend, "GCB=Prepend");
2186}
2187
2188bool
2189Perl_is_utf8_X_non_hangul(pTHX_ const U8 *p)
2190{
2191 dVAR;
2192
2193 PERL_ARGS_ASSERT_IS_UTF8_X_NON_HANGUL;
2194
2195 return is_utf8_common(p, &PL_utf8_X_non_hangul, "HST=Not_Applicable");
2196}
2197
2198bool
2199Perl_is_utf8_X_L(pTHX_ const U8 *p)
2200{
2201 dVAR;
2202
2203 PERL_ARGS_ASSERT_IS_UTF8_X_L;
2204
2205 return is_utf8_common(p, &PL_utf8_X_L, "GCB=L");
2206}
2207
2208bool
2209Perl_is_utf8_X_LV(pTHX_ const U8 *p)
2210{
2211 dVAR;
2212
2213 PERL_ARGS_ASSERT_IS_UTF8_X_LV;
2214
2215 return is_utf8_common(p, &PL_utf8_X_LV, "GCB=LV");
2216}
2217
2218bool
2219Perl_is_utf8_X_LVT(pTHX_ const U8 *p)
2220{
2221 dVAR;
2222
2223 PERL_ARGS_ASSERT_IS_UTF8_X_LVT;
2224
2225 return is_utf8_common(p, &PL_utf8_X_LVT, "GCB=LVT");
2226}
2227
2228bool
2229Perl_is_utf8_X_T(pTHX_ const U8 *p)
2230{
2231 dVAR;
2232
2233 PERL_ARGS_ASSERT_IS_UTF8_X_T;
2234
2235 return is_utf8_common(p, &PL_utf8_X_T, "GCB=T");
2236}
2237
2238bool
2239Perl_is_utf8_X_V(pTHX_ const U8 *p)
2240{
2241 dVAR;
2242
2243 PERL_ARGS_ASSERT_IS_UTF8_X_V;
2244
2245 return is_utf8_common(p, &PL_utf8_X_V, "GCB=V");
2246}
2247
2248bool
2249Perl_is_utf8_X_LV_LVT_V(pTHX_ const U8 *p)
2250{
2251 dVAR;
2252
2253 PERL_ARGS_ASSERT_IS_UTF8_X_LV_LVT_V;
2254
2255 return is_utf8_common(p, &PL_utf8_X_LV_LVT_V, "_X_LV_LVT_V");
2256}
2257
2e2b2571
KW
2258bool
2259Perl__is_utf8_quotemeta(pTHX_ const U8 *p)
2260{
2261 /* For exclusive use of pp_quotemeta() */
2262
2263 dVAR;
2264
2265 PERL_ARGS_ASSERT__IS_UTF8_QUOTEMETA;
2266
2267 return is_utf8_common(p, &PL_utf8_quotemeta, "_Perl_Quotemeta");
2268}
2269
6b5c0936 2270/*
87cea99e 2271=for apidoc to_utf8_case
6b5c0936 2272
a1433954
KW
2273The C<p> contains the pointer to the UTF-8 string encoding
2274the character that is being converted. This routine assumes that the character
2275at C<p> is well-formed.
6b5c0936 2276
a1433954
KW
2277The C<ustrp> is a pointer to the character buffer to put the
2278conversion result to. The C<lenp> is a pointer to the length
6b5c0936
JH
2279of the result.
2280
a1433954 2281The C<swashp> is a pointer to the swash to use.
6b5c0936 2282
a1433954
KW
2283Both the special and normal mappings are stored in F<lib/unicore/To/Foo.pl>,
2284and loaded by SWASHNEW, using F<lib/utf8_heavy.pl>. The C<special> (usually,
0134edef 2285but not always, a multicharacter mapping), is tried first.
6b5c0936 2286
a1433954 2287The C<special> is a string like "utf8::ToSpecLower", which means the
0134edef
JH
2288hash %utf8::ToSpecLower. The access to the hash is through
2289Perl_to_utf8_case().
6b5c0936 2290
a1433954 2291The C<normal> is a string like "ToLower" which means the swash
0134edef
JH
2292%utf8::ToLower.
2293
2294=cut */
6b5c0936 2295
2104c8d9 2296UV
9a957fbc
AL
2297Perl_to_utf8_case(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp,
2298 SV **swashp, const char *normal, const char *special)
a0ed51b3 2299{
97aff369 2300 dVAR;
89ebb4a3 2301 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
0134edef 2302 STRLEN len = 0;
4b88fb76 2303 const UV uv0 = valid_utf8_to_uvchr(p, NULL);
1feea2c7
JH
2304 /* The NATIVE_TO_UNI() and UNI_TO_NATIVE() mappings
2305 * are necessary in EBCDIC, they are redundant no-ops
2306 * in ASCII-ish platforms, and hopefully optimized away. */
f54cb97a 2307 const UV uv1 = NATIVE_TO_UNI(uv0);
7918f24d
NC
2308
2309 PERL_ARGS_ASSERT_TO_UTF8_CASE;
2310
9ae3ac1a
KW
2311 /* Note that swash_fetch() doesn't output warnings for these because it
2312 * assumes we will */
8457b38f 2313 if (uv1 >= UNICODE_SURROGATE_FIRST) {
9ae3ac1a 2314 if (uv1 <= UNICODE_SURROGATE_LAST) {
8457b38f
KW
2315 if (ckWARN_d(WARN_SURROGATE)) {
2316 const char* desc = (PL_op) ? OP_DESC(PL_op) : normal;
2317 Perl_warner(aTHX_ packWARN(WARN_SURROGATE),
2318 "Operation \"%s\" returns its argument for UTF-16 surrogate U+%04"UVXf"", desc, uv1);
2319 }
9ae3ac1a
KW
2320 }
2321 else if (UNICODE_IS_SUPER(uv1)) {
8457b38f
KW
2322 if (ckWARN_d(WARN_NON_UNICODE)) {
2323 const char* desc = (PL_op) ? OP_DESC(PL_op) : normal;
2324 Perl_warner(aTHX_ packWARN(WARN_NON_UNICODE),
2325 "Operation \"%s\" returns its argument for non-Unicode code point 0x%04"UVXf"", desc, uv1);
2326 }
9ae3ac1a
KW
2327 }
2328
2329 /* Note that non-characters are perfectly legal, so no warning should
2330 * be given */
2331 }
2332
1feea2c7 2333 uvuni_to_utf8(tmpbuf, uv1);
0134edef
JH
2334
2335 if (!*swashp) /* load on-demand */
2336 *swashp = swash_init("utf8", normal, &PL_sv_undef, 4, 0);
2337
a6f87d8c 2338 if (special) {
0134edef 2339 /* It might be "special" (sometimes, but not always,
2a37f04d 2340 * a multicharacter mapping) */
6673a63c 2341 HV * const hv = get_hv(special, 0);
b08cf34e
JH
2342 SV **svp;
2343
35da51f7 2344 if (hv &&
b08cf34e
JH
2345 (svp = hv_fetch(hv, (const char*)tmpbuf, UNISKIP(uv1), FALSE)) &&
2346 (*svp)) {
cfd0369c 2347 const char *s;
47654450 2348
cfd0369c 2349 s = SvPV_const(*svp, len);
47654450
JH
2350 if (len == 1)
2351 len = uvuni_to_utf8(ustrp, NATIVE_TO_UNI(*(U8*)s)) - ustrp;
2a37f04d 2352 else {
2f9475ad
JH
2353#ifdef EBCDIC
2354 /* If we have EBCDIC we need to remap the characters
2355 * since any characters in the low 256 are Unicode
2356 * code points, not EBCDIC. */
7cda7a3d 2357 U8 *t = (U8*)s, *tend = t + len, *d;
2f9475ad
JH
2358
2359 d = tmpbuf;
b08cf34e 2360 if (SvUTF8(*svp)) {
2f9475ad
JH
2361 STRLEN tlen = 0;
2362
2363 while (t < tend) {
6bd1c396 2364 const UV c = utf8_to_uvchr_buf(t, tend, &tlen);
2f9475ad
JH
2365 if (tlen > 0) {
2366 d = uvchr_to_utf8(d, UNI_TO_NATIVE(c));
2367 t += tlen;
2368 }
2369 else
2370 break;
2371 }
2372 }
2373 else {
36fec512
JH
2374 while (t < tend) {
2375 d = uvchr_to_utf8(d, UNI_TO_NATIVE(*t));
2376 t++;
2377 }
2f9475ad
JH
2378 }
2379 len = d - tmpbuf;
2380 Copy(tmpbuf, ustrp, len, U8);
2381#else
d2dcd0fb 2382 Copy(s, ustrp, len, U8);
2f9475ad 2383#endif
29e98929 2384 }
983ffd37 2385 }
0134edef
JH
2386 }
2387
2388 if (!len && *swashp) {
d4c19fe8
AL
2389 const UV uv2 = swash_fetch(*swashp, tmpbuf, TRUE);
2390
0134edef
JH
2391 if (uv2) {
2392 /* It was "normal" (a single character mapping). */
d4c19fe8 2393 const UV uv3 = UNI_TO_NATIVE(uv2);
e9101d72 2394 len = uvchr_to_utf8(ustrp, uv3) - ustrp;
2a37f04d
JH
2395 }
2396 }
1feea2c7 2397
37e2e78e
KW
2398 if (!len) /* Neither: just copy. In other words, there was no mapping
2399 defined, which means that the code point maps to itself */
0134edef
JH
2400 len = uvchr_to_utf8(ustrp, uv0) - ustrp;
2401
2a37f04d
JH
2402 if (lenp)
2403 *lenp = len;
2404
6bd1c396 2405 return len ? valid_utf8_to_uvchr(ustrp, 0) : 0;
a0ed51b3
LW
2406}
2407
051a06d4
KW
2408STATIC UV
2409S_check_locale_boundary_crossing(pTHX_ const U8* const p, const UV result, U8* const ustrp, STRLEN *lenp)
2410{
2411 /* This is called when changing the case of a utf8-encoded character above
2412 * the Latin1 range, and the operation is in locale. If the result
2413 * contains a character that crosses the 255/256 boundary, disallow the
2414 * change, and return the original code point. See L<perlfunc/lc> for why;
2415 *
a1433954
KW
2416 * p points to the original string whose case was changed; assumed
2417 * by this routine to be well-formed
051a06d4
KW
2418 * result the code point of the first character in the changed-case string
2419 * ustrp points to the changed-case string (<result> represents its first char)
2420 * lenp points to the length of <ustrp> */
2421
2422 UV original; /* To store the first code point of <p> */
2423
2424 PERL_ARGS_ASSERT_CHECK_LOCALE_BOUNDARY_CROSSING;
2425
2426 assert(! UTF8_IS_INVARIANT(*p) && ! UTF8_IS_DOWNGRADEABLE_START(*p));
2427
2428 /* We know immediately if the first character in the string crosses the
2429 * boundary, so can skip */
2430 if (result > 255) {
2431
2432 /* Look at every character in the result; if any cross the
2433 * boundary, the whole thing is disallowed */
2434 U8* s = ustrp + UTF8SKIP(ustrp);
2435 U8* e = ustrp + *lenp;
2436 while (s < e) {
2437 if (UTF8_IS_INVARIANT(*s) || UTF8_IS_DOWNGRADEABLE_START(*s))
2438 {
2439 goto bad_crossing;
2440 }
2441 s += UTF8SKIP(s);
2442 }
2443
2444 /* Here, no characters crossed, result is ok as-is */
2445 return result;
2446 }
2447
2448bad_crossing:
2449
2450 /* Failed, have to return the original */
4b88fb76 2451 original = valid_utf8_to_uvchr(p, lenp);
051a06d4
KW
2452 Copy(p, ustrp, *lenp, char);
2453 return original;
2454}
2455
d3e79532 2456/*
87cea99e 2457=for apidoc to_utf8_upper
d3e79532 2458
a1433954
KW
2459Convert the UTF-8 encoded character at C<p> to its uppercase version and
2460store that in UTF-8 in C<ustrp> and its length in bytes in C<lenp>. Note
89ebb4a3
JH
2461that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since
2462the uppercase version may be longer than the original character.
d3e79532
JH
2463
2464The first character of the uppercased version is returned
2465(but note, as explained above, that there may be more.)
2466
a1433954
KW
2467The character at C<p> is assumed by this routine to be well-formed.
2468
d3e79532
JH
2469=cut */
2470
051a06d4
KW
2471/* Not currently externally documented, and subject to change:
2472 * <flags> is set iff locale semantics are to be used for code points < 256
2473 * <tainted_ptr> if non-null, *tainted_ptr will be set TRUE iff locale rules
2474 * were used in the calculation; otherwise unchanged. */
2475
2104c8d9 2476UV
051a06d4 2477Perl__to_utf8_upper_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, const bool flags, bool* tainted_ptr)
a0ed51b3 2478{
97aff369 2479 dVAR;
7918f24d 2480
051a06d4
KW
2481 UV result;
2482
2483 PERL_ARGS_ASSERT__TO_UTF8_UPPER_FLAGS;
7918f24d 2484
3a4c58c9 2485 if (UTF8_IS_INVARIANT(*p)) {
051a06d4
KW
2486 if (flags) {
2487 result = toUPPER_LC(*p);
2488 }
2489 else {
81c6c7ce 2490 return _to_upper_title_latin1(*p, ustrp, lenp, 'S');
051a06d4 2491 }
3a4c58c9
KW
2492 }
2493 else if UTF8_IS_DOWNGRADEABLE_START(*p) {
051a06d4
KW
2494 if (flags) {
2495 result = toUPPER_LC(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)));
2496 }
2497 else {
81c6c7ce
KW
2498 return _to_upper_title_latin1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)),
2499 ustrp, lenp, 'S');
051a06d4
KW
2500 }
2501 }
2502 else { /* utf8, ord above 255 */
2503 result = CALL_UPPER_CASE(p, ustrp, lenp);
2504
2505 if (flags) {
2506 result = check_locale_boundary_crossing(p, result, ustrp, lenp);
2507 }
2508 return result;
2509 }
2510
2511 /* Here, used locale rules. Convert back to utf8 */
2512 if (UTF8_IS_INVARIANT(result)) {
2513 *ustrp = (U8) result;
2514 *lenp = 1;
2515 }
2516 else {
2517 *ustrp = UTF8_EIGHT_BIT_HI(result);
2518 *(ustrp + 1) = UTF8_EIGHT_BIT_LO(result);
2519 *lenp = 2;
3a4c58c9
KW
2520 }
2521
051a06d4
KW
2522 if (tainted_ptr) {
2523 *tainted_ptr = TRUE;
2524 }
2525 return result;
983ffd37 2526}
a0ed51b3 2527
d3e79532 2528/*
87cea99e 2529=for apidoc to_utf8_title
d3e79532 2530
a1433954
KW
2531Convert the UTF-8 encoded character at C<p> to its titlecase version and
2532store that in UTF-8 in C<ustrp> and its length in bytes in C<lenp>. Note
2533that the C<ustrp> needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
89ebb4a3 2534titlecase version may be longer than the original character.
d3e79532
JH
2535
2536The first character of the titlecased version is returned
2537(but note, as explained above, that there may be more.)
2538
a1433954
KW
2539The character at C<p> is assumed by this routine to be well-formed.
2540
d3e79532
JH
2541=cut */
2542
051a06d4
KW
2543/* Not currently externally documented, and subject to change:
2544 * <flags> is set iff locale semantics are to be used for code points < 256
2545 * Since titlecase is not defined in POSIX, uppercase is used instead
2546 * for these/
2547 * <tainted_ptr> if non-null, *tainted_ptr will be set TRUE iff locale rules
2548 * were used in the calculation; otherwise unchanged. */
2549
983ffd37 2550UV
051a06d4 2551Perl__to_utf8_title_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, const bool flags, bool* tainted_ptr)
983ffd37 2552{
97aff369 2553 dVAR;
7918f24d 2554
051a06d4
KW
2555 UV result;
2556
2557 PERL_ARGS_ASSERT__TO_UTF8_TITLE_FLAGS;
7918f24d 2558
3a4c58c9 2559 if (UTF8_IS_INVARIANT(*p)) {
051a06d4
KW
2560 if (flags) {
2561 result = toUPPER_LC(*p);
2562 }
2563 else {
81c6c7ce 2564 return _to_upper_title_latin1(*p, ustrp, lenp, 's');
051a06d4 2565 }
3a4c58c9
KW
2566 }
2567 else if UTF8_IS_DOWNGRADEABLE_START(*p) {
051a06d4
KW
2568 if (flags) {
2569 result = toUPPER_LC(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)));
2570 }
2571 else {
81c6c7ce
KW
2572 return _to_upper_title_latin1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)),
2573 ustrp, lenp, 's');
051a06d4
KW
2574 }
2575 }
2576 else { /* utf8, ord above 255 */
2577 result = CALL_TITLE_CASE(p, ustrp, lenp);
2578
2579 if (flags) {
2580 result = check_locale_boundary_crossing(p, result, ustrp, lenp);
2581 }
2582 return result;
2583 }
2584
2585 /* Here, used locale rules. Convert back to utf8 */
2586 if (UTF8_IS_INVARIANT(result)) {
2587 *ustrp = (U8) result;
2588 *lenp = 1;
2589 }
2590 else {
2591 *ustrp = UTF8_EIGHT_BIT_HI(result);
2592 *(ustrp + 1) = UTF8_EIGHT_BIT_LO(result);
2593 *lenp = 2;
3a4c58c9
KW
2594 }
2595
051a06d4
KW
2596 if (tainted_ptr) {
2597 *tainted_ptr = TRUE;
2598 }
2599 return result;
a0ed51b3
LW
2600}
2601
d3e79532 2602/*
87cea99e 2603=for apidoc to_utf8_lower
d3e79532 2604
a1433954
KW
2605Convert the UTF-8 encoded character at C<p> to its lowercase version and
2606store that in UTF-8 in ustrp and its length in bytes in C<lenp>. Note
2607that the C<ustrp> needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
89ebb4a3 2608lowercase version may be longer than the original character.
d3e79532
JH
2609
2610The first character of the lowercased version is returned
2611(but note, as explained above, that there may be more.)
2612
a1433954
KW
2613The character at C<p> is assumed by this routine to be well-formed.
2614
d3e79532
JH
2615=cut */
2616
051a06d4
KW
2617/* Not currently externally documented, and subject to change:
2618 * <flags> is set iff locale semantics are to be used for code points < 256
2619 * <tainted_ptr> if non-null, *tainted_ptr will be set TRUE iff locale rules
2620 * were used in the calculation; otherwise unchanged. */
2621
2104c8d9 2622UV
051a06d4 2623Perl__to_utf8_lower_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, const bool flags, bool* tainted_ptr)
a0ed51b3 2624{
051a06d4
KW
2625 UV result;
2626
97aff369 2627 dVAR;
7918f24d 2628
051a06d4 2629 PERL_ARGS_ASSERT__TO_UTF8_LOWER_FLAGS;
7918f24d 2630
968c5e6a 2631 if (UTF8_IS_INVARIANT(*p)) {
051a06d4
KW
2632 if (flags) {
2633 result = toLOWER_LC(*p);
2634 }
2635 else {
81c6c7ce 2636 return to_lower_latin1(*p, ustrp, lenp);
051a06d4 2637 }
968c5e6a
KW
2638 }
2639 else if UTF8_IS_DOWNGRADEABLE_START(*p) {
051a06d4
KW
2640 if (flags) {
2641 result = toLOWER_LC(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)));
2642 }
2643 else {
81c6c7ce
KW
2644 return to_lower_latin1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)),
2645 ustrp, lenp);
051a06d4 2646 }
968c5e6a 2647 }
051a06d4
KW
2648 else { /* utf8, ord above 255 */
2649 result = CALL_LOWER_CASE(p, ustrp, lenp);
2650
2651 if (flags) {
2652 result = check_locale_boundary_crossing(p, result, ustrp, lenp);
2653 }
968c5e6a 2654
051a06d4
KW
2655 return result;
2656 }
2657
2658 /* Here, used locale rules. Convert back to utf8 */
2659 if (UTF8_IS_INVARIANT(result)) {
2660 *ustrp = (U8) result;
2661 *lenp = 1;
2662 }
2663 else {
2664 *ustrp = UTF8_EIGHT_BIT_HI(result);
2665 *(ustrp + 1) = UTF8_EIGHT_BIT_LO(result);
2666 *lenp = 2;
2667 }
2668
2669 if (tainted_ptr) {
2670 *tainted_ptr = TRUE;
2671 }
2672 return result;
b4e400f9
JH
2673}
2674
d3e79532 2675/*
87cea99e 2676=for apidoc to_utf8_fold
d3e79532 2677
a1433954
KW
2678Convert the UTF-8 encoded character at C<p> to its foldcase version and
2679store that in UTF-8 in C<ustrp> and its length in bytes in C<lenp>. Note
2680that the C<ustrp> needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
d3e79532
JH
2681foldcase version may be longer than the original character (up to
2682three characters).
2683
2684The first character of the foldcased version is returned
2685(but note, as explained above, that there may be more.)
2686
a1433954
KW
2687The character at C<p> is assumed by this routine to be well-formed.
2688
d3e79532
JH
2689=cut */
2690
051a06d4
KW
2691/* Not currently externally documented, and subject to change,
2692 * in <flags>
2693 * bit FOLD_FLAGS_LOCALE is set iff locale semantics are to be used for code
2694 * points < 256. Since foldcase is not defined in
2695 * POSIX, lowercase is used instead
2696 * bit FOLD_FLAGS_FULL is set iff full case folds are to be used;
2697 * otherwise simple folds
2698 * <tainted_ptr> if non-null, *tainted_ptr will be set TRUE iff locale rules
2699 * were used in the calculation; otherwise unchanged. */
36bb2ab6 2700
b4e400f9 2701UV
051a06d4 2702Perl__to_utf8_fold_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, U8 flags, bool* tainted_ptr)
b4e400f9 2703{
97aff369 2704 dVAR;
7918f24d 2705
051a06d4
KW
2706 UV result;
2707
36bb2ab6 2708 PERL_ARGS_ASSERT__TO_UTF8_FOLD_FLAGS;
7918f24d 2709
50ba90ff
KW
2710 assert(p != ustrp); /* Otherwise overwrites */
2711
a1dde8de 2712 if (UTF8_IS_INVARIANT(*p)) {
051a06d4
KW
2713 if (flags & FOLD_FLAGS_LOCALE) {
2714 result = toLOWER_LC(*p);
2715 }
2716 else {
81c6c7ce
KW
2717 return _to_fold_latin1(*p, ustrp, lenp,
2718 cBOOL(flags & FOLD_FLAGS_FULL));
051a06d4 2719 }
a1dde8de
KW
2720 }
2721 else if UTF8_IS_DOWNGRADEABLE_START(*p) {
051a06d4
KW
2722 if (flags & FOLD_FLAGS_LOCALE) {
2723 result = toLOWER_LC(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)));
2724 }
2725 else {
81c6c7ce 2726 return _to_fold_latin1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)),
051a06d4
KW
2727 ustrp, lenp, cBOOL(flags & FOLD_FLAGS_FULL));
2728 }
a1dde8de 2729 }
051a06d4
KW
2730 else { /* utf8, ord above 255 */
2731 result = CALL_FOLD_CASE(p, ustrp, lenp, flags);
a1dde8de 2732
051a06d4
KW
2733 if ((flags & FOLD_FLAGS_LOCALE)) {
2734 result = check_locale_boundary_crossing(p, result, ustrp, lenp);
2735 }
2736
2737 return result;
2738 }
2739
2740 /* Here, used locale rules. Convert back to utf8 */
2741 if (UTF8_IS_INVARIANT(result)) {
2742 *ustrp = (U8) result;
2743 *lenp = 1;
2744 }
2745 else {
2746 *ustrp = UTF8_EIGHT_BIT_HI(result);
2747 *(ustrp + 1) = UTF8_EIGHT_BIT_LO(result);
2748 *lenp = 2;
2749 }
2750
2751 if (tainted_ptr) {
2752 *tainted_ptr = TRUE;
2753 }
2754 return result;
a0ed51b3
LW
2755}
2756
711a919c 2757/* Note:
f90a9a02 2758 * Returns a "swash" which is a hash described in utf8.c:Perl_swash_fetch().
711a919c
ST
2759 * C<pkg> is a pointer to a package name for SWASHNEW, should be "utf8".
2760 * For other parameters, see utf8::SWASHNEW in lib/utf8_heavy.pl.
2761 */
c4a5db0c 2762
a0ed51b3 2763SV*
7fc63493 2764Perl_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv, I32 minbits, I32 none)
a0ed51b3 2765{
c4a5db0c
KW
2766 PERL_ARGS_ASSERT_SWASH_INIT;
2767
2768 /* Returns a copy of a swash initiated by the called function. This is the
2769 * public interface, and returning a copy prevents others from doing
2770 * mischief on the original */
2771
9a53f6cf 2772 return newSVsv(_core_swash_init(pkg, name, listsv, minbits, none, FALSE, NULL, FALSE));
c4a5db0c
KW
2773}
2774
2775SV*
9a53f6cf 2776Perl__core_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv, I32 minbits, I32 none, bool return_if_undef, SV* invlist, bool passed_in_invlist_has_user_defined_property)
c4a5db0c
KW
2777{
2778 /* Initialize and return a swash, creating it if necessary. It does this
9a53f6cf 2779 * by calling utf8_heavy.pl in the general case.
c4a5db0c
KW
2780 *
2781 * This interface should only be used by functions that won't destroy or
2782 * adversely change the swash, as doing so affects all other uses of the
2783 * swash in the program; the general public should use 'Perl_swash_init'
2784 * instead.
2785 *
2786 * pkg is the name of the package that <name> should be in.
2787 * name is the name of the swash to find. Typically it is a Unicode
2788 * property name, including user-defined ones
2789 * listsv is a string to initialize the swash with. It must be of the form
2790 * documented as the subroutine return value in
2791 * L<perlunicode/User-Defined Character Properties>
2792 * minbits is the number of bits required to represent each data element.
2793 * It is '1' for binary properties.
2794 * none I (khw) do not understand this one, but it is used only in tr///.
9a53f6cf
KW
2795 * return_if_undef is TRUE if the routine shouldn't croak if it can't find
2796 * the requested property
2797 * invlist is an inversion list to initialize the swash with (or NULL)
2798 * has_user_defined_property is TRUE if <invlist> has some component that
2799 * came from a user-defined property
2800 *
2801 * Thus there are three possible inputs to find the swash: <name>,
2802 * <listsv>, and <invlist>. At least one must be specified. The result
2803 * will be the union of the specified ones, although <listsv>'s various
2804 * actions can intersect, etc. what <name> gives.
2805 *
2806 * <invlist> is only valid for binary properties */
c4a5db0c 2807
27da23d5 2808 dVAR;
c4a5db0c 2809 SV* retval = &PL_sv_undef;
9a53f6cf
KW
2810
2811 assert(listsv != &PL_sv_undef || strNE(name, "") || invlist);
2812 assert(! invlist || minbits == 1);
2813
2814 /* If data was passed in to go out to utf8_heavy to find the swash of, do
2815 * so */
2816 if (listsv != &PL_sv_undef || strNE(name, "")) {
69794297
KW
2817 dSP;
2818 const size_t pkg_len = strlen(pkg);
2819 const size_t name_len = strlen(name);
2820 HV * const stash = gv_stashpvn(pkg, pkg_len, 0);
2821 SV* errsv_save;
2822 GV *method;
2823
2824 PERL_ARGS_ASSERT__CORE_SWASH_INIT;
2825
2826 PUSHSTACKi(PERLSI_MAGIC);
ce3b816e 2827 ENTER;
69794297
KW
2828 SAVEHINTS();
2829 save_re_context();
2830 if (PL_parser && PL_parser->error_count)
2831 SAVEI8(PL_parser->error_count), PL_parser->error_count = 0;
2832 method = gv_fetchmeth(stash, "SWASHNEW", 8, -1);
2833 if (!method) { /* demand load utf8 */
2834 ENTER;
2835 errsv_save = newSVsv(ERRSV);
2836 /* It is assumed that callers of this routine are not passing in
2837 * any user derived data. */
2838 /* Need to do this after save_re_context() as it will set
2839 * PL_tainted to 1 while saving $1 etc (see the code after getrx:
2840 * in Perl_magic_get). Even line to create errsv_save can turn on
2841 * PL_tainted. */
2842 SAVEBOOL(PL_tainted);
2843 PL_tainted = 0;
2844 Perl_load_module(aTHX_ PERL_LOADMOD_NOIMPORT, newSVpvn(pkg,pkg_len),
2845 NULL);
2846 if (!SvTRUE(ERRSV))
2847 sv_setsv(ERRSV, errsv_save);
2848 SvREFCNT_dec(errsv_save);
2849 LEAVE;
2850 }
2851 SPAGAIN;
2852 PUSHMARK(SP);
2853 EXTEND(SP,5);
2854 mPUSHp(pkg, pkg_len);
2855 mPUSHp(name, name_len);
2856 PUSHs(listsv);
2857 mPUSHi(minbits);
2858 mPUSHi(none);
2859 PUTBACK;
f8be5cf0 2860 errsv_save = newSVsv(ERRSV);
69794297
KW
2861 /* If we already have a pointer to the method, no need to use
2862 * call_method() to repeat the lookup. */
2863 if (method ? call_sv(MUTABLE_SV(method), G_SCALAR)
2864 : call_sv(newSVpvs_flags("SWASHNEW", SVs_TEMP), G_SCALAR | G_METHOD))
2865 {
2866 retval = *PL_stack_sp--;
2867 SvREFCNT_inc(retval);
2868 }
f8be5cf0
JH
2869 if (!SvTRUE(ERRSV))
2870 sv_setsv(ERRSV, errsv_save);
2871 SvREFCNT_dec(errsv_save);
ce3b816e 2872 LEAVE;
69794297
KW
2873 POPSTACK;
2874 if (IN_PERL_COMPILETIME) {
2875 CopHINTS_set(PL_curcop, PL_hints);
2876 }
2877 if (!SvROK(retval) || SvTYPE(SvRV(retval)) != SVt_PVHV) {
2878 if (SvPOK(retval))
2879
2880 /* If caller wants to handle missing properties, let them */
2881 if (return_if_undef) {
2882 return NULL;
2883 }
2884 Perl_croak(aTHX_
2885 "Can't find Unicode property definition \"%"SVf"\"",
2886 SVfARG(retval));
2887 Perl_croak(aTHX_ "SWASHNEW didn't return an HV ref");
2888 }
9a53f6cf 2889 } /* End of calling the module to find the swash */
36eb48b4
KW
2890
2891 /* Make sure there is an inversion list for binary properties */
2892 if (minbits == 1) {
2893 SV** swash_invlistsvp = NULL;
2894 SV* swash_invlist = NULL;
9a53f6cf 2895 bool invlist_in_swash_is_valid = FALSE;
88d45d28 2896 HV* swash_hv = NULL;
36eb48b4 2897
9a53f6cf
KW
2898 /* If this operation fetched a swash, get its already existing
2899 * inversion list or create one for it */
2900 if (retval != &PL_sv_undef) {
36eb48b4
KW
2901 swash_hv = MUTABLE_HV(SvRV(retval));
2902
2903 swash_invlistsvp = hv_fetchs(swash_hv, "INVLIST", FALSE);
9a53f6cf
KW
2904 if (swash_invlistsvp) {
2905 swash_invlist = *swash_invlistsvp;
2906 invlist_in_swash_is_valid = TRUE;
2907 }
2908 else {
36eb48b4 2909 swash_invlist = _swash_to_invlist(retval);
9a53f6cf
KW
2910 }
2911 }
2912
2913 /* If an inversion list was passed in, have to include it */
2914 if (invlist) {
2915
2916 /* Any fetched swash will by now have an inversion list in it;
2917 * otherwise <swash_invlist> will be NULL, indicating that we
2918 * didn't fetch a swash */
2919 if (swash_invlist) {
2920
2921 /* Add the passed-in inversion list, which invalidates the one
2922 * already stored in the swash */
2923 invlist_in_swash_is_valid = FALSE;
2924 _invlist_union(invlist, swash_invlist, &swash_invlist);
2925 }
2926 else {
2927
2928 /* Here, there is no swash already. Set up a minimal one */
2929 swash_hv = newHV();
2930 retval = newRV_inc(MUTABLE_SV(swash_hv));
2931 swash_invlist = invlist;
2932 }
2933
2934 if (passed_in_invlist_has_user_defined_property) {
2935 if (! hv_stores(swash_hv, "USER_DEFINED", newSVuv(1))) {
2936 Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed");
2937 }
2938 }
2939 }
2940
2941 /* Here, we have computed the union of all the passed-in data. It may
2942 * be that there was an inversion list in the swash which didn't get
2943 * touched; otherwise save the one computed one */
2944 if (! invlist_in_swash_is_valid) {
69794297
KW
2945 if (! hv_stores(MUTABLE_HV(SvRV(retval)), "INVLIST", swash_invlist))
2946 {
2947 Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed");
2948 }
9a53f6cf 2949 }
36eb48b4
KW
2950 }
2951
a0ed51b3
LW
2952 return retval;
2953}
2954
035d37be
JH
2955
2956/* This API is wrong for special case conversions since we may need to
2957 * return several Unicode characters for a single Unicode character
2958 * (see lib/unicore/SpecCase.txt) The SWASHGET in lib/utf8_heavy.pl is
2959 * the lower-level routine, and it is similarly broken for returning
38684baa
KW
2960 * multiple values. --jhi
2961 * For those, you should use to_utf8_case() instead */
b0e3252e 2962/* Now SWASHGET is recasted into S_swatch_get in this file. */
680c470c
ST
2963
2964/* Note:
2965 * Returns the value of property/mapping C<swash> for the first character
2966 * of the string C<ptr>. If C<do_utf8> is true, the string C<ptr> is
2967 * assumed to be in utf8. If C<do_utf8> is false, the string C<ptr> is
2968 * assumed to be in native 8-bit encoding. Caches the swatch in C<swash>.
af2af982
KW
2969 *
2970 * A "swash" is a hash which contains initially the keys/values set up by
2971 * SWASHNEW. The purpose is to be able to completely represent a Unicode
2972 * property for all possible code points. Things are stored in a compact form
2973 * (see utf8_heavy.pl) so that calculation is required to find the actual
2974 * property value for a given code point. As code points are looked up, new
2975 * key/value pairs are added to the hash, so that the calculation doesn't have
2976 * to ever be re-done. Further, each calculation is done, not just for the
2977 * desired one, but for a whole block of code points adjacent to that one.
2978 * For binary properties on ASCII machines, the block is usually for 64 code
2979 * points, starting with a code point evenly divisible by 64. Thus if the
2980 * property value for code point 257 is requested, the code goes out and
2981 * calculates the property values for all 64 code points between 256 and 319,
2982 * and stores these as a single 64-bit long bit vector, called a "swatch",
2983 * under the key for code point 256. The key is the UTF-8 encoding for code
2984 * point 256, minus the final byte. Thus, if the length of the UTF-8 encoding
2985 * for a code point is 13 bytes, the key will be 12 bytes long. If the value
2986 * for code point 258 is then requested, this code realizes that it would be
2987 * stored under the key for 256, and would find that value and extract the
2988 * relevant bit, offset from 256.
2989 *
2990 * Non-binary properties are stored in as many bits as necessary to represent
2991 * their values (32 currently, though the code is more general than that), not
2992 * as single bits, but the principal is the same: the value for each key is a
2993 * vector that encompasses the property values for all code points whose UTF-8
2994 * representations are represented by the key. That is, for all code points
2995 * whose UTF-8 representations are length N bytes, and the key is the first N-1
2996 * bytes of that.
680c470c 2997 */
a0ed51b3 2998UV
680c470c 2999Perl_swash_fetch(pTHX_ SV *swash, const U8 *ptr, bool do_utf8)
a0ed51b3 3000{
27da23d5 3001 dVAR;
ef8f7699 3002 HV *const hv = MUTABLE_HV(SvRV(swash));
3568d838
JH
3003 U32 klen;
3004 U32 off;
a0ed51b3 3005 STRLEN slen;
7d85a32c 3006 STRLEN needents;
cfd0369c 3007 const U8 *tmps = NULL;
a0ed51b3 3008 U32 bit;
979f2922 3009 SV *swatch;
3568d838 3010 U8 tmputf8[2];
35da51f7 3011 const UV c = NATIVE_TO_ASCII(*ptr);
3568d838 3012
7918f24d
NC
3013 PERL_ARGS_ASSERT_SWASH_FETCH;
3014
dbe7a391 3015 /* Convert to utf8 if not already */
3568d838 3016 if (!do_utf8 && !UNI_IS_INVARIANT(c)) {
979f2922
ST
3017 tmputf8[0] = (U8)UTF8_EIGHT_BIT_HI(c);
3018 tmputf8[1] = (U8)UTF8_EIGHT_BIT_LO(c);
3019 ptr = tmputf8;
3568d838
JH
3020 }
3021 /* Given a UTF-X encoded char 0xAA..0xYY,0xZZ
37e2e78e 3022 * then the "swatch" is a vec() for all the chars which start
3568d838
JH
3023 * with 0xAA..0xYY
3024 * So the key in the hash (klen) is length of encoded char -1
3025 */
3026 klen = UTF8SKIP(ptr) - 1;
3027 off = ptr[klen];
a0ed51b3 3028
979f2922 3029 if (klen == 0) {
37e2e78e 3030 /* If char is invariant then swatch is for all the invariant chars
1e54db1a 3031 * In both UTF-8 and UTF-8-MOD that happens to be UTF_CONTINUATION_MARK
7d85a32c 3032 */
979f2922
ST
3033 needents = UTF_CONTINUATION_MARK;
3034 off = NATIVE_TO_UTF(ptr[klen]);
3035 }
3036 else {
7d85a32c 3037 /* If char is encoded then swatch is for the prefix */
979f2922
ST
3038 needents = (1 << UTF_ACCUMULATION_SHIFT);
3039 off = NATIVE_TO_UTF(ptr[klen]) & UTF_CONTINUATION_MASK;
3040 }
7d85a32c 3041
a0ed51b3
LW
3042 /*
3043 * This single-entry cache saves about 1/3 of the utf8 overhead in test
3044 * suite. (That is, only 7-8% overall over just a hash cache. Still,
3045 * it's nothing to sniff at.) Pity we usually come through at least
3046 * two function calls to get here...
3047 *
3048 * NB: this code assumes that swatches are never modified, once generated!
3049 */
3050
3568d838 3051 if (hv == PL_last_swash_hv &&
a0ed51b3 3052 klen == PL_last_swash_klen &&
27da23d5 3053 (!klen || memEQ((char *)ptr, (char *)PL_last_swash_key, klen)) )
a0ed51b3
LW
3054 {
3055 tmps = PL_last_swash_tmps;
3056 slen = PL_last_swash_slen;
3057 }
3058 else {
3059 /* Try our second-level swatch cache, kept in a hash. */
e1ec3a88 3060 SV** svp = hv_fetch(hv, (const char*)ptr, klen, FALSE);
a0ed51b3 3061
b0e3252e 3062 /* If not cached, generate it via swatch_get */
979f2922
ST
3063 if (!svp || !SvPOK(*svp)
3064 || !(tmps = (const U8*)SvPV_const(*svp, slen))) {
2b9d42f0
NIS
3065 /* We use utf8n_to_uvuni() as we want an index into
3066 Unicode tables, not a native character number.
3067 */
aec46f14 3068 const UV code_point = utf8n_to_uvuni(ptr, UTF8_MAXBYTES, 0,
872c91ae
JH
3069 ckWARN(WARN_UTF8) ?
3070 0 : UTF8_ALLOW_ANY);
b0e3252e 3071 swatch = swatch_get(swash,
979f2922 3072 /* On EBCDIC & ~(0xA0-1) isn't a useful thing to do */
361ee0fe 3073 (klen) ? (code_point & ~((UV)needents - 1)) : 0,
979f2922
ST
3074 needents);
3075
923e4eb5 3076 if (IN_PERL_COMPILETIME)
623e6609 3077 CopHINTS_set(PL_curcop, PL_hints);
a0ed51b3 3078
979f2922 3079 svp = hv_store(hv, (const char *)ptr, klen, swatch, 0);
a0ed51b3 3080
979f2922
ST
3081 if (!svp || !(tmps = (U8*)SvPV(*svp, slen))
3082 || (slen << 3) < needents)
5637ef5b
NC
3083 Perl_croak(aTHX_ "panic: swash_fetch got improper swatch, "
3084 "svp=%p, tmps=%p, slen=%"UVuf", needents=%"UVuf,
3085 svp, tmps, (UV)slen, (UV)needents);
a0ed51b3
LW
3086 }
3087
3088 PL_last_swash_hv = hv;
16d8f38a 3089 assert(klen <= sizeof(PL_last_swash_key));
eac04b2e 3090 PL_last_swash_klen = (U8)klen;
cfd0369c
NC
3091 /* FIXME change interpvar.h? */
3092 PL_last_swash_tmps = (U8 *) tmps;
a0ed51b3
LW
3093 PL_last_swash_slen = slen;
3094 if (klen)
3095 Copy(ptr, PL_last_swash_key, klen, U8);
3096 }
3097
a410ec23
KW
3098 if (UTF8_IS_SUPER(ptr) && ckWARN_d(WARN_NON_UNICODE)) {
3099 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
3100
3101 /* This outputs warnings for binary properties only, assuming that
3102 * to_utf8_case() will output any for non-binary. Also, surrogates
3103 * aren't checked for, as that would warn on things like /\p{Gc=Cs}/ */
3104
9a53f6cf 3105 if (! bitssvp || SvUV(*bitssvp) == 1) {
a410ec23
KW
3106 /* User-defined properties can silently match above-Unicode */
3107 SV** const user_defined_svp = hv_fetchs(hv, "USER_DEFINED", FALSE);
3108 if (! user_defined_svp || ! SvUV(*user_defined_svp)) {
3109 const UV code_point = utf8n_to_uvuni(ptr, UTF8_MAXBYTES, 0, 0);
3110 Perl_warner(aTHX_ packWARN(WARN_NON_UNICODE),
3111 "Code point 0x%04"UVXf" is not Unicode, all \\p{} matches fail; all \\P{} matches succeed", code_point);
3112 }
3113 }
3114 }
3115
9faf8d75 3116 switch ((int)((slen << 3) / needents)) {
a0ed51b3
LW
3117 case 1:
3118 bit = 1 << (off & 7);
3119 off >>= 3;
3120 return (tmps[off] & bit) != 0;
3121 case 8:
3122 return tmps[off];
3123 case 16:
3124 off <<= 1;
3125 return (tmps[off] << 8) + tmps[off + 1] ;
3126 case 32:
3127 off <<= 2;
3128 return (tmps[off] << 24) + (tmps[off+1] << 16) + (tmps[off+2] << 8) + tmps[off + 3] ;
3129 }
5637ef5b
NC
3130 Perl_croak(aTHX_ "panic: swash_fetch got swatch of unexpected bit width, "
3131 "slen=%"UVuf", needents=%"UVuf, (UV)slen, (UV)needents);
670f1322 3132 NORETURN_FUNCTION_END;
a0ed51b3 3133}
2b9d42f0 3134
319009ee
KW
3135/* Read a single line of the main body of the swash input text. These are of
3136 * the form:
3137 * 0053 0056 0073
3138 * where each number is hex. The first two numbers form the minimum and
3139 * maximum of a range, and the third is the value associated with the range.
3140 * Not all swashes should have a third number
3141 *
3142 * On input: l points to the beginning of the line to be examined; it points
3143 * to somewhere in the string of the whole input text, and is
3144 * terminated by a \n or the null string terminator.
3145 * lend points to the null terminator of that string
3146 * wants_value is non-zero if the swash expects a third number
3147 * typestr is the name of the swash's mapping, like 'ToLower'
3148 * On output: *min, *max, and *val are set to the values read from the line.
3149 * returns a pointer just beyond the line examined. If there was no
3150 * valid min number on the line, returns lend+1
3151 */
3152
3153STATIC U8*
3154S_swash_scan_list_line(pTHX_ U8* l, U8* const lend, UV* min, UV* max, UV* val,
3155 const bool wants_value, const U8* const typestr)
3156{
3157 const int typeto = typestr[0] == 'T' && typestr[1] == 'o';
3158 STRLEN numlen; /* Length of the number */
02470786
KW
3159 I32 flags = PERL_SCAN_SILENT_ILLDIGIT
3160 | PERL_SCAN_DISALLOW_PREFIX
3161 | PERL_SCAN_SILENT_NON_PORTABLE;
319009ee
KW
3162
3163 /* nl points to the next \n in the scan */
3164 U8* const nl = (U8*)memchr(l, '\n', lend - l);
3165
3166 /* Get the first number on the line: the range minimum */
3167 numlen = lend - l;
3168 *min = grok_hex((char *)l, &numlen, &flags, NULL);
3169 if (numlen) /* If found a hex number, position past it */
3170 l += numlen;
3171 else if (nl) { /* Else, go handle next line, if any */
3172 return nl + 1; /* 1 is length of "\n" */
3173 }
3174 else { /* Else, no next line */
3175 return lend + 1; /* to LIST's end at which \n is not found */
3176 }
3177
3178 /* The max range value follows, separated by a BLANK */
3179 if (isBLANK(*l)) {
3180 ++l;
02470786
KW
3181 flags = PERL_SCAN_SILENT_ILLDIGIT
3182 | PERL_SCAN_DISALLOW_PREFIX
3183 | PERL_SCAN_SILENT_NON_PORTABLE;
319009ee
KW
3184 numlen = lend - l;
3185 *max = grok_hex((char *)l, &numlen, &flags, NULL);
3186 if (numlen)
3187 l += numlen;
3188 else /* If no value here, it is a single element range */
3189 *max = *min;
3190
3191 /* Non-binary tables have a third entry: what the first element of the
3192 * range maps to */
3193 if (wants_value) {
3194 if (isBLANK(*l)) {
3195 ++l;
f90a9a02
KW
3196
3197 /* The ToLc, etc table mappings are not in hex, and must be
3198 * corrected by adding the code point to them */
3199 if (typeto) {
3200 char *after_strtol = (char *) lend;
3201 *val = Strtol((char *)l, &after_strtol, 10);
3202 l = (U8 *) after_strtol;
f90a9a02
KW
3203 }
3204 else { /* Other tables are in hex, and are the correct result
3205 without tweaking */
a9d188b3
KW
3206 flags = PERL_SCAN_SILENT_ILLDIGIT
3207 | PERL_SCAN_DISALLOW_PREFIX
3208 | PERL_SCAN_SILENT_NON_PORTABLE;
3209 numlen = lend - l;
3210 *val = grok_hex((char *)l, &numlen, &flags, NULL);
3211 if (numlen)
3212 l += numlen;
3213 else
3214 *val = 0;
f90a9a02 3215 }
319009ee
KW
3216 }
3217 else {
3218 *val = 0;
3219 if (typeto) {
dcbac5bb 3220 /* diag_listed_as: To%s: illegal mapping '%s' */
319009ee
KW
3221 Perl_croak(aTHX_ "%s: illegal mapping '%s'",
3222 typestr, l);
3223 }
3224 }
3225 }
3226 else
3227 *val = 0; /* bits == 1, then any val should be ignored */
3228 }
3229 else { /* Nothing following range min, should be single element with no
3230 mapping expected */
3231 *max = *min;
3232 if (wants_value) {
3233 *val = 0;
3234 if (typeto) {
dcbac5bb 3235 /* diag_listed_as: To%s: illegal mapping '%s' */
319009ee
KW
3236 Perl_croak(aTHX_ "%s: illegal mapping '%s'", typestr, l);
3237 }
3238 }
3239 else
3240 *val = 0; /* bits == 1, then val should be ignored */
3241 }
3242
3243 /* Position to next line if any, or EOF */
3244 if (nl)
3245 l = nl + 1;
3246 else
3247 l = lend;
3248
3249 return l;
3250}
3251
979f2922
ST
3252/* Note:
3253 * Returns a swatch (a bit vector string) for a code point sequence
3254 * that starts from the value C<start> and comprises the number C<span>.
3255 * A C<swash> must be an object created by SWASHNEW (see lib/utf8_heavy.pl).
3256 * Should be used via swash_fetch, which will cache the swatch in C<swash>.
3257 */
3258STATIC SV*
b0e3252e 3259S_swatch_get(pTHX_ SV* swash, UV start, UV span)
979f2922
ST
3260{
3261 SV *swatch;
77f9f126 3262 U8 *l, *lend, *x, *xend, *s, *send;
979f2922 3263 STRLEN lcur, xcur, scur;
ef8f7699 3264 HV *const hv = MUTABLE_HV(SvRV(swash));
36eb48b4
KW
3265 SV** const invlistsvp = hv_fetchs(hv, "INVLIST", FALSE);
3266
88d45d28
KW
3267 SV** listsvp = NULL; /* The string containing the main body of the table */
3268 SV** extssvp = NULL;
3269 SV** invert_it_svp = NULL;
3270 U8* typestr = NULL;
786861f5
KW
3271 STRLEN bits;
3272 STRLEN octets; /* if bits == 1, then octets == 0 */
3273 UV none;
3274 UV end = start + span;
972dd592 3275
36eb48b4 3276 if (invlistsvp == NULL) {
786861f5
KW
3277 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
3278 SV** const nonesvp = hv_fetchs(hv, "NONE", FALSE);
3279 SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE);
3280 extssvp = hv_fetchs(hv, "EXTRAS", FALSE);
3281 listsvp = hv_fetchs(hv, "LIST", FALSE);
3282 invert_it_svp = hv_fetchs(hv, "INVERT_IT", FALSE);
3283
3284 bits = SvUV(*bitssvp);
3285 none = SvUV(*nonesvp);
3286 typestr = (U8*)SvPV_nolen(*typesvp);
3287 }
36eb48b4
KW
3288 else {
3289 bits = 1;
3290 none = 0;
3291 }
786861f5 3292 octets = bits >> 3; /* if bits == 1, then octets == 0 */
979f2922 3293
b0e3252e 3294 PERL_ARGS_ASSERT_SWATCH_GET;
7918f24d 3295
979f2922 3296 if (bits != 1 && bits != 8 && bits != 16 && bits != 32) {
b0e3252e 3297 Perl_croak(aTHX_ "panic: swatch_get doesn't expect bits %"UVuf,
660a4616 3298 (UV)bits);
979f2922
ST
3299 }
3300
84ea5ef6
KW
3301 /* If overflowed, use the max possible */
3302 if (end < start) {
3303 end = UV_MAX;
3304 span = end - start;
3305 }
3306
979f2922 3307 /* create and initialize $swatch */
979f2922 3308 scur = octets ? (span * octets) : (span + 7) / 8;
e524fe40
NC
3309 swatch = newSV(scur);
3310 SvPOK_on(swatch);
979f2922
ST
3311 s = (U8*)SvPVX(swatch);
3312 if (octets && none) {
0bd48802 3313 const U8* const e = s + scur;
979f2922
ST
3314 while (s < e) {
3315 if (bits == 8)
3316 *s++ = (U8)(none & 0xff);
3317 else if (bits == 16) {
3318 *s++ = (U8)((none >> 8) & 0xff);
3319 *s++ = (U8)( none & 0xff);
3320 }
3321 else if (bits == 32) {
3322 *s++ = (U8)((none >> 24) & 0xff);
3323 *s++ = (U8)((none >> 16) & 0xff);
3324 *s++ = (U8)((none >> 8) & 0xff);
3325 *s++ = (U8)( none & 0xff);
3326 }
3327 }
3328 *s = '\0';
3329 }
3330 else {
3331 (void)memzero((U8*)s, scur + 1);
3332 }
3333 SvCUR_set(swatch, scur);
3334 s = (U8*)SvPVX(swatch);
3335
36eb48b4
KW
3336 if (invlistsvp) { /* If has an inversion list set up use that */
3337 _invlist_populate_swatch(*invlistsvp, start, end, s);
3338 return swatch;
3339 }
3340
3341 /* read $swash->{LIST} */
979f2922
ST
3342 l = (U8*)SvPV(*listsvp, lcur);
3343 lend = l + lcur;
3344 while (l < lend) {
8ed25d53 3345 UV min, max, val, upper;
319009ee
KW
3346 l = S_swash_scan_list_line(aTHX_ l, lend, &min, &max, &val,
3347 cBOOL(octets), typestr);
3348 if (l > lend) {
979f2922
ST
3349 break;
3350 }
3351
972dd592 3352 /* If looking for something beyond this range, go try the next one */
979f2922
ST
3353 if (max < start)
3354 continue;
3355
8ed25d53
KW
3356 /* <end> is generally 1 beyond where we want to set things, but at the
3357 * platform's infinity, where we can't go any higher, we want to
3358 * include the code point at <end> */
3359 upper = (max < end)
3360 ? max
3361 : (max != UV_MAX || end != UV_MAX)
3362 ? end - 1
3363 : end;
3364
979f2922 3365 if (octets) {
35da51f7 3366 UV key;
979f2922
ST
3367 if (min < start) {
3368 if (!none || val < none) {
3369 val += start - min;
3370 }
3371 min = start;
3372 }
8ed25d53 3373 for (key = min; key <= upper; key++) {
979f2922 3374 STRLEN offset;
979f2922
ST
3375 /* offset must be non-negative (start <= min <= key < end) */
3376 offset = octets * (key - start);
3377 if (bits == 8)
3378 s[offset] = (U8)(val & 0xff);
3379 else if (bits == 16) {
3380 s[offset ] = (U8)((val >> 8) & 0xff);
3381 s[offset + 1] = (U8)( val & 0xff);
3382 }
3383 else if (bits == 32) {
3384 s[offset ] = (U8)((val >> 24) & 0xff);
3385 s[offset + 1] = (U8)((val >> 16) & 0xff);
3386 s[offset + 2] = (U8)((val >> 8) & 0xff);
3387 s[offset + 3] = (U8)( val & 0xff);
3388 }
3389
3390 if (!none || val < none)
3391 ++val;
3392 }
3393 }
711a919c 3394 else { /* bits == 1, then val should be ignored */
35da51f7 3395 UV key;
979f2922
ST
3396 if (min < start)
3397 min = start;
6cb05c12 3398
8ed25d53 3399 for (key = min; key <= upper; key++) {
0bd48802 3400 const STRLEN offset = (STRLEN)(key - start);
979f2922
ST
3401 s[offset >> 3] |= 1 << (offset & 7);
3402 }
3403 }
3404 } /* while */
979f2922 3405
9479a769 3406 /* Invert if the data says it should be. Assumes that bits == 1 */
77f9f126 3407 if (invert_it_svp && SvUV(*invert_it_svp)) {
0bda3001
KW
3408
3409 /* Unicode properties should come with all bits above PERL_UNICODE_MAX
3410 * be 0, and their inversion should also be 0, as we don't succeed any
3411 * Unicode property matches for non-Unicode code points */
3412 if (start <= PERL_UNICODE_MAX) {
3413
3414 /* The code below assumes that we never cross the
3415 * Unicode/above-Unicode boundary in a range, as otherwise we would
3416 * have to figure out where to stop flipping the bits. Since this
3417 * boundary is divisible by a large power of 2, and swatches comes
3418 * in small powers of 2, this should be a valid assumption */
3419 assert(start + span - 1 <= PERL_UNICODE_MAX);
3420
507a8485
KW
3421 send = s + scur;
3422 while (s < send) {
3423 *s = ~(*s);
3424 s++;
3425 }
0bda3001 3426 }
77f9f126
KW
3427 }
3428
d73c39c5
KW
3429 /* read $swash->{EXTRAS}
3430 * This code also copied to swash_to_invlist() below */
979f2922
ST
3431 x = (U8*)SvPV(*extssvp, xcur);
3432 xend = x + xcur;
3433 while (x < xend) {
3434 STRLEN namelen;
3435 U8 *namestr;
3436 SV** othersvp;
3437 HV* otherhv;
3438 STRLEN otherbits;
3439 SV **otherbitssvp, *other;
711a919c 3440 U8 *s, *o, *nl;
979f2922
ST
3441 STRLEN slen, olen;
3442
35da51f7 3443 const U8 opc = *x++;
979f2922
ST
3444 if (opc == '\n')
3445 continue;
3446
3447 nl = (U8*)memchr(x, '\n', xend - x);
3448
3449 if (opc != '-' && opc != '+' && opc != '!' && opc != '&') {
3450 if (nl) {
3451 x = nl + 1; /* 1 is length of "\n" */
3452 continue;
3453 }
3454 else {
3455 x = xend; /* to EXTRAS' end at which \n is not found */
3456 break;
3457 }
3458 }
3459
3460 namestr = x;
3461 if (nl) {
3462 namelen = nl - namestr;
3463 x = nl + 1;
3464 }
3465 else {
3466 namelen = xend - namestr;
3467 x = xend;
3468 }
3469
3470 othersvp = hv_fetch(hv, (char *)namestr, namelen, FALSE);
ef8f7699 3471 otherhv = MUTABLE_HV(SvRV(*othersvp));
017a3ce5 3472 otherbitssvp = hv_fetchs(otherhv, "BITS", FALSE);
979f2922
ST
3473 otherbits = (STRLEN)SvUV(*otherbitssvp);
3474 if (bits < otherbits)
5637ef5b
NC
3475 Perl_croak(aTHX_ "panic: swatch_get found swatch size mismatch, "
3476 "bits=%"UVuf", otherbits=%"UVuf, (UV)bits, (UV)otherbits);
979f2922
ST
3477
3478 /* The "other" swatch must be destroyed after. */
b0e3252e 3479 other = swatch_get(*othersvp, start, span);
979f2922
ST
3480 o = (U8*)SvPV(other, olen);
3481
3482 if (!olen)
b0e3252e 3483 Perl_croak(aTHX_ "panic: swatch_get got improper swatch");
979f2922
ST
3484
3485 s = (U8*)SvPV(swatch, slen);
3486 if (bits == 1 && otherbits == 1) {
3487 if (slen != olen)
5637ef5b
NC
3488 Perl_croak(aTHX_ "panic: swatch_get found swatch length "
3489 "mismatch, slen=%"UVuf", olen=%"UVuf,
3490 (UV)slen, (UV)olen);
979f2922
ST
3491
3492 switch (opc) {
3493 case '+':
3494 while (slen--)
3495 *s++ |= *o++;
3496 break;
3497 case '!':
3498 while (slen--)
3499 *s++ |= ~*o++;
3500 break;
3501 case '-':
3502 while (slen--)
3503 *s++ &= ~*o++;
3504 break;
3505 case '&':
3506 while (slen--)
3507 *s++ &= *o++;
3508 break;
3509 default:
3510 break;
3511 }
3512 }
711a919c 3513 else {
979f2922
ST
3514 STRLEN otheroctets = otherbits >> 3;
3515 STRLEN offset = 0;
35da51f7 3516 U8* const send = s + slen;
979f2922
ST
3517
3518 while (s < send) {
3519 UV otherval = 0;
3520
3521 if (otherbits == 1) {
3522 otherval = (o[offset >> 3] >> (offset & 7)) & 1;
3523 ++offset;
3524 }
3525 else {
3526 STRLEN vlen = otheroctets;
3527 otherval = *o++;
3528 while (--vlen) {
3529 otherval <<= 8;
3530 otherval |= *o++;
3531 }
3532 }
3533
711a919c 3534 if (opc == '+' && otherval)
6f207bd3 3535 NOOP; /* replace with otherval */
979f2922
ST
3536 else if (opc == '!' && !otherval)
3537 otherval = 1;
3538 else if (opc == '-' && otherval)
3539 otherval = 0;
3540 else if (opc == '&' && !otherval)
3541 otherval = 0;
3542 else {
711a919c 3543 s += octets; /* no replacement */
979f2922
ST
3544 continue;
3545 }
3546
3547 if (bits == 8)
3548 *s++ = (U8)( otherval & 0xff);
3549 else if (bits == 16) {
3550 *s++ = (U8)((otherval >> 8) & 0xff);
3551 *s++ = (U8)( otherval & 0xff);
3552 }
3553 else if (bits == 32) {
3554 *s++ = (U8)((otherval >> 24) & 0xff);
3555 *s++ = (U8)((otherval >> 16) & 0xff);
3556 *s++ = (U8)((otherval >> 8) & 0xff);
3557 *s++ = (U8)( otherval & 0xff);
3558 }
3559 }
3560 }
3561 sv_free(other); /* through with it! */
3562 } /* while */
3563 return swatch;
3564}
3565
064c021d 3566HV*
4c2e1131 3567Perl__swash_inversion_hash(pTHX_ SV* const swash)
064c021d
KW
3568{
3569
5662e334
KW
3570 /* Subject to change or removal. For use only in one place in regcomp.c.
3571 * Can't be used on a property that is subject to user override, as it
3572 * relies on the value of SPECIALS in the swash which would be set by
3573 * utf8_heavy.pl to the hash in the non-overriden file, and hence is not set
3574 * for overridden properties
064c021d
KW
3575 *
3576 * Returns a hash which is the inversion and closure of a swash mapping.
3577 * For example, consider the input lines:
3578 * 004B 006B
3579 * 004C 006C
3580 * 212A 006B
3581 *
3582 * The returned hash would have two keys, the utf8 for 006B and the utf8 for
3583 * 006C. The value for each key is an array. For 006C, the array would
3584 * have a two elements, the utf8 for itself, and for 004C. For 006B, there
3585 * would be three elements in its array, the utf8 for 006B, 004B and 212A.
3586 *
3587 * Essentially, for any code point, it gives all the code points that map to
3588 * it, or the list of 'froms' for that point.
3589 *
5662e334
KW
3590 * Currently it ignores any additions or deletions from other swashes,
3591 * looking at just the main body of the swash, and if there are SPECIALS
3592 * in the swash, at that hash
3593 *
3594 * The specials hash can be extra code points, and most likely consists of
3595 * maps from single code points to multiple ones (each expressed as a string
3596 * of utf8 characters). This function currently returns only 1-1 mappings.
3597 * However consider this possible input in the specials hash:
3598 * "\xEF\xAC\x85" => "\x{0073}\x{0074}", # U+FB05 => 0073 0074
3599 * "\xEF\xAC\x86" => "\x{0073}\x{0074}", # U+FB06 => 0073 0074
3600 *
3601 * Both FB05 and FB06 map to the same multi-char sequence, which we don't
3602 * currently handle. But it also means that FB05 and FB06 are equivalent in
3603 * a 1-1 mapping which we should handle, and this relationship may not be in
3604 * the main table. Therefore this function examines all the multi-char
3605 * sequences and adds the 1-1 mappings that come out of that. */
064c021d
KW
3606
3607 U8 *l, *lend;
3608 STRLEN lcur;
3609 HV *const hv = MUTABLE_HV(SvRV(swash));
3610
3611 /* The string containing the main body of the table */
3612 SV** const listsvp = hv_fetchs(hv, "LIST", FALSE);
3613
3614 SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE);
3615 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
3616 SV** const nonesvp = hv_fetchs(hv, "NONE", FALSE);
3617 /*SV** const extssvp = hv_fetchs(hv, "EXTRAS", FALSE);*/
3618 const U8* const typestr = (U8*)SvPV_nolen(*typesvp);
3619 const STRLEN bits = SvUV(*bitssvp);
3620 const STRLEN octets = bits >> 3; /* if bits == 1, then octets == 0 */
3621 const UV none = SvUV(*nonesvp);
5662e334 3622 SV **specials_p = hv_fetchs(hv, "SPECIALS", 0);
064c021d
KW
3623
3624 HV* ret = newHV();
3625
3626 PERL_ARGS_ASSERT__SWASH_INVERSION_HASH;
3627
3628 /* Must have at least 8 bits to get the mappings */
3629 if (bits != 8 && bits != 16 && bits != 32) {
3630 Perl_croak(aTHX_ "panic: swash_inversion_hash doesn't expect bits %"UVuf,
3631 (UV)bits);
3632 }
3633
5662e334
KW
3634 if (specials_p) { /* It might be "special" (sometimes, but not always, a
3635 mapping to more than one character */
3636
3637 /* Construct an inverse mapping hash for the specials */
3638 HV * const specials_hv = MUTABLE_HV(SvRV(*specials_p));
3639 HV * specials_inverse = newHV();
3640 char *char_from; /* the lhs of the map */
3641 I32 from_len; /* its byte length */
3642 char *char_to; /* the rhs of the map */
3643 I32 to_len; /* its byte length */
3644 SV *sv_to; /* and in a sv */
3645 AV* from_list; /* list of things that map to each 'to' */
3646
3647 hv_iterinit(specials_hv);
3648
3649 /* The keys are the characters (in utf8) that map to the corresponding
3650 * utf8 string value. Iterate through the list creating the inverse
3651 * list. */
3652 while ((sv_to = hv_iternextsv(specials_hv, &char_from, &from_len))) {
3653 SV** listp;
3654 if (! SvPOK(sv_to)) {
5637ef5b
NC
3655 Perl_croak(aTHX_ "panic: value returned from hv_iternextsv() "
3656 "unexpectedly is not a string, flags=%lu",
3657 (unsigned long)SvFLAGS(sv_to));
5662e334 3658 }
4b88fb76 3659 /*DEBUG_U(PerlIO_printf(Perl_debug_log, "Found mapping from %"UVXf", First char of to is %"UVXf"\n", valid_utf8_to_uvchr((U8*) char_from, 0), valid_utf8_to_uvchr((U8*) SvPVX(sv_to), 0)));*/
5662e334
KW
3660
3661 /* Each key in the inverse list is a mapped-to value, and the key's
3662 * hash value is a list of the strings (each in utf8) that map to
3663 * it. Those strings are all one character long */
3664 if ((listp = hv_fetch(specials_inverse,
3665 SvPVX(sv_to),
3666 SvCUR(sv_to), 0)))
3667 {
3668 from_list = (AV*) *listp;
3669 }
3670 else { /* No entry yet for it: create one */
3671 from_list = newAV();
3672 if (! hv_store(specials_inverse,
3673 SvPVX(sv_to),
3674 SvCUR(sv_to),
3675 (SV*) from_list, 0))
3676 {
3677 Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed");
3678 }
3679 }
3680
3681 /* Here have the list associated with this 'to' (perhaps newly
3682 * created and empty). Just add to it. Note that we ASSUME that
3683 * the input is guaranteed to not have duplications, so we don't
3684 * check for that. Duplications just slow down execution time. */
3685 av_push(from_list, newSVpvn_utf8(char_from, from_len, TRUE));
3686 }
3687
3688 /* Here, 'specials_inverse' contains the inverse mapping. Go through
3689 * it looking for cases like the FB05/FB06 examples above. There would
3690 * be an entry in the hash like
3691 * 'st' => [ FB05, FB06 ]
3692 * In this example we will create two lists that get stored in the
3693 * returned hash, 'ret':
3694 * FB05 => [ FB05, FB06 ]
3695 * FB06 => [ FB05, FB06 ]
3696 *
3697 * Note that there is nothing to do if the array only has one element.
3698 * (In the normal 1-1 case handled below, we don't have to worry about
3699 * two lists, as everything gets tied to the single list that is
3700 * generated for the single character 'to'. But here, we are omitting
3701 * that list, ('st' in the example), so must have multiple lists.) */
3702 while ((from_list = (AV *) hv_iternextsv(specials_inverse,
3703 &char_to, &to_len)))
3704 {
3705 if (av_len(from_list) > 0) {
3706 int i;
3707
3708 /* We iterate over all combinations of i,j to place each code
3709 * point on each list */
3710 for (i = 0; i <= av_len(from_list); i++) {
3711 int j;
3712 AV* i_list = newAV();
3713 SV** entryp = av_fetch(from_list, i, FALSE);
3714 if (entryp == NULL) {
3715 Perl_croak(aTHX_ "panic: av_fetch() unexpectedly failed");
3716 }
3717 if (hv_fetch(ret, SvPVX(*entryp), SvCUR(*entryp), FALSE)) {
3718 Perl_croak(aTHX_ "panic: unexpected entry for %s", SvPVX(*entryp));
3719 }
3720 if (! hv_store(ret, SvPVX(*entryp), SvCUR(*entryp),
3721 (SV*) i_list, FALSE))
3722 {
3723 Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed");
3724 }
3725
4b88fb76 3726 /* For debugging: UV u = valid_utf8_to_uvchr((U8*) SvPVX(*entryp), 0);*/
5662e334
KW
3727 for (j = 0; j <= av_len(from_list); j++) {
3728 entryp = av_fetch(from_list, j, FALSE);
3729 if (entryp == NULL) {
3730 Perl_croak(aTHX_ "panic: av_fetch() unexpectedly failed");
3731 }
3732
3733 /* When i==j this adds itself to the list */
4b88fb76
KW
3734 av_push(i_list, newSVuv(utf8_to_uvchr_buf(
3735 (U8*) SvPVX(*entryp),
3736 (U8*) SvPVX(*entryp) + SvCUR(*entryp),
3737 0)));
3738 /*DEBUG_U(PerlIO_printf(Perl_debug_log, "Adding %"UVXf" to list for %"UVXf"\n", valid_utf8_to_uvchr((U8*) SvPVX(*entryp), 0), u));*/
5662e334
KW
3739 }
3740 }
3741 }
3742 }
3743 SvREFCNT_dec(specials_inverse); /* done with it */
3744 } /* End of specials */
3745
064c021d
KW
3746 /* read $swash->{LIST} */
3747 l = (U8*)SvPV(*listsvp, lcur);
3748 lend = l + lcur;
3749
3750 /* Go through each input line */
3751 while (l < lend) {
3752 UV min, max, val;
3753 UV inverse;
3754 l = S_swash_scan_list_line(aTHX_ l, lend, &min, &max, &val,
3755 cBOOL(octets), typestr);
3756 if (l > lend) {
3757 break;
3758 }
3759
3760 /* Each element in the range is to be inverted */
3761 for (inverse = min; inverse <= max; inverse++) {
3762 AV* list;
064c021d
KW
3763 SV** listp;
3764 IV i;
3765 bool found_key = FALSE;
5662e334 3766 bool found_inverse = FALSE;
064c021d
KW
3767
3768 /* The key is the inverse mapping */
3769 char key[UTF8_MAXBYTES+1];
3770 char* key_end = (char *) uvuni_to_utf8((U8*) key, val);
3771 STRLEN key_len = key_end - key;
3772
064c021d
KW
3773 /* Get the list for the map */
3774 if ((listp = hv_fetch(ret, key, key_len, FALSE))) {
3775 list = (AV*) *listp;
3776 }
3777 else { /* No entry yet for it: create one */
3778 list = newAV();
3779 if (! hv_store(ret, key, key_len, (SV*) list, FALSE)) {
3780 Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed");
3781 }
3782 }
3783
5662e334
KW
3784 /* Look through list to see if this inverse mapping already is
3785