This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
utf8.h: Add parameter checking to some macros in DEBUGGING builds
[perl5.git] / utf8.c
CommitLineData
a0ed51b3
LW
1/* utf8.c
2 *
1129b882 3 * Copyright (C) 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008
b94e2f88 4 * by Larry Wall and others
a0ed51b3
LW
5 *
6 * You may distribute under the terms of either the GNU General Public
7 * License or the Artistic License, as specified in the README file.
8 *
9 */
10
11/*
4ac71550
TC
12 * 'What a fix!' said Sam. 'That's the one place in all the lands we've ever
13 * heard of that we don't want to see any closer; and that's the one place
14 * we're trying to get to! And that's just where we can't get, nohow.'
15 *
cdad3b53 16 * [p.603 of _The Lord of the Rings_, IV/I: "The Taming of Sméagol"]
a0ed51b3
LW
17 *
18 * 'Well do I understand your speech,' he answered in the same language;
19 * 'yet few strangers do so. Why then do you not speak in the Common Tongue,
4ac71550 20 * as is the custom in the West, if you wish to be answered?'
cdad3b53 21 * --Gandalf, addressing Théoden's door wardens
4ac71550
TC
22 *
23 * [p.508 of _The Lord of the Rings_, III/vi: "The King of the Golden Hall"]
a0ed51b3
LW
24 *
25 * ...the travellers perceived that the floor was paved with stones of many
26 * hues; branching runes and strange devices intertwined beneath their feet.
4ac71550
TC
27 *
28 * [p.512 of _The Lord of the Rings_, III/vi: "The King of the Golden Hall"]
a0ed51b3
LW
29 */
30
31#include "EXTERN.h"
864dbfa3 32#define PERL_IN_UTF8_C
a0ed51b3 33#include "perl.h"
81e983c1 34#include "inline_invlist.c"
a0ed51b3 35
27da23d5
JH
36static const char unees[] =
37 "Malformed UTF-8 character (unexpected end of string)";
901b21bf 38
48ef279e 39/*
ccfc67b7 40=head1 Unicode Support
a0ed51b3 41
166f8a29
DM
42This file contains various utility functions for manipulating UTF8-encoded
43strings. For the uninitiated, this is a method of representing arbitrary
61296642 44Unicode characters as a variable number of bytes, in such a way that
56da48f7
DM
45characters in the ASCII range are unmodified, and a zero byte never appears
46within non-zero characters.
166f8a29 47
eaf7a4d2
CS
48=cut
49*/
50
51/*
52=for apidoc is_ascii_string
53
a1433954 54Returns true if the first C<len> bytes of the string C<s> are the same whether
970ea3cb
KW
55or not the string is encoded in UTF-8 (or UTF-EBCDIC on EBCDIC machines). That
56is, if they are invariant. On ASCII-ish machines, only ASCII characters
57fit this definition, hence the function's name.
eaf7a4d2 58
9f7e3d64
MH
59If C<len> is 0, it will be calculated using C<strlen(s)>.
60
a1433954 61See also L</is_utf8_string>(), L</is_utf8_string_loclen>(), and L</is_utf8_string_loc>().
eaf7a4d2
CS
62
63=cut
64*/
65
66bool
668b6d8d 67Perl_is_ascii_string(const U8 *s, STRLEN len)
eaf7a4d2
CS
68{
69 const U8* const send = s + (len ? len : strlen((const char *)s));
70 const U8* x = s;
71
72 PERL_ARGS_ASSERT_IS_ASCII_STRING;
eaf7a4d2
CS
73
74 for (; x < send; ++x) {
75 if (!UTF8_IS_INVARIANT(*x))
76 break;
77 }
78
79 return x == send;
80}
81
82/*
378516de 83=for apidoc uvoffuni_to_utf8_flags
eebe1485 84
a27992cc 85THIS FUNCTION SHOULD BE USED IN ONLY VERY SPECIALIZED CIRCUMSTANCES.
de69f3af
KW
86Instead, B<Almost all code should use L</uvchr_to_utf8> or
87L</uvchr_to_utf8_flags>>.
a27992cc 88
de69f3af
KW
89This function is like them, but the input is a strict Unicode
90(as opposed to native) code point. Only in very rare circumstances should code
91not be using the native code point.
949cf498 92
de69f3af 93For details, see the description for L</uvchr_to_utf8_flags>>.
949cf498 94
eebe1485
SC
95=cut
96*/
97
dfe13c55 98U8 *
378516de 99Perl_uvoffuni_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags)
a0ed51b3 100{
378516de 101 PERL_ARGS_ASSERT_UVOFFUNI_TO_UTF8_FLAGS;
7918f24d 102
d9432125
KW
103 if (UNI_IS_INVARIANT(uv)) {
104 *d++ = (U8) LATIN1_TO_NATIVE(uv);
105 return d;
106 }
107
979f77b6
KW
108 /* The first problematic code point is the first surrogate */
109 if (uv >= UNICODE_SURROGATE_FIRST
110 && ckWARN4_d(WARN_UTF8, WARN_SURROGATE, WARN_NON_UNICODE, WARN_NONCHAR))
111 {
949cf498
KW
112 if (UNICODE_IS_SURROGATE(uv)) {
113 if (flags & UNICODE_WARN_SURROGATE) {
8457b38f 114 Perl_ck_warner_d(aTHX_ packWARN(WARN_SURROGATE),
949cf498
KW
115 "UTF-16 surrogate U+%04"UVXf, uv);
116 }
117 if (flags & UNICODE_DISALLOW_SURROGATE) {
118 return NULL;
119 }
120 }
121 else if (UNICODE_IS_SUPER(uv)) {
122 if (flags & UNICODE_WARN_SUPER
123 || (UNICODE_IS_FE_FF(uv) && (flags & UNICODE_WARN_FE_FF)))
124 {
8457b38f 125 Perl_ck_warner_d(aTHX_ packWARN(WARN_NON_UNICODE),
949cf498
KW
126 "Code point 0x%04"UVXf" is not Unicode, may not be portable", uv);
127 }
128 if (flags & UNICODE_DISALLOW_SUPER
129 || (UNICODE_IS_FE_FF(uv) && (flags & UNICODE_DISALLOW_FE_FF)))
130 {
131 return NULL;
132 }
133 }
134 else if (UNICODE_IS_NONCHAR(uv)) {
135 if (flags & UNICODE_WARN_NONCHAR) {
8457b38f 136 Perl_ck_warner_d(aTHX_ packWARN(WARN_NONCHAR),
949cf498
KW
137 "Unicode non-character U+%04"UVXf" is illegal for open interchange",
138 uv);
139 }
140 if (flags & UNICODE_DISALLOW_NONCHAR) {
141 return NULL;
142 }
143 }
507b9800 144 }
d9432125 145
2d331972 146#if defined(EBCDIC)
d9432125 147 {
5aaebcb3 148 STRLEN len = OFFUNISKIP(uv);
1d72bdf6
NIS
149 U8 *p = d+len-1;
150 while (p > d) {
bc3632a8 151 *p-- = (U8) I8_TO_NATIVE_UTF8((uv & UTF_CONTINUATION_MASK) | UTF_CONTINUATION_MARK);
1d72bdf6
NIS
152 uv >>= UTF_ACCUMULATION_SHIFT;
153 }
bc3632a8 154 *p = (U8) I8_TO_NATIVE_UTF8((uv & UTF_START_MASK(len)) | UTF_START_MARK(len));
1d72bdf6
NIS
155 return d+len;
156 }
157#else /* Non loop style */
a0ed51b3 158 if (uv < 0x800) {
eb160463
GS
159 *d++ = (U8)(( uv >> 6) | 0xc0);
160 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
161 return d;
162 }
163 if (uv < 0x10000) {
eb160463
GS
164 *d++ = (U8)(( uv >> 12) | 0xe0);
165 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
166 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
167 return d;
168 }
169 if (uv < 0x200000) {
eb160463
GS
170 *d++ = (U8)(( uv >> 18) | 0xf0);
171 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
172 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
173 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
174 return d;
175 }
176 if (uv < 0x4000000) {
eb160463
GS
177 *d++ = (U8)(( uv >> 24) | 0xf8);
178 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
179 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
180 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
181 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
182 return d;
183 }
184 if (uv < 0x80000000) {
eb160463
GS
185 *d++ = (U8)(( uv >> 30) | 0xfc);
186 *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80);
187 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
188 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
189 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
190 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
191 return d;
192 }
6588300d 193#ifdef UTF8_QUAD_MAX
d7578b48 194 if (uv < UTF8_QUAD_MAX)
a0ed51b3
LW
195#endif
196 {
eb160463
GS
197 *d++ = 0xfe; /* Can't match U+FEFF! */
198 *d++ = (U8)(((uv >> 30) & 0x3f) | 0x80);
199 *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80);
200 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
201 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
202 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
203 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
204 return d;
205 }
6588300d 206#ifdef UTF8_QUAD_MAX
a0ed51b3 207 {
eb160463
GS
208 *d++ = 0xff; /* Can't match U+FFFE! */
209 *d++ = 0x80; /* 6 Reserved bits */
210 *d++ = (U8)(((uv >> 60) & 0x0f) | 0x80); /* 2 Reserved bits */
211 *d++ = (U8)(((uv >> 54) & 0x3f) | 0x80);
212 *d++ = (U8)(((uv >> 48) & 0x3f) | 0x80);
213 *d++ = (U8)(((uv >> 42) & 0x3f) | 0x80);
214 *d++ = (U8)(((uv >> 36) & 0x3f) | 0x80);
215 *d++ = (U8)(((uv >> 30) & 0x3f) | 0x80);
216 *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80);
217 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
218 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
219 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
220 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
221 return d;
222 }
223#endif
537124e4 224#endif /* Non loop style */
a0ed51b3 225}
646ca15d 226/*
07693fe6
KW
227=for apidoc uvchr_to_utf8
228
bcb1a2d4 229Adds the UTF-8 representation of the native code point C<uv> to the end
07693fe6
KW
230of the string C<d>; C<d> should have at least C<UTF8_MAXBYTES+1> free
231bytes available. The return value is the pointer to the byte after the
232end of the new character. In other words,
233
234 d = uvchr_to_utf8(d, uv);
235
236is the recommended wide native character-aware way of saying
237
238 *(d++) = uv;
239
de69f3af
KW
240This function accepts any UV as input. To forbid or warn on non-Unicode code
241points, or those that may be problematic, see L</uvchr_to_utf8_flags>.
242
07693fe6
KW
243=cut
244*/
245
de69f3af
KW
246/* This is also a macro */
247PERL_CALLCONV U8* Perl_uvchr_to_utf8(pTHX_ U8 *d, UV uv);
248
07693fe6
KW
249U8 *
250Perl_uvchr_to_utf8(pTHX_ U8 *d, UV uv)
251{
de69f3af 252 return uvchr_to_utf8(d, uv);
07693fe6
KW
253}
254
de69f3af
KW
255/*
256=for apidoc uvchr_to_utf8_flags
257
258Adds the UTF-8 representation of the native code point C<uv> to the end
259of the string C<d>; C<d> should have at least C<UTF8_MAXBYTES+1> free
260bytes available. The return value is the pointer to the byte after the
261end of the new character. In other words,
262
263 d = uvchr_to_utf8_flags(d, uv, flags);
264
265or, in most cases,
266
267 d = uvchr_to_utf8_flags(d, uv, 0);
268
269This is the Unicode-aware way of saying
270
271 *(d++) = uv;
272
273This function will convert to UTF-8 (and not warn) even code points that aren't
274legal Unicode or are problematic, unless C<flags> contains one or more of the
275following flags:
276
277If C<uv> is a Unicode surrogate code point and UNICODE_WARN_SURROGATE is set,
278the function will raise a warning, provided UTF8 warnings are enabled. If instead
279UNICODE_DISALLOW_SURROGATE is set, the function will fail and return NULL.
280If both flags are set, the function will both warn and return NULL.
281
4c3cfd5d 282The UNICODE_WARN_NONCHAR and UNICODE_DISALLOW_NONCHAR flags
de69f3af 283affect how the function handles a Unicode non-character. And likewise, the
4c3cfd5d 284UNICODE_WARN_SUPER and UNICODE_DISALLOW_SUPER flags affect the handling of
de69f3af
KW
285code points that are
286above the Unicode maximum of 0x10FFFF. Code points above 0x7FFF_FFFF (which are
287even less portable) can be warned and/or disallowed even if other above-Unicode
288code points are accepted, by the UNICODE_WARN_FE_FF and UNICODE_DISALLOW_FE_FF
289flags.
290
291And finally, the flag UNICODE_WARN_ILLEGAL_INTERCHANGE selects all four of the
292above WARN flags; and UNICODE_DISALLOW_ILLEGAL_INTERCHANGE selects all four
293DISALLOW flags.
294
295=cut
296*/
297
298/* This is also a macro */
299PERL_CALLCONV U8* Perl_uvchr_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags);
300
07693fe6
KW
301U8 *
302Perl_uvchr_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags)
303{
de69f3af 304 return uvchr_to_utf8_flags(d, uv, flags);
07693fe6
KW
305}
306
307/*
646ca15d 308
f7d739d1 309Tests if the first C<len> bytes of string C<s> form a valid UTF-8
bcb1a2d4
KW
310character. Note that an INVARIANT (i.e. ASCII on non-EBCDIC) character is a
311valid UTF-8 character. The number of bytes in the UTF-8 character
646ca15d
JH
312will be returned if it is valid, otherwise 0.
313
314This is the "slow" version as opposed to the "fast" version which is
315the "unrolled" IS_UTF8_CHAR(). E.g. for t/uni/class.t the speed
316difference is a factor of 2 to 3. For lengths (UTF8SKIP(s)) of four
317or less you should use the IS_UTF8_CHAR(), for lengths of five or more
318you should use the _slow(). In practice this means that the _slow()
319will be used very rarely, since the maximum Unicode code point (as of
320Unicode 4.1) is U+10FFFF, which encodes in UTF-8 to four bytes. Only
537124e4 321the "Perl extended UTF-8" (e.g, the infamous 'v-strings') will encode into
646ca15d
JH
322five bytes or more.
323
324=cut */
7af276bc 325PERL_STATIC_INLINE STRLEN
5f66b61c 326S_is_utf8_char_slow(const U8 *s, const STRLEN len)
646ca15d 327{
cd7e6c88 328 dTHX; /* The function called below requires thread context */
646ca15d 329
cd7e6c88 330 STRLEN actual_len;
646ca15d 331
cd7e6c88 332 PERL_ARGS_ASSERT_IS_UTF8_CHAR_SLOW;
646ca15d 333
18712bce 334 utf8n_to_uvchr(s, len, &actual_len, UTF8_CHECK_ONLY);
646ca15d 335
cd7e6c88 336 return (actual_len == (STRLEN) -1) ? 0 : actual_len;
646ca15d 337}
9041c2e3
NIS
338
339/*
492a624f
KW
340=for apidoc is_utf8_char_buf
341
342Returns the number of bytes that comprise the first UTF-8 encoded character in
343buffer C<buf>. C<buf_end> should point to one position beyond the end of the
344buffer. 0 is returned if C<buf> does not point to a complete, valid UTF-8
345encoded character.
346
347Note that an INVARIANT character (i.e. ASCII on non-EBCDIC
348machines) is a valid UTF-8 character.
349
350=cut */
351
352STRLEN
353Perl_is_utf8_char_buf(const U8 *buf, const U8* buf_end)
354{
355
356 STRLEN len;
357
358 PERL_ARGS_ASSERT_IS_UTF8_CHAR_BUF;
359
360 if (buf_end <= buf) {
361 return 0;
362 }
363
364 len = buf_end - buf;
365 if (len > UTF8SKIP(buf)) {
366 len = UTF8SKIP(buf);
367 }
368
492a624f
KW
369 if (IS_UTF8_CHAR_FAST(len))
370 return IS_UTF8_CHAR(buf, len) ? len : 0;
492a624f
KW
371 return is_utf8_char_slow(buf, len);
372}
373
374/*
87cea99e 375=for apidoc is_utf8_char
eebe1485 376
5da9da9e 377Tests if some arbitrary number of bytes begins in a valid UTF-8
2bbc8d55
SP
378character. Note that an INVARIANT (i.e. ASCII on non-EBCDIC machines)
379character is a valid UTF-8 character. The actual number of bytes in the UTF-8
380character will be returned if it is valid, otherwise 0.
9041c2e3 381
76848387 382This function is deprecated due to the possibility that malformed input could
a1433954 383cause reading beyond the end of the input buffer. Use L</is_utf8_char_buf>
76848387 384instead.
e0328548 385
82686b01 386=cut */
76848387 387
067a85ef 388STRLEN
668b6d8d 389Perl_is_utf8_char(const U8 *s)
386d01d6 390{
7918f24d 391 PERL_ARGS_ASSERT_IS_UTF8_CHAR;
492a624f 392
76848387 393 /* Assumes we have enough space, which is why this is deprecated */
492a624f 394 return is_utf8_char_buf(s, s + UTF8SKIP(s));
386d01d6
GS
395}
396
eaf7a4d2 397
6662521e 398/*
87cea99e 399=for apidoc is_utf8_string
6662521e 400
a1433954 401Returns true if the first C<len> bytes of string C<s> form a valid
9f7e3d64 402UTF-8 string, false otherwise. If C<len> is 0, it will be calculated
e0328548
KW
403using C<strlen(s)> (which means if you use this option, that C<s> has to have a
404terminating NUL byte). Note that all characters being ASCII constitute 'a
405valid UTF-8 string'.
6662521e 406
a1433954 407See also L</is_ascii_string>(), L</is_utf8_string_loclen>(), and L</is_utf8_string_loc>().
768c67ee 408
6662521e
GS
409=cut
410*/
411
8e84507e 412bool
668b6d8d 413Perl_is_utf8_string(const U8 *s, STRLEN len)
6662521e 414{
35da51f7 415 const U8* const send = s + (len ? len : strlen((const char *)s));
7fc63493 416 const U8* x = s;
067a85ef 417
7918f24d 418 PERL_ARGS_ASSERT_IS_UTF8_STRING;
1aa99e6b 419
6662521e 420 while (x < send) {
1acdb0da 421 /* Inline the easy bits of is_utf8_char() here for speed... */
e0328548
KW
422 if (UTF8_IS_INVARIANT(*x)) {
423 x++;
424 }
1acdb0da
JH
425 else {
426 /* ... and call is_utf8_char() only if really needed. */
e0328548
KW
427 const STRLEN c = UTF8SKIP(x);
428 const U8* const next_char_ptr = x + c;
429
430 if (next_char_ptr > send) {
431 return FALSE;
432 }
433
768c67ee
JH
434 if (IS_UTF8_CHAR_FAST(c)) {
435 if (!IS_UTF8_CHAR(x, c))
e0328548 436 return FALSE;
3c614e38 437 }
e0328548
KW
438 else if (! is_utf8_char_slow(x, c)) {
439 return FALSE;
440 }
441 x = next_char_ptr;
1acdb0da 442 }
6662521e 443 }
768c67ee 444
067a85ef 445 return TRUE;
6662521e
GS
446}
447
67e989fb 448/*
814fafa7
NC
449Implemented as a macro in utf8.h
450
87cea99e 451=for apidoc is_utf8_string_loc
814fafa7 452
a1433954
KW
453Like L</is_utf8_string> but stores the location of the failure (in the
454case of "utf8ness failure") or the location C<s>+C<len> (in the case of
814fafa7
NC
455"utf8ness success") in the C<ep>.
456
a1433954 457See also L</is_utf8_string_loclen>() and L</is_utf8_string>().
814fafa7 458
87cea99e 459=for apidoc is_utf8_string_loclen
81cd54e3 460
a1433954
KW
461Like L</is_utf8_string>() but stores the location of the failure (in the
462case of "utf8ness failure") or the location C<s>+C<len> (in the case of
768c67ee
JH
463"utf8ness success") in the C<ep>, and the number of UTF-8
464encoded characters in the C<el>.
465
a1433954 466See also L</is_utf8_string_loc>() and L</is_utf8_string>().
81cd54e3
JH
467
468=cut
469*/
470
471bool
668b6d8d 472Perl_is_utf8_string_loclen(const U8 *s, STRLEN len, const U8 **ep, STRLEN *el)
81cd54e3 473{
35da51f7 474 const U8* const send = s + (len ? len : strlen((const char *)s));
7fc63493 475 const U8* x = s;
81cd54e3 476 STRLEN c;
3ebfea28 477 STRLEN outlen = 0;
7918f24d
NC
478
479 PERL_ARGS_ASSERT_IS_UTF8_STRING_LOCLEN;
81cd54e3 480
81cd54e3 481 while (x < send) {
e0328548
KW
482 const U8* next_char_ptr;
483
81cd54e3
JH
484 /* Inline the easy bits of is_utf8_char() here for speed... */
485 if (UTF8_IS_INVARIANT(*x))
e0328548 486 next_char_ptr = x + 1;
81cd54e3 487 else {
768c67ee 488 /* ... and call is_utf8_char() only if really needed. */
768c67ee 489 c = UTF8SKIP(x);
e0328548
KW
490 next_char_ptr = c + x;
491 if (next_char_ptr > send) {
492 goto out;
493 }
768c67ee
JH
494 if (IS_UTF8_CHAR_FAST(c)) {
495 if (!IS_UTF8_CHAR(x, c))
496 c = 0;
497 } else
498 c = is_utf8_char_slow(x, c);
768c67ee
JH
499 if (!c)
500 goto out;
81cd54e3 501 }
e0328548 502 x = next_char_ptr;
3ebfea28 503 outlen++;
81cd54e3 504 }
768c67ee
JH
505
506 out:
3ebfea28
AL
507 if (el)
508 *el = outlen;
509
768c67ee
JH
510 if (ep)
511 *ep = x;
3ebfea28 512 return (x == send);
81cd54e3
JH
513}
514
515/*
768c67ee 516
de69f3af 517=for apidoc utf8n_to_uvchr
378516de
KW
518
519THIS FUNCTION SHOULD BE USED IN ONLY VERY SPECIALIZED CIRCUMSTANCES.
de69f3af 520Most code should use L</utf8_to_uvchr_buf>() rather than call this directly.
67e989fb 521
9041c2e3 522Bottom level UTF-8 decode routine.
de69f3af 523Returns the native code point value of the first character in the string C<s>,
746afd53
KW
524which is assumed to be in UTF-8 (or UTF-EBCDIC) encoding, and no longer than
525C<curlen> bytes; C<*retlen> (if C<retlen> isn't NULL) will be set to
526the length, in bytes, of that character.
949cf498
KW
527
528The value of C<flags> determines the behavior when C<s> does not point to a
529well-formed UTF-8 character. If C<flags> is 0, when a malformation is found,
524080c4
KW
530zero is returned and C<*retlen> is set so that (S<C<s> + C<*retlen>>) is the
531next possible position in C<s> that could begin a non-malformed character.
532Also, if UTF-8 warnings haven't been lexically disabled, a warning is raised.
949cf498
KW
533
534Various ALLOW flags can be set in C<flags> to allow (and not warn on)
535individual types of malformations, such as the sequence being overlong (that
536is, when there is a shorter sequence that can express the same code point;
537overlong sequences are expressly forbidden in the UTF-8 standard due to
538potential security issues). Another malformation example is the first byte of
539a character not being a legal first byte. See F<utf8.h> for the list of such
524080c4
KW
540flags. For allowed 0 length strings, this function returns 0; for allowed
541overlong sequences, the computed code point is returned; for all other allowed
542malformations, the Unicode REPLACEMENT CHARACTER is returned, as these have no
543determinable reasonable value.
949cf498
KW
544
545The UTF8_CHECK_ONLY flag overrides the behavior when a non-allowed (by other
546flags) malformation is found. If this flag is set, the routine assumes that
547the caller will raise a warning, and this function will silently just set
d088425d
KW
548C<retlen> to C<-1> (cast to C<STRLEN>) and return zero.
549
550Note that this API requires disambiguation between successful decoding a NUL
551character, and an error return (unless the UTF8_CHECK_ONLY flag is set), as
552in both cases, 0 is returned. To disambiguate, upon a zero return, see if the
553first byte of C<s> is 0 as well. If so, the input was a NUL; if not, the input
554had an error.
949cf498
KW
555
556Certain code points are considered problematic. These are Unicode surrogates,
746afd53 557Unicode non-characters, and code points above the Unicode maximum of 0x10FFFF.
949cf498 558By default these are considered regular code points, but certain situations
5eafe189 559warrant special handling for them. If C<flags> contains
949cf498
KW
560UTF8_DISALLOW_ILLEGAL_INTERCHANGE, all three classes are treated as
561malformations and handled as such. The flags UTF8_DISALLOW_SURROGATE,
562UTF8_DISALLOW_NONCHAR, and UTF8_DISALLOW_SUPER (meaning above the legal Unicode
563maximum) can be set to disallow these categories individually.
564
565The flags UTF8_WARN_ILLEGAL_INTERCHANGE, UTF8_WARN_SURROGATE,
566UTF8_WARN_NONCHAR, and UTF8_WARN_SUPER will cause warning messages to be raised
567for their respective categories, but otherwise the code points are considered
568valid (not malformations). To get a category to both be treated as a
569malformation and raise a warning, specify both the WARN and DISALLOW flags.
570(But note that warnings are not raised if lexically disabled nor if
571UTF8_CHECK_ONLY is also specified.)
572
573Very large code points (above 0x7FFF_FFFF) are considered more problematic than
574the others that are above the Unicode legal maximum. There are several
eb83ed87
KW
575reasons: they requre at least 32 bits to represent them on ASCII platforms, are
576not representable at all on EBCDIC platforms, and the original UTF-8
577specification never went above this number (the current 0x10FFFF limit was
578imposed later). (The smaller ones, those that fit into 32 bits, are
579representable by a UV on ASCII platforms, but not by an IV, which means that
580the number of operations that can be performed on them is quite restricted.)
581The UTF-8 encoding on ASCII platforms for these large code points begins with a
582byte containing 0xFE or 0xFF. The UTF8_DISALLOW_FE_FF flag will cause them to
583be treated as malformations, while allowing smaller above-Unicode code points.
584(Of course UTF8_DISALLOW_SUPER will treat all above-Unicode code points,
585including these, as malformations.) Similarly, UTF8_WARN_FE_FF acts just like
586the other WARN flags, but applies just to these code points.
949cf498
KW
587
588All other code points corresponding to Unicode characters, including private
589use and those yet to be assigned, are never considered malformed and never
590warn.
67e989fb 591
37607a96
PK
592=cut
593*/
67e989fb 594
a0ed51b3 595UV
de69f3af 596Perl_utf8n_to_uvchr(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen, U32 flags)
a0ed51b3 597{
97aff369 598 dVAR;
d4c19fe8 599 const U8 * const s0 = s;
eb83ed87 600 U8 overflow_byte = '\0'; /* Save byte in case of overflow */
0b8d30e8 601 U8 * send;
eb83ed87
KW
602 UV uv = *s;
603 STRLEN expectlen;
949cf498 604 SV* sv = NULL;
eb83ed87
KW
605 UV outlier_ret = 0; /* return value when input is in error or problematic
606 */
607 UV pack_warn = 0; /* Save result of packWARN() for later */
608 bool unexpected_non_continuation = FALSE;
609 bool overflowed = FALSE;
2f8f112e 610 bool do_overlong_test = TRUE; /* May have to skip this test */
a0dbb045 611
eb83ed87 612 const char* const malformed_text = "Malformed UTF-8 character";
7918f24d 613
de69f3af 614 PERL_ARGS_ASSERT_UTF8N_TO_UVCHR;
a0dbb045 615
eb83ed87
KW
616 /* The order of malformation tests here is important. We should consume as
617 * few bytes as possible in order to not skip any valid character. This is
618 * required by the Unicode Standard (section 3.9 of Unicode 6.0); see also
619 * http://unicode.org/reports/tr36 for more discussion as to why. For
620 * example, once we've done a UTF8SKIP, we can tell the expected number of
621 * bytes, and could fail right off the bat if the input parameters indicate
622 * that there are too few available. But it could be that just that first
623 * byte is garbled, and the intended character occupies fewer bytes. If we
624 * blindly assumed that the first byte is correct, and skipped based on
625 * that number, we could skip over a valid input character. So instead, we
626 * always examine the sequence byte-by-byte.
627 *
628 * We also should not consume too few bytes, otherwise someone could inject
629 * things. For example, an input could be deliberately designed to
630 * overflow, and if this code bailed out immediately upon discovering that,
e2660c54 631 * returning to the caller C<*retlen> pointing to the very next byte (one
eb83ed87
KW
632 * which is actually part of of the overflowing sequence), that could look
633 * legitimate to the caller, which could discard the initial partial
634 * sequence and process the rest, inappropriately */
635
636 /* Zero length strings, if allowed, of necessity are zero */
b5b9af04 637 if (UNLIKELY(curlen == 0)) {
eb83ed87
KW
638 if (retlen) {
639 *retlen = 0;
640 }
a0dbb045 641
eb83ed87
KW
642 if (flags & UTF8_ALLOW_EMPTY) {
643 return 0;
644 }
645 if (! (flags & UTF8_CHECK_ONLY)) {
646 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (empty string)", malformed_text));
647 }
0c443dc2
JH
648 goto malformed;
649 }
650
eb83ed87
KW
651 expectlen = UTF8SKIP(s);
652
653 /* A well-formed UTF-8 character, as the vast majority of calls to this
654 * function will be for, has this expected length. For efficiency, set
655 * things up here to return it. It will be overriden only in those rare
656 * cases where a malformation is found */
657 if (retlen) {
658 *retlen = expectlen;
659 }
660
661 /* An invariant is trivially well-formed */
1d72bdf6 662 if (UTF8_IS_INVARIANT(uv)) {
de69f3af 663 return uv;
a0ed51b3 664 }
67e989fb 665
eb83ed87 666 /* A continuation character can't start a valid sequence */
b5b9af04 667 if (UNLIKELY(UTF8_IS_CONTINUATION(uv))) {
eb83ed87
KW
668 if (flags & UTF8_ALLOW_CONTINUATION) {
669 if (retlen) {
670 *retlen = 1;
671 }
672 return UNICODE_REPLACEMENT;
673 }
ba210ebe 674
eb83ed87
KW
675 if (! (flags & UTF8_CHECK_ONLY)) {
676 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (unexpected continuation byte 0x%02x, with no preceding start byte)", malformed_text, *s0));
677 }
678 curlen = 1;
ba210ebe
JH
679 goto malformed;
680 }
9041c2e3 681
dcd27b3c
KW
682 /* Here is not a continuation byte, nor an invariant. The only thing left
683 * is a start byte (possibly for an overlong) */
684
1d72bdf6 685#ifdef EBCDIC
bc3632a8 686 uv = NATIVE_UTF8_TO_I8(uv);
1d72bdf6
NIS
687#endif
688
eb83ed87
KW
689 /* Remove the leading bits that indicate the number of bytes in the
690 * character's whole UTF-8 sequence, leaving just the bits that are part of
691 * the value */
692 uv &= UTF_START_MASK(expectlen);
ba210ebe 693
eb83ed87
KW
694 /* Now, loop through the remaining bytes in the character's sequence,
695 * accumulating each into the working value as we go. Be sure to not look
696 * past the end of the input string */
0b8d30e8
KW
697 send = (U8*) s0 + ((expectlen <= curlen) ? expectlen : curlen);
698
eb83ed87 699 for (s = s0 + 1; s < send; s++) {
b5b9af04 700 if (LIKELY(UTF8_IS_CONTINUATION(*s))) {
eb83ed87
KW
701#ifndef EBCDIC /* Can't overflow in EBCDIC */
702 if (uv & UTF_ACCUMULATION_OVERFLOW_MASK) {
703
704 /* The original implementors viewed this malformation as more
705 * serious than the others (though I, khw, don't understand
706 * why, since other malformations also give very very wrong
707 * results), so there is no way to turn off checking for it.
708 * Set a flag, but keep going in the loop, so that we absorb
709 * the rest of the bytes that comprise the character. */
710 overflowed = TRUE;
711 overflow_byte = *s; /* Save for warning message's use */
712 }
713#endif
8850bf83 714 uv = UTF8_ACCUMULATE(uv, *s);
eb83ed87
KW
715 }
716 else {
717 /* Here, found a non-continuation before processing all expected
718 * bytes. This byte begins a new character, so quit, even if
719 * allowing this malformation. */
720 unexpected_non_continuation = TRUE;
721 break;
722 }
723 } /* End of loop through the character's bytes */
724
725 /* Save how many bytes were actually in the character */
726 curlen = s - s0;
727
728 /* The loop above finds two types of malformations: non-continuation and/or
729 * overflow. The non-continuation malformation is really a too-short
730 * malformation, as it means that the current character ended before it was
731 * expected to (being terminated prematurely by the beginning of the next
732 * character, whereas in the too-short malformation there just are too few
733 * bytes available to hold the character. In both cases, the check below
734 * that we have found the expected number of bytes would fail if executed.)
735 * Thus the non-continuation malformation is really unnecessary, being a
736 * subset of the too-short malformation. But there may be existing
737 * applications that are expecting the non-continuation type, so we retain
738 * it, and return it in preference to the too-short malformation. (If this
739 * code were being written from scratch, the two types might be collapsed
740 * into one.) I, khw, am also giving priority to returning the
741 * non-continuation and too-short malformations over overflow when multiple
742 * ones are present. I don't know of any real reason to prefer one over
743 * the other, except that it seems to me that multiple-byte errors trumps
744 * errors from a single byte */
b5b9af04 745 if (UNLIKELY(unexpected_non_continuation)) {
eb83ed87
KW
746 if (!(flags & UTF8_ALLOW_NON_CONTINUATION)) {
747 if (! (flags & UTF8_CHECK_ONLY)) {
748 if (curlen == 1) {
749 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (unexpected non-continuation byte 0x%02x, immediately after start byte 0x%02x)", malformed_text, *s, *s0));
750 }
751 else {
752 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
753 }
754 }
eb83ed87
KW
755 goto malformed;
756 }
757 uv = UNICODE_REPLACEMENT;
2f8f112e
KW
758
759 /* Skip testing for overlongs, as the REPLACEMENT may not be the same
760 * as what the original expectations were. */
761 do_overlong_test = FALSE;
eb83ed87
KW
762 if (retlen) {
763 *retlen = curlen;
764 }
765 }
b5b9af04 766 else if (UNLIKELY(curlen < expectlen)) {
eb83ed87
KW
767 if (! (flags & UTF8_ALLOW_SHORT)) {
768 if (! (flags & UTF8_CHECK_ONLY)) {
769 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 770 }
eb83ed87
KW
771 goto malformed;
772 }
773 uv = UNICODE_REPLACEMENT;
2f8f112e 774 do_overlong_test = FALSE;
eb83ed87
KW
775 if (retlen) {
776 *retlen = curlen;
777 }
778 }
779
780#ifndef EBCDIC /* EBCDIC allows FE, FF, can't overflow */
2f8f112e 781 if ((*s0 & 0xFE) == 0xFE /* matches both FE, FF */
eb83ed87
KW
782 && (flags & (UTF8_WARN_FE_FF|UTF8_DISALLOW_FE_FF)))
783 {
784 /* By adding UTF8_CHECK_ONLY to the test, we avoid unnecessary
785 * generation of the sv, since no warnings are raised under CHECK */
786 if ((flags & (UTF8_WARN_FE_FF|UTF8_CHECK_ONLY)) == UTF8_WARN_FE_FF
787 && ckWARN_d(WARN_UTF8))
788 {
42303544
KW
789 /* This message is deliberately not of the same syntax as the other
790 * messages for malformations, for backwards compatibility in the
791 * unlikely event that code is relying on its precise earlier text
792 */
eb83ed87
KW
793 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s Code point beginning with byte 0x%02X is not Unicode, and not portable", malformed_text, *s0));
794 pack_warn = packWARN(WARN_UTF8);
795 }
796 if (flags & UTF8_DISALLOW_FE_FF) {
797 goto malformed;
ba210ebe 798 }
ba210ebe 799 }
b5b9af04 800 if (UNLIKELY(overflowed)) {
ba210ebe 801
eb83ed87
KW
802 /* If the first byte is FF, it will overflow a 32-bit word. If the
803 * first byte is FE, it will overflow a signed 32-bit word. The
804 * above preserves backward compatibility, since its message was used
805 * in earlier versions of this code in preference to overflow */
806 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (overflow at byte 0x%02x, after start byte 0x%02x)", malformed_text, overflow_byte, *s0));
ba210ebe 807 goto malformed;
eb83ed87
KW
808 }
809#endif
810
2f8f112e 811 if (do_overlong_test
5aaebcb3 812 && expectlen > (STRLEN) OFFUNISKIP(uv)
2f8f112e
KW
813 && ! (flags & UTF8_ALLOW_LONG))
814 {
eb83ed87
KW
815 /* The overlong malformation has lower precedence than the others.
816 * Note that if this malformation is allowed, we return the actual
817 * value, instead of the replacement character. This is because this
818 * value is actually well-defined. */
819 if (! (flags & UTF8_CHECK_ONLY)) {
5aaebcb3 820 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (%d byte%s, need %d, after start byte 0x%02x)", malformed_text, (int)expectlen, expectlen == 1 ? "": "s", OFFUNISKIP(uv), *s0));
eb83ed87
KW
821 }
822 goto malformed;
823 }
824
825 /* Here, the input is considered to be well-formed , but could be a
826 * problematic code point that is not allowed by the input parameters. */
827 if (uv >= UNICODE_SURROGATE_FIRST /* isn't problematic if < this */
828 && (flags & (UTF8_DISALLOW_ILLEGAL_INTERCHANGE
829 |UTF8_WARN_ILLEGAL_INTERCHANGE)))
830 {
949cf498 831 if (UNICODE_IS_SURROGATE(uv)) {
eb83ed87
KW
832 if ((flags & (UTF8_WARN_SURROGATE|UTF8_CHECK_ONLY)) == UTF8_WARN_SURROGATE
833 && ckWARN2_d(WARN_UTF8, WARN_SURROGATE))
834 {
111d382d 835 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "UTF-16 surrogate U+%04"UVXf"", uv));
eb83ed87 836 pack_warn = packWARN2(WARN_UTF8, WARN_SURROGATE);
949cf498
KW
837 }
838 if (flags & UTF8_DISALLOW_SURROGATE) {
839 goto disallowed;
840 }
841 }
949cf498 842 else if ((uv > PERL_UNICODE_MAX)) {
eb83ed87
KW
843 if ((flags & (UTF8_WARN_SUPER|UTF8_CHECK_ONLY)) == UTF8_WARN_SUPER
844 && ckWARN2_d(WARN_UTF8, WARN_NON_UNICODE))
845 {
111d382d 846 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "Code point 0x%04"UVXf" is not Unicode, may not be portable", uv));
eb83ed87 847 pack_warn = packWARN2(WARN_UTF8, WARN_NON_UNICODE);
949cf498
KW
848 }
849 if (flags & UTF8_DISALLOW_SUPER) {
850 goto disallowed;
851 }
852 }
4190d317
KW
853 else if (UNICODE_IS_NONCHAR(uv)) {
854 if ((flags & (UTF8_WARN_NONCHAR|UTF8_CHECK_ONLY)) == UTF8_WARN_NONCHAR
855 && ckWARN2_d(WARN_UTF8, WARN_NONCHAR))
856 {
857 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "Unicode non-character U+%04"UVXf" is illegal for open interchange", uv));
858 pack_warn = packWARN2(WARN_UTF8, WARN_NONCHAR);
859 }
860 if (flags & UTF8_DISALLOW_NONCHAR) {
861 goto disallowed;
862 }
863 }
949cf498 864
eb83ed87 865 if (sv) {
de69f3af
KW
866 outlier_ret = uv; /* Note we don't bother to convert to native,
867 as all the outlier code points are the same
868 in both ASCII and EBCDIC */
eb83ed87
KW
869 goto do_warn;
870 }
871
949cf498
KW
872 /* Here, this is not considered a malformed character, so drop through
873 * to return it */
a0ed51b3 874 }
ba210ebe 875
de69f3af 876 return UNI_TO_NATIVE(uv);
ba210ebe 877
eb83ed87
KW
878 /* There are three cases which get to beyond this point. In all 3 cases:
879 * <sv> if not null points to a string to print as a warning.
880 * <curlen> is what <*retlen> should be set to if UTF8_CHECK_ONLY isn't
881 * set.
882 * <outlier_ret> is what return value to use if UTF8_CHECK_ONLY isn't set.
883 * This is done by initializing it to 0, and changing it only
884 * for case 1).
885 * The 3 cases are:
886 * 1) The input is valid but problematic, and to be warned about. The
887 * return value is the resultant code point; <*retlen> is set to
888 * <curlen>, the number of bytes that comprise the code point.
889 * <pack_warn> contains the result of packWARN() for the warning
890 * types. The entry point for this case is the label <do_warn>;
891 * 2) The input is a valid code point but disallowed by the parameters to
892 * this function. The return value is 0. If UTF8_CHECK_ONLY is set,
893 * <*relen> is -1; otherwise it is <curlen>, the number of bytes that
894 * comprise the code point. <pack_warn> contains the result of
895 * packWARN() for the warning types. The entry point for this case is
896 * the label <disallowed>.
897 * 3) The input is malformed. The return value is 0. If UTF8_CHECK_ONLY
898 * is set, <*relen> is -1; otherwise it is <curlen>, the number of
899 * bytes that comprise the malformation. All such malformations are
900 * assumed to be warning type <utf8>. The entry point for this case
901 * is the label <malformed>.
902 */
949cf498 903
ba210ebe
JH
904malformed:
905
eb83ed87
KW
906 if (sv && ckWARN_d(WARN_UTF8)) {
907 pack_warn = packWARN(WARN_UTF8);
908 }
909
910disallowed:
911
fcc8fcf6 912 if (flags & UTF8_CHECK_ONLY) {
ba210ebe 913 if (retlen)
10edeb5d 914 *retlen = ((STRLEN) -1);
ba210ebe
JH
915 return 0;
916 }
917
eb83ed87 918do_warn:
5b311467 919
eb83ed87
KW
920 if (pack_warn) { /* <pack_warn> was initialized to 0, and changed only
921 if warnings are to be raised. */
f555bc63 922 const char * const string = SvPVX_const(sv);
a0dbb045 923
f555bc63
KW
924 if (PL_op)
925 Perl_warner(aTHX_ pack_warn, "%s in %s", string, OP_DESC(PL_op));
926 else
927 Perl_warner(aTHX_ pack_warn, "%s", string);
a0dbb045
JH
928 }
929
eb83ed87
KW
930 if (retlen) {
931 *retlen = curlen;
932 }
ba210ebe 933
eb83ed87 934 return outlier_ret;
a0ed51b3
LW
935}
936
8e84507e 937/*
ec5f19d0
KW
938=for apidoc utf8_to_uvchr_buf
939
940Returns the native code point of the first character in the string C<s> which
941is assumed to be in UTF-8 encoding; C<send> points to 1 beyond the end of C<s>.
524080c4 942C<*retlen> will be set to the length, in bytes, of that character.
ec5f19d0 943
524080c4
KW
944If C<s> does not point to a well-formed UTF-8 character and UTF8 warnings are
945enabled, zero is returned and C<*retlen> is set (if C<retlen> isn't
173db420
KW
946NULL) to -1. If those warnings are off, the computed value, if well-defined
947(or the Unicode REPLACEMENT CHARACTER if not), is silently returned, and
948C<*retlen> is set (if C<retlen> isn't NULL) so that (S<C<s> + C<*retlen>>) is
949the next possible position in C<s> that could begin a non-malformed character.
de69f3af 950See L</utf8n_to_uvchr> for details on when the REPLACEMENT CHARACTER is
173db420 951returned.
ec5f19d0
KW
952
953=cut
954*/
955
956
957UV
958Perl_utf8_to_uvchr_buf(pTHX_ const U8 *s, const U8 *send, STRLEN *retlen)
959{
ec5f19d0
KW
960 assert(s < send);
961
962 return utf8n_to_uvchr(s, send - s, retlen,
963 ckWARN_d(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
964}
965
27d6c58a 966/* Like L</utf8_to_uvchr_buf>(), but should only be called when it is known that
3986bb7c 967 * there are no malformations in the input UTF-8 string C<s>. surrogates,
57b0056d 968 * non-character code points, and non-Unicode code points are allowed. */
27d6c58a
KW
969
970UV
971Perl_valid_utf8_to_uvchr(pTHX_ const U8 *s, STRLEN *retlen)
972{
010ab96b
KW
973 UV expectlen = UTF8SKIP(s);
974 const U8* send = s + expectlen;
9ff2f0f7 975 UV uv = *s;
3986bb7c 976
27d6c58a
KW
977 PERL_ARGS_ASSERT_VALID_UTF8_TO_UVCHR;
978
010ab96b
KW
979 if (retlen) {
980 *retlen = expectlen;
981 }
982
983 /* An invariant is trivially returned */
984 if (expectlen == 1) {
9ff2f0f7 985 return uv;
010ab96b
KW
986 }
987
9ff2f0f7
KW
988#ifdef EBCDIC
989 uv = NATIVE_UTF8_TO_I8(uv);
990#endif
991
010ab96b
KW
992 /* Remove the leading bits that indicate the number of bytes, leaving just
993 * the bits that are part of the value */
994 uv &= UTF_START_MASK(expectlen);
995
996 /* Now, loop through the remaining bytes, accumulating each into the
997 * working total as we go. (I khw tried unrolling the loop for up to 4
998 * bytes, but there was no performance improvement) */
999 for (++s; s < send; s++) {
1000 uv = UTF8_ACCUMULATE(uv, *s);
1001 }
1002
3986bb7c 1003 return UNI_TO_NATIVE(uv);
010ab96b 1004
27d6c58a
KW
1005}
1006
ec5f19d0 1007/*
87cea99e 1008=for apidoc utf8_to_uvchr
9041c2e3 1009
6ee84de2 1010Returns the native code point of the first character in the string C<s>
1e54db1a 1011which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
9041c2e3
NIS
1012length, in bytes, of that character.
1013
4b88fb76 1014Some, but not all, UTF-8 malformations are detected, and in fact, some
977c1d31
KW
1015malformed input could cause reading beyond the end of the input buffer, which
1016is why this function is deprecated. Use L</utf8_to_uvchr_buf> instead.
4b88fb76 1017
524080c4
KW
1018If C<s> points to one of the detected malformations, and UTF8 warnings are
1019enabled, zero is returned and C<*retlen> is set (if C<retlen> isn't
1020NULL) to -1. If those warnings are off, the computed value if well-defined (or
1021the Unicode REPLACEMENT CHARACTER, if not) is silently returned, and C<*retlen>
1022is set (if C<retlen> isn't NULL) so that (S<C<s> + C<*retlen>>) is the
1023next possible position in C<s> that could begin a non-malformed character.
de69f3af 1024See L</utf8n_to_uvchr> for details on when the REPLACEMENT CHARACTER is returned.
9041c2e3
NIS
1025
1026=cut
1027*/
1028
1029UV
7fc63493 1030Perl_utf8_to_uvchr(pTHX_ const U8 *s, STRLEN *retlen)
9041c2e3 1031{
7918f24d
NC
1032 PERL_ARGS_ASSERT_UTF8_TO_UVCHR;
1033
2ff6c191 1034 return utf8_to_uvchr_buf(s, s + UTF8_MAXBYTES, retlen);
9041c2e3
NIS
1035}
1036
1037/*
ec5f19d0
KW
1038=for apidoc utf8_to_uvuni_buf
1039
de69f3af
KW
1040Only in very rare circumstances should code need to be dealing in Unicode
1041(as opposed to native) code points. In those few cases, use
1042C<L<NATIVE_TO_UNI(utf8_to_uvchr_buf(...))|/utf8_to_uvchr_buf>> instead.
4f83cdcd
KW
1043
1044Returns the Unicode (not-native) code point of the first character in the
1045string C<s> which
ec5f19d0
KW
1046is assumed to be in UTF-8 encoding; C<send> points to 1 beyond the end of C<s>.
1047C<retlen> will be set to the length, in bytes, of that character.
1048
524080c4
KW
1049If C<s> does not point to a well-formed UTF-8 character and UTF8 warnings are
1050enabled, zero is returned and C<*retlen> is set (if C<retlen> isn't
1051NULL) to -1. If those warnings are off, the computed value if well-defined (or
1052the Unicode REPLACEMENT CHARACTER, if not) is silently returned, and C<*retlen>
1053is set (if C<retlen> isn't NULL) so that (S<C<s> + C<*retlen>>) is the
1054next possible position in C<s> that could begin a non-malformed character.
de69f3af 1055See L</utf8n_to_uvchr> for details on when the REPLACEMENT CHARACTER is returned.
ec5f19d0
KW
1056
1057=cut
1058*/
1059
1060UV
1061Perl_utf8_to_uvuni_buf(pTHX_ const U8 *s, const U8 *send, STRLEN *retlen)
1062{
1063 PERL_ARGS_ASSERT_UTF8_TO_UVUNI_BUF;
1064
1065 assert(send > s);
1066
1067 /* Call the low level routine asking for checks */
de69f3af
KW
1068 return NATIVE_TO_UNI(Perl_utf8n_to_uvchr(aTHX_ s, send -s, retlen,
1069 ckWARN_d(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY));
ec5f19d0
KW
1070}
1071
5495102a
KW
1072/* DEPRECATED!
1073 * Like L</utf8_to_uvuni_buf>(), but should only be called when it is known that
2114036c 1074 * there are no malformations in the input UTF-8 string C<s>. Surrogates,
3986bb7c 1075 * non-character code points, and non-Unicode code points are allowed */
27d6c58a
KW
1076
1077UV
1078Perl_valid_utf8_to_uvuni(pTHX_ const U8 *s, STRLEN *retlen)
1079{
1080 PERL_ARGS_ASSERT_VALID_UTF8_TO_UVUNI;
1081
010ab96b 1082 return NATIVE_TO_UNI(valid_utf8_to_uvchr(s, retlen));
27d6c58a
KW
1083}
1084
ec5f19d0 1085/*
87cea99e 1086=for apidoc utf8_to_uvuni
9041c2e3
NIS
1087
1088Returns the Unicode code point of the first character in the string C<s>
1e54db1a 1089which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
9041c2e3
NIS
1090length, in bytes, of that character.
1091
4b88fb76 1092Some, but not all, UTF-8 malformations are detected, and in fact, some
977c1d31 1093malformed input could cause reading beyond the end of the input buffer, which
4f83cdcd
KW
1094is one reason why this function is deprecated. The other is that only in
1095extremely limited circumstances should the Unicode versus native code point be
de69f3af 1096of any interest to you. See L</utf8_to_uvuni_buf> for alternatives.
9041c2e3 1097
524080c4
KW
1098If C<s> points to one of the detected malformations, and UTF8 warnings are
1099enabled, zero is returned and C<*retlen> is set (if C<retlen> doesn't point to
1100NULL) to -1. If those warnings are off, the computed value if well-defined (or
1101the Unicode REPLACEMENT CHARACTER, if not) is silently returned, and C<*retlen>
1102is set (if C<retlen> isn't NULL) so that (S<C<s> + C<*retlen>>) is the
1103next possible position in C<s> that could begin a non-malformed character.
de69f3af 1104See L</utf8n_to_uvchr> for details on when the REPLACEMENT CHARACTER is returned.
8e84507e
NIS
1105
1106=cut
1107*/
1108
1109UV
7fc63493 1110Perl_utf8_to_uvuni(pTHX_ const U8 *s, STRLEN *retlen)
8e84507e 1111{
7918f24d
NC
1112 PERL_ARGS_ASSERT_UTF8_TO_UVUNI;
1113
5495102a 1114 return NATIVE_TO_UNI(valid_utf8_to_uvchr(s, retlen));
8e84507e
NIS
1115}
1116
b76347f2 1117/*
87cea99e 1118=for apidoc utf8_length
b76347f2
JH
1119
1120Return the length of the UTF-8 char encoded string C<s> in characters.
02eb7b47
JH
1121Stops at C<e> (inclusive). If C<e E<lt> s> or if the scan would end
1122up past C<e>, croaks.
b76347f2
JH
1123
1124=cut
1125*/
1126
1127STRLEN
35a4481c 1128Perl_utf8_length(pTHX_ const U8 *s, const U8 *e)
b76347f2 1129{
97aff369 1130 dVAR;
b76347f2
JH
1131 STRLEN len = 0;
1132
7918f24d
NC
1133 PERL_ARGS_ASSERT_UTF8_LENGTH;
1134
8850bf83
JH
1135 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g.
1136 * the bitops (especially ~) can create illegal UTF-8.
1137 * In other words: in Perl UTF-8 is not just for Unicode. */
1138
a3b680e6
AL
1139 if (e < s)
1140 goto warn_and_return;
b76347f2 1141 while (s < e) {
4cbf4130 1142 s += UTF8SKIP(s);
8e91ec7f
AV
1143 len++;
1144 }
1145
1146 if (e != s) {
1147 len--;
1148 warn_and_return:
9b387841
NC
1149 if (PL_op)
1150 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8),
1151 "%s in %s", unees, OP_DESC(PL_op));
1152 else
61a12c31 1153 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), "%s", unees);
b76347f2
JH
1154 }
1155
1156 return len;
1157}
1158
b06226ff 1159/*
87cea99e 1160=for apidoc utf8_distance
b06226ff 1161
1e54db1a 1162Returns the number of UTF-8 characters between the UTF-8 pointers C<a>
b06226ff
JH
1163and C<b>.
1164
1165WARNING: use only if you *know* that the pointers point inside the
1166same UTF-8 buffer.
1167
37607a96
PK
1168=cut
1169*/
a0ed51b3 1170
02eb7b47 1171IV
35a4481c 1172Perl_utf8_distance(pTHX_ const U8 *a, const U8 *b)
a0ed51b3 1173{
7918f24d
NC
1174 PERL_ARGS_ASSERT_UTF8_DISTANCE;
1175
bf1665bc 1176 return (a < b) ? -1 * (IV) utf8_length(a, b) : (IV) utf8_length(b, a);
a0ed51b3
LW
1177}
1178
b06226ff 1179/*
87cea99e 1180=for apidoc utf8_hop
b06226ff 1181
8850bf83
JH
1182Return the UTF-8 pointer C<s> displaced by C<off> characters, either
1183forward or backward.
b06226ff
JH
1184
1185WARNING: do not use the following unless you *know* C<off> is within
8850bf83
JH
1186the UTF-8 data pointed to by C<s> *and* that on entry C<s> is aligned
1187on the first byte of character or just after the last byte of a character.
b06226ff 1188
37607a96
PK
1189=cut
1190*/
a0ed51b3
LW
1191
1192U8 *
4373e329 1193Perl_utf8_hop(pTHX_ const U8 *s, I32 off)
a0ed51b3 1194{
7918f24d
NC
1195 PERL_ARGS_ASSERT_UTF8_HOP;
1196
96a5add6 1197 PERL_UNUSED_CONTEXT;
8850bf83
JH
1198 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g
1199 * the bitops (especially ~) can create illegal UTF-8.
1200 * In other words: in Perl UTF-8 is not just for Unicode. */
1201
a0ed51b3
LW
1202 if (off >= 0) {
1203 while (off--)
1204 s += UTF8SKIP(s);
1205 }
1206 else {
1207 while (off++) {
1208 s--;
8850bf83
JH
1209 while (UTF8_IS_CONTINUATION(*s))
1210 s--;
a0ed51b3
LW
1211 }
1212 }
4373e329 1213 return (U8 *)s;
a0ed51b3
LW
1214}
1215
6940069f 1216/*
fed3ba5d
NC
1217=for apidoc bytes_cmp_utf8
1218
a1433954
KW
1219Compares the sequence of characters (stored as octets) in C<b>, C<blen> with the
1220sequence of characters (stored as UTF-8) in C<u>, C<ulen>. Returns 0 if they are
fed3ba5d
NC
1221equal, -1 or -2 if the first string is less than the second string, +1 or +2
1222if the first string is greater than the second string.
1223
1224-1 or +1 is returned if the shorter string was identical to the start of the
1225longer string. -2 or +2 is returned if the was a difference between characters
1226within the strings.
1227
1228=cut
1229*/
1230
1231int
1232Perl_bytes_cmp_utf8(pTHX_ const U8 *b, STRLEN blen, const U8 *u, STRLEN ulen)
1233{
1234 const U8 *const bend = b + blen;
1235 const U8 *const uend = u + ulen;
1236
1237 PERL_ARGS_ASSERT_BYTES_CMP_UTF8;
1238
1239 PERL_UNUSED_CONTEXT;
1240
1241 while (b < bend && u < uend) {
1242 U8 c = *u++;
1243 if (!UTF8_IS_INVARIANT(c)) {
1244 if (UTF8_IS_DOWNGRADEABLE_START(c)) {
1245 if (u < uend) {
1246 U8 c1 = *u++;
1247 if (UTF8_IS_CONTINUATION(c1)) {
94bb8c36 1248 c = TWO_BYTE_UTF8_TO_NATIVE(c, c1);
fed3ba5d
NC
1249 } else {
1250 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8),
1251 "Malformed UTF-8 character "
1252 "(unexpected non-continuation byte 0x%02x"
1253 ", immediately after start byte 0x%02x)"
1254 /* Dear diag.t, it's in the pod. */
1255 "%s%s", c1, c,
1256 PL_op ? " in " : "",
1257 PL_op ? OP_DESC(PL_op) : "");
1258 return -2;
1259 }
1260 } else {
1261 if (PL_op)
1262 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8),
1263 "%s in %s", unees, OP_DESC(PL_op));
1264 else
61a12c31 1265 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), "%s", unees);
fed3ba5d
NC
1266 return -2; /* Really want to return undef :-) */
1267 }
1268 } else {
1269 return -2;
1270 }
1271 }
1272 if (*b != c) {
1273 return *b < c ? -2 : +2;
1274 }
1275 ++b;
1276 }
1277
1278 if (b == bend && u == uend)
1279 return 0;
1280
1281 return b < bend ? +1 : -1;
1282}
1283
1284/*
87cea99e 1285=for apidoc utf8_to_bytes
6940069f 1286
2bbc8d55 1287Converts a string C<s> of length C<len> from UTF-8 into native byte encoding.
a1433954
KW
1288Unlike L</bytes_to_utf8>, this over-writes the original string, and
1289updates C<len> to contain the new length.
67e989fb 1290Returns zero on failure, setting C<len> to -1.
6940069f 1291
a1433954 1292If you need a copy of the string, see L</bytes_from_utf8>.
95be277c 1293
6940069f
GS
1294=cut
1295*/
1296
1297U8 *
37607a96 1298Perl_utf8_to_bytes(pTHX_ U8 *s, STRLEN *len)
6940069f 1299{
d4c19fe8
AL
1300 U8 * const save = s;
1301 U8 * const send = s + *len;
6940069f 1302 U8 *d;
246fae53 1303
7918f24d
NC
1304 PERL_ARGS_ASSERT_UTF8_TO_BYTES;
1305
1e54db1a 1306 /* ensure valid UTF-8 and chars < 256 before updating string */
d4c19fe8 1307 while (s < send) {
d59937ca
KW
1308 if (! UTF8_IS_INVARIANT(*s)) {
1309 if (! UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(s, send)) {
1310 *len = ((STRLEN) -1);
1311 return 0;
1312 }
1313 s++;
dcad2880 1314 }
d59937ca 1315 s++;
246fae53 1316 }
dcad2880
JH
1317
1318 d = s = save;
6940069f 1319 while (s < send) {
80e0b38f
KW
1320 U8 c = *s++;
1321 if (! UTF8_IS_INVARIANT(c)) {
1322 /* Then it is two-byte encoded */
1323 c = TWO_BYTE_UTF8_TO_NATIVE(c, *s);
1324 s++;
1325 }
1326 *d++ = c;
6940069f
GS
1327 }
1328 *d = '\0';
246fae53 1329 *len = d - save;
6940069f
GS
1330 return save;
1331}
1332
1333/*
87cea99e 1334=for apidoc bytes_from_utf8
f9a63242 1335
2bbc8d55 1336Converts a string C<s> of length C<len> from UTF-8 into native byte encoding.
a1433954 1337Unlike L</utf8_to_bytes> but like L</bytes_to_utf8>, returns a pointer to
ef9edfd0
JH
1338the newly-created string, and updates C<len> to contain the new
1339length. Returns the original string if no conversion occurs, C<len>
1340is unchanged. Do nothing if C<is_utf8> points to 0. Sets C<is_utf8> to
2bbc8d55
SP
13410 if C<s> is converted or consisted entirely of characters that are invariant
1342in utf8 (i.e., US-ASCII on non-EBCDIC machines).
f9a63242 1343
37607a96
PK
1344=cut
1345*/
f9a63242
JH
1346
1347U8 *
e1ec3a88 1348Perl_bytes_from_utf8(pTHX_ const U8 *s, STRLEN *len, bool *is_utf8)
f9a63242 1349{
f9a63242 1350 U8 *d;
e1ec3a88
AL
1351 const U8 *start = s;
1352 const U8 *send;
f9a63242
JH
1353 I32 count = 0;
1354
7918f24d
NC
1355 PERL_ARGS_ASSERT_BYTES_FROM_UTF8;
1356
96a5add6 1357 PERL_UNUSED_CONTEXT;
f9a63242 1358 if (!*is_utf8)
73d840c0 1359 return (U8 *)start;
f9a63242 1360
1e54db1a 1361 /* ensure valid UTF-8 and chars < 256 before converting string */
f9a63242 1362 for (send = s + *len; s < send;) {
d59937ca
KW
1363 if (! UTF8_IS_INVARIANT(*s)) {
1364 if (! UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(s, send)) {
73d840c0 1365 return (U8 *)start;
d59937ca
KW
1366 }
1367 count++;
1368 s++;
db42d148 1369 }
d59937ca 1370 s++;
f9a63242
JH
1371 }
1372
35da51f7 1373 *is_utf8 = FALSE;
f9a63242 1374
212542aa 1375 Newx(d, (*len) - count + 1, U8);
ef9edfd0 1376 s = start; start = d;
f9a63242
JH
1377 while (s < send) {
1378 U8 c = *s++;
1a91c45d 1379 if (! UTF8_IS_INVARIANT(c)) {
c4d5f83a 1380 /* Then it is two-byte encoded */
1a91c45d
KW
1381 c = TWO_BYTE_UTF8_TO_NATIVE(c, *s);
1382 s++;
c4d5f83a
NIS
1383 }
1384 *d++ = c;
f9a63242
JH
1385 }
1386 *d = '\0';
1387 *len = d - start;
73d840c0 1388 return (U8 *)start;
f9a63242
JH
1389}
1390
1391/*
87cea99e 1392=for apidoc bytes_to_utf8
6940069f 1393
ff97e5cf
KW
1394Converts a string C<s> of length C<len> bytes from the native encoding into
1395UTF-8.
6662521e 1396Returns a pointer to the newly-created string, and sets C<len> to
ff97e5cf 1397reflect the new length in bytes.
6940069f 1398
2bbc8d55
SP
1399A NUL character will be written after the end of the string.
1400
1401If you want to convert to UTF-8 from encodings other than
1402the native (Latin1 or EBCDIC),
a1433954 1403see L</sv_recode_to_utf8>().
c9ada85f 1404
497711e7 1405=cut
6940069f
GS
1406*/
1407
c682ebef
FC
1408/* This logic is duplicated in sv_catpvn_flags, so any bug fixes will
1409 likewise need duplication. */
1410
6940069f 1411U8*
35a4481c 1412Perl_bytes_to_utf8(pTHX_ const U8 *s, STRLEN *len)
6940069f 1413{
35a4481c 1414 const U8 * const send = s + (*len);
6940069f
GS
1415 U8 *d;
1416 U8 *dst;
7918f24d
NC
1417
1418 PERL_ARGS_ASSERT_BYTES_TO_UTF8;
96a5add6 1419 PERL_UNUSED_CONTEXT;
6940069f 1420
212542aa 1421 Newx(d, (*len) * 2 + 1, U8);
6940069f
GS
1422 dst = d;
1423
1424 while (s < send) {
55d09dc8
KW
1425 append_utf8_from_native_byte(*s, &d);
1426 s++;
6940069f
GS
1427 }
1428 *d = '\0';
6662521e 1429 *len = d-dst;
6940069f
GS
1430 return dst;
1431}
1432
a0ed51b3 1433/*
dea0fc0b 1434 * Convert native (big-endian) or reversed (little-endian) UTF-16 to UTF-8.
a0ed51b3
LW
1435 *
1436 * Destination must be pre-extended to 3/2 source. Do not use in-place.
1437 * We optimize for native, for obvious reasons. */
1438
1439U8*
dea0fc0b 1440Perl_utf16_to_utf8(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen)
a0ed51b3 1441{
dea0fc0b
JH
1442 U8* pend;
1443 U8* dstart = d;
1444
7918f24d
NC
1445 PERL_ARGS_ASSERT_UTF16_TO_UTF8;
1446
dea0fc0b 1447 if (bytelen & 1)
f5992bc4 1448 Perl_croak(aTHX_ "panic: utf16_to_utf8: odd bytelen %"UVuf, (UV)bytelen);
dea0fc0b
JH
1449
1450 pend = p + bytelen;
1451
a0ed51b3 1452 while (p < pend) {
dea0fc0b
JH
1453 UV uv = (p[0] << 8) + p[1]; /* UTF-16BE */
1454 p += 2;
56d37426
KW
1455 if (UNI_IS_INVARIANT(uv)) {
1456 *d++ = LATIN1_TO_NATIVE((U8) uv);
a0ed51b3
LW
1457 continue;
1458 }
56d37426
KW
1459 if (uv <= MAX_UTF8_TWO_BYTE) {
1460 *d++ = UTF8_TWO_BYTE_HI(UNI_TO_NATIVE(uv));
1461 *d++ = UTF8_TWO_BYTE_LO(UNI_TO_NATIVE(uv));
a0ed51b3
LW
1462 continue;
1463 }
46956fad
KW
1464#define FIRST_HIGH_SURROGATE UNICODE_SURROGATE_FIRST
1465#define LAST_HIGH_SURROGATE 0xDBFF
1466#define FIRST_LOW_SURROGATE 0xDC00
1467#define LAST_LOW_SURROGATE UNICODE_SURROGATE_LAST
1468 if (uv >= FIRST_HIGH_SURROGATE && uv <= LAST_HIGH_SURROGATE) {
01ea242b 1469 if (p >= pend) {
dea0fc0b 1470 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
01ea242b
NC
1471 } else {
1472 UV low = (p[0] << 8) + p[1];
1473 p += 2;
46956fad 1474 if (low < FIRST_LOW_SURROGATE || low > LAST_LOW_SURROGATE)
01ea242b 1475 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
46956fad
KW
1476 uv = ((uv - FIRST_HIGH_SURROGATE) << 10)
1477 + (low - FIRST_LOW_SURROGATE) + 0x10000;
01ea242b 1478 }
46956fad 1479 } else if (uv >= FIRST_LOW_SURROGATE && uv <= LAST_LOW_SURROGATE) {
dbde1951 1480 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
a0ed51b3 1481 }
56d37426
KW
1482#ifdef EBCDIC
1483 d = uvoffuni_to_utf8_flags(d, uv, 0);
1484#else
a0ed51b3 1485 if (uv < 0x10000) {
eb160463
GS
1486 *d++ = (U8)(( uv >> 12) | 0xe0);
1487 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
1488 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
1489 continue;
1490 }
1491 else {
eb160463
GS
1492 *d++ = (U8)(( uv >> 18) | 0xf0);
1493 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
1494 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
1495 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
1496 continue;
1497 }
56d37426 1498#endif
a0ed51b3 1499 }
dea0fc0b 1500 *newlen = d - dstart;
a0ed51b3
LW
1501 return d;
1502}
1503
1504/* Note: this one is slightly destructive of the source. */
1505
1506U8*
dea0fc0b 1507Perl_utf16_to_utf8_reversed(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen)
a0ed51b3
LW
1508{
1509 U8* s = (U8*)p;
d4c19fe8 1510 U8* const send = s + bytelen;
7918f24d
NC
1511
1512 PERL_ARGS_ASSERT_UTF16_TO_UTF8_REVERSED;
1513
e0ea5e2d
NC
1514 if (bytelen & 1)
1515 Perl_croak(aTHX_ "panic: utf16_to_utf8_reversed: odd bytelen %"UVuf,
1516 (UV)bytelen);
1517
a0ed51b3 1518 while (s < send) {
d4c19fe8 1519 const U8 tmp = s[0];
a0ed51b3
LW
1520 s[0] = s[1];
1521 s[1] = tmp;
1522 s += 2;
1523 }
dea0fc0b 1524 return utf16_to_utf8(p, d, bytelen, newlen);
a0ed51b3
LW
1525}
1526
922e8cb4
KW
1527bool
1528Perl__is_uni_FOO(pTHX_ const U8 classnum, const UV c)
1529{
1530 U8 tmpbuf[UTF8_MAXBYTES+1];
1531 uvchr_to_utf8(tmpbuf, c);
1532 return _is_utf8_FOO(classnum, tmpbuf);
1533}
1534
c3fd2246
KW
1535/* for now these are all defined (inefficiently) in terms of the utf8 versions.
1536 * Note that the macros in handy.h that call these short-circuit calling them
1537 * for Latin-1 range inputs */
a0ed51b3
LW
1538
1539bool
84afefe6 1540Perl_is_uni_alnum(pTHX_ UV c)
a0ed51b3 1541{
89ebb4a3 1542 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1543 uvchr_to_utf8(tmpbuf, c);
922e8cb4 1544 return _is_utf8_FOO(_CC_WORDCHAR, tmpbuf);
a0ed51b3
LW
1545}
1546
1547bool
5092f92a
KW
1548Perl_is_uni_alnumc(pTHX_ UV c)
1549{
1550 U8 tmpbuf[UTF8_MAXBYTES+1];
1551 uvchr_to_utf8(tmpbuf, c);
922e8cb4 1552 return _is_utf8_FOO(_CC_ALPHANUMERIC, tmpbuf);
5092f92a
KW
1553}
1554
f9ae8fb6
JD
1555/* Internal function so we can deprecate the external one, and call
1556 this one from other deprecated functions in this file */
1557
1558PERL_STATIC_INLINE bool
61b19385
KW
1559S_is_utf8_idfirst(pTHX_ const U8 *p)
1560{
1561 dVAR;
1562
1563 if (*p == '_')
1564 return TRUE;
1565 /* is_utf8_idstart would be more logical. */
1566 return is_utf8_common(p, &PL_utf8_idstart, "IdStart");
1567}
1568
5092f92a 1569bool
84afefe6 1570Perl_is_uni_idfirst(pTHX_ UV c)
a0ed51b3 1571{
89ebb4a3 1572 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1573 uvchr_to_utf8(tmpbuf, c);
61b19385 1574 return S_is_utf8_idfirst(aTHX_ tmpbuf);
a0ed51b3
LW
1575}
1576
1577bool
eba68aa0
KW
1578Perl__is_uni_perl_idcont(pTHX_ UV c)
1579{
1580 U8 tmpbuf[UTF8_MAXBYTES+1];
1581 uvchr_to_utf8(tmpbuf, c);
1582 return _is_utf8_perl_idcont(tmpbuf);
1583}
1584
1585bool
f91dcd13
KW
1586Perl__is_uni_perl_idstart(pTHX_ UV c)
1587{
1588 U8 tmpbuf[UTF8_MAXBYTES+1];
1589 uvchr_to_utf8(tmpbuf, c);
1590 return _is_utf8_perl_idstart(tmpbuf);
1591}
1592
1593bool
84afefe6 1594Perl_is_uni_alpha(pTHX_ UV c)
a0ed51b3 1595{
89ebb4a3 1596 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1597 uvchr_to_utf8(tmpbuf, c);
922e8cb4 1598 return _is_utf8_FOO(_CC_ALPHA, tmpbuf);
a0ed51b3
LW
1599}
1600
1601bool
84afefe6 1602Perl_is_uni_ascii(pTHX_ UV c)
4d61ec05 1603{
bc39fe24 1604 return isASCII(c);
4d61ec05
GS
1605}
1606
1607bool
bdd8600f
KW
1608Perl_is_uni_blank(pTHX_ UV c)
1609{
2cafb56b 1610 return isBLANK_uni(c);
bdd8600f
KW
1611}
1612
1613bool
84afefe6 1614Perl_is_uni_space(pTHX_ UV c)
a0ed51b3 1615{
add4123a 1616 return isSPACE_uni(c);
a0ed51b3
LW
1617}
1618
1619bool
84afefe6 1620Perl_is_uni_digit(pTHX_ UV c)
a0ed51b3 1621{
89ebb4a3 1622 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1623 uvchr_to_utf8(tmpbuf, c);
922e8cb4 1624 return _is_utf8_FOO(_CC_DIGIT, tmpbuf);
a0ed51b3
LW
1625}
1626
1627bool
84afefe6 1628Perl_is_uni_upper(pTHX_ UV c)
a0ed51b3 1629{
89ebb4a3 1630 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1631 uvchr_to_utf8(tmpbuf, c);
922e8cb4 1632 return _is_utf8_FOO(_CC_UPPER, tmpbuf);
a0ed51b3
LW
1633}
1634
1635bool
84afefe6 1636Perl_is_uni_lower(pTHX_ UV c)
a0ed51b3 1637{
89ebb4a3 1638 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1639 uvchr_to_utf8(tmpbuf, c);
922e8cb4 1640 return _is_utf8_FOO(_CC_LOWER, tmpbuf);
a0ed51b3
LW
1641}
1642
1643bool
84afefe6 1644Perl_is_uni_cntrl(pTHX_ UV c)
b8c5462f 1645{
7b952154 1646 return isCNTRL_L1(c);
b8c5462f
JH
1647}
1648
1649bool
84afefe6 1650Perl_is_uni_graph(pTHX_ UV c)
b8c5462f 1651{
89ebb4a3 1652 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1653 uvchr_to_utf8(tmpbuf, c);
922e8cb4 1654 return _is_utf8_FOO(_CC_GRAPH, tmpbuf);
b8c5462f
JH
1655}
1656
1657bool
84afefe6 1658Perl_is_uni_print(pTHX_ UV c)
a0ed51b3 1659{
89ebb4a3 1660 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1661 uvchr_to_utf8(tmpbuf, c);
922e8cb4 1662 return _is_utf8_FOO(_CC_PRINT, tmpbuf);
a0ed51b3
LW
1663}
1664
b8c5462f 1665bool
84afefe6 1666Perl_is_uni_punct(pTHX_ UV c)
b8c5462f 1667{
89ebb4a3 1668 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1669 uvchr_to_utf8(tmpbuf, c);
922e8cb4 1670 return _is_utf8_FOO(_CC_PUNCT, tmpbuf);
b8c5462f
JH
1671}
1672
4d61ec05 1673bool
84afefe6 1674Perl_is_uni_xdigit(pTHX_ UV c)
4d61ec05 1675{
4ac6419d 1676 return isXDIGIT_uni(c);
4d61ec05
GS
1677}
1678
3a4c58c9
KW
1679UV
1680Perl__to_upper_title_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp, const char S_or_s)
1681{
1682 /* We have the latin1-range values compiled into the core, so just use
1683 * those, converting the result to utf8. The only difference between upper
1684 * and title case in this range is that LATIN_SMALL_LETTER_SHARP_S is
1685 * either "SS" or "Ss". Which one to use is passed into the routine in
1686 * 'S_or_s' to avoid a test */
1687
1688 UV converted = toUPPER_LATIN1_MOD(c);
1689
1690 PERL_ARGS_ASSERT__TO_UPPER_TITLE_LATIN1;
1691
1692 assert(S_or_s == 'S' || S_or_s == 's');
1693
6f2d5cbc 1694 if (UVCHR_IS_INVARIANT(converted)) { /* No difference between the two for
f4cd282c 1695 characters in this range */
3a4c58c9
KW
1696 *p = (U8) converted;
1697 *lenp = 1;
1698 return converted;
1699 }
1700
1701 /* toUPPER_LATIN1_MOD gives the correct results except for three outliers,
1702 * which it maps to one of them, so as to only have to have one check for
1703 * it in the main case */
1704 if (UNLIKELY(converted == LATIN_SMALL_LETTER_Y_WITH_DIAERESIS)) {
1705 switch (c) {
1706 case LATIN_SMALL_LETTER_Y_WITH_DIAERESIS:
1707 converted = LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS;
1708 break;
1709 case MICRO_SIGN:
1710 converted = GREEK_CAPITAL_LETTER_MU;
1711 break;
1712 case LATIN_SMALL_LETTER_SHARP_S:
1713 *(p)++ = 'S';
1714 *p = S_or_s;
1715 *lenp = 2;
1716 return 'S';
1717 default:
1718 Perl_croak(aTHX_ "panic: to_upper_title_latin1 did not expect '%c' to map to '%c'", c, LATIN_SMALL_LETTER_Y_WITH_DIAERESIS);
118e2215 1719 assert(0); /* NOTREACHED */
3a4c58c9
KW
1720 }
1721 }
1722
1723 *(p)++ = UTF8_TWO_BYTE_HI(converted);
1724 *p = UTF8_TWO_BYTE_LO(converted);
1725 *lenp = 2;
1726
1727 return converted;
1728}
1729
50bda2c3
KW
1730/* Call the function to convert a UTF-8 encoded character to the specified case.
1731 * Note that there may be more than one character in the result.
1732 * INP is a pointer to the first byte of the input character
1733 * OUTP will be set to the first byte of the string of changed characters. It
1734 * needs to have space for UTF8_MAXBYTES_CASE+1 bytes
1735 * LENP will be set to the length in bytes of the string of changed characters
1736 *
1737 * The functions return the ordinal of the first character in the string of OUTP */
4a8240a3
KW
1738#define CALL_UPPER_CASE(INP, OUTP, LENP) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_toupper, "ToUc", "")
1739#define CALL_TITLE_CASE(INP, OUTP, LENP) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_totitle, "ToTc", "")
1740#define CALL_LOWER_CASE(INP, OUTP, LENP) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_tolower, "ToLc", "")
50bda2c3
KW
1741
1742/* This additionally has the input parameter SPECIALS, which if non-zero will
1743 * cause this to use the SPECIALS hash for folding (meaning get full case
1744 * folding); otherwise, when zero, this implies a simple case fold */
4a8240a3 1745#define CALL_FOLD_CASE(INP, OUTP, LENP, SPECIALS) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_tofold, "ToCf", (SPECIALS) ? "" : NULL)
c3fd2246 1746
84afefe6
JH
1747UV
1748Perl_to_uni_upper(pTHX_ UV c, U8* p, STRLEN *lenp)
a0ed51b3 1749{
3a4c58c9
KW
1750 dVAR;
1751
a1433954
KW
1752 /* Convert the Unicode character whose ordinal is <c> to its uppercase
1753 * version and store that in UTF-8 in <p> and its length in bytes in <lenp>.
1754 * Note that the <p> needs to be at least UTF8_MAXBYTES_CASE+1 bytes since
c3fd2246
KW
1755 * the changed version may be longer than the original character.
1756 *
1757 * The ordinal of the first character of the changed version is returned
1758 * (but note, as explained above, that there may be more.) */
1759
7918f24d
NC
1760 PERL_ARGS_ASSERT_TO_UNI_UPPER;
1761
3a4c58c9
KW
1762 if (c < 256) {
1763 return _to_upper_title_latin1((U8) c, p, lenp, 'S');
1764 }
1765
0ebc6274 1766 uvchr_to_utf8(p, c);
3a4c58c9 1767 return CALL_UPPER_CASE(p, p, lenp);
a0ed51b3
LW
1768}
1769
84afefe6
JH
1770UV
1771Perl_to_uni_title(pTHX_ UV c, U8* p, STRLEN *lenp)
a0ed51b3 1772{
3a4c58c9
KW
1773 dVAR;
1774
7918f24d
NC
1775 PERL_ARGS_ASSERT_TO_UNI_TITLE;
1776
3a4c58c9
KW
1777 if (c < 256) {
1778 return _to_upper_title_latin1((U8) c, p, lenp, 's');
1779 }
1780
0ebc6274 1781 uvchr_to_utf8(p, c);
3a4c58c9 1782 return CALL_TITLE_CASE(p, p, lenp);
a0ed51b3
LW
1783}
1784
afc16117
KW
1785STATIC U8
1786S_to_lower_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp)
1787{
1788 /* We have the latin1-range values compiled into the core, so just use
1789 * those, converting the result to utf8. Since the result is always just
a1433954 1790 * one character, we allow <p> to be NULL */
afc16117
KW
1791
1792 U8 converted = toLOWER_LATIN1(c);
1793
1794 if (p != NULL) {
6f2d5cbc 1795 if (NATIVE_BYTE_IS_INVARIANT(converted)) {
afc16117
KW
1796 *p = converted;
1797 *lenp = 1;
1798 }
1799 else {
1800 *p = UTF8_TWO_BYTE_HI(converted);
1801 *(p+1) = UTF8_TWO_BYTE_LO(converted);
1802 *lenp = 2;
1803 }
1804 }
1805 return converted;
1806}
1807
84afefe6
JH
1808UV
1809Perl_to_uni_lower(pTHX_ UV c, U8* p, STRLEN *lenp)
a0ed51b3 1810{
968c5e6a
KW
1811 dVAR;
1812
7918f24d
NC
1813 PERL_ARGS_ASSERT_TO_UNI_LOWER;
1814
afc16117
KW
1815 if (c < 256) {
1816 return to_lower_latin1((U8) c, p, lenp);
bca00c02
KW
1817 }
1818
afc16117 1819 uvchr_to_utf8(p, c);
968c5e6a 1820 return CALL_LOWER_CASE(p, p, lenp);
a0ed51b3
LW
1821}
1822
84afefe6 1823UV
51910141 1824Perl__to_fold_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp, const unsigned int flags)
a1dde8de 1825{
51910141 1826 /* Corresponds to to_lower_latin1(); <flags> bits meanings:
1ca267a5 1827 * FOLD_FLAGS_NOMIX_ASCII iff non-ASCII to ASCII folds are prohibited
51910141 1828 * FOLD_FLAGS_FULL iff full folding is to be used;
1ca267a5
KW
1829 *
1830 * Not to be used for locale folds
51910141 1831 */
f673fad4 1832
a1dde8de
KW
1833 UV converted;
1834
1835 PERL_ARGS_ASSERT__TO_FOLD_LATIN1;
1836
1ca267a5
KW
1837 assert (! (flags & FOLD_FLAGS_LOCALE));
1838
a1dde8de
KW
1839 if (c == MICRO_SIGN) {
1840 converted = GREEK_SMALL_LETTER_MU;
1841 }
51910141 1842 else if ((flags & FOLD_FLAGS_FULL) && c == LATIN_SMALL_LETTER_SHARP_S) {
1ca267a5
KW
1843
1844 /* If can't cross 127/128 boundary, can't return "ss"; instead return
1845 * two U+017F characters, as fc("\df") should eq fc("\x{17f}\x{17f}")
1846 * under those circumstances. */
1847 if (flags & FOLD_FLAGS_NOMIX_ASCII) {
1848 *lenp = 2 * sizeof(LATIN_SMALL_LETTER_LONG_S_UTF8) - 2;
1849 Copy(LATIN_SMALL_LETTER_LONG_S_UTF8 LATIN_SMALL_LETTER_LONG_S_UTF8,
1850 p, *lenp, U8);
1851 return LATIN_SMALL_LETTER_LONG_S;
1852 }
1853 else {
4f489194
KW
1854 *(p)++ = 's';
1855 *p = 's';
1856 *lenp = 2;
1857 return 's';
1ca267a5 1858 }
a1dde8de
KW
1859 }
1860 else { /* In this range the fold of all other characters is their lower
1861 case */
1862 converted = toLOWER_LATIN1(c);
1863 }
1864
6f2d5cbc 1865 if (UVCHR_IS_INVARIANT(converted)) {
a1dde8de
KW
1866 *p = (U8) converted;
1867 *lenp = 1;
1868 }
1869 else {
1870 *(p)++ = UTF8_TWO_BYTE_HI(converted);
1871 *p = UTF8_TWO_BYTE_LO(converted);
1872 *lenp = 2;
1873 }
1874
1875 return converted;
1876}
1877
1878UV
a0270393 1879Perl__to_uni_fold_flags(pTHX_ UV c, U8* p, STRLEN *lenp, const U8 flags)
84afefe6 1880{
4b593389 1881
a0270393
KW
1882 /* Not currently externally documented, and subject to change
1883 * <flags> bits meanings:
1884 * FOLD_FLAGS_FULL iff full folding is to be used;
1885 * FOLD_FLAGS_LOCALE iff in locale
1886 * FOLD_FLAGS_NOMIX_ASCII iff non-ASCII to ASCII folds are prohibited
1887 */
4b593389 1888
36bb2ab6 1889 PERL_ARGS_ASSERT__TO_UNI_FOLD_FLAGS;
7918f24d 1890
a1dde8de 1891 if (c < 256) {
a0270393 1892 UV result = _to_fold_latin1((U8) c, p, lenp,
1ca267a5 1893 flags & (FOLD_FLAGS_FULL | FOLD_FLAGS_NOMIX_ASCII));
a0270393
KW
1894 /* It is illegal for the fold to cross the 255/256 boundary under
1895 * locale; in this case return the original */
1896 return (result > 256 && flags & FOLD_FLAGS_LOCALE)
1897 ? c
1898 : result;
a1dde8de
KW
1899 }
1900
a0270393
KW
1901 /* If no special needs, just use the macro */
1902 if ( ! (flags & (FOLD_FLAGS_LOCALE|FOLD_FLAGS_NOMIX_ASCII))) {
1903 uvchr_to_utf8(p, c);
1904 return CALL_FOLD_CASE(p, p, lenp, flags & FOLD_FLAGS_FULL);
1905 }
1906 else { /* Otherwise, _to_utf8_fold_flags has the intelligence to deal with
1907 the special flags. */
1908 U8 utf8_c[UTF8_MAXBYTES + 1];
1909 uvchr_to_utf8(utf8_c, c);
1910 return _to_utf8_fold_flags(utf8_c, p, lenp, flags, NULL);
1911 }
84afefe6
JH
1912}
1913
a0ed51b3 1914bool
84afefe6 1915Perl_is_uni_alnum_lc(pTHX_ UV c)
a0ed51b3 1916{
edfb3318 1917 if (c < 256) {
f4cd282c 1918 return isALNUM_LC(c);
edfb3318 1919 }
922e8cb4 1920 return _is_uni_FOO(_CC_WORDCHAR, c);
a0ed51b3
LW
1921}
1922
1923bool
5092f92a
KW
1924Perl_is_uni_alnumc_lc(pTHX_ UV c)
1925{
1926 if (c < 256) {
f4cd282c 1927 return isALPHANUMERIC_LC(c);
5092f92a 1928 }
922e8cb4 1929 return _is_uni_FOO(_CC_ALPHANUMERIC, c);
5092f92a
KW
1930}
1931
1932bool
84afefe6 1933Perl_is_uni_idfirst_lc(pTHX_ UV c)
a0ed51b3 1934{
edfb3318 1935 if (c < 256) {
f4cd282c 1936 return isIDFIRST_LC(c);
edfb3318
KW
1937 }
1938 return _is_uni_perl_idstart(c);
a0ed51b3
LW
1939}
1940
1941bool
84afefe6 1942Perl_is_uni_alpha_lc(pTHX_ UV c)
a0ed51b3 1943{
edfb3318 1944 if (c < 256) {
f4cd282c 1945 return isALPHA_LC(c);
edfb3318 1946 }
922e8cb4 1947 return _is_uni_FOO(_CC_ALPHA, c);
a0ed51b3
LW
1948}
1949
1950bool
84afefe6 1951Perl_is_uni_ascii_lc(pTHX_ UV c)
4d61ec05 1952{
edfb3318 1953 if (c < 256) {
f4cd282c 1954 return isASCII_LC(c);
edfb3318
KW
1955 }
1956 return 0;
4d61ec05
GS
1957}
1958
1959bool
bdd8600f
KW
1960Perl_is_uni_blank_lc(pTHX_ UV c)
1961{
edfb3318 1962 if (c < 256) {
f4cd282c 1963 return isBLANK_LC(c);
edfb3318 1964 }
61b19385 1965 return isBLANK_uni(c);
bdd8600f
KW
1966}
1967
1968bool
84afefe6 1969Perl_is_uni_space_lc(pTHX_ UV c)
a0ed51b3 1970{
edfb3318 1971 if (c < 256) {
f4cd282c 1972 return isSPACE_LC(c);
edfb3318 1973 }
61b19385 1974 return isSPACE_uni(c);
a0ed51b3
LW
1975}
1976
1977bool
84afefe6 1978Perl_is_uni_digit_lc(pTHX_ UV c)
a0ed51b3 1979{
edfb3318 1980 if (c < 256) {
f4cd282c 1981 return isDIGIT_LC(c);
edfb3318 1982 }
922e8cb4 1983 return _is_uni_FOO(_CC_DIGIT, c);
a0ed51b3
LW
1984}
1985
1986bool
84afefe6 1987Perl_is_uni_upper_lc(pTHX_ UV c)
a0ed51b3 1988{
edfb3318 1989 if (c < 256) {
f4cd282c 1990 return isUPPER_LC(c);
edfb3318 1991 }
922e8cb4 1992 return _is_uni_FOO(_CC_UPPER, c);
a0ed51b3
LW
1993}
1994
1995bool
84afefe6 1996Perl_is_uni_lower_lc(pTHX_ UV c)
a0ed51b3 1997{
edfb3318 1998 if (c < 256) {
f4cd282c 1999 return isLOWER_LC(c);
edfb3318 2000 }
922e8cb4 2001 return _is_uni_FOO(_CC_LOWER, c);
a0ed51b3
LW
2002}
2003
2004bool
84afefe6 2005Perl_is_uni_cntrl_lc(pTHX_ UV c)
b8c5462f 2006{
edfb3318 2007 if (c < 256) {
f4cd282c 2008 return isCNTRL_LC(c);
edfb3318 2009 }
61b19385 2010 return 0;
b8c5462f
JH
2011}
2012
2013bool
84afefe6 2014Perl_is_uni_graph_lc(pTHX_ UV c)
b8c5462f 2015{
edfb3318 2016 if (c < 256) {
f4cd282c 2017 return isGRAPH_LC(c);
edfb3318 2018 }
922e8cb4 2019 return _is_uni_FOO(_CC_GRAPH, c);
b8c5462f
JH
2020}
2021
2022bool
84afefe6 2023Perl_is_uni_print_lc(pTHX_ UV c)
a0ed51b3 2024{
edfb3318 2025 if (c < 256) {
f4cd282c 2026 return isPRINT_LC(c);
edfb3318 2027 }
922e8cb4 2028 return _is_uni_FOO(_CC_PRINT, c);
a0ed51b3
LW
2029}
2030
b8c5462f 2031bool
84afefe6 2032Perl_is_uni_punct_lc(pTHX_ UV c)
b8c5462f 2033{
edfb3318 2034 if (c < 256) {
f4cd282c 2035 return isPUNCT_LC(c);
edfb3318 2036 }
922e8cb4 2037 return _is_uni_FOO(_CC_PUNCT, c);
b8c5462f
JH
2038}
2039
4d61ec05 2040bool
84afefe6 2041Perl_is_uni_xdigit_lc(pTHX_ UV c)
4d61ec05 2042{
edfb3318 2043 if (c < 256) {
f4cd282c 2044 return isXDIGIT_LC(c);
edfb3318 2045 }
61b19385 2046 return isXDIGIT_uni(c);
4d61ec05
GS
2047}
2048
b7ac61fa
JH
2049U32
2050Perl_to_uni_upper_lc(pTHX_ U32 c)
2051{
ee099d14
JH
2052 /* XXX returns only the first character -- do not use XXX */
2053 /* XXX no locale support yet */
2054 STRLEN len;
89ebb4a3 2055 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
ee099d14 2056 return (U32)to_uni_upper(c, tmpbuf, &len);
b7ac61fa
JH
2057}
2058
2059U32
2060Perl_to_uni_title_lc(pTHX_ U32 c)
2061{
ee099d14
JH
2062 /* XXX returns only the first character XXX -- do not use XXX */
2063 /* XXX no locale support yet */
2064 STRLEN len;
89ebb4a3 2065 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
ee099d14 2066 return (U32)to_uni_title(c, tmpbuf, &len);
b7ac61fa
JH
2067}
2068
2069U32
2070Perl_to_uni_lower_lc(pTHX_ U32 c)
2071{
ee099d14
JH
2072 /* XXX returns only the first character -- do not use XXX */
2073 /* XXX no locale support yet */
2074 STRLEN len;
89ebb4a3 2075 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
ee099d14 2076 return (U32)to_uni_lower(c, tmpbuf, &len);
b7ac61fa
JH
2077}
2078
26483009 2079PERL_STATIC_INLINE bool
5141f98e 2080S_is_utf8_common(pTHX_ const U8 *const p, SV **swash,
bde6a22d
NC
2081 const char *const swashname)
2082{
ea317ccb
KW
2083 /* returns a boolean giving whether or not the UTF8-encoded character that
2084 * starts at <p> is in the swash indicated by <swashname>. <swash>
2085 * contains a pointer to where the swash indicated by <swashname>
2086 * is to be stored; which this routine will do, so that future calls will
2087 * look at <*swash> and only generate a swash if it is not null
2088 *
2089 * Note that it is assumed that the buffer length of <p> is enough to
2090 * contain all the bytes that comprise the character. Thus, <*p> should
2091 * have been checked before this call for mal-formedness enough to assure
2092 * that. */
2093
97aff369 2094 dVAR;
7918f24d
NC
2095
2096 PERL_ARGS_ASSERT_IS_UTF8_COMMON;
2097
492a624f 2098 /* The API should have included a length for the UTF-8 character in <p>,
28123549 2099 * but it doesn't. We therefore assume that p has been validated at least
492a624f
KW
2100 * as far as there being enough bytes available in it to accommodate the
2101 * character without reading beyond the end, and pass that number on to the
2102 * validating routine */
28123549
KW
2103 if (! is_utf8_char_buf(p, p + UTF8SKIP(p))) {
2104 if (ckWARN_d(WARN_UTF8)) {
2105 Perl_warner(aTHX_ packWARN2(WARN_DEPRECATED,WARN_UTF8),
9816f121 2106 "Passing malformed UTF-8 to \"%s\" is deprecated", swashname);
28123549
KW
2107 if (ckWARN(WARN_UTF8)) { /* This will output details as to the
2108 what the malformation is */
2109 utf8_to_uvchr_buf(p, p + UTF8SKIP(p), NULL);
2110 }
2111 }
2112 return FALSE;
2113 }
87367d5f
KW
2114 if (!*swash) {
2115 U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST;
2116 *swash = _core_swash_init("utf8", swashname, &PL_sv_undef, 1, 0, NULL, &flags);
2117 }
28123549 2118
bde6a22d
NC
2119 return swash_fetch(*swash, p, TRUE) != 0;
2120}
2121
2122bool
922e8cb4
KW
2123Perl__is_utf8_FOO(pTHX_ const U8 classnum, const U8 *p)
2124{
2125 dVAR;
2126
2127 PERL_ARGS_ASSERT__IS_UTF8_FOO;
2128
2129 assert(classnum < _FIRST_NON_SWASH_CC);
2130
2131 return is_utf8_common(p, &PL_utf8_swash_ptrs[classnum], swash_property_names[classnum]);
2132}
2133
2134bool
7fc63493 2135Perl_is_utf8_alnum(pTHX_ const U8 *p)
a0ed51b3 2136{
97aff369 2137 dVAR;
7918f24d
NC
2138
2139 PERL_ARGS_ASSERT_IS_UTF8_ALNUM;
2140
671c33bf
NC
2141 /* NOTE: "IsWord", not "IsAlnum", since Alnum is a true
2142 * descendant of isalnum(3), in other words, it doesn't
2143 * contain the '_'. --jhi */
03940dc2 2144 return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_WORDCHAR], "IsWord");
a0ed51b3
LW
2145}
2146
2147bool
5092f92a
KW
2148Perl_is_utf8_alnumc(pTHX_ const U8 *p)
2149{
2150 dVAR;
2151
2152 PERL_ARGS_ASSERT_IS_UTF8_ALNUMC;
2153
03940dc2 2154 return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_ALPHANUMERIC], "IsAlnum");
5092f92a
KW
2155}
2156
2157bool
7fc63493 2158Perl_is_utf8_idfirst(pTHX_ const U8 *p) /* The naming is historical. */
a0ed51b3 2159{
97aff369 2160 dVAR;
7918f24d
NC
2161
2162 PERL_ARGS_ASSERT_IS_UTF8_IDFIRST;
2163
61b19385 2164 return S_is_utf8_idfirst(aTHX_ p);
82686b01
JH
2165}
2166
2167bool
c11ff943
KW
2168Perl_is_utf8_xidfirst(pTHX_ const U8 *p) /* The naming is historical. */
2169{
2170 dVAR;
2171
2172 PERL_ARGS_ASSERT_IS_UTF8_XIDFIRST;
2173
2174 if (*p == '_')
2175 return TRUE;
2176 /* is_utf8_idstart would be more logical. */
2177 return is_utf8_common(p, &PL_utf8_xidstart, "XIdStart");
2178}
2179
2180bool
d65654cb 2181Perl__is_utf8_perl_idstart(pTHX_ const U8 *p)
b6912c02
KW
2182{
2183 dVAR;
2184
d65654cb 2185 PERL_ARGS_ASSERT__IS_UTF8_PERL_IDSTART;
b6912c02
KW
2186
2187 return is_utf8_common(p, &PL_utf8_perl_idstart, "_Perl_IDStart");
2188}
2189
2190bool
eba68aa0
KW
2191Perl__is_utf8_perl_idcont(pTHX_ const U8 *p)
2192{
2193 dVAR;
2194
2195 PERL_ARGS_ASSERT__IS_UTF8_PERL_IDCONT;
2196
2197 return is_utf8_common(p, &PL_utf8_perl_idcont, "_Perl_IDCont");
2198}
2199
2200
2201bool
7fc63493 2202Perl_is_utf8_idcont(pTHX_ const U8 *p)
82686b01 2203{
97aff369 2204 dVAR;
7918f24d
NC
2205
2206 PERL_ARGS_ASSERT_IS_UTF8_IDCONT;
2207
d4c19fe8 2208 return is_utf8_common(p, &PL_utf8_idcont, "IdContinue");
a0ed51b3
LW
2209}
2210
2211bool
c11ff943
KW
2212Perl_is_utf8_xidcont(pTHX_ const U8 *p)
2213{
2214 dVAR;
2215
2216 PERL_ARGS_ASSERT_IS_UTF8_XIDCONT;
2217
c11ff943
KW
2218 return is_utf8_common(p, &PL_utf8_idcont, "XIdContinue");
2219}
2220
2221bool
7fc63493 2222Perl_is_utf8_alpha(pTHX_ const U8 *p)
a0ed51b3 2223{
97aff369 2224 dVAR;
7918f24d
NC
2225
2226 PERL_ARGS_ASSERT_IS_UTF8_ALPHA;
2227
03940dc2 2228 return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_ALPHA], "IsAlpha");
a0ed51b3
LW
2229}
2230
2231bool
7fc63493 2232Perl_is_utf8_ascii(pTHX_ const U8 *p)
b8c5462f 2233{
97aff369 2234 dVAR;
7918f24d
NC
2235
2236 PERL_ARGS_ASSERT_IS_UTF8_ASCII;
2237
bc39fe24
KW
2238 /* ASCII characters are the same whether in utf8 or not. So the macro
2239 * works on both utf8 and non-utf8 representations. */
2240 return isASCII(*p);
b8c5462f
JH
2241}
2242
2243bool
bdd8600f
KW
2244Perl_is_utf8_blank(pTHX_ const U8 *p)
2245{
2246 dVAR;
2247
2248 PERL_ARGS_ASSERT_IS_UTF8_BLANK;
2249
2cafb56b 2250 return isBLANK_utf8(p);
bdd8600f
KW
2251}
2252
2253bool
7fc63493 2254Perl_is_utf8_space(pTHX_ const U8 *p)
a0ed51b3 2255{
97aff369 2256 dVAR;
7918f24d
NC
2257
2258 PERL_ARGS_ASSERT_IS_UTF8_SPACE;
2259
add4123a 2260 return isSPACE_utf8(p);
a0ed51b3
LW
2261}
2262
2263bool
d1eb3177
YO
2264Perl_is_utf8_perl_space(pTHX_ const U8 *p)
2265{
2266 dVAR;
2267
2268 PERL_ARGS_ASSERT_IS_UTF8_PERL_SPACE;
2269
c4428693
KW
2270 /* Only true if is an ASCII space-like character, and ASCII is invariant
2271 * under utf8, so can just use the macro */
2272 return isSPACE_A(*p);
d1eb3177
YO
2273}
2274
2275bool
2276Perl_is_utf8_perl_word(pTHX_ const U8 *p)
2277{
2278 dVAR;
2279
2280 PERL_ARGS_ASSERT_IS_UTF8_PERL_WORD;
2281
c4428693
KW
2282 /* Only true if is an ASCII word character, and ASCII is invariant
2283 * under utf8, so can just use the macro */
2284 return isWORDCHAR_A(*p);
d1eb3177
YO
2285}
2286
2287bool
7fc63493 2288Perl_is_utf8_digit(pTHX_ const U8 *p)
a0ed51b3 2289{
97aff369 2290 dVAR;
7918f24d
NC
2291
2292 PERL_ARGS_ASSERT_IS_UTF8_DIGIT;
2293
03940dc2 2294 return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_DIGIT], "IsDigit");
a0ed51b3
LW
2295}
2296
2297bool
d1eb3177
YO
2298Perl_is_utf8_posix_digit(pTHX_ const U8 *p)
2299{
2300 dVAR;
2301
2302 PERL_ARGS_ASSERT_IS_UTF8_POSIX_DIGIT;
2303
c4428693
KW
2304 /* Only true if is an ASCII digit character, and ASCII is invariant
2305 * under utf8, so can just use the macro */
2306 return isDIGIT_A(*p);
d1eb3177
YO
2307}
2308
2309bool
7fc63493 2310Perl_is_utf8_upper(pTHX_ const U8 *p)
a0ed51b3 2311{
97aff369 2312 dVAR;
7918f24d
NC
2313
2314 PERL_ARGS_ASSERT_IS_UTF8_UPPER;
2315
03940dc2 2316 return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_UPPER], "IsUppercase");
a0ed51b3
LW
2317}
2318
2319bool
7fc63493 2320Perl_is_utf8_lower(pTHX_ const U8 *p)
a0ed51b3 2321{
97aff369 2322 dVAR;
7918f24d
NC
2323
2324 PERL_ARGS_ASSERT_IS_UTF8_LOWER;
2325
03940dc2 2326 return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_LOWER], "IsLowercase");
a0ed51b3
LW
2327}
2328
2329bool
7fc63493 2330Perl_is_utf8_cntrl(pTHX_ const U8 *p)
b8c5462f 2331{
97aff369 2332 dVAR;
7918f24d
NC
2333
2334 PERL_ARGS_ASSERT_IS_UTF8_CNTRL;
2335
a35d759a 2336 return isCNTRL_utf8(p);
b8c5462f
JH
2337}
2338
2339bool
7fc63493 2340Perl_is_utf8_graph(pTHX_ const U8 *p)
b8c5462f 2341{
97aff369 2342 dVAR;
7918f24d
NC
2343
2344 PERL_ARGS_ASSERT_IS_UTF8_GRAPH;
2345
03940dc2 2346 return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_GRAPH], "IsGraph");
b8c5462f
JH
2347}
2348
2349bool
7fc63493 2350Perl_is_utf8_print(pTHX_ const U8 *p)
a0ed51b3 2351{
97aff369 2352 dVAR;
7918f24d
NC
2353
2354 PERL_ARGS_ASSERT_IS_UTF8_PRINT;
2355
03940dc2 2356 return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_PRINT], "IsPrint");
a0ed51b3
LW
2357}
2358
2359bool
7fc63493 2360Perl_is_utf8_punct(pTHX_ const U8 *p)
b8c5462f 2361{
97aff369 2362 dVAR;
7918f24d
NC
2363
2364 PERL_ARGS_ASSERT_IS_UTF8_PUNCT;
2365
03940dc2 2366 return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_PUNCT], "IsPunct");
b8c5462f
JH
2367}
2368
2369bool
7fc63493 2370Perl_is_utf8_xdigit(pTHX_ const U8 *p)
b8c5462f 2371{
97aff369 2372 dVAR;
7918f24d
NC
2373
2374 PERL_ARGS_ASSERT_IS_UTF8_XDIGIT;
2375
4ac6419d 2376 return is_XDIGIT_utf8(p);
b8c5462f
JH
2377}
2378
2379bool
7dbf68d2
KW
2380Perl__is_utf8_mark(pTHX_ const U8 *p)
2381{
2382 dVAR;
2383
2384 PERL_ARGS_ASSERT__IS_UTF8_MARK;
2385
2386 return is_utf8_common(p, &PL_utf8_mark, "IsM");
2387}
2388
2389
2390bool
7fc63493 2391Perl_is_utf8_mark(pTHX_ const U8 *p)
a0ed51b3 2392{
97aff369 2393 dVAR;
7918f24d
NC
2394
2395 PERL_ARGS_ASSERT_IS_UTF8_MARK;
2396
7dbf68d2 2397 return _is_utf8_mark(p);
a0ed51b3
LW
2398}
2399
6b5c0936 2400/*
87cea99e 2401=for apidoc to_utf8_case
6b5c0936 2402
6fae5207 2403C<p> contains the pointer to the UTF-8 string encoding
a1433954
KW
2404the character that is being converted. This routine assumes that the character
2405at C<p> is well-formed.
6b5c0936 2406
6fae5207
KW
2407C<ustrp> is a pointer to the character buffer to put the
2408conversion result to. C<lenp> is a pointer to the length
6b5c0936
JH
2409of the result.
2410
6fae5207 2411C<swashp> is a pointer to the swash to use.
6b5c0936 2412
a1433954 2413Both the special and normal mappings are stored in F<lib/unicore/To/Foo.pl>,
6fae5207 2414and loaded by SWASHNEW, using F<lib/utf8_heavy.pl>. C<special> (usually,
0134edef 2415but not always, a multicharacter mapping), is tried first.
6b5c0936 2416
4a8240a3
KW
2417C<special> is a string, normally C<NULL> or C<"">. C<NULL> means to not use
2418any special mappings; C<""> means to use the special mappings. Values other
2419than these two are treated as the name of the hash containing the special
2420mappings, like C<"utf8::ToSpecLower">.
6b5c0936 2421
6fae5207 2422C<normal> is a string like "ToLower" which means the swash
0134edef
JH
2423%utf8::ToLower.
2424
2425=cut */
6b5c0936 2426
2104c8d9 2427UV
9a957fbc
AL
2428Perl_to_utf8_case(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp,
2429 SV **swashp, const char *normal, const char *special)
a0ed51b3 2430{
97aff369 2431 dVAR;
0134edef 2432 STRLEN len = 0;
f4cd282c 2433 const UV uv1 = valid_utf8_to_uvchr(p, NULL);
7918f24d
NC
2434
2435 PERL_ARGS_ASSERT_TO_UTF8_CASE;
2436
9ae3ac1a
KW
2437 /* Note that swash_fetch() doesn't output warnings for these because it
2438 * assumes we will */
8457b38f 2439 if (uv1 >= UNICODE_SURROGATE_FIRST) {
9ae3ac1a 2440 if (uv1 <= UNICODE_SURROGATE_LAST) {
8457b38f
KW
2441 if (ckWARN_d(WARN_SURROGATE)) {
2442 const char* desc = (PL_op) ? OP_DESC(PL_op) : normal;
2443 Perl_warner(aTHX_ packWARN(WARN_SURROGATE),
2444 "Operation \"%s\" returns its argument for UTF-16 surrogate U+%04"UVXf"", desc, uv1);
2445 }
9ae3ac1a
KW
2446 }
2447 else if (UNICODE_IS_SUPER(uv1)) {
8457b38f
KW
2448 if (ckWARN_d(WARN_NON_UNICODE)) {
2449 const char* desc = (PL_op) ? OP_DESC(PL_op) : normal;
2450 Perl_warner(aTHX_ packWARN(WARN_NON_UNICODE),
2451 "Operation \"%s\" returns its argument for non-Unicode code point 0x%04"UVXf"", desc, uv1);
2452 }
9ae3ac1a
KW
2453 }
2454
2455 /* Note that non-characters are perfectly legal, so no warning should
2456 * be given */
2457 }
2458
0134edef 2459 if (!*swashp) /* load on-demand */
5ab9d2ef 2460 *swashp = _core_swash_init("utf8", normal, &PL_sv_undef, 4, 0, NULL, NULL);
0134edef 2461
a6f87d8c 2462 if (special) {
0134edef 2463 /* It might be "special" (sometimes, but not always,
2a37f04d 2464 * a multicharacter mapping) */
4a8240a3 2465 HV *hv = NULL;
b08cf34e
JH
2466 SV **svp;
2467
4a8240a3
KW
2468 /* If passed in the specials name, use that; otherwise use any
2469 * given in the swash */
2470 if (*special != '\0') {
2471 hv = get_hv(special, 0);
2472 }
2473 else {
2474 svp = hv_fetchs(MUTABLE_HV(SvRV(*swashp)), "SPECIALS", 0);
2475 if (svp) {
2476 hv = MUTABLE_HV(SvRV(*svp));
2477 }
2478 }
2479
35da51f7 2480 if (hv &&
f4cd282c 2481 (svp = hv_fetch(hv, (const char*)p, UNISKIP(uv1), FALSE)) &&
b08cf34e 2482 (*svp)) {
cfd0369c 2483 const char *s;
47654450 2484
cfd0369c 2485 s = SvPV_const(*svp, len);
47654450 2486 if (len == 1)
f4cd282c 2487 /* EIGHTBIT */
c80e42f3 2488 len = uvchr_to_utf8(ustrp, *(U8*)s) - ustrp;
2a37f04d 2489 else {
d2dcd0fb 2490 Copy(s, ustrp, len, U8);
29e98929 2491 }
983ffd37 2492 }
0134edef
JH
2493 }
2494
2495 if (!len && *swashp) {
f4cd282c 2496 const UV uv2 = swash_fetch(*swashp, p, TRUE /* => is utf8 */);
d4c19fe8 2497
0134edef
JH
2498 if (uv2) {
2499 /* It was "normal" (a single character mapping). */
f4cd282c 2500 len = uvchr_to_utf8(ustrp, uv2) - ustrp;
2a37f04d
JH
2501 }
2502 }
1feea2c7 2503
cbe07460
KW
2504 if (len) {
2505 if (lenp) {
2506 *lenp = len;
2507 }
2508 return valid_utf8_to_uvchr(ustrp, 0);
2509 }
2510
2511 /* Here, there was no mapping defined, which means that the code point maps
2512 * to itself. Return the inputs */
bfdf22ec 2513 len = UTF8SKIP(p);
ca9fab46
KW
2514 if (p != ustrp) { /* Don't copy onto itself */
2515 Copy(p, ustrp, len, U8);
2516 }
0134edef 2517
2a37f04d
JH
2518 if (lenp)
2519 *lenp = len;
2520
f4cd282c 2521 return uv1;
cbe07460 2522
a0ed51b3
LW
2523}
2524
051a06d4
KW
2525STATIC UV
2526S_check_locale_boundary_crossing(pTHX_ const U8* const p, const UV result, U8* const ustrp, STRLEN *lenp)
2527{
2528 /* This is called when changing the case of a utf8-encoded character above
2529 * the Latin1 range, and the operation is in locale. If the result
2530 * contains a character that crosses the 255/256 boundary, disallow the
2531 * change, and return the original code point. See L<perlfunc/lc> for why;
2532 *
a1433954
KW
2533 * p points to the original string whose case was changed; assumed
2534 * by this routine to be well-formed
051a06d4
KW
2535 * result the code point of the first character in the changed-case string
2536 * ustrp points to the changed-case string (<result> represents its first char)
2537 * lenp points to the length of <ustrp> */
2538
2539 UV original; /* To store the first code point of <p> */
2540
2541 PERL_ARGS_ASSERT_CHECK_LOCALE_BOUNDARY_CROSSING;
2542
a4f12ed7 2543 assert(UTF8_IS_ABOVE_LATIN1(*p));
051a06d4
KW
2544
2545 /* We know immediately if the first character in the string crosses the
2546 * boundary, so can skip */
2547 if (result > 255) {
2548
2549 /* Look at every character in the result; if any cross the
2550 * boundary, the whole thing is disallowed */
2551 U8* s = ustrp + UTF8SKIP(ustrp);
2552 U8* e = ustrp + *lenp;
2553 while (s < e) {
a4f12ed7 2554 if (! UTF8_IS_ABOVE_LATIN1(*s)) {
051a06d4
KW
2555 goto bad_crossing;
2556 }
2557 s += UTF8SKIP(s);
2558 }
2559
2560 /* Here, no characters crossed, result is ok as-is */
2561 return result;
2562 }
2563
2564bad_crossing:
2565
2566 /* Failed, have to return the original */
4b88fb76 2567 original = valid_utf8_to_uvchr(p, lenp);
051a06d4
KW
2568 Copy(p, ustrp, *lenp, char);
2569 return original;
2570}
2571
d3e79532 2572/*
87cea99e 2573=for apidoc to_utf8_upper
d3e79532 2574
1f607577 2575Instead use L</toUPPER_utf8>.
a1433954 2576
d3e79532
JH
2577=cut */
2578
051a06d4
KW
2579/* Not currently externally documented, and subject to change:
2580 * <flags> is set iff locale semantics are to be used for code points < 256
2581 * <tainted_ptr> if non-null, *tainted_ptr will be set TRUE iff locale rules
2582 * were used in the calculation; otherwise unchanged. */
2583
2104c8d9 2584UV
051a06d4 2585Perl__to_utf8_upper_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, const bool flags, bool* tainted_ptr)
a0ed51b3 2586{
97aff369 2587 dVAR;
7918f24d 2588
051a06d4
KW
2589 UV result;
2590
2591 PERL_ARGS_ASSERT__TO_UTF8_UPPER_FLAGS;
7918f24d 2592
3a4c58c9 2593 if (UTF8_IS_INVARIANT(*p)) {
051a06d4
KW
2594 if (flags) {
2595 result = toUPPER_LC(*p);
2596 }
2597 else {
81c6c7ce 2598 return _to_upper_title_latin1(*p, ustrp, lenp, 'S');
051a06d4 2599 }
3a4c58c9
KW
2600 }
2601 else if UTF8_IS_DOWNGRADEABLE_START(*p) {
051a06d4 2602 if (flags) {
a6d8b88b 2603 U8 c = TWO_BYTE_UTF8_TO_NATIVE(*p, *(p+1));
68067e4e 2604 result = toUPPER_LC(c);
051a06d4
KW
2605 }
2606 else {
94bb8c36 2607 return _to_upper_title_latin1(TWO_BYTE_UTF8_TO_NATIVE(*p, *(p+1)),
81c6c7ce 2608 ustrp, lenp, 'S');
051a06d4
KW
2609 }
2610 }
2611 else { /* utf8, ord above 255 */
2612 result = CALL_UPPER_CASE(p, ustrp, lenp);
2613
2614 if (flags) {
2615 result = check_locale_boundary_crossing(p, result, ustrp, lenp);
2616 }
2617 return result;
2618 }
2619
2620 /* Here, used locale rules. Convert back to utf8 */
2621 if (UTF8_IS_INVARIANT(result)) {
2622 *ustrp = (U8) result;
2623 *lenp = 1;
2624 }
2625 else {
62cb07ea
KW
2626 *ustrp = UTF8_EIGHT_BIT_HI((U8) result);
2627 *(ustrp + 1) = UTF8_EIGHT_BIT_LO((U8) result);
051a06d4 2628 *lenp = 2;
3a4c58c9
KW
2629 }
2630
051a06d4
KW
2631 if (tainted_ptr) {
2632 *tainted_ptr = TRUE;
2633 }
2634 return result;
983ffd37 2635}
a0ed51b3 2636
d3e79532 2637/*
87cea99e 2638=for apidoc to_utf8_title
d3e79532 2639
1f607577 2640Instead use L</toTITLE_utf8>.
a1433954 2641
d3e79532
JH
2642=cut */
2643
051a06d4
KW
2644/* Not currently externally documented, and subject to change:
2645 * <flags> is set iff locale semantics are to be used for code points < 256
2646 * Since titlecase is not defined in POSIX, uppercase is used instead
2647 * for these/
2648 * <tainted_ptr> if non-null, *tainted_ptr will be set TRUE iff locale rules
2649 * were used in the calculation; otherwise unchanged. */
2650
983ffd37 2651UV
051a06d4 2652Perl__to_utf8_title_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, const bool flags, bool* tainted_ptr)
983ffd37 2653{
97aff369 2654 dVAR;
7918f24d 2655
051a06d4
KW
2656 UV result;
2657
2658 PERL_ARGS_ASSERT__TO_UTF8_TITLE_FLAGS;
7918f24d 2659
3a4c58c9 2660 if (UTF8_IS_INVARIANT(*p)) {
051a06d4
KW
2661 if (flags) {
2662 result = toUPPER_LC(*p);
2663 }
2664 else {
81c6c7ce 2665 return _to_upper_title_latin1(*p, ustrp, lenp, 's');
051a06d4 2666 }
3a4c58c9
KW
2667 }
2668 else if UTF8_IS_DOWNGRADEABLE_START(*p) {
051a06d4 2669 if (flags) {
a6d8b88b 2670 U8 c = TWO_BYTE_UTF8_TO_NATIVE(*p, *(p+1));
68067e4e 2671 result = toUPPER_LC(c);
051a06d4
KW
2672 }
2673 else {
94bb8c36 2674 return _to_upper_title_latin1(TWO_BYTE_UTF8_TO_NATIVE(*p, *(p+1)),
81c6c7ce 2675 ustrp, lenp, 's');
051a06d4
KW
2676 }
2677 }
2678 else { /* utf8, ord above 255 */
2679 result = CALL_TITLE_CASE(p, ustrp, lenp);
2680
2681 if (flags) {
2682 result = check_locale_boundary_crossing(p, result, ustrp, lenp);
2683 }
2684 return result;
2685 }
2686
2687 /* Here, used locale rules. Convert back to utf8 */
2688 if (UTF8_IS_INVARIANT(result)) {
2689 *ustrp = (U8) result;
2690 *lenp = 1;
2691 }
2692 else {
62cb07ea
KW
2693 *ustrp = UTF8_EIGHT_BIT_HI((U8) result);
2694 *(ustrp + 1) = UTF8_EIGHT_BIT_LO((U8) result);
051a06d4 2695 *lenp = 2;
3a4c58c9
KW
2696 }
2697
051a06d4
KW
2698 if (tainted_ptr) {
2699 *tainted_ptr = TRUE;
2700 }
2701 return result;
a0ed51b3
LW
2702}
2703
d3e79532 2704/*
87cea99e 2705=for apidoc to_utf8_lower
d3e79532 2706
1f607577 2707Instead use L</toLOWER_utf8>.
a1433954 2708
d3e79532
JH
2709=cut */
2710
051a06d4
KW
2711/* Not currently externally documented, and subject to change:
2712 * <flags> is set iff locale semantics are to be used for code points < 256
2713 * <tainted_ptr> if non-null, *tainted_ptr will be set TRUE iff locale rules
2714 * were used in the calculation; otherwise unchanged. */
2715
2104c8d9 2716UV
051a06d4 2717Perl__to_utf8_lower_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, const bool flags, bool* tainted_ptr)
a0ed51b3 2718{
051a06d4
KW
2719 UV result;
2720
97aff369 2721 dVAR;
7918f24d 2722
051a06d4 2723 PERL_ARGS_ASSERT__TO_UTF8_LOWER_FLAGS;
7918f24d 2724
968c5e6a 2725 if (UTF8_IS_INVARIANT(*p)) {
051a06d4
KW
2726 if (flags) {
2727 result = toLOWER_LC(*p);
2728 }
2729 else {
81c6c7ce 2730 return to_lower_latin1(*p, ustrp, lenp);
051a06d4 2731 }
968c5e6a
KW
2732 }
2733 else if UTF8_IS_DOWNGRADEABLE_START(*p) {
051a06d4 2734 if (flags) {
a6d8b88b 2735 U8 c = TWO_BYTE_UTF8_TO_NATIVE(*p, *(p+1));
68067e4e 2736 result = toLOWER_LC(c);
051a06d4
KW
2737 }
2738 else {
94bb8c36 2739 return to_lower_latin1(TWO_BYTE_UTF8_TO_NATIVE(*p, *(p+1)),
81c6c7ce 2740 ustrp, lenp);
051a06d4 2741 }
968c5e6a 2742 }
051a06d4
KW
2743 else { /* utf8, ord above 255 */
2744 result = CALL_LOWER_CASE(p, ustrp, lenp);
2745
2746 if (flags) {
2747 result = check_locale_boundary_crossing(p, result, ustrp, lenp);
2748 }
968c5e6a 2749
051a06d4
KW
2750 return result;
2751 }
2752
2753 /* Here, used locale rules. Convert back to utf8 */
2754 if (UTF8_IS_INVARIANT(result)) {
2755 *ustrp = (U8) result;
2756 *lenp = 1;
2757 }
2758 else {
62cb07ea
KW
2759 *ustrp = UTF8_EIGHT_BIT_HI((U8) result);
2760 *(ustrp + 1) = UTF8_EIGHT_BIT_LO((U8) result);
051a06d4
KW
2761 *lenp = 2;
2762 }
2763
2764 if (tainted_ptr) {
2765 *tainted_ptr = TRUE;
2766 }
2767 return result;
b4e400f9
JH
2768}
2769
d3e79532 2770/*
87cea99e 2771=for apidoc to_utf8_fold
d3e79532 2772
1f607577 2773Instead use L</toFOLD_utf8>.
a1433954 2774
d3e79532
JH
2775=cut */
2776
051a06d4
KW
2777/* Not currently externally documented, and subject to change,
2778 * in <flags>
2779 * bit FOLD_FLAGS_LOCALE is set iff locale semantics are to be used for code
2780 * points < 256. Since foldcase is not defined in
2781 * POSIX, lowercase is used instead
2782 * bit FOLD_FLAGS_FULL is set iff full case folds are to be used;
2783 * otherwise simple folds
a0270393
KW
2784 * bit FOLD_FLAGS_NOMIX_ASCII is set iff folds of non-ASCII to ASCII are
2785 * prohibited
051a06d4
KW
2786 * <tainted_ptr> if non-null, *tainted_ptr will be set TRUE iff locale rules
2787 * were used in the calculation; otherwise unchanged. */
36bb2ab6 2788
b4e400f9 2789UV
051a06d4 2790Perl__to_utf8_fold_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, U8 flags, bool* tainted_ptr)
b4e400f9 2791{
97aff369 2792 dVAR;
7918f24d 2793
051a06d4
KW
2794 UV result;
2795
36bb2ab6 2796 PERL_ARGS_ASSERT__TO_UTF8_FOLD_FLAGS;
7918f24d 2797
a0270393
KW
2798 /* These are mutually exclusive */
2799 assert (! ((flags & FOLD_FLAGS_LOCALE) && (flags & FOLD_FLAGS_NOMIX_ASCII)));
2800
50ba90ff
KW
2801 assert(p != ustrp); /* Otherwise overwrites */
2802
a1dde8de 2803 if (UTF8_IS_INVARIANT(*p)) {
051a06d4 2804 if (flags & FOLD_FLAGS_LOCALE) {
d22b930b 2805 result = toFOLD_LC(*p);
051a06d4
KW
2806 }
2807 else {
81c6c7ce 2808 return _to_fold_latin1(*p, ustrp, lenp,
1ca267a5 2809 flags & (FOLD_FLAGS_FULL | FOLD_FLAGS_NOMIX_ASCII));
051a06d4 2810 }
a1dde8de
KW
2811 }
2812 else if UTF8_IS_DOWNGRADEABLE_START(*p) {
051a06d4 2813 if (flags & FOLD_FLAGS_LOCALE) {
a6d8b88b 2814 U8 c = TWO_BYTE_UTF8_TO_NATIVE(*p, *(p+1));
68067e4e 2815 result = toFOLD_LC(c);
051a06d4
KW
2816 }
2817 else {
94bb8c36 2818 return _to_fold_latin1(TWO_BYTE_UTF8_TO_NATIVE(*p, *(p+1)),
51910141 2819 ustrp, lenp,
1ca267a5 2820 flags & (FOLD_FLAGS_FULL | FOLD_FLAGS_NOMIX_ASCII));
051a06d4 2821 }
a1dde8de 2822 }
051a06d4 2823 else { /* utf8, ord above 255 */
a0270393 2824 result = CALL_FOLD_CASE(p, ustrp, lenp, flags & FOLD_FLAGS_FULL);
a1dde8de 2825
1ca267a5
KW
2826 if (flags & FOLD_FLAGS_LOCALE) {
2827
9fc2026f 2828 /* Special case these characters, as what normally gets returned
1ca267a5
KW
2829 * under locale doesn't work */
2830 if (UTF8SKIP(p) == sizeof(LATIN_CAPITAL_LETTER_SHARP_S_UTF8) - 1
2831 && memEQ((char *) p, LATIN_CAPITAL_LETTER_SHARP_S_UTF8,
2832 sizeof(LATIN_CAPITAL_LETTER_SHARP_S_UTF8) - 1))
2833 {
2834 goto return_long_s;
2835 }
9fc2026f
KW
2836 else if (UTF8SKIP(p) == sizeof(LATIN_SMALL_LIGATURE_LONG_S_T) - 1
2837 && memEQ((char *) p, LATIN_SMALL_LIGATURE_LONG_S_T_UTF8,
2838 sizeof(LATIN_SMALL_LIGATURE_LONG_S_T_UTF8) - 1))
2839 {
2840 goto return_ligature_st;
2841 }
a0270393 2842 return check_locale_boundary_crossing(p, result, ustrp, lenp);
051a06d4 2843 }
a0270393
KW
2844 else if (! (flags & FOLD_FLAGS_NOMIX_ASCII)) {
2845 return result;
2846 }
2847 else {
2848 /* This is called when changing the case of a utf8-encoded
9fc2026f
KW
2849 * character above the ASCII range, and the result should not
2850 * contain an ASCII character. */
a0270393
KW
2851
2852 UV original; /* To store the first code point of <p> */
2853
2854 /* Look at every character in the result; if any cross the
2855 * boundary, the whole thing is disallowed */
2856 U8* s = ustrp;
2857 U8* e = ustrp + *lenp;
2858 while (s < e) {
2859 if (isASCII(*s)) {
2860 /* Crossed, have to return the original */
2861 original = valid_utf8_to_uvchr(p, lenp);
1ca267a5 2862
9fc2026f 2863 /* But in these instances, there is an alternative we can
1ca267a5 2864 * return that is valid */
9fc2026f
KW
2865 if (original == LATIN_CAPITAL_LETTER_SHARP_S
2866 || original == LATIN_SMALL_LETTER_SHARP_S)
2867 {
1ca267a5
KW
2868 goto return_long_s;
2869 }
9fc2026f
KW
2870 else if (original == LATIN_SMALL_LIGATURE_LONG_S_T) {
2871 goto return_ligature_st;
2872 }
a0270393
KW
2873 Copy(p, ustrp, *lenp, char);
2874 return original;
2875 }
2876 s += UTF8SKIP(s);
2877 }
051a06d4 2878
a0270393
KW
2879 /* Here, no characters crossed, result is ok as-is */
2880 return result;
2881 }
051a06d4
KW
2882 }
2883
2884 /* Here, used locale rules. Convert back to utf8 */
2885 if (UTF8_IS_INVARIANT(result)) {
2886 *ustrp = (U8) result;
2887 *lenp = 1;
2888 }
2889 else {
62cb07ea
KW
2890 *ustrp = UTF8_EIGHT_BIT_HI((U8) result);
2891 *(ustrp + 1) = UTF8_EIGHT_BIT_LO((U8) result);
051a06d4
KW
2892 *lenp = 2;
2893 }
2894
2895 if (tainted_ptr) {
2896 *tainted_ptr = TRUE;
2897 }
2898 return result;
1ca267a5
KW
2899
2900 return_long_s:
2901 /* Certain folds to 'ss' are prohibited by the options, but they do allow
2902 * folds to a string of two of these characters. By returning this
2903 * instead, then, e.g.,
2904 * fc("\x{1E9E}") eq fc("\x{17F}\x{17F}")
2905 * works. */
2906
2907 *lenp = 2 * sizeof(LATIN_SMALL_LETTER_LONG_S_UTF8) - 2;
2908 Copy(LATIN_SMALL_LETTER_LONG_S_UTF8 LATIN_SMALL_LETTER_LONG_S_UTF8,
2909 ustrp, *lenp, U8);
2910 return LATIN_SMALL_LETTER_LONG_S;
9fc2026f
KW
2911
2912 return_ligature_st:
2913 /* Two folds to 'st' are prohibited by the options; instead we pick one and
2914 * have the other one fold to it */
2915
2916 *lenp = sizeof(LATIN_SMALL_LIGATURE_ST_UTF8) - 1;
2917 Copy(LATIN_SMALL_LIGATURE_ST_UTF8, ustrp, *lenp, U8);
2918 return LATIN_SMALL_LIGATURE_ST;
a0ed51b3
LW
2919}
2920
711a919c 2921/* Note:
f90a9a02 2922 * Returns a "swash" which is a hash described in utf8.c:Perl_swash_fetch().
711a919c
ST
2923 * C<pkg> is a pointer to a package name for SWASHNEW, should be "utf8".
2924 * For other parameters, see utf8::SWASHNEW in lib/utf8_heavy.pl.
2925 */
c4a5db0c 2926
a0ed51b3 2927SV*
7fc63493 2928Perl_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv, I32 minbits, I32 none)
a0ed51b3 2929{
c4a5db0c
KW
2930 PERL_ARGS_ASSERT_SWASH_INIT;
2931
2932 /* Returns a copy of a swash initiated by the called function. This is the
2933 * public interface, and returning a copy prevents others from doing
2934 * mischief on the original */
2935
5d3d13d1 2936 return newSVsv(_core_swash_init(pkg, name, listsv, minbits, none, NULL, NULL));
c4a5db0c
KW
2937}
2938
2939SV*
5d3d13d1 2940Perl__core_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv, I32 minbits, I32 none, SV* invlist, U8* const flags_p)
c4a5db0c
KW
2941{
2942 /* Initialize and return a swash, creating it if necessary. It does this
87367d5f
KW
2943 * by calling utf8_heavy.pl in the general case. The returned value may be
2944 * the swash's inversion list instead if the input parameters allow it.
2945 * Which is returned should be immaterial to callers, as the only
923b6d4e
KW
2946 * operations permitted on a swash, swash_fetch(), _get_swash_invlist(),
2947 * and swash_to_invlist() handle both these transparently.
c4a5db0c
KW
2948 *
2949 * This interface should only be used by functions that won't destroy or
2950 * adversely change the swash, as doing so affects all other uses of the
2951 * swash in the program; the general public should use 'Perl_swash_init'
2952 * instead.
2953 *
2954 * pkg is the name of the package that <name> should be in.
2955 * name is the name of the swash to find. Typically it is a Unicode
2956 * property name, including user-defined ones
2957 * listsv is a string to initialize the swash with. It must be of the form
2958 * documented as the subroutine return value in
2959 * L<perlunicode/User-Defined Character Properties>
2960 * minbits is the number of bits required to represent each data element.
2961 * It is '1' for binary properties.
2962 * none I (khw) do not understand this one, but it is used only in tr///.
9a53f6cf 2963 * invlist is an inversion list to initialize the swash with (or NULL)
83199d38
KW
2964 * flags_p if non-NULL is the address of various input and output flag bits
2965 * to the routine, as follows: ('I' means is input to the routine;
2966 * 'O' means output from the routine. Only flags marked O are
2967 * meaningful on return.)
2968 * _CORE_SWASH_INIT_USER_DEFINED_PROPERTY indicates if the swash
2969 * came from a user-defined property. (I O)
5d3d13d1
KW
2970 * _CORE_SWASH_INIT_RETURN_IF_UNDEF indicates that instead of croaking
2971 * when the swash cannot be located, to simply return NULL. (I)
87367d5f
KW
2972 * _CORE_SWASH_INIT_ACCEPT_INVLIST indicates that the caller will accept a
2973 * return of an inversion list instead of a swash hash if this routine
2974 * thinks that would result in faster execution of swash_fetch() later
2975 * on. (I)
9a53f6cf
KW
2976 *
2977 * Thus there are three possible inputs to find the swash: <name>,
2978 * <listsv>, and <invlist>. At least one must be specified. The result
2979 * will be the union of the specified ones, although <listsv>'s various
2980 * actions can intersect, etc. what <name> gives.
2981 *
2982 * <invlist> is only valid for binary properties */
c4a5db0c 2983
27da23d5 2984 dVAR;
c4a5db0c 2985 SV* retval = &PL_sv_undef;
83199d38 2986 HV* swash_hv = NULL;
87367d5f
KW
2987 const int invlist_swash_boundary =
2988 (flags_p && *flags_p & _CORE_SWASH_INIT_ACCEPT_INVLIST)
2989 ? 512 /* Based on some benchmarking, but not extensive, see commit
2990 message */
2991 : -1; /* Never return just an inversion list */
9a53f6cf
KW
2992
2993 assert(listsv != &PL_sv_undef || strNE(name, "") || invlist);
2994 assert(! invlist || minbits == 1);
2995
2996 /* If data was passed in to go out to utf8_heavy to find the swash of, do
2997 * so */
2998 if (listsv != &PL_sv_undef || strNE(name, "")) {
69794297
KW
2999 dSP;
3000 const size_t pkg_len = strlen(pkg);
3001 const size_t name_len = strlen(name);
3002 HV * const stash = gv_stashpvn(pkg, pkg_len, 0);
3003 SV* errsv_save;
3004 GV *method;
3005
3006 PERL_ARGS_ASSERT__CORE_SWASH_INIT;
3007
3008 PUSHSTACKi(PERLSI_MAGIC);
ce3b816e 3009 ENTER;
69794297
KW
3010 SAVEHINTS();
3011 save_re_context();
650f067c
JL
3012 /* We might get here via a subroutine signature which uses a utf8
3013 * parameter name, at which point PL_subname will have been set
3014 * but not yet used. */
3015 save_item(PL_subname);
69794297
KW
3016 if (PL_parser && PL_parser->error_count)
3017 SAVEI8(PL_parser->error_count), PL_parser->error_count = 0;
3018 method = gv_fetchmeth(stash, "SWASHNEW", 8, -1);
3019 if (!method) { /* demand load utf8 */
3020 ENTER;
db2c6cb3
FC
3021 if ((errsv_save = GvSV(PL_errgv))) SAVEFREESV(errsv_save);
3022 GvSV(PL_errgv) = NULL;
69794297
KW
3023 /* It is assumed that callers of this routine are not passing in
3024 * any user derived data. */
3025 /* Need to do this after save_re_context() as it will set
3026 * PL_tainted to 1 while saving $1 etc (see the code after getrx:
3027 * in Perl_magic_get). Even line to create errsv_save can turn on
3028 * PL_tainted. */
284167a5
SM
3029#ifndef NO_TAINT_SUPPORT
3030 SAVEBOOL(TAINT_get);
3031 TAINT_NOT;
3032#endif
69794297
KW
3033 Perl_load_module(aTHX_ PERL_LOADMOD_NOIMPORT, newSVpvn(pkg,pkg_len),
3034 NULL);
eed484f9 3035 {
db2c6cb3
FC
3036 /* Not ERRSV, as there is no need to vivify a scalar we are
3037 about to discard. */
3038 SV * const errsv = GvSV(PL_errgv);
3039 if (!SvTRUE(errsv)) {
3040 GvSV(PL_errgv) = SvREFCNT_inc_simple(errsv_save);
3041 SvREFCNT_dec(errsv);
3042 }
eed484f9 3043 }
69794297
KW
3044 LEAVE;
3045 }
3046 SPAGAIN;
3047 PUSHMARK(SP);
3048 EXTEND(SP,5);
3049 mPUSHp(pkg, pkg_len);
3050 mPUSHp(name, name_len);
3051 PUSHs(listsv);
3052 mPUSHi(minbits);
3053 mPUSHi(none);
3054 PUTBACK;
db2c6cb3
FC
3055 if ((errsv_save = GvSV(PL_errgv))) SAVEFREESV(errsv_save);
3056 GvSV(PL_errgv) = NULL;
69794297
KW
3057 /* If we already have a pointer to the method, no need to use
3058 * call_method() to repeat the lookup. */
c41800a8
KW
3059 if (method
3060 ? call_sv(MUTABLE_SV(method), G_SCALAR)
69794297
KW
3061 : call_sv(newSVpvs_flags("SWASHNEW", SVs_TEMP), G_SCALAR | G_METHOD))
3062 {
3063 retval = *PL_stack_sp--;
3064 SvREFCNT_inc(retval);
3065 }
eed484f9 3066 {
db2c6cb3
FC
3067 /* Not ERRSV. See above. */
3068 SV * const errsv = GvSV(PL_errgv);
3069 if (!SvTRUE(errsv)) {
3070 GvSV(PL_errgv) = SvREFCNT_inc_simple(errsv_save);
3071 SvREFCNT_dec(errsv);
3072 }
eed484f9 3073 }
ce3b816e 3074 LEAVE;
69794297
KW
3075 POPSTACK;
3076 if (IN_PERL_COMPILETIME) {
3077 CopHINTS_set(PL_curcop, PL_hints);
3078 }
3079 if (!SvROK(retval) || SvTYPE(SvRV(retval)) != SVt_PVHV) {
3080 if (SvPOK(retval))
3081
3082 /* If caller wants to handle missing properties, let them */
5d3d13d1 3083 if (flags_p && *flags_p & _CORE_SWASH_INIT_RETURN_IF_UNDEF) {
69794297
KW
3084 return NULL;
3085 }
3086 Perl_croak(aTHX_
3087 "Can't find Unicode property definition \"%"SVf"\"",
3088 SVfARG(retval));
3089 Perl_croak(aTHX_ "SWASHNEW didn't return an HV ref");
3090 }
9a53f6cf 3091 } /* End of calling the module to find the swash */
36eb48b4 3092
83199d38
KW
3093 /* If this operation fetched a swash, and we will need it later, get it */
3094 if (retval != &PL_sv_undef
3095 && (minbits == 1 || (flags_p
3096 && ! (*flags_p
3097 & _CORE_SWASH_INIT_USER_DEFINED_PROPERTY))))
3098 {
3099 swash_hv = MUTABLE_HV(SvRV(retval));
3100
3101 /* If we don't already know that there is a user-defined component to
3102 * this swash, and the user has indicated they wish to know if there is
3103 * one (by passing <flags_p>), find out */
3104 if (flags_p && ! (*flags_p & _CORE_SWASH_INIT_USER_DEFINED_PROPERTY)) {
3105 SV** user_defined = hv_fetchs(swash_hv, "USER_DEFINED", FALSE);
3106 if (user_defined && SvUV(*user_defined)) {
3107 *flags_p |= _CORE_SWASH_INIT_USER_DEFINED_PROPERTY;
3108 }
3109 }
3110 }
3111
36eb48b4
KW
3112 /* Make sure there is an inversion list for binary properties */
3113 if (minbits == 1) {
3114 SV** swash_invlistsvp = NULL;
3115 SV* swash_invlist = NULL;
9a53f6cf 3116 bool invlist_in_swash_is_valid = FALSE;
02c85471
FC
3117 bool swash_invlist_unclaimed = FALSE; /* whether swash_invlist has
3118 an unclaimed reference count */
36eb48b4 3119
9a53f6cf 3120 /* If this operation fetched a swash, get its already existing
83199d38 3121 * inversion list, or create one for it */
36eb48b4 3122
83199d38 3123 if (swash_hv) {
5c9f4bd2 3124 swash_invlistsvp = hv_fetchs(swash_hv, "V", FALSE);
9a53f6cf
KW
3125 if (swash_invlistsvp) {
3126 swash_invlist = *swash_invlistsvp;
3127 invlist_in_swash_is_valid = TRUE;
3128 }
3129 else {
36eb48b4 3130 swash_invlist = _swash_to_invlist(retval);
02c85471 3131 swash_invlist_unclaimed = TRUE;
9a53f6cf
KW
3132 }
3133 }
3134
3135 /* If an inversion list was passed in, have to include it */
3136 if (invlist) {
3137
3138 /* Any fetched swash will by now have an inversion list in it;
3139 * otherwise <swash_invlist> will be NULL, indicating that we
3140 * didn't fetch a swash */
3141 if (swash_invlist) {
3142
3143 /* Add the passed-in inversion list, which invalidates the one
3144 * already stored in the swash */
3145 invlist_in_swash_is_valid = FALSE;
3146 _invlist_union(invlist, swash_invlist, &swash_invlist);
3147 }
3148 else {
3149
87367d5f
KW
3150 /* Here, there is no swash already. Set up a minimal one, if
3151 * we are going to return a swash */
3152 if ((int) _invlist_len(invlist) > invlist_swash_boundary) {
971d486f 3153 swash_hv = newHV();
4aca0fe6 3154 retval = newRV_noinc(MUTABLE_SV(swash_hv));
87367d5f 3155 }
9a53f6cf
KW
3156 swash_invlist = invlist;
3157 }
9a53f6cf
KW
3158 }
3159
3160 /* Here, we have computed the union of all the passed-in data. It may
3161 * be that there was an inversion list in the swash which didn't get
3162 * touched; otherwise save the one computed one */
87367d5f
KW
3163 if (! invlist_in_swash_is_valid
3164 && (int) _invlist_len(swash_invlist) > invlist_swash_boundary)
3165 {
5c9f4bd2 3166 if (! hv_stores(MUTABLE_HV(SvRV(retval)), "V", swash_invlist))
69794297
KW
3167 {
3168 Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed");
3169 }
cc34d8c5
FC
3170 /* We just stole a reference count. */
3171 if (swash_invlist_unclaimed) swash_invlist_unclaimed = FALSE;
3172 else SvREFCNT_inc_simple_void_NN(swash_invlist);
9a53f6cf 3173 }
87367d5f 3174
c41800a8 3175 /* Use the inversion list stand-alone if small enough */
87367d5f
KW
3176 if ((int) _invlist_len(swash_invlist) <= invlist_swash_boundary) {
3177 SvREFCNT_dec(retval);
02c85471
FC
3178 if (!swash_invlist_unclaimed)
3179 SvREFCNT_inc_simple_void_NN(swash_invlist);
3180 retval = newRV_noinc(swash_invlist);
87367d5f 3181 }
36eb48b4
KW
3182 }
3183
a0ed51b3
LW
3184 return retval;
3185}
3186
035d37be
JH
3187
3188/* This API is wrong for special case conversions since we may need to
3189 * return several Unicode characters for a single Unicode character
3190 * (see lib/unicore/SpecCase.txt) The SWASHGET in lib/utf8_heavy.pl is
3191 * the lower-level routine, and it is similarly broken for returning
38684baa
KW
3192 * multiple values. --jhi
3193 * For those, you should use to_utf8_case() instead */
b0e3252e 3194/* Now SWASHGET is recasted into S_swatch_get in this file. */
680c470c
ST
3195
3196/* Note:
3197 * Returns the value of property/mapping C<swash> for the first character
3198 * of the string C<ptr>. If C<do_utf8> is true, the string C<ptr> is
3d0f8846
KW
3199 * assumed to be in well-formed utf8. If C<do_utf8> is false, the string C<ptr>
3200 * is assumed to be in native 8-bit encoding. Caches the swatch in C<swash>.
af2af982
KW
3201 *
3202 * A "swash" is a hash which contains initially the keys/values set up by
3203 * SWASHNEW. The purpose is to be able to completely represent a Unicode
3204 * property for all possible code points. Things are stored in a compact form
3205 * (see utf8_heavy.pl) so that calculation is required to find the actual
3206 * property value for a given code point. As code points are looked up, new
3207 * key/value pairs are added to the hash, so that the calculation doesn't have
3208 * to ever be re-done. Further, each calculation is done, not just for the
3209 * desired one, but for a whole block of code points adjacent to that one.
3210 * For binary properties on ASCII machines, the block is usually for 64 code
3211 * points, starting with a code point evenly divisible by 64. Thus if the
3212 * property value for code point 257 is requested, the code goes out and
3213 * calculates the property values for all 64 code points between 256 and 319,
3214 * and stores these as a single 64-bit long bit vector, called a "swatch",
3215 * under the key for code point 256. The key is the UTF-8 encoding for code
3216 * point 256, minus the final byte. Thus, if the length of the UTF-8 encoding
3217 * for a code point is 13 bytes, the key will be 12 bytes long. If the value
3218 * for code point 258 is then requested, this code realizes that it would be
3219 * stored under the key for 256, and would find that value and extract the
3220 * relevant bit, offset from 256.
3221 *
3222 * Non-binary properties are stored in as many bits as necessary to represent
3223 * their values (32 currently, though the code is more general than that), not
3224 * as single bits, but the principal is the same: the value for each key is a
3225 * vector that encompasses the property values for all code points whose UTF-8
3226 * representations are represented by the key. That is, for all code points
3227 * whose UTF-8 representations are length N bytes, and the key is the first N-1
3228 * bytes of that.
680c470c 3229 */
a0ed51b3 3230UV
680c470c 3231Perl_swash_fetch(pTHX_ SV *swash, const U8 *ptr, bool do_utf8)
a0ed51b3 3232{
27da23d5 3233 dVAR;
ef8f7699 3234 HV *const hv = MUTABLE_HV(SvRV(swash));
3568d838
JH
3235 U32 klen;
3236 U32 off;
a0ed51b3 3237 STRLEN slen;
7d85a32c 3238 STRLEN needents;
cfd0369c 3239 const U8 *tmps = NULL;
a0ed51b3 3240 U32 bit;
979f2922 3241 SV *swatch;
08fb1ac5 3242 const U8 c = *ptr;
3568d838 3243
7918f24d
NC
3244 PERL_ARGS_ASSERT_SWASH_FETCH;
3245
87367d5f
KW
3246 /* If it really isn't a hash, it isn't really swash; must be an inversion
3247 * list */
3248 if (SvTYPE(hv) != SVt_PVHV) {
3249 return _invlist_contains_cp((SV*)hv,
3250 (do_utf8)
3251 ? valid_utf8_to_uvchr(ptr, NULL)
3252 : c);
3253 }
3254
08fb1ac5
KW
3255 /* We store the values in a "swatch" which is a vec() value in a swash
3256 * hash. Code points 0-255 are a single vec() stored with key length
3257 * (klen) 0. All other code points have a UTF-8 representation
3258 * 0xAA..0xYY,0xZZ. A vec() is constructed containing all of them which
3259 * share 0xAA..0xYY, which is the key in the hash to that vec. So the key
3260 * length for them is the length of the encoded char - 1. ptr[klen] is the
3261 * final byte in the sequence representing the character */
3262 if (!do_utf8 || UTF8_IS_INVARIANT(c)) {
3263 klen = 0;
3264 needents = 256;
3265 off = c;
3568d838 3266 }
08fb1ac5
KW
3267 else if (UTF8_IS_DOWNGRADEABLE_START(c)) {
3268 klen = 0;
3269 needents = 256;
3270 off = TWO_BYTE_UTF8_TO_NATIVE(c, *(ptr + 1));
979f2922
ST
3271 }
3272 else {
08fb1ac5
KW
3273 klen = UTF8SKIP(ptr) - 1;
3274
3275 /* Each vec() stores 2**UTF_ACCUMULATION_SHIFT values. The offset into
3276 * the vec is the final byte in the sequence. (In EBCDIC this is
3277 * converted to I8 to get consecutive values.) To help you visualize
3278 * all this:
3279 * Straight 1047 After final byte
3280 * UTF-8 UTF-EBCDIC I8 transform
3281 * U+0400: \xD0\x80 \xB8\x41\x41 \xB8\x41\xA0
3282 * U+0401: \xD0\x81 \xB8\x41\x42 \xB8\x41\xA1
3283 * ...
3284 * U+0409: \xD0\x89 \xB8\x41\x4A \xB8\x41\xA9
3285 * U+040A: \xD0\x8A \xB8\x41\x51 \xB8\x41\xAA
3286 * ...
3287 * U+0412: \xD0\x92 \xB8\x41\x59 \xB8\x41\xB2
3288 * U+0413: \xD0\x93 \xB8\x41\x62 \xB8\x41\xB3
3289 * ...
3290 * U+041B: \xD0\x9B \xB8\x41\x6A \xB8\x41\xBB
3291 * U+041C: \xD0\x9C \xB8\x41\x70 \xB8\x41\xBC
3292 * ...
3293 * U+041F: \xD0\x9F \xB8\x41\x73 \xB8\x41\xBF
3294 * U+0420: \xD0\xA0 \xB8\x42\x41 \xB8\x42\x41
3295 *
3296 * (There are no discontinuities in the elided (...) entries.)
3297 * The UTF-8 key for these 33 code points is '\xD0' (which also is the
3298 * key for the next 31, up through U+043F, whose UTF-8 final byte is
3299 * \xBF). Thus in UTF-8, each key is for a vec() for 64 code points.
3300 * The final UTF-8 byte, which ranges between \x80 and \xBF, is an
3301 * index into the vec() swatch (after subtracting 0x80, which we
3302 * actually do with an '&').
3303 * In UTF-EBCDIC, each key is for a 32 code point vec(). The first 32
3304 * code points above have key '\xB8\x41'. The final UTF-EBCDIC byte has
3305 * dicontinuities which go away by transforming it into I8, and we
3306 * effectively subtract 0xA0 to get the index. */
979f2922 3307 needents = (1 << UTF_ACCUMULATION_SHIFT);
bc3632a8 3308 off = NATIVE_UTF8_TO_I8(ptr[klen]) & UTF_CONTINUATION_MASK;
979f2922 3309 }
7d85a32c 3310
a0ed51b3
LW
3311 /*
3312 * This single-entry cache saves about 1/3 of the utf8 overhead in test
3313 * suite. (That is, only 7-8% overall over just a hash cache. Still,
3314 * it's nothing to sniff at.) Pity we usually come through at least
3315 * two function calls to get here...
3316 *
3317 * NB: this code assumes that swatches are never modified, once generated!
3318 */
3319
3568d838 3320 if (hv == PL_last_swash_hv &&
a0ed51b3 3321 klen == PL_last_swash_klen &&
27da23d5 3322 (!klen || memEQ((char *)ptr, (char *)PL_last_swash_key, klen)) )
a0ed51b3
LW
3323 {
3324 tmps = PL_last_swash_tmps;
3325 slen = PL_last_swash_slen;
3326 }
3327 else {
3328 /* Try our second-level swatch cache, kept in a hash. */
e1ec3a88 3329 SV** svp = hv_fetch(hv, (const char*)ptr, klen, FALSE);
a0ed51b3 3330
b0e3252e 3331 /* If not cached, generate it via swatch_get */
979f2922 3332 if (!svp || !SvPOK(*svp)
08fb1ac5
KW
3333 || !(tmps = (const U8*)SvPV_const(*svp, slen)))
3334 {
3335 if (klen) {
3336 const UV code_point = valid_utf8_to_uvchr(ptr, NULL);
3337 swatch = swatch_get(swash,
3338 code_point & ~((UV)needents - 1),
3339 needents);
3340 }
3341 else { /* For the first 256 code points, the swatch has a key of
3342 length 0 */
3343 swatch = swatch_get(swash, 0, needents);
3344 }
979f2922 3345
923e4eb5 3346 if (IN_PERL_COMPILETIME)
623e6609 3347 CopHINTS_set(PL_curcop, PL_hints);
a0ed51b3 3348
979f2922 3349 svp = hv_store(hv, (const char *)ptr, klen, swatch, 0);
a0ed51b3 3350
979f2922
ST
3351 if (!svp || !(tmps = (U8*)SvPV(*svp, slen))
3352 || (slen << 3) < needents)
5637ef5b
NC
3353 Perl_croak(aTHX_ "panic: swash_fetch got improper swatch, "
3354 "svp=%p, tmps=%p, slen=%"UVuf", needents=%"UVuf,
3355 svp, tmps, (UV)slen, (UV)needents);
a0ed51b3
LW
3356 }
3357
3358 PL_last_swash_hv = hv;
16d8f38a 3359 assert(klen <= sizeof(PL_last_swash_key));
eac04b2e 3360 PL_last_swash_klen = (U8)klen;
cfd0369c
NC
3361 /* FIXME change interpvar.h? */
3362 PL_last_swash_tmps = (U8 *) tmps;
a0ed51b3
LW
3363 PL_last_swash_slen = slen;
3364 if (klen)
3365 Copy(ptr, PL_last_swash_key, klen, U8);
3366 }
3367
9faf8d75 3368 switch ((int)((slen << 3) / needents)) {
a0ed51b3
LW
3369 case 1:
3370 bit = 1 << (off & 7);
3371 off >>= 3;
3372 return (tmps[off] & bit) != 0;
3373 case 8:
3374 return tmps[off];
3375 case 16:
3376 off <<= 1;
3377 return (tmps[off] << 8) + tmps[off + 1] ;
3378 case 32:
3379 off <<= 2;
3380 return (tmps[off] << 24) + (tmps[off+1] << 16) + (tmps[off+2] << 8) + tmps[off + 3] ;
3381 }
5637ef5b
NC
3382 Perl_croak(aTHX_ "panic: swash_fetch got swatch of unexpected bit width, "
3383 "slen=%"UVuf", needents=%"UVuf, (UV)slen, (UV)needents);
670f1322 3384 NORETURN_FUNCTION_END;
a0ed51b3 3385}
2b9d42f0 3386
319009ee
KW
3387/* Read a single line of the main body of the swash input text. These are of
3388 * the form:
3389 * 0053 0056 0073
3390 * where each number is hex. The first two numbers form the minimum and
3391 * maximum of a range, and the third is the value associated with the range.
3392 * Not all swashes should have a third number
3393 *
3394 * On input: l points to the beginning of the line to be examined; it points
3395 * to somewhere in the string of the whole input text, and is
3396 * terminated by a \n or the null string terminator.
3397 * lend points to the null terminator of that string
3398 * wants_value is non-zero if the swash expects a third number
3399 * typestr is the name of the swash's mapping, like 'ToLower'
3400 * On output: *min, *max, and *val are set to the values read from the line.
3401 * returns a pointer just beyond the line examined. If there was no
3402 * valid min number on the line, returns lend+1
3403 */
3404
3405STATIC U8*
3406S_swash_scan_list_line(pTHX_ U8* l, U8* const lend, UV* min, UV* max, UV* val,
3407 const bool wants_value, const U8* const typestr)
3408{
3409 const int typeto = typestr[0] == 'T' && typestr[1] == 'o';
3410 STRLEN numlen; /* Length of the number */
02470786
KW
3411 I32 flags = PERL_SCAN_SILENT_ILLDIGIT
3412 | PERL_SCAN_DISALLOW_PREFIX
3413 | PERL_SCAN_SILENT_NON_PORTABLE;
319009ee
KW
3414
3415 /* nl points to the next \n in the scan */
3416 U8* const nl = (U8*)memchr(l, '\n', lend - l);
3417
3418 /* Get the first number on the line: the range minimum */
3419 numlen = lend - l;
3420 *min = grok_hex((char *)l, &numlen, &flags, NULL);
3421 if (numlen) /* If found a hex number, position past it */
3422 l += numlen;
3423 else if (nl) { /* Else, go handle next line, if any */
3424 return nl + 1; /* 1 is length of "\n" */
3425 }
3426 else { /* Else, no next line */
3427 return lend + 1; /* to LIST's end at which \n is not found */
3428 }
3429
3430 /* The max range value follows, separated by a BLANK */
3431 if (isBLANK(*l)) {
3432 ++l;
02470786
KW
3433 flags = PERL_SCAN_SILENT_ILLDIGIT
3434 | PERL_SCAN_DISALLOW_PREFIX
3435 | PERL_SCAN_SILENT_NON_PORTABLE;
319009ee
KW
3436 numlen = lend - l;
3437 *max = grok_hex((char *)l, &numlen, &flags, NULL);
3438 if (numlen)
3439 l += numlen;
3440 else /* If no value here, it is a single element range */
3441 *max = *min;
3442
3443 /* Non-binary tables have a third entry: what the first element of the
24303724 3444 * range maps to. The map for those currently read here is in hex */
319009ee
KW
3445 if (wants_value) {
3446 if (isBLANK(*l)) {
3447 ++l;
f2a7d0fc
KW
3448 flags = PERL_SCAN_SILENT_ILLDIGIT
3449 | PERL_SCAN_DISALLOW_PREFIX
3450 | PERL_SCAN_SILENT_NON_PORTABLE;
3451 numlen = lend - l;
3452 *val = grok_hex((char *)l, &numlen, &flags, NULL);
3453 if (numlen)
3454 l += numlen;
3455 else
3456 *val = 0;
319009ee
KW
3457 }
3458 else {
3459 *val = 0;
3460 if (typeto) {
dcbac5bb 3461 /* diag_listed_as: To%s: illegal mapping '%s' */
319009ee
KW
3462 Perl_croak(aTHX_ "%s: illegal mapping '%s'",
3463 typestr, l);
3464 }
3465 }
3466 }
3467 else
3468 *val = 0; /* bits == 1, then any val should be ignored */
3469 }
3470 else { /* Nothing following range min, should be single element with no
3471 mapping expected */
3472 *max = *min;
3473 if (wants_value) {
3474 *val = 0;
3475 if (typeto) {
dcbac5bb 3476 /* diag_listed_as: To%s: illegal mapping '%s' */
319009ee
KW
3477 Perl_croak(aTHX_ "%s: illegal mapping '%s'", typestr, l);
3478 }
3479 }
3480 else
3481 *val = 0; /* bits == 1, then val should be ignored */
3482 }
3483
3484 /* Position to next line if any, or EOF */
3485 if (nl)
3486 l = nl + 1;
3487 else
3488 l = lend;
3489
3490 return l;
3491}
3492
979f2922
ST
3493/* Note:
3494 * Returns a swatch (a bit vector string) for a code point sequence
3495 * that starts from the value C<start> and comprises the number C<span>.
3496 * A C<swash> must be an object created by SWASHNEW (see lib/utf8_heavy.pl).
3497 * Should be used via swash_fetch, which will cache the swatch in C<swash>.
3498 */
3499STATIC SV*
b0e3252e 3500S_swatch_get(pTHX_ SV* swash, UV start, UV span)
979f2922
ST
3501{
3502 SV *swatch;
77f9f126 3503 U8 *l, *lend, *x, *xend, *s, *send;
979f2922 3504 STRLEN lcur, xcur, scur;
ef8f7699 3505 HV *const hv = MUTABLE_HV(SvRV(swash));
5c9f4bd2 3506 SV** const invlistsvp = hv_fetchs(hv, "V", FALSE);
36eb48b4 3507
88d45d28
KW
3508 SV** listsvp = NULL; /* The string containing the main body of the table */
3509 SV** extssvp = NULL;
3510 SV** invert_it_svp = NULL;
3511 U8* typestr = NULL;
786861f5
KW
3512 STRLEN bits;
3513 STRLEN octets; /* if bits == 1, then octets == 0 */
3514 UV none;
3515 UV end = start + span;
972dd592 3516
36eb48b4 3517 if (invlistsvp == NULL) {
786861f5
KW
3518 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
3519 SV** const nonesvp = hv_fetchs(hv, "NONE", FALSE);
3520 SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE);
3521 extssvp = hv_fetchs(hv, "EXTRAS", FALSE);
3522 listsvp = hv_fetchs(hv, "LIST", FALSE);
3523 invert_it_svp = hv_fetchs(hv, "INVERT_IT", FALSE);
3524
3525 bits = SvUV(*bitssvp);
3526 none = SvUV(*nonesvp);
3527 typestr = (U8*)SvPV_nolen(*typesvp);
3528 }
36eb48b4
KW
3529 else {
3530 bits = 1;
3531 none = 0;
3532 }
786861f5 3533 octets = bits >> 3; /* if bits == 1, then octets == 0 */
979f2922 3534
b0e3252e 3535 PERL_ARGS_ASSERT_SWATCH_GET;
7918f24d 3536
979f2922 3537 if (bits != 1 && bits != 8 && bits != 16 && bits != 32) {
b0e3252e 3538 Perl_croak(aTHX_ "panic: swatch_get doesn't expect bits %"UVuf,
660a4616 3539 (UV)bits);
979f2922
ST
3540 }
3541
84ea5ef6
KW
3542 /* If overflowed, use the max possible */
3543 if (end < start) {
3544 end = UV_MAX;
3545 span = end - start;
3546 }
3547
979f2922 3548 /* create and initialize $swatch */
979f2922 3549 scur = octets ? (span * octets) : (span + 7) / 8;
e524fe40
NC
3550 swatch = newSV(scur);
3551 SvPOK_on(swatch);
979f2922
ST
3552 s = (U8*)SvPVX(swatch);
3553 if (octets && none) {
0bd48802 3554 const U8* const e = s + scur;
979f2922
ST
3555 while (s < e) {
3556 if (bits == 8)
3557 *s++ = (U8)(none & 0xff);
3558 else if (bits == 16) {
3559 *s++ = (U8)((none >> 8) & 0xff);
3560 *s++ = (U8)( none & 0xff);
3561 }
3562 else if (bits == 32) {
3563 *s++ = (U8)((none >> 24) & 0xff);
3564 *s++ = (U8)((none >> 16) & 0xff);
3565 *s++ = (U8)((none >> 8) & 0xff);
3566 *s++ = (U8)( none & 0xff);
3567 }
3568 }
3569 *s = '\0';
3570 }
3571 else {
3572 (void)memzero((U8*)s, scur + 1);
3573 }
3574 SvCUR_set(swatch, scur);
3575 s = (U8*)SvPVX(swatch);
3576
36eb48b4
KW
3577 if (invlistsvp) { /* If has an inversion list set up use that */
3578 _invlist_populate_swatch(*invlistsvp, start, end, s);
3579 return swatch;
3580 }
3581
3582 /* read $swash->{LIST} */
979f2922
ST
3583 l = (U8*)SvPV(*listsvp, lcur);
3584 lend = l + lcur;
3585 while (l < lend) {
8ed25d53 3586 UV min, max, val, upper;
319009ee
KW
3587 l = S_swash_scan_list_line(aTHX_ l, lend, &min, &max, &val,
3588 cBOOL(octets), typestr);
3589 if (l > lend) {
979f2922
ST
3590 break;
3591 }
3592
972dd592 3593 /* If looking for something beyond this range, go try the next one */
979f2922
ST
3594 if (max < start)
3595 continue;
3596
8ed25d53
KW
3597 /* <end> is generally 1 beyond where we want to set things, but at the
3598 * platform's infinity, where we can't go any higher, we want to
3599 * include the code point at <end> */
3600 upper = (max < end)
3601 ? max
3602 : (max != UV_MAX || end != UV_MAX)
3603 ? end - 1
3604 : end;
3605
979f2922 3606 if (octets) {
35da51f7 3607 UV key;
979f2922
ST
3608 if (min < start) {
3609 if (!none || val < none) {
3610 val += start - min;
3611 }
3612 min = start;
3613 }
8ed25d53 3614 for (key = min; key <= upper; key++) {
979f2922 3615 STRLEN offset;
979f2922
ST
3616 /* offset must be non-negative (start <= min <= key < end) */
3617 offset = octets * (key - start);
3618 if (bits == 8)
3619 s[offset] = (U8)(val & 0xff);
3620 else if (bits == 16) {
3621 s[offset ] = (U8)((val >> 8) & 0xff);
3622 s[offset + 1] = (U8)( val & 0xff);
3623 }
3624 else if (bits == 32) {
3625 s[offset ] = (U8)((val >> 24) & 0xff);
3626 s[offset + 1] = (U8)((val >> 16) & 0xff);
3627 s[offset + 2] = (U8)((val >> 8) & 0xff);
3628 s[offset + 3] = (U8)( val & 0xff);
3629 }
3630
3631 if (!none || val < none)
3632 ++val;
3633 }
3634 }
711a919c 3635 else { /* bits == 1, then val should be ignored */
35da51f7 3636 UV key;
979f2922
ST
3637 if (min < start)
3638 min = start;
6cb05c12 3639
8ed25d53 3640 for (key = min; key <= upper; key++) {