This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
utf8.c: Add comment
[perl5.git] / utf8.c
CommitLineData
a0ed51b3
LW
1/* utf8.c
2 *
1129b882 3 * Copyright (C) 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008
b94e2f88 4 * by Larry Wall and others
a0ed51b3
LW
5 *
6 * You may distribute under the terms of either the GNU General Public
7 * License or the Artistic License, as specified in the README file.
8 *
9 */
10
11/*
4ac71550
TC
12 * 'What a fix!' said Sam. 'That's the one place in all the lands we've ever
13 * heard of that we don't want to see any closer; and that's the one place
14 * we're trying to get to! And that's just where we can't get, nohow.'
15 *
cdad3b53 16 * [p.603 of _The Lord of the Rings_, IV/I: "The Taming of Sméagol"]
a0ed51b3
LW
17 *
18 * 'Well do I understand your speech,' he answered in the same language;
19 * 'yet few strangers do so. Why then do you not speak in the Common Tongue,
4ac71550 20 * as is the custom in the West, if you wish to be answered?'
cdad3b53 21 * --Gandalf, addressing Théoden's door wardens
4ac71550
TC
22 *
23 * [p.508 of _The Lord of the Rings_, III/vi: "The King of the Golden Hall"]
a0ed51b3
LW
24 *
25 * ...the travellers perceived that the floor was paved with stones of many
26 * hues; branching runes and strange devices intertwined beneath their feet.
4ac71550
TC
27 *
28 * [p.512 of _The Lord of the Rings_, III/vi: "The King of the Golden Hall"]
a0ed51b3
LW
29 */
30
31#include "EXTERN.h"
864dbfa3 32#define PERL_IN_UTF8_C
a0ed51b3
LW
33#include "perl.h"
34
a0c21aa1 35#ifndef EBCDIC
970ea3cb 36/* Separate prototypes needed because in ASCII systems these are
a0c21aa1
JH
37 * usually macros but they still are compiled as code, too. */
38PERL_CALLCONV UV Perl_utf8n_to_uvchr(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen, U32 flags);
39PERL_CALLCONV U8* Perl_uvchr_to_utf8(pTHX_ U8 *d, UV uv);
40#endif
41
27da23d5
JH
42static const char unees[] =
43 "Malformed UTF-8 character (unexpected end of string)";
901b21bf 44
48ef279e 45/*
ccfc67b7 46=head1 Unicode Support
a0ed51b3 47
166f8a29
DM
48This file contains various utility functions for manipulating UTF8-encoded
49strings. For the uninitiated, this is a method of representing arbitrary
61296642 50Unicode characters as a variable number of bytes, in such a way that
56da48f7
DM
51characters in the ASCII range are unmodified, and a zero byte never appears
52within non-zero characters.
166f8a29 53
eaf7a4d2
CS
54=cut
55*/
56
57/*
58=for apidoc is_ascii_string
59
970ea3cb
KW
60Returns true if the first C<len> bytes of the given string are the same whether
61or not the string is encoded in UTF-8 (or UTF-EBCDIC on EBCDIC machines). That
62is, if they are invariant. On ASCII-ish machines, only ASCII characters
63fit this definition, hence the function's name.
eaf7a4d2 64
9f7e3d64
MH
65If C<len> is 0, it will be calculated using C<strlen(s)>.
66
eaf7a4d2
CS
67See also is_utf8_string(), is_utf8_string_loclen(), and is_utf8_string_loc().
68
69=cut
70*/
71
72bool
668b6d8d 73Perl_is_ascii_string(const U8 *s, STRLEN len)
eaf7a4d2
CS
74{
75 const U8* const send = s + (len ? len : strlen((const char *)s));
76 const U8* x = s;
77
78 PERL_ARGS_ASSERT_IS_ASCII_STRING;
eaf7a4d2
CS
79
80 for (; x < send; ++x) {
81 if (!UTF8_IS_INVARIANT(*x))
82 break;
83 }
84
85 return x == send;
86}
87
88/*
87cea99e 89=for apidoc uvuni_to_utf8_flags
eebe1485 90
6ee84de2
KW
91Adds the UTF-8 representation of the code point C<uv> to the end
92of the string C<d>; C<d> should have at least C<UTF8_MAXBYTES+1> free
eebe1485 93bytes available. The return value is the pointer to the byte after the
9041c2e3 94end of the new character. In other words,
eebe1485 95
b851fbc1
JH
96 d = uvuni_to_utf8_flags(d, uv, flags);
97
98or, in most cases,
99
9041c2e3 100 d = uvuni_to_utf8(d, uv);
eebe1485 101
b851fbc1
JH
102(which is equivalent to)
103
104 d = uvuni_to_utf8_flags(d, uv, 0);
105
949cf498 106This is the recommended Unicode-aware way of saying
eebe1485
SC
107
108 *(d++) = uv;
109
949cf498
KW
110This function will convert to UTF-8 (and not warn) even code points that aren't
111legal Unicode or are problematic, unless C<flags> contains one or more of the
112following flags.
113If C<uv> is a Unicode surrogate code point and UNICODE_WARN_SURROGATE is set,
114the function will raise a warning, provided UTF8 warnings are enabled. If instead
115UNICODE_DISALLOW_SURROGATE is set, the function will fail and return NULL.
116If both flags are set, the function will both warn and return NULL.
117
118The UNICODE_WARN_NONCHAR and UNICODE_DISALLOW_NONCHAR flags correspondingly
119affect how the function handles a Unicode non-character. And, likewise for the
120UNICODE_WARN_SUPER and UNICODE_DISALLOW_SUPER flags, and code points that are
121above the Unicode maximum of 0x10FFFF. Code points above 0x7FFF_FFFF (which are
122even less portable) can be warned and/or disallowed even if other above-Unicode
123code points are accepted by the UNICODE_WARN_FE_FF and UNICODE_DISALLOW_FE_FF
124flags.
125
126And finally, the flag UNICODE_WARN_ILLEGAL_INTERCHANGE selects all four of the
127above WARN flags; and UNICODE_DISALLOW_ILLEGAL_INTERCHANGE selects all four
128DISALLOW flags.
129
130
eebe1485
SC
131=cut
132*/
133
dfe13c55 134U8 *
b851fbc1 135Perl_uvuni_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags)
a0ed51b3 136{
7918f24d
NC
137 PERL_ARGS_ASSERT_UVUNI_TO_UTF8_FLAGS;
138
949cf498
KW
139 if (ckWARN_d(WARN_UTF8)) {
140 if (UNICODE_IS_SURROGATE(uv)) {
141 if (flags & UNICODE_WARN_SURROGATE) {
8457b38f 142 Perl_ck_warner_d(aTHX_ packWARN(WARN_SURROGATE),
949cf498
KW
143 "UTF-16 surrogate U+%04"UVXf, uv);
144 }
145 if (flags & UNICODE_DISALLOW_SURROGATE) {
146 return NULL;
147 }
148 }
149 else if (UNICODE_IS_SUPER(uv)) {
150 if (flags & UNICODE_WARN_SUPER
151 || (UNICODE_IS_FE_FF(uv) && (flags & UNICODE_WARN_FE_FF)))
152 {
8457b38f 153 Perl_ck_warner_d(aTHX_ packWARN(WARN_NON_UNICODE),
949cf498
KW
154 "Code point 0x%04"UVXf" is not Unicode, may not be portable", uv);
155 }
156 if (flags & UNICODE_DISALLOW_SUPER
157 || (UNICODE_IS_FE_FF(uv) && (flags & UNICODE_DISALLOW_FE_FF)))
158 {
159 return NULL;
160 }
161 }
162 else if (UNICODE_IS_NONCHAR(uv)) {
163 if (flags & UNICODE_WARN_NONCHAR) {
8457b38f 164 Perl_ck_warner_d(aTHX_ packWARN(WARN_NONCHAR),
949cf498
KW
165 "Unicode non-character U+%04"UVXf" is illegal for open interchange",
166 uv);
167 }
168 if (flags & UNICODE_DISALLOW_NONCHAR) {
169 return NULL;
170 }
171 }
507b9800 172 }
c4d5f83a 173 if (UNI_IS_INVARIANT(uv)) {
eb160463 174 *d++ = (U8)UTF_TO_NATIVE(uv);
a0ed51b3
LW
175 return d;
176 }
2d331972 177#if defined(EBCDIC)
1d72bdf6
NIS
178 else {
179 STRLEN len = UNISKIP(uv);
180 U8 *p = d+len-1;
181 while (p > d) {
eb160463 182 *p-- = (U8)UTF_TO_NATIVE((uv & UTF_CONTINUATION_MASK) | UTF_CONTINUATION_MARK);
1d72bdf6
NIS
183 uv >>= UTF_ACCUMULATION_SHIFT;
184 }
eb160463 185 *p = (U8)UTF_TO_NATIVE((uv & UTF_START_MASK(len)) | UTF_START_MARK(len));
1d72bdf6
NIS
186 return d+len;
187 }
188#else /* Non loop style */
a0ed51b3 189 if (uv < 0x800) {
eb160463
GS
190 *d++ = (U8)(( uv >> 6) | 0xc0);
191 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
192 return d;
193 }
194 if (uv < 0x10000) {
eb160463
GS
195 *d++ = (U8)(( uv >> 12) | 0xe0);
196 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
197 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
198 return d;
199 }
200 if (uv < 0x200000) {
eb160463
GS
201 *d++ = (U8)(( uv >> 18) | 0xf0);
202 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
203 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
204 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
205 return d;
206 }
207 if (uv < 0x4000000) {
eb160463
GS
208 *d++ = (U8)(( uv >> 24) | 0xf8);
209 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
210 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
211 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
212 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
213 return d;
214 }
215 if (uv < 0x80000000) {
eb160463
GS
216 *d++ = (U8)(( uv >> 30) | 0xfc);
217 *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80);
218 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
219 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
220 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
221 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
222 return d;
223 }
6b8eaf93 224#ifdef HAS_QUAD
d7578b48 225 if (uv < UTF8_QUAD_MAX)
a0ed51b3
LW
226#endif
227 {
eb160463
GS
228 *d++ = 0xfe; /* Can't match U+FEFF! */
229 *d++ = (U8)(((uv >> 30) & 0x3f) | 0x80);
230 *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80);
231 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
232 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
233 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
234 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
235 return d;
236 }
6b8eaf93 237#ifdef HAS_QUAD
a0ed51b3 238 {
eb160463
GS
239 *d++ = 0xff; /* Can't match U+FFFE! */
240 *d++ = 0x80; /* 6 Reserved bits */
241 *d++ = (U8)(((uv >> 60) & 0x0f) | 0x80); /* 2 Reserved bits */
242 *d++ = (U8)(((uv >> 54) & 0x3f) | 0x80);
243 *d++ = (U8)(((uv >> 48) & 0x3f) | 0x80);
244 *d++ = (U8)(((uv >> 42) & 0x3f) | 0x80);
245 *d++ = (U8)(((uv >> 36) & 0x3f) | 0x80);
246 *d++ = (U8)(((uv >> 30) & 0x3f) | 0x80);
247 *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80);
248 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
249 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
250 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
251 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
252 return d;
253 }
254#endif
1d72bdf6 255#endif /* Loop style */
a0ed51b3 256}
9041c2e3 257
646ca15d
JH
258/*
259
260Tests if some arbitrary number of bytes begins in a valid UTF-8
261character. Note that an INVARIANT (i.e. ASCII) character is a valid
262UTF-8 character. The actual number of bytes in the UTF-8 character
263will be returned if it is valid, otherwise 0.
264
265This is the "slow" version as opposed to the "fast" version which is
266the "unrolled" IS_UTF8_CHAR(). E.g. for t/uni/class.t the speed
267difference is a factor of 2 to 3. For lengths (UTF8SKIP(s)) of four
268or less you should use the IS_UTF8_CHAR(), for lengths of five or more
269you should use the _slow(). In practice this means that the _slow()
270will be used very rarely, since the maximum Unicode code point (as of
271Unicode 4.1) is U+10FFFF, which encodes in UTF-8 to four bytes. Only
272the "Perl extended UTF-8" (the infamous 'v-strings') will encode into
273five bytes or more.
274
275=cut */
c053b435 276STATIC STRLEN
5f66b61c 277S_is_utf8_char_slow(const U8 *s, const STRLEN len)
646ca15d
JH
278{
279 U8 u = *s;
280 STRLEN slen;
281 UV uv, ouv;
282
7918f24d
NC
283 PERL_ARGS_ASSERT_IS_UTF8_CHAR_SLOW;
284
646ca15d
JH
285 if (UTF8_IS_INVARIANT(u))
286 return 1;
287
288 if (!UTF8_IS_START(u))
289 return 0;
290
291 if (len < 2 || !UTF8_IS_CONTINUATION(s[1]))
292 return 0;
293
294 slen = len - 1;
295 s++;
77263263
TS
296#ifdef EBCDIC
297 u = NATIVE_TO_UTF(u);
298#endif
646ca15d
JH
299 u &= UTF_START_MASK(len);
300 uv = u;
301 ouv = uv;
302 while (slen--) {
303 if (!UTF8_IS_CONTINUATION(*s))
304 return 0;
305 uv = UTF8_ACCUMULATE(uv, *s);
48ef279e 306 if (uv < ouv)
646ca15d
JH
307 return 0;
308 ouv = uv;
309 s++;
310 }
311
312 if ((STRLEN)UNISKIP(uv) < len)
313 return 0;
314
315 return len;
316}
9041c2e3
NIS
317
318/*
87cea99e 319=for apidoc is_utf8_char
eebe1485 320
5da9da9e 321Tests if some arbitrary number of bytes begins in a valid UTF-8
2bbc8d55
SP
322character. Note that an INVARIANT (i.e. ASCII on non-EBCDIC machines)
323character is a valid UTF-8 character. The actual number of bytes in the UTF-8
324character will be returned if it is valid, otherwise 0.
9041c2e3 325
e0328548
KW
326WARNING: use only if you *know* that C<s> has at least either UTF8_MAXBYTES or
327UTF8SKIP(s) bytes.
328
82686b01 329=cut */
067a85ef 330STRLEN
668b6d8d 331Perl_is_utf8_char(const U8 *s)
386d01d6 332{
44f8325f 333 const STRLEN len = UTF8SKIP(s);
7918f24d
NC
334
335 PERL_ARGS_ASSERT_IS_UTF8_CHAR;
3b0fc154 336#ifdef IS_UTF8_CHAR
768c67ee 337 if (IS_UTF8_CHAR_FAST(len))
3b0fc154
JH
338 return IS_UTF8_CHAR(s, len) ? len : 0;
339#endif /* #ifdef IS_UTF8_CHAR */
2c0c5f92 340 return is_utf8_char_slow(s, len);
386d01d6
GS
341}
342
eaf7a4d2 343
6662521e 344/*
87cea99e 345=for apidoc is_utf8_string
6662521e 346
c9ada85f 347Returns true if first C<len> bytes of the given string form a valid
9f7e3d64 348UTF-8 string, false otherwise. If C<len> is 0, it will be calculated
e0328548
KW
349using C<strlen(s)> (which means if you use this option, that C<s> has to have a
350terminating NUL byte). Note that all characters being ASCII constitute 'a
351valid UTF-8 string'.
6662521e 352
eaf7a4d2 353See also is_ascii_string(), is_utf8_string_loclen(), and is_utf8_string_loc().
768c67ee 354
6662521e
GS
355=cut
356*/
357
8e84507e 358bool
668b6d8d 359Perl_is_utf8_string(const U8 *s, STRLEN len)
6662521e 360{
35da51f7 361 const U8* const send = s + (len ? len : strlen((const char *)s));
7fc63493 362 const U8* x = s;
067a85ef 363
7918f24d 364 PERL_ARGS_ASSERT_IS_UTF8_STRING;
1aa99e6b 365
6662521e 366 while (x < send) {
1acdb0da 367 /* Inline the easy bits of is_utf8_char() here for speed... */
e0328548
KW
368 if (UTF8_IS_INVARIANT(*x)) {
369 x++;
370 }
1acdb0da 371 else if (!UTF8_IS_START(*x))
e0328548 372 return FALSE;
1acdb0da
JH
373 else {
374 /* ... and call is_utf8_char() only if really needed. */
e0328548
KW
375 const STRLEN c = UTF8SKIP(x);
376 const U8* const next_char_ptr = x + c;
377
378 if (next_char_ptr > send) {
379 return FALSE;
380 }
381
768c67ee
JH
382 if (IS_UTF8_CHAR_FAST(c)) {
383 if (!IS_UTF8_CHAR(x, c))
e0328548 384 return FALSE;
3c614e38 385 }
e0328548
KW
386 else if (! is_utf8_char_slow(x, c)) {
387 return FALSE;
388 }
389 x = next_char_ptr;
1acdb0da 390 }
6662521e 391 }
768c67ee 392
067a85ef 393 return TRUE;
6662521e
GS
394}
395
67e989fb 396/*
814fafa7
NC
397Implemented as a macro in utf8.h
398
87cea99e 399=for apidoc is_utf8_string_loc
814fafa7
NC
400
401Like is_utf8_string() but stores the location of the failure (in the
402case of "utf8ness failure") or the location s+len (in the case of
403"utf8ness success") in the C<ep>.
404
405See also is_utf8_string_loclen() and is_utf8_string().
406
87cea99e 407=for apidoc is_utf8_string_loclen
81cd54e3 408
e3e4599f 409Like is_utf8_string() but stores the location of the failure (in the
768c67ee
JH
410case of "utf8ness failure") or the location s+len (in the case of
411"utf8ness success") in the C<ep>, and the number of UTF-8
412encoded characters in the C<el>.
413
414See also is_utf8_string_loc() and is_utf8_string().
81cd54e3
JH
415
416=cut
417*/
418
419bool
668b6d8d 420Perl_is_utf8_string_loclen(const U8 *s, STRLEN len, const U8 **ep, STRLEN *el)
81cd54e3 421{
35da51f7 422 const U8* const send = s + (len ? len : strlen((const char *)s));
7fc63493 423 const U8* x = s;
81cd54e3 424 STRLEN c;
3ebfea28 425 STRLEN outlen = 0;
7918f24d
NC
426
427 PERL_ARGS_ASSERT_IS_UTF8_STRING_LOCLEN;
81cd54e3 428
81cd54e3 429 while (x < send) {
e0328548
KW
430 const U8* next_char_ptr;
431
81cd54e3
JH
432 /* Inline the easy bits of is_utf8_char() here for speed... */
433 if (UTF8_IS_INVARIANT(*x))
e0328548 434 next_char_ptr = x + 1;
768c67ee
JH
435 else if (!UTF8_IS_START(*x))
436 goto out;
81cd54e3 437 else {
768c67ee 438 /* ... and call is_utf8_char() only if really needed. */
768c67ee 439 c = UTF8SKIP(x);
e0328548
KW
440 next_char_ptr = c + x;
441 if (next_char_ptr > send) {
442 goto out;
443 }
768c67ee
JH
444 if (IS_UTF8_CHAR_FAST(c)) {
445 if (!IS_UTF8_CHAR(x, c))
446 c = 0;
447 } else
448 c = is_utf8_char_slow(x, c);
768c67ee
JH
449 if (!c)
450 goto out;
81cd54e3 451 }
e0328548 452 x = next_char_ptr;
3ebfea28 453 outlen++;
81cd54e3 454 }
768c67ee
JH
455
456 out:
3ebfea28
AL
457 if (el)
458 *el = outlen;
459
768c67ee
JH
460 if (ep)
461 *ep = x;
3ebfea28 462 return (x == send);
81cd54e3
JH
463}
464
465/*
768c67ee 466
87cea99e 467=for apidoc utf8n_to_uvuni
67e989fb 468
9041c2e3 469Bottom level UTF-8 decode routine.
949cf498
KW
470Returns the code point value of the first character in the string C<s>
471which is assumed to be in UTF-8 (or UTF-EBCDIC) encoding and no longer than
472C<curlen> bytes; C<retlen> will be set to the length, in bytes, of that
473character.
474
475The value of C<flags> determines the behavior when C<s> does not point to a
476well-formed UTF-8 character. If C<flags> is 0, when a malformation is found,
477C<retlen> is set to the expected length of the UTF-8 character in bytes, zero
478is returned, and if UTF-8 warnings haven't been lexically disabled, a warning
479is raised.
480
481Various ALLOW flags can be set in C<flags> to allow (and not warn on)
482individual types of malformations, such as the sequence being overlong (that
483is, when there is a shorter sequence that can express the same code point;
484overlong sequences are expressly forbidden in the UTF-8 standard due to
485potential security issues). Another malformation example is the first byte of
486a character not being a legal first byte. See F<utf8.h> for the list of such
487flags. Of course, the value returned by this function under such conditions is
488not reliable.
489
490The UTF8_CHECK_ONLY flag overrides the behavior when a non-allowed (by other
491flags) malformation is found. If this flag is set, the routine assumes that
492the caller will raise a warning, and this function will silently just set
493C<retlen> to C<-1> and return zero.
494
495Certain code points are considered problematic. These are Unicode surrogates,
496Unicode non-characters, and code points above the Unicode maximum of 0x10FFF.
497By default these are considered regular code points, but certain situations
5eafe189 498warrant special handling for them. If C<flags> contains
949cf498
KW
499UTF8_DISALLOW_ILLEGAL_INTERCHANGE, all three classes are treated as
500malformations and handled as such. The flags UTF8_DISALLOW_SURROGATE,
501UTF8_DISALLOW_NONCHAR, and UTF8_DISALLOW_SUPER (meaning above the legal Unicode
502maximum) can be set to disallow these categories individually.
503
504The flags UTF8_WARN_ILLEGAL_INTERCHANGE, UTF8_WARN_SURROGATE,
505UTF8_WARN_NONCHAR, and UTF8_WARN_SUPER will cause warning messages to be raised
506for their respective categories, but otherwise the code points are considered
507valid (not malformations). To get a category to both be treated as a
508malformation and raise a warning, specify both the WARN and DISALLOW flags.
509(But note that warnings are not raised if lexically disabled nor if
510UTF8_CHECK_ONLY is also specified.)
511
512Very large code points (above 0x7FFF_FFFF) are considered more problematic than
513the others that are above the Unicode legal maximum. There are several
514reasons, one of which is that the original UTF-8 specification never went above
515this number (the current 0x10FFF limit was imposed later). The UTF-8 encoding
5eafe189 516on ASCII platforms for these large code points begins with a byte containing
949cf498
KW
5170xFE or 0xFF. The UTF8_DISALLOW_FE_FF flag will cause them to be treated as
518malformations, while allowing smaller above-Unicode code points. (Of course
519UTF8_DISALLOW_SUPER will treat all above-Unicode code points, including these,
520as malformations.) Similarly, UTF8_WARN_FE_FF acts just like the other WARN
521flags, but applies just to these code points.
522
523All other code points corresponding to Unicode characters, including private
524use and those yet to be assigned, are never considered malformed and never
525warn.
67e989fb 526
9041c2e3
NIS
527Most code should use utf8_to_uvchr() rather than call this directly.
528
37607a96
PK
529=cut
530*/
67e989fb 531
a0ed51b3 532UV
7fc63493 533Perl_utf8n_to_uvuni(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen, U32 flags)
a0ed51b3 534{
97aff369 535 dVAR;
d4c19fe8 536 const U8 * const s0 = s;
9c5ffd7c 537 UV uv = *s, ouv = 0;
ba210ebe 538 STRLEN len = 1;
949cf498 539 bool dowarn = ckWARN_d(WARN_UTF8);
7fc63493 540 const UV startbyte = *s;
ba210ebe 541 STRLEN expectlen = 0;
a0dbb045 542 U32 warning = 0;
949cf498 543 SV* sv = NULL;
a0dbb045 544
7918f24d
NC
545 PERL_ARGS_ASSERT_UTF8N_TO_UVUNI;
546
949cf498 547/* This list is a superset of the UTF8_ALLOW_XXX. */
a0dbb045
JH
548
549#define UTF8_WARN_EMPTY 1
550#define UTF8_WARN_CONTINUATION 2
551#define UTF8_WARN_NON_CONTINUATION 3
f9bc8ed7
KW
552#define UTF8_WARN_SHORT 4
553#define UTF8_WARN_OVERFLOW 5
554#define UTF8_WARN_LONG 6
a0dbb045
JH
555
556 if (curlen == 0 &&
557 !(flags & UTF8_ALLOW_EMPTY)) {
558 warning = UTF8_WARN_EMPTY;
0c443dc2
JH
559 goto malformed;
560 }
561
1d72bdf6 562 if (UTF8_IS_INVARIANT(uv)) {
a0ed51b3
LW
563 if (retlen)
564 *retlen = 1;
c4d5f83a 565 return (UV) (NATIVE_TO_UTF(*s));
a0ed51b3 566 }
67e989fb 567
421a8bf2 568 if (UTF8_IS_CONTINUATION(uv) &&
fcc8fcf6 569 !(flags & UTF8_ALLOW_CONTINUATION)) {
a0dbb045 570 warning = UTF8_WARN_CONTINUATION;
ba210ebe
JH
571 goto malformed;
572 }
573
421a8bf2 574 if (UTF8_IS_START(uv) && curlen > 1 && !UTF8_IS_CONTINUATION(s[1]) &&
fcc8fcf6 575 !(flags & UTF8_ALLOW_NON_CONTINUATION)) {
a0dbb045 576 warning = UTF8_WARN_NON_CONTINUATION;
ba210ebe
JH
577 goto malformed;
578 }
9041c2e3 579
1d72bdf6 580#ifdef EBCDIC
75383841 581 uv = NATIVE_TO_UTF(uv);
1d72bdf6 582#else
949cf498
KW
583 if (uv == 0xfe || uv == 0xff) {
584 if (flags & (UTF8_WARN_SUPER|UTF8_WARN_FE_FF)) {
111d382d 585 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "Code point beginning with byte 0x%02"UVXf" is not Unicode, and not portable", uv));
949cf498
KW
586 flags &= ~UTF8_WARN_SUPER; /* Only warn once on this problem */
587 }
588 if (flags & (UTF8_DISALLOW_SUPER|UTF8_DISALLOW_FE_FF)) {
589 goto malformed;
590 }
a0ed51b3 591 }
1d72bdf6
NIS
592#endif
593
ba210ebe
JH
594 if (!(uv & 0x20)) { len = 2; uv &= 0x1f; }
595 else if (!(uv & 0x10)) { len = 3; uv &= 0x0f; }
596 else if (!(uv & 0x08)) { len = 4; uv &= 0x07; }
597 else if (!(uv & 0x04)) { len = 5; uv &= 0x03; }
1d72bdf6
NIS
598#ifdef EBCDIC
599 else if (!(uv & 0x02)) { len = 6; uv &= 0x01; }
600 else { len = 7; uv &= 0x01; }
601#else
ba210ebe
JH
602 else if (!(uv & 0x02)) { len = 6; uv &= 0x01; }
603 else if (!(uv & 0x01)) { len = 7; uv = 0; }
1d72bdf6
NIS
604 else { len = 13; uv = 0; } /* whoa! */
605#endif
606
a0ed51b3
LW
607 if (retlen)
608 *retlen = len;
9041c2e3 609
ba210ebe
JH
610 expectlen = len;
611
fcc8fcf6
JH
612 if ((curlen < expectlen) &&
613 !(flags & UTF8_ALLOW_SHORT)) {
a0dbb045 614 warning = UTF8_WARN_SHORT;
ba210ebe
JH
615 goto malformed;
616 }
617
618 len--;
a0ed51b3 619 s++;
949cf498 620 ouv = uv; /* ouv is the value from the previous iteration */
ba210ebe 621
a0ed51b3 622 while (len--) {
421a8bf2
JH
623 if (!UTF8_IS_CONTINUATION(*s) &&
624 !(flags & UTF8_ALLOW_NON_CONTINUATION)) {
a0dbb045
JH
625 s--;
626 warning = UTF8_WARN_NON_CONTINUATION;
ba210ebe 627 goto malformed;
a0ed51b3
LW
628 }
629 else
8850bf83 630 uv = UTF8_ACCUMULATE(uv, *s);
949cf498
KW
631 if (!(uv > ouv)) { /* If the value didn't grow from the previous
632 iteration, something is horribly wrong */
a0dbb045
JH
633 /* These cannot be allowed. */
634 if (uv == ouv) {
75dbc644 635 if (expectlen != 13 && !(flags & UTF8_ALLOW_LONG)) {
a0dbb045
JH
636 warning = UTF8_WARN_LONG;
637 goto malformed;
638 }
639 }
640 else { /* uv < ouv */
641 /* This cannot be allowed. */
642 warning = UTF8_WARN_OVERFLOW;
643 goto malformed;
644 }
ba210ebe
JH
645 }
646 s++;
647 ouv = uv;
648 }
649
949cf498 650 if ((expectlen > (STRLEN)UNISKIP(uv)) && !(flags & UTF8_ALLOW_LONG)) {
a0dbb045 651 warning = UTF8_WARN_LONG;
ba210ebe 652 goto malformed;
949cf498
KW
653 } else if (flags & (UTF8_DISALLOW_ILLEGAL_INTERCHANGE|UTF8_WARN_ILLEGAL_INTERCHANGE)) {
654 if (UNICODE_IS_SURROGATE(uv)) {
655 if ((flags & (UTF8_WARN_SURROGATE|UTF8_CHECK_ONLY)) == UTF8_WARN_SURROGATE) {
111d382d 656 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "UTF-16 surrogate U+%04"UVXf"", uv));
949cf498
KW
657 }
658 if (flags & UTF8_DISALLOW_SURROGATE) {
659 goto disallowed;
660 }
661 }
662 else if (UNICODE_IS_NONCHAR(uv)) {
663 if ((flags & (UTF8_WARN_NONCHAR|UTF8_CHECK_ONLY)) == UTF8_WARN_NONCHAR ) {
111d382d 664 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "Unicode non-character U+%04"UVXf" is illegal for open interchange", uv));
949cf498
KW
665 }
666 if (flags & UTF8_DISALLOW_NONCHAR) {
667 goto disallowed;
668 }
669 }
670 else if ((uv > PERL_UNICODE_MAX)) {
671 if ((flags & (UTF8_WARN_SUPER|UTF8_CHECK_ONLY)) == UTF8_WARN_SUPER) {
111d382d 672 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "Code point 0x%04"UVXf" is not Unicode, may not be portable", uv));
949cf498
KW
673 }
674 if (flags & UTF8_DISALLOW_SUPER) {
675 goto disallowed;
676 }
677 }
678
679 /* Here, this is not considered a malformed character, so drop through
680 * to return it */
a0ed51b3 681 }
ba210ebe 682
a0ed51b3 683 return uv;
ba210ebe 684
949cf498
KW
685disallowed: /* Is disallowed, but otherwise not malformed. 'sv' will have been
686 set if there is to be a warning. */
687 if (!sv) {
688 dowarn = 0;
689 }
690
ba210ebe
JH
691malformed:
692
fcc8fcf6 693 if (flags & UTF8_CHECK_ONLY) {
ba210ebe 694 if (retlen)
10edeb5d 695 *retlen = ((STRLEN) -1);
ba210ebe
JH
696 return 0;
697 }
698
a0dbb045 699 if (dowarn) {
949cf498 700 if (! sv) {
5b311467 701 sv = newSVpvs_flags("Malformed UTF-8 character ", SVs_TEMP);
949cf498 702 }
5b311467 703
5e3d7cf5
KW
704 switch (warning) {
705 case 0: /* Intentionally empty. */ break;
706 case UTF8_WARN_EMPTY:
707 sv_catpvs(sv, "(empty string)");
708 break;
709 case UTF8_WARN_CONTINUATION:
710 Perl_sv_catpvf(aTHX_ sv, "(unexpected continuation byte 0x%02"UVxf", with no preceding start byte)", uv);
711 break;
712 case UTF8_WARN_NON_CONTINUATION:
713 if (s == s0)
714 Perl_sv_catpvf(aTHX_ sv, "(unexpected non-continuation byte 0x%02"UVxf", immediately after start byte 0x%02"UVxf")",
715 (UV)s[1], startbyte);
716 else {
717 const int len = (int)(s-s0);
718 Perl_sv_catpvf(aTHX_ sv, "(unexpected non-continuation byte 0x%02"UVxf", %d byte%s after start byte 0x%02"UVxf", expected %d bytes)",
719 (UV)s[1], len, len > 1 ? "s" : "", startbyte, (int)expectlen);
720 }
721
722 break;
723 case UTF8_WARN_SHORT:
724 Perl_sv_catpvf(aTHX_ sv, "(%d byte%s, need %d, after start byte 0x%02"UVxf")",
725 (int)curlen, curlen == 1 ? "" : "s", (int)expectlen, startbyte);
726 expectlen = curlen; /* distance for caller to skip */
727 break;
728 case UTF8_WARN_OVERFLOW:
729 Perl_sv_catpvf(aTHX_ sv, "(overflow at 0x%"UVxf", byte 0x%02x, after start byte 0x%02"UVxf")",
730 ouv, *s, startbyte);
731 break;
732 case UTF8_WARN_LONG:
733 Perl_sv_catpvf(aTHX_ sv, "(%d byte%s, need %d, after start byte 0x%02"UVxf")",
734 (int)expectlen, expectlen == 1 ? "": "s", UNISKIP(uv), startbyte);
735 break;
736 default:
737 sv_catpvs(sv, "(unknown reason)");
738 break;
739 }
a0dbb045 740
949cf498 741 if (sv) {
44f8325f 742 const char * const s = SvPVX_const(sv);
a0dbb045
JH
743
744 if (PL_op)
9014280d 745 Perl_warner(aTHX_ packWARN(WARN_UTF8),
53e06cf0 746 "%s in %s", s, OP_DESC(PL_op));
a0dbb045 747 else
9014280d 748 Perl_warner(aTHX_ packWARN(WARN_UTF8), "%s", s);
a0dbb045
JH
749 }
750 }
751
ba210ebe 752 if (retlen)
28d3d195 753 *retlen = expectlen ? expectlen : len;
ba210ebe 754
28d3d195 755 return 0;
a0ed51b3
LW
756}
757
8e84507e 758/*
87cea99e 759=for apidoc utf8_to_uvchr
9041c2e3 760
6ee84de2 761Returns the native code point of the first character in the string C<s>
1e54db1a 762which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
9041c2e3
NIS
763length, in bytes, of that character.
764
1e54db1a 765If C<s> does not point to a well-formed UTF-8 character, zero is
9041c2e3
NIS
766returned and retlen is set, if possible, to -1.
767
768=cut
769*/
770
ae0e24b1 771
9041c2e3 772UV
7fc63493 773Perl_utf8_to_uvchr(pTHX_ const U8 *s, STRLEN *retlen)
9041c2e3 774{
7918f24d
NC
775 PERL_ARGS_ASSERT_UTF8_TO_UVCHR;
776
1754c1a1 777 return utf8n_to_uvchr(s, UTF8_MAXBYTES, retlen,
ae0e24b1 778 ckWARN_d(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
9041c2e3
NIS
779}
780
781/*
87cea99e 782=for apidoc utf8_to_uvuni
9041c2e3
NIS
783
784Returns the Unicode code point of the first character in the string C<s>
1e54db1a 785which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
9041c2e3
NIS
786length, in bytes, of that character.
787
2bbc8d55 788This function should only be used when the returned UV is considered
9041c2e3
NIS
789an index into the Unicode semantic tables (e.g. swashes).
790
1e54db1a 791If C<s> does not point to a well-formed UTF-8 character, zero is
ba210ebe 792returned and retlen is set, if possible, to -1.
8e84507e
NIS
793
794=cut
795*/
796
797UV
7fc63493 798Perl_utf8_to_uvuni(pTHX_ const U8 *s, STRLEN *retlen)
8e84507e 799{
7918f24d
NC
800 PERL_ARGS_ASSERT_UTF8_TO_UVUNI;
801
9041c2e3 802 /* Call the low level routine asking for checks */
89ebb4a3 803 return Perl_utf8n_to_uvuni(aTHX_ s, UTF8_MAXBYTES, retlen,
ae0e24b1 804 ckWARN_d(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
8e84507e
NIS
805}
806
b76347f2 807/*
87cea99e 808=for apidoc utf8_length
b76347f2
JH
809
810Return the length of the UTF-8 char encoded string C<s> in characters.
02eb7b47
JH
811Stops at C<e> (inclusive). If C<e E<lt> s> or if the scan would end
812up past C<e>, croaks.
b76347f2
JH
813
814=cut
815*/
816
817STRLEN
35a4481c 818Perl_utf8_length(pTHX_ const U8 *s, const U8 *e)
b76347f2 819{
97aff369 820 dVAR;
b76347f2
JH
821 STRLEN len = 0;
822
7918f24d
NC
823 PERL_ARGS_ASSERT_UTF8_LENGTH;
824
8850bf83
JH
825 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g.
826 * the bitops (especially ~) can create illegal UTF-8.
827 * In other words: in Perl UTF-8 is not just for Unicode. */
828
a3b680e6
AL
829 if (e < s)
830 goto warn_and_return;
b76347f2 831 while (s < e) {
8e91ec7f
AV
832 if (!UTF8_IS_INVARIANT(*s))
833 s += UTF8SKIP(s);
834 else
835 s++;
836 len++;
837 }
838
839 if (e != s) {
840 len--;
841 warn_and_return:
9b387841
NC
842 if (PL_op)
843 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8),
844 "%s in %s", unees, OP_DESC(PL_op));
845 else
61a12c31 846 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), "%s", unees);
b76347f2
JH
847 }
848
849 return len;
850}
851
b06226ff 852/*
87cea99e 853=for apidoc utf8_distance
b06226ff 854
1e54db1a 855Returns the number of UTF-8 characters between the UTF-8 pointers C<a>
b06226ff
JH
856and C<b>.
857
858WARNING: use only if you *know* that the pointers point inside the
859same UTF-8 buffer.
860
37607a96
PK
861=cut
862*/
a0ed51b3 863
02eb7b47 864IV
35a4481c 865Perl_utf8_distance(pTHX_ const U8 *a, const U8 *b)
a0ed51b3 866{
7918f24d
NC
867 PERL_ARGS_ASSERT_UTF8_DISTANCE;
868
bf1665bc 869 return (a < b) ? -1 * (IV) utf8_length(a, b) : (IV) utf8_length(b, a);
a0ed51b3
LW
870}
871
b06226ff 872/*
87cea99e 873=for apidoc utf8_hop
b06226ff 874
8850bf83
JH
875Return the UTF-8 pointer C<s> displaced by C<off> characters, either
876forward or backward.
b06226ff
JH
877
878WARNING: do not use the following unless you *know* C<off> is within
8850bf83
JH
879the UTF-8 data pointed to by C<s> *and* that on entry C<s> is aligned
880on the first byte of character or just after the last byte of a character.
b06226ff 881
37607a96
PK
882=cut
883*/
a0ed51b3
LW
884
885U8 *
4373e329 886Perl_utf8_hop(pTHX_ const U8 *s, I32 off)
a0ed51b3 887{
7918f24d
NC
888 PERL_ARGS_ASSERT_UTF8_HOP;
889
96a5add6 890 PERL_UNUSED_CONTEXT;
8850bf83
JH
891 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g
892 * the bitops (especially ~) can create illegal UTF-8.
893 * In other words: in Perl UTF-8 is not just for Unicode. */
894
a0ed51b3
LW
895 if (off >= 0) {
896 while (off--)
897 s += UTF8SKIP(s);
898 }
899 else {
900 while (off++) {
901 s--;
8850bf83
JH
902 while (UTF8_IS_CONTINUATION(*s))
903 s--;
a0ed51b3
LW
904 }
905 }
4373e329 906 return (U8 *)s;
a0ed51b3
LW
907}
908
6940069f 909/*
fed3ba5d
NC
910=for apidoc bytes_cmp_utf8
911
912Compares the sequence of characters (stored as octets) in b, blen with the
913sequence of characters (stored as UTF-8) in u, ulen. Returns 0 if they are
914equal, -1 or -2 if the first string is less than the second string, +1 or +2
915if the first string is greater than the second string.
916
917-1 or +1 is returned if the shorter string was identical to the start of the
918longer string. -2 or +2 is returned if the was a difference between characters
919within the strings.
920
921=cut
922*/
923
924int
925Perl_bytes_cmp_utf8(pTHX_ const U8 *b, STRLEN blen, const U8 *u, STRLEN ulen)
926{
927 const U8 *const bend = b + blen;
928 const U8 *const uend = u + ulen;
929
930 PERL_ARGS_ASSERT_BYTES_CMP_UTF8;
931
932 PERL_UNUSED_CONTEXT;
933
934 while (b < bend && u < uend) {
935 U8 c = *u++;
936 if (!UTF8_IS_INVARIANT(c)) {
937 if (UTF8_IS_DOWNGRADEABLE_START(c)) {
938 if (u < uend) {
939 U8 c1 = *u++;
940 if (UTF8_IS_CONTINUATION(c1)) {
356979f4 941 c = UNI_TO_NATIVE(TWO_BYTE_UTF8_TO_UNI(c, c1));
fed3ba5d
NC
942 } else {
943 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8),
944 "Malformed UTF-8 character "
945 "(unexpected non-continuation byte 0x%02x"
946 ", immediately after start byte 0x%02x)"
947 /* Dear diag.t, it's in the pod. */
948 "%s%s", c1, c,
949 PL_op ? " in " : "",
950 PL_op ? OP_DESC(PL_op) : "");
951 return -2;
952 }
953 } else {
954 if (PL_op)
955 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8),
956 "%s in %s", unees, OP_DESC(PL_op));
957 else
61a12c31 958 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), "%s", unees);
fed3ba5d
NC
959 return -2; /* Really want to return undef :-) */
960 }
961 } else {
962 return -2;
963 }
964 }
965 if (*b != c) {
966 return *b < c ? -2 : +2;
967 }
968 ++b;
969 }
970
971 if (b == bend && u == uend)
972 return 0;
973
974 return b < bend ? +1 : -1;
975}
976
977/*
87cea99e 978=for apidoc utf8_to_bytes
6940069f 979
2bbc8d55 980Converts a string C<s> of length C<len> from UTF-8 into native byte encoding.
246fae53
MG
981Unlike C<bytes_to_utf8>, this over-writes the original string, and
982updates len to contain the new length.
67e989fb 983Returns zero on failure, setting C<len> to -1.
6940069f 984
95be277c
NC
985If you need a copy of the string, see C<bytes_from_utf8>.
986
6940069f
GS
987=cut
988*/
989
990U8 *
37607a96 991Perl_utf8_to_bytes(pTHX_ U8 *s, STRLEN *len)
6940069f 992{
d4c19fe8
AL
993 U8 * const save = s;
994 U8 * const send = s + *len;
6940069f 995 U8 *d;
246fae53 996
7918f24d
NC
997 PERL_ARGS_ASSERT_UTF8_TO_BYTES;
998
1e54db1a 999 /* ensure valid UTF-8 and chars < 256 before updating string */
d4c19fe8 1000 while (s < send) {
dcad2880
JH
1001 U8 c = *s++;
1002
1d72bdf6
NIS
1003 if (!UTF8_IS_INVARIANT(c) &&
1004 (!UTF8_IS_DOWNGRADEABLE_START(c) || (s >= send)
1005 || !(c = *s++) || !UTF8_IS_CONTINUATION(c))) {
10edeb5d 1006 *len = ((STRLEN) -1);
dcad2880
JH
1007 return 0;
1008 }
246fae53 1009 }
dcad2880
JH
1010
1011 d = s = save;
6940069f 1012 while (s < send) {
ed646e6e 1013 STRLEN ulen;
9041c2e3 1014 *d++ = (U8)utf8_to_uvchr(s, &ulen);
ed646e6e 1015 s += ulen;
6940069f
GS
1016 }
1017 *d = '\0';
246fae53 1018 *len = d - save;
6940069f
GS
1019 return save;
1020}
1021
1022/*
87cea99e 1023=for apidoc bytes_from_utf8
f9a63242 1024
2bbc8d55 1025Converts a string C<s> of length C<len> from UTF-8 into native byte encoding.
35a4481c 1026Unlike C<utf8_to_bytes> but like C<bytes_to_utf8>, returns a pointer to
ef9edfd0
JH
1027the newly-created string, and updates C<len> to contain the new
1028length. Returns the original string if no conversion occurs, C<len>
1029is unchanged. Do nothing if C<is_utf8> points to 0. Sets C<is_utf8> to
2bbc8d55
SP
10300 if C<s> is converted or consisted entirely of characters that are invariant
1031in utf8 (i.e., US-ASCII on non-EBCDIC machines).
f9a63242 1032
37607a96
PK
1033=cut
1034*/
f9a63242
JH
1035
1036U8 *
e1ec3a88 1037Perl_bytes_from_utf8(pTHX_ const U8 *s, STRLEN *len, bool *is_utf8)
f9a63242 1038{
f9a63242 1039 U8 *d;
e1ec3a88
AL
1040 const U8 *start = s;
1041 const U8 *send;
f9a63242
JH
1042 I32 count = 0;
1043
7918f24d
NC
1044 PERL_ARGS_ASSERT_BYTES_FROM_UTF8;
1045
96a5add6 1046 PERL_UNUSED_CONTEXT;
f9a63242 1047 if (!*is_utf8)
73d840c0 1048 return (U8 *)start;
f9a63242 1049
1e54db1a 1050 /* ensure valid UTF-8 and chars < 256 before converting string */
f9a63242 1051 for (send = s + *len; s < send;) {
e1ec3a88 1052 U8 c = *s++;
1d72bdf6 1053 if (!UTF8_IS_INVARIANT(c)) {
db42d148
NIS
1054 if (UTF8_IS_DOWNGRADEABLE_START(c) && s < send &&
1055 (c = *s++) && UTF8_IS_CONTINUATION(c))
1056 count++;
1057 else
73d840c0 1058 return (U8 *)start;
db42d148 1059 }
f9a63242
JH
1060 }
1061
35da51f7 1062 *is_utf8 = FALSE;
f9a63242 1063
212542aa 1064 Newx(d, (*len) - count + 1, U8);
ef9edfd0 1065 s = start; start = d;
f9a63242
JH
1066 while (s < send) {
1067 U8 c = *s++;
c4d5f83a
NIS
1068 if (!UTF8_IS_INVARIANT(c)) {
1069 /* Then it is two-byte encoded */
356979f4 1070 c = UNI_TO_NATIVE(TWO_BYTE_UTF8_TO_UNI(c, *s++));
c4d5f83a
NIS
1071 }
1072 *d++ = c;
f9a63242
JH
1073 }
1074 *d = '\0';
1075 *len = d - start;
73d840c0 1076 return (U8 *)start;
f9a63242
JH
1077}
1078
1079/*
87cea99e 1080=for apidoc bytes_to_utf8
6940069f 1081
ff97e5cf
KW
1082Converts a string C<s> of length C<len> bytes from the native encoding into
1083UTF-8.
6662521e 1084Returns a pointer to the newly-created string, and sets C<len> to
ff97e5cf 1085reflect the new length in bytes.
6940069f 1086
2bbc8d55
SP
1087A NUL character will be written after the end of the string.
1088
1089If you want to convert to UTF-8 from encodings other than
1090the native (Latin1 or EBCDIC),
c9ada85f
JH
1091see sv_recode_to_utf8().
1092
497711e7 1093=cut
6940069f
GS
1094*/
1095
c682ebef
FC
1096/* This logic is duplicated in sv_catpvn_flags, so any bug fixes will
1097 likewise need duplication. */
1098
6940069f 1099U8*
35a4481c 1100Perl_bytes_to_utf8(pTHX_ const U8 *s, STRLEN *len)
6940069f 1101{
35a4481c 1102 const U8 * const send = s + (*len);
6940069f
GS
1103 U8 *d;
1104 U8 *dst;
7918f24d
NC
1105
1106 PERL_ARGS_ASSERT_BYTES_TO_UTF8;
96a5add6 1107 PERL_UNUSED_CONTEXT;
6940069f 1108
212542aa 1109 Newx(d, (*len) * 2 + 1, U8);
6940069f
GS
1110 dst = d;
1111
1112 while (s < send) {
35a4481c 1113 const UV uv = NATIVE_TO_ASCII(*s++);
c4d5f83a 1114 if (UNI_IS_INVARIANT(uv))
eb160463 1115 *d++ = (U8)UTF_TO_NATIVE(uv);
6940069f 1116 else {
eb160463
GS
1117 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
1118 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
6940069f
GS
1119 }
1120 }
1121 *d = '\0';
6662521e 1122 *len = d-dst;
6940069f
GS
1123 return dst;
1124}
1125
a0ed51b3 1126/*
dea0fc0b 1127 * Convert native (big-endian) or reversed (little-endian) UTF-16 to UTF-8.
a0ed51b3
LW
1128 *
1129 * Destination must be pre-extended to 3/2 source. Do not use in-place.
1130 * We optimize for native, for obvious reasons. */
1131
1132U8*
dea0fc0b 1133Perl_utf16_to_utf8(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen)
a0ed51b3 1134{
dea0fc0b
JH
1135 U8* pend;
1136 U8* dstart = d;
1137
7918f24d
NC
1138 PERL_ARGS_ASSERT_UTF16_TO_UTF8;
1139
dea0fc0b 1140 if (bytelen & 1)
f5992bc4 1141 Perl_croak(aTHX_ "panic: utf16_to_utf8: odd bytelen %"UVuf, (UV)bytelen);
dea0fc0b
JH
1142
1143 pend = p + bytelen;
1144
a0ed51b3 1145 while (p < pend) {
dea0fc0b
JH
1146 UV uv = (p[0] << 8) + p[1]; /* UTF-16BE */
1147 p += 2;
a0ed51b3 1148 if (uv < 0x80) {
e294cc5d
JH
1149#ifdef EBCDIC
1150 *d++ = UNI_TO_NATIVE(uv);
1151#else
eb160463 1152 *d++ = (U8)uv;
e294cc5d 1153#endif
a0ed51b3
LW
1154 continue;
1155 }
1156 if (uv < 0x800) {
eb160463
GS
1157 *d++ = (U8)(( uv >> 6) | 0xc0);
1158 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
1159 continue;
1160 }
52b9aa85 1161 if (uv >= 0xd800 && uv <= 0xdbff) { /* surrogates */
01ea242b 1162 if (p >= pend) {
dea0fc0b 1163 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
01ea242b
NC
1164 } else {
1165 UV low = (p[0] << 8) + p[1];
1166 p += 2;
52b9aa85 1167 if (low < 0xdc00 || low > 0xdfff)
01ea242b
NC
1168 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
1169 uv = ((uv - 0xd800) << 10) + (low - 0xdc00) + 0x10000;
1170 }
dbde1951
NC
1171 } else if (uv >= 0xdc00 && uv <= 0xdfff) {
1172 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
a0ed51b3
LW
1173 }
1174 if (uv < 0x10000) {
eb160463
GS
1175 *d++ = (U8)(( uv >> 12) | 0xe0);
1176 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
1177 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
1178 continue;
1179 }
1180 else {
eb160463
GS
1181 *d++ = (U8)(( uv >> 18) | 0xf0);
1182 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
1183 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
1184 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
1185 continue;
1186 }
1187 }
dea0fc0b 1188 *newlen = d - dstart;
a0ed51b3
LW
1189 return d;
1190}
1191
1192/* Note: this one is slightly destructive of the source. */
1193
1194U8*
dea0fc0b 1195Perl_utf16_to_utf8_reversed(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen)
a0ed51b3
LW
1196{
1197 U8* s = (U8*)p;
d4c19fe8 1198 U8* const send = s + bytelen;
7918f24d
NC
1199
1200 PERL_ARGS_ASSERT_UTF16_TO_UTF8_REVERSED;
1201
e0ea5e2d
NC
1202 if (bytelen & 1)
1203 Perl_croak(aTHX_ "panic: utf16_to_utf8_reversed: odd bytelen %"UVuf,
1204 (UV)bytelen);
1205
a0ed51b3 1206 while (s < send) {
d4c19fe8 1207 const U8 tmp = s[0];
a0ed51b3
LW
1208 s[0] = s[1];
1209 s[1] = tmp;
1210 s += 2;
1211 }
dea0fc0b 1212 return utf16_to_utf8(p, d, bytelen, newlen);
a0ed51b3
LW
1213}
1214
c3fd2246
KW
1215/* for now these are all defined (inefficiently) in terms of the utf8 versions.
1216 * Note that the macros in handy.h that call these short-circuit calling them
1217 * for Latin-1 range inputs */
a0ed51b3
LW
1218
1219bool
84afefe6 1220Perl_is_uni_alnum(pTHX_ UV c)
a0ed51b3 1221{
89ebb4a3 1222 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1223 uvchr_to_utf8(tmpbuf, c);
a0ed51b3
LW
1224 return is_utf8_alnum(tmpbuf);
1225}
1226
1227bool
84afefe6 1228Perl_is_uni_idfirst(pTHX_ UV c)
a0ed51b3 1229{
89ebb4a3 1230 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1231 uvchr_to_utf8(tmpbuf, c);
a0ed51b3
LW
1232 return is_utf8_idfirst(tmpbuf);
1233}
1234
1235bool
84afefe6 1236Perl_is_uni_alpha(pTHX_ UV c)
a0ed51b3 1237{
89ebb4a3 1238 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1239 uvchr_to_utf8(tmpbuf, c);
a0ed51b3
LW
1240 return is_utf8_alpha(tmpbuf);
1241}
1242
1243bool
84afefe6 1244Perl_is_uni_ascii(pTHX_ UV c)
4d61ec05 1245{
bc39fe24 1246 return isASCII(c);
4d61ec05
GS
1247}
1248
1249bool
84afefe6 1250Perl_is_uni_space(pTHX_ UV c)
a0ed51b3 1251{
89ebb4a3 1252 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1253 uvchr_to_utf8(tmpbuf, c);
a0ed51b3
LW
1254 return is_utf8_space(tmpbuf);
1255}
1256
1257bool
84afefe6 1258Perl_is_uni_digit(pTHX_ UV c)
a0ed51b3 1259{
89ebb4a3 1260 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1261 uvchr_to_utf8(tmpbuf, c);
a0ed51b3
LW
1262 return is_utf8_digit(tmpbuf);
1263}
1264
1265bool
84afefe6 1266Perl_is_uni_upper(pTHX_ UV c)
a0ed51b3 1267{
89ebb4a3 1268 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1269 uvchr_to_utf8(tmpbuf, c);
a0ed51b3
LW
1270 return is_utf8_upper(tmpbuf);
1271}
1272
1273bool
84afefe6 1274Perl_is_uni_lower(pTHX_ UV c)
a0ed51b3 1275{
89ebb4a3 1276 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1277 uvchr_to_utf8(tmpbuf, c);
a0ed51b3
LW
1278 return is_utf8_lower(tmpbuf);
1279}
1280
1281bool
84afefe6 1282Perl_is_uni_cntrl(pTHX_ UV c)
b8c5462f 1283{
7b952154 1284 return isCNTRL_L1(c);
b8c5462f
JH
1285}
1286
1287bool
84afefe6 1288Perl_is_uni_graph(pTHX_ UV c)
b8c5462f 1289{
89ebb4a3 1290 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1291 uvchr_to_utf8(tmpbuf, c);
b8c5462f
JH
1292 return is_utf8_graph(tmpbuf);
1293}
1294
1295bool
84afefe6 1296Perl_is_uni_print(pTHX_ UV c)
a0ed51b3 1297{
89ebb4a3 1298 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1299 uvchr_to_utf8(tmpbuf, c);
a0ed51b3
LW
1300 return is_utf8_print(tmpbuf);
1301}
1302
b8c5462f 1303bool
84afefe6 1304Perl_is_uni_punct(pTHX_ UV c)
b8c5462f 1305{
89ebb4a3 1306 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1307 uvchr_to_utf8(tmpbuf, c);
b8c5462f
JH
1308 return is_utf8_punct(tmpbuf);
1309}
1310
4d61ec05 1311bool
84afefe6 1312Perl_is_uni_xdigit(pTHX_ UV c)
4d61ec05 1313{
89ebb4a3 1314 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
230880c1 1315 uvchr_to_utf8(tmpbuf, c);
4d61ec05
GS
1316 return is_utf8_xdigit(tmpbuf);
1317}
1318
3a4c58c9
KW
1319UV
1320Perl__to_upper_title_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp, const char S_or_s)
1321{
1322 /* We have the latin1-range values compiled into the core, so just use
1323 * those, converting the result to utf8. The only difference between upper
1324 * and title case in this range is that LATIN_SMALL_LETTER_SHARP_S is
1325 * either "SS" or "Ss". Which one to use is passed into the routine in
1326 * 'S_or_s' to avoid a test */
1327
1328 UV converted = toUPPER_LATIN1_MOD(c);
1329
1330 PERL_ARGS_ASSERT__TO_UPPER_TITLE_LATIN1;
1331
1332 assert(S_or_s == 'S' || S_or_s == 's');
1333
1334 if (UNI_IS_INVARIANT(converted)) { /* No difference between the two for
1335 characters in this range */
1336 *p = (U8) converted;
1337 *lenp = 1;
1338 return converted;
1339 }
1340
1341 /* toUPPER_LATIN1_MOD gives the correct results except for three outliers,
1342 * which it maps to one of them, so as to only have to have one check for
1343 * it in the main case */
1344 if (UNLIKELY(converted == LATIN_SMALL_LETTER_Y_WITH_DIAERESIS)) {
1345 switch (c) {
1346 case LATIN_SMALL_LETTER_Y_WITH_DIAERESIS:
1347 converted = LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS;
1348 break;
1349 case MICRO_SIGN:
1350 converted = GREEK_CAPITAL_LETTER_MU;
1351 break;
1352 case LATIN_SMALL_LETTER_SHARP_S:
1353 *(p)++ = 'S';
1354 *p = S_or_s;
1355 *lenp = 2;
1356 return 'S';
1357 default:
1358 Perl_croak(aTHX_ "panic: to_upper_title_latin1 did not expect '%c' to map to '%c'", c, LATIN_SMALL_LETTER_Y_WITH_DIAERESIS);
1359 /* NOTREACHED */
1360 }
1361 }
1362
1363 *(p)++ = UTF8_TWO_BYTE_HI(converted);
1364 *p = UTF8_TWO_BYTE_LO(converted);
1365 *lenp = 2;
1366
1367 return converted;
1368}
1369
50bda2c3
KW
1370/* Call the function to convert a UTF-8 encoded character to the specified case.
1371 * Note that there may be more than one character in the result.
1372 * INP is a pointer to the first byte of the input character
1373 * OUTP will be set to the first byte of the string of changed characters. It
1374 * needs to have space for UTF8_MAXBYTES_CASE+1 bytes
1375 * LENP will be set to the length in bytes of the string of changed characters
1376 *
1377 * The functions return the ordinal of the first character in the string of OUTP */
1378#define CALL_UPPER_CASE(INP, OUTP, LENP) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_toupper, "ToUc", "utf8::ToSpecUpper")
1379#define CALL_TITLE_CASE(INP, OUTP, LENP) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_totitle, "ToTc", "utf8::ToSpecTitle")
1380#define CALL_LOWER_CASE(INP, OUTP, LENP) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_tolower, "ToLc", "utf8::ToSpecLower")
1381
1382/* This additionally has the input parameter SPECIALS, which if non-zero will
1383 * cause this to use the SPECIALS hash for folding (meaning get full case
1384 * folding); otherwise, when zero, this implies a simple case fold */
1385#define CALL_FOLD_CASE(INP, OUTP, LENP, SPECIALS) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_tofold, "ToCf", (SPECIALS) ? "utf8::ToSpecFold" : NULL)
c3fd2246 1386
84afefe6
JH
1387UV
1388Perl_to_uni_upper(pTHX_ UV c, U8* p, STRLEN *lenp)
a0ed51b3 1389{
3a4c58c9
KW
1390 dVAR;
1391
c3fd2246
KW
1392 /* Convert the Unicode character whose ordinal is c to its uppercase
1393 * version and store that in UTF-8 in p and its length in bytes in lenp.
1394 * Note that the p needs to be at least UTF8_MAXBYTES_CASE+1 bytes since
1395 * the changed version may be longer than the original character.
1396 *
1397 * The ordinal of the first character of the changed version is returned
1398 * (but note, as explained above, that there may be more.) */
1399
7918f24d
NC
1400 PERL_ARGS_ASSERT_TO_UNI_UPPER;
1401
3a4c58c9
KW
1402 if (c < 256) {
1403 return _to_upper_title_latin1((U8) c, p, lenp, 'S');
1404 }
1405
0ebc6274 1406 uvchr_to_utf8(p, c);
3a4c58c9 1407 return CALL_UPPER_CASE(p, p, lenp);
a0ed51b3
LW
1408}
1409
84afefe6
JH
1410UV
1411Perl_to_uni_title(pTHX_ UV c, U8* p, STRLEN *lenp)
a0ed51b3 1412{
3a4c58c9
KW
1413 dVAR;
1414
7918f24d
NC
1415 PERL_ARGS_ASSERT_TO_UNI_TITLE;
1416
3a4c58c9
KW
1417 if (c < 256) {
1418 return _to_upper_title_latin1((U8) c, p, lenp, 's');
1419 }
1420
0ebc6274 1421 uvchr_to_utf8(p, c);
3a4c58c9 1422 return CALL_TITLE_CASE(p, p, lenp);
a0ed51b3
LW
1423}
1424
afc16117
KW
1425STATIC U8
1426S_to_lower_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp)
1427{
1428 /* We have the latin1-range values compiled into the core, so just use
1429 * those, converting the result to utf8. Since the result is always just
1430 * one character, we allow p to be NULL */
1431
1432 U8 converted = toLOWER_LATIN1(c);
1433
1434 if (p != NULL) {
1435 if (UNI_IS_INVARIANT(converted)) {
1436 *p = converted;
1437 *lenp = 1;
1438 }
1439 else {
1440 *p = UTF8_TWO_BYTE_HI(converted);
1441 *(p+1) = UTF8_TWO_BYTE_LO(converted);
1442 *lenp = 2;
1443 }
1444 }
1445 return converted;
1446}
1447
84afefe6
JH
1448UV
1449Perl_to_uni_lower(pTHX_ UV c, U8* p, STRLEN *lenp)
a0ed51b3 1450{
968c5e6a
KW
1451 dVAR;
1452
7918f24d
NC
1453 PERL_ARGS_ASSERT_TO_UNI_LOWER;
1454
afc16117
KW
1455 if (c < 256) {
1456 return to_lower_latin1((U8) c, p, lenp);
bca00c02
KW
1457 }
1458
afc16117 1459 uvchr_to_utf8(p, c);
968c5e6a 1460 return CALL_LOWER_CASE(p, p, lenp);
a0ed51b3
LW
1461}
1462
84afefe6 1463UV
a1dde8de
KW
1464Perl__to_fold_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp, const U8 flags)
1465{
1466 UV converted;
1467
1468 PERL_ARGS_ASSERT__TO_FOLD_LATIN1;
1469
1470 if (c == MICRO_SIGN) {
1471 converted = GREEK_SMALL_LETTER_MU;
1472 }
1473 else if (flags && c == LATIN_SMALL_LETTER_SHARP_S) {
1474 *(p)++ = 's';
1475 *p = 's';
1476 *lenp = 2;
1477 return 's';
1478 }
1479 else { /* In this range the fold of all other characters is their lower
1480 case */
1481 converted = toLOWER_LATIN1(c);
1482 }
1483
1484 if (UNI_IS_INVARIANT(converted)) {
1485 *p = (U8) converted;
1486 *lenp = 1;
1487 }
1488 else {
1489 *(p)++ = UTF8_TWO_BYTE_HI(converted);
1490 *p = UTF8_TWO_BYTE_LO(converted);
1491 *lenp = 2;
1492 }
1493
1494 return converted;
1495}
1496
1497UV
36bb2ab6 1498Perl__to_uni_fold_flags(pTHX_ UV c, U8* p, STRLEN *lenp, U8 flags)
84afefe6 1499{
4b593389
KW
1500
1501 /* Not currently externally documented, and subject to change, <flags> is
1502 * non-zero iff full folding is to be used */
1503
36bb2ab6 1504 PERL_ARGS_ASSERT__TO_UNI_FOLD_FLAGS;
7918f24d 1505
a1dde8de
KW
1506 if (c < 256) {
1507 return _to_fold_latin1((U8) c, p, lenp, flags);
1508 }
1509
0ebc6274 1510 uvchr_to_utf8(p, c);
a1dde8de 1511 return CALL_FOLD_CASE(p, p, lenp, flags);
84afefe6
JH
1512}
1513
a0ed51b3
LW
1514/* for now these all assume no locale info available for Unicode > 255 */
1515
1516bool
84afefe6 1517Perl_is_uni_alnum_lc(pTHX_ UV c)
a0ed51b3
LW
1518{
1519 return is_uni_alnum(c); /* XXX no locale support yet */
1520}
1521
1522bool
84afefe6 1523Perl_is_uni_idfirst_lc(pTHX_ UV c)
a0ed51b3
LW
1524{
1525 return is_uni_idfirst(c); /* XXX no locale support yet */
1526}
1527
1528bool
84afefe6 1529Perl_is_uni_alpha_lc(pTHX_ UV c)
a0ed51b3
LW
1530{
1531 return is_uni_alpha(c); /* XXX no locale support yet */
1532}
1533
1534bool
84afefe6 1535Perl_is_uni_ascii_lc(pTHX_ UV c)
4d61ec05
GS
1536{
1537 return is_uni_ascii(c); /* XXX no locale support yet */
1538}
1539
1540bool
84afefe6 1541Perl_is_uni_space_lc(pTHX_ UV c)
a0ed51b3
LW
1542{
1543 return is_uni_space(c); /* XXX no locale support yet */
1544}
1545
1546bool
84afefe6 1547Perl_is_uni_digit_lc(pTHX_ UV c)
a0ed51b3
LW
1548{
1549 return is_uni_digit(c); /* XXX no locale support yet */
1550}
1551
1552bool
84afefe6 1553Perl_is_uni_upper_lc(pTHX_ UV c)
a0ed51b3
LW
1554{
1555 return is_uni_upper(c); /* XXX no locale support yet */
1556}
1557
1558bool
84afefe6 1559Perl_is_uni_lower_lc(pTHX_ UV c)
a0ed51b3
LW
1560{
1561 return is_uni_lower(c); /* XXX no locale support yet */
1562}
1563
1564bool
84afefe6 1565Perl_is_uni_cntrl_lc(pTHX_ UV c)
b8c5462f
JH
1566{
1567 return is_uni_cntrl(c); /* XXX no locale support yet */
1568}
1569
1570bool
84afefe6 1571Perl_is_uni_graph_lc(pTHX_ UV c)
b8c5462f
JH
1572{
1573 return is_uni_graph(c); /* XXX no locale support yet */
1574}
1575
1576bool
84afefe6 1577Perl_is_uni_print_lc(pTHX_ UV c)
a0ed51b3
LW
1578{
1579 return is_uni_print(c); /* XXX no locale support yet */
1580}
1581
b8c5462f 1582bool
84afefe6 1583Perl_is_uni_punct_lc(pTHX_ UV c)
b8c5462f
JH
1584{
1585 return is_uni_punct(c); /* XXX no locale support yet */
1586}
1587
4d61ec05 1588bool
84afefe6 1589Perl_is_uni_xdigit_lc(pTHX_ UV c)
4d61ec05
GS
1590{
1591 return is_uni_xdigit(c); /* XXX no locale support yet */
1592}
1593
b7ac61fa
JH
1594U32
1595Perl_to_uni_upper_lc(pTHX_ U32 c)
1596{
ee099d14
JH
1597 /* XXX returns only the first character -- do not use XXX */
1598 /* XXX no locale support yet */
1599 STRLEN len;
89ebb4a3 1600 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
ee099d14 1601 return (U32)to_uni_upper(c, tmpbuf, &len);
b7ac61fa
JH
1602}
1603
1604U32
1605Perl_to_uni_title_lc(pTHX_ U32 c)
1606{
ee099d14
JH
1607 /* XXX returns only the first character XXX -- do not use XXX */
1608 /* XXX no locale support yet */
1609 STRLEN len;
89ebb4a3 1610 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
ee099d14 1611 return (U32)to_uni_title(c, tmpbuf, &len);
b7ac61fa
JH
1612}
1613
1614U32
1615Perl_to_uni_lower_lc(pTHX_ U32 c)
1616{
ee099d14
JH
1617 /* XXX returns only the first character -- do not use XXX */
1618 /* XXX no locale support yet */
1619 STRLEN len;
89ebb4a3 1620 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
ee099d14 1621 return (U32)to_uni_lower(c, tmpbuf, &len);
b7ac61fa
JH
1622}
1623
7452cf6a 1624static bool
5141f98e 1625S_is_utf8_common(pTHX_ const U8 *const p, SV **swash,
bde6a22d
NC
1626 const char *const swashname)
1627{
97aff369 1628 dVAR;
7918f24d
NC
1629
1630 PERL_ARGS_ASSERT_IS_UTF8_COMMON;
1631
bde6a22d
NC
1632 if (!is_utf8_char(p))
1633 return FALSE;
1634 if (!*swash)
711a919c 1635 *swash = swash_init("utf8", swashname, &PL_sv_undef, 1, 0);
bde6a22d
NC
1636 return swash_fetch(*swash, p, TRUE) != 0;
1637}
1638
1639bool
7fc63493 1640Perl_is_utf8_alnum(pTHX_ const U8 *p)
a0ed51b3 1641{
97aff369 1642 dVAR;
7918f24d
NC
1643
1644 PERL_ARGS_ASSERT_IS_UTF8_ALNUM;
1645
671c33bf
NC
1646 /* NOTE: "IsWord", not "IsAlnum", since Alnum is a true
1647 * descendant of isalnum(3), in other words, it doesn't
1648 * contain the '_'. --jhi */
d4c19fe8 1649 return is_utf8_common(p, &PL_utf8_alnum, "IsWord");
a0ed51b3
LW
1650}
1651
1652bool
7fc63493 1653Perl_is_utf8_idfirst(pTHX_ const U8 *p) /* The naming is historical. */
a0ed51b3 1654{
97aff369 1655 dVAR;
7918f24d
NC
1656
1657 PERL_ARGS_ASSERT_IS_UTF8_IDFIRST;
1658
82686b01
JH
1659 if (*p == '_')
1660 return TRUE;
bde6a22d 1661 /* is_utf8_idstart would be more logical. */
d4c19fe8 1662 return is_utf8_common(p, &PL_utf8_idstart, "IdStart");
82686b01
JH
1663}
1664
1665bool
c11ff943
KW
1666Perl_is_utf8_xidfirst(pTHX_ const U8 *p) /* The naming is historical. */
1667{
1668 dVAR;
1669
1670 PERL_ARGS_ASSERT_IS_UTF8_XIDFIRST;
1671
1672 if (*p == '_')
1673 return TRUE;
1674 /* is_utf8_idstart would be more logical. */
1675 return is_utf8_common(p, &PL_utf8_xidstart, "XIdStart");
1676}
1677
1678bool
b6912c02
KW
1679Perl__is_utf8__perl_idstart(pTHX_ const U8 *p)
1680{
1681 dVAR;
1682
1683 PERL_ARGS_ASSERT__IS_UTF8__PERL_IDSTART;
1684
1685 return is_utf8_common(p, &PL_utf8_perl_idstart, "_Perl_IDStart");
1686}
1687
1688bool
7fc63493 1689Perl_is_utf8_idcont(pTHX_ const U8 *p)
82686b01 1690{
97aff369 1691 dVAR;
7918f24d
NC
1692
1693 PERL_ARGS_ASSERT_IS_UTF8_IDCONT;
1694
d4c19fe8 1695 return is_utf8_common(p, &PL_utf8_idcont, "IdContinue");
a0ed51b3
LW
1696}
1697
1698bool
c11ff943
KW
1699Perl_is_utf8_xidcont(pTHX_ const U8 *p)
1700{
1701 dVAR;
1702
1703 PERL_ARGS_ASSERT_IS_UTF8_XIDCONT;
1704
c11ff943
KW
1705 return is_utf8_common(p, &PL_utf8_idcont, "XIdContinue");
1706}
1707
1708bool
7fc63493 1709Perl_is_utf8_alpha(pTHX_ const U8 *p)
a0ed51b3 1710{
97aff369 1711 dVAR;
7918f24d
NC
1712
1713 PERL_ARGS_ASSERT_IS_UTF8_ALPHA;
1714
d4c19fe8 1715 return is_utf8_common(p, &PL_utf8_alpha, "IsAlpha");
a0ed51b3
LW
1716}
1717
1718bool
7fc63493 1719Perl_is_utf8_ascii(pTHX_ const U8 *p)
b8c5462f 1720{
97aff369 1721 dVAR;
7918f24d
NC
1722
1723 PERL_ARGS_ASSERT_IS_UTF8_ASCII;
1724
bc39fe24
KW
1725 /* ASCII characters are the same whether in utf8 or not. So the macro
1726 * works on both utf8 and non-utf8 representations. */
1727 return isASCII(*p);
b8c5462f
JH
1728}
1729
1730bool
7fc63493 1731Perl_is_utf8_space(pTHX_ const U8 *p)
a0ed51b3 1732{
97aff369 1733 dVAR;
7918f24d
NC
1734
1735 PERL_ARGS_ASSERT_IS_UTF8_SPACE;
1736
a34094a9 1737 return is_utf8_common(p, &PL_utf8_space, "IsXPerlSpace");
a0ed51b3
LW
1738}
1739
1740bool
d1eb3177
YO
1741Perl_is_utf8_perl_space(pTHX_ const U8 *p)
1742{
1743 dVAR;
1744
1745 PERL_ARGS_ASSERT_IS_UTF8_PERL_SPACE;
1746
c4428693
KW
1747 /* Only true if is an ASCII space-like character, and ASCII is invariant
1748 * under utf8, so can just use the macro */
1749 return isSPACE_A(*p);
d1eb3177
YO
1750}
1751
1752bool
1753Perl_is_utf8_perl_word(pTHX_ const U8 *p)
1754{
1755 dVAR;
1756
1757 PERL_ARGS_ASSERT_IS_UTF8_PERL_WORD;
1758
c4428693
KW
1759 /* Only true if is an ASCII word character, and ASCII is invariant
1760 * under utf8, so can just use the macro */
1761 return isWORDCHAR_A(*p);
d1eb3177
YO
1762}
1763
1764bool
7fc63493 1765Perl_is_utf8_digit(pTHX_ const U8 *p)
a0ed51b3 1766{
97aff369 1767 dVAR;
7918f24d
NC
1768
1769 PERL_ARGS_ASSERT_IS_UTF8_DIGIT;
1770
d4c19fe8 1771 return is_utf8_common(p, &PL_utf8_digit, "IsDigit");
a0ed51b3
LW
1772}
1773
1774bool
d1eb3177
YO
1775Perl_is_utf8_posix_digit(pTHX_ const U8 *p)
1776{
1777 dVAR;
1778
1779 PERL_ARGS_ASSERT_IS_UTF8_POSIX_DIGIT;
1780
c4428693
KW
1781 /* Only true if is an ASCII digit character, and ASCII is invariant
1782 * under utf8, so can just use the macro */
1783 return isDIGIT_A(*p);
d1eb3177
YO
1784}
1785
1786bool
7fc63493 1787Perl_is_utf8_upper(pTHX_ const U8 *p)
a0ed51b3 1788{
97aff369 1789 dVAR;
7918f24d
NC
1790
1791 PERL_ARGS_ASSERT_IS_UTF8_UPPER;
1792
d4c19fe8 1793 return is_utf8_common(p, &PL_utf8_upper, "IsUppercase");
a0ed51b3
LW
1794}
1795
1796bool
7fc63493 1797Perl_is_utf8_lower(pTHX_ const U8 *p)
a0ed51b3 1798{
97aff369 1799 dVAR;
7918f24d
NC
1800
1801 PERL_ARGS_ASSERT_IS_UTF8_LOWER;
1802
d4c19fe8 1803 return is_utf8_common(p, &PL_utf8_lower, "IsLowercase");
a0ed51b3
LW
1804}
1805
1806bool
7fc63493 1807Perl_is_utf8_cntrl(pTHX_ const U8 *p)
b8c5462f 1808{
97aff369 1809 dVAR;
7918f24d
NC
1810
1811 PERL_ARGS_ASSERT_IS_UTF8_CNTRL;
1812
7b952154
KW
1813 if (isASCII(*p)) {
1814 return isCNTRL_A(*p);
1815 }
1816
1817 /* All controls are in Latin1 */
1818 if (! UTF8_IS_DOWNGRADEABLE_START(*p)) {
1819 return 0;
1820 }
1821 return isCNTRL_L1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)));
b8c5462f
JH
1822}
1823
1824bool
7fc63493 1825Perl_is_utf8_graph(pTHX_ const U8 *p)
b8c5462f 1826{
97aff369 1827 dVAR;
7918f24d
NC
1828
1829 PERL_ARGS_ASSERT_IS_UTF8_GRAPH;
1830
d4c19fe8 1831 return is_utf8_common(p, &PL_utf8_graph, "IsGraph");
b8c5462f
JH
1832}
1833
1834bool
7fc63493 1835Perl_is_utf8_print(pTHX_ const U8 *p)
a0ed51b3 1836{
97aff369 1837 dVAR;
7918f24d
NC
1838
1839 PERL_ARGS_ASSERT_IS_UTF8_PRINT;
1840
d4c19fe8 1841 return is_utf8_common(p, &PL_utf8_print, "IsPrint");
a0ed51b3
LW
1842}
1843
1844bool
7fc63493 1845Perl_is_utf8_punct(pTHX_ const U8 *p)
b8c5462f 1846{
97aff369 1847 dVAR;
7918f24d
NC
1848
1849 PERL_ARGS_ASSERT_IS_UTF8_PUNCT;
1850
d4c19fe8 1851 return is_utf8_common(p, &PL_utf8_punct, "IsPunct");
b8c5462f
JH
1852}
1853
1854bool
7fc63493 1855Perl_is_utf8_xdigit(pTHX_ const U8 *p)
b8c5462f 1856{
97aff369 1857 dVAR;
7918f24d
NC
1858
1859 PERL_ARGS_ASSERT_IS_UTF8_XDIGIT;
1860
d1eb3177 1861 return is_utf8_common(p, &PL_utf8_xdigit, "IsXDigit");
b8c5462f
JH
1862}
1863
1864bool
7fc63493 1865Perl_is_utf8_mark(pTHX_ const U8 *p)
a0ed51b3 1866{
97aff369 1867 dVAR;
7918f24d
NC
1868
1869 PERL_ARGS_ASSERT_IS_UTF8_MARK;
1870
d4c19fe8 1871 return is_utf8_common(p, &PL_utf8_mark, "IsM");
a0ed51b3
LW
1872}
1873
37e2e78e
KW
1874bool
1875Perl_is_utf8_X_begin(pTHX_ const U8 *p)
1876{
1877 dVAR;
1878
1879 PERL_ARGS_ASSERT_IS_UTF8_X_BEGIN;
1880
1881 return is_utf8_common(p, &PL_utf8_X_begin, "_X_Begin");
1882}
1883
1884bool
1885Perl_is_utf8_X_extend(pTHX_ const U8 *p)
1886{
1887 dVAR;
1888
1889 PERL_ARGS_ASSERT_IS_UTF8_X_EXTEND;
1890
1891 return is_utf8_common(p, &PL_utf8_X_extend, "_X_Extend");
1892}
1893
1894bool
1895Perl_is_utf8_X_prepend(pTHX_ const U8 *p)
1896{
1897 dVAR;
1898
1899 PERL_ARGS_ASSERT_IS_UTF8_X_PREPEND;
1900
1901 return is_utf8_common(p, &PL_utf8_X_prepend, "GCB=Prepend");
1902}
1903
1904bool
1905Perl_is_utf8_X_non_hangul(pTHX_ const U8 *p)
1906{
1907 dVAR;
1908
1909 PERL_ARGS_ASSERT_IS_UTF8_X_NON_HANGUL;
1910
1911 return is_utf8_common(p, &PL_utf8_X_non_hangul, "HST=Not_Applicable");
1912}
1913
1914bool
1915Perl_is_utf8_X_L(pTHX_ const U8 *p)
1916{
1917 dVAR;
1918
1919 PERL_ARGS_ASSERT_IS_UTF8_X_L;
1920
1921 return is_utf8_common(p, &PL_utf8_X_L, "GCB=L");
1922}
1923
1924bool
1925Perl_is_utf8_X_LV(pTHX_ const U8 *p)
1926{
1927 dVAR;
1928
1929 PERL_ARGS_ASSERT_IS_UTF8_X_LV;
1930
1931 return is_utf8_common(p, &PL_utf8_X_LV, "GCB=LV");
1932}
1933
1934bool
1935Perl_is_utf8_X_LVT(pTHX_ const U8 *p)
1936{
1937 dVAR;
1938
1939 PERL_ARGS_ASSERT_IS_UTF8_X_LVT;
1940
1941 return is_utf8_common(p, &PL_utf8_X_LVT, "GCB=LVT");
1942}
1943
1944bool
1945Perl_is_utf8_X_T(pTHX_ const U8 *p)
1946{
1947 dVAR;
1948
1949 PERL_ARGS_ASSERT_IS_UTF8_X_T;
1950
1951 return is_utf8_common(p, &PL_utf8_X_T, "GCB=T");
1952}
1953
1954bool
1955Perl_is_utf8_X_V(pTHX_ const U8 *p)
1956{
1957 dVAR;
1958
1959 PERL_ARGS_ASSERT_IS_UTF8_X_V;
1960
1961 return is_utf8_common(p, &PL_utf8_X_V, "GCB=V");
1962}
1963
1964bool
1965Perl_is_utf8_X_LV_LVT_V(pTHX_ const U8 *p)
1966{
1967 dVAR;
1968
1969 PERL_ARGS_ASSERT_IS_UTF8_X_LV_LVT_V;
1970
1971 return is_utf8_common(p, &PL_utf8_X_LV_LVT_V, "_X_LV_LVT_V");
1972}
1973
6b5c0936 1974/*
87cea99e 1975=for apidoc to_utf8_case
6b5c0936
JH
1976
1977The "p" contains the pointer to the UTF-8 string encoding
1978the character that is being converted.
1979
1980The "ustrp" is a pointer to the character buffer to put the
1981conversion result to. The "lenp" is a pointer to the length
1982of the result.
1983
0134edef 1984The "swashp" is a pointer to the swash to use.
6b5c0936 1985
36bb2ab6 1986Both the special and normal mappings are stored in lib/unicore/To/Foo.pl,
8fe4d5b2 1987and loaded by SWASHNEW, using lib/utf8_heavy.pl. The special (usually,
0134edef 1988but not always, a multicharacter mapping), is tried first.
6b5c0936 1989
0134edef
JH
1990The "special" is a string like "utf8::ToSpecLower", which means the
1991hash %utf8::ToSpecLower. The access to the hash is through
1992Perl_to_utf8_case().
6b5c0936 1993
0134edef
JH
1994The "normal" is a string like "ToLower" which means the swash
1995%utf8::ToLower.
1996
1997=cut */
6b5c0936 1998
2104c8d9 1999UV
9a957fbc
AL
2000Perl_to_utf8_case(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp,
2001 SV **swashp, const char *normal, const char *special)
a0ed51b3 2002{
97aff369 2003 dVAR;
89ebb4a3 2004 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
0134edef 2005 STRLEN len = 0;
aec46f14 2006 const UV uv0 = utf8_to_uvchr(p, NULL);
1feea2c7
JH
2007 /* The NATIVE_TO_UNI() and UNI_TO_NATIVE() mappings
2008 * are necessary in EBCDIC, they are redundant no-ops
2009 * in ASCII-ish platforms, and hopefully optimized away. */
f54cb97a 2010 const UV uv1 = NATIVE_TO_UNI(uv0);
7918f24d
NC
2011
2012 PERL_ARGS_ASSERT_TO_UTF8_CASE;
2013
9ae3ac1a
KW
2014 /* Note that swash_fetch() doesn't output warnings for these because it
2015 * assumes we will */
8457b38f 2016 if (uv1 >= UNICODE_SURROGATE_FIRST) {
9ae3ac1a 2017 if (uv1 <= UNICODE_SURROGATE_LAST) {
8457b38f
KW
2018 if (ckWARN_d(WARN_SURROGATE)) {
2019 const char* desc = (PL_op) ? OP_DESC(PL_op) : normal;
2020 Perl_warner(aTHX_ packWARN(WARN_SURROGATE),
2021 "Operation \"%s\" returns its argument for UTF-16 surrogate U+%04"UVXf"", desc, uv1);
2022 }
9ae3ac1a
KW
2023 }
2024 else if (UNICODE_IS_SUPER(uv1)) {
8457b38f
KW
2025 if (ckWARN_d(WARN_NON_UNICODE)) {
2026 const char* desc = (PL_op) ? OP_DESC(PL_op) : normal;
2027 Perl_warner(aTHX_ packWARN(WARN_NON_UNICODE),
2028 "Operation \"%s\" returns its argument for non-Unicode code point 0x%04"UVXf"", desc, uv1);
2029 }
9ae3ac1a
KW
2030 }
2031
2032 /* Note that non-characters are perfectly legal, so no warning should
2033 * be given */
2034 }
2035
1feea2c7 2036 uvuni_to_utf8(tmpbuf, uv1);
0134edef
JH
2037
2038 if (!*swashp) /* load on-demand */
2039 *swashp = swash_init("utf8", normal, &PL_sv_undef, 4, 0);
2040
a6f87d8c 2041 if (special) {
0134edef 2042 /* It might be "special" (sometimes, but not always,
2a37f04d 2043 * a multicharacter mapping) */
6673a63c 2044 HV * const hv = get_hv(special, 0);
b08cf34e
JH
2045 SV **svp;
2046
35da51f7 2047 if (hv &&
b08cf34e
JH
2048 (svp = hv_fetch(hv, (const char*)tmpbuf, UNISKIP(uv1), FALSE)) &&
2049 (*svp)) {
cfd0369c 2050 const char *s;
47654450 2051
cfd0369c 2052 s = SvPV_const(*svp, len);
47654450
JH
2053 if (len == 1)
2054 len = uvuni_to_utf8(ustrp, NATIVE_TO_UNI(*(U8*)s)) - ustrp;
2a37f04d 2055 else {
2f9475ad
JH
2056#ifdef EBCDIC
2057 /* If we have EBCDIC we need to remap the characters
2058 * since any characters in the low 256 are Unicode
2059 * code points, not EBCDIC. */
7cda7a3d 2060 U8 *t = (U8*)s, *tend = t + len, *d;
2f9475ad
JH
2061
2062 d = tmpbuf;
b08cf34e 2063 if (SvUTF8(*svp)) {
2f9475ad
JH
2064 STRLEN tlen = 0;
2065
2066 while (t < tend) {
d4c19fe8 2067 const UV c = utf8_to_uvchr(t, &tlen);
2f9475ad
JH
2068 if (tlen > 0) {
2069 d = uvchr_to_utf8(d, UNI_TO_NATIVE(c));
2070 t += tlen;
2071 }
2072 else
2073 break;
2074 }
2075 }
2076 else {
36fec512
JH
2077 while (t < tend) {
2078 d = uvchr_to_utf8(d, UNI_TO_NATIVE(*t));
2079 t++;
2080 }
2f9475ad
JH
2081 }
2082 len = d - tmpbuf;
2083 Copy(tmpbuf, ustrp, len, U8);
2084#else
d2dcd0fb 2085 Copy(s, ustrp, len, U8);
2f9475ad 2086#endif
29e98929 2087 }
983ffd37 2088 }
0134edef
JH
2089 }
2090
2091 if (!len && *swashp) {
d4c19fe8
AL
2092 const UV uv2 = swash_fetch(*swashp, tmpbuf, TRUE);
2093
0134edef
JH
2094 if (uv2) {
2095 /* It was "normal" (a single character mapping). */
d4c19fe8 2096 const UV uv3 = UNI_TO_NATIVE(uv2);
e9101d72 2097 len = uvchr_to_utf8(ustrp, uv3) - ustrp;
2a37f04d
JH
2098 }
2099 }
1feea2c7 2100
37e2e78e
KW
2101 if (!len) /* Neither: just copy. In other words, there was no mapping
2102 defined, which means that the code point maps to itself */
0134edef
JH
2103 len = uvchr_to_utf8(ustrp, uv0) - ustrp;
2104
2a37f04d
JH
2105 if (lenp)
2106 *lenp = len;
2107
0134edef 2108 return len ? utf8_to_uvchr(ustrp, 0) : 0;
a0ed51b3
LW
2109}
2110
d3e79532 2111/*
87cea99e 2112=for apidoc to_utf8_upper
d3e79532
JH
2113
2114Convert the UTF-8 encoded character at p to its uppercase version and
2115store that in UTF-8 in ustrp and its length in bytes in lenp. Note
89ebb4a3
JH
2116that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since
2117the uppercase version may be longer than the original character.
d3e79532
JH
2118
2119The first character of the uppercased version is returned
2120(but note, as explained above, that there may be more.)
2121
2122=cut */
2123
2104c8d9 2124UV
7fc63493 2125Perl_to_utf8_upper(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
a0ed51b3 2126{
97aff369 2127 dVAR;
7918f24d
NC
2128
2129 PERL_ARGS_ASSERT_TO_UTF8_UPPER;
2130
3a4c58c9
KW
2131 if (UTF8_IS_INVARIANT(*p)) {
2132 return _to_upper_title_latin1(*p, ustrp, lenp, 'S');
2133 }
2134 else if UTF8_IS_DOWNGRADEABLE_START(*p) {
2135 return _to_upper_title_latin1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)),
2136 ustrp, lenp, 'S');
2137 }
2138
50bda2c3 2139 return CALL_UPPER_CASE(p, ustrp, lenp);
983ffd37 2140}
a0ed51b3 2141
d3e79532 2142/*
87cea99e 2143=for apidoc to_utf8_title
d3e79532
JH
2144
2145Convert the UTF-8 encoded character at p to its titlecase version and
2146store that in UTF-8 in ustrp and its length in bytes in lenp. Note
89ebb4a3
JH
2147that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
2148titlecase version may be longer than the original character.
d3e79532
JH
2149
2150The first character of the titlecased version is returned
2151(but note, as explained above, that there may be more.)
2152
2153=cut */
2154
983ffd37 2155UV
7fc63493 2156Perl_to_utf8_title(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
983ffd37 2157{
97aff369 2158 dVAR;
7918f24d
NC
2159
2160 PERL_ARGS_ASSERT_TO_UTF8_TITLE;
2161
3a4c58c9
KW
2162 if (UTF8_IS_INVARIANT(*p)) {
2163 return _to_upper_title_latin1(*p, ustrp, lenp, 's');
2164 }
2165 else if UTF8_IS_DOWNGRADEABLE_START(*p) {
2166 return _to_upper_title_latin1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)),
2167 ustrp, lenp, 's');
2168 }
2169
50bda2c3 2170 return CALL_TITLE_CASE(p, ustrp, lenp);
a0ed51b3
LW
2171}
2172
d3e79532 2173/*
87cea99e 2174=for apidoc to_utf8_lower
d3e79532
JH
2175
2176Convert the UTF-8 encoded character at p to its lowercase version and
2177store that in UTF-8 in ustrp and its length in bytes in lenp. Note
89ebb4a3
JH
2178that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
2179lowercase version may be longer than the original character.
d3e79532
JH
2180
2181The first character of the lowercased version is returned
2182(but note, as explained above, that there may be more.)
2183
2184=cut */
2185
2104c8d9 2186UV
7fc63493 2187Perl_to_utf8_lower(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
a0ed51b3 2188{
97aff369 2189 dVAR;
7918f24d
NC
2190
2191 PERL_ARGS_ASSERT_TO_UTF8_LOWER;
2192
968c5e6a
KW
2193 if (UTF8_IS_INVARIANT(*p)) {
2194 return to_lower_latin1(*p, ustrp, lenp);
2195 }
2196 else if UTF8_IS_DOWNGRADEABLE_START(*p) {
2197 return to_lower_latin1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)), ustrp, lenp);
2198 }
2199
50bda2c3 2200 return CALL_LOWER_CASE(p, ustrp, lenp);
b4e400f9
JH
2201}
2202
d3e79532 2203/*
87cea99e 2204=for apidoc to_utf8_fold
d3e79532
JH
2205
2206Convert the UTF-8 encoded character at p to its foldcase version and
2207store that in UTF-8 in ustrp and its length in bytes in lenp. Note
89ebb4a3 2208that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
d3e79532
JH
2209foldcase version may be longer than the original character (up to
2210three characters).
2211
2212The first character of the foldcased version is returned
2213(but note, as explained above, that there may be more.)
2214
2215=cut */
2216
36bb2ab6
KW
2217/* Not currently externally documented is 'flags', which currently is non-zero
2218 * if full case folds are to be used; otherwise simple folds */
2219
b4e400f9 2220UV
36bb2ab6 2221Perl__to_utf8_fold_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, U8 flags)
b4e400f9 2222{
97aff369 2223 dVAR;
7918f24d 2224
36bb2ab6 2225 PERL_ARGS_ASSERT__TO_UTF8_FOLD_FLAGS;
7918f24d 2226
a1dde8de
KW
2227 if (UTF8_IS_INVARIANT(*p)) {
2228 return _to_fold_latin1(*p, ustrp, lenp, flags);
2229 }
2230 else if UTF8_IS_DOWNGRADEABLE_START(*p) {
2231 return _to_fold_latin1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)),
2232 ustrp, lenp, flags);
2233 }
2234
50bda2c3 2235 return CALL_FOLD_CASE(p, ustrp, lenp, flags);
a0ed51b3
LW
2236}
2237
711a919c 2238/* Note:
ed8b0411 2239 * Returns a "swash" which is a hash described in utf8.c:S_swash_fetch().
711a919c
TS
2240 * C<pkg> is a pointer to a package name for SWASHNEW, should be "utf8".
2241 * For other parameters, see utf8::SWASHNEW in lib/utf8_heavy.pl.
2242 */
a0ed51b3 2243SV*
7fc63493 2244Perl_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv, I32 minbits, I32 none)
a0ed51b3 2245{
27da23d5 2246 dVAR;
a0ed51b3 2247 SV* retval;
8e84507e 2248 dSP;
7fc63493
AL
2249 const size_t pkg_len = strlen(pkg);
2250 const size_t name_len = strlen(name);
da51bb9b 2251 HV * const stash = gv_stashpvn(pkg, pkg_len, 0);
f8be5cf0 2252 SV* errsv_save;
88c57d91 2253 GV *method;
ce3b816e 2254
7918f24d
NC
2255 PERL_ARGS_ASSERT_SWASH_INIT;
2256
96ca9f55
DM
2257 PUSHSTACKi(PERLSI_MAGIC);
2258 ENTER;
ec34a119 2259 SAVEHINTS();
96ca9f55 2260 save_re_context();
ba6ff154
FC
2261 if (PL_parser && PL_parser->error_count)
2262 SAVEI8(PL_parser->error_count), PL_parser->error_count = 0;
88c57d91
NC
2263 method = gv_fetchmeth(stash, "SWASHNEW", 8, -1);
2264 if (!method) { /* demand load utf8 */
ce3b816e 2265 ENTER;
f8be5cf0 2266 errsv_save = newSVsv(ERRSV);
dc0c6abb
NC
2267 /* It is assumed that callers of this routine are not passing in any
2268 user derived data. */
2269 /* Need to do this after save_re_context() as it will set PL_tainted to
2270 1 while saving $1 etc (see the code after getrx: in Perl_magic_get).
2271 Even line to create errsv_save can turn on PL_tainted. */
2272 SAVEBOOL(PL_tainted);
2273 PL_tainted = 0;
71bed85a 2274 Perl_load_module(aTHX_ PERL_LOADMOD_NOIMPORT, newSVpvn(pkg,pkg_len),
a0714e2c 2275 NULL);
f8be5cf0
JH
2276 if (!SvTRUE(ERRSV))
2277 sv_setsv(ERRSV, errsv_save);
2278 SvREFCNT_dec(errsv_save);
ce3b816e
GS
2279 LEAVE;
2280 }
2281 SPAGAIN;
a0ed51b3
LW
2282 PUSHMARK(SP);
2283 EXTEND(SP,5);
6e449a3a
MHM
2284 mPUSHp(pkg, pkg_len);
2285 mPUSHp(name, name_len);
a0ed51b3 2286 PUSHs(listsv);
6e449a3a
MHM
2287 mPUSHi(minbits);
2288 mPUSHi(none);
a0ed51b3 2289 PUTBACK;
f8be5cf0 2290 errsv_save = newSVsv(ERRSV);
88c57d91
NC
2291 /* If we already have a pointer to the method, no need to use call_method()
2292 to repeat the lookup. */
2293 if (method ? call_sv(MUTABLE_SV(method), G_SCALAR)
8b563eef 2294 : call_sv(newSVpvs_flags("SWASHNEW", SVs_TEMP), G_SCALAR | G_METHOD))
8e84507e 2295 retval = newSVsv(*PL_stack_sp--);
a0ed51b3 2296 else
e24b16f9 2297 retval = &PL_sv_undef;
f8be5cf0
JH
2298 if (!SvTRUE(ERRSV))
2299 sv_setsv(ERRSV, errsv_save);
2300 SvREFCNT_dec(errsv_save);
a0ed51b3
LW
2301 LEAVE;
2302 POPSTACK;
923e4eb5 2303 if (IN_PERL_COMPILETIME) {
623e6609 2304 CopHINTS_set(PL_curcop, PL_hints);
a0ed51b3 2305 }
bc45ce41
JH
2306 if (!SvROK(retval) || SvTYPE(SvRV(retval)) != SVt_PVHV) {
2307 if (SvPOK(retval))
35c1215d 2308 Perl_croak(aTHX_ "Can't find Unicode property definition \"%"SVf"\"",
be2597df 2309 SVfARG(retval));
cea2e8a9 2310 Perl_croak(aTHX_ "SWASHNEW didn't return an HV ref");
bc45ce41 2311 }
a0ed51b3
LW
2312 return retval;
2313}
2314
035d37be
JH
2315
2316/* This API is wrong for special case conversions since we may need to
2317 * return several Unicode characters for a single Unicode character
2318 * (see lib/unicore/SpecCase.txt) The SWASHGET in lib/utf8_heavy.pl is
2319 * the lower-level routine, and it is similarly broken for returning
38684baa
KW
2320 * multiple values. --jhi
2321 * For those, you should use to_utf8_case() instead */
979f2922 2322/* Now SWASHGET is recasted into S_swash_get in this file. */
680c470c
TS
2323
2324/* Note:
2325 * Returns the value of property/mapping C<swash> for the first character
2326 * of the string C<ptr>. If C<do_utf8> is true, the string C<ptr> is
2327 * assumed to be in utf8. If C<do_utf8> is false, the string C<ptr> is
2328 * assumed to be in native 8-bit encoding. Caches the swatch in C<swash>.
af2af982
KW
2329 *
2330 * A "swash" is a hash which contains initially the keys/values set up by
2331 * SWASHNEW. The purpose is to be able to completely represent a Unicode
2332 * property for all possible code points. Things are stored in a compact form
2333 * (see utf8_heavy.pl) so that calculation is required to find the actual
2334 * property value for a given code point. As code points are looked up, new
2335 * key/value pairs are added to the hash, so that the calculation doesn't have
2336 * to ever be re-done. Further, each calculation is done, not just for the
2337 * desired one, but for a whole block of code points adjacent to that one.
2338 * For binary properties on ASCII machines, the block is usually for 64 code
2339 * points, starting with a code point evenly divisible by 64. Thus if the
2340 * property value for code point 257 is requested, the code goes out and
2341 * calculates the property values for all 64 code points between 256 and 319,
2342 * and stores these as a single 64-bit long bit vector, called a "swatch",
2343 * under the key for code point 256. The key is the UTF-8 encoding for code
2344 * point 256, minus the final byte. Thus, if the length of the UTF-8 encoding
2345 * for a code point is 13 bytes, the key will be 12 bytes long. If the value
2346 * for code point 258 is then requested, this code realizes that it would be
2347 * stored under the key for 256, and would find that value and extract the
2348 * relevant bit, offset from 256.
2349 *
2350 * Non-binary properties are stored in as many bits as necessary to represent
2351 * their values (32 currently, though the code is more general than that), not
2352 * as single bits, but the principal is the same: the value for each key is a
2353 * vector that encompasses the property values for all code points whose UTF-8
2354 * representations are represented by the key. That is, for all code points
2355 * whose UTF-8 representations are length N bytes, and the key is the first N-1
2356 * bytes of that.
680c470c 2357 */
a0ed51b3 2358UV
680c470c 2359Perl_swash_fetch(pTHX_ SV *swash, const U8 *ptr, bool do_utf8)
a0ed51b3 2360{
27da23d5 2361 dVAR;
ef8f7699 2362 HV *const hv = MUTABLE_HV(SvRV(swash));
3568d838
JH
2363 U32 klen;
2364 U32 off;
a0ed51b3 2365 STRLEN slen;
7d85a32c 2366 STRLEN needents;
cfd0369c 2367 const U8 *tmps = NULL;
a0ed51b3 2368 U32 bit;
979f2922 2369 SV *swatch;
3568d838 2370 U8 tmputf8[2];
35da51f7 2371 const UV c = NATIVE_TO_ASCII(*ptr);
3568d838 2372
7918f24d
NC
2373 PERL_ARGS_ASSERT_SWASH_FETCH;
2374
3568d838 2375 if (!do_utf8 && !UNI_IS_INVARIANT(c)) {
979f2922
TS
2376 tmputf8[0] = (U8)UTF8_EIGHT_BIT_HI(c);
2377 tmputf8[1] = (U8)UTF8_EIGHT_BIT_LO(c);
2378 ptr = tmputf8;
3568d838
JH
2379 }
2380 /* Given a UTF-X encoded char 0xAA..0xYY,0xZZ
37e2e78e 2381 * then the "swatch" is a vec() for all the chars which start
3568d838
JH
2382 * with 0xAA..0xYY
2383 * So the key in the hash (klen) is length of encoded char -1
2384 */
2385 klen = UTF8SKIP(ptr) - 1;
2386 off = ptr[klen];
a0ed51b3 2387
979f2922 2388 if (klen == 0) {
37e2e78e 2389 /* If char is invariant then swatch is for all the invariant chars
1e54db1a 2390 * In both UTF-8 and UTF-8-MOD that happens to be UTF_CONTINUATION_MARK
7d85a32c 2391 */
979f2922
TS
2392 needents = UTF_CONTINUATION_MARK;
2393 off = NATIVE_TO_UTF(ptr[klen]);
2394 }
2395 else {
7d85a32c 2396 /* If char is encoded then swatch is for the prefix */
979f2922
TS
2397 needents = (1 << UTF_ACCUMULATION_SHIFT);
2398 off = NATIVE_TO_UTF(ptr[klen]) & UTF_CONTINUATION_MASK;
2399 }
7d85a32c 2400
a0ed51b3
LW
2401 /*
2402 * This single-entry cache saves about 1/3 of the utf8 overhead in test
2403 * suite. (That is, only 7-8% overall over just a hash cache. Still,
2404 * it's nothing to sniff at.) Pity we usually come through at least
2405 * two function calls to get here...
2406 *
2407 * NB: this code assumes that swatches are never modified, once generated!
2408 */
2409
3568d838 2410 if (hv == PL_last_swash_hv &&
a0ed51b3 2411 klen == PL_last_swash_klen &&
27da23d5 2412 (!klen || memEQ((char *)ptr, (char *)PL_last_swash_key, klen)) )
a0ed51b3
LW
2413 {
2414 tmps = PL_last_swash_tmps;
2415 slen = PL_last_swash_slen;
2416 }
2417 else {
2418 /* Try our second-level swatch cache, kept in a hash. */
e1ec3a88 2419 SV** svp = hv_fetch(hv, (const char*)ptr, klen, FALSE);
a0ed51b3 2420
979f2922
TS
2421 /* If not cached, generate it via swash_get */
2422 if (!svp || !SvPOK(*svp)
2423 || !(tmps = (const U8*)SvPV_const(*svp, slen))) {
2b9d42f0
NIS
2424 /* We use utf8n_to_uvuni() as we want an index into
2425 Unicode tables, not a native character number.
2426 */
aec46f14 2427 const UV code_point = utf8n_to_uvuni(ptr, UTF8_MAXBYTES, 0,
872c91ae
JH
2428 ckWARN(WARN_UTF8) ?
2429 0 : UTF8_ALLOW_ANY);
680c470c 2430 swatch = swash_get(swash,
979f2922 2431 /* On EBCDIC & ~(0xA0-1) isn't a useful thing to do */
361ee0fe 2432 (klen) ? (code_point & ~((UV)needents - 1)) : 0,
979f2922
TS
2433 needents);
2434
923e4eb5 2435 if (IN_PERL_COMPILETIME)
623e6609 2436 CopHINTS_set(PL_curcop, PL_hints);
a0ed51b3 2437
979f2922 2438 svp = hv_store(hv, (const char *)ptr, klen, swatch, 0);
a0ed51b3 2439
979f2922
TS
2440 if (!svp || !(tmps = (U8*)SvPV(*svp, slen))
2441 || (slen << 3) < needents)
660a4616 2442 Perl_croak(aTHX_ "panic: swash_fetch got improper swatch");
a0ed51b3
LW
2443 }
2444
2445 PL_last_swash_hv = hv;
16d8f38a 2446 assert(klen <= sizeof(PL_last_swash_key));
eac04b2e 2447 PL_last_swash_klen = (U8)klen;
cfd0369c
NC
2448 /* FIXME change interpvar.h? */
2449 PL_last_swash_tmps = (U8 *) tmps;
a0ed51b3
LW
2450 PL_last_swash_slen = slen;
2451 if (klen)
2452 Copy(ptr, PL_last_swash_key, klen, U8);
2453 }
2454
a410ec23
KW
2455 if (UTF8_IS_SUPER(ptr) && ckWARN_d(WARN_NON_UNICODE)) {
2456 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
2457
2458 /* This outputs warnings for binary properties only, assuming that
2459 * to_utf8_case() will output any for non-binary. Also, surrogates
2460 * aren't checked for, as that would warn on things like /\p{Gc=Cs}/ */
2461
2462 if (SvUV(*bitssvp) == 1) {
2463 /* User-defined properties can silently match above-Unicode */
2464 SV** const user_defined_svp = hv_fetchs(hv, "USER_DEFINED", FALSE);
2465 if (! user_defined_svp || ! SvUV(*user_defined_svp)) {
2466 const UV code_point = utf8n_to_uvuni(ptr, UTF8_MAXBYTES, 0, 0);
2467 Perl_warner(aTHX_ packWARN(WARN_NON_UNICODE),
2468 "Code point 0x%04"UVXf" is not Unicode, all \\p{} matches fail; all \\P{} matches succeed", code_point);
2469 }
2470 }
2471 }
2472
9faf8d75 2473 switch ((int)((slen << 3) / needents)) {
a0ed51b3
LW
2474 case 1:
2475 bit = 1 << (off & 7);
2476 off >>= 3;
2477 return (tmps[off] & bit) != 0;
2478 case 8:
2479 return tmps[off];
2480 case 16:
2481 off <<= 1;
2482 return (tmps[off] << 8) + tmps[off + 1] ;
2483 case 32:
2484 off <<= 2;
2485 return (tmps[off] << 24) + (tmps[off+1] << 16) + (tmps[off+2] << 8) + tmps[off + 3] ;
2486 }
660a4616 2487 Perl_croak(aTHX_ "panic: swash_fetch got swatch of unexpected bit width");
670f1322 2488 NORETURN_FUNCTION_END;
a0ed51b3 2489}
2b9d42f0 2490
319009ee
KW
2491/* Read a single line of the main body of the swash input text. These are of
2492 * the form:
2493 * 0053 0056 0073
2494 * where each number is hex. The first two numbers form the minimum and
2495 * maximum of a range, and the third is the value associated with the range.
2496 * Not all swashes should have a third number
2497 *
2498 * On input: l points to the beginning of the line to be examined; it points
2499 * to somewhere in the string of the whole input text, and is
2500 * terminated by a \n or the null string terminator.
2501 * lend points to the null terminator of that string
2502 * wants_value is non-zero if the swash expects a third number
2503 * typestr is the name of the swash's mapping, like 'ToLower'
2504 * On output: *min, *max, and *val are set to the values read from the line.
2505 * returns a pointer just beyond the line examined. If there was no
2506 * valid min number on the line, returns lend+1
2507 */
2508
2509STATIC U8*
2510S_swash_scan_list_line(pTHX_ U8* l, U8* const lend, UV* min, UV* max, UV* val,
2511 const bool wants_value, const U8* const typestr)
2512{
2513 const int typeto = typestr[0] == 'T' && typestr[1] == 'o';
2514 STRLEN numlen; /* Length of the number */
02470786
KW
2515 I32 flags = PERL_SCAN_SILENT_ILLDIGIT
2516 | PERL_SCAN_DISALLOW_PREFIX
2517 | PERL_SCAN_SILENT_NON_PORTABLE;
319009ee
KW
2518
2519 /* nl points to the next \n in the scan */
2520 U8* const nl = (U8*)memchr(l, '\n', lend - l);
2521
2522 /* Get the first number on the line: the range minimum */
2523 numlen = lend - l;
2524 *min = grok_hex((char *)l, &numlen, &flags, NULL);
2525 if (numlen) /* If found a hex number, position past it */
2526 l += numlen;
2527 else if (nl) { /* Else, go handle next line, if any */
2528 return nl + 1; /* 1 is length of "\n" */
2529 }
2530 else { /* Else, no next line */
2531 return lend + 1; /* to LIST's end at which \n is not found */
2532 }
2533
2534 /* The max range value follows, separated by a BLANK */
2535 if (isBLANK(*l)) {
2536 ++l;
02470786
KW
2537 flags = PERL_SCAN_SILENT_ILLDIGIT
2538 | PERL_SCAN_DISALLOW_PREFIX
2539 | PERL_SCAN_SILENT_NON_PORTABLE;
319009ee
KW
2540 numlen = lend - l;
2541 *max = grok_hex((char *)l, &numlen, &flags, NULL);
2542 if (numlen)
2543 l += numlen;
2544 else /* If no value here, it is a single element range */
2545 *max = *min;
2546
2547 /* Non-binary tables have a third entry: what the first element of the
2548 * range maps to */
2549 if (wants_value) {
2550 if (isBLANK(*l)) {
2551 ++l;
02470786
KW
2552 flags = PERL_SCAN_SILENT_ILLDIGIT
2553 | PERL_SCAN_DISALLOW_PREFIX
2554 | PERL_SCAN_SILENT_NON_PORTABLE;
319009ee
KW
2555 numlen = lend - l;
2556 *val = grok_hex((char *)l, &numlen, &flags, NULL);
2557 if (numlen)
2558 l += numlen;
2559 else
2560 *val = 0;
2561 }
2562 else {
2563 *val = 0;
2564 if (typeto) {
2565 Perl_croak(aTHX_ "%s: illegal mapping '%s'",
2566 typestr, l);
2567 }
2568 }
2569 }
2570 else
2571 *val = 0; /* bits == 1, then any val should be ignored */
2572 }
2573 else { /* Nothing following range min, should be single element with no
2574 mapping expected */
2575 *max = *min;
2576 if (wants_value) {
2577 *val = 0;
2578 if (typeto) {
2579 Perl_croak(aTHX_ "%s: illegal mapping '%s'", typestr, l);
2580 }
2581 }
2582 else
2583 *val = 0; /* bits == 1, then val should be ignored */
2584 }
2585
2586 /* Position to next line if any, or EOF */
2587 if (nl)
2588 l = nl + 1;
2589 else
2590 l = lend;
2591
2592 return l;
2593}
2594
979f2922
TS
2595/* Note:
2596 * Returns a swatch (a bit vector string) for a code point sequence
2597 * that starts from the value C<start> and comprises the number C<span>.
2598 * A C<swash> must be an object created by SWASHNEW (see lib/utf8_heavy.pl).
2599 * Should be used via swash_fetch, which will cache the swatch in C<swash>.
2600 */
2601STATIC SV*
2602S_swash_get(pTHX_ SV* swash, UV start, UV span)
2603{
2604 SV *swatch;
77f9f126 2605 U8 *l, *lend, *x, *xend, *s, *send;
979f2922 2606 STRLEN lcur, xcur, scur;
ef8f7699 2607 HV *const hv = MUTABLE_HV(SvRV(swash));
972dd592
KW
2608
2609 /* The string containing the main body of the table */
017a3ce5 2610 SV** const listsvp = hv_fetchs(hv, "LIST", FALSE);
972dd592 2611
017a3ce5
GA
2612 SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE);
2613 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
2614 SV** const nonesvp = hv_fetchs(hv, "NONE", FALSE);
2615 SV** const extssvp = hv_fetchs(hv, "EXTRAS", FALSE);
77f9f126 2616 SV** const invert_it_svp = hv_fetchs(hv, "INVERT_IT", FALSE);
0bd48802 2617 const U8* const typestr = (U8*)SvPV_nolen(*typesvp);
0bd48802
AL
2618 const STRLEN bits = SvUV(*bitssvp);
2619 const STRLEN octets = bits >> 3; /* if bits == 1, then octets == 0 */
2620 const UV none = SvUV(*nonesvp);
84ea5ef6 2621 UV end = start + span;
979f2922 2622
7918f24d
NC
2623 PERL_ARGS_ASSERT_SWASH_GET;
2624
979f2922 2625 if (bits != 1 && bits != 8 && bits != 16 && bits != 32) {
660a4616
TS
2626 Perl_croak(aTHX_ "panic: swash_get doesn't expect bits %"UVuf,
2627 (UV)bits);
979f2922
TS
2628 }
2629
84ea5ef6
KW
2630 /* If overflowed, use the max possible */
2631 if (end < start) {
2632 end = UV_MAX;
2633 span = end - start;
2634 }
2635
979f2922 2636 /* create and initialize $swatch */
979f2922 2637 scur = octets ? (span * octets) : (span + 7) / 8;
e524fe40
NC
2638 swatch = newSV(scur);
2639 SvPOK_on(swatch);
979f2922
TS
2640 s = (U8*)SvPVX(swatch);
2641 if (octets && none) {
0bd48802 2642 const U8* const e = s + scur;
979f2922
TS
2643 while (s < e) {
2644 if (bits == 8)
2645 *s++ = (U8)(none & 0xff);
2646 else if (bits == 16) {
2647 *s++ = (U8)((none >> 8) & 0xff);
2648 *s++ = (U8)( none & 0xff);
2649 }
2650 else if (bits == 32) {
2651 *s++ = (U8)((none >> 24) & 0xff);
2652 *s++ = (U8)((none >> 16) & 0xff);
2653 *s++ = (U8)((none >> 8) & 0xff);
2654 *s++ = (U8)( none & 0xff);
2655 }
2656 }
2657 *s = '\0';
2658 }
2659 else {
2660 (void)memzero((U8*)s, scur + 1);
2661 }
2662 SvCUR_set(swatch, scur);
2663 s = (U8*)SvPVX(swatch);
2664
af2af982
KW
2665 /* read $swash->{LIST}. XXX Note that this is a linear scan through a
2666 * sorted list. A binary search would be much more efficient */
979f2922
TS
2667 l = (U8*)SvPV(*listsvp, lcur);
2668 lend = l + lcur;
2669 while (l < lend) {
35da51f7 2670 UV min, max, val;
319009ee
KW
2671 l = S_swash_scan_list_line(aTHX_ l, lend, &min, &max, &val,
2672 cBOOL(octets), typestr);
2673 if (l > lend) {
979f2922
TS
2674 break;
2675 }
2676
972dd592 2677 /* If looking for something beyond this range, go try the next one */
979f2922
TS
2678 if (max < start)
2679 continue;
2680
2681 if (octets) {
35da51f7 2682 UV key;
979f2922
TS
2683 if (min < start) {
2684 if (!none || val < none) {
2685 val += start - min;
2686 }
2687 min = start;
2688 }
2689 for (key = min; key <= max; key++) {
2690 STRLEN offset;
2691 if (key >= end)
2692 goto go_out_list;
6cb05c12
KW
2693 /* XXX If it should ever happen (very unlikely) that we would
2694 * want a non-binary result for the code point at UV_MAX,
2695 * special handling would need to be inserted here, as is done
2696 * below for the binary case */
979f2922
TS
2697 /* offset must be non-negative (start <= min <= key < end) */
2698 offset = octets * (key - start);
2699 if (bits == 8)
2700 s[offset] = (U8)(val & 0xff);
2701 else if (bits == 16) {
2702 s[offset ] = (U8)((val >> 8) & 0xff);
2703 s[offset + 1] = (U8)( val & 0xff);
2704 }
2705 else if (bits == 32) {
2706 s[offset ] = (U8)((val >> 24) & 0xff);
2707 s[offset + 1] = (U8)((val >> 16) & 0xff);
2708 s[offset + 2] = (U8)((val >> 8) & 0xff);
2709 s[offset + 3] = (U8)( val & 0xff);
2710 }
2711
2712 if (!none || val < none)
2713 ++val;
2714 }
2715 }
711a919c 2716 else { /* bits == 1, then val should be ignored */
35da51f7 2717 UV key;
979f2922
TS
2718 if (min < start)
2719 min = start;
6cb05c12
KW
2720
2721 /* Special case when the upper-end is the highest possible code
2722 * point representable on the platform. Otherwise, the code below
2723 * exits before setting this bit. Done here to avoid testing for
2724 * this extremely unlikely possibility in the loop */
2725 if (UNLIKELY(end == UV_MAX && max == UV_MAX)) {
2726 const STRLEN offset = (STRLEN)(max - start);
2727 s[offset >> 3] |= 1 << (offset & 7);
2728 }
979f2922 2729 for (key = min; key <= max; key++) {
0bd48802 2730 const STRLEN offset = (STRLEN)(key - start);
979f2922
TS
2731 if (key >= end)
2732 goto go_out_list;
2733 s[offset >> 3] |= 1 << (offset & 7);
2734 }
2735 }
2736 } /* while */
2737 go_out_list:
2738
9479a769 2739 /* Invert if the data says it should be. Assumes that bits == 1 */
77f9f126 2740 if (invert_it_svp && SvUV(*invert_it_svp)) {
0bda3001
KW
2741
2742 /* Unicode properties should come with all bits above PERL_UNICODE_MAX
2743 * be 0, and their inversion should also be 0, as we don't succeed any
2744 * Unicode property matches for non-Unicode code points */
2745 if (start <= PERL_UNICODE_MAX) {
2746
2747 /* The code below assumes that we never cross the
2748 * Unicode/above-Unicode boundary in a range, as otherwise we would
2749 * have to figure out where to stop flipping the bits. Since this
2750 * boundary is divisible by a large power of 2, and swatches comes
2751 * in small powers of 2, this should be a valid assumption */
2752 assert(start + span - 1 <= PERL_UNICODE_MAX);
2753
507a8485
KW
2754 send = s + scur;
2755 while (s < send) {
2756 *s = ~(*s);
2757 s++;
2758 }
0bda3001 2759 }
77f9f126
KW
2760 }
2761
d73c39c5
KW
2762 /* read $swash->{EXTRAS}
2763 * This code also copied to swash_to_invlist() below */
979f2922
TS
2764 x = (U8*)SvPV(*extssvp, xcur);
2765 xend = x + xcur;
2766 while (x < xend) {
2767 STRLEN namelen;
2768 U8 *namestr;
2769 SV** othersvp;
2770 HV* otherhv;
2771 STRLEN otherbits;
2772 SV **otherbitssvp, *other;
711a919c 2773 U8 *s, *o, *nl;
979f2922
TS
2774 STRLEN slen, olen;
2775
35da51f7 2776 const U8 opc = *x++;
979f2922
TS
2777 if (opc == '\n')
2778 continue;
2779
2780 nl = (U8*)memchr(x, '\n', xend - x);
2781
2782 if (opc != '-' && opc != '+' && opc != '!' && opc != '&') {
2783 if (nl) {
2784 x = nl + 1; /* 1 is length of "\n" */
2785 continue;
2786 }
2787 else {
2788 x = xend; /* to EXTRAS' end at which \n is not found */
2789 break;
2790 }
2791 }
2792
2793 namestr = x;
2794 if (nl) {
2795 namelen = nl - namestr;
2796 x = nl + 1;
2797 }
2798 else {
2799 namelen = xend - namestr;
2800 x = xend;
2801 }
2802
2803 othersvp = hv_fetch(hv, (char *)namestr, namelen, FALSE);
ef8f7699 2804 otherhv = MUTABLE_HV(SvRV(*othersvp));
017a3ce5 2805 otherbitssvp = hv_fetchs(otherhv, "BITS", FALSE);
979f2922
TS
2806 otherbits = (STRLEN)SvUV(*otherbitssvp);
2807 if (bits < otherbits)
660a4616 2808 Perl_croak(aTHX_ "panic: swash_get found swatch size mismatch");
979f2922
TS
2809
2810 /* The "other" swatch must be destroyed after. */
2811 other = swash_get(*othersvp, start, span);
2812 o = (U8*)SvPV(other, olen);
2813
2814 if (!olen)
660a4616 2815 Perl_croak(aTHX_ "panic: swash_get got improper swatch");
979f2922
TS
2816
2817 s = (U8*)SvPV(swatch, slen);
2818 if (bits == 1 && otherbits == 1) {
2819 if (slen != olen)
660a4616 2820 Perl_croak(aTHX_ "panic: swash_get found swatch length mismatch");
979f2922
TS
2821
2822 switch (opc) {
2823 case '+':
2824 while (slen--)
2825 *s++ |= *o++;
2826 break;
2827 case '!':
2828 while (slen--)
2829 *s++ |= ~*o++;
2830 break;
2831 case '-':
2832 while (slen--)
2833 *s++ &= ~*o++;
2834 break;
2835 case '&':
2836 while (slen--)
2837 *s++ &= *o++;
2838 break;
2839 default:
2840 break;
2841 }
2842 }
711a919c 2843 else {
979f2922
TS
2844 STRLEN otheroctets = otherbits >> 3;
2845 STRLEN offset = 0;
35da51f7 2846 U8* const send = s + slen;
979f2922
TS
2847
2848 while (s < send) {
2849 UV otherval = 0;
2850
2851 if (otherbits == 1) {
2852 otherval = (o[offset >> 3] >> (offset & 7)) & 1;
2853 ++offset;
2854 }
2855 else {
2856 STRLEN vlen = otheroctets;
2857 otherval = *o++;
2858 while (--vlen) {
2859 otherval <<= 8;
2860 otherval |= *o++;
2861 }
2862 }
2863
711a919c 2864 if (opc == '+' && otherval)
6f207bd3 2865 NOOP; /* replace with otherval */
979f2922
TS
2866 else if (opc == '!' && !otherval)
2867 otherval = 1;
2868 else if (opc == '-' && otherval)
2869 otherval = 0;
2870 else if (opc == '&' && !otherval)
2871 otherval = 0;
2872 else {
711a919c 2873 s += octets; /* no replacement */
979f2922
TS
2874 continue;
2875 }
2876
2877 if (bits == 8)
2878 *s++ = (U8)( otherval & 0xff);
2879 else if (bits == 16) {
2880 *s++ = (U8)((otherval >> 8) & 0xff);
2881 *s++ = (U8)( otherval & 0xff);
2882 }
2883 else if (bits == 32) {
2884 *s++ = (U8)((otherval >> 24) & 0xff);
2885 *s++ = (U8)((otherval >> 16) & 0xff);
2886 *s++ = (U8)((otherval >> 8) & 0xff);
2887 *s++ = (U8)( otherval & 0xff);
2888 }
2889 }
2890 }
2891 sv_free(other); /* through with it! */
2892 } /* while */
2893 return swatch;
2894}
2895
064c021d 2896HV*
4c2e1131 2897Perl__swash_inversion_hash(pTHX_ SV* const swash)
064c021d
KW
2898{
2899
5662e334
KW
2900 /* Subject to change or removal. For use only in one place in regcomp.c.
2901 * Can't be used on a property that is subject to user override, as it
2902 * relies on the value of SPECIALS in the swash which would be set by
2903 * utf8_heavy.pl to the hash in the non-overriden file, and hence is not set
2904 * for overridden properties
064c021d
KW
2905 *
2906 * Returns a hash which is the inversion and closure of a swash mapping.
2907 * For example, consider the input lines:
2908 * 004B 006B
2909 * 004C 006C
2910 * 212A 006B
2911 *
2912 * The returned hash would have two keys, the utf8 for 006B and the utf8 for
2913 * 006C. The value for each key is an array. For 006C, the array would
2914 * have a two elements, the utf8 for itself, and for 004C. For 006B, there
2915 * would be three elements in its array, the utf8 for 006B, 004B and 212A.
2916 *
2917 * Essentially, for any code point, it gives all the code points that map to
2918 * it, or the list of 'froms' for that point.
2919 *
5662e334
KW
2920 * Currently it ignores any additions or deletions from other swashes,
2921 * looking at just the main body of the swash, and if there are SPECIALS
2922 * in the swash, at that hash
2923 *
2924 * The specials hash can be extra code points, and most likely consists of
2925 * maps from single code points to multiple ones (each expressed as a string
2926 * of utf8 characters). This function currently returns only 1-1 mappings.
2927 * However consider this possible input in the specials hash:
2928 * "\xEF\xAC\x85" => "\x{0073}\x{0074}", # U+FB05 => 0073 0074
2929 * "\xEF\xAC\x86" => "\x{0073}\x{0074}", # U+FB06 => 0073 0074
2930 *
2931 * Both FB05 and FB06 map to the same multi-char sequence, which we don't
2932 * currently handle. But it also means that FB05 and FB06 are equivalent in
2933 * a 1-1 mapping which we should handle, and this relationship may not be in
2934 * the main table. Therefore this function examines all the multi-char
2935 * sequences and adds the 1-1 mappings that come out of that. */
064c021d
KW
2936
2937 U8 *l, *lend;
2938 STRLEN lcur;
2939 HV *const hv = MUTABLE_HV(SvRV(swash));
2940
2941 /* The string containing the main body of the table */
2942 SV** const listsvp = hv_fetchs(hv, "LIST", FALSE);
2943
2944 SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE);
2945 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
2946 SV** const nonesvp = hv_fetchs(hv, "NONE", FALSE);
2947 /*SV** const extssvp = hv_fetchs(hv, "EXTRAS", FALSE);*/
2948 const U8* const typestr = (U8*)SvPV_nolen(*typesvp);
2949 const STRLEN bits = SvUV(*bitssvp);
2950 const STRLEN octets = bits >> 3; /* if bits == 1, then octets == 0 */
2951 const UV none = SvUV(*nonesvp);
5662e334 2952 SV **specials_p = hv_fetchs(hv, "SPECIALS", 0);
064c021d
KW
2953
2954 HV* ret = newHV();
2955
2956 PERL_ARGS_ASSERT__SWASH_INVERSION_HASH;
2957
2958 /* Must have at least 8 bits to get the mappings */
2959 if (bits != 8 && bits != 16 && bits != 32) {
2960 Perl_croak(aTHX_ "panic: swash_inversion_hash doesn't expect bits %"UVuf,
2961 (UV)bits);
2962 }
2963
5662e334
KW
2964 if (specials_p) { /* It might be "special" (sometimes, but not always, a
2965 mapping to more than one character */
2966
2967 /* Construct an inverse mapping hash for the specials */
2968 HV * const specials_hv = MUTABLE_HV(SvRV(*specials_p));
2969 HV * specials_inverse = newHV();
2970 char *char_from; /* the lhs of the map */
2971 I32 from_len; /* its byte length */
2972 char *char_to; /* the rhs of the map */
2973 I32 to_len; /* its byte length */
2974 SV *sv_to; /* and in a sv */
2975 AV* from_list; /* list of things that map to each 'to' */
2976
2977 hv_iterinit(specials_hv);
2978
2979 /* The keys are the characters (in utf8) that map to the corresponding
2980 * utf8 string value. Iterate through the list creating the inverse
2981 * list. */
2982 while ((sv_to = hv_iternextsv(specials_hv, &char_from, &from_len))) {
2983 SV** listp;
2984 if (! SvPOK(sv_to)) {
2985 Perl_croak(aTHX_ "panic: value returned from hv_iternextsv() unexpectedly is not a string");
2986 }
2987 /*DEBUG_U(PerlIO_printf(Perl_debug_log, "Found mapping from %"UVXf", First char of to is %"UVXf"\n", utf8_to_uvchr((U8*) char_from, 0), utf8_to_uvchr((U8*) SvPVX(sv_to), 0)));*/
2988
2989 /* Each key in the inverse list is a mapped-to value, and the key's
2990 * hash value is a list of the strings (each in utf8) that map to
2991 * it. Those strings are all one character long */
2992 if ((listp = hv_fetch(specials_inverse,
2993 SvPVX(sv_to),
2994 SvCUR(sv_to), 0)))
2995 {
2996 from_list = (AV*) *listp;
2997 }
2998 else { /* No entry yet for it: create one */
2999 from_list = newAV();
3000 if (! hv_store(specials_inverse,
3001 SvPVX(sv_to),
3002 SvCUR(sv_to),
3003 (SV*) from_list, 0))
3004 {
3005 Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed");
3006 }
3007 }
3008
3009 /* Here have the list associated with this 'to' (perhaps newly
3010 * created and empty). Just add to it. Note that we ASSUME that
3011 * the input is guaranteed to not have duplications, so we don't
3012 * check for that. Duplications just slow down execution time. */
3013 av_push(from_list, newSVpvn_utf8(char_from, from_len, TRUE));
3014 }
3015
3016 /* Here, 'specials_inverse' contains the inverse mapping. Go through
3017 * it looking for cases like the FB05/FB06 examples above. There would
3018 * be an entry in the hash like
3019 * 'st' => [ FB05, FB06 ]
3020 * In this example we will create two lists that get stored in the
3021 * returned hash, 'ret':
3022 * FB05 => [ FB05, FB06 ]
3023 * FB06 => [ FB05, FB06 ]
3024 *
3025 * Note that there is nothing to do if the array only has one element.
3026 * (In the normal 1-1 case handled below, we don't have to worry about
3027 * two lists, as everything gets tied to the single list that is
3028 * generated for the single character 'to'. But here, we are omitting
3029 * that list, ('st' in the example), so must have multiple lists.) */
3030 while ((from_list = (AV *) hv_iternextsv(specials_inverse,
3031 &char_to, &to_len)))
3032 {
3033 if (av_len(from_list) > 0) {
3034 int i;
3035
3036 /* We iterate over all combinations of i,j to place each code
3037 * point on each list */
3038 for (i = 0; i <= av_len(from_list); i++) {
3039 int j;
3040 AV* i_list = newAV();
3041 SV** entryp = av_fetch(from_list, i, FALSE);
3042 if (entryp == NULL) {
3043 Perl_croak(aTHX_ "panic: av_fetch() unexpectedly failed");
3044 }
3045 if (hv_fetch(ret, SvPVX(*entryp), SvCUR(*entryp), FALSE)) {
3046 Perl_croak(aTHX_ "panic: unexpected entry for %s", SvPVX(*entryp));
3047 }
3048 if (! hv_store(ret, SvPVX(*entryp), SvCUR(*entryp),
3049 (SV*) i_list, FALSE))
3050 {
3051 Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed");
3052 }
3053
3054 /* For debugging: UV u = utf8_to_uvchr((U8*) SvPVX(*entryp), 0);*/
3055 for (j = 0; j <= av_len(from_list); j++) {
3056 entryp = av_fetch(from_list, j, FALSE);
3057 if (entryp == NULL) {
3058 Perl_croak(aTHX_ "panic: av_fetch() unexpectedly failed");
3059 }
3060
3061 /* When i==j this adds itself to the list */
3062 av_push(i_list, newSVuv(utf8_to_uvchr(
3063 (U8*) SvPVX(*entryp), 0)));
3064 /*DEBUG_U(PerlIO_printf(Perl_debug_log, "Adding %"UVXf" to list for %"UVXf"\n", utf8_to_uvchr((U8*) SvPVX(*entryp), 0), u));*/
3065 }
3066 }
3067 }
3068 }
3069 SvREFCNT_dec(specials_inverse); /* done with it */
3070 } /* End of specials */
3071
064c021d
KW
3072 /* read $swash->{LIST} */
3073 l = (U8*)SvPV(*listsvp, lcur);
3074 lend = l + lcur;
3075
3076 /* Go through each input line */
3077 while (l < lend) {
3078 UV min, max, val;
3079 UV inverse;
3080 l = S_swash_scan_list_line(aTHX_ l, lend, &min, &max, &val,
3081 cBOOL(octets), typestr);
3082 if (l > lend) {
3083 break;
3084 }
3085
3086 /* Each element in the range is to be inverted */
3087 for (inverse = min; inverse <= max; inverse++) {
3088 AV* list;
064c021d
KW
3089 SV** listp;
3090 IV i;
3091 bool found_key = FALSE;
5662e334 3092 bool found_inverse = FALSE;
064c021d
KW
3093
3094 /* The key is the inverse mapping */
3095 char key[UTF8_MAXBYTES+1];
3096 char* key_end = (char *) uvuni_to_utf8((U8*) key, val);
3097 STRLEN key_len = key_end - key;
3098
064c021d
KW
3099 /* Get the list for the map */
3100 if ((listp = hv_fetch(ret, key, key_len, FALSE))) {
3101 list = (AV*) *listp;
3102 }
3103 else { /* No entry yet for it: create one */
3104 list = newAV();
3105 if (! hv_store(ret, key, key_len, (SV*) list, FALSE)) {
3106 Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed");
3107 }
3108 }
3109
5662e334
KW
3110 /* Look through list to see if this inverse mapping already is
3111 * listed, or if there is a mapping to itself already */
508f7cfa 3112 for (i = 0; i <= av_len(list); i++) {
064c021d
KW
3113 SV** entryp = av_fetch(list, i, FALSE);
3114 SV* entry;
3115 if (entryp == NULL) {
3116 Perl_croak(aTHX_ "panic: av_fetch() unexpectedly failed");
3117 }
3118 entry = *entryp;
5662e334 3119 /*DEBUG_U(PerlIO_printf(Perl_debug_log, "list for %"UVXf" contains %"UVXf"\n", val, SvUV(entry)));*/
56ca34ca 3120 if (SvUV(entry) == val) {
064c021d 3121 found_key = TRUE;
5662e334
KW
3122 }
3123 if (SvUV(entry) == inverse) {
3124 found_inverse = TRUE;
3125 }
3126
3127 /* No need to continue searching if found everything we are
3128 * looking for */
3129 if (found_key && found_inverse) {
064c021d
KW
3130 break;
3131 }
3132 }
56ca34ca
KW
3133
3134 /* Make sure there is a mapping to itself on the list */
064c021d 3135 if (! found_key) {
d397ff6a 3136 av_push(list, newSVuv(val));
5662e334 3137 /*DEBUG_U(PerlIO_printf(Perl_debug_log, "Adding %"UVXf" to list for %"UVXf"\n", val, val));*/
064c021d
KW
3138 }
3139
3140
3141 /* Simply add the value to the list */
5662e334
KW
3142 if (! found_inverse) {
3143 av_push(list, newSVuv(inverse));
3144 /*DEBUG_U(PerlIO_printf(Perl_debug_log, "Adding %"UVXf" to list for %"UVXf"\n", inverse, val));*/
3145 }
064c021d
KW
3146
3147 /* swash_get() increments the value of val for each element in the
3148 * range. That makes more compact tables possible. You can
3149 * express the capitalization, for example, of all consecutive
3150 * letters with a single line: 0061\t007A\t0041 This maps 0061 to
3151 * 0041, 0062 to 0042, etc. I (khw) have never understood 'none',
bd3f2f94
KW
3152 * and it's not documented; it appears to be used only in
3153 * implementing tr//; I copied the semantics from swash_get(), just
3154 * in case */
064c021d
KW
3155 if (!none || val < none) {
3156 ++val;
3157 }
3158 }
3159 }
3160
3161 return ret;
3162}
3163
a25abddc 3164SV*
d764b54e
KW
3165Perl__swash_to_invlist(pTHX_ SV* const swash)
3166{
3167
3168 /* Subject to change or removal. For use only in one place in regcomp.c */
3169
3170 U8 *l, *lend;
3171 char *loc;
3172 STRLEN lcur;
3173 HV *const hv = MUTABLE_HV(SvRV(swash));
3174 UV elements = 0; /* Number of elements in the inversion list */
b443038a 3175 U8 empty[] = "";
d764b54e
KW
3176
3177 /* The string containing the main body of the table */
3178 SV** const listsvp = hv_fetchs(hv, "LIST", FALSE);
3179 SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE);
3180 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
d73c39c5 3181 SV** const extssvp = hv_fetchs(hv, "EXTRAS", FALSE);
77f9f126 3182 SV** const invert_it_svp = hv_fetchs(hv, "INVERT_IT", FALSE);
d764b54e
KW
3183
3184 const U8* const typestr = (U8*)SvPV_nolen(*typesvp);
3185 const STRLEN bits = SvUV(*bitssvp);
3186 const STRLEN octets = bits >> 3; /* if bits == 1, then octets == 0 */
d73c39c5
KW
3187 U8 *x, *xend;
3188 STRLEN xcur;
d764b54e 3189
a25abddc 3190 SV* invlist;
d764b54e
KW
3191
3192 PERL_ARGS_ASSERT__SWASH_TO_INVLIST;
3193
3194 /* read $swash->{LIST} */
b443038a
KW
3195 if (SvPOK(*listsvp)) {
3196 l = (U8*)SvPV(*listsvp, lcur);
3197 }
3198 else {
3199 /* LIST legitimately doesn't contain a string during compilation phases
3200 * of Perl itself, before the Unicode tables are generated. In this
3201 * case, just fake things up by creating an empty list */
3202 l = empty;
3203 lcur = 0;
3204 }
d764b54e
KW
3205 loc = (char *) l;
3206 lend = l + lcur;
3207
3208 /* Scan the input to count the number of lines to preallocate array size
3209 * based on worst possible case, which is each line in the input creates 2
3210 * elements in the inversion list: 1) the beginning of a range in the list;
3211 * 2) the beginning of a range not in the list. */
3212 while ((loc = (strchr(loc, '\n'))) != NULL) {
3213 elements += 2;
3214 loc++;
3215 }
3216
3217 /* If the ending is somehow corrupt and isn't a new line, add another
3218 * element for the final range that isn't in the inversion list */
3219 if (! (*lend == '\n' || (*lend == '\0' && *(lend - 1) == '\n'))) {
3220 elements++;
3221 }
3222
3223 invlist = _new_invlist(elements);
3224
3225 /* Now go through the input again, adding each range to the list */
3226 while (l < lend) {
3227 UV start, end;
3228 UV val; /* Not used by this function */
3229
3230 l = S_swash_scan_list_line(aTHX_ l, lend, &start, &end, &val,
3231 cBOOL(octets), typestr);
3232
3233 if (l > lend) {
3234 break;
3235 }
3236
3237 _append_range_to_invlist(invlist, start, end);
3238 }
3239
77f9f126
KW
3240 /* Invert if the data says it should be */
3241 if (invert_it_svp && SvUV(*invert_it_svp)) {
45bb2768 3242 _invlist_invert_prop(invlist);
77f9f126
KW
3243 }
3244
d73c39c5
KW
3245 /* This code is copied from swash_get()
3246 * read $swash->{EXTRAS} */
3247 x = (U8*)SvPV(*extssvp, xcur);
3248 xend = x + xcur;
3249 while (x < xend) {
3250 STRLEN namelen;
3251 U8 *namestr;
3252 SV** othersvp;
3253 HV* otherhv;
3254 STRLEN otherbits;
3255 SV **otherbitssvp, *other;
3256 U8 *nl;
3257
3258 const U8 opc = *x++;
3259 if (opc == '\n')
3260 continue;
3261
3262 nl = (U8*)memchr(x, '\n', xend - x);
3263
3264 if (opc != '-' && opc != '+' && opc != '!' && opc != '&') {
3265 if (nl) {
3266 x = nl + 1; /* 1 is length of "\n" */
3267 continue;
3268 }
3269 else {
3270 x = xend; /* to EXTRAS' end at which \n is not found */
3271 break;
3272 }
3273 }
3274
3275 namestr = x;
3276 if (nl) {
3277 namelen = nl - namestr;
3278 x = nl + 1;
3279 }
3280 else {
3281 namelen = xend - namestr;
3282 x = xend;
3283 }
3284
3285 othersvp = hv_fetch(hv, (char *)namestr, namelen, FALSE);
3286 otherhv = MUTABLE_HV(SvRV(*othersvp));
3287 otherbitssvp = hv_fetchs(otherhv, "BITS", FALSE);
3288 otherbits = (STRLEN)SvUV(*otherbitssvp);
3289
3290 if (bits != otherbits || bits != 1) {
3291 Perl_croak(aTHX_ "panic: _swash_to_invlist only operates on boolean properties");
3292 }
3293
3294 /* The "other" swatch must be destroyed after. */
3295 other = _swash_to_invlist((SV *)*othersvp);
3296
3297 /* End of code copied from swash_get() */
3298 switch (opc) {
3299 case '+':
3300 _invlist_union(invlist, other, &invlist);
3301 break;
3302 case '!':
3303 _invlist_invert(other);
3304 _invlist_union(invlist, other, &invlist);
3305 break;
3306 case '-':
3307 _invlist_subtract(invlist, other, &invlist);
3308 break;
3309 case '&':
3310 _invlist_intersection(invlist, other, &invlist);
3311 break;
3312 default:
3313 break;
3314 }
3315 sv_free(other); /* through with it! */
3316 }
3317
d764b54e
KW
3318 return invlist;
3319}
3320
0f830e0b 3321/*
87cea99e 3322=for apidoc uvchr_to_utf8
0f830e0b 3323
6ee84de2 3324Adds the UTF-8 representation of the Native code point C<uv> to the end
0f830e0b
NC
3325of the string C<d>; C<d> should be have at least C<UTF8_MAXBYTES+1> free
3326bytes available. The return value is the pointer to the byte after the
3327end of the new character. In other words,
3328
3329 d = uvchr_to_utf8(d, uv);
3330
3331is the recommended wide native character-aware way of saying
3332
3333 *(d++) = uv;
3334
3335=cut
3336*/
3337
3338/* On ASCII machines this is normally a macro but we want a
3339 real function in case XS code wants it
3340*/
0f830e0b
NC
3341U8 *
3342Perl_uvchr_to_utf8(pTHX_ U8 *d, UV uv)
3343{
7918f24d
NC
3344 PERL_ARGS_ASSERT_UVCHR_TO_UTF8;
3345
0f830e0b
NC
3346 return Perl_uvuni_to_utf8_flags(aTHX_ d, NATIVE_TO_UNI(uv), 0);
3347}
3348
b851fbc1
JH
3349U8 *
3350Perl_uvchr_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags)
3351{
7918f24d
NC
3352 PERL_ARGS_ASSERT_UVCHR_TO_UTF8_FLAGS;
3353
b851fbc1
JH
3354 return Perl_uvuni_to_utf8_flags(aTHX_ d, NATIVE_TO_UNI(uv), flags);
3355}
2b9d42f0
NIS
3356
3357/*
87cea99e 3358=for apidoc utf8n_to_uvchr
0f830e0b 3359
48ef279e 3360Returns the native character value of the first character in the string
0f830e0b
NC
3361C<s>
3362which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
3363length, in bytes, of that character.
3364
6ee84de2 3365length and flags are the same as utf8n_to_uvuni().
0f830e0b
NC
3366
3367=cut
3368*/
3369/* On ASCII machines this is normally a macro but we want
3370 a real function in case XS code wants it
3371*/
0f830e0b 3372UV
48ef279e 3373Perl_utf8n_to_uvchr(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen,
0f830e0b
NC
3374U32 flags)
3375{
3376 const UV uv = Perl_utf8n_to_uvuni(aTHX_ s, curlen, retlen, flags);
7918f24d
NC
3377
3378 PERL_ARGS_ASSERT_UTF8N_TO_UVCHR;
3379
0f830e0b
NC
3380 return UNI_TO_NATIVE(uv);
3381}
3382
0876b9a0
KW
3383bool
3384Perl_check_utf8_print(pTHX_ register const U8* s, const STRLEN len)
3385{
3386 /* May change: warns if surrogates, non-character code points, or
af2af982
KW
3387 * non-Unicode code points are in s which has length len bytes. Returns
3388 * TRUE if none found; FALSE otherwise. The only other validity check is
3389 * to make sure that this won't exceed the string's length */
0876b9a0
KW
3390
3391 const U8* const e = s + len;
3392 bool ok = TRUE;
3393
3394 PERL_ARGS_ASSERT_CHECK_UTF8_PRINT;
3395
3396 while (s < e) {
3397 if (UTF8SKIP(s) > len) {
3398 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8),
3399 "%s in %s", unees, PL_op ? OP_DESC(PL_op) : "print");
3400 return FALSE;
3401 }
732fbc05 3402 if (UNLIKELY(*s >= UTF8_FIRST_PROBLEMATIC_CODE_POINT_FIRST_BYTE)) {
0876b9a0
KW
3403 STRLEN char_len;
3404 if (UTF8_IS_SUPER(s)) {
8457b38f
KW
3405 if (ckWARN_d(WARN_NON_UNICODE)) {
3406 UV uv = utf8_to_uvchr(s, &char_len);
3407 Perl_warner(aTHX_ packWARN(WARN_NON_UNICODE),
3408 "Code point 0x%04"UVXf" is not Unicode, may not be portable", uv);
3409 ok = FALSE;
3410 }
0876b9a0
KW
3411 }
3412 else if (UTF8_IS_SURROGATE(s)) {
8457b38f
KW
3413 if (ckWARN_d(WARN_SURROGATE)) {
3414 UV uv = utf8_to_uvchr(s, &char_len);
3415 Perl_warner(aTHX_ packWARN(WARN_SURROGATE),
3416 "Unicode surrogate U+%04"UVXf" is illegal in UTF-8", uv);
3417 ok = FALSE;
3418 }
0876b9a0
KW
3419 }
3420 else if
8457b38f
KW
3421 ((UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s))
3422 && (ckWARN_d(WARN_NONCHAR)))
0876b9a0
KW
3423 {
3424 UV uv = utf8_to_uvchr(s, &char_len);
8457b38f 3425 Perl_warner(aTHX_ packWARN(WARN_NONCHAR),
0876b9a0
KW
3426 "Unicode non-character U+%04"UVXf" is illegal for open interchange", uv);
3427 ok = FALSE;
3428 }
3429 }
3430 s += UTF8SKIP(s);
3431 }
3432
3433 return ok;
3434}
3435
0f830e0b 3436/*
87cea99e 3437=for apidoc pv_uni_display
d2cc3551
JH
3438
3439Build to the scalar dsv a displayable version of the string spv,
3440length len, the displayable version being at most pvlim bytes long
3441(if longer, the rest is truncated and "..." will be appended).
0a2ef054 3442
9e55ce06 3443The flags argument can have UNI_DISPLAY_ISPRINT set to display
00e86452 3444isPRINT()able characters as themselves, UNI_DISPLAY_BACKSLASH
0a2ef054
JH
3445to display the \\[nrfta\\] as the backslashed versions (like '\n')
3446(UNI_DISPLAY_BACKSLASH is preferred over UNI_DISPLAY_ISPRINT for \\).
3447UNI_DISPLAY_QQ (and its alias UNI_DISPLAY_REGEX) have both
3448UNI_DISPLAY_BACKSLASH and UNI_DISPLAY_ISPRINT turned on.
3449
d2cc3551
JH
3450The pointer to the PV of the dsv is returned.
3451
3452=cut */
e6b2e755 3453char *
e1ec3a88 3454Perl_pv_uni_display(pTHX_ SV *dsv, const U8 *spv, STRLEN len, STRLEN pvlim, UV flags)
e6b2e755
JH
3455{
3456 int truncated = 0;
e1ec3a88 3457 const char *s, *e;
e6b2e755 3458
7918f24d
NC
3459 PERL_ARGS_ASSERT_PV_UNI_DISPLAY;
3460
76f68e9b 3461 sv_setpvs(dsv, "");
7fddd944 3462 SvUTF8_off(dsv);
e1ec3a88 3463 for (s = (const char *)spv, e = s + len; s < e; s += UTF8SKIP(s)) {
e6b2e755 3464 UV u;
a49f32c6
NC
3465 /* This serves double duty as a flag and a character to print after
3466 a \ when flags & UNI_DISPLAY_BACKSLASH is true.
3467 */
3468 char ok = 0;
c728cb41 3469
e6b2e755
JH
3470 if (pvlim && SvCUR(dsv) >= pvlim) {
3471 truncated++;
3472 break;
3473 }
3474 u = utf8_to_uvchr((U8*)s, 0);
c728cb41 3475 if (u < 256) {
a3b680e6 3476 const unsigned char c = (unsigned char)u & 0xFF;
0bd48802 3477 if (flags & UNI_DISPLAY_BACKSLASH) {
a49f32c6 3478 switch (c) {
c728cb41 3479 case '\n':
a49f32c6 3480 ok = 'n'; break;
c728cb41 3481 case '\r':
a49f32c6 3482 ok = 'r'; break;
c728cb41 3483 case '\t':
a49f32c6 3484 ok = 't'; break;
c728cb41 3485 case '\f':
a49f32c6 3486 ok = 'f'; break;
c728cb41 3487 case '\a':
a49f32c6 3488 ok = 'a'; break;
c728cb41 3489 case '\\':
a49f32c6 3490 ok = '\\'; break;
c728cb41
JH
3491 default: break;
3492 }
a49f32c6 3493 if (ok) {
88c9ea1e 3494 const char string = ok;
76f68e9b 3495 sv_catpvs(dsv, "\\");
5e7aa789 3496 sv_catpvn(dsv, &string, 1);
a49f32c6 3497 }
c728cb41 3498 }
00e86452 3499 /* isPRINT() is the locale-blind version. */
a49f32c6 3500 if (!ok && (flags & UNI_DISPLAY_ISPRINT) && isPRINT(c)) {
88c9ea1e 3501 const char string = c;
5e7aa789 3502 sv_catpvn(dsv, &string, 1);
a49f32c6 3503 ok = 1;
0a2ef054 3504 }
c728cb41
JH
3505 }
3506 if (!ok)
9e55ce06 3507 Perl_sv_catpvf(aTHX_ dsv, "\\x{%"UVxf"}", u);
e6b2e755
JH
3508 }
3509 if (truncated)
396482e1 3510 sv_catpvs(dsv, "...");
48ef279e 3511
e6b2e755
JH
3512 return SvPVX(dsv);
3513}
2b9d42f0 3514
d2cc3551 3515/*
87cea99e 3516=for apidoc sv_uni_display
d2cc3551
JH
3517
3518Build to the scalar dsv a displayable version of the scalar sv,
0a2ef054 3519the displayable version being at most pvlim bytes long
d2cc3551 3520(if longer, the rest is truncated and "..." will be appended).
0a2ef054
JH
3521
3522The flags argument is as in pv_uni_display().
3523
d2cc3551
JH
3524The pointer to the PV of the dsv is returned.
3525
d4c19fe8
AL
3526=cut
3527*/
e6b2e755
JH
3528char *
3529Perl_sv_uni_display(pTHX_ SV *dsv, SV *ssv, STRLEN pvlim, UV flags)
3530{
7918f24d
NC
3531 PERL_ARGS_ASSERT_SV_UNI_DISPLAY;
3532
cfd0369c
NC
3533 return Perl_pv_uni_display(aTHX_ dsv, (const U8*)SvPVX_const(ssv),
3534 SvCUR(ssv), pvlim, flags);
701a277b
JH
3535}
3536
d2cc3551 3537/*
e6226b18 3538=for apidoc foldEQ_utf8
d2cc3551 3539
d51c1b21 3540Returns true if the leading portions of the strings s1 and s2 (either or both
e6226b18 3541of which may be in UTF-8) are the same case-insensitively; false otherwise.
d51c1b21 3542How far into the strings to compare is determined by other input parameters.
8b35872c
KW
3543
3544If u1 is true, the string s1 is assumed to be in UTF-8-encoded Unicode;
3545otherwise it is assumed to be in native 8-bit encoding. Correspondingly for u2
3546with respect to s2.
3547
d51c1b21
KW
3548If the byte length l1 is non-zero, it says how far into s1 to check for fold
3549equality. In other words, s1+l1 will be used as a goal to reach. The
8b35872c
KW
3550scan will not be considered to be a match unless the goal is reached, and
3551scanning won't continue past that goal. Correspondingly for l2 with respect to
3552s2.
3553
3554If pe1 is non-NULL and the pointer it points to is not NULL, that pointer is
3555considered an end pointer beyond which scanning of s1 will not continue under
3556any circumstances. This means that if both l1 and pe1 are specified, and pe1
3557is less than s1+l1, the match will never be successful because it can never
d51c1b21
KW
3558get as far as its goal (and in fact is asserted against). Correspondingly for
3559pe2 with respect to s2.
8b35872c 3560
d51c1b21
KW
3561At least one of s1 and s2 must have a goal (at least one of l1 and l2 must be
3562non-zero), and if both do, both have to be
8b35872c
KW
3563reached for a successful match. Also, if the fold of a character is multiple
3564characters, all of them must be matched (see tr21 reference below for
3565'folding').
3566
e6226b18 3567Upon a successful match, if pe1 is non-NULL,
8b35872c
KW
3568it will be set to point to the beginning of the I<next> character of s1 beyond
3569what was matched. Correspondingly for pe2 and s2.
d2cc3551
JH
3570
3571For case-insensitiveness, the "casefolding" of Unicode is used
3572instead of upper/lowercasing both the characters, see
3573http://www.unicode.org/unicode/reports/tr21/ (Case Mappings).
3574
3575=cut */
a33c29bc
KW
3576
3577/* A flags parameter has been added which may change, and hence isn't
3578 * externally documented. Currently it is:
3579 * 0 for as-documented above
3580 * FOLDEQ_UTF8_NOMIX_ASCII meaning that if a non-ASCII character folds to an
3581 ASCII one, to not match
5e64d0fa
KW
3582 * FOLDEQ_UTF8_LOCALE meaning that locale rules are to be used for code
3583 * points below 256; unicode rules for above 255; and
3584 * folds that cross those boundaries are disallowed,
3585 * like the NOMIX_ASCII option
18f762c3
KW
3586 * FOLDEQ_S1_ALREADY_FOLDED s1 has already been folded before calling this
3587 * routine. This allows that step to be skipped.
3588 * FOLDEQ_S2_ALREADY_FOLDED Similarly.
a33c29bc 3589 */
701a277b 3590I32
eda9cac1 3591Perl_foldEQ_utf8_flags(pTHX_ const char *s1, char **pe1, register UV l1, bool u1, const char *s2, char **pe2, register UV l2, bool u2, U32 flags)
332ddc25 3592{
8b35872c
KW
3593 dVAR;
3594 register const U8 *p1 = (const U8*)s1; /* Point to current char */
3595 register const U8 *p2 = (const U8*)s2;
48ef279e 3596 register const U8 *g1 = NULL; /* goal for s1 */
8b35872c 3597 register const U8 *g2 = NULL;
48ef279e
KW
3598 register const U8 *e1 = NULL; /* Don't scan s1 past this */
3599 register U8 *f1 = NULL; /* Point to current folded */
8b35872c
KW
3600 register const U8 *e2 = NULL;
3601 register U8 *f2 = NULL;
48ef279e 3602 STRLEN n1 = 0, n2 = 0; /* Number of bytes in current char */
8b35872c
KW
3603 U8 foldbuf1[UTF8_MAXBYTES_CASE+1];
3604 U8 foldbuf2[UTF8_MAXBYTES_CASE+1];
8b35872c 3605
eda9cac1 3606 PERL_ARGS_ASSERT_FOLDEQ_UTF8_FLAGS;
8b35872c 3607
18f762c3
KW
3608 /* The algorithm requires that input with the flags on the first line of
3609 * the assert not be pre-folded. */
3610 assert( ! ((flags & (FOLDEQ_UTF8_NOMIX_ASCII | FOLDEQ_UTF8_LOCALE))
3611 && (flags & (FOLDEQ_S1_ALREADY_FOLDED | FOLDEQ_S2_ALREADY_FOLDED))));
3612
8b35872c 3613 if (pe1) {
48ef279e 3614 e1 = *(U8**)pe1;
8b35872c
KW
3615 }
3616
3617 if (l1) {
48ef279e 3618 g1 = (const U8*)s1 + l1;
8b35872c
KW
3619 }
3620
3621 if (pe2) {
48ef279e 3622 e2 = *(U8**)pe2;
8b35872c
KW
3623 }
3624
3625 if (l2) {
48ef279e 3626 g2 = (const U8*)s2 + l2;
8b35872c
KW
3627 }
3628
3629 /* Must have at least one goal */
3630 assert(g1 || g2);
3631
3632 if (g1) {
3633
48ef279e
KW
3634 /* Will never match if goal is out-of-bounds */
3635 assert(! e1 || e1 >= g1);
8b35872c 3636
48ef279e
KW
3637 /* Here, there isn't an end pointer, or it is beyond the goal. We
3638 * only go as far as the goal */
3639 e1 = g1;
8b35872c 3640 }
313b38e5
NC
3641 else {
3642 assert(e1); /* Must have an end for looking at s1 */
3643 }
8b35872c
KW
3644
3645 /* Same for goal for s2 */
3646 if (g2) {
48ef279e
KW
3647 assert(! e2 || e2 >= g2);
3648 e2 = g2;
8b35872c 3649 }
313b38e5
NC
3650 else {
3651 assert(e2);
3652 }
8b35872c 3653
18f762c3
KW
3654 /* If both operands are already folded, we could just do a memEQ on the
3655 * whole strings at once, but it would be better if the caller realized
3656 * this and didn't even call us */
3657
8b35872c
KW
3658 /* Look through both strings, a character at a time */
3659 while (p1 < e1 && p2 < e2) {
3660
d51c1b21 3661 /* If at the beginning of a new character in s1, get its fold to use
5e64d0fa
KW
3662 * and the length of the fold. (exception: locale rules just get the
3663 * character to a single byte) */
48ef279e 3664 if (n1 == 0) {
18f762c3
KW
3665 if (flags & FOLDEQ_S1_ALREADY_FOLDED) {
3666 f1 = (U8 *) p1;
3667 n1 = UTF8SKIP(f1);
a33c29bc 3668
5e64d0fa
KW
3669 /* If in locale matching, we use two sets of rules, depending on if
3670 * the code point is above or below 255. Here, we test for and
3671 * handle locale rules */
18f762c3
KW
3672 }
3673 else {
1b4059d5
KW
3674 if ((flags & FOLDEQ_UTF8_LOCALE)
3675 && (! u1 || UTF8_IS_INVARIANT(*p1)
3676 || UTF8_IS_DOWNGRADEABLE_START(*p1)))
5e64d0fa 3677 {
1b4059d5
KW
3678 /* There is no mixing of code points above and below 255. */
3679 if (u2 && (! UTF8_IS_INVARIANT(*p2)
3680 && ! UTF8_IS_DOWNGRADEABLE_START(*p2)))
3681 {
3682 return 0;
3683 }
5e64d0fa 3684
1b4059d5
KW
3685 /* We handle locale rules by converting, if necessary, the
3686 * code point to a single byte. */
3687 if (! u1 || UTF8_IS_INVARIANT(*p1)) {
3688 *foldbuf1 = *p1;
3689 }
3690 else {
3691 *foldbuf1 = TWO_BYTE_UTF8_TO_UNI(*p1, *(p1 + 1));
3692 }
3693 n1 = 1;
5e64d0fa 3694 }
1b4059d5
KW
3695 else if (isASCII(*p1)) { /* Note, that here won't be
3696 both ASCII and using locale
3697 rules */
3698
3699 /* If trying to mix non- with ASCII, and not supposed to,
3700 * fail */
3701 if ((flags & FOLDEQ_UTF8_NOMIX_ASCII) && ! isASCII(*p2)) {
3702 return 0;
3703 }
3704 n1 = 1;
3705 *foldbuf1 = toLOWER(*p1); /* Folds in the ASCII range are
3706 just lowercased */
5e64d0fa 3707 }
1b4059d5
KW
3708 else if (u1) {
3709 to_utf8_fold(p1, foldbuf1, &n1);
a33c29bc 3710 }
9cef8533
KW
3711 else { /* Not utf8, get utf8 fold */
3712 to_uni_fold(NATIVE_TO_UNI(*p1), foldbuf1, &n1);
1b4059d5
KW
3713 }
3714 f1 = foldbuf1;
18f762c3 3715 }
48ef279e 3716 }
8b35872c 3717
48ef279e 3718 if (n2 == 0) { /* Same for s2 */
18f762c3
KW
3719 if (flags & FOLDEQ_S2_ALREADY_FOLDED) {
3720 f2 = (U8 *) p2;
3721 n2 = UTF8SKIP(f2);
3722 }
3723 else {
227968da
KW
3724 if ((flags & FOLDEQ_UTF8_LOCALE)
3725 && (! u2 || UTF8_IS_INVARIANT(*p2) || UTF8_IS_DOWNGRADEABLE_START(*p2)))
5e64d0fa 3726 {
227968da
KW
3727 /* Here, the next char in s2 is < 256. We've already
3728 * worked on s1, and if it isn't also < 256, can't match */
3729 if (u1 && (! UTF8_IS_INVARIANT(*p1)
3730 && ! UTF8_IS_DOWNGRADEABLE_START(*p1)))
3731 {
3732 return 0;
3733 }
3734 if (! u2 || UTF8_IS_INVARIANT(*p2)) {
3735 *foldbuf2 = *p2;
3736 }
3737 else {
3738 *foldbuf2 = TWO_BYTE_UTF8_TO_UNI(*p2, *(p2 + 1));
3739 }
3740
3741 /* Use another function to handle locale rules. We've made
3742 * sure that both characters to compare are single bytes */
3743 if (! foldEQ_locale((char *) f1, (char *) foldbuf2, 1)) {
3744 return 0;
3745 }
3746 n1 = n2 = 0;
5e64d0fa 3747 }
227968da
KW
3748 else if (isASCII(*p2)) {
3749 if (flags && ! isASCII(*p1)) {
3750 return 0;
3751 }
3752 n2 = 1;
3753 *foldbuf2 = toLOWER(*p2);
5e64d0fa 3754 }
227968da
KW
3755 else if (u2) {
3756 to_utf8_fold(p2, foldbuf2, &n2);
5001101e 3757 }
227968da 3758 else {
9cef8533 3759 to_uni_fold(NATIVE_TO_UNI(*p2), foldbuf2, &n2);
a33c29bc 3760 }
227968da 3761 f2 = foldbuf2;
18f762c3 3762 }
48ef279e 3763 }
8b35872c 3764
5001101e 3765 /* Here f1 and f2 point to the beginning of the strings to compare.
227968da
KW
3766 * These strings are the folds of the next character from each input
3767 * string, stored in utf