This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
ExtUtils::Install: Fix POD error
[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
ST
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 */
f90a9a02
KW
1378#define CALL_UPPER_CASE(INP, OUTP, LENP) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_toupper, "ToUc", "utf8::ToSpecUc")
1379#define CALL_TITLE_CASE(INP, OUTP, LENP) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_totitle, "ToTc", "utf8::ToSpecTc")
1380#define CALL_LOWER_CASE(INP, OUTP, LENP) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_tolower, "ToLc", "utf8::ToSpecLc")
50bda2c3
KW
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 */
f90a9a02 1385#define CALL_FOLD_CASE(INP, OUTP, LENP, SPECIALS) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_tofold, "ToCf", (SPECIALS) ? "utf8::ToSpecCf" : NULL)
c3fd2246 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
f673fad4 1464Perl__to_fold_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp, const bool flags)
a1dde8de 1465{
f673fad4
KW
1466 /* Corresponds to to_lower_latin1(), flags is TRUE if to use full case
1467 * folding */
1468
a1dde8de
KW
1469 UV converted;
1470
1471 PERL_ARGS_ASSERT__TO_FOLD_LATIN1;
1472
1473 if (c == MICRO_SIGN) {
1474 converted = GREEK_SMALL_LETTER_MU;
1475 }
1476 else if (flags && c == LATIN_SMALL_LETTER_SHARP_S) {
1477 *(p)++ = 's';
1478 *p = 's';
1479 *lenp = 2;
1480 return 's';
1481 }
1482 else { /* In this range the fold of all other characters is their lower
1483 case */
1484 converted = toLOWER_LATIN1(c);
1485 }
1486
1487 if (UNI_IS_INVARIANT(converted)) {
1488 *p = (U8) converted;
1489 *lenp = 1;
1490 }
1491 else {
1492 *(p)++ = UTF8_TWO_BYTE_HI(converted);
1493 *p = UTF8_TWO_BYTE_LO(converted);
1494 *lenp = 2;
1495 }
1496
1497 return converted;
1498}
1499
1500UV
f673fad4 1501Perl__to_uni_fold_flags(pTHX_ UV c, U8* p, STRLEN *lenp, const bool flags)
84afefe6 1502{
4b593389
KW
1503
1504 /* Not currently externally documented, and subject to change, <flags> is
f673fad4 1505 * TRUE iff full folding is to be used */
4b593389 1506
36bb2ab6 1507 PERL_ARGS_ASSERT__TO_UNI_FOLD_FLAGS;
7918f24d 1508
a1dde8de
KW
1509 if (c < 256) {
1510 return _to_fold_latin1((U8) c, p, lenp, flags);
1511 }
1512
0ebc6274 1513 uvchr_to_utf8(p, c);
a1dde8de 1514 return CALL_FOLD_CASE(p, p, lenp, flags);
84afefe6
JH
1515}
1516
ea317ccb
KW
1517/* for now these all assume no locale info available for Unicode > 255; and
1518 * the corresponding macros in handy.h (like isALNUM_LC_uvchr) should have been
1519 * called instead, so that these don't get called for < 255 */
a0ed51b3
LW
1520
1521bool
84afefe6 1522Perl_is_uni_alnum_lc(pTHX_ UV c)
a0ed51b3
LW
1523{
1524 return is_uni_alnum(c); /* XXX no locale support yet */
1525}
1526
1527bool
84afefe6 1528Perl_is_uni_idfirst_lc(pTHX_ UV c)
a0ed51b3
LW
1529{
1530 return is_uni_idfirst(c); /* XXX no locale support yet */
1531}
1532
1533bool
84afefe6 1534Perl_is_uni_alpha_lc(pTHX_ UV c)
a0ed51b3
LW
1535{
1536 return is_uni_alpha(c); /* XXX no locale support yet */
1537}
1538
1539bool
84afefe6 1540Perl_is_uni_ascii_lc(pTHX_ UV c)
4d61ec05
GS
1541{
1542 return is_uni_ascii(c); /* XXX no locale support yet */
1543}
1544
1545bool
84afefe6 1546Perl_is_uni_space_lc(pTHX_ UV c)
a0ed51b3
LW
1547{
1548 return is_uni_space(c); /* XXX no locale support yet */
1549}
1550
1551bool
84afefe6 1552Perl_is_uni_digit_lc(pTHX_ UV c)
a0ed51b3
LW
1553{
1554 return is_uni_digit(c); /* XXX no locale support yet */
1555}
1556
1557bool
84afefe6 1558Perl_is_uni_upper_lc(pTHX_ UV c)
a0ed51b3
LW
1559{
1560 return is_uni_upper(c); /* XXX no locale support yet */
1561}
1562
1563bool
84afefe6 1564Perl_is_uni_lower_lc(pTHX_ UV c)
a0ed51b3
LW
1565{
1566 return is_uni_lower(c); /* XXX no locale support yet */
1567}
1568
1569bool
84afefe6 1570Perl_is_uni_cntrl_lc(pTHX_ UV c)
b8c5462f
JH
1571{
1572 return is_uni_cntrl(c); /* XXX no locale support yet */
1573}
1574
1575bool
84afefe6 1576Perl_is_uni_graph_lc(pTHX_ UV c)
b8c5462f
JH
1577{
1578 return is_uni_graph(c); /* XXX no locale support yet */
1579}
1580
1581bool
84afefe6 1582Perl_is_uni_print_lc(pTHX_ UV c)
a0ed51b3
LW
1583{
1584 return is_uni_print(c); /* XXX no locale support yet */
1585}
1586
b8c5462f 1587bool
84afefe6 1588Perl_is_uni_punct_lc(pTHX_ UV c)
b8c5462f
JH
1589{
1590 return is_uni_punct(c); /* XXX no locale support yet */
1591}
1592
4d61ec05 1593bool
84afefe6 1594Perl_is_uni_xdigit_lc(pTHX_ UV c)
4d61ec05
GS
1595{
1596 return is_uni_xdigit(c); /* XXX no locale support yet */
1597}
1598
b7ac61fa
JH
1599U32
1600Perl_to_uni_upper_lc(pTHX_ U32 c)
1601{
ee099d14
JH
1602 /* XXX returns only the first character -- do not use XXX */
1603 /* XXX no locale support yet */
1604 STRLEN len;
89ebb4a3 1605 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
ee099d14 1606 return (U32)to_uni_upper(c, tmpbuf, &len);
b7ac61fa
JH
1607}
1608
1609U32
1610Perl_to_uni_title_lc(pTHX_ U32 c)
1611{
ee099d14
JH
1612 /* XXX returns only the first character XXX -- do not use XXX */
1613 /* XXX no locale support yet */
1614 STRLEN len;
89ebb4a3 1615 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
ee099d14 1616 return (U32)to_uni_title(c, tmpbuf, &len);
b7ac61fa
JH
1617}
1618
1619U32
1620Perl_to_uni_lower_lc(pTHX_ U32 c)
1621{
ee099d14
JH
1622 /* XXX returns only the first character -- do not use XXX */
1623 /* XXX no locale support yet */
1624 STRLEN len;
89ebb4a3 1625 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
ee099d14 1626 return (U32)to_uni_lower(c, tmpbuf, &len);
b7ac61fa
JH
1627}
1628
7452cf6a 1629static bool
5141f98e 1630S_is_utf8_common(pTHX_ const U8 *const p, SV **swash,
bde6a22d
NC
1631 const char *const swashname)
1632{
ea317ccb
KW
1633 /* returns a boolean giving whether or not the UTF8-encoded character that
1634 * starts at <p> is in the swash indicated by <swashname>. <swash>
1635 * contains a pointer to where the swash indicated by <swashname>
1636 * is to be stored; which this routine will do, so that future calls will
1637 * look at <*swash> and only generate a swash if it is not null
1638 *
1639 * Note that it is assumed that the buffer length of <p> is enough to
1640 * contain all the bytes that comprise the character. Thus, <*p> should
1641 * have been checked before this call for mal-formedness enough to assure
1642 * that. */
1643
97aff369 1644 dVAR;
7918f24d
NC
1645
1646 PERL_ARGS_ASSERT_IS_UTF8_COMMON;
1647
bde6a22d
NC
1648 if (!is_utf8_char(p))
1649 return FALSE;
1650 if (!*swash)
711a919c 1651 *swash = swash_init("utf8", swashname, &PL_sv_undef, 1, 0);
bde6a22d
NC
1652 return swash_fetch(*swash, p, TRUE) != 0;
1653}
1654
1655bool
7fc63493 1656Perl_is_utf8_alnum(pTHX_ const U8 *p)
a0ed51b3 1657{
97aff369 1658 dVAR;
7918f24d
NC
1659
1660 PERL_ARGS_ASSERT_IS_UTF8_ALNUM;
1661
671c33bf
NC
1662 /* NOTE: "IsWord", not "IsAlnum", since Alnum is a true
1663 * descendant of isalnum(3), in other words, it doesn't
1664 * contain the '_'. --jhi */
d4c19fe8 1665 return is_utf8_common(p, &PL_utf8_alnum, "IsWord");
a0ed51b3
LW
1666}
1667
1668bool
7fc63493 1669Perl_is_utf8_idfirst(pTHX_ const U8 *p) /* The naming is historical. */
a0ed51b3 1670{
97aff369 1671 dVAR;
7918f24d
NC
1672
1673 PERL_ARGS_ASSERT_IS_UTF8_IDFIRST;
1674
82686b01
JH
1675 if (*p == '_')
1676 return TRUE;
bde6a22d 1677 /* is_utf8_idstart would be more logical. */
d4c19fe8 1678 return is_utf8_common(p, &PL_utf8_idstart, "IdStart");
82686b01
JH
1679}
1680
1681bool
c11ff943
KW
1682Perl_is_utf8_xidfirst(pTHX_ const U8 *p) /* The naming is historical. */
1683{
1684 dVAR;
1685
1686 PERL_ARGS_ASSERT_IS_UTF8_XIDFIRST;
1687
1688 if (*p == '_')
1689 return TRUE;
1690 /* is_utf8_idstart would be more logical. */
1691 return is_utf8_common(p, &PL_utf8_xidstart, "XIdStart");
1692}
1693
1694bool
b6912c02
KW
1695Perl__is_utf8__perl_idstart(pTHX_ const U8 *p)
1696{
1697 dVAR;
1698
1699 PERL_ARGS_ASSERT__IS_UTF8__PERL_IDSTART;
1700
1701 return is_utf8_common(p, &PL_utf8_perl_idstart, "_Perl_IDStart");
1702}
1703
1704bool
7fc63493 1705Perl_is_utf8_idcont(pTHX_ const U8 *p)
82686b01 1706{
97aff369 1707 dVAR;
7918f24d
NC
1708
1709 PERL_ARGS_ASSERT_IS_UTF8_IDCONT;
1710
d4c19fe8 1711 return is_utf8_common(p, &PL_utf8_idcont, "IdContinue");
a0ed51b3
LW
1712}
1713
1714bool
c11ff943
KW
1715Perl_is_utf8_xidcont(pTHX_ const U8 *p)
1716{
1717 dVAR;
1718
1719 PERL_ARGS_ASSERT_IS_UTF8_XIDCONT;
1720
c11ff943
KW
1721 return is_utf8_common(p, &PL_utf8_idcont, "XIdContinue");
1722}
1723
1724bool
7fc63493 1725Perl_is_utf8_alpha(pTHX_ const U8 *p)
a0ed51b3 1726{
97aff369 1727 dVAR;
7918f24d
NC
1728
1729 PERL_ARGS_ASSERT_IS_UTF8_ALPHA;
1730
d4c19fe8 1731 return is_utf8_common(p, &PL_utf8_alpha, "IsAlpha");
a0ed51b3
LW
1732}
1733
1734bool
7fc63493 1735Perl_is_utf8_ascii(pTHX_ const U8 *p)
b8c5462f 1736{
97aff369 1737 dVAR;
7918f24d
NC
1738
1739 PERL_ARGS_ASSERT_IS_UTF8_ASCII;
1740
bc39fe24
KW
1741 /* ASCII characters are the same whether in utf8 or not. So the macro
1742 * works on both utf8 and non-utf8 representations. */
1743 return isASCII(*p);
b8c5462f
JH
1744}
1745
1746bool
7fc63493 1747Perl_is_utf8_space(pTHX_ const U8 *p)
a0ed51b3 1748{
97aff369 1749 dVAR;
7918f24d
NC
1750
1751 PERL_ARGS_ASSERT_IS_UTF8_SPACE;
1752
a34094a9 1753 return is_utf8_common(p, &PL_utf8_space, "IsXPerlSpace");
a0ed51b3
LW
1754}
1755
1756bool
d1eb3177
YO
1757Perl_is_utf8_perl_space(pTHX_ const U8 *p)
1758{
1759 dVAR;
1760
1761 PERL_ARGS_ASSERT_IS_UTF8_PERL_SPACE;
1762
c4428693
KW
1763 /* Only true if is an ASCII space-like character, and ASCII is invariant
1764 * under utf8, so can just use the macro */
1765 return isSPACE_A(*p);
d1eb3177
YO
1766}
1767
1768bool
1769Perl_is_utf8_perl_word(pTHX_ const U8 *p)
1770{
1771 dVAR;
1772
1773 PERL_ARGS_ASSERT_IS_UTF8_PERL_WORD;
1774
c4428693
KW
1775 /* Only true if is an ASCII word character, and ASCII is invariant
1776 * under utf8, so can just use the macro */
1777 return isWORDCHAR_A(*p);
d1eb3177
YO
1778}
1779
1780bool
7fc63493 1781Perl_is_utf8_digit(pTHX_ const U8 *p)
a0ed51b3 1782{
97aff369 1783 dVAR;
7918f24d
NC
1784
1785 PERL_ARGS_ASSERT_IS_UTF8_DIGIT;
1786
d4c19fe8 1787 return is_utf8_common(p, &PL_utf8_digit, "IsDigit");
a0ed51b3
LW
1788}
1789
1790bool
d1eb3177
YO
1791Perl_is_utf8_posix_digit(pTHX_ const U8 *p)
1792{
1793 dVAR;
1794
1795 PERL_ARGS_ASSERT_IS_UTF8_POSIX_DIGIT;
1796
c4428693
KW
1797 /* Only true if is an ASCII digit character, and ASCII is invariant
1798 * under utf8, so can just use the macro */
1799 return isDIGIT_A(*p);
d1eb3177
YO
1800}
1801
1802bool
7fc63493 1803Perl_is_utf8_upper(pTHX_ const U8 *p)
a0ed51b3 1804{
97aff369 1805 dVAR;
7918f24d
NC
1806
1807 PERL_ARGS_ASSERT_IS_UTF8_UPPER;
1808
d4c19fe8 1809 return is_utf8_common(p, &PL_utf8_upper, "IsUppercase");
a0ed51b3
LW
1810}
1811
1812bool
7fc63493 1813Perl_is_utf8_lower(pTHX_ const U8 *p)
a0ed51b3 1814{
97aff369 1815 dVAR;
7918f24d
NC
1816
1817 PERL_ARGS_ASSERT_IS_UTF8_LOWER;
1818
d4c19fe8 1819 return is_utf8_common(p, &PL_utf8_lower, "IsLowercase");
a0ed51b3
LW
1820}
1821
1822bool
7fc63493 1823Perl_is_utf8_cntrl(pTHX_ const U8 *p)
b8c5462f 1824{
97aff369 1825 dVAR;
7918f24d
NC
1826
1827 PERL_ARGS_ASSERT_IS_UTF8_CNTRL;
1828
7b952154
KW
1829 if (isASCII(*p)) {
1830 return isCNTRL_A(*p);
1831 }
1832
1833 /* All controls are in Latin1 */
1834 if (! UTF8_IS_DOWNGRADEABLE_START(*p)) {
1835 return 0;
1836 }
1837 return isCNTRL_L1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)));
b8c5462f
JH
1838}
1839
1840bool
7fc63493 1841Perl_is_utf8_graph(pTHX_ const U8 *p)
b8c5462f 1842{
97aff369 1843 dVAR;
7918f24d
NC
1844
1845 PERL_ARGS_ASSERT_IS_UTF8_GRAPH;
1846
d4c19fe8 1847 return is_utf8_common(p, &PL_utf8_graph, "IsGraph");
b8c5462f
JH
1848}
1849
1850bool
7fc63493 1851Perl_is_utf8_print(pTHX_ const U8 *p)
a0ed51b3 1852{
97aff369 1853 dVAR;
7918f24d
NC
1854
1855 PERL_ARGS_ASSERT_IS_UTF8_PRINT;
1856
d4c19fe8 1857 return is_utf8_common(p, &PL_utf8_print, "IsPrint");
a0ed51b3
LW
1858}
1859
1860bool
7fc63493 1861Perl_is_utf8_punct(pTHX_ const U8 *p)
b8c5462f 1862{
97aff369 1863 dVAR;
7918f24d
NC
1864
1865 PERL_ARGS_ASSERT_IS_UTF8_PUNCT;
1866
d4c19fe8 1867 return is_utf8_common(p, &PL_utf8_punct, "IsPunct");
b8c5462f
JH
1868}
1869
1870bool
7fc63493 1871Perl_is_utf8_xdigit(pTHX_ const U8 *p)
b8c5462f 1872{
97aff369 1873 dVAR;
7918f24d
NC
1874
1875 PERL_ARGS_ASSERT_IS_UTF8_XDIGIT;
1876
d1eb3177 1877 return is_utf8_common(p, &PL_utf8_xdigit, "IsXDigit");
b8c5462f
JH
1878}
1879
1880bool
7fc63493 1881Perl_is_utf8_mark(pTHX_ const U8 *p)
a0ed51b3 1882{
97aff369 1883 dVAR;
7918f24d
NC
1884
1885 PERL_ARGS_ASSERT_IS_UTF8_MARK;
1886
d4c19fe8 1887 return is_utf8_common(p, &PL_utf8_mark, "IsM");
a0ed51b3
LW
1888}
1889
37e2e78e
KW
1890bool
1891Perl_is_utf8_X_begin(pTHX_ const U8 *p)
1892{
1893 dVAR;
1894
1895 PERL_ARGS_ASSERT_IS_UTF8_X_BEGIN;
1896
1897 return is_utf8_common(p, &PL_utf8_X_begin, "_X_Begin");
1898}
1899
1900bool
1901Perl_is_utf8_X_extend(pTHX_ const U8 *p)
1902{
1903 dVAR;
1904
1905 PERL_ARGS_ASSERT_IS_UTF8_X_EXTEND;
1906
1907 return is_utf8_common(p, &PL_utf8_X_extend, "_X_Extend");
1908}
1909
1910bool
1911Perl_is_utf8_X_prepend(pTHX_ const U8 *p)
1912{
1913 dVAR;
1914
1915 PERL_ARGS_ASSERT_IS_UTF8_X_PREPEND;
1916
1917 return is_utf8_common(p, &PL_utf8_X_prepend, "GCB=Prepend");
1918}
1919
1920bool
1921Perl_is_utf8_X_non_hangul(pTHX_ const U8 *p)
1922{
1923 dVAR;
1924
1925 PERL_ARGS_ASSERT_IS_UTF8_X_NON_HANGUL;
1926
1927 return is_utf8_common(p, &PL_utf8_X_non_hangul, "HST=Not_Applicable");
1928}
1929
1930bool
1931Perl_is_utf8_X_L(pTHX_ const U8 *p)
1932{
1933 dVAR;
1934
1935 PERL_ARGS_ASSERT_IS_UTF8_X_L;
1936
1937 return is_utf8_common(p, &PL_utf8_X_L, "GCB=L");
1938}
1939
1940bool
1941Perl_is_utf8_X_LV(pTHX_ const U8 *p)
1942{
1943 dVAR;
1944
1945 PERL_ARGS_ASSERT_IS_UTF8_X_LV;
1946
1947 return is_utf8_common(p, &PL_utf8_X_LV, "GCB=LV");
1948}
1949
1950bool
1951Perl_is_utf8_X_LVT(pTHX_ const U8 *p)
1952{
1953 dVAR;
1954
1955 PERL_ARGS_ASSERT_IS_UTF8_X_LVT;
1956
1957 return is_utf8_common(p, &PL_utf8_X_LVT, "GCB=LVT");
1958}
1959
1960bool
1961Perl_is_utf8_X_T(pTHX_ const U8 *p)
1962{
1963 dVAR;
1964
1965 PERL_ARGS_ASSERT_IS_UTF8_X_T;
1966
1967 return is_utf8_common(p, &PL_utf8_X_T, "GCB=T");
1968}
1969
1970bool
1971Perl_is_utf8_X_V(pTHX_ const U8 *p)
1972{
1973 dVAR;
1974
1975 PERL_ARGS_ASSERT_IS_UTF8_X_V;
1976
1977 return is_utf8_common(p, &PL_utf8_X_V, "GCB=V");
1978}
1979
1980bool
1981Perl_is_utf8_X_LV_LVT_V(pTHX_ const U8 *p)
1982{
1983 dVAR;
1984
1985 PERL_ARGS_ASSERT_IS_UTF8_X_LV_LVT_V;
1986
1987 return is_utf8_common(p, &PL_utf8_X_LV_LVT_V, "_X_LV_LVT_V");
1988}
1989
6b5c0936 1990/*
87cea99e 1991=for apidoc to_utf8_case
6b5c0936
JH
1992
1993The "p" contains the pointer to the UTF-8 string encoding
1994the character that is being converted.
1995
1996The "ustrp" is a pointer to the character buffer to put the
1997conversion result to. The "lenp" is a pointer to the length
1998of the result.
1999
0134edef 2000The "swashp" is a pointer to the swash to use.
6b5c0936 2001
36bb2ab6 2002Both the special and normal mappings are stored in lib/unicore/To/Foo.pl,
8fe4d5b2 2003and loaded by SWASHNEW, using lib/utf8_heavy.pl. The special (usually,
0134edef 2004but not always, a multicharacter mapping), is tried first.
6b5c0936 2005
0134edef
JH
2006The "special" is a string like "utf8::ToSpecLower", which means the
2007hash %utf8::ToSpecLower. The access to the hash is through
2008Perl_to_utf8_case().
6b5c0936 2009
0134edef
JH
2010The "normal" is a string like "ToLower" which means the swash
2011%utf8::ToLower.
2012
2013=cut */
6b5c0936 2014
2104c8d9 2015UV
9a957fbc
AL
2016Perl_to_utf8_case(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp,
2017 SV **swashp, const char *normal, const char *special)
a0ed51b3 2018{
97aff369 2019 dVAR;
89ebb4a3 2020 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
0134edef 2021 STRLEN len = 0;
aec46f14 2022 const UV uv0 = utf8_to_uvchr(p, NULL);
1feea2c7
JH
2023 /* The NATIVE_TO_UNI() and UNI_TO_NATIVE() mappings
2024 * are necessary in EBCDIC, they are redundant no-ops
2025 * in ASCII-ish platforms, and hopefully optimized away. */
f54cb97a 2026 const UV uv1 = NATIVE_TO_UNI(uv0);
7918f24d
NC
2027
2028 PERL_ARGS_ASSERT_TO_UTF8_CASE;
2029
9ae3ac1a
KW
2030 /* Note that swash_fetch() doesn't output warnings for these because it
2031 * assumes we will */
8457b38f 2032 if (uv1 >= UNICODE_SURROGATE_FIRST) {
9ae3ac1a 2033 if (uv1 <= UNICODE_SURROGATE_LAST) {
8457b38f
KW
2034 if (ckWARN_d(WARN_SURROGATE)) {
2035 const char* desc = (PL_op) ? OP_DESC(PL_op) : normal;
2036 Perl_warner(aTHX_ packWARN(WARN_SURROGATE),
2037 "Operation \"%s\" returns its argument for UTF-16 surrogate U+%04"UVXf"", desc, uv1);
2038 }
9ae3ac1a
KW
2039 }
2040 else if (UNICODE_IS_SUPER(uv1)) {
8457b38f
KW
2041 if (ckWARN_d(WARN_NON_UNICODE)) {
2042 const char* desc = (PL_op) ? OP_DESC(PL_op) : normal;
2043 Perl_warner(aTHX_ packWARN(WARN_NON_UNICODE),
2044 "Operation \"%s\" returns its argument for non-Unicode code point 0x%04"UVXf"", desc, uv1);
2045 }
9ae3ac1a
KW
2046 }
2047
2048 /* Note that non-characters are perfectly legal, so no warning should
2049 * be given */
2050 }
2051
1feea2c7 2052 uvuni_to_utf8(tmpbuf, uv1);
0134edef
JH
2053
2054 if (!*swashp) /* load on-demand */
2055 *swashp = swash_init("utf8", normal, &PL_sv_undef, 4, 0);
2056
a6f87d8c 2057 if (special) {
0134edef 2058 /* It might be "special" (sometimes, but not always,
2a37f04d 2059 * a multicharacter mapping) */
6673a63c 2060 HV * const hv = get_hv(special, 0);
b08cf34e
JH
2061 SV **svp;
2062
35da51f7 2063 if (hv &&
b08cf34e
JH
2064 (svp = hv_fetch(hv, (const char*)tmpbuf, UNISKIP(uv1), FALSE)) &&
2065 (*svp)) {
cfd0369c 2066 const char *s;
47654450 2067
cfd0369c 2068 s = SvPV_const(*svp, len);
47654450
JH
2069 if (len == 1)
2070 len = uvuni_to_utf8(ustrp, NATIVE_TO_UNI(*(U8*)s)) - ustrp;
2a37f04d 2071 else {
2f9475ad
JH
2072#ifdef EBCDIC
2073 /* If we have EBCDIC we need to remap the characters
2074 * since any characters in the low 256 are Unicode
2075 * code points, not EBCDIC. */
7cda7a3d 2076 U8 *t = (U8*)s, *tend = t + len, *d;
2f9475ad
JH
2077
2078 d = tmpbuf;
b08cf34e 2079 if (SvUTF8(*svp)) {
2f9475ad
JH
2080 STRLEN tlen = 0;
2081
2082 while (t < tend) {
d4c19fe8 2083 const UV c = utf8_to_uvchr(t, &tlen);
2f9475ad
JH
2084 if (tlen > 0) {
2085 d = uvchr_to_utf8(d, UNI_TO_NATIVE(c));
2086 t += tlen;
2087 }
2088 else
2089 break;
2090 }
2091 }
2092 else {
36fec512
JH
2093 while (t < tend) {
2094 d = uvchr_to_utf8(d, UNI_TO_NATIVE(*t));
2095 t++;
2096 }
2f9475ad
JH
2097 }
2098 len = d - tmpbuf;
2099 Copy(tmpbuf, ustrp, len, U8);
2100#else
d2dcd0fb 2101 Copy(s, ustrp, len, U8);
2f9475ad 2102#endif
29e98929 2103 }
983ffd37 2104 }
0134edef
JH
2105 }
2106
2107 if (!len && *swashp) {
d4c19fe8
AL
2108 const UV uv2 = swash_fetch(*swashp, tmpbuf, TRUE);
2109
0134edef
JH
2110 if (uv2) {
2111 /* It was "normal" (a single character mapping). */
d4c19fe8 2112 const UV uv3 = UNI_TO_NATIVE(uv2);
e9101d72 2113 len = uvchr_to_utf8(ustrp, uv3) - ustrp;
2a37f04d
JH
2114 }
2115 }
1feea2c7 2116
37e2e78e
KW
2117 if (!len) /* Neither: just copy. In other words, there was no mapping
2118 defined, which means that the code point maps to itself */
0134edef
JH
2119 len = uvchr_to_utf8(ustrp, uv0) - ustrp;
2120
2a37f04d
JH
2121 if (lenp)
2122 *lenp = len;
2123
0134edef 2124 return len ? utf8_to_uvchr(ustrp, 0) : 0;
a0ed51b3
LW
2125}
2126
051a06d4
KW
2127STATIC UV
2128S_check_locale_boundary_crossing(pTHX_ const U8* const p, const UV result, U8* const ustrp, STRLEN *lenp)
2129{
2130 /* This is called when changing the case of a utf8-encoded character above
2131 * the Latin1 range, and the operation is in locale. If the result
2132 * contains a character that crosses the 255/256 boundary, disallow the
2133 * change, and return the original code point. See L<perlfunc/lc> for why;
2134 *
2135 * p points to the original string whose case was changed
2136 * result the code point of the first character in the changed-case string
2137 * ustrp points to the changed-case string (<result> represents its first char)
2138 * lenp points to the length of <ustrp> */
2139
2140 UV original; /* To store the first code point of <p> */
2141
2142 PERL_ARGS_ASSERT_CHECK_LOCALE_BOUNDARY_CROSSING;
2143
2144 assert(! UTF8_IS_INVARIANT(*p) && ! UTF8_IS_DOWNGRADEABLE_START(*p));
2145
2146 /* We know immediately if the first character in the string crosses the
2147 * boundary, so can skip */
2148 if (result > 255) {
2149
2150 /* Look at every character in the result; if any cross the
2151 * boundary, the whole thing is disallowed */
2152 U8* s = ustrp + UTF8SKIP(ustrp);
2153 U8* e = ustrp + *lenp;
2154 while (s < e) {
2155 if (UTF8_IS_INVARIANT(*s) || UTF8_IS_DOWNGRADEABLE_START(*s))
2156 {
2157 goto bad_crossing;
2158 }
2159 s += UTF8SKIP(s);
2160 }
2161
2162 /* Here, no characters crossed, result is ok as-is */
2163 return result;
2164 }
2165
2166bad_crossing:
2167
2168 /* Failed, have to return the original */
2169 original = utf8_to_uvchr(p, lenp);
2170 Copy(p, ustrp, *lenp, char);
2171 return original;
2172}
2173
d3e79532 2174/*
87cea99e 2175=for apidoc to_utf8_upper
d3e79532
JH
2176
2177Convert the UTF-8 encoded character at p to its uppercase version and
2178store that in UTF-8 in ustrp and its length in bytes in lenp. Note
89ebb4a3
JH
2179that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since
2180the uppercase version may be longer than the original character.
d3e79532
JH
2181
2182The first character of the uppercased version is returned
2183(but note, as explained above, that there may be more.)
2184
2185=cut */
2186
051a06d4
KW
2187/* Not currently externally documented, and subject to change:
2188 * <flags> is set iff locale semantics are to be used for code points < 256
2189 * <tainted_ptr> if non-null, *tainted_ptr will be set TRUE iff locale rules
2190 * were used in the calculation; otherwise unchanged. */
2191
2104c8d9 2192UV
051a06d4 2193Perl__to_utf8_upper_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, const bool flags, bool* tainted_ptr)
a0ed51b3 2194{
97aff369 2195 dVAR;
7918f24d 2196
051a06d4
KW
2197 UV result;
2198
2199 PERL_ARGS_ASSERT__TO_UTF8_UPPER_FLAGS;
7918f24d 2200
3a4c58c9 2201 if (UTF8_IS_INVARIANT(*p)) {
051a06d4
KW
2202 if (flags) {
2203 result = toUPPER_LC(*p);
2204 }
2205 else {
81c6c7ce 2206 return _to_upper_title_latin1(*p, ustrp, lenp, 'S');
051a06d4 2207 }
3a4c58c9
KW
2208 }
2209 else if UTF8_IS_DOWNGRADEABLE_START(*p) {
051a06d4
KW
2210 if (flags) {
2211 result = toUPPER_LC(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)));
2212 }
2213 else {
81c6c7ce
KW
2214 return _to_upper_title_latin1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)),
2215 ustrp, lenp, 'S');
051a06d4
KW
2216 }
2217 }
2218 else { /* utf8, ord above 255 */
2219 result = CALL_UPPER_CASE(p, ustrp, lenp);
2220
2221 if (flags) {
2222 result = check_locale_boundary_crossing(p, result, ustrp, lenp);
2223 }
2224 return result;
2225 }
2226
2227 /* Here, used locale rules. Convert back to utf8 */
2228 if (UTF8_IS_INVARIANT(result)) {
2229 *ustrp = (U8) result;
2230 *lenp = 1;
2231 }
2232 else {
2233 *ustrp = UTF8_EIGHT_BIT_HI(result);
2234 *(ustrp + 1) = UTF8_EIGHT_BIT_LO(result);
2235 *lenp = 2;
3a4c58c9
KW
2236 }
2237
051a06d4
KW
2238 if (tainted_ptr) {
2239 *tainted_ptr = TRUE;
2240 }
2241 return result;
983ffd37 2242}
a0ed51b3 2243
d3e79532 2244/*
87cea99e 2245=for apidoc to_utf8_title
d3e79532
JH
2246
2247Convert the UTF-8 encoded character at p to its titlecase version and
2248store that in UTF-8 in ustrp and its length in bytes in lenp. Note
89ebb4a3
JH
2249that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
2250titlecase version may be longer than the original character.
d3e79532
JH
2251
2252The first character of the titlecased version is returned
2253(but note, as explained above, that there may be more.)
2254
2255=cut */
2256
051a06d4
KW
2257/* Not currently externally documented, and subject to change:
2258 * <flags> is set iff locale semantics are to be used for code points < 256
2259 * Since titlecase is not defined in POSIX, uppercase is used instead
2260 * for these/
2261 * <tainted_ptr> if non-null, *tainted_ptr will be set TRUE iff locale rules
2262 * were used in the calculation; otherwise unchanged. */
2263
983ffd37 2264UV
051a06d4 2265Perl__to_utf8_title_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, const bool flags, bool* tainted_ptr)
983ffd37 2266{
97aff369 2267 dVAR;
7918f24d 2268
051a06d4
KW
2269 UV result;
2270
2271 PERL_ARGS_ASSERT__TO_UTF8_TITLE_FLAGS;
7918f24d 2272
3a4c58c9 2273 if (UTF8_IS_INVARIANT(*p)) {
051a06d4
KW
2274 if (flags) {
2275 result = toUPPER_LC(*p);
2276 }
2277 else {
81c6c7ce 2278 return _to_upper_title_latin1(*p, ustrp, lenp, 's');
051a06d4 2279 }
3a4c58c9
KW
2280 }
2281 else if UTF8_IS_DOWNGRADEABLE_START(*p) {
051a06d4
KW
2282 if (flags) {
2283 result = toUPPER_LC(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)));
2284 }
2285 else {
81c6c7ce
KW
2286 return _to_upper_title_latin1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)),
2287 ustrp, lenp, 's');
051a06d4
KW
2288 }
2289 }
2290 else { /* utf8, ord above 255 */
2291 result = CALL_TITLE_CASE(p, ustrp, lenp);
2292
2293 if (flags) {
2294 result = check_locale_boundary_crossing(p, result, ustrp, lenp);
2295 }
2296 return result;
2297 }
2298
2299 /* Here, used locale rules. Convert back to utf8 */
2300 if (UTF8_IS_INVARIANT(result)) {
2301 *ustrp = (U8) result;
2302 *lenp = 1;
2303 }
2304 else {
2305 *ustrp = UTF8_EIGHT_BIT_HI(result);
2306 *(ustrp + 1) = UTF8_EIGHT_BIT_LO(result);
2307 *lenp = 2;
3a4c58c9
KW
2308 }
2309
051a06d4
KW
2310 if (tainted_ptr) {
2311 *tainted_ptr = TRUE;
2312 }
2313 return result;
a0ed51b3
LW
2314}
2315
d3e79532 2316/*
87cea99e 2317=for apidoc to_utf8_lower
d3e79532
JH
2318
2319Convert the UTF-8 encoded character at p to its lowercase version and
2320store that in UTF-8 in ustrp and its length in bytes in lenp. Note
89ebb4a3
JH
2321that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
2322lowercase version may be longer than the original character.
d3e79532
JH
2323
2324The first character of the lowercased version is returned
2325(but note, as explained above, that there may be more.)
2326
2327=cut */
2328
051a06d4
KW
2329/* Not currently externally documented, and subject to change:
2330 * <flags> is set iff locale semantics are to be used for code points < 256
2331 * <tainted_ptr> if non-null, *tainted_ptr will be set TRUE iff locale rules
2332 * were used in the calculation; otherwise unchanged. */
2333
2104c8d9 2334UV
051a06d4 2335Perl__to_utf8_lower_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, const bool flags, bool* tainted_ptr)
a0ed51b3 2336{
051a06d4
KW
2337 UV result;
2338
97aff369 2339 dVAR;
7918f24d 2340
051a06d4 2341 PERL_ARGS_ASSERT__TO_UTF8_LOWER_FLAGS;
7918f24d 2342
968c5e6a 2343 if (UTF8_IS_INVARIANT(*p)) {
051a06d4
KW
2344 if (flags) {
2345 result = toLOWER_LC(*p);
2346 }
2347 else {
81c6c7ce 2348 return to_lower_latin1(*p, ustrp, lenp);
051a06d4 2349 }
968c5e6a
KW
2350 }
2351 else if UTF8_IS_DOWNGRADEABLE_START(*p) {
051a06d4
KW
2352 if (flags) {
2353 result = toLOWER_LC(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)));
2354 }
2355 else {
81c6c7ce
KW
2356 return to_lower_latin1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)),
2357 ustrp, lenp);
051a06d4 2358 }
968c5e6a 2359 }
051a06d4
KW
2360 else { /* utf8, ord above 255 */
2361 result = CALL_LOWER_CASE(p, ustrp, lenp);
2362
2363 if (flags) {
2364 result = check_locale_boundary_crossing(p, result, ustrp, lenp);
2365 }
968c5e6a 2366
051a06d4
KW
2367 return result;
2368 }
2369
2370 /* Here, used locale rules. Convert back to utf8 */
2371 if (UTF8_IS_INVARIANT(result)) {
2372 *ustrp = (U8) result;
2373 *lenp = 1;
2374 }
2375 else {
2376 *ustrp = UTF8_EIGHT_BIT_HI(result);
2377 *(ustrp + 1) = UTF8_EIGHT_BIT_LO(result);
2378 *lenp = 2;
2379 }
2380
2381 if (tainted_ptr) {
2382 *tainted_ptr = TRUE;
2383 }
2384 return result;
b4e400f9
JH
2385}
2386
d3e79532 2387/*
87cea99e 2388=for apidoc to_utf8_fold
d3e79532
JH
2389
2390Convert the UTF-8 encoded character at p to its foldcase version and
2391store that in UTF-8 in ustrp and its length in bytes in lenp. Note
89ebb4a3 2392that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
d3e79532
JH
2393foldcase version may be longer than the original character (up to
2394three characters).
2395
2396The first character of the foldcased version is returned
2397(but note, as explained above, that there may be more.)
2398
2399=cut */
2400
051a06d4
KW
2401/* Not currently externally documented, and subject to change,
2402 * in <flags>
2403 * bit FOLD_FLAGS_LOCALE is set iff locale semantics are to be used for code
2404 * points < 256. Since foldcase is not defined in
2405 * POSIX, lowercase is used instead
2406 * bit FOLD_FLAGS_FULL is set iff full case folds are to be used;
2407 * otherwise simple folds
2408 * <tainted_ptr> if non-null, *tainted_ptr will be set TRUE iff locale rules
2409 * were used in the calculation; otherwise unchanged. */
36bb2ab6 2410
b4e400f9 2411UV
051a06d4 2412Perl__to_utf8_fold_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, U8 flags, bool* tainted_ptr)
b4e400f9 2413{
97aff369 2414 dVAR;
7918f24d 2415
051a06d4
KW
2416 UV result;
2417
36bb2ab6 2418 PERL_ARGS_ASSERT__TO_UTF8_FOLD_FLAGS;
7918f24d 2419
a1dde8de 2420 if (UTF8_IS_INVARIANT(*p)) {
051a06d4
KW
2421 if (flags & FOLD_FLAGS_LOCALE) {
2422 result = toLOWER_LC(*p);
2423 }
2424 else {
81c6c7ce
KW
2425 return _to_fold_latin1(*p, ustrp, lenp,
2426 cBOOL(flags & FOLD_FLAGS_FULL));
051a06d4 2427 }
a1dde8de
KW
2428 }
2429 else if UTF8_IS_DOWNGRADEABLE_START(*p) {
051a06d4
KW
2430 if (flags & FOLD_FLAGS_LOCALE) {
2431 result = toLOWER_LC(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)));
2432 }
2433 else {
81c6c7ce 2434 return _to_fold_latin1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)),
051a06d4
KW
2435 ustrp, lenp, cBOOL(flags & FOLD_FLAGS_FULL));
2436 }
a1dde8de 2437 }
051a06d4
KW
2438 else { /* utf8, ord above 255 */
2439 result = CALL_FOLD_CASE(p, ustrp, lenp, flags);
a1dde8de 2440
051a06d4
KW
2441 if ((flags & FOLD_FLAGS_LOCALE)) {
2442 result = check_locale_boundary_crossing(p, result, ustrp, lenp);
2443 }
2444
2445 return result;
2446 }
2447
2448 /* Here, used locale rules. Convert back to utf8 */
2449 if (UTF8_IS_INVARIANT(result)) {
2450 *ustrp = (U8) result;
2451 *lenp = 1;
2452 }
2453 else {
2454 *ustrp = UTF8_EIGHT_BIT_HI(result);
2455 *(ustrp + 1) = UTF8_EIGHT_BIT_LO(result);
2456 *lenp = 2;
2457 }
2458
2459 if (tainted_ptr) {
2460 *tainted_ptr = TRUE;
2461 }
2462 return result;
a0ed51b3
LW
2463}
2464
711a919c 2465/* Note:
f90a9a02 2466 * Returns a "swash" which is a hash described in utf8.c:Perl_swash_fetch().
711a919c
ST
2467 * C<pkg> is a pointer to a package name for SWASHNEW, should be "utf8".
2468 * For other parameters, see utf8::SWASHNEW in lib/utf8_heavy.pl.
2469 */
c4a5db0c 2470
a0ed51b3 2471SV*
7fc63493 2472Perl_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv, I32 minbits, I32 none)
a0ed51b3 2473{
c4a5db0c
KW
2474 PERL_ARGS_ASSERT_SWASH_INIT;
2475
2476 /* Returns a copy of a swash initiated by the called function. This is the
2477 * public interface, and returning a copy prevents others from doing
2478 * mischief on the original */
2479
9a53f6cf 2480 return newSVsv(_core_swash_init(pkg, name, listsv, minbits, none, FALSE, NULL, FALSE));
c4a5db0c
KW
2481}
2482
2483SV*
9a53f6cf 2484Perl__core_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv, I32 minbits, I32 none, bool return_if_undef, SV* invlist, bool passed_in_invlist_has_user_defined_property)
c4a5db0c
KW
2485{
2486 /* Initialize and return a swash, creating it if necessary. It does this
9a53f6cf 2487 * by calling utf8_heavy.pl in the general case.
c4a5db0c
KW
2488 *
2489 * This interface should only be used by functions that won't destroy or
2490 * adversely change the swash, as doing so affects all other uses of the
2491 * swash in the program; the general public should use 'Perl_swash_init'
2492 * instead.
2493 *
2494 * pkg is the name of the package that <name> should be in.
2495 * name is the name of the swash to find. Typically it is a Unicode
2496 * property name, including user-defined ones
2497 * listsv is a string to initialize the swash with. It must be of the form
2498 * documented as the subroutine return value in
2499 * L<perlunicode/User-Defined Character Properties>
2500 * minbits is the number of bits required to represent each data element.
2501 * It is '1' for binary properties.
2502 * none I (khw) do not understand this one, but it is used only in tr///.
9a53f6cf
KW
2503 * return_if_undef is TRUE if the routine shouldn't croak if it can't find
2504 * the requested property
2505 * invlist is an inversion list to initialize the swash with (or NULL)
2506 * has_user_defined_property is TRUE if <invlist> has some component that
2507 * came from a user-defined property
2508 *
2509 * Thus there are three possible inputs to find the swash: <name>,
2510 * <listsv>, and <invlist>. At least one must be specified. The result
2511 * will be the union of the specified ones, although <listsv>'s various
2512 * actions can intersect, etc. what <name> gives.
2513 *
2514 * <invlist> is only valid for binary properties */
c4a5db0c 2515
27da23d5 2516 dVAR;
c4a5db0c 2517 SV* retval = &PL_sv_undef;
9a53f6cf
KW
2518
2519 assert(listsv != &PL_sv_undef || strNE(name, "") || invlist);
2520 assert(! invlist || minbits == 1);
2521
2522 /* If data was passed in to go out to utf8_heavy to find the swash of, do
2523 * so */
2524 if (listsv != &PL_sv_undef || strNE(name, "")) {
69794297
KW
2525 dSP;
2526 const size_t pkg_len = strlen(pkg);
2527 const size_t name_len = strlen(name);
2528 HV * const stash = gv_stashpvn(pkg, pkg_len, 0);
2529 SV* errsv_save;
2530 GV *method;
2531
2532 PERL_ARGS_ASSERT__CORE_SWASH_INIT;
2533
2534 PUSHSTACKi(PERLSI_MAGIC);
ce3b816e 2535 ENTER;
69794297
KW
2536 SAVEHINTS();
2537 save_re_context();
2538 if (PL_parser && PL_parser->error_count)
2539 SAVEI8(PL_parser->error_count), PL_parser->error_count = 0;
2540 method = gv_fetchmeth(stash, "SWASHNEW", 8, -1);
2541 if (!method) { /* demand load utf8 */
2542 ENTER;
2543 errsv_save = newSVsv(ERRSV);
2544 /* It is assumed that callers of this routine are not passing in
2545 * any user derived data. */
2546 /* Need to do this after save_re_context() as it will set
2547 * PL_tainted to 1 while saving $1 etc (see the code after getrx:
2548 * in Perl_magic_get). Even line to create errsv_save can turn on
2549 * PL_tainted. */
2550 SAVEBOOL(PL_tainted);
2551 PL_tainted = 0;
2552 Perl_load_module(aTHX_ PERL_LOADMOD_NOIMPORT, newSVpvn(pkg,pkg_len),
2553 NULL);
2554 if (!SvTRUE(ERRSV))
2555 sv_setsv(ERRSV, errsv_save);
2556 SvREFCNT_dec(errsv_save);
2557 LEAVE;
2558 }
2559 SPAGAIN;
2560 PUSHMARK(SP);
2561 EXTEND(SP,5);
2562 mPUSHp(pkg, pkg_len);
2563 mPUSHp(name, name_len);
2564 PUSHs(listsv);
2565 mPUSHi(minbits);
2566 mPUSHi(none);
2567 PUTBACK;
f8be5cf0 2568 errsv_save = newSVsv(ERRSV);
69794297
KW
2569 /* If we already have a pointer to the method, no need to use
2570 * call_method() to repeat the lookup. */
2571 if (method ? call_sv(MUTABLE_SV(method), G_SCALAR)
2572 : call_sv(newSVpvs_flags("SWASHNEW", SVs_TEMP), G_SCALAR | G_METHOD))
2573 {
2574 retval = *PL_stack_sp--;
2575 SvREFCNT_inc(retval);
2576 }
f8be5cf0
JH
2577 if (!SvTRUE(ERRSV))
2578 sv_setsv(ERRSV, errsv_save);
2579 SvREFCNT_dec(errsv_save);
ce3b816e 2580 LEAVE;
69794297
KW
2581 POPSTACK;
2582 if (IN_PERL_COMPILETIME) {
2583 CopHINTS_set(PL_curcop, PL_hints);
2584 }
2585 if (!SvROK(retval) || SvTYPE(SvRV(retval)) != SVt_PVHV) {
2586 if (SvPOK(retval))
2587
2588 /* If caller wants to handle missing properties, let them */
2589 if (return_if_undef) {
2590 return NULL;
2591 }
2592 Perl_croak(aTHX_
2593 "Can't find Unicode property definition \"%"SVf"\"",
2594 SVfARG(retval));
2595 Perl_croak(aTHX_ "SWASHNEW didn't return an HV ref");
2596 }
9a53f6cf 2597 } /* End of calling the module to find the swash */
36eb48b4
KW
2598
2599 /* Make sure there is an inversion list for binary properties */
2600 if (minbits == 1) {
2601 SV** swash_invlistsvp = NULL;
2602 SV* swash_invlist = NULL;
9a53f6cf 2603 bool invlist_in_swash_is_valid = FALSE;
88d45d28 2604 HV* swash_hv = NULL;
36eb48b4 2605
9a53f6cf
KW
2606 /* If this operation fetched a swash, get its already existing
2607 * inversion list or create one for it */
2608 if (retval != &PL_sv_undef) {
36eb48b4
KW
2609 swash_hv = MUTABLE_HV(SvRV(retval));
2610
2611 swash_invlistsvp = hv_fetchs(swash_hv, "INVLIST", FALSE);
9a53f6cf
KW
2612 if (swash_invlistsvp) {
2613 swash_invlist = *swash_invlistsvp;
2614 invlist_in_swash_is_valid = TRUE;
2615 }
2616 else {
36eb48b4 2617 swash_invlist = _swash_to_invlist(retval);
9a53f6cf
KW
2618 }
2619 }
2620
2621 /* If an inversion list was passed in, have to include it */
2622 if (invlist) {
2623
2624 /* Any fetched swash will by now have an inversion list in it;
2625 * otherwise <swash_invlist> will be NULL, indicating that we
2626 * didn't fetch a swash */
2627 if (swash_invlist) {
2628
2629 /* Add the passed-in inversion list, which invalidates the one
2630 * already stored in the swash */
2631 invlist_in_swash_is_valid = FALSE;
2632 _invlist_union(invlist, swash_invlist, &swash_invlist);
2633 }
2634 else {
2635
2636 /* Here, there is no swash already. Set up a minimal one */
2637 swash_hv = newHV();
2638 retval = newRV_inc(MUTABLE_SV(swash_hv));
2639 swash_invlist = invlist;
2640 }
2641
2642 if (passed_in_invlist_has_user_defined_property) {
2643 if (! hv_stores(swash_hv, "USER_DEFINED", newSVuv(1))) {
2644 Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed");
2645 }
2646 }
2647 }
2648
2649 /* Here, we have computed the union of all the passed-in data. It may
2650 * be that there was an inversion list in the swash which didn't get
2651 * touched; otherwise save the one computed one */
2652 if (! invlist_in_swash_is_valid) {
69794297
KW
2653 if (! hv_stores(MUTABLE_HV(SvRV(retval)), "INVLIST", swash_invlist))
2654 {
2655 Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed");
2656 }
9a53f6cf 2657 }
36eb48b4
KW
2658 }
2659
a0ed51b3
LW
2660 return retval;
2661}
2662
035d37be
JH
2663
2664/* This API is wrong for special case conversions since we may need to
2665 * return several Unicode characters for a single Unicode character
2666 * (see lib/unicore/SpecCase.txt) The SWASHGET in lib/utf8_heavy.pl is
2667 * the lower-level routine, and it is similarly broken for returning
38684baa
KW
2668 * multiple values. --jhi
2669 * For those, you should use to_utf8_case() instead */
b0e3252e 2670/* Now SWASHGET is recasted into S_swatch_get in this file. */
680c470c
ST
2671
2672/* Note:
2673 * Returns the value of property/mapping C<swash> for the first character
2674 * of the string C<ptr>. If C<do_utf8> is true, the string C<ptr> is
2675 * assumed to be in utf8. If C<do_utf8> is false, the string C<ptr> is
2676 * assumed to be in native 8-bit encoding. Caches the swatch in C<swash>.
af2af982
KW
2677 *
2678 * A "swash" is a hash which contains initially the keys/values set up by
2679 * SWASHNEW. The purpose is to be able to completely represent a Unicode
2680 * property for all possible code points. Things are stored in a compact form
2681 * (see utf8_heavy.pl) so that calculation is required to find the actual
2682 * property value for a given code point. As code points are looked up, new
2683 * key/value pairs are added to the hash, so that the calculation doesn't have
2684 * to ever be re-done. Further, each calculation is done, not just for the
2685 * desired one, but for a whole block of code points adjacent to that one.
2686 * For binary properties on ASCII machines, the block is usually for 64 code
2687 * points, starting with a code point evenly divisible by 64. Thus if the
2688 * property value for code point 257 is requested, the code goes out and
2689 * calculates the property values for all 64 code points between 256 and 319,
2690 * and stores these as a single 64-bit long bit vector, called a "swatch",
2691 * under the key for code point 256. The key is the UTF-8 encoding for code
2692 * point 256, minus the final byte. Thus, if the length of the UTF-8 encoding
2693 * for a code point is 13 bytes, the key will be 12 bytes long. If the value
2694 * for code point 258 is then requested, this code realizes that it would be
2695 * stored under the key for 256, and would find that value and extract the
2696 * relevant bit, offset from 256.
2697 *
2698 * Non-binary properties are stored in as many bits as necessary to represent
2699 * their values (32 currently, though the code is more general than that), not
2700 * as single bits, but the principal is the same: the value for each key is a
2701 * vector that encompasses the property values for all code points whose UTF-8
2702 * representations are represented by the key. That is, for all code points
2703 * whose UTF-8 representations are length N bytes, and the key is the first N-1
2704 * bytes of that.
680c470c 2705 */
a0ed51b3 2706UV
680c470c 2707Perl_swash_fetch(pTHX_ SV *swash, const U8 *ptr, bool do_utf8)
a0ed51b3 2708{
27da23d5 2709 dVAR;
ef8f7699 2710 HV *const hv = MUTABLE_HV(SvRV(swash));
3568d838
JH
2711 U32 klen;
2712 U32 off;
a0ed51b3 2713 STRLEN slen;
7d85a32c 2714 STRLEN needents;
cfd0369c 2715 const U8 *tmps = NULL;
a0ed51b3 2716 U32 bit;
979f2922 2717 SV *swatch;
3568d838 2718 U8 tmputf8[2];
35da51f7 2719 const UV c = NATIVE_TO_ASCII(*ptr);
3568d838 2720
7918f24d
NC
2721 PERL_ARGS_ASSERT_SWASH_FETCH;
2722
dbe7a391 2723 /* Convert to utf8 if not already */
3568d838 2724 if (!do_utf8 && !UNI_IS_INVARIANT(c)) {
979f2922
ST
2725 tmputf8[0] = (U8)UTF8_EIGHT_BIT_HI(c);
2726 tmputf8[1] = (U8)UTF8_EIGHT_BIT_LO(c);
2727 ptr = tmputf8;
3568d838
JH
2728 }
2729 /* Given a UTF-X encoded char 0xAA..0xYY,0xZZ
37e2e78e 2730 * then the "swatch" is a vec() for all the chars which start
3568d838
JH
2731 * with 0xAA..0xYY
2732 * So the key in the hash (klen) is length of encoded char -1
2733 */
2734 klen = UTF8SKIP(ptr) - 1;
2735 off = ptr[klen];
a0ed51b3 2736
979f2922 2737 if (klen == 0) {
37e2e78e 2738 /* If char is invariant then swatch is for all the invariant chars
1e54db1a 2739 * In both UTF-8 and UTF-8-MOD that happens to be UTF_CONTINUATION_MARK
7d85a32c 2740 */
979f2922
ST
2741 needents = UTF_CONTINUATION_MARK;
2742 off = NATIVE_TO_UTF(ptr[klen]);
2743 }
2744 else {
7d85a32c 2745 /* If char is encoded then swatch is for the prefix */
979f2922
ST
2746 needents = (1 << UTF_ACCUMULATION_SHIFT);
2747 off = NATIVE_TO_UTF(ptr[klen]) & UTF_CONTINUATION_MASK;
2748 }
7d85a32c 2749
a0ed51b3
LW
2750 /*
2751 * This single-entry cache saves about 1/3 of the utf8 overhead in test
2752 * suite. (That is, only 7-8% overall over just a hash cache. Still,
2753 * it's nothing to sniff at.) Pity we usually come through at least
2754 * two function calls to get here...
2755 *
2756 * NB: this code assumes that swatches are never modified, once generated!
2757 */
2758
3568d838 2759 if (hv == PL_last_swash_hv &&
a0ed51b3 2760 klen == PL_last_swash_klen &&
27da23d5 2761 (!klen || memEQ((char *)ptr, (char *)PL_last_swash_key, klen)) )
a0ed51b3
LW
2762 {
2763 tmps = PL_last_swash_tmps;
2764 slen = PL_last_swash_slen;
2765 }
2766 else {
2767 /* Try our second-level swatch cache, kept in a hash. */
e1ec3a88 2768 SV** svp = hv_fetch(hv, (const char*)ptr, klen, FALSE);
a0ed51b3 2769
b0e3252e 2770 /* If not cached, generate it via swatch_get */
979f2922
ST
2771 if (!svp || !SvPOK(*svp)
2772 || !(tmps = (const U8*)SvPV_const(*svp, slen))) {
2b9d42f0
NIS
2773 /* We use utf8n_to_uvuni() as we want an index into
2774 Unicode tables, not a native character number.
2775 */
aec46f14 2776 const UV code_point = utf8n_to_uvuni(ptr, UTF8_MAXBYTES, 0,
872c91ae
JH
2777 ckWARN(WARN_UTF8) ?
2778 0 : UTF8_ALLOW_ANY);
b0e3252e 2779 swatch = swatch_get(swash,
979f2922 2780 /* On EBCDIC & ~(0xA0-1) isn't a useful thing to do */
361ee0fe 2781 (klen) ? (code_point & ~((UV)needents - 1)) : 0,
979f2922
ST
2782 needents);
2783
923e4eb5 2784 if (IN_PERL_COMPILETIME)
623e6609 2785 CopHINTS_set(PL_curcop, PL_hints);
a0ed51b3 2786
979f2922 2787 svp = hv_store(hv, (const char *)ptr, klen, swatch, 0);
a0ed51b3 2788
979f2922
ST
2789 if (!svp || !(tmps = (U8*)SvPV(*svp, slen))
2790 || (slen << 3) < needents)
5637ef5b
NC
2791 Perl_croak(aTHX_ "panic: swash_fetch got improper swatch, "
2792 "svp=%p, tmps=%p, slen=%"UVuf", needents=%"UVuf,
2793 svp, tmps, (UV)slen, (UV)needents);
a0ed51b3
LW
2794 }
2795
2796 PL_last_swash_hv = hv;
16d8f38a 2797 assert(klen <= sizeof(PL_last_swash_key));
eac04b2e 2798 PL_last_swash_klen = (U8)klen;
cfd0369c
NC
2799 /* FIXME change interpvar.h? */
2800 PL_last_swash_tmps = (U8 *) tmps;
a0ed51b3
LW
2801 PL_last_swash_slen = slen;
2802 if (klen)
2803 Copy(ptr, PL_last_swash_key, klen, U8);
2804 }
2805
a410ec23
KW
2806 if (UTF8_IS_SUPER(ptr) && ckWARN_d(WARN_NON_UNICODE)) {
2807 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
2808
2809 /* This outputs warnings for binary properties only, assuming that
2810 * to_utf8_case() will output any for non-binary. Also, surrogates
2811 * aren't checked for, as that would warn on things like /\p{Gc=Cs}/ */
2812
9a53f6cf 2813 if (! bitssvp || SvUV(*bitssvp) == 1) {
a410ec23
KW
2814 /* User-defined properties can silently match above-Unicode */
2815 SV** const user_defined_svp = hv_fetchs(hv, "USER_DEFINED", FALSE);
2816 if (! user_defined_svp || ! SvUV(*user_defined_svp)) {
2817 const UV code_point = utf8n_to_uvuni(ptr, UTF8_MAXBYTES, 0, 0);
2818 Perl_warner(aTHX_ packWARN(WARN_NON_UNICODE),
2819 "Code point 0x%04"UVXf" is not Unicode, all \\p{} matches fail; all \\P{} matches succeed", code_point);
2820 }
2821 }
2822 }
2823
9faf8d75 2824 switch ((int)((slen << 3) / needents)) {
a0ed51b3
LW
2825 case 1:
2826 bit = 1 << (off & 7);
2827 off >>= 3;
2828 return (tmps[off] & bit) != 0;
2829 case 8:
2830 return tmps[off];
2831 case 16:
2832 off <<= 1;
2833 return (tmps[off] << 8) + tmps[off + 1] ;
2834 case 32:
2835 off <<= 2;
2836 return (tmps[off] << 24) + (tmps[off+1] << 16) + (tmps[off+2] << 8) + tmps[off + 3] ;
2837 }
5637ef5b
NC
2838 Perl_croak(aTHX_ "panic: swash_fetch got swatch of unexpected bit width, "
2839 "slen=%"UVuf", needents=%"UVuf, (UV)slen, (UV)needents);
670f1322 2840 NORETURN_FUNCTION_END;
a0ed51b3 2841}
2b9d42f0 2842
319009ee
KW
2843/* Read a single line of the main body of the swash input text. These are of
2844 * the form:
2845 * 0053 0056 0073
2846 * where each number is hex. The first two numbers form the minimum and
2847 * maximum of a range, and the third is the value associated with the range.
2848 * Not all swashes should have a third number
2849 *
2850 * On input: l points to the beginning of the line to be examined; it points
2851 * to somewhere in the string of the whole input text, and is
2852 * terminated by a \n or the null string terminator.
2853 * lend points to the null terminator of that string
2854 * wants_value is non-zero if the swash expects a third number
2855 * typestr is the name of the swash's mapping, like 'ToLower'
2856 * On output: *min, *max, and *val are set to the values read from the line.
2857 * returns a pointer just beyond the line examined. If there was no
2858 * valid min number on the line, returns lend+1
2859 */
2860
2861STATIC U8*
2862S_swash_scan_list_line(pTHX_ U8* l, U8* const lend, UV* min, UV* max, UV* val,
2863 const bool wants_value, const U8* const typestr)
2864{
2865 const int typeto = typestr[0] == 'T' && typestr[1] == 'o';
2866 STRLEN numlen; /* Length of the number */
02470786
KW
2867 I32 flags = PERL_SCAN_SILENT_ILLDIGIT
2868 | PERL_SCAN_DISALLOW_PREFIX
2869 | PERL_SCAN_SILENT_NON_PORTABLE;
319009ee
KW
2870
2871 /* nl points to the next \n in the scan */
2872 U8* const nl = (U8*)memchr(l, '\n', lend - l);
2873
2874 /* Get the first number on the line: the range minimum */
2875 numlen = lend - l;
2876 *min = grok_hex((char *)l, &numlen, &flags, NULL);
2877 if (numlen) /* If found a hex number, position past it */
2878 l += numlen;
2879 else if (nl) { /* Else, go handle next line, if any */
2880 return nl + 1; /* 1 is length of "\n" */
2881 }
2882 else { /* Else, no next line */
2883 return lend + 1; /* to LIST's end at which \n is not found */
2884 }
2885
2886 /* The max range value follows, separated by a BLANK */
2887 if (isBLANK(*l)) {
2888 ++l;
02470786
KW
2889 flags = PERL_SCAN_SILENT_ILLDIGIT
2890 | PERL_SCAN_DISALLOW_PREFIX
2891 | PERL_SCAN_SILENT_NON_PORTABLE;
319009ee
KW
2892 numlen = lend - l;
2893 *max = grok_hex((char *)l, &numlen, &flags, NULL);
2894 if (numlen)
2895 l += numlen;
2896 else /* If no value here, it is a single element range */
2897 *max = *min;
2898
2899 /* Non-binary tables have a third entry: what the first element of the
2900 * range maps to */
2901 if (wants_value) {
2902 if (isBLANK(*l)) {
2903 ++l;
f90a9a02
KW
2904
2905 /* The ToLc, etc table mappings are not in hex, and must be
2906 * corrected by adding the code point to them */
2907 if (typeto) {
2908 char *after_strtol = (char *) lend;
2909 *val = Strtol((char *)l, &after_strtol, 10);
2910 l = (U8 *) after_strtol;
f90a9a02
KW
2911 }
2912 else { /* Other tables are in hex, and are the correct result
2913 without tweaking */
a9d188b3
KW
2914 flags = PERL_SCAN_SILENT_ILLDIGIT
2915 | PERL_SCAN_DISALLOW_PREFIX
2916 | PERL_SCAN_SILENT_NON_PORTABLE;
2917 numlen = lend - l;
2918 *val = grok_hex((char *)l, &numlen, &flags, NULL);
2919 if (numlen)
2920 l += numlen;
2921 else
2922 *val = 0;
f90a9a02 2923 }
319009ee
KW
2924 }
2925 else {
2926 *val = 0;
2927 if (typeto) {
dcbac5bb 2928 /* diag_listed_as: To%s: illegal mapping '%s' */
319009ee
KW
2929 Perl_croak(aTHX_ "%s: illegal mapping '%s'",
2930 typestr, l);
2931 }
2932 }
2933 }
2934 else
2935 *val = 0; /* bits == 1, then any val should be ignored */
2936 }
2937 else { /* Nothing following range min, should be single element with no
2938 mapping expected */
2939 *max = *min;
2940 if (wants_value) {
2941 *val = 0;
2942 if (typeto) {
dcbac5bb 2943 /* diag_listed_as: To%s: illegal mapping '%s' */
319009ee
KW
2944 Perl_croak(aTHX_ "%s: illegal mapping '%s'", typestr, l);
2945 }
2946 }
2947 else
2948 *val = 0; /* bits == 1, then val should be ignored */
2949 }
2950
2951 /* Position to next line if any, or EOF */
2952 if (nl)
2953 l = nl + 1;
2954 else
2955 l = lend;
2956
2957 return l;
2958}
2959
979f2922
ST
2960/* Note:
2961 * Returns a swatch (a bit vector string) for a code point sequence
2962 * that starts from the value C<start> and comprises the number C<span>.
2963 * A C<swash> must be an object created by SWASHNEW (see lib/utf8_heavy.pl).
2964 * Should be used via swash_fetch, which will cache the swatch in C<swash>.
2965 */
2966STATIC SV*
b0e3252e 2967S_swatch_get(pTHX_ SV* swash, UV start, UV span)
979f2922
ST
2968{
2969 SV *swatch;
77f9f126 2970 U8 *l, *lend, *x, *xend, *s, *send;
979f2922 2971 STRLEN lcur, xcur, scur;
ef8f7699 2972 HV *const hv = MUTABLE_HV(SvRV(swash));
36eb48b4
KW
2973 SV** const invlistsvp = hv_fetchs(hv, "INVLIST", FALSE);
2974
88d45d28
KW
2975 SV** listsvp = NULL; /* The string containing the main body of the table */
2976 SV** extssvp = NULL;
2977 SV** invert_it_svp = NULL;
2978 U8* typestr = NULL;
786861f5
KW
2979 STRLEN bits;
2980 STRLEN octets; /* if bits == 1, then octets == 0 */
2981 UV none;
2982 UV end = start + span;
972dd592 2983
36eb48b4 2984 if (invlistsvp == NULL) {
786861f5
KW
2985 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
2986 SV** const nonesvp = hv_fetchs(hv, "NONE", FALSE);
2987 SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE);
2988 extssvp = hv_fetchs(hv, "EXTRAS", FALSE);
2989 listsvp = hv_fetchs(hv, "LIST", FALSE);
2990 invert_it_svp = hv_fetchs(hv, "INVERT_IT", FALSE);
2991
2992 bits = SvUV(*bitssvp);
2993 none = SvUV(*nonesvp);
2994 typestr = (U8*)SvPV_nolen(*typesvp);
2995 }
36eb48b4
KW
2996 else {
2997 bits = 1;
2998 none = 0;
2999 }
786861f5 3000 octets = bits >> 3; /* if bits == 1, then octets == 0 */
979f2922 3001
b0e3252e 3002 PERL_ARGS_ASSERT_SWATCH_GET;
7918f24d 3003
979f2922 3004 if (bits != 1 && bits != 8 && bits != 16 && bits != 32) {
b0e3252e 3005 Perl_croak(aTHX_ "panic: swatch_get doesn't expect bits %"UVuf,
660a4616 3006 (UV)bits);
979f2922
ST
3007 }
3008
84ea5ef6
KW
3009 /* If overflowed, use the max possible */
3010 if (end < start) {
3011 end = UV_MAX;
3012 span = end - start;
3013 }
3014
979f2922 3015 /* create and initialize $swatch */
979f2922 3016 scur = octets ? (span * octets) : (span + 7) / 8;
e524fe40
NC
3017 swatch = newSV(scur);
3018 SvPOK_on(swatch);
979f2922
ST
3019 s = (U8*)SvPVX(swatch);
3020 if (octets && none) {
0bd48802 3021 const U8* const e = s + scur;
979f2922
ST
3022 while (s < e) {
3023 if (bits == 8)
3024 *s++ = (U8)(none & 0xff);
3025 else if (bits == 16) {
3026 *s++ = (U8)((none >> 8) & 0xff);
3027 *s++ = (U8)( none & 0xff);
3028 }
3029 else if (bits == 32) {
3030 *s++ = (U8)((none >> 24) & 0xff);
3031 *s++ = (U8)((none >> 16) & 0xff);
3032 *s++ = (U8)((none >> 8) & 0xff);
3033 *s++ = (U8)( none & 0xff);
3034 }
3035 }
3036 *s = '\0';
3037 }
3038 else {
3039 (void)memzero((U8*)s, scur + 1);
3040 }
3041 SvCUR_set(swatch, scur);
3042 s = (U8*)SvPVX(swatch);
3043
36eb48b4
KW
3044 if (invlistsvp) { /* If has an inversion list set up use that */
3045 _invlist_populate_swatch(*invlistsvp, start, end, s);
3046 return swatch;
3047 }
3048
3049 /* read $swash->{LIST} */
979f2922
ST
3050 l = (U8*)SvPV(*listsvp, lcur);
3051 lend = l + lcur;
3052 while (l < lend) {
8ed25d53 3053 UV min, max, val, upper;
319009ee
KW
3054 l = S_swash_scan_list_line(aTHX_ l, lend, &min, &max, &val,
3055 cBOOL(octets), typestr);
3056 if (l > lend) {
979f2922
ST
3057 break;
3058 }
3059
972dd592 3060 /* If looking for something beyond this range, go try the next one */
979f2922
ST
3061 if (max < start)
3062 continue;
3063
8ed25d53
KW
3064 /* <end> is generally 1 beyond where we want to set things, but at the
3065 * platform's infinity, where we can't go any higher, we want to
3066 * include the code point at <end> */
3067 upper = (max < end)
3068 ? max
3069 : (max != UV_MAX || end != UV_MAX)
3070 ? end - 1
3071 : end;
3072
979f2922 3073 if (octets) {
35da51f7 3074 UV key;
979f2922
ST
3075 if (min < start) {
3076 if (!none || val < none) {
3077 val += start - min;
3078 }
3079 min = start;
3080 }
8ed25d53 3081 for (key = min; key <= upper; key++) {
979f2922 3082 STRLEN offset;
979f2922
ST
3083 /* offset must be non-negative (start <= min <= key < end) */
3084 offset = octets * (key - start);
3085 if (bits == 8)
3086 s[offset] = (U8)(val & 0xff);
3087 else if (bits == 16) {
3088 s[offset ] = (U8)((val >> 8) & 0xff);
3089 s[offset + 1] = (U8)( val & 0xff);
3090 }
3091 else if (bits == 32) {
3092 s[offset ] = (U8)((val >> 24) & 0xff);
3093 s[offset + 1] = (U8)((val >> 16) & 0xff);
3094 s[offset + 2] = (U8)((val >> 8) & 0xff);
3095 s[offset + 3] = (U8)( val & 0xff);
3096 }
3097
3098 if (!none || val < none)
3099 ++val;
3100 }
3101 }
711a919c 3102 else { /* bits == 1, then val should be ignored */
35da51f7 3103 UV key;
979f2922
ST
3104 if (min < start)
3105 min = start;
6cb05c12 3106
8ed25d53 3107 for (key = min; key <= upper; key++) {
0bd48802 3108 const STRLEN offset = (STRLEN)(key - start);
979f2922
ST
3109 s[offset >> 3] |= 1 << (offset & 7);
3110 }
3111 }
3112 } /* while */
979f2922 3113
9479a769 3114 /* Invert if the data says it should be. Assumes that bits == 1 */
77f9f126 3115 if (invert_it_svp && SvUV(*invert_it_svp)) {
0bda3001
KW
3116
3117 /* Unicode properties should come with all bits above PERL_UNICODE_MAX
3118 * be 0, and their inversion should also be 0, as we don't succeed any
3119 * Unicode property matches for non-Unicode code points */
3120 if (start <= PERL_UNICODE_MAX) {
3121
3122 /* The code below assumes that we never cross the
3123 * Unicode/above-Unicode boundary in a range, as otherwise we would
3124 * have to figure out where to stop flipping the bits. Since this
3125 * boundary is divisible by a large power of 2, and swatches comes
3126 * in small powers of 2, this should be a valid assumption */
3127 assert(start + span - 1 <= PERL_UNICODE_MAX);
3128
507a8485
KW
3129 send = s + scur;
3130 while (s < send) {
3131 *s = ~(*s);
3132 s++;
3133 }
0bda3001 3134 }
77f9f126
KW
3135 }
3136
d73c39c5
KW
3137 /* read $swash->{EXTRAS}
3138 * This code also copied to swash_to_invlist() below */
979f2922
ST
3139 x = (U8*)SvPV(*extssvp, xcur);
3140 xend = x + xcur;
3141 while (x < xend) {
3142 STRLEN namelen;
3143 U8 *namestr;
3144 SV** othersvp;
3145 HV* otherhv;
3146 STRLEN otherbits;
3147 SV **otherbitssvp, *other;
711a919c 3148 U8 *s, *o, *nl;
979f2922
ST
3149 STRLEN slen, olen;
3150
35da51f7 3151 const U8 opc = *x++;
979f2922
ST
3152 if (opc == '\n')
3153 continue;
3154
3155 nl = (U8*)memchr(x, '\n', xend - x);
3156
3157 if (opc != '-' && opc != '+' && opc != '!' && opc != '&') {
3158 if (nl) {
3159 x = nl + 1; /* 1 is length of "\n" */
3160 continue;
3161 }
3162 else {
3163 x = xend; /* to EXTRAS' end at which \n is not found */
3164 break;
3165 }
3166 }
3167
3168 namestr = x;
3169 if (nl) {
3170 namelen = nl - namestr;
3171 x = nl + 1;
3172 }
3173 else {
3174 namelen = xend - namestr;
3175 x = xend;
3176 }
3177
3178 othersvp = hv_fetch(hv, (char *)namestr, namelen, FALSE);
ef8f7699 3179 otherhv = MUTABLE_HV(SvRV(*othersvp));
017a3ce5 3180 otherbitssvp = hv_fetchs(otherhv, "BITS", FALSE);
979f2922
ST
3181 otherbits = (STRLEN)SvUV(*otherbitssvp);
3182 if (bits < otherbits)
5637ef5b
NC
3183 Perl_croak(aTHX_ "panic: swatch_get found swatch size mismatch, "
3184 "bits=%"UVuf", otherbits=%"UVuf, (UV)bits, (UV)otherbits);
979f2922
ST
3185
3186 /* The "other" swatch must be destroyed after. */
b0e3252e 3187 other = swatch_get(*othersvp, start, span);
979f2922
ST
3188 o = (U8*)SvPV(other, olen);
3189
3190 if (!olen)
b0e3252e 3191 Perl_croak(aTHX_ "panic: swatch_get got improper swatch");
979f2922
ST
3192
3193 s = (U8*)SvPV(swatch, slen);
3194 if (bits == 1 && otherbits == 1) {
3195 if (slen != olen)
5637ef5b
NC
3196 Perl_croak(aTHX_ "panic: swatch_get found swatch length "
3197 "mismatch, slen=%"UVuf", olen=%"UVuf,
3198 (UV)slen, (UV)olen);
979f2922
ST
3199
3200 switch (opc) {
3201 case '+':
3202 while (slen--)
3203 *s++ |= *o++;
3204 break;
3205 case '!':
3206 while (slen--)
3207 *s++ |= ~*o++;
3208 break;
3209 case '-':
3210 while (slen--)
3211 *s++ &= ~*o++;
3212 break;
3213 case '&':
3214 while (slen--)
3215 *s++ &= *o++;
3216 break;
3217 default:
3218 break;
3219 }
3220 }
711a919c 3221 else {
979f2922
ST
3222 STRLEN otheroctets = otherbits >> 3;
3223 STRLEN offset = 0;
35da51f7 3224 U8* const send = s + slen;
979f2922
ST
3225
3226 while (s < send) {
3227 UV otherval = 0;
3228
3229 if (otherbits == 1) {
3230 otherval = (o[offset >> 3] >> (offset & 7)) & 1;
3231 ++offset;
3232 }
3233 else {
3234 STRLEN vlen = otheroctets;
3235 otherval = *o++;
3236 while (--vlen) {
3237 otherval <<= 8;
3238 otherval |= *o++;
3239 }
3240 }
3241
711a919c 3242 if (opc == '+' && otherval)
6f207bd3 3243 NOOP; /* replace with otherval */
979f2922
ST
3244 else if (opc == '!' && !otherval)
3245 otherval = 1;
3246 else if (opc == '-' && otherval)
3247 otherval = 0;
3248 else if (opc == '&' && !otherval)
3249 otherval = 0;
3250 else {
711a919c 3251 s += octets; /* no replacement */
979f2922
ST
3252 continue;
3253 }
3254
3255 if (bits == 8)
3256 *s++ = (U8)( otherval & 0xff);
3257 else if (bits == 16) {
3258 *s++ = (U8)((otherval >> 8) & 0xff);
3259 *s++ = (U8)( otherval & 0xff);
3260 }
3261 else if (bits == 32) {
3262 *s++ = (U8)((otherval >> 24) & 0xff);
3263 *s++ = (U8)((otherval >> 16) & 0xff);
3264 *s++ = (U8)((otherval >> 8) & 0xff);
3265 *s++ = (U8)( otherval & 0xff);
3266 }
3267 }
3268 }
3269 sv_free(other); /* through with it! */
3270 } /* while */
3271 return swatch;
3272}
3273
064c021d 3274HV*
4c2e1131 3275Perl__swash_inversion_hash(pTHX_ SV* const swash)
064c021d
KW
3276{
3277
5662e334
KW
3278 /* Subject to change or removal. For use only in one place in regcomp.c.
3279 * Can't be used on a property that is subject to user override, as it
3280 * relies on the value of SPECIALS in the swash which would be set by
3281 * utf8_heavy.pl to the hash in the non-overriden file, and hence is not set
3282 * for overridden properties
064c021d
KW
3283 *
3284 * Returns a hash which is the inversion and closure of a swash mapping.
3285 * For example, consider the input lines:
3286 * 004B 006B
3287 * 004C 006C
3288 * 212A 006B
3289 *
3290 * The returned hash would have two keys, the utf8 for 006B and the utf8 for
3291 * 006C. The value for each key is an array. For 006C, the array would
3292 * have a two elements, the utf8 for itself, and for 004C. For 006B, there
3293 * would be three elements in its array, the utf8 for 006B, 004B and 212A.
3294 *
3295 * Essentially, for any code point, it gives all the code points that map to
3296 * it, or the list of 'froms' for that point.
3297 *
5662e334
KW
3298 * Currently it ignores any additions or deletions from other swashes,
3299 * looking at just the main body of the swash, and if there are SPECIALS
3300 * in the swash, at that hash
3301 *
3302 * The specials hash can be extra code points, and most likely consists of
3303 * maps from single code points to multiple ones (each expressed as a string
3304 * of utf8 characters). This function currently returns only 1-1 mappings.
3305 * However consider this possible input in the specials hash:
3306 * "\xEF\xAC\x85" => "\x{0073}\x{0074}", # U+FB05 => 0073 0074
3307 * "\xEF\xAC\x86" => "\x{0073}\x{0074}", # U+FB06 => 0073 0074
3308 *
3309 * Both FB05 and FB06 map to the same multi-char sequence, which we don't
3310 * currently handle. But it also means that FB05 and FB06 are equivalent in
3311 * a 1-1 mapping which we should handle, and this relationship may not be in
3312 * the main table. Therefore this function examines all the multi-char
3313 * sequences and adds the 1-1 mappings that come out of that. */
064c021d
KW
3314
3315 U8 *l, *lend;
3316 STRLEN lcur;
3317 HV *const hv = MUTABLE_HV(SvRV(swash));
3318
3319 /* The string containing the main body of the table */
3320 SV** const listsvp = hv_fetchs(hv, "LIST", FALSE);
3321
3322 SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE);
3323 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
3324 SV** const nonesvp = hv_fetchs(hv, "NONE", FALSE);
3325 /*SV** const extssvp = hv_fetchs(hv, "EXTRAS", FALSE);*/
3326 const U8* const typestr = (U8*)SvPV_nolen(*typesvp);
3327 const STRLEN bits = SvUV(*bitssvp);
3328 const STRLEN octets = bits >> 3; /* if bits == 1, then octets == 0 */
3329 const UV none = SvUV(*nonesvp);
5662e334 3330 SV **specials_p = hv_fetchs(hv, "SPECIALS", 0);
064c021d
KW
3331
3332 HV* ret = newHV();
3333
3334 PERL_ARGS_ASSERT__SWASH_INVERSION_HASH;
3335
3336 /* Must have at least 8 bits to get the mappings */
3337 if (bits != 8 && bits != 16 && bits != 32) {
3338 Perl_croak(aTHX_ "panic: swash_inversion_hash doesn't expect bits %"UVuf,
3339 (UV)bits);
3340 }
3341
5662e334
KW
3342 if (specials_p) { /* It might be "special" (sometimes, but not always, a
3343 mapping to more than one character */
3344
3345 /* Construct an inverse mapping hash for the specials */
3346 HV * const specials_hv = MUTABLE_HV(SvRV(*specials_p));
3347 HV * specials_inverse = newHV();
3348 char *char_from; /* the lhs of the map */
3349 I32 from_len; /* its byte length */
3350 char *char_to; /* the rhs of the map */
3351 I32 to_len; /* its byte length */
3352 SV *sv_to; /* and in a sv */
3353 AV* from_list; /* list of things that map to each 'to' */
3354
3355 hv_iterinit(specials_hv);
3356
3357 /* The keys are the characters (in utf8) that map to the corresponding
3358 * utf8 string value. Iterate through the list creating the inverse
3359 * list. */
3360 while ((sv_to = hv_iternextsv(specials_hv, &char_from, &from_len))) {
3361 SV** listp;
3362 if (! SvPOK(sv_to)) {
5637ef5b
NC
3363 Perl_croak(aTHX_ "panic: value returned from hv_iternextsv() "
3364 "unexpectedly is not a string, flags=%lu",
3365 (unsigned long)SvFLAGS(sv_to));
5662e334
KW
3366 }
3367 /*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)));*/
3368
3369 /* Each key in the inverse list is a mapped-to value, and the key's
3370 * hash value is a list of the strings (each in utf8) that map to
3371 * it. Those strings are all one character long */
3372 if ((listp = hv_fetch(specials_inverse,
3373 SvPVX(sv_to),
3374 SvCUR(sv_to), 0)))
3375 {
3376 from_list = (AV*) *listp;
3377 }
3378 else { /* No entry yet for it: create one */
3379 from_list = newAV();
3380 if (! hv_store(specials_inverse,
3381 SvPVX(sv_to),
3382 SvCUR(sv_to),
3383 (SV*) from_list, 0))
3384 {
3385 Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed");
3386 }
3387 }
3388
3389 /* Here have the list associated with this 'to' (perhaps newly
3390 * created and empty). Just add to it. Note that we ASSUME that
3391 * the input is guaranteed to not have duplications, so we don't
3392 * check for that. Duplications just slow down execution time. */
3393 av_push(from_list, newSVpvn_utf8(char_from, from_len, TRUE));
3394 }
3395
3396 /* Here, 'specials_inverse' contains the inverse mapping. Go through
3397 * it looking for cases like the FB05/FB06 examples above. There would
3398 * be an entry in the hash like
3399 * 'st' => [ FB05, FB06 ]
3400 * In this example we will create two lists that get stored in the
3401 * returned hash, 'ret':
3402 * FB05 => [ FB05, FB06 ]
3403 * FB06 => [ FB05, FB06 ]
3404 *
3405 * Note that there is nothing to do if the array only has one element.
3406 * (In the normal 1-1 case handled below, we don't have to worry about
3407 * two lists, as everything gets tied to the single list that is
3408 * generated for the single character 'to'. But here, we are omitting
3409 * that list, ('st' in the example), so must have multiple lists.) */
3410 while ((from_list = (AV *) hv_iternextsv(specials_inverse,
3411 &char_to, &to_len)))
3412 {
3413 if (av_len(from_list) > 0) {
3414 int i;
3415
3416 /* We iterate over all combinations of i,j to place each code
3417 * point on each list */
3418 for (i = 0; i <= av_len(from_list); i++) {
3419 int j;
3420 AV* i_list = newAV();
3421 SV** entryp = av_fetch(from_list, i, FALSE);
3422 if (entryp == NULL) {
3423 Perl_croak(aTHX_ "panic: av_fetch() unexpectedly failed");
3424 }
3425 if (hv_fetch(ret, SvPVX(*entryp), SvCUR(*entryp), FALSE)) {
3426 Perl_croak(aTHX_ "panic: unexpected entry for %s", SvPVX(*entryp));
3427 }
3428 if (! hv_store(ret, SvPVX(*entryp), SvCUR(*entryp),
3429 (SV*) i_list, FALSE))
3430 {
3431 Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed");
3432 }
3433
3434 /* For debugging: UV u = utf8_to_uvchr((U8*) SvPVX(*entryp), 0);*/
3435 for (j = 0; j <= av_len(from_list); j++) {
3436 entryp = av_fetch(from_list, j, FALSE);
3437 if (entryp == NULL) {
3438 Perl_croak(aTHX_ "panic: av_fetch() unexpectedly failed");
3439 }
3440
3441 /* When i==j this adds itself to the list */
3442 av_push(i_list, newSVuv(utf8_to_uvchr(
3443 (U8*) SvPVX(*entryp), 0)));
3444 /*DEBUG_U(PerlIO_printf(Perl_debug_log, "Adding %"UVXf" to list for %"UVXf"\n", utf8_to_uvchr((U8*) SvPVX(*entryp), 0), u));*/
3445 }
3446 }
3447 }
3448 }
3449 SvREFCNT_dec(specials_inverse); /* done with it */
3450 } /* End of specials */
3451
064c021d
KW
3452 /* read $swash->{LIST} */
3453 l = (U8*)SvPV(*listsvp, lcur);
3454 lend = l + lcur;
3455
3456 /* Go through each input line */
3457 while (l < lend) {
3458 UV min, max, val;
3459 UV inverse;
3460 l = S_swash_scan_list_line(aTHX_ l, lend, &min, &max, &val,
3461 cBOOL(octets), typestr);
3462 if (l > lend) {
3463 break;
3464 }
3465
3466 /* Each element in the range is to be inverted */
3467 for (inverse = min; inverse <= max; inverse++) {
3468 AV* list;
064c021d
KW
3469 SV** listp;
3470 IV i;
3471 bool found_key = FALSE;
5662e334 3472 bool found_inverse = FALSE;
064c021d
KW
3473
3474 /* The key is the inverse mapping */
3475 char key[UTF8_MAXBYTES+1];
3476 char* key_end = (char *) uvuni_to_utf8((U8*) key, val);
3477 STRLEN key_len = key_end - key;
3478
064c021d
KW
3479 /* Get the list for the map */
3480 if ((listp = hv_fetch(ret, key, key_len, FALSE))) {
3481 list = (AV*) *listp;
3482 }
3483 else { /* No entry yet for it: create one */
3484 list = newAV();
3485 if (! hv_store(ret, key, key_len, (SV*) list, FALSE)) {
3486 Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed");
3487 }
3488 }
3489
5662e334
KW
3490 /* Look through list to see if this inverse mapping already is
3491 * listed, or if there is a mapping to itself already */
508f7cfa 3492 for (i = 0; i <= av_len(list); i++) {
064c021d
KW
3493 SV** entryp = av_fetch(list, i, FALSE);
3494 SV* entry;
3495 if (entryp == NULL) {
3496 Perl_croak(aTHX_ "panic: av_fetch() unexpectedly failed");
3497 }
3498 entry = *entryp;
5662e334 3499 /*DEBUG_U(PerlIO_printf(Perl_debug_log, "list for %"UVXf" contains %"UVXf"\n", val, SvUV(entry)));*/
56ca34ca 3500 if (SvUV(entry) == val) {
064c021d 3501 found_key = TRUE;
5662e334
KW
3502 }
3503 if (SvUV(entry) == inverse) {
3504 found_inverse = TRUE;
3505 }
3506
3507 /* No need to continue searching if found everything we are
3508 * looking for */
3509 if (found_key && found_inverse) {
064c021d
KW
3510 break;
3511 }
3512 }
56ca34ca
KW
3513
3514 /* Make sure there is a mapping to itself on the list */
064c021d 3515 if (! found_key) {
d397ff6a 3516 av_push(list, newSVuv(val));
5662e334 3517 /*DEBUG_U(PerlIO_printf(Perl_debug_log, "Adding %"UVXf" to list for %"UVXf"\n", val, val));*/
064c021d
KW
3518 }
3519
3520
3521 /* Simply add the value to the list */
5662e334
KW
3522 if (! found_inverse) {
3523 av_push(list, newSVuv(inverse));
3524 /*DEBUG_U(PerlIO_printf(Perl_debug_log, "Adding %"UVXf" to list for %"UVXf"\n", inverse, val));*/
3525 }
064c021d 3526
b0e3252e 3527 /* swatch_get() increments the value of val for each element in the
064c021d
KW
3528 * range. That makes more compact tables possible. You can
3529 * express the capitalization, for example, of all consecutive
3530 * letters with a single line: 0061\t007A\t0041 This maps 0061 to
3531 * 0041, 0062 to 0042, etc. I (khw) have never understood 'none',
bd3f2f94 3532 * and it's not documented; it appears to be used only in
b0e3252e 3533 * implementing tr//; I copied the semantics from swatch_get(), just
bd3f2f94 3534 * in case */
064c021d
KW
3535 if (!none || val < none) {
3536 ++val;
3537 }
3538 }
3539 }
3540
3541 return ret;
3542}
3543
a25abddc 3544SV*
d764b54e
KW
3545Perl__swash_to_invlist(pTHX_ SV* const swash)
3546{
3547
3548 /* Subject to change or removal. For use only in one place in regcomp.c */
3549
3550 U8 *l, *lend;
3551 char *loc;
3552 STRLEN lcur;
3553 HV *const hv = MUTABLE_HV(SvRV(swash));
3554 UV elements = 0; /* Number of elements in the inversion list */
b443038a 3555 U8 empty[] = "";
d764b54e
KW
3556
3557 /* The string containing the main body of the table */
3558 SV** const listsvp = hv_fetchs(hv, "LIST", FALSE);
3559 SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE);
3560 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
d73c39c5 3561 SV** const extssvp = hv_fetchs(hv, "EXTRAS", FALSE);
77f9f126 3562 SV** const invert_it_svp = hv_fetchs(hv, "INVERT_IT", FALSE);
d764b54e
KW
3563
3564 const U8* const typestr = (U8*)SvPV_nolen(*typesvp);
3565 const STRLEN bits = SvUV(*bitssvp);
3566 const STRLEN octets = bits >> 3; /* if bits == 1, then octets == 0 */
d73c39c5
KW
3567 U8 *x, *xend;
3568 STRLEN xcur;
d764b54e 3569
a25abddc 3570 SV* invlist;
d764b54e
KW
3571
3572 PERL_ARGS_ASSERT__SWASH_TO_INVLIST;
3573
3574 /* read $swash->{LIST} */
b443038a
KW
3575 if (SvPOK(*listsvp)) {
3576 l = (U8*)SvPV(*listsvp, lcur);
3577 }
3578 else {
3579 /* LIST legitimately doesn't contain a string during compilation phases
3580 * of Perl itself, before the Unicode tables are generated. In this
3581 * case, just fake things up by creating an empty list */
3582 l = empty;
3583 lcur = 0;
3584 }
d764b54e
KW
3585 loc = (char *) l;
3586 lend = l + lcur;
3587
3588 /* Scan the input to count the number of lines to preallocate array size
3589 * based on worst possible case, which is each line in the input creates 2
3590 * elements in the inversion list: 1) the beginning of a range in the list;
3591 * 2) the beginning of a range not in the list. */
3592 while ((loc = (strchr(loc, '\n'))) != NULL) {
3593 elements += 2;
3594 loc++;
3595 }
3596
3597 /* If the ending is somehow corrupt and isn't a new line, add another
3598 * element for the final range that isn't in the inversion list */
fd05e003
KW
3599 if (! (*lend == '\n'
3600 || (*lend == '\0' && (lcur == 0 || *(lend - 1) == '\n'))))
3601 {
d764b54e
KW
3602 elements++;
3603 }
3604
3605 invlist = _new_invlist(elements);
3606
3607 /* Now go through the input again, adding each range to the list */
3608 while (l < lend) {
3609 UV start, end;
3610 UV val; /* Not used by this function */
3611
3612 l = S_swash_scan_list_line(aTHX_ l, lend, &start, &end, &val,
3613 cBOOL(octets), typestr);
3614
3615 if (l > lend) {
3616 break;
3617 }
3618
3619 _append_range_to_invlist(invlist, start, end);
3620 }
3621
77f9f126
KW
3622 /* Invert if the data says it should be */
3623 if (invert_it_svp && SvUV(*invert_it_svp)) {
45bb2768 3624 _invlist_invert_prop(invlist);
77f9f126
KW
3625 }
3626
b0e3252e 3627 /* This code is copied from swatch_get()
d73c39c5
KW
3628 * read $swash->{EXTRAS} */
3629 x = (U8*)SvPV(*extssvp, xcur);
3630 xend = x + xcur;
3631 while (x < xend) {
3632 STRLEN namelen;
3633 U8 *namestr;
3634 SV** othersvp;
3635 HV* otherhv;
3636 STRLEN otherbits;
3637 SV **otherbitssvp, *other;
3638 U8 *nl;
3639
3640 const U8 opc = *x++;
3641 if (opc == '\n')
3642 continue;
3643
3644 nl = (U8*)memchr(x, '\n', xend - x);
3645
3646 if (opc != '-' && opc != '+' && opc != '!' && opc != '&') {
3647 if (nl) {
3648 x = nl + 1; /* 1 is length of "\n" */
3649 continue;
3650 }
3651 else {
3652 x = xend; /* to EXTRAS' end at which \n is not found */
3653 break;
3654 }
3655 }
3656
3657 namestr = x;
3658 if (nl) {
3659 namelen = nl - namestr;
3660 x = nl + 1;
3661 }
3662 else {
3663 namelen = xend - namestr;
3664 x = xend;
3665 }
3666
3667 othersvp = hv_fetch(hv, (char *)namestr, namelen, FALSE);
3668 otherhv = MUTABLE_HV(SvRV(*othersvp));
3669 otherbitssvp = hv_fetchs(otherhv, "BITS", FALSE);
3670 otherbits = (STRLEN)SvUV(*otherbitssvp);
3671
3672 if (bits != otherbits || bits != 1) {
5637ef5b
NC
3673 Perl_croak(aTHX_ "panic: _swash_to_invlist only operates on boolean "
3674 "properties, bits=%"UVuf", otherbits=%"UVuf,
3675 (UV)bits, (UV)otherbits);
d73c39c5
KW
3676 }
3677
3678 /* The "other" swatch must be destroyed after. */
3679 other = _swash_to_invlist((SV *)*othersvp);
3680
b0e3252e 3681 /* End of code copied from swatch_get() */
d73c39c5
KW
3682 switch (opc) {
3683 case '+':
3684 _invlist_union(invlist, other, &invlist);
3685 break;
3686 case '!':
3687 _invlist_invert(other);
3688 _invlist_union(invlist, other, &invlist);
3689 break;
3690 case '-':
3691 _invlist_subtract(invlist, other, &invlist);
3692 break;
3693 case '&':
3694 _invlist_intersection(invlist, other, &invlist);
3695 break;
3696 default:
3697 break;
3698 }
3699 sv_free(other); /* through with it! */
3700 }
3701
d764b54e
KW
3702 return invlist;
3703}
3704
0f830e0b 3705/*
87cea99e 3706=for apidoc uvchr_to_utf8
0f830e0b 3707
6ee84de2 3708Adds the UTF-8 representation of the Native code point C<uv> to the end
e0aa61c6 3709of the string C<d>; C<d> should have at least C<UTF8_MAXBYTES+1> free
0f830e0b
NC
3710bytes available. The return value is the pointer to the byte after the
3711end of the new character. In other words,
3712
3713 d = uvchr_to_utf8(d, uv);
3714
3715is the recommended wide native character-aware way of saying
3716
3717 *(d++) = uv;
3718
3719=cut
3720*/
3721
3722/* On ASCII machines this is normally a macro but we want a
3723 real function in case XS code wants it
3724*/
0f830e0b
NC
3725U8 *
3726Perl_uvchr_to_utf8(pTHX_ U8 *d, UV uv)
3727{
7918f24d
NC
3728 PERL_ARGS_ASSERT_UVCHR_TO_UTF8;
3729
0f830e0b
NC
3730 return Perl_uvuni_to_utf8_flags(aTHX_ d, NATIVE_TO_UNI(uv), 0);
3731}
3732
b851fbc1
JH
3733U8 *
3734Perl_uvchr_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags)
3735{
7918f24d
NC
3736 PERL_ARGS_ASSERT_UVCHR_TO_UTF8_FLAGS;
3737
b851fbc1
JH
3738 return Perl_uvuni_to_utf8_flags(aTHX_ d, NATIVE_TO_UNI(uv), flags);
3739}
2b9d42f0
NIS
3740
3741/*
87cea99e 3742=for apidoc utf8n_to_uvchr
0f830e0b 3743
48ef279e 3744Returns the native character value of the first character in the string
0f830e0b
NC
3745C<s>
3746which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
3747length, in bytes, of that character.
3748
6ee84de2 3749length and flags are the same as utf8n_to_uvuni().
0f830e0b
NC
3750
3751=cut
3752*/
3753/* On ASCII machines this is normally a macro but we want
3754 a real function in case XS code wants it
3755*/
0f830e0b 3756UV
48ef279e 3757Perl_utf8n_to_uvchr(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen,
0f830e0b
NC
3758U32 flags)
3759{
3760 const UV uv = Perl_utf8n_to_uvuni(aTHX_ s, curlen, retlen, flags);
7918f24d
NC
3761
3762 PERL_ARGS_ASSERT_UTF8N_TO_UVCHR;
3763
0f830e0b
NC
3764 return UNI_TO_NATIVE(uv);
3765}
3766
0876b9a0
KW
3767bool
3768Perl_check_utf8_print(pTHX_ register const U8* s, const STRLEN len)
3769{
3770 /* May change: warns if surrogates, non-character code points, or
af2af982
KW
3771 * non-Unicode code points are in s which has length len bytes. Returns
3772 * TRUE if none found; FALSE otherwise. The only other validity check is
3773 * to make sure that this won't exceed the string's length */
0876b9a0
KW
3774
3775 const U8* const e = s + len;
3776 bool ok = TRUE;
3777
3778 PERL_ARGS_ASSERT_CHECK_UTF8_PRINT;
3779
3780 while (s < e) {
3781 if (UTF8SKIP(s) > len) {
3782 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8),
3783 "%s in %s", unees, PL_op ? OP_DESC(PL_op) : "print");
3784 return FALSE;
3785 }
732fbc05 3786 if (UNLIKELY(*s >= UTF8_FIRST_PROBLEMATIC_CODE_POINT_FIRST_BYTE)) {
0876b9a0
KW
3787 STRLEN char_len;
3788 if (UTF8_IS_SUPER(s)) {
8457b38f
KW
3789 if (ckWARN_d(WARN_NON_UNICODE)) {
3790 UV uv = utf8_to_uvchr(s, &char_len);
3791 Perl_warner(aTHX_ packWARN(WARN_NON_UNICODE),
3792 "Code point 0x%04"UVXf" is not Unicode, may not be portable", uv);
3793 ok = FALSE;
3794 }
0876b9a0
KW
3795 }
3796 else if (UTF8_IS_SURROGATE(s)) {
8457b38f
KW
3797 if (ckWARN_d(WARN_SURROGATE)) {
3798 UV uv = utf8_to_uvchr(s, &char_len);
3799 Perl_warner(aTHX_ packWARN(WARN_SURROGATE),
3800 "Unicode surrogate U+%04"UVXf" is illegal in UTF-8", uv);
3801 ok = FALSE;
3802 }
0876b9a0
KW
3803 }
3804 else if
8457b38f
KW
3805 ((UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s))
3806 && (ckWARN_d(WARN_NONCHAR)))
0876b9a0
KW
3807 {
3808 UV uv = utf8_to_uvchr(s, &char_len);
8457b38f 3809 Perl_warner(aTHX_ packWARN(WARN_NONCHAR),
0876b9a0
KW
3810 "Unicode non-character U+%04"UVXf" is illegal for open interchange", uv);
3811 ok = FALSE;
3812 }
3813 }
3814 s += UTF8SKIP(s);
3815 }
3816
3817 return ok;
3818}
3819
0f830e0b 3820/*
87cea99e 3821=for apidoc pv_uni_display
d2cc3551
JH
3822
3823Build to the scalar dsv a displayable version of the string spv,
3824length len, the displayable version being at most pvlim bytes long
3825</