This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
regcomp.c: Silence compiler warning about longjump
[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
82686b01 326=cut */
067a85ef 327STRLEN
668b6d8d 328Perl_is_utf8_char(const U8 *s)
386d01d6 329{
44f8325f 330 const STRLEN len = UTF8SKIP(s);
7918f24d
NC
331
332 PERL_ARGS_ASSERT_IS_UTF8_CHAR;
3b0fc154 333#ifdef IS_UTF8_CHAR
768c67ee 334 if (IS_UTF8_CHAR_FAST(len))
3b0fc154
JH
335 return IS_UTF8_CHAR(s, len) ? len : 0;
336#endif /* #ifdef IS_UTF8_CHAR */
2c0c5f92 337 return is_utf8_char_slow(s, len);
386d01d6
GS
338}
339
eaf7a4d2 340
6662521e 341/*
87cea99e 342=for apidoc is_utf8_string
6662521e 343
c9ada85f 344Returns true if first C<len> bytes of the given string form a valid
9f7e3d64
MH
345UTF-8 string, false otherwise. If C<len> is 0, it will be calculated
346using C<strlen(s)>. Note that 'a valid UTF-8 string' does not mean 'a
347string that contains code points above 0x7F encoded in UTF-8' because a
348valid ASCII string is a valid UTF-8 string.
6662521e 349
eaf7a4d2 350See also is_ascii_string(), is_utf8_string_loclen(), and is_utf8_string_loc().
768c67ee 351
6662521e
GS
352=cut
353*/
354
8e84507e 355bool
668b6d8d 356Perl_is_utf8_string(const U8 *s, STRLEN len)
6662521e 357{
35da51f7 358 const U8* const send = s + (len ? len : strlen((const char *)s));
7fc63493 359 const U8* x = s;
067a85ef 360
7918f24d 361 PERL_ARGS_ASSERT_IS_UTF8_STRING;
1aa99e6b 362
6662521e 363 while (x < send) {
a3b680e6 364 STRLEN c;
1acdb0da
JH
365 /* Inline the easy bits of is_utf8_char() here for speed... */
366 if (UTF8_IS_INVARIANT(*x))
367 c = 1;
368 else if (!UTF8_IS_START(*x))
768c67ee 369 goto out;
1acdb0da
JH
370 else {
371 /* ... and call is_utf8_char() only if really needed. */
646ca15d
JH
372#ifdef IS_UTF8_CHAR
373 c = UTF8SKIP(x);
768c67ee
JH
374 if (IS_UTF8_CHAR_FAST(c)) {
375 if (!IS_UTF8_CHAR(x, c))
3c614e38
AD
376 c = 0;
377 }
378 else
379 c = is_utf8_char_slow(x, c);
646ca15d
JH
380#else
381 c = is_utf8_char(x);
382#endif /* #ifdef IS_UTF8_CHAR */
1acdb0da 383 if (!c)
768c67ee 384 goto out;
1acdb0da 385 }
6662521e 386 x += c;
6662521e 387 }
768c67ee
JH
388
389 out:
60006e79
JH
390 if (x != send)
391 return FALSE;
067a85ef
A
392
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
JH
429 while (x < send) {
430 /* Inline the easy bits of is_utf8_char() here for speed... */
431 if (UTF8_IS_INVARIANT(*x))
768c67ee
JH
432 c = 1;
433 else if (!UTF8_IS_START(*x))
434 goto out;
81cd54e3 435 else {
768c67ee
JH
436 /* ... and call is_utf8_char() only if really needed. */
437#ifdef IS_UTF8_CHAR
438 c = UTF8SKIP(x);
439 if (IS_UTF8_CHAR_FAST(c)) {
440 if (!IS_UTF8_CHAR(x, c))
441 c = 0;
442 } else
443 c = is_utf8_char_slow(x, c);
444#else
445 c = is_utf8_char(x);
446#endif /* #ifdef IS_UTF8_CHAR */
447 if (!c)
448 goto out;
81cd54e3 449 }
768c67ee 450 x += c;
3ebfea28 451 outlen++;
81cd54e3 452 }
768c67ee
JH
453
454 out:
3ebfea28
AL
455 if (el)
456 *el = outlen;
457
768c67ee
JH
458 if (ep)
459 *ep = x;
3ebfea28 460 return (x == send);
81cd54e3
JH
461}
462
463/*
768c67ee 464
87cea99e 465=for apidoc utf8n_to_uvuni
67e989fb 466
9041c2e3 467Bottom level UTF-8 decode routine.
949cf498
KW
468Returns the code point value of the first character in the string C<s>
469which is assumed to be in UTF-8 (or UTF-EBCDIC) encoding and no longer than
470C<curlen> bytes; C<retlen> will be set to the length, in bytes, of that
471character.
472
473The value of C<flags> determines the behavior when C<s> does not point to a
474well-formed UTF-8 character. If C<flags> is 0, when a malformation is found,
475C<retlen> is set to the expected length of the UTF-8 character in bytes, zero
476is returned, and if UTF-8 warnings haven't been lexically disabled, a warning
477is raised.
478
479Various ALLOW flags can be set in C<flags> to allow (and not warn on)
480individual types of malformations, such as the sequence being overlong (that
481is, when there is a shorter sequence that can express the same code point;
482overlong sequences are expressly forbidden in the UTF-8 standard due to
483potential security issues). Another malformation example is the first byte of
484a character not being a legal first byte. See F<utf8.h> for the list of such
485flags. Of course, the value returned by this function under such conditions is
486not reliable.
487
488The UTF8_CHECK_ONLY flag overrides the behavior when a non-allowed (by other
489flags) malformation is found. If this flag is set, the routine assumes that
490the caller will raise a warning, and this function will silently just set
491C<retlen> to C<-1> and return zero.
492
493Certain code points are considered problematic. These are Unicode surrogates,
494Unicode non-characters, and code points above the Unicode maximum of 0x10FFF.
495By default these are considered regular code points, but certain situations
496warrant special handling for them. if C<flags> contains
497UTF8_DISALLOW_ILLEGAL_INTERCHANGE, all three classes are treated as
498malformations and handled as such. The flags UTF8_DISALLOW_SURROGATE,
499UTF8_DISALLOW_NONCHAR, and UTF8_DISALLOW_SUPER (meaning above the legal Unicode
500maximum) can be set to disallow these categories individually.
501
502The flags UTF8_WARN_ILLEGAL_INTERCHANGE, UTF8_WARN_SURROGATE,
503UTF8_WARN_NONCHAR, and UTF8_WARN_SUPER will cause warning messages to be raised
504for their respective categories, but otherwise the code points are considered
505valid (not malformations). To get a category to both be treated as a
506malformation and raise a warning, specify both the WARN and DISALLOW flags.
507(But note that warnings are not raised if lexically disabled nor if
508UTF8_CHECK_ONLY is also specified.)
509
510Very large code points (above 0x7FFF_FFFF) are considered more problematic than
511the others that are above the Unicode legal maximum. There are several
512reasons, one of which is that the original UTF-8 specification never went above
513this number (the current 0x10FFF limit was imposed later). The UTF-8 encoding
514on ASCII platforms for these large code point begins with a byte containing
5150xFE or 0xFF. The UTF8_DISALLOW_FE_FF flag will cause them to be treated as
516malformations, while allowing smaller above-Unicode code points. (Of course
517UTF8_DISALLOW_SUPER will treat all above-Unicode code points, including these,
518as malformations.) Similarly, UTF8_WARN_FE_FF acts just like the other WARN
519flags, but applies just to these code points.
520
521All other code points corresponding to Unicode characters, including private
522use and those yet to be assigned, are never considered malformed and never
523warn.
67e989fb 524
9041c2e3
NIS
525Most code should use utf8_to_uvchr() rather than call this directly.
526
37607a96
PK
527=cut
528*/
67e989fb 529
a0ed51b3 530UV
7fc63493 531Perl_utf8n_to_uvuni(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen, U32 flags)
a0ed51b3 532{
97aff369 533 dVAR;
d4c19fe8 534 const U8 * const s0 = s;
9c5ffd7c 535 UV uv = *s, ouv = 0;
ba210ebe 536 STRLEN len = 1;
949cf498 537 bool dowarn = ckWARN_d(WARN_UTF8);
7fc63493 538 const UV startbyte = *s;
ba210ebe 539 STRLEN expectlen = 0;
a0dbb045 540 U32 warning = 0;
949cf498 541 SV* sv = NULL;
a0dbb045 542
7918f24d
NC
543 PERL_ARGS_ASSERT_UTF8N_TO_UVUNI;
544
949cf498 545/* This list is a superset of the UTF8_ALLOW_XXX. */
a0dbb045
JH
546
547#define UTF8_WARN_EMPTY 1
548#define UTF8_WARN_CONTINUATION 2
549#define UTF8_WARN_NON_CONTINUATION 3
f9bc8ed7
KW
550#define UTF8_WARN_SHORT 4
551#define UTF8_WARN_OVERFLOW 5
552#define UTF8_WARN_LONG 6
a0dbb045
JH
553
554 if (curlen == 0 &&
555 !(flags & UTF8_ALLOW_EMPTY)) {
556 warning = UTF8_WARN_EMPTY;
0c443dc2
JH
557 goto malformed;
558 }
559
1d72bdf6 560 if (UTF8_IS_INVARIANT(uv)) {
a0ed51b3
LW
561 if (retlen)
562 *retlen = 1;
c4d5f83a 563 return (UV) (NATIVE_TO_UTF(*s));
a0ed51b3 564 }
67e989fb 565
421a8bf2 566 if (UTF8_IS_CONTINUATION(uv) &&
fcc8fcf6 567 !(flags & UTF8_ALLOW_CONTINUATION)) {
a0dbb045 568 warning = UTF8_WARN_CONTINUATION;
ba210ebe
JH
569 goto malformed;
570 }
571
421a8bf2 572 if (UTF8_IS_START(uv) && curlen > 1 && !UTF8_IS_CONTINUATION(s[1]) &&
fcc8fcf6 573 !(flags & UTF8_ALLOW_NON_CONTINUATION)) {
a0dbb045 574 warning = UTF8_WARN_NON_CONTINUATION;
ba210ebe
JH
575 goto malformed;
576 }
9041c2e3 577
1d72bdf6 578#ifdef EBCDIC
75383841 579 uv = NATIVE_TO_UTF(uv);
1d72bdf6 580#else
949cf498
KW
581 if (uv == 0xfe || uv == 0xff) {
582 if (flags & (UTF8_WARN_SUPER|UTF8_WARN_FE_FF)) {
111d382d 583 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "Code point beginning with byte 0x%02"UVXf" is not Unicode, and not portable", uv));
949cf498
KW
584 flags &= ~UTF8_WARN_SUPER; /* Only warn once on this problem */
585 }
586 if (flags & (UTF8_DISALLOW_SUPER|UTF8_DISALLOW_FE_FF)) {
587 goto malformed;
588 }
a0ed51b3 589 }
1d72bdf6
NIS
590#endif
591
ba210ebe
JH
592 if (!(uv & 0x20)) { len = 2; uv &= 0x1f; }
593 else if (!(uv & 0x10)) { len = 3; uv &= 0x0f; }
594 else if (!(uv & 0x08)) { len = 4; uv &= 0x07; }
595 else if (!(uv & 0x04)) { len = 5; uv &= 0x03; }
1d72bdf6
NIS
596#ifdef EBCDIC
597 else if (!(uv & 0x02)) { len = 6; uv &= 0x01; }
598 else { len = 7; uv &= 0x01; }
599#else
ba210ebe
JH
600 else if (!(uv & 0x02)) { len = 6; uv &= 0x01; }
601 else if (!(uv & 0x01)) { len = 7; uv = 0; }
1d72bdf6
NIS
602 else { len = 13; uv = 0; } /* whoa! */
603#endif
604
a0ed51b3
LW
605 if (retlen)
606 *retlen = len;
9041c2e3 607
ba210ebe
JH
608 expectlen = len;
609
fcc8fcf6
JH
610 if ((curlen < expectlen) &&
611 !(flags & UTF8_ALLOW_SHORT)) {
a0dbb045 612 warning = UTF8_WARN_SHORT;
ba210ebe
JH
613 goto malformed;
614 }
615
616 len--;
a0ed51b3 617 s++;
949cf498 618 ouv = uv; /* ouv is the value from the previous iteration */
ba210ebe 619
a0ed51b3 620 while (len--) {
421a8bf2
JH
621 if (!UTF8_IS_CONTINUATION(*s) &&
622 !(flags & UTF8_ALLOW_NON_CONTINUATION)) {
a0dbb045
JH
623 s--;
624 warning = UTF8_WARN_NON_CONTINUATION;
ba210ebe 625 goto malformed;
a0ed51b3
LW
626 }
627 else
8850bf83 628 uv = UTF8_ACCUMULATE(uv, *s);
949cf498
KW
629 if (!(uv > ouv)) { /* If the value didn't grow from the previous
630 iteration, something is horribly wrong */
a0dbb045
JH
631 /* These cannot be allowed. */
632 if (uv == ouv) {
75dbc644 633 if (expectlen != 13 && !(flags & UTF8_ALLOW_LONG)) {
a0dbb045
JH
634 warning = UTF8_WARN_LONG;
635 goto malformed;
636 }
637 }
638 else { /* uv < ouv */
639 /* This cannot be allowed. */
640 warning = UTF8_WARN_OVERFLOW;
641 goto malformed;
642 }
ba210ebe
JH
643 }
644 s++;
645 ouv = uv;
646 }
647
949cf498 648 if ((expectlen > (STRLEN)UNISKIP(uv)) && !(flags & UTF8_ALLOW_LONG)) {
a0dbb045 649 warning = UTF8_WARN_LONG;
ba210ebe 650 goto malformed;
949cf498
KW
651 } else if (flags & (UTF8_DISALLOW_ILLEGAL_INTERCHANGE|UTF8_WARN_ILLEGAL_INTERCHANGE)) {
652 if (UNICODE_IS_SURROGATE(uv)) {
653 if ((flags & (UTF8_WARN_SURROGATE|UTF8_CHECK_ONLY)) == UTF8_WARN_SURROGATE) {
111d382d 654 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "UTF-16 surrogate U+%04"UVXf"", uv));
949cf498
KW
655 }
656 if (flags & UTF8_DISALLOW_SURROGATE) {
657 goto disallowed;
658 }
659 }
660 else if (UNICODE_IS_NONCHAR(uv)) {
661 if ((flags & (UTF8_WARN_NONCHAR|UTF8_CHECK_ONLY)) == UTF8_WARN_NONCHAR ) {
111d382d 662 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "Unicode non-character U+%04"UVXf" is illegal for open interchange", uv));
949cf498
KW
663 }
664 if (flags & UTF8_DISALLOW_NONCHAR) {
665 goto disallowed;
666 }
667 }
668 else if ((uv > PERL_UNICODE_MAX)) {
669 if ((flags & (UTF8_WARN_SUPER|UTF8_CHECK_ONLY)) == UTF8_WARN_SUPER) {
111d382d 670 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "Code point 0x%04"UVXf" is not Unicode, may not be portable", uv));
949cf498
KW
671 }
672 if (flags & UTF8_DISALLOW_SUPER) {
673 goto disallowed;
674 }
675 }
676
677 /* Here, this is not considered a malformed character, so drop through
678 * to return it */
a0ed51b3 679 }
ba210ebe 680
a0ed51b3 681 return uv;
ba210ebe 682
949cf498
KW
683disallowed: /* Is disallowed, but otherwise not malformed. 'sv' will have been
684 set if there is to be a warning. */
685 if (!sv) {
686 dowarn = 0;
687 }
688
ba210ebe
JH
689malformed:
690
fcc8fcf6 691 if (flags & UTF8_CHECK_ONLY) {
ba210ebe 692 if (retlen)
10edeb5d 693 *retlen = ((STRLEN) -1);
ba210ebe
JH
694 return 0;
695 }
696
a0dbb045 697 if (dowarn) {
949cf498 698 if (! sv) {
5b311467 699 sv = newSVpvs_flags("Malformed UTF-8 character ", SVs_TEMP);
949cf498 700 }
5b311467 701
5e3d7cf5
KW
702 switch (warning) {
703 case 0: /* Intentionally empty. */ break;
704 case UTF8_WARN_EMPTY:
705 sv_catpvs(sv, "(empty string)");
706 break;
707 case UTF8_WARN_CONTINUATION:
708 Perl_sv_catpvf(aTHX_ sv, "(unexpected continuation byte 0x%02"UVxf", with no preceding start byte)", uv);
709 break;
710 case UTF8_WARN_NON_CONTINUATION:
711 if (s == s0)
712 Perl_sv_catpvf(aTHX_ sv, "(unexpected non-continuation byte 0x%02"UVxf", immediately after start byte 0x%02"UVxf")",
713 (UV)s[1], startbyte);
714 else {
715 const int len = (int)(s-s0);
716 Perl_sv_catpvf(aTHX_ sv, "(unexpected non-continuation byte 0x%02"UVxf", %d byte%s after start byte 0x%02"UVxf", expected %d bytes)",
717 (UV)s[1], len, len > 1 ? "s" : "", startbyte, (int)expectlen);
718 }
719
720 break;
721 case UTF8_WARN_SHORT:
722 Perl_sv_catpvf(aTHX_ sv, "(%d byte%s, need %d, after start byte 0x%02"UVxf")",
723 (int)curlen, curlen == 1 ? "" : "s", (int)expectlen, startbyte);
724 expectlen = curlen; /* distance for caller to skip */
725 break;
726 case UTF8_WARN_OVERFLOW:
727 Perl_sv_catpvf(aTHX_ sv, "(overflow at 0x%"UVxf", byte 0x%02x, after start byte 0x%02"UVxf")",
728 ouv, *s, startbyte);
729 break;
730 case UTF8_WARN_LONG:
731 Perl_sv_catpvf(aTHX_ sv, "(%d byte%s, need %d, after start byte 0x%02"UVxf")",
732 (int)expectlen, expectlen == 1 ? "": "s", UNISKIP(uv), startbyte);
733 break;
734 default:
735 sv_catpvs(sv, "(unknown reason)");
736 break;
737 }
a0dbb045 738
949cf498 739 if (sv) {
44f8325f 740 const char * const s = SvPVX_const(sv);
a0dbb045
JH
741
742 if (PL_op)
9014280d 743 Perl_warner(aTHX_ packWARN(WARN_UTF8),
53e06cf0 744 "%s in %s", s, OP_DESC(PL_op));
a0dbb045 745 else
9014280d 746 Perl_warner(aTHX_ packWARN(WARN_UTF8), "%s", s);
a0dbb045
JH
747 }
748 }
749
ba210ebe 750 if (retlen)
28d3d195 751 *retlen = expectlen ? expectlen : len;
ba210ebe 752
28d3d195 753 return 0;
a0ed51b3
LW
754}
755
8e84507e 756/*
87cea99e 757=for apidoc utf8_to_uvchr
9041c2e3 758
6ee84de2 759Returns the native code point of the first character in the string C<s>
1e54db1a 760which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
9041c2e3
NIS
761length, in bytes, of that character.
762
1e54db1a 763If C<s> does not point to a well-formed UTF-8 character, zero is
9041c2e3
NIS
764returned and retlen is set, if possible, to -1.
765
766=cut
767*/
768
ae0e24b1 769
9041c2e3 770UV
7fc63493 771Perl_utf8_to_uvchr(pTHX_ const U8 *s, STRLEN *retlen)
9041c2e3 772{
7918f24d
NC
773 PERL_ARGS_ASSERT_UTF8_TO_UVCHR;
774
1754c1a1 775 return utf8n_to_uvchr(s, UTF8_MAXBYTES, retlen,
ae0e24b1 776 ckWARN_d(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
9041c2e3
NIS
777}
778
779/*
87cea99e 780=for apidoc utf8_to_uvuni
9041c2e3
NIS
781
782Returns the Unicode code point of the first character in the string C<s>
1e54db1a 783which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
9041c2e3
NIS
784length, in bytes, of that character.
785
2bbc8d55 786This function should only be used when the returned UV is considered
9041c2e3
NIS
787an index into the Unicode semantic tables (e.g. swashes).
788
1e54db1a 789If C<s> does not point to a well-formed UTF-8 character, zero is
ba210ebe 790returned and retlen is set, if possible, to -1.
8e84507e
NIS
791
792=cut
793*/
794
795UV
7fc63493 796Perl_utf8_to_uvuni(pTHX_ const U8 *s, STRLEN *retlen)
8e84507e 797{
7918f24d
NC
798 PERL_ARGS_ASSERT_UTF8_TO_UVUNI;
799
9041c2e3 800 /* Call the low level routine asking for checks */
89ebb4a3 801 return Perl_utf8n_to_uvuni(aTHX_ s, UTF8_MAXBYTES, retlen,
ae0e24b1 802 ckWARN_d(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
8e84507e
NIS
803}
804
b76347f2 805/*
87cea99e 806=for apidoc utf8_length
b76347f2
JH
807
808Return the length of the UTF-8 char encoded string C<s> in characters.
02eb7b47
JH
809Stops at C<e> (inclusive). If C<e E<lt> s> or if the scan would end
810up past C<e>, croaks.
b76347f2
JH
811
812=cut
813*/
814
815STRLEN
35a4481c 816Perl_utf8_length(pTHX_ const U8 *s, const U8 *e)
b76347f2 817{
97aff369 818 dVAR;
b76347f2
JH
819 STRLEN len = 0;
820
7918f24d
NC
821 PERL_ARGS_ASSERT_UTF8_LENGTH;
822
8850bf83
JH
823 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g.
824 * the bitops (especially ~) can create illegal UTF-8.
825 * In other words: in Perl UTF-8 is not just for Unicode. */
826
a3b680e6
AL
827 if (e < s)
828 goto warn_and_return;
b76347f2 829 while (s < e) {
8e91ec7f
AV
830 if (!UTF8_IS_INVARIANT(*s))
831 s += UTF8SKIP(s);
832 else
833 s++;
834 len++;
835 }
836
837 if (e != s) {
838 len--;
839 warn_and_return:
9b387841
NC
840 if (PL_op)
841 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8),
842 "%s in %s", unees, OP_DESC(PL_op));
843 else
61a12c31 844 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), "%s", unees);
b76347f2
JH
845 }
846
847 return len;
848}
849
b06226ff 850/*
87cea99e 851=for apidoc utf8_distance
b06226ff 852
1e54db1a 853Returns the number of UTF-8 characters between the UTF-8 pointers C<a>
b06226ff
JH
854and C<b>.
855
856WARNING: use only if you *know* that the pointers point inside the
857same UTF-8 buffer.
858
37607a96
PK
859=cut
860*/
a0ed51b3 861
02eb7b47 862IV
35a4481c 863Perl_utf8_distance(pTHX_ const U8 *a, const U8 *b)
a0ed51b3 864{
7918f24d
NC
865 PERL_ARGS_ASSERT_UTF8_DISTANCE;
866
bf1665bc 867 return (a < b) ? -1 * (IV) utf8_length(a, b) : (IV) utf8_length(b, a);
a0ed51b3
LW
868}
869
b06226ff 870/*
87cea99e 871=for apidoc utf8_hop
b06226ff 872
8850bf83
JH
873Return the UTF-8 pointer C<s> displaced by C<off> characters, either
874forward or backward.
b06226ff
JH
875
876WARNING: do not use the following unless you *know* C<off> is within
8850bf83
JH
877the UTF-8 data pointed to by C<s> *and* that on entry C<s> is aligned
878on the first byte of character or just after the last byte of a character.
b06226ff 879
37607a96
PK
880=cut
881*/
a0ed51b3
LW
882
883U8 *
4373e329 884Perl_utf8_hop(pTHX_ const U8 *s, I32 off)
a0ed51b3 885{
7918f24d
NC
886 PERL_ARGS_ASSERT_UTF8_HOP;
887
96a5add6 888 PERL_UNUSED_CONTEXT;
8850bf83
JH
889 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g
890 * the bitops (especially ~) can create illegal UTF-8.
891 * In other words: in Perl UTF-8 is not just for Unicode. */
892
a0ed51b3
LW
893 if (off >= 0) {
894 while (off--)
895 s += UTF8SKIP(s);
896 }
897 else {
898 while (off++) {
899 s--;
8850bf83
JH
900 while (UTF8_IS_CONTINUATION(*s))
901 s--;
a0ed51b3
LW
902 }
903 }
4373e329 904 return (U8 *)s;
a0ed51b3
LW
905}
906
6940069f 907/*
fed3ba5d
NC
908=for apidoc bytes_cmp_utf8
909
910Compares the sequence of characters (stored as octets) in b, blen with the
911sequence of characters (stored as UTF-8) in u, ulen. Returns 0 if they are
912equal, -1 or -2 if the first string is less than the second string, +1 or +2
913if the first string is greater than the second string.
914
915-1 or +1 is returned if the shorter string was identical to the start of the
916longer string. -2 or +2 is returned if the was a difference between characters
917within the strings.
918
919=cut
920*/
921
922int
923Perl_bytes_cmp_utf8(pTHX_ const U8 *b, STRLEN blen, const U8 *u, STRLEN ulen)
924{
925 const U8 *const bend = b + blen;
926 const U8 *const uend = u + ulen;
927
928 PERL_ARGS_ASSERT_BYTES_CMP_UTF8;
929
930 PERL_UNUSED_CONTEXT;
931
932 while (b < bend && u < uend) {
933 U8 c = *u++;
934 if (!UTF8_IS_INVARIANT(c)) {
935 if (UTF8_IS_DOWNGRADEABLE_START(c)) {
936 if (u < uend) {
937 U8 c1 = *u++;
938 if (UTF8_IS_CONTINUATION(c1)) {
356979f4 939 c = UNI_TO_NATIVE(TWO_BYTE_UTF8_TO_UNI(c, c1));
fed3ba5d
NC
940 } else {
941 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8),
942 "Malformed UTF-8 character "
943 "(unexpected non-continuation byte 0x%02x"
944 ", immediately after start byte 0x%02x)"
945 /* Dear diag.t, it's in the pod. */
946 "%s%s", c1, c,
947 PL_op ? " in " : "",
948 PL_op ? OP_DESC(PL_op) : "");
949 return -2;
950 }
951 } else {
952 if (PL_op)
953 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8),
954 "%s in %s", unees, OP_DESC(PL_op));
955 else
61a12c31 956 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), "%s", unees);
fed3ba5d
NC
957 return -2; /* Really want to return undef :-) */
958 }
959 } else {
960 return -2;
961 }
962 }
963 if (*b != c) {
964 return *b < c ? -2 : +2;
965 }
966 ++b;
967 }
968
969 if (b == bend && u == uend)
970 return 0;
971
972 return b < bend ? +1 : -1;
973}
974
975/*
87cea99e 976=for apidoc utf8_to_bytes
6940069f 977
2bbc8d55 978Converts a string C<s> of length C<len> from UTF-8 into native byte encoding.
246fae53
MG
979Unlike C<bytes_to_utf8>, this over-writes the original string, and
980updates len to contain the new length.
67e989fb 981Returns zero on failure, setting C<len> to -1.
6940069f 982
95be277c
NC
983If you need a copy of the string, see C<bytes_from_utf8>.
984
6940069f
GS
985=cut
986*/
987
988U8 *
37607a96 989Perl_utf8_to_bytes(pTHX_ U8 *s, STRLEN *len)
6940069f 990{
d4c19fe8
AL
991 U8 * const save = s;
992 U8 * const send = s + *len;
6940069f 993 U8 *d;
246fae53 994
7918f24d
NC
995 PERL_ARGS_ASSERT_UTF8_TO_BYTES;
996
1e54db1a 997 /* ensure valid UTF-8 and chars < 256 before updating string */
d4c19fe8 998 while (s < send) {
dcad2880
JH
999 U8 c = *s++;
1000
1d72bdf6
NIS
1001 if (!UTF8_IS_INVARIANT(c) &&
1002 (!UTF8_IS_DOWNGRADEABLE_START(c) || (s >= send)
1003 || !(c = *s++) || !UTF8_IS_CONTINUATION(c))) {
10edeb5d 1004 *len = ((STRLEN) -1);
dcad2880
JH
1005 return 0;
1006 }
246fae53 1007 }
dcad2880
JH
1008
1009 d = s = save;
6940069f 1010 while (s < send) {
ed646e6e 1011 STRLEN ulen;
9041c2e3 1012 *d++ = (U8)utf8_to_uvchr(s, &ulen);
ed646e6e 1013 s += ulen;
6940069f
GS
1014 }
1015 *d = '\0';
246fae53 1016 *len = d - save;
6940069f
GS
1017 return save;
1018}
1019
1020/*
87cea99e 1021=for apidoc bytes_from_utf8
f9a63242 1022
2bbc8d55 1023Converts a string C<s> of length C<len> from UTF-8 into native byte encoding.
35a4481c 1024Unlike C<utf8_to_bytes> but like C<bytes_to_utf8>, returns a pointer to
ef9edfd0
JH
1025the newly-created string, and updates C<len> to contain the new
1026length. Returns the original string if no conversion occurs, C<len>
1027is unchanged. Do nothing if C<is_utf8> points to 0. Sets C<is_utf8> to
2bbc8d55
SP
10280 if C<s> is converted or consisted entirely of characters that are invariant
1029in utf8 (i.e., US-ASCII on non-EBCDIC machines).
f9a63242 1030
37607a96
PK
1031=cut
1032*/
f9a63242
JH
1033
1034U8 *
e1ec3a88 1035Perl_bytes_from_utf8(pTHX_ const U8 *s, STRLEN *len, bool *is_utf8)
f9a63242 1036{
f9a63242 1037 U8 *d;
e1ec3a88
AL
1038 const U8 *start = s;
1039 const U8 *send;
f9a63242
JH
1040 I32 count = 0;
1041
7918f24d
NC
1042 PERL_ARGS_ASSERT_BYTES_FROM_UTF8;
1043
96a5add6 1044 PERL_UNUSED_CONTEXT;
f9a63242 1045 if (!*is_utf8)
73d840c0 1046 return (U8 *)start;
f9a63242 1047
1e54db1a 1048 /* ensure valid UTF-8 and chars < 256 before converting string */
f9a63242 1049 for (send = s + *len; s < send;) {
e1ec3a88 1050 U8 c = *s++;
1d72bdf6 1051 if (!UTF8_IS_INVARIANT(c)) {
db42d148
NIS
1052 if (UTF8_IS_DOWNGRADEABLE_START(c) && s < send &&
1053 (c = *s++) && UTF8_IS_CONTINUATION(c))
1054 count++;
1055 else
73d840c0 1056 return (U8 *)start;
db42d148 1057 }
f9a63242
JH
1058 }
1059
35da51f7 1060 *is_utf8 = FALSE;
f9a63242 1061
212542aa 1062 Newx(d, (*len) - count + 1, U8);
ef9edfd0 1063 s = start; start = d;
f9a63242
JH
1064 while (s < send) {
1065 U8 c = *s++;
c4d5f83a
NIS
1066 if (!UTF8_IS_INVARIANT(c)) {
1067 /* Then it is two-byte encoded */
356979f4 1068 c = UNI_TO_NATIVE(TWO_BYTE_UTF8_TO_UNI(c, *s++));
c4d5f83a
NIS
1069 }
1070 *d++ = c;
f9a63242
JH
1071 }
1072 *d = '\0';
1073 *len = d - start;
73d840c0 1074 return (U8 *)start;
f9a63242
JH
1075}
1076
1077/*
87cea99e 1078=for apidoc bytes_to_utf8
6940069f 1079
ff97e5cf
KW
1080Converts a string C<s> of length C<len> bytes from the native encoding into
1081UTF-8.
6662521e 1082Returns a pointer to the newly-created string, and sets C<len> to
ff97e5cf 1083reflect the new length in bytes.
6940069f 1084
2bbc8d55
SP
1085A NUL character will be written after the end of the string.
1086
1087If you want to convert to UTF-8 from encodings other than
1088the native (Latin1 or EBCDIC),
c9ada85f
JH
1089see sv_recode_to_utf8().
1090
497711e7 1091=cut
6940069f
GS
1092*/
1093
c682ebef
FC
1094/* This logic is duplicated in sv_catpvn_flags, so any bug fixes will
1095 likewise need duplication. */
1096
6940069f 1097U8*
35a4481c 1098Perl_bytes_to_utf8(pTHX_ const U8 *s, STRLEN *len)
6940069f 1099{
35a4481c 1100 const U8 * const send = s + (*len);
6940069f
GS
1101 U8 *d;
1102 U8 *dst;
7918f24d
NC
1103
1104 PERL_ARGS_ASSERT_BYTES_TO_UTF8;
96a5add6 1105 PERL_UNUSED_CONTEXT;
6940069f 1106
212542aa 1107 Newx(d, (*len) * 2 + 1, U8);
6940069f
GS
1108 dst = d;
1109
1110 while (s < send) {
35a4481c 1111 const UV uv = NATIVE_TO_ASCII(*s++);
c4d5f83a 1112 if (UNI_IS_INVARIANT(uv))
eb160463 1113 *d++ = (U8)UTF_TO_NATIVE(uv);
6940069f 1114 else {
eb160463
GS
1115 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
1116 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
6940069f
GS
1117 }
1118 }
1119 *d = '\0';
6662521e 1120 *len = d-dst;
6940069f
GS
1121 return dst;
1122}
1123
a0ed51b3 1124/*
dea0fc0b 1125 * Convert native (big-endian) or reversed (little-endian) UTF-16 to UTF-8.
a0ed51b3
LW
1126 *
1127 * Destination must be pre-extended to 3/2 source. Do not use in-place.
1128 * We optimize for native, for obvious reasons. */
1129
1130U8*
dea0fc0b 1131Perl_utf16_to_utf8(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen)
a0ed51b3 1132{
dea0fc0b
JH
1133 U8* pend;
1134 U8* dstart = d;
1135
7918f24d
NC
1136 PERL_ARGS_ASSERT_UTF16_TO_UTF8;
1137
dea0fc0b 1138 if (bytelen & 1)
f5992bc4 1139 Perl_croak(aTHX_ "panic: utf16_to_utf8: odd bytelen %"UVuf, (UV)bytelen);
dea0fc0b
JH
1140
1141 pend = p + bytelen;
1142
a0ed51b3 1143 while (p < pend) {
dea0fc0b
JH
1144 UV uv = (p[0] << 8) + p[1]; /* UTF-16BE */
1145 p += 2;
a0ed51b3 1146 if (uv < 0x80) {
e294cc5d
JH
1147#ifdef EBCDIC
1148 *d++ = UNI_TO_NATIVE(uv);
1149#else
eb160463 1150 *d++ = (U8)uv;
e294cc5d 1151#endif
a0ed51b3
LW
1152 continue;
1153 }
1154 if (uv < 0x800) {
eb160463
GS
1155 *d++ = (U8)(( uv >> 6) | 0xc0);
1156 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
1157 continue;
1158 }
52b9aa85 1159 if (uv >= 0xd800 && uv <= 0xdbff) { /* surrogates */
01ea242b 1160 if (p >= pend) {
dea0fc0b 1161 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
01ea242b
NC
1162 } else {
1163 UV low = (p[0] << 8) + p[1];
1164 p += 2;
52b9aa85 1165 if (low < 0xdc00 || low > 0xdfff)
01ea242b
NC
1166 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
1167 uv = ((uv - 0xd800) << 10) + (low - 0xdc00) + 0x10000;
1168 }
dbde1951
NC
1169 } else if (uv >= 0xdc00 && uv <= 0xdfff) {
1170 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
a0ed51b3
LW
1171 }
1172 if (uv < 0x10000) {
eb160463
GS
1173 *d++ = (U8)(( uv >> 12) | 0xe0);
1174 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
1175 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
1176 continue;
1177 }
1178 else {
eb160463
GS
1179 *d++ = (U8)(( uv >> 18) | 0xf0);
1180 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
1181 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
1182 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
1183 continue;
1184 }
1185 }
dea0fc0b 1186 *newlen = d - dstart;
a0ed51b3
LW
1187 return d;
1188}
1189
1190/* Note: this one is slightly destructive of the source. */
1191
1192U8*
dea0fc0b 1193Perl_utf16_to_utf8_reversed(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen)
a0ed51b3
LW
1194{
1195 U8* s = (U8*)p;
d4c19fe8 1196 U8* const send = s + bytelen;
7918f24d
NC
1197
1198 PERL_ARGS_ASSERT_UTF16_TO_UTF8_REVERSED;
1199
e0ea5e2d
NC
1200 if (bytelen & 1)
1201 Perl_croak(aTHX_ "panic: utf16_to_utf8_reversed: odd bytelen %"UVuf,
1202 (UV)bytelen);
1203
a0ed51b3 1204 while (s < send) {
d4c19fe8 1205 const U8 tmp = s[0];
a0ed51b3
LW
1206 s[0] = s[1];
1207 s[1] = tmp;
1208 s += 2;
1209 }
dea0fc0b 1210 return utf16_to_utf8(p, d, bytelen, newlen);
a0ed51b3
LW
1211}
1212
c3fd2246
KW
1213/* for now these are all defined (inefficiently) in terms of the utf8 versions.
1214 * Note that the macros in handy.h that call these short-circuit calling them
1215 * for Latin-1 range inputs */
a0ed51b3
LW
1216
1217bool
84afefe6 1218Perl_is_uni_alnum(pTHX_ UV c)
a0ed51b3 1219{
89ebb4a3 1220 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1221 uvchr_to_utf8(tmpbuf, c);
a0ed51b3
LW
1222 return is_utf8_alnum(tmpbuf);
1223}
1224
1225bool
84afefe6 1226Perl_is_uni_idfirst(pTHX_ UV c)
a0ed51b3 1227{
89ebb4a3 1228 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1229 uvchr_to_utf8(tmpbuf, c);
a0ed51b3
LW
1230 return is_utf8_idfirst(tmpbuf);
1231}
1232
1233bool
84afefe6 1234Perl_is_uni_alpha(pTHX_ UV c)
a0ed51b3 1235{
89ebb4a3 1236 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1237 uvchr_to_utf8(tmpbuf, c);
a0ed51b3
LW
1238 return is_utf8_alpha(tmpbuf);
1239}
1240
1241bool
84afefe6 1242Perl_is_uni_ascii(pTHX_ UV c)
4d61ec05 1243{
bc39fe24 1244 return isASCII(c);
4d61ec05
GS
1245}
1246
1247bool
84afefe6 1248Perl_is_uni_space(pTHX_ UV c)
a0ed51b3 1249{
89ebb4a3 1250 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1251 uvchr_to_utf8(tmpbuf, c);
a0ed51b3
LW
1252 return is_utf8_space(tmpbuf);
1253}
1254
1255bool
84afefe6 1256Perl_is_uni_digit(pTHX_ UV c)
a0ed51b3 1257{
89ebb4a3 1258 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1259 uvchr_to_utf8(tmpbuf, c);
a0ed51b3
LW
1260 return is_utf8_digit(tmpbuf);
1261}
1262
1263bool
84afefe6 1264Perl_is_uni_upper(pTHX_ UV c)
a0ed51b3 1265{
89ebb4a3 1266 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1267 uvchr_to_utf8(tmpbuf, c);
a0ed51b3
LW
1268 return is_utf8_upper(tmpbuf);
1269}
1270
1271bool
84afefe6 1272Perl_is_uni_lower(pTHX_ UV c)
a0ed51b3 1273{
89ebb4a3 1274 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1275 uvchr_to_utf8(tmpbuf, c);
a0ed51b3
LW
1276 return is_utf8_lower(tmpbuf);
1277}
1278
1279bool
84afefe6 1280Perl_is_uni_cntrl(pTHX_ UV c)
b8c5462f 1281{
7b952154 1282 return isCNTRL_L1(c);
b8c5462f
JH
1283}
1284
1285bool
84afefe6 1286Perl_is_uni_graph(pTHX_ UV c)
b8c5462f 1287{
89ebb4a3 1288 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1289 uvchr_to_utf8(tmpbuf, c);
b8c5462f
JH
1290 return is_utf8_graph(tmpbuf);
1291}
1292
1293bool
84afefe6 1294Perl_is_uni_print(pTHX_ UV c)
a0ed51b3 1295{
89ebb4a3 1296 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1297 uvchr_to_utf8(tmpbuf, c);
a0ed51b3
LW
1298 return is_utf8_print(tmpbuf);
1299}
1300
b8c5462f 1301bool
84afefe6 1302Perl_is_uni_punct(pTHX_ UV c)
b8c5462f 1303{
89ebb4a3 1304 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1305 uvchr_to_utf8(tmpbuf, c);
b8c5462f
JH
1306 return is_utf8_punct(tmpbuf);
1307}
1308
4d61ec05 1309bool
84afefe6 1310Perl_is_uni_xdigit(pTHX_ UV c)
4d61ec05 1311{
89ebb4a3 1312 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
230880c1 1313 uvchr_to_utf8(tmpbuf, c);
4d61ec05
GS
1314 return is_utf8_xdigit(tmpbuf);
1315}
1316
c3fd2246 1317
84afefe6
JH
1318UV
1319Perl_to_uni_upper(pTHX_ UV c, U8* p, STRLEN *lenp)
a0ed51b3 1320{
c3fd2246
KW
1321 /* Convert the Unicode character whose ordinal is c to its uppercase
1322 * version and store that in UTF-8 in p and its length in bytes in lenp.
1323 * Note that the p needs to be at least UTF8_MAXBYTES_CASE+1 bytes since
1324 * the changed version may be longer than the original character.
1325 *
1326 * The ordinal of the first character of the changed version is returned
1327 * (but note, as explained above, that there may be more.) */
1328
7918f24d
NC
1329 PERL_ARGS_ASSERT_TO_UNI_UPPER;
1330
0ebc6274
JH
1331 uvchr_to_utf8(p, c);
1332 return to_utf8_upper(p, p, lenp);
a0ed51b3
LW
1333}
1334
84afefe6
JH
1335UV
1336Perl_to_uni_title(pTHX_ UV c, U8* p, STRLEN *lenp)
a0ed51b3 1337{
7918f24d
NC
1338 PERL_ARGS_ASSERT_TO_UNI_TITLE;
1339
0ebc6274
JH
1340 uvchr_to_utf8(p, c);
1341 return to_utf8_title(p, p, lenp);
a0ed51b3
LW
1342}
1343
84afefe6
JH
1344UV
1345Perl_to_uni_lower(pTHX_ UV c, U8* p, STRLEN *lenp)
a0ed51b3 1346{
7918f24d
NC
1347 PERL_ARGS_ASSERT_TO_UNI_LOWER;
1348
bca00c02
KW
1349 if (c > 255) {
1350 uvchr_to_utf8(p, c);
1351 return to_utf8_lower(p, p, lenp);
1352 }
1353
1354 /* We have the latin1-range values compiled into the core, so just use
1355 * those, converting the result to utf8 */
1356 c = toLOWER_LATIN1(c);
1357 if (UNI_IS_INVARIANT(c)) {
1358 *p = c;
1359 *lenp = 1;
1360 }
1361 else {
1362 *p = UTF8_TWO_BYTE_HI(c);
1363 *(p+1) = UTF8_TWO_BYTE_LO(c);
1364 *lenp = 2;
1365 }
1366 return c;
a0ed51b3
LW
1367}
1368
84afefe6 1369UV
36bb2ab6 1370Perl__to_uni_fold_flags(pTHX_ UV c, U8* p, STRLEN *lenp, U8 flags)
84afefe6 1371{
36bb2ab6 1372 PERL_ARGS_ASSERT__TO_UNI_FOLD_FLAGS;
7918f24d 1373
0ebc6274 1374 uvchr_to_utf8(p, c);
36bb2ab6 1375 return _to_utf8_fold_flags(p, p, lenp, flags);
84afefe6
JH
1376}
1377
a0ed51b3
LW
1378/* for now these all assume no locale info available for Unicode > 255 */
1379
1380bool
84afefe6 1381Perl_is_uni_alnum_lc(pTHX_ UV c)
a0ed51b3
LW
1382{
1383 return is_uni_alnum(c); /* XXX no locale support yet */
1384}
1385
1386bool
84afefe6 1387Perl_is_uni_idfirst_lc(pTHX_ UV c)
a0ed51b3
LW
1388{
1389 return is_uni_idfirst(c); /* XXX no locale support yet */
1390}
1391
1392bool
84afefe6 1393Perl_is_uni_alpha_lc(pTHX_ UV c)
a0ed51b3
LW
1394{
1395 return is_uni_alpha(c); /* XXX no locale support yet */
1396}
1397
1398bool
84afefe6 1399Perl_is_uni_ascii_lc(pTHX_ UV c)
4d61ec05
GS
1400{
1401 return is_uni_ascii(c); /* XXX no locale support yet */
1402}
1403
1404bool
84afefe6 1405Perl_is_uni_space_lc(pTHX_ UV c)
a0ed51b3
LW
1406{
1407 return is_uni_space(c); /* XXX no locale support yet */
1408}
1409
1410bool
84afefe6 1411Perl_is_uni_digit_lc(pTHX_ UV c)
a0ed51b3
LW
1412{
1413 return is_uni_digit(c); /* XXX no locale support yet */
1414}
1415
1416bool
84afefe6 1417Perl_is_uni_upper_lc(pTHX_ UV c)
a0ed51b3
LW
1418{
1419 return is_uni_upper(c); /* XXX no locale support yet */
1420}
1421
1422bool
84afefe6 1423Perl_is_uni_lower_lc(pTHX_ UV c)
a0ed51b3
LW
1424{
1425 return is_uni_lower(c); /* XXX no locale support yet */
1426}
1427
1428bool
84afefe6 1429Perl_is_uni_cntrl_lc(pTHX_ UV c)
b8c5462f
JH
1430{
1431 return is_uni_cntrl(c); /* XXX no locale support yet */
1432}
1433
1434bool
84afefe6 1435Perl_is_uni_graph_lc(pTHX_ UV c)
b8c5462f
JH
1436{
1437 return is_uni_graph(c); /* XXX no locale support yet */
1438}
1439
1440bool
84afefe6 1441Perl_is_uni_print_lc(pTHX_ UV c)
a0ed51b3
LW
1442{
1443 return is_uni_print(c); /* XXX no locale support yet */
1444}
1445
b8c5462f 1446bool
84afefe6 1447Perl_is_uni_punct_lc(pTHX_ UV c)
b8c5462f
JH
1448{
1449 return is_uni_punct(c); /* XXX no locale support yet */
1450}
1451
4d61ec05 1452bool
84afefe6 1453Perl_is_uni_xdigit_lc(pTHX_ UV c)
4d61ec05
GS
1454{
1455 return is_uni_xdigit(c); /* XXX no locale support yet */
1456}
1457
b7ac61fa
JH
1458U32
1459Perl_to_uni_upper_lc(pTHX_ U32 c)
1460{
ee099d14
JH
1461 /* XXX returns only the first character -- do not use XXX */
1462 /* XXX no locale support yet */
1463 STRLEN len;
89ebb4a3 1464 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
ee099d14 1465 return (U32)to_uni_upper(c, tmpbuf, &len);
b7ac61fa
JH
1466}
1467
1468U32
1469Perl_to_uni_title_lc(pTHX_ U32 c)
1470{
ee099d14
JH
1471 /* XXX returns only the first character XXX -- do not use XXX */
1472 /* XXX no locale support yet */
1473 STRLEN len;
89ebb4a3 1474 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
ee099d14 1475 return (U32)to_uni_title(c, tmpbuf, &len);
b7ac61fa
JH
1476}
1477
1478U32
1479Perl_to_uni_lower_lc(pTHX_ U32 c)
1480{
ee099d14
JH
1481 /* XXX returns only the first character -- do not use XXX */
1482 /* XXX no locale support yet */
1483 STRLEN len;
89ebb4a3 1484 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
ee099d14 1485 return (U32)to_uni_lower(c, tmpbuf, &len);
b7ac61fa
JH
1486}
1487
7452cf6a 1488static bool
5141f98e 1489S_is_utf8_common(pTHX_ const U8 *const p, SV **swash,
bde6a22d
NC
1490 const char *const swashname)
1491{
97aff369 1492 dVAR;
7918f24d
NC
1493
1494 PERL_ARGS_ASSERT_IS_UTF8_COMMON;
1495
bde6a22d
NC
1496 if (!is_utf8_char(p))
1497 return FALSE;
1498 if (!*swash)
711a919c 1499 *swash = swash_init("utf8", swashname, &PL_sv_undef, 1, 0);
bde6a22d
NC
1500 return swash_fetch(*swash, p, TRUE) != 0;
1501}
1502
1503bool
7fc63493 1504Perl_is_utf8_alnum(pTHX_ const U8 *p)
a0ed51b3 1505{
97aff369 1506 dVAR;
7918f24d
NC
1507
1508 PERL_ARGS_ASSERT_IS_UTF8_ALNUM;
1509
671c33bf
NC
1510 /* NOTE: "IsWord", not "IsAlnum", since Alnum is a true
1511 * descendant of isalnum(3), in other words, it doesn't
1512 * contain the '_'. --jhi */
d4c19fe8 1513 return is_utf8_common(p, &PL_utf8_alnum, "IsWord");
a0ed51b3
LW
1514}
1515
1516bool
7fc63493 1517Perl_is_utf8_idfirst(pTHX_ const U8 *p) /* The naming is historical. */
a0ed51b3 1518{
97aff369 1519 dVAR;
7918f24d
NC
1520
1521 PERL_ARGS_ASSERT_IS_UTF8_IDFIRST;
1522
82686b01
JH
1523 if (*p == '_')
1524 return TRUE;
bde6a22d 1525 /* is_utf8_idstart would be more logical. */
d4c19fe8 1526 return is_utf8_common(p, &PL_utf8_idstart, "IdStart");
82686b01
JH
1527}
1528
1529bool
c11ff943
KW
1530Perl_is_utf8_xidfirst(pTHX_ const U8 *p) /* The naming is historical. */
1531{
1532 dVAR;
1533
1534 PERL_ARGS_ASSERT_IS_UTF8_XIDFIRST;
1535
1536 if (*p == '_')
1537 return TRUE;
1538 /* is_utf8_idstart would be more logical. */
1539 return is_utf8_common(p, &PL_utf8_xidstart, "XIdStart");
1540}
1541
1542bool
b6912c02
KW
1543Perl__is_utf8__perl_idstart(pTHX_ const U8 *p)
1544{
1545 dVAR;
1546
1547 PERL_ARGS_ASSERT__IS_UTF8__PERL_IDSTART;
1548
1549 return is_utf8_common(p, &PL_utf8_perl_idstart, "_Perl_IDStart");
1550}
1551
1552bool
7fc63493 1553Perl_is_utf8_idcont(pTHX_ const U8 *p)
82686b01 1554{
97aff369 1555 dVAR;
7918f24d
NC
1556
1557 PERL_ARGS_ASSERT_IS_UTF8_IDCONT;
1558
d4c19fe8 1559 return is_utf8_common(p, &PL_utf8_idcont, "IdContinue");
a0ed51b3
LW
1560}
1561
1562bool
c11ff943
KW
1563Perl_is_utf8_xidcont(pTHX_ const U8 *p)
1564{
1565 dVAR;
1566
1567 PERL_ARGS_ASSERT_IS_UTF8_XIDCONT;
1568
c11ff943
KW
1569 return is_utf8_common(p, &PL_utf8_idcont, "XIdContinue");
1570}
1571
1572bool
7fc63493 1573Perl_is_utf8_alpha(pTHX_ const U8 *p)
a0ed51b3 1574{
97aff369 1575 dVAR;
7918f24d
NC
1576
1577 PERL_ARGS_ASSERT_IS_UTF8_ALPHA;
1578
d4c19fe8 1579 return is_utf8_common(p, &PL_utf8_alpha, "IsAlpha");
a0ed51b3
LW
1580}
1581
1582bool
7fc63493 1583Perl_is_utf8_ascii(pTHX_ const U8 *p)
b8c5462f 1584{
97aff369 1585 dVAR;
7918f24d
NC
1586
1587 PERL_ARGS_ASSERT_IS_UTF8_ASCII;
1588
bc39fe24
KW
1589 /* ASCII characters are the same whether in utf8 or not. So the macro
1590 * works on both utf8 and non-utf8 representations. */
1591 return isASCII(*p);
b8c5462f
JH
1592}
1593
1594bool
7fc63493 1595Perl_is_utf8_space(pTHX_ const U8 *p)
a0ed51b3 1596{
97aff369 1597 dVAR;
7918f24d
NC
1598
1599 PERL_ARGS_ASSERT_IS_UTF8_SPACE;
1600
a34094a9 1601 return is_utf8_common(p, &PL_utf8_space, "IsXPerlSpace");
a0ed51b3
LW
1602}
1603
1604bool
d1eb3177
YO
1605Perl_is_utf8_perl_space(pTHX_ const U8 *p)
1606{
1607 dVAR;
1608
1609 PERL_ARGS_ASSERT_IS_UTF8_PERL_SPACE;
1610
c4428693
KW
1611 /* Only true if is an ASCII space-like character, and ASCII is invariant
1612 * under utf8, so can just use the macro */
1613 return isSPACE_A(*p);
d1eb3177
YO
1614}
1615
1616bool
1617Perl_is_utf8_perl_word(pTHX_ const U8 *p)
1618{
1619 dVAR;
1620
1621 PERL_ARGS_ASSERT_IS_UTF8_PERL_WORD;
1622
c4428693
KW
1623 /* Only true if is an ASCII word character, and ASCII is invariant
1624 * under utf8, so can just use the macro */
1625 return isWORDCHAR_A(*p);
d1eb3177
YO
1626}
1627
1628bool
7fc63493 1629Perl_is_utf8_digit(pTHX_ const U8 *p)
a0ed51b3 1630{
97aff369 1631 dVAR;
7918f24d
NC
1632
1633 PERL_ARGS_ASSERT_IS_UTF8_DIGIT;
1634
d4c19fe8 1635 return is_utf8_common(p, &PL_utf8_digit, "IsDigit");
a0ed51b3
LW
1636}
1637
1638bool
d1eb3177
YO
1639Perl_is_utf8_posix_digit(pTHX_ const U8 *p)
1640{
1641 dVAR;
1642
1643 PERL_ARGS_ASSERT_IS_UTF8_POSIX_DIGIT;
1644
c4428693
KW
1645 /* Only true if is an ASCII digit character, and ASCII is invariant
1646 * under utf8, so can just use the macro */
1647 return isDIGIT_A(*p);
d1eb3177
YO
1648}
1649
1650bool
7fc63493 1651Perl_is_utf8_upper(pTHX_ const U8 *p)
a0ed51b3 1652{
97aff369 1653 dVAR;
7918f24d
NC
1654
1655 PERL_ARGS_ASSERT_IS_UTF8_UPPER;
1656
d4c19fe8 1657 return is_utf8_common(p, &PL_utf8_upper, "IsUppercase");
a0ed51b3
LW
1658}
1659
1660bool
7fc63493 1661Perl_is_utf8_lower(pTHX_ const U8 *p)
a0ed51b3 1662{
97aff369 1663 dVAR;
7918f24d
NC
1664
1665 PERL_ARGS_ASSERT_IS_UTF8_LOWER;
1666
d4c19fe8 1667 return is_utf8_common(p, &PL_utf8_lower, "IsLowercase");
a0ed51b3
LW
1668}
1669
1670bool
7fc63493 1671Perl_is_utf8_cntrl(pTHX_ const U8 *p)
b8c5462f 1672{
97aff369 1673 dVAR;
7918f24d
NC
1674
1675 PERL_ARGS_ASSERT_IS_UTF8_CNTRL;
1676
7b952154
KW
1677 if (isASCII(*p)) {
1678 return isCNTRL_A(*p);
1679 }
1680
1681 /* All controls are in Latin1 */
1682 if (! UTF8_IS_DOWNGRADEABLE_START(*p)) {
1683 return 0;
1684 }
1685 return isCNTRL_L1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)));
b8c5462f
JH
1686}
1687
1688bool
7fc63493 1689Perl_is_utf8_graph(pTHX_ const U8 *p)
b8c5462f 1690{
97aff369 1691 dVAR;
7918f24d
NC
1692
1693 PERL_ARGS_ASSERT_IS_UTF8_GRAPH;
1694
d4c19fe8 1695 return is_utf8_common(p, &PL_utf8_graph, "IsGraph");
b8c5462f
JH
1696}
1697
1698bool
7fc63493 1699Perl_is_utf8_print(pTHX_ const U8 *p)
a0ed51b3 1700{
97aff369 1701 dVAR;
7918f24d
NC
1702
1703 PERL_ARGS_ASSERT_IS_UTF8_PRINT;
1704
d4c19fe8 1705 return is_utf8_common(p, &PL_utf8_print, "IsPrint");
a0ed51b3
LW
1706}
1707
1708bool
7fc63493 1709Perl_is_utf8_punct(pTHX_ const U8 *p)
b8c5462f 1710{
97aff369 1711 dVAR;
7918f24d
NC
1712
1713 PERL_ARGS_ASSERT_IS_UTF8_PUNCT;
1714
d4c19fe8 1715 return is_utf8_common(p, &PL_utf8_punct, "IsPunct");
b8c5462f
JH
1716}
1717
1718bool
7fc63493 1719Perl_is_utf8_xdigit(pTHX_ const U8 *p)
b8c5462f 1720{
97aff369 1721 dVAR;
7918f24d
NC
1722
1723 PERL_ARGS_ASSERT_IS_UTF8_XDIGIT;
1724
d1eb3177 1725 return is_utf8_common(p, &PL_utf8_xdigit, "IsXDigit");
b8c5462f
JH
1726}
1727
1728bool
7fc63493 1729Perl_is_utf8_mark(pTHX_ const U8 *p)
a0ed51b3 1730{
97aff369 1731 dVAR;
7918f24d
NC
1732
1733 PERL_ARGS_ASSERT_IS_UTF8_MARK;
1734
d4c19fe8 1735 return is_utf8_common(p, &PL_utf8_mark, "IsM");
a0ed51b3
LW
1736}
1737
37e2e78e
KW
1738bool
1739Perl_is_utf8_X_begin(pTHX_ const U8 *p)
1740{
1741 dVAR;
1742
1743 PERL_ARGS_ASSERT_IS_UTF8_X_BEGIN;
1744
1745 return is_utf8_common(p, &PL_utf8_X_begin, "_X_Begin");
1746}
1747
1748bool
1749Perl_is_utf8_X_extend(pTHX_ const U8 *p)
1750{
1751 dVAR;
1752
1753 PERL_ARGS_ASSERT_IS_UTF8_X_EXTEND;
1754
1755 return is_utf8_common(p, &PL_utf8_X_extend, "_X_Extend");
1756}
1757
1758bool
1759Perl_is_utf8_X_prepend(pTHX_ const U8 *p)
1760{
1761 dVAR;
1762
1763 PERL_ARGS_ASSERT_IS_UTF8_X_PREPEND;
1764
1765 return is_utf8_common(p, &PL_utf8_X_prepend, "GCB=Prepend");
1766}
1767
1768bool
1769Perl_is_utf8_X_non_hangul(pTHX_ const U8 *p)
1770{
1771 dVAR;
1772
1773 PERL_ARGS_ASSERT_IS_UTF8_X_NON_HANGUL;
1774
1775 return is_utf8_common(p, &PL_utf8_X_non_hangul, "HST=Not_Applicable");
1776}
1777
1778bool
1779Perl_is_utf8_X_L(pTHX_ const U8 *p)
1780{
1781 dVAR;
1782
1783 PERL_ARGS_ASSERT_IS_UTF8_X_L;
1784
1785 return is_utf8_common(p, &PL_utf8_X_L, "GCB=L");
1786}
1787
1788bool
1789Perl_is_utf8_X_LV(pTHX_ const U8 *p)
1790{
1791 dVAR;
1792
1793 PERL_ARGS_ASSERT_IS_UTF8_X_LV;
1794
1795 return is_utf8_common(p, &PL_utf8_X_LV, "GCB=LV");
1796}
1797
1798bool
1799Perl_is_utf8_X_LVT(pTHX_ const U8 *p)
1800{
1801 dVAR;
1802
1803 PERL_ARGS_ASSERT_IS_UTF8_X_LVT;
1804
1805 return is_utf8_common(p, &PL_utf8_X_LVT, "GCB=LVT");
1806}
1807
1808bool
1809Perl_is_utf8_X_T(pTHX_ const U8 *p)
1810{
1811 dVAR;
1812
1813 PERL_ARGS_ASSERT_IS_UTF8_X_T;
1814
1815 return is_utf8_common(p, &PL_utf8_X_T, "GCB=T");
1816}
1817
1818bool
1819Perl_is_utf8_X_V(pTHX_ const U8 *p)
1820{
1821 dVAR;
1822
1823 PERL_ARGS_ASSERT_IS_UTF8_X_V;
1824
1825 return is_utf8_common(p, &PL_utf8_X_V, "GCB=V");
1826}
1827
1828bool
1829Perl_is_utf8_X_LV_LVT_V(pTHX_ const U8 *p)
1830{
1831 dVAR;
1832
1833 PERL_ARGS_ASSERT_IS_UTF8_X_LV_LVT_V;
1834
1835 return is_utf8_common(p, &PL_utf8_X_LV_LVT_V, "_X_LV_LVT_V");
1836}
1837
6b5c0936 1838/*
87cea99e 1839=for apidoc to_utf8_case
6b5c0936
JH
1840
1841The "p" contains the pointer to the UTF-8 string encoding
1842the character that is being converted.
1843
1844The "ustrp" is a pointer to the character buffer to put the
1845conversion result to. The "lenp" is a pointer to the length
1846of the result.
1847
0134edef 1848The "swashp" is a pointer to the swash to use.
6b5c0936 1849
36bb2ab6 1850Both the special and normal mappings are stored in lib/unicore/To/Foo.pl,
8fe4d5b2 1851and loaded by SWASHNEW, using lib/utf8_heavy.pl. The special (usually,
0134edef 1852but not always, a multicharacter mapping), is tried first.
6b5c0936 1853
0134edef
JH
1854The "special" is a string like "utf8::ToSpecLower", which means the
1855hash %utf8::ToSpecLower. The access to the hash is through
1856Perl_to_utf8_case().
6b5c0936 1857
0134edef
JH
1858The "normal" is a string like "ToLower" which means the swash
1859%utf8::ToLower.
1860
1861=cut */
6b5c0936 1862
2104c8d9 1863UV
9a957fbc
AL
1864Perl_to_utf8_case(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp,
1865 SV **swashp, const char *normal, const char *special)
a0ed51b3 1866{
97aff369 1867 dVAR;
89ebb4a3 1868 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
0134edef 1869 STRLEN len = 0;
aec46f14 1870 const UV uv0 = utf8_to_uvchr(p, NULL);
1feea2c7
JH
1871 /* The NATIVE_TO_UNI() and UNI_TO_NATIVE() mappings
1872 * are necessary in EBCDIC, they are redundant no-ops
1873 * in ASCII-ish platforms, and hopefully optimized away. */
f54cb97a 1874 const UV uv1 = NATIVE_TO_UNI(uv0);
7918f24d
NC
1875
1876 PERL_ARGS_ASSERT_TO_UTF8_CASE;
1877
9ae3ac1a
KW
1878 /* Note that swash_fetch() doesn't output warnings for these because it
1879 * assumes we will */
8457b38f 1880 if (uv1 >= UNICODE_SURROGATE_FIRST) {
9ae3ac1a 1881 if (uv1 <= UNICODE_SURROGATE_LAST) {
8457b38f
KW
1882 if (ckWARN_d(WARN_SURROGATE)) {
1883 const char* desc = (PL_op) ? OP_DESC(PL_op) : normal;
1884 Perl_warner(aTHX_ packWARN(WARN_SURROGATE),
1885 "Operation \"%s\" returns its argument for UTF-16 surrogate U+%04"UVXf"", desc, uv1);
1886 }
9ae3ac1a
KW
1887 }
1888 else if (UNICODE_IS_SUPER(uv1)) {
8457b38f
KW
1889 if (ckWARN_d(WARN_NON_UNICODE)) {
1890 const char* desc = (PL_op) ? OP_DESC(PL_op) : normal;
1891 Perl_warner(aTHX_ packWARN(WARN_NON_UNICODE),
1892 "Operation \"%s\" returns its argument for non-Unicode code point 0x%04"UVXf"", desc, uv1);
1893 }
9ae3ac1a
KW
1894 }
1895
1896 /* Note that non-characters are perfectly legal, so no warning should
1897 * be given */
1898 }
1899
1feea2c7 1900 uvuni_to_utf8(tmpbuf, uv1);
0134edef
JH
1901
1902 if (!*swashp) /* load on-demand */
1903 *swashp = swash_init("utf8", normal, &PL_sv_undef, 4, 0);
1904
a6f87d8c 1905 if (special) {
0134edef 1906 /* It might be "special" (sometimes, but not always,
2a37f04d 1907 * a multicharacter mapping) */
6673a63c 1908 HV * const hv = get_hv(special, 0);
b08cf34e
JH
1909 SV **svp;
1910
35da51f7 1911 if (hv &&
b08cf34e
JH
1912 (svp = hv_fetch(hv, (const char*)tmpbuf, UNISKIP(uv1), FALSE)) &&
1913 (*svp)) {
cfd0369c 1914 const char *s;
47654450 1915
cfd0369c 1916 s = SvPV_const(*svp, len);
47654450
JH
1917 if (len == 1)
1918 len = uvuni_to_utf8(ustrp, NATIVE_TO_UNI(*(U8*)s)) - ustrp;
2a37f04d 1919 else {
2f9475ad
JH
1920#ifdef EBCDIC
1921 /* If we have EBCDIC we need to remap the characters
1922 * since any characters in the low 256 are Unicode
1923 * code points, not EBCDIC. */
7cda7a3d 1924 U8 *t = (U8*)s, *tend = t + len, *d;
2f9475ad
JH
1925
1926 d = tmpbuf;
b08cf34e 1927 if (SvUTF8(*svp)) {
2f9475ad
JH
1928 STRLEN tlen = 0;
1929
1930 while (t < tend) {
d4c19fe8 1931 const UV c = utf8_to_uvchr(t, &tlen);
2f9475ad
JH
1932 if (tlen > 0) {
1933 d = uvchr_to_utf8(d, UNI_TO_NATIVE(c));
1934 t += tlen;
1935 }
1936 else
1937 break;
1938 }
1939 }
1940 else {
36fec512
JH
1941 while (t < tend) {
1942 d = uvchr_to_utf8(d, UNI_TO_NATIVE(*t));
1943 t++;
1944 }
2f9475ad
JH
1945 }
1946 len = d - tmpbuf;
1947 Copy(tmpbuf, ustrp, len, U8);
1948#else
d2dcd0fb 1949 Copy(s, ustrp, len, U8);
2f9475ad 1950#endif
29e98929 1951 }
983ffd37 1952 }
0134edef
JH
1953 }
1954
1955 if (!len && *swashp) {
d4c19fe8
AL
1956 const UV uv2 = swash_fetch(*swashp, tmpbuf, TRUE);
1957
0134edef
JH
1958 if (uv2) {
1959 /* It was "normal" (a single character mapping). */
d4c19fe8 1960 const UV uv3 = UNI_TO_NATIVE(uv2);
e9101d72 1961 len = uvchr_to_utf8(ustrp, uv3) - ustrp;
2a37f04d
JH
1962 }
1963 }
1feea2c7 1964
37e2e78e
KW
1965 if (!len) /* Neither: just copy. In other words, there was no mapping
1966 defined, which means that the code point maps to itself */
0134edef
JH
1967 len = uvchr_to_utf8(ustrp, uv0) - ustrp;
1968
2a37f04d
JH
1969 if (lenp)
1970 *lenp = len;
1971
0134edef 1972 return len ? utf8_to_uvchr(ustrp, 0) : 0;
a0ed51b3
LW
1973}
1974
d3e79532 1975/*
87cea99e 1976=for apidoc to_utf8_upper
d3e79532
JH
1977
1978Convert the UTF-8 encoded character at p to its uppercase version and
1979store that in UTF-8 in ustrp and its length in bytes in lenp. Note
89ebb4a3
JH
1980that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since
1981the uppercase version may be longer than the original character.
d3e79532
JH
1982
1983The first character of the uppercased version is returned
1984(but note, as explained above, that there may be more.)
1985
1986=cut */
1987
2104c8d9 1988UV
7fc63493 1989Perl_to_utf8_upper(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
a0ed51b3 1990{
97aff369 1991 dVAR;
7918f24d
NC
1992
1993 PERL_ARGS_ASSERT_TO_UTF8_UPPER;
1994
983ffd37 1995 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
7dd21db7 1996 &PL_utf8_toupper, "Touc", "utf8::ToSpecUpper");
983ffd37 1997}
a0ed51b3 1998
d3e79532 1999/*
87cea99e 2000=for apidoc to_utf8_title
d3e79532
JH
2001
2002Convert the UTF-8 encoded character at p to its titlecase version and
2003store that in UTF-8 in ustrp and its length in bytes in lenp. Note
89ebb4a3
JH
2004that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
2005titlecase version may be longer than the original character.
d3e79532
JH
2006
2007The first character of the titlecased version is returned
2008(but note, as explained above, that there may be more.)
2009
2010=cut */
2011
983ffd37 2012UV
7fc63493 2013Perl_to_utf8_title(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
983ffd37 2014{
97aff369 2015 dVAR;
7918f24d
NC
2016
2017 PERL_ARGS_ASSERT_TO_UTF8_TITLE;
2018
983ffd37 2019 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
7dd21db7 2020 &PL_utf8_totitle, "ToTc", "utf8::ToSpecTitle");
a0ed51b3
LW
2021}
2022
d3e79532 2023/*
87cea99e 2024=for apidoc to_utf8_lower
d3e79532
JH
2025
2026Convert the UTF-8 encoded character at p to its lowercase version and
2027store that in UTF-8 in ustrp and its length in bytes in lenp. Note
89ebb4a3
JH
2028that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
2029lowercase version may be longer than the original character.
d3e79532
JH
2030
2031The first character of the lowercased version is returned
2032(but note, as explained above, that there may be more.)
2033
2034=cut */
2035
2104c8d9 2036UV
7fc63493 2037Perl_to_utf8_lower(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
a0ed51b3 2038{
97aff369 2039 dVAR;
7918f24d
NC
2040
2041 PERL_ARGS_ASSERT_TO_UTF8_LOWER;
2042
983ffd37 2043 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
7dd21db7 2044 &PL_utf8_tolower, "Tolc", "utf8::ToSpecLower");
b4e400f9
JH
2045}
2046
d3e79532 2047/*
87cea99e 2048=for apidoc to_utf8_fold
d3e79532
JH
2049
2050Convert the UTF-8 encoded character at p to its foldcase version and
2051store that in UTF-8 in ustrp and its length in bytes in lenp. Note
89ebb4a3 2052that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
d3e79532
JH
2053foldcase version may be longer than the original character (up to
2054three characters).
2055
2056The first character of the foldcased version is returned
2057(but note, as explained above, that there may be more.)
2058
2059=cut */
2060
36bb2ab6
KW
2061/* Not currently externally documented is 'flags', which currently is non-zero
2062 * if full case folds are to be used; otherwise simple folds */
2063
b4e400f9 2064UV
36bb2ab6 2065Perl__to_utf8_fold_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, U8 flags)
b4e400f9 2066{
36bb2ab6
KW
2067 const char *specials = (flags) ? "utf8::ToSpecFold" : NULL;
2068
97aff369 2069 dVAR;
7918f24d 2070
36bb2ab6 2071 PERL_ARGS_ASSERT__TO_UTF8_FOLD_FLAGS;
7918f24d 2072
b4e400f9 2073 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
7dd21db7 2074 &PL_utf8_tofold, "Tocf", specials);
a0ed51b3
LW
2075}
2076
711a919c
ST
2077/* Note:
2078 * A "swash" is a swatch hash.
2079 * A "swatch" is a bit vector generated by utf8.c:S_swash_get().
2080 * C<pkg> is a pointer to a package name for SWASHNEW, should be "utf8".
2081 * For other parameters, see utf8::SWASHNEW in lib/utf8_heavy.pl.
2082 */
a0ed51b3 2083SV*
7fc63493 2084Perl_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv, I32 minbits, I32 none)
a0ed51b3 2085{
27da23d5 2086 dVAR;
a0ed51b3 2087 SV* retval;
8e84507e 2088 dSP;
7fc63493
AL
2089 const size_t pkg_len = strlen(pkg);
2090 const size_t name_len = strlen(name);
da51bb9b 2091 HV * const stash = gv_stashpvn(pkg, pkg_len, 0);
f8be5cf0 2092 SV* errsv_save;
88c57d91 2093 GV *method;
ce3b816e 2094
7918f24d
NC
2095 PERL_ARGS_ASSERT_SWASH_INIT;
2096
96ca9f55
DM
2097 PUSHSTACKi(PERLSI_MAGIC);
2098 ENTER;
ec34a119 2099 SAVEHINTS();
96ca9f55 2100 save_re_context();
ba6ff154
FC
2101 if (PL_parser && PL_parser->error_count)
2102 SAVEI8(PL_parser->error_count), PL_parser->error_count = 0;
88c57d91
NC
2103 method = gv_fetchmeth(stash, "SWASHNEW", 8, -1);
2104 if (!method) { /* demand load utf8 */
ce3b816e 2105 ENTER;
f8be5cf0 2106 errsv_save = newSVsv(ERRSV);
dc0c6abb
NC
2107 /* It is assumed that callers of this routine are not passing in any
2108 user derived data. */
2109 /* Need to do this after save_re_context() as it will set PL_tainted to
2110 1 while saving $1 etc (see the code after getrx: in Perl_magic_get).
2111 Even line to create errsv_save can turn on PL_tainted. */
2112 SAVEBOOL(PL_tainted);
2113 PL_tainted = 0;
71bed85a 2114 Perl_load_module(aTHX_ PERL_LOADMOD_NOIMPORT, newSVpvn(pkg,pkg_len),
a0714e2c 2115 NULL);
f8be5cf0
JH
2116 if (!SvTRUE(ERRSV))
2117 sv_setsv(ERRSV, errsv_save);
2118 SvREFCNT_dec(errsv_save);
ce3b816e
GS
2119 LEAVE;
2120 }
2121 SPAGAIN;
a0ed51b3
LW
2122 PUSHMARK(SP);
2123 EXTEND(SP,5);
6e449a3a
MHM
2124 mPUSHp(pkg, pkg_len);
2125 mPUSHp(name, name_len);
a0ed51b3 2126 PUSHs(listsv);
6e449a3a
MHM
2127 mPUSHi(minbits);
2128 mPUSHi(none);
a0ed51b3 2129 PUTBACK;
f8be5cf0 2130 errsv_save = newSVsv(ERRSV);
88c57d91
NC
2131 /* If we already have a pointer to the method, no need to use call_method()
2132 to repeat the lookup. */
2133 if (method ? call_sv(MUTABLE_SV(method), G_SCALAR)
8b563eef 2134 : call_sv(newSVpvs_flags("SWASHNEW", SVs_TEMP), G_SCALAR | G_METHOD))
8e84507e 2135 retval = newSVsv(*PL_stack_sp--);
a0ed51b3 2136 else
e24b16f9 2137 retval = &PL_sv_undef;
f8be5cf0
JH
2138 if (!SvTRUE(ERRSV))
2139 sv_setsv(ERRSV, errsv_save);
2140 SvREFCNT_dec(errsv_save);
a0ed51b3
LW
2141 LEAVE;
2142 POPSTACK;
923e4eb5 2143 if (IN_PERL_COMPILETIME) {
623e6609 2144 CopHINTS_set(PL_curcop, PL_hints);
a0ed51b3 2145 }
bc45ce41
JH
2146 if (!SvROK(retval) || SvTYPE(SvRV(retval)) != SVt_PVHV) {
2147 if (SvPOK(retval))
35c1215d 2148 Perl_croak(aTHX_ "Can't find Unicode property definition \"%"SVf"\"",
be2597df 2149 SVfARG(retval));
cea2e8a9 2150 Perl_croak(aTHX_ "SWASHNEW didn't return an HV ref");
bc45ce41 2151 }
a0ed51b3
LW
2152 return retval;
2153}
2154
035d37be
JH
2155
2156/* This API is wrong for special case conversions since we may need to
2157 * return several Unicode characters for a single Unicode character
2158 * (see lib/unicore/SpecCase.txt) The SWASHGET in lib/utf8_heavy.pl is
2159 * the lower-level routine, and it is similarly broken for returning
38684baa
KW
2160 * multiple values. --jhi
2161 * For those, you should use to_utf8_case() instead */
979f2922 2162/* Now SWASHGET is recasted into S_swash_get in this file. */
680c470c
ST
2163
2164/* Note:
2165 * Returns the value of property/mapping C<swash> for the first character
2166 * of the string C<ptr>. If C<do_utf8> is true, the string C<ptr> is
2167 * assumed to be in utf8. If C<do_utf8> is false, the string C<ptr> is
2168 * assumed to be in native 8-bit encoding. Caches the swatch in C<swash>.
2169 */
a0ed51b3 2170UV
680c470c 2171Perl_swash_fetch(pTHX_ SV *swash, const U8 *ptr, bool do_utf8)
a0ed51b3 2172{
27da23d5 2173 dVAR;
ef8f7699 2174 HV *const hv = MUTABLE_HV(SvRV(swash));
3568d838
JH
2175 U32 klen;
2176 U32 off;
a0ed51b3 2177 STRLEN slen;
7d85a32c 2178 STRLEN needents;
cfd0369c 2179 const U8 *tmps = NULL;
a0ed51b3 2180 U32 bit;
979f2922 2181 SV *swatch;
3568d838 2182 U8 tmputf8[2];
35da51f7 2183 const UV c = NATIVE_TO_ASCII(*ptr);
3568d838 2184
7918f24d
NC
2185 PERL_ARGS_ASSERT_SWASH_FETCH;
2186
3568d838 2187 if (!do_utf8 && !UNI_IS_INVARIANT(c)) {
979f2922
ST
2188 tmputf8[0] = (U8)UTF8_EIGHT_BIT_HI(c);
2189 tmputf8[1] = (U8)UTF8_EIGHT_BIT_LO(c);
2190 ptr = tmputf8;
3568d838
JH
2191 }
2192 /* Given a UTF-X encoded char 0xAA..0xYY,0xZZ
37e2e78e 2193 * then the "swatch" is a vec() for all the chars which start
3568d838
JH
2194 * with 0xAA..0xYY
2195 * So the key in the hash (klen) is length of encoded char -1
2196 */
2197 klen = UTF8SKIP(ptr) - 1;
2198 off = ptr[klen];
a0ed51b3 2199
979f2922 2200 if (klen == 0) {
37e2e78e 2201 /* If char is invariant then swatch is for all the invariant chars
1e54db1a 2202 * In both UTF-8 and UTF-8-MOD that happens to be UTF_CONTINUATION_MARK
7d85a32c 2203 */
979f2922
ST
2204 needents = UTF_CONTINUATION_MARK;
2205 off = NATIVE_TO_UTF(ptr[klen]);
2206 }
2207 else {
7d85a32c 2208 /* If char is encoded then swatch is for the prefix */
979f2922
ST
2209 needents = (1 << UTF_ACCUMULATION_SHIFT);
2210 off = NATIVE_TO_UTF(ptr[klen]) & UTF_CONTINUATION_MASK;
8457b38f 2211 if (UTF8_IS_SUPER(ptr) && ckWARN_d(WARN_NON_UNICODE)) {
9ae3ac1a
KW
2212 const UV code_point = utf8n_to_uvuni(ptr, UTF8_MAXBYTES, 0, 0);
2213
2214 /* This outputs warnings for binary properties only, assuming that
9479a769
KW
2215 * to_utf8_case() will output any for non-binary. Also, surrogates
2216 * aren't checked for, as that would warn on things like
2217 * /\p{Gc=Cs}/ */
9ae3ac1a
KW
2218 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
2219 if (SvUV(*bitssvp) == 1) {
8457b38f 2220 Perl_warner(aTHX_ packWARN(WARN_NON_UNICODE),
c634fdd3 2221 "Code point 0x%04"UVXf" is not Unicode, all \\p{} matches fail; all \\P{} matches succeed", code_point);
9ae3ac1a
KW
2222 }
2223 }
979f2922 2224 }
7d85a32c 2225
a0ed51b3
LW
2226 /*
2227 * This single-entry cache saves about 1/3 of the utf8 overhead in test
2228 * suite. (That is, only 7-8% overall over just a hash cache. Still,
2229 * it's nothing to sniff at.) Pity we usually come through at least
2230 * two function calls to get here...
2231 *
2232 * NB: this code assumes that swatches are never modified, once generated!
2233 */
2234
3568d838 2235 if (hv == PL_last_swash_hv &&
a0ed51b3 2236 klen == PL_last_swash_klen &&
27da23d5 2237 (!klen || memEQ((char *)ptr, (char *)PL_last_swash_key, klen)) )
a0ed51b3
LW
2238 {
2239 tmps = PL_last_swash_tmps;
2240 slen = PL_last_swash_slen;
2241 }
2242 else {
2243 /* Try our second-level swatch cache, kept in a hash. */
e1ec3a88 2244 SV** svp = hv_fetch(hv, (const char*)ptr, klen, FALSE);
a0ed51b3 2245
979f2922
ST
2246 /* If not cached, generate it via swash_get */
2247 if (!svp || !SvPOK(*svp)
2248 || !(tmps = (const U8*)SvPV_const(*svp, slen))) {
2b9d42f0
NIS
2249 /* We use utf8n_to_uvuni() as we want an index into
2250 Unicode tables, not a native character number.
2251 */
aec46f14 2252 const UV code_point = utf8n_to_uvuni(ptr, UTF8_MAXBYTES, 0,
872c91ae
JH
2253 ckWARN(WARN_UTF8) ?
2254 0 : UTF8_ALLOW_ANY);
680c470c 2255 swatch = swash_get(swash,
979f2922
ST
2256 /* On EBCDIC & ~(0xA0-1) isn't a useful thing to do */
2257 (klen) ? (code_point & ~(needents - 1)) : 0,
2258 needents);
2259
923e4eb5 2260 if (IN_PERL_COMPILETIME)
623e6609 2261 CopHINTS_set(PL_curcop, PL_hints);
a0ed51b3 2262
979f2922 2263 svp = hv_store(hv, (const char *)ptr, klen, swatch, 0);
a0ed51b3 2264
979f2922
ST
2265 if (!svp || !(tmps = (U8*)SvPV(*svp, slen))
2266 || (slen << 3) < needents)
660a4616 2267 Perl_croak(aTHX_ "panic: swash_fetch got improper swatch");
a0ed51b3
LW
2268 }
2269
2270 PL_last_swash_hv = hv;
16d8f38a 2271 assert(klen <= sizeof(PL_last_swash_key));
eac04b2e 2272 PL_last_swash_klen = (U8)klen;
cfd0369c
NC
2273 /* FIXME change interpvar.h? */
2274 PL_last_swash_tmps = (U8 *) tmps;
a0ed51b3
LW
2275 PL_last_swash_slen = slen;
2276 if (klen)
2277 Copy(ptr, PL_last_swash_key, klen, U8);
2278 }
2279
9faf8d75 2280 switch ((int)((slen << 3) / needents)) {
a0ed51b3
LW
2281 case 1:
2282 bit = 1 << (off & 7);
2283 off >>= 3;
2284 return (tmps[off] & bit) != 0;
2285 case 8:
2286 return tmps[off];
2287 case 16:
2288 off <<= 1;
2289 return (tmps[off] << 8) + tmps[off + 1] ;
2290 case 32:
2291 off <<= 2;
2292 return (tmps[off] << 24) + (tmps[off+1] << 16) + (tmps[off+2] << 8) + tmps[off + 3] ;
2293 }
660a4616 2294 Perl_croak(aTHX_ "panic: swash_fetch got swatch of unexpected bit width");
670f1322 2295 NORETURN_FUNCTION_END;
a0ed51b3 2296}
2b9d42f0 2297
319009ee
KW
2298/* Read a single line of the main body of the swash input text. These are of
2299 * the form:
2300 * 0053 0056 0073
2301 * where each number is hex. The first two numbers form the minimum and
2302 * maximum of a range, and the third is the value associated with the range.
2303 * Not all swashes should have a third number
2304 *
2305 * On input: l points to the beginning of the line to be examined; it points
2306 * to somewhere in the string of the whole input text, and is
2307 * terminated by a \n or the null string terminator.
2308 * lend points to the null terminator of that string
2309 * wants_value is non-zero if the swash expects a third number
2310 * typestr is the name of the swash's mapping, like 'ToLower'
2311 * On output: *min, *max, and *val are set to the values read from the line.
2312 * returns a pointer just beyond the line examined. If there was no
2313 * valid min number on the line, returns lend+1
2314 */
2315
2316STATIC U8*
2317S_swash_scan_list_line(pTHX_ U8* l, U8* const lend, UV* min, UV* max, UV* val,
2318 const bool wants_value, const U8* const typestr)
2319{
2320 const int typeto = typestr[0] == 'T' && typestr[1] == 'o';
2321 STRLEN numlen; /* Length of the number */
02470786
KW
2322 I32 flags = PERL_SCAN_SILENT_ILLDIGIT
2323 | PERL_SCAN_DISALLOW_PREFIX
2324 | PERL_SCAN_SILENT_NON_PORTABLE;
319009ee
KW
2325
2326 /* nl points to the next \n in the scan */
2327 U8* const nl = (U8*)memchr(l, '\n', lend - l);
2328
2329 /* Get the first number on the line: the range minimum */
2330 numlen = lend - l;
2331 *min = grok_hex((char *)l, &numlen, &flags, NULL);
2332 if (numlen) /* If found a hex number, position past it */
2333 l += numlen;
2334 else if (nl) { /* Else, go handle next line, if any */
2335 return nl + 1; /* 1 is length of "\n" */
2336 }
2337 else { /* Else, no next line */
2338 return lend + 1; /* to LIST's end at which \n is not found */
2339 }
2340
2341 /* The max range value follows, separated by a BLANK */
2342 if (isBLANK(*l)) {
2343 ++l;
02470786
KW
2344 flags = PERL_SCAN_SILENT_ILLDIGIT
2345 | PERL_SCAN_DISALLOW_PREFIX
2346 | PERL_SCAN_SILENT_NON_PORTABLE;
319009ee
KW
2347 numlen = lend - l;
2348 *max = grok_hex((char *)l, &numlen, &flags, NULL);
2349 if (numlen)
2350 l += numlen;
2351 else /* If no value here, it is a single element range */
2352 *max = *min;
2353
2354 /* Non-binary tables have a third entry: what the first element of the
2355 * range maps to */
2356 if (wants_value) {
2357 if (isBLANK(*l)) {
2358 ++l;
02470786
KW
2359 flags = PERL_SCAN_SILENT_ILLDIGIT
2360 | PERL_SCAN_DISALLOW_PREFIX
2361 | PERL_SCAN_SILENT_NON_PORTABLE;
319009ee
KW
2362 numlen = lend - l;
2363 *val = grok_hex((char *)l, &numlen, &flags, NULL);
2364 if (numlen)
2365 l += numlen;
2366 else
2367 *val = 0;
2368 }
2369 else {
2370 *val = 0;
2371 if (typeto) {
2372 Perl_croak(aTHX_ "%s: illegal mapping '%s'",
2373 typestr, l);
2374 }
2375 }
2376 }
2377 else
2378 *val = 0; /* bits == 1, then any val should be ignored */
2379 }
2380 else { /* Nothing following range min, should be single element with no
2381 mapping expected */
2382 *max = *min;
2383 if (wants_value) {
2384 *val = 0;
2385 if (typeto) {
2386 Perl_croak(aTHX_ "%s: illegal mapping '%s'", typestr, l);
2387 }
2388 }
2389 else
2390 *val = 0; /* bits == 1, then val should be ignored */
2391 }
2392
2393 /* Position to next line if any, or EOF */
2394 if (nl)
2395 l = nl + 1;
2396 else
2397 l = lend;
2398
2399 return l;
2400}
2401
979f2922
ST
2402/* Note:
2403 * Returns a swatch (a bit vector string) for a code point sequence
2404 * that starts from the value C<start> and comprises the number C<span>.
2405 * A C<swash> must be an object created by SWASHNEW (see lib/utf8_heavy.pl).
2406 * Should be used via swash_fetch, which will cache the swatch in C<swash>.
2407 */
2408STATIC SV*
2409S_swash_get(pTHX_ SV* swash, UV start, UV span)
2410{
2411 SV *swatch;
77f9f126 2412 U8 *l, *lend, *x, *xend, *s, *send;
979f2922 2413 STRLEN lcur, xcur, scur;
ef8f7699 2414 HV *const hv = MUTABLE_HV(SvRV(swash));
972dd592
KW
2415
2416 /* The string containing the main body of the table */
017a3ce5 2417 SV** const listsvp = hv_fetchs(hv, "LIST", FALSE);
972dd592 2418
017a3ce5
GA
2419 SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE);
2420 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
2421 SV** const nonesvp = hv_fetchs(hv, "NONE", FALSE);
2422 SV** const extssvp = hv_fetchs(hv, "EXTRAS", FALSE);
77f9f126 2423 SV** const invert_it_svp = hv_fetchs(hv, "INVERT_IT", FALSE);
0bd48802 2424 const U8* const typestr = (U8*)SvPV_nolen(*typesvp);
0bd48802
AL
2425 const STRLEN bits = SvUV(*bitssvp);
2426 const STRLEN octets = bits >> 3; /* if bits == 1, then octets == 0 */
2427 const UV none = SvUV(*nonesvp);
2428 const UV end = start + span;
979f2922 2429
7918f24d
NC
2430 PERL_ARGS_ASSERT_SWASH_GET;
2431
979f2922 2432 if (bits != 1 && bits != 8 && bits != 16 && bits != 32) {
660a4616
ST
2433 Perl_croak(aTHX_ "panic: swash_get doesn't expect bits %"UVuf,
2434 (UV)bits);
979f2922
ST
2435 }
2436
2437 /* create and initialize $swatch */
979f2922 2438 scur = octets ? (span * octets) : (span + 7) / 8;
e524fe40
NC
2439 swatch = newSV(scur);
2440 SvPOK_on(swatch);
979f2922
ST
2441 s = (U8*)SvPVX(swatch);
2442 if (octets && none) {
0bd48802 2443 const U8* const e = s + scur;
979f2922
ST
2444 while (s < e) {
2445 if (bits == 8)
2446 *s++ = (U8)(none & 0xff);
2447 else if (bits == 16) {
2448 *s++ = (U8)((none >> 8) & 0xff);
2449 *s++ = (U8)( none & 0xff);
2450 }
2451 else if (bits == 32) {
2452 *s++ = (U8)((none >> 24) & 0xff);
2453 *s++ = (U8)((none >> 16) & 0xff);
2454 *s++ = (U8)((none >> 8) & 0xff);
2455 *s++ = (U8)( none & 0xff);
2456 }
2457 }
2458 *s = '\0';
2459 }
2460 else {
2461 (void)memzero((U8*)s, scur + 1);
2462 }
2463 SvCUR_set(swatch, scur);
2464 s = (U8*)SvPVX(swatch);
2465
2466 /* read $swash->{LIST} */
2467 l = (U8*)SvPV(*listsvp, lcur);
2468 lend = l + lcur;
2469 while (l < lend) {
35da51f7 2470 UV min, max, val;
319009ee
KW
2471 l = S_swash_scan_list_line(aTHX_ l, lend, &min, &max, &val,
2472 cBOOL(octets), typestr);
2473 if (l > lend) {
979f2922
ST
2474 break;
2475 }
2476
972dd592 2477 /* If looking for something beyond this range, go try the next one */
979f2922
ST
2478 if (max < start)
2479 continue;
2480
2481 if (octets) {
35da51f7 2482 UV key;
979f2922
ST
2483 if (min < start) {
2484 if (!none || val < none) {
2485 val += start - min;
2486 }
2487 min = start;
2488 }
2489 for (key = min; key <= max; key++) {
2490 STRLEN offset;
2491 if (key >= end)
2492 goto go_out_list;
2493 /* offset must be non-negative (start <= min <= key < end) */
2494 offset = octets * (key - start);
2495 if (bits == 8)
2496 s[offset] = (U8)(val & 0xff);
2497 else if (bits == 16) {
2498 s[offset ] = (U8)((val >> 8) & 0xff);
2499 s[offset + 1] = (U8)( val & 0xff);
2500 }
2501 else if (bits == 32) {
2502 s[offset ] = (U8)((val >> 24) & 0xff);
2503 s[offset + 1] = (U8)((val >> 16) & 0xff);
2504 s[offset + 2] = (U8)((val >> 8) & 0xff);
2505 s[offset + 3] = (U8)( val & 0xff);
2506 }
2507
2508 if (!none || val < none)
2509 ++val;
2510 }
2511 }
711a919c 2512 else { /* bits == 1, then val should be ignored */
35da51f7 2513 UV key;
979f2922
ST
2514 if (min < start)
2515 min = start;
2516 for (key = min; key <= max; key++) {
0bd48802 2517 const STRLEN offset = (STRLEN)(key - start);
979f2922
ST
2518 if (key >= end)
2519 goto go_out_list;
2520 s[offset >> 3] |= 1 << (offset & 7);
2521 }
2522 }
2523 } /* while */
2524 go_out_list:
2525
9479a769 2526 /* Invert if the data says it should be. Assumes that bits == 1 */
77f9f126 2527 if (invert_it_svp && SvUV(*invert_it_svp)) {
0bda3001
KW
2528
2529 /* Unicode properties should come with all bits above PERL_UNICODE_MAX
2530 * be 0, and their inversion should also be 0, as we don't succeed any
2531 * Unicode property matches for non-Unicode code points */
2532 if (start <= PERL_UNICODE_MAX) {
2533
2534 /* The code below assumes that we never cross the
2535 * Unicode/above-Unicode boundary in a range, as otherwise we would
2536 * have to figure out where to stop flipping the bits. Since this
2537 * boundary is divisible by a large power of 2, and swatches comes
2538 * in small powers of 2, this should be a valid assumption */
2539 assert(start + span - 1 <= PERL_UNICODE_MAX);
2540
507a8485
KW
2541 send = s + scur;
2542 while (s < send) {
2543 *s = ~(*s);
2544 s++;
2545 }
0bda3001 2546 }
77f9f126
KW
2547 }
2548
d73c39c5
KW
2549 /* read $swash->{EXTRAS}
2550 * This code also copied to swash_to_invlist() below */
979f2922
ST
2551 x = (U8*)SvPV(*extssvp, xcur);
2552 xend = x + xcur;
2553 while (x < xend) {
2554 STRLEN namelen;
2555 U8 *namestr;
2556 SV** othersvp;
2557 HV* otherhv;
2558 STRLEN otherbits;
2559 SV **otherbitssvp, *other;
711a919c 2560 U8 *s, *o, *nl;
979f2922
ST
2561 STRLEN slen, olen;
2562
35da51f7 2563 const U8 opc = *x++;
979f2922
ST
2564 if (opc == '\n')
2565 continue;
2566
2567 nl = (U8*)memchr(x, '\n', xend - x);
2568
2569 if (opc != '-' && opc != '+' && opc != '!' && opc != '&') {
2570 if (nl) {
2571 x = nl + 1; /* 1 is length of "\n" */
2572 continue;
2573 }
2574 else {
2575 x = xend; /* to EXTRAS' end at which \n is not found */
2576 break;
2577 }
2578 }
2579
2580 namestr = x;
2581 if (nl) {
2582 namelen = nl - namestr;
2583 x = nl + 1;
2584 }
2585 else {
2586 namelen = xend - namestr;
2587 x = xend;
2588 }
2589
2590 othersvp = hv_fetch(hv, (char *)namestr, namelen, FALSE);
ef8f7699 2591 otherhv = MUTABLE_HV(SvRV(*othersvp));
017a3ce5 2592 otherbitssvp = hv_fetchs(otherhv, "BITS", FALSE);
979f2922
ST
2593 otherbits = (STRLEN)SvUV(*otherbitssvp);
2594 if (bits < otherbits)
660a4616 2595 Perl_croak(aTHX_ "panic: swash_get found swatch size mismatch");
979f2922
ST
2596
2597 /* The "other" swatch must be destroyed after. */
2598 other = swash_get(*othersvp, start, span);
2599 o = (U8*)SvPV(other, olen);
2600
2601 if (!olen)
660a4616 2602 Perl_croak(aTHX_ "panic: swash_get got improper swatch");
979f2922
ST
2603
2604 s = (U8*)SvPV(swatch, slen);
2605 if (bits == 1 && otherbits == 1) {
2606 if (slen != olen)
660a4616 2607 Perl_croak(aTHX_ "panic: swash_get found swatch length mismatch");
979f2922
ST
2608
2609 switch (opc) {
2610 case '+':
2611 while (slen--)
2612 *s++ |= *o++;
2613 break;
2614 case '!':
2615 while (slen--)
2616 *s++ |= ~*o++;
2617 break;
2618 case '-':
2619 while (slen--)
2620 *s++ &= ~*o++;
2621 break;
2622 case '&':
2623 while (slen--)
2624 *s++ &= *o++;
2625 break;
2626 default:
2627 break;
2628 }
2629 }
711a919c 2630 else {
979f2922
ST
2631 STRLEN otheroctets = otherbits >> 3;
2632 STRLEN offset = 0;
35da51f7 2633 U8* const send = s + slen;
979f2922
ST
2634
2635 while (s < send) {
2636 UV otherval = 0;
2637
2638 if (otherbits == 1) {
2639 otherval = (o[offset >> 3] >> (offset & 7)) & 1;
2640 ++offset;
2641 }
2642 else {
2643 STRLEN vlen = otheroctets;
2644 otherval = *o++;
2645 while (--vlen) {
2646 otherval <<= 8;
2647 otherval |= *o++;
2648 }
2649 }
2650
711a919c 2651 if (opc == '+' && otherval)
6f207bd3 2652 NOOP; /* replace with otherval */
979f2922
ST
2653 else if (opc == '!' && !otherval)
2654 otherval = 1;
2655 else if (opc == '-' && otherval)
2656 otherval = 0;
2657 else if (opc == '&' && !otherval)
2658 otherval = 0;
2659 else {
711a919c 2660 s += octets; /* no replacement */
979f2922
ST
2661 continue;
2662 }
2663
2664 if (bits == 8)
2665 *s++ = (U8)( otherval & 0xff);
2666 else if (bits == 16) {
2667 *s++ = (U8)((otherval >> 8) & 0xff);
2668 *s++ = (U8)( otherval & 0xff);
2669 }
2670 else if (bits == 32) {
2671 *s++ = (U8)((otherval >> 24) & 0xff);
2672 *s++ = (U8)((otherval >> 16) & 0xff);
2673 *s++ = (U8)((otherval >> 8) & 0xff);
2674 *s++ = (U8)( otherval & 0xff);
2675 }
2676 }
2677 }
2678 sv_free(other); /* through with it! */
2679 } /* while */
2680 return swatch;
2681}
2682
064c021d 2683HV*
4c2e1131 2684Perl__swash_inversion_hash(pTHX_ SV* const swash)
064c021d
KW
2685{
2686
5662e334
KW
2687 /* Subject to change or removal. For use only in one place in regcomp.c.
2688 * Can't be used on a property that is subject to user override, as it
2689 * relies on the value of SPECIALS in the swash which would be set by
2690 * utf8_heavy.pl to the hash in the non-overriden file, and hence is not set
2691 * for overridden properties
064c021d
KW
2692 *
2693 * Returns a hash which is the inversion and closure of a swash mapping.
2694 * For example, consider the input lines:
2695 * 004B 006B
2696 * 004C 006C
2697 * 212A 006B
2698 *
2699 * The returned hash would have two keys, the utf8 for 006B and the utf8 for
2700 * 006C. The value for each key is an array. For 006C, the array would
2701 * have a two elements, the utf8 for itself, and for 004C. For 006B, there
2702 * would be three elements in its array, the utf8 for 006B, 004B and 212A.
2703 *
2704 * Essentially, for any code point, it gives all the code points that map to
2705 * it, or the list of 'froms' for that point.
2706 *
5662e334
KW
2707 * Currently it ignores any additions or deletions from other swashes,
2708 * looking at just the main body of the swash, and if there are SPECIALS
2709 * in the swash, at that hash
2710 *
2711 * The specials hash can be extra code points, and most likely consists of
2712 * maps from single code points to multiple ones (each expressed as a string
2713 * of utf8 characters). This function currently returns only 1-1 mappings.
2714 * However consider this possible input in the specials hash:
2715 * "\xEF\xAC\x85" => "\x{0073}\x{0074}", # U+FB05 => 0073 0074
2716 * "\xEF\xAC\x86" => "\x{0073}\x{0074}", # U+FB06 => 0073 0074
2717 *
2718 * Both FB05 and FB06 map to the same multi-char sequence, which we don't
2719 * currently handle. But it also means that FB05 and FB06 are equivalent in
2720 * a 1-1 mapping which we should handle, and this relationship may not be in
2721 * the main table. Therefore this function examines all the multi-char
2722 * sequences and adds the 1-1 mappings that come out of that. */
064c021d
KW
2723
2724 U8 *l, *lend;
2725 STRLEN lcur;
2726 HV *const hv = MUTABLE_HV(SvRV(swash));
2727
2728 /* The string containing the main body of the table */
2729 SV** const listsvp = hv_fetchs(hv, "LIST", FALSE);
2730
2731 SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE);
2732 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
2733 SV** const nonesvp = hv_fetchs(hv, "NONE", FALSE);
2734 /*SV** const extssvp = hv_fetchs(hv, "EXTRAS", FALSE);*/
2735 const U8* const typestr = (U8*)SvPV_nolen(*typesvp);
2736 const STRLEN bits = SvUV(*bitssvp);
2737 const STRLEN octets = bits >> 3; /* if bits == 1, then octets == 0 */
2738 const UV none = SvUV(*nonesvp);
5662e334 2739 SV **specials_p = hv_fetchs(hv, "SPECIALS", 0);
064c021d
KW
2740
2741 HV* ret = newHV();
2742
2743 PERL_ARGS_ASSERT__SWASH_INVERSION_HASH;
2744
2745 /* Must have at least 8 bits to get the mappings */
2746 if (bits != 8 && bits != 16 && bits != 32) {
2747 Perl_croak(aTHX_ "panic: swash_inversion_hash doesn't expect bits %"UVuf,
2748 (UV)bits);
2749 }
2750
5662e334
KW
2751 if (specials_p) { /* It might be "special" (sometimes, but not always, a
2752 mapping to more than one character */
2753
2754 /* Construct an inverse mapping hash for the specials */
2755 HV * const specials_hv = MUTABLE_HV(SvRV(*specials_p));
2756 HV * specials_inverse = newHV();
2757 char *char_from; /* the lhs of the map */
2758 I32 from_len; /* its byte length */
2759 char *char_to; /* the rhs of the map */
2760 I32 to_len; /* its byte length */
2761 SV *sv_to; /* and in a sv */
2762 AV* from_list; /* list of things that map to each 'to' */
2763
2764 hv_iterinit(specials_hv);
2765
2766 /* The keys are the characters (in utf8) that map to the corresponding
2767 * utf8 string value. Iterate through the list creating the inverse
2768 * list. */
2769 while ((sv_to = hv_iternextsv(specials_hv, &char_from, &from_len))) {
2770 SV** listp;
2771 if (! SvPOK(sv_to)) {
2772 Perl_croak(aTHX_ "panic: value returned from hv_iternextsv() unexpectedly is not a string");
2773 }
2774 /*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)));*/
2775
2776 /* Each key in the inverse list is a mapped-to value, and the key's
2777 * hash value is a list of the strings (each in utf8) that map to
2778 * it. Those strings are all one character long */
2779 if ((listp = hv_fetch(specials_inverse,
2780 SvPVX(sv_to),
2781 SvCUR(sv_to), 0)))
2782 {
2783 from_list = (AV*) *listp;
2784 }
2785 else { /* No entry yet for it: create one */
2786 from_list = newAV();
2787 if (! hv_store(specials_inverse,
2788 SvPVX(sv_to),
2789 SvCUR(sv_to),
2790 (SV*) from_list, 0))
2791 {
2792 Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed");
2793 }
2794 }
2795
2796 /* Here have the list associated with this 'to' (perhaps newly
2797 * created and empty). Just add to it. Note that we ASSUME that
2798 * the input is guaranteed to not have duplications, so we don't
2799 * check for that. Duplications just slow down execution time. */
2800 av_push(from_list, newSVpvn_utf8(char_from, from_len, TRUE));
2801 }
2802
2803 /* Here, 'specials_inverse' contains the inverse mapping. Go through
2804 * it looking for cases like the FB05/FB06 examples above. There would
2805 * be an entry in the hash like
2806 * 'st' => [ FB05, FB06 ]
2807 * In this example we will create two lists that get stored in the
2808 * returned hash, 'ret':
2809 * FB05 => [ FB05, FB06 ]
2810 * FB06 => [ FB05, FB06 ]
2811 *
2812 * Note that there is nothing to do if the array only has one element.
2813 * (In the normal 1-1 case handled below, we don't have to worry about
2814 * two lists, as everything gets tied to the single list that is
2815 * generated for the single character 'to'. But here, we are omitting
2816 * that list, ('st' in the example), so must have multiple lists.) */
2817 while ((from_list = (AV *) hv_iternextsv(specials_inverse,
2818 &char_to, &to_len)))
2819 {
2820 if (av_len(from_list) > 0) {
2821 int i;
2822
2823 /* We iterate over all combinations of i,j to place each code
2824 * point on each list */
2825 for (i = 0; i <= av_len(from_list); i++) {
2826 int j;
2827 AV* i_list = newAV();
2828 SV** entryp = av_fetch(from_list, i, FALSE);
2829 if (entryp == NULL) {
2830 Perl_croak(aTHX_ "panic: av_fetch() unexpectedly failed");
2831 }
2832 if (hv_fetch(ret, SvPVX(*entryp), SvCUR(*entryp), FALSE)) {
2833 Perl_croak(aTHX_ "panic: unexpected entry for %s", SvPVX(*entryp));
2834 }
2835 if (! hv_store(ret, SvPVX(*entryp), SvCUR(*entryp),
2836 (SV*) i_list, FALSE))
2837 {
2838 Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed");
2839 }
2840
2841 /* For debugging: UV u = utf8_to_uvchr((U8*) SvPVX(*entryp), 0);*/
2842 for (j = 0; j <= av_len(from_list); j++) {
2843 entryp = av_fetch(from_list, j, FALSE);
2844 if (entryp == NULL) {
2845 Perl_croak(aTHX_ "panic: av_fetch() unexpectedly failed");
2846 }
2847
2848 /* When i==j this adds itself to the list */
2849 av_push(i_list, newSVuv(utf8_to_uvchr(
2850 (U8*) SvPVX(*entryp), 0)));
2851 /*DEBUG_U(PerlIO_printf(Perl_debug_log, "Adding %"UVXf" to list for %"UVXf"\n", utf8_to_uvchr((U8*) SvPVX(*entryp), 0), u));*/
2852 }
2853 }
2854 }
2855 }
2856 SvREFCNT_dec(specials_inverse); /* done with it */
2857 } /* End of specials */
2858
064c021d
KW
2859 /* read $swash->{LIST} */
2860 l = (U8*)SvPV(*listsvp, lcur);
2861 lend = l + lcur;
2862
2863 /* Go through each input line */
2864 while (l < lend) {
2865 UV min, max, val;
2866 UV inverse;
2867 l = S_swash_scan_list_line(aTHX_ l, lend, &min, &max, &val,
2868 cBOOL(octets), typestr);
2869 if (l > lend) {
2870 break;
2871 }
2872
2873 /* Each element in the range is to be inverted */
2874 for (inverse = min; inverse <= max; inverse++) {
2875 AV* list;
064c021d
KW
2876 SV** listp;
2877 IV i;
2878 bool found_key = FALSE;
5662e334 2879 bool found_inverse = FALSE;
064c021d
KW
2880
2881 /* The key is the inverse mapping */
2882 char key[UTF8_MAXBYTES+1];
2883 char* key_end = (char *) uvuni_to_utf8((U8*) key, val);
2884 STRLEN key_len = key_end - key;
2885
064c021d
KW
2886 /* Get the list for the map */
2887 if ((listp = hv_fetch(ret, key, key_len, FALSE))) {
2888 list = (AV*) *listp;
2889 }
2890 else { /* No entry yet for it: create one */
2891 list = newAV();
2892 if (! hv_store(ret, key, key_len, (SV*) list, FALSE)) {
2893 Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed");
2894 }
2895 }
2896
5662e334
KW
2897 /* Look through list to see if this inverse mapping already is
2898 * listed, or if there is a mapping to itself already */
508f7cfa 2899 for (i = 0; i <= av_len(list); i++) {
064c021d
KW
2900 SV** entryp = av_fetch(list, i, FALSE);
2901 SV* entry;
2902 if (entryp == NULL) {
2903 Perl_croak(aTHX_ "panic: av_fetch() unexpectedly failed");
2904 }
2905 entry = *entryp;
5662e334 2906 /*DEBUG_U(PerlIO_printf(Perl_debug_log, "list for %"UVXf" contains %"UVXf"\n", val, SvUV(entry)));*/
56ca34ca 2907 if (SvUV(entry) == val) {
064c021d 2908 found_key = TRUE;
5662e334
KW
2909 }
2910 if (SvUV(entry) == inverse) {
2911 found_inverse = TRUE;
2912 }
2913
2914 /* No need to continue searching if found everything we are
2915 * looking for */
2916 if (found_key && found_inverse) {
064c021d
KW
2917 break;
2918 }
2919 }
56ca34ca
KW
2920
2921 /* Make sure there is a mapping to itself on the list */
064c021d 2922 if (! found_key) {
d397ff6a 2923 av_push(list, newSVuv(val));
5662e334 2924 /*DEBUG_U(PerlIO_printf(Perl_debug_log, "Adding %"UVXf" to list for %"UVXf"\n", val, val));*/
064c021d
KW
2925 }
2926
2927
2928 /* Simply add the value to the list */
5662e334
KW
2929 if (! found_inverse) {
2930 av_push(list, newSVuv(inverse));
2931 /*DEBUG_U(PerlIO_printf(Perl_debug_log, "Adding %"UVXf" to list for %"UVXf"\n", inverse, val));*/
2932 }
064c021d
KW
2933
2934 /* swash_get() increments the value of val for each element in the
2935 * range. That makes more compact tables possible. You can
2936 * express the capitalization, for example, of all consecutive
2937 * letters with a single line: 0061\t007A\t0041 This maps 0061 to
2938 * 0041, 0062 to 0042, etc. I (khw) have never understood 'none',
bd3f2f94
KW
2939 * and it's not documented; it appears to be used only in
2940 * implementing tr//; I copied the semantics from swash_get(), just
2941 * in case */
064c021d
KW
2942 if (!none || val < none) {
2943 ++val;
2944 }
2945 }
2946 }
2947
2948 return ret;
2949}
2950
a25abddc 2951SV*
d764b54e
KW
2952Perl__swash_to_invlist(pTHX_ SV* const swash)
2953{
2954
2955 /* Subject to change or removal. For use only in one place in regcomp.c */
2956
2957 U8 *l, *lend;
2958 char *loc;
2959 STRLEN lcur;
2960 HV *const hv = MUTABLE_HV(SvRV(swash));
2961 UV elements = 0; /* Number of elements in the inversion list */
b443038a 2962 U8 empty[] = "";
d764b54e
KW
2963
2964 /* The string containing the main body of the table */
2965 SV** const listsvp = hv_fetchs(hv, "LIST", FALSE);
2966 SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE);
2967 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
d73c39c5 2968 SV** const extssvp = hv_fetchs(hv, "EXTRAS", FALSE);
77f9f126 2969 SV** const invert_it_svp = hv_fetchs(hv, "INVERT_IT", FALSE);
d764b54e
KW
2970
2971 const U8* const typestr = (U8*)SvPV_nolen(*typesvp);
2972 const STRLEN bits = SvUV(*bitssvp);
2973 const STRLEN octets = bits >> 3; /* if bits == 1, then octets == 0 */
d73c39c5
KW
2974 U8 *x, *xend;
2975 STRLEN xcur;
d764b54e 2976
a25abddc 2977 SV* invlist;
d764b54e
KW
2978
2979 PERL_ARGS_ASSERT__SWASH_TO_INVLIST;
2980
2981 /* read $swash->{LIST} */
b443038a
KW
2982 if (SvPOK(*listsvp)) {
2983 l = (U8*)SvPV(*listsvp, lcur);
2984 }
2985 else {
2986 /* LIST legitimately doesn't contain a string during compilation phases
2987 * of Perl itself, before the Unicode tables are generated. In this
2988 * case, just fake things up by creating an empty list */
2989 l = empty;
2990 lcur = 0;
2991 }
d764b54e
KW
2992 loc = (char *) l;
2993 lend = l + lcur;
2994
2995 /* Scan the input to count the number of lines to preallocate array size
2996 * based on worst possible case, which is each line in the input creates 2
2997 * elements in the inversion list: 1) the beginning of a range in the list;
2998 * 2) the beginning of a range not in the list. */
2999 while ((loc = (strchr(loc, '\n'))) != NULL) {
3000 elements += 2;
3001 loc++;
3002 }
3003
3004 /* If the ending is somehow corrupt and isn't a new line, add another
3005 * element for the final range that isn't in the inversion list */
3006 if (! (*lend == '\n' || (*lend == '\0' && *(lend - 1) == '\n'))) {
3007 elements++;
3008 }
3009
3010 invlist = _new_invlist(elements);
3011
3012 /* Now go through the input again, adding each range to the list */
3013 while (l < lend) {
3014 UV start, end;
3015 UV val; /* Not used by this function */
3016
3017 l = S_swash_scan_list_line(aTHX_ l, lend, &start, &end, &val,
3018 cBOOL(octets), typestr);
3019
3020 if (l > lend) {
3021 break;
3022 }
3023
3024 _append_range_to_invlist(invlist, start, end);
3025 }
3026
77f9f126
KW
3027 /* Invert if the data says it should be */
3028 if (invert_it_svp && SvUV(*invert_it_svp)) {
45bb2768 3029 _invlist_invert_prop(invlist);
77f9f126
KW
3030 }
3031
d73c39c5
KW
3032 /* This code is copied from swash_get()
3033 * read $swash->{EXTRAS} */
3034 x = (U8*)SvPV(*extssvp, xcur);
3035 xend = x + xcur;
3036 while (x < xend) {
3037 STRLEN namelen;
3038 U8 *namestr;
3039 SV** othersvp;
3040 HV* otherhv;
3041 STRLEN otherbits;
3042 SV **otherbitssvp, *other;
3043 U8 *nl;
3044
3045 const U8 opc = *x++;
3046 if (opc == '\n')
3047 continue;
3048
3049 nl = (U8*)memchr(x, '\n', xend - x);
3050
3051 if (opc != '-' && opc != '+' && opc != '!' && opc != '&') {
3052 if (nl) {
3053 x = nl + 1; /* 1 is length of "\n" */
3054 continue;
3055 }
3056 else {
3057 x = xend; /* to EXTRAS' end at which \n is not found */
3058 break;
3059 }
3060 }
3061
3062 namestr = x;
3063 if (nl) {
3064 namelen = nl - namestr;
3065 x = nl + 1;
3066 }
3067 else {
3068 namelen = xend - namestr;
3069 x = xend;
3070 }
3071
3072 othersvp = hv_fetch(hv, (char *)namestr, namelen, FALSE);
3073 otherhv = MUTABLE_HV(SvRV(*othersvp));
3074 otherbitssvp = hv_fetchs(otherhv, "BITS", FALSE);
3075 otherbits = (STRLEN)SvUV(*otherbitssvp);
3076
3077 if (bits != otherbits || bits != 1) {
3078 Perl_croak(aTHX_ "panic: _swash_to_invlist only operates on boolean properties");
3079 }
3080
3081 /* The "other" swatch must be destroyed after. */
3082 other = _swash_to_invlist((SV *)*othersvp);
3083
3084 /* End of code copied from swash_get() */
3085 switch (opc) {
3086 case '+':
3087 _invlist_union(invlist, other, &invlist);
3088 break;
3089 case '!':
3090 _invlist_invert(other);
3091 _invlist_union(invlist, other, &invlist);
3092 break;
3093 case '-':
3094 _invlist_subtract(invlist, other, &invlist);
3095 break;
3096 case '&':
3097 _invlist_intersection(invlist, other, &invlist);
3098 break;
3099 default:
3100 break;
3101 }
3102 sv_free(other); /* through with it! */
3103 }
3104
d764b54e
KW
3105 return invlist;
3106}
3107
0f830e0b 3108/*
87cea99e 3109=for apidoc uvchr_to_utf8
0f830e0b 3110
6ee84de2 3111Adds the UTF-8 representation of the Native code point C<uv> to the end
0f830e0b
NC
3112of the string C<d>; C<d> should be have at least C<UTF8_MAXBYTES+1> free
3113bytes available. The return value is the pointer to the byte after the
3114end of the new character. In other words,
3115
3116 d = uvchr_to_utf8(d, uv);
3117
3118is the recommended wide native character-aware way of saying
3119
3120 *(d++) = uv;
3121
3122=cut
3123*/
3124
3125/* On ASCII machines this is normally a macro but we want a
3126 real function in case XS code wants it
3127*/
0f830e0b
NC
3128U8 *
3129Perl_uvchr_to_utf8(pTHX_ U8 *d, UV uv)
3130{
7918f24d
NC
3131 PERL_ARGS_ASSERT_UVCHR_TO_UTF8;
3132
0f830e0b
NC
3133 return Perl_uvuni_to_utf8_flags(aTHX_ d, NATIVE_TO_UNI(uv), 0);
3134}
3135
b851fbc1
JH
3136U8 *
3137Perl_uvchr_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags)
3138{
7918f24d
NC
3139 PERL_ARGS_ASSERT_UVCHR_TO_UTF8_FLAGS;
3140
b851fbc1
JH
3141 return Perl_uvuni_to_utf8_flags(aTHX_ d, NATIVE_TO_UNI(uv), flags);
3142}
2b9d42f0
NIS
3143
3144/*
87cea99e 3145=for apidoc utf8n_to_uvchr
0f830e0b 3146
48ef279e 3147Returns the native character value of the first character in the string
0f830e0b
NC
3148C<s>
3149which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
3150length, in bytes, of that character.
3151
6ee84de2 3152length and flags are the same as utf8n_to_uvuni().
0f830e0b
NC
3153
3154=cut
3155*/
3156/* On ASCII machines this is normally a macro but we want
3157 a real function in case XS code wants it
3158*/
0f830e0b 3159UV
48ef279e 3160Perl_utf8n_to_uvchr(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen,
0f830e0b
NC
3161U32 flags)
3162{
3163 const UV uv = Perl_utf8n_to_uvuni(aTHX_ s, curlen, retlen, flags);
7918f24d
NC
3164
3165 PERL_ARGS_ASSERT_UTF8N_TO_UVCHR;
3166
0f830e0b
NC
3167 return UNI_TO_NATIVE(uv);
3168}
3169
0876b9a0
KW
3170bool
3171Perl_check_utf8_print(pTHX_ register const U8* s, const STRLEN len)
3172{
3173 /* May change: warns if surrogates, non-character code points, or
3174 * non-Unicode code points are in s which has length len. Returns TRUE if
3175 * none found; FALSE otherwise. The only other validity check is to make
3176 * sure that this won't exceed the string's length */
3177
3178 const U8* const e = s + len;
3179 bool ok = TRUE;
3180
3181 PERL_ARGS_ASSERT_CHECK_UTF8_PRINT;
3182
3183 while (s < e) {
3184 if (UTF8SKIP(s) > len) {
3185 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8),
3186 "%s in %s", unees, PL_op ? OP_DESC(PL_op) : "print");
3187 return FALSE;
3188 }
3189 if (*s >= UTF8_FIRST_PROBLEMATIC_CODE_POINT_FIRST_BYTE) {
3190 STRLEN char_len;
3191 if (UTF8_IS_SUPER(s)) {
8457b38f
KW
3192 if (ckWARN_d(WARN_NON_UNICODE)) {
3193 UV uv = utf8_to_uvchr(s, &char_len);
3194 Perl_warner(aTHX_ packWARN(WARN_NON_UNICODE),
3195 "Code point 0x%04"UVXf" is not Unicode, may not be portable", uv);
3196 ok = FALSE;
3197 }
0876b9a0
KW
3198 }
3199 else if (UTF8_IS_SURROGATE(s)) {
8457b38f
KW
3200 if (ckWARN_d(WARN_SURROGATE)) {
3201 UV uv = utf8_to_uvchr(s, &char_len);
3202 Perl_warner(aTHX_ packWARN(WARN_SURROGATE),
3203 "Unicode surrogate U+%04"UVXf" is illegal in UTF-8", uv);
3204 ok = FALSE;
3205 }
0876b9a0
KW
3206 }
3207 else if
8457b38f
KW
3208 ((UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s))
3209 && (ckWARN_d(WARN_NONCHAR)))
0876b9a0
KW
3210 {
3211 UV uv = utf8_to_uvchr(s, &char_len);
8457b38f 3212 Perl_warner(aTHX_ packWARN(WARN_NONCHAR),
0876b9a0
KW
3213 "Unicode non-character U+%04"UVXf" is illegal for open interchange", uv);
3214 ok = FALSE;
3215 }
3216 }
3217 s += UTF8SKIP(s);
3218 }
3219
3220 return ok;
3221}
3222
0f830e0b 3223/*
87cea99e 3224=for apidoc pv_uni_display
d2cc3551
JH
3225
3226Build to the scalar dsv a displayable version of the string spv,
3227length len, the displayable version being at most pvlim bytes long
3228(if longer, the rest is truncated and "..." will be appended).
0a2ef054 3229
9e55ce06 3230The flags argument can have UNI_DISPLAY_ISPRINT set to display
00e86452 3231isPRINT()able characters as themselves, UNI_DISPLAY_BACKSLASH
0a2ef054
JH
3232to display the \\[nrfta\\] as the backslashed versions (like '\n')
3233(UNI_DISPLAY_BACKSLASH is preferred over UNI_DISPLAY_ISPRINT for \\).
3234UNI_DISPLAY_QQ (and its alias UNI_DISPLAY_REGEX) have both
3235UNI_DISPLAY_BACKSLASH and UNI_DISPLAY_ISPRINT turned on.
3236
d2cc3551
JH
3237The pointer to the PV of the dsv is returned.
3238
3239=cut */
e6b2e755 3240char *
e1ec3a88 3241Perl_pv_uni_display(pTHX_ SV *dsv, const U8 *spv, STRLEN len, STRLEN pvlim, UV flags)
e6b2e755
JH
3242{
3243 int truncated = 0;
e1ec3a88 3244 const char *s, *e;
e6b2e755 3245
7918f24d
NC
3246 PERL_ARGS_ASSERT_PV_UNI_DISPLAY;
3247
76f68e9b 3248 sv_setpvs(dsv, "");
7fddd944 3249 SvUTF8_off(dsv);
e1ec3a88 3250 for (s = (const char *)spv, e = s + len; s < e; s += UTF8SKIP(s)) {
e6b2e755 3251 UV u;
a49f32c6
NC
3252 /* This serves double duty as a flag and a character to print after
3253 a \ when flags & UNI_DISPLAY_BACKSLASH is true.
3254 */
3255 char ok = 0;
c728cb41 3256
e6b2e755
JH
3257 if (pvlim && SvCUR(dsv) >= pvlim) {
3258 truncated++;
3259 break;
3260 }
3261 u = utf8_to_uvchr((U8*)s, 0);
c728cb41 3262 if (u < 256) {
a3b680e6 3263 const unsigned char c = (unsigned char)u & 0xFF;
0bd48802 3264 if (flags & UNI_DISPLAY_BACKSLASH) {
a49f32c6 3265 switch (c) {
c728cb41 3266 case '\n':
a49f32c6 3267 ok = 'n'; break;
c728cb41 3268 case '\r':
a49f32c6 3269 ok = 'r'; break;
c728cb41 3270 case '\t':
a49f32c6 3271 ok = 't'; break;
c728cb41 3272 case '\f':
a49f32c6 3273 ok = 'f'; break;
c728cb41 3274 case '\a':
a49f32c6 3275 ok = 'a'; break;
c728cb41 3276 case '\\':
a49f32c6 3277 ok = '\\'; break;
c728cb41
JH
3278 default: break;
3279 }
a49f32c6 3280 if (ok) {
88c9ea1e 3281 const char string = ok;
76f68e9b 3282 sv_catpvs(dsv, "\\");
5e7aa789 3283 sv_catpvn(dsv, &string, 1);
a49f32c6 3284 }
c728cb41 3285 }
00e86452 3286 /* isPRINT() is the locale-blind version. */
a49f32c6 3287 if (!ok && (flags & UNI_DISPLAY_ISPRINT) && isPRINT(c)) {
88c9ea1e 3288 const char string = c;
5e7aa789 3289 sv_catpvn(dsv, &string, 1);
a49f32c6 3290 ok = 1;
0a2ef054 3291 }
c728cb41
JH
3292 }
3293 if (!ok)
9e55ce06 3294 Perl_sv_catpvf(aTHX_ dsv, "\\x{%"UVxf"}", u);
e6b2e755
JH
3295 }
3296 if (truncated)
396482e1 3297 sv_catpvs(dsv, "...");
48ef279e 3298
e6b2e755
JH
3299 return SvPVX(dsv);
3300}
2b9d42f0 3301
d2cc3551 3302/*
87cea99e 3303=for apidoc sv_uni_display
d2cc3551
JH
3304
3305Build to the scalar dsv a displayable version of the scalar sv,
0a2ef054 3306the displayable version being at most pvlim bytes long
d2cc3551 3307(if longer, the rest is truncated and "..." will be appended).
0a2ef054
JH
3308
3309The flags argument is as in pv_uni_display().
3310
d2cc3551
JH
3311The pointer to the PV of the dsv is returned.
3312
d4c19fe8
AL
3313=cut
3314*/
e6b2e755
JH
3315char *
3316Perl_sv_uni_display(pTHX_ SV *dsv, SV *ssv, STRLEN pvlim, UV flags)
3317{
7918f24d
NC
3318 PERL_ARGS_ASSERT_SV_UNI_DISPLAY;
3319
cfd0369c
NC
3320 return Perl_pv_uni_display(aTHX_ dsv, (const U8*)SvPVX_const(ssv),
3321 SvCUR(ssv), pvlim, flags);
701a277b
JH
3322}
3323
d2cc3551 3324/*
e6226b18 3325=for apidoc foldEQ_utf8
d2cc3551 3326
d51c1b21 3327Returns true if the leading portions of the strings s1 and s2 (either or both
e6226b18 3328of which may be in UTF-8) are the same case-insensitively; false otherwise.
d51c1b21 3329How far into the strings to compare is determined by other input parameters.
8b35872c
KW
3330
3331If u1 is true, the string s1 is assumed to be in UTF-8-encoded Unicode;
3332otherwise it is assumed to be in native 8-bit encoding. Correspondingly for u2
3333with respect to s2.
3334
d51c1b21
KW
3335If the byte length l1 is non-zero, it says how far into s1 to check for fold
3336equality. In other words, s1+l1 will be used as a goal to reach. The
8b35872c
KW
3337scan will not be considered to be a match unless the goal is reached, and
3338scanning won't continue past that goal. Correspondingly for l2 with respect to
3339s2.
3340
3341If pe1 is non-NULL and the pointer it points to is not NULL, that pointer is
3342considered an end pointer beyond which scanning of s1 will not continue under
3343any circumstances. This means that if both l1 and pe1 are specified, and pe1
3344is less than s1+l1, the match will never be successful because it can never
d51c1b21
KW
3345get as far as its goal (and in fact is asserted against). Correspondingly for
3346pe2 with respect to s2.
8b35872c 3347
d51c1b21
KW
3348At least one of s1 and s2 must have a goal (at least one of l1 and l2 must be
3349non-zero), and if both do, both have to be
8b35872c
KW
3350reached for a successful match. Also, if the fold of a character is multiple
3351characters, all of them must be matched (see tr21 reference below for
3352'folding').
3353
e6226b18 3354Upon a successful match, if pe1 is non-NULL,
8b35872c
KW
3355it will be set to point to the beginning of the I<next> character of s1 beyond
3356what was matched. Correspondingly for pe2 and s2.
d2cc3551
JH
3357
3358For case-insensitiveness, the "casefolding" of Unicode is used
3359instead of upper/lowercasing both the characters, see
3360http://www.unicode.org/unicode/reports/tr21/ (Case Mappings).
3361
3362=cut */
a33c29bc
KW
3363
3364/* A flags parameter has been added which may change, and hence isn't
3365 * externally documented. Currently it is:
3366 * 0 for as-documented above
3367 * FOLDEQ_UTF8_NOMIX_ASCII meaning that if a non-ASCII character folds to an
3368 ASCII one, to not match
5e64d0fa
KW
3369 * FOLDEQ_UTF8_LOCALE meaning that locale rules are to be used for code
3370 * points below 256; unicode rules for above 255; and
3371 * folds that cross those boundaries are disallowed,
3372 * like the NOMIX_ASCII option
18f762c3
KW
3373 * FOLDEQ_S1_ALREADY_FOLDED s1 has already been folded before calling this
3374 * routine. This allows that step to be skipped.
3375 * FOLDEQ_S2_ALREADY_FOLDED Similarly.
a33c29bc 3376 */
701a277b 3377I32
eda9cac1 3378Perl_foldEQ_utf8_flags(pTHX_ const char *s1, char **pe1, register UV l1, bool u1, const char *s2, char **pe2, register UV l2, bool u2, U32 flags)
332ddc25 3379{
8b35872c
KW
3380 dVAR;
3381 register const U8 *p1 = (const U8*)s1; /* Point to current char */
3382 register const U8 *p2 = (const U8*)s2;
48ef279e 3383 register const U8 *g1 = NULL; /* goal for s1 */
8b35872c 3384 register const U8 *g2 = NULL;
48ef279e
KW
3385 register const U8 *e1 = NULL; /* Don't scan s1 past this */
3386 register U8 *f1 = NULL; /* Point to current folded */
8b35872c
KW
3387 register const U8 *e2 = NULL;
3388 register U8 *f2 = NULL;
48ef279e 3389 STRLEN n1 = 0, n2 = 0; /* Number of bytes in current char */
8b35872c
KW
3390 U8 foldbuf1[UTF8_MAXBYTES_CASE+1];
3391 U8 foldbuf2[UTF8_MAXBYTES_CASE+1];
48ef279e
KW
3392 U8 natbuf[2]; /* Holds native 8-bit char converted to utf8;
3393 these always fit in 2 bytes */
8b35872c 3394
eda9cac1 3395 PERL_ARGS_ASSERT_FOLDEQ_UTF8_FLAGS;
8b35872c 3396
18f762c3
KW
3397 /* The algorithm requires that input with the flags on the first line of
3398 * the assert not be pre-folded. */
3399 assert( ! ((flags & (FOLDEQ_UTF8_NOMIX_ASCII | FOLDEQ_UTF8_LOCALE))
3400 && (flags & (FOLDEQ_S1_ALREADY_FOLDED | FOLDEQ_S2_ALREADY_FOLDED))));
3401
8b35872c 3402 if (pe1) {
48ef279e 3403 e1 = *(U8**)pe1;
8b35872c
KW
3404 }
3405
3406 if (l1) {
48ef279e 3407 g1 = (const U8*)s1 + l1;
8b35872c
KW
3408 }
3409
3410 if (pe2) {
48ef279e 3411 e2 = *(U8**)pe2;
8b35872c
KW
3412 }
3413
3414 if (l2) {
48ef279e 3415 g2 = (const U8*)s2 + l2;
8b35872c
KW
3416 }
3417
3418 /* Must have at least one goal */
3419 assert(g1 || g2);
3420
3421 if (g1) {
3422
48ef279e
KW
3423 /* Will never match if goal is out-of-bounds */
3424 assert(! e1 || e1 >= g1);
8b35872c 3425
48ef279e
KW
3426 /* Here, there isn't an end pointer, or it is beyond the goal. We
3427 * only go as far as the goal */
3428 e1 = g1;
8b35872c 3429 }
313b38e5
NC
3430 else {
3431 assert(e1); /* Must have an end for looking at s1 */
3432 }
8b35872c
KW
3433
3434 /* Same for goal for s2 */
3435 if (g2) {
48ef279e
KW
3436 assert(! e2 || e2 >= g2);
3437 e2 = g2;
8b35872c 3438 }
313b38e5
NC
3439 else {
3440 assert(e2);
3441 }
8b35872c 3442
18f762c3
KW
3443 /* If both operands are already folded, we could just do a memEQ on the
3444 * whole strings at once, but it would be better if the caller realized
3445 * this and didn't even call us */
3446
8b35872c
KW
3447 /* Look through both strings, a character at a time */
3448 while (p1 < e1 && p2 < e2) {
3449
d51c1b21 3450 /* If at the beginning of a new character in s1, get its fold to use
5e64d0fa
KW
3451 * and the length of the fold. (exception: locale rules just get the
3452 * character to a single byte) */
48ef279e 3453 if (n1 == 0) {
18f762c3
KW
3454 if (flags & FOLDEQ_S1_ALREADY_FOLDED) {
3455 f1 = (U8 *) p1;
3456 n1 = UTF8SKIP(f1);
a33c29bc 3457
5e64d0fa
KW
3458 /* If in locale matching, we use two sets of rules, depending on if
3459 * the code point is above or below 255. Here, we test for and
3460 * handle locale rules */
18f762c3
KW
3461 }
3462 else {
1b4059d5
KW
3463 if ((flags & FOLDEQ_UTF8_LOCALE)
3464 && (! u1 || UTF8_IS_INVARIANT(*p1)
3465 || UTF8_IS_DOWNGRADEABLE_START(*p1)))
5e64d0fa 3466 {
1b4059d5
KW
3467 /* There is no mixing of code points above and below 255. */
3468 if (u2 && (! UTF8_IS_INVARIANT(*p2)
3469 && ! UTF8_IS_DOWNGRADEABLE_START(*p2)))
3470 {
3471 return 0;
3472 }
5e64d0fa 3473
1b4059d5
KW
3474 /* We handle locale rules by converting, if necessary, the
3475 * code point to a single byte. */
3476 if (! u1 || UTF8_IS_INVARIANT(*p1)) {
3477 *foldbuf1 = *p1;
3478 }
3479 else {
3480 *foldbuf1 = TWO_BYTE_UTF8_TO_UNI(*p1, *(p1 + 1));
3481 }
3482 n1 = 1;
5e64d0fa 3483 }
1b4059d5
KW
3484 else if (isASCII(*p1)) { /* Note, that here won't be
3485 both ASCII and using locale
3486 rules */
3487
3488 /* If trying to mix non- with ASCII, and not supposed to,
3489 * fail */
3490 if ((flags & FOLDEQ_UTF8_NOMIX_ASCII) && ! isASCII(*p2)) {
3491 return 0;
3492 }
3493 n1 = 1;
3494 *foldbuf1 = toLOWER(*p1); /* Folds in the ASCII range are
3495 just lowercased */
5e64d0fa 3496 }
1b4059d5
KW
3497 else if (u1) {
3498 to_utf8_fold(p1, foldbuf1, &n1);
a33c29bc 3499 }
1b4059d5
KW
3500 else { /* Not utf8, convert to it first and then get fold */
3501 uvuni_to_utf8(natbuf, (UV) NATIVE_TO_UNI(((UV)*p1)));
3502 to_utf8_fold(natbuf, foldbuf1, &n1);
3503 }
3504 f1 = foldbuf1;
18f762c3 3505 }
48ef279e 3506 }
8b35872c 3507
48ef279e 3508 if (n2 == 0) { /* Same for s2 */
18f762c3
KW
3509 if (flags & FOLDEQ_S2_ALREADY_FOLDED) {
3510 f2 = (U8 *) p2;
3511 n2 = UTF8SKIP(f2);
3512 }
3513 else {
227968da
KW
3514 if ((flags & FOLDEQ_UTF8_LOCALE)
3515 && (! u2 || UTF8_IS_INVARIANT(*p2) || UTF8_IS_DOWNGRADEABLE_START(*p2)))
5e64d0fa 3516 {
227968da
KW
3517 /* Here, the next char in s2 is < 256. We've already
3518 * worked on s1, and if it isn't also < 256, can't match */
3519 if (u1 && (! UTF8_IS_INVARIANT(*p1)
3520 && ! UTF8_IS_DOWNGRADEABLE_START(*p1)))
3521 {
3522 return 0;
3523 }
3524 if (! u2 || UTF8_IS_INVARIANT(*p2)) {
3525 *foldbuf2 = *p2;
3526 }
3527 else {
3528 *foldbuf2 = TWO_BYTE_UTF8_TO_UNI(*p2, *(p2 + 1));
3529 }
3530
3531 /* Use another function to handle locale rules. We've made
3532 * sure that both characters to compare are single bytes */
3533 if (! foldEQ_locale((char *) f1, (char *) foldbuf2, 1)) {
3534 return 0;
3535 }
3536 n1 = n2 = 0;
5e64d0fa 3537 }
227968da
KW
3538 else if (isASCII(*p2)) {
3539 if (flags && ! isASCII(*p1)) {
3540 return 0;
3541 }
3542 n2 = 1;
3543 *foldbuf2 = toLOWER(*p2);
5e64d0fa 3544 }
227968da
KW
3545 else if (u2) {
3546 to_utf8_fold(p2, foldbuf2, &n2);
5001101e 3547 }
227968da
KW
3548 else {
3549 uvuni_to_utf8(natbuf, (UV) NATIVE_TO_UNI(((UV)*p2)));
3550 to_utf8_fold(natbuf, foldbuf2, &n2);
a33c29bc 3551 }
227968da 3552 f2 = foldbuf2;
18f762c3 3553 }
48ef279e 3554 }
8b35872c 3555
5001101e 3556 /* Here f1 and f2 point to the beginning of the strings to compare.
227968da
KW
3557 * These strings are the folds of the next character from each input
3558 * string, stored in utf8. */
5e64d0fa 3559
48ef279e
KW
3560 /* While there is more to look for in both folds, see if they
3561 * continue to match */
3562 while (n1 && n2) {
3563 U8 fold_length = UTF8SKIP(f1);
3564 if (fold_length != UTF8SKIP(f2)
3565 || (fold_length == 1 && *f1 != *f2) /* Short circuit memNE
3566 function call for single
3567 character */
3568 || memNE((char*)f1, (char*)f2, fold_length))
3569 {
e6226b18 3570 return 0; /* mismatch */
48ef279e
KW
3571 }
3572
3573 /* Here, they matched, advance past them */
3574 n1 -= fold_length;
3575 f1 += fold_length;
3576 n2 -= fold_length;
3577 f2 += fold_length;
3578 }
8b35872c 3579
48ef279e
KW
3580 /* When reach the end of any fold, advance the input past it */
3581 if (n1 == 0) {
3582 p1 += u1 ? UTF8SKIP(p1) : 1;
3583 }
3584 if (n2 == 0) {
3585 p2 += u2 ? UTF8SKIP(p2) : 1;
3586 }
8b35872c
KW
3587 } /* End of loop through both strings */
3588
3589 /* A match is defined by each scan that specified an explicit length
3590 * reaching its final goal, and the other not having matched a partial
3591 * character (which can happen when the fold of a character is more than one
3592 * character). */
3593 if (! ((g1 == 0 || p1 == g1) && (g2 == 0 || p2 == g2)) || n1 || n2) {
e6226b18 3594 return 0;
8b35872c
KW
3595 }
3596
3597 /* Successful match. Set output pointers */
3598 if (pe1) {
48ef279e 3599 *pe1 = (char*)p1;
8b35872c
KW
3600 }
3601 if (pe2) {
48ef279e 3602 *pe2 = (char*)p2;
8b35872c 3603 }
e6226b18 3604 return 1;
e6b2e755 3605}
701a277b 3606
a49f32c6
NC
3607/*
3608 * Local variables:
3609 * c-indentation-style: bsd
3610 * c-basic-offset: 4
3611 * indent-tabs-mode: t
3612 * End:
3613 *
37442d52
RGS
3614 * ex: set ts=8 sts=4 sw=4 noet:
3615 */