This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
utf8n_to_uvuni: Avoid reading outside of buffer
[perl5.git] / utf8.c
CommitLineData
a0ed51b3
LW
1/* utf8.c
2 *
1129b882 3 * Copyright (C) 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008
b94e2f88 4 * by Larry Wall and others
a0ed51b3
LW
5 *
6 * You may distribute under the terms of either the GNU General Public
7 * License or the Artistic License, as specified in the README file.
8 *
9 */
10
11/*
4ac71550
TC
12 * 'What a fix!' said Sam. 'That's the one place in all the lands we've ever
13 * heard of that we don't want to see any closer; and that's the one place
14 * we're trying to get to! And that's just where we can't get, nohow.'
15 *
cdad3b53 16 * [p.603 of _The Lord of the Rings_, IV/I: "The Taming of Sméagol"]
a0ed51b3
LW
17 *
18 * 'Well do I understand your speech,' he answered in the same language;
19 * 'yet few strangers do so. Why then do you not speak in the Common Tongue,
4ac71550 20 * as is the custom in the West, if you wish to be answered?'
cdad3b53 21 * --Gandalf, addressing Théoden's door wardens
4ac71550
TC
22 *
23 * [p.508 of _The Lord of the Rings_, III/vi: "The King of the Golden Hall"]
a0ed51b3
LW
24 *
25 * ...the travellers perceived that the floor was paved with stones of many
26 * hues; branching runes and strange devices intertwined beneath their feet.
4ac71550
TC
27 *
28 * [p.512 of _The Lord of the Rings_, III/vi: "The King of the Golden Hall"]
a0ed51b3
LW
29 */
30
31#include "EXTERN.h"
864dbfa3 32#define PERL_IN_UTF8_C
a0ed51b3
LW
33#include "perl.h"
34
a0c21aa1 35#ifndef EBCDIC
970ea3cb 36/* Separate prototypes needed because in ASCII systems these are
a0c21aa1
JH
37 * usually macros but they still are compiled as code, too. */
38PERL_CALLCONV UV Perl_utf8n_to_uvchr(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen, U32 flags);
39PERL_CALLCONV U8* Perl_uvchr_to_utf8(pTHX_ U8 *d, UV uv);
40#endif
41
27da23d5
JH
42static const char unees[] =
43 "Malformed UTF-8 character (unexpected end of string)";
901b21bf 44
48ef279e 45/*
ccfc67b7 46=head1 Unicode Support
a0ed51b3 47
166f8a29
DM
48This file contains various utility functions for manipulating UTF8-encoded
49strings. For the uninitiated, this is a method of representing arbitrary
61296642 50Unicode characters as a variable number of bytes, in such a way that
56da48f7
DM
51characters in the ASCII range are unmodified, and a zero byte never appears
52within non-zero characters.
166f8a29 53
eaf7a4d2
CS
54=cut
55*/
56
57/*
58=for apidoc is_ascii_string
59
a1433954 60Returns true if the first C<len> bytes of the string C<s> are the same whether
970ea3cb
KW
61or not the string is encoded in UTF-8 (or UTF-EBCDIC on EBCDIC machines). That
62is, if they are invariant. On ASCII-ish machines, only ASCII characters
63fit this definition, hence the function's name.
eaf7a4d2 64
9f7e3d64
MH
65If C<len> is 0, it will be calculated using C<strlen(s)>.
66
a1433954 67See also L</is_utf8_string>(), L</is_utf8_string_loclen>(), and L</is_utf8_string_loc>().
eaf7a4d2
CS
68
69=cut
70*/
71
72bool
668b6d8d 73Perl_is_ascii_string(const U8 *s, STRLEN len)
eaf7a4d2
CS
74{
75 const U8* const send = s + (len ? len : strlen((const char *)s));
76 const U8* x = s;
77
78 PERL_ARGS_ASSERT_IS_ASCII_STRING;
eaf7a4d2
CS
79
80 for (; x < send; ++x) {
81 if (!UTF8_IS_INVARIANT(*x))
82 break;
83 }
84
85 return x == send;
86}
87
88/*
87cea99e 89=for apidoc uvuni_to_utf8_flags
eebe1485 90
6ee84de2
KW
91Adds the UTF-8 representation of the code point C<uv> to the end
92of the string C<d>; C<d> should have at least C<UTF8_MAXBYTES+1> free
eebe1485 93bytes available. The return value is the pointer to the byte after the
9041c2e3 94end of the new character. In other words,
eebe1485 95
b851fbc1
JH
96 d = uvuni_to_utf8_flags(d, uv, flags);
97
98or, in most cases,
99
9041c2e3 100 d = uvuni_to_utf8(d, uv);
eebe1485 101
b851fbc1
JH
102(which is equivalent to)
103
104 d = uvuni_to_utf8_flags(d, uv, 0);
105
949cf498 106This is the recommended Unicode-aware way of saying
eebe1485
SC
107
108 *(d++) = uv;
109
949cf498
KW
110This function will convert to UTF-8 (and not warn) even code points that aren't
111legal Unicode or are problematic, unless C<flags> contains one or more of the
a1433954
KW
112following flags:
113
949cf498
KW
114If C<uv> is a Unicode surrogate code point and UNICODE_WARN_SURROGATE is set,
115the function will raise a warning, provided UTF8 warnings are enabled. If instead
116UNICODE_DISALLOW_SURROGATE is set, the function will fail and return NULL.
117If both flags are set, the function will both warn and return NULL.
118
119The UNICODE_WARN_NONCHAR and UNICODE_DISALLOW_NONCHAR flags correspondingly
120affect how the function handles a Unicode non-character. And, likewise for the
121UNICODE_WARN_SUPER and UNICODE_DISALLOW_SUPER flags, and code points that are
122above the Unicode maximum of 0x10FFFF. Code points above 0x7FFF_FFFF (which are
123even less portable) can be warned and/or disallowed even if other above-Unicode
124code points are accepted by the UNICODE_WARN_FE_FF and UNICODE_DISALLOW_FE_FF
125flags.
126
127And finally, the flag UNICODE_WARN_ILLEGAL_INTERCHANGE selects all four of the
128above WARN flags; and UNICODE_DISALLOW_ILLEGAL_INTERCHANGE selects all four
129DISALLOW flags.
130
131
eebe1485
SC
132=cut
133*/
134
dfe13c55 135U8 *
b851fbc1 136Perl_uvuni_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags)
a0ed51b3 137{
7918f24d
NC
138 PERL_ARGS_ASSERT_UVUNI_TO_UTF8_FLAGS;
139
a099aed4 140 if (ckWARN4_d(WARN_UTF8, WARN_SURROGATE, WARN_NON_UNICODE, WARN_NONCHAR)) {
949cf498
KW
141 if (UNICODE_IS_SURROGATE(uv)) {
142 if (flags & UNICODE_WARN_SURROGATE) {
8457b38f 143 Perl_ck_warner_d(aTHX_ packWARN(WARN_SURROGATE),
949cf498
KW
144 "UTF-16 surrogate U+%04"UVXf, uv);
145 }
146 if (flags & UNICODE_DISALLOW_SURROGATE) {
147 return NULL;
148 }
149 }
150 else if (UNICODE_IS_SUPER(uv)) {
151 if (flags & UNICODE_WARN_SUPER
152 || (UNICODE_IS_FE_FF(uv) && (flags & UNICODE_WARN_FE_FF)))
153 {
8457b38f 154 Perl_ck_warner_d(aTHX_ packWARN(WARN_NON_UNICODE),
949cf498
KW
155 "Code point 0x%04"UVXf" is not Unicode, may not be portable", uv);
156 }
157 if (flags & UNICODE_DISALLOW_SUPER
158 || (UNICODE_IS_FE_FF(uv) && (flags & UNICODE_DISALLOW_FE_FF)))
159 {
160 return NULL;
161 }
162 }
163 else if (UNICODE_IS_NONCHAR(uv)) {
164 if (flags & UNICODE_WARN_NONCHAR) {
8457b38f 165 Perl_ck_warner_d(aTHX_ packWARN(WARN_NONCHAR),
949cf498
KW
166 "Unicode non-character U+%04"UVXf" is illegal for open interchange",
167 uv);
168 }
169 if (flags & UNICODE_DISALLOW_NONCHAR) {
170 return NULL;
171 }
172 }
507b9800 173 }
c4d5f83a 174 if (UNI_IS_INVARIANT(uv)) {
eb160463 175 *d++ = (U8)UTF_TO_NATIVE(uv);
a0ed51b3
LW
176 return d;
177 }
2d331972 178#if defined(EBCDIC)
1d72bdf6
NIS
179 else {
180 STRLEN len = UNISKIP(uv);
181 U8 *p = d+len-1;
182 while (p > d) {
eb160463 183 *p-- = (U8)UTF_TO_NATIVE((uv & UTF_CONTINUATION_MASK) | UTF_CONTINUATION_MARK);
1d72bdf6
NIS
184 uv >>= UTF_ACCUMULATION_SHIFT;
185 }
eb160463 186 *p = (U8)UTF_TO_NATIVE((uv & UTF_START_MASK(len)) | UTF_START_MARK(len));
1d72bdf6
NIS
187 return d+len;
188 }
189#else /* Non loop style */
a0ed51b3 190 if (uv < 0x800) {
eb160463
GS
191 *d++ = (U8)(( uv >> 6) | 0xc0);
192 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
193 return d;
194 }
195 if (uv < 0x10000) {
eb160463
GS
196 *d++ = (U8)(( uv >> 12) | 0xe0);
197 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
198 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
199 return d;
200 }
201 if (uv < 0x200000) {
eb160463
GS
202 *d++ = (U8)(( uv >> 18) | 0xf0);
203 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
204 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
205 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
206 return d;
207 }
208 if (uv < 0x4000000) {
eb160463
GS
209 *d++ = (U8)(( uv >> 24) | 0xf8);
210 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
211 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
212 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
213 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
214 return d;
215 }
216 if (uv < 0x80000000) {
eb160463
GS
217 *d++ = (U8)(( uv >> 30) | 0xfc);
218 *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80);
219 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
220 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
221 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
222 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
223 return d;
224 }
6b8eaf93 225#ifdef HAS_QUAD
d7578b48 226 if (uv < UTF8_QUAD_MAX)
a0ed51b3
LW
227#endif
228 {
eb160463
GS
229 *d++ = 0xfe; /* Can't match U+FEFF! */
230 *d++ = (U8)(((uv >> 30) & 0x3f) | 0x80);
231 *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80);
232 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
233 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
234 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
235 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
236 return d;
237 }
6b8eaf93 238#ifdef HAS_QUAD
a0ed51b3 239 {
eb160463
GS
240 *d++ = 0xff; /* Can't match U+FFFE! */
241 *d++ = 0x80; /* 6 Reserved bits */
242 *d++ = (U8)(((uv >> 60) & 0x0f) | 0x80); /* 2 Reserved bits */
243 *d++ = (U8)(((uv >> 54) & 0x3f) | 0x80);
244 *d++ = (U8)(((uv >> 48) & 0x3f) | 0x80);
245 *d++ = (U8)(((uv >> 42) & 0x3f) | 0x80);
246 *d++ = (U8)(((uv >> 36) & 0x3f) | 0x80);
247 *d++ = (U8)(((uv >> 30) & 0x3f) | 0x80);
248 *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80);
249 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
250 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
251 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
252 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
253 return d;
254 }
255#endif
1d72bdf6 256#endif /* Loop style */
a0ed51b3 257}
9041c2e3 258
646ca15d
JH
259/*
260
f7d739d1 261Tests if the first C<len> bytes of string C<s> form a valid UTF-8
646ca15d 262character. Note that an INVARIANT (i.e. ASCII) character is a valid
f7d739d1 263UTF-8 character. The number of bytes in the UTF-8 character
646ca15d
JH
264will be returned if it is valid, otherwise 0.
265
266This is the "slow" version as opposed to the "fast" version which is
267the "unrolled" IS_UTF8_CHAR(). E.g. for t/uni/class.t the speed
268difference is a factor of 2 to 3. For lengths (UTF8SKIP(s)) of four
269or less you should use the IS_UTF8_CHAR(), for lengths of five or more
270you should use the _slow(). In practice this means that the _slow()
271will be used very rarely, since the maximum Unicode code point (as of
272Unicode 4.1) is U+10FFFF, which encodes in UTF-8 to four bytes. Only
273the "Perl extended UTF-8" (the infamous 'v-strings') will encode into
274five bytes or more.
275
276=cut */
c053b435 277STATIC STRLEN
5f66b61c 278S_is_utf8_char_slow(const U8 *s, const STRLEN len)
646ca15d
JH
279{
280 U8 u = *s;
281 STRLEN slen;
282 UV uv, ouv;
283
7918f24d
NC
284 PERL_ARGS_ASSERT_IS_UTF8_CHAR_SLOW;
285
646ca15d 286 if (UTF8_IS_INVARIANT(u))
f7d739d1 287 return len == 1;
646ca15d
JH
288
289 if (!UTF8_IS_START(u))
290 return 0;
291
292 if (len < 2 || !UTF8_IS_CONTINUATION(s[1]))
293 return 0;
294
295 slen = len - 1;
296 s++;
77263263
ST
297#ifdef EBCDIC
298 u = NATIVE_TO_UTF(u);
299#endif
646ca15d
JH
300 u &= UTF_START_MASK(len);
301 uv = u;
302 ouv = uv;
303 while (slen--) {
304 if (!UTF8_IS_CONTINUATION(*s))
305 return 0;
306 uv = UTF8_ACCUMULATE(uv, *s);
48ef279e 307 if (uv < ouv)
646ca15d
JH
308 return 0;
309 ouv = uv;
310 s++;
311 }
312
313 if ((STRLEN)UNISKIP(uv) < len)
314 return 0;
315
316 return len;
317}
9041c2e3
NIS
318
319/*
492a624f
KW
320=for apidoc is_utf8_char_buf
321
322Returns the number of bytes that comprise the first UTF-8 encoded character in
323buffer C<buf>. C<buf_end> should point to one position beyond the end of the
324buffer. 0 is returned if C<buf> does not point to a complete, valid UTF-8
325encoded character.
326
327Note that an INVARIANT character (i.e. ASCII on non-EBCDIC
328machines) is a valid UTF-8 character.
329
330=cut */
331
332STRLEN
333Perl_is_utf8_char_buf(const U8 *buf, const U8* buf_end)
334{
335
336 STRLEN len;
337
338 PERL_ARGS_ASSERT_IS_UTF8_CHAR_BUF;
339
340 if (buf_end <= buf) {
341 return 0;
342 }
343
344 len = buf_end - buf;
345 if (len > UTF8SKIP(buf)) {
346 len = UTF8SKIP(buf);
347 }
348
349#ifdef IS_UTF8_CHAR
350 if (IS_UTF8_CHAR_FAST(len))
351 return IS_UTF8_CHAR(buf, len) ? len : 0;
352#endif /* #ifdef IS_UTF8_CHAR */
353 return is_utf8_char_slow(buf, len);
354}
355
356/*
87cea99e 357=for apidoc is_utf8_char
eebe1485 358
76848387
KW
359DEPRECATED!
360
5da9da9e 361Tests if some arbitrary number of bytes begins in a valid UTF-8
2bbc8d55
SP
362character. Note that an INVARIANT (i.e. ASCII on non-EBCDIC machines)
363character is a valid UTF-8 character. The actual number of bytes in the UTF-8
364character will be returned if it is valid, otherwise 0.
9041c2e3 365
76848387 366This function is deprecated due to the possibility that malformed input could
a1433954 367cause reading beyond the end of the input buffer. Use L</is_utf8_char_buf>
76848387 368instead.
e0328548 369
82686b01 370=cut */
76848387 371
067a85ef 372STRLEN
668b6d8d 373Perl_is_utf8_char(const U8 *s)
386d01d6 374{
7918f24d 375 PERL_ARGS_ASSERT_IS_UTF8_CHAR;
492a624f 376
76848387 377 /* Assumes we have enough space, which is why this is deprecated */
492a624f 378 return is_utf8_char_buf(s, s + UTF8SKIP(s));
386d01d6
GS
379}
380
eaf7a4d2 381
6662521e 382/*
87cea99e 383=for apidoc is_utf8_string
6662521e 384
a1433954 385Returns true if the first C<len> bytes of string C<s> form a valid
9f7e3d64 386UTF-8 string, false otherwise. If C<len> is 0, it will be calculated
e0328548
KW
387using C<strlen(s)> (which means if you use this option, that C<s> has to have a
388terminating NUL byte). Note that all characters being ASCII constitute 'a
389valid UTF-8 string'.
6662521e 390
a1433954 391See also L</is_ascii_string>(), L</is_utf8_string_loclen>(), and L</is_utf8_string_loc>().
768c67ee 392
6662521e
GS
393=cut
394*/
395
8e84507e 396bool
668b6d8d 397Perl_is_utf8_string(const U8 *s, STRLEN len)
6662521e 398{
35da51f7 399 const U8* const send = s + (len ? len : strlen((const char *)s));
7fc63493 400 const U8* x = s;
067a85ef 401
7918f24d 402 PERL_ARGS_ASSERT_IS_UTF8_STRING;
1aa99e6b 403
6662521e 404 while (x < send) {
1acdb0da 405 /* Inline the easy bits of is_utf8_char() here for speed... */
e0328548
KW
406 if (UTF8_IS_INVARIANT(*x)) {
407 x++;
408 }
1acdb0da 409 else if (!UTF8_IS_START(*x))
e0328548 410 return FALSE;
1acdb0da
JH
411 else {
412 /* ... and call is_utf8_char() only if really needed. */
e0328548
KW
413 const STRLEN c = UTF8SKIP(x);
414 const U8* const next_char_ptr = x + c;
415
416 if (next_char_ptr > send) {
417 return FALSE;
418 }
419
768c67ee
JH
420 if (IS_UTF8_CHAR_FAST(c)) {
421 if (!IS_UTF8_CHAR(x, c))
e0328548 422 return FALSE;
3c614e38 423 }
e0328548
KW
424 else if (! is_utf8_char_slow(x, c)) {
425 return FALSE;
426 }
427 x = next_char_ptr;
1acdb0da 428 }
6662521e 429 }
768c67ee 430
067a85ef 431 return TRUE;
6662521e
GS
432}
433
67e989fb 434/*
814fafa7
NC
435Implemented as a macro in utf8.h
436
87cea99e 437=for apidoc is_utf8_string_loc
814fafa7 438
a1433954
KW
439Like L</is_utf8_string> but stores the location of the failure (in the
440case of "utf8ness failure") or the location C<s>+C<len> (in the case of
814fafa7
NC
441"utf8ness success") in the C<ep>.
442
a1433954 443See also L</is_utf8_string_loclen>() and L</is_utf8_string>().
814fafa7 444
87cea99e 445=for apidoc is_utf8_string_loclen
81cd54e3 446
a1433954
KW
447Like L</is_utf8_string>() but stores the location of the failure (in the
448case of "utf8ness failure") or the location C<s>+C<len> (in the case of
768c67ee
JH
449"utf8ness success") in the C<ep>, and the number of UTF-8
450encoded characters in the C<el>.
451
a1433954 452See also L</is_utf8_string_loc>() and L</is_utf8_string>().
81cd54e3
JH
453
454=cut
455*/
456
457bool
668b6d8d 458Perl_is_utf8_string_loclen(const U8 *s, STRLEN len, const U8 **ep, STRLEN *el)
81cd54e3 459{
35da51f7 460 const U8* const send = s + (len ? len : strlen((const char *)s));
7fc63493 461 const U8* x = s;
81cd54e3 462 STRLEN c;
3ebfea28 463 STRLEN outlen = 0;
7918f24d
NC
464
465 PERL_ARGS_ASSERT_IS_UTF8_STRING_LOCLEN;
81cd54e3 466
81cd54e3 467 while (x < send) {
e0328548
KW
468 const U8* next_char_ptr;
469
81cd54e3
JH
470 /* Inline the easy bits of is_utf8_char() here for speed... */
471 if (UTF8_IS_INVARIANT(*x))
e0328548 472 next_char_ptr = x + 1;
768c67ee
JH
473 else if (!UTF8_IS_START(*x))
474 goto out;
81cd54e3 475 else {
768c67ee 476 /* ... and call is_utf8_char() only if really needed. */
768c67ee 477 c = UTF8SKIP(x);
e0328548
KW
478 next_char_ptr = c + x;
479 if (next_char_ptr > send) {
480 goto out;
481 }
768c67ee
JH
482 if (IS_UTF8_CHAR_FAST(c)) {
483 if (!IS_UTF8_CHAR(x, c))
484 c = 0;
485 } else
486 c = is_utf8_char_slow(x, c);
768c67ee
JH
487 if (!c)
488 goto out;
81cd54e3 489 }
e0328548 490 x = next_char_ptr;
3ebfea28 491 outlen++;
81cd54e3 492 }
768c67ee
JH
493
494 out:
3ebfea28
AL
495 if (el)
496 *el = outlen;
497
768c67ee
JH
498 if (ep)
499 *ep = x;
3ebfea28 500 return (x == send);
81cd54e3
JH
501}
502
503/*
768c67ee 504
87cea99e 505=for apidoc utf8n_to_uvuni
67e989fb 506
9041c2e3 507Bottom level UTF-8 decode routine.
746afd53
KW
508Returns the code point value of the first character in the string C<s>,
509which is assumed to be in UTF-8 (or UTF-EBCDIC) encoding, and no longer than
510C<curlen> bytes; C<*retlen> (if C<retlen> isn't NULL) will be set to
511the length, in bytes, of that character.
949cf498
KW
512
513The value of C<flags> determines the behavior when C<s> does not point to a
514well-formed UTF-8 character. If C<flags> is 0, when a malformation is found,
515C<retlen> is set to the expected length of the UTF-8 character in bytes, zero
516is returned, and if UTF-8 warnings haven't been lexically disabled, a warning
517is raised.
518
519Various ALLOW flags can be set in C<flags> to allow (and not warn on)
520individual types of malformations, such as the sequence being overlong (that
521is, when there is a shorter sequence that can express the same code point;
522overlong sequences are expressly forbidden in the UTF-8 standard due to
523potential security issues). Another malformation example is the first byte of
524a character not being a legal first byte. See F<utf8.h> for the list of such
525flags. Of course, the value returned by this function under such conditions is
526not reliable.
527
528The UTF8_CHECK_ONLY flag overrides the behavior when a non-allowed (by other
529flags) malformation is found. If this flag is set, the routine assumes that
530the caller will raise a warning, and this function will silently just set
531C<retlen> to C<-1> and return zero.
532
533Certain code points are considered problematic. These are Unicode surrogates,
746afd53 534Unicode non-characters, and code points above the Unicode maximum of 0x10FFFF.
949cf498 535By default these are considered regular code points, but certain situations
5eafe189 536warrant special handling for them. If C<flags> contains
949cf498
KW
537UTF8_DISALLOW_ILLEGAL_INTERCHANGE, all three classes are treated as
538malformations and handled as such. The flags UTF8_DISALLOW_SURROGATE,
539UTF8_DISALLOW_NONCHAR, and UTF8_DISALLOW_SUPER (meaning above the legal Unicode
540maximum) can be set to disallow these categories individually.
541
542The flags UTF8_WARN_ILLEGAL_INTERCHANGE, UTF8_WARN_SURROGATE,
543UTF8_WARN_NONCHAR, and UTF8_WARN_SUPER will cause warning messages to be raised
544for their respective categories, but otherwise the code points are considered
545valid (not malformations). To get a category to both be treated as a
546malformation and raise a warning, specify both the WARN and DISALLOW flags.
547(But note that warnings are not raised if lexically disabled nor if
548UTF8_CHECK_ONLY is also specified.)
549
550Very large code points (above 0x7FFF_FFFF) are considered more problematic than
551the others that are above the Unicode legal maximum. There are several
d0460f30
KW
552reasons: they do not fit into a 32-bit word, are not representable on EBCDIC
553platforms, and the original UTF-8 specification never went above
746afd53 554this number (the current 0x10FFFF limit was imposed later). The UTF-8 encoding
5eafe189 555on ASCII platforms for these large code points begins with a byte containing
949cf498
KW
5560xFE or 0xFF. The UTF8_DISALLOW_FE_FF flag will cause them to be treated as
557malformations, while allowing smaller above-Unicode code points. (Of course
558UTF8_DISALLOW_SUPER will treat all above-Unicode code points, including these,
559as malformations.) Similarly, UTF8_WARN_FE_FF acts just like the other WARN
560flags, but applies just to these code points.
561
562All other code points corresponding to Unicode characters, including private
563use and those yet to be assigned, are never considered malformed and never
564warn.
67e989fb 565
ec5f19d0 566Most code should use L</utf8_to_uvchr_buf>() rather than call this directly.
9041c2e3 567
37607a96
PK
568=cut
569*/
67e989fb 570
a0ed51b3 571UV
7fc63493 572Perl_utf8n_to_uvuni(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen, U32 flags)
a0ed51b3 573{
97aff369 574 dVAR;
d4c19fe8 575 const U8 * const s0 = s;
0b8d30e8 576 U8 * send;
9c5ffd7c 577 UV uv = *s, ouv = 0;
ba210ebe 578 STRLEN len = 1;
949cf498 579 bool dowarn = ckWARN_d(WARN_UTF8);
7fc63493 580 const UV startbyte = *s;
ba210ebe 581 STRLEN expectlen = 0;
a0dbb045 582 U32 warning = 0;
949cf498 583 SV* sv = NULL;
a0dbb045 584
7918f24d
NC
585 PERL_ARGS_ASSERT_UTF8N_TO_UVUNI;
586
949cf498 587/* This list is a superset of the UTF8_ALLOW_XXX. */
a0dbb045
JH
588
589#define UTF8_WARN_EMPTY 1
590#define UTF8_WARN_CONTINUATION 2
591#define UTF8_WARN_NON_CONTINUATION 3
f9bc8ed7
KW
592#define UTF8_WARN_SHORT 4
593#define UTF8_WARN_OVERFLOW 5
594#define UTF8_WARN_LONG 6
a0dbb045
JH
595
596 if (curlen == 0 &&
597 !(flags & UTF8_ALLOW_EMPTY)) {
598 warning = UTF8_WARN_EMPTY;
0c443dc2
JH
599 goto malformed;
600 }
601
1d72bdf6 602 if (UTF8_IS_INVARIANT(uv)) {
a0ed51b3
LW
603 if (retlen)
604 *retlen = 1;
c4d5f83a 605 return (UV) (NATIVE_TO_UTF(*s));
a0ed51b3 606 }
67e989fb 607
421a8bf2 608 if (UTF8_IS_CONTINUATION(uv) &&
fcc8fcf6 609 !(flags & UTF8_ALLOW_CONTINUATION)) {
a0dbb045 610 warning = UTF8_WARN_CONTINUATION;
ba210ebe
JH
611 goto malformed;
612 }
613
421a8bf2 614 if (UTF8_IS_START(uv) && curlen > 1 && !UTF8_IS_CONTINUATION(s[1]) &&
fcc8fcf6 615 !(flags & UTF8_ALLOW_NON_CONTINUATION)) {
a0dbb045 616 warning = UTF8_WARN_NON_CONTINUATION;
ba210ebe
JH
617 goto malformed;
618 }
9041c2e3 619
1d72bdf6 620#ifdef EBCDIC
75383841 621 uv = NATIVE_TO_UTF(uv);
1d72bdf6 622#else
949cf498
KW
623 if (uv == 0xfe || uv == 0xff) {
624 if (flags & (UTF8_WARN_SUPER|UTF8_WARN_FE_FF)) {
111d382d 625 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "Code point beginning with byte 0x%02"UVXf" is not Unicode, and not portable", uv));
949cf498
KW
626 flags &= ~UTF8_WARN_SUPER; /* Only warn once on this problem */
627 }
628 if (flags & (UTF8_DISALLOW_SUPER|UTF8_DISALLOW_FE_FF)) {
629 goto malformed;
630 }
a0ed51b3 631 }
1d72bdf6
NIS
632#endif
633
99ee1dcd
KW
634 len = UTF8SKIP(s);
635 uv &= UTF_START_MASK(len);
1d72bdf6 636
a0ed51b3
LW
637 if (retlen)
638 *retlen = len;
9041c2e3 639
ba210ebe
JH
640 expectlen = len;
641
fcc8fcf6
JH
642 if ((curlen < expectlen) &&
643 !(flags & UTF8_ALLOW_SHORT)) {
a0dbb045 644 warning = UTF8_WARN_SHORT;
ba210ebe
JH
645 goto malformed;
646 }
647
0b8d30e8
KW
648 send = (U8*) s0 + ((expectlen <= curlen) ? expectlen : curlen);
649
949cf498 650 ouv = uv; /* ouv is the value from the previous iteration */
ba210ebe 651
0b8d30e8 652 for (++s; s < send; s++) {
421a8bf2
JH
653 if (!UTF8_IS_CONTINUATION(*s) &&
654 !(flags & UTF8_ALLOW_NON_CONTINUATION)) {
a0dbb045
JH
655 s--;
656 warning = UTF8_WARN_NON_CONTINUATION;
ba210ebe 657 goto malformed;
a0ed51b3
LW
658 }
659 else
8850bf83 660 uv = UTF8_ACCUMULATE(uv, *s);
949cf498
KW
661 if (!(uv > ouv)) { /* If the value didn't grow from the previous
662 iteration, something is horribly wrong */
a0dbb045
JH
663 /* These cannot be allowed. */
664 if (uv == ouv) {
75dbc644 665 if (expectlen != 13 && !(flags & UTF8_ALLOW_LONG)) {
a0dbb045
JH
666 warning = UTF8_WARN_LONG;
667 goto malformed;
668 }
669 }
670 else { /* uv < ouv */
671 /* This cannot be allowed. */
672 warning = UTF8_WARN_OVERFLOW;
673 goto malformed;
674 }
ba210ebe 675 }
ba210ebe
JH
676 ouv = uv;
677 }
678
949cf498 679 if ((expectlen > (STRLEN)UNISKIP(uv)) && !(flags & UTF8_ALLOW_LONG)) {
a0dbb045 680 warning = UTF8_WARN_LONG;
ba210ebe 681 goto malformed;
949cf498
KW
682 } else if (flags & (UTF8_DISALLOW_ILLEGAL_INTERCHANGE|UTF8_WARN_ILLEGAL_INTERCHANGE)) {
683 if (UNICODE_IS_SURROGATE(uv)) {
684 if ((flags & (UTF8_WARN_SURROGATE|UTF8_CHECK_ONLY)) == UTF8_WARN_SURROGATE) {
111d382d 685 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "UTF-16 surrogate U+%04"UVXf"", uv));
949cf498
KW
686 }
687 if (flags & UTF8_DISALLOW_SURROGATE) {
688 goto disallowed;
689 }
690 }
691 else if (UNICODE_IS_NONCHAR(uv)) {
692 if ((flags & (UTF8_WARN_NONCHAR|UTF8_CHECK_ONLY)) == UTF8_WARN_NONCHAR ) {
111d382d 693 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "Unicode non-character U+%04"UVXf" is illegal for open interchange", uv));
949cf498
KW
694 }
695 if (flags & UTF8_DISALLOW_NONCHAR) {
696 goto disallowed;
697 }
698 }
699 else if ((uv > PERL_UNICODE_MAX)) {
700 if ((flags & (UTF8_WARN_SUPER|UTF8_CHECK_ONLY)) == UTF8_WARN_SUPER) {
111d382d 701 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "Code point 0x%04"UVXf" is not Unicode, may not be portable", uv));
949cf498
KW
702 }
703 if (flags & UTF8_DISALLOW_SUPER) {
704 goto disallowed;
705 }
706 }
707
708 /* Here, this is not considered a malformed character, so drop through
709 * to return it */
a0ed51b3 710 }
ba210ebe 711
a0ed51b3 712 return uv;
ba210ebe 713
949cf498
KW
714disallowed: /* Is disallowed, but otherwise not malformed. 'sv' will have been
715 set if there is to be a warning. */
716 if (!sv) {
717 dowarn = 0;
718 }
719
ba210ebe
JH
720malformed:
721
fcc8fcf6 722 if (flags & UTF8_CHECK_ONLY) {
ba210ebe 723 if (retlen)
10edeb5d 724 *retlen = ((STRLEN) -1);
ba210ebe
JH
725 return 0;
726 }
727
a0dbb045 728 if (dowarn) {
949cf498 729 if (! sv) {
5b311467 730 sv = newSVpvs_flags("Malformed UTF-8 character ", SVs_TEMP);
949cf498 731 }
5b311467 732
5e3d7cf5
KW
733 switch (warning) {
734 case 0: /* Intentionally empty. */ break;
735 case UTF8_WARN_EMPTY:
736 sv_catpvs(sv, "(empty string)");
737 break;
738 case UTF8_WARN_CONTINUATION:
739 Perl_sv_catpvf(aTHX_ sv, "(unexpected continuation byte 0x%02"UVxf", with no preceding start byte)", uv);
740 break;
741 case UTF8_WARN_NON_CONTINUATION:
742 if (s == s0)
743 Perl_sv_catpvf(aTHX_ sv, "(unexpected non-continuation byte 0x%02"UVxf", immediately after start byte 0x%02"UVxf")",
744 (UV)s[1], startbyte);
745 else {
746 const int len = (int)(s-s0);
747 Perl_sv_catpvf(aTHX_ sv, "(unexpected non-continuation byte 0x%02"UVxf", %d byte%s after start byte 0x%02"UVxf", expected %d bytes)",
748 (UV)s[1], len, len > 1 ? "s" : "", startbyte, (int)expectlen);
749 }
750
751 break;
752 case UTF8_WARN_SHORT:
753 Perl_sv_catpvf(aTHX_ sv, "(%d byte%s, need %d, after start byte 0x%02"UVxf")",
754 (int)curlen, curlen == 1 ? "" : "s", (int)expectlen, startbyte);
755 expectlen = curlen; /* distance for caller to skip */
756 break;
757 case UTF8_WARN_OVERFLOW:
758 Perl_sv_catpvf(aTHX_ sv, "(overflow at 0x%"UVxf", byte 0x%02x, after start byte 0x%02"UVxf")",
759 ouv, *s, startbyte);
760 break;
761 case UTF8_WARN_LONG:
762 Perl_sv_catpvf(aTHX_ sv, "(%d byte%s, need %d, after start byte 0x%02"UVxf")",
763 (int)expectlen, expectlen == 1 ? "": "s", UNISKIP(uv), startbyte);
764 break;
765 default:
766 sv_catpvs(sv, "(unknown reason)");
767 break;
768 }
a0dbb045 769
949cf498 770 if (sv) {
44f8325f 771 const char * const s = SvPVX_const(sv);
a0dbb045
JH
772
773 if (PL_op)
9014280d 774 Perl_warner(aTHX_ packWARN(WARN_UTF8),
53e06cf0 775 "%s in %s", s, OP_DESC(PL_op));
a0dbb045 776 else
9014280d 777 Perl_warner(aTHX_ packWARN(WARN_UTF8), "%s", s);
a0dbb045
JH
778 }
779 }
780
ba210ebe 781 if (retlen)
28d3d195 782 *retlen = expectlen ? expectlen : len;
ba210ebe 783
28d3d195 784 return 0;
a0ed51b3
LW
785}
786
8e84507e 787/*
ec5f19d0
KW
788=for apidoc utf8_to_uvchr_buf
789
790Returns the native code point of the first character in the string C<s> which
791is assumed to be in UTF-8 encoding; C<send> points to 1 beyond the end of C<s>.
792C<retlen> will be set to the length, in bytes, of that character.
793
794If C<s> does not point to a well-formed UTF-8 character, zero is
795returned and C<retlen> is set, if possible, to -1.
796
797=cut
798*/
799
800
801UV
802Perl_utf8_to_uvchr_buf(pTHX_ const U8 *s, const U8 *send, STRLEN *retlen)
803{
804 PERL_ARGS_ASSERT_UTF8_TO_UVCHR_BUF;
805
806 assert(s < send);
807
808 return utf8n_to_uvchr(s, send - s, retlen,
809 ckWARN_d(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
810}
811
27d6c58a
KW
812/* Like L</utf8_to_uvchr_buf>(), but should only be called when it is known that
813 * there are no malformations in the input UTF-8 string C<s>. Currently, some
814 * malformations are checked for, but this checking likely will be removed in
815 * the future */
816
817UV
818Perl_valid_utf8_to_uvchr(pTHX_ const U8 *s, STRLEN *retlen)
819{
820 PERL_ARGS_ASSERT_VALID_UTF8_TO_UVCHR;
821
822 return utf8_to_uvchr_buf(s, s + UTF8_MAXBYTES, retlen);
823}
824
ec5f19d0 825/*
87cea99e 826=for apidoc utf8_to_uvchr
9041c2e3 827
977c1d31
KW
828DEPRECATED!
829
6ee84de2 830Returns the native code point of the first character in the string C<s>
1e54db1a 831which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
9041c2e3
NIS
832length, in bytes, of that character.
833
4b88fb76 834Some, but not all, UTF-8 malformations are detected, and in fact, some
977c1d31
KW
835malformed input could cause reading beyond the end of the input buffer, which
836is why this function is deprecated. Use L</utf8_to_uvchr_buf> instead.
4b88fb76
KW
837
838If C<s> points to one of the detected malformations, zero is
a1433954 839returned and C<retlen> is set, if possible, to -1.
9041c2e3
NIS
840
841=cut
842*/
843
844UV
7fc63493 845Perl_utf8_to_uvchr(pTHX_ const U8 *s, STRLEN *retlen)
9041c2e3 846{
7918f24d
NC
847 PERL_ARGS_ASSERT_UTF8_TO_UVCHR;
848
4b88fb76 849 return valid_utf8_to_uvchr(s, retlen);
9041c2e3
NIS
850}
851
852/*
ec5f19d0
KW
853=for apidoc utf8_to_uvuni_buf
854
855Returns the Unicode code point of the first character in the string C<s> which
856is assumed to be in UTF-8 encoding; C<send> points to 1 beyond the end of C<s>.
857C<retlen> will be set to the length, in bytes, of that character.
858
859This function should only be used when the returned UV is considered
860an index into the Unicode semantic tables (e.g. swashes).
861
862If C<s> does not point to a well-formed UTF-8 character, zero is
863returned and C<retlen> is set, if possible, to -1.
864
865=cut
866*/
867
868UV
869Perl_utf8_to_uvuni_buf(pTHX_ const U8 *s, const U8 *send, STRLEN *retlen)
870{
871 PERL_ARGS_ASSERT_UTF8_TO_UVUNI_BUF;
872
873 assert(send > s);
874
875 /* Call the low level routine asking for checks */
876 return Perl_utf8n_to_uvuni(aTHX_ s, send -s, retlen,
877 ckWARN_d(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
878}
879
27d6c58a
KW
880/* Like L</utf8_to_uvuni_buf>(), but should only be called when it is known that
881 * there are no malformations in the input UTF-8 string C<s>. Currently, some
882 * malformations are checked for, but this checking likely will be removed in
883 * the future */
884
885UV
886Perl_valid_utf8_to_uvuni(pTHX_ const U8 *s, STRLEN *retlen)
887{
888 PERL_ARGS_ASSERT_VALID_UTF8_TO_UVUNI;
889
890 return utf8_to_uvuni_buf(s, s + UTF8_MAXBYTES, retlen);
891}
892
ec5f19d0 893/*
87cea99e 894=for apidoc utf8_to_uvuni
9041c2e3 895
977c1d31
KW
896DEPRECATED!
897
9041c2e3 898Returns the Unicode code point of the first character in the string C<s>
1e54db1a 899which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
9041c2e3
NIS
900length, in bytes, of that character.
901
39e518fd
KW
902This function should only be used when the returned UV is considered
903an index into the Unicode semantic tables (e.g. swashes).
904
4b88fb76 905Some, but not all, UTF-8 malformations are detected, and in fact, some
977c1d31
KW
906malformed input could cause reading beyond the end of the input buffer, which
907is why this function is deprecated. Use L</utf8_to_uvuni_buf> instead.
9041c2e3 908
4b88fb76 909If C<s> points to one of the detected malformations, zero is
a1433954 910returned and C<retlen> is set, if possible, to -1.
8e84507e
NIS
911
912=cut
913*/
914
915UV
7fc63493 916Perl_utf8_to_uvuni(pTHX_ const U8 *s, STRLEN *retlen)
8e84507e 917{
7918f24d
NC
918 PERL_ARGS_ASSERT_UTF8_TO_UVUNI;
919
4b88fb76 920 return valid_utf8_to_uvuni(s, retlen);
8e84507e
NIS
921}
922
b76347f2 923/*
87cea99e 924=for apidoc utf8_length
b76347f2
JH
925
926Return the length of the UTF-8 char encoded string C<s> in characters.
02eb7b47
JH
927Stops at C<e> (inclusive). If C<e E<lt> s> or if the scan would end
928up past C<e>, croaks.
b76347f2
JH
929
930=cut
931*/
932
933STRLEN
35a4481c 934Perl_utf8_length(pTHX_ const U8 *s, const U8 *e)
b76347f2 935{
97aff369 936 dVAR;
b76347f2
JH
937 STRLEN len = 0;
938
7918f24d
NC
939 PERL_ARGS_ASSERT_UTF8_LENGTH;
940
8850bf83
JH
941 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g.
942 * the bitops (especially ~) can create illegal UTF-8.
943 * In other words: in Perl UTF-8 is not just for Unicode. */
944
a3b680e6
AL
945 if (e < s)
946 goto warn_and_return;
b76347f2 947 while (s < e) {
8e91ec7f
AV
948 if (!UTF8_IS_INVARIANT(*s))
949 s += UTF8SKIP(s);
950 else
951 s++;
952 len++;
953 }
954
955 if (e != s) {
956 len--;
957 warn_and_return:
9b387841
NC
958 if (PL_op)
959 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8),
960 "%s in %s", unees, OP_DESC(PL_op));
961 else
61a12c31 962 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), "%s", unees);
b76347f2
JH
963 }
964
965 return len;
966}
967
b06226ff 968/*
87cea99e 969=for apidoc utf8_distance
b06226ff 970
1e54db1a 971Returns the number of UTF-8 characters between the UTF-8 pointers C<a>
b06226ff
JH
972and C<b>.
973
974WARNING: use only if you *know* that the pointers point inside the
975same UTF-8 buffer.
976
37607a96
PK
977=cut
978*/
a0ed51b3 979
02eb7b47 980IV
35a4481c 981Perl_utf8_distance(pTHX_ const U8 *a, const U8 *b)
a0ed51b3 982{
7918f24d
NC
983 PERL_ARGS_ASSERT_UTF8_DISTANCE;
984
bf1665bc 985 return (a < b) ? -1 * (IV) utf8_length(a, b) : (IV) utf8_length(b, a);
a0ed51b3
LW
986}
987
b06226ff 988/*
87cea99e 989=for apidoc utf8_hop
b06226ff 990
8850bf83
JH
991Return the UTF-8 pointer C<s> displaced by C<off> characters, either
992forward or backward.
b06226ff
JH
993
994WARNING: do not use the following unless you *know* C<off> is within
8850bf83
JH
995the UTF-8 data pointed to by C<s> *and* that on entry C<s> is aligned
996on the first byte of character or just after the last byte of a character.
b06226ff 997
37607a96
PK
998=cut
999*/
a0ed51b3
LW
1000
1001U8 *
4373e329 1002Perl_utf8_hop(pTHX_ const U8 *s, I32 off)
a0ed51b3 1003{
7918f24d
NC
1004 PERL_ARGS_ASSERT_UTF8_HOP;
1005
96a5add6 1006 PERL_UNUSED_CONTEXT;
8850bf83
JH
1007 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g
1008 * the bitops (especially ~) can create illegal UTF-8.
1009 * In other words: in Perl UTF-8 is not just for Unicode. */
1010
a0ed51b3
LW
1011 if (off >= 0) {
1012 while (off--)
1013 s += UTF8SKIP(s);
1014 }
1015 else {
1016 while (off++) {
1017 s--;
8850bf83
JH
1018 while (UTF8_IS_CONTINUATION(*s))
1019 s--;
a0ed51b3
LW
1020 }
1021 }
4373e329 1022 return (U8 *)s;
a0ed51b3
LW
1023}
1024
6940069f 1025/*
fed3ba5d
NC
1026=for apidoc bytes_cmp_utf8
1027
a1433954
KW
1028Compares the sequence of characters (stored as octets) in C<b>, C<blen> with the
1029sequence of characters (stored as UTF-8) in C<u>, C<ulen>. Returns 0 if they are
fed3ba5d
NC
1030equal, -1 or -2 if the first string is less than the second string, +1 or +2
1031if the first string is greater than the second string.
1032
1033-1 or +1 is returned if the shorter string was identical to the start of the
1034longer string. -2 or +2 is returned if the was a difference between characters
1035within the strings.
1036
1037=cut
1038*/
1039
1040int
1041Perl_bytes_cmp_utf8(pTHX_ const U8 *b, STRLEN blen, const U8 *u, STRLEN ulen)
1042{
1043 const U8 *const bend = b + blen;
1044 const U8 *const uend = u + ulen;
1045
1046 PERL_ARGS_ASSERT_BYTES_CMP_UTF8;
1047
1048 PERL_UNUSED_CONTEXT;
1049
1050 while (b < bend && u < uend) {
1051 U8 c = *u++;
1052 if (!UTF8_IS_INVARIANT(c)) {
1053 if (UTF8_IS_DOWNGRADEABLE_START(c)) {
1054 if (u < uend) {
1055 U8 c1 = *u++;
1056 if (UTF8_IS_CONTINUATION(c1)) {
356979f4 1057 c = UNI_TO_NATIVE(TWO_BYTE_UTF8_TO_UNI(c, c1));
fed3ba5d
NC
1058 } else {
1059 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8),
1060 "Malformed UTF-8 character "
1061 "(unexpected non-continuation byte 0x%02x"
1062 ", immediately after start byte 0x%02x)"
1063 /* Dear diag.t, it's in the pod. */
1064 "%s%s", c1, c,
1065 PL_op ? " in " : "",
1066 PL_op ? OP_DESC(PL_op) : "");
1067 return -2;
1068 }
1069 } else {
1070 if (PL_op)
1071 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8),
1072 "%s in %s", unees, OP_DESC(PL_op));
1073 else
61a12c31 1074 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), "%s", unees);
fed3ba5d
NC
1075 return -2; /* Really want to return undef :-) */
1076 }
1077 } else {
1078 return -2;
1079 }
1080 }
1081 if (*b != c) {
1082 return *b < c ? -2 : +2;
1083 }
1084 ++b;
1085 }
1086
1087 if (b == bend && u == uend)
1088 return 0;
1089
1090 return b < bend ? +1 : -1;
1091}
1092
1093/*
87cea99e 1094=for apidoc utf8_to_bytes
6940069f 1095
2bbc8d55 1096Converts a string C<s> of length C<len> from UTF-8 into native byte encoding.
a1433954
KW
1097Unlike L</bytes_to_utf8>, this over-writes the original string, and
1098updates C<len> to contain the new length.
67e989fb 1099Returns zero on failure, setting C<len> to -1.
6940069f 1100
a1433954 1101If you need a copy of the string, see L</bytes_from_utf8>.
95be277c 1102
6940069f
GS
1103=cut
1104*/
1105
1106U8 *
37607a96 1107Perl_utf8_to_bytes(pTHX_ U8 *s, STRLEN *len)
6940069f 1108{
d4c19fe8
AL
1109 U8 * const save = s;
1110 U8 * const send = s + *len;
6940069f 1111 U8 *d;
246fae53 1112
7918f24d
NC
1113 PERL_ARGS_ASSERT_UTF8_TO_BYTES;
1114
1e54db1a 1115 /* ensure valid UTF-8 and chars < 256 before updating string */
d4c19fe8 1116 while (s < send) {
dcad2880
JH
1117 U8 c = *s++;
1118
1d72bdf6
NIS
1119 if (!UTF8_IS_INVARIANT(c) &&
1120 (!UTF8_IS_DOWNGRADEABLE_START(c) || (s >= send)
1121 || !(c = *s++) || !UTF8_IS_CONTINUATION(c))) {
10edeb5d 1122 *len = ((STRLEN) -1);
dcad2880
JH
1123 return 0;
1124 }
246fae53 1125 }
dcad2880
JH
1126
1127 d = s = save;
6940069f 1128 while (s < send) {
ed646e6e 1129 STRLEN ulen;
4b88fb76 1130 *d++ = (U8)utf8_to_uvchr_buf(s, send, &ulen);
ed646e6e 1131 s += ulen;
6940069f
GS
1132 }
1133 *d = '\0';
246fae53 1134 *len = d - save;
6940069f
GS
1135 return save;
1136}
1137
1138/*
87cea99e 1139=for apidoc bytes_from_utf8
f9a63242 1140
2bbc8d55 1141Converts a string C<s> of length C<len> from UTF-8 into native byte encoding.
a1433954 1142Unlike L</utf8_to_bytes> but like L</bytes_to_utf8>, returns a pointer to
ef9edfd0
JH
1143the newly-created string, and updates C<len> to contain the new
1144length. Returns the original string if no conversion occurs, C<len>
1145is unchanged. Do nothing if C<is_utf8> points to 0. Sets C<is_utf8> to
2bbc8d55
SP
11460 if C<s> is converted or consisted entirely of characters that are invariant
1147in utf8 (i.e., US-ASCII on non-EBCDIC machines).
f9a63242 1148
37607a96
PK
1149=cut
1150*/
f9a63242
JH
1151
1152U8 *
e1ec3a88 1153Perl_bytes_from_utf8(pTHX_ const U8 *s, STRLEN *len, bool *is_utf8)
f9a63242 1154{
f9a63242 1155 U8 *d;
e1ec3a88
AL
1156 const U8 *start = s;
1157 const U8 *send;
f9a63242
JH
1158 I32 count = 0;
1159
7918f24d
NC
1160 PERL_ARGS_ASSERT_BYTES_FROM_UTF8;
1161
96a5add6 1162 PERL_UNUSED_CONTEXT;
f9a63242 1163 if (!*is_utf8)
73d840c0 1164 return (U8 *)start;
f9a63242 1165
1e54db1a 1166 /* ensure valid UTF-8 and chars < 256 before converting string */
f9a63242 1167 for (send = s + *len; s < send;) {
e1ec3a88 1168 U8 c = *s++;
1d72bdf6 1169 if (!UTF8_IS_INVARIANT(c)) {
db42d148
NIS
1170 if (UTF8_IS_DOWNGRADEABLE_START(c) && s < send &&
1171 (c = *s++) && UTF8_IS_CONTINUATION(c))
1172 count++;
1173 else
73d840c0 1174 return (U8 *)start;
db42d148 1175 }
f9a63242
JH
1176 }
1177
35da51f7 1178 *is_utf8 = FALSE;
f9a63242 1179
212542aa 1180 Newx(d, (*len) - count + 1, U8);
ef9edfd0 1181 s = start; start = d;
f9a63242
JH
1182 while (s < send) {
1183 U8 c = *s++;
c4d5f83a
NIS
1184 if (!UTF8_IS_INVARIANT(c)) {
1185 /* Then it is two-byte encoded */
356979f4 1186 c = UNI_TO_NATIVE(TWO_BYTE_UTF8_TO_UNI(c, *s++));
c4d5f83a
NIS
1187 }
1188 *d++ = c;
f9a63242
JH
1189 }
1190 *d = '\0';
1191 *len = d - start;
73d840c0 1192 return (U8 *)start;
f9a63242
JH
1193}
1194
1195/*
87cea99e 1196=for apidoc bytes_to_utf8
6940069f 1197
ff97e5cf
KW
1198Converts a string C<s> of length C<len> bytes from the native encoding into
1199UTF-8.
6662521e 1200Returns a pointer to the newly-created string, and sets C<len> to
ff97e5cf 1201reflect the new length in bytes.
6940069f 1202
2bbc8d55
SP
1203A NUL character will be written after the end of the string.
1204
1205If you want to convert to UTF-8 from encodings other than
1206the native (Latin1 or EBCDIC),
a1433954 1207see L</sv_recode_to_utf8>().
c9ada85f 1208
497711e7 1209=cut
6940069f
GS
1210*/
1211
c682ebef
FC
1212/* This logic is duplicated in sv_catpvn_flags, so any bug fixes will
1213 likewise need duplication. */
1214
6940069f 1215U8*
35a4481c 1216Perl_bytes_to_utf8(pTHX_ const U8 *s, STRLEN *len)
6940069f 1217{
35a4481c 1218 const U8 * const send = s + (*len);
6940069f
GS
1219 U8 *d;
1220 U8 *dst;
7918f24d
NC
1221
1222 PERL_ARGS_ASSERT_BYTES_TO_UTF8;
96a5add6 1223 PERL_UNUSED_CONTEXT;
6940069f 1224
212542aa 1225 Newx(d, (*len) * 2 + 1, U8);
6940069f
GS
1226 dst = d;
1227
1228 while (s < send) {
35a4481c 1229 const UV uv = NATIVE_TO_ASCII(*s++);
c4d5f83a 1230 if (UNI_IS_INVARIANT(uv))
eb160463 1231 *d++ = (U8)UTF_TO_NATIVE(uv);
6940069f 1232 else {
eb160463
GS
1233 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
1234 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
6940069f
GS
1235 }
1236 }
1237 *d = '\0';
6662521e 1238 *len = d-dst;
6940069f
GS
1239 return dst;
1240}
1241
a0ed51b3 1242/*
dea0fc0b 1243 * Convert native (big-endian) or reversed (little-endian) UTF-16 to UTF-8.
a0ed51b3
LW
1244 *
1245 * Destination must be pre-extended to 3/2 source. Do not use in-place.
1246 * We optimize for native, for obvious reasons. */
1247
1248U8*
dea0fc0b 1249Perl_utf16_to_utf8(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen)
a0ed51b3 1250{
dea0fc0b
JH
1251 U8* pend;
1252 U8* dstart = d;
1253
7918f24d
NC
1254 PERL_ARGS_ASSERT_UTF16_TO_UTF8;
1255
dea0fc0b 1256 if (bytelen & 1)
f5992bc4 1257 Perl_croak(aTHX_ "panic: utf16_to_utf8: odd bytelen %"UVuf, (UV)bytelen);
dea0fc0b
JH
1258
1259 pend = p + bytelen;
1260
a0ed51b3 1261 while (p < pend) {
dea0fc0b
JH
1262 UV uv = (p[0] << 8) + p[1]; /* UTF-16BE */
1263 p += 2;
a0ed51b3 1264 if (uv < 0x80) {
e294cc5d
JH
1265#ifdef EBCDIC
1266 *d++ = UNI_TO_NATIVE(uv);
1267#else
eb160463 1268 *d++ = (U8)uv;
e294cc5d 1269#endif
a0ed51b3
LW
1270 continue;
1271 }
1272 if (uv < 0x800) {
eb160463
GS
1273 *d++ = (U8)(( uv >> 6) | 0xc0);
1274 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
1275 continue;
1276 }
52b9aa85 1277 if (uv >= 0xd800 && uv <= 0xdbff) { /* surrogates */
01ea242b 1278 if (p >= pend) {
dea0fc0b 1279 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
01ea242b
NC
1280 } else {
1281 UV low = (p[0] << 8) + p[1];
1282 p += 2;
52b9aa85 1283 if (low < 0xdc00 || low > 0xdfff)
01ea242b
NC
1284 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
1285 uv = ((uv - 0xd800) << 10) + (low - 0xdc00) + 0x10000;
1286 }
dbde1951
NC
1287 } else if (uv >= 0xdc00 && uv <= 0xdfff) {
1288 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
a0ed51b3
LW
1289 }
1290 if (uv < 0x10000) {
eb160463
GS
1291 *d++ = (U8)(( uv >> 12) | 0xe0);
1292 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
1293 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
1294 continue;
1295 }
1296 else {
eb160463
GS
1297 *d++ = (U8)(( uv >> 18) | 0xf0);
1298 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
1299 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
1300 *d++ = (U8)(( uv & 0x3f) | 0x80);
a0ed51b3
LW
1301 continue;
1302 }
1303 }
dea0fc0b 1304 *newlen = d - dstart;
a0ed51b3
LW
1305 return d;
1306}
1307
1308/* Note: this one is slightly destructive of the source. */
1309
1310U8*
dea0fc0b 1311Perl_utf16_to_utf8_reversed(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen)
a0ed51b3
LW
1312{
1313 U8* s = (U8*)p;
d4c19fe8 1314 U8* const send = s + bytelen;
7918f24d
NC
1315
1316 PERL_ARGS_ASSERT_UTF16_TO_UTF8_REVERSED;
1317
e0ea5e2d
NC
1318 if (bytelen & 1)
1319 Perl_croak(aTHX_ "panic: utf16_to_utf8_reversed: odd bytelen %"UVuf,
1320 (UV)bytelen);
1321
a0ed51b3 1322 while (s < send) {
d4c19fe8 1323 const U8 tmp = s[0];
a0ed51b3
LW
1324 s[0] = s[1];
1325 s[1] = tmp;
1326 s += 2;
1327 }
dea0fc0b 1328 return utf16_to_utf8(p, d, bytelen, newlen);
a0ed51b3
LW
1329}
1330
c3fd2246
KW
1331/* for now these are all defined (inefficiently) in terms of the utf8 versions.
1332 * Note that the macros in handy.h that call these short-circuit calling them
1333 * for Latin-1 range inputs */
a0ed51b3
LW
1334
1335bool
84afefe6 1336Perl_is_uni_alnum(pTHX_ UV c)
a0ed51b3 1337{
89ebb4a3 1338 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1339 uvchr_to_utf8(tmpbuf, c);
a0ed51b3
LW
1340 return is_utf8_alnum(tmpbuf);
1341}
1342
1343bool
84afefe6 1344Perl_is_uni_idfirst(pTHX_ UV c)
a0ed51b3 1345{
89ebb4a3 1346 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1347 uvchr_to_utf8(tmpbuf, c);
a0ed51b3
LW
1348 return is_utf8_idfirst(tmpbuf);
1349}
1350
1351bool
84afefe6 1352Perl_is_uni_alpha(pTHX_ UV c)
a0ed51b3 1353{
89ebb4a3 1354 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1355 uvchr_to_utf8(tmpbuf, c);
a0ed51b3
LW
1356 return is_utf8_alpha(tmpbuf);
1357}
1358
1359bool
84afefe6 1360Perl_is_uni_ascii(pTHX_ UV c)
4d61ec05 1361{
bc39fe24 1362 return isASCII(c);
4d61ec05
GS
1363}
1364
1365bool
84afefe6 1366Perl_is_uni_space(pTHX_ UV c)
a0ed51b3 1367{
89ebb4a3 1368 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1369 uvchr_to_utf8(tmpbuf, c);
a0ed51b3
LW
1370 return is_utf8_space(tmpbuf);
1371}
1372
1373bool
84afefe6 1374Perl_is_uni_digit(pTHX_ UV c)
a0ed51b3 1375{
89ebb4a3 1376 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1377 uvchr_to_utf8(tmpbuf, c);
a0ed51b3
LW
1378 return is_utf8_digit(tmpbuf);
1379}
1380
1381bool
84afefe6 1382Perl_is_uni_upper(pTHX_ UV c)
a0ed51b3 1383{
89ebb4a3 1384 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1385 uvchr_to_utf8(tmpbuf, c);
a0ed51b3
LW
1386 return is_utf8_upper(tmpbuf);
1387}
1388
1389bool
84afefe6 1390Perl_is_uni_lower(pTHX_ UV c)
a0ed51b3 1391{
89ebb4a3 1392 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1393 uvchr_to_utf8(tmpbuf, c);
a0ed51b3
LW
1394 return is_utf8_lower(tmpbuf);
1395}
1396
1397bool
84afefe6 1398Perl_is_uni_cntrl(pTHX_ UV c)
b8c5462f 1399{
7b952154 1400 return isCNTRL_L1(c);
b8c5462f
JH
1401}
1402
1403bool
84afefe6 1404Perl_is_uni_graph(pTHX_ UV c)
b8c5462f 1405{
89ebb4a3 1406 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1407 uvchr_to_utf8(tmpbuf, c);
b8c5462f
JH
1408 return is_utf8_graph(tmpbuf);
1409}
1410
1411bool
84afefe6 1412Perl_is_uni_print(pTHX_ UV c)
a0ed51b3 1413{
89ebb4a3 1414 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1415 uvchr_to_utf8(tmpbuf, c);
a0ed51b3
LW
1416 return is_utf8_print(tmpbuf);
1417}
1418
b8c5462f 1419bool
84afefe6 1420Perl_is_uni_punct(pTHX_ UV c)
b8c5462f 1421{
89ebb4a3 1422 U8 tmpbuf[UTF8_MAXBYTES+1];
230880c1 1423 uvchr_to_utf8(tmpbuf, c);
b8c5462f
JH
1424 return is_utf8_punct(tmpbuf);
1425}
1426
4d61ec05 1427bool
84afefe6 1428Perl_is_uni_xdigit(pTHX_ UV c)
4d61ec05 1429{
89ebb4a3 1430 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
230880c1 1431 uvchr_to_utf8(tmpbuf, c);
4d61ec05
GS
1432 return is_utf8_xdigit(tmpbuf);
1433}
1434
3a4c58c9
KW
1435UV
1436Perl__to_upper_title_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp, const char S_or_s)
1437{
1438 /* We have the latin1-range values compiled into the core, so just use
1439 * those, converting the result to utf8. The only difference between upper
1440 * and title case in this range is that LATIN_SMALL_LETTER_SHARP_S is
1441 * either "SS" or "Ss". Which one to use is passed into the routine in
1442 * 'S_or_s' to avoid a test */
1443
1444 UV converted = toUPPER_LATIN1_MOD(c);
1445
1446 PERL_ARGS_ASSERT__TO_UPPER_TITLE_LATIN1;
1447
1448 assert(S_or_s == 'S' || S_or_s == 's');
1449
1450 if (UNI_IS_INVARIANT(converted)) { /* No difference between the two for
1451 characters in this range */
1452 *p = (U8) converted;
1453 *lenp = 1;
1454 return converted;
1455 }
1456
1457 /* toUPPER_LATIN1_MOD gives the correct results except for three outliers,
1458 * which it maps to one of them, so as to only have to have one check for
1459 * it in the main case */
1460 if (UNLIKELY(converted == LATIN_SMALL_LETTER_Y_WITH_DIAERESIS)) {
1461 switch (c) {
1462 case LATIN_SMALL_LETTER_Y_WITH_DIAERESIS:
1463 converted = LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS;
1464 break;
1465 case MICRO_SIGN:
1466 converted = GREEK_CAPITAL_LETTER_MU;
1467 break;
1468 case LATIN_SMALL_LETTER_SHARP_S:
1469 *(p)++ = 'S';
1470 *p = S_or_s;
1471 *lenp = 2;
1472 return 'S';
1473 default:
1474 Perl_croak(aTHX_ "panic: to_upper_title_latin1 did not expect '%c' to map to '%c'", c, LATIN_SMALL_LETTER_Y_WITH_DIAERESIS);
1475 /* NOTREACHED */
1476 }
1477 }
1478
1479 *(p)++ = UTF8_TWO_BYTE_HI(converted);
1480 *p = UTF8_TWO_BYTE_LO(converted);
1481 *lenp = 2;
1482
1483 return converted;
1484}
1485
50bda2c3
KW
1486/* Call the function to convert a UTF-8 encoded character to the specified case.
1487 * Note that there may be more than one character in the result.
1488 * INP is a pointer to the first byte of the input character
1489 * OUTP will be set to the first byte of the string of changed characters. It
1490 * needs to have space for UTF8_MAXBYTES_CASE+1 bytes
1491 * LENP will be set to the length in bytes of the string of changed characters
1492 *
1493 * The functions return the ordinal of the first character in the string of OUTP */
f90a9a02
KW
1494#define CALL_UPPER_CASE(INP, OUTP, LENP) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_toupper, "ToUc", "utf8::ToSpecUc")
1495#define CALL_TITLE_CASE(INP, OUTP, LENP) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_totitle, "ToTc", "utf8::ToSpecTc")
1496#define CALL_LOWER_CASE(INP, OUTP, LENP) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_tolower, "ToLc", "utf8::ToSpecLc")
50bda2c3
KW
1497
1498/* This additionally has the input parameter SPECIALS, which if non-zero will
1499 * cause this to use the SPECIALS hash for folding (meaning get full case
1500 * folding); otherwise, when zero, this implies a simple case fold */
f90a9a02 1501#define CALL_FOLD_CASE(INP, OUTP, LENP, SPECIALS) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_tofold, "ToCf", (SPECIALS) ? "utf8::ToSpecCf" : NULL)
c3fd2246 1502
84afefe6
JH
1503UV
1504Perl_to_uni_upper(pTHX_ UV c, U8* p, STRLEN *lenp)
a0ed51b3 1505{
3a4c58c9
KW
1506 dVAR;
1507
a1433954
KW
1508 /* Convert the Unicode character whose ordinal is <c> to its uppercase
1509 * version and store that in UTF-8 in <p> and its length in bytes in <lenp>.
1510 * Note that the <p> needs to be at least UTF8_MAXBYTES_CASE+1 bytes since
c3fd2246
KW
1511 * the changed version may be longer than the original character.
1512 *
1513 * The ordinal of the first character of the changed version is returned
1514 * (but note, as explained above, that there may be more.) */
1515
7918f24d
NC
1516 PERL_ARGS_ASSERT_TO_UNI_UPPER;
1517
3a4c58c9
KW
1518 if (c < 256) {
1519 return _to_upper_title_latin1((U8) c, p, lenp, 'S');
1520 }
1521
0ebc6274 1522 uvchr_to_utf8(p, c);
3a4c58c9 1523 return CALL_UPPER_CASE(p, p, lenp);
a0ed51b3
LW
1524}
1525
84afefe6
JH
1526UV
1527Perl_to_uni_title(pTHX_ UV c, U8* p, STRLEN *lenp)
a0ed51b3 1528{
3a4c58c9
KW
1529 dVAR;
1530
7918f24d
NC
1531 PERL_ARGS_ASSERT_TO_UNI_TITLE;
1532
3a4c58c9
KW
1533 if (c < 256) {
1534 return _to_upper_title_latin1((U8) c, p, lenp, 's');
1535 }
1536
0ebc6274 1537 uvchr_to_utf8(p, c);
3a4c58c9 1538 return CALL_TITLE_CASE(p, p, lenp);
a0ed51b3
LW
1539}
1540
afc16117
KW
1541STATIC U8
1542S_to_lower_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp)
1543{
1544 /* We have the latin1-range values compiled into the core, so just use
1545 * those, converting the result to utf8. Since the result is always just
a1433954 1546 * one character, we allow <p> to be NULL */
afc16117
KW
1547
1548 U8 converted = toLOWER_LATIN1(c);
1549
1550 if (p != NULL) {
1551 if (UNI_IS_INVARIANT(converted)) {
1552 *p = converted;
1553 *lenp = 1;
1554 }
1555 else {
1556 *p = UTF8_TWO_BYTE_HI(converted);
1557 *(p+1) = UTF8_TWO_BYTE_LO(converted);
1558 *lenp = 2;
1559 }
1560 }
1561 return converted;
1562}
1563
84afefe6
JH
1564UV
1565Perl_to_uni_lower(pTHX_ UV c, U8* p, STRLEN *lenp)
a0ed51b3 1566{
968c5e6a
KW
1567 dVAR;
1568
7918f24d
NC
1569 PERL_ARGS_ASSERT_TO_UNI_LOWER;
1570
afc16117
KW
1571 if (c < 256) {
1572 return to_lower_latin1((U8) c, p, lenp);
bca00c02
KW
1573 }
1574
afc16117 1575 uvchr_to_utf8(p, c);
968c5e6a 1576 return CALL_LOWER_CASE(p, p, lenp);
a0ed51b3
LW
1577}
1578
84afefe6 1579UV
f673fad4 1580Perl__to_fold_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp, const bool flags)
a1dde8de 1581{
a1433954 1582 /* Corresponds to to_lower_latin1(), <flags> is TRUE if to use full case
f673fad4
KW
1583 * folding */
1584
a1dde8de
KW
1585 UV converted;
1586
1587 PERL_ARGS_ASSERT__TO_FOLD_LATIN1;
1588
1589 if (c == MICRO_SIGN) {
1590 converted = GREEK_SMALL_LETTER_MU;
1591 }
1592 else if (flags && c == LATIN_SMALL_LETTER_SHARP_S) {
1593 *(p)++ = 's';
1594 *p = 's';
1595 *lenp = 2;
1596 return 's';
1597 }
1598 else { /* In this range the fold of all other characters is their lower
1599 case */
1600 converted = toLOWER_LATIN1(c);
1601 }
1602
1603 if (UNI_IS_INVARIANT(converted)) {
1604 *p = (U8) converted;
1605 *lenp = 1;
1606 }
1607 else {
1608 *(p)++ = UTF8_TWO_BYTE_HI(converted);
1609 *p = UTF8_TWO_BYTE_LO(converted);
1610 *lenp = 2;
1611 }
1612
1613 return converted;
1614}
1615
1616UV
f673fad4 1617Perl__to_uni_fold_flags(pTHX_ UV c, U8* p, STRLEN *lenp, const bool flags)
84afefe6 1618{
4b593389
KW
1619
1620 /* Not currently externally documented, and subject to change, <flags> is
f673fad4 1621 * TRUE iff full folding is to be used */
4b593389 1622
36bb2ab6 1623 PERL_ARGS_ASSERT__TO_UNI_FOLD_FLAGS;
7918f24d 1624
a1dde8de
KW
1625 if (c < 256) {
1626 return _to_fold_latin1((U8) c, p, lenp, flags);
1627 }
1628
0ebc6274 1629 uvchr_to_utf8(p, c);
a1dde8de 1630 return CALL_FOLD_CASE(p, p, lenp, flags);
84afefe6
JH
1631}
1632
ea317ccb
KW
1633/* for now these all assume no locale info available for Unicode > 255; and
1634 * the corresponding macros in handy.h (like isALNUM_LC_uvchr) should have been
1635 * called instead, so that these don't get called for < 255 */
a0ed51b3
LW
1636
1637bool
84afefe6 1638Perl_is_uni_alnum_lc(pTHX_ UV c)
a0ed51b3
LW
1639{
1640 return is_uni_alnum(c); /* XXX no locale support yet */
1641}
1642
1643bool
84afefe6 1644Perl_is_uni_idfirst_lc(pTHX_ UV c)
a0ed51b3
LW
1645{
1646 return is_uni_idfirst(c); /* XXX no locale support yet */
1647}
1648
1649bool
84afefe6 1650Perl_is_uni_alpha_lc(pTHX_ UV c)
a0ed51b3
LW
1651{
1652 return is_uni_alpha(c); /* XXX no locale support yet */
1653}
1654
1655bool
84afefe6 1656Perl_is_uni_ascii_lc(pTHX_ UV c)
4d61ec05
GS
1657{
1658 return is_uni_ascii(c); /* XXX no locale support yet */
1659}
1660
1661bool
84afefe6 1662Perl_is_uni_space_lc(pTHX_ UV c)
a0ed51b3
LW
1663{
1664 return is_uni_space(c); /* XXX no locale support yet */
1665}
1666
1667bool
84afefe6 1668Perl_is_uni_digit_lc(pTHX_ UV c)
a0ed51b3
LW
1669{
1670 return is_uni_digit(c); /* XXX no locale support yet */
1671}
1672
1673bool
84afefe6 1674Perl_is_uni_upper_lc(pTHX_ UV c)
a0ed51b3
LW
1675{
1676 return is_uni_upper(c); /* XXX no locale support yet */
1677}
1678
1679bool
84afefe6 1680Perl_is_uni_lower_lc(pTHX_ UV c)
a0ed51b3
LW
1681{
1682 return is_uni_lower(c); /* XXX no locale support yet */
1683}
1684
1685bool
84afefe6 1686Perl_is_uni_cntrl_lc(pTHX_ UV c)
b8c5462f
JH
1687{
1688 return is_uni_cntrl(c); /* XXX no locale support yet */
1689}
1690
1691bool
84afefe6 1692Perl_is_uni_graph_lc(pTHX_ UV c)
b8c5462f
JH
1693{
1694 return is_uni_graph(c); /* XXX no locale support yet */
1695}
1696
1697bool
84afefe6 1698Perl_is_uni_print_lc(pTHX_ UV c)
a0ed51b3
LW
1699{
1700 return is_uni_print(c); /* XXX no locale support yet */
1701}
1702
b8c5462f 1703bool
84afefe6 1704Perl_is_uni_punct_lc(pTHX_ UV c)
b8c5462f
JH
1705{
1706 return is_uni_punct(c); /* XXX no locale support yet */
1707}
1708
4d61ec05 1709bool
84afefe6 1710Perl_is_uni_xdigit_lc(pTHX_ UV c)
4d61ec05
GS
1711{
1712 return is_uni_xdigit(c); /* XXX no locale support yet */
1713}
1714
b7ac61fa
JH
1715U32
1716Perl_to_uni_upper_lc(pTHX_ U32 c)
1717{
ee099d14
JH
1718 /* XXX returns only the first character -- do not use XXX */
1719 /* XXX no locale support yet */
1720 STRLEN len;
89ebb4a3 1721 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
ee099d14 1722 return (U32)to_uni_upper(c, tmpbuf, &len);
b7ac61fa
JH
1723}
1724
1725U32
1726Perl_to_uni_title_lc(pTHX_ U32 c)
1727{
ee099d14
JH
1728 /* XXX returns only the first character XXX -- do not use XXX */
1729 /* XXX no locale support yet */
1730 STRLEN len;
89ebb4a3 1731 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
ee099d14 1732 return (U32)to_uni_title(c, tmpbuf, &len);
b7ac61fa
JH
1733}
1734
1735U32
1736Perl_to_uni_lower_lc(pTHX_ U32 c)
1737{
ee099d14
JH
1738 /* XXX returns only the first character -- do not use XXX */
1739 /* XXX no locale support yet */
1740 STRLEN len;
89ebb4a3 1741 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
ee099d14 1742 return (U32)to_uni_lower(c, tmpbuf, &len);
b7ac61fa
JH
1743}
1744
7452cf6a 1745static bool
5141f98e 1746S_is_utf8_common(pTHX_ const U8 *const p, SV **swash,
bde6a22d
NC
1747 const char *const swashname)
1748{
ea317ccb
KW
1749 /* returns a boolean giving whether or not the UTF8-encoded character that
1750 * starts at <p> is in the swash indicated by <swashname>. <swash>
1751 * contains a pointer to where the swash indicated by <swashname>
1752 * is to be stored; which this routine will do, so that future calls will
1753 * look at <*swash> and only generate a swash if it is not null
1754 *
1755 * Note that it is assumed that the buffer length of <p> is enough to
1756 * contain all the bytes that comprise the character. Thus, <*p> should
1757 * have been checked before this call for mal-formedness enough to assure
1758 * that. */
1759
97aff369 1760 dVAR;
7918f24d
NC
1761
1762 PERL_ARGS_ASSERT_IS_UTF8_COMMON;
1763
492a624f
KW
1764 /* The API should have included a length for the UTF-8 character in <p>,
1765 * but it doesn't. We therefor assume that p has been validated at least
1766 * as far as there being enough bytes available in it to accommodate the
1767 * character without reading beyond the end, and pass that number on to the
1768 * validating routine */
1769 if (!is_utf8_char_buf(p, p + UTF8SKIP(p)))
bde6a22d
NC
1770 return FALSE;
1771 if (!*swash)
711a919c 1772 *swash = swash_init("utf8", swashname, &PL_sv_undef, 1, 0);
bde6a22d
NC
1773 return swash_fetch(*swash, p, TRUE) != 0;
1774}
1775
1776bool
7fc63493 1777Perl_is_utf8_alnum(pTHX_ const U8 *p)
a0ed51b3 1778{
97aff369 1779 dVAR;
7918f24d
NC
1780
1781 PERL_ARGS_ASSERT_IS_UTF8_ALNUM;
1782
671c33bf
NC
1783 /* NOTE: "IsWord", not "IsAlnum", since Alnum is a true
1784 * descendant of isalnum(3), in other words, it doesn't
1785 * contain the '_'. --jhi */
d4c19fe8 1786 return is_utf8_common(p, &PL_utf8_alnum, "IsWord");
a0ed51b3
LW
1787}
1788
1789bool
7fc63493 1790Perl_is_utf8_idfirst(pTHX_ const U8 *p) /* The naming is historical. */
a0ed51b3 1791{
97aff369 1792 dVAR;
7918f24d
NC
1793
1794 PERL_ARGS_ASSERT_IS_UTF8_IDFIRST;
1795
82686b01
JH
1796 if (*p == '_')
1797 return TRUE;
bde6a22d 1798 /* is_utf8_idstart would be more logical. */
d4c19fe8 1799 return is_utf8_common(p, &PL_utf8_idstart, "IdStart");
82686b01
JH
1800}
1801
1802bool
c11ff943
KW
1803Perl_is_utf8_xidfirst(pTHX_ const U8 *p) /* The naming is historical. */
1804{
1805 dVAR;
1806
1807 PERL_ARGS_ASSERT_IS_UTF8_XIDFIRST;
1808
1809 if (*p == '_')
1810 return TRUE;
1811 /* is_utf8_idstart would be more logical. */
1812 return is_utf8_common(p, &PL_utf8_xidstart, "XIdStart");
1813}
1814
1815bool
b6912c02
KW
1816Perl__is_utf8__perl_idstart(pTHX_ const U8 *p)
1817{
1818 dVAR;
1819
1820 PERL_ARGS_ASSERT__IS_UTF8__PERL_IDSTART;
1821
1822 return is_utf8_common(p, &PL_utf8_perl_idstart, "_Perl_IDStart");
1823}
1824
1825bool
7fc63493 1826Perl_is_utf8_idcont(pTHX_ const U8 *p)
82686b01 1827{
97aff369 1828 dVAR;
7918f24d
NC
1829
1830 PERL_ARGS_ASSERT_IS_UTF8_IDCONT;
1831
d4c19fe8 1832 return is_utf8_common(p, &PL_utf8_idcont, "IdContinue");
a0ed51b3
LW
1833}
1834
1835bool
c11ff943
KW
1836Perl_is_utf8_xidcont(pTHX_ const U8 *p)
1837{
1838 dVAR;
1839
1840 PERL_ARGS_ASSERT_IS_UTF8_XIDCONT;
1841
c11ff943
KW
1842 return is_utf8_common(p, &PL_utf8_idcont, "XIdContinue");
1843}
1844
1845bool
7fc63493 1846Perl_is_utf8_alpha(pTHX_ const U8 *p)
a0ed51b3 1847{
97aff369 1848 dVAR;
7918f24d
NC
1849
1850 PERL_ARGS_ASSERT_IS_UTF8_ALPHA;
1851
d4c19fe8 1852 return is_utf8_common(p, &PL_utf8_alpha, "IsAlpha");
a0ed51b3
LW
1853}
1854
1855bool
7fc63493 1856Perl_is_utf8_ascii(pTHX_ const U8 *p)
b8c5462f 1857{
97aff369 1858 dVAR;
7918f24d
NC
1859
1860 PERL_ARGS_ASSERT_IS_UTF8_ASCII;
1861
bc39fe24
KW
1862 /* ASCII characters are the same whether in utf8 or not. So the macro
1863 * works on both utf8 and non-utf8 representations. */
1864 return isASCII(*p);
b8c5462f
JH
1865}
1866
1867bool
7fc63493 1868Perl_is_utf8_space(pTHX_ const U8 *p)
a0ed51b3 1869{
97aff369 1870 dVAR;
7918f24d
NC
1871
1872 PERL_ARGS_ASSERT_IS_UTF8_SPACE;
1873
a34094a9 1874 return is_utf8_common(p, &PL_utf8_space, "IsXPerlSpace");
a0ed51b3
LW
1875}
1876
1877bool
d1eb3177
YO
1878Perl_is_utf8_perl_space(pTHX_ const U8 *p)
1879{
1880 dVAR;
1881
1882 PERL_ARGS_ASSERT_IS_UTF8_PERL_SPACE;
1883
c4428693
KW
1884 /* Only true if is an ASCII space-like character, and ASCII is invariant
1885 * under utf8, so can just use the macro */
1886 return isSPACE_A(*p);
d1eb3177
YO
1887}
1888
1889bool
1890Perl_is_utf8_perl_word(pTHX_ const U8 *p)
1891{
1892 dVAR;
1893
1894 PERL_ARGS_ASSERT_IS_UTF8_PERL_WORD;
1895
c4428693
KW
1896 /* Only true if is an ASCII word character, and ASCII is invariant
1897 * under utf8, so can just use the macro */
1898 return isWORDCHAR_A(*p);
d1eb3177
YO
1899}
1900
1901bool
7fc63493 1902Perl_is_utf8_digit(pTHX_ const U8 *p)
a0ed51b3 1903{
97aff369 1904 dVAR;
7918f24d
NC
1905
1906 PERL_ARGS_ASSERT_IS_UTF8_DIGIT;
1907
d4c19fe8 1908 return is_utf8_common(p, &PL_utf8_digit, "IsDigit");
a0ed51b3
LW
1909}
1910
1911bool
d1eb3177
YO
1912Perl_is_utf8_posix_digit(pTHX_ const U8 *p)
1913{
1914 dVAR;
1915
1916 PERL_ARGS_ASSERT_IS_UTF8_POSIX_DIGIT;
1917
c4428693
KW
1918 /* Only true if is an ASCII digit character, and ASCII is invariant
1919 * under utf8, so can just use the macro */
1920 return isDIGIT_A(*p);
d1eb3177
YO
1921}
1922
1923bool
7fc63493 1924Perl_is_utf8_upper(pTHX_ const U8 *p)
a0ed51b3 1925{
97aff369 1926 dVAR;
7918f24d
NC
1927
1928 PERL_ARGS_ASSERT_IS_UTF8_UPPER;
1929
d4c19fe8 1930 return is_utf8_common(p, &PL_utf8_upper, "IsUppercase");
a0ed51b3
LW
1931}
1932
1933bool
7fc63493 1934Perl_is_utf8_lower(pTHX_ const U8 *p)
a0ed51b3 1935{
97aff369 1936 dVAR;
7918f24d
NC
1937
1938 PERL_ARGS_ASSERT_IS_UTF8_LOWER;
1939
d4c19fe8 1940 return is_utf8_common(p, &PL_utf8_lower, "IsLowercase");
a0ed51b3
LW
1941}
1942
1943bool
7fc63493 1944Perl_is_utf8_cntrl(pTHX_ const U8 *p)
b8c5462f 1945{
97aff369 1946 dVAR;
7918f24d
NC
1947
1948 PERL_ARGS_ASSERT_IS_UTF8_CNTRL;
1949
7b952154
KW
1950 if (isASCII(*p)) {
1951 return isCNTRL_A(*p);
1952 }
1953
1954 /* All controls are in Latin1 */
1955 if (! UTF8_IS_DOWNGRADEABLE_START(*p)) {
1956 return 0;
1957 }
1958 return isCNTRL_L1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)));
b8c5462f
JH
1959}
1960
1961bool
7fc63493 1962Perl_is_utf8_graph(pTHX_ const U8 *p)
b8c5462f 1963{
97aff369 1964 dVAR;
7918f24d
NC
1965
1966 PERL_ARGS_ASSERT_IS_UTF8_GRAPH;
1967
d4c19fe8 1968 return is_utf8_common(p, &PL_utf8_graph, "IsGraph");
b8c5462f
JH
1969}
1970
1971bool
7fc63493 1972Perl_is_utf8_print(pTHX_ const U8 *p)
a0ed51b3 1973{
97aff369 1974 dVAR;
7918f24d
NC
1975
1976 PERL_ARGS_ASSERT_IS_UTF8_PRINT;
1977
d4c19fe8 1978 return is_utf8_common(p, &PL_utf8_print, "IsPrint");
a0ed51b3
LW
1979}
1980
1981bool
7fc63493 1982Perl_is_utf8_punct(pTHX_ const U8 *p)
b8c5462f 1983{
97aff369 1984 dVAR;
7918f24d
NC
1985
1986 PERL_ARGS_ASSERT_IS_UTF8_PUNCT;
1987
d4c19fe8 1988 return is_utf8_common(p, &PL_utf8_punct, "IsPunct");
b8c5462f
JH
1989}
1990
1991bool
7fc63493 1992Perl_is_utf8_xdigit(pTHX_ const U8 *p)
b8c5462f 1993{
97aff369 1994 dVAR;
7918f24d
NC
1995
1996 PERL_ARGS_ASSERT_IS_UTF8_XDIGIT;
1997
d1eb3177 1998 return is_utf8_common(p, &PL_utf8_xdigit, "IsXDigit");
b8c5462f
JH
1999}
2000
2001bool
7fc63493 2002Perl_is_utf8_mark(pTHX_ const U8 *p)
a0ed51b3 2003{
97aff369 2004 dVAR;
7918f24d
NC
2005
2006 PERL_ARGS_ASSERT_IS_UTF8_MARK;
2007
d4c19fe8 2008 return is_utf8_common(p, &PL_utf8_mark, "IsM");
a0ed51b3
LW
2009}
2010
37e2e78e
KW
2011bool
2012Perl_is_utf8_X_begin(pTHX_ const U8 *p)
2013{
2014 dVAR;
2015
2016 PERL_ARGS_ASSERT_IS_UTF8_X_BEGIN;
2017
2018 return is_utf8_common(p, &PL_utf8_X_begin, "_X_Begin");
2019}
2020
2021bool
2022Perl_is_utf8_X_extend(pTHX_ const U8 *p)
2023{
2024 dVAR;
2025
2026 PERL_ARGS_ASSERT_IS_UTF8_X_EXTEND;
2027
2028 return is_utf8_common(p, &PL_utf8_X_extend, "_X_Extend");
2029}
2030
2031bool
2032Perl_is_utf8_X_prepend(pTHX_ const U8 *p)
2033{
2034 dVAR;
2035
2036 PERL_ARGS_ASSERT_IS_UTF8_X_PREPEND;
2037
2038 return is_utf8_common(p, &PL_utf8_X_prepend, "GCB=Prepend");
2039}
2040
2041bool
2042Perl_is_utf8_X_non_hangul(pTHX_ const U8 *p)
2043{
2044 dVAR;
2045
2046 PERL_ARGS_ASSERT_IS_UTF8_X_NON_HANGUL;
2047
2048 return is_utf8_common(p, &PL_utf8_X_non_hangul, "HST=Not_Applicable");
2049}
2050
2051bool
2052Perl_is_utf8_X_L(pTHX_ const U8 *p)
2053{
2054 dVAR;
2055
2056 PERL_ARGS_ASSERT_IS_UTF8_X_L;
2057
2058 return is_utf8_common(p, &PL_utf8_X_L, "GCB=L");
2059}
2060
2061bool
2062Perl_is_utf8_X_LV(pTHX_ const U8 *p)
2063{
2064 dVAR;
2065
2066 PERL_ARGS_ASSERT_IS_UTF8_X_LV;
2067
2068 return is_utf8_common(p, &PL_utf8_X_LV, "GCB=LV");
2069}
2070
2071bool
2072Perl_is_utf8_X_LVT(pTHX_ const U8 *p)
2073{
2074 dVAR;
2075
2076 PERL_ARGS_ASSERT_IS_UTF8_X_LVT;
2077
2078 return is_utf8_common(p, &PL_utf8_X_LVT, "GCB=LVT");
2079}
2080
2081bool
2082Perl_is_utf8_X_T(pTHX_ const U8 *p)
2083{
2084 dVAR;
2085
2086 PERL_ARGS_ASSERT_IS_UTF8_X_T;
2087
2088 return is_utf8_common(p, &PL_utf8_X_T, "GCB=T");
2089}
2090
2091bool
2092Perl_is_utf8_X_V(pTHX_ const U8 *p)
2093{
2094 dVAR;
2095
2096 PERL_ARGS_ASSERT_IS_UTF8_X_V;
2097
2098 return is_utf8_common(p, &PL_utf8_X_V, "GCB=V");
2099}
2100
2101bool
2102Perl_is_utf8_X_LV_LVT_V(pTHX_ const U8 *p)
2103{
2104 dVAR;
2105
2106 PERL_ARGS_ASSERT_IS_UTF8_X_LV_LVT_V;
2107
2108 return is_utf8_common(p, &PL_utf8_X_LV_LVT_V, "_X_LV_LVT_V");
2109}
2110
2e2b2571
KW
2111bool
2112Perl__is_utf8_quotemeta(pTHX_ const U8 *p)
2113{
2114 /* For exclusive use of pp_quotemeta() */
2115
2116 dVAR;
2117
2118 PERL_ARGS_ASSERT__IS_UTF8_QUOTEMETA;
2119
2120 return is_utf8_common(p, &PL_utf8_quotemeta, "_Perl_Quotemeta");
2121}
2122
6b5c0936 2123/*
87cea99e 2124=for apidoc to_utf8_case
6b5c0936 2125
a1433954
KW
2126The C<p> contains the pointer to the UTF-8 string encoding
2127the character that is being converted. This routine assumes that the character
2128at C<p> is well-formed.
6b5c0936 2129
a1433954
KW
2130The C<ustrp> is a pointer to the character buffer to put the
2131conversion result to. The C<lenp> is a pointer to the length
6b5c0936
JH
2132of the result.
2133
a1433954 2134The C<swashp> is a pointer to the swash to use.
6b5c0936 2135
a1433954
KW
2136Both the special and normal mappings are stored in F<lib/unicore/To/Foo.pl>,
2137and loaded by SWASHNEW, using F<lib/utf8_heavy.pl>. The C<special> (usually,
0134edef 2138but not always, a multicharacter mapping), is tried first.
6b5c0936 2139
a1433954 2140The C<special> is a string like "utf8::ToSpecLower", which means the
0134edef
JH
2141hash %utf8::ToSpecLower. The access to the hash is through
2142Perl_to_utf8_case().
6b5c0936 2143
a1433954 2144The C<normal> is a string like "ToLower" which means the swash
0134edef
JH
2145%utf8::ToLower.
2146
2147=cut */
6b5c0936 2148
2104c8d9 2149UV
9a957fbc
AL
2150Perl_to_utf8_case(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp,
2151 SV **swashp, const char *normal, const char *special)
a0ed51b3 2152{
97aff369 2153 dVAR;
89ebb4a3 2154 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
0134edef 2155 STRLEN len = 0;
4b88fb76 2156 const UV uv0 = valid_utf8_to_uvchr(p, NULL);
1feea2c7
JH
2157 /* The NATIVE_TO_UNI() and UNI_TO_NATIVE() mappings
2158 * are necessary in EBCDIC, they are redundant no-ops
2159 * in ASCII-ish platforms, and hopefully optimized away. */
f54cb97a 2160 const UV uv1 = NATIVE_TO_UNI(uv0);
7918f24d
NC
2161
2162 PERL_ARGS_ASSERT_TO_UTF8_CASE;
2163
9ae3ac1a
KW
2164 /* Note that swash_fetch() doesn't output warnings for these because it
2165 * assumes we will */
8457b38f 2166 if (uv1 >= UNICODE_SURROGATE_FIRST) {
9ae3ac1a 2167 if (uv1 <= UNICODE_SURROGATE_LAST) {
8457b38f
KW
2168 if (ckWARN_d(WARN_SURROGATE)) {
2169 const char* desc = (PL_op) ? OP_DESC(PL_op) : normal;
2170 Perl_warner(aTHX_ packWARN(WARN_SURROGATE),
2171 "Operation \"%s\" returns its argument for UTF-16 surrogate U+%04"UVXf"", desc, uv1);
2172 }
9ae3ac1a
KW
2173 }
2174 else if (UNICODE_IS_SUPER(uv1)) {
8457b38f
KW
2175 if (ckWARN_d(WARN_NON_UNICODE)) {
2176 const char* desc = (PL_op) ? OP_DESC(PL_op) : normal;
2177 Perl_warner(aTHX_ packWARN(WARN_NON_UNICODE),
2178 "Operation \"%s\" returns its argument for non-Unicode code point 0x%04"UVXf"", desc, uv1);
2179 }
9ae3ac1a
KW
2180 }
2181
2182 /* Note that non-characters are perfectly legal, so no warning should
2183 * be given */
2184 }
2185
1feea2c7 2186 uvuni_to_utf8(tmpbuf, uv1);
0134edef
JH
2187
2188 if (!*swashp) /* load on-demand */
2189 *swashp = swash_init("utf8", normal, &PL_sv_undef, 4, 0);
2190
a6f87d8c 2191 if (special) {
0134edef 2192 /* It might be "special" (sometimes, but not always,
2a37f04d 2193 * a multicharacter mapping) */
6673a63c 2194 HV * const hv = get_hv(special, 0);
b08cf34e
JH
2195 SV **svp;
2196
35da51f7 2197 if (hv &&
b08cf34e
JH
2198 (svp = hv_fetch(hv, (const char*)tmpbuf, UNISKIP(uv1), FALSE)) &&
2199 (*svp)) {
cfd0369c 2200 const char *s;
47654450 2201
cfd0369c 2202 s = SvPV_const(*svp, len);
47654450
JH
2203 if (len == 1)
2204 len = uvuni_to_utf8(ustrp, NATIVE_TO_UNI(*(U8*)s)) - ustrp;
2a37f04d 2205 else {
2f9475ad
JH
2206#ifdef EBCDIC
2207 /* If we have EBCDIC we need to remap the characters
2208 * since any characters in the low 256 are Unicode
2209 * code points, not EBCDIC. */
7cda7a3d 2210 U8 *t = (U8*)s, *tend = t + len, *d;
2f9475ad
JH
2211
2212 d = tmpbuf;
b08cf34e 2213 if (SvUTF8(*svp)) {
2f9475ad
JH
2214 STRLEN tlen = 0;
2215
2216 while (t < tend) {
6bd1c396 2217 const UV c = utf8_to_uvchr_buf(t, tend, &tlen);
2f9475ad
JH
2218 if (tlen > 0) {
2219 d = uvchr_to_utf8(d, UNI_TO_NATIVE(c));
2220 t += tlen;
2221 }
2222 else
2223 break;
2224 }
2225 }
2226 else {
36fec512
JH
2227 while (t < tend) {
2228 d = uvchr_to_utf8(d, UNI_TO_NATIVE(*t));
2229 t++;
2230 }
2f9475ad
JH
2231 }
2232 len = d - tmpbuf;
2233 Copy(tmpbuf, ustrp, len, U8);
2234#else
d2dcd0fb 2235 Copy(s, ustrp, len, U8);
2f9475ad 2236#endif
29e98929 2237 }
983ffd37 2238 }
0134edef
JH
2239 }
2240
2241 if (!len && *swashp) {
d4c19fe8
AL
2242 const UV uv2 = swash_fetch(*swashp, tmpbuf, TRUE);
2243
0134edef
JH
2244 if (uv2) {
2245 /* It was "normal" (a single character mapping). */
d4c19fe8 2246 const UV uv3 = UNI_TO_NATIVE(uv2);
e9101d72 2247 len = uvchr_to_utf8(ustrp, uv3) - ustrp;
2a37f04d
JH
2248 }
2249 }
1feea2c7 2250
37e2e78e
KW
2251 if (!len) /* Neither: just copy. In other words, there was no mapping
2252 defined, which means that the code point maps to itself */
0134edef
JH
2253 len = uvchr_to_utf8(ustrp, uv0) - ustrp;
2254
2a37f04d
JH
2255 if (lenp)
2256 *lenp = len;
2257
6bd1c396 2258 return len ? valid_utf8_to_uvchr(ustrp, 0) : 0;
a0ed51b3
LW
2259}
2260
051a06d4
KW
2261STATIC UV
2262S_check_locale_boundary_crossing(pTHX_ const U8* const p, const UV result, U8* const ustrp, STRLEN *lenp)
2263{
2264 /* This is called when changing the case of a utf8-encoded character above
2265 * the Latin1 range, and the operation is in locale. If the result
2266 * contains a character that crosses the 255/256 boundary, disallow the
2267 * change, and return the original code point. See L<perlfunc/lc> for why;
2268 *
a1433954
KW
2269 * p points to the original string whose case was changed; assumed
2270 * by this routine to be well-formed
051a06d4
KW
2271 * result the code point of the first character in the changed-case string
2272 * ustrp points to the changed-case string (<result> represents its first char)
2273 * lenp points to the length of <ustrp> */
2274
2275 UV original; /* To store the first code point of <p> */
2276
2277 PERL_ARGS_ASSERT_CHECK_LOCALE_BOUNDARY_CROSSING;
2278
2279 assert(! UTF8_IS_INVARIANT(*p) && ! UTF8_IS_DOWNGRADEABLE_START(*p));
2280
2281 /* We know immediately if the first character in the string crosses the
2282 * boundary, so can skip */
2283 if (result > 255) {
2284
2285 /* Look at every character in the result; if any cross the
2286 * boundary, the whole thing is disallowed */
2287 U8* s = ustrp + UTF8SKIP(ustrp);
2288 U8* e = ustrp + *lenp;
2289 while (s < e) {
2290 if (UTF8_IS_INVARIANT(*s) || UTF8_IS_DOWNGRADEABLE_START(*s))
2291 {
2292 goto bad_crossing;
2293 }
2294 s += UTF8SKIP(s);
2295 }
2296
2297 /* Here, no characters crossed, result is ok as-is */
2298 return result;
2299 }
2300
2301bad_crossing:
2302
2303 /* Failed, have to return the original */
4b88fb76 2304 original = valid_utf8_to_uvchr(p, lenp);
051a06d4
KW
2305 Copy(p, ustrp, *lenp, char);
2306 return original;
2307}
2308
d3e79532 2309/*
87cea99e 2310=for apidoc to_utf8_upper
d3e79532 2311
a1433954
KW
2312Convert the UTF-8 encoded character at C<p> to its uppercase version and
2313store that in UTF-8 in C<ustrp> and its length in bytes in C<lenp>. Note
89ebb4a3
JH
2314that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since
2315the uppercase version may be longer than the original character.
d3e79532
JH
2316
2317The first character of the uppercased version is returned
2318(but note, as explained above, that there may be more.)
2319
a1433954
KW
2320The character at C<p> is assumed by this routine to be well-formed.
2321
d3e79532
JH
2322=cut */
2323
051a06d4
KW
2324/* Not currently externally documented, and subject to change:
2325 * <flags> is set iff locale semantics are to be used for code points < 256
2326 * <tainted_ptr> if non-null, *tainted_ptr will be set TRUE iff locale rules
2327 * were used in the calculation; otherwise unchanged. */
2328
2104c8d9 2329UV
051a06d4 2330Perl__to_utf8_upper_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, const bool flags, bool* tainted_ptr)
a0ed51b3 2331{
97aff369 2332 dVAR;
7918f24d 2333
051a06d4
KW
2334 UV result;
2335
2336 PERL_ARGS_ASSERT__TO_UTF8_UPPER_FLAGS;
7918f24d 2337
3a4c58c9 2338 if (UTF8_IS_INVARIANT(*p)) {
051a06d4
KW
2339 if (flags) {
2340 result = toUPPER_LC(*p);
2341 }
2342 else {
81c6c7ce 2343 return _to_upper_title_latin1(*p, ustrp, lenp, 'S');
051a06d4 2344 }
3a4c58c9
KW
2345 }
2346 else if UTF8_IS_DOWNGRADEABLE_START(*p) {
051a06d4
KW
2347 if (flags) {
2348 result = toUPPER_LC(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)));
2349 }
2350 else {
81c6c7ce
KW
2351 return _to_upper_title_latin1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)),
2352 ustrp, lenp, 'S');
051a06d4
KW
2353 }
2354 }
2355 else { /* utf8, ord above 255 */
2356 result = CALL_UPPER_CASE(p, ustrp, lenp);
2357
2358 if (flags) {
2359 result = check_locale_boundary_crossing(p, result, ustrp, lenp);
2360 }
2361 return result;
2362 }
2363
2364 /* Here, used locale rules. Convert back to utf8 */
2365 if (UTF8_IS_INVARIANT(result)) {
2366 *ustrp = (U8) result;
2367 *lenp = 1;
2368 }
2369 else {
2370 *ustrp = UTF8_EIGHT_BIT_HI(result);
2371 *(ustrp + 1) = UTF8_EIGHT_BIT_LO(result);
2372 *lenp = 2;
3a4c58c9
KW
2373 }
2374
051a06d4
KW
2375 if (tainted_ptr) {
2376 *tainted_ptr = TRUE;
2377 }
2378 return result;
983ffd37 2379}
a0ed51b3 2380
d3e79532 2381/*
87cea99e 2382=for apidoc to_utf8_title
d3e79532 2383
a1433954
KW
2384Convert the UTF-8 encoded character at C<p> to its titlecase version and
2385store that in UTF-8 in C<ustrp> and its length in bytes in C<lenp>. Note
2386that the C<ustrp> needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
89ebb4a3 2387titlecase version may be longer than the original character.
d3e79532
JH
2388
2389The first character of the titlecased version is returned
2390(but note, as explained above, that there may be more.)
2391
a1433954
KW
2392The character at C<p> is assumed by this routine to be well-formed.
2393
d3e79532
JH
2394=cut */
2395
051a06d4
KW
2396/* Not currently externally documented, and subject to change:
2397 * <flags> is set iff locale semantics are to be used for code points < 256
2398 * Since titlecase is not defined in POSIX, uppercase is used instead
2399 * for these/
2400 * <tainted_ptr> if non-null, *tainted_ptr will be set TRUE iff locale rules
2401 * were used in the calculation; otherwise unchanged. */
2402
983ffd37 2403UV
051a06d4 2404Perl__to_utf8_title_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, const bool flags, bool* tainted_ptr)
983ffd37 2405{
97aff369 2406 dVAR;
7918f24d 2407
051a06d4
KW
2408 UV result;
2409
2410 PERL_ARGS_ASSERT__TO_UTF8_TITLE_FLAGS;
7918f24d 2411
3a4c58c9 2412 if (UTF8_IS_INVARIANT(*p)) {
051a06d4
KW
2413 if (flags) {
2414 result = toUPPER_LC(*p);
2415 }
2416 else {
81c6c7ce 2417 return _to_upper_title_latin1(*p, ustrp, lenp, 's');
051a06d4 2418 }
3a4c58c9
KW
2419 }
2420 else if UTF8_IS_DOWNGRADEABLE_START(*p) {
051a06d4
KW
2421 if (flags) {
2422 result = toUPPER_LC(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)));
2423 }
2424 else {
81c6c7ce
KW
2425 return _to_upper_title_latin1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)),
2426 ustrp, lenp, 's');
051a06d4
KW
2427 }
2428 }
2429 else { /* utf8, ord above 255 */
2430 result = CALL_TITLE_CASE(p, ustrp, lenp);
2431
2432 if (flags) {
2433 result = check_locale_boundary_crossing(p, result, ustrp, lenp);
2434 }
2435 return result;
2436 }
2437
2438 /* Here, used locale rules. Convert back to utf8 */
2439 if (UTF8_IS_INVARIANT(result)) {
2440 *ustrp = (U8) result;
2441 *lenp = 1;
2442 }
2443 else {
2444 *ustrp = UTF8_EIGHT_BIT_HI(result);
2445 *(ustrp + 1) = UTF8_EIGHT_BIT_LO(result);
2446 *lenp = 2;
3a4c58c9
KW
2447 }
2448
051a06d4
KW
2449 if (tainted_ptr) {
2450 *tainted_ptr = TRUE;
2451 }
2452 return result;
a0ed51b3
LW
2453}
2454
d3e79532 2455/*
87cea99e 2456=for apidoc to_utf8_lower
d3e79532 2457
a1433954
KW
2458Convert the UTF-8 encoded character at C<p> to its lowercase version and
2459store that in UTF-8 in ustrp and its length in bytes in C<lenp>. Note
2460that the C<ustrp> needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
89ebb4a3 2461lowercase version may be longer than the original character.
d3e79532
JH
2462
2463The first character of the lowercased version is returned
2464(but note, as explained above, that there may be more.)
2465
a1433954
KW
2466The character at C<p> is assumed by this routine to be well-formed.
2467
d3e79532
JH
2468=cut */
2469
051a06d4
KW
2470/* Not currently externally documented, and subject to change:
2471 * <flags> is set iff locale semantics are to be used for code points < 256
2472 * <tainted_ptr> if non-null, *tainted_ptr will be set TRUE iff locale rules
2473 * were used in the calculation; otherwise unchanged. */
2474
2104c8d9 2475UV
051a06d4 2476Perl__to_utf8_lower_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, const bool flags, bool* tainted_ptr)
a0ed51b3 2477{
051a06d4
KW
2478 UV result;
2479
97aff369 2480 dVAR;
7918f24d 2481
051a06d4 2482 PERL_ARGS_ASSERT__TO_UTF8_LOWER_FLAGS;
7918f24d 2483
968c5e6a 2484 if (UTF8_IS_INVARIANT(*p)) {
051a06d4
KW
2485 if (flags) {
2486 result = toLOWER_LC(*p);
2487 }
2488 else {
81c6c7ce 2489 return to_lower_latin1(*p, ustrp, lenp);
051a06d4 2490 }
968c5e6a
KW
2491 }
2492 else if UTF8_IS_DOWNGRADEABLE_START(*p) {
051a06d4
KW
2493 if (flags) {
2494 result = toLOWER_LC(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)));
2495 }
2496 else {
81c6c7ce
KW
2497 return to_lower_latin1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)),
2498 ustrp, lenp);
051a06d4 2499 }
968c5e6a 2500 }
051a06d4
KW
2501 else { /* utf8, ord above 255 */
2502 result = CALL_LOWER_CASE(p, ustrp, lenp);
2503
2504 if (flags) {
2505 result = check_locale_boundary_crossing(p, result, ustrp, lenp);
2506 }
968c5e6a 2507
051a06d4
KW
2508 return result;
2509 }
2510
2511 /* Here, used locale rules. Convert back to utf8 */
2512 if (UTF8_IS_INVARIANT(result)) {
2513 *ustrp = (U8) result;
2514 *lenp = 1;
2515 }
2516 else {
2517 *ustrp = UTF8_EIGHT_BIT_HI(result);
2518 *(ustrp + 1) = UTF8_EIGHT_BIT_LO(result);
2519 *lenp = 2;
2520 }
2521
2522 if (tainted_ptr) {
2523 *tainted_ptr = TRUE;
2524 }
2525 return result;
b4e400f9
JH
2526}
2527
d3e79532 2528/*
87cea99e 2529=for apidoc to_utf8_fold
d3e79532 2530
a1433954
KW
2531Convert the UTF-8 encoded character at C<p> to its foldcase version and
2532store that in UTF-8 in C<ustrp> and its length in bytes in C<lenp>. Note
2533that the C<ustrp> needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
d3e79532
JH
2534foldcase version may be longer than the original character (up to
2535three characters).
2536
2537The first character of the foldcased version is returned
2538(but note, as explained above, that there may be more.)
2539
a1433954
KW
2540The character at C<p> is assumed by this routine to be well-formed.
2541
d3e79532
JH
2542=cut */
2543
051a06d4
KW
2544/* Not currently externally documented, and subject to change,
2545 * in <flags>
2546 * bit FOLD_FLAGS_LOCALE is set iff locale semantics are to be used for code
2547 * points < 256. Since foldcase is not defined in
2548 * POSIX, lowercase is used instead
2549 * bit FOLD_FLAGS_FULL is set iff full case folds are to be used;
2550 * otherwise simple folds
2551 * <tainted_ptr> if non-null, *tainted_ptr will be set TRUE iff locale rules
2552 * were used in the calculation; otherwise unchanged. */
36bb2ab6 2553
b4e400f9 2554UV
051a06d4 2555Perl__to_utf8_fold_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, U8 flags, bool* tainted_ptr)
b4e400f9 2556{
97aff369 2557 dVAR;
7918f24d 2558
051a06d4
KW
2559 UV result;
2560
36bb2ab6 2561 PERL_ARGS_ASSERT__TO_UTF8_FOLD_FLAGS;
7918f24d 2562
a1dde8de 2563 if (UTF8_IS_INVARIANT(*p)) {
051a06d4
KW
2564 if (flags & FOLD_FLAGS_LOCALE) {
2565 result = toLOWER_LC(*p);
2566 }
2567 else {
81c6c7ce
KW
2568 return _to_fold_latin1(*p, ustrp, lenp,
2569 cBOOL(flags & FOLD_FLAGS_FULL));
051a06d4 2570 }
a1dde8de
KW
2571 }
2572 else if UTF8_IS_DOWNGRADEABLE_START(*p) {
051a06d4
KW
2573 if (flags & FOLD_FLAGS_LOCALE) {
2574 result = toLOWER_LC(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)));
2575 }
2576 else {
81c6c7ce 2577 return _to_fold_latin1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)),
051a06d4
KW
2578 ustrp, lenp, cBOOL(flags & FOLD_FLAGS_FULL));
2579 }
a1dde8de 2580 }
051a06d4
KW
2581 else { /* utf8, ord above 255 */
2582 result = CALL_FOLD_CASE(p, ustrp, lenp, flags);
a1dde8de 2583
051a06d4
KW
2584 if ((flags & FOLD_FLAGS_LOCALE)) {
2585 result = check_locale_boundary_crossing(p, result, ustrp, lenp);
2586 }
2587
2588 return result;
2589 }
2590
2591 /* Here, used locale rules. Convert back to utf8 */
2592 if (UTF8_IS_INVARIANT(result)) {
2593 *ustrp = (U8) result;
2594 *lenp = 1;
2595 }
2596 else {
2597 *ustrp = UTF8_EIGHT_BIT_HI(result);
2598 *(ustrp + 1) = UTF8_EIGHT_BIT_LO(result);
2599 *lenp = 2;
2600 }
2601
2602 if (tainted_ptr) {
2603 *tainted_ptr = TRUE;
2604 }
2605 return result;
a0ed51b3
LW
2606}
2607
711a919c 2608/* Note:
f90a9a02 2609 * Returns a "swash" which is a hash described in utf8.c:Perl_swash_fetch().
711a919c
ST
2610 * C<pkg> is a pointer to a package name for SWASHNEW, should be "utf8".
2611 * For other parameters, see utf8::SWASHNEW in lib/utf8_heavy.pl.
2612 */
c4a5db0c 2613
a0ed51b3 2614SV*
7fc63493 2615Perl_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv, I32 minbits, I32 none)
a0ed51b3 2616{
c4a5db0c
KW
2617 PERL_ARGS_ASSERT_SWASH_INIT;
2618
2619 /* Returns a copy of a swash initiated by the called function. This is the
2620 * public interface, and returning a copy prevents others from doing
2621 * mischief on the original */
2622
9a53f6cf 2623 return newSVsv(_core_swash_init(pkg, name, listsv, minbits, none, FALSE, NULL, FALSE));
c4a5db0c
KW
2624}
2625
2626SV*
9a53f6cf 2627Perl__core_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv, I32 minbits, I32 none, bool return_if_undef, SV* invlist, bool passed_in_invlist_has_user_defined_property)
c4a5db0c
KW
2628{
2629 /* Initialize and return a swash, creating it if necessary. It does this
9a53f6cf 2630 * by calling utf8_heavy.pl in the general case.
c4a5db0c
KW
2631 *
2632 * This interface should only be used by functions that won't destroy or
2633 * adversely change the swash, as doing so affects all other uses of the
2634 * swash in the program; the general public should use 'Perl_swash_init'
2635 * instead.
2636 *
2637 * pkg is the name of the package that <name> should be in.
2638 * name is the name of the swash to find. Typically it is a Unicode
2639 * property name, including user-defined ones
2640 * listsv is a string to initialize the swash with. It must be of the form
2641 * documented as the subroutine return value in
2642 * L<perlunicode/User-Defined Character Properties>
2643 * minbits is the number of bits required to represent each data element.
2644 * It is '1' for binary properties.
2645 * none I (khw) do not understand this one, but it is used only in tr///.
9a53f6cf
KW
2646 * return_if_undef is TRUE if the routine shouldn't croak if it can't find
2647 * the requested property
2648 * invlist is an inversion list to initialize the swash with (or NULL)
2649 * has_user_defined_property is TRUE if <invlist> has some component that
2650 * came from a user-defined property
2651 *
2652 * Thus there are three possible inputs to find the swash: <name>,
2653 * <listsv>, and <invlist>. At least one must be specified. The result
2654 * will be the union of the specified ones, although <listsv>'s various
2655 * actions can intersect, etc. what <name> gives.
2656 *
2657 * <invlist> is only valid for binary properties */
c4a5db0c 2658
27da23d5 2659 dVAR;
c4a5db0c 2660 SV* retval = &PL_sv_undef;
9a53f6cf
KW
2661
2662 assert(listsv != &PL_sv_undef || strNE(name, "") || invlist);
2663 assert(! invlist || minbits == 1);
2664
2665 /* If data was passed in to go out to utf8_heavy to find the swash of, do
2666 * so */
2667 if (listsv != &PL_sv_undef || strNE(name, "")) {
69794297
KW
2668 dSP;
2669 const size_t pkg_len = strlen(pkg);
2670 const size_t name_len = strlen(name);
2671 HV * const stash = gv_stashpvn(pkg, pkg_len, 0);
2672 SV* errsv_save;
2673 GV *method;
2674
2675 PERL_ARGS_ASSERT__CORE_SWASH_INIT;
2676
2677 PUSHSTACKi(PERLSI_MAGIC);
ce3b816e 2678 ENTER;
69794297
KW
2679 SAVEHINTS();
2680 save_re_context();
2681 if (PL_parser && PL_parser->error_count)
2682 SAVEI8(PL_parser->error_count), PL_parser->error_count = 0;
2683 method = gv_fetchmeth(stash, "SWASHNEW", 8, -1);
2684 if (!method) { /* demand load utf8 */
2685 ENTER;
2686 errsv_save = newSVsv(ERRSV);
2687 /* It is assumed that callers of this routine are not passing in
2688 * any user derived data. */
2689 /* Need to do this after save_re_context() as it will set
2690 * PL_tainted to 1 while saving $1 etc (see the code after getrx:
2691 * in Perl_magic_get). Even line to create errsv_save can turn on
2692 * PL_tainted. */
2693 SAVEBOOL(PL_tainted);
2694 PL_tainted = 0;
2695 Perl_load_module(aTHX_ PERL_LOADMOD_NOIMPORT, newSVpvn(pkg,pkg_len),
2696 NULL);
2697 if (!SvTRUE(ERRSV))
2698 sv_setsv(ERRSV, errsv_save);
2699 SvREFCNT_dec(errsv_save);
2700 LEAVE;
2701 }
2702 SPAGAIN;
2703 PUSHMARK(SP);
2704 EXTEND(SP,5);
2705 mPUSHp(pkg, pkg_len);
2706 mPUSHp(name, name_len);
2707 PUSHs(listsv);
2708 mPUSHi(minbits);
2709 mPUSHi(none);
2710 PUTBACK;
f8be5cf0 2711 errsv_save = newSVsv(ERRSV);
69794297
KW
2712 /* If we already have a pointer to the method, no need to use
2713 * call_method() to repeat the lookup. */
2714 if (method ? call_sv(MUTABLE_SV(method), G_SCALAR)
2715 : call_sv(newSVpvs_flags("SWASHNEW", SVs_TEMP), G_SCALAR | G_METHOD))
2716 {
2717 retval = *PL_stack_sp--;
2718 SvREFCNT_inc(retval);
2719 }
f8be5cf0
JH
2720 if (!SvTRUE(ERRSV))
2721 sv_setsv(ERRSV, errsv_save);
2722 SvREFCNT_dec(errsv_save);
ce3b816e 2723 LEAVE;
69794297
KW
2724 POPSTACK;
2725 if (IN_PERL_COMPILETIME) {
2726 CopHINTS_set(PL_curcop, PL_hints);
2727 }
2728 if (!SvROK(retval) || SvTYPE(SvRV(retval)) != SVt_PVHV) {
2729 if (SvPOK(retval))
2730
2731 /* If caller wants to handle missing properties, let them */
2732 if (return_if_undef) {
2733 return NULL;
2734 }
2735 Perl_croak(aTHX_
2736 "Can't find Unicode property definition \"%"SVf"\"",
2737 SVfARG(retval));
2738 Perl_croak(aTHX_ "SWASHNEW didn't return an HV ref");
2739 }
9a53f6cf 2740 } /* End of calling the module to find the swash */
36eb48b4
KW
2741
2742 /* Make sure there is an inversion list for binary properties */
2743 if (minbits == 1) {
2744 SV** swash_invlistsvp = NULL;
2745 SV* swash_invlist = NULL;
9a53f6cf 2746 bool invlist_in_swash_is_valid = FALSE;
88d45d28 2747 HV* swash_hv = NULL;
36eb48b4 2748
9a53f6cf
KW
2749 /* If this operation fetched a swash, get its already existing
2750 * inversion list or create one for it */
2751 if (retval != &PL_sv_undef) {
36eb48b4
KW
2752 swash_hv = MUTABLE_HV(SvRV(retval));
2753
2754 swash_invlistsvp = hv_fetchs(swash_hv, "INVLIST", FALSE);
9a53f6cf
KW
2755 if (swash_invlistsvp) {
2756 swash_invlist = *swash_invlistsvp;
2757 invlist_in_swash_is_valid = TRUE;
2758 }
2759 else {
36eb48b4 2760 swash_invlist = _swash_to_invlist(retval);
9a53f6cf
KW
2761 }
2762 }
2763
2764 /* If an inversion list was passed in, have to include it */
2765 if (invlist) {
2766
2767 /* Any fetched swash will by now have an inversion list in it;
2768 * otherwise <swash_invlist> will be NULL, indicating that we
2769 * didn't fetch a swash */
2770 if (swash_invlist) {
2771
2772 /* Add the passed-in inversion list, which invalidates the one
2773 * already stored in the swash */
2774 invlist_in_swash_is_valid = FALSE;
2775 _invlist_union(invlist, swash_invlist, &swash_invlist);
2776 }
2777 else {
2778
2779 /* Here, there is no swash already. Set up a minimal one */
2780 swash_hv = newHV();
2781 retval = newRV_inc(MUTABLE_SV(swash_hv));
2782 swash_invlist = invlist;
2783 }
2784
2785 if (passed_in_invlist_has_user_defined_property) {
2786 if (! hv_stores(swash_hv, "USER_DEFINED", newSVuv(1))) {
2787 Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed");
2788 }
2789 }
2790 }
2791
2792 /* Here, we have computed the union of all the passed-in data. It may
2793 * be that there was an inversion list in the swash which didn't get
2794 * touched; otherwise save the one computed one */
2795 if (! invlist_in_swash_is_valid) {
69794297
KW
2796 if (! hv_stores(MUTABLE_HV(SvRV(retval)), "INVLIST", swash_invlist))
2797 {
2798 Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed");
2799 }
9a53f6cf 2800 }
36eb48b4
KW
2801 }
2802
a0ed51b3
LW
2803 return retval;
2804}
2805
035d37be
JH
2806
2807/* This API is wrong for special case conversions since we may need to
2808 * return several Unicode characters for a single Unicode character
2809 * (see lib/unicore/SpecCase.txt) The SWASHGET in lib/utf8_heavy.pl is
2810 * the lower-level routine, and it is similarly broken for returning
38684baa
KW
2811 * multiple values. --jhi
2812 * For those, you should use to_utf8_case() instead */
b0e3252e 2813/* Now SWASHGET is recasted into S_swatch_get in this file. */
680c470c
ST
2814
2815/* Note:
2816 * Returns the value of property/mapping C<swash> for the first character
2817 * of the string C<ptr>. If C<do_utf8> is true, the string C<ptr> is
2818 * assumed to be in utf8. If C<do_utf8> is false, the string C<ptr> is
2819 * assumed to be in native 8-bit encoding. Caches the swatch in C<swash>.
af2af982
KW
2820 *
2821 * A "swash" is a hash which contains initially the keys/values set up by
2822 * SWASHNEW. The purpose is to be able to completely represent a Unicode
2823 * property for all possible code points. Things are stored in a compact form
2824 * (see utf8_heavy.pl) so that calculation is required to find the actual
2825 * property value for a given code point. As code points are looked up, new
2826 * key/value pairs are added to the hash, so that the calculation doesn't have
2827 * to ever be re-done. Further, each calculation is done, not just for the
2828 * desired one, but for a whole block of code points adjacent to that one.
2829 * For binary properties on ASCII machines, the block is usually for 64 code
2830 * points, starting with a code point evenly divisible by 64. Thus if the
2831 * property value for code point 257 is requested, the code goes out and
2832 * calculates the property values for all 64 code points between 256 and 319,
2833 * and stores these as a single 64-bit long bit vector, called a "swatch",
2834 * under the key for code point 256. The key is the UTF-8 encoding for code
2835 * point 256, minus the final byte. Thus, if the length of the UTF-8 encoding
2836 * for a code point is 13 bytes, the key will be 12 bytes long. If the value
2837 * for code point 258 is then requested, this code realizes that it would be
2838 * stored under the key for 256, and would find that value and extract the
2839 * relevant bit, offset from 256.
2840 *
2841 * Non-binary properties are stored in as many bits as necessary to represent
2842 * their values (32 currently, though the code is more general than that), not
2843 * as single bits, but the principal is the same: the value for each key is a
2844 * vector that encompasses the property values for all code points whose UTF-8
2845 * representations are represented by the key. That is, for all code points
2846 * whose UTF-8 representations are length N bytes, and the key is the first N-1
2847 * bytes of that.
680c470c 2848 */
a0ed51b3 2849UV
680c470c 2850Perl_swash_fetch(pTHX_ SV *swash, const U8 *ptr, bool do_utf8)
a0ed51b3 2851{
27da23d5 2852 dVAR;
ef8f7699 2853 HV *const hv = MUTABLE_HV(SvRV(swash));
3568d838
JH
2854 U32 klen;
2855 U32 off;
a0ed51b3 2856 STRLEN slen;
7d85a32c 2857 STRLEN needents;
cfd0369c 2858 const U8 *tmps = NULL;
a0ed51b3 2859 U32 bit;
979f2922 2860 SV *swatch;
3568d838 2861 U8 tmputf8[2];
35da51f7 2862 const UV c = NATIVE_TO_ASCII(*ptr);
3568d838 2863
7918f24d
NC
2864 PERL_ARGS_ASSERT_SWASH_FETCH;
2865
dbe7a391 2866 /* Convert to utf8 if not already */
3568d838 2867 if (!do_utf8 && !UNI_IS_INVARIANT(c)) {
979f2922
ST
2868 tmputf8[0] = (U8)UTF8_EIGHT_BIT_HI(c);
2869 tmputf8[1] = (U8)UTF8_EIGHT_BIT_LO(c);
2870 ptr = tmputf8;
3568d838
JH
2871 }
2872 /* Given a UTF-X encoded char 0xAA..0xYY,0xZZ
37e2e78e 2873 * then the "swatch" is a vec() for all the chars which start
3568d838
JH
2874 * with 0xAA..0xYY
2875 * So the key in the hash (klen) is length of encoded char -1
2876 */
2877 klen = UTF8SKIP(ptr) - 1;
2878 off = ptr[klen];
a0ed51b3 2879
979f2922 2880 if (klen == 0) {
37e2e78e 2881 /* If char is invariant then swatch is for all the invariant chars
1e54db1a 2882 * In both UTF-8 and UTF-8-MOD that happens to be UTF_CONTINUATION_MARK
7d85a32c 2883 */
979f2922
ST
2884 needents = UTF_CONTINUATION_MARK;
2885 off = NATIVE_TO_UTF(ptr[klen]);
2886 }
2887 else {
7d85a32c 2888 /* If char is encoded then swatch is for the prefix */
979f2922
ST
2889 needents = (1 << UTF_ACCUMULATION_SHIFT);
2890 off = NATIVE_TO_UTF(ptr[klen]) & UTF_CONTINUATION_MASK;
2891 }
7d85a32c 2892
a0ed51b3
LW
2893 /*
2894 * This single-entry cache saves about 1/3 of the utf8 overhead in test
2895 * suite. (That is, only 7-8% overall over just a hash cache. Still,
2896 * it's nothing to sniff at.) Pity we usually come through at least
2897 * two function calls to get here...
2898 *
2899 * NB: this code assumes that swatches are never modified, once generated!
2900 */
2901
3568d838 2902 if (hv == PL_last_swash_hv &&
a0ed51b3 2903 klen == PL_last_swash_klen &&
27da23d5 2904 (!klen || memEQ((char *)ptr, (char *)PL_last_swash_key, klen)) )
a0ed51b3
LW
2905 {
2906 tmps = PL_last_swash_tmps;
2907 slen = PL_last_swash_slen;
2908 }
2909 else {
2910 /* Try our second-level swatch cache, kept in a hash. */
e1ec3a88 2911 SV** svp = hv_fetch(hv, (const char*)ptr, klen, FALSE);
a0ed51b3 2912
b0e3252e 2913 /* If not cached, generate it via swatch_get */
979f2922
ST
2914 if (!svp || !SvPOK(*svp)
2915 || !(tmps = (const U8*)SvPV_const(*svp, slen))) {
2b9d42f0
NIS
2916 /* We use utf8n_to_uvuni() as we want an index into
2917 Unicode tables, not a native character number.
2918 */
aec46f14 2919 const UV code_point = utf8n_to_uvuni(ptr, UTF8_MAXBYTES, 0,
872c91ae
JH
2920 ckWARN(WARN_UTF8) ?
2921 0 : UTF8_ALLOW_ANY);
b0e3252e 2922 swatch = swatch_get(swash,
979f2922 2923 /* On EBCDIC & ~(0xA0-1) isn't a useful thing to do */
361ee0fe 2924 (klen) ? (code_point & ~((UV)needents - 1)) : 0,
979f2922
ST
2925 needents);
2926
923e4eb5 2927 if (IN_PERL_COMPILETIME)
623e6609 2928 CopHINTS_set(PL_curcop, PL_hints);
a0ed51b3 2929
979f2922 2930 svp = hv_store(hv, (const char *)ptr, klen, swatch, 0);
a0ed51b3 2931
979f2922
ST
2932 if (!svp || !(tmps = (U8*)SvPV(*svp, slen))
2933 || (slen << 3) < needents)
5637ef5b
NC
2934 Perl_croak(aTHX_ "panic: swash_fetch got improper swatch, "
2935 "svp=%p, tmps=%p, slen=%"UVuf", needents=%"UVuf,
2936 svp, tmps, (UV)slen, (UV)needents);
a0ed51b3
LW
2937 }
2938
2939 PL_last_swash_hv = hv;
16d8f38a 2940 assert(klen <= sizeof(PL_last_swash_key));
eac04b2e 2941 PL_last_swash_klen = (U8)klen;
cfd0369c
NC
2942 /* FIXME change interpvar.h? */
2943 PL_last_swash_tmps = (U8 *) tmps;
a0ed51b3
LW
2944 PL_last_swash_slen = slen;
2945 if (klen)
2946 Copy(ptr, PL_last_swash_key, klen, U8);
2947 }
2948
a410ec23
KW
2949 if (UTF8_IS_SUPER(ptr) && ckWARN_d(WARN_NON_UNICODE)) {
2950 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
2951
2952 /* This outputs warnings for binary properties only, assuming that
2953 * to_utf8_case() will output any for non-binary. Also, surrogates
2954 * aren't checked for, as that would warn on things like /\p{Gc=Cs}/ */
2955
9a53f6cf 2956 if (! bitssvp || SvUV(*bitssvp) == 1) {
a410ec23
KW
2957 /* User-defined properties can silently match above-Unicode */
2958 SV** const user_defined_svp = hv_fetchs(hv, "USER_DEFINED", FALSE);
2959 if (! user_defined_svp || ! SvUV(*user_defined_svp)) {
2960 const UV code_point = utf8n_to_uvuni(ptr, UTF8_MAXBYTES, 0, 0);
2961 Perl_warner(aTHX_ packWARN(WARN_NON_UNICODE),
2962 "Code point 0x%04"UVXf" is not Unicode, all \\p{} matches fail; all \\P{} matches succeed", code_point);
2963 }
2964 }
2965 }
2966
9faf8d75 2967 switch ((int)((slen << 3) / needents)) {
a0ed51b3
LW
2968 case 1:
2969 bit = 1 << (off & 7);
2970 off >>= 3;
2971 return (tmps[off] & bit) != 0;
2972 case 8:
2973 return tmps[off];
2974 case 16:
2975 off <<= 1;
2976 return (tmps[off] << 8) + tmps[off + 1] ;
2977 case 32:
2978 off <<= 2;
2979 return (tmps[off] << 24) + (tmps[off+1] << 16) + (tmps[off+2] << 8) + tmps[off + 3] ;
2980 }
5637ef5b
NC
2981 Perl_croak(aTHX_ "panic: swash_fetch got swatch of unexpected bit width, "
2982 "slen=%"UVuf", needents=%"UVuf, (UV)slen, (UV)needents);
670f1322 2983 NORETURN_FUNCTION_END;
a0ed51b3 2984}
2b9d42f0 2985
319009ee
KW
2986/* Read a single line of the main body of the swash input text. These are of
2987 * the form:
2988 * 0053 0056 0073
2989 * where each number is hex. The first two numbers form the minimum and
2990 * maximum of a range, and the third is the value associated with the range.
2991 * Not all swashes should have a third number
2992 *
2993 * On input: l points to the beginning of the line to be examined; it points
2994 * to somewhere in the string of the whole input text, and is
2995 * terminated by a \n or the null string terminator.
2996 * lend points to the null terminator of that string
2997 * wants_value is non-zero if the swash expects a third number
2998 * typestr is the name of the swash's mapping, like 'ToLower'
2999 * On output: *min, *max, and *val are set to the values read from the line.
3000 * returns a pointer just beyond the line examined. If there was no
3001 * valid min number on the line, returns lend+1
3002 */
3003
3004STATIC U8*
3005S_swash_scan_list_line(pTHX_ U8* l, U8* const lend, UV* min, UV* max, UV* val,
3006 const bool wants_value, const U8* const typestr)
3007{
3008 const int typeto = typestr[0] == 'T' && typestr[1] == 'o';
3009 STRLEN numlen; /* Length of the number */
02470786
KW
3010 I32 flags = PERL_SCAN_SILENT_ILLDIGIT
3011 | PERL_SCAN_DISALLOW_PREFIX
3012 | PERL_SCAN_SILENT_NON_PORTABLE;
319009ee
KW
3013
3014 /* nl points to the next \n in the scan */
3015 U8* const nl = (U8*)memchr(l, '\n', lend - l);
3016
3017 /* Get the first number on the line: the range minimum */
3018 numlen = lend - l;
3019 *min = grok_hex((char *)l, &numlen, &flags, NULL);
3020 if (numlen) /* If found a hex number, position past it */
3021 l += numlen;
3022 else if (nl) { /* Else, go handle next line, if any */
3023 return nl + 1; /* 1 is length of "\n" */
3024 }
3025 else { /* Else, no next line */
3026 return lend + 1; /* to LIST's end at which \n is not found */
3027 }
3028
3029 /* The max range value follows, separated by a BLANK */
3030 if (isBLANK(*l)) {
3031 ++l;
02470786
KW
3032 flags = PERL_SCAN_SILENT_ILLDIGIT
3033 | PERL_SCAN_DISALLOW_PREFIX
3034 | PERL_SCAN_SILENT_NON_PORTABLE;
319009ee
KW
3035 numlen = lend - l;
3036 *max = grok_hex((char *)l, &numlen, &flags, NULL);
3037 if (numlen)
3038 l += numlen;
3039 else /* If no value here, it is a single element range */
3040 *max = *min;
3041
3042 /* Non-binary tables have a third entry: what the first element of the
3043 * range maps to */
3044 if (wants_value) {
3045 if (isBLANK(*l)) {
3046 ++l;
f90a9a02
KW
3047
3048 /* The ToLc, etc table mappings are not in hex, and must be
3049 * corrected by adding the code point to them */
3050 if (typeto) {
3051 char *after_strtol = (char *) lend;
3052 *val = Strtol((char *)l, &after_strtol, 10);
3053 l = (U8 *) after_strtol;
f90a9a02
KW
3054 }
3055 else { /* Other tables are in hex, and are the correct result
3056 without tweaking */
a9d188b3
KW
3057 flags = PERL_SCAN_SILENT_ILLDIGIT
3058 | PERL_SCAN_DISALLOW_PREFIX
3059 | PERL_SCAN_SILENT_NON_PORTABLE;
3060 numlen = lend - l;
3061 *val = grok_hex((char *)l, &numlen, &flags, NULL);
3062 if (numlen)
3063 l += numlen;
3064 else
3065 *val = 0;
f90a9a02 3066 }
319009ee
KW
3067 }
3068 else {
3069 *val = 0;
3070 if (typeto) {
dcbac5bb 3071 /* diag_listed_as: To%s: illegal mapping '%s' */
319009ee
KW
3072 Perl_croak(aTHX_ "%s: illegal mapping '%s'",
3073 typestr, l);
3074 }
3075 }
3076 }
3077 else
3078 *val = 0; /* bits == 1, then any val should be ignored */
3079 }
3080 else { /* Nothing following range min, should be single element with no
3081 mapping expected */
3082 *max = *min;
3083 if (wants_value) {
3084 *val = 0;
3085 if (typeto) {
dcbac5bb 3086 /* diag_listed_as: To%s: illegal mapping '%s' */
319009ee
KW
3087 Perl_croak(aTHX_ "%s: illegal mapping '%s'", typestr, l);
3088 }
3089 }
3090 else
3091 *val = 0; /* bits == 1, then val should be ignored */
3092 }
3093
3094 /* Position to next line if any, or EOF */
3095 if (nl)
3096 l = nl + 1;
3097 else
3098 l = lend;
3099
3100 return l;
3101}
3102
979f2922
ST
3103/* Note:
3104 * Returns a swatch (a bit vector string) for a code point sequence
3105 * that starts from the value C<start> and comprises the number C<span>.
3106 * A C<swash> must be an object created by SWASHNEW (see lib/utf8_heavy.pl).
3107 * Should be used via swash_fetch, which will cache the swatch in C<swash>.
3108 */
3109STATIC SV*
b0e3252e 3110S_swatch_get(pTHX_ SV* swash, UV start, UV span)
979f2922
ST
3111{
3112 SV *swatch;
77f9f126 3113 U8 *l, *lend, *x, *xend, *s, *send;
979f2922 3114 STRLEN lcur, xcur, scur;
ef8f7699 3115 HV *const hv = MUTABLE_HV(SvRV(swash));
36eb48b4
KW
3116 SV** const invlistsvp = hv_fetchs(hv, "INVLIST", FALSE);
3117
88d45d28
KW
3118 SV** listsvp = NULL; /* The string containing the main body of the table */
3119 SV** extssvp = NULL;
3120 SV** invert_it_svp = NULL;
3121 U8* typestr = NULL;
786861f5
KW
3122 STRLEN bits;
3123 STRLEN octets; /* if bits == 1, then octets == 0 */
3124 UV none;
3125 UV end = start + span;
972dd592 3126
36eb48b4 3127 if (invlistsvp == NULL) {
786861f5
KW
3128 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
3129 SV** const nonesvp = hv_fetchs(hv, "NONE", FALSE);
3130 SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE);
3131 extssvp = hv_fetchs(hv, "EXTRAS", FALSE);
3132 listsvp = hv_fetchs(hv, "LIST", FALSE);
3133 invert_it_svp = hv_fetchs(hv, "INVERT_IT", FALSE);
3134
3135 bits = SvUV(*bitssvp);
3136 none = SvUV(*nonesvp);
3137 typestr = (U8*)SvPV_nolen(*typesvp);
3138 }
36eb48b4
KW
3139 else {
3140 bits = 1;
3141 none = 0;
3142 }
786861f5 3143 octets = bits >> 3; /* if bits == 1, then octets == 0 */
979f2922 3144
b0e3252e 3145 PERL_ARGS_ASSERT_SWATCH_GET;
7918f24d 3146
979f2922 3147 if (bits != 1 && bits != 8 && bits != 16 && bits != 32) {
b0e3252e 3148 Perl_croak(aTHX_ "panic: swatch_get doesn't expect bits %"UVuf,
660a4616 3149 (UV)bits);
979f2922
ST
3150 }
3151
84ea5ef6
KW
3152 /* If overflowed, use the max possible */
3153 if (end < start) {
3154 end = UV_MAX;
3155 span = end - start;
3156 }
3157
979f2922 3158 /* create and initialize $swatch */
979f2922 3159 scur = octets ? (span * octets) : (span + 7) / 8;
e524fe40
NC
3160 swatch = newSV(scur);
3161 SvPOK_on(swatch);
979f2922
ST
3162 s = (U8*)SvPVX(swatch);
3163 if (octets && none) {
0bd48802 3164 const U8* const e = s + scur;
979f2922
ST
3165 while (s < e) {
3166 if (bits == 8)
3167 *s++ = (U8)(none & 0xff);
3168 else if (bits == 16) {
3169 *s++ = (U8)((none >> 8) & 0xff);
3170 *s++ = (U8)( none & 0xff);
3171 }
3172 else if (bits == 32) {
3173 *s++ = (U8)((none >> 24) & 0xff);
3174 *s++ = (U8)((none >> 16) & 0xff);
3175 *s++ = (U8)((none >> 8) & 0xff);
3176 *s++ = (U8)( none & 0xff);
3177 }
3178 }
3179 *s = '\0';
3180 }
3181 else {
3182 (void)memzero((U8*)s, scur + 1);
3183 }
3184 SvCUR_set(swatch, scur);
3185 s = (U8*)SvPVX(swatch);
3186
36eb48b4
KW
3187 if (invlistsvp) { /* If has an inversion list set up use that */
3188 _invlist_populate_swatch(*invlistsvp, start, end, s);
3189 return swatch;
3190 }
3191
3192 /* read $swash->{LIST} */
979f2922
ST
3193 l = (U8*)SvPV(*listsvp, lcur);
3194 lend = l + lcur;
3195 while (l < lend) {
8ed25d53 3196 UV min, max, val, upper;
319009ee
KW
3197 l = S_swash_scan_list_line(aTHX_ l, lend, &min, &max, &val,
3198 cBOOL(octets), typestr);
3199 if (l > lend) {
979f2922
ST
3200 break;
3201 }
3202
972dd592 3203 /* If looking for something beyond this range, go try the next one */
979f2922
ST
3204 if (max < start)
3205 continue;
3206
8ed25d53
KW
3207 /* <end> is generally 1 beyond where we want to set things, but at the
3208 * platform's infinity, where we can't go any higher, we want to
3209 * include the code point at <end> */
3210 upper = (max < end)
3211 ? max
3212 : (max != UV_MAX || end != UV_MAX)
3213 ? end - 1
3214 : end;
3215
979f2922 3216 if (octets) {
35da51f7 3217 UV key;
979f2922
ST
3218 if (min < start) {
3219 if (!none || val < none) {
3220 val += start - min;
3221 }
3222 min = start;
3223 }
8ed25d53 3224 for (key = min; key <= upper; key++) {
979f2922 3225 STRLEN offset;
979f2922
ST
3226 /* offset must be non-negative (start <= min <= key < end) */
3227 offset = octets * (key - start);
3228 if (bits == 8)
3229 s[offset] = (U8)(val & 0xff);
3230 else if (bits == 16) {
3231 s[offset ] = (U8)((val >> 8) & 0xff);
3232 s[offset + 1] = (U8)( val & 0xff);
3233 }
3234 else if (bits == 32) {
3235 s[offset ] = (U8)((val >> 24) & 0xff);
3236 s[offset + 1] = (U8)((val >> 16) & 0xff);
3237 s[offset + 2] = (U8)((val >> 8) & 0xff);
3238 s[offset + 3] = (U8)( val & 0xff);
3239 }
3240
3241 if (!none || val < none)
3242 ++val;
3243 }
3244 }
711a919c 3245 else { /* bits == 1, then val should be ignored */
35da51f7 3246 UV key;
979f2922
ST
3247 if (min < start)
3248 min = start;
6cb05c12 3249
8ed25d53 3250 for (key = min; key <= upper; key++) {
0bd48802 3251 const STRLEN offset = (STRLEN)(key - start);
979f2922
ST
3252 s[offset >> 3] |= 1 << (offset & 7);
3253 }
3254 }
3255 } /* while */
979f2922 3256
9479a769 3257 /* Invert if the data says it should be. Assumes that bits == 1 */
77f9f126 3258 if (invert_it_svp && SvUV(*invert_it_svp)) {
0bda3001
KW
3259
3260 /* Unicode properties should come with all bits above PERL_UNICODE_MAX
3261 * be 0, and their inversion should also be 0, as we don't succeed any
3262 * Unicode property matches for non-Unicode code points */
3263 if (start <= PERL_UNICODE_MAX) {
3264
3265 /* The code below assumes that we never cross the
3266 * Unicode/above-Unicode boundary in a range, as otherwise we would
3267 * have to figure out where to stop flipping the bits. Since this
3268 * boundary is divisible by a large power of 2, and swatches comes
3269 * in small powers of 2, this should be a valid assumption */
3270 assert(start + span - 1 <= PERL_UNICODE_MAX);
3271
507a8485
KW
3272 send = s + scur;
3273 while (s < send) {
3274 *s = ~(*s);
3275 s++;
3276 }
0bda3001 3277 }
77f9f126
KW
3278 }
3279
d73c39c5
KW
3280 /* read $swash->{EXTRAS}
3281 * This code also copied to swash_to_invlist() below */
979f2922
ST
3282 x = (U8*)SvPV(*extssvp, xcur);
3283 xend = x + xcur;
3284 while (x < xend) {
3285 STRLEN namelen;
3286 U8 *namestr;
3287 SV** othersvp;
3288 HV* otherhv;
3289 STRLEN otherbits;
3290 SV **otherbitssvp, *other;
711a919c 3291 U8 *s, *o, *nl;
979f2922
ST
3292 STRLEN slen, olen;
3293
35da51f7 3294 const U8 opc = *x++;
979f2922
ST
3295 if (opc == '\n')
3296 continue;
3297
3298 nl = (U8*)memchr(x, '\n', xend - x);
3299
3300 if (opc != '-' && opc != '+' && opc != '!' && opc != '&') {
3301 if (nl) {
3302 x = nl + 1; /* 1 is length of "\n" */
3303 continue;
3304 }
3305 else {
3306 x = xend; /* to EXTRAS' end at which \n is not found */
3307 break;
3308 }
3309 }
3310
3311 namestr = x;
3312 if (nl) {
3313 namelen = nl - namestr;
3314 x = nl + 1;
3315 }
3316 else {
3317 namelen = xend - namestr;
3318 x = xend;
3319 }
3320
3321 othersvp = hv_fetch(hv, (char *)namestr, namelen, FALSE);
ef8f7699 3322 otherhv = MUTABLE_HV(SvRV(*othersvp));
017a3ce5 3323 otherbitssvp = hv_fetchs(otherhv, "BITS", FALSE);
979f2922
ST
3324 otherbits = (STRLEN)SvUV(*otherbitssvp);
3325 if (bits < otherbits)
5637ef5b
NC
3326 Perl_croak(aTHX_ "panic: swatch_get found swatch size mismatch, "
3327 "bits=%"UVuf", otherbits=%"UVuf, (UV)bits, (UV)otherbits);
979f2922
ST
3328
3329 /* The "other" swatch must be destroyed after. */
b0e3252e 3330 other = swatch_get(*othersvp, start, span);
979f2922
ST
3331 o = (U8*)SvPV(other, olen);
3332
3333 if (!olen)
b0e3252e 3334 Perl_croak(aTHX_ "panic: swatch_get got improper swatch");
979f2922
ST
3335
3336 s = (U8*)SvPV(swatch, slen);
3337 if (bits == 1 && otherbits == 1) {
3338 if (slen != olen)
5637ef5b
NC
3339 Perl_croak(aTHX_ "panic: swatch_get found swatch length "
3340 "mismatch, slen=%"UVuf", olen=%"UVuf,
3341 (UV)slen, (UV)olen);
979f2922
ST
3342
3343 switch (opc) {
3344 case '+':
3345 while (slen--)
3346 *s++ |= *o++;
3347 break;
3348 case '!':
3349 while (slen--)
3350 *s++ |= ~*o++;
3351 break;
3352 case '-':
3353 while (slen--)
3354 *s++ &= ~*o++;
3355 break;
3356 case '&':
3357 while (slen--)
3358 *s++ &= *o++;
3359 break;
3360 default:
3361 break;
3362 }
3363 }
711a919c 3364 else {
979f2922
ST
3365 STRLEN otheroctets = otherbits >> 3;
3366 STRLEN offset = 0;
35da51f7 3367 U8* const send = s + slen;
979f2922
ST
3368
3369 while (s < send) {
3370 UV otherval = 0;
3371
3372 if (otherbits == 1) {
3373 otherval = (o[offset >> 3] >> (offset & 7)) & 1;
3374 ++offset;
3375 }
3376 else {
3377 STRLEN vlen = otheroctets;
3378 otherval = *o++;
3379 while (--vlen) {
3380 otherval <<= 8;
3381 otherval |= *o++;
3382 }
3383 }
3384
711a919c 3385 if (opc == '+' && otherval)
6f207bd3 3386 NOOP; /* replace with otherval */
979f2922
ST
3387 else if (opc == '!' && !otherval)
3388 otherval = 1;
3389 else if (opc == '-' && otherval)
3390 otherval = 0;
3391 else if (opc == '&' && !otherval)
3392 otherval = 0;
3393 else {
711a919c 3394 s += octets; /* no replacement */
979f2922
ST
3395 continue;
3396 }
3397
3398 if (bits == 8)
3399 *s++ = (U8)( otherval & 0xff);
3400 else if (bits == 16) {
3401 *s++ = (U8)((otherval >> 8) & 0xff);
3402 *s++ = (U8)( otherval & 0xff);
3403 }
3404 else if (bits == 32) {
3405 *s++ = (U8)((otherval >> 24) & 0xff);
3406 *s++ = (U8)((otherval >> 16) & 0xff);
3407 *s++ = (U8)((otherval >> 8) & 0xff);
3408 *s++ = (U8)( otherval & 0xff);
3409 }
3410 }
3411 }
3412 sv_free(other); /* through with it! */
3413 } /* while */
3414 return swatch;
3415}
3416
064c021d 3417HV*
4c2e1131 3418Perl__swash_inversion_hash(pTHX_ SV* const swash)
064c021d
KW
3419{
3420
5662e334
KW
3421 /* Subject to change or removal. For use only in one place in regcomp.c.
3422 * Can't be used on a property that is subject to user override, as it
3423 * relies on the value of SPECIALS in the swash which would be set by
3424 * utf8_heavy.pl to the hash in the non-overriden file, and hence is not set
3425 * for overridden properties
064c021d
KW
3426 *
3427 * Returns a hash which is the inversion and closure of a swash mapping.
3428 * For example, consider the input lines:
3429 * 004B 006B
3430 * 004C 006C
3431 * 212A 006B
3432 *
3433 * The returned hash would have two keys, the utf8 for 006B and the utf8 for
3434 * 006C. The value for each key is an array. For 006C, the array would
3435 * have a two elements, the utf8 for itself, and for 004C. For 006B, there
3436 * would be three elements in its array, the utf8 for 006B, 004B and 212A.
3437 *
3438 * Essentially, for any code point, it gives all the code points that map to
3439 * it, or the list of 'froms' for that point.
3440 *
5662e334
KW
3441 * Currently it ignores any additions or deletions from other swashes,
3442 * looking at just the main body of the swash, and if there are SPECIALS
3443 * in the swash, at that hash
3444 *
3445 * The specials hash can be extra code points, and most likely consists of
3446 * maps from single code points to multiple ones (each expressed as a string
3447 * of utf8 characters). This function currently returns only 1-1 mappings.
3448 * However consider this possible input in the specials hash:
3449 * "\xEF\xAC\x85" => "\x{0073}\x{0074}", # U+FB05 => 0073 0074
3450 * "\xEF\xAC\x86" => "\x{0073}\x{0074}", # U+FB06 => 0073 0074
3451 *
3452 * Both FB05 and FB06 map to the same multi-char sequence, which we don't
3453 * currently handle. But it also means that FB05 and FB06 are equivalent in
3454 * a 1-1 mapping which we should handle, and this relationship may not be in
3455 * the main table. Therefore this function examines all the multi-char
3456 * sequences and adds the 1-1 mappings that come out of that. */
064c021d
KW
3457
3458 U8 *l, *lend;
3459 STRLEN lcur;
3460 HV *const hv = MUTABLE_HV(SvRV(swash));
3461
3462 /* The string containing the main body of the table */
3463 SV** const listsvp = hv_fetchs(hv, "LIST", FALSE);
3464
3465 SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE);
3466 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
3467 SV** const nonesvp = hv_fetchs(hv, "NONE", FALSE);
3468 /*SV** const extssvp = hv_fetchs(hv, "EXTRAS", FALSE);*/
3469 const U8* const typestr = (U8*)SvPV_nolen(*typesvp);
3470 const STRLEN bits = SvUV(*bitssvp);
3471 const STRLEN octets = bits >> 3; /* if bits == 1, then octets == 0 */
3472 const UV none = SvUV(*nonesvp);
5662e334 3473 SV **specials_p = hv_fetchs(hv, "SPECIALS", 0);
064c021d
KW
3474
3475 HV* ret = newHV();
3476
3477 PERL_ARGS_ASSERT__SWASH_INVERSION_HASH;
3478
3479 /* Must have at least 8 bits to get the mappings */
3480 if (bits != 8 && bits != 16 && bits != 32) {
3481 Perl_croak(aTHX_ "panic: swash_inversion_hash doesn't expect bits %"UVuf,
3482 (UV)bits);
3483 }
3484
5662e334
KW
3485 if (specials_p) { /* It might be "special" (sometimes, but not always, a
3486 mapping to more than one character */
3487
3488 /* Construct an inverse mapping hash for the specials */
3489 HV * const specials_hv = MUTABLE_HV(SvRV(*specials_p));
3490 HV * specials_inverse = newHV();
3491 char *char_from; /* the lhs of the map */
3492 I32 from_len; /* its byte length */
3493 char *char_to; /* the rhs of the map */
3494 I32 to_len; /* its byte length */
3495 SV *sv_to; /* and in a sv */
3496 AV* from_list; /* list of things that map to each 'to' */
3497
3498 hv_iterinit(specials_hv);
3499
3500 /* The keys are the characters (in utf8) that map to the corresponding
3501 * utf8 string value. Iterate through the list creating the inverse
3502 * list. */
3503 while ((sv_to = hv_iternextsv(specials_hv, &char_from, &from_len))) {
3504 SV** listp;
3505 if (! SvPOK(sv_to)) {
5637ef5b
NC
3506 Perl_croak(aTHX_ "panic: value returned from hv_iternextsv() "
3507 "unexpectedly is not a string, flags=%lu",
3508 (unsigned long)SvFLAGS(sv_to));
5662e334 3509 }
4b88fb76 3510 /*DEBUG_U(PerlIO_printf(Perl_debug_log, "Found mapping from %"UVXf", First char of to is %"UVXf"\n", valid_utf8_to_uvchr((U8*) char_from, 0), valid_utf8_to_uvchr((U8*) SvPVX(sv_to), 0)));*/
5662e334
KW
3511
3512 /* Each key in the inverse list is a mapped-to value, and the key's
3513 * hash value is a list of the strings (each in utf8) that map to
3514 * it. Those strings are all one character long */
3515 if ((listp = hv_fetch(specials_inverse,
3516 SvPVX(sv_to),
3517 SvCUR(sv_to), 0)))
3518 {
3519 from_list = (AV*) *listp;
3520 }
3521 else { /* No entry yet for it: create one */
3522 from_list = newAV();
3523 if (! hv_store(specials_inverse,
3524 SvPVX(sv_to),
3525 SvCUR(sv_to),
3526 (SV*) from_list, 0))
3527 {
3528 Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed");
3529 }
3530 }
3531
3532 /* Here have the list associated with this 'to' (perhaps newly
3533 * created and empty). Just add to it. Note that we ASSUME that
3534 * the input is guaranteed to not have duplications, so we don't
3535 * check for that. Duplications just slow down execution time. */
3536 av_push(from_list, newSVpvn_utf8(char_from, from_len, TRUE));
3537 }
3538
3539 /* Here, 'specials_inverse' contains the inverse mapping. Go through
3540 * it looking for cases like the FB05/FB06 examples above. There would
3541 * be an entry in the hash like
3542 * 'st' => [ FB05, FB06 ]
3543 * In this example we will create two lists that get stored in the
3544 * returned hash, 'ret':
3545 * FB05 => [ FB05, FB06 ]
3546 * FB06 => [ FB05, FB06 ]
3547 *
3548 * Note that there is nothing to do if the array only has one element.
3549 * (In the normal 1-1 case handled below, we don't have to worry about
3550 * two lists, as everything gets tied to the single list that is
3551 * generated for the single character 'to'. But here, we are omitting
3552 * that list, ('st' in the example), so must have multiple lists.) */
3553 while ((from_list = (AV *) hv_iternextsv(specials_inverse,
3554 &char_to, &to_len)))
3555 {
3556 if (av_len(from_list) > 0) {
3557 int i;
3558
3559 /* We iterate over all combinations of i,j to place each code
3560 * point on each list */
3561 for (i = 0; i <= av_len(from_list); i++) {
3562 int j;
3563 AV* i_list = newAV();
3564 SV** entryp = av_fetch(from_list, i, FALSE);
3565 if (entryp == NULL) {
3566 Perl_croak(aTHX_ "panic: av_fetch() unexpectedly failed");
3567 }
3568 if (hv_fetch(ret, SvPVX(*entryp), SvCUR(*entryp), FALSE)) {
3569 Perl_croak(aTHX_ "panic: unexpected entry for %s", SvPVX(*entryp));
3570 }
3571 if (! hv_store(ret, SvPVX(*entryp), SvCUR(*entryp),
3572 (SV*) i_list, FALSE))
3573 {
3574 Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed");
3575 }
3576
4b88fb76 3577 /* For debugging: UV u = valid_utf8_to_uvchr((U8*) SvPVX(*entryp), 0);*/
5662e334
KW
3578 for (j = 0; j <= av_len(from_list); j++) {
3579 entryp = av_fetch(from_list, j, FALSE);
3580 if (entryp == NULL) {
3581 Perl_croak(aTHX_ "panic: av_fetch() unexpectedly failed");
3582 }
3583
3584 /* When i==j this adds itself to the list */
4b88fb76
KW
3585 av_push(i_list, newSVuv(utf8_to_uvchr_buf(
3586 (U8*) SvPVX(*entryp),
3587 (U8*) SvPVX(*entryp) + SvCUR(*entryp),
3588 0)));
3589 /*DEBUG_U(PerlIO_printf(Perl_debug_log, "Adding %"UVXf" to list for %"UVXf"\n", valid_utf8_to_uvchr((U8*) SvPVX(*entryp), 0), u));*/
5662e334
KW
3590 }
3591 }
3592 }
3593 }
3594 SvREFCNT_dec(specials_inverse); /* done with it */
3595 } /* End of specials */
3596
064c021d
KW
3597 /* read $swash->{LIST} */
3598 l = (U8*)SvPV(*listsvp, lcur);
3599 lend = l + lcur;
3600
3601 /* Go through each input line */
3602 while (l < lend) {
3603 UV min, max, val;
3604 UV inverse;
3605 l = S_swash_scan_list_line(aTHX_ l, lend, &min, &max, &val,
3606 cBOOL(octets), typestr);
3607 if (l > lend) {
3608 break;
3609 }
3610
3611 /* Each element in the range is to be inverted */
3612 for (inverse = min; inverse <= max; inverse++) {
3613 AV* list;
064c021d
KW
3614 SV** listp;
3615 IV i;
3616 bool found_key = FALSE;
5662e334 3617 bool found_inverse = FALSE;
064c021d
KW
3618
3619 /* The key is the inverse mapping */
3620 char key[UTF8_MAXBYTES+1];
3621 char* key_end = (char *) uvuni_to_utf8((U8*) key, val);
3622 STRLEN key_len = key_end - key;
3623
064c021d
KW
3624 /* Get the list for the map */
3625 if ((listp = hv_fetch(ret, key, key_len, FALSE))) {
3626 list = (AV*) *listp;
3627 }
3628 else { /* No entry yet for it: create one */
3629 list = newAV();
3630 if (! hv_store(ret, key, key_len, (SV*) list, FALSE)) {
3631 Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed");
3632 }
3633 }
3634
5662e334
KW
3635 /* Look through list to see if this inverse mapping already is
3636 * listed, or if there is a mapping to itself already */
508f7cfa 3637 for (i = 0; i <= av_len(list); i++) {
064c021d
KW
3638 SV** entryp = av_fetch(list, i, FALSE);
3639 SV* entry;
3640 if (entryp == NULL) {
3641 Perl_croak(aTHX_ "panic: av_fetch() unexpectedly failed");
3642 }
3643 entry = *entryp;
5662e334 3644 /*DEBUG_U(PerlIO_printf(Perl_debug_log, "list for %"UVXf" contains %"UVXf"\n", val, SvUV(entry)));*/
56ca34ca 3645 if (SvUV(entry) == val) {
064c021d 3646 found_key = TRUE;
5662e334
KW
3647 }
3648 if (SvUV(entry) == inverse) {
3649 found_inverse = TRUE;
3650 }
3651
3652 /* No need to continue searching if found everything we are
3653 * looking for */
3654 if (found_key && found_inverse) {
064c021d
KW
3655 break;
3656 }
3657 }
56ca34ca
KW
3658
3659 /* Make sure there is a mapping to itself on the list */
064c021d 3660 if (! found_key) {
d397ff6a 3661 av_push(list, newSVuv(val));
5662e334 3662 /*DEBUG_U(PerlIO_printf(Perl_debug_log, "Adding %"UVXf" to list for %"UVXf"\n", val, val));*/
064c021d
KW
3663 }
3664
3665
3666 /* Simply add the value to the list */
5662e334
KW
3667 if (! found_inverse) {
3668 av_push(list, newSVuv(inverse));
3669 /*DEBUG_U(PerlIO_printf(Perl_debug_log, "Adding %"UVXf" to list for %"UVXf"\n", inverse, val));*/
3670 }
064c021d 3671
b0e3252e 3672 /* swatch_get() increments the value of val for each element in the
064c021d
KW
3673 * range. That makes more compact tables possible. You can
3674 * express the capitalization, for example, of all consecutive
3675 * letters with a single line: 0061\t007A\t0041 This maps 0061 to
3676 * 0041, 0062 to 0042, etc. I (khw) have never understood 'none',
bd3f2f94 3677 * and it's not documented; it appears to be used only in
b0e3252e 3678 * implementing tr//; I copied the semantics from swatch_get(), just
bd3f2f94 3679 * in case */
064c021d
KW
3680 if (!none || val < none) {
3681 ++val;
3682 }
3683 }
3684 }
3685
3686 return ret;
3687}
3688
a25abddc 3689SV*
d764b54e
KW
3690Perl__swash_to_invlist(pTHX_ SV* const swash)
3691{
3692
3693 /* Subject to change or removal. For use only in one place in regcomp.c */
3694
3695 U8 *l, *lend;
3696 char *loc;
3697 STRLEN lcur;
3698 HV *const hv = MUTABLE_HV(SvRV(swash));
3699 UV elements = 0; /* Number of elements in the inversion list */
b443038a 3700 U8 empty[] = "";
d764b54e
KW
3701
3702 /* The string containing the main body of the table */
3703 SV** const listsvp = hv_fetchs(hv, "LIST", FALSE);
3704 SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE);
3705 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
d73c39c5 3706 SV** const extssvp = hv_fetchs(hv, "EXTRAS", FALSE);
77f9f126 3707 SV** const invert_it_svp = hv_fetchs(hv, "INVERT_IT", FALSE);
d764b54e
KW
3708
3709 const U8* const typestr = (U8*)SvPV_nolen(*typesvp);
3710 const STRLEN bits = SvUV(*bitssvp);
3711 const STRLEN octets = bits >> 3; /* if bits == 1, then octets == 0 */
d73c39c5
KW
3712 U8 *x, *xend;
3713 STRLEN xcur;
d764b54e 3714
a25abddc 3715 SV* invlist;
d764b54e
KW
3716
3717 PERL_ARGS_ASSERT__SWASH_TO_INVLIST;
3718
3719 /* read $swash->{LIST} */
b443038a
KW
3720 if (SvPOK(*listsvp)) {
3721 l = (U8*)SvPV(*listsvp, lcur);
3722 }
3723 else {
3724 /* LIST legitimately doesn't contain a string during compilation phases
3725 * of Perl itself, before the Unicode tables are generated. In this
3726 * case, just fake things up by creating an empty list */
3727 l = empty;
3728 lcur = 0;
3729 }
d764b54e
KW
3730 loc = (char *) l;
3731 lend = l + lcur;
3732
3733 /* Scan the input to count the number of lines to preallocate array size
3734 * based on worst possible case, which is each line in the input creates 2
3735 * elements in the inversion list: 1) the beginning of a range in the list;
3736 * 2) the beginning of a range not in the list. */
3737 while ((loc = (strchr(loc, '\n'))) != NULL) {
3738 elements += 2;
3739 loc++;
3740 }
3741
3742 /* If the ending is somehow corrupt and isn't a new line, add another
3743 * element for the final range that isn't in the inversion list */
fd05e003
KW
3744 if (! (*lend == '\n'
3745 || (*lend == '\0' && (lcur == 0 || *(lend - 1) == '\n'))))
3746 {
d764b54e
KW
3747 elements++;
3748 }
3749
3750 invlist = _new_invlist(elements);
3751
3752 /* Now go through the input again, adding each range to the list */
3753 while (l < lend) {
3754 UV start, end;
3755 UV val; /* Not used by this function */
3756
3757 l = S_swash_scan_list_line(aTHX_ l, lend, &start, &end, &val,
3758 cBOOL(octets), typestr);
3759
3760 if (l > lend) {
3761 break;
3762 }
3763
9d501133 3764 invlist = _add_range_to_invlist(invlist, start, end);
d764b54e
KW
3765 }
3766
77f9f126
KW
3767 /* Invert if the data says it should be */
3768 if (invert_it_svp && SvUV(*invert_it_svp)) {
45bb2768 3769 _invlist_invert_prop(invlist);
77f9f126
KW
3770 }
3771
b0e3252e 3772 /* This code is copied from swatch_get()
d73c39c5
KW
3773 * read $swash->{EXTRAS} */
3774 x = (U8*)SvPV(*extssvp, xcur);
3775 xend = x + xcur;
3776 while (x < xend) {
3777 STRLEN namelen;
3778 U8 *namestr;
3779 SV** othersvp;
3780 HV* otherhv;
3781 STRLEN otherbits;
3782 SV **otherbitssvp, *other;
3783 U8 *nl;
3784
3785 const U8 opc = *x++;
3786 if (opc == '\n')
3787 continue;
3788
3789 nl = (U8*)memchr(x, '\n', xend - x);
3790
3791 if (opc != '-' && opc != '+' && opc != '!' && opc != '&') {
3792 if (nl) {
3793 x = nl + 1; /* 1 is length of "\n" */
3794 continue;
3795 }
3796 else {
3797 x = xend; /* to EXTRAS' end at which \n is not found */
3798 break;