3 * Copyright (C) 2000, 2001, 2002, 2003, 2004, 2005, 2006,
4 * by Larry Wall and others
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.
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.'
16 * 'Well do I understand your speech,' he answered in the same language;
17 * 'yet few strangers do so. Why then do you not speak in the Common Tongue,
18 * as is the custom in the West, if you wish to be answered?'
20 * ...the travellers perceived that the floor was paved with stones of many
21 * hues; branching runes and strange devices intertwined beneath their feet.
25 #define PERL_IN_UTF8_C
28 static const char unees[] =
29 "Malformed UTF-8 character (unexpected end of string)";
32 =head1 Unicode Support
34 This file contains various utility functions for manipulating UTF8-encoded
35 strings. For the uninitiated, this is a method of representing arbitrary
36 Unicode characters as a variable number of bytes, in such a way that
37 characters in the ASCII range are unmodified, and a zero byte never appears
38 within non-zero characters.
40 =for apidoc A|U8 *|uvuni_to_utf8_flags|U8 *d|UV uv|UV flags
42 Adds the UTF-8 representation of the Unicode codepoint C<uv> to the end
43 of the string C<d>; C<d> should be have at least C<UTF8_MAXBYTES+1> free
44 bytes available. The return value is the pointer to the byte after the
45 end of the new character. In other words,
47 d = uvuni_to_utf8_flags(d, uv, flags);
51 d = uvuni_to_utf8(d, uv);
53 (which is equivalent to)
55 d = uvuni_to_utf8_flags(d, uv, 0);
57 is the recommended Unicode-aware way of saying
65 Perl_uvuni_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags)
67 if (ckWARN(WARN_UTF8)) {
68 if (UNICODE_IS_SURROGATE(uv) &&
69 !(flags & UNICODE_ALLOW_SURROGATE))
70 Perl_warner(aTHX_ packWARN(WARN_UTF8), "UTF-16 surrogate 0x%04"UVxf, uv);
72 ((uv >= 0xFDD0 && uv <= 0xFDEF &&
73 !(flags & UNICODE_ALLOW_FDD0))
75 ((uv & 0xFFFE) == 0xFFFE && /* Either FFFE or FFFF. */
76 !(flags & UNICODE_ALLOW_FFFF))) &&
77 /* UNICODE_ALLOW_SUPER includes
78 * FFFEs and FFFFs beyond 0x10FFFF. */
79 ((uv <= PERL_UNICODE_MAX) ||
80 !(flags & UNICODE_ALLOW_SUPER))
82 Perl_warner(aTHX_ packWARN(WARN_UTF8),
83 "Unicode character 0x%04"UVxf" is illegal", uv);
85 if (UNI_IS_INVARIANT(uv)) {
86 *d++ = (U8)UTF_TO_NATIVE(uv);
91 STRLEN len = UNISKIP(uv);
94 *p-- = (U8)UTF_TO_NATIVE((uv & UTF_CONTINUATION_MASK) | UTF_CONTINUATION_MARK);
95 uv >>= UTF_ACCUMULATION_SHIFT;
97 *p = (U8)UTF_TO_NATIVE((uv & UTF_START_MASK(len)) | UTF_START_MARK(len));
100 #else /* Non loop style */
102 *d++ = (U8)(( uv >> 6) | 0xc0);
103 *d++ = (U8)(( uv & 0x3f) | 0x80);
107 *d++ = (U8)(( uv >> 12) | 0xe0);
108 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
109 *d++ = (U8)(( uv & 0x3f) | 0x80);
113 *d++ = (U8)(( uv >> 18) | 0xf0);
114 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
115 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
116 *d++ = (U8)(( uv & 0x3f) | 0x80);
119 if (uv < 0x4000000) {
120 *d++ = (U8)(( uv >> 24) | 0xf8);
121 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
122 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
123 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
124 *d++ = (U8)(( uv & 0x3f) | 0x80);
127 if (uv < 0x80000000) {
128 *d++ = (U8)(( uv >> 30) | 0xfc);
129 *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80);
130 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
131 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
132 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
133 *d++ = (U8)(( uv & 0x3f) | 0x80);
137 if (uv < UTF8_QUAD_MAX)
140 *d++ = 0xfe; /* Can't match U+FEFF! */
141 *d++ = (U8)(((uv >> 30) & 0x3f) | 0x80);
142 *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80);
143 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
144 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
145 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
146 *d++ = (U8)(( uv & 0x3f) | 0x80);
151 *d++ = 0xff; /* Can't match U+FFFE! */
152 *d++ = 0x80; /* 6 Reserved bits */
153 *d++ = (U8)(((uv >> 60) & 0x0f) | 0x80); /* 2 Reserved bits */
154 *d++ = (U8)(((uv >> 54) & 0x3f) | 0x80);
155 *d++ = (U8)(((uv >> 48) & 0x3f) | 0x80);
156 *d++ = (U8)(((uv >> 42) & 0x3f) | 0x80);
157 *d++ = (U8)(((uv >> 36) & 0x3f) | 0x80);
158 *d++ = (U8)(((uv >> 30) & 0x3f) | 0x80);
159 *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80);
160 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
161 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
162 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
163 *d++ = (U8)(( uv & 0x3f) | 0x80);
167 #endif /* Loop style */
172 Tests if some arbitrary number of bytes begins in a valid UTF-8
173 character. Note that an INVARIANT (i.e. ASCII) character is a valid
174 UTF-8 character. The actual number of bytes in the UTF-8 character
175 will be returned if it is valid, otherwise 0.
177 This is the "slow" version as opposed to the "fast" version which is
178 the "unrolled" IS_UTF8_CHAR(). E.g. for t/uni/class.t the speed
179 difference is a factor of 2 to 3. For lengths (UTF8SKIP(s)) of four
180 or less you should use the IS_UTF8_CHAR(), for lengths of five or more
181 you should use the _slow(). In practice this means that the _slow()
182 will be used very rarely, since the maximum Unicode code point (as of
183 Unicode 4.1) is U+10FFFF, which encodes in UTF-8 to four bytes. Only
184 the "Perl extended UTF-8" (the infamous 'v-strings') will encode into
189 S_is_utf8_char_slow(pTHX_ const U8 *s, const STRLEN len)
195 if (UTF8_IS_INVARIANT(u))
198 if (!UTF8_IS_START(u))
201 if (len < 2 || !UTF8_IS_CONTINUATION(s[1]))
207 u = NATIVE_TO_UTF(u);
209 u &= UTF_START_MASK(len);
213 if (!UTF8_IS_CONTINUATION(*s))
215 uv = UTF8_ACCUMULATE(uv, *s);
222 if ((STRLEN)UNISKIP(uv) < len)
229 =for apidoc A|STRLEN|is_utf8_char|const U8 *s
231 Tests if some arbitrary number of bytes begins in a valid UTF-8
232 character. Note that an INVARIANT (i.e. ASCII) character is a valid
233 UTF-8 character. The actual number of bytes in the UTF-8 character
234 will be returned if it is valid, otherwise 0.
238 Perl_is_utf8_char(pTHX_ const U8 *s)
240 const STRLEN len = UTF8SKIP(s);
242 if (IS_UTF8_CHAR_FAST(len))
243 return IS_UTF8_CHAR(s, len) ? len : 0;
244 #endif /* #ifdef IS_UTF8_CHAR */
245 return is_utf8_char_slow(s, len);
249 =for apidoc A|bool|is_utf8_string|const U8 *s|STRLEN len
251 Returns true if first C<len> bytes of the given string form a valid
252 UTF-8 string, false otherwise. Note that 'a valid UTF-8 string' does
253 not mean 'a string that contains code points above 0x7F encoded in UTF-8'
254 because a valid ASCII string is a valid UTF-8 string.
256 See also is_utf8_string_loclen() and is_utf8_string_loc().
262 Perl_is_utf8_string(pTHX_ const U8 *s, STRLEN len)
268 len = strlen((const char *)s);
273 /* Inline the easy bits of is_utf8_char() here for speed... */
274 if (UTF8_IS_INVARIANT(*x))
276 else if (!UTF8_IS_START(*x))
279 /* ... and call is_utf8_char() only if really needed. */
282 if (IS_UTF8_CHAR_FAST(c)) {
283 if (!IS_UTF8_CHAR(x, c))
285 } else if (!is_utf8_char_slow(x, c))
289 #endif /* #ifdef IS_UTF8_CHAR */
304 Implemented as a macro in utf8.h
306 =for apidoc A|bool|is_utf8_string_loc|const U8 *s|STRLEN len|const U8 **ep
308 Like is_utf8_string() but stores the location of the failure (in the
309 case of "utf8ness failure") or the location s+len (in the case of
310 "utf8ness success") in the C<ep>.
312 See also is_utf8_string_loclen() and is_utf8_string().
314 =for apidoc A|bool|is_utf8_string_loclen|const U8 *s|STRLEN len|const U8 **ep|const STRLEN *el
316 Like is_utf8_string() but stores the location of the failure (in the
317 case of "utf8ness failure") or the location s+len (in the case of
318 "utf8ness success") in the C<ep>, and the number of UTF-8
319 encoded characters in the C<el>.
321 See also is_utf8_string_loc() and is_utf8_string().
327 Perl_is_utf8_string_loclen(pTHX_ const U8 *s, STRLEN len, const U8 **ep, STRLEN *el)
334 len = strlen((const char *)s);
340 /* Inline the easy bits of is_utf8_char() here for speed... */
341 if (UTF8_IS_INVARIANT(*x))
343 else if (!UTF8_IS_START(*x))
346 /* ... and call is_utf8_char() only if really needed. */
349 if (IS_UTF8_CHAR_FAST(c)) {
350 if (!IS_UTF8_CHAR(x, c))
353 c = is_utf8_char_slow(x, c);
356 #endif /* #ifdef IS_UTF8_CHAR */
376 =for apidoc A|UV|utf8n_to_uvuni|const U8 *s|STRLEN curlen|STRLEN *retlen|U32 flags
378 Bottom level UTF-8 decode routine.
379 Returns the unicode code point value of the first character in the string C<s>
380 which is assumed to be in UTF-8 encoding and no longer than C<curlen>;
381 C<retlen> will be set to the length, in bytes, of that character.
383 If C<s> does not point to a well-formed UTF-8 character, the behaviour
384 is dependent on the value of C<flags>: if it contains UTF8_CHECK_ONLY,
385 it is assumed that the caller will raise a warning, and this function
386 will silently just set C<retlen> to C<-1> and return zero. If the
387 C<flags> does not contain UTF8_CHECK_ONLY, warnings about
388 malformations will be given, C<retlen> will be set to the expected
389 length of the UTF-8 character in bytes, and zero will be returned.
391 The C<flags> can also contain various flags to allow deviations from
392 the strict UTF-8 encoding (see F<utf8.h>).
394 Most code should use utf8_to_uvchr() rather than call this directly.
400 Perl_utf8n_to_uvuni(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen, U32 flags)
406 const bool dowarn = ckWARN_d(WARN_UTF8);
407 const UV startbyte = *s;
408 STRLEN expectlen = 0;
411 /* This list is a superset of the UTF8_ALLOW_XXX. */
413 #define UTF8_WARN_EMPTY 1
414 #define UTF8_WARN_CONTINUATION 2
415 #define UTF8_WARN_NON_CONTINUATION 3
416 #define UTF8_WARN_FE_FF 4
417 #define UTF8_WARN_SHORT 5
418 #define UTF8_WARN_OVERFLOW 6
419 #define UTF8_WARN_SURROGATE 7
420 #define UTF8_WARN_LONG 8
421 #define UTF8_WARN_FFFF 9 /* Also FFFE. */
424 !(flags & UTF8_ALLOW_EMPTY)) {
425 warning = UTF8_WARN_EMPTY;
429 if (UTF8_IS_INVARIANT(uv)) {
432 return (UV) (NATIVE_TO_UTF(*s));
435 if (UTF8_IS_CONTINUATION(uv) &&
436 !(flags & UTF8_ALLOW_CONTINUATION)) {
437 warning = UTF8_WARN_CONTINUATION;
441 if (UTF8_IS_START(uv) && curlen > 1 && !UTF8_IS_CONTINUATION(s[1]) &&
442 !(flags & UTF8_ALLOW_NON_CONTINUATION)) {
443 warning = UTF8_WARN_NON_CONTINUATION;
448 uv = NATIVE_TO_UTF(uv);
450 if ((uv == 0xfe || uv == 0xff) &&
451 !(flags & UTF8_ALLOW_FE_FF)) {
452 warning = UTF8_WARN_FE_FF;
457 if (!(uv & 0x20)) { len = 2; uv &= 0x1f; }
458 else if (!(uv & 0x10)) { len = 3; uv &= 0x0f; }
459 else if (!(uv & 0x08)) { len = 4; uv &= 0x07; }
460 else if (!(uv & 0x04)) { len = 5; uv &= 0x03; }
462 else if (!(uv & 0x02)) { len = 6; uv &= 0x01; }
463 else { len = 7; uv &= 0x01; }
465 else if (!(uv & 0x02)) { len = 6; uv &= 0x01; }
466 else if (!(uv & 0x01)) { len = 7; uv = 0; }
467 else { len = 13; uv = 0; } /* whoa! */
475 if ((curlen < expectlen) &&
476 !(flags & UTF8_ALLOW_SHORT)) {
477 warning = UTF8_WARN_SHORT;
486 if (!UTF8_IS_CONTINUATION(*s) &&
487 !(flags & UTF8_ALLOW_NON_CONTINUATION)) {
489 warning = UTF8_WARN_NON_CONTINUATION;
493 uv = UTF8_ACCUMULATE(uv, *s);
495 /* These cannot be allowed. */
497 if (expectlen != 13 && !(flags & UTF8_ALLOW_LONG)) {
498 warning = UTF8_WARN_LONG;
502 else { /* uv < ouv */
503 /* This cannot be allowed. */
504 warning = UTF8_WARN_OVERFLOW;
512 if (UNICODE_IS_SURROGATE(uv) &&
513 !(flags & UTF8_ALLOW_SURROGATE)) {
514 warning = UTF8_WARN_SURROGATE;
516 } else if ((expectlen > (STRLEN)UNISKIP(uv)) &&
517 !(flags & UTF8_ALLOW_LONG)) {
518 warning = UTF8_WARN_LONG;
520 } else if (UNICODE_IS_ILLEGAL(uv) &&
521 !(flags & UTF8_ALLOW_FFFF)) {
522 warning = UTF8_WARN_FFFF;
530 if (flags & UTF8_CHECK_ONLY) {
537 SV* const sv = sv_2mortal(newSVpvs("Malformed UTF-8 character "));
540 case 0: /* Intentionally empty. */ break;
541 case UTF8_WARN_EMPTY:
542 sv_catpvs(sv, "(empty string)");
544 case UTF8_WARN_CONTINUATION:
545 Perl_sv_catpvf(aTHX_ sv, "(unexpected continuation byte 0x%02"UVxf", with no preceding start byte)", uv);
547 case UTF8_WARN_NON_CONTINUATION:
549 Perl_sv_catpvf(aTHX_ sv, "(unexpected non-continuation byte 0x%02"UVxf", immediately after start byte 0x%02"UVxf")",
550 (UV)s[1], startbyte);
552 const int len = (int)(s-s0);
553 Perl_sv_catpvf(aTHX_ sv, "(unexpected non-continuation byte 0x%02"UVxf", %d byte%s after start byte 0x%02"UVxf", expected %d bytes)",
554 (UV)s[1], len, len > 1 ? "s" : "", startbyte, (int)expectlen);
558 case UTF8_WARN_FE_FF:
559 Perl_sv_catpvf(aTHX_ sv, "(byte 0x%02"UVxf")", uv);
561 case UTF8_WARN_SHORT:
562 Perl_sv_catpvf(aTHX_ sv, "(%d byte%s, need %d, after start byte 0x%02"UVxf")",
563 (int)curlen, curlen == 1 ? "" : "s", (int)expectlen, startbyte);
564 expectlen = curlen; /* distance for caller to skip */
566 case UTF8_WARN_OVERFLOW:
567 Perl_sv_catpvf(aTHX_ sv, "(overflow at 0x%"UVxf", byte 0x%02x, after start byte 0x%02"UVxf")",
570 case UTF8_WARN_SURROGATE:
571 Perl_sv_catpvf(aTHX_ sv, "(UTF-16 surrogate 0x%04"UVxf")", uv);
574 Perl_sv_catpvf(aTHX_ sv, "(%d byte%s, need %d, after start byte 0x%02"UVxf")",
575 (int)expectlen, expectlen == 1 ? "": "s", UNISKIP(uv), startbyte);
578 Perl_sv_catpvf(aTHX_ sv, "(character 0x%04"UVxf")", uv);
581 sv_catpvs(sv, "(unknown reason)");
586 const char * const s = SvPVX_const(sv);
589 Perl_warner(aTHX_ packWARN(WARN_UTF8),
590 "%s in %s", s, OP_DESC(PL_op));
592 Perl_warner(aTHX_ packWARN(WARN_UTF8), "%s", s);
597 *retlen = expectlen ? expectlen : len;
603 =for apidoc A|UV|utf8_to_uvchr|const U8 *s|STRLEN *retlen
605 Returns the native character value of the first character in the string C<s>
606 which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
607 length, in bytes, of that character.
609 If C<s> does not point to a well-formed UTF-8 character, zero is
610 returned and retlen is set, if possible, to -1.
616 Perl_utf8_to_uvchr(pTHX_ const U8 *s, STRLEN *retlen)
618 return utf8n_to_uvchr(s, UTF8_MAXBYTES, retlen,
619 ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
623 =for apidoc A|UV|utf8_to_uvuni|const U8 *s|STRLEN *retlen
625 Returns the Unicode code point of the first character in the string C<s>
626 which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
627 length, in bytes, of that character.
629 This function should only be used when returned UV is considered
630 an index into the Unicode semantic tables (e.g. swashes).
632 If C<s> does not point to a well-formed UTF-8 character, zero is
633 returned and retlen is set, if possible, to -1.
639 Perl_utf8_to_uvuni(pTHX_ const U8 *s, STRLEN *retlen)
641 /* Call the low level routine asking for checks */
642 return Perl_utf8n_to_uvuni(aTHX_ s, UTF8_MAXBYTES, retlen,
643 ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
647 =for apidoc A|STRLEN|utf8_length|const U8 *s|const U8 *e
649 Return the length of the UTF-8 char encoded string C<s> in characters.
650 Stops at C<e> (inclusive). If C<e E<lt> s> or if the scan would end
651 up past C<e>, croaks.
657 Perl_utf8_length(pTHX_ const U8 *s, const U8 *e)
662 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g.
663 * the bitops (especially ~) can create illegal UTF-8.
664 * In other words: in Perl UTF-8 is not just for Unicode. */
667 goto warn_and_return;
669 const U8 t = UTF8SKIP(s);
672 if (ckWARN_d(WARN_UTF8)) {
674 Perl_warner(aTHX_ packWARN(WARN_UTF8),
675 "%s in %s", unees, OP_DESC(PL_op));
677 Perl_warner(aTHX_ packWARN(WARN_UTF8), unees);
689 =for apidoc A|IV|utf8_distance|const U8 *a|const U8 *b
691 Returns the number of UTF-8 characters between the UTF-8 pointers C<a>
694 WARNING: use only if you *know* that the pointers point inside the
701 Perl_utf8_distance(pTHX_ const U8 *a, const U8 *b)
706 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g.
707 * the bitops (especially ~) can create illegal UTF-8.
708 * In other words: in Perl UTF-8 is not just for Unicode. */
712 const U8 c = UTF8SKIP(a);
714 goto warn_and_return;
721 const U8 c = UTF8SKIP(b);
725 if (ckWARN_d(WARN_UTF8)) {
727 Perl_warner(aTHX_ packWARN(WARN_UTF8),
728 "%s in %s", unees, OP_DESC(PL_op));
730 Perl_warner(aTHX_ packWARN(WARN_UTF8), unees);
743 =for apidoc A|U8 *|utf8_hop|U8 *s|I32 off
745 Return the UTF-8 pointer C<s> displaced by C<off> characters, either
748 WARNING: do not use the following unless you *know* C<off> is within
749 the UTF-8 data pointed to by C<s> *and* that on entry C<s> is aligned
750 on the first byte of character or just after the last byte of a character.
756 Perl_utf8_hop(pTHX_ const U8 *s, I32 off)
758 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g
759 * the bitops (especially ~) can create illegal UTF-8.
760 * In other words: in Perl UTF-8 is not just for Unicode. */
769 while (UTF8_IS_CONTINUATION(*s))
777 =for apidoc A|U8 *|utf8_to_bytes|U8 *s|STRLEN *len
779 Converts a string C<s> of length C<len> from UTF-8 into byte encoding.
780 Unlike C<bytes_to_utf8>, this over-writes the original string, and
781 updates len to contain the new length.
782 Returns zero on failure, setting C<len> to -1.
788 Perl_utf8_to_bytes(pTHX_ U8 *s, STRLEN *len)
794 /* ensure valid UTF-8 and chars < 256 before updating string */
795 for (send = s + *len; s < send; ) {
798 if (!UTF8_IS_INVARIANT(c) &&
799 (!UTF8_IS_DOWNGRADEABLE_START(c) || (s >= send)
800 || !(c = *s++) || !UTF8_IS_CONTINUATION(c))) {
809 *d++ = (U8)utf8_to_uvchr(s, &ulen);
818 =for apidoc A|U8 *|bytes_from_utf8|const U8 *s|STRLEN *len|bool *is_utf8
820 Converts a string C<s> of length C<len> from UTF-8 into byte encoding.
821 Unlike C<utf8_to_bytes> but like C<bytes_to_utf8>, returns a pointer to
822 the newly-created string, and updates C<len> to contain the new
823 length. Returns the original string if no conversion occurs, C<len>
824 is unchanged. Do nothing if C<is_utf8> points to 0. Sets C<is_utf8> to
825 0 if C<s> is converted or contains all 7bit characters.
831 Perl_bytes_from_utf8(pTHX_ const U8 *s, STRLEN *len, bool *is_utf8)
841 /* ensure valid UTF-8 and chars < 256 before converting string */
842 for (send = s + *len; s < send;) {
844 if (!UTF8_IS_INVARIANT(c)) {
845 if (UTF8_IS_DOWNGRADEABLE_START(c) && s < send &&
846 (c = *s++) && UTF8_IS_CONTINUATION(c))
855 Newxz(d, (*len) - count + 1, U8);
856 s = start; start = d;
859 if (!UTF8_IS_INVARIANT(c)) {
860 /* Then it is two-byte encoded */
861 c = UTF8_ACCUMULATE(NATIVE_TO_UTF(c), *s++);
862 c = ASCII_TO_NATIVE(c);
872 =for apidoc A|U8 *|bytes_to_utf8|const U8 *s|STRLEN *len
874 Converts a string C<s> of length C<len> from ASCII into UTF-8 encoding.
875 Returns a pointer to the newly-created string, and sets C<len> to
876 reflect the new length.
878 If you want to convert to UTF-8 from other encodings than ASCII,
879 see sv_recode_to_utf8().
885 Perl_bytes_to_utf8(pTHX_ const U8 *s, STRLEN *len)
887 const U8 * const send = s + (*len);
891 Newxz(d, (*len) * 2 + 1, U8);
895 const UV uv = NATIVE_TO_ASCII(*s++);
896 if (UNI_IS_INVARIANT(uv))
897 *d++ = (U8)UTF_TO_NATIVE(uv);
899 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
900 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
909 * Convert native (big-endian) or reversed (little-endian) UTF-16 to UTF-8.
911 * Destination must be pre-extended to 3/2 source. Do not use in-place.
912 * We optimize for native, for obvious reasons. */
915 Perl_utf16_to_utf8(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen)
920 if (bytelen == 1 && p[0] == 0) { /* Be understanding. */
927 Perl_croak(aTHX_ "panic: utf16_to_utf8: odd bytelen %"UVf, (UV)bytelen);
932 UV uv = (p[0] << 8) + p[1]; /* UTF-16BE */
939 *d++ = (U8)(( uv >> 6) | 0xc0);
940 *d++ = (U8)(( uv & 0x3f) | 0x80);
943 if (uv >= 0xd800 && uv < 0xdbff) { /* surrogates */
944 UV low = (p[0] << 8) + p[1];
946 if (low < 0xdc00 || low >= 0xdfff)
947 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
948 uv = ((uv - 0xd800) << 10) + (low - 0xdc00) + 0x10000;
951 *d++ = (U8)(( uv >> 12) | 0xe0);
952 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
953 *d++ = (U8)(( uv & 0x3f) | 0x80);
957 *d++ = (U8)(( uv >> 18) | 0xf0);
958 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
959 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
960 *d++ = (U8)(( uv & 0x3f) | 0x80);
964 *newlen = d - dstart;
968 /* Note: this one is slightly destructive of the source. */
971 Perl_utf16_to_utf8_reversed(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen)
974 U8* send = s + bytelen;
981 return utf16_to_utf8(p, d, bytelen, newlen);
984 /* for now these are all defined (inefficiently) in terms of the utf8 versions */
987 Perl_is_uni_alnum(pTHX_ UV c)
989 U8 tmpbuf[UTF8_MAXBYTES+1];
990 uvchr_to_utf8(tmpbuf, c);
991 return is_utf8_alnum(tmpbuf);
995 Perl_is_uni_alnumc(pTHX_ UV c)
997 U8 tmpbuf[UTF8_MAXBYTES+1];
998 uvchr_to_utf8(tmpbuf, c);
999 return is_utf8_alnumc(tmpbuf);
1003 Perl_is_uni_idfirst(pTHX_ UV c)
1005 U8 tmpbuf[UTF8_MAXBYTES+1];
1006 uvchr_to_utf8(tmpbuf, c);
1007 return is_utf8_idfirst(tmpbuf);
1011 Perl_is_uni_alpha(pTHX_ UV c)
1013 U8 tmpbuf[UTF8_MAXBYTES+1];
1014 uvchr_to_utf8(tmpbuf, c);
1015 return is_utf8_alpha(tmpbuf);
1019 Perl_is_uni_ascii(pTHX_ UV c)
1021 U8 tmpbuf[UTF8_MAXBYTES+1];
1022 uvchr_to_utf8(tmpbuf, c);
1023 return is_utf8_ascii(tmpbuf);
1027 Perl_is_uni_space(pTHX_ UV c)
1029 U8 tmpbuf[UTF8_MAXBYTES+1];
1030 uvchr_to_utf8(tmpbuf, c);
1031 return is_utf8_space(tmpbuf);
1035 Perl_is_uni_digit(pTHX_ UV c)
1037 U8 tmpbuf[UTF8_MAXBYTES+1];
1038 uvchr_to_utf8(tmpbuf, c);
1039 return is_utf8_digit(tmpbuf);
1043 Perl_is_uni_upper(pTHX_ UV c)
1045 U8 tmpbuf[UTF8_MAXBYTES+1];
1046 uvchr_to_utf8(tmpbuf, c);
1047 return is_utf8_upper(tmpbuf);
1051 Perl_is_uni_lower(pTHX_ UV c)
1053 U8 tmpbuf[UTF8_MAXBYTES+1];
1054 uvchr_to_utf8(tmpbuf, c);
1055 return is_utf8_lower(tmpbuf);
1059 Perl_is_uni_cntrl(pTHX_ UV c)
1061 U8 tmpbuf[UTF8_MAXBYTES+1];
1062 uvchr_to_utf8(tmpbuf, c);
1063 return is_utf8_cntrl(tmpbuf);
1067 Perl_is_uni_graph(pTHX_ UV c)
1069 U8 tmpbuf[UTF8_MAXBYTES+1];
1070 uvchr_to_utf8(tmpbuf, c);
1071 return is_utf8_graph(tmpbuf);
1075 Perl_is_uni_print(pTHX_ UV c)
1077 U8 tmpbuf[UTF8_MAXBYTES+1];
1078 uvchr_to_utf8(tmpbuf, c);
1079 return is_utf8_print(tmpbuf);
1083 Perl_is_uni_punct(pTHX_ UV c)
1085 U8 tmpbuf[UTF8_MAXBYTES+1];
1086 uvchr_to_utf8(tmpbuf, c);
1087 return is_utf8_punct(tmpbuf);
1091 Perl_is_uni_xdigit(pTHX_ UV c)
1093 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1094 uvchr_to_utf8(tmpbuf, c);
1095 return is_utf8_xdigit(tmpbuf);
1099 Perl_to_uni_upper(pTHX_ UV c, U8* p, STRLEN *lenp)
1101 uvchr_to_utf8(p, c);
1102 return to_utf8_upper(p, p, lenp);
1106 Perl_to_uni_title(pTHX_ UV c, U8* p, STRLEN *lenp)
1108 uvchr_to_utf8(p, c);
1109 return to_utf8_title(p, p, lenp);
1113 Perl_to_uni_lower(pTHX_ UV c, U8* p, STRLEN *lenp)
1115 uvchr_to_utf8(p, c);
1116 return to_utf8_lower(p, p, lenp);
1120 Perl_to_uni_fold(pTHX_ UV c, U8* p, STRLEN *lenp)
1122 uvchr_to_utf8(p, c);
1123 return to_utf8_fold(p, p, lenp);
1126 /* for now these all assume no locale info available for Unicode > 255 */
1129 Perl_is_uni_alnum_lc(pTHX_ UV c)
1131 return is_uni_alnum(c); /* XXX no locale support yet */
1135 Perl_is_uni_alnumc_lc(pTHX_ UV c)
1137 return is_uni_alnumc(c); /* XXX no locale support yet */
1141 Perl_is_uni_idfirst_lc(pTHX_ UV c)
1143 return is_uni_idfirst(c); /* XXX no locale support yet */
1147 Perl_is_uni_alpha_lc(pTHX_ UV c)
1149 return is_uni_alpha(c); /* XXX no locale support yet */
1153 Perl_is_uni_ascii_lc(pTHX_ UV c)
1155 return is_uni_ascii(c); /* XXX no locale support yet */
1159 Perl_is_uni_space_lc(pTHX_ UV c)
1161 return is_uni_space(c); /* XXX no locale support yet */
1165 Perl_is_uni_digit_lc(pTHX_ UV c)
1167 return is_uni_digit(c); /* XXX no locale support yet */
1171 Perl_is_uni_upper_lc(pTHX_ UV c)
1173 return is_uni_upper(c); /* XXX no locale support yet */
1177 Perl_is_uni_lower_lc(pTHX_ UV c)
1179 return is_uni_lower(c); /* XXX no locale support yet */
1183 Perl_is_uni_cntrl_lc(pTHX_ UV c)
1185 return is_uni_cntrl(c); /* XXX no locale support yet */
1189 Perl_is_uni_graph_lc(pTHX_ UV c)
1191 return is_uni_graph(c); /* XXX no locale support yet */
1195 Perl_is_uni_print_lc(pTHX_ UV c)
1197 return is_uni_print(c); /* XXX no locale support yet */
1201 Perl_is_uni_punct_lc(pTHX_ UV c)
1203 return is_uni_punct(c); /* XXX no locale support yet */
1207 Perl_is_uni_xdigit_lc(pTHX_ UV c)
1209 return is_uni_xdigit(c); /* XXX no locale support yet */
1213 Perl_to_uni_upper_lc(pTHX_ U32 c)
1215 /* XXX returns only the first character -- do not use XXX */
1216 /* XXX no locale support yet */
1218 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1219 return (U32)to_uni_upper(c, tmpbuf, &len);
1223 Perl_to_uni_title_lc(pTHX_ U32 c)
1225 /* XXX returns only the first character XXX -- do not use XXX */
1226 /* XXX no locale support yet */
1228 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1229 return (U32)to_uni_title(c, tmpbuf, &len);
1233 Perl_to_uni_lower_lc(pTHX_ U32 c)
1235 /* XXX returns only the first character -- do not use XXX */
1236 /* XXX no locale support yet */
1238 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1239 return (U32)to_uni_lower(c, tmpbuf, &len);
1243 S_is_utf8_common(pTHX_ const U8 *const p, SV **swash,
1244 const char *const swashname)
1247 if (!is_utf8_char(p))
1250 *swash = swash_init("utf8", swashname, &PL_sv_undef, 1, 0);
1251 return swash_fetch(*swash, p, TRUE) != 0;
1255 Perl_is_utf8_alnum(pTHX_ const U8 *p)
1258 /* NOTE: "IsWord", not "IsAlnum", since Alnum is a true
1259 * descendant of isalnum(3), in other words, it doesn't
1260 * contain the '_'. --jhi */
1261 return S_is_utf8_common(aTHX_ p, &PL_utf8_alnum, "IsWord");
1265 Perl_is_utf8_alnumc(pTHX_ const U8 *p)
1268 return S_is_utf8_common(aTHX_ p, &PL_utf8_alnumc, "IsAlnumC");
1272 Perl_is_utf8_idfirst(pTHX_ const U8 *p) /* The naming is historical. */
1277 /* is_utf8_idstart would be more logical. */
1278 return S_is_utf8_common(aTHX_ p, &PL_utf8_idstart, "IdStart");
1282 Perl_is_utf8_idcont(pTHX_ const U8 *p)
1287 return S_is_utf8_common(aTHX_ p, &PL_utf8_idcont, "IdContinue");
1291 Perl_is_utf8_alpha(pTHX_ const U8 *p)
1294 return S_is_utf8_common(aTHX_ p, &PL_utf8_alpha, "IsAlpha");
1298 Perl_is_utf8_ascii(pTHX_ const U8 *p)
1301 return S_is_utf8_common(aTHX_ p, &PL_utf8_ascii, "IsAscii");
1305 Perl_is_utf8_space(pTHX_ const U8 *p)
1308 return S_is_utf8_common(aTHX_ p, &PL_utf8_space, "IsSpacePerl");
1312 Perl_is_utf8_digit(pTHX_ const U8 *p)
1315 return S_is_utf8_common(aTHX_ p, &PL_utf8_digit, "IsDigit");
1319 Perl_is_utf8_upper(pTHX_ const U8 *p)
1322 return S_is_utf8_common(aTHX_ p, &PL_utf8_upper, "IsUppercase");
1326 Perl_is_utf8_lower(pTHX_ const U8 *p)
1329 return S_is_utf8_common(aTHX_ p, &PL_utf8_lower, "IsLowercase");
1333 Perl_is_utf8_cntrl(pTHX_ const U8 *p)
1336 return S_is_utf8_common(aTHX_ p, &PL_utf8_cntrl, "IsCntrl");
1340 Perl_is_utf8_graph(pTHX_ const U8 *p)
1343 return S_is_utf8_common(aTHX_ p, &PL_utf8_graph, "IsGraph");
1347 Perl_is_utf8_print(pTHX_ const U8 *p)
1350 return S_is_utf8_common(aTHX_ p, &PL_utf8_print, "IsPrint");
1354 Perl_is_utf8_punct(pTHX_ const U8 *p)
1357 return S_is_utf8_common(aTHX_ p, &PL_utf8_punct, "IsPunct");
1361 Perl_is_utf8_xdigit(pTHX_ const U8 *p)
1364 return S_is_utf8_common(aTHX_ p, &PL_utf8_xdigit, "Isxdigit");
1368 Perl_is_utf8_mark(pTHX_ const U8 *p)
1371 return S_is_utf8_common(aTHX_ p, &PL_utf8_mark, "IsM");
1375 =for apidoc A|UV|to_utf8_case|U8 *p|U8* ustrp|STRLEN *lenp|SV **swash|char *normal|char *special
1377 The "p" contains the pointer to the UTF-8 string encoding
1378 the character that is being converted.
1380 The "ustrp" is a pointer to the character buffer to put the
1381 conversion result to. The "lenp" is a pointer to the length
1384 The "swashp" is a pointer to the swash to use.
1386 Both the special and normal mappings are stored lib/unicore/To/Foo.pl,
1387 and loaded by SWASHNEW, using lib/utf8_heavy.pl. The special (usually,
1388 but not always, a multicharacter mapping), is tried first.
1390 The "special" is a string like "utf8::ToSpecLower", which means the
1391 hash %utf8::ToSpecLower. The access to the hash is through
1392 Perl_to_utf8_case().
1394 The "normal" is a string like "ToLower" which means the swash
1400 Perl_to_utf8_case(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp,
1401 SV **swashp, const char *normal, const char *special)
1404 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1407 const UV uv0 = utf8_to_uvchr(p, NULL);
1408 /* The NATIVE_TO_UNI() and UNI_TO_NATIVE() mappings
1409 * are necessary in EBCDIC, they are redundant no-ops
1410 * in ASCII-ish platforms, and hopefully optimized away. */
1411 const UV uv1 = NATIVE_TO_UNI(uv0);
1412 uvuni_to_utf8(tmpbuf, uv1);
1414 if (!*swashp) /* load on-demand */
1415 *swashp = swash_init("utf8", normal, &PL_sv_undef, 4, 0);
1417 /* The 0xDF is the only special casing Unicode code point below 0x100. */
1418 if (special && (uv1 == 0xDF || uv1 > 0xFF)) {
1419 /* It might be "special" (sometimes, but not always,
1420 * a multicharacter mapping) */
1424 if ((hv = get_hv(special, FALSE)) &&
1425 (svp = hv_fetch(hv, (const char*)tmpbuf, UNISKIP(uv1), FALSE)) &&
1429 s = SvPV_const(*svp, len);
1431 len = uvuni_to_utf8(ustrp, NATIVE_TO_UNI(*(U8*)s)) - ustrp;
1434 /* If we have EBCDIC we need to remap the characters
1435 * since any characters in the low 256 are Unicode
1436 * code points, not EBCDIC. */
1437 U8 *t = (U8*)s, *tend = t + len, *d;
1444 UV c = utf8_to_uvchr(t, &tlen);
1446 d = uvchr_to_utf8(d, UNI_TO_NATIVE(c));
1455 d = uvchr_to_utf8(d, UNI_TO_NATIVE(*t));
1460 Copy(tmpbuf, ustrp, len, U8);
1462 Copy(s, ustrp, len, U8);
1468 if (!len && *swashp) {
1469 UV uv2 = swash_fetch(*swashp, tmpbuf, TRUE);
1472 /* It was "normal" (a single character mapping). */
1473 UV uv3 = UNI_TO_NATIVE(uv2);
1475 len = uvchr_to_utf8(ustrp, uv3) - ustrp;
1479 if (!len) /* Neither: just copy. */
1480 len = uvchr_to_utf8(ustrp, uv0) - ustrp;
1485 return len ? utf8_to_uvchr(ustrp, 0) : 0;
1489 =for apidoc A|UV|to_utf8_upper|const U8 *p|U8 *ustrp|STRLEN *lenp
1491 Convert the UTF-8 encoded character at p to its uppercase version and
1492 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
1493 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since
1494 the uppercase version may be longer than the original character.
1496 The first character of the uppercased version is returned
1497 (but note, as explained above, that there may be more.)
1502 Perl_to_utf8_upper(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
1505 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
1506 &PL_utf8_toupper, "ToUpper", "utf8::ToSpecUpper");
1510 =for apidoc A|UV|to_utf8_title|const U8 *p|U8 *ustrp|STRLEN *lenp
1512 Convert the UTF-8 encoded character at p to its titlecase version and
1513 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
1514 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
1515 titlecase version may be longer than the original character.
1517 The first character of the titlecased version is returned
1518 (but note, as explained above, that there may be more.)
1523 Perl_to_utf8_title(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
1526 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
1527 &PL_utf8_totitle, "ToTitle", "utf8::ToSpecTitle");
1531 =for apidoc A|UV|to_utf8_lower|const U8 *p|U8 *ustrp|STRLEN *lenp
1533 Convert the UTF-8 encoded character at p to its lowercase version and
1534 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
1535 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
1536 lowercase version may be longer than the original character.
1538 The first character of the lowercased version is returned
1539 (but note, as explained above, that there may be more.)
1544 Perl_to_utf8_lower(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
1547 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
1548 &PL_utf8_tolower, "ToLower", "utf8::ToSpecLower");
1552 =for apidoc A|UV|to_utf8_fold|const U8 *p|U8 *ustrp|STRLEN *lenp
1554 Convert the UTF-8 encoded character at p to its foldcase version and
1555 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
1556 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
1557 foldcase version may be longer than the original character (up to
1560 The first character of the foldcased version is returned
1561 (but note, as explained above, that there may be more.)
1566 Perl_to_utf8_fold(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
1569 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
1570 &PL_utf8_tofold, "ToFold", "utf8::ToSpecFold");
1574 * A "swash" is a swatch hash.
1575 * A "swatch" is a bit vector generated by utf8.c:S_swash_get().
1576 * C<pkg> is a pointer to a package name for SWASHNEW, should be "utf8".
1577 * For other parameters, see utf8::SWASHNEW in lib/utf8_heavy.pl.
1580 Perl_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv, I32 minbits, I32 none)
1584 SV* const tokenbufsv = sv_newmortal();
1586 const size_t pkg_len = strlen(pkg);
1587 const size_t name_len = strlen(name);
1588 HV * const stash = gv_stashpvn(pkg, pkg_len, FALSE);
1591 PUSHSTACKi(PERLSI_MAGIC);
1596 if (!gv_fetchmeth(stash, "SWASHNEW", 8, -1)) { /* demand load utf8 */
1598 errsv_save = newSVsv(ERRSV);
1599 Perl_load_module(aTHX_ PERL_LOADMOD_NOIMPORT, newSVpvn(pkg,pkg_len),
1602 sv_setsv(ERRSV, errsv_save);
1603 SvREFCNT_dec(errsv_save);
1609 PUSHs(sv_2mortal(newSVpvn(pkg, pkg_len)));
1610 PUSHs(sv_2mortal(newSVpvn(name, name_len)));
1612 PUSHs(sv_2mortal(newSViv(minbits)));
1613 PUSHs(sv_2mortal(newSViv(none)));
1615 if (IN_PERL_COMPILETIME) {
1616 /* XXX ought to be handled by lex_start */
1619 sv_setpv(tokenbufsv, PL_tokenbuf);
1621 errsv_save = newSVsv(ERRSV);
1622 if (call_method("SWASHNEW", G_SCALAR))
1623 retval = newSVsv(*PL_stack_sp--);
1625 retval = &PL_sv_undef;
1627 sv_setsv(ERRSV, errsv_save);
1628 SvREFCNT_dec(errsv_save);
1631 if (IN_PERL_COMPILETIME) {
1633 const char* const pv = SvPV_const(tokenbufsv, len);
1635 Copy(pv, PL_tokenbuf, len+1, char);
1636 PL_curcop->op_private = (U8)(PL_hints & HINT_PRIVATE_MASK);
1638 if (!SvROK(retval) || SvTYPE(SvRV(retval)) != SVt_PVHV) {
1640 Perl_croak(aTHX_ "Can't find Unicode property definition \"%"SVf"\"",
1642 Perl_croak(aTHX_ "SWASHNEW didn't return an HV ref");
1648 /* This API is wrong for special case conversions since we may need to
1649 * return several Unicode characters for a single Unicode character
1650 * (see lib/unicore/SpecCase.txt) The SWASHGET in lib/utf8_heavy.pl is
1651 * the lower-level routine, and it is similarly broken for returning
1652 * multiple values. --jhi */
1653 /* Now SWASHGET is recasted into S_swash_get in this file. */
1656 * Returns the value of property/mapping C<swash> for the first character
1657 * of the string C<ptr>. If C<do_utf8> is true, the string C<ptr> is
1658 * assumed to be in utf8. If C<do_utf8> is false, the string C<ptr> is
1659 * assumed to be in native 8-bit encoding. Caches the swatch in C<swash>.
1662 Perl_swash_fetch(pTHX_ SV *swash, const U8 *ptr, bool do_utf8)
1665 HV* const hv = (HV*)SvRV(swash);
1670 const U8 *tmps = NULL;
1674 UV c = NATIVE_TO_ASCII(*ptr);
1676 if (!do_utf8 && !UNI_IS_INVARIANT(c)) {
1677 tmputf8[0] = (U8)UTF8_EIGHT_BIT_HI(c);
1678 tmputf8[1] = (U8)UTF8_EIGHT_BIT_LO(c);
1681 /* Given a UTF-X encoded char 0xAA..0xYY,0xZZ
1682 * then the "swatch" is a vec() for al the chars which start
1684 * So the key in the hash (klen) is length of encoded char -1
1686 klen = UTF8SKIP(ptr) - 1;
1690 /* If char in invariant then swatch is for all the invariant chars
1691 * In both UTF-8 and UTF-8-MOD that happens to be UTF_CONTINUATION_MARK
1693 needents = UTF_CONTINUATION_MARK;
1694 off = NATIVE_TO_UTF(ptr[klen]);
1697 /* If char is encoded then swatch is for the prefix */
1698 needents = (1 << UTF_ACCUMULATION_SHIFT);
1699 off = NATIVE_TO_UTF(ptr[klen]) & UTF_CONTINUATION_MASK;
1703 * This single-entry cache saves about 1/3 of the utf8 overhead in test
1704 * suite. (That is, only 7-8% overall over just a hash cache. Still,
1705 * it's nothing to sniff at.) Pity we usually come through at least
1706 * two function calls to get here...
1708 * NB: this code assumes that swatches are never modified, once generated!
1711 if (hv == PL_last_swash_hv &&
1712 klen == PL_last_swash_klen &&
1713 (!klen || memEQ((char *)ptr, (char *)PL_last_swash_key, klen)) )
1715 tmps = PL_last_swash_tmps;
1716 slen = PL_last_swash_slen;
1719 /* Try our second-level swatch cache, kept in a hash. */
1720 SV** svp = hv_fetch(hv, (const char*)ptr, klen, FALSE);
1722 /* If not cached, generate it via swash_get */
1723 if (!svp || !SvPOK(*svp)
1724 || !(tmps = (const U8*)SvPV_const(*svp, slen))) {
1725 /* We use utf8n_to_uvuni() as we want an index into
1726 Unicode tables, not a native character number.
1728 const UV code_point = utf8n_to_uvuni(ptr, UTF8_MAXBYTES, 0,
1730 0 : UTF8_ALLOW_ANY);
1731 swatch = swash_get(swash,
1732 /* On EBCDIC & ~(0xA0-1) isn't a useful thing to do */
1733 (klen) ? (code_point & ~(needents - 1)) : 0,
1736 if (IN_PERL_COMPILETIME)
1737 PL_curcop->op_private = (U8)(PL_hints & HINT_PRIVATE_MASK);
1739 svp = hv_store(hv, (const char *)ptr, klen, swatch, 0);
1741 if (!svp || !(tmps = (U8*)SvPV(*svp, slen))
1742 || (slen << 3) < needents)
1743 Perl_croak(aTHX_ "panic: swash_fetch got improper swatch");
1746 PL_last_swash_hv = hv;
1747 PL_last_swash_klen = klen;
1748 /* FIXME change interpvar.h? */
1749 PL_last_swash_tmps = (U8 *) tmps;
1750 PL_last_swash_slen = slen;
1752 Copy(ptr, PL_last_swash_key, klen, U8);
1755 switch ((int)((slen << 3) / needents)) {
1757 bit = 1 << (off & 7);
1759 return (tmps[off] & bit) != 0;
1764 return (tmps[off] << 8) + tmps[off + 1] ;
1767 return (tmps[off] << 24) + (tmps[off+1] << 16) + (tmps[off+2] << 8) + tmps[off + 3] ;
1769 Perl_croak(aTHX_ "panic: swash_fetch got swatch of unexpected bit width");
1774 * Returns a swatch (a bit vector string) for a code point sequence
1775 * that starts from the value C<start> and comprises the number C<span>.
1776 * A C<swash> must be an object created by SWASHNEW (see lib/utf8_heavy.pl).
1777 * Should be used via swash_fetch, which will cache the swatch in C<swash>.
1780 S_swash_get(pTHX_ SV* swash, UV start, UV span)
1783 U8 *l, *lend, *x, *xend, *s;
1784 STRLEN lcur, xcur, scur;
1786 HV* const hv = (HV*)SvRV(swash);
1787 SV** const listsvp = hv_fetchs(hv, "LIST", FALSE);
1788 SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE);
1789 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
1790 SV** const nonesvp = hv_fetchs(hv, "NONE", FALSE);
1791 SV** const extssvp = hv_fetchs(hv, "EXTRAS", FALSE);
1792 const U8* const typestr = (U8*)SvPV_nolen(*typesvp);
1793 const int typeto = typestr[0] == 'T' && typestr[1] == 'o';
1794 const STRLEN bits = SvUV(*bitssvp);
1795 const STRLEN octets = bits >> 3; /* if bits == 1, then octets == 0 */
1796 const UV none = SvUV(*nonesvp);
1797 const UV end = start + span;
1799 if (bits != 1 && bits != 8 && bits != 16 && bits != 32) {
1800 Perl_croak(aTHX_ "panic: swash_get doesn't expect bits %"UVuf,
1804 /* create and initialize $swatch */
1805 swatch = newSVpvs("");
1806 scur = octets ? (span * octets) : (span + 7) / 8;
1807 SvGROW(swatch, scur + 1);
1808 s = (U8*)SvPVX(swatch);
1809 if (octets && none) {
1810 const U8* const e = s + scur;
1813 *s++ = (U8)(none & 0xff);
1814 else if (bits == 16) {
1815 *s++ = (U8)((none >> 8) & 0xff);
1816 *s++ = (U8)( none & 0xff);
1818 else if (bits == 32) {
1819 *s++ = (U8)((none >> 24) & 0xff);
1820 *s++ = (U8)((none >> 16) & 0xff);
1821 *s++ = (U8)((none >> 8) & 0xff);
1822 *s++ = (U8)( none & 0xff);
1828 (void)memzero((U8*)s, scur + 1);
1830 SvCUR_set(swatch, scur);
1831 s = (U8*)SvPVX(swatch);
1833 /* read $swash->{LIST} */
1834 l = (U8*)SvPV(*listsvp, lcur);
1837 UV min, max, val, key;
1839 I32 flags = PERL_SCAN_SILENT_ILLDIGIT | PERL_SCAN_DISALLOW_PREFIX;
1841 U8* const nl = (U8*)memchr(l, '\n', lend - l);
1844 min = grok_hex((char *)l, &numlen, &flags, NULL);
1848 l = nl + 1; /* 1 is length of "\n" */
1852 l = lend; /* to LIST's end at which \n is not found */
1858 flags = PERL_SCAN_SILENT_ILLDIGIT | PERL_SCAN_DISALLOW_PREFIX;
1860 max = grok_hex((char *)l, &numlen, &flags, NULL);
1869 flags = PERL_SCAN_SILENT_ILLDIGIT |
1870 PERL_SCAN_DISALLOW_PREFIX;
1872 val = grok_hex((char *)l, &numlen, &flags, NULL);
1881 Perl_croak(aTHX_ "%s: illegal mapping '%s'",
1887 val = 0; /* bits == 1, then val should be ignored */
1894 Perl_croak(aTHX_ "%s: illegal mapping '%s'", typestr, l);
1898 val = 0; /* bits == 1, then val should be ignored */
1911 if (!none || val < none) {
1916 for (key = min; key <= max; key++) {
1920 /* offset must be non-negative (start <= min <= key < end) */
1921 offset = octets * (key - start);
1923 s[offset] = (U8)(val & 0xff);
1924 else if (bits == 16) {
1925 s[offset ] = (U8)((val >> 8) & 0xff);
1926 s[offset + 1] = (U8)( val & 0xff);
1928 else if (bits == 32) {
1929 s[offset ] = (U8)((val >> 24) & 0xff);
1930 s[offset + 1] = (U8)((val >> 16) & 0xff);
1931 s[offset + 2] = (U8)((val >> 8) & 0xff);
1932 s[offset + 3] = (U8)( val & 0xff);
1935 if (!none || val < none)
1939 else { /* bits == 1, then val should be ignored */
1942 for (key = min; key <= max; key++) {
1943 const STRLEN offset = (STRLEN)(key - start);
1946 s[offset >> 3] |= 1 << (offset & 7);
1952 /* read $swash->{EXTRAS} */
1953 x = (U8*)SvPV(*extssvp, xcur);
1961 SV **otherbitssvp, *other;
1969 nl = (U8*)memchr(x, '\n', xend - x);
1971 if (opc != '-' && opc != '+' && opc != '!' && opc != '&') {
1973 x = nl + 1; /* 1 is length of "\n" */
1977 x = xend; /* to EXTRAS' end at which \n is not found */
1984 namelen = nl - namestr;
1988 namelen = xend - namestr;
1992 othersvp = hv_fetch(hv, (char *)namestr, namelen, FALSE);
1993 otherhv = (HV*)SvRV(*othersvp);
1994 otherbitssvp = hv_fetchs(otherhv, "BITS", FALSE);
1995 otherbits = (STRLEN)SvUV(*otherbitssvp);
1996 if (bits < otherbits)
1997 Perl_croak(aTHX_ "panic: swash_get found swatch size mismatch");
1999 /* The "other" swatch must be destroyed after. */
2000 other = swash_get(*othersvp, start, span);
2001 o = (U8*)SvPV(other, olen);
2004 Perl_croak(aTHX_ "panic: swash_get got improper swatch");
2006 s = (U8*)SvPV(swatch, slen);
2007 if (bits == 1 && otherbits == 1) {
2009 Perl_croak(aTHX_ "panic: swash_get found swatch length mismatch");
2033 STRLEN otheroctets = otherbits >> 3;
2035 U8* send = s + slen;
2040 if (otherbits == 1) {
2041 otherval = (o[offset >> 3] >> (offset & 7)) & 1;
2045 STRLEN vlen = otheroctets;
2053 if (opc == '+' && otherval)
2054 ; /* replace with otherval */
2055 else if (opc == '!' && !otherval)
2057 else if (opc == '-' && otherval)
2059 else if (opc == '&' && !otherval)
2062 s += octets; /* no replacement */
2067 *s++ = (U8)( otherval & 0xff);
2068 else if (bits == 16) {
2069 *s++ = (U8)((otherval >> 8) & 0xff);
2070 *s++ = (U8)( otherval & 0xff);
2072 else if (bits == 32) {
2073 *s++ = (U8)((otherval >> 24) & 0xff);
2074 *s++ = (U8)((otherval >> 16) & 0xff);
2075 *s++ = (U8)((otherval >> 8) & 0xff);
2076 *s++ = (U8)( otherval & 0xff);
2080 sv_free(other); /* through with it! */
2086 =for apidoc A|U8 *|uvchr_to_utf8|U8 *d|UV uv
2088 Adds the UTF-8 representation of the Native codepoint C<uv> to the end
2089 of the string C<d>; C<d> should be have at least C<UTF8_MAXBYTES+1> free
2090 bytes available. The return value is the pointer to the byte after the
2091 end of the new character. In other words,
2093 d = uvchr_to_utf8(d, uv);
2095 is the recommended wide native character-aware way of saying
2102 /* On ASCII machines this is normally a macro but we want a
2103 real function in case XS code wants it
2106 Perl_uvchr_to_utf8(pTHX_ U8 *d, UV uv)
2108 return Perl_uvuni_to_utf8_flags(aTHX_ d, NATIVE_TO_UNI(uv), 0);
2112 Perl_uvchr_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags)
2114 return Perl_uvuni_to_utf8_flags(aTHX_ d, NATIVE_TO_UNI(uv), flags);
2118 =for apidoc A|UV|utf8n_to_uvchr|U8 *s|STRLEN curlen|STRLEN *retlen|U32
2121 Returns the native character value of the first character in the string
2123 which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
2124 length, in bytes, of that character.
2126 Allows length and flags to be passed to low level routine.
2130 /* On ASCII machines this is normally a macro but we want
2131 a real function in case XS code wants it
2134 Perl_utf8n_to_uvchr(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen,
2137 const UV uv = Perl_utf8n_to_uvuni(aTHX_ s, curlen, retlen, flags);
2138 return UNI_TO_NATIVE(uv);
2142 =for apidoc A|char *|pv_uni_display|SV *dsv|U8 *spv|STRLEN len|STRLEN pvlim|UV flags
2144 Build to the scalar dsv a displayable version of the string spv,
2145 length len, the displayable version being at most pvlim bytes long
2146 (if longer, the rest is truncated and "..." will be appended).
2148 The flags argument can have UNI_DISPLAY_ISPRINT set to display
2149 isPRINT()able characters as themselves, UNI_DISPLAY_BACKSLASH
2150 to display the \\[nrfta\\] as the backslashed versions (like '\n')
2151 (UNI_DISPLAY_BACKSLASH is preferred over UNI_DISPLAY_ISPRINT for \\).
2152 UNI_DISPLAY_QQ (and its alias UNI_DISPLAY_REGEX) have both
2153 UNI_DISPLAY_BACKSLASH and UNI_DISPLAY_ISPRINT turned on.
2155 The pointer to the PV of the dsv is returned.
2159 Perl_pv_uni_display(pTHX_ SV *dsv, const U8 *spv, STRLEN len, STRLEN pvlim, UV flags)
2164 sv_setpvn(dsv, "", 0);
2165 for (s = (const char *)spv, e = s + len; s < e; s += UTF8SKIP(s)) {
2167 /* This serves double duty as a flag and a character to print after
2168 a \ when flags & UNI_DISPLAY_BACKSLASH is true.
2172 if (pvlim && SvCUR(dsv) >= pvlim) {
2176 u = utf8_to_uvchr((U8*)s, 0);
2178 const unsigned char c = (unsigned char)u & 0xFF;
2179 if (flags & UNI_DISPLAY_BACKSLASH) {
2196 Perl_sv_catpvf(aTHX_ dsv, "\\%c", ok);
2199 /* isPRINT() is the locale-blind version. */
2200 if (!ok && (flags & UNI_DISPLAY_ISPRINT) && isPRINT(c)) {
2201 Perl_sv_catpvf(aTHX_ dsv, "%c", c);
2206 Perl_sv_catpvf(aTHX_ dsv, "\\x{%"UVxf"}", u);
2209 sv_catpvs(dsv, "...");
2215 =for apidoc A|char *|sv_uni_display|SV *dsv|SV *ssv|STRLEN pvlim|UV flags
2217 Build to the scalar dsv a displayable version of the scalar sv,
2218 the displayable version being at most pvlim bytes long
2219 (if longer, the rest is truncated and "..." will be appended).
2221 The flags argument is as in pv_uni_display().
2223 The pointer to the PV of the dsv is returned.
2227 Perl_sv_uni_display(pTHX_ SV *dsv, SV *ssv, STRLEN pvlim, UV flags)
2229 return Perl_pv_uni_display(aTHX_ dsv, (const U8*)SvPVX_const(ssv),
2230 SvCUR(ssv), pvlim, flags);
2234 =for apidoc A|I32|ibcmp_utf8|const char *s1|char **pe1|register UV l1|bool u1|const char *s2|char **pe2|register UV l2|bool u2
2236 Return true if the strings s1 and s2 differ case-insensitively, false
2237 if not (if they are equal case-insensitively). If u1 is true, the
2238 string s1 is assumed to be in UTF-8-encoded Unicode. If u2 is true,
2239 the string s2 is assumed to be in UTF-8-encoded Unicode. If u1 or u2
2240 are false, the respective string is assumed to be in native 8-bit
2243 If the pe1 and pe2 are non-NULL, the scanning pointers will be copied
2244 in there (they will point at the beginning of the I<next> character).
2245 If the pointers behind pe1 or pe2 are non-NULL, they are the end
2246 pointers beyond which scanning will not continue under any
2247 circumstances. If the byte lengths l1 and l2 are non-zero, s1+l1 and
2248 s2+l2 will be used as goal end pointers that will also stop the scan,
2249 and which qualify towards defining a successful match: all the scans
2250 that define an explicit length must reach their goal pointers for
2251 a match to succeed).
2253 For case-insensitiveness, the "casefolding" of Unicode is used
2254 instead of upper/lowercasing both the characters, see
2255 http://www.unicode.org/unicode/reports/tr21/ (Case Mappings).
2259 Perl_ibcmp_utf8(pTHX_ const char *s1, char **pe1, register UV l1, bool u1, const char *s2, char **pe2, register UV l2, bool u2)
2262 register const U8 *p1 = (const U8*)s1;
2263 register const U8 *p2 = (const U8*)s2;
2264 register const U8 *f1 = NULL;
2265 register const U8 *f2 = NULL;
2266 register U8 *e1 = NULL;
2267 register U8 *q1 = NULL;
2268 register U8 *e2 = NULL;
2269 register U8 *q2 = NULL;
2270 STRLEN n1 = 0, n2 = 0;
2271 U8 foldbuf1[UTF8_MAXBYTES_CASE+1];
2272 U8 foldbuf2[UTF8_MAXBYTES_CASE+1];
2274 STRLEN foldlen1, foldlen2;
2279 if (e1 == 0 || (l1 && l1 < (UV)(e1 - (const U8*)s1)))
2280 f1 = (const U8*)s1 + l1;
2283 if (e2 == 0 || (l2 && l2 < (UV)(e2 - (const U8*)s2)))
2284 f2 = (const U8*)s2 + l2;
2286 if ((e1 == 0 && f1 == 0) || (e2 == 0 && f2 == 0) || (f1 == 0 && f2 == 0))
2287 return 1; /* mismatch; possible infinite loop or false positive */
2290 natbuf[1] = 0; /* Need to terminate the buffer. */
2292 while ((e1 == 0 || p1 < e1) &&
2293 (f1 == 0 || p1 < f1) &&
2294 (e2 == 0 || p2 < e2) &&
2295 (f2 == 0 || p2 < f2)) {
2298 to_utf8_fold(p1, foldbuf1, &foldlen1);
2300 uvuni_to_utf8(natbuf, (UV) NATIVE_TO_UNI(((UV)*p1)));
2301 to_utf8_fold(natbuf, foldbuf1, &foldlen1);
2308 to_utf8_fold(p2, foldbuf2, &foldlen2);
2310 uvuni_to_utf8(natbuf, (UV) NATIVE_TO_UNI(((UV)*p2)));
2311 to_utf8_fold(natbuf, foldbuf2, &foldlen2);
2317 if ( UTF8SKIP(q1) != UTF8SKIP(q2) ||
2318 (UTF8SKIP(q1) == 1 && *q1 != *q2) ||
2319 memNE((char*)q1, (char*)q2, UTF8SKIP(q1)) )
2320 return 1; /* mismatch */
2327 p1 += u1 ? UTF8SKIP(p1) : 1;
2329 p2 += u2 ? UTF8SKIP(p2) : 1;
2333 /* A match is defined by all the scans that specified
2334 * an explicit length reaching their final goals. */
2335 match = (f1 == 0 || p1 == f1) && (f2 == 0 || p2 == f2);
2344 return match ? 0 : 1; /* 0 match, 1 mismatch */
2349 * c-indentation-style: bsd
2351 * indent-tabs-mode: t
2354 * ex: set ts=8 sts=4 sw=4 noet: