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(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);
243 if (IS_UTF8_CHAR_FAST(len))
244 return IS_UTF8_CHAR(s, len) ? len : 0;
245 #endif /* #ifdef IS_UTF8_CHAR */
246 return is_utf8_char_slow(s, len);
250 =for apidoc A|bool|is_utf8_string|const U8 *s|STRLEN len
252 Returns true if first C<len> bytes of the given string form a valid
253 UTF-8 string, false otherwise. Note that 'a valid UTF-8 string' does
254 not mean 'a string that contains code points above 0x7F encoded in UTF-8'
255 because a valid ASCII string is a valid UTF-8 string.
257 See also is_utf8_string_loclen() and is_utf8_string_loc().
263 Perl_is_utf8_string(pTHX_ const U8 *s, STRLEN len)
265 const U8* const send = s + (len ? len : strlen((const char *)s));
272 /* Inline the easy bits of is_utf8_char() here for speed... */
273 if (UTF8_IS_INVARIANT(*x))
275 else if (!UTF8_IS_START(*x))
278 /* ... and call is_utf8_char() only if really needed. */
281 if (IS_UTF8_CHAR_FAST(c)) {
282 if (!IS_UTF8_CHAR(x, c))
286 c = 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)
329 const U8* const send = s + (len ? len : strlen((const char *)s));
336 /* Inline the easy bits of is_utf8_char() here for speed... */
337 if (UTF8_IS_INVARIANT(*x))
339 else if (!UTF8_IS_START(*x))
342 /* ... and call is_utf8_char() only if really needed. */
345 if (IS_UTF8_CHAR_FAST(c)) {
346 if (!IS_UTF8_CHAR(x, c))
349 c = is_utf8_char_slow(x, c);
352 #endif /* #ifdef IS_UTF8_CHAR */
371 =for apidoc A|UV|utf8n_to_uvuni|const U8 *s|STRLEN curlen|STRLEN *retlen|U32 flags
373 Bottom level UTF-8 decode routine.
374 Returns the unicode code point value of the first character in the string C<s>
375 which is assumed to be in UTF-8 encoding and no longer than C<curlen>;
376 C<retlen> will be set to the length, in bytes, of that character.
378 If C<s> does not point to a well-formed UTF-8 character, the behaviour
379 is dependent on the value of C<flags>: if it contains UTF8_CHECK_ONLY,
380 it is assumed that the caller will raise a warning, and this function
381 will silently just set C<retlen> to C<-1> and return zero. If the
382 C<flags> does not contain UTF8_CHECK_ONLY, warnings about
383 malformations will be given, C<retlen> will be set to the expected
384 length of the UTF-8 character in bytes, and zero will be returned.
386 The C<flags> can also contain various flags to allow deviations from
387 the strict UTF-8 encoding (see F<utf8.h>).
389 Most code should use utf8_to_uvchr() rather than call this directly.
395 Perl_utf8n_to_uvuni(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen, U32 flags)
398 const U8 * const s0 = s;
401 const bool dowarn = ckWARN_d(WARN_UTF8);
402 const UV startbyte = *s;
403 STRLEN expectlen = 0;
406 /* This list is a superset of the UTF8_ALLOW_XXX. */
408 #define UTF8_WARN_EMPTY 1
409 #define UTF8_WARN_CONTINUATION 2
410 #define UTF8_WARN_NON_CONTINUATION 3
411 #define UTF8_WARN_FE_FF 4
412 #define UTF8_WARN_SHORT 5
413 #define UTF8_WARN_OVERFLOW 6
414 #define UTF8_WARN_SURROGATE 7
415 #define UTF8_WARN_LONG 8
416 #define UTF8_WARN_FFFF 9 /* Also FFFE. */
419 !(flags & UTF8_ALLOW_EMPTY)) {
420 warning = UTF8_WARN_EMPTY;
424 if (UTF8_IS_INVARIANT(uv)) {
427 return (UV) (NATIVE_TO_UTF(*s));
430 if (UTF8_IS_CONTINUATION(uv) &&
431 !(flags & UTF8_ALLOW_CONTINUATION)) {
432 warning = UTF8_WARN_CONTINUATION;
436 if (UTF8_IS_START(uv) && curlen > 1 && !UTF8_IS_CONTINUATION(s[1]) &&
437 !(flags & UTF8_ALLOW_NON_CONTINUATION)) {
438 warning = UTF8_WARN_NON_CONTINUATION;
443 uv = NATIVE_TO_UTF(uv);
445 if ((uv == 0xfe || uv == 0xff) &&
446 !(flags & UTF8_ALLOW_FE_FF)) {
447 warning = UTF8_WARN_FE_FF;
452 if (!(uv & 0x20)) { len = 2; uv &= 0x1f; }
453 else if (!(uv & 0x10)) { len = 3; uv &= 0x0f; }
454 else if (!(uv & 0x08)) { len = 4; uv &= 0x07; }
455 else if (!(uv & 0x04)) { len = 5; uv &= 0x03; }
457 else if (!(uv & 0x02)) { len = 6; uv &= 0x01; }
458 else { len = 7; uv &= 0x01; }
460 else if (!(uv & 0x02)) { len = 6; uv &= 0x01; }
461 else if (!(uv & 0x01)) { len = 7; uv = 0; }
462 else { len = 13; uv = 0; } /* whoa! */
470 if ((curlen < expectlen) &&
471 !(flags & UTF8_ALLOW_SHORT)) {
472 warning = UTF8_WARN_SHORT;
481 if (!UTF8_IS_CONTINUATION(*s) &&
482 !(flags & UTF8_ALLOW_NON_CONTINUATION)) {
484 warning = UTF8_WARN_NON_CONTINUATION;
488 uv = UTF8_ACCUMULATE(uv, *s);
490 /* These cannot be allowed. */
492 if (expectlen != 13 && !(flags & UTF8_ALLOW_LONG)) {
493 warning = UTF8_WARN_LONG;
497 else { /* uv < ouv */
498 /* This cannot be allowed. */
499 warning = UTF8_WARN_OVERFLOW;
507 if (UNICODE_IS_SURROGATE(uv) &&
508 !(flags & UTF8_ALLOW_SURROGATE)) {
509 warning = UTF8_WARN_SURROGATE;
511 } else if ((expectlen > (STRLEN)UNISKIP(uv)) &&
512 !(flags & UTF8_ALLOW_LONG)) {
513 warning = UTF8_WARN_LONG;
515 } else if (UNICODE_IS_ILLEGAL(uv) &&
516 !(flags & UTF8_ALLOW_FFFF)) {
517 warning = UTF8_WARN_FFFF;
525 if (flags & UTF8_CHECK_ONLY) {
527 *retlen = ((STRLEN) -1);
532 SV* const sv = sv_2mortal(newSVpvs("Malformed UTF-8 character "));
535 case 0: /* Intentionally empty. */ break;
536 case UTF8_WARN_EMPTY:
537 sv_catpvs(sv, "(empty string)");
539 case UTF8_WARN_CONTINUATION:
540 Perl_sv_catpvf(aTHX_ sv, "(unexpected continuation byte 0x%02"UVxf", with no preceding start byte)", uv);
542 case UTF8_WARN_NON_CONTINUATION:
544 Perl_sv_catpvf(aTHX_ sv, "(unexpected non-continuation byte 0x%02"UVxf", immediately after start byte 0x%02"UVxf")",
545 (UV)s[1], startbyte);
547 const int len = (int)(s-s0);
548 Perl_sv_catpvf(aTHX_ sv, "(unexpected non-continuation byte 0x%02"UVxf", %d byte%s after start byte 0x%02"UVxf", expected %d bytes)",
549 (UV)s[1], len, len > 1 ? "s" : "", startbyte, (int)expectlen);
553 case UTF8_WARN_FE_FF:
554 Perl_sv_catpvf(aTHX_ sv, "(byte 0x%02"UVxf")", uv);
556 case UTF8_WARN_SHORT:
557 Perl_sv_catpvf(aTHX_ sv, "(%d byte%s, need %d, after start byte 0x%02"UVxf")",
558 (int)curlen, curlen == 1 ? "" : "s", (int)expectlen, startbyte);
559 expectlen = curlen; /* distance for caller to skip */
561 case UTF8_WARN_OVERFLOW:
562 Perl_sv_catpvf(aTHX_ sv, "(overflow at 0x%"UVxf", byte 0x%02x, after start byte 0x%02"UVxf")",
565 case UTF8_WARN_SURROGATE:
566 Perl_sv_catpvf(aTHX_ sv, "(UTF-16 surrogate 0x%04"UVxf")", uv);
569 Perl_sv_catpvf(aTHX_ sv, "(%d byte%s, need %d, after start byte 0x%02"UVxf")",
570 (int)expectlen, expectlen == 1 ? "": "s", UNISKIP(uv), startbyte);
573 Perl_sv_catpvf(aTHX_ sv, "(character 0x%04"UVxf")", uv);
576 sv_catpvs(sv, "(unknown reason)");
581 const char * const s = SvPVX_const(sv);
584 Perl_warner(aTHX_ packWARN(WARN_UTF8),
585 "%s in %s", s, OP_DESC(PL_op));
587 Perl_warner(aTHX_ packWARN(WARN_UTF8), "%s", s);
592 *retlen = expectlen ? expectlen : len;
598 =for apidoc A|UV|utf8_to_uvchr|const U8 *s|STRLEN *retlen
600 Returns the native character value of the first character in the string C<s>
601 which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
602 length, in bytes, of that character.
604 If C<s> does not point to a well-formed UTF-8 character, zero is
605 returned and retlen is set, if possible, to -1.
611 Perl_utf8_to_uvchr(pTHX_ const U8 *s, STRLEN *retlen)
613 return utf8n_to_uvchr(s, UTF8_MAXBYTES, retlen,
614 ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
618 =for apidoc A|UV|utf8_to_uvuni|const U8 *s|STRLEN *retlen
620 Returns the Unicode code point of the first character in the string C<s>
621 which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
622 length, in bytes, of that character.
624 This function should only be used when returned UV is considered
625 an index into the Unicode semantic tables (e.g. swashes).
627 If C<s> does not point to a well-formed UTF-8 character, zero is
628 returned and retlen is set, if possible, to -1.
634 Perl_utf8_to_uvuni(pTHX_ const U8 *s, STRLEN *retlen)
636 /* Call the low level routine asking for checks */
637 return Perl_utf8n_to_uvuni(aTHX_ s, UTF8_MAXBYTES, retlen,
638 ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
642 =for apidoc A|STRLEN|utf8_length|const U8 *s|const U8 *e
644 Return the length of the UTF-8 char encoded string C<s> in characters.
645 Stops at C<e> (inclusive). If C<e E<lt> s> or if the scan would end
646 up past C<e>, croaks.
652 Perl_utf8_length(pTHX_ const U8 *s, const U8 *e)
658 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g.
659 * the bitops (especially ~) can create illegal UTF-8.
660 * In other words: in Perl UTF-8 is not just for Unicode. */
663 goto warn_and_return;
668 if (ckWARN_d(WARN_UTF8)) {
670 Perl_warner(aTHX_ packWARN(WARN_UTF8),
671 "%s in %s", unees, OP_DESC(PL_op));
673 Perl_warner(aTHX_ packWARN(WARN_UTF8), unees);
685 =for apidoc A|IV|utf8_distance|const U8 *a|const U8 *b
687 Returns the number of UTF-8 characters between the UTF-8 pointers C<a>
690 WARNING: use only if you *know* that the pointers point inside the
697 Perl_utf8_distance(pTHX_ const U8 *a, const U8 *b)
699 return (a < b) ? -1 * (IV) utf8_length(a, b) : (IV) utf8_length(b, a);
703 =for apidoc A|U8 *|utf8_hop|U8 *s|I32 off
705 Return the UTF-8 pointer C<s> displaced by C<off> characters, either
708 WARNING: do not use the following unless you *know* C<off> is within
709 the UTF-8 data pointed to by C<s> *and* that on entry C<s> is aligned
710 on the first byte of character or just after the last byte of a character.
716 Perl_utf8_hop(pTHX_ const U8 *s, I32 off)
719 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g
720 * the bitops (especially ~) can create illegal UTF-8.
721 * In other words: in Perl UTF-8 is not just for Unicode. */
730 while (UTF8_IS_CONTINUATION(*s))
738 =for apidoc A|U8 *|utf8_to_bytes|U8 *s|STRLEN *len
740 Converts a string C<s> of length C<len> from UTF-8 into byte encoding.
741 Unlike C<bytes_to_utf8>, this over-writes the original string, and
742 updates len to contain the new length.
743 Returns zero on failure, setting C<len> to -1.
745 If you need a copy of the string, see C<bytes_from_utf8>.
751 Perl_utf8_to_bytes(pTHX_ U8 *s, STRLEN *len)
754 U8 * const send = s + *len;
757 /* ensure valid UTF-8 and chars < 256 before updating string */
761 if (!UTF8_IS_INVARIANT(c) &&
762 (!UTF8_IS_DOWNGRADEABLE_START(c) || (s >= send)
763 || !(c = *s++) || !UTF8_IS_CONTINUATION(c))) {
764 *len = ((STRLEN) -1);
772 *d++ = (U8)utf8_to_uvchr(s, &ulen);
781 =for apidoc A|U8 *|bytes_from_utf8|const U8 *s|STRLEN *len|bool *is_utf8
783 Converts a string C<s> of length C<len> from UTF-8 into byte encoding.
784 Unlike C<utf8_to_bytes> but like C<bytes_to_utf8>, returns a pointer to
785 the newly-created string, and updates C<len> to contain the new
786 length. Returns the original string if no conversion occurs, C<len>
787 is unchanged. Do nothing if C<is_utf8> points to 0. Sets C<is_utf8> to
788 0 if C<s> is converted or contains all 7bit characters.
794 Perl_bytes_from_utf8(pTHX_ const U8 *s, STRLEN *len, bool *is_utf8)
805 /* ensure valid UTF-8 and chars < 256 before converting string */
806 for (send = s + *len; s < send;) {
808 if (!UTF8_IS_INVARIANT(c)) {
809 if (UTF8_IS_DOWNGRADEABLE_START(c) && s < send &&
810 (c = *s++) && UTF8_IS_CONTINUATION(c))
819 Newx(d, (*len) - count + 1, U8);
820 s = start; start = d;
823 if (!UTF8_IS_INVARIANT(c)) {
824 /* Then it is two-byte encoded */
825 c = UTF8_ACCUMULATE(NATIVE_TO_UTF(c), *s++);
826 c = ASCII_TO_NATIVE(c);
836 =for apidoc A|U8 *|bytes_to_utf8|const U8 *s|STRLEN *len
838 Converts a string C<s> of length C<len> from ASCII into UTF-8 encoding.
839 Returns a pointer to the newly-created string, and sets C<len> to
840 reflect the new length.
842 If you want to convert to UTF-8 from other encodings than ASCII,
843 see sv_recode_to_utf8().
849 Perl_bytes_to_utf8(pTHX_ const U8 *s, STRLEN *len)
851 const U8 * const send = s + (*len);
856 Newx(d, (*len) * 2 + 1, U8);
860 const UV uv = NATIVE_TO_ASCII(*s++);
861 if (UNI_IS_INVARIANT(uv))
862 *d++ = (U8)UTF_TO_NATIVE(uv);
864 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
865 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
874 * Convert native (big-endian) or reversed (little-endian) UTF-16 to UTF-8.
876 * Destination must be pre-extended to 3/2 source. Do not use in-place.
877 * We optimize for native, for obvious reasons. */
880 Perl_utf16_to_utf8(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen)
885 if (bytelen == 1 && p[0] == 0) { /* Be understanding. */
892 Perl_croak(aTHX_ "panic: utf16_to_utf8: odd bytelen %"UVf, (UV)bytelen);
897 UV uv = (p[0] << 8) + p[1]; /* UTF-16BE */
901 *d++ = UNI_TO_NATIVE(uv);
908 *d++ = (U8)(( uv >> 6) | 0xc0);
909 *d++ = (U8)(( uv & 0x3f) | 0x80);
912 if (uv >= 0xd800 && uv < 0xdbff) { /* surrogates */
913 UV low = (p[0] << 8) + p[1];
915 if (low < 0xdc00 || low >= 0xdfff)
916 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
917 uv = ((uv - 0xd800) << 10) + (low - 0xdc00) + 0x10000;
920 *d++ = (U8)(( uv >> 12) | 0xe0);
921 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
922 *d++ = (U8)(( uv & 0x3f) | 0x80);
926 *d++ = (U8)(( uv >> 18) | 0xf0);
927 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
928 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
929 *d++ = (U8)(( uv & 0x3f) | 0x80);
933 *newlen = d - dstart;
937 /* Note: this one is slightly destructive of the source. */
940 Perl_utf16_to_utf8_reversed(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen)
943 U8* const send = s + bytelen;
950 return utf16_to_utf8(p, d, bytelen, newlen);
953 /* for now these are all defined (inefficiently) in terms of the utf8 versions */
956 Perl_is_uni_alnum(pTHX_ UV c)
958 U8 tmpbuf[UTF8_MAXBYTES+1];
959 uvchr_to_utf8(tmpbuf, c);
960 return is_utf8_alnum(tmpbuf);
964 Perl_is_uni_alnumc(pTHX_ UV c)
966 U8 tmpbuf[UTF8_MAXBYTES+1];
967 uvchr_to_utf8(tmpbuf, c);
968 return is_utf8_alnumc(tmpbuf);
972 Perl_is_uni_idfirst(pTHX_ UV c)
974 U8 tmpbuf[UTF8_MAXBYTES+1];
975 uvchr_to_utf8(tmpbuf, c);
976 return is_utf8_idfirst(tmpbuf);
980 Perl_is_uni_alpha(pTHX_ UV c)
982 U8 tmpbuf[UTF8_MAXBYTES+1];
983 uvchr_to_utf8(tmpbuf, c);
984 return is_utf8_alpha(tmpbuf);
988 Perl_is_uni_ascii(pTHX_ UV c)
990 U8 tmpbuf[UTF8_MAXBYTES+1];
991 uvchr_to_utf8(tmpbuf, c);
992 return is_utf8_ascii(tmpbuf);
996 Perl_is_uni_space(pTHX_ UV c)
998 U8 tmpbuf[UTF8_MAXBYTES+1];
999 uvchr_to_utf8(tmpbuf, c);
1000 return is_utf8_space(tmpbuf);
1004 Perl_is_uni_digit(pTHX_ UV c)
1006 U8 tmpbuf[UTF8_MAXBYTES+1];
1007 uvchr_to_utf8(tmpbuf, c);
1008 return is_utf8_digit(tmpbuf);
1012 Perl_is_uni_upper(pTHX_ UV c)
1014 U8 tmpbuf[UTF8_MAXBYTES+1];
1015 uvchr_to_utf8(tmpbuf, c);
1016 return is_utf8_upper(tmpbuf);
1020 Perl_is_uni_lower(pTHX_ UV c)
1022 U8 tmpbuf[UTF8_MAXBYTES+1];
1023 uvchr_to_utf8(tmpbuf, c);
1024 return is_utf8_lower(tmpbuf);
1028 Perl_is_uni_cntrl(pTHX_ UV c)
1030 U8 tmpbuf[UTF8_MAXBYTES+1];
1031 uvchr_to_utf8(tmpbuf, c);
1032 return is_utf8_cntrl(tmpbuf);
1036 Perl_is_uni_graph(pTHX_ UV c)
1038 U8 tmpbuf[UTF8_MAXBYTES+1];
1039 uvchr_to_utf8(tmpbuf, c);
1040 return is_utf8_graph(tmpbuf);
1044 Perl_is_uni_print(pTHX_ UV c)
1046 U8 tmpbuf[UTF8_MAXBYTES+1];
1047 uvchr_to_utf8(tmpbuf, c);
1048 return is_utf8_print(tmpbuf);
1052 Perl_is_uni_punct(pTHX_ UV c)
1054 U8 tmpbuf[UTF8_MAXBYTES+1];
1055 uvchr_to_utf8(tmpbuf, c);
1056 return is_utf8_punct(tmpbuf);
1060 Perl_is_uni_xdigit(pTHX_ UV c)
1062 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1063 uvchr_to_utf8(tmpbuf, c);
1064 return is_utf8_xdigit(tmpbuf);
1068 Perl_to_uni_upper(pTHX_ UV c, U8* p, STRLEN *lenp)
1070 uvchr_to_utf8(p, c);
1071 return to_utf8_upper(p, p, lenp);
1075 Perl_to_uni_title(pTHX_ UV c, U8* p, STRLEN *lenp)
1077 uvchr_to_utf8(p, c);
1078 return to_utf8_title(p, p, lenp);
1082 Perl_to_uni_lower(pTHX_ UV c, U8* p, STRLEN *lenp)
1084 uvchr_to_utf8(p, c);
1085 return to_utf8_lower(p, p, lenp);
1089 Perl_to_uni_fold(pTHX_ UV c, U8* p, STRLEN *lenp)
1091 uvchr_to_utf8(p, c);
1092 return to_utf8_fold(p, p, lenp);
1095 /* for now these all assume no locale info available for Unicode > 255 */
1098 Perl_is_uni_alnum_lc(pTHX_ UV c)
1100 return is_uni_alnum(c); /* XXX no locale support yet */
1104 Perl_is_uni_alnumc_lc(pTHX_ UV c)
1106 return is_uni_alnumc(c); /* XXX no locale support yet */
1110 Perl_is_uni_idfirst_lc(pTHX_ UV c)
1112 return is_uni_idfirst(c); /* XXX no locale support yet */
1116 Perl_is_uni_alpha_lc(pTHX_ UV c)
1118 return is_uni_alpha(c); /* XXX no locale support yet */
1122 Perl_is_uni_ascii_lc(pTHX_ UV c)
1124 return is_uni_ascii(c); /* XXX no locale support yet */
1128 Perl_is_uni_space_lc(pTHX_ UV c)
1130 return is_uni_space(c); /* XXX no locale support yet */
1134 Perl_is_uni_digit_lc(pTHX_ UV c)
1136 return is_uni_digit(c); /* XXX no locale support yet */
1140 Perl_is_uni_upper_lc(pTHX_ UV c)
1142 return is_uni_upper(c); /* XXX no locale support yet */
1146 Perl_is_uni_lower_lc(pTHX_ UV c)
1148 return is_uni_lower(c); /* XXX no locale support yet */
1152 Perl_is_uni_cntrl_lc(pTHX_ UV c)
1154 return is_uni_cntrl(c); /* XXX no locale support yet */
1158 Perl_is_uni_graph_lc(pTHX_ UV c)
1160 return is_uni_graph(c); /* XXX no locale support yet */
1164 Perl_is_uni_print_lc(pTHX_ UV c)
1166 return is_uni_print(c); /* XXX no locale support yet */
1170 Perl_is_uni_punct_lc(pTHX_ UV c)
1172 return is_uni_punct(c); /* XXX no locale support yet */
1176 Perl_is_uni_xdigit_lc(pTHX_ UV c)
1178 return is_uni_xdigit(c); /* XXX no locale support yet */
1182 Perl_to_uni_upper_lc(pTHX_ U32 c)
1184 /* XXX returns only the first character -- do not use XXX */
1185 /* XXX no locale support yet */
1187 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1188 return (U32)to_uni_upper(c, tmpbuf, &len);
1192 Perl_to_uni_title_lc(pTHX_ U32 c)
1194 /* XXX returns only the first character XXX -- do not use XXX */
1195 /* XXX no locale support yet */
1197 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1198 return (U32)to_uni_title(c, tmpbuf, &len);
1202 Perl_to_uni_lower_lc(pTHX_ U32 c)
1204 /* XXX returns only the first character -- do not use XXX */
1205 /* XXX no locale support yet */
1207 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1208 return (U32)to_uni_lower(c, tmpbuf, &len);
1212 S_is_utf8_common(pTHX_ const U8 *const p, SV **swash,
1213 const char *const swashname)
1216 if (!is_utf8_char(p))
1219 *swash = swash_init("utf8", swashname, &PL_sv_undef, 1, 0);
1220 return swash_fetch(*swash, p, TRUE) != 0;
1224 Perl_is_utf8_alnum(pTHX_ const U8 *p)
1227 /* NOTE: "IsWord", not "IsAlnum", since Alnum is a true
1228 * descendant of isalnum(3), in other words, it doesn't
1229 * contain the '_'. --jhi */
1230 return is_utf8_common(p, &PL_utf8_alnum, "IsWord");
1234 Perl_is_utf8_alnumc(pTHX_ const U8 *p)
1237 return is_utf8_common(p, &PL_utf8_alnumc, "IsAlnumC");
1241 Perl_is_utf8_idfirst(pTHX_ const U8 *p) /* The naming is historical. */
1246 /* is_utf8_idstart would be more logical. */
1247 return is_utf8_common(p, &PL_utf8_idstart, "IdStart");
1251 Perl_is_utf8_idcont(pTHX_ const U8 *p)
1256 return is_utf8_common(p, &PL_utf8_idcont, "IdContinue");
1260 Perl_is_utf8_alpha(pTHX_ const U8 *p)
1263 return is_utf8_common(p, &PL_utf8_alpha, "IsAlpha");
1267 Perl_is_utf8_ascii(pTHX_ const U8 *p)
1270 return is_utf8_common(p, &PL_utf8_ascii, "IsAscii");
1274 Perl_is_utf8_space(pTHX_ const U8 *p)
1277 return is_utf8_common(p, &PL_utf8_space, "IsSpacePerl");
1281 Perl_is_utf8_digit(pTHX_ const U8 *p)
1284 return is_utf8_common(p, &PL_utf8_digit, "IsDigit");
1288 Perl_is_utf8_upper(pTHX_ const U8 *p)
1291 return is_utf8_common(p, &PL_utf8_upper, "IsUppercase");
1295 Perl_is_utf8_lower(pTHX_ const U8 *p)
1298 return is_utf8_common(p, &PL_utf8_lower, "IsLowercase");
1302 Perl_is_utf8_cntrl(pTHX_ const U8 *p)
1305 return is_utf8_common(p, &PL_utf8_cntrl, "IsCntrl");
1309 Perl_is_utf8_graph(pTHX_ const U8 *p)
1312 return is_utf8_common(p, &PL_utf8_graph, "IsGraph");
1316 Perl_is_utf8_print(pTHX_ const U8 *p)
1319 return is_utf8_common(p, &PL_utf8_print, "IsPrint");
1323 Perl_is_utf8_punct(pTHX_ const U8 *p)
1326 return is_utf8_common(p, &PL_utf8_punct, "IsPunct");
1330 Perl_is_utf8_xdigit(pTHX_ const U8 *p)
1333 return is_utf8_common(p, &PL_utf8_xdigit, "Isxdigit");
1337 Perl_is_utf8_mark(pTHX_ const U8 *p)
1340 return is_utf8_common(p, &PL_utf8_mark, "IsM");
1344 =for apidoc A|UV|to_utf8_case|U8 *p|U8* ustrp|STRLEN *lenp|SV **swash|char *normal|char *special
1346 The "p" contains the pointer to the UTF-8 string encoding
1347 the character that is being converted.
1349 The "ustrp" is a pointer to the character buffer to put the
1350 conversion result to. The "lenp" is a pointer to the length
1353 The "swashp" is a pointer to the swash to use.
1355 Both the special and normal mappings are stored lib/unicore/To/Foo.pl,
1356 and loaded by SWASHNEW, using lib/utf8_heavy.pl. The special (usually,
1357 but not always, a multicharacter mapping), is tried first.
1359 The "special" is a string like "utf8::ToSpecLower", which means the
1360 hash %utf8::ToSpecLower. The access to the hash is through
1361 Perl_to_utf8_case().
1363 The "normal" is a string like "ToLower" which means the swash
1369 Perl_to_utf8_case(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp,
1370 SV **swashp, const char *normal, const char *special)
1373 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1376 const UV uv0 = utf8_to_uvchr(p, NULL);
1377 /* The NATIVE_TO_UNI() and UNI_TO_NATIVE() mappings
1378 * are necessary in EBCDIC, they are redundant no-ops
1379 * in ASCII-ish platforms, and hopefully optimized away. */
1380 const UV uv1 = NATIVE_TO_UNI(uv0);
1381 uvuni_to_utf8(tmpbuf, uv1);
1383 if (!*swashp) /* load on-demand */
1384 *swashp = swash_init("utf8", normal, &PL_sv_undef, 4, 0);
1386 /* The 0xDF is the only special casing Unicode code point below 0x100. */
1387 if (special && (uv1 == 0xDF || uv1 > 0xFF)) {
1388 /* It might be "special" (sometimes, but not always,
1389 * a multicharacter mapping) */
1390 HV * const hv = get_hv(special, FALSE);
1394 (svp = hv_fetch(hv, (const char*)tmpbuf, UNISKIP(uv1), FALSE)) &&
1398 s = SvPV_const(*svp, len);
1400 len = uvuni_to_utf8(ustrp, NATIVE_TO_UNI(*(U8*)s)) - ustrp;
1403 /* If we have EBCDIC we need to remap the characters
1404 * since any characters in the low 256 are Unicode
1405 * code points, not EBCDIC. */
1406 U8 *t = (U8*)s, *tend = t + len, *d;
1413 const UV c = utf8_to_uvchr(t, &tlen);
1415 d = uvchr_to_utf8(d, UNI_TO_NATIVE(c));
1424 d = uvchr_to_utf8(d, UNI_TO_NATIVE(*t));
1429 Copy(tmpbuf, ustrp, len, U8);
1431 Copy(s, ustrp, len, U8);
1437 if (!len && *swashp) {
1438 const UV uv2 = swash_fetch(*swashp, tmpbuf, TRUE);
1441 /* It was "normal" (a single character mapping). */
1442 const UV uv3 = UNI_TO_NATIVE(uv2);
1443 len = uvchr_to_utf8(ustrp, uv3) - ustrp;
1447 if (!len) /* Neither: just copy. */
1448 len = uvchr_to_utf8(ustrp, uv0) - ustrp;
1453 return len ? utf8_to_uvchr(ustrp, 0) : 0;
1457 =for apidoc A|UV|to_utf8_upper|const U8 *p|U8 *ustrp|STRLEN *lenp
1459 Convert the UTF-8 encoded character at p to its uppercase version and
1460 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
1461 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since
1462 the uppercase version may be longer than the original character.
1464 The first character of the uppercased version is returned
1465 (but note, as explained above, that there may be more.)
1470 Perl_to_utf8_upper(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
1473 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
1474 &PL_utf8_toupper, "ToUpper", "utf8::ToSpecUpper");
1478 =for apidoc A|UV|to_utf8_title|const U8 *p|U8 *ustrp|STRLEN *lenp
1480 Convert the UTF-8 encoded character at p to its titlecase version and
1481 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
1482 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
1483 titlecase version may be longer than the original character.
1485 The first character of the titlecased version is returned
1486 (but note, as explained above, that there may be more.)
1491 Perl_to_utf8_title(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
1494 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
1495 &PL_utf8_totitle, "ToTitle", "utf8::ToSpecTitle");
1499 =for apidoc A|UV|to_utf8_lower|const U8 *p|U8 *ustrp|STRLEN *lenp
1501 Convert the UTF-8 encoded character at p to its lowercase version and
1502 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
1503 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
1504 lowercase version may be longer than the original character.
1506 The first character of the lowercased version is returned
1507 (but note, as explained above, that there may be more.)
1512 Perl_to_utf8_lower(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
1515 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
1516 &PL_utf8_tolower, "ToLower", "utf8::ToSpecLower");
1520 =for apidoc A|UV|to_utf8_fold|const U8 *p|U8 *ustrp|STRLEN *lenp
1522 Convert the UTF-8 encoded character at p to its foldcase version and
1523 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
1524 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
1525 foldcase version may be longer than the original character (up to
1528 The first character of the foldcased version is returned
1529 (but note, as explained above, that there may be more.)
1534 Perl_to_utf8_fold(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
1537 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
1538 &PL_utf8_tofold, "ToFold", "utf8::ToSpecFold");
1542 * A "swash" is a swatch hash.
1543 * A "swatch" is a bit vector generated by utf8.c:S_swash_get().
1544 * C<pkg> is a pointer to a package name for SWASHNEW, should be "utf8".
1545 * For other parameters, see utf8::SWASHNEW in lib/utf8_heavy.pl.
1548 Perl_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv, I32 minbits, I32 none)
1552 SV* const tokenbufsv = sv_newmortal();
1554 const size_t pkg_len = strlen(pkg);
1555 const size_t name_len = strlen(name);
1556 HV * const stash = gv_stashpvn(pkg, pkg_len, FALSE);
1559 PUSHSTACKi(PERLSI_MAGIC);
1564 if (!gv_fetchmeth(stash, "SWASHNEW", 8, -1)) { /* demand load utf8 */
1566 errsv_save = newSVsv(ERRSV);
1567 /* It is assumed that callers of this routine are not passing in any
1568 user derived data. */
1569 /* Need to do this after save_re_context() as it will set PL_tainted to
1570 1 while saving $1 etc (see the code after getrx: in Perl_magic_get).
1571 Even line to create errsv_save can turn on PL_tainted. */
1572 SAVEBOOL(PL_tainted);
1574 Perl_load_module(aTHX_ PERL_LOADMOD_NOIMPORT, newSVpvn(pkg,pkg_len),
1577 sv_setsv(ERRSV, errsv_save);
1578 SvREFCNT_dec(errsv_save);
1584 PUSHs(sv_2mortal(newSVpvn(pkg, pkg_len)));
1585 PUSHs(sv_2mortal(newSVpvn(name, name_len)));
1587 PUSHs(sv_2mortal(newSViv(minbits)));
1588 PUSHs(sv_2mortal(newSViv(none)));
1590 if (IN_PERL_COMPILETIME) {
1591 /* XXX ought to be handled by lex_start */
1594 sv_setpv(tokenbufsv, PL_tokenbuf);
1596 errsv_save = newSVsv(ERRSV);
1597 if (call_method("SWASHNEW", G_SCALAR))
1598 retval = newSVsv(*PL_stack_sp--);
1600 retval = &PL_sv_undef;
1602 sv_setsv(ERRSV, errsv_save);
1603 SvREFCNT_dec(errsv_save);
1606 if (IN_PERL_COMPILETIME) {
1608 const char* const pv = SvPV_const(tokenbufsv, len);
1610 Copy(pv, PL_tokenbuf, len+1, char);
1611 CopHINTS_set(PL_curcop, PL_hints);
1613 if (!SvROK(retval) || SvTYPE(SvRV(retval)) != SVt_PVHV) {
1615 Perl_croak(aTHX_ "Can't find Unicode property definition \"%"SVf"\"",
1617 Perl_croak(aTHX_ "SWASHNEW didn't return an HV ref");
1623 /* This API is wrong for special case conversions since we may need to
1624 * return several Unicode characters for a single Unicode character
1625 * (see lib/unicore/SpecCase.txt) The SWASHGET in lib/utf8_heavy.pl is
1626 * the lower-level routine, and it is similarly broken for returning
1627 * multiple values. --jhi */
1628 /* Now SWASHGET is recasted into S_swash_get in this file. */
1631 * Returns the value of property/mapping C<swash> for the first character
1632 * of the string C<ptr>. If C<do_utf8> is true, the string C<ptr> is
1633 * assumed to be in utf8. If C<do_utf8> is false, the string C<ptr> is
1634 * assumed to be in native 8-bit encoding. Caches the swatch in C<swash>.
1637 Perl_swash_fetch(pTHX_ SV *swash, const U8 *ptr, bool do_utf8)
1640 HV* const hv = (HV*)SvRV(swash);
1645 const U8 *tmps = NULL;
1649 const UV c = NATIVE_TO_ASCII(*ptr);
1651 if (!do_utf8 && !UNI_IS_INVARIANT(c)) {
1652 tmputf8[0] = (U8)UTF8_EIGHT_BIT_HI(c);
1653 tmputf8[1] = (U8)UTF8_EIGHT_BIT_LO(c);
1656 /* Given a UTF-X encoded char 0xAA..0xYY,0xZZ
1657 * then the "swatch" is a vec() for al the chars which start
1659 * So the key in the hash (klen) is length of encoded char -1
1661 klen = UTF8SKIP(ptr) - 1;
1665 /* If char in invariant then swatch is for all the invariant chars
1666 * In both UTF-8 and UTF-8-MOD that happens to be UTF_CONTINUATION_MARK
1668 needents = UTF_CONTINUATION_MARK;
1669 off = NATIVE_TO_UTF(ptr[klen]);
1672 /* If char is encoded then swatch is for the prefix */
1673 needents = (1 << UTF_ACCUMULATION_SHIFT);
1674 off = NATIVE_TO_UTF(ptr[klen]) & UTF_CONTINUATION_MASK;
1678 * This single-entry cache saves about 1/3 of the utf8 overhead in test
1679 * suite. (That is, only 7-8% overall over just a hash cache. Still,
1680 * it's nothing to sniff at.) Pity we usually come through at least
1681 * two function calls to get here...
1683 * NB: this code assumes that swatches are never modified, once generated!
1686 if (hv == PL_last_swash_hv &&
1687 klen == PL_last_swash_klen &&
1688 (!klen || memEQ((char *)ptr, (char *)PL_last_swash_key, klen)) )
1690 tmps = PL_last_swash_tmps;
1691 slen = PL_last_swash_slen;
1694 /* Try our second-level swatch cache, kept in a hash. */
1695 SV** svp = hv_fetch(hv, (const char*)ptr, klen, FALSE);
1697 /* If not cached, generate it via swash_get */
1698 if (!svp || !SvPOK(*svp)
1699 || !(tmps = (const U8*)SvPV_const(*svp, slen))) {
1700 /* We use utf8n_to_uvuni() as we want an index into
1701 Unicode tables, not a native character number.
1703 const UV code_point = utf8n_to_uvuni(ptr, UTF8_MAXBYTES, 0,
1705 0 : UTF8_ALLOW_ANY);
1706 swatch = swash_get(swash,
1707 /* On EBCDIC & ~(0xA0-1) isn't a useful thing to do */
1708 (klen) ? (code_point & ~(needents - 1)) : 0,
1711 if (IN_PERL_COMPILETIME)
1712 CopHINTS_set(PL_curcop, PL_hints);
1714 svp = hv_store(hv, (const char *)ptr, klen, swatch, 0);
1716 if (!svp || !(tmps = (U8*)SvPV(*svp, slen))
1717 || (slen << 3) < needents)
1718 Perl_croak(aTHX_ "panic: swash_fetch got improper swatch");
1721 PL_last_swash_hv = hv;
1722 PL_last_swash_klen = klen;
1723 /* FIXME change interpvar.h? */
1724 PL_last_swash_tmps = (U8 *) tmps;
1725 PL_last_swash_slen = slen;
1727 Copy(ptr, PL_last_swash_key, klen, U8);
1730 switch ((int)((slen << 3) / needents)) {
1732 bit = 1 << (off & 7);
1734 return (tmps[off] & bit) != 0;
1739 return (tmps[off] << 8) + tmps[off + 1] ;
1742 return (tmps[off] << 24) + (tmps[off+1] << 16) + (tmps[off+2] << 8) + tmps[off + 3] ;
1744 Perl_croak(aTHX_ "panic: swash_fetch got swatch of unexpected bit width");
1745 NORETURN_FUNCTION_END;
1749 * Returns a swatch (a bit vector string) for a code point sequence
1750 * that starts from the value C<start> and comprises the number C<span>.
1751 * A C<swash> must be an object created by SWASHNEW (see lib/utf8_heavy.pl).
1752 * Should be used via swash_fetch, which will cache the swatch in C<swash>.
1755 S_swash_get(pTHX_ SV* swash, UV start, UV span)
1758 U8 *l, *lend, *x, *xend, *s;
1759 STRLEN lcur, xcur, scur;
1761 HV* const hv = (HV*)SvRV(swash);
1762 SV** const listsvp = hv_fetchs(hv, "LIST", FALSE);
1763 SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE);
1764 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
1765 SV** const nonesvp = hv_fetchs(hv, "NONE", FALSE);
1766 SV** const extssvp = hv_fetchs(hv, "EXTRAS", FALSE);
1767 const U8* const typestr = (U8*)SvPV_nolen(*typesvp);
1768 const int typeto = typestr[0] == 'T' && typestr[1] == 'o';
1769 const STRLEN bits = SvUV(*bitssvp);
1770 const STRLEN octets = bits >> 3; /* if bits == 1, then octets == 0 */
1771 const UV none = SvUV(*nonesvp);
1772 const UV end = start + span;
1774 if (bits != 1 && bits != 8 && bits != 16 && bits != 32) {
1775 Perl_croak(aTHX_ "panic: swash_get doesn't expect bits %"UVuf,
1779 /* create and initialize $swatch */
1780 swatch = newSVpvs("");
1781 scur = octets ? (span * octets) : (span + 7) / 8;
1782 SvGROW(swatch, scur + 1);
1783 s = (U8*)SvPVX(swatch);
1784 if (octets && none) {
1785 const U8* const e = s + scur;
1788 *s++ = (U8)(none & 0xff);
1789 else if (bits == 16) {
1790 *s++ = (U8)((none >> 8) & 0xff);
1791 *s++ = (U8)( none & 0xff);
1793 else if (bits == 32) {
1794 *s++ = (U8)((none >> 24) & 0xff);
1795 *s++ = (U8)((none >> 16) & 0xff);
1796 *s++ = (U8)((none >> 8) & 0xff);
1797 *s++ = (U8)( none & 0xff);
1803 (void)memzero((U8*)s, scur + 1);
1805 SvCUR_set(swatch, scur);
1806 s = (U8*)SvPVX(swatch);
1808 /* read $swash->{LIST} */
1809 l = (U8*)SvPV(*listsvp, lcur);
1814 I32 flags = PERL_SCAN_SILENT_ILLDIGIT | PERL_SCAN_DISALLOW_PREFIX;
1816 U8* const nl = (U8*)memchr(l, '\n', lend - l);
1819 min = grok_hex((char *)l, &numlen, &flags, NULL);
1823 l = nl + 1; /* 1 is length of "\n" */
1827 l = lend; /* to LIST's end at which \n is not found */
1833 flags = PERL_SCAN_SILENT_ILLDIGIT | PERL_SCAN_DISALLOW_PREFIX;
1835 max = grok_hex((char *)l, &numlen, &flags, NULL);
1844 flags = PERL_SCAN_SILENT_ILLDIGIT |
1845 PERL_SCAN_DISALLOW_PREFIX;
1847 val = grok_hex((char *)l, &numlen, &flags, NULL);
1856 Perl_croak(aTHX_ "%s: illegal mapping '%s'",
1862 val = 0; /* bits == 1, then val should be ignored */
1869 Perl_croak(aTHX_ "%s: illegal mapping '%s'", typestr, l);
1873 val = 0; /* bits == 1, then val should be ignored */
1887 if (!none || val < none) {
1892 for (key = min; key <= max; key++) {
1896 /* offset must be non-negative (start <= min <= key < end) */
1897 offset = octets * (key - start);
1899 s[offset] = (U8)(val & 0xff);
1900 else if (bits == 16) {
1901 s[offset ] = (U8)((val >> 8) & 0xff);
1902 s[offset + 1] = (U8)( val & 0xff);
1904 else if (bits == 32) {
1905 s[offset ] = (U8)((val >> 24) & 0xff);
1906 s[offset + 1] = (U8)((val >> 16) & 0xff);
1907 s[offset + 2] = (U8)((val >> 8) & 0xff);
1908 s[offset + 3] = (U8)( val & 0xff);
1911 if (!none || val < none)
1915 else { /* bits == 1, then val should be ignored */
1919 for (key = min; key <= max; key++) {
1920 const STRLEN offset = (STRLEN)(key - start);
1923 s[offset >> 3] |= 1 << (offset & 7);
1929 /* read $swash->{EXTRAS} */
1930 x = (U8*)SvPV(*extssvp, xcur);
1938 SV **otherbitssvp, *other;
1942 const U8 opc = *x++;
1946 nl = (U8*)memchr(x, '\n', xend - x);
1948 if (opc != '-' && opc != '+' && opc != '!' && opc != '&') {
1950 x = nl + 1; /* 1 is length of "\n" */
1954 x = xend; /* to EXTRAS' end at which \n is not found */
1961 namelen = nl - namestr;
1965 namelen = xend - namestr;
1969 othersvp = hv_fetch(hv, (char *)namestr, namelen, FALSE);
1970 otherhv = (HV*)SvRV(*othersvp);
1971 otherbitssvp = hv_fetchs(otherhv, "BITS", FALSE);
1972 otherbits = (STRLEN)SvUV(*otherbitssvp);
1973 if (bits < otherbits)
1974 Perl_croak(aTHX_ "panic: swash_get found swatch size mismatch");
1976 /* The "other" swatch must be destroyed after. */
1977 other = swash_get(*othersvp, start, span);
1978 o = (U8*)SvPV(other, olen);
1981 Perl_croak(aTHX_ "panic: swash_get got improper swatch");
1983 s = (U8*)SvPV(swatch, slen);
1984 if (bits == 1 && otherbits == 1) {
1986 Perl_croak(aTHX_ "panic: swash_get found swatch length mismatch");
2010 STRLEN otheroctets = otherbits >> 3;
2012 U8* const send = s + slen;
2017 if (otherbits == 1) {
2018 otherval = (o[offset >> 3] >> (offset & 7)) & 1;
2022 STRLEN vlen = otheroctets;
2030 if (opc == '+' && otherval)
2031 NOOP; /* replace with otherval */
2032 else if (opc == '!' && !otherval)
2034 else if (opc == '-' && otherval)
2036 else if (opc == '&' && !otherval)
2039 s += octets; /* no replacement */
2044 *s++ = (U8)( otherval & 0xff);
2045 else if (bits == 16) {
2046 *s++ = (U8)((otherval >> 8) & 0xff);
2047 *s++ = (U8)( otherval & 0xff);
2049 else if (bits == 32) {
2050 *s++ = (U8)((otherval >> 24) & 0xff);
2051 *s++ = (U8)((otherval >> 16) & 0xff);
2052 *s++ = (U8)((otherval >> 8) & 0xff);
2053 *s++ = (U8)( otherval & 0xff);
2057 sv_free(other); /* through with it! */
2063 =for apidoc A|U8 *|uvchr_to_utf8|U8 *d|UV uv
2065 Adds the UTF-8 representation of the Native codepoint C<uv> to the end
2066 of the string C<d>; C<d> should be have at least C<UTF8_MAXBYTES+1> free
2067 bytes available. The return value is the pointer to the byte after the
2068 end of the new character. In other words,
2070 d = uvchr_to_utf8(d, uv);
2072 is the recommended wide native character-aware way of saying
2079 /* On ASCII machines this is normally a macro but we want a
2080 real function in case XS code wants it
2083 Perl_uvchr_to_utf8(pTHX_ U8 *d, UV uv)
2085 return Perl_uvuni_to_utf8_flags(aTHX_ d, NATIVE_TO_UNI(uv), 0);
2089 Perl_uvchr_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags)
2091 return Perl_uvuni_to_utf8_flags(aTHX_ d, NATIVE_TO_UNI(uv), flags);
2095 =for apidoc A|UV|utf8n_to_uvchr|U8 *s|STRLEN curlen|STRLEN *retlen|U32
2098 Returns the native character value of the first character in the string
2100 which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
2101 length, in bytes, of that character.
2103 Allows length and flags to be passed to low level routine.
2107 /* On ASCII machines this is normally a macro but we want
2108 a real function in case XS code wants it
2111 Perl_utf8n_to_uvchr(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen,
2114 const UV uv = Perl_utf8n_to_uvuni(aTHX_ s, curlen, retlen, flags);
2115 return UNI_TO_NATIVE(uv);
2119 =for apidoc A|char *|pv_uni_display|SV *dsv|U8 *spv|STRLEN len|STRLEN pvlim|UV flags
2121 Build to the scalar dsv a displayable version of the string spv,
2122 length len, the displayable version being at most pvlim bytes long
2123 (if longer, the rest is truncated and "..." will be appended).
2125 The flags argument can have UNI_DISPLAY_ISPRINT set to display
2126 isPRINT()able characters as themselves, UNI_DISPLAY_BACKSLASH
2127 to display the \\[nrfta\\] as the backslashed versions (like '\n')
2128 (UNI_DISPLAY_BACKSLASH is preferred over UNI_DISPLAY_ISPRINT for \\).
2129 UNI_DISPLAY_QQ (and its alias UNI_DISPLAY_REGEX) have both
2130 UNI_DISPLAY_BACKSLASH and UNI_DISPLAY_ISPRINT turned on.
2132 The pointer to the PV of the dsv is returned.
2136 Perl_pv_uni_display(pTHX_ SV *dsv, const U8 *spv, STRLEN len, STRLEN pvlim, UV flags)
2141 sv_setpvn(dsv, "", 0);
2142 for (s = (const char *)spv, e = s + len; s < e; s += UTF8SKIP(s)) {
2144 /* This serves double duty as a flag and a character to print after
2145 a \ when flags & UNI_DISPLAY_BACKSLASH is true.
2149 if (pvlim && SvCUR(dsv) >= pvlim) {
2153 u = utf8_to_uvchr((U8*)s, 0);
2155 const unsigned char c = (unsigned char)u & 0xFF;
2156 if (flags & UNI_DISPLAY_BACKSLASH) {
2173 Perl_sv_catpvf(aTHX_ dsv, "\\%c", ok);
2176 /* isPRINT() is the locale-blind version. */
2177 if (!ok && (flags & UNI_DISPLAY_ISPRINT) && isPRINT(c)) {
2178 Perl_sv_catpvf(aTHX_ dsv, "%c", c);
2183 Perl_sv_catpvf(aTHX_ dsv, "\\x{%"UVxf"}", u);
2186 sv_catpvs(dsv, "...");
2192 =for apidoc A|char *|sv_uni_display|SV *dsv|SV *ssv|STRLEN pvlim|UV flags
2194 Build to the scalar dsv a displayable version of the scalar sv,
2195 the displayable version being at most pvlim bytes long
2196 (if longer, the rest is truncated and "..." will be appended).
2198 The flags argument is as in pv_uni_display().
2200 The pointer to the PV of the dsv is returned.
2205 Perl_sv_uni_display(pTHX_ SV *dsv, SV *ssv, STRLEN pvlim, UV flags)
2207 return Perl_pv_uni_display(aTHX_ dsv, (const U8*)SvPVX_const(ssv),
2208 SvCUR(ssv), pvlim, flags);
2212 =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
2214 Return true if the strings s1 and s2 differ case-insensitively, false
2215 if not (if they are equal case-insensitively). If u1 is true, the
2216 string s1 is assumed to be in UTF-8-encoded Unicode. If u2 is true,
2217 the string s2 is assumed to be in UTF-8-encoded Unicode. If u1 or u2
2218 are false, the respective string is assumed to be in native 8-bit
2221 If the pe1 and pe2 are non-NULL, the scanning pointers will be copied
2222 in there (they will point at the beginning of the I<next> character).
2223 If the pointers behind pe1 or pe2 are non-NULL, they are the end
2224 pointers beyond which scanning will not continue under any
2225 circumstances. If the byte lengths l1 and l2 are non-zero, s1+l1 and
2226 s2+l2 will be used as goal end pointers that will also stop the scan,
2227 and which qualify towards defining a successful match: all the scans
2228 that define an explicit length must reach their goal pointers for
2229 a match to succeed).
2231 For case-insensitiveness, the "casefolding" of Unicode is used
2232 instead of upper/lowercasing both the characters, see
2233 http://www.unicode.org/unicode/reports/tr21/ (Case Mappings).
2237 Perl_ibcmp_utf8(pTHX_ const char *s1, char **pe1, register UV l1, bool u1, const char *s2, char **pe2, register UV l2, bool u2)
2240 register const U8 *p1 = (const U8*)s1;
2241 register const U8 *p2 = (const U8*)s2;
2242 register const U8 *f1 = NULL;
2243 register const U8 *f2 = NULL;
2244 register U8 *e1 = NULL;
2245 register U8 *q1 = NULL;
2246 register U8 *e2 = NULL;
2247 register U8 *q2 = NULL;
2248 STRLEN n1 = 0, n2 = 0;
2249 U8 foldbuf1[UTF8_MAXBYTES_CASE+1];
2250 U8 foldbuf2[UTF8_MAXBYTES_CASE+1];
2252 STRLEN foldlen1, foldlen2;
2257 if (e1 == 0 || (l1 && l1 < (UV)(e1 - (const U8*)s1)))
2258 f1 = (const U8*)s1 + l1;
2261 if (e2 == 0 || (l2 && l2 < (UV)(e2 - (const U8*)s2)))
2262 f2 = (const U8*)s2 + l2;
2264 if ((e1 == 0 && f1 == 0) || (e2 == 0 && f2 == 0) || (f1 == 0 && f2 == 0))
2265 return 1; /* mismatch; possible infinite loop or false positive */
2268 natbuf[1] = 0; /* Need to terminate the buffer. */
2270 while ((e1 == 0 || p1 < e1) &&
2271 (f1 == 0 || p1 < f1) &&
2272 (e2 == 0 || p2 < e2) &&
2273 (f2 == 0 || p2 < f2)) {
2276 to_utf8_fold(p1, foldbuf1, &foldlen1);
2278 uvuni_to_utf8(natbuf, (UV) NATIVE_TO_UNI(((UV)*p1)));
2279 to_utf8_fold(natbuf, foldbuf1, &foldlen1);
2286 to_utf8_fold(p2, foldbuf2, &foldlen2);
2288 uvuni_to_utf8(natbuf, (UV) NATIVE_TO_UNI(((UV)*p2)));
2289 to_utf8_fold(natbuf, foldbuf2, &foldlen2);
2295 if ( UTF8SKIP(q1) != UTF8SKIP(q2) ||
2296 (UTF8SKIP(q1) == 1 && *q1 != *q2) ||
2297 memNE((char*)q1, (char*)q2, UTF8SKIP(q1)) )
2298 return 1; /* mismatch */
2305 p1 += u1 ? UTF8SKIP(p1) : 1;
2307 p2 += u2 ? UTF8SKIP(p2) : 1;
2311 /* A match is defined by all the scans that specified
2312 * an explicit length reaching their final goals. */
2313 match = (f1 == 0 || p1 == f1) && (f2 == 0 || p2 == f2);
2322 return match ? 0 : 1; /* 0 match, 1 mismatch */
2327 * c-indentation-style: bsd
2329 * indent-tabs-mode: t
2332 * ex: set ts=8 sts=4 sw=4 noet: