3 * Copyright (C) 2000, 2001, 2002, 2003, 2004, 2005 by Larry Wall and
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)
405 const bool dowarn = ckWARN_d(WARN_UTF8);
406 const UV startbyte = *s;
407 STRLEN expectlen = 0;
410 /* This list is a superset of the UTF8_ALLOW_XXX. */
412 #define UTF8_WARN_EMPTY 1
413 #define UTF8_WARN_CONTINUATION 2
414 #define UTF8_WARN_NON_CONTINUATION 3
415 #define UTF8_WARN_FE_FF 4
416 #define UTF8_WARN_SHORT 5
417 #define UTF8_WARN_OVERFLOW 6
418 #define UTF8_WARN_SURROGATE 7
419 #define UTF8_WARN_LONG 8
420 #define UTF8_WARN_FFFF 9 /* Also FFFE. */
423 !(flags & UTF8_ALLOW_EMPTY)) {
424 warning = UTF8_WARN_EMPTY;
428 if (UTF8_IS_INVARIANT(uv)) {
431 return (UV) (NATIVE_TO_UTF(*s));
434 if (UTF8_IS_CONTINUATION(uv) &&
435 !(flags & UTF8_ALLOW_CONTINUATION)) {
436 warning = UTF8_WARN_CONTINUATION;
440 if (UTF8_IS_START(uv) && curlen > 1 && !UTF8_IS_CONTINUATION(s[1]) &&
441 !(flags & UTF8_ALLOW_NON_CONTINUATION)) {
442 warning = UTF8_WARN_NON_CONTINUATION;
447 uv = NATIVE_TO_UTF(uv);
449 if ((uv == 0xfe || uv == 0xff) &&
450 !(flags & UTF8_ALLOW_FE_FF)) {
451 warning = UTF8_WARN_FE_FF;
456 if (!(uv & 0x20)) { len = 2; uv &= 0x1f; }
457 else if (!(uv & 0x10)) { len = 3; uv &= 0x0f; }
458 else if (!(uv & 0x08)) { len = 4; uv &= 0x07; }
459 else if (!(uv & 0x04)) { len = 5; uv &= 0x03; }
461 else if (!(uv & 0x02)) { len = 6; uv &= 0x01; }
462 else { len = 7; uv &= 0x01; }
464 else if (!(uv & 0x02)) { len = 6; uv &= 0x01; }
465 else if (!(uv & 0x01)) { len = 7; uv = 0; }
466 else { len = 13; uv = 0; } /* whoa! */
474 if ((curlen < expectlen) &&
475 !(flags & UTF8_ALLOW_SHORT)) {
476 warning = UTF8_WARN_SHORT;
485 if (!UTF8_IS_CONTINUATION(*s) &&
486 !(flags & UTF8_ALLOW_NON_CONTINUATION)) {
488 warning = UTF8_WARN_NON_CONTINUATION;
492 uv = UTF8_ACCUMULATE(uv, *s);
494 /* These cannot be allowed. */
496 if (expectlen != 13 && !(flags & UTF8_ALLOW_LONG)) {
497 warning = UTF8_WARN_LONG;
501 else { /* uv < ouv */
502 /* This cannot be allowed. */
503 warning = UTF8_WARN_OVERFLOW;
511 if (UNICODE_IS_SURROGATE(uv) &&
512 !(flags & UTF8_ALLOW_SURROGATE)) {
513 warning = UTF8_WARN_SURROGATE;
515 } else if ((expectlen > (STRLEN)UNISKIP(uv)) &&
516 !(flags & UTF8_ALLOW_LONG)) {
517 warning = UTF8_WARN_LONG;
519 } else if (UNICODE_IS_ILLEGAL(uv) &&
520 !(flags & UTF8_ALLOW_FFFF)) {
521 warning = UTF8_WARN_FFFF;
529 if (flags & UTF8_CHECK_ONLY) {
536 SV* const sv = sv_2mortal(newSVpv("Malformed UTF-8 character ", 0));
539 case 0: /* Intentionally empty. */ break;
540 case UTF8_WARN_EMPTY:
541 Perl_sv_catpv(aTHX_ sv, "(empty string)");
543 case UTF8_WARN_CONTINUATION:
544 Perl_sv_catpvf(aTHX_ sv, "(unexpected continuation byte 0x%02"UVxf", with no preceding start byte)", uv);
546 case UTF8_WARN_NON_CONTINUATION:
548 Perl_sv_catpvf(aTHX_ sv, "(unexpected non-continuation byte 0x%02"UVxf", immediately after start byte 0x%02"UVxf")",
549 (UV)s[1], startbyte);
551 const int len = (int)(s-s0);
552 Perl_sv_catpvf(aTHX_ sv, "(unexpected non-continuation byte 0x%02"UVxf", %d byte%s after start byte 0x%02"UVxf", expected %d bytes)",
553 (UV)s[1], len, len > 1 ? "s" : "", startbyte, (int)expectlen);
557 case UTF8_WARN_FE_FF:
558 Perl_sv_catpvf(aTHX_ sv, "(byte 0x%02"UVxf")", uv);
560 case UTF8_WARN_SHORT:
561 Perl_sv_catpvf(aTHX_ sv, "(%d byte%s, need %d, after start byte 0x%02"UVxf")",
562 (int)curlen, curlen == 1 ? "" : "s", (int)expectlen, startbyte);
563 expectlen = curlen; /* distance for caller to skip */
565 case UTF8_WARN_OVERFLOW:
566 Perl_sv_catpvf(aTHX_ sv, "(overflow at 0x%"UVxf", byte 0x%02x, after start byte 0x%02"UVxf")",
569 case UTF8_WARN_SURROGATE:
570 Perl_sv_catpvf(aTHX_ sv, "(UTF-16 surrogate 0x%04"UVxf")", uv);
573 Perl_sv_catpvf(aTHX_ sv, "(%d byte%s, need %d, after start byte 0x%02"UVxf")",
574 (int)expectlen, expectlen == 1 ? "": "s", UNISKIP(uv), startbyte);
577 Perl_sv_catpvf(aTHX_ sv, "(character 0x%04"UVxf")", uv);
580 Perl_sv_catpv(aTHX_ sv, "(unknown reason)");
585 const char * const s = SvPVX_const(sv);
588 Perl_warner(aTHX_ packWARN(WARN_UTF8),
589 "%s in %s", s, OP_DESC(PL_op));
591 Perl_warner(aTHX_ packWARN(WARN_UTF8), "%s", s);
596 *retlen = expectlen ? expectlen : len;
602 =for apidoc A|UV|utf8_to_uvchr|const U8 *s|STRLEN *retlen
604 Returns the native character value of the first character in the string C<s>
605 which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
606 length, in bytes, of that character.
608 If C<s> does not point to a well-formed UTF-8 character, zero is
609 returned and retlen is set, if possible, to -1.
615 Perl_utf8_to_uvchr(pTHX_ const U8 *s, STRLEN *retlen)
617 return utf8n_to_uvchr(s, UTF8_MAXBYTES, retlen,
618 ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
622 =for apidoc A|UV|utf8_to_uvuni|const U8 *s|STRLEN *retlen
624 Returns the Unicode code point of the first character in the string C<s>
625 which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
626 length, in bytes, of that character.
628 This function should only be used when returned UV is considered
629 an index into the Unicode semantic tables (e.g. swashes).
631 If C<s> does not point to a well-formed UTF-8 character, zero is
632 returned and retlen is set, if possible, to -1.
638 Perl_utf8_to_uvuni(pTHX_ const U8 *s, STRLEN *retlen)
640 /* Call the low level routine asking for checks */
641 return Perl_utf8n_to_uvuni(aTHX_ s, UTF8_MAXBYTES, retlen,
642 ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
646 =for apidoc A|STRLEN|utf8_length|const U8 *s|const U8 *e
648 Return the length of the UTF-8 char encoded string C<s> in characters.
649 Stops at C<e> (inclusive). If C<e E<lt> s> or if the scan would end
650 up past C<e>, croaks.
656 Perl_utf8_length(pTHX_ const U8 *s, const U8 *e)
660 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g.
661 * the bitops (especially ~) can create illegal UTF-8.
662 * In other words: in Perl UTF-8 is not just for Unicode. */
665 goto warn_and_return;
667 const U8 t = UTF8SKIP(s);
670 if (ckWARN_d(WARN_UTF8)) {
672 Perl_warner(aTHX_ packWARN(WARN_UTF8),
673 "%s in %s", unees, OP_DESC(PL_op));
675 Perl_warner(aTHX_ packWARN(WARN_UTF8), unees);
687 =for apidoc A|IV|utf8_distance|const U8 *a|const U8 *b
689 Returns the number of UTF-8 characters between the UTF-8 pointers C<a>
692 WARNING: use only if you *know* that the pointers point inside the
699 Perl_utf8_distance(pTHX_ const U8 *a, const U8 *b)
703 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g.
704 * the bitops (especially ~) can create illegal UTF-8.
705 * In other words: in Perl UTF-8 is not just for Unicode. */
709 const U8 c = UTF8SKIP(a);
711 goto warn_and_return;
718 const U8 c = UTF8SKIP(b);
722 if (ckWARN_d(WARN_UTF8)) {
724 Perl_warner(aTHX_ packWARN(WARN_UTF8),
725 "%s in %s", unees, OP_DESC(PL_op));
727 Perl_warner(aTHX_ packWARN(WARN_UTF8), unees);
740 =for apidoc A|U8 *|utf8_hop|U8 *s|I32 off
742 Return the UTF-8 pointer C<s> displaced by C<off> characters, either
745 WARNING: do not use the following unless you *know* C<off> is within
746 the UTF-8 data pointed to by C<s> *and* that on entry C<s> is aligned
747 on the first byte of character or just after the last byte of a character.
753 Perl_utf8_hop(pTHX_ const U8 *s, I32 off)
755 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g
756 * the bitops (especially ~) can create illegal UTF-8.
757 * In other words: in Perl UTF-8 is not just for Unicode. */
766 while (UTF8_IS_CONTINUATION(*s))
774 =for apidoc A|U8 *|utf8_to_bytes|U8 *s|STRLEN *len
776 Converts a string C<s> of length C<len> from UTF-8 into byte encoding.
777 Unlike C<bytes_to_utf8>, this over-writes the original string, and
778 updates len to contain the new length.
779 Returns zero on failure, setting C<len> to -1.
785 Perl_utf8_to_bytes(pTHX_ U8 *s, STRLEN *len)
791 /* ensure valid UTF-8 and chars < 256 before updating string */
792 for (send = s + *len; s < send; ) {
795 if (!UTF8_IS_INVARIANT(c) &&
796 (!UTF8_IS_DOWNGRADEABLE_START(c) || (s >= send)
797 || !(c = *s++) || !UTF8_IS_CONTINUATION(c))) {
806 *d++ = (U8)utf8_to_uvchr(s, &ulen);
815 =for apidoc A|U8 *|bytes_from_utf8|const U8 *s|STRLEN *len|bool *is_utf8
817 Converts a string C<s> of length C<len> from UTF-8 into byte encoding.
818 Unlike C<utf8_to_bytes> but like C<bytes_to_utf8>, returns a pointer to
819 the newly-created string, and updates C<len> to contain the new
820 length. Returns the original string if no conversion occurs, C<len>
821 is unchanged. Do nothing if C<is_utf8> points to 0. Sets C<is_utf8> to
822 0 if C<s> is converted or contains all 7bit characters.
828 Perl_bytes_from_utf8(pTHX_ const U8 *s, STRLEN *len, bool *is_utf8)
838 /* ensure valid UTF-8 and chars < 256 before converting string */
839 for (send = s + *len; s < send;) {
841 if (!UTF8_IS_INVARIANT(c)) {
842 if (UTF8_IS_DOWNGRADEABLE_START(c) && s < send &&
843 (c = *s++) && UTF8_IS_CONTINUATION(c))
852 Newxz(d, (*len) - count + 1, U8);
853 s = start; start = d;
856 if (!UTF8_IS_INVARIANT(c)) {
857 /* Then it is two-byte encoded */
858 c = UTF8_ACCUMULATE(NATIVE_TO_UTF(c), *s++);
859 c = ASCII_TO_NATIVE(c);
869 =for apidoc A|U8 *|bytes_to_utf8|const U8 *s|STRLEN *len
871 Converts a string C<s> of length C<len> from ASCII into UTF-8 encoding.
872 Returns a pointer to the newly-created string, and sets C<len> to
873 reflect the new length.
875 If you want to convert to UTF-8 from other encodings than ASCII,
876 see sv_recode_to_utf8().
882 Perl_bytes_to_utf8(pTHX_ const U8 *s, STRLEN *len)
884 const U8 * const send = s + (*len);
888 Newxz(d, (*len) * 2 + 1, U8);
892 const UV uv = NATIVE_TO_ASCII(*s++);
893 if (UNI_IS_INVARIANT(uv))
894 *d++ = (U8)UTF_TO_NATIVE(uv);
896 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
897 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
906 * Convert native (big-endian) or reversed (little-endian) UTF-16 to UTF-8.
908 * Destination must be pre-extended to 3/2 source. Do not use in-place.
909 * We optimize for native, for obvious reasons. */
912 Perl_utf16_to_utf8(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen)
917 if (bytelen == 1 && p[0] == 0) { /* Be understanding. */
924 Perl_croak(aTHX_ "panic: utf16_to_utf8: odd bytelen %"UVf, (UV)bytelen);
929 UV uv = (p[0] << 8) + p[1]; /* UTF-16BE */
936 *d++ = (U8)(( uv >> 6) | 0xc0);
937 *d++ = (U8)(( uv & 0x3f) | 0x80);
940 if (uv >= 0xd800 && uv < 0xdbff) { /* surrogates */
941 UV low = (p[0] << 8) + p[1];
943 if (low < 0xdc00 || low >= 0xdfff)
944 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
945 uv = ((uv - 0xd800) << 10) + (low - 0xdc00) + 0x10000;
948 *d++ = (U8)(( uv >> 12) | 0xe0);
949 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
950 *d++ = (U8)(( uv & 0x3f) | 0x80);
954 *d++ = (U8)(( uv >> 18) | 0xf0);
955 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
956 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
957 *d++ = (U8)(( uv & 0x3f) | 0x80);
961 *newlen = d - dstart;
965 /* Note: this one is slightly destructive of the source. */
968 Perl_utf16_to_utf8_reversed(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen)
971 U8* send = s + bytelen;
978 return utf16_to_utf8(p, d, bytelen, newlen);
981 /* for now these are all defined (inefficiently) in terms of the utf8 versions */
984 Perl_is_uni_alnum(pTHX_ UV c)
986 U8 tmpbuf[UTF8_MAXBYTES+1];
987 uvchr_to_utf8(tmpbuf, c);
988 return is_utf8_alnum(tmpbuf);
992 Perl_is_uni_alnumc(pTHX_ UV c)
994 U8 tmpbuf[UTF8_MAXBYTES+1];
995 uvchr_to_utf8(tmpbuf, c);
996 return is_utf8_alnumc(tmpbuf);
1000 Perl_is_uni_idfirst(pTHX_ UV c)
1002 U8 tmpbuf[UTF8_MAXBYTES+1];
1003 uvchr_to_utf8(tmpbuf, c);
1004 return is_utf8_idfirst(tmpbuf);
1008 Perl_is_uni_alpha(pTHX_ UV c)
1010 U8 tmpbuf[UTF8_MAXBYTES+1];
1011 uvchr_to_utf8(tmpbuf, c);
1012 return is_utf8_alpha(tmpbuf);
1016 Perl_is_uni_ascii(pTHX_ UV c)
1018 U8 tmpbuf[UTF8_MAXBYTES+1];
1019 uvchr_to_utf8(tmpbuf, c);
1020 return is_utf8_ascii(tmpbuf);
1024 Perl_is_uni_space(pTHX_ UV c)
1026 U8 tmpbuf[UTF8_MAXBYTES+1];
1027 uvchr_to_utf8(tmpbuf, c);
1028 return is_utf8_space(tmpbuf);
1032 Perl_is_uni_digit(pTHX_ UV c)
1034 U8 tmpbuf[UTF8_MAXBYTES+1];
1035 uvchr_to_utf8(tmpbuf, c);
1036 return is_utf8_digit(tmpbuf);
1040 Perl_is_uni_upper(pTHX_ UV c)
1042 U8 tmpbuf[UTF8_MAXBYTES+1];
1043 uvchr_to_utf8(tmpbuf, c);
1044 return is_utf8_upper(tmpbuf);
1048 Perl_is_uni_lower(pTHX_ UV c)
1050 U8 tmpbuf[UTF8_MAXBYTES+1];
1051 uvchr_to_utf8(tmpbuf, c);
1052 return is_utf8_lower(tmpbuf);
1056 Perl_is_uni_cntrl(pTHX_ UV c)
1058 U8 tmpbuf[UTF8_MAXBYTES+1];
1059 uvchr_to_utf8(tmpbuf, c);
1060 return is_utf8_cntrl(tmpbuf);
1064 Perl_is_uni_graph(pTHX_ UV c)
1066 U8 tmpbuf[UTF8_MAXBYTES+1];
1067 uvchr_to_utf8(tmpbuf, c);
1068 return is_utf8_graph(tmpbuf);
1072 Perl_is_uni_print(pTHX_ UV c)
1074 U8 tmpbuf[UTF8_MAXBYTES+1];
1075 uvchr_to_utf8(tmpbuf, c);
1076 return is_utf8_print(tmpbuf);
1080 Perl_is_uni_punct(pTHX_ UV c)
1082 U8 tmpbuf[UTF8_MAXBYTES+1];
1083 uvchr_to_utf8(tmpbuf, c);
1084 return is_utf8_punct(tmpbuf);
1088 Perl_is_uni_xdigit(pTHX_ UV c)
1090 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1091 uvchr_to_utf8(tmpbuf, c);
1092 return is_utf8_xdigit(tmpbuf);
1096 Perl_to_uni_upper(pTHX_ UV c, U8* p, STRLEN *lenp)
1098 uvchr_to_utf8(p, c);
1099 return to_utf8_upper(p, p, lenp);
1103 Perl_to_uni_title(pTHX_ UV c, U8* p, STRLEN *lenp)
1105 uvchr_to_utf8(p, c);
1106 return to_utf8_title(p, p, lenp);
1110 Perl_to_uni_lower(pTHX_ UV c, U8* p, STRLEN *lenp)
1112 uvchr_to_utf8(p, c);
1113 return to_utf8_lower(p, p, lenp);
1117 Perl_to_uni_fold(pTHX_ UV c, U8* p, STRLEN *lenp)
1119 uvchr_to_utf8(p, c);
1120 return to_utf8_fold(p, p, lenp);
1123 /* for now these all assume no locale info available for Unicode > 255 */
1126 Perl_is_uni_alnum_lc(pTHX_ UV c)
1128 return is_uni_alnum(c); /* XXX no locale support yet */
1132 Perl_is_uni_alnumc_lc(pTHX_ UV c)
1134 return is_uni_alnumc(c); /* XXX no locale support yet */
1138 Perl_is_uni_idfirst_lc(pTHX_ UV c)
1140 return is_uni_idfirst(c); /* XXX no locale support yet */
1144 Perl_is_uni_alpha_lc(pTHX_ UV c)
1146 return is_uni_alpha(c); /* XXX no locale support yet */
1150 Perl_is_uni_ascii_lc(pTHX_ UV c)
1152 return is_uni_ascii(c); /* XXX no locale support yet */
1156 Perl_is_uni_space_lc(pTHX_ UV c)
1158 return is_uni_space(c); /* XXX no locale support yet */
1162 Perl_is_uni_digit_lc(pTHX_ UV c)
1164 return is_uni_digit(c); /* XXX no locale support yet */
1168 Perl_is_uni_upper_lc(pTHX_ UV c)
1170 return is_uni_upper(c); /* XXX no locale support yet */
1174 Perl_is_uni_lower_lc(pTHX_ UV c)
1176 return is_uni_lower(c); /* XXX no locale support yet */
1180 Perl_is_uni_cntrl_lc(pTHX_ UV c)
1182 return is_uni_cntrl(c); /* XXX no locale support yet */
1186 Perl_is_uni_graph_lc(pTHX_ UV c)
1188 return is_uni_graph(c); /* XXX no locale support yet */
1192 Perl_is_uni_print_lc(pTHX_ UV c)
1194 return is_uni_print(c); /* XXX no locale support yet */
1198 Perl_is_uni_punct_lc(pTHX_ UV c)
1200 return is_uni_punct(c); /* XXX no locale support yet */
1204 Perl_is_uni_xdigit_lc(pTHX_ UV c)
1206 return is_uni_xdigit(c); /* XXX no locale support yet */
1210 Perl_to_uni_upper_lc(pTHX_ U32 c)
1212 /* XXX returns only the first character -- do not use XXX */
1213 /* XXX no locale support yet */
1215 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1216 return (U32)to_uni_upper(c, tmpbuf, &len);
1220 Perl_to_uni_title_lc(pTHX_ U32 c)
1222 /* XXX returns only the first character XXX -- do not use XXX */
1223 /* XXX no locale support yet */
1225 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1226 return (U32)to_uni_title(c, tmpbuf, &len);
1230 Perl_to_uni_lower_lc(pTHX_ U32 c)
1232 /* XXX returns only the first character -- do not use XXX */
1233 /* XXX no locale support yet */
1235 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1236 return (U32)to_uni_lower(c, tmpbuf, &len);
1240 S_is_utf8_common(pTHX_ const U8 *const p, SV **swash,
1241 const char *const swashname)
1243 if (!is_utf8_char(p))
1246 *swash = swash_init("utf8", swashname, &PL_sv_undef, 1, 0);
1247 return swash_fetch(*swash, p, TRUE) != 0;
1251 Perl_is_utf8_alnum(pTHX_ const U8 *p)
1253 /* NOTE: "IsWord", not "IsAlnum", since Alnum is a true
1254 * descendant of isalnum(3), in other words, it doesn't
1255 * contain the '_'. --jhi */
1256 return S_is_utf8_common(aTHX_ p, &PL_utf8_alnum, "IsWord");
1260 Perl_is_utf8_alnumc(pTHX_ const U8 *p)
1262 return S_is_utf8_common(aTHX_ p, &PL_utf8_alnumc, "IsAlnumC");
1266 Perl_is_utf8_idfirst(pTHX_ const U8 *p) /* The naming is historical. */
1270 /* is_utf8_idstart would be more logical. */
1271 return S_is_utf8_common(aTHX_ p, &PL_utf8_idstart, "IdStart");
1275 Perl_is_utf8_idcont(pTHX_ const U8 *p)
1279 return S_is_utf8_common(aTHX_ p, &PL_utf8_idcont, "IdContinue");
1283 Perl_is_utf8_alpha(pTHX_ const U8 *p)
1285 return S_is_utf8_common(aTHX_ p, &PL_utf8_alpha, "IsAlpha");
1289 Perl_is_utf8_ascii(pTHX_ const U8 *p)
1291 return S_is_utf8_common(aTHX_ p, &PL_utf8_ascii, "IsAscii");
1295 Perl_is_utf8_space(pTHX_ const U8 *p)
1297 return S_is_utf8_common(aTHX_ p, &PL_utf8_space, "IsSpacePerl");
1301 Perl_is_utf8_digit(pTHX_ const U8 *p)
1303 return S_is_utf8_common(aTHX_ p, &PL_utf8_digit, "IsDigit");
1307 Perl_is_utf8_upper(pTHX_ const U8 *p)
1309 return S_is_utf8_common(aTHX_ p, &PL_utf8_upper, "IsUppercase");
1313 Perl_is_utf8_lower(pTHX_ const U8 *p)
1315 return S_is_utf8_common(aTHX_ p, &PL_utf8_lower, "IsLowercase");
1319 Perl_is_utf8_cntrl(pTHX_ const U8 *p)
1321 return S_is_utf8_common(aTHX_ p, &PL_utf8_cntrl, "IsCntrl");
1325 Perl_is_utf8_graph(pTHX_ const U8 *p)
1327 return S_is_utf8_common(aTHX_ p, &PL_utf8_graph, "IsGraph");
1331 Perl_is_utf8_print(pTHX_ const U8 *p)
1333 return S_is_utf8_common(aTHX_ p, &PL_utf8_print, "IsPrint");
1337 Perl_is_utf8_punct(pTHX_ const U8 *p)
1339 return S_is_utf8_common(aTHX_ p, &PL_utf8_punct, "IsPunct");
1343 Perl_is_utf8_xdigit(pTHX_ const U8 *p)
1345 return S_is_utf8_common(aTHX_ p, &PL_utf8_xdigit, "Isxdigit");
1349 Perl_is_utf8_mark(pTHX_ const U8 *p)
1351 return S_is_utf8_common(aTHX_ p, &PL_utf8_mark, "IsM");
1355 =for apidoc A|UV|to_utf8_case|U8 *p|U8* ustrp|STRLEN *lenp|SV **swash|char *normal|char *special
1357 The "p" contains the pointer to the UTF-8 string encoding
1358 the character that is being converted.
1360 The "ustrp" is a pointer to the character buffer to put the
1361 conversion result to. The "lenp" is a pointer to the length
1364 The "swashp" is a pointer to the swash to use.
1366 Both the special and normal mappings are stored lib/unicore/To/Foo.pl,
1367 and loaded by SWASHNEW, using lib/utf8_heavy.pl. The special (usually,
1368 but not always, a multicharacter mapping), is tried first.
1370 The "special" is a string like "utf8::ToSpecLower", which means the
1371 hash %utf8::ToSpecLower. The access to the hash is through
1372 Perl_to_utf8_case().
1374 The "normal" is a string like "ToLower" which means the swash
1380 Perl_to_utf8_case(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp,
1381 SV **swashp, const char *normal, const char *special)
1383 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1386 const UV uv0 = utf8_to_uvchr(p, NULL);
1387 /* The NATIVE_TO_UNI() and UNI_TO_NATIVE() mappings
1388 * are necessary in EBCDIC, they are redundant no-ops
1389 * in ASCII-ish platforms, and hopefully optimized away. */
1390 const UV uv1 = NATIVE_TO_UNI(uv0);
1391 uvuni_to_utf8(tmpbuf, uv1);
1393 if (!*swashp) /* load on-demand */
1394 *swashp = swash_init("utf8", normal, &PL_sv_undef, 4, 0);
1396 /* The 0xDF is the only special casing Unicode code point below 0x100. */
1397 if (special && (uv1 == 0xDF || uv1 > 0xFF)) {
1398 /* It might be "special" (sometimes, but not always,
1399 * a multicharacter mapping) */
1403 if ((hv = get_hv(special, FALSE)) &&
1404 (svp = hv_fetch(hv, (const char*)tmpbuf, UNISKIP(uv1), FALSE)) &&
1408 s = SvPV_const(*svp, len);
1410 len = uvuni_to_utf8(ustrp, NATIVE_TO_UNI(*(U8*)s)) - ustrp;
1413 /* If we have EBCDIC we need to remap the characters
1414 * since any characters in the low 256 are Unicode
1415 * code points, not EBCDIC. */
1416 U8 *t = (U8*)s, *tend = t + len, *d;
1423 UV c = utf8_to_uvchr(t, &tlen);
1425 d = uvchr_to_utf8(d, UNI_TO_NATIVE(c));
1434 d = uvchr_to_utf8(d, UNI_TO_NATIVE(*t));
1439 Copy(tmpbuf, ustrp, len, U8);
1441 Copy(s, ustrp, len, U8);
1447 if (!len && *swashp) {
1448 UV uv2 = swash_fetch(*swashp, tmpbuf, TRUE);
1451 /* It was "normal" (a single character mapping). */
1452 UV uv3 = UNI_TO_NATIVE(uv2);
1454 len = uvchr_to_utf8(ustrp, uv3) - ustrp;
1458 if (!len) /* Neither: just copy. */
1459 len = uvchr_to_utf8(ustrp, uv0) - ustrp;
1464 return len ? utf8_to_uvchr(ustrp, 0) : 0;
1468 =for apidoc A|UV|to_utf8_upper|const U8 *p|U8 *ustrp|STRLEN *lenp
1470 Convert the UTF-8 encoded character at p to its uppercase version and
1471 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
1472 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since
1473 the uppercase version may be longer than the original character.
1475 The first character of the uppercased version is returned
1476 (but note, as explained above, that there may be more.)
1481 Perl_to_utf8_upper(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
1483 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
1484 &PL_utf8_toupper, "ToUpper", "utf8::ToSpecUpper");
1488 =for apidoc A|UV|to_utf8_title|const U8 *p|U8 *ustrp|STRLEN *lenp
1490 Convert the UTF-8 encoded character at p to its titlecase version and
1491 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
1492 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
1493 titlecase version may be longer than the original character.
1495 The first character of the titlecased version is returned
1496 (but note, as explained above, that there may be more.)
1501 Perl_to_utf8_title(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
1503 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
1504 &PL_utf8_totitle, "ToTitle", "utf8::ToSpecTitle");
1508 =for apidoc A|UV|to_utf8_lower|const U8 *p|U8 *ustrp|STRLEN *lenp
1510 Convert the UTF-8 encoded character at p to its lowercase version and
1511 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
1512 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
1513 lowercase version may be longer than the original character.
1515 The first character of the lowercased version is returned
1516 (but note, as explained above, that there may be more.)
1521 Perl_to_utf8_lower(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
1523 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
1524 &PL_utf8_tolower, "ToLower", "utf8::ToSpecLower");
1528 =for apidoc A|UV|to_utf8_fold|const U8 *p|U8 *ustrp|STRLEN *lenp
1530 Convert the UTF-8 encoded character at p to its foldcase version and
1531 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
1532 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
1533 foldcase version may be longer than the original character (up to
1536 The first character of the foldcased version is returned
1537 (but note, as explained above, that there may be more.)
1542 Perl_to_utf8_fold(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
1544 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
1545 &PL_utf8_tofold, "ToFold", "utf8::ToSpecFold");
1549 * A "swash" is a swatch hash.
1550 * A "swatch" is a bit vector generated by utf8.c:S_swash_get().
1551 * C<pkg> is a pointer to a package name for SWASHNEW, should be "utf8".
1552 * For other parameters, see utf8::SWASHNEW in lib/utf8_heavy.pl.
1555 Perl_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv, I32 minbits, I32 none)
1559 SV* const tokenbufsv = sv_newmortal();
1561 const size_t pkg_len = strlen(pkg);
1562 const size_t name_len = strlen(name);
1563 HV * const stash = gv_stashpvn(pkg, pkg_len, FALSE);
1566 PUSHSTACKi(PERLSI_MAGIC);
1571 if (!gv_fetchmeth(stash, "SWASHNEW", 8, -1)) { /* demand load utf8 */
1573 errsv_save = newSVsv(ERRSV);
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 PL_curcop->op_private = (U8)(PL_hints & HINT_PRIVATE_MASK);
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. */
1630 Perl_swash_fetch(pTHX_ SV *sv, const U8 *ptr, bool do_utf8)
1633 HV* const hv = (HV*)SvRV(sv);
1638 const U8 *tmps = NULL;
1642 UV c = NATIVE_TO_ASCII(*ptr);
1644 if (!do_utf8 && !UNI_IS_INVARIANT(c)) {
1645 tmputf8[0] = (U8)UTF8_EIGHT_BIT_HI(c);
1646 tmputf8[1] = (U8)UTF8_EIGHT_BIT_LO(c);
1649 /* Given a UTF-X encoded char 0xAA..0xYY,0xZZ
1650 * then the "swatch" is a vec() for al the chars which start
1652 * So the key in the hash (klen) is length of encoded char -1
1654 klen = UTF8SKIP(ptr) - 1;
1658 /* If char in invariant then swatch is for all the invariant chars
1659 * In both UTF-8 and UTF-8-MOD that happens to be UTF_CONTINUATION_MARK
1661 needents = UTF_CONTINUATION_MARK;
1662 off = NATIVE_TO_UTF(ptr[klen]);
1665 /* If char is encoded then swatch is for the prefix */
1666 needents = (1 << UTF_ACCUMULATION_SHIFT);
1667 off = NATIVE_TO_UTF(ptr[klen]) & UTF_CONTINUATION_MASK;
1671 * This single-entry cache saves about 1/3 of the utf8 overhead in test
1672 * suite. (That is, only 7-8% overall over just a hash cache. Still,
1673 * it's nothing to sniff at.) Pity we usually come through at least
1674 * two function calls to get here...
1676 * NB: this code assumes that swatches are never modified, once generated!
1679 if (hv == PL_last_swash_hv &&
1680 klen == PL_last_swash_klen &&
1681 (!klen || memEQ((char *)ptr, (char *)PL_last_swash_key, klen)) )
1683 tmps = PL_last_swash_tmps;
1684 slen = PL_last_swash_slen;
1687 /* Try our second-level swatch cache, kept in a hash. */
1688 SV** svp = hv_fetch(hv, (const char*)ptr, klen, FALSE);
1690 /* If not cached, generate it via swash_get */
1691 if (!svp || !SvPOK(*svp)
1692 || !(tmps = (const U8*)SvPV_const(*svp, slen))) {
1693 /* We use utf8n_to_uvuni() as we want an index into
1694 Unicode tables, not a native character number.
1696 const UV code_point = utf8n_to_uvuni(ptr, UTF8_MAXBYTES, 0,
1698 0 : UTF8_ALLOW_ANY);
1699 swatch = swash_get(sv,
1700 /* On EBCDIC & ~(0xA0-1) isn't a useful thing to do */
1701 (klen) ? (code_point & ~(needents - 1)) : 0,
1704 if (IN_PERL_COMPILETIME)
1705 PL_curcop->op_private = (U8)(PL_hints & HINT_PRIVATE_MASK);
1707 svp = hv_store(hv, (const char *)ptr, klen, swatch, 0);
1709 if (!svp || !(tmps = (U8*)SvPV(*svp, slen))
1710 || (slen << 3) < needents)
1711 Perl_croak(aTHX_ "panic: swash_fetch got improper swatch");
1714 PL_last_swash_hv = hv;
1715 PL_last_swash_klen = klen;
1716 /* FIXME change interpvar.h? */
1717 PL_last_swash_tmps = (U8 *) tmps;
1718 PL_last_swash_slen = slen;
1720 Copy(ptr, PL_last_swash_key, klen, U8);
1723 switch ((int)((slen << 3) / needents)) {
1725 bit = 1 << (off & 7);
1727 return (tmps[off] & bit) != 0;
1732 return (tmps[off] << 8) + tmps[off + 1] ;
1735 return (tmps[off] << 24) + (tmps[off+1] << 16) + (tmps[off+2] << 8) + tmps[off + 3] ;
1737 Perl_croak(aTHX_ "panic: swash_fetch got swatch of unexpected bit width");
1742 * Returns a swatch (a bit vector string) for a code point sequence
1743 * that starts from the value C<start> and comprises the number C<span>.
1744 * A C<swash> must be an object created by SWASHNEW (see lib/utf8_heavy.pl).
1745 * Should be used via swash_fetch, which will cache the swatch in C<swash>.
1748 S_swash_get(pTHX_ SV* swash, UV start, UV span)
1751 U8 *l, *lend, *x, *xend, *s;
1752 STRLEN lcur, xcur, scur;
1754 HV* const hv = (HV*)SvRV(swash);
1755 SV** listsvp = hv_fetch(hv, "LIST", 4, FALSE);
1756 SV** typesvp = hv_fetch(hv, "TYPE", 4, FALSE);
1757 SV** bitssvp = hv_fetch(hv, "BITS", 4, FALSE);
1758 SV** nonesvp = hv_fetch(hv, "NONE", 4, FALSE);
1759 SV** extssvp = hv_fetch(hv, "EXTRAS", 6, FALSE);
1760 U8* typestr = (U8*)SvPV_nolen(*typesvp);
1761 int typeto = typestr[0] == 'T' && typestr[1] == 'o';
1762 STRLEN bits = SvUV(*bitssvp);
1763 STRLEN octets = bits >> 3; /* if bits == 1, then octets == 0 */
1764 UV none = SvUV(*nonesvp);
1765 UV end = start + span;
1767 if (bits != 1 && bits != 8 && bits != 16 && bits != 32) {
1768 Perl_croak(aTHX_ "panic: swash_get doesn't expect bits %"UVuf,
1772 /* create and initialize $swatch */
1773 swatch = newSVpvn("",0);
1774 scur = octets ? (span * octets) : (span + 7) / 8;
1775 SvGROW(swatch, scur + 1);
1776 s = (U8*)SvPVX(swatch);
1777 if (octets && none) {
1778 const U8* e = s + scur;
1781 *s++ = (U8)(none & 0xff);
1782 else if (bits == 16) {
1783 *s++ = (U8)((none >> 8) & 0xff);
1784 *s++ = (U8)( none & 0xff);
1786 else if (bits == 32) {
1787 *s++ = (U8)((none >> 24) & 0xff);
1788 *s++ = (U8)((none >> 16) & 0xff);
1789 *s++ = (U8)((none >> 8) & 0xff);
1790 *s++ = (U8)( none & 0xff);
1796 (void)memzero((U8*)s, scur + 1);
1798 SvCUR_set(swatch, scur);
1799 s = (U8*)SvPVX(swatch);
1801 /* read $swash->{LIST} */
1802 l = (U8*)SvPV(*listsvp, lcur);
1805 UV min, max, val, key;
1807 I32 flags = PERL_SCAN_SILENT_ILLDIGIT | PERL_SCAN_DISALLOW_PREFIX;
1809 U8* nl = (U8*)memchr(l, '\n', lend - l);
1812 min = grok_hex((char *)l, &numlen, &flags, NULL);
1816 l = nl + 1; /* 1 is length of "\n" */
1820 l = lend; /* to LIST's end at which \n is not found */
1826 flags = PERL_SCAN_SILENT_ILLDIGIT | PERL_SCAN_DISALLOW_PREFIX;
1828 max = grok_hex((char *)l, &numlen, &flags, NULL);
1837 flags = PERL_SCAN_SILENT_ILLDIGIT |
1838 PERL_SCAN_DISALLOW_PREFIX;
1840 val = grok_hex((char *)l, &numlen, &flags, NULL);
1849 Perl_croak(aTHX_ "%s: illegal mapping '%s'",
1855 val = 0; /* bits == 1, then val should be ignored */
1862 Perl_croak(aTHX_ "%s: illegal mapping '%s'", typestr, l);
1866 val = 0; /* bits == 1, then val should be ignored */
1879 if (!none || val < none) {
1884 for (key = min; key <= max; key++) {
1888 /* offset must be non-negative (start <= min <= key < end) */
1889 offset = octets * (key - start);
1891 s[offset] = (U8)(val & 0xff);
1892 else if (bits == 16) {
1893 s[offset ] = (U8)((val >> 8) & 0xff);
1894 s[offset + 1] = (U8)( val & 0xff);
1896 else if (bits == 32) {
1897 s[offset ] = (U8)((val >> 24) & 0xff);
1898 s[offset + 1] = (U8)((val >> 16) & 0xff);
1899 s[offset + 2] = (U8)((val >> 8) & 0xff);
1900 s[offset + 3] = (U8)( val & 0xff);
1903 if (!none || val < none)
1907 else { /* bits == 1, then val should be ignored */
1910 for (key = min; key <= max; key++) {
1911 STRLEN offset = (STRLEN)(key - start);
1914 s[offset >> 3] |= 1 << (offset & 7);
1920 /* read $swash->{EXTRAS} */
1921 x = (U8*)SvPV(*extssvp, xcur);
1929 SV **otherbitssvp, *other;
1937 nl = (U8*)memchr(x, '\n', xend - x);
1939 if (opc != '-' && opc != '+' && opc != '!' && opc != '&') {
1941 x = nl + 1; /* 1 is length of "\n" */
1945 x = xend; /* to EXTRAS' end at which \n is not found */
1952 namelen = nl - namestr;
1956 namelen = xend - namestr;
1960 othersvp = hv_fetch(hv, (char *)namestr, namelen, FALSE);
1961 otherhv = (HV*)SvRV(*othersvp);
1962 otherbitssvp = hv_fetch(otherhv, "BITS", 4, FALSE);
1963 otherbits = (STRLEN)SvUV(*otherbitssvp);
1964 if (bits < otherbits)
1965 Perl_croak(aTHX_ "panic: swash_get found swatch size mismatch");
1967 /* The "other" swatch must be destroyed after. */
1968 other = swash_get(*othersvp, start, span);
1969 o = (U8*)SvPV(other, olen);
1972 Perl_croak(aTHX_ "panic: swash_get got improper swatch");
1974 s = (U8*)SvPV(swatch, slen);
1975 if (bits == 1 && otherbits == 1) {
1977 Perl_croak(aTHX_ "panic: swash_get found swatch length mismatch");
2001 STRLEN otheroctets = otherbits >> 3;
2003 U8* send = s + slen;
2008 if (otherbits == 1) {
2009 otherval = (o[offset >> 3] >> (offset & 7)) & 1;
2013 STRLEN vlen = otheroctets;
2021 if (opc == '+' && otherval)
2022 ; /* replace with otherval */
2023 else if (opc == '!' && !otherval)
2025 else if (opc == '-' && otherval)
2027 else if (opc == '&' && !otherval)
2030 s += octets; /* no replacement */
2035 *s++ = (U8)( otherval & 0xff);
2036 else if (bits == 16) {
2037 *s++ = (U8)((otherval >> 8) & 0xff);
2038 *s++ = (U8)( otherval & 0xff);
2040 else if (bits == 32) {
2041 *s++ = (U8)((otherval >> 24) & 0xff);
2042 *s++ = (U8)((otherval >> 16) & 0xff);
2043 *s++ = (U8)((otherval >> 8) & 0xff);
2044 *s++ = (U8)( otherval & 0xff);
2048 sv_free(other); /* through with it! */
2054 =for apidoc A|U8 *|uvchr_to_utf8|U8 *d|UV uv
2056 Adds the UTF-8 representation of the Native codepoint C<uv> to the end
2057 of the string C<d>; C<d> should be have at least C<UTF8_MAXBYTES+1> free
2058 bytes available. The return value is the pointer to the byte after the
2059 end of the new character. In other words,
2061 d = uvchr_to_utf8(d, uv);
2063 is the recommended wide native character-aware way of saying
2070 /* On ASCII machines this is normally a macro but we want a
2071 real function in case XS code wants it
2074 Perl_uvchr_to_utf8(pTHX_ U8 *d, UV uv)
2076 return Perl_uvuni_to_utf8_flags(aTHX_ d, NATIVE_TO_UNI(uv), 0);
2080 Perl_uvchr_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags)
2082 return Perl_uvuni_to_utf8_flags(aTHX_ d, NATIVE_TO_UNI(uv), flags);
2086 =for apidoc A|UV|utf8n_to_uvchr|U8 *s|STRLEN curlen|STRLEN *retlen|U32
2089 Returns the native character value of the first character in the string
2091 which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
2092 length, in bytes, of that character.
2094 Allows length and flags to be passed to low level routine.
2098 /* On ASCII machines this is normally a macro but we want
2099 a real function in case XS code wants it
2102 Perl_utf8n_to_uvchr(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen,
2105 const UV uv = Perl_utf8n_to_uvuni(aTHX_ s, curlen, retlen, flags);
2106 return UNI_TO_NATIVE(uv);
2110 =for apidoc A|char *|pv_uni_display|SV *dsv|U8 *spv|STRLEN len|STRLEN pvlim|UV flags
2112 Build to the scalar dsv a displayable version of the string spv,
2113 length len, the displayable version being at most pvlim bytes long
2114 (if longer, the rest is truncated and "..." will be appended).
2116 The flags argument can have UNI_DISPLAY_ISPRINT set to display
2117 isPRINT()able characters as themselves, UNI_DISPLAY_BACKSLASH
2118 to display the \\[nrfta\\] as the backslashed versions (like '\n')
2119 (UNI_DISPLAY_BACKSLASH is preferred over UNI_DISPLAY_ISPRINT for \\).
2120 UNI_DISPLAY_QQ (and its alias UNI_DISPLAY_REGEX) have both
2121 UNI_DISPLAY_BACKSLASH and UNI_DISPLAY_ISPRINT turned on.
2123 The pointer to the PV of the dsv is returned.
2127 Perl_pv_uni_display(pTHX_ SV *dsv, const U8 *spv, STRLEN len, STRLEN pvlim, UV flags)
2132 sv_setpvn(dsv, "", 0);
2133 for (s = (const char *)spv, e = s + len; s < e; s += UTF8SKIP(s)) {
2135 /* This serves double duty as a flag and a character to print after
2136 a \ when flags & UNI_DISPLAY_BACKSLASH is true.
2140 if (pvlim && SvCUR(dsv) >= pvlim) {
2144 u = utf8_to_uvchr((U8*)s, 0);
2146 const unsigned char c = (unsigned char)u & 0xFF;
2147 if (!ok && (flags & UNI_DISPLAY_BACKSLASH)) {
2164 Perl_sv_catpvf(aTHX_ dsv, "\\%c", ok);
2167 /* isPRINT() is the locale-blind version. */
2168 if (!ok && (flags & UNI_DISPLAY_ISPRINT) && isPRINT(c)) {
2169 Perl_sv_catpvf(aTHX_ dsv, "%c", c);
2174 Perl_sv_catpvf(aTHX_ dsv, "\\x{%"UVxf"}", u);
2177 sv_catpvn(dsv, "...", 3);
2183 =for apidoc A|char *|sv_uni_display|SV *dsv|SV *ssv|STRLEN pvlim|UV flags
2185 Build to the scalar dsv a displayable version of the scalar sv,
2186 the displayable version being at most pvlim bytes long
2187 (if longer, the rest is truncated and "..." will be appended).
2189 The flags argument is as in pv_uni_display().
2191 The pointer to the PV of the dsv is returned.
2195 Perl_sv_uni_display(pTHX_ SV *dsv, SV *ssv, STRLEN pvlim, UV flags)
2197 return Perl_pv_uni_display(aTHX_ dsv, (const U8*)SvPVX_const(ssv),
2198 SvCUR(ssv), pvlim, flags);
2202 =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
2204 Return true if the strings s1 and s2 differ case-insensitively, false
2205 if not (if they are equal case-insensitively). If u1 is true, the
2206 string s1 is assumed to be in UTF-8-encoded Unicode. If u2 is true,
2207 the string s2 is assumed to be in UTF-8-encoded Unicode. If u1 or u2
2208 are false, the respective string is assumed to be in native 8-bit
2211 If the pe1 and pe2 are non-NULL, the scanning pointers will be copied
2212 in there (they will point at the beginning of the I<next> character).
2213 If the pointers behind pe1 or pe2 are non-NULL, they are the end
2214 pointers beyond which scanning will not continue under any
2215 circumstances. If the byte lengths l1 and l2 are non-zero, s1+l1 and
2216 s2+l2 will be used as goal end pointers that will also stop the scan,
2217 and which qualify towards defining a successful match: all the scans
2218 that define an explicit length must reach their goal pointers for
2219 a match to succeed).
2221 For case-insensitiveness, the "casefolding" of Unicode is used
2222 instead of upper/lowercasing both the characters, see
2223 http://www.unicode.org/unicode/reports/tr21/ (Case Mappings).
2227 Perl_ibcmp_utf8(pTHX_ const char *s1, char **pe1, register UV l1, bool u1, const char *s2, char **pe2, register UV l2, bool u2)
2229 register const U8 *p1 = (const U8*)s1;
2230 register const U8 *p2 = (const U8*)s2;
2231 register const U8 *f1 = 0, *f2 = 0;
2232 register U8 *e1 = 0, *q1 = 0;
2233 register U8 *e2 = 0, *q2 = 0;
2234 STRLEN n1 = 0, n2 = 0;
2235 U8 foldbuf1[UTF8_MAXBYTES_CASE+1];
2236 U8 foldbuf2[UTF8_MAXBYTES_CASE+1];
2238 STRLEN foldlen1, foldlen2;
2243 if (e1 == 0 || (l1 && l1 < (UV)(e1 - (const U8*)s1)))
2244 f1 = (const U8*)s1 + l1;
2247 if (e2 == 0 || (l2 && l2 < (UV)(e2 - (const U8*)s2)))
2248 f2 = (const U8*)s2 + l2;
2250 if ((e1 == 0 && f1 == 0) || (e2 == 0 && f2 == 0) || (f1 == 0 && f2 == 0))
2251 return 1; /* mismatch; possible infinite loop or false positive */
2254 natbuf[1] = 0; /* Need to terminate the buffer. */
2256 while ((e1 == 0 || p1 < e1) &&
2257 (f1 == 0 || p1 < f1) &&
2258 (e2 == 0 || p2 < e2) &&
2259 (f2 == 0 || p2 < f2)) {
2262 to_utf8_fold(p1, foldbuf1, &foldlen1);
2264 uvuni_to_utf8(natbuf, (UV) NATIVE_TO_UNI(((UV)*p1)));
2265 to_utf8_fold(natbuf, foldbuf1, &foldlen1);
2272 to_utf8_fold(p2, foldbuf2, &foldlen2);
2274 uvuni_to_utf8(natbuf, (UV) NATIVE_TO_UNI(((UV)*p2)));
2275 to_utf8_fold(natbuf, foldbuf2, &foldlen2);
2281 if ( UTF8SKIP(q1) != UTF8SKIP(q2) ||
2282 (UTF8SKIP(q1) == 1 && *q1 != *q2) ||
2283 memNE((char*)q1, (char*)q2, UTF8SKIP(q1)) )
2284 return 1; /* mismatch */
2291 p1 += u1 ? UTF8SKIP(p1) : 1;
2293 p2 += u2 ? UTF8SKIP(p2) : 1;
2297 /* A match is defined by all the scans that specified
2298 * an explicit length reaching their final goals. */
2299 match = (f1 == 0 || p1 == f1) && (f2 == 0 || p2 == f2);
2308 return match ? 0 : 1; /* 0 match, 1 mismatch */
2313 * c-indentation-style: bsd
2315 * indent-tabs-mode: t
2318 * ex: set ts=8 sts=4 sw=4 noet: