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)
270 len = strlen((const char *)s);
275 /* Inline the easy bits of is_utf8_char() here for speed... */
276 if (UTF8_IS_INVARIANT(*x))
278 else if (!UTF8_IS_START(*x))
281 /* ... and call is_utf8_char() only if really needed. */
284 if (IS_UTF8_CHAR_FAST(c)) {
285 if (!IS_UTF8_CHAR(x, c))
287 } else if (!is_utf8_char_slow(x, c))
291 #endif /* #ifdef IS_UTF8_CHAR */
306 Implemented as a macro in utf8.h
308 =for apidoc A|bool|is_utf8_string_loc|const U8 *s|STRLEN len|const U8 **ep
310 Like is_utf8_string() but stores the location of the failure (in the
311 case of "utf8ness failure") or the location s+len (in the case of
312 "utf8ness success") in the C<ep>.
314 See also is_utf8_string_loclen() and is_utf8_string().
316 =for apidoc A|bool|is_utf8_string_loclen|const U8 *s|STRLEN len|const U8 **ep|const STRLEN *el
318 Like is_utf8_string() but stores the location of the failure (in the
319 case of "utf8ness failure") or the location s+len (in the case of
320 "utf8ness success") in the C<ep>, and the number of UTF-8
321 encoded characters in the C<el>.
323 See also is_utf8_string_loc() and is_utf8_string().
329 Perl_is_utf8_string_loclen(pTHX_ const U8 *s, STRLEN len, const U8 **ep, STRLEN *el)
337 len = strlen((const char *)s);
343 /* Inline the easy bits of is_utf8_char() here for speed... */
344 if (UTF8_IS_INVARIANT(*x))
346 else if (!UTF8_IS_START(*x))
349 /* ... and call is_utf8_char() only if really needed. */
352 if (IS_UTF8_CHAR_FAST(c)) {
353 if (!IS_UTF8_CHAR(x, c))
356 c = is_utf8_char_slow(x, c);
359 #endif /* #ifdef IS_UTF8_CHAR */
379 =for apidoc A|UV|utf8n_to_uvuni|const U8 *s|STRLEN curlen|STRLEN *retlen|U32 flags
381 Bottom level UTF-8 decode routine.
382 Returns the unicode code point value of the first character in the string C<s>
383 which is assumed to be in UTF-8 encoding and no longer than C<curlen>;
384 C<retlen> will be set to the length, in bytes, of that character.
386 If C<s> does not point to a well-formed UTF-8 character, the behaviour
387 is dependent on the value of C<flags>: if it contains UTF8_CHECK_ONLY,
388 it is assumed that the caller will raise a warning, and this function
389 will silently just set C<retlen> to C<-1> and return zero. If the
390 C<flags> does not contain UTF8_CHECK_ONLY, warnings about
391 malformations will be given, C<retlen> will be set to the expected
392 length of the UTF-8 character in bytes, and zero will be returned.
394 The C<flags> can also contain various flags to allow deviations from
395 the strict UTF-8 encoding (see F<utf8.h>).
397 Most code should use utf8_to_uvchr() rather than call this directly.
403 Perl_utf8n_to_uvuni(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen, U32 flags)
409 const bool dowarn = ckWARN_d(WARN_UTF8);
410 const UV startbyte = *s;
411 STRLEN expectlen = 0;
414 /* This list is a superset of the UTF8_ALLOW_XXX. */
416 #define UTF8_WARN_EMPTY 1
417 #define UTF8_WARN_CONTINUATION 2
418 #define UTF8_WARN_NON_CONTINUATION 3
419 #define UTF8_WARN_FE_FF 4
420 #define UTF8_WARN_SHORT 5
421 #define UTF8_WARN_OVERFLOW 6
422 #define UTF8_WARN_SURROGATE 7
423 #define UTF8_WARN_LONG 8
424 #define UTF8_WARN_FFFF 9 /* Also FFFE. */
427 !(flags & UTF8_ALLOW_EMPTY)) {
428 warning = UTF8_WARN_EMPTY;
432 if (UTF8_IS_INVARIANT(uv)) {
435 return (UV) (NATIVE_TO_UTF(*s));
438 if (UTF8_IS_CONTINUATION(uv) &&
439 !(flags & UTF8_ALLOW_CONTINUATION)) {
440 warning = UTF8_WARN_CONTINUATION;
444 if (UTF8_IS_START(uv) && curlen > 1 && !UTF8_IS_CONTINUATION(s[1]) &&
445 !(flags & UTF8_ALLOW_NON_CONTINUATION)) {
446 warning = UTF8_WARN_NON_CONTINUATION;
451 uv = NATIVE_TO_UTF(uv);
453 if ((uv == 0xfe || uv == 0xff) &&
454 !(flags & UTF8_ALLOW_FE_FF)) {
455 warning = UTF8_WARN_FE_FF;
460 if (!(uv & 0x20)) { len = 2; uv &= 0x1f; }
461 else if (!(uv & 0x10)) { len = 3; uv &= 0x0f; }
462 else if (!(uv & 0x08)) { len = 4; uv &= 0x07; }
463 else if (!(uv & 0x04)) { len = 5; uv &= 0x03; }
465 else if (!(uv & 0x02)) { len = 6; uv &= 0x01; }
466 else { len = 7; uv &= 0x01; }
468 else if (!(uv & 0x02)) { len = 6; uv &= 0x01; }
469 else if (!(uv & 0x01)) { len = 7; uv = 0; }
470 else { len = 13; uv = 0; } /* whoa! */
478 if ((curlen < expectlen) &&
479 !(flags & UTF8_ALLOW_SHORT)) {
480 warning = UTF8_WARN_SHORT;
489 if (!UTF8_IS_CONTINUATION(*s) &&
490 !(flags & UTF8_ALLOW_NON_CONTINUATION)) {
492 warning = UTF8_WARN_NON_CONTINUATION;
496 uv = UTF8_ACCUMULATE(uv, *s);
498 /* These cannot be allowed. */
500 if (expectlen != 13 && !(flags & UTF8_ALLOW_LONG)) {
501 warning = UTF8_WARN_LONG;
505 else { /* uv < ouv */
506 /* This cannot be allowed. */
507 warning = UTF8_WARN_OVERFLOW;
515 if (UNICODE_IS_SURROGATE(uv) &&
516 !(flags & UTF8_ALLOW_SURROGATE)) {
517 warning = UTF8_WARN_SURROGATE;
519 } else if ((expectlen > (STRLEN)UNISKIP(uv)) &&
520 !(flags & UTF8_ALLOW_LONG)) {
521 warning = UTF8_WARN_LONG;
523 } else if (UNICODE_IS_ILLEGAL(uv) &&
524 !(flags & UTF8_ALLOW_FFFF)) {
525 warning = UTF8_WARN_FFFF;
533 if (flags & UTF8_CHECK_ONLY) {
540 SV* const sv = sv_2mortal(newSVpvs("Malformed UTF-8 character "));
543 case 0: /* Intentionally empty. */ break;
544 case UTF8_WARN_EMPTY:
545 sv_catpvs(sv, "(empty string)");
547 case UTF8_WARN_CONTINUATION:
548 Perl_sv_catpvf(aTHX_ sv, "(unexpected continuation byte 0x%02"UVxf", with no preceding start byte)", uv);
550 case UTF8_WARN_NON_CONTINUATION:
552 Perl_sv_catpvf(aTHX_ sv, "(unexpected non-continuation byte 0x%02"UVxf", immediately after start byte 0x%02"UVxf")",
553 (UV)s[1], startbyte);
555 const int len = (int)(s-s0);
556 Perl_sv_catpvf(aTHX_ sv, "(unexpected non-continuation byte 0x%02"UVxf", %d byte%s after start byte 0x%02"UVxf", expected %d bytes)",
557 (UV)s[1], len, len > 1 ? "s" : "", startbyte, (int)expectlen);
561 case UTF8_WARN_FE_FF:
562 Perl_sv_catpvf(aTHX_ sv, "(byte 0x%02"UVxf")", uv);
564 case UTF8_WARN_SHORT:
565 Perl_sv_catpvf(aTHX_ sv, "(%d byte%s, need %d, after start byte 0x%02"UVxf")",
566 (int)curlen, curlen == 1 ? "" : "s", (int)expectlen, startbyte);
567 expectlen = curlen; /* distance for caller to skip */
569 case UTF8_WARN_OVERFLOW:
570 Perl_sv_catpvf(aTHX_ sv, "(overflow at 0x%"UVxf", byte 0x%02x, after start byte 0x%02"UVxf")",
573 case UTF8_WARN_SURROGATE:
574 Perl_sv_catpvf(aTHX_ sv, "(UTF-16 surrogate 0x%04"UVxf")", uv);
577 Perl_sv_catpvf(aTHX_ sv, "(%d byte%s, need %d, after start byte 0x%02"UVxf")",
578 (int)expectlen, expectlen == 1 ? "": "s", UNISKIP(uv), startbyte);
581 Perl_sv_catpvf(aTHX_ sv, "(character 0x%04"UVxf")", uv);
584 sv_catpvs(sv, "(unknown reason)");
589 const char * const s = SvPVX_const(sv);
592 Perl_warner(aTHX_ packWARN(WARN_UTF8),
593 "%s in %s", s, OP_DESC(PL_op));
595 Perl_warner(aTHX_ packWARN(WARN_UTF8), "%s", s);
600 *retlen = expectlen ? expectlen : len;
606 =for apidoc A|UV|utf8_to_uvchr|const U8 *s|STRLEN *retlen
608 Returns the native character value of the first character in the string C<s>
609 which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
610 length, in bytes, of that character.
612 If C<s> does not point to a well-formed UTF-8 character, zero is
613 returned and retlen is set, if possible, to -1.
619 Perl_utf8_to_uvchr(pTHX_ const U8 *s, STRLEN *retlen)
621 return utf8n_to_uvchr(s, UTF8_MAXBYTES, retlen,
622 ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
626 =for apidoc A|UV|utf8_to_uvuni|const U8 *s|STRLEN *retlen
628 Returns the Unicode code point of the first character in the string C<s>
629 which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
630 length, in bytes, of that character.
632 This function should only be used when returned UV is considered
633 an index into the Unicode semantic tables (e.g. swashes).
635 If C<s> does not point to a well-formed UTF-8 character, zero is
636 returned and retlen is set, if possible, to -1.
642 Perl_utf8_to_uvuni(pTHX_ const U8 *s, STRLEN *retlen)
644 /* Call the low level routine asking for checks */
645 return Perl_utf8n_to_uvuni(aTHX_ s, UTF8_MAXBYTES, retlen,
646 ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
650 =for apidoc A|STRLEN|utf8_length|const U8 *s|const U8 *e
652 Return the length of the UTF-8 char encoded string C<s> in characters.
653 Stops at C<e> (inclusive). If C<e E<lt> s> or if the scan would end
654 up past C<e>, croaks.
660 Perl_utf8_length(pTHX_ const U8 *s, const U8 *e)
665 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g.
666 * the bitops (especially ~) can create illegal UTF-8.
667 * In other words: in Perl UTF-8 is not just for Unicode. */
670 goto warn_and_return;
672 const U8 t = UTF8SKIP(s);
675 if (ckWARN_d(WARN_UTF8)) {
677 Perl_warner(aTHX_ packWARN(WARN_UTF8),
678 "%s in %s", unees, OP_DESC(PL_op));
680 Perl_warner(aTHX_ packWARN(WARN_UTF8), unees);
692 =for apidoc A|IV|utf8_distance|const U8 *a|const U8 *b
694 Returns the number of UTF-8 characters between the UTF-8 pointers C<a>
697 WARNING: use only if you *know* that the pointers point inside the
704 Perl_utf8_distance(pTHX_ const U8 *a, const U8 *b)
709 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g.
710 * the bitops (especially ~) can create illegal UTF-8.
711 * In other words: in Perl UTF-8 is not just for Unicode. */
715 const U8 c = UTF8SKIP(a);
717 goto warn_and_return;
724 const U8 c = UTF8SKIP(b);
728 if (ckWARN_d(WARN_UTF8)) {
730 Perl_warner(aTHX_ packWARN(WARN_UTF8),
731 "%s in %s", unees, OP_DESC(PL_op));
733 Perl_warner(aTHX_ packWARN(WARN_UTF8), unees);
746 =for apidoc A|U8 *|utf8_hop|U8 *s|I32 off
748 Return the UTF-8 pointer C<s> displaced by C<off> characters, either
751 WARNING: do not use the following unless you *know* C<off> is within
752 the UTF-8 data pointed to by C<s> *and* that on entry C<s> is aligned
753 on the first byte of character or just after the last byte of a character.
759 Perl_utf8_hop(pTHX_ const U8 *s, I32 off)
762 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g
763 * the bitops (especially ~) can create illegal UTF-8.
764 * In other words: in Perl UTF-8 is not just for Unicode. */
773 while (UTF8_IS_CONTINUATION(*s))
781 =for apidoc A|U8 *|utf8_to_bytes|U8 *s|STRLEN *len
783 Converts a string C<s> of length C<len> from UTF-8 into byte encoding.
784 Unlike C<bytes_to_utf8>, this over-writes the original string, and
785 updates len to contain the new length.
786 Returns zero on failure, setting C<len> to -1.
792 Perl_utf8_to_bytes(pTHX_ U8 *s, STRLEN *len)
798 /* ensure valid UTF-8 and chars < 256 before updating string */
799 for (send = s + *len; s < send; ) {
802 if (!UTF8_IS_INVARIANT(c) &&
803 (!UTF8_IS_DOWNGRADEABLE_START(c) || (s >= send)
804 || !(c = *s++) || !UTF8_IS_CONTINUATION(c))) {
813 *d++ = (U8)utf8_to_uvchr(s, &ulen);
822 =for apidoc A|U8 *|bytes_from_utf8|const U8 *s|STRLEN *len|bool *is_utf8
824 Converts a string C<s> of length C<len> from UTF-8 into byte encoding.
825 Unlike C<utf8_to_bytes> but like C<bytes_to_utf8>, returns a pointer to
826 the newly-created string, and updates C<len> to contain the new
827 length. Returns the original string if no conversion occurs, C<len>
828 is unchanged. Do nothing if C<is_utf8> points to 0. Sets C<is_utf8> to
829 0 if C<s> is converted or contains all 7bit characters.
835 Perl_bytes_from_utf8(pTHX_ const U8 *s, STRLEN *len, bool *is_utf8)
846 /* ensure valid UTF-8 and chars < 256 before converting string */
847 for (send = s + *len; s < send;) {
849 if (!UTF8_IS_INVARIANT(c)) {
850 if (UTF8_IS_DOWNGRADEABLE_START(c) && s < send &&
851 (c = *s++) && UTF8_IS_CONTINUATION(c))
860 Newx(d, (*len) - count + 1, U8);
861 s = start; start = d;
864 if (!UTF8_IS_INVARIANT(c)) {
865 /* Then it is two-byte encoded */
866 c = UTF8_ACCUMULATE(NATIVE_TO_UTF(c), *s++);
867 c = ASCII_TO_NATIVE(c);
877 =for apidoc A|U8 *|bytes_to_utf8|const U8 *s|STRLEN *len
879 Converts a string C<s> of length C<len> from ASCII into UTF-8 encoding.
880 Returns a pointer to the newly-created string, and sets C<len> to
881 reflect the new length.
883 If you want to convert to UTF-8 from other encodings than ASCII,
884 see sv_recode_to_utf8().
890 Perl_bytes_to_utf8(pTHX_ const U8 *s, STRLEN *len)
892 const U8 * const send = s + (*len);
897 Newx(d, (*len) * 2 + 1, U8);
901 const UV uv = NATIVE_TO_ASCII(*s++);
902 if (UNI_IS_INVARIANT(uv))
903 *d++ = (U8)UTF_TO_NATIVE(uv);
905 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
906 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
915 * Convert native (big-endian) or reversed (little-endian) UTF-16 to UTF-8.
917 * Destination must be pre-extended to 3/2 source. Do not use in-place.
918 * We optimize for native, for obvious reasons. */
921 Perl_utf16_to_utf8(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen)
926 if (bytelen == 1 && p[0] == 0) { /* Be understanding. */
933 Perl_croak(aTHX_ "panic: utf16_to_utf8: odd bytelen %"UVf, (UV)bytelen);
938 UV uv = (p[0] << 8) + p[1]; /* UTF-16BE */
945 *d++ = (U8)(( uv >> 6) | 0xc0);
946 *d++ = (U8)(( uv & 0x3f) | 0x80);
949 if (uv >= 0xd800 && uv < 0xdbff) { /* surrogates */
950 UV low = (p[0] << 8) + p[1];
952 if (low < 0xdc00 || low >= 0xdfff)
953 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
954 uv = ((uv - 0xd800) << 10) + (low - 0xdc00) + 0x10000;
957 *d++ = (U8)(( uv >> 12) | 0xe0);
958 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
959 *d++ = (U8)(( uv & 0x3f) | 0x80);
963 *d++ = (U8)(( uv >> 18) | 0xf0);
964 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
965 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
966 *d++ = (U8)(( uv & 0x3f) | 0x80);
970 *newlen = d - dstart;
974 /* Note: this one is slightly destructive of the source. */
977 Perl_utf16_to_utf8_reversed(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen)
980 U8* send = s + bytelen;
987 return utf16_to_utf8(p, d, bytelen, newlen);
990 /* for now these are all defined (inefficiently) in terms of the utf8 versions */
993 Perl_is_uni_alnum(pTHX_ UV c)
995 U8 tmpbuf[UTF8_MAXBYTES+1];
996 uvchr_to_utf8(tmpbuf, c);
997 return is_utf8_alnum(tmpbuf);
1001 Perl_is_uni_alnumc(pTHX_ UV c)
1003 U8 tmpbuf[UTF8_MAXBYTES+1];
1004 uvchr_to_utf8(tmpbuf, c);
1005 return is_utf8_alnumc(tmpbuf);
1009 Perl_is_uni_idfirst(pTHX_ UV c)
1011 U8 tmpbuf[UTF8_MAXBYTES+1];
1012 uvchr_to_utf8(tmpbuf, c);
1013 return is_utf8_idfirst(tmpbuf);
1017 Perl_is_uni_alpha(pTHX_ UV c)
1019 U8 tmpbuf[UTF8_MAXBYTES+1];
1020 uvchr_to_utf8(tmpbuf, c);
1021 return is_utf8_alpha(tmpbuf);
1025 Perl_is_uni_ascii(pTHX_ UV c)
1027 U8 tmpbuf[UTF8_MAXBYTES+1];
1028 uvchr_to_utf8(tmpbuf, c);
1029 return is_utf8_ascii(tmpbuf);
1033 Perl_is_uni_space(pTHX_ UV c)
1035 U8 tmpbuf[UTF8_MAXBYTES+1];
1036 uvchr_to_utf8(tmpbuf, c);
1037 return is_utf8_space(tmpbuf);
1041 Perl_is_uni_digit(pTHX_ UV c)
1043 U8 tmpbuf[UTF8_MAXBYTES+1];
1044 uvchr_to_utf8(tmpbuf, c);
1045 return is_utf8_digit(tmpbuf);
1049 Perl_is_uni_upper(pTHX_ UV c)
1051 U8 tmpbuf[UTF8_MAXBYTES+1];
1052 uvchr_to_utf8(tmpbuf, c);
1053 return is_utf8_upper(tmpbuf);
1057 Perl_is_uni_lower(pTHX_ UV c)
1059 U8 tmpbuf[UTF8_MAXBYTES+1];
1060 uvchr_to_utf8(tmpbuf, c);
1061 return is_utf8_lower(tmpbuf);
1065 Perl_is_uni_cntrl(pTHX_ UV c)
1067 U8 tmpbuf[UTF8_MAXBYTES+1];
1068 uvchr_to_utf8(tmpbuf, c);
1069 return is_utf8_cntrl(tmpbuf);
1073 Perl_is_uni_graph(pTHX_ UV c)
1075 U8 tmpbuf[UTF8_MAXBYTES+1];
1076 uvchr_to_utf8(tmpbuf, c);
1077 return is_utf8_graph(tmpbuf);
1081 Perl_is_uni_print(pTHX_ UV c)
1083 U8 tmpbuf[UTF8_MAXBYTES+1];
1084 uvchr_to_utf8(tmpbuf, c);
1085 return is_utf8_print(tmpbuf);
1089 Perl_is_uni_punct(pTHX_ UV c)
1091 U8 tmpbuf[UTF8_MAXBYTES+1];
1092 uvchr_to_utf8(tmpbuf, c);
1093 return is_utf8_punct(tmpbuf);
1097 Perl_is_uni_xdigit(pTHX_ UV c)
1099 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1100 uvchr_to_utf8(tmpbuf, c);
1101 return is_utf8_xdigit(tmpbuf);
1105 Perl_to_uni_upper(pTHX_ UV c, U8* p, STRLEN *lenp)
1107 uvchr_to_utf8(p, c);
1108 return to_utf8_upper(p, p, lenp);
1112 Perl_to_uni_title(pTHX_ UV c, U8* p, STRLEN *lenp)
1114 uvchr_to_utf8(p, c);
1115 return to_utf8_title(p, p, lenp);
1119 Perl_to_uni_lower(pTHX_ UV c, U8* p, STRLEN *lenp)
1121 uvchr_to_utf8(p, c);
1122 return to_utf8_lower(p, p, lenp);
1126 Perl_to_uni_fold(pTHX_ UV c, U8* p, STRLEN *lenp)
1128 uvchr_to_utf8(p, c);
1129 return to_utf8_fold(p, p, lenp);
1132 /* for now these all assume no locale info available for Unicode > 255 */
1135 Perl_is_uni_alnum_lc(pTHX_ UV c)
1137 return is_uni_alnum(c); /* XXX no locale support yet */
1141 Perl_is_uni_alnumc_lc(pTHX_ UV c)
1143 return is_uni_alnumc(c); /* XXX no locale support yet */
1147 Perl_is_uni_idfirst_lc(pTHX_ UV c)
1149 return is_uni_idfirst(c); /* XXX no locale support yet */
1153 Perl_is_uni_alpha_lc(pTHX_ UV c)
1155 return is_uni_alpha(c); /* XXX no locale support yet */
1159 Perl_is_uni_ascii_lc(pTHX_ UV c)
1161 return is_uni_ascii(c); /* XXX no locale support yet */
1165 Perl_is_uni_space_lc(pTHX_ UV c)
1167 return is_uni_space(c); /* XXX no locale support yet */
1171 Perl_is_uni_digit_lc(pTHX_ UV c)
1173 return is_uni_digit(c); /* XXX no locale support yet */
1177 Perl_is_uni_upper_lc(pTHX_ UV c)
1179 return is_uni_upper(c); /* XXX no locale support yet */
1183 Perl_is_uni_lower_lc(pTHX_ UV c)
1185 return is_uni_lower(c); /* XXX no locale support yet */
1189 Perl_is_uni_cntrl_lc(pTHX_ UV c)
1191 return is_uni_cntrl(c); /* XXX no locale support yet */
1195 Perl_is_uni_graph_lc(pTHX_ UV c)
1197 return is_uni_graph(c); /* XXX no locale support yet */
1201 Perl_is_uni_print_lc(pTHX_ UV c)
1203 return is_uni_print(c); /* XXX no locale support yet */
1207 Perl_is_uni_punct_lc(pTHX_ UV c)
1209 return is_uni_punct(c); /* XXX no locale support yet */
1213 Perl_is_uni_xdigit_lc(pTHX_ UV c)
1215 return is_uni_xdigit(c); /* XXX no locale support yet */
1219 Perl_to_uni_upper_lc(pTHX_ U32 c)
1221 /* XXX returns only the first character -- do not use XXX */
1222 /* XXX no locale support yet */
1224 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1225 return (U32)to_uni_upper(c, tmpbuf, &len);
1229 Perl_to_uni_title_lc(pTHX_ U32 c)
1231 /* XXX returns only the first character XXX -- do not use XXX */
1232 /* XXX no locale support yet */
1234 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1235 return (U32)to_uni_title(c, tmpbuf, &len);
1239 Perl_to_uni_lower_lc(pTHX_ U32 c)
1241 /* XXX returns only the first character -- do not use XXX */
1242 /* XXX no locale support yet */
1244 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1245 return (U32)to_uni_lower(c, tmpbuf, &len);
1249 S_is_utf8_common(pTHX_ const U8 *const p, SV **swash,
1250 const char *const swashname)
1253 if (!is_utf8_char(p))
1256 *swash = swash_init("utf8", swashname, &PL_sv_undef, 1, 0);
1257 return swash_fetch(*swash, p, TRUE) != 0;
1261 Perl_is_utf8_alnum(pTHX_ const U8 *p)
1264 /* NOTE: "IsWord", not "IsAlnum", since Alnum is a true
1265 * descendant of isalnum(3), in other words, it doesn't
1266 * contain the '_'. --jhi */
1267 return S_is_utf8_common(aTHX_ p, &PL_utf8_alnum, "IsWord");
1271 Perl_is_utf8_alnumc(pTHX_ const U8 *p)
1274 return S_is_utf8_common(aTHX_ p, &PL_utf8_alnumc, "IsAlnumC");
1278 Perl_is_utf8_idfirst(pTHX_ const U8 *p) /* The naming is historical. */
1283 /* is_utf8_idstart would be more logical. */
1284 return S_is_utf8_common(aTHX_ p, &PL_utf8_idstart, "IdStart");
1288 Perl_is_utf8_idcont(pTHX_ const U8 *p)
1293 return S_is_utf8_common(aTHX_ p, &PL_utf8_idcont, "IdContinue");
1297 Perl_is_utf8_alpha(pTHX_ const U8 *p)
1300 return S_is_utf8_common(aTHX_ p, &PL_utf8_alpha, "IsAlpha");
1304 Perl_is_utf8_ascii(pTHX_ const U8 *p)
1307 return S_is_utf8_common(aTHX_ p, &PL_utf8_ascii, "IsAscii");
1311 Perl_is_utf8_space(pTHX_ const U8 *p)
1314 return S_is_utf8_common(aTHX_ p, &PL_utf8_space, "IsSpacePerl");
1318 Perl_is_utf8_digit(pTHX_ const U8 *p)
1321 return S_is_utf8_common(aTHX_ p, &PL_utf8_digit, "IsDigit");
1325 Perl_is_utf8_upper(pTHX_ const U8 *p)
1328 return S_is_utf8_common(aTHX_ p, &PL_utf8_upper, "IsUppercase");
1332 Perl_is_utf8_lower(pTHX_ const U8 *p)
1335 return S_is_utf8_common(aTHX_ p, &PL_utf8_lower, "IsLowercase");
1339 Perl_is_utf8_cntrl(pTHX_ const U8 *p)
1342 return S_is_utf8_common(aTHX_ p, &PL_utf8_cntrl, "IsCntrl");
1346 Perl_is_utf8_graph(pTHX_ const U8 *p)
1349 return S_is_utf8_common(aTHX_ p, &PL_utf8_graph, "IsGraph");
1353 Perl_is_utf8_print(pTHX_ const U8 *p)
1356 return S_is_utf8_common(aTHX_ p, &PL_utf8_print, "IsPrint");
1360 Perl_is_utf8_punct(pTHX_ const U8 *p)
1363 return S_is_utf8_common(aTHX_ p, &PL_utf8_punct, "IsPunct");
1367 Perl_is_utf8_xdigit(pTHX_ const U8 *p)
1370 return S_is_utf8_common(aTHX_ p, &PL_utf8_xdigit, "Isxdigit");
1374 Perl_is_utf8_mark(pTHX_ const U8 *p)
1377 return S_is_utf8_common(aTHX_ p, &PL_utf8_mark, "IsM");
1381 =for apidoc A|UV|to_utf8_case|U8 *p|U8* ustrp|STRLEN *lenp|SV **swash|char *normal|char *special
1383 The "p" contains the pointer to the UTF-8 string encoding
1384 the character that is being converted.
1386 The "ustrp" is a pointer to the character buffer to put the
1387 conversion result to. The "lenp" is a pointer to the length
1390 The "swashp" is a pointer to the swash to use.
1392 Both the special and normal mappings are stored lib/unicore/To/Foo.pl,
1393 and loaded by SWASHNEW, using lib/utf8_heavy.pl. The special (usually,
1394 but not always, a multicharacter mapping), is tried first.
1396 The "special" is a string like "utf8::ToSpecLower", which means the
1397 hash %utf8::ToSpecLower. The access to the hash is through
1398 Perl_to_utf8_case().
1400 The "normal" is a string like "ToLower" which means the swash
1406 Perl_to_utf8_case(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp,
1407 SV **swashp, const char *normal, const char *special)
1410 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1413 const UV uv0 = utf8_to_uvchr(p, NULL);
1414 /* The NATIVE_TO_UNI() and UNI_TO_NATIVE() mappings
1415 * are necessary in EBCDIC, they are redundant no-ops
1416 * in ASCII-ish platforms, and hopefully optimized away. */
1417 const UV uv1 = NATIVE_TO_UNI(uv0);
1418 uvuni_to_utf8(tmpbuf, uv1);
1420 if (!*swashp) /* load on-demand */
1421 *swashp = swash_init("utf8", normal, &PL_sv_undef, 4, 0);
1423 /* The 0xDF is the only special casing Unicode code point below 0x100. */
1424 if (special && (uv1 == 0xDF || uv1 > 0xFF)) {
1425 /* It might be "special" (sometimes, but not always,
1426 * a multicharacter mapping) */
1430 if ((hv = get_hv(special, FALSE)) &&
1431 (svp = hv_fetch(hv, (const char*)tmpbuf, UNISKIP(uv1), FALSE)) &&
1435 s = SvPV_const(*svp, len);
1437 len = uvuni_to_utf8(ustrp, NATIVE_TO_UNI(*(U8*)s)) - ustrp;
1440 /* If we have EBCDIC we need to remap the characters
1441 * since any characters in the low 256 are Unicode
1442 * code points, not EBCDIC. */
1443 U8 *t = (U8*)s, *tend = t + len, *d;
1450 UV c = utf8_to_uvchr(t, &tlen);
1452 d = uvchr_to_utf8(d, UNI_TO_NATIVE(c));
1461 d = uvchr_to_utf8(d, UNI_TO_NATIVE(*t));
1466 Copy(tmpbuf, ustrp, len, U8);
1468 Copy(s, ustrp, len, U8);
1474 if (!len && *swashp) {
1475 UV uv2 = swash_fetch(*swashp, tmpbuf, TRUE);
1478 /* It was "normal" (a single character mapping). */
1479 UV uv3 = UNI_TO_NATIVE(uv2);
1481 len = uvchr_to_utf8(ustrp, uv3) - ustrp;
1485 if (!len) /* Neither: just copy. */
1486 len = uvchr_to_utf8(ustrp, uv0) - ustrp;
1491 return len ? utf8_to_uvchr(ustrp, 0) : 0;
1495 =for apidoc A|UV|to_utf8_upper|const U8 *p|U8 *ustrp|STRLEN *lenp
1497 Convert the UTF-8 encoded character at p to its uppercase version and
1498 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
1499 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since
1500 the uppercase version may be longer than the original character.
1502 The first character of the uppercased version is returned
1503 (but note, as explained above, that there may be more.)
1508 Perl_to_utf8_upper(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
1511 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
1512 &PL_utf8_toupper, "ToUpper", "utf8::ToSpecUpper");
1516 =for apidoc A|UV|to_utf8_title|const U8 *p|U8 *ustrp|STRLEN *lenp
1518 Convert the UTF-8 encoded character at p to its titlecase version and
1519 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
1520 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
1521 titlecase version may be longer than the original character.
1523 The first character of the titlecased version is returned
1524 (but note, as explained above, that there may be more.)
1529 Perl_to_utf8_title(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
1532 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
1533 &PL_utf8_totitle, "ToTitle", "utf8::ToSpecTitle");
1537 =for apidoc A|UV|to_utf8_lower|const U8 *p|U8 *ustrp|STRLEN *lenp
1539 Convert the UTF-8 encoded character at p to its lowercase version and
1540 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
1541 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
1542 lowercase version may be longer than the original character.
1544 The first character of the lowercased version is returned
1545 (but note, as explained above, that there may be more.)
1550 Perl_to_utf8_lower(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
1553 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
1554 &PL_utf8_tolower, "ToLower", "utf8::ToSpecLower");
1558 =for apidoc A|UV|to_utf8_fold|const U8 *p|U8 *ustrp|STRLEN *lenp
1560 Convert the UTF-8 encoded character at p to its foldcase version and
1561 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
1562 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
1563 foldcase version may be longer than the original character (up to
1566 The first character of the foldcased version is returned
1567 (but note, as explained above, that there may be more.)
1572 Perl_to_utf8_fold(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
1575 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
1576 &PL_utf8_tofold, "ToFold", "utf8::ToSpecFold");
1580 * A "swash" is a swatch hash.
1581 * A "swatch" is a bit vector generated by utf8.c:S_swash_get().
1582 * C<pkg> is a pointer to a package name for SWASHNEW, should be "utf8".
1583 * For other parameters, see utf8::SWASHNEW in lib/utf8_heavy.pl.
1586 Perl_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv, I32 minbits, I32 none)
1590 SV* const tokenbufsv = sv_newmortal();
1592 const size_t pkg_len = strlen(pkg);
1593 const size_t name_len = strlen(name);
1594 HV * const stash = gv_stashpvn(pkg, pkg_len, FALSE);
1597 PUSHSTACKi(PERLSI_MAGIC);
1602 if (!gv_fetchmeth(stash, "SWASHNEW", 8, -1)) { /* demand load utf8 */
1604 errsv_save = newSVsv(ERRSV);
1605 /* It is assumed that callers of this routine are not passing in any
1606 user derived data. */
1607 /* Need to do this after save_re_context() as it will set PL_tainted to
1608 1 while saving $1 etc (see the code after getrx: in Perl_magic_get).
1609 Even line to create errsv_save can turn on PL_tainted. */
1610 SAVEBOOL(PL_tainted);
1612 Perl_load_module(aTHX_ PERL_LOADMOD_NOIMPORT, newSVpvn(pkg,pkg_len),
1615 sv_setsv(ERRSV, errsv_save);
1616 SvREFCNT_dec(errsv_save);
1622 PUSHs(sv_2mortal(newSVpvn(pkg, pkg_len)));
1623 PUSHs(sv_2mortal(newSVpvn(name, name_len)));
1625 PUSHs(sv_2mortal(newSViv(minbits)));
1626 PUSHs(sv_2mortal(newSViv(none)));
1628 if (IN_PERL_COMPILETIME) {
1629 /* XXX ought to be handled by lex_start */
1632 sv_setpv(tokenbufsv, PL_tokenbuf);
1634 errsv_save = newSVsv(ERRSV);
1635 if (call_method("SWASHNEW", G_SCALAR))
1636 retval = newSVsv(*PL_stack_sp--);
1638 retval = &PL_sv_undef;
1640 sv_setsv(ERRSV, errsv_save);
1641 SvREFCNT_dec(errsv_save);
1644 if (IN_PERL_COMPILETIME) {
1646 const char* const pv = SvPV_const(tokenbufsv, len);
1648 Copy(pv, PL_tokenbuf, len+1, char);
1649 PL_curcop->op_private = (U8)(PL_hints & HINT_PRIVATE_MASK);
1651 if (!SvROK(retval) || SvTYPE(SvRV(retval)) != SVt_PVHV) {
1653 Perl_croak(aTHX_ "Can't find Unicode property definition \"%"SVf"\"",
1655 Perl_croak(aTHX_ "SWASHNEW didn't return an HV ref");
1661 /* This API is wrong for special case conversions since we may need to
1662 * return several Unicode characters for a single Unicode character
1663 * (see lib/unicore/SpecCase.txt) The SWASHGET in lib/utf8_heavy.pl is
1664 * the lower-level routine, and it is similarly broken for returning
1665 * multiple values. --jhi */
1666 /* Now SWASHGET is recasted into S_swash_get in this file. */
1669 * Returns the value of property/mapping C<swash> for the first character
1670 * of the string C<ptr>. If C<do_utf8> is true, the string C<ptr> is
1671 * assumed to be in utf8. If C<do_utf8> is false, the string C<ptr> is
1672 * assumed to be in native 8-bit encoding. Caches the swatch in C<swash>.
1675 Perl_swash_fetch(pTHX_ SV *swash, const U8 *ptr, bool do_utf8)
1678 HV* const hv = (HV*)SvRV(swash);
1683 const U8 *tmps = NULL;
1687 UV c = NATIVE_TO_ASCII(*ptr);
1689 if (!do_utf8 && !UNI_IS_INVARIANT(c)) {
1690 tmputf8[0] = (U8)UTF8_EIGHT_BIT_HI(c);
1691 tmputf8[1] = (U8)UTF8_EIGHT_BIT_LO(c);
1694 /* Given a UTF-X encoded char 0xAA..0xYY,0xZZ
1695 * then the "swatch" is a vec() for al the chars which start
1697 * So the key in the hash (klen) is length of encoded char -1
1699 klen = UTF8SKIP(ptr) - 1;
1703 /* If char in invariant then swatch is for all the invariant chars
1704 * In both UTF-8 and UTF-8-MOD that happens to be UTF_CONTINUATION_MARK
1706 needents = UTF_CONTINUATION_MARK;
1707 off = NATIVE_TO_UTF(ptr[klen]);
1710 /* If char is encoded then swatch is for the prefix */
1711 needents = (1 << UTF_ACCUMULATION_SHIFT);
1712 off = NATIVE_TO_UTF(ptr[klen]) & UTF_CONTINUATION_MASK;
1716 * This single-entry cache saves about 1/3 of the utf8 overhead in test
1717 * suite. (That is, only 7-8% overall over just a hash cache. Still,
1718 * it's nothing to sniff at.) Pity we usually come through at least
1719 * two function calls to get here...
1721 * NB: this code assumes that swatches are never modified, once generated!
1724 if (hv == PL_last_swash_hv &&
1725 klen == PL_last_swash_klen &&
1726 (!klen || memEQ((char *)ptr, (char *)PL_last_swash_key, klen)) )
1728 tmps = PL_last_swash_tmps;
1729 slen = PL_last_swash_slen;
1732 /* Try our second-level swatch cache, kept in a hash. */
1733 SV** svp = hv_fetch(hv, (const char*)ptr, klen, FALSE);
1735 /* If not cached, generate it via swash_get */
1736 if (!svp || !SvPOK(*svp)
1737 || !(tmps = (const U8*)SvPV_const(*svp, slen))) {
1738 /* We use utf8n_to_uvuni() as we want an index into
1739 Unicode tables, not a native character number.
1741 const UV code_point = utf8n_to_uvuni(ptr, UTF8_MAXBYTES, 0,
1743 0 : UTF8_ALLOW_ANY);
1744 swatch = swash_get(swash,
1745 /* On EBCDIC & ~(0xA0-1) isn't a useful thing to do */
1746 (klen) ? (code_point & ~(needents - 1)) : 0,
1749 if (IN_PERL_COMPILETIME)
1750 PL_curcop->op_private = (U8)(PL_hints & HINT_PRIVATE_MASK);
1752 svp = hv_store(hv, (const char *)ptr, klen, swatch, 0);
1754 if (!svp || !(tmps = (U8*)SvPV(*svp, slen))
1755 || (slen << 3) < needents)
1756 Perl_croak(aTHX_ "panic: swash_fetch got improper swatch");
1759 PL_last_swash_hv = hv;
1760 PL_last_swash_klen = klen;
1761 /* FIXME change interpvar.h? */
1762 PL_last_swash_tmps = (U8 *) tmps;
1763 PL_last_swash_slen = slen;
1765 Copy(ptr, PL_last_swash_key, klen, U8);
1768 switch ((int)((slen << 3) / needents)) {
1770 bit = 1 << (off & 7);
1772 return (tmps[off] & bit) != 0;
1777 return (tmps[off] << 8) + tmps[off + 1] ;
1780 return (tmps[off] << 24) + (tmps[off+1] << 16) + (tmps[off+2] << 8) + tmps[off + 3] ;
1782 Perl_croak(aTHX_ "panic: swash_fetch got swatch of unexpected bit width");
1787 * Returns a swatch (a bit vector string) for a code point sequence
1788 * that starts from the value C<start> and comprises the number C<span>.
1789 * A C<swash> must be an object created by SWASHNEW (see lib/utf8_heavy.pl).
1790 * Should be used via swash_fetch, which will cache the swatch in C<swash>.
1793 S_swash_get(pTHX_ SV* swash, UV start, UV span)
1796 U8 *l, *lend, *x, *xend, *s;
1797 STRLEN lcur, xcur, scur;
1799 HV* const hv = (HV*)SvRV(swash);
1800 SV** const listsvp = hv_fetchs(hv, "LIST", FALSE);
1801 SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE);
1802 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
1803 SV** const nonesvp = hv_fetchs(hv, "NONE", FALSE);
1804 SV** const extssvp = hv_fetchs(hv, "EXTRAS", FALSE);
1805 const U8* const typestr = (U8*)SvPV_nolen(*typesvp);
1806 const int typeto = typestr[0] == 'T' && typestr[1] == 'o';
1807 const STRLEN bits = SvUV(*bitssvp);
1808 const STRLEN octets = bits >> 3; /* if bits == 1, then octets == 0 */
1809 const UV none = SvUV(*nonesvp);
1810 const UV end = start + span;
1812 if (bits != 1 && bits != 8 && bits != 16 && bits != 32) {
1813 Perl_croak(aTHX_ "panic: swash_get doesn't expect bits %"UVuf,
1817 /* create and initialize $swatch */
1818 swatch = newSVpvs("");
1819 scur = octets ? (span * octets) : (span + 7) / 8;
1820 SvGROW(swatch, scur + 1);
1821 s = (U8*)SvPVX(swatch);
1822 if (octets && none) {
1823 const U8* const e = s + scur;
1826 *s++ = (U8)(none & 0xff);
1827 else if (bits == 16) {
1828 *s++ = (U8)((none >> 8) & 0xff);
1829 *s++ = (U8)( none & 0xff);
1831 else if (bits == 32) {
1832 *s++ = (U8)((none >> 24) & 0xff);
1833 *s++ = (U8)((none >> 16) & 0xff);
1834 *s++ = (U8)((none >> 8) & 0xff);
1835 *s++ = (U8)( none & 0xff);
1841 (void)memzero((U8*)s, scur + 1);
1843 SvCUR_set(swatch, scur);
1844 s = (U8*)SvPVX(swatch);
1846 /* read $swash->{LIST} */
1847 l = (U8*)SvPV(*listsvp, lcur);
1850 UV min, max, val, key;
1852 I32 flags = PERL_SCAN_SILENT_ILLDIGIT | PERL_SCAN_DISALLOW_PREFIX;
1854 U8* const nl = (U8*)memchr(l, '\n', lend - l);
1857 min = grok_hex((char *)l, &numlen, &flags, NULL);
1861 l = nl + 1; /* 1 is length of "\n" */
1865 l = lend; /* to LIST's end at which \n is not found */
1871 flags = PERL_SCAN_SILENT_ILLDIGIT | PERL_SCAN_DISALLOW_PREFIX;
1873 max = grok_hex((char *)l, &numlen, &flags, NULL);
1882 flags = PERL_SCAN_SILENT_ILLDIGIT |
1883 PERL_SCAN_DISALLOW_PREFIX;
1885 val = grok_hex((char *)l, &numlen, &flags, NULL);
1894 Perl_croak(aTHX_ "%s: illegal mapping '%s'",
1900 val = 0; /* bits == 1, then val should be ignored */
1907 Perl_croak(aTHX_ "%s: illegal mapping '%s'", typestr, l);
1911 val = 0; /* bits == 1, then val should be ignored */
1924 if (!none || val < none) {
1929 for (key = min; key <= max; key++) {
1933 /* offset must be non-negative (start <= min <= key < end) */
1934 offset = octets * (key - start);
1936 s[offset] = (U8)(val & 0xff);
1937 else if (bits == 16) {
1938 s[offset ] = (U8)((val >> 8) & 0xff);
1939 s[offset + 1] = (U8)( val & 0xff);
1941 else if (bits == 32) {
1942 s[offset ] = (U8)((val >> 24) & 0xff);
1943 s[offset + 1] = (U8)((val >> 16) & 0xff);
1944 s[offset + 2] = (U8)((val >> 8) & 0xff);
1945 s[offset + 3] = (U8)( val & 0xff);
1948 if (!none || val < none)
1952 else { /* bits == 1, then val should be ignored */
1955 for (key = min; key <= max; key++) {
1956 const STRLEN offset = (STRLEN)(key - start);
1959 s[offset >> 3] |= 1 << (offset & 7);
1965 /* read $swash->{EXTRAS} */
1966 x = (U8*)SvPV(*extssvp, xcur);
1974 SV **otherbitssvp, *other;
1982 nl = (U8*)memchr(x, '\n', xend - x);
1984 if (opc != '-' && opc != '+' && opc != '!' && opc != '&') {
1986 x = nl + 1; /* 1 is length of "\n" */
1990 x = xend; /* to EXTRAS' end at which \n is not found */
1997 namelen = nl - namestr;
2001 namelen = xend - namestr;
2005 othersvp = hv_fetch(hv, (char *)namestr, namelen, FALSE);
2006 otherhv = (HV*)SvRV(*othersvp);
2007 otherbitssvp = hv_fetchs(otherhv, "BITS", FALSE);
2008 otherbits = (STRLEN)SvUV(*otherbitssvp);
2009 if (bits < otherbits)
2010 Perl_croak(aTHX_ "panic: swash_get found swatch size mismatch");
2012 /* The "other" swatch must be destroyed after. */
2013 other = swash_get(*othersvp, start, span);
2014 o = (U8*)SvPV(other, olen);
2017 Perl_croak(aTHX_ "panic: swash_get got improper swatch");
2019 s = (U8*)SvPV(swatch, slen);
2020 if (bits == 1 && otherbits == 1) {
2022 Perl_croak(aTHX_ "panic: swash_get found swatch length mismatch");
2046 STRLEN otheroctets = otherbits >> 3;
2048 U8* send = s + slen;
2053 if (otherbits == 1) {
2054 otherval = (o[offset >> 3] >> (offset & 7)) & 1;
2058 STRLEN vlen = otheroctets;
2066 if (opc == '+' && otherval)
2067 /*EMPTY*/; /* replace with otherval */
2068 else if (opc == '!' && !otherval)
2070 else if (opc == '-' && otherval)
2072 else if (opc == '&' && !otherval)
2075 s += octets; /* no replacement */
2080 *s++ = (U8)( otherval & 0xff);
2081 else if (bits == 16) {
2082 *s++ = (U8)((otherval >> 8) & 0xff);
2083 *s++ = (U8)( otherval & 0xff);
2085 else if (bits == 32) {
2086 *s++ = (U8)((otherval >> 24) & 0xff);
2087 *s++ = (U8)((otherval >> 16) & 0xff);
2088 *s++ = (U8)((otherval >> 8) & 0xff);
2089 *s++ = (U8)( otherval & 0xff);
2093 sv_free(other); /* through with it! */
2099 =for apidoc A|U8 *|uvchr_to_utf8|U8 *d|UV uv
2101 Adds the UTF-8 representation of the Native codepoint C<uv> to the end
2102 of the string C<d>; C<d> should be have at least C<UTF8_MAXBYTES+1> free
2103 bytes available. The return value is the pointer to the byte after the
2104 end of the new character. In other words,
2106 d = uvchr_to_utf8(d, uv);
2108 is the recommended wide native character-aware way of saying
2115 /* On ASCII machines this is normally a macro but we want a
2116 real function in case XS code wants it
2119 Perl_uvchr_to_utf8(pTHX_ U8 *d, UV uv)
2121 return Perl_uvuni_to_utf8_flags(aTHX_ d, NATIVE_TO_UNI(uv), 0);
2125 Perl_uvchr_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags)
2127 return Perl_uvuni_to_utf8_flags(aTHX_ d, NATIVE_TO_UNI(uv), flags);
2131 =for apidoc A|UV|utf8n_to_uvchr|U8 *s|STRLEN curlen|STRLEN *retlen|U32
2134 Returns the native character value of the first character in the string
2136 which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
2137 length, in bytes, of that character.
2139 Allows length and flags to be passed to low level routine.
2143 /* On ASCII machines this is normally a macro but we want
2144 a real function in case XS code wants it
2147 Perl_utf8n_to_uvchr(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen,
2150 const UV uv = Perl_utf8n_to_uvuni(aTHX_ s, curlen, retlen, flags);
2151 return UNI_TO_NATIVE(uv);
2155 =for apidoc A|char *|pv_uni_display|SV *dsv|U8 *spv|STRLEN len|STRLEN pvlim|UV flags
2157 Build to the scalar dsv a displayable version of the string spv,
2158 length len, the displayable version being at most pvlim bytes long
2159 (if longer, the rest is truncated and "..." will be appended).
2161 The flags argument can have UNI_DISPLAY_ISPRINT set to display
2162 isPRINT()able characters as themselves, UNI_DISPLAY_BACKSLASH
2163 to display the \\[nrfta\\] as the backslashed versions (like '\n')
2164 (UNI_DISPLAY_BACKSLASH is preferred over UNI_DISPLAY_ISPRINT for \\).
2165 UNI_DISPLAY_QQ (and its alias UNI_DISPLAY_REGEX) have both
2166 UNI_DISPLAY_BACKSLASH and UNI_DISPLAY_ISPRINT turned on.
2168 The pointer to the PV of the dsv is returned.
2172 Perl_pv_uni_display(pTHX_ SV *dsv, const U8 *spv, STRLEN len, STRLEN pvlim, UV flags)
2177 sv_setpvn(dsv, "", 0);
2178 for (s = (const char *)spv, e = s + len; s < e; s += UTF8SKIP(s)) {
2180 /* This serves double duty as a flag and a character to print after
2181 a \ when flags & UNI_DISPLAY_BACKSLASH is true.
2185 if (pvlim && SvCUR(dsv) >= pvlim) {
2189 u = utf8_to_uvchr((U8*)s, 0);
2191 const unsigned char c = (unsigned char)u & 0xFF;
2192 if (flags & UNI_DISPLAY_BACKSLASH) {
2209 Perl_sv_catpvf(aTHX_ dsv, "\\%c", ok);
2212 /* isPRINT() is the locale-blind version. */
2213 if (!ok && (flags & UNI_DISPLAY_ISPRINT) && isPRINT(c)) {
2214 Perl_sv_catpvf(aTHX_ dsv, "%c", c);
2219 Perl_sv_catpvf(aTHX_ dsv, "\\x{%"UVxf"}", u);
2222 sv_catpvs(dsv, "...");
2228 =for apidoc A|char *|sv_uni_display|SV *dsv|SV *ssv|STRLEN pvlim|UV flags
2230 Build to the scalar dsv a displayable version of the scalar sv,
2231 the displayable version being at most pvlim bytes long
2232 (if longer, the rest is truncated and "..." will be appended).
2234 The flags argument is as in pv_uni_display().
2236 The pointer to the PV of the dsv is returned.
2240 Perl_sv_uni_display(pTHX_ SV *dsv, SV *ssv, STRLEN pvlim, UV flags)
2242 return Perl_pv_uni_display(aTHX_ dsv, (const U8*)SvPVX_const(ssv),
2243 SvCUR(ssv), pvlim, flags);
2247 =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
2249 Return true if the strings s1 and s2 differ case-insensitively, false
2250 if not (if they are equal case-insensitively). If u1 is true, the
2251 string s1 is assumed to be in UTF-8-encoded Unicode. If u2 is true,
2252 the string s2 is assumed to be in UTF-8-encoded Unicode. If u1 or u2
2253 are false, the respective string is assumed to be in native 8-bit
2256 If the pe1 and pe2 are non-NULL, the scanning pointers will be copied
2257 in there (they will point at the beginning of the I<next> character).
2258 If the pointers behind pe1 or pe2 are non-NULL, they are the end
2259 pointers beyond which scanning will not continue under any
2260 circumstances. If the byte lengths l1 and l2 are non-zero, s1+l1 and
2261 s2+l2 will be used as goal end pointers that will also stop the scan,
2262 and which qualify towards defining a successful match: all the scans
2263 that define an explicit length must reach their goal pointers for
2264 a match to succeed).
2266 For case-insensitiveness, the "casefolding" of Unicode is used
2267 instead of upper/lowercasing both the characters, see
2268 http://www.unicode.org/unicode/reports/tr21/ (Case Mappings).
2272 Perl_ibcmp_utf8(pTHX_ const char *s1, char **pe1, register UV l1, bool u1, const char *s2, char **pe2, register UV l2, bool u2)
2275 register const U8 *p1 = (const U8*)s1;
2276 register const U8 *p2 = (const U8*)s2;
2277 register const U8 *f1 = NULL;
2278 register const U8 *f2 = NULL;
2279 register U8 *e1 = NULL;
2280 register U8 *q1 = NULL;
2281 register U8 *e2 = NULL;
2282 register U8 *q2 = NULL;
2283 STRLEN n1 = 0, n2 = 0;
2284 U8 foldbuf1[UTF8_MAXBYTES_CASE+1];
2285 U8 foldbuf2[UTF8_MAXBYTES_CASE+1];
2287 STRLEN foldlen1, foldlen2;
2292 if (e1 == 0 || (l1 && l1 < (UV)(e1 - (const U8*)s1)))
2293 f1 = (const U8*)s1 + l1;
2296 if (e2 == 0 || (l2 && l2 < (UV)(e2 - (const U8*)s2)))
2297 f2 = (const U8*)s2 + l2;
2299 if ((e1 == 0 && f1 == 0) || (e2 == 0 && f2 == 0) || (f1 == 0 && f2 == 0))
2300 return 1; /* mismatch; possible infinite loop or false positive */
2303 natbuf[1] = 0; /* Need to terminate the buffer. */
2305 while ((e1 == 0 || p1 < e1) &&
2306 (f1 == 0 || p1 < f1) &&
2307 (e2 == 0 || p2 < e2) &&
2308 (f2 == 0 || p2 < f2)) {
2311 to_utf8_fold(p1, foldbuf1, &foldlen1);
2313 uvuni_to_utf8(natbuf, (UV) NATIVE_TO_UNI(((UV)*p1)));
2314 to_utf8_fold(natbuf, foldbuf1, &foldlen1);
2321 to_utf8_fold(p2, foldbuf2, &foldlen2);
2323 uvuni_to_utf8(natbuf, (UV) NATIVE_TO_UNI(((UV)*p2)));
2324 to_utf8_fold(natbuf, foldbuf2, &foldlen2);
2330 if ( UTF8SKIP(q1) != UTF8SKIP(q2) ||
2331 (UTF8SKIP(q1) == 1 && *q1 != *q2) ||
2332 memNE((char*)q1, (char*)q2, UTF8SKIP(q1)) )
2333 return 1; /* mismatch */
2340 p1 += u1 ? UTF8SKIP(p1) : 1;
2342 p2 += u2 ? UTF8SKIP(p2) : 1;
2346 /* A match is defined by all the scans that specified
2347 * an explicit length reaching their final goals. */
2348 match = (f1 == 0 || p1 == f1) && (f2 == 0 || p2 == f2);
2357 return match ? 0 : 1; /* 0 match, 1 mismatch */
2362 * c-indentation-style: bsd
2364 * indent-tabs-mode: t
2367 * ex: set ts=8 sts=4 sw=4 noet: