3 * Copyright (C) 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008
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 * [p.603 of _The Lord of the Rings_, IV/I: "The Taming of Sméagol"]
18 * 'Well do I understand your speech,' he answered in the same language;
19 * 'yet few strangers do so. Why then do you not speak in the Common Tongue,
20 * as is the custom in the West, if you wish to be answered?'
21 * --Gandalf, addressing Théoden's door wardens
23 * [p.508 of _The Lord of the Rings_, III/vi: "The King of the Golden Hall"]
25 * ...the travellers perceived that the floor was paved with stones of many
26 * hues; branching runes and strange devices intertwined beneath their feet.
28 * [p.512 of _The Lord of the Rings_, III/vi: "The King of the Golden Hall"]
32 #define PERL_IN_UTF8_C
36 /* Separate prototypes needed because in ASCII systems these
37 * usually macros but they still are compiled as code, too. */
38 PERL_CALLCONV UV Perl_utf8n_to_uvchr(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen, U32 flags);
39 PERL_CALLCONV U8* Perl_uvchr_to_utf8(pTHX_ U8 *d, UV uv);
42 static const char unees[] =
43 "Malformed UTF-8 character (unexpected end of string)";
46 =head1 Unicode Support
48 This file contains various utility functions for manipulating UTF8-encoded
49 strings. For the uninitiated, this is a method of representing arbitrary
50 Unicode characters as a variable number of bytes, in such a way that
51 characters in the ASCII range are unmodified, and a zero byte never appears
52 within non-zero characters.
58 =for apidoc is_ascii_string
60 Returns true if first C<len> bytes of the given string are ASCII (i.e. none
61 of them even raise the question of UTF-8-ness).
63 See also is_utf8_string(), is_utf8_string_loclen(), and is_utf8_string_loc().
69 Perl_is_ascii_string(const U8 *s, STRLEN len)
71 const U8* const send = s + (len ? len : strlen((const char *)s));
74 PERL_ARGS_ASSERT_IS_ASCII_STRING;
76 for (; x < send; ++x) {
77 if (!UTF8_IS_INVARIANT(*x))
85 =for apidoc uvuni_to_utf8_flags
87 Adds the UTF-8 representation of the Unicode codepoint C<uv> to the end
88 of the string C<d>; C<d> should be have at least C<UTF8_MAXBYTES+1> free
89 bytes available. The return value is the pointer to the byte after the
90 end of the new character. In other words,
92 d = uvuni_to_utf8_flags(d, uv, flags);
96 d = uvuni_to_utf8(d, uv);
98 (which is equivalent to)
100 d = uvuni_to_utf8_flags(d, uv, 0);
102 is the recommended Unicode-aware way of saying
110 Perl_uvuni_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags)
112 PERL_ARGS_ASSERT_UVUNI_TO_UTF8_FLAGS;
114 if (ckWARN(WARN_UTF8)) {
115 if (UNICODE_IS_SURROGATE(uv) &&
116 !(flags & UNICODE_ALLOW_SURROGATE))
117 Perl_warner(aTHX_ packWARN(WARN_UTF8), "UTF-16 surrogate 0x%04"UVxf, uv);
119 ((uv >= 0xFDD0 && uv <= 0xFDEF &&
120 !(flags & UNICODE_ALLOW_FDD0))
122 ((uv & 0xFFFE) == 0xFFFE && /* Either FFFE or FFFF. */
123 !(flags & UNICODE_ALLOW_FFFF))) &&
124 /* UNICODE_ALLOW_SUPER includes
125 * FFFEs and FFFFs beyond 0x10FFFF. */
126 ((uv <= PERL_UNICODE_MAX) ||
127 !(flags & UNICODE_ALLOW_SUPER))
129 Perl_warner(aTHX_ packWARN(WARN_UTF8),
130 "Unicode character 0x%04"UVxf" is illegal", uv);
132 if (UNI_IS_INVARIANT(uv)) {
133 *d++ = (U8)UTF_TO_NATIVE(uv);
138 STRLEN len = UNISKIP(uv);
141 *p-- = (U8)UTF_TO_NATIVE((uv & UTF_CONTINUATION_MASK) | UTF_CONTINUATION_MARK);
142 uv >>= UTF_ACCUMULATION_SHIFT;
144 *p = (U8)UTF_TO_NATIVE((uv & UTF_START_MASK(len)) | UTF_START_MARK(len));
147 #else /* Non loop style */
149 *d++ = (U8)(( uv >> 6) | 0xc0);
150 *d++ = (U8)(( uv & 0x3f) | 0x80);
154 *d++ = (U8)(( uv >> 12) | 0xe0);
155 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
156 *d++ = (U8)(( uv & 0x3f) | 0x80);
160 *d++ = (U8)(( uv >> 18) | 0xf0);
161 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
162 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
163 *d++ = (U8)(( uv & 0x3f) | 0x80);
166 if (uv < 0x4000000) {
167 *d++ = (U8)(( uv >> 24) | 0xf8);
168 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
169 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
170 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
171 *d++ = (U8)(( uv & 0x3f) | 0x80);
174 if (uv < 0x80000000) {
175 *d++ = (U8)(( uv >> 30) | 0xfc);
176 *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80);
177 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
178 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
179 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
180 *d++ = (U8)(( uv & 0x3f) | 0x80);
184 if (uv < UTF8_QUAD_MAX)
187 *d++ = 0xfe; /* Can't match U+FEFF! */
188 *d++ = (U8)(((uv >> 30) & 0x3f) | 0x80);
189 *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80);
190 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
191 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
192 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
193 *d++ = (U8)(( uv & 0x3f) | 0x80);
198 *d++ = 0xff; /* Can't match U+FFFE! */
199 *d++ = 0x80; /* 6 Reserved bits */
200 *d++ = (U8)(((uv >> 60) & 0x0f) | 0x80); /* 2 Reserved bits */
201 *d++ = (U8)(((uv >> 54) & 0x3f) | 0x80);
202 *d++ = (U8)(((uv >> 48) & 0x3f) | 0x80);
203 *d++ = (U8)(((uv >> 42) & 0x3f) | 0x80);
204 *d++ = (U8)(((uv >> 36) & 0x3f) | 0x80);
205 *d++ = (U8)(((uv >> 30) & 0x3f) | 0x80);
206 *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80);
207 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
208 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
209 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
210 *d++ = (U8)(( uv & 0x3f) | 0x80);
214 #endif /* Loop style */
219 Tests if some arbitrary number of bytes begins in a valid UTF-8
220 character. Note that an INVARIANT (i.e. ASCII) character is a valid
221 UTF-8 character. The actual number of bytes in the UTF-8 character
222 will be returned if it is valid, otherwise 0.
224 This is the "slow" version as opposed to the "fast" version which is
225 the "unrolled" IS_UTF8_CHAR(). E.g. for t/uni/class.t the speed
226 difference is a factor of 2 to 3. For lengths (UTF8SKIP(s)) of four
227 or less you should use the IS_UTF8_CHAR(), for lengths of five or more
228 you should use the _slow(). In practice this means that the _slow()
229 will be used very rarely, since the maximum Unicode code point (as of
230 Unicode 4.1) is U+10FFFF, which encodes in UTF-8 to four bytes. Only
231 the "Perl extended UTF-8" (the infamous 'v-strings') will encode into
236 S_is_utf8_char_slow(const U8 *s, const STRLEN len)
242 PERL_ARGS_ASSERT_IS_UTF8_CHAR_SLOW;
244 if (UTF8_IS_INVARIANT(u))
247 if (!UTF8_IS_START(u))
250 if (len < 2 || !UTF8_IS_CONTINUATION(s[1]))
256 u = NATIVE_TO_UTF(u);
258 u &= UTF_START_MASK(len);
262 if (!UTF8_IS_CONTINUATION(*s))
264 uv = UTF8_ACCUMULATE(uv, *s);
271 if ((STRLEN)UNISKIP(uv) < len)
278 =for apidoc is_utf8_char
280 Tests if some arbitrary number of bytes begins in a valid UTF-8
281 character. Note that an INVARIANT (i.e. ASCII on non-EBCDIC machines)
282 character is a valid UTF-8 character. The actual number of bytes in the UTF-8
283 character will be returned if it is valid, otherwise 0.
287 Perl_is_utf8_char(const U8 *s)
289 const STRLEN len = UTF8SKIP(s);
291 PERL_ARGS_ASSERT_IS_UTF8_CHAR;
293 if (IS_UTF8_CHAR_FAST(len))
294 return IS_UTF8_CHAR(s, len) ? len : 0;
295 #endif /* #ifdef IS_UTF8_CHAR */
296 return is_utf8_char_slow(s, len);
301 =for apidoc is_utf8_string
303 Returns true if first C<len> bytes of the given string form a valid
304 UTF-8 string, false otherwise. Note that 'a valid UTF-8 string' does
305 not mean 'a string that contains code points above 0x7F encoded in UTF-8'
306 because a valid ASCII string is a valid UTF-8 string.
308 See also is_ascii_string(), is_utf8_string_loclen(), and is_utf8_string_loc().
314 Perl_is_utf8_string(const U8 *s, STRLEN len)
316 const U8* const send = s + (len ? len : strlen((const char *)s));
319 PERL_ARGS_ASSERT_IS_UTF8_STRING;
323 /* Inline the easy bits of is_utf8_char() here for speed... */
324 if (UTF8_IS_INVARIANT(*x))
326 else if (!UTF8_IS_START(*x))
329 /* ... and call is_utf8_char() only if really needed. */
332 if (IS_UTF8_CHAR_FAST(c)) {
333 if (!IS_UTF8_CHAR(x, c))
337 c = is_utf8_char_slow(x, c);
340 #endif /* #ifdef IS_UTF8_CHAR */
355 Implemented as a macro in utf8.h
357 =for apidoc is_utf8_string_loc
359 Like is_utf8_string() but stores the location of the failure (in the
360 case of "utf8ness failure") or the location s+len (in the case of
361 "utf8ness success") in the C<ep>.
363 See also is_utf8_string_loclen() and is_utf8_string().
365 =for apidoc is_utf8_string_loclen
367 Like is_utf8_string() but stores the location of the failure (in the
368 case of "utf8ness failure") or the location s+len (in the case of
369 "utf8ness success") in the C<ep>, and the number of UTF-8
370 encoded characters in the C<el>.
372 See also is_utf8_string_loc() and is_utf8_string().
378 Perl_is_utf8_string_loclen(const U8 *s, STRLEN len, const U8 **ep, STRLEN *el)
380 const U8* const send = s + (len ? len : strlen((const char *)s));
385 PERL_ARGS_ASSERT_IS_UTF8_STRING_LOCLEN;
388 /* Inline the easy bits of is_utf8_char() here for speed... */
389 if (UTF8_IS_INVARIANT(*x))
391 else if (!UTF8_IS_START(*x))
394 /* ... and call is_utf8_char() only if really needed. */
397 if (IS_UTF8_CHAR_FAST(c)) {
398 if (!IS_UTF8_CHAR(x, c))
401 c = is_utf8_char_slow(x, c);
404 #endif /* #ifdef IS_UTF8_CHAR */
423 =for apidoc utf8n_to_uvuni
425 Bottom level UTF-8 decode routine.
426 Returns the Unicode code point value of the first character in the string C<s>
427 which is assumed to be in UTF-8 encoding and no longer than C<curlen>;
428 C<retlen> will be set to the length, in bytes, of that character.
430 If C<s> does not point to a well-formed UTF-8 character, the behaviour
431 is dependent on the value of C<flags>: if it contains UTF8_CHECK_ONLY,
432 it is assumed that the caller will raise a warning, and this function
433 will silently just set C<retlen> to C<-1> and return zero. If the
434 C<flags> does not contain UTF8_CHECK_ONLY, warnings about
435 malformations will be given, C<retlen> will be set to the expected
436 length of the UTF-8 character in bytes, and zero will be returned.
438 The C<flags> can also contain various flags to allow deviations from
439 the strict UTF-8 encoding (see F<utf8.h>).
441 Most code should use utf8_to_uvchr() rather than call this directly.
447 Perl_utf8n_to_uvuni(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen, U32 flags)
450 const U8 * const s0 = s;
453 const bool dowarn = ckWARN_d(WARN_UTF8);
454 const UV startbyte = *s;
455 STRLEN expectlen = 0;
458 PERL_ARGS_ASSERT_UTF8N_TO_UVUNI;
460 /* This list is a superset of the UTF8_ALLOW_XXX. */
462 #define UTF8_WARN_EMPTY 1
463 #define UTF8_WARN_CONTINUATION 2
464 #define UTF8_WARN_NON_CONTINUATION 3
465 #define UTF8_WARN_FE_FF 4
466 #define UTF8_WARN_SHORT 5
467 #define UTF8_WARN_OVERFLOW 6
468 #define UTF8_WARN_SURROGATE 7
469 #define UTF8_WARN_LONG 8
470 #define UTF8_WARN_FFFF 9 /* Also FFFE. */
473 !(flags & UTF8_ALLOW_EMPTY)) {
474 warning = UTF8_WARN_EMPTY;
478 if (UTF8_IS_INVARIANT(uv)) {
481 return (UV) (NATIVE_TO_UTF(*s));
484 if (UTF8_IS_CONTINUATION(uv) &&
485 !(flags & UTF8_ALLOW_CONTINUATION)) {
486 warning = UTF8_WARN_CONTINUATION;
490 if (UTF8_IS_START(uv) && curlen > 1 && !UTF8_IS_CONTINUATION(s[1]) &&
491 !(flags & UTF8_ALLOW_NON_CONTINUATION)) {
492 warning = UTF8_WARN_NON_CONTINUATION;
497 uv = NATIVE_TO_UTF(uv);
499 if ((uv == 0xfe || uv == 0xff) &&
500 !(flags & UTF8_ALLOW_FE_FF)) {
501 warning = UTF8_WARN_FE_FF;
506 if (!(uv & 0x20)) { len = 2; uv &= 0x1f; }
507 else if (!(uv & 0x10)) { len = 3; uv &= 0x0f; }
508 else if (!(uv & 0x08)) { len = 4; uv &= 0x07; }
509 else if (!(uv & 0x04)) { len = 5; uv &= 0x03; }
511 else if (!(uv & 0x02)) { len = 6; uv &= 0x01; }
512 else { len = 7; uv &= 0x01; }
514 else if (!(uv & 0x02)) { len = 6; uv &= 0x01; }
515 else if (!(uv & 0x01)) { len = 7; uv = 0; }
516 else { len = 13; uv = 0; } /* whoa! */
524 if ((curlen < expectlen) &&
525 !(flags & UTF8_ALLOW_SHORT)) {
526 warning = UTF8_WARN_SHORT;
535 if (!UTF8_IS_CONTINUATION(*s) &&
536 !(flags & UTF8_ALLOW_NON_CONTINUATION)) {
538 warning = UTF8_WARN_NON_CONTINUATION;
542 uv = UTF8_ACCUMULATE(uv, *s);
544 /* These cannot be allowed. */
546 if (expectlen != 13 && !(flags & UTF8_ALLOW_LONG)) {
547 warning = UTF8_WARN_LONG;
551 else { /* uv < ouv */
552 /* This cannot be allowed. */
553 warning = UTF8_WARN_OVERFLOW;
561 if (UNICODE_IS_SURROGATE(uv) &&
562 !(flags & UTF8_ALLOW_SURROGATE)) {
563 warning = UTF8_WARN_SURROGATE;
565 } else if ((expectlen > (STRLEN)UNISKIP(uv)) &&
566 !(flags & UTF8_ALLOW_LONG)) {
567 warning = UTF8_WARN_LONG;
569 } else if (UNICODE_IS_ILLEGAL(uv) &&
570 !(flags & UTF8_ALLOW_FFFF)) {
571 warning = UTF8_WARN_FFFF;
579 if (flags & UTF8_CHECK_ONLY) {
581 *retlen = ((STRLEN) -1);
586 SV* const sv = newSVpvs_flags("Malformed UTF-8 character ", SVs_TEMP);
589 case 0: /* Intentionally empty. */ break;
590 case UTF8_WARN_EMPTY:
591 sv_catpvs(sv, "(empty string)");
593 case UTF8_WARN_CONTINUATION:
594 Perl_sv_catpvf(aTHX_ sv, "(unexpected continuation byte 0x%02"UVxf", with no preceding start byte)", uv);
596 case UTF8_WARN_NON_CONTINUATION:
598 Perl_sv_catpvf(aTHX_ sv, "(unexpected non-continuation byte 0x%02"UVxf", immediately after start byte 0x%02"UVxf")",
599 (UV)s[1], startbyte);
601 const int len = (int)(s-s0);
602 Perl_sv_catpvf(aTHX_ sv, "(unexpected non-continuation byte 0x%02"UVxf", %d byte%s after start byte 0x%02"UVxf", expected %d bytes)",
603 (UV)s[1], len, len > 1 ? "s" : "", startbyte, (int)expectlen);
607 case UTF8_WARN_FE_FF:
608 Perl_sv_catpvf(aTHX_ sv, "(byte 0x%02"UVxf")", uv);
610 case UTF8_WARN_SHORT:
611 Perl_sv_catpvf(aTHX_ sv, "(%d byte%s, need %d, after start byte 0x%02"UVxf")",
612 (int)curlen, curlen == 1 ? "" : "s", (int)expectlen, startbyte);
613 expectlen = curlen; /* distance for caller to skip */
615 case UTF8_WARN_OVERFLOW:
616 Perl_sv_catpvf(aTHX_ sv, "(overflow at 0x%"UVxf", byte 0x%02x, after start byte 0x%02"UVxf")",
619 case UTF8_WARN_SURROGATE:
620 Perl_sv_catpvf(aTHX_ sv, "(UTF-16 surrogate 0x%04"UVxf")", uv);
623 Perl_sv_catpvf(aTHX_ sv, "(%d byte%s, need %d, after start byte 0x%02"UVxf")",
624 (int)expectlen, expectlen == 1 ? "": "s", UNISKIP(uv), startbyte);
627 Perl_sv_catpvf(aTHX_ sv, "(character 0x%04"UVxf")", uv);
630 sv_catpvs(sv, "(unknown reason)");
635 const char * const s = SvPVX_const(sv);
638 Perl_warner(aTHX_ packWARN(WARN_UTF8),
639 "%s in %s", s, OP_DESC(PL_op));
641 Perl_warner(aTHX_ packWARN(WARN_UTF8), "%s", s);
646 *retlen = expectlen ? expectlen : len;
652 =for apidoc utf8_to_uvchr
654 Returns the native character value of the first character in the string C<s>
655 which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
656 length, in bytes, of that character.
658 If C<s> does not point to a well-formed UTF-8 character, zero is
659 returned and retlen is set, if possible, to -1.
665 Perl_utf8_to_uvchr(pTHX_ const U8 *s, STRLEN *retlen)
667 PERL_ARGS_ASSERT_UTF8_TO_UVCHR;
669 return utf8n_to_uvchr(s, UTF8_MAXBYTES, retlen,
670 ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
674 =for apidoc utf8_to_uvuni
676 Returns the Unicode code point of the first character in the string C<s>
677 which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
678 length, in bytes, of that character.
680 This function should only be used when the returned UV is considered
681 an index into the Unicode semantic tables (e.g. swashes).
683 If C<s> does not point to a well-formed UTF-8 character, zero is
684 returned and retlen is set, if possible, to -1.
690 Perl_utf8_to_uvuni(pTHX_ const U8 *s, STRLEN *retlen)
692 PERL_ARGS_ASSERT_UTF8_TO_UVUNI;
694 /* Call the low level routine asking for checks */
695 return Perl_utf8n_to_uvuni(aTHX_ s, UTF8_MAXBYTES, retlen,
696 ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
700 =for apidoc utf8_length
702 Return the length of the UTF-8 char encoded string C<s> in characters.
703 Stops at C<e> (inclusive). If C<e E<lt> s> or if the scan would end
704 up past C<e>, croaks.
710 Perl_utf8_length(pTHX_ const U8 *s, const U8 *e)
715 PERL_ARGS_ASSERT_UTF8_LENGTH;
717 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g.
718 * the bitops (especially ~) can create illegal UTF-8.
719 * In other words: in Perl UTF-8 is not just for Unicode. */
722 goto warn_and_return;
724 if (!UTF8_IS_INVARIANT(*s))
734 if (ckWARN_d(WARN_UTF8)) {
736 Perl_warner(aTHX_ packWARN(WARN_UTF8),
737 "%s in %s", unees, OP_DESC(PL_op));
739 Perl_warner(aTHX_ packWARN(WARN_UTF8), unees);
747 =for apidoc utf8_distance
749 Returns the number of UTF-8 characters between the UTF-8 pointers C<a>
752 WARNING: use only if you *know* that the pointers point inside the
759 Perl_utf8_distance(pTHX_ const U8 *a, const U8 *b)
761 PERL_ARGS_ASSERT_UTF8_DISTANCE;
763 return (a < b) ? -1 * (IV) utf8_length(a, b) : (IV) utf8_length(b, a);
769 Return the UTF-8 pointer C<s> displaced by C<off> characters, either
772 WARNING: do not use the following unless you *know* C<off> is within
773 the UTF-8 data pointed to by C<s> *and* that on entry C<s> is aligned
774 on the first byte of character or just after the last byte of a character.
780 Perl_utf8_hop(pTHX_ const U8 *s, I32 off)
782 PERL_ARGS_ASSERT_UTF8_HOP;
785 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g
786 * the bitops (especially ~) can create illegal UTF-8.
787 * In other words: in Perl UTF-8 is not just for Unicode. */
796 while (UTF8_IS_CONTINUATION(*s))
804 =for apidoc utf8_to_bytes
806 Converts a string C<s> of length C<len> from UTF-8 into native byte encoding.
807 Unlike C<bytes_to_utf8>, this over-writes the original string, and
808 updates len to contain the new length.
809 Returns zero on failure, setting C<len> to -1.
811 If you need a copy of the string, see C<bytes_from_utf8>.
817 Perl_utf8_to_bytes(pTHX_ U8 *s, STRLEN *len)
820 U8 * const send = s + *len;
823 PERL_ARGS_ASSERT_UTF8_TO_BYTES;
825 /* ensure valid UTF-8 and chars < 256 before updating string */
829 if (!UTF8_IS_INVARIANT(c) &&
830 (!UTF8_IS_DOWNGRADEABLE_START(c) || (s >= send)
831 || !(c = *s++) || !UTF8_IS_CONTINUATION(c))) {
832 *len = ((STRLEN) -1);
840 *d++ = (U8)utf8_to_uvchr(s, &ulen);
849 =for apidoc bytes_from_utf8
851 Converts a string C<s> of length C<len> from UTF-8 into native byte encoding.
852 Unlike C<utf8_to_bytes> but like C<bytes_to_utf8>, returns a pointer to
853 the newly-created string, and updates C<len> to contain the new
854 length. Returns the original string if no conversion occurs, C<len>
855 is unchanged. Do nothing if C<is_utf8> points to 0. Sets C<is_utf8> to
856 0 if C<s> is converted or consisted entirely of characters that are invariant
857 in utf8 (i.e., US-ASCII on non-EBCDIC machines).
863 Perl_bytes_from_utf8(pTHX_ const U8 *s, STRLEN *len, bool *is_utf8)
870 PERL_ARGS_ASSERT_BYTES_FROM_UTF8;
876 /* ensure valid UTF-8 and chars < 256 before converting string */
877 for (send = s + *len; s < send;) {
879 if (!UTF8_IS_INVARIANT(c)) {
880 if (UTF8_IS_DOWNGRADEABLE_START(c) && s < send &&
881 (c = *s++) && UTF8_IS_CONTINUATION(c))
890 Newx(d, (*len) - count + 1, U8);
891 s = start; start = d;
894 if (!UTF8_IS_INVARIANT(c)) {
895 /* Then it is two-byte encoded */
896 c = UTF8_ACCUMULATE(NATIVE_TO_UTF(c), *s++);
897 c = ASCII_TO_NATIVE(c);
907 =for apidoc bytes_to_utf8
909 Converts a string C<s> of length C<len> from the native encoding into UTF-8.
910 Returns a pointer to the newly-created string, and sets C<len> to
911 reflect the new length.
913 A NUL character will be written after the end of the string.
915 If you want to convert to UTF-8 from encodings other than
916 the native (Latin1 or EBCDIC),
917 see sv_recode_to_utf8().
923 Perl_bytes_to_utf8(pTHX_ const U8 *s, STRLEN *len)
925 const U8 * const send = s + (*len);
929 PERL_ARGS_ASSERT_BYTES_TO_UTF8;
932 Newx(d, (*len) * 2 + 1, U8);
936 const UV uv = NATIVE_TO_ASCII(*s++);
937 if (UNI_IS_INVARIANT(uv))
938 *d++ = (U8)UTF_TO_NATIVE(uv);
940 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
941 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
950 * Convert native (big-endian) or reversed (little-endian) UTF-16 to UTF-8.
952 * Destination must be pre-extended to 3/2 source. Do not use in-place.
953 * We optimize for native, for obvious reasons. */
956 Perl_utf16_to_utf8(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen)
961 PERL_ARGS_ASSERT_UTF16_TO_UTF8;
963 if (bytelen == 1 && p[0] == 0) { /* Be understanding. */
970 Perl_croak(aTHX_ "panic: utf16_to_utf8: odd bytelen %"UVuf, (UV)bytelen);
975 UV uv = (p[0] << 8) + p[1]; /* UTF-16BE */
979 *d++ = UNI_TO_NATIVE(uv);
986 *d++ = (U8)(( uv >> 6) | 0xc0);
987 *d++ = (U8)(( uv & 0x3f) | 0x80);
990 if (uv >= 0xd800 && uv < 0xdbff) { /* surrogates */
991 UV low = (p[0] << 8) + p[1];
993 if (low < 0xdc00 || low >= 0xdfff)
994 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
995 uv = ((uv - 0xd800) << 10) + (low - 0xdc00) + 0x10000;
998 *d++ = (U8)(( uv >> 12) | 0xe0);
999 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
1000 *d++ = (U8)(( uv & 0x3f) | 0x80);
1004 *d++ = (U8)(( uv >> 18) | 0xf0);
1005 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
1006 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
1007 *d++ = (U8)(( uv & 0x3f) | 0x80);
1011 *newlen = d - dstart;
1015 /* Note: this one is slightly destructive of the source. */
1018 Perl_utf16_to_utf8_reversed(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen)
1021 U8* const send = s + bytelen;
1023 PERL_ARGS_ASSERT_UTF16_TO_UTF8_REVERSED;
1026 const U8 tmp = s[0];
1031 return utf16_to_utf8(p, d, bytelen, newlen);
1034 /* for now these are all defined (inefficiently) in terms of the utf8 versions */
1037 Perl_is_uni_alnum(pTHX_ UV c)
1039 U8 tmpbuf[UTF8_MAXBYTES+1];
1040 uvchr_to_utf8(tmpbuf, c);
1041 return is_utf8_alnum(tmpbuf);
1045 Perl_is_uni_alnumc(pTHX_ UV c)
1047 U8 tmpbuf[UTF8_MAXBYTES+1];
1048 uvchr_to_utf8(tmpbuf, c);
1049 return is_utf8_alnumc(tmpbuf);
1053 Perl_is_uni_idfirst(pTHX_ UV c)
1055 U8 tmpbuf[UTF8_MAXBYTES+1];
1056 uvchr_to_utf8(tmpbuf, c);
1057 return is_utf8_idfirst(tmpbuf);
1061 Perl_is_uni_alpha(pTHX_ UV c)
1063 U8 tmpbuf[UTF8_MAXBYTES+1];
1064 uvchr_to_utf8(tmpbuf, c);
1065 return is_utf8_alpha(tmpbuf);
1069 Perl_is_uni_ascii(pTHX_ UV c)
1071 U8 tmpbuf[UTF8_MAXBYTES+1];
1072 uvchr_to_utf8(tmpbuf, c);
1073 return is_utf8_ascii(tmpbuf);
1077 Perl_is_uni_space(pTHX_ UV c)
1079 U8 tmpbuf[UTF8_MAXBYTES+1];
1080 uvchr_to_utf8(tmpbuf, c);
1081 return is_utf8_space(tmpbuf);
1085 Perl_is_uni_digit(pTHX_ UV c)
1087 U8 tmpbuf[UTF8_MAXBYTES+1];
1088 uvchr_to_utf8(tmpbuf, c);
1089 return is_utf8_digit(tmpbuf);
1093 Perl_is_uni_upper(pTHX_ UV c)
1095 U8 tmpbuf[UTF8_MAXBYTES+1];
1096 uvchr_to_utf8(tmpbuf, c);
1097 return is_utf8_upper(tmpbuf);
1101 Perl_is_uni_lower(pTHX_ UV c)
1103 U8 tmpbuf[UTF8_MAXBYTES+1];
1104 uvchr_to_utf8(tmpbuf, c);
1105 return is_utf8_lower(tmpbuf);
1109 Perl_is_uni_cntrl(pTHX_ UV c)
1111 U8 tmpbuf[UTF8_MAXBYTES+1];
1112 uvchr_to_utf8(tmpbuf, c);
1113 return is_utf8_cntrl(tmpbuf);
1117 Perl_is_uni_graph(pTHX_ UV c)
1119 U8 tmpbuf[UTF8_MAXBYTES+1];
1120 uvchr_to_utf8(tmpbuf, c);
1121 return is_utf8_graph(tmpbuf);
1125 Perl_is_uni_print(pTHX_ UV c)
1127 U8 tmpbuf[UTF8_MAXBYTES+1];
1128 uvchr_to_utf8(tmpbuf, c);
1129 return is_utf8_print(tmpbuf);
1133 Perl_is_uni_punct(pTHX_ UV c)
1135 U8 tmpbuf[UTF8_MAXBYTES+1];
1136 uvchr_to_utf8(tmpbuf, c);
1137 return is_utf8_punct(tmpbuf);
1141 Perl_is_uni_xdigit(pTHX_ UV c)
1143 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1144 uvchr_to_utf8(tmpbuf, c);
1145 return is_utf8_xdigit(tmpbuf);
1149 Perl_to_uni_upper(pTHX_ UV c, U8* p, STRLEN *lenp)
1151 PERL_ARGS_ASSERT_TO_UNI_UPPER;
1153 uvchr_to_utf8(p, c);
1154 return to_utf8_upper(p, p, lenp);
1158 Perl_to_uni_title(pTHX_ UV c, U8* p, STRLEN *lenp)
1160 PERL_ARGS_ASSERT_TO_UNI_TITLE;
1162 uvchr_to_utf8(p, c);
1163 return to_utf8_title(p, p, lenp);
1167 Perl_to_uni_lower(pTHX_ UV c, U8* p, STRLEN *lenp)
1169 PERL_ARGS_ASSERT_TO_UNI_LOWER;
1171 uvchr_to_utf8(p, c);
1172 return to_utf8_lower(p, p, lenp);
1176 Perl_to_uni_fold(pTHX_ UV c, U8* p, STRLEN *lenp)
1178 PERL_ARGS_ASSERT_TO_UNI_FOLD;
1180 uvchr_to_utf8(p, c);
1181 return to_utf8_fold(p, p, lenp);
1184 /* for now these all assume no locale info available for Unicode > 255 */
1187 Perl_is_uni_alnum_lc(pTHX_ UV c)
1189 return is_uni_alnum(c); /* XXX no locale support yet */
1193 Perl_is_uni_alnumc_lc(pTHX_ UV c)
1195 return is_uni_alnumc(c); /* XXX no locale support yet */
1199 Perl_is_uni_idfirst_lc(pTHX_ UV c)
1201 return is_uni_idfirst(c); /* XXX no locale support yet */
1205 Perl_is_uni_alpha_lc(pTHX_ UV c)
1207 return is_uni_alpha(c); /* XXX no locale support yet */
1211 Perl_is_uni_ascii_lc(pTHX_ UV c)
1213 return is_uni_ascii(c); /* XXX no locale support yet */
1217 Perl_is_uni_space_lc(pTHX_ UV c)
1219 return is_uni_space(c); /* XXX no locale support yet */
1223 Perl_is_uni_digit_lc(pTHX_ UV c)
1225 return is_uni_digit(c); /* XXX no locale support yet */
1229 Perl_is_uni_upper_lc(pTHX_ UV c)
1231 return is_uni_upper(c); /* XXX no locale support yet */
1235 Perl_is_uni_lower_lc(pTHX_ UV c)
1237 return is_uni_lower(c); /* XXX no locale support yet */
1241 Perl_is_uni_cntrl_lc(pTHX_ UV c)
1243 return is_uni_cntrl(c); /* XXX no locale support yet */
1247 Perl_is_uni_graph_lc(pTHX_ UV c)
1249 return is_uni_graph(c); /* XXX no locale support yet */
1253 Perl_is_uni_print_lc(pTHX_ UV c)
1255 return is_uni_print(c); /* XXX no locale support yet */
1259 Perl_is_uni_punct_lc(pTHX_ UV c)
1261 return is_uni_punct(c); /* XXX no locale support yet */
1265 Perl_is_uni_xdigit_lc(pTHX_ UV c)
1267 return is_uni_xdigit(c); /* XXX no locale support yet */
1271 Perl_to_uni_upper_lc(pTHX_ U32 c)
1273 /* XXX returns only the first character -- do not use XXX */
1274 /* XXX no locale support yet */
1276 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1277 return (U32)to_uni_upper(c, tmpbuf, &len);
1281 Perl_to_uni_title_lc(pTHX_ U32 c)
1283 /* XXX returns only the first character XXX -- do not use XXX */
1284 /* XXX no locale support yet */
1286 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1287 return (U32)to_uni_title(c, tmpbuf, &len);
1291 Perl_to_uni_lower_lc(pTHX_ U32 c)
1293 /* XXX returns only the first character -- do not use XXX */
1294 /* XXX no locale support yet */
1296 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1297 return (U32)to_uni_lower(c, tmpbuf, &len);
1301 S_is_utf8_common(pTHX_ const U8 *const p, SV **swash,
1302 const char *const swashname)
1306 PERL_ARGS_ASSERT_IS_UTF8_COMMON;
1308 if (!is_utf8_char(p))
1311 *swash = swash_init("utf8", swashname, &PL_sv_undef, 1, 0);
1312 return swash_fetch(*swash, p, TRUE) != 0;
1316 Perl_is_utf8_alnum(pTHX_ const U8 *p)
1320 PERL_ARGS_ASSERT_IS_UTF8_ALNUM;
1322 /* NOTE: "IsWord", not "IsAlnum", since Alnum is a true
1323 * descendant of isalnum(3), in other words, it doesn't
1324 * contain the '_'. --jhi */
1325 return is_utf8_common(p, &PL_utf8_alnum, "IsWord");
1329 Perl_is_utf8_alnumc(pTHX_ const U8 *p)
1333 PERL_ARGS_ASSERT_IS_UTF8_ALNUMC;
1335 return is_utf8_common(p, &PL_utf8_alnumc, "IsAlnumC");
1339 Perl_is_utf8_idfirst(pTHX_ const U8 *p) /* The naming is historical. */
1343 PERL_ARGS_ASSERT_IS_UTF8_IDFIRST;
1347 /* is_utf8_idstart would be more logical. */
1348 return is_utf8_common(p, &PL_utf8_idstart, "IdStart");
1352 Perl_is_utf8_idcont(pTHX_ const U8 *p)
1356 PERL_ARGS_ASSERT_IS_UTF8_IDCONT;
1360 return is_utf8_common(p, &PL_utf8_idcont, "IdContinue");
1364 Perl_is_utf8_alpha(pTHX_ const U8 *p)
1368 PERL_ARGS_ASSERT_IS_UTF8_ALPHA;
1370 return is_utf8_common(p, &PL_utf8_alpha, "IsAlpha");
1374 Perl_is_utf8_ascii(pTHX_ const U8 *p)
1378 PERL_ARGS_ASSERT_IS_UTF8_ASCII;
1380 return is_utf8_common(p, &PL_utf8_ascii, "IsAscii");
1384 Perl_is_utf8_space(pTHX_ const U8 *p)
1388 PERL_ARGS_ASSERT_IS_UTF8_SPACE;
1390 return is_utf8_common(p, &PL_utf8_space, "IsSpacePerl");
1394 Perl_is_utf8_digit(pTHX_ const U8 *p)
1398 PERL_ARGS_ASSERT_IS_UTF8_DIGIT;
1400 return is_utf8_common(p, &PL_utf8_digit, "IsDigit");
1404 Perl_is_utf8_upper(pTHX_ const U8 *p)
1408 PERL_ARGS_ASSERT_IS_UTF8_UPPER;
1410 return is_utf8_common(p, &PL_utf8_upper, "IsUppercase");
1414 Perl_is_utf8_lower(pTHX_ const U8 *p)
1418 PERL_ARGS_ASSERT_IS_UTF8_LOWER;
1420 return is_utf8_common(p, &PL_utf8_lower, "IsLowercase");
1424 Perl_is_utf8_cntrl(pTHX_ const U8 *p)
1428 PERL_ARGS_ASSERT_IS_UTF8_CNTRL;
1430 return is_utf8_common(p, &PL_utf8_cntrl, "IsCntrl");
1434 Perl_is_utf8_graph(pTHX_ const U8 *p)
1438 PERL_ARGS_ASSERT_IS_UTF8_GRAPH;
1440 return is_utf8_common(p, &PL_utf8_graph, "IsGraph");
1444 Perl_is_utf8_print(pTHX_ const U8 *p)
1448 PERL_ARGS_ASSERT_IS_UTF8_PRINT;
1450 return is_utf8_common(p, &PL_utf8_print, "IsPrint");
1454 Perl_is_utf8_punct(pTHX_ const U8 *p)
1458 PERL_ARGS_ASSERT_IS_UTF8_PUNCT;
1460 return is_utf8_common(p, &PL_utf8_punct, "IsPunct");
1464 Perl_is_utf8_xdigit(pTHX_ const U8 *p)
1468 PERL_ARGS_ASSERT_IS_UTF8_XDIGIT;
1470 return is_utf8_common(p, &PL_utf8_xdigit, "Isxdigit");
1474 Perl_is_utf8_mark(pTHX_ const U8 *p)
1478 PERL_ARGS_ASSERT_IS_UTF8_MARK;
1480 return is_utf8_common(p, &PL_utf8_mark, "IsM");
1484 =for apidoc to_utf8_case
1486 The "p" contains the pointer to the UTF-8 string encoding
1487 the character that is being converted.
1489 The "ustrp" is a pointer to the character buffer to put the
1490 conversion result to. The "lenp" is a pointer to the length
1493 The "swashp" is a pointer to the swash to use.
1495 Both the special and normal mappings are stored lib/unicore/To/Foo.pl,
1496 and loaded by SWASHNEW, using lib/utf8_heavy.pl. The special (usually,
1497 but not always, a multicharacter mapping), is tried first.
1499 The "special" is a string like "utf8::ToSpecLower", which means the
1500 hash %utf8::ToSpecLower. The access to the hash is through
1501 Perl_to_utf8_case().
1503 The "normal" is a string like "ToLower" which means the swash
1509 Perl_to_utf8_case(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp,
1510 SV **swashp, const char *normal, const char *special)
1513 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1515 const UV uv0 = utf8_to_uvchr(p, NULL);
1516 /* The NATIVE_TO_UNI() and UNI_TO_NATIVE() mappings
1517 * are necessary in EBCDIC, they are redundant no-ops
1518 * in ASCII-ish platforms, and hopefully optimized away. */
1519 const UV uv1 = NATIVE_TO_UNI(uv0);
1521 PERL_ARGS_ASSERT_TO_UTF8_CASE;
1523 uvuni_to_utf8(tmpbuf, uv1);
1525 if (!*swashp) /* load on-demand */
1526 *swashp = swash_init("utf8", normal, &PL_sv_undef, 4, 0);
1528 /* The 0xDF is the only special casing Unicode code point below 0x100. */
1529 if (special && (uv1 == 0xDF || uv1 > 0xFF)) {
1530 /* It might be "special" (sometimes, but not always,
1531 * a multicharacter mapping) */
1532 HV * const hv = get_hv(special, 0);
1536 (svp = hv_fetch(hv, (const char*)tmpbuf, UNISKIP(uv1), FALSE)) &&
1540 s = SvPV_const(*svp, len);
1542 len = uvuni_to_utf8(ustrp, NATIVE_TO_UNI(*(U8*)s)) - ustrp;
1545 /* If we have EBCDIC we need to remap the characters
1546 * since any characters in the low 256 are Unicode
1547 * code points, not EBCDIC. */
1548 U8 *t = (U8*)s, *tend = t + len, *d;
1555 const UV c = utf8_to_uvchr(t, &tlen);
1557 d = uvchr_to_utf8(d, UNI_TO_NATIVE(c));
1566 d = uvchr_to_utf8(d, UNI_TO_NATIVE(*t));
1571 Copy(tmpbuf, ustrp, len, U8);
1573 Copy(s, ustrp, len, U8);
1579 if (!len && *swashp) {
1580 const UV uv2 = swash_fetch(*swashp, tmpbuf, TRUE);
1583 /* It was "normal" (a single character mapping). */
1584 const UV uv3 = UNI_TO_NATIVE(uv2);
1585 len = uvchr_to_utf8(ustrp, uv3) - ustrp;
1589 if (!len) /* Neither: just copy. */
1590 len = uvchr_to_utf8(ustrp, uv0) - ustrp;
1595 return len ? utf8_to_uvchr(ustrp, 0) : 0;
1599 =for apidoc to_utf8_upper
1601 Convert the UTF-8 encoded character at p to its uppercase version and
1602 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
1603 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since
1604 the uppercase version may be longer than the original character.
1606 The first character of the uppercased version is returned
1607 (but note, as explained above, that there may be more.)
1612 Perl_to_utf8_upper(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
1616 PERL_ARGS_ASSERT_TO_UTF8_UPPER;
1618 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
1619 &PL_utf8_toupper, "ToUpper", "utf8::ToSpecUpper");
1623 =for apidoc to_utf8_title
1625 Convert the UTF-8 encoded character at p to its titlecase version and
1626 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
1627 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
1628 titlecase version may be longer than the original character.
1630 The first character of the titlecased version is returned
1631 (but note, as explained above, that there may be more.)
1636 Perl_to_utf8_title(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
1640 PERL_ARGS_ASSERT_TO_UTF8_TITLE;
1642 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
1643 &PL_utf8_totitle, "ToTitle", "utf8::ToSpecTitle");
1647 =for apidoc to_utf8_lower
1649 Convert the UTF-8 encoded character at p to its lowercase version and
1650 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
1651 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
1652 lowercase version may be longer than the original character.
1654 The first character of the lowercased version is returned
1655 (but note, as explained above, that there may be more.)
1660 Perl_to_utf8_lower(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
1664 PERL_ARGS_ASSERT_TO_UTF8_LOWER;
1666 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
1667 &PL_utf8_tolower, "ToLower", "utf8::ToSpecLower");
1671 =for apidoc to_utf8_fold
1673 Convert the UTF-8 encoded character at p to its foldcase version and
1674 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
1675 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
1676 foldcase version may be longer than the original character (up to
1679 The first character of the foldcased version is returned
1680 (but note, as explained above, that there may be more.)
1685 Perl_to_utf8_fold(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
1689 PERL_ARGS_ASSERT_TO_UTF8_FOLD;
1691 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
1692 &PL_utf8_tofold, "ToFold", "utf8::ToSpecFold");
1696 * A "swash" is a swatch hash.
1697 * A "swatch" is a bit vector generated by utf8.c:S_swash_get().
1698 * C<pkg> is a pointer to a package name for SWASHNEW, should be "utf8".
1699 * For other parameters, see utf8::SWASHNEW in lib/utf8_heavy.pl.
1702 Perl_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv, I32 minbits, I32 none)
1707 const size_t pkg_len = strlen(pkg);
1708 const size_t name_len = strlen(name);
1709 HV * const stash = gv_stashpvn(pkg, pkg_len, 0);
1712 PERL_ARGS_ASSERT_SWASH_INIT;
1714 PUSHSTACKi(PERLSI_MAGIC);
1719 if (!gv_fetchmeth(stash, "SWASHNEW", 8, -1)) { /* demand load utf8 */
1721 errsv_save = newSVsv(ERRSV);
1722 /* It is assumed that callers of this routine are not passing in any
1723 user derived data. */
1724 /* Need to do this after save_re_context() as it will set PL_tainted to
1725 1 while saving $1 etc (see the code after getrx: in Perl_magic_get).
1726 Even line to create errsv_save can turn on PL_tainted. */
1727 SAVEBOOL(PL_tainted);
1729 Perl_load_module(aTHX_ PERL_LOADMOD_NOIMPORT, newSVpvn(pkg,pkg_len),
1732 sv_setsv(ERRSV, errsv_save);
1733 SvREFCNT_dec(errsv_save);
1739 mPUSHp(pkg, pkg_len);
1740 mPUSHp(name, name_len);
1745 errsv_save = newSVsv(ERRSV);
1746 if (call_method("SWASHNEW", G_SCALAR))
1747 retval = newSVsv(*PL_stack_sp--);
1749 retval = &PL_sv_undef;
1751 sv_setsv(ERRSV, errsv_save);
1752 SvREFCNT_dec(errsv_save);
1755 if (IN_PERL_COMPILETIME) {
1756 CopHINTS_set(PL_curcop, PL_hints);
1758 if (!SvROK(retval) || SvTYPE(SvRV(retval)) != SVt_PVHV) {
1760 Perl_croak(aTHX_ "Can't find Unicode property definition \"%"SVf"\"",
1762 Perl_croak(aTHX_ "SWASHNEW didn't return an HV ref");
1768 /* This API is wrong for special case conversions since we may need to
1769 * return several Unicode characters for a single Unicode character
1770 * (see lib/unicore/SpecCase.txt) The SWASHGET in lib/utf8_heavy.pl is
1771 * the lower-level routine, and it is similarly broken for returning
1772 * multiple values. --jhi */
1773 /* Now SWASHGET is recasted into S_swash_get in this file. */
1776 * Returns the value of property/mapping C<swash> for the first character
1777 * of the string C<ptr>. If C<do_utf8> is true, the string C<ptr> is
1778 * assumed to be in utf8. If C<do_utf8> is false, the string C<ptr> is
1779 * assumed to be in native 8-bit encoding. Caches the swatch in C<swash>.
1782 Perl_swash_fetch(pTHX_ SV *swash, const U8 *ptr, bool do_utf8)
1785 HV *const hv = MUTABLE_HV(SvRV(swash));
1790 const U8 *tmps = NULL;
1794 const UV c = NATIVE_TO_ASCII(*ptr);
1796 PERL_ARGS_ASSERT_SWASH_FETCH;
1798 if (!do_utf8 && !UNI_IS_INVARIANT(c)) {
1799 tmputf8[0] = (U8)UTF8_EIGHT_BIT_HI(c);
1800 tmputf8[1] = (U8)UTF8_EIGHT_BIT_LO(c);
1803 /* Given a UTF-X encoded char 0xAA..0xYY,0xZZ
1804 * then the "swatch" is a vec() for al the chars which start
1806 * So the key in the hash (klen) is length of encoded char -1
1808 klen = UTF8SKIP(ptr) - 1;
1812 /* If char in invariant then swatch is for all the invariant chars
1813 * In both UTF-8 and UTF-8-MOD that happens to be UTF_CONTINUATION_MARK
1815 needents = UTF_CONTINUATION_MARK;
1816 off = NATIVE_TO_UTF(ptr[klen]);
1819 /* If char is encoded then swatch is for the prefix */
1820 needents = (1 << UTF_ACCUMULATION_SHIFT);
1821 off = NATIVE_TO_UTF(ptr[klen]) & UTF_CONTINUATION_MASK;
1825 * This single-entry cache saves about 1/3 of the utf8 overhead in test
1826 * suite. (That is, only 7-8% overall over just a hash cache. Still,
1827 * it's nothing to sniff at.) Pity we usually come through at least
1828 * two function calls to get here...
1830 * NB: this code assumes that swatches are never modified, once generated!
1833 if (hv == PL_last_swash_hv &&
1834 klen == PL_last_swash_klen &&
1835 (!klen || memEQ((char *)ptr, (char *)PL_last_swash_key, klen)) )
1837 tmps = PL_last_swash_tmps;
1838 slen = PL_last_swash_slen;
1841 /* Try our second-level swatch cache, kept in a hash. */
1842 SV** svp = hv_fetch(hv, (const char*)ptr, klen, FALSE);
1844 /* If not cached, generate it via swash_get */
1845 if (!svp || !SvPOK(*svp)
1846 || !(tmps = (const U8*)SvPV_const(*svp, slen))) {
1847 /* We use utf8n_to_uvuni() as we want an index into
1848 Unicode tables, not a native character number.
1850 const UV code_point = utf8n_to_uvuni(ptr, UTF8_MAXBYTES, 0,
1852 0 : UTF8_ALLOW_ANY);
1853 swatch = swash_get(swash,
1854 /* On EBCDIC & ~(0xA0-1) isn't a useful thing to do */
1855 (klen) ? (code_point & ~(needents - 1)) : 0,
1858 if (IN_PERL_COMPILETIME)
1859 CopHINTS_set(PL_curcop, PL_hints);
1861 svp = hv_store(hv, (const char *)ptr, klen, swatch, 0);
1863 if (!svp || !(tmps = (U8*)SvPV(*svp, slen))
1864 || (slen << 3) < needents)
1865 Perl_croak(aTHX_ "panic: swash_fetch got improper swatch");
1868 PL_last_swash_hv = hv;
1869 assert(klen <= sizeof(PL_last_swash_key));
1870 PL_last_swash_klen = (U8)klen;
1871 /* FIXME change interpvar.h? */
1872 PL_last_swash_tmps = (U8 *) tmps;
1873 PL_last_swash_slen = slen;
1875 Copy(ptr, PL_last_swash_key, klen, U8);
1878 switch ((int)((slen << 3) / needents)) {
1880 bit = 1 << (off & 7);
1882 return (tmps[off] & bit) != 0;
1887 return (tmps[off] << 8) + tmps[off + 1] ;
1890 return (tmps[off] << 24) + (tmps[off+1] << 16) + (tmps[off+2] << 8) + tmps[off + 3] ;
1892 Perl_croak(aTHX_ "panic: swash_fetch got swatch of unexpected bit width");
1893 NORETURN_FUNCTION_END;
1897 * Returns a swatch (a bit vector string) for a code point sequence
1898 * that starts from the value C<start> and comprises the number C<span>.
1899 * A C<swash> must be an object created by SWASHNEW (see lib/utf8_heavy.pl).
1900 * Should be used via swash_fetch, which will cache the swatch in C<swash>.
1903 S_swash_get(pTHX_ SV* swash, UV start, UV span)
1906 U8 *l, *lend, *x, *xend, *s;
1907 STRLEN lcur, xcur, scur;
1908 HV *const hv = MUTABLE_HV(SvRV(swash));
1909 SV** const listsvp = hv_fetchs(hv, "LIST", FALSE);
1910 SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE);
1911 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
1912 SV** const nonesvp = hv_fetchs(hv, "NONE", FALSE);
1913 SV** const extssvp = hv_fetchs(hv, "EXTRAS", FALSE);
1914 const U8* const typestr = (U8*)SvPV_nolen(*typesvp);
1915 const int typeto = typestr[0] == 'T' && typestr[1] == 'o';
1916 const STRLEN bits = SvUV(*bitssvp);
1917 const STRLEN octets = bits >> 3; /* if bits == 1, then octets == 0 */
1918 const UV none = SvUV(*nonesvp);
1919 const UV end = start + span;
1921 PERL_ARGS_ASSERT_SWASH_GET;
1923 if (bits != 1 && bits != 8 && bits != 16 && bits != 32) {
1924 Perl_croak(aTHX_ "panic: swash_get doesn't expect bits %"UVuf,
1928 /* create and initialize $swatch */
1929 scur = octets ? (span * octets) : (span + 7) / 8;
1930 swatch = newSV(scur);
1932 s = (U8*)SvPVX(swatch);
1933 if (octets && none) {
1934 const U8* const e = s + scur;
1937 *s++ = (U8)(none & 0xff);
1938 else if (bits == 16) {
1939 *s++ = (U8)((none >> 8) & 0xff);
1940 *s++ = (U8)( none & 0xff);
1942 else if (bits == 32) {
1943 *s++ = (U8)((none >> 24) & 0xff);
1944 *s++ = (U8)((none >> 16) & 0xff);
1945 *s++ = (U8)((none >> 8) & 0xff);
1946 *s++ = (U8)( none & 0xff);
1952 (void)memzero((U8*)s, scur + 1);
1954 SvCUR_set(swatch, scur);
1955 s = (U8*)SvPVX(swatch);
1957 /* read $swash->{LIST} */
1958 l = (U8*)SvPV(*listsvp, lcur);
1963 I32 flags = PERL_SCAN_SILENT_ILLDIGIT | PERL_SCAN_DISALLOW_PREFIX;
1965 U8* const nl = (U8*)memchr(l, '\n', lend - l);
1968 min = grok_hex((char *)l, &numlen, &flags, NULL);
1972 l = nl + 1; /* 1 is length of "\n" */
1976 l = lend; /* to LIST's end at which \n is not found */
1982 flags = PERL_SCAN_SILENT_ILLDIGIT | PERL_SCAN_DISALLOW_PREFIX;
1984 max = grok_hex((char *)l, &numlen, &flags, NULL);
1993 flags = PERL_SCAN_SILENT_ILLDIGIT |
1994 PERL_SCAN_DISALLOW_PREFIX;
1996 val = grok_hex((char *)l, &numlen, &flags, NULL);
2005 Perl_croak(aTHX_ "%s: illegal mapping '%s'",
2011 val = 0; /* bits == 1, then val should be ignored */
2018 Perl_croak(aTHX_ "%s: illegal mapping '%s'", typestr, l);
2022 val = 0; /* bits == 1, then val should be ignored */
2036 if (!none || val < none) {
2041 for (key = min; key <= max; key++) {
2045 /* offset must be non-negative (start <= min <= key < end) */
2046 offset = octets * (key - start);
2048 s[offset] = (U8)(val & 0xff);
2049 else if (bits == 16) {
2050 s[offset ] = (U8)((val >> 8) & 0xff);
2051 s[offset + 1] = (U8)( val & 0xff);
2053 else if (bits == 32) {
2054 s[offset ] = (U8)((val >> 24) & 0xff);
2055 s[offset + 1] = (U8)((val >> 16) & 0xff);
2056 s[offset + 2] = (U8)((val >> 8) & 0xff);
2057 s[offset + 3] = (U8)( val & 0xff);
2060 if (!none || val < none)
2064 else { /* bits == 1, then val should be ignored */
2068 for (key = min; key <= max; key++) {
2069 const STRLEN offset = (STRLEN)(key - start);
2072 s[offset >> 3] |= 1 << (offset & 7);
2078 /* read $swash->{EXTRAS} */
2079 x = (U8*)SvPV(*extssvp, xcur);
2087 SV **otherbitssvp, *other;
2091 const U8 opc = *x++;
2095 nl = (U8*)memchr(x, '\n', xend - x);
2097 if (opc != '-' && opc != '+' && opc != '!' && opc != '&') {
2099 x = nl + 1; /* 1 is length of "\n" */
2103 x = xend; /* to EXTRAS' end at which \n is not found */
2110 namelen = nl - namestr;
2114 namelen = xend - namestr;
2118 othersvp = hv_fetch(hv, (char *)namestr, namelen, FALSE);
2119 otherhv = MUTABLE_HV(SvRV(*othersvp));
2120 otherbitssvp = hv_fetchs(otherhv, "BITS", FALSE);
2121 otherbits = (STRLEN)SvUV(*otherbitssvp);
2122 if (bits < otherbits)
2123 Perl_croak(aTHX_ "panic: swash_get found swatch size mismatch");
2125 /* The "other" swatch must be destroyed after. */
2126 other = swash_get(*othersvp, start, span);
2127 o = (U8*)SvPV(other, olen);
2130 Perl_croak(aTHX_ "panic: swash_get got improper swatch");
2132 s = (U8*)SvPV(swatch, slen);
2133 if (bits == 1 && otherbits == 1) {
2135 Perl_croak(aTHX_ "panic: swash_get found swatch length mismatch");
2159 STRLEN otheroctets = otherbits >> 3;
2161 U8* const send = s + slen;
2166 if (otherbits == 1) {
2167 otherval = (o[offset >> 3] >> (offset & 7)) & 1;
2171 STRLEN vlen = otheroctets;
2179 if (opc == '+' && otherval)
2180 NOOP; /* replace with otherval */
2181 else if (opc == '!' && !otherval)
2183 else if (opc == '-' && otherval)
2185 else if (opc == '&' && !otherval)
2188 s += octets; /* no replacement */
2193 *s++ = (U8)( otherval & 0xff);
2194 else if (bits == 16) {
2195 *s++ = (U8)((otherval >> 8) & 0xff);
2196 *s++ = (U8)( otherval & 0xff);
2198 else if (bits == 32) {
2199 *s++ = (U8)((otherval >> 24) & 0xff);
2200 *s++ = (U8)((otherval >> 16) & 0xff);
2201 *s++ = (U8)((otherval >> 8) & 0xff);
2202 *s++ = (U8)( otherval & 0xff);
2206 sv_free(other); /* through with it! */
2212 =for apidoc uvchr_to_utf8
2214 Adds the UTF-8 representation of the Native codepoint C<uv> to the end
2215 of the string C<d>; C<d> should be have at least C<UTF8_MAXBYTES+1> free
2216 bytes available. The return value is the pointer to the byte after the
2217 end of the new character. In other words,
2219 d = uvchr_to_utf8(d, uv);
2221 is the recommended wide native character-aware way of saying
2228 /* On ASCII machines this is normally a macro but we want a
2229 real function in case XS code wants it
2232 Perl_uvchr_to_utf8(pTHX_ U8 *d, UV uv)
2234 PERL_ARGS_ASSERT_UVCHR_TO_UTF8;
2236 return Perl_uvuni_to_utf8_flags(aTHX_ d, NATIVE_TO_UNI(uv), 0);
2240 Perl_uvchr_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags)
2242 PERL_ARGS_ASSERT_UVCHR_TO_UTF8_FLAGS;
2244 return Perl_uvuni_to_utf8_flags(aTHX_ d, NATIVE_TO_UNI(uv), flags);
2248 =for apidoc utf8n_to_uvchr
2251 Returns the native character value of the first character in the string
2253 which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
2254 length, in bytes, of that character.
2256 Allows length and flags to be passed to low level routine.
2260 /* On ASCII machines this is normally a macro but we want
2261 a real function in case XS code wants it
2264 Perl_utf8n_to_uvchr(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen,
2267 const UV uv = Perl_utf8n_to_uvuni(aTHX_ s, curlen, retlen, flags);
2269 PERL_ARGS_ASSERT_UTF8N_TO_UVCHR;
2271 return UNI_TO_NATIVE(uv);
2275 =for apidoc pv_uni_display
2277 Build to the scalar dsv a displayable version of the string spv,
2278 length len, the displayable version being at most pvlim bytes long
2279 (if longer, the rest is truncated and "..." will be appended).
2281 The flags argument can have UNI_DISPLAY_ISPRINT set to display
2282 isPRINT()able characters as themselves, UNI_DISPLAY_BACKSLASH
2283 to display the \\[nrfta\\] as the backslashed versions (like '\n')
2284 (UNI_DISPLAY_BACKSLASH is preferred over UNI_DISPLAY_ISPRINT for \\).
2285 UNI_DISPLAY_QQ (and its alias UNI_DISPLAY_REGEX) have both
2286 UNI_DISPLAY_BACKSLASH and UNI_DISPLAY_ISPRINT turned on.
2288 The pointer to the PV of the dsv is returned.
2292 Perl_pv_uni_display(pTHX_ SV *dsv, const U8 *spv, STRLEN len, STRLEN pvlim, UV flags)
2297 PERL_ARGS_ASSERT_PV_UNI_DISPLAY;
2301 for (s = (const char *)spv, e = s + len; s < e; s += UTF8SKIP(s)) {
2303 /* This serves double duty as a flag and a character to print after
2304 a \ when flags & UNI_DISPLAY_BACKSLASH is true.
2308 if (pvlim && SvCUR(dsv) >= pvlim) {
2312 u = utf8_to_uvchr((U8*)s, 0);
2314 const unsigned char c = (unsigned char)u & 0xFF;
2315 if (flags & UNI_DISPLAY_BACKSLASH) {
2332 const char string = ok;
2333 sv_catpvs(dsv, "\\");
2334 sv_catpvn(dsv, &string, 1);
2337 /* isPRINT() is the locale-blind version. */
2338 if (!ok && (flags & UNI_DISPLAY_ISPRINT) && isPRINT(c)) {
2339 const char string = c;
2340 sv_catpvn(dsv, &string, 1);
2345 Perl_sv_catpvf(aTHX_ dsv, "\\x{%"UVxf"}", u);
2348 sv_catpvs(dsv, "...");
2354 =for apidoc sv_uni_display
2356 Build to the scalar dsv a displayable version of the scalar sv,
2357 the displayable version being at most pvlim bytes long
2358 (if longer, the rest is truncated and "..." will be appended).
2360 The flags argument is as in pv_uni_display().
2362 The pointer to the PV of the dsv is returned.
2367 Perl_sv_uni_display(pTHX_ SV *dsv, SV *ssv, STRLEN pvlim, UV flags)
2369 PERL_ARGS_ASSERT_SV_UNI_DISPLAY;
2371 return Perl_pv_uni_display(aTHX_ dsv, (const U8*)SvPVX_const(ssv),
2372 SvCUR(ssv), pvlim, flags);
2376 =for apidoc ibcmp_utf8
2378 Return true if the strings s1 and s2 differ case-insensitively, false
2379 if not (if they are equal case-insensitively). If u1 is true, the
2380 string s1 is assumed to be in UTF-8-encoded Unicode. If u2 is true,
2381 the string s2 is assumed to be in UTF-8-encoded Unicode. If u1 or u2
2382 are false, the respective string is assumed to be in native 8-bit
2385 If the pe1 and pe2 are non-NULL, the scanning pointers will be copied
2386 in there (they will point at the beginning of the I<next> character).
2387 If the pointers behind pe1 or pe2 are non-NULL, they are the end
2388 pointers beyond which scanning will not continue under any
2389 circumstances. If the byte lengths l1 and l2 are non-zero, s1+l1 and
2390 s2+l2 will be used as goal end pointers that will also stop the scan,
2391 and which qualify towards defining a successful match: all the scans
2392 that define an explicit length must reach their goal pointers for
2393 a match to succeed).
2395 For case-insensitiveness, the "casefolding" of Unicode is used
2396 instead of upper/lowercasing both the characters, see
2397 http://www.unicode.org/unicode/reports/tr21/ (Case Mappings).
2401 Perl_ibcmp_utf8(pTHX_ const char *s1, char **pe1, register UV l1, bool u1, const char *s2, char **pe2, register UV l2, bool u2)
2404 register const U8 *p1 = (const U8*)s1;
2405 register const U8 *p2 = (const U8*)s2;
2406 register const U8 *f1 = NULL;
2407 register const U8 *f2 = NULL;
2408 register U8 *e1 = NULL;
2409 register U8 *q1 = NULL;
2410 register U8 *e2 = NULL;
2411 register U8 *q2 = NULL;
2412 STRLEN n1 = 0, n2 = 0;
2413 U8 foldbuf1[UTF8_MAXBYTES_CASE+1];
2414 U8 foldbuf2[UTF8_MAXBYTES_CASE+1];
2416 STRLEN foldlen1, foldlen2;
2419 PERL_ARGS_ASSERT_IBCMP_UTF8;
2423 /* assert(e1 || l1); */
2424 if (e1 == 0 || (l1 && l1 < (UV)(e1 - (const U8*)s1)))
2425 f1 = (const U8*)s1 + l1;
2428 /* assert(e2 || l2); */
2429 if (e2 == 0 || (l2 && l2 < (UV)(e2 - (const U8*)s2)))
2430 f2 = (const U8*)s2 + l2;
2432 /* This shouldn't happen. However, putting an assert() there makes some
2434 /* assert((e1 == 0 && f1 == 0) || (e2 == 0 && f2 == 0) || (f1 == 0 && f2 == 0)); */
2435 if ((e1 == 0 && f1 == 0) || (e2 == 0 && f2 == 0) || (f1 == 0 && f2 == 0))
2436 return 1; /* mismatch; possible infinite loop or false positive */
2439 natbuf[1] = 0; /* Need to terminate the buffer. */
2441 while ((e1 == 0 || p1 < e1) &&
2442 (f1 == 0 || p1 < f1) &&
2443 (e2 == 0 || p2 < e2) &&
2444 (f2 == 0 || p2 < f2)) {
2447 to_utf8_fold(p1, foldbuf1, &foldlen1);
2449 uvuni_to_utf8(natbuf, (UV) NATIVE_TO_UNI(((UV)*p1)));
2450 to_utf8_fold(natbuf, foldbuf1, &foldlen1);
2457 to_utf8_fold(p2, foldbuf2, &foldlen2);
2459 uvuni_to_utf8(natbuf, (UV) NATIVE_TO_UNI(((UV)*p2)));
2460 to_utf8_fold(natbuf, foldbuf2, &foldlen2);
2466 if ( UTF8SKIP(q1) != UTF8SKIP(q2) ||
2467 (UTF8SKIP(q1) == 1 && *q1 != *q2) ||
2468 memNE((char*)q1, (char*)q2, UTF8SKIP(q1)) )
2469 return 1; /* mismatch */
2476 p1 += u1 ? UTF8SKIP(p1) : 1;
2478 p2 += u2 ? UTF8SKIP(p2) : 1;
2482 /* A match is defined by all the scans that specified
2483 * an explicit length reaching their final goals. */
2484 match = (f1 == 0 || p1 == f1) && (f2 == 0 || p2 == f2);
2493 return match ? 0 : 1; /* 0 match, 1 mismatch */
2498 * c-indentation-style: bsd
2500 * indent-tabs-mode: t
2503 * ex: set ts=8 sts=4 sw=4 noet: