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 non-character 0x%04"UVxf" is illegal for interchange", 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))
735 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8),
736 "%s in %s", unees, OP_DESC(PL_op));
738 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), unees);
745 =for apidoc utf8_distance
747 Returns the number of UTF-8 characters between the UTF-8 pointers C<a>
750 WARNING: use only if you *know* that the pointers point inside the
757 Perl_utf8_distance(pTHX_ const U8 *a, const U8 *b)
759 PERL_ARGS_ASSERT_UTF8_DISTANCE;
761 return (a < b) ? -1 * (IV) utf8_length(a, b) : (IV) utf8_length(b, a);
767 Return the UTF-8 pointer C<s> displaced by C<off> characters, either
770 WARNING: do not use the following unless you *know* C<off> is within
771 the UTF-8 data pointed to by C<s> *and* that on entry C<s> is aligned
772 on the first byte of character or just after the last byte of a character.
778 Perl_utf8_hop(pTHX_ const U8 *s, I32 off)
780 PERL_ARGS_ASSERT_UTF8_HOP;
783 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g
784 * the bitops (especially ~) can create illegal UTF-8.
785 * In other words: in Perl UTF-8 is not just for Unicode. */
794 while (UTF8_IS_CONTINUATION(*s))
802 =for apidoc utf8_to_bytes
804 Converts a string C<s> of length C<len> from UTF-8 into native byte encoding.
805 Unlike C<bytes_to_utf8>, this over-writes the original string, and
806 updates len to contain the new length.
807 Returns zero on failure, setting C<len> to -1.
809 If you need a copy of the string, see C<bytes_from_utf8>.
815 Perl_utf8_to_bytes(pTHX_ U8 *s, STRLEN *len)
818 U8 * const send = s + *len;
821 PERL_ARGS_ASSERT_UTF8_TO_BYTES;
823 /* ensure valid UTF-8 and chars < 256 before updating string */
827 if (!UTF8_IS_INVARIANT(c) &&
828 (!UTF8_IS_DOWNGRADEABLE_START(c) || (s >= send)
829 || !(c = *s++) || !UTF8_IS_CONTINUATION(c))) {
830 *len = ((STRLEN) -1);
838 *d++ = (U8)utf8_to_uvchr(s, &ulen);
847 =for apidoc bytes_from_utf8
849 Converts a string C<s> of length C<len> from UTF-8 into native byte encoding.
850 Unlike C<utf8_to_bytes> but like C<bytes_to_utf8>, returns a pointer to
851 the newly-created string, and updates C<len> to contain the new
852 length. Returns the original string if no conversion occurs, C<len>
853 is unchanged. Do nothing if C<is_utf8> points to 0. Sets C<is_utf8> to
854 0 if C<s> is converted or consisted entirely of characters that are invariant
855 in utf8 (i.e., US-ASCII on non-EBCDIC machines).
861 Perl_bytes_from_utf8(pTHX_ const U8 *s, STRLEN *len, bool *is_utf8)
868 PERL_ARGS_ASSERT_BYTES_FROM_UTF8;
874 /* ensure valid UTF-8 and chars < 256 before converting string */
875 for (send = s + *len; s < send;) {
877 if (!UTF8_IS_INVARIANT(c)) {
878 if (UTF8_IS_DOWNGRADEABLE_START(c) && s < send &&
879 (c = *s++) && UTF8_IS_CONTINUATION(c))
888 Newx(d, (*len) - count + 1, U8);
889 s = start; start = d;
892 if (!UTF8_IS_INVARIANT(c)) {
893 /* Then it is two-byte encoded */
894 c = UTF8_ACCUMULATE(NATIVE_TO_UTF(c), *s++);
895 c = ASCII_TO_NATIVE(c);
905 =for apidoc bytes_to_utf8
907 Converts a string C<s> of length C<len> from the native encoding into UTF-8.
908 Returns a pointer to the newly-created string, and sets C<len> to
909 reflect the new length.
911 A NUL character will be written after the end of the string.
913 If you want to convert to UTF-8 from encodings other than
914 the native (Latin1 or EBCDIC),
915 see sv_recode_to_utf8().
921 Perl_bytes_to_utf8(pTHX_ const U8 *s, STRLEN *len)
923 const U8 * const send = s + (*len);
927 PERL_ARGS_ASSERT_BYTES_TO_UTF8;
930 Newx(d, (*len) * 2 + 1, U8);
934 const UV uv = NATIVE_TO_ASCII(*s++);
935 if (UNI_IS_INVARIANT(uv))
936 *d++ = (U8)UTF_TO_NATIVE(uv);
938 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
939 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
948 * Convert native (big-endian) or reversed (little-endian) UTF-16 to UTF-8.
950 * Destination must be pre-extended to 3/2 source. Do not use in-place.
951 * We optimize for native, for obvious reasons. */
954 Perl_utf16_to_utf8(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen)
959 PERL_ARGS_ASSERT_UTF16_TO_UTF8;
961 if (bytelen == 1 && p[0] == 0) { /* Be understanding. */
968 Perl_croak(aTHX_ "panic: utf16_to_utf8: odd bytelen %"UVuf, (UV)bytelen);
973 UV uv = (p[0] << 8) + p[1]; /* UTF-16BE */
977 *d++ = UNI_TO_NATIVE(uv);
984 *d++ = (U8)(( uv >> 6) | 0xc0);
985 *d++ = (U8)(( uv & 0x3f) | 0x80);
988 if (uv >= 0xd800 && uv < 0xdbff) { /* surrogates */
990 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
992 UV low = (p[0] << 8) + p[1];
994 if (low < 0xdc00 || low >= 0xdfff)
995 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
996 uv = ((uv - 0xd800) << 10) + (low - 0xdc00) + 0x10000;
1000 *d++ = (U8)(( uv >> 12) | 0xe0);
1001 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
1002 *d++ = (U8)(( uv & 0x3f) | 0x80);
1006 *d++ = (U8)(( uv >> 18) | 0xf0);
1007 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
1008 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
1009 *d++ = (U8)(( uv & 0x3f) | 0x80);
1013 *newlen = d - dstart;
1017 /* Note: this one is slightly destructive of the source. */
1020 Perl_utf16_to_utf8_reversed(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen)
1023 U8* const send = s + bytelen;
1025 PERL_ARGS_ASSERT_UTF16_TO_UTF8_REVERSED;
1028 Perl_croak(aTHX_ "panic: utf16_to_utf8_reversed: odd bytelen %"UVuf,
1032 const U8 tmp = s[0];
1037 return utf16_to_utf8(p, d, bytelen, newlen);
1040 /* for now these are all defined (inefficiently) in terms of the utf8 versions */
1043 Perl_is_uni_alnum(pTHX_ UV c)
1045 U8 tmpbuf[UTF8_MAXBYTES+1];
1046 uvchr_to_utf8(tmpbuf, c);
1047 return is_utf8_alnum(tmpbuf);
1051 Perl_is_uni_idfirst(pTHX_ UV c)
1053 U8 tmpbuf[UTF8_MAXBYTES+1];
1054 uvchr_to_utf8(tmpbuf, c);
1055 return is_utf8_idfirst(tmpbuf);
1059 Perl_is_uni_alpha(pTHX_ UV c)
1061 U8 tmpbuf[UTF8_MAXBYTES+1];
1062 uvchr_to_utf8(tmpbuf, c);
1063 return is_utf8_alpha(tmpbuf);
1067 Perl_is_uni_ascii(pTHX_ UV c)
1069 U8 tmpbuf[UTF8_MAXBYTES+1];
1070 uvchr_to_utf8(tmpbuf, c);
1071 return is_utf8_ascii(tmpbuf);
1075 Perl_is_uni_space(pTHX_ UV c)
1077 U8 tmpbuf[UTF8_MAXBYTES+1];
1078 uvchr_to_utf8(tmpbuf, c);
1079 return is_utf8_space(tmpbuf);
1083 Perl_is_uni_digit(pTHX_ UV c)
1085 U8 tmpbuf[UTF8_MAXBYTES+1];
1086 uvchr_to_utf8(tmpbuf, c);
1087 return is_utf8_digit(tmpbuf);
1091 Perl_is_uni_upper(pTHX_ UV c)
1093 U8 tmpbuf[UTF8_MAXBYTES+1];
1094 uvchr_to_utf8(tmpbuf, c);
1095 return is_utf8_upper(tmpbuf);
1099 Perl_is_uni_lower(pTHX_ UV c)
1101 U8 tmpbuf[UTF8_MAXBYTES+1];
1102 uvchr_to_utf8(tmpbuf, c);
1103 return is_utf8_lower(tmpbuf);
1107 Perl_is_uni_cntrl(pTHX_ UV c)
1109 U8 tmpbuf[UTF8_MAXBYTES+1];
1110 uvchr_to_utf8(tmpbuf, c);
1111 return is_utf8_cntrl(tmpbuf);
1115 Perl_is_uni_graph(pTHX_ UV c)
1117 U8 tmpbuf[UTF8_MAXBYTES+1];
1118 uvchr_to_utf8(tmpbuf, c);
1119 return is_utf8_graph(tmpbuf);
1123 Perl_is_uni_print(pTHX_ UV c)
1125 U8 tmpbuf[UTF8_MAXBYTES+1];
1126 uvchr_to_utf8(tmpbuf, c);
1127 return is_utf8_print(tmpbuf);
1131 Perl_is_uni_punct(pTHX_ UV c)
1133 U8 tmpbuf[UTF8_MAXBYTES+1];
1134 uvchr_to_utf8(tmpbuf, c);
1135 return is_utf8_punct(tmpbuf);
1139 Perl_is_uni_xdigit(pTHX_ UV c)
1141 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1142 uvchr_to_utf8(tmpbuf, c);
1143 return is_utf8_xdigit(tmpbuf);
1147 Perl_to_uni_upper(pTHX_ UV c, U8* p, STRLEN *lenp)
1149 PERL_ARGS_ASSERT_TO_UNI_UPPER;
1151 uvchr_to_utf8(p, c);
1152 return to_utf8_upper(p, p, lenp);
1156 Perl_to_uni_title(pTHX_ UV c, U8* p, STRLEN *lenp)
1158 PERL_ARGS_ASSERT_TO_UNI_TITLE;
1160 uvchr_to_utf8(p, c);
1161 return to_utf8_title(p, p, lenp);
1165 Perl_to_uni_lower(pTHX_ UV c, U8* p, STRLEN *lenp)
1167 PERL_ARGS_ASSERT_TO_UNI_LOWER;
1169 uvchr_to_utf8(p, c);
1170 return to_utf8_lower(p, p, lenp);
1174 Perl_to_uni_fold(pTHX_ UV c, U8* p, STRLEN *lenp)
1176 PERL_ARGS_ASSERT_TO_UNI_FOLD;
1178 uvchr_to_utf8(p, c);
1179 return to_utf8_fold(p, p, lenp);
1182 /* for now these all assume no locale info available for Unicode > 255 */
1185 Perl_is_uni_alnum_lc(pTHX_ UV c)
1187 return is_uni_alnum(c); /* XXX no locale support yet */
1191 Perl_is_uni_idfirst_lc(pTHX_ UV c)
1193 return is_uni_idfirst(c); /* XXX no locale support yet */
1197 Perl_is_uni_alpha_lc(pTHX_ UV c)
1199 return is_uni_alpha(c); /* XXX no locale support yet */
1203 Perl_is_uni_ascii_lc(pTHX_ UV c)
1205 return is_uni_ascii(c); /* XXX no locale support yet */
1209 Perl_is_uni_space_lc(pTHX_ UV c)
1211 return is_uni_space(c); /* XXX no locale support yet */
1215 Perl_is_uni_digit_lc(pTHX_ UV c)
1217 return is_uni_digit(c); /* XXX no locale support yet */
1221 Perl_is_uni_upper_lc(pTHX_ UV c)
1223 return is_uni_upper(c); /* XXX no locale support yet */
1227 Perl_is_uni_lower_lc(pTHX_ UV c)
1229 return is_uni_lower(c); /* XXX no locale support yet */
1233 Perl_is_uni_cntrl_lc(pTHX_ UV c)
1235 return is_uni_cntrl(c); /* XXX no locale support yet */
1239 Perl_is_uni_graph_lc(pTHX_ UV c)
1241 return is_uni_graph(c); /* XXX no locale support yet */
1245 Perl_is_uni_print_lc(pTHX_ UV c)
1247 return is_uni_print(c); /* XXX no locale support yet */
1251 Perl_is_uni_punct_lc(pTHX_ UV c)
1253 return is_uni_punct(c); /* XXX no locale support yet */
1257 Perl_is_uni_xdigit_lc(pTHX_ UV c)
1259 return is_uni_xdigit(c); /* XXX no locale support yet */
1263 Perl_to_uni_upper_lc(pTHX_ U32 c)
1265 /* XXX returns only the first character -- do not use XXX */
1266 /* XXX no locale support yet */
1268 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1269 return (U32)to_uni_upper(c, tmpbuf, &len);
1273 Perl_to_uni_title_lc(pTHX_ U32 c)
1275 /* XXX returns only the first character XXX -- do not use XXX */
1276 /* XXX no locale support yet */
1278 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1279 return (U32)to_uni_title(c, tmpbuf, &len);
1283 Perl_to_uni_lower_lc(pTHX_ U32 c)
1285 /* XXX returns only the first character -- do not use XXX */
1286 /* XXX no locale support yet */
1288 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1289 return (U32)to_uni_lower(c, tmpbuf, &len);
1293 S_is_utf8_common(pTHX_ const U8 *const p, SV **swash,
1294 const char *const swashname)
1298 PERL_ARGS_ASSERT_IS_UTF8_COMMON;
1300 if (!is_utf8_char(p))
1303 *swash = swash_init("utf8", swashname, &PL_sv_undef, 1, 0);
1304 return swash_fetch(*swash, p, TRUE) != 0;
1308 Perl_is_utf8_alnum(pTHX_ const U8 *p)
1312 PERL_ARGS_ASSERT_IS_UTF8_ALNUM;
1314 /* NOTE: "IsWord", not "IsAlnum", since Alnum is a true
1315 * descendant of isalnum(3), in other words, it doesn't
1316 * contain the '_'. --jhi */
1317 return is_utf8_common(p, &PL_utf8_alnum, "IsWord");
1321 Perl_is_utf8_idfirst(pTHX_ const U8 *p) /* The naming is historical. */
1325 PERL_ARGS_ASSERT_IS_UTF8_IDFIRST;
1329 /* is_utf8_idstart would be more logical. */
1330 return is_utf8_common(p, &PL_utf8_idstart, "IdStart");
1334 Perl_is_utf8_idcont(pTHX_ const U8 *p)
1338 PERL_ARGS_ASSERT_IS_UTF8_IDCONT;
1342 return is_utf8_common(p, &PL_utf8_idcont, "IdContinue");
1346 Perl_is_utf8_alpha(pTHX_ const U8 *p)
1350 PERL_ARGS_ASSERT_IS_UTF8_ALPHA;
1352 return is_utf8_common(p, &PL_utf8_alpha, "IsAlpha");
1356 Perl_is_utf8_ascii(pTHX_ const U8 *p)
1360 PERL_ARGS_ASSERT_IS_UTF8_ASCII;
1362 return is_utf8_common(p, &PL_utf8_ascii, "IsAscii");
1366 Perl_is_utf8_space(pTHX_ const U8 *p)
1370 PERL_ARGS_ASSERT_IS_UTF8_SPACE;
1372 return is_utf8_common(p, &PL_utf8_space, "IsSpacePerl");
1376 Perl_is_utf8_digit(pTHX_ const U8 *p)
1380 PERL_ARGS_ASSERT_IS_UTF8_DIGIT;
1382 return is_utf8_common(p, &PL_utf8_digit, "IsDigit");
1386 Perl_is_utf8_upper(pTHX_ const U8 *p)
1390 PERL_ARGS_ASSERT_IS_UTF8_UPPER;
1392 return is_utf8_common(p, &PL_utf8_upper, "IsUppercase");
1396 Perl_is_utf8_lower(pTHX_ const U8 *p)
1400 PERL_ARGS_ASSERT_IS_UTF8_LOWER;
1402 return is_utf8_common(p, &PL_utf8_lower, "IsLowercase");
1406 Perl_is_utf8_cntrl(pTHX_ const U8 *p)
1410 PERL_ARGS_ASSERT_IS_UTF8_CNTRL;
1412 return is_utf8_common(p, &PL_utf8_cntrl, "IsCntrl");
1416 Perl_is_utf8_graph(pTHX_ const U8 *p)
1420 PERL_ARGS_ASSERT_IS_UTF8_GRAPH;
1422 return is_utf8_common(p, &PL_utf8_graph, "IsGraph");
1426 Perl_is_utf8_print(pTHX_ const U8 *p)
1430 PERL_ARGS_ASSERT_IS_UTF8_PRINT;
1432 return is_utf8_common(p, &PL_utf8_print, "IsPrint");
1436 Perl_is_utf8_punct(pTHX_ const U8 *p)
1440 PERL_ARGS_ASSERT_IS_UTF8_PUNCT;
1442 return is_utf8_common(p, &PL_utf8_punct, "IsPunct");
1446 Perl_is_utf8_xdigit(pTHX_ const U8 *p)
1450 PERL_ARGS_ASSERT_IS_UTF8_XDIGIT;
1452 return is_utf8_common(p, &PL_utf8_xdigit, "Isxdigit");
1456 Perl_is_utf8_mark(pTHX_ const U8 *p)
1460 PERL_ARGS_ASSERT_IS_UTF8_MARK;
1462 return is_utf8_common(p, &PL_utf8_mark, "IsM");
1466 =for apidoc to_utf8_case
1468 The "p" contains the pointer to the UTF-8 string encoding
1469 the character that is being converted.
1471 The "ustrp" is a pointer to the character buffer to put the
1472 conversion result to. The "lenp" is a pointer to the length
1475 The "swashp" is a pointer to the swash to use.
1477 Both the special and normal mappings are stored lib/unicore/To/Foo.pl,
1478 and loaded by SWASHNEW, using lib/utf8_heavy.pl. The special (usually,
1479 but not always, a multicharacter mapping), is tried first.
1481 The "special" is a string like "utf8::ToSpecLower", which means the
1482 hash %utf8::ToSpecLower. The access to the hash is through
1483 Perl_to_utf8_case().
1485 The "normal" is a string like "ToLower" which means the swash
1491 Perl_to_utf8_case(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp,
1492 SV **swashp, const char *normal, const char *special)
1495 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1497 const UV uv0 = utf8_to_uvchr(p, NULL);
1498 /* The NATIVE_TO_UNI() and UNI_TO_NATIVE() mappings
1499 * are necessary in EBCDIC, they are redundant no-ops
1500 * in ASCII-ish platforms, and hopefully optimized away. */
1501 const UV uv1 = NATIVE_TO_UNI(uv0);
1503 PERL_ARGS_ASSERT_TO_UTF8_CASE;
1505 uvuni_to_utf8(tmpbuf, uv1);
1507 if (!*swashp) /* load on-demand */
1508 *swashp = swash_init("utf8", normal, &PL_sv_undef, 4, 0);
1510 /* The 0xDF is the only special casing Unicode code point below 0x100. */
1511 if (special && (uv1 == 0xDF || uv1 > 0xFF)) {
1512 /* It might be "special" (sometimes, but not always,
1513 * a multicharacter mapping) */
1514 HV * const hv = get_hv(special, 0);
1518 (svp = hv_fetch(hv, (const char*)tmpbuf, UNISKIP(uv1), FALSE)) &&
1522 s = SvPV_const(*svp, len);
1524 len = uvuni_to_utf8(ustrp, NATIVE_TO_UNI(*(U8*)s)) - ustrp;
1527 /* If we have EBCDIC we need to remap the characters
1528 * since any characters in the low 256 are Unicode
1529 * code points, not EBCDIC. */
1530 U8 *t = (U8*)s, *tend = t + len, *d;
1537 const UV c = utf8_to_uvchr(t, &tlen);
1539 d = uvchr_to_utf8(d, UNI_TO_NATIVE(c));
1548 d = uvchr_to_utf8(d, UNI_TO_NATIVE(*t));
1553 Copy(tmpbuf, ustrp, len, U8);
1555 Copy(s, ustrp, len, U8);
1561 if (!len && *swashp) {
1562 const UV uv2 = swash_fetch(*swashp, tmpbuf, TRUE);
1565 /* It was "normal" (a single character mapping). */
1566 const UV uv3 = UNI_TO_NATIVE(uv2);
1567 len = uvchr_to_utf8(ustrp, uv3) - ustrp;
1571 if (!len) /* Neither: just copy. */
1572 len = uvchr_to_utf8(ustrp, uv0) - ustrp;
1577 return len ? utf8_to_uvchr(ustrp, 0) : 0;
1581 =for apidoc to_utf8_upper
1583 Convert the UTF-8 encoded character at p to its uppercase version and
1584 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
1585 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since
1586 the uppercase version may be longer than the original character.
1588 The first character of the uppercased version is returned
1589 (but note, as explained above, that there may be more.)
1594 Perl_to_utf8_upper(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
1598 PERL_ARGS_ASSERT_TO_UTF8_UPPER;
1600 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
1601 &PL_utf8_toupper, "ToUpper", "utf8::ToSpecUpper");
1605 =for apidoc to_utf8_title
1607 Convert the UTF-8 encoded character at p to its titlecase version and
1608 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
1609 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
1610 titlecase version may be longer than the original character.
1612 The first character of the titlecased version is returned
1613 (but note, as explained above, that there may be more.)
1618 Perl_to_utf8_title(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
1622 PERL_ARGS_ASSERT_TO_UTF8_TITLE;
1624 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
1625 &PL_utf8_totitle, "ToTitle", "utf8::ToSpecTitle");
1629 =for apidoc to_utf8_lower
1631 Convert the UTF-8 encoded character at p to its lowercase version and
1632 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
1633 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
1634 lowercase version may be longer than the original character.
1636 The first character of the lowercased version is returned
1637 (but note, as explained above, that there may be more.)
1642 Perl_to_utf8_lower(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
1646 PERL_ARGS_ASSERT_TO_UTF8_LOWER;
1648 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
1649 &PL_utf8_tolower, "ToLower", "utf8::ToSpecLower");
1653 =for apidoc to_utf8_fold
1655 Convert the UTF-8 encoded character at p to its foldcase version and
1656 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
1657 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
1658 foldcase version may be longer than the original character (up to
1661 The first character of the foldcased version is returned
1662 (but note, as explained above, that there may be more.)
1667 Perl_to_utf8_fold(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
1671 PERL_ARGS_ASSERT_TO_UTF8_FOLD;
1673 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
1674 &PL_utf8_tofold, "ToFold", "utf8::ToSpecFold");
1678 * A "swash" is a swatch hash.
1679 * A "swatch" is a bit vector generated by utf8.c:S_swash_get().
1680 * C<pkg> is a pointer to a package name for SWASHNEW, should be "utf8".
1681 * For other parameters, see utf8::SWASHNEW in lib/utf8_heavy.pl.
1684 Perl_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv, I32 minbits, I32 none)
1689 const size_t pkg_len = strlen(pkg);
1690 const size_t name_len = strlen(name);
1691 HV * const stash = gv_stashpvn(pkg, pkg_len, 0);
1694 PERL_ARGS_ASSERT_SWASH_INIT;
1696 PUSHSTACKi(PERLSI_MAGIC);
1701 if (!gv_fetchmeth(stash, "SWASHNEW", 8, -1)) { /* demand load utf8 */
1703 errsv_save = newSVsv(ERRSV);
1704 /* It is assumed that callers of this routine are not passing in any
1705 user derived data. */
1706 /* Need to do this after save_re_context() as it will set PL_tainted to
1707 1 while saving $1 etc (see the code after getrx: in Perl_magic_get).
1708 Even line to create errsv_save can turn on PL_tainted. */
1709 SAVEBOOL(PL_tainted);
1711 Perl_load_module(aTHX_ PERL_LOADMOD_NOIMPORT, newSVpvn(pkg,pkg_len),
1714 sv_setsv(ERRSV, errsv_save);
1715 SvREFCNT_dec(errsv_save);
1721 mPUSHp(pkg, pkg_len);
1722 mPUSHp(name, name_len);
1727 errsv_save = newSVsv(ERRSV);
1728 if (call_method("SWASHNEW", G_SCALAR))
1729 retval = newSVsv(*PL_stack_sp--);
1731 retval = &PL_sv_undef;
1733 sv_setsv(ERRSV, errsv_save);
1734 SvREFCNT_dec(errsv_save);
1737 if (IN_PERL_COMPILETIME) {
1738 CopHINTS_set(PL_curcop, PL_hints);
1740 if (!SvROK(retval) || SvTYPE(SvRV(retval)) != SVt_PVHV) {
1742 Perl_croak(aTHX_ "Can't find Unicode property definition \"%"SVf"\"",
1744 Perl_croak(aTHX_ "SWASHNEW didn't return an HV ref");
1750 /* This API is wrong for special case conversions since we may need to
1751 * return several Unicode characters for a single Unicode character
1752 * (see lib/unicore/SpecCase.txt) The SWASHGET in lib/utf8_heavy.pl is
1753 * the lower-level routine, and it is similarly broken for returning
1754 * multiple values. --jhi */
1755 /* Now SWASHGET is recasted into S_swash_get in this file. */
1758 * Returns the value of property/mapping C<swash> for the first character
1759 * of the string C<ptr>. If C<do_utf8> is true, the string C<ptr> is
1760 * assumed to be in utf8. If C<do_utf8> is false, the string C<ptr> is
1761 * assumed to be in native 8-bit encoding. Caches the swatch in C<swash>.
1764 Perl_swash_fetch(pTHX_ SV *swash, const U8 *ptr, bool do_utf8)
1767 HV *const hv = MUTABLE_HV(SvRV(swash));
1772 const U8 *tmps = NULL;
1776 const UV c = NATIVE_TO_ASCII(*ptr);
1778 PERL_ARGS_ASSERT_SWASH_FETCH;
1780 if (!do_utf8 && !UNI_IS_INVARIANT(c)) {
1781 tmputf8[0] = (U8)UTF8_EIGHT_BIT_HI(c);
1782 tmputf8[1] = (U8)UTF8_EIGHT_BIT_LO(c);
1785 /* Given a UTF-X encoded char 0xAA..0xYY,0xZZ
1786 * then the "swatch" is a vec() for al the chars which start
1788 * So the key in the hash (klen) is length of encoded char -1
1790 klen = UTF8SKIP(ptr) - 1;
1794 /* If char in invariant then swatch is for all the invariant chars
1795 * In both UTF-8 and UTF-8-MOD that happens to be UTF_CONTINUATION_MARK
1797 needents = UTF_CONTINUATION_MARK;
1798 off = NATIVE_TO_UTF(ptr[klen]);
1801 /* If char is encoded then swatch is for the prefix */
1802 needents = (1 << UTF_ACCUMULATION_SHIFT);
1803 off = NATIVE_TO_UTF(ptr[klen]) & UTF_CONTINUATION_MASK;
1807 * This single-entry cache saves about 1/3 of the utf8 overhead in test
1808 * suite. (That is, only 7-8% overall over just a hash cache. Still,
1809 * it's nothing to sniff at.) Pity we usually come through at least
1810 * two function calls to get here...
1812 * NB: this code assumes that swatches are never modified, once generated!
1815 if (hv == PL_last_swash_hv &&
1816 klen == PL_last_swash_klen &&
1817 (!klen || memEQ((char *)ptr, (char *)PL_last_swash_key, klen)) )
1819 tmps = PL_last_swash_tmps;
1820 slen = PL_last_swash_slen;
1823 /* Try our second-level swatch cache, kept in a hash. */
1824 SV** svp = hv_fetch(hv, (const char*)ptr, klen, FALSE);
1826 /* If not cached, generate it via swash_get */
1827 if (!svp || !SvPOK(*svp)
1828 || !(tmps = (const U8*)SvPV_const(*svp, slen))) {
1829 /* We use utf8n_to_uvuni() as we want an index into
1830 Unicode tables, not a native character number.
1832 const UV code_point = utf8n_to_uvuni(ptr, UTF8_MAXBYTES, 0,
1834 0 : UTF8_ALLOW_ANY);
1835 swatch = swash_get(swash,
1836 /* On EBCDIC & ~(0xA0-1) isn't a useful thing to do */
1837 (klen) ? (code_point & ~(needents - 1)) : 0,
1840 if (IN_PERL_COMPILETIME)
1841 CopHINTS_set(PL_curcop, PL_hints);
1843 svp = hv_store(hv, (const char *)ptr, klen, swatch, 0);
1845 if (!svp || !(tmps = (U8*)SvPV(*svp, slen))
1846 || (slen << 3) < needents)
1847 Perl_croak(aTHX_ "panic: swash_fetch got improper swatch");
1850 PL_last_swash_hv = hv;
1851 assert(klen <= sizeof(PL_last_swash_key));
1852 PL_last_swash_klen = (U8)klen;
1853 /* FIXME change interpvar.h? */
1854 PL_last_swash_tmps = (U8 *) tmps;
1855 PL_last_swash_slen = slen;
1857 Copy(ptr, PL_last_swash_key, klen, U8);
1860 switch ((int)((slen << 3) / needents)) {
1862 bit = 1 << (off & 7);
1864 return (tmps[off] & bit) != 0;
1869 return (tmps[off] << 8) + tmps[off + 1] ;
1872 return (tmps[off] << 24) + (tmps[off+1] << 16) + (tmps[off+2] << 8) + tmps[off + 3] ;
1874 Perl_croak(aTHX_ "panic: swash_fetch got swatch of unexpected bit width");
1875 NORETURN_FUNCTION_END;
1879 * Returns a swatch (a bit vector string) for a code point sequence
1880 * that starts from the value C<start> and comprises the number C<span>.
1881 * A C<swash> must be an object created by SWASHNEW (see lib/utf8_heavy.pl).
1882 * Should be used via swash_fetch, which will cache the swatch in C<swash>.
1885 S_swash_get(pTHX_ SV* swash, UV start, UV span)
1888 U8 *l, *lend, *x, *xend, *s;
1889 STRLEN lcur, xcur, scur;
1890 HV *const hv = MUTABLE_HV(SvRV(swash));
1891 SV** const listsvp = hv_fetchs(hv, "LIST", FALSE);
1892 SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE);
1893 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
1894 SV** const nonesvp = hv_fetchs(hv, "NONE", FALSE);
1895 SV** const extssvp = hv_fetchs(hv, "EXTRAS", FALSE);
1896 const U8* const typestr = (U8*)SvPV_nolen(*typesvp);
1897 const int typeto = typestr[0] == 'T' && typestr[1] == 'o';
1898 const STRLEN bits = SvUV(*bitssvp);
1899 const STRLEN octets = bits >> 3; /* if bits == 1, then octets == 0 */
1900 const UV none = SvUV(*nonesvp);
1901 const UV end = start + span;
1903 PERL_ARGS_ASSERT_SWASH_GET;
1905 if (bits != 1 && bits != 8 && bits != 16 && bits != 32) {
1906 Perl_croak(aTHX_ "panic: swash_get doesn't expect bits %"UVuf,
1910 /* create and initialize $swatch */
1911 scur = octets ? (span * octets) : (span + 7) / 8;
1912 swatch = newSV(scur);
1914 s = (U8*)SvPVX(swatch);
1915 if (octets && none) {
1916 const U8* const e = s + scur;
1919 *s++ = (U8)(none & 0xff);
1920 else if (bits == 16) {
1921 *s++ = (U8)((none >> 8) & 0xff);
1922 *s++ = (U8)( none & 0xff);
1924 else if (bits == 32) {
1925 *s++ = (U8)((none >> 24) & 0xff);
1926 *s++ = (U8)((none >> 16) & 0xff);
1927 *s++ = (U8)((none >> 8) & 0xff);
1928 *s++ = (U8)( none & 0xff);
1934 (void)memzero((U8*)s, scur + 1);
1936 SvCUR_set(swatch, scur);
1937 s = (U8*)SvPVX(swatch);
1939 /* read $swash->{LIST} */
1940 l = (U8*)SvPV(*listsvp, lcur);
1945 I32 flags = PERL_SCAN_SILENT_ILLDIGIT | PERL_SCAN_DISALLOW_PREFIX;
1947 U8* const nl = (U8*)memchr(l, '\n', lend - l);
1950 min = grok_hex((char *)l, &numlen, &flags, NULL);
1954 l = nl + 1; /* 1 is length of "\n" */
1958 l = lend; /* to LIST's end at which \n is not found */
1964 flags = PERL_SCAN_SILENT_ILLDIGIT | PERL_SCAN_DISALLOW_PREFIX;
1966 max = grok_hex((char *)l, &numlen, &flags, NULL);
1975 flags = PERL_SCAN_SILENT_ILLDIGIT |
1976 PERL_SCAN_DISALLOW_PREFIX;
1978 val = grok_hex((char *)l, &numlen, &flags, NULL);
1987 Perl_croak(aTHX_ "%s: illegal mapping '%s'",
1993 val = 0; /* bits == 1, then val should be ignored */
2000 Perl_croak(aTHX_ "%s: illegal mapping '%s'", typestr, l);
2004 val = 0; /* bits == 1, then val should be ignored */
2018 if (!none || val < none) {
2023 for (key = min; key <= max; key++) {
2027 /* offset must be non-negative (start <= min <= key < end) */
2028 offset = octets * (key - start);
2030 s[offset] = (U8)(val & 0xff);
2031 else if (bits == 16) {
2032 s[offset ] = (U8)((val >> 8) & 0xff);
2033 s[offset + 1] = (U8)( val & 0xff);
2035 else if (bits == 32) {
2036 s[offset ] = (U8)((val >> 24) & 0xff);
2037 s[offset + 1] = (U8)((val >> 16) & 0xff);
2038 s[offset + 2] = (U8)((val >> 8) & 0xff);
2039 s[offset + 3] = (U8)( val & 0xff);
2042 if (!none || val < none)
2046 else { /* bits == 1, then val should be ignored */
2050 for (key = min; key <= max; key++) {
2051 const STRLEN offset = (STRLEN)(key - start);
2054 s[offset >> 3] |= 1 << (offset & 7);
2060 /* read $swash->{EXTRAS} */
2061 x = (U8*)SvPV(*extssvp, xcur);
2069 SV **otherbitssvp, *other;
2073 const U8 opc = *x++;
2077 nl = (U8*)memchr(x, '\n', xend - x);
2079 if (opc != '-' && opc != '+' && opc != '!' && opc != '&') {
2081 x = nl + 1; /* 1 is length of "\n" */
2085 x = xend; /* to EXTRAS' end at which \n is not found */
2092 namelen = nl - namestr;
2096 namelen = xend - namestr;
2100 othersvp = hv_fetch(hv, (char *)namestr, namelen, FALSE);
2101 otherhv = MUTABLE_HV(SvRV(*othersvp));
2102 otherbitssvp = hv_fetchs(otherhv, "BITS", FALSE);
2103 otherbits = (STRLEN)SvUV(*otherbitssvp);
2104 if (bits < otherbits)
2105 Perl_croak(aTHX_ "panic: swash_get found swatch size mismatch");
2107 /* The "other" swatch must be destroyed after. */
2108 other = swash_get(*othersvp, start, span);
2109 o = (U8*)SvPV(other, olen);
2112 Perl_croak(aTHX_ "panic: swash_get got improper swatch");
2114 s = (U8*)SvPV(swatch, slen);
2115 if (bits == 1 && otherbits == 1) {
2117 Perl_croak(aTHX_ "panic: swash_get found swatch length mismatch");
2141 STRLEN otheroctets = otherbits >> 3;
2143 U8* const send = s + slen;
2148 if (otherbits == 1) {
2149 otherval = (o[offset >> 3] >> (offset & 7)) & 1;
2153 STRLEN vlen = otheroctets;
2161 if (opc == '+' && otherval)
2162 NOOP; /* replace with otherval */
2163 else if (opc == '!' && !otherval)
2165 else if (opc == '-' && otherval)
2167 else if (opc == '&' && !otherval)
2170 s += octets; /* no replacement */
2175 *s++ = (U8)( otherval & 0xff);
2176 else if (bits == 16) {
2177 *s++ = (U8)((otherval >> 8) & 0xff);
2178 *s++ = (U8)( otherval & 0xff);
2180 else if (bits == 32) {
2181 *s++ = (U8)((otherval >> 24) & 0xff);
2182 *s++ = (U8)((otherval >> 16) & 0xff);
2183 *s++ = (U8)((otherval >> 8) & 0xff);
2184 *s++ = (U8)( otherval & 0xff);
2188 sv_free(other); /* through with it! */
2194 =for apidoc uvchr_to_utf8
2196 Adds the UTF-8 representation of the Native codepoint C<uv> to the end
2197 of the string C<d>; C<d> should be have at least C<UTF8_MAXBYTES+1> free
2198 bytes available. The return value is the pointer to the byte after the
2199 end of the new character. In other words,
2201 d = uvchr_to_utf8(d, uv);
2203 is the recommended wide native character-aware way of saying
2210 /* On ASCII machines this is normally a macro but we want a
2211 real function in case XS code wants it
2214 Perl_uvchr_to_utf8(pTHX_ U8 *d, UV uv)
2216 PERL_ARGS_ASSERT_UVCHR_TO_UTF8;
2218 return Perl_uvuni_to_utf8_flags(aTHX_ d, NATIVE_TO_UNI(uv), 0);
2222 Perl_uvchr_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags)
2224 PERL_ARGS_ASSERT_UVCHR_TO_UTF8_FLAGS;
2226 return Perl_uvuni_to_utf8_flags(aTHX_ d, NATIVE_TO_UNI(uv), flags);
2230 =for apidoc utf8n_to_uvchr
2233 Returns the native character value of the first character in the string
2235 which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
2236 length, in bytes, of that character.
2238 Allows length and flags to be passed to low level routine.
2242 /* On ASCII machines this is normally a macro but we want
2243 a real function in case XS code wants it
2246 Perl_utf8n_to_uvchr(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen,
2249 const UV uv = Perl_utf8n_to_uvuni(aTHX_ s, curlen, retlen, flags);
2251 PERL_ARGS_ASSERT_UTF8N_TO_UVCHR;
2253 return UNI_TO_NATIVE(uv);
2257 =for apidoc pv_uni_display
2259 Build to the scalar dsv a displayable version of the string spv,
2260 length len, the displayable version being at most pvlim bytes long
2261 (if longer, the rest is truncated and "..." will be appended).
2263 The flags argument can have UNI_DISPLAY_ISPRINT set to display
2264 isPRINT()able characters as themselves, UNI_DISPLAY_BACKSLASH
2265 to display the \\[nrfta\\] as the backslashed versions (like '\n')
2266 (UNI_DISPLAY_BACKSLASH is preferred over UNI_DISPLAY_ISPRINT for \\).
2267 UNI_DISPLAY_QQ (and its alias UNI_DISPLAY_REGEX) have both
2268 UNI_DISPLAY_BACKSLASH and UNI_DISPLAY_ISPRINT turned on.
2270 The pointer to the PV of the dsv is returned.
2274 Perl_pv_uni_display(pTHX_ SV *dsv, const U8 *spv, STRLEN len, STRLEN pvlim, UV flags)
2279 PERL_ARGS_ASSERT_PV_UNI_DISPLAY;
2283 for (s = (const char *)spv, e = s + len; s < e; s += UTF8SKIP(s)) {
2285 /* This serves double duty as a flag and a character to print after
2286 a \ when flags & UNI_DISPLAY_BACKSLASH is true.
2290 if (pvlim && SvCUR(dsv) >= pvlim) {
2294 u = utf8_to_uvchr((U8*)s, 0);
2296 const unsigned char c = (unsigned char)u & 0xFF;
2297 if (flags & UNI_DISPLAY_BACKSLASH) {
2314 const char string = ok;
2315 sv_catpvs(dsv, "\\");
2316 sv_catpvn(dsv, &string, 1);
2319 /* isPRINT() is the locale-blind version. */
2320 if (!ok && (flags & UNI_DISPLAY_ISPRINT) && isPRINT(c)) {
2321 const char string = c;
2322 sv_catpvn(dsv, &string, 1);
2327 Perl_sv_catpvf(aTHX_ dsv, "\\x{%"UVxf"}", u);
2330 sv_catpvs(dsv, "...");
2336 =for apidoc sv_uni_display
2338 Build to the scalar dsv a displayable version of the scalar sv,
2339 the displayable version being at most pvlim bytes long
2340 (if longer, the rest is truncated and "..." will be appended).
2342 The flags argument is as in pv_uni_display().
2344 The pointer to the PV of the dsv is returned.
2349 Perl_sv_uni_display(pTHX_ SV *dsv, SV *ssv, STRLEN pvlim, UV flags)
2351 PERL_ARGS_ASSERT_SV_UNI_DISPLAY;
2353 return Perl_pv_uni_display(aTHX_ dsv, (const U8*)SvPVX_const(ssv),
2354 SvCUR(ssv), pvlim, flags);
2358 =for apidoc ibcmp_utf8
2360 Return true if the strings s1 and s2 differ case-insensitively, false
2361 if not (if they are equal case-insensitively). If u1 is true, the
2362 string s1 is assumed to be in UTF-8-encoded Unicode. If u2 is true,
2363 the string s2 is assumed to be in UTF-8-encoded Unicode. If u1 or u2
2364 are false, the respective string is assumed to be in native 8-bit
2367 If the pe1 and pe2 are non-NULL, the scanning pointers will be copied
2368 in there (they will point at the beginning of the I<next> character).
2369 If the pointers behind pe1 or pe2 are non-NULL, they are the end
2370 pointers beyond which scanning will not continue under any
2371 circumstances. If the byte lengths l1 and l2 are non-zero, s1+l1 and
2372 s2+l2 will be used as goal end pointers that will also stop the scan,
2373 and which qualify towards defining a successful match: all the scans
2374 that define an explicit length must reach their goal pointers for
2375 a match to succeed).
2377 For case-insensitiveness, the "casefolding" of Unicode is used
2378 instead of upper/lowercasing both the characters, see
2379 http://www.unicode.org/unicode/reports/tr21/ (Case Mappings).
2383 Perl_ibcmp_utf8(pTHX_ const char *s1, char **pe1, register UV l1, bool u1, const char *s2, char **pe2, register UV l2, bool u2)
2386 register const U8 *p1 = (const U8*)s1;
2387 register const U8 *p2 = (const U8*)s2;
2388 register const U8 *f1 = NULL;
2389 register const U8 *f2 = NULL;
2390 register U8 *e1 = NULL;
2391 register U8 *q1 = NULL;
2392 register U8 *e2 = NULL;
2393 register U8 *q2 = NULL;
2394 STRLEN n1 = 0, n2 = 0;
2395 U8 foldbuf1[UTF8_MAXBYTES_CASE+1];
2396 U8 foldbuf2[UTF8_MAXBYTES_CASE+1];
2398 STRLEN foldlen1, foldlen2;
2401 PERL_ARGS_ASSERT_IBCMP_UTF8;
2405 /* assert(e1 || l1); */
2406 if (e1 == 0 || (l1 && l1 < (UV)(e1 - (const U8*)s1)))
2407 f1 = (const U8*)s1 + l1;
2410 /* assert(e2 || l2); */
2411 if (e2 == 0 || (l2 && l2 < (UV)(e2 - (const U8*)s2)))
2412 f2 = (const U8*)s2 + l2;
2414 /* This shouldn't happen. However, putting an assert() there makes some
2416 /* assert((e1 == 0 && f1 == 0) || (e2 == 0 && f2 == 0) || (f1 == 0 && f2 == 0)); */
2417 if ((e1 == 0 && f1 == 0) || (e2 == 0 && f2 == 0) || (f1 == 0 && f2 == 0))
2418 return 1; /* mismatch; possible infinite loop or false positive */
2421 natbuf[1] = 0; /* Need to terminate the buffer. */
2423 while ((e1 == 0 || p1 < e1) &&
2424 (f1 == 0 || p1 < f1) &&
2425 (e2 == 0 || p2 < e2) &&
2426 (f2 == 0 || p2 < f2)) {
2429 to_utf8_fold(p1, foldbuf1, &foldlen1);
2431 uvuni_to_utf8(natbuf, (UV) NATIVE_TO_UNI(((UV)*p1)));
2432 to_utf8_fold(natbuf, foldbuf1, &foldlen1);
2439 to_utf8_fold(p2, foldbuf2, &foldlen2);
2441 uvuni_to_utf8(natbuf, (UV) NATIVE_TO_UNI(((UV)*p2)));
2442 to_utf8_fold(natbuf, foldbuf2, &foldlen2);
2448 if ( UTF8SKIP(q1) != UTF8SKIP(q2) ||
2449 (UTF8SKIP(q1) == 1 && *q1 != *q2) ||
2450 memNE((char*)q1, (char*)q2, UTF8SKIP(q1)) )
2451 return 1; /* mismatch */
2458 p1 += u1 ? UTF8SKIP(p1) : 1;
2460 p2 += u2 ? UTF8SKIP(p2) : 1;
2464 /* A match is defined by all the scans that specified
2465 * an explicit length reaching their final goals. */
2466 match = (f1 == 0 || p1 == f1) && (f2 == 0 || p2 == f2);
2475 return match ? 0 : 1; /* 0 match, 1 mismatch */
2480 * c-indentation-style: bsd
2482 * indent-tabs-mode: t
2485 * ex: set ts=8 sts=4 sw=4 noet: