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 */
989 UV low = (p[0] << 8) + p[1];
991 if (low < 0xdc00 || low >= 0xdfff)
992 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
993 uv = ((uv - 0xd800) << 10) + (low - 0xdc00) + 0x10000;
996 *d++ = (U8)(( uv >> 12) | 0xe0);
997 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
998 *d++ = (U8)(( uv & 0x3f) | 0x80);
1002 *d++ = (U8)(( uv >> 18) | 0xf0);
1003 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
1004 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
1005 *d++ = (U8)(( uv & 0x3f) | 0x80);
1009 *newlen = d - dstart;
1013 /* Note: this one is slightly destructive of the source. */
1016 Perl_utf16_to_utf8_reversed(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen)
1019 U8* const send = s + bytelen;
1021 PERL_ARGS_ASSERT_UTF16_TO_UTF8_REVERSED;
1024 const U8 tmp = s[0];
1029 return utf16_to_utf8(p, d, bytelen, newlen);
1032 /* for now these are all defined (inefficiently) in terms of the utf8 versions */
1035 Perl_is_uni_alnum(pTHX_ UV c)
1037 U8 tmpbuf[UTF8_MAXBYTES+1];
1038 uvchr_to_utf8(tmpbuf, c);
1039 return is_utf8_alnum(tmpbuf);
1043 Perl_is_uni_idfirst(pTHX_ UV c)
1045 U8 tmpbuf[UTF8_MAXBYTES+1];
1046 uvchr_to_utf8(tmpbuf, c);
1047 return is_utf8_idfirst(tmpbuf);
1051 Perl_is_uni_alpha(pTHX_ UV c)
1053 U8 tmpbuf[UTF8_MAXBYTES+1];
1054 uvchr_to_utf8(tmpbuf, c);
1055 return is_utf8_alpha(tmpbuf);
1059 Perl_is_uni_ascii(pTHX_ UV c)
1061 U8 tmpbuf[UTF8_MAXBYTES+1];
1062 uvchr_to_utf8(tmpbuf, c);
1063 return is_utf8_ascii(tmpbuf);
1067 Perl_is_uni_space(pTHX_ UV c)
1069 U8 tmpbuf[UTF8_MAXBYTES+1];
1070 uvchr_to_utf8(tmpbuf, c);
1071 return is_utf8_space(tmpbuf);
1075 Perl_is_uni_digit(pTHX_ UV c)
1077 U8 tmpbuf[UTF8_MAXBYTES+1];
1078 uvchr_to_utf8(tmpbuf, c);
1079 return is_utf8_digit(tmpbuf);
1083 Perl_is_uni_upper(pTHX_ UV c)
1085 U8 tmpbuf[UTF8_MAXBYTES+1];
1086 uvchr_to_utf8(tmpbuf, c);
1087 return is_utf8_upper(tmpbuf);
1091 Perl_is_uni_lower(pTHX_ UV c)
1093 U8 tmpbuf[UTF8_MAXBYTES+1];
1094 uvchr_to_utf8(tmpbuf, c);
1095 return is_utf8_lower(tmpbuf);
1099 Perl_is_uni_cntrl(pTHX_ UV c)
1101 U8 tmpbuf[UTF8_MAXBYTES+1];
1102 uvchr_to_utf8(tmpbuf, c);
1103 return is_utf8_cntrl(tmpbuf);
1107 Perl_is_uni_graph(pTHX_ UV c)
1109 U8 tmpbuf[UTF8_MAXBYTES+1];
1110 uvchr_to_utf8(tmpbuf, c);
1111 return is_utf8_graph(tmpbuf);
1115 Perl_is_uni_print(pTHX_ UV c)
1117 U8 tmpbuf[UTF8_MAXBYTES+1];
1118 uvchr_to_utf8(tmpbuf, c);
1119 return is_utf8_print(tmpbuf);
1123 Perl_is_uni_punct(pTHX_ UV c)
1125 U8 tmpbuf[UTF8_MAXBYTES+1];
1126 uvchr_to_utf8(tmpbuf, c);
1127 return is_utf8_punct(tmpbuf);
1131 Perl_is_uni_xdigit(pTHX_ UV c)
1133 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1134 uvchr_to_utf8(tmpbuf, c);
1135 return is_utf8_xdigit(tmpbuf);
1139 Perl_to_uni_upper(pTHX_ UV c, U8* p, STRLEN *lenp)
1141 PERL_ARGS_ASSERT_TO_UNI_UPPER;
1143 uvchr_to_utf8(p, c);
1144 return to_utf8_upper(p, p, lenp);
1148 Perl_to_uni_title(pTHX_ UV c, U8* p, STRLEN *lenp)
1150 PERL_ARGS_ASSERT_TO_UNI_TITLE;
1152 uvchr_to_utf8(p, c);
1153 return to_utf8_title(p, p, lenp);
1157 Perl_to_uni_lower(pTHX_ UV c, U8* p, STRLEN *lenp)
1159 PERL_ARGS_ASSERT_TO_UNI_LOWER;
1161 uvchr_to_utf8(p, c);
1162 return to_utf8_lower(p, p, lenp);
1166 Perl_to_uni_fold(pTHX_ UV c, U8* p, STRLEN *lenp)
1168 PERL_ARGS_ASSERT_TO_UNI_FOLD;
1170 uvchr_to_utf8(p, c);
1171 return to_utf8_fold(p, p, lenp);
1174 /* for now these all assume no locale info available for Unicode > 255 */
1177 Perl_is_uni_alnum_lc(pTHX_ UV c)
1179 return is_uni_alnum(c); /* XXX no locale support yet */
1183 Perl_is_uni_idfirst_lc(pTHX_ UV c)
1185 return is_uni_idfirst(c); /* XXX no locale support yet */
1189 Perl_is_uni_alpha_lc(pTHX_ UV c)
1191 return is_uni_alpha(c); /* XXX no locale support yet */
1195 Perl_is_uni_ascii_lc(pTHX_ UV c)
1197 return is_uni_ascii(c); /* XXX no locale support yet */
1201 Perl_is_uni_space_lc(pTHX_ UV c)
1203 return is_uni_space(c); /* XXX no locale support yet */
1207 Perl_is_uni_digit_lc(pTHX_ UV c)
1209 return is_uni_digit(c); /* XXX no locale support yet */
1213 Perl_is_uni_upper_lc(pTHX_ UV c)
1215 return is_uni_upper(c); /* XXX no locale support yet */
1219 Perl_is_uni_lower_lc(pTHX_ UV c)
1221 return is_uni_lower(c); /* XXX no locale support yet */
1225 Perl_is_uni_cntrl_lc(pTHX_ UV c)
1227 return is_uni_cntrl(c); /* XXX no locale support yet */
1231 Perl_is_uni_graph_lc(pTHX_ UV c)
1233 return is_uni_graph(c); /* XXX no locale support yet */
1237 Perl_is_uni_print_lc(pTHX_ UV c)
1239 return is_uni_print(c); /* XXX no locale support yet */
1243 Perl_is_uni_punct_lc(pTHX_ UV c)
1245 return is_uni_punct(c); /* XXX no locale support yet */
1249 Perl_is_uni_xdigit_lc(pTHX_ UV c)
1251 return is_uni_xdigit(c); /* XXX no locale support yet */
1255 Perl_to_uni_upper_lc(pTHX_ U32 c)
1257 /* XXX returns only the first character -- do not use XXX */
1258 /* XXX no locale support yet */
1260 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1261 return (U32)to_uni_upper(c, tmpbuf, &len);
1265 Perl_to_uni_title_lc(pTHX_ U32 c)
1267 /* XXX returns only the first character XXX -- do not use XXX */
1268 /* XXX no locale support yet */
1270 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1271 return (U32)to_uni_title(c, tmpbuf, &len);
1275 Perl_to_uni_lower_lc(pTHX_ U32 c)
1277 /* XXX returns only the first character -- do not use XXX */
1278 /* XXX no locale support yet */
1280 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1281 return (U32)to_uni_lower(c, tmpbuf, &len);
1285 S_is_utf8_common(pTHX_ const U8 *const p, SV **swash,
1286 const char *const swashname)
1290 PERL_ARGS_ASSERT_IS_UTF8_COMMON;
1292 if (!is_utf8_char(p))
1295 *swash = swash_init("utf8", swashname, &PL_sv_undef, 1, 0);
1296 return swash_fetch(*swash, p, TRUE) != 0;
1300 Perl_is_utf8_alnum(pTHX_ const U8 *p)
1304 PERL_ARGS_ASSERT_IS_UTF8_ALNUM;
1306 /* NOTE: "IsWord", not "IsAlnum", since Alnum is a true
1307 * descendant of isalnum(3), in other words, it doesn't
1308 * contain the '_'. --jhi */
1309 return is_utf8_common(p, &PL_utf8_alnum, "IsWord");
1313 Perl_is_utf8_idfirst(pTHX_ const U8 *p) /* The naming is historical. */
1317 PERL_ARGS_ASSERT_IS_UTF8_IDFIRST;
1321 /* is_utf8_idstart would be more logical. */
1322 return is_utf8_common(p, &PL_utf8_idstart, "IdStart");
1326 Perl_is_utf8_idcont(pTHX_ const U8 *p)
1330 PERL_ARGS_ASSERT_IS_UTF8_IDCONT;
1334 return is_utf8_common(p, &PL_utf8_idcont, "IdContinue");
1338 Perl_is_utf8_alpha(pTHX_ const U8 *p)
1342 PERL_ARGS_ASSERT_IS_UTF8_ALPHA;
1344 return is_utf8_common(p, &PL_utf8_alpha, "IsAlpha");
1348 Perl_is_utf8_ascii(pTHX_ const U8 *p)
1352 PERL_ARGS_ASSERT_IS_UTF8_ASCII;
1354 return is_utf8_common(p, &PL_utf8_ascii, "IsAscii");
1358 Perl_is_utf8_space(pTHX_ const U8 *p)
1362 PERL_ARGS_ASSERT_IS_UTF8_SPACE;
1364 return is_utf8_common(p, &PL_utf8_space, "IsSpacePerl");
1368 Perl_is_utf8_digit(pTHX_ const U8 *p)
1372 PERL_ARGS_ASSERT_IS_UTF8_DIGIT;
1374 return is_utf8_common(p, &PL_utf8_digit, "IsDigit");
1378 Perl_is_utf8_upper(pTHX_ const U8 *p)
1382 PERL_ARGS_ASSERT_IS_UTF8_UPPER;
1384 return is_utf8_common(p, &PL_utf8_upper, "IsUppercase");
1388 Perl_is_utf8_lower(pTHX_ const U8 *p)
1392 PERL_ARGS_ASSERT_IS_UTF8_LOWER;
1394 return is_utf8_common(p, &PL_utf8_lower, "IsLowercase");
1398 Perl_is_utf8_cntrl(pTHX_ const U8 *p)
1402 PERL_ARGS_ASSERT_IS_UTF8_CNTRL;
1404 return is_utf8_common(p, &PL_utf8_cntrl, "IsCntrl");
1408 Perl_is_utf8_graph(pTHX_ const U8 *p)
1412 PERL_ARGS_ASSERT_IS_UTF8_GRAPH;
1414 return is_utf8_common(p, &PL_utf8_graph, "IsGraph");
1418 Perl_is_utf8_print(pTHX_ const U8 *p)
1422 PERL_ARGS_ASSERT_IS_UTF8_PRINT;
1424 return is_utf8_common(p, &PL_utf8_print, "IsPrint");
1428 Perl_is_utf8_punct(pTHX_ const U8 *p)
1432 PERL_ARGS_ASSERT_IS_UTF8_PUNCT;
1434 return is_utf8_common(p, &PL_utf8_punct, "IsPunct");
1438 Perl_is_utf8_xdigit(pTHX_ const U8 *p)
1442 PERL_ARGS_ASSERT_IS_UTF8_XDIGIT;
1444 return is_utf8_common(p, &PL_utf8_xdigit, "Isxdigit");
1448 Perl_is_utf8_mark(pTHX_ const U8 *p)
1452 PERL_ARGS_ASSERT_IS_UTF8_MARK;
1454 return is_utf8_common(p, &PL_utf8_mark, "IsM");
1458 =for apidoc to_utf8_case
1460 The "p" contains the pointer to the UTF-8 string encoding
1461 the character that is being converted.
1463 The "ustrp" is a pointer to the character buffer to put the
1464 conversion result to. The "lenp" is a pointer to the length
1467 The "swashp" is a pointer to the swash to use.
1469 Both the special and normal mappings are stored lib/unicore/To/Foo.pl,
1470 and loaded by SWASHNEW, using lib/utf8_heavy.pl. The special (usually,
1471 but not always, a multicharacter mapping), is tried first.
1473 The "special" is a string like "utf8::ToSpecLower", which means the
1474 hash %utf8::ToSpecLower. The access to the hash is through
1475 Perl_to_utf8_case().
1477 The "normal" is a string like "ToLower" which means the swash
1483 Perl_to_utf8_case(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp,
1484 SV **swashp, const char *normal, const char *special)
1487 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1489 const UV uv0 = utf8_to_uvchr(p, NULL);
1490 /* The NATIVE_TO_UNI() and UNI_TO_NATIVE() mappings
1491 * are necessary in EBCDIC, they are redundant no-ops
1492 * in ASCII-ish platforms, and hopefully optimized away. */
1493 const UV uv1 = NATIVE_TO_UNI(uv0);
1495 PERL_ARGS_ASSERT_TO_UTF8_CASE;
1497 uvuni_to_utf8(tmpbuf, uv1);
1499 if (!*swashp) /* load on-demand */
1500 *swashp = swash_init("utf8", normal, &PL_sv_undef, 4, 0);
1502 /* The 0xDF is the only special casing Unicode code point below 0x100. */
1503 if (special && (uv1 == 0xDF || uv1 > 0xFF)) {
1504 /* It might be "special" (sometimes, but not always,
1505 * a multicharacter mapping) */
1506 HV * const hv = get_hv(special, 0);
1510 (svp = hv_fetch(hv, (const char*)tmpbuf, UNISKIP(uv1), FALSE)) &&
1514 s = SvPV_const(*svp, len);
1516 len = uvuni_to_utf8(ustrp, NATIVE_TO_UNI(*(U8*)s)) - ustrp;
1519 /* If we have EBCDIC we need to remap the characters
1520 * since any characters in the low 256 are Unicode
1521 * code points, not EBCDIC. */
1522 U8 *t = (U8*)s, *tend = t + len, *d;
1529 const UV c = utf8_to_uvchr(t, &tlen);
1531 d = uvchr_to_utf8(d, UNI_TO_NATIVE(c));
1540 d = uvchr_to_utf8(d, UNI_TO_NATIVE(*t));
1545 Copy(tmpbuf, ustrp, len, U8);
1547 Copy(s, ustrp, len, U8);
1553 if (!len && *swashp) {
1554 const UV uv2 = swash_fetch(*swashp, tmpbuf, TRUE);
1557 /* It was "normal" (a single character mapping). */
1558 const UV uv3 = UNI_TO_NATIVE(uv2);
1559 len = uvchr_to_utf8(ustrp, uv3) - ustrp;
1563 if (!len) /* Neither: just copy. */
1564 len = uvchr_to_utf8(ustrp, uv0) - ustrp;
1569 return len ? utf8_to_uvchr(ustrp, 0) : 0;
1573 =for apidoc to_utf8_upper
1575 Convert the UTF-8 encoded character at p to its uppercase version and
1576 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
1577 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since
1578 the uppercase version may be longer than the original character.
1580 The first character of the uppercased version is returned
1581 (but note, as explained above, that there may be more.)
1586 Perl_to_utf8_upper(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
1590 PERL_ARGS_ASSERT_TO_UTF8_UPPER;
1592 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
1593 &PL_utf8_toupper, "ToUpper", "utf8::ToSpecUpper");
1597 =for apidoc to_utf8_title
1599 Convert the UTF-8 encoded character at p to its titlecase version and
1600 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
1601 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
1602 titlecase version may be longer than the original character.
1604 The first character of the titlecased version is returned
1605 (but note, as explained above, that there may be more.)
1610 Perl_to_utf8_title(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
1614 PERL_ARGS_ASSERT_TO_UTF8_TITLE;
1616 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
1617 &PL_utf8_totitle, "ToTitle", "utf8::ToSpecTitle");
1621 =for apidoc to_utf8_lower
1623 Convert the UTF-8 encoded character at p to its lowercase version and
1624 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
1625 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
1626 lowercase version may be longer than the original character.
1628 The first character of the lowercased version is returned
1629 (but note, as explained above, that there may be more.)
1634 Perl_to_utf8_lower(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
1638 PERL_ARGS_ASSERT_TO_UTF8_LOWER;
1640 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
1641 &PL_utf8_tolower, "ToLower", "utf8::ToSpecLower");
1645 =for apidoc to_utf8_fold
1647 Convert the UTF-8 encoded character at p to its foldcase version and
1648 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
1649 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
1650 foldcase version may be longer than the original character (up to
1653 The first character of the foldcased version is returned
1654 (but note, as explained above, that there may be more.)
1659 Perl_to_utf8_fold(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
1663 PERL_ARGS_ASSERT_TO_UTF8_FOLD;
1665 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
1666 &PL_utf8_tofold, "ToFold", "utf8::ToSpecFold");
1670 * A "swash" is a swatch hash.
1671 * A "swatch" is a bit vector generated by utf8.c:S_swash_get().
1672 * C<pkg> is a pointer to a package name for SWASHNEW, should be "utf8".
1673 * For other parameters, see utf8::SWASHNEW in lib/utf8_heavy.pl.
1676 Perl_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv, I32 minbits, I32 none)
1681 const size_t pkg_len = strlen(pkg);
1682 const size_t name_len = strlen(name);
1683 HV * const stash = gv_stashpvn(pkg, pkg_len, 0);
1686 PERL_ARGS_ASSERT_SWASH_INIT;
1688 PUSHSTACKi(PERLSI_MAGIC);
1693 if (!gv_fetchmeth(stash, "SWASHNEW", 8, -1)) { /* demand load utf8 */
1695 errsv_save = newSVsv(ERRSV);
1696 /* It is assumed that callers of this routine are not passing in any
1697 user derived data. */
1698 /* Need to do this after save_re_context() as it will set PL_tainted to
1699 1 while saving $1 etc (see the code after getrx: in Perl_magic_get).
1700 Even line to create errsv_save can turn on PL_tainted. */
1701 SAVEBOOL(PL_tainted);
1703 Perl_load_module(aTHX_ PERL_LOADMOD_NOIMPORT, newSVpvn(pkg,pkg_len),
1706 sv_setsv(ERRSV, errsv_save);
1707 SvREFCNT_dec(errsv_save);
1713 mPUSHp(pkg, pkg_len);
1714 mPUSHp(name, name_len);
1719 errsv_save = newSVsv(ERRSV);
1720 if (call_method("SWASHNEW", G_SCALAR))
1721 retval = newSVsv(*PL_stack_sp--);
1723 retval = &PL_sv_undef;
1725 sv_setsv(ERRSV, errsv_save);
1726 SvREFCNT_dec(errsv_save);
1729 if (IN_PERL_COMPILETIME) {
1730 CopHINTS_set(PL_curcop, PL_hints);
1732 if (!SvROK(retval) || SvTYPE(SvRV(retval)) != SVt_PVHV) {
1734 Perl_croak(aTHX_ "Can't find Unicode property definition \"%"SVf"\"",
1736 Perl_croak(aTHX_ "SWASHNEW didn't return an HV ref");
1742 /* This API is wrong for special case conversions since we may need to
1743 * return several Unicode characters for a single Unicode character
1744 * (see lib/unicore/SpecCase.txt) The SWASHGET in lib/utf8_heavy.pl is
1745 * the lower-level routine, and it is similarly broken for returning
1746 * multiple values. --jhi */
1747 /* Now SWASHGET is recasted into S_swash_get in this file. */
1750 * Returns the value of property/mapping C<swash> for the first character
1751 * of the string C<ptr>. If C<do_utf8> is true, the string C<ptr> is
1752 * assumed to be in utf8. If C<do_utf8> is false, the string C<ptr> is
1753 * assumed to be in native 8-bit encoding. Caches the swatch in C<swash>.
1756 Perl_swash_fetch(pTHX_ SV *swash, const U8 *ptr, bool do_utf8)
1759 HV *const hv = MUTABLE_HV(SvRV(swash));
1764 const U8 *tmps = NULL;
1768 const UV c = NATIVE_TO_ASCII(*ptr);
1770 PERL_ARGS_ASSERT_SWASH_FETCH;
1772 if (!do_utf8 && !UNI_IS_INVARIANT(c)) {
1773 tmputf8[0] = (U8)UTF8_EIGHT_BIT_HI(c);
1774 tmputf8[1] = (U8)UTF8_EIGHT_BIT_LO(c);
1777 /* Given a UTF-X encoded char 0xAA..0xYY,0xZZ
1778 * then the "swatch" is a vec() for al the chars which start
1780 * So the key in the hash (klen) is length of encoded char -1
1782 klen = UTF8SKIP(ptr) - 1;
1786 /* If char in invariant then swatch is for all the invariant chars
1787 * In both UTF-8 and UTF-8-MOD that happens to be UTF_CONTINUATION_MARK
1789 needents = UTF_CONTINUATION_MARK;
1790 off = NATIVE_TO_UTF(ptr[klen]);
1793 /* If char is encoded then swatch is for the prefix */
1794 needents = (1 << UTF_ACCUMULATION_SHIFT);
1795 off = NATIVE_TO_UTF(ptr[klen]) & UTF_CONTINUATION_MASK;
1799 * This single-entry cache saves about 1/3 of the utf8 overhead in test
1800 * suite. (That is, only 7-8% overall over just a hash cache. Still,
1801 * it's nothing to sniff at.) Pity we usually come through at least
1802 * two function calls to get here...
1804 * NB: this code assumes that swatches are never modified, once generated!
1807 if (hv == PL_last_swash_hv &&
1808 klen == PL_last_swash_klen &&
1809 (!klen || memEQ((char *)ptr, (char *)PL_last_swash_key, klen)) )
1811 tmps = PL_last_swash_tmps;
1812 slen = PL_last_swash_slen;
1815 /* Try our second-level swatch cache, kept in a hash. */
1816 SV** svp = hv_fetch(hv, (const char*)ptr, klen, FALSE);
1818 /* If not cached, generate it via swash_get */
1819 if (!svp || !SvPOK(*svp)
1820 || !(tmps = (const U8*)SvPV_const(*svp, slen))) {
1821 /* We use utf8n_to_uvuni() as we want an index into
1822 Unicode tables, not a native character number.
1824 const UV code_point = utf8n_to_uvuni(ptr, UTF8_MAXBYTES, 0,
1826 0 : UTF8_ALLOW_ANY);
1827 swatch = swash_get(swash,
1828 /* On EBCDIC & ~(0xA0-1) isn't a useful thing to do */
1829 (klen) ? (code_point & ~(needents - 1)) : 0,
1832 if (IN_PERL_COMPILETIME)
1833 CopHINTS_set(PL_curcop, PL_hints);
1835 svp = hv_store(hv, (const char *)ptr, klen, swatch, 0);
1837 if (!svp || !(tmps = (U8*)SvPV(*svp, slen))
1838 || (slen << 3) < needents)
1839 Perl_croak(aTHX_ "panic: swash_fetch got improper swatch");
1842 PL_last_swash_hv = hv;
1843 assert(klen <= sizeof(PL_last_swash_key));
1844 PL_last_swash_klen = (U8)klen;
1845 /* FIXME change interpvar.h? */
1846 PL_last_swash_tmps = (U8 *) tmps;
1847 PL_last_swash_slen = slen;
1849 Copy(ptr, PL_last_swash_key, klen, U8);
1852 switch ((int)((slen << 3) / needents)) {
1854 bit = 1 << (off & 7);
1856 return (tmps[off] & bit) != 0;
1861 return (tmps[off] << 8) + tmps[off + 1] ;
1864 return (tmps[off] << 24) + (tmps[off+1] << 16) + (tmps[off+2] << 8) + tmps[off + 3] ;
1866 Perl_croak(aTHX_ "panic: swash_fetch got swatch of unexpected bit width");
1867 NORETURN_FUNCTION_END;
1871 * Returns a swatch (a bit vector string) for a code point sequence
1872 * that starts from the value C<start> and comprises the number C<span>.
1873 * A C<swash> must be an object created by SWASHNEW (see lib/utf8_heavy.pl).
1874 * Should be used via swash_fetch, which will cache the swatch in C<swash>.
1877 S_swash_get(pTHX_ SV* swash, UV start, UV span)
1880 U8 *l, *lend, *x, *xend, *s;
1881 STRLEN lcur, xcur, scur;
1882 HV *const hv = MUTABLE_HV(SvRV(swash));
1883 SV** const listsvp = hv_fetchs(hv, "LIST", FALSE);
1884 SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE);
1885 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
1886 SV** const nonesvp = hv_fetchs(hv, "NONE", FALSE);
1887 SV** const extssvp = hv_fetchs(hv, "EXTRAS", FALSE);
1888 const U8* const typestr = (U8*)SvPV_nolen(*typesvp);
1889 const int typeto = typestr[0] == 'T' && typestr[1] == 'o';
1890 const STRLEN bits = SvUV(*bitssvp);
1891 const STRLEN octets = bits >> 3; /* if bits == 1, then octets == 0 */
1892 const UV none = SvUV(*nonesvp);
1893 const UV end = start + span;
1895 PERL_ARGS_ASSERT_SWASH_GET;
1897 if (bits != 1 && bits != 8 && bits != 16 && bits != 32) {
1898 Perl_croak(aTHX_ "panic: swash_get doesn't expect bits %"UVuf,
1902 /* create and initialize $swatch */
1903 scur = octets ? (span * octets) : (span + 7) / 8;
1904 swatch = newSV(scur);
1906 s = (U8*)SvPVX(swatch);
1907 if (octets && none) {
1908 const U8* const e = s + scur;
1911 *s++ = (U8)(none & 0xff);
1912 else if (bits == 16) {
1913 *s++ = (U8)((none >> 8) & 0xff);
1914 *s++ = (U8)( none & 0xff);
1916 else if (bits == 32) {
1917 *s++ = (U8)((none >> 24) & 0xff);
1918 *s++ = (U8)((none >> 16) & 0xff);
1919 *s++ = (U8)((none >> 8) & 0xff);
1920 *s++ = (U8)( none & 0xff);
1926 (void)memzero((U8*)s, scur + 1);
1928 SvCUR_set(swatch, scur);
1929 s = (U8*)SvPVX(swatch);
1931 /* read $swash->{LIST} */
1932 l = (U8*)SvPV(*listsvp, lcur);
1937 I32 flags = PERL_SCAN_SILENT_ILLDIGIT | PERL_SCAN_DISALLOW_PREFIX;
1939 U8* const nl = (U8*)memchr(l, '\n', lend - l);
1942 min = grok_hex((char *)l, &numlen, &flags, NULL);
1946 l = nl + 1; /* 1 is length of "\n" */
1950 l = lend; /* to LIST's end at which \n is not found */
1956 flags = PERL_SCAN_SILENT_ILLDIGIT | PERL_SCAN_DISALLOW_PREFIX;
1958 max = grok_hex((char *)l, &numlen, &flags, NULL);
1967 flags = PERL_SCAN_SILENT_ILLDIGIT |
1968 PERL_SCAN_DISALLOW_PREFIX;
1970 val = grok_hex((char *)l, &numlen, &flags, NULL);
1979 Perl_croak(aTHX_ "%s: illegal mapping '%s'",
1985 val = 0; /* bits == 1, then val should be ignored */
1992 Perl_croak(aTHX_ "%s: illegal mapping '%s'", typestr, l);
1996 val = 0; /* bits == 1, then val should be ignored */
2010 if (!none || val < none) {
2015 for (key = min; key <= max; key++) {
2019 /* offset must be non-negative (start <= min <= key < end) */
2020 offset = octets * (key - start);
2022 s[offset] = (U8)(val & 0xff);
2023 else if (bits == 16) {
2024 s[offset ] = (U8)((val >> 8) & 0xff);
2025 s[offset + 1] = (U8)( val & 0xff);
2027 else if (bits == 32) {
2028 s[offset ] = (U8)((val >> 24) & 0xff);
2029 s[offset + 1] = (U8)((val >> 16) & 0xff);
2030 s[offset + 2] = (U8)((val >> 8) & 0xff);
2031 s[offset + 3] = (U8)( val & 0xff);
2034 if (!none || val < none)
2038 else { /* bits == 1, then val should be ignored */
2042 for (key = min; key <= max; key++) {
2043 const STRLEN offset = (STRLEN)(key - start);
2046 s[offset >> 3] |= 1 << (offset & 7);
2052 /* read $swash->{EXTRAS} */
2053 x = (U8*)SvPV(*extssvp, xcur);
2061 SV **otherbitssvp, *other;
2065 const U8 opc = *x++;
2069 nl = (U8*)memchr(x, '\n', xend - x);
2071 if (opc != '-' && opc != '+' && opc != '!' && opc != '&') {
2073 x = nl + 1; /* 1 is length of "\n" */
2077 x = xend; /* to EXTRAS' end at which \n is not found */
2084 namelen = nl - namestr;
2088 namelen = xend - namestr;
2092 othersvp = hv_fetch(hv, (char *)namestr, namelen, FALSE);
2093 otherhv = MUTABLE_HV(SvRV(*othersvp));
2094 otherbitssvp = hv_fetchs(otherhv, "BITS", FALSE);
2095 otherbits = (STRLEN)SvUV(*otherbitssvp);
2096 if (bits < otherbits)
2097 Perl_croak(aTHX_ "panic: swash_get found swatch size mismatch");
2099 /* The "other" swatch must be destroyed after. */
2100 other = swash_get(*othersvp, start, span);
2101 o = (U8*)SvPV(other, olen);
2104 Perl_croak(aTHX_ "panic: swash_get got improper swatch");
2106 s = (U8*)SvPV(swatch, slen);
2107 if (bits == 1 && otherbits == 1) {
2109 Perl_croak(aTHX_ "panic: swash_get found swatch length mismatch");
2133 STRLEN otheroctets = otherbits >> 3;
2135 U8* const send = s + slen;
2140 if (otherbits == 1) {
2141 otherval = (o[offset >> 3] >> (offset & 7)) & 1;
2145 STRLEN vlen = otheroctets;
2153 if (opc == '+' && otherval)
2154 NOOP; /* replace with otherval */
2155 else if (opc == '!' && !otherval)
2157 else if (opc == '-' && otherval)
2159 else if (opc == '&' && !otherval)
2162 s += octets; /* no replacement */
2167 *s++ = (U8)( otherval & 0xff);
2168 else if (bits == 16) {
2169 *s++ = (U8)((otherval >> 8) & 0xff);
2170 *s++ = (U8)( otherval & 0xff);
2172 else if (bits == 32) {
2173 *s++ = (U8)((otherval >> 24) & 0xff);
2174 *s++ = (U8)((otherval >> 16) & 0xff);
2175 *s++ = (U8)((otherval >> 8) & 0xff);
2176 *s++ = (U8)( otherval & 0xff);
2180 sv_free(other); /* through with it! */
2186 =for apidoc uvchr_to_utf8
2188 Adds the UTF-8 representation of the Native codepoint C<uv> to the end
2189 of the string C<d>; C<d> should be have at least C<UTF8_MAXBYTES+1> free
2190 bytes available. The return value is the pointer to the byte after the
2191 end of the new character. In other words,
2193 d = uvchr_to_utf8(d, uv);
2195 is the recommended wide native character-aware way of saying
2202 /* On ASCII machines this is normally a macro but we want a
2203 real function in case XS code wants it
2206 Perl_uvchr_to_utf8(pTHX_ U8 *d, UV uv)
2208 PERL_ARGS_ASSERT_UVCHR_TO_UTF8;
2210 return Perl_uvuni_to_utf8_flags(aTHX_ d, NATIVE_TO_UNI(uv), 0);
2214 Perl_uvchr_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags)
2216 PERL_ARGS_ASSERT_UVCHR_TO_UTF8_FLAGS;
2218 return Perl_uvuni_to_utf8_flags(aTHX_ d, NATIVE_TO_UNI(uv), flags);
2222 =for apidoc utf8n_to_uvchr
2225 Returns the native character value of the first character in the string
2227 which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
2228 length, in bytes, of that character.
2230 Allows length and flags to be passed to low level routine.
2234 /* On ASCII machines this is normally a macro but we want
2235 a real function in case XS code wants it
2238 Perl_utf8n_to_uvchr(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen,
2241 const UV uv = Perl_utf8n_to_uvuni(aTHX_ s, curlen, retlen, flags);
2243 PERL_ARGS_ASSERT_UTF8N_TO_UVCHR;
2245 return UNI_TO_NATIVE(uv);
2249 =for apidoc pv_uni_display
2251 Build to the scalar dsv a displayable version of the string spv,
2252 length len, the displayable version being at most pvlim bytes long
2253 (if longer, the rest is truncated and "..." will be appended).
2255 The flags argument can have UNI_DISPLAY_ISPRINT set to display
2256 isPRINT()able characters as themselves, UNI_DISPLAY_BACKSLASH
2257 to display the \\[nrfta\\] as the backslashed versions (like '\n')
2258 (UNI_DISPLAY_BACKSLASH is preferred over UNI_DISPLAY_ISPRINT for \\).
2259 UNI_DISPLAY_QQ (and its alias UNI_DISPLAY_REGEX) have both
2260 UNI_DISPLAY_BACKSLASH and UNI_DISPLAY_ISPRINT turned on.
2262 The pointer to the PV of the dsv is returned.
2266 Perl_pv_uni_display(pTHX_ SV *dsv, const U8 *spv, STRLEN len, STRLEN pvlim, UV flags)
2271 PERL_ARGS_ASSERT_PV_UNI_DISPLAY;
2275 for (s = (const char *)spv, e = s + len; s < e; s += UTF8SKIP(s)) {
2277 /* This serves double duty as a flag and a character to print after
2278 a \ when flags & UNI_DISPLAY_BACKSLASH is true.
2282 if (pvlim && SvCUR(dsv) >= pvlim) {
2286 u = utf8_to_uvchr((U8*)s, 0);
2288 const unsigned char c = (unsigned char)u & 0xFF;
2289 if (flags & UNI_DISPLAY_BACKSLASH) {
2306 const char string = ok;
2307 sv_catpvs(dsv, "\\");
2308 sv_catpvn(dsv, &string, 1);
2311 /* isPRINT() is the locale-blind version. */
2312 if (!ok && (flags & UNI_DISPLAY_ISPRINT) && isPRINT(c)) {
2313 const char string = c;
2314 sv_catpvn(dsv, &string, 1);
2319 Perl_sv_catpvf(aTHX_ dsv, "\\x{%"UVxf"}", u);
2322 sv_catpvs(dsv, "...");
2328 =for apidoc sv_uni_display
2330 Build to the scalar dsv a displayable version of the scalar sv,
2331 the displayable version being at most pvlim bytes long
2332 (if longer, the rest is truncated and "..." will be appended).
2334 The flags argument is as in pv_uni_display().
2336 The pointer to the PV of the dsv is returned.
2341 Perl_sv_uni_display(pTHX_ SV *dsv, SV *ssv, STRLEN pvlim, UV flags)
2343 PERL_ARGS_ASSERT_SV_UNI_DISPLAY;
2345 return Perl_pv_uni_display(aTHX_ dsv, (const U8*)SvPVX_const(ssv),
2346 SvCUR(ssv), pvlim, flags);
2350 =for apidoc ibcmp_utf8
2352 Return true if the strings s1 and s2 differ case-insensitively, false
2353 if not (if they are equal case-insensitively). If u1 is true, the
2354 string s1 is assumed to be in UTF-8-encoded Unicode. If u2 is true,
2355 the string s2 is assumed to be in UTF-8-encoded Unicode. If u1 or u2
2356 are false, the respective string is assumed to be in native 8-bit
2359 If the pe1 and pe2 are non-NULL, the scanning pointers will be copied
2360 in there (they will point at the beginning of the I<next> character).
2361 If the pointers behind pe1 or pe2 are non-NULL, they are the end
2362 pointers beyond which scanning will not continue under any
2363 circumstances. If the byte lengths l1 and l2 are non-zero, s1+l1 and
2364 s2+l2 will be used as goal end pointers that will also stop the scan,
2365 and which qualify towards defining a successful match: all the scans
2366 that define an explicit length must reach their goal pointers for
2367 a match to succeed).
2369 For case-insensitiveness, the "casefolding" of Unicode is used
2370 instead of upper/lowercasing both the characters, see
2371 http://www.unicode.org/unicode/reports/tr21/ (Case Mappings).
2375 Perl_ibcmp_utf8(pTHX_ const char *s1, char **pe1, register UV l1, bool u1, const char *s2, char **pe2, register UV l2, bool u2)
2378 register const U8 *p1 = (const U8*)s1;
2379 register const U8 *p2 = (const U8*)s2;
2380 register const U8 *f1 = NULL;
2381 register const U8 *f2 = NULL;
2382 register U8 *e1 = NULL;
2383 register U8 *q1 = NULL;
2384 register U8 *e2 = NULL;
2385 register U8 *q2 = NULL;
2386 STRLEN n1 = 0, n2 = 0;
2387 U8 foldbuf1[UTF8_MAXBYTES_CASE+1];
2388 U8 foldbuf2[UTF8_MAXBYTES_CASE+1];
2390 STRLEN foldlen1, foldlen2;
2393 PERL_ARGS_ASSERT_IBCMP_UTF8;
2397 /* assert(e1 || l1); */
2398 if (e1 == 0 || (l1 && l1 < (UV)(e1 - (const U8*)s1)))
2399 f1 = (const U8*)s1 + l1;
2402 /* assert(e2 || l2); */
2403 if (e2 == 0 || (l2 && l2 < (UV)(e2 - (const U8*)s2)))
2404 f2 = (const U8*)s2 + l2;
2406 /* This shouldn't happen. However, putting an assert() there makes some
2408 /* assert((e1 == 0 && f1 == 0) || (e2 == 0 && f2 == 0) || (f1 == 0 && f2 == 0)); */
2409 if ((e1 == 0 && f1 == 0) || (e2 == 0 && f2 == 0) || (f1 == 0 && f2 == 0))
2410 return 1; /* mismatch; possible infinite loop or false positive */
2413 natbuf[1] = 0; /* Need to terminate the buffer. */
2415 while ((e1 == 0 || p1 < e1) &&
2416 (f1 == 0 || p1 < f1) &&
2417 (e2 == 0 || p2 < e2) &&
2418 (f2 == 0 || p2 < f2)) {
2421 to_utf8_fold(p1, foldbuf1, &foldlen1);
2423 uvuni_to_utf8(natbuf, (UV) NATIVE_TO_UNI(((UV)*p1)));
2424 to_utf8_fold(natbuf, foldbuf1, &foldlen1);
2431 to_utf8_fold(p2, foldbuf2, &foldlen2);
2433 uvuni_to_utf8(natbuf, (UV) NATIVE_TO_UNI(((UV)*p2)));
2434 to_utf8_fold(natbuf, foldbuf2, &foldlen2);
2440 if ( UTF8SKIP(q1) != UTF8SKIP(q2) ||
2441 (UTF8SKIP(q1) == 1 && *q1 != *q2) ||
2442 memNE((char*)q1, (char*)q2, UTF8SKIP(q1)) )
2443 return 1; /* mismatch */
2450 p1 += u1 ? UTF8SKIP(p1) : 1;
2452 p2 += u2 ? UTF8SKIP(p2) : 1;
2456 /* A match is defined by all the scans that specified
2457 * an explicit length reaching their final goals. */
2458 match = (f1 == 0 || p1 == f1) && (f2 == 0 || p2 == f2);
2467 return match ? 0 : 1; /* 0 match, 1 mismatch */
2472 * c-indentation-style: bsd
2474 * indent-tabs-mode: t
2477 * ex: set ts=8 sts=4 sw=4 noet: