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.
54 =for apidoc uvuni_to_utf8_flags
56 Adds the UTF-8 representation of the Unicode codepoint C<uv> to the end
57 of the string C<d>; C<d> should be have at least C<UTF8_MAXBYTES+1> free
58 bytes available. The return value is the pointer to the byte after the
59 end of the new character. In other words,
61 d = uvuni_to_utf8_flags(d, uv, flags);
65 d = uvuni_to_utf8(d, uv);
67 (which is equivalent to)
69 d = uvuni_to_utf8_flags(d, uv, 0);
71 is the recommended Unicode-aware way of saying
79 Perl_uvuni_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags)
81 PERL_ARGS_ASSERT_UVUNI_TO_UTF8_FLAGS;
83 if (ckWARN(WARN_UTF8)) {
84 if (UNICODE_IS_SURROGATE(uv) &&
85 !(flags & UNICODE_ALLOW_SURROGATE))
86 Perl_warner(aTHX_ packWARN(WARN_UTF8), "UTF-16 surrogate 0x%04"UVxf, uv);
88 ((uv >= 0xFDD0 && uv <= 0xFDEF &&
89 !(flags & UNICODE_ALLOW_FDD0))
91 ((uv & 0xFFFE) == 0xFFFE && /* Either FFFE or FFFF. */
92 !(flags & UNICODE_ALLOW_FFFF))) &&
93 /* UNICODE_ALLOW_SUPER includes
94 * FFFEs and FFFFs beyond 0x10FFFF. */
95 ((uv <= PERL_UNICODE_MAX) ||
96 !(flags & UNICODE_ALLOW_SUPER))
98 Perl_warner(aTHX_ packWARN(WARN_UTF8),
99 "Unicode character 0x%04"UVxf" is illegal", uv);
101 if (UNI_IS_INVARIANT(uv)) {
102 *d++ = (U8)UTF_TO_NATIVE(uv);
107 STRLEN len = UNISKIP(uv);
110 *p-- = (U8)UTF_TO_NATIVE((uv & UTF_CONTINUATION_MASK) | UTF_CONTINUATION_MARK);
111 uv >>= UTF_ACCUMULATION_SHIFT;
113 *p = (U8)UTF_TO_NATIVE((uv & UTF_START_MASK(len)) | UTF_START_MARK(len));
116 #else /* Non loop style */
118 *d++ = (U8)(( uv >> 6) | 0xc0);
119 *d++ = (U8)(( uv & 0x3f) | 0x80);
123 *d++ = (U8)(( uv >> 12) | 0xe0);
124 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
125 *d++ = (U8)(( uv & 0x3f) | 0x80);
129 *d++ = (U8)(( uv >> 18) | 0xf0);
130 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
131 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
132 *d++ = (U8)(( uv & 0x3f) | 0x80);
135 if (uv < 0x4000000) {
136 *d++ = (U8)(( uv >> 24) | 0xf8);
137 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
138 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
139 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
140 *d++ = (U8)(( uv & 0x3f) | 0x80);
143 if (uv < 0x80000000) {
144 *d++ = (U8)(( uv >> 30) | 0xfc);
145 *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80);
146 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
147 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
148 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
149 *d++ = (U8)(( uv & 0x3f) | 0x80);
153 if (uv < UTF8_QUAD_MAX)
156 *d++ = 0xfe; /* Can't match U+FEFF! */
157 *d++ = (U8)(((uv >> 30) & 0x3f) | 0x80);
158 *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80);
159 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
160 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
161 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
162 *d++ = (U8)(( uv & 0x3f) | 0x80);
167 *d++ = 0xff; /* Can't match U+FFFE! */
168 *d++ = 0x80; /* 6 Reserved bits */
169 *d++ = (U8)(((uv >> 60) & 0x0f) | 0x80); /* 2 Reserved bits */
170 *d++ = (U8)(((uv >> 54) & 0x3f) | 0x80);
171 *d++ = (U8)(((uv >> 48) & 0x3f) | 0x80);
172 *d++ = (U8)(((uv >> 42) & 0x3f) | 0x80);
173 *d++ = (U8)(((uv >> 36) & 0x3f) | 0x80);
174 *d++ = (U8)(((uv >> 30) & 0x3f) | 0x80);
175 *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80);
176 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
177 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
178 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
179 *d++ = (U8)(( uv & 0x3f) | 0x80);
183 #endif /* Loop style */
188 Tests if some arbitrary number of bytes begins in a valid UTF-8
189 character. Note that an INVARIANT (i.e. ASCII) character is a valid
190 UTF-8 character. The actual number of bytes in the UTF-8 character
191 will be returned if it is valid, otherwise 0.
193 This is the "slow" version as opposed to the "fast" version which is
194 the "unrolled" IS_UTF8_CHAR(). E.g. for t/uni/class.t the speed
195 difference is a factor of 2 to 3. For lengths (UTF8SKIP(s)) of four
196 or less you should use the IS_UTF8_CHAR(), for lengths of five or more
197 you should use the _slow(). In practice this means that the _slow()
198 will be used very rarely, since the maximum Unicode code point (as of
199 Unicode 4.1) is U+10FFFF, which encodes in UTF-8 to four bytes. Only
200 the "Perl extended UTF-8" (the infamous 'v-strings') will encode into
205 S_is_utf8_char_slow(const U8 *s, const STRLEN len)
211 PERL_ARGS_ASSERT_IS_UTF8_CHAR_SLOW;
213 if (UTF8_IS_INVARIANT(u))
216 if (!UTF8_IS_START(u))
219 if (len < 2 || !UTF8_IS_CONTINUATION(s[1]))
225 u = NATIVE_TO_UTF(u);
227 u &= UTF_START_MASK(len);
231 if (!UTF8_IS_CONTINUATION(*s))
233 uv = UTF8_ACCUMULATE(uv, *s);
240 if ((STRLEN)UNISKIP(uv) < len)
247 =for apidoc is_utf8_char
249 Tests if some arbitrary number of bytes begins in a valid UTF-8
250 character. Note that an INVARIANT (i.e. ASCII on non-EBCDIC machines)
251 character is a valid UTF-8 character. The actual number of bytes in the UTF-8
252 character will be returned if it is valid, otherwise 0.
256 Perl_is_utf8_char(pTHX_ const U8 *s)
258 const STRLEN len = UTF8SKIP(s);
260 PERL_ARGS_ASSERT_IS_UTF8_CHAR;
263 if (IS_UTF8_CHAR_FAST(len))
264 return IS_UTF8_CHAR(s, len) ? len : 0;
265 #endif /* #ifdef IS_UTF8_CHAR */
266 return is_utf8_char_slow(s, len);
270 =for apidoc is_utf8_string
272 Returns true if first C<len> bytes of the given string form a valid
273 UTF-8 string, false otherwise. Note that 'a valid UTF-8 string' does
274 not mean 'a string that contains code points above 0x7F encoded in UTF-8'
275 because a valid ASCII string is a valid UTF-8 string.
277 See also is_utf8_string_loclen() and is_utf8_string_loc().
283 Perl_is_utf8_string(pTHX_ const U8 *s, STRLEN len)
285 const U8* const send = s + (len ? len : strlen((const char *)s));
288 PERL_ARGS_ASSERT_IS_UTF8_STRING;
293 /* Inline the easy bits of is_utf8_char() here for speed... */
294 if (UTF8_IS_INVARIANT(*x))
296 else if (!UTF8_IS_START(*x))
299 /* ... and call is_utf8_char() only if really needed. */
302 if (IS_UTF8_CHAR_FAST(c)) {
303 if (!IS_UTF8_CHAR(x, c))
307 c = is_utf8_char_slow(x, c);
310 #endif /* #ifdef IS_UTF8_CHAR */
325 Implemented as a macro in utf8.h
327 =for apidoc is_utf8_string_loc
329 Like is_utf8_string() but stores the location of the failure (in the
330 case of "utf8ness failure") or the location s+len (in the case of
331 "utf8ness success") in the C<ep>.
333 See also is_utf8_string_loclen() and is_utf8_string().
335 =for apidoc is_utf8_string_loclen
337 Like is_utf8_string() but stores the location of the failure (in the
338 case of "utf8ness failure") or the location s+len (in the case of
339 "utf8ness success") in the C<ep>, and the number of UTF-8
340 encoded characters in the C<el>.
342 See also is_utf8_string_loc() and is_utf8_string().
348 Perl_is_utf8_string_loclen(pTHX_ const U8 *s, STRLEN len, const U8 **ep, STRLEN *el)
350 const U8* const send = s + (len ? len : strlen((const char *)s));
355 PERL_ARGS_ASSERT_IS_UTF8_STRING_LOCLEN;
359 /* Inline the easy bits of is_utf8_char() here for speed... */
360 if (UTF8_IS_INVARIANT(*x))
362 else if (!UTF8_IS_START(*x))
365 /* ... and call is_utf8_char() only if really needed. */
368 if (IS_UTF8_CHAR_FAST(c)) {
369 if (!IS_UTF8_CHAR(x, c))
372 c = is_utf8_char_slow(x, c);
375 #endif /* #ifdef IS_UTF8_CHAR */
394 =for apidoc utf8n_to_uvuni
396 Bottom level UTF-8 decode routine.
397 Returns the Unicode code point value of the first character in the string C<s>
398 which is assumed to be in UTF-8 encoding and no longer than C<curlen>;
399 C<retlen> will be set to the length, in bytes, of that character.
401 If C<s> does not point to a well-formed UTF-8 character, the behaviour
402 is dependent on the value of C<flags>: if it contains UTF8_CHECK_ONLY,
403 it is assumed that the caller will raise a warning, and this function
404 will silently just set C<retlen> to C<-1> and return zero. If the
405 C<flags> does not contain UTF8_CHECK_ONLY, warnings about
406 malformations will be given, C<retlen> will be set to the expected
407 length of the UTF-8 character in bytes, and zero will be returned.
409 The C<flags> can also contain various flags to allow deviations from
410 the strict UTF-8 encoding (see F<utf8.h>).
412 Most code should use utf8_to_uvchr() rather than call this directly.
418 Perl_utf8n_to_uvuni(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen, U32 flags)
421 const U8 * const s0 = s;
424 const bool dowarn = ckWARN_d(WARN_UTF8);
425 const UV startbyte = *s;
426 STRLEN expectlen = 0;
429 PERL_ARGS_ASSERT_UTF8N_TO_UVUNI;
431 /* This list is a superset of the UTF8_ALLOW_XXX. */
433 #define UTF8_WARN_EMPTY 1
434 #define UTF8_WARN_CONTINUATION 2
435 #define UTF8_WARN_NON_CONTINUATION 3
436 #define UTF8_WARN_FE_FF 4
437 #define UTF8_WARN_SHORT 5
438 #define UTF8_WARN_OVERFLOW 6
439 #define UTF8_WARN_SURROGATE 7
440 #define UTF8_WARN_LONG 8
441 #define UTF8_WARN_FFFF 9 /* Also FFFE. */
444 !(flags & UTF8_ALLOW_EMPTY)) {
445 warning = UTF8_WARN_EMPTY;
449 if (UTF8_IS_INVARIANT(uv)) {
452 return (UV) (NATIVE_TO_UTF(*s));
455 if (UTF8_IS_CONTINUATION(uv) &&
456 !(flags & UTF8_ALLOW_CONTINUATION)) {
457 warning = UTF8_WARN_CONTINUATION;
461 if (UTF8_IS_START(uv) && curlen > 1 && !UTF8_IS_CONTINUATION(s[1]) &&
462 !(flags & UTF8_ALLOW_NON_CONTINUATION)) {
463 warning = UTF8_WARN_NON_CONTINUATION;
468 uv = NATIVE_TO_UTF(uv);
470 if ((uv == 0xfe || uv == 0xff) &&
471 !(flags & UTF8_ALLOW_FE_FF)) {
472 warning = UTF8_WARN_FE_FF;
477 if (!(uv & 0x20)) { len = 2; uv &= 0x1f; }
478 else if (!(uv & 0x10)) { len = 3; uv &= 0x0f; }
479 else if (!(uv & 0x08)) { len = 4; uv &= 0x07; }
480 else if (!(uv & 0x04)) { len = 5; uv &= 0x03; }
482 else if (!(uv & 0x02)) { len = 6; uv &= 0x01; }
483 else { len = 7; uv &= 0x01; }
485 else if (!(uv & 0x02)) { len = 6; uv &= 0x01; }
486 else if (!(uv & 0x01)) { len = 7; uv = 0; }
487 else { len = 13; uv = 0; } /* whoa! */
495 if ((curlen < expectlen) &&
496 !(flags & UTF8_ALLOW_SHORT)) {
497 warning = UTF8_WARN_SHORT;
506 if (!UTF8_IS_CONTINUATION(*s) &&
507 !(flags & UTF8_ALLOW_NON_CONTINUATION)) {
509 warning = UTF8_WARN_NON_CONTINUATION;
513 uv = UTF8_ACCUMULATE(uv, *s);
515 /* These cannot be allowed. */
517 if (expectlen != 13 && !(flags & UTF8_ALLOW_LONG)) {
518 warning = UTF8_WARN_LONG;
522 else { /* uv < ouv */
523 /* This cannot be allowed. */
524 warning = UTF8_WARN_OVERFLOW;
532 if (UNICODE_IS_SURROGATE(uv) &&
533 !(flags & UTF8_ALLOW_SURROGATE)) {
534 warning = UTF8_WARN_SURROGATE;
536 } else if ((expectlen > (STRLEN)UNISKIP(uv)) &&
537 !(flags & UTF8_ALLOW_LONG)) {
538 warning = UTF8_WARN_LONG;
540 } else if (UNICODE_IS_ILLEGAL(uv) &&
541 !(flags & UTF8_ALLOW_FFFF)) {
542 warning = UTF8_WARN_FFFF;
550 if (flags & UTF8_CHECK_ONLY) {
552 *retlen = ((STRLEN) -1);
557 SV* const sv = newSVpvs_flags("Malformed UTF-8 character ", SVs_TEMP);
560 case 0: /* Intentionally empty. */ break;
561 case UTF8_WARN_EMPTY:
562 sv_catpvs(sv, "(empty string)");
564 case UTF8_WARN_CONTINUATION:
565 Perl_sv_catpvf(aTHX_ sv, "(unexpected continuation byte 0x%02"UVxf", with no preceding start byte)", uv);
567 case UTF8_WARN_NON_CONTINUATION:
569 Perl_sv_catpvf(aTHX_ sv, "(unexpected non-continuation byte 0x%02"UVxf", immediately after start byte 0x%02"UVxf")",
570 (UV)s[1], startbyte);
572 const int len = (int)(s-s0);
573 Perl_sv_catpvf(aTHX_ sv, "(unexpected non-continuation byte 0x%02"UVxf", %d byte%s after start byte 0x%02"UVxf", expected %d bytes)",
574 (UV)s[1], len, len > 1 ? "s" : "", startbyte, (int)expectlen);
578 case UTF8_WARN_FE_FF:
579 Perl_sv_catpvf(aTHX_ sv, "(byte 0x%02"UVxf")", uv);
581 case UTF8_WARN_SHORT:
582 Perl_sv_catpvf(aTHX_ sv, "(%d byte%s, need %d, after start byte 0x%02"UVxf")",
583 (int)curlen, curlen == 1 ? "" : "s", (int)expectlen, startbyte);
584 expectlen = curlen; /* distance for caller to skip */
586 case UTF8_WARN_OVERFLOW:
587 Perl_sv_catpvf(aTHX_ sv, "(overflow at 0x%"UVxf", byte 0x%02x, after start byte 0x%02"UVxf")",
590 case UTF8_WARN_SURROGATE:
591 Perl_sv_catpvf(aTHX_ sv, "(UTF-16 surrogate 0x%04"UVxf")", uv);
594 Perl_sv_catpvf(aTHX_ sv, "(%d byte%s, need %d, after start byte 0x%02"UVxf")",
595 (int)expectlen, expectlen == 1 ? "": "s", UNISKIP(uv), startbyte);
598 Perl_sv_catpvf(aTHX_ sv, "(character 0x%04"UVxf")", uv);
601 sv_catpvs(sv, "(unknown reason)");
606 const char * const s = SvPVX_const(sv);
609 Perl_warner(aTHX_ packWARN(WARN_UTF8),
610 "%s in %s", s, OP_DESC(PL_op));
612 Perl_warner(aTHX_ packWARN(WARN_UTF8), "%s", s);
617 *retlen = expectlen ? expectlen : len;
623 =for apidoc utf8_to_uvchr
625 Returns the native character value of the first character in the string C<s>
626 which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
627 length, in bytes, of that character.
629 If C<s> does not point to a well-formed UTF-8 character, zero is
630 returned and retlen is set, if possible, to -1.
636 Perl_utf8_to_uvchr(pTHX_ const U8 *s, STRLEN *retlen)
638 PERL_ARGS_ASSERT_UTF8_TO_UVCHR;
640 return utf8n_to_uvchr(s, UTF8_MAXBYTES, retlen,
641 ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
645 =for apidoc utf8_to_uvuni
647 Returns the Unicode code point of the first character in the string C<s>
648 which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
649 length, in bytes, of that character.
651 This function should only be used when the returned UV is considered
652 an index into the Unicode semantic tables (e.g. swashes).
654 If C<s> does not point to a well-formed UTF-8 character, zero is
655 returned and retlen is set, if possible, to -1.
661 Perl_utf8_to_uvuni(pTHX_ const U8 *s, STRLEN *retlen)
663 PERL_ARGS_ASSERT_UTF8_TO_UVUNI;
665 /* Call the low level routine asking for checks */
666 return Perl_utf8n_to_uvuni(aTHX_ s, UTF8_MAXBYTES, retlen,
667 ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
671 =for apidoc utf8_length
673 Return the length of the UTF-8 char encoded string C<s> in characters.
674 Stops at C<e> (inclusive). If C<e E<lt> s> or if the scan would end
675 up past C<e>, croaks.
681 Perl_utf8_length(pTHX_ const U8 *s, const U8 *e)
687 PERL_ARGS_ASSERT_UTF8_LENGTH;
689 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g.
690 * the bitops (especially ~) can create illegal UTF-8.
691 * In other words: in Perl UTF-8 is not just for Unicode. */
694 goto warn_and_return;
699 if (ckWARN_d(WARN_UTF8)) {
701 Perl_warner(aTHX_ packWARN(WARN_UTF8),
702 "%s in %s", unees, OP_DESC(PL_op));
704 Perl_warner(aTHX_ packWARN(WARN_UTF8), unees);
716 =for apidoc utf8_distance
718 Returns the number of UTF-8 characters between the UTF-8 pointers C<a>
721 WARNING: use only if you *know* that the pointers point inside the
728 Perl_utf8_distance(pTHX_ const U8 *a, const U8 *b)
730 PERL_ARGS_ASSERT_UTF8_DISTANCE;
732 return (a < b) ? -1 * (IV) utf8_length(a, b) : (IV) utf8_length(b, a);
738 Return the UTF-8 pointer C<s> displaced by C<off> characters, either
741 WARNING: do not use the following unless you *know* C<off> is within
742 the UTF-8 data pointed to by C<s> *and* that on entry C<s> is aligned
743 on the first byte of character or just after the last byte of a character.
749 Perl_utf8_hop(pTHX_ const U8 *s, I32 off)
751 PERL_ARGS_ASSERT_UTF8_HOP;
754 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g
755 * the bitops (especially ~) can create illegal UTF-8.
756 * In other words: in Perl UTF-8 is not just for Unicode. */
765 while (UTF8_IS_CONTINUATION(*s))
773 =for apidoc utf8_to_bytes
775 Converts a string C<s> of length C<len> from UTF-8 into native byte encoding.
776 Unlike C<bytes_to_utf8>, this over-writes the original string, and
777 updates len to contain the new length.
778 Returns zero on failure, setting C<len> to -1.
780 If you need a copy of the string, see C<bytes_from_utf8>.
786 Perl_utf8_to_bytes(pTHX_ U8 *s, STRLEN *len)
789 U8 * const send = s + *len;
792 PERL_ARGS_ASSERT_UTF8_TO_BYTES;
794 /* ensure valid UTF-8 and chars < 256 before updating string */
798 if (!UTF8_IS_INVARIANT(c) &&
799 (!UTF8_IS_DOWNGRADEABLE_START(c) || (s >= send)
800 || !(c = *s++) || !UTF8_IS_CONTINUATION(c))) {
801 *len = ((STRLEN) -1);
809 *d++ = (U8)utf8_to_uvchr(s, &ulen);
818 =for apidoc bytes_from_utf8
820 Converts a string C<s> of length C<len> from UTF-8 into native byte encoding.
821 Unlike C<utf8_to_bytes> but like C<bytes_to_utf8>, returns a pointer to
822 the newly-created string, and updates C<len> to contain the new
823 length. Returns the original string if no conversion occurs, C<len>
824 is unchanged. Do nothing if C<is_utf8> points to 0. Sets C<is_utf8> to
825 0 if C<s> is converted or consisted entirely of characters that are invariant
826 in utf8 (i.e., US-ASCII on non-EBCDIC machines).
832 Perl_bytes_from_utf8(pTHX_ const U8 *s, STRLEN *len, bool *is_utf8)
839 PERL_ARGS_ASSERT_BYTES_FROM_UTF8;
845 /* ensure valid UTF-8 and chars < 256 before converting string */
846 for (send = s + *len; s < send;) {
848 if (!UTF8_IS_INVARIANT(c)) {
849 if (UTF8_IS_DOWNGRADEABLE_START(c) && s < send &&
850 (c = *s++) && UTF8_IS_CONTINUATION(c))
859 Newx(d, (*len) - count + 1, U8);
860 s = start; start = d;
863 if (!UTF8_IS_INVARIANT(c)) {
864 /* Then it is two-byte encoded */
865 c = UTF8_ACCUMULATE(NATIVE_TO_UTF(c), *s++);
866 c = ASCII_TO_NATIVE(c);
876 =for apidoc bytes_to_utf8
878 Converts a string C<s> of length C<len> from the native encoding into UTF-8.
879 Returns a pointer to the newly-created string, and sets C<len> to
880 reflect the new length.
882 A NUL character will be written after the end of the string.
884 If you want to convert to UTF-8 from encodings other than
885 the native (Latin1 or EBCDIC),
886 see sv_recode_to_utf8().
892 Perl_bytes_to_utf8(pTHX_ const U8 *s, STRLEN *len)
894 const U8 * const send = s + (*len);
898 PERL_ARGS_ASSERT_BYTES_TO_UTF8;
901 Newx(d, (*len) * 2 + 1, U8);
905 const UV uv = NATIVE_TO_ASCII(*s++);
906 if (UNI_IS_INVARIANT(uv))
907 *d++ = (U8)UTF_TO_NATIVE(uv);
909 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
910 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
919 * Convert native (big-endian) or reversed (little-endian) UTF-16 to UTF-8.
921 * Destination must be pre-extended to 3/2 source. Do not use in-place.
922 * We optimize for native, for obvious reasons. */
925 Perl_utf16_to_utf8(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen)
930 PERL_ARGS_ASSERT_UTF16_TO_UTF8;
932 if (bytelen == 1 && p[0] == 0) { /* Be understanding. */
939 Perl_croak(aTHX_ "panic: utf16_to_utf8: odd bytelen %"UVuf, (UV)bytelen);
944 UV uv = (p[0] << 8) + p[1]; /* UTF-16BE */
948 *d++ = UNI_TO_NATIVE(uv);
955 *d++ = (U8)(( uv >> 6) | 0xc0);
956 *d++ = (U8)(( uv & 0x3f) | 0x80);
959 if (uv >= 0xd800 && uv < 0xdbff) { /* surrogates */
960 UV low = (p[0] << 8) + p[1];
962 if (low < 0xdc00 || low >= 0xdfff)
963 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
964 uv = ((uv - 0xd800) << 10) + (low - 0xdc00) + 0x10000;
967 *d++ = (U8)(( uv >> 12) | 0xe0);
968 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
969 *d++ = (U8)(( uv & 0x3f) | 0x80);
973 *d++ = (U8)(( uv >> 18) | 0xf0);
974 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
975 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
976 *d++ = (U8)(( uv & 0x3f) | 0x80);
980 *newlen = d - dstart;
984 /* Note: this one is slightly destructive of the source. */
987 Perl_utf16_to_utf8_reversed(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen)
990 U8* const send = s + bytelen;
992 PERL_ARGS_ASSERT_UTF16_TO_UTF8_REVERSED;
1000 return utf16_to_utf8(p, d, bytelen, newlen);
1003 /* for now these are all defined (inefficiently) in terms of the utf8 versions */
1006 Perl_is_uni_alnum(pTHX_ UV c)
1008 U8 tmpbuf[UTF8_MAXBYTES+1];
1009 uvchr_to_utf8(tmpbuf, c);
1010 return is_utf8_alnum(tmpbuf);
1014 Perl_is_uni_alnumc(pTHX_ UV c)
1016 U8 tmpbuf[UTF8_MAXBYTES+1];
1017 uvchr_to_utf8(tmpbuf, c);
1018 return is_utf8_alnumc(tmpbuf);
1022 Perl_is_uni_idfirst(pTHX_ UV c)
1024 U8 tmpbuf[UTF8_MAXBYTES+1];
1025 uvchr_to_utf8(tmpbuf, c);
1026 return is_utf8_idfirst(tmpbuf);
1030 Perl_is_uni_alpha(pTHX_ UV c)
1032 U8 tmpbuf[UTF8_MAXBYTES+1];
1033 uvchr_to_utf8(tmpbuf, c);
1034 return is_utf8_alpha(tmpbuf);
1038 Perl_is_uni_ascii(pTHX_ UV c)
1040 U8 tmpbuf[UTF8_MAXBYTES+1];
1041 uvchr_to_utf8(tmpbuf, c);
1042 return is_utf8_ascii(tmpbuf);
1046 Perl_is_uni_space(pTHX_ UV c)
1048 U8 tmpbuf[UTF8_MAXBYTES+1];
1049 uvchr_to_utf8(tmpbuf, c);
1050 return is_utf8_space(tmpbuf);
1054 Perl_is_uni_digit(pTHX_ UV c)
1056 U8 tmpbuf[UTF8_MAXBYTES+1];
1057 uvchr_to_utf8(tmpbuf, c);
1058 return is_utf8_digit(tmpbuf);
1062 Perl_is_uni_upper(pTHX_ UV c)
1064 U8 tmpbuf[UTF8_MAXBYTES+1];
1065 uvchr_to_utf8(tmpbuf, c);
1066 return is_utf8_upper(tmpbuf);
1070 Perl_is_uni_lower(pTHX_ UV c)
1072 U8 tmpbuf[UTF8_MAXBYTES+1];
1073 uvchr_to_utf8(tmpbuf, c);
1074 return is_utf8_lower(tmpbuf);
1078 Perl_is_uni_cntrl(pTHX_ UV c)
1080 U8 tmpbuf[UTF8_MAXBYTES+1];
1081 uvchr_to_utf8(tmpbuf, c);
1082 return is_utf8_cntrl(tmpbuf);
1086 Perl_is_uni_graph(pTHX_ UV c)
1088 U8 tmpbuf[UTF8_MAXBYTES+1];
1089 uvchr_to_utf8(tmpbuf, c);
1090 return is_utf8_graph(tmpbuf);
1094 Perl_is_uni_print(pTHX_ UV c)
1096 U8 tmpbuf[UTF8_MAXBYTES+1];
1097 uvchr_to_utf8(tmpbuf, c);
1098 return is_utf8_print(tmpbuf);
1102 Perl_is_uni_punct(pTHX_ UV c)
1104 U8 tmpbuf[UTF8_MAXBYTES+1];
1105 uvchr_to_utf8(tmpbuf, c);
1106 return is_utf8_punct(tmpbuf);
1110 Perl_is_uni_xdigit(pTHX_ UV c)
1112 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1113 uvchr_to_utf8(tmpbuf, c);
1114 return is_utf8_xdigit(tmpbuf);
1118 Perl_to_uni_upper(pTHX_ UV c, U8* p, STRLEN *lenp)
1120 PERL_ARGS_ASSERT_TO_UNI_UPPER;
1122 uvchr_to_utf8(p, c);
1123 return to_utf8_upper(p, p, lenp);
1127 Perl_to_uni_title(pTHX_ UV c, U8* p, STRLEN *lenp)
1129 PERL_ARGS_ASSERT_TO_UNI_TITLE;
1131 uvchr_to_utf8(p, c);
1132 return to_utf8_title(p, p, lenp);
1136 Perl_to_uni_lower(pTHX_ UV c, U8* p, STRLEN *lenp)
1138 PERL_ARGS_ASSERT_TO_UNI_LOWER;
1140 uvchr_to_utf8(p, c);
1141 return to_utf8_lower(p, p, lenp);
1145 Perl_to_uni_fold(pTHX_ UV c, U8* p, STRLEN *lenp)
1147 PERL_ARGS_ASSERT_TO_UNI_FOLD;
1149 uvchr_to_utf8(p, c);
1150 return to_utf8_fold(p, p, lenp);
1153 /* for now these all assume no locale info available for Unicode > 255 */
1156 Perl_is_uni_alnum_lc(pTHX_ UV c)
1158 return is_uni_alnum(c); /* XXX no locale support yet */
1162 Perl_is_uni_alnumc_lc(pTHX_ UV c)
1164 return is_uni_alnumc(c); /* XXX no locale support yet */
1168 Perl_is_uni_idfirst_lc(pTHX_ UV c)
1170 return is_uni_idfirst(c); /* XXX no locale support yet */
1174 Perl_is_uni_alpha_lc(pTHX_ UV c)
1176 return is_uni_alpha(c); /* XXX no locale support yet */
1180 Perl_is_uni_ascii_lc(pTHX_ UV c)
1182 return is_uni_ascii(c); /* XXX no locale support yet */
1186 Perl_is_uni_space_lc(pTHX_ UV c)
1188 return is_uni_space(c); /* XXX no locale support yet */
1192 Perl_is_uni_digit_lc(pTHX_ UV c)
1194 return is_uni_digit(c); /* XXX no locale support yet */
1198 Perl_is_uni_upper_lc(pTHX_ UV c)
1200 return is_uni_upper(c); /* XXX no locale support yet */
1204 Perl_is_uni_lower_lc(pTHX_ UV c)
1206 return is_uni_lower(c); /* XXX no locale support yet */
1210 Perl_is_uni_cntrl_lc(pTHX_ UV c)
1212 return is_uni_cntrl(c); /* XXX no locale support yet */
1216 Perl_is_uni_graph_lc(pTHX_ UV c)
1218 return is_uni_graph(c); /* XXX no locale support yet */
1222 Perl_is_uni_print_lc(pTHX_ UV c)
1224 return is_uni_print(c); /* XXX no locale support yet */
1228 Perl_is_uni_punct_lc(pTHX_ UV c)
1230 return is_uni_punct(c); /* XXX no locale support yet */
1234 Perl_is_uni_xdigit_lc(pTHX_ UV c)
1236 return is_uni_xdigit(c); /* XXX no locale support yet */
1240 Perl_to_uni_upper_lc(pTHX_ U32 c)
1242 /* XXX returns only the first character -- do not use XXX */
1243 /* XXX no locale support yet */
1245 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1246 return (U32)to_uni_upper(c, tmpbuf, &len);
1250 Perl_to_uni_title_lc(pTHX_ U32 c)
1252 /* XXX returns only the first character XXX -- do not use XXX */
1253 /* XXX no locale support yet */
1255 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1256 return (U32)to_uni_title(c, tmpbuf, &len);
1260 Perl_to_uni_lower_lc(pTHX_ U32 c)
1262 /* XXX returns only the first character -- do not use XXX */
1263 /* XXX no locale support yet */
1265 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1266 return (U32)to_uni_lower(c, tmpbuf, &len);
1270 S_is_utf8_common(pTHX_ const U8 *const p, SV **swash,
1271 const char *const swashname)
1275 PERL_ARGS_ASSERT_IS_UTF8_COMMON;
1277 if (!is_utf8_char(p))
1280 *swash = swash_init("utf8", swashname, &PL_sv_undef, 1, 0);
1281 return swash_fetch(*swash, p, TRUE) != 0;
1285 Perl_is_utf8_alnum(pTHX_ const U8 *p)
1289 PERL_ARGS_ASSERT_IS_UTF8_ALNUM;
1291 /* NOTE: "IsWord", not "IsAlnum", since Alnum is a true
1292 * descendant of isalnum(3), in other words, it doesn't
1293 * contain the '_'. --jhi */
1294 return is_utf8_common(p, &PL_utf8_alnum, "IsWord");
1298 Perl_is_utf8_alnumc(pTHX_ const U8 *p)
1302 PERL_ARGS_ASSERT_IS_UTF8_ALNUMC;
1304 return is_utf8_common(p, &PL_utf8_alnumc, "IsAlnumC");
1308 Perl_is_utf8_idfirst(pTHX_ const U8 *p) /* The naming is historical. */
1312 PERL_ARGS_ASSERT_IS_UTF8_IDFIRST;
1316 /* is_utf8_idstart would be more logical. */
1317 return is_utf8_common(p, &PL_utf8_idstart, "IdStart");
1321 Perl_is_utf8_idcont(pTHX_ const U8 *p)
1325 PERL_ARGS_ASSERT_IS_UTF8_IDCONT;
1329 return is_utf8_common(p, &PL_utf8_idcont, "IdContinue");
1333 Perl_is_utf8_alpha(pTHX_ const U8 *p)
1337 PERL_ARGS_ASSERT_IS_UTF8_ALPHA;
1339 return is_utf8_common(p, &PL_utf8_alpha, "IsAlpha");
1343 Perl_is_utf8_ascii(pTHX_ const U8 *p)
1347 PERL_ARGS_ASSERT_IS_UTF8_ASCII;
1349 return is_utf8_common(p, &PL_utf8_ascii, "IsAscii");
1353 Perl_is_utf8_space(pTHX_ const U8 *p)
1357 PERL_ARGS_ASSERT_IS_UTF8_SPACE;
1359 return is_utf8_common(p, &PL_utf8_space, "IsSpacePerl");
1363 Perl_is_utf8_digit(pTHX_ const U8 *p)
1367 PERL_ARGS_ASSERT_IS_UTF8_DIGIT;
1369 return is_utf8_common(p, &PL_utf8_digit, "IsDigit");
1373 Perl_is_utf8_upper(pTHX_ const U8 *p)
1377 PERL_ARGS_ASSERT_IS_UTF8_UPPER;
1379 return is_utf8_common(p, &PL_utf8_upper, "IsUppercase");
1383 Perl_is_utf8_lower(pTHX_ const U8 *p)
1387 PERL_ARGS_ASSERT_IS_UTF8_LOWER;
1389 return is_utf8_common(p, &PL_utf8_lower, "IsLowercase");
1393 Perl_is_utf8_cntrl(pTHX_ const U8 *p)
1397 PERL_ARGS_ASSERT_IS_UTF8_CNTRL;
1399 return is_utf8_common(p, &PL_utf8_cntrl, "IsCntrl");
1403 Perl_is_utf8_graph(pTHX_ const U8 *p)
1407 PERL_ARGS_ASSERT_IS_UTF8_GRAPH;
1409 return is_utf8_common(p, &PL_utf8_graph, "IsGraph");
1413 Perl_is_utf8_print(pTHX_ const U8 *p)
1417 PERL_ARGS_ASSERT_IS_UTF8_PRINT;
1419 return is_utf8_common(p, &PL_utf8_print, "IsPrint");
1423 Perl_is_utf8_punct(pTHX_ const U8 *p)
1427 PERL_ARGS_ASSERT_IS_UTF8_PUNCT;
1429 return is_utf8_common(p, &PL_utf8_punct, "IsPunct");
1433 Perl_is_utf8_xdigit(pTHX_ const U8 *p)
1437 PERL_ARGS_ASSERT_IS_UTF8_XDIGIT;
1439 return is_utf8_common(p, &PL_utf8_xdigit, "Isxdigit");
1443 Perl_is_utf8_mark(pTHX_ const U8 *p)
1447 PERL_ARGS_ASSERT_IS_UTF8_MARK;
1449 return is_utf8_common(p, &PL_utf8_mark, "IsM");
1453 =for apidoc to_utf8_case
1455 The "p" contains the pointer to the UTF-8 string encoding
1456 the character that is being converted.
1458 The "ustrp" is a pointer to the character buffer to put the
1459 conversion result to. The "lenp" is a pointer to the length
1462 The "swashp" is a pointer to the swash to use.
1464 Both the special and normal mappings are stored lib/unicore/To/Foo.pl,
1465 and loaded by SWASHNEW, using lib/utf8_heavy.pl. The special (usually,
1466 but not always, a multicharacter mapping), is tried first.
1468 The "special" is a string like "utf8::ToSpecLower", which means the
1469 hash %utf8::ToSpecLower. The access to the hash is through
1470 Perl_to_utf8_case().
1472 The "normal" is a string like "ToLower" which means the swash
1478 Perl_to_utf8_case(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp,
1479 SV **swashp, const char *normal, const char *special)
1482 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1484 const UV uv0 = utf8_to_uvchr(p, NULL);
1485 /* The NATIVE_TO_UNI() and UNI_TO_NATIVE() mappings
1486 * are necessary in EBCDIC, they are redundant no-ops
1487 * in ASCII-ish platforms, and hopefully optimized away. */
1488 const UV uv1 = NATIVE_TO_UNI(uv0);
1490 PERL_ARGS_ASSERT_TO_UTF8_CASE;
1492 uvuni_to_utf8(tmpbuf, uv1);
1494 if (!*swashp) /* load on-demand */
1495 *swashp = swash_init("utf8", normal, &PL_sv_undef, 4, 0);
1497 /* The 0xDF is the only special casing Unicode code point below 0x100. */
1498 if (special && (uv1 == 0xDF || uv1 > 0xFF)) {
1499 /* It might be "special" (sometimes, but not always,
1500 * a multicharacter mapping) */
1501 HV * const hv = get_hv(special, 0);
1505 (svp = hv_fetch(hv, (const char*)tmpbuf, UNISKIP(uv1), FALSE)) &&
1509 s = SvPV_const(*svp, len);
1511 len = uvuni_to_utf8(ustrp, NATIVE_TO_UNI(*(U8*)s)) - ustrp;
1514 /* If we have EBCDIC we need to remap the characters
1515 * since any characters in the low 256 are Unicode
1516 * code points, not EBCDIC. */
1517 U8 *t = (U8*)s, *tend = t + len, *d;
1524 const UV c = utf8_to_uvchr(t, &tlen);
1526 d = uvchr_to_utf8(d, UNI_TO_NATIVE(c));
1535 d = uvchr_to_utf8(d, UNI_TO_NATIVE(*t));
1540 Copy(tmpbuf, ustrp, len, U8);
1542 Copy(s, ustrp, len, U8);
1548 if (!len && *swashp) {
1549 const UV uv2 = swash_fetch(*swashp, tmpbuf, TRUE);
1552 /* It was "normal" (a single character mapping). */
1553 const UV uv3 = UNI_TO_NATIVE(uv2);
1554 len = uvchr_to_utf8(ustrp, uv3) - ustrp;
1558 if (!len) /* Neither: just copy. */
1559 len = uvchr_to_utf8(ustrp, uv0) - ustrp;
1564 return len ? utf8_to_uvchr(ustrp, 0) : 0;
1568 =for apidoc to_utf8_upper
1570 Convert the UTF-8 encoded character at p to its uppercase version and
1571 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
1572 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since
1573 the uppercase version may be longer than the original character.
1575 The first character of the uppercased version is returned
1576 (but note, as explained above, that there may be more.)
1581 Perl_to_utf8_upper(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
1585 PERL_ARGS_ASSERT_TO_UTF8_UPPER;
1587 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
1588 &PL_utf8_toupper, "ToUpper", "utf8::ToSpecUpper");
1592 =for apidoc to_utf8_title
1594 Convert the UTF-8 encoded character at p to its titlecase version and
1595 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
1596 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
1597 titlecase version may be longer than the original character.
1599 The first character of the titlecased version is returned
1600 (but note, as explained above, that there may be more.)
1605 Perl_to_utf8_title(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
1609 PERL_ARGS_ASSERT_TO_UTF8_TITLE;
1611 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
1612 &PL_utf8_totitle, "ToTitle", "utf8::ToSpecTitle");
1616 =for apidoc to_utf8_lower
1618 Convert the UTF-8 encoded character at p to its lowercase version and
1619 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
1620 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
1621 lowercase version may be longer than the original character.
1623 The first character of the lowercased version is returned
1624 (but note, as explained above, that there may be more.)
1629 Perl_to_utf8_lower(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
1633 PERL_ARGS_ASSERT_TO_UTF8_LOWER;
1635 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
1636 &PL_utf8_tolower, "ToLower", "utf8::ToSpecLower");
1640 =for apidoc to_utf8_fold
1642 Convert the UTF-8 encoded character at p to its foldcase version and
1643 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
1644 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
1645 foldcase version may be longer than the original character (up to
1648 The first character of the foldcased version is returned
1649 (but note, as explained above, that there may be more.)
1654 Perl_to_utf8_fold(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
1658 PERL_ARGS_ASSERT_TO_UTF8_FOLD;
1660 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
1661 &PL_utf8_tofold, "ToFold", "utf8::ToSpecFold");
1665 * A "swash" is a swatch hash.
1666 * A "swatch" is a bit vector generated by utf8.c:S_swash_get().
1667 * C<pkg> is a pointer to a package name for SWASHNEW, should be "utf8".
1668 * For other parameters, see utf8::SWASHNEW in lib/utf8_heavy.pl.
1671 Perl_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv, I32 minbits, I32 none)
1676 const size_t pkg_len = strlen(pkg);
1677 const size_t name_len = strlen(name);
1678 HV * const stash = gv_stashpvn(pkg, pkg_len, 0);
1681 PERL_ARGS_ASSERT_SWASH_INIT;
1683 PUSHSTACKi(PERLSI_MAGIC);
1688 if (!gv_fetchmeth(stash, "SWASHNEW", 8, -1)) { /* demand load utf8 */
1690 errsv_save = newSVsv(ERRSV);
1691 /* It is assumed that callers of this routine are not passing in any
1692 user derived data. */
1693 /* Need to do this after save_re_context() as it will set PL_tainted to
1694 1 while saving $1 etc (see the code after getrx: in Perl_magic_get).
1695 Even line to create errsv_save can turn on PL_tainted. */
1696 SAVEBOOL(PL_tainted);
1698 Perl_load_module(aTHX_ PERL_LOADMOD_NOIMPORT, newSVpvn(pkg,pkg_len),
1701 sv_setsv(ERRSV, errsv_save);
1702 SvREFCNT_dec(errsv_save);
1708 mPUSHp(pkg, pkg_len);
1709 mPUSHp(name, name_len);
1714 errsv_save = newSVsv(ERRSV);
1715 if (call_method("SWASHNEW", G_SCALAR))
1716 retval = newSVsv(*PL_stack_sp--);
1718 retval = &PL_sv_undef;
1720 sv_setsv(ERRSV, errsv_save);
1721 SvREFCNT_dec(errsv_save);
1724 if (IN_PERL_COMPILETIME) {
1725 CopHINTS_set(PL_curcop, PL_hints);
1727 if (!SvROK(retval) || SvTYPE(SvRV(retval)) != SVt_PVHV) {
1729 Perl_croak(aTHX_ "Can't find Unicode property definition \"%"SVf"\"",
1731 Perl_croak(aTHX_ "SWASHNEW didn't return an HV ref");
1737 /* This API is wrong for special case conversions since we may need to
1738 * return several Unicode characters for a single Unicode character
1739 * (see lib/unicore/SpecCase.txt) The SWASHGET in lib/utf8_heavy.pl is
1740 * the lower-level routine, and it is similarly broken for returning
1741 * multiple values. --jhi */
1742 /* Now SWASHGET is recasted into S_swash_get in this file. */
1745 * Returns the value of property/mapping C<swash> for the first character
1746 * of the string C<ptr>. If C<do_utf8> is true, the string C<ptr> is
1747 * assumed to be in utf8. If C<do_utf8> is false, the string C<ptr> is
1748 * assumed to be in native 8-bit encoding. Caches the swatch in C<swash>.
1751 Perl_swash_fetch(pTHX_ SV *swash, const U8 *ptr, bool do_utf8)
1754 HV *const hv = MUTABLE_HV(SvRV(swash));
1759 const U8 *tmps = NULL;
1763 const UV c = NATIVE_TO_ASCII(*ptr);
1765 PERL_ARGS_ASSERT_SWASH_FETCH;
1767 if (!do_utf8 && !UNI_IS_INVARIANT(c)) {
1768 tmputf8[0] = (U8)UTF8_EIGHT_BIT_HI(c);
1769 tmputf8[1] = (U8)UTF8_EIGHT_BIT_LO(c);
1772 /* Given a UTF-X encoded char 0xAA..0xYY,0xZZ
1773 * then the "swatch" is a vec() for al the chars which start
1775 * So the key in the hash (klen) is length of encoded char -1
1777 klen = UTF8SKIP(ptr) - 1;
1781 /* If char in invariant then swatch is for all the invariant chars
1782 * In both UTF-8 and UTF-8-MOD that happens to be UTF_CONTINUATION_MARK
1784 needents = UTF_CONTINUATION_MARK;
1785 off = NATIVE_TO_UTF(ptr[klen]);
1788 /* If char is encoded then swatch is for the prefix */
1789 needents = (1 << UTF_ACCUMULATION_SHIFT);
1790 off = NATIVE_TO_UTF(ptr[klen]) & UTF_CONTINUATION_MASK;
1794 * This single-entry cache saves about 1/3 of the utf8 overhead in test
1795 * suite. (That is, only 7-8% overall over just a hash cache. Still,
1796 * it's nothing to sniff at.) Pity we usually come through at least
1797 * two function calls to get here...
1799 * NB: this code assumes that swatches are never modified, once generated!
1802 if (hv == PL_last_swash_hv &&
1803 klen == PL_last_swash_klen &&
1804 (!klen || memEQ((char *)ptr, (char *)PL_last_swash_key, klen)) )
1806 tmps = PL_last_swash_tmps;
1807 slen = PL_last_swash_slen;
1810 /* Try our second-level swatch cache, kept in a hash. */
1811 SV** svp = hv_fetch(hv, (const char*)ptr, klen, FALSE);
1813 /* If not cached, generate it via swash_get */
1814 if (!svp || !SvPOK(*svp)
1815 || !(tmps = (const U8*)SvPV_const(*svp, slen))) {
1816 /* We use utf8n_to_uvuni() as we want an index into
1817 Unicode tables, not a native character number.
1819 const UV code_point = utf8n_to_uvuni(ptr, UTF8_MAXBYTES, 0,
1821 0 : UTF8_ALLOW_ANY);
1822 swatch = swash_get(swash,
1823 /* On EBCDIC & ~(0xA0-1) isn't a useful thing to do */
1824 (klen) ? (code_point & ~(needents - 1)) : 0,
1827 if (IN_PERL_COMPILETIME)
1828 CopHINTS_set(PL_curcop, PL_hints);
1830 svp = hv_store(hv, (const char *)ptr, klen, swatch, 0);
1832 if (!svp || !(tmps = (U8*)SvPV(*svp, slen))
1833 || (slen << 3) < needents)
1834 Perl_croak(aTHX_ "panic: swash_fetch got improper swatch");
1837 PL_last_swash_hv = hv;
1838 assert(klen <= sizeof(PL_last_swash_key));
1839 PL_last_swash_klen = (U8)klen;
1840 /* FIXME change interpvar.h? */
1841 PL_last_swash_tmps = (U8 *) tmps;
1842 PL_last_swash_slen = slen;
1844 Copy(ptr, PL_last_swash_key, klen, U8);
1847 switch ((int)((slen << 3) / needents)) {
1849 bit = 1 << (off & 7);
1851 return (tmps[off] & bit) != 0;
1856 return (tmps[off] << 8) + tmps[off + 1] ;
1859 return (tmps[off] << 24) + (tmps[off+1] << 16) + (tmps[off+2] << 8) + tmps[off + 3] ;
1861 Perl_croak(aTHX_ "panic: swash_fetch got swatch of unexpected bit width");
1862 NORETURN_FUNCTION_END;
1866 * Returns a swatch (a bit vector string) for a code point sequence
1867 * that starts from the value C<start> and comprises the number C<span>.
1868 * A C<swash> must be an object created by SWASHNEW (see lib/utf8_heavy.pl).
1869 * Should be used via swash_fetch, which will cache the swatch in C<swash>.
1872 S_swash_get(pTHX_ SV* swash, UV start, UV span)
1875 U8 *l, *lend, *x, *xend, *s;
1876 STRLEN lcur, xcur, scur;
1877 HV *const hv = MUTABLE_HV(SvRV(swash));
1878 SV** const listsvp = hv_fetchs(hv, "LIST", FALSE);
1879 SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE);
1880 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
1881 SV** const nonesvp = hv_fetchs(hv, "NONE", FALSE);
1882 SV** const extssvp = hv_fetchs(hv, "EXTRAS", FALSE);
1883 const U8* const typestr = (U8*)SvPV_nolen(*typesvp);
1884 const int typeto = typestr[0] == 'T' && typestr[1] == 'o';
1885 const STRLEN bits = SvUV(*bitssvp);
1886 const STRLEN octets = bits >> 3; /* if bits == 1, then octets == 0 */
1887 const UV none = SvUV(*nonesvp);
1888 const UV end = start + span;
1890 PERL_ARGS_ASSERT_SWASH_GET;
1892 if (bits != 1 && bits != 8 && bits != 16 && bits != 32) {
1893 Perl_croak(aTHX_ "panic: swash_get doesn't expect bits %"UVuf,
1897 /* create and initialize $swatch */
1898 scur = octets ? (span * octets) : (span + 7) / 8;
1899 swatch = newSV(scur);
1901 s = (U8*)SvPVX(swatch);
1902 if (octets && none) {
1903 const U8* const e = s + scur;
1906 *s++ = (U8)(none & 0xff);
1907 else if (bits == 16) {
1908 *s++ = (U8)((none >> 8) & 0xff);
1909 *s++ = (U8)( none & 0xff);
1911 else if (bits == 32) {
1912 *s++ = (U8)((none >> 24) & 0xff);
1913 *s++ = (U8)((none >> 16) & 0xff);
1914 *s++ = (U8)((none >> 8) & 0xff);
1915 *s++ = (U8)( none & 0xff);
1921 (void)memzero((U8*)s, scur + 1);
1923 SvCUR_set(swatch, scur);
1924 s = (U8*)SvPVX(swatch);
1926 /* read $swash->{LIST} */
1927 l = (U8*)SvPV(*listsvp, lcur);
1932 I32 flags = PERL_SCAN_SILENT_ILLDIGIT | PERL_SCAN_DISALLOW_PREFIX;
1934 U8* const nl = (U8*)memchr(l, '\n', lend - l);
1937 min = grok_hex((char *)l, &numlen, &flags, NULL);
1941 l = nl + 1; /* 1 is length of "\n" */
1945 l = lend; /* to LIST's end at which \n is not found */
1951 flags = PERL_SCAN_SILENT_ILLDIGIT | PERL_SCAN_DISALLOW_PREFIX;
1953 max = grok_hex((char *)l, &numlen, &flags, NULL);
1962 flags = PERL_SCAN_SILENT_ILLDIGIT |
1963 PERL_SCAN_DISALLOW_PREFIX;
1965 val = grok_hex((char *)l, &numlen, &flags, NULL);
1974 Perl_croak(aTHX_ "%s: illegal mapping '%s'",
1980 val = 0; /* bits == 1, then val should be ignored */
1987 Perl_croak(aTHX_ "%s: illegal mapping '%s'", typestr, l);
1991 val = 0; /* bits == 1, then val should be ignored */
2005 if (!none || val < none) {
2010 for (key = min; key <= max; key++) {
2014 /* offset must be non-negative (start <= min <= key < end) */
2015 offset = octets * (key - start);
2017 s[offset] = (U8)(val & 0xff);
2018 else if (bits == 16) {
2019 s[offset ] = (U8)((val >> 8) & 0xff);
2020 s[offset + 1] = (U8)( val & 0xff);
2022 else if (bits == 32) {
2023 s[offset ] = (U8)((val >> 24) & 0xff);
2024 s[offset + 1] = (U8)((val >> 16) & 0xff);
2025 s[offset + 2] = (U8)((val >> 8) & 0xff);
2026 s[offset + 3] = (U8)( val & 0xff);
2029 if (!none || val < none)
2033 else { /* bits == 1, then val should be ignored */
2037 for (key = min; key <= max; key++) {
2038 const STRLEN offset = (STRLEN)(key - start);
2041 s[offset >> 3] |= 1 << (offset & 7);
2047 /* read $swash->{EXTRAS} */
2048 x = (U8*)SvPV(*extssvp, xcur);
2056 SV **otherbitssvp, *other;
2060 const U8 opc = *x++;
2064 nl = (U8*)memchr(x, '\n', xend - x);
2066 if (opc != '-' && opc != '+' && opc != '!' && opc != '&') {
2068 x = nl + 1; /* 1 is length of "\n" */
2072 x = xend; /* to EXTRAS' end at which \n is not found */
2079 namelen = nl - namestr;
2083 namelen = xend - namestr;
2087 othersvp = hv_fetch(hv, (char *)namestr, namelen, FALSE);
2088 otherhv = MUTABLE_HV(SvRV(*othersvp));
2089 otherbitssvp = hv_fetchs(otherhv, "BITS", FALSE);
2090 otherbits = (STRLEN)SvUV(*otherbitssvp);
2091 if (bits < otherbits)
2092 Perl_croak(aTHX_ "panic: swash_get found swatch size mismatch");
2094 /* The "other" swatch must be destroyed after. */
2095 other = swash_get(*othersvp, start, span);
2096 o = (U8*)SvPV(other, olen);
2099 Perl_croak(aTHX_ "panic: swash_get got improper swatch");
2101 s = (U8*)SvPV(swatch, slen);
2102 if (bits == 1 && otherbits == 1) {
2104 Perl_croak(aTHX_ "panic: swash_get found swatch length mismatch");
2128 STRLEN otheroctets = otherbits >> 3;
2130 U8* const send = s + slen;
2135 if (otherbits == 1) {
2136 otherval = (o[offset >> 3] >> (offset & 7)) & 1;
2140 STRLEN vlen = otheroctets;
2148 if (opc == '+' && otherval)
2149 NOOP; /* replace with otherval */
2150 else if (opc == '!' && !otherval)
2152 else if (opc == '-' && otherval)
2154 else if (opc == '&' && !otherval)
2157 s += octets; /* no replacement */
2162 *s++ = (U8)( otherval & 0xff);
2163 else if (bits == 16) {
2164 *s++ = (U8)((otherval >> 8) & 0xff);
2165 *s++ = (U8)( otherval & 0xff);
2167 else if (bits == 32) {
2168 *s++ = (U8)((otherval >> 24) & 0xff);
2169 *s++ = (U8)((otherval >> 16) & 0xff);
2170 *s++ = (U8)((otherval >> 8) & 0xff);
2171 *s++ = (U8)( otherval & 0xff);
2175 sv_free(other); /* through with it! */
2181 =for apidoc uvchr_to_utf8
2183 Adds the UTF-8 representation of the Native codepoint C<uv> to the end
2184 of the string C<d>; C<d> should be have at least C<UTF8_MAXBYTES+1> free
2185 bytes available. The return value is the pointer to the byte after the
2186 end of the new character. In other words,
2188 d = uvchr_to_utf8(d, uv);
2190 is the recommended wide native character-aware way of saying
2197 /* On ASCII machines this is normally a macro but we want a
2198 real function in case XS code wants it
2201 Perl_uvchr_to_utf8(pTHX_ U8 *d, UV uv)
2203 PERL_ARGS_ASSERT_UVCHR_TO_UTF8;
2205 return Perl_uvuni_to_utf8_flags(aTHX_ d, NATIVE_TO_UNI(uv), 0);
2209 Perl_uvchr_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags)
2211 PERL_ARGS_ASSERT_UVCHR_TO_UTF8_FLAGS;
2213 return Perl_uvuni_to_utf8_flags(aTHX_ d, NATIVE_TO_UNI(uv), flags);
2217 =for apidoc utf8n_to_uvchr
2220 Returns the native character value of the first character in the string
2222 which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
2223 length, in bytes, of that character.
2225 Allows length and flags to be passed to low level routine.
2229 /* On ASCII machines this is normally a macro but we want
2230 a real function in case XS code wants it
2233 Perl_utf8n_to_uvchr(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen,
2236 const UV uv = Perl_utf8n_to_uvuni(aTHX_ s, curlen, retlen, flags);
2238 PERL_ARGS_ASSERT_UTF8N_TO_UVCHR;
2240 return UNI_TO_NATIVE(uv);
2244 =for apidoc pv_uni_display
2246 Build to the scalar dsv a displayable version of the string spv,
2247 length len, the displayable version being at most pvlim bytes long
2248 (if longer, the rest is truncated and "..." will be appended).
2250 The flags argument can have UNI_DISPLAY_ISPRINT set to display
2251 isPRINT()able characters as themselves, UNI_DISPLAY_BACKSLASH
2252 to display the \\[nrfta\\] as the backslashed versions (like '\n')
2253 (UNI_DISPLAY_BACKSLASH is preferred over UNI_DISPLAY_ISPRINT for \\).
2254 UNI_DISPLAY_QQ (and its alias UNI_DISPLAY_REGEX) have both
2255 UNI_DISPLAY_BACKSLASH and UNI_DISPLAY_ISPRINT turned on.
2257 The pointer to the PV of the dsv is returned.
2261 Perl_pv_uni_display(pTHX_ SV *dsv, const U8 *spv, STRLEN len, STRLEN pvlim, UV flags)
2266 PERL_ARGS_ASSERT_PV_UNI_DISPLAY;
2270 for (s = (const char *)spv, e = s + len; s < e; s += UTF8SKIP(s)) {
2272 /* This serves double duty as a flag and a character to print after
2273 a \ when flags & UNI_DISPLAY_BACKSLASH is true.
2277 if (pvlim && SvCUR(dsv) >= pvlim) {
2281 u = utf8_to_uvchr((U8*)s, 0);
2283 const unsigned char c = (unsigned char)u & 0xFF;
2284 if (flags & UNI_DISPLAY_BACKSLASH) {
2301 const char string = ok;
2302 sv_catpvs(dsv, "\\");
2303 sv_catpvn(dsv, &string, 1);
2306 /* isPRINT() is the locale-blind version. */
2307 if (!ok && (flags & UNI_DISPLAY_ISPRINT) && isPRINT(c)) {
2308 const char string = c;
2309 sv_catpvn(dsv, &string, 1);
2314 Perl_sv_catpvf(aTHX_ dsv, "\\x{%"UVxf"}", u);
2317 sv_catpvs(dsv, "...");
2323 =for apidoc sv_uni_display
2325 Build to the scalar dsv a displayable version of the scalar sv,
2326 the displayable version being at most pvlim bytes long
2327 (if longer, the rest is truncated and "..." will be appended).
2329 The flags argument is as in pv_uni_display().
2331 The pointer to the PV of the dsv is returned.
2336 Perl_sv_uni_display(pTHX_ SV *dsv, SV *ssv, STRLEN pvlim, UV flags)
2338 PERL_ARGS_ASSERT_SV_UNI_DISPLAY;
2340 return Perl_pv_uni_display(aTHX_ dsv, (const U8*)SvPVX_const(ssv),
2341 SvCUR(ssv), pvlim, flags);
2345 =for apidoc ibcmp_utf8
2347 Return true if the strings s1 and s2 differ case-insensitively, false
2348 if not (if they are equal case-insensitively). If u1 is true, the
2349 string s1 is assumed to be in UTF-8-encoded Unicode. If u2 is true,
2350 the string s2 is assumed to be in UTF-8-encoded Unicode. If u1 or u2
2351 are false, the respective string is assumed to be in native 8-bit
2354 If the pe1 and pe2 are non-NULL, the scanning pointers will be copied
2355 in there (they will point at the beginning of the I<next> character).
2356 If the pointers behind pe1 or pe2 are non-NULL, they are the end
2357 pointers beyond which scanning will not continue under any
2358 circumstances. If the byte lengths l1 and l2 are non-zero, s1+l1 and
2359 s2+l2 will be used as goal end pointers that will also stop the scan,
2360 and which qualify towards defining a successful match: all the scans
2361 that define an explicit length must reach their goal pointers for
2362 a match to succeed).
2364 For case-insensitiveness, the "casefolding" of Unicode is used
2365 instead of upper/lowercasing both the characters, see
2366 http://www.unicode.org/unicode/reports/tr21/ (Case Mappings).
2370 Perl_ibcmp_utf8(pTHX_ const char *s1, char **pe1, register UV l1, bool u1, const char *s2, char **pe2, register UV l2, bool u2)
2373 register const U8 *p1 = (const U8*)s1;
2374 register const U8 *p2 = (const U8*)s2;
2375 register const U8 *f1 = NULL;
2376 register const U8 *f2 = NULL;
2377 register U8 *e1 = NULL;
2378 register U8 *q1 = NULL;
2379 register U8 *e2 = NULL;
2380 register U8 *q2 = NULL;
2381 STRLEN n1 = 0, n2 = 0;
2382 U8 foldbuf1[UTF8_MAXBYTES_CASE+1];
2383 U8 foldbuf2[UTF8_MAXBYTES_CASE+1];
2385 STRLEN foldlen1, foldlen2;
2388 PERL_ARGS_ASSERT_IBCMP_UTF8;
2392 /* assert(e1 || l1); */
2393 if (e1 == 0 || (l1 && l1 < (UV)(e1 - (const U8*)s1)))
2394 f1 = (const U8*)s1 + l1;
2397 /* assert(e2 || l2); */
2398 if (e2 == 0 || (l2 && l2 < (UV)(e2 - (const U8*)s2)))
2399 f2 = (const U8*)s2 + l2;
2401 /* This shouldn't happen. However, putting an assert() there makes some
2403 /* assert((e1 == 0 && f1 == 0) || (e2 == 0 && f2 == 0) || (f1 == 0 && f2 == 0)); */
2404 if ((e1 == 0 && f1 == 0) || (e2 == 0 && f2 == 0) || (f1 == 0 && f2 == 0))
2405 return 1; /* mismatch; possible infinite loop or false positive */
2408 natbuf[1] = 0; /* Need to terminate the buffer. */
2410 while ((e1 == 0 || p1 < e1) &&
2411 (f1 == 0 || p1 < f1) &&
2412 (e2 == 0 || p2 < e2) &&
2413 (f2 == 0 || p2 < f2)) {
2416 to_utf8_fold(p1, foldbuf1, &foldlen1);
2418 uvuni_to_utf8(natbuf, (UV) NATIVE_TO_UNI(((UV)*p1)));
2419 to_utf8_fold(natbuf, foldbuf1, &foldlen1);
2426 to_utf8_fold(p2, foldbuf2, &foldlen2);
2428 uvuni_to_utf8(natbuf, (UV) NATIVE_TO_UNI(((UV)*p2)));
2429 to_utf8_fold(natbuf, foldbuf2, &foldlen2);
2435 if ( UTF8SKIP(q1) != UTF8SKIP(q2) ||
2436 (UTF8SKIP(q1) == 1 && *q1 != *q2) ||
2437 memNE((char*)q1, (char*)q2, UTF8SKIP(q1)) )
2438 return 1; /* mismatch */
2445 p1 += u1 ? UTF8SKIP(p1) : 1;
2447 p2 += u2 ? UTF8SKIP(p2) : 1;
2451 /* A match is defined by all the scans that specified
2452 * an explicit length reaching their final goals. */
2453 match = (f1 == 0 || p1 == f1) && (f2 == 0 || p2 == f2);
2462 return match ? 0 : 1; /* 0 match, 1 mismatch */
2467 * c-indentation-style: bsd
2469 * indent-tabs-mode: t
2472 * ex: set ts=8 sts=4 sw=4 noet: