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 are
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 the first C<len> bytes of the given string are the same whether
61 or not the string is encoded in UTF-8 (or UTF-EBCDIC on EBCDIC machines). That
62 is, if they are invariant. On ASCII-ish machines, only ASCII characters
63 fit this definition, hence the function's name.
65 If C<len> is 0, it will be calculated using C<strlen(s)>.
67 See also is_utf8_string(), is_utf8_string_loclen(), and is_utf8_string_loc().
73 Perl_is_ascii_string(const U8 *s, STRLEN len)
75 const U8* const send = s + (len ? len : strlen((const char *)s));
78 PERL_ARGS_ASSERT_IS_ASCII_STRING;
80 for (; x < send; ++x) {
81 if (!UTF8_IS_INVARIANT(*x))
89 =for apidoc uvuni_to_utf8_flags
91 Adds the UTF-8 representation of the code point C<uv> to the end
92 of the string C<d>; C<d> should have at least C<UTF8_MAXBYTES+1> free
93 bytes available. The return value is the pointer to the byte after the
94 end of the new character. In other words,
96 d = uvuni_to_utf8_flags(d, uv, flags);
100 d = uvuni_to_utf8(d, uv);
102 (which is equivalent to)
104 d = uvuni_to_utf8_flags(d, uv, 0);
106 This is the recommended Unicode-aware way of saying
110 This function will convert to UTF-8 (and not warn) even code points that aren't
111 legal Unicode or are problematic, unless C<flags> contains one or more of the
113 If C<uv> is a Unicode surrogate code point and UNICODE_WARN_SURROGATE is set,
114 the function will raise a warning, provided UTF8 warnings are enabled. If instead
115 UNICODE_DISALLOW_SURROGATE is set, the function will fail and return NULL.
116 If both flags are set, the function will both warn and return NULL.
118 The UNICODE_WARN_NONCHAR and UNICODE_DISALLOW_NONCHAR flags correspondingly
119 affect how the function handles a Unicode non-character. And, likewise for the
120 UNICODE_WARN_SUPER and UNICODE_DISALLOW_SUPER flags, and code points that are
121 above the Unicode maximum of 0x10FFFF. Code points above 0x7FFF_FFFF (which are
122 even less portable) can be warned and/or disallowed even if other above-Unicode
123 code points are accepted by the UNICODE_WARN_FE_FF and UNICODE_DISALLOW_FE_FF
126 And finally, the flag UNICODE_WARN_ILLEGAL_INTERCHANGE selects all four of the
127 above WARN flags; and UNICODE_DISALLOW_ILLEGAL_INTERCHANGE selects all four
135 Perl_uvuni_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags)
137 PERL_ARGS_ASSERT_UVUNI_TO_UTF8_FLAGS;
139 if (ckWARN_d(WARN_UTF8)) {
140 if (UNICODE_IS_SURROGATE(uv)) {
141 if (flags & UNICODE_WARN_SURROGATE) {
142 Perl_ck_warner_d(aTHX_ packWARN(WARN_SURROGATE),
143 "UTF-16 surrogate U+%04"UVXf, uv);
145 if (flags & UNICODE_DISALLOW_SURROGATE) {
149 else if (UNICODE_IS_SUPER(uv)) {
150 if (flags & UNICODE_WARN_SUPER
151 || (UNICODE_IS_FE_FF(uv) && (flags & UNICODE_WARN_FE_FF)))
153 Perl_ck_warner_d(aTHX_ packWARN(WARN_NON_UNICODE),
154 "Code point 0x%04"UVXf" is not Unicode, may not be portable", uv);
156 if (flags & UNICODE_DISALLOW_SUPER
157 || (UNICODE_IS_FE_FF(uv) && (flags & UNICODE_DISALLOW_FE_FF)))
162 else if (UNICODE_IS_NONCHAR(uv)) {
163 if (flags & UNICODE_WARN_NONCHAR) {
164 Perl_ck_warner_d(aTHX_ packWARN(WARN_NONCHAR),
165 "Unicode non-character U+%04"UVXf" is illegal for open interchange",
168 if (flags & UNICODE_DISALLOW_NONCHAR) {
173 if (UNI_IS_INVARIANT(uv)) {
174 *d++ = (U8)UTF_TO_NATIVE(uv);
179 STRLEN len = UNISKIP(uv);
182 *p-- = (U8)UTF_TO_NATIVE((uv & UTF_CONTINUATION_MASK) | UTF_CONTINUATION_MARK);
183 uv >>= UTF_ACCUMULATION_SHIFT;
185 *p = (U8)UTF_TO_NATIVE((uv & UTF_START_MASK(len)) | UTF_START_MARK(len));
188 #else /* Non loop style */
190 *d++ = (U8)(( uv >> 6) | 0xc0);
191 *d++ = (U8)(( uv & 0x3f) | 0x80);
195 *d++ = (U8)(( uv >> 12) | 0xe0);
196 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
197 *d++ = (U8)(( uv & 0x3f) | 0x80);
201 *d++ = (U8)(( uv >> 18) | 0xf0);
202 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
203 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
204 *d++ = (U8)(( uv & 0x3f) | 0x80);
207 if (uv < 0x4000000) {
208 *d++ = (U8)(( uv >> 24) | 0xf8);
209 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
210 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
211 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
212 *d++ = (U8)(( uv & 0x3f) | 0x80);
215 if (uv < 0x80000000) {
216 *d++ = (U8)(( uv >> 30) | 0xfc);
217 *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80);
218 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
219 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
220 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
221 *d++ = (U8)(( uv & 0x3f) | 0x80);
225 if (uv < UTF8_QUAD_MAX)
228 *d++ = 0xfe; /* Can't match U+FEFF! */
229 *d++ = (U8)(((uv >> 30) & 0x3f) | 0x80);
230 *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80);
231 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
232 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
233 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
234 *d++ = (U8)(( uv & 0x3f) | 0x80);
239 *d++ = 0xff; /* Can't match U+FFFE! */
240 *d++ = 0x80; /* 6 Reserved bits */
241 *d++ = (U8)(((uv >> 60) & 0x0f) | 0x80); /* 2 Reserved bits */
242 *d++ = (U8)(((uv >> 54) & 0x3f) | 0x80);
243 *d++ = (U8)(((uv >> 48) & 0x3f) | 0x80);
244 *d++ = (U8)(((uv >> 42) & 0x3f) | 0x80);
245 *d++ = (U8)(((uv >> 36) & 0x3f) | 0x80);
246 *d++ = (U8)(((uv >> 30) & 0x3f) | 0x80);
247 *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80);
248 *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
249 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
250 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
251 *d++ = (U8)(( uv & 0x3f) | 0x80);
255 #endif /* Loop style */
260 Tests if some arbitrary number of bytes begins in a valid UTF-8
261 character. Note that an INVARIANT (i.e. ASCII) character is a valid
262 UTF-8 character. The actual number of bytes in the UTF-8 character
263 will be returned if it is valid, otherwise 0.
265 This is the "slow" version as opposed to the "fast" version which is
266 the "unrolled" IS_UTF8_CHAR(). E.g. for t/uni/class.t the speed
267 difference is a factor of 2 to 3. For lengths (UTF8SKIP(s)) of four
268 or less you should use the IS_UTF8_CHAR(), for lengths of five or more
269 you should use the _slow(). In practice this means that the _slow()
270 will be used very rarely, since the maximum Unicode code point (as of
271 Unicode 4.1) is U+10FFFF, which encodes in UTF-8 to four bytes. Only
272 the "Perl extended UTF-8" (the infamous 'v-strings') will encode into
277 S_is_utf8_char_slow(const U8 *s, const STRLEN len)
283 PERL_ARGS_ASSERT_IS_UTF8_CHAR_SLOW;
285 if (UTF8_IS_INVARIANT(u))
288 if (!UTF8_IS_START(u))
291 if (len < 2 || !UTF8_IS_CONTINUATION(s[1]))
297 u = NATIVE_TO_UTF(u);
299 u &= UTF_START_MASK(len);
303 if (!UTF8_IS_CONTINUATION(*s))
305 uv = UTF8_ACCUMULATE(uv, *s);
312 if ((STRLEN)UNISKIP(uv) < len)
319 =for apidoc is_utf8_char
321 Tests if some arbitrary number of bytes begins in a valid UTF-8
322 character. Note that an INVARIANT (i.e. ASCII on non-EBCDIC machines)
323 character is a valid UTF-8 character. The actual number of bytes in the UTF-8
324 character will be returned if it is valid, otherwise 0.
328 Perl_is_utf8_char(const U8 *s)
330 const STRLEN len = UTF8SKIP(s);
332 PERL_ARGS_ASSERT_IS_UTF8_CHAR;
334 if (IS_UTF8_CHAR_FAST(len))
335 return IS_UTF8_CHAR(s, len) ? len : 0;
336 #endif /* #ifdef IS_UTF8_CHAR */
337 return is_utf8_char_slow(s, len);
342 =for apidoc is_utf8_string
344 Returns true if first C<len> bytes of the given string form a valid
345 UTF-8 string, false otherwise. If C<len> is 0, it will be calculated
346 using C<strlen(s)>. Note that 'a valid UTF-8 string' does not mean 'a
347 string that contains code points above 0x7F encoded in UTF-8' because a
348 valid ASCII string is a valid UTF-8 string.
350 See also is_ascii_string(), is_utf8_string_loclen(), and is_utf8_string_loc().
356 Perl_is_utf8_string(const U8 *s, STRLEN len)
358 const U8* const send = s + (len ? len : strlen((const char *)s));
361 PERL_ARGS_ASSERT_IS_UTF8_STRING;
365 /* Inline the easy bits of is_utf8_char() here for speed... */
366 if (UTF8_IS_INVARIANT(*x))
368 else if (!UTF8_IS_START(*x))
371 /* ... and call is_utf8_char() only if really needed. */
374 if (IS_UTF8_CHAR_FAST(c)) {
375 if (!IS_UTF8_CHAR(x, c))
379 c = is_utf8_char_slow(x, c);
382 #endif /* #ifdef IS_UTF8_CHAR */
397 Implemented as a macro in utf8.h
399 =for apidoc is_utf8_string_loc
401 Like is_utf8_string() but stores the location of the failure (in the
402 case of "utf8ness failure") or the location s+len (in the case of
403 "utf8ness success") in the C<ep>.
405 See also is_utf8_string_loclen() and is_utf8_string().
407 =for apidoc is_utf8_string_loclen
409 Like is_utf8_string() but stores the location of the failure (in the
410 case of "utf8ness failure") or the location s+len (in the case of
411 "utf8ness success") in the C<ep>, and the number of UTF-8
412 encoded characters in the C<el>.
414 See also is_utf8_string_loc() and is_utf8_string().
420 Perl_is_utf8_string_loclen(const U8 *s, STRLEN len, const U8 **ep, STRLEN *el)
422 const U8* const send = s + (len ? len : strlen((const char *)s));
427 PERL_ARGS_ASSERT_IS_UTF8_STRING_LOCLEN;
430 /* Inline the easy bits of is_utf8_char() here for speed... */
431 if (UTF8_IS_INVARIANT(*x))
433 else if (!UTF8_IS_START(*x))
436 /* ... and call is_utf8_char() only if really needed. */
439 if (IS_UTF8_CHAR_FAST(c)) {
440 if (!IS_UTF8_CHAR(x, c))
443 c = is_utf8_char_slow(x, c);
446 #endif /* #ifdef IS_UTF8_CHAR */
465 =for apidoc utf8n_to_uvuni
467 Bottom level UTF-8 decode routine.
468 Returns the code point value of the first character in the string C<s>
469 which is assumed to be in UTF-8 (or UTF-EBCDIC) encoding and no longer than
470 C<curlen> bytes; C<retlen> will be set to the length, in bytes, of that
473 The value of C<flags> determines the behavior when C<s> does not point to a
474 well-formed UTF-8 character. If C<flags> is 0, when a malformation is found,
475 C<retlen> is set to the expected length of the UTF-8 character in bytes, zero
476 is returned, and if UTF-8 warnings haven't been lexically disabled, a warning
479 Various ALLOW flags can be set in C<flags> to allow (and not warn on)
480 individual types of malformations, such as the sequence being overlong (that
481 is, when there is a shorter sequence that can express the same code point;
482 overlong sequences are expressly forbidden in the UTF-8 standard due to
483 potential security issues). Another malformation example is the first byte of
484 a character not being a legal first byte. See F<utf8.h> for the list of such
485 flags. Of course, the value returned by this function under such conditions is
488 The UTF8_CHECK_ONLY flag overrides the behavior when a non-allowed (by other
489 flags) malformation is found. If this flag is set, the routine assumes that
490 the caller will raise a warning, and this function will silently just set
491 C<retlen> to C<-1> and return zero.
493 Certain code points are considered problematic. These are Unicode surrogates,
494 Unicode non-characters, and code points above the Unicode maximum of 0x10FFF.
495 By default these are considered regular code points, but certain situations
496 warrant special handling for them. if C<flags> contains
497 UTF8_DISALLOW_ILLEGAL_INTERCHANGE, all three classes are treated as
498 malformations and handled as such. The flags UTF8_DISALLOW_SURROGATE,
499 UTF8_DISALLOW_NONCHAR, and UTF8_DISALLOW_SUPER (meaning above the legal Unicode
500 maximum) can be set to disallow these categories individually.
502 The flags UTF8_WARN_ILLEGAL_INTERCHANGE, UTF8_WARN_SURROGATE,
503 UTF8_WARN_NONCHAR, and UTF8_WARN_SUPER will cause warning messages to be raised
504 for their respective categories, but otherwise the code points are considered
505 valid (not malformations). To get a category to both be treated as a
506 malformation and raise a warning, specify both the WARN and DISALLOW flags.
507 (But note that warnings are not raised if lexically disabled nor if
508 UTF8_CHECK_ONLY is also specified.)
510 Very large code points (above 0x7FFF_FFFF) are considered more problematic than
511 the others that are above the Unicode legal maximum. There are several
512 reasons, one of which is that the original UTF-8 specification never went above
513 this number (the current 0x10FFF limit was imposed later). The UTF-8 encoding
514 on ASCII platforms for these large code point begins with a byte containing
515 0xFE or 0xFF. The UTF8_DISALLOW_FE_FF flag will cause them to be treated as
516 malformations, while allowing smaller above-Unicode code points. (Of course
517 UTF8_DISALLOW_SUPER will treat all above-Unicode code points, including these,
518 as malformations.) Similarly, UTF8_WARN_FE_FF acts just like the other WARN
519 flags, but applies just to these code points.
521 All other code points corresponding to Unicode characters, including private
522 use and those yet to be assigned, are never considered malformed and never
525 Most code should use utf8_to_uvchr() rather than call this directly.
531 Perl_utf8n_to_uvuni(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen, U32 flags)
534 const U8 * const s0 = s;
537 bool dowarn = ckWARN_d(WARN_UTF8);
538 const UV startbyte = *s;
539 STRLEN expectlen = 0;
543 PERL_ARGS_ASSERT_UTF8N_TO_UVUNI;
545 /* This list is a superset of the UTF8_ALLOW_XXX. */
547 #define UTF8_WARN_EMPTY 1
548 #define UTF8_WARN_CONTINUATION 2
549 #define UTF8_WARN_NON_CONTINUATION 3
550 #define UTF8_WARN_SHORT 4
551 #define UTF8_WARN_OVERFLOW 5
552 #define UTF8_WARN_LONG 6
555 !(flags & UTF8_ALLOW_EMPTY)) {
556 warning = UTF8_WARN_EMPTY;
560 if (UTF8_IS_INVARIANT(uv)) {
563 return (UV) (NATIVE_TO_UTF(*s));
566 if (UTF8_IS_CONTINUATION(uv) &&
567 !(flags & UTF8_ALLOW_CONTINUATION)) {
568 warning = UTF8_WARN_CONTINUATION;
572 if (UTF8_IS_START(uv) && curlen > 1 && !UTF8_IS_CONTINUATION(s[1]) &&
573 !(flags & UTF8_ALLOW_NON_CONTINUATION)) {
574 warning = UTF8_WARN_NON_CONTINUATION;
579 uv = NATIVE_TO_UTF(uv);
581 if (uv == 0xfe || uv == 0xff) {
582 if (flags & (UTF8_WARN_SUPER|UTF8_WARN_FE_FF)) {
583 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "Code point beginning with byte 0x%02"UVXf" is not Unicode, and not portable", uv));
584 flags &= ~UTF8_WARN_SUPER; /* Only warn once on this problem */
586 if (flags & (UTF8_DISALLOW_SUPER|UTF8_DISALLOW_FE_FF)) {
592 if (!(uv & 0x20)) { len = 2; uv &= 0x1f; }
593 else if (!(uv & 0x10)) { len = 3; uv &= 0x0f; }
594 else if (!(uv & 0x08)) { len = 4; uv &= 0x07; }
595 else if (!(uv & 0x04)) { len = 5; uv &= 0x03; }
597 else if (!(uv & 0x02)) { len = 6; uv &= 0x01; }
598 else { len = 7; uv &= 0x01; }
600 else if (!(uv & 0x02)) { len = 6; uv &= 0x01; }
601 else if (!(uv & 0x01)) { len = 7; uv = 0; }
602 else { len = 13; uv = 0; } /* whoa! */
610 if ((curlen < expectlen) &&
611 !(flags & UTF8_ALLOW_SHORT)) {
612 warning = UTF8_WARN_SHORT;
618 ouv = uv; /* ouv is the value from the previous iteration */
621 if (!UTF8_IS_CONTINUATION(*s) &&
622 !(flags & UTF8_ALLOW_NON_CONTINUATION)) {
624 warning = UTF8_WARN_NON_CONTINUATION;
628 uv = UTF8_ACCUMULATE(uv, *s);
629 if (!(uv > ouv)) { /* If the value didn't grow from the previous
630 iteration, something is horribly wrong */
631 /* These cannot be allowed. */
633 if (expectlen != 13 && !(flags & UTF8_ALLOW_LONG)) {
634 warning = UTF8_WARN_LONG;
638 else { /* uv < ouv */
639 /* This cannot be allowed. */
640 warning = UTF8_WARN_OVERFLOW;
648 if ((expectlen > (STRLEN)UNISKIP(uv)) && !(flags & UTF8_ALLOW_LONG)) {
649 warning = UTF8_WARN_LONG;
651 } else if (flags & (UTF8_DISALLOW_ILLEGAL_INTERCHANGE|UTF8_WARN_ILLEGAL_INTERCHANGE)) {
652 if (UNICODE_IS_SURROGATE(uv)) {
653 if ((flags & (UTF8_WARN_SURROGATE|UTF8_CHECK_ONLY)) == UTF8_WARN_SURROGATE) {
654 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "UTF-16 surrogate U+%04"UVXf"", uv));
656 if (flags & UTF8_DISALLOW_SURROGATE) {
660 else if (UNICODE_IS_NONCHAR(uv)) {
661 if ((flags & (UTF8_WARN_NONCHAR|UTF8_CHECK_ONLY)) == UTF8_WARN_NONCHAR ) {
662 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "Unicode non-character U+%04"UVXf" is illegal for open interchange", uv));
664 if (flags & UTF8_DISALLOW_NONCHAR) {
668 else if ((uv > PERL_UNICODE_MAX)) {
669 if ((flags & (UTF8_WARN_SUPER|UTF8_CHECK_ONLY)) == UTF8_WARN_SUPER) {
670 sv = sv_2mortal(Perl_newSVpvf(aTHX_ "Code point 0x%04"UVXf" is not Unicode, may not be portable", uv));
672 if (flags & UTF8_DISALLOW_SUPER) {
677 /* Here, this is not considered a malformed character, so drop through
683 disallowed: /* Is disallowed, but otherwise not malformed. 'sv' will have been
684 set if there is to be a warning. */
691 if (flags & UTF8_CHECK_ONLY) {
693 *retlen = ((STRLEN) -1);
699 sv = newSVpvs_flags("Malformed UTF-8 character ", SVs_TEMP);
703 case 0: /* Intentionally empty. */ break;
704 case UTF8_WARN_EMPTY:
705 sv_catpvs(sv, "(empty string)");
707 case UTF8_WARN_CONTINUATION:
708 Perl_sv_catpvf(aTHX_ sv, "(unexpected continuation byte 0x%02"UVxf", with no preceding start byte)", uv);
710 case UTF8_WARN_NON_CONTINUATION:
712 Perl_sv_catpvf(aTHX_ sv, "(unexpected non-continuation byte 0x%02"UVxf", immediately after start byte 0x%02"UVxf")",
713 (UV)s[1], startbyte);
715 const int len = (int)(s-s0);
716 Perl_sv_catpvf(aTHX_ sv, "(unexpected non-continuation byte 0x%02"UVxf", %d byte%s after start byte 0x%02"UVxf", expected %d bytes)",
717 (UV)s[1], len, len > 1 ? "s" : "", startbyte, (int)expectlen);
721 case UTF8_WARN_SHORT:
722 Perl_sv_catpvf(aTHX_ sv, "(%d byte%s, need %d, after start byte 0x%02"UVxf")",
723 (int)curlen, curlen == 1 ? "" : "s", (int)expectlen, startbyte);
724 expectlen = curlen; /* distance for caller to skip */
726 case UTF8_WARN_OVERFLOW:
727 Perl_sv_catpvf(aTHX_ sv, "(overflow at 0x%"UVxf", byte 0x%02x, after start byte 0x%02"UVxf")",
731 Perl_sv_catpvf(aTHX_ sv, "(%d byte%s, need %d, after start byte 0x%02"UVxf")",
732 (int)expectlen, expectlen == 1 ? "": "s", UNISKIP(uv), startbyte);
735 sv_catpvs(sv, "(unknown reason)");
740 const char * const s = SvPVX_const(sv);
743 Perl_warner(aTHX_ packWARN(WARN_UTF8),
744 "%s in %s", s, OP_DESC(PL_op));
746 Perl_warner(aTHX_ packWARN(WARN_UTF8), "%s", s);
751 *retlen = expectlen ? expectlen : len;
757 =for apidoc utf8_to_uvchr
759 Returns the native code point of the first character in the string C<s>
760 which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
761 length, in bytes, of that character.
763 If C<s> does not point to a well-formed UTF-8 character, zero is
764 returned and retlen is set, if possible, to -1.
771 Perl_utf8_to_uvchr(pTHX_ const U8 *s, STRLEN *retlen)
773 PERL_ARGS_ASSERT_UTF8_TO_UVCHR;
775 return utf8n_to_uvchr(s, UTF8_MAXBYTES, retlen,
776 ckWARN_d(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
780 =for apidoc utf8_to_uvuni
782 Returns the Unicode code point of the first character in the string C<s>
783 which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
784 length, in bytes, of that character.
786 This function should only be used when the returned UV is considered
787 an index into the Unicode semantic tables (e.g. swashes).
789 If C<s> does not point to a well-formed UTF-8 character, zero is
790 returned and retlen is set, if possible, to -1.
796 Perl_utf8_to_uvuni(pTHX_ const U8 *s, STRLEN *retlen)
798 PERL_ARGS_ASSERT_UTF8_TO_UVUNI;
800 /* Call the low level routine asking for checks */
801 return Perl_utf8n_to_uvuni(aTHX_ s, UTF8_MAXBYTES, retlen,
802 ckWARN_d(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
806 =for apidoc utf8_length
808 Return the length of the UTF-8 char encoded string C<s> in characters.
809 Stops at C<e> (inclusive). If C<e E<lt> s> or if the scan would end
810 up past C<e>, croaks.
816 Perl_utf8_length(pTHX_ const U8 *s, const U8 *e)
821 PERL_ARGS_ASSERT_UTF8_LENGTH;
823 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g.
824 * the bitops (especially ~) can create illegal UTF-8.
825 * In other words: in Perl UTF-8 is not just for Unicode. */
828 goto warn_and_return;
830 if (!UTF8_IS_INVARIANT(*s))
841 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8),
842 "%s in %s", unees, OP_DESC(PL_op));
844 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), unees);
851 =for apidoc utf8_distance
853 Returns the number of UTF-8 characters between the UTF-8 pointers C<a>
856 WARNING: use only if you *know* that the pointers point inside the
863 Perl_utf8_distance(pTHX_ const U8 *a, const U8 *b)
865 PERL_ARGS_ASSERT_UTF8_DISTANCE;
867 return (a < b) ? -1 * (IV) utf8_length(a, b) : (IV) utf8_length(b, a);
873 Return the UTF-8 pointer C<s> displaced by C<off> characters, either
876 WARNING: do not use the following unless you *know* C<off> is within
877 the UTF-8 data pointed to by C<s> *and* that on entry C<s> is aligned
878 on the first byte of character or just after the last byte of a character.
884 Perl_utf8_hop(pTHX_ const U8 *s, I32 off)
886 PERL_ARGS_ASSERT_UTF8_HOP;
889 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g
890 * the bitops (especially ~) can create illegal UTF-8.
891 * In other words: in Perl UTF-8 is not just for Unicode. */
900 while (UTF8_IS_CONTINUATION(*s))
908 =for apidoc bytes_cmp_utf8
910 Compares the sequence of characters (stored as octets) in b, blen with the
911 sequence of characters (stored as UTF-8) in u, ulen. Returns 0 if they are
912 equal, -1 or -2 if the first string is less than the second string, +1 or +2
913 if the first string is greater than the second string.
915 -1 or +1 is returned if the shorter string was identical to the start of the
916 longer string. -2 or +2 is returned if the was a difference between characters
923 Perl_bytes_cmp_utf8(pTHX_ const U8 *b, STRLEN blen, const U8 *u, STRLEN ulen)
925 const U8 *const bend = b + blen;
926 const U8 *const uend = u + ulen;
928 PERL_ARGS_ASSERT_BYTES_CMP_UTF8;
932 while (b < bend && u < uend) {
934 if (!UTF8_IS_INVARIANT(c)) {
935 if (UTF8_IS_DOWNGRADEABLE_START(c)) {
938 if (UTF8_IS_CONTINUATION(c1)) {
939 c = UNI_TO_NATIVE(TWO_BYTE_UTF8_TO_UNI(c, c1));
941 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8),
942 "Malformed UTF-8 character "
943 "(unexpected non-continuation byte 0x%02x"
944 ", immediately after start byte 0x%02x)"
945 /* Dear diag.t, it's in the pod. */
948 PL_op ? OP_DESC(PL_op) : "");
953 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8),
954 "%s in %s", unees, OP_DESC(PL_op));
956 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), unees);
957 return -2; /* Really want to return undef :-) */
964 return *b < c ? -2 : +2;
969 if (b == bend && u == uend)
972 return b < bend ? +1 : -1;
976 =for apidoc utf8_to_bytes
978 Converts a string C<s> of length C<len> from UTF-8 into native byte encoding.
979 Unlike C<bytes_to_utf8>, this over-writes the original string, and
980 updates len to contain the new length.
981 Returns zero on failure, setting C<len> to -1.
983 If you need a copy of the string, see C<bytes_from_utf8>.
989 Perl_utf8_to_bytes(pTHX_ U8 *s, STRLEN *len)
992 U8 * const send = s + *len;
995 PERL_ARGS_ASSERT_UTF8_TO_BYTES;
997 /* ensure valid UTF-8 and chars < 256 before updating string */
1001 if (!UTF8_IS_INVARIANT(c) &&
1002 (!UTF8_IS_DOWNGRADEABLE_START(c) || (s >= send)
1003 || !(c = *s++) || !UTF8_IS_CONTINUATION(c))) {
1004 *len = ((STRLEN) -1);
1012 *d++ = (U8)utf8_to_uvchr(s, &ulen);
1021 =for apidoc bytes_from_utf8
1023 Converts a string C<s> of length C<len> from UTF-8 into native byte encoding.
1024 Unlike C<utf8_to_bytes> but like C<bytes_to_utf8>, returns a pointer to
1025 the newly-created string, and updates C<len> to contain the new
1026 length. Returns the original string if no conversion occurs, C<len>
1027 is unchanged. Do nothing if C<is_utf8> points to 0. Sets C<is_utf8> to
1028 0 if C<s> is converted or consisted entirely of characters that are invariant
1029 in utf8 (i.e., US-ASCII on non-EBCDIC machines).
1035 Perl_bytes_from_utf8(pTHX_ const U8 *s, STRLEN *len, bool *is_utf8)
1038 const U8 *start = s;
1042 PERL_ARGS_ASSERT_BYTES_FROM_UTF8;
1044 PERL_UNUSED_CONTEXT;
1048 /* ensure valid UTF-8 and chars < 256 before converting string */
1049 for (send = s + *len; s < send;) {
1051 if (!UTF8_IS_INVARIANT(c)) {
1052 if (UTF8_IS_DOWNGRADEABLE_START(c) && s < send &&
1053 (c = *s++) && UTF8_IS_CONTINUATION(c))
1062 Newx(d, (*len) - count + 1, U8);
1063 s = start; start = d;
1066 if (!UTF8_IS_INVARIANT(c)) {
1067 /* Then it is two-byte encoded */
1068 c = UNI_TO_NATIVE(TWO_BYTE_UTF8_TO_UNI(c, *s++));
1078 =for apidoc bytes_to_utf8
1080 Converts a string C<s> of length C<len> bytes from the native encoding into
1082 Returns a pointer to the newly-created string, and sets C<len> to
1083 reflect the new length in bytes.
1085 A NUL character will be written after the end of the string.
1087 If you want to convert to UTF-8 from encodings other than
1088 the native (Latin1 or EBCDIC),
1089 see sv_recode_to_utf8().
1095 Perl_bytes_to_utf8(pTHX_ const U8 *s, STRLEN *len)
1097 const U8 * const send = s + (*len);
1101 PERL_ARGS_ASSERT_BYTES_TO_UTF8;
1102 PERL_UNUSED_CONTEXT;
1104 Newx(d, (*len) * 2 + 1, U8);
1108 const UV uv = NATIVE_TO_ASCII(*s++);
1109 if (UNI_IS_INVARIANT(uv))
1110 *d++ = (U8)UTF_TO_NATIVE(uv);
1112 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
1113 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
1122 * Convert native (big-endian) or reversed (little-endian) UTF-16 to UTF-8.
1124 * Destination must be pre-extended to 3/2 source. Do not use in-place.
1125 * We optimize for native, for obvious reasons. */
1128 Perl_utf16_to_utf8(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen)
1133 PERL_ARGS_ASSERT_UTF16_TO_UTF8;
1136 Perl_croak(aTHX_ "panic: utf16_to_utf8: odd bytelen %"UVuf, (UV)bytelen);
1141 UV uv = (p[0] << 8) + p[1]; /* UTF-16BE */
1145 *d++ = UNI_TO_NATIVE(uv);
1152 *d++ = (U8)(( uv >> 6) | 0xc0);
1153 *d++ = (U8)(( uv & 0x3f) | 0x80);
1156 if (uv >= 0xd800 && uv <= 0xdbff) { /* surrogates */
1158 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
1160 UV low = (p[0] << 8) + p[1];
1162 if (low < 0xdc00 || low > 0xdfff)
1163 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
1164 uv = ((uv - 0xd800) << 10) + (low - 0xdc00) + 0x10000;
1166 } else if (uv >= 0xdc00 && uv <= 0xdfff) {
1167 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
1170 *d++ = (U8)(( uv >> 12) | 0xe0);
1171 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
1172 *d++ = (U8)(( uv & 0x3f) | 0x80);
1176 *d++ = (U8)(( uv >> 18) | 0xf0);
1177 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
1178 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
1179 *d++ = (U8)(( uv & 0x3f) | 0x80);
1183 *newlen = d - dstart;
1187 /* Note: this one is slightly destructive of the source. */
1190 Perl_utf16_to_utf8_reversed(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen)
1193 U8* const send = s + bytelen;
1195 PERL_ARGS_ASSERT_UTF16_TO_UTF8_REVERSED;
1198 Perl_croak(aTHX_ "panic: utf16_to_utf8_reversed: odd bytelen %"UVuf,
1202 const U8 tmp = s[0];
1207 return utf16_to_utf8(p, d, bytelen, newlen);
1210 /* for now these are all defined (inefficiently) in terms of the utf8 versions */
1213 Perl_is_uni_alnum(pTHX_ UV c)
1215 U8 tmpbuf[UTF8_MAXBYTES+1];
1216 uvchr_to_utf8(tmpbuf, c);
1217 return is_utf8_alnum(tmpbuf);
1221 Perl_is_uni_idfirst(pTHX_ UV c)
1223 U8 tmpbuf[UTF8_MAXBYTES+1];
1224 uvchr_to_utf8(tmpbuf, c);
1225 return is_utf8_idfirst(tmpbuf);
1229 Perl_is_uni_alpha(pTHX_ UV c)
1231 U8 tmpbuf[UTF8_MAXBYTES+1];
1232 uvchr_to_utf8(tmpbuf, c);
1233 return is_utf8_alpha(tmpbuf);
1237 Perl_is_uni_ascii(pTHX_ UV c)
1239 U8 tmpbuf[UTF8_MAXBYTES+1];
1240 uvchr_to_utf8(tmpbuf, c);
1241 return is_utf8_ascii(tmpbuf);
1245 Perl_is_uni_space(pTHX_ UV c)
1247 U8 tmpbuf[UTF8_MAXBYTES+1];
1248 uvchr_to_utf8(tmpbuf, c);
1249 return is_utf8_space(tmpbuf);
1253 Perl_is_uni_digit(pTHX_ UV c)
1255 U8 tmpbuf[UTF8_MAXBYTES+1];
1256 uvchr_to_utf8(tmpbuf, c);
1257 return is_utf8_digit(tmpbuf);
1261 Perl_is_uni_upper(pTHX_ UV c)
1263 U8 tmpbuf[UTF8_MAXBYTES+1];
1264 uvchr_to_utf8(tmpbuf, c);
1265 return is_utf8_upper(tmpbuf);
1269 Perl_is_uni_lower(pTHX_ UV c)
1271 U8 tmpbuf[UTF8_MAXBYTES+1];
1272 uvchr_to_utf8(tmpbuf, c);
1273 return is_utf8_lower(tmpbuf);
1277 Perl_is_uni_cntrl(pTHX_ UV c)
1279 U8 tmpbuf[UTF8_MAXBYTES+1];
1280 uvchr_to_utf8(tmpbuf, c);
1281 return is_utf8_cntrl(tmpbuf);
1285 Perl_is_uni_graph(pTHX_ UV c)
1287 U8 tmpbuf[UTF8_MAXBYTES+1];
1288 uvchr_to_utf8(tmpbuf, c);
1289 return is_utf8_graph(tmpbuf);
1293 Perl_is_uni_print(pTHX_ UV c)
1295 U8 tmpbuf[UTF8_MAXBYTES+1];
1296 uvchr_to_utf8(tmpbuf, c);
1297 return is_utf8_print(tmpbuf);
1301 Perl_is_uni_punct(pTHX_ UV c)
1303 U8 tmpbuf[UTF8_MAXBYTES+1];
1304 uvchr_to_utf8(tmpbuf, c);
1305 return is_utf8_punct(tmpbuf);
1309 Perl_is_uni_xdigit(pTHX_ UV c)
1311 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1312 uvchr_to_utf8(tmpbuf, c);
1313 return is_utf8_xdigit(tmpbuf);
1317 Perl_to_uni_upper(pTHX_ UV c, U8* p, STRLEN *lenp)
1319 PERL_ARGS_ASSERT_TO_UNI_UPPER;
1321 uvchr_to_utf8(p, c);
1322 return to_utf8_upper(p, p, lenp);
1326 Perl_to_uni_title(pTHX_ UV c, U8* p, STRLEN *lenp)
1328 PERL_ARGS_ASSERT_TO_UNI_TITLE;
1330 uvchr_to_utf8(p, c);
1331 return to_utf8_title(p, p, lenp);
1335 Perl_to_uni_lower(pTHX_ UV c, U8* p, STRLEN *lenp)
1337 PERL_ARGS_ASSERT_TO_UNI_LOWER;
1339 uvchr_to_utf8(p, c);
1340 return to_utf8_lower(p, p, lenp);
1344 Perl_to_uni_fold(pTHX_ UV c, U8* p, STRLEN *lenp)
1346 PERL_ARGS_ASSERT_TO_UNI_FOLD;
1348 uvchr_to_utf8(p, c);
1349 return to_utf8_fold(p, p, lenp);
1352 /* for now these all assume no locale info available for Unicode > 255 */
1355 Perl_is_uni_alnum_lc(pTHX_ UV c)
1357 return is_uni_alnum(c); /* XXX no locale support yet */
1361 Perl_is_uni_idfirst_lc(pTHX_ UV c)
1363 return is_uni_idfirst(c); /* XXX no locale support yet */
1367 Perl_is_uni_alpha_lc(pTHX_ UV c)
1369 return is_uni_alpha(c); /* XXX no locale support yet */
1373 Perl_is_uni_ascii_lc(pTHX_ UV c)
1375 return is_uni_ascii(c); /* XXX no locale support yet */
1379 Perl_is_uni_space_lc(pTHX_ UV c)
1381 return is_uni_space(c); /* XXX no locale support yet */
1385 Perl_is_uni_digit_lc(pTHX_ UV c)
1387 return is_uni_digit(c); /* XXX no locale support yet */
1391 Perl_is_uni_upper_lc(pTHX_ UV c)
1393 return is_uni_upper(c); /* XXX no locale support yet */
1397 Perl_is_uni_lower_lc(pTHX_ UV c)
1399 return is_uni_lower(c); /* XXX no locale support yet */
1403 Perl_is_uni_cntrl_lc(pTHX_ UV c)
1405 return is_uni_cntrl(c); /* XXX no locale support yet */
1409 Perl_is_uni_graph_lc(pTHX_ UV c)
1411 return is_uni_graph(c); /* XXX no locale support yet */
1415 Perl_is_uni_print_lc(pTHX_ UV c)
1417 return is_uni_print(c); /* XXX no locale support yet */
1421 Perl_is_uni_punct_lc(pTHX_ UV c)
1423 return is_uni_punct(c); /* XXX no locale support yet */
1427 Perl_is_uni_xdigit_lc(pTHX_ UV c)
1429 return is_uni_xdigit(c); /* XXX no locale support yet */
1433 Perl_to_uni_upper_lc(pTHX_ U32 c)
1435 /* XXX returns only the first character -- do not use XXX */
1436 /* XXX no locale support yet */
1438 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1439 return (U32)to_uni_upper(c, tmpbuf, &len);
1443 Perl_to_uni_title_lc(pTHX_ U32 c)
1445 /* XXX returns only the first character XXX -- do not use XXX */
1446 /* XXX no locale support yet */
1448 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1449 return (U32)to_uni_title(c, tmpbuf, &len);
1453 Perl_to_uni_lower_lc(pTHX_ U32 c)
1455 /* XXX returns only the first character -- do not use XXX */
1456 /* XXX no locale support yet */
1458 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1459 return (U32)to_uni_lower(c, tmpbuf, &len);
1463 S_is_utf8_common(pTHX_ const U8 *const p, SV **swash,
1464 const char *const swashname)
1468 PERL_ARGS_ASSERT_IS_UTF8_COMMON;
1470 if (!is_utf8_char(p))
1473 *swash = swash_init("utf8", swashname, &PL_sv_undef, 1, 0);
1474 return swash_fetch(*swash, p, TRUE) != 0;
1478 Perl_is_utf8_alnum(pTHX_ const U8 *p)
1482 PERL_ARGS_ASSERT_IS_UTF8_ALNUM;
1484 /* NOTE: "IsWord", not "IsAlnum", since Alnum is a true
1485 * descendant of isalnum(3), in other words, it doesn't
1486 * contain the '_'. --jhi */
1487 return is_utf8_common(p, &PL_utf8_alnum, "IsWord");
1491 Perl_is_utf8_idfirst(pTHX_ const U8 *p) /* The naming is historical. */
1495 PERL_ARGS_ASSERT_IS_UTF8_IDFIRST;
1499 /* is_utf8_idstart would be more logical. */
1500 return is_utf8_common(p, &PL_utf8_idstart, "IdStart");
1504 Perl_is_utf8_xidfirst(pTHX_ const U8 *p) /* The naming is historical. */
1508 PERL_ARGS_ASSERT_IS_UTF8_XIDFIRST;
1512 /* is_utf8_idstart would be more logical. */
1513 return is_utf8_common(p, &PL_utf8_xidstart, "XIdStart");
1517 Perl_is_utf8_idcont(pTHX_ const U8 *p)
1521 PERL_ARGS_ASSERT_IS_UTF8_IDCONT;
1525 return is_utf8_common(p, &PL_utf8_idcont, "IdContinue");
1529 Perl_is_utf8_xidcont(pTHX_ const U8 *p)
1533 PERL_ARGS_ASSERT_IS_UTF8_XIDCONT;
1537 return is_utf8_common(p, &PL_utf8_idcont, "XIdContinue");
1541 Perl_is_utf8_alpha(pTHX_ const U8 *p)
1545 PERL_ARGS_ASSERT_IS_UTF8_ALPHA;
1547 return is_utf8_common(p, &PL_utf8_alpha, "IsAlpha");
1551 Perl_is_utf8_ascii(pTHX_ const U8 *p)
1555 PERL_ARGS_ASSERT_IS_UTF8_ASCII;
1557 return is_utf8_common(p, &PL_utf8_ascii, "IsAscii");
1561 Perl_is_utf8_space(pTHX_ const U8 *p)
1565 PERL_ARGS_ASSERT_IS_UTF8_SPACE;
1567 return is_utf8_common(p, &PL_utf8_space, "IsSpacePerl");
1571 Perl_is_utf8_perl_space(pTHX_ const U8 *p)
1575 PERL_ARGS_ASSERT_IS_UTF8_PERL_SPACE;
1577 return is_utf8_common(p, &PL_utf8_perl_space, "IsPerlSpace");
1581 Perl_is_utf8_perl_word(pTHX_ const U8 *p)
1585 PERL_ARGS_ASSERT_IS_UTF8_PERL_WORD;
1587 return is_utf8_common(p, &PL_utf8_perl_word, "IsPerlWord");
1591 Perl_is_utf8_digit(pTHX_ const U8 *p)
1595 PERL_ARGS_ASSERT_IS_UTF8_DIGIT;
1597 return is_utf8_common(p, &PL_utf8_digit, "IsDigit");
1601 Perl_is_utf8_posix_digit(pTHX_ const U8 *p)
1605 PERL_ARGS_ASSERT_IS_UTF8_POSIX_DIGIT;
1607 return is_utf8_common(p, &PL_utf8_posix_digit, "IsPosixDigit");
1611 Perl_is_utf8_upper(pTHX_ const U8 *p)
1615 PERL_ARGS_ASSERT_IS_UTF8_UPPER;
1617 return is_utf8_common(p, &PL_utf8_upper, "IsUppercase");
1621 Perl_is_utf8_lower(pTHX_ const U8 *p)
1625 PERL_ARGS_ASSERT_IS_UTF8_LOWER;
1627 return is_utf8_common(p, &PL_utf8_lower, "IsLowercase");
1631 Perl_is_utf8_cntrl(pTHX_ const U8 *p)
1635 PERL_ARGS_ASSERT_IS_UTF8_CNTRL;
1637 return is_utf8_common(p, &PL_utf8_cntrl, "IsCntrl");
1641 Perl_is_utf8_graph(pTHX_ const U8 *p)
1645 PERL_ARGS_ASSERT_IS_UTF8_GRAPH;
1647 return is_utf8_common(p, &PL_utf8_graph, "IsGraph");
1651 Perl_is_utf8_print(pTHX_ const U8 *p)
1655 PERL_ARGS_ASSERT_IS_UTF8_PRINT;
1657 return is_utf8_common(p, &PL_utf8_print, "IsPrint");
1661 Perl_is_utf8_punct(pTHX_ const U8 *p)
1665 PERL_ARGS_ASSERT_IS_UTF8_PUNCT;
1667 return is_utf8_common(p, &PL_utf8_punct, "IsPunct");
1671 Perl_is_utf8_xdigit(pTHX_ const U8 *p)
1675 PERL_ARGS_ASSERT_IS_UTF8_XDIGIT;
1677 return is_utf8_common(p, &PL_utf8_xdigit, "IsXDigit");
1681 Perl_is_utf8_mark(pTHX_ const U8 *p)
1685 PERL_ARGS_ASSERT_IS_UTF8_MARK;
1687 return is_utf8_common(p, &PL_utf8_mark, "IsM");
1691 Perl_is_utf8_X_begin(pTHX_ const U8 *p)
1695 PERL_ARGS_ASSERT_IS_UTF8_X_BEGIN;
1697 return is_utf8_common(p, &PL_utf8_X_begin, "_X_Begin");
1701 Perl_is_utf8_X_extend(pTHX_ const U8 *p)
1705 PERL_ARGS_ASSERT_IS_UTF8_X_EXTEND;
1707 return is_utf8_common(p, &PL_utf8_X_extend, "_X_Extend");
1711 Perl_is_utf8_X_prepend(pTHX_ const U8 *p)
1715 PERL_ARGS_ASSERT_IS_UTF8_X_PREPEND;
1717 return is_utf8_common(p, &PL_utf8_X_prepend, "GCB=Prepend");
1721 Perl_is_utf8_X_non_hangul(pTHX_ const U8 *p)
1725 PERL_ARGS_ASSERT_IS_UTF8_X_NON_HANGUL;
1727 return is_utf8_common(p, &PL_utf8_X_non_hangul, "HST=Not_Applicable");
1731 Perl_is_utf8_X_L(pTHX_ const U8 *p)
1735 PERL_ARGS_ASSERT_IS_UTF8_X_L;
1737 return is_utf8_common(p, &PL_utf8_X_L, "GCB=L");
1741 Perl_is_utf8_X_LV(pTHX_ const U8 *p)
1745 PERL_ARGS_ASSERT_IS_UTF8_X_LV;
1747 return is_utf8_common(p, &PL_utf8_X_LV, "GCB=LV");
1751 Perl_is_utf8_X_LVT(pTHX_ const U8 *p)
1755 PERL_ARGS_ASSERT_IS_UTF8_X_LVT;
1757 return is_utf8_common(p, &PL_utf8_X_LVT, "GCB=LVT");
1761 Perl_is_utf8_X_T(pTHX_ const U8 *p)
1765 PERL_ARGS_ASSERT_IS_UTF8_X_T;
1767 return is_utf8_common(p, &PL_utf8_X_T, "GCB=T");
1771 Perl_is_utf8_X_V(pTHX_ const U8 *p)
1775 PERL_ARGS_ASSERT_IS_UTF8_X_V;
1777 return is_utf8_common(p, &PL_utf8_X_V, "GCB=V");
1781 Perl_is_utf8_X_LV_LVT_V(pTHX_ const U8 *p)
1785 PERL_ARGS_ASSERT_IS_UTF8_X_LV_LVT_V;
1787 return is_utf8_common(p, &PL_utf8_X_LV_LVT_V, "_X_LV_LVT_V");
1791 =for apidoc to_utf8_case
1793 The "p" contains the pointer to the UTF-8 string encoding
1794 the character that is being converted.
1796 The "ustrp" is a pointer to the character buffer to put the
1797 conversion result to. The "lenp" is a pointer to the length
1800 The "swashp" is a pointer to the swash to use.
1802 Both the special and normal mappings are stored lib/unicore/To/Foo.pl,
1803 and loaded by SWASHNEW, using lib/utf8_heavy.pl. The special (usually,
1804 but not always, a multicharacter mapping), is tried first.
1806 The "special" is a string like "utf8::ToSpecLower", which means the
1807 hash %utf8::ToSpecLower. The access to the hash is through
1808 Perl_to_utf8_case().
1810 The "normal" is a string like "ToLower" which means the swash
1816 Perl_to_utf8_case(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp,
1817 SV **swashp, const char *normal, const char *special)
1820 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1822 const UV uv0 = utf8_to_uvchr(p, NULL);
1823 /* The NATIVE_TO_UNI() and UNI_TO_NATIVE() mappings
1824 * are necessary in EBCDIC, they are redundant no-ops
1825 * in ASCII-ish platforms, and hopefully optimized away. */
1826 const UV uv1 = NATIVE_TO_UNI(uv0);
1828 PERL_ARGS_ASSERT_TO_UTF8_CASE;
1830 /* Note that swash_fetch() doesn't output warnings for these because it
1831 * assumes we will */
1832 if (uv1 >= UNICODE_SURROGATE_FIRST) {
1833 if (uv1 <= UNICODE_SURROGATE_LAST) {
1834 if (ckWARN_d(WARN_SURROGATE)) {
1835 const char* desc = (PL_op) ? OP_DESC(PL_op) : normal;
1836 Perl_warner(aTHX_ packWARN(WARN_SURROGATE),
1837 "Operation \"%s\" returns its argument for UTF-16 surrogate U+%04"UVXf"", desc, uv1);
1840 else if (UNICODE_IS_SUPER(uv1)) {
1841 if (ckWARN_d(WARN_NON_UNICODE)) {
1842 const char* desc = (PL_op) ? OP_DESC(PL_op) : normal;
1843 Perl_warner(aTHX_ packWARN(WARN_NON_UNICODE),
1844 "Operation \"%s\" returns its argument for non-Unicode code point 0x%04"UVXf"", desc, uv1);
1848 /* Note that non-characters are perfectly legal, so no warning should
1852 uvuni_to_utf8(tmpbuf, uv1);
1854 if (!*swashp) /* load on-demand */
1855 *swashp = swash_init("utf8", normal, &PL_sv_undef, 4, 0);
1856 /* This is the beginnings of a skeleton of code to read the info section
1857 * that is in all the swashes in case we ever want to do that, so one can
1858 * read things whose maps aren't code points, and whose default if missing
1859 * is not to the code point itself. This was just to see if it actually
1860 * worked. Details on what the possibilities are are in perluniprops.pod
1861 HV * const hv = get_hv("utf8::SwashInfo", 0);
1864 svp = hv_fetch(hv, (const char*)normal, strlen(normal), FALSE);
1867 HV * const this_hash = SvRV(*svp);
1868 svp = hv_fetch(this_hash, "type", strlen("type"), FALSE);
1869 s = SvPV_const(*svp, len);
1874 /* It might be "special" (sometimes, but not always,
1875 * a multicharacter mapping) */
1876 HV * const hv = get_hv(special, 0);
1880 (svp = hv_fetch(hv, (const char*)tmpbuf, UNISKIP(uv1), FALSE)) &&
1884 s = SvPV_const(*svp, len);
1886 len = uvuni_to_utf8(ustrp, NATIVE_TO_UNI(*(U8*)s)) - ustrp;
1889 /* If we have EBCDIC we need to remap the characters
1890 * since any characters in the low 256 are Unicode
1891 * code points, not EBCDIC. */
1892 U8 *t = (U8*)s, *tend = t + len, *d;
1899 const UV c = utf8_to_uvchr(t, &tlen);
1901 d = uvchr_to_utf8(d, UNI_TO_NATIVE(c));
1910 d = uvchr_to_utf8(d, UNI_TO_NATIVE(*t));
1915 Copy(tmpbuf, ustrp, len, U8);
1917 Copy(s, ustrp, len, U8);
1923 if (!len && *swashp) {
1924 const UV uv2 = swash_fetch(*swashp, tmpbuf, TRUE);
1927 /* It was "normal" (a single character mapping). */
1928 const UV uv3 = UNI_TO_NATIVE(uv2);
1929 len = uvchr_to_utf8(ustrp, uv3) - ustrp;
1933 if (!len) /* Neither: just copy. In other words, there was no mapping
1934 defined, which means that the code point maps to itself */
1935 len = uvchr_to_utf8(ustrp, uv0) - ustrp;
1940 return len ? utf8_to_uvchr(ustrp, 0) : 0;
1944 =for apidoc to_utf8_upper
1946 Convert the UTF-8 encoded character at p to its uppercase version and
1947 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
1948 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since
1949 the uppercase version may be longer than the original character.
1951 The first character of the uppercased version is returned
1952 (but note, as explained above, that there may be more.)
1957 Perl_to_utf8_upper(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
1961 PERL_ARGS_ASSERT_TO_UTF8_UPPER;
1963 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
1964 &PL_utf8_toupper, "ToUpper", "utf8::ToSpecUpper");
1968 =for apidoc to_utf8_title
1970 Convert the UTF-8 encoded character at p to its titlecase version and
1971 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
1972 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
1973 titlecase version may be longer than the original character.
1975 The first character of the titlecased version is returned
1976 (but note, as explained above, that there may be more.)
1981 Perl_to_utf8_title(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
1985 PERL_ARGS_ASSERT_TO_UTF8_TITLE;
1987 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
1988 &PL_utf8_totitle, "ToTitle", "utf8::ToSpecTitle");
1992 =for apidoc to_utf8_lower
1994 Convert the UTF-8 encoded character at p to its lowercase version and
1995 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
1996 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
1997 lowercase version may be longer than the original character.
1999 The first character of the lowercased version is returned
2000 (but note, as explained above, that there may be more.)
2005 Perl_to_utf8_lower(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
2009 PERL_ARGS_ASSERT_TO_UTF8_LOWER;
2011 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
2012 &PL_utf8_tolower, "ToLower", "utf8::ToSpecLower");
2016 =for apidoc to_utf8_fold
2018 Convert the UTF-8 encoded character at p to its foldcase version and
2019 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
2020 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
2021 foldcase version may be longer than the original character (up to
2024 The first character of the foldcased version is returned
2025 (but note, as explained above, that there may be more.)
2030 Perl_to_utf8_fold(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
2034 PERL_ARGS_ASSERT_TO_UTF8_FOLD;
2036 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
2037 &PL_utf8_tofold, "ToFold", "utf8::ToSpecFold");
2041 * A "swash" is a swatch hash.
2042 * A "swatch" is a bit vector generated by utf8.c:S_swash_get().
2043 * C<pkg> is a pointer to a package name for SWASHNEW, should be "utf8".
2044 * For other parameters, see utf8::SWASHNEW in lib/utf8_heavy.pl.
2047 Perl_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv, I32 minbits, I32 none)
2052 const size_t pkg_len = strlen(pkg);
2053 const size_t name_len = strlen(name);
2054 HV * const stash = gv_stashpvn(pkg, pkg_len, 0);
2058 PERL_ARGS_ASSERT_SWASH_INIT;
2060 PUSHSTACKi(PERLSI_MAGIC);
2064 method = gv_fetchmeth(stash, "SWASHNEW", 8, -1);
2065 if (!method) { /* demand load utf8 */
2067 errsv_save = newSVsv(ERRSV);
2068 /* It is assumed that callers of this routine are not passing in any
2069 user derived data. */
2070 /* Need to do this after save_re_context() as it will set PL_tainted to
2071 1 while saving $1 etc (see the code after getrx: in Perl_magic_get).
2072 Even line to create errsv_save can turn on PL_tainted. */
2073 SAVEBOOL(PL_tainted);
2075 Perl_load_module(aTHX_ PERL_LOADMOD_NOIMPORT, newSVpvn(pkg,pkg_len),
2078 sv_setsv(ERRSV, errsv_save);
2079 SvREFCNT_dec(errsv_save);
2085 mPUSHp(pkg, pkg_len);
2086 mPUSHp(name, name_len);
2091 errsv_save = newSVsv(ERRSV);
2092 /* If we already have a pointer to the method, no need to use call_method()
2093 to repeat the lookup. */
2094 if (method ? call_sv(MUTABLE_SV(method), G_SCALAR)
2095 : call_sv(newSVpvs_flags("SWASHNEW", SVs_TEMP), G_SCALAR | G_METHOD))
2096 retval = newSVsv(*PL_stack_sp--);
2098 retval = &PL_sv_undef;
2100 sv_setsv(ERRSV, errsv_save);
2101 SvREFCNT_dec(errsv_save);
2104 if (IN_PERL_COMPILETIME) {
2105 CopHINTS_set(PL_curcop, PL_hints);
2107 if (!SvROK(retval) || SvTYPE(SvRV(retval)) != SVt_PVHV) {
2109 Perl_croak(aTHX_ "Can't find Unicode property definition \"%"SVf"\"",
2111 Perl_croak(aTHX_ "SWASHNEW didn't return an HV ref");
2117 /* This API is wrong for special case conversions since we may need to
2118 * return several Unicode characters for a single Unicode character
2119 * (see lib/unicore/SpecCase.txt) The SWASHGET in lib/utf8_heavy.pl is
2120 * the lower-level routine, and it is similarly broken for returning
2121 * multiple values. --jhi
2122 * For those, you should use to_utf8_case() instead */
2123 /* Now SWASHGET is recasted into S_swash_get in this file. */
2126 * Returns the value of property/mapping C<swash> for the first character
2127 * of the string C<ptr>. If C<do_utf8> is true, the string C<ptr> is
2128 * assumed to be in utf8. If C<do_utf8> is false, the string C<ptr> is
2129 * assumed to be in native 8-bit encoding. Caches the swatch in C<swash>.
2132 Perl_swash_fetch(pTHX_ SV *swash, const U8 *ptr, bool do_utf8)
2135 HV *const hv = MUTABLE_HV(SvRV(swash));
2140 const U8 *tmps = NULL;
2144 const UV c = NATIVE_TO_ASCII(*ptr);
2146 PERL_ARGS_ASSERT_SWASH_FETCH;
2148 if (!do_utf8 && !UNI_IS_INVARIANT(c)) {
2149 tmputf8[0] = (U8)UTF8_EIGHT_BIT_HI(c);
2150 tmputf8[1] = (U8)UTF8_EIGHT_BIT_LO(c);
2153 /* Given a UTF-X encoded char 0xAA..0xYY,0xZZ
2154 * then the "swatch" is a vec() for all the chars which start
2156 * So the key in the hash (klen) is length of encoded char -1
2158 klen = UTF8SKIP(ptr) - 1;
2162 /* If char is invariant then swatch is for all the invariant chars
2163 * In both UTF-8 and UTF-8-MOD that happens to be UTF_CONTINUATION_MARK
2165 needents = UTF_CONTINUATION_MARK;
2166 off = NATIVE_TO_UTF(ptr[klen]);
2169 /* If char is encoded then swatch is for the prefix */
2170 needents = (1 << UTF_ACCUMULATION_SHIFT);
2171 off = NATIVE_TO_UTF(ptr[klen]) & UTF_CONTINUATION_MASK;
2172 if (UTF8_IS_SUPER(ptr) && ckWARN_d(WARN_NON_UNICODE)) {
2173 const UV code_point = utf8n_to_uvuni(ptr, UTF8_MAXBYTES, 0, 0);
2175 /* This outputs warnings for binary properties only, assuming that
2176 * to_utf8_case() will output any. Also, surrogates aren't checked
2177 * for, as that would warn on things like /\p{Gc=Cs}/ */
2178 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
2179 if (SvUV(*bitssvp) == 1) {
2180 Perl_warner(aTHX_ packWARN(WARN_NON_UNICODE),
2181 "Code point 0x%04"UVXf" is not Unicode, no properties match it; all inverse properties do", code_point);
2187 * This single-entry cache saves about 1/3 of the utf8 overhead in test
2188 * suite. (That is, only 7-8% overall over just a hash cache. Still,
2189 * it's nothing to sniff at.) Pity we usually come through at least
2190 * two function calls to get here...
2192 * NB: this code assumes that swatches are never modified, once generated!
2195 if (hv == PL_last_swash_hv &&
2196 klen == PL_last_swash_klen &&
2197 (!klen || memEQ((char *)ptr, (char *)PL_last_swash_key, klen)) )
2199 tmps = PL_last_swash_tmps;
2200 slen = PL_last_swash_slen;
2203 /* Try our second-level swatch cache, kept in a hash. */
2204 SV** svp = hv_fetch(hv, (const char*)ptr, klen, FALSE);
2206 /* If not cached, generate it via swash_get */
2207 if (!svp || !SvPOK(*svp)
2208 || !(tmps = (const U8*)SvPV_const(*svp, slen))) {
2209 /* We use utf8n_to_uvuni() as we want an index into
2210 Unicode tables, not a native character number.
2212 const UV code_point = utf8n_to_uvuni(ptr, UTF8_MAXBYTES, 0,
2214 0 : UTF8_ALLOW_ANY);
2215 swatch = swash_get(swash,
2216 /* On EBCDIC & ~(0xA0-1) isn't a useful thing to do */
2217 (klen) ? (code_point & ~(needents - 1)) : 0,
2220 if (IN_PERL_COMPILETIME)
2221 CopHINTS_set(PL_curcop, PL_hints);
2223 svp = hv_store(hv, (const char *)ptr, klen, swatch, 0);
2225 if (!svp || !(tmps = (U8*)SvPV(*svp, slen))
2226 || (slen << 3) < needents)
2227 Perl_croak(aTHX_ "panic: swash_fetch got improper swatch");
2230 PL_last_swash_hv = hv;
2231 assert(klen <= sizeof(PL_last_swash_key));
2232 PL_last_swash_klen = (U8)klen;
2233 /* FIXME change interpvar.h? */
2234 PL_last_swash_tmps = (U8 *) tmps;
2235 PL_last_swash_slen = slen;
2237 Copy(ptr, PL_last_swash_key, klen, U8);
2240 switch ((int)((slen << 3) / needents)) {
2242 bit = 1 << (off & 7);
2244 return (tmps[off] & bit) != 0;
2249 return (tmps[off] << 8) + tmps[off + 1] ;
2252 return (tmps[off] << 24) + (tmps[off+1] << 16) + (tmps[off+2] << 8) + tmps[off + 3] ;
2254 Perl_croak(aTHX_ "panic: swash_fetch got swatch of unexpected bit width");
2255 NORETURN_FUNCTION_END;
2258 /* Read a single line of the main body of the swash input text. These are of
2261 * where each number is hex. The first two numbers form the minimum and
2262 * maximum of a range, and the third is the value associated with the range.
2263 * Not all swashes should have a third number
2265 * On input: l points to the beginning of the line to be examined; it points
2266 * to somewhere in the string of the whole input text, and is
2267 * terminated by a \n or the null string terminator.
2268 * lend points to the null terminator of that string
2269 * wants_value is non-zero if the swash expects a third number
2270 * typestr is the name of the swash's mapping, like 'ToLower'
2271 * On output: *min, *max, and *val are set to the values read from the line.
2272 * returns a pointer just beyond the line examined. If there was no
2273 * valid min number on the line, returns lend+1
2277 S_swash_scan_list_line(pTHX_ U8* l, U8* const lend, UV* min, UV* max, UV* val,
2278 const bool wants_value, const U8* const typestr)
2280 const int typeto = typestr[0] == 'T' && typestr[1] == 'o';
2281 STRLEN numlen; /* Length of the number */
2282 I32 flags = PERL_SCAN_SILENT_ILLDIGIT | PERL_SCAN_DISALLOW_PREFIX;
2284 /* nl points to the next \n in the scan */
2285 U8* const nl = (U8*)memchr(l, '\n', lend - l);
2287 /* Get the first number on the line: the range minimum */
2289 *min = grok_hex((char *)l, &numlen, &flags, NULL);
2290 if (numlen) /* If found a hex number, position past it */
2292 else if (nl) { /* Else, go handle next line, if any */
2293 return nl + 1; /* 1 is length of "\n" */
2295 else { /* Else, no next line */
2296 return lend + 1; /* to LIST's end at which \n is not found */
2299 /* The max range value follows, separated by a BLANK */
2302 flags = PERL_SCAN_SILENT_ILLDIGIT | PERL_SCAN_DISALLOW_PREFIX;
2304 *max = grok_hex((char *)l, &numlen, &flags, NULL);
2307 else /* If no value here, it is a single element range */
2310 /* Non-binary tables have a third entry: what the first element of the
2315 flags = PERL_SCAN_SILENT_ILLDIGIT |
2316 PERL_SCAN_DISALLOW_PREFIX;
2318 *val = grok_hex((char *)l, &numlen, &flags, NULL);
2327 Perl_croak(aTHX_ "%s: illegal mapping '%s'",
2333 *val = 0; /* bits == 1, then any val should be ignored */
2335 else { /* Nothing following range min, should be single element with no
2341 Perl_croak(aTHX_ "%s: illegal mapping '%s'", typestr, l);
2345 *val = 0; /* bits == 1, then val should be ignored */
2348 /* Position to next line if any, or EOF */
2358 * Returns a swatch (a bit vector string) for a code point sequence
2359 * that starts from the value C<start> and comprises the number C<span>.
2360 * A C<swash> must be an object created by SWASHNEW (see lib/utf8_heavy.pl).
2361 * Should be used via swash_fetch, which will cache the swatch in C<swash>.
2364 S_swash_get(pTHX_ SV* swash, UV start, UV span)
2367 U8 *l, *lend, *x, *xend, *s;
2368 STRLEN lcur, xcur, scur;
2369 HV *const hv = MUTABLE_HV(SvRV(swash));
2371 /* The string containing the main body of the table */
2372 SV** const listsvp = hv_fetchs(hv, "LIST", FALSE);
2374 SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE);
2375 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
2376 SV** const nonesvp = hv_fetchs(hv, "NONE", FALSE);
2377 SV** const extssvp = hv_fetchs(hv, "EXTRAS", FALSE);
2378 const U8* const typestr = (U8*)SvPV_nolen(*typesvp);
2379 const STRLEN bits = SvUV(*bitssvp);
2380 const STRLEN octets = bits >> 3; /* if bits == 1, then octets == 0 */
2381 const UV none = SvUV(*nonesvp);
2382 const UV end = start + span;
2384 PERL_ARGS_ASSERT_SWASH_GET;
2386 if (bits != 1 && bits != 8 && bits != 16 && bits != 32) {
2387 Perl_croak(aTHX_ "panic: swash_get doesn't expect bits %"UVuf,
2391 /* create and initialize $swatch */
2392 scur = octets ? (span * octets) : (span + 7) / 8;
2393 swatch = newSV(scur);
2395 s = (U8*)SvPVX(swatch);
2396 if (octets && none) {
2397 const U8* const e = s + scur;
2400 *s++ = (U8)(none & 0xff);
2401 else if (bits == 16) {
2402 *s++ = (U8)((none >> 8) & 0xff);
2403 *s++ = (U8)( none & 0xff);
2405 else if (bits == 32) {
2406 *s++ = (U8)((none >> 24) & 0xff);
2407 *s++ = (U8)((none >> 16) & 0xff);
2408 *s++ = (U8)((none >> 8) & 0xff);
2409 *s++ = (U8)( none & 0xff);
2415 (void)memzero((U8*)s, scur + 1);
2417 SvCUR_set(swatch, scur);
2418 s = (U8*)SvPVX(swatch);
2420 /* read $swash->{LIST} */
2421 l = (U8*)SvPV(*listsvp, lcur);
2425 l = S_swash_scan_list_line(aTHX_ l, lend, &min, &max, &val,
2426 cBOOL(octets), typestr);
2431 /* If looking for something beyond this range, go try the next one */
2438 if (!none || val < none) {
2443 for (key = min; key <= max; key++) {
2447 /* offset must be non-negative (start <= min <= key < end) */
2448 offset = octets * (key - start);
2450 s[offset] = (U8)(val & 0xff);
2451 else if (bits == 16) {
2452 s[offset ] = (U8)((val >> 8) & 0xff);
2453 s[offset + 1] = (U8)( val & 0xff);
2455 else if (bits == 32) {
2456 s[offset ] = (U8)((val >> 24) & 0xff);
2457 s[offset + 1] = (U8)((val >> 16) & 0xff);
2458 s[offset + 2] = (U8)((val >> 8) & 0xff);
2459 s[offset + 3] = (U8)( val & 0xff);
2462 if (!none || val < none)
2466 else { /* bits == 1, then val should be ignored */
2470 for (key = min; key <= max; key++) {
2471 const STRLEN offset = (STRLEN)(key - start);
2474 s[offset >> 3] |= 1 << (offset & 7);
2480 /* read $swash->{EXTRAS} */
2481 x = (U8*)SvPV(*extssvp, xcur);
2489 SV **otherbitssvp, *other;
2493 const U8 opc = *x++;
2497 nl = (U8*)memchr(x, '\n', xend - x);
2499 if (opc != '-' && opc != '+' && opc != '!' && opc != '&') {
2501 x = nl + 1; /* 1 is length of "\n" */
2505 x = xend; /* to EXTRAS' end at which \n is not found */
2512 namelen = nl - namestr;
2516 namelen = xend - namestr;
2520 othersvp = hv_fetch(hv, (char *)namestr, namelen, FALSE);
2521 otherhv = MUTABLE_HV(SvRV(*othersvp));
2522 otherbitssvp = hv_fetchs(otherhv, "BITS", FALSE);
2523 otherbits = (STRLEN)SvUV(*otherbitssvp);
2524 if (bits < otherbits)
2525 Perl_croak(aTHX_ "panic: swash_get found swatch size mismatch");
2527 /* The "other" swatch must be destroyed after. */
2528 other = swash_get(*othersvp, start, span);
2529 o = (U8*)SvPV(other, olen);
2532 Perl_croak(aTHX_ "panic: swash_get got improper swatch");
2534 s = (U8*)SvPV(swatch, slen);
2535 if (bits == 1 && otherbits == 1) {
2537 Perl_croak(aTHX_ "panic: swash_get found swatch length mismatch");
2561 STRLEN otheroctets = otherbits >> 3;
2563 U8* const send = s + slen;
2568 if (otherbits == 1) {
2569 otherval = (o[offset >> 3] >> (offset & 7)) & 1;
2573 STRLEN vlen = otheroctets;
2581 if (opc == '+' && otherval)
2582 NOOP; /* replace with otherval */
2583 else if (opc == '!' && !otherval)
2585 else if (opc == '-' && otherval)
2587 else if (opc == '&' && !otherval)
2590 s += octets; /* no replacement */
2595 *s++ = (U8)( otherval & 0xff);
2596 else if (bits == 16) {
2597 *s++ = (U8)((otherval >> 8) & 0xff);
2598 *s++ = (U8)( otherval & 0xff);
2600 else if (bits == 32) {
2601 *s++ = (U8)((otherval >> 24) & 0xff);
2602 *s++ = (U8)((otherval >> 16) & 0xff);
2603 *s++ = (U8)((otherval >> 8) & 0xff);
2604 *s++ = (U8)( otherval & 0xff);
2608 sv_free(other); /* through with it! */
2614 Perl__swash_inversion_hash(pTHX_ SV* const swash)
2617 /* Subject to change or removal. For use only in one place in regexec.c
2619 * Returns a hash which is the inversion and closure of a swash mapping.
2620 * For example, consider the input lines:
2625 * The returned hash would have two keys, the utf8 for 006B and the utf8 for
2626 * 006C. The value for each key is an array. For 006C, the array would
2627 * have a two elements, the utf8 for itself, and for 004C. For 006B, there
2628 * would be three elements in its array, the utf8 for 006B, 004B and 212A.
2630 * Essentially, for any code point, it gives all the code points that map to
2631 * it, or the list of 'froms' for that point.
2633 * Currently it only looks at the main body of the swash, and ignores any
2634 * additions or deletions from other swashes */
2638 HV *const hv = MUTABLE_HV(SvRV(swash));
2640 /* The string containing the main body of the table */
2641 SV** const listsvp = hv_fetchs(hv, "LIST", FALSE);
2643 SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE);
2644 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
2645 SV** const nonesvp = hv_fetchs(hv, "NONE", FALSE);
2646 /*SV** const extssvp = hv_fetchs(hv, "EXTRAS", FALSE);*/
2647 const U8* const typestr = (U8*)SvPV_nolen(*typesvp);
2648 const STRLEN bits = SvUV(*bitssvp);
2649 const STRLEN octets = bits >> 3; /* if bits == 1, then octets == 0 */
2650 const UV none = SvUV(*nonesvp);
2654 PERL_ARGS_ASSERT__SWASH_INVERSION_HASH;
2656 /* Must have at least 8 bits to get the mappings */
2657 if (bits != 8 && bits != 16 && bits != 32) {
2658 Perl_croak(aTHX_ "panic: swash_inversion_hash doesn't expect bits %"UVuf,
2662 /* read $swash->{LIST} */
2663 l = (U8*)SvPV(*listsvp, lcur);
2666 /* Go through each input line */
2670 l = S_swash_scan_list_line(aTHX_ l, lend, &min, &max, &val,
2671 cBOOL(octets), typestr);
2676 /* Each element in the range is to be inverted */
2677 for (inverse = min; inverse <= max; inverse++) {
2682 bool found_key = FALSE;
2684 /* The key is the inverse mapping */
2685 char key[UTF8_MAXBYTES+1];
2686 char* key_end = (char *) uvuni_to_utf8((U8*) key, val);
2687 STRLEN key_len = key_end - key;
2689 /* Get the list for the map */
2690 if ((listp = hv_fetch(ret, key, key_len, FALSE))) {
2691 list = (AV*) *listp;
2693 else { /* No entry yet for it: create one */
2695 if (! hv_store(ret, key, key_len, (SV*) list, FALSE)) {
2696 Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed");
2700 for (i = 0; i < av_len(list); i++) {
2701 SV** entryp = av_fetch(list, i, FALSE);
2703 if (entryp == NULL) {
2704 Perl_croak(aTHX_ "panic: av_fetch() unexpectedly failed");
2707 if (SvUV(entry) == val) {
2713 /* Make sure there is a mapping to itself on the list */
2715 element = newSVuv(val);
2716 av_push(list, element);
2720 /* Simply add the value to the list */
2721 element = newSVuv(inverse);
2722 av_push(list, element);
2724 /* swash_get() increments the value of val for each element in the
2725 * range. That makes more compact tables possible. You can
2726 * express the capitalization, for example, of all consecutive
2727 * letters with a single line: 0061\t007A\t0041 This maps 0061 to
2728 * 0041, 0062 to 0042, etc. I (khw) have never understood 'none',
2729 * and it's not documented, and perhaps not even currently used,
2730 * but I copied the semantics from swash_get(), just in case */
2731 if (!none || val < none) {
2741 Perl__swash_to_invlist(pTHX_ SV* const swash)
2744 /* Subject to change or removal. For use only in one place in regcomp.c */
2749 HV *const hv = MUTABLE_HV(SvRV(swash));
2750 UV elements = 0; /* Number of elements in the inversion list */
2753 /* The string containing the main body of the table */
2754 SV** const listsvp = hv_fetchs(hv, "LIST", FALSE);
2755 SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE);
2756 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
2758 const U8* const typestr = (U8*)SvPV_nolen(*typesvp);
2759 const STRLEN bits = SvUV(*bitssvp);
2760 const STRLEN octets = bits >> 3; /* if bits == 1, then octets == 0 */
2764 PERL_ARGS_ASSERT__SWASH_TO_INVLIST;
2766 /* read $swash->{LIST} */
2767 if (SvPOK(*listsvp)) {
2768 l = (U8*)SvPV(*listsvp, lcur);
2771 /* LIST legitimately doesn't contain a string during compilation phases
2772 * of Perl itself, before the Unicode tables are generated. In this
2773 * case, just fake things up by creating an empty list */
2780 /* Scan the input to count the number of lines to preallocate array size
2781 * based on worst possible case, which is each line in the input creates 2
2782 * elements in the inversion list: 1) the beginning of a range in the list;
2783 * 2) the beginning of a range not in the list. */
2784 while ((loc = (strchr(loc, '\n'))) != NULL) {
2789 /* If the ending is somehow corrupt and isn't a new line, add another
2790 * element for the final range that isn't in the inversion list */
2791 if (! (*lend == '\n' || (*lend == '\0' && *(lend - 1) == '\n'))) {
2795 invlist = _new_invlist(elements);
2797 /* Now go through the input again, adding each range to the list */
2800 UV val; /* Not used by this function */
2802 l = S_swash_scan_list_line(aTHX_ l, lend, &start, &end, &val,
2803 cBOOL(octets), typestr);
2809 _append_range_to_invlist(invlist, start, end);
2816 =for apidoc uvchr_to_utf8
2818 Adds the UTF-8 representation of the Native code point C<uv> to the end
2819 of the string C<d>; C<d> should be have at least C<UTF8_MAXBYTES+1> free
2820 bytes available. The return value is the pointer to the byte after the
2821 end of the new character. In other words,
2823 d = uvchr_to_utf8(d, uv);
2825 is the recommended wide native character-aware way of saying
2832 /* On ASCII machines this is normally a macro but we want a
2833 real function in case XS code wants it
2836 Perl_uvchr_to_utf8(pTHX_ U8 *d, UV uv)
2838 PERL_ARGS_ASSERT_UVCHR_TO_UTF8;
2840 return Perl_uvuni_to_utf8_flags(aTHX_ d, NATIVE_TO_UNI(uv), 0);
2844 Perl_uvchr_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags)
2846 PERL_ARGS_ASSERT_UVCHR_TO_UTF8_FLAGS;
2848 return Perl_uvuni_to_utf8_flags(aTHX_ d, NATIVE_TO_UNI(uv), flags);
2852 =for apidoc utf8n_to_uvchr
2854 Returns the native character value of the first character in the string
2856 which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
2857 length, in bytes, of that character.
2859 length and flags are the same as utf8n_to_uvuni().
2863 /* On ASCII machines this is normally a macro but we want
2864 a real function in case XS code wants it
2867 Perl_utf8n_to_uvchr(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen,
2870 const UV uv = Perl_utf8n_to_uvuni(aTHX_ s, curlen, retlen, flags);
2872 PERL_ARGS_ASSERT_UTF8N_TO_UVCHR;
2874 return UNI_TO_NATIVE(uv);
2878 Perl_check_utf8_print(pTHX_ register const U8* s, const STRLEN len)
2880 /* May change: warns if surrogates, non-character code points, or
2881 * non-Unicode code points are in s which has length len. Returns TRUE if
2882 * none found; FALSE otherwise. The only other validity check is to make
2883 * sure that this won't exceed the string's length */
2885 const U8* const e = s + len;
2888 PERL_ARGS_ASSERT_CHECK_UTF8_PRINT;
2891 if (UTF8SKIP(s) > len) {
2892 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8),
2893 "%s in %s", unees, PL_op ? OP_DESC(PL_op) : "print");
2896 if (*s >= UTF8_FIRST_PROBLEMATIC_CODE_POINT_FIRST_BYTE) {
2898 if (UTF8_IS_SUPER(s)) {
2899 if (ckWARN_d(WARN_NON_UNICODE)) {
2900 UV uv = utf8_to_uvchr(s, &char_len);
2901 Perl_warner(aTHX_ packWARN(WARN_NON_UNICODE),
2902 "Code point 0x%04"UVXf" is not Unicode, may not be portable", uv);
2906 else if (UTF8_IS_SURROGATE(s)) {
2907 if (ckWARN_d(WARN_SURROGATE)) {
2908 UV uv = utf8_to_uvchr(s, &char_len);
2909 Perl_warner(aTHX_ packWARN(WARN_SURROGATE),
2910 "Unicode surrogate U+%04"UVXf" is illegal in UTF-8", uv);
2915 ((UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s))
2916 && (ckWARN_d(WARN_NONCHAR)))
2918 UV uv = utf8_to_uvchr(s, &char_len);
2919 Perl_warner(aTHX_ packWARN(WARN_NONCHAR),
2920 "Unicode non-character U+%04"UVXf" is illegal for open interchange", uv);
2931 =for apidoc pv_uni_display
2933 Build to the scalar dsv a displayable version of the string spv,
2934 length len, the displayable version being at most pvlim bytes long
2935 (if longer, the rest is truncated and "..." will be appended).
2937 The flags argument can have UNI_DISPLAY_ISPRINT set to display
2938 isPRINT()able characters as themselves, UNI_DISPLAY_BACKSLASH
2939 to display the \\[nrfta\\] as the backslashed versions (like '\n')
2940 (UNI_DISPLAY_BACKSLASH is preferred over UNI_DISPLAY_ISPRINT for \\).
2941 UNI_DISPLAY_QQ (and its alias UNI_DISPLAY_REGEX) have both
2942 UNI_DISPLAY_BACKSLASH and UNI_DISPLAY_ISPRINT turned on.
2944 The pointer to the PV of the dsv is returned.
2948 Perl_pv_uni_display(pTHX_ SV *dsv, const U8 *spv, STRLEN len, STRLEN pvlim, UV flags)
2953 PERL_ARGS_ASSERT_PV_UNI_DISPLAY;
2957 for (s = (const char *)spv, e = s + len; s < e; s += UTF8SKIP(s)) {
2959 /* This serves double duty as a flag and a character to print after
2960 a \ when flags & UNI_DISPLAY_BACKSLASH is true.
2964 if (pvlim && SvCUR(dsv) >= pvlim) {
2968 u = utf8_to_uvchr((U8*)s, 0);
2970 const unsigned char c = (unsigned char)u & 0xFF;
2971 if (flags & UNI_DISPLAY_BACKSLASH) {
2988 const char string = ok;
2989 sv_catpvs(dsv, "\\");
2990 sv_catpvn(dsv, &string, 1);
2993 /* isPRINT() is the locale-blind version. */
2994 if (!ok && (flags & UNI_DISPLAY_ISPRINT) && isPRINT(c)) {
2995 const char string = c;
2996 sv_catpvn(dsv, &string, 1);
3001 Perl_sv_catpvf(aTHX_ dsv, "\\x{%"UVxf"}", u);
3004 sv_catpvs(dsv, "...");
3010 =for apidoc sv_uni_display
3012 Build to the scalar dsv a displayable version of the scalar sv,
3013 the displayable version being at most pvlim bytes long
3014 (if longer, the rest is truncated and "..." will be appended).
3016 The flags argument is as in pv_uni_display().
3018 The pointer to the PV of the dsv is returned.
3023 Perl_sv_uni_display(pTHX_ SV *dsv, SV *ssv, STRLEN pvlim, UV flags)
3025 PERL_ARGS_ASSERT_SV_UNI_DISPLAY;
3027 return Perl_pv_uni_display(aTHX_ dsv, (const U8*)SvPVX_const(ssv),
3028 SvCUR(ssv), pvlim, flags);
3032 =for apidoc foldEQ_utf8
3034 Returns true if the leading portions of the strings s1 and s2 (either or both
3035 of which may be in UTF-8) are the same case-insensitively; false otherwise.
3036 How far into the strings to compare is determined by other input parameters.
3038 If u1 is true, the string s1 is assumed to be in UTF-8-encoded Unicode;
3039 otherwise it is assumed to be in native 8-bit encoding. Correspondingly for u2
3042 If the byte length l1 is non-zero, it says how far into s1 to check for fold
3043 equality. In other words, s1+l1 will be used as a goal to reach. The
3044 scan will not be considered to be a match unless the goal is reached, and
3045 scanning won't continue past that goal. Correspondingly for l2 with respect to
3048 If pe1 is non-NULL and the pointer it points to is not NULL, that pointer is
3049 considered an end pointer beyond which scanning of s1 will not continue under
3050 any circumstances. This means that if both l1 and pe1 are specified, and pe1
3051 is less than s1+l1, the match will never be successful because it can never
3052 get as far as its goal (and in fact is asserted against). Correspondingly for
3053 pe2 with respect to s2.
3055 At least one of s1 and s2 must have a goal (at least one of l1 and l2 must be
3056 non-zero), and if both do, both have to be
3057 reached for a successful match. Also, if the fold of a character is multiple
3058 characters, all of them must be matched (see tr21 reference below for
3061 Upon a successful match, if pe1 is non-NULL,
3062 it will be set to point to the beginning of the I<next> character of s1 beyond
3063 what was matched. Correspondingly for pe2 and s2.
3065 For case-insensitiveness, the "casefolding" of Unicode is used
3066 instead of upper/lowercasing both the characters, see
3067 http://www.unicode.org/unicode/reports/tr21/ (Case Mappings).
3071 /* A flags parameter has been added which may change, and hence isn't
3072 * externally documented. Currently it is:
3073 * 0 for as-documented above
3074 * FOLDEQ_UTF8_NOMIX_ASCII meaning that if a non-ASCII character folds to an
3075 ASCII one, to not match
3076 * FOLDEQ_UTF8_LOCALE meaning that locale rules are to be used for code
3077 * points below 256; unicode rules for above 255; and
3078 * folds that cross those boundaries are disallowed,
3079 * like the NOMIX_ASCII option
3082 Perl_foldEQ_utf8_flags(pTHX_ const char *s1, char **pe1, register UV l1, bool u1, const char *s2, char **pe2, register UV l2, bool u2, U32 flags)
3085 register const U8 *p1 = (const U8*)s1; /* Point to current char */
3086 register const U8 *p2 = (const U8*)s2;
3087 register const U8 *g1 = NULL; /* goal for s1 */
3088 register const U8 *g2 = NULL;
3089 register const U8 *e1 = NULL; /* Don't scan s1 past this */
3090 register U8 *f1 = NULL; /* Point to current folded */
3091 register const U8 *e2 = NULL;
3092 register U8 *f2 = NULL;
3093 STRLEN n1 = 0, n2 = 0; /* Number of bytes in current char */
3094 U8 foldbuf1[UTF8_MAXBYTES_CASE+1];
3095 U8 foldbuf2[UTF8_MAXBYTES_CASE+1];
3096 U8 natbuf[2]; /* Holds native 8-bit char converted to utf8;
3097 these always fit in 2 bytes */
3099 PERL_ARGS_ASSERT_FOLDEQ_UTF8_FLAGS;
3106 g1 = (const U8*)s1 + l1;
3114 g2 = (const U8*)s2 + l2;
3117 /* Must have at least one goal */
3122 /* Will never match if goal is out-of-bounds */
3123 assert(! e1 || e1 >= g1);
3125 /* Here, there isn't an end pointer, or it is beyond the goal. We
3126 * only go as far as the goal */
3130 assert(e1); /* Must have an end for looking at s1 */
3133 /* Same for goal for s2 */
3135 assert(! e2 || e2 >= g2);
3142 /* Look through both strings, a character at a time */
3143 while (p1 < e1 && p2 < e2) {
3145 /* If at the beginning of a new character in s1, get its fold to use
3146 * and the length of the fold. (exception: locale rules just get the
3147 * character to a single byte) */
3150 /* If in locale matching, we use two sets of rules, depending on if
3151 * the code point is above or below 255. Here, we test for and
3152 * handle locale rules */
3153 if ((flags & FOLDEQ_UTF8_LOCALE)
3154 && (! u1 || UTF8_IS_INVARIANT(*p1) || UTF8_IS_DOWNGRADEABLE_START(*p1)))
3156 /* There is no mixing of code points above and below 255. */
3157 if (u2 && (! UTF8_IS_INVARIANT(*p2)
3158 && ! UTF8_IS_DOWNGRADEABLE_START(*p2)))
3163 /* We handle locale rules by converting, if necessary, the code
3164 * point to a single byte. */
3165 if (! u1 || UTF8_IS_INVARIANT(*p1)) {
3169 *foldbuf1 = TWO_BYTE_UTF8_TO_UNI(*p1, *(p1 + 1));
3173 else if (isASCII(*p1)) { /* Note, that here won't be both ASCII
3174 and using locale rules */
3176 /* If trying to mix non- with ASCII, and not supposed to, fail */
3177 if ((flags & FOLDEQ_UTF8_NOMIX_ASCII) && ! isASCII(*p2)) {
3181 *foldbuf1 = toLOWER(*p1); /* Folds in the ASCII range are
3185 to_utf8_fold(p1, foldbuf1, &n1);
3187 else { /* Not utf8, convert to it first and then get fold */
3188 uvuni_to_utf8(natbuf, (UV) NATIVE_TO_UNI(((UV)*p1)));
3189 to_utf8_fold(natbuf, foldbuf1, &n1);
3194 if (n2 == 0) { /* Same for s2 */
3195 if ((flags & FOLDEQ_UTF8_LOCALE)
3196 && (! u2 || UTF8_IS_INVARIANT(*p2) || UTF8_IS_DOWNGRADEABLE_START(*p2)))
3198 /* Here, the next char in s2 is < 256. We've already worked on
3199 * s1, and if it isn't also < 256, can't match */
3200 if (u1 && (! UTF8_IS_INVARIANT(*p1)
3201 && ! UTF8_IS_DOWNGRADEABLE_START(*p1)))
3205 if (! u2 || UTF8_IS_INVARIANT(*p2)) {
3209 *foldbuf2 = TWO_BYTE_UTF8_TO_UNI(*p2, *(p2 + 1));
3212 /* Use another function to handle locale rules. We've made
3213 * sure that both characters to compare are single bytes */
3214 if (! foldEQ_locale((char *) f1, (char *) foldbuf2, 1)) {
3219 else if (isASCII(*p2)) {
3220 if (flags && ! isASCII(*p1)) {
3224 *foldbuf2 = toLOWER(*p2);
3227 to_utf8_fold(p2, foldbuf2, &n2);
3230 uvuni_to_utf8(natbuf, (UV) NATIVE_TO_UNI(((UV)*p2)));
3231 to_utf8_fold(natbuf, foldbuf2, &n2);
3236 /* Here f1 and f2 point to the beginning of the strings to compare.
3237 * These strings are the folds of the input characters, stored in utf8.
3240 /* While there is more to look for in both folds, see if they
3241 * continue to match */
3243 U8 fold_length = UTF8SKIP(f1);
3244 if (fold_length != UTF8SKIP(f2)
3245 || (fold_length == 1 && *f1 != *f2) /* Short circuit memNE
3246 function call for single
3248 || memNE((char*)f1, (char*)f2, fold_length))
3250 return 0; /* mismatch */
3253 /* Here, they matched, advance past them */
3260 /* When reach the end of any fold, advance the input past it */
3262 p1 += u1 ? UTF8SKIP(p1) : 1;
3265 p2 += u2 ? UTF8SKIP(p2) : 1;
3267 } /* End of loop through both strings */
3269 /* A match is defined by each scan that specified an explicit length
3270 * reaching its final goal, and the other not having matched a partial
3271 * character (which can happen when the fold of a character is more than one
3273 if (! ((g1 == 0 || p1 == g1) && (g2 == 0 || p2 == g2)) || n1 || n2) {
3277 /* Successful match. Set output pointers */
3289 * c-indentation-style: bsd
3291 * indent-tabs-mode: t
3294 * ex: set ts=8 sts=4 sw=4 noet: