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), "%s", 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), "%s", 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().
1094 /* This logic is duplicated in sv_catpvn_flags, so any bug fixes will
1095 likewise need duplication. */
1098 Perl_bytes_to_utf8(pTHX_ const U8 *s, STRLEN *len)
1100 const U8 * const send = s + (*len);
1104 PERL_ARGS_ASSERT_BYTES_TO_UTF8;
1105 PERL_UNUSED_CONTEXT;
1107 Newx(d, (*len) * 2 + 1, U8);
1111 const UV uv = NATIVE_TO_ASCII(*s++);
1112 if (UNI_IS_INVARIANT(uv))
1113 *d++ = (U8)UTF_TO_NATIVE(uv);
1115 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
1116 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
1125 * Convert native (big-endian) or reversed (little-endian) UTF-16 to UTF-8.
1127 * Destination must be pre-extended to 3/2 source. Do not use in-place.
1128 * We optimize for native, for obvious reasons. */
1131 Perl_utf16_to_utf8(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen)
1136 PERL_ARGS_ASSERT_UTF16_TO_UTF8;
1139 Perl_croak(aTHX_ "panic: utf16_to_utf8: odd bytelen %"UVuf, (UV)bytelen);
1144 UV uv = (p[0] << 8) + p[1]; /* UTF-16BE */
1148 *d++ = UNI_TO_NATIVE(uv);
1155 *d++ = (U8)(( uv >> 6) | 0xc0);
1156 *d++ = (U8)(( uv & 0x3f) | 0x80);
1159 if (uv >= 0xd800 && uv <= 0xdbff) { /* surrogates */
1161 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
1163 UV low = (p[0] << 8) + p[1];
1165 if (low < 0xdc00 || low > 0xdfff)
1166 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
1167 uv = ((uv - 0xd800) << 10) + (low - 0xdc00) + 0x10000;
1169 } else if (uv >= 0xdc00 && uv <= 0xdfff) {
1170 Perl_croak(aTHX_ "Malformed UTF-16 surrogate");
1173 *d++ = (U8)(( uv >> 12) | 0xe0);
1174 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
1175 *d++ = (U8)(( uv & 0x3f) | 0x80);
1179 *d++ = (U8)(( uv >> 18) | 0xf0);
1180 *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
1181 *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
1182 *d++ = (U8)(( uv & 0x3f) | 0x80);
1186 *newlen = d - dstart;
1190 /* Note: this one is slightly destructive of the source. */
1193 Perl_utf16_to_utf8_reversed(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen)
1196 U8* const send = s + bytelen;
1198 PERL_ARGS_ASSERT_UTF16_TO_UTF8_REVERSED;
1201 Perl_croak(aTHX_ "panic: utf16_to_utf8_reversed: odd bytelen %"UVuf,
1205 const U8 tmp = s[0];
1210 return utf16_to_utf8(p, d, bytelen, newlen);
1213 /* for now these are all defined (inefficiently) in terms of the utf8 versions */
1216 Perl_is_uni_alnum(pTHX_ UV c)
1218 U8 tmpbuf[UTF8_MAXBYTES+1];
1219 uvchr_to_utf8(tmpbuf, c);
1220 return is_utf8_alnum(tmpbuf);
1224 Perl_is_uni_idfirst(pTHX_ UV c)
1226 U8 tmpbuf[UTF8_MAXBYTES+1];
1227 uvchr_to_utf8(tmpbuf, c);
1228 return is_utf8_idfirst(tmpbuf);
1232 Perl_is_uni_alpha(pTHX_ UV c)
1234 U8 tmpbuf[UTF8_MAXBYTES+1];
1235 uvchr_to_utf8(tmpbuf, c);
1236 return is_utf8_alpha(tmpbuf);
1240 Perl_is_uni_ascii(pTHX_ UV c)
1246 Perl_is_uni_space(pTHX_ UV c)
1248 U8 tmpbuf[UTF8_MAXBYTES+1];
1249 uvchr_to_utf8(tmpbuf, c);
1250 return is_utf8_space(tmpbuf);
1254 Perl_is_uni_digit(pTHX_ UV c)
1256 U8 tmpbuf[UTF8_MAXBYTES+1];
1257 uvchr_to_utf8(tmpbuf, c);
1258 return is_utf8_digit(tmpbuf);
1262 Perl_is_uni_upper(pTHX_ UV c)
1264 U8 tmpbuf[UTF8_MAXBYTES+1];
1265 uvchr_to_utf8(tmpbuf, c);
1266 return is_utf8_upper(tmpbuf);
1270 Perl_is_uni_lower(pTHX_ UV c)
1272 U8 tmpbuf[UTF8_MAXBYTES+1];
1273 uvchr_to_utf8(tmpbuf, c);
1274 return is_utf8_lower(tmpbuf);
1278 Perl_is_uni_cntrl(pTHX_ UV c)
1280 return isCNTRL_L1(c);
1284 Perl_is_uni_graph(pTHX_ UV c)
1286 U8 tmpbuf[UTF8_MAXBYTES+1];
1287 uvchr_to_utf8(tmpbuf, c);
1288 return is_utf8_graph(tmpbuf);
1292 Perl_is_uni_print(pTHX_ UV c)
1294 U8 tmpbuf[UTF8_MAXBYTES+1];
1295 uvchr_to_utf8(tmpbuf, c);
1296 return is_utf8_print(tmpbuf);
1300 Perl_is_uni_punct(pTHX_ UV c)
1302 U8 tmpbuf[UTF8_MAXBYTES+1];
1303 uvchr_to_utf8(tmpbuf, c);
1304 return is_utf8_punct(tmpbuf);
1308 Perl_is_uni_xdigit(pTHX_ UV c)
1310 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1311 uvchr_to_utf8(tmpbuf, c);
1312 return is_utf8_xdigit(tmpbuf);
1316 Perl_to_uni_upper(pTHX_ UV c, U8* p, STRLEN *lenp)
1318 PERL_ARGS_ASSERT_TO_UNI_UPPER;
1320 uvchr_to_utf8(p, c);
1321 return to_utf8_upper(p, p, lenp);
1325 Perl_to_uni_title(pTHX_ UV c, U8* p, STRLEN *lenp)
1327 PERL_ARGS_ASSERT_TO_UNI_TITLE;
1329 uvchr_to_utf8(p, c);
1330 return to_utf8_title(p, p, lenp);
1334 Perl_to_uni_lower(pTHX_ UV c, U8* p, STRLEN *lenp)
1336 PERL_ARGS_ASSERT_TO_UNI_LOWER;
1338 uvchr_to_utf8(p, c);
1339 return to_utf8_lower(p, p, lenp);
1343 Perl__to_uni_fold_flags(pTHX_ UV c, U8* p, STRLEN *lenp, U8 flags)
1345 PERL_ARGS_ASSERT__TO_UNI_FOLD_FLAGS;
1347 uvchr_to_utf8(p, c);
1348 return _to_utf8_fold_flags(p, p, lenp, flags);
1351 /* for now these all assume no locale info available for Unicode > 255 */
1354 Perl_is_uni_alnum_lc(pTHX_ UV c)
1356 return is_uni_alnum(c); /* XXX no locale support yet */
1360 Perl_is_uni_idfirst_lc(pTHX_ UV c)
1362 return is_uni_idfirst(c); /* XXX no locale support yet */
1366 Perl_is_uni_alpha_lc(pTHX_ UV c)
1368 return is_uni_alpha(c); /* XXX no locale support yet */
1372 Perl_is_uni_ascii_lc(pTHX_ UV c)
1374 return is_uni_ascii(c); /* XXX no locale support yet */
1378 Perl_is_uni_space_lc(pTHX_ UV c)
1380 return is_uni_space(c); /* XXX no locale support yet */
1384 Perl_is_uni_digit_lc(pTHX_ UV c)
1386 return is_uni_digit(c); /* XXX no locale support yet */
1390 Perl_is_uni_upper_lc(pTHX_ UV c)
1392 return is_uni_upper(c); /* XXX no locale support yet */
1396 Perl_is_uni_lower_lc(pTHX_ UV c)
1398 return is_uni_lower(c); /* XXX no locale support yet */
1402 Perl_is_uni_cntrl_lc(pTHX_ UV c)
1404 return is_uni_cntrl(c); /* XXX no locale support yet */
1408 Perl_is_uni_graph_lc(pTHX_ UV c)
1410 return is_uni_graph(c); /* XXX no locale support yet */
1414 Perl_is_uni_print_lc(pTHX_ UV c)
1416 return is_uni_print(c); /* XXX no locale support yet */
1420 Perl_is_uni_punct_lc(pTHX_ UV c)
1422 return is_uni_punct(c); /* XXX no locale support yet */
1426 Perl_is_uni_xdigit_lc(pTHX_ UV c)
1428 return is_uni_xdigit(c); /* XXX no locale support yet */
1432 Perl_to_uni_upper_lc(pTHX_ U32 c)
1434 /* XXX returns only the first character -- do not use XXX */
1435 /* XXX no locale support yet */
1437 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1438 return (U32)to_uni_upper(c, tmpbuf, &len);
1442 Perl_to_uni_title_lc(pTHX_ U32 c)
1444 /* XXX returns only the first character XXX -- do not use XXX */
1445 /* XXX no locale support yet */
1447 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1448 return (U32)to_uni_title(c, tmpbuf, &len);
1452 Perl_to_uni_lower_lc(pTHX_ U32 c)
1454 /* XXX returns only the first character -- do not use XXX */
1455 /* XXX no locale support yet */
1457 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1458 return (U32)to_uni_lower(c, tmpbuf, &len);
1462 S_is_utf8_common(pTHX_ const U8 *const p, SV **swash,
1463 const char *const swashname)
1467 PERL_ARGS_ASSERT_IS_UTF8_COMMON;
1469 if (!is_utf8_char(p))
1472 *swash = swash_init("utf8", swashname, &PL_sv_undef, 1, 0);
1473 return swash_fetch(*swash, p, TRUE) != 0;
1477 Perl_is_utf8_alnum(pTHX_ const U8 *p)
1481 PERL_ARGS_ASSERT_IS_UTF8_ALNUM;
1483 /* NOTE: "IsWord", not "IsAlnum", since Alnum is a true
1484 * descendant of isalnum(3), in other words, it doesn't
1485 * contain the '_'. --jhi */
1486 return is_utf8_common(p, &PL_utf8_alnum, "IsWord");
1490 Perl_is_utf8_idfirst(pTHX_ const U8 *p) /* The naming is historical. */
1494 PERL_ARGS_ASSERT_IS_UTF8_IDFIRST;
1498 /* is_utf8_idstart would be more logical. */
1499 return is_utf8_common(p, &PL_utf8_idstart, "IdStart");
1503 Perl_is_utf8_xidfirst(pTHX_ const U8 *p) /* The naming is historical. */
1507 PERL_ARGS_ASSERT_IS_UTF8_XIDFIRST;
1511 /* is_utf8_idstart would be more logical. */
1512 return is_utf8_common(p, &PL_utf8_xidstart, "XIdStart");
1516 Perl__is_utf8__perl_idstart(pTHX_ const U8 *p)
1520 PERL_ARGS_ASSERT__IS_UTF8__PERL_IDSTART;
1522 return is_utf8_common(p, &PL_utf8_perl_idstart, "_Perl_IDStart");
1526 Perl_is_utf8_idcont(pTHX_ const U8 *p)
1530 PERL_ARGS_ASSERT_IS_UTF8_IDCONT;
1532 return is_utf8_common(p, &PL_utf8_idcont, "IdContinue");
1536 Perl_is_utf8_xidcont(pTHX_ const U8 *p)
1540 PERL_ARGS_ASSERT_IS_UTF8_XIDCONT;
1542 return is_utf8_common(p, &PL_utf8_idcont, "XIdContinue");
1546 Perl_is_utf8_alpha(pTHX_ const U8 *p)
1550 PERL_ARGS_ASSERT_IS_UTF8_ALPHA;
1552 return is_utf8_common(p, &PL_utf8_alpha, "IsAlpha");
1556 Perl_is_utf8_ascii(pTHX_ const U8 *p)
1560 PERL_ARGS_ASSERT_IS_UTF8_ASCII;
1562 /* ASCII characters are the same whether in utf8 or not. So the macro
1563 * works on both utf8 and non-utf8 representations. */
1568 Perl_is_utf8_space(pTHX_ const U8 *p)
1572 PERL_ARGS_ASSERT_IS_UTF8_SPACE;
1574 return is_utf8_common(p, &PL_utf8_space, "IsXPerlSpace");
1578 Perl_is_utf8_perl_space(pTHX_ const U8 *p)
1582 PERL_ARGS_ASSERT_IS_UTF8_PERL_SPACE;
1584 /* Only true if is an ASCII space-like character, and ASCII is invariant
1585 * under utf8, so can just use the macro */
1586 return isSPACE_A(*p);
1590 Perl_is_utf8_perl_word(pTHX_ const U8 *p)
1594 PERL_ARGS_ASSERT_IS_UTF8_PERL_WORD;
1596 /* Only true if is an ASCII word character, and ASCII is invariant
1597 * under utf8, so can just use the macro */
1598 return isWORDCHAR_A(*p);
1602 Perl_is_utf8_digit(pTHX_ const U8 *p)
1606 PERL_ARGS_ASSERT_IS_UTF8_DIGIT;
1608 return is_utf8_common(p, &PL_utf8_digit, "IsDigit");
1612 Perl_is_utf8_posix_digit(pTHX_ const U8 *p)
1616 PERL_ARGS_ASSERT_IS_UTF8_POSIX_DIGIT;
1618 /* Only true if is an ASCII digit character, and ASCII is invariant
1619 * under utf8, so can just use the macro */
1620 return isDIGIT_A(*p);
1624 Perl_is_utf8_upper(pTHX_ const U8 *p)
1628 PERL_ARGS_ASSERT_IS_UTF8_UPPER;
1630 return is_utf8_common(p, &PL_utf8_upper, "IsUppercase");
1634 Perl_is_utf8_lower(pTHX_ const U8 *p)
1638 PERL_ARGS_ASSERT_IS_UTF8_LOWER;
1640 return is_utf8_common(p, &PL_utf8_lower, "IsLowercase");
1644 Perl_is_utf8_cntrl(pTHX_ const U8 *p)
1648 PERL_ARGS_ASSERT_IS_UTF8_CNTRL;
1651 return isCNTRL_A(*p);
1654 /* All controls are in Latin1 */
1655 if (! UTF8_IS_DOWNGRADEABLE_START(*p)) {
1658 return isCNTRL_L1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)));
1662 Perl_is_utf8_graph(pTHX_ const U8 *p)
1666 PERL_ARGS_ASSERT_IS_UTF8_GRAPH;
1668 return is_utf8_common(p, &PL_utf8_graph, "IsGraph");
1672 Perl_is_utf8_print(pTHX_ const U8 *p)
1676 PERL_ARGS_ASSERT_IS_UTF8_PRINT;
1678 return is_utf8_common(p, &PL_utf8_print, "IsPrint");
1682 Perl_is_utf8_punct(pTHX_ const U8 *p)
1686 PERL_ARGS_ASSERT_IS_UTF8_PUNCT;
1688 return is_utf8_common(p, &PL_utf8_punct, "IsPunct");
1692 Perl_is_utf8_xdigit(pTHX_ const U8 *p)
1696 PERL_ARGS_ASSERT_IS_UTF8_XDIGIT;
1698 return is_utf8_common(p, &PL_utf8_xdigit, "IsXDigit");
1702 Perl_is_utf8_mark(pTHX_ const U8 *p)
1706 PERL_ARGS_ASSERT_IS_UTF8_MARK;
1708 return is_utf8_common(p, &PL_utf8_mark, "IsM");
1712 Perl_is_utf8_X_begin(pTHX_ const U8 *p)
1716 PERL_ARGS_ASSERT_IS_UTF8_X_BEGIN;
1718 return is_utf8_common(p, &PL_utf8_X_begin, "_X_Begin");
1722 Perl_is_utf8_X_extend(pTHX_ const U8 *p)
1726 PERL_ARGS_ASSERT_IS_UTF8_X_EXTEND;
1728 return is_utf8_common(p, &PL_utf8_X_extend, "_X_Extend");
1732 Perl_is_utf8_X_prepend(pTHX_ const U8 *p)
1736 PERL_ARGS_ASSERT_IS_UTF8_X_PREPEND;
1738 return is_utf8_common(p, &PL_utf8_X_prepend, "GCB=Prepend");
1742 Perl_is_utf8_X_non_hangul(pTHX_ const U8 *p)
1746 PERL_ARGS_ASSERT_IS_UTF8_X_NON_HANGUL;
1748 return is_utf8_common(p, &PL_utf8_X_non_hangul, "HST=Not_Applicable");
1752 Perl_is_utf8_X_L(pTHX_ const U8 *p)
1756 PERL_ARGS_ASSERT_IS_UTF8_X_L;
1758 return is_utf8_common(p, &PL_utf8_X_L, "GCB=L");
1762 Perl_is_utf8_X_LV(pTHX_ const U8 *p)
1766 PERL_ARGS_ASSERT_IS_UTF8_X_LV;
1768 return is_utf8_common(p, &PL_utf8_X_LV, "GCB=LV");
1772 Perl_is_utf8_X_LVT(pTHX_ const U8 *p)
1776 PERL_ARGS_ASSERT_IS_UTF8_X_LVT;
1778 return is_utf8_common(p, &PL_utf8_X_LVT, "GCB=LVT");
1782 Perl_is_utf8_X_T(pTHX_ const U8 *p)
1786 PERL_ARGS_ASSERT_IS_UTF8_X_T;
1788 return is_utf8_common(p, &PL_utf8_X_T, "GCB=T");
1792 Perl_is_utf8_X_V(pTHX_ const U8 *p)
1796 PERL_ARGS_ASSERT_IS_UTF8_X_V;
1798 return is_utf8_common(p, &PL_utf8_X_V, "GCB=V");
1802 Perl_is_utf8_X_LV_LVT_V(pTHX_ const U8 *p)
1806 PERL_ARGS_ASSERT_IS_UTF8_X_LV_LVT_V;
1808 return is_utf8_common(p, &PL_utf8_X_LV_LVT_V, "_X_LV_LVT_V");
1812 =for apidoc to_utf8_case
1814 The "p" contains the pointer to the UTF-8 string encoding
1815 the character that is being converted.
1817 The "ustrp" is a pointer to the character buffer to put the
1818 conversion result to. The "lenp" is a pointer to the length
1821 The "swashp" is a pointer to the swash to use.
1823 Both the special and normal mappings are stored in lib/unicore/To/Foo.pl,
1824 and loaded by SWASHNEW, using lib/utf8_heavy.pl. The special (usually,
1825 but not always, a multicharacter mapping), is tried first.
1827 The "special" is a string like "utf8::ToSpecLower", which means the
1828 hash %utf8::ToSpecLower. The access to the hash is through
1829 Perl_to_utf8_case().
1831 The "normal" is a string like "ToLower" which means the swash
1837 Perl_to_utf8_case(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp,
1838 SV **swashp, const char *normal, const char *special)
1841 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1843 const UV uv0 = utf8_to_uvchr(p, NULL);
1844 /* The NATIVE_TO_UNI() and UNI_TO_NATIVE() mappings
1845 * are necessary in EBCDIC, they are redundant no-ops
1846 * in ASCII-ish platforms, and hopefully optimized away. */
1847 const UV uv1 = NATIVE_TO_UNI(uv0);
1849 PERL_ARGS_ASSERT_TO_UTF8_CASE;
1851 /* Note that swash_fetch() doesn't output warnings for these because it
1852 * assumes we will */
1853 if (uv1 >= UNICODE_SURROGATE_FIRST) {
1854 if (uv1 <= UNICODE_SURROGATE_LAST) {
1855 if (ckWARN_d(WARN_SURROGATE)) {
1856 const char* desc = (PL_op) ? OP_DESC(PL_op) : normal;
1857 Perl_warner(aTHX_ packWARN(WARN_SURROGATE),
1858 "Operation \"%s\" returns its argument for UTF-16 surrogate U+%04"UVXf"", desc, uv1);
1861 else if (UNICODE_IS_SUPER(uv1)) {
1862 if (ckWARN_d(WARN_NON_UNICODE)) {
1863 const char* desc = (PL_op) ? OP_DESC(PL_op) : normal;
1864 Perl_warner(aTHX_ packWARN(WARN_NON_UNICODE),
1865 "Operation \"%s\" returns its argument for non-Unicode code point 0x%04"UVXf"", desc, uv1);
1869 /* Note that non-characters are perfectly legal, so no warning should
1873 uvuni_to_utf8(tmpbuf, uv1);
1875 if (!*swashp) /* load on-demand */
1876 *swashp = swash_init("utf8", normal, &PL_sv_undef, 4, 0);
1879 /* It might be "special" (sometimes, but not always,
1880 * a multicharacter mapping) */
1881 HV * const hv = get_hv(special, 0);
1885 (svp = hv_fetch(hv, (const char*)tmpbuf, UNISKIP(uv1), FALSE)) &&
1889 s = SvPV_const(*svp, len);
1891 len = uvuni_to_utf8(ustrp, NATIVE_TO_UNI(*(U8*)s)) - ustrp;
1894 /* If we have EBCDIC we need to remap the characters
1895 * since any characters in the low 256 are Unicode
1896 * code points, not EBCDIC. */
1897 U8 *t = (U8*)s, *tend = t + len, *d;
1904 const UV c = utf8_to_uvchr(t, &tlen);
1906 d = uvchr_to_utf8(d, UNI_TO_NATIVE(c));
1915 d = uvchr_to_utf8(d, UNI_TO_NATIVE(*t));
1920 Copy(tmpbuf, ustrp, len, U8);
1922 Copy(s, ustrp, len, U8);
1928 if (!len && *swashp) {
1929 const UV uv2 = swash_fetch(*swashp, tmpbuf, TRUE);
1932 /* It was "normal" (a single character mapping). */
1933 const UV uv3 = UNI_TO_NATIVE(uv2);
1934 len = uvchr_to_utf8(ustrp, uv3) - ustrp;
1938 if (!len) /* Neither: just copy. In other words, there was no mapping
1939 defined, which means that the code point maps to itself */
1940 len = uvchr_to_utf8(ustrp, uv0) - ustrp;
1945 return len ? utf8_to_uvchr(ustrp, 0) : 0;
1949 =for apidoc to_utf8_upper
1951 Convert the UTF-8 encoded character at p to its uppercase version and
1952 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
1953 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since
1954 the uppercase version may be longer than the original character.
1956 The first character of the uppercased version is returned
1957 (but note, as explained above, that there may be more.)
1962 Perl_to_utf8_upper(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
1966 PERL_ARGS_ASSERT_TO_UTF8_UPPER;
1968 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
1969 &PL_utf8_toupper, "ToUpper", "utf8::ToSpecUpper");
1973 =for apidoc to_utf8_title
1975 Convert the UTF-8 encoded character at p to its titlecase version and
1976 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
1977 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
1978 titlecase version may be longer than the original character.
1980 The first character of the titlecased version is returned
1981 (but note, as explained above, that there may be more.)
1986 Perl_to_utf8_title(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
1990 PERL_ARGS_ASSERT_TO_UTF8_TITLE;
1992 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
1993 &PL_utf8_totitle, "ToTitle", "utf8::ToSpecTitle");
1997 =for apidoc to_utf8_lower
1999 Convert the UTF-8 encoded character at p to its lowercase version and
2000 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
2001 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
2002 lowercase version may be longer than the original character.
2004 The first character of the lowercased version is returned
2005 (but note, as explained above, that there may be more.)
2010 Perl_to_utf8_lower(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
2014 PERL_ARGS_ASSERT_TO_UTF8_LOWER;
2016 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
2017 &PL_utf8_tolower, "ToLower", "utf8::ToSpecLower");
2021 =for apidoc to_utf8_fold
2023 Convert the UTF-8 encoded character at p to its foldcase version and
2024 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
2025 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
2026 foldcase version may be longer than the original character (up to
2029 The first character of the foldcased version is returned
2030 (but note, as explained above, that there may be more.)
2034 /* Not currently externally documented is 'flags', which currently is non-zero
2035 * if full case folds are to be used; otherwise simple folds */
2038 Perl__to_utf8_fold_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, U8 flags)
2040 const char *specials = (flags) ? "utf8::ToSpecFold" : NULL;
2044 PERL_ARGS_ASSERT__TO_UTF8_FOLD_FLAGS;
2046 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
2047 &PL_utf8_tofold, "ToFold", specials);
2051 * A "swash" is a swatch hash.
2052 * A "swatch" is a bit vector generated by utf8.c:S_swash_get().
2053 * C<pkg> is a pointer to a package name for SWASHNEW, should be "utf8".
2054 * For other parameters, see utf8::SWASHNEW in lib/utf8_heavy.pl.
2057 Perl_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv, I32 minbits, I32 none)
2062 const size_t pkg_len = strlen(pkg);
2063 const size_t name_len = strlen(name);
2064 HV * const stash = gv_stashpvn(pkg, pkg_len, 0);
2068 PERL_ARGS_ASSERT_SWASH_INIT;
2070 PUSHSTACKi(PERLSI_MAGIC);
2074 if (PL_parser && PL_parser->error_count)
2075 SAVEI8(PL_parser->error_count), PL_parser->error_count = 0;
2076 method = gv_fetchmeth(stash, "SWASHNEW", 8, -1);
2077 if (!method) { /* demand load utf8 */
2079 errsv_save = newSVsv(ERRSV);
2080 /* It is assumed that callers of this routine are not passing in any
2081 user derived data. */
2082 /* Need to do this after save_re_context() as it will set PL_tainted to
2083 1 while saving $1 etc (see the code after getrx: in Perl_magic_get).
2084 Even line to create errsv_save can turn on PL_tainted. */
2085 SAVEBOOL(PL_tainted);
2087 Perl_load_module(aTHX_ PERL_LOADMOD_NOIMPORT, newSVpvn(pkg,pkg_len),
2090 sv_setsv(ERRSV, errsv_save);
2091 SvREFCNT_dec(errsv_save);
2097 mPUSHp(pkg, pkg_len);
2098 mPUSHp(name, name_len);
2103 errsv_save = newSVsv(ERRSV);
2104 /* If we already have a pointer to the method, no need to use call_method()
2105 to repeat the lookup. */
2106 if (method ? call_sv(MUTABLE_SV(method), G_SCALAR)
2107 : call_sv(newSVpvs_flags("SWASHNEW", SVs_TEMP), G_SCALAR | G_METHOD))
2108 retval = newSVsv(*PL_stack_sp--);
2110 retval = &PL_sv_undef;
2112 sv_setsv(ERRSV, errsv_save);
2113 SvREFCNT_dec(errsv_save);
2116 if (IN_PERL_COMPILETIME) {
2117 CopHINTS_set(PL_curcop, PL_hints);
2119 if (!SvROK(retval) || SvTYPE(SvRV(retval)) != SVt_PVHV) {
2121 Perl_croak(aTHX_ "Can't find Unicode property definition \"%"SVf"\"",
2123 Perl_croak(aTHX_ "SWASHNEW didn't return an HV ref");
2129 /* This API is wrong for special case conversions since we may need to
2130 * return several Unicode characters for a single Unicode character
2131 * (see lib/unicore/SpecCase.txt) The SWASHGET in lib/utf8_heavy.pl is
2132 * the lower-level routine, and it is similarly broken for returning
2133 * multiple values. --jhi
2134 * For those, you should use to_utf8_case() instead */
2135 /* Now SWASHGET is recasted into S_swash_get in this file. */
2138 * Returns the value of property/mapping C<swash> for the first character
2139 * of the string C<ptr>. If C<do_utf8> is true, the string C<ptr> is
2140 * assumed to be in utf8. If C<do_utf8> is false, the string C<ptr> is
2141 * assumed to be in native 8-bit encoding. Caches the swatch in C<swash>.
2144 Perl_swash_fetch(pTHX_ SV *swash, const U8 *ptr, bool do_utf8)
2147 HV *const hv = MUTABLE_HV(SvRV(swash));
2152 const U8 *tmps = NULL;
2156 const UV c = NATIVE_TO_ASCII(*ptr);
2158 PERL_ARGS_ASSERT_SWASH_FETCH;
2160 if (!do_utf8 && !UNI_IS_INVARIANT(c)) {
2161 tmputf8[0] = (U8)UTF8_EIGHT_BIT_HI(c);
2162 tmputf8[1] = (U8)UTF8_EIGHT_BIT_LO(c);
2165 /* Given a UTF-X encoded char 0xAA..0xYY,0xZZ
2166 * then the "swatch" is a vec() for all the chars which start
2168 * So the key in the hash (klen) is length of encoded char -1
2170 klen = UTF8SKIP(ptr) - 1;
2174 /* If char is invariant then swatch is for all the invariant chars
2175 * In both UTF-8 and UTF-8-MOD that happens to be UTF_CONTINUATION_MARK
2177 needents = UTF_CONTINUATION_MARK;
2178 off = NATIVE_TO_UTF(ptr[klen]);
2181 /* If char is encoded then swatch is for the prefix */
2182 needents = (1 << UTF_ACCUMULATION_SHIFT);
2183 off = NATIVE_TO_UTF(ptr[klen]) & UTF_CONTINUATION_MASK;
2184 if (UTF8_IS_SUPER(ptr) && ckWARN_d(WARN_NON_UNICODE)) {
2185 const UV code_point = utf8n_to_uvuni(ptr, UTF8_MAXBYTES, 0, 0);
2187 /* This outputs warnings for binary properties only, assuming that
2188 * to_utf8_case() will output any for non-binary. Also, surrogates
2189 * aren't checked for, as that would warn on things like
2191 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
2192 if (SvUV(*bitssvp) == 1) {
2193 Perl_warner(aTHX_ packWARN(WARN_NON_UNICODE),
2194 "Code point 0x%04"UVXf" is not Unicode, all \\p{} matches fail; all \\P{} matches succeed", code_point);
2200 * This single-entry cache saves about 1/3 of the utf8 overhead in test
2201 * suite. (That is, only 7-8% overall over just a hash cache. Still,
2202 * it's nothing to sniff at.) Pity we usually come through at least
2203 * two function calls to get here...
2205 * NB: this code assumes that swatches are never modified, once generated!
2208 if (hv == PL_last_swash_hv &&
2209 klen == PL_last_swash_klen &&
2210 (!klen || memEQ((char *)ptr, (char *)PL_last_swash_key, klen)) )
2212 tmps = PL_last_swash_tmps;
2213 slen = PL_last_swash_slen;
2216 /* Try our second-level swatch cache, kept in a hash. */
2217 SV** svp = hv_fetch(hv, (const char*)ptr, klen, FALSE);
2219 /* If not cached, generate it via swash_get */
2220 if (!svp || !SvPOK(*svp)
2221 || !(tmps = (const U8*)SvPV_const(*svp, slen))) {
2222 /* We use utf8n_to_uvuni() as we want an index into
2223 Unicode tables, not a native character number.
2225 const UV code_point = utf8n_to_uvuni(ptr, UTF8_MAXBYTES, 0,
2227 0 : UTF8_ALLOW_ANY);
2228 swatch = swash_get(swash,
2229 /* On EBCDIC & ~(0xA0-1) isn't a useful thing to do */
2230 (klen) ? (code_point & ~(needents - 1)) : 0,
2233 if (IN_PERL_COMPILETIME)
2234 CopHINTS_set(PL_curcop, PL_hints);
2236 svp = hv_store(hv, (const char *)ptr, klen, swatch, 0);
2238 if (!svp || !(tmps = (U8*)SvPV(*svp, slen))
2239 || (slen << 3) < needents)
2240 Perl_croak(aTHX_ "panic: swash_fetch got improper swatch");
2243 PL_last_swash_hv = hv;
2244 assert(klen <= sizeof(PL_last_swash_key));
2245 PL_last_swash_klen = (U8)klen;
2246 /* FIXME change interpvar.h? */
2247 PL_last_swash_tmps = (U8 *) tmps;
2248 PL_last_swash_slen = slen;
2250 Copy(ptr, PL_last_swash_key, klen, U8);
2253 switch ((int)((slen << 3) / needents)) {
2255 bit = 1 << (off & 7);
2257 return (tmps[off] & bit) != 0;
2262 return (tmps[off] << 8) + tmps[off + 1] ;
2265 return (tmps[off] << 24) + (tmps[off+1] << 16) + (tmps[off+2] << 8) + tmps[off + 3] ;
2267 Perl_croak(aTHX_ "panic: swash_fetch got swatch of unexpected bit width");
2268 NORETURN_FUNCTION_END;
2271 /* Read a single line of the main body of the swash input text. These are of
2274 * where each number is hex. The first two numbers form the minimum and
2275 * maximum of a range, and the third is the value associated with the range.
2276 * Not all swashes should have a third number
2278 * On input: l points to the beginning of the line to be examined; it points
2279 * to somewhere in the string of the whole input text, and is
2280 * terminated by a \n or the null string terminator.
2281 * lend points to the null terminator of that string
2282 * wants_value is non-zero if the swash expects a third number
2283 * typestr is the name of the swash's mapping, like 'ToLower'
2284 * On output: *min, *max, and *val are set to the values read from the line.
2285 * returns a pointer just beyond the line examined. If there was no
2286 * valid min number on the line, returns lend+1
2290 S_swash_scan_list_line(pTHX_ U8* l, U8* const lend, UV* min, UV* max, UV* val,
2291 const bool wants_value, const U8* const typestr)
2293 const int typeto = typestr[0] == 'T' && typestr[1] == 'o';
2294 STRLEN numlen; /* Length of the number */
2295 I32 flags = PERL_SCAN_SILENT_ILLDIGIT
2296 | PERL_SCAN_DISALLOW_PREFIX
2297 | PERL_SCAN_SILENT_NON_PORTABLE;
2299 /* nl points to the next \n in the scan */
2300 U8* const nl = (U8*)memchr(l, '\n', lend - l);
2302 /* Get the first number on the line: the range minimum */
2304 *min = grok_hex((char *)l, &numlen, &flags, NULL);
2305 if (numlen) /* If found a hex number, position past it */
2307 else if (nl) { /* Else, go handle next line, if any */
2308 return nl + 1; /* 1 is length of "\n" */
2310 else { /* Else, no next line */
2311 return lend + 1; /* to LIST's end at which \n is not found */
2314 /* The max range value follows, separated by a BLANK */
2317 flags = PERL_SCAN_SILENT_ILLDIGIT
2318 | PERL_SCAN_DISALLOW_PREFIX
2319 | PERL_SCAN_SILENT_NON_PORTABLE;
2321 *max = grok_hex((char *)l, &numlen, &flags, NULL);
2324 else /* If no value here, it is a single element range */
2327 /* Non-binary tables have a third entry: what the first element of the
2332 flags = PERL_SCAN_SILENT_ILLDIGIT
2333 | PERL_SCAN_DISALLOW_PREFIX
2334 | PERL_SCAN_SILENT_NON_PORTABLE;
2336 *val = grok_hex((char *)l, &numlen, &flags, NULL);
2345 Perl_croak(aTHX_ "%s: illegal mapping '%s'",
2351 *val = 0; /* bits == 1, then any val should be ignored */
2353 else { /* Nothing following range min, should be single element with no
2359 Perl_croak(aTHX_ "%s: illegal mapping '%s'", typestr, l);
2363 *val = 0; /* bits == 1, then val should be ignored */
2366 /* Position to next line if any, or EOF */
2376 * Returns a swatch (a bit vector string) for a code point sequence
2377 * that starts from the value C<start> and comprises the number C<span>.
2378 * A C<swash> must be an object created by SWASHNEW (see lib/utf8_heavy.pl).
2379 * Should be used via swash_fetch, which will cache the swatch in C<swash>.
2382 S_swash_get(pTHX_ SV* swash, UV start, UV span)
2385 U8 *l, *lend, *x, *xend, *s, *send;
2386 STRLEN lcur, xcur, scur;
2387 HV *const hv = MUTABLE_HV(SvRV(swash));
2389 /* The string containing the main body of the table */
2390 SV** const listsvp = hv_fetchs(hv, "LIST", FALSE);
2392 SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE);
2393 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
2394 SV** const nonesvp = hv_fetchs(hv, "NONE", FALSE);
2395 SV** const extssvp = hv_fetchs(hv, "EXTRAS", FALSE);
2396 SV** const invert_it_svp = hv_fetchs(hv, "INVERT_IT", FALSE);
2397 const U8* const typestr = (U8*)SvPV_nolen(*typesvp);
2398 const STRLEN bits = SvUV(*bitssvp);
2399 const STRLEN octets = bits >> 3; /* if bits == 1, then octets == 0 */
2400 const UV none = SvUV(*nonesvp);
2401 const UV end = start + span;
2403 PERL_ARGS_ASSERT_SWASH_GET;
2405 if (bits != 1 && bits != 8 && bits != 16 && bits != 32) {
2406 Perl_croak(aTHX_ "panic: swash_get doesn't expect bits %"UVuf,
2410 /* create and initialize $swatch */
2411 scur = octets ? (span * octets) : (span + 7) / 8;
2412 swatch = newSV(scur);
2414 s = (U8*)SvPVX(swatch);
2415 if (octets && none) {
2416 const U8* const e = s + scur;
2419 *s++ = (U8)(none & 0xff);
2420 else if (bits == 16) {
2421 *s++ = (U8)((none >> 8) & 0xff);
2422 *s++ = (U8)( none & 0xff);
2424 else if (bits == 32) {
2425 *s++ = (U8)((none >> 24) & 0xff);
2426 *s++ = (U8)((none >> 16) & 0xff);
2427 *s++ = (U8)((none >> 8) & 0xff);
2428 *s++ = (U8)( none & 0xff);
2434 (void)memzero((U8*)s, scur + 1);
2436 SvCUR_set(swatch, scur);
2437 s = (U8*)SvPVX(swatch);
2439 /* read $swash->{LIST} */
2440 l = (U8*)SvPV(*listsvp, lcur);
2444 l = S_swash_scan_list_line(aTHX_ l, lend, &min, &max, &val,
2445 cBOOL(octets), typestr);
2450 /* If looking for something beyond this range, go try the next one */
2457 if (!none || val < none) {
2462 for (key = min; key <= max; key++) {
2466 /* offset must be non-negative (start <= min <= key < end) */
2467 offset = octets * (key - start);
2469 s[offset] = (U8)(val & 0xff);
2470 else if (bits == 16) {
2471 s[offset ] = (U8)((val >> 8) & 0xff);
2472 s[offset + 1] = (U8)( val & 0xff);
2474 else if (bits == 32) {
2475 s[offset ] = (U8)((val >> 24) & 0xff);
2476 s[offset + 1] = (U8)((val >> 16) & 0xff);
2477 s[offset + 2] = (U8)((val >> 8) & 0xff);
2478 s[offset + 3] = (U8)( val & 0xff);
2481 if (!none || val < none)
2485 else { /* bits == 1, then val should be ignored */
2489 for (key = min; key <= max; key++) {
2490 const STRLEN offset = (STRLEN)(key - start);
2493 s[offset >> 3] |= 1 << (offset & 7);
2499 /* Invert if the data says it should be. Assumes that bits == 1 */
2500 if (invert_it_svp && SvUV(*invert_it_svp)) {
2502 /* Unicode properties should come with all bits above PERL_UNICODE_MAX
2503 * be 0, and their inversion should also be 0, as we don't succeed any
2504 * Unicode property matches for non-Unicode code points */
2505 if (start <= PERL_UNICODE_MAX) {
2507 /* The code below assumes that we never cross the
2508 * Unicode/above-Unicode boundary in a range, as otherwise we would
2509 * have to figure out where to stop flipping the bits. Since this
2510 * boundary is divisible by a large power of 2, and swatches comes
2511 * in small powers of 2, this should be a valid assumption */
2512 assert(start + span - 1 <= PERL_UNICODE_MAX);
2522 /* read $swash->{EXTRAS}
2523 * This code also copied to swash_to_invlist() below */
2524 x = (U8*)SvPV(*extssvp, xcur);
2532 SV **otherbitssvp, *other;
2536 const U8 opc = *x++;
2540 nl = (U8*)memchr(x, '\n', xend - x);
2542 if (opc != '-' && opc != '+' && opc != '!' && opc != '&') {
2544 x = nl + 1; /* 1 is length of "\n" */
2548 x = xend; /* to EXTRAS' end at which \n is not found */
2555 namelen = nl - namestr;
2559 namelen = xend - namestr;
2563 othersvp = hv_fetch(hv, (char *)namestr, namelen, FALSE);
2564 otherhv = MUTABLE_HV(SvRV(*othersvp));
2565 otherbitssvp = hv_fetchs(otherhv, "BITS", FALSE);
2566 otherbits = (STRLEN)SvUV(*otherbitssvp);
2567 if (bits < otherbits)
2568 Perl_croak(aTHX_ "panic: swash_get found swatch size mismatch");
2570 /* The "other" swatch must be destroyed after. */
2571 other = swash_get(*othersvp, start, span);
2572 o = (U8*)SvPV(other, olen);
2575 Perl_croak(aTHX_ "panic: swash_get got improper swatch");
2577 s = (U8*)SvPV(swatch, slen);
2578 if (bits == 1 && otherbits == 1) {
2580 Perl_croak(aTHX_ "panic: swash_get found swatch length mismatch");
2604 STRLEN otheroctets = otherbits >> 3;
2606 U8* const send = s + slen;
2611 if (otherbits == 1) {
2612 otherval = (o[offset >> 3] >> (offset & 7)) & 1;
2616 STRLEN vlen = otheroctets;
2624 if (opc == '+' && otherval)
2625 NOOP; /* replace with otherval */
2626 else if (opc == '!' && !otherval)
2628 else if (opc == '-' && otherval)
2630 else if (opc == '&' && !otherval)
2633 s += octets; /* no replacement */
2638 *s++ = (U8)( otherval & 0xff);
2639 else if (bits == 16) {
2640 *s++ = (U8)((otherval >> 8) & 0xff);
2641 *s++ = (U8)( otherval & 0xff);
2643 else if (bits == 32) {
2644 *s++ = (U8)((otherval >> 24) & 0xff);
2645 *s++ = (U8)((otherval >> 16) & 0xff);
2646 *s++ = (U8)((otherval >> 8) & 0xff);
2647 *s++ = (U8)( otherval & 0xff);
2651 sv_free(other); /* through with it! */
2657 Perl__swash_inversion_hash(pTHX_ SV* const swash)
2660 /* Subject to change or removal. For use only in one place in regcomp.c.
2661 * Can't be used on a property that is subject to user override, as it
2662 * relies on the value of SPECIALS in the swash which would be set by
2663 * utf8_heavy.pl to the hash in the non-overriden file, and hence is not set
2664 * for overridden properties
2666 * Returns a hash which is the inversion and closure of a swash mapping.
2667 * For example, consider the input lines:
2672 * The returned hash would have two keys, the utf8 for 006B and the utf8 for
2673 * 006C. The value for each key is an array. For 006C, the array would
2674 * have a two elements, the utf8 for itself, and for 004C. For 006B, there
2675 * would be three elements in its array, the utf8 for 006B, 004B and 212A.
2677 * Essentially, for any code point, it gives all the code points that map to
2678 * it, or the list of 'froms' for that point.
2680 * Currently it ignores any additions or deletions from other swashes,
2681 * looking at just the main body of the swash, and if there are SPECIALS
2682 * in the swash, at that hash
2684 * The specials hash can be extra code points, and most likely consists of
2685 * maps from single code points to multiple ones (each expressed as a string
2686 * of utf8 characters). This function currently returns only 1-1 mappings.
2687 * However consider this possible input in the specials hash:
2688 * "\xEF\xAC\x85" => "\x{0073}\x{0074}", # U+FB05 => 0073 0074
2689 * "\xEF\xAC\x86" => "\x{0073}\x{0074}", # U+FB06 => 0073 0074
2691 * Both FB05 and FB06 map to the same multi-char sequence, which we don't
2692 * currently handle. But it also means that FB05 and FB06 are equivalent in
2693 * a 1-1 mapping which we should handle, and this relationship may not be in
2694 * the main table. Therefore this function examines all the multi-char
2695 * sequences and adds the 1-1 mappings that come out of that. */
2699 HV *const hv = MUTABLE_HV(SvRV(swash));
2701 /* The string containing the main body of the table */
2702 SV** const listsvp = hv_fetchs(hv, "LIST", FALSE);
2704 SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE);
2705 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
2706 SV** const nonesvp = hv_fetchs(hv, "NONE", FALSE);
2707 /*SV** const extssvp = hv_fetchs(hv, "EXTRAS", FALSE);*/
2708 const U8* const typestr = (U8*)SvPV_nolen(*typesvp);
2709 const STRLEN bits = SvUV(*bitssvp);
2710 const STRLEN octets = bits >> 3; /* if bits == 1, then octets == 0 */
2711 const UV none = SvUV(*nonesvp);
2712 SV **specials_p = hv_fetchs(hv, "SPECIALS", 0);
2716 PERL_ARGS_ASSERT__SWASH_INVERSION_HASH;
2718 /* Must have at least 8 bits to get the mappings */
2719 if (bits != 8 && bits != 16 && bits != 32) {
2720 Perl_croak(aTHX_ "panic: swash_inversion_hash doesn't expect bits %"UVuf,
2724 if (specials_p) { /* It might be "special" (sometimes, but not always, a
2725 mapping to more than one character */
2727 /* Construct an inverse mapping hash for the specials */
2728 HV * const specials_hv = MUTABLE_HV(SvRV(*specials_p));
2729 HV * specials_inverse = newHV();
2730 char *char_from; /* the lhs of the map */
2731 I32 from_len; /* its byte length */
2732 char *char_to; /* the rhs of the map */
2733 I32 to_len; /* its byte length */
2734 SV *sv_to; /* and in a sv */
2735 AV* from_list; /* list of things that map to each 'to' */
2737 hv_iterinit(specials_hv);
2739 /* The keys are the characters (in utf8) that map to the corresponding
2740 * utf8 string value. Iterate through the list creating the inverse
2742 while ((sv_to = hv_iternextsv(specials_hv, &char_from, &from_len))) {
2744 if (! SvPOK(sv_to)) {
2745 Perl_croak(aTHX_ "panic: value returned from hv_iternextsv() unexpectedly is not a string");
2747 /*DEBUG_U(PerlIO_printf(Perl_debug_log, "Found mapping from %"UVXf", First char of to is %"UVXf"\n", utf8_to_uvchr((U8*) char_from, 0), utf8_to_uvchr((U8*) SvPVX(sv_to), 0)));*/
2749 /* Each key in the inverse list is a mapped-to value, and the key's
2750 * hash value is a list of the strings (each in utf8) that map to
2751 * it. Those strings are all one character long */
2752 if ((listp = hv_fetch(specials_inverse,
2756 from_list = (AV*) *listp;
2758 else { /* No entry yet for it: create one */
2759 from_list = newAV();
2760 if (! hv_store(specials_inverse,
2763 (SV*) from_list, 0))
2765 Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed");
2769 /* Here have the list associated with this 'to' (perhaps newly
2770 * created and empty). Just add to it. Note that we ASSUME that
2771 * the input is guaranteed to not have duplications, so we don't
2772 * check for that. Duplications just slow down execution time. */
2773 av_push(from_list, newSVpvn_utf8(char_from, from_len, TRUE));
2776 /* Here, 'specials_inverse' contains the inverse mapping. Go through
2777 * it looking for cases like the FB05/FB06 examples above. There would
2778 * be an entry in the hash like
2779 * 'st' => [ FB05, FB06 ]
2780 * In this example we will create two lists that get stored in the
2781 * returned hash, 'ret':
2782 * FB05 => [ FB05, FB06 ]
2783 * FB06 => [ FB05, FB06 ]
2785 * Note that there is nothing to do if the array only has one element.
2786 * (In the normal 1-1 case handled below, we don't have to worry about
2787 * two lists, as everything gets tied to the single list that is
2788 * generated for the single character 'to'. But here, we are omitting
2789 * that list, ('st' in the example), so must have multiple lists.) */
2790 while ((from_list = (AV *) hv_iternextsv(specials_inverse,
2791 &char_to, &to_len)))
2793 if (av_len(from_list) > 0) {
2796 /* We iterate over all combinations of i,j to place each code
2797 * point on each list */
2798 for (i = 0; i <= av_len(from_list); i++) {
2800 AV* i_list = newAV();
2801 SV** entryp = av_fetch(from_list, i, FALSE);
2802 if (entryp == NULL) {
2803 Perl_croak(aTHX_ "panic: av_fetch() unexpectedly failed");
2805 if (hv_fetch(ret, SvPVX(*entryp), SvCUR(*entryp), FALSE)) {
2806 Perl_croak(aTHX_ "panic: unexpected entry for %s", SvPVX(*entryp));
2808 if (! hv_store(ret, SvPVX(*entryp), SvCUR(*entryp),
2809 (SV*) i_list, FALSE))
2811 Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed");
2814 /* For debugging: UV u = utf8_to_uvchr((U8*) SvPVX(*entryp), 0);*/
2815 for (j = 0; j <= av_len(from_list); j++) {
2816 entryp = av_fetch(from_list, j, FALSE);
2817 if (entryp == NULL) {
2818 Perl_croak(aTHX_ "panic: av_fetch() unexpectedly failed");
2821 /* When i==j this adds itself to the list */
2822 av_push(i_list, newSVuv(utf8_to_uvchr(
2823 (U8*) SvPVX(*entryp), 0)));
2824 /*DEBUG_U(PerlIO_printf(Perl_debug_log, "Adding %"UVXf" to list for %"UVXf"\n", utf8_to_uvchr((U8*) SvPVX(*entryp), 0), u));*/
2829 SvREFCNT_dec(specials_inverse); /* done with it */
2830 } /* End of specials */
2832 /* read $swash->{LIST} */
2833 l = (U8*)SvPV(*listsvp, lcur);
2836 /* Go through each input line */
2840 l = S_swash_scan_list_line(aTHX_ l, lend, &min, &max, &val,
2841 cBOOL(octets), typestr);
2846 /* Each element in the range is to be inverted */
2847 for (inverse = min; inverse <= max; inverse++) {
2851 bool found_key = FALSE;
2852 bool found_inverse = FALSE;
2854 /* The key is the inverse mapping */
2855 char key[UTF8_MAXBYTES+1];
2856 char* key_end = (char *) uvuni_to_utf8((U8*) key, val);
2857 STRLEN key_len = key_end - key;
2859 /* Get the list for the map */
2860 if ((listp = hv_fetch(ret, key, key_len, FALSE))) {
2861 list = (AV*) *listp;
2863 else { /* No entry yet for it: create one */
2865 if (! hv_store(ret, key, key_len, (SV*) list, FALSE)) {
2866 Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed");
2870 /* Look through list to see if this inverse mapping already is
2871 * listed, or if there is a mapping to itself already */
2872 for (i = 0; i <= av_len(list); i++) {
2873 SV** entryp = av_fetch(list, i, FALSE);
2875 if (entryp == NULL) {
2876 Perl_croak(aTHX_ "panic: av_fetch() unexpectedly failed");
2879 /*DEBUG_U(PerlIO_printf(Perl_debug_log, "list for %"UVXf" contains %"UVXf"\n", val, SvUV(entry)));*/
2880 if (SvUV(entry) == val) {
2883 if (SvUV(entry) == inverse) {
2884 found_inverse = TRUE;
2887 /* No need to continue searching if found everything we are
2889 if (found_key && found_inverse) {
2894 /* Make sure there is a mapping to itself on the list */
2896 av_push(list, newSVuv(val));
2897 /*DEBUG_U(PerlIO_printf(Perl_debug_log, "Adding %"UVXf" to list for %"UVXf"\n", val, val));*/
2901 /* Simply add the value to the list */
2902 if (! found_inverse) {
2903 av_push(list, newSVuv(inverse));
2904 /*DEBUG_U(PerlIO_printf(Perl_debug_log, "Adding %"UVXf" to list for %"UVXf"\n", inverse, val));*/
2907 /* swash_get() increments the value of val for each element in the
2908 * range. That makes more compact tables possible. You can
2909 * express the capitalization, for example, of all consecutive
2910 * letters with a single line: 0061\t007A\t0041 This maps 0061 to
2911 * 0041, 0062 to 0042, etc. I (khw) have never understood 'none',
2912 * and it's not documented; it appears to be used only in
2913 * implementing tr//; I copied the semantics from swash_get(), just
2915 if (!none || val < none) {
2925 Perl__swash_to_invlist(pTHX_ SV* const swash)
2928 /* Subject to change or removal. For use only in one place in regcomp.c */
2933 HV *const hv = MUTABLE_HV(SvRV(swash));
2934 UV elements = 0; /* Number of elements in the inversion list */
2937 /* The string containing the main body of the table */
2938 SV** const listsvp = hv_fetchs(hv, "LIST", FALSE);
2939 SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE);
2940 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
2941 SV** const extssvp = hv_fetchs(hv, "EXTRAS", FALSE);
2942 SV** const invert_it_svp = hv_fetchs(hv, "INVERT_IT", FALSE);
2944 const U8* const typestr = (U8*)SvPV_nolen(*typesvp);
2945 const STRLEN bits = SvUV(*bitssvp);
2946 const STRLEN octets = bits >> 3; /* if bits == 1, then octets == 0 */
2952 PERL_ARGS_ASSERT__SWASH_TO_INVLIST;
2954 /* read $swash->{LIST} */
2955 if (SvPOK(*listsvp)) {
2956 l = (U8*)SvPV(*listsvp, lcur);
2959 /* LIST legitimately doesn't contain a string during compilation phases
2960 * of Perl itself, before the Unicode tables are generated. In this
2961 * case, just fake things up by creating an empty list */
2968 /* Scan the input to count the number of lines to preallocate array size
2969 * based on worst possible case, which is each line in the input creates 2
2970 * elements in the inversion list: 1) the beginning of a range in the list;
2971 * 2) the beginning of a range not in the list. */
2972 while ((loc = (strchr(loc, '\n'))) != NULL) {
2977 /* If the ending is somehow corrupt and isn't a new line, add another
2978 * element for the final range that isn't in the inversion list */
2979 if (! (*lend == '\n' || (*lend == '\0' && *(lend - 1) == '\n'))) {
2983 invlist = _new_invlist(elements);
2985 /* Now go through the input again, adding each range to the list */
2988 UV val; /* Not used by this function */
2990 l = S_swash_scan_list_line(aTHX_ l, lend, &start, &end, &val,
2991 cBOOL(octets), typestr);
2997 _append_range_to_invlist(invlist, start, end);
3000 /* Invert if the data says it should be */
3001 if (invert_it_svp && SvUV(*invert_it_svp)) {
3002 _invlist_invert_prop(invlist);
3005 /* This code is copied from swash_get()
3006 * read $swash->{EXTRAS} */
3007 x = (U8*)SvPV(*extssvp, xcur);
3015 SV **otherbitssvp, *other;
3018 const U8 opc = *x++;
3022 nl = (U8*)memchr(x, '\n', xend - x);
3024 if (opc != '-' && opc != '+' && opc != '!' && opc != '&') {
3026 x = nl + 1; /* 1 is length of "\n" */
3030 x = xend; /* to EXTRAS' end at which \n is not found */
3037 namelen = nl - namestr;
3041 namelen = xend - namestr;
3045 othersvp = hv_fetch(hv, (char *)namestr, namelen, FALSE);
3046 otherhv = MUTABLE_HV(SvRV(*othersvp));
3047 otherbitssvp = hv_fetchs(otherhv, "BITS", FALSE);
3048 otherbits = (STRLEN)SvUV(*otherbitssvp);
3050 if (bits != otherbits || bits != 1) {
3051 Perl_croak(aTHX_ "panic: _swash_to_invlist only operates on boolean properties");
3054 /* The "other" swatch must be destroyed after. */
3055 other = _swash_to_invlist((SV *)*othersvp);
3057 /* End of code copied from swash_get() */
3060 _invlist_union(invlist, other, &invlist);
3063 _invlist_invert(other);
3064 _invlist_union(invlist, other, &invlist);
3067 _invlist_subtract(invlist, other, &invlist);
3070 _invlist_intersection(invlist, other, &invlist);
3075 sv_free(other); /* through with it! */
3082 =for apidoc uvchr_to_utf8
3084 Adds the UTF-8 representation of the Native code point C<uv> to the end
3085 of the string C<d>; C<d> should be have at least C<UTF8_MAXBYTES+1> free
3086 bytes available. The return value is the pointer to the byte after the
3087 end of the new character. In other words,
3089 d = uvchr_to_utf8(d, uv);
3091 is the recommended wide native character-aware way of saying
3098 /* On ASCII machines this is normally a macro but we want a
3099 real function in case XS code wants it
3102 Perl_uvchr_to_utf8(pTHX_ U8 *d, UV uv)
3104 PERL_ARGS_ASSERT_UVCHR_TO_UTF8;
3106 return Perl_uvuni_to_utf8_flags(aTHX_ d, NATIVE_TO_UNI(uv), 0);
3110 Perl_uvchr_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags)
3112 PERL_ARGS_ASSERT_UVCHR_TO_UTF8_FLAGS;
3114 return Perl_uvuni_to_utf8_flags(aTHX_ d, NATIVE_TO_UNI(uv), flags);
3118 =for apidoc utf8n_to_uvchr
3120 Returns the native character value of the first character in the string
3122 which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
3123 length, in bytes, of that character.
3125 length and flags are the same as utf8n_to_uvuni().
3129 /* On ASCII machines this is normally a macro but we want
3130 a real function in case XS code wants it
3133 Perl_utf8n_to_uvchr(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen,
3136 const UV uv = Perl_utf8n_to_uvuni(aTHX_ s, curlen, retlen, flags);
3138 PERL_ARGS_ASSERT_UTF8N_TO_UVCHR;
3140 return UNI_TO_NATIVE(uv);
3144 Perl_check_utf8_print(pTHX_ register const U8* s, const STRLEN len)
3146 /* May change: warns if surrogates, non-character code points, or
3147 * non-Unicode code points are in s which has length len. Returns TRUE if
3148 * none found; FALSE otherwise. The only other validity check is to make
3149 * sure that this won't exceed the string's length */
3151 const U8* const e = s + len;
3154 PERL_ARGS_ASSERT_CHECK_UTF8_PRINT;
3157 if (UTF8SKIP(s) > len) {
3158 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8),
3159 "%s in %s", unees, PL_op ? OP_DESC(PL_op) : "print");
3162 if (*s >= UTF8_FIRST_PROBLEMATIC_CODE_POINT_FIRST_BYTE) {
3164 if (UTF8_IS_SUPER(s)) {
3165 if (ckWARN_d(WARN_NON_UNICODE)) {
3166 UV uv = utf8_to_uvchr(s, &char_len);
3167 Perl_warner(aTHX_ packWARN(WARN_NON_UNICODE),
3168 "Code point 0x%04"UVXf" is not Unicode, may not be portable", uv);
3172 else if (UTF8_IS_SURROGATE(s)) {
3173 if (ckWARN_d(WARN_SURROGATE)) {
3174 UV uv = utf8_to_uvchr(s, &char_len);
3175 Perl_warner(aTHX_ packWARN(WARN_SURROGATE),
3176 "Unicode surrogate U+%04"UVXf" is illegal in UTF-8", uv);
3181 ((UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s))
3182 && (ckWARN_d(WARN_NONCHAR)))
3184 UV uv = utf8_to_uvchr(s, &char_len);
3185 Perl_warner(aTHX_ packWARN(WARN_NONCHAR),
3186 "Unicode non-character U+%04"UVXf" is illegal for open interchange", uv);
3197 =for apidoc pv_uni_display
3199 Build to the scalar dsv a displayable version of the string spv,
3200 length len, the displayable version being at most pvlim bytes long
3201 (if longer, the rest is truncated and "..." will be appended).
3203 The flags argument can have UNI_DISPLAY_ISPRINT set to display
3204 isPRINT()able characters as themselves, UNI_DISPLAY_BACKSLASH
3205 to display the \\[nrfta\\] as the backslashed versions (like '\n')
3206 (UNI_DISPLAY_BACKSLASH is preferred over UNI_DISPLAY_ISPRINT for \\).
3207 UNI_DISPLAY_QQ (and its alias UNI_DISPLAY_REGEX) have both
3208 UNI_DISPLAY_BACKSLASH and UNI_DISPLAY_ISPRINT turned on.
3210 The pointer to the PV of the dsv is returned.
3214 Perl_pv_uni_display(pTHX_ SV *dsv, const U8 *spv, STRLEN len, STRLEN pvlim, UV flags)
3219 PERL_ARGS_ASSERT_PV_UNI_DISPLAY;
3223 for (s = (const char *)spv, e = s + len; s < e; s += UTF8SKIP(s)) {
3225 /* This serves double duty as a flag and a character to print after
3226 a \ when flags & UNI_DISPLAY_BACKSLASH is true.
3230 if (pvlim && SvCUR(dsv) >= pvlim) {
3234 u = utf8_to_uvchr((U8*)s, 0);
3236 const unsigned char c = (unsigned char)u & 0xFF;
3237 if (flags & UNI_DISPLAY_BACKSLASH) {
3254 const char string = ok;
3255 sv_catpvs(dsv, "\\");
3256 sv_catpvn(dsv, &string, 1);
3259 /* isPRINT() is the locale-blind version. */
3260 if (!ok && (flags & UNI_DISPLAY_ISPRINT) && isPRINT(c)) {
3261 const char string = c;
3262 sv_catpvn(dsv, &string, 1);
3267 Perl_sv_catpvf(aTHX_ dsv, "\\x{%"UVxf"}", u);
3270 sv_catpvs(dsv, "...");
3276 =for apidoc sv_uni_display
3278 Build to the scalar dsv a displayable version of the scalar sv,
3279 the displayable version being at most pvlim bytes long
3280 (if longer, the rest is truncated and "..." will be appended).
3282 The flags argument is as in pv_uni_display().
3284 The pointer to the PV of the dsv is returned.
3289 Perl_sv_uni_display(pTHX_ SV *dsv, SV *ssv, STRLEN pvlim, UV flags)
3291 PERL_ARGS_ASSERT_SV_UNI_DISPLAY;
3293 return Perl_pv_uni_display(aTHX_ dsv, (const U8*)SvPVX_const(ssv),
3294 SvCUR(ssv), pvlim, flags);
3298 =for apidoc foldEQ_utf8
3300 Returns true if the leading portions of the strings s1 and s2 (either or both
3301 of which may be in UTF-8) are the same case-insensitively; false otherwise.
3302 How far into the strings to compare is determined by other input parameters.
3304 If u1 is true, the string s1 is assumed to be in UTF-8-encoded Unicode;
3305 otherwise it is assumed to be in native 8-bit encoding. Correspondingly for u2
3308 If the byte length l1 is non-zero, it says how far into s1 to check for fold
3309 equality. In other words, s1+l1 will be used as a goal to reach. The
3310 scan will not be considered to be a match unless the goal is reached, and
3311 scanning won't continue past that goal. Correspondingly for l2 with respect to
3314 If pe1 is non-NULL and the pointer it points to is not NULL, that pointer is
3315 considered an end pointer beyond which scanning of s1 will not continue under
3316 any circumstances. This means that if both l1 and pe1 are specified, and pe1
3317 is less than s1+l1, the match will never be successful because it can never
3318 get as far as its goal (and in fact is asserted against). Correspondingly for
3319 pe2 with respect to s2.
3321 At least one of s1 and s2 must have a goal (at least one of l1 and l2 must be
3322 non-zero), and if both do, both have to be
3323 reached for a successful match. Also, if the fold of a character is multiple
3324 characters, all of them must be matched (see tr21 reference below for
3327 Upon a successful match, if pe1 is non-NULL,
3328 it will be set to point to the beginning of the I<next> character of s1 beyond
3329 what was matched. Correspondingly for pe2 and s2.
3331 For case-insensitiveness, the "casefolding" of Unicode is used
3332 instead of upper/lowercasing both the characters, see
3333 http://www.unicode.org/unicode/reports/tr21/ (Case Mappings).
3337 /* A flags parameter has been added which may change, and hence isn't
3338 * externally documented. Currently it is:
3339 * 0 for as-documented above
3340 * FOLDEQ_UTF8_NOMIX_ASCII meaning that if a non-ASCII character folds to an
3341 ASCII one, to not match
3342 * FOLDEQ_UTF8_LOCALE meaning that locale rules are to be used for code
3343 * points below 256; unicode rules for above 255; and
3344 * folds that cross those boundaries are disallowed,
3345 * like the NOMIX_ASCII option
3348 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)
3351 register const U8 *p1 = (const U8*)s1; /* Point to current char */
3352 register const U8 *p2 = (const U8*)s2;
3353 register const U8 *g1 = NULL; /* goal for s1 */
3354 register const U8 *g2 = NULL;
3355 register const U8 *e1 = NULL; /* Don't scan s1 past this */
3356 register U8 *f1 = NULL; /* Point to current folded */
3357 register const U8 *e2 = NULL;
3358 register U8 *f2 = NULL;
3359 STRLEN n1 = 0, n2 = 0; /* Number of bytes in current char */
3360 U8 foldbuf1[UTF8_MAXBYTES_CASE+1];
3361 U8 foldbuf2[UTF8_MAXBYTES_CASE+1];
3362 U8 natbuf[2]; /* Holds native 8-bit char converted to utf8;
3363 these always fit in 2 bytes */
3365 PERL_ARGS_ASSERT_FOLDEQ_UTF8_FLAGS;
3372 g1 = (const U8*)s1 + l1;
3380 g2 = (const U8*)s2 + l2;
3383 /* Must have at least one goal */
3388 /* Will never match if goal is out-of-bounds */
3389 assert(! e1 || e1 >= g1);
3391 /* Here, there isn't an end pointer, or it is beyond the goal. We
3392 * only go as far as the goal */
3396 assert(e1); /* Must have an end for looking at s1 */
3399 /* Same for goal for s2 */
3401 assert(! e2 || e2 >= g2);
3408 /* Look through both strings, a character at a time */
3409 while (p1 < e1 && p2 < e2) {
3411 /* If at the beginning of a new character in s1, get its fold to use
3412 * and the length of the fold. (exception: locale rules just get the
3413 * character to a single byte) */
3416 /* If in locale matching, we use two sets of rules, depending on if
3417 * the code point is above or below 255. Here, we test for and
3418 * handle locale rules */
3419 if ((flags & FOLDEQ_UTF8_LOCALE)
3420 && (! u1 || UTF8_IS_INVARIANT(*p1) || UTF8_IS_DOWNGRADEABLE_START(*p1)))
3422 /* There is no mixing of code points above and below 255. */
3423 if (u2 && (! UTF8_IS_INVARIANT(*p2)
3424 && ! UTF8_IS_DOWNGRADEABLE_START(*p2)))
3429 /* We handle locale rules by converting, if necessary, the code
3430 * point to a single byte. */
3431 if (! u1 || UTF8_IS_INVARIANT(*p1)) {
3435 *foldbuf1 = TWO_BYTE_UTF8_TO_UNI(*p1, *(p1 + 1));
3439 else if (isASCII(*p1)) { /* Note, that here won't be both ASCII
3440 and using locale rules */
3442 /* If trying to mix non- with ASCII, and not supposed to, fail */
3443 if ((flags & FOLDEQ_UTF8_NOMIX_ASCII) && ! isASCII(*p2)) {
3447 *foldbuf1 = toLOWER(*p1); /* Folds in the ASCII range are
3451 to_utf8_fold(p1, foldbuf1, &n1);
3453 else { /* Not utf8, convert to it first and then get fold */
3454 uvuni_to_utf8(natbuf, (UV) NATIVE_TO_UNI(((UV)*p1)));
3455 to_utf8_fold(natbuf, foldbuf1, &n1);
3460 if (n2 == 0) { /* Same for s2 */
3461 if ((flags & FOLDEQ_UTF8_LOCALE)
3462 && (! u2 || UTF8_IS_INVARIANT(*p2) || UTF8_IS_DOWNGRADEABLE_START(*p2)))
3464 /* Here, the next char in s2 is < 256. We've already worked on
3465 * s1, and if it isn't also < 256, can't match */
3466 if (u1 && (! UTF8_IS_INVARIANT(*p1)
3467 && ! UTF8_IS_DOWNGRADEABLE_START(*p1)))
3471 if (! u2 || UTF8_IS_INVARIANT(*p2)) {
3475 *foldbuf2 = TWO_BYTE_UTF8_TO_UNI(*p2, *(p2 + 1));
3478 /* Use another function to handle locale rules. We've made
3479 * sure that both characters to compare are single bytes */
3480 if (! foldEQ_locale((char *) f1, (char *) foldbuf2, 1)) {
3485 else if (isASCII(*p2)) {
3486 if (flags && ! isASCII(*p1)) {
3490 *foldbuf2 = toLOWER(*p2);
3493 to_utf8_fold(p2, foldbuf2, &n2);
3496 uvuni_to_utf8(natbuf, (UV) NATIVE_TO_UNI(((UV)*p2)));
3497 to_utf8_fold(natbuf, foldbuf2, &n2);
3502 /* Here f1 and f2 point to the beginning of the strings to compare.
3503 * These strings are the folds of the input characters, stored in utf8.
3506 /* While there is more to look for in both folds, see if they
3507 * continue to match */
3509 U8 fold_length = UTF8SKIP(f1);
3510 if (fold_length != UTF8SKIP(f2)
3511 || (fold_length == 1 && *f1 != *f2) /* Short circuit memNE
3512 function call for single
3514 || memNE((char*)f1, (char*)f2, fold_length))
3516 return 0; /* mismatch */
3519 /* Here, they matched, advance past them */
3526 /* When reach the end of any fold, advance the input past it */
3528 p1 += u1 ? UTF8SKIP(p1) : 1;
3531 p2 += u2 ? UTF8SKIP(p2) : 1;
3533 } /* End of loop through both strings */
3535 /* A match is defined by each scan that specified an explicit length
3536 * reaching its final goal, and the other not having matched a partial
3537 * character (which can happen when the fold of a character is more than one
3539 if (! ((g1 == 0 || p1 == g1) && (g2 == 0 || p2 == g2)) || n1 || n2) {
3543 /* Successful match. Set output pointers */
3555 * c-indentation-style: bsd
3557 * indent-tabs-mode: t
3560 * ex: set ts=8 sts=4 sw=4 noet: