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.
1214 * Note that the macros in handy.h that call these short-circuit calling them
1215 * for Latin-1 range inputs */
1218 Perl_is_uni_alnum(pTHX_ UV c)
1220 U8 tmpbuf[UTF8_MAXBYTES+1];
1221 uvchr_to_utf8(tmpbuf, c);
1222 return is_utf8_alnum(tmpbuf);
1226 Perl_is_uni_idfirst(pTHX_ UV c)
1228 U8 tmpbuf[UTF8_MAXBYTES+1];
1229 uvchr_to_utf8(tmpbuf, c);
1230 return is_utf8_idfirst(tmpbuf);
1234 Perl_is_uni_alpha(pTHX_ UV c)
1236 U8 tmpbuf[UTF8_MAXBYTES+1];
1237 uvchr_to_utf8(tmpbuf, c);
1238 return is_utf8_alpha(tmpbuf);
1242 Perl_is_uni_ascii(pTHX_ UV c)
1248 Perl_is_uni_space(pTHX_ UV c)
1250 U8 tmpbuf[UTF8_MAXBYTES+1];
1251 uvchr_to_utf8(tmpbuf, c);
1252 return is_utf8_space(tmpbuf);
1256 Perl_is_uni_digit(pTHX_ UV c)
1258 U8 tmpbuf[UTF8_MAXBYTES+1];
1259 uvchr_to_utf8(tmpbuf, c);
1260 return is_utf8_digit(tmpbuf);
1264 Perl_is_uni_upper(pTHX_ UV c)
1266 U8 tmpbuf[UTF8_MAXBYTES+1];
1267 uvchr_to_utf8(tmpbuf, c);
1268 return is_utf8_upper(tmpbuf);
1272 Perl_is_uni_lower(pTHX_ UV c)
1274 U8 tmpbuf[UTF8_MAXBYTES+1];
1275 uvchr_to_utf8(tmpbuf, c);
1276 return is_utf8_lower(tmpbuf);
1280 Perl_is_uni_cntrl(pTHX_ UV c)
1282 return isCNTRL_L1(c);
1286 Perl_is_uni_graph(pTHX_ UV c)
1288 U8 tmpbuf[UTF8_MAXBYTES+1];
1289 uvchr_to_utf8(tmpbuf, c);
1290 return is_utf8_graph(tmpbuf);
1294 Perl_is_uni_print(pTHX_ UV c)
1296 U8 tmpbuf[UTF8_MAXBYTES+1];
1297 uvchr_to_utf8(tmpbuf, c);
1298 return is_utf8_print(tmpbuf);
1302 Perl_is_uni_punct(pTHX_ UV c)
1304 U8 tmpbuf[UTF8_MAXBYTES+1];
1305 uvchr_to_utf8(tmpbuf, c);
1306 return is_utf8_punct(tmpbuf);
1310 Perl_is_uni_xdigit(pTHX_ UV c)
1312 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1313 uvchr_to_utf8(tmpbuf, c);
1314 return is_utf8_xdigit(tmpbuf);
1319 Perl_to_uni_upper(pTHX_ UV c, U8* p, STRLEN *lenp)
1321 /* Convert the Unicode character whose ordinal is c to its uppercase
1322 * version and store that in UTF-8 in p and its length in bytes in lenp.
1323 * Note that the p needs to be at least UTF8_MAXBYTES_CASE+1 bytes since
1324 * the changed version may be longer than the original character.
1326 * The ordinal of the first character of the changed version is returned
1327 * (but note, as explained above, that there may be more.) */
1329 PERL_ARGS_ASSERT_TO_UNI_UPPER;
1331 uvchr_to_utf8(p, c);
1332 return to_utf8_upper(p, p, lenp);
1336 Perl_to_uni_title(pTHX_ UV c, U8* p, STRLEN *lenp)
1338 PERL_ARGS_ASSERT_TO_UNI_TITLE;
1340 uvchr_to_utf8(p, c);
1341 return to_utf8_title(p, p, lenp);
1345 Perl_to_uni_lower(pTHX_ UV c, U8* p, STRLEN *lenp)
1347 PERL_ARGS_ASSERT_TO_UNI_LOWER;
1350 uvchr_to_utf8(p, c);
1351 return to_utf8_lower(p, p, lenp);
1354 /* We have the latin1-range values compiled into the core, so just use
1355 * those, converting the result to utf8 */
1356 c = toLOWER_LATIN1(c);
1357 if (UNI_IS_INVARIANT(c)) {
1362 *p = UTF8_TWO_BYTE_HI(c);
1363 *(p+1) = UTF8_TWO_BYTE_LO(c);
1370 Perl__to_uni_fold_flags(pTHX_ UV c, U8* p, STRLEN *lenp, U8 flags)
1372 PERL_ARGS_ASSERT__TO_UNI_FOLD_FLAGS;
1374 uvchr_to_utf8(p, c);
1375 return _to_utf8_fold_flags(p, p, lenp, flags);
1378 /* for now these all assume no locale info available for Unicode > 255 */
1381 Perl_is_uni_alnum_lc(pTHX_ UV c)
1383 return is_uni_alnum(c); /* XXX no locale support yet */
1387 Perl_is_uni_idfirst_lc(pTHX_ UV c)
1389 return is_uni_idfirst(c); /* XXX no locale support yet */
1393 Perl_is_uni_alpha_lc(pTHX_ UV c)
1395 return is_uni_alpha(c); /* XXX no locale support yet */
1399 Perl_is_uni_ascii_lc(pTHX_ UV c)
1401 return is_uni_ascii(c); /* XXX no locale support yet */
1405 Perl_is_uni_space_lc(pTHX_ UV c)
1407 return is_uni_space(c); /* XXX no locale support yet */
1411 Perl_is_uni_digit_lc(pTHX_ UV c)
1413 return is_uni_digit(c); /* XXX no locale support yet */
1417 Perl_is_uni_upper_lc(pTHX_ UV c)
1419 return is_uni_upper(c); /* XXX no locale support yet */
1423 Perl_is_uni_lower_lc(pTHX_ UV c)
1425 return is_uni_lower(c); /* XXX no locale support yet */
1429 Perl_is_uni_cntrl_lc(pTHX_ UV c)
1431 return is_uni_cntrl(c); /* XXX no locale support yet */
1435 Perl_is_uni_graph_lc(pTHX_ UV c)
1437 return is_uni_graph(c); /* XXX no locale support yet */
1441 Perl_is_uni_print_lc(pTHX_ UV c)
1443 return is_uni_print(c); /* XXX no locale support yet */
1447 Perl_is_uni_punct_lc(pTHX_ UV c)
1449 return is_uni_punct(c); /* XXX no locale support yet */
1453 Perl_is_uni_xdigit_lc(pTHX_ UV c)
1455 return is_uni_xdigit(c); /* XXX no locale support yet */
1459 Perl_to_uni_upper_lc(pTHX_ U32 c)
1461 /* XXX returns only the first character -- do not use XXX */
1462 /* XXX no locale support yet */
1464 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1465 return (U32)to_uni_upper(c, tmpbuf, &len);
1469 Perl_to_uni_title_lc(pTHX_ U32 c)
1471 /* XXX returns only the first character XXX -- do not use XXX */
1472 /* XXX no locale support yet */
1474 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1475 return (U32)to_uni_title(c, tmpbuf, &len);
1479 Perl_to_uni_lower_lc(pTHX_ U32 c)
1481 /* XXX returns only the first character -- do not use XXX */
1482 /* XXX no locale support yet */
1484 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1485 return (U32)to_uni_lower(c, tmpbuf, &len);
1489 S_is_utf8_common(pTHX_ const U8 *const p, SV **swash,
1490 const char *const swashname)
1494 PERL_ARGS_ASSERT_IS_UTF8_COMMON;
1496 if (!is_utf8_char(p))
1499 *swash = swash_init("utf8", swashname, &PL_sv_undef, 1, 0);
1500 return swash_fetch(*swash, p, TRUE) != 0;
1504 Perl_is_utf8_alnum(pTHX_ const U8 *p)
1508 PERL_ARGS_ASSERT_IS_UTF8_ALNUM;
1510 /* NOTE: "IsWord", not "IsAlnum", since Alnum is a true
1511 * descendant of isalnum(3), in other words, it doesn't
1512 * contain the '_'. --jhi */
1513 return is_utf8_common(p, &PL_utf8_alnum, "IsWord");
1517 Perl_is_utf8_idfirst(pTHX_ const U8 *p) /* The naming is historical. */
1521 PERL_ARGS_ASSERT_IS_UTF8_IDFIRST;
1525 /* is_utf8_idstart would be more logical. */
1526 return is_utf8_common(p, &PL_utf8_idstart, "IdStart");
1530 Perl_is_utf8_xidfirst(pTHX_ const U8 *p) /* The naming is historical. */
1534 PERL_ARGS_ASSERT_IS_UTF8_XIDFIRST;
1538 /* is_utf8_idstart would be more logical. */
1539 return is_utf8_common(p, &PL_utf8_xidstart, "XIdStart");
1543 Perl__is_utf8__perl_idstart(pTHX_ const U8 *p)
1547 PERL_ARGS_ASSERT__IS_UTF8__PERL_IDSTART;
1549 return is_utf8_common(p, &PL_utf8_perl_idstart, "_Perl_IDStart");
1553 Perl_is_utf8_idcont(pTHX_ const U8 *p)
1557 PERL_ARGS_ASSERT_IS_UTF8_IDCONT;
1559 return is_utf8_common(p, &PL_utf8_idcont, "IdContinue");
1563 Perl_is_utf8_xidcont(pTHX_ const U8 *p)
1567 PERL_ARGS_ASSERT_IS_UTF8_XIDCONT;
1569 return is_utf8_common(p, &PL_utf8_idcont, "XIdContinue");
1573 Perl_is_utf8_alpha(pTHX_ const U8 *p)
1577 PERL_ARGS_ASSERT_IS_UTF8_ALPHA;
1579 return is_utf8_common(p, &PL_utf8_alpha, "IsAlpha");
1583 Perl_is_utf8_ascii(pTHX_ const U8 *p)
1587 PERL_ARGS_ASSERT_IS_UTF8_ASCII;
1589 /* ASCII characters are the same whether in utf8 or not. So the macro
1590 * works on both utf8 and non-utf8 representations. */
1595 Perl_is_utf8_space(pTHX_ const U8 *p)
1599 PERL_ARGS_ASSERT_IS_UTF8_SPACE;
1601 return is_utf8_common(p, &PL_utf8_space, "IsXPerlSpace");
1605 Perl_is_utf8_perl_space(pTHX_ const U8 *p)
1609 PERL_ARGS_ASSERT_IS_UTF8_PERL_SPACE;
1611 /* Only true if is an ASCII space-like character, and ASCII is invariant
1612 * under utf8, so can just use the macro */
1613 return isSPACE_A(*p);
1617 Perl_is_utf8_perl_word(pTHX_ const U8 *p)
1621 PERL_ARGS_ASSERT_IS_UTF8_PERL_WORD;
1623 /* Only true if is an ASCII word character, and ASCII is invariant
1624 * under utf8, so can just use the macro */
1625 return isWORDCHAR_A(*p);
1629 Perl_is_utf8_digit(pTHX_ const U8 *p)
1633 PERL_ARGS_ASSERT_IS_UTF8_DIGIT;
1635 return is_utf8_common(p, &PL_utf8_digit, "IsDigit");
1639 Perl_is_utf8_posix_digit(pTHX_ const U8 *p)
1643 PERL_ARGS_ASSERT_IS_UTF8_POSIX_DIGIT;
1645 /* Only true if is an ASCII digit character, and ASCII is invariant
1646 * under utf8, so can just use the macro */
1647 return isDIGIT_A(*p);
1651 Perl_is_utf8_upper(pTHX_ const U8 *p)
1655 PERL_ARGS_ASSERT_IS_UTF8_UPPER;
1657 return is_utf8_common(p, &PL_utf8_upper, "IsUppercase");
1661 Perl_is_utf8_lower(pTHX_ const U8 *p)
1665 PERL_ARGS_ASSERT_IS_UTF8_LOWER;
1667 return is_utf8_common(p, &PL_utf8_lower, "IsLowercase");
1671 Perl_is_utf8_cntrl(pTHX_ const U8 *p)
1675 PERL_ARGS_ASSERT_IS_UTF8_CNTRL;
1678 return isCNTRL_A(*p);
1681 /* All controls are in Latin1 */
1682 if (! UTF8_IS_DOWNGRADEABLE_START(*p)) {
1685 return isCNTRL_L1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)));
1689 Perl_is_utf8_graph(pTHX_ const U8 *p)
1693 PERL_ARGS_ASSERT_IS_UTF8_GRAPH;
1695 return is_utf8_common(p, &PL_utf8_graph, "IsGraph");
1699 Perl_is_utf8_print(pTHX_ const U8 *p)
1703 PERL_ARGS_ASSERT_IS_UTF8_PRINT;
1705 return is_utf8_common(p, &PL_utf8_print, "IsPrint");
1709 Perl_is_utf8_punct(pTHX_ const U8 *p)
1713 PERL_ARGS_ASSERT_IS_UTF8_PUNCT;
1715 return is_utf8_common(p, &PL_utf8_punct, "IsPunct");
1719 Perl_is_utf8_xdigit(pTHX_ const U8 *p)
1723 PERL_ARGS_ASSERT_IS_UTF8_XDIGIT;
1725 return is_utf8_common(p, &PL_utf8_xdigit, "IsXDigit");
1729 Perl_is_utf8_mark(pTHX_ const U8 *p)
1733 PERL_ARGS_ASSERT_IS_UTF8_MARK;
1735 return is_utf8_common(p, &PL_utf8_mark, "IsM");
1739 Perl_is_utf8_X_begin(pTHX_ const U8 *p)
1743 PERL_ARGS_ASSERT_IS_UTF8_X_BEGIN;
1745 return is_utf8_common(p, &PL_utf8_X_begin, "_X_Begin");
1749 Perl_is_utf8_X_extend(pTHX_ const U8 *p)
1753 PERL_ARGS_ASSERT_IS_UTF8_X_EXTEND;
1755 return is_utf8_common(p, &PL_utf8_X_extend, "_X_Extend");
1759 Perl_is_utf8_X_prepend(pTHX_ const U8 *p)
1763 PERL_ARGS_ASSERT_IS_UTF8_X_PREPEND;
1765 return is_utf8_common(p, &PL_utf8_X_prepend, "GCB=Prepend");
1769 Perl_is_utf8_X_non_hangul(pTHX_ const U8 *p)
1773 PERL_ARGS_ASSERT_IS_UTF8_X_NON_HANGUL;
1775 return is_utf8_common(p, &PL_utf8_X_non_hangul, "HST=Not_Applicable");
1779 Perl_is_utf8_X_L(pTHX_ const U8 *p)
1783 PERL_ARGS_ASSERT_IS_UTF8_X_L;
1785 return is_utf8_common(p, &PL_utf8_X_L, "GCB=L");
1789 Perl_is_utf8_X_LV(pTHX_ const U8 *p)
1793 PERL_ARGS_ASSERT_IS_UTF8_X_LV;
1795 return is_utf8_common(p, &PL_utf8_X_LV, "GCB=LV");
1799 Perl_is_utf8_X_LVT(pTHX_ const U8 *p)
1803 PERL_ARGS_ASSERT_IS_UTF8_X_LVT;
1805 return is_utf8_common(p, &PL_utf8_X_LVT, "GCB=LVT");
1809 Perl_is_utf8_X_T(pTHX_ const U8 *p)
1813 PERL_ARGS_ASSERT_IS_UTF8_X_T;
1815 return is_utf8_common(p, &PL_utf8_X_T, "GCB=T");
1819 Perl_is_utf8_X_V(pTHX_ const U8 *p)
1823 PERL_ARGS_ASSERT_IS_UTF8_X_V;
1825 return is_utf8_common(p, &PL_utf8_X_V, "GCB=V");
1829 Perl_is_utf8_X_LV_LVT_V(pTHX_ const U8 *p)
1833 PERL_ARGS_ASSERT_IS_UTF8_X_LV_LVT_V;
1835 return is_utf8_common(p, &PL_utf8_X_LV_LVT_V, "_X_LV_LVT_V");
1839 =for apidoc to_utf8_case
1841 The "p" contains the pointer to the UTF-8 string encoding
1842 the character that is being converted.
1844 The "ustrp" is a pointer to the character buffer to put the
1845 conversion result to. The "lenp" is a pointer to the length
1848 The "swashp" is a pointer to the swash to use.
1850 Both the special and normal mappings are stored in lib/unicore/To/Foo.pl,
1851 and loaded by SWASHNEW, using lib/utf8_heavy.pl. The special (usually,
1852 but not always, a multicharacter mapping), is tried first.
1854 The "special" is a string like "utf8::ToSpecLower", which means the
1855 hash %utf8::ToSpecLower. The access to the hash is through
1856 Perl_to_utf8_case().
1858 The "normal" is a string like "ToLower" which means the swash
1864 Perl_to_utf8_case(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp,
1865 SV **swashp, const char *normal, const char *special)
1868 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1870 const UV uv0 = utf8_to_uvchr(p, NULL);
1871 /* The NATIVE_TO_UNI() and UNI_TO_NATIVE() mappings
1872 * are necessary in EBCDIC, they are redundant no-ops
1873 * in ASCII-ish platforms, and hopefully optimized away. */
1874 const UV uv1 = NATIVE_TO_UNI(uv0);
1876 PERL_ARGS_ASSERT_TO_UTF8_CASE;
1878 /* Note that swash_fetch() doesn't output warnings for these because it
1879 * assumes we will */
1880 if (uv1 >= UNICODE_SURROGATE_FIRST) {
1881 if (uv1 <= UNICODE_SURROGATE_LAST) {
1882 if (ckWARN_d(WARN_SURROGATE)) {
1883 const char* desc = (PL_op) ? OP_DESC(PL_op) : normal;
1884 Perl_warner(aTHX_ packWARN(WARN_SURROGATE),
1885 "Operation \"%s\" returns its argument for UTF-16 surrogate U+%04"UVXf"", desc, uv1);
1888 else if (UNICODE_IS_SUPER(uv1)) {
1889 if (ckWARN_d(WARN_NON_UNICODE)) {
1890 const char* desc = (PL_op) ? OP_DESC(PL_op) : normal;
1891 Perl_warner(aTHX_ packWARN(WARN_NON_UNICODE),
1892 "Operation \"%s\" returns its argument for non-Unicode code point 0x%04"UVXf"", desc, uv1);
1896 /* Note that non-characters are perfectly legal, so no warning should
1900 uvuni_to_utf8(tmpbuf, uv1);
1902 if (!*swashp) /* load on-demand */
1903 *swashp = swash_init("utf8", normal, &PL_sv_undef, 4, 0);
1906 /* It might be "special" (sometimes, but not always,
1907 * a multicharacter mapping) */
1908 HV * const hv = get_hv(special, 0);
1912 (svp = hv_fetch(hv, (const char*)tmpbuf, UNISKIP(uv1), FALSE)) &&
1916 s = SvPV_const(*svp, len);
1918 len = uvuni_to_utf8(ustrp, NATIVE_TO_UNI(*(U8*)s)) - ustrp;
1921 /* If we have EBCDIC we need to remap the characters
1922 * since any characters in the low 256 are Unicode
1923 * code points, not EBCDIC. */
1924 U8 *t = (U8*)s, *tend = t + len, *d;
1931 const UV c = utf8_to_uvchr(t, &tlen);
1933 d = uvchr_to_utf8(d, UNI_TO_NATIVE(c));
1942 d = uvchr_to_utf8(d, UNI_TO_NATIVE(*t));
1947 Copy(tmpbuf, ustrp, len, U8);
1949 Copy(s, ustrp, len, U8);
1955 if (!len && *swashp) {
1956 const UV uv2 = swash_fetch(*swashp, tmpbuf, TRUE);
1959 /* It was "normal" (a single character mapping). */
1960 const UV uv3 = UNI_TO_NATIVE(uv2);
1961 len = uvchr_to_utf8(ustrp, uv3) - ustrp;
1965 if (!len) /* Neither: just copy. In other words, there was no mapping
1966 defined, which means that the code point maps to itself */
1967 len = uvchr_to_utf8(ustrp, uv0) - ustrp;
1972 return len ? utf8_to_uvchr(ustrp, 0) : 0;
1976 =for apidoc to_utf8_upper
1978 Convert the UTF-8 encoded character at p to its uppercase version and
1979 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
1980 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since
1981 the uppercase version may be longer than the original character.
1983 The first character of the uppercased version is returned
1984 (but note, as explained above, that there may be more.)
1989 Perl_to_utf8_upper(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
1993 PERL_ARGS_ASSERT_TO_UTF8_UPPER;
1995 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
1996 &PL_utf8_toupper, "ToUpper", "utf8::ToSpecUpper");
2000 =for apidoc to_utf8_title
2002 Convert the UTF-8 encoded character at p to its titlecase version and
2003 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
2004 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
2005 titlecase version may be longer than the original character.
2007 The first character of the titlecased version is returned
2008 (but note, as explained above, that there may be more.)
2013 Perl_to_utf8_title(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
2017 PERL_ARGS_ASSERT_TO_UTF8_TITLE;
2019 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
2020 &PL_utf8_totitle, "ToTitle", "utf8::ToSpecTitle");
2024 =for apidoc to_utf8_lower
2026 Convert the UTF-8 encoded character at p to its lowercase version and
2027 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
2028 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
2029 lowercase version may be longer than the original character.
2031 The first character of the lowercased version is returned
2032 (but note, as explained above, that there may be more.)
2037 Perl_to_utf8_lower(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
2041 PERL_ARGS_ASSERT_TO_UTF8_LOWER;
2043 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
2044 &PL_utf8_tolower, "ToLower", "utf8::ToSpecLower");
2048 =for apidoc to_utf8_fold
2050 Convert the UTF-8 encoded character at p to its foldcase version and
2051 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
2052 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
2053 foldcase version may be longer than the original character (up to
2056 The first character of the foldcased version is returned
2057 (but note, as explained above, that there may be more.)
2061 /* Not currently externally documented is 'flags', which currently is non-zero
2062 * if full case folds are to be used; otherwise simple folds */
2065 Perl__to_utf8_fold_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, U8 flags)
2067 const char *specials = (flags) ? "utf8::ToSpecFold" : NULL;
2071 PERL_ARGS_ASSERT__TO_UTF8_FOLD_FLAGS;
2073 return Perl_to_utf8_case(aTHX_ p, ustrp, lenp,
2074 &PL_utf8_tofold, "ToFold", specials);
2078 * A "swash" is a swatch hash.
2079 * A "swatch" is a bit vector generated by utf8.c:S_swash_get().
2080 * C<pkg> is a pointer to a package name for SWASHNEW, should be "utf8".
2081 * For other parameters, see utf8::SWASHNEW in lib/utf8_heavy.pl.
2084 Perl_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv, I32 minbits, I32 none)
2089 const size_t pkg_len = strlen(pkg);
2090 const size_t name_len = strlen(name);
2091 HV * const stash = gv_stashpvn(pkg, pkg_len, 0);
2095 PERL_ARGS_ASSERT_SWASH_INIT;
2097 PUSHSTACKi(PERLSI_MAGIC);
2101 if (PL_parser && PL_parser->error_count)
2102 SAVEI8(PL_parser->error_count), PL_parser->error_count = 0;
2103 method = gv_fetchmeth(stash, "SWASHNEW", 8, -1);
2104 if (!method) { /* demand load utf8 */
2106 errsv_save = newSVsv(ERRSV);
2107 /* It is assumed that callers of this routine are not passing in any
2108 user derived data. */
2109 /* Need to do this after save_re_context() as it will set PL_tainted to
2110 1 while saving $1 etc (see the code after getrx: in Perl_magic_get).
2111 Even line to create errsv_save can turn on PL_tainted. */
2112 SAVEBOOL(PL_tainted);
2114 Perl_load_module(aTHX_ PERL_LOADMOD_NOIMPORT, newSVpvn(pkg,pkg_len),
2117 sv_setsv(ERRSV, errsv_save);
2118 SvREFCNT_dec(errsv_save);
2124 mPUSHp(pkg, pkg_len);
2125 mPUSHp(name, name_len);
2130 errsv_save = newSVsv(ERRSV);
2131 /* If we already have a pointer to the method, no need to use call_method()
2132 to repeat the lookup. */
2133 if (method ? call_sv(MUTABLE_SV(method), G_SCALAR)
2134 : call_sv(newSVpvs_flags("SWASHNEW", SVs_TEMP), G_SCALAR | G_METHOD))
2135 retval = newSVsv(*PL_stack_sp--);
2137 retval = &PL_sv_undef;
2139 sv_setsv(ERRSV, errsv_save);
2140 SvREFCNT_dec(errsv_save);
2143 if (IN_PERL_COMPILETIME) {
2144 CopHINTS_set(PL_curcop, PL_hints);
2146 if (!SvROK(retval) || SvTYPE(SvRV(retval)) != SVt_PVHV) {
2148 Perl_croak(aTHX_ "Can't find Unicode property definition \"%"SVf"\"",
2150 Perl_croak(aTHX_ "SWASHNEW didn't return an HV ref");
2156 /* This API is wrong for special case conversions since we may need to
2157 * return several Unicode characters for a single Unicode character
2158 * (see lib/unicore/SpecCase.txt) The SWASHGET in lib/utf8_heavy.pl is
2159 * the lower-level routine, and it is similarly broken for returning
2160 * multiple values. --jhi
2161 * For those, you should use to_utf8_case() instead */
2162 /* Now SWASHGET is recasted into S_swash_get in this file. */
2165 * Returns the value of property/mapping C<swash> for the first character
2166 * of the string C<ptr>. If C<do_utf8> is true, the string C<ptr> is
2167 * assumed to be in utf8. If C<do_utf8> is false, the string C<ptr> is
2168 * assumed to be in native 8-bit encoding. Caches the swatch in C<swash>.
2171 Perl_swash_fetch(pTHX_ SV *swash, const U8 *ptr, bool do_utf8)
2174 HV *const hv = MUTABLE_HV(SvRV(swash));
2179 const U8 *tmps = NULL;
2183 const UV c = NATIVE_TO_ASCII(*ptr);
2185 PERL_ARGS_ASSERT_SWASH_FETCH;
2187 if (!do_utf8 && !UNI_IS_INVARIANT(c)) {
2188 tmputf8[0] = (U8)UTF8_EIGHT_BIT_HI(c);
2189 tmputf8[1] = (U8)UTF8_EIGHT_BIT_LO(c);
2192 /* Given a UTF-X encoded char 0xAA..0xYY,0xZZ
2193 * then the "swatch" is a vec() for all the chars which start
2195 * So the key in the hash (klen) is length of encoded char -1
2197 klen = UTF8SKIP(ptr) - 1;
2201 /* If char is invariant then swatch is for all the invariant chars
2202 * In both UTF-8 and UTF-8-MOD that happens to be UTF_CONTINUATION_MARK
2204 needents = UTF_CONTINUATION_MARK;
2205 off = NATIVE_TO_UTF(ptr[klen]);
2208 /* If char is encoded then swatch is for the prefix */
2209 needents = (1 << UTF_ACCUMULATION_SHIFT);
2210 off = NATIVE_TO_UTF(ptr[klen]) & UTF_CONTINUATION_MASK;
2211 if (UTF8_IS_SUPER(ptr) && ckWARN_d(WARN_NON_UNICODE)) {
2212 const UV code_point = utf8n_to_uvuni(ptr, UTF8_MAXBYTES, 0, 0);
2214 /* This outputs warnings for binary properties only, assuming that
2215 * to_utf8_case() will output any for non-binary. Also, surrogates
2216 * aren't checked for, as that would warn on things like
2218 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
2219 if (SvUV(*bitssvp) == 1) {
2220 Perl_warner(aTHX_ packWARN(WARN_NON_UNICODE),
2221 "Code point 0x%04"UVXf" is not Unicode, all \\p{} matches fail; all \\P{} matches succeed", code_point);
2227 * This single-entry cache saves about 1/3 of the utf8 overhead in test
2228 * suite. (That is, only 7-8% overall over just a hash cache. Still,
2229 * it's nothing to sniff at.) Pity we usually come through at least
2230 * two function calls to get here...
2232 * NB: this code assumes that swatches are never modified, once generated!
2235 if (hv == PL_last_swash_hv &&
2236 klen == PL_last_swash_klen &&
2237 (!klen || memEQ((char *)ptr, (char *)PL_last_swash_key, klen)) )
2239 tmps = PL_last_swash_tmps;
2240 slen = PL_last_swash_slen;
2243 /* Try our second-level swatch cache, kept in a hash. */
2244 SV** svp = hv_fetch(hv, (const char*)ptr, klen, FALSE);
2246 /* If not cached, generate it via swash_get */
2247 if (!svp || !SvPOK(*svp)
2248 || !(tmps = (const U8*)SvPV_const(*svp, slen))) {
2249 /* We use utf8n_to_uvuni() as we want an index into
2250 Unicode tables, not a native character number.
2252 const UV code_point = utf8n_to_uvuni(ptr, UTF8_MAXBYTES, 0,
2254 0 : UTF8_ALLOW_ANY);
2255 swatch = swash_get(swash,
2256 /* On EBCDIC & ~(0xA0-1) isn't a useful thing to do */
2257 (klen) ? (code_point & ~(needents - 1)) : 0,
2260 if (IN_PERL_COMPILETIME)
2261 CopHINTS_set(PL_curcop, PL_hints);
2263 svp = hv_store(hv, (const char *)ptr, klen, swatch, 0);
2265 if (!svp || !(tmps = (U8*)SvPV(*svp, slen))
2266 || (slen << 3) < needents)
2267 Perl_croak(aTHX_ "panic: swash_fetch got improper swatch");
2270 PL_last_swash_hv = hv;
2271 assert(klen <= sizeof(PL_last_swash_key));
2272 PL_last_swash_klen = (U8)klen;
2273 /* FIXME change interpvar.h? */
2274 PL_last_swash_tmps = (U8 *) tmps;
2275 PL_last_swash_slen = slen;
2277 Copy(ptr, PL_last_swash_key, klen, U8);
2280 switch ((int)((slen << 3) / needents)) {
2282 bit = 1 << (off & 7);
2284 return (tmps[off] & bit) != 0;
2289 return (tmps[off] << 8) + tmps[off + 1] ;
2292 return (tmps[off] << 24) + (tmps[off+1] << 16) + (tmps[off+2] << 8) + tmps[off + 3] ;
2294 Perl_croak(aTHX_ "panic: swash_fetch got swatch of unexpected bit width");
2295 NORETURN_FUNCTION_END;
2298 /* Read a single line of the main body of the swash input text. These are of
2301 * where each number is hex. The first two numbers form the minimum and
2302 * maximum of a range, and the third is the value associated with the range.
2303 * Not all swashes should have a third number
2305 * On input: l points to the beginning of the line to be examined; it points
2306 * to somewhere in the string of the whole input text, and is
2307 * terminated by a \n or the null string terminator.
2308 * lend points to the null terminator of that string
2309 * wants_value is non-zero if the swash expects a third number
2310 * typestr is the name of the swash's mapping, like 'ToLower'
2311 * On output: *min, *max, and *val are set to the values read from the line.
2312 * returns a pointer just beyond the line examined. If there was no
2313 * valid min number on the line, returns lend+1
2317 S_swash_scan_list_line(pTHX_ U8* l, U8* const lend, UV* min, UV* max, UV* val,
2318 const bool wants_value, const U8* const typestr)
2320 const int typeto = typestr[0] == 'T' && typestr[1] == 'o';
2321 STRLEN numlen; /* Length of the number */
2322 I32 flags = PERL_SCAN_SILENT_ILLDIGIT
2323 | PERL_SCAN_DISALLOW_PREFIX
2324 | PERL_SCAN_SILENT_NON_PORTABLE;
2326 /* nl points to the next \n in the scan */
2327 U8* const nl = (U8*)memchr(l, '\n', lend - l);
2329 /* Get the first number on the line: the range minimum */
2331 *min = grok_hex((char *)l, &numlen, &flags, NULL);
2332 if (numlen) /* If found a hex number, position past it */
2334 else if (nl) { /* Else, go handle next line, if any */
2335 return nl + 1; /* 1 is length of "\n" */
2337 else { /* Else, no next line */
2338 return lend + 1; /* to LIST's end at which \n is not found */
2341 /* The max range value follows, separated by a BLANK */
2344 flags = PERL_SCAN_SILENT_ILLDIGIT
2345 | PERL_SCAN_DISALLOW_PREFIX
2346 | PERL_SCAN_SILENT_NON_PORTABLE;
2348 *max = grok_hex((char *)l, &numlen, &flags, NULL);
2351 else /* If no value here, it is a single element range */
2354 /* Non-binary tables have a third entry: what the first element of the
2359 flags = PERL_SCAN_SILENT_ILLDIGIT
2360 | PERL_SCAN_DISALLOW_PREFIX
2361 | PERL_SCAN_SILENT_NON_PORTABLE;
2363 *val = grok_hex((char *)l, &numlen, &flags, NULL);
2372 Perl_croak(aTHX_ "%s: illegal mapping '%s'",
2378 *val = 0; /* bits == 1, then any val should be ignored */
2380 else { /* Nothing following range min, should be single element with no
2386 Perl_croak(aTHX_ "%s: illegal mapping '%s'", typestr, l);
2390 *val = 0; /* bits == 1, then val should be ignored */
2393 /* Position to next line if any, or EOF */
2403 * Returns a swatch (a bit vector string) for a code point sequence
2404 * that starts from the value C<start> and comprises the number C<span>.
2405 * A C<swash> must be an object created by SWASHNEW (see lib/utf8_heavy.pl).
2406 * Should be used via swash_fetch, which will cache the swatch in C<swash>.
2409 S_swash_get(pTHX_ SV* swash, UV start, UV span)
2412 U8 *l, *lend, *x, *xend, *s, *send;
2413 STRLEN lcur, xcur, scur;
2414 HV *const hv = MUTABLE_HV(SvRV(swash));
2416 /* The string containing the main body of the table */
2417 SV** const listsvp = hv_fetchs(hv, "LIST", FALSE);
2419 SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE);
2420 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
2421 SV** const nonesvp = hv_fetchs(hv, "NONE", FALSE);
2422 SV** const extssvp = hv_fetchs(hv, "EXTRAS", FALSE);
2423 SV** const invert_it_svp = hv_fetchs(hv, "INVERT_IT", FALSE);
2424 const U8* const typestr = (U8*)SvPV_nolen(*typesvp);
2425 const STRLEN bits = SvUV(*bitssvp);
2426 const STRLEN octets = bits >> 3; /* if bits == 1, then octets == 0 */
2427 const UV none = SvUV(*nonesvp);
2428 const UV end = start + span;
2430 PERL_ARGS_ASSERT_SWASH_GET;
2432 if (bits != 1 && bits != 8 && bits != 16 && bits != 32) {
2433 Perl_croak(aTHX_ "panic: swash_get doesn't expect bits %"UVuf,
2437 /* create and initialize $swatch */
2438 scur = octets ? (span * octets) : (span + 7) / 8;
2439 swatch = newSV(scur);
2441 s = (U8*)SvPVX(swatch);
2442 if (octets && none) {
2443 const U8* const e = s + scur;
2446 *s++ = (U8)(none & 0xff);
2447 else if (bits == 16) {
2448 *s++ = (U8)((none >> 8) & 0xff);
2449 *s++ = (U8)( none & 0xff);
2451 else if (bits == 32) {
2452 *s++ = (U8)((none >> 24) & 0xff);
2453 *s++ = (U8)((none >> 16) & 0xff);
2454 *s++ = (U8)((none >> 8) & 0xff);
2455 *s++ = (U8)( none & 0xff);
2461 (void)memzero((U8*)s, scur + 1);
2463 SvCUR_set(swatch, scur);
2464 s = (U8*)SvPVX(swatch);
2466 /* read $swash->{LIST} */
2467 l = (U8*)SvPV(*listsvp, lcur);
2471 l = S_swash_scan_list_line(aTHX_ l, lend, &min, &max, &val,
2472 cBOOL(octets), typestr);
2477 /* If looking for something beyond this range, go try the next one */
2484 if (!none || val < none) {
2489 for (key = min; key <= max; key++) {
2493 /* offset must be non-negative (start <= min <= key < end) */
2494 offset = octets * (key - start);
2496 s[offset] = (U8)(val & 0xff);
2497 else if (bits == 16) {
2498 s[offset ] = (U8)((val >> 8) & 0xff);
2499 s[offset + 1] = (U8)( val & 0xff);
2501 else if (bits == 32) {
2502 s[offset ] = (U8)((val >> 24) & 0xff);
2503 s[offset + 1] = (U8)((val >> 16) & 0xff);
2504 s[offset + 2] = (U8)((val >> 8) & 0xff);
2505 s[offset + 3] = (U8)( val & 0xff);
2508 if (!none || val < none)
2512 else { /* bits == 1, then val should be ignored */
2516 for (key = min; key <= max; key++) {
2517 const STRLEN offset = (STRLEN)(key - start);
2520 s[offset >> 3] |= 1 << (offset & 7);
2526 /* Invert if the data says it should be. Assumes that bits == 1 */
2527 if (invert_it_svp && SvUV(*invert_it_svp)) {
2529 /* Unicode properties should come with all bits above PERL_UNICODE_MAX
2530 * be 0, and their inversion should also be 0, as we don't succeed any
2531 * Unicode property matches for non-Unicode code points */
2532 if (start <= PERL_UNICODE_MAX) {
2534 /* The code below assumes that we never cross the
2535 * Unicode/above-Unicode boundary in a range, as otherwise we would
2536 * have to figure out where to stop flipping the bits. Since this
2537 * boundary is divisible by a large power of 2, and swatches comes
2538 * in small powers of 2, this should be a valid assumption */
2539 assert(start + span - 1 <= PERL_UNICODE_MAX);
2549 /* read $swash->{EXTRAS}
2550 * This code also copied to swash_to_invlist() below */
2551 x = (U8*)SvPV(*extssvp, xcur);
2559 SV **otherbitssvp, *other;
2563 const U8 opc = *x++;
2567 nl = (U8*)memchr(x, '\n', xend - x);
2569 if (opc != '-' && opc != '+' && opc != '!' && opc != '&') {
2571 x = nl + 1; /* 1 is length of "\n" */
2575 x = xend; /* to EXTRAS' end at which \n is not found */
2582 namelen = nl - namestr;
2586 namelen = xend - namestr;
2590 othersvp = hv_fetch(hv, (char *)namestr, namelen, FALSE);
2591 otherhv = MUTABLE_HV(SvRV(*othersvp));
2592 otherbitssvp = hv_fetchs(otherhv, "BITS", FALSE);
2593 otherbits = (STRLEN)SvUV(*otherbitssvp);
2594 if (bits < otherbits)
2595 Perl_croak(aTHX_ "panic: swash_get found swatch size mismatch");
2597 /* The "other" swatch must be destroyed after. */
2598 other = swash_get(*othersvp, start, span);
2599 o = (U8*)SvPV(other, olen);
2602 Perl_croak(aTHX_ "panic: swash_get got improper swatch");
2604 s = (U8*)SvPV(swatch, slen);
2605 if (bits == 1 && otherbits == 1) {
2607 Perl_croak(aTHX_ "panic: swash_get found swatch length mismatch");
2631 STRLEN otheroctets = otherbits >> 3;
2633 U8* const send = s + slen;
2638 if (otherbits == 1) {
2639 otherval = (o[offset >> 3] >> (offset & 7)) & 1;
2643 STRLEN vlen = otheroctets;
2651 if (opc == '+' && otherval)
2652 NOOP; /* replace with otherval */
2653 else if (opc == '!' && !otherval)
2655 else if (opc == '-' && otherval)
2657 else if (opc == '&' && !otherval)
2660 s += octets; /* no replacement */
2665 *s++ = (U8)( otherval & 0xff);
2666 else if (bits == 16) {
2667 *s++ = (U8)((otherval >> 8) & 0xff);
2668 *s++ = (U8)( otherval & 0xff);
2670 else if (bits == 32) {
2671 *s++ = (U8)((otherval >> 24) & 0xff);
2672 *s++ = (U8)((otherval >> 16) & 0xff);
2673 *s++ = (U8)((otherval >> 8) & 0xff);
2674 *s++ = (U8)( otherval & 0xff);
2678 sv_free(other); /* through with it! */
2684 Perl__swash_inversion_hash(pTHX_ SV* const swash)
2687 /* Subject to change or removal. For use only in one place in regcomp.c.
2688 * Can't be used on a property that is subject to user override, as it
2689 * relies on the value of SPECIALS in the swash which would be set by
2690 * utf8_heavy.pl to the hash in the non-overriden file, and hence is not set
2691 * for overridden properties
2693 * Returns a hash which is the inversion and closure of a swash mapping.
2694 * For example, consider the input lines:
2699 * The returned hash would have two keys, the utf8 for 006B and the utf8 for
2700 * 006C. The value for each key is an array. For 006C, the array would
2701 * have a two elements, the utf8 for itself, and for 004C. For 006B, there
2702 * would be three elements in its array, the utf8 for 006B, 004B and 212A.
2704 * Essentially, for any code point, it gives all the code points that map to
2705 * it, or the list of 'froms' for that point.
2707 * Currently it ignores any additions or deletions from other swashes,
2708 * looking at just the main body of the swash, and if there are SPECIALS
2709 * in the swash, at that hash
2711 * The specials hash can be extra code points, and most likely consists of
2712 * maps from single code points to multiple ones (each expressed as a string
2713 * of utf8 characters). This function currently returns only 1-1 mappings.
2714 * However consider this possible input in the specials hash:
2715 * "\xEF\xAC\x85" => "\x{0073}\x{0074}", # U+FB05 => 0073 0074
2716 * "\xEF\xAC\x86" => "\x{0073}\x{0074}", # U+FB06 => 0073 0074
2718 * Both FB05 and FB06 map to the same multi-char sequence, which we don't
2719 * currently handle. But it also means that FB05 and FB06 are equivalent in
2720 * a 1-1 mapping which we should handle, and this relationship may not be in
2721 * the main table. Therefore this function examines all the multi-char
2722 * sequences and adds the 1-1 mappings that come out of that. */
2726 HV *const hv = MUTABLE_HV(SvRV(swash));
2728 /* The string containing the main body of the table */
2729 SV** const listsvp = hv_fetchs(hv, "LIST", FALSE);
2731 SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE);
2732 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
2733 SV** const nonesvp = hv_fetchs(hv, "NONE", FALSE);
2734 /*SV** const extssvp = hv_fetchs(hv, "EXTRAS", FALSE);*/
2735 const U8* const typestr = (U8*)SvPV_nolen(*typesvp);
2736 const STRLEN bits = SvUV(*bitssvp);
2737 const STRLEN octets = bits >> 3; /* if bits == 1, then octets == 0 */
2738 const UV none = SvUV(*nonesvp);
2739 SV **specials_p = hv_fetchs(hv, "SPECIALS", 0);
2743 PERL_ARGS_ASSERT__SWASH_INVERSION_HASH;
2745 /* Must have at least 8 bits to get the mappings */
2746 if (bits != 8 && bits != 16 && bits != 32) {
2747 Perl_croak(aTHX_ "panic: swash_inversion_hash doesn't expect bits %"UVuf,
2751 if (specials_p) { /* It might be "special" (sometimes, but not always, a
2752 mapping to more than one character */
2754 /* Construct an inverse mapping hash for the specials */
2755 HV * const specials_hv = MUTABLE_HV(SvRV(*specials_p));
2756 HV * specials_inverse = newHV();
2757 char *char_from; /* the lhs of the map */
2758 I32 from_len; /* its byte length */
2759 char *char_to; /* the rhs of the map */
2760 I32 to_len; /* its byte length */
2761 SV *sv_to; /* and in a sv */
2762 AV* from_list; /* list of things that map to each 'to' */
2764 hv_iterinit(specials_hv);
2766 /* The keys are the characters (in utf8) that map to the corresponding
2767 * utf8 string value. Iterate through the list creating the inverse
2769 while ((sv_to = hv_iternextsv(specials_hv, &char_from, &from_len))) {
2771 if (! SvPOK(sv_to)) {
2772 Perl_croak(aTHX_ "panic: value returned from hv_iternextsv() unexpectedly is not a string");
2774 /*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)));*/
2776 /* Each key in the inverse list is a mapped-to value, and the key's
2777 * hash value is a list of the strings (each in utf8) that map to
2778 * it. Those strings are all one character long */
2779 if ((listp = hv_fetch(specials_inverse,
2783 from_list = (AV*) *listp;
2785 else { /* No entry yet for it: create one */
2786 from_list = newAV();
2787 if (! hv_store(specials_inverse,
2790 (SV*) from_list, 0))
2792 Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed");
2796 /* Here have the list associated with this 'to' (perhaps newly
2797 * created and empty). Just add to it. Note that we ASSUME that
2798 * the input is guaranteed to not have duplications, so we don't
2799 * check for that. Duplications just slow down execution time. */
2800 av_push(from_list, newSVpvn_utf8(char_from, from_len, TRUE));
2803 /* Here, 'specials_inverse' contains the inverse mapping. Go through
2804 * it looking for cases like the FB05/FB06 examples above. There would
2805 * be an entry in the hash like
2806 * 'st' => [ FB05, FB06 ]
2807 * In this example we will create two lists that get stored in the
2808 * returned hash, 'ret':
2809 * FB05 => [ FB05, FB06 ]
2810 * FB06 => [ FB05, FB06 ]
2812 * Note that there is nothing to do if the array only has one element.
2813 * (In the normal 1-1 case handled below, we don't have to worry about
2814 * two lists, as everything gets tied to the single list that is
2815 * generated for the single character 'to'. But here, we are omitting
2816 * that list, ('st' in the example), so must have multiple lists.) */
2817 while ((from_list = (AV *) hv_iternextsv(specials_inverse,
2818 &char_to, &to_len)))
2820 if (av_len(from_list) > 0) {
2823 /* We iterate over all combinations of i,j to place each code
2824 * point on each list */
2825 for (i = 0; i <= av_len(from_list); i++) {
2827 AV* i_list = newAV();
2828 SV** entryp = av_fetch(from_list, i, FALSE);
2829 if (entryp == NULL) {
2830 Perl_croak(aTHX_ "panic: av_fetch() unexpectedly failed");
2832 if (hv_fetch(ret, SvPVX(*entryp), SvCUR(*entryp), FALSE)) {
2833 Perl_croak(aTHX_ "panic: unexpected entry for %s", SvPVX(*entryp));
2835 if (! hv_store(ret, SvPVX(*entryp), SvCUR(*entryp),
2836 (SV*) i_list, FALSE))
2838 Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed");
2841 /* For debugging: UV u = utf8_to_uvchr((U8*) SvPVX(*entryp), 0);*/
2842 for (j = 0; j <= av_len(from_list); j++) {
2843 entryp = av_fetch(from_list, j, FALSE);
2844 if (entryp == NULL) {
2845 Perl_croak(aTHX_ "panic: av_fetch() unexpectedly failed");
2848 /* When i==j this adds itself to the list */
2849 av_push(i_list, newSVuv(utf8_to_uvchr(
2850 (U8*) SvPVX(*entryp), 0)));
2851 /*DEBUG_U(PerlIO_printf(Perl_debug_log, "Adding %"UVXf" to list for %"UVXf"\n", utf8_to_uvchr((U8*) SvPVX(*entryp), 0), u));*/
2856 SvREFCNT_dec(specials_inverse); /* done with it */
2857 } /* End of specials */
2859 /* read $swash->{LIST} */
2860 l = (U8*)SvPV(*listsvp, lcur);
2863 /* Go through each input line */
2867 l = S_swash_scan_list_line(aTHX_ l, lend, &min, &max, &val,
2868 cBOOL(octets), typestr);
2873 /* Each element in the range is to be inverted */
2874 for (inverse = min; inverse <= max; inverse++) {
2878 bool found_key = FALSE;
2879 bool found_inverse = FALSE;
2881 /* The key is the inverse mapping */
2882 char key[UTF8_MAXBYTES+1];
2883 char* key_end = (char *) uvuni_to_utf8((U8*) key, val);
2884 STRLEN key_len = key_end - key;
2886 /* Get the list for the map */
2887 if ((listp = hv_fetch(ret, key, key_len, FALSE))) {
2888 list = (AV*) *listp;
2890 else { /* No entry yet for it: create one */
2892 if (! hv_store(ret, key, key_len, (SV*) list, FALSE)) {
2893 Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed");
2897 /* Look through list to see if this inverse mapping already is
2898 * listed, or if there is a mapping to itself already */
2899 for (i = 0; i <= av_len(list); i++) {
2900 SV** entryp = av_fetch(list, i, FALSE);
2902 if (entryp == NULL) {
2903 Perl_croak(aTHX_ "panic: av_fetch() unexpectedly failed");
2906 /*DEBUG_U(PerlIO_printf(Perl_debug_log, "list for %"UVXf" contains %"UVXf"\n", val, SvUV(entry)));*/
2907 if (SvUV(entry) == val) {
2910 if (SvUV(entry) == inverse) {
2911 found_inverse = TRUE;
2914 /* No need to continue searching if found everything we are
2916 if (found_key && found_inverse) {
2921 /* Make sure there is a mapping to itself on the list */
2923 av_push(list, newSVuv(val));
2924 /*DEBUG_U(PerlIO_printf(Perl_debug_log, "Adding %"UVXf" to list for %"UVXf"\n", val, val));*/
2928 /* Simply add the value to the list */
2929 if (! found_inverse) {
2930 av_push(list, newSVuv(inverse));
2931 /*DEBUG_U(PerlIO_printf(Perl_debug_log, "Adding %"UVXf" to list for %"UVXf"\n", inverse, val));*/
2934 /* swash_get() increments the value of val for each element in the
2935 * range. That makes more compact tables possible. You can
2936 * express the capitalization, for example, of all consecutive
2937 * letters with a single line: 0061\t007A\t0041 This maps 0061 to
2938 * 0041, 0062 to 0042, etc. I (khw) have never understood 'none',
2939 * and it's not documented; it appears to be used only in
2940 * implementing tr//; I copied the semantics from swash_get(), just
2942 if (!none || val < none) {
2952 Perl__swash_to_invlist(pTHX_ SV* const swash)
2955 /* Subject to change or removal. For use only in one place in regcomp.c */
2960 HV *const hv = MUTABLE_HV(SvRV(swash));
2961 UV elements = 0; /* Number of elements in the inversion list */
2964 /* The string containing the main body of the table */
2965 SV** const listsvp = hv_fetchs(hv, "LIST", FALSE);
2966 SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE);
2967 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
2968 SV** const extssvp = hv_fetchs(hv, "EXTRAS", FALSE);
2969 SV** const invert_it_svp = hv_fetchs(hv, "INVERT_IT", FALSE);
2971 const U8* const typestr = (U8*)SvPV_nolen(*typesvp);
2972 const STRLEN bits = SvUV(*bitssvp);
2973 const STRLEN octets = bits >> 3; /* if bits == 1, then octets == 0 */
2979 PERL_ARGS_ASSERT__SWASH_TO_INVLIST;
2981 /* read $swash->{LIST} */
2982 if (SvPOK(*listsvp)) {
2983 l = (U8*)SvPV(*listsvp, lcur);
2986 /* LIST legitimately doesn't contain a string during compilation phases
2987 * of Perl itself, before the Unicode tables are generated. In this
2988 * case, just fake things up by creating an empty list */
2995 /* Scan the input to count the number of lines to preallocate array size
2996 * based on worst possible case, which is each line in the input creates 2
2997 * elements in the inversion list: 1) the beginning of a range in the list;
2998 * 2) the beginning of a range not in the list. */
2999 while ((loc = (strchr(loc, '\n'))) != NULL) {
3004 /* If the ending is somehow corrupt and isn't a new line, add another
3005 * element for the final range that isn't in the inversion list */
3006 if (! (*lend == '\n' || (*lend == '\0' && *(lend - 1) == '\n'))) {
3010 invlist = _new_invlist(elements);
3012 /* Now go through the input again, adding each range to the list */
3015 UV val; /* Not used by this function */
3017 l = S_swash_scan_list_line(aTHX_ l, lend, &start, &end, &val,
3018 cBOOL(octets), typestr);
3024 _append_range_to_invlist(invlist, start, end);
3027 /* Invert if the data says it should be */
3028 if (invert_it_svp && SvUV(*invert_it_svp)) {
3029 _invlist_invert_prop(invlist);
3032 /* This code is copied from swash_get()
3033 * read $swash->{EXTRAS} */
3034 x = (U8*)SvPV(*extssvp, xcur);
3042 SV **otherbitssvp, *other;
3045 const U8 opc = *x++;
3049 nl = (U8*)memchr(x, '\n', xend - x);
3051 if (opc != '-' && opc != '+' && opc != '!' && opc != '&') {
3053 x = nl + 1; /* 1 is length of "\n" */
3057 x = xend; /* to EXTRAS' end at which \n is not found */
3064 namelen = nl - namestr;
3068 namelen = xend - namestr;
3072 othersvp = hv_fetch(hv, (char *)namestr, namelen, FALSE);
3073 otherhv = MUTABLE_HV(SvRV(*othersvp));
3074 otherbitssvp = hv_fetchs(otherhv, "BITS", FALSE);
3075 otherbits = (STRLEN)SvUV(*otherbitssvp);
3077 if (bits != otherbits || bits != 1) {
3078 Perl_croak(aTHX_ "panic: _swash_to_invlist only operates on boolean properties");
3081 /* The "other" swatch must be destroyed after. */
3082 other = _swash_to_invlist((SV *)*othersvp);
3084 /* End of code copied from swash_get() */
3087 _invlist_union(invlist, other, &invlist);
3090 _invlist_invert(other);
3091 _invlist_union(invlist, other, &invlist);
3094 _invlist_subtract(invlist, other, &invlist);
3097 _invlist_intersection(invlist, other, &invlist);
3102 sv_free(other); /* through with it! */
3109 =for apidoc uvchr_to_utf8
3111 Adds the UTF-8 representation of the Native code point C<uv> to the end
3112 of the string C<d>; C<d> should be have at least C<UTF8_MAXBYTES+1> free
3113 bytes available. The return value is the pointer to the byte after the
3114 end of the new character. In other words,
3116 d = uvchr_to_utf8(d, uv);
3118 is the recommended wide native character-aware way of saying
3125 /* On ASCII machines this is normally a macro but we want a
3126 real function in case XS code wants it
3129 Perl_uvchr_to_utf8(pTHX_ U8 *d, UV uv)
3131 PERL_ARGS_ASSERT_UVCHR_TO_UTF8;
3133 return Perl_uvuni_to_utf8_flags(aTHX_ d, NATIVE_TO_UNI(uv), 0);
3137 Perl_uvchr_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags)
3139 PERL_ARGS_ASSERT_UVCHR_TO_UTF8_FLAGS;
3141 return Perl_uvuni_to_utf8_flags(aTHX_ d, NATIVE_TO_UNI(uv), flags);
3145 =for apidoc utf8n_to_uvchr
3147 Returns the native character value of the first character in the string
3149 which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
3150 length, in bytes, of that character.
3152 length and flags are the same as utf8n_to_uvuni().
3156 /* On ASCII machines this is normally a macro but we want
3157 a real function in case XS code wants it
3160 Perl_utf8n_to_uvchr(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen,
3163 const UV uv = Perl_utf8n_to_uvuni(aTHX_ s, curlen, retlen, flags);
3165 PERL_ARGS_ASSERT_UTF8N_TO_UVCHR;
3167 return UNI_TO_NATIVE(uv);
3171 Perl_check_utf8_print(pTHX_ register const U8* s, const STRLEN len)
3173 /* May change: warns if surrogates, non-character code points, or
3174 * non-Unicode code points are in s which has length len. Returns TRUE if
3175 * none found; FALSE otherwise. The only other validity check is to make
3176 * sure that this won't exceed the string's length */
3178 const U8* const e = s + len;
3181 PERL_ARGS_ASSERT_CHECK_UTF8_PRINT;
3184 if (UTF8SKIP(s) > len) {
3185 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8),
3186 "%s in %s", unees, PL_op ? OP_DESC(PL_op) : "print");
3189 if (*s >= UTF8_FIRST_PROBLEMATIC_CODE_POINT_FIRST_BYTE) {
3191 if (UTF8_IS_SUPER(s)) {
3192 if (ckWARN_d(WARN_NON_UNICODE)) {
3193 UV uv = utf8_to_uvchr(s, &char_len);
3194 Perl_warner(aTHX_ packWARN(WARN_NON_UNICODE),
3195 "Code point 0x%04"UVXf" is not Unicode, may not be portable", uv);
3199 else if (UTF8_IS_SURROGATE(s)) {
3200 if (ckWARN_d(WARN_SURROGATE)) {
3201 UV uv = utf8_to_uvchr(s, &char_len);
3202 Perl_warner(aTHX_ packWARN(WARN_SURROGATE),
3203 "Unicode surrogate U+%04"UVXf" is illegal in UTF-8", uv);
3208 ((UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s))
3209 && (ckWARN_d(WARN_NONCHAR)))
3211 UV uv = utf8_to_uvchr(s, &char_len);
3212 Perl_warner(aTHX_ packWARN(WARN_NONCHAR),
3213 "Unicode non-character U+%04"UVXf" is illegal for open interchange", uv);
3224 =for apidoc pv_uni_display
3226 Build to the scalar dsv a displayable version of the string spv,
3227 length len, the displayable version being at most pvlim bytes long
3228 (if longer, the rest is truncated and "..." will be appended).
3230 The flags argument can have UNI_DISPLAY_ISPRINT set to display
3231 isPRINT()able characters as themselves, UNI_DISPLAY_BACKSLASH
3232 to display the \\[nrfta\\] as the backslashed versions (like '\n')
3233 (UNI_DISPLAY_BACKSLASH is preferred over UNI_DISPLAY_ISPRINT for \\).
3234 UNI_DISPLAY_QQ (and its alias UNI_DISPLAY_REGEX) have both
3235 UNI_DISPLAY_BACKSLASH and UNI_DISPLAY_ISPRINT turned on.
3237 The pointer to the PV of the dsv is returned.
3241 Perl_pv_uni_display(pTHX_ SV *dsv, const U8 *spv, STRLEN len, STRLEN pvlim, UV flags)
3246 PERL_ARGS_ASSERT_PV_UNI_DISPLAY;
3250 for (s = (const char *)spv, e = s + len; s < e; s += UTF8SKIP(s)) {
3252 /* This serves double duty as a flag and a character to print after
3253 a \ when flags & UNI_DISPLAY_BACKSLASH is true.
3257 if (pvlim && SvCUR(dsv) >= pvlim) {
3261 u = utf8_to_uvchr((U8*)s, 0);
3263 const unsigned char c = (unsigned char)u & 0xFF;
3264 if (flags & UNI_DISPLAY_BACKSLASH) {
3281 const char string = ok;
3282 sv_catpvs(dsv, "\\");
3283 sv_catpvn(dsv, &string, 1);
3286 /* isPRINT() is the locale-blind version. */
3287 if (!ok && (flags & UNI_DISPLAY_ISPRINT) && isPRINT(c)) {
3288 const char string = c;
3289 sv_catpvn(dsv, &string, 1);
3294 Perl_sv_catpvf(aTHX_ dsv, "\\x{%"UVxf"}", u);
3297 sv_catpvs(dsv, "...");
3303 =for apidoc sv_uni_display
3305 Build to the scalar dsv a displayable version of the scalar sv,
3306 the displayable version being at most pvlim bytes long
3307 (if longer, the rest is truncated and "..." will be appended).
3309 The flags argument is as in pv_uni_display().
3311 The pointer to the PV of the dsv is returned.
3316 Perl_sv_uni_display(pTHX_ SV *dsv, SV *ssv, STRLEN pvlim, UV flags)
3318 PERL_ARGS_ASSERT_SV_UNI_DISPLAY;
3320 return Perl_pv_uni_display(aTHX_ dsv, (const U8*)SvPVX_const(ssv),
3321 SvCUR(ssv), pvlim, flags);
3325 =for apidoc foldEQ_utf8
3327 Returns true if the leading portions of the strings s1 and s2 (either or both
3328 of which may be in UTF-8) are the same case-insensitively; false otherwise.
3329 How far into the strings to compare is determined by other input parameters.
3331 If u1 is true, the string s1 is assumed to be in UTF-8-encoded Unicode;
3332 otherwise it is assumed to be in native 8-bit encoding. Correspondingly for u2
3335 If the byte length l1 is non-zero, it says how far into s1 to check for fold
3336 equality. In other words, s1+l1 will be used as a goal to reach. The
3337 scan will not be considered to be a match unless the goal is reached, and
3338 scanning won't continue past that goal. Correspondingly for l2 with respect to
3341 If pe1 is non-NULL and the pointer it points to is not NULL, that pointer is
3342 considered an end pointer beyond which scanning of s1 will not continue under
3343 any circumstances. This means that if both l1 and pe1 are specified, and pe1
3344 is less than s1+l1, the match will never be successful because it can never
3345 get as far as its goal (and in fact is asserted against). Correspondingly for
3346 pe2 with respect to s2.
3348 At least one of s1 and s2 must have a goal (at least one of l1 and l2 must be
3349 non-zero), and if both do, both have to be
3350 reached for a successful match. Also, if the fold of a character is multiple
3351 characters, all of them must be matched (see tr21 reference below for
3354 Upon a successful match, if pe1 is non-NULL,
3355 it will be set to point to the beginning of the I<next> character of s1 beyond
3356 what was matched. Correspondingly for pe2 and s2.
3358 For case-insensitiveness, the "casefolding" of Unicode is used
3359 instead of upper/lowercasing both the characters, see
3360 http://www.unicode.org/unicode/reports/tr21/ (Case Mappings).
3364 /* A flags parameter has been added which may change, and hence isn't
3365 * externally documented. Currently it is:
3366 * 0 for as-documented above
3367 * FOLDEQ_UTF8_NOMIX_ASCII meaning that if a non-ASCII character folds to an
3368 ASCII one, to not match
3369 * FOLDEQ_UTF8_LOCALE meaning that locale rules are to be used for code
3370 * points below 256; unicode rules for above 255; and
3371 * folds that cross those boundaries are disallowed,
3372 * like the NOMIX_ASCII option
3373 * FOLDEQ_S1_ALREADY_FOLDED s1 has already been folded before calling this
3374 * routine. This allows that step to be skipped.
3375 * FOLDEQ_S2_ALREADY_FOLDED Similarly.
3378 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)
3381 register const U8 *p1 = (const U8*)s1; /* Point to current char */
3382 register const U8 *p2 = (const U8*)s2;
3383 register const U8 *g1 = NULL; /* goal for s1 */
3384 register const U8 *g2 = NULL;
3385 register const U8 *e1 = NULL; /* Don't scan s1 past this */
3386 register U8 *f1 = NULL; /* Point to current folded */
3387 register const U8 *e2 = NULL;
3388 register U8 *f2 = NULL;
3389 STRLEN n1 = 0, n2 = 0; /* Number of bytes in current char */
3390 U8 foldbuf1[UTF8_MAXBYTES_CASE+1];
3391 U8 foldbuf2[UTF8_MAXBYTES_CASE+1];
3392 U8 natbuf[2]; /* Holds native 8-bit char converted to utf8;
3393 these always fit in 2 bytes */
3395 PERL_ARGS_ASSERT_FOLDEQ_UTF8_FLAGS;
3397 /* The algorithm requires that input with the flags on the first line of
3398 * the assert not be pre-folded. */
3399 assert( ! ((flags & (FOLDEQ_UTF8_NOMIX_ASCII | FOLDEQ_UTF8_LOCALE))
3400 && (flags & (FOLDEQ_S1_ALREADY_FOLDED | FOLDEQ_S2_ALREADY_FOLDED))));
3407 g1 = (const U8*)s1 + l1;
3415 g2 = (const U8*)s2 + l2;
3418 /* Must have at least one goal */
3423 /* Will never match if goal is out-of-bounds */
3424 assert(! e1 || e1 >= g1);
3426 /* Here, there isn't an end pointer, or it is beyond the goal. We
3427 * only go as far as the goal */
3431 assert(e1); /* Must have an end for looking at s1 */
3434 /* Same for goal for s2 */
3436 assert(! e2 || e2 >= g2);
3443 /* If both operands are already folded, we could just do a memEQ on the
3444 * whole strings at once, but it would be better if the caller realized
3445 * this and didn't even call us */
3447 /* Look through both strings, a character at a time */
3448 while (p1 < e1 && p2 < e2) {
3450 /* If at the beginning of a new character in s1, get its fold to use
3451 * and the length of the fold. (exception: locale rules just get the
3452 * character to a single byte) */
3454 if (flags & FOLDEQ_S1_ALREADY_FOLDED) {
3458 /* If in locale matching, we use two sets of rules, depending on if
3459 * the code point is above or below 255. Here, we test for and
3460 * handle locale rules */
3463 if ((flags & FOLDEQ_UTF8_LOCALE)
3464 && (! u1 || UTF8_IS_INVARIANT(*p1)
3465 || UTF8_IS_DOWNGRADEABLE_START(*p1)))
3467 /* There is no mixing of code points above and below 255. */
3468 if (u2 && (! UTF8_IS_INVARIANT(*p2)
3469 && ! UTF8_IS_DOWNGRADEABLE_START(*p2)))
3474 /* We handle locale rules by converting, if necessary, the
3475 * code point to a single byte. */
3476 if (! u1 || UTF8_IS_INVARIANT(*p1)) {
3480 *foldbuf1 = TWO_BYTE_UTF8_TO_UNI(*p1, *(p1 + 1));
3484 else if (isASCII(*p1)) { /* Note, that here won't be
3485 both ASCII and using locale
3488 /* If trying to mix non- with ASCII, and not supposed to,
3490 if ((flags & FOLDEQ_UTF8_NOMIX_ASCII) && ! isASCII(*p2)) {
3494 *foldbuf1 = toLOWER(*p1); /* Folds in the ASCII range are
3498 to_utf8_fold(p1, foldbuf1, &n1);
3500 else { /* Not utf8, convert to it first and then get fold */
3501 uvuni_to_utf8(natbuf, (UV) NATIVE_TO_UNI(((UV)*p1)));
3502 to_utf8_fold(natbuf, foldbuf1, &n1);
3508 if (n2 == 0) { /* Same for s2 */
3509 if (flags & FOLDEQ_S2_ALREADY_FOLDED) {
3514 if ((flags & FOLDEQ_UTF8_LOCALE)
3515 && (! u2 || UTF8_IS_INVARIANT(*p2) || UTF8_IS_DOWNGRADEABLE_START(*p2)))
3517 /* Here, the next char in s2 is < 256. We've already
3518 * worked on s1, and if it isn't also < 256, can't match */
3519 if (u1 && (! UTF8_IS_INVARIANT(*p1)
3520 && ! UTF8_IS_DOWNGRADEABLE_START(*p1)))
3524 if (! u2 || UTF8_IS_INVARIANT(*p2)) {
3528 *foldbuf2 = TWO_BYTE_UTF8_TO_UNI(*p2, *(p2 + 1));
3531 /* Use another function to handle locale rules. We've made
3532 * sure that both characters to compare are single bytes */
3533 if (! foldEQ_locale((char *) f1, (char *) foldbuf2, 1)) {
3538 else if (isASCII(*p2)) {
3539 if (flags && ! isASCII(*p1)) {
3543 *foldbuf2 = toLOWER(*p2);
3546 to_utf8_fold(p2, foldbuf2, &n2);
3549 uvuni_to_utf8(natbuf, (UV) NATIVE_TO_UNI(((UV)*p2)));
3550 to_utf8_fold(natbuf, foldbuf2, &n2);
3556 /* Here f1 and f2 point to the beginning of the strings to compare.
3557 * These strings are the folds of the next character from each input
3558 * string, stored in utf8. */
3560 /* While there is more to look for in both folds, see if they
3561 * continue to match */
3563 U8 fold_length = UTF8SKIP(f1);
3564 if (fold_length != UTF8SKIP(f2)
3565 || (fold_length == 1 && *f1 != *f2) /* Short circuit memNE
3566 function call for single
3568 || memNE((char*)f1, (char*)f2, fold_length))
3570 return 0; /* mismatch */
3573 /* Here, they matched, advance past them */
3580 /* When reach the end of any fold, advance the input past it */
3582 p1 += u1 ? UTF8SKIP(p1) : 1;
3585 p2 += u2 ? UTF8SKIP(p2) : 1;
3587 } /* End of loop through both strings */
3589 /* A match is defined by each scan that specified an explicit length
3590 * reaching its final goal, and the other not having matched a partial
3591 * character (which can happen when the fold of a character is more than one
3593 if (! ((g1 == 0 || p1 == g1) && (g2 == 0 || p2 == g2)) || n1 || n2) {
3597 /* Successful match. Set output pointers */
3609 * c-indentation-style: bsd
3611 * indent-tabs-mode: t
3614 * ex: set ts=8 sts=4 sw=4 noet: