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);
1318 Perl__to_upper_title_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp, const char S_or_s)
1320 /* We have the latin1-range values compiled into the core, so just use
1321 * those, converting the result to utf8. The only difference between upper
1322 * and title case in this range is that LATIN_SMALL_LETTER_SHARP_S is
1323 * either "SS" or "Ss". Which one to use is passed into the routine in
1324 * 'S_or_s' to avoid a test */
1326 UV converted = toUPPER_LATIN1_MOD(c);
1328 PERL_ARGS_ASSERT__TO_UPPER_TITLE_LATIN1;
1330 assert(S_or_s == 'S' || S_or_s == 's');
1332 if (UNI_IS_INVARIANT(converted)) { /* No difference between the two for
1333 characters in this range */
1334 *p = (U8) converted;
1339 /* toUPPER_LATIN1_MOD gives the correct results except for three outliers,
1340 * which it maps to one of them, so as to only have to have one check for
1341 * it in the main case */
1342 if (UNLIKELY(converted == LATIN_SMALL_LETTER_Y_WITH_DIAERESIS)) {
1344 case LATIN_SMALL_LETTER_Y_WITH_DIAERESIS:
1345 converted = LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS;
1348 converted = GREEK_CAPITAL_LETTER_MU;
1350 case LATIN_SMALL_LETTER_SHARP_S:
1356 Perl_croak(aTHX_ "panic: to_upper_title_latin1 did not expect '%c' to map to '%c'", c, LATIN_SMALL_LETTER_Y_WITH_DIAERESIS);
1361 *(p)++ = UTF8_TWO_BYTE_HI(converted);
1362 *p = UTF8_TWO_BYTE_LO(converted);
1368 /* Call the function to convert a UTF-8 encoded character to the specified case.
1369 * Note that there may be more than one character in the result.
1370 * INP is a pointer to the first byte of the input character
1371 * OUTP will be set to the first byte of the string of changed characters. It
1372 * needs to have space for UTF8_MAXBYTES_CASE+1 bytes
1373 * LENP will be set to the length in bytes of the string of changed characters
1375 * The functions return the ordinal of the first character in the string of OUTP */
1376 #define CALL_UPPER_CASE(INP, OUTP, LENP) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_toupper, "ToUc", "utf8::ToSpecUpper")
1377 #define CALL_TITLE_CASE(INP, OUTP, LENP) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_totitle, "ToTc", "utf8::ToSpecTitle")
1378 #define CALL_LOWER_CASE(INP, OUTP, LENP) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_tolower, "ToLc", "utf8::ToSpecLower")
1380 /* This additionally has the input parameter SPECIALS, which if non-zero will
1381 * cause this to use the SPECIALS hash for folding (meaning get full case
1382 * folding); otherwise, when zero, this implies a simple case fold */
1383 #define CALL_FOLD_CASE(INP, OUTP, LENP, SPECIALS) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_tofold, "ToCf", (SPECIALS) ? "utf8::ToSpecFold" : NULL)
1386 Perl_to_uni_upper(pTHX_ UV c, U8* p, STRLEN *lenp)
1390 /* Convert the Unicode character whose ordinal is c to its uppercase
1391 * version and store that in UTF-8 in p and its length in bytes in lenp.
1392 * Note that the p needs to be at least UTF8_MAXBYTES_CASE+1 bytes since
1393 * the changed version may be longer than the original character.
1395 * The ordinal of the first character of the changed version is returned
1396 * (but note, as explained above, that there may be more.) */
1398 PERL_ARGS_ASSERT_TO_UNI_UPPER;
1401 return _to_upper_title_latin1((U8) c, p, lenp, 'S');
1404 uvchr_to_utf8(p, c);
1405 return CALL_UPPER_CASE(p, p, lenp);
1409 Perl_to_uni_title(pTHX_ UV c, U8* p, STRLEN *lenp)
1413 PERL_ARGS_ASSERT_TO_UNI_TITLE;
1416 return _to_upper_title_latin1((U8) c, p, lenp, 's');
1419 uvchr_to_utf8(p, c);
1420 return CALL_TITLE_CASE(p, p, lenp);
1424 S_to_lower_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp)
1426 /* We have the latin1-range values compiled into the core, so just use
1427 * those, converting the result to utf8. Since the result is always just
1428 * one character, we allow p to be NULL */
1430 U8 converted = toLOWER_LATIN1(c);
1433 if (UNI_IS_INVARIANT(converted)) {
1438 *p = UTF8_TWO_BYTE_HI(converted);
1439 *(p+1) = UTF8_TWO_BYTE_LO(converted);
1447 Perl_to_uni_lower(pTHX_ UV c, U8* p, STRLEN *lenp)
1451 PERL_ARGS_ASSERT_TO_UNI_LOWER;
1454 return to_lower_latin1((U8) c, p, lenp);
1457 uvchr_to_utf8(p, c);
1458 return CALL_LOWER_CASE(p, p, lenp);
1462 Perl__to_fold_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp, const U8 flags)
1466 PERL_ARGS_ASSERT__TO_FOLD_LATIN1;
1468 if (c == MICRO_SIGN) {
1469 converted = GREEK_SMALL_LETTER_MU;
1471 else if (flags && c == LATIN_SMALL_LETTER_SHARP_S) {
1477 else { /* In this range the fold of all other characters is their lower
1479 converted = toLOWER_LATIN1(c);
1482 if (UNI_IS_INVARIANT(converted)) {
1483 *p = (U8) converted;
1487 *(p)++ = UTF8_TWO_BYTE_HI(converted);
1488 *p = UTF8_TWO_BYTE_LO(converted);
1496 Perl__to_uni_fold_flags(pTHX_ UV c, U8* p, STRLEN *lenp, U8 flags)
1498 PERL_ARGS_ASSERT__TO_UNI_FOLD_FLAGS;
1501 return _to_fold_latin1((U8) c, p, lenp, flags);
1504 uvchr_to_utf8(p, c);
1505 return CALL_FOLD_CASE(p, p, lenp, flags);
1508 /* for now these all assume no locale info available for Unicode > 255 */
1511 Perl_is_uni_alnum_lc(pTHX_ UV c)
1513 return is_uni_alnum(c); /* XXX no locale support yet */
1517 Perl_is_uni_idfirst_lc(pTHX_ UV c)
1519 return is_uni_idfirst(c); /* XXX no locale support yet */
1523 Perl_is_uni_alpha_lc(pTHX_ UV c)
1525 return is_uni_alpha(c); /* XXX no locale support yet */
1529 Perl_is_uni_ascii_lc(pTHX_ UV c)
1531 return is_uni_ascii(c); /* XXX no locale support yet */
1535 Perl_is_uni_space_lc(pTHX_ UV c)
1537 return is_uni_space(c); /* XXX no locale support yet */
1541 Perl_is_uni_digit_lc(pTHX_ UV c)
1543 return is_uni_digit(c); /* XXX no locale support yet */
1547 Perl_is_uni_upper_lc(pTHX_ UV c)
1549 return is_uni_upper(c); /* XXX no locale support yet */
1553 Perl_is_uni_lower_lc(pTHX_ UV c)
1555 return is_uni_lower(c); /* XXX no locale support yet */
1559 Perl_is_uni_cntrl_lc(pTHX_ UV c)
1561 return is_uni_cntrl(c); /* XXX no locale support yet */
1565 Perl_is_uni_graph_lc(pTHX_ UV c)
1567 return is_uni_graph(c); /* XXX no locale support yet */
1571 Perl_is_uni_print_lc(pTHX_ UV c)
1573 return is_uni_print(c); /* XXX no locale support yet */
1577 Perl_is_uni_punct_lc(pTHX_ UV c)
1579 return is_uni_punct(c); /* XXX no locale support yet */
1583 Perl_is_uni_xdigit_lc(pTHX_ UV c)
1585 return is_uni_xdigit(c); /* XXX no locale support yet */
1589 Perl_to_uni_upper_lc(pTHX_ U32 c)
1591 /* XXX returns only the first character -- do not use XXX */
1592 /* XXX no locale support yet */
1594 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1595 return (U32)to_uni_upper(c, tmpbuf, &len);
1599 Perl_to_uni_title_lc(pTHX_ U32 c)
1601 /* XXX returns only the first character XXX -- do not use XXX */
1602 /* XXX no locale support yet */
1604 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1605 return (U32)to_uni_title(c, tmpbuf, &len);
1609 Perl_to_uni_lower_lc(pTHX_ U32 c)
1611 /* XXX returns only the first character -- do not use XXX */
1612 /* XXX no locale support yet */
1614 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
1615 return (U32)to_uni_lower(c, tmpbuf, &len);
1619 S_is_utf8_common(pTHX_ const U8 *const p, SV **swash,
1620 const char *const swashname)
1624 PERL_ARGS_ASSERT_IS_UTF8_COMMON;
1626 if (!is_utf8_char(p))
1629 *swash = swash_init("utf8", swashname, &PL_sv_undef, 1, 0);
1630 return swash_fetch(*swash, p, TRUE) != 0;
1634 Perl_is_utf8_alnum(pTHX_ const U8 *p)
1638 PERL_ARGS_ASSERT_IS_UTF8_ALNUM;
1640 /* NOTE: "IsWord", not "IsAlnum", since Alnum is a true
1641 * descendant of isalnum(3), in other words, it doesn't
1642 * contain the '_'. --jhi */
1643 return is_utf8_common(p, &PL_utf8_alnum, "IsWord");
1647 Perl_is_utf8_idfirst(pTHX_ const U8 *p) /* The naming is historical. */
1651 PERL_ARGS_ASSERT_IS_UTF8_IDFIRST;
1655 /* is_utf8_idstart would be more logical. */
1656 return is_utf8_common(p, &PL_utf8_idstart, "IdStart");
1660 Perl_is_utf8_xidfirst(pTHX_ const U8 *p) /* The naming is historical. */
1664 PERL_ARGS_ASSERT_IS_UTF8_XIDFIRST;
1668 /* is_utf8_idstart would be more logical. */
1669 return is_utf8_common(p, &PL_utf8_xidstart, "XIdStart");
1673 Perl__is_utf8__perl_idstart(pTHX_ const U8 *p)
1677 PERL_ARGS_ASSERT__IS_UTF8__PERL_IDSTART;
1679 return is_utf8_common(p, &PL_utf8_perl_idstart, "_Perl_IDStart");
1683 Perl_is_utf8_idcont(pTHX_ const U8 *p)
1687 PERL_ARGS_ASSERT_IS_UTF8_IDCONT;
1689 return is_utf8_common(p, &PL_utf8_idcont, "IdContinue");
1693 Perl_is_utf8_xidcont(pTHX_ const U8 *p)
1697 PERL_ARGS_ASSERT_IS_UTF8_XIDCONT;
1699 return is_utf8_common(p, &PL_utf8_idcont, "XIdContinue");
1703 Perl_is_utf8_alpha(pTHX_ const U8 *p)
1707 PERL_ARGS_ASSERT_IS_UTF8_ALPHA;
1709 return is_utf8_common(p, &PL_utf8_alpha, "IsAlpha");
1713 Perl_is_utf8_ascii(pTHX_ const U8 *p)
1717 PERL_ARGS_ASSERT_IS_UTF8_ASCII;
1719 /* ASCII characters are the same whether in utf8 or not. So the macro
1720 * works on both utf8 and non-utf8 representations. */
1725 Perl_is_utf8_space(pTHX_ const U8 *p)
1729 PERL_ARGS_ASSERT_IS_UTF8_SPACE;
1731 return is_utf8_common(p, &PL_utf8_space, "IsXPerlSpace");
1735 Perl_is_utf8_perl_space(pTHX_ const U8 *p)
1739 PERL_ARGS_ASSERT_IS_UTF8_PERL_SPACE;
1741 /* Only true if is an ASCII space-like character, and ASCII is invariant
1742 * under utf8, so can just use the macro */
1743 return isSPACE_A(*p);
1747 Perl_is_utf8_perl_word(pTHX_ const U8 *p)
1751 PERL_ARGS_ASSERT_IS_UTF8_PERL_WORD;
1753 /* Only true if is an ASCII word character, and ASCII is invariant
1754 * under utf8, so can just use the macro */
1755 return isWORDCHAR_A(*p);
1759 Perl_is_utf8_digit(pTHX_ const U8 *p)
1763 PERL_ARGS_ASSERT_IS_UTF8_DIGIT;
1765 return is_utf8_common(p, &PL_utf8_digit, "IsDigit");
1769 Perl_is_utf8_posix_digit(pTHX_ const U8 *p)
1773 PERL_ARGS_ASSERT_IS_UTF8_POSIX_DIGIT;
1775 /* Only true if is an ASCII digit character, and ASCII is invariant
1776 * under utf8, so can just use the macro */
1777 return isDIGIT_A(*p);
1781 Perl_is_utf8_upper(pTHX_ const U8 *p)
1785 PERL_ARGS_ASSERT_IS_UTF8_UPPER;
1787 return is_utf8_common(p, &PL_utf8_upper, "IsUppercase");
1791 Perl_is_utf8_lower(pTHX_ const U8 *p)
1795 PERL_ARGS_ASSERT_IS_UTF8_LOWER;
1797 return is_utf8_common(p, &PL_utf8_lower, "IsLowercase");
1801 Perl_is_utf8_cntrl(pTHX_ const U8 *p)
1805 PERL_ARGS_ASSERT_IS_UTF8_CNTRL;
1808 return isCNTRL_A(*p);
1811 /* All controls are in Latin1 */
1812 if (! UTF8_IS_DOWNGRADEABLE_START(*p)) {
1815 return isCNTRL_L1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)));
1819 Perl_is_utf8_graph(pTHX_ const U8 *p)
1823 PERL_ARGS_ASSERT_IS_UTF8_GRAPH;
1825 return is_utf8_common(p, &PL_utf8_graph, "IsGraph");
1829 Perl_is_utf8_print(pTHX_ const U8 *p)
1833 PERL_ARGS_ASSERT_IS_UTF8_PRINT;
1835 return is_utf8_common(p, &PL_utf8_print, "IsPrint");
1839 Perl_is_utf8_punct(pTHX_ const U8 *p)
1843 PERL_ARGS_ASSERT_IS_UTF8_PUNCT;
1845 return is_utf8_common(p, &PL_utf8_punct, "IsPunct");
1849 Perl_is_utf8_xdigit(pTHX_ const U8 *p)
1853 PERL_ARGS_ASSERT_IS_UTF8_XDIGIT;
1855 return is_utf8_common(p, &PL_utf8_xdigit, "IsXDigit");
1859 Perl_is_utf8_mark(pTHX_ const U8 *p)
1863 PERL_ARGS_ASSERT_IS_UTF8_MARK;
1865 return is_utf8_common(p, &PL_utf8_mark, "IsM");
1869 Perl_is_utf8_X_begin(pTHX_ const U8 *p)
1873 PERL_ARGS_ASSERT_IS_UTF8_X_BEGIN;
1875 return is_utf8_common(p, &PL_utf8_X_begin, "_X_Begin");
1879 Perl_is_utf8_X_extend(pTHX_ const U8 *p)
1883 PERL_ARGS_ASSERT_IS_UTF8_X_EXTEND;
1885 return is_utf8_common(p, &PL_utf8_X_extend, "_X_Extend");
1889 Perl_is_utf8_X_prepend(pTHX_ const U8 *p)
1893 PERL_ARGS_ASSERT_IS_UTF8_X_PREPEND;
1895 return is_utf8_common(p, &PL_utf8_X_prepend, "GCB=Prepend");
1899 Perl_is_utf8_X_non_hangul(pTHX_ const U8 *p)
1903 PERL_ARGS_ASSERT_IS_UTF8_X_NON_HANGUL;
1905 return is_utf8_common(p, &PL_utf8_X_non_hangul, "HST=Not_Applicable");
1909 Perl_is_utf8_X_L(pTHX_ const U8 *p)
1913 PERL_ARGS_ASSERT_IS_UTF8_X_L;
1915 return is_utf8_common(p, &PL_utf8_X_L, "GCB=L");
1919 Perl_is_utf8_X_LV(pTHX_ const U8 *p)
1923 PERL_ARGS_ASSERT_IS_UTF8_X_LV;
1925 return is_utf8_common(p, &PL_utf8_X_LV, "GCB=LV");
1929 Perl_is_utf8_X_LVT(pTHX_ const U8 *p)
1933 PERL_ARGS_ASSERT_IS_UTF8_X_LVT;
1935 return is_utf8_common(p, &PL_utf8_X_LVT, "GCB=LVT");
1939 Perl_is_utf8_X_T(pTHX_ const U8 *p)
1943 PERL_ARGS_ASSERT_IS_UTF8_X_T;
1945 return is_utf8_common(p, &PL_utf8_X_T, "GCB=T");
1949 Perl_is_utf8_X_V(pTHX_ const U8 *p)
1953 PERL_ARGS_ASSERT_IS_UTF8_X_V;
1955 return is_utf8_common(p, &PL_utf8_X_V, "GCB=V");
1959 Perl_is_utf8_X_LV_LVT_V(pTHX_ const U8 *p)
1963 PERL_ARGS_ASSERT_IS_UTF8_X_LV_LVT_V;
1965 return is_utf8_common(p, &PL_utf8_X_LV_LVT_V, "_X_LV_LVT_V");
1969 =for apidoc to_utf8_case
1971 The "p" contains the pointer to the UTF-8 string encoding
1972 the character that is being converted.
1974 The "ustrp" is a pointer to the character buffer to put the
1975 conversion result to. The "lenp" is a pointer to the length
1978 The "swashp" is a pointer to the swash to use.
1980 Both the special and normal mappings are stored in lib/unicore/To/Foo.pl,
1981 and loaded by SWASHNEW, using lib/utf8_heavy.pl. The special (usually,
1982 but not always, a multicharacter mapping), is tried first.
1984 The "special" is a string like "utf8::ToSpecLower", which means the
1985 hash %utf8::ToSpecLower. The access to the hash is through
1986 Perl_to_utf8_case().
1988 The "normal" is a string like "ToLower" which means the swash
1994 Perl_to_utf8_case(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp,
1995 SV **swashp, const char *normal, const char *special)
1998 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
2000 const UV uv0 = utf8_to_uvchr(p, NULL);
2001 /* The NATIVE_TO_UNI() and UNI_TO_NATIVE() mappings
2002 * are necessary in EBCDIC, they are redundant no-ops
2003 * in ASCII-ish platforms, and hopefully optimized away. */
2004 const UV uv1 = NATIVE_TO_UNI(uv0);
2006 PERL_ARGS_ASSERT_TO_UTF8_CASE;
2008 /* Note that swash_fetch() doesn't output warnings for these because it
2009 * assumes we will */
2010 if (uv1 >= UNICODE_SURROGATE_FIRST) {
2011 if (uv1 <= UNICODE_SURROGATE_LAST) {
2012 if (ckWARN_d(WARN_SURROGATE)) {
2013 const char* desc = (PL_op) ? OP_DESC(PL_op) : normal;
2014 Perl_warner(aTHX_ packWARN(WARN_SURROGATE),
2015 "Operation \"%s\" returns its argument for UTF-16 surrogate U+%04"UVXf"", desc, uv1);
2018 else if (UNICODE_IS_SUPER(uv1)) {
2019 if (ckWARN_d(WARN_NON_UNICODE)) {
2020 const char* desc = (PL_op) ? OP_DESC(PL_op) : normal;
2021 Perl_warner(aTHX_ packWARN(WARN_NON_UNICODE),
2022 "Operation \"%s\" returns its argument for non-Unicode code point 0x%04"UVXf"", desc, uv1);
2026 /* Note that non-characters are perfectly legal, so no warning should
2030 uvuni_to_utf8(tmpbuf, uv1);
2032 if (!*swashp) /* load on-demand */
2033 *swashp = swash_init("utf8", normal, &PL_sv_undef, 4, 0);
2036 /* It might be "special" (sometimes, but not always,
2037 * a multicharacter mapping) */
2038 HV * const hv = get_hv(special, 0);
2042 (svp = hv_fetch(hv, (const char*)tmpbuf, UNISKIP(uv1), FALSE)) &&
2046 s = SvPV_const(*svp, len);
2048 len = uvuni_to_utf8(ustrp, NATIVE_TO_UNI(*(U8*)s)) - ustrp;
2051 /* If we have EBCDIC we need to remap the characters
2052 * since any characters in the low 256 are Unicode
2053 * code points, not EBCDIC. */
2054 U8 *t = (U8*)s, *tend = t + len, *d;
2061 const UV c = utf8_to_uvchr(t, &tlen);
2063 d = uvchr_to_utf8(d, UNI_TO_NATIVE(c));
2072 d = uvchr_to_utf8(d, UNI_TO_NATIVE(*t));
2077 Copy(tmpbuf, ustrp, len, U8);
2079 Copy(s, ustrp, len, U8);
2085 if (!len && *swashp) {
2086 const UV uv2 = swash_fetch(*swashp, tmpbuf, TRUE);
2089 /* It was "normal" (a single character mapping). */
2090 const UV uv3 = UNI_TO_NATIVE(uv2);
2091 len = uvchr_to_utf8(ustrp, uv3) - ustrp;
2095 if (!len) /* Neither: just copy. In other words, there was no mapping
2096 defined, which means that the code point maps to itself */
2097 len = uvchr_to_utf8(ustrp, uv0) - ustrp;
2102 return len ? utf8_to_uvchr(ustrp, 0) : 0;
2106 =for apidoc to_utf8_upper
2108 Convert the UTF-8 encoded character at p to its uppercase version and
2109 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
2110 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since
2111 the uppercase version may be longer than the original character.
2113 The first character of the uppercased version is returned
2114 (but note, as explained above, that there may be more.)
2119 Perl_to_utf8_upper(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
2123 PERL_ARGS_ASSERT_TO_UTF8_UPPER;
2125 if (UTF8_IS_INVARIANT(*p)) {
2126 return _to_upper_title_latin1(*p, ustrp, lenp, 'S');
2128 else if UTF8_IS_DOWNGRADEABLE_START(*p) {
2129 return _to_upper_title_latin1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)),
2133 return CALL_UPPER_CASE(p, ustrp, lenp);
2137 =for apidoc to_utf8_title
2139 Convert the UTF-8 encoded character at p to its titlecase version and
2140 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
2141 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
2142 titlecase version may be longer than the original character.
2144 The first character of the titlecased version is returned
2145 (but note, as explained above, that there may be more.)
2150 Perl_to_utf8_title(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
2154 PERL_ARGS_ASSERT_TO_UTF8_TITLE;
2156 if (UTF8_IS_INVARIANT(*p)) {
2157 return _to_upper_title_latin1(*p, ustrp, lenp, 's');
2159 else if UTF8_IS_DOWNGRADEABLE_START(*p) {
2160 return _to_upper_title_latin1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)),
2164 return CALL_TITLE_CASE(p, ustrp, lenp);
2168 =for apidoc to_utf8_lower
2170 Convert the UTF-8 encoded character at p to its lowercase version and
2171 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
2172 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
2173 lowercase version may be longer than the original character.
2175 The first character of the lowercased version is returned
2176 (but note, as explained above, that there may be more.)
2181 Perl_to_utf8_lower(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp)
2185 PERL_ARGS_ASSERT_TO_UTF8_LOWER;
2187 if (UTF8_IS_INVARIANT(*p)) {
2188 return to_lower_latin1(*p, ustrp, lenp);
2190 else if UTF8_IS_DOWNGRADEABLE_START(*p) {
2191 return to_lower_latin1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)), ustrp, lenp);
2194 return CALL_LOWER_CASE(p, ustrp, lenp);
2198 =for apidoc to_utf8_fold
2200 Convert the UTF-8 encoded character at p to its foldcase version and
2201 store that in UTF-8 in ustrp and its length in bytes in lenp. Note
2202 that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the
2203 foldcase version may be longer than the original character (up to
2206 The first character of the foldcased version is returned
2207 (but note, as explained above, that there may be more.)
2211 /* Not currently externally documented is 'flags', which currently is non-zero
2212 * if full case folds are to be used; otherwise simple folds */
2215 Perl__to_utf8_fold_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, U8 flags)
2219 PERL_ARGS_ASSERT__TO_UTF8_FOLD_FLAGS;
2221 if (UTF8_IS_INVARIANT(*p)) {
2222 return _to_fold_latin1(*p, ustrp, lenp, flags);
2224 else if UTF8_IS_DOWNGRADEABLE_START(*p) {
2225 return _to_fold_latin1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)),
2226 ustrp, lenp, flags);
2229 return CALL_FOLD_CASE(p, ustrp, lenp, flags);
2233 * Returns a "swash" which is a hash described in utf8.c:S_swash_fetch().
2234 * C<pkg> is a pointer to a package name for SWASHNEW, should be "utf8".
2235 * For other parameters, see utf8::SWASHNEW in lib/utf8_heavy.pl.
2238 Perl_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv, I32 minbits, I32 none)
2243 const size_t pkg_len = strlen(pkg);
2244 const size_t name_len = strlen(name);
2245 HV * const stash = gv_stashpvn(pkg, pkg_len, 0);
2249 PERL_ARGS_ASSERT_SWASH_INIT;
2251 PUSHSTACKi(PERLSI_MAGIC);
2255 if (PL_parser && PL_parser->error_count)
2256 SAVEI8(PL_parser->error_count), PL_parser->error_count = 0;
2257 method = gv_fetchmeth(stash, "SWASHNEW", 8, -1);
2258 if (!method) { /* demand load utf8 */
2260 errsv_save = newSVsv(ERRSV);
2261 /* It is assumed that callers of this routine are not passing in any
2262 user derived data. */
2263 /* Need to do this after save_re_context() as it will set PL_tainted to
2264 1 while saving $1 etc (see the code after getrx: in Perl_magic_get).
2265 Even line to create errsv_save can turn on PL_tainted. */
2266 SAVEBOOL(PL_tainted);
2268 Perl_load_module(aTHX_ PERL_LOADMOD_NOIMPORT, newSVpvn(pkg,pkg_len),
2271 sv_setsv(ERRSV, errsv_save);
2272 SvREFCNT_dec(errsv_save);
2278 mPUSHp(pkg, pkg_len);
2279 mPUSHp(name, name_len);
2284 errsv_save = newSVsv(ERRSV);
2285 /* If we already have a pointer to the method, no need to use call_method()
2286 to repeat the lookup. */
2287 if (method ? call_sv(MUTABLE_SV(method), G_SCALAR)
2288 : call_sv(newSVpvs_flags("SWASHNEW", SVs_TEMP), G_SCALAR | G_METHOD))
2289 retval = newSVsv(*PL_stack_sp--);
2291 retval = &PL_sv_undef;
2293 sv_setsv(ERRSV, errsv_save);
2294 SvREFCNT_dec(errsv_save);
2297 if (IN_PERL_COMPILETIME) {
2298 CopHINTS_set(PL_curcop, PL_hints);
2300 if (!SvROK(retval) || SvTYPE(SvRV(retval)) != SVt_PVHV) {
2302 Perl_croak(aTHX_ "Can't find Unicode property definition \"%"SVf"\"",
2304 Perl_croak(aTHX_ "SWASHNEW didn't return an HV ref");
2310 /* This API is wrong for special case conversions since we may need to
2311 * return several Unicode characters for a single Unicode character
2312 * (see lib/unicore/SpecCase.txt) The SWASHGET in lib/utf8_heavy.pl is
2313 * the lower-level routine, and it is similarly broken for returning
2314 * multiple values. --jhi
2315 * For those, you should use to_utf8_case() instead */
2316 /* Now SWASHGET is recasted into S_swash_get in this file. */
2319 * Returns the value of property/mapping C<swash> for the first character
2320 * of the string C<ptr>. If C<do_utf8> is true, the string C<ptr> is
2321 * assumed to be in utf8. If C<do_utf8> is false, the string C<ptr> is
2322 * assumed to be in native 8-bit encoding. Caches the swatch in C<swash>.
2324 * A "swash" is a hash which contains initially the keys/values set up by
2325 * SWASHNEW. The purpose is to be able to completely represent a Unicode
2326 * property for all possible code points. Things are stored in a compact form
2327 * (see utf8_heavy.pl) so that calculation is required to find the actual
2328 * property value for a given code point. As code points are looked up, new
2329 * key/value pairs are added to the hash, so that the calculation doesn't have
2330 * to ever be re-done. Further, each calculation is done, not just for the
2331 * desired one, but for a whole block of code points adjacent to that one.
2332 * For binary properties on ASCII machines, the block is usually for 64 code
2333 * points, starting with a code point evenly divisible by 64. Thus if the
2334 * property value for code point 257 is requested, the code goes out and
2335 * calculates the property values for all 64 code points between 256 and 319,
2336 * and stores these as a single 64-bit long bit vector, called a "swatch",
2337 * under the key for code point 256. The key is the UTF-8 encoding for code
2338 * point 256, minus the final byte. Thus, if the length of the UTF-8 encoding
2339 * for a code point is 13 bytes, the key will be 12 bytes long. If the value
2340 * for code point 258 is then requested, this code realizes that it would be
2341 * stored under the key for 256, and would find that value and extract the
2342 * relevant bit, offset from 256.
2344 * Non-binary properties are stored in as many bits as necessary to represent
2345 * their values (32 currently, though the code is more general than that), not
2346 * as single bits, but the principal is the same: the value for each key is a
2347 * vector that encompasses the property values for all code points whose UTF-8
2348 * representations are represented by the key. That is, for all code points
2349 * whose UTF-8 representations are length N bytes, and the key is the first N-1
2353 Perl_swash_fetch(pTHX_ SV *swash, const U8 *ptr, bool do_utf8)
2356 HV *const hv = MUTABLE_HV(SvRV(swash));
2361 const U8 *tmps = NULL;
2365 const UV c = NATIVE_TO_ASCII(*ptr);
2367 PERL_ARGS_ASSERT_SWASH_FETCH;
2369 if (!do_utf8 && !UNI_IS_INVARIANT(c)) {
2370 tmputf8[0] = (U8)UTF8_EIGHT_BIT_HI(c);
2371 tmputf8[1] = (U8)UTF8_EIGHT_BIT_LO(c);
2374 /* Given a UTF-X encoded char 0xAA..0xYY,0xZZ
2375 * then the "swatch" is a vec() for all the chars which start
2377 * So the key in the hash (klen) is length of encoded char -1
2379 klen = UTF8SKIP(ptr) - 1;
2383 /* If char is invariant then swatch is for all the invariant chars
2384 * In both UTF-8 and UTF-8-MOD that happens to be UTF_CONTINUATION_MARK
2386 needents = UTF_CONTINUATION_MARK;
2387 off = NATIVE_TO_UTF(ptr[klen]);
2390 /* If char is encoded then swatch is for the prefix */
2391 needents = (1 << UTF_ACCUMULATION_SHIFT);
2392 off = NATIVE_TO_UTF(ptr[klen]) & UTF_CONTINUATION_MASK;
2396 * This single-entry cache saves about 1/3 of the utf8 overhead in test
2397 * suite. (That is, only 7-8% overall over just a hash cache. Still,
2398 * it's nothing to sniff at.) Pity we usually come through at least
2399 * two function calls to get here...
2401 * NB: this code assumes that swatches are never modified, once generated!
2404 if (hv == PL_last_swash_hv &&
2405 klen == PL_last_swash_klen &&
2406 (!klen || memEQ((char *)ptr, (char *)PL_last_swash_key, klen)) )
2408 tmps = PL_last_swash_tmps;
2409 slen = PL_last_swash_slen;
2412 /* Try our second-level swatch cache, kept in a hash. */
2413 SV** svp = hv_fetch(hv, (const char*)ptr, klen, FALSE);
2415 /* If not cached, generate it via swash_get */
2416 if (!svp || !SvPOK(*svp)
2417 || !(tmps = (const U8*)SvPV_const(*svp, slen))) {
2418 /* We use utf8n_to_uvuni() as we want an index into
2419 Unicode tables, not a native character number.
2421 const UV code_point = utf8n_to_uvuni(ptr, UTF8_MAXBYTES, 0,
2423 0 : UTF8_ALLOW_ANY);
2424 swatch = swash_get(swash,
2425 /* On EBCDIC & ~(0xA0-1) isn't a useful thing to do */
2426 (klen) ? (code_point & ~((UV)needents - 1)) : 0,
2429 if (IN_PERL_COMPILETIME)
2430 CopHINTS_set(PL_curcop, PL_hints);
2432 svp = hv_store(hv, (const char *)ptr, klen, swatch, 0);
2434 if (!svp || !(tmps = (U8*)SvPV(*svp, slen))
2435 || (slen << 3) < needents)
2436 Perl_croak(aTHX_ "panic: swash_fetch got improper swatch");
2439 PL_last_swash_hv = hv;
2440 assert(klen <= sizeof(PL_last_swash_key));
2441 PL_last_swash_klen = (U8)klen;
2442 /* FIXME change interpvar.h? */
2443 PL_last_swash_tmps = (U8 *) tmps;
2444 PL_last_swash_slen = slen;
2446 Copy(ptr, PL_last_swash_key, klen, U8);
2449 if (UTF8_IS_SUPER(ptr) && ckWARN_d(WARN_NON_UNICODE)) {
2450 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
2452 /* This outputs warnings for binary properties only, assuming that
2453 * to_utf8_case() will output any for non-binary. Also, surrogates
2454 * aren't checked for, as that would warn on things like /\p{Gc=Cs}/ */
2456 if (SvUV(*bitssvp) == 1) {
2457 /* User-defined properties can silently match above-Unicode */
2458 SV** const user_defined_svp = hv_fetchs(hv, "USER_DEFINED", FALSE);
2459 if (! user_defined_svp || ! SvUV(*user_defined_svp)) {
2460 const UV code_point = utf8n_to_uvuni(ptr, UTF8_MAXBYTES, 0, 0);
2461 Perl_warner(aTHX_ packWARN(WARN_NON_UNICODE),
2462 "Code point 0x%04"UVXf" is not Unicode, all \\p{} matches fail; all \\P{} matches succeed", code_point);
2467 switch ((int)((slen << 3) / needents)) {
2469 bit = 1 << (off & 7);
2471 return (tmps[off] & bit) != 0;
2476 return (tmps[off] << 8) + tmps[off + 1] ;
2479 return (tmps[off] << 24) + (tmps[off+1] << 16) + (tmps[off+2] << 8) + tmps[off + 3] ;
2481 Perl_croak(aTHX_ "panic: swash_fetch got swatch of unexpected bit width");
2482 NORETURN_FUNCTION_END;
2485 /* Read a single line of the main body of the swash input text. These are of
2488 * where each number is hex. The first two numbers form the minimum and
2489 * maximum of a range, and the third is the value associated with the range.
2490 * Not all swashes should have a third number
2492 * On input: l points to the beginning of the line to be examined; it points
2493 * to somewhere in the string of the whole input text, and is
2494 * terminated by a \n or the null string terminator.
2495 * lend points to the null terminator of that string
2496 * wants_value is non-zero if the swash expects a third number
2497 * typestr is the name of the swash's mapping, like 'ToLower'
2498 * On output: *min, *max, and *val are set to the values read from the line.
2499 * returns a pointer just beyond the line examined. If there was no
2500 * valid min number on the line, returns lend+1
2504 S_swash_scan_list_line(pTHX_ U8* l, U8* const lend, UV* min, UV* max, UV* val,
2505 const bool wants_value, const U8* const typestr)
2507 const int typeto = typestr[0] == 'T' && typestr[1] == 'o';
2508 STRLEN numlen; /* Length of the number */
2509 I32 flags = PERL_SCAN_SILENT_ILLDIGIT
2510 | PERL_SCAN_DISALLOW_PREFIX
2511 | PERL_SCAN_SILENT_NON_PORTABLE;
2513 /* nl points to the next \n in the scan */
2514 U8* const nl = (U8*)memchr(l, '\n', lend - l);
2516 /* Get the first number on the line: the range minimum */
2518 *min = grok_hex((char *)l, &numlen, &flags, NULL);
2519 if (numlen) /* If found a hex number, position past it */
2521 else if (nl) { /* Else, go handle next line, if any */
2522 return nl + 1; /* 1 is length of "\n" */
2524 else { /* Else, no next line */
2525 return lend + 1; /* to LIST's end at which \n is not found */
2528 /* The max range value follows, separated by a BLANK */
2531 flags = PERL_SCAN_SILENT_ILLDIGIT
2532 | PERL_SCAN_DISALLOW_PREFIX
2533 | PERL_SCAN_SILENT_NON_PORTABLE;
2535 *max = grok_hex((char *)l, &numlen, &flags, NULL);
2538 else /* If no value here, it is a single element range */
2541 /* Non-binary tables have a third entry: what the first element of the
2546 flags = PERL_SCAN_SILENT_ILLDIGIT
2547 | PERL_SCAN_DISALLOW_PREFIX
2548 | PERL_SCAN_SILENT_NON_PORTABLE;
2550 *val = grok_hex((char *)l, &numlen, &flags, NULL);
2559 Perl_croak(aTHX_ "%s: illegal mapping '%s'",
2565 *val = 0; /* bits == 1, then any val should be ignored */
2567 else { /* Nothing following range min, should be single element with no
2573 Perl_croak(aTHX_ "%s: illegal mapping '%s'", typestr, l);
2577 *val = 0; /* bits == 1, then val should be ignored */
2580 /* Position to next line if any, or EOF */
2590 * Returns a swatch (a bit vector string) for a code point sequence
2591 * that starts from the value C<start> and comprises the number C<span>.
2592 * A C<swash> must be an object created by SWASHNEW (see lib/utf8_heavy.pl).
2593 * Should be used via swash_fetch, which will cache the swatch in C<swash>.
2596 S_swash_get(pTHX_ SV* swash, UV start, UV span)
2599 U8 *l, *lend, *x, *xend, *s, *send;
2600 STRLEN lcur, xcur, scur;
2601 HV *const hv = MUTABLE_HV(SvRV(swash));
2603 /* The string containing the main body of the table */
2604 SV** const listsvp = hv_fetchs(hv, "LIST", FALSE);
2606 SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE);
2607 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
2608 SV** const nonesvp = hv_fetchs(hv, "NONE", FALSE);
2609 SV** const extssvp = hv_fetchs(hv, "EXTRAS", FALSE);
2610 SV** const invert_it_svp = hv_fetchs(hv, "INVERT_IT", FALSE);
2611 const U8* const typestr = (U8*)SvPV_nolen(*typesvp);
2612 const STRLEN bits = SvUV(*bitssvp);
2613 const STRLEN octets = bits >> 3; /* if bits == 1, then octets == 0 */
2614 const UV none = SvUV(*nonesvp);
2615 UV end = start + span;
2617 PERL_ARGS_ASSERT_SWASH_GET;
2619 if (bits != 1 && bits != 8 && bits != 16 && bits != 32) {
2620 Perl_croak(aTHX_ "panic: swash_get doesn't expect bits %"UVuf,
2624 /* If overflowed, use the max possible */
2630 /* create and initialize $swatch */
2631 scur = octets ? (span * octets) : (span + 7) / 8;
2632 swatch = newSV(scur);
2634 s = (U8*)SvPVX(swatch);
2635 if (octets && none) {
2636 const U8* const e = s + scur;
2639 *s++ = (U8)(none & 0xff);
2640 else if (bits == 16) {
2641 *s++ = (U8)((none >> 8) & 0xff);
2642 *s++ = (U8)( none & 0xff);
2644 else if (bits == 32) {
2645 *s++ = (U8)((none >> 24) & 0xff);
2646 *s++ = (U8)((none >> 16) & 0xff);
2647 *s++ = (U8)((none >> 8) & 0xff);
2648 *s++ = (U8)( none & 0xff);
2654 (void)memzero((U8*)s, scur + 1);
2656 SvCUR_set(swatch, scur);
2657 s = (U8*)SvPVX(swatch);
2659 /* read $swash->{LIST}. XXX Note that this is a linear scan through a
2660 * sorted list. A binary search would be much more efficient */
2661 l = (U8*)SvPV(*listsvp, lcur);
2665 l = S_swash_scan_list_line(aTHX_ l, lend, &min, &max, &val,
2666 cBOOL(octets), typestr);
2671 /* If looking for something beyond this range, go try the next one */
2678 if (!none || val < none) {
2683 for (key = min; key <= max; key++) {
2687 /* XXX If it should ever happen (very unlikely) that we would
2688 * want a non-binary result for the code point at UV_MAX,
2689 * special handling would need to be inserted here, as is done
2690 * below for the binary case */
2691 /* offset must be non-negative (start <= min <= key < end) */
2692 offset = octets * (key - start);
2694 s[offset] = (U8)(val & 0xff);
2695 else if (bits == 16) {
2696 s[offset ] = (U8)((val >> 8) & 0xff);
2697 s[offset + 1] = (U8)( val & 0xff);
2699 else if (bits == 32) {
2700 s[offset ] = (U8)((val >> 24) & 0xff);
2701 s[offset + 1] = (U8)((val >> 16) & 0xff);
2702 s[offset + 2] = (U8)((val >> 8) & 0xff);
2703 s[offset + 3] = (U8)( val & 0xff);
2706 if (!none || val < none)
2710 else { /* bits == 1, then val should be ignored */
2715 /* Special case when the upper-end is the highest possible code
2716 * point representable on the platform. Otherwise, the code below
2717 * exits before setting this bit. Done here to avoid testing for
2718 * this extremely unlikely possibility in the loop */
2719 if (UNLIKELY(end == UV_MAX && max == UV_MAX)) {
2720 const STRLEN offset = (STRLEN)(max - start);
2721 s[offset >> 3] |= 1 << (offset & 7);
2723 for (key = min; key <= max; key++) {
2724 const STRLEN offset = (STRLEN)(key - start);
2727 s[offset >> 3] |= 1 << (offset & 7);
2733 /* Invert if the data says it should be. Assumes that bits == 1 */
2734 if (invert_it_svp && SvUV(*invert_it_svp)) {
2736 /* Unicode properties should come with all bits above PERL_UNICODE_MAX
2737 * be 0, and their inversion should also be 0, as we don't succeed any
2738 * Unicode property matches for non-Unicode code points */
2739 if (start <= PERL_UNICODE_MAX) {
2741 /* The code below assumes that we never cross the
2742 * Unicode/above-Unicode boundary in a range, as otherwise we would
2743 * have to figure out where to stop flipping the bits. Since this
2744 * boundary is divisible by a large power of 2, and swatches comes
2745 * in small powers of 2, this should be a valid assumption */
2746 assert(start + span - 1 <= PERL_UNICODE_MAX);
2756 /* read $swash->{EXTRAS}
2757 * This code also copied to swash_to_invlist() below */
2758 x = (U8*)SvPV(*extssvp, xcur);
2766 SV **otherbitssvp, *other;
2770 const U8 opc = *x++;
2774 nl = (U8*)memchr(x, '\n', xend - x);
2776 if (opc != '-' && opc != '+' && opc != '!' && opc != '&') {
2778 x = nl + 1; /* 1 is length of "\n" */
2782 x = xend; /* to EXTRAS' end at which \n is not found */
2789 namelen = nl - namestr;
2793 namelen = xend - namestr;
2797 othersvp = hv_fetch(hv, (char *)namestr, namelen, FALSE);
2798 otherhv = MUTABLE_HV(SvRV(*othersvp));
2799 otherbitssvp = hv_fetchs(otherhv, "BITS", FALSE);
2800 otherbits = (STRLEN)SvUV(*otherbitssvp);
2801 if (bits < otherbits)
2802 Perl_croak(aTHX_ "panic: swash_get found swatch size mismatch");
2804 /* The "other" swatch must be destroyed after. */
2805 other = swash_get(*othersvp, start, span);
2806 o = (U8*)SvPV(other, olen);
2809 Perl_croak(aTHX_ "panic: swash_get got improper swatch");
2811 s = (U8*)SvPV(swatch, slen);
2812 if (bits == 1 && otherbits == 1) {
2814 Perl_croak(aTHX_ "panic: swash_get found swatch length mismatch");
2838 STRLEN otheroctets = otherbits >> 3;
2840 U8* const send = s + slen;
2845 if (otherbits == 1) {
2846 otherval = (o[offset >> 3] >> (offset & 7)) & 1;
2850 STRLEN vlen = otheroctets;
2858 if (opc == '+' && otherval)
2859 NOOP; /* replace with otherval */
2860 else if (opc == '!' && !otherval)
2862 else if (opc == '-' && otherval)
2864 else if (opc == '&' && !otherval)
2867 s += octets; /* no replacement */
2872 *s++ = (U8)( otherval & 0xff);
2873 else if (bits == 16) {
2874 *s++ = (U8)((otherval >> 8) & 0xff);
2875 *s++ = (U8)( otherval & 0xff);
2877 else if (bits == 32) {
2878 *s++ = (U8)((otherval >> 24) & 0xff);
2879 *s++ = (U8)((otherval >> 16) & 0xff);
2880 *s++ = (U8)((otherval >> 8) & 0xff);
2881 *s++ = (U8)( otherval & 0xff);
2885 sv_free(other); /* through with it! */
2891 Perl__swash_inversion_hash(pTHX_ SV* const swash)
2894 /* Subject to change or removal. For use only in one place in regcomp.c.
2895 * Can't be used on a property that is subject to user override, as it
2896 * relies on the value of SPECIALS in the swash which would be set by
2897 * utf8_heavy.pl to the hash in the non-overriden file, and hence is not set
2898 * for overridden properties
2900 * Returns a hash which is the inversion and closure of a swash mapping.
2901 * For example, consider the input lines:
2906 * The returned hash would have two keys, the utf8 for 006B and the utf8 for
2907 * 006C. The value for each key is an array. For 006C, the array would
2908 * have a two elements, the utf8 for itself, and for 004C. For 006B, there
2909 * would be three elements in its array, the utf8 for 006B, 004B and 212A.
2911 * Essentially, for any code point, it gives all the code points that map to
2912 * it, or the list of 'froms' for that point.
2914 * Currently it ignores any additions or deletions from other swashes,
2915 * looking at just the main body of the swash, and if there are SPECIALS
2916 * in the swash, at that hash
2918 * The specials hash can be extra code points, and most likely consists of
2919 * maps from single code points to multiple ones (each expressed as a string
2920 * of utf8 characters). This function currently returns only 1-1 mappings.
2921 * However consider this possible input in the specials hash:
2922 * "\xEF\xAC\x85" => "\x{0073}\x{0074}", # U+FB05 => 0073 0074
2923 * "\xEF\xAC\x86" => "\x{0073}\x{0074}", # U+FB06 => 0073 0074
2925 * Both FB05 and FB06 map to the same multi-char sequence, which we don't
2926 * currently handle. But it also means that FB05 and FB06 are equivalent in
2927 * a 1-1 mapping which we should handle, and this relationship may not be in
2928 * the main table. Therefore this function examines all the multi-char
2929 * sequences and adds the 1-1 mappings that come out of that. */
2933 HV *const hv = MUTABLE_HV(SvRV(swash));
2935 /* The string containing the main body of the table */
2936 SV** const listsvp = hv_fetchs(hv, "LIST", FALSE);
2938 SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE);
2939 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
2940 SV** const nonesvp = hv_fetchs(hv, "NONE", FALSE);
2941 /*SV** const extssvp = hv_fetchs(hv, "EXTRAS", FALSE);*/
2942 const U8* const typestr = (U8*)SvPV_nolen(*typesvp);
2943 const STRLEN bits = SvUV(*bitssvp);
2944 const STRLEN octets = bits >> 3; /* if bits == 1, then octets == 0 */
2945 const UV none = SvUV(*nonesvp);
2946 SV **specials_p = hv_fetchs(hv, "SPECIALS", 0);
2950 PERL_ARGS_ASSERT__SWASH_INVERSION_HASH;
2952 /* Must have at least 8 bits to get the mappings */
2953 if (bits != 8 && bits != 16 && bits != 32) {
2954 Perl_croak(aTHX_ "panic: swash_inversion_hash doesn't expect bits %"UVuf,
2958 if (specials_p) { /* It might be "special" (sometimes, but not always, a
2959 mapping to more than one character */
2961 /* Construct an inverse mapping hash for the specials */
2962 HV * const specials_hv = MUTABLE_HV(SvRV(*specials_p));
2963 HV * specials_inverse = newHV();
2964 char *char_from; /* the lhs of the map */
2965 I32 from_len; /* its byte length */
2966 char *char_to; /* the rhs of the map */
2967 I32 to_len; /* its byte length */
2968 SV *sv_to; /* and in a sv */
2969 AV* from_list; /* list of things that map to each 'to' */
2971 hv_iterinit(specials_hv);
2973 /* The keys are the characters (in utf8) that map to the corresponding
2974 * utf8 string value. Iterate through the list creating the inverse
2976 while ((sv_to = hv_iternextsv(specials_hv, &char_from, &from_len))) {
2978 if (! SvPOK(sv_to)) {
2979 Perl_croak(aTHX_ "panic: value returned from hv_iternextsv() unexpectedly is not a string");
2981 /*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)));*/
2983 /* Each key in the inverse list is a mapped-to value, and the key's
2984 * hash value is a list of the strings (each in utf8) that map to
2985 * it. Those strings are all one character long */
2986 if ((listp = hv_fetch(specials_inverse,
2990 from_list = (AV*) *listp;
2992 else { /* No entry yet for it: create one */
2993 from_list = newAV();
2994 if (! hv_store(specials_inverse,
2997 (SV*) from_list, 0))
2999 Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed");
3003 /* Here have the list associated with this 'to' (perhaps newly
3004 * created and empty). Just add to it. Note that we ASSUME that
3005 * the input is guaranteed to not have duplications, so we don't
3006 * check for that. Duplications just slow down execution time. */
3007 av_push(from_list, newSVpvn_utf8(char_from, from_len, TRUE));
3010 /* Here, 'specials_inverse' contains the inverse mapping. Go through
3011 * it looking for cases like the FB05/FB06 examples above. There would
3012 * be an entry in the hash like
3013 * 'st' => [ FB05, FB06 ]
3014 * In this example we will create two lists that get stored in the
3015 * returned hash, 'ret':
3016 * FB05 => [ FB05, FB06 ]
3017 * FB06 => [ FB05, FB06 ]
3019 * Note that there is nothing to do if the array only has one element.
3020 * (In the normal 1-1 case handled below, we don't have to worry about
3021 * two lists, as everything gets tied to the single list that is
3022 * generated for the single character 'to'. But here, we are omitting
3023 * that list, ('st' in the example), so must have multiple lists.) */
3024 while ((from_list = (AV *) hv_iternextsv(specials_inverse,
3025 &char_to, &to_len)))
3027 if (av_len(from_list) > 0) {
3030 /* We iterate over all combinations of i,j to place each code
3031 * point on each list */
3032 for (i = 0; i <= av_len(from_list); i++) {
3034 AV* i_list = newAV();
3035 SV** entryp = av_fetch(from_list, i, FALSE);
3036 if (entryp == NULL) {
3037 Perl_croak(aTHX_ "panic: av_fetch() unexpectedly failed");
3039 if (hv_fetch(ret, SvPVX(*entryp), SvCUR(*entryp), FALSE)) {
3040 Perl_croak(aTHX_ "panic: unexpected entry for %s", SvPVX(*entryp));
3042 if (! hv_store(ret, SvPVX(*entryp), SvCUR(*entryp),
3043 (SV*) i_list, FALSE))
3045 Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed");
3048 /* For debugging: UV u = utf8_to_uvchr((U8*) SvPVX(*entryp), 0);*/
3049 for (j = 0; j <= av_len(from_list); j++) {
3050 entryp = av_fetch(from_list, j, FALSE);
3051 if (entryp == NULL) {
3052 Perl_croak(aTHX_ "panic: av_fetch() unexpectedly failed");
3055 /* When i==j this adds itself to the list */
3056 av_push(i_list, newSVuv(utf8_to_uvchr(
3057 (U8*) SvPVX(*entryp), 0)));
3058 /*DEBUG_U(PerlIO_printf(Perl_debug_log, "Adding %"UVXf" to list for %"UVXf"\n", utf8_to_uvchr((U8*) SvPVX(*entryp), 0), u));*/
3063 SvREFCNT_dec(specials_inverse); /* done with it */
3064 } /* End of specials */
3066 /* read $swash->{LIST} */
3067 l = (U8*)SvPV(*listsvp, lcur);
3070 /* Go through each input line */
3074 l = S_swash_scan_list_line(aTHX_ l, lend, &min, &max, &val,
3075 cBOOL(octets), typestr);
3080 /* Each element in the range is to be inverted */
3081 for (inverse = min; inverse <= max; inverse++) {
3085 bool found_key = FALSE;
3086 bool found_inverse = FALSE;
3088 /* The key is the inverse mapping */
3089 char key[UTF8_MAXBYTES+1];
3090 char* key_end = (char *) uvuni_to_utf8((U8*) key, val);
3091 STRLEN key_len = key_end - key;
3093 /* Get the list for the map */
3094 if ((listp = hv_fetch(ret, key, key_len, FALSE))) {
3095 list = (AV*) *listp;
3097 else { /* No entry yet for it: create one */
3099 if (! hv_store(ret, key, key_len, (SV*) list, FALSE)) {
3100 Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed");
3104 /* Look through list to see if this inverse mapping already is
3105 * listed, or if there is a mapping to itself already */
3106 for (i = 0; i <= av_len(list); i++) {
3107 SV** entryp = av_fetch(list, i, FALSE);
3109 if (entryp == NULL) {
3110 Perl_croak(aTHX_ "panic: av_fetch() unexpectedly failed");
3113 /*DEBUG_U(PerlIO_printf(Perl_debug_log, "list for %"UVXf" contains %"UVXf"\n", val, SvUV(entry)));*/
3114 if (SvUV(entry) == val) {
3117 if (SvUV(entry) == inverse) {
3118 found_inverse = TRUE;
3121 /* No need to continue searching if found everything we are
3123 if (found_key && found_inverse) {
3128 /* Make sure there is a mapping to itself on the list */
3130 av_push(list, newSVuv(val));
3131 /*DEBUG_U(PerlIO_printf(Perl_debug_log, "Adding %"UVXf" to list for %"UVXf"\n", val, val));*/
3135 /* Simply add the value to the list */
3136 if (! found_inverse) {
3137 av_push(list, newSVuv(inverse));
3138 /*DEBUG_U(PerlIO_printf(Perl_debug_log, "Adding %"UVXf" to list for %"UVXf"\n", inverse, val));*/
3141 /* swash_get() increments the value of val for each element in the
3142 * range. That makes more compact tables possible. You can
3143 * express the capitalization, for example, of all consecutive
3144 * letters with a single line: 0061\t007A\t0041 This maps 0061 to
3145 * 0041, 0062 to 0042, etc. I (khw) have never understood 'none',
3146 * and it's not documented; it appears to be used only in
3147 * implementing tr//; I copied the semantics from swash_get(), just
3149 if (!none || val < none) {
3159 Perl__swash_to_invlist(pTHX_ SV* const swash)
3162 /* Subject to change or removal. For use only in one place in regcomp.c */
3167 HV *const hv = MUTABLE_HV(SvRV(swash));
3168 UV elements = 0; /* Number of elements in the inversion list */
3171 /* The string containing the main body of the table */
3172 SV** const listsvp = hv_fetchs(hv, "LIST", FALSE);
3173 SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE);
3174 SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
3175 SV** const extssvp = hv_fetchs(hv, "EXTRAS", FALSE);
3176 SV** const invert_it_svp = hv_fetchs(hv, "INVERT_IT", FALSE);
3178 const U8* const typestr = (U8*)SvPV_nolen(*typesvp);
3179 const STRLEN bits = SvUV(*bitssvp);
3180 const STRLEN octets = bits >> 3; /* if bits == 1, then octets == 0 */
3186 PERL_ARGS_ASSERT__SWASH_TO_INVLIST;
3188 /* read $swash->{LIST} */
3189 if (SvPOK(*listsvp)) {
3190 l = (U8*)SvPV(*listsvp, lcur);
3193 /* LIST legitimately doesn't contain a string during compilation phases
3194 * of Perl itself, before the Unicode tables are generated. In this
3195 * case, just fake things up by creating an empty list */
3202 /* Scan the input to count the number of lines to preallocate array size
3203 * based on worst possible case, which is each line in the input creates 2
3204 * elements in the inversion list: 1) the beginning of a range in the list;
3205 * 2) the beginning of a range not in the list. */
3206 while ((loc = (strchr(loc, '\n'))) != NULL) {
3211 /* If the ending is somehow corrupt and isn't a new line, add another
3212 * element for the final range that isn't in the inversion list */
3213 if (! (*lend == '\n' || (*lend == '\0' && *(lend - 1) == '\n'))) {
3217 invlist = _new_invlist(elements);
3219 /* Now go through the input again, adding each range to the list */
3222 UV val; /* Not used by this function */
3224 l = S_swash_scan_list_line(aTHX_ l, lend, &start, &end, &val,
3225 cBOOL(octets), typestr);
3231 _append_range_to_invlist(invlist, start, end);
3234 /* Invert if the data says it should be */
3235 if (invert_it_svp && SvUV(*invert_it_svp)) {
3236 _invlist_invert_prop(invlist);
3239 /* This code is copied from swash_get()
3240 * read $swash->{EXTRAS} */
3241 x = (U8*)SvPV(*extssvp, xcur);
3249 SV **otherbitssvp, *other;
3252 const U8 opc = *x++;
3256 nl = (U8*)memchr(x, '\n', xend - x);
3258 if (opc != '-' && opc != '+' && opc != '!' && opc != '&') {
3260 x = nl + 1; /* 1 is length of "\n" */
3264 x = xend; /* to EXTRAS' end at which \n is not found */
3271 namelen = nl - namestr;
3275 namelen = xend - namestr;
3279 othersvp = hv_fetch(hv, (char *)namestr, namelen, FALSE);
3280 otherhv = MUTABLE_HV(SvRV(*othersvp));
3281 otherbitssvp = hv_fetchs(otherhv, "BITS", FALSE);
3282 otherbits = (STRLEN)SvUV(*otherbitssvp);
3284 if (bits != otherbits || bits != 1) {
3285 Perl_croak(aTHX_ "panic: _swash_to_invlist only operates on boolean properties");
3288 /* The "other" swatch must be destroyed after. */
3289 other = _swash_to_invlist((SV *)*othersvp);
3291 /* End of code copied from swash_get() */
3294 _invlist_union(invlist, other, &invlist);
3297 _invlist_invert(other);
3298 _invlist_union(invlist, other, &invlist);
3301 _invlist_subtract(invlist, other, &invlist);
3304 _invlist_intersection(invlist, other, &invlist);
3309 sv_free(other); /* through with it! */
3316 =for apidoc uvchr_to_utf8
3318 Adds the UTF-8 representation of the Native code point C<uv> to the end
3319 of the string C<d>; C<d> should be have at least C<UTF8_MAXBYTES+1> free
3320 bytes available. The return value is the pointer to the byte after the
3321 end of the new character. In other words,
3323 d = uvchr_to_utf8(d, uv);
3325 is the recommended wide native character-aware way of saying
3332 /* On ASCII machines this is normally a macro but we want a
3333 real function in case XS code wants it
3336 Perl_uvchr_to_utf8(pTHX_ U8 *d, UV uv)
3338 PERL_ARGS_ASSERT_UVCHR_TO_UTF8;
3340 return Perl_uvuni_to_utf8_flags(aTHX_ d, NATIVE_TO_UNI(uv), 0);
3344 Perl_uvchr_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags)
3346 PERL_ARGS_ASSERT_UVCHR_TO_UTF8_FLAGS;
3348 return Perl_uvuni_to_utf8_flags(aTHX_ d, NATIVE_TO_UNI(uv), flags);
3352 =for apidoc utf8n_to_uvchr
3354 Returns the native character value of the first character in the string
3356 which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
3357 length, in bytes, of that character.
3359 length and flags are the same as utf8n_to_uvuni().
3363 /* On ASCII machines this is normally a macro but we want
3364 a real function in case XS code wants it
3367 Perl_utf8n_to_uvchr(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen,
3370 const UV uv = Perl_utf8n_to_uvuni(aTHX_ s, curlen, retlen, flags);
3372 PERL_ARGS_ASSERT_UTF8N_TO_UVCHR;
3374 return UNI_TO_NATIVE(uv);
3378 Perl_check_utf8_print(pTHX_ register const U8* s, const STRLEN len)
3380 /* May change: warns if surrogates, non-character code points, or
3381 * non-Unicode code points are in s which has length len bytes. Returns
3382 * TRUE if none found; FALSE otherwise. The only other validity check is
3383 * to make sure that this won't exceed the string's length */
3385 const U8* const e = s + len;
3388 PERL_ARGS_ASSERT_CHECK_UTF8_PRINT;
3391 if (UTF8SKIP(s) > len) {
3392 Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8),
3393 "%s in %s", unees, PL_op ? OP_DESC(PL_op) : "print");
3396 if (UNLIKELY(*s >= UTF8_FIRST_PROBLEMATIC_CODE_POINT_FIRST_BYTE)) {
3398 if (UTF8_IS_SUPER(s)) {
3399 if (ckWARN_d(WARN_NON_UNICODE)) {
3400 UV uv = utf8_to_uvchr(s, &char_len);
3401 Perl_warner(aTHX_ packWARN(WARN_NON_UNICODE),
3402 "Code point 0x%04"UVXf" is not Unicode, may not be portable", uv);
3406 else if (UTF8_IS_SURROGATE(s)) {
3407 if (ckWARN_d(WARN_SURROGATE)) {
3408 UV uv = utf8_to_uvchr(s, &char_len);
3409 Perl_warner(aTHX_ packWARN(WARN_SURROGATE),
3410 "Unicode surrogate U+%04"UVXf" is illegal in UTF-8", uv);
3415 ((UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s))
3416 && (ckWARN_d(WARN_NONCHAR)))
3418 UV uv = utf8_to_uvchr(s, &char_len);
3419 Perl_warner(aTHX_ packWARN(WARN_NONCHAR),
3420 "Unicode non-character U+%04"UVXf" is illegal for open interchange", uv);
3431 =for apidoc pv_uni_display
3433 Build to the scalar dsv a displayable version of the string spv,
3434 length len, the displayable version being at most pvlim bytes long
3435 (if longer, the rest is truncated and "..." will be appended).
3437 The flags argument can have UNI_DISPLAY_ISPRINT set to display
3438 isPRINT()able characters as themselves, UNI_DISPLAY_BACKSLASH
3439 to display the \\[nrfta\\] as the backslashed versions (like '\n')
3440 (UNI_DISPLAY_BACKSLASH is preferred over UNI_DISPLAY_ISPRINT for \\).
3441 UNI_DISPLAY_QQ (and its alias UNI_DISPLAY_REGEX) have both
3442 UNI_DISPLAY_BACKSLASH and UNI_DISPLAY_ISPRINT turned on.
3444 The pointer to the PV of the dsv is returned.
3448 Perl_pv_uni_display(pTHX_ SV *dsv, const U8 *spv, STRLEN len, STRLEN pvlim, UV flags)
3453 PERL_ARGS_ASSERT_PV_UNI_DISPLAY;
3457 for (s = (const char *)spv, e = s + len; s < e; s += UTF8SKIP(s)) {
3459 /* This serves double duty as a flag and a character to print after
3460 a \ when flags & UNI_DISPLAY_BACKSLASH is true.
3464 if (pvlim && SvCUR(dsv) >= pvlim) {
3468 u = utf8_to_uvchr((U8*)s, 0);
3470 const unsigned char c = (unsigned char)u & 0xFF;
3471 if (flags & UNI_DISPLAY_BACKSLASH) {
3488 const char string = ok;
3489 sv_catpvs(dsv, "\\");
3490 sv_catpvn(dsv, &string, 1);
3493 /* isPRINT() is the locale-blind version. */
3494 if (!ok && (flags & UNI_DISPLAY_ISPRINT) && isPRINT(c)) {
3495 const char string = c;
3496 sv_catpvn(dsv, &string, 1);
3501 Perl_sv_catpvf(aTHX_ dsv, "\\x{%"UVxf"}", u);
3504 sv_catpvs(dsv, "...");
3510 =for apidoc sv_uni_display
3512 Build to the scalar dsv a displayable version of the scalar sv,
3513 the displayable version being at most pvlim bytes long
3514 (if longer, the rest is truncated and "..." will be appended).
3516 The flags argument is as in pv_uni_display().
3518 The pointer to the PV of the dsv is returned.
3523 Perl_sv_uni_display(pTHX_ SV *dsv, SV *ssv, STRLEN pvlim, UV flags)
3525 PERL_ARGS_ASSERT_SV_UNI_DISPLAY;
3527 return Perl_pv_uni_display(aTHX_ dsv, (const U8*)SvPVX_const(ssv),
3528 SvCUR(ssv), pvlim, flags);
3532 =for apidoc foldEQ_utf8
3534 Returns true if the leading portions of the strings s1 and s2 (either or both
3535 of which may be in UTF-8) are the same case-insensitively; false otherwise.
3536 How far into the strings to compare is determined by other input parameters.
3538 If u1 is true, the string s1 is assumed to be in UTF-8-encoded Unicode;
3539 otherwise it is assumed to be in native 8-bit encoding. Correspondingly for u2
3542 If the byte length l1 is non-zero, it says how far into s1 to check for fold
3543 equality. In other words, s1+l1 will be used as a goal to reach. The
3544 scan will not be considered to be a match unless the goal is reached, and
3545 scanning won't continue past that goal. Correspondingly for l2 with respect to
3548 If pe1 is non-NULL and the pointer it points to is not NULL, that pointer is
3549 considered an end pointer beyond which scanning of s1 will not continue under
3550 any circumstances. This means that if both l1 and pe1 are specified, and pe1
3551 is less than s1+l1, the match will never be successful because it can never
3552 get as far as its goal (and in fact is asserted against). Correspondingly for
3553 pe2 with respect to s2.
3555 At least one of s1 and s2 must have a goal (at least one of l1 and l2 must be
3556 non-zero), and if both do, both have to be
3557 reached for a successful match. Also, if the fold of a character is multiple
3558 characters, all of them must be matched (see tr21 reference below for
3561 Upon a successful match, if pe1 is non-NULL,
3562 it will be set to point to the beginning of the I<next> character of s1 beyond
3563 what was matched. Correspondingly for pe2 and s2.
3565 For case-insensitiveness, the "casefolding" of Unicode is used
3566 instead of upper/lowercasing both the characters, see
3567 http://www.unicode.org/unicode/reports/tr21/ (Case Mappings).
3571 /* A flags parameter has been added which may change, and hence isn't
3572 * externally documented. Currently it is:
3573 * 0 for as-documented above
3574 * FOLDEQ_UTF8_NOMIX_ASCII meaning that if a non-ASCII character folds to an
3575 ASCII one, to not match
3576 * FOLDEQ_UTF8_LOCALE meaning that locale rules are to be used for code
3577 * points below 256; unicode rules for above 255; and
3578 * folds that cross those boundaries are disallowed,
3579 * like the NOMIX_ASCII option
3580 * FOLDEQ_S1_ALREADY_FOLDED s1 has already been folded before calling this
3581 * routine. This allows that step to be skipped.
3582 * FOLDEQ_S2_ALREADY_FOLDED Similarly.
3585 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)
3588 register const U8 *p1 = (const U8*)s1; /* Point to current char */
3589 register const U8 *p2 = (const U8*)s2;
3590 register const U8 *g1 = NULL; /* goal for s1 */
3591 register const U8 *g2 = NULL;
3592 register const U8 *e1 = NULL; /* Don't scan s1 past this */
3593 register U8 *f1 = NULL; /* Point to current folded */
3594 register const U8 *e2 = NULL;
3595 register U8 *f2 = NULL;
3596 STRLEN n1 = 0, n2 = 0; /* Number of bytes in current char */
3597 U8 foldbuf1[UTF8_MAXBYTES_CASE+1];
3598 U8 foldbuf2[UTF8_MAXBYTES_CASE+1];
3600 PERL_ARGS_ASSERT_FOLDEQ_UTF8_FLAGS;
3602 /* The algorithm requires that input with the flags on the first line of
3603 * the assert not be pre-folded. */
3604 assert( ! ((flags & (FOLDEQ_UTF8_NOMIX_ASCII | FOLDEQ_UTF8_LOCALE))
3605 && (flags & (FOLDEQ_S1_ALREADY_FOLDED | FOLDEQ_S2_ALREADY_FOLDED))));
3612 g1 = (const U8*)s1 + l1;
3620 g2 = (const U8*)s2 + l2;
3623 /* Must have at least one goal */
3628 /* Will never match if goal is out-of-bounds */
3629 assert(! e1 || e1 >= g1);
3631 /* Here, there isn't an end pointer, or it is beyond the goal. We
3632 * only go as far as the goal */
3636 assert(e1); /* Must have an end for looking at s1 */
3639 /* Same for goal for s2 */
3641 assert(! e2 || e2 >= g2);
3648 /* If both operands are already folded, we could just do a memEQ on the
3649 * whole strings at once, but it would be better if the caller realized
3650 * this and didn't even call us */
3652 /* Look through both strings, a character at a time */
3653 while (p1 < e1 && p2 < e2) {
3655 /* If at the beginning of a new character in s1, get its fold to use
3656 * and the length of the fold. (exception: locale rules just get the
3657 * character to a single byte) */
3659 if (flags & FOLDEQ_S1_ALREADY_FOLDED) {
3663 /* If in locale matching, we use two sets of rules, depending on if
3664 * the code point is above or below 255. Here, we test for and
3665 * handle locale rules */
3668 if ((flags & FOLDEQ_UTF8_LOCALE)
3669 && (! u1 || UTF8_IS_INVARIANT(*p1)
3670 || UTF8_IS_DOWNGRADEABLE_START(*p1)))
3672 /* There is no mixing of code points above and below 255. */
3673 if (u2 && (! UTF8_IS_INVARIANT(*p2)
3674 && ! UTF8_IS_DOWNGRADEABLE_START(*p2)))
3679 /* We handle locale rules by converting, if necessary, the
3680 * code point to a single byte. */
3681 if (! u1 || UTF8_IS_INVARIANT(*p1)) {
3685 *foldbuf1 = TWO_BYTE_UTF8_TO_UNI(*p1, *(p1 + 1));
3689 else if (isASCII(*p1)) { /* Note, that here won't be
3690 both ASCII and using locale
3693 /* If trying to mix non- with ASCII, and not supposed to,
3695 if ((flags & FOLDEQ_UTF8_NOMIX_ASCII) && ! isASCII(*p2)) {
3699 *foldbuf1 = toLOWER(*p1); /* Folds in the ASCII range are
3703 to_utf8_fold(p1, foldbuf1, &n1);
3705 else { /* Not utf8, get utf8 fold */
3706 to_uni_fold(NATIVE_TO_UNI(*p1), foldbuf1, &n1);
3712 if (n2 == 0) { /* Same for s2 */
3713 if (flags & FOLDEQ_S2_ALREADY_FOLDED) {
3718 if ((flags & FOLDEQ_UTF8_LOCALE)
3719 && (! u2 || UTF8_IS_INVARIANT(*p2) || UTF8_IS_DOWNGRADEABLE_START(*p2)))
3721 /* Here, the next char in s2 is < 256. We've already
3722 * worked on s1, and if it isn't also < 256, can't match */
3723 if (u1 && (! UTF8_IS_INVARIANT(*p1)
3724 && ! UTF8_IS_DOWNGRADEABLE_START(*p1)))
3728 if (! u2 || UTF8_IS_INVARIANT(*p2)) {
3732 *foldbuf2 = TWO_BYTE_UTF8_TO_UNI(*p2, *(p2 + 1));
3735 /* Use another function to handle locale rules. We've made
3736 * sure that both characters to compare are single bytes */
3737 if (! foldEQ_locale((char *) f1, (char *) foldbuf2, 1)) {
3742 else if (isASCII(*p2)) {
3743 if (flags && ! isASCII(*p1)) {
3747 *foldbuf2 = toLOWER(*p2);
3750 to_utf8_fold(p2, foldbuf2, &n2);
3753 to_uni_fold(NATIVE_TO_UNI(*p2), foldbuf2, &n2);
3759 /* Here f1 and f2 point to the beginning of the strings to compare.
3760 * These strings are the folds of the next character from each input
3761 * string, stored in utf8. */
3763 /* While there is more to look for in both folds, see if they
3764 * continue to match */
3766 U8 fold_length = UTF8SKIP(f1);
3767 if (fold_length != UTF8SKIP(f2)
3768 || (fold_length == 1 && *f1 != *f2) /* Short circuit memNE
3769 function call for single
3771 || memNE((char*)f1, (char*)f2, fold_length))
3773 return 0; /* mismatch */
3776 /* Here, they matched, advance past them */
3783 /* When reach the end of any fold, advance the input past it */
3785 p1 += u1 ? UTF8SKIP(p1) : 1;
3788 p2 += u2 ? UTF8SKIP(p2) : 1;
3790 } /* End of loop through both strings */
3792 /* A match is defined by each scan that specified an explicit length
3793 * reaching its final goal, and the other not having matched a partial
3794 * character (which can happen when the fold of a character is more than one
3796 if (! ((g1 == 0 || p1 == g1) && (g2 == 0 || p2 == g2)) || n1 || n2) {
3800 /* Successful match. Set output pointers */
3812 * c-indentation-style: bsd
3814 * indent-tabs-mode: t
3817 * ex: set ts=8 sts=4 sw=4 noet: