This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
https://perl5.git.perl.org / perl5.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
5.26.2 today
[perl5.git] / utf8.h
CommitLineData
a0ed51b3
LW		 1/* utf8.h
2 *
f3cb6f94
KW		 3 * This file contains definitions for use with the UTF-8 encoding. It
4 * actually also works with the variant UTF-8 encoding called UTF-EBCDIC, and
5 * hides almost all of the differences between these from the caller. In other
6 * words, someone should #include this file, and if the code is being compiled
7 * on an EBCDIC platform, things should mostly just work.
8 *
2eee27d7
SS		 9 * Copyright (C) 2000, 2001, 2002, 2005, 2006, 2007, 2009,
10 * 2010, 2011 by Larry Wall and others
a0ed51b3
LW		 11 *
12 * You may distribute under the terms of either the GNU General Public
13 * License or the Artistic License, as specified in the README file.
14 *
15 */
16
6a5bc5ac
KW		 17#ifndef PERL_UTF8_H_ /* Guard against recursive inclusion */
18#define PERL_UTF8_H_ 1
57f0e7e2 19
39e02b42 20/* Use UTF-8 as the default script encoding?
1e54db1a 21 * Turning this on will break scripts having non-UTF-8 binary
39e02b42
JH		 22 * data (such as Latin-1) in string literals. */
23#ifdef USE_UTF8_SCRIPTS
24# define USE_UTF8_IN_NAMES (!IN_BYTES)
25#else
26# define USE_UTF8_IN_NAMES (PL_hints & HINT_UTF8)
27#endif
28
3cd96634
KW		 29#include "regcharclass.h"
30#include "unicode_constants.h"
31
051a06d4 32/* For to_utf8_fold_flags, q.v. */
e4f4ef45
KW		 33#define FOLD_FLAGS_LOCALE 0x1
34#define FOLD_FLAGS_FULL 0x2
35#define FOLD_FLAGS_NOMIX_ASCII 0x4
051a06d4 36
83199d38
KW		 37/* For _core_swash_init(), internal core use only */
38#define _CORE_SWASH_INIT_USER_DEFINED_PROPERTY 0x1
5d3d13d1 39#define _CORE_SWASH_INIT_RETURN_IF_UNDEF 0x2
87367d5f 40#define _CORE_SWASH_INIT_ACCEPT_INVLIST 0x4
83199d38 41
7bbfa158
KW		 42/*
43=head1 Unicode Support
8cca77bc
KW		 44L<perlguts/Unicode Support> has an introduction to this API.
45
46See also L</Character classification>,
47and L</Character case changing>.
48Various functions outside this section also work specially with Unicode.
49Search for the string "utf8" in this document.
7bbfa158
KW		 50
51=for apidoc is_ascii_string
52
8871a094 53This is a misleadingly-named synonym for L</is_utf8_invariant_string>.
7bbfa158
KW		 54On ASCII-ish platforms, the name isn't misleading: the ASCII-range characters
55are exactly the UTF-8 invariants. But EBCDIC machines have more invariants
8871a094
KW		 56than just the ASCII characters, so C<is_utf8_invariant_string> is preferred.
57
58=for apidoc is_invariant_string
59
60This is a somewhat misleadingly-named synonym for L</is_utf8_invariant_string>.
61C<is_utf8_invariant_string> is preferred, as it indicates under what conditions
62the string is invariant.
7bbfa158
KW		 63
64=cut
65*/
8871a094
KW		 66#define is_ascii_string(s, len) is_utf8_invariant_string(s, len)
67#define is_invariant_string(s, len) is_utf8_invariant_string(s, len)
7bbfa158 68
33f38593
KW		 69#define uvoffuni_to_utf8_flags(d,uv,flags) \
70 uvoffuni_to_utf8_flags_msgs(d, uv, flags, 0)
de69f3af
KW		 71#define uvchr_to_utf8(a,b) uvchr_to_utf8_flags(a,b,0)
72#define uvchr_to_utf8_flags(d,uv,flags) \
33f38593
KW		 73 uvchr_to_utf8_flags_msgs(d,uv,flags, 0)
74#define uvchr_to_utf8_flags_msgs(d,uv,flags,msgs) \
75 uvoffuni_to_utf8_flags_msgs(d,NATIVE_TO_UNI(uv),flags, msgs)
de69f3af 76#define utf8_to_uvchr_buf(s, e, lenp) \
842991ae 77 utf8n_to_uvchr(s, (U8*)(e) - (U8*)(s), lenp, \
de69f3af 78 ckWARN_d(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY)
f9380377
KW		 79#define utf8n_to_uvchr(s, len, lenp, flags) \
80 utf8n_to_uvchr_error(s, len, lenp, flags, 0)
37657a5b
KW		 81#define utf8n_to_uvchr_error(s, len, lenp, flags, errors) \
82 utf8n_to_uvchr_msgs(s, len, lenp, flags, errors, 0)
de69f3af 83
a0270393 84#define to_uni_fold(c, p, lenp) _to_uni_fold_flags(c, p, lenp, FOLD_FLAGS_FULL)
a239b1e2
KW		 85
86#define to_utf8_fold(s, r, lenr) \
607313a1 87 _to_utf8_fold_flags (s, NULL, r, lenr, FOLD_FLAGS_FULL, __FILE__, __LINE__)
a239b1e2 88#define to_utf8_lower(s, r, lenr) \
607313a1 89 _to_utf8_lower_flags(s, NULL, r ,lenr, 0, __FILE__, __LINE__)
a239b1e2 90#define to_utf8_upper(s, r, lenr) \
607313a1 91 _to_utf8_upper_flags(s, NULL, r, lenr, 0, __FILE__, __LINE__)
a239b1e2 92#define to_utf8_title(s, r, lenr) \
607313a1 93 _to_utf8_title_flags(s, NULL, r, lenr ,0, __FILE__, __LINE__)
36bb2ab6 94
eda9cac1
KW		 95#define foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \
96 foldEQ_utf8_flags(s1, pe1, l1, u1, s2, pe2, l2, u2, 0)
baa60164 97#define FOLDEQ_UTF8_NOMIX_ASCII (1 << 0)
cea315b6 98#define FOLDEQ_LOCALE (1 << 1)
18f762c3
KW		 99#define FOLDEQ_S1_ALREADY_FOLDED (1 << 2)
100#define FOLDEQ_S2_ALREADY_FOLDED (1 << 3)
d635b710
KW		 101#define FOLDEQ_S1_FOLDS_SANE (1 << 4)
102#define FOLDEQ_S2_FOLDS_SANE (1 << 5)
a33c29bc 103
e6226b18
KW		 104#define ibcmp_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \
105 cBOOL(! foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2))
106
1d72bdf6
NIS		 107#ifdef EBCDIC
108/* The equivalent of these macros but implementing UTF-EBCDIC
109 are in the following header file:
110 */
111
112#include "utfebcdic.h"
fd7cb289 113
d06134e5 114#else /* ! EBCDIC */
73c4f7a1
GS		 115START_EXTERN_C
116
111e8ed9
KW		 117/* How wide can a single UTF-8 encoded character become in bytes. */
118/* NOTE: Strictly speaking Perl's UTF-8 should not be called UTF-8 since UTF-8
119 * is an encoding of Unicode, and Unicode's upper limit, 0x10FFFF, can be
120 * expressed with 4 bytes. However, Perl thinks of UTF-8 as a way to encode
121 * non-negative integers in a binary format, even those above Unicode */
122#define UTF8_MAXBYTES 13
123
a0ed51b3 124#ifdef DOINIT
6f06b55f 125EXTCONST unsigned char PL_utf8skip[] = {
b2635aa8
KW		 126/* 0x00 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
127/* 0x10 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
128/* 0x20 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
129/* 0x30 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
130/* 0x40 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
131/* 0x50 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
132/* 0x60 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
133/* 0x70 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
134/* 0x80 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
135/* 0x90 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
136/* 0xA0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
137/* 0xB0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
138/* 0xC0 */ 2,2, /* overlong */
1ff3baa2 139/* 0xC2 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0080 to U+03FF */
b2635aa8
KW		 140/* 0xD0 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0400 to U+07FF */
141/* 0xE0 */ 3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3, /* U+0800 to U+FFFF */
142/* 0xF0 */ 4,4,4,4,4,4,4,4,5,5,5,5,6,6, /* above BMP to 2**31 - 1 */
6937f885
KW		 143 /* Perl extended (never was official UTF-8). Up to 36 bit */
144/* 0xFE */ 7,
145 /* More extended, Up to 72 bits (64-bit + reserved) */
111e8ed9 146/* 0xFF */ UTF8_MAXBYTES
a0ed51b3
LW		 147};
148#else
6f06b55f 149EXTCONST unsigned char PL_utf8skip[];
a0ed51b3
LW		 150#endif
151
73c4f7a1 152END_EXTERN_C
7e2040f0 153
1a3756de 154#if defined(_MSC_VER) && _MSC_VER < 1400
6f6d1bab
TC		 155/* older MSVC versions have a smallish macro buffer */
156#define PERL_SMALL_MACRO_BUFFER
157#endif
158
59a449d5
KW		 159/* Native character to/from iso-8859-1. Are the identity functions on ASCII
160 * platforms */
6f6d1bab 161#ifdef PERL_SMALL_MACRO_BUFFER
1a3756de
TC		 162#define NATIVE_TO_LATIN1(ch) ((U8)(ch))
163#define LATIN1_TO_NATIVE(ch) ((U8)(ch))
6f6d1bab 164#else
4c8cd605
KW		 165#define NATIVE_TO_LATIN1(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch)))
166#define LATIN1_TO_NATIVE(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch)))
6f6d1bab 167#endif
59a449d5
KW		 168
169/* I8 is an intermediate version of UTF-8 used only in UTF-EBCDIC. We thus
170 * consider it to be identical to UTF-8 on ASCII platforms. Strictly speaking
171 * UTF-8 and UTF-EBCDIC are two different things, but we often conflate them
172 * because they are 8-bit encodings that serve the same purpose in Perl, and
173 * rarely do we need to distinguish them. The term "NATIVE_UTF8" applies to
174 * whichever one is applicable on the current platform */
6f6d1bab 175#ifdef PERL_SMALL_MACRO_BUFFER
1d4ea287
DD		 176#define NATIVE_UTF8_TO_I8(ch) ((U8) (ch))
177#define I8_TO_NATIVE_UTF8(ch) ((U8) (ch))
6f6d1bab 178#else
4c8cd605
KW		 179#define NATIVE_UTF8_TO_I8(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch)))
180#define I8_TO_NATIVE_UTF8(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch)))
6f6d1bab 181#endif
59a449d5 182
1d72bdf6 183/* Transforms in wide UV chars */
4c8cd605
KW		 184#define UNI_TO_NATIVE(ch) ((UV) (ch))
185#define NATIVE_TO_UNI(ch) ((UV) (ch))
d7578b48 186
877d9f0d 187/*
9041c2e3 188
a14e0a36
KW		 189 The following table is from Unicode 3.2, plus the Perl extensions for above
190 U+10FFFF
877d9f0d 191
a14e0a36 192 Code Points 1st Byte 2nd Byte 3rd 4th 5th 6th 7th 8th-13th
877d9f0d 193
375122d7 194 U+0000..U+007F 00..7F
e1b711da 195 U+0080..U+07FF * C2..DF 80..BF
a14e0a36
KW		 196 U+0800..U+0FFF E0 * A0..BF 80..BF
197 U+1000..U+CFFF E1..EC 80..BF 80..BF
198 U+D000..U+D7FF ED 80..9F 80..BF
199 U+D800..U+DFFF ED A0..BF 80..BF (surrogates)
200 U+E000..U+FFFF EE..EF 80..BF 80..BF
201 U+10000..U+3FFFF F0 * 90..BF 80..BF 80..BF
202 U+40000..U+FFFFF F1..F3 80..BF 80..BF 80..BF
203 U+100000..U+10FFFF F4 80..8F 80..BF 80..BF
204 Below are above-Unicode code points
205 U+110000..U+13FFFF F4 90..BF 80..BF 80..BF
206 U+110000..U+1FFFFF F5..F7 80..BF 80..BF 80..BF
207 U+200000..U+FFFFFF F8 * 88..BF 80..BF 80..BF 80..BF
208U+1000000..U+3FFFFFF F9..FB 80..BF 80..BF 80..BF 80..BF
209U+4000000..U+3FFFFFFF FC * 84..BF 80..BF 80..BF 80..BF 80..BF
210U+40000000..U+7FFFFFFF FD 80..BF 80..BF 80..BF 80..BF 80..BF
211U+80000000..U+FFFFFFFFF FE * 82..BF 80..BF 80..BF 80..BF 80..BF 80..BF
212U+1000000000.. FF 80..BF 80..BF 80..BF 80..BF 80..BF * 81..BF 80..BF
877d9f0d 213
e1b711da 214Note the gaps before several of the byte entries above marked by '*'. These are
37e2e78e
KW		 215caused by legal UTF-8 avoiding non-shortest encodings: it is technically
216possible to UTF-8-encode a single code point in different ways, but that is
217explicitly forbidden, and the shortest possible encoding should always be used
15824458 218(and that is what Perl does). The non-shortest ones are called 'overlongs'.
8c007b5a 219
877d9f0d
JH		 220 */
221
8c007b5a
JH		 222/*
223 Another way to look at it, as bits:
224
b2635aa8 225 Code Points 1st Byte 2nd Byte 3rd Byte 4th Byte
8c007b5a 226
b2635aa8
KW		 227 0aaa aaaa 0aaa aaaa
228 0000 0bbb bbaa aaaa 110b bbbb 10aa aaaa
229 cccc bbbb bbaa aaaa 1110 cccc 10bb bbbb 10aa aaaa
230 00 000d ddcc cccc bbbb bbaa aaaa 1111 0ddd 10cc cccc 10bb bbbb 10aa aaaa
8c007b5a
JH		 231
232As you can see, the continuation bytes all begin with C<10>, and the
e1b711da 233leading bits of the start byte tell how many bytes there are in the
8c007b5a
JH		 234encoded character.
235
df863e43
KW		 236Perl's extended UTF-8 means we can have start bytes up through FF, though any
237beginning with FF yields a code point that is too large for 32-bit ASCII
238platforms. FF signals to use 13 bytes for the encoded character. This breaks
239the paradigm that the number of leading bits gives how many total bytes there
240are in the character.
65ab9279 241
8c007b5a
JH		 242*/
243
6c88483e 244/* Is the representation of the Unicode code point 'cp' the same regardless of
15824458 245 * being encoded in UTF-8 or not? */
2d1545e5 246#define OFFUNI_IS_INVARIANT(cp) isASCII(cp)
15824458 247
c2b32798
KW		 248/*
249=for apidoc Am|bool|UVCHR_IS_INVARIANT|UV cp
250
251Evaluates to 1 if the representation of code point C<cp> is the same whether or
252not it is encoded in UTF-8; otherwise evaluates to 0. UTF-8 invariant
253characters can be copied as-is when converting to/from UTF-8, saving time.
254C<cp> is Unicode if above 255; otherwise is platform-native.
255
256=cut
257 */
258
cf1be84e 259#define UVCHR_IS_INVARIANT(cp) OFFUNI_IS_INVARIANT(cp)
38953e5a 260
c9264833
KW		 261/* This defines the bits that are to be in the continuation bytes of a multi-byte
262 * UTF-8 encoded character that mark it is a continuation byte. */
263#define UTF_CONTINUATION_MARK 0x80
264
a95ec4fb 265/* Misleadingly named: is the UTF8-encoded byte 'c' part of a variant sequence
5c06326b
KW		 266 * in UTF-8? This is the inverse of UTF8_IS_INVARIANT. The |0 makes sure this
267 * isn't mistakenly called with a ptr argument */
a6951642
KW		 268#define UTF8_IS_CONTINUED(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
269 ((U8)((c) | 0)) & UTF_CONTINUATION_MARK)
15824458
KW		 270
271/* Is the byte 'c' the first byte of a multi-byte UTF8-8 encoded sequence?
272 * This doesn't catch invariants (they are single-byte). It also excludes the
5c06326b
KW		 273 * illegal overlong sequences that begin with C0 and C1. The |0 makes sure
274 * this isn't mistakenly called with a ptr argument */
a6951642
KW		 275#define UTF8_IS_START(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
276 ((U8)((c) | 0)) >= 0xc2)
15824458 277
858cd8ab
KW		 278/* For use in UTF8_IS_CONTINUATION() below */
279#define UTF_IS_CONTINUATION_MASK 0xC0
280
15824458 281/* Is the byte 'c' part of a multi-byte UTF8-8 encoded sequence, and not the
5c06326b
KW		 282 * first byte thereof? The |0 makes sure this isn't mistakenly called with a
283 * ptr argument */
a6951642
KW		 284#define UTF8_IS_CONTINUATION(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
285 (((U8)((c) | 0)) & UTF_IS_CONTINUATION_MASK) == UTF_CONTINUATION_MARK)
0ae1fa71 286
15824458
KW		 287/* Is the UTF8-encoded byte 'c' the first byte of a two byte sequence? Use
288 * UTF8_IS_NEXT_CHAR_DOWNGRADEABLE() instead if the input isn't known to
289 * be well-formed. Masking with 0xfe allows the low bit to be 0 or 1; thus
5c06326b
KW		 290 * this matches 0xc[23]. The |0 makes sure this isn't mistakenly called with a
291 * ptr argument */
a6951642
KW		 292#define UTF8_IS_DOWNGRADEABLE_START(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
293 (((U8)((c) | 0)) & 0xfe) == 0xc2)
4ab10950 294
15824458 295/* Is the UTF8-encoded byte 'c' the first byte of a sequence of bytes that
5c06326b
KW		 296 * represent a code point > 255? The |0 makes sure this isn't mistakenly
297 * called with a ptr argument */
a6951642
KW		 298#define UTF8_IS_ABOVE_LATIN1(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
299 ((U8)((c) | 0)) >= 0xc4)
8850bf83 300
15824458
KW		 301/* This is the number of low-order bits a continuation byte in a UTF-8 encoded
302 * sequence contributes to the specification of the code point. In the bit
303 * maps above, you see that the first 2 bits are a constant '10', leaving 6 of
304 * real information */
1d72bdf6 305#define UTF_ACCUMULATION_SHIFT 6
b2635aa8 306
fed423a5
KW		 307/* ^? is defined to be DEL on ASCII systems. See the definition of toCTRL()
308 * for more */
309#define QUESTION_MARK_CTRL DEL_NATIVE
310
311/* Surrogates, non-character code points and above-Unicode code points are
312 * problematic in some contexts. This allows code that needs to check for
313 * those to to quickly exclude the vast majority of code points it will
314 * encounter */
a6951642
KW		 315#define isUTF8_POSSIBLY_PROBLEMATIC(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
316 (U8) c >= 0xED)
fed423a5 317
784d4f31
KW		 318/* A helper macro for isUTF8_CHAR, so use that one instead of this. This was
319 * generated by regen/regcharclass.pl, and then moved here. Then it was
320 * hand-edited to add some LIKELY() calls, presuming that malformations are
321 * unlikely. The lines that generated it were then commented out. This was
322 * done because it takes on the order of 10 minutes to generate, and is never
323 * going to change, unless the generated code is improved, and figuring out
324 * the LIKELYs there would be hard.
325 *
326 UTF8_CHAR: Matches legal UTF-8 variant code points up through 0x1FFFFFF
327
328 0x80 - 0x1FFFFF
329*/
330/*** GENERATED CODE ***/
331#define is_UTF8_CHAR_utf8_no_length_checks(s) \
9f2eed98
P		 332( ( 0xC2 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xDF ) ? \
333 ( LIKELY( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) ? 2 : 0 ) \
334: ( 0xE0 == ((const U8*)s)[0] ) ? \
335 ( LIKELY( ( ( ((const U8*)s)[1] & 0xE0 ) == 0xA0 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
336: ( 0xE1 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xEF ) ? \
337 ( LIKELY( ( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
338: ( 0xF0 == ((const U8*)s)[0] ) ? \
339 ( LIKELY( ( ( 0x90 <= ((const U8*)s)[1] && ((const U8*)s)[1] <= 0xBF ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) && ( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\
340: ( ( ( ( 0xF1 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xF7 ) && LIKELY( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) ) && LIKELY( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) && LIKELY( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )
784d4f31
KW		 341
342/* The above macro handles UTF-8 that has this start byte as the maximum */
343#define _IS_UTF8_CHAR_HIGHEST_START_BYTE 0xF7
344
e23e8bc1
KW		 345/* A helper macro for isSTRICT_UTF8_CHAR, so use that one instead of this.
346 * Like is_UTF8_CHAR_utf8_no_length_checks(), this was moved here and LIKELYs
347 * added manually.
348 *
349 STRICT_UTF8_CHAR: Matches legal Unicode UTF-8 variant code points, no
350 surrrogates nor non-character code points
351*/
352/*** GENERATED CODE ***/
353#define is_STRICT_UTF8_CHAR_utf8_no_length_checks(s) \
9f2eed98
P		 354( ( 0xC2 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xDF ) ? \
355 ( LIKELY( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) ? 2 : 0 ) \
356: ( 0xE0 == ((const U8*)s)[0] ) ? \
357 ( LIKELY( ( ( ((const U8*)s)[1] & 0xE0 ) == 0xA0 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
358: ( ( 0xE1 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xEC ) || 0xEE == ((const U8*)s)[0] ) ?\
359 ( ( ( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
360: ( 0xED == ((const U8*)s)[0] ) ? \
361 ( LIKELY( ( ( ((const U8*)s)[1] & 0xE0 ) == 0x80 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
362: ( 0xEF == ((const U8*)s)[0] ) ? \
363 ( ( ( 0x80 <= ((const U8*)s)[1] && ((const U8*)s)[1] <= 0xB6 ) || ( 0xB8 <= ((const U8*)s)[1] && ((const U8*)s)[1] <= 0xBE ) ) ?\
364 ( LIKELY( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ? 3 : 0 ) \
365 : ( 0xB7 == ((const U8*)s)[1] ) ? \
366 ( LIKELY( ( ((const U8*)s)[2] & 0xF0 ) == 0x80 || ( ((const U8*)s)[2] & 0xF0 ) == 0xB0 ) ? 3 : 0 )\
367 : ( ( 0xBF == ((const U8*)s)[1] ) && ( 0x80 <= ((const U8*)s)[2] && ((const U8*)s)[2] <= 0xBD ) ) ? 3 : 0 )\
368: ( 0xF0 == ((const U8*)s)[0] ) ? \
369 ( ( ( 0x90 <= ((const U8*)s)[1] && ((const U8*)s)[1] <= 0x9E ) || ( 0xA0 <= ((const U8*)s)[1] && ((const U8*)s)[1] <= 0xAE ) || ( 0xB0 <= ((const U8*)s)[1] && ((const U8*)s)[1] <= 0xBE ) ) ?\
370 ( LIKELY( ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\
371 : ( ((const U8*)s)[1] == 0x9F || ( ( ((const U8*)s)[1] & 0xEF ) == 0xAF ) ) ? \
372 ( ( 0x80 <= ((const U8*)s)[2] && ((const U8*)s)[2] <= 0xBE ) ? \
373 ( LIKELY( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ? 4 : 0 ) \
374 : LIKELY( ( 0xBF == ((const U8*)s)[2] ) && ( 0x80 <= ((const U8*)s)[3] && ((const U8*)s)[3] <= 0xBD ) ) ? 4 : 0 )\
e23e8bc1 375 : 0 ) \
9f2eed98
P		 376: ( 0xF1 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xF3 ) ? \
377 ( ( ( ( ((const U8*)s)[1] & 0xC8 ) == 0x80 ) || ( ( ((const U8*)s)[1] & 0xCC ) == 0x88 ) || ( ( ((const U8*)s)[1] & 0xCE ) == 0x8C ) || ( ( ((const U8*)s)[1] & 0xCF ) == 0x8E ) ) ?\
378 ( LIKELY( ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\
379 : ( ( ((const U8*)s)[1] & 0xCF ) == 0x8F ) ? \
380 ( ( 0x80 <= ((const U8*)s)[2] && ((const U8*)s)[2] <= 0xBE ) ? \
381 ( LIKELY( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ? 4 : 0 ) \
382 : LIKELY( ( 0xBF == ((const U8*)s)[2] ) && ( 0x80 <= ((const U8*)s)[3] && ((const U8*)s)[3] <= 0xBD ) ) ? 4 : 0 )\
e23e8bc1 383 : 0 ) \
9f2eed98
P		 384: ( 0xF4 == ((const U8*)s)[0] ) ? \
385 ( ( 0x80 <= ((const U8*)s)[1] && ((const U8*)s)[1] <= 0x8E ) ? \
386 ( LIKELY( ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\
387 : ( 0x8F == ((const U8*)s)[1] ) ? \
388 ( ( 0x80 <= ((const U8*)s)[2] && ((const U8*)s)[2] <= 0xBE ) ? \
389 ( LIKELY( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ? 4 : 0 ) \
390 : LIKELY( ( 0xBF == ((const U8*)s)[2] ) && ( 0x80 <= ((const U8*)s)[3] && ((const U8*)s)[3] <= 0xBD ) ) ? 4 : 0 )\
e23e8bc1
KW		 391 : 0 ) \
392: 0 )
393
a82be82b
KW		 394/* Similarly,
395 C9_STRICT_UTF8_CHAR: Matches legal Unicode UTF-8 variant code
396 points, no surrogates
397 0x0080 - 0xD7FF
398 0xE000 - 0x10FFFF
399*/
400/*** GENERATED CODE ***/
401#define is_C9_STRICT_UTF8_CHAR_utf8_no_length_checks(s) \
9f2eed98
P		 402( ( 0xC2 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xDF ) ? \
403 ( LIKELY( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) ? 2 : 0 ) \
404: ( 0xE0 == ((const U8*)s)[0] ) ? \
405 ( LIKELY( ( ( ((const U8*)s)[1] & 0xE0 ) == 0xA0 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
406: ( ( 0xE1 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xEC ) || ( ((const U8*)s)[0] & 0xFE ) == 0xEE ) ?\
407 ( LIKELY( ( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
408: ( 0xED == ((const U8*)s)[0] ) ? \
409 ( LIKELY( ( ( ((const U8*)s)[1] & 0xE0 ) == 0x80 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
410: ( 0xF0 == ((const U8*)s)[0] ) ? \
411 ( LIKELY( ( ( 0x90 <= ((const U8*)s)[1] && ((const U8*)s)[1] <= 0xBF ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) && ( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\
412: ( 0xF1 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xF3 ) ? \
413 ( LIKELY( ( ( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) && ( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\
414: LIKELY( ( ( ( 0xF4 == ((const U8*)s)[0] ) && ( ( ((const U8*)s)[1] & 0xF0 ) == 0x80 ) ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) && ( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )
a82be82b 415
57ff5f59
KW		 416#define UNICODE_IS_PERL_EXTENDED(uv) UNLIKELY((UV) (uv) > 0x7FFFFFFF)
417
fed423a5
KW		 418#endif /* EBCDIC vs ASCII */
419
420/* 2**UTF_ACCUMULATION_SHIFT - 1 */
421#define UTF_CONTINUATION_MASK ((U8) ((1U << UTF_ACCUMULATION_SHIFT) - 1))
422
7028aeba
KW		 423/* Internal macro to be used only in this file to aid in constructing other
424 * publicly accessible macros.
425 * The number of bytes required to express this uv in UTF-8, for just those
426 * uv's requiring 2 through 6 bytes, as these are common to all platforms and
427 * word sizes. The number of bytes needed is given by the number of leading 1
428 * bits in the start byte. There are 32 start bytes that have 2 initial 1 bits
429 * (C0-DF); there are 16 that have 3 initial 1 bits (E0-EF); 8 that have 4
430 * initial 1 bits (F0-F8); 4 that have 5 initial 1 bits (F9-FB), and 2 that
431 * have 6 initial 1 bits (FC-FD). The largest number a string of n bytes can
432 * represent is (the number of possible start bytes for 'n')
433 * * (the number of possiblities for each start byte
434 * The latter in turn is
435 * 2 ** ( (how many continuation bytes there are)
436 * * (the number of bits of information each
437 * continuation byte holds))
438 *
439 * If we were on a platform where we could use a fast find first set bit
440 * instruction (or count leading zeros instruction) this could be replaced by
441 * using that to find the log2 of the uv, and divide that by the number of bits
442 * of information in each continuation byte, adjusting for large cases and how
443 * much information is in a start byte for that length */
72164d3a 444#define __COMMON_UNI_SKIP(uv) \
7028aeba
KW		 445 (UV) (uv) < (32 * (1U << ( UTF_ACCUMULATION_SHIFT))) ? 2 : \
446 (UV) (uv) < (16 * (1U << (2 * UTF_ACCUMULATION_SHIFT))) ? 3 : \
447 (UV) (uv) < ( 8 * (1U << (3 * UTF_ACCUMULATION_SHIFT))) ? 4 : \
448 (UV) (uv) < ( 4 * (1U << (4 * UTF_ACCUMULATION_SHIFT))) ? 5 : \
449 (UV) (uv) < ( 2 * (1U << (5 * UTF_ACCUMULATION_SHIFT))) ? 6 :
72164d3a
KW		 450
451/* Internal macro to be used only in this file.
452 * This adds to __COMMON_UNI_SKIP the details at this platform's upper range.
fed423a5 453 * For any-sized EBCDIC platforms, or 64-bit ASCII ones, we need one more test
72164d3a
KW		 454 * to see if just 7 bytes is needed, or if the maximum is needed. For 32-bit
455 * ASCII platforms, everything is representable by 7 bytes */
fed423a5 456#if defined(UV_IS_QUAD) || defined(EBCDIC)
72164d3a 457# define __BASE_UNI_SKIP(uv) (__COMMON_UNI_SKIP(uv) \
7028aeba 458 (UV) (uv) < ((UV) 1U << (6 * UTF_ACCUMULATION_SHIFT)) ? 7 : UTF8_MAXBYTES)
1d68d6cd 459#else
72164d3a 460# define __BASE_UNI_SKIP(uv) (__COMMON_UNI_SKIP(uv) 7)
1d68d6cd
SC		 461#endif
462
fed423a5
KW		 463/* The next two macros use the base macro defined above, and add in the tests
464 * at the low-end of the range, for just 1 byte, yielding complete macros,
465 * publicly accessible. */
466
467/* Input is a true Unicode (not-native) code point */
468#define OFFUNISKIP(uv) (OFFUNI_IS_INVARIANT(uv) ? 1 : __BASE_UNI_SKIP(uv))
2084b489 469
5352a763
KW		 470/*
471
472=for apidoc Am|STRLEN|UVCHR_SKIP|UV cp
473returns the number of bytes required to represent the code point C<cp> when
474encoded as UTF-8. C<cp> is a native (ASCII or EBCDIC) code point if less than
475255; a Unicode code point otherwise.
476
477=cut
478 */
fdb6583d 479#define UVCHR_SKIP(uv) ( UVCHR_IS_INVARIANT(uv) ? 1 : __BASE_UNI_SKIP(uv))
5352a763 480
b651802e
KW		 481/* The largest code point representable by two UTF-8 bytes on this platform.
482 * As explained in the comments for __COMMON_UNI_SKIP, 32 start bytes with
fed423a5 483 * UTF_ACCUMULATION_SHIFT bits of information each */
aa206fb7
KW		 484#define MAX_UTF8_TWO_BYTE (32 * (1U << UTF_ACCUMULATION_SHIFT) - 1)
485
b651802e
KW		 486/* The largest code point representable by two UTF-8 bytes on any platform that
487 * Perl runs on. This value is constrained by EBCDIC which has 5 bits per
488 * continuation byte */
aa206fb7
KW		 489#define MAX_PORTABLE_UTF8_TWO_BYTE (32 * (1U << 5) - 1)
490
c03c0950
KW		 491/* The maximum number of UTF-8 bytes a single Unicode character can
492 * uppercase/lowercase/fold into. Unicode guarantees that the maximum
493 * expansion is UTF8_MAX_FOLD_CHAR_EXPAND characters, but any above-Unicode
494 * code point will fold to itself, so we only have to look at the expansion of
495 * the maximum Unicode code point. But this number may be less than the space
496 * occupied by a very large code point under Perl's extended UTF-8. We have to
497 * make it large enough to fit any single character. (It turns out that ASCII
498 * and EBCDIC differ in which is larger) */
499#define UTF8_MAXBYTES_CASE \
500 (UTF8_MAXBYTES >= (UTF8_MAX_FOLD_CHAR_EXPAND * OFFUNISKIP(0x10FFFF)) \
501 ? UTF8_MAXBYTES \
502 : (UTF8_MAX_FOLD_CHAR_EXPAND * OFFUNISKIP(0x10FFFF)))
503
d06134e5
KW		 504/* Rest of these are attributes of Unicode and perl's internals rather than the
505 * encoding, or happen to be the same in both ASCII and EBCDIC (at least at
506 * this level; the macros that some of these call may have different
507 * definitions in the two encodings */
508
59a449d5
KW		 509/* In domain restricted to ASCII, these may make more sense to the reader than
510 * the ones with Latin1 in the name */
511#define NATIVE_TO_ASCII(ch) NATIVE_TO_LATIN1(ch)
512#define ASCII_TO_NATIVE(ch) LATIN1_TO_NATIVE(ch)
513
514/* More or less misleadingly-named defines, retained for back compat */
515#define NATIVE_TO_UTF(ch) NATIVE_UTF8_TO_I8(ch)
516#define NATIVE_TO_I8(ch) NATIVE_UTF8_TO_I8(ch)
517#define UTF_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch)
518#define I8_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch)
519#define NATIVE8_TO_UNI(ch) NATIVE_TO_LATIN1(ch)
d06134e5 520
c0236afe 521/* This defines the 1-bits that are to be in the first byte of a multi-byte
97d0ceda
KW		 522 * UTF-8 encoded character that mark it as a start byte and give the number of
523 * bytes that comprise the character. 'len' is the number of bytes in the
524 * multi-byte sequence. */
c0236afe
KW		 525#define UTF_START_MARK(len) (((len) > 7) ? 0xFF : (0xFF & (0xFE << (7-(len)))))
526
527/* Masks out the initial one bits in a start byte, leaving the real data ones.
528 * Doesn't work on an invariant byte. 'len' is the number of bytes in the
529 * multi-byte sequence that comprises the character. */
530#define UTF_START_MASK(len) (((len) >= 7) ? 0x00 : (0x1F >> ((len)-2)))
531
537124e4
KW		 532/* Adds a UTF8 continuation byte 'new' of information to a running total code
533 * point 'old' of all the continuation bytes so far. This is designed to be
155d2738
KW		 534 * used in a loop to convert from UTF-8 to the code point represented. Note
535 * that this is asymmetric on EBCDIC platforms, in that the 'new' parameter is
536 * the UTF-EBCDIC byte, whereas the 'old' parameter is a Unicode (not EBCDIC)
537 * code point in process of being generated */
a6951642
KW		 538#define UTF8_ACCUMULATE(old, new) (__ASSERT_(FITS_IN_8_BITS(new)) \
539 ((old) << UTF_ACCUMULATION_SHIFT) \
155d2738
KW		 540 | ((NATIVE_UTF8_TO_I8((U8)new)) \
541 & UTF_CONTINUATION_MASK))
d06134e5 542
4ab10950 543/* This works in the face of malformed UTF-8. */
4e1ed312
KW		 544#define UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(s, e) \
545 ( UTF8_IS_DOWNGRADEABLE_START(*(s)) \
546 && ( (e) - (s) > 1) \
547 && UTF8_IS_CONTINUATION(*((s)+1)))
4ab10950 548
5aaebcb3 549/* Number of bytes a code point occupies in UTF-8. */
5352a763 550#define NATIVE_SKIP(uv) UVCHR_SKIP(uv)
bd18bd40 551
5aaebcb3
KW		 552/* Most code which says UNISKIP is really thinking in terms of native code
553 * points (0-255) plus all those beyond. This is an imprecise term, but having
2accb712 554 * it means existing code continues to work. For precision, use UVCHR_SKIP,
5352a763
KW		 555 * NATIVE_SKIP, or OFFUNISKIP */
556#define UNISKIP(uv) UVCHR_SKIP(uv)
5aaebcb3 557
3c0792e4
KW		 558/* Longer, but more accurate name */
559#define UTF8_IS_ABOVE_LATIN1_START(c) UTF8_IS_ABOVE_LATIN1(c)
560
a62b247b
KW		 561/* Convert a UTF-8 variant Latin1 character to a native code point value.
562 * Needs just one iteration of accumulate. Should be used only if it is known
563 * that the code point is < 256, and is not UTF-8 invariant. Use the slower
564 * but more general TWO_BYTE_UTF8_TO_NATIVE() which handles any code point
565 * representable by two bytes (which turns out to be up through
566 * MAX_PORTABLE_UTF8_TWO_BYTE). The two parameters are:
567 * HI: a downgradable start byte;
568 * LO: continuation.
569 * */
570#define EIGHT_BIT_UTF8_TO_NATIVE(HI, LO) \
571 ( __ASSERT_(UTF8_IS_DOWNGRADEABLE_START(HI)) \
572 __ASSERT_(UTF8_IS_CONTINUATION(LO)) \
573 LATIN1_TO_NATIVE(UTF8_ACCUMULATE(( \
574 NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), (LO))))
575
94bb8c36 576/* Convert a two (not one) byte utf8 character to a native code point value.
2950f2a7
KW		 577 * Needs just one iteration of accumulate. Should not be used unless it is
578 * known that the two bytes are legal: 1) two-byte start, and 2) continuation.
579 * Note that the result can be larger than 255 if the input character is not
580 * downgradable */
94bb8c36 581#define TWO_BYTE_UTF8_TO_NATIVE(HI, LO) \
a6951642
KW		 582 (__ASSERT_(FITS_IN_8_BITS(HI)) \
583 __ASSERT_(FITS_IN_8_BITS(LO)) \
584 __ASSERT_(PL_utf8skip[HI] == 2) \
585 __ASSERT_(UTF8_IS_CONTINUATION(LO)) \
94bb8c36 586 UNI_TO_NATIVE(UTF8_ACCUMULATE((NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), \
635e76f5 587 (LO))))
94bb8c36
KW		 588
589/* Should never be used, and be deprecated */
590#define TWO_BYTE_UTF8_TO_UNI(HI, LO) NATIVE_TO_UNI(TWO_BYTE_UTF8_TO_NATIVE(HI, LO))
2950f2a7 591
bd18bd40
KW		 592/*
593
594=for apidoc Am|STRLEN|UTF8SKIP|char* s
595returns the number of bytes in the UTF-8 encoded character whose first (perhaps
596only) byte is pointed to by C<s>.
597
598=cut
599 */
2a70536e
KW		 600#define UTF8SKIP(s) PL_utf8skip[*(const U8*)(s)]
601#define UTF8_SKIP(s) UTF8SKIP(s)
d06134e5 602
2d1545e5
KW		 603/* Most code that says 'UNI_' really means the native value for code points up
604 * through 255 */
605#define UNI_IS_INVARIANT(cp) UVCHR_IS_INVARIANT(cp)
606
c2b32798
KW		 607/*
608=for apidoc Am|bool|UTF8_IS_INVARIANT|char c
609
610Evaluates to 1 if the byte C<c> represents the same character when encoded in
611UTF-8 as when not; otherwise evaluates to 0. UTF-8 invariant characters can be
612copied as-is when converting to/from UTF-8, saving time.
613
614In spite of the name, this macro gives the correct result if the input string
615from which C<c> comes is not encoded in UTF-8.
616
617See C<L</UVCHR_IS_INVARIANT>> for checking if a UV is invariant.
618
619=cut
620
621The reason it works on both UTF-8 encoded strings and non-UTF-8 encoded, is
622that it returns TRUE in each for the exact same set of bit patterns. It is
623valid on a subset of what UVCHR_IS_INVARIANT is valid on, so can just use that;
624and the compiler should optimize out anything extraneous given the
625implementation of the latter. The |0 makes sure this isn't mistakenly called
626with a ptr argument.
627*/
5c06326b 628#define UTF8_IS_INVARIANT(c) UVCHR_IS_INVARIANT((c) | 0)
5fc230f1
KW		 629
630/* Like the above, but its name implies a non-UTF8 input, which as the comments
631 * above show, doesn't matter as to its implementation */
38953e5a 632#define NATIVE_BYTE_IS_INVARIANT(c) UVCHR_IS_INVARIANT(c)
d06134e5 633
48ccf5e1
KW		 634/* The macros in the next 4 sets are used to generate the two utf8 or utfebcdic
635 * bytes from an ordinal that is known to fit into exactly two (not one) bytes;
636 * it must be less than 0x3FF to work across both encodings. */
637
638/* These two are helper macros for the other three sets, and should not be used
639 * directly anywhere else. 'translate_function' is either NATIVE_TO_LATIN1
1ff3baa2
KW		 640 * (which works for code points up through 0xFF) or NATIVE_TO_UNI which works
641 * for any code point */
48ccf5e1 642#define __BASE_TWO_BYTE_HI(c, translate_function) \
2863dafa 643 (__ASSERT_(! UVCHR_IS_INVARIANT(c)) \
48ccf5e1 644 I8_TO_NATIVE_UTF8((translate_function(c) >> UTF_ACCUMULATION_SHIFT) \
2863dafa 645 | UTF_START_MARK(2)))
48ccf5e1 646#define __BASE_TWO_BYTE_LO(c, translate_function) \
2863dafa 647 (__ASSERT_(! UVCHR_IS_INVARIANT(c)) \
48ccf5e1 648 I8_TO_NATIVE_UTF8((translate_function(c) & UTF_CONTINUATION_MASK) \
2863dafa 649 | UTF_CONTINUATION_MARK))
48ccf5e1 650
48ccf5e1
KW		 651/* The next two macros should not be used. They were designed to be usable as
652 * the case label of a switch statement, but this doesn't work for EBCDIC. Use
9d0d3a03 653 * regen/unicode_constants.pl instead */
48ccf5e1
KW		 654#define UTF8_TWO_BYTE_HI_nocast(c) __BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI)
655#define UTF8_TWO_BYTE_LO_nocast(c) __BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI)
656
657/* The next two macros are used when the source should be a single byte
658 * character; checked for under DEBUGGING */
659#define UTF8_EIGHT_BIT_HI(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
4c8cd605 660 ( __BASE_TWO_BYTE_HI(c, NATIVE_TO_LATIN1)))
48ccf5e1 661#define UTF8_EIGHT_BIT_LO(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
4c8cd605 662 (__BASE_TWO_BYTE_LO(c, NATIVE_TO_LATIN1)))
48ccf5e1
KW		 663
664/* These final two macros in the series are used when the source can be any
665 * code point whose UTF-8 is known to occupy 2 bytes; they are less efficient
666 * than the EIGHT_BIT versions on EBCDIC platforms. We use the logical '~'
667 * operator instead of "<=" to avoid getting compiler warnings.
d52b8576 668 * MAX_UTF8_TWO_BYTE should be exactly all one bits in the lower few
48ccf5e1
KW		 669 * places, so the ~ works */
670#define UTF8_TWO_BYTE_HI(c) \
671 (__ASSERT_((sizeof(c) == 1) \
d52b8576 672 || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \
4c8cd605 673 (__BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI)))
48ccf5e1
KW		 674#define UTF8_TWO_BYTE_LO(c) \
675 (__ASSERT_((sizeof(c) == 1) \
d52b8576 676 || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \
4c8cd605 677 (__BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI)))
d06134e5 678
e7214ce8
KW		 679/* This is illegal in any well-formed UTF-8 in both EBCDIC and ASCII
680 * as it is only in overlongs. */
681#define ILLEGAL_UTF8_BYTE I8_TO_NATIVE_UTF8(0xC1)
682
7e2040f0 683/*
e3036cf4 684 * 'UTF' is whether or not p is encoded in UTF8. The names 'foo_lazy_if' stem
20df05f4
KW		 685 * from an earlier version of these macros in which they didn't call the
686 * foo_utf8() macros (i.e. were 'lazy') unless they decided that *p is the
687 * beginning of a utf8 character. Now that foo_utf8() determines that itself,
688 * no need to do it again here
7e2040f0 689 */
34aeb2e9
KW		 690#define isIDFIRST_lazy_if(p,UTF) \
691 _is_utf8_FOO(_CC_IDFIRST, (const U8 *) p, "isIDFIRST_lazy_if", \
692 "isIDFIRST_lazy_if_safe", \
693 cBOOL(UTF && ! IN_BYTES), 0, __FILE__,__LINE__)
1d72bdf6 694
da8c1a98
KW		 695#define isIDFIRST_lazy_if_safe(p, e, UTF) \
696 ((IN_BYTES || !UTF) \
697 ? isIDFIRST(*(p)) \
698 : isIDFIRST_utf8_safe(p, e))
699
34aeb2e9
KW		 700#define isWORDCHAR_lazy_if(p,UTF) \
701 _is_utf8_FOO(_CC_IDFIRST, (const U8 *) p, "isWORDCHAR_lazy_if", \
702 "isWORDCHAR_lazy_if_safe", \
703 cBOOL(UTF && ! IN_BYTES), 0, __FILE__,__LINE__)
704
da8c1a98
KW		 705#define isWORDCHAR_lazy_if_safe(p, e, UTF) \
706 ((IN_BYTES || !UTF) \
707 ? isWORDCHAR(*(p)) \
708 : isWORDCHAR_utf8_safe((U8 *) p, (U8 *) e))
709
34aeb2e9
KW		 710#define isALNUM_lazy_if(p,UTF) \
711 _is_utf8_FOO(_CC_IDFIRST, (const U8 *) p, "isALNUM_lazy_if", \
712 "isWORDCHAR_lazy_if_safe", \
713 cBOOL(UTF && ! IN_BYTES), 0, __FILE__,__LINE__)
da8c1a98 714
89ebb4a3
JH		 715#define UTF8_MAXLEN UTF8_MAXBYTES
716
8cb75cc8
KW		 717/* A Unicode character can fold to up to 3 characters */
718#define UTF8_MAX_FOLD_CHAR_EXPAND 3
719
d3481830 720#define IN_BYTES UNLIKELY(CopHINTS_get(PL_curcop) & HINT_BYTES)
bd18bd40
KW		 721
722/*
723
724=for apidoc Am|bool|DO_UTF8|SV* sv
725Returns a bool giving whether or not the PV in C<sv> is to be treated as being
726encoded in UTF-8.
727
728You should use this I<after> a call to C<SvPV()> or one of its variants, in
729case any call to string overloading updates the internal UTF-8 encoding flag.
730
731=cut
732*/
0064a8a9 733#define DO_UTF8(sv) (SvUTF8(sv) && !IN_BYTES)
1ff3baa2
KW		 734
735/* Should all strings be treated as Unicode, and not just UTF-8 encoded ones?
736 * Is so within 'feature unicode_strings' or 'locale :not_characters', and not
737 * within 'use bytes'. UTF-8 locales are not tested for here, but perhaps
738 * could be */
70844984
KW		 739#define IN_UNI_8_BIT \
740 (( ( (CopHINTS_get(PL_curcop) & HINT_UNI_8_BIT)) \
741 || ( CopHINTS_get(PL_curcop) & HINT_LOCALE_PARTIAL \
742 /* -1 below is for :not_characters */ \
743 && _is_in_locale_category(FALSE, -1))) \
744 && (! IN_BYTES))
b36bf33f 745
1d72bdf6 746
c76687c5 747#define UTF8_ALLOW_EMPTY 0x0001 /* Allow a zero length string */
2b5e7bc2 748#define UTF8_GOT_EMPTY UTF8_ALLOW_EMPTY
c76687c5
KW		 749
750/* Allow first byte to be a continuation byte */
1d72bdf6 751#define UTF8_ALLOW_CONTINUATION 0x0002
2b5e7bc2 752#define UTF8_GOT_CONTINUATION UTF8_ALLOW_CONTINUATION
c76687c5 753
2b5e7bc2 754/* Unexpected continuation byte */
1d72bdf6 755#define UTF8_ALLOW_NON_CONTINUATION 0x0004
2b5e7bc2 756#define UTF8_GOT_NON_CONTINUATION UTF8_ALLOW_NON_CONTINUATION
949cf498
KW		 757
758/* expecting more bytes than were available in the string */
759#define UTF8_ALLOW_SHORT 0x0008
2b5e7bc2 760#define UTF8_GOT_SHORT UTF8_ALLOW_SHORT
949cf498 761
94953955
KW		 762/* Overlong sequence; i.e., the code point can be specified in fewer bytes.
763 * First one will convert the overlong to the REPLACEMENT CHARACTER; second
764 * will return what the overlong evaluates to */
949cf498 765#define UTF8_ALLOW_LONG 0x0010
94953955 766#define UTF8_ALLOW_LONG_AND_ITS_VALUE (UTF8_ALLOW_LONG|0x0020)
2b5e7bc2
KW		 767#define UTF8_GOT_LONG UTF8_ALLOW_LONG
768
d60baaa7
KW		 769#define UTF8_ALLOW_OVERFLOW 0x0080
770#define UTF8_GOT_OVERFLOW UTF8_ALLOW_OVERFLOW
949cf498 771
f180b292 772#define UTF8_DISALLOW_SURROGATE 0x0100 /* Unicode surrogates */
2b5e7bc2 773#define UTF8_GOT_SURROGATE UTF8_DISALLOW_SURROGATE
f180b292 774#define UTF8_WARN_SURROGATE 0x0200
949cf498 775
c4e96019
KW		 776/* Unicode non-character code points */
777#define UTF8_DISALLOW_NONCHAR 0x0400
2b5e7bc2 778#define UTF8_GOT_NONCHAR UTF8_DISALLOW_NONCHAR
c4e96019 779#define UTF8_WARN_NONCHAR 0x0800
949cf498 780
c4e96019
KW		 781/* Super-set of Unicode: code points above the legal max */
782#define UTF8_DISALLOW_SUPER 0x1000
2b5e7bc2 783#define UTF8_GOT_SUPER UTF8_DISALLOW_SUPER
c4e96019
KW		 784#define UTF8_WARN_SUPER 0x2000
785
786/* The original UTF-8 standard did not define UTF-8 with start bytes of 0xFE or
787 * 0xFF, though UTF-EBCDIC did. This allowed both versions to represent code
788 * points up to 2 ** 31 - 1. Perl extends UTF-8 so that 0xFE and 0xFF are
789 * usable on ASCII platforms, and 0xFF means something different than
790 * UTF-EBCDIC defines. These changes allow code points of 64 bits (actually
791 * somewhat more) to be represented on both platforms. But these are Perl
792 * extensions, and not likely to be interchangeable with other languages. Note
793 * that on ASCII platforms, FE overflows a signed 32-bit word, and FF an
794 * unsigned one. */
d044b7a7
KW		 795#define UTF8_DISALLOW_PERL_EXTENDED 0x4000
796#define UTF8_GOT_PERL_EXTENDED UTF8_DISALLOW_PERL_EXTENDED
797#define UTF8_WARN_PERL_EXTENDED 0x8000
d35f2ca5 798
57ff5f59
KW		 799/* For back compat, these old names are misleading for overlongs and
800 * UTF_EBCDIC. */
d044b7a7
KW		 801#define UTF8_DISALLOW_ABOVE_31_BIT UTF8_DISALLOW_PERL_EXTENDED
802#define UTF8_GOT_ABOVE_31_BIT UTF8_GOT_PERL_EXTENDED
803#define UTF8_WARN_ABOVE_31_BIT UTF8_WARN_PERL_EXTENDED
804#define UTF8_DISALLOW_FE_FF UTF8_DISALLOW_PERL_EXTENDED
805#define UTF8_WARN_FE_FF UTF8_WARN_PERL_EXTENDED
949cf498 806
f180b292 807#define UTF8_CHECK_ONLY 0x10000
99a765e9 808#define _UTF8_NO_CONFIDENCE_IN_CURLEN 0x20000 /* Internal core use only */
949cf498
KW		 809
810/* For backwards source compatibility. They do nothing, as the default now
811 * includes what they used to mean. The first one's meaning was to allow the
812 * just the single non-character 0xFFFF */
813#define UTF8_ALLOW_FFFF 0
c825ef8c 814#define UTF8_ALLOW_FE_FF 0
949cf498
KW		 815#define UTF8_ALLOW_SURROGATE 0
816
ecc1615f
KW		 817/* C9 refers to Unicode Corrigendum #9: allows but discourages non-chars */
818#define UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE \
819 (UTF8_DISALLOW_SUPER|UTF8_DISALLOW_SURROGATE)
820#define UTF8_WARN_ILLEGAL_C9_INTERCHANGE (UTF8_WARN_SUPER|UTF8_WARN_SURROGATE)
821
d35f2ca5 822#define UTF8_DISALLOW_ILLEGAL_INTERCHANGE \
ecc1615f 823 (UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE|UTF8_DISALLOW_NONCHAR)
949cf498 824#define UTF8_WARN_ILLEGAL_INTERCHANGE \
ecc1615f
KW		 825 (UTF8_WARN_ILLEGAL_C9_INTERCHANGE|UTF8_WARN_NONCHAR)
826
0eb3d6a0
KW		 827/* This is typically used for code that processes UTF-8 input and doesn't want
828 * to have to deal with any malformations that might be present. All such will
829 * be safely replaced by the REPLACEMENT CHARACTER, unless other flags
830 * overriding this are also present. */
2d532c27
KW		 831#define UTF8_ALLOW_ANY ( UTF8_ALLOW_CONTINUATION \
832 |UTF8_ALLOW_NON_CONTINUATION \
833 |UTF8_ALLOW_SHORT \
d60baaa7
KW		 834 |UTF8_ALLOW_LONG \
835 |UTF8_ALLOW_OVERFLOW)
2d532c27
KW		 836
837/* Accept any Perl-extended UTF-8 that evaluates to any UV on the platform, but
838 * not any malformed. This is the default. (Note that UVs above IV_MAX are
839 * deprecated. */
840#define UTF8_ALLOW_ANYUV 0
841#define UTF8_ALLOW_DEFAULT UTF8_ALLOW_ANYUV
1d72bdf6 842
89d986df
KW		 843/*
844=for apidoc Am|bool|UTF8_IS_SURROGATE|const U8 *s|const U8 *e
845
846Evaluates to non-zero if the first few bytes of the string starting at C<s> and
847looking no further than S<C<e - 1>> are well-formed UTF-8 that represents one
848of the Unicode surrogate code points; otherwise it evaluates to 0. If
849non-zero, the value gives how many bytes starting at C<s> comprise the code
850point's representation.
851
852=cut
853 */
854#define UTF8_IS_SURROGATE(s, e) is_SURROGATE_utf8_safe(s, e)
855
856
857#define UTF8_IS_REPLACEMENT(s, send) is_REPLACEMENT_utf8_safe(s,send)
858
859/*
860=for apidoc Am|bool|UTF8_IS_SUPER|const U8 *s|const U8 *e
861
862Recall that Perl recognizes an extension to UTF-8 that can encode code
863points larger than the ones defined by Unicode, which are 0..0x10FFFF.
864
865This macro evaluates to non-zero if the first few bytes of the string starting
866at C<s> and looking no further than S<C<e - 1>> are from this UTF-8 extension;
867otherwise it evaluates to 0. If non-zero, the value gives how many bytes
868starting at C<s> comprise the code point's representation.
869
8700 is returned if the bytes are not well-formed extended UTF-8, or if they
871represent a code point that cannot fit in a UV on the current platform. Hence
872this macro can give different results when run on a 64-bit word machine than on
873one with a 32-bit word size.
0c58a72b 874
89d986df
KW		 875Note that it is deprecated to have code points that are larger than what can
876fit in an IV on the current machine.
7131f24d 877
89d986df
KW		 878=cut
879
880 * ASCII EBCDIC I8
7131f24d
KW		 881 * U+10FFFF: \xF4\x8F\xBF\xBF \xF9\xA1\xBF\xBF\xBF max legal Unicode
882 * U+110000: \xF4\x90\x80\x80 \xF9\xA2\xA0\xA0\xA0
883 * U+110001: \xF4\x90\x80\x81 \xF9\xA2\xA0\xA0\xA1
89d986df
KW		 884 */
885#ifdef EBCDIC
a14e0a36 886# define UTF8_IS_SUPER(s, e) \
89d986df
KW		 887 (( LIKELY((e) > (s) + 4) \
888 && NATIVE_UTF8_TO_I8(*(s)) >= 0xF9 \
889 && ( NATIVE_UTF8_TO_I8(*(s)) > 0xF9 \
3d42f267 890 || (NATIVE_UTF8_TO_I8(*((s) + 1)) >= 0xA2)) \
89d986df 891 && LIKELY((s) + UTF8SKIP(s) <= (e))) \
2b479609 892 ? _is_utf8_char_helper(s, s + UTF8SKIP(s), 0) : 0)
7131f24d 893#else
a14e0a36 894# define UTF8_IS_SUPER(s, e) \
89d986df
KW		 895 (( LIKELY((e) > (s) + 3) \
896 && (*(U8*) (s)) >= 0xF4 \
897 && ((*(U8*) (s)) > 0xF4 || (*((U8*) (s) + 1) >= 0x90))\
898 && LIKELY((s) + UTF8SKIP(s) <= (e))) \
2b479609 899 ? _is_utf8_char_helper(s, s + UTF8SKIP(s), 0) : 0)
7131f24d
KW		 900#endif
901
b96a92fb
KW		 902/* These are now machine generated, and the 'given' clause is no longer
903 * applicable */
0c58a72b 904#define UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s, e) \
89d986df
KW		 905 cBOOL(is_NONCHAR_utf8_safe(s,e))
906
907/*
908=for apidoc Am|bool|UTF8_IS_NONCHAR|const U8 *s|const U8 *e
909
910Evaluates to non-zero if the first few bytes of the string starting at C<s> and
911looking no further than S<C<e - 1>> are well-formed UTF-8 that represents one
912of the Unicode non-character code points; otherwise it evaluates to 0. If
913non-zero, the value gives how many bytes starting at C<s> comprise the code
914point's representation.
915
916=cut
917 */
0c58a72b
KW		 918#define UTF8_IS_NONCHAR(s, e) \
919 UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s, e)
7131f24d 920
c867b360
JH		 921#define UNICODE_SURROGATE_FIRST 0xD800
922#define UNICODE_SURROGATE_LAST 0xDFFF
923#define UNICODE_REPLACEMENT 0xFFFD
924#define UNICODE_BYTE_ORDER_MARK 0xFEFF
1d72bdf6 925
b851fbc1 926/* Though our UTF-8 encoding can go beyond this,
c76687c5 927 * let's be conservative and do as Unicode says. */
b851fbc1
JH		 928#define PERL_UNICODE_MAX 0x10FFFF
929
d044b7a7
KW		 930#define UNICODE_WARN_SURROGATE 0x0001 /* UTF-16 surrogates */
931#define UNICODE_WARN_NONCHAR 0x0002 /* Non-char code points */
932#define UNICODE_WARN_SUPER 0x0004 /* Above 0x10FFFF */
933#define UNICODE_WARN_PERL_EXTENDED 0x0008 /* Above 0x7FFF_FFFF */
934#define UNICODE_WARN_ABOVE_31_BIT UNICODE_WARN_PERL_EXTENDED
935#define UNICODE_DISALLOW_SURROGATE 0x0010
936#define UNICODE_DISALLOW_NONCHAR 0x0020
937#define UNICODE_DISALLOW_SUPER 0x0040
938#define UNICODE_DISALLOW_PERL_EXTENDED 0x0080
939#define UNICODE_DISALLOW_ABOVE_31_BIT UNICODE_DISALLOW_PERL_EXTENDED
33f38593
KW		 940
941#define UNICODE_GOT_SURROGATE UNICODE_DISALLOW_SURROGATE
942#define UNICODE_GOT_NONCHAR UNICODE_DISALLOW_NONCHAR
943#define UNICODE_GOT_SUPER UNICODE_DISALLOW_SUPER
944#define UNICODE_GOT_PERL_EXTENDED UNICODE_DISALLOW_PERL_EXTENDED
945
ecc1615f
KW		 946#define UNICODE_WARN_ILLEGAL_C9_INTERCHANGE \
947 (UNICODE_WARN_SURROGATE|UNICODE_WARN_SUPER)
bb88be5f 948#define UNICODE_WARN_ILLEGAL_INTERCHANGE \
ecc1615f
KW		 949 (UNICODE_WARN_ILLEGAL_C9_INTERCHANGE|UNICODE_WARN_NONCHAR)
950#define UNICODE_DISALLOW_ILLEGAL_C9_INTERCHANGE \
951 (UNICODE_DISALLOW_SURROGATE|UNICODE_DISALLOW_SUPER)
bb88be5f 952#define UNICODE_DISALLOW_ILLEGAL_INTERCHANGE \
ecc1615f 953 (UNICODE_DISALLOW_ILLEGAL_C9_INTERCHANGE|UNICODE_DISALLOW_NONCHAR)
949cf498
KW		 954
955/* For backward source compatibility, as are now the default */
956#define UNICODE_ALLOW_SURROGATE 0
957#define UNICODE_ALLOW_SUPER 0
958#define UNICODE_ALLOW_ANY 0
b851fbc1 959
2d6b3d38
KW		 960/* This matches the 2048 code points between UNICODE_SURROGATE_FIRST (0xD800) and
961 * UNICODE_SURROGATE_LAST (0xDFFF) */
962#define UNICODE_IS_SURROGATE(uv) (((UV) (uv) & (~0xFFFF | 0xF800)) \
963 == 0xD800)
964
646d1759
KW		 965#define UNICODE_IS_REPLACEMENT(uv) ((UV) (uv) == UNICODE_REPLACEMENT)
966#define UNICODE_IS_BYTE_ORDER_MARK(uv) ((UV) (uv) == UNICODE_BYTE_ORDER_MARK)
c149ab20
KW		 967
968/* Is 'uv' one of the 32 contiguous-range noncharacters? */
969#define UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv) ((UV) (uv) >= 0xFDD0 \
970 && (UV) (uv) <= 0xFDEF)
971
972/* Is 'uv' one of the 34 plane-ending noncharacters 0xFFFE, 0xFFFF, 0x1FFFE,
973 * 0x1FFFF, ... 0x10FFFE, 0x10FFFF, given that we know that 'uv' is not above
974 * the Unicode legal max */
975#define UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv) \
976 (((UV) (uv) & 0xFFFE) == 0xFFFE)
977
978#define UNICODE_IS_NONCHAR(uv) \
979 ( UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv) \
980 || ( LIKELY( ! UNICODE_IS_SUPER(uv)) \
981 && UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv)))
982
983#define UNICODE_IS_SUPER(uv) ((UV) (uv) > PERL_UNICODE_MAX)
1d72bdf6 984
ec34087a
KW		 985#define LATIN_SMALL_LETTER_SHARP_S LATIN_SMALL_LETTER_SHARP_S_NATIVE
986#define LATIN_SMALL_LETTER_Y_WITH_DIAERESIS \
987 LATIN_SMALL_LETTER_Y_WITH_DIAERESIS_NATIVE
988#define MICRO_SIGN MICRO_SIGN_NATIVE
989#define LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE \
990 LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE_NATIVE
991#define LATIN_SMALL_LETTER_A_WITH_RING_ABOVE \
992 LATIN_SMALL_LETTER_A_WITH_RING_ABOVE_NATIVE
09091399
JH		 993#define UNICODE_GREEK_CAPITAL_LETTER_SIGMA 0x03A3
994#define UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA 0x03C2
995#define UNICODE_GREEK_SMALL_LETTER_SIGMA 0x03C3
9dcbe121 996#define GREEK_SMALL_LETTER_MU 0x03BC
9e682c18
KW		 997#define GREEK_CAPITAL_LETTER_MU 0x039C /* Upper and title case
998 of MICRON */
999#define LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS 0x0178 /* Also is title case */
0766489e
KW		 1000#ifdef LATIN_CAPITAL_LETTER_SHARP_S_UTF8
1001# define LATIN_CAPITAL_LETTER_SHARP_S 0x1E9E
1002#endif
74894415
KW		 1003#define LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE 0x130
1004#define LATIN_SMALL_LETTER_DOTLESS_I 0x131
9e682c18 1005#define LATIN_SMALL_LETTER_LONG_S 0x017F
a9f50d33
KW		 1006#define LATIN_SMALL_LIGATURE_LONG_S_T 0xFB05
1007#define LATIN_SMALL_LIGATURE_ST 0xFB06
9e682c18
KW		 1008#define KELVIN_SIGN 0x212A
1009#define ANGSTROM_SIGN 0x212B
09091399 1010
9e55ce06 1011#define UNI_DISPLAY_ISPRINT 0x0001
c728cb41
JH		 1012#define UNI_DISPLAY_BACKSLASH 0x0002
1013#define UNI_DISPLAY_QQ (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH)
1014#define UNI_DISPLAY_REGEX (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH)
9e55ce06 1015
5cd46e1f
KW		 1016#define ANYOF_FOLD_SHARP_S(node, input, end) \
1017 (ANYOF_BITMAP_TEST(node, LATIN_SMALL_LETTER_SHARP_S) && \
137165a6 1018 (ANYOF_NONBITMAP(node)) && \
39065660 1019 (ANYOF_FLAGS(node) & ANYOF_LOC_NONBITMAP_FOLD) && \
07b6858f 1020 ((end) > (input) + 1) && \
305b8651 1021 isALPHA_FOLD_EQ((input)[0], 's'))
6302f837 1022
ebc501f0 1023#define SHARP_S_SKIP 2
3b0fc154 1024
6302f837 1025/*
5dca9278
KW		 1026
1027=for apidoc Am|STRLEN|isUTF8_CHAR|const U8 *s|const U8 *e
1028
35f8c9bd 1029Evaluates to non-zero if the first few bytes of the string starting at C<s> and
2c6ed66c
KW		 1030looking no further than S<C<e - 1>> are well-formed UTF-8, as extended by Perl,
1031that represents some code point; otherwise it evaluates to 0. If non-zero, the
1032value gives how many bytes starting at C<s> comprise the code point's
2717076a
KW		 1033representation. Any bytes remaining before C<e>, but beyond the ones needed to
1034form the first code point in C<s>, are not examined.
6302f837 1035
35f8c9bd
KW		 1036The code point can be any that will fit in a UV on this machine, using Perl's
1037extension to official UTF-8 to represent those higher than the Unicode maximum
1038of 0x10FFFF. That means that this macro is used to efficiently decide if the
2717076a
KW		 1039next few bytes in C<s> is legal UTF-8 for a single character.
1040
1041Use C<L</isSTRICT_UTF8_CHAR>> to restrict the acceptable code points to those
1042defined by Unicode to be fully interchangeable across applications;
1043C<L</isC9_STRICT_UTF8_CHAR>> to use the L<Unicode Corrigendum
1044#9|http://www.unicode.org/versions/corrigendum9.html> definition of allowable
1045code points; and C<L</isUTF8_CHAR_flags>> for a more customized definition.
1046
1047Use C<L</is_utf8_string>>, C<L</is_utf8_string_loc>>, and
1048C<L</is_utf8_string_loclen>> to check entire strings.
35f8c9bd
KW		 1049
1050Note that it is deprecated to use code points higher than what will fit in an
1051IV. This macro does not raise any warnings for such code points, treating them
1052as valid.
1053
1054Note also that a UTF-8 INVARIANT character (i.e. ASCII on non-EBCDIC machines)
1055is a valid UTF-8 character.
5dca9278
KW		 1056
1057=cut
1058*/
6302f837 1059
784d4f31
KW		 1060#define isUTF8_CHAR(s, e) \
1061 (UNLIKELY((e) <= (s)) \
1062 ? 0 \
1063 : (UTF8_IS_INVARIANT(*s)) \
df863e43
KW		 1064 ? 1 \
1065 : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \
1066 ? 0 \
1067 : LIKELY(NATIVE_UTF8_TO_I8(*s) <= _IS_UTF8_CHAR_HIGHEST_START_BYTE) \
1068 ? is_UTF8_CHAR_utf8_no_length_checks(s) \
1069 : _is_utf8_char_helper(s, e, 0))
6302f837 1070
3cedd9d9 1071#define is_utf8_char_buf(buf, buf_end) isUTF8_CHAR(buf, buf_end)
976c1b08
KW		 1072#define bytes_from_utf8(s, lenp, is_utf8p) \
1073 bytes_from_utf8_loc(s, lenp, is_utf8p, 0)
3cedd9d9 1074
e23e8bc1
KW		 1075/*
1076
1077=for apidoc Am|STRLEN|isSTRICT_UTF8_CHAR|const U8 *s|const U8 *e
1078
1079Evaluates to non-zero if the first few bytes of the string starting at C<s> and
1080looking no further than S<C<e - 1>> are well-formed UTF-8 that represents some
1081Unicode code point completely acceptable for open interchange between all
1082applications; otherwise it evaluates to 0. If non-zero, the value gives how
2717076a
KW		 1083many bytes starting at C<s> comprise the code point's representation. Any
1084bytes remaining before C<e>, but beyond the ones needed to form the first code
1085point in C<s>, are not examined.
e23e8bc1
KW		 1086
1087The largest acceptable code point is the Unicode maximum 0x10FFFF, and must not
1088be a surrogate nor a non-character code point. Thus this excludes any code
1089point from Perl's extended UTF-8.
1090
1091This is used to efficiently decide if the next few bytes in C<s> is
2717076a
KW		 1092legal Unicode-acceptable UTF-8 for a single character.
1093
1094Use C<L</isC9_STRICT_UTF8_CHAR>> to use the L<Unicode Corrigendum
1095#9|http://www.unicode.org/versions/corrigendum9.html> definition of allowable
1096code points; C<L</isUTF8_CHAR>> to check for Perl's extended UTF-8;
1097and C<L</isUTF8_CHAR_flags>> for a more customized definition.
e23e8bc1 1098
9f2abfde
KW		 1099Use C<L</is_strict_utf8_string>>, C<L</is_strict_utf8_string_loc>>, and
1100C<L</is_strict_utf8_string_loclen>> to check entire strings.
1101
e23e8bc1
KW		 1102=cut
1103*/
1104
1105#define isSTRICT_UTF8_CHAR(s, e) \
1106 (UNLIKELY((e) <= (s)) \
1107 ? 0 \
1108 : (UTF8_IS_INVARIANT(*s)) \
1109 ? 1 \
1110 : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \
1111 ? 0 \
1112 : is_STRICT_UTF8_CHAR_utf8_no_length_checks(s))
1113
a82be82b
KW		 1114/*
1115
1116=for apidoc Am|STRLEN|isC9_STRICT_UTF8_CHAR|const U8 *s|const U8 *e
1117
1118Evaluates to non-zero if the first few bytes of the string starting at C<s> and
1119looking no further than S<C<e - 1>> are well-formed UTF-8 that represents some
1120Unicode non-surrogate code point; otherwise it evaluates to 0. If non-zero,
1121the value gives how many bytes starting at C<s> comprise the code point's
2717076a
KW		 1122representation. Any bytes remaining before C<e>, but beyond the ones needed to
1123form the first code point in C<s>, are not examined.
a82be82b
KW		 1124
1125The largest acceptable code point is the Unicode maximum 0x10FFFF. This
1126differs from C<L</isSTRICT_UTF8_CHAR>> only in that it accepts non-character
1127code points. This corresponds to
1128L<Unicode Corrigendum #9|http://www.unicode.org/versions/corrigendum9.html>.
1129which said that non-character code points are merely discouraged rather than
1130completely forbidden in open interchange. See
1131L<perlunicode/Noncharacter code points>.
1132
2717076a
KW		 1133Use C<L</isUTF8_CHAR>> to check for Perl's extended UTF-8; and
1134C<L</isUTF8_CHAR_flags>> for a more customized definition.
1135
9f2abfde
KW		 1136Use C<L</is_c9strict_utf8_string>>, C<L</is_c9strict_utf8_string_loc>>, and
1137C<L</is_c9strict_utf8_string_loclen>> to check entire strings.
1138
a82be82b
KW		 1139=cut
1140*/
1141
1142#define isC9_STRICT_UTF8_CHAR(s, e) \
1143 (UNLIKELY((e) <= (s)) \
1144 ? 0 \
1145 : (UTF8_IS_INVARIANT(*s)) \
1146 ? 1 \
1147 : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \
1148 ? 0 \
1149 : is_C9_STRICT_UTF8_CHAR_utf8_no_length_checks(s))
1150
25e3a4e0
KW		 1151/*
1152
1153=for apidoc Am|STRLEN|isUTF8_CHAR_flags|const U8 *s|const U8 *e| const U32 flags
1154
1155Evaluates to non-zero if the first few bytes of the string starting at C<s> and
1156looking no further than S<C<e - 1>> are well-formed UTF-8, as extended by Perl,
1157that represents some code point, subject to the restrictions given by C<flags>;
1158otherwise it evaluates to 0. If non-zero, the value gives how many bytes
2717076a
KW		 1159starting at C<s> comprise the code point's representation. Any bytes remaining
1160before C<e>, but beyond the ones needed to form the first code point in C<s>,
1161are not examined.
25e3a4e0
KW		 1162
1163If C<flags> is 0, this gives the same results as C<L</isUTF8_CHAR>>;
1164if C<flags> is C<UTF8_DISALLOW_ILLEGAL_INTERCHANGE>, this gives the same results
1165as C<L</isSTRICT_UTF8_CHAR>>;
1166and if C<flags> is C<UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE>, this gives
1167the same results as C<L</isC9_STRICT_UTF8_CHAR>>.
1168Otherwise C<flags> may be any combination of the C<UTF8_DISALLOW_I<foo>> flags
1169understood by C<L</utf8n_to_uvchr>>, with the same meanings.
1170
1171The three alternative macros are for the most commonly needed validations; they
1172are likely to run somewhat faster than this more general one, as they can be
1173inlined into your code.
1174
9f2abfde
KW		 1175Use L</is_utf8_string_flags>, L</is_utf8_string_loc_flags>, and
1176L</is_utf8_string_loclen_flags> to check entire strings.
1177
25e3a4e0
KW		 1178=cut
1179*/
1180
1181#define isUTF8_CHAR_flags(s, e, flags) \
1182 (UNLIKELY((e) <= (s)) \
1183 ? 0 \
1184 : (UTF8_IS_INVARIANT(*s)) \
1185 ? 1 \
1186 : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \
1187 ? 0 \
1188 : _is_utf8_char_helper(s, e, flags))
1189
6302f837
KW		 1190/* Do not use; should be deprecated. Use isUTF8_CHAR() instead; this is
1191 * retained solely for backwards compatibility */
1192#define IS_UTF8_CHAR(p, n) (isUTF8_CHAR(p, (p) + (n)) == n)
e9a8c099 1193
6a5bc5ac 1194#endif /* PERL_UTF8_H_ */
57f0e7e2 1195
e9a8c099 1196/*
14d04a33 1197 * ex: set ts=8 sts=4 sw=4 et:
e9a8c099 1198 */
Unnamed repository; edit this file 'description' to name the repository.