This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Make isUTF8_char() an inline function
[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) \
fdb72414
KW
77 (__ASSERT_((U8*) (e) > (U8*) (s)) \
78 utf8n_to_uvchr(s, (U8*)(e) - (U8*)(s), lenp, \
79 ckWARN_d(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY))
f9380377
KW
80#define utf8n_to_uvchr(s, len, lenp, flags) \
81 utf8n_to_uvchr_error(s, len, lenp, flags, 0)
37657a5b
KW
82#define utf8n_to_uvchr_error(s, len, lenp, flags, errors) \
83 utf8n_to_uvchr_msgs(s, len, lenp, flags, errors, 0)
de69f3af 84
a0270393 85#define to_uni_fold(c, p, lenp) _to_uni_fold_flags(c, p, lenp, FOLD_FLAGS_FULL)
a239b1e2
KW
86
87#define to_utf8_fold(s, r, lenr) \
607313a1 88 _to_utf8_fold_flags (s, NULL, r, lenr, FOLD_FLAGS_FULL, __FILE__, __LINE__)
a239b1e2 89#define to_utf8_lower(s, r, lenr) \
607313a1 90 _to_utf8_lower_flags(s, NULL, r ,lenr, 0, __FILE__, __LINE__)
a239b1e2 91#define to_utf8_upper(s, r, lenr) \
607313a1 92 _to_utf8_upper_flags(s, NULL, r, lenr, 0, __FILE__, __LINE__)
a239b1e2 93#define to_utf8_title(s, r, lenr) \
607313a1 94 _to_utf8_title_flags(s, NULL, r, lenr ,0, __FILE__, __LINE__)
36bb2ab6 95
eda9cac1
KW
96#define foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \
97 foldEQ_utf8_flags(s1, pe1, l1, u1, s2, pe2, l2, u2, 0)
baa60164 98#define FOLDEQ_UTF8_NOMIX_ASCII (1 << 0)
cea315b6 99#define FOLDEQ_LOCALE (1 << 1)
18f762c3
KW
100#define FOLDEQ_S1_ALREADY_FOLDED (1 << 2)
101#define FOLDEQ_S2_ALREADY_FOLDED (1 << 3)
d635b710
KW
102#define FOLDEQ_S1_FOLDS_SANE (1 << 4)
103#define FOLDEQ_S2_FOLDS_SANE (1 << 5)
a33c29bc 104
e6226b18
KW
105#define ibcmp_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \
106 cBOOL(! foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2))
107
1d72bdf6
NIS
108#ifdef EBCDIC
109/* The equivalent of these macros but implementing UTF-EBCDIC
110 are in the following header file:
111 */
112
113#include "utfebcdic.h"
fd7cb289 114
d06134e5 115#else /* ! EBCDIC */
73c4f7a1
GS
116START_EXTERN_C
117
111e8ed9
KW
118/* How wide can a single UTF-8 encoded character become in bytes. */
119/* NOTE: Strictly speaking Perl's UTF-8 should not be called UTF-8 since UTF-8
120 * is an encoding of Unicode, and Unicode's upper limit, 0x10FFFF, can be
121 * expressed with 4 bytes. However, Perl thinks of UTF-8 as a way to encode
122 * non-negative integers in a binary format, even those above Unicode */
123#define UTF8_MAXBYTES 13
124
a0ed51b3 125#ifdef DOINIT
6f06b55f 126EXTCONST unsigned char PL_utf8skip[] = {
b2635aa8
KW
127/* 0x00 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
128/* 0x10 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
129/* 0x20 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
130/* 0x30 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
131/* 0x40 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
132/* 0x50 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
133/* 0x60 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
134/* 0x70 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
135/* 0x80 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
136/* 0x90 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
137/* 0xA0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
138/* 0xB0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
139/* 0xC0 */ 2,2, /* overlong */
1ff3baa2 140/* 0xC2 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0080 to U+03FF */
b2635aa8
KW
141/* 0xD0 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0400 to U+07FF */
142/* 0xE0 */ 3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3, /* U+0800 to U+FFFF */
143/* 0xF0 */ 4,4,4,4,4,4,4,4,5,5,5,5,6,6, /* above BMP to 2**31 - 1 */
6937f885
KW
144 /* Perl extended (never was official UTF-8). Up to 36 bit */
145/* 0xFE */ 7,
146 /* More extended, Up to 72 bits (64-bit + reserved) */
111e8ed9 147/* 0xFF */ UTF8_MAXBYTES
a0ed51b3
LW
148};
149#else
6f06b55f 150EXTCONST unsigned char PL_utf8skip[];
a0ed51b3
LW
151#endif
152
73c4f7a1 153END_EXTERN_C
7e2040f0 154
1a3756de 155#if defined(_MSC_VER) && _MSC_VER < 1400
6f6d1bab
TC
156/* older MSVC versions have a smallish macro buffer */
157#define PERL_SMALL_MACRO_BUFFER
158#endif
159
59a449d5
KW
160/* Native character to/from iso-8859-1. Are the identity functions on ASCII
161 * platforms */
6f6d1bab 162#ifdef PERL_SMALL_MACRO_BUFFER
1a3756de
TC
163#define NATIVE_TO_LATIN1(ch) ((U8)(ch))
164#define LATIN1_TO_NATIVE(ch) ((U8)(ch))
6f6d1bab 165#else
4c8cd605
KW
166#define NATIVE_TO_LATIN1(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch)))
167#define LATIN1_TO_NATIVE(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch)))
6f6d1bab 168#endif
59a449d5
KW
169
170/* I8 is an intermediate version of UTF-8 used only in UTF-EBCDIC. We thus
171 * consider it to be identical to UTF-8 on ASCII platforms. Strictly speaking
172 * UTF-8 and UTF-EBCDIC are two different things, but we often conflate them
173 * because they are 8-bit encodings that serve the same purpose in Perl, and
174 * rarely do we need to distinguish them. The term "NATIVE_UTF8" applies to
175 * whichever one is applicable on the current platform */
6f6d1bab 176#ifdef PERL_SMALL_MACRO_BUFFER
1d4ea287
DD
177#define NATIVE_UTF8_TO_I8(ch) ((U8) (ch))
178#define I8_TO_NATIVE_UTF8(ch) ((U8) (ch))
6f6d1bab 179#else
4c8cd605
KW
180#define NATIVE_UTF8_TO_I8(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch)))
181#define I8_TO_NATIVE_UTF8(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch)))
6f6d1bab 182#endif
59a449d5 183
1d72bdf6 184/* Transforms in wide UV chars */
4c8cd605
KW
185#define UNI_TO_NATIVE(ch) ((UV) (ch))
186#define NATIVE_TO_UNI(ch) ((UV) (ch))
d7578b48 187
877d9f0d 188/*
9041c2e3 189
a14e0a36
KW
190 The following table is from Unicode 3.2, plus the Perl extensions for above
191 U+10FFFF
877d9f0d 192
a14e0a36 193 Code Points 1st Byte 2nd Byte 3rd 4th 5th 6th 7th 8th-13th
877d9f0d 194
375122d7 195 U+0000..U+007F 00..7F
e1b711da 196 U+0080..U+07FF * C2..DF 80..BF
a14e0a36
KW
197 U+0800..U+0FFF E0 * A0..BF 80..BF
198 U+1000..U+CFFF E1..EC 80..BF 80..BF
199 U+D000..U+D7FF ED 80..9F 80..BF
200 U+D800..U+DFFF ED A0..BF 80..BF (surrogates)
201 U+E000..U+FFFF EE..EF 80..BF 80..BF
202 U+10000..U+3FFFF F0 * 90..BF 80..BF 80..BF
203 U+40000..U+FFFFF F1..F3 80..BF 80..BF 80..BF
204 U+100000..U+10FFFF F4 80..8F 80..BF 80..BF
205 Below are above-Unicode code points
206 U+110000..U+13FFFF F4 90..BF 80..BF 80..BF
207 U+110000..U+1FFFFF F5..F7 80..BF 80..BF 80..BF
208 U+200000..U+FFFFFF F8 * 88..BF 80..BF 80..BF 80..BF
209U+1000000..U+3FFFFFF F9..FB 80..BF 80..BF 80..BF 80..BF
210U+4000000..U+3FFFFFFF FC * 84..BF 80..BF 80..BF 80..BF 80..BF
211U+40000000..U+7FFFFFFF FD 80..BF 80..BF 80..BF 80..BF 80..BF
212U+80000000..U+FFFFFFFFF FE * 82..BF 80..BF 80..BF 80..BF 80..BF 80..BF
213U+1000000000.. FF 80..BF 80..BF 80..BF 80..BF 80..BF * 81..BF 80..BF
877d9f0d 214
e1b711da 215Note the gaps before several of the byte entries above marked by '*'. These are
37e2e78e
KW
216caused by legal UTF-8 avoiding non-shortest encodings: it is technically
217possible to UTF-8-encode a single code point in different ways, but that is
218explicitly forbidden, and the shortest possible encoding should always be used
15824458 219(and that is what Perl does). The non-shortest ones are called 'overlongs'.
8c007b5a 220
877d9f0d
JH
221 */
222
8c007b5a
JH
223/*
224 Another way to look at it, as bits:
225
b2635aa8 226 Code Points 1st Byte 2nd Byte 3rd Byte 4th Byte
8c007b5a 227
b2635aa8
KW
228 0aaa aaaa 0aaa aaaa
229 0000 0bbb bbaa aaaa 110b bbbb 10aa aaaa
230 cccc bbbb bbaa aaaa 1110 cccc 10bb bbbb 10aa aaaa
231 00 000d ddcc cccc bbbb bbaa aaaa 1111 0ddd 10cc cccc 10bb bbbb 10aa aaaa
8c007b5a
JH
232
233As you can see, the continuation bytes all begin with C<10>, and the
e1b711da 234leading bits of the start byte tell how many bytes there are in the
8c007b5a
JH
235encoded character.
236
df863e43
KW
237Perl's extended UTF-8 means we can have start bytes up through FF, though any
238beginning with FF yields a code point that is too large for 32-bit ASCII
239platforms. FF signals to use 13 bytes for the encoded character. This breaks
240the paradigm that the number of leading bits gives how many total bytes there
241are in the character.
65ab9279 242
8c007b5a
JH
243*/
244
6c88483e 245/* Is the representation of the Unicode code point 'cp' the same regardless of
15824458 246 * being encoded in UTF-8 or not? */
2d1545e5 247#define OFFUNI_IS_INVARIANT(cp) isASCII(cp)
15824458 248
c2b32798
KW
249/*
250=for apidoc Am|bool|UVCHR_IS_INVARIANT|UV cp
251
252Evaluates to 1 if the representation of code point C<cp> is the same whether or
253not it is encoded in UTF-8; otherwise evaluates to 0. UTF-8 invariant
254characters can be copied as-is when converting to/from UTF-8, saving time.
255C<cp> is Unicode if above 255; otherwise is platform-native.
256
257=cut
258 */
259
cf1be84e 260#define UVCHR_IS_INVARIANT(cp) OFFUNI_IS_INVARIANT(cp)
38953e5a 261
c9264833
KW
262/* This defines the bits that are to be in the continuation bytes of a multi-byte
263 * UTF-8 encoded character that mark it is a continuation byte. */
264#define UTF_CONTINUATION_MARK 0x80
265
a95ec4fb 266/* Misleadingly named: is the UTF8-encoded byte 'c' part of a variant sequence
5c06326b
KW
267 * in UTF-8? This is the inverse of UTF8_IS_INVARIANT. The |0 makes sure this
268 * isn't mistakenly called with a ptr argument */
a6951642
KW
269#define UTF8_IS_CONTINUED(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
270 ((U8)((c) | 0)) & UTF_CONTINUATION_MARK)
15824458
KW
271
272/* Is the byte 'c' the first byte of a multi-byte UTF8-8 encoded sequence?
273 * This doesn't catch invariants (they are single-byte). It also excludes the
5c06326b
KW
274 * illegal overlong sequences that begin with C0 and C1. The |0 makes sure
275 * this isn't mistakenly called with a ptr argument */
a6951642
KW
276#define UTF8_IS_START(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
277 ((U8)((c) | 0)) >= 0xc2)
15824458 278
858cd8ab
KW
279/* For use in UTF8_IS_CONTINUATION() below */
280#define UTF_IS_CONTINUATION_MASK 0xC0
281
15824458 282/* Is the byte 'c' part of a multi-byte UTF8-8 encoded sequence, and not the
5c06326b
KW
283 * first byte thereof? The |0 makes sure this isn't mistakenly called with a
284 * ptr argument */
a6951642
KW
285#define UTF8_IS_CONTINUATION(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
286 (((U8)((c) | 0)) & UTF_IS_CONTINUATION_MASK) == UTF_CONTINUATION_MARK)
0ae1fa71 287
15824458
KW
288/* Is the UTF8-encoded byte 'c' the first byte of a two byte sequence? Use
289 * UTF8_IS_NEXT_CHAR_DOWNGRADEABLE() instead if the input isn't known to
290 * be well-formed. Masking with 0xfe allows the low bit to be 0 or 1; thus
5c06326b
KW
291 * this matches 0xc[23]. The |0 makes sure this isn't mistakenly called with a
292 * ptr argument */
a6951642
KW
293#define UTF8_IS_DOWNGRADEABLE_START(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
294 (((U8)((c) | 0)) & 0xfe) == 0xc2)
4ab10950 295
15824458 296/* Is the UTF8-encoded byte 'c' the first byte of a sequence of bytes that
5c06326b
KW
297 * represent a code point > 255? The |0 makes sure this isn't mistakenly
298 * called with a ptr argument */
a6951642
KW
299#define UTF8_IS_ABOVE_LATIN1(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
300 ((U8)((c) | 0)) >= 0xc4)
8850bf83 301
15824458
KW
302/* This is the number of low-order bits a continuation byte in a UTF-8 encoded
303 * sequence contributes to the specification of the code point. In the bit
304 * maps above, you see that the first 2 bits are a constant '10', leaving 6 of
305 * real information */
1d72bdf6 306#define UTF_ACCUMULATION_SHIFT 6
b2635aa8 307
fed423a5
KW
308/* ^? is defined to be DEL on ASCII systems. See the definition of toCTRL()
309 * for more */
310#define QUESTION_MARK_CTRL DEL_NATIVE
311
312/* Surrogates, non-character code points and above-Unicode code points are
313 * problematic in some contexts. This allows code that needs to check for
314 * those to to quickly exclude the vast majority of code points it will
315 * encounter */
a6951642
KW
316#define isUTF8_POSSIBLY_PROBLEMATIC(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
317 (U8) c >= 0xED)
fed423a5 318
e23e8bc1
KW
319/* A helper macro for isSTRICT_UTF8_CHAR, so use that one instead of this.
320 * Like is_UTF8_CHAR_utf8_no_length_checks(), this was moved here and LIKELYs
321 * added manually.
322 *
323 STRICT_UTF8_CHAR: Matches legal Unicode UTF-8 variant code points, no
324 surrrogates nor non-character code points
325*/
326/*** GENERATED CODE ***/
327#define is_STRICT_UTF8_CHAR_utf8_no_length_checks(s) \
9f2eed98
P
328( ( 0xC2 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xDF ) ? \
329 ( LIKELY( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) ? 2 : 0 ) \
330: ( 0xE0 == ((const U8*)s)[0] ) ? \
331 ( LIKELY( ( ( ((const U8*)s)[1] & 0xE0 ) == 0xA0 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
332: ( ( 0xE1 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xEC ) || 0xEE == ((const U8*)s)[0] ) ?\
333 ( ( ( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
334: ( 0xED == ((const U8*)s)[0] ) ? \
335 ( LIKELY( ( ( ((const U8*)s)[1] & 0xE0 ) == 0x80 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
336: ( 0xEF == ((const U8*)s)[0] ) ? \
337 ( ( ( 0x80 <= ((const U8*)s)[1] && ((const U8*)s)[1] <= 0xB6 ) || ( 0xB8 <= ((const U8*)s)[1] && ((const U8*)s)[1] <= 0xBE ) ) ?\
338 ( LIKELY( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ? 3 : 0 ) \
339 : ( 0xB7 == ((const U8*)s)[1] ) ? \
340 ( LIKELY( ( ((const U8*)s)[2] & 0xF0 ) == 0x80 || ( ((const U8*)s)[2] & 0xF0 ) == 0xB0 ) ? 3 : 0 )\
341 : ( ( 0xBF == ((const U8*)s)[1] ) && ( 0x80 <= ((const U8*)s)[2] && ((const U8*)s)[2] <= 0xBD ) ) ? 3 : 0 )\
342: ( 0xF0 == ((const U8*)s)[0] ) ? \
343 ( ( ( 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 ) ) ?\
344 ( LIKELY( ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\
345 : ( ((const U8*)s)[1] == 0x9F || ( ( ((const U8*)s)[1] & 0xEF ) == 0xAF ) ) ? \
346 ( ( 0x80 <= ((const U8*)s)[2] && ((const U8*)s)[2] <= 0xBE ) ? \
347 ( LIKELY( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ? 4 : 0 ) \
348 : LIKELY( ( 0xBF == ((const U8*)s)[2] ) && ( 0x80 <= ((const U8*)s)[3] && ((const U8*)s)[3] <= 0xBD ) ) ? 4 : 0 )\
e23e8bc1 349 : 0 ) \
9f2eed98
P
350: ( 0xF1 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xF3 ) ? \
351 ( ( ( ( ((const U8*)s)[1] & 0xC8 ) == 0x80 ) || ( ( ((const U8*)s)[1] & 0xCC ) == 0x88 ) || ( ( ((const U8*)s)[1] & 0xCE ) == 0x8C ) || ( ( ((const U8*)s)[1] & 0xCF ) == 0x8E ) ) ?\
352 ( LIKELY( ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\
353 : ( ( ((const U8*)s)[1] & 0xCF ) == 0x8F ) ? \
354 ( ( 0x80 <= ((const U8*)s)[2] && ((const U8*)s)[2] <= 0xBE ) ? \
355 ( LIKELY( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ? 4 : 0 ) \
356 : LIKELY( ( 0xBF == ((const U8*)s)[2] ) && ( 0x80 <= ((const U8*)s)[3] && ((const U8*)s)[3] <= 0xBD ) ) ? 4 : 0 )\
e23e8bc1 357 : 0 ) \
9f2eed98
P
358: ( 0xF4 == ((const U8*)s)[0] ) ? \
359 ( ( 0x80 <= ((const U8*)s)[1] && ((const U8*)s)[1] <= 0x8E ) ? \
360 ( LIKELY( ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\
361 : ( 0x8F == ((const U8*)s)[1] ) ? \
362 ( ( 0x80 <= ((const U8*)s)[2] && ((const U8*)s)[2] <= 0xBE ) ? \
363 ( LIKELY( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ? 4 : 0 ) \
364 : LIKELY( ( 0xBF == ((const U8*)s)[2] ) && ( 0x80 <= ((const U8*)s)[3] && ((const U8*)s)[3] <= 0xBD ) ) ? 4 : 0 )\
e23e8bc1
KW
365 : 0 ) \
366: 0 )
367
a82be82b
KW
368/* Similarly,
369 C9_STRICT_UTF8_CHAR: Matches legal Unicode UTF-8 variant code
370 points, no surrogates
371 0x0080 - 0xD7FF
372 0xE000 - 0x10FFFF
373*/
374/*** GENERATED CODE ***/
375#define is_C9_STRICT_UTF8_CHAR_utf8_no_length_checks(s) \
9f2eed98
P
376( ( 0xC2 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xDF ) ? \
377 ( LIKELY( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) ? 2 : 0 ) \
378: ( 0xE0 == ((const U8*)s)[0] ) ? \
379 ( LIKELY( ( ( ((const U8*)s)[1] & 0xE0 ) == 0xA0 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
380: ( ( 0xE1 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xEC ) || ( ((const U8*)s)[0] & 0xFE ) == 0xEE ) ?\
381 ( LIKELY( ( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
382: ( 0xED == ((const U8*)s)[0] ) ? \
383 ( LIKELY( ( ( ((const U8*)s)[1] & 0xE0 ) == 0x80 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
384: ( 0xF0 == ((const U8*)s)[0] ) ? \
385 ( 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 )\
386: ( 0xF1 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xF3 ) ? \
387 ( LIKELY( ( ( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) && ( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\
388: 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 389
57ff5f59
KW
390#define UNICODE_IS_PERL_EXTENDED(uv) UNLIKELY((UV) (uv) > 0x7FFFFFFF)
391
fed423a5
KW
392#endif /* EBCDIC vs ASCII */
393
394/* 2**UTF_ACCUMULATION_SHIFT - 1 */
395#define UTF_CONTINUATION_MASK ((U8) ((1U << UTF_ACCUMULATION_SHIFT) - 1))
396
7028aeba
KW
397/* Internal macro to be used only in this file to aid in constructing other
398 * publicly accessible macros.
399 * The number of bytes required to express this uv in UTF-8, for just those
400 * uv's requiring 2 through 6 bytes, as these are common to all platforms and
401 * word sizes. The number of bytes needed is given by the number of leading 1
402 * bits in the start byte. There are 32 start bytes that have 2 initial 1 bits
403 * (C0-DF); there are 16 that have 3 initial 1 bits (E0-EF); 8 that have 4
404 * initial 1 bits (F0-F8); 4 that have 5 initial 1 bits (F9-FB), and 2 that
405 * have 6 initial 1 bits (FC-FD). The largest number a string of n bytes can
406 * represent is (the number of possible start bytes for 'n')
407 * * (the number of possiblities for each start byte
408 * The latter in turn is
409 * 2 ** ( (how many continuation bytes there are)
410 * * (the number of bits of information each
411 * continuation byte holds))
412 *
413 * If we were on a platform where we could use a fast find first set bit
414 * instruction (or count leading zeros instruction) this could be replaced by
415 * using that to find the log2 of the uv, and divide that by the number of bits
416 * of information in each continuation byte, adjusting for large cases and how
417 * much information is in a start byte for that length */
72164d3a 418#define __COMMON_UNI_SKIP(uv) \
7028aeba
KW
419 (UV) (uv) < (32 * (1U << ( UTF_ACCUMULATION_SHIFT))) ? 2 : \
420 (UV) (uv) < (16 * (1U << (2 * UTF_ACCUMULATION_SHIFT))) ? 3 : \
421 (UV) (uv) < ( 8 * (1U << (3 * UTF_ACCUMULATION_SHIFT))) ? 4 : \
422 (UV) (uv) < ( 4 * (1U << (4 * UTF_ACCUMULATION_SHIFT))) ? 5 : \
423 (UV) (uv) < ( 2 * (1U << (5 * UTF_ACCUMULATION_SHIFT))) ? 6 :
72164d3a
KW
424
425/* Internal macro to be used only in this file.
426 * This adds to __COMMON_UNI_SKIP the details at this platform's upper range.
fed423a5 427 * For any-sized EBCDIC platforms, or 64-bit ASCII ones, we need one more test
72164d3a
KW
428 * to see if just 7 bytes is needed, or if the maximum is needed. For 32-bit
429 * ASCII platforms, everything is representable by 7 bytes */
fed423a5 430#if defined(UV_IS_QUAD) || defined(EBCDIC)
72164d3a 431# define __BASE_UNI_SKIP(uv) (__COMMON_UNI_SKIP(uv) \
7028aeba 432 (UV) (uv) < ((UV) 1U << (6 * UTF_ACCUMULATION_SHIFT)) ? 7 : UTF8_MAXBYTES)
1d68d6cd 433#else
72164d3a 434# define __BASE_UNI_SKIP(uv) (__COMMON_UNI_SKIP(uv) 7)
1d68d6cd
SC
435#endif
436
fed423a5
KW
437/* The next two macros use the base macro defined above, and add in the tests
438 * at the low-end of the range, for just 1 byte, yielding complete macros,
439 * publicly accessible. */
440
441/* Input is a true Unicode (not-native) code point */
442#define OFFUNISKIP(uv) (OFFUNI_IS_INVARIANT(uv) ? 1 : __BASE_UNI_SKIP(uv))
2084b489 443
5352a763
KW
444/*
445
446=for apidoc Am|STRLEN|UVCHR_SKIP|UV cp
447returns the number of bytes required to represent the code point C<cp> when
448encoded as UTF-8. C<cp> is a native (ASCII or EBCDIC) code point if less than
449255; a Unicode code point otherwise.
450
451=cut
452 */
fdb6583d 453#define UVCHR_SKIP(uv) ( UVCHR_IS_INVARIANT(uv) ? 1 : __BASE_UNI_SKIP(uv))
5352a763 454
b651802e
KW
455/* The largest code point representable by two UTF-8 bytes on this platform.
456 * As explained in the comments for __COMMON_UNI_SKIP, 32 start bytes with
fed423a5 457 * UTF_ACCUMULATION_SHIFT bits of information each */
aa206fb7
KW
458#define MAX_UTF8_TWO_BYTE (32 * (1U << UTF_ACCUMULATION_SHIFT) - 1)
459
b651802e
KW
460/* The largest code point representable by two UTF-8 bytes on any platform that
461 * Perl runs on. This value is constrained by EBCDIC which has 5 bits per
462 * continuation byte */
aa206fb7
KW
463#define MAX_PORTABLE_UTF8_TWO_BYTE (32 * (1U << 5) - 1)
464
c03c0950
KW
465/* The maximum number of UTF-8 bytes a single Unicode character can
466 * uppercase/lowercase/fold into. Unicode guarantees that the maximum
467 * expansion is UTF8_MAX_FOLD_CHAR_EXPAND characters, but any above-Unicode
468 * code point will fold to itself, so we only have to look at the expansion of
469 * the maximum Unicode code point. But this number may be less than the space
470 * occupied by a very large code point under Perl's extended UTF-8. We have to
471 * make it large enough to fit any single character. (It turns out that ASCII
472 * and EBCDIC differ in which is larger) */
473#define UTF8_MAXBYTES_CASE \
474 (UTF8_MAXBYTES >= (UTF8_MAX_FOLD_CHAR_EXPAND * OFFUNISKIP(0x10FFFF)) \
475 ? UTF8_MAXBYTES \
476 : (UTF8_MAX_FOLD_CHAR_EXPAND * OFFUNISKIP(0x10FFFF)))
477
d06134e5
KW
478/* Rest of these are attributes of Unicode and perl's internals rather than the
479 * encoding, or happen to be the same in both ASCII and EBCDIC (at least at
480 * this level; the macros that some of these call may have different
481 * definitions in the two encodings */
482
59a449d5
KW
483/* In domain restricted to ASCII, these may make more sense to the reader than
484 * the ones with Latin1 in the name */
485#define NATIVE_TO_ASCII(ch) NATIVE_TO_LATIN1(ch)
486#define ASCII_TO_NATIVE(ch) LATIN1_TO_NATIVE(ch)
487
488/* More or less misleadingly-named defines, retained for back compat */
489#define NATIVE_TO_UTF(ch) NATIVE_UTF8_TO_I8(ch)
490#define NATIVE_TO_I8(ch) NATIVE_UTF8_TO_I8(ch)
491#define UTF_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch)
492#define I8_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch)
493#define NATIVE8_TO_UNI(ch) NATIVE_TO_LATIN1(ch)
d06134e5 494
c0236afe 495/* This defines the 1-bits that are to be in the first byte of a multi-byte
97d0ceda
KW
496 * UTF-8 encoded character that mark it as a start byte and give the number of
497 * bytes that comprise the character. 'len' is the number of bytes in the
498 * multi-byte sequence. */
c0236afe
KW
499#define UTF_START_MARK(len) (((len) > 7) ? 0xFF : (0xFF & (0xFE << (7-(len)))))
500
501/* Masks out the initial one bits in a start byte, leaving the real data ones.
502 * Doesn't work on an invariant byte. 'len' is the number of bytes in the
503 * multi-byte sequence that comprises the character. */
504#define UTF_START_MASK(len) (((len) >= 7) ? 0x00 : (0x1F >> ((len)-2)))
505
537124e4
KW
506/* Adds a UTF8 continuation byte 'new' of information to a running total code
507 * point 'old' of all the continuation bytes so far. This is designed to be
155d2738
KW
508 * used in a loop to convert from UTF-8 to the code point represented. Note
509 * that this is asymmetric on EBCDIC platforms, in that the 'new' parameter is
510 * the UTF-EBCDIC byte, whereas the 'old' parameter is a Unicode (not EBCDIC)
511 * code point in process of being generated */
a6951642
KW
512#define UTF8_ACCUMULATE(old, new) (__ASSERT_(FITS_IN_8_BITS(new)) \
513 ((old) << UTF_ACCUMULATION_SHIFT) \
155d2738
KW
514 | ((NATIVE_UTF8_TO_I8((U8)new)) \
515 & UTF_CONTINUATION_MASK))
d06134e5 516
4ab10950 517/* This works in the face of malformed UTF-8. */
4e1ed312
KW
518#define UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(s, e) \
519 ( UTF8_IS_DOWNGRADEABLE_START(*(s)) \
520 && ( (e) - (s) > 1) \
521 && UTF8_IS_CONTINUATION(*((s)+1)))
4ab10950 522
5aaebcb3 523/* Number of bytes a code point occupies in UTF-8. */
5352a763 524#define NATIVE_SKIP(uv) UVCHR_SKIP(uv)
bd18bd40 525
5aaebcb3
KW
526/* Most code which says UNISKIP is really thinking in terms of native code
527 * points (0-255) plus all those beyond. This is an imprecise term, but having
2accb712 528 * it means existing code continues to work. For precision, use UVCHR_SKIP,
5352a763
KW
529 * NATIVE_SKIP, or OFFUNISKIP */
530#define UNISKIP(uv) UVCHR_SKIP(uv)
5aaebcb3 531
3c0792e4
KW
532/* Longer, but more accurate name */
533#define UTF8_IS_ABOVE_LATIN1_START(c) UTF8_IS_ABOVE_LATIN1(c)
534
a62b247b
KW
535/* Convert a UTF-8 variant Latin1 character to a native code point value.
536 * Needs just one iteration of accumulate. Should be used only if it is known
537 * that the code point is < 256, and is not UTF-8 invariant. Use the slower
538 * but more general TWO_BYTE_UTF8_TO_NATIVE() which handles any code point
539 * representable by two bytes (which turns out to be up through
540 * MAX_PORTABLE_UTF8_TWO_BYTE). The two parameters are:
541 * HI: a downgradable start byte;
542 * LO: continuation.
543 * */
544#define EIGHT_BIT_UTF8_TO_NATIVE(HI, LO) \
545 ( __ASSERT_(UTF8_IS_DOWNGRADEABLE_START(HI)) \
546 __ASSERT_(UTF8_IS_CONTINUATION(LO)) \
547 LATIN1_TO_NATIVE(UTF8_ACCUMULATE(( \
548 NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), (LO))))
549
94bb8c36 550/* Convert a two (not one) byte utf8 character to a native code point value.
2950f2a7
KW
551 * Needs just one iteration of accumulate. Should not be used unless it is
552 * known that the two bytes are legal: 1) two-byte start, and 2) continuation.
553 * Note that the result can be larger than 255 if the input character is not
554 * downgradable */
94bb8c36 555#define TWO_BYTE_UTF8_TO_NATIVE(HI, LO) \
a6951642
KW
556 (__ASSERT_(FITS_IN_8_BITS(HI)) \
557 __ASSERT_(FITS_IN_8_BITS(LO)) \
558 __ASSERT_(PL_utf8skip[HI] == 2) \
559 __ASSERT_(UTF8_IS_CONTINUATION(LO)) \
94bb8c36 560 UNI_TO_NATIVE(UTF8_ACCUMULATE((NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), \
635e76f5 561 (LO))))
94bb8c36
KW
562
563/* Should never be used, and be deprecated */
564#define TWO_BYTE_UTF8_TO_UNI(HI, LO) NATIVE_TO_UNI(TWO_BYTE_UTF8_TO_NATIVE(HI, LO))
2950f2a7 565
bd18bd40
KW
566/*
567
568=for apidoc Am|STRLEN|UTF8SKIP|char* s
569returns the number of bytes in the UTF-8 encoded character whose first (perhaps
570only) byte is pointed to by C<s>.
571
572=cut
573 */
2a70536e
KW
574#define UTF8SKIP(s) PL_utf8skip[*(const U8*)(s)]
575#define UTF8_SKIP(s) UTF8SKIP(s)
d06134e5 576
2d1545e5
KW
577/* Most code that says 'UNI_' really means the native value for code points up
578 * through 255 */
579#define UNI_IS_INVARIANT(cp) UVCHR_IS_INVARIANT(cp)
580
c2b32798
KW
581/*
582=for apidoc Am|bool|UTF8_IS_INVARIANT|char c
583
584Evaluates to 1 if the byte C<c> represents the same character when encoded in
585UTF-8 as when not; otherwise evaluates to 0. UTF-8 invariant characters can be
586copied as-is when converting to/from UTF-8, saving time.
587
588In spite of the name, this macro gives the correct result if the input string
589from which C<c> comes is not encoded in UTF-8.
590
591See C<L</UVCHR_IS_INVARIANT>> for checking if a UV is invariant.
592
593=cut
594
595The reason it works on both UTF-8 encoded strings and non-UTF-8 encoded, is
596that it returns TRUE in each for the exact same set of bit patterns. It is
597valid on a subset of what UVCHR_IS_INVARIANT is valid on, so can just use that;
598and the compiler should optimize out anything extraneous given the
599implementation of the latter. The |0 makes sure this isn't mistakenly called
600with a ptr argument.
601*/
5c06326b 602#define UTF8_IS_INVARIANT(c) UVCHR_IS_INVARIANT((c) | 0)
5fc230f1
KW
603
604/* Like the above, but its name implies a non-UTF8 input, which as the comments
605 * above show, doesn't matter as to its implementation */
38953e5a 606#define NATIVE_BYTE_IS_INVARIANT(c) UVCHR_IS_INVARIANT(c)
d06134e5 607
48ccf5e1
KW
608/* The macros in the next 4 sets are used to generate the two utf8 or utfebcdic
609 * bytes from an ordinal that is known to fit into exactly two (not one) bytes;
610 * it must be less than 0x3FF to work across both encodings. */
611
612/* These two are helper macros for the other three sets, and should not be used
613 * directly anywhere else. 'translate_function' is either NATIVE_TO_LATIN1
1ff3baa2
KW
614 * (which works for code points up through 0xFF) or NATIVE_TO_UNI which works
615 * for any code point */
48ccf5e1 616#define __BASE_TWO_BYTE_HI(c, translate_function) \
2863dafa 617 (__ASSERT_(! UVCHR_IS_INVARIANT(c)) \
48ccf5e1 618 I8_TO_NATIVE_UTF8((translate_function(c) >> UTF_ACCUMULATION_SHIFT) \
2863dafa 619 | UTF_START_MARK(2)))
48ccf5e1 620#define __BASE_TWO_BYTE_LO(c, translate_function) \
2863dafa 621 (__ASSERT_(! UVCHR_IS_INVARIANT(c)) \
48ccf5e1 622 I8_TO_NATIVE_UTF8((translate_function(c) & UTF_CONTINUATION_MASK) \
2863dafa 623 | UTF_CONTINUATION_MARK))
48ccf5e1 624
48ccf5e1
KW
625/* The next two macros should not be used. They were designed to be usable as
626 * the case label of a switch statement, but this doesn't work for EBCDIC. Use
9d0d3a03 627 * regen/unicode_constants.pl instead */
48ccf5e1
KW
628#define UTF8_TWO_BYTE_HI_nocast(c) __BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI)
629#define UTF8_TWO_BYTE_LO_nocast(c) __BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI)
630
631/* The next two macros are used when the source should be a single byte
632 * character; checked for under DEBUGGING */
633#define UTF8_EIGHT_BIT_HI(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
4c8cd605 634 ( __BASE_TWO_BYTE_HI(c, NATIVE_TO_LATIN1)))
48ccf5e1 635#define UTF8_EIGHT_BIT_LO(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
4c8cd605 636 (__BASE_TWO_BYTE_LO(c, NATIVE_TO_LATIN1)))
48ccf5e1
KW
637
638/* These final two macros in the series are used when the source can be any
639 * code point whose UTF-8 is known to occupy 2 bytes; they are less efficient
640 * than the EIGHT_BIT versions on EBCDIC platforms. We use the logical '~'
641 * operator instead of "<=" to avoid getting compiler warnings.
d52b8576 642 * MAX_UTF8_TWO_BYTE should be exactly all one bits in the lower few
48ccf5e1
KW
643 * places, so the ~ works */
644#define UTF8_TWO_BYTE_HI(c) \
645 (__ASSERT_((sizeof(c) == 1) \
d52b8576 646 || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \
4c8cd605 647 (__BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI)))
48ccf5e1
KW
648#define UTF8_TWO_BYTE_LO(c) \
649 (__ASSERT_((sizeof(c) == 1) \
d52b8576 650 || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \
4c8cd605 651 (__BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI)))
d06134e5 652
e7214ce8
KW
653/* This is illegal in any well-formed UTF-8 in both EBCDIC and ASCII
654 * as it is only in overlongs. */
655#define ILLEGAL_UTF8_BYTE I8_TO_NATIVE_UTF8(0xC1)
656
7e2040f0 657/*
e3036cf4 658 * 'UTF' is whether or not p is encoded in UTF8. The names 'foo_lazy_if' stem
20df05f4
KW
659 * from an earlier version of these macros in which they didn't call the
660 * foo_utf8() macros (i.e. were 'lazy') unless they decided that *p is the
661 * beginning of a utf8 character. Now that foo_utf8() determines that itself,
662 * no need to do it again here
7e2040f0 663 */
34aeb2e9
KW
664#define isIDFIRST_lazy_if(p,UTF) \
665 _is_utf8_FOO(_CC_IDFIRST, (const U8 *) p, "isIDFIRST_lazy_if", \
666 "isIDFIRST_lazy_if_safe", \
667 cBOOL(UTF && ! IN_BYTES), 0, __FILE__,__LINE__)
1d72bdf6 668
da8c1a98
KW
669#define isIDFIRST_lazy_if_safe(p, e, UTF) \
670 ((IN_BYTES || !UTF) \
671 ? isIDFIRST(*(p)) \
672 : isIDFIRST_utf8_safe(p, e))
673
34aeb2e9
KW
674#define isWORDCHAR_lazy_if(p,UTF) \
675 _is_utf8_FOO(_CC_IDFIRST, (const U8 *) p, "isWORDCHAR_lazy_if", \
676 "isWORDCHAR_lazy_if_safe", \
677 cBOOL(UTF && ! IN_BYTES), 0, __FILE__,__LINE__)
678
da8c1a98
KW
679#define isWORDCHAR_lazy_if_safe(p, e, UTF) \
680 ((IN_BYTES || !UTF) \
681 ? isWORDCHAR(*(p)) \
682 : isWORDCHAR_utf8_safe((U8 *) p, (U8 *) e))
683
34aeb2e9
KW
684#define isALNUM_lazy_if(p,UTF) \
685 _is_utf8_FOO(_CC_IDFIRST, (const U8 *) p, "isALNUM_lazy_if", \
686 "isWORDCHAR_lazy_if_safe", \
687 cBOOL(UTF && ! IN_BYTES), 0, __FILE__,__LINE__)
da8c1a98 688
89ebb4a3
JH
689#define UTF8_MAXLEN UTF8_MAXBYTES
690
8cb75cc8
KW
691/* A Unicode character can fold to up to 3 characters */
692#define UTF8_MAX_FOLD_CHAR_EXPAND 3
693
d3481830 694#define IN_BYTES UNLIKELY(CopHINTS_get(PL_curcop) & HINT_BYTES)
bd18bd40
KW
695
696/*
697
698=for apidoc Am|bool|DO_UTF8|SV* sv
699Returns a bool giving whether or not the PV in C<sv> is to be treated as being
700encoded in UTF-8.
701
702You should use this I<after> a call to C<SvPV()> or one of its variants, in
703case any call to string overloading updates the internal UTF-8 encoding flag.
704
705=cut
706*/
0064a8a9 707#define DO_UTF8(sv) (SvUTF8(sv) && !IN_BYTES)
1ff3baa2
KW
708
709/* Should all strings be treated as Unicode, and not just UTF-8 encoded ones?
710 * Is so within 'feature unicode_strings' or 'locale :not_characters', and not
711 * within 'use bytes'. UTF-8 locales are not tested for here, but perhaps
712 * could be */
70844984
KW
713#define IN_UNI_8_BIT \
714 (( ( (CopHINTS_get(PL_curcop) & HINT_UNI_8_BIT)) \
715 || ( CopHINTS_get(PL_curcop) & HINT_LOCALE_PARTIAL \
716 /* -1 below is for :not_characters */ \
717 && _is_in_locale_category(FALSE, -1))) \
718 && (! IN_BYTES))
b36bf33f 719
1d72bdf6 720
c76687c5 721#define UTF8_ALLOW_EMPTY 0x0001 /* Allow a zero length string */
2b5e7bc2 722#define UTF8_GOT_EMPTY UTF8_ALLOW_EMPTY
c76687c5
KW
723
724/* Allow first byte to be a continuation byte */
1d72bdf6 725#define UTF8_ALLOW_CONTINUATION 0x0002
2b5e7bc2 726#define UTF8_GOT_CONTINUATION UTF8_ALLOW_CONTINUATION
c76687c5 727
cd01d3b1 728/* Unexpected non-continuation byte */
1d72bdf6 729#define UTF8_ALLOW_NON_CONTINUATION 0x0004
2b5e7bc2 730#define UTF8_GOT_NON_CONTINUATION UTF8_ALLOW_NON_CONTINUATION
949cf498
KW
731
732/* expecting more bytes than were available in the string */
733#define UTF8_ALLOW_SHORT 0x0008
2b5e7bc2 734#define UTF8_GOT_SHORT UTF8_ALLOW_SHORT
949cf498 735
94953955
KW
736/* Overlong sequence; i.e., the code point can be specified in fewer bytes.
737 * First one will convert the overlong to the REPLACEMENT CHARACTER; second
738 * will return what the overlong evaluates to */
949cf498 739#define UTF8_ALLOW_LONG 0x0010
94953955 740#define UTF8_ALLOW_LONG_AND_ITS_VALUE (UTF8_ALLOW_LONG|0x0020)
2b5e7bc2
KW
741#define UTF8_GOT_LONG UTF8_ALLOW_LONG
742
d60baaa7
KW
743#define UTF8_ALLOW_OVERFLOW 0x0080
744#define UTF8_GOT_OVERFLOW UTF8_ALLOW_OVERFLOW
949cf498 745
f180b292 746#define UTF8_DISALLOW_SURROGATE 0x0100 /* Unicode surrogates */
2b5e7bc2 747#define UTF8_GOT_SURROGATE UTF8_DISALLOW_SURROGATE
f180b292 748#define UTF8_WARN_SURROGATE 0x0200
949cf498 749
c4e96019
KW
750/* Unicode non-character code points */
751#define UTF8_DISALLOW_NONCHAR 0x0400
2b5e7bc2 752#define UTF8_GOT_NONCHAR UTF8_DISALLOW_NONCHAR
c4e96019 753#define UTF8_WARN_NONCHAR 0x0800
949cf498 754
c4e96019
KW
755/* Super-set of Unicode: code points above the legal max */
756#define UTF8_DISALLOW_SUPER 0x1000
2b5e7bc2 757#define UTF8_GOT_SUPER UTF8_DISALLOW_SUPER
c4e96019
KW
758#define UTF8_WARN_SUPER 0x2000
759
760/* The original UTF-8 standard did not define UTF-8 with start bytes of 0xFE or
761 * 0xFF, though UTF-EBCDIC did. This allowed both versions to represent code
762 * points up to 2 ** 31 - 1. Perl extends UTF-8 so that 0xFE and 0xFF are
763 * usable on ASCII platforms, and 0xFF means something different than
764 * UTF-EBCDIC defines. These changes allow code points of 64 bits (actually
765 * somewhat more) to be represented on both platforms. But these are Perl
766 * extensions, and not likely to be interchangeable with other languages. Note
767 * that on ASCII platforms, FE overflows a signed 32-bit word, and FF an
768 * unsigned one. */
d044b7a7
KW
769#define UTF8_DISALLOW_PERL_EXTENDED 0x4000
770#define UTF8_GOT_PERL_EXTENDED UTF8_DISALLOW_PERL_EXTENDED
771#define UTF8_WARN_PERL_EXTENDED 0x8000
d35f2ca5 772
57ff5f59
KW
773/* For back compat, these old names are misleading for overlongs and
774 * UTF_EBCDIC. */
d044b7a7
KW
775#define UTF8_DISALLOW_ABOVE_31_BIT UTF8_DISALLOW_PERL_EXTENDED
776#define UTF8_GOT_ABOVE_31_BIT UTF8_GOT_PERL_EXTENDED
777#define UTF8_WARN_ABOVE_31_BIT UTF8_WARN_PERL_EXTENDED
778#define UTF8_DISALLOW_FE_FF UTF8_DISALLOW_PERL_EXTENDED
779#define UTF8_WARN_FE_FF UTF8_WARN_PERL_EXTENDED
949cf498 780
f180b292 781#define UTF8_CHECK_ONLY 0x10000
99a765e9 782#define _UTF8_NO_CONFIDENCE_IN_CURLEN 0x20000 /* Internal core use only */
949cf498
KW
783
784/* For backwards source compatibility. They do nothing, as the default now
785 * includes what they used to mean. The first one's meaning was to allow the
786 * just the single non-character 0xFFFF */
787#define UTF8_ALLOW_FFFF 0
c825ef8c 788#define UTF8_ALLOW_FE_FF 0
949cf498
KW
789#define UTF8_ALLOW_SURROGATE 0
790
ecc1615f
KW
791/* C9 refers to Unicode Corrigendum #9: allows but discourages non-chars */
792#define UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE \
793 (UTF8_DISALLOW_SUPER|UTF8_DISALLOW_SURROGATE)
794#define UTF8_WARN_ILLEGAL_C9_INTERCHANGE (UTF8_WARN_SUPER|UTF8_WARN_SURROGATE)
795
d35f2ca5 796#define UTF8_DISALLOW_ILLEGAL_INTERCHANGE \
ecc1615f 797 (UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE|UTF8_DISALLOW_NONCHAR)
949cf498 798#define UTF8_WARN_ILLEGAL_INTERCHANGE \
ecc1615f
KW
799 (UTF8_WARN_ILLEGAL_C9_INTERCHANGE|UTF8_WARN_NONCHAR)
800
0eb3d6a0
KW
801/* This is typically used for code that processes UTF-8 input and doesn't want
802 * to have to deal with any malformations that might be present. All such will
803 * be safely replaced by the REPLACEMENT CHARACTER, unless other flags
804 * overriding this are also present. */
2d532c27
KW
805#define UTF8_ALLOW_ANY ( UTF8_ALLOW_CONTINUATION \
806 |UTF8_ALLOW_NON_CONTINUATION \
807 |UTF8_ALLOW_SHORT \
d60baaa7
KW
808 |UTF8_ALLOW_LONG \
809 |UTF8_ALLOW_OVERFLOW)
2d532c27
KW
810
811/* Accept any Perl-extended UTF-8 that evaluates to any UV on the platform, but
cd01d3b1 812 * not any malformed. This is the default. */
2d532c27
KW
813#define UTF8_ALLOW_ANYUV 0
814#define UTF8_ALLOW_DEFAULT UTF8_ALLOW_ANYUV
1d72bdf6 815
89d986df
KW
816/*
817=for apidoc Am|bool|UTF8_IS_SURROGATE|const U8 *s|const U8 *e
818
819Evaluates to non-zero if the first few bytes of the string starting at C<s> and
820looking no further than S<C<e - 1>> are well-formed UTF-8 that represents one
821of the Unicode surrogate code points; otherwise it evaluates to 0. If
822non-zero, the value gives how many bytes starting at C<s> comprise the code
823point's representation.
824
825=cut
826 */
827#define UTF8_IS_SURROGATE(s, e) is_SURROGATE_utf8_safe(s, e)
828
829
830#define UTF8_IS_REPLACEMENT(s, send) is_REPLACEMENT_utf8_safe(s,send)
831
832/*
833=for apidoc Am|bool|UTF8_IS_SUPER|const U8 *s|const U8 *e
834
835Recall that Perl recognizes an extension to UTF-8 that can encode code
836points larger than the ones defined by Unicode, which are 0..0x10FFFF.
837
838This macro evaluates to non-zero if the first few bytes of the string starting
839at C<s> and looking no further than S<C<e - 1>> are from this UTF-8 extension;
840otherwise it evaluates to 0. If non-zero, the value gives how many bytes
841starting at C<s> comprise the code point's representation.
842
8430 is returned if the bytes are not well-formed extended UTF-8, or if they
844represent a code point that cannot fit in a UV on the current platform. Hence
845this macro can give different results when run on a 64-bit word machine than on
846one with a 32-bit word size.
0c58a72b 847
89d986df
KW
848Note that it is deprecated to have code points that are larger than what can
849fit in an IV on the current machine.
7131f24d 850
89d986df
KW
851=cut
852
853 * ASCII EBCDIC I8
7131f24d
KW
854 * U+10FFFF: \xF4\x8F\xBF\xBF \xF9\xA1\xBF\xBF\xBF max legal Unicode
855 * U+110000: \xF4\x90\x80\x80 \xF9\xA2\xA0\xA0\xA0
856 * U+110001: \xF4\x90\x80\x81 \xF9\xA2\xA0\xA0\xA1
89d986df
KW
857 */
858#ifdef EBCDIC
a14e0a36 859# define UTF8_IS_SUPER(s, e) \
89d986df
KW
860 (( LIKELY((e) > (s) + 4) \
861 && NATIVE_UTF8_TO_I8(*(s)) >= 0xF9 \
862 && ( NATIVE_UTF8_TO_I8(*(s)) > 0xF9 \
3d42f267 863 || (NATIVE_UTF8_TO_I8(*((s) + 1)) >= 0xA2)) \
89d986df 864 && LIKELY((s) + UTF8SKIP(s) <= (e))) \
2b479609 865 ? _is_utf8_char_helper(s, s + UTF8SKIP(s), 0) : 0)
7131f24d 866#else
a14e0a36 867# define UTF8_IS_SUPER(s, e) \
89d986df
KW
868 (( LIKELY((e) > (s) + 3) \
869 && (*(U8*) (s)) >= 0xF4 \
870 && ((*(U8*) (s)) > 0xF4 || (*((U8*) (s) + 1) >= 0x90))\
871 && LIKELY((s) + UTF8SKIP(s) <= (e))) \
2b479609 872 ? _is_utf8_char_helper(s, s + UTF8SKIP(s), 0) : 0)
7131f24d
KW
873#endif
874
b96a92fb
KW
875/* These are now machine generated, and the 'given' clause is no longer
876 * applicable */
0c58a72b 877#define UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s, e) \
89d986df
KW
878 cBOOL(is_NONCHAR_utf8_safe(s,e))
879
880/*
881=for apidoc Am|bool|UTF8_IS_NONCHAR|const U8 *s|const U8 *e
882
883Evaluates to non-zero if the first few bytes of the string starting at C<s> and
884looking no further than S<C<e - 1>> are well-formed UTF-8 that represents one
885of the Unicode non-character code points; otherwise it evaluates to 0. If
886non-zero, the value gives how many bytes starting at C<s> comprise the code
887point's representation.
888
889=cut
890 */
0c58a72b
KW
891#define UTF8_IS_NONCHAR(s, e) \
892 UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s, e)
7131f24d 893
c867b360
JH
894#define UNICODE_SURROGATE_FIRST 0xD800
895#define UNICODE_SURROGATE_LAST 0xDFFF
896#define UNICODE_REPLACEMENT 0xFFFD
897#define UNICODE_BYTE_ORDER_MARK 0xFEFF
1d72bdf6 898
b851fbc1 899/* Though our UTF-8 encoding can go beyond this,
c76687c5 900 * let's be conservative and do as Unicode says. */
b851fbc1
JH
901#define PERL_UNICODE_MAX 0x10FFFF
902
d044b7a7
KW
903#define UNICODE_WARN_SURROGATE 0x0001 /* UTF-16 surrogates */
904#define UNICODE_WARN_NONCHAR 0x0002 /* Non-char code points */
905#define UNICODE_WARN_SUPER 0x0004 /* Above 0x10FFFF */
906#define UNICODE_WARN_PERL_EXTENDED 0x0008 /* Above 0x7FFF_FFFF */
907#define UNICODE_WARN_ABOVE_31_BIT UNICODE_WARN_PERL_EXTENDED
908#define UNICODE_DISALLOW_SURROGATE 0x0010
909#define UNICODE_DISALLOW_NONCHAR 0x0020
910#define UNICODE_DISALLOW_SUPER 0x0040
911#define UNICODE_DISALLOW_PERL_EXTENDED 0x0080
912#define UNICODE_DISALLOW_ABOVE_31_BIT UNICODE_DISALLOW_PERL_EXTENDED
33f38593
KW
913
914#define UNICODE_GOT_SURROGATE UNICODE_DISALLOW_SURROGATE
915#define UNICODE_GOT_NONCHAR UNICODE_DISALLOW_NONCHAR
916#define UNICODE_GOT_SUPER UNICODE_DISALLOW_SUPER
917#define UNICODE_GOT_PERL_EXTENDED UNICODE_DISALLOW_PERL_EXTENDED
918
ecc1615f
KW
919#define UNICODE_WARN_ILLEGAL_C9_INTERCHANGE \
920 (UNICODE_WARN_SURROGATE|UNICODE_WARN_SUPER)
bb88be5f 921#define UNICODE_WARN_ILLEGAL_INTERCHANGE \
ecc1615f
KW
922 (UNICODE_WARN_ILLEGAL_C9_INTERCHANGE|UNICODE_WARN_NONCHAR)
923#define UNICODE_DISALLOW_ILLEGAL_C9_INTERCHANGE \
924 (UNICODE_DISALLOW_SURROGATE|UNICODE_DISALLOW_SUPER)
bb88be5f 925#define UNICODE_DISALLOW_ILLEGAL_INTERCHANGE \
ecc1615f 926 (UNICODE_DISALLOW_ILLEGAL_C9_INTERCHANGE|UNICODE_DISALLOW_NONCHAR)
949cf498
KW
927
928/* For backward source compatibility, as are now the default */
929#define UNICODE_ALLOW_SURROGATE 0
930#define UNICODE_ALLOW_SUPER 0
931#define UNICODE_ALLOW_ANY 0
b851fbc1 932
2d6b3d38
KW
933/* This matches the 2048 code points between UNICODE_SURROGATE_FIRST (0xD800) and
934 * UNICODE_SURROGATE_LAST (0xDFFF) */
935#define UNICODE_IS_SURROGATE(uv) (((UV) (uv) & (~0xFFFF | 0xF800)) \
936 == 0xD800)
937
646d1759
KW
938#define UNICODE_IS_REPLACEMENT(uv) ((UV) (uv) == UNICODE_REPLACEMENT)
939#define UNICODE_IS_BYTE_ORDER_MARK(uv) ((UV) (uv) == UNICODE_BYTE_ORDER_MARK)
c149ab20
KW
940
941/* Is 'uv' one of the 32 contiguous-range noncharacters? */
942#define UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv) ((UV) (uv) >= 0xFDD0 \
943 && (UV) (uv) <= 0xFDEF)
944
945/* Is 'uv' one of the 34 plane-ending noncharacters 0xFFFE, 0xFFFF, 0x1FFFE,
946 * 0x1FFFF, ... 0x10FFFE, 0x10FFFF, given that we know that 'uv' is not above
947 * the Unicode legal max */
948#define UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv) \
949 (((UV) (uv) & 0xFFFE) == 0xFFFE)
950
951#define UNICODE_IS_NONCHAR(uv) \
952 ( UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv) \
953 || ( LIKELY( ! UNICODE_IS_SUPER(uv)) \
954 && UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv)))
955
956#define UNICODE_IS_SUPER(uv) ((UV) (uv) > PERL_UNICODE_MAX)
1d72bdf6 957
ec34087a
KW
958#define LATIN_SMALL_LETTER_SHARP_S LATIN_SMALL_LETTER_SHARP_S_NATIVE
959#define LATIN_SMALL_LETTER_Y_WITH_DIAERESIS \
960 LATIN_SMALL_LETTER_Y_WITH_DIAERESIS_NATIVE
961#define MICRO_SIGN MICRO_SIGN_NATIVE
962#define LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE \
963 LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE_NATIVE
964#define LATIN_SMALL_LETTER_A_WITH_RING_ABOVE \
965 LATIN_SMALL_LETTER_A_WITH_RING_ABOVE_NATIVE
09091399
JH
966#define UNICODE_GREEK_CAPITAL_LETTER_SIGMA 0x03A3
967#define UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA 0x03C2
968#define UNICODE_GREEK_SMALL_LETTER_SIGMA 0x03C3
9dcbe121 969#define GREEK_SMALL_LETTER_MU 0x03BC
9e682c18
KW
970#define GREEK_CAPITAL_LETTER_MU 0x039C /* Upper and title case
971 of MICRON */
972#define LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS 0x0178 /* Also is title case */
0766489e
KW
973#ifdef LATIN_CAPITAL_LETTER_SHARP_S_UTF8
974# define LATIN_CAPITAL_LETTER_SHARP_S 0x1E9E
975#endif
74894415
KW
976#define LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE 0x130
977#define LATIN_SMALL_LETTER_DOTLESS_I 0x131
9e682c18 978#define LATIN_SMALL_LETTER_LONG_S 0x017F
a9f50d33
KW
979#define LATIN_SMALL_LIGATURE_LONG_S_T 0xFB05
980#define LATIN_SMALL_LIGATURE_ST 0xFB06
9e682c18
KW
981#define KELVIN_SIGN 0x212A
982#define ANGSTROM_SIGN 0x212B
09091399 983
9e55ce06 984#define UNI_DISPLAY_ISPRINT 0x0001
c728cb41
JH
985#define UNI_DISPLAY_BACKSLASH 0x0002
986#define UNI_DISPLAY_QQ (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH)
987#define UNI_DISPLAY_REGEX (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH)
9e55ce06 988
5cd46e1f
KW
989#define ANYOF_FOLD_SHARP_S(node, input, end) \
990 (ANYOF_BITMAP_TEST(node, LATIN_SMALL_LETTER_SHARP_S) && \
137165a6 991 (ANYOF_NONBITMAP(node)) && \
39065660 992 (ANYOF_FLAGS(node) & ANYOF_LOC_NONBITMAP_FOLD) && \
07b6858f 993 ((end) > (input) + 1) && \
305b8651 994 isALPHA_FOLD_EQ((input)[0], 's'))
6302f837 995
ebc501f0 996#define SHARP_S_SKIP 2
3b0fc154 997
3cedd9d9 998#define is_utf8_char_buf(buf, buf_end) isUTF8_CHAR(buf, buf_end)
976c1b08
KW
999#define bytes_from_utf8(s, lenp, is_utf8p) \
1000 bytes_from_utf8_loc(s, lenp, is_utf8p, 0)
3cedd9d9 1001
e23e8bc1
KW
1002/*
1003
1004=for apidoc Am|STRLEN|isSTRICT_UTF8_CHAR|const U8 *s|const U8 *e
1005
1006Evaluates to non-zero if the first few bytes of the string starting at C<s> and
1007looking no further than S<C<e - 1>> are well-formed UTF-8 that represents some
1008Unicode code point completely acceptable for open interchange between all
1009applications; otherwise it evaluates to 0. If non-zero, the value gives how
2717076a
KW
1010many bytes starting at C<s> comprise the code point's representation. Any
1011bytes remaining before C<e>, but beyond the ones needed to form the first code
1012point in C<s>, are not examined.
e23e8bc1
KW
1013
1014The largest acceptable code point is the Unicode maximum 0x10FFFF, and must not
1015be a surrogate nor a non-character code point. Thus this excludes any code
1016point from Perl's extended UTF-8.
1017
1018This is used to efficiently decide if the next few bytes in C<s> is
2717076a
KW
1019legal Unicode-acceptable UTF-8 for a single character.
1020
1021Use C<L</isC9_STRICT_UTF8_CHAR>> to use the L<Unicode Corrigendum
1022#9|http://www.unicode.org/versions/corrigendum9.html> definition of allowable
1023code points; C<L</isUTF8_CHAR>> to check for Perl's extended UTF-8;
1024and C<L</isUTF8_CHAR_flags>> for a more customized definition.
e23e8bc1 1025
9f2abfde
KW
1026Use C<L</is_strict_utf8_string>>, C<L</is_strict_utf8_string_loc>>, and
1027C<L</is_strict_utf8_string_loclen>> to check entire strings.
1028
e23e8bc1
KW
1029=cut
1030*/
1031
1032#define isSTRICT_UTF8_CHAR(s, e) \
1033 (UNLIKELY((e) <= (s)) \
1034 ? 0 \
1035 : (UTF8_IS_INVARIANT(*s)) \
1036 ? 1 \
1037 : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \
1038 ? 0 \
1039 : is_STRICT_UTF8_CHAR_utf8_no_length_checks(s))
1040
a82be82b
KW
1041/*
1042
1043=for apidoc Am|STRLEN|isC9_STRICT_UTF8_CHAR|const U8 *s|const U8 *e
1044
1045Evaluates to non-zero if the first few bytes of the string starting at C<s> and
1046looking no further than S<C<e - 1>> are well-formed UTF-8 that represents some
1047Unicode non-surrogate code point; otherwise it evaluates to 0. If non-zero,
1048the value gives how many bytes starting at C<s> comprise the code point's
2717076a
KW
1049representation. Any bytes remaining before C<e>, but beyond the ones needed to
1050form the first code point in C<s>, are not examined.
a82be82b
KW
1051
1052The largest acceptable code point is the Unicode maximum 0x10FFFF. This
1053differs from C<L</isSTRICT_UTF8_CHAR>> only in that it accepts non-character
1054code points. This corresponds to
1055L<Unicode Corrigendum #9|http://www.unicode.org/versions/corrigendum9.html>.
1056which said that non-character code points are merely discouraged rather than
1057completely forbidden in open interchange. See
1058L<perlunicode/Noncharacter code points>.
1059
2717076a
KW
1060Use C<L</isUTF8_CHAR>> to check for Perl's extended UTF-8; and
1061C<L</isUTF8_CHAR_flags>> for a more customized definition.
1062
9f2abfde
KW
1063Use C<L</is_c9strict_utf8_string>>, C<L</is_c9strict_utf8_string_loc>>, and
1064C<L</is_c9strict_utf8_string_loclen>> to check entire strings.
1065
a82be82b
KW
1066=cut
1067*/
1068
1069#define isC9_STRICT_UTF8_CHAR(s, e) \
1070 (UNLIKELY((e) <= (s)) \
1071 ? 0 \
1072 : (UTF8_IS_INVARIANT(*s)) \
1073 ? 1 \
1074 : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \
1075 ? 0 \
1076 : is_C9_STRICT_UTF8_CHAR_utf8_no_length_checks(s))
1077
25e3a4e0
KW
1078/*
1079
1080=for apidoc Am|STRLEN|isUTF8_CHAR_flags|const U8 *s|const U8 *e| const U32 flags
1081
1082Evaluates to non-zero if the first few bytes of the string starting at C<s> and
1083looking no further than S<C<e - 1>> are well-formed UTF-8, as extended by Perl,
1084that represents some code point, subject to the restrictions given by C<flags>;
1085otherwise it evaluates to 0. If non-zero, the value gives how many bytes
2717076a
KW
1086starting at C<s> comprise the code point's representation. Any bytes remaining
1087before C<e>, but beyond the ones needed to form the first code point in C<s>,
1088are not examined.
25e3a4e0
KW
1089
1090If C<flags> is 0, this gives the same results as C<L</isUTF8_CHAR>>;
1091if C<flags> is C<UTF8_DISALLOW_ILLEGAL_INTERCHANGE>, this gives the same results
1092as C<L</isSTRICT_UTF8_CHAR>>;
1093and if C<flags> is C<UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE>, this gives
1094the same results as C<L</isC9_STRICT_UTF8_CHAR>>.
1095Otherwise C<flags> may be any combination of the C<UTF8_DISALLOW_I<foo>> flags
1096understood by C<L</utf8n_to_uvchr>>, with the same meanings.
1097
1098The three alternative macros are for the most commonly needed validations; they
1099are likely to run somewhat faster than this more general one, as they can be
1100inlined into your code.
1101
9f2abfde
KW
1102Use L</is_utf8_string_flags>, L</is_utf8_string_loc_flags>, and
1103L</is_utf8_string_loclen_flags> to check entire strings.
1104
25e3a4e0
KW
1105=cut
1106*/
1107
1108#define isUTF8_CHAR_flags(s, e, flags) \
1109 (UNLIKELY((e) <= (s)) \
1110 ? 0 \
1111 : (UTF8_IS_INVARIANT(*s)) \
1112 ? 1 \
1113 : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \
1114 ? 0 \
1115 : _is_utf8_char_helper(s, e, flags))
1116
6302f837
KW
1117/* Do not use; should be deprecated. Use isUTF8_CHAR() instead; this is
1118 * retained solely for backwards compatibility */
1119#define IS_UTF8_CHAR(p, n) (isUTF8_CHAR(p, (p) + (n)) == n)
e9a8c099 1120
6a5bc5ac 1121#endif /* PERL_UTF8_H_ */
57f0e7e2 1122
e9a8c099 1123/*
14d04a33 1124 * ex: set ts=8 sts=4 sw=4 et:
e9a8c099 1125 */