This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
APItest:t/utf8_warn_base.pl: Clarify some comments
[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
de69f3af
KW
69#define uvchr_to_utf8(a,b) uvchr_to_utf8_flags(a,b,0)
70#define uvchr_to_utf8_flags(d,uv,flags) \
71 uvoffuni_to_utf8_flags(d,NATIVE_TO_UNI(uv),flags)
72#define utf8_to_uvchr_buf(s, e, lenp) \
842991ae 73 utf8n_to_uvchr(s, (U8*)(e) - (U8*)(s), lenp, \
de69f3af 74 ckWARN_d(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY)
f9380377
KW
75#define utf8n_to_uvchr(s, len, lenp, flags) \
76 utf8n_to_uvchr_error(s, len, lenp, flags, 0)
37657a5b
KW
77#define utf8n_to_uvchr_error(s, len, lenp, flags, errors) \
78 utf8n_to_uvchr_msgs(s, len, lenp, flags, errors, 0)
de69f3af 79
a0270393 80#define to_uni_fold(c, p, lenp) _to_uni_fold_flags(c, p, lenp, FOLD_FLAGS_FULL)
a239b1e2
KW
81
82#define to_utf8_fold(s, r, lenr) \
607313a1 83 _to_utf8_fold_flags (s, NULL, r, lenr, FOLD_FLAGS_FULL, __FILE__, __LINE__)
a239b1e2 84#define to_utf8_lower(s, r, lenr) \
607313a1 85 _to_utf8_lower_flags(s, NULL, r ,lenr, 0, __FILE__, __LINE__)
a239b1e2 86#define to_utf8_upper(s, r, lenr) \
607313a1 87 _to_utf8_upper_flags(s, NULL, r, lenr, 0, __FILE__, __LINE__)
a239b1e2 88#define to_utf8_title(s, r, lenr) \
607313a1 89 _to_utf8_title_flags(s, NULL, r, lenr ,0, __FILE__, __LINE__)
36bb2ab6 90
eda9cac1
KW
91#define foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \
92 foldEQ_utf8_flags(s1, pe1, l1, u1, s2, pe2, l2, u2, 0)
baa60164 93#define FOLDEQ_UTF8_NOMIX_ASCII (1 << 0)
cea315b6 94#define FOLDEQ_LOCALE (1 << 1)
18f762c3
KW
95#define FOLDEQ_S1_ALREADY_FOLDED (1 << 2)
96#define FOLDEQ_S2_ALREADY_FOLDED (1 << 3)
d635b710
KW
97#define FOLDEQ_S1_FOLDS_SANE (1 << 4)
98#define FOLDEQ_S2_FOLDS_SANE (1 << 5)
a33c29bc 99
e6226b18
KW
100#define ibcmp_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \
101 cBOOL(! foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2))
102
1d72bdf6
NIS
103#ifdef EBCDIC
104/* The equivalent of these macros but implementing UTF-EBCDIC
105 are in the following header file:
106 */
107
108#include "utfebcdic.h"
fd7cb289 109
d06134e5 110#else /* ! EBCDIC */
73c4f7a1
GS
111START_EXTERN_C
112
111e8ed9
KW
113/* How wide can a single UTF-8 encoded character become in bytes. */
114/* NOTE: Strictly speaking Perl's UTF-8 should not be called UTF-8 since UTF-8
115 * is an encoding of Unicode, and Unicode's upper limit, 0x10FFFF, can be
116 * expressed with 4 bytes. However, Perl thinks of UTF-8 as a way to encode
117 * non-negative integers in a binary format, even those above Unicode */
118#define UTF8_MAXBYTES 13
119
a0ed51b3 120#ifdef DOINIT
6f06b55f 121EXTCONST unsigned char PL_utf8skip[] = {
b2635aa8
KW
122/* 0x00 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
123/* 0x10 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
124/* 0x20 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
125/* 0x30 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
126/* 0x40 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
127/* 0x50 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
128/* 0x60 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
129/* 0x70 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
130/* 0x80 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
131/* 0x90 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
132/* 0xA0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
133/* 0xB0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
134/* 0xC0 */ 2,2, /* overlong */
1ff3baa2 135/* 0xC2 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0080 to U+03FF */
b2635aa8
KW
136/* 0xD0 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0400 to U+07FF */
137/* 0xE0 */ 3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3, /* U+0800 to U+FFFF */
138/* 0xF0 */ 4,4,4,4,4,4,4,4,5,5,5,5,6,6, /* above BMP to 2**31 - 1 */
6937f885
KW
139 /* Perl extended (never was official UTF-8). Up to 36 bit */
140/* 0xFE */ 7,
141 /* More extended, Up to 72 bits (64-bit + reserved) */
111e8ed9 142/* 0xFF */ UTF8_MAXBYTES
a0ed51b3
LW
143};
144#else
6f06b55f 145EXTCONST unsigned char PL_utf8skip[];
a0ed51b3
LW
146#endif
147
73c4f7a1 148END_EXTERN_C
7e2040f0 149
1a3756de 150#if defined(_MSC_VER) && _MSC_VER < 1400
6f6d1bab
TC
151/* older MSVC versions have a smallish macro buffer */
152#define PERL_SMALL_MACRO_BUFFER
153#endif
154
59a449d5
KW
155/* Native character to/from iso-8859-1. Are the identity functions on ASCII
156 * platforms */
6f6d1bab 157#ifdef PERL_SMALL_MACRO_BUFFER
1a3756de
TC
158#define NATIVE_TO_LATIN1(ch) ((U8)(ch))
159#define LATIN1_TO_NATIVE(ch) ((U8)(ch))
6f6d1bab 160#else
4c8cd605
KW
161#define NATIVE_TO_LATIN1(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch)))
162#define LATIN1_TO_NATIVE(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch)))
6f6d1bab 163#endif
59a449d5
KW
164
165/* I8 is an intermediate version of UTF-8 used only in UTF-EBCDIC. We thus
166 * consider it to be identical to UTF-8 on ASCII platforms. Strictly speaking
167 * UTF-8 and UTF-EBCDIC are two different things, but we often conflate them
168 * because they are 8-bit encodings that serve the same purpose in Perl, and
169 * rarely do we need to distinguish them. The term "NATIVE_UTF8" applies to
170 * whichever one is applicable on the current platform */
6f6d1bab 171#ifdef PERL_SMALL_MACRO_BUFFER
1d4ea287
DD
172#define NATIVE_UTF8_TO_I8(ch) ((U8) (ch))
173#define I8_TO_NATIVE_UTF8(ch) ((U8) (ch))
6f6d1bab 174#else
4c8cd605
KW
175#define NATIVE_UTF8_TO_I8(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch)))
176#define I8_TO_NATIVE_UTF8(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch)))
6f6d1bab 177#endif
59a449d5 178
1d72bdf6 179/* Transforms in wide UV chars */
4c8cd605
KW
180#define UNI_TO_NATIVE(ch) ((UV) (ch))
181#define NATIVE_TO_UNI(ch) ((UV) (ch))
d7578b48 182
877d9f0d 183/*
9041c2e3 184
a14e0a36
KW
185 The following table is from Unicode 3.2, plus the Perl extensions for above
186 U+10FFFF
877d9f0d 187
a14e0a36 188 Code Points 1st Byte 2nd Byte 3rd 4th 5th 6th 7th 8th-13th
877d9f0d 189
375122d7 190 U+0000..U+007F 00..7F
e1b711da 191 U+0080..U+07FF * C2..DF 80..BF
a14e0a36
KW
192 U+0800..U+0FFF E0 * A0..BF 80..BF
193 U+1000..U+CFFF E1..EC 80..BF 80..BF
194 U+D000..U+D7FF ED 80..9F 80..BF
195 U+D800..U+DFFF ED A0..BF 80..BF (surrogates)
196 U+E000..U+FFFF EE..EF 80..BF 80..BF
197 U+10000..U+3FFFF F0 * 90..BF 80..BF 80..BF
198 U+40000..U+FFFFF F1..F3 80..BF 80..BF 80..BF
199 U+100000..U+10FFFF F4 80..8F 80..BF 80..BF
200 Below are above-Unicode code points
201 U+110000..U+13FFFF F4 90..BF 80..BF 80..BF
202 U+110000..U+1FFFFF F5..F7 80..BF 80..BF 80..BF
203 U+200000..U+FFFFFF F8 * 88..BF 80..BF 80..BF 80..BF
204U+1000000..U+3FFFFFF F9..FB 80..BF 80..BF 80..BF 80..BF
205U+4000000..U+3FFFFFFF FC * 84..BF 80..BF 80..BF 80..BF 80..BF
206U+40000000..U+7FFFFFFF FD 80..BF 80..BF 80..BF 80..BF 80..BF
207U+80000000..U+FFFFFFFFF FE * 82..BF 80..BF 80..BF 80..BF 80..BF 80..BF
208U+1000000000.. FF 80..BF 80..BF 80..BF 80..BF 80..BF * 81..BF 80..BF
877d9f0d 209
e1b711da 210Note the gaps before several of the byte entries above marked by '*'. These are
37e2e78e
KW
211caused by legal UTF-8 avoiding non-shortest encodings: it is technically
212possible to UTF-8-encode a single code point in different ways, but that is
213explicitly forbidden, and the shortest possible encoding should always be used
15824458 214(and that is what Perl does). The non-shortest ones are called 'overlongs'.
8c007b5a 215
877d9f0d
JH
216 */
217
8c007b5a
JH
218/*
219 Another way to look at it, as bits:
220
b2635aa8 221 Code Points 1st Byte 2nd Byte 3rd Byte 4th Byte
8c007b5a 222
b2635aa8
KW
223 0aaa aaaa 0aaa aaaa
224 0000 0bbb bbaa aaaa 110b bbbb 10aa aaaa
225 cccc bbbb bbaa aaaa 1110 cccc 10bb bbbb 10aa aaaa
226 00 000d ddcc cccc bbbb bbaa aaaa 1111 0ddd 10cc cccc 10bb bbbb 10aa aaaa
8c007b5a
JH
227
228As you can see, the continuation bytes all begin with C<10>, and the
e1b711da 229leading bits of the start byte tell how many bytes there are in the
8c007b5a
JH
230encoded character.
231
df863e43
KW
232Perl's extended UTF-8 means we can have start bytes up through FF, though any
233beginning with FF yields a code point that is too large for 32-bit ASCII
234platforms. FF signals to use 13 bytes for the encoded character. This breaks
235the paradigm that the number of leading bits gives how many total bytes there
236are in the character.
65ab9279 237
8c007b5a
JH
238*/
239
6c88483e 240/* Is the representation of the Unicode code point 'cp' the same regardless of
15824458 241 * being encoded in UTF-8 or not? */
2d1545e5 242#define OFFUNI_IS_INVARIANT(cp) isASCII(cp)
15824458 243
c2b32798
KW
244/*
245=for apidoc Am|bool|UVCHR_IS_INVARIANT|UV cp
246
247Evaluates to 1 if the representation of code point C<cp> is the same whether or
248not it is encoded in UTF-8; otherwise evaluates to 0. UTF-8 invariant
249characters can be copied as-is when converting to/from UTF-8, saving time.
250C<cp> is Unicode if above 255; otherwise is platform-native.
251
252=cut
253 */
254
cf1be84e 255#define UVCHR_IS_INVARIANT(cp) OFFUNI_IS_INVARIANT(cp)
38953e5a 256
c9264833
KW
257/* This defines the bits that are to be in the continuation bytes of a multi-byte
258 * UTF-8 encoded character that mark it is a continuation byte. */
259#define UTF_CONTINUATION_MARK 0x80
260
a95ec4fb 261/* Misleadingly named: is the UTF8-encoded byte 'c' part of a variant sequence
5c06326b
KW
262 * in UTF-8? This is the inverse of UTF8_IS_INVARIANT. The |0 makes sure this
263 * isn't mistakenly called with a ptr argument */
a6951642
KW
264#define UTF8_IS_CONTINUED(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
265 ((U8)((c) | 0)) & UTF_CONTINUATION_MARK)
15824458
KW
266
267/* Is the byte 'c' the first byte of a multi-byte UTF8-8 encoded sequence?
268 * This doesn't catch invariants (they are single-byte). It also excludes the
5c06326b
KW
269 * illegal overlong sequences that begin with C0 and C1. The |0 makes sure
270 * this isn't mistakenly called with a ptr argument */
a6951642
KW
271#define UTF8_IS_START(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
272 ((U8)((c) | 0)) >= 0xc2)
15824458 273
858cd8ab
KW
274/* For use in UTF8_IS_CONTINUATION() below */
275#define UTF_IS_CONTINUATION_MASK 0xC0
276
15824458 277/* Is the byte 'c' part of a multi-byte UTF8-8 encoded sequence, and not the
5c06326b
KW
278 * first byte thereof? The |0 makes sure this isn't mistakenly called with a
279 * ptr argument */
a6951642
KW
280#define UTF8_IS_CONTINUATION(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
281 (((U8)((c) | 0)) & UTF_IS_CONTINUATION_MASK) == UTF_CONTINUATION_MARK)
0ae1fa71 282
15824458
KW
283/* Is the UTF8-encoded byte 'c' the first byte of a two byte sequence? Use
284 * UTF8_IS_NEXT_CHAR_DOWNGRADEABLE() instead if the input isn't known to
285 * be well-formed. Masking with 0xfe allows the low bit to be 0 or 1; thus
5c06326b
KW
286 * this matches 0xc[23]. The |0 makes sure this isn't mistakenly called with a
287 * ptr argument */
a6951642
KW
288#define UTF8_IS_DOWNGRADEABLE_START(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
289 (((U8)((c) | 0)) & 0xfe) == 0xc2)
4ab10950 290
15824458 291/* Is the UTF8-encoded byte 'c' the first byte of a sequence of bytes that
5c06326b
KW
292 * represent a code point > 255? The |0 makes sure this isn't mistakenly
293 * called with a ptr argument */
a6951642
KW
294#define UTF8_IS_ABOVE_LATIN1(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
295 ((U8)((c) | 0)) >= 0xc4)
8850bf83 296
15824458
KW
297/* This is the number of low-order bits a continuation byte in a UTF-8 encoded
298 * sequence contributes to the specification of the code point. In the bit
299 * maps above, you see that the first 2 bits are a constant '10', leaving 6 of
300 * real information */
1d72bdf6 301#define UTF_ACCUMULATION_SHIFT 6
b2635aa8 302
fed423a5
KW
303/* ^? is defined to be DEL on ASCII systems. See the definition of toCTRL()
304 * for more */
305#define QUESTION_MARK_CTRL DEL_NATIVE
306
307/* Surrogates, non-character code points and above-Unicode code points are
308 * problematic in some contexts. This allows code that needs to check for
309 * those to to quickly exclude the vast majority of code points it will
310 * encounter */
a6951642
KW
311#define isUTF8_POSSIBLY_PROBLEMATIC(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
312 (U8) c >= 0xED)
fed423a5 313
784d4f31
KW
314/* A helper macro for isUTF8_CHAR, so use that one instead of this. This was
315 * generated by regen/regcharclass.pl, and then moved here. Then it was
316 * hand-edited to add some LIKELY() calls, presuming that malformations are
317 * unlikely. The lines that generated it were then commented out. This was
318 * done because it takes on the order of 10 minutes to generate, and is never
319 * going to change, unless the generated code is improved, and figuring out
320 * the LIKELYs there would be hard.
321 *
322 UTF8_CHAR: Matches legal UTF-8 variant code points up through 0x1FFFFFF
323
324 0x80 - 0x1FFFFF
325*/
326/*** GENERATED CODE ***/
327#define is_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] <= 0xEF ) ? \
333 ( LIKELY( ( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
334: ( 0xF0 == ((const U8*)s)[0] ) ? \
335 ( 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 )\
336: ( ( ( ( 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
337
338/* The above macro handles UTF-8 that has this start byte as the maximum */
339#define _IS_UTF8_CHAR_HIGHEST_START_BYTE 0xF7
340
e23e8bc1
KW
341/* A helper macro for isSTRICT_UTF8_CHAR, so use that one instead of this.
342 * Like is_UTF8_CHAR_utf8_no_length_checks(), this was moved here and LIKELYs
343 * added manually.
344 *
345 STRICT_UTF8_CHAR: Matches legal Unicode UTF-8 variant code points, no
346 surrrogates nor non-character code points
347*/
348/*** GENERATED CODE ***/
349#define is_STRICT_UTF8_CHAR_utf8_no_length_checks(s) \
9f2eed98
P
350( ( 0xC2 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xDF ) ? \
351 ( LIKELY( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) ? 2 : 0 ) \
352: ( 0xE0 == ((const U8*)s)[0] ) ? \
353 ( LIKELY( ( ( ((const U8*)s)[1] & 0xE0 ) == 0xA0 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
354: ( ( 0xE1 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xEC ) || 0xEE == ((const U8*)s)[0] ) ?\
355 ( ( ( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
356: ( 0xED == ((const U8*)s)[0] ) ? \
357 ( LIKELY( ( ( ((const U8*)s)[1] & 0xE0 ) == 0x80 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
358: ( 0xEF == ((const U8*)s)[0] ) ? \
359 ( ( ( 0x80 <= ((const U8*)s)[1] && ((const U8*)s)[1] <= 0xB6 ) || ( 0xB8 <= ((const U8*)s)[1] && ((const U8*)s)[1] <= 0xBE ) ) ?\
360 ( LIKELY( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ? 3 : 0 ) \
361 : ( 0xB7 == ((const U8*)s)[1] ) ? \
362 ( LIKELY( ( ((const U8*)s)[2] & 0xF0 ) == 0x80 || ( ((const U8*)s)[2] & 0xF0 ) == 0xB0 ) ? 3 : 0 )\
363 : ( ( 0xBF == ((const U8*)s)[1] ) && ( 0x80 <= ((const U8*)s)[2] && ((const U8*)s)[2] <= 0xBD ) ) ? 3 : 0 )\
364: ( 0xF0 == ((const U8*)s)[0] ) ? \
365 ( ( ( 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 ) ) ?\
366 ( LIKELY( ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\
367 : ( ((const U8*)s)[1] == 0x9F || ( ( ((const U8*)s)[1] & 0xEF ) == 0xAF ) ) ? \
368 ( ( 0x80 <= ((const U8*)s)[2] && ((const U8*)s)[2] <= 0xBE ) ? \
369 ( LIKELY( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ? 4 : 0 ) \
370 : LIKELY( ( 0xBF == ((const U8*)s)[2] ) && ( 0x80 <= ((const U8*)s)[3] && ((const U8*)s)[3] <= 0xBD ) ) ? 4 : 0 )\
e23e8bc1 371 : 0 ) \
9f2eed98
P
372: ( 0xF1 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xF3 ) ? \
373 ( ( ( ( ((const U8*)s)[1] & 0xC8 ) == 0x80 ) || ( ( ((const U8*)s)[1] & 0xCC ) == 0x88 ) || ( ( ((const U8*)s)[1] & 0xCE ) == 0x8C ) || ( ( ((const U8*)s)[1] & 0xCF ) == 0x8E ) ) ?\
374 ( LIKELY( ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\
375 : ( ( ((const U8*)s)[1] & 0xCF ) == 0x8F ) ? \
376 ( ( 0x80 <= ((const U8*)s)[2] && ((const U8*)s)[2] <= 0xBE ) ? \
377 ( LIKELY( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ? 4 : 0 ) \
378 : LIKELY( ( 0xBF == ((const U8*)s)[2] ) && ( 0x80 <= ((const U8*)s)[3] && ((const U8*)s)[3] <= 0xBD ) ) ? 4 : 0 )\
e23e8bc1 379 : 0 ) \
9f2eed98
P
380: ( 0xF4 == ((const U8*)s)[0] ) ? \
381 ( ( 0x80 <= ((const U8*)s)[1] && ((const U8*)s)[1] <= 0x8E ) ? \
382 ( LIKELY( ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\
383 : ( 0x8F == ((const U8*)s)[1] ) ? \
384 ( ( 0x80 <= ((const U8*)s)[2] && ((const U8*)s)[2] <= 0xBE ) ? \
385 ( LIKELY( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ? 4 : 0 ) \
386 : LIKELY( ( 0xBF == ((const U8*)s)[2] ) && ( 0x80 <= ((const U8*)s)[3] && ((const U8*)s)[3] <= 0xBD ) ) ? 4 : 0 )\
e23e8bc1
KW
387 : 0 ) \
388: 0 )
389
a82be82b
KW
390/* Similarly,
391 C9_STRICT_UTF8_CHAR: Matches legal Unicode UTF-8 variant code
392 points, no surrogates
393 0x0080 - 0xD7FF
394 0xE000 - 0x10FFFF
395*/
396/*** GENERATED CODE ***/
397#define is_C9_STRICT_UTF8_CHAR_utf8_no_length_checks(s) \
9f2eed98
P
398( ( 0xC2 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xDF ) ? \
399 ( LIKELY( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) ? 2 : 0 ) \
400: ( 0xE0 == ((const U8*)s)[0] ) ? \
401 ( LIKELY( ( ( ((const U8*)s)[1] & 0xE0 ) == 0xA0 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
402: ( ( 0xE1 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xEC ) || ( ((const U8*)s)[0] & 0xFE ) == 0xEE ) ?\
403 ( LIKELY( ( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
404: ( 0xED == ((const U8*)s)[0] ) ? \
405 ( LIKELY( ( ( ((const U8*)s)[1] & 0xE0 ) == 0x80 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
406: ( 0xF0 == ((const U8*)s)[0] ) ? \
407 ( 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 )\
408: ( 0xF1 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xF3 ) ? \
409 ( LIKELY( ( ( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) && ( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\
410: 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 411
57ff5f59
KW
412#define UNICODE_IS_PERL_EXTENDED(uv) UNLIKELY((UV) (uv) > 0x7FFFFFFF)
413
fed423a5
KW
414#endif /* EBCDIC vs ASCII */
415
416/* 2**UTF_ACCUMULATION_SHIFT - 1 */
417#define UTF_CONTINUATION_MASK ((U8) ((1U << UTF_ACCUMULATION_SHIFT) - 1))
418
7028aeba
KW
419/* Internal macro to be used only in this file to aid in constructing other
420 * publicly accessible macros.
421 * The number of bytes required to express this uv in UTF-8, for just those
422 * uv's requiring 2 through 6 bytes, as these are common to all platforms and
423 * word sizes. The number of bytes needed is given by the number of leading 1
424 * bits in the start byte. There are 32 start bytes that have 2 initial 1 bits
425 * (C0-DF); there are 16 that have 3 initial 1 bits (E0-EF); 8 that have 4
426 * initial 1 bits (F0-F8); 4 that have 5 initial 1 bits (F9-FB), and 2 that
427 * have 6 initial 1 bits (FC-FD). The largest number a string of n bytes can
428 * represent is (the number of possible start bytes for 'n')
429 * * (the number of possiblities for each start byte
430 * The latter in turn is
431 * 2 ** ( (how many continuation bytes there are)
432 * * (the number of bits of information each
433 * continuation byte holds))
434 *
435 * If we were on a platform where we could use a fast find first set bit
436 * instruction (or count leading zeros instruction) this could be replaced by
437 * using that to find the log2 of the uv, and divide that by the number of bits
438 * of information in each continuation byte, adjusting for large cases and how
439 * much information is in a start byte for that length */
72164d3a 440#define __COMMON_UNI_SKIP(uv) \
7028aeba
KW
441 (UV) (uv) < (32 * (1U << ( UTF_ACCUMULATION_SHIFT))) ? 2 : \
442 (UV) (uv) < (16 * (1U << (2 * UTF_ACCUMULATION_SHIFT))) ? 3 : \
443 (UV) (uv) < ( 8 * (1U << (3 * UTF_ACCUMULATION_SHIFT))) ? 4 : \
444 (UV) (uv) < ( 4 * (1U << (4 * UTF_ACCUMULATION_SHIFT))) ? 5 : \
445 (UV) (uv) < ( 2 * (1U << (5 * UTF_ACCUMULATION_SHIFT))) ? 6 :
72164d3a
KW
446
447/* Internal macro to be used only in this file.
448 * This adds to __COMMON_UNI_SKIP the details at this platform's upper range.
fed423a5 449 * For any-sized EBCDIC platforms, or 64-bit ASCII ones, we need one more test
72164d3a
KW
450 * to see if just 7 bytes is needed, or if the maximum is needed. For 32-bit
451 * ASCII platforms, everything is representable by 7 bytes */
fed423a5 452#if defined(UV_IS_QUAD) || defined(EBCDIC)
72164d3a 453# define __BASE_UNI_SKIP(uv) (__COMMON_UNI_SKIP(uv) \
7028aeba 454 (UV) (uv) < ((UV) 1U << (6 * UTF_ACCUMULATION_SHIFT)) ? 7 : UTF8_MAXBYTES)
1d68d6cd 455#else
72164d3a 456# define __BASE_UNI_SKIP(uv) (__COMMON_UNI_SKIP(uv) 7)
1d68d6cd
SC
457#endif
458
fed423a5
KW
459/* The next two macros use the base macro defined above, and add in the tests
460 * at the low-end of the range, for just 1 byte, yielding complete macros,
461 * publicly accessible. */
462
463/* Input is a true Unicode (not-native) code point */
464#define OFFUNISKIP(uv) (OFFUNI_IS_INVARIANT(uv) ? 1 : __BASE_UNI_SKIP(uv))
2084b489 465
5352a763
KW
466/*
467
468=for apidoc Am|STRLEN|UVCHR_SKIP|UV cp
469returns the number of bytes required to represent the code point C<cp> when
470encoded as UTF-8. C<cp> is a native (ASCII or EBCDIC) code point if less than
471255; a Unicode code point otherwise.
472
473=cut
474 */
fdb6583d 475#define UVCHR_SKIP(uv) ( UVCHR_IS_INVARIANT(uv) ? 1 : __BASE_UNI_SKIP(uv))
5352a763 476
b651802e
KW
477/* The largest code point representable by two UTF-8 bytes on this platform.
478 * As explained in the comments for __COMMON_UNI_SKIP, 32 start bytes with
fed423a5 479 * UTF_ACCUMULATION_SHIFT bits of information each */
aa206fb7
KW
480#define MAX_UTF8_TWO_BYTE (32 * (1U << UTF_ACCUMULATION_SHIFT) - 1)
481
b651802e
KW
482/* The largest code point representable by two UTF-8 bytes on any platform that
483 * Perl runs on. This value is constrained by EBCDIC which has 5 bits per
484 * continuation byte */
aa206fb7
KW
485#define MAX_PORTABLE_UTF8_TWO_BYTE (32 * (1U << 5) - 1)
486
c03c0950
KW
487/* The maximum number of UTF-8 bytes a single Unicode character can
488 * uppercase/lowercase/fold into. Unicode guarantees that the maximum
489 * expansion is UTF8_MAX_FOLD_CHAR_EXPAND characters, but any above-Unicode
490 * code point will fold to itself, so we only have to look at the expansion of
491 * the maximum Unicode code point. But this number may be less than the space
492 * occupied by a very large code point under Perl's extended UTF-8. We have to
493 * make it large enough to fit any single character. (It turns out that ASCII
494 * and EBCDIC differ in which is larger) */
495#define UTF8_MAXBYTES_CASE \
496 (UTF8_MAXBYTES >= (UTF8_MAX_FOLD_CHAR_EXPAND * OFFUNISKIP(0x10FFFF)) \
497 ? UTF8_MAXBYTES \
498 : (UTF8_MAX_FOLD_CHAR_EXPAND * OFFUNISKIP(0x10FFFF)))
499
d06134e5
KW
500/* Rest of these are attributes of Unicode and perl's internals rather than the
501 * encoding, or happen to be the same in both ASCII and EBCDIC (at least at
502 * this level; the macros that some of these call may have different
503 * definitions in the two encodings */
504
59a449d5
KW
505/* In domain restricted to ASCII, these may make more sense to the reader than
506 * the ones with Latin1 in the name */
507#define NATIVE_TO_ASCII(ch) NATIVE_TO_LATIN1(ch)
508#define ASCII_TO_NATIVE(ch) LATIN1_TO_NATIVE(ch)
509
510/* More or less misleadingly-named defines, retained for back compat */
511#define NATIVE_TO_UTF(ch) NATIVE_UTF8_TO_I8(ch)
512#define NATIVE_TO_I8(ch) NATIVE_UTF8_TO_I8(ch)
513#define UTF_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch)
514#define I8_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch)
515#define NATIVE8_TO_UNI(ch) NATIVE_TO_LATIN1(ch)
d06134e5 516
c0236afe 517/* This defines the 1-bits that are to be in the first byte of a multi-byte
97d0ceda
KW
518 * UTF-8 encoded character that mark it as a start byte and give the number of
519 * bytes that comprise the character. 'len' is the number of bytes in the
520 * multi-byte sequence. */
c0236afe
KW
521#define UTF_START_MARK(len) (((len) > 7) ? 0xFF : (0xFF & (0xFE << (7-(len)))))
522
523/* Masks out the initial one bits in a start byte, leaving the real data ones.
524 * Doesn't work on an invariant byte. 'len' is the number of bytes in the
525 * multi-byte sequence that comprises the character. */
526#define UTF_START_MASK(len) (((len) >= 7) ? 0x00 : (0x1F >> ((len)-2)))
527
537124e4
KW
528/* Adds a UTF8 continuation byte 'new' of information to a running total code
529 * point 'old' of all the continuation bytes so far. This is designed to be
155d2738
KW
530 * used in a loop to convert from UTF-8 to the code point represented. Note
531 * that this is asymmetric on EBCDIC platforms, in that the 'new' parameter is
532 * the UTF-EBCDIC byte, whereas the 'old' parameter is a Unicode (not EBCDIC)
533 * code point in process of being generated */
a6951642
KW
534#define UTF8_ACCUMULATE(old, new) (__ASSERT_(FITS_IN_8_BITS(new)) \
535 ((old) << UTF_ACCUMULATION_SHIFT) \
155d2738
KW
536 | ((NATIVE_UTF8_TO_I8((U8)new)) \
537 & UTF_CONTINUATION_MASK))
d06134e5 538
4ab10950 539/* This works in the face of malformed UTF-8. */
4e1ed312
KW
540#define UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(s, e) \
541 ( UTF8_IS_DOWNGRADEABLE_START(*(s)) \
542 && ( (e) - (s) > 1) \
543 && UTF8_IS_CONTINUATION(*((s)+1)))
4ab10950 544
5aaebcb3 545/* Number of bytes a code point occupies in UTF-8. */
5352a763 546#define NATIVE_SKIP(uv) UVCHR_SKIP(uv)
bd18bd40 547
5aaebcb3
KW
548/* Most code which says UNISKIP is really thinking in terms of native code
549 * points (0-255) plus all those beyond. This is an imprecise term, but having
2accb712 550 * it means existing code continues to work. For precision, use UVCHR_SKIP,
5352a763
KW
551 * NATIVE_SKIP, or OFFUNISKIP */
552#define UNISKIP(uv) UVCHR_SKIP(uv)
5aaebcb3 553
3c0792e4
KW
554/* Longer, but more accurate name */
555#define UTF8_IS_ABOVE_LATIN1_START(c) UTF8_IS_ABOVE_LATIN1(c)
556
a62b247b
KW
557/* Convert a UTF-8 variant Latin1 character to a native code point value.
558 * Needs just one iteration of accumulate. Should be used only if it is known
559 * that the code point is < 256, and is not UTF-8 invariant. Use the slower
560 * but more general TWO_BYTE_UTF8_TO_NATIVE() which handles any code point
561 * representable by two bytes (which turns out to be up through
562 * MAX_PORTABLE_UTF8_TWO_BYTE). The two parameters are:
563 * HI: a downgradable start byte;
564 * LO: continuation.
565 * */
566#define EIGHT_BIT_UTF8_TO_NATIVE(HI, LO) \
567 ( __ASSERT_(UTF8_IS_DOWNGRADEABLE_START(HI)) \
568 __ASSERT_(UTF8_IS_CONTINUATION(LO)) \
569 LATIN1_TO_NATIVE(UTF8_ACCUMULATE(( \
570 NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), (LO))))
571
94bb8c36 572/* Convert a two (not one) byte utf8 character to a native code point value.
2950f2a7
KW
573 * Needs just one iteration of accumulate. Should not be used unless it is
574 * known that the two bytes are legal: 1) two-byte start, and 2) continuation.
575 * Note that the result can be larger than 255 if the input character is not
576 * downgradable */
94bb8c36 577#define TWO_BYTE_UTF8_TO_NATIVE(HI, LO) \
a6951642
KW
578 (__ASSERT_(FITS_IN_8_BITS(HI)) \
579 __ASSERT_(FITS_IN_8_BITS(LO)) \
580 __ASSERT_(PL_utf8skip[HI] == 2) \
581 __ASSERT_(UTF8_IS_CONTINUATION(LO)) \
94bb8c36 582 UNI_TO_NATIVE(UTF8_ACCUMULATE((NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), \
635e76f5 583 (LO))))
94bb8c36
KW
584
585/* Should never be used, and be deprecated */
586#define TWO_BYTE_UTF8_TO_UNI(HI, LO) NATIVE_TO_UNI(TWO_BYTE_UTF8_TO_NATIVE(HI, LO))
2950f2a7 587
bd18bd40
KW
588/*
589
590=for apidoc Am|STRLEN|UTF8SKIP|char* s
591returns the number of bytes in the UTF-8 encoded character whose first (perhaps
592only) byte is pointed to by C<s>.
593
594=cut
595 */
2a70536e
KW
596#define UTF8SKIP(s) PL_utf8skip[*(const U8*)(s)]
597#define UTF8_SKIP(s) UTF8SKIP(s)
d06134e5 598
2d1545e5
KW
599/* Most code that says 'UNI_' really means the native value for code points up
600 * through 255 */
601#define UNI_IS_INVARIANT(cp) UVCHR_IS_INVARIANT(cp)
602
c2b32798
KW
603/*
604=for apidoc Am|bool|UTF8_IS_INVARIANT|char c
605
606Evaluates to 1 if the byte C<c> represents the same character when encoded in
607UTF-8 as when not; otherwise evaluates to 0. UTF-8 invariant characters can be
608copied as-is when converting to/from UTF-8, saving time.
609
610In spite of the name, this macro gives the correct result if the input string
611from which C<c> comes is not encoded in UTF-8.
612
613See C<L</UVCHR_IS_INVARIANT>> for checking if a UV is invariant.
614
615=cut
616
617The reason it works on both UTF-8 encoded strings and non-UTF-8 encoded, is
618that it returns TRUE in each for the exact same set of bit patterns. It is
619valid on a subset of what UVCHR_IS_INVARIANT is valid on, so can just use that;
620and the compiler should optimize out anything extraneous given the
621implementation of the latter. The |0 makes sure this isn't mistakenly called
622with a ptr argument.
623*/
5c06326b 624#define UTF8_IS_INVARIANT(c) UVCHR_IS_INVARIANT((c) | 0)
5fc230f1
KW
625
626/* Like the above, but its name implies a non-UTF8 input, which as the comments
627 * above show, doesn't matter as to its implementation */
38953e5a 628#define NATIVE_BYTE_IS_INVARIANT(c) UVCHR_IS_INVARIANT(c)
d06134e5 629
48ccf5e1
KW
630/* The macros in the next 4 sets are used to generate the two utf8 or utfebcdic
631 * bytes from an ordinal that is known to fit into exactly two (not one) bytes;
632 * it must be less than 0x3FF to work across both encodings. */
633
634/* These two are helper macros for the other three sets, and should not be used
635 * directly anywhere else. 'translate_function' is either NATIVE_TO_LATIN1
1ff3baa2
KW
636 * (which works for code points up through 0xFF) or NATIVE_TO_UNI which works
637 * for any code point */
48ccf5e1 638#define __BASE_TWO_BYTE_HI(c, translate_function) \
2863dafa 639 (__ASSERT_(! UVCHR_IS_INVARIANT(c)) \
48ccf5e1 640 I8_TO_NATIVE_UTF8((translate_function(c) >> UTF_ACCUMULATION_SHIFT) \
2863dafa 641 | UTF_START_MARK(2)))
48ccf5e1 642#define __BASE_TWO_BYTE_LO(c, translate_function) \
2863dafa 643 (__ASSERT_(! UVCHR_IS_INVARIANT(c)) \
48ccf5e1 644 I8_TO_NATIVE_UTF8((translate_function(c) & UTF_CONTINUATION_MASK) \
2863dafa 645 | UTF_CONTINUATION_MARK))
48ccf5e1 646
48ccf5e1
KW
647/* The next two macros should not be used. They were designed to be usable as
648 * the case label of a switch statement, but this doesn't work for EBCDIC. Use
9d0d3a03 649 * regen/unicode_constants.pl instead */
48ccf5e1
KW
650#define UTF8_TWO_BYTE_HI_nocast(c) __BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI)
651#define UTF8_TWO_BYTE_LO_nocast(c) __BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI)
652
653/* The next two macros are used when the source should be a single byte
654 * character; checked for under DEBUGGING */
655#define UTF8_EIGHT_BIT_HI(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
4c8cd605 656 ( __BASE_TWO_BYTE_HI(c, NATIVE_TO_LATIN1)))
48ccf5e1 657#define UTF8_EIGHT_BIT_LO(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
4c8cd605 658 (__BASE_TWO_BYTE_LO(c, NATIVE_TO_LATIN1)))
48ccf5e1
KW
659
660/* These final two macros in the series are used when the source can be any
661 * code point whose UTF-8 is known to occupy 2 bytes; they are less efficient
662 * than the EIGHT_BIT versions on EBCDIC platforms. We use the logical '~'
663 * operator instead of "<=" to avoid getting compiler warnings.
d52b8576 664 * MAX_UTF8_TWO_BYTE should be exactly all one bits in the lower few
48ccf5e1
KW
665 * places, so the ~ works */
666#define UTF8_TWO_BYTE_HI(c) \
667 (__ASSERT_((sizeof(c) == 1) \
d52b8576 668 || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \
4c8cd605 669 (__BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI)))
48ccf5e1
KW
670#define UTF8_TWO_BYTE_LO(c) \
671 (__ASSERT_((sizeof(c) == 1) \
d52b8576 672 || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \
4c8cd605 673 (__BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI)))
d06134e5 674
e7214ce8
KW
675/* This is illegal in any well-formed UTF-8 in both EBCDIC and ASCII
676 * as it is only in overlongs. */
677#define ILLEGAL_UTF8_BYTE I8_TO_NATIVE_UTF8(0xC1)
678
7e2040f0 679/*
e3036cf4 680 * 'UTF' is whether or not p is encoded in UTF8. The names 'foo_lazy_if' stem
20df05f4
KW
681 * from an earlier version of these macros in which they didn't call the
682 * foo_utf8() macros (i.e. were 'lazy') unless they decided that *p is the
683 * beginning of a utf8 character. Now that foo_utf8() determines that itself,
684 * no need to do it again here
7e2040f0 685 */
34aeb2e9
KW
686#define isIDFIRST_lazy_if(p,UTF) \
687 _is_utf8_FOO(_CC_IDFIRST, (const U8 *) p, "isIDFIRST_lazy_if", \
688 "isIDFIRST_lazy_if_safe", \
689 cBOOL(UTF && ! IN_BYTES), 0, __FILE__,__LINE__)
1d72bdf6 690
da8c1a98
KW
691#define isIDFIRST_lazy_if_safe(p, e, UTF) \
692 ((IN_BYTES || !UTF) \
693 ? isIDFIRST(*(p)) \
694 : isIDFIRST_utf8_safe(p, e))
695
34aeb2e9
KW
696#define isWORDCHAR_lazy_if(p,UTF) \
697 _is_utf8_FOO(_CC_IDFIRST, (const U8 *) p, "isWORDCHAR_lazy_if", \
698 "isWORDCHAR_lazy_if_safe", \
699 cBOOL(UTF && ! IN_BYTES), 0, __FILE__,__LINE__)
700
da8c1a98
KW
701#define isWORDCHAR_lazy_if_safe(p, e, UTF) \
702 ((IN_BYTES || !UTF) \
703 ? isWORDCHAR(*(p)) \
704 : isWORDCHAR_utf8_safe((U8 *) p, (U8 *) e))
705
34aeb2e9
KW
706#define isALNUM_lazy_if(p,UTF) \
707 _is_utf8_FOO(_CC_IDFIRST, (const U8 *) p, "isALNUM_lazy_if", \
708 "isWORDCHAR_lazy_if_safe", \
709 cBOOL(UTF && ! IN_BYTES), 0, __FILE__,__LINE__)
da8c1a98 710
89ebb4a3
JH
711#define UTF8_MAXLEN UTF8_MAXBYTES
712
8cb75cc8
KW
713/* A Unicode character can fold to up to 3 characters */
714#define UTF8_MAX_FOLD_CHAR_EXPAND 3
715
d3481830 716#define IN_BYTES UNLIKELY(CopHINTS_get(PL_curcop) & HINT_BYTES)
bd18bd40
KW
717
718/*
719
720=for apidoc Am|bool|DO_UTF8|SV* sv
721Returns a bool giving whether or not the PV in C<sv> is to be treated as being
722encoded in UTF-8.
723
724You should use this I<after> a call to C<SvPV()> or one of its variants, in
725case any call to string overloading updates the internal UTF-8 encoding flag.
726
727=cut
728*/
0064a8a9 729#define DO_UTF8(sv) (SvUTF8(sv) && !IN_BYTES)
1ff3baa2
KW
730
731/* Should all strings be treated as Unicode, and not just UTF-8 encoded ones?
732 * Is so within 'feature unicode_strings' or 'locale :not_characters', and not
733 * within 'use bytes'. UTF-8 locales are not tested for here, but perhaps
734 * could be */
70844984
KW
735#define IN_UNI_8_BIT \
736 (( ( (CopHINTS_get(PL_curcop) & HINT_UNI_8_BIT)) \
737 || ( CopHINTS_get(PL_curcop) & HINT_LOCALE_PARTIAL \
738 /* -1 below is for :not_characters */ \
739 && _is_in_locale_category(FALSE, -1))) \
740 && (! IN_BYTES))
b36bf33f 741
1d72bdf6 742
c76687c5 743#define UTF8_ALLOW_EMPTY 0x0001 /* Allow a zero length string */
2b5e7bc2 744#define UTF8_GOT_EMPTY UTF8_ALLOW_EMPTY
c76687c5
KW
745
746/* Allow first byte to be a continuation byte */
1d72bdf6 747#define UTF8_ALLOW_CONTINUATION 0x0002
2b5e7bc2 748#define UTF8_GOT_CONTINUATION UTF8_ALLOW_CONTINUATION
c76687c5 749
2b5e7bc2 750/* Unexpected continuation byte */
1d72bdf6 751#define UTF8_ALLOW_NON_CONTINUATION 0x0004
2b5e7bc2 752#define UTF8_GOT_NON_CONTINUATION UTF8_ALLOW_NON_CONTINUATION
949cf498
KW
753
754/* expecting more bytes than were available in the string */
755#define UTF8_ALLOW_SHORT 0x0008
2b5e7bc2 756#define UTF8_GOT_SHORT UTF8_ALLOW_SHORT
949cf498 757
94953955
KW
758/* Overlong sequence; i.e., the code point can be specified in fewer bytes.
759 * First one will convert the overlong to the REPLACEMENT CHARACTER; second
760 * will return what the overlong evaluates to */
949cf498 761#define UTF8_ALLOW_LONG 0x0010
94953955 762#define UTF8_ALLOW_LONG_AND_ITS_VALUE (UTF8_ALLOW_LONG|0x0020)
2b5e7bc2
KW
763#define UTF8_GOT_LONG UTF8_ALLOW_LONG
764
d60baaa7
KW
765#define UTF8_ALLOW_OVERFLOW 0x0080
766#define UTF8_GOT_OVERFLOW UTF8_ALLOW_OVERFLOW
949cf498 767
f180b292 768#define UTF8_DISALLOW_SURROGATE 0x0100 /* Unicode surrogates */
2b5e7bc2 769#define UTF8_GOT_SURROGATE UTF8_DISALLOW_SURROGATE
f180b292 770#define UTF8_WARN_SURROGATE 0x0200
949cf498 771
c4e96019
KW
772/* Unicode non-character code points */
773#define UTF8_DISALLOW_NONCHAR 0x0400
2b5e7bc2 774#define UTF8_GOT_NONCHAR UTF8_DISALLOW_NONCHAR
c4e96019 775#define UTF8_WARN_NONCHAR 0x0800
949cf498 776
c4e96019
KW
777/* Super-set of Unicode: code points above the legal max */
778#define UTF8_DISALLOW_SUPER 0x1000
2b5e7bc2 779#define UTF8_GOT_SUPER UTF8_DISALLOW_SUPER
c4e96019
KW
780#define UTF8_WARN_SUPER 0x2000
781
782/* The original UTF-8 standard did not define UTF-8 with start bytes of 0xFE or
783 * 0xFF, though UTF-EBCDIC did. This allowed both versions to represent code
784 * points up to 2 ** 31 - 1. Perl extends UTF-8 so that 0xFE and 0xFF are
785 * usable on ASCII platforms, and 0xFF means something different than
786 * UTF-EBCDIC defines. These changes allow code points of 64 bits (actually
787 * somewhat more) to be represented on both platforms. But these are Perl
788 * extensions, and not likely to be interchangeable with other languages. Note
789 * that on ASCII platforms, FE overflows a signed 32-bit word, and FF an
790 * unsigned one. */
d044b7a7
KW
791#define UTF8_DISALLOW_PERL_EXTENDED 0x4000
792#define UTF8_GOT_PERL_EXTENDED UTF8_DISALLOW_PERL_EXTENDED
793#define UTF8_WARN_PERL_EXTENDED 0x8000
d35f2ca5 794
57ff5f59
KW
795/* For back compat, these old names are misleading for overlongs and
796 * UTF_EBCDIC. */
d044b7a7
KW
797#define UTF8_DISALLOW_ABOVE_31_BIT UTF8_DISALLOW_PERL_EXTENDED
798#define UTF8_GOT_ABOVE_31_BIT UTF8_GOT_PERL_EXTENDED
799#define UTF8_WARN_ABOVE_31_BIT UTF8_WARN_PERL_EXTENDED
800#define UTF8_DISALLOW_FE_FF UTF8_DISALLOW_PERL_EXTENDED
801#define UTF8_WARN_FE_FF UTF8_WARN_PERL_EXTENDED
949cf498 802
f180b292 803#define UTF8_CHECK_ONLY 0x10000
99a765e9 804#define _UTF8_NO_CONFIDENCE_IN_CURLEN 0x20000 /* Internal core use only */
949cf498
KW
805
806/* For backwards source compatibility. They do nothing, as the default now
807 * includes what they used to mean. The first one's meaning was to allow the
808 * just the single non-character 0xFFFF */
809#define UTF8_ALLOW_FFFF 0
810#define UTF8_ALLOW_SURROGATE 0
811
ecc1615f
KW
812/* C9 refers to Unicode Corrigendum #9: allows but discourages non-chars */
813#define UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE \
814 (UTF8_DISALLOW_SUPER|UTF8_DISALLOW_SURROGATE)
815#define UTF8_WARN_ILLEGAL_C9_INTERCHANGE (UTF8_WARN_SUPER|UTF8_WARN_SURROGATE)
816
d35f2ca5 817#define UTF8_DISALLOW_ILLEGAL_INTERCHANGE \
ecc1615f 818 (UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE|UTF8_DISALLOW_NONCHAR)
949cf498 819#define UTF8_WARN_ILLEGAL_INTERCHANGE \
ecc1615f
KW
820 (UTF8_WARN_ILLEGAL_C9_INTERCHANGE|UTF8_WARN_NONCHAR)
821
0eb3d6a0
KW
822/* This is typically used for code that processes UTF-8 input and doesn't want
823 * to have to deal with any malformations that might be present. All such will
824 * be safely replaced by the REPLACEMENT CHARACTER, unless other flags
825 * overriding this are also present. */
2d532c27
KW
826#define UTF8_ALLOW_ANY ( UTF8_ALLOW_CONTINUATION \
827 |UTF8_ALLOW_NON_CONTINUATION \
828 |UTF8_ALLOW_SHORT \
d60baaa7
KW
829 |UTF8_ALLOW_LONG \
830 |UTF8_ALLOW_OVERFLOW)
2d532c27
KW
831
832/* Accept any Perl-extended UTF-8 that evaluates to any UV on the platform, but
833 * not any malformed. This is the default. (Note that UVs above IV_MAX are
834 * deprecated. */
835#define UTF8_ALLOW_ANYUV 0
836#define UTF8_ALLOW_DEFAULT UTF8_ALLOW_ANYUV
1d72bdf6 837
89d986df
KW
838/*
839=for apidoc Am|bool|UTF8_IS_SURROGATE|const U8 *s|const U8 *e
840
841Evaluates to non-zero if the first few bytes of the string starting at C<s> and
842looking no further than S<C<e - 1>> are well-formed UTF-8 that represents one
843of the Unicode surrogate code points; otherwise it evaluates to 0. If
844non-zero, the value gives how many bytes starting at C<s> comprise the code
845point's representation.
846
847=cut
848 */
849#define UTF8_IS_SURROGATE(s, e) is_SURROGATE_utf8_safe(s, e)
850
851
852#define UTF8_IS_REPLACEMENT(s, send) is_REPLACEMENT_utf8_safe(s,send)
853
854/*
855=for apidoc Am|bool|UTF8_IS_SUPER|const U8 *s|const U8 *e
856
857Recall that Perl recognizes an extension to UTF-8 that can encode code
858points larger than the ones defined by Unicode, which are 0..0x10FFFF.
859
860This macro evaluates to non-zero if the first few bytes of the string starting
861at C<s> and looking no further than S<C<e - 1>> are from this UTF-8 extension;
862otherwise it evaluates to 0. If non-zero, the value gives how many bytes
863starting at C<s> comprise the code point's representation.
864
8650 is returned if the bytes are not well-formed extended UTF-8, or if they
866represent a code point that cannot fit in a UV on the current platform. Hence
867this macro can give different results when run on a 64-bit word machine than on
868one with a 32-bit word size.
0c58a72b 869
89d986df
KW
870Note that it is deprecated to have code points that are larger than what can
871fit in an IV on the current machine.
7131f24d 872
89d986df
KW
873=cut
874
875 * ASCII EBCDIC I8
7131f24d
KW
876 * U+10FFFF: \xF4\x8F\xBF\xBF \xF9\xA1\xBF\xBF\xBF max legal Unicode
877 * U+110000: \xF4\x90\x80\x80 \xF9\xA2\xA0\xA0\xA0
878 * U+110001: \xF4\x90\x80\x81 \xF9\xA2\xA0\xA0\xA1
89d986df
KW
879 */
880#ifdef EBCDIC
a14e0a36 881# define UTF8_IS_SUPER(s, e) \
89d986df
KW
882 (( LIKELY((e) > (s) + 4) \
883 && NATIVE_UTF8_TO_I8(*(s)) >= 0xF9 \
884 && ( NATIVE_UTF8_TO_I8(*(s)) > 0xF9 \
3d42f267 885 || (NATIVE_UTF8_TO_I8(*((s) + 1)) >= 0xA2)) \
89d986df 886 && LIKELY((s) + UTF8SKIP(s) <= (e))) \
2b479609 887 ? _is_utf8_char_helper(s, s + UTF8SKIP(s), 0) : 0)
7131f24d 888#else
a14e0a36 889# define UTF8_IS_SUPER(s, e) \
89d986df
KW
890 (( LIKELY((e) > (s) + 3) \
891 && (*(U8*) (s)) >= 0xF4 \
892 && ((*(U8*) (s)) > 0xF4 || (*((U8*) (s) + 1) >= 0x90))\
893 && LIKELY((s) + UTF8SKIP(s) <= (e))) \
2b479609 894 ? _is_utf8_char_helper(s, s + UTF8SKIP(s), 0) : 0)
7131f24d
KW
895#endif
896
b96a92fb
KW
897/* These are now machine generated, and the 'given' clause is no longer
898 * applicable */
0c58a72b 899#define UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s, e) \
89d986df
KW
900 cBOOL(is_NONCHAR_utf8_safe(s,e))
901
902/*
903=for apidoc Am|bool|UTF8_IS_NONCHAR|const U8 *s|const U8 *e
904
905Evaluates to non-zero if the first few bytes of the string starting at C<s> and
906looking no further than S<C<e - 1>> are well-formed UTF-8 that represents one
907of the Unicode non-character code points; otherwise it evaluates to 0. If
908non-zero, the value gives how many bytes starting at C<s> comprise the code
909point's representation.
910
911=cut
912 */
0c58a72b
KW
913#define UTF8_IS_NONCHAR(s, e) \
914 UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s, e)
7131f24d 915
c867b360
JH
916#define UNICODE_SURROGATE_FIRST 0xD800
917#define UNICODE_SURROGATE_LAST 0xDFFF
918#define UNICODE_REPLACEMENT 0xFFFD
919#define UNICODE_BYTE_ORDER_MARK 0xFEFF
1d72bdf6 920
b851fbc1 921/* Though our UTF-8 encoding can go beyond this,
c76687c5 922 * let's be conservative and do as Unicode says. */
b851fbc1
JH
923#define PERL_UNICODE_MAX 0x10FFFF
924
d044b7a7
KW
925#define UNICODE_WARN_SURROGATE 0x0001 /* UTF-16 surrogates */
926#define UNICODE_WARN_NONCHAR 0x0002 /* Non-char code points */
927#define UNICODE_WARN_SUPER 0x0004 /* Above 0x10FFFF */
928#define UNICODE_WARN_PERL_EXTENDED 0x0008 /* Above 0x7FFF_FFFF */
929#define UNICODE_WARN_ABOVE_31_BIT UNICODE_WARN_PERL_EXTENDED
930#define UNICODE_DISALLOW_SURROGATE 0x0010
931#define UNICODE_DISALLOW_NONCHAR 0x0020
932#define UNICODE_DISALLOW_SUPER 0x0040
933#define UNICODE_DISALLOW_PERL_EXTENDED 0x0080
934#define UNICODE_DISALLOW_ABOVE_31_BIT UNICODE_DISALLOW_PERL_EXTENDED
ecc1615f
KW
935#define UNICODE_WARN_ILLEGAL_C9_INTERCHANGE \
936 (UNICODE_WARN_SURROGATE|UNICODE_WARN_SUPER)
bb88be5f 937#define UNICODE_WARN_ILLEGAL_INTERCHANGE \
ecc1615f
KW
938 (UNICODE_WARN_ILLEGAL_C9_INTERCHANGE|UNICODE_WARN_NONCHAR)
939#define UNICODE_DISALLOW_ILLEGAL_C9_INTERCHANGE \
940 (UNICODE_DISALLOW_SURROGATE|UNICODE_DISALLOW_SUPER)
bb88be5f 941#define UNICODE_DISALLOW_ILLEGAL_INTERCHANGE \
ecc1615f 942 (UNICODE_DISALLOW_ILLEGAL_C9_INTERCHANGE|UNICODE_DISALLOW_NONCHAR)
949cf498
KW
943
944/* For backward source compatibility, as are now the default */
945#define UNICODE_ALLOW_SURROGATE 0
946#define UNICODE_ALLOW_SUPER 0
947#define UNICODE_ALLOW_ANY 0
b851fbc1 948
2d6b3d38
KW
949/* This matches the 2048 code points between UNICODE_SURROGATE_FIRST (0xD800) and
950 * UNICODE_SURROGATE_LAST (0xDFFF) */
951#define UNICODE_IS_SURROGATE(uv) (((UV) (uv) & (~0xFFFF | 0xF800)) \
952 == 0xD800)
953
646d1759
KW
954#define UNICODE_IS_REPLACEMENT(uv) ((UV) (uv) == UNICODE_REPLACEMENT)
955#define UNICODE_IS_BYTE_ORDER_MARK(uv) ((UV) (uv) == UNICODE_BYTE_ORDER_MARK)
c149ab20
KW
956
957/* Is 'uv' one of the 32 contiguous-range noncharacters? */
958#define UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv) ((UV) (uv) >= 0xFDD0 \
959 && (UV) (uv) <= 0xFDEF)
960
961/* Is 'uv' one of the 34 plane-ending noncharacters 0xFFFE, 0xFFFF, 0x1FFFE,
962 * 0x1FFFF, ... 0x10FFFE, 0x10FFFF, given that we know that 'uv' is not above
963 * the Unicode legal max */
964#define UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv) \
965 (((UV) (uv) & 0xFFFE) == 0xFFFE)
966
967#define UNICODE_IS_NONCHAR(uv) \
968 ( UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv) \
969 || ( LIKELY( ! UNICODE_IS_SUPER(uv)) \
970 && UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv)))
971
972#define UNICODE_IS_SUPER(uv) ((UV) (uv) > PERL_UNICODE_MAX)
1d72bdf6 973
ec34087a
KW
974#define LATIN_SMALL_LETTER_SHARP_S LATIN_SMALL_LETTER_SHARP_S_NATIVE
975#define LATIN_SMALL_LETTER_Y_WITH_DIAERESIS \
976 LATIN_SMALL_LETTER_Y_WITH_DIAERESIS_NATIVE
977#define MICRO_SIGN MICRO_SIGN_NATIVE
978#define LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE \
979 LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE_NATIVE
980#define LATIN_SMALL_LETTER_A_WITH_RING_ABOVE \
981 LATIN_SMALL_LETTER_A_WITH_RING_ABOVE_NATIVE
09091399
JH
982#define UNICODE_GREEK_CAPITAL_LETTER_SIGMA 0x03A3
983#define UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA 0x03C2
984#define UNICODE_GREEK_SMALL_LETTER_SIGMA 0x03C3
9dcbe121 985#define GREEK_SMALL_LETTER_MU 0x03BC
9e682c18
KW
986#define GREEK_CAPITAL_LETTER_MU 0x039C /* Upper and title case
987 of MICRON */
988#define LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS 0x0178 /* Also is title case */
0766489e
KW
989#ifdef LATIN_CAPITAL_LETTER_SHARP_S_UTF8
990# define LATIN_CAPITAL_LETTER_SHARP_S 0x1E9E
991#endif
74894415
KW
992#define LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE 0x130
993#define LATIN_SMALL_LETTER_DOTLESS_I 0x131
9e682c18 994#define LATIN_SMALL_LETTER_LONG_S 0x017F
a9f50d33
KW
995#define LATIN_SMALL_LIGATURE_LONG_S_T 0xFB05
996#define LATIN_SMALL_LIGATURE_ST 0xFB06
9e682c18
KW
997#define KELVIN_SIGN 0x212A
998#define ANGSTROM_SIGN 0x212B
09091399 999
9e55ce06 1000#define UNI_DISPLAY_ISPRINT 0x0001
c728cb41
JH
1001#define UNI_DISPLAY_BACKSLASH 0x0002
1002#define UNI_DISPLAY_QQ (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH)
1003#define UNI_DISPLAY_REGEX (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH)
9e55ce06 1004
5cd46e1f
KW
1005#define ANYOF_FOLD_SHARP_S(node, input, end) \
1006 (ANYOF_BITMAP_TEST(node, LATIN_SMALL_LETTER_SHARP_S) && \
137165a6 1007 (ANYOF_NONBITMAP(node)) && \
39065660 1008 (ANYOF_FLAGS(node) & ANYOF_LOC_NONBITMAP_FOLD) && \
07b6858f 1009 ((end) > (input) + 1) && \
305b8651 1010 isALPHA_FOLD_EQ((input)[0], 's'))
6302f837 1011
ebc501f0 1012#define SHARP_S_SKIP 2
3b0fc154 1013
6302f837 1014/*
5dca9278
KW
1015
1016=for apidoc Am|STRLEN|isUTF8_CHAR|const U8 *s|const U8 *e
1017
35f8c9bd 1018Evaluates to non-zero if the first few bytes of the string starting at C<s> and
2c6ed66c
KW
1019looking no further than S<C<e - 1>> are well-formed UTF-8, as extended by Perl,
1020that represents some code point; otherwise it evaluates to 0. If non-zero, the
1021value gives how many bytes starting at C<s> comprise the code point's
2717076a
KW
1022representation. Any bytes remaining before C<e>, but beyond the ones needed to
1023form the first code point in C<s>, are not examined.
6302f837 1024
35f8c9bd
KW
1025The code point can be any that will fit in a UV on this machine, using Perl's
1026extension to official UTF-8 to represent those higher than the Unicode maximum
1027of 0x10FFFF. That means that this macro is used to efficiently decide if the
2717076a
KW
1028next few bytes in C<s> is legal UTF-8 for a single character.
1029
1030Use C<L</isSTRICT_UTF8_CHAR>> to restrict the acceptable code points to those
1031defined by Unicode to be fully interchangeable across applications;
1032C<L</isC9_STRICT_UTF8_CHAR>> to use the L<Unicode Corrigendum
1033#9|http://www.unicode.org/versions/corrigendum9.html> definition of allowable
1034code points; and C<L</isUTF8_CHAR_flags>> for a more customized definition.
1035
1036Use C<L</is_utf8_string>>, C<L</is_utf8_string_loc>>, and
1037C<L</is_utf8_string_loclen>> to check entire strings.
35f8c9bd
KW
1038
1039Note that it is deprecated to use code points higher than what will fit in an
1040IV. This macro does not raise any warnings for such code points, treating them
1041as valid.
1042
1043Note also that a UTF-8 INVARIANT character (i.e. ASCII on non-EBCDIC machines)
1044is a valid UTF-8 character.
5dca9278
KW
1045
1046=cut
1047*/
6302f837 1048
784d4f31
KW
1049#define isUTF8_CHAR(s, e) \
1050 (UNLIKELY((e) <= (s)) \
1051 ? 0 \
1052 : (UTF8_IS_INVARIANT(*s)) \
df863e43
KW
1053 ? 1 \
1054 : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \
1055 ? 0 \
1056 : LIKELY(NATIVE_UTF8_TO_I8(*s) <= _IS_UTF8_CHAR_HIGHEST_START_BYTE) \
1057 ? is_UTF8_CHAR_utf8_no_length_checks(s) \
1058 : _is_utf8_char_helper(s, e, 0))
6302f837 1059
3cedd9d9 1060#define is_utf8_char_buf(buf, buf_end) isUTF8_CHAR(buf, buf_end)
976c1b08
KW
1061#define bytes_from_utf8(s, lenp, is_utf8p) \
1062 bytes_from_utf8_loc(s, lenp, is_utf8p, 0)
3cedd9d9 1063
e23e8bc1
KW
1064/*
1065
1066=for apidoc Am|STRLEN|isSTRICT_UTF8_CHAR|const U8 *s|const U8 *e
1067
1068Evaluates to non-zero if the first few bytes of the string starting at C<s> and
1069looking no further than S<C<e - 1>> are well-formed UTF-8 that represents some
1070Unicode code point completely acceptable for open interchange between all
1071applications; otherwise it evaluates to 0. If non-zero, the value gives how
2717076a
KW
1072many bytes starting at C<s> comprise the code point's representation. Any
1073bytes remaining before C<e>, but beyond the ones needed to form the first code
1074point in C<s>, are not examined.
e23e8bc1
KW
1075
1076The largest acceptable code point is the Unicode maximum 0x10FFFF, and must not
1077be a surrogate nor a non-character code point. Thus this excludes any code
1078point from Perl's extended UTF-8.
1079
1080This is used to efficiently decide if the next few bytes in C<s> is
2717076a
KW
1081legal Unicode-acceptable UTF-8 for a single character.
1082
1083Use C<L</isC9_STRICT_UTF8_CHAR>> to use the L<Unicode Corrigendum
1084#9|http://www.unicode.org/versions/corrigendum9.html> definition of allowable
1085code points; C<L</isUTF8_CHAR>> to check for Perl's extended UTF-8;
1086and C<L</isUTF8_CHAR_flags>> for a more customized definition.
e23e8bc1 1087
9f2abfde
KW
1088Use C<L</is_strict_utf8_string>>, C<L</is_strict_utf8_string_loc>>, and
1089C<L</is_strict_utf8_string_loclen>> to check entire strings.
1090
e23e8bc1
KW
1091=cut
1092*/
1093
1094#define isSTRICT_UTF8_CHAR(s, e) \
1095 (UNLIKELY((e) <= (s)) \
1096 ? 0 \
1097 : (UTF8_IS_INVARIANT(*s)) \
1098 ? 1 \
1099 : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \
1100 ? 0 \
1101 : is_STRICT_UTF8_CHAR_utf8_no_length_checks(s))
1102
a82be82b
KW
1103/*
1104
1105=for apidoc Am|STRLEN|isC9_STRICT_UTF8_CHAR|const U8 *s|const U8 *e
1106
1107Evaluates to non-zero if the first few bytes of the string starting at C<s> and
1108looking no further than S<C<e - 1>> are well-formed UTF-8 that represents some
1109Unicode non-surrogate code point; otherwise it evaluates to 0. If non-zero,
1110the value gives how many bytes starting at C<s> comprise the code point's
2717076a
KW
1111representation. Any bytes remaining before C<e>, but beyond the ones needed to
1112form the first code point in C<s>, are not examined.
a82be82b
KW
1113
1114The largest acceptable code point is the Unicode maximum 0x10FFFF. This
1115differs from C<L</isSTRICT_UTF8_CHAR>> only in that it accepts non-character
1116code points. This corresponds to
1117L<Unicode Corrigendum #9|http://www.unicode.org/versions/corrigendum9.html>.
1118which said that non-character code points are merely discouraged rather than
1119completely forbidden in open interchange. See
1120L<perlunicode/Noncharacter code points>.
1121
2717076a
KW
1122Use C<L</isUTF8_CHAR>> to check for Perl's extended UTF-8; and
1123C<L</isUTF8_CHAR_flags>> for a more customized definition.
1124
9f2abfde
KW
1125Use C<L</is_c9strict_utf8_string>>, C<L</is_c9strict_utf8_string_loc>>, and
1126C<L</is_c9strict_utf8_string_loclen>> to check entire strings.
1127
a82be82b
KW
1128=cut
1129*/
1130
1131#define isC9_STRICT_UTF8_CHAR(s, e) \
1132 (UNLIKELY((e) <= (s)) \
1133 ? 0 \
1134 : (UTF8_IS_INVARIANT(*s)) \
1135 ? 1 \
1136 : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \
1137 ? 0 \
1138 : is_C9_STRICT_UTF8_CHAR_utf8_no_length_checks(s))
1139
25e3a4e0
KW
1140/*
1141
1142=for apidoc Am|STRLEN|isUTF8_CHAR_flags|const U8 *s|const U8 *e| const U32 flags
1143
1144Evaluates to non-zero if the first few bytes of the string starting at C<s> and
1145looking no further than S<C<e - 1>> are well-formed UTF-8, as extended by Perl,
1146that represents some code point, subject to the restrictions given by C<flags>;
1147otherwise it evaluates to 0. If non-zero, the value gives how many bytes
2717076a
KW
1148starting at C<s> comprise the code point's representation. Any bytes remaining
1149before C<e>, but beyond the ones needed to form the first code point in C<s>,
1150are not examined.
25e3a4e0
KW
1151
1152If C<flags> is 0, this gives the same results as C<L</isUTF8_CHAR>>;
1153if C<flags> is C<UTF8_DISALLOW_ILLEGAL_INTERCHANGE>, this gives the same results
1154as C<L</isSTRICT_UTF8_CHAR>>;
1155and if C<flags> is C<UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE>, this gives
1156the same results as C<L</isC9_STRICT_UTF8_CHAR>>.
1157Otherwise C<flags> may be any combination of the C<UTF8_DISALLOW_I<foo>> flags
1158understood by C<L</utf8n_to_uvchr>>, with the same meanings.
1159
1160The three alternative macros are for the most commonly needed validations; they
1161are likely to run somewhat faster than this more general one, as they can be
1162inlined into your code.
1163
9f2abfde
KW
1164Use L</is_utf8_string_flags>, L</is_utf8_string_loc_flags>, and
1165L</is_utf8_string_loclen_flags> to check entire strings.
1166
25e3a4e0
KW
1167=cut
1168*/
1169
1170#define isUTF8_CHAR_flags(s, e, flags) \
1171 (UNLIKELY((e) <= (s)) \
1172 ? 0 \
1173 : (UTF8_IS_INVARIANT(*s)) \
1174 ? 1 \
1175 : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \
1176 ? 0 \
1177 : _is_utf8_char_helper(s, e, flags))
1178
6302f837
KW
1179/* Do not use; should be deprecated. Use isUTF8_CHAR() instead; this is
1180 * retained solely for backwards compatibility */
1181#define IS_UTF8_CHAR(p, n) (isUTF8_CHAR(p, (p) + (n)) == n)
e9a8c099 1182
6a5bc5ac 1183#endif /* PERL_UTF8_H_ */
57f0e7e2 1184
e9a8c099 1185/*
14d04a33 1186 * ex: set ts=8 sts=4 sw=4 et:
e9a8c099 1187 */