This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
APItest/t/utf8_warn_base.pl: Generate smaller overlongs
[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)
de69f3af 77
a0270393 78#define to_uni_fold(c, p, lenp) _to_uni_fold_flags(c, p, lenp, FOLD_FLAGS_FULL)
a239b1e2
KW
79
80#define to_utf8_fold(s, r, lenr) \
607313a1 81 _to_utf8_fold_flags (s, NULL, r, lenr, FOLD_FLAGS_FULL, __FILE__, __LINE__)
a239b1e2 82#define to_utf8_lower(s, r, lenr) \
607313a1 83 _to_utf8_lower_flags(s, NULL, r ,lenr, 0, __FILE__, __LINE__)
a239b1e2 84#define to_utf8_upper(s, r, lenr) \
607313a1 85 _to_utf8_upper_flags(s, NULL, r, lenr, 0, __FILE__, __LINE__)
a239b1e2 86#define to_utf8_title(s, r, lenr) \
607313a1 87 _to_utf8_title_flags(s, NULL, r, lenr ,0, __FILE__, __LINE__)
36bb2ab6 88
eda9cac1
KW
89#define foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \
90 foldEQ_utf8_flags(s1, pe1, l1, u1, s2, pe2, l2, u2, 0)
baa60164 91#define FOLDEQ_UTF8_NOMIX_ASCII (1 << 0)
cea315b6 92#define FOLDEQ_LOCALE (1 << 1)
18f762c3
KW
93#define FOLDEQ_S1_ALREADY_FOLDED (1 << 2)
94#define FOLDEQ_S2_ALREADY_FOLDED (1 << 3)
d635b710
KW
95#define FOLDEQ_S1_FOLDS_SANE (1 << 4)
96#define FOLDEQ_S2_FOLDS_SANE (1 << 5)
a33c29bc 97
e6226b18
KW
98#define ibcmp_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \
99 cBOOL(! foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2))
100
1d72bdf6
NIS
101#ifdef EBCDIC
102/* The equivalent of these macros but implementing UTF-EBCDIC
103 are in the following header file:
104 */
105
106#include "utfebcdic.h"
fd7cb289 107
d06134e5 108#else /* ! EBCDIC */
73c4f7a1
GS
109START_EXTERN_C
110
111e8ed9
KW
111/* How wide can a single UTF-8 encoded character become in bytes. */
112/* NOTE: Strictly speaking Perl's UTF-8 should not be called UTF-8 since UTF-8
113 * is an encoding of Unicode, and Unicode's upper limit, 0x10FFFF, can be
114 * expressed with 4 bytes. However, Perl thinks of UTF-8 as a way to encode
115 * non-negative integers in a binary format, even those above Unicode */
116#define UTF8_MAXBYTES 13
117
a0ed51b3 118#ifdef DOINIT
6f06b55f 119EXTCONST unsigned char PL_utf8skip[] = {
b2635aa8
KW
120/* 0x00 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
121/* 0x10 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
122/* 0x20 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
123/* 0x30 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
124/* 0x40 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
125/* 0x50 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
126/* 0x60 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
127/* 0x70 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
128/* 0x80 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
129/* 0x90 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
130/* 0xA0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
131/* 0xB0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
132/* 0xC0 */ 2,2, /* overlong */
1ff3baa2 133/* 0xC2 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0080 to U+03FF */
b2635aa8
KW
134/* 0xD0 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0400 to U+07FF */
135/* 0xE0 */ 3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3, /* U+0800 to U+FFFF */
136/* 0xF0 */ 4,4,4,4,4,4,4,4,5,5,5,5,6,6, /* above BMP to 2**31 - 1 */
6937f885
KW
137 /* Perl extended (never was official UTF-8). Up to 36 bit */
138/* 0xFE */ 7,
139 /* More extended, Up to 72 bits (64-bit + reserved) */
111e8ed9 140/* 0xFF */ UTF8_MAXBYTES
a0ed51b3
LW
141};
142#else
6f06b55f 143EXTCONST unsigned char PL_utf8skip[];
a0ed51b3
LW
144#endif
145
73c4f7a1 146END_EXTERN_C
7e2040f0 147
1a3756de 148#if defined(_MSC_VER) && _MSC_VER < 1400
6f6d1bab
TC
149/* older MSVC versions have a smallish macro buffer */
150#define PERL_SMALL_MACRO_BUFFER
151#endif
152
59a449d5
KW
153/* Native character to/from iso-8859-1. Are the identity functions on ASCII
154 * platforms */
6f6d1bab 155#ifdef PERL_SMALL_MACRO_BUFFER
1a3756de
TC
156#define NATIVE_TO_LATIN1(ch) ((U8)(ch))
157#define LATIN1_TO_NATIVE(ch) ((U8)(ch))
6f6d1bab 158#else
4c8cd605
KW
159#define NATIVE_TO_LATIN1(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch)))
160#define LATIN1_TO_NATIVE(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch)))
6f6d1bab 161#endif
59a449d5
KW
162
163/* I8 is an intermediate version of UTF-8 used only in UTF-EBCDIC. We thus
164 * consider it to be identical to UTF-8 on ASCII platforms. Strictly speaking
165 * UTF-8 and UTF-EBCDIC are two different things, but we often conflate them
166 * because they are 8-bit encodings that serve the same purpose in Perl, and
167 * rarely do we need to distinguish them. The term "NATIVE_UTF8" applies to
168 * whichever one is applicable on the current platform */
6f6d1bab 169#ifdef PERL_SMALL_MACRO_BUFFER
1d4ea287
DD
170#define NATIVE_UTF8_TO_I8(ch) ((U8) (ch))
171#define I8_TO_NATIVE_UTF8(ch) ((U8) (ch))
6f6d1bab 172#else
4c8cd605
KW
173#define NATIVE_UTF8_TO_I8(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch)))
174#define I8_TO_NATIVE_UTF8(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch)))
6f6d1bab 175#endif
59a449d5 176
1d72bdf6 177/* Transforms in wide UV chars */
4c8cd605
KW
178#define UNI_TO_NATIVE(ch) ((UV) (ch))
179#define NATIVE_TO_UNI(ch) ((UV) (ch))
d7578b48 180
877d9f0d 181/*
9041c2e3 182
a14e0a36
KW
183 The following table is from Unicode 3.2, plus the Perl extensions for above
184 U+10FFFF
877d9f0d 185
a14e0a36 186 Code Points 1st Byte 2nd Byte 3rd 4th 5th 6th 7th 8th-13th
877d9f0d 187
375122d7 188 U+0000..U+007F 00..7F
e1b711da 189 U+0080..U+07FF * C2..DF 80..BF
a14e0a36
KW
190 U+0800..U+0FFF E0 * A0..BF 80..BF
191 U+1000..U+CFFF E1..EC 80..BF 80..BF
192 U+D000..U+D7FF ED 80..9F 80..BF
193 U+D800..U+DFFF ED A0..BF 80..BF (surrogates)
194 U+E000..U+FFFF EE..EF 80..BF 80..BF
195 U+10000..U+3FFFF F0 * 90..BF 80..BF 80..BF
196 U+40000..U+FFFFF F1..F3 80..BF 80..BF 80..BF
197 U+100000..U+10FFFF F4 80..8F 80..BF 80..BF
198 Below are above-Unicode code points
199 U+110000..U+13FFFF F4 90..BF 80..BF 80..BF
200 U+110000..U+1FFFFF F5..F7 80..BF 80..BF 80..BF
201 U+200000..U+FFFFFF F8 * 88..BF 80..BF 80..BF 80..BF
202U+1000000..U+3FFFFFF F9..FB 80..BF 80..BF 80..BF 80..BF
203U+4000000..U+3FFFFFFF FC * 84..BF 80..BF 80..BF 80..BF 80..BF
204U+40000000..U+7FFFFFFF FD 80..BF 80..BF 80..BF 80..BF 80..BF
205U+80000000..U+FFFFFFFFF FE * 82..BF 80..BF 80..BF 80..BF 80..BF 80..BF
206U+1000000000.. FF 80..BF 80..BF 80..BF 80..BF 80..BF * 81..BF 80..BF
877d9f0d 207
e1b711da 208Note the gaps before several of the byte entries above marked by '*'. These are
37e2e78e
KW
209caused by legal UTF-8 avoiding non-shortest encodings: it is technically
210possible to UTF-8-encode a single code point in different ways, but that is
211explicitly forbidden, and the shortest possible encoding should always be used
15824458 212(and that is what Perl does). The non-shortest ones are called 'overlongs'.
8c007b5a 213
877d9f0d
JH
214 */
215
8c007b5a
JH
216/*
217 Another way to look at it, as bits:
218
b2635aa8 219 Code Points 1st Byte 2nd Byte 3rd Byte 4th Byte
8c007b5a 220
b2635aa8
KW
221 0aaa aaaa 0aaa aaaa
222 0000 0bbb bbaa aaaa 110b bbbb 10aa aaaa
223 cccc bbbb bbaa aaaa 1110 cccc 10bb bbbb 10aa aaaa
224 00 000d ddcc cccc bbbb bbaa aaaa 1111 0ddd 10cc cccc 10bb bbbb 10aa aaaa
8c007b5a
JH
225
226As you can see, the continuation bytes all begin with C<10>, and the
e1b711da 227leading bits of the start byte tell how many bytes there are in the
8c007b5a
JH
228encoded character.
229
df863e43
KW
230Perl's extended UTF-8 means we can have start bytes up through FF, though any
231beginning with FF yields a code point that is too large for 32-bit ASCII
232platforms. FF signals to use 13 bytes for the encoded character. This breaks
233the paradigm that the number of leading bits gives how many total bytes there
234are in the character.
65ab9279 235
8c007b5a
JH
236*/
237
6c88483e 238/* Is the representation of the Unicode code point 'cp' the same regardless of
15824458 239 * being encoded in UTF-8 or not? */
2d1545e5 240#define OFFUNI_IS_INVARIANT(cp) isASCII(cp)
15824458 241
c2b32798
KW
242/*
243=for apidoc Am|bool|UVCHR_IS_INVARIANT|UV cp
244
245Evaluates to 1 if the representation of code point C<cp> is the same whether or
246not it is encoded in UTF-8; otherwise evaluates to 0. UTF-8 invariant
247characters can be copied as-is when converting to/from UTF-8, saving time.
248C<cp> is Unicode if above 255; otherwise is platform-native.
249
250=cut
251 */
252
cf1be84e 253#define UVCHR_IS_INVARIANT(cp) OFFUNI_IS_INVARIANT(cp)
38953e5a 254
c9264833
KW
255/* This defines the bits that are to be in the continuation bytes of a multi-byte
256 * UTF-8 encoded character that mark it is a continuation byte. */
257#define UTF_CONTINUATION_MARK 0x80
258
a95ec4fb 259/* Misleadingly named: is the UTF8-encoded byte 'c' part of a variant sequence
5c06326b
KW
260 * in UTF-8? This is the inverse of UTF8_IS_INVARIANT. The |0 makes sure this
261 * isn't mistakenly called with a ptr argument */
a6951642
KW
262#define UTF8_IS_CONTINUED(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
263 ((U8)((c) | 0)) & UTF_CONTINUATION_MARK)
15824458
KW
264
265/* Is the byte 'c' the first byte of a multi-byte UTF8-8 encoded sequence?
266 * This doesn't catch invariants (they are single-byte). It also excludes the
5c06326b
KW
267 * illegal overlong sequences that begin with C0 and C1. The |0 makes sure
268 * this isn't mistakenly called with a ptr argument */
a6951642
KW
269#define UTF8_IS_START(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
270 ((U8)((c) | 0)) >= 0xc2)
15824458 271
858cd8ab
KW
272/* For use in UTF8_IS_CONTINUATION() below */
273#define UTF_IS_CONTINUATION_MASK 0xC0
274
15824458 275/* Is the byte 'c' part of a multi-byte UTF8-8 encoded sequence, and not the
5c06326b
KW
276 * first byte thereof? The |0 makes sure this isn't mistakenly called with a
277 * ptr argument */
a6951642
KW
278#define UTF8_IS_CONTINUATION(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
279 (((U8)((c) | 0)) & UTF_IS_CONTINUATION_MASK) == UTF_CONTINUATION_MARK)
0ae1fa71 280
15824458
KW
281/* Is the UTF8-encoded byte 'c' the first byte of a two byte sequence? Use
282 * UTF8_IS_NEXT_CHAR_DOWNGRADEABLE() instead if the input isn't known to
283 * be well-formed. Masking with 0xfe allows the low bit to be 0 or 1; thus
5c06326b
KW
284 * this matches 0xc[23]. The |0 makes sure this isn't mistakenly called with a
285 * ptr argument */
a6951642
KW
286#define UTF8_IS_DOWNGRADEABLE_START(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
287 (((U8)((c) | 0)) & 0xfe) == 0xc2)
4ab10950 288
15824458 289/* Is the UTF8-encoded byte 'c' the first byte of a sequence of bytes that
5c06326b
KW
290 * represent a code point > 255? The |0 makes sure this isn't mistakenly
291 * called with a ptr argument */
a6951642
KW
292#define UTF8_IS_ABOVE_LATIN1(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
293 ((U8)((c) | 0)) >= 0xc4)
8850bf83 294
15824458
KW
295/* This is the number of low-order bits a continuation byte in a UTF-8 encoded
296 * sequence contributes to the specification of the code point. In the bit
297 * maps above, you see that the first 2 bits are a constant '10', leaving 6 of
298 * real information */
1d72bdf6 299#define UTF_ACCUMULATION_SHIFT 6
b2635aa8 300
fed423a5
KW
301/* ^? is defined to be DEL on ASCII systems. See the definition of toCTRL()
302 * for more */
303#define QUESTION_MARK_CTRL DEL_NATIVE
304
305/* Surrogates, non-character code points and above-Unicode code points are
306 * problematic in some contexts. This allows code that needs to check for
307 * those to to quickly exclude the vast majority of code points it will
308 * encounter */
a6951642
KW
309#define isUTF8_POSSIBLY_PROBLEMATIC(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
310 (U8) c >= 0xED)
fed423a5 311
784d4f31
KW
312/* A helper macro for isUTF8_CHAR, so use that one instead of this. This was
313 * generated by regen/regcharclass.pl, and then moved here. Then it was
314 * hand-edited to add some LIKELY() calls, presuming that malformations are
315 * unlikely. The lines that generated it were then commented out. This was
316 * done because it takes on the order of 10 minutes to generate, and is never
317 * going to change, unless the generated code is improved, and figuring out
318 * the LIKELYs there would be hard.
319 *
320 UTF8_CHAR: Matches legal UTF-8 variant code points up through 0x1FFFFFF
321
322 0x80 - 0x1FFFFF
323*/
324/*** GENERATED CODE ***/
325#define is_UTF8_CHAR_utf8_no_length_checks(s) \
9f2eed98
P
326( ( 0xC2 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xDF ) ? \
327 ( LIKELY( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) ? 2 : 0 ) \
328: ( 0xE0 == ((const U8*)s)[0] ) ? \
329 ( LIKELY( ( ( ((const U8*)s)[1] & 0xE0 ) == 0xA0 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
330: ( 0xE1 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xEF ) ? \
331 ( LIKELY( ( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
332: ( 0xF0 == ((const U8*)s)[0] ) ? \
333 ( 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 )\
334: ( ( ( ( 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
335
336/* The above macro handles UTF-8 that has this start byte as the maximum */
337#define _IS_UTF8_CHAR_HIGHEST_START_BYTE 0xF7
338
e23e8bc1
KW
339/* A helper macro for isSTRICT_UTF8_CHAR, so use that one instead of this.
340 * Like is_UTF8_CHAR_utf8_no_length_checks(), this was moved here and LIKELYs
341 * added manually.
342 *
343 STRICT_UTF8_CHAR: Matches legal Unicode UTF-8 variant code points, no
344 surrrogates nor non-character code points
345*/
346/*** GENERATED CODE ***/
347#define is_STRICT_UTF8_CHAR_utf8_no_length_checks(s) \
9f2eed98
P
348( ( 0xC2 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xDF ) ? \
349 ( LIKELY( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) ? 2 : 0 ) \
350: ( 0xE0 == ((const U8*)s)[0] ) ? \
351 ( LIKELY( ( ( ((const U8*)s)[1] & 0xE0 ) == 0xA0 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
352: ( ( 0xE1 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xEC ) || 0xEE == ((const U8*)s)[0] ) ?\
353 ( ( ( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
354: ( 0xED == ((const U8*)s)[0] ) ? \
355 ( LIKELY( ( ( ((const U8*)s)[1] & 0xE0 ) == 0x80 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
356: ( 0xEF == ((const U8*)s)[0] ) ? \
357 ( ( ( 0x80 <= ((const U8*)s)[1] && ((const U8*)s)[1] <= 0xB6 ) || ( 0xB8 <= ((const U8*)s)[1] && ((const U8*)s)[1] <= 0xBE ) ) ?\
358 ( LIKELY( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ? 3 : 0 ) \
359 : ( 0xB7 == ((const U8*)s)[1] ) ? \
360 ( LIKELY( ( ((const U8*)s)[2] & 0xF0 ) == 0x80 || ( ((const U8*)s)[2] & 0xF0 ) == 0xB0 ) ? 3 : 0 )\
361 : ( ( 0xBF == ((const U8*)s)[1] ) && ( 0x80 <= ((const U8*)s)[2] && ((const U8*)s)[2] <= 0xBD ) ) ? 3 : 0 )\
362: ( 0xF0 == ((const U8*)s)[0] ) ? \
363 ( ( ( 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 ) ) ?\
364 ( LIKELY( ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\
365 : ( ((const U8*)s)[1] == 0x9F || ( ( ((const U8*)s)[1] & 0xEF ) == 0xAF ) ) ? \
366 ( ( 0x80 <= ((const U8*)s)[2] && ((const U8*)s)[2] <= 0xBE ) ? \
367 ( LIKELY( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ? 4 : 0 ) \
368 : LIKELY( ( 0xBF == ((const U8*)s)[2] ) && ( 0x80 <= ((const U8*)s)[3] && ((const U8*)s)[3] <= 0xBD ) ) ? 4 : 0 )\
e23e8bc1 369 : 0 ) \
9f2eed98
P
370: ( 0xF1 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xF3 ) ? \
371 ( ( ( ( ((const U8*)s)[1] & 0xC8 ) == 0x80 ) || ( ( ((const U8*)s)[1] & 0xCC ) == 0x88 ) || ( ( ((const U8*)s)[1] & 0xCE ) == 0x8C ) || ( ( ((const U8*)s)[1] & 0xCF ) == 0x8E ) ) ?\
372 ( LIKELY( ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\
373 : ( ( ((const U8*)s)[1] & 0xCF ) == 0x8F ) ? \
374 ( ( 0x80 <= ((const U8*)s)[2] && ((const U8*)s)[2] <= 0xBE ) ? \
375 ( LIKELY( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ? 4 : 0 ) \
376 : LIKELY( ( 0xBF == ((const U8*)s)[2] ) && ( 0x80 <= ((const U8*)s)[3] && ((const U8*)s)[3] <= 0xBD ) ) ? 4 : 0 )\
e23e8bc1 377 : 0 ) \
9f2eed98
P
378: ( 0xF4 == ((const U8*)s)[0] ) ? \
379 ( ( 0x80 <= ((const U8*)s)[1] && ((const U8*)s)[1] <= 0x8E ) ? \
380 ( LIKELY( ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\
381 : ( 0x8F == ((const U8*)s)[1] ) ? \
382 ( ( 0x80 <= ((const U8*)s)[2] && ((const U8*)s)[2] <= 0xBE ) ? \
383 ( LIKELY( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ? 4 : 0 ) \
384 : LIKELY( ( 0xBF == ((const U8*)s)[2] ) && ( 0x80 <= ((const U8*)s)[3] && ((const U8*)s)[3] <= 0xBD ) ) ? 4 : 0 )\
e23e8bc1
KW
385 : 0 ) \
386: 0 )
387
a82be82b
KW
388/* Similarly,
389 C9_STRICT_UTF8_CHAR: Matches legal Unicode UTF-8 variant code
390 points, no surrogates
391 0x0080 - 0xD7FF
392 0xE000 - 0x10FFFF
393*/
394/*** GENERATED CODE ***/
395#define is_C9_STRICT_UTF8_CHAR_utf8_no_length_checks(s) \
9f2eed98
P
396( ( 0xC2 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xDF ) ? \
397 ( LIKELY( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) ? 2 : 0 ) \
398: ( 0xE0 == ((const U8*)s)[0] ) ? \
399 ( LIKELY( ( ( ((const U8*)s)[1] & 0xE0 ) == 0xA0 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
400: ( ( 0xE1 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xEC ) || ( ((const U8*)s)[0] & 0xFE ) == 0xEE ) ?\
401 ( LIKELY( ( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
402: ( 0xED == ((const U8*)s)[0] ) ? \
403 ( LIKELY( ( ( ((const U8*)s)[1] & 0xE0 ) == 0x80 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
404: ( 0xF0 == ((const U8*)s)[0] ) ? \
405 ( 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 )\
406: ( 0xF1 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xF3 ) ? \
407 ( LIKELY( ( ( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) && ( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\
408: 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 409
fed423a5
KW
410#endif /* EBCDIC vs ASCII */
411
412/* 2**UTF_ACCUMULATION_SHIFT - 1 */
413#define UTF_CONTINUATION_MASK ((U8) ((1U << UTF_ACCUMULATION_SHIFT) - 1))
414
7028aeba
KW
415/* Internal macro to be used only in this file to aid in constructing other
416 * publicly accessible macros.
417 * The number of bytes required to express this uv in UTF-8, for just those
418 * uv's requiring 2 through 6 bytes, as these are common to all platforms and
419 * word sizes. The number of bytes needed is given by the number of leading 1
420 * bits in the start byte. There are 32 start bytes that have 2 initial 1 bits
421 * (C0-DF); there are 16 that have 3 initial 1 bits (E0-EF); 8 that have 4
422 * initial 1 bits (F0-F8); 4 that have 5 initial 1 bits (F9-FB), and 2 that
423 * have 6 initial 1 bits (FC-FD). The largest number a string of n bytes can
424 * represent is (the number of possible start bytes for 'n')
425 * * (the number of possiblities for each start byte
426 * The latter in turn is
427 * 2 ** ( (how many continuation bytes there are)
428 * * (the number of bits of information each
429 * continuation byte holds))
430 *
431 * If we were on a platform where we could use a fast find first set bit
432 * instruction (or count leading zeros instruction) this could be replaced by
433 * using that to find the log2 of the uv, and divide that by the number of bits
434 * of information in each continuation byte, adjusting for large cases and how
435 * much information is in a start byte for that length */
72164d3a 436#define __COMMON_UNI_SKIP(uv) \
7028aeba
KW
437 (UV) (uv) < (32 * (1U << ( UTF_ACCUMULATION_SHIFT))) ? 2 : \
438 (UV) (uv) < (16 * (1U << (2 * UTF_ACCUMULATION_SHIFT))) ? 3 : \
439 (UV) (uv) < ( 8 * (1U << (3 * UTF_ACCUMULATION_SHIFT))) ? 4 : \
440 (UV) (uv) < ( 4 * (1U << (4 * UTF_ACCUMULATION_SHIFT))) ? 5 : \
441 (UV) (uv) < ( 2 * (1U << (5 * UTF_ACCUMULATION_SHIFT))) ? 6 :
72164d3a
KW
442
443/* Internal macro to be used only in this file.
444 * This adds to __COMMON_UNI_SKIP the details at this platform's upper range.
fed423a5 445 * For any-sized EBCDIC platforms, or 64-bit ASCII ones, we need one more test
72164d3a
KW
446 * to see if just 7 bytes is needed, or if the maximum is needed. For 32-bit
447 * ASCII platforms, everything is representable by 7 bytes */
fed423a5 448#if defined(UV_IS_QUAD) || defined(EBCDIC)
72164d3a 449# define __BASE_UNI_SKIP(uv) (__COMMON_UNI_SKIP(uv) \
7028aeba 450 (UV) (uv) < ((UV) 1U << (6 * UTF_ACCUMULATION_SHIFT)) ? 7 : UTF8_MAXBYTES)
1d68d6cd 451#else
72164d3a 452# define __BASE_UNI_SKIP(uv) (__COMMON_UNI_SKIP(uv) 7)
1d68d6cd
SC
453#endif
454
fed423a5
KW
455/* The next two macros use the base macro defined above, and add in the tests
456 * at the low-end of the range, for just 1 byte, yielding complete macros,
457 * publicly accessible. */
458
459/* Input is a true Unicode (not-native) code point */
460#define OFFUNISKIP(uv) (OFFUNI_IS_INVARIANT(uv) ? 1 : __BASE_UNI_SKIP(uv))
2084b489 461
5352a763
KW
462/*
463
464=for apidoc Am|STRLEN|UVCHR_SKIP|UV cp
465returns the number of bytes required to represent the code point C<cp> when
466encoded as UTF-8. C<cp> is a native (ASCII or EBCDIC) code point if less than
467255; a Unicode code point otherwise.
468
469=cut
470 */
fdb6583d 471#define UVCHR_SKIP(uv) ( UVCHR_IS_INVARIANT(uv) ? 1 : __BASE_UNI_SKIP(uv))
5352a763 472
b651802e
KW
473/* The largest code point representable by two UTF-8 bytes on this platform.
474 * As explained in the comments for __COMMON_UNI_SKIP, 32 start bytes with
fed423a5 475 * UTF_ACCUMULATION_SHIFT bits of information each */
aa206fb7
KW
476#define MAX_UTF8_TWO_BYTE (32 * (1U << UTF_ACCUMULATION_SHIFT) - 1)
477
b651802e
KW
478/* The largest code point representable by two UTF-8 bytes on any platform that
479 * Perl runs on. This value is constrained by EBCDIC which has 5 bits per
480 * continuation byte */
aa206fb7
KW
481#define MAX_PORTABLE_UTF8_TWO_BYTE (32 * (1U << 5) - 1)
482
c03c0950
KW
483/* The maximum number of UTF-8 bytes a single Unicode character can
484 * uppercase/lowercase/fold into. Unicode guarantees that the maximum
485 * expansion is UTF8_MAX_FOLD_CHAR_EXPAND characters, but any above-Unicode
486 * code point will fold to itself, so we only have to look at the expansion of
487 * the maximum Unicode code point. But this number may be less than the space
488 * occupied by a very large code point under Perl's extended UTF-8. We have to
489 * make it large enough to fit any single character. (It turns out that ASCII
490 * and EBCDIC differ in which is larger) */
491#define UTF8_MAXBYTES_CASE \
492 (UTF8_MAXBYTES >= (UTF8_MAX_FOLD_CHAR_EXPAND * OFFUNISKIP(0x10FFFF)) \
493 ? UTF8_MAXBYTES \
494 : (UTF8_MAX_FOLD_CHAR_EXPAND * OFFUNISKIP(0x10FFFF)))
495
d06134e5
KW
496/* Rest of these are attributes of Unicode and perl's internals rather than the
497 * encoding, or happen to be the same in both ASCII and EBCDIC (at least at
498 * this level; the macros that some of these call may have different
499 * definitions in the two encodings */
500
59a449d5
KW
501/* In domain restricted to ASCII, these may make more sense to the reader than
502 * the ones with Latin1 in the name */
503#define NATIVE_TO_ASCII(ch) NATIVE_TO_LATIN1(ch)
504#define ASCII_TO_NATIVE(ch) LATIN1_TO_NATIVE(ch)
505
506/* More or less misleadingly-named defines, retained for back compat */
507#define NATIVE_TO_UTF(ch) NATIVE_UTF8_TO_I8(ch)
508#define NATIVE_TO_I8(ch) NATIVE_UTF8_TO_I8(ch)
509#define UTF_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch)
510#define I8_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch)
511#define NATIVE8_TO_UNI(ch) NATIVE_TO_LATIN1(ch)
d06134e5 512
c0236afe 513/* This defines the 1-bits that are to be in the first byte of a multi-byte
97d0ceda
KW
514 * UTF-8 encoded character that mark it as a start byte and give the number of
515 * bytes that comprise the character. 'len' is the number of bytes in the
516 * multi-byte sequence. */
c0236afe
KW
517#define UTF_START_MARK(len) (((len) > 7) ? 0xFF : (0xFF & (0xFE << (7-(len)))))
518
519/* Masks out the initial one bits in a start byte, leaving the real data ones.
520 * Doesn't work on an invariant byte. 'len' is the number of bytes in the
521 * multi-byte sequence that comprises the character. */
522#define UTF_START_MASK(len) (((len) >= 7) ? 0x00 : (0x1F >> ((len)-2)))
523
537124e4
KW
524/* Adds a UTF8 continuation byte 'new' of information to a running total code
525 * point 'old' of all the continuation bytes so far. This is designed to be
155d2738
KW
526 * used in a loop to convert from UTF-8 to the code point represented. Note
527 * that this is asymmetric on EBCDIC platforms, in that the 'new' parameter is
528 * the UTF-EBCDIC byte, whereas the 'old' parameter is a Unicode (not EBCDIC)
529 * code point in process of being generated */
a6951642
KW
530#define UTF8_ACCUMULATE(old, new) (__ASSERT_(FITS_IN_8_BITS(new)) \
531 ((old) << UTF_ACCUMULATION_SHIFT) \
155d2738
KW
532 | ((NATIVE_UTF8_TO_I8((U8)new)) \
533 & UTF_CONTINUATION_MASK))
d06134e5 534
4ab10950 535/* This works in the face of malformed UTF-8. */
4e1ed312
KW
536#define UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(s, e) \
537 ( UTF8_IS_DOWNGRADEABLE_START(*(s)) \
538 && ( (e) - (s) > 1) \
539 && UTF8_IS_CONTINUATION(*((s)+1)))
4ab10950 540
5aaebcb3 541/* Number of bytes a code point occupies in UTF-8. */
5352a763 542#define NATIVE_SKIP(uv) UVCHR_SKIP(uv)
bd18bd40 543
5aaebcb3
KW
544/* Most code which says UNISKIP is really thinking in terms of native code
545 * points (0-255) plus all those beyond. This is an imprecise term, but having
2accb712 546 * it means existing code continues to work. For precision, use UVCHR_SKIP,
5352a763
KW
547 * NATIVE_SKIP, or OFFUNISKIP */
548#define UNISKIP(uv) UVCHR_SKIP(uv)
5aaebcb3 549
3c0792e4
KW
550/* Longer, but more accurate name */
551#define UTF8_IS_ABOVE_LATIN1_START(c) UTF8_IS_ABOVE_LATIN1(c)
552
a62b247b
KW
553/* Convert a UTF-8 variant Latin1 character to a native code point value.
554 * Needs just one iteration of accumulate. Should be used only if it is known
555 * that the code point is < 256, and is not UTF-8 invariant. Use the slower
556 * but more general TWO_BYTE_UTF8_TO_NATIVE() which handles any code point
557 * representable by two bytes (which turns out to be up through
558 * MAX_PORTABLE_UTF8_TWO_BYTE). The two parameters are:
559 * HI: a downgradable start byte;
560 * LO: continuation.
561 * */
562#define EIGHT_BIT_UTF8_TO_NATIVE(HI, LO) \
563 ( __ASSERT_(UTF8_IS_DOWNGRADEABLE_START(HI)) \
564 __ASSERT_(UTF8_IS_CONTINUATION(LO)) \
565 LATIN1_TO_NATIVE(UTF8_ACCUMULATE(( \
566 NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), (LO))))
567
94bb8c36 568/* Convert a two (not one) byte utf8 character to a native code point value.
2950f2a7
KW
569 * Needs just one iteration of accumulate. Should not be used unless it is
570 * known that the two bytes are legal: 1) two-byte start, and 2) continuation.
571 * Note that the result can be larger than 255 if the input character is not
572 * downgradable */
94bb8c36 573#define TWO_BYTE_UTF8_TO_NATIVE(HI, LO) \
a6951642
KW
574 (__ASSERT_(FITS_IN_8_BITS(HI)) \
575 __ASSERT_(FITS_IN_8_BITS(LO)) \
576 __ASSERT_(PL_utf8skip[HI] == 2) \
577 __ASSERT_(UTF8_IS_CONTINUATION(LO)) \
94bb8c36 578 UNI_TO_NATIVE(UTF8_ACCUMULATE((NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), \
635e76f5 579 (LO))))
94bb8c36
KW
580
581/* Should never be used, and be deprecated */
582#define TWO_BYTE_UTF8_TO_UNI(HI, LO) NATIVE_TO_UNI(TWO_BYTE_UTF8_TO_NATIVE(HI, LO))
2950f2a7 583
bd18bd40
KW
584/*
585
586=for apidoc Am|STRLEN|UTF8SKIP|char* s
587returns the number of bytes in the UTF-8 encoded character whose first (perhaps
588only) byte is pointed to by C<s>.
589
590=cut
591 */
2a70536e
KW
592#define UTF8SKIP(s) PL_utf8skip[*(const U8*)(s)]
593#define UTF8_SKIP(s) UTF8SKIP(s)
d06134e5 594
2d1545e5
KW
595/* Most code that says 'UNI_' really means the native value for code points up
596 * through 255 */
597#define UNI_IS_INVARIANT(cp) UVCHR_IS_INVARIANT(cp)
598
c2b32798
KW
599/*
600=for apidoc Am|bool|UTF8_IS_INVARIANT|char c
601
602Evaluates to 1 if the byte C<c> represents the same character when encoded in
603UTF-8 as when not; otherwise evaluates to 0. UTF-8 invariant characters can be
604copied as-is when converting to/from UTF-8, saving time.
605
606In spite of the name, this macro gives the correct result if the input string
607from which C<c> comes is not encoded in UTF-8.
608
609See C<L</UVCHR_IS_INVARIANT>> for checking if a UV is invariant.
610
611=cut
612
613The reason it works on both UTF-8 encoded strings and non-UTF-8 encoded, is
614that it returns TRUE in each for the exact same set of bit patterns. It is
615valid on a subset of what UVCHR_IS_INVARIANT is valid on, so can just use that;
616and the compiler should optimize out anything extraneous given the
617implementation of the latter. The |0 makes sure this isn't mistakenly called
618with a ptr argument.
619*/
5c06326b 620#define UTF8_IS_INVARIANT(c) UVCHR_IS_INVARIANT((c) | 0)
5fc230f1
KW
621
622/* Like the above, but its name implies a non-UTF8 input, which as the comments
623 * above show, doesn't matter as to its implementation */
38953e5a 624#define NATIVE_BYTE_IS_INVARIANT(c) UVCHR_IS_INVARIANT(c)
d06134e5 625
48ccf5e1
KW
626/* The macros in the next 4 sets are used to generate the two utf8 or utfebcdic
627 * bytes from an ordinal that is known to fit into exactly two (not one) bytes;
628 * it must be less than 0x3FF to work across both encodings. */
629
630/* These two are helper macros for the other three sets, and should not be used
631 * directly anywhere else. 'translate_function' is either NATIVE_TO_LATIN1
1ff3baa2
KW
632 * (which works for code points up through 0xFF) or NATIVE_TO_UNI which works
633 * for any code point */
48ccf5e1 634#define __BASE_TWO_BYTE_HI(c, translate_function) \
2863dafa 635 (__ASSERT_(! UVCHR_IS_INVARIANT(c)) \
48ccf5e1 636 I8_TO_NATIVE_UTF8((translate_function(c) >> UTF_ACCUMULATION_SHIFT) \
2863dafa 637 | UTF_START_MARK(2)))
48ccf5e1 638#define __BASE_TWO_BYTE_LO(c, translate_function) \
2863dafa 639 (__ASSERT_(! UVCHR_IS_INVARIANT(c)) \
48ccf5e1 640 I8_TO_NATIVE_UTF8((translate_function(c) & UTF_CONTINUATION_MASK) \
2863dafa 641 | UTF_CONTINUATION_MARK))
48ccf5e1 642
48ccf5e1
KW
643/* The next two macros should not be used. They were designed to be usable as
644 * the case label of a switch statement, but this doesn't work for EBCDIC. Use
9d0d3a03 645 * regen/unicode_constants.pl instead */
48ccf5e1
KW
646#define UTF8_TWO_BYTE_HI_nocast(c) __BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI)
647#define UTF8_TWO_BYTE_LO_nocast(c) __BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI)
648
649/* The next two macros are used when the source should be a single byte
650 * character; checked for under DEBUGGING */
651#define UTF8_EIGHT_BIT_HI(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
4c8cd605 652 ( __BASE_TWO_BYTE_HI(c, NATIVE_TO_LATIN1)))
48ccf5e1 653#define UTF8_EIGHT_BIT_LO(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
4c8cd605 654 (__BASE_TWO_BYTE_LO(c, NATIVE_TO_LATIN1)))
48ccf5e1
KW
655
656/* These final two macros in the series are used when the source can be any
657 * code point whose UTF-8 is known to occupy 2 bytes; they are less efficient
658 * than the EIGHT_BIT versions on EBCDIC platforms. We use the logical '~'
659 * operator instead of "<=" to avoid getting compiler warnings.
d52b8576 660 * MAX_UTF8_TWO_BYTE should be exactly all one bits in the lower few
48ccf5e1
KW
661 * places, so the ~ works */
662#define UTF8_TWO_BYTE_HI(c) \
663 (__ASSERT_((sizeof(c) == 1) \
d52b8576 664 || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \
4c8cd605 665 (__BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI)))
48ccf5e1
KW
666#define UTF8_TWO_BYTE_LO(c) \
667 (__ASSERT_((sizeof(c) == 1) \
d52b8576 668 || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \
4c8cd605 669 (__BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI)))
d06134e5 670
e7214ce8
KW
671/* This is illegal in any well-formed UTF-8 in both EBCDIC and ASCII
672 * as it is only in overlongs. */
673#define ILLEGAL_UTF8_BYTE I8_TO_NATIVE_UTF8(0xC1)
674
7e2040f0 675/*
e3036cf4 676 * 'UTF' is whether or not p is encoded in UTF8. The names 'foo_lazy_if' stem
20df05f4
KW
677 * from an earlier version of these macros in which they didn't call the
678 * foo_utf8() macros (i.e. were 'lazy') unless they decided that *p is the
679 * beginning of a utf8 character. Now that foo_utf8() determines that itself,
680 * no need to do it again here
7e2040f0 681 */
34aeb2e9
KW
682#define isIDFIRST_lazy_if(p,UTF) \
683 _is_utf8_FOO(_CC_IDFIRST, (const U8 *) p, "isIDFIRST_lazy_if", \
684 "isIDFIRST_lazy_if_safe", \
685 cBOOL(UTF && ! IN_BYTES), 0, __FILE__,__LINE__)
1d72bdf6 686
da8c1a98
KW
687#define isIDFIRST_lazy_if_safe(p, e, UTF) \
688 ((IN_BYTES || !UTF) \
689 ? isIDFIRST(*(p)) \
690 : isIDFIRST_utf8_safe(p, e))
691
34aeb2e9
KW
692#define isWORDCHAR_lazy_if(p,UTF) \
693 _is_utf8_FOO(_CC_IDFIRST, (const U8 *) p, "isWORDCHAR_lazy_if", \
694 "isWORDCHAR_lazy_if_safe", \
695 cBOOL(UTF && ! IN_BYTES), 0, __FILE__,__LINE__)
696
da8c1a98
KW
697#define isWORDCHAR_lazy_if_safe(p, e, UTF) \
698 ((IN_BYTES || !UTF) \
699 ? isWORDCHAR(*(p)) \
700 : isWORDCHAR_utf8_safe((U8 *) p, (U8 *) e))
701
34aeb2e9
KW
702#define isALNUM_lazy_if(p,UTF) \
703 _is_utf8_FOO(_CC_IDFIRST, (const U8 *) p, "isALNUM_lazy_if", \
704 "isWORDCHAR_lazy_if_safe", \
705 cBOOL(UTF && ! IN_BYTES), 0, __FILE__,__LINE__)
da8c1a98 706
89ebb4a3
JH
707#define UTF8_MAXLEN UTF8_MAXBYTES
708
8cb75cc8
KW
709/* A Unicode character can fold to up to 3 characters */
710#define UTF8_MAX_FOLD_CHAR_EXPAND 3
711
d3481830 712#define IN_BYTES UNLIKELY(CopHINTS_get(PL_curcop) & HINT_BYTES)
bd18bd40
KW
713
714/*
715
716=for apidoc Am|bool|DO_UTF8|SV* sv
717Returns a bool giving whether or not the PV in C<sv> is to be treated as being
718encoded in UTF-8.
719
720You should use this I<after> a call to C<SvPV()> or one of its variants, in
721case any call to string overloading updates the internal UTF-8 encoding flag.
722
723=cut
724*/
0064a8a9 725#define DO_UTF8(sv) (SvUTF8(sv) && !IN_BYTES)
1ff3baa2
KW
726
727/* Should all strings be treated as Unicode, and not just UTF-8 encoded ones?
728 * Is so within 'feature unicode_strings' or 'locale :not_characters', and not
729 * within 'use bytes'. UTF-8 locales are not tested for here, but perhaps
730 * could be */
70844984
KW
731#define IN_UNI_8_BIT \
732 (( ( (CopHINTS_get(PL_curcop) & HINT_UNI_8_BIT)) \
733 || ( CopHINTS_get(PL_curcop) & HINT_LOCALE_PARTIAL \
734 /* -1 below is for :not_characters */ \
735 && _is_in_locale_category(FALSE, -1))) \
736 && (! IN_BYTES))
b36bf33f 737
1d72bdf6 738
c76687c5 739#define UTF8_ALLOW_EMPTY 0x0001 /* Allow a zero length string */
2b5e7bc2 740#define UTF8_GOT_EMPTY UTF8_ALLOW_EMPTY
c76687c5
KW
741
742/* Allow first byte to be a continuation byte */
1d72bdf6 743#define UTF8_ALLOW_CONTINUATION 0x0002
2b5e7bc2 744#define UTF8_GOT_CONTINUATION UTF8_ALLOW_CONTINUATION
c76687c5 745
2b5e7bc2 746/* Unexpected continuation byte */
1d72bdf6 747#define UTF8_ALLOW_NON_CONTINUATION 0x0004
2b5e7bc2 748#define UTF8_GOT_NON_CONTINUATION UTF8_ALLOW_NON_CONTINUATION
949cf498
KW
749
750/* expecting more bytes than were available in the string */
751#define UTF8_ALLOW_SHORT 0x0008
2b5e7bc2 752#define UTF8_GOT_SHORT UTF8_ALLOW_SHORT
949cf498 753
94953955
KW
754/* Overlong sequence; i.e., the code point can be specified in fewer bytes.
755 * First one will convert the overlong to the REPLACEMENT CHARACTER; second
756 * will return what the overlong evaluates to */
949cf498 757#define UTF8_ALLOW_LONG 0x0010
94953955 758#define UTF8_ALLOW_LONG_AND_ITS_VALUE (UTF8_ALLOW_LONG|0x0020)
2b5e7bc2
KW
759#define UTF8_GOT_LONG UTF8_ALLOW_LONG
760
d60baaa7
KW
761#define UTF8_ALLOW_OVERFLOW 0x0080
762#define UTF8_GOT_OVERFLOW UTF8_ALLOW_OVERFLOW
949cf498 763
f180b292 764#define UTF8_DISALLOW_SURROGATE 0x0100 /* Unicode surrogates */
2b5e7bc2 765#define UTF8_GOT_SURROGATE UTF8_DISALLOW_SURROGATE
f180b292 766#define UTF8_WARN_SURROGATE 0x0200
949cf498 767
f180b292 768#define UTF8_DISALLOW_NONCHAR 0x0400 /* Unicode non-character */
2b5e7bc2 769#define UTF8_GOT_NONCHAR UTF8_DISALLOW_NONCHAR
f180b292 770#define UTF8_WARN_NONCHAR 0x0800 /* code points */
949cf498 771
f180b292 772#define UTF8_DISALLOW_SUPER 0x1000 /* Super-set of Unicode: code */
2b5e7bc2 773#define UTF8_GOT_SUPER UTF8_DISALLOW_SUPER
f180b292 774#define UTF8_WARN_SUPER 0x2000 /* points above the legal max */
949cf498 775
d35f2ca5
KW
776/* Code points which never were part of the original UTF-8 standard, which only
777 * went up to 2 ** 31 - 1. Note that these all overflow a signed 32-bit word,
778 * The first byte of these code points is FE or FF on ASCII platforms. If the
779 * first byte is FF, it will overflow a 32-bit word. */
f180b292 780#define UTF8_DISALLOW_ABOVE_31_BIT 0x4000
2b5e7bc2 781#define UTF8_GOT_ABOVE_31_BIT UTF8_DISALLOW_ABOVE_31_BIT
f180b292 782#define UTF8_WARN_ABOVE_31_BIT 0x8000
d35f2ca5
KW
783
784/* For back compat, these old names are misleading for UTF_EBCDIC */
785#define UTF8_DISALLOW_FE_FF UTF8_DISALLOW_ABOVE_31_BIT
786#define UTF8_WARN_FE_FF UTF8_WARN_ABOVE_31_BIT
949cf498 787
f180b292 788#define UTF8_CHECK_ONLY 0x10000
99a765e9 789#define _UTF8_NO_CONFIDENCE_IN_CURLEN 0x20000 /* Internal core use only */
949cf498
KW
790
791/* For backwards source compatibility. They do nothing, as the default now
792 * includes what they used to mean. The first one's meaning was to allow the
793 * just the single non-character 0xFFFF */
794#define UTF8_ALLOW_FFFF 0
795#define UTF8_ALLOW_SURROGATE 0
796
ecc1615f
KW
797/* C9 refers to Unicode Corrigendum #9: allows but discourages non-chars */
798#define UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE \
799 (UTF8_DISALLOW_SUPER|UTF8_DISALLOW_SURROGATE)
800#define UTF8_WARN_ILLEGAL_C9_INTERCHANGE (UTF8_WARN_SUPER|UTF8_WARN_SURROGATE)
801
d35f2ca5 802#define UTF8_DISALLOW_ILLEGAL_INTERCHANGE \
ecc1615f 803 (UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE|UTF8_DISALLOW_NONCHAR)
949cf498 804#define UTF8_WARN_ILLEGAL_INTERCHANGE \
ecc1615f
KW
805 (UTF8_WARN_ILLEGAL_C9_INTERCHANGE|UTF8_WARN_NONCHAR)
806
0eb3d6a0
KW
807/* This is typically used for code that processes UTF-8 input and doesn't want
808 * to have to deal with any malformations that might be present. All such will
809 * be safely replaced by the REPLACEMENT CHARACTER, unless other flags
810 * overriding this are also present. */
2d532c27
KW
811#define UTF8_ALLOW_ANY ( UTF8_ALLOW_CONTINUATION \
812 |UTF8_ALLOW_NON_CONTINUATION \
813 |UTF8_ALLOW_SHORT \
d60baaa7
KW
814 |UTF8_ALLOW_LONG \
815 |UTF8_ALLOW_OVERFLOW)
2d532c27
KW
816
817/* Accept any Perl-extended UTF-8 that evaluates to any UV on the platform, but
818 * not any malformed. This is the default. (Note that UVs above IV_MAX are
819 * deprecated. */
820#define UTF8_ALLOW_ANYUV 0
821#define UTF8_ALLOW_DEFAULT UTF8_ALLOW_ANYUV
1d72bdf6 822
89d986df
KW
823/*
824=for apidoc Am|bool|UTF8_IS_SURROGATE|const U8 *s|const U8 *e
825
826Evaluates to non-zero if the first few bytes of the string starting at C<s> and
827looking no further than S<C<e - 1>> are well-formed UTF-8 that represents one
828of the Unicode surrogate code points; otherwise it evaluates to 0. If
829non-zero, the value gives how many bytes starting at C<s> comprise the code
830point's representation.
831
832=cut
833 */
834#define UTF8_IS_SURROGATE(s, e) is_SURROGATE_utf8_safe(s, e)
835
836
837#define UTF8_IS_REPLACEMENT(s, send) is_REPLACEMENT_utf8_safe(s,send)
838
839/*
840=for apidoc Am|bool|UTF8_IS_SUPER|const U8 *s|const U8 *e
841
842Recall that Perl recognizes an extension to UTF-8 that can encode code
843points larger than the ones defined by Unicode, which are 0..0x10FFFF.
844
845This macro evaluates to non-zero if the first few bytes of the string starting
846at C<s> and looking no further than S<C<e - 1>> are from this UTF-8 extension;
847otherwise it evaluates to 0. If non-zero, the value gives how many bytes
848starting at C<s> comprise the code point's representation.
849
8500 is returned if the bytes are not well-formed extended UTF-8, or if they
851represent a code point that cannot fit in a UV on the current platform. Hence
852this macro can give different results when run on a 64-bit word machine than on
853one with a 32-bit word size.
0c58a72b 854
89d986df
KW
855Note that it is deprecated to have code points that are larger than what can
856fit in an IV on the current machine.
7131f24d 857
89d986df
KW
858=cut
859
860 * ASCII EBCDIC I8
7131f24d
KW
861 * U+10FFFF: \xF4\x8F\xBF\xBF \xF9\xA1\xBF\xBF\xBF max legal Unicode
862 * U+110000: \xF4\x90\x80\x80 \xF9\xA2\xA0\xA0\xA0
863 * U+110001: \xF4\x90\x80\x81 \xF9\xA2\xA0\xA0\xA1
89d986df
KW
864 */
865#ifdef EBCDIC
a14e0a36 866# define UTF8_IS_SUPER(s, e) \
89d986df
KW
867 (( LIKELY((e) > (s) + 4) \
868 && NATIVE_UTF8_TO_I8(*(s)) >= 0xF9 \
869 && ( NATIVE_UTF8_TO_I8(*(s)) > 0xF9 \
3d42f267 870 || (NATIVE_UTF8_TO_I8(*((s) + 1)) >= 0xA2)) \
89d986df 871 && LIKELY((s) + UTF8SKIP(s) <= (e))) \
2b479609 872 ? _is_utf8_char_helper(s, s + UTF8SKIP(s), 0) : 0)
7131f24d 873#else
a14e0a36 874# define UTF8_IS_SUPER(s, e) \
89d986df
KW
875 (( LIKELY((e) > (s) + 3) \
876 && (*(U8*) (s)) >= 0xF4 \
877 && ((*(U8*) (s)) > 0xF4 || (*((U8*) (s) + 1) >= 0x90))\
878 && LIKELY((s) + UTF8SKIP(s) <= (e))) \
2b479609 879 ? _is_utf8_char_helper(s, s + UTF8SKIP(s), 0) : 0)
7131f24d
KW
880#endif
881
b96a92fb
KW
882/* These are now machine generated, and the 'given' clause is no longer
883 * applicable */
0c58a72b 884#define UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s, e) \
89d986df
KW
885 cBOOL(is_NONCHAR_utf8_safe(s,e))
886
887/*
888=for apidoc Am|bool|UTF8_IS_NONCHAR|const U8 *s|const U8 *e
889
890Evaluates to non-zero if the first few bytes of the string starting at C<s> and
891looking no further than S<C<e - 1>> are well-formed UTF-8 that represents one
892of the Unicode non-character code points; otherwise it evaluates to 0. If
893non-zero, the value gives how many bytes starting at C<s> comprise the code
894point's representation.
895
896=cut
897 */
0c58a72b
KW
898#define UTF8_IS_NONCHAR(s, e) \
899 UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s, e)
7131f24d 900
c867b360
JH
901#define UNICODE_SURROGATE_FIRST 0xD800
902#define UNICODE_SURROGATE_LAST 0xDFFF
903#define UNICODE_REPLACEMENT 0xFFFD
904#define UNICODE_BYTE_ORDER_MARK 0xFEFF
1d72bdf6 905
b851fbc1 906/* Though our UTF-8 encoding can go beyond this,
c76687c5 907 * let's be conservative and do as Unicode says. */
b851fbc1
JH
908#define PERL_UNICODE_MAX 0x10FFFF
909
d35f2ca5
KW
910#define UNICODE_WARN_SURROGATE 0x0001 /* UTF-16 surrogates */
911#define UNICODE_WARN_NONCHAR 0x0002 /* Non-char code points */
912#define UNICODE_WARN_SUPER 0x0004 /* Above 0x10FFFF */
913#define UNICODE_WARN_ABOVE_31_BIT 0x0008 /* Above 0x7FFF_FFFF */
914#define UNICODE_DISALLOW_SURROGATE 0x0010
915#define UNICODE_DISALLOW_NONCHAR 0x0020
916#define UNICODE_DISALLOW_SUPER 0x0040
917#define UNICODE_DISALLOW_ABOVE_31_BIT 0x0080
ecc1615f
KW
918#define UNICODE_WARN_ILLEGAL_C9_INTERCHANGE \
919 (UNICODE_WARN_SURROGATE|UNICODE_WARN_SUPER)
bb88be5f 920#define UNICODE_WARN_ILLEGAL_INTERCHANGE \
ecc1615f
KW
921 (UNICODE_WARN_ILLEGAL_C9_INTERCHANGE|UNICODE_WARN_NONCHAR)
922#define UNICODE_DISALLOW_ILLEGAL_C9_INTERCHANGE \
923 (UNICODE_DISALLOW_SURROGATE|UNICODE_DISALLOW_SUPER)
bb88be5f 924#define UNICODE_DISALLOW_ILLEGAL_INTERCHANGE \
ecc1615f 925 (UNICODE_DISALLOW_ILLEGAL_C9_INTERCHANGE|UNICODE_DISALLOW_NONCHAR)
949cf498
KW
926
927/* For backward source compatibility, as are now the default */
928#define UNICODE_ALLOW_SURROGATE 0
929#define UNICODE_ALLOW_SUPER 0
930#define UNICODE_ALLOW_ANY 0
b851fbc1 931
2d6b3d38
KW
932/* This matches the 2048 code points between UNICODE_SURROGATE_FIRST (0xD800) and
933 * UNICODE_SURROGATE_LAST (0xDFFF) */
934#define UNICODE_IS_SURROGATE(uv) (((UV) (uv) & (~0xFFFF | 0xF800)) \
935 == 0xD800)
936
646d1759
KW
937#define UNICODE_IS_REPLACEMENT(uv) ((UV) (uv) == UNICODE_REPLACEMENT)
938#define UNICODE_IS_BYTE_ORDER_MARK(uv) ((UV) (uv) == UNICODE_BYTE_ORDER_MARK)
c149ab20
KW
939
940/* Is 'uv' one of the 32 contiguous-range noncharacters? */
941#define UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv) ((UV) (uv) >= 0xFDD0 \
942 && (UV) (uv) <= 0xFDEF)
943
944/* Is 'uv' one of the 34 plane-ending noncharacters 0xFFFE, 0xFFFF, 0x1FFFE,
945 * 0x1FFFF, ... 0x10FFFE, 0x10FFFF, given that we know that 'uv' is not above
946 * the Unicode legal max */
947#define UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv) \
948 (((UV) (uv) & 0xFFFE) == 0xFFFE)
949
950#define UNICODE_IS_NONCHAR(uv) \
951 ( UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv) \
952 || ( LIKELY( ! UNICODE_IS_SUPER(uv)) \
953 && UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv)))
954
955#define UNICODE_IS_SUPER(uv) ((UV) (uv) > PERL_UNICODE_MAX)
d35f2ca5 956#define UNICODE_IS_ABOVE_31_BIT(uv) ((UV) (uv) > 0x7FFFFFFF)
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
6302f837 998/*
5dca9278
KW
999
1000=for apidoc Am|STRLEN|isUTF8_CHAR|const U8 *s|const U8 *e
1001
35f8c9bd 1002Evaluates to non-zero if the first few bytes of the string starting at C<s> and
2c6ed66c
KW
1003looking no further than S<C<e - 1>> are well-formed UTF-8, as extended by Perl,
1004that represents some code point; otherwise it evaluates to 0. If non-zero, the
1005value gives how many bytes starting at C<s> comprise the code point's
2717076a
KW
1006representation. Any bytes remaining before C<e>, but beyond the ones needed to
1007form the first code point in C<s>, are not examined.
6302f837 1008
35f8c9bd
KW
1009The code point can be any that will fit in a UV on this machine, using Perl's
1010extension to official UTF-8 to represent those higher than the Unicode maximum
1011of 0x10FFFF. That means that this macro is used to efficiently decide if the
2717076a
KW
1012next few bytes in C<s> is legal UTF-8 for a single character.
1013
1014Use C<L</isSTRICT_UTF8_CHAR>> to restrict the acceptable code points to those
1015defined by Unicode to be fully interchangeable across applications;
1016C<L</isC9_STRICT_UTF8_CHAR>> to use the L<Unicode Corrigendum
1017#9|http://www.unicode.org/versions/corrigendum9.html> definition of allowable
1018code points; and C<L</isUTF8_CHAR_flags>> for a more customized definition.
1019
1020Use C<L</is_utf8_string>>, C<L</is_utf8_string_loc>>, and
1021C<L</is_utf8_string_loclen>> to check entire strings.
35f8c9bd
KW
1022
1023Note that it is deprecated to use code points higher than what will fit in an
1024IV. This macro does not raise any warnings for such code points, treating them
1025as valid.
1026
1027Note also that a UTF-8 INVARIANT character (i.e. ASCII on non-EBCDIC machines)
1028is a valid UTF-8 character.
5dca9278
KW
1029
1030=cut
1031*/
6302f837 1032
784d4f31
KW
1033#define isUTF8_CHAR(s, e) \
1034 (UNLIKELY((e) <= (s)) \
1035 ? 0 \
1036 : (UTF8_IS_INVARIANT(*s)) \
df863e43
KW
1037 ? 1 \
1038 : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \
1039 ? 0 \
1040 : LIKELY(NATIVE_UTF8_TO_I8(*s) <= _IS_UTF8_CHAR_HIGHEST_START_BYTE) \
1041 ? is_UTF8_CHAR_utf8_no_length_checks(s) \
1042 : _is_utf8_char_helper(s, e, 0))
6302f837 1043
3cedd9d9 1044#define is_utf8_char_buf(buf, buf_end) isUTF8_CHAR(buf, buf_end)
976c1b08
KW
1045#define bytes_from_utf8(s, lenp, is_utf8p) \
1046 bytes_from_utf8_loc(s, lenp, is_utf8p, 0)
3cedd9d9 1047
e23e8bc1
KW
1048/*
1049
1050=for apidoc Am|STRLEN|isSTRICT_UTF8_CHAR|const U8 *s|const U8 *e
1051
1052Evaluates to non-zero if the first few bytes of the string starting at C<s> and
1053looking no further than S<C<e - 1>> are well-formed UTF-8 that represents some
1054Unicode code point completely acceptable for open interchange between all
1055applications; otherwise it evaluates to 0. If non-zero, the value gives how
2717076a
KW
1056many bytes starting at C<s> comprise the code point's representation. Any
1057bytes remaining before C<e>, but beyond the ones needed to form the first code
1058point in C<s>, are not examined.
e23e8bc1
KW
1059
1060The largest acceptable code point is the Unicode maximum 0x10FFFF, and must not
1061be a surrogate nor a non-character code point. Thus this excludes any code
1062point from Perl's extended UTF-8.
1063
1064This is used to efficiently decide if the next few bytes in C<s> is
2717076a
KW
1065legal Unicode-acceptable UTF-8 for a single character.
1066
1067Use C<L</isC9_STRICT_UTF8_CHAR>> to use the L<Unicode Corrigendum
1068#9|http://www.unicode.org/versions/corrigendum9.html> definition of allowable
1069code points; C<L</isUTF8_CHAR>> to check for Perl's extended UTF-8;
1070and C<L</isUTF8_CHAR_flags>> for a more customized definition.
e23e8bc1 1071
9f2abfde
KW
1072Use C<L</is_strict_utf8_string>>, C<L</is_strict_utf8_string_loc>>, and
1073C<L</is_strict_utf8_string_loclen>> to check entire strings.
1074
e23e8bc1
KW
1075=cut
1076*/
1077
1078#define isSTRICT_UTF8_CHAR(s, e) \
1079 (UNLIKELY((e) <= (s)) \
1080 ? 0 \
1081 : (UTF8_IS_INVARIANT(*s)) \
1082 ? 1 \
1083 : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \
1084 ? 0 \
1085 : is_STRICT_UTF8_CHAR_utf8_no_length_checks(s))
1086
a82be82b
KW
1087/*
1088
1089=for apidoc Am|STRLEN|isC9_STRICT_UTF8_CHAR|const U8 *s|const U8 *e
1090
1091Evaluates to non-zero if the first few bytes of the string starting at C<s> and
1092looking no further than S<C<e - 1>> are well-formed UTF-8 that represents some
1093Unicode non-surrogate code point; otherwise it evaluates to 0. If non-zero,
1094the value gives how many bytes starting at C<s> comprise the code point's
2717076a
KW
1095representation. Any bytes remaining before C<e>, but beyond the ones needed to
1096form the first code point in C<s>, are not examined.
a82be82b
KW
1097
1098The largest acceptable code point is the Unicode maximum 0x10FFFF. This
1099differs from C<L</isSTRICT_UTF8_CHAR>> only in that it accepts non-character
1100code points. This corresponds to
1101L<Unicode Corrigendum #9|http://www.unicode.org/versions/corrigendum9.html>.
1102which said that non-character code points are merely discouraged rather than
1103completely forbidden in open interchange. See
1104L<perlunicode/Noncharacter code points>.
1105
2717076a
KW
1106Use C<L</isUTF8_CHAR>> to check for Perl's extended UTF-8; and
1107C<L</isUTF8_CHAR_flags>> for a more customized definition.
1108
9f2abfde
KW
1109Use C<L</is_c9strict_utf8_string>>, C<L</is_c9strict_utf8_string_loc>>, and
1110C<L</is_c9strict_utf8_string_loclen>> to check entire strings.
1111
a82be82b
KW
1112=cut
1113*/
1114
1115#define isC9_STRICT_UTF8_CHAR(s, e) \
1116 (UNLIKELY((e) <= (s)) \
1117 ? 0 \
1118 : (UTF8_IS_INVARIANT(*s)) \
1119 ? 1 \
1120 : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \
1121 ? 0 \
1122 : is_C9_STRICT_UTF8_CHAR_utf8_no_length_checks(s))
1123
25e3a4e0
KW
1124/*
1125
1126=for apidoc Am|STRLEN|isUTF8_CHAR_flags|const U8 *s|const U8 *e| const U32 flags
1127
1128Evaluates to non-zero if the first few bytes of the string starting at C<s> and
1129looking no further than S<C<e - 1>> are well-formed UTF-8, as extended by Perl,
1130that represents some code point, subject to the restrictions given by C<flags>;
1131otherwise it evaluates to 0. If non-zero, the value gives how many bytes
2717076a
KW
1132starting at C<s> comprise the code point's representation. Any bytes remaining
1133before C<e>, but beyond the ones needed to form the first code point in C<s>,
1134are not examined.
25e3a4e0
KW
1135
1136If C<flags> is 0, this gives the same results as C<L</isUTF8_CHAR>>;
1137if C<flags> is C<UTF8_DISALLOW_ILLEGAL_INTERCHANGE>, this gives the same results
1138as C<L</isSTRICT_UTF8_CHAR>>;
1139and if C<flags> is C<UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE>, this gives
1140the same results as C<L</isC9_STRICT_UTF8_CHAR>>.
1141Otherwise C<flags> may be any combination of the C<UTF8_DISALLOW_I<foo>> flags
1142understood by C<L</utf8n_to_uvchr>>, with the same meanings.
1143
1144The three alternative macros are for the most commonly needed validations; they
1145are likely to run somewhat faster than this more general one, as they can be
1146inlined into your code.
1147
9f2abfde
KW
1148Use L</is_utf8_string_flags>, L</is_utf8_string_loc_flags>, and
1149L</is_utf8_string_loclen_flags> to check entire strings.
1150
25e3a4e0
KW
1151=cut
1152*/
1153
1154#define isUTF8_CHAR_flags(s, e, flags) \
1155 (UNLIKELY((e) <= (s)) \
1156 ? 0 \
1157 : (UTF8_IS_INVARIANT(*s)) \
1158 ? 1 \
1159 : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \
1160 ? 0 \
1161 : _is_utf8_char_helper(s, e, flags))
1162
6302f837
KW
1163/* Do not use; should be deprecated. Use isUTF8_CHAR() instead; this is
1164 * retained solely for backwards compatibility */
1165#define IS_UTF8_CHAR(p, n) (isUTF8_CHAR(p, (p) + (n)) == n)
e9a8c099 1166
6a5bc5ac 1167#endif /* PERL_UTF8_H_ */
57f0e7e2 1168
e9a8c099 1169/*
14d04a33 1170 * ex: set ts=8 sts=4 sw=4 et:
e9a8c099 1171 */