This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Deprecate isFOO_utf8() macros
[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
57f0e7e2
KW
17#ifndef H_UTF8 /* Guard against recursive inclusion */
18#define H_UTF8 1
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)
445bf929
KW
79#define to_utf8_fold(c, p, lenp) _to_utf8_fold_flags(c, p, lenp, FOLD_FLAGS_FULL)
80#define to_utf8_lower(a,b,c) _to_utf8_lower_flags(a,b,c,0)
81#define to_utf8_upper(a,b,c) _to_utf8_upper_flags(a,b,c,0)
82#define to_utf8_title(a,b,c) _to_utf8_title_flags(a,b,c,0)
36bb2ab6 83
eda9cac1
KW
84#define foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \
85 foldEQ_utf8_flags(s1, pe1, l1, u1, s2, pe2, l2, u2, 0)
baa60164 86#define FOLDEQ_UTF8_NOMIX_ASCII (1 << 0)
cea315b6 87#define FOLDEQ_LOCALE (1 << 1)
18f762c3
KW
88#define FOLDEQ_S1_ALREADY_FOLDED (1 << 2)
89#define FOLDEQ_S2_ALREADY_FOLDED (1 << 3)
d635b710
KW
90#define FOLDEQ_S1_FOLDS_SANE (1 << 4)
91#define FOLDEQ_S2_FOLDS_SANE (1 << 5)
a33c29bc 92
e6226b18
KW
93#define ibcmp_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \
94 cBOOL(! foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2))
95
1d72bdf6
NIS
96#ifdef EBCDIC
97/* The equivalent of these macros but implementing UTF-EBCDIC
98 are in the following header file:
99 */
100
101#include "utfebcdic.h"
fd7cb289 102
d06134e5 103#else /* ! EBCDIC */
73c4f7a1
GS
104START_EXTERN_C
105
111e8ed9
KW
106/* How wide can a single UTF-8 encoded character become in bytes. */
107/* NOTE: Strictly speaking Perl's UTF-8 should not be called UTF-8 since UTF-8
108 * is an encoding of Unicode, and Unicode's upper limit, 0x10FFFF, can be
109 * expressed with 4 bytes. However, Perl thinks of UTF-8 as a way to encode
110 * non-negative integers in a binary format, even those above Unicode */
111#define UTF8_MAXBYTES 13
112
a0ed51b3 113#ifdef DOINIT
6f06b55f 114EXTCONST unsigned char PL_utf8skip[] = {
b2635aa8
KW
115/* 0x00 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
116/* 0x10 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
117/* 0x20 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
118/* 0x30 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
119/* 0x40 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
120/* 0x50 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
121/* 0x60 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
122/* 0x70 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
123/* 0x80 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
124/* 0x90 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
125/* 0xA0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
126/* 0xB0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
127/* 0xC0 */ 2,2, /* overlong */
1ff3baa2 128/* 0xC2 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0080 to U+03FF */
b2635aa8
KW
129/* 0xD0 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0400 to U+07FF */
130/* 0xE0 */ 3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3, /* U+0800 to U+FFFF */
131/* 0xF0 */ 4,4,4,4,4,4,4,4,5,5,5,5,6,6, /* above BMP to 2**31 - 1 */
6937f885
KW
132 /* Perl extended (never was official UTF-8). Up to 36 bit */
133/* 0xFE */ 7,
134 /* More extended, Up to 72 bits (64-bit + reserved) */
111e8ed9 135/* 0xFF */ UTF8_MAXBYTES
a0ed51b3
LW
136};
137#else
6f06b55f 138EXTCONST unsigned char PL_utf8skip[];
a0ed51b3
LW
139#endif
140
73c4f7a1 141END_EXTERN_C
7e2040f0 142
1a3756de 143#if defined(_MSC_VER) && _MSC_VER < 1400
6f6d1bab
TC
144/* older MSVC versions have a smallish macro buffer */
145#define PERL_SMALL_MACRO_BUFFER
146#endif
147
59a449d5
KW
148/* Native character to/from iso-8859-1. Are the identity functions on ASCII
149 * platforms */
6f6d1bab 150#ifdef PERL_SMALL_MACRO_BUFFER
1a3756de
TC
151#define NATIVE_TO_LATIN1(ch) ((U8)(ch))
152#define LATIN1_TO_NATIVE(ch) ((U8)(ch))
6f6d1bab 153#else
4c8cd605
KW
154#define NATIVE_TO_LATIN1(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch)))
155#define LATIN1_TO_NATIVE(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch)))
6f6d1bab 156#endif
59a449d5
KW
157
158/* I8 is an intermediate version of UTF-8 used only in UTF-EBCDIC. We thus
159 * consider it to be identical to UTF-8 on ASCII platforms. Strictly speaking
160 * UTF-8 and UTF-EBCDIC are two different things, but we often conflate them
161 * because they are 8-bit encodings that serve the same purpose in Perl, and
162 * rarely do we need to distinguish them. The term "NATIVE_UTF8" applies to
163 * whichever one is applicable on the current platform */
6f6d1bab 164#ifdef PERL_SMALL_MACRO_BUFFER
1d4ea287
DD
165#define NATIVE_UTF8_TO_I8(ch) ((U8) (ch))
166#define I8_TO_NATIVE_UTF8(ch) ((U8) (ch))
6f6d1bab 167#else
4c8cd605
KW
168#define NATIVE_UTF8_TO_I8(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch)))
169#define I8_TO_NATIVE_UTF8(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch)))
6f6d1bab 170#endif
59a449d5 171
1d72bdf6 172/* Transforms in wide UV chars */
4c8cd605
KW
173#define UNI_TO_NATIVE(ch) ((UV) (ch))
174#define NATIVE_TO_UNI(ch) ((UV) (ch))
d7578b48 175
877d9f0d 176/*
9041c2e3 177
a14e0a36
KW
178 The following table is from Unicode 3.2, plus the Perl extensions for above
179 U+10FFFF
877d9f0d 180
a14e0a36 181 Code Points 1st Byte 2nd Byte 3rd 4th 5th 6th 7th 8th-13th
877d9f0d 182
375122d7 183 U+0000..U+007F 00..7F
e1b711da 184 U+0080..U+07FF * C2..DF 80..BF
a14e0a36
KW
185 U+0800..U+0FFF E0 * A0..BF 80..BF
186 U+1000..U+CFFF E1..EC 80..BF 80..BF
187 U+D000..U+D7FF ED 80..9F 80..BF
188 U+D800..U+DFFF ED A0..BF 80..BF (surrogates)
189 U+E000..U+FFFF EE..EF 80..BF 80..BF
190 U+10000..U+3FFFF F0 * 90..BF 80..BF 80..BF
191 U+40000..U+FFFFF F1..F3 80..BF 80..BF 80..BF
192 U+100000..U+10FFFF F4 80..8F 80..BF 80..BF
193 Below are above-Unicode code points
194 U+110000..U+13FFFF F4 90..BF 80..BF 80..BF
195 U+110000..U+1FFFFF F5..F7 80..BF 80..BF 80..BF
196 U+200000..U+FFFFFF F8 * 88..BF 80..BF 80..BF 80..BF
197U+1000000..U+3FFFFFF F9..FB 80..BF 80..BF 80..BF 80..BF
198U+4000000..U+3FFFFFFF FC * 84..BF 80..BF 80..BF 80..BF 80..BF
199U+40000000..U+7FFFFFFF FD 80..BF 80..BF 80..BF 80..BF 80..BF
200U+80000000..U+FFFFFFFFF FE * 82..BF 80..BF 80..BF 80..BF 80..BF 80..BF
201U+1000000000.. FF 80..BF 80..BF 80..BF 80..BF 80..BF * 81..BF 80..BF
877d9f0d 202
e1b711da 203Note the gaps before several of the byte entries above marked by '*'. These are
37e2e78e
KW
204caused by legal UTF-8 avoiding non-shortest encodings: it is technically
205possible to UTF-8-encode a single code point in different ways, but that is
206explicitly forbidden, and the shortest possible encoding should always be used
15824458 207(and that is what Perl does). The non-shortest ones are called 'overlongs'.
8c007b5a 208
877d9f0d
JH
209 */
210
8c007b5a
JH
211/*
212 Another way to look at it, as bits:
213
b2635aa8 214 Code Points 1st Byte 2nd Byte 3rd Byte 4th Byte
8c007b5a 215
b2635aa8
KW
216 0aaa aaaa 0aaa aaaa
217 0000 0bbb bbaa aaaa 110b bbbb 10aa aaaa
218 cccc bbbb bbaa aaaa 1110 cccc 10bb bbbb 10aa aaaa
219 00 000d ddcc cccc bbbb bbaa aaaa 1111 0ddd 10cc cccc 10bb bbbb 10aa aaaa
8c007b5a
JH
220
221As you can see, the continuation bytes all begin with C<10>, and the
e1b711da 222leading bits of the start byte tell how many bytes there are in the
8c007b5a
JH
223encoded character.
224
df863e43
KW
225Perl's extended UTF-8 means we can have start bytes up through FF, though any
226beginning with FF yields a code point that is too large for 32-bit ASCII
227platforms. FF signals to use 13 bytes for the encoded character. This breaks
228the paradigm that the number of leading bits gives how many total bytes there
229are in the character.
65ab9279 230
8c007b5a
JH
231*/
232
eda91ad7
KW
233/* Anything larger than this will overflow the word if it were converted into a UV */
234#if defined(UV_IS_QUAD)
235# define HIGHEST_REPRESENTABLE_UTF8 "\xFF\x80\x8F\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF"
236#else
237# define HIGHEST_REPRESENTABLE_UTF8 "\xFE\x83\xBF\xBF\xBF\xBF\xBF"
238#endif
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 */
264#define UTF8_IS_CONTINUED(c) (((U8)((c) | 0)) & UTF_CONTINUATION_MARK)
15824458
KW
265
266/* Is the byte 'c' the first byte of a multi-byte UTF8-8 encoded sequence?
267 * This doesn't catch invariants (they are single-byte). It also excludes the
5c06326b
KW
268 * illegal overlong sequences that begin with C0 and C1. The |0 makes sure
269 * this isn't mistakenly called with a ptr argument */
270#define UTF8_IS_START(c) (((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 */
858cd8ab 278#define UTF8_IS_CONTINUATION(c) \
5c06326b 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 */
286#define UTF8_IS_DOWNGRADEABLE_START(c) ((((U8)((c) | 0)) & 0xfe) == 0xc2)
4ab10950 287
15824458 288/* Is the UTF8-encoded byte 'c' the first byte of a sequence of bytes that
5c06326b
KW
289 * represent a code point > 255? The |0 makes sure this isn't mistakenly
290 * called with a ptr argument */
291#define UTF8_IS_ABOVE_LATIN1(c) (((U8)((c) | 0)) >= 0xc4)
8850bf83 292
15824458
KW
293/* This is the number of low-order bits a continuation byte in a UTF-8 encoded
294 * sequence contributes to the specification of the code point. In the bit
295 * maps above, you see that the first 2 bits are a constant '10', leaving 6 of
296 * real information */
1d72bdf6 297#define UTF_ACCUMULATION_SHIFT 6
b2635aa8 298
fed423a5
KW
299/* ^? is defined to be DEL on ASCII systems. See the definition of toCTRL()
300 * for more */
301#define QUESTION_MARK_CTRL DEL_NATIVE
302
303/* Surrogates, non-character code points and above-Unicode code points are
304 * problematic in some contexts. This allows code that needs to check for
305 * those to to quickly exclude the vast majority of code points it will
306 * encounter */
307#define isUTF8_POSSIBLY_PROBLEMATIC(c) ((U8) c >= 0xED)
308
784d4f31
KW
309/* A helper macro for isUTF8_CHAR, so use that one instead of this. This was
310 * generated by regen/regcharclass.pl, and then moved here. Then it was
311 * hand-edited to add some LIKELY() calls, presuming that malformations are
312 * unlikely. The lines that generated it were then commented out. This was
313 * done because it takes on the order of 10 minutes to generate, and is never
314 * going to change, unless the generated code is improved, and figuring out
315 * the LIKELYs there would be hard.
316 *
317 UTF8_CHAR: Matches legal UTF-8 variant code points up through 0x1FFFFFF
318
319 0x80 - 0x1FFFFF
320*/
321/*** GENERATED CODE ***/
322#define is_UTF8_CHAR_utf8_no_length_checks(s) \
9f2eed98
P
323( ( 0xC2 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xDF ) ? \
324 ( LIKELY( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) ? 2 : 0 ) \
325: ( 0xE0 == ((const U8*)s)[0] ) ? \
326 ( LIKELY( ( ( ((const U8*)s)[1] & 0xE0 ) == 0xA0 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
327: ( 0xE1 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xEF ) ? \
328 ( LIKELY( ( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
329: ( 0xF0 == ((const U8*)s)[0] ) ? \
330 ( 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 )\
331: ( ( ( ( 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
332
333/* The above macro handles UTF-8 that has this start byte as the maximum */
334#define _IS_UTF8_CHAR_HIGHEST_START_BYTE 0xF7
335
e23e8bc1
KW
336/* A helper macro for isSTRICT_UTF8_CHAR, so use that one instead of this.
337 * Like is_UTF8_CHAR_utf8_no_length_checks(), this was moved here and LIKELYs
338 * added manually.
339 *
340 STRICT_UTF8_CHAR: Matches legal Unicode UTF-8 variant code points, no
341 surrrogates nor non-character code points
342*/
343/*** GENERATED CODE ***/
344#define is_STRICT_UTF8_CHAR_utf8_no_length_checks(s) \
9f2eed98
P
345( ( 0xC2 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xDF ) ? \
346 ( LIKELY( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) ? 2 : 0 ) \
347: ( 0xE0 == ((const U8*)s)[0] ) ? \
348 ( LIKELY( ( ( ((const U8*)s)[1] & 0xE0 ) == 0xA0 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
349: ( ( 0xE1 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xEC ) || 0xEE == ((const U8*)s)[0] ) ?\
350 ( ( ( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
351: ( 0xED == ((const U8*)s)[0] ) ? \
352 ( LIKELY( ( ( ((const U8*)s)[1] & 0xE0 ) == 0x80 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
353: ( 0xEF == ((const U8*)s)[0] ) ? \
354 ( ( ( 0x80 <= ((const U8*)s)[1] && ((const U8*)s)[1] <= 0xB6 ) || ( 0xB8 <= ((const U8*)s)[1] && ((const U8*)s)[1] <= 0xBE ) ) ?\
355 ( LIKELY( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ? 3 : 0 ) \
356 : ( 0xB7 == ((const U8*)s)[1] ) ? \
357 ( LIKELY( ( ((const U8*)s)[2] & 0xF0 ) == 0x80 || ( ((const U8*)s)[2] & 0xF0 ) == 0xB0 ) ? 3 : 0 )\
358 : ( ( 0xBF == ((const U8*)s)[1] ) && ( 0x80 <= ((const U8*)s)[2] && ((const U8*)s)[2] <= 0xBD ) ) ? 3 : 0 )\
359: ( 0xF0 == ((const U8*)s)[0] ) ? \
360 ( ( ( 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 ) ) ?\
361 ( LIKELY( ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\
362 : ( ((const U8*)s)[1] == 0x9F || ( ( ((const U8*)s)[1] & 0xEF ) == 0xAF ) ) ? \
363 ( ( 0x80 <= ((const U8*)s)[2] && ((const U8*)s)[2] <= 0xBE ) ? \
364 ( LIKELY( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ? 4 : 0 ) \
365 : LIKELY( ( 0xBF == ((const U8*)s)[2] ) && ( 0x80 <= ((const U8*)s)[3] && ((const U8*)s)[3] <= 0xBD ) ) ? 4 : 0 )\
e23e8bc1 366 : 0 ) \
9f2eed98
P
367: ( 0xF1 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xF3 ) ? \
368 ( ( ( ( ((const U8*)s)[1] & 0xC8 ) == 0x80 ) || ( ( ((const U8*)s)[1] & 0xCC ) == 0x88 ) || ( ( ((const U8*)s)[1] & 0xCE ) == 0x8C ) || ( ( ((const U8*)s)[1] & 0xCF ) == 0x8E ) ) ?\
369 ( LIKELY( ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\
370 : ( ( ((const U8*)s)[1] & 0xCF ) == 0x8F ) ? \
371 ( ( 0x80 <= ((const U8*)s)[2] && ((const U8*)s)[2] <= 0xBE ) ? \
372 ( LIKELY( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ? 4 : 0 ) \
373 : LIKELY( ( 0xBF == ((const U8*)s)[2] ) && ( 0x80 <= ((const U8*)s)[3] && ((const U8*)s)[3] <= 0xBD ) ) ? 4 : 0 )\
e23e8bc1 374 : 0 ) \
9f2eed98
P
375: ( 0xF4 == ((const U8*)s)[0] ) ? \
376 ( ( 0x80 <= ((const U8*)s)[1] && ((const U8*)s)[1] <= 0x8E ) ? \
377 ( LIKELY( ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\
378 : ( 0x8F == ((const U8*)s)[1] ) ? \
379 ( ( 0x80 <= ((const U8*)s)[2] && ((const U8*)s)[2] <= 0xBE ) ? \
380 ( LIKELY( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ? 4 : 0 ) \
381 : LIKELY( ( 0xBF == ((const U8*)s)[2] ) && ( 0x80 <= ((const U8*)s)[3] && ((const U8*)s)[3] <= 0xBD ) ) ? 4 : 0 )\
e23e8bc1
KW
382 : 0 ) \
383: 0 )
384
a82be82b
KW
385/* Similarly,
386 C9_STRICT_UTF8_CHAR: Matches legal Unicode UTF-8 variant code
387 points, no surrogates
388 0x0080 - 0xD7FF
389 0xE000 - 0x10FFFF
390*/
391/*** GENERATED CODE ***/
392#define is_C9_STRICT_UTF8_CHAR_utf8_no_length_checks(s) \
9f2eed98
P
393( ( 0xC2 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xDF ) ? \
394 ( LIKELY( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) ? 2 : 0 ) \
395: ( 0xE0 == ((const U8*)s)[0] ) ? \
396 ( LIKELY( ( ( ((const U8*)s)[1] & 0xE0 ) == 0xA0 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
397: ( ( 0xE1 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xEC ) || ( ((const U8*)s)[0] & 0xFE ) == 0xEE ) ?\
398 ( LIKELY( ( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
399: ( 0xED == ((const U8*)s)[0] ) ? \
400 ( LIKELY( ( ( ((const U8*)s)[1] & 0xE0 ) == 0x80 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
401: ( 0xF0 == ((const U8*)s)[0] ) ? \
402 ( 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 )\
403: ( 0xF1 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xF3 ) ? \
404 ( LIKELY( ( ( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) && ( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\
405: 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 406
fed423a5
KW
407#endif /* EBCDIC vs ASCII */
408
409/* 2**UTF_ACCUMULATION_SHIFT - 1 */
410#define UTF_CONTINUATION_MASK ((U8) ((1U << UTF_ACCUMULATION_SHIFT) - 1))
411
7028aeba
KW
412/* Internal macro to be used only in this file to aid in constructing other
413 * publicly accessible macros.
414 * The number of bytes required to express this uv in UTF-8, for just those
415 * uv's requiring 2 through 6 bytes, as these are common to all platforms and
416 * word sizes. The number of bytes needed is given by the number of leading 1
417 * bits in the start byte. There are 32 start bytes that have 2 initial 1 bits
418 * (C0-DF); there are 16 that have 3 initial 1 bits (E0-EF); 8 that have 4
419 * initial 1 bits (F0-F8); 4 that have 5 initial 1 bits (F9-FB), and 2 that
420 * have 6 initial 1 bits (FC-FD). The largest number a string of n bytes can
421 * represent is (the number of possible start bytes for 'n')
422 * * (the number of possiblities for each start byte
423 * The latter in turn is
424 * 2 ** ( (how many continuation bytes there are)
425 * * (the number of bits of information each
426 * continuation byte holds))
427 *
428 * If we were on a platform where we could use a fast find first set bit
429 * instruction (or count leading zeros instruction) this could be replaced by
430 * using that to find the log2 of the uv, and divide that by the number of bits
431 * of information in each continuation byte, adjusting for large cases and how
432 * much information is in a start byte for that length */
72164d3a 433#define __COMMON_UNI_SKIP(uv) \
7028aeba
KW
434 (UV) (uv) < (32 * (1U << ( UTF_ACCUMULATION_SHIFT))) ? 2 : \
435 (UV) (uv) < (16 * (1U << (2 * UTF_ACCUMULATION_SHIFT))) ? 3 : \
436 (UV) (uv) < ( 8 * (1U << (3 * UTF_ACCUMULATION_SHIFT))) ? 4 : \
437 (UV) (uv) < ( 4 * (1U << (4 * UTF_ACCUMULATION_SHIFT))) ? 5 : \
438 (UV) (uv) < ( 2 * (1U << (5 * UTF_ACCUMULATION_SHIFT))) ? 6 :
72164d3a
KW
439
440/* Internal macro to be used only in this file.
441 * This adds to __COMMON_UNI_SKIP the details at this platform's upper range.
fed423a5 442 * For any-sized EBCDIC platforms, or 64-bit ASCII ones, we need one more test
72164d3a
KW
443 * to see if just 7 bytes is needed, or if the maximum is needed. For 32-bit
444 * ASCII platforms, everything is representable by 7 bytes */
fed423a5 445#if defined(UV_IS_QUAD) || defined(EBCDIC)
72164d3a 446# define __BASE_UNI_SKIP(uv) (__COMMON_UNI_SKIP(uv) \
7028aeba 447 (UV) (uv) < ((UV) 1U << (6 * UTF_ACCUMULATION_SHIFT)) ? 7 : UTF8_MAXBYTES)
1d68d6cd 448#else
72164d3a 449# define __BASE_UNI_SKIP(uv) (__COMMON_UNI_SKIP(uv) 7)
1d68d6cd
SC
450#endif
451
fed423a5
KW
452/* The next two macros use the base macro defined above, and add in the tests
453 * at the low-end of the range, for just 1 byte, yielding complete macros,
454 * publicly accessible. */
455
456/* Input is a true Unicode (not-native) code point */
457#define OFFUNISKIP(uv) (OFFUNI_IS_INVARIANT(uv) ? 1 : __BASE_UNI_SKIP(uv))
2084b489 458
5352a763
KW
459/*
460
461=for apidoc Am|STRLEN|UVCHR_SKIP|UV cp
462returns the number of bytes required to represent the code point C<cp> when
463encoded as UTF-8. C<cp> is a native (ASCII or EBCDIC) code point if less than
464255; a Unicode code point otherwise.
465
466=cut
467 */
fdb6583d 468#define UVCHR_SKIP(uv) ( UVCHR_IS_INVARIANT(uv) ? 1 : __BASE_UNI_SKIP(uv))
5352a763 469
b651802e
KW
470/* The largest code point representable by two UTF-8 bytes on this platform.
471 * As explained in the comments for __COMMON_UNI_SKIP, 32 start bytes with
fed423a5 472 * UTF_ACCUMULATION_SHIFT bits of information each */
aa206fb7
KW
473#define MAX_UTF8_TWO_BYTE (32 * (1U << UTF_ACCUMULATION_SHIFT) - 1)
474
b651802e
KW
475/* The largest code point representable by two UTF-8 bytes on any platform that
476 * Perl runs on. This value is constrained by EBCDIC which has 5 bits per
477 * continuation byte */
aa206fb7
KW
478#define MAX_PORTABLE_UTF8_TWO_BYTE (32 * (1U << 5) - 1)
479
c03c0950
KW
480/* The maximum number of UTF-8 bytes a single Unicode character can
481 * uppercase/lowercase/fold into. Unicode guarantees that the maximum
482 * expansion is UTF8_MAX_FOLD_CHAR_EXPAND characters, but any above-Unicode
483 * code point will fold to itself, so we only have to look at the expansion of
484 * the maximum Unicode code point. But this number may be less than the space
485 * occupied by a very large code point under Perl's extended UTF-8. We have to
486 * make it large enough to fit any single character. (It turns out that ASCII
487 * and EBCDIC differ in which is larger) */
488#define UTF8_MAXBYTES_CASE \
489 (UTF8_MAXBYTES >= (UTF8_MAX_FOLD_CHAR_EXPAND * OFFUNISKIP(0x10FFFF)) \
490 ? UTF8_MAXBYTES \
491 : (UTF8_MAX_FOLD_CHAR_EXPAND * OFFUNISKIP(0x10FFFF)))
492
d06134e5
KW
493/* Rest of these are attributes of Unicode and perl's internals rather than the
494 * encoding, or happen to be the same in both ASCII and EBCDIC (at least at
495 * this level; the macros that some of these call may have different
496 * definitions in the two encodings */
497
59a449d5
KW
498/* In domain restricted to ASCII, these may make more sense to the reader than
499 * the ones with Latin1 in the name */
500#define NATIVE_TO_ASCII(ch) NATIVE_TO_LATIN1(ch)
501#define ASCII_TO_NATIVE(ch) LATIN1_TO_NATIVE(ch)
502
503/* More or less misleadingly-named defines, retained for back compat */
504#define NATIVE_TO_UTF(ch) NATIVE_UTF8_TO_I8(ch)
505#define NATIVE_TO_I8(ch) NATIVE_UTF8_TO_I8(ch)
506#define UTF_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch)
507#define I8_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch)
508#define NATIVE8_TO_UNI(ch) NATIVE_TO_LATIN1(ch)
d06134e5 509
c0236afe 510/* This defines the 1-bits that are to be in the first byte of a multi-byte
97d0ceda
KW
511 * UTF-8 encoded character that mark it as a start byte and give the number of
512 * bytes that comprise the character. 'len' is the number of bytes in the
513 * multi-byte sequence. */
c0236afe
KW
514#define UTF_START_MARK(len) (((len) > 7) ? 0xFF : (0xFF & (0xFE << (7-(len)))))
515
516/* Masks out the initial one bits in a start byte, leaving the real data ones.
517 * Doesn't work on an invariant byte. 'len' is the number of bytes in the
518 * multi-byte sequence that comprises the character. */
519#define UTF_START_MASK(len) (((len) >= 7) ? 0x00 : (0x1F >> ((len)-2)))
520
537124e4
KW
521/* Adds a UTF8 continuation byte 'new' of information to a running total code
522 * point 'old' of all the continuation bytes so far. This is designed to be
155d2738
KW
523 * used in a loop to convert from UTF-8 to the code point represented. Note
524 * that this is asymmetric on EBCDIC platforms, in that the 'new' parameter is
525 * the UTF-EBCDIC byte, whereas the 'old' parameter is a Unicode (not EBCDIC)
526 * code point in process of being generated */
527#define UTF8_ACCUMULATE(old, new) (((old) << UTF_ACCUMULATION_SHIFT) \
528 | ((NATIVE_UTF8_TO_I8((U8)new)) \
529 & UTF_CONTINUATION_MASK))
d06134e5 530
4ab10950
KW
531/* This works in the face of malformed UTF-8. */
532#define UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(s, e) (UTF8_IS_DOWNGRADEABLE_START(*s) \
533 && ( (e) - (s) > 1) \
534 && UTF8_IS_CONTINUATION(*((s)+1)))
535
5aaebcb3 536/* Number of bytes a code point occupies in UTF-8. */
5352a763 537#define NATIVE_SKIP(uv) UVCHR_SKIP(uv)
bd18bd40 538
5aaebcb3
KW
539/* Most code which says UNISKIP is really thinking in terms of native code
540 * points (0-255) plus all those beyond. This is an imprecise term, but having
2accb712 541 * it means existing code continues to work. For precision, use UVCHR_SKIP,
5352a763
KW
542 * NATIVE_SKIP, or OFFUNISKIP */
543#define UNISKIP(uv) UVCHR_SKIP(uv)
5aaebcb3 544
3c0792e4
KW
545/* Longer, but more accurate name */
546#define UTF8_IS_ABOVE_LATIN1_START(c) UTF8_IS_ABOVE_LATIN1(c)
547
a62b247b
KW
548/* Convert a UTF-8 variant Latin1 character to a native code point value.
549 * Needs just one iteration of accumulate. Should be used only if it is known
550 * that the code point is < 256, and is not UTF-8 invariant. Use the slower
551 * but more general TWO_BYTE_UTF8_TO_NATIVE() which handles any code point
552 * representable by two bytes (which turns out to be up through
553 * MAX_PORTABLE_UTF8_TWO_BYTE). The two parameters are:
554 * HI: a downgradable start byte;
555 * LO: continuation.
556 * */
557#define EIGHT_BIT_UTF8_TO_NATIVE(HI, LO) \
558 ( __ASSERT_(UTF8_IS_DOWNGRADEABLE_START(HI)) \
559 __ASSERT_(UTF8_IS_CONTINUATION(LO)) \
560 LATIN1_TO_NATIVE(UTF8_ACCUMULATE(( \
561 NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), (LO))))
562
94bb8c36 563/* Convert a two (not one) byte utf8 character to a native code point value.
2950f2a7
KW
564 * Needs just one iteration of accumulate. Should not be used unless it is
565 * known that the two bytes are legal: 1) two-byte start, and 2) continuation.
566 * Note that the result can be larger than 255 if the input character is not
567 * downgradable */
94bb8c36 568#define TWO_BYTE_UTF8_TO_NATIVE(HI, LO) \
b18b15ad 569 ( __ASSERT_(PL_utf8skip[HI] == 2) \
635e76f5 570 __ASSERT_(UTF8_IS_CONTINUATION(LO)) \
94bb8c36 571 UNI_TO_NATIVE(UTF8_ACCUMULATE((NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), \
635e76f5 572 (LO))))
94bb8c36
KW
573
574/* Should never be used, and be deprecated */
575#define TWO_BYTE_UTF8_TO_UNI(HI, LO) NATIVE_TO_UNI(TWO_BYTE_UTF8_TO_NATIVE(HI, LO))
2950f2a7 576
bd18bd40
KW
577/*
578
579=for apidoc Am|STRLEN|UTF8SKIP|char* s
580returns the number of bytes in the UTF-8 encoded character whose first (perhaps
581only) byte is pointed to by C<s>.
582
583=cut
584 */
2a70536e
KW
585#define UTF8SKIP(s) PL_utf8skip[*(const U8*)(s)]
586#define UTF8_SKIP(s) UTF8SKIP(s)
d06134e5 587
2d1545e5
KW
588/* Most code that says 'UNI_' really means the native value for code points up
589 * through 255 */
590#define UNI_IS_INVARIANT(cp) UVCHR_IS_INVARIANT(cp)
591
c2b32798
KW
592/*
593=for apidoc Am|bool|UTF8_IS_INVARIANT|char c
594
595Evaluates to 1 if the byte C<c> represents the same character when encoded in
596UTF-8 as when not; otherwise evaluates to 0. UTF-8 invariant characters can be
597copied as-is when converting to/from UTF-8, saving time.
598
599In spite of the name, this macro gives the correct result if the input string
600from which C<c> comes is not encoded in UTF-8.
601
602See C<L</UVCHR_IS_INVARIANT>> for checking if a UV is invariant.
603
604=cut
605
606The reason it works on both UTF-8 encoded strings and non-UTF-8 encoded, is
607that it returns TRUE in each for the exact same set of bit patterns. It is
608valid on a subset of what UVCHR_IS_INVARIANT is valid on, so can just use that;
609and the compiler should optimize out anything extraneous given the
610implementation of the latter. The |0 makes sure this isn't mistakenly called
611with a ptr argument.
612*/
5c06326b 613#define UTF8_IS_INVARIANT(c) UVCHR_IS_INVARIANT((c) | 0)
5fc230f1
KW
614
615/* Like the above, but its name implies a non-UTF8 input, which as the comments
616 * above show, doesn't matter as to its implementation */
38953e5a 617#define NATIVE_BYTE_IS_INVARIANT(c) UVCHR_IS_INVARIANT(c)
d06134e5 618
48ccf5e1
KW
619/* The macros in the next 4 sets are used to generate the two utf8 or utfebcdic
620 * bytes from an ordinal that is known to fit into exactly two (not one) bytes;
621 * it must be less than 0x3FF to work across both encodings. */
622
623/* These two are helper macros for the other three sets, and should not be used
624 * directly anywhere else. 'translate_function' is either NATIVE_TO_LATIN1
1ff3baa2
KW
625 * (which works for code points up through 0xFF) or NATIVE_TO_UNI which works
626 * for any code point */
48ccf5e1 627#define __BASE_TWO_BYTE_HI(c, translate_function) \
2863dafa 628 (__ASSERT_(! UVCHR_IS_INVARIANT(c)) \
48ccf5e1 629 I8_TO_NATIVE_UTF8((translate_function(c) >> UTF_ACCUMULATION_SHIFT) \
2863dafa 630 | UTF_START_MARK(2)))
48ccf5e1 631#define __BASE_TWO_BYTE_LO(c, translate_function) \
2863dafa 632 (__ASSERT_(! UVCHR_IS_INVARIANT(c)) \
48ccf5e1 633 I8_TO_NATIVE_UTF8((translate_function(c) & UTF_CONTINUATION_MASK) \
2863dafa 634 | UTF_CONTINUATION_MARK))
48ccf5e1 635
48ccf5e1
KW
636/* The next two macros should not be used. They were designed to be usable as
637 * the case label of a switch statement, but this doesn't work for EBCDIC. Use
9d0d3a03 638 * regen/unicode_constants.pl instead */
48ccf5e1
KW
639#define UTF8_TWO_BYTE_HI_nocast(c) __BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI)
640#define UTF8_TWO_BYTE_LO_nocast(c) __BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI)
641
642/* The next two macros are used when the source should be a single byte
643 * character; checked for under DEBUGGING */
644#define UTF8_EIGHT_BIT_HI(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
4c8cd605 645 ( __BASE_TWO_BYTE_HI(c, NATIVE_TO_LATIN1)))
48ccf5e1 646#define UTF8_EIGHT_BIT_LO(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
4c8cd605 647 (__BASE_TWO_BYTE_LO(c, NATIVE_TO_LATIN1)))
48ccf5e1
KW
648
649/* These final two macros in the series are used when the source can be any
650 * code point whose UTF-8 is known to occupy 2 bytes; they are less efficient
651 * than the EIGHT_BIT versions on EBCDIC platforms. We use the logical '~'
652 * operator instead of "<=" to avoid getting compiler warnings.
d52b8576 653 * MAX_UTF8_TWO_BYTE should be exactly all one bits in the lower few
48ccf5e1
KW
654 * places, so the ~ works */
655#define UTF8_TWO_BYTE_HI(c) \
656 (__ASSERT_((sizeof(c) == 1) \
d52b8576 657 || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \
4c8cd605 658 (__BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI)))
48ccf5e1
KW
659#define UTF8_TWO_BYTE_LO(c) \
660 (__ASSERT_((sizeof(c) == 1) \
d52b8576 661 || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \
4c8cd605 662 (__BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI)))
d06134e5 663
e7214ce8
KW
664/* This is illegal in any well-formed UTF-8 in both EBCDIC and ASCII
665 * as it is only in overlongs. */
666#define ILLEGAL_UTF8_BYTE I8_TO_NATIVE_UTF8(0xC1)
667
7e2040f0 668/*
e3036cf4 669 * 'UTF' is whether or not p is encoded in UTF8. The names 'foo_lazy_if' stem
20df05f4
KW
670 * from an earlier version of these macros in which they didn't call the
671 * foo_utf8() macros (i.e. were 'lazy') unless they decided that *p is the
672 * beginning of a utf8 character. Now that foo_utf8() determines that itself,
673 * no need to do it again here
7e2040f0 674 */
34aeb2e9
KW
675#define isIDFIRST_lazy_if(p,UTF) \
676 _is_utf8_FOO(_CC_IDFIRST, (const U8 *) p, "isIDFIRST_lazy_if", \
677 "isIDFIRST_lazy_if_safe", \
678 cBOOL(UTF && ! IN_BYTES), 0, __FILE__,__LINE__)
1d72bdf6 679
da8c1a98
KW
680#define isIDFIRST_lazy_if_safe(p, e, UTF) \
681 ((IN_BYTES || !UTF) \
682 ? isIDFIRST(*(p)) \
683 : isIDFIRST_utf8_safe(p, e))
684
34aeb2e9
KW
685#define isWORDCHAR_lazy_if(p,UTF) \
686 _is_utf8_FOO(_CC_IDFIRST, (const U8 *) p, "isWORDCHAR_lazy_if", \
687 "isWORDCHAR_lazy_if_safe", \
688 cBOOL(UTF && ! IN_BYTES), 0, __FILE__,__LINE__)
689
da8c1a98
KW
690#define isWORDCHAR_lazy_if_safe(p, e, UTF) \
691 ((IN_BYTES || !UTF) \
692 ? isWORDCHAR(*(p)) \
693 : isWORDCHAR_utf8_safe((U8 *) p, (U8 *) e))
694
34aeb2e9
KW
695#define isALNUM_lazy_if(p,UTF) \
696 _is_utf8_FOO(_CC_IDFIRST, (const U8 *) p, "isALNUM_lazy_if", \
697 "isWORDCHAR_lazy_if_safe", \
698 cBOOL(UTF && ! IN_BYTES), 0, __FILE__,__LINE__)
da8c1a98 699
89ebb4a3
JH
700#define UTF8_MAXLEN UTF8_MAXBYTES
701
8cb75cc8
KW
702/* A Unicode character can fold to up to 3 characters */
703#define UTF8_MAX_FOLD_CHAR_EXPAND 3
704
a98fe34d 705#define IN_BYTES (CopHINTS_get(PL_curcop) & HINT_BYTES)
bd18bd40
KW
706
707/*
708
709=for apidoc Am|bool|DO_UTF8|SV* sv
710Returns a bool giving whether or not the PV in C<sv> is to be treated as being
711encoded in UTF-8.
712
713You should use this I<after> a call to C<SvPV()> or one of its variants, in
714case any call to string overloading updates the internal UTF-8 encoding flag.
715
716=cut
717*/
0064a8a9 718#define DO_UTF8(sv) (SvUTF8(sv) && !IN_BYTES)
1ff3baa2
KW
719
720/* Should all strings be treated as Unicode, and not just UTF-8 encoded ones?
721 * Is so within 'feature unicode_strings' or 'locale :not_characters', and not
722 * within 'use bytes'. UTF-8 locales are not tested for here, but perhaps
723 * could be */
724#define IN_UNI_8_BIT \
d6ded950
KW
725 (((CopHINTS_get(PL_curcop) & (HINT_UNI_8_BIT)) \
726 || (CopHINTS_get(PL_curcop) & HINT_LOCALE_PARTIAL \
727 /* -1 below is for :not_characters */ \
728 && _is_in_locale_category(FALSE, -1))) \
729 && ! IN_BYTES)
b36bf33f 730
1d72bdf6 731
c76687c5 732#define UTF8_ALLOW_EMPTY 0x0001 /* Allow a zero length string */
2b5e7bc2 733#define UTF8_GOT_EMPTY UTF8_ALLOW_EMPTY
c76687c5
KW
734
735/* Allow first byte to be a continuation byte */
1d72bdf6 736#define UTF8_ALLOW_CONTINUATION 0x0002
2b5e7bc2 737#define UTF8_GOT_CONTINUATION UTF8_ALLOW_CONTINUATION
c76687c5 738
2b5e7bc2 739/* Unexpected continuation byte */
1d72bdf6 740#define UTF8_ALLOW_NON_CONTINUATION 0x0004
2b5e7bc2 741#define UTF8_GOT_NON_CONTINUATION UTF8_ALLOW_NON_CONTINUATION
949cf498
KW
742
743/* expecting more bytes than were available in the string */
744#define UTF8_ALLOW_SHORT 0x0008
2b5e7bc2 745#define UTF8_GOT_SHORT UTF8_ALLOW_SHORT
949cf498 746
94953955
KW
747/* Overlong sequence; i.e., the code point can be specified in fewer bytes.
748 * First one will convert the overlong to the REPLACEMENT CHARACTER; second
749 * will return what the overlong evaluates to */
949cf498 750#define UTF8_ALLOW_LONG 0x0010
94953955 751#define UTF8_ALLOW_LONG_AND_ITS_VALUE (UTF8_ALLOW_LONG|0x0020)
2b5e7bc2
KW
752#define UTF8_GOT_LONG UTF8_ALLOW_LONG
753
d60baaa7
KW
754#define UTF8_ALLOW_OVERFLOW 0x0080
755#define UTF8_GOT_OVERFLOW UTF8_ALLOW_OVERFLOW
949cf498 756
f180b292 757#define UTF8_DISALLOW_SURROGATE 0x0100 /* Unicode surrogates */
2b5e7bc2 758#define UTF8_GOT_SURROGATE UTF8_DISALLOW_SURROGATE
f180b292 759#define UTF8_WARN_SURROGATE 0x0200
949cf498 760
f180b292 761#define UTF8_DISALLOW_NONCHAR 0x0400 /* Unicode non-character */
2b5e7bc2 762#define UTF8_GOT_NONCHAR UTF8_DISALLOW_NONCHAR
f180b292 763#define UTF8_WARN_NONCHAR 0x0800 /* code points */
949cf498 764
f180b292 765#define UTF8_DISALLOW_SUPER 0x1000 /* Super-set of Unicode: code */
2b5e7bc2 766#define UTF8_GOT_SUPER UTF8_DISALLOW_SUPER
f180b292 767#define UTF8_WARN_SUPER 0x2000 /* points above the legal max */
949cf498 768
d35f2ca5
KW
769/* Code points which never were part of the original UTF-8 standard, which only
770 * went up to 2 ** 31 - 1. Note that these all overflow a signed 32-bit word,
771 * The first byte of these code points is FE or FF on ASCII platforms. If the
772 * first byte is FF, it will overflow a 32-bit word. */
f180b292 773#define UTF8_DISALLOW_ABOVE_31_BIT 0x4000
2b5e7bc2 774#define UTF8_GOT_ABOVE_31_BIT UTF8_DISALLOW_ABOVE_31_BIT
f180b292 775#define UTF8_WARN_ABOVE_31_BIT 0x8000
d35f2ca5
KW
776
777/* For back compat, these old names are misleading for UTF_EBCDIC */
778#define UTF8_DISALLOW_FE_FF UTF8_DISALLOW_ABOVE_31_BIT
779#define UTF8_WARN_FE_FF UTF8_WARN_ABOVE_31_BIT
949cf498 780
f180b292 781#define UTF8_CHECK_ONLY 0x10000
949cf498
KW
782
783/* For backwards source compatibility. They do nothing, as the default now
784 * includes what they used to mean. The first one's meaning was to allow the
785 * just the single non-character 0xFFFF */
786#define UTF8_ALLOW_FFFF 0
787#define UTF8_ALLOW_SURROGATE 0
788
ecc1615f
KW
789/* C9 refers to Unicode Corrigendum #9: allows but discourages non-chars */
790#define UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE \
791 (UTF8_DISALLOW_SUPER|UTF8_DISALLOW_SURROGATE)
792#define UTF8_WARN_ILLEGAL_C9_INTERCHANGE (UTF8_WARN_SUPER|UTF8_WARN_SURROGATE)
793
d35f2ca5 794#define UTF8_DISALLOW_ILLEGAL_INTERCHANGE \
ecc1615f 795 (UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE|UTF8_DISALLOW_NONCHAR)
949cf498 796#define UTF8_WARN_ILLEGAL_INTERCHANGE \
ecc1615f
KW
797 (UTF8_WARN_ILLEGAL_C9_INTERCHANGE|UTF8_WARN_NONCHAR)
798
d60baaa7
KW
799/* This is used typically for code that is willing to accept inputs of
800 * illformed UTF-8 sequences, for whatever reason. However, all such sequences
801 * evaluate to the REPLACEMENT CHARACTER unless other flags overriding this are
802 * also present. */
2d532c27
KW
803#define UTF8_ALLOW_ANY ( UTF8_ALLOW_CONTINUATION \
804 |UTF8_ALLOW_NON_CONTINUATION \
805 |UTF8_ALLOW_SHORT \
d60baaa7
KW
806 |UTF8_ALLOW_LONG \
807 |UTF8_ALLOW_OVERFLOW)
2d532c27
KW
808
809/* Accept any Perl-extended UTF-8 that evaluates to any UV on the platform, but
810 * not any malformed. This is the default. (Note that UVs above IV_MAX are
811 * deprecated. */
812#define UTF8_ALLOW_ANYUV 0
813#define UTF8_ALLOW_DEFAULT UTF8_ALLOW_ANYUV
1d72bdf6 814
89d986df
KW
815/*
816=for apidoc Am|bool|UTF8_IS_SURROGATE|const U8 *s|const U8 *e
817
818Evaluates to non-zero if the first few bytes of the string starting at C<s> and
819looking no further than S<C<e - 1>> are well-formed UTF-8 that represents one
820of the Unicode surrogate code points; otherwise it evaluates to 0. If
821non-zero, the value gives how many bytes starting at C<s> comprise the code
822point's representation.
823
824=cut
825 */
826#define UTF8_IS_SURROGATE(s, e) is_SURROGATE_utf8_safe(s, e)
827
828
829#define UTF8_IS_REPLACEMENT(s, send) is_REPLACEMENT_utf8_safe(s,send)
830
831/*
832=for apidoc Am|bool|UTF8_IS_SUPER|const U8 *s|const U8 *e
833
834Recall that Perl recognizes an extension to UTF-8 that can encode code
835points larger than the ones defined by Unicode, which are 0..0x10FFFF.
836
837This macro evaluates to non-zero if the first few bytes of the string starting
838at C<s> and looking no further than S<C<e - 1>> are from this UTF-8 extension;
839otherwise it evaluates to 0. If non-zero, the value gives how many bytes
840starting at C<s> comprise the code point's representation.
841
8420 is returned if the bytes are not well-formed extended UTF-8, or if they
843represent a code point that cannot fit in a UV on the current platform. Hence
844this macro can give different results when run on a 64-bit word machine than on
845one with a 32-bit word size.
0c58a72b 846
89d986df
KW
847Note that it is deprecated to have code points that are larger than what can
848fit in an IV on the current machine.
7131f24d 849
89d986df
KW
850=cut
851
852 * ASCII EBCDIC I8
7131f24d
KW
853 * U+10FFFF: \xF4\x8F\xBF\xBF \xF9\xA1\xBF\xBF\xBF max legal Unicode
854 * U+110000: \xF4\x90\x80\x80 \xF9\xA2\xA0\xA0\xA0
855 * U+110001: \xF4\x90\x80\x81 \xF9\xA2\xA0\xA0\xA1
89d986df
KW
856 */
857#ifdef EBCDIC
a14e0a36 858# define UTF8_IS_SUPER(s, e) \
89d986df
KW
859 (( LIKELY((e) > (s) + 4) \
860 && NATIVE_UTF8_TO_I8(*(s)) >= 0xF9 \
861 && ( NATIVE_UTF8_TO_I8(*(s)) > 0xF9 \
3d42f267 862 || (NATIVE_UTF8_TO_I8(*((s) + 1)) >= 0xA2)) \
89d986df 863 && LIKELY((s) + UTF8SKIP(s) <= (e))) \
2b479609 864 ? _is_utf8_char_helper(s, s + UTF8SKIP(s), 0) : 0)
7131f24d 865#else
a14e0a36 866# define UTF8_IS_SUPER(s, e) \
89d986df
KW
867 (( LIKELY((e) > (s) + 3) \
868 && (*(U8*) (s)) >= 0xF4 \
869 && ((*(U8*) (s)) > 0xF4 || (*((U8*) (s) + 1) >= 0x90))\
870 && LIKELY((s) + UTF8SKIP(s) <= (e))) \
2b479609 871 ? _is_utf8_char_helper(s, s + UTF8SKIP(s), 0) : 0)
7131f24d
KW
872#endif
873
b96a92fb
KW
874/* These are now machine generated, and the 'given' clause is no longer
875 * applicable */
0c58a72b 876#define UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s, e) \
89d986df
KW
877 cBOOL(is_NONCHAR_utf8_safe(s,e))
878
879/*
880=for apidoc Am|bool|UTF8_IS_NONCHAR|const U8 *s|const U8 *e
881
882Evaluates to non-zero if the first few bytes of the string starting at C<s> and
883looking no further than S<C<e - 1>> are well-formed UTF-8 that represents one
884of the Unicode non-character code points; otherwise it evaluates to 0. If
885non-zero, the value gives how many bytes starting at C<s> comprise the code
886point's representation.
887
888=cut
889 */
0c58a72b
KW
890#define UTF8_IS_NONCHAR(s, e) \
891 UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s, e)
7131f24d 892
c867b360
JH
893#define UNICODE_SURROGATE_FIRST 0xD800
894#define UNICODE_SURROGATE_LAST 0xDFFF
895#define UNICODE_REPLACEMENT 0xFFFD
896#define UNICODE_BYTE_ORDER_MARK 0xFEFF
1d72bdf6 897
b851fbc1 898/* Though our UTF-8 encoding can go beyond this,
c76687c5 899 * let's be conservative and do as Unicode says. */
b851fbc1
JH
900#define PERL_UNICODE_MAX 0x10FFFF
901
d35f2ca5
KW
902#define UNICODE_WARN_SURROGATE 0x0001 /* UTF-16 surrogates */
903#define UNICODE_WARN_NONCHAR 0x0002 /* Non-char code points */
904#define UNICODE_WARN_SUPER 0x0004 /* Above 0x10FFFF */
905#define UNICODE_WARN_ABOVE_31_BIT 0x0008 /* Above 0x7FFF_FFFF */
906#define UNICODE_DISALLOW_SURROGATE 0x0010
907#define UNICODE_DISALLOW_NONCHAR 0x0020
908#define UNICODE_DISALLOW_SUPER 0x0040
909#define UNICODE_DISALLOW_ABOVE_31_BIT 0x0080
ecc1615f
KW
910#define UNICODE_WARN_ILLEGAL_C9_INTERCHANGE \
911 (UNICODE_WARN_SURROGATE|UNICODE_WARN_SUPER)
bb88be5f 912#define UNICODE_WARN_ILLEGAL_INTERCHANGE \
ecc1615f
KW
913 (UNICODE_WARN_ILLEGAL_C9_INTERCHANGE|UNICODE_WARN_NONCHAR)
914#define UNICODE_DISALLOW_ILLEGAL_C9_INTERCHANGE \
915 (UNICODE_DISALLOW_SURROGATE|UNICODE_DISALLOW_SUPER)
bb88be5f 916#define UNICODE_DISALLOW_ILLEGAL_INTERCHANGE \
ecc1615f 917 (UNICODE_DISALLOW_ILLEGAL_C9_INTERCHANGE|UNICODE_DISALLOW_NONCHAR)
949cf498
KW
918
919/* For backward source compatibility, as are now the default */
920#define UNICODE_ALLOW_SURROGATE 0
921#define UNICODE_ALLOW_SUPER 0
922#define UNICODE_ALLOW_ANY 0
b851fbc1 923
2d6b3d38
KW
924/* This matches the 2048 code points between UNICODE_SURROGATE_FIRST (0xD800) and
925 * UNICODE_SURROGATE_LAST (0xDFFF) */
926#define UNICODE_IS_SURROGATE(uv) (((UV) (uv) & (~0xFFFF | 0xF800)) \
927 == 0xD800)
928
646d1759
KW
929#define UNICODE_IS_REPLACEMENT(uv) ((UV) (uv) == UNICODE_REPLACEMENT)
930#define UNICODE_IS_BYTE_ORDER_MARK(uv) ((UV) (uv) == UNICODE_BYTE_ORDER_MARK)
c149ab20
KW
931
932/* Is 'uv' one of the 32 contiguous-range noncharacters? */
933#define UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv) ((UV) (uv) >= 0xFDD0 \
934 && (UV) (uv) <= 0xFDEF)
935
936/* Is 'uv' one of the 34 plane-ending noncharacters 0xFFFE, 0xFFFF, 0x1FFFE,
937 * 0x1FFFF, ... 0x10FFFE, 0x10FFFF, given that we know that 'uv' is not above
938 * the Unicode legal max */
939#define UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv) \
940 (((UV) (uv) & 0xFFFE) == 0xFFFE)
941
942#define UNICODE_IS_NONCHAR(uv) \
943 ( UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv) \
944 || ( LIKELY( ! UNICODE_IS_SUPER(uv)) \
945 && UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv)))
946
947#define UNICODE_IS_SUPER(uv) ((UV) (uv) > PERL_UNICODE_MAX)
d35f2ca5 948#define UNICODE_IS_ABOVE_31_BIT(uv) ((UV) (uv) > 0x7FFFFFFF)
1d72bdf6 949
ec34087a
KW
950#define LATIN_SMALL_LETTER_SHARP_S LATIN_SMALL_LETTER_SHARP_S_NATIVE
951#define LATIN_SMALL_LETTER_Y_WITH_DIAERESIS \
952 LATIN_SMALL_LETTER_Y_WITH_DIAERESIS_NATIVE
953#define MICRO_SIGN MICRO_SIGN_NATIVE
954#define LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE \
955 LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE_NATIVE
956#define LATIN_SMALL_LETTER_A_WITH_RING_ABOVE \
957 LATIN_SMALL_LETTER_A_WITH_RING_ABOVE_NATIVE
09091399
JH
958#define UNICODE_GREEK_CAPITAL_LETTER_SIGMA 0x03A3
959#define UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA 0x03C2
960#define UNICODE_GREEK_SMALL_LETTER_SIGMA 0x03C3
9dcbe121 961#define GREEK_SMALL_LETTER_MU 0x03BC
9e682c18
KW
962#define GREEK_CAPITAL_LETTER_MU 0x039C /* Upper and title case
963 of MICRON */
964#define LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS 0x0178 /* Also is title case */
0766489e
KW
965#ifdef LATIN_CAPITAL_LETTER_SHARP_S_UTF8
966# define LATIN_CAPITAL_LETTER_SHARP_S 0x1E9E
967#endif
74894415
KW
968#define LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE 0x130
969#define LATIN_SMALL_LETTER_DOTLESS_I 0x131
9e682c18 970#define LATIN_SMALL_LETTER_LONG_S 0x017F
a9f50d33
KW
971#define LATIN_SMALL_LIGATURE_LONG_S_T 0xFB05
972#define LATIN_SMALL_LIGATURE_ST 0xFB06
9e682c18
KW
973#define KELVIN_SIGN 0x212A
974#define ANGSTROM_SIGN 0x212B
09091399 975
9e55ce06 976#define UNI_DISPLAY_ISPRINT 0x0001
c728cb41
JH
977#define UNI_DISPLAY_BACKSLASH 0x0002
978#define UNI_DISPLAY_QQ (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH)
979#define UNI_DISPLAY_REGEX (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH)
9e55ce06 980
5cd46e1f
KW
981#define ANYOF_FOLD_SHARP_S(node, input, end) \
982 (ANYOF_BITMAP_TEST(node, LATIN_SMALL_LETTER_SHARP_S) && \
137165a6 983 (ANYOF_NONBITMAP(node)) && \
39065660 984 (ANYOF_FLAGS(node) & ANYOF_LOC_NONBITMAP_FOLD) && \
07b6858f 985 ((end) > (input) + 1) && \
305b8651 986 isALPHA_FOLD_EQ((input)[0], 's'))
6302f837 987
ebc501f0 988#define SHARP_S_SKIP 2
3b0fc154 989
6302f837 990/*
5dca9278
KW
991
992=for apidoc Am|STRLEN|isUTF8_CHAR|const U8 *s|const U8 *e
993
35f8c9bd 994Evaluates to non-zero if the first few bytes of the string starting at C<s> and
2c6ed66c
KW
995looking no further than S<C<e - 1>> are well-formed UTF-8, as extended by Perl,
996that represents some code point; otherwise it evaluates to 0. If non-zero, the
997value gives how many bytes starting at C<s> comprise the code point's
2717076a
KW
998representation. Any bytes remaining before C<e>, but beyond the ones needed to
999form the first code point in C<s>, are not examined.
6302f837 1000
35f8c9bd
KW
1001The code point can be any that will fit in a UV on this machine, using Perl's
1002extension to official UTF-8 to represent those higher than the Unicode maximum
1003of 0x10FFFF. That means that this macro is used to efficiently decide if the
2717076a
KW
1004next few bytes in C<s> is legal UTF-8 for a single character.
1005
1006Use C<L</isSTRICT_UTF8_CHAR>> to restrict the acceptable code points to those
1007defined by Unicode to be fully interchangeable across applications;
1008C<L</isC9_STRICT_UTF8_CHAR>> to use the L<Unicode Corrigendum
1009#9|http://www.unicode.org/versions/corrigendum9.html> definition of allowable
1010code points; and C<L</isUTF8_CHAR_flags>> for a more customized definition.
1011
1012Use C<L</is_utf8_string>>, C<L</is_utf8_string_loc>>, and
1013C<L</is_utf8_string_loclen>> to check entire strings.
35f8c9bd
KW
1014
1015Note that it is deprecated to use code points higher than what will fit in an
1016IV. This macro does not raise any warnings for such code points, treating them
1017as valid.
1018
1019Note also that a UTF-8 INVARIANT character (i.e. ASCII on non-EBCDIC machines)
1020is a valid UTF-8 character.
5dca9278
KW
1021
1022=cut
1023*/
6302f837 1024
784d4f31
KW
1025#define isUTF8_CHAR(s, e) \
1026 (UNLIKELY((e) <= (s)) \
1027 ? 0 \
1028 : (UTF8_IS_INVARIANT(*s)) \
df863e43
KW
1029 ? 1 \
1030 : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \
1031 ? 0 \
1032 : LIKELY(NATIVE_UTF8_TO_I8(*s) <= _IS_UTF8_CHAR_HIGHEST_START_BYTE) \
1033 ? is_UTF8_CHAR_utf8_no_length_checks(s) \
1034 : _is_utf8_char_helper(s, e, 0))
6302f837 1035
3cedd9d9
KW
1036#define is_utf8_char_buf(buf, buf_end) isUTF8_CHAR(buf, buf_end)
1037
e23e8bc1
KW
1038/*
1039
1040=for apidoc Am|STRLEN|isSTRICT_UTF8_CHAR|const U8 *s|const U8 *e
1041
1042Evaluates to non-zero if the first few bytes of the string starting at C<s> and
1043looking no further than S<C<e - 1>> are well-formed UTF-8 that represents some
1044Unicode code point completely acceptable for open interchange between all
1045applications; otherwise it evaluates to 0. If non-zero, the value gives how
2717076a
KW
1046many bytes starting at C<s> comprise the code point's representation. Any
1047bytes remaining before C<e>, but beyond the ones needed to form the first code
1048point in C<s>, are not examined.
e23e8bc1
KW
1049
1050The largest acceptable code point is the Unicode maximum 0x10FFFF, and must not
1051be a surrogate nor a non-character code point. Thus this excludes any code
1052point from Perl's extended UTF-8.
1053
1054This is used to efficiently decide if the next few bytes in C<s> is
2717076a
KW
1055legal Unicode-acceptable UTF-8 for a single character.
1056
1057Use C<L</isC9_STRICT_UTF8_CHAR>> to use the L<Unicode Corrigendum
1058#9|http://www.unicode.org/versions/corrigendum9.html> definition of allowable
1059code points; C<L</isUTF8_CHAR>> to check for Perl's extended UTF-8;
1060and C<L</isUTF8_CHAR_flags>> for a more customized definition.
e23e8bc1 1061
9f2abfde
KW
1062Use C<L</is_strict_utf8_string>>, C<L</is_strict_utf8_string_loc>>, and
1063C<L</is_strict_utf8_string_loclen>> to check entire strings.
1064
e23e8bc1
KW
1065=cut
1066*/
1067
1068#define isSTRICT_UTF8_CHAR(s, e) \
1069 (UNLIKELY((e) <= (s)) \
1070 ? 0 \
1071 : (UTF8_IS_INVARIANT(*s)) \
1072 ? 1 \
1073 : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \
1074 ? 0 \
1075 : is_STRICT_UTF8_CHAR_utf8_no_length_checks(s))
1076
a82be82b
KW
1077/*
1078
1079=for apidoc Am|STRLEN|isC9_STRICT_UTF8_CHAR|const U8 *s|const U8 *e
1080
1081Evaluates to non-zero if the first few bytes of the string starting at C<s> and
1082looking no further than S<C<e - 1>> are well-formed UTF-8 that represents some
1083Unicode non-surrogate code point; otherwise it evaluates to 0. If non-zero,
1084the value gives how many bytes starting at C<s> comprise the code point's
2717076a
KW
1085representation. Any bytes remaining before C<e>, but beyond the ones needed to
1086form the first code point in C<s>, are not examined.
a82be82b
KW
1087
1088The largest acceptable code point is the Unicode maximum 0x10FFFF. This
1089differs from C<L</isSTRICT_UTF8_CHAR>> only in that it accepts non-character
1090code points. This corresponds to
1091L<Unicode Corrigendum #9|http://www.unicode.org/versions/corrigendum9.html>.
1092which said that non-character code points are merely discouraged rather than
1093completely forbidden in open interchange. See
1094L<perlunicode/Noncharacter code points>.
1095
2717076a
KW
1096Use C<L</isUTF8_CHAR>> to check for Perl's extended UTF-8; and
1097C<L</isUTF8_CHAR_flags>> for a more customized definition.
1098
9f2abfde
KW
1099Use C<L</is_c9strict_utf8_string>>, C<L</is_c9strict_utf8_string_loc>>, and
1100C<L</is_c9strict_utf8_string_loclen>> to check entire strings.
1101
a82be82b
KW
1102=cut
1103*/
1104
1105#define isC9_STRICT_UTF8_CHAR(s, e) \
1106 (UNLIKELY((e) <= (s)) \
1107 ? 0 \
1108 : (UTF8_IS_INVARIANT(*s)) \
1109 ? 1 \
1110 : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \
1111 ? 0 \
1112 : is_C9_STRICT_UTF8_CHAR_utf8_no_length_checks(s))
1113
25e3a4e0
KW
1114/*
1115
1116=for apidoc Am|STRLEN|isUTF8_CHAR_flags|const U8 *s|const U8 *e| const U32 flags
1117
1118Evaluates to non-zero if the first few bytes of the string starting at C<s> and
1119looking no further than S<C<e - 1>> are well-formed UTF-8, as extended by Perl,
1120that represents some code point, subject to the restrictions given by C<flags>;
1121otherwise it evaluates to 0. If non-zero, the value gives how many bytes
2717076a
KW
1122starting at C<s> comprise the code point's representation. Any bytes remaining
1123before C<e>, but beyond the ones needed to form the first code point in C<s>,
1124are not examined.
25e3a4e0
KW
1125
1126If C<flags> is 0, this gives the same results as C<L</isUTF8_CHAR>>;
1127if C<flags> is C<UTF8_DISALLOW_ILLEGAL_INTERCHANGE>, this gives the same results
1128as C<L</isSTRICT_UTF8_CHAR>>;
1129and if C<flags> is C<UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE>, this gives
1130the same results as C<L</isC9_STRICT_UTF8_CHAR>>.
1131Otherwise C<flags> may be any combination of the C<UTF8_DISALLOW_I<foo>> flags
1132understood by C<L</utf8n_to_uvchr>>, with the same meanings.
1133
1134The three alternative macros are for the most commonly needed validations; they
1135are likely to run somewhat faster than this more general one, as they can be
1136inlined into your code.
1137
9f2abfde
KW
1138Use L</is_utf8_string_flags>, L</is_utf8_string_loc_flags>, and
1139L</is_utf8_string_loclen_flags> to check entire strings.
1140
25e3a4e0
KW
1141=cut
1142*/
1143
1144#define isUTF8_CHAR_flags(s, e, flags) \
1145 (UNLIKELY((e) <= (s)) \
1146 ? 0 \
1147 : (UTF8_IS_INVARIANT(*s)) \
1148 ? 1 \
1149 : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \
1150 ? 0 \
1151 : _is_utf8_char_helper(s, e, flags))
1152
6302f837
KW
1153/* Do not use; should be deprecated. Use isUTF8_CHAR() instead; this is
1154 * retained solely for backwards compatibility */
1155#define IS_UTF8_CHAR(p, n) (isUTF8_CHAR(p, (p) + (n)) == n)
e9a8c099 1156
57f0e7e2
KW
1157#endif /* H_UTF8 */
1158
e9a8c099 1159/*
14d04a33 1160 * ex: set ts=8 sts=4 sw=4 et:
e9a8c099 1161 */