This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Add macro for determining if UTF-8 is Unicode-strict
[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
KW
74 ckWARN_d(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY)
75
a0270393 76#define to_uni_fold(c, p, lenp) _to_uni_fold_flags(c, p, lenp, FOLD_FLAGS_FULL)
445bf929
KW
77#define to_utf8_fold(c, p, lenp) _to_utf8_fold_flags(c, p, lenp, FOLD_FLAGS_FULL)
78#define to_utf8_lower(a,b,c) _to_utf8_lower_flags(a,b,c,0)
79#define to_utf8_upper(a,b,c) _to_utf8_upper_flags(a,b,c,0)
80#define to_utf8_title(a,b,c) _to_utf8_title_flags(a,b,c,0)
36bb2ab6 81
fd7cb289 82/* Source backward compatibility. */
fd7cb289
RGS
83#define is_utf8_string_loc(s, len, ep) is_utf8_string_loclen(s, len, ep, 0)
84
eda9cac1
KW
85#define foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \
86 foldEQ_utf8_flags(s1, pe1, l1, u1, s2, pe2, l2, u2, 0)
baa60164 87#define FOLDEQ_UTF8_NOMIX_ASCII (1 << 0)
cea315b6 88#define FOLDEQ_LOCALE (1 << 1)
18f762c3
KW
89#define FOLDEQ_S1_ALREADY_FOLDED (1 << 2)
90#define FOLDEQ_S2_ALREADY_FOLDED (1 << 3)
d635b710
KW
91#define FOLDEQ_S1_FOLDS_SANE (1 << 4)
92#define FOLDEQ_S2_FOLDS_SANE (1 << 5)
a33c29bc 93
e6226b18
KW
94#define ibcmp_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \
95 cBOOL(! foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2))
96
1d72bdf6
NIS
97#ifdef EBCDIC
98/* The equivalent of these macros but implementing UTF-EBCDIC
99 are in the following header file:
100 */
101
102#include "utfebcdic.h"
fd7cb289 103
d06134e5 104#else /* ! EBCDIC */
73c4f7a1
GS
105START_EXTERN_C
106
111e8ed9
KW
107/* How wide can a single UTF-8 encoded character become in bytes. */
108/* NOTE: Strictly speaking Perl's UTF-8 should not be called UTF-8 since UTF-8
109 * is an encoding of Unicode, and Unicode's upper limit, 0x10FFFF, can be
110 * expressed with 4 bytes. However, Perl thinks of UTF-8 as a way to encode
111 * non-negative integers in a binary format, even those above Unicode */
112#define UTF8_MAXBYTES 13
113
a0ed51b3 114#ifdef DOINIT
6f06b55f 115EXTCONST unsigned char PL_utf8skip[] = {
b2635aa8
KW
116/* 0x00 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
117/* 0x10 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
118/* 0x20 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
119/* 0x30 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
120/* 0x40 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
121/* 0x50 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
122/* 0x60 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
123/* 0x70 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
124/* 0x80 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
125/* 0x90 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
126/* 0xA0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
127/* 0xB0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
128/* 0xC0 */ 2,2, /* overlong */
1ff3baa2 129/* 0xC2 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0080 to U+03FF */
b2635aa8
KW
130/* 0xD0 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0400 to U+07FF */
131/* 0xE0 */ 3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3, /* U+0800 to U+FFFF */
132/* 0xF0 */ 4,4,4,4,4,4,4,4,5,5,5,5,6,6, /* above BMP to 2**31 - 1 */
6937f885
KW
133 /* Perl extended (never was official UTF-8). Up to 36 bit */
134/* 0xFE */ 7,
135 /* More extended, Up to 72 bits (64-bit + reserved) */
111e8ed9 136/* 0xFF */ UTF8_MAXBYTES
a0ed51b3
LW
137};
138#else
6f06b55f 139EXTCONST unsigned char PL_utf8skip[];
a0ed51b3
LW
140#endif
141
73c4f7a1 142END_EXTERN_C
7e2040f0 143
1a3756de 144#if defined(_MSC_VER) && _MSC_VER < 1400
6f6d1bab
TC
145/* older MSVC versions have a smallish macro buffer */
146#define PERL_SMALL_MACRO_BUFFER
147#endif
148
59a449d5
KW
149/* Native character to/from iso-8859-1. Are the identity functions on ASCII
150 * platforms */
6f6d1bab 151#ifdef PERL_SMALL_MACRO_BUFFER
1a3756de
TC
152#define NATIVE_TO_LATIN1(ch) ((U8)(ch))
153#define LATIN1_TO_NATIVE(ch) ((U8)(ch))
6f6d1bab 154#else
4c8cd605
KW
155#define NATIVE_TO_LATIN1(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch)))
156#define LATIN1_TO_NATIVE(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch)))
6f6d1bab 157#endif
59a449d5
KW
158
159/* I8 is an intermediate version of UTF-8 used only in UTF-EBCDIC. We thus
160 * consider it to be identical to UTF-8 on ASCII platforms. Strictly speaking
161 * UTF-8 and UTF-EBCDIC are two different things, but we often conflate them
162 * because they are 8-bit encodings that serve the same purpose in Perl, and
163 * rarely do we need to distinguish them. The term "NATIVE_UTF8" applies to
164 * whichever one is applicable on the current platform */
6f6d1bab 165#ifdef PERL_SMALL_MACRO_BUFFER
1d4ea287
DD
166#define NATIVE_UTF8_TO_I8(ch) ((U8) (ch))
167#define I8_TO_NATIVE_UTF8(ch) ((U8) (ch))
6f6d1bab 168#else
4c8cd605
KW
169#define NATIVE_UTF8_TO_I8(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch)))
170#define I8_TO_NATIVE_UTF8(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch)))
6f6d1bab 171#endif
59a449d5 172
1d72bdf6 173/* Transforms in wide UV chars */
4c8cd605
KW
174#define UNI_TO_NATIVE(ch) ((UV) (ch))
175#define NATIVE_TO_UNI(ch) ((UV) (ch))
d7578b48 176
877d9f0d 177/*
9041c2e3 178
a14e0a36
KW
179 The following table is from Unicode 3.2, plus the Perl extensions for above
180 U+10FFFF
877d9f0d 181
a14e0a36 182 Code Points 1st Byte 2nd Byte 3rd 4th 5th 6th 7th 8th-13th
877d9f0d 183
375122d7 184 U+0000..U+007F 00..7F
e1b711da 185 U+0080..U+07FF * C2..DF 80..BF
a14e0a36
KW
186 U+0800..U+0FFF E0 * A0..BF 80..BF
187 U+1000..U+CFFF E1..EC 80..BF 80..BF
188 U+D000..U+D7FF ED 80..9F 80..BF
189 U+D800..U+DFFF ED A0..BF 80..BF (surrogates)
190 U+E000..U+FFFF EE..EF 80..BF 80..BF
191 U+10000..U+3FFFF F0 * 90..BF 80..BF 80..BF
192 U+40000..U+FFFFF F1..F3 80..BF 80..BF 80..BF
193 U+100000..U+10FFFF F4 80..8F 80..BF 80..BF
194 Below are above-Unicode code points
195 U+110000..U+13FFFF F4 90..BF 80..BF 80..BF
196 U+110000..U+1FFFFF F5..F7 80..BF 80..BF 80..BF
197 U+200000..U+FFFFFF F8 * 88..BF 80..BF 80..BF 80..BF
198U+1000000..U+3FFFFFF F9..FB 80..BF 80..BF 80..BF 80..BF
199U+4000000..U+3FFFFFFF FC * 84..BF 80..BF 80..BF 80..BF 80..BF
200U+40000000..U+7FFFFFFF FD 80..BF 80..BF 80..BF 80..BF 80..BF
201U+80000000..U+FFFFFFFFF FE * 82..BF 80..BF 80..BF 80..BF 80..BF 80..BF
202U+1000000000.. FF 80..BF 80..BF 80..BF 80..BF 80..BF * 81..BF 80..BF
877d9f0d 203
e1b711da 204Note the gaps before several of the byte entries above marked by '*'. These are
37e2e78e
KW
205caused by legal UTF-8 avoiding non-shortest encodings: it is technically
206possible to UTF-8-encode a single code point in different ways, but that is
207explicitly forbidden, and the shortest possible encoding should always be used
15824458 208(and that is what Perl does). The non-shortest ones are called 'overlongs'.
8c007b5a 209
877d9f0d
JH
210 */
211
8c007b5a
JH
212/*
213 Another way to look at it, as bits:
214
b2635aa8 215 Code Points 1st Byte 2nd Byte 3rd Byte 4th Byte
8c007b5a 216
b2635aa8
KW
217 0aaa aaaa 0aaa aaaa
218 0000 0bbb bbaa aaaa 110b bbbb 10aa aaaa
219 cccc bbbb bbaa aaaa 1110 cccc 10bb bbbb 10aa aaaa
220 00 000d ddcc cccc bbbb bbaa aaaa 1111 0ddd 10cc cccc 10bb bbbb 10aa aaaa
8c007b5a
JH
221
222As you can see, the continuation bytes all begin with C<10>, and the
e1b711da 223leading bits of the start byte tell how many bytes there are in the
8c007b5a
JH
224encoded character.
225
df863e43
KW
226Perl's extended UTF-8 means we can have start bytes up through FF, though any
227beginning with FF yields a code point that is too large for 32-bit ASCII
228platforms. FF signals to use 13 bytes for the encoded character. This breaks
229the paradigm that the number of leading bits gives how many total bytes there
230are in the character.
65ab9279 231
8c007b5a
JH
232*/
233
eda91ad7
KW
234/* Anything larger than this will overflow the word if it were converted into a UV */
235#if defined(UV_IS_QUAD)
236# define HIGHEST_REPRESENTABLE_UTF8 "\xFF\x80\x8F\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF"
237#else
238# define HIGHEST_REPRESENTABLE_UTF8 "\xFE\x83\xBF\xBF\xBF\xBF\xBF"
239#endif
240
6c88483e 241/* Is the representation of the Unicode code point 'cp' the same regardless of
15824458 242 * being encoded in UTF-8 or not? */
2d1545e5 243#define OFFUNI_IS_INVARIANT(cp) isASCII(cp)
15824458 244
c2b32798
KW
245/*
246=for apidoc Am|bool|UVCHR_IS_INVARIANT|UV cp
247
248Evaluates to 1 if the representation of code point C<cp> is the same whether or
249not it is encoded in UTF-8; otherwise evaluates to 0. UTF-8 invariant
250characters can be copied as-is when converting to/from UTF-8, saving time.
251C<cp> is Unicode if above 255; otherwise is platform-native.
252
253=cut
254 */
255
cf1be84e 256#define UVCHR_IS_INVARIANT(cp) OFFUNI_IS_INVARIANT(cp)
38953e5a 257
c9264833
KW
258/* This defines the bits that are to be in the continuation bytes of a multi-byte
259 * UTF-8 encoded character that mark it is a continuation byte. */
260#define UTF_CONTINUATION_MARK 0x80
261
a95ec4fb 262/* Misleadingly named: is the UTF8-encoded byte 'c' part of a variant sequence
5c06326b
KW
263 * in UTF-8? This is the inverse of UTF8_IS_INVARIANT. The |0 makes sure this
264 * isn't mistakenly called with a ptr argument */
265#define UTF8_IS_CONTINUED(c) (((U8)((c) | 0)) & UTF_CONTINUATION_MARK)
15824458
KW
266
267/* Is the byte 'c' the first byte of a multi-byte UTF8-8 encoded sequence?
268 * This doesn't catch invariants (they are single-byte). It also excludes the
5c06326b
KW
269 * illegal overlong sequences that begin with C0 and C1. The |0 makes sure
270 * this isn't mistakenly called with a ptr argument */
271#define UTF8_IS_START(c) (((U8)((c) | 0)) >= 0xc2)
15824458 272
858cd8ab
KW
273/* For use in UTF8_IS_CONTINUATION() below */
274#define UTF_IS_CONTINUATION_MASK 0xC0
275
15824458 276/* Is the byte 'c' part of a multi-byte UTF8-8 encoded sequence, and not the
5c06326b
KW
277 * first byte thereof? The |0 makes sure this isn't mistakenly called with a
278 * ptr argument */
858cd8ab 279#define UTF8_IS_CONTINUATION(c) \
5c06326b 280 ((((U8)((c) | 0)) & UTF_IS_CONTINUATION_MASK) == UTF_CONTINUATION_MARK)
0ae1fa71 281
15824458
KW
282/* Is the UTF8-encoded byte 'c' the first byte of a two byte sequence? Use
283 * UTF8_IS_NEXT_CHAR_DOWNGRADEABLE() instead if the input isn't known to
284 * be well-formed. Masking with 0xfe allows the low bit to be 0 or 1; thus
5c06326b
KW
285 * this matches 0xc[23]. The |0 makes sure this isn't mistakenly called with a
286 * ptr argument */
287#define UTF8_IS_DOWNGRADEABLE_START(c) ((((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 */
292#define UTF8_IS_ABOVE_LATIN1(c) (((U8)((c) | 0)) >= 0xc4)
8850bf83 293
15824458
KW
294/* This is the number of low-order bits a continuation byte in a UTF-8 encoded
295 * sequence contributes to the specification of the code point. In the bit
296 * maps above, you see that the first 2 bits are a constant '10', leaving 6 of
297 * real information */
1d72bdf6 298#define UTF_ACCUMULATION_SHIFT 6
b2635aa8 299
fed423a5
KW
300/* ^? is defined to be DEL on ASCII systems. See the definition of toCTRL()
301 * for more */
302#define QUESTION_MARK_CTRL DEL_NATIVE
303
304/* Surrogates, non-character code points and above-Unicode code points are
305 * problematic in some contexts. This allows code that needs to check for
306 * those to to quickly exclude the vast majority of code points it will
307 * encounter */
308#define isUTF8_POSSIBLY_PROBLEMATIC(c) ((U8) c >= 0xED)
309
784d4f31
KW
310/* A helper macro for isUTF8_CHAR, so use that one instead of this. This was
311 * generated by regen/regcharclass.pl, and then moved here. Then it was
312 * hand-edited to add some LIKELY() calls, presuming that malformations are
313 * unlikely. The lines that generated it were then commented out. This was
314 * done because it takes on the order of 10 minutes to generate, and is never
315 * going to change, unless the generated code is improved, and figuring out
316 * the LIKELYs there would be hard.
317 *
318 UTF8_CHAR: Matches legal UTF-8 variant code points up through 0x1FFFFFF
319
320 0x80 - 0x1FFFFF
321*/
322/*** GENERATED CODE ***/
323#define is_UTF8_CHAR_utf8_no_length_checks(s) \
324( ( 0xC2 <= ((U8*)s)[0] && ((U8*)s)[0] <= 0xDF ) ? \
325 ( LIKELY( ( ((U8*)s)[1] & 0xC0 ) == 0x80 ) ? 2 : 0 ) \
326: ( 0xE0 == ((U8*)s)[0] ) ? \
327 ( LIKELY( ( ( ((U8*)s)[1] & 0xE0 ) == 0xA0 ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
328: ( 0xE1 <= ((U8*)s)[0] && ((U8*)s)[0] <= 0xEF ) ? \
329 ( LIKELY( ( ( ((U8*)s)[1] & 0xC0 ) == 0x80 ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
330: ( 0xF0 == ((U8*)s)[0] ) ? \
331 ( LIKELY( ( ( 0x90 <= ((U8*)s)[1] && ((U8*)s)[1] <= 0xBF ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) && ( ( ((U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\
332: ( ( ( ( 0xF1 <= ((U8*)s)[0] && ((U8*)s)[0] <= 0xF7 ) && LIKELY( ( ((U8*)s)[1] & 0xC0 ) == 0x80 ) ) && LIKELY( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) && LIKELY( ( ((U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )
333
334/* The above macro handles UTF-8 that has this start byte as the maximum */
335#define _IS_UTF8_CHAR_HIGHEST_START_BYTE 0xF7
336
e23e8bc1
KW
337/* A helper macro for isSTRICT_UTF8_CHAR, so use that one instead of this.
338 * Like is_UTF8_CHAR_utf8_no_length_checks(), this was moved here and LIKELYs
339 * added manually.
340 *
341 STRICT_UTF8_CHAR: Matches legal Unicode UTF-8 variant code points, no
342 surrrogates nor non-character code points
343*/
344/*** GENERATED CODE ***/
345#define is_STRICT_UTF8_CHAR_utf8_no_length_checks(s) \
346( ( 0xC2 <= ((U8*)s)[0] && ((U8*)s)[0] <= 0xDF ) ? \
347 ( LIKELY( ( ((U8*)s)[1] & 0xC0 ) == 0x80 ) ? 2 : 0 ) \
348: ( 0xE0 == ((U8*)s)[0] ) ? \
349 ( LIKELY( ( ( ((U8*)s)[1] & 0xE0 ) == 0xA0 ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
350: ( ( 0xE1 <= ((U8*)s)[0] && ((U8*)s)[0] <= 0xEC ) || 0xEE == ((U8*)s)[0] ) ?\
351 ( ( ( ( ((U8*)s)[1] & 0xC0 ) == 0x80 ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
352: ( 0xED == ((U8*)s)[0] ) ? \
353 ( LIKELY( ( ( ((U8*)s)[1] & 0xE0 ) == 0x80 ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
354: ( 0xEF == ((U8*)s)[0] ) ? \
355 ( ( ( 0x80 <= ((U8*)s)[1] && ((U8*)s)[1] <= 0xB6 ) || ( 0xB8 <= ((U8*)s)[1] && ((U8*)s)[1] <= 0xBE ) ) ?\
356 ( LIKELY( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ? 3 : 0 ) \
357 : ( 0xB7 == ((U8*)s)[1] ) ? \
358 ( LIKELY( ( ((U8*)s)[2] & 0xF0 ) == 0x80 || ( ((U8*)s)[2] & 0xF0 ) == 0xB0 ) ? 3 : 0 )\
359 : ( ( 0xBF == ((U8*)s)[1] ) && ( 0x80 <= ((U8*)s)[2] && ((U8*)s)[2] <= 0xBD ) ) ? 3 : 0 )\
360: ( 0xF0 == ((U8*)s)[0] ) ? \
361 ( ( ( 0x90 <= ((U8*)s)[1] && ((U8*)s)[1] <= 0x9E ) || ( 0xA0 <= ((U8*)s)[1] && ((U8*)s)[1] <= 0xAE ) || ( 0xB0 <= ((U8*)s)[1] && ((U8*)s)[1] <= 0xBE ) ) ?\
362 ( LIKELY( ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) && ( ( ((U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\
363 : ( ((U8*)s)[1] == 0x9F || ( ( ((U8*)s)[1] & 0xEF ) == 0xAF ) ) ? \
364 ( ( 0x80 <= ((U8*)s)[2] && ((U8*)s)[2] <= 0xBE ) ? \
365 ( LIKELY( ( ((U8*)s)[3] & 0xC0 ) == 0x80 ) ? 4 : 0 ) \
366 : LIKELY( ( 0xBF == ((U8*)s)[2] ) && ( 0x80 <= ((U8*)s)[3] && ((U8*)s)[3] <= 0xBD ) ) ? 4 : 0 )\
367 : 0 ) \
368: ( 0xF1 <= ((U8*)s)[0] && ((U8*)s)[0] <= 0xF3 ) ? \
369 ( ( ( ( ((U8*)s)[1] & 0xC8 ) == 0x80 ) || ( ( ((U8*)s)[1] & 0xCC ) == 0x88 ) || ( ( ((U8*)s)[1] & 0xCE ) == 0x8C ) || ( ( ((U8*)s)[1] & 0xCF ) == 0x8E ) ) ?\
370 ( LIKELY( ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) && ( ( ((U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\
371 : ( ( ((U8*)s)[1] & 0xCF ) == 0x8F ) ? \
372 ( ( 0x80 <= ((U8*)s)[2] && ((U8*)s)[2] <= 0xBE ) ? \
373 ( LIKELY( ( ((U8*)s)[3] & 0xC0 ) == 0x80 ) ? 4 : 0 ) \
374 : LIKELY( ( 0xBF == ((U8*)s)[2] ) && ( 0x80 <= ((U8*)s)[3] && ((U8*)s)[3] <= 0xBD ) ) ? 4 : 0 )\
375 : 0 ) \
376: ( 0xF4 == ((U8*)s)[0] ) ? \
377 ( ( 0x80 <= ((U8*)s)[1] && ((U8*)s)[1] <= 0x8E ) ? \
378 ( LIKELY( ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) && ( ( ((U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\
379 : ( 0x8F == ((U8*)s)[1] ) ? \
380 ( ( 0x80 <= ((U8*)s)[2] && ((U8*)s)[2] <= 0xBE ) ? \
381 ( LIKELY( ( ((U8*)s)[3] & 0xC0 ) == 0x80 ) ? 4 : 0 ) \
382 : LIKELY( ( 0xBF == ((U8*)s)[2] ) && ( 0x80 <= ((U8*)s)[3] && ((U8*)s)[3] <= 0xBD ) ) ? 4 : 0 )\
383 : 0 ) \
384: 0 )
385
fed423a5
KW
386#endif /* EBCDIC vs ASCII */
387
388/* 2**UTF_ACCUMULATION_SHIFT - 1 */
389#define UTF_CONTINUATION_MASK ((U8) ((1U << UTF_ACCUMULATION_SHIFT) - 1))
390
7028aeba
KW
391/* Internal macro to be used only in this file to aid in constructing other
392 * publicly accessible macros.
393 * The number of bytes required to express this uv in UTF-8, for just those
394 * uv's requiring 2 through 6 bytes, as these are common to all platforms and
395 * word sizes. The number of bytes needed is given by the number of leading 1
396 * bits in the start byte. There are 32 start bytes that have 2 initial 1 bits
397 * (C0-DF); there are 16 that have 3 initial 1 bits (E0-EF); 8 that have 4
398 * initial 1 bits (F0-F8); 4 that have 5 initial 1 bits (F9-FB), and 2 that
399 * have 6 initial 1 bits (FC-FD). The largest number a string of n bytes can
400 * represent is (the number of possible start bytes for 'n')
401 * * (the number of possiblities for each start byte
402 * The latter in turn is
403 * 2 ** ( (how many continuation bytes there are)
404 * * (the number of bits of information each
405 * continuation byte holds))
406 *
407 * If we were on a platform where we could use a fast find first set bit
408 * instruction (or count leading zeros instruction) this could be replaced by
409 * using that to find the log2 of the uv, and divide that by the number of bits
410 * of information in each continuation byte, adjusting for large cases and how
411 * much information is in a start byte for that length */
72164d3a 412#define __COMMON_UNI_SKIP(uv) \
7028aeba
KW
413 (UV) (uv) < (32 * (1U << ( UTF_ACCUMULATION_SHIFT))) ? 2 : \
414 (UV) (uv) < (16 * (1U << (2 * UTF_ACCUMULATION_SHIFT))) ? 3 : \
415 (UV) (uv) < ( 8 * (1U << (3 * UTF_ACCUMULATION_SHIFT))) ? 4 : \
416 (UV) (uv) < ( 4 * (1U << (4 * UTF_ACCUMULATION_SHIFT))) ? 5 : \
417 (UV) (uv) < ( 2 * (1U << (5 * UTF_ACCUMULATION_SHIFT))) ? 6 :
72164d3a
KW
418
419/* Internal macro to be used only in this file.
420 * This adds to __COMMON_UNI_SKIP the details at this platform's upper range.
fed423a5 421 * For any-sized EBCDIC platforms, or 64-bit ASCII ones, we need one more test
72164d3a
KW
422 * to see if just 7 bytes is needed, or if the maximum is needed. For 32-bit
423 * ASCII platforms, everything is representable by 7 bytes */
fed423a5 424#if defined(UV_IS_QUAD) || defined(EBCDIC)
72164d3a 425# define __BASE_UNI_SKIP(uv) (__COMMON_UNI_SKIP(uv) \
7028aeba 426 (UV) (uv) < ((UV) 1U << (6 * UTF_ACCUMULATION_SHIFT)) ? 7 : UTF8_MAXBYTES)
1d68d6cd 427#else
72164d3a 428# define __BASE_UNI_SKIP(uv) (__COMMON_UNI_SKIP(uv) 7)
1d68d6cd
SC
429#endif
430
fed423a5
KW
431/* The next two macros use the base macro defined above, and add in the tests
432 * at the low-end of the range, for just 1 byte, yielding complete macros,
433 * publicly accessible. */
434
435/* Input is a true Unicode (not-native) code point */
436#define OFFUNISKIP(uv) (OFFUNI_IS_INVARIANT(uv) ? 1 : __BASE_UNI_SKIP(uv))
2084b489 437
5352a763
KW
438/*
439
440=for apidoc Am|STRLEN|UVCHR_SKIP|UV cp
441returns the number of bytes required to represent the code point C<cp> when
442encoded as UTF-8. C<cp> is a native (ASCII or EBCDIC) code point if less than
443255; a Unicode code point otherwise.
444
445=cut
446 */
fdb6583d 447#define UVCHR_SKIP(uv) ( UVCHR_IS_INVARIANT(uv) ? 1 : __BASE_UNI_SKIP(uv))
5352a763 448
b651802e
KW
449/* The largest code point representable by two UTF-8 bytes on this platform.
450 * As explained in the comments for __COMMON_UNI_SKIP, 32 start bytes with
fed423a5 451 * UTF_ACCUMULATION_SHIFT bits of information each */
aa206fb7
KW
452#define MAX_UTF8_TWO_BYTE (32 * (1U << UTF_ACCUMULATION_SHIFT) - 1)
453
b651802e
KW
454/* The largest code point representable by two UTF-8 bytes on any platform that
455 * Perl runs on. This value is constrained by EBCDIC which has 5 bits per
456 * continuation byte */
aa206fb7
KW
457#define MAX_PORTABLE_UTF8_TWO_BYTE (32 * (1U << 5) - 1)
458
c03c0950
KW
459/* The maximum number of UTF-8 bytes a single Unicode character can
460 * uppercase/lowercase/fold into. Unicode guarantees that the maximum
461 * expansion is UTF8_MAX_FOLD_CHAR_EXPAND characters, but any above-Unicode
462 * code point will fold to itself, so we only have to look at the expansion of
463 * the maximum Unicode code point. But this number may be less than the space
464 * occupied by a very large code point under Perl's extended UTF-8. We have to
465 * make it large enough to fit any single character. (It turns out that ASCII
466 * and EBCDIC differ in which is larger) */
467#define UTF8_MAXBYTES_CASE \
468 (UTF8_MAXBYTES >= (UTF8_MAX_FOLD_CHAR_EXPAND * OFFUNISKIP(0x10FFFF)) \
469 ? UTF8_MAXBYTES \
470 : (UTF8_MAX_FOLD_CHAR_EXPAND * OFFUNISKIP(0x10FFFF)))
471
d06134e5
KW
472/* Rest of these are attributes of Unicode and perl's internals rather than the
473 * encoding, or happen to be the same in both ASCII and EBCDIC (at least at
474 * this level; the macros that some of these call may have different
475 * definitions in the two encodings */
476
59a449d5
KW
477/* In domain restricted to ASCII, these may make more sense to the reader than
478 * the ones with Latin1 in the name */
479#define NATIVE_TO_ASCII(ch) NATIVE_TO_LATIN1(ch)
480#define ASCII_TO_NATIVE(ch) LATIN1_TO_NATIVE(ch)
481
482/* More or less misleadingly-named defines, retained for back compat */
483#define NATIVE_TO_UTF(ch) NATIVE_UTF8_TO_I8(ch)
484#define NATIVE_TO_I8(ch) NATIVE_UTF8_TO_I8(ch)
485#define UTF_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch)
486#define I8_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch)
487#define NATIVE8_TO_UNI(ch) NATIVE_TO_LATIN1(ch)
d06134e5 488
c0236afe 489/* This defines the 1-bits that are to be in the first byte of a multi-byte
97d0ceda
KW
490 * UTF-8 encoded character that mark it as a start byte and give the number of
491 * bytes that comprise the character. 'len' is the number of bytes in the
492 * multi-byte sequence. */
c0236afe
KW
493#define UTF_START_MARK(len) (((len) > 7) ? 0xFF : (0xFF & (0xFE << (7-(len)))))
494
495/* Masks out the initial one bits in a start byte, leaving the real data ones.
496 * Doesn't work on an invariant byte. 'len' is the number of bytes in the
497 * multi-byte sequence that comprises the character. */
498#define UTF_START_MASK(len) (((len) >= 7) ? 0x00 : (0x1F >> ((len)-2)))
499
537124e4
KW
500/* Adds a UTF8 continuation byte 'new' of information to a running total code
501 * point 'old' of all the continuation bytes so far. This is designed to be
155d2738
KW
502 * used in a loop to convert from UTF-8 to the code point represented. Note
503 * that this is asymmetric on EBCDIC platforms, in that the 'new' parameter is
504 * the UTF-EBCDIC byte, whereas the 'old' parameter is a Unicode (not EBCDIC)
505 * code point in process of being generated */
506#define UTF8_ACCUMULATE(old, new) (((old) << UTF_ACCUMULATION_SHIFT) \
507 | ((NATIVE_UTF8_TO_I8((U8)new)) \
508 & UTF_CONTINUATION_MASK))
d06134e5 509
7c560c3b
KW
510/* If a value is anded with this, and the result is non-zero, then using the
511 * original value in UTF8_ACCUMULATE will overflow, shifting bits off the left
512 * */
513#define UTF_ACCUMULATION_OVERFLOW_MASK \
514 (((UV) UTF_CONTINUATION_MASK) << ((sizeof(UV) * CHARBITS) \
515 - UTF_ACCUMULATION_SHIFT))
516
4ab10950
KW
517/* This works in the face of malformed UTF-8. */
518#define UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(s, e) (UTF8_IS_DOWNGRADEABLE_START(*s) \
519 && ( (e) - (s) > 1) \
520 && UTF8_IS_CONTINUATION(*((s)+1)))
521
5aaebcb3 522/* Number of bytes a code point occupies in UTF-8. */
5352a763 523#define NATIVE_SKIP(uv) UVCHR_SKIP(uv)
bd18bd40 524
5aaebcb3
KW
525/* Most code which says UNISKIP is really thinking in terms of native code
526 * points (0-255) plus all those beyond. This is an imprecise term, but having
2accb712 527 * it means existing code continues to work. For precision, use UVCHR_SKIP,
5352a763
KW
528 * NATIVE_SKIP, or OFFUNISKIP */
529#define UNISKIP(uv) UVCHR_SKIP(uv)
5aaebcb3 530
3c0792e4
KW
531/* Longer, but more accurate name */
532#define UTF8_IS_ABOVE_LATIN1_START(c) UTF8_IS_ABOVE_LATIN1(c)
533
a62b247b
KW
534/* Convert a UTF-8 variant Latin1 character to a native code point value.
535 * Needs just one iteration of accumulate. Should be used only if it is known
536 * that the code point is < 256, and is not UTF-8 invariant. Use the slower
537 * but more general TWO_BYTE_UTF8_TO_NATIVE() which handles any code point
538 * representable by two bytes (which turns out to be up through
539 * MAX_PORTABLE_UTF8_TWO_BYTE). The two parameters are:
540 * HI: a downgradable start byte;
541 * LO: continuation.
542 * */
543#define EIGHT_BIT_UTF8_TO_NATIVE(HI, LO) \
544 ( __ASSERT_(UTF8_IS_DOWNGRADEABLE_START(HI)) \
545 __ASSERT_(UTF8_IS_CONTINUATION(LO)) \
546 LATIN1_TO_NATIVE(UTF8_ACCUMULATE(( \
547 NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), (LO))))
548
94bb8c36 549/* Convert a two (not one) byte utf8 character to a native code point value.
2950f2a7
KW
550 * Needs just one iteration of accumulate. Should not be used unless it is
551 * known that the two bytes are legal: 1) two-byte start, and 2) continuation.
552 * Note that the result can be larger than 255 if the input character is not
553 * downgradable */
94bb8c36 554#define TWO_BYTE_UTF8_TO_NATIVE(HI, LO) \
b18b15ad 555 ( __ASSERT_(PL_utf8skip[HI] == 2) \
635e76f5 556 __ASSERT_(UTF8_IS_CONTINUATION(LO)) \
94bb8c36 557 UNI_TO_NATIVE(UTF8_ACCUMULATE((NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), \
635e76f5 558 (LO))))
94bb8c36
KW
559
560/* Should never be used, and be deprecated */
561#define TWO_BYTE_UTF8_TO_UNI(HI, LO) NATIVE_TO_UNI(TWO_BYTE_UTF8_TO_NATIVE(HI, LO))
2950f2a7 562
bd18bd40
KW
563/*
564
565=for apidoc Am|STRLEN|UTF8SKIP|char* s
566returns the number of bytes in the UTF-8 encoded character whose first (perhaps
567only) byte is pointed to by C<s>.
568
569=cut
570 */
2a70536e
KW
571#define UTF8SKIP(s) PL_utf8skip[*(const U8*)(s)]
572#define UTF8_SKIP(s) UTF8SKIP(s)
d06134e5 573
2d1545e5
KW
574/* Most code that says 'UNI_' really means the native value for code points up
575 * through 255 */
576#define UNI_IS_INVARIANT(cp) UVCHR_IS_INVARIANT(cp)
577
c2b32798
KW
578/*
579=for apidoc Am|bool|UTF8_IS_INVARIANT|char c
580
581Evaluates to 1 if the byte C<c> represents the same character when encoded in
582UTF-8 as when not; otherwise evaluates to 0. UTF-8 invariant characters can be
583copied as-is when converting to/from UTF-8, saving time.
584
585In spite of the name, this macro gives the correct result if the input string
586from which C<c> comes is not encoded in UTF-8.
587
588See C<L</UVCHR_IS_INVARIANT>> for checking if a UV is invariant.
589
590=cut
591
592The reason it works on both UTF-8 encoded strings and non-UTF-8 encoded, is
593that it returns TRUE in each for the exact same set of bit patterns. It is
594valid on a subset of what UVCHR_IS_INVARIANT is valid on, so can just use that;
595and the compiler should optimize out anything extraneous given the
596implementation of the latter. The |0 makes sure this isn't mistakenly called
597with a ptr argument.
598*/
5c06326b 599#define UTF8_IS_INVARIANT(c) UVCHR_IS_INVARIANT((c) | 0)
5fc230f1
KW
600
601/* Like the above, but its name implies a non-UTF8 input, which as the comments
602 * above show, doesn't matter as to its implementation */
38953e5a 603#define NATIVE_BYTE_IS_INVARIANT(c) UVCHR_IS_INVARIANT(c)
d06134e5 604
48ccf5e1
KW
605/* The macros in the next 4 sets are used to generate the two utf8 or utfebcdic
606 * bytes from an ordinal that is known to fit into exactly two (not one) bytes;
607 * it must be less than 0x3FF to work across both encodings. */
608
609/* These two are helper macros for the other three sets, and should not be used
610 * directly anywhere else. 'translate_function' is either NATIVE_TO_LATIN1
1ff3baa2
KW
611 * (which works for code points up through 0xFF) or NATIVE_TO_UNI which works
612 * for any code point */
48ccf5e1 613#define __BASE_TWO_BYTE_HI(c, translate_function) \
2863dafa 614 (__ASSERT_(! UVCHR_IS_INVARIANT(c)) \
48ccf5e1 615 I8_TO_NATIVE_UTF8((translate_function(c) >> UTF_ACCUMULATION_SHIFT) \
2863dafa 616 | UTF_START_MARK(2)))
48ccf5e1 617#define __BASE_TWO_BYTE_LO(c, translate_function) \
2863dafa 618 (__ASSERT_(! UVCHR_IS_INVARIANT(c)) \
48ccf5e1 619 I8_TO_NATIVE_UTF8((translate_function(c) & UTF_CONTINUATION_MASK) \
2863dafa 620 | UTF_CONTINUATION_MARK))
48ccf5e1 621
48ccf5e1
KW
622/* The next two macros should not be used. They were designed to be usable as
623 * the case label of a switch statement, but this doesn't work for EBCDIC. Use
9d0d3a03 624 * regen/unicode_constants.pl instead */
48ccf5e1
KW
625#define UTF8_TWO_BYTE_HI_nocast(c) __BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI)
626#define UTF8_TWO_BYTE_LO_nocast(c) __BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI)
627
628/* The next two macros are used when the source should be a single byte
629 * character; checked for under DEBUGGING */
630#define UTF8_EIGHT_BIT_HI(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
4c8cd605 631 ( __BASE_TWO_BYTE_HI(c, NATIVE_TO_LATIN1)))
48ccf5e1 632#define UTF8_EIGHT_BIT_LO(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
4c8cd605 633 (__BASE_TWO_BYTE_LO(c, NATIVE_TO_LATIN1)))
48ccf5e1
KW
634
635/* These final two macros in the series are used when the source can be any
636 * code point whose UTF-8 is known to occupy 2 bytes; they are less efficient
637 * than the EIGHT_BIT versions on EBCDIC platforms. We use the logical '~'
638 * operator instead of "<=" to avoid getting compiler warnings.
d52b8576 639 * MAX_UTF8_TWO_BYTE should be exactly all one bits in the lower few
48ccf5e1
KW
640 * places, so the ~ works */
641#define UTF8_TWO_BYTE_HI(c) \
642 (__ASSERT_((sizeof(c) == 1) \
d52b8576 643 || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \
4c8cd605 644 (__BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI)))
48ccf5e1
KW
645#define UTF8_TWO_BYTE_LO(c) \
646 (__ASSERT_((sizeof(c) == 1) \
d52b8576 647 || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \
4c8cd605 648 (__BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI)))
d06134e5 649
e7214ce8
KW
650/* This is illegal in any well-formed UTF-8 in both EBCDIC and ASCII
651 * as it is only in overlongs. */
652#define ILLEGAL_UTF8_BYTE I8_TO_NATIVE_UTF8(0xC1)
653
7e2040f0 654/*
e3036cf4 655 * 'UTF' is whether or not p is encoded in UTF8. The names 'foo_lazy_if' stem
20df05f4
KW
656 * from an earlier version of these macros in which they didn't call the
657 * foo_utf8() macros (i.e. were 'lazy') unless they decided that *p is the
658 * beginning of a utf8 character. Now that foo_utf8() determines that itself,
659 * no need to do it again here
7e2040f0 660 */
97d0ceda
KW
661#define isIDFIRST_lazy_if(p,UTF) ((IN_BYTES || !UTF) \
662 ? isIDFIRST(*(p)) \
e3036cf4 663 : isIDFIRST_utf8((const U8*)p))
97d0ceda
KW
664#define isWORDCHAR_lazy_if(p,UTF) ((IN_BYTES || (!UTF)) \
665 ? isWORDCHAR(*(p)) \
32636478
KW
666 : isWORDCHAR_utf8((const U8*)p))
667#define isALNUM_lazy_if(p,UTF) isWORDCHAR_lazy_if(p,UTF)
1d72bdf6 668
89ebb4a3
JH
669#define UTF8_MAXLEN UTF8_MAXBYTES
670
8cb75cc8
KW
671/* A Unicode character can fold to up to 3 characters */
672#define UTF8_MAX_FOLD_CHAR_EXPAND 3
673
a98fe34d 674#define IN_BYTES (CopHINTS_get(PL_curcop) & HINT_BYTES)
bd18bd40
KW
675
676/*
677
678=for apidoc Am|bool|DO_UTF8|SV* sv
679Returns a bool giving whether or not the PV in C<sv> is to be treated as being
680encoded in UTF-8.
681
682You should use this I<after> a call to C<SvPV()> or one of its variants, in
683case any call to string overloading updates the internal UTF-8 encoding flag.
684
685=cut
686*/
0064a8a9 687#define DO_UTF8(sv) (SvUTF8(sv) && !IN_BYTES)
1ff3baa2
KW
688
689/* Should all strings be treated as Unicode, and not just UTF-8 encoded ones?
690 * Is so within 'feature unicode_strings' or 'locale :not_characters', and not
691 * within 'use bytes'. UTF-8 locales are not tested for here, but perhaps
692 * could be */
693#define IN_UNI_8_BIT \
d6ded950
KW
694 (((CopHINTS_get(PL_curcop) & (HINT_UNI_8_BIT)) \
695 || (CopHINTS_get(PL_curcop) & HINT_LOCALE_PARTIAL \
696 /* -1 below is for :not_characters */ \
697 && _is_in_locale_category(FALSE, -1))) \
698 && ! IN_BYTES)
b36bf33f 699
1d72bdf6 700
c76687c5
KW
701#define UTF8_ALLOW_EMPTY 0x0001 /* Allow a zero length string */
702
703/* Allow first byte to be a continuation byte */
1d72bdf6 704#define UTF8_ALLOW_CONTINUATION 0x0002
c76687c5
KW
705
706/* Allow second... bytes to be non-continuation bytes */
1d72bdf6 707#define UTF8_ALLOW_NON_CONTINUATION 0x0004
949cf498
KW
708
709/* expecting more bytes than were available in the string */
710#define UTF8_ALLOW_SHORT 0x0008
711
712/* Overlong sequence; i.e., the code point can be specified in fewer bytes. */
713#define UTF8_ALLOW_LONG 0x0010
714
715#define UTF8_DISALLOW_SURROGATE 0x0020 /* Unicode surrogates */
716#define UTF8_WARN_SURROGATE 0x0040
717
718#define UTF8_DISALLOW_NONCHAR 0x0080 /* Unicode non-character */
719#define UTF8_WARN_NONCHAR 0x0100 /* code points */
720
721#define UTF8_DISALLOW_SUPER 0x0200 /* Super-set of Unicode: code */
722#define UTF8_WARN_SUPER 0x0400 /* points above the legal max */
723
d35f2ca5
KW
724/* Code points which never were part of the original UTF-8 standard, which only
725 * went up to 2 ** 31 - 1. Note that these all overflow a signed 32-bit word,
726 * The first byte of these code points is FE or FF on ASCII platforms. If the
727 * first byte is FF, it will overflow a 32-bit word. */
728#define UTF8_DISALLOW_ABOVE_31_BIT 0x0800
729#define UTF8_WARN_ABOVE_31_BIT 0x1000
730
731/* For back compat, these old names are misleading for UTF_EBCDIC */
732#define UTF8_DISALLOW_FE_FF UTF8_DISALLOW_ABOVE_31_BIT
733#define UTF8_WARN_FE_FF UTF8_WARN_ABOVE_31_BIT
949cf498
KW
734
735#define UTF8_CHECK_ONLY 0x2000
736
737/* For backwards source compatibility. They do nothing, as the default now
738 * includes what they used to mean. The first one's meaning was to allow the
739 * just the single non-character 0xFFFF */
740#define UTF8_ALLOW_FFFF 0
741#define UTF8_ALLOW_SURROGATE 0
742
ecc1615f
KW
743/* C9 refers to Unicode Corrigendum #9: allows but discourages non-chars */
744#define UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE \
745 (UTF8_DISALLOW_SUPER|UTF8_DISALLOW_SURROGATE)
746#define UTF8_WARN_ILLEGAL_C9_INTERCHANGE (UTF8_WARN_SUPER|UTF8_WARN_SURROGATE)
747
d35f2ca5 748#define UTF8_DISALLOW_ILLEGAL_INTERCHANGE \
ecc1615f 749 (UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE|UTF8_DISALLOW_NONCHAR)
949cf498 750#define UTF8_WARN_ILLEGAL_INTERCHANGE \
ecc1615f
KW
751 (UTF8_WARN_ILLEGAL_C9_INTERCHANGE|UTF8_WARN_NONCHAR)
752
93e6dbd6
KW
753#define UTF8_ALLOW_ANY \
754 (~( UTF8_DISALLOW_ILLEGAL_INTERCHANGE|UTF8_DISALLOW_ABOVE_31_BIT \
755 |UTF8_WARN_ILLEGAL_INTERCHANGE|UTF8_WARN_ABOVE_31_BIT))
949cf498
KW
756#define UTF8_ALLOW_ANYUV \
757 (UTF8_ALLOW_EMPTY \
758 & ~(UTF8_DISALLOW_ILLEGAL_INTERCHANGE|UTF8_WARN_ILLEGAL_INTERCHANGE))
9f7f3913
ST
759#define UTF8_ALLOW_DEFAULT (ckWARN(WARN_UTF8) ? 0 : \
760 UTF8_ALLOW_ANYUV)
1d72bdf6 761
89d986df
KW
762/*
763=for apidoc Am|bool|UTF8_IS_SURROGATE|const U8 *s|const U8 *e
764
765Evaluates to non-zero if the first few bytes of the string starting at C<s> and
766looking no further than S<C<e - 1>> are well-formed UTF-8 that represents one
767of the Unicode surrogate code points; otherwise it evaluates to 0. If
768non-zero, the value gives how many bytes starting at C<s> comprise the code
769point's representation.
770
771=cut
772 */
773#define UTF8_IS_SURROGATE(s, e) is_SURROGATE_utf8_safe(s, e)
774
775
776#define UTF8_IS_REPLACEMENT(s, send) is_REPLACEMENT_utf8_safe(s,send)
777
778/*
779=for apidoc Am|bool|UTF8_IS_SUPER|const U8 *s|const U8 *e
780
781Recall that Perl recognizes an extension to UTF-8 that can encode code
782points larger than the ones defined by Unicode, which are 0..0x10FFFF.
783
784This macro evaluates to non-zero if the first few bytes of the string starting
785at C<s> and looking no further than S<C<e - 1>> are from this UTF-8 extension;
786otherwise it evaluates to 0. If non-zero, the value gives how many bytes
787starting at C<s> comprise the code point's representation.
788
7890 is returned if the bytes are not well-formed extended UTF-8, or if they
790represent a code point that cannot fit in a UV on the current platform. Hence
791this macro can give different results when run on a 64-bit word machine than on
792one with a 32-bit word size.
0c58a72b 793
89d986df
KW
794Note that it is deprecated to have code points that are larger than what can
795fit in an IV on the current machine.
7131f24d 796
89d986df
KW
797=cut
798
799 * ASCII EBCDIC I8
7131f24d
KW
800 * U+10FFFF: \xF4\x8F\xBF\xBF \xF9\xA1\xBF\xBF\xBF max legal Unicode
801 * U+110000: \xF4\x90\x80\x80 \xF9\xA2\xA0\xA0\xA0
802 * U+110001: \xF4\x90\x80\x81 \xF9\xA2\xA0\xA0\xA1
89d986df
KW
803 */
804#ifdef EBCDIC
a14e0a36 805# define UTF8_IS_SUPER(s, e) \
89d986df
KW
806 (( LIKELY((e) > (s) + 4) \
807 && NATIVE_UTF8_TO_I8(*(s)) >= 0xF9 \
808 && ( NATIVE_UTF8_TO_I8(*(s)) > 0xF9 \
809 || (NATIVE_UTF8_TO_I8(*(s) + 1) >= 0xA2)) \
810 && LIKELY((s) + UTF8SKIP(s) <= (e))) \
2b479609 811 ? _is_utf8_char_helper(s, s + UTF8SKIP(s), 0) : 0)
7131f24d 812#else
a14e0a36 813# define UTF8_IS_SUPER(s, e) \
89d986df
KW
814 (( LIKELY((e) > (s) + 3) \
815 && (*(U8*) (s)) >= 0xF4 \
816 && ((*(U8*) (s)) > 0xF4 || (*((U8*) (s) + 1) >= 0x90))\
817 && LIKELY((s) + UTF8SKIP(s) <= (e))) \
2b479609 818 ? _is_utf8_char_helper(s, s + UTF8SKIP(s), 0) : 0)
7131f24d
KW
819#endif
820
b96a92fb
KW
821/* These are now machine generated, and the 'given' clause is no longer
822 * applicable */
0c58a72b 823#define UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s, e) \
89d986df
KW
824 cBOOL(is_NONCHAR_utf8_safe(s,e))
825
826/*
827=for apidoc Am|bool|UTF8_IS_NONCHAR|const U8 *s|const U8 *e
828
829Evaluates to non-zero if the first few bytes of the string starting at C<s> and
830looking no further than S<C<e - 1>> are well-formed UTF-8 that represents one
831of the Unicode non-character code points; otherwise it evaluates to 0. If
832non-zero, the value gives how many bytes starting at C<s> comprise the code
833point's representation.
834
835=cut
836 */
0c58a72b
KW
837#define UTF8_IS_NONCHAR(s, e) \
838 UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s, e)
7131f24d 839
c867b360
JH
840#define UNICODE_SURROGATE_FIRST 0xD800
841#define UNICODE_SURROGATE_LAST 0xDFFF
842#define UNICODE_REPLACEMENT 0xFFFD
843#define UNICODE_BYTE_ORDER_MARK 0xFEFF
1d72bdf6 844
b851fbc1 845/* Though our UTF-8 encoding can go beyond this,
c76687c5 846 * let's be conservative and do as Unicode says. */
b851fbc1
JH
847#define PERL_UNICODE_MAX 0x10FFFF
848
d35f2ca5
KW
849#define UNICODE_WARN_SURROGATE 0x0001 /* UTF-16 surrogates */
850#define UNICODE_WARN_NONCHAR 0x0002 /* Non-char code points */
851#define UNICODE_WARN_SUPER 0x0004 /* Above 0x10FFFF */
852#define UNICODE_WARN_ABOVE_31_BIT 0x0008 /* Above 0x7FFF_FFFF */
853#define UNICODE_DISALLOW_SURROGATE 0x0010
854#define UNICODE_DISALLOW_NONCHAR 0x0020
855#define UNICODE_DISALLOW_SUPER 0x0040
856#define UNICODE_DISALLOW_ABOVE_31_BIT 0x0080
ecc1615f
KW
857#define UNICODE_WARN_ILLEGAL_C9_INTERCHANGE \
858 (UNICODE_WARN_SURROGATE|UNICODE_WARN_SUPER)
bb88be5f 859#define UNICODE_WARN_ILLEGAL_INTERCHANGE \
ecc1615f
KW
860 (UNICODE_WARN_ILLEGAL_C9_INTERCHANGE|UNICODE_WARN_NONCHAR)
861#define UNICODE_DISALLOW_ILLEGAL_C9_INTERCHANGE \
862 (UNICODE_DISALLOW_SURROGATE|UNICODE_DISALLOW_SUPER)
bb88be5f 863#define UNICODE_DISALLOW_ILLEGAL_INTERCHANGE \
ecc1615f 864 (UNICODE_DISALLOW_ILLEGAL_C9_INTERCHANGE|UNICODE_DISALLOW_NONCHAR)
949cf498
KW
865
866/* For backward source compatibility, as are now the default */
867#define UNICODE_ALLOW_SURROGATE 0
868#define UNICODE_ALLOW_SUPER 0
869#define UNICODE_ALLOW_ANY 0
b851fbc1 870
2d6b3d38
KW
871/* This matches the 2048 code points between UNICODE_SURROGATE_FIRST (0xD800) and
872 * UNICODE_SURROGATE_LAST (0xDFFF) */
873#define UNICODE_IS_SURROGATE(uv) (((UV) (uv) & (~0xFFFF | 0xF800)) \
874 == 0xD800)
875
646d1759
KW
876#define UNICODE_IS_REPLACEMENT(uv) ((UV) (uv) == UNICODE_REPLACEMENT)
877#define UNICODE_IS_BYTE_ORDER_MARK(uv) ((UV) (uv) == UNICODE_BYTE_ORDER_MARK)
c149ab20
KW
878
879/* Is 'uv' one of the 32 contiguous-range noncharacters? */
880#define UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv) ((UV) (uv) >= 0xFDD0 \
881 && (UV) (uv) <= 0xFDEF)
882
883/* Is 'uv' one of the 34 plane-ending noncharacters 0xFFFE, 0xFFFF, 0x1FFFE,
884 * 0x1FFFF, ... 0x10FFFE, 0x10FFFF, given that we know that 'uv' is not above
885 * the Unicode legal max */
886#define UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv) \
887 (((UV) (uv) & 0xFFFE) == 0xFFFE)
888
889#define UNICODE_IS_NONCHAR(uv) \
890 ( UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv) \
891 || ( LIKELY( ! UNICODE_IS_SUPER(uv)) \
892 && UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv)))
893
894#define UNICODE_IS_SUPER(uv) ((UV) (uv) > PERL_UNICODE_MAX)
d35f2ca5 895#define UNICODE_IS_ABOVE_31_BIT(uv) ((UV) (uv) > 0x7FFFFFFF)
1d72bdf6 896
ec34087a
KW
897#define LATIN_SMALL_LETTER_SHARP_S LATIN_SMALL_LETTER_SHARP_S_NATIVE
898#define LATIN_SMALL_LETTER_Y_WITH_DIAERESIS \
899 LATIN_SMALL_LETTER_Y_WITH_DIAERESIS_NATIVE
900#define MICRO_SIGN MICRO_SIGN_NATIVE
901#define LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE \
902 LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE_NATIVE
903#define LATIN_SMALL_LETTER_A_WITH_RING_ABOVE \
904 LATIN_SMALL_LETTER_A_WITH_RING_ABOVE_NATIVE
09091399
JH
905#define UNICODE_GREEK_CAPITAL_LETTER_SIGMA 0x03A3
906#define UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA 0x03C2
907#define UNICODE_GREEK_SMALL_LETTER_SIGMA 0x03C3
9dcbe121 908#define GREEK_SMALL_LETTER_MU 0x03BC
9e682c18
KW
909#define GREEK_CAPITAL_LETTER_MU 0x039C /* Upper and title case
910 of MICRON */
911#define LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS 0x0178 /* Also is title case */
0766489e
KW
912#ifdef LATIN_CAPITAL_LETTER_SHARP_S_UTF8
913# define LATIN_CAPITAL_LETTER_SHARP_S 0x1E9E
914#endif
74894415
KW
915#define LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE 0x130
916#define LATIN_SMALL_LETTER_DOTLESS_I 0x131
9e682c18 917#define LATIN_SMALL_LETTER_LONG_S 0x017F
a9f50d33
KW
918#define LATIN_SMALL_LIGATURE_LONG_S_T 0xFB05
919#define LATIN_SMALL_LIGATURE_ST 0xFB06
9e682c18
KW
920#define KELVIN_SIGN 0x212A
921#define ANGSTROM_SIGN 0x212B
09091399 922
9e55ce06 923#define UNI_DISPLAY_ISPRINT 0x0001
c728cb41
JH
924#define UNI_DISPLAY_BACKSLASH 0x0002
925#define UNI_DISPLAY_QQ (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH)
926#define UNI_DISPLAY_REGEX (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH)
9e55ce06 927
5cd46e1f
KW
928#define ANYOF_FOLD_SHARP_S(node, input, end) \
929 (ANYOF_BITMAP_TEST(node, LATIN_SMALL_LETTER_SHARP_S) && \
137165a6 930 (ANYOF_NONBITMAP(node)) && \
39065660 931 (ANYOF_FLAGS(node) & ANYOF_LOC_NONBITMAP_FOLD) && \
07b6858f 932 ((end) > (input) + 1) && \
305b8651 933 isALPHA_FOLD_EQ((input)[0], 's'))
6302f837 934
ebc501f0 935#define SHARP_S_SKIP 2
3b0fc154 936
6302f837 937/*
5dca9278
KW
938
939=for apidoc Am|STRLEN|isUTF8_CHAR|const U8 *s|const U8 *e
940
35f8c9bd 941Evaluates to non-zero if the first few bytes of the string starting at C<s> and
2c6ed66c
KW
942looking no further than S<C<e - 1>> are well-formed UTF-8, as extended by Perl,
943that represents some code point; otherwise it evaluates to 0. If non-zero, the
944value gives how many bytes starting at C<s> comprise the code point's
945representation.
6302f837 946
35f8c9bd
KW
947The code point can be any that will fit in a UV on this machine, using Perl's
948extension to official UTF-8 to represent those higher than the Unicode maximum
949of 0x10FFFF. That means that this macro is used to efficiently decide if the
950next few bytes in C<s> is legal UTF-8 for a single character. Use
951L</is_utf8_string>(), L</is_utf8_string_loclen>(), and
952L</is_utf8_string_loc>() to check entire strings.
953
954Note that it is deprecated to use code points higher than what will fit in an
955IV. This macro does not raise any warnings for such code points, treating them
956as valid.
957
958Note also that a UTF-8 INVARIANT character (i.e. ASCII on non-EBCDIC machines)
959is a valid UTF-8 character.
5dca9278
KW
960
961=cut
962*/
6302f837 963
784d4f31
KW
964#define isUTF8_CHAR(s, e) \
965 (UNLIKELY((e) <= (s)) \
966 ? 0 \
967 : (UTF8_IS_INVARIANT(*s)) \
df863e43
KW
968 ? 1 \
969 : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \
970 ? 0 \
971 : LIKELY(NATIVE_UTF8_TO_I8(*s) <= _IS_UTF8_CHAR_HIGHEST_START_BYTE) \
972 ? is_UTF8_CHAR_utf8_no_length_checks(s) \
973 : _is_utf8_char_helper(s, e, 0))
6302f837 974
3cedd9d9
KW
975#define is_utf8_char_buf(buf, buf_end) isUTF8_CHAR(buf, buf_end)
976
e23e8bc1
KW
977/*
978
979=for apidoc Am|STRLEN|isSTRICT_UTF8_CHAR|const U8 *s|const U8 *e
980
981Evaluates to non-zero if the first few bytes of the string starting at C<s> and
982looking no further than S<C<e - 1>> are well-formed UTF-8 that represents some
983Unicode code point completely acceptable for open interchange between all
984applications; otherwise it evaluates to 0. If non-zero, the value gives how
985many bytes starting at C<s> comprise the code point's representation.
986
987The largest acceptable code point is the Unicode maximum 0x10FFFF, and must not
988be a surrogate nor a non-character code point. Thus this excludes any code
989point from Perl's extended UTF-8.
990
991This is used to efficiently decide if the next few bytes in C<s> is
992legal Unicode-acceptable UTF-8 for a single character.
993
994=cut
995*/
996
997#define isSTRICT_UTF8_CHAR(s, e) \
998 (UNLIKELY((e) <= (s)) \
999 ? 0 \
1000 : (UTF8_IS_INVARIANT(*s)) \
1001 ? 1 \
1002 : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \
1003 ? 0 \
1004 : is_STRICT_UTF8_CHAR_utf8_no_length_checks(s))
1005
6302f837
KW
1006/* Do not use; should be deprecated. Use isUTF8_CHAR() instead; this is
1007 * retained solely for backwards compatibility */
1008#define IS_UTF8_CHAR(p, n) (isUTF8_CHAR(p, (p) + (n)) == n)
e9a8c099 1009
57f0e7e2
KW
1010#endif /* H_UTF8 */
1011
e9a8c099 1012/*
14d04a33 1013 * ex: set ts=8 sts=4 sw=4 et:
e9a8c099 1014 */