This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
perldiag: A deprecation is now fatal
[perl5.git] / utf8.h
CommitLineData
a0ed51b3
LW
1/* utf8.h
2 *
f3cb6f94
KW
3 * This file contains definitions for use with the UTF-8 encoding. It
4 * actually also works with the variant UTF-8 encoding called UTF-EBCDIC, and
5 * hides almost all of the differences between these from the caller. In other
6 * words, someone should #include this file, and if the code is being compiled
7 * on an EBCDIC platform, things should mostly just work.
8 *
2eee27d7
SS
9 * Copyright (C) 2000, 2001, 2002, 2005, 2006, 2007, 2009,
10 * 2010, 2011 by Larry Wall and others
a0ed51b3
LW
11 *
12 * You may distribute under the terms of either the GNU General Public
13 * License or the Artistic License, as specified in the README file.
14 *
15 */
16
6a5bc5ac
KW
17#ifndef PERL_UTF8_H_ /* Guard against recursive inclusion */
18#define PERL_UTF8_H_ 1
57f0e7e2 19
39e02b42 20/* Use UTF-8 as the default script encoding?
1e54db1a 21 * Turning this on will break scripts having non-UTF-8 binary
39e02b42
JH
22 * data (such as Latin-1) in string literals. */
23#ifdef USE_UTF8_SCRIPTS
24# define USE_UTF8_IN_NAMES (!IN_BYTES)
25#else
26# define USE_UTF8_IN_NAMES (PL_hints & HINT_UTF8)
27#endif
28
3cd96634
KW
29#include "regcharclass.h"
30#include "unicode_constants.h"
31
051a06d4 32/* For to_utf8_fold_flags, q.v. */
e4f4ef45
KW
33#define FOLD_FLAGS_LOCALE 0x1
34#define FOLD_FLAGS_FULL 0x2
35#define FOLD_FLAGS_NOMIX_ASCII 0x4
051a06d4 36
7bbfa158
KW
37/*
38=head1 Unicode Support
8cca77bc
KW
39L<perlguts/Unicode Support> has an introduction to this API.
40
41See also L</Character classification>,
42and L</Character case changing>.
43Various functions outside this section also work specially with Unicode.
44Search for the string "utf8" in this document.
7bbfa158
KW
45
46=for apidoc is_ascii_string
47
8871a094 48This is a misleadingly-named synonym for L</is_utf8_invariant_string>.
7bbfa158
KW
49On ASCII-ish platforms, the name isn't misleading: the ASCII-range characters
50are exactly the UTF-8 invariants. But EBCDIC machines have more invariants
8871a094
KW
51than just the ASCII characters, so C<is_utf8_invariant_string> is preferred.
52
53=for apidoc is_invariant_string
54
55This is a somewhat misleadingly-named synonym for L</is_utf8_invariant_string>.
56C<is_utf8_invariant_string> is preferred, as it indicates under what conditions
57the string is invariant.
7bbfa158
KW
58
59=cut
60*/
8871a094
KW
61#define is_ascii_string(s, len) is_utf8_invariant_string(s, len)
62#define is_invariant_string(s, len) is_utf8_invariant_string(s, len)
7bbfa158 63
33f38593
KW
64#define uvoffuni_to_utf8_flags(d,uv,flags) \
65 uvoffuni_to_utf8_flags_msgs(d, uv, flags, 0)
de69f3af
KW
66#define uvchr_to_utf8(a,b) uvchr_to_utf8_flags(a,b,0)
67#define uvchr_to_utf8_flags(d,uv,flags) \
33f38593
KW
68 uvchr_to_utf8_flags_msgs(d,uv,flags, 0)
69#define uvchr_to_utf8_flags_msgs(d,uv,flags,msgs) \
70 uvoffuni_to_utf8_flags_msgs(d,NATIVE_TO_UNI(uv),flags, msgs)
de69f3af 71#define utf8_to_uvchr_buf(s, e, lenp) \
fdb72414
KW
72 (__ASSERT_((U8*) (e) > (U8*) (s)) \
73 utf8n_to_uvchr(s, (U8*)(e) - (U8*)(s), lenp, \
74 ckWARN_d(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY))
f9380377
KW
75#define utf8n_to_uvchr(s, len, lenp, flags) \
76 utf8n_to_uvchr_error(s, len, lenp, flags, 0)
37657a5b
KW
77#define utf8n_to_uvchr_error(s, len, lenp, flags, errors) \
78 utf8n_to_uvchr_msgs(s, len, lenp, flags, errors, 0)
de69f3af 79
a0270393 80#define to_uni_fold(c, p, lenp) _to_uni_fold_flags(c, p, lenp, FOLD_FLAGS_FULL)
a239b1e2
KW
81
82#define to_utf8_fold(s, r, lenr) \
607313a1 83 _to_utf8_fold_flags (s, NULL, r, lenr, FOLD_FLAGS_FULL, __FILE__, __LINE__)
a239b1e2 84#define to_utf8_lower(s, r, lenr) \
607313a1 85 _to_utf8_lower_flags(s, NULL, r ,lenr, 0, __FILE__, __LINE__)
a239b1e2 86#define to_utf8_upper(s, r, lenr) \
607313a1 87 _to_utf8_upper_flags(s, NULL, r, lenr, 0, __FILE__, __LINE__)
a239b1e2 88#define to_utf8_title(s, r, lenr) \
607313a1 89 _to_utf8_title_flags(s, NULL, r, lenr ,0, __FILE__, __LINE__)
36bb2ab6 90
eda9cac1
KW
91#define foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \
92 foldEQ_utf8_flags(s1, pe1, l1, u1, s2, pe2, l2, u2, 0)
baa60164 93#define FOLDEQ_UTF8_NOMIX_ASCII (1 << 0)
cea315b6 94#define FOLDEQ_LOCALE (1 << 1)
18f762c3
KW
95#define FOLDEQ_S1_ALREADY_FOLDED (1 << 2)
96#define FOLDEQ_S2_ALREADY_FOLDED (1 << 3)
d635b710
KW
97#define FOLDEQ_S1_FOLDS_SANE (1 << 4)
98#define FOLDEQ_S2_FOLDS_SANE (1 << 5)
a33c29bc 99
e6226b18
KW
100#define ibcmp_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \
101 cBOOL(! foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2))
102
1d72bdf6
NIS
103#ifdef EBCDIC
104/* The equivalent of these macros but implementing UTF-EBCDIC
105 are in the following header file:
106 */
107
108#include "utfebcdic.h"
fd7cb289 109
d06134e5 110#else /* ! EBCDIC */
73c4f7a1
GS
111START_EXTERN_C
112
111e8ed9
KW
113/* How wide can a single UTF-8 encoded character become in bytes. */
114/* NOTE: Strictly speaking Perl's UTF-8 should not be called UTF-8 since UTF-8
115 * is an encoding of Unicode, and Unicode's upper limit, 0x10FFFF, can be
116 * expressed with 4 bytes. However, Perl thinks of UTF-8 as a way to encode
117 * non-negative integers in a binary format, even those above Unicode */
118#define UTF8_MAXBYTES 13
119
a0ed51b3 120#ifdef DOINIT
6f06b55f 121EXTCONST unsigned char PL_utf8skip[] = {
b2635aa8
KW
122/* 0x00 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
123/* 0x10 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
124/* 0x20 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
125/* 0x30 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
126/* 0x40 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
127/* 0x50 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
128/* 0x60 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
129/* 0x70 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
130/* 0x80 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
131/* 0x90 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
132/* 0xA0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
133/* 0xB0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
134/* 0xC0 */ 2,2, /* overlong */
1ff3baa2 135/* 0xC2 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0080 to U+03FF */
b2635aa8
KW
136/* 0xD0 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0400 to U+07FF */
137/* 0xE0 */ 3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3, /* U+0800 to U+FFFF */
138/* 0xF0 */ 4,4,4,4,4,4,4,4,5,5,5,5,6,6, /* above BMP to 2**31 - 1 */
6937f885
KW
139 /* Perl extended (never was official UTF-8). Up to 36 bit */
140/* 0xFE */ 7,
141 /* More extended, Up to 72 bits (64-bit + reserved) */
111e8ed9 142/* 0xFF */ UTF8_MAXBYTES
a0ed51b3
LW
143};
144#else
6f06b55f 145EXTCONST unsigned char PL_utf8skip[];
a0ed51b3
LW
146#endif
147
73c4f7a1 148END_EXTERN_C
7e2040f0 149
1a3756de 150#if defined(_MSC_VER) && _MSC_VER < 1400
6f6d1bab
TC
151/* older MSVC versions have a smallish macro buffer */
152#define PERL_SMALL_MACRO_BUFFER
153#endif
154
59a449d5
KW
155/* Native character to/from iso-8859-1. Are the identity functions on ASCII
156 * platforms */
6f6d1bab 157#ifdef PERL_SMALL_MACRO_BUFFER
1a3756de
TC
158#define NATIVE_TO_LATIN1(ch) ((U8)(ch))
159#define LATIN1_TO_NATIVE(ch) ((U8)(ch))
6f6d1bab 160#else
4c8cd605
KW
161#define NATIVE_TO_LATIN1(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch)))
162#define LATIN1_TO_NATIVE(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch)))
6f6d1bab 163#endif
59a449d5
KW
164
165/* I8 is an intermediate version of UTF-8 used only in UTF-EBCDIC. We thus
166 * consider it to be identical to UTF-8 on ASCII platforms. Strictly speaking
167 * UTF-8 and UTF-EBCDIC are two different things, but we often conflate them
168 * because they are 8-bit encodings that serve the same purpose in Perl, and
169 * rarely do we need to distinguish them. The term "NATIVE_UTF8" applies to
170 * whichever one is applicable on the current platform */
6f6d1bab 171#ifdef PERL_SMALL_MACRO_BUFFER
1d4ea287
DD
172#define NATIVE_UTF8_TO_I8(ch) ((U8) (ch))
173#define I8_TO_NATIVE_UTF8(ch) ((U8) (ch))
6f6d1bab 174#else
4c8cd605
KW
175#define NATIVE_UTF8_TO_I8(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch)))
176#define I8_TO_NATIVE_UTF8(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch)))
6f6d1bab 177#endif
59a449d5 178
1d72bdf6 179/* Transforms in wide UV chars */
4c8cd605
KW
180#define UNI_TO_NATIVE(ch) ((UV) (ch))
181#define NATIVE_TO_UNI(ch) ((UV) (ch))
d7578b48 182
877d9f0d 183/*
9041c2e3 184
a14e0a36
KW
185 The following table is from Unicode 3.2, plus the Perl extensions for above
186 U+10FFFF
877d9f0d 187
a14e0a36 188 Code Points 1st Byte 2nd Byte 3rd 4th 5th 6th 7th 8th-13th
877d9f0d 189
375122d7 190 U+0000..U+007F 00..7F
e1b711da 191 U+0080..U+07FF * C2..DF 80..BF
a14e0a36
KW
192 U+0800..U+0FFF E0 * A0..BF 80..BF
193 U+1000..U+CFFF E1..EC 80..BF 80..BF
194 U+D000..U+D7FF ED 80..9F 80..BF
195 U+D800..U+DFFF ED A0..BF 80..BF (surrogates)
196 U+E000..U+FFFF EE..EF 80..BF 80..BF
197 U+10000..U+3FFFF F0 * 90..BF 80..BF 80..BF
198 U+40000..U+FFFFF F1..F3 80..BF 80..BF 80..BF
199 U+100000..U+10FFFF F4 80..8F 80..BF 80..BF
200 Below are above-Unicode code points
201 U+110000..U+13FFFF F4 90..BF 80..BF 80..BF
202 U+110000..U+1FFFFF F5..F7 80..BF 80..BF 80..BF
203 U+200000..U+FFFFFF F8 * 88..BF 80..BF 80..BF 80..BF
204U+1000000..U+3FFFFFF F9..FB 80..BF 80..BF 80..BF 80..BF
205U+4000000..U+3FFFFFFF FC * 84..BF 80..BF 80..BF 80..BF 80..BF
206U+40000000..U+7FFFFFFF FD 80..BF 80..BF 80..BF 80..BF 80..BF
207U+80000000..U+FFFFFFFFF FE * 82..BF 80..BF 80..BF 80..BF 80..BF 80..BF
208U+1000000000.. FF 80..BF 80..BF 80..BF 80..BF 80..BF * 81..BF 80..BF
877d9f0d 209
e1b711da 210Note the gaps before several of the byte entries above marked by '*'. These are
37e2e78e
KW
211caused by legal UTF-8 avoiding non-shortest encodings: it is technically
212possible to UTF-8-encode a single code point in different ways, but that is
213explicitly forbidden, and the shortest possible encoding should always be used
15824458 214(and that is what Perl does). The non-shortest ones are called 'overlongs'.
8c007b5a 215
877d9f0d
JH
216 */
217
8c007b5a
JH
218/*
219 Another way to look at it, as bits:
220
b2635aa8 221 Code Points 1st Byte 2nd Byte 3rd Byte 4th Byte
8c007b5a 222
b2635aa8
KW
223 0aaa aaaa 0aaa aaaa
224 0000 0bbb bbaa aaaa 110b bbbb 10aa aaaa
225 cccc bbbb bbaa aaaa 1110 cccc 10bb bbbb 10aa aaaa
226 00 000d ddcc cccc bbbb bbaa aaaa 1111 0ddd 10cc cccc 10bb bbbb 10aa aaaa
8c007b5a
JH
227
228As you can see, the continuation bytes all begin with C<10>, and the
e1b711da 229leading bits of the start byte tell how many bytes there are in the
8c007b5a
JH
230encoded character.
231
df863e43
KW
232Perl's extended UTF-8 means we can have start bytes up through FF, though any
233beginning with FF yields a code point that is too large for 32-bit ASCII
234platforms. FF signals to use 13 bytes for the encoded character. This breaks
235the paradigm that the number of leading bits gives how many total bytes there
236are in the character.
65ab9279 237
8c007b5a
JH
238*/
239
6c88483e 240/* Is the representation of the Unicode code point 'cp' the same regardless of
15824458 241 * being encoded in UTF-8 or not? */
2d1545e5 242#define OFFUNI_IS_INVARIANT(cp) isASCII(cp)
15824458 243
c2b32798
KW
244/*
245=for apidoc Am|bool|UVCHR_IS_INVARIANT|UV cp
246
247Evaluates to 1 if the representation of code point C<cp> is the same whether or
248not it is encoded in UTF-8; otherwise evaluates to 0. UTF-8 invariant
249characters can be copied as-is when converting to/from UTF-8, saving time.
250C<cp> is Unicode if above 255; otherwise is platform-native.
251
252=cut
253 */
254
cf1be84e 255#define UVCHR_IS_INVARIANT(cp) OFFUNI_IS_INVARIANT(cp)
38953e5a 256
c9264833
KW
257/* This defines the bits that are to be in the continuation bytes of a multi-byte
258 * UTF-8 encoded character that mark it is a continuation byte. */
259#define UTF_CONTINUATION_MARK 0x80
260
a95ec4fb 261/* Misleadingly named: is the UTF8-encoded byte 'c' part of a variant sequence
5c06326b
KW
262 * in UTF-8? This is the inverse of UTF8_IS_INVARIANT. The |0 makes sure this
263 * isn't mistakenly called with a ptr argument */
a6951642
KW
264#define UTF8_IS_CONTINUED(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
265 ((U8)((c) | 0)) & UTF_CONTINUATION_MARK)
15824458
KW
266
267/* Is the byte 'c' the first byte of a multi-byte UTF8-8 encoded sequence?
268 * This doesn't catch invariants (they are single-byte). It also excludes the
5c06326b
KW
269 * illegal overlong sequences that begin with C0 and C1. The |0 makes sure
270 * this isn't mistakenly called with a ptr argument */
a6951642
KW
271#define UTF8_IS_START(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
272 ((U8)((c) | 0)) >= 0xc2)
15824458 273
858cd8ab
KW
274/* For use in UTF8_IS_CONTINUATION() below */
275#define UTF_IS_CONTINUATION_MASK 0xC0
276
15824458 277/* Is the byte 'c' part of a multi-byte UTF8-8 encoded sequence, and not the
5c06326b
KW
278 * first byte thereof? The |0 makes sure this isn't mistakenly called with a
279 * ptr argument */
a6951642
KW
280#define UTF8_IS_CONTINUATION(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
281 (((U8)((c) | 0)) & UTF_IS_CONTINUATION_MASK) == UTF_CONTINUATION_MARK)
0ae1fa71 282
15824458
KW
283/* Is the UTF8-encoded byte 'c' the first byte of a two byte sequence? Use
284 * UTF8_IS_NEXT_CHAR_DOWNGRADEABLE() instead if the input isn't known to
285 * be well-formed. Masking with 0xfe allows the low bit to be 0 or 1; thus
5c06326b
KW
286 * this matches 0xc[23]. The |0 makes sure this isn't mistakenly called with a
287 * ptr argument */
a6951642
KW
288#define UTF8_IS_DOWNGRADEABLE_START(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
289 (((U8)((c) | 0)) & 0xfe) == 0xc2)
4ab10950 290
15824458 291/* Is the UTF8-encoded byte 'c' the first byte of a sequence of bytes that
5c06326b
KW
292 * represent a code point > 255? The |0 makes sure this isn't mistakenly
293 * called with a ptr argument */
a6951642
KW
294#define UTF8_IS_ABOVE_LATIN1(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
295 ((U8)((c) | 0)) >= 0xc4)
8850bf83 296
15824458
KW
297/* This is the number of low-order bits a continuation byte in a UTF-8 encoded
298 * sequence contributes to the specification of the code point. In the bit
299 * maps above, you see that the first 2 bits are a constant '10', leaving 6 of
300 * real information */
1d72bdf6 301#define UTF_ACCUMULATION_SHIFT 6
b2635aa8 302
fed423a5
KW
303/* ^? is defined to be DEL on ASCII systems. See the definition of toCTRL()
304 * for more */
305#define QUESTION_MARK_CTRL DEL_NATIVE
306
307/* Surrogates, non-character code points and above-Unicode code points are
308 * problematic in some contexts. This allows code that needs to check for
309 * those to to quickly exclude the vast majority of code points it will
310 * encounter */
a6951642
KW
311#define isUTF8_POSSIBLY_PROBLEMATIC(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
312 (U8) c >= 0xED)
fed423a5 313
57ff5f59
KW
314#define UNICODE_IS_PERL_EXTENDED(uv) UNLIKELY((UV) (uv) > 0x7FFFFFFF)
315
fed423a5
KW
316#endif /* EBCDIC vs ASCII */
317
318/* 2**UTF_ACCUMULATION_SHIFT - 1 */
319#define UTF_CONTINUATION_MASK ((U8) ((1U << UTF_ACCUMULATION_SHIFT) - 1))
320
7028aeba
KW
321/* Internal macro to be used only in this file to aid in constructing other
322 * publicly accessible macros.
323 * The number of bytes required to express this uv in UTF-8, for just those
324 * uv's requiring 2 through 6 bytes, as these are common to all platforms and
325 * word sizes. The number of bytes needed is given by the number of leading 1
326 * bits in the start byte. There are 32 start bytes that have 2 initial 1 bits
327 * (C0-DF); there are 16 that have 3 initial 1 bits (E0-EF); 8 that have 4
328 * initial 1 bits (F0-F8); 4 that have 5 initial 1 bits (F9-FB), and 2 that
329 * have 6 initial 1 bits (FC-FD). The largest number a string of n bytes can
330 * represent is (the number of possible start bytes for 'n')
331 * * (the number of possiblities for each start byte
332 * The latter in turn is
333 * 2 ** ( (how many continuation bytes there are)
334 * * (the number of bits of information each
335 * continuation byte holds))
336 *
337 * If we were on a platform where we could use a fast find first set bit
338 * instruction (or count leading zeros instruction) this could be replaced by
339 * using that to find the log2 of the uv, and divide that by the number of bits
340 * of information in each continuation byte, adjusting for large cases and how
341 * much information is in a start byte for that length */
72164d3a 342#define __COMMON_UNI_SKIP(uv) \
7028aeba
KW
343 (UV) (uv) < (32 * (1U << ( UTF_ACCUMULATION_SHIFT))) ? 2 : \
344 (UV) (uv) < (16 * (1U << (2 * UTF_ACCUMULATION_SHIFT))) ? 3 : \
345 (UV) (uv) < ( 8 * (1U << (3 * UTF_ACCUMULATION_SHIFT))) ? 4 : \
346 (UV) (uv) < ( 4 * (1U << (4 * UTF_ACCUMULATION_SHIFT))) ? 5 : \
347 (UV) (uv) < ( 2 * (1U << (5 * UTF_ACCUMULATION_SHIFT))) ? 6 :
72164d3a
KW
348
349/* Internal macro to be used only in this file.
350 * This adds to __COMMON_UNI_SKIP the details at this platform's upper range.
fed423a5 351 * For any-sized EBCDIC platforms, or 64-bit ASCII ones, we need one more test
72164d3a
KW
352 * to see if just 7 bytes is needed, or if the maximum is needed. For 32-bit
353 * ASCII platforms, everything is representable by 7 bytes */
fed423a5 354#if defined(UV_IS_QUAD) || defined(EBCDIC)
72164d3a 355# define __BASE_UNI_SKIP(uv) (__COMMON_UNI_SKIP(uv) \
7028aeba 356 (UV) (uv) < ((UV) 1U << (6 * UTF_ACCUMULATION_SHIFT)) ? 7 : UTF8_MAXBYTES)
1d68d6cd 357#else
72164d3a 358# define __BASE_UNI_SKIP(uv) (__COMMON_UNI_SKIP(uv) 7)
1d68d6cd
SC
359#endif
360
fed423a5
KW
361/* The next two macros use the base macro defined above, and add in the tests
362 * at the low-end of the range, for just 1 byte, yielding complete macros,
363 * publicly accessible. */
364
365/* Input is a true Unicode (not-native) code point */
366#define OFFUNISKIP(uv) (OFFUNI_IS_INVARIANT(uv) ? 1 : __BASE_UNI_SKIP(uv))
2084b489 367
5352a763
KW
368/*
369
370=for apidoc Am|STRLEN|UVCHR_SKIP|UV cp
371returns the number of bytes required to represent the code point C<cp> when
372encoded as UTF-8. C<cp> is a native (ASCII or EBCDIC) code point if less than
373255; a Unicode code point otherwise.
374
375=cut
376 */
fdb6583d 377#define UVCHR_SKIP(uv) ( UVCHR_IS_INVARIANT(uv) ? 1 : __BASE_UNI_SKIP(uv))
5352a763 378
b651802e
KW
379/* The largest code point representable by two UTF-8 bytes on this platform.
380 * As explained in the comments for __COMMON_UNI_SKIP, 32 start bytes with
fed423a5 381 * UTF_ACCUMULATION_SHIFT bits of information each */
aa206fb7
KW
382#define MAX_UTF8_TWO_BYTE (32 * (1U << UTF_ACCUMULATION_SHIFT) - 1)
383
b651802e
KW
384/* The largest code point representable by two UTF-8 bytes on any platform that
385 * Perl runs on. This value is constrained by EBCDIC which has 5 bits per
386 * continuation byte */
aa206fb7
KW
387#define MAX_PORTABLE_UTF8_TWO_BYTE (32 * (1U << 5) - 1)
388
c03c0950
KW
389/* The maximum number of UTF-8 bytes a single Unicode character can
390 * uppercase/lowercase/fold into. Unicode guarantees that the maximum
391 * expansion is UTF8_MAX_FOLD_CHAR_EXPAND characters, but any above-Unicode
392 * code point will fold to itself, so we only have to look at the expansion of
393 * the maximum Unicode code point. But this number may be less than the space
394 * occupied by a very large code point under Perl's extended UTF-8. We have to
395 * make it large enough to fit any single character. (It turns out that ASCII
396 * and EBCDIC differ in which is larger) */
397#define UTF8_MAXBYTES_CASE \
398 (UTF8_MAXBYTES >= (UTF8_MAX_FOLD_CHAR_EXPAND * OFFUNISKIP(0x10FFFF)) \
399 ? UTF8_MAXBYTES \
400 : (UTF8_MAX_FOLD_CHAR_EXPAND * OFFUNISKIP(0x10FFFF)))
401
d06134e5
KW
402/* Rest of these are attributes of Unicode and perl's internals rather than the
403 * encoding, or happen to be the same in both ASCII and EBCDIC (at least at
404 * this level; the macros that some of these call may have different
405 * definitions in the two encodings */
406
59a449d5
KW
407/* In domain restricted to ASCII, these may make more sense to the reader than
408 * the ones with Latin1 in the name */
409#define NATIVE_TO_ASCII(ch) NATIVE_TO_LATIN1(ch)
410#define ASCII_TO_NATIVE(ch) LATIN1_TO_NATIVE(ch)
411
412/* More or less misleadingly-named defines, retained for back compat */
413#define NATIVE_TO_UTF(ch) NATIVE_UTF8_TO_I8(ch)
414#define NATIVE_TO_I8(ch) NATIVE_UTF8_TO_I8(ch)
415#define UTF_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch)
416#define I8_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch)
417#define NATIVE8_TO_UNI(ch) NATIVE_TO_LATIN1(ch)
d06134e5 418
c0236afe 419/* This defines the 1-bits that are to be in the first byte of a multi-byte
97d0ceda
KW
420 * UTF-8 encoded character that mark it as a start byte and give the number of
421 * bytes that comprise the character. 'len' is the number of bytes in the
422 * multi-byte sequence. */
c0236afe
KW
423#define UTF_START_MARK(len) (((len) > 7) ? 0xFF : (0xFF & (0xFE << (7-(len)))))
424
425/* Masks out the initial one bits in a start byte, leaving the real data ones.
426 * Doesn't work on an invariant byte. 'len' is the number of bytes in the
427 * multi-byte sequence that comprises the character. */
428#define UTF_START_MASK(len) (((len) >= 7) ? 0x00 : (0x1F >> ((len)-2)))
429
537124e4
KW
430/* Adds a UTF8 continuation byte 'new' of information to a running total code
431 * point 'old' of all the continuation bytes so far. This is designed to be
155d2738
KW
432 * used in a loop to convert from UTF-8 to the code point represented. Note
433 * that this is asymmetric on EBCDIC platforms, in that the 'new' parameter is
434 * the UTF-EBCDIC byte, whereas the 'old' parameter is a Unicode (not EBCDIC)
435 * code point in process of being generated */
a6951642
KW
436#define UTF8_ACCUMULATE(old, new) (__ASSERT_(FITS_IN_8_BITS(new)) \
437 ((old) << UTF_ACCUMULATION_SHIFT) \
155d2738
KW
438 | ((NATIVE_UTF8_TO_I8((U8)new)) \
439 & UTF_CONTINUATION_MASK))
d06134e5 440
4ab10950 441/* This works in the face of malformed UTF-8. */
4e1ed312
KW
442#define UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(s, e) \
443 ( UTF8_IS_DOWNGRADEABLE_START(*(s)) \
444 && ( (e) - (s) > 1) \
445 && UTF8_IS_CONTINUATION(*((s)+1)))
4ab10950 446
5aaebcb3 447/* Number of bytes a code point occupies in UTF-8. */
5352a763 448#define NATIVE_SKIP(uv) UVCHR_SKIP(uv)
bd18bd40 449
5aaebcb3
KW
450/* Most code which says UNISKIP is really thinking in terms of native code
451 * points (0-255) plus all those beyond. This is an imprecise term, but having
2accb712 452 * it means existing code continues to work. For precision, use UVCHR_SKIP,
5352a763
KW
453 * NATIVE_SKIP, or OFFUNISKIP */
454#define UNISKIP(uv) UVCHR_SKIP(uv)
5aaebcb3 455
3c0792e4
KW
456/* Longer, but more accurate name */
457#define UTF8_IS_ABOVE_LATIN1_START(c) UTF8_IS_ABOVE_LATIN1(c)
458
a62b247b
KW
459/* Convert a UTF-8 variant Latin1 character to a native code point value.
460 * Needs just one iteration of accumulate. Should be used only if it is known
461 * that the code point is < 256, and is not UTF-8 invariant. Use the slower
462 * but more general TWO_BYTE_UTF8_TO_NATIVE() which handles any code point
463 * representable by two bytes (which turns out to be up through
464 * MAX_PORTABLE_UTF8_TWO_BYTE). The two parameters are:
465 * HI: a downgradable start byte;
466 * LO: continuation.
467 * */
468#define EIGHT_BIT_UTF8_TO_NATIVE(HI, LO) \
469 ( __ASSERT_(UTF8_IS_DOWNGRADEABLE_START(HI)) \
470 __ASSERT_(UTF8_IS_CONTINUATION(LO)) \
471 LATIN1_TO_NATIVE(UTF8_ACCUMULATE(( \
472 NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), (LO))))
473
94bb8c36 474/* Convert a two (not one) byte utf8 character to a native code point value.
2950f2a7
KW
475 * Needs just one iteration of accumulate. Should not be used unless it is
476 * known that the two bytes are legal: 1) two-byte start, and 2) continuation.
477 * Note that the result can be larger than 255 if the input character is not
478 * downgradable */
94bb8c36 479#define TWO_BYTE_UTF8_TO_NATIVE(HI, LO) \
a6951642
KW
480 (__ASSERT_(FITS_IN_8_BITS(HI)) \
481 __ASSERT_(FITS_IN_8_BITS(LO)) \
482 __ASSERT_(PL_utf8skip[HI] == 2) \
483 __ASSERT_(UTF8_IS_CONTINUATION(LO)) \
94bb8c36 484 UNI_TO_NATIVE(UTF8_ACCUMULATE((NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), \
635e76f5 485 (LO))))
94bb8c36
KW
486
487/* Should never be used, and be deprecated */
488#define TWO_BYTE_UTF8_TO_UNI(HI, LO) NATIVE_TO_UNI(TWO_BYTE_UTF8_TO_NATIVE(HI, LO))
2950f2a7 489
bd18bd40
KW
490/*
491
492=for apidoc Am|STRLEN|UTF8SKIP|char* s
493returns the number of bytes in the UTF-8 encoded character whose first (perhaps
494only) byte is pointed to by C<s>.
495
496=cut
497 */
2a70536e
KW
498#define UTF8SKIP(s) PL_utf8skip[*(const U8*)(s)]
499#define UTF8_SKIP(s) UTF8SKIP(s)
d06134e5 500
2d1545e5
KW
501/* Most code that says 'UNI_' really means the native value for code points up
502 * through 255 */
503#define UNI_IS_INVARIANT(cp) UVCHR_IS_INVARIANT(cp)
504
c2b32798
KW
505/*
506=for apidoc Am|bool|UTF8_IS_INVARIANT|char c
507
508Evaluates to 1 if the byte C<c> represents the same character when encoded in
509UTF-8 as when not; otherwise evaluates to 0. UTF-8 invariant characters can be
510copied as-is when converting to/from UTF-8, saving time.
511
512In spite of the name, this macro gives the correct result if the input string
513from which C<c> comes is not encoded in UTF-8.
514
515See C<L</UVCHR_IS_INVARIANT>> for checking if a UV is invariant.
516
517=cut
518
519The reason it works on both UTF-8 encoded strings and non-UTF-8 encoded, is
520that it returns TRUE in each for the exact same set of bit patterns. It is
521valid on a subset of what UVCHR_IS_INVARIANT is valid on, so can just use that;
522and the compiler should optimize out anything extraneous given the
523implementation of the latter. The |0 makes sure this isn't mistakenly called
524with a ptr argument.
525*/
5c06326b 526#define UTF8_IS_INVARIANT(c) UVCHR_IS_INVARIANT((c) | 0)
5fc230f1
KW
527
528/* Like the above, but its name implies a non-UTF8 input, which as the comments
529 * above show, doesn't matter as to its implementation */
38953e5a 530#define NATIVE_BYTE_IS_INVARIANT(c) UVCHR_IS_INVARIANT(c)
d06134e5 531
48ccf5e1
KW
532/* The macros in the next 4 sets are used to generate the two utf8 or utfebcdic
533 * bytes from an ordinal that is known to fit into exactly two (not one) bytes;
534 * it must be less than 0x3FF to work across both encodings. */
535
536/* These two are helper macros for the other three sets, and should not be used
537 * directly anywhere else. 'translate_function' is either NATIVE_TO_LATIN1
1ff3baa2
KW
538 * (which works for code points up through 0xFF) or NATIVE_TO_UNI which works
539 * for any code point */
48ccf5e1 540#define __BASE_TWO_BYTE_HI(c, translate_function) \
2863dafa 541 (__ASSERT_(! UVCHR_IS_INVARIANT(c)) \
48ccf5e1 542 I8_TO_NATIVE_UTF8((translate_function(c) >> UTF_ACCUMULATION_SHIFT) \
2863dafa 543 | UTF_START_MARK(2)))
48ccf5e1 544#define __BASE_TWO_BYTE_LO(c, translate_function) \
2863dafa 545 (__ASSERT_(! UVCHR_IS_INVARIANT(c)) \
48ccf5e1 546 I8_TO_NATIVE_UTF8((translate_function(c) & UTF_CONTINUATION_MASK) \
2863dafa 547 | UTF_CONTINUATION_MARK))
48ccf5e1 548
48ccf5e1
KW
549/* The next two macros should not be used. They were designed to be usable as
550 * the case label of a switch statement, but this doesn't work for EBCDIC. Use
9d0d3a03 551 * regen/unicode_constants.pl instead */
48ccf5e1
KW
552#define UTF8_TWO_BYTE_HI_nocast(c) __BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI)
553#define UTF8_TWO_BYTE_LO_nocast(c) __BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI)
554
555/* The next two macros are used when the source should be a single byte
556 * character; checked for under DEBUGGING */
557#define UTF8_EIGHT_BIT_HI(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
4c8cd605 558 ( __BASE_TWO_BYTE_HI(c, NATIVE_TO_LATIN1)))
48ccf5e1 559#define UTF8_EIGHT_BIT_LO(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
4c8cd605 560 (__BASE_TWO_BYTE_LO(c, NATIVE_TO_LATIN1)))
48ccf5e1
KW
561
562/* These final two macros in the series are used when the source can be any
563 * code point whose UTF-8 is known to occupy 2 bytes; they are less efficient
564 * than the EIGHT_BIT versions on EBCDIC platforms. We use the logical '~'
565 * operator instead of "<=" to avoid getting compiler warnings.
d52b8576 566 * MAX_UTF8_TWO_BYTE should be exactly all one bits in the lower few
48ccf5e1
KW
567 * places, so the ~ works */
568#define UTF8_TWO_BYTE_HI(c) \
569 (__ASSERT_((sizeof(c) == 1) \
d52b8576 570 || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \
4c8cd605 571 (__BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI)))
48ccf5e1
KW
572#define UTF8_TWO_BYTE_LO(c) \
573 (__ASSERT_((sizeof(c) == 1) \
d52b8576 574 || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \
4c8cd605 575 (__BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI)))
d06134e5 576
e7214ce8
KW
577/* This is illegal in any well-formed UTF-8 in both EBCDIC and ASCII
578 * as it is only in overlongs. */
579#define ILLEGAL_UTF8_BYTE I8_TO_NATIVE_UTF8(0xC1)
580
7e2040f0 581/*
e3036cf4 582 * 'UTF' is whether or not p is encoded in UTF8. The names 'foo_lazy_if' stem
20df05f4
KW
583 * from an earlier version of these macros in which they didn't call the
584 * foo_utf8() macros (i.e. were 'lazy') unless they decided that *p is the
585 * beginning of a utf8 character. Now that foo_utf8() determines that itself,
586 * no need to do it again here
7e2040f0 587 */
34aeb2e9
KW
588#define isIDFIRST_lazy_if(p,UTF) \
589 _is_utf8_FOO(_CC_IDFIRST, (const U8 *) p, "isIDFIRST_lazy_if", \
590 "isIDFIRST_lazy_if_safe", \
591 cBOOL(UTF && ! IN_BYTES), 0, __FILE__,__LINE__)
1d72bdf6 592
da8c1a98
KW
593#define isIDFIRST_lazy_if_safe(p, e, UTF) \
594 ((IN_BYTES || !UTF) \
595 ? isIDFIRST(*(p)) \
596 : isIDFIRST_utf8_safe(p, e))
597
34aeb2e9
KW
598#define isWORDCHAR_lazy_if(p,UTF) \
599 _is_utf8_FOO(_CC_IDFIRST, (const U8 *) p, "isWORDCHAR_lazy_if", \
600 "isWORDCHAR_lazy_if_safe", \
601 cBOOL(UTF && ! IN_BYTES), 0, __FILE__,__LINE__)
602
da8c1a98
KW
603#define isWORDCHAR_lazy_if_safe(p, e, UTF) \
604 ((IN_BYTES || !UTF) \
605 ? isWORDCHAR(*(p)) \
606 : isWORDCHAR_utf8_safe((U8 *) p, (U8 *) e))
607
34aeb2e9
KW
608#define isALNUM_lazy_if(p,UTF) \
609 _is_utf8_FOO(_CC_IDFIRST, (const U8 *) p, "isALNUM_lazy_if", \
610 "isWORDCHAR_lazy_if_safe", \
611 cBOOL(UTF && ! IN_BYTES), 0, __FILE__,__LINE__)
da8c1a98 612
89ebb4a3
JH
613#define UTF8_MAXLEN UTF8_MAXBYTES
614
8cb75cc8
KW
615/* A Unicode character can fold to up to 3 characters */
616#define UTF8_MAX_FOLD_CHAR_EXPAND 3
617
d3481830 618#define IN_BYTES UNLIKELY(CopHINTS_get(PL_curcop) & HINT_BYTES)
bd18bd40
KW
619
620/*
621
622=for apidoc Am|bool|DO_UTF8|SV* sv
623Returns a bool giving whether or not the PV in C<sv> is to be treated as being
624encoded in UTF-8.
625
626You should use this I<after> a call to C<SvPV()> or one of its variants, in
627case any call to string overloading updates the internal UTF-8 encoding flag.
628
629=cut
630*/
0064a8a9 631#define DO_UTF8(sv) (SvUTF8(sv) && !IN_BYTES)
1ff3baa2
KW
632
633/* Should all strings be treated as Unicode, and not just UTF-8 encoded ones?
634 * Is so within 'feature unicode_strings' or 'locale :not_characters', and not
635 * within 'use bytes'. UTF-8 locales are not tested for here, but perhaps
636 * could be */
70844984
KW
637#define IN_UNI_8_BIT \
638 (( ( (CopHINTS_get(PL_curcop) & HINT_UNI_8_BIT)) \
639 || ( CopHINTS_get(PL_curcop) & HINT_LOCALE_PARTIAL \
640 /* -1 below is for :not_characters */ \
641 && _is_in_locale_category(FALSE, -1))) \
642 && (! IN_BYTES))
b36bf33f 643
1d72bdf6 644
c76687c5 645#define UTF8_ALLOW_EMPTY 0x0001 /* Allow a zero length string */
2b5e7bc2 646#define UTF8_GOT_EMPTY UTF8_ALLOW_EMPTY
c76687c5
KW
647
648/* Allow first byte to be a continuation byte */
1d72bdf6 649#define UTF8_ALLOW_CONTINUATION 0x0002
2b5e7bc2 650#define UTF8_GOT_CONTINUATION UTF8_ALLOW_CONTINUATION
c76687c5 651
cd01d3b1 652/* Unexpected non-continuation byte */
1d72bdf6 653#define UTF8_ALLOW_NON_CONTINUATION 0x0004
2b5e7bc2 654#define UTF8_GOT_NON_CONTINUATION UTF8_ALLOW_NON_CONTINUATION
949cf498
KW
655
656/* expecting more bytes than were available in the string */
657#define UTF8_ALLOW_SHORT 0x0008
2b5e7bc2 658#define UTF8_GOT_SHORT UTF8_ALLOW_SHORT
949cf498 659
94953955
KW
660/* Overlong sequence; i.e., the code point can be specified in fewer bytes.
661 * First one will convert the overlong to the REPLACEMENT CHARACTER; second
662 * will return what the overlong evaluates to */
949cf498 663#define UTF8_ALLOW_LONG 0x0010
94953955 664#define UTF8_ALLOW_LONG_AND_ITS_VALUE (UTF8_ALLOW_LONG|0x0020)
2b5e7bc2
KW
665#define UTF8_GOT_LONG UTF8_ALLOW_LONG
666
d60baaa7
KW
667#define UTF8_ALLOW_OVERFLOW 0x0080
668#define UTF8_GOT_OVERFLOW UTF8_ALLOW_OVERFLOW
949cf498 669
f180b292 670#define UTF8_DISALLOW_SURROGATE 0x0100 /* Unicode surrogates */
2b5e7bc2 671#define UTF8_GOT_SURROGATE UTF8_DISALLOW_SURROGATE
f180b292 672#define UTF8_WARN_SURROGATE 0x0200
949cf498 673
c4e96019
KW
674/* Unicode non-character code points */
675#define UTF8_DISALLOW_NONCHAR 0x0400
2b5e7bc2 676#define UTF8_GOT_NONCHAR UTF8_DISALLOW_NONCHAR
c4e96019 677#define UTF8_WARN_NONCHAR 0x0800
949cf498 678
c4e96019
KW
679/* Super-set of Unicode: code points above the legal max */
680#define UTF8_DISALLOW_SUPER 0x1000
2b5e7bc2 681#define UTF8_GOT_SUPER UTF8_DISALLOW_SUPER
c4e96019
KW
682#define UTF8_WARN_SUPER 0x2000
683
684/* The original UTF-8 standard did not define UTF-8 with start bytes of 0xFE or
685 * 0xFF, though UTF-EBCDIC did. This allowed both versions to represent code
686 * points up to 2 ** 31 - 1. Perl extends UTF-8 so that 0xFE and 0xFF are
687 * usable on ASCII platforms, and 0xFF means something different than
688 * UTF-EBCDIC defines. These changes allow code points of 64 bits (actually
689 * somewhat more) to be represented on both platforms. But these are Perl
690 * extensions, and not likely to be interchangeable with other languages. Note
691 * that on ASCII platforms, FE overflows a signed 32-bit word, and FF an
692 * unsigned one. */
d044b7a7
KW
693#define UTF8_DISALLOW_PERL_EXTENDED 0x4000
694#define UTF8_GOT_PERL_EXTENDED UTF8_DISALLOW_PERL_EXTENDED
695#define UTF8_WARN_PERL_EXTENDED 0x8000
d35f2ca5 696
57ff5f59
KW
697/* For back compat, these old names are misleading for overlongs and
698 * UTF_EBCDIC. */
d044b7a7
KW
699#define UTF8_DISALLOW_ABOVE_31_BIT UTF8_DISALLOW_PERL_EXTENDED
700#define UTF8_GOT_ABOVE_31_BIT UTF8_GOT_PERL_EXTENDED
701#define UTF8_WARN_ABOVE_31_BIT UTF8_WARN_PERL_EXTENDED
702#define UTF8_DISALLOW_FE_FF UTF8_DISALLOW_PERL_EXTENDED
703#define UTF8_WARN_FE_FF UTF8_WARN_PERL_EXTENDED
949cf498 704
f180b292 705#define UTF8_CHECK_ONLY 0x10000
99a765e9 706#define _UTF8_NO_CONFIDENCE_IN_CURLEN 0x20000 /* Internal core use only */
949cf498
KW
707
708/* For backwards source compatibility. They do nothing, as the default now
709 * includes what they used to mean. The first one's meaning was to allow the
710 * just the single non-character 0xFFFF */
711#define UTF8_ALLOW_FFFF 0
c825ef8c 712#define UTF8_ALLOW_FE_FF 0
949cf498
KW
713#define UTF8_ALLOW_SURROGATE 0
714
ecc1615f
KW
715/* C9 refers to Unicode Corrigendum #9: allows but discourages non-chars */
716#define UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE \
717 (UTF8_DISALLOW_SUPER|UTF8_DISALLOW_SURROGATE)
718#define UTF8_WARN_ILLEGAL_C9_INTERCHANGE (UTF8_WARN_SUPER|UTF8_WARN_SURROGATE)
719
d35f2ca5 720#define UTF8_DISALLOW_ILLEGAL_INTERCHANGE \
ecc1615f 721 (UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE|UTF8_DISALLOW_NONCHAR)
949cf498 722#define UTF8_WARN_ILLEGAL_INTERCHANGE \
ecc1615f
KW
723 (UTF8_WARN_ILLEGAL_C9_INTERCHANGE|UTF8_WARN_NONCHAR)
724
0eb3d6a0
KW
725/* This is typically used for code that processes UTF-8 input and doesn't want
726 * to have to deal with any malformations that might be present. All such will
727 * be safely replaced by the REPLACEMENT CHARACTER, unless other flags
728 * overriding this are also present. */
2d532c27
KW
729#define UTF8_ALLOW_ANY ( UTF8_ALLOW_CONTINUATION \
730 |UTF8_ALLOW_NON_CONTINUATION \
731 |UTF8_ALLOW_SHORT \
d60baaa7
KW
732 |UTF8_ALLOW_LONG \
733 |UTF8_ALLOW_OVERFLOW)
2d532c27
KW
734
735/* Accept any Perl-extended UTF-8 that evaluates to any UV on the platform, but
cd01d3b1 736 * not any malformed. This is the default. */
2d532c27
KW
737#define UTF8_ALLOW_ANYUV 0
738#define UTF8_ALLOW_DEFAULT UTF8_ALLOW_ANYUV
1d72bdf6 739
89d986df
KW
740/*
741=for apidoc Am|bool|UTF8_IS_SURROGATE|const U8 *s|const U8 *e
742
743Evaluates to non-zero if the first few bytes of the string starting at C<s> and
744looking no further than S<C<e - 1>> are well-formed UTF-8 that represents one
745of the Unicode surrogate code points; otherwise it evaluates to 0. If
746non-zero, the value gives how many bytes starting at C<s> comprise the code
747point's representation.
748
749=cut
750 */
751#define UTF8_IS_SURROGATE(s, e) is_SURROGATE_utf8_safe(s, e)
752
753
754#define UTF8_IS_REPLACEMENT(s, send) is_REPLACEMENT_utf8_safe(s,send)
755
285aa1f0 756#define MAX_LEGAL_CP ((UV)IV_MAX)
40606899 757
89d986df
KW
758/*
759=for apidoc Am|bool|UTF8_IS_SUPER|const U8 *s|const U8 *e
760
761Recall that Perl recognizes an extension to UTF-8 that can encode code
762points larger than the ones defined by Unicode, which are 0..0x10FFFF.
763
764This macro evaluates to non-zero if the first few bytes of the string starting
765at C<s> and looking no further than S<C<e - 1>> are from this UTF-8 extension;
766otherwise it evaluates to 0. If non-zero, the value gives how many bytes
767starting at C<s> comprise the code point's representation.
768
7690 is returned if the bytes are not well-formed extended UTF-8, or if they
770represent a code point that cannot fit in a UV on the current platform. Hence
771this macro can give different results when run on a 64-bit word machine than on
772one with a 32-bit word size.
0c58a72b 773
891fd405 774Note that it is illegal to have code points that are larger than what can
89d986df 775fit in an IV on the current machine.
7131f24d 776
89d986df
KW
777=cut
778
779 * ASCII EBCDIC I8
7131f24d
KW
780 * U+10FFFF: \xF4\x8F\xBF\xBF \xF9\xA1\xBF\xBF\xBF max legal Unicode
781 * U+110000: \xF4\x90\x80\x80 \xF9\xA2\xA0\xA0\xA0
782 * U+110001: \xF4\x90\x80\x81 \xF9\xA2\xA0\xA0\xA1
89d986df
KW
783 */
784#ifdef EBCDIC
a14e0a36 785# define UTF8_IS_SUPER(s, e) \
89d986df
KW
786 (( LIKELY((e) > (s) + 4) \
787 && NATIVE_UTF8_TO_I8(*(s)) >= 0xF9 \
788 && ( NATIVE_UTF8_TO_I8(*(s)) > 0xF9 \
3d42f267 789 || (NATIVE_UTF8_TO_I8(*((s) + 1)) >= 0xA2)) \
89d986df 790 && LIKELY((s) + UTF8SKIP(s) <= (e))) \
2b479609 791 ? _is_utf8_char_helper(s, s + UTF8SKIP(s), 0) : 0)
7131f24d 792#else
a14e0a36 793# define UTF8_IS_SUPER(s, e) \
89d986df
KW
794 (( LIKELY((e) > (s) + 3) \
795 && (*(U8*) (s)) >= 0xF4 \
796 && ((*(U8*) (s)) > 0xF4 || (*((U8*) (s) + 1) >= 0x90))\
797 && LIKELY((s) + UTF8SKIP(s) <= (e))) \
2b479609 798 ? _is_utf8_char_helper(s, s + UTF8SKIP(s), 0) : 0)
7131f24d
KW
799#endif
800
b96a92fb
KW
801/* These are now machine generated, and the 'given' clause is no longer
802 * applicable */
0c58a72b 803#define UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s, e) \
89d986df
KW
804 cBOOL(is_NONCHAR_utf8_safe(s,e))
805
806/*
807=for apidoc Am|bool|UTF8_IS_NONCHAR|const U8 *s|const U8 *e
808
809Evaluates to non-zero if the first few bytes of the string starting at C<s> and
810looking no further than S<C<e - 1>> are well-formed UTF-8 that represents one
811of the Unicode non-character code points; otherwise it evaluates to 0. If
812non-zero, the value gives how many bytes starting at C<s> comprise the code
813point's representation.
814
815=cut
816 */
0c58a72b
KW
817#define UTF8_IS_NONCHAR(s, e) \
818 UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s, e)
7131f24d 819
c867b360
JH
820#define UNICODE_SURROGATE_FIRST 0xD800
821#define UNICODE_SURROGATE_LAST 0xDFFF
822#define UNICODE_REPLACEMENT 0xFFFD
823#define UNICODE_BYTE_ORDER_MARK 0xFEFF
1d72bdf6 824
b851fbc1 825/* Though our UTF-8 encoding can go beyond this,
c76687c5 826 * let's be conservative and do as Unicode says. */
b851fbc1
JH
827#define PERL_UNICODE_MAX 0x10FFFF
828
d044b7a7
KW
829#define UNICODE_WARN_SURROGATE 0x0001 /* UTF-16 surrogates */
830#define UNICODE_WARN_NONCHAR 0x0002 /* Non-char code points */
831#define UNICODE_WARN_SUPER 0x0004 /* Above 0x10FFFF */
832#define UNICODE_WARN_PERL_EXTENDED 0x0008 /* Above 0x7FFF_FFFF */
833#define UNICODE_WARN_ABOVE_31_BIT UNICODE_WARN_PERL_EXTENDED
834#define UNICODE_DISALLOW_SURROGATE 0x0010
835#define UNICODE_DISALLOW_NONCHAR 0x0020
836#define UNICODE_DISALLOW_SUPER 0x0040
837#define UNICODE_DISALLOW_PERL_EXTENDED 0x0080
838#define UNICODE_DISALLOW_ABOVE_31_BIT UNICODE_DISALLOW_PERL_EXTENDED
33f38593
KW
839
840#define UNICODE_GOT_SURROGATE UNICODE_DISALLOW_SURROGATE
841#define UNICODE_GOT_NONCHAR UNICODE_DISALLOW_NONCHAR
842#define UNICODE_GOT_SUPER UNICODE_DISALLOW_SUPER
843#define UNICODE_GOT_PERL_EXTENDED UNICODE_DISALLOW_PERL_EXTENDED
844
ecc1615f
KW
845#define UNICODE_WARN_ILLEGAL_C9_INTERCHANGE \
846 (UNICODE_WARN_SURROGATE|UNICODE_WARN_SUPER)
bb88be5f 847#define UNICODE_WARN_ILLEGAL_INTERCHANGE \
ecc1615f
KW
848 (UNICODE_WARN_ILLEGAL_C9_INTERCHANGE|UNICODE_WARN_NONCHAR)
849#define UNICODE_DISALLOW_ILLEGAL_C9_INTERCHANGE \
850 (UNICODE_DISALLOW_SURROGATE|UNICODE_DISALLOW_SUPER)
bb88be5f 851#define UNICODE_DISALLOW_ILLEGAL_INTERCHANGE \
ecc1615f 852 (UNICODE_DISALLOW_ILLEGAL_C9_INTERCHANGE|UNICODE_DISALLOW_NONCHAR)
949cf498
KW
853
854/* For backward source compatibility, as are now the default */
855#define UNICODE_ALLOW_SURROGATE 0
856#define UNICODE_ALLOW_SUPER 0
857#define UNICODE_ALLOW_ANY 0
b851fbc1 858
2d6b3d38
KW
859/* This matches the 2048 code points between UNICODE_SURROGATE_FIRST (0xD800) and
860 * UNICODE_SURROGATE_LAST (0xDFFF) */
861#define UNICODE_IS_SURROGATE(uv) (((UV) (uv) & (~0xFFFF | 0xF800)) \
862 == 0xD800)
863
646d1759
KW
864#define UNICODE_IS_REPLACEMENT(uv) ((UV) (uv) == UNICODE_REPLACEMENT)
865#define UNICODE_IS_BYTE_ORDER_MARK(uv) ((UV) (uv) == UNICODE_BYTE_ORDER_MARK)
c149ab20
KW
866
867/* Is 'uv' one of the 32 contiguous-range noncharacters? */
868#define UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv) ((UV) (uv) >= 0xFDD0 \
869 && (UV) (uv) <= 0xFDEF)
870
871/* Is 'uv' one of the 34 plane-ending noncharacters 0xFFFE, 0xFFFF, 0x1FFFE,
872 * 0x1FFFF, ... 0x10FFFE, 0x10FFFF, given that we know that 'uv' is not above
873 * the Unicode legal max */
874#define UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv) \
875 (((UV) (uv) & 0xFFFE) == 0xFFFE)
876
877#define UNICODE_IS_NONCHAR(uv) \
878 ( UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv) \
879 || ( LIKELY( ! UNICODE_IS_SUPER(uv)) \
880 && UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv)))
881
882#define UNICODE_IS_SUPER(uv) ((UV) (uv) > PERL_UNICODE_MAX)
1d72bdf6 883
ec34087a
KW
884#define LATIN_SMALL_LETTER_SHARP_S LATIN_SMALL_LETTER_SHARP_S_NATIVE
885#define LATIN_SMALL_LETTER_Y_WITH_DIAERESIS \
886 LATIN_SMALL_LETTER_Y_WITH_DIAERESIS_NATIVE
887#define MICRO_SIGN MICRO_SIGN_NATIVE
888#define LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE \
889 LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE_NATIVE
890#define LATIN_SMALL_LETTER_A_WITH_RING_ABOVE \
891 LATIN_SMALL_LETTER_A_WITH_RING_ABOVE_NATIVE
09091399
JH
892#define UNICODE_GREEK_CAPITAL_LETTER_SIGMA 0x03A3
893#define UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA 0x03C2
894#define UNICODE_GREEK_SMALL_LETTER_SIGMA 0x03C3
9dcbe121 895#define GREEK_SMALL_LETTER_MU 0x03BC
9e682c18
KW
896#define GREEK_CAPITAL_LETTER_MU 0x039C /* Upper and title case
897 of MICRON */
898#define LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS 0x0178 /* Also is title case */
0766489e
KW
899#ifdef LATIN_CAPITAL_LETTER_SHARP_S_UTF8
900# define LATIN_CAPITAL_LETTER_SHARP_S 0x1E9E
901#endif
74894415
KW
902#define LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE 0x130
903#define LATIN_SMALL_LETTER_DOTLESS_I 0x131
9e682c18 904#define LATIN_SMALL_LETTER_LONG_S 0x017F
a9f50d33
KW
905#define LATIN_SMALL_LIGATURE_LONG_S_T 0xFB05
906#define LATIN_SMALL_LIGATURE_ST 0xFB06
9e682c18
KW
907#define KELVIN_SIGN 0x212A
908#define ANGSTROM_SIGN 0x212B
09091399 909
9e55ce06 910#define UNI_DISPLAY_ISPRINT 0x0001
c728cb41
JH
911#define UNI_DISPLAY_BACKSLASH 0x0002
912#define UNI_DISPLAY_QQ (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH)
913#define UNI_DISPLAY_REGEX (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH)
9e55ce06 914
5cd46e1f
KW
915#define ANYOF_FOLD_SHARP_S(node, input, end) \
916 (ANYOF_BITMAP_TEST(node, LATIN_SMALL_LETTER_SHARP_S) && \
137165a6 917 (ANYOF_NONBITMAP(node)) && \
39065660 918 (ANYOF_FLAGS(node) & ANYOF_LOC_NONBITMAP_FOLD) && \
07b6858f 919 ((end) > (input) + 1) && \
305b8651 920 isALPHA_FOLD_EQ((input)[0], 's'))
6302f837 921
ebc501f0 922#define SHARP_S_SKIP 2
3b0fc154 923
3cedd9d9 924#define is_utf8_char_buf(buf, buf_end) isUTF8_CHAR(buf, buf_end)
976c1b08
KW
925#define bytes_from_utf8(s, lenp, is_utf8p) \
926 bytes_from_utf8_loc(s, lenp, is_utf8p, 0)
3cedd9d9 927
e23e8bc1
KW
928/*
929
25e3a4e0
KW
930=for apidoc Am|STRLEN|isUTF8_CHAR_flags|const U8 *s|const U8 *e| const U32 flags
931
932Evaluates to non-zero if the first few bytes of the string starting at C<s> and
933looking no further than S<C<e - 1>> are well-formed UTF-8, as extended by Perl,
934that represents some code point, subject to the restrictions given by C<flags>;
935otherwise it evaluates to 0. If non-zero, the value gives how many bytes
2717076a
KW
936starting at C<s> comprise the code point's representation. Any bytes remaining
937before C<e>, but beyond the ones needed to form the first code point in C<s>,
938are not examined.
25e3a4e0
KW
939
940If C<flags> is 0, this gives the same results as C<L</isUTF8_CHAR>>;
941if C<flags> is C<UTF8_DISALLOW_ILLEGAL_INTERCHANGE>, this gives the same results
942as C<L</isSTRICT_UTF8_CHAR>>;
943and if C<flags> is C<UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE>, this gives
944the same results as C<L</isC9_STRICT_UTF8_CHAR>>.
945Otherwise C<flags> may be any combination of the C<UTF8_DISALLOW_I<foo>> flags
946understood by C<L</utf8n_to_uvchr>>, with the same meanings.
947
948The three alternative macros are for the most commonly needed validations; they
949are likely to run somewhat faster than this more general one, as they can be
950inlined into your code.
951
9f2abfde
KW
952Use L</is_utf8_string_flags>, L</is_utf8_string_loc_flags>, and
953L</is_utf8_string_loclen_flags> to check entire strings.
954
25e3a4e0
KW
955=cut
956*/
957
958#define isUTF8_CHAR_flags(s, e, flags) \
959 (UNLIKELY((e) <= (s)) \
960 ? 0 \
961 : (UTF8_IS_INVARIANT(*s)) \
962 ? 1 \
963 : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \
964 ? 0 \
965 : _is_utf8_char_helper(s, e, flags))
966
6302f837
KW
967/* Do not use; should be deprecated. Use isUTF8_CHAR() instead; this is
968 * retained solely for backwards compatibility */
969#define IS_UTF8_CHAR(p, n) (isUTF8_CHAR(p, (p) + (n)) == n)
e9a8c099 970
6a5bc5ac 971#endif /* PERL_UTF8_H_ */
57f0e7e2 972
e9a8c099 973/*
14d04a33 974 * ex: set ts=8 sts=4 sw=4 et:
e9a8c099 975 */