This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
utf8_to_uvchr_buf() make behavior match docs
[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) \
82651abe 72 _utf8_to_uvchr_buf((const U8 *) (s), (const U8 *) e, lenp)
f9380377
KW
73#define utf8n_to_uvchr(s, len, lenp, flags) \
74 utf8n_to_uvchr_error(s, len, lenp, flags, 0)
37657a5b
KW
75#define utf8n_to_uvchr_error(s, len, lenp, flags, errors) \
76 utf8n_to_uvchr_msgs(s, len, lenp, flags, errors, 0)
de69f3af 77
a0270393 78#define to_uni_fold(c, p, lenp) _to_uni_fold_flags(c, p, lenp, FOLD_FLAGS_FULL)
a239b1e2
KW
79
80#define to_utf8_fold(s, r, lenr) \
607313a1 81 _to_utf8_fold_flags (s, NULL, r, lenr, FOLD_FLAGS_FULL, __FILE__, __LINE__)
a239b1e2 82#define to_utf8_lower(s, r, lenr) \
607313a1 83 _to_utf8_lower_flags(s, NULL, r ,lenr, 0, __FILE__, __LINE__)
a239b1e2 84#define to_utf8_upper(s, r, lenr) \
607313a1 85 _to_utf8_upper_flags(s, NULL, r, lenr, 0, __FILE__, __LINE__)
a239b1e2 86#define to_utf8_title(s, r, lenr) \
607313a1 87 _to_utf8_title_flags(s, NULL, r, lenr ,0, __FILE__, __LINE__)
36bb2ab6 88
eda9cac1
KW
89#define foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \
90 foldEQ_utf8_flags(s1, pe1, l1, u1, s2, pe2, l2, u2, 0)
baa60164 91#define FOLDEQ_UTF8_NOMIX_ASCII (1 << 0)
cea315b6 92#define FOLDEQ_LOCALE (1 << 1)
18f762c3
KW
93#define FOLDEQ_S1_ALREADY_FOLDED (1 << 2)
94#define FOLDEQ_S2_ALREADY_FOLDED (1 << 3)
d635b710
KW
95#define FOLDEQ_S1_FOLDS_SANE (1 << 4)
96#define FOLDEQ_S2_FOLDS_SANE (1 << 5)
a33c29bc 97
e6226b18
KW
98#define ibcmp_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \
99 cBOOL(! foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2))
100
1d72bdf6
NIS
101#ifdef EBCDIC
102/* The equivalent of these macros but implementing UTF-EBCDIC
103 are in the following header file:
104 */
105
106#include "utfebcdic.h"
fd7cb289 107
d06134e5 108#else /* ! EBCDIC */
73c4f7a1
GS
109START_EXTERN_C
110
111e8ed9
KW
111/* How wide can a single UTF-8 encoded character become in bytes. */
112/* NOTE: Strictly speaking Perl's UTF-8 should not be called UTF-8 since UTF-8
113 * is an encoding of Unicode, and Unicode's upper limit, 0x10FFFF, can be
114 * expressed with 4 bytes. However, Perl thinks of UTF-8 as a way to encode
115 * non-negative integers in a binary format, even those above Unicode */
116#define UTF8_MAXBYTES 13
117
a0ed51b3 118#ifdef DOINIT
6f06b55f 119EXTCONST unsigned char PL_utf8skip[] = {
b2635aa8
KW
120/* 0x00 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
121/* 0x10 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
122/* 0x20 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
123/* 0x30 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
124/* 0x40 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
125/* 0x50 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
126/* 0x60 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
127/* 0x70 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
128/* 0x80 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
129/* 0x90 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
130/* 0xA0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
131/* 0xB0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
132/* 0xC0 */ 2,2, /* overlong */
1ff3baa2 133/* 0xC2 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0080 to U+03FF */
b2635aa8
KW
134/* 0xD0 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0400 to U+07FF */
135/* 0xE0 */ 3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3, /* U+0800 to U+FFFF */
136/* 0xF0 */ 4,4,4,4,4,4,4,4,5,5,5,5,6,6, /* above BMP to 2**31 - 1 */
6937f885
KW
137 /* Perl extended (never was official UTF-8). Up to 36 bit */
138/* 0xFE */ 7,
139 /* More extended, Up to 72 bits (64-bit + reserved) */
111e8ed9 140/* 0xFF */ UTF8_MAXBYTES
a0ed51b3
LW
141};
142#else
6f06b55f 143EXTCONST unsigned char PL_utf8skip[];
a0ed51b3
LW
144#endif
145
73c4f7a1 146END_EXTERN_C
7e2040f0 147
1a3756de 148#if defined(_MSC_VER) && _MSC_VER < 1400
6f6d1bab
TC
149/* older MSVC versions have a smallish macro buffer */
150#define PERL_SMALL_MACRO_BUFFER
151#endif
152
59a449d5
KW
153/* Native character to/from iso-8859-1. Are the identity functions on ASCII
154 * platforms */
6f6d1bab 155#ifdef PERL_SMALL_MACRO_BUFFER
1a3756de
TC
156#define NATIVE_TO_LATIN1(ch) ((U8)(ch))
157#define LATIN1_TO_NATIVE(ch) ((U8)(ch))
6f6d1bab 158#else
4c8cd605
KW
159#define NATIVE_TO_LATIN1(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch)))
160#define LATIN1_TO_NATIVE(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch)))
6f6d1bab 161#endif
59a449d5
KW
162
163/* I8 is an intermediate version of UTF-8 used only in UTF-EBCDIC. We thus
164 * consider it to be identical to UTF-8 on ASCII platforms. Strictly speaking
165 * UTF-8 and UTF-EBCDIC are two different things, but we often conflate them
166 * because they are 8-bit encodings that serve the same purpose in Perl, and
167 * rarely do we need to distinguish them. The term "NATIVE_UTF8" applies to
168 * whichever one is applicable on the current platform */
6f6d1bab 169#ifdef PERL_SMALL_MACRO_BUFFER
1d4ea287
DD
170#define NATIVE_UTF8_TO_I8(ch) ((U8) (ch))
171#define I8_TO_NATIVE_UTF8(ch) ((U8) (ch))
6f6d1bab 172#else
4c8cd605
KW
173#define NATIVE_UTF8_TO_I8(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch)))
174#define I8_TO_NATIVE_UTF8(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch)))
6f6d1bab 175#endif
59a449d5 176
1d72bdf6 177/* Transforms in wide UV chars */
4c8cd605
KW
178#define UNI_TO_NATIVE(ch) ((UV) (ch))
179#define NATIVE_TO_UNI(ch) ((UV) (ch))
d7578b48 180
877d9f0d 181/*
9041c2e3 182
a14e0a36
KW
183 The following table is from Unicode 3.2, plus the Perl extensions for above
184 U+10FFFF
877d9f0d 185
a14e0a36 186 Code Points 1st Byte 2nd Byte 3rd 4th 5th 6th 7th 8th-13th
877d9f0d 187
375122d7 188 U+0000..U+007F 00..7F
e1b711da 189 U+0080..U+07FF * C2..DF 80..BF
a14e0a36
KW
190 U+0800..U+0FFF E0 * A0..BF 80..BF
191 U+1000..U+CFFF E1..EC 80..BF 80..BF
192 U+D000..U+D7FF ED 80..9F 80..BF
193 U+D800..U+DFFF ED A0..BF 80..BF (surrogates)
194 U+E000..U+FFFF EE..EF 80..BF 80..BF
195 U+10000..U+3FFFF F0 * 90..BF 80..BF 80..BF
196 U+40000..U+FFFFF F1..F3 80..BF 80..BF 80..BF
197 U+100000..U+10FFFF F4 80..8F 80..BF 80..BF
198 Below are above-Unicode code points
199 U+110000..U+13FFFF F4 90..BF 80..BF 80..BF
200 U+110000..U+1FFFFF F5..F7 80..BF 80..BF 80..BF
201 U+200000..U+FFFFFF F8 * 88..BF 80..BF 80..BF 80..BF
202U+1000000..U+3FFFFFF F9..FB 80..BF 80..BF 80..BF 80..BF
203U+4000000..U+3FFFFFFF FC * 84..BF 80..BF 80..BF 80..BF 80..BF
204U+40000000..U+7FFFFFFF FD 80..BF 80..BF 80..BF 80..BF 80..BF
205U+80000000..U+FFFFFFFFF FE * 82..BF 80..BF 80..BF 80..BF 80..BF 80..BF
206U+1000000000.. FF 80..BF 80..BF 80..BF 80..BF 80..BF * 81..BF 80..BF
877d9f0d 207
e1b711da 208Note the gaps before several of the byte entries above marked by '*'. These are
37e2e78e
KW
209caused by legal UTF-8 avoiding non-shortest encodings: it is technically
210possible to UTF-8-encode a single code point in different ways, but that is
211explicitly forbidden, and the shortest possible encoding should always be used
15824458 212(and that is what Perl does). The non-shortest ones are called 'overlongs'.
8c007b5a 213
877d9f0d
JH
214 */
215
8c007b5a
JH
216/*
217 Another way to look at it, as bits:
218
b2635aa8 219 Code Points 1st Byte 2nd Byte 3rd Byte 4th Byte
8c007b5a 220
b2635aa8
KW
221 0aaa aaaa 0aaa aaaa
222 0000 0bbb bbaa aaaa 110b bbbb 10aa aaaa
223 cccc bbbb bbaa aaaa 1110 cccc 10bb bbbb 10aa aaaa
224 00 000d ddcc cccc bbbb bbaa aaaa 1111 0ddd 10cc cccc 10bb bbbb 10aa aaaa
8c007b5a
JH
225
226As you can see, the continuation bytes all begin with C<10>, and the
e1b711da 227leading bits of the start byte tell how many bytes there are in the
8c007b5a
JH
228encoded character.
229
df863e43
KW
230Perl's extended UTF-8 means we can have start bytes up through FF, though any
231beginning with FF yields a code point that is too large for 32-bit ASCII
232platforms. FF signals to use 13 bytes for the encoded character. This breaks
233the paradigm that the number of leading bits gives how many total bytes there
234are in the character.
65ab9279 235
8c007b5a
JH
236*/
237
6c88483e 238/* Is the representation of the Unicode code point 'cp' the same regardless of
15824458 239 * being encoded in UTF-8 or not? */
2d1545e5 240#define OFFUNI_IS_INVARIANT(cp) isASCII(cp)
15824458 241
c2b32798
KW
242/*
243=for apidoc Am|bool|UVCHR_IS_INVARIANT|UV cp
244
245Evaluates to 1 if the representation of code point C<cp> is the same whether or
246not it is encoded in UTF-8; otherwise evaluates to 0. UTF-8 invariant
247characters can be copied as-is when converting to/from UTF-8, saving time.
248C<cp> is Unicode if above 255; otherwise is platform-native.
249
250=cut
251 */
252
cf1be84e 253#define UVCHR_IS_INVARIANT(cp) OFFUNI_IS_INVARIANT(cp)
38953e5a 254
c9264833
KW
255/* This defines the bits that are to be in the continuation bytes of a multi-byte
256 * UTF-8 encoded character that mark it is a continuation byte. */
257#define UTF_CONTINUATION_MARK 0x80
258
a95ec4fb 259/* Misleadingly named: is the UTF8-encoded byte 'c' part of a variant sequence
5c06326b
KW
260 * in UTF-8? This is the inverse of UTF8_IS_INVARIANT. The |0 makes sure this
261 * isn't mistakenly called with a ptr argument */
a6951642
KW
262#define UTF8_IS_CONTINUED(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
263 ((U8)((c) | 0)) & UTF_CONTINUATION_MARK)
15824458
KW
264
265/* Is the byte 'c' the first byte of a multi-byte UTF8-8 encoded sequence?
266 * This doesn't catch invariants (they are single-byte). It also excludes the
5c06326b
KW
267 * illegal overlong sequences that begin with C0 and C1. The |0 makes sure
268 * this isn't mistakenly called with a ptr argument */
a6951642
KW
269#define UTF8_IS_START(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
270 ((U8)((c) | 0)) >= 0xc2)
15824458 271
858cd8ab
KW
272/* For use in UTF8_IS_CONTINUATION() below */
273#define UTF_IS_CONTINUATION_MASK 0xC0
274
15824458 275/* Is the byte 'c' part of a multi-byte UTF8-8 encoded sequence, and not the
5c06326b
KW
276 * first byte thereof? The |0 makes sure this isn't mistakenly called with a
277 * ptr argument */
a6951642
KW
278#define UTF8_IS_CONTINUATION(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
279 (((U8)((c) | 0)) & UTF_IS_CONTINUATION_MASK) == UTF_CONTINUATION_MARK)
0ae1fa71 280
15824458
KW
281/* Is the UTF8-encoded byte 'c' the first byte of a two byte sequence? Use
282 * UTF8_IS_NEXT_CHAR_DOWNGRADEABLE() instead if the input isn't known to
283 * be well-formed. Masking with 0xfe allows the low bit to be 0 or 1; thus
5c06326b
KW
284 * this matches 0xc[23]. The |0 makes sure this isn't mistakenly called with a
285 * ptr argument */
a6951642
KW
286#define UTF8_IS_DOWNGRADEABLE_START(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
287 (((U8)((c) | 0)) & 0xfe) == 0xc2)
4ab10950 288
15824458 289/* Is the UTF8-encoded byte 'c' the first byte of a sequence of bytes that
5c06326b
KW
290 * represent a code point > 255? The |0 makes sure this isn't mistakenly
291 * called with a ptr argument */
a6951642
KW
292#define UTF8_IS_ABOVE_LATIN1(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
293 ((U8)((c) | 0)) >= 0xc4)
8850bf83 294
15824458
KW
295/* This is the number of low-order bits a continuation byte in a UTF-8 encoded
296 * sequence contributes to the specification of the code point. In the bit
297 * maps above, you see that the first 2 bits are a constant '10', leaving 6 of
298 * real information */
1d72bdf6 299#define UTF_ACCUMULATION_SHIFT 6
b2635aa8 300
fed423a5
KW
301/* ^? is defined to be DEL on ASCII systems. See the definition of toCTRL()
302 * for more */
303#define QUESTION_MARK_CTRL DEL_NATIVE
304
305/* Surrogates, non-character code points and above-Unicode code points are
306 * problematic in some contexts. This allows code that needs to check for
307 * those to to quickly exclude the vast majority of code points it will
308 * encounter */
a6951642
KW
309#define isUTF8_POSSIBLY_PROBLEMATIC(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
310 (U8) c >= 0xED)
fed423a5 311
57ff5f59
KW
312#define UNICODE_IS_PERL_EXTENDED(uv) UNLIKELY((UV) (uv) > 0x7FFFFFFF)
313
fed423a5
KW
314#endif /* EBCDIC vs ASCII */
315
316/* 2**UTF_ACCUMULATION_SHIFT - 1 */
317#define UTF_CONTINUATION_MASK ((U8) ((1U << UTF_ACCUMULATION_SHIFT) - 1))
318
7028aeba
KW
319/* Internal macro to be used only in this file to aid in constructing other
320 * publicly accessible macros.
321 * The number of bytes required to express this uv in UTF-8, for just those
322 * uv's requiring 2 through 6 bytes, as these are common to all platforms and
323 * word sizes. The number of bytes needed is given by the number of leading 1
324 * bits in the start byte. There are 32 start bytes that have 2 initial 1 bits
325 * (C0-DF); there are 16 that have 3 initial 1 bits (E0-EF); 8 that have 4
326 * initial 1 bits (F0-F8); 4 that have 5 initial 1 bits (F9-FB), and 2 that
327 * have 6 initial 1 bits (FC-FD). The largest number a string of n bytes can
328 * represent is (the number of possible start bytes for 'n')
329 * * (the number of possiblities for each start byte
330 * The latter in turn is
331 * 2 ** ( (how many continuation bytes there are)
332 * * (the number of bits of information each
333 * continuation byte holds))
334 *
335 * If we were on a platform where we could use a fast find first set bit
336 * instruction (or count leading zeros instruction) this could be replaced by
337 * using that to find the log2 of the uv, and divide that by the number of bits
338 * of information in each continuation byte, adjusting for large cases and how
339 * much information is in a start byte for that length */
72164d3a 340#define __COMMON_UNI_SKIP(uv) \
7028aeba
KW
341 (UV) (uv) < (32 * (1U << ( UTF_ACCUMULATION_SHIFT))) ? 2 : \
342 (UV) (uv) < (16 * (1U << (2 * UTF_ACCUMULATION_SHIFT))) ? 3 : \
343 (UV) (uv) < ( 8 * (1U << (3 * UTF_ACCUMULATION_SHIFT))) ? 4 : \
344 (UV) (uv) < ( 4 * (1U << (4 * UTF_ACCUMULATION_SHIFT))) ? 5 : \
345 (UV) (uv) < ( 2 * (1U << (5 * UTF_ACCUMULATION_SHIFT))) ? 6 :
72164d3a
KW
346
347/* Internal macro to be used only in this file.
348 * This adds to __COMMON_UNI_SKIP the details at this platform's upper range.
fed423a5 349 * For any-sized EBCDIC platforms, or 64-bit ASCII ones, we need one more test
72164d3a
KW
350 * to see if just 7 bytes is needed, or if the maximum is needed. For 32-bit
351 * ASCII platforms, everything is representable by 7 bytes */
fed423a5 352#if defined(UV_IS_QUAD) || defined(EBCDIC)
72164d3a 353# define __BASE_UNI_SKIP(uv) (__COMMON_UNI_SKIP(uv) \
7028aeba 354 (UV) (uv) < ((UV) 1U << (6 * UTF_ACCUMULATION_SHIFT)) ? 7 : UTF8_MAXBYTES)
1d68d6cd 355#else
72164d3a 356# define __BASE_UNI_SKIP(uv) (__COMMON_UNI_SKIP(uv) 7)
1d68d6cd
SC
357#endif
358
fed423a5
KW
359/* The next two macros use the base macro defined above, and add in the tests
360 * at the low-end of the range, for just 1 byte, yielding complete macros,
361 * publicly accessible. */
362
363/* Input is a true Unicode (not-native) code point */
364#define OFFUNISKIP(uv) (OFFUNI_IS_INVARIANT(uv) ? 1 : __BASE_UNI_SKIP(uv))
2084b489 365
5352a763
KW
366/*
367
368=for apidoc Am|STRLEN|UVCHR_SKIP|UV cp
369returns the number of bytes required to represent the code point C<cp> when
370encoded as UTF-8. C<cp> is a native (ASCII or EBCDIC) code point if less than
371255; a Unicode code point otherwise.
372
373=cut
374 */
fdb6583d 375#define UVCHR_SKIP(uv) ( UVCHR_IS_INVARIANT(uv) ? 1 : __BASE_UNI_SKIP(uv))
5352a763 376
b651802e
KW
377/* The largest code point representable by two UTF-8 bytes on this platform.
378 * As explained in the comments for __COMMON_UNI_SKIP, 32 start bytes with
fed423a5 379 * UTF_ACCUMULATION_SHIFT bits of information each */
aa206fb7
KW
380#define MAX_UTF8_TWO_BYTE (32 * (1U << UTF_ACCUMULATION_SHIFT) - 1)
381
b651802e
KW
382/* The largest code point representable by two UTF-8 bytes on any platform that
383 * Perl runs on. This value is constrained by EBCDIC which has 5 bits per
384 * continuation byte */
aa206fb7
KW
385#define MAX_PORTABLE_UTF8_TWO_BYTE (32 * (1U << 5) - 1)
386
c03c0950
KW
387/* The maximum number of UTF-8 bytes a single Unicode character can
388 * uppercase/lowercase/fold into. Unicode guarantees that the maximum
389 * expansion is UTF8_MAX_FOLD_CHAR_EXPAND characters, but any above-Unicode
390 * code point will fold to itself, so we only have to look at the expansion of
391 * the maximum Unicode code point. But this number may be less than the space
392 * occupied by a very large code point under Perl's extended UTF-8. We have to
393 * make it large enough to fit any single character. (It turns out that ASCII
394 * and EBCDIC differ in which is larger) */
395#define UTF8_MAXBYTES_CASE \
396 (UTF8_MAXBYTES >= (UTF8_MAX_FOLD_CHAR_EXPAND * OFFUNISKIP(0x10FFFF)) \
397 ? UTF8_MAXBYTES \
398 : (UTF8_MAX_FOLD_CHAR_EXPAND * OFFUNISKIP(0x10FFFF)))
399
d06134e5
KW
400/* Rest of these are attributes of Unicode and perl's internals rather than the
401 * encoding, or happen to be the same in both ASCII and EBCDIC (at least at
402 * this level; the macros that some of these call may have different
403 * definitions in the two encodings */
404
59a449d5
KW
405/* In domain restricted to ASCII, these may make more sense to the reader than
406 * the ones with Latin1 in the name */
407#define NATIVE_TO_ASCII(ch) NATIVE_TO_LATIN1(ch)
408#define ASCII_TO_NATIVE(ch) LATIN1_TO_NATIVE(ch)
409
410/* More or less misleadingly-named defines, retained for back compat */
411#define NATIVE_TO_UTF(ch) NATIVE_UTF8_TO_I8(ch)
412#define NATIVE_TO_I8(ch) NATIVE_UTF8_TO_I8(ch)
413#define UTF_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch)
414#define I8_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch)
415#define NATIVE8_TO_UNI(ch) NATIVE_TO_LATIN1(ch)
d06134e5 416
c0236afe 417/* This defines the 1-bits that are to be in the first byte of a multi-byte
97d0ceda
KW
418 * UTF-8 encoded character that mark it as a start byte and give the number of
419 * bytes that comprise the character. 'len' is the number of bytes in the
420 * multi-byte sequence. */
c0236afe
KW
421#define UTF_START_MARK(len) (((len) > 7) ? 0xFF : (0xFF & (0xFE << (7-(len)))))
422
423/* Masks out the initial one bits in a start byte, leaving the real data ones.
424 * Doesn't work on an invariant byte. 'len' is the number of bytes in the
425 * multi-byte sequence that comprises the character. */
426#define UTF_START_MASK(len) (((len) >= 7) ? 0x00 : (0x1F >> ((len)-2)))
427
537124e4
KW
428/* Adds a UTF8 continuation byte 'new' of information to a running total code
429 * point 'old' of all the continuation bytes so far. This is designed to be
155d2738
KW
430 * used in a loop to convert from UTF-8 to the code point represented. Note
431 * that this is asymmetric on EBCDIC platforms, in that the 'new' parameter is
432 * the UTF-EBCDIC byte, whereas the 'old' parameter is a Unicode (not EBCDIC)
433 * code point in process of being generated */
a6951642
KW
434#define UTF8_ACCUMULATE(old, new) (__ASSERT_(FITS_IN_8_BITS(new)) \
435 ((old) << UTF_ACCUMULATION_SHIFT) \
155d2738
KW
436 | ((NATIVE_UTF8_TO_I8((U8)new)) \
437 & UTF_CONTINUATION_MASK))
d06134e5 438
4ab10950 439/* This works in the face of malformed UTF-8. */
4e1ed312
KW
440#define UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(s, e) \
441 ( UTF8_IS_DOWNGRADEABLE_START(*(s)) \
442 && ( (e) - (s) > 1) \
443 && UTF8_IS_CONTINUATION(*((s)+1)))
4ab10950 444
5aaebcb3 445/* Number of bytes a code point occupies in UTF-8. */
5352a763 446#define NATIVE_SKIP(uv) UVCHR_SKIP(uv)
bd18bd40 447
5aaebcb3
KW
448/* Most code which says UNISKIP is really thinking in terms of native code
449 * points (0-255) plus all those beyond. This is an imprecise term, but having
2accb712 450 * it means existing code continues to work. For precision, use UVCHR_SKIP,
5352a763
KW
451 * NATIVE_SKIP, or OFFUNISKIP */
452#define UNISKIP(uv) UVCHR_SKIP(uv)
5aaebcb3 453
3c0792e4
KW
454/* Longer, but more accurate name */
455#define UTF8_IS_ABOVE_LATIN1_START(c) UTF8_IS_ABOVE_LATIN1(c)
456
a62b247b
KW
457/* Convert a UTF-8 variant Latin1 character to a native code point value.
458 * Needs just one iteration of accumulate. Should be used only if it is known
459 * that the code point is < 256, and is not UTF-8 invariant. Use the slower
460 * but more general TWO_BYTE_UTF8_TO_NATIVE() which handles any code point
461 * representable by two bytes (which turns out to be up through
462 * MAX_PORTABLE_UTF8_TWO_BYTE). The two parameters are:
463 * HI: a downgradable start byte;
464 * LO: continuation.
465 * */
466#define EIGHT_BIT_UTF8_TO_NATIVE(HI, LO) \
467 ( __ASSERT_(UTF8_IS_DOWNGRADEABLE_START(HI)) \
468 __ASSERT_(UTF8_IS_CONTINUATION(LO)) \
469 LATIN1_TO_NATIVE(UTF8_ACCUMULATE(( \
470 NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), (LO))))
471
94bb8c36 472/* Convert a two (not one) byte utf8 character to a native code point value.
2950f2a7
KW
473 * Needs just one iteration of accumulate. Should not be used unless it is
474 * known that the two bytes are legal: 1) two-byte start, and 2) continuation.
475 * Note that the result can be larger than 255 if the input character is not
476 * downgradable */
94bb8c36 477#define TWO_BYTE_UTF8_TO_NATIVE(HI, LO) \
a6951642
KW
478 (__ASSERT_(FITS_IN_8_BITS(HI)) \
479 __ASSERT_(FITS_IN_8_BITS(LO)) \
480 __ASSERT_(PL_utf8skip[HI] == 2) \
481 __ASSERT_(UTF8_IS_CONTINUATION(LO)) \
94bb8c36 482 UNI_TO_NATIVE(UTF8_ACCUMULATE((NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), \
635e76f5 483 (LO))))
94bb8c36
KW
484
485/* Should never be used, and be deprecated */
486#define TWO_BYTE_UTF8_TO_UNI(HI, LO) NATIVE_TO_UNI(TWO_BYTE_UTF8_TO_NATIVE(HI, LO))
2950f2a7 487
bd18bd40
KW
488/*
489
490=for apidoc Am|STRLEN|UTF8SKIP|char* s
491returns the number of bytes in the UTF-8 encoded character whose first (perhaps
492only) byte is pointed to by C<s>.
493
494=cut
495 */
2a70536e
KW
496#define UTF8SKIP(s) PL_utf8skip[*(const U8*)(s)]
497#define UTF8_SKIP(s) UTF8SKIP(s)
d06134e5 498
85fcc8f2
KW
499/*
500
501=for apidoc Am|STRLEN|UTF8_SAFE_SKIP|char* s|char* e
45671da2
KW
502returns 0 if S<C<s E<gt>= e>>; otherwise returns the number of bytes in the
503UTF-8 encoded character whose first byte is pointed to by C<s>. But it never
504returns beyond C<e>. On DEBUGGING builds, it asserts that S<C<s E<lt>= e>>.
85fcc8f2
KW
505
506=cut
507 */
45671da2
KW
508#define UTF8_SAFE_SKIP(s, e) (__ASSERT_((e) >= (s)) \
509 ((e) - (s)) <= 0 \
510 ? 0 \
511 : MIN(((e) - (s)), UTF8_SKIP(s)))
85fcc8f2 512
2d1545e5
KW
513/* Most code that says 'UNI_' really means the native value for code points up
514 * through 255 */
515#define UNI_IS_INVARIANT(cp) UVCHR_IS_INVARIANT(cp)
516
c2b32798
KW
517/*
518=for apidoc Am|bool|UTF8_IS_INVARIANT|char c
519
520Evaluates to 1 if the byte C<c> represents the same character when encoded in
521UTF-8 as when not; otherwise evaluates to 0. UTF-8 invariant characters can be
522copied as-is when converting to/from UTF-8, saving time.
523
524In spite of the name, this macro gives the correct result if the input string
525from which C<c> comes is not encoded in UTF-8.
526
527See C<L</UVCHR_IS_INVARIANT>> for checking if a UV is invariant.
528
529=cut
530
531The reason it works on both UTF-8 encoded strings and non-UTF-8 encoded, is
532that it returns TRUE in each for the exact same set of bit patterns. It is
533valid on a subset of what UVCHR_IS_INVARIANT is valid on, so can just use that;
534and the compiler should optimize out anything extraneous given the
535implementation of the latter. The |0 makes sure this isn't mistakenly called
536with a ptr argument.
537*/
5c06326b 538#define UTF8_IS_INVARIANT(c) UVCHR_IS_INVARIANT((c) | 0)
5fc230f1
KW
539
540/* Like the above, but its name implies a non-UTF8 input, which as the comments
541 * above show, doesn't matter as to its implementation */
38953e5a 542#define NATIVE_BYTE_IS_INVARIANT(c) UVCHR_IS_INVARIANT(c)
d06134e5 543
48ccf5e1
KW
544/* The macros in the next 4 sets are used to generate the two utf8 or utfebcdic
545 * bytes from an ordinal that is known to fit into exactly two (not one) bytes;
546 * it must be less than 0x3FF to work across both encodings. */
547
548/* These two are helper macros for the other three sets, and should not be used
549 * directly anywhere else. 'translate_function' is either NATIVE_TO_LATIN1
1ff3baa2
KW
550 * (which works for code points up through 0xFF) or NATIVE_TO_UNI which works
551 * for any code point */
48ccf5e1 552#define __BASE_TWO_BYTE_HI(c, translate_function) \
2863dafa 553 (__ASSERT_(! UVCHR_IS_INVARIANT(c)) \
48ccf5e1 554 I8_TO_NATIVE_UTF8((translate_function(c) >> UTF_ACCUMULATION_SHIFT) \
2863dafa 555 | UTF_START_MARK(2)))
48ccf5e1 556#define __BASE_TWO_BYTE_LO(c, translate_function) \
2863dafa 557 (__ASSERT_(! UVCHR_IS_INVARIANT(c)) \
48ccf5e1 558 I8_TO_NATIVE_UTF8((translate_function(c) & UTF_CONTINUATION_MASK) \
2863dafa 559 | UTF_CONTINUATION_MARK))
48ccf5e1 560
48ccf5e1
KW
561/* The next two macros should not be used. They were designed to be usable as
562 * the case label of a switch statement, but this doesn't work for EBCDIC. Use
9d0d3a03 563 * regen/unicode_constants.pl instead */
48ccf5e1
KW
564#define UTF8_TWO_BYTE_HI_nocast(c) __BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI)
565#define UTF8_TWO_BYTE_LO_nocast(c) __BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI)
566
567/* The next two macros are used when the source should be a single byte
568 * character; checked for under DEBUGGING */
569#define UTF8_EIGHT_BIT_HI(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
4c8cd605 570 ( __BASE_TWO_BYTE_HI(c, NATIVE_TO_LATIN1)))
48ccf5e1 571#define UTF8_EIGHT_BIT_LO(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
4c8cd605 572 (__BASE_TWO_BYTE_LO(c, NATIVE_TO_LATIN1)))
48ccf5e1
KW
573
574/* These final two macros in the series are used when the source can be any
575 * code point whose UTF-8 is known to occupy 2 bytes; they are less efficient
576 * than the EIGHT_BIT versions on EBCDIC platforms. We use the logical '~'
577 * operator instead of "<=" to avoid getting compiler warnings.
d52b8576 578 * MAX_UTF8_TWO_BYTE should be exactly all one bits in the lower few
48ccf5e1
KW
579 * places, so the ~ works */
580#define UTF8_TWO_BYTE_HI(c) \
581 (__ASSERT_((sizeof(c) == 1) \
d52b8576 582 || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \
4c8cd605 583 (__BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI)))
48ccf5e1
KW
584#define UTF8_TWO_BYTE_LO(c) \
585 (__ASSERT_((sizeof(c) == 1) \
d52b8576 586 || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \
4c8cd605 587 (__BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI)))
d06134e5 588
e7214ce8
KW
589/* This is illegal in any well-formed UTF-8 in both EBCDIC and ASCII
590 * as it is only in overlongs. */
591#define ILLEGAL_UTF8_BYTE I8_TO_NATIVE_UTF8(0xC1)
592
7e2040f0 593/*
e3036cf4 594 * 'UTF' is whether or not p is encoded in UTF8. The names 'foo_lazy_if' stem
20df05f4
KW
595 * from an earlier version of these macros in which they didn't call the
596 * foo_utf8() macros (i.e. were 'lazy') unless they decided that *p is the
597 * beginning of a utf8 character. Now that foo_utf8() determines that itself,
598 * no need to do it again here
7e2040f0 599 */
34aeb2e9
KW
600#define isIDFIRST_lazy_if(p,UTF) \
601 _is_utf8_FOO(_CC_IDFIRST, (const U8 *) p, "isIDFIRST_lazy_if", \
602 "isIDFIRST_lazy_if_safe", \
603 cBOOL(UTF && ! IN_BYTES), 0, __FILE__,__LINE__)
1d72bdf6 604
da8c1a98
KW
605#define isIDFIRST_lazy_if_safe(p, e, UTF) \
606 ((IN_BYTES || !UTF) \
607 ? isIDFIRST(*(p)) \
608 : isIDFIRST_utf8_safe(p, e))
609
34aeb2e9
KW
610#define isWORDCHAR_lazy_if(p,UTF) \
611 _is_utf8_FOO(_CC_IDFIRST, (const U8 *) p, "isWORDCHAR_lazy_if", \
612 "isWORDCHAR_lazy_if_safe", \
613 cBOOL(UTF && ! IN_BYTES), 0, __FILE__,__LINE__)
614
da8c1a98
KW
615#define isWORDCHAR_lazy_if_safe(p, e, UTF) \
616 ((IN_BYTES || !UTF) \
617 ? isWORDCHAR(*(p)) \
618 : isWORDCHAR_utf8_safe((U8 *) p, (U8 *) e))
619
34aeb2e9
KW
620#define isALNUM_lazy_if(p,UTF) \
621 _is_utf8_FOO(_CC_IDFIRST, (const U8 *) p, "isALNUM_lazy_if", \
622 "isWORDCHAR_lazy_if_safe", \
623 cBOOL(UTF && ! IN_BYTES), 0, __FILE__,__LINE__)
da8c1a98 624
89ebb4a3
JH
625#define UTF8_MAXLEN UTF8_MAXBYTES
626
8cb75cc8
KW
627/* A Unicode character can fold to up to 3 characters */
628#define UTF8_MAX_FOLD_CHAR_EXPAND 3
629
d3481830 630#define IN_BYTES UNLIKELY(CopHINTS_get(PL_curcop) & HINT_BYTES)
bd18bd40
KW
631
632/*
633
634=for apidoc Am|bool|DO_UTF8|SV* sv
635Returns a bool giving whether or not the PV in C<sv> is to be treated as being
636encoded in UTF-8.
637
638You should use this I<after> a call to C<SvPV()> or one of its variants, in
639case any call to string overloading updates the internal UTF-8 encoding flag.
640
641=cut
642*/
0064a8a9 643#define DO_UTF8(sv) (SvUTF8(sv) && !IN_BYTES)
1ff3baa2
KW
644
645/* Should all strings be treated as Unicode, and not just UTF-8 encoded ones?
646 * Is so within 'feature unicode_strings' or 'locale :not_characters', and not
647 * within 'use bytes'. UTF-8 locales are not tested for here, but perhaps
648 * could be */
70844984
KW
649#define IN_UNI_8_BIT \
650 (( ( (CopHINTS_get(PL_curcop) & HINT_UNI_8_BIT)) \
651 || ( CopHINTS_get(PL_curcop) & HINT_LOCALE_PARTIAL \
652 /* -1 below is for :not_characters */ \
653 && _is_in_locale_category(FALSE, -1))) \
654 && (! IN_BYTES))
b36bf33f 655
1d72bdf6 656
c76687c5 657#define UTF8_ALLOW_EMPTY 0x0001 /* Allow a zero length string */
2b5e7bc2 658#define UTF8_GOT_EMPTY UTF8_ALLOW_EMPTY
c76687c5
KW
659
660/* Allow first byte to be a continuation byte */
1d72bdf6 661#define UTF8_ALLOW_CONTINUATION 0x0002
2b5e7bc2 662#define UTF8_GOT_CONTINUATION UTF8_ALLOW_CONTINUATION
c76687c5 663
cd01d3b1 664/* Unexpected non-continuation byte */
1d72bdf6 665#define UTF8_ALLOW_NON_CONTINUATION 0x0004
2b5e7bc2 666#define UTF8_GOT_NON_CONTINUATION UTF8_ALLOW_NON_CONTINUATION
949cf498
KW
667
668/* expecting more bytes than were available in the string */
669#define UTF8_ALLOW_SHORT 0x0008
2b5e7bc2 670#define UTF8_GOT_SHORT UTF8_ALLOW_SHORT
949cf498 671
94953955
KW
672/* Overlong sequence; i.e., the code point can be specified in fewer bytes.
673 * First one will convert the overlong to the REPLACEMENT CHARACTER; second
674 * will return what the overlong evaluates to */
949cf498 675#define UTF8_ALLOW_LONG 0x0010
94953955 676#define UTF8_ALLOW_LONG_AND_ITS_VALUE (UTF8_ALLOW_LONG|0x0020)
2b5e7bc2
KW
677#define UTF8_GOT_LONG UTF8_ALLOW_LONG
678
d60baaa7
KW
679#define UTF8_ALLOW_OVERFLOW 0x0080
680#define UTF8_GOT_OVERFLOW UTF8_ALLOW_OVERFLOW
949cf498 681
f180b292 682#define UTF8_DISALLOW_SURROGATE 0x0100 /* Unicode surrogates */
2b5e7bc2 683#define UTF8_GOT_SURROGATE UTF8_DISALLOW_SURROGATE
f180b292 684#define UTF8_WARN_SURROGATE 0x0200
949cf498 685
c4e96019
KW
686/* Unicode non-character code points */
687#define UTF8_DISALLOW_NONCHAR 0x0400
2b5e7bc2 688#define UTF8_GOT_NONCHAR UTF8_DISALLOW_NONCHAR
c4e96019 689#define UTF8_WARN_NONCHAR 0x0800
949cf498 690
c4e96019
KW
691/* Super-set of Unicode: code points above the legal max */
692#define UTF8_DISALLOW_SUPER 0x1000
2b5e7bc2 693#define UTF8_GOT_SUPER UTF8_DISALLOW_SUPER
c4e96019
KW
694#define UTF8_WARN_SUPER 0x2000
695
696/* The original UTF-8 standard did not define UTF-8 with start bytes of 0xFE or
697 * 0xFF, though UTF-EBCDIC did. This allowed both versions to represent code
698 * points up to 2 ** 31 - 1. Perl extends UTF-8 so that 0xFE and 0xFF are
699 * usable on ASCII platforms, and 0xFF means something different than
700 * UTF-EBCDIC defines. These changes allow code points of 64 bits (actually
701 * somewhat more) to be represented on both platforms. But these are Perl
702 * extensions, and not likely to be interchangeable with other languages. Note
703 * that on ASCII platforms, FE overflows a signed 32-bit word, and FF an
704 * unsigned one. */
d044b7a7
KW
705#define UTF8_DISALLOW_PERL_EXTENDED 0x4000
706#define UTF8_GOT_PERL_EXTENDED UTF8_DISALLOW_PERL_EXTENDED
707#define UTF8_WARN_PERL_EXTENDED 0x8000
d35f2ca5 708
57ff5f59
KW
709/* For back compat, these old names are misleading for overlongs and
710 * UTF_EBCDIC. */
d044b7a7
KW
711#define UTF8_DISALLOW_ABOVE_31_BIT UTF8_DISALLOW_PERL_EXTENDED
712#define UTF8_GOT_ABOVE_31_BIT UTF8_GOT_PERL_EXTENDED
713#define UTF8_WARN_ABOVE_31_BIT UTF8_WARN_PERL_EXTENDED
714#define UTF8_DISALLOW_FE_FF UTF8_DISALLOW_PERL_EXTENDED
715#define UTF8_WARN_FE_FF UTF8_WARN_PERL_EXTENDED
949cf498 716
f180b292 717#define UTF8_CHECK_ONLY 0x10000
99a765e9 718#define _UTF8_NO_CONFIDENCE_IN_CURLEN 0x20000 /* Internal core use only */
949cf498
KW
719
720/* For backwards source compatibility. They do nothing, as the default now
721 * includes what they used to mean. The first one's meaning was to allow the
722 * just the single non-character 0xFFFF */
723#define UTF8_ALLOW_FFFF 0
c825ef8c 724#define UTF8_ALLOW_FE_FF 0
949cf498
KW
725#define UTF8_ALLOW_SURROGATE 0
726
ecc1615f
KW
727/* C9 refers to Unicode Corrigendum #9: allows but discourages non-chars */
728#define UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE \
729 (UTF8_DISALLOW_SUPER|UTF8_DISALLOW_SURROGATE)
730#define UTF8_WARN_ILLEGAL_C9_INTERCHANGE (UTF8_WARN_SUPER|UTF8_WARN_SURROGATE)
731
d35f2ca5 732#define UTF8_DISALLOW_ILLEGAL_INTERCHANGE \
ecc1615f 733 (UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE|UTF8_DISALLOW_NONCHAR)
949cf498 734#define UTF8_WARN_ILLEGAL_INTERCHANGE \
ecc1615f
KW
735 (UTF8_WARN_ILLEGAL_C9_INTERCHANGE|UTF8_WARN_NONCHAR)
736
0eb3d6a0
KW
737/* This is typically used for code that processes UTF-8 input and doesn't want
738 * to have to deal with any malformations that might be present. All such will
739 * be safely replaced by the REPLACEMENT CHARACTER, unless other flags
740 * overriding this are also present. */
2d532c27
KW
741#define UTF8_ALLOW_ANY ( UTF8_ALLOW_CONTINUATION \
742 |UTF8_ALLOW_NON_CONTINUATION \
743 |UTF8_ALLOW_SHORT \
d60baaa7
KW
744 |UTF8_ALLOW_LONG \
745 |UTF8_ALLOW_OVERFLOW)
2d532c27
KW
746
747/* Accept any Perl-extended UTF-8 that evaluates to any UV on the platform, but
cd01d3b1 748 * not any malformed. This is the default. */
2d532c27
KW
749#define UTF8_ALLOW_ANYUV 0
750#define UTF8_ALLOW_DEFAULT UTF8_ALLOW_ANYUV
1d72bdf6 751
89d986df
KW
752/*
753=for apidoc Am|bool|UTF8_IS_SURROGATE|const U8 *s|const U8 *e
754
755Evaluates to non-zero if the first few bytes of the string starting at C<s> and
756looking no further than S<C<e - 1>> are well-formed UTF-8 that represents one
757of the Unicode surrogate code points; otherwise it evaluates to 0. If
758non-zero, the value gives how many bytes starting at C<s> comprise the code
759point's representation.
760
761=cut
762 */
763#define UTF8_IS_SURROGATE(s, e) is_SURROGATE_utf8_safe(s, e)
764
765
766#define UTF8_IS_REPLACEMENT(s, send) is_REPLACEMENT_utf8_safe(s,send)
767
285aa1f0 768#define MAX_LEGAL_CP ((UV)IV_MAX)
40606899 769
89d986df
KW
770/*
771=for apidoc Am|bool|UTF8_IS_SUPER|const U8 *s|const U8 *e
772
773Recall that Perl recognizes an extension to UTF-8 that can encode code
774points larger than the ones defined by Unicode, which are 0..0x10FFFF.
775
776This macro evaluates to non-zero if the first few bytes of the string starting
777at C<s> and looking no further than S<C<e - 1>> are from this UTF-8 extension;
778otherwise it evaluates to 0. If non-zero, the value gives how many bytes
779starting at C<s> comprise the code point's representation.
780
7810 is returned if the bytes are not well-formed extended UTF-8, or if they
782represent a code point that cannot fit in a UV on the current platform. Hence
783this macro can give different results when run on a 64-bit word machine than on
784one with a 32-bit word size.
0c58a72b 785
891fd405 786Note that it is illegal to have code points that are larger than what can
89d986df 787fit in an IV on the current machine.
7131f24d 788
89d986df
KW
789=cut
790
791 * ASCII EBCDIC I8
7131f24d
KW
792 * U+10FFFF: \xF4\x8F\xBF\xBF \xF9\xA1\xBF\xBF\xBF max legal Unicode
793 * U+110000: \xF4\x90\x80\x80 \xF9\xA2\xA0\xA0\xA0
794 * U+110001: \xF4\x90\x80\x81 \xF9\xA2\xA0\xA0\xA1
89d986df
KW
795 */
796#ifdef EBCDIC
a14e0a36 797# define UTF8_IS_SUPER(s, e) \
89d986df
KW
798 (( LIKELY((e) > (s) + 4) \
799 && NATIVE_UTF8_TO_I8(*(s)) >= 0xF9 \
800 && ( NATIVE_UTF8_TO_I8(*(s)) > 0xF9 \
3d42f267 801 || (NATIVE_UTF8_TO_I8(*((s) + 1)) >= 0xA2)) \
89d986df 802 && LIKELY((s) + UTF8SKIP(s) <= (e))) \
2b479609 803 ? _is_utf8_char_helper(s, s + UTF8SKIP(s), 0) : 0)
7131f24d 804#else
a14e0a36 805# define UTF8_IS_SUPER(s, e) \
89d986df
KW
806 (( LIKELY((e) > (s) + 3) \
807 && (*(U8*) (s)) >= 0xF4 \
808 && ((*(U8*) (s)) > 0xF4 || (*((U8*) (s) + 1) >= 0x90))\
809 && LIKELY((s) + UTF8SKIP(s) <= (e))) \
2b479609 810 ? _is_utf8_char_helper(s, s + UTF8SKIP(s), 0) : 0)
7131f24d
KW
811#endif
812
b96a92fb
KW
813/* These are now machine generated, and the 'given' clause is no longer
814 * applicable */
0c58a72b 815#define UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s, e) \
89d986df
KW
816 cBOOL(is_NONCHAR_utf8_safe(s,e))
817
818/*
819=for apidoc Am|bool|UTF8_IS_NONCHAR|const U8 *s|const U8 *e
820
821Evaluates to non-zero if the first few bytes of the string starting at C<s> and
822looking no further than S<C<e - 1>> are well-formed UTF-8 that represents one
823of the Unicode non-character code points; otherwise it evaluates to 0. If
824non-zero, the value gives how many bytes starting at C<s> comprise the code
825point's representation.
826
827=cut
828 */
0c58a72b
KW
829#define UTF8_IS_NONCHAR(s, e) \
830 UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s, e)
7131f24d 831
c867b360
JH
832#define UNICODE_SURROGATE_FIRST 0xD800
833#define UNICODE_SURROGATE_LAST 0xDFFF
834#define UNICODE_REPLACEMENT 0xFFFD
835#define UNICODE_BYTE_ORDER_MARK 0xFEFF
1d72bdf6 836
b851fbc1 837/* Though our UTF-8 encoding can go beyond this,
c76687c5 838 * let's be conservative and do as Unicode says. */
b851fbc1
JH
839#define PERL_UNICODE_MAX 0x10FFFF
840
d044b7a7
KW
841#define UNICODE_WARN_SURROGATE 0x0001 /* UTF-16 surrogates */
842#define UNICODE_WARN_NONCHAR 0x0002 /* Non-char code points */
843#define UNICODE_WARN_SUPER 0x0004 /* Above 0x10FFFF */
844#define UNICODE_WARN_PERL_EXTENDED 0x0008 /* Above 0x7FFF_FFFF */
845#define UNICODE_WARN_ABOVE_31_BIT UNICODE_WARN_PERL_EXTENDED
846#define UNICODE_DISALLOW_SURROGATE 0x0010
847#define UNICODE_DISALLOW_NONCHAR 0x0020
848#define UNICODE_DISALLOW_SUPER 0x0040
849#define UNICODE_DISALLOW_PERL_EXTENDED 0x0080
850#define UNICODE_DISALLOW_ABOVE_31_BIT UNICODE_DISALLOW_PERL_EXTENDED
33f38593
KW
851
852#define UNICODE_GOT_SURROGATE UNICODE_DISALLOW_SURROGATE
853#define UNICODE_GOT_NONCHAR UNICODE_DISALLOW_NONCHAR
854#define UNICODE_GOT_SUPER UNICODE_DISALLOW_SUPER
855#define UNICODE_GOT_PERL_EXTENDED UNICODE_DISALLOW_PERL_EXTENDED
856
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)
1d72bdf6 895
ec34087a
KW
896#define LATIN_SMALL_LETTER_SHARP_S LATIN_SMALL_LETTER_SHARP_S_NATIVE
897#define LATIN_SMALL_LETTER_Y_WITH_DIAERESIS \
898 LATIN_SMALL_LETTER_Y_WITH_DIAERESIS_NATIVE
899#define MICRO_SIGN MICRO_SIGN_NATIVE
900#define LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE \
901 LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE_NATIVE
902#define LATIN_SMALL_LETTER_A_WITH_RING_ABOVE \
903 LATIN_SMALL_LETTER_A_WITH_RING_ABOVE_NATIVE
09091399
JH
904#define UNICODE_GREEK_CAPITAL_LETTER_SIGMA 0x03A3
905#define UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA 0x03C2
906#define UNICODE_GREEK_SMALL_LETTER_SIGMA 0x03C3
9dcbe121 907#define GREEK_SMALL_LETTER_MU 0x03BC
9e682c18
KW
908#define GREEK_CAPITAL_LETTER_MU 0x039C /* Upper and title case
909 of MICRON */
910#define LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS 0x0178 /* Also is title case */
0766489e
KW
911#ifdef LATIN_CAPITAL_LETTER_SHARP_S_UTF8
912# define LATIN_CAPITAL_LETTER_SHARP_S 0x1E9E
913#endif
74894415
KW
914#define LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE 0x130
915#define LATIN_SMALL_LETTER_DOTLESS_I 0x131
9e682c18 916#define LATIN_SMALL_LETTER_LONG_S 0x017F
a9f50d33
KW
917#define LATIN_SMALL_LIGATURE_LONG_S_T 0xFB05
918#define LATIN_SMALL_LIGATURE_ST 0xFB06
9e682c18
KW
919#define KELVIN_SIGN 0x212A
920#define ANGSTROM_SIGN 0x212B
09091399 921
9e55ce06 922#define UNI_DISPLAY_ISPRINT 0x0001
c728cb41
JH
923#define UNI_DISPLAY_BACKSLASH 0x0002
924#define UNI_DISPLAY_QQ (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH)
925#define UNI_DISPLAY_REGEX (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH)
9e55ce06 926
5cd46e1f
KW
927#define ANYOF_FOLD_SHARP_S(node, input, end) \
928 (ANYOF_BITMAP_TEST(node, LATIN_SMALL_LETTER_SHARP_S) && \
137165a6 929 (ANYOF_NONBITMAP(node)) && \
39065660 930 (ANYOF_FLAGS(node) & ANYOF_LOC_NONBITMAP_FOLD) && \
07b6858f 931 ((end) > (input) + 1) && \
305b8651 932 isALPHA_FOLD_EQ((input)[0], 's'))
6302f837 933
ebc501f0 934#define SHARP_S_SKIP 2
3b0fc154 935
3cedd9d9 936#define is_utf8_char_buf(buf, buf_end) isUTF8_CHAR(buf, buf_end)
976c1b08
KW
937#define bytes_from_utf8(s, lenp, is_utf8p) \
938 bytes_from_utf8_loc(s, lenp, is_utf8p, 0)
3cedd9d9 939
e23e8bc1
KW
940/*
941
25e3a4e0
KW
942=for apidoc Am|STRLEN|isUTF8_CHAR_flags|const U8 *s|const U8 *e| const U32 flags
943
944Evaluates to non-zero if the first few bytes of the string starting at C<s> and
945looking no further than S<C<e - 1>> are well-formed UTF-8, as extended by Perl,
946that represents some code point, subject to the restrictions given by C<flags>;
947otherwise it evaluates to 0. If non-zero, the value gives how many bytes
2717076a
KW
948starting at C<s> comprise the code point's representation. Any bytes remaining
949before C<e>, but beyond the ones needed to form the first code point in C<s>,
950are not examined.
25e3a4e0
KW
951
952If C<flags> is 0, this gives the same results as C<L</isUTF8_CHAR>>;
953if C<flags> is C<UTF8_DISALLOW_ILLEGAL_INTERCHANGE>, this gives the same results
954as C<L</isSTRICT_UTF8_CHAR>>;
955and if C<flags> is C<UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE>, this gives
956the same results as C<L</isC9_STRICT_UTF8_CHAR>>.
957Otherwise C<flags> may be any combination of the C<UTF8_DISALLOW_I<foo>> flags
958understood by C<L</utf8n_to_uvchr>>, with the same meanings.
959
960The three alternative macros are for the most commonly needed validations; they
961are likely to run somewhat faster than this more general one, as they can be
962inlined into your code.
963
9f2abfde
KW
964Use L</is_utf8_string_flags>, L</is_utf8_string_loc_flags>, and
965L</is_utf8_string_loclen_flags> to check entire strings.
966
25e3a4e0
KW
967=cut
968*/
969
970#define isUTF8_CHAR_flags(s, e, flags) \
971 (UNLIKELY((e) <= (s)) \
972 ? 0 \
973 : (UTF8_IS_INVARIANT(*s)) \
974 ? 1 \
975 : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \
976 ? 0 \
977 : _is_utf8_char_helper(s, e, flags))
978
6302f837
KW
979/* Do not use; should be deprecated. Use isUTF8_CHAR() instead; this is
980 * retained solely for backwards compatibility */
981#define IS_UTF8_CHAR(p, n) (isUTF8_CHAR(p, (p) + (n)) == n)
e9a8c099 982
6a5bc5ac 983#endif /* PERL_UTF8_H_ */
57f0e7e2 984
e9a8c099 985/*
14d04a33 986 * ex: set ts=8 sts=4 sw=4 et:
e9a8c099 987 */