This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Update IO-Compress to CPAN version 2.093
[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) \
9a9a6c98 72 utf8_to_uvchr_buf_helper((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 79
eda9cac1
KW
80#define foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \
81 foldEQ_utf8_flags(s1, pe1, l1, u1, s2, pe2, l2, u2, 0)
baa60164 82#define FOLDEQ_UTF8_NOMIX_ASCII (1 << 0)
cea315b6 83#define FOLDEQ_LOCALE (1 << 1)
18f762c3
KW
84#define FOLDEQ_S1_ALREADY_FOLDED (1 << 2)
85#define FOLDEQ_S2_ALREADY_FOLDED (1 << 3)
d635b710
KW
86#define FOLDEQ_S1_FOLDS_SANE (1 << 4)
87#define FOLDEQ_S2_FOLDS_SANE (1 << 5)
a33c29bc 88
e6226b18
KW
89#define ibcmp_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \
90 cBOOL(! foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2))
91
1d72bdf6
NIS
92#ifdef EBCDIC
93/* The equivalent of these macros but implementing UTF-EBCDIC
94 are in the following header file:
95 */
96
97#include "utfebcdic.h"
fd7cb289 98
d06134e5 99#else /* ! EBCDIC */
73c4f7a1
GS
100START_EXTERN_C
101
f4d83e55
KW
102/*
103
104=for apidoc AmnU|STRLEN|UTF8_MAXBYTES
105
106The maximum width of a single UTF-8 encoded character, in bytes.
107
108NOTE: Strictly speaking Perl's UTF-8 should not be called UTF-8 since UTF-8
109is an encoding of Unicode, and Unicode's upper limit, 0x10FFFF, can be
110expressed with 4 bytes. However, Perl thinks of UTF-8 as a way to encode
111non-negative integers in a binary format, even those above Unicode.
112
113=cut
114 */
111e8ed9
KW
115#define UTF8_MAXBYTES 13
116
a0ed51b3 117#ifdef DOINIT
6f06b55f 118EXTCONST unsigned char PL_utf8skip[] = {
b2635aa8
KW
119/* 0x00 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
120/* 0x10 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
121/* 0x20 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
122/* 0x30 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
123/* 0x40 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
124/* 0x50 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
125/* 0x60 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
126/* 0x70 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
127/* 0x80 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
128/* 0x90 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
129/* 0xA0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
130/* 0xB0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
131/* 0xC0 */ 2,2, /* overlong */
1ff3baa2 132/* 0xC2 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0080 to U+03FF */
b2635aa8
KW
133/* 0xD0 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0400 to U+07FF */
134/* 0xE0 */ 3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3, /* U+0800 to U+FFFF */
135/* 0xF0 */ 4,4,4,4,4,4,4,4,5,5,5,5,6,6, /* above BMP to 2**31 - 1 */
6937f885
KW
136 /* Perl extended (never was official UTF-8). Up to 36 bit */
137/* 0xFE */ 7,
138 /* More extended, Up to 72 bits (64-bit + reserved) */
111e8ed9 139/* 0xFF */ UTF8_MAXBYTES
a0ed51b3
LW
140};
141#else
6f06b55f 142EXTCONST unsigned char PL_utf8skip[];
a0ed51b3
LW
143#endif
144
73c4f7a1 145END_EXTERN_C
7e2040f0 146
ef2a4c8f
KW
147/*
148
149=for apidoc Am|U8|NATIVE_TO_LATIN1|U8 ch
150
151Returns the Latin-1 (including ASCII and control characters) equivalent of the
152input native code point given by C<ch>. Thus, C<NATIVE_TO_LATIN1(193)> on
153EBCDIC platforms returns 65. These each represent the character C<"A"> on
154their respective platforms. On ASCII platforms no conversion is needed, so
155this macro expands to just its input, adding no time nor space requirements to
156the implementation.
157
158For conversion of code points potentially larger than will fit in a character,
159use L</NATIVE_TO_UNI>.
160
161=for apidoc Am|U8|LATIN1_TO_NATIVE|U8 ch
162
163Returns the native equivalent of the input Latin-1 code point (including ASCII
164and control characters) given by C<ch>. Thus, C<LATIN1_TO_NATIVE(66)> on
165EBCDIC platforms returns 194. These each represent the character C<"B"> on
166their respective platforms. On ASCII platforms no conversion is needed, so
167this macro expands to just its input, adding no time nor space requirements to
168the implementation.
169
170For conversion of code points potentially larger than will fit in a character,
171use L</UNI_TO_NATIVE>.
172
173=for apidoc Am|UV|NATIVE_TO_UNI|UV ch
174
175Returns the Unicode equivalent of the input native code point given by C<ch>.
176Thus, C<NATIVE_TO_UNI(195)> on EBCDIC platforms returns 67. These each
177represent the character C<"C"> on their respective platforms. On ASCII
178platforms no conversion is needed, so this macro expands to just its input,
179adding no time nor space requirements to the implementation.
180
181=for apidoc Am|UV|UNI_TO_NATIVE|UV ch
182
183Returns the native equivalent of the input Unicode code point given by C<ch>.
184Thus, C<UNI_TO_NATIVE(68)> on EBCDIC platforms returns 196. These each
185represent the character C<"D"> on their respective platforms. On ASCII
186platforms no conversion is needed, so this macro expands to just its input,
187adding no time nor space requirements to the implementation.
188
189=cut
190*/
191
e7ae132e
KW
192#define NATIVE_TO_LATIN1(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) ((ch) | 0)))
193#define LATIN1_TO_NATIVE(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) ((ch) | 0)))
59a449d5
KW
194
195/* I8 is an intermediate version of UTF-8 used only in UTF-EBCDIC. We thus
196 * consider it to be identical to UTF-8 on ASCII platforms. Strictly speaking
197 * UTF-8 and UTF-EBCDIC are two different things, but we often conflate them
198 * because they are 8-bit encodings that serve the same purpose in Perl, and
199 * rarely do we need to distinguish them. The term "NATIVE_UTF8" applies to
200 * whichever one is applicable on the current platform */
cb15eeb2
KW
201#define NATIVE_UTF8_TO_I8(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) ((ch) | 0)))
202#define I8_TO_NATIVE_UTF8(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) ((ch) | 0)))
59a449d5 203
cb15eeb2
KW
204#define UNI_TO_NATIVE(ch) ((UV) ((ch) | 0))
205#define NATIVE_TO_UNI(ch) ((UV) ((ch) | 0))
d7578b48 206
877d9f0d 207/*
9041c2e3 208
a14e0a36
KW
209 The following table is from Unicode 3.2, plus the Perl extensions for above
210 U+10FFFF
877d9f0d 211
a14e0a36 212 Code Points 1st Byte 2nd Byte 3rd 4th 5th 6th 7th 8th-13th
877d9f0d 213
375122d7 214 U+0000..U+007F 00..7F
e1b711da 215 U+0080..U+07FF * C2..DF 80..BF
a14e0a36
KW
216 U+0800..U+0FFF E0 * A0..BF 80..BF
217 U+1000..U+CFFF E1..EC 80..BF 80..BF
218 U+D000..U+D7FF ED 80..9F 80..BF
219 U+D800..U+DFFF ED A0..BF 80..BF (surrogates)
220 U+E000..U+FFFF EE..EF 80..BF 80..BF
221 U+10000..U+3FFFF F0 * 90..BF 80..BF 80..BF
222 U+40000..U+FFFFF F1..F3 80..BF 80..BF 80..BF
223 U+100000..U+10FFFF F4 80..8F 80..BF 80..BF
224 Below are above-Unicode code points
225 U+110000..U+13FFFF F4 90..BF 80..BF 80..BF
226 U+110000..U+1FFFFF F5..F7 80..BF 80..BF 80..BF
227 U+200000..U+FFFFFF F8 * 88..BF 80..BF 80..BF 80..BF
228U+1000000..U+3FFFFFF F9..FB 80..BF 80..BF 80..BF 80..BF
229U+4000000..U+3FFFFFFF FC * 84..BF 80..BF 80..BF 80..BF 80..BF
230U+40000000..U+7FFFFFFF FD 80..BF 80..BF 80..BF 80..BF 80..BF
231U+80000000..U+FFFFFFFFF FE * 82..BF 80..BF 80..BF 80..BF 80..BF 80..BF
232U+1000000000.. FF 80..BF 80..BF 80..BF 80..BF 80..BF * 81..BF 80..BF
877d9f0d 233
e1b711da 234Note the gaps before several of the byte entries above marked by '*'. These are
37e2e78e
KW
235caused by legal UTF-8 avoiding non-shortest encodings: it is technically
236possible to UTF-8-encode a single code point in different ways, but that is
237explicitly forbidden, and the shortest possible encoding should always be used
15824458 238(and that is what Perl does). The non-shortest ones are called 'overlongs'.
8c007b5a 239
877d9f0d
JH
240 */
241
8c007b5a
JH
242/*
243 Another way to look at it, as bits:
244
b2635aa8 245 Code Points 1st Byte 2nd Byte 3rd Byte 4th Byte
8c007b5a 246
b2635aa8
KW
247 0aaa aaaa 0aaa aaaa
248 0000 0bbb bbaa aaaa 110b bbbb 10aa aaaa
249 cccc bbbb bbaa aaaa 1110 cccc 10bb bbbb 10aa aaaa
250 00 000d ddcc cccc bbbb bbaa aaaa 1111 0ddd 10cc cccc 10bb bbbb 10aa aaaa
8c007b5a
JH
251
252As you can see, the continuation bytes all begin with C<10>, and the
e1b711da 253leading bits of the start byte tell how many bytes there are in the
8c007b5a
JH
254encoded character.
255
df863e43
KW
256Perl's extended UTF-8 means we can have start bytes up through FF, though any
257beginning with FF yields a code point that is too large for 32-bit ASCII
258platforms. FF signals to use 13 bytes for the encoded character. This breaks
259the paradigm that the number of leading bits gives how many total bytes there
ab2e28c2 260are in the character. */
38953e5a 261
15824458
KW
262/* This is the number of low-order bits a continuation byte in a UTF-8 encoded
263 * sequence contributes to the specification of the code point. In the bit
264 * maps above, you see that the first 2 bits are a constant '10', leaving 6 of
265 * real information */
1d72bdf6 266#define UTF_ACCUMULATION_SHIFT 6
b2635aa8 267
fed423a5
KW
268/* ^? is defined to be DEL on ASCII systems. See the definition of toCTRL()
269 * for more */
270#define QUESTION_MARK_CTRL DEL_NATIVE
271
272/* Surrogates, non-character code points and above-Unicode code points are
273 * problematic in some contexts. This allows code that needs to check for
274 * those to to quickly exclude the vast majority of code points it will
275 * encounter */
a6951642
KW
276#define isUTF8_POSSIBLY_PROBLEMATIC(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
277 (U8) c >= 0xED)
fed423a5 278
57ff5f59
KW
279#define UNICODE_IS_PERL_EXTENDED(uv) UNLIKELY((UV) (uv) > 0x7FFFFFFF)
280
fed423a5
KW
281#endif /* EBCDIC vs ASCII */
282
d7bcd45a
KW
283/* 2**UTF_ACCUMULATION_SHIFT - 1. This masks out all but the bits that carry
284 * real information in a continuation byte. This turns out to be 0x3F in
285 * UTF-8, 0x1F in UTF-EBCDIC. */
fed423a5
KW
286#define UTF_CONTINUATION_MASK ((U8) ((1U << UTF_ACCUMULATION_SHIFT) - 1))
287
9f3cfb7a
KW
288/* For use in UTF8_IS_CONTINUATION(). This turns out to be 0xC0 in UTF-8,
289 * E0 in UTF-EBCDIC */
290#define UTF_IS_CONTINUATION_MASK ((U8) (0xFF << UTF_ACCUMULATION_SHIFT))
291
38f458ff
KW
292/* This defines the bits that are to be in the continuation bytes of a
293 * multi-byte UTF-8 encoded character that mark it is a continuation byte.
294 * This turns out to be 0x80 in UTF-8, 0xA0 in UTF-EBCDIC. (khw doesn't know
295 * the underlying reason that B0 works here) */
296#define UTF_CONTINUATION_MARK (UTF_IS_CONTINUATION_MASK & 0xB0)
297
f4225fa0
KW
298/* Is the byte 'c' part of a multi-byte UTF8-8 encoded sequence, and not the
299 * first byte thereof? */
300#define UTF8_IS_CONTINUATION(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
301 (((NATIVE_UTF8_TO_I8(c) & UTF_IS_CONTINUATION_MASK) \
302 == UTF_CONTINUATION_MARK)))
303
2dc97505
KW
304/* Is the representation of the Unicode code point 'cp' the same regardless of
305 * being encoded in UTF-8 or not? This is a fundamental property of
306 * UTF-8,EBCDIC */
307#define OFFUNI_IS_INVARIANT(c) (((WIDEST_UTYPE)(c)) < UTF_CONTINUATION_MARK)
308
ab2e28c2
KW
309/*
310=for apidoc Am|bool|UVCHR_IS_INVARIANT|UV cp
311
312Evaluates to 1 if the representation of code point C<cp> is the same whether or
313not it is encoded in UTF-8; otherwise evaluates to 0. UTF-8 invariant
314characters can be copied as-is when converting to/from UTF-8, saving time.
315C<cp> is Unicode if above 255; otherwise is platform-native.
316
317=cut
318 */
319#define UVCHR_IS_INVARIANT(cp) (OFFUNI_IS_INVARIANT(NATIVE_TO_UNI(cp)))
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
4bab39bc
KW
379#define UTF_MIN_START_BYTE \
380 ((UTF_CONTINUATION_MARK >> UTF_ACCUMULATION_SHIFT) | UTF_START_MARK(2))
381
382/* Is the byte 'c' the first byte of a multi-byte UTF8-8 encoded sequence?
383 * This doesn't catch invariants (they are single-byte). It also excludes the
384 * illegal overlong sequences that begin with C0 and C1 on ASCII platforms, and
385 * C0-C4 I8 start bytes on EBCDIC ones */
386#define UTF8_IS_START(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
387 (NATIVE_UTF8_TO_I8(c) >= UTF_MIN_START_BYTE))
388
1df63428
KW
389#define UTF_MIN_ABOVE_LATIN1_BYTE \
390 ((0x100 >> UTF_ACCUMULATION_SHIFT) | UTF_START_MARK(2))
391
392/* Is the UTF8-encoded byte 'c' the first byte of a sequence of bytes that
393 * represent a code point > 255? */
394#define UTF8_IS_ABOVE_LATIN1(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
395 (NATIVE_UTF8_TO_I8(c) >= UTF_MIN_ABOVE_LATIN1_BYTE))
396
7c88d61e
KW
397/* Is the UTF8-encoded byte 'c' the first byte of a two byte sequence? Use
398 * UTF8_IS_NEXT_CHAR_DOWNGRADEABLE() instead if the input isn't known to
399 * be well-formed. */
400#define UTF8_IS_DOWNGRADEABLE_START(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
401 inRANGE(NATIVE_UTF8_TO_I8(c), \
402 UTF_MIN_START_BYTE, UTF_MIN_ABOVE_LATIN1_BYTE - 1))
403
b651802e
KW
404/* The largest code point representable by two UTF-8 bytes on this platform.
405 * As explained in the comments for __COMMON_UNI_SKIP, 32 start bytes with
fed423a5 406 * UTF_ACCUMULATION_SHIFT bits of information each */
aa206fb7
KW
407#define MAX_UTF8_TWO_BYTE (32 * (1U << UTF_ACCUMULATION_SHIFT) - 1)
408
b651802e
KW
409/* The largest code point representable by two UTF-8 bytes on any platform that
410 * Perl runs on. This value is constrained by EBCDIC which has 5 bits per
411 * continuation byte */
aa206fb7
KW
412#define MAX_PORTABLE_UTF8_TWO_BYTE (32 * (1U << 5) - 1)
413
f2c50040
KW
414/*
415
416=for apidoc AmnU|STRLEN|UTF8_MAXBYTES_CASE
417
418The maximum number of UTF-8 bytes a single Unicode character can
419uppercase/lowercase/titlecase/fold into.
420
421=cut
422
423 * Unicode guarantees that the maximum expansion is UTF8_MAX_FOLD_CHAR_EXPAND
424 * characters, but any above-Unicode code point will fold to itself, so we only
425 * have to look at the expansion of the maximum Unicode code point. But this
426 * number may be less than the space occupied by a very large code point under
427 * Perl's extended UTF-8. We have to make it large enough to fit any single
428 * character. (It turns out that ASCII and EBCDIC differ in which is larger)
429 *
430=cut
431*/
c03c0950 432#define UTF8_MAXBYTES_CASE \
ae9a9edb 433 MAX(UTF8_MAXBYTES, UTF8_MAX_FOLD_CHAR_EXPAND * OFFUNISKIP(0x10FFFF))
c03c0950 434
d06134e5
KW
435/* Rest of these are attributes of Unicode and perl's internals rather than the
436 * encoding, or happen to be the same in both ASCII and EBCDIC (at least at
437 * this level; the macros that some of these call may have different
438 * definitions in the two encodings */
439
59a449d5
KW
440/* In domain restricted to ASCII, these may make more sense to the reader than
441 * the ones with Latin1 in the name */
442#define NATIVE_TO_ASCII(ch) NATIVE_TO_LATIN1(ch)
443#define ASCII_TO_NATIVE(ch) LATIN1_TO_NATIVE(ch)
444
445/* More or less misleadingly-named defines, retained for back compat */
446#define NATIVE_TO_UTF(ch) NATIVE_UTF8_TO_I8(ch)
447#define NATIVE_TO_I8(ch) NATIVE_UTF8_TO_I8(ch)
448#define UTF_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch)
449#define I8_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch)
450#define NATIVE8_TO_UNI(ch) NATIVE_TO_LATIN1(ch)
d06134e5 451
c0236afe 452/* This defines the 1-bits that are to be in the first byte of a multi-byte
97d0ceda
KW
453 * UTF-8 encoded character that mark it as a start byte and give the number of
454 * bytes that comprise the character. 'len' is the number of bytes in the
455 * multi-byte sequence. */
b5ad20ca 456#define UTF_START_MARK(len) (((len) > 7) ? 0xFF : ((U8) (0xFE << (7-(len)))))
c0236afe
KW
457
458/* Masks out the initial one bits in a start byte, leaving the real data ones.
459 * Doesn't work on an invariant byte. 'len' is the number of bytes in the
460 * multi-byte sequence that comprises the character. */
461#define UTF_START_MASK(len) (((len) >= 7) ? 0x00 : (0x1F >> ((len)-2)))
462
537124e4
KW
463/* Adds a UTF8 continuation byte 'new' of information to a running total code
464 * point 'old' of all the continuation bytes so far. This is designed to be
155d2738
KW
465 * used in a loop to convert from UTF-8 to the code point represented. Note
466 * that this is asymmetric on EBCDIC platforms, in that the 'new' parameter is
467 * the UTF-EBCDIC byte, whereas the 'old' parameter is a Unicode (not EBCDIC)
468 * code point in process of being generated */
a6951642
KW
469#define UTF8_ACCUMULATE(old, new) (__ASSERT_(FITS_IN_8_BITS(new)) \
470 ((old) << UTF_ACCUMULATION_SHIFT) \
009097b1 471 | ((NATIVE_UTF8_TO_I8(new)) \
155d2738 472 & UTF_CONTINUATION_MASK))
d06134e5 473
4ab10950 474/* This works in the face of malformed UTF-8. */
4e1ed312
KW
475#define UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(s, e) \
476 ( UTF8_IS_DOWNGRADEABLE_START(*(s)) \
477 && ( (e) - (s) > 1) \
478 && UTF8_IS_CONTINUATION(*((s)+1)))
4ab10950 479
5aaebcb3 480/* Number of bytes a code point occupies in UTF-8. */
5352a763 481#define NATIVE_SKIP(uv) UVCHR_SKIP(uv)
bd18bd40 482
5aaebcb3
KW
483/* Most code which says UNISKIP is really thinking in terms of native code
484 * points (0-255) plus all those beyond. This is an imprecise term, but having
2accb712 485 * it means existing code continues to work. For precision, use UVCHR_SKIP,
5352a763
KW
486 * NATIVE_SKIP, or OFFUNISKIP */
487#define UNISKIP(uv) UVCHR_SKIP(uv)
5aaebcb3 488
3c0792e4
KW
489/* Longer, but more accurate name */
490#define UTF8_IS_ABOVE_LATIN1_START(c) UTF8_IS_ABOVE_LATIN1(c)
491
a62b247b
KW
492/* Convert a UTF-8 variant Latin1 character to a native code point value.
493 * Needs just one iteration of accumulate. Should be used only if it is known
494 * that the code point is < 256, and is not UTF-8 invariant. Use the slower
495 * but more general TWO_BYTE_UTF8_TO_NATIVE() which handles any code point
496 * representable by two bytes (which turns out to be up through
497 * MAX_PORTABLE_UTF8_TWO_BYTE). The two parameters are:
498 * HI: a downgradable start byte;
499 * LO: continuation.
500 * */
501#define EIGHT_BIT_UTF8_TO_NATIVE(HI, LO) \
502 ( __ASSERT_(UTF8_IS_DOWNGRADEABLE_START(HI)) \
503 __ASSERT_(UTF8_IS_CONTINUATION(LO)) \
504 LATIN1_TO_NATIVE(UTF8_ACCUMULATE(( \
505 NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), (LO))))
506
94bb8c36 507/* Convert a two (not one) byte utf8 character to a native code point value.
2950f2a7
KW
508 * Needs just one iteration of accumulate. Should not be used unless it is
509 * known that the two bytes are legal: 1) two-byte start, and 2) continuation.
510 * Note that the result can be larger than 255 if the input character is not
511 * downgradable */
94bb8c36 512#define TWO_BYTE_UTF8_TO_NATIVE(HI, LO) \
a6951642
KW
513 (__ASSERT_(FITS_IN_8_BITS(HI)) \
514 __ASSERT_(FITS_IN_8_BITS(LO)) \
515 __ASSERT_(PL_utf8skip[HI] == 2) \
516 __ASSERT_(UTF8_IS_CONTINUATION(LO)) \
94bb8c36 517 UNI_TO_NATIVE(UTF8_ACCUMULATE((NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), \
635e76f5 518 (LO))))
94bb8c36
KW
519
520/* Should never be used, and be deprecated */
521#define TWO_BYTE_UTF8_TO_UNI(HI, LO) NATIVE_TO_UNI(TWO_BYTE_UTF8_TO_NATIVE(HI, LO))
2950f2a7 522
bd18bd40
KW
523/*
524
525=for apidoc Am|STRLEN|UTF8SKIP|char* s
ee0ff0f5
KW
526returns the number of bytes a non-malformed UTF-8 encoded character whose first
527(perhaps only) byte is pointed to by C<s>.
528
529If there is a possibility of malformed input, use instead:
530
531=over
532
533=item L</C<UTF8_SAFE_SKIP>> if you know the maximum ending pointer in the
534buffer pointed to by C<s>; or
535
536=item L</C<UTF8_CHK_SKIP>> if you don't know it.
537
538=back
539
540It is better to restructure your code so the end pointer is passed down so that
541you know what it actually is at the point of this call, but if that isn't
542possible, L</C<UTF8_CHK_SKIP>> can minimize the chance of accessing beyond the end
543of the input buffer.
bd18bd40
KW
544
545=cut
546 */
2a70536e 547#define UTF8SKIP(s) PL_utf8skip[*(const U8*)(s)]
a281f16c
KW
548
549/*
550=for apidoc Am|STRLEN|UTF8_SKIP|char* s
551This is a synonym for L</C<UTF8SKIP>>
552
553=cut
554*/
555
2a70536e 556#define UTF8_SKIP(s) UTF8SKIP(s)
d06134e5 557
85fcc8f2 558/*
ee0ff0f5
KW
559=for apidoc Am|STRLEN|UTF8_CHK_SKIP|char* s
560
561This is a safer version of L</C<UTF8SKIP>>, but still not as safe as
562L</C<UTF8_SAFE_SKIP>>. This version doesn't blindly assume that the input
563string pointed to by C<s> is well-formed, but verifies that there isn't a NUL
564terminating character before the expected end of the next character in C<s>.
565The length C<UTF8_CHK_SKIP> returns stops just before any such NUL.
566
567Perl tends to add NULs, as an insurance policy, after the end of strings in
568SV's, so it is likely that using this macro will prevent inadvertent reading
569beyond the end of the input buffer, even if it is malformed UTF-8.
570
571This macro is intended to be used by XS modules where the inputs could be
572malformed, and it isn't feasible to restructure to use the safer
573L</C<UTF8_SAFE_SKIP>>, for example when interfacing with a C library.
574
575=cut
576*/
577
578#define UTF8_CHK_SKIP(s) \
f87d8789
KW
579 (s[0] == '\0' ? 1 : MIN(UTF8SKIP(s), \
580 my_strnlen((char *) (s), UTF8SKIP(s))))
ee0ff0f5 581/*
85fcc8f2
KW
582
583=for apidoc Am|STRLEN|UTF8_SAFE_SKIP|char* s|char* e
45671da2
KW
584returns 0 if S<C<s E<gt>= e>>; otherwise returns the number of bytes in the
585UTF-8 encoded character whose first byte is pointed to by C<s>. But it never
586returns beyond C<e>. On DEBUGGING builds, it asserts that S<C<s E<lt>= e>>.
85fcc8f2
KW
587
588=cut
589 */
45671da2
KW
590#define UTF8_SAFE_SKIP(s, e) (__ASSERT_((e) >= (s)) \
591 ((e) - (s)) <= 0 \
592 ? 0 \
593 : MIN(((e) - (s)), UTF8_SKIP(s)))
85fcc8f2 594
2d1545e5
KW
595/* Most code that says 'UNI_' really means the native value for code points up
596 * through 255 */
597#define UNI_IS_INVARIANT(cp) UVCHR_IS_INVARIANT(cp)
598
c2b32798
KW
599/*
600=for apidoc Am|bool|UTF8_IS_INVARIANT|char c
601
602Evaluates to 1 if the byte C<c> represents the same character when encoded in
603UTF-8 as when not; otherwise evaluates to 0. UTF-8 invariant characters can be
604copied as-is when converting to/from UTF-8, saving time.
605
606In spite of the name, this macro gives the correct result if the input string
607from which C<c> comes is not encoded in UTF-8.
608
609See C<L</UVCHR_IS_INVARIANT>> for checking if a UV is invariant.
610
611=cut
612
613The reason it works on both UTF-8 encoded strings and non-UTF-8 encoded, is
614that it returns TRUE in each for the exact same set of bit patterns. It is
615valid on a subset of what UVCHR_IS_INVARIANT is valid on, so can just use that;
616and the compiler should optimize out anything extraneous given the
617implementation of the latter. The |0 makes sure this isn't mistakenly called
618with a ptr argument.
619*/
5c06326b 620#define UTF8_IS_INVARIANT(c) UVCHR_IS_INVARIANT((c) | 0)
5fc230f1
KW
621
622/* Like the above, but its name implies a non-UTF8 input, which as the comments
623 * above show, doesn't matter as to its implementation */
38953e5a 624#define NATIVE_BYTE_IS_INVARIANT(c) UVCHR_IS_INVARIANT(c)
d06134e5 625
2c03e801
KW
626/* Misleadingly named: is the UTF8-encoded byte 'c' part of a variant sequence
627 * in UTF-8? This is the inverse of UTF8_IS_INVARIANT. */
628#define UTF8_IS_CONTINUED(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
629 (! UTF8_IS_INVARIANT(c)))
630
48ccf5e1
KW
631/* The macros in the next 4 sets are used to generate the two utf8 or utfebcdic
632 * bytes from an ordinal that is known to fit into exactly two (not one) bytes;
633 * it must be less than 0x3FF to work across both encodings. */
634
635/* These two are helper macros for the other three sets, and should not be used
636 * directly anywhere else. 'translate_function' is either NATIVE_TO_LATIN1
1ff3baa2
KW
637 * (which works for code points up through 0xFF) or NATIVE_TO_UNI which works
638 * for any code point */
48ccf5e1 639#define __BASE_TWO_BYTE_HI(c, translate_function) \
2863dafa 640 (__ASSERT_(! UVCHR_IS_INVARIANT(c)) \
48ccf5e1 641 I8_TO_NATIVE_UTF8((translate_function(c) >> UTF_ACCUMULATION_SHIFT) \
2863dafa 642 | UTF_START_MARK(2)))
48ccf5e1 643#define __BASE_TWO_BYTE_LO(c, translate_function) \
2863dafa 644 (__ASSERT_(! UVCHR_IS_INVARIANT(c)) \
48ccf5e1 645 I8_TO_NATIVE_UTF8((translate_function(c) & UTF_CONTINUATION_MASK) \
2863dafa 646 | UTF_CONTINUATION_MARK))
48ccf5e1 647
48ccf5e1
KW
648/* The next two macros should not be used. They were designed to be usable as
649 * the case label of a switch statement, but this doesn't work for EBCDIC. Use
9d0d3a03 650 * regen/unicode_constants.pl instead */
48ccf5e1
KW
651#define UTF8_TWO_BYTE_HI_nocast(c) __BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI)
652#define UTF8_TWO_BYTE_LO_nocast(c) __BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI)
653
654/* The next two macros are used when the source should be a single byte
655 * character; checked for under DEBUGGING */
656#define UTF8_EIGHT_BIT_HI(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
4c8cd605 657 ( __BASE_TWO_BYTE_HI(c, NATIVE_TO_LATIN1)))
48ccf5e1 658#define UTF8_EIGHT_BIT_LO(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
4c8cd605 659 (__BASE_TWO_BYTE_LO(c, NATIVE_TO_LATIN1)))
48ccf5e1
KW
660
661/* These final two macros in the series are used when the source can be any
662 * code point whose UTF-8 is known to occupy 2 bytes; they are less efficient
663 * than the EIGHT_BIT versions on EBCDIC platforms. We use the logical '~'
664 * operator instead of "<=" to avoid getting compiler warnings.
d52b8576 665 * MAX_UTF8_TWO_BYTE should be exactly all one bits in the lower few
48ccf5e1
KW
666 * places, so the ~ works */
667#define UTF8_TWO_BYTE_HI(c) \
668 (__ASSERT_((sizeof(c) == 1) \
d52b8576 669 || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \
4c8cd605 670 (__BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI)))
48ccf5e1
KW
671#define UTF8_TWO_BYTE_LO(c) \
672 (__ASSERT_((sizeof(c) == 1) \
d52b8576 673 || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \
4c8cd605 674 (__BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI)))
d06134e5 675
e7214ce8
KW
676/* This is illegal in any well-formed UTF-8 in both EBCDIC and ASCII
677 * as it is only in overlongs. */
678#define ILLEGAL_UTF8_BYTE I8_TO_NATIVE_UTF8(0xC1)
679
7e2040f0 680/*
e3036cf4 681 * 'UTF' is whether or not p is encoded in UTF8. The names 'foo_lazy_if' stem
20df05f4
KW
682 * from an earlier version of these macros in which they didn't call the
683 * foo_utf8() macros (i.e. were 'lazy') unless they decided that *p is the
684 * beginning of a utf8 character. Now that foo_utf8() determines that itself,
685 * no need to do it again here
7e2040f0 686 */
da8c1a98
KW
687#define isIDFIRST_lazy_if_safe(p, e, UTF) \
688 ((IN_BYTES || !UTF) \
689 ? isIDFIRST(*(p)) \
690 : isIDFIRST_utf8_safe(p, e))
da8c1a98
KW
691#define isWORDCHAR_lazy_if_safe(p, e, UTF) \
692 ((IN_BYTES || !UTF) \
693 ? isWORDCHAR(*(p)) \
694 : isWORDCHAR_utf8_safe((U8 *) p, (U8 *) e))
4c1d9526 695#define isALNUM_lazy_if_safe(p, e, UTF) isWORDCHAR_lazy_if_safe(p, e, UTF)
da8c1a98 696
89ebb4a3
JH
697#define UTF8_MAXLEN UTF8_MAXBYTES
698
8cb75cc8
KW
699/* A Unicode character can fold to up to 3 characters */
700#define UTF8_MAX_FOLD_CHAR_EXPAND 3
701
d3481830 702#define IN_BYTES UNLIKELY(CopHINTS_get(PL_curcop) & HINT_BYTES)
bd18bd40
KW
703
704/*
705
706=for apidoc Am|bool|DO_UTF8|SV* sv
707Returns a bool giving whether or not the PV in C<sv> is to be treated as being
708encoded in UTF-8.
709
710You should use this I<after> a call to C<SvPV()> or one of its variants, in
711case any call to string overloading updates the internal UTF-8 encoding flag.
712
713=cut
714*/
0064a8a9 715#define DO_UTF8(sv) (SvUTF8(sv) && !IN_BYTES)
1ff3baa2
KW
716
717/* Should all strings be treated as Unicode, and not just UTF-8 encoded ones?
718 * Is so within 'feature unicode_strings' or 'locale :not_characters', and not
719 * within 'use bytes'. UTF-8 locales are not tested for here, but perhaps
720 * could be */
70844984
KW
721#define IN_UNI_8_BIT \
722 (( ( (CopHINTS_get(PL_curcop) & HINT_UNI_8_BIT)) \
723 || ( CopHINTS_get(PL_curcop) & HINT_LOCALE_PARTIAL \
724 /* -1 below is for :not_characters */ \
725 && _is_in_locale_category(FALSE, -1))) \
726 && (! IN_BYTES))
b36bf33f 727
1d72bdf6 728
c76687c5 729#define UTF8_ALLOW_EMPTY 0x0001 /* Allow a zero length string */
2b5e7bc2 730#define UTF8_GOT_EMPTY UTF8_ALLOW_EMPTY
c76687c5
KW
731
732/* Allow first byte to be a continuation byte */
1d72bdf6 733#define UTF8_ALLOW_CONTINUATION 0x0002
2b5e7bc2 734#define UTF8_GOT_CONTINUATION UTF8_ALLOW_CONTINUATION
c76687c5 735
cd01d3b1 736/* Unexpected non-continuation byte */
1d72bdf6 737#define UTF8_ALLOW_NON_CONTINUATION 0x0004
2b5e7bc2 738#define UTF8_GOT_NON_CONTINUATION UTF8_ALLOW_NON_CONTINUATION
949cf498
KW
739
740/* expecting more bytes than were available in the string */
741#define UTF8_ALLOW_SHORT 0x0008
2b5e7bc2 742#define UTF8_GOT_SHORT UTF8_ALLOW_SHORT
949cf498 743
94953955
KW
744/* Overlong sequence; i.e., the code point can be specified in fewer bytes.
745 * First one will convert the overlong to the REPLACEMENT CHARACTER; second
746 * will return what the overlong evaluates to */
949cf498 747#define UTF8_ALLOW_LONG 0x0010
94953955 748#define UTF8_ALLOW_LONG_AND_ITS_VALUE (UTF8_ALLOW_LONG|0x0020)
2b5e7bc2
KW
749#define UTF8_GOT_LONG UTF8_ALLOW_LONG
750
d60baaa7
KW
751#define UTF8_ALLOW_OVERFLOW 0x0080
752#define UTF8_GOT_OVERFLOW UTF8_ALLOW_OVERFLOW
949cf498 753
f180b292 754#define UTF8_DISALLOW_SURROGATE 0x0100 /* Unicode surrogates */
2b5e7bc2 755#define UTF8_GOT_SURROGATE UTF8_DISALLOW_SURROGATE
f180b292 756#define UTF8_WARN_SURROGATE 0x0200
949cf498 757
c4e96019
KW
758/* Unicode non-character code points */
759#define UTF8_DISALLOW_NONCHAR 0x0400
2b5e7bc2 760#define UTF8_GOT_NONCHAR UTF8_DISALLOW_NONCHAR
c4e96019 761#define UTF8_WARN_NONCHAR 0x0800
949cf498 762
c4e96019
KW
763/* Super-set of Unicode: code points above the legal max */
764#define UTF8_DISALLOW_SUPER 0x1000
2b5e7bc2 765#define UTF8_GOT_SUPER UTF8_DISALLOW_SUPER
c4e96019
KW
766#define UTF8_WARN_SUPER 0x2000
767
768/* The original UTF-8 standard did not define UTF-8 with start bytes of 0xFE or
769 * 0xFF, though UTF-EBCDIC did. This allowed both versions to represent code
770 * points up to 2 ** 31 - 1. Perl extends UTF-8 so that 0xFE and 0xFF are
771 * usable on ASCII platforms, and 0xFF means something different than
772 * UTF-EBCDIC defines. These changes allow code points of 64 bits (actually
773 * somewhat more) to be represented on both platforms. But these are Perl
774 * extensions, and not likely to be interchangeable with other languages. Note
775 * that on ASCII platforms, FE overflows a signed 32-bit word, and FF an
776 * unsigned one. */
d044b7a7
KW
777#define UTF8_DISALLOW_PERL_EXTENDED 0x4000
778#define UTF8_GOT_PERL_EXTENDED UTF8_DISALLOW_PERL_EXTENDED
779#define UTF8_WARN_PERL_EXTENDED 0x8000
d35f2ca5 780
57ff5f59
KW
781/* For back compat, these old names are misleading for overlongs and
782 * UTF_EBCDIC. */
d044b7a7
KW
783#define UTF8_DISALLOW_ABOVE_31_BIT UTF8_DISALLOW_PERL_EXTENDED
784#define UTF8_GOT_ABOVE_31_BIT UTF8_GOT_PERL_EXTENDED
785#define UTF8_WARN_ABOVE_31_BIT UTF8_WARN_PERL_EXTENDED
786#define UTF8_DISALLOW_FE_FF UTF8_DISALLOW_PERL_EXTENDED
787#define UTF8_WARN_FE_FF UTF8_WARN_PERL_EXTENDED
949cf498 788
f180b292 789#define UTF8_CHECK_ONLY 0x10000
99a765e9 790#define _UTF8_NO_CONFIDENCE_IN_CURLEN 0x20000 /* Internal core use only */
949cf498
KW
791
792/* For backwards source compatibility. They do nothing, as the default now
793 * includes what they used to mean. The first one's meaning was to allow the
794 * just the single non-character 0xFFFF */
795#define UTF8_ALLOW_FFFF 0
c825ef8c 796#define UTF8_ALLOW_FE_FF 0
949cf498
KW
797#define UTF8_ALLOW_SURROGATE 0
798
ecc1615f
KW
799/* C9 refers to Unicode Corrigendum #9: allows but discourages non-chars */
800#define UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE \
801 (UTF8_DISALLOW_SUPER|UTF8_DISALLOW_SURROGATE)
802#define UTF8_WARN_ILLEGAL_C9_INTERCHANGE (UTF8_WARN_SUPER|UTF8_WARN_SURROGATE)
803
d35f2ca5 804#define UTF8_DISALLOW_ILLEGAL_INTERCHANGE \
ecc1615f 805 (UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE|UTF8_DISALLOW_NONCHAR)
949cf498 806#define UTF8_WARN_ILLEGAL_INTERCHANGE \
ecc1615f
KW
807 (UTF8_WARN_ILLEGAL_C9_INTERCHANGE|UTF8_WARN_NONCHAR)
808
0eb3d6a0
KW
809/* This is typically used for code that processes UTF-8 input and doesn't want
810 * to have to deal with any malformations that might be present. All such will
811 * be safely replaced by the REPLACEMENT CHARACTER, unless other flags
812 * overriding this are also present. */
2d532c27
KW
813#define UTF8_ALLOW_ANY ( UTF8_ALLOW_CONTINUATION \
814 |UTF8_ALLOW_NON_CONTINUATION \
815 |UTF8_ALLOW_SHORT \
d60baaa7
KW
816 |UTF8_ALLOW_LONG \
817 |UTF8_ALLOW_OVERFLOW)
2d532c27
KW
818
819/* Accept any Perl-extended UTF-8 that evaluates to any UV on the platform, but
cd01d3b1 820 * not any malformed. This is the default. */
2d532c27
KW
821#define UTF8_ALLOW_ANYUV 0
822#define UTF8_ALLOW_DEFAULT UTF8_ALLOW_ANYUV
1d72bdf6 823
89d986df
KW
824/*
825=for apidoc Am|bool|UTF8_IS_SURROGATE|const U8 *s|const U8 *e
826
827Evaluates to non-zero if the first few bytes of the string starting at C<s> and
828looking no further than S<C<e - 1>> are well-formed UTF-8 that represents one
829of the Unicode surrogate code points; otherwise it evaluates to 0. If
830non-zero, the value gives how many bytes starting at C<s> comprise the code
831point's representation.
832
833=cut
834 */
835#define UTF8_IS_SURROGATE(s, e) is_SURROGATE_utf8_safe(s, e)
836
837
838#define UTF8_IS_REPLACEMENT(s, send) is_REPLACEMENT_utf8_safe(s,send)
839
285aa1f0 840#define MAX_LEGAL_CP ((UV)IV_MAX)
40606899 841
89d986df
KW
842/*
843=for apidoc Am|bool|UTF8_IS_SUPER|const U8 *s|const U8 *e
844
845Recall that Perl recognizes an extension to UTF-8 that can encode code
846points larger than the ones defined by Unicode, which are 0..0x10FFFF.
847
848This macro evaluates to non-zero if the first few bytes of the string starting
849at C<s> and looking no further than S<C<e - 1>> are from this UTF-8 extension;
850otherwise it evaluates to 0. If non-zero, the value gives how many bytes
851starting at C<s> comprise the code point's representation.
852
8530 is returned if the bytes are not well-formed extended UTF-8, or if they
854represent a code point that cannot fit in a UV on the current platform. Hence
855this macro can give different results when run on a 64-bit word machine than on
856one with a 32-bit word size.
0c58a72b 857
891fd405 858Note that it is illegal to have code points that are larger than what can
89d986df 859fit in an IV on the current machine.
7131f24d 860
89d986df
KW
861=cut
862
863 * ASCII EBCDIC I8
7131f24d
KW
864 * U+10FFFF: \xF4\x8F\xBF\xBF \xF9\xA1\xBF\xBF\xBF max legal Unicode
865 * U+110000: \xF4\x90\x80\x80 \xF9\xA2\xA0\xA0\xA0
866 * U+110001: \xF4\x90\x80\x81 \xF9\xA2\xA0\xA0\xA1
89d986df
KW
867 */
868#ifdef EBCDIC
a14e0a36 869# define UTF8_IS_SUPER(s, e) \
89d986df
KW
870 (( LIKELY((e) > (s) + 4) \
871 && NATIVE_UTF8_TO_I8(*(s)) >= 0xF9 \
872 && ( NATIVE_UTF8_TO_I8(*(s)) > 0xF9 \
3d42f267 873 || (NATIVE_UTF8_TO_I8(*((s) + 1)) >= 0xA2)) \
89d986df 874 && LIKELY((s) + UTF8SKIP(s) <= (e))) \
1376b35c 875 ? is_utf8_char_helper(s, s + UTF8SKIP(s), 0) : 0)
7131f24d 876#else
a14e0a36 877# define UTF8_IS_SUPER(s, e) \
89d986df
KW
878 (( LIKELY((e) > (s) + 3) \
879 && (*(U8*) (s)) >= 0xF4 \
880 && ((*(U8*) (s)) > 0xF4 || (*((U8*) (s) + 1) >= 0x90))\
881 && LIKELY((s) + UTF8SKIP(s) <= (e))) \
1376b35c 882 ? is_utf8_char_helper(s, s + UTF8SKIP(s), 0) : 0)
7131f24d
KW
883#endif
884
b96a92fb
KW
885/* These are now machine generated, and the 'given' clause is no longer
886 * applicable */
0c58a72b 887#define UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s, e) \
89d986df
KW
888 cBOOL(is_NONCHAR_utf8_safe(s,e))
889
890/*
891=for apidoc Am|bool|UTF8_IS_NONCHAR|const U8 *s|const U8 *e
892
893Evaluates to non-zero if the first few bytes of the string starting at C<s> and
894looking no further than S<C<e - 1>> are well-formed UTF-8 that represents one
895of the Unicode non-character code points; otherwise it evaluates to 0. If
896non-zero, the value gives how many bytes starting at C<s> comprise the code
897point's representation.
898
d296fe14
KW
899=for apidoc AmnU|UV|UNICODE_REPLACEMENT
900
901Evaluates to 0xFFFD, the code point of the Unicode REPLACEMENT CHARACTER
902
89d986df
KW
903=cut
904 */
0c58a72b
KW
905#define UTF8_IS_NONCHAR(s, e) \
906 UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s, e)
7131f24d 907
c867b360
JH
908#define UNICODE_SURROGATE_FIRST 0xD800
909#define UNICODE_SURROGATE_LAST 0xDFFF
910#define UNICODE_REPLACEMENT 0xFFFD
911#define UNICODE_BYTE_ORDER_MARK 0xFEFF
1d72bdf6 912
b851fbc1 913/* Though our UTF-8 encoding can go beyond this,
c76687c5 914 * let's be conservative and do as Unicode says. */
b851fbc1
JH
915#define PERL_UNICODE_MAX 0x10FFFF
916
d044b7a7
KW
917#define UNICODE_WARN_SURROGATE 0x0001 /* UTF-16 surrogates */
918#define UNICODE_WARN_NONCHAR 0x0002 /* Non-char code points */
919#define UNICODE_WARN_SUPER 0x0004 /* Above 0x10FFFF */
920#define UNICODE_WARN_PERL_EXTENDED 0x0008 /* Above 0x7FFF_FFFF */
921#define UNICODE_WARN_ABOVE_31_BIT UNICODE_WARN_PERL_EXTENDED
922#define UNICODE_DISALLOW_SURROGATE 0x0010
923#define UNICODE_DISALLOW_NONCHAR 0x0020
924#define UNICODE_DISALLOW_SUPER 0x0040
925#define UNICODE_DISALLOW_PERL_EXTENDED 0x0080
24b4c303
KW
926
927#ifdef PERL_CORE
928# define UNICODE_ALLOW_ABOVE_IV_MAX 0x0100
929#endif
d044b7a7 930#define UNICODE_DISALLOW_ABOVE_31_BIT UNICODE_DISALLOW_PERL_EXTENDED
33f38593
KW
931
932#define UNICODE_GOT_SURROGATE UNICODE_DISALLOW_SURROGATE
933#define UNICODE_GOT_NONCHAR UNICODE_DISALLOW_NONCHAR
934#define UNICODE_GOT_SUPER UNICODE_DISALLOW_SUPER
935#define UNICODE_GOT_PERL_EXTENDED UNICODE_DISALLOW_PERL_EXTENDED
936
ecc1615f
KW
937#define UNICODE_WARN_ILLEGAL_C9_INTERCHANGE \
938 (UNICODE_WARN_SURROGATE|UNICODE_WARN_SUPER)
bb88be5f 939#define UNICODE_WARN_ILLEGAL_INTERCHANGE \
ecc1615f
KW
940 (UNICODE_WARN_ILLEGAL_C9_INTERCHANGE|UNICODE_WARN_NONCHAR)
941#define UNICODE_DISALLOW_ILLEGAL_C9_INTERCHANGE \
942 (UNICODE_DISALLOW_SURROGATE|UNICODE_DISALLOW_SUPER)
bb88be5f 943#define UNICODE_DISALLOW_ILLEGAL_INTERCHANGE \
ecc1615f 944 (UNICODE_DISALLOW_ILLEGAL_C9_INTERCHANGE|UNICODE_DISALLOW_NONCHAR)
949cf498
KW
945
946/* For backward source compatibility, as are now the default */
947#define UNICODE_ALLOW_SURROGATE 0
948#define UNICODE_ALLOW_SUPER 0
949#define UNICODE_ALLOW_ANY 0
b851fbc1 950
2d6b3d38
KW
951/* This matches the 2048 code points between UNICODE_SURROGATE_FIRST (0xD800) and
952 * UNICODE_SURROGATE_LAST (0xDFFF) */
953#define UNICODE_IS_SURROGATE(uv) (((UV) (uv) & (~0xFFFF | 0xF800)) \
954 == 0xD800)
955
646d1759
KW
956#define UNICODE_IS_REPLACEMENT(uv) ((UV) (uv) == UNICODE_REPLACEMENT)
957#define UNICODE_IS_BYTE_ORDER_MARK(uv) ((UV) (uv) == UNICODE_BYTE_ORDER_MARK)
c149ab20
KW
958
959/* Is 'uv' one of the 32 contiguous-range noncharacters? */
960#define UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv) ((UV) (uv) >= 0xFDD0 \
961 && (UV) (uv) <= 0xFDEF)
962
963/* Is 'uv' one of the 34 plane-ending noncharacters 0xFFFE, 0xFFFF, 0x1FFFE,
964 * 0x1FFFF, ... 0x10FFFE, 0x10FFFF, given that we know that 'uv' is not above
965 * the Unicode legal max */
966#define UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv) \
967 (((UV) (uv) & 0xFFFE) == 0xFFFE)
968
969#define UNICODE_IS_NONCHAR(uv) \
970 ( UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv) \
971 || ( LIKELY( ! UNICODE_IS_SUPER(uv)) \
972 && UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv)))
973
974#define UNICODE_IS_SUPER(uv) ((UV) (uv) > PERL_UNICODE_MAX)
1d72bdf6 975
ec34087a
KW
976#define LATIN_SMALL_LETTER_SHARP_S LATIN_SMALL_LETTER_SHARP_S_NATIVE
977#define LATIN_SMALL_LETTER_Y_WITH_DIAERESIS \
978 LATIN_SMALL_LETTER_Y_WITH_DIAERESIS_NATIVE
979#define MICRO_SIGN MICRO_SIGN_NATIVE
980#define LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE \
981 LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE_NATIVE
982#define LATIN_SMALL_LETTER_A_WITH_RING_ABOVE \
983 LATIN_SMALL_LETTER_A_WITH_RING_ABOVE_NATIVE
09091399
JH
984#define UNICODE_GREEK_CAPITAL_LETTER_SIGMA 0x03A3
985#define UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA 0x03C2
986#define UNICODE_GREEK_SMALL_LETTER_SIGMA 0x03C3
9dcbe121 987#define GREEK_SMALL_LETTER_MU 0x03BC
9e682c18
KW
988#define GREEK_CAPITAL_LETTER_MU 0x039C /* Upper and title case
989 of MICRON */
990#define LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS 0x0178 /* Also is title case */
0766489e
KW
991#ifdef LATIN_CAPITAL_LETTER_SHARP_S_UTF8
992# define LATIN_CAPITAL_LETTER_SHARP_S 0x1E9E
993#endif
74894415
KW
994#define LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE 0x130
995#define LATIN_SMALL_LETTER_DOTLESS_I 0x131
9e682c18 996#define LATIN_SMALL_LETTER_LONG_S 0x017F
a9f50d33
KW
997#define LATIN_SMALL_LIGATURE_LONG_S_T 0xFB05
998#define LATIN_SMALL_LIGATURE_ST 0xFB06
9e682c18
KW
999#define KELVIN_SIGN 0x212A
1000#define ANGSTROM_SIGN 0x212B
09091399 1001
9e55ce06 1002#define UNI_DISPLAY_ISPRINT 0x0001
c728cb41
JH
1003#define UNI_DISPLAY_BACKSLASH 0x0002
1004#define UNI_DISPLAY_QQ (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH)
1005#define UNI_DISPLAY_REGEX (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH)
9e55ce06 1006
5cd46e1f
KW
1007#define ANYOF_FOLD_SHARP_S(node, input, end) \
1008 (ANYOF_BITMAP_TEST(node, LATIN_SMALL_LETTER_SHARP_S) && \
137165a6 1009 (ANYOF_NONBITMAP(node)) && \
39065660 1010 (ANYOF_FLAGS(node) & ANYOF_LOC_NONBITMAP_FOLD) && \
07b6858f 1011 ((end) > (input) + 1) && \
305b8651 1012 isALPHA_FOLD_EQ((input)[0], 's'))
6302f837 1013
ebc501f0 1014#define SHARP_S_SKIP 2
3b0fc154 1015
3cedd9d9 1016#define is_utf8_char_buf(buf, buf_end) isUTF8_CHAR(buf, buf_end)
976c1b08
KW
1017#define bytes_from_utf8(s, lenp, is_utf8p) \
1018 bytes_from_utf8_loc(s, lenp, is_utf8p, 0)
3cedd9d9 1019
e23e8bc1
KW
1020/*
1021
25e3a4e0
KW
1022=for apidoc Am|STRLEN|isUTF8_CHAR_flags|const U8 *s|const U8 *e| const U32 flags
1023
1024Evaluates to non-zero if the first few bytes of the string starting at C<s> and
1025looking no further than S<C<e - 1>> are well-formed UTF-8, as extended by Perl,
1026that represents some code point, subject to the restrictions given by C<flags>;
1027otherwise it evaluates to 0. If non-zero, the value gives how many bytes
2717076a
KW
1028starting at C<s> comprise the code point's representation. Any bytes remaining
1029before C<e>, but beyond the ones needed to form the first code point in C<s>,
1030are not examined.
25e3a4e0
KW
1031
1032If C<flags> is 0, this gives the same results as C<L</isUTF8_CHAR>>;
1033if C<flags> is C<UTF8_DISALLOW_ILLEGAL_INTERCHANGE>, this gives the same results
1034as C<L</isSTRICT_UTF8_CHAR>>;
1035and if C<flags> is C<UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE>, this gives
1036the same results as C<L</isC9_STRICT_UTF8_CHAR>>.
1037Otherwise C<flags> may be any combination of the C<UTF8_DISALLOW_I<foo>> flags
1038understood by C<L</utf8n_to_uvchr>>, with the same meanings.
1039
1040The three alternative macros are for the most commonly needed validations; they
1041are likely to run somewhat faster than this more general one, as they can be
1042inlined into your code.
1043
9f2abfde
KW
1044Use L</is_utf8_string_flags>, L</is_utf8_string_loc_flags>, and
1045L</is_utf8_string_loclen_flags> to check entire strings.
1046
25e3a4e0
KW
1047=cut
1048*/
1049
1050#define isUTF8_CHAR_flags(s, e, flags) \
1051 (UNLIKELY((e) <= (s)) \
1052 ? 0 \
1053 : (UTF8_IS_INVARIANT(*s)) \
1054 ? 1 \
1055 : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \
1056 ? 0 \
1376b35c 1057 : is_utf8_char_helper(s, e, flags))
25e3a4e0 1058
6302f837
KW
1059/* Do not use; should be deprecated. Use isUTF8_CHAR() instead; this is
1060 * retained solely for backwards compatibility */
1061#define IS_UTF8_CHAR(p, n) (isUTF8_CHAR(p, (p) + (n)) == n)
e9a8c099 1062
6a5bc5ac 1063#endif /* PERL_UTF8_H_ */
57f0e7e2 1064
e9a8c099 1065/*
14d04a33 1066 * ex: set ts=8 sts=4 sw=4 et:
e9a8c099 1067 */