This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Document UTF8_MAXBYTES_CASE
[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
1a3756de 147#if defined(_MSC_VER) && _MSC_VER < 1400
6f6d1bab
TC
148/* older MSVC versions have a smallish macro buffer */
149#define PERL_SMALL_MACRO_BUFFER
150#endif
151
ef2a4c8f
KW
152/*
153
154=for apidoc Am|U8|NATIVE_TO_LATIN1|U8 ch
155
156Returns the Latin-1 (including ASCII and control characters) equivalent of the
157input native code point given by C<ch>. Thus, C<NATIVE_TO_LATIN1(193)> on
158EBCDIC platforms returns 65. These each represent the character C<"A"> on
159their respective platforms. On ASCII platforms no conversion is needed, so
160this macro expands to just its input, adding no time nor space requirements to
161the implementation.
162
163For conversion of code points potentially larger than will fit in a character,
164use L</NATIVE_TO_UNI>.
165
166=for apidoc Am|U8|LATIN1_TO_NATIVE|U8 ch
167
168Returns the native equivalent of the input Latin-1 code point (including ASCII
169and control characters) given by C<ch>. Thus, C<LATIN1_TO_NATIVE(66)> on
170EBCDIC platforms returns 194. These each represent the character C<"B"> on
171their respective platforms. On ASCII platforms no conversion is needed, so
172this macro expands to just its input, adding no time nor space requirements to
173the implementation.
174
175For conversion of code points potentially larger than will fit in a character,
176use L</UNI_TO_NATIVE>.
177
178=for apidoc Am|UV|NATIVE_TO_UNI|UV ch
179
180Returns the Unicode equivalent of the input native code point given by C<ch>.
181Thus, C<NATIVE_TO_UNI(195)> on EBCDIC platforms returns 67. These each
182represent the character C<"C"> on their respective platforms. On ASCII
183platforms no conversion is needed, so this macro expands to just its input,
184adding no time nor space requirements to the implementation.
185
186=for apidoc Am|UV|UNI_TO_NATIVE|UV ch
187
188Returns the native equivalent of the input Unicode code point given by C<ch>.
189Thus, C<UNI_TO_NATIVE(68)> on EBCDIC platforms returns 196. These each
190represent the character C<"D"> on their respective platforms. On ASCII
191platforms no conversion is needed, so this macro expands to just its input,
192adding no time nor space requirements to the implementation.
193
194=cut
195*/
196
6f6d1bab 197#ifdef PERL_SMALL_MACRO_BUFFER
cb15eeb2
KW
198# define NATIVE_TO_LATIN1(ch) ((U8)(ch))
199# define LATIN1_TO_NATIVE(ch) ((U8)(ch))
6f6d1bab 200#else
cb15eeb2
KW
201# define NATIVE_TO_LATIN1(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) ((ch) | 0)))
202# define LATIN1_TO_NATIVE(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) ((ch) | 0)))
6f6d1bab 203#endif
59a449d5
KW
204
205/* I8 is an intermediate version of UTF-8 used only in UTF-EBCDIC. We thus
206 * consider it to be identical to UTF-8 on ASCII platforms. Strictly speaking
207 * UTF-8 and UTF-EBCDIC are two different things, but we often conflate them
208 * because they are 8-bit encodings that serve the same purpose in Perl, and
209 * rarely do we need to distinguish them. The term "NATIVE_UTF8" applies to
210 * whichever one is applicable on the current platform */
6f6d1bab 211#ifdef PERL_SMALL_MACRO_BUFFER
1d4ea287
DD
212#define NATIVE_UTF8_TO_I8(ch) ((U8) (ch))
213#define I8_TO_NATIVE_UTF8(ch) ((U8) (ch))
6f6d1bab 214#else
cb15eeb2
KW
215#define NATIVE_UTF8_TO_I8(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) ((ch) | 0)))
216#define I8_TO_NATIVE_UTF8(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) ((ch) | 0)))
6f6d1bab 217#endif
59a449d5 218
cb15eeb2
KW
219#define UNI_TO_NATIVE(ch) ((UV) ((ch) | 0))
220#define NATIVE_TO_UNI(ch) ((UV) ((ch) | 0))
d7578b48 221
877d9f0d 222/*
9041c2e3 223
a14e0a36
KW
224 The following table is from Unicode 3.2, plus the Perl extensions for above
225 U+10FFFF
877d9f0d 226
a14e0a36 227 Code Points 1st Byte 2nd Byte 3rd 4th 5th 6th 7th 8th-13th
877d9f0d 228
375122d7 229 U+0000..U+007F 00..7F
e1b711da 230 U+0080..U+07FF * C2..DF 80..BF
a14e0a36
KW
231 U+0800..U+0FFF E0 * A0..BF 80..BF
232 U+1000..U+CFFF E1..EC 80..BF 80..BF
233 U+D000..U+D7FF ED 80..9F 80..BF
234 U+D800..U+DFFF ED A0..BF 80..BF (surrogates)
235 U+E000..U+FFFF EE..EF 80..BF 80..BF
236 U+10000..U+3FFFF F0 * 90..BF 80..BF 80..BF
237 U+40000..U+FFFFF F1..F3 80..BF 80..BF 80..BF
238 U+100000..U+10FFFF F4 80..8F 80..BF 80..BF
239 Below are above-Unicode code points
240 U+110000..U+13FFFF F4 90..BF 80..BF 80..BF
241 U+110000..U+1FFFFF F5..F7 80..BF 80..BF 80..BF
242 U+200000..U+FFFFFF F8 * 88..BF 80..BF 80..BF 80..BF
243U+1000000..U+3FFFFFF F9..FB 80..BF 80..BF 80..BF 80..BF
244U+4000000..U+3FFFFFFF FC * 84..BF 80..BF 80..BF 80..BF 80..BF
245U+40000000..U+7FFFFFFF FD 80..BF 80..BF 80..BF 80..BF 80..BF
246U+80000000..U+FFFFFFFFF FE * 82..BF 80..BF 80..BF 80..BF 80..BF 80..BF
247U+1000000000.. FF 80..BF 80..BF 80..BF 80..BF 80..BF * 81..BF 80..BF
877d9f0d 248
e1b711da 249Note the gaps before several of the byte entries above marked by '*'. These are
37e2e78e
KW
250caused by legal UTF-8 avoiding non-shortest encodings: it is technically
251possible to UTF-8-encode a single code point in different ways, but that is
252explicitly forbidden, and the shortest possible encoding should always be used
15824458 253(and that is what Perl does). The non-shortest ones are called 'overlongs'.
8c007b5a 254
877d9f0d
JH
255 */
256
8c007b5a
JH
257/*
258 Another way to look at it, as bits:
259
b2635aa8 260 Code Points 1st Byte 2nd Byte 3rd Byte 4th Byte
8c007b5a 261
b2635aa8
KW
262 0aaa aaaa 0aaa aaaa
263 0000 0bbb bbaa aaaa 110b bbbb 10aa aaaa
264 cccc bbbb bbaa aaaa 1110 cccc 10bb bbbb 10aa aaaa
265 00 000d ddcc cccc bbbb bbaa aaaa 1111 0ddd 10cc cccc 10bb bbbb 10aa aaaa
8c007b5a
JH
266
267As you can see, the continuation bytes all begin with C<10>, and the
e1b711da 268leading bits of the start byte tell how many bytes there are in the
8c007b5a
JH
269encoded character.
270
df863e43
KW
271Perl's extended UTF-8 means we can have start bytes up through FF, though any
272beginning with FF yields a code point that is too large for 32-bit ASCII
273platforms. FF signals to use 13 bytes for the encoded character. This breaks
274the paradigm that the number of leading bits gives how many total bytes there
ab2e28c2 275are in the character. */
38953e5a 276
15824458
KW
277/* This is the number of low-order bits a continuation byte in a UTF-8 encoded
278 * sequence contributes to the specification of the code point. In the bit
279 * maps above, you see that the first 2 bits are a constant '10', leaving 6 of
280 * real information */
1d72bdf6 281#define UTF_ACCUMULATION_SHIFT 6
b2635aa8 282
fed423a5
KW
283/* ^? is defined to be DEL on ASCII systems. See the definition of toCTRL()
284 * for more */
285#define QUESTION_MARK_CTRL DEL_NATIVE
286
287/* Surrogates, non-character code points and above-Unicode code points are
288 * problematic in some contexts. This allows code that needs to check for
289 * those to to quickly exclude the vast majority of code points it will
290 * encounter */
a6951642
KW
291#define isUTF8_POSSIBLY_PROBLEMATIC(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
292 (U8) c >= 0xED)
fed423a5 293
57ff5f59
KW
294#define UNICODE_IS_PERL_EXTENDED(uv) UNLIKELY((UV) (uv) > 0x7FFFFFFF)
295
fed423a5
KW
296#endif /* EBCDIC vs ASCII */
297
d7bcd45a
KW
298/* 2**UTF_ACCUMULATION_SHIFT - 1. This masks out all but the bits that carry
299 * real information in a continuation byte. This turns out to be 0x3F in
300 * UTF-8, 0x1F in UTF-EBCDIC. */
fed423a5
KW
301#define UTF_CONTINUATION_MASK ((U8) ((1U << UTF_ACCUMULATION_SHIFT) - 1))
302
9f3cfb7a
KW
303/* For use in UTF8_IS_CONTINUATION(). This turns out to be 0xC0 in UTF-8,
304 * E0 in UTF-EBCDIC */
305#define UTF_IS_CONTINUATION_MASK ((U8) (0xFF << UTF_ACCUMULATION_SHIFT))
306
38f458ff
KW
307/* This defines the bits that are to be in the continuation bytes of a
308 * multi-byte UTF-8 encoded character that mark it is a continuation byte.
309 * This turns out to be 0x80 in UTF-8, 0xA0 in UTF-EBCDIC. (khw doesn't know
310 * the underlying reason that B0 works here) */
311#define UTF_CONTINUATION_MARK (UTF_IS_CONTINUATION_MASK & 0xB0)
312
f4225fa0
KW
313/* Is the byte 'c' part of a multi-byte UTF8-8 encoded sequence, and not the
314 * first byte thereof? */
315#define UTF8_IS_CONTINUATION(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
316 (((NATIVE_UTF8_TO_I8(c) & UTF_IS_CONTINUATION_MASK) \
317 == UTF_CONTINUATION_MARK)))
318
2dc97505
KW
319/* Is the representation of the Unicode code point 'cp' the same regardless of
320 * being encoded in UTF-8 or not? This is a fundamental property of
321 * UTF-8,EBCDIC */
322#define OFFUNI_IS_INVARIANT(c) (((WIDEST_UTYPE)(c)) < UTF_CONTINUATION_MARK)
323
ab2e28c2
KW
324/*
325=for apidoc Am|bool|UVCHR_IS_INVARIANT|UV cp
326
327Evaluates to 1 if the representation of code point C<cp> is the same whether or
328not it is encoded in UTF-8; otherwise evaluates to 0. UTF-8 invariant
329characters can be copied as-is when converting to/from UTF-8, saving time.
330C<cp> is Unicode if above 255; otherwise is platform-native.
331
332=cut
333 */
334#define UVCHR_IS_INVARIANT(cp) (OFFUNI_IS_INVARIANT(NATIVE_TO_UNI(cp)))
335
7028aeba
KW
336/* Internal macro to be used only in this file to aid in constructing other
337 * publicly accessible macros.
338 * The number of bytes required to express this uv in UTF-8, for just those
339 * uv's requiring 2 through 6 bytes, as these are common to all platforms and
340 * word sizes. The number of bytes needed is given by the number of leading 1
341 * bits in the start byte. There are 32 start bytes that have 2 initial 1 bits
342 * (C0-DF); there are 16 that have 3 initial 1 bits (E0-EF); 8 that have 4
343 * initial 1 bits (F0-F8); 4 that have 5 initial 1 bits (F9-FB), and 2 that
344 * have 6 initial 1 bits (FC-FD). The largest number a string of n bytes can
345 * represent is (the number of possible start bytes for 'n')
346 * * (the number of possiblities for each start byte
347 * The latter in turn is
348 * 2 ** ( (how many continuation bytes there are)
349 * * (the number of bits of information each
350 * continuation byte holds))
351 *
352 * If we were on a platform where we could use a fast find first set bit
353 * instruction (or count leading zeros instruction) this could be replaced by
354 * using that to find the log2 of the uv, and divide that by the number of bits
355 * of information in each continuation byte, adjusting for large cases and how
356 * much information is in a start byte for that length */
72164d3a 357#define __COMMON_UNI_SKIP(uv) \
7028aeba
KW
358 (UV) (uv) < (32 * (1U << ( UTF_ACCUMULATION_SHIFT))) ? 2 : \
359 (UV) (uv) < (16 * (1U << (2 * UTF_ACCUMULATION_SHIFT))) ? 3 : \
360 (UV) (uv) < ( 8 * (1U << (3 * UTF_ACCUMULATION_SHIFT))) ? 4 : \
361 (UV) (uv) < ( 4 * (1U << (4 * UTF_ACCUMULATION_SHIFT))) ? 5 : \
362 (UV) (uv) < ( 2 * (1U << (5 * UTF_ACCUMULATION_SHIFT))) ? 6 :
72164d3a
KW
363
364/* Internal macro to be used only in this file.
365 * This adds to __COMMON_UNI_SKIP the details at this platform's upper range.
fed423a5 366 * For any-sized EBCDIC platforms, or 64-bit ASCII ones, we need one more test
72164d3a
KW
367 * to see if just 7 bytes is needed, or if the maximum is needed. For 32-bit
368 * ASCII platforms, everything is representable by 7 bytes */
fed423a5 369#if defined(UV_IS_QUAD) || defined(EBCDIC)
72164d3a 370# define __BASE_UNI_SKIP(uv) (__COMMON_UNI_SKIP(uv) \
7028aeba 371 (UV) (uv) < ((UV) 1U << (6 * UTF_ACCUMULATION_SHIFT)) ? 7 : UTF8_MAXBYTES)
1d68d6cd 372#else
72164d3a 373# define __BASE_UNI_SKIP(uv) (__COMMON_UNI_SKIP(uv) 7)
1d68d6cd
SC
374#endif
375
fed423a5
KW
376/* The next two macros use the base macro defined above, and add in the tests
377 * at the low-end of the range, for just 1 byte, yielding complete macros,
378 * publicly accessible. */
379
380/* Input is a true Unicode (not-native) code point */
381#define OFFUNISKIP(uv) (OFFUNI_IS_INVARIANT(uv) ? 1 : __BASE_UNI_SKIP(uv))
2084b489 382
5352a763
KW
383/*
384
385=for apidoc Am|STRLEN|UVCHR_SKIP|UV cp
386returns the number of bytes required to represent the code point C<cp> when
387encoded as UTF-8. C<cp> is a native (ASCII or EBCDIC) code point if less than
388255; a Unicode code point otherwise.
389
390=cut
391 */
fdb6583d 392#define UVCHR_SKIP(uv) ( UVCHR_IS_INVARIANT(uv) ? 1 : __BASE_UNI_SKIP(uv))
5352a763 393
4bab39bc
KW
394#define UTF_MIN_START_BYTE \
395 ((UTF_CONTINUATION_MARK >> UTF_ACCUMULATION_SHIFT) | UTF_START_MARK(2))
396
397/* Is the byte 'c' the first byte of a multi-byte UTF8-8 encoded sequence?
398 * This doesn't catch invariants (they are single-byte). It also excludes the
399 * illegal overlong sequences that begin with C0 and C1 on ASCII platforms, and
400 * C0-C4 I8 start bytes on EBCDIC ones */
401#define UTF8_IS_START(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
402 (NATIVE_UTF8_TO_I8(c) >= UTF_MIN_START_BYTE))
403
1df63428
KW
404#define UTF_MIN_ABOVE_LATIN1_BYTE \
405 ((0x100 >> UTF_ACCUMULATION_SHIFT) | UTF_START_MARK(2))
406
407/* Is the UTF8-encoded byte 'c' the first byte of a sequence of bytes that
408 * represent a code point > 255? */
409#define UTF8_IS_ABOVE_LATIN1(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
410 (NATIVE_UTF8_TO_I8(c) >= UTF_MIN_ABOVE_LATIN1_BYTE))
411
7c88d61e
KW
412/* Is the UTF8-encoded byte 'c' the first byte of a two byte sequence? Use
413 * UTF8_IS_NEXT_CHAR_DOWNGRADEABLE() instead if the input isn't known to
414 * be well-formed. */
415#define UTF8_IS_DOWNGRADEABLE_START(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
416 inRANGE(NATIVE_UTF8_TO_I8(c), \
417 UTF_MIN_START_BYTE, UTF_MIN_ABOVE_LATIN1_BYTE - 1))
418
b651802e
KW
419/* The largest code point representable by two UTF-8 bytes on this platform.
420 * As explained in the comments for __COMMON_UNI_SKIP, 32 start bytes with
fed423a5 421 * UTF_ACCUMULATION_SHIFT bits of information each */
aa206fb7
KW
422#define MAX_UTF8_TWO_BYTE (32 * (1U << UTF_ACCUMULATION_SHIFT) - 1)
423
b651802e
KW
424/* The largest code point representable by two UTF-8 bytes on any platform that
425 * Perl runs on. This value is constrained by EBCDIC which has 5 bits per
426 * continuation byte */
aa206fb7
KW
427#define MAX_PORTABLE_UTF8_TWO_BYTE (32 * (1U << 5) - 1)
428
f2c50040
KW
429/*
430
431=for apidoc AmnU|STRLEN|UTF8_MAXBYTES_CASE
432
433The maximum number of UTF-8 bytes a single Unicode character can
434uppercase/lowercase/titlecase/fold into.
435
436=cut
437
438 * Unicode guarantees that the maximum expansion is UTF8_MAX_FOLD_CHAR_EXPAND
439 * characters, but any above-Unicode code point will fold to itself, so we only
440 * have to look at the expansion of the maximum Unicode code point. But this
441 * number may be less than the space occupied by a very large code point under
442 * Perl's extended UTF-8. We have to make it large enough to fit any single
443 * character. (It turns out that ASCII and EBCDIC differ in which is larger)
444 *
445=cut
446*/
c03c0950
KW
447#define UTF8_MAXBYTES_CASE \
448 (UTF8_MAXBYTES >= (UTF8_MAX_FOLD_CHAR_EXPAND * OFFUNISKIP(0x10FFFF)) \
449 ? UTF8_MAXBYTES \
450 : (UTF8_MAX_FOLD_CHAR_EXPAND * OFFUNISKIP(0x10FFFF)))
451
d06134e5
KW
452/* Rest of these are attributes of Unicode and perl's internals rather than the
453 * encoding, or happen to be the same in both ASCII and EBCDIC (at least at
454 * this level; the macros that some of these call may have different
455 * definitions in the two encodings */
456
59a449d5
KW
457/* In domain restricted to ASCII, these may make more sense to the reader than
458 * the ones with Latin1 in the name */
459#define NATIVE_TO_ASCII(ch) NATIVE_TO_LATIN1(ch)
460#define ASCII_TO_NATIVE(ch) LATIN1_TO_NATIVE(ch)
461
462/* More or less misleadingly-named defines, retained for back compat */
463#define NATIVE_TO_UTF(ch) NATIVE_UTF8_TO_I8(ch)
464#define NATIVE_TO_I8(ch) NATIVE_UTF8_TO_I8(ch)
465#define UTF_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch)
466#define I8_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch)
467#define NATIVE8_TO_UNI(ch) NATIVE_TO_LATIN1(ch)
d06134e5 468
c0236afe 469/* This defines the 1-bits that are to be in the first byte of a multi-byte
97d0ceda
KW
470 * UTF-8 encoded character that mark it as a start byte and give the number of
471 * bytes that comprise the character. 'len' is the number of bytes in the
472 * multi-byte sequence. */
c0236afe
KW
473#define UTF_START_MARK(len) (((len) > 7) ? 0xFF : (0xFF & (0xFE << (7-(len)))))
474
475/* Masks out the initial one bits in a start byte, leaving the real data ones.
476 * Doesn't work on an invariant byte. 'len' is the number of bytes in the
477 * multi-byte sequence that comprises the character. */
478#define UTF_START_MASK(len) (((len) >= 7) ? 0x00 : (0x1F >> ((len)-2)))
479
537124e4
KW
480/* Adds a UTF8 continuation byte 'new' of information to a running total code
481 * point 'old' of all the continuation bytes so far. This is designed to be
155d2738
KW
482 * used in a loop to convert from UTF-8 to the code point represented. Note
483 * that this is asymmetric on EBCDIC platforms, in that the 'new' parameter is
484 * the UTF-EBCDIC byte, whereas the 'old' parameter is a Unicode (not EBCDIC)
485 * code point in process of being generated */
a6951642
KW
486#define UTF8_ACCUMULATE(old, new) (__ASSERT_(FITS_IN_8_BITS(new)) \
487 ((old) << UTF_ACCUMULATION_SHIFT) \
009097b1 488 | ((NATIVE_UTF8_TO_I8(new)) \
155d2738 489 & UTF_CONTINUATION_MASK))
d06134e5 490
4ab10950 491/* This works in the face of malformed UTF-8. */
4e1ed312
KW
492#define UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(s, e) \
493 ( UTF8_IS_DOWNGRADEABLE_START(*(s)) \
494 && ( (e) - (s) > 1) \
495 && UTF8_IS_CONTINUATION(*((s)+1)))
4ab10950 496
5aaebcb3 497/* Number of bytes a code point occupies in UTF-8. */
5352a763 498#define NATIVE_SKIP(uv) UVCHR_SKIP(uv)
bd18bd40 499
5aaebcb3
KW
500/* Most code which says UNISKIP is really thinking in terms of native code
501 * points (0-255) plus all those beyond. This is an imprecise term, but having
2accb712 502 * it means existing code continues to work. For precision, use UVCHR_SKIP,
5352a763
KW
503 * NATIVE_SKIP, or OFFUNISKIP */
504#define UNISKIP(uv) UVCHR_SKIP(uv)
5aaebcb3 505
3c0792e4
KW
506/* Longer, but more accurate name */
507#define UTF8_IS_ABOVE_LATIN1_START(c) UTF8_IS_ABOVE_LATIN1(c)
508
a62b247b
KW
509/* Convert a UTF-8 variant Latin1 character to a native code point value.
510 * Needs just one iteration of accumulate. Should be used only if it is known
511 * that the code point is < 256, and is not UTF-8 invariant. Use the slower
512 * but more general TWO_BYTE_UTF8_TO_NATIVE() which handles any code point
513 * representable by two bytes (which turns out to be up through
514 * MAX_PORTABLE_UTF8_TWO_BYTE). The two parameters are:
515 * HI: a downgradable start byte;
516 * LO: continuation.
517 * */
518#define EIGHT_BIT_UTF8_TO_NATIVE(HI, LO) \
519 ( __ASSERT_(UTF8_IS_DOWNGRADEABLE_START(HI)) \
520 __ASSERT_(UTF8_IS_CONTINUATION(LO)) \
521 LATIN1_TO_NATIVE(UTF8_ACCUMULATE(( \
522 NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), (LO))))
523
94bb8c36 524/* Convert a two (not one) byte utf8 character to a native code point value.
2950f2a7
KW
525 * Needs just one iteration of accumulate. Should not be used unless it is
526 * known that the two bytes are legal: 1) two-byte start, and 2) continuation.
527 * Note that the result can be larger than 255 if the input character is not
528 * downgradable */
94bb8c36 529#define TWO_BYTE_UTF8_TO_NATIVE(HI, LO) \
a6951642
KW
530 (__ASSERT_(FITS_IN_8_BITS(HI)) \
531 __ASSERT_(FITS_IN_8_BITS(LO)) \
532 __ASSERT_(PL_utf8skip[HI] == 2) \
533 __ASSERT_(UTF8_IS_CONTINUATION(LO)) \
94bb8c36 534 UNI_TO_NATIVE(UTF8_ACCUMULATE((NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), \
635e76f5 535 (LO))))
94bb8c36
KW
536
537/* Should never be used, and be deprecated */
538#define TWO_BYTE_UTF8_TO_UNI(HI, LO) NATIVE_TO_UNI(TWO_BYTE_UTF8_TO_NATIVE(HI, LO))
2950f2a7 539
bd18bd40
KW
540/*
541
542=for apidoc Am|STRLEN|UTF8SKIP|char* s
ee0ff0f5
KW
543returns the number of bytes a non-malformed UTF-8 encoded character whose first
544(perhaps only) byte is pointed to by C<s>.
545
546If there is a possibility of malformed input, use instead:
547
548=over
549
550=item L</C<UTF8_SAFE_SKIP>> if you know the maximum ending pointer in the
551buffer pointed to by C<s>; or
552
553=item L</C<UTF8_CHK_SKIP>> if you don't know it.
554
555=back
556
557It is better to restructure your code so the end pointer is passed down so that
558you know what it actually is at the point of this call, but if that isn't
559possible, L</C<UTF8_CHK_SKIP>> can minimize the chance of accessing beyond the end
560of the input buffer.
bd18bd40
KW
561
562=cut
563 */
2a70536e 564#define UTF8SKIP(s) PL_utf8skip[*(const U8*)(s)]
a281f16c
KW
565
566/*
567=for apidoc Am|STRLEN|UTF8_SKIP|char* s
568This is a synonym for L</C<UTF8SKIP>>
569
570=cut
571*/
572
2a70536e 573#define UTF8_SKIP(s) UTF8SKIP(s)
d06134e5 574
85fcc8f2 575/*
ee0ff0f5
KW
576=for apidoc Am|STRLEN|UTF8_CHK_SKIP|char* s
577
578This is a safer version of L</C<UTF8SKIP>>, but still not as safe as
579L</C<UTF8_SAFE_SKIP>>. This version doesn't blindly assume that the input
580string pointed to by C<s> is well-formed, but verifies that there isn't a NUL
581terminating character before the expected end of the next character in C<s>.
582The length C<UTF8_CHK_SKIP> returns stops just before any such NUL.
583
584Perl tends to add NULs, as an insurance policy, after the end of strings in
585SV's, so it is likely that using this macro will prevent inadvertent reading
586beyond the end of the input buffer, even if it is malformed UTF-8.
587
588This macro is intended to be used by XS modules where the inputs could be
589malformed, and it isn't feasible to restructure to use the safer
590L</C<UTF8_SAFE_SKIP>>, for example when interfacing with a C library.
591
592=cut
593*/
594
595#define UTF8_CHK_SKIP(s) \
f87d8789
KW
596 (s[0] == '\0' ? 1 : MIN(UTF8SKIP(s), \
597 my_strnlen((char *) (s), UTF8SKIP(s))))
ee0ff0f5 598/*
85fcc8f2
KW
599
600=for apidoc Am|STRLEN|UTF8_SAFE_SKIP|char* s|char* e
45671da2
KW
601returns 0 if S<C<s E<gt>= e>>; otherwise returns the number of bytes in the
602UTF-8 encoded character whose first byte is pointed to by C<s>. But it never
603returns beyond C<e>. On DEBUGGING builds, it asserts that S<C<s E<lt>= e>>.
85fcc8f2
KW
604
605=cut
606 */
45671da2
KW
607#define UTF8_SAFE_SKIP(s, e) (__ASSERT_((e) >= (s)) \
608 ((e) - (s)) <= 0 \
609 ? 0 \
610 : MIN(((e) - (s)), UTF8_SKIP(s)))
85fcc8f2 611
2d1545e5
KW
612/* Most code that says 'UNI_' really means the native value for code points up
613 * through 255 */
614#define UNI_IS_INVARIANT(cp) UVCHR_IS_INVARIANT(cp)
615
c2b32798
KW
616/*
617=for apidoc Am|bool|UTF8_IS_INVARIANT|char c
618
619Evaluates to 1 if the byte C<c> represents the same character when encoded in
620UTF-8 as when not; otherwise evaluates to 0. UTF-8 invariant characters can be
621copied as-is when converting to/from UTF-8, saving time.
622
623In spite of the name, this macro gives the correct result if the input string
624from which C<c> comes is not encoded in UTF-8.
625
626See C<L</UVCHR_IS_INVARIANT>> for checking if a UV is invariant.
627
628=cut
629
630The reason it works on both UTF-8 encoded strings and non-UTF-8 encoded, is
631that it returns TRUE in each for the exact same set of bit patterns. It is
632valid on a subset of what UVCHR_IS_INVARIANT is valid on, so can just use that;
633and the compiler should optimize out anything extraneous given the
634implementation of the latter. The |0 makes sure this isn't mistakenly called
635with a ptr argument.
636*/
5c06326b 637#define UTF8_IS_INVARIANT(c) UVCHR_IS_INVARIANT((c) | 0)
5fc230f1
KW
638
639/* Like the above, but its name implies a non-UTF8 input, which as the comments
640 * above show, doesn't matter as to its implementation */
38953e5a 641#define NATIVE_BYTE_IS_INVARIANT(c) UVCHR_IS_INVARIANT(c)
d06134e5 642
2c03e801
KW
643/* Misleadingly named: is the UTF8-encoded byte 'c' part of a variant sequence
644 * in UTF-8? This is the inverse of UTF8_IS_INVARIANT. */
645#define UTF8_IS_CONTINUED(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
646 (! UTF8_IS_INVARIANT(c)))
647
48ccf5e1
KW
648/* The macros in the next 4 sets are used to generate the two utf8 or utfebcdic
649 * bytes from an ordinal that is known to fit into exactly two (not one) bytes;
650 * it must be less than 0x3FF to work across both encodings. */
651
652/* These two are helper macros for the other three sets, and should not be used
653 * directly anywhere else. 'translate_function' is either NATIVE_TO_LATIN1
1ff3baa2
KW
654 * (which works for code points up through 0xFF) or NATIVE_TO_UNI which works
655 * for any code point */
48ccf5e1 656#define __BASE_TWO_BYTE_HI(c, translate_function) \
2863dafa 657 (__ASSERT_(! UVCHR_IS_INVARIANT(c)) \
48ccf5e1 658 I8_TO_NATIVE_UTF8((translate_function(c) >> UTF_ACCUMULATION_SHIFT) \
2863dafa 659 | UTF_START_MARK(2)))
48ccf5e1 660#define __BASE_TWO_BYTE_LO(c, translate_function) \
2863dafa 661 (__ASSERT_(! UVCHR_IS_INVARIANT(c)) \
48ccf5e1 662 I8_TO_NATIVE_UTF8((translate_function(c) & UTF_CONTINUATION_MASK) \
2863dafa 663 | UTF_CONTINUATION_MARK))
48ccf5e1 664
48ccf5e1
KW
665/* The next two macros should not be used. They were designed to be usable as
666 * the case label of a switch statement, but this doesn't work for EBCDIC. Use
9d0d3a03 667 * regen/unicode_constants.pl instead */
48ccf5e1
KW
668#define UTF8_TWO_BYTE_HI_nocast(c) __BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI)
669#define UTF8_TWO_BYTE_LO_nocast(c) __BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI)
670
671/* The next two macros are used when the source should be a single byte
672 * character; checked for under DEBUGGING */
673#define UTF8_EIGHT_BIT_HI(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
4c8cd605 674 ( __BASE_TWO_BYTE_HI(c, NATIVE_TO_LATIN1)))
48ccf5e1 675#define UTF8_EIGHT_BIT_LO(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
4c8cd605 676 (__BASE_TWO_BYTE_LO(c, NATIVE_TO_LATIN1)))
48ccf5e1
KW
677
678/* These final two macros in the series are used when the source can be any
679 * code point whose UTF-8 is known to occupy 2 bytes; they are less efficient
680 * than the EIGHT_BIT versions on EBCDIC platforms. We use the logical '~'
681 * operator instead of "<=" to avoid getting compiler warnings.
d52b8576 682 * MAX_UTF8_TWO_BYTE should be exactly all one bits in the lower few
48ccf5e1
KW
683 * places, so the ~ works */
684#define UTF8_TWO_BYTE_HI(c) \
685 (__ASSERT_((sizeof(c) == 1) \
d52b8576 686 || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \
4c8cd605 687 (__BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI)))
48ccf5e1
KW
688#define UTF8_TWO_BYTE_LO(c) \
689 (__ASSERT_((sizeof(c) == 1) \
d52b8576 690 || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \
4c8cd605 691 (__BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI)))
d06134e5 692
e7214ce8
KW
693/* This is illegal in any well-formed UTF-8 in both EBCDIC and ASCII
694 * as it is only in overlongs. */
695#define ILLEGAL_UTF8_BYTE I8_TO_NATIVE_UTF8(0xC1)
696
7e2040f0 697/*
e3036cf4 698 * 'UTF' is whether or not p is encoded in UTF8. The names 'foo_lazy_if' stem
20df05f4
KW
699 * from an earlier version of these macros in which they didn't call the
700 * foo_utf8() macros (i.e. were 'lazy') unless they decided that *p is the
701 * beginning of a utf8 character. Now that foo_utf8() determines that itself,
702 * no need to do it again here
7e2040f0 703 */
34aeb2e9
KW
704#define isIDFIRST_lazy_if(p,UTF) \
705 _is_utf8_FOO(_CC_IDFIRST, (const U8 *) p, "isIDFIRST_lazy_if", \
706 "isIDFIRST_lazy_if_safe", \
707 cBOOL(UTF && ! IN_BYTES), 0, __FILE__,__LINE__)
1d72bdf6 708
da8c1a98
KW
709#define isIDFIRST_lazy_if_safe(p, e, UTF) \
710 ((IN_BYTES || !UTF) \
711 ? isIDFIRST(*(p)) \
712 : isIDFIRST_utf8_safe(p, e))
713
34aeb2e9
KW
714#define isWORDCHAR_lazy_if(p,UTF) \
715 _is_utf8_FOO(_CC_IDFIRST, (const U8 *) p, "isWORDCHAR_lazy_if", \
716 "isWORDCHAR_lazy_if_safe", \
717 cBOOL(UTF && ! IN_BYTES), 0, __FILE__,__LINE__)
718
da8c1a98
KW
719#define isWORDCHAR_lazy_if_safe(p, e, UTF) \
720 ((IN_BYTES || !UTF) \
721 ? isWORDCHAR(*(p)) \
722 : isWORDCHAR_utf8_safe((U8 *) p, (U8 *) e))
723
34aeb2e9
KW
724#define isALNUM_lazy_if(p,UTF) \
725 _is_utf8_FOO(_CC_IDFIRST, (const U8 *) p, "isALNUM_lazy_if", \
726 "isWORDCHAR_lazy_if_safe", \
727 cBOOL(UTF && ! IN_BYTES), 0, __FILE__,__LINE__)
da8c1a98 728
89ebb4a3
JH
729#define UTF8_MAXLEN UTF8_MAXBYTES
730
8cb75cc8
KW
731/* A Unicode character can fold to up to 3 characters */
732#define UTF8_MAX_FOLD_CHAR_EXPAND 3
733
d3481830 734#define IN_BYTES UNLIKELY(CopHINTS_get(PL_curcop) & HINT_BYTES)
bd18bd40
KW
735
736/*
737
738=for apidoc Am|bool|DO_UTF8|SV* sv
739Returns a bool giving whether or not the PV in C<sv> is to be treated as being
740encoded in UTF-8.
741
742You should use this I<after> a call to C<SvPV()> or one of its variants, in
743case any call to string overloading updates the internal UTF-8 encoding flag.
744
745=cut
746*/
0064a8a9 747#define DO_UTF8(sv) (SvUTF8(sv) && !IN_BYTES)
1ff3baa2
KW
748
749/* Should all strings be treated as Unicode, and not just UTF-8 encoded ones?
750 * Is so within 'feature unicode_strings' or 'locale :not_characters', and not
751 * within 'use bytes'. UTF-8 locales are not tested for here, but perhaps
752 * could be */
70844984
KW
753#define IN_UNI_8_BIT \
754 (( ( (CopHINTS_get(PL_curcop) & HINT_UNI_8_BIT)) \
755 || ( CopHINTS_get(PL_curcop) & HINT_LOCALE_PARTIAL \
756 /* -1 below is for :not_characters */ \
757 && _is_in_locale_category(FALSE, -1))) \
758 && (! IN_BYTES))
b36bf33f 759
1d72bdf6 760
c76687c5 761#define UTF8_ALLOW_EMPTY 0x0001 /* Allow a zero length string */
2b5e7bc2 762#define UTF8_GOT_EMPTY UTF8_ALLOW_EMPTY
c76687c5
KW
763
764/* Allow first byte to be a continuation byte */
1d72bdf6 765#define UTF8_ALLOW_CONTINUATION 0x0002
2b5e7bc2 766#define UTF8_GOT_CONTINUATION UTF8_ALLOW_CONTINUATION
c76687c5 767
cd01d3b1 768/* Unexpected non-continuation byte */
1d72bdf6 769#define UTF8_ALLOW_NON_CONTINUATION 0x0004
2b5e7bc2 770#define UTF8_GOT_NON_CONTINUATION UTF8_ALLOW_NON_CONTINUATION
949cf498
KW
771
772/* expecting more bytes than were available in the string */
773#define UTF8_ALLOW_SHORT 0x0008
2b5e7bc2 774#define UTF8_GOT_SHORT UTF8_ALLOW_SHORT
949cf498 775
94953955
KW
776/* Overlong sequence; i.e., the code point can be specified in fewer bytes.
777 * First one will convert the overlong to the REPLACEMENT CHARACTER; second
778 * will return what the overlong evaluates to */
949cf498 779#define UTF8_ALLOW_LONG 0x0010
94953955 780#define UTF8_ALLOW_LONG_AND_ITS_VALUE (UTF8_ALLOW_LONG|0x0020)
2b5e7bc2
KW
781#define UTF8_GOT_LONG UTF8_ALLOW_LONG
782
d60baaa7
KW
783#define UTF8_ALLOW_OVERFLOW 0x0080
784#define UTF8_GOT_OVERFLOW UTF8_ALLOW_OVERFLOW
949cf498 785
f180b292 786#define UTF8_DISALLOW_SURROGATE 0x0100 /* Unicode surrogates */
2b5e7bc2 787#define UTF8_GOT_SURROGATE UTF8_DISALLOW_SURROGATE
f180b292 788#define UTF8_WARN_SURROGATE 0x0200
949cf498 789
c4e96019
KW
790/* Unicode non-character code points */
791#define UTF8_DISALLOW_NONCHAR 0x0400
2b5e7bc2 792#define UTF8_GOT_NONCHAR UTF8_DISALLOW_NONCHAR
c4e96019 793#define UTF8_WARN_NONCHAR 0x0800
949cf498 794
c4e96019
KW
795/* Super-set of Unicode: code points above the legal max */
796#define UTF8_DISALLOW_SUPER 0x1000
2b5e7bc2 797#define UTF8_GOT_SUPER UTF8_DISALLOW_SUPER
c4e96019
KW
798#define UTF8_WARN_SUPER 0x2000
799
800/* The original UTF-8 standard did not define UTF-8 with start bytes of 0xFE or
801 * 0xFF, though UTF-EBCDIC did. This allowed both versions to represent code
802 * points up to 2 ** 31 - 1. Perl extends UTF-8 so that 0xFE and 0xFF are
803 * usable on ASCII platforms, and 0xFF means something different than
804 * UTF-EBCDIC defines. These changes allow code points of 64 bits (actually
805 * somewhat more) to be represented on both platforms. But these are Perl
806 * extensions, and not likely to be interchangeable with other languages. Note
807 * that on ASCII platforms, FE overflows a signed 32-bit word, and FF an
808 * unsigned one. */
d044b7a7
KW
809#define UTF8_DISALLOW_PERL_EXTENDED 0x4000
810#define UTF8_GOT_PERL_EXTENDED UTF8_DISALLOW_PERL_EXTENDED
811#define UTF8_WARN_PERL_EXTENDED 0x8000
d35f2ca5 812
57ff5f59
KW
813/* For back compat, these old names are misleading for overlongs and
814 * UTF_EBCDIC. */
d044b7a7
KW
815#define UTF8_DISALLOW_ABOVE_31_BIT UTF8_DISALLOW_PERL_EXTENDED
816#define UTF8_GOT_ABOVE_31_BIT UTF8_GOT_PERL_EXTENDED
817#define UTF8_WARN_ABOVE_31_BIT UTF8_WARN_PERL_EXTENDED
818#define UTF8_DISALLOW_FE_FF UTF8_DISALLOW_PERL_EXTENDED
819#define UTF8_WARN_FE_FF UTF8_WARN_PERL_EXTENDED
949cf498 820
f180b292 821#define UTF8_CHECK_ONLY 0x10000
99a765e9 822#define _UTF8_NO_CONFIDENCE_IN_CURLEN 0x20000 /* Internal core use only */
949cf498
KW
823
824/* For backwards source compatibility. They do nothing, as the default now
825 * includes what they used to mean. The first one's meaning was to allow the
826 * just the single non-character 0xFFFF */
827#define UTF8_ALLOW_FFFF 0
c825ef8c 828#define UTF8_ALLOW_FE_FF 0
949cf498
KW
829#define UTF8_ALLOW_SURROGATE 0
830
ecc1615f
KW
831/* C9 refers to Unicode Corrigendum #9: allows but discourages non-chars */
832#define UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE \
833 (UTF8_DISALLOW_SUPER|UTF8_DISALLOW_SURROGATE)
834#define UTF8_WARN_ILLEGAL_C9_INTERCHANGE (UTF8_WARN_SUPER|UTF8_WARN_SURROGATE)
835
d35f2ca5 836#define UTF8_DISALLOW_ILLEGAL_INTERCHANGE \
ecc1615f 837 (UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE|UTF8_DISALLOW_NONCHAR)
949cf498 838#define UTF8_WARN_ILLEGAL_INTERCHANGE \
ecc1615f
KW
839 (UTF8_WARN_ILLEGAL_C9_INTERCHANGE|UTF8_WARN_NONCHAR)
840
0eb3d6a0
KW
841/* This is typically used for code that processes UTF-8 input and doesn't want
842 * to have to deal with any malformations that might be present. All such will
843 * be safely replaced by the REPLACEMENT CHARACTER, unless other flags
844 * overriding this are also present. */
2d532c27
KW
845#define UTF8_ALLOW_ANY ( UTF8_ALLOW_CONTINUATION \
846 |UTF8_ALLOW_NON_CONTINUATION \
847 |UTF8_ALLOW_SHORT \
d60baaa7
KW
848 |UTF8_ALLOW_LONG \
849 |UTF8_ALLOW_OVERFLOW)
2d532c27
KW
850
851/* Accept any Perl-extended UTF-8 that evaluates to any UV on the platform, but
cd01d3b1 852 * not any malformed. This is the default. */
2d532c27
KW
853#define UTF8_ALLOW_ANYUV 0
854#define UTF8_ALLOW_DEFAULT UTF8_ALLOW_ANYUV
1d72bdf6 855
89d986df
KW
856/*
857=for apidoc Am|bool|UTF8_IS_SURROGATE|const U8 *s|const U8 *e
858
859Evaluates to non-zero if the first few bytes of the string starting at C<s> and
860looking no further than S<C<e - 1>> are well-formed UTF-8 that represents one
861of the Unicode surrogate code points; otherwise it evaluates to 0. If
862non-zero, the value gives how many bytes starting at C<s> comprise the code
863point's representation.
864
865=cut
866 */
867#define UTF8_IS_SURROGATE(s, e) is_SURROGATE_utf8_safe(s, e)
868
869
870#define UTF8_IS_REPLACEMENT(s, send) is_REPLACEMENT_utf8_safe(s,send)
871
285aa1f0 872#define MAX_LEGAL_CP ((UV)IV_MAX)
40606899 873
89d986df
KW
874/*
875=for apidoc Am|bool|UTF8_IS_SUPER|const U8 *s|const U8 *e
876
877Recall that Perl recognizes an extension to UTF-8 that can encode code
878points larger than the ones defined by Unicode, which are 0..0x10FFFF.
879
880This macro evaluates to non-zero if the first few bytes of the string starting
881at C<s> and looking no further than S<C<e - 1>> are from this UTF-8 extension;
882otherwise it evaluates to 0. If non-zero, the value gives how many bytes
883starting at C<s> comprise the code point's representation.
884
8850 is returned if the bytes are not well-formed extended UTF-8, or if they
886represent a code point that cannot fit in a UV on the current platform. Hence
887this macro can give different results when run on a 64-bit word machine than on
888one with a 32-bit word size.
0c58a72b 889
891fd405 890Note that it is illegal to have code points that are larger than what can
89d986df 891fit in an IV on the current machine.
7131f24d 892
89d986df
KW
893=cut
894
895 * ASCII EBCDIC I8
7131f24d
KW
896 * U+10FFFF: \xF4\x8F\xBF\xBF \xF9\xA1\xBF\xBF\xBF max legal Unicode
897 * U+110000: \xF4\x90\x80\x80 \xF9\xA2\xA0\xA0\xA0
898 * U+110001: \xF4\x90\x80\x81 \xF9\xA2\xA0\xA0\xA1
89d986df
KW
899 */
900#ifdef EBCDIC
a14e0a36 901# define UTF8_IS_SUPER(s, e) \
89d986df
KW
902 (( LIKELY((e) > (s) + 4) \
903 && NATIVE_UTF8_TO_I8(*(s)) >= 0xF9 \
904 && ( NATIVE_UTF8_TO_I8(*(s)) > 0xF9 \
3d42f267 905 || (NATIVE_UTF8_TO_I8(*((s) + 1)) >= 0xA2)) \
89d986df 906 && LIKELY((s) + UTF8SKIP(s) <= (e))) \
1376b35c 907 ? is_utf8_char_helper(s, s + UTF8SKIP(s), 0) : 0)
7131f24d 908#else
a14e0a36 909# define UTF8_IS_SUPER(s, e) \
89d986df
KW
910 (( LIKELY((e) > (s) + 3) \
911 && (*(U8*) (s)) >= 0xF4 \
912 && ((*(U8*) (s)) > 0xF4 || (*((U8*) (s) + 1) >= 0x90))\
913 && LIKELY((s) + UTF8SKIP(s) <= (e))) \
1376b35c 914 ? is_utf8_char_helper(s, s + UTF8SKIP(s), 0) : 0)
7131f24d
KW
915#endif
916
b96a92fb
KW
917/* These are now machine generated, and the 'given' clause is no longer
918 * applicable */
0c58a72b 919#define UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s, e) \
89d986df
KW
920 cBOOL(is_NONCHAR_utf8_safe(s,e))
921
922/*
923=for apidoc Am|bool|UTF8_IS_NONCHAR|const U8 *s|const U8 *e
924
925Evaluates to non-zero if the first few bytes of the string starting at C<s> and
926looking no further than S<C<e - 1>> are well-formed UTF-8 that represents one
927of the Unicode non-character code points; otherwise it evaluates to 0. If
928non-zero, the value gives how many bytes starting at C<s> comprise the code
929point's representation.
930
d296fe14
KW
931=for apidoc AmnU|UV|UNICODE_REPLACEMENT
932
933Evaluates to 0xFFFD, the code point of the Unicode REPLACEMENT CHARACTER
934
89d986df
KW
935=cut
936 */
0c58a72b
KW
937#define UTF8_IS_NONCHAR(s, e) \
938 UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s, e)
7131f24d 939
c867b360
JH
940#define UNICODE_SURROGATE_FIRST 0xD800
941#define UNICODE_SURROGATE_LAST 0xDFFF
942#define UNICODE_REPLACEMENT 0xFFFD
943#define UNICODE_BYTE_ORDER_MARK 0xFEFF
1d72bdf6 944
b851fbc1 945/* Though our UTF-8 encoding can go beyond this,
c76687c5 946 * let's be conservative and do as Unicode says. */
b851fbc1
JH
947#define PERL_UNICODE_MAX 0x10FFFF
948
d044b7a7
KW
949#define UNICODE_WARN_SURROGATE 0x0001 /* UTF-16 surrogates */
950#define UNICODE_WARN_NONCHAR 0x0002 /* Non-char code points */
951#define UNICODE_WARN_SUPER 0x0004 /* Above 0x10FFFF */
952#define UNICODE_WARN_PERL_EXTENDED 0x0008 /* Above 0x7FFF_FFFF */
953#define UNICODE_WARN_ABOVE_31_BIT UNICODE_WARN_PERL_EXTENDED
954#define UNICODE_DISALLOW_SURROGATE 0x0010
955#define UNICODE_DISALLOW_NONCHAR 0x0020
956#define UNICODE_DISALLOW_SUPER 0x0040
957#define UNICODE_DISALLOW_PERL_EXTENDED 0x0080
958#define UNICODE_DISALLOW_ABOVE_31_BIT UNICODE_DISALLOW_PERL_EXTENDED
33f38593
KW
959
960#define UNICODE_GOT_SURROGATE UNICODE_DISALLOW_SURROGATE
961#define UNICODE_GOT_NONCHAR UNICODE_DISALLOW_NONCHAR
962#define UNICODE_GOT_SUPER UNICODE_DISALLOW_SUPER
963#define UNICODE_GOT_PERL_EXTENDED UNICODE_DISALLOW_PERL_EXTENDED
964
ecc1615f
KW
965#define UNICODE_WARN_ILLEGAL_C9_INTERCHANGE \
966 (UNICODE_WARN_SURROGATE|UNICODE_WARN_SUPER)
bb88be5f 967#define UNICODE_WARN_ILLEGAL_INTERCHANGE \
ecc1615f
KW
968 (UNICODE_WARN_ILLEGAL_C9_INTERCHANGE|UNICODE_WARN_NONCHAR)
969#define UNICODE_DISALLOW_ILLEGAL_C9_INTERCHANGE \
970 (UNICODE_DISALLOW_SURROGATE|UNICODE_DISALLOW_SUPER)
bb88be5f 971#define UNICODE_DISALLOW_ILLEGAL_INTERCHANGE \
ecc1615f 972 (UNICODE_DISALLOW_ILLEGAL_C9_INTERCHANGE|UNICODE_DISALLOW_NONCHAR)
949cf498
KW
973
974/* For backward source compatibility, as are now the default */
975#define UNICODE_ALLOW_SURROGATE 0
976#define UNICODE_ALLOW_SUPER 0
977#define UNICODE_ALLOW_ANY 0
b851fbc1 978
2d6b3d38
KW
979/* This matches the 2048 code points between UNICODE_SURROGATE_FIRST (0xD800) and
980 * UNICODE_SURROGATE_LAST (0xDFFF) */
981#define UNICODE_IS_SURROGATE(uv) (((UV) (uv) & (~0xFFFF | 0xF800)) \
982 == 0xD800)
983
646d1759
KW
984#define UNICODE_IS_REPLACEMENT(uv) ((UV) (uv) == UNICODE_REPLACEMENT)
985#define UNICODE_IS_BYTE_ORDER_MARK(uv) ((UV) (uv) == UNICODE_BYTE_ORDER_MARK)
c149ab20
KW
986
987/* Is 'uv' one of the 32 contiguous-range noncharacters? */
988#define UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv) ((UV) (uv) >= 0xFDD0 \
989 && (UV) (uv) <= 0xFDEF)
990
991/* Is 'uv' one of the 34 plane-ending noncharacters 0xFFFE, 0xFFFF, 0x1FFFE,
992 * 0x1FFFF, ... 0x10FFFE, 0x10FFFF, given that we know that 'uv' is not above
993 * the Unicode legal max */
994#define UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv) \
995 (((UV) (uv) & 0xFFFE) == 0xFFFE)
996
997#define UNICODE_IS_NONCHAR(uv) \
998 ( UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv) \
999 || ( LIKELY( ! UNICODE_IS_SUPER(uv)) \
1000 && UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv)))
1001
1002#define UNICODE_IS_SUPER(uv) ((UV) (uv) > PERL_UNICODE_MAX)
1d72bdf6 1003
ec34087a
KW
1004#define LATIN_SMALL_LETTER_SHARP_S LATIN_SMALL_LETTER_SHARP_S_NATIVE
1005#define LATIN_SMALL_LETTER_Y_WITH_DIAERESIS \
1006 LATIN_SMALL_LETTER_Y_WITH_DIAERESIS_NATIVE
1007#define MICRO_SIGN MICRO_SIGN_NATIVE
1008#define LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE \
1009 LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE_NATIVE
1010#define LATIN_SMALL_LETTER_A_WITH_RING_ABOVE \
1011 LATIN_SMALL_LETTER_A_WITH_RING_ABOVE_NATIVE
09091399
JH
1012#define UNICODE_GREEK_CAPITAL_LETTER_SIGMA 0x03A3
1013#define UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA 0x03C2
1014#define UNICODE_GREEK_SMALL_LETTER_SIGMA 0x03C3
9dcbe121 1015#define GREEK_SMALL_LETTER_MU 0x03BC
9e682c18
KW
1016#define GREEK_CAPITAL_LETTER_MU 0x039C /* Upper and title case
1017 of MICRON */
1018#define LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS 0x0178 /* Also is title case */
0766489e
KW
1019#ifdef LATIN_CAPITAL_LETTER_SHARP_S_UTF8
1020# define LATIN_CAPITAL_LETTER_SHARP_S 0x1E9E
1021#endif
74894415
KW
1022#define LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE 0x130
1023#define LATIN_SMALL_LETTER_DOTLESS_I 0x131
9e682c18 1024#define LATIN_SMALL_LETTER_LONG_S 0x017F
a9f50d33
KW
1025#define LATIN_SMALL_LIGATURE_LONG_S_T 0xFB05
1026#define LATIN_SMALL_LIGATURE_ST 0xFB06
9e682c18
KW
1027#define KELVIN_SIGN 0x212A
1028#define ANGSTROM_SIGN 0x212B
09091399 1029
9e55ce06 1030#define UNI_DISPLAY_ISPRINT 0x0001
c728cb41
JH
1031#define UNI_DISPLAY_BACKSLASH 0x0002
1032#define UNI_DISPLAY_QQ (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH)
1033#define UNI_DISPLAY_REGEX (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH)
9e55ce06 1034
5cd46e1f
KW
1035#define ANYOF_FOLD_SHARP_S(node, input, end) \
1036 (ANYOF_BITMAP_TEST(node, LATIN_SMALL_LETTER_SHARP_S) && \
137165a6 1037 (ANYOF_NONBITMAP(node)) && \
39065660 1038 (ANYOF_FLAGS(node) & ANYOF_LOC_NONBITMAP_FOLD) && \
07b6858f 1039 ((end) > (input) + 1) && \
305b8651 1040 isALPHA_FOLD_EQ((input)[0], 's'))
6302f837 1041
ebc501f0 1042#define SHARP_S_SKIP 2
3b0fc154 1043
3cedd9d9 1044#define is_utf8_char_buf(buf, buf_end) isUTF8_CHAR(buf, buf_end)
976c1b08
KW
1045#define bytes_from_utf8(s, lenp, is_utf8p) \
1046 bytes_from_utf8_loc(s, lenp, is_utf8p, 0)
3cedd9d9 1047
e23e8bc1
KW
1048/*
1049
25e3a4e0
KW
1050=for apidoc Am|STRLEN|isUTF8_CHAR_flags|const U8 *s|const U8 *e| const U32 flags
1051
1052Evaluates to non-zero if the first few bytes of the string starting at C<s> and
1053looking no further than S<C<e - 1>> are well-formed UTF-8, as extended by Perl,
1054that represents some code point, subject to the restrictions given by C<flags>;
1055otherwise it evaluates to 0. If non-zero, the value gives how many bytes
2717076a
KW
1056starting at C<s> comprise the code point's representation. Any bytes remaining
1057before C<e>, but beyond the ones needed to form the first code point in C<s>,
1058are not examined.
25e3a4e0
KW
1059
1060If C<flags> is 0, this gives the same results as C<L</isUTF8_CHAR>>;
1061if C<flags> is C<UTF8_DISALLOW_ILLEGAL_INTERCHANGE>, this gives the same results
1062as C<L</isSTRICT_UTF8_CHAR>>;
1063and if C<flags> is C<UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE>, this gives
1064the same results as C<L</isC9_STRICT_UTF8_CHAR>>.
1065Otherwise C<flags> may be any combination of the C<UTF8_DISALLOW_I<foo>> flags
1066understood by C<L</utf8n_to_uvchr>>, with the same meanings.
1067
1068The three alternative macros are for the most commonly needed validations; they
1069are likely to run somewhat faster than this more general one, as they can be
1070inlined into your code.
1071
9f2abfde
KW
1072Use L</is_utf8_string_flags>, L</is_utf8_string_loc_flags>, and
1073L</is_utf8_string_loclen_flags> to check entire strings.
1074
25e3a4e0
KW
1075=cut
1076*/
1077
1078#define isUTF8_CHAR_flags(s, e, flags) \
1079 (UNLIKELY((e) <= (s)) \
1080 ? 0 \
1081 : (UTF8_IS_INVARIANT(*s)) \
1082 ? 1 \
1083 : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \
1084 ? 0 \
1376b35c 1085 : is_utf8_char_helper(s, e, flags))
25e3a4e0 1086
6302f837
KW
1087/* Do not use; should be deprecated. Use isUTF8_CHAR() instead; this is
1088 * retained solely for backwards compatibility */
1089#define IS_UTF8_CHAR(p, n) (isUTF8_CHAR(p, (p) + (n)) == n)
e9a8c099 1090
6a5bc5ac 1091#endif /* PERL_UTF8_H_ */
57f0e7e2 1092
e9a8c099 1093/*
14d04a33 1094 * ex: set ts=8 sts=4 sw=4 et:
e9a8c099 1095 */