This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Document UTF8_MAXBYTES
[perl5.git] / utf8.h
CommitLineData
a0ed51b3
LW
1/* utf8.h
2 *
f3cb6f94
KW
3 * This file contains definitions for use with the UTF-8 encoding. It
4 * actually also works with the variant UTF-8 encoding called UTF-EBCDIC, and
5 * hides almost all of the differences between these from the caller. In other
6 * words, someone should #include this file, and if the code is being compiled
7 * on an EBCDIC platform, things should mostly just work.
8 *
2eee27d7
SS
9 * Copyright (C) 2000, 2001, 2002, 2005, 2006, 2007, 2009,
10 * 2010, 2011 by Larry Wall and others
a0ed51b3
LW
11 *
12 * You may distribute under the terms of either the GNU General Public
13 * License or the Artistic License, as specified in the README file.
14 *
15 */
16
6a5bc5ac
KW
17#ifndef PERL_UTF8_H_ /* Guard against recursive inclusion */
18#define PERL_UTF8_H_ 1
57f0e7e2 19
39e02b42 20/* Use UTF-8 as the default script encoding?
1e54db1a 21 * Turning this on will break scripts having non-UTF-8 binary
39e02b42
JH
22 * data (such as Latin-1) in string literals. */
23#ifdef USE_UTF8_SCRIPTS
24# define USE_UTF8_IN_NAMES (!IN_BYTES)
25#else
26# define USE_UTF8_IN_NAMES (PL_hints & HINT_UTF8)
27#endif
28
3cd96634
KW
29#include "regcharclass.h"
30#include "unicode_constants.h"
31
051a06d4 32/* For to_utf8_fold_flags, q.v. */
e4f4ef45
KW
33#define FOLD_FLAGS_LOCALE 0x1
34#define FOLD_FLAGS_FULL 0x2
35#define FOLD_FLAGS_NOMIX_ASCII 0x4
051a06d4 36
7bbfa158
KW
37/*
38=head1 Unicode Support
8cca77bc
KW
39L<perlguts/Unicode Support> has an introduction to this API.
40
41See also L</Character classification>,
42and L</Character case changing>.
43Various functions outside this section also work specially with Unicode.
44Search for the string "utf8" in this document.
7bbfa158
KW
45
46=for apidoc is_ascii_string
47
8871a094 48This is a misleadingly-named synonym for L</is_utf8_invariant_string>.
7bbfa158
KW
49On ASCII-ish platforms, the name isn't misleading: the ASCII-range characters
50are exactly the UTF-8 invariants. But EBCDIC machines have more invariants
8871a094
KW
51than just the ASCII characters, so C<is_utf8_invariant_string> is preferred.
52
53=for apidoc is_invariant_string
54
55This is a somewhat misleadingly-named synonym for L</is_utf8_invariant_string>.
56C<is_utf8_invariant_string> is preferred, as it indicates under what conditions
57the string is invariant.
7bbfa158
KW
58
59=cut
60*/
8871a094
KW
61#define is_ascii_string(s, len) is_utf8_invariant_string(s, len)
62#define is_invariant_string(s, len) is_utf8_invariant_string(s, len)
7bbfa158 63
33f38593
KW
64#define uvoffuni_to_utf8_flags(d,uv,flags) \
65 uvoffuni_to_utf8_flags_msgs(d, uv, flags, 0)
de69f3af
KW
66#define uvchr_to_utf8(a,b) uvchr_to_utf8_flags(a,b,0)
67#define uvchr_to_utf8_flags(d,uv,flags) \
33f38593
KW
68 uvchr_to_utf8_flags_msgs(d,uv,flags, 0)
69#define uvchr_to_utf8_flags_msgs(d,uv,flags,msgs) \
70 uvoffuni_to_utf8_flags_msgs(d,NATIVE_TO_UNI(uv),flags, msgs)
de69f3af 71#define utf8_to_uvchr_buf(s, e, lenp) \
82651abe 72 _utf8_to_uvchr_buf((const U8 *) (s), (const U8 *) e, lenp)
f9380377
KW
73#define utf8n_to_uvchr(s, len, lenp, flags) \
74 utf8n_to_uvchr_error(s, len, lenp, flags, 0)
37657a5b
KW
75#define utf8n_to_uvchr_error(s, len, lenp, flags, errors) \
76 utf8n_to_uvchr_msgs(s, len, lenp, flags, errors, 0)
de69f3af 77
a0270393 78#define to_uni_fold(c, p, lenp) _to_uni_fold_flags(c, p, lenp, FOLD_FLAGS_FULL)
a239b1e2
KW
79
80#define to_utf8_fold(s, r, lenr) \
607313a1 81 _to_utf8_fold_flags (s, NULL, r, lenr, FOLD_FLAGS_FULL, __FILE__, __LINE__)
a239b1e2 82#define to_utf8_lower(s, r, lenr) \
607313a1 83 _to_utf8_lower_flags(s, NULL, r ,lenr, 0, __FILE__, __LINE__)
a239b1e2 84#define to_utf8_upper(s, r, lenr) \
607313a1 85 _to_utf8_upper_flags(s, NULL, r, lenr, 0, __FILE__, __LINE__)
a239b1e2 86#define to_utf8_title(s, r, lenr) \
607313a1 87 _to_utf8_title_flags(s, NULL, r, lenr ,0, __FILE__, __LINE__)
36bb2ab6 88
eda9cac1
KW
89#define foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \
90 foldEQ_utf8_flags(s1, pe1, l1, u1, s2, pe2, l2, u2, 0)
baa60164 91#define FOLDEQ_UTF8_NOMIX_ASCII (1 << 0)
cea315b6 92#define FOLDEQ_LOCALE (1 << 1)
18f762c3
KW
93#define FOLDEQ_S1_ALREADY_FOLDED (1 << 2)
94#define FOLDEQ_S2_ALREADY_FOLDED (1 << 3)
d635b710
KW
95#define FOLDEQ_S1_FOLDS_SANE (1 << 4)
96#define FOLDEQ_S2_FOLDS_SANE (1 << 5)
a33c29bc 97
e6226b18
KW
98#define ibcmp_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \
99 cBOOL(! foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2))
100
1d72bdf6
NIS
101#ifdef EBCDIC
102/* The equivalent of these macros but implementing UTF-EBCDIC
103 are in the following header file:
104 */
105
106#include "utfebcdic.h"
fd7cb289 107
d06134e5 108#else /* ! EBCDIC */
73c4f7a1
GS
109START_EXTERN_C
110
f4d83e55
KW
111/*
112
113=for apidoc AmnU|STRLEN|UTF8_MAXBYTES
114
115The maximum width of a single UTF-8 encoded character, in bytes.
116
117NOTE: Strictly speaking Perl's UTF-8 should not be called UTF-8 since UTF-8
118is an encoding of Unicode, and Unicode's upper limit, 0x10FFFF, can be
119expressed with 4 bytes. However, Perl thinks of UTF-8 as a way to encode
120non-negative integers in a binary format, even those above Unicode.
121
122=cut
123 */
111e8ed9
KW
124#define UTF8_MAXBYTES 13
125
a0ed51b3 126#ifdef DOINIT
6f06b55f 127EXTCONST unsigned char PL_utf8skip[] = {
b2635aa8
KW
128/* 0x00 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
129/* 0x10 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
130/* 0x20 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
131/* 0x30 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
132/* 0x40 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
133/* 0x50 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
134/* 0x60 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
135/* 0x70 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
136/* 0x80 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
137/* 0x90 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
138/* 0xA0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
139/* 0xB0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
140/* 0xC0 */ 2,2, /* overlong */
1ff3baa2 141/* 0xC2 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0080 to U+03FF */
b2635aa8
KW
142/* 0xD0 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0400 to U+07FF */
143/* 0xE0 */ 3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3, /* U+0800 to U+FFFF */
144/* 0xF0 */ 4,4,4,4,4,4,4,4,5,5,5,5,6,6, /* above BMP to 2**31 - 1 */
6937f885
KW
145 /* Perl extended (never was official UTF-8). Up to 36 bit */
146/* 0xFE */ 7,
147 /* More extended, Up to 72 bits (64-bit + reserved) */
111e8ed9 148/* 0xFF */ UTF8_MAXBYTES
a0ed51b3
LW
149};
150#else
6f06b55f 151EXTCONST unsigned char PL_utf8skip[];
a0ed51b3
LW
152#endif
153
73c4f7a1 154END_EXTERN_C
7e2040f0 155
1a3756de 156#if defined(_MSC_VER) && _MSC_VER < 1400
6f6d1bab
TC
157/* older MSVC versions have a smallish macro buffer */
158#define PERL_SMALL_MACRO_BUFFER
159#endif
160
ef2a4c8f
KW
161/*
162
163=for apidoc Am|U8|NATIVE_TO_LATIN1|U8 ch
164
165Returns the Latin-1 (including ASCII and control characters) equivalent of the
166input native code point given by C<ch>. Thus, C<NATIVE_TO_LATIN1(193)> on
167EBCDIC platforms returns 65. These each represent the character C<"A"> on
168their respective platforms. On ASCII platforms no conversion is needed, so
169this macro expands to just its input, adding no time nor space requirements to
170the implementation.
171
172For conversion of code points potentially larger than will fit in a character,
173use L</NATIVE_TO_UNI>.
174
175=for apidoc Am|U8|LATIN1_TO_NATIVE|U8 ch
176
177Returns the native equivalent of the input Latin-1 code point (including ASCII
178and control characters) given by C<ch>. Thus, C<LATIN1_TO_NATIVE(66)> on
179EBCDIC platforms returns 194. These each represent the character C<"B"> on
180their respective platforms. On ASCII platforms no conversion is needed, so
181this macro expands to just its input, adding no time nor space requirements to
182the implementation.
183
184For conversion of code points potentially larger than will fit in a character,
185use L</UNI_TO_NATIVE>.
186
187=for apidoc Am|UV|NATIVE_TO_UNI|UV ch
188
189Returns the Unicode equivalent of the input native code point given by C<ch>.
190Thus, C<NATIVE_TO_UNI(195)> on EBCDIC platforms returns 67. These each
191represent the character C<"C"> on their respective platforms. On ASCII
192platforms no conversion is needed, so this macro expands to just its input,
193adding no time nor space requirements to the implementation.
194
195=for apidoc Am|UV|UNI_TO_NATIVE|UV ch
196
197Returns the native equivalent of the input Unicode code point given by C<ch>.
198Thus, C<UNI_TO_NATIVE(68)> on EBCDIC platforms returns 196. These each
199represent the character C<"D"> on their respective platforms. On ASCII
200platforms no conversion is needed, so this macro expands to just its input,
201adding no time nor space requirements to the implementation.
202
203=cut
204*/
205
6f6d1bab 206#ifdef PERL_SMALL_MACRO_BUFFER
1a3756de
TC
207#define NATIVE_TO_LATIN1(ch) ((U8)(ch))
208#define LATIN1_TO_NATIVE(ch) ((U8)(ch))
6f6d1bab 209#else
4c8cd605
KW
210#define NATIVE_TO_LATIN1(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch)))
211#define LATIN1_TO_NATIVE(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch)))
6f6d1bab 212#endif
59a449d5
KW
213
214/* I8 is an intermediate version of UTF-8 used only in UTF-EBCDIC. We thus
215 * consider it to be identical to UTF-8 on ASCII platforms. Strictly speaking
216 * UTF-8 and UTF-EBCDIC are two different things, but we often conflate them
217 * because they are 8-bit encodings that serve the same purpose in Perl, and
218 * rarely do we need to distinguish them. The term "NATIVE_UTF8" applies to
219 * whichever one is applicable on the current platform */
6f6d1bab 220#ifdef PERL_SMALL_MACRO_BUFFER
1d4ea287
DD
221#define NATIVE_UTF8_TO_I8(ch) ((U8) (ch))
222#define I8_TO_NATIVE_UTF8(ch) ((U8) (ch))
6f6d1bab 223#else
4c8cd605
KW
224#define NATIVE_UTF8_TO_I8(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch)))
225#define I8_TO_NATIVE_UTF8(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch)))
6f6d1bab 226#endif
59a449d5 227
4c8cd605
KW
228#define UNI_TO_NATIVE(ch) ((UV) (ch))
229#define NATIVE_TO_UNI(ch) ((UV) (ch))
d7578b48 230
877d9f0d 231/*
9041c2e3 232
a14e0a36
KW
233 The following table is from Unicode 3.2, plus the Perl extensions for above
234 U+10FFFF
877d9f0d 235
a14e0a36 236 Code Points 1st Byte 2nd Byte 3rd 4th 5th 6th 7th 8th-13th
877d9f0d 237
375122d7 238 U+0000..U+007F 00..7F
e1b711da 239 U+0080..U+07FF * C2..DF 80..BF
a14e0a36
KW
240 U+0800..U+0FFF E0 * A0..BF 80..BF
241 U+1000..U+CFFF E1..EC 80..BF 80..BF
242 U+D000..U+D7FF ED 80..9F 80..BF
243 U+D800..U+DFFF ED A0..BF 80..BF (surrogates)
244 U+E000..U+FFFF EE..EF 80..BF 80..BF
245 U+10000..U+3FFFF F0 * 90..BF 80..BF 80..BF
246 U+40000..U+FFFFF F1..F3 80..BF 80..BF 80..BF
247 U+100000..U+10FFFF F4 80..8F 80..BF 80..BF
248 Below are above-Unicode code points
249 U+110000..U+13FFFF F4 90..BF 80..BF 80..BF
250 U+110000..U+1FFFFF F5..F7 80..BF 80..BF 80..BF
251 U+200000..U+FFFFFF F8 * 88..BF 80..BF 80..BF 80..BF
252U+1000000..U+3FFFFFF F9..FB 80..BF 80..BF 80..BF 80..BF
253U+4000000..U+3FFFFFFF FC * 84..BF 80..BF 80..BF 80..BF 80..BF
254U+40000000..U+7FFFFFFF FD 80..BF 80..BF 80..BF 80..BF 80..BF
255U+80000000..U+FFFFFFFFF FE * 82..BF 80..BF 80..BF 80..BF 80..BF 80..BF
256U+1000000000.. FF 80..BF 80..BF 80..BF 80..BF 80..BF * 81..BF 80..BF
877d9f0d 257
e1b711da 258Note the gaps before several of the byte entries above marked by '*'. These are
37e2e78e
KW
259caused by legal UTF-8 avoiding non-shortest encodings: it is technically
260possible to UTF-8-encode a single code point in different ways, but that is
261explicitly forbidden, and the shortest possible encoding should always be used
15824458 262(and that is what Perl does). The non-shortest ones are called 'overlongs'.
8c007b5a 263
877d9f0d
JH
264 */
265
8c007b5a
JH
266/*
267 Another way to look at it, as bits:
268
b2635aa8 269 Code Points 1st Byte 2nd Byte 3rd Byte 4th Byte
8c007b5a 270
b2635aa8
KW
271 0aaa aaaa 0aaa aaaa
272 0000 0bbb bbaa aaaa 110b bbbb 10aa aaaa
273 cccc bbbb bbaa aaaa 1110 cccc 10bb bbbb 10aa aaaa
274 00 000d ddcc cccc bbbb bbaa aaaa 1111 0ddd 10cc cccc 10bb bbbb 10aa aaaa
8c007b5a
JH
275
276As you can see, the continuation bytes all begin with C<10>, and the
e1b711da 277leading bits of the start byte tell how many bytes there are in the
8c007b5a
JH
278encoded character.
279
df863e43
KW
280Perl's extended UTF-8 means we can have start bytes up through FF, though any
281beginning with FF yields a code point that is too large for 32-bit ASCII
282platforms. FF signals to use 13 bytes for the encoded character. This breaks
283the paradigm that the number of leading bits gives how many total bytes there
284are in the character.
65ab9279 285
d296fe14 286=cut
8c007b5a
JH
287*/
288
6c88483e 289/* Is the representation of the Unicode code point 'cp' the same regardless of
15824458 290 * being encoded in UTF-8 or not? */
2d1545e5 291#define OFFUNI_IS_INVARIANT(cp) isASCII(cp)
15824458 292
c2b32798
KW
293/*
294=for apidoc Am|bool|UVCHR_IS_INVARIANT|UV cp
295
296Evaluates to 1 if the representation of code point C<cp> is the same whether or
297not it is encoded in UTF-8; otherwise evaluates to 0. UTF-8 invariant
298characters can be copied as-is when converting to/from UTF-8, saving time.
299C<cp> is Unicode if above 255; otherwise is platform-native.
300
301=cut
302 */
303
cf1be84e 304#define UVCHR_IS_INVARIANT(cp) OFFUNI_IS_INVARIANT(cp)
38953e5a 305
c9264833
KW
306/* This defines the bits that are to be in the continuation bytes of a multi-byte
307 * UTF-8 encoded character that mark it is a continuation byte. */
308#define UTF_CONTINUATION_MARK 0x80
309
a95ec4fb 310/* Misleadingly named: is the UTF8-encoded byte 'c' part of a variant sequence
5c06326b
KW
311 * in UTF-8? This is the inverse of UTF8_IS_INVARIANT. The |0 makes sure this
312 * isn't mistakenly called with a ptr argument */
a6951642
KW
313#define UTF8_IS_CONTINUED(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
314 ((U8)((c) | 0)) & UTF_CONTINUATION_MARK)
15824458
KW
315
316/* Is the byte 'c' the first byte of a multi-byte UTF8-8 encoded sequence?
317 * This doesn't catch invariants (they are single-byte). It also excludes the
5c06326b
KW
318 * illegal overlong sequences that begin with C0 and C1. The |0 makes sure
319 * this isn't mistakenly called with a ptr argument */
a6951642
KW
320#define UTF8_IS_START(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
321 ((U8)((c) | 0)) >= 0xc2)
15824458 322
858cd8ab
KW
323/* For use in UTF8_IS_CONTINUATION() below */
324#define UTF_IS_CONTINUATION_MASK 0xC0
325
15824458 326/* Is the byte 'c' part of a multi-byte UTF8-8 encoded sequence, and not the
5c06326b
KW
327 * first byte thereof? The |0 makes sure this isn't mistakenly called with a
328 * ptr argument */
a6951642
KW
329#define UTF8_IS_CONTINUATION(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
330 (((U8)((c) | 0)) & UTF_IS_CONTINUATION_MASK) == UTF_CONTINUATION_MARK)
0ae1fa71 331
15824458
KW
332/* Is the UTF8-encoded byte 'c' the first byte of a two byte sequence? Use
333 * UTF8_IS_NEXT_CHAR_DOWNGRADEABLE() instead if the input isn't known to
334 * be well-formed. Masking with 0xfe allows the low bit to be 0 or 1; thus
5c06326b
KW
335 * this matches 0xc[23]. The |0 makes sure this isn't mistakenly called with a
336 * ptr argument */
a6951642
KW
337#define UTF8_IS_DOWNGRADEABLE_START(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
338 (((U8)((c) | 0)) & 0xfe) == 0xc2)
4ab10950 339
15824458 340/* Is the UTF8-encoded byte 'c' the first byte of a sequence of bytes that
5c06326b
KW
341 * represent a code point > 255? The |0 makes sure this isn't mistakenly
342 * called with a ptr argument */
a6951642
KW
343#define UTF8_IS_ABOVE_LATIN1(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
344 ((U8)((c) | 0)) >= 0xc4)
8850bf83 345
15824458
KW
346/* This is the number of low-order bits a continuation byte in a UTF-8 encoded
347 * sequence contributes to the specification of the code point. In the bit
348 * maps above, you see that the first 2 bits are a constant '10', leaving 6 of
349 * real information */
1d72bdf6 350#define UTF_ACCUMULATION_SHIFT 6
b2635aa8 351
fed423a5
KW
352/* ^? is defined to be DEL on ASCII systems. See the definition of toCTRL()
353 * for more */
354#define QUESTION_MARK_CTRL DEL_NATIVE
355
356/* Surrogates, non-character code points and above-Unicode code points are
357 * problematic in some contexts. This allows code that needs to check for
358 * those to to quickly exclude the vast majority of code points it will
359 * encounter */
a6951642
KW
360#define isUTF8_POSSIBLY_PROBLEMATIC(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
361 (U8) c >= 0xED)
fed423a5 362
57ff5f59
KW
363#define UNICODE_IS_PERL_EXTENDED(uv) UNLIKELY((UV) (uv) > 0x7FFFFFFF)
364
fed423a5
KW
365#endif /* EBCDIC vs ASCII */
366
367/* 2**UTF_ACCUMULATION_SHIFT - 1 */
368#define UTF_CONTINUATION_MASK ((U8) ((1U << UTF_ACCUMULATION_SHIFT) - 1))
369
7028aeba
KW
370/* Internal macro to be used only in this file to aid in constructing other
371 * publicly accessible macros.
372 * The number of bytes required to express this uv in UTF-8, for just those
373 * uv's requiring 2 through 6 bytes, as these are common to all platforms and
374 * word sizes. The number of bytes needed is given by the number of leading 1
375 * bits in the start byte. There are 32 start bytes that have 2 initial 1 bits
376 * (C0-DF); there are 16 that have 3 initial 1 bits (E0-EF); 8 that have 4
377 * initial 1 bits (F0-F8); 4 that have 5 initial 1 bits (F9-FB), and 2 that
378 * have 6 initial 1 bits (FC-FD). The largest number a string of n bytes can
379 * represent is (the number of possible start bytes for 'n')
380 * * (the number of possiblities for each start byte
381 * The latter in turn is
382 * 2 ** ( (how many continuation bytes there are)
383 * * (the number of bits of information each
384 * continuation byte holds))
385 *
386 * If we were on a platform where we could use a fast find first set bit
387 * instruction (or count leading zeros instruction) this could be replaced by
388 * using that to find the log2 of the uv, and divide that by the number of bits
389 * of information in each continuation byte, adjusting for large cases and how
390 * much information is in a start byte for that length */
72164d3a 391#define __COMMON_UNI_SKIP(uv) \
7028aeba
KW
392 (UV) (uv) < (32 * (1U << ( UTF_ACCUMULATION_SHIFT))) ? 2 : \
393 (UV) (uv) < (16 * (1U << (2 * UTF_ACCUMULATION_SHIFT))) ? 3 : \
394 (UV) (uv) < ( 8 * (1U << (3 * UTF_ACCUMULATION_SHIFT))) ? 4 : \
395 (UV) (uv) < ( 4 * (1U << (4 * UTF_ACCUMULATION_SHIFT))) ? 5 : \
396 (UV) (uv) < ( 2 * (1U << (5 * UTF_ACCUMULATION_SHIFT))) ? 6 :
72164d3a
KW
397
398/* Internal macro to be used only in this file.
399 * This adds to __COMMON_UNI_SKIP the details at this platform's upper range.
fed423a5 400 * For any-sized EBCDIC platforms, or 64-bit ASCII ones, we need one more test
72164d3a
KW
401 * to see if just 7 bytes is needed, or if the maximum is needed. For 32-bit
402 * ASCII platforms, everything is representable by 7 bytes */
fed423a5 403#if defined(UV_IS_QUAD) || defined(EBCDIC)
72164d3a 404# define __BASE_UNI_SKIP(uv) (__COMMON_UNI_SKIP(uv) \
7028aeba 405 (UV) (uv) < ((UV) 1U << (6 * UTF_ACCUMULATION_SHIFT)) ? 7 : UTF8_MAXBYTES)
1d68d6cd 406#else
72164d3a 407# define __BASE_UNI_SKIP(uv) (__COMMON_UNI_SKIP(uv) 7)
1d68d6cd
SC
408#endif
409
fed423a5
KW
410/* The next two macros use the base macro defined above, and add in the tests
411 * at the low-end of the range, for just 1 byte, yielding complete macros,
412 * publicly accessible. */
413
414/* Input is a true Unicode (not-native) code point */
415#define OFFUNISKIP(uv) (OFFUNI_IS_INVARIANT(uv) ? 1 : __BASE_UNI_SKIP(uv))
2084b489 416
5352a763
KW
417/*
418
419=for apidoc Am|STRLEN|UVCHR_SKIP|UV cp
420returns the number of bytes required to represent the code point C<cp> when
421encoded as UTF-8. C<cp> is a native (ASCII or EBCDIC) code point if less than
422255; a Unicode code point otherwise.
423
424=cut
425 */
fdb6583d 426#define UVCHR_SKIP(uv) ( UVCHR_IS_INVARIANT(uv) ? 1 : __BASE_UNI_SKIP(uv))
5352a763 427
b651802e
KW
428/* The largest code point representable by two UTF-8 bytes on this platform.
429 * As explained in the comments for __COMMON_UNI_SKIP, 32 start bytes with
fed423a5 430 * UTF_ACCUMULATION_SHIFT bits of information each */
aa206fb7
KW
431#define MAX_UTF8_TWO_BYTE (32 * (1U << UTF_ACCUMULATION_SHIFT) - 1)
432
b651802e
KW
433/* The largest code point representable by two UTF-8 bytes on any platform that
434 * Perl runs on. This value is constrained by EBCDIC which has 5 bits per
435 * continuation byte */
aa206fb7
KW
436#define MAX_PORTABLE_UTF8_TWO_BYTE (32 * (1U << 5) - 1)
437
c03c0950
KW
438/* The maximum number of UTF-8 bytes a single Unicode character can
439 * uppercase/lowercase/fold into. Unicode guarantees that the maximum
440 * expansion is UTF8_MAX_FOLD_CHAR_EXPAND characters, but any above-Unicode
441 * code point will fold to itself, so we only have to look at the expansion of
442 * the maximum Unicode code point. But this number may be less than the space
443 * occupied by a very large code point under Perl's extended UTF-8. We have to
444 * make it large enough to fit any single character. (It turns out that ASCII
445 * and EBCDIC differ in which is larger) */
446#define UTF8_MAXBYTES_CASE \
447 (UTF8_MAXBYTES >= (UTF8_MAX_FOLD_CHAR_EXPAND * OFFUNISKIP(0x10FFFF)) \
448 ? UTF8_MAXBYTES \
449 : (UTF8_MAX_FOLD_CHAR_EXPAND * OFFUNISKIP(0x10FFFF)))
450
d06134e5
KW
451/* Rest of these are attributes of Unicode and perl's internals rather than the
452 * encoding, or happen to be the same in both ASCII and EBCDIC (at least at
453 * this level; the macros that some of these call may have different
454 * definitions in the two encodings */
455
59a449d5
KW
456/* In domain restricted to ASCII, these may make more sense to the reader than
457 * the ones with Latin1 in the name */
458#define NATIVE_TO_ASCII(ch) NATIVE_TO_LATIN1(ch)
459#define ASCII_TO_NATIVE(ch) LATIN1_TO_NATIVE(ch)
460
461/* More or less misleadingly-named defines, retained for back compat */
462#define NATIVE_TO_UTF(ch) NATIVE_UTF8_TO_I8(ch)
463#define NATIVE_TO_I8(ch) NATIVE_UTF8_TO_I8(ch)
464#define UTF_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch)
465#define I8_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch)
466#define NATIVE8_TO_UNI(ch) NATIVE_TO_LATIN1(ch)
d06134e5 467
c0236afe 468/* This defines the 1-bits that are to be in the first byte of a multi-byte
97d0ceda
KW
469 * UTF-8 encoded character that mark it as a start byte and give the number of
470 * bytes that comprise the character. 'len' is the number of bytes in the
471 * multi-byte sequence. */
c0236afe
KW
472#define UTF_START_MARK(len) (((len) > 7) ? 0xFF : (0xFF & (0xFE << (7-(len)))))
473
474/* Masks out the initial one bits in a start byte, leaving the real data ones.
475 * Doesn't work on an invariant byte. 'len' is the number of bytes in the
476 * multi-byte sequence that comprises the character. */
477#define UTF_START_MASK(len) (((len) >= 7) ? 0x00 : (0x1F >> ((len)-2)))
478
537124e4
KW
479/* Adds a UTF8 continuation byte 'new' of information to a running total code
480 * point 'old' of all the continuation bytes so far. This is designed to be
155d2738
KW
481 * used in a loop to convert from UTF-8 to the code point represented. Note
482 * that this is asymmetric on EBCDIC platforms, in that the 'new' parameter is
483 * the UTF-EBCDIC byte, whereas the 'old' parameter is a Unicode (not EBCDIC)
484 * code point in process of being generated */
a6951642
KW
485#define UTF8_ACCUMULATE(old, new) (__ASSERT_(FITS_IN_8_BITS(new)) \
486 ((old) << UTF_ACCUMULATION_SHIFT) \
155d2738
KW
487 | ((NATIVE_UTF8_TO_I8((U8)new)) \
488 & UTF_CONTINUATION_MASK))
d06134e5 489
4ab10950 490/* This works in the face of malformed UTF-8. */
4e1ed312
KW
491#define UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(s, e) \
492 ( UTF8_IS_DOWNGRADEABLE_START(*(s)) \
493 && ( (e) - (s) > 1) \
494 && UTF8_IS_CONTINUATION(*((s)+1)))
4ab10950 495
5aaebcb3 496/* Number of bytes a code point occupies in UTF-8. */
5352a763 497#define NATIVE_SKIP(uv) UVCHR_SKIP(uv)
bd18bd40 498
5aaebcb3
KW
499/* Most code which says UNISKIP is really thinking in terms of native code
500 * points (0-255) plus all those beyond. This is an imprecise term, but having
2accb712 501 * it means existing code continues to work. For precision, use UVCHR_SKIP,
5352a763
KW
502 * NATIVE_SKIP, or OFFUNISKIP */
503#define UNISKIP(uv) UVCHR_SKIP(uv)
5aaebcb3 504
3c0792e4
KW
505/* Longer, but more accurate name */
506#define UTF8_IS_ABOVE_LATIN1_START(c) UTF8_IS_ABOVE_LATIN1(c)
507
a62b247b
KW
508/* Convert a UTF-8 variant Latin1 character to a native code point value.
509 * Needs just one iteration of accumulate. Should be used only if it is known
510 * that the code point is < 256, and is not UTF-8 invariant. Use the slower
511 * but more general TWO_BYTE_UTF8_TO_NATIVE() which handles any code point
512 * representable by two bytes (which turns out to be up through
513 * MAX_PORTABLE_UTF8_TWO_BYTE). The two parameters are:
514 * HI: a downgradable start byte;
515 * LO: continuation.
516 * */
517#define EIGHT_BIT_UTF8_TO_NATIVE(HI, LO) \
518 ( __ASSERT_(UTF8_IS_DOWNGRADEABLE_START(HI)) \
519 __ASSERT_(UTF8_IS_CONTINUATION(LO)) \
520 LATIN1_TO_NATIVE(UTF8_ACCUMULATE(( \
521 NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), (LO))))
522
94bb8c36 523/* Convert a two (not one) byte utf8 character to a native code point value.
2950f2a7
KW
524 * Needs just one iteration of accumulate. Should not be used unless it is
525 * known that the two bytes are legal: 1) two-byte start, and 2) continuation.
526 * Note that the result can be larger than 255 if the input character is not
527 * downgradable */
94bb8c36 528#define TWO_BYTE_UTF8_TO_NATIVE(HI, LO) \
a6951642
KW
529 (__ASSERT_(FITS_IN_8_BITS(HI)) \
530 __ASSERT_(FITS_IN_8_BITS(LO)) \
531 __ASSERT_(PL_utf8skip[HI] == 2) \
532 __ASSERT_(UTF8_IS_CONTINUATION(LO)) \
94bb8c36 533 UNI_TO_NATIVE(UTF8_ACCUMULATE((NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), \
635e76f5 534 (LO))))
94bb8c36
KW
535
536/* Should never be used, and be deprecated */
537#define TWO_BYTE_UTF8_TO_UNI(HI, LO) NATIVE_TO_UNI(TWO_BYTE_UTF8_TO_NATIVE(HI, LO))
2950f2a7 538
bd18bd40
KW
539/*
540
541=for apidoc Am|STRLEN|UTF8SKIP|char* s
542returns the number of bytes in the UTF-8 encoded character whose first (perhaps
543only) byte is pointed to by C<s>.
544
545=cut
546 */
2a70536e
KW
547#define UTF8SKIP(s) PL_utf8skip[*(const U8*)(s)]
548#define UTF8_SKIP(s) UTF8SKIP(s)
d06134e5 549
85fcc8f2
KW
550/*
551
552=for apidoc Am|STRLEN|UTF8_SAFE_SKIP|char* s|char* e
45671da2
KW
553returns 0 if S<C<s E<gt>= e>>; otherwise returns the number of bytes in the
554UTF-8 encoded character whose first byte is pointed to by C<s>. But it never
555returns beyond C<e>. On DEBUGGING builds, it asserts that S<C<s E<lt>= e>>.
85fcc8f2
KW
556
557=cut
558 */
45671da2
KW
559#define UTF8_SAFE_SKIP(s, e) (__ASSERT_((e) >= (s)) \
560 ((e) - (s)) <= 0 \
561 ? 0 \
562 : MIN(((e) - (s)), UTF8_SKIP(s)))
85fcc8f2 563
2d1545e5
KW
564/* Most code that says 'UNI_' really means the native value for code points up
565 * through 255 */
566#define UNI_IS_INVARIANT(cp) UVCHR_IS_INVARIANT(cp)
567
c2b32798
KW
568/*
569=for apidoc Am|bool|UTF8_IS_INVARIANT|char c
570
571Evaluates to 1 if the byte C<c> represents the same character when encoded in
572UTF-8 as when not; otherwise evaluates to 0. UTF-8 invariant characters can be
573copied as-is when converting to/from UTF-8, saving time.
574
575In spite of the name, this macro gives the correct result if the input string
576from which C<c> comes is not encoded in UTF-8.
577
578See C<L</UVCHR_IS_INVARIANT>> for checking if a UV is invariant.
579
580=cut
581
582The reason it works on both UTF-8 encoded strings and non-UTF-8 encoded, is
583that it returns TRUE in each for the exact same set of bit patterns. It is
584valid on a subset of what UVCHR_IS_INVARIANT is valid on, so can just use that;
585and the compiler should optimize out anything extraneous given the
586implementation of the latter. The |0 makes sure this isn't mistakenly called
587with a ptr argument.
588*/
5c06326b 589#define UTF8_IS_INVARIANT(c) UVCHR_IS_INVARIANT((c) | 0)
5fc230f1
KW
590
591/* Like the above, but its name implies a non-UTF8 input, which as the comments
592 * above show, doesn't matter as to its implementation */
38953e5a 593#define NATIVE_BYTE_IS_INVARIANT(c) UVCHR_IS_INVARIANT(c)
d06134e5 594
48ccf5e1
KW
595/* The macros in the next 4 sets are used to generate the two utf8 or utfebcdic
596 * bytes from an ordinal that is known to fit into exactly two (not one) bytes;
597 * it must be less than 0x3FF to work across both encodings. */
598
599/* These two are helper macros for the other three sets, and should not be used
600 * directly anywhere else. 'translate_function' is either NATIVE_TO_LATIN1
1ff3baa2
KW
601 * (which works for code points up through 0xFF) or NATIVE_TO_UNI which works
602 * for any code point */
48ccf5e1 603#define __BASE_TWO_BYTE_HI(c, translate_function) \
2863dafa 604 (__ASSERT_(! UVCHR_IS_INVARIANT(c)) \
48ccf5e1 605 I8_TO_NATIVE_UTF8((translate_function(c) >> UTF_ACCUMULATION_SHIFT) \
2863dafa 606 | UTF_START_MARK(2)))
48ccf5e1 607#define __BASE_TWO_BYTE_LO(c, translate_function) \
2863dafa 608 (__ASSERT_(! UVCHR_IS_INVARIANT(c)) \
48ccf5e1 609 I8_TO_NATIVE_UTF8((translate_function(c) & UTF_CONTINUATION_MASK) \
2863dafa 610 | UTF_CONTINUATION_MARK))
48ccf5e1 611
48ccf5e1
KW
612/* The next two macros should not be used. They were designed to be usable as
613 * the case label of a switch statement, but this doesn't work for EBCDIC. Use
9d0d3a03 614 * regen/unicode_constants.pl instead */
48ccf5e1
KW
615#define UTF8_TWO_BYTE_HI_nocast(c) __BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI)
616#define UTF8_TWO_BYTE_LO_nocast(c) __BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI)
617
618/* The next two macros are used when the source should be a single byte
619 * character; checked for under DEBUGGING */
620#define UTF8_EIGHT_BIT_HI(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
4c8cd605 621 ( __BASE_TWO_BYTE_HI(c, NATIVE_TO_LATIN1)))
48ccf5e1 622#define UTF8_EIGHT_BIT_LO(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
4c8cd605 623 (__BASE_TWO_BYTE_LO(c, NATIVE_TO_LATIN1)))
48ccf5e1
KW
624
625/* These final two macros in the series are used when the source can be any
626 * code point whose UTF-8 is known to occupy 2 bytes; they are less efficient
627 * than the EIGHT_BIT versions on EBCDIC platforms. We use the logical '~'
628 * operator instead of "<=" to avoid getting compiler warnings.
d52b8576 629 * MAX_UTF8_TWO_BYTE should be exactly all one bits in the lower few
48ccf5e1
KW
630 * places, so the ~ works */
631#define UTF8_TWO_BYTE_HI(c) \
632 (__ASSERT_((sizeof(c) == 1) \
d52b8576 633 || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \
4c8cd605 634 (__BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI)))
48ccf5e1
KW
635#define UTF8_TWO_BYTE_LO(c) \
636 (__ASSERT_((sizeof(c) == 1) \
d52b8576 637 || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \
4c8cd605 638 (__BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI)))
d06134e5 639
e7214ce8
KW
640/* This is illegal in any well-formed UTF-8 in both EBCDIC and ASCII
641 * as it is only in overlongs. */
642#define ILLEGAL_UTF8_BYTE I8_TO_NATIVE_UTF8(0xC1)
643
7e2040f0 644/*
e3036cf4 645 * 'UTF' is whether or not p is encoded in UTF8. The names 'foo_lazy_if' stem
20df05f4
KW
646 * from an earlier version of these macros in which they didn't call the
647 * foo_utf8() macros (i.e. were 'lazy') unless they decided that *p is the
648 * beginning of a utf8 character. Now that foo_utf8() determines that itself,
649 * no need to do it again here
7e2040f0 650 */
34aeb2e9
KW
651#define isIDFIRST_lazy_if(p,UTF) \
652 _is_utf8_FOO(_CC_IDFIRST, (const U8 *) p, "isIDFIRST_lazy_if", \
653 "isIDFIRST_lazy_if_safe", \
654 cBOOL(UTF && ! IN_BYTES), 0, __FILE__,__LINE__)
1d72bdf6 655
da8c1a98
KW
656#define isIDFIRST_lazy_if_safe(p, e, UTF) \
657 ((IN_BYTES || !UTF) \
658 ? isIDFIRST(*(p)) \
659 : isIDFIRST_utf8_safe(p, e))
660
34aeb2e9
KW
661#define isWORDCHAR_lazy_if(p,UTF) \
662 _is_utf8_FOO(_CC_IDFIRST, (const U8 *) p, "isWORDCHAR_lazy_if", \
663 "isWORDCHAR_lazy_if_safe", \
664 cBOOL(UTF && ! IN_BYTES), 0, __FILE__,__LINE__)
665
da8c1a98
KW
666#define isWORDCHAR_lazy_if_safe(p, e, UTF) \
667 ((IN_BYTES || !UTF) \
668 ? isWORDCHAR(*(p)) \
669 : isWORDCHAR_utf8_safe((U8 *) p, (U8 *) e))
670
34aeb2e9
KW
671#define isALNUM_lazy_if(p,UTF) \
672 _is_utf8_FOO(_CC_IDFIRST, (const U8 *) p, "isALNUM_lazy_if", \
673 "isWORDCHAR_lazy_if_safe", \
674 cBOOL(UTF && ! IN_BYTES), 0, __FILE__,__LINE__)
da8c1a98 675
89ebb4a3
JH
676#define UTF8_MAXLEN UTF8_MAXBYTES
677
8cb75cc8
KW
678/* A Unicode character can fold to up to 3 characters */
679#define UTF8_MAX_FOLD_CHAR_EXPAND 3
680
d3481830 681#define IN_BYTES UNLIKELY(CopHINTS_get(PL_curcop) & HINT_BYTES)
bd18bd40
KW
682
683/*
684
685=for apidoc Am|bool|DO_UTF8|SV* sv
686Returns a bool giving whether or not the PV in C<sv> is to be treated as being
687encoded in UTF-8.
688
689You should use this I<after> a call to C<SvPV()> or one of its variants, in
690case any call to string overloading updates the internal UTF-8 encoding flag.
691
692=cut
693*/
0064a8a9 694#define DO_UTF8(sv) (SvUTF8(sv) && !IN_BYTES)
1ff3baa2
KW
695
696/* Should all strings be treated as Unicode, and not just UTF-8 encoded ones?
697 * Is so within 'feature unicode_strings' or 'locale :not_characters', and not
698 * within 'use bytes'. UTF-8 locales are not tested for here, but perhaps
699 * could be */
70844984
KW
700#define IN_UNI_8_BIT \
701 (( ( (CopHINTS_get(PL_curcop) & HINT_UNI_8_BIT)) \
702 || ( CopHINTS_get(PL_curcop) & HINT_LOCALE_PARTIAL \
703 /* -1 below is for :not_characters */ \
704 && _is_in_locale_category(FALSE, -1))) \
705 && (! IN_BYTES))
b36bf33f 706
1d72bdf6 707
c76687c5 708#define UTF8_ALLOW_EMPTY 0x0001 /* Allow a zero length string */
2b5e7bc2 709#define UTF8_GOT_EMPTY UTF8_ALLOW_EMPTY
c76687c5
KW
710
711/* Allow first byte to be a continuation byte */
1d72bdf6 712#define UTF8_ALLOW_CONTINUATION 0x0002
2b5e7bc2 713#define UTF8_GOT_CONTINUATION UTF8_ALLOW_CONTINUATION
c76687c5 714
cd01d3b1 715/* Unexpected non-continuation byte */
1d72bdf6 716#define UTF8_ALLOW_NON_CONTINUATION 0x0004
2b5e7bc2 717#define UTF8_GOT_NON_CONTINUATION UTF8_ALLOW_NON_CONTINUATION
949cf498
KW
718
719/* expecting more bytes than were available in the string */
720#define UTF8_ALLOW_SHORT 0x0008
2b5e7bc2 721#define UTF8_GOT_SHORT UTF8_ALLOW_SHORT
949cf498 722
94953955
KW
723/* Overlong sequence; i.e., the code point can be specified in fewer bytes.
724 * First one will convert the overlong to the REPLACEMENT CHARACTER; second
725 * will return what the overlong evaluates to */
949cf498 726#define UTF8_ALLOW_LONG 0x0010
94953955 727#define UTF8_ALLOW_LONG_AND_ITS_VALUE (UTF8_ALLOW_LONG|0x0020)
2b5e7bc2
KW
728#define UTF8_GOT_LONG UTF8_ALLOW_LONG
729
d60baaa7
KW
730#define UTF8_ALLOW_OVERFLOW 0x0080
731#define UTF8_GOT_OVERFLOW UTF8_ALLOW_OVERFLOW
949cf498 732
f180b292 733#define UTF8_DISALLOW_SURROGATE 0x0100 /* Unicode surrogates */
2b5e7bc2 734#define UTF8_GOT_SURROGATE UTF8_DISALLOW_SURROGATE
f180b292 735#define UTF8_WARN_SURROGATE 0x0200
949cf498 736
c4e96019
KW
737/* Unicode non-character code points */
738#define UTF8_DISALLOW_NONCHAR 0x0400
2b5e7bc2 739#define UTF8_GOT_NONCHAR UTF8_DISALLOW_NONCHAR
c4e96019 740#define UTF8_WARN_NONCHAR 0x0800
949cf498 741
c4e96019
KW
742/* Super-set of Unicode: code points above the legal max */
743#define UTF8_DISALLOW_SUPER 0x1000
2b5e7bc2 744#define UTF8_GOT_SUPER UTF8_DISALLOW_SUPER
c4e96019
KW
745#define UTF8_WARN_SUPER 0x2000
746
747/* The original UTF-8 standard did not define UTF-8 with start bytes of 0xFE or
748 * 0xFF, though UTF-EBCDIC did. This allowed both versions to represent code
749 * points up to 2 ** 31 - 1. Perl extends UTF-8 so that 0xFE and 0xFF are
750 * usable on ASCII platforms, and 0xFF means something different than
751 * UTF-EBCDIC defines. These changes allow code points of 64 bits (actually
752 * somewhat more) to be represented on both platforms. But these are Perl
753 * extensions, and not likely to be interchangeable with other languages. Note
754 * that on ASCII platforms, FE overflows a signed 32-bit word, and FF an
755 * unsigned one. */
d044b7a7
KW
756#define UTF8_DISALLOW_PERL_EXTENDED 0x4000
757#define UTF8_GOT_PERL_EXTENDED UTF8_DISALLOW_PERL_EXTENDED
758#define UTF8_WARN_PERL_EXTENDED 0x8000
d35f2ca5 759
57ff5f59
KW
760/* For back compat, these old names are misleading for overlongs and
761 * UTF_EBCDIC. */
d044b7a7
KW
762#define UTF8_DISALLOW_ABOVE_31_BIT UTF8_DISALLOW_PERL_EXTENDED
763#define UTF8_GOT_ABOVE_31_BIT UTF8_GOT_PERL_EXTENDED
764#define UTF8_WARN_ABOVE_31_BIT UTF8_WARN_PERL_EXTENDED
765#define UTF8_DISALLOW_FE_FF UTF8_DISALLOW_PERL_EXTENDED
766#define UTF8_WARN_FE_FF UTF8_WARN_PERL_EXTENDED
949cf498 767
f180b292 768#define UTF8_CHECK_ONLY 0x10000
99a765e9 769#define _UTF8_NO_CONFIDENCE_IN_CURLEN 0x20000 /* Internal core use only */
949cf498
KW
770
771/* For backwards source compatibility. They do nothing, as the default now
772 * includes what they used to mean. The first one's meaning was to allow the
773 * just the single non-character 0xFFFF */
774#define UTF8_ALLOW_FFFF 0
c825ef8c 775#define UTF8_ALLOW_FE_FF 0
949cf498
KW
776#define UTF8_ALLOW_SURROGATE 0
777
ecc1615f
KW
778/* C9 refers to Unicode Corrigendum #9: allows but discourages non-chars */
779#define UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE \
780 (UTF8_DISALLOW_SUPER|UTF8_DISALLOW_SURROGATE)
781#define UTF8_WARN_ILLEGAL_C9_INTERCHANGE (UTF8_WARN_SUPER|UTF8_WARN_SURROGATE)
782
d35f2ca5 783#define UTF8_DISALLOW_ILLEGAL_INTERCHANGE \
ecc1615f 784 (UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE|UTF8_DISALLOW_NONCHAR)
949cf498 785#define UTF8_WARN_ILLEGAL_INTERCHANGE \
ecc1615f
KW
786 (UTF8_WARN_ILLEGAL_C9_INTERCHANGE|UTF8_WARN_NONCHAR)
787
0eb3d6a0
KW
788/* This is typically used for code that processes UTF-8 input and doesn't want
789 * to have to deal with any malformations that might be present. All such will
790 * be safely replaced by the REPLACEMENT CHARACTER, unless other flags
791 * overriding this are also present. */
2d532c27
KW
792#define UTF8_ALLOW_ANY ( UTF8_ALLOW_CONTINUATION \
793 |UTF8_ALLOW_NON_CONTINUATION \
794 |UTF8_ALLOW_SHORT \
d60baaa7
KW
795 |UTF8_ALLOW_LONG \
796 |UTF8_ALLOW_OVERFLOW)
2d532c27
KW
797
798/* Accept any Perl-extended UTF-8 that evaluates to any UV on the platform, but
cd01d3b1 799 * not any malformed. This is the default. */
2d532c27
KW
800#define UTF8_ALLOW_ANYUV 0
801#define UTF8_ALLOW_DEFAULT UTF8_ALLOW_ANYUV
1d72bdf6 802
89d986df
KW
803/*
804=for apidoc Am|bool|UTF8_IS_SURROGATE|const U8 *s|const U8 *e
805
806Evaluates to non-zero if the first few bytes of the string starting at C<s> and
807looking no further than S<C<e - 1>> are well-formed UTF-8 that represents one
808of the Unicode surrogate code points; otherwise it evaluates to 0. If
809non-zero, the value gives how many bytes starting at C<s> comprise the code
810point's representation.
811
812=cut
813 */
814#define UTF8_IS_SURROGATE(s, e) is_SURROGATE_utf8_safe(s, e)
815
816
817#define UTF8_IS_REPLACEMENT(s, send) is_REPLACEMENT_utf8_safe(s,send)
818
285aa1f0 819#define MAX_LEGAL_CP ((UV)IV_MAX)
40606899 820
89d986df
KW
821/*
822=for apidoc Am|bool|UTF8_IS_SUPER|const U8 *s|const U8 *e
823
824Recall that Perl recognizes an extension to UTF-8 that can encode code
825points larger than the ones defined by Unicode, which are 0..0x10FFFF.
826
827This macro evaluates to non-zero if the first few bytes of the string starting
828at C<s> and looking no further than S<C<e - 1>> are from this UTF-8 extension;
829otherwise it evaluates to 0. If non-zero, the value gives how many bytes
830starting at C<s> comprise the code point's representation.
831
8320 is returned if the bytes are not well-formed extended UTF-8, or if they
833represent a code point that cannot fit in a UV on the current platform. Hence
834this macro can give different results when run on a 64-bit word machine than on
835one with a 32-bit word size.
0c58a72b 836
891fd405 837Note that it is illegal to have code points that are larger than what can
89d986df 838fit in an IV on the current machine.
7131f24d 839
89d986df
KW
840=cut
841
842 * ASCII EBCDIC I8
7131f24d
KW
843 * U+10FFFF: \xF4\x8F\xBF\xBF \xF9\xA1\xBF\xBF\xBF max legal Unicode
844 * U+110000: \xF4\x90\x80\x80 \xF9\xA2\xA0\xA0\xA0
845 * U+110001: \xF4\x90\x80\x81 \xF9\xA2\xA0\xA0\xA1
89d986df
KW
846 */
847#ifdef EBCDIC
a14e0a36 848# define UTF8_IS_SUPER(s, e) \
89d986df
KW
849 (( LIKELY((e) > (s) + 4) \
850 && NATIVE_UTF8_TO_I8(*(s)) >= 0xF9 \
851 && ( NATIVE_UTF8_TO_I8(*(s)) > 0xF9 \
3d42f267 852 || (NATIVE_UTF8_TO_I8(*((s) + 1)) >= 0xA2)) \
89d986df 853 && LIKELY((s) + UTF8SKIP(s) <= (e))) \
2b479609 854 ? _is_utf8_char_helper(s, s + UTF8SKIP(s), 0) : 0)
7131f24d 855#else
a14e0a36 856# define UTF8_IS_SUPER(s, e) \
89d986df
KW
857 (( LIKELY((e) > (s) + 3) \
858 && (*(U8*) (s)) >= 0xF4 \
859 && ((*(U8*) (s)) > 0xF4 || (*((U8*) (s) + 1) >= 0x90))\
860 && LIKELY((s) + UTF8SKIP(s) <= (e))) \
2b479609 861 ? _is_utf8_char_helper(s, s + UTF8SKIP(s), 0) : 0)
7131f24d
KW
862#endif
863
b96a92fb
KW
864/* These are now machine generated, and the 'given' clause is no longer
865 * applicable */
0c58a72b 866#define UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s, e) \
89d986df
KW
867 cBOOL(is_NONCHAR_utf8_safe(s,e))
868
869/*
870=for apidoc Am|bool|UTF8_IS_NONCHAR|const U8 *s|const U8 *e
871
872Evaluates to non-zero if the first few bytes of the string starting at C<s> and
873looking no further than S<C<e - 1>> are well-formed UTF-8 that represents one
874of the Unicode non-character code points; otherwise it evaluates to 0. If
875non-zero, the value gives how many bytes starting at C<s> comprise the code
876point's representation.
877
d296fe14
KW
878=for apidoc AmnU|UV|UNICODE_REPLACEMENT
879
880Evaluates to 0xFFFD, the code point of the Unicode REPLACEMENT CHARACTER
881
89d986df
KW
882=cut
883 */
0c58a72b
KW
884#define UTF8_IS_NONCHAR(s, e) \
885 UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s, e)
7131f24d 886
c867b360
JH
887#define UNICODE_SURROGATE_FIRST 0xD800
888#define UNICODE_SURROGATE_LAST 0xDFFF
889#define UNICODE_REPLACEMENT 0xFFFD
890#define UNICODE_BYTE_ORDER_MARK 0xFEFF
1d72bdf6 891
b851fbc1 892/* Though our UTF-8 encoding can go beyond this,
c76687c5 893 * let's be conservative and do as Unicode says. */
b851fbc1
JH
894#define PERL_UNICODE_MAX 0x10FFFF
895
d044b7a7
KW
896#define UNICODE_WARN_SURROGATE 0x0001 /* UTF-16 surrogates */
897#define UNICODE_WARN_NONCHAR 0x0002 /* Non-char code points */
898#define UNICODE_WARN_SUPER 0x0004 /* Above 0x10FFFF */
899#define UNICODE_WARN_PERL_EXTENDED 0x0008 /* Above 0x7FFF_FFFF */
900#define UNICODE_WARN_ABOVE_31_BIT UNICODE_WARN_PERL_EXTENDED
901#define UNICODE_DISALLOW_SURROGATE 0x0010
902#define UNICODE_DISALLOW_NONCHAR 0x0020
903#define UNICODE_DISALLOW_SUPER 0x0040
904#define UNICODE_DISALLOW_PERL_EXTENDED 0x0080
905#define UNICODE_DISALLOW_ABOVE_31_BIT UNICODE_DISALLOW_PERL_EXTENDED
33f38593
KW
906
907#define UNICODE_GOT_SURROGATE UNICODE_DISALLOW_SURROGATE
908#define UNICODE_GOT_NONCHAR UNICODE_DISALLOW_NONCHAR
909#define UNICODE_GOT_SUPER UNICODE_DISALLOW_SUPER
910#define UNICODE_GOT_PERL_EXTENDED UNICODE_DISALLOW_PERL_EXTENDED
911
ecc1615f
KW
912#define UNICODE_WARN_ILLEGAL_C9_INTERCHANGE \
913 (UNICODE_WARN_SURROGATE|UNICODE_WARN_SUPER)
bb88be5f 914#define UNICODE_WARN_ILLEGAL_INTERCHANGE \
ecc1615f
KW
915 (UNICODE_WARN_ILLEGAL_C9_INTERCHANGE|UNICODE_WARN_NONCHAR)
916#define UNICODE_DISALLOW_ILLEGAL_C9_INTERCHANGE \
917 (UNICODE_DISALLOW_SURROGATE|UNICODE_DISALLOW_SUPER)
bb88be5f 918#define UNICODE_DISALLOW_ILLEGAL_INTERCHANGE \
ecc1615f 919 (UNICODE_DISALLOW_ILLEGAL_C9_INTERCHANGE|UNICODE_DISALLOW_NONCHAR)
949cf498
KW
920
921/* For backward source compatibility, as are now the default */
922#define UNICODE_ALLOW_SURROGATE 0
923#define UNICODE_ALLOW_SUPER 0
924#define UNICODE_ALLOW_ANY 0
b851fbc1 925
2d6b3d38
KW
926/* This matches the 2048 code points between UNICODE_SURROGATE_FIRST (0xD800) and
927 * UNICODE_SURROGATE_LAST (0xDFFF) */
928#define UNICODE_IS_SURROGATE(uv) (((UV) (uv) & (~0xFFFF | 0xF800)) \
929 == 0xD800)
930
646d1759
KW
931#define UNICODE_IS_REPLACEMENT(uv) ((UV) (uv) == UNICODE_REPLACEMENT)
932#define UNICODE_IS_BYTE_ORDER_MARK(uv) ((UV) (uv) == UNICODE_BYTE_ORDER_MARK)
c149ab20
KW
933
934/* Is 'uv' one of the 32 contiguous-range noncharacters? */
935#define UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv) ((UV) (uv) >= 0xFDD0 \
936 && (UV) (uv) <= 0xFDEF)
937
938/* Is 'uv' one of the 34 plane-ending noncharacters 0xFFFE, 0xFFFF, 0x1FFFE,
939 * 0x1FFFF, ... 0x10FFFE, 0x10FFFF, given that we know that 'uv' is not above
940 * the Unicode legal max */
941#define UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv) \
942 (((UV) (uv) & 0xFFFE) == 0xFFFE)
943
944#define UNICODE_IS_NONCHAR(uv) \
945 ( UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv) \
946 || ( LIKELY( ! UNICODE_IS_SUPER(uv)) \
947 && UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv)))
948
949#define UNICODE_IS_SUPER(uv) ((UV) (uv) > PERL_UNICODE_MAX)
1d72bdf6 950
ec34087a
KW
951#define LATIN_SMALL_LETTER_SHARP_S LATIN_SMALL_LETTER_SHARP_S_NATIVE
952#define LATIN_SMALL_LETTER_Y_WITH_DIAERESIS \
953 LATIN_SMALL_LETTER_Y_WITH_DIAERESIS_NATIVE
954#define MICRO_SIGN MICRO_SIGN_NATIVE
955#define LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE \
956 LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE_NATIVE
957#define LATIN_SMALL_LETTER_A_WITH_RING_ABOVE \
958 LATIN_SMALL_LETTER_A_WITH_RING_ABOVE_NATIVE
09091399
JH
959#define UNICODE_GREEK_CAPITAL_LETTER_SIGMA 0x03A3
960#define UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA 0x03C2
961#define UNICODE_GREEK_SMALL_LETTER_SIGMA 0x03C3
9dcbe121 962#define GREEK_SMALL_LETTER_MU 0x03BC
9e682c18
KW
963#define GREEK_CAPITAL_LETTER_MU 0x039C /* Upper and title case
964 of MICRON */
965#define LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS 0x0178 /* Also is title case */
0766489e
KW
966#ifdef LATIN_CAPITAL_LETTER_SHARP_S_UTF8
967# define LATIN_CAPITAL_LETTER_SHARP_S 0x1E9E
968#endif
74894415
KW
969#define LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE 0x130
970#define LATIN_SMALL_LETTER_DOTLESS_I 0x131
9e682c18 971#define LATIN_SMALL_LETTER_LONG_S 0x017F
a9f50d33
KW
972#define LATIN_SMALL_LIGATURE_LONG_S_T 0xFB05
973#define LATIN_SMALL_LIGATURE_ST 0xFB06
9e682c18
KW
974#define KELVIN_SIGN 0x212A
975#define ANGSTROM_SIGN 0x212B
09091399 976
9e55ce06 977#define UNI_DISPLAY_ISPRINT 0x0001
c728cb41
JH
978#define UNI_DISPLAY_BACKSLASH 0x0002
979#define UNI_DISPLAY_QQ (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH)
980#define UNI_DISPLAY_REGEX (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH)
9e55ce06 981
5cd46e1f
KW
982#define ANYOF_FOLD_SHARP_S(node, input, end) \
983 (ANYOF_BITMAP_TEST(node, LATIN_SMALL_LETTER_SHARP_S) && \
137165a6 984 (ANYOF_NONBITMAP(node)) && \
39065660 985 (ANYOF_FLAGS(node) & ANYOF_LOC_NONBITMAP_FOLD) && \
07b6858f 986 ((end) > (input) + 1) && \
305b8651 987 isALPHA_FOLD_EQ((input)[0], 's'))
6302f837 988
ebc501f0 989#define SHARP_S_SKIP 2
3b0fc154 990
3cedd9d9 991#define is_utf8_char_buf(buf, buf_end) isUTF8_CHAR(buf, buf_end)
976c1b08
KW
992#define bytes_from_utf8(s, lenp, is_utf8p) \
993 bytes_from_utf8_loc(s, lenp, is_utf8p, 0)
3cedd9d9 994
e23e8bc1
KW
995/*
996
25e3a4e0
KW
997=for apidoc Am|STRLEN|isUTF8_CHAR_flags|const U8 *s|const U8 *e| const U32 flags
998
999Evaluates to non-zero if the first few bytes of the string starting at C<s> and
1000looking no further than S<C<e - 1>> are well-formed UTF-8, as extended by Perl,
1001that represents some code point, subject to the restrictions given by C<flags>;
1002otherwise it evaluates to 0. If non-zero, the value gives how many bytes
2717076a
KW
1003starting at C<s> comprise the code point's representation. Any bytes remaining
1004before C<e>, but beyond the ones needed to form the first code point in C<s>,
1005are not examined.
25e3a4e0
KW
1006
1007If C<flags> is 0, this gives the same results as C<L</isUTF8_CHAR>>;
1008if C<flags> is C<UTF8_DISALLOW_ILLEGAL_INTERCHANGE>, this gives the same results
1009as C<L</isSTRICT_UTF8_CHAR>>;
1010and if C<flags> is C<UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE>, this gives
1011the same results as C<L</isC9_STRICT_UTF8_CHAR>>.
1012Otherwise C<flags> may be any combination of the C<UTF8_DISALLOW_I<foo>> flags
1013understood by C<L</utf8n_to_uvchr>>, with the same meanings.
1014
1015The three alternative macros are for the most commonly needed validations; they
1016are likely to run somewhat faster than this more general one, as they can be
1017inlined into your code.
1018
9f2abfde
KW
1019Use L</is_utf8_string_flags>, L</is_utf8_string_loc_flags>, and
1020L</is_utf8_string_loclen_flags> to check entire strings.
1021
25e3a4e0
KW
1022=cut
1023*/
1024
1025#define isUTF8_CHAR_flags(s, e, flags) \
1026 (UNLIKELY((e) <= (s)) \
1027 ? 0 \
1028 : (UTF8_IS_INVARIANT(*s)) \
1029 ? 1 \
1030 : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \
1031 ? 0 \
1032 : _is_utf8_char_helper(s, e, flags))
1033
6302f837
KW
1034/* Do not use; should be deprecated. Use isUTF8_CHAR() instead; this is
1035 * retained solely for backwards compatibility */
1036#define IS_UTF8_CHAR(p, n) (isUTF8_CHAR(p, (p) + (n)) == n)
e9a8c099 1037
6a5bc5ac 1038#endif /* PERL_UTF8_H_ */
57f0e7e2 1039
e9a8c099 1040/*
14d04a33 1041 * ex: set ts=8 sts=4 sw=4 et:
e9a8c099 1042 */