This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
perlapi: Fix typo
[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
1d72bdf6
NIS
89#ifdef EBCDIC
90/* The equivalent of these macros but implementing UTF-EBCDIC
91 are in the following header file:
92 */
93
94#include "utfebcdic.h"
fd7cb289 95
d06134e5 96#else /* ! EBCDIC */
73c4f7a1
GS
97START_EXTERN_C
98
f4d83e55
KW
99/*
100
101=for apidoc AmnU|STRLEN|UTF8_MAXBYTES
102
103The maximum width of a single UTF-8 encoded character, in bytes.
104
105NOTE: Strictly speaking Perl's UTF-8 should not be called UTF-8 since UTF-8
106is an encoding of Unicode, and Unicode's upper limit, 0x10FFFF, can be
107expressed with 4 bytes. However, Perl thinks of UTF-8 as a way to encode
108non-negative integers in a binary format, even those above Unicode.
109
110=cut
111 */
111e8ed9
KW
112#define UTF8_MAXBYTES 13
113
a0ed51b3 114#ifdef DOINIT
6f06b55f 115EXTCONST unsigned char PL_utf8skip[] = {
b2635aa8
KW
116/* 0x00 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
117/* 0x10 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
118/* 0x20 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
119/* 0x30 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
120/* 0x40 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
121/* 0x50 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
122/* 0x60 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
123/* 0x70 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
124/* 0x80 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
125/* 0x90 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
126/* 0xA0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
127/* 0xB0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
128/* 0xC0 */ 2,2, /* overlong */
1ff3baa2 129/* 0xC2 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0080 to U+03FF */
b2635aa8
KW
130/* 0xD0 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0400 to U+07FF */
131/* 0xE0 */ 3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3, /* U+0800 to U+FFFF */
132/* 0xF0 */ 4,4,4,4,4,4,4,4,5,5,5,5,6,6, /* above BMP to 2**31 - 1 */
6937f885
KW
133 /* Perl extended (never was official UTF-8). Up to 36 bit */
134/* 0xFE */ 7,
135 /* More extended, Up to 72 bits (64-bit + reserved) */
111e8ed9 136/* 0xFF */ UTF8_MAXBYTES
a0ed51b3
LW
137};
138#else
6f06b55f 139EXTCONST unsigned char PL_utf8skip[];
a0ed51b3
LW
140#endif
141
73c4f7a1 142END_EXTERN_C
7e2040f0 143
ef2a4c8f
KW
144/*
145
146=for apidoc Am|U8|NATIVE_TO_LATIN1|U8 ch
147
148Returns the Latin-1 (including ASCII and control characters) equivalent of the
149input native code point given by C<ch>. Thus, C<NATIVE_TO_LATIN1(193)> on
150EBCDIC platforms returns 65. These each represent the character C<"A"> on
151their respective platforms. On ASCII platforms no conversion is needed, so
152this macro expands to just its input, adding no time nor space requirements to
153the implementation.
154
155For conversion of code points potentially larger than will fit in a character,
156use L</NATIVE_TO_UNI>.
157
158=for apidoc Am|U8|LATIN1_TO_NATIVE|U8 ch
159
160Returns the native equivalent of the input Latin-1 code point (including ASCII
161and control characters) given by C<ch>. Thus, C<LATIN1_TO_NATIVE(66)> on
162EBCDIC platforms returns 194. These each represent the character C<"B"> on
163their respective platforms. On ASCII platforms no conversion is needed, so
164this macro expands to just its input, adding no time nor space requirements to
165the implementation.
166
167For conversion of code points potentially larger than will fit in a character,
168use L</UNI_TO_NATIVE>.
169
170=for apidoc Am|UV|NATIVE_TO_UNI|UV ch
171
172Returns the Unicode equivalent of the input native code point given by C<ch>.
173Thus, C<NATIVE_TO_UNI(195)> on EBCDIC platforms returns 67. These each
174represent the character C<"C"> on their respective platforms. On ASCII
175platforms no conversion is needed, so this macro expands to just its input,
176adding no time nor space requirements to the implementation.
177
178=for apidoc Am|UV|UNI_TO_NATIVE|UV ch
179
180Returns the native equivalent of the input Unicode code point given by C<ch>.
181Thus, C<UNI_TO_NATIVE(68)> on EBCDIC platforms returns 196. These each
182represent the character C<"D"> 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=cut
187*/
188
e7ae132e
KW
189#define NATIVE_TO_LATIN1(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) ((ch) | 0)))
190#define LATIN1_TO_NATIVE(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) ((ch) | 0)))
59a449d5
KW
191
192/* I8 is an intermediate version of UTF-8 used only in UTF-EBCDIC. We thus
193 * consider it to be identical to UTF-8 on ASCII platforms. Strictly speaking
194 * UTF-8 and UTF-EBCDIC are two different things, but we often conflate them
195 * because they are 8-bit encodings that serve the same purpose in Perl, and
196 * rarely do we need to distinguish them. The term "NATIVE_UTF8" applies to
197 * whichever one is applicable on the current platform */
cb15eeb2
KW
198#define NATIVE_UTF8_TO_I8(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) ((ch) | 0)))
199#define I8_TO_NATIVE_UTF8(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) ((ch) | 0)))
59a449d5 200
cb15eeb2
KW
201#define UNI_TO_NATIVE(ch) ((UV) ((ch) | 0))
202#define NATIVE_TO_UNI(ch) ((UV) ((ch) | 0))
d7578b48 203
877d9f0d 204/*
9041c2e3 205
a14e0a36
KW
206 The following table is from Unicode 3.2, plus the Perl extensions for above
207 U+10FFFF
877d9f0d 208
a14e0a36 209 Code Points 1st Byte 2nd Byte 3rd 4th 5th 6th 7th 8th-13th
877d9f0d 210
375122d7 211 U+0000..U+007F 00..7F
e1b711da 212 U+0080..U+07FF * C2..DF 80..BF
a14e0a36
KW
213 U+0800..U+0FFF E0 * A0..BF 80..BF
214 U+1000..U+CFFF E1..EC 80..BF 80..BF
215 U+D000..U+D7FF ED 80..9F 80..BF
216 U+D800..U+DFFF ED A0..BF 80..BF (surrogates)
217 U+E000..U+FFFF EE..EF 80..BF 80..BF
218 U+10000..U+3FFFF F0 * 90..BF 80..BF 80..BF
219 U+40000..U+FFFFF F1..F3 80..BF 80..BF 80..BF
220 U+100000..U+10FFFF F4 80..8F 80..BF 80..BF
221 Below are above-Unicode code points
222 U+110000..U+13FFFF F4 90..BF 80..BF 80..BF
223 U+110000..U+1FFFFF F5..F7 80..BF 80..BF 80..BF
224 U+200000..U+FFFFFF F8 * 88..BF 80..BF 80..BF 80..BF
225U+1000000..U+3FFFFFF F9..FB 80..BF 80..BF 80..BF 80..BF
226U+4000000..U+3FFFFFFF FC * 84..BF 80..BF 80..BF 80..BF 80..BF
227U+40000000..U+7FFFFFFF FD 80..BF 80..BF 80..BF 80..BF 80..BF
228U+80000000..U+FFFFFFFFF FE * 82..BF 80..BF 80..BF 80..BF 80..BF 80..BF
229U+1000000000.. FF 80..BF 80..BF 80..BF 80..BF 80..BF * 81..BF 80..BF
877d9f0d 230
e1b711da 231Note the gaps before several of the byte entries above marked by '*'. These are
37e2e78e
KW
232caused by legal UTF-8 avoiding non-shortest encodings: it is technically
233possible to UTF-8-encode a single code point in different ways, but that is
234explicitly forbidden, and the shortest possible encoding should always be used
15824458 235(and that is what Perl does). The non-shortest ones are called 'overlongs'.
8c007b5a 236
877d9f0d
JH
237 */
238
8c007b5a
JH
239/*
240 Another way to look at it, as bits:
241
b2635aa8 242 Code Points 1st Byte 2nd Byte 3rd Byte 4th Byte
8c007b5a 243
b2635aa8
KW
244 0aaa aaaa 0aaa aaaa
245 0000 0bbb bbaa aaaa 110b bbbb 10aa aaaa
246 cccc bbbb bbaa aaaa 1110 cccc 10bb bbbb 10aa aaaa
247 00 000d ddcc cccc bbbb bbaa aaaa 1111 0ddd 10cc cccc 10bb bbbb 10aa aaaa
8c007b5a
JH
248
249As you can see, the continuation bytes all begin with C<10>, and the
e1b711da 250leading bits of the start byte tell how many bytes there are in the
8c007b5a
JH
251encoded character.
252
df863e43
KW
253Perl's extended UTF-8 means we can have start bytes up through FF, though any
254beginning with FF yields a code point that is too large for 32-bit ASCII
255platforms. FF signals to use 13 bytes for the encoded character. This breaks
256the paradigm that the number of leading bits gives how many total bytes there
ab2e28c2 257are in the character. */
38953e5a 258
15824458
KW
259/* This is the number of low-order bits a continuation byte in a UTF-8 encoded
260 * sequence contributes to the specification of the code point. In the bit
261 * maps above, you see that the first 2 bits are a constant '10', leaving 6 of
262 * real information */
1d72bdf6 263#define UTF_ACCUMULATION_SHIFT 6
b2635aa8 264
fed423a5
KW
265/* ^? is defined to be DEL on ASCII systems. See the definition of toCTRL()
266 * for more */
267#define QUESTION_MARK_CTRL DEL_NATIVE
268
269/* Surrogates, non-character code points and above-Unicode code points are
270 * problematic in some contexts. This allows code that needs to check for
a3815e44 271 * those to quickly exclude the vast majority of code points it will
fed423a5 272 * encounter */
a6951642
KW
273#define isUTF8_POSSIBLY_PROBLEMATIC(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
274 (U8) c >= 0xED)
fed423a5 275
57ff5f59
KW
276#define UNICODE_IS_PERL_EXTENDED(uv) UNLIKELY((UV) (uv) > 0x7FFFFFFF)
277
fed423a5
KW
278#endif /* EBCDIC vs ASCII */
279
d7bcd45a
KW
280/* 2**UTF_ACCUMULATION_SHIFT - 1. This masks out all but the bits that carry
281 * real information in a continuation byte. This turns out to be 0x3F in
282 * UTF-8, 0x1F in UTF-EBCDIC. */
d223e1ea 283#define UTF_CONTINUATION_MASK ((U8) (nBIT_MASK(UTF_ACCUMULATION_SHIFT)))
fed423a5 284
9f3cfb7a
KW
285/* For use in UTF8_IS_CONTINUATION(). This turns out to be 0xC0 in UTF-8,
286 * E0 in UTF-EBCDIC */
287#define UTF_IS_CONTINUATION_MASK ((U8) (0xFF << UTF_ACCUMULATION_SHIFT))
288
38f458ff
KW
289/* This defines the bits that are to be in the continuation bytes of a
290 * multi-byte UTF-8 encoded character that mark it is a continuation byte.
291 * This turns out to be 0x80 in UTF-8, 0xA0 in UTF-EBCDIC. (khw doesn't know
292 * the underlying reason that B0 works here) */
293#define UTF_CONTINUATION_MARK (UTF_IS_CONTINUATION_MASK & 0xB0)
294
f4225fa0
KW
295/* Is the byte 'c' part of a multi-byte UTF8-8 encoded sequence, and not the
296 * first byte thereof? */
297#define UTF8_IS_CONTINUATION(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
298 (((NATIVE_UTF8_TO_I8(c) & UTF_IS_CONTINUATION_MASK) \
299 == UTF_CONTINUATION_MARK)))
300
2dc97505
KW
301/* Is the representation of the Unicode code point 'cp' the same regardless of
302 * being encoded in UTF-8 or not? This is a fundamental property of
303 * UTF-8,EBCDIC */
304#define OFFUNI_IS_INVARIANT(c) (((WIDEST_UTYPE)(c)) < UTF_CONTINUATION_MARK)
305
ab2e28c2
KW
306/*
307=for apidoc Am|bool|UVCHR_IS_INVARIANT|UV cp
308
309Evaluates to 1 if the representation of code point C<cp> is the same whether or
310not it is encoded in UTF-8; otherwise evaluates to 0. UTF-8 invariant
311characters can be copied as-is when converting to/from UTF-8, saving time.
312C<cp> is Unicode if above 255; otherwise is platform-native.
313
314=cut
315 */
316#define UVCHR_IS_INVARIANT(cp) (OFFUNI_IS_INVARIANT(NATIVE_TO_UNI(cp)))
317
7028aeba
KW
318/* Internal macro to be used only in this file to aid in constructing other
319 * publicly accessible macros.
320 * The number of bytes required to express this uv in UTF-8, for just those
321 * uv's requiring 2 through 6 bytes, as these are common to all platforms and
322 * word sizes. The number of bytes needed is given by the number of leading 1
323 * bits in the start byte. There are 32 start bytes that have 2 initial 1 bits
324 * (C0-DF); there are 16 that have 3 initial 1 bits (E0-EF); 8 that have 4
325 * initial 1 bits (F0-F8); 4 that have 5 initial 1 bits (F9-FB), and 2 that
326 * have 6 initial 1 bits (FC-FD). The largest number a string of n bytes can
327 * represent is (the number of possible start bytes for 'n')
328 * * (the number of possiblities for each start byte
329 * The latter in turn is
330 * 2 ** ( (how many continuation bytes there are)
331 * * (the number of bits of information each
332 * continuation byte holds))
333 *
334 * If we were on a platform where we could use a fast find first set bit
335 * instruction (or count leading zeros instruction) this could be replaced by
336 * using that to find the log2 of the uv, and divide that by the number of bits
337 * of information in each continuation byte, adjusting for large cases and how
338 * much information is in a start byte for that length */
72164d3a 339#define __COMMON_UNI_SKIP(uv) \
7028aeba
KW
340 (UV) (uv) < (32 * (1U << ( UTF_ACCUMULATION_SHIFT))) ? 2 : \
341 (UV) (uv) < (16 * (1U << (2 * UTF_ACCUMULATION_SHIFT))) ? 3 : \
342 (UV) (uv) < ( 8 * (1U << (3 * UTF_ACCUMULATION_SHIFT))) ? 4 : \
343 (UV) (uv) < ( 4 * (1U << (4 * UTF_ACCUMULATION_SHIFT))) ? 5 : \
344 (UV) (uv) < ( 2 * (1U << (5 * UTF_ACCUMULATION_SHIFT))) ? 6 :
72164d3a
KW
345
346/* Internal macro to be used only in this file.
347 * This adds to __COMMON_UNI_SKIP the details at this platform's upper range.
fed423a5 348 * For any-sized EBCDIC platforms, or 64-bit ASCII ones, we need one more test
72164d3a
KW
349 * to see if just 7 bytes is needed, or if the maximum is needed. For 32-bit
350 * ASCII platforms, everything is representable by 7 bytes */
fed423a5 351#if defined(UV_IS_QUAD) || defined(EBCDIC)
72164d3a 352# define __BASE_UNI_SKIP(uv) (__COMMON_UNI_SKIP(uv) \
8974941d
KW
353 LIKELY((UV) (uv) < ((UV) 1U << (6 * UTF_ACCUMULATION_SHIFT))) \
354 ? 7 \
355 : UTF8_MAXBYTES)
1d68d6cd 356#else
72164d3a 357# define __BASE_UNI_SKIP(uv) (__COMMON_UNI_SKIP(uv) 7)
1d68d6cd
SC
358#endif
359
fed423a5
KW
360/* The next two macros use the base macro defined above, and add in the tests
361 * at the low-end of the range, for just 1 byte, yielding complete macros,
362 * publicly accessible. */
363
364/* Input is a true Unicode (not-native) code point */
365#define OFFUNISKIP(uv) (OFFUNI_IS_INVARIANT(uv) ? 1 : __BASE_UNI_SKIP(uv))
2084b489 366
5352a763
KW
367/*
368
369=for apidoc Am|STRLEN|UVCHR_SKIP|UV cp
370returns the number of bytes required to represent the code point C<cp> when
371encoded as UTF-8. C<cp> is a native (ASCII or EBCDIC) code point if less than
372255; a Unicode code point otherwise.
373
374=cut
375 */
fdb6583d 376#define UVCHR_SKIP(uv) ( UVCHR_IS_INVARIANT(uv) ? 1 : __BASE_UNI_SKIP(uv))
5352a763 377
4bab39bc
KW
378#define UTF_MIN_START_BYTE \
379 ((UTF_CONTINUATION_MARK >> UTF_ACCUMULATION_SHIFT) | UTF_START_MARK(2))
380
381/* Is the byte 'c' the first byte of a multi-byte UTF8-8 encoded sequence?
59645eb1 382 * This excludes invariants (they are single-byte). It also excludes the
4bab39bc 383 * illegal overlong sequences that begin with C0 and C1 on ASCII platforms, and
59645eb1
KW
384 * C0-C4 I8 start bytes on EBCDIC ones. On EBCDIC E0 can't start a
385 * non-overlong sequence, so we define a base macro and for those platforms,
386 * extend it to also exclude E0 */
387#define UTF8_IS_START_base(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
4bab39bc 388 (NATIVE_UTF8_TO_I8(c) >= UTF_MIN_START_BYTE))
59645eb1
KW
389#ifdef EBCDIC
390# define UTF8_IS_START(c) \
391 (UTF8_IS_START_base(c) && (c) != I8_TO_NATIVE_UTF8(0xE0))
392#else
393# define UTF8_IS_START(c) UTF8_IS_START_base(c)
394#endif
4bab39bc 395
1df63428
KW
396#define UTF_MIN_ABOVE_LATIN1_BYTE \
397 ((0x100 >> UTF_ACCUMULATION_SHIFT) | UTF_START_MARK(2))
398
399/* Is the UTF8-encoded byte 'c' the first byte of a sequence of bytes that
400 * represent a code point > 255? */
401#define UTF8_IS_ABOVE_LATIN1(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
402 (NATIVE_UTF8_TO_I8(c) >= UTF_MIN_ABOVE_LATIN1_BYTE))
403
7c88d61e
KW
404/* Is the UTF8-encoded byte 'c' the first byte of a two byte sequence? Use
405 * UTF8_IS_NEXT_CHAR_DOWNGRADEABLE() instead if the input isn't known to
406 * be well-formed. */
407#define UTF8_IS_DOWNGRADEABLE_START(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
408 inRANGE(NATIVE_UTF8_TO_I8(c), \
409 UTF_MIN_START_BYTE, UTF_MIN_ABOVE_LATIN1_BYTE - 1))
410
b651802e
KW
411/* The largest code point representable by two UTF-8 bytes on this platform.
412 * As explained in the comments for __COMMON_UNI_SKIP, 32 start bytes with
fed423a5 413 * UTF_ACCUMULATION_SHIFT bits of information each */
aa206fb7
KW
414#define MAX_UTF8_TWO_BYTE (32 * (1U << UTF_ACCUMULATION_SHIFT) - 1)
415
b651802e
KW
416/* The largest code point representable by two UTF-8 bytes on any platform that
417 * Perl runs on. This value is constrained by EBCDIC which has 5 bits per
418 * continuation byte */
924b0bfd 419#define MAX_PORTABLE_UTF8_TWO_BYTE (32 * nBIT_UMAX(5))
aa206fb7 420
f2c50040
KW
421/*
422
423=for apidoc AmnU|STRLEN|UTF8_MAXBYTES_CASE
424
425The maximum number of UTF-8 bytes a single Unicode character can
426uppercase/lowercase/titlecase/fold into.
427
428=cut
429
430 * Unicode guarantees that the maximum expansion is UTF8_MAX_FOLD_CHAR_EXPAND
431 * characters, but any above-Unicode code point will fold to itself, so we only
432 * have to look at the expansion of the maximum Unicode code point. But this
433 * number may be less than the space occupied by a very large code point under
434 * Perl's extended UTF-8. We have to make it large enough to fit any single
435 * character. (It turns out that ASCII and EBCDIC differ in which is larger)
436 *
437=cut
438*/
c03c0950 439#define UTF8_MAXBYTES_CASE \
ae9a9edb 440 MAX(UTF8_MAXBYTES, UTF8_MAX_FOLD_CHAR_EXPAND * OFFUNISKIP(0x10FFFF))
c03c0950 441
d06134e5
KW
442/* Rest of these are attributes of Unicode and perl's internals rather than the
443 * encoding, or happen to be the same in both ASCII and EBCDIC (at least at
444 * this level; the macros that some of these call may have different
445 * definitions in the two encodings */
446
59a449d5
KW
447/* In domain restricted to ASCII, these may make more sense to the reader than
448 * the ones with Latin1 in the name */
449#define NATIVE_TO_ASCII(ch) NATIVE_TO_LATIN1(ch)
450#define ASCII_TO_NATIVE(ch) LATIN1_TO_NATIVE(ch)
451
452/* More or less misleadingly-named defines, retained for back compat */
453#define NATIVE_TO_UTF(ch) NATIVE_UTF8_TO_I8(ch)
454#define NATIVE_TO_I8(ch) NATIVE_UTF8_TO_I8(ch)
455#define UTF_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch)
456#define I8_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch)
457#define NATIVE8_TO_UNI(ch) NATIVE_TO_LATIN1(ch)
d06134e5 458
c0236afe 459/* This defines the 1-bits that are to be in the first byte of a multi-byte
97d0ceda
KW
460 * UTF-8 encoded character that mark it as a start byte and give the number of
461 * bytes that comprise the character. 'len' is the number of bytes in the
462 * multi-byte sequence. */
8974941d
KW
463#define UTF_START_MARK(len) (UNLIKELY((len) > 7) \
464 ? 0xFF \
465 : ((U8) (0xFE << (7-(len)))))
c0236afe
KW
466
467/* Masks out the initial one bits in a start byte, leaving the real data ones.
468 * Doesn't work on an invariant byte. 'len' is the number of bytes in the
469 * multi-byte sequence that comprises the character. */
8974941d 470#define UTF_START_MASK(len) (UNLIKELY((len) >= 7) ? 0x00 : (0x1F >> ((len)-2)))
c0236afe 471
537124e4
KW
472/* Adds a UTF8 continuation byte 'new' of information to a running total code
473 * point 'old' of all the continuation bytes so far. This is designed to be
155d2738
KW
474 * used in a loop to convert from UTF-8 to the code point represented. Note
475 * that this is asymmetric on EBCDIC platforms, in that the 'new' parameter is
476 * the UTF-EBCDIC byte, whereas the 'old' parameter is a Unicode (not EBCDIC)
477 * code point in process of being generated */
a6951642
KW
478#define UTF8_ACCUMULATE(old, new) (__ASSERT_(FITS_IN_8_BITS(new)) \
479 ((old) << UTF_ACCUMULATION_SHIFT) \
009097b1 480 | ((NATIVE_UTF8_TO_I8(new)) \
155d2738 481 & UTF_CONTINUATION_MASK))
d06134e5 482
4ab10950 483/* This works in the face of malformed UTF-8. */
4e1ed312
KW
484#define UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(s, e) \
485 ( UTF8_IS_DOWNGRADEABLE_START(*(s)) \
486 && ( (e) - (s) > 1) \
487 && UTF8_IS_CONTINUATION(*((s)+1)))
4ab10950 488
5aaebcb3 489/* Number of bytes a code point occupies in UTF-8. */
5352a763 490#define NATIVE_SKIP(uv) UVCHR_SKIP(uv)
bd18bd40 491
5aaebcb3
KW
492/* Most code which says UNISKIP is really thinking in terms of native code
493 * points (0-255) plus all those beyond. This is an imprecise term, but having
2accb712 494 * it means existing code continues to work. For precision, use UVCHR_SKIP,
5352a763
KW
495 * NATIVE_SKIP, or OFFUNISKIP */
496#define UNISKIP(uv) UVCHR_SKIP(uv)
5aaebcb3 497
3c0792e4
KW
498/* Longer, but more accurate name */
499#define UTF8_IS_ABOVE_LATIN1_START(c) UTF8_IS_ABOVE_LATIN1(c)
500
a62b247b
KW
501/* Convert a UTF-8 variant Latin1 character to a native code point value.
502 * Needs just one iteration of accumulate. Should be used only if it is known
503 * that the code point is < 256, and is not UTF-8 invariant. Use the slower
504 * but more general TWO_BYTE_UTF8_TO_NATIVE() which handles any code point
505 * representable by two bytes (which turns out to be up through
506 * MAX_PORTABLE_UTF8_TWO_BYTE). The two parameters are:
507 * HI: a downgradable start byte;
508 * LO: continuation.
509 * */
510#define EIGHT_BIT_UTF8_TO_NATIVE(HI, LO) \
511 ( __ASSERT_(UTF8_IS_DOWNGRADEABLE_START(HI)) \
512 __ASSERT_(UTF8_IS_CONTINUATION(LO)) \
513 LATIN1_TO_NATIVE(UTF8_ACCUMULATE(( \
514 NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), (LO))))
515
94bb8c36 516/* Convert a two (not one) byte utf8 character to a native code point value.
2950f2a7
KW
517 * Needs just one iteration of accumulate. Should not be used unless it is
518 * known that the two bytes are legal: 1) two-byte start, and 2) continuation.
519 * Note that the result can be larger than 255 if the input character is not
520 * downgradable */
94bb8c36 521#define TWO_BYTE_UTF8_TO_NATIVE(HI, LO) \
a6951642
KW
522 (__ASSERT_(FITS_IN_8_BITS(HI)) \
523 __ASSERT_(FITS_IN_8_BITS(LO)) \
524 __ASSERT_(PL_utf8skip[HI] == 2) \
525 __ASSERT_(UTF8_IS_CONTINUATION(LO)) \
94bb8c36 526 UNI_TO_NATIVE(UTF8_ACCUMULATE((NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), \
635e76f5 527 (LO))))
94bb8c36
KW
528
529/* Should never be used, and be deprecated */
530#define TWO_BYTE_UTF8_TO_UNI(HI, LO) NATIVE_TO_UNI(TWO_BYTE_UTF8_TO_NATIVE(HI, LO))
2950f2a7 531
bd18bd40
KW
532/*
533
534=for apidoc Am|STRLEN|UTF8SKIP|char* s
ee0ff0f5
KW
535returns the number of bytes a non-malformed UTF-8 encoded character whose first
536(perhaps only) byte is pointed to by C<s>.
537
538If there is a possibility of malformed input, use instead:
539
540=over
541
542=item L</C<UTF8_SAFE_SKIP>> if you know the maximum ending pointer in the
543buffer pointed to by C<s>; or
544
545=item L</C<UTF8_CHK_SKIP>> if you don't know it.
546
547=back
548
549It is better to restructure your code so the end pointer is passed down so that
550you know what it actually is at the point of this call, but if that isn't
551possible, L</C<UTF8_CHK_SKIP>> can minimize the chance of accessing beyond the end
552of the input buffer.
bd18bd40
KW
553
554=cut
555 */
2a70536e 556#define UTF8SKIP(s) PL_utf8skip[*(const U8*)(s)]
a281f16c
KW
557
558/*
559=for apidoc Am|STRLEN|UTF8_SKIP|char* s
560This is a synonym for L</C<UTF8SKIP>>
561
562=cut
563*/
564
2a70536e 565#define UTF8_SKIP(s) UTF8SKIP(s)
d06134e5 566
85fcc8f2 567/*
ee0ff0f5
KW
568=for apidoc Am|STRLEN|UTF8_CHK_SKIP|char* s
569
570This is a safer version of L</C<UTF8SKIP>>, but still not as safe as
571L</C<UTF8_SAFE_SKIP>>. This version doesn't blindly assume that the input
572string pointed to by C<s> is well-formed, but verifies that there isn't a NUL
573terminating character before the expected end of the next character in C<s>.
574The length C<UTF8_CHK_SKIP> returns stops just before any such NUL.
575
576Perl tends to add NULs, as an insurance policy, after the end of strings in
577SV's, so it is likely that using this macro will prevent inadvertent reading
578beyond the end of the input buffer, even if it is malformed UTF-8.
579
580This macro is intended to be used by XS modules where the inputs could be
581malformed, and it isn't feasible to restructure to use the safer
582L</C<UTF8_SAFE_SKIP>>, for example when interfacing with a C library.
583
584=cut
585*/
586
587#define UTF8_CHK_SKIP(s) \
8974941d 588 (UNLIKELY(s[0] == '\0') ? 1 : MIN(UTF8SKIP(s), \
f87d8789 589 my_strnlen((char *) (s), UTF8SKIP(s))))
ee0ff0f5 590/*
85fcc8f2
KW
591
592=for apidoc Am|STRLEN|UTF8_SAFE_SKIP|char* s|char* e
45671da2
KW
593returns 0 if S<C<s E<gt>= e>>; otherwise returns the number of bytes in the
594UTF-8 encoded character whose first byte is pointed to by C<s>. But it never
595returns beyond C<e>. On DEBUGGING builds, it asserts that S<C<s E<lt>= e>>.
85fcc8f2
KW
596
597=cut
598 */
45671da2 599#define UTF8_SAFE_SKIP(s, e) (__ASSERT_((e) >= (s)) \
8974941d 600 UNLIKELY(((e) - (s)) <= 0) \
45671da2
KW
601 ? 0 \
602 : MIN(((e) - (s)), UTF8_SKIP(s)))
85fcc8f2 603
2d1545e5
KW
604/* Most code that says 'UNI_' really means the native value for code points up
605 * through 255 */
606#define UNI_IS_INVARIANT(cp) UVCHR_IS_INVARIANT(cp)
607
c2b32798
KW
608/*
609=for apidoc Am|bool|UTF8_IS_INVARIANT|char c
610
611Evaluates to 1 if the byte C<c> represents the same character when encoded in
612UTF-8 as when not; otherwise evaluates to 0. UTF-8 invariant characters can be
613copied as-is when converting to/from UTF-8, saving time.
614
615In spite of the name, this macro gives the correct result if the input string
616from which C<c> comes is not encoded in UTF-8.
617
618See C<L</UVCHR_IS_INVARIANT>> for checking if a UV is invariant.
619
620=cut
621
622The reason it works on both UTF-8 encoded strings and non-UTF-8 encoded, is
623that it returns TRUE in each for the exact same set of bit patterns. It is
624valid on a subset of what UVCHR_IS_INVARIANT is valid on, so can just use that;
625and the compiler should optimize out anything extraneous given the
626implementation of the latter. The |0 makes sure this isn't mistakenly called
627with a ptr argument.
628*/
5c06326b 629#define UTF8_IS_INVARIANT(c) UVCHR_IS_INVARIANT((c) | 0)
5fc230f1
KW
630
631/* Like the above, but its name implies a non-UTF8 input, which as the comments
632 * above show, doesn't matter as to its implementation */
38953e5a 633#define NATIVE_BYTE_IS_INVARIANT(c) UVCHR_IS_INVARIANT(c)
d06134e5 634
2c03e801
KW
635/* Misleadingly named: is the UTF8-encoded byte 'c' part of a variant sequence
636 * in UTF-8? This is the inverse of UTF8_IS_INVARIANT. */
637#define UTF8_IS_CONTINUED(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
638 (! UTF8_IS_INVARIANT(c)))
639
48ccf5e1
KW
640/* The macros in the next 4 sets are used to generate the two utf8 or utfebcdic
641 * bytes from an ordinal that is known to fit into exactly two (not one) bytes;
642 * it must be less than 0x3FF to work across both encodings. */
643
644/* These two are helper macros for the other three sets, and should not be used
645 * directly anywhere else. 'translate_function' is either NATIVE_TO_LATIN1
1ff3baa2
KW
646 * (which works for code points up through 0xFF) or NATIVE_TO_UNI which works
647 * for any code point */
48ccf5e1 648#define __BASE_TWO_BYTE_HI(c, translate_function) \
2863dafa 649 (__ASSERT_(! UVCHR_IS_INVARIANT(c)) \
48ccf5e1 650 I8_TO_NATIVE_UTF8((translate_function(c) >> UTF_ACCUMULATION_SHIFT) \
2863dafa 651 | UTF_START_MARK(2)))
48ccf5e1 652#define __BASE_TWO_BYTE_LO(c, translate_function) \
2863dafa 653 (__ASSERT_(! UVCHR_IS_INVARIANT(c)) \
48ccf5e1 654 I8_TO_NATIVE_UTF8((translate_function(c) & UTF_CONTINUATION_MASK) \
2863dafa 655 | UTF_CONTINUATION_MARK))
48ccf5e1 656
48ccf5e1
KW
657/* The next two macros should not be used. They were designed to be usable as
658 * the case label of a switch statement, but this doesn't work for EBCDIC. Use
9d0d3a03 659 * regen/unicode_constants.pl instead */
48ccf5e1
KW
660#define UTF8_TWO_BYTE_HI_nocast(c) __BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI)
661#define UTF8_TWO_BYTE_LO_nocast(c) __BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI)
662
663/* The next two macros are used when the source should be a single byte
664 * character; checked for under DEBUGGING */
665#define UTF8_EIGHT_BIT_HI(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
4c8cd605 666 ( __BASE_TWO_BYTE_HI(c, NATIVE_TO_LATIN1)))
48ccf5e1 667#define UTF8_EIGHT_BIT_LO(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
4c8cd605 668 (__BASE_TWO_BYTE_LO(c, NATIVE_TO_LATIN1)))
48ccf5e1
KW
669
670/* These final two macros in the series are used when the source can be any
671 * code point whose UTF-8 is known to occupy 2 bytes; they are less efficient
672 * than the EIGHT_BIT versions on EBCDIC platforms. We use the logical '~'
673 * operator instead of "<=" to avoid getting compiler warnings.
d52b8576 674 * MAX_UTF8_TWO_BYTE should be exactly all one bits in the lower few
48ccf5e1
KW
675 * places, so the ~ works */
676#define UTF8_TWO_BYTE_HI(c) \
677 (__ASSERT_((sizeof(c) == 1) \
d52b8576 678 || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \
4c8cd605 679 (__BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI)))
48ccf5e1
KW
680#define UTF8_TWO_BYTE_LO(c) \
681 (__ASSERT_((sizeof(c) == 1) \
d52b8576 682 || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \
4c8cd605 683 (__BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI)))
d06134e5 684
e7214ce8
KW
685/* This is illegal in any well-formed UTF-8 in both EBCDIC and ASCII
686 * as it is only in overlongs. */
687#define ILLEGAL_UTF8_BYTE I8_TO_NATIVE_UTF8(0xC1)
688
7e2040f0 689/*
e3036cf4 690 * 'UTF' is whether or not p is encoded in UTF8. The names 'foo_lazy_if' stem
20df05f4
KW
691 * from an earlier version of these macros in which they didn't call the
692 * foo_utf8() macros (i.e. were 'lazy') unless they decided that *p is the
693 * beginning of a utf8 character. Now that foo_utf8() determines that itself,
694 * no need to do it again here
7e2040f0 695 */
da8c1a98
KW
696#define isIDFIRST_lazy_if_safe(p, e, UTF) \
697 ((IN_BYTES || !UTF) \
698 ? isIDFIRST(*(p)) \
699 : isIDFIRST_utf8_safe(p, e))
da8c1a98
KW
700#define isWORDCHAR_lazy_if_safe(p, e, UTF) \
701 ((IN_BYTES || !UTF) \
702 ? isWORDCHAR(*(p)) \
703 : isWORDCHAR_utf8_safe((U8 *) p, (U8 *) e))
4c1d9526 704#define isALNUM_lazy_if_safe(p, e, UTF) isWORDCHAR_lazy_if_safe(p, e, UTF)
da8c1a98 705
89ebb4a3
JH
706#define UTF8_MAXLEN UTF8_MAXBYTES
707
8cb75cc8
KW
708/* A Unicode character can fold to up to 3 characters */
709#define UTF8_MAX_FOLD_CHAR_EXPAND 3
710
d3481830 711#define IN_BYTES UNLIKELY(CopHINTS_get(PL_curcop) & HINT_BYTES)
bd18bd40
KW
712
713/*
714
715=for apidoc Am|bool|DO_UTF8|SV* sv
716Returns a bool giving whether or not the PV in C<sv> is to be treated as being
717encoded in UTF-8.
718
719You should use this I<after> a call to C<SvPV()> or one of its variants, in
720case any call to string overloading updates the internal UTF-8 encoding flag.
721
722=cut
723*/
0064a8a9 724#define DO_UTF8(sv) (SvUTF8(sv) && !IN_BYTES)
1ff3baa2
KW
725
726/* Should all strings be treated as Unicode, and not just UTF-8 encoded ones?
727 * Is so within 'feature unicode_strings' or 'locale :not_characters', and not
728 * within 'use bytes'. UTF-8 locales are not tested for here, but perhaps
729 * could be */
70844984
KW
730#define IN_UNI_8_BIT \
731 (( ( (CopHINTS_get(PL_curcop) & HINT_UNI_8_BIT)) \
732 || ( CopHINTS_get(PL_curcop) & HINT_LOCALE_PARTIAL \
733 /* -1 below is for :not_characters */ \
734 && _is_in_locale_category(FALSE, -1))) \
735 && (! IN_BYTES))
b36bf33f 736
1d72bdf6 737
c76687c5 738#define UTF8_ALLOW_EMPTY 0x0001 /* Allow a zero length string */
2b5e7bc2 739#define UTF8_GOT_EMPTY UTF8_ALLOW_EMPTY
c76687c5
KW
740
741/* Allow first byte to be a continuation byte */
1d72bdf6 742#define UTF8_ALLOW_CONTINUATION 0x0002
2b5e7bc2 743#define UTF8_GOT_CONTINUATION UTF8_ALLOW_CONTINUATION
c76687c5 744
cd01d3b1 745/* Unexpected non-continuation byte */
1d72bdf6 746#define UTF8_ALLOW_NON_CONTINUATION 0x0004
2b5e7bc2 747#define UTF8_GOT_NON_CONTINUATION UTF8_ALLOW_NON_CONTINUATION
949cf498
KW
748
749/* expecting more bytes than were available in the string */
750#define UTF8_ALLOW_SHORT 0x0008
2b5e7bc2 751#define UTF8_GOT_SHORT UTF8_ALLOW_SHORT
949cf498 752
94953955
KW
753/* Overlong sequence; i.e., the code point can be specified in fewer bytes.
754 * First one will convert the overlong to the REPLACEMENT CHARACTER; second
755 * will return what the overlong evaluates to */
949cf498 756#define UTF8_ALLOW_LONG 0x0010
94953955 757#define UTF8_ALLOW_LONG_AND_ITS_VALUE (UTF8_ALLOW_LONG|0x0020)
2b5e7bc2
KW
758#define UTF8_GOT_LONG UTF8_ALLOW_LONG
759
d60baaa7
KW
760#define UTF8_ALLOW_OVERFLOW 0x0080
761#define UTF8_GOT_OVERFLOW UTF8_ALLOW_OVERFLOW
949cf498 762
f180b292 763#define UTF8_DISALLOW_SURROGATE 0x0100 /* Unicode surrogates */
2b5e7bc2 764#define UTF8_GOT_SURROGATE UTF8_DISALLOW_SURROGATE
f180b292 765#define UTF8_WARN_SURROGATE 0x0200
949cf498 766
c4e96019
KW
767/* Unicode non-character code points */
768#define UTF8_DISALLOW_NONCHAR 0x0400
2b5e7bc2 769#define UTF8_GOT_NONCHAR UTF8_DISALLOW_NONCHAR
c4e96019 770#define UTF8_WARN_NONCHAR 0x0800
949cf498 771
c4e96019
KW
772/* Super-set of Unicode: code points above the legal max */
773#define UTF8_DISALLOW_SUPER 0x1000
2b5e7bc2 774#define UTF8_GOT_SUPER UTF8_DISALLOW_SUPER
c4e96019
KW
775#define UTF8_WARN_SUPER 0x2000
776
777/* The original UTF-8 standard did not define UTF-8 with start bytes of 0xFE or
778 * 0xFF, though UTF-EBCDIC did. This allowed both versions to represent code
779 * points up to 2 ** 31 - 1. Perl extends UTF-8 so that 0xFE and 0xFF are
780 * usable on ASCII platforms, and 0xFF means something different than
781 * UTF-EBCDIC defines. These changes allow code points of 64 bits (actually
782 * somewhat more) to be represented on both platforms. But these are Perl
783 * extensions, and not likely to be interchangeable with other languages. Note
784 * that on ASCII platforms, FE overflows a signed 32-bit word, and FF an
785 * unsigned one. */
d044b7a7
KW
786#define UTF8_DISALLOW_PERL_EXTENDED 0x4000
787#define UTF8_GOT_PERL_EXTENDED UTF8_DISALLOW_PERL_EXTENDED
788#define UTF8_WARN_PERL_EXTENDED 0x8000
d35f2ca5 789
57ff5f59
KW
790/* For back compat, these old names are misleading for overlongs and
791 * UTF_EBCDIC. */
d044b7a7
KW
792#define UTF8_DISALLOW_ABOVE_31_BIT UTF8_DISALLOW_PERL_EXTENDED
793#define UTF8_GOT_ABOVE_31_BIT UTF8_GOT_PERL_EXTENDED
794#define UTF8_WARN_ABOVE_31_BIT UTF8_WARN_PERL_EXTENDED
795#define UTF8_DISALLOW_FE_FF UTF8_DISALLOW_PERL_EXTENDED
796#define UTF8_WARN_FE_FF UTF8_WARN_PERL_EXTENDED
949cf498 797
f180b292 798#define UTF8_CHECK_ONLY 0x10000
99a765e9 799#define _UTF8_NO_CONFIDENCE_IN_CURLEN 0x20000 /* Internal core use only */
949cf498
KW
800
801/* For backwards source compatibility. They do nothing, as the default now
802 * includes what they used to mean. The first one's meaning was to allow the
803 * just the single non-character 0xFFFF */
804#define UTF8_ALLOW_FFFF 0
c825ef8c 805#define UTF8_ALLOW_FE_FF 0
949cf498
KW
806#define UTF8_ALLOW_SURROGATE 0
807
ecc1615f
KW
808/* C9 refers to Unicode Corrigendum #9: allows but discourages non-chars */
809#define UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE \
810 (UTF8_DISALLOW_SUPER|UTF8_DISALLOW_SURROGATE)
811#define UTF8_WARN_ILLEGAL_C9_INTERCHANGE (UTF8_WARN_SUPER|UTF8_WARN_SURROGATE)
812
d35f2ca5 813#define UTF8_DISALLOW_ILLEGAL_INTERCHANGE \
ecc1615f 814 (UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE|UTF8_DISALLOW_NONCHAR)
949cf498 815#define UTF8_WARN_ILLEGAL_INTERCHANGE \
ecc1615f
KW
816 (UTF8_WARN_ILLEGAL_C9_INTERCHANGE|UTF8_WARN_NONCHAR)
817
0eb3d6a0
KW
818/* This is typically used for code that processes UTF-8 input and doesn't want
819 * to have to deal with any malformations that might be present. All such will
820 * be safely replaced by the REPLACEMENT CHARACTER, unless other flags
821 * overriding this are also present. */
2d532c27
KW
822#define UTF8_ALLOW_ANY ( UTF8_ALLOW_CONTINUATION \
823 |UTF8_ALLOW_NON_CONTINUATION \
824 |UTF8_ALLOW_SHORT \
d60baaa7
KW
825 |UTF8_ALLOW_LONG \
826 |UTF8_ALLOW_OVERFLOW)
2d532c27
KW
827
828/* Accept any Perl-extended UTF-8 that evaluates to any UV on the platform, but
cd01d3b1 829 * not any malformed. This is the default. */
2d532c27
KW
830#define UTF8_ALLOW_ANYUV 0
831#define UTF8_ALLOW_DEFAULT UTF8_ALLOW_ANYUV
1d72bdf6 832
89d986df
KW
833/*
834=for apidoc Am|bool|UTF8_IS_SURROGATE|const U8 *s|const U8 *e
835
836Evaluates to non-zero if the first few bytes of the string starting at C<s> and
837looking no further than S<C<e - 1>> are well-formed UTF-8 that represents one
838of the Unicode surrogate code points; otherwise it evaluates to 0. If
839non-zero, the value gives how many bytes starting at C<s> comprise the code
840point's representation.
841
842=cut
843 */
844#define UTF8_IS_SURROGATE(s, e) is_SURROGATE_utf8_safe(s, e)
845
846
847#define UTF8_IS_REPLACEMENT(s, send) is_REPLACEMENT_utf8_safe(s,send)
848
285aa1f0 849#define MAX_LEGAL_CP ((UV)IV_MAX)
40606899 850
89d986df
KW
851/*
852=for apidoc Am|bool|UTF8_IS_SUPER|const U8 *s|const U8 *e
853
854Recall that Perl recognizes an extension to UTF-8 that can encode code
855points larger than the ones defined by Unicode, which are 0..0x10FFFF.
856
857This macro evaluates to non-zero if the first few bytes of the string starting
858at C<s> and looking no further than S<C<e - 1>> are from this UTF-8 extension;
859otherwise it evaluates to 0. If non-zero, the value gives how many bytes
860starting at C<s> comprise the code point's representation.
861
8620 is returned if the bytes are not well-formed extended UTF-8, or if they
863represent a code point that cannot fit in a UV on the current platform. Hence
864this macro can give different results when run on a 64-bit word machine than on
865one with a 32-bit word size.
0c58a72b 866
891fd405 867Note that it is illegal to have code points that are larger than what can
89d986df 868fit in an IV on the current machine.
7131f24d 869
89d986df
KW
870=cut
871
872 * ASCII EBCDIC I8
7131f24d
KW
873 * U+10FFFF: \xF4\x8F\xBF\xBF \xF9\xA1\xBF\xBF\xBF max legal Unicode
874 * U+110000: \xF4\x90\x80\x80 \xF9\xA2\xA0\xA0\xA0
875 * U+110001: \xF4\x90\x80\x81 \xF9\xA2\xA0\xA0\xA1
89d986df
KW
876 */
877#ifdef EBCDIC
a14e0a36 878# define UTF8_IS_SUPER(s, e) \
8974941d
KW
879 (( ((e) > (s) + 4) \
880 && (NATIVE_UTF8_TO_I8(*(s)) >= 0xF9) \
881 && UNLIKELY( NATIVE_UTF8_TO_I8(*(s)) > 0xF9 \
882 || (NATIVE_UTF8_TO_I8(*((s) + 1)) >= 0xA2)) \
883 && LIKELY((s) + UTF8SKIP(s) <= (e))) \
884 ? is_utf8_char_helper(s, s + UTF8SKIP(s), 0) : 0
7131f24d 885#else
a14e0a36 886# define UTF8_IS_SUPER(s, e) \
8974941d 887 (( ((e) > (s) + 3) \
89d986df 888 && (*(U8*) (s)) >= 0xF4 \
8974941d
KW
889 && (UNLIKELY( ((*(U8*) (s)) > 0xF4) \
890 || (*((U8*) (s) + 1) >= 0x90))) \
89d986df 891 && LIKELY((s) + UTF8SKIP(s) <= (e))) \
1376b35c 892 ? is_utf8_char_helper(s, s + UTF8SKIP(s), 0) : 0)
7131f24d
KW
893#endif
894
b96a92fb
KW
895/* These are now machine generated, and the 'given' clause is no longer
896 * applicable */
0c58a72b 897#define UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s, e) \
89d986df
KW
898 cBOOL(is_NONCHAR_utf8_safe(s,e))
899
900/*
901=for apidoc Am|bool|UTF8_IS_NONCHAR|const U8 *s|const U8 *e
902
903Evaluates to non-zero if the first few bytes of the string starting at C<s> and
904looking no further than S<C<e - 1>> are well-formed UTF-8 that represents one
905of the Unicode non-character code points; otherwise it evaluates to 0. If
906non-zero, the value gives how many bytes starting at C<s> comprise the code
907point's representation.
908
d296fe14
KW
909=for apidoc AmnU|UV|UNICODE_REPLACEMENT
910
911Evaluates to 0xFFFD, the code point of the Unicode REPLACEMENT CHARACTER
912
89d986df
KW
913=cut
914 */
0c58a72b
KW
915#define UTF8_IS_NONCHAR(s, e) \
916 UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s, e)
7131f24d 917
c867b360
JH
918#define UNICODE_SURROGATE_FIRST 0xD800
919#define UNICODE_SURROGATE_LAST 0xDFFF
920#define UNICODE_REPLACEMENT 0xFFFD
921#define UNICODE_BYTE_ORDER_MARK 0xFEFF
1d72bdf6 922
b851fbc1 923/* Though our UTF-8 encoding can go beyond this,
c76687c5 924 * let's be conservative and do as Unicode says. */
b851fbc1
JH
925#define PERL_UNICODE_MAX 0x10FFFF
926
d044b7a7
KW
927#define UNICODE_WARN_SURROGATE 0x0001 /* UTF-16 surrogates */
928#define UNICODE_WARN_NONCHAR 0x0002 /* Non-char code points */
929#define UNICODE_WARN_SUPER 0x0004 /* Above 0x10FFFF */
930#define UNICODE_WARN_PERL_EXTENDED 0x0008 /* Above 0x7FFF_FFFF */
931#define UNICODE_WARN_ABOVE_31_BIT UNICODE_WARN_PERL_EXTENDED
932#define UNICODE_DISALLOW_SURROGATE 0x0010
933#define UNICODE_DISALLOW_NONCHAR 0x0020
934#define UNICODE_DISALLOW_SUPER 0x0040
935#define UNICODE_DISALLOW_PERL_EXTENDED 0x0080
24b4c303
KW
936
937#ifdef PERL_CORE
938# define UNICODE_ALLOW_ABOVE_IV_MAX 0x0100
939#endif
d044b7a7 940#define UNICODE_DISALLOW_ABOVE_31_BIT UNICODE_DISALLOW_PERL_EXTENDED
33f38593
KW
941
942#define UNICODE_GOT_SURROGATE UNICODE_DISALLOW_SURROGATE
943#define UNICODE_GOT_NONCHAR UNICODE_DISALLOW_NONCHAR
944#define UNICODE_GOT_SUPER UNICODE_DISALLOW_SUPER
945#define UNICODE_GOT_PERL_EXTENDED UNICODE_DISALLOW_PERL_EXTENDED
946
ecc1615f
KW
947#define UNICODE_WARN_ILLEGAL_C9_INTERCHANGE \
948 (UNICODE_WARN_SURROGATE|UNICODE_WARN_SUPER)
bb88be5f 949#define UNICODE_WARN_ILLEGAL_INTERCHANGE \
ecc1615f
KW
950 (UNICODE_WARN_ILLEGAL_C9_INTERCHANGE|UNICODE_WARN_NONCHAR)
951#define UNICODE_DISALLOW_ILLEGAL_C9_INTERCHANGE \
952 (UNICODE_DISALLOW_SURROGATE|UNICODE_DISALLOW_SUPER)
bb88be5f 953#define UNICODE_DISALLOW_ILLEGAL_INTERCHANGE \
ecc1615f 954 (UNICODE_DISALLOW_ILLEGAL_C9_INTERCHANGE|UNICODE_DISALLOW_NONCHAR)
949cf498
KW
955
956/* For backward source compatibility, as are now the default */
957#define UNICODE_ALLOW_SURROGATE 0
958#define UNICODE_ALLOW_SUPER 0
959#define UNICODE_ALLOW_ANY 0
b851fbc1 960
2d6b3d38
KW
961/* This matches the 2048 code points between UNICODE_SURROGATE_FIRST (0xD800) and
962 * UNICODE_SURROGATE_LAST (0xDFFF) */
8974941d 963#define UNICODE_IS_SURROGATE(uv) UNLIKELY(((UV) (uv) & (~0xFFFF | 0xF800)) \
2d6b3d38
KW
964 == 0xD800)
965
8974941d
KW
966#define UNICODE_IS_REPLACEMENT(uv) UNLIKELY((UV) (uv) == UNICODE_REPLACEMENT)
967#define UNICODE_IS_BYTE_ORDER_MARK(uv) UNLIKELY((UV) (uv) \
968 == UNICODE_BYTE_ORDER_MARK)
c149ab20
KW
969
970/* Is 'uv' one of the 32 contiguous-range noncharacters? */
8974941d
KW
971#define UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv) UNLIKELY((UV) (uv) >= 0xFDD0 \
972 && (UV) (uv) <= 0xFDEF)
c149ab20
KW
973
974/* Is 'uv' one of the 34 plane-ending noncharacters 0xFFFE, 0xFFFF, 0x1FFFE,
975 * 0x1FFFF, ... 0x10FFFE, 0x10FFFF, given that we know that 'uv' is not above
976 * the Unicode legal max */
977#define UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv) \
8974941d 978 UNLIKELY(((UV) (uv) & 0xFFFE) == 0xFFFE)
c149ab20
KW
979
980#define UNICODE_IS_NONCHAR(uv) \
981 ( UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv) \
982 || ( LIKELY( ! UNICODE_IS_SUPER(uv)) \
983 && UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv)))
984
8974941d 985#define UNICODE_IS_SUPER(uv) UNLIKELY((UV) (uv) > PERL_UNICODE_MAX)
1d72bdf6 986
ec34087a
KW
987#define LATIN_SMALL_LETTER_SHARP_S LATIN_SMALL_LETTER_SHARP_S_NATIVE
988#define LATIN_SMALL_LETTER_Y_WITH_DIAERESIS \
989 LATIN_SMALL_LETTER_Y_WITH_DIAERESIS_NATIVE
990#define MICRO_SIGN MICRO_SIGN_NATIVE
991#define LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE \
992 LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE_NATIVE
993#define LATIN_SMALL_LETTER_A_WITH_RING_ABOVE \
994 LATIN_SMALL_LETTER_A_WITH_RING_ABOVE_NATIVE
09091399
JH
995#define UNICODE_GREEK_CAPITAL_LETTER_SIGMA 0x03A3
996#define UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA 0x03C2
997#define UNICODE_GREEK_SMALL_LETTER_SIGMA 0x03C3
9dcbe121 998#define GREEK_SMALL_LETTER_MU 0x03BC
9e682c18
KW
999#define GREEK_CAPITAL_LETTER_MU 0x039C /* Upper and title case
1000 of MICRON */
1001#define LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS 0x0178 /* Also is title case */
0766489e
KW
1002#ifdef LATIN_CAPITAL_LETTER_SHARP_S_UTF8
1003# define LATIN_CAPITAL_LETTER_SHARP_S 0x1E9E
1004#endif
74894415
KW
1005#define LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE 0x130
1006#define LATIN_SMALL_LETTER_DOTLESS_I 0x131
9e682c18 1007#define LATIN_SMALL_LETTER_LONG_S 0x017F
a9f50d33
KW
1008#define LATIN_SMALL_LIGATURE_LONG_S_T 0xFB05
1009#define LATIN_SMALL_LIGATURE_ST 0xFB06
9e682c18
KW
1010#define KELVIN_SIGN 0x212A
1011#define ANGSTROM_SIGN 0x212B
09091399 1012
9e55ce06 1013#define UNI_DISPLAY_ISPRINT 0x0001
c728cb41 1014#define UNI_DISPLAY_BACKSLASH 0x0002
daf6caf1
KW
1015#define UNI_DISPLAY_BACKSPACE 0x0004 /* Allow \b when also
1016 UNI_DISPLAY_BACKSLASH */
1017#define UNI_DISPLAY_QQ (UNI_DISPLAY_ISPRINT \
1018 |UNI_DISPLAY_BACKSLASH \
1019 |UNI_DISPLAY_BACKSPACE)
1020
1021/* Character classes could also allow \b, but not patterns in general */
c728cb41 1022#define UNI_DISPLAY_REGEX (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH)
9e55ce06 1023
e0ffa6d6 1024/* Should be removed; maybe deprecated, but not used in CPAN */
ebc501f0 1025#define SHARP_S_SKIP 2
3b0fc154 1026
3cedd9d9 1027#define is_utf8_char_buf(buf, buf_end) isUTF8_CHAR(buf, buf_end)
976c1b08
KW
1028#define bytes_from_utf8(s, lenp, is_utf8p) \
1029 bytes_from_utf8_loc(s, lenp, is_utf8p, 0)
3cedd9d9 1030
e23e8bc1
KW
1031/*
1032
25e3a4e0
KW
1033=for apidoc Am|STRLEN|isUTF8_CHAR_flags|const U8 *s|const U8 *e| const U32 flags
1034
1035Evaluates to non-zero if the first few bytes of the string starting at C<s> and
1036looking no further than S<C<e - 1>> are well-formed UTF-8, as extended by Perl,
1037that represents some code point, subject to the restrictions given by C<flags>;
1038otherwise it evaluates to 0. If non-zero, the value gives how many bytes
2717076a
KW
1039starting at C<s> comprise the code point's representation. Any bytes remaining
1040before C<e>, but beyond the ones needed to form the first code point in C<s>,
1041are not examined.
25e3a4e0
KW
1042
1043If C<flags> is 0, this gives the same results as C<L</isUTF8_CHAR>>;
1044if C<flags> is C<UTF8_DISALLOW_ILLEGAL_INTERCHANGE>, this gives the same results
1045as C<L</isSTRICT_UTF8_CHAR>>;
1046and if C<flags> is C<UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE>, this gives
1047the same results as C<L</isC9_STRICT_UTF8_CHAR>>.
1048Otherwise C<flags> may be any combination of the C<UTF8_DISALLOW_I<foo>> flags
1049understood by C<L</utf8n_to_uvchr>>, with the same meanings.
1050
1051The three alternative macros are for the most commonly needed validations; they
1052are likely to run somewhat faster than this more general one, as they can be
1053inlined into your code.
1054
9f2abfde
KW
1055Use L</is_utf8_string_flags>, L</is_utf8_string_loc_flags>, and
1056L</is_utf8_string_loclen_flags> to check entire strings.
1057
25e3a4e0
KW
1058=cut
1059*/
1060
1061#define isUTF8_CHAR_flags(s, e, flags) \
1062 (UNLIKELY((e) <= (s)) \
1063 ? 0 \
1064 : (UTF8_IS_INVARIANT(*s)) \
1065 ? 1 \
1066 : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \
1067 ? 0 \
1376b35c 1068 : is_utf8_char_helper(s, e, flags))
25e3a4e0 1069
6302f837
KW
1070/* Do not use; should be deprecated. Use isUTF8_CHAR() instead; this is
1071 * retained solely for backwards compatibility */
1072#define IS_UTF8_CHAR(p, n) (isUTF8_CHAR(p, (p) + (n)) == n)
e9a8c099 1073
6a5bc5ac 1074#endif /* PERL_UTF8_H_ */
57f0e7e2 1075
e9a8c099 1076/*
14d04a33 1077 * ex: set ts=8 sts=4 sw=4 et:
e9a8c099 1078 */