This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Make defn of UTF8_IS_START common
[perl5.git] / utf8.h
CommitLineData
a0ed51b3
LW
1/* utf8.h
2 *
f3cb6f94
KW
3 * This file contains definitions for use with the UTF-8 encoding. It
4 * actually also works with the variant UTF-8 encoding called UTF-EBCDIC, and
5 * hides almost all of the differences between these from the caller. In other
6 * words, someone should #include this file, and if the code is being compiled
7 * on an EBCDIC platform, things should mostly just work.
8 *
2eee27d7
SS
9 * Copyright (C) 2000, 2001, 2002, 2005, 2006, 2007, 2009,
10 * 2010, 2011 by Larry Wall and others
a0ed51b3
LW
11 *
12 * You may distribute under the terms of either the GNU General Public
13 * License or the Artistic License, as specified in the README file.
14 *
15 */
16
6a5bc5ac
KW
17#ifndef PERL_UTF8_H_ /* Guard against recursive inclusion */
18#define PERL_UTF8_H_ 1
57f0e7e2 19
39e02b42 20/* Use UTF-8 as the default script encoding?
1e54db1a 21 * Turning this on will break scripts having non-UTF-8 binary
39e02b42
JH
22 * data (such as Latin-1) in string literals. */
23#ifdef USE_UTF8_SCRIPTS
24# define USE_UTF8_IN_NAMES (!IN_BYTES)
25#else
26# define USE_UTF8_IN_NAMES (PL_hints & HINT_UTF8)
27#endif
28
3cd96634
KW
29#include "regcharclass.h"
30#include "unicode_constants.h"
31
051a06d4 32/* For to_utf8_fold_flags, q.v. */
e4f4ef45
KW
33#define FOLD_FLAGS_LOCALE 0x1
34#define FOLD_FLAGS_FULL 0x2
35#define FOLD_FLAGS_NOMIX_ASCII 0x4
051a06d4 36
7bbfa158
KW
37/*
38=head1 Unicode Support
8cca77bc
KW
39L<perlguts/Unicode Support> has an introduction to this API.
40
41See also L</Character classification>,
42and L</Character case changing>.
43Various functions outside this section also work specially with Unicode.
44Search for the string "utf8" in this document.
7bbfa158
KW
45
46=for apidoc is_ascii_string
47
8871a094 48This is a misleadingly-named synonym for L</is_utf8_invariant_string>.
7bbfa158
KW
49On ASCII-ish platforms, the name isn't misleading: the ASCII-range characters
50are exactly the UTF-8 invariants. But EBCDIC machines have more invariants
8871a094
KW
51than just the ASCII characters, so C<is_utf8_invariant_string> is preferred.
52
53=for apidoc is_invariant_string
54
55This is a somewhat misleadingly-named synonym for L</is_utf8_invariant_string>.
56C<is_utf8_invariant_string> is preferred, as it indicates under what conditions
57the string is invariant.
7bbfa158
KW
58
59=cut
60*/
8871a094
KW
61#define is_ascii_string(s, len) is_utf8_invariant_string(s, len)
62#define is_invariant_string(s, len) is_utf8_invariant_string(s, len)
7bbfa158 63
33f38593
KW
64#define uvoffuni_to_utf8_flags(d,uv,flags) \
65 uvoffuni_to_utf8_flags_msgs(d, uv, flags, 0)
de69f3af
KW
66#define uvchr_to_utf8(a,b) uvchr_to_utf8_flags(a,b,0)
67#define uvchr_to_utf8_flags(d,uv,flags) \
33f38593
KW
68 uvchr_to_utf8_flags_msgs(d,uv,flags, 0)
69#define uvchr_to_utf8_flags_msgs(d,uv,flags,msgs) \
70 uvoffuni_to_utf8_flags_msgs(d,NATIVE_TO_UNI(uv),flags, msgs)
de69f3af 71#define utf8_to_uvchr_buf(s, e, lenp) \
9a9a6c98 72 utf8_to_uvchr_buf_helper((const U8 *) (s), (const U8 *) e, lenp)
f9380377
KW
73#define utf8n_to_uvchr(s, len, lenp, flags) \
74 utf8n_to_uvchr_error(s, len, lenp, flags, 0)
37657a5b
KW
75#define utf8n_to_uvchr_error(s, len, lenp, flags, errors) \
76 utf8n_to_uvchr_msgs(s, len, lenp, flags, errors, 0)
de69f3af 77
a0270393 78#define to_uni_fold(c, p, lenp) _to_uni_fold_flags(c, p, lenp, FOLD_FLAGS_FULL)
a239b1e2 79
eda9cac1
KW
80#define foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \
81 foldEQ_utf8_flags(s1, pe1, l1, u1, s2, pe2, l2, u2, 0)
baa60164 82#define FOLDEQ_UTF8_NOMIX_ASCII (1 << 0)
cea315b6 83#define FOLDEQ_LOCALE (1 << 1)
18f762c3
KW
84#define FOLDEQ_S1_ALREADY_FOLDED (1 << 2)
85#define FOLDEQ_S2_ALREADY_FOLDED (1 << 3)
d635b710
KW
86#define FOLDEQ_S1_FOLDS_SANE (1 << 4)
87#define FOLDEQ_S2_FOLDS_SANE (1 << 5)
a33c29bc 88
e6226b18
KW
89#define ibcmp_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \
90 cBOOL(! foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2))
91
1d72bdf6
NIS
92#ifdef EBCDIC
93/* The equivalent of these macros but implementing UTF-EBCDIC
94 are in the following header file:
95 */
96
97#include "utfebcdic.h"
fd7cb289 98
d06134e5 99#else /* ! EBCDIC */
73c4f7a1
GS
100START_EXTERN_C
101
f4d83e55
KW
102/*
103
104=for apidoc AmnU|STRLEN|UTF8_MAXBYTES
105
106The maximum width of a single UTF-8 encoded character, in bytes.
107
108NOTE: Strictly speaking Perl's UTF-8 should not be called UTF-8 since UTF-8
109is an encoding of Unicode, and Unicode's upper limit, 0x10FFFF, can be
110expressed with 4 bytes. However, Perl thinks of UTF-8 as a way to encode
111non-negative integers in a binary format, even those above Unicode.
112
113=cut
114 */
111e8ed9
KW
115#define UTF8_MAXBYTES 13
116
a0ed51b3 117#ifdef DOINIT
6f06b55f 118EXTCONST unsigned char PL_utf8skip[] = {
b2635aa8
KW
119/* 0x00 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
120/* 0x10 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
121/* 0x20 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
122/* 0x30 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
123/* 0x40 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
124/* 0x50 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
125/* 0x60 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
126/* 0x70 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
127/* 0x80 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
128/* 0x90 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
129/* 0xA0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
130/* 0xB0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
131/* 0xC0 */ 2,2, /* overlong */
1ff3baa2 132/* 0xC2 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0080 to U+03FF */
b2635aa8
KW
133/* 0xD0 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0400 to U+07FF */
134/* 0xE0 */ 3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3, /* U+0800 to U+FFFF */
135/* 0xF0 */ 4,4,4,4,4,4,4,4,5,5,5,5,6,6, /* above BMP to 2**31 - 1 */
6937f885
KW
136 /* Perl extended (never was official UTF-8). Up to 36 bit */
137/* 0xFE */ 7,
138 /* More extended, Up to 72 bits (64-bit + reserved) */
111e8ed9 139/* 0xFF */ UTF8_MAXBYTES
a0ed51b3
LW
140};
141#else
6f06b55f 142EXTCONST unsigned char PL_utf8skip[];
a0ed51b3
LW
143#endif
144
73c4f7a1 145END_EXTERN_C
7e2040f0 146
1a3756de 147#if defined(_MSC_VER) && _MSC_VER < 1400
6f6d1bab
TC
148/* older MSVC versions have a smallish macro buffer */
149#define PERL_SMALL_MACRO_BUFFER
150#endif
151
ef2a4c8f
KW
152/*
153
154=for apidoc Am|U8|NATIVE_TO_LATIN1|U8 ch
155
156Returns the Latin-1 (including ASCII and control characters) equivalent of the
157input native code point given by C<ch>. Thus, C<NATIVE_TO_LATIN1(193)> on
158EBCDIC platforms returns 65. These each represent the character C<"A"> on
159their respective platforms. On ASCII platforms no conversion is needed, so
160this macro expands to just its input, adding no time nor space requirements to
161the implementation.
162
163For conversion of code points potentially larger than will fit in a character,
164use L</NATIVE_TO_UNI>.
165
166=for apidoc Am|U8|LATIN1_TO_NATIVE|U8 ch
167
168Returns the native equivalent of the input Latin-1 code point (including ASCII
169and control characters) given by C<ch>. Thus, C<LATIN1_TO_NATIVE(66)> on
170EBCDIC platforms returns 194. These each represent the character C<"B"> on
171their respective platforms. On ASCII platforms no conversion is needed, so
172this macro expands to just its input, adding no time nor space requirements to
173the implementation.
174
175For conversion of code points potentially larger than will fit in a character,
176use L</UNI_TO_NATIVE>.
177
178=for apidoc Am|UV|NATIVE_TO_UNI|UV ch
179
180Returns the Unicode equivalent of the input native code point given by C<ch>.
181Thus, C<NATIVE_TO_UNI(195)> on EBCDIC platforms returns 67. These each
182represent the character C<"C"> on their respective platforms. On ASCII
183platforms no conversion is needed, so this macro expands to just its input,
184adding no time nor space requirements to the implementation.
185
186=for apidoc Am|UV|UNI_TO_NATIVE|UV ch
187
188Returns the native equivalent of the input Unicode code point given by C<ch>.
189Thus, C<UNI_TO_NATIVE(68)> on EBCDIC platforms returns 196. These each
190represent the character C<"D"> on their respective platforms. On ASCII
191platforms no conversion is needed, so this macro expands to just its input,
192adding no time nor space requirements to the implementation.
193
194=cut
195*/
196
6f6d1bab 197#ifdef PERL_SMALL_MACRO_BUFFER
cb15eeb2
KW
198# define NATIVE_TO_LATIN1(ch) ((U8)(ch))
199# define LATIN1_TO_NATIVE(ch) ((U8)(ch))
6f6d1bab 200#else
cb15eeb2
KW
201# define NATIVE_TO_LATIN1(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) ((ch) | 0)))
202# define LATIN1_TO_NATIVE(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) ((ch) | 0)))
6f6d1bab 203#endif
59a449d5
KW
204
205/* I8 is an intermediate version of UTF-8 used only in UTF-EBCDIC. We thus
206 * consider it to be identical to UTF-8 on ASCII platforms. Strictly speaking
207 * UTF-8 and UTF-EBCDIC are two different things, but we often conflate them
208 * because they are 8-bit encodings that serve the same purpose in Perl, and
209 * rarely do we need to distinguish them. The term "NATIVE_UTF8" applies to
210 * whichever one is applicable on the current platform */
6f6d1bab 211#ifdef PERL_SMALL_MACRO_BUFFER
1d4ea287
DD
212#define NATIVE_UTF8_TO_I8(ch) ((U8) (ch))
213#define I8_TO_NATIVE_UTF8(ch) ((U8) (ch))
6f6d1bab 214#else
cb15eeb2
KW
215#define NATIVE_UTF8_TO_I8(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) ((ch) | 0)))
216#define I8_TO_NATIVE_UTF8(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) ((ch) | 0)))
6f6d1bab 217#endif
59a449d5 218
cb15eeb2
KW
219#define UNI_TO_NATIVE(ch) ((UV) ((ch) | 0))
220#define NATIVE_TO_UNI(ch) ((UV) ((ch) | 0))
d7578b48 221
877d9f0d 222/*
9041c2e3 223
a14e0a36
KW
224 The following table is from Unicode 3.2, plus the Perl extensions for above
225 U+10FFFF
877d9f0d 226
a14e0a36 227 Code Points 1st Byte 2nd Byte 3rd 4th 5th 6th 7th 8th-13th
877d9f0d 228
375122d7 229 U+0000..U+007F 00..7F
e1b711da 230 U+0080..U+07FF * C2..DF 80..BF
a14e0a36
KW
231 U+0800..U+0FFF E0 * A0..BF 80..BF
232 U+1000..U+CFFF E1..EC 80..BF 80..BF
233 U+D000..U+D7FF ED 80..9F 80..BF
234 U+D800..U+DFFF ED A0..BF 80..BF (surrogates)
235 U+E000..U+FFFF EE..EF 80..BF 80..BF
236 U+10000..U+3FFFF F0 * 90..BF 80..BF 80..BF
237 U+40000..U+FFFFF F1..F3 80..BF 80..BF 80..BF
238 U+100000..U+10FFFF F4 80..8F 80..BF 80..BF
239 Below are above-Unicode code points
240 U+110000..U+13FFFF F4 90..BF 80..BF 80..BF
241 U+110000..U+1FFFFF F5..F7 80..BF 80..BF 80..BF
242 U+200000..U+FFFFFF F8 * 88..BF 80..BF 80..BF 80..BF
243U+1000000..U+3FFFFFF F9..FB 80..BF 80..BF 80..BF 80..BF
244U+4000000..U+3FFFFFFF FC * 84..BF 80..BF 80..BF 80..BF 80..BF
245U+40000000..U+7FFFFFFF FD 80..BF 80..BF 80..BF 80..BF 80..BF
246U+80000000..U+FFFFFFFFF FE * 82..BF 80..BF 80..BF 80..BF 80..BF 80..BF
247U+1000000000.. FF 80..BF 80..BF 80..BF 80..BF 80..BF * 81..BF 80..BF
877d9f0d 248
e1b711da 249Note the gaps before several of the byte entries above marked by '*'. These are
37e2e78e
KW
250caused by legal UTF-8 avoiding non-shortest encodings: it is technically
251possible to UTF-8-encode a single code point in different ways, but that is
252explicitly forbidden, and the shortest possible encoding should always be used
15824458 253(and that is what Perl does). The non-shortest ones are called 'overlongs'.
8c007b5a 254
877d9f0d
JH
255 */
256
8c007b5a
JH
257/*
258 Another way to look at it, as bits:
259
b2635aa8 260 Code Points 1st Byte 2nd Byte 3rd Byte 4th Byte
8c007b5a 261
b2635aa8
KW
262 0aaa aaaa 0aaa aaaa
263 0000 0bbb bbaa aaaa 110b bbbb 10aa aaaa
264 cccc bbbb bbaa aaaa 1110 cccc 10bb bbbb 10aa aaaa
265 00 000d ddcc cccc bbbb bbaa aaaa 1111 0ddd 10cc cccc 10bb bbbb 10aa aaaa
8c007b5a
JH
266
267As you can see, the continuation bytes all begin with C<10>, and the
e1b711da 268leading bits of the start byte tell how many bytes there are in the
8c007b5a
JH
269encoded character.
270
df863e43
KW
271Perl's extended UTF-8 means we can have start bytes up through FF, though any
272beginning with FF yields a code point that is too large for 32-bit ASCII
273platforms. FF signals to use 13 bytes for the encoded character. This breaks
274the paradigm that the number of leading bits gives how many total bytes there
275are in the character.
65ab9279 276
d296fe14 277=cut
8c007b5a
JH
278*/
279
6c88483e 280/* Is the representation of the Unicode code point 'cp' the same regardless of
15824458 281 * being encoded in UTF-8 or not? */
2d1545e5 282#define OFFUNI_IS_INVARIANT(cp) isASCII(cp)
15824458 283
c2b32798
KW
284/*
285=for apidoc Am|bool|UVCHR_IS_INVARIANT|UV cp
286
287Evaluates to 1 if the representation of code point C<cp> is the same whether or
288not it is encoded in UTF-8; otherwise evaluates to 0. UTF-8 invariant
289characters can be copied as-is when converting to/from UTF-8, saving time.
290C<cp> is Unicode if above 255; otherwise is platform-native.
291
292=cut
293 */
294
cf1be84e 295#define UVCHR_IS_INVARIANT(cp) OFFUNI_IS_INVARIANT(cp)
38953e5a 296
a95ec4fb 297/* Misleadingly named: is the UTF8-encoded byte 'c' part of a variant sequence
5c06326b
KW
298 * in UTF-8? This is the inverse of UTF8_IS_INVARIANT. The |0 makes sure this
299 * isn't mistakenly called with a ptr argument */
a6951642
KW
300#define UTF8_IS_CONTINUED(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
301 ((U8)((c) | 0)) & UTF_CONTINUATION_MARK)
15824458 302
15824458
KW
303/* Is the UTF8-encoded byte 'c' the first byte of a two byte sequence? Use
304 * UTF8_IS_NEXT_CHAR_DOWNGRADEABLE() instead if the input isn't known to
305 * be well-formed. Masking with 0xfe allows the low bit to be 0 or 1; thus
5c06326b
KW
306 * this matches 0xc[23]. The |0 makes sure this isn't mistakenly called with a
307 * ptr argument */
a6951642
KW
308#define UTF8_IS_DOWNGRADEABLE_START(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
309 (((U8)((c) | 0)) & 0xfe) == 0xc2)
4ab10950 310
15824458 311/* Is the UTF8-encoded byte 'c' the first byte of a sequence of bytes that
5c06326b
KW
312 * represent a code point > 255? The |0 makes sure this isn't mistakenly
313 * called with a ptr argument */
a6951642
KW
314#define UTF8_IS_ABOVE_LATIN1(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
315 ((U8)((c) | 0)) >= 0xc4)
8850bf83 316
15824458
KW
317/* This is the number of low-order bits a continuation byte in a UTF-8 encoded
318 * sequence contributes to the specification of the code point. In the bit
319 * maps above, you see that the first 2 bits are a constant '10', leaving 6 of
320 * real information */
1d72bdf6 321#define UTF_ACCUMULATION_SHIFT 6
b2635aa8 322
fed423a5
KW
323/* ^? is defined to be DEL on ASCII systems. See the definition of toCTRL()
324 * for more */
325#define QUESTION_MARK_CTRL DEL_NATIVE
326
327/* Surrogates, non-character code points and above-Unicode code points are
328 * problematic in some contexts. This allows code that needs to check for
329 * those to to quickly exclude the vast majority of code points it will
330 * encounter */
a6951642
KW
331#define isUTF8_POSSIBLY_PROBLEMATIC(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
332 (U8) c >= 0xED)
fed423a5 333
57ff5f59
KW
334#define UNICODE_IS_PERL_EXTENDED(uv) UNLIKELY((UV) (uv) > 0x7FFFFFFF)
335
fed423a5
KW
336#endif /* EBCDIC vs ASCII */
337
d7bcd45a
KW
338/* 2**UTF_ACCUMULATION_SHIFT - 1. This masks out all but the bits that carry
339 * real information in a continuation byte. This turns out to be 0x3F in
340 * UTF-8, 0x1F in UTF-EBCDIC. */
fed423a5
KW
341#define UTF_CONTINUATION_MASK ((U8) ((1U << UTF_ACCUMULATION_SHIFT) - 1))
342
9f3cfb7a
KW
343/* For use in UTF8_IS_CONTINUATION(). This turns out to be 0xC0 in UTF-8,
344 * E0 in UTF-EBCDIC */
345#define UTF_IS_CONTINUATION_MASK ((U8) (0xFF << UTF_ACCUMULATION_SHIFT))
346
38f458ff
KW
347/* This defines the bits that are to be in the continuation bytes of a
348 * multi-byte UTF-8 encoded character that mark it is a continuation byte.
349 * This turns out to be 0x80 in UTF-8, 0xA0 in UTF-EBCDIC. (khw doesn't know
350 * the underlying reason that B0 works here) */
351#define UTF_CONTINUATION_MARK (UTF_IS_CONTINUATION_MASK & 0xB0)
352
f4225fa0
KW
353/* Is the byte 'c' part of a multi-byte UTF8-8 encoded sequence, and not the
354 * first byte thereof? */
355#define UTF8_IS_CONTINUATION(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
356 (((NATIVE_UTF8_TO_I8(c) & UTF_IS_CONTINUATION_MASK) \
357 == UTF_CONTINUATION_MARK)))
358
7028aeba
KW
359/* Internal macro to be used only in this file to aid in constructing other
360 * publicly accessible macros.
361 * The number of bytes required to express this uv in UTF-8, for just those
362 * uv's requiring 2 through 6 bytes, as these are common to all platforms and
363 * word sizes. The number of bytes needed is given by the number of leading 1
364 * bits in the start byte. There are 32 start bytes that have 2 initial 1 bits
365 * (C0-DF); there are 16 that have 3 initial 1 bits (E0-EF); 8 that have 4
366 * initial 1 bits (F0-F8); 4 that have 5 initial 1 bits (F9-FB), and 2 that
367 * have 6 initial 1 bits (FC-FD). The largest number a string of n bytes can
368 * represent is (the number of possible start bytes for 'n')
369 * * (the number of possiblities for each start byte
370 * The latter in turn is
371 * 2 ** ( (how many continuation bytes there are)
372 * * (the number of bits of information each
373 * continuation byte holds))
374 *
375 * If we were on a platform where we could use a fast find first set bit
376 * instruction (or count leading zeros instruction) this could be replaced by
377 * using that to find the log2 of the uv, and divide that by the number of bits
378 * of information in each continuation byte, adjusting for large cases and how
379 * much information is in a start byte for that length */
72164d3a 380#define __COMMON_UNI_SKIP(uv) \
7028aeba
KW
381 (UV) (uv) < (32 * (1U << ( UTF_ACCUMULATION_SHIFT))) ? 2 : \
382 (UV) (uv) < (16 * (1U << (2 * UTF_ACCUMULATION_SHIFT))) ? 3 : \
383 (UV) (uv) < ( 8 * (1U << (3 * UTF_ACCUMULATION_SHIFT))) ? 4 : \
384 (UV) (uv) < ( 4 * (1U << (4 * UTF_ACCUMULATION_SHIFT))) ? 5 : \
385 (UV) (uv) < ( 2 * (1U << (5 * UTF_ACCUMULATION_SHIFT))) ? 6 :
72164d3a
KW
386
387/* Internal macro to be used only in this file.
388 * This adds to __COMMON_UNI_SKIP the details at this platform's upper range.
fed423a5 389 * For any-sized EBCDIC platforms, or 64-bit ASCII ones, we need one more test
72164d3a
KW
390 * to see if just 7 bytes is needed, or if the maximum is needed. For 32-bit
391 * ASCII platforms, everything is representable by 7 bytes */
fed423a5 392#if defined(UV_IS_QUAD) || defined(EBCDIC)
72164d3a 393# define __BASE_UNI_SKIP(uv) (__COMMON_UNI_SKIP(uv) \
7028aeba 394 (UV) (uv) < ((UV) 1U << (6 * UTF_ACCUMULATION_SHIFT)) ? 7 : UTF8_MAXBYTES)
1d68d6cd 395#else
72164d3a 396# define __BASE_UNI_SKIP(uv) (__COMMON_UNI_SKIP(uv) 7)
1d68d6cd
SC
397#endif
398
fed423a5
KW
399/* The next two macros use the base macro defined above, and add in the tests
400 * at the low-end of the range, for just 1 byte, yielding complete macros,
401 * publicly accessible. */
402
403/* Input is a true Unicode (not-native) code point */
404#define OFFUNISKIP(uv) (OFFUNI_IS_INVARIANT(uv) ? 1 : __BASE_UNI_SKIP(uv))
2084b489 405
5352a763
KW
406/*
407
408=for apidoc Am|STRLEN|UVCHR_SKIP|UV cp
409returns the number of bytes required to represent the code point C<cp> when
410encoded as UTF-8. C<cp> is a native (ASCII or EBCDIC) code point if less than
411255; a Unicode code point otherwise.
412
413=cut
414 */
fdb6583d 415#define UVCHR_SKIP(uv) ( UVCHR_IS_INVARIANT(uv) ? 1 : __BASE_UNI_SKIP(uv))
5352a763 416
4bab39bc
KW
417#define UTF_MIN_START_BYTE \
418 ((UTF_CONTINUATION_MARK >> UTF_ACCUMULATION_SHIFT) | UTF_START_MARK(2))
419
420/* Is the byte 'c' the first byte of a multi-byte UTF8-8 encoded sequence?
421 * This doesn't catch invariants (they are single-byte). It also excludes the
422 * illegal overlong sequences that begin with C0 and C1 on ASCII platforms, and
423 * C0-C4 I8 start bytes on EBCDIC ones */
424#define UTF8_IS_START(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
425 (NATIVE_UTF8_TO_I8(c) >= UTF_MIN_START_BYTE))
426
b651802e
KW
427/* The largest code point representable by two UTF-8 bytes on this platform.
428 * As explained in the comments for __COMMON_UNI_SKIP, 32 start bytes with
fed423a5 429 * UTF_ACCUMULATION_SHIFT bits of information each */
aa206fb7
KW
430#define MAX_UTF8_TWO_BYTE (32 * (1U << UTF_ACCUMULATION_SHIFT) - 1)
431
b651802e
KW
432/* The largest code point representable by two UTF-8 bytes on any platform that
433 * Perl runs on. This value is constrained by EBCDIC which has 5 bits per
434 * continuation byte */
aa206fb7
KW
435#define MAX_PORTABLE_UTF8_TWO_BYTE (32 * (1U << 5) - 1)
436
c03c0950
KW
437/* The maximum number of UTF-8 bytes a single Unicode character can
438 * uppercase/lowercase/fold into. Unicode guarantees that the maximum
439 * expansion is UTF8_MAX_FOLD_CHAR_EXPAND characters, but any above-Unicode
440 * code point will fold to itself, so we only have to look at the expansion of
441 * the maximum Unicode code point. But this number may be less than the space
442 * occupied by a very large code point under Perl's extended UTF-8. We have to
443 * make it large enough to fit any single character. (It turns out that ASCII
444 * and EBCDIC differ in which is larger) */
445#define UTF8_MAXBYTES_CASE \
446 (UTF8_MAXBYTES >= (UTF8_MAX_FOLD_CHAR_EXPAND * OFFUNISKIP(0x10FFFF)) \
447 ? UTF8_MAXBYTES \
448 : (UTF8_MAX_FOLD_CHAR_EXPAND * OFFUNISKIP(0x10FFFF)))
449
d06134e5
KW
450/* Rest of these are attributes of Unicode and perl's internals rather than the
451 * encoding, or happen to be the same in both ASCII and EBCDIC (at least at
452 * this level; the macros that some of these call may have different
453 * definitions in the two encodings */
454
59a449d5
KW
455/* In domain restricted to ASCII, these may make more sense to the reader than
456 * the ones with Latin1 in the name */
457#define NATIVE_TO_ASCII(ch) NATIVE_TO_LATIN1(ch)
458#define ASCII_TO_NATIVE(ch) LATIN1_TO_NATIVE(ch)
459
460/* More or less misleadingly-named defines, retained for back compat */
461#define NATIVE_TO_UTF(ch) NATIVE_UTF8_TO_I8(ch)
462#define NATIVE_TO_I8(ch) NATIVE_UTF8_TO_I8(ch)
463#define UTF_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch)
464#define I8_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch)
465#define NATIVE8_TO_UNI(ch) NATIVE_TO_LATIN1(ch)
d06134e5 466
c0236afe 467/* This defines the 1-bits that are to be in the first byte of a multi-byte
97d0ceda
KW
468 * UTF-8 encoded character that mark it as a start byte and give the number of
469 * bytes that comprise the character. 'len' is the number of bytes in the
470 * multi-byte sequence. */
c0236afe
KW
471#define UTF_START_MARK(len) (((len) > 7) ? 0xFF : (0xFF & (0xFE << (7-(len)))))
472
473/* Masks out the initial one bits in a start byte, leaving the real data ones.
474 * Doesn't work on an invariant byte. 'len' is the number of bytes in the
475 * multi-byte sequence that comprises the character. */
476#define UTF_START_MASK(len) (((len) >= 7) ? 0x00 : (0x1F >> ((len)-2)))
477
537124e4
KW
478/* Adds a UTF8 continuation byte 'new' of information to a running total code
479 * point 'old' of all the continuation bytes so far. This is designed to be
155d2738
KW
480 * used in a loop to convert from UTF-8 to the code point represented. Note
481 * that this is asymmetric on EBCDIC platforms, in that the 'new' parameter is
482 * the UTF-EBCDIC byte, whereas the 'old' parameter is a Unicode (not EBCDIC)
483 * code point in process of being generated */
a6951642
KW
484#define UTF8_ACCUMULATE(old, new) (__ASSERT_(FITS_IN_8_BITS(new)) \
485 ((old) << UTF_ACCUMULATION_SHIFT) \
009097b1 486 | ((NATIVE_UTF8_TO_I8(new)) \
155d2738 487 & UTF_CONTINUATION_MASK))
d06134e5 488
4ab10950 489/* This works in the face of malformed UTF-8. */
4e1ed312
KW
490#define UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(s, e) \
491 ( UTF8_IS_DOWNGRADEABLE_START(*(s)) \
492 && ( (e) - (s) > 1) \
493 && UTF8_IS_CONTINUATION(*((s)+1)))
4ab10950 494
5aaebcb3 495/* Number of bytes a code point occupies in UTF-8. */
5352a763 496#define NATIVE_SKIP(uv) UVCHR_SKIP(uv)
bd18bd40 497
5aaebcb3
KW
498/* Most code which says UNISKIP is really thinking in terms of native code
499 * points (0-255) plus all those beyond. This is an imprecise term, but having
2accb712 500 * it means existing code continues to work. For precision, use UVCHR_SKIP,
5352a763
KW
501 * NATIVE_SKIP, or OFFUNISKIP */
502#define UNISKIP(uv) UVCHR_SKIP(uv)
5aaebcb3 503
3c0792e4
KW
504/* Longer, but more accurate name */
505#define UTF8_IS_ABOVE_LATIN1_START(c) UTF8_IS_ABOVE_LATIN1(c)
506
a62b247b
KW
507/* Convert a UTF-8 variant Latin1 character to a native code point value.
508 * Needs just one iteration of accumulate. Should be used only if it is known
509 * that the code point is < 256, and is not UTF-8 invariant. Use the slower
510 * but more general TWO_BYTE_UTF8_TO_NATIVE() which handles any code point
511 * representable by two bytes (which turns out to be up through
512 * MAX_PORTABLE_UTF8_TWO_BYTE). The two parameters are:
513 * HI: a downgradable start byte;
514 * LO: continuation.
515 * */
516#define EIGHT_BIT_UTF8_TO_NATIVE(HI, LO) \
517 ( __ASSERT_(UTF8_IS_DOWNGRADEABLE_START(HI)) \
518 __ASSERT_(UTF8_IS_CONTINUATION(LO)) \
519 LATIN1_TO_NATIVE(UTF8_ACCUMULATE(( \
520 NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), (LO))))
521
94bb8c36 522/* Convert a two (not one) byte utf8 character to a native code point value.
2950f2a7
KW
523 * Needs just one iteration of accumulate. Should not be used unless it is
524 * known that the two bytes are legal: 1) two-byte start, and 2) continuation.
525 * Note that the result can be larger than 255 if the input character is not
526 * downgradable */
94bb8c36 527#define TWO_BYTE_UTF8_TO_NATIVE(HI, LO) \
a6951642
KW
528 (__ASSERT_(FITS_IN_8_BITS(HI)) \
529 __ASSERT_(FITS_IN_8_BITS(LO)) \
530 __ASSERT_(PL_utf8skip[HI] == 2) \
531 __ASSERT_(UTF8_IS_CONTINUATION(LO)) \
94bb8c36 532 UNI_TO_NATIVE(UTF8_ACCUMULATE((NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), \
635e76f5 533 (LO))))
94bb8c36
KW
534
535/* Should never be used, and be deprecated */
536#define TWO_BYTE_UTF8_TO_UNI(HI, LO) NATIVE_TO_UNI(TWO_BYTE_UTF8_TO_NATIVE(HI, LO))
2950f2a7 537
bd18bd40
KW
538/*
539
540=for apidoc Am|STRLEN|UTF8SKIP|char* s
541returns the number of bytes in the UTF-8 encoded character whose first (perhaps
542only) byte is pointed to by C<s>.
543
544=cut
545 */
2a70536e
KW
546#define UTF8SKIP(s) PL_utf8skip[*(const U8*)(s)]
547#define UTF8_SKIP(s) UTF8SKIP(s)
d06134e5 548
85fcc8f2
KW
549/*
550
551=for apidoc Am|STRLEN|UTF8_SAFE_SKIP|char* s|char* e
45671da2
KW
552returns 0 if S<C<s E<gt>= e>>; otherwise returns the number of bytes in the
553UTF-8 encoded character whose first byte is pointed to by C<s>. But it never
554returns beyond C<e>. On DEBUGGING builds, it asserts that S<C<s E<lt>= e>>.
85fcc8f2
KW
555
556=cut
557 */
45671da2
KW
558#define UTF8_SAFE_SKIP(s, e) (__ASSERT_((e) >= (s)) \
559 ((e) - (s)) <= 0 \
560 ? 0 \
561 : MIN(((e) - (s)), UTF8_SKIP(s)))
85fcc8f2 562
2d1545e5
KW
563/* Most code that says 'UNI_' really means the native value for code points up
564 * through 255 */
565#define UNI_IS_INVARIANT(cp) UVCHR_IS_INVARIANT(cp)
566
c2b32798
KW
567/*
568=for apidoc Am|bool|UTF8_IS_INVARIANT|char c
569
570Evaluates to 1 if the byte C<c> represents the same character when encoded in
571UTF-8 as when not; otherwise evaluates to 0. UTF-8 invariant characters can be
572copied as-is when converting to/from UTF-8, saving time.
573
574In spite of the name, this macro gives the correct result if the input string
575from which C<c> comes is not encoded in UTF-8.
576
577See C<L</UVCHR_IS_INVARIANT>> for checking if a UV is invariant.
578
579=cut
580
581The reason it works on both UTF-8 encoded strings and non-UTF-8 encoded, is
582that it returns TRUE in each for the exact same set of bit patterns. It is
583valid on a subset of what UVCHR_IS_INVARIANT is valid on, so can just use that;
584and the compiler should optimize out anything extraneous given the
585implementation of the latter. The |0 makes sure this isn't mistakenly called
586with a ptr argument.
587*/
5c06326b 588#define UTF8_IS_INVARIANT(c) UVCHR_IS_INVARIANT((c) | 0)
5fc230f1
KW
589
590/* Like the above, but its name implies a non-UTF8 input, which as the comments
591 * above show, doesn't matter as to its implementation */
38953e5a 592#define NATIVE_BYTE_IS_INVARIANT(c) UVCHR_IS_INVARIANT(c)
d06134e5 593
48ccf5e1
KW
594/* The macros in the next 4 sets are used to generate the two utf8 or utfebcdic
595 * bytes from an ordinal that is known to fit into exactly two (not one) bytes;
596 * it must be less than 0x3FF to work across both encodings. */
597
598/* These two are helper macros for the other three sets, and should not be used
599 * directly anywhere else. 'translate_function' is either NATIVE_TO_LATIN1
1ff3baa2
KW
600 * (which works for code points up through 0xFF) or NATIVE_TO_UNI which works
601 * for any code point */
48ccf5e1 602#define __BASE_TWO_BYTE_HI(c, translate_function) \
2863dafa 603 (__ASSERT_(! UVCHR_IS_INVARIANT(c)) \
48ccf5e1 604 I8_TO_NATIVE_UTF8((translate_function(c) >> UTF_ACCUMULATION_SHIFT) \
2863dafa 605 | UTF_START_MARK(2)))
48ccf5e1 606#define __BASE_TWO_BYTE_LO(c, translate_function) \
2863dafa 607 (__ASSERT_(! UVCHR_IS_INVARIANT(c)) \
48ccf5e1 608 I8_TO_NATIVE_UTF8((translate_function(c) & UTF_CONTINUATION_MASK) \
2863dafa 609 | UTF_CONTINUATION_MARK))
48ccf5e1 610
48ccf5e1
KW
611/* The next two macros should not be used. They were designed to be usable as
612 * the case label of a switch statement, but this doesn't work for EBCDIC. Use
9d0d3a03 613 * regen/unicode_constants.pl instead */
48ccf5e1
KW
614#define UTF8_TWO_BYTE_HI_nocast(c) __BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI)
615#define UTF8_TWO_BYTE_LO_nocast(c) __BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI)
616
617/* The next two macros are used when the source should be a single byte
618 * character; checked for under DEBUGGING */
619#define UTF8_EIGHT_BIT_HI(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
4c8cd605 620 ( __BASE_TWO_BYTE_HI(c, NATIVE_TO_LATIN1)))
48ccf5e1 621#define UTF8_EIGHT_BIT_LO(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
4c8cd605 622 (__BASE_TWO_BYTE_LO(c, NATIVE_TO_LATIN1)))
48ccf5e1
KW
623
624/* These final two macros in the series are used when the source can be any
625 * code point whose UTF-8 is known to occupy 2 bytes; they are less efficient
626 * than the EIGHT_BIT versions on EBCDIC platforms. We use the logical '~'
627 * operator instead of "<=" to avoid getting compiler warnings.
d52b8576 628 * MAX_UTF8_TWO_BYTE should be exactly all one bits in the lower few
48ccf5e1
KW
629 * places, so the ~ works */
630#define UTF8_TWO_BYTE_HI(c) \
631 (__ASSERT_((sizeof(c) == 1) \
d52b8576 632 || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \
4c8cd605 633 (__BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI)))
48ccf5e1
KW
634#define UTF8_TWO_BYTE_LO(c) \
635 (__ASSERT_((sizeof(c) == 1) \
d52b8576 636 || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \
4c8cd605 637 (__BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI)))
d06134e5 638
e7214ce8
KW
639/* This is illegal in any well-formed UTF-8 in both EBCDIC and ASCII
640 * as it is only in overlongs. */
641#define ILLEGAL_UTF8_BYTE I8_TO_NATIVE_UTF8(0xC1)
642
7e2040f0 643/*
e3036cf4 644 * 'UTF' is whether or not p is encoded in UTF8. The names 'foo_lazy_if' stem
20df05f4
KW
645 * from an earlier version of these macros in which they didn't call the
646 * foo_utf8() macros (i.e. were 'lazy') unless they decided that *p is the
647 * beginning of a utf8 character. Now that foo_utf8() determines that itself,
648 * no need to do it again here
7e2040f0 649 */
34aeb2e9
KW
650#define isIDFIRST_lazy_if(p,UTF) \
651 _is_utf8_FOO(_CC_IDFIRST, (const U8 *) p, "isIDFIRST_lazy_if", \
652 "isIDFIRST_lazy_if_safe", \
653 cBOOL(UTF && ! IN_BYTES), 0, __FILE__,__LINE__)
1d72bdf6 654
da8c1a98
KW
655#define isIDFIRST_lazy_if_safe(p, e, UTF) \
656 ((IN_BYTES || !UTF) \
657 ? isIDFIRST(*(p)) \
658 : isIDFIRST_utf8_safe(p, e))
659
34aeb2e9
KW
660#define isWORDCHAR_lazy_if(p,UTF) \
661 _is_utf8_FOO(_CC_IDFIRST, (const U8 *) p, "isWORDCHAR_lazy_if", \
662 "isWORDCHAR_lazy_if_safe", \
663 cBOOL(UTF && ! IN_BYTES), 0, __FILE__,__LINE__)
664
da8c1a98
KW
665#define isWORDCHAR_lazy_if_safe(p, e, UTF) \
666 ((IN_BYTES || !UTF) \
667 ? isWORDCHAR(*(p)) \
668 : isWORDCHAR_utf8_safe((U8 *) p, (U8 *) e))
669
34aeb2e9
KW
670#define isALNUM_lazy_if(p,UTF) \
671 _is_utf8_FOO(_CC_IDFIRST, (const U8 *) p, "isALNUM_lazy_if", \
672 "isWORDCHAR_lazy_if_safe", \
673 cBOOL(UTF && ! IN_BYTES), 0, __FILE__,__LINE__)
da8c1a98 674
89ebb4a3
JH
675#define UTF8_MAXLEN UTF8_MAXBYTES
676
8cb75cc8
KW
677/* A Unicode character can fold to up to 3 characters */
678#define UTF8_MAX_FOLD_CHAR_EXPAND 3
679
d3481830 680#define IN_BYTES UNLIKELY(CopHINTS_get(PL_curcop) & HINT_BYTES)
bd18bd40
KW
681
682/*
683
684=for apidoc Am|bool|DO_UTF8|SV* sv
685Returns a bool giving whether or not the PV in C<sv> is to be treated as being
686encoded in UTF-8.
687
688You should use this I<after> a call to C<SvPV()> or one of its variants, in
689case any call to string overloading updates the internal UTF-8 encoding flag.
690
691=cut
692*/
0064a8a9 693#define DO_UTF8(sv) (SvUTF8(sv) && !IN_BYTES)
1ff3baa2
KW
694
695/* Should all strings be treated as Unicode, and not just UTF-8 encoded ones?
696 * Is so within 'feature unicode_strings' or 'locale :not_characters', and not
697 * within 'use bytes'. UTF-8 locales are not tested for here, but perhaps
698 * could be */
70844984
KW
699#define IN_UNI_8_BIT \
700 (( ( (CopHINTS_get(PL_curcop) & HINT_UNI_8_BIT)) \
701 || ( CopHINTS_get(PL_curcop) & HINT_LOCALE_PARTIAL \
702 /* -1 below is for :not_characters */ \
703 && _is_in_locale_category(FALSE, -1))) \
704 && (! IN_BYTES))
b36bf33f 705
1d72bdf6 706
c76687c5 707#define UTF8_ALLOW_EMPTY 0x0001 /* Allow a zero length string */
2b5e7bc2 708#define UTF8_GOT_EMPTY UTF8_ALLOW_EMPTY
c76687c5
KW
709
710/* Allow first byte to be a continuation byte */
1d72bdf6 711#define UTF8_ALLOW_CONTINUATION 0x0002
2b5e7bc2 712#define UTF8_GOT_CONTINUATION UTF8_ALLOW_CONTINUATION
c76687c5 713
cd01d3b1 714/* Unexpected non-continuation byte */
1d72bdf6 715#define UTF8_ALLOW_NON_CONTINUATION 0x0004
2b5e7bc2 716#define UTF8_GOT_NON_CONTINUATION UTF8_ALLOW_NON_CONTINUATION
949cf498
KW
717
718/* expecting more bytes than were available in the string */
719#define UTF8_ALLOW_SHORT 0x0008
2b5e7bc2 720#define UTF8_GOT_SHORT UTF8_ALLOW_SHORT
949cf498 721
94953955
KW
722/* Overlong sequence; i.e., the code point can be specified in fewer bytes.
723 * First one will convert the overlong to the REPLACEMENT CHARACTER; second
724 * will return what the overlong evaluates to */
949cf498 725#define UTF8_ALLOW_LONG 0x0010
94953955 726#define UTF8_ALLOW_LONG_AND_ITS_VALUE (UTF8_ALLOW_LONG|0x0020)
2b5e7bc2
KW
727#define UTF8_GOT_LONG UTF8_ALLOW_LONG
728
d60baaa7
KW
729#define UTF8_ALLOW_OVERFLOW 0x0080
730#define UTF8_GOT_OVERFLOW UTF8_ALLOW_OVERFLOW
949cf498 731
f180b292 732#define UTF8_DISALLOW_SURROGATE 0x0100 /* Unicode surrogates */
2b5e7bc2 733#define UTF8_GOT_SURROGATE UTF8_DISALLOW_SURROGATE
f180b292 734#define UTF8_WARN_SURROGATE 0x0200
949cf498 735
c4e96019
KW
736/* Unicode non-character code points */
737#define UTF8_DISALLOW_NONCHAR 0x0400
2b5e7bc2 738#define UTF8_GOT_NONCHAR UTF8_DISALLOW_NONCHAR
c4e96019 739#define UTF8_WARN_NONCHAR 0x0800
949cf498 740
c4e96019
KW
741/* Super-set of Unicode: code points above the legal max */
742#define UTF8_DISALLOW_SUPER 0x1000
2b5e7bc2 743#define UTF8_GOT_SUPER UTF8_DISALLOW_SUPER
c4e96019
KW
744#define UTF8_WARN_SUPER 0x2000
745
746/* The original UTF-8 standard did not define UTF-8 with start bytes of 0xFE or
747 * 0xFF, though UTF-EBCDIC did. This allowed both versions to represent code
748 * points up to 2 ** 31 - 1. Perl extends UTF-8 so that 0xFE and 0xFF are
749 * usable on ASCII platforms, and 0xFF means something different than
750 * UTF-EBCDIC defines. These changes allow code points of 64 bits (actually
751 * somewhat more) to be represented on both platforms. But these are Perl
752 * extensions, and not likely to be interchangeable with other languages. Note
753 * that on ASCII platforms, FE overflows a signed 32-bit word, and FF an
754 * unsigned one. */
d044b7a7
KW
755#define UTF8_DISALLOW_PERL_EXTENDED 0x4000
756#define UTF8_GOT_PERL_EXTENDED UTF8_DISALLOW_PERL_EXTENDED
757#define UTF8_WARN_PERL_EXTENDED 0x8000
d35f2ca5 758
57ff5f59
KW
759/* For back compat, these old names are misleading for overlongs and
760 * UTF_EBCDIC. */
d044b7a7
KW
761#define UTF8_DISALLOW_ABOVE_31_BIT UTF8_DISALLOW_PERL_EXTENDED
762#define UTF8_GOT_ABOVE_31_BIT UTF8_GOT_PERL_EXTENDED
763#define UTF8_WARN_ABOVE_31_BIT UTF8_WARN_PERL_EXTENDED
764#define UTF8_DISALLOW_FE_FF UTF8_DISALLOW_PERL_EXTENDED
765#define UTF8_WARN_FE_FF UTF8_WARN_PERL_EXTENDED
949cf498 766
f180b292 767#define UTF8_CHECK_ONLY 0x10000
99a765e9 768#define _UTF8_NO_CONFIDENCE_IN_CURLEN 0x20000 /* Internal core use only */
949cf498
KW
769
770/* For backwards source compatibility. They do nothing, as the default now
771 * includes what they used to mean. The first one's meaning was to allow the
772 * just the single non-character 0xFFFF */
773#define UTF8_ALLOW_FFFF 0
c825ef8c 774#define UTF8_ALLOW_FE_FF 0
949cf498
KW
775#define UTF8_ALLOW_SURROGATE 0
776
ecc1615f
KW
777/* C9 refers to Unicode Corrigendum #9: allows but discourages non-chars */
778#define UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE \
779 (UTF8_DISALLOW_SUPER|UTF8_DISALLOW_SURROGATE)
780#define UTF8_WARN_ILLEGAL_C9_INTERCHANGE (UTF8_WARN_SUPER|UTF8_WARN_SURROGATE)
781
d35f2ca5 782#define UTF8_DISALLOW_ILLEGAL_INTERCHANGE \
ecc1615f 783 (UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE|UTF8_DISALLOW_NONCHAR)
949cf498 784#define UTF8_WARN_ILLEGAL_INTERCHANGE \
ecc1615f
KW
785 (UTF8_WARN_ILLEGAL_C9_INTERCHANGE|UTF8_WARN_NONCHAR)
786
0eb3d6a0
KW
787/* This is typically used for code that processes UTF-8 input and doesn't want
788 * to have to deal with any malformations that might be present. All such will
789 * be safely replaced by the REPLACEMENT CHARACTER, unless other flags
790 * overriding this are also present. */
2d532c27
KW
791#define UTF8_ALLOW_ANY ( UTF8_ALLOW_CONTINUATION \
792 |UTF8_ALLOW_NON_CONTINUATION \
793 |UTF8_ALLOW_SHORT \
d60baaa7
KW
794 |UTF8_ALLOW_LONG \
795 |UTF8_ALLOW_OVERFLOW)
2d532c27
KW
796
797/* Accept any Perl-extended UTF-8 that evaluates to any UV on the platform, but
cd01d3b1 798 * not any malformed. This is the default. */
2d532c27
KW
799#define UTF8_ALLOW_ANYUV 0
800#define UTF8_ALLOW_DEFAULT UTF8_ALLOW_ANYUV
1d72bdf6 801
89d986df
KW
802/*
803=for apidoc Am|bool|UTF8_IS_SURROGATE|const U8 *s|const U8 *e
804
805Evaluates to non-zero if the first few bytes of the string starting at C<s> and
806looking no further than S<C<e - 1>> are well-formed UTF-8 that represents one
807of the Unicode surrogate code points; otherwise it evaluates to 0. If
808non-zero, the value gives how many bytes starting at C<s> comprise the code
809point's representation.
810
811=cut
812 */
813#define UTF8_IS_SURROGATE(s, e) is_SURROGATE_utf8_safe(s, e)
814
815
816#define UTF8_IS_REPLACEMENT(s, send) is_REPLACEMENT_utf8_safe(s,send)
817
285aa1f0 818#define MAX_LEGAL_CP ((UV)IV_MAX)
40606899 819
89d986df
KW
820/*
821=for apidoc Am|bool|UTF8_IS_SUPER|const U8 *s|const U8 *e
822
823Recall that Perl recognizes an extension to UTF-8 that can encode code
824points larger than the ones defined by Unicode, which are 0..0x10FFFF.
825
826This macro evaluates to non-zero if the first few bytes of the string starting
827at C<s> and looking no further than S<C<e - 1>> are from this UTF-8 extension;
828otherwise it evaluates to 0. If non-zero, the value gives how many bytes
829starting at C<s> comprise the code point's representation.
830
8310 is returned if the bytes are not well-formed extended UTF-8, or if they
832represent a code point that cannot fit in a UV on the current platform. Hence
833this macro can give different results when run on a 64-bit word machine than on
834one with a 32-bit word size.
0c58a72b 835
891fd405 836Note that it is illegal to have code points that are larger than what can
89d986df 837fit in an IV on the current machine.
7131f24d 838
89d986df
KW
839=cut
840
841 * ASCII EBCDIC I8
7131f24d
KW
842 * U+10FFFF: \xF4\x8F\xBF\xBF \xF9\xA1\xBF\xBF\xBF max legal Unicode
843 * U+110000: \xF4\x90\x80\x80 \xF9\xA2\xA0\xA0\xA0
844 * U+110001: \xF4\x90\x80\x81 \xF9\xA2\xA0\xA0\xA1
89d986df
KW
845 */
846#ifdef EBCDIC
a14e0a36 847# define UTF8_IS_SUPER(s, e) \
89d986df
KW
848 (( LIKELY((e) > (s) + 4) \
849 && NATIVE_UTF8_TO_I8(*(s)) >= 0xF9 \
850 && ( NATIVE_UTF8_TO_I8(*(s)) > 0xF9 \
3d42f267 851 || (NATIVE_UTF8_TO_I8(*((s) + 1)) >= 0xA2)) \
89d986df 852 && LIKELY((s) + UTF8SKIP(s) <= (e))) \
1376b35c 853 ? is_utf8_char_helper(s, s + UTF8SKIP(s), 0) : 0)
7131f24d 854#else
a14e0a36 855# define UTF8_IS_SUPER(s, e) \
89d986df
KW
856 (( LIKELY((e) > (s) + 3) \
857 && (*(U8*) (s)) >= 0xF4 \
858 && ((*(U8*) (s)) > 0xF4 || (*((U8*) (s) + 1) >= 0x90))\
859 && LIKELY((s) + UTF8SKIP(s) <= (e))) \
1376b35c 860 ? is_utf8_char_helper(s, s + UTF8SKIP(s), 0) : 0)
7131f24d
KW
861#endif
862
b96a92fb
KW
863/* These are now machine generated, and the 'given' clause is no longer
864 * applicable */
0c58a72b 865#define UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s, e) \
89d986df
KW
866 cBOOL(is_NONCHAR_utf8_safe(s,e))
867
868/*
869=for apidoc Am|bool|UTF8_IS_NONCHAR|const U8 *s|const U8 *e
870
871Evaluates to non-zero if the first few bytes of the string starting at C<s> and
872looking no further than S<C<e - 1>> are well-formed UTF-8 that represents one
873of the Unicode non-character code points; otherwise it evaluates to 0. If
874non-zero, the value gives how many bytes starting at C<s> comprise the code
875point's representation.
876
d296fe14
KW
877=for apidoc AmnU|UV|UNICODE_REPLACEMENT
878
879Evaluates to 0xFFFD, the code point of the Unicode REPLACEMENT CHARACTER
880
89d986df
KW
881=cut
882 */
0c58a72b
KW
883#define UTF8_IS_NONCHAR(s, e) \
884 UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s, e)
7131f24d 885
c867b360
JH
886#define UNICODE_SURROGATE_FIRST 0xD800
887#define UNICODE_SURROGATE_LAST 0xDFFF
888#define UNICODE_REPLACEMENT 0xFFFD
889#define UNICODE_BYTE_ORDER_MARK 0xFEFF
1d72bdf6 890
b851fbc1 891/* Though our UTF-8 encoding can go beyond this,
c76687c5 892 * let's be conservative and do as Unicode says. */
b851fbc1
JH
893#define PERL_UNICODE_MAX 0x10FFFF
894
d044b7a7
KW
895#define UNICODE_WARN_SURROGATE 0x0001 /* UTF-16 surrogates */
896#define UNICODE_WARN_NONCHAR 0x0002 /* Non-char code points */
897#define UNICODE_WARN_SUPER 0x0004 /* Above 0x10FFFF */
898#define UNICODE_WARN_PERL_EXTENDED 0x0008 /* Above 0x7FFF_FFFF */
899#define UNICODE_WARN_ABOVE_31_BIT UNICODE_WARN_PERL_EXTENDED
900#define UNICODE_DISALLOW_SURROGATE 0x0010
901#define UNICODE_DISALLOW_NONCHAR 0x0020
902#define UNICODE_DISALLOW_SUPER 0x0040
903#define UNICODE_DISALLOW_PERL_EXTENDED 0x0080
904#define UNICODE_DISALLOW_ABOVE_31_BIT UNICODE_DISALLOW_PERL_EXTENDED
33f38593
KW
905
906#define UNICODE_GOT_SURROGATE UNICODE_DISALLOW_SURROGATE
907#define UNICODE_GOT_NONCHAR UNICODE_DISALLOW_NONCHAR
908#define UNICODE_GOT_SUPER UNICODE_DISALLOW_SUPER
909#define UNICODE_GOT_PERL_EXTENDED UNICODE_DISALLOW_PERL_EXTENDED
910
ecc1615f
KW
911#define UNICODE_WARN_ILLEGAL_C9_INTERCHANGE \
912 (UNICODE_WARN_SURROGATE|UNICODE_WARN_SUPER)
bb88be5f 913#define UNICODE_WARN_ILLEGAL_INTERCHANGE \
ecc1615f
KW
914 (UNICODE_WARN_ILLEGAL_C9_INTERCHANGE|UNICODE_WARN_NONCHAR)
915#define UNICODE_DISALLOW_ILLEGAL_C9_INTERCHANGE \
916 (UNICODE_DISALLOW_SURROGATE|UNICODE_DISALLOW_SUPER)
bb88be5f 917#define UNICODE_DISALLOW_ILLEGAL_INTERCHANGE \
ecc1615f 918 (UNICODE_DISALLOW_ILLEGAL_C9_INTERCHANGE|UNICODE_DISALLOW_NONCHAR)
949cf498
KW
919
920/* For backward source compatibility, as are now the default */
921#define UNICODE_ALLOW_SURROGATE 0
922#define UNICODE_ALLOW_SUPER 0
923#define UNICODE_ALLOW_ANY 0
b851fbc1 924
2d6b3d38
KW
925/* This matches the 2048 code points between UNICODE_SURROGATE_FIRST (0xD800) and
926 * UNICODE_SURROGATE_LAST (0xDFFF) */
927#define UNICODE_IS_SURROGATE(uv) (((UV) (uv) & (~0xFFFF | 0xF800)) \
928 == 0xD800)
929
646d1759
KW
930#define UNICODE_IS_REPLACEMENT(uv) ((UV) (uv) == UNICODE_REPLACEMENT)
931#define UNICODE_IS_BYTE_ORDER_MARK(uv) ((UV) (uv) == UNICODE_BYTE_ORDER_MARK)
c149ab20
KW
932
933/* Is 'uv' one of the 32 contiguous-range noncharacters? */
934#define UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv) ((UV) (uv) >= 0xFDD0 \
935 && (UV) (uv) <= 0xFDEF)
936
937/* Is 'uv' one of the 34 plane-ending noncharacters 0xFFFE, 0xFFFF, 0x1FFFE,
938 * 0x1FFFF, ... 0x10FFFE, 0x10FFFF, given that we know that 'uv' is not above
939 * the Unicode legal max */
940#define UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv) \
941 (((UV) (uv) & 0xFFFE) == 0xFFFE)
942
943#define UNICODE_IS_NONCHAR(uv) \
944 ( UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv) \
945 || ( LIKELY( ! UNICODE_IS_SUPER(uv)) \
946 && UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv)))
947
948#define UNICODE_IS_SUPER(uv) ((UV) (uv) > PERL_UNICODE_MAX)
1d72bdf6 949
ec34087a
KW
950#define LATIN_SMALL_LETTER_SHARP_S LATIN_SMALL_LETTER_SHARP_S_NATIVE
951#define LATIN_SMALL_LETTER_Y_WITH_DIAERESIS \
952 LATIN_SMALL_LETTER_Y_WITH_DIAERESIS_NATIVE
953#define MICRO_SIGN MICRO_SIGN_NATIVE
954#define LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE \
955 LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE_NATIVE
956#define LATIN_SMALL_LETTER_A_WITH_RING_ABOVE \
957 LATIN_SMALL_LETTER_A_WITH_RING_ABOVE_NATIVE
09091399
JH
958#define UNICODE_GREEK_CAPITAL_LETTER_SIGMA 0x03A3
959#define UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA 0x03C2
960#define UNICODE_GREEK_SMALL_LETTER_SIGMA 0x03C3
9dcbe121 961#define GREEK_SMALL_LETTER_MU 0x03BC
9e682c18
KW
962#define GREEK_CAPITAL_LETTER_MU 0x039C /* Upper and title case
963 of MICRON */
964#define LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS 0x0178 /* Also is title case */
0766489e
KW
965#ifdef LATIN_CAPITAL_LETTER_SHARP_S_UTF8
966# define LATIN_CAPITAL_LETTER_SHARP_S 0x1E9E
967#endif
74894415
KW
968#define LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE 0x130
969#define LATIN_SMALL_LETTER_DOTLESS_I 0x131
9e682c18 970#define LATIN_SMALL_LETTER_LONG_S 0x017F
a9f50d33
KW
971#define LATIN_SMALL_LIGATURE_LONG_S_T 0xFB05
972#define LATIN_SMALL_LIGATURE_ST 0xFB06
9e682c18
KW
973#define KELVIN_SIGN 0x212A
974#define ANGSTROM_SIGN 0x212B
09091399 975
9e55ce06 976#define UNI_DISPLAY_ISPRINT 0x0001
c728cb41
JH
977#define UNI_DISPLAY_BACKSLASH 0x0002
978#define UNI_DISPLAY_QQ (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH)
979#define UNI_DISPLAY_REGEX (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH)
9e55ce06 980
5cd46e1f
KW
981#define ANYOF_FOLD_SHARP_S(node, input, end) \
982 (ANYOF_BITMAP_TEST(node, LATIN_SMALL_LETTER_SHARP_S) && \
137165a6 983 (ANYOF_NONBITMAP(node)) && \
39065660 984 (ANYOF_FLAGS(node) & ANYOF_LOC_NONBITMAP_FOLD) && \
07b6858f 985 ((end) > (input) + 1) && \
305b8651 986 isALPHA_FOLD_EQ((input)[0], 's'))
6302f837 987
ebc501f0 988#define SHARP_S_SKIP 2
3b0fc154 989
3cedd9d9 990#define is_utf8_char_buf(buf, buf_end) isUTF8_CHAR(buf, buf_end)
976c1b08
KW
991#define bytes_from_utf8(s, lenp, is_utf8p) \
992 bytes_from_utf8_loc(s, lenp, is_utf8p, 0)
3cedd9d9 993
e23e8bc1
KW
994/*
995
25e3a4e0
KW
996=for apidoc Am|STRLEN|isUTF8_CHAR_flags|const U8 *s|const U8 *e| const U32 flags
997
998Evaluates to non-zero if the first few bytes of the string starting at C<s> and
999looking no further than S<C<e - 1>> are well-formed UTF-8, as extended by Perl,
1000that represents some code point, subject to the restrictions given by C<flags>;
1001otherwise it evaluates to 0. If non-zero, the value gives how many bytes
2717076a
KW
1002starting at C<s> comprise the code point's representation. Any bytes remaining
1003before C<e>, but beyond the ones needed to form the first code point in C<s>,
1004are not examined.
25e3a4e0
KW
1005
1006If C<flags> is 0, this gives the same results as C<L</isUTF8_CHAR>>;
1007if C<flags> is C<UTF8_DISALLOW_ILLEGAL_INTERCHANGE>, this gives the same results
1008as C<L</isSTRICT_UTF8_CHAR>>;
1009and if C<flags> is C<UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE>, this gives
1010the same results as C<L</isC9_STRICT_UTF8_CHAR>>.
1011Otherwise C<flags> may be any combination of the C<UTF8_DISALLOW_I<foo>> flags
1012understood by C<L</utf8n_to_uvchr>>, with the same meanings.
1013
1014The three alternative macros are for the most commonly needed validations; they
1015are likely to run somewhat faster than this more general one, as they can be
1016inlined into your code.
1017
9f2abfde
KW
1018Use L</is_utf8_string_flags>, L</is_utf8_string_loc_flags>, and
1019L</is_utf8_string_loclen_flags> to check entire strings.
1020
25e3a4e0
KW
1021=cut
1022*/
1023
1024#define isUTF8_CHAR_flags(s, e, flags) \
1025 (UNLIKELY((e) <= (s)) \
1026 ? 0 \
1027 : (UTF8_IS_INVARIANT(*s)) \
1028 ? 1 \
1029 : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \
1030 ? 0 \
1376b35c 1031 : is_utf8_char_helper(s, e, flags))
25e3a4e0 1032
6302f837
KW
1033/* Do not use; should be deprecated. Use isUTF8_CHAR() instead; this is
1034 * retained solely for backwards compatibility */
1035#define IS_UTF8_CHAR(p, n) (isUTF8_CHAR(p, (p) + (n)) == n)
e9a8c099 1036
6a5bc5ac 1037#endif /* PERL_UTF8_H_ */
57f0e7e2 1038
e9a8c099 1039/*
14d04a33 1040 * ex: set ts=8 sts=4 sw=4 et:
e9a8c099 1041 */