This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Inline is_utf8_string() and is_utf8_stringloclen()
[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
57f0e7e2
KW
17#ifndef H_UTF8 /* Guard against recursive inclusion */
18#define H_UTF8 1
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
83199d38
KW
37/* For _core_swash_init(), internal core use only */
38#define _CORE_SWASH_INIT_USER_DEFINED_PROPERTY 0x1
5d3d13d1 39#define _CORE_SWASH_INIT_RETURN_IF_UNDEF 0x2
87367d5f 40#define _CORE_SWASH_INIT_ACCEPT_INVLIST 0x4
83199d38 41
7bbfa158
KW
42/*
43=head1 Unicode Support
8cca77bc
KW
44L<perlguts/Unicode Support> has an introduction to this API.
45
46See also L</Character classification>,
47and L</Character case changing>.
48Various functions outside this section also work specially with Unicode.
49Search for the string "utf8" in this document.
7bbfa158
KW
50
51=for apidoc is_ascii_string
52
8871a094 53This is a misleadingly-named synonym for L</is_utf8_invariant_string>.
7bbfa158
KW
54On ASCII-ish platforms, the name isn't misleading: the ASCII-range characters
55are exactly the UTF-8 invariants. But EBCDIC machines have more invariants
8871a094
KW
56than just the ASCII characters, so C<is_utf8_invariant_string> is preferred.
57
58=for apidoc is_invariant_string
59
60This is a somewhat misleadingly-named synonym for L</is_utf8_invariant_string>.
61C<is_utf8_invariant_string> is preferred, as it indicates under what conditions
62the string is invariant.
7bbfa158
KW
63
64=cut
65*/
8871a094
KW
66#define is_ascii_string(s, len) is_utf8_invariant_string(s, len)
67#define is_invariant_string(s, len) is_utf8_invariant_string(s, len)
7bbfa158 68
de69f3af
KW
69#define uvchr_to_utf8(a,b) uvchr_to_utf8_flags(a,b,0)
70#define uvchr_to_utf8_flags(d,uv,flags) \
71 uvoffuni_to_utf8_flags(d,NATIVE_TO_UNI(uv),flags)
72#define utf8_to_uvchr_buf(s, e, lenp) \
842991ae 73 utf8n_to_uvchr(s, (U8*)(e) - (U8*)(s), lenp, \
de69f3af
KW
74 ckWARN_d(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY)
75
a0270393 76#define to_uni_fold(c, p, lenp) _to_uni_fold_flags(c, p, lenp, FOLD_FLAGS_FULL)
445bf929
KW
77#define to_utf8_fold(c, p, lenp) _to_utf8_fold_flags(c, p, lenp, FOLD_FLAGS_FULL)
78#define to_utf8_lower(a,b,c) _to_utf8_lower_flags(a,b,c,0)
79#define to_utf8_upper(a,b,c) _to_utf8_upper_flags(a,b,c,0)
80#define to_utf8_title(a,b,c) _to_utf8_title_flags(a,b,c,0)
36bb2ab6 81
fd7cb289 82/* Source backward compatibility. */
fd7cb289
RGS
83#define is_utf8_string_loc(s, len, ep) is_utf8_string_loclen(s, len, ep, 0)
84
eda9cac1
KW
85#define foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \
86 foldEQ_utf8_flags(s1, pe1, l1, u1, s2, pe2, l2, u2, 0)
baa60164 87#define FOLDEQ_UTF8_NOMIX_ASCII (1 << 0)
cea315b6 88#define FOLDEQ_LOCALE (1 << 1)
18f762c3
KW
89#define FOLDEQ_S1_ALREADY_FOLDED (1 << 2)
90#define FOLDEQ_S2_ALREADY_FOLDED (1 << 3)
d635b710
KW
91#define FOLDEQ_S1_FOLDS_SANE (1 << 4)
92#define FOLDEQ_S2_FOLDS_SANE (1 << 5)
a33c29bc 93
e6226b18
KW
94#define ibcmp_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \
95 cBOOL(! foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2))
96
1d72bdf6
NIS
97#ifdef EBCDIC
98/* The equivalent of these macros but implementing UTF-EBCDIC
99 are in the following header file:
100 */
101
102#include "utfebcdic.h"
fd7cb289 103
d06134e5 104#else /* ! EBCDIC */
73c4f7a1
GS
105START_EXTERN_C
106
111e8ed9
KW
107/* How wide can a single UTF-8 encoded character become in bytes. */
108/* NOTE: Strictly speaking Perl's UTF-8 should not be called UTF-8 since UTF-8
109 * is an encoding of Unicode, and Unicode's upper limit, 0x10FFFF, can be
110 * expressed with 4 bytes. However, Perl thinks of UTF-8 as a way to encode
111 * non-negative integers in a binary format, even those above Unicode */
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
1a3756de 144#if defined(_MSC_VER) && _MSC_VER < 1400
6f6d1bab
TC
145/* older MSVC versions have a smallish macro buffer */
146#define PERL_SMALL_MACRO_BUFFER
147#endif
148
59a449d5
KW
149/* Native character to/from iso-8859-1. Are the identity functions on ASCII
150 * platforms */
6f6d1bab 151#ifdef PERL_SMALL_MACRO_BUFFER
1a3756de
TC
152#define NATIVE_TO_LATIN1(ch) ((U8)(ch))
153#define LATIN1_TO_NATIVE(ch) ((U8)(ch))
6f6d1bab 154#else
4c8cd605
KW
155#define NATIVE_TO_LATIN1(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch)))
156#define LATIN1_TO_NATIVE(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch)))
6f6d1bab 157#endif
59a449d5
KW
158
159/* I8 is an intermediate version of UTF-8 used only in UTF-EBCDIC. We thus
160 * consider it to be identical to UTF-8 on ASCII platforms. Strictly speaking
161 * UTF-8 and UTF-EBCDIC are two different things, but we often conflate them
162 * because they are 8-bit encodings that serve the same purpose in Perl, and
163 * rarely do we need to distinguish them. The term "NATIVE_UTF8" applies to
164 * whichever one is applicable on the current platform */
6f6d1bab 165#ifdef PERL_SMALL_MACRO_BUFFER
1d4ea287
DD
166#define NATIVE_UTF8_TO_I8(ch) ((U8) (ch))
167#define I8_TO_NATIVE_UTF8(ch) ((U8) (ch))
6f6d1bab 168#else
4c8cd605
KW
169#define NATIVE_UTF8_TO_I8(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch)))
170#define I8_TO_NATIVE_UTF8(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch)))
6f6d1bab 171#endif
59a449d5 172
1d72bdf6 173/* Transforms in wide UV chars */
4c8cd605
KW
174#define UNI_TO_NATIVE(ch) ((UV) (ch))
175#define NATIVE_TO_UNI(ch) ((UV) (ch))
d7578b48 176
877d9f0d 177/*
9041c2e3 178
8c007b5a 179 The following table is from Unicode 3.2.
877d9f0d
JH
180
181 Code Points 1st Byte 2nd Byte 3rd Byte 4th Byte
182
375122d7 183 U+0000..U+007F 00..7F
e1b711da 184 U+0080..U+07FF * C2..DF 80..BF
37e2e78e 185 U+0800..U+0FFF E0 * A0..BF 80..BF
375122d7 186 U+1000..U+CFFF E1..EC 80..BF 80..BF
e1b711da 187 U+D000..U+D7FF ED 80..9F 80..BF
537124e4 188 U+D800..U+DFFF ED A0..BF 80..BF (surrogates)
375122d7 189 U+E000..U+FFFF EE..EF 80..BF 80..BF
37e2e78e 190 U+10000..U+3FFFF F0 * 90..BF 80..BF 80..BF
877d9f0d
JH
191 U+40000..U+FFFFF F1..F3 80..BF 80..BF 80..BF
192 U+100000..U+10FFFF F4 80..8F 80..BF 80..BF
b2635aa8
KW
193 Below are non-Unicode code points
194 U+110000..U+13FFFF F4 90..BF 80..BF 80..BF
195 U+110000..U+1FFFFF F5..F7 80..BF 80..BF 80..BF
537124e4 196 U+200000..: F8.. * 88..BF 80..BF 80..BF 80..BF
877d9f0d 197
e1b711da 198Note the gaps before several of the byte entries above marked by '*'. These are
37e2e78e
KW
199caused by legal UTF-8 avoiding non-shortest encodings: it is technically
200possible to UTF-8-encode a single code point in different ways, but that is
201explicitly forbidden, and the shortest possible encoding should always be used
15824458 202(and that is what Perl does). The non-shortest ones are called 'overlongs'.
8c007b5a 203
877d9f0d
JH
204 */
205
8c007b5a
JH
206/*
207 Another way to look at it, as bits:
208
b2635aa8 209 Code Points 1st Byte 2nd Byte 3rd Byte 4th Byte
8c007b5a 210
b2635aa8
KW
211 0aaa aaaa 0aaa aaaa
212 0000 0bbb bbaa aaaa 110b bbbb 10aa aaaa
213 cccc bbbb bbaa aaaa 1110 cccc 10bb bbbb 10aa aaaa
214 00 000d ddcc cccc bbbb bbaa aaaa 1111 0ddd 10cc cccc 10bb bbbb 10aa aaaa
8c007b5a
JH
215
216As you can see, the continuation bytes all begin with C<10>, and the
e1b711da 217leading bits of the start byte tell how many bytes there are in the
8c007b5a
JH
218encoded character.
219
65ab9279
TC
220Perl's extended UTF-8 means we can have start bytes up to FF.
221
8c007b5a
JH
222*/
223
6c88483e 224/* Is the representation of the Unicode code point 'cp' the same regardless of
15824458 225 * being encoded in UTF-8 or not? */
2d1545e5 226#define OFFUNI_IS_INVARIANT(cp) isASCII(cp)
15824458 227
38953e5a
KW
228/* Is the representation of the code point 'cp' the same regardless of
229 * being encoded in UTF-8 or not? 'cp' is native if < 256; Unicode otherwise
230 * */
cf1be84e 231#define UVCHR_IS_INVARIANT(cp) OFFUNI_IS_INVARIANT(cp)
38953e5a 232
c9264833
KW
233/* This defines the bits that are to be in the continuation bytes of a multi-byte
234 * UTF-8 encoded character that mark it is a continuation byte. */
235#define UTF_CONTINUATION_MARK 0x80
236
a95ec4fb 237/* Misleadingly named: is the UTF8-encoded byte 'c' part of a variant sequence
5c06326b
KW
238 * in UTF-8? This is the inverse of UTF8_IS_INVARIANT. The |0 makes sure this
239 * isn't mistakenly called with a ptr argument */
240#define UTF8_IS_CONTINUED(c) (((U8)((c) | 0)) & UTF_CONTINUATION_MARK)
15824458
KW
241
242/* Is the byte 'c' the first byte of a multi-byte UTF8-8 encoded sequence?
243 * This doesn't catch invariants (they are single-byte). It also excludes the
5c06326b
KW
244 * illegal overlong sequences that begin with C0 and C1. The |0 makes sure
245 * this isn't mistakenly called with a ptr argument */
246#define UTF8_IS_START(c) (((U8)((c) | 0)) >= 0xc2)
15824458 247
858cd8ab
KW
248/* For use in UTF8_IS_CONTINUATION() below */
249#define UTF_IS_CONTINUATION_MASK 0xC0
250
15824458 251/* Is the byte 'c' part of a multi-byte UTF8-8 encoded sequence, and not the
5c06326b
KW
252 * first byte thereof? The |0 makes sure this isn't mistakenly called with a
253 * ptr argument */
858cd8ab 254#define UTF8_IS_CONTINUATION(c) \
5c06326b 255 ((((U8)((c) | 0)) & UTF_IS_CONTINUATION_MASK) == UTF_CONTINUATION_MARK)
0ae1fa71 256
15824458
KW
257/* Is the UTF8-encoded byte 'c' the first byte of a two byte sequence? Use
258 * UTF8_IS_NEXT_CHAR_DOWNGRADEABLE() instead if the input isn't known to
259 * be well-formed. Masking with 0xfe allows the low bit to be 0 or 1; thus
5c06326b
KW
260 * this matches 0xc[23]. The |0 makes sure this isn't mistakenly called with a
261 * ptr argument */
262#define UTF8_IS_DOWNGRADEABLE_START(c) ((((U8)((c) | 0)) & 0xfe) == 0xc2)
4ab10950 263
15824458 264/* Is the UTF8-encoded byte 'c' the first byte of a sequence of bytes that
5c06326b
KW
265 * represent a code point > 255? The |0 makes sure this isn't mistakenly
266 * called with a ptr argument */
267#define UTF8_IS_ABOVE_LATIN1(c) (((U8)((c) | 0)) >= 0xc4)
8850bf83 268
15824458
KW
269/* This is the number of low-order bits a continuation byte in a UTF-8 encoded
270 * sequence contributes to the specification of the code point. In the bit
271 * maps above, you see that the first 2 bits are a constant '10', leaving 6 of
272 * real information */
1d72bdf6 273#define UTF_ACCUMULATION_SHIFT 6
b2635aa8 274
fed423a5
KW
275/* ^? is defined to be DEL on ASCII systems. See the definition of toCTRL()
276 * for more */
277#define QUESTION_MARK_CTRL DEL_NATIVE
278
279/* Surrogates, non-character code points and above-Unicode code points are
280 * problematic in some contexts. This allows code that needs to check for
281 * those to to quickly exclude the vast majority of code points it will
282 * encounter */
283#define isUTF8_POSSIBLY_PROBLEMATIC(c) ((U8) c >= 0xED)
284
285#endif /* EBCDIC vs ASCII */
286
287/* 2**UTF_ACCUMULATION_SHIFT - 1 */
288#define UTF_CONTINUATION_MASK ((U8) ((1U << UTF_ACCUMULATION_SHIFT) - 1))
289
7028aeba
KW
290/* Internal macro to be used only in this file to aid in constructing other
291 * publicly accessible macros.
292 * The number of bytes required to express this uv in UTF-8, for just those
293 * uv's requiring 2 through 6 bytes, as these are common to all platforms and
294 * word sizes. The number of bytes needed is given by the number of leading 1
295 * bits in the start byte. There are 32 start bytes that have 2 initial 1 bits
296 * (C0-DF); there are 16 that have 3 initial 1 bits (E0-EF); 8 that have 4
297 * initial 1 bits (F0-F8); 4 that have 5 initial 1 bits (F9-FB), and 2 that
298 * have 6 initial 1 bits (FC-FD). The largest number a string of n bytes can
299 * represent is (the number of possible start bytes for 'n')
300 * * (the number of possiblities for each start byte
301 * The latter in turn is
302 * 2 ** ( (how many continuation bytes there are)
303 * * (the number of bits of information each
304 * continuation byte holds))
305 *
306 * If we were on a platform where we could use a fast find first set bit
307 * instruction (or count leading zeros instruction) this could be replaced by
308 * using that to find the log2 of the uv, and divide that by the number of bits
309 * of information in each continuation byte, adjusting for large cases and how
310 * much information is in a start byte for that length */
72164d3a 311#define __COMMON_UNI_SKIP(uv) \
7028aeba
KW
312 (UV) (uv) < (32 * (1U << ( UTF_ACCUMULATION_SHIFT))) ? 2 : \
313 (UV) (uv) < (16 * (1U << (2 * UTF_ACCUMULATION_SHIFT))) ? 3 : \
314 (UV) (uv) < ( 8 * (1U << (3 * UTF_ACCUMULATION_SHIFT))) ? 4 : \
315 (UV) (uv) < ( 4 * (1U << (4 * UTF_ACCUMULATION_SHIFT))) ? 5 : \
316 (UV) (uv) < ( 2 * (1U << (5 * UTF_ACCUMULATION_SHIFT))) ? 6 :
72164d3a
KW
317
318/* Internal macro to be used only in this file.
319 * This adds to __COMMON_UNI_SKIP the details at this platform's upper range.
fed423a5 320 * For any-sized EBCDIC platforms, or 64-bit ASCII ones, we need one more test
72164d3a
KW
321 * to see if just 7 bytes is needed, or if the maximum is needed. For 32-bit
322 * ASCII platforms, everything is representable by 7 bytes */
fed423a5 323#if defined(UV_IS_QUAD) || defined(EBCDIC)
72164d3a 324# define __BASE_UNI_SKIP(uv) (__COMMON_UNI_SKIP(uv) \
7028aeba 325 (UV) (uv) < ((UV) 1U << (6 * UTF_ACCUMULATION_SHIFT)) ? 7 : UTF8_MAXBYTES)
1d68d6cd 326#else
72164d3a 327# define __BASE_UNI_SKIP(uv) (__COMMON_UNI_SKIP(uv) 7)
1d68d6cd
SC
328#endif
329
fed423a5
KW
330/* The next two macros use the base macro defined above, and add in the tests
331 * at the low-end of the range, for just 1 byte, yielding complete macros,
332 * publicly accessible. */
333
334/* Input is a true Unicode (not-native) code point */
335#define OFFUNISKIP(uv) (OFFUNI_IS_INVARIANT(uv) ? 1 : __BASE_UNI_SKIP(uv))
2084b489 336
5352a763
KW
337/*
338
339=for apidoc Am|STRLEN|UVCHR_SKIP|UV cp
340returns the number of bytes required to represent the code point C<cp> when
341encoded as UTF-8. C<cp> is a native (ASCII or EBCDIC) code point if less than
342255; a Unicode code point otherwise.
343
344=cut
345 */
fdb6583d 346#define UVCHR_SKIP(uv) ( UVCHR_IS_INVARIANT(uv) ? 1 : __BASE_UNI_SKIP(uv))
5352a763 347
b651802e
KW
348/* The largest code point representable by two UTF-8 bytes on this platform.
349 * As explained in the comments for __COMMON_UNI_SKIP, 32 start bytes with
fed423a5 350 * UTF_ACCUMULATION_SHIFT bits of information each */
aa206fb7
KW
351#define MAX_UTF8_TWO_BYTE (32 * (1U << UTF_ACCUMULATION_SHIFT) - 1)
352
b651802e
KW
353/* The largest code point representable by two UTF-8 bytes on any platform that
354 * Perl runs on. This value is constrained by EBCDIC which has 5 bits per
355 * continuation byte */
aa206fb7
KW
356#define MAX_PORTABLE_UTF8_TWO_BYTE (32 * (1U << 5) - 1)
357
c03c0950
KW
358/* The maximum number of UTF-8 bytes a single Unicode character can
359 * uppercase/lowercase/fold into. Unicode guarantees that the maximum
360 * expansion is UTF8_MAX_FOLD_CHAR_EXPAND characters, but any above-Unicode
361 * code point will fold to itself, so we only have to look at the expansion of
362 * the maximum Unicode code point. But this number may be less than the space
363 * occupied by a very large code point under Perl's extended UTF-8. We have to
364 * make it large enough to fit any single character. (It turns out that ASCII
365 * and EBCDIC differ in which is larger) */
366#define UTF8_MAXBYTES_CASE \
367 (UTF8_MAXBYTES >= (UTF8_MAX_FOLD_CHAR_EXPAND * OFFUNISKIP(0x10FFFF)) \
368 ? UTF8_MAXBYTES \
369 : (UTF8_MAX_FOLD_CHAR_EXPAND * OFFUNISKIP(0x10FFFF)))
370
d06134e5
KW
371/* Rest of these are attributes of Unicode and perl's internals rather than the
372 * encoding, or happen to be the same in both ASCII and EBCDIC (at least at
373 * this level; the macros that some of these call may have different
374 * definitions in the two encodings */
375
59a449d5
KW
376/* In domain restricted to ASCII, these may make more sense to the reader than
377 * the ones with Latin1 in the name */
378#define NATIVE_TO_ASCII(ch) NATIVE_TO_LATIN1(ch)
379#define ASCII_TO_NATIVE(ch) LATIN1_TO_NATIVE(ch)
380
381/* More or less misleadingly-named defines, retained for back compat */
382#define NATIVE_TO_UTF(ch) NATIVE_UTF8_TO_I8(ch)
383#define NATIVE_TO_I8(ch) NATIVE_UTF8_TO_I8(ch)
384#define UTF_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch)
385#define I8_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch)
386#define NATIVE8_TO_UNI(ch) NATIVE_TO_LATIN1(ch)
d06134e5 387
c0236afe 388/* This defines the 1-bits that are to be in the first byte of a multi-byte
97d0ceda
KW
389 * UTF-8 encoded character that mark it as a start byte and give the number of
390 * bytes that comprise the character. 'len' is the number of bytes in the
391 * multi-byte sequence. */
c0236afe
KW
392#define UTF_START_MARK(len) (((len) > 7) ? 0xFF : (0xFF & (0xFE << (7-(len)))))
393
394/* Masks out the initial one bits in a start byte, leaving the real data ones.
395 * Doesn't work on an invariant byte. 'len' is the number of bytes in the
396 * multi-byte sequence that comprises the character. */
397#define UTF_START_MASK(len) (((len) >= 7) ? 0x00 : (0x1F >> ((len)-2)))
398
537124e4
KW
399/* Adds a UTF8 continuation byte 'new' of information to a running total code
400 * point 'old' of all the continuation bytes so far. This is designed to be
155d2738
KW
401 * used in a loop to convert from UTF-8 to the code point represented. Note
402 * that this is asymmetric on EBCDIC platforms, in that the 'new' parameter is
403 * the UTF-EBCDIC byte, whereas the 'old' parameter is a Unicode (not EBCDIC)
404 * code point in process of being generated */
405#define UTF8_ACCUMULATE(old, new) (((old) << UTF_ACCUMULATION_SHIFT) \
406 | ((NATIVE_UTF8_TO_I8((U8)new)) \
407 & UTF_CONTINUATION_MASK))
d06134e5 408
7c560c3b
KW
409/* If a value is anded with this, and the result is non-zero, then using the
410 * original value in UTF8_ACCUMULATE will overflow, shifting bits off the left
411 * */
412#define UTF_ACCUMULATION_OVERFLOW_MASK \
413 (((UV) UTF_CONTINUATION_MASK) << ((sizeof(UV) * CHARBITS) \
414 - UTF_ACCUMULATION_SHIFT))
415
4ab10950
KW
416/* This works in the face of malformed UTF-8. */
417#define UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(s, e) (UTF8_IS_DOWNGRADEABLE_START(*s) \
418 && ( (e) - (s) > 1) \
419 && UTF8_IS_CONTINUATION(*((s)+1)))
420
5aaebcb3 421/* Number of bytes a code point occupies in UTF-8. */
5352a763 422#define NATIVE_SKIP(uv) UVCHR_SKIP(uv)
bd18bd40 423
5aaebcb3
KW
424/* Most code which says UNISKIP is really thinking in terms of native code
425 * points (0-255) plus all those beyond. This is an imprecise term, but having
2accb712 426 * it means existing code continues to work. For precision, use UVCHR_SKIP,
5352a763
KW
427 * NATIVE_SKIP, or OFFUNISKIP */
428#define UNISKIP(uv) UVCHR_SKIP(uv)
5aaebcb3 429
3c0792e4
KW
430/* Longer, but more accurate name */
431#define UTF8_IS_ABOVE_LATIN1_START(c) UTF8_IS_ABOVE_LATIN1(c)
432
a62b247b
KW
433/* Convert a UTF-8 variant Latin1 character to a native code point value.
434 * Needs just one iteration of accumulate. Should be used only if it is known
435 * that the code point is < 256, and is not UTF-8 invariant. Use the slower
436 * but more general TWO_BYTE_UTF8_TO_NATIVE() which handles any code point
437 * representable by two bytes (which turns out to be up through
438 * MAX_PORTABLE_UTF8_TWO_BYTE). The two parameters are:
439 * HI: a downgradable start byte;
440 * LO: continuation.
441 * */
442#define EIGHT_BIT_UTF8_TO_NATIVE(HI, LO) \
443 ( __ASSERT_(UTF8_IS_DOWNGRADEABLE_START(HI)) \
444 __ASSERT_(UTF8_IS_CONTINUATION(LO)) \
445 LATIN1_TO_NATIVE(UTF8_ACCUMULATE(( \
446 NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), (LO))))
447
94bb8c36 448/* Convert a two (not one) byte utf8 character to a native code point value.
2950f2a7
KW
449 * Needs just one iteration of accumulate. Should not be used unless it is
450 * known that the two bytes are legal: 1) two-byte start, and 2) continuation.
451 * Note that the result can be larger than 255 if the input character is not
452 * downgradable */
94bb8c36 453#define TWO_BYTE_UTF8_TO_NATIVE(HI, LO) \
b18b15ad 454 ( __ASSERT_(PL_utf8skip[HI] == 2) \
635e76f5 455 __ASSERT_(UTF8_IS_CONTINUATION(LO)) \
94bb8c36 456 UNI_TO_NATIVE(UTF8_ACCUMULATE((NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), \
635e76f5 457 (LO))))
94bb8c36
KW
458
459/* Should never be used, and be deprecated */
460#define TWO_BYTE_UTF8_TO_UNI(HI, LO) NATIVE_TO_UNI(TWO_BYTE_UTF8_TO_NATIVE(HI, LO))
2950f2a7 461
bd18bd40
KW
462/*
463
464=for apidoc Am|STRLEN|UTF8SKIP|char* s
465returns the number of bytes in the UTF-8 encoded character whose first (perhaps
466only) byte is pointed to by C<s>.
467
468=cut
469 */
2a70536e
KW
470#define UTF8SKIP(s) PL_utf8skip[*(const U8*)(s)]
471#define UTF8_SKIP(s) UTF8SKIP(s)
d06134e5 472
2d1545e5
KW
473/* Most code that says 'UNI_' really means the native value for code points up
474 * through 255 */
475#define UNI_IS_INVARIANT(cp) UVCHR_IS_INVARIANT(cp)
476
15824458
KW
477/* Is the byte 'c' the same character when encoded in UTF-8 as when not. This
478 * works on both UTF-8 encoded strings and non-encoded, as it returns TRUE in
5fc230f1
KW
479 * each for the exact same set of bit patterns. It is valid on a subset of
480 * what UVCHR_IS_INVARIANT is valid on, so can just use that; and the compiler
481 * should optimize out anything extraneous given the implementation of the
5c06326b
KW
482 * latter. The |0 makes sure this isn't mistakenly called with a ptr argument.
483 * */
484#define UTF8_IS_INVARIANT(c) UVCHR_IS_INVARIANT((c) | 0)
5fc230f1
KW
485
486/* Like the above, but its name implies a non-UTF8 input, which as the comments
487 * above show, doesn't matter as to its implementation */
38953e5a 488#define NATIVE_BYTE_IS_INVARIANT(c) UVCHR_IS_INVARIANT(c)
d06134e5 489
48ccf5e1
KW
490/* The macros in the next 4 sets are used to generate the two utf8 or utfebcdic
491 * bytes from an ordinal that is known to fit into exactly two (not one) bytes;
492 * it must be less than 0x3FF to work across both encodings. */
493
494/* These two are helper macros for the other three sets, and should not be used
495 * directly anywhere else. 'translate_function' is either NATIVE_TO_LATIN1
1ff3baa2
KW
496 * (which works for code points up through 0xFF) or NATIVE_TO_UNI which works
497 * for any code point */
48ccf5e1 498#define __BASE_TWO_BYTE_HI(c, translate_function) \
2863dafa 499 (__ASSERT_(! UVCHR_IS_INVARIANT(c)) \
48ccf5e1 500 I8_TO_NATIVE_UTF8((translate_function(c) >> UTF_ACCUMULATION_SHIFT) \
2863dafa 501 | UTF_START_MARK(2)))
48ccf5e1 502#define __BASE_TWO_BYTE_LO(c, translate_function) \
2863dafa 503 (__ASSERT_(! UVCHR_IS_INVARIANT(c)) \
48ccf5e1 504 I8_TO_NATIVE_UTF8((translate_function(c) & UTF_CONTINUATION_MASK) \
2863dafa 505 | UTF_CONTINUATION_MARK))
48ccf5e1 506
48ccf5e1
KW
507/* The next two macros should not be used. They were designed to be usable as
508 * the case label of a switch statement, but this doesn't work for EBCDIC. Use
9d0d3a03 509 * regen/unicode_constants.pl instead */
48ccf5e1
KW
510#define UTF8_TWO_BYTE_HI_nocast(c) __BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI)
511#define UTF8_TWO_BYTE_LO_nocast(c) __BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI)
512
513/* The next two macros are used when the source should be a single byte
514 * character; checked for under DEBUGGING */
515#define UTF8_EIGHT_BIT_HI(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
4c8cd605 516 ( __BASE_TWO_BYTE_HI(c, NATIVE_TO_LATIN1)))
48ccf5e1 517#define UTF8_EIGHT_BIT_LO(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
4c8cd605 518 (__BASE_TWO_BYTE_LO(c, NATIVE_TO_LATIN1)))
48ccf5e1
KW
519
520/* These final two macros in the series are used when the source can be any
521 * code point whose UTF-8 is known to occupy 2 bytes; they are less efficient
522 * than the EIGHT_BIT versions on EBCDIC platforms. We use the logical '~'
523 * operator instead of "<=" to avoid getting compiler warnings.
d52b8576 524 * MAX_UTF8_TWO_BYTE should be exactly all one bits in the lower few
48ccf5e1
KW
525 * places, so the ~ works */
526#define UTF8_TWO_BYTE_HI(c) \
527 (__ASSERT_((sizeof(c) == 1) \
d52b8576 528 || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \
4c8cd605 529 (__BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI)))
48ccf5e1
KW
530#define UTF8_TWO_BYTE_LO(c) \
531 (__ASSERT_((sizeof(c) == 1) \
d52b8576 532 || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \
4c8cd605 533 (__BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI)))
d06134e5 534
e7214ce8
KW
535/* This is illegal in any well-formed UTF-8 in both EBCDIC and ASCII
536 * as it is only in overlongs. */
537#define ILLEGAL_UTF8_BYTE I8_TO_NATIVE_UTF8(0xC1)
538
7e2040f0 539/*
e3036cf4 540 * 'UTF' is whether or not p is encoded in UTF8. The names 'foo_lazy_if' stem
20df05f4
KW
541 * from an earlier version of these macros in which they didn't call the
542 * foo_utf8() macros (i.e. were 'lazy') unless they decided that *p is the
543 * beginning of a utf8 character. Now that foo_utf8() determines that itself,
544 * no need to do it again here
7e2040f0 545 */
97d0ceda
KW
546#define isIDFIRST_lazy_if(p,UTF) ((IN_BYTES || !UTF) \
547 ? isIDFIRST(*(p)) \
e3036cf4 548 : isIDFIRST_utf8((const U8*)p))
97d0ceda
KW
549#define isWORDCHAR_lazy_if(p,UTF) ((IN_BYTES || (!UTF)) \
550 ? isWORDCHAR(*(p)) \
32636478
KW
551 : isWORDCHAR_utf8((const U8*)p))
552#define isALNUM_lazy_if(p,UTF) isWORDCHAR_lazy_if(p,UTF)
1d72bdf6 553
89ebb4a3
JH
554#define UTF8_MAXLEN UTF8_MAXBYTES
555
8cb75cc8
KW
556/* A Unicode character can fold to up to 3 characters */
557#define UTF8_MAX_FOLD_CHAR_EXPAND 3
558
a98fe34d 559#define IN_BYTES (CopHINTS_get(PL_curcop) & HINT_BYTES)
bd18bd40
KW
560
561/*
562
563=for apidoc Am|bool|DO_UTF8|SV* sv
564Returns a bool giving whether or not the PV in C<sv> is to be treated as being
565encoded in UTF-8.
566
567You should use this I<after> a call to C<SvPV()> or one of its variants, in
568case any call to string overloading updates the internal UTF-8 encoding flag.
569
570=cut
571*/
0064a8a9 572#define DO_UTF8(sv) (SvUTF8(sv) && !IN_BYTES)
1ff3baa2
KW
573
574/* Should all strings be treated as Unicode, and not just UTF-8 encoded ones?
575 * Is so within 'feature unicode_strings' or 'locale :not_characters', and not
576 * within 'use bytes'. UTF-8 locales are not tested for here, but perhaps
577 * could be */
578#define IN_UNI_8_BIT \
d6ded950
KW
579 (((CopHINTS_get(PL_curcop) & (HINT_UNI_8_BIT)) \
580 || (CopHINTS_get(PL_curcop) & HINT_LOCALE_PARTIAL \
581 /* -1 below is for :not_characters */ \
582 && _is_in_locale_category(FALSE, -1))) \
583 && ! IN_BYTES)
b36bf33f 584
1d72bdf6 585
c76687c5
KW
586#define UTF8_ALLOW_EMPTY 0x0001 /* Allow a zero length string */
587
588/* Allow first byte to be a continuation byte */
1d72bdf6 589#define UTF8_ALLOW_CONTINUATION 0x0002
c76687c5
KW
590
591/* Allow second... bytes to be non-continuation bytes */
1d72bdf6 592#define UTF8_ALLOW_NON_CONTINUATION 0x0004
949cf498
KW
593
594/* expecting more bytes than were available in the string */
595#define UTF8_ALLOW_SHORT 0x0008
596
597/* Overlong sequence; i.e., the code point can be specified in fewer bytes. */
598#define UTF8_ALLOW_LONG 0x0010
599
600#define UTF8_DISALLOW_SURROGATE 0x0020 /* Unicode surrogates */
601#define UTF8_WARN_SURROGATE 0x0040
602
603#define UTF8_DISALLOW_NONCHAR 0x0080 /* Unicode non-character */
604#define UTF8_WARN_NONCHAR 0x0100 /* code points */
605
606#define UTF8_DISALLOW_SUPER 0x0200 /* Super-set of Unicode: code */
607#define UTF8_WARN_SUPER 0x0400 /* points above the legal max */
608
d35f2ca5
KW
609/* Code points which never were part of the original UTF-8 standard, which only
610 * went up to 2 ** 31 - 1. Note that these all overflow a signed 32-bit word,
611 * The first byte of these code points is FE or FF on ASCII platforms. If the
612 * first byte is FF, it will overflow a 32-bit word. */
613#define UTF8_DISALLOW_ABOVE_31_BIT 0x0800
614#define UTF8_WARN_ABOVE_31_BIT 0x1000
615
616/* For back compat, these old names are misleading for UTF_EBCDIC */
617#define UTF8_DISALLOW_FE_FF UTF8_DISALLOW_ABOVE_31_BIT
618#define UTF8_WARN_FE_FF UTF8_WARN_ABOVE_31_BIT
949cf498
KW
619
620#define UTF8_CHECK_ONLY 0x2000
621
622/* For backwards source compatibility. They do nothing, as the default now
623 * includes what they used to mean. The first one's meaning was to allow the
624 * just the single non-character 0xFFFF */
625#define UTF8_ALLOW_FFFF 0
626#define UTF8_ALLOW_SURROGATE 0
627
d35f2ca5 628#define UTF8_DISALLOW_ILLEGAL_INTERCHANGE \
93e6dbd6
KW
629 ( UTF8_DISALLOW_SUPER|UTF8_DISALLOW_NONCHAR \
630 |UTF8_DISALLOW_SURROGATE)
949cf498 631#define UTF8_WARN_ILLEGAL_INTERCHANGE \
93e6dbd6
KW
632 (UTF8_WARN_SUPER|UTF8_WARN_NONCHAR|UTF8_WARN_SURROGATE)
633#define UTF8_ALLOW_ANY \
634 (~( UTF8_DISALLOW_ILLEGAL_INTERCHANGE|UTF8_DISALLOW_ABOVE_31_BIT \
635 |UTF8_WARN_ILLEGAL_INTERCHANGE|UTF8_WARN_ABOVE_31_BIT))
949cf498
KW
636#define UTF8_ALLOW_ANYUV \
637 (UTF8_ALLOW_EMPTY \
638 & ~(UTF8_DISALLOW_ILLEGAL_INTERCHANGE|UTF8_WARN_ILLEGAL_INTERCHANGE))
9f7f3913
ST
639#define UTF8_ALLOW_DEFAULT (ckWARN(WARN_UTF8) ? 0 : \
640 UTF8_ALLOW_ANYUV)
1d72bdf6 641
0c58a72b
KW
642/* Several of the macros below have a second parameter that is currently
643 * unused; but could be used in the future to make sure that the input is
644 * well-formed. */
645
646#define UTF8_IS_SURROGATE(s, e) cBOOL(is_SURROGATE_utf8(s))
b96a92fb 647#define UTF8_IS_REPLACEMENT(s, send) cBOOL(is_REPLACEMENT_utf8_safe(s,send))
7131f24d
KW
648
649/* ASCII EBCDIC I8
650 * U+10FFFF: \xF4\x8F\xBF\xBF \xF9\xA1\xBF\xBF\xBF max legal Unicode
651 * U+110000: \xF4\x90\x80\x80 \xF9\xA2\xA0\xA0\xA0
652 * U+110001: \xF4\x90\x80\x81 \xF9\xA2\xA0\xA0\xA1
a1776718
KW
653 *
654 * BE AWARE that this test doesn't rule out malformed code points, in
655 * particular overlongs */
7131f24d 656#ifdef EBCDIC /* Both versions assume well-formed UTF8 */
0c58a72b 657# define UTF8_IS_SUPER(s, e) (NATIVE_UTF8_TO_I8(* (U8*) (s)) >= 0xF9 \
bc3632a8 658 && (NATIVE_UTF8_TO_I8(* (U8*) (s)) > 0xF9 \
a3481822 659 || (NATIVE_UTF8_TO_I8(* ((U8*) (s) + 1)) >= 0xA2)))
7131f24d 660#else
0c58a72b
KW
661# define UTF8_IS_SUPER(s, e) (*(U8*) (s) >= 0xF4 \
662 && (*(U8*) (s) > 0xF4 || (*((U8*) (s) + 1) >= 0x90)))
7131f24d
KW
663#endif
664
b96a92fb
KW
665/* These are now machine generated, and the 'given' clause is no longer
666 * applicable */
0c58a72b 667#define UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s, e) \
b96a92fb 668 cBOOL(is_NONCHAR_utf8(s))
0c58a72b
KW
669#define UTF8_IS_NONCHAR(s, e) \
670 UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s, e)
7131f24d 671
c867b360
JH
672#define UNICODE_SURROGATE_FIRST 0xD800
673#define UNICODE_SURROGATE_LAST 0xDFFF
674#define UNICODE_REPLACEMENT 0xFFFD
675#define UNICODE_BYTE_ORDER_MARK 0xFEFF
1d72bdf6 676
b851fbc1 677/* Though our UTF-8 encoding can go beyond this,
c76687c5 678 * let's be conservative and do as Unicode says. */
b851fbc1
JH
679#define PERL_UNICODE_MAX 0x10FFFF
680
d35f2ca5
KW
681#define UNICODE_WARN_SURROGATE 0x0001 /* UTF-16 surrogates */
682#define UNICODE_WARN_NONCHAR 0x0002 /* Non-char code points */
683#define UNICODE_WARN_SUPER 0x0004 /* Above 0x10FFFF */
684#define UNICODE_WARN_ABOVE_31_BIT 0x0008 /* Above 0x7FFF_FFFF */
685#define UNICODE_DISALLOW_SURROGATE 0x0010
686#define UNICODE_DISALLOW_NONCHAR 0x0020
687#define UNICODE_DISALLOW_SUPER 0x0040
688#define UNICODE_DISALLOW_ABOVE_31_BIT 0x0080
bb88be5f
KW
689#define UNICODE_WARN_ILLEGAL_INTERCHANGE \
690 (UNICODE_WARN_SURROGATE|UNICODE_WARN_NONCHAR|UNICODE_WARN_SUPER)
691#define UNICODE_DISALLOW_ILLEGAL_INTERCHANGE \
692 (UNICODE_DISALLOW_SURROGATE|UNICODE_DISALLOW_NONCHAR|UNICODE_DISALLOW_SUPER)
949cf498
KW
693
694/* For backward source compatibility, as are now the default */
695#define UNICODE_ALLOW_SURROGATE 0
696#define UNICODE_ALLOW_SUPER 0
697#define UNICODE_ALLOW_ANY 0
b851fbc1 698
2d6b3d38
KW
699/* This matches the 2048 code points between UNICODE_SURROGATE_FIRST (0xD800) and
700 * UNICODE_SURROGATE_LAST (0xDFFF) */
701#define UNICODE_IS_SURROGATE(uv) (((UV) (uv) & (~0xFFFF | 0xF800)) \
702 == 0xD800)
703
646d1759
KW
704#define UNICODE_IS_REPLACEMENT(uv) ((UV) (uv) == UNICODE_REPLACEMENT)
705#define UNICODE_IS_BYTE_ORDER_MARK(uv) ((UV) (uv) == UNICODE_BYTE_ORDER_MARK)
c149ab20
KW
706
707/* Is 'uv' one of the 32 contiguous-range noncharacters? */
708#define UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv) ((UV) (uv) >= 0xFDD0 \
709 && (UV) (uv) <= 0xFDEF)
710
711/* Is 'uv' one of the 34 plane-ending noncharacters 0xFFFE, 0xFFFF, 0x1FFFE,
712 * 0x1FFFF, ... 0x10FFFE, 0x10FFFF, given that we know that 'uv' is not above
713 * the Unicode legal max */
714#define UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv) \
715 (((UV) (uv) & 0xFFFE) == 0xFFFE)
716
717#define UNICODE_IS_NONCHAR(uv) \
718 ( UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv) \
719 || ( LIKELY( ! UNICODE_IS_SUPER(uv)) \
720 && UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv)))
721
722#define UNICODE_IS_SUPER(uv) ((UV) (uv) > PERL_UNICODE_MAX)
d35f2ca5 723#define UNICODE_IS_ABOVE_31_BIT(uv) ((UV) (uv) > 0x7FFFFFFF)
1d72bdf6 724
ec34087a
KW
725#define LATIN_SMALL_LETTER_SHARP_S LATIN_SMALL_LETTER_SHARP_S_NATIVE
726#define LATIN_SMALL_LETTER_Y_WITH_DIAERESIS \
727 LATIN_SMALL_LETTER_Y_WITH_DIAERESIS_NATIVE
728#define MICRO_SIGN MICRO_SIGN_NATIVE
729#define LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE \
730 LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE_NATIVE
731#define LATIN_SMALL_LETTER_A_WITH_RING_ABOVE \
732 LATIN_SMALL_LETTER_A_WITH_RING_ABOVE_NATIVE
09091399
JH
733#define UNICODE_GREEK_CAPITAL_LETTER_SIGMA 0x03A3
734#define UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA 0x03C2
735#define UNICODE_GREEK_SMALL_LETTER_SIGMA 0x03C3
9dcbe121 736#define GREEK_SMALL_LETTER_MU 0x03BC
9e682c18
KW
737#define GREEK_CAPITAL_LETTER_MU 0x039C /* Upper and title case
738 of MICRON */
739#define LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS 0x0178 /* Also is title case */
0766489e
KW
740#ifdef LATIN_CAPITAL_LETTER_SHARP_S_UTF8
741# define LATIN_CAPITAL_LETTER_SHARP_S 0x1E9E
742#endif
74894415
KW
743#define LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE 0x130
744#define LATIN_SMALL_LETTER_DOTLESS_I 0x131
9e682c18 745#define LATIN_SMALL_LETTER_LONG_S 0x017F
a9f50d33
KW
746#define LATIN_SMALL_LIGATURE_LONG_S_T 0xFB05
747#define LATIN_SMALL_LIGATURE_ST 0xFB06
9e682c18
KW
748#define KELVIN_SIGN 0x212A
749#define ANGSTROM_SIGN 0x212B
09091399 750
9e55ce06 751#define UNI_DISPLAY_ISPRINT 0x0001
c728cb41
JH
752#define UNI_DISPLAY_BACKSLASH 0x0002
753#define UNI_DISPLAY_QQ (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH)
754#define UNI_DISPLAY_REGEX (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH)
9e55ce06 755
5cd46e1f
KW
756#define ANYOF_FOLD_SHARP_S(node, input, end) \
757 (ANYOF_BITMAP_TEST(node, LATIN_SMALL_LETTER_SHARP_S) && \
137165a6 758 (ANYOF_NONBITMAP(node)) && \
39065660 759 (ANYOF_FLAGS(node) & ANYOF_LOC_NONBITMAP_FOLD) && \
07b6858f 760 ((end) > (input) + 1) && \
305b8651 761 isALPHA_FOLD_EQ((input)[0], 's'))
6302f837 762
ebc501f0 763#define SHARP_S_SKIP 2
3b0fc154 764
a4f7a67c
KW
765/* If you want to exclude surrogates, and beyond legal Unicode, see the blame
766 * log for earlier versions which gave details for these */
4d646140 767
6302f837
KW
768/* A helper macro for isUTF8_CHAR, so use that one, and not this one. This is
769 * retained solely for backwards compatibility and may be deprecated and
770 * removed in a future Perl version.
771 *
772 * regen/regcharclass.pl generates is_UTF8_CHAR_utf8() macros for up to these
d9f92374
KW
773 * number of bytes. So this has to be coordinated with that file */
774#ifdef EBCDIC
775# define IS_UTF8_CHAR_FAST(n) ((n) <= 3)
776#else
777# define IS_UTF8_CHAR_FAST(n) ((n) <= 4)
778#endif
779
4d646140 780#ifndef EBCDIC
6302f837
KW
781/* A helper macro for isUTF8_CHAR, so use that one instead of this. This was
782 * generated by regen/regcharclass.pl, and then moved here. The lines that
783 * generated it were then commented out. This was done solely because it takes
784 * on the order of 10 minutes to generate, and is never going to change, unless
785 * the generated code is improved.
786 *
1ff3baa2
KW
787 * The EBCDIC versions have been cut to not cover all of legal Unicode,
788 * otherwise they take too long to generate; besides there is a separate one
789 * for each code page, so they are in regcharclass.h instead of here */
39a0f513 790/*
5dca9278 791 UTF8_CHAR: Matches legal UTF-8 encoded characters from 2 through 4 bytes
39a0f513 792
5dca9278 793 0x80 - 0x1FFFFF
39a0f513 794*/
4d646140 795/*** GENERATED CODE ***/
5dca9278
KW
796#define is_UTF8_CHAR_utf8_no_length_checks(s) \
797( ( 0xC2 <= ((U8*)s)[0] && ((U8*)s)[0] <= 0xDF ) ? \
798 ( ( ( ((U8*)s)[1] & 0xC0 ) == 0x80 ) ? 2 : 0 ) \
799: ( 0xE0 == ((U8*)s)[0] ) ? \
800 ( ( ( ( ((U8*)s)[1] & 0xE0 ) == 0xA0 ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
801: ( 0xE1 <= ((U8*)s)[0] && ((U8*)s)[0] <= 0xEF ) ? \
802 ( ( ( ( ((U8*)s)[1] & 0xC0 ) == 0x80 ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
803: ( 0xF0 == ((U8*)s)[0] ) ? \
804 ( ( ( ( 0x90 <= ((U8*)s)[1] && ((U8*)s)[1] <= 0xBF ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) && ( ( ((U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\
805: ( ( ( ( 0xF1 <= ((U8*)s)[0] && ((U8*)s)[0] <= 0xF7 ) && ( ( ((U8*)s)[1] & 0xC0 ) == 0x80 ) ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) && ( ( ((U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )
4d646140 806#endif
3b0fc154 807
6302f837 808/*
5dca9278
KW
809
810=for apidoc Am|STRLEN|isUTF8_CHAR|const U8 *s|const U8 *e
811
812Returns the number of bytes beginning at C<s> which form a legal UTF-8 (or
61b16eb9
KW
813UTF-EBCDIC) encoded character, looking no further than S<C<e - s>> bytes into
814C<s>. Returns 0 if the sequence starting at C<s> through S<C<e - 1>> is not
815well-formed UTF-8.
6302f837
KW
816
817Note that an INVARIANT character (i.e. ASCII on non-EBCDIC
5dca9278
KW
818machines) is a valid UTF-8 character.
819
820=cut
821*/
6302f837 822
dd9bc2b0 823#define isUTF8_CHAR(s, e) (UNLIKELY((e) <= (s)) \
6302f837
KW
824 ? 0 \
825 : (UTF8_IS_INVARIANT(*s)) \
826 ? 1 \
dd9bc2b0 827 : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \
6302f837 828 ? 0 \
dd9bc2b0 829 : LIKELY(IS_UTF8_CHAR_FAST(UTF8SKIP(s))) \
5dca9278 830 ? is_UTF8_CHAR_utf8_no_length_checks(s) \
6302f837
KW
831 : _is_utf8_char_slow(s, e))
832
3cedd9d9
KW
833#define is_utf8_char_buf(buf, buf_end) isUTF8_CHAR(buf, buf_end)
834
6302f837
KW
835/* Do not use; should be deprecated. Use isUTF8_CHAR() instead; this is
836 * retained solely for backwards compatibility */
837#define IS_UTF8_CHAR(p, n) (isUTF8_CHAR(p, (p) + (n)) == n)
e9a8c099 838
57f0e7e2
KW
839#endif /* H_UTF8 */
840
e9a8c099 841/*
14d04a33 842 * ex: set ts=8 sts=4 sw=4 et:
e9a8c099 843 */