This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
utf8.h: Add a #define
[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
KW
32/* For to_utf8_fold_flags, q.v. */
33#define FOLD_FLAGS_LOCALE 0x1
34#define FOLD_FLAGS_FULL 0x2
a0270393 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
44
45=for apidoc is_ascii_string
46
47This is a misleadingly-named synonym for L</is_invariant_string>.
48On ASCII-ish platforms, the name isn't misleading: the ASCII-range characters
49are exactly the UTF-8 invariants. But EBCDIC machines have more invariants
50than just the ASCII characters, so C<is_invariant_string> is preferred.
51
52=cut
53*/
54#define is_ascii_string(s, len) is_invariant_string(s, len)
55
de69f3af
KW
56#define uvchr_to_utf8(a,b) uvchr_to_utf8_flags(a,b,0)
57#define uvchr_to_utf8_flags(d,uv,flags) \
58 uvoffuni_to_utf8_flags(d,NATIVE_TO_UNI(uv),flags)
59#define utf8_to_uvchr_buf(s, e, lenp) \
842991ae 60 utf8n_to_uvchr(s, (U8*)(e) - (U8*)(s), lenp, \
de69f3af
KW
61 ckWARN_d(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY)
62
a0270393 63#define to_uni_fold(c, p, lenp) _to_uni_fold_flags(c, p, lenp, FOLD_FLAGS_FULL)
445bf929
KW
64#define to_utf8_fold(c, p, lenp) _to_utf8_fold_flags(c, p, lenp, FOLD_FLAGS_FULL)
65#define to_utf8_lower(a,b,c) _to_utf8_lower_flags(a,b,c,0)
66#define to_utf8_upper(a,b,c) _to_utf8_upper_flags(a,b,c,0)
67#define to_utf8_title(a,b,c) _to_utf8_title_flags(a,b,c,0)
36bb2ab6 68
fd7cb289 69/* Source backward compatibility. */
fd7cb289
RGS
70#define is_utf8_string_loc(s, len, ep) is_utf8_string_loclen(s, len, ep, 0)
71
eda9cac1
KW
72#define foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \
73 foldEQ_utf8_flags(s1, pe1, l1, u1, s2, pe2, l2, u2, 0)
baa60164 74#define FOLDEQ_UTF8_NOMIX_ASCII (1 << 0)
cea315b6 75#define FOLDEQ_LOCALE (1 << 1)
18f762c3
KW
76#define FOLDEQ_S1_ALREADY_FOLDED (1 << 2)
77#define FOLDEQ_S2_ALREADY_FOLDED (1 << 3)
d635b710
KW
78#define FOLDEQ_S1_FOLDS_SANE (1 << 4)
79#define FOLDEQ_S2_FOLDS_SANE (1 << 5)
a33c29bc 80
e6226b18
KW
81#define ibcmp_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \
82 cBOOL(! foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2))
83
1d72bdf6
NIS
84#ifdef EBCDIC
85/* The equivalent of these macros but implementing UTF-EBCDIC
86 are in the following header file:
87 */
88
89#include "utfebcdic.h"
fd7cb289 90
d06134e5 91#else /* ! EBCDIC */
73c4f7a1
GS
92START_EXTERN_C
93
a0ed51b3 94#ifdef DOINIT
6f06b55f 95EXTCONST unsigned char PL_utf8skip[] = {
b2635aa8
KW
96/* 0x00 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
97/* 0x10 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
98/* 0x20 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
99/* 0x30 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
100/* 0x40 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
101/* 0x50 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
102/* 0x60 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
103/* 0x70 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
104/* 0x80 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
105/* 0x90 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
106/* 0xA0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
107/* 0xB0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
108/* 0xC0 */ 2,2, /* overlong */
109/* 0xC2 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0080 to U+03FF */
110/* 0xD0 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0400 to U+07FF */
111/* 0xE0 */ 3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3, /* U+0800 to U+FFFF */
112/* 0xF0 */ 4,4,4,4,4,4,4,4,5,5,5,5,6,6, /* above BMP to 2**31 - 1 */
113/* 0xFE */ 7,13, /* Perl extended (never was official UTF-8). Up to 72bit
114 allowed (64-bit + reserved). */
a0ed51b3
LW
115};
116#else
6f06b55f 117EXTCONST unsigned char PL_utf8skip[];
a0ed51b3
LW
118#endif
119
73c4f7a1 120END_EXTERN_C
7e2040f0 121
59a449d5
KW
122/* Native character to/from iso-8859-1. Are the identity functions on ASCII
123 * platforms */
124#define NATIVE_TO_LATIN1(ch) (ch)
125#define LATIN1_TO_NATIVE(ch) (ch)
126
127/* I8 is an intermediate version of UTF-8 used only in UTF-EBCDIC. We thus
128 * consider it to be identical to UTF-8 on ASCII platforms. Strictly speaking
129 * UTF-8 and UTF-EBCDIC are two different things, but we often conflate them
130 * because they are 8-bit encodings that serve the same purpose in Perl, and
131 * rarely do we need to distinguish them. The term "NATIVE_UTF8" applies to
132 * whichever one is applicable on the current platform */
133#define NATIVE_UTF8_TO_I8(ch) (ch)
134#define I8_TO_NATIVE_UTF8(ch) (ch)
135
1d72bdf6
NIS
136/* Transforms in wide UV chars */
137#define UNI_TO_NATIVE(ch) (ch)
138#define NATIVE_TO_UNI(ch) (ch)
d7578b48 139
877d9f0d 140/*
9041c2e3 141
8c007b5a 142 The following table is from Unicode 3.2.
877d9f0d
JH
143
144 Code Points 1st Byte 2nd Byte 3rd Byte 4th Byte
145
375122d7 146 U+0000..U+007F 00..7F
e1b711da 147 U+0080..U+07FF * C2..DF 80..BF
37e2e78e 148 U+0800..U+0FFF E0 * A0..BF 80..BF
375122d7 149 U+1000..U+CFFF E1..EC 80..BF 80..BF
e1b711da 150 U+D000..U+D7FF ED 80..9F 80..BF
537124e4 151 U+D800..U+DFFF ED A0..BF 80..BF (surrogates)
375122d7 152 U+E000..U+FFFF EE..EF 80..BF 80..BF
37e2e78e 153 U+10000..U+3FFFF F0 * 90..BF 80..BF 80..BF
877d9f0d
JH
154 U+40000..U+FFFFF F1..F3 80..BF 80..BF 80..BF
155 U+100000..U+10FFFF F4 80..8F 80..BF 80..BF
b2635aa8
KW
156 Below are non-Unicode code points
157 U+110000..U+13FFFF F4 90..BF 80..BF 80..BF
158 U+110000..U+1FFFFF F5..F7 80..BF 80..BF 80..BF
537124e4 159 U+200000..: F8.. * 88..BF 80..BF 80..BF 80..BF
877d9f0d 160
e1b711da 161Note the gaps before several of the byte entries above marked by '*'. These are
37e2e78e
KW
162caused by legal UTF-8 avoiding non-shortest encodings: it is technically
163possible to UTF-8-encode a single code point in different ways, but that is
164explicitly forbidden, and the shortest possible encoding should always be used
15824458 165(and that is what Perl does). The non-shortest ones are called 'overlongs'.
8c007b5a 166
877d9f0d
JH
167 */
168
8c007b5a
JH
169/*
170 Another way to look at it, as bits:
171
b2635aa8 172 Code Points 1st Byte 2nd Byte 3rd Byte 4th Byte
8c007b5a 173
b2635aa8
KW
174 0aaa aaaa 0aaa aaaa
175 0000 0bbb bbaa aaaa 110b bbbb 10aa aaaa
176 cccc bbbb bbaa aaaa 1110 cccc 10bb bbbb 10aa aaaa
177 00 000d ddcc cccc bbbb bbaa aaaa 1111 0ddd 10cc cccc 10bb bbbb 10aa aaaa
8c007b5a
JH
178
179As you can see, the continuation bytes all begin with C<10>, and the
e1b711da 180leading bits of the start byte tell how many bytes there are in the
8c007b5a
JH
181encoded character.
182
65ab9279
TC
183Perl's extended UTF-8 means we can have start bytes up to FF.
184
8c007b5a
JH
185*/
186
15824458
KW
187/* Is the representation of the Unicode code point 'c' the same regardless of
188 * being encoded in UTF-8 or not? */
c4d5f83a 189#define UNI_IS_INVARIANT(c) (((UV)c) < 0x80)
15824458
KW
190
191/* Is the UTF8-encoded byte 'c' part of a variant sequence in UTF-8? This is
192 * the inverse of UTF8_IS_INVARIANT */
193#define UTF8_IS_CONTINUED(c) (((U8)c) & 0x80)
194
195/* Is the byte 'c' the first byte of a multi-byte UTF8-8 encoded sequence?
196 * This doesn't catch invariants (they are single-byte). It also excludes the
197 * illegal overlong sequences that begin with C0 and C1. */
0447e8df 198#define UTF8_IS_START(c) (((U8)c) >= 0xc2)
15824458
KW
199
200/* Is the byte 'c' part of a multi-byte UTF8-8 encoded sequence, and not the
201 * first byte thereof? */
e021c6e6 202#define UTF8_IS_CONTINUATION(c) ((((U8)c) & 0xC0) == 0x80)
0ae1fa71 203
15824458
KW
204/* Is the UTF8-encoded byte 'c' the first byte of a two byte sequence? Use
205 * UTF8_IS_NEXT_CHAR_DOWNGRADEABLE() instead if the input isn't known to
206 * be well-formed. Masking with 0xfe allows the low bit to be 0 or 1; thus
207 * this matches 0xc[23]. */
559c7f10 208#define UTF8_IS_DOWNGRADEABLE_START(c) (((U8)(c) & 0xfe) == 0xc2)
4ab10950 209
15824458
KW
210/* Is the UTF8-encoded byte 'c' the first byte of a sequence of bytes that
211 * represent a code point > 255? */
7b4252f4 212#define UTF8_IS_ABOVE_LATIN1(c) ((U8)(c) >= 0xc4)
8850bf83 213
15824458
KW
214/* This defines the 1-bits that are to be in the first byte of a multi-byte
215 * UTF-8 encoded character that give the number of bytes that comprise the
f710bf0c 216 * character. 'len' is the number of bytes in the multi-byte sequence. */
ee372ee9 217#define UTF_START_MARK(len) (((len) > 7) ? 0xFF : (0xFF & (0xFE << (7-(len)))))
b2635aa8
KW
218
219/* Masks out the initial one bits in a start byte, leaving the real data ones.
f710bf0c
KW
220 * Doesn't work on an invariant byte. 'len' is the number of bytes in the
221 * multi-byte sequence that comprises the character. */
22901f30 222#define UTF_START_MASK(len) (((len) >= 7) ? 0x00 : (0x1F >> ((len)-2)))
1d72bdf6 223
15824458
KW
224/* This defines the bits that are to be in the continuation bytes of a multi-byte
225 * UTF-8 encoded character that indicate it is a continuation byte. */
1d72bdf6 226#define UTF_CONTINUATION_MARK 0x80
15824458
KW
227
228/* This is the number of low-order bits a continuation byte in a UTF-8 encoded
229 * sequence contributes to the specification of the code point. In the bit
230 * maps above, you see that the first 2 bits are a constant '10', leaving 6 of
231 * real information */
1d72bdf6 232#define UTF_ACCUMULATION_SHIFT 6
b2635aa8
KW
233
234/* 2**UTF_ACCUMULATION_SHIFT - 1 */
1d72bdf6 235#define UTF_CONTINUATION_MASK ((U8)0x3f)
c512ce4f 236
15824458
KW
237/* If a value is anded with this, and the result is non-zero, then using the
238 * original value in UTF8_ACCUMULATE will overflow, shifting bits off the left
239 * */
eb83ed87 240#define UTF_ACCUMULATION_OVERFLOW_MASK \
bb88be5f
KW
241 (((UV) UTF_CONTINUATION_MASK) << ((sizeof(UV) * CHARBITS) \
242 - UTF_ACCUMULATION_SHIFT))
eb83ed87 243
6588300d
NC
244#if UVSIZE >= 8
245# define UTF8_QUAD_MAX UINT64_C(0x1000000000)
246
5aaebcb3
KW
247/* Input is a true Unicode (not-native) code point */
248#define OFFUNISKIP(uv) ( (uv) < 0x80 ? 1 : \
1d68d6cd
SC
249 (uv) < 0x800 ? 2 : \
250 (uv) < 0x10000 ? 3 : \
251 (uv) < 0x200000 ? 4 : \
252 (uv) < 0x4000000 ? 5 : \
253 (uv) < 0x80000000 ? 6 : \
9041c2e3 254 (uv) < UTF8_QUAD_MAX ? 7 : 13 )
1d68d6cd
SC
255#else
256/* No, I'm not even going to *TRY* putting #ifdef inside a #define */
5aaebcb3 257#define OFFUNISKIP(uv) ( (uv) < 0x80 ? 1 : \
1d68d6cd
SC
258 (uv) < 0x800 ? 2 : \
259 (uv) < 0x10000 ? 3 : \
260 (uv) < 0x200000 ? 4 : \
261 (uv) < 0x4000000 ? 5 : \
262 (uv) < 0x80000000 ? 6 : 7 )
263#endif
264
03c76984
KW
265/* How wide can a single UTF-8 encoded character become in bytes. */
266/* NOTE: Strictly speaking Perl's UTF-8 should not be called UTF-8 since UTF-8
267 * is an encoding of Unicode, and Unicode's upper limit, 0x10FFFF, can be
268 * expressed with 4 bytes. However, Perl thinks of UTF-8 as a way to encode
269 * non-negative integers in a binary format, even those above Unicode */
270#define UTF8_MAXBYTES 13
271
272/* The maximum number of UTF-8 bytes a single Unicode character can
273 * uppercase/lowercase/fold into. Unicode guarantees that the maximum
274 * expansion is 3 characters. On ASCIIish platforms, the highest Unicode
275 * character occupies 4 bytes, therefore this number would be 12, but this is
276 * smaller than the maximum width a single above-Unicode character can occupy,
277 * so use that instead */
278#if UTF8_MAXBYTES < 12
279#error UTF8_MAXBYTES must be at least 12
280#endif
281
0ed2b00b
KW
282/* ^? is defined to be DEL on ASCII systems. See the definition of toCTRL()
283 * for more */
284#define QUESTION_MARK_CTRL DEL_NATIVE
285
843a4590
KW
286#define MAX_UTF8_TWO_BYTE 0x7FF
287
03c76984
KW
288#define UTF8_MAXBYTES_CASE UTF8_MAXBYTES
289
d06134e5
KW
290#endif /* EBCDIC vs ASCII */
291
292/* Rest of these are attributes of Unicode and perl's internals rather than the
293 * encoding, or happen to be the same in both ASCII and EBCDIC (at least at
294 * this level; the macros that some of these call may have different
295 * definitions in the two encodings */
296
59a449d5
KW
297/* In domain restricted to ASCII, these may make more sense to the reader than
298 * the ones with Latin1 in the name */
299#define NATIVE_TO_ASCII(ch) NATIVE_TO_LATIN1(ch)
300#define ASCII_TO_NATIVE(ch) LATIN1_TO_NATIVE(ch)
301
302/* More or less misleadingly-named defines, retained for back compat */
303#define NATIVE_TO_UTF(ch) NATIVE_UTF8_TO_I8(ch)
304#define NATIVE_TO_I8(ch) NATIVE_UTF8_TO_I8(ch)
305#define UTF_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch)
306#define I8_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch)
307#define NATIVE8_TO_UNI(ch) NATIVE_TO_LATIN1(ch)
d06134e5 308
537124e4
KW
309/* Adds a UTF8 continuation byte 'new' of information to a running total code
310 * point 'old' of all the continuation bytes so far. This is designed to be
155d2738
KW
311 * used in a loop to convert from UTF-8 to the code point represented. Note
312 * that this is asymmetric on EBCDIC platforms, in that the 'new' parameter is
313 * the UTF-EBCDIC byte, whereas the 'old' parameter is a Unicode (not EBCDIC)
314 * code point in process of being generated */
315#define UTF8_ACCUMULATE(old, new) (((old) << UTF_ACCUMULATION_SHIFT) \
316 | ((NATIVE_UTF8_TO_I8((U8)new)) \
317 & UTF_CONTINUATION_MASK))
d06134e5 318
4ab10950
KW
319/* This works in the face of malformed UTF-8. */
320#define UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(s, e) (UTF8_IS_DOWNGRADEABLE_START(*s) \
321 && ( (e) - (s) > 1) \
322 && UTF8_IS_CONTINUATION(*((s)+1)))
323
5aaebcb3
KW
324/* Number of bytes a code point occupies in UTF-8. */
325#define NATIVE_SKIP(uv) OFFUNISKIP(NATIVE_TO_UNI(uv))
326
327/* Most code which says UNISKIP is really thinking in terms of native code
328 * points (0-255) plus all those beyond. This is an imprecise term, but having
2accb712
KW
329 * it means existing code continues to work. For precision, use UVCHR_SKIP,
330 * NATIVE_SKIP, and OFFUNISKIP */
5aaebcb3 331#define UNISKIP(uv) NATIVE_SKIP(uv)
2accb712 332#define UVCHR_SKIP(uv) NATIVE_SKIP(uv)
5aaebcb3 333
94bb8c36 334/* Convert a two (not one) byte utf8 character to a native code point value.
2950f2a7
KW
335 * Needs just one iteration of accumulate. Should not be used unless it is
336 * known that the two bytes are legal: 1) two-byte start, and 2) continuation.
337 * Note that the result can be larger than 255 if the input character is not
338 * downgradable */
94bb8c36
KW
339#define TWO_BYTE_UTF8_TO_NATIVE(HI, LO) \
340 UNI_TO_NATIVE(UTF8_ACCUMULATE((NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), \
155d2738 341 (LO)))
94bb8c36
KW
342
343/* Should never be used, and be deprecated */
344#define TWO_BYTE_UTF8_TO_UNI(HI, LO) NATIVE_TO_UNI(TWO_BYTE_UTF8_TO_NATIVE(HI, LO))
2950f2a7 345
15824458
KW
346/* How many bytes in the UTF-8 encoded character whose first (perhaps only)
347 * byte is pointed to by 's' */
d06134e5
KW
348#define UTF8SKIP(s) PL_utf8skip[*(const U8*)(s)]
349
15824458
KW
350/* Is the byte 'c' the same character when encoded in UTF-8 as when not. This
351 * works on both UTF-8 encoded strings and non-encoded, as it returns TRUE in
352 * each for the exact same set of bit patterns. (And it works on any byte in a
353 * UTF-8 encoded string) */
bc3632a8 354#define UTF8_IS_INVARIANT(c) UNI_IS_INVARIANT(NATIVE_UTF8_TO_I8(c))
15824458 355
6f2d5cbc
KW
356/* Like the above, but its name implies a non-UTF8 input, and is implemented
357 * differently (for no particular reason) */
358#define NATIVE_BYTE_IS_INVARIANT(c) UNI_IS_INVARIANT(NATIVE_TO_LATIN1(c))
359
360/* Like the above, but accepts any UV as input */
361#define UVCHR_IS_INVARIANT(uv) UNI_IS_INVARIANT(NATIVE_TO_UNI(uv))
d06134e5
KW
362
363#define MAX_PORTABLE_UTF8_TWO_BYTE 0x3FF /* constrained by EBCDIC */
364
48ccf5e1
KW
365/* The macros in the next 4 sets are used to generate the two utf8 or utfebcdic
366 * bytes from an ordinal that is known to fit into exactly two (not one) bytes;
367 * it must be less than 0x3FF to work across both encodings. */
368
369/* These two are helper macros for the other three sets, and should not be used
370 * directly anywhere else. 'translate_function' is either NATIVE_TO_LATIN1
371 * (which works for code points up to 0xFF) or NATIVE_TO_UNI which works for any
372 * code point */
373#define __BASE_TWO_BYTE_HI(c, translate_function) \
374 I8_TO_NATIVE_UTF8((translate_function(c) >> UTF_ACCUMULATION_SHIFT) \
375 | UTF_START_MARK(2))
376#define __BASE_TWO_BYTE_LO(c, translate_function) \
377 I8_TO_NATIVE_UTF8((translate_function(c) & UTF_CONTINUATION_MASK) \
378 | UTF_CONTINUATION_MARK)
379
48ccf5e1
KW
380/* The next two macros should not be used. They were designed to be usable as
381 * the case label of a switch statement, but this doesn't work for EBCDIC. Use
9d0d3a03 382 * regen/unicode_constants.pl instead */
48ccf5e1
KW
383#define UTF8_TWO_BYTE_HI_nocast(c) __BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI)
384#define UTF8_TWO_BYTE_LO_nocast(c) __BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI)
385
386/* The next two macros are used when the source should be a single byte
387 * character; checked for under DEBUGGING */
388#define UTF8_EIGHT_BIT_HI(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
389 ((U8) __BASE_TWO_BYTE_HI(c, NATIVE_TO_LATIN1)))
390#define UTF8_EIGHT_BIT_LO(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
391 ((U8) __BASE_TWO_BYTE_LO(c, NATIVE_TO_LATIN1)))
392
393/* These final two macros in the series are used when the source can be any
394 * code point whose UTF-8 is known to occupy 2 bytes; they are less efficient
395 * than the EIGHT_BIT versions on EBCDIC platforms. We use the logical '~'
396 * operator instead of "<=" to avoid getting compiler warnings.
d52b8576 397 * MAX_UTF8_TWO_BYTE should be exactly all one bits in the lower few
48ccf5e1
KW
398 * places, so the ~ works */
399#define UTF8_TWO_BYTE_HI(c) \
400 (__ASSERT_((sizeof(c) == 1) \
d52b8576 401 || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \
d9759938 402 ((U8) __BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI)))
48ccf5e1
KW
403#define UTF8_TWO_BYTE_LO(c) \
404 (__ASSERT_((sizeof(c) == 1) \
d52b8576 405 || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \
d9759938 406 ((U8) __BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI)))
d06134e5 407
e7214ce8
KW
408/* This is illegal in any well-formed UTF-8 in both EBCDIC and ASCII
409 * as it is only in overlongs. */
410#define ILLEGAL_UTF8_BYTE I8_TO_NATIVE_UTF8(0xC1)
411
7e2040f0 412/*
e3036cf4 413 * 'UTF' is whether or not p is encoded in UTF8. The names 'foo_lazy_if' stem
20df05f4
KW
414 * from an earlier version of these macros in which they didn't call the
415 * foo_utf8() macros (i.e. were 'lazy') unless they decided that *p is the
416 * beginning of a utf8 character. Now that foo_utf8() determines that itself,
417 * no need to do it again here
7e2040f0 418 */
e3036cf4
KW
419#define isIDFIRST_lazy_if(p,UTF) ((IN_BYTES || !UTF ) \
420 ? isIDFIRST(*(p)) \
421 : isIDFIRST_utf8((const U8*)p))
32636478
KW
422#define isWORDCHAR_lazy_if(p,UTF) ((IN_BYTES || (!UTF )) \
423 ? isWORDCHAR(*(p)) \
424 : isWORDCHAR_utf8((const U8*)p))
425#define isALNUM_lazy_if(p,UTF) isWORDCHAR_lazy_if(p,UTF)
1d72bdf6 426
89ebb4a3
JH
427#define UTF8_MAXLEN UTF8_MAXBYTES
428
8cb75cc8
KW
429/* A Unicode character can fold to up to 3 characters */
430#define UTF8_MAX_FOLD_CHAR_EXPAND 3
431
a98fe34d 432#define IN_BYTES (CopHINTS_get(PL_curcop) & HINT_BYTES)
0064a8a9 433#define DO_UTF8(sv) (SvUTF8(sv) && !IN_BYTES)
66cbab2c 434#define IN_UNI_8_BIT \
d6ded950
KW
435 (((CopHINTS_get(PL_curcop) & (HINT_UNI_8_BIT)) \
436 || (CopHINTS_get(PL_curcop) & HINT_LOCALE_PARTIAL \
437 /* -1 below is for :not_characters */ \
438 && _is_in_locale_category(FALSE, -1))) \
439 && ! IN_BYTES)
b36bf33f 440
1d72bdf6 441
c76687c5
KW
442#define UTF8_ALLOW_EMPTY 0x0001 /* Allow a zero length string */
443
444/* Allow first byte to be a continuation byte */
1d72bdf6 445#define UTF8_ALLOW_CONTINUATION 0x0002
c76687c5
KW
446
447/* Allow second... bytes to be non-continuation bytes */
1d72bdf6 448#define UTF8_ALLOW_NON_CONTINUATION 0x0004
949cf498
KW
449
450/* expecting more bytes than were available in the string */
451#define UTF8_ALLOW_SHORT 0x0008
452
453/* Overlong sequence; i.e., the code point can be specified in fewer bytes. */
454#define UTF8_ALLOW_LONG 0x0010
455
456#define UTF8_DISALLOW_SURROGATE 0x0020 /* Unicode surrogates */
457#define UTF8_WARN_SURROGATE 0x0040
458
459#define UTF8_DISALLOW_NONCHAR 0x0080 /* Unicode non-character */
460#define UTF8_WARN_NONCHAR 0x0100 /* code points */
461
462#define UTF8_DISALLOW_SUPER 0x0200 /* Super-set of Unicode: code */
463#define UTF8_WARN_SUPER 0x0400 /* points above the legal max */
464
465/* Code points which never were part of the original UTF-8 standard, the first
ea5ced44
KW
466 * byte of which is a FE or FF on ASCII platforms. If the first byte is FF, it
467 * will overflow a 32-bit word. If the first byte is FE, it will overflow a
468 * signed 32-bit word. */
949cf498
KW
469#define UTF8_DISALLOW_FE_FF 0x0800
470#define UTF8_WARN_FE_FF 0x1000
471
472#define UTF8_CHECK_ONLY 0x2000
473
474/* For backwards source compatibility. They do nothing, as the default now
475 * includes what they used to mean. The first one's meaning was to allow the
476 * just the single non-character 0xFFFF */
477#define UTF8_ALLOW_FFFF 0
478#define UTF8_ALLOW_SURROGATE 0
479
33d9abfb 480#define UTF8_DISALLOW_ILLEGAL_INTERCHANGE (UTF8_DISALLOW_SUPER|UTF8_DISALLOW_NONCHAR|UTF8_DISALLOW_SURROGATE|UTF8_DISALLOW_FE_FF)
949cf498 481#define UTF8_WARN_ILLEGAL_INTERCHANGE \
33d9abfb 482 (UTF8_WARN_SUPER|UTF8_WARN_NONCHAR|UTF8_WARN_SURROGATE|UTF8_WARN_FE_FF)
949cf498
KW
483#define UTF8_ALLOW_ANY \
484 (~(UTF8_DISALLOW_ILLEGAL_INTERCHANGE|UTF8_WARN_ILLEGAL_INTERCHANGE))
485#define UTF8_ALLOW_ANYUV \
486 (UTF8_ALLOW_EMPTY \
487 & ~(UTF8_DISALLOW_ILLEGAL_INTERCHANGE|UTF8_WARN_ILLEGAL_INTERCHANGE))
9f7f3913
ST
488#define UTF8_ALLOW_DEFAULT (ckWARN(WARN_UTF8) ? 0 : \
489 UTF8_ALLOW_ANYUV)
1d72bdf6 490
7131f24d
KW
491/* Surrogates, non-character code points and above-Unicode code points are
492 * problematic in some contexts. This allows code that needs to check for
493 * those to to quickly exclude the vast majority of code points it will
494 * encounter */
b96a92fb
KW
495#define UTF8_FIRST_PROBLEMATIC_CODE_POINT_FIRST_BYTE \
496 FIRST_SURROGATE_UTF8_FIRST_BYTE
7131f24d 497
b96a92fb
KW
498#define UTF8_IS_SURROGATE(s) cBOOL(is_SURROGATE_utf8(s))
499#define UTF8_IS_REPLACEMENT(s, send) cBOOL(is_REPLACEMENT_utf8_safe(s,send))
7131f24d
KW
500
501/* ASCII EBCDIC I8
502 * U+10FFFF: \xF4\x8F\xBF\xBF \xF9\xA1\xBF\xBF\xBF max legal Unicode
503 * U+110000: \xF4\x90\x80\x80 \xF9\xA2\xA0\xA0\xA0
504 * U+110001: \xF4\x90\x80\x81 \xF9\xA2\xA0\xA0\xA1
a1776718
KW
505 *
506 * BE AWARE that this test doesn't rule out malformed code points, in
507 * particular overlongs */
7131f24d 508#ifdef EBCDIC /* Both versions assume well-formed UTF8 */
bc3632a8
KW
509# define UTF8_IS_SUPER(s) (NATIVE_UTF8_TO_I8(* (U8*) (s)) >= 0xF9 \
510 && (NATIVE_UTF8_TO_I8(* (U8*) (s)) > 0xF9 \
a3481822 511 || (NATIVE_UTF8_TO_I8(* ((U8*) (s) + 1)) >= 0xA2)))
7131f24d 512#else
0404c92a
KW
513# define UTF8_IS_SUPER(s) (*(U8*) (s) >= 0xF4 \
514 && (*(U8*) (s) > 0xF4 || (*((U8*) (s) + 1) >= 0x90)))
7131f24d
KW
515#endif
516
b96a92fb
KW
517/* These are now machine generated, and the 'given' clause is no longer
518 * applicable */
519#define UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s) \
520 cBOOL(is_NONCHAR_utf8(s))
521#define UTF8_IS_NONCHAR_(s) \
522 UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s)
7131f24d 523
c867b360
JH
524#define UNICODE_SURROGATE_FIRST 0xD800
525#define UNICODE_SURROGATE_LAST 0xDFFF
526#define UNICODE_REPLACEMENT 0xFFFD
527#define UNICODE_BYTE_ORDER_MARK 0xFEFF
1d72bdf6 528
b851fbc1 529/* Though our UTF-8 encoding can go beyond this,
c76687c5 530 * let's be conservative and do as Unicode says. */
b851fbc1
JH
531#define PERL_UNICODE_MAX 0x10FFFF
532
949cf498
KW
533#define UNICODE_WARN_SURROGATE 0x0001 /* UTF-16 surrogates */
534#define UNICODE_WARN_NONCHAR 0x0002 /* Non-char code points */
535#define UNICODE_WARN_SUPER 0x0004 /* Above 0x10FFFF */
536#define UNICODE_WARN_FE_FF 0x0008 /* Above 0x10FFFF */
537#define UNICODE_DISALLOW_SURROGATE 0x0010
538#define UNICODE_DISALLOW_NONCHAR 0x0020
539#define UNICODE_DISALLOW_SUPER 0x0040
540#define UNICODE_DISALLOW_FE_FF 0x0080
bb88be5f
KW
541#define UNICODE_WARN_ILLEGAL_INTERCHANGE \
542 (UNICODE_WARN_SURROGATE|UNICODE_WARN_NONCHAR|UNICODE_WARN_SUPER)
543#define UNICODE_DISALLOW_ILLEGAL_INTERCHANGE \
544 (UNICODE_DISALLOW_SURROGATE|UNICODE_DISALLOW_NONCHAR|UNICODE_DISALLOW_SUPER)
949cf498
KW
545
546/* For backward source compatibility, as are now the default */
547#define UNICODE_ALLOW_SURROGATE 0
548#define UNICODE_ALLOW_SUPER 0
549#define UNICODE_ALLOW_ANY 0
b851fbc1 550
1d72bdf6
NIS
551#define UNICODE_IS_SURROGATE(c) ((c) >= UNICODE_SURROGATE_FIRST && \
552 (c) <= UNICODE_SURROGATE_LAST)
a10ec373 553#define UNICODE_IS_REPLACEMENT(c) ((c) == UNICODE_REPLACEMENT)
872c91ae 554#define UNICODE_IS_BYTE_ORDER_MARK(c) ((c) == UNICODE_BYTE_ORDER_MARK)
7131f24d
KW
555#define UNICODE_IS_NONCHAR(c) ((c >= 0xFDD0 && c <= 0xFDEF) \
556 /* The other noncharacters end in FFFE or FFFF, which \
557 * the mask below catches both of, but beyond the last \
558 * official unicode code point, they aren't \
559 * noncharacters, since those aren't Unicode \
560 * characters at all */ \
561 || ((((c & 0xFFFE) == 0xFFFE)) && ! UNICODE_IS_SUPER(c)))
562#define UNICODE_IS_SUPER(c) ((c) > PERL_UNICODE_MAX)
563#define UNICODE_IS_FE_FF(c) ((c) > 0x7FFFFFFF)
1d72bdf6 564
ec34087a
KW
565#define LATIN_SMALL_LETTER_SHARP_S LATIN_SMALL_LETTER_SHARP_S_NATIVE
566#define LATIN_SMALL_LETTER_Y_WITH_DIAERESIS \
567 LATIN_SMALL_LETTER_Y_WITH_DIAERESIS_NATIVE
568#define MICRO_SIGN MICRO_SIGN_NATIVE
569#define LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE \
570 LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE_NATIVE
571#define LATIN_SMALL_LETTER_A_WITH_RING_ABOVE \
572 LATIN_SMALL_LETTER_A_WITH_RING_ABOVE_NATIVE
09091399
JH
573#define UNICODE_GREEK_CAPITAL_LETTER_SIGMA 0x03A3
574#define UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA 0x03C2
575#define UNICODE_GREEK_SMALL_LETTER_SIGMA 0x03C3
9dcbe121 576#define GREEK_SMALL_LETTER_MU 0x03BC
9e682c18
KW
577#define GREEK_CAPITAL_LETTER_MU 0x039C /* Upper and title case
578 of MICRON */
579#define LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS 0x0178 /* Also is title case */
580#define LATIN_CAPITAL_LETTER_SHARP_S 0x1E9E
581#define LATIN_SMALL_LETTER_LONG_S 0x017F
a9f50d33
KW
582#define LATIN_SMALL_LIGATURE_LONG_S_T 0xFB05
583#define LATIN_SMALL_LIGATURE_ST 0xFB06
9e682c18
KW
584#define KELVIN_SIGN 0x212A
585#define ANGSTROM_SIGN 0x212B
09091399 586
9e55ce06 587#define UNI_DISPLAY_ISPRINT 0x0001
c728cb41
JH
588#define UNI_DISPLAY_BACKSLASH 0x0002
589#define UNI_DISPLAY_QQ (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH)
590#define UNI_DISPLAY_REGEX (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH)
9e55ce06 591
5cd46e1f
KW
592#define ANYOF_FOLD_SHARP_S(node, input, end) \
593 (ANYOF_BITMAP_TEST(node, LATIN_SMALL_LETTER_SHARP_S) && \
137165a6 594 (ANYOF_NONBITMAP(node)) && \
39065660 595 (ANYOF_FLAGS(node) & ANYOF_LOC_NONBITMAP_FOLD) && \
07b6858f 596 ((end) > (input) + 1) && \
305b8651 597 isALPHA_FOLD_EQ((input)[0], 's'))
6302f837 598
ebc501f0 599#define SHARP_S_SKIP 2
3b0fc154 600
a4f7a67c
KW
601/* If you want to exclude surrogates, and beyond legal Unicode, see the blame
602 * log for earlier versions which gave details for these */
4d646140 603
6302f837
KW
604/* A helper macro for isUTF8_CHAR, so use that one, and not this one. This is
605 * retained solely for backwards compatibility and may be deprecated and
606 * removed in a future Perl version.
607 *
608 * regen/regcharclass.pl generates is_UTF8_CHAR_utf8() macros for up to these
d9f92374
KW
609 * number of bytes. So this has to be coordinated with that file */
610#ifdef EBCDIC
611# define IS_UTF8_CHAR_FAST(n) ((n) <= 3)
612#else
613# define IS_UTF8_CHAR_FAST(n) ((n) <= 4)
614#endif
615
4d646140 616#ifndef EBCDIC
6302f837
KW
617/* A helper macro for isUTF8_CHAR, so use that one instead of this. This was
618 * generated by regen/regcharclass.pl, and then moved here. The lines that
619 * generated it were then commented out. This was done solely because it takes
620 * on the order of 10 minutes to generate, and is never going to change, unless
621 * the generated code is improved.
622 *
623 * The EBCDIC versions have been cut to not cover all of legal Unicode, so
624 * don't take too long to generate, and there is a separate one for each code
625 * page, so they are in regcharclass.h instead of here */
39a0f513 626/*
5dca9278 627 UTF8_CHAR: Matches legal UTF-8 encoded characters from 2 through 4 bytes
39a0f513 628
5dca9278 629 0x80 - 0x1FFFFF
39a0f513 630*/
4d646140 631/*** GENERATED CODE ***/
5dca9278
KW
632#define is_UTF8_CHAR_utf8_no_length_checks(s) \
633( ( 0xC2 <= ((U8*)s)[0] && ((U8*)s)[0] <= 0xDF ) ? \
634 ( ( ( ((U8*)s)[1] & 0xC0 ) == 0x80 ) ? 2 : 0 ) \
635: ( 0xE0 == ((U8*)s)[0] ) ? \
636 ( ( ( ( ((U8*)s)[1] & 0xE0 ) == 0xA0 ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
637: ( 0xE1 <= ((U8*)s)[0] && ((U8*)s)[0] <= 0xEF ) ? \
638 ( ( ( ( ((U8*)s)[1] & 0xC0 ) == 0x80 ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
639: ( 0xF0 == ((U8*)s)[0] ) ? \
640 ( ( ( ( 0x90 <= ((U8*)s)[1] && ((U8*)s)[1] <= 0xBF ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) && ( ( ((U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\
641: ( ( ( ( 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 642#endif
3b0fc154 643
6302f837 644/*
5dca9278
KW
645
646=for apidoc Am|STRLEN|isUTF8_CHAR|const U8 *s|const U8 *e
647
648Returns the number of bytes beginning at C<s> which form a legal UTF-8 (or
649UTF-EBCDIC) encoded character, looking no further than C<e - s> bytes into
650C<s>. Returns 0 if the sequence starting at C<s> through C<e - 1> is not
651well-formed UTF-8
6302f837
KW
652
653Note that an INVARIANT character (i.e. ASCII on non-EBCDIC
5dca9278
KW
654machines) is a valid UTF-8 character.
655
656=cut
657*/
6302f837 658
dd9bc2b0 659#define isUTF8_CHAR(s, e) (UNLIKELY((e) <= (s)) \
6302f837
KW
660 ? 0 \
661 : (UTF8_IS_INVARIANT(*s)) \
662 ? 1 \
dd9bc2b0 663 : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \
6302f837 664 ? 0 \
dd9bc2b0 665 : LIKELY(IS_UTF8_CHAR_FAST(UTF8SKIP(s))) \
5dca9278 666 ? is_UTF8_CHAR_utf8_no_length_checks(s) \
6302f837
KW
667 : _is_utf8_char_slow(s, e))
668
3cedd9d9
KW
669#define is_utf8_char_buf(buf, buf_end) isUTF8_CHAR(buf, buf_end)
670
6302f837
KW
671/* Do not use; should be deprecated. Use isUTF8_CHAR() instead; this is
672 * retained solely for backwards compatibility */
673#define IS_UTF8_CHAR(p, n) (isUTF8_CHAR(p, (p) + (n)) == n)
e9a8c099 674
57f0e7e2
KW
675#endif /* H_UTF8 */
676
e9a8c099 677/*
14d04a33 678 * ex: set ts=8 sts=4 sw=4 et:
e9a8c099 679 */