This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
regen/regcharclass.pl: Add new macro type with intermed checking
[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
de69f3af
KW
42#define uvchr_to_utf8(a,b) uvchr_to_utf8_flags(a,b,0)
43#define uvchr_to_utf8_flags(d,uv,flags) \
44 uvoffuni_to_utf8_flags(d,NATIVE_TO_UNI(uv),flags)
45#define utf8_to_uvchr_buf(s, e, lenp) \
842991ae 46 utf8n_to_uvchr(s, (U8*)(e) - (U8*)(s), lenp, \
de69f3af
KW
47 ckWARN_d(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY)
48
a0270393 49#define to_uni_fold(c, p, lenp) _to_uni_fold_flags(c, p, lenp, FOLD_FLAGS_FULL)
445bf929
KW
50#define to_utf8_fold(c, p, lenp) _to_utf8_fold_flags(c, p, lenp, FOLD_FLAGS_FULL)
51#define to_utf8_lower(a,b,c) _to_utf8_lower_flags(a,b,c,0)
52#define to_utf8_upper(a,b,c) _to_utf8_upper_flags(a,b,c,0)
53#define to_utf8_title(a,b,c) _to_utf8_title_flags(a,b,c,0)
36bb2ab6 54
fd7cb289 55/* Source backward compatibility. */
fd7cb289
RGS
56#define is_utf8_string_loc(s, len, ep) is_utf8_string_loclen(s, len, ep, 0)
57
eda9cac1
KW
58#define foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \
59 foldEQ_utf8_flags(s1, pe1, l1, u1, s2, pe2, l2, u2, 0)
baa60164 60#define FOLDEQ_UTF8_NOMIX_ASCII (1 << 0)
cea315b6 61#define FOLDEQ_LOCALE (1 << 1)
18f762c3
KW
62#define FOLDEQ_S1_ALREADY_FOLDED (1 << 2)
63#define FOLDEQ_S2_ALREADY_FOLDED (1 << 3)
a33c29bc 64
e6226b18
KW
65/*
66=for apidoc ibcmp_utf8
67
68This is a synonym for (! foldEQ_utf8())
69
70=cut
71*/
72#define ibcmp_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \
73 cBOOL(! foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2))
74
1d72bdf6
NIS
75#ifdef EBCDIC
76/* The equivalent of these macros but implementing UTF-EBCDIC
77 are in the following header file:
78 */
79
80#include "utfebcdic.h"
fd7cb289 81
d06134e5 82#else /* ! EBCDIC */
73c4f7a1
GS
83START_EXTERN_C
84
a0ed51b3 85#ifdef DOINIT
6f06b55f 86EXTCONST unsigned char PL_utf8skip[] = {
b2635aa8
KW
87/* 0x00 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
88/* 0x10 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
89/* 0x20 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
90/* 0x30 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
91/* 0x40 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
92/* 0x50 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
93/* 0x60 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
94/* 0x70 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
95/* 0x80 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
96/* 0x90 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
97/* 0xA0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
98/* 0xB0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
99/* 0xC0 */ 2,2, /* overlong */
100/* 0xC2 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0080 to U+03FF */
101/* 0xD0 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0400 to U+07FF */
102/* 0xE0 */ 3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3, /* U+0800 to U+FFFF */
103/* 0xF0 */ 4,4,4,4,4,4,4,4,5,5,5,5,6,6, /* above BMP to 2**31 - 1 */
104/* 0xFE */ 7,13, /* Perl extended (never was official UTF-8). Up to 72bit
105 allowed (64-bit + reserved). */
a0ed51b3
LW
106};
107#else
6f06b55f 108EXTCONST unsigned char PL_utf8skip[];
a0ed51b3
LW
109#endif
110
73c4f7a1 111END_EXTERN_C
7e2040f0 112
59a449d5
KW
113/* Native character to/from iso-8859-1. Are the identity functions on ASCII
114 * platforms */
115#define NATIVE_TO_LATIN1(ch) (ch)
116#define LATIN1_TO_NATIVE(ch) (ch)
117
118/* I8 is an intermediate version of UTF-8 used only in UTF-EBCDIC. We thus
119 * consider it to be identical to UTF-8 on ASCII platforms. Strictly speaking
120 * UTF-8 and UTF-EBCDIC are two different things, but we often conflate them
121 * because they are 8-bit encodings that serve the same purpose in Perl, and
122 * rarely do we need to distinguish them. The term "NATIVE_UTF8" applies to
123 * whichever one is applicable on the current platform */
124#define NATIVE_UTF8_TO_I8(ch) (ch)
125#define I8_TO_NATIVE_UTF8(ch) (ch)
126
1d72bdf6
NIS
127/* Transforms in wide UV chars */
128#define UNI_TO_NATIVE(ch) (ch)
129#define NATIVE_TO_UNI(ch) (ch)
d7578b48 130
877d9f0d 131/*
9041c2e3 132
8c007b5a 133 The following table is from Unicode 3.2.
877d9f0d
JH
134
135 Code Points 1st Byte 2nd Byte 3rd Byte 4th Byte
136
375122d7 137 U+0000..U+007F 00..7F
e1b711da 138 U+0080..U+07FF * C2..DF 80..BF
37e2e78e 139 U+0800..U+0FFF E0 * A0..BF 80..BF
375122d7 140 U+1000..U+CFFF E1..EC 80..BF 80..BF
e1b711da 141 U+D000..U+D7FF ED 80..9F 80..BF
537124e4 142 U+D800..U+DFFF ED A0..BF 80..BF (surrogates)
375122d7 143 U+E000..U+FFFF EE..EF 80..BF 80..BF
37e2e78e 144 U+10000..U+3FFFF F0 * 90..BF 80..BF 80..BF
877d9f0d
JH
145 U+40000..U+FFFFF F1..F3 80..BF 80..BF 80..BF
146 U+100000..U+10FFFF F4 80..8F 80..BF 80..BF
b2635aa8
KW
147 Below are non-Unicode code points
148 U+110000..U+13FFFF F4 90..BF 80..BF 80..BF
149 U+110000..U+1FFFFF F5..F7 80..BF 80..BF 80..BF
537124e4 150 U+200000..: F8.. * 88..BF 80..BF 80..BF 80..BF
877d9f0d 151
e1b711da 152Note the gaps before several of the byte entries above marked by '*'. These are
37e2e78e
KW
153caused by legal UTF-8 avoiding non-shortest encodings: it is technically
154possible to UTF-8-encode a single code point in different ways, but that is
155explicitly forbidden, and the shortest possible encoding should always be used
15824458 156(and that is what Perl does). The non-shortest ones are called 'overlongs'.
8c007b5a 157
877d9f0d
JH
158 */
159
8c007b5a
JH
160/*
161 Another way to look at it, as bits:
162
b2635aa8 163 Code Points 1st Byte 2nd Byte 3rd Byte 4th Byte
8c007b5a 164
b2635aa8
KW
165 0aaa aaaa 0aaa aaaa
166 0000 0bbb bbaa aaaa 110b bbbb 10aa aaaa
167 cccc bbbb bbaa aaaa 1110 cccc 10bb bbbb 10aa aaaa
168 00 000d ddcc cccc bbbb bbaa aaaa 1111 0ddd 10cc cccc 10bb bbbb 10aa aaaa
8c007b5a
JH
169
170As you can see, the continuation bytes all begin with C<10>, and the
e1b711da 171leading bits of the start byte tell how many bytes there are in the
8c007b5a
JH
172encoded character.
173
65ab9279
TC
174Perl's extended UTF-8 means we can have start bytes up to FF.
175
8c007b5a
JH
176*/
177
15824458
KW
178/* Is the representation of the Unicode code point 'c' the same regardless of
179 * being encoded in UTF-8 or not? */
c4d5f83a 180#define UNI_IS_INVARIANT(c) (((UV)c) < 0x80)
15824458
KW
181
182/* Is the UTF8-encoded byte 'c' part of a variant sequence in UTF-8? This is
183 * the inverse of UTF8_IS_INVARIANT */
184#define UTF8_IS_CONTINUED(c) (((U8)c) & 0x80)
185
186/* Is the byte 'c' the first byte of a multi-byte UTF8-8 encoded sequence?
187 * This doesn't catch invariants (they are single-byte). It also excludes the
188 * illegal overlong sequences that begin with C0 and C1. */
0447e8df 189#define UTF8_IS_START(c) (((U8)c) >= 0xc2)
15824458
KW
190
191/* Is the byte 'c' part of a multi-byte UTF8-8 encoded sequence, and not the
192 * first byte thereof? */
e021c6e6 193#define UTF8_IS_CONTINUATION(c) ((((U8)c) & 0xC0) == 0x80)
0ae1fa71 194
15824458
KW
195/* Is the UTF8-encoded byte 'c' the first byte of a two byte sequence? Use
196 * UTF8_IS_NEXT_CHAR_DOWNGRADEABLE() instead if the input isn't known to
197 * be well-formed. Masking with 0xfe allows the low bit to be 0 or 1; thus
198 * this matches 0xc[23]. */
559c7f10 199#define UTF8_IS_DOWNGRADEABLE_START(c) (((U8)(c) & 0xfe) == 0xc2)
4ab10950 200
15824458
KW
201/* Is the UTF8-encoded byte 'c' the first byte of a sequence of bytes that
202 * represent a code point > 255? */
7b4252f4 203#define UTF8_IS_ABOVE_LATIN1(c) ((U8)(c) >= 0xc4)
8850bf83 204
15824458
KW
205/* This defines the 1-bits that are to be in the first byte of a multi-byte
206 * UTF-8 encoded character that give the number of bytes that comprise the
f710bf0c 207 * character. 'len' is the number of bytes in the multi-byte sequence. */
ee372ee9 208#define UTF_START_MARK(len) (((len) > 7) ? 0xFF : (0xFF & (0xFE << (7-(len)))))
b2635aa8
KW
209
210/* Masks out the initial one bits in a start byte, leaving the real data ones.
f710bf0c
KW
211 * Doesn't work on an invariant byte. 'len' is the number of bytes in the
212 * multi-byte sequence that comprises the character. */
22901f30 213#define UTF_START_MASK(len) (((len) >= 7) ? 0x00 : (0x1F >> ((len)-2)))
1d72bdf6 214
15824458
KW
215/* This defines the bits that are to be in the continuation bytes of a multi-byte
216 * UTF-8 encoded character that indicate it is a continuation byte. */
1d72bdf6 217#define UTF_CONTINUATION_MARK 0x80
15824458
KW
218
219/* This is the number of low-order bits a continuation byte in a UTF-8 encoded
220 * sequence contributes to the specification of the code point. In the bit
221 * maps above, you see that the first 2 bits are a constant '10', leaving 6 of
222 * real information */
1d72bdf6 223#define UTF_ACCUMULATION_SHIFT 6
b2635aa8
KW
224
225/* 2**UTF_ACCUMULATION_SHIFT - 1 */
1d72bdf6 226#define UTF_CONTINUATION_MASK ((U8)0x3f)
c512ce4f 227
15824458
KW
228/* If a value is anded with this, and the result is non-zero, then using the
229 * original value in UTF8_ACCUMULATE will overflow, shifting bits off the left
230 * */
eb83ed87 231#define UTF_ACCUMULATION_OVERFLOW_MASK \
bb88be5f
KW
232 (((UV) UTF_CONTINUATION_MASK) << ((sizeof(UV) * CHARBITS) \
233 - UTF_ACCUMULATION_SHIFT))
eb83ed87 234
6588300d
NC
235#if UVSIZE >= 8
236# define UTF8_QUAD_MAX UINT64_C(0x1000000000)
237
5aaebcb3
KW
238/* Input is a true Unicode (not-native) code point */
239#define OFFUNISKIP(uv) ( (uv) < 0x80 ? 1 : \
1d68d6cd
SC
240 (uv) < 0x800 ? 2 : \
241 (uv) < 0x10000 ? 3 : \
242 (uv) < 0x200000 ? 4 : \
243 (uv) < 0x4000000 ? 5 : \
244 (uv) < 0x80000000 ? 6 : \
9041c2e3 245 (uv) < UTF8_QUAD_MAX ? 7 : 13 )
1d68d6cd
SC
246#else
247/* No, I'm not even going to *TRY* putting #ifdef inside a #define */
5aaebcb3 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 : 7 )
254#endif
255
03c76984
KW
256/* How wide can a single UTF-8 encoded character become in bytes. */
257/* NOTE: Strictly speaking Perl's UTF-8 should not be called UTF-8 since UTF-8
258 * is an encoding of Unicode, and Unicode's upper limit, 0x10FFFF, can be
259 * expressed with 4 bytes. However, Perl thinks of UTF-8 as a way to encode
260 * non-negative integers in a binary format, even those above Unicode */
261#define UTF8_MAXBYTES 13
262
263/* The maximum number of UTF-8 bytes a single Unicode character can
264 * uppercase/lowercase/fold into. Unicode guarantees that the maximum
265 * expansion is 3 characters. On ASCIIish platforms, the highest Unicode
266 * character occupies 4 bytes, therefore this number would be 12, but this is
267 * smaller than the maximum width a single above-Unicode character can occupy,
268 * so use that instead */
269#if UTF8_MAXBYTES < 12
270#error UTF8_MAXBYTES must be at least 12
271#endif
272
843a4590
KW
273#define MAX_UTF8_TWO_BYTE 0x7FF
274
03c76984
KW
275#define UTF8_MAXBYTES_CASE UTF8_MAXBYTES
276
d06134e5
KW
277#endif /* EBCDIC vs ASCII */
278
279/* Rest of these are attributes of Unicode and perl's internals rather than the
280 * encoding, or happen to be the same in both ASCII and EBCDIC (at least at
281 * this level; the macros that some of these call may have different
282 * definitions in the two encodings */
283
59a449d5
KW
284/* In domain restricted to ASCII, these may make more sense to the reader than
285 * the ones with Latin1 in the name */
286#define NATIVE_TO_ASCII(ch) NATIVE_TO_LATIN1(ch)
287#define ASCII_TO_NATIVE(ch) LATIN1_TO_NATIVE(ch)
288
289/* More or less misleadingly-named defines, retained for back compat */
290#define NATIVE_TO_UTF(ch) NATIVE_UTF8_TO_I8(ch)
291#define NATIVE_TO_I8(ch) NATIVE_UTF8_TO_I8(ch)
292#define UTF_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch)
293#define I8_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch)
294#define NATIVE8_TO_UNI(ch) NATIVE_TO_LATIN1(ch)
d06134e5 295
537124e4
KW
296/* Adds a UTF8 continuation byte 'new' of information to a running total code
297 * point 'old' of all the continuation bytes so far. This is designed to be
155d2738
KW
298 * used in a loop to convert from UTF-8 to the code point represented. Note
299 * that this is asymmetric on EBCDIC platforms, in that the 'new' parameter is
300 * the UTF-EBCDIC byte, whereas the 'old' parameter is a Unicode (not EBCDIC)
301 * code point in process of being generated */
302#define UTF8_ACCUMULATE(old, new) (((old) << UTF_ACCUMULATION_SHIFT) \
303 | ((NATIVE_UTF8_TO_I8((U8)new)) \
304 & UTF_CONTINUATION_MASK))
d06134e5 305
4ab10950
KW
306/* This works in the face of malformed UTF-8. */
307#define UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(s, e) (UTF8_IS_DOWNGRADEABLE_START(*s) \
308 && ( (e) - (s) > 1) \
309 && UTF8_IS_CONTINUATION(*((s)+1)))
310
5aaebcb3
KW
311/* Number of bytes a code point occupies in UTF-8. */
312#define NATIVE_SKIP(uv) OFFUNISKIP(NATIVE_TO_UNI(uv))
313
314/* Most code which says UNISKIP is really thinking in terms of native code
315 * points (0-255) plus all those beyond. This is an imprecise term, but having
316 * it means existing code continues to work. For precision, use NATIVE_SKIP
317 * and OFFUNISKIP */
318#define UNISKIP(uv) NATIVE_SKIP(uv)
319
94bb8c36 320/* Convert a two (not one) byte utf8 character to a native code point value.
2950f2a7
KW
321 * Needs just one iteration of accumulate. Should not be used unless it is
322 * known that the two bytes are legal: 1) two-byte start, and 2) continuation.
323 * Note that the result can be larger than 255 if the input character is not
324 * downgradable */
94bb8c36
KW
325#define TWO_BYTE_UTF8_TO_NATIVE(HI, LO) \
326 UNI_TO_NATIVE(UTF8_ACCUMULATE((NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), \
155d2738 327 (LO)))
94bb8c36
KW
328
329/* Should never be used, and be deprecated */
330#define TWO_BYTE_UTF8_TO_UNI(HI, LO) NATIVE_TO_UNI(TWO_BYTE_UTF8_TO_NATIVE(HI, LO))
2950f2a7 331
15824458
KW
332/* How many bytes in the UTF-8 encoded character whose first (perhaps only)
333 * byte is pointed to by 's' */
d06134e5
KW
334#define UTF8SKIP(s) PL_utf8skip[*(const U8*)(s)]
335
15824458
KW
336/* Is the byte 'c' the same character when encoded in UTF-8 as when not. This
337 * works on both UTF-8 encoded strings and non-encoded, as it returns TRUE in
338 * each for the exact same set of bit patterns. (And it works on any byte in a
339 * UTF-8 encoded string) */
bc3632a8 340#define UTF8_IS_INVARIANT(c) UNI_IS_INVARIANT(NATIVE_UTF8_TO_I8(c))
15824458 341
6f2d5cbc
KW
342/* Like the above, but its name implies a non-UTF8 input, and is implemented
343 * differently (for no particular reason) */
344#define NATIVE_BYTE_IS_INVARIANT(c) UNI_IS_INVARIANT(NATIVE_TO_LATIN1(c))
345
346/* Like the above, but accepts any UV as input */
347#define UVCHR_IS_INVARIANT(uv) UNI_IS_INVARIANT(NATIVE_TO_UNI(uv))
d06134e5
KW
348
349#define MAX_PORTABLE_UTF8_TWO_BYTE 0x3FF /* constrained by EBCDIC */
350
48ccf5e1
KW
351/* The macros in the next 4 sets are used to generate the two utf8 or utfebcdic
352 * bytes from an ordinal that is known to fit into exactly two (not one) bytes;
353 * it must be less than 0x3FF to work across both encodings. */
354
355/* These two are helper macros for the other three sets, and should not be used
356 * directly anywhere else. 'translate_function' is either NATIVE_TO_LATIN1
357 * (which works for code points up to 0xFF) or NATIVE_TO_UNI which works for any
358 * code point */
359#define __BASE_TWO_BYTE_HI(c, translate_function) \
360 I8_TO_NATIVE_UTF8((translate_function(c) >> UTF_ACCUMULATION_SHIFT) \
361 | UTF_START_MARK(2))
362#define __BASE_TWO_BYTE_LO(c, translate_function) \
363 I8_TO_NATIVE_UTF8((translate_function(c) & UTF_CONTINUATION_MASK) \
364 | UTF_CONTINUATION_MARK)
365
48ccf5e1
KW
366/* The next two macros should not be used. They were designed to be usable as
367 * the case label of a switch statement, but this doesn't work for EBCDIC. Use
9d0d3a03 368 * regen/unicode_constants.pl instead */
48ccf5e1
KW
369#define UTF8_TWO_BYTE_HI_nocast(c) __BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI)
370#define UTF8_TWO_BYTE_LO_nocast(c) __BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI)
371
372/* The next two macros are used when the source should be a single byte
373 * character; checked for under DEBUGGING */
374#define UTF8_EIGHT_BIT_HI(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
375 ((U8) __BASE_TWO_BYTE_HI(c, NATIVE_TO_LATIN1)))
376#define UTF8_EIGHT_BIT_LO(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
377 ((U8) __BASE_TWO_BYTE_LO(c, NATIVE_TO_LATIN1)))
378
379/* These final two macros in the series are used when the source can be any
380 * code point whose UTF-8 is known to occupy 2 bytes; they are less efficient
381 * than the EIGHT_BIT versions on EBCDIC platforms. We use the logical '~'
382 * operator instead of "<=" to avoid getting compiler warnings.
383 * MAX_PORTABLE_UTF8_TWO_BYTE should be exactly all one bits in the lower few
384 * places, so the ~ works */
385#define UTF8_TWO_BYTE_HI(c) \
386 (__ASSERT_((sizeof(c) == 1) \
387 || !(((WIDEST_UTYPE)(c)) & ~MAX_PORTABLE_UTF8_TWO_BYTE)) \
388 ((U8) __BASE_TWO_BYTE_HI(c, NATIVE_TO_LATIN1)))
389#define UTF8_TWO_BYTE_LO(c) \
390 (__ASSERT_((sizeof(c) == 1) \
391 || !(((WIDEST_UTYPE)(c)) & ~MAX_PORTABLE_UTF8_TWO_BYTE)) \
392 ((U8) __BASE_TWO_BYTE_LO(c, NATIVE_TO_LATIN1)))
d06134e5 393
e7214ce8
KW
394/* This is illegal in any well-formed UTF-8 in both EBCDIC and ASCII
395 * as it is only in overlongs. */
396#define ILLEGAL_UTF8_BYTE I8_TO_NATIVE_UTF8(0xC1)
397
7e2040f0 398/*
e3036cf4 399 * 'UTF' is whether or not p is encoded in UTF8. The names 'foo_lazy_if' stem
20df05f4
KW
400 * from an earlier version of these macros in which they didn't call the
401 * foo_utf8() macros (i.e. were 'lazy') unless they decided that *p is the
402 * beginning of a utf8 character. Now that foo_utf8() determines that itself,
403 * no need to do it again here
7e2040f0 404 */
e3036cf4
KW
405#define isIDFIRST_lazy_if(p,UTF) ((IN_BYTES || !UTF ) \
406 ? isIDFIRST(*(p)) \
407 : isIDFIRST_utf8((const U8*)p))
32636478
KW
408#define isWORDCHAR_lazy_if(p,UTF) ((IN_BYTES || (!UTF )) \
409 ? isWORDCHAR(*(p)) \
410 : isWORDCHAR_utf8((const U8*)p))
411#define isALNUM_lazy_if(p,UTF) isWORDCHAR_lazy_if(p,UTF)
1d72bdf6 412
89ebb4a3
JH
413#define UTF8_MAXLEN UTF8_MAXBYTES
414
8cb75cc8
KW
415/* A Unicode character can fold to up to 3 characters */
416#define UTF8_MAX_FOLD_CHAR_EXPAND 3
417
a98fe34d 418#define IN_BYTES (CopHINTS_get(PL_curcop) & HINT_BYTES)
0064a8a9 419#define DO_UTF8(sv) (SvUTF8(sv) && !IN_BYTES)
66cbab2c 420#define IN_UNI_8_BIT \
b36bf33f
KW
421 (CopHINTS_get(PL_curcop) & (HINT_UNI_8_BIT|HINT_LOCALE_NOT_CHARS) \
422 && ! IN_LOCALE_RUNTIME && ! IN_BYTES)
423
1d72bdf6 424
c76687c5
KW
425#define UTF8_ALLOW_EMPTY 0x0001 /* Allow a zero length string */
426
427/* Allow first byte to be a continuation byte */
1d72bdf6 428#define UTF8_ALLOW_CONTINUATION 0x0002
c76687c5
KW
429
430/* Allow second... bytes to be non-continuation bytes */
1d72bdf6 431#define UTF8_ALLOW_NON_CONTINUATION 0x0004
949cf498
KW
432
433/* expecting more bytes than were available in the string */
434#define UTF8_ALLOW_SHORT 0x0008
435
436/* Overlong sequence; i.e., the code point can be specified in fewer bytes. */
437#define UTF8_ALLOW_LONG 0x0010
438
439#define UTF8_DISALLOW_SURROGATE 0x0020 /* Unicode surrogates */
440#define UTF8_WARN_SURROGATE 0x0040
441
442#define UTF8_DISALLOW_NONCHAR 0x0080 /* Unicode non-character */
443#define UTF8_WARN_NONCHAR 0x0100 /* code points */
444
445#define UTF8_DISALLOW_SUPER 0x0200 /* Super-set of Unicode: code */
446#define UTF8_WARN_SUPER 0x0400 /* points above the legal max */
447
448/* Code points which never were part of the original UTF-8 standard, the first
ea5ced44
KW
449 * byte of which is a FE or FF on ASCII platforms. If the first byte is FF, it
450 * will overflow a 32-bit word. If the first byte is FE, it will overflow a
451 * signed 32-bit word. */
949cf498
KW
452#define UTF8_DISALLOW_FE_FF 0x0800
453#define UTF8_WARN_FE_FF 0x1000
454
455#define UTF8_CHECK_ONLY 0x2000
456
457/* For backwards source compatibility. They do nothing, as the default now
458 * includes what they used to mean. The first one's meaning was to allow the
459 * just the single non-character 0xFFFF */
460#define UTF8_ALLOW_FFFF 0
461#define UTF8_ALLOW_SURROGATE 0
462
33d9abfb 463#define UTF8_DISALLOW_ILLEGAL_INTERCHANGE (UTF8_DISALLOW_SUPER|UTF8_DISALLOW_NONCHAR|UTF8_DISALLOW_SURROGATE|UTF8_DISALLOW_FE_FF)
949cf498 464#define UTF8_WARN_ILLEGAL_INTERCHANGE \
33d9abfb 465 (UTF8_WARN_SUPER|UTF8_WARN_NONCHAR|UTF8_WARN_SURROGATE|UTF8_WARN_FE_FF)
949cf498
KW
466#define UTF8_ALLOW_ANY \
467 (~(UTF8_DISALLOW_ILLEGAL_INTERCHANGE|UTF8_WARN_ILLEGAL_INTERCHANGE))
468#define UTF8_ALLOW_ANYUV \
469 (UTF8_ALLOW_EMPTY \
470 & ~(UTF8_DISALLOW_ILLEGAL_INTERCHANGE|UTF8_WARN_ILLEGAL_INTERCHANGE))
9f7f3913
ST
471#define UTF8_ALLOW_DEFAULT (ckWARN(WARN_UTF8) ? 0 : \
472 UTF8_ALLOW_ANYUV)
1d72bdf6 473
7131f24d
KW
474/* Surrogates, non-character code points and above-Unicode code points are
475 * problematic in some contexts. This allows code that needs to check for
476 * those to to quickly exclude the vast majority of code points it will
477 * encounter */
b96a92fb
KW
478#define UTF8_FIRST_PROBLEMATIC_CODE_POINT_FIRST_BYTE \
479 FIRST_SURROGATE_UTF8_FIRST_BYTE
7131f24d 480
b96a92fb
KW
481#define UTF8_IS_SURROGATE(s) cBOOL(is_SURROGATE_utf8(s))
482#define UTF8_IS_REPLACEMENT(s, send) cBOOL(is_REPLACEMENT_utf8_safe(s,send))
7131f24d
KW
483
484/* ASCII EBCDIC I8
485 * U+10FFFF: \xF4\x8F\xBF\xBF \xF9\xA1\xBF\xBF\xBF max legal Unicode
486 * U+110000: \xF4\x90\x80\x80 \xF9\xA2\xA0\xA0\xA0
487 * U+110001: \xF4\x90\x80\x81 \xF9\xA2\xA0\xA0\xA1
488 */
489#ifdef EBCDIC /* Both versions assume well-formed UTF8 */
bc3632a8
KW
490# define UTF8_IS_SUPER(s) (NATIVE_UTF8_TO_I8(* (U8*) (s)) >= 0xF9 \
491 && (NATIVE_UTF8_TO_I8(* (U8*) (s)) > 0xF9 \
a3481822 492 || (NATIVE_UTF8_TO_I8(* ((U8*) (s) + 1)) >= 0xA2)))
7131f24d 493#else
0404c92a
KW
494# define UTF8_IS_SUPER(s) (*(U8*) (s) >= 0xF4 \
495 && (*(U8*) (s) > 0xF4 || (*((U8*) (s) + 1) >= 0x90)))
7131f24d
KW
496#endif
497
b96a92fb
KW
498/* These are now machine generated, and the 'given' clause is no longer
499 * applicable */
500#define UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s) \
501 cBOOL(is_NONCHAR_utf8(s))
502#define UTF8_IS_NONCHAR_(s) \
503 UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s)
7131f24d 504
c867b360
JH
505#define UNICODE_SURROGATE_FIRST 0xD800
506#define UNICODE_SURROGATE_LAST 0xDFFF
507#define UNICODE_REPLACEMENT 0xFFFD
508#define UNICODE_BYTE_ORDER_MARK 0xFEFF
1d72bdf6 509
b851fbc1 510/* Though our UTF-8 encoding can go beyond this,
c76687c5 511 * let's be conservative and do as Unicode says. */
b851fbc1
JH
512#define PERL_UNICODE_MAX 0x10FFFF
513
949cf498
KW
514#define UNICODE_WARN_SURROGATE 0x0001 /* UTF-16 surrogates */
515#define UNICODE_WARN_NONCHAR 0x0002 /* Non-char code points */
516#define UNICODE_WARN_SUPER 0x0004 /* Above 0x10FFFF */
517#define UNICODE_WARN_FE_FF 0x0008 /* Above 0x10FFFF */
518#define UNICODE_DISALLOW_SURROGATE 0x0010
519#define UNICODE_DISALLOW_NONCHAR 0x0020
520#define UNICODE_DISALLOW_SUPER 0x0040
521#define UNICODE_DISALLOW_FE_FF 0x0080
bb88be5f
KW
522#define UNICODE_WARN_ILLEGAL_INTERCHANGE \
523 (UNICODE_WARN_SURROGATE|UNICODE_WARN_NONCHAR|UNICODE_WARN_SUPER)
524#define UNICODE_DISALLOW_ILLEGAL_INTERCHANGE \
525 (UNICODE_DISALLOW_SURROGATE|UNICODE_DISALLOW_NONCHAR|UNICODE_DISALLOW_SUPER)
949cf498
KW
526
527/* For backward source compatibility, as are now the default */
528#define UNICODE_ALLOW_SURROGATE 0
529#define UNICODE_ALLOW_SUPER 0
530#define UNICODE_ALLOW_ANY 0
b851fbc1 531
1d72bdf6
NIS
532#define UNICODE_IS_SURROGATE(c) ((c) >= UNICODE_SURROGATE_FIRST && \
533 (c) <= UNICODE_SURROGATE_LAST)
a10ec373 534#define UNICODE_IS_REPLACEMENT(c) ((c) == UNICODE_REPLACEMENT)
872c91ae 535#define UNICODE_IS_BYTE_ORDER_MARK(c) ((c) == UNICODE_BYTE_ORDER_MARK)
7131f24d
KW
536#define UNICODE_IS_NONCHAR(c) ((c >= 0xFDD0 && c <= 0xFDEF) \
537 /* The other noncharacters end in FFFE or FFFF, which \
538 * the mask below catches both of, but beyond the last \
539 * official unicode code point, they aren't \
540 * noncharacters, since those aren't Unicode \
541 * characters at all */ \
542 || ((((c & 0xFFFE) == 0xFFFE)) && ! UNICODE_IS_SUPER(c)))
543#define UNICODE_IS_SUPER(c) ((c) > PERL_UNICODE_MAX)
544#define UNICODE_IS_FE_FF(c) ((c) > 0x7FFFFFFF)
1d72bdf6 545
ec34087a
KW
546#define LATIN_SMALL_LETTER_SHARP_S LATIN_SMALL_LETTER_SHARP_S_NATIVE
547#define LATIN_SMALL_LETTER_Y_WITH_DIAERESIS \
548 LATIN_SMALL_LETTER_Y_WITH_DIAERESIS_NATIVE
549#define MICRO_SIGN MICRO_SIGN_NATIVE
550#define LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE \
551 LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE_NATIVE
552#define LATIN_SMALL_LETTER_A_WITH_RING_ABOVE \
553 LATIN_SMALL_LETTER_A_WITH_RING_ABOVE_NATIVE
09091399
JH
554#define UNICODE_GREEK_CAPITAL_LETTER_SIGMA 0x03A3
555#define UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA 0x03C2
556#define UNICODE_GREEK_SMALL_LETTER_SIGMA 0x03C3
9dcbe121 557#define GREEK_SMALL_LETTER_MU 0x03BC
9e682c18
KW
558#define GREEK_CAPITAL_LETTER_MU 0x039C /* Upper and title case
559 of MICRON */
560#define LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS 0x0178 /* Also is title case */
561#define LATIN_CAPITAL_LETTER_SHARP_S 0x1E9E
562#define LATIN_SMALL_LETTER_LONG_S 0x017F
a9f50d33
KW
563#define LATIN_SMALL_LIGATURE_LONG_S_T 0xFB05
564#define LATIN_SMALL_LIGATURE_ST 0xFB06
9e682c18
KW
565#define KELVIN_SIGN 0x212A
566#define ANGSTROM_SIGN 0x212B
09091399 567
9e55ce06 568#define UNI_DISPLAY_ISPRINT 0x0001
c728cb41
JH
569#define UNI_DISPLAY_BACKSLASH 0x0002
570#define UNI_DISPLAY_QQ (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH)
571#define UNI_DISPLAY_REGEX (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH)
9e55ce06 572
5cd46e1f
KW
573#define ANYOF_FOLD_SHARP_S(node, input, end) \
574 (ANYOF_BITMAP_TEST(node, LATIN_SMALL_LETTER_SHARP_S) && \
137165a6 575 (ANYOF_NONBITMAP(node)) && \
39065660 576 (ANYOF_FLAGS(node) & ANYOF_LOC_NONBITMAP_FOLD) && \
07b6858f 577 ((end) > (input) + 1) && \
d22b930b
KW
578 toFOLD((input)[0]) == 's' && \
579 toFOLD((input)[1]) == 's')
6302f837 580
ebc501f0 581#define SHARP_S_SKIP 2
3b0fc154 582
a4f7a67c
KW
583/* If you want to exclude surrogates, and beyond legal Unicode, see the blame
584 * log for earlier versions which gave details for these */
4d646140 585
6302f837
KW
586/* A helper macro for isUTF8_CHAR, so use that one, and not this one. This is
587 * retained solely for backwards compatibility and may be deprecated and
588 * removed in a future Perl version.
589 *
590 * regen/regcharclass.pl generates is_UTF8_CHAR_utf8() macros for up to these
d9f92374
KW
591 * number of bytes. So this has to be coordinated with that file */
592#ifdef EBCDIC
593# define IS_UTF8_CHAR_FAST(n) ((n) <= 3)
594#else
595# define IS_UTF8_CHAR_FAST(n) ((n) <= 4)
596#endif
597
4d646140 598#ifndef EBCDIC
6302f837
KW
599/* A helper macro for isUTF8_CHAR, so use that one instead of this. This was
600 * generated by regen/regcharclass.pl, and then moved here. The lines that
601 * generated it were then commented out. This was done solely because it takes
602 * on the order of 10 minutes to generate, and is never going to change, unless
603 * the generated code is improved.
604 *
605 * The EBCDIC versions have been cut to not cover all of legal Unicode, so
606 * don't take too long to generate, and there is a separate one for each code
607 * page, so they are in regcharclass.h instead of here */
39a0f513
YO
608/*
609 UTF8_CHAR: Matches utf8 from 1 to 4 bytes
610
611 0x0 - 0x1FFFFF
612*/
4d646140
KW
613/*** GENERATED CODE ***/
614#define is_UTF8_CHAR_utf8_safe(s,e) \
615( ((e)-(s) > 3) ? \
616 ( ( ( ((U8*)s)[0] & 0x80 ) == 0x00 ) ? 1 \
617 : ( 0xC2 <= ((U8*)s)[0] && ((U8*)s)[0] <= 0xDF ) ? \
618 ( ( ( ((U8*)s)[1] & 0xC0 ) == 0x80 ) ? 2 : 0 ) \
619 : ( 0xE0 == ((U8*)s)[0] ) ? \
620 ( ( ( ( ((U8*)s)[1] & 0xE0 ) == 0xA0 ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
621 : ( 0xE1 <= ((U8*)s)[0] && ((U8*)s)[0] <= 0xEF ) ? \
622 ( ( ( ( ((U8*)s)[1] & 0xC0 ) == 0x80 ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
623 : ( 0xF0 == ((U8*)s)[0] ) ? \
624 ( ( ( ( 0x90 <= ((U8*)s)[1] && ((U8*)s)[1] <= 0xBF ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) && ( ( ((U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\
39a0f513 625 : ( ( ( ( 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
KW
626: ((e)-(s) > 2) ? \
627 ( ( ( ((U8*)s)[0] & 0x80 ) == 0x00 ) ? 1 \
628 : ( 0xC2 <= ((U8*)s)[0] && ((U8*)s)[0] <= 0xDF ) ? \
629 ( ( ( ((U8*)s)[1] & 0xC0 ) == 0x80 ) ? 2 : 0 ) \
630 : ( 0xE0 == ((U8*)s)[0] ) ? \
631 ( ( ( ( ((U8*)s)[1] & 0xE0 ) == 0xA0 ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
39a0f513 632 : ( ( ( 0xE1 <= ((U8*)s)[0] && ((U8*)s)[0] <= 0xEF ) && ( ( ((U8*)s)[1] & 0xC0 ) == 0x80 ) ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
4d646140
KW
633: ((e)-(s) > 1) ? \
634 ( ( ( ((U8*)s)[0] & 0x80 ) == 0x00 ) ? 1 \
39a0f513 635 : ( ( 0xC2 <= ((U8*)s)[0] && ((U8*)s)[0] <= 0xDF ) && ( ( ((U8*)s)[1] & 0xC0 ) == 0x80 ) ) ? 2 : 0 )\
4d646140
KW
636: ((e)-(s) > 0) ? \
637 ( ( ((U8*)s)[0] & 0x80 ) == 0x00 ) \
638: 0 )
639#endif
3b0fc154 640
6302f837
KW
641/*
642 * =for apidoc isUTF8_CHAR
643 *
644 * Returns the number of bytes beginning at C<s> which form a legal UTF-8 (or
645 * UTF-EBCDIC) encoded character, looking no further than C<e - s> bytes into
646 * C<s>. Returns 0 if the sequence starting at C<s> through C<e - 1> is not
647 * well-formed UTF-8
648
649Note that an INVARIANT character (i.e. ASCII on non-EBCDIC
650machines) is a valid UTF-8 character. */
651
652#define isUTF8_CHAR(s, e) (((e) <= (s)) \
653 ? 0 \
654 : (UTF8_IS_INVARIANT(*s)) \
655 ? 1 \
656 : (((e) - (s)) < UTF8SKIP(s)) \
657 ? 0 \
658 : (IS_UTF8_CHAR_FAST(UTF8SKIP(s))) \
659 ? is_UTF8_CHAR_utf8_safe(s,e) \
660 : _is_utf8_char_slow(s, e))
661
662/* Do not use; should be deprecated. Use isUTF8_CHAR() instead; this is
663 * retained solely for backwards compatibility */
664#define IS_UTF8_CHAR(p, n) (isUTF8_CHAR(p, (p) + (n)) == n)
e9a8c099 665
57f0e7e2
KW
666#endif /* H_UTF8 */
667
e9a8c099
MHM
668/*
669 * Local variables:
670 * c-indentation-style: bsd
671 * c-basic-offset: 4
14d04a33 672 * indent-tabs-mode: nil
e9a8c099
MHM
673 * End:
674 *
14d04a33 675 * ex: set ts=8 sts=4 sw=4 et:
e9a8c099 676 */