This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
fold_grind.t: Use LC_CTYPE, not LC_ALL
[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)
051a06d4
KW
50#define to_utf8_fold(c, p, lenp) _to_utf8_fold_flags(c, p, lenp, \
51 FOLD_FLAGS_FULL, NULL)
52#define to_utf8_lower(a,b,c) _to_utf8_lower_flags(a,b,c,0, NULL)
53#define to_utf8_upper(a,b,c) _to_utf8_upper_flags(a,b,c,0, NULL)
54#define to_utf8_title(a,b,c) _to_utf8_title_flags(a,b,c,0, NULL)
36bb2ab6 55
fd7cb289 56/* Source backward compatibility. */
fd7cb289
RGS
57#define is_utf8_string_loc(s, len, ep) is_utf8_string_loclen(s, len, ep, 0)
58
eda9cac1
KW
59#define foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \
60 foldEQ_utf8_flags(s1, pe1, l1, u1, s2, pe2, l2, u2, 0)
a33c29bc 61#define FOLDEQ_UTF8_NOMIX_ASCII (1 << 0)
5e64d0fa 62#define FOLDEQ_UTF8_LOCALE (1 << 1)
18f762c3
KW
63#define FOLDEQ_S1_ALREADY_FOLDED (1 << 2)
64#define FOLDEQ_S2_ALREADY_FOLDED (1 << 3)
a33c29bc 65
e6226b18
KW
66/*
67=for apidoc ibcmp_utf8
68
69This is a synonym for (! foldEQ_utf8())
70
71=cut
72*/
73#define ibcmp_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \
74 cBOOL(! foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2))
75
1d72bdf6
NIS
76#ifdef EBCDIC
77/* The equivalent of these macros but implementing UTF-EBCDIC
78 are in the following header file:
79 */
80
81#include "utfebcdic.h"
fd7cb289 82
d06134e5 83#else /* ! EBCDIC */
73c4f7a1
GS
84START_EXTERN_C
85
a0ed51b3 86#ifdef DOINIT
6f06b55f 87EXTCONST unsigned char PL_utf8skip[] = {
b2635aa8
KW
88/* 0x00 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
89/* 0x10 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
90/* 0x20 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
91/* 0x30 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
92/* 0x40 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
93/* 0x50 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
94/* 0x60 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
95/* 0x70 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
96/* 0x80 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
97/* 0x90 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
98/* 0xA0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
99/* 0xB0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
100/* 0xC0 */ 2,2, /* overlong */
101/* 0xC2 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0080 to U+03FF */
102/* 0xD0 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0400 to U+07FF */
103/* 0xE0 */ 3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3, /* U+0800 to U+FFFF */
104/* 0xF0 */ 4,4,4,4,4,4,4,4,5,5,5,5,6,6, /* above BMP to 2**31 - 1 */
105/* 0xFE */ 7,13, /* Perl extended (never was official UTF-8). Up to 72bit
106 allowed (64-bit + reserved). */
a0ed51b3
LW
107};
108#else
6f06b55f 109EXTCONST unsigned char PL_utf8skip[];
a0ed51b3
LW
110#endif
111
73c4f7a1 112END_EXTERN_C
7e2040f0 113
59a449d5
KW
114/* Native character to/from iso-8859-1. Are the identity functions on ASCII
115 * platforms */
116#define NATIVE_TO_LATIN1(ch) (ch)
117#define LATIN1_TO_NATIVE(ch) (ch)
118
119/* I8 is an intermediate version of UTF-8 used only in UTF-EBCDIC. We thus
120 * consider it to be identical to UTF-8 on ASCII platforms. Strictly speaking
121 * UTF-8 and UTF-EBCDIC are two different things, but we often conflate them
122 * because they are 8-bit encodings that serve the same purpose in Perl, and
123 * rarely do we need to distinguish them. The term "NATIVE_UTF8" applies to
124 * whichever one is applicable on the current platform */
125#define NATIVE_UTF8_TO_I8(ch) (ch)
126#define I8_TO_NATIVE_UTF8(ch) (ch)
127
1d72bdf6
NIS
128/* Transforms in wide UV chars */
129#define UNI_TO_NATIVE(ch) (ch)
130#define NATIVE_TO_UNI(ch) (ch)
d7578b48 131
877d9f0d 132/*
9041c2e3 133
8c007b5a 134 The following table is from Unicode 3.2.
877d9f0d
JH
135
136 Code Points 1st Byte 2nd Byte 3rd Byte 4th Byte
137
375122d7 138 U+0000..U+007F 00..7F
e1b711da 139 U+0080..U+07FF * C2..DF 80..BF
37e2e78e 140 U+0800..U+0FFF E0 * A0..BF 80..BF
375122d7 141 U+1000..U+CFFF E1..EC 80..BF 80..BF
e1b711da 142 U+D000..U+D7FF ED 80..9F 80..BF
537124e4 143 U+D800..U+DFFF ED A0..BF 80..BF (surrogates)
375122d7 144 U+E000..U+FFFF EE..EF 80..BF 80..BF
37e2e78e 145 U+10000..U+3FFFF F0 * 90..BF 80..BF 80..BF
877d9f0d
JH
146 U+40000..U+FFFFF F1..F3 80..BF 80..BF 80..BF
147 U+100000..U+10FFFF F4 80..8F 80..BF 80..BF
b2635aa8
KW
148 Below are non-Unicode code points
149 U+110000..U+13FFFF F4 90..BF 80..BF 80..BF
150 U+110000..U+1FFFFF F5..F7 80..BF 80..BF 80..BF
537124e4 151 U+200000..: F8.. * 88..BF 80..BF 80..BF 80..BF
877d9f0d 152
e1b711da 153Note the gaps before several of the byte entries above marked by '*'. These are
37e2e78e
KW
154caused by legal UTF-8 avoiding non-shortest encodings: it is technically
155possible to UTF-8-encode a single code point in different ways, but that is
156explicitly forbidden, and the shortest possible encoding should always be used
15824458 157(and that is what Perl does). The non-shortest ones are called 'overlongs'.
8c007b5a 158
877d9f0d
JH
159 */
160
8c007b5a
JH
161/*
162 Another way to look at it, as bits:
163
b2635aa8 164 Code Points 1st Byte 2nd Byte 3rd Byte 4th Byte
8c007b5a 165
b2635aa8
KW
166 0aaa aaaa 0aaa aaaa
167 0000 0bbb bbaa aaaa 110b bbbb 10aa aaaa
168 cccc bbbb bbaa aaaa 1110 cccc 10bb bbbb 10aa aaaa
169 00 000d ddcc cccc bbbb bbaa aaaa 1111 0ddd 10cc cccc 10bb bbbb 10aa aaaa
8c007b5a
JH
170
171As you can see, the continuation bytes all begin with C<10>, and the
e1b711da 172leading bits of the start byte tell how many bytes there are in the
8c007b5a
JH
173encoded character.
174
65ab9279
TC
175Perl's extended UTF-8 means we can have start bytes up to FF.
176
8c007b5a
JH
177*/
178
15824458
KW
179/* Is the representation of the Unicode code point 'c' the same regardless of
180 * being encoded in UTF-8 or not? */
c4d5f83a 181#define UNI_IS_INVARIANT(c) (((UV)c) < 0x80)
15824458
KW
182
183/* Is the UTF8-encoded byte 'c' part of a variant sequence in UTF-8? This is
184 * the inverse of UTF8_IS_INVARIANT */
185#define UTF8_IS_CONTINUED(c) (((U8)c) & 0x80)
186
187/* Is the byte 'c' the first byte of a multi-byte UTF8-8 encoded sequence?
188 * This doesn't catch invariants (they are single-byte). It also excludes the
189 * illegal overlong sequences that begin with C0 and C1. */
0447e8df 190#define UTF8_IS_START(c) (((U8)c) >= 0xc2)
15824458
KW
191
192/* Is the byte 'c' part of a multi-byte UTF8-8 encoded sequence, and not the
193 * first byte thereof? */
e021c6e6 194#define UTF8_IS_CONTINUATION(c) ((((U8)c) & 0xC0) == 0x80)
0ae1fa71 195
15824458
KW
196/* Is the UTF8-encoded byte 'c' the first byte of a two byte sequence? Use
197 * UTF8_IS_NEXT_CHAR_DOWNGRADEABLE() instead if the input isn't known to
198 * be well-formed. Masking with 0xfe allows the low bit to be 0 or 1; thus
199 * this matches 0xc[23]. */
559c7f10 200#define UTF8_IS_DOWNGRADEABLE_START(c) (((U8)(c) & 0xfe) == 0xc2)
4ab10950 201
15824458
KW
202/* Is the UTF8-encoded byte 'c' the first byte of a sequence of bytes that
203 * represent a code point > 255? */
7b4252f4 204#define UTF8_IS_ABOVE_LATIN1(c) ((U8)(c) >= 0xc4)
8850bf83 205
15824458
KW
206/* This defines the 1-bits that are to be in the first byte of a multi-byte
207 * UTF-8 encoded character that give the number of bytes that comprise the
f710bf0c 208 * character. 'len' is the number of bytes in the multi-byte sequence. */
ee372ee9 209#define UTF_START_MARK(len) (((len) > 7) ? 0xFF : (0xFF & (0xFE << (7-(len)))))
b2635aa8
KW
210
211/* Masks out the initial one bits in a start byte, leaving the real data ones.
f710bf0c
KW
212 * Doesn't work on an invariant byte. 'len' is the number of bytes in the
213 * multi-byte sequence that comprises the character. */
22901f30 214#define UTF_START_MASK(len) (((len) >= 7) ? 0x00 : (0x1F >> ((len)-2)))
1d72bdf6 215
15824458
KW
216/* This defines the bits that are to be in the continuation bytes of a multi-byte
217 * UTF-8 encoded character that indicate it is a continuation byte. */
1d72bdf6 218#define UTF_CONTINUATION_MARK 0x80
15824458
KW
219
220/* This is the number of low-order bits a continuation byte in a UTF-8 encoded
221 * sequence contributes to the specification of the code point. In the bit
222 * maps above, you see that the first 2 bits are a constant '10', leaving 6 of
223 * real information */
1d72bdf6 224#define UTF_ACCUMULATION_SHIFT 6
b2635aa8
KW
225
226/* 2**UTF_ACCUMULATION_SHIFT - 1 */
1d72bdf6 227#define UTF_CONTINUATION_MASK ((U8)0x3f)
c512ce4f 228
15824458
KW
229/* If a value is anded with this, and the result is non-zero, then using the
230 * original value in UTF8_ACCUMULATE will overflow, shifting bits off the left
231 * */
eb83ed87 232#define UTF_ACCUMULATION_OVERFLOW_MASK \
bb88be5f
KW
233 (((UV) UTF_CONTINUATION_MASK) << ((sizeof(UV) * CHARBITS) \
234 - UTF_ACCUMULATION_SHIFT))
eb83ed87 235
6588300d
NC
236#if UVSIZE >= 8
237# define UTF8_QUAD_MAX UINT64_C(0x1000000000)
238
5aaebcb3
KW
239/* Input is a true Unicode (not-native) code point */
240#define OFFUNISKIP(uv) ( (uv) < 0x80 ? 1 : \
1d68d6cd
SC
241 (uv) < 0x800 ? 2 : \
242 (uv) < 0x10000 ? 3 : \
243 (uv) < 0x200000 ? 4 : \
244 (uv) < 0x4000000 ? 5 : \
245 (uv) < 0x80000000 ? 6 : \
9041c2e3 246 (uv) < UTF8_QUAD_MAX ? 7 : 13 )
1d68d6cd
SC
247#else
248/* No, I'm not even going to *TRY* putting #ifdef inside a #define */
5aaebcb3 249#define OFFUNISKIP(uv) ( (uv) < 0x80 ? 1 : \
1d68d6cd
SC
250 (uv) < 0x800 ? 2 : \
251 (uv) < 0x10000 ? 3 : \
252 (uv) < 0x200000 ? 4 : \
253 (uv) < 0x4000000 ? 5 : \
254 (uv) < 0x80000000 ? 6 : 7 )
255#endif
256
03c76984
KW
257/* How wide can a single UTF-8 encoded character become in bytes. */
258/* NOTE: Strictly speaking Perl's UTF-8 should not be called UTF-8 since UTF-8
259 * is an encoding of Unicode, and Unicode's upper limit, 0x10FFFF, can be
260 * expressed with 4 bytes. However, Perl thinks of UTF-8 as a way to encode
261 * non-negative integers in a binary format, even those above Unicode */
262#define UTF8_MAXBYTES 13
263
264/* The maximum number of UTF-8 bytes a single Unicode character can
265 * uppercase/lowercase/fold into. Unicode guarantees that the maximum
266 * expansion is 3 characters. On ASCIIish platforms, the highest Unicode
267 * character occupies 4 bytes, therefore this number would be 12, but this is
268 * smaller than the maximum width a single above-Unicode character can occupy,
269 * so use that instead */
270#if UTF8_MAXBYTES < 12
271#error UTF8_MAXBYTES must be at least 12
272#endif
273
843a4590
KW
274#define MAX_UTF8_TWO_BYTE 0x7FF
275
03c76984
KW
276#define UTF8_MAXBYTES_CASE UTF8_MAXBYTES
277
d06134e5
KW
278#endif /* EBCDIC vs ASCII */
279
280/* Rest of these are attributes of Unicode and perl's internals rather than the
281 * encoding, or happen to be the same in both ASCII and EBCDIC (at least at
282 * this level; the macros that some of these call may have different
283 * definitions in the two encodings */
284
59a449d5
KW
285/* In domain restricted to ASCII, these may make more sense to the reader than
286 * the ones with Latin1 in the name */
287#define NATIVE_TO_ASCII(ch) NATIVE_TO_LATIN1(ch)
288#define ASCII_TO_NATIVE(ch) LATIN1_TO_NATIVE(ch)
289
290/* More or less misleadingly-named defines, retained for back compat */
291#define NATIVE_TO_UTF(ch) NATIVE_UTF8_TO_I8(ch)
292#define NATIVE_TO_I8(ch) NATIVE_UTF8_TO_I8(ch)
293#define UTF_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch)
294#define I8_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch)
295#define NATIVE8_TO_UNI(ch) NATIVE_TO_LATIN1(ch)
d06134e5 296
537124e4
KW
297/* Adds a UTF8 continuation byte 'new' of information to a running total code
298 * point 'old' of all the continuation bytes so far. This is designed to be
155d2738
KW
299 * used in a loop to convert from UTF-8 to the code point represented. Note
300 * that this is asymmetric on EBCDIC platforms, in that the 'new' parameter is
301 * the UTF-EBCDIC byte, whereas the 'old' parameter is a Unicode (not EBCDIC)
302 * code point in process of being generated */
303#define UTF8_ACCUMULATE(old, new) (((old) << UTF_ACCUMULATION_SHIFT) \
304 | ((NATIVE_UTF8_TO_I8((U8)new)) \
305 & UTF_CONTINUATION_MASK))
d06134e5 306
4ab10950
KW
307/* This works in the face of malformed UTF-8. */
308#define UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(s, e) (UTF8_IS_DOWNGRADEABLE_START(*s) \
309 && ( (e) - (s) > 1) \
310 && UTF8_IS_CONTINUATION(*((s)+1)))
311
5aaebcb3
KW
312/* Number of bytes a code point occupies in UTF-8. */
313#define NATIVE_SKIP(uv) OFFUNISKIP(NATIVE_TO_UNI(uv))
314
315/* Most code which says UNISKIP is really thinking in terms of native code
316 * points (0-255) plus all those beyond. This is an imprecise term, but having
317 * it means existing code continues to work. For precision, use NATIVE_SKIP
318 * and OFFUNISKIP */
319#define UNISKIP(uv) NATIVE_SKIP(uv)
320
94bb8c36 321/* Convert a two (not one) byte utf8 character to a native code point value.
2950f2a7
KW
322 * Needs just one iteration of accumulate. Should not be used unless it is
323 * known that the two bytes are legal: 1) two-byte start, and 2) continuation.
324 * Note that the result can be larger than 255 if the input character is not
325 * downgradable */
94bb8c36
KW
326#define TWO_BYTE_UTF8_TO_NATIVE(HI, LO) \
327 UNI_TO_NATIVE(UTF8_ACCUMULATE((NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), \
155d2738 328 (LO)))
94bb8c36
KW
329
330/* Should never be used, and be deprecated */
331#define TWO_BYTE_UTF8_TO_UNI(HI, LO) NATIVE_TO_UNI(TWO_BYTE_UTF8_TO_NATIVE(HI, LO))
2950f2a7 332
15824458
KW
333/* How many bytes in the UTF-8 encoded character whose first (perhaps only)
334 * byte is pointed to by 's' */
d06134e5
KW
335#define UTF8SKIP(s) PL_utf8skip[*(const U8*)(s)]
336
15824458
KW
337/* Is the byte 'c' the same character when encoded in UTF-8 as when not. This
338 * works on both UTF-8 encoded strings and non-encoded, as it returns TRUE in
339 * each for the exact same set of bit patterns. (And it works on any byte in a
340 * UTF-8 encoded string) */
bc3632a8 341#define UTF8_IS_INVARIANT(c) UNI_IS_INVARIANT(NATIVE_UTF8_TO_I8(c))
15824458 342
6f2d5cbc
KW
343/* Like the above, but its name implies a non-UTF8 input, and is implemented
344 * differently (for no particular reason) */
345#define NATIVE_BYTE_IS_INVARIANT(c) UNI_IS_INVARIANT(NATIVE_TO_LATIN1(c))
346
347/* Like the above, but accepts any UV as input */
348#define UVCHR_IS_INVARIANT(uv) UNI_IS_INVARIANT(NATIVE_TO_UNI(uv))
d06134e5
KW
349
350#define MAX_PORTABLE_UTF8_TWO_BYTE 0x3FF /* constrained by EBCDIC */
351
48ccf5e1
KW
352/* The macros in the next 4 sets are used to generate the two utf8 or utfebcdic
353 * bytes from an ordinal that is known to fit into exactly two (not one) bytes;
354 * it must be less than 0x3FF to work across both encodings. */
355
356/* These two are helper macros for the other three sets, and should not be used
357 * directly anywhere else. 'translate_function' is either NATIVE_TO_LATIN1
358 * (which works for code points up to 0xFF) or NATIVE_TO_UNI which works for any
359 * code point */
360#define __BASE_TWO_BYTE_HI(c, translate_function) \
361 I8_TO_NATIVE_UTF8((translate_function(c) >> UTF_ACCUMULATION_SHIFT) \
362 | UTF_START_MARK(2))
363#define __BASE_TWO_BYTE_LO(c, translate_function) \
364 I8_TO_NATIVE_UTF8((translate_function(c) & UTF_CONTINUATION_MASK) \
365 | UTF_CONTINUATION_MARK)
366
48ccf5e1
KW
367/* The next two macros should not be used. They were designed to be usable as
368 * the case label of a switch statement, but this doesn't work for EBCDIC. Use
9d0d3a03 369 * regen/unicode_constants.pl instead */
48ccf5e1
KW
370#define UTF8_TWO_BYTE_HI_nocast(c) __BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI)
371#define UTF8_TWO_BYTE_LO_nocast(c) __BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI)
372
373/* The next two macros are used when the source should be a single byte
374 * character; checked for under DEBUGGING */
375#define UTF8_EIGHT_BIT_HI(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
376 ((U8) __BASE_TWO_BYTE_HI(c, NATIVE_TO_LATIN1)))
377#define UTF8_EIGHT_BIT_LO(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
378 ((U8) __BASE_TWO_BYTE_LO(c, NATIVE_TO_LATIN1)))
379
380/* These final two macros in the series are used when the source can be any
381 * code point whose UTF-8 is known to occupy 2 bytes; they are less efficient
382 * than the EIGHT_BIT versions on EBCDIC platforms. We use the logical '~'
383 * operator instead of "<=" to avoid getting compiler warnings.
384 * MAX_PORTABLE_UTF8_TWO_BYTE should be exactly all one bits in the lower few
385 * places, so the ~ works */
386#define UTF8_TWO_BYTE_HI(c) \
387 (__ASSERT_((sizeof(c) == 1) \
388 || !(((WIDEST_UTYPE)(c)) & ~MAX_PORTABLE_UTF8_TWO_BYTE)) \
389 ((U8) __BASE_TWO_BYTE_HI(c, NATIVE_TO_LATIN1)))
390#define UTF8_TWO_BYTE_LO(c) \
391 (__ASSERT_((sizeof(c) == 1) \
392 || !(((WIDEST_UTYPE)(c)) & ~MAX_PORTABLE_UTF8_TWO_BYTE)) \
393 ((U8) __BASE_TWO_BYTE_LO(c, NATIVE_TO_LATIN1)))
d06134e5 394
e7214ce8
KW
395/* This is illegal in any well-formed UTF-8 in both EBCDIC and ASCII
396 * as it is only in overlongs. */
397#define ILLEGAL_UTF8_BYTE I8_TO_NATIVE_UTF8(0xC1)
398
7e2040f0 399/*
e3036cf4 400 * 'UTF' is whether or not p is encoded in UTF8. The names 'foo_lazy_if' stem
20df05f4
KW
401 * from an earlier version of these macros in which they didn't call the
402 * foo_utf8() macros (i.e. were 'lazy') unless they decided that *p is the
403 * beginning of a utf8 character. Now that foo_utf8() determines that itself,
404 * no need to do it again here
7e2040f0 405 */
e3036cf4
KW
406#define isIDFIRST_lazy_if(p,UTF) ((IN_BYTES || !UTF ) \
407 ? isIDFIRST(*(p)) \
408 : isIDFIRST_utf8((const U8*)p))
32636478
KW
409#define isWORDCHAR_lazy_if(p,UTF) ((IN_BYTES || (!UTF )) \
410 ? isWORDCHAR(*(p)) \
411 : isWORDCHAR_utf8((const U8*)p))
412#define isALNUM_lazy_if(p,UTF) isWORDCHAR_lazy_if(p,UTF)
1d72bdf6 413
89ebb4a3
JH
414#define UTF8_MAXLEN UTF8_MAXBYTES
415
8cb75cc8
KW
416/* A Unicode character can fold to up to 3 characters */
417#define UTF8_MAX_FOLD_CHAR_EXPAND 3
418
a98fe34d 419#define IN_BYTES (CopHINTS_get(PL_curcop) & HINT_BYTES)
0064a8a9 420#define DO_UTF8(sv) (SvUTF8(sv) && !IN_BYTES)
66cbab2c 421#define IN_UNI_8_BIT \
b36bf33f
KW
422 (CopHINTS_get(PL_curcop) & (HINT_UNI_8_BIT|HINT_LOCALE_NOT_CHARS) \
423 && ! IN_LOCALE_RUNTIME && ! IN_BYTES)
424
1d72bdf6 425
c76687c5
KW
426#define UTF8_ALLOW_EMPTY 0x0001 /* Allow a zero length string */
427
428/* Allow first byte to be a continuation byte */
1d72bdf6 429#define UTF8_ALLOW_CONTINUATION 0x0002
c76687c5
KW
430
431/* Allow second... bytes to be non-continuation bytes */
1d72bdf6 432#define UTF8_ALLOW_NON_CONTINUATION 0x0004
949cf498
KW
433
434/* expecting more bytes than were available in the string */
435#define UTF8_ALLOW_SHORT 0x0008
436
437/* Overlong sequence; i.e., the code point can be specified in fewer bytes. */
438#define UTF8_ALLOW_LONG 0x0010
439
440#define UTF8_DISALLOW_SURROGATE 0x0020 /* Unicode surrogates */
441#define UTF8_WARN_SURROGATE 0x0040
442
443#define UTF8_DISALLOW_NONCHAR 0x0080 /* Unicode non-character */
444#define UTF8_WARN_NONCHAR 0x0100 /* code points */
445
446#define UTF8_DISALLOW_SUPER 0x0200 /* Super-set of Unicode: code */
447#define UTF8_WARN_SUPER 0x0400 /* points above the legal max */
448
449/* Code points which never were part of the original UTF-8 standard, the first
ea5ced44
KW
450 * byte of which is a FE or FF on ASCII platforms. If the first byte is FF, it
451 * will overflow a 32-bit word. If the first byte is FE, it will overflow a
452 * signed 32-bit word. */
949cf498
KW
453#define UTF8_DISALLOW_FE_FF 0x0800
454#define UTF8_WARN_FE_FF 0x1000
455
456#define UTF8_CHECK_ONLY 0x2000
457
458/* For backwards source compatibility. They do nothing, as the default now
459 * includes what they used to mean. The first one's meaning was to allow the
460 * just the single non-character 0xFFFF */
461#define UTF8_ALLOW_FFFF 0
462#define UTF8_ALLOW_SURROGATE 0
463
33d9abfb 464#define UTF8_DISALLOW_ILLEGAL_INTERCHANGE (UTF8_DISALLOW_SUPER|UTF8_DISALLOW_NONCHAR|UTF8_DISALLOW_SURROGATE|UTF8_DISALLOW_FE_FF)
949cf498 465#define UTF8_WARN_ILLEGAL_INTERCHANGE \
33d9abfb 466 (UTF8_WARN_SUPER|UTF8_WARN_NONCHAR|UTF8_WARN_SURROGATE|UTF8_WARN_FE_FF)
949cf498
KW
467#define UTF8_ALLOW_ANY \
468 (~(UTF8_DISALLOW_ILLEGAL_INTERCHANGE|UTF8_WARN_ILLEGAL_INTERCHANGE))
469#define UTF8_ALLOW_ANYUV \
470 (UTF8_ALLOW_EMPTY \
471 & ~(UTF8_DISALLOW_ILLEGAL_INTERCHANGE|UTF8_WARN_ILLEGAL_INTERCHANGE))
9f7f3913
ST
472#define UTF8_ALLOW_DEFAULT (ckWARN(WARN_UTF8) ? 0 : \
473 UTF8_ALLOW_ANYUV)
1d72bdf6 474
7131f24d
KW
475/* Surrogates, non-character code points and above-Unicode code points are
476 * problematic in some contexts. This allows code that needs to check for
477 * those to to quickly exclude the vast majority of code points it will
478 * encounter */
b96a92fb
KW
479#define UTF8_FIRST_PROBLEMATIC_CODE_POINT_FIRST_BYTE \
480 FIRST_SURROGATE_UTF8_FIRST_BYTE
7131f24d 481
b96a92fb
KW
482#define UTF8_IS_SURROGATE(s) cBOOL(is_SURROGATE_utf8(s))
483#define UTF8_IS_REPLACEMENT(s, send) cBOOL(is_REPLACEMENT_utf8_safe(s,send))
7131f24d
KW
484
485/* ASCII EBCDIC I8
486 * U+10FFFF: \xF4\x8F\xBF\xBF \xF9\xA1\xBF\xBF\xBF max legal Unicode
487 * U+110000: \xF4\x90\x80\x80 \xF9\xA2\xA0\xA0\xA0
488 * U+110001: \xF4\x90\x80\x81 \xF9\xA2\xA0\xA0\xA1
489 */
490#ifdef EBCDIC /* Both versions assume well-formed UTF8 */
bc3632a8
KW
491# define UTF8_IS_SUPER(s) (NATIVE_UTF8_TO_I8(* (U8*) (s)) >= 0xF9 \
492 && (NATIVE_UTF8_TO_I8(* (U8*) (s)) > 0xF9 \
a3481822 493 || (NATIVE_UTF8_TO_I8(* ((U8*) (s) + 1)) >= 0xA2)))
7131f24d 494#else
0404c92a
KW
495# define UTF8_IS_SUPER(s) (*(U8*) (s) >= 0xF4 \
496 && (*(U8*) (s) > 0xF4 || (*((U8*) (s) + 1) >= 0x90)))
7131f24d
KW
497#endif
498
b96a92fb
KW
499/* These are now machine generated, and the 'given' clause is no longer
500 * applicable */
501#define UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s) \
502 cBOOL(is_NONCHAR_utf8(s))
503#define UTF8_IS_NONCHAR_(s) \
504 UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s)
7131f24d 505
c867b360
JH
506#define UNICODE_SURROGATE_FIRST 0xD800
507#define UNICODE_SURROGATE_LAST 0xDFFF
508#define UNICODE_REPLACEMENT 0xFFFD
509#define UNICODE_BYTE_ORDER_MARK 0xFEFF
1d72bdf6 510
b851fbc1 511/* Though our UTF-8 encoding can go beyond this,
c76687c5 512 * let's be conservative and do as Unicode says. */
b851fbc1
JH
513#define PERL_UNICODE_MAX 0x10FFFF
514
949cf498
KW
515#define UNICODE_WARN_SURROGATE 0x0001 /* UTF-16 surrogates */
516#define UNICODE_WARN_NONCHAR 0x0002 /* Non-char code points */
517#define UNICODE_WARN_SUPER 0x0004 /* Above 0x10FFFF */
518#define UNICODE_WARN_FE_FF 0x0008 /* Above 0x10FFFF */
519#define UNICODE_DISALLOW_SURROGATE 0x0010
520#define UNICODE_DISALLOW_NONCHAR 0x0020
521#define UNICODE_DISALLOW_SUPER 0x0040
522#define UNICODE_DISALLOW_FE_FF 0x0080
bb88be5f
KW
523#define UNICODE_WARN_ILLEGAL_INTERCHANGE \
524 (UNICODE_WARN_SURROGATE|UNICODE_WARN_NONCHAR|UNICODE_WARN_SUPER)
525#define UNICODE_DISALLOW_ILLEGAL_INTERCHANGE \
526 (UNICODE_DISALLOW_SURROGATE|UNICODE_DISALLOW_NONCHAR|UNICODE_DISALLOW_SUPER)
949cf498
KW
527
528/* For backward source compatibility, as are now the default */
529#define UNICODE_ALLOW_SURROGATE 0
530#define UNICODE_ALLOW_SUPER 0
531#define UNICODE_ALLOW_ANY 0
b851fbc1 532
1d72bdf6
NIS
533#define UNICODE_IS_SURROGATE(c) ((c) >= UNICODE_SURROGATE_FIRST && \
534 (c) <= UNICODE_SURROGATE_LAST)
a10ec373 535#define UNICODE_IS_REPLACEMENT(c) ((c) == UNICODE_REPLACEMENT)
872c91ae 536#define UNICODE_IS_BYTE_ORDER_MARK(c) ((c) == UNICODE_BYTE_ORDER_MARK)
7131f24d
KW
537#define UNICODE_IS_NONCHAR(c) ((c >= 0xFDD0 && c <= 0xFDEF) \
538 /* The other noncharacters end in FFFE or FFFF, which \
539 * the mask below catches both of, but beyond the last \
540 * official unicode code point, they aren't \
541 * noncharacters, since those aren't Unicode \
542 * characters at all */ \
543 || ((((c & 0xFFFE) == 0xFFFE)) && ! UNICODE_IS_SUPER(c)))
544#define UNICODE_IS_SUPER(c) ((c) > PERL_UNICODE_MAX)
545#define UNICODE_IS_FE_FF(c) ((c) > 0x7FFFFFFF)
1d72bdf6 546
ec34087a
KW
547#define LATIN_SMALL_LETTER_SHARP_S LATIN_SMALL_LETTER_SHARP_S_NATIVE
548#define LATIN_SMALL_LETTER_Y_WITH_DIAERESIS \
549 LATIN_SMALL_LETTER_Y_WITH_DIAERESIS_NATIVE
550#define MICRO_SIGN MICRO_SIGN_NATIVE
551#define LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE \
552 LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE_NATIVE
553#define LATIN_SMALL_LETTER_A_WITH_RING_ABOVE \
554 LATIN_SMALL_LETTER_A_WITH_RING_ABOVE_NATIVE
09091399
JH
555#define UNICODE_GREEK_CAPITAL_LETTER_SIGMA 0x03A3
556#define UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA 0x03C2
557#define UNICODE_GREEK_SMALL_LETTER_SIGMA 0x03C3
9dcbe121 558#define GREEK_SMALL_LETTER_MU 0x03BC
9e682c18
KW
559#define GREEK_CAPITAL_LETTER_MU 0x039C /* Upper and title case
560 of MICRON */
561#define LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS 0x0178 /* Also is title case */
562#define LATIN_CAPITAL_LETTER_SHARP_S 0x1E9E
563#define LATIN_SMALL_LETTER_LONG_S 0x017F
a9f50d33
KW
564#define LATIN_SMALL_LIGATURE_LONG_S_T 0xFB05
565#define LATIN_SMALL_LIGATURE_ST 0xFB06
9e682c18
KW
566#define KELVIN_SIGN 0x212A
567#define ANGSTROM_SIGN 0x212B
09091399 568
9e55ce06 569#define UNI_DISPLAY_ISPRINT 0x0001
c728cb41
JH
570#define UNI_DISPLAY_BACKSLASH 0x0002
571#define UNI_DISPLAY_QQ (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH)
572#define UNI_DISPLAY_REGEX (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH)
9e55ce06 573
5cd46e1f
KW
574#define ANYOF_FOLD_SHARP_S(node, input, end) \
575 (ANYOF_BITMAP_TEST(node, LATIN_SMALL_LETTER_SHARP_S) && \
137165a6 576 (ANYOF_NONBITMAP(node)) && \
39065660 577 (ANYOF_FLAGS(node) & ANYOF_LOC_NONBITMAP_FOLD) && \
07b6858f 578 ((end) > (input) + 1) && \
d22b930b
KW
579 toFOLD((input)[0]) == 's' && \
580 toFOLD((input)[1]) == 's')
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
KW
585
586#ifndef EBCDIC
587/* This was generated by regen/regcharclass.pl, and then moved here. The lines
588 * that generated it were then commented out. This was done solely because it
589 * takes on the order of 10 minutes to generate, and is never going to change.
590 * The EBCDIC equivalent hasn't been commented out in regcharclass.pl, so it
591 * should generate and run the correct stuff */
39a0f513
YO
592/*
593 UTF8_CHAR: Matches utf8 from 1 to 4 bytes
594
595 0x0 - 0x1FFFFF
596*/
4d646140
KW
597/*** GENERATED CODE ***/
598#define is_UTF8_CHAR_utf8_safe(s,e) \
599( ((e)-(s) > 3) ? \
600 ( ( ( ((U8*)s)[0] & 0x80 ) == 0x00 ) ? 1 \
601 : ( 0xC2 <= ((U8*)s)[0] && ((U8*)s)[0] <= 0xDF ) ? \
602 ( ( ( ((U8*)s)[1] & 0xC0 ) == 0x80 ) ? 2 : 0 ) \
603 : ( 0xE0 == ((U8*)s)[0] ) ? \
604 ( ( ( ( ((U8*)s)[1] & 0xE0 ) == 0xA0 ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
605 : ( 0xE1 <= ((U8*)s)[0] && ((U8*)s)[0] <= 0xEF ) ? \
606 ( ( ( ( ((U8*)s)[1] & 0xC0 ) == 0x80 ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
607 : ( 0xF0 == ((U8*)s)[0] ) ? \
608 ( ( ( ( 0x90 <= ((U8*)s)[1] && ((U8*)s)[1] <= 0xBF ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) && ( ( ((U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\
39a0f513 609 : ( ( ( ( 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
610: ((e)-(s) > 2) ? \
611 ( ( ( ((U8*)s)[0] & 0x80 ) == 0x00 ) ? 1 \
612 : ( 0xC2 <= ((U8*)s)[0] && ((U8*)s)[0] <= 0xDF ) ? \
613 ( ( ( ((U8*)s)[1] & 0xC0 ) == 0x80 ) ? 2 : 0 ) \
614 : ( 0xE0 == ((U8*)s)[0] ) ? \
615 ( ( ( ( ((U8*)s)[1] & 0xE0 ) == 0xA0 ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
39a0f513 616 : ( ( ( 0xE1 <= ((U8*)s)[0] && ((U8*)s)[0] <= 0xEF ) && ( ( ((U8*)s)[1] & 0xC0 ) == 0x80 ) ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
4d646140
KW
617: ((e)-(s) > 1) ? \
618 ( ( ( ((U8*)s)[0] & 0x80 ) == 0x00 ) ? 1 \
39a0f513 619 : ( ( 0xC2 <= ((U8*)s)[0] && ((U8*)s)[0] <= 0xDF ) && ( ( ((U8*)s)[1] & 0xC0 ) == 0x80 ) ) ? 2 : 0 )\
4d646140
KW
620: ((e)-(s) > 0) ? \
621 ( ( ((U8*)s)[0] & 0x80 ) == 0x00 ) \
622: 0 )
623#endif
3b0fc154
JH
624
625/* IS_UTF8_CHAR(p) is strictly speaking wrong (not UTF-8) because it
626 * (1) allows UTF-8 encoded UTF-16 surrogates
627 * (2) it allows code points past U+10FFFF.
628 * The Perl_is_utf8_char() full "slow" code will handle the Perl
629 * "extended UTF-8". */
4d646140 630#define IS_UTF8_CHAR(p, n) (is_UTF8_CHAR_utf8_safe(p, (p) + (n)) == n)
2f454f11 631
4d646140
KW
632/* regen/regcharclass.pl generates is_UTF8_CHAR_utf8_safe() macros for up to
633 * these number of bytes. So this has to be coordinated with it */
634#ifdef EBCDIC
2f454f11 635# define IS_UTF8_CHAR_FAST(n) ((n) <= 5)
4d646140
KW
636#else
637# define IS_UTF8_CHAR_FAST(n) ((n) <= 4)
638#endif
e9a8c099 639
57f0e7e2
KW
640#endif /* H_UTF8 */
641
e9a8c099
MHM
642/*
643 * Local variables:
644 * c-indentation-style: bsd
645 * c-basic-offset: 4
14d04a33 646 * indent-tabs-mode: nil
e9a8c099
MHM
647 * End:
648 *
14d04a33 649 * ex: set ts=8 sts=4 sw=4 et:
e9a8c099 650 */