This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Improve API pod of is_ascii_string
[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#define ibcmp_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \
66 cBOOL(! foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2))
67
1d72bdf6
NIS
68#ifdef EBCDIC
69/* The equivalent of these macros but implementing UTF-EBCDIC
70 are in the following header file:
71 */
72
73#include "utfebcdic.h"
fd7cb289 74
d06134e5 75#else /* ! EBCDIC */
73c4f7a1
GS
76START_EXTERN_C
77
a0ed51b3 78#ifdef DOINIT
6f06b55f 79EXTCONST unsigned char PL_utf8skip[] = {
b2635aa8
KW
80/* 0x00 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
81/* 0x10 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
82/* 0x20 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
83/* 0x30 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
84/* 0x40 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
85/* 0x50 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
86/* 0x60 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
87/* 0x70 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
88/* 0x80 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
89/* 0x90 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
90/* 0xA0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
91/* 0xB0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
92/* 0xC0 */ 2,2, /* overlong */
93/* 0xC2 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0080 to U+03FF */
94/* 0xD0 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0400 to U+07FF */
95/* 0xE0 */ 3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3, /* U+0800 to U+FFFF */
96/* 0xF0 */ 4,4,4,4,4,4,4,4,5,5,5,5,6,6, /* above BMP to 2**31 - 1 */
97/* 0xFE */ 7,13, /* Perl extended (never was official UTF-8). Up to 72bit
98 allowed (64-bit + reserved). */
a0ed51b3
LW
99};
100#else
6f06b55f 101EXTCONST unsigned char PL_utf8skip[];
a0ed51b3
LW
102#endif
103
73c4f7a1 104END_EXTERN_C
7e2040f0 105
59a449d5
KW
106/* Native character to/from iso-8859-1. Are the identity functions on ASCII
107 * platforms */
108#define NATIVE_TO_LATIN1(ch) (ch)
109#define LATIN1_TO_NATIVE(ch) (ch)
110
111/* I8 is an intermediate version of UTF-8 used only in UTF-EBCDIC. We thus
112 * consider it to be identical to UTF-8 on ASCII platforms. Strictly speaking
113 * UTF-8 and UTF-EBCDIC are two different things, but we often conflate them
114 * because they are 8-bit encodings that serve the same purpose in Perl, and
115 * rarely do we need to distinguish them. The term "NATIVE_UTF8" applies to
116 * whichever one is applicable on the current platform */
117#define NATIVE_UTF8_TO_I8(ch) (ch)
118#define I8_TO_NATIVE_UTF8(ch) (ch)
119
1d72bdf6
NIS
120/* Transforms in wide UV chars */
121#define UNI_TO_NATIVE(ch) (ch)
122#define NATIVE_TO_UNI(ch) (ch)
d7578b48 123
877d9f0d 124/*
9041c2e3 125
8c007b5a 126 The following table is from Unicode 3.2.
877d9f0d
JH
127
128 Code Points 1st Byte 2nd Byte 3rd Byte 4th Byte
129
375122d7 130 U+0000..U+007F 00..7F
e1b711da 131 U+0080..U+07FF * C2..DF 80..BF
37e2e78e 132 U+0800..U+0FFF E0 * A0..BF 80..BF
375122d7 133 U+1000..U+CFFF E1..EC 80..BF 80..BF
e1b711da 134 U+D000..U+D7FF ED 80..9F 80..BF
537124e4 135 U+D800..U+DFFF ED A0..BF 80..BF (surrogates)
375122d7 136 U+E000..U+FFFF EE..EF 80..BF 80..BF
37e2e78e 137 U+10000..U+3FFFF F0 * 90..BF 80..BF 80..BF
877d9f0d
JH
138 U+40000..U+FFFFF F1..F3 80..BF 80..BF 80..BF
139 U+100000..U+10FFFF F4 80..8F 80..BF 80..BF
b2635aa8
KW
140 Below are non-Unicode code points
141 U+110000..U+13FFFF F4 90..BF 80..BF 80..BF
142 U+110000..U+1FFFFF F5..F7 80..BF 80..BF 80..BF
537124e4 143 U+200000..: F8.. * 88..BF 80..BF 80..BF 80..BF
877d9f0d 144
e1b711da 145Note the gaps before several of the byte entries above marked by '*'. These are
37e2e78e
KW
146caused by legal UTF-8 avoiding non-shortest encodings: it is technically
147possible to UTF-8-encode a single code point in different ways, but that is
148explicitly forbidden, and the shortest possible encoding should always be used
15824458 149(and that is what Perl does). The non-shortest ones are called 'overlongs'.
8c007b5a 150
877d9f0d
JH
151 */
152
8c007b5a
JH
153/*
154 Another way to look at it, as bits:
155
b2635aa8 156 Code Points 1st Byte 2nd Byte 3rd Byte 4th Byte
8c007b5a 157
b2635aa8
KW
158 0aaa aaaa 0aaa aaaa
159 0000 0bbb bbaa aaaa 110b bbbb 10aa aaaa
160 cccc bbbb bbaa aaaa 1110 cccc 10bb bbbb 10aa aaaa
161 00 000d ddcc cccc bbbb bbaa aaaa 1111 0ddd 10cc cccc 10bb bbbb 10aa aaaa
8c007b5a
JH
162
163As you can see, the continuation bytes all begin with C<10>, and the
e1b711da 164leading bits of the start byte tell how many bytes there are in the
8c007b5a
JH
165encoded character.
166
65ab9279
TC
167Perl's extended UTF-8 means we can have start bytes up to FF.
168
8c007b5a
JH
169*/
170
15824458
KW
171/* Is the representation of the Unicode code point 'c' the same regardless of
172 * being encoded in UTF-8 or not? */
c4d5f83a 173#define UNI_IS_INVARIANT(c) (((UV)c) < 0x80)
15824458
KW
174
175/* Is the UTF8-encoded byte 'c' part of a variant sequence in UTF-8? This is
176 * the inverse of UTF8_IS_INVARIANT */
177#define UTF8_IS_CONTINUED(c) (((U8)c) & 0x80)
178
179/* Is the byte 'c' the first byte of a multi-byte UTF8-8 encoded sequence?
180 * This doesn't catch invariants (they are single-byte). It also excludes the
181 * illegal overlong sequences that begin with C0 and C1. */
0447e8df 182#define UTF8_IS_START(c) (((U8)c) >= 0xc2)
15824458
KW
183
184/* Is the byte 'c' part of a multi-byte UTF8-8 encoded sequence, and not the
185 * first byte thereof? */
e021c6e6 186#define UTF8_IS_CONTINUATION(c) ((((U8)c) & 0xC0) == 0x80)
0ae1fa71 187
15824458
KW
188/* Is the UTF8-encoded byte 'c' the first byte of a two byte sequence? Use
189 * UTF8_IS_NEXT_CHAR_DOWNGRADEABLE() instead if the input isn't known to
190 * be well-formed. Masking with 0xfe allows the low bit to be 0 or 1; thus
191 * this matches 0xc[23]. */
559c7f10 192#define UTF8_IS_DOWNGRADEABLE_START(c) (((U8)(c) & 0xfe) == 0xc2)
4ab10950 193
15824458
KW
194/* Is the UTF8-encoded byte 'c' the first byte of a sequence of bytes that
195 * represent a code point > 255? */
7b4252f4 196#define UTF8_IS_ABOVE_LATIN1(c) ((U8)(c) >= 0xc4)
8850bf83 197
15824458
KW
198/* This defines the 1-bits that are to be in the first byte of a multi-byte
199 * UTF-8 encoded character that give the number of bytes that comprise the
f710bf0c 200 * character. 'len' is the number of bytes in the multi-byte sequence. */
ee372ee9 201#define UTF_START_MARK(len) (((len) > 7) ? 0xFF : (0xFF & (0xFE << (7-(len)))))
b2635aa8
KW
202
203/* Masks out the initial one bits in a start byte, leaving the real data ones.
f710bf0c
KW
204 * Doesn't work on an invariant byte. 'len' is the number of bytes in the
205 * multi-byte sequence that comprises the character. */
22901f30 206#define UTF_START_MASK(len) (((len) >= 7) ? 0x00 : (0x1F >> ((len)-2)))
1d72bdf6 207
15824458
KW
208/* This defines the bits that are to be in the continuation bytes of a multi-byte
209 * UTF-8 encoded character that indicate it is a continuation byte. */
1d72bdf6 210#define UTF_CONTINUATION_MARK 0x80
15824458
KW
211
212/* This is the number of low-order bits a continuation byte in a UTF-8 encoded
213 * sequence contributes to the specification of the code point. In the bit
214 * maps above, you see that the first 2 bits are a constant '10', leaving 6 of
215 * real information */
1d72bdf6 216#define UTF_ACCUMULATION_SHIFT 6
b2635aa8
KW
217
218/* 2**UTF_ACCUMULATION_SHIFT - 1 */
1d72bdf6 219#define UTF_CONTINUATION_MASK ((U8)0x3f)
c512ce4f 220
15824458
KW
221/* If a value is anded with this, and the result is non-zero, then using the
222 * original value in UTF8_ACCUMULATE will overflow, shifting bits off the left
223 * */
eb83ed87 224#define UTF_ACCUMULATION_OVERFLOW_MASK \
bb88be5f
KW
225 (((UV) UTF_CONTINUATION_MASK) << ((sizeof(UV) * CHARBITS) \
226 - UTF_ACCUMULATION_SHIFT))
eb83ed87 227
6588300d
NC
228#if UVSIZE >= 8
229# define UTF8_QUAD_MAX UINT64_C(0x1000000000)
230
5aaebcb3
KW
231/* Input is a true Unicode (not-native) code point */
232#define OFFUNISKIP(uv) ( (uv) < 0x80 ? 1 : \
1d68d6cd
SC
233 (uv) < 0x800 ? 2 : \
234 (uv) < 0x10000 ? 3 : \
235 (uv) < 0x200000 ? 4 : \
236 (uv) < 0x4000000 ? 5 : \
237 (uv) < 0x80000000 ? 6 : \
9041c2e3 238 (uv) < UTF8_QUAD_MAX ? 7 : 13 )
1d68d6cd
SC
239#else
240/* No, I'm not even going to *TRY* putting #ifdef inside a #define */
5aaebcb3 241#define OFFUNISKIP(uv) ( (uv) < 0x80 ? 1 : \
1d68d6cd
SC
242 (uv) < 0x800 ? 2 : \
243 (uv) < 0x10000 ? 3 : \
244 (uv) < 0x200000 ? 4 : \
245 (uv) < 0x4000000 ? 5 : \
246 (uv) < 0x80000000 ? 6 : 7 )
247#endif
248
03c76984
KW
249/* How wide can a single UTF-8 encoded character become in bytes. */
250/* NOTE: Strictly speaking Perl's UTF-8 should not be called UTF-8 since UTF-8
251 * is an encoding of Unicode, and Unicode's upper limit, 0x10FFFF, can be
252 * expressed with 4 bytes. However, Perl thinks of UTF-8 as a way to encode
253 * non-negative integers in a binary format, even those above Unicode */
254#define UTF8_MAXBYTES 13
255
256/* The maximum number of UTF-8 bytes a single Unicode character can
257 * uppercase/lowercase/fold into. Unicode guarantees that the maximum
258 * expansion is 3 characters. On ASCIIish platforms, the highest Unicode
259 * character occupies 4 bytes, therefore this number would be 12, but this is
260 * smaller than the maximum width a single above-Unicode character can occupy,
261 * so use that instead */
262#if UTF8_MAXBYTES < 12
263#error UTF8_MAXBYTES must be at least 12
264#endif
265
0ed2b00b
KW
266/* ^? is defined to be DEL on ASCII systems. See the definition of toCTRL()
267 * for more */
268#define QUESTION_MARK_CTRL DEL_NATIVE
269
843a4590
KW
270#define MAX_UTF8_TWO_BYTE 0x7FF
271
03c76984
KW
272#define UTF8_MAXBYTES_CASE UTF8_MAXBYTES
273
d06134e5
KW
274#endif /* EBCDIC vs ASCII */
275
276/* Rest of these are attributes of Unicode and perl's internals rather than the
277 * encoding, or happen to be the same in both ASCII and EBCDIC (at least at
278 * this level; the macros that some of these call may have different
279 * definitions in the two encodings */
280
59a449d5
KW
281/* In domain restricted to ASCII, these may make more sense to the reader than
282 * the ones with Latin1 in the name */
283#define NATIVE_TO_ASCII(ch) NATIVE_TO_LATIN1(ch)
284#define ASCII_TO_NATIVE(ch) LATIN1_TO_NATIVE(ch)
285
286/* More or less misleadingly-named defines, retained for back compat */
287#define NATIVE_TO_UTF(ch) NATIVE_UTF8_TO_I8(ch)
288#define NATIVE_TO_I8(ch) NATIVE_UTF8_TO_I8(ch)
289#define UTF_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch)
290#define I8_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch)
291#define NATIVE8_TO_UNI(ch) NATIVE_TO_LATIN1(ch)
d06134e5 292
537124e4
KW
293/* Adds a UTF8 continuation byte 'new' of information to a running total code
294 * point 'old' of all the continuation bytes so far. This is designed to be
155d2738
KW
295 * used in a loop to convert from UTF-8 to the code point represented. Note
296 * that this is asymmetric on EBCDIC platforms, in that the 'new' parameter is
297 * the UTF-EBCDIC byte, whereas the 'old' parameter is a Unicode (not EBCDIC)
298 * code point in process of being generated */
299#define UTF8_ACCUMULATE(old, new) (((old) << UTF_ACCUMULATION_SHIFT) \
300 | ((NATIVE_UTF8_TO_I8((U8)new)) \
301 & UTF_CONTINUATION_MASK))
d06134e5 302
4ab10950
KW
303/* This works in the face of malformed UTF-8. */
304#define UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(s, e) (UTF8_IS_DOWNGRADEABLE_START(*s) \
305 && ( (e) - (s) > 1) \
306 && UTF8_IS_CONTINUATION(*((s)+1)))
307
5aaebcb3
KW
308/* Number of bytes a code point occupies in UTF-8. */
309#define NATIVE_SKIP(uv) OFFUNISKIP(NATIVE_TO_UNI(uv))
310
311/* Most code which says UNISKIP is really thinking in terms of native code
312 * points (0-255) plus all those beyond. This is an imprecise term, but having
313 * it means existing code continues to work. For precision, use NATIVE_SKIP
314 * and OFFUNISKIP */
315#define UNISKIP(uv) NATIVE_SKIP(uv)
316
94bb8c36 317/* Convert a two (not one) byte utf8 character to a native code point value.
2950f2a7
KW
318 * Needs just one iteration of accumulate. Should not be used unless it is
319 * known that the two bytes are legal: 1) two-byte start, and 2) continuation.
320 * Note that the result can be larger than 255 if the input character is not
321 * downgradable */
94bb8c36
KW
322#define TWO_BYTE_UTF8_TO_NATIVE(HI, LO) \
323 UNI_TO_NATIVE(UTF8_ACCUMULATE((NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), \
155d2738 324 (LO)))
94bb8c36
KW
325
326/* Should never be used, and be deprecated */
327#define TWO_BYTE_UTF8_TO_UNI(HI, LO) NATIVE_TO_UNI(TWO_BYTE_UTF8_TO_NATIVE(HI, LO))
2950f2a7 328
15824458
KW
329/* How many bytes in the UTF-8 encoded character whose first (perhaps only)
330 * byte is pointed to by 's' */
d06134e5
KW
331#define UTF8SKIP(s) PL_utf8skip[*(const U8*)(s)]
332
15824458
KW
333/* Is the byte 'c' the same character when encoded in UTF-8 as when not. This
334 * works on both UTF-8 encoded strings and non-encoded, as it returns TRUE in
335 * each for the exact same set of bit patterns. (And it works on any byte in a
336 * UTF-8 encoded string) */
bc3632a8 337#define UTF8_IS_INVARIANT(c) UNI_IS_INVARIANT(NATIVE_UTF8_TO_I8(c))
15824458 338
6f2d5cbc
KW
339/* Like the above, but its name implies a non-UTF8 input, and is implemented
340 * differently (for no particular reason) */
341#define NATIVE_BYTE_IS_INVARIANT(c) UNI_IS_INVARIANT(NATIVE_TO_LATIN1(c))
342
343/* Like the above, but accepts any UV as input */
344#define UVCHR_IS_INVARIANT(uv) UNI_IS_INVARIANT(NATIVE_TO_UNI(uv))
d06134e5
KW
345
346#define MAX_PORTABLE_UTF8_TWO_BYTE 0x3FF /* constrained by EBCDIC */
347
48ccf5e1
KW
348/* The macros in the next 4 sets are used to generate the two utf8 or utfebcdic
349 * bytes from an ordinal that is known to fit into exactly two (not one) bytes;
350 * it must be less than 0x3FF to work across both encodings. */
351
352/* These two are helper macros for the other three sets, and should not be used
353 * directly anywhere else. 'translate_function' is either NATIVE_TO_LATIN1
354 * (which works for code points up to 0xFF) or NATIVE_TO_UNI which works for any
355 * code point */
356#define __BASE_TWO_BYTE_HI(c, translate_function) \
357 I8_TO_NATIVE_UTF8((translate_function(c) >> UTF_ACCUMULATION_SHIFT) \
358 | UTF_START_MARK(2))
359#define __BASE_TWO_BYTE_LO(c, translate_function) \
360 I8_TO_NATIVE_UTF8((translate_function(c) & UTF_CONTINUATION_MASK) \
361 | UTF_CONTINUATION_MARK)
362
48ccf5e1
KW
363/* The next two macros should not be used. They were designed to be usable as
364 * the case label of a switch statement, but this doesn't work for EBCDIC. Use
9d0d3a03 365 * regen/unicode_constants.pl instead */
48ccf5e1
KW
366#define UTF8_TWO_BYTE_HI_nocast(c) __BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI)
367#define UTF8_TWO_BYTE_LO_nocast(c) __BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI)
368
369/* The next two macros are used when the source should be a single byte
370 * character; checked for under DEBUGGING */
371#define UTF8_EIGHT_BIT_HI(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
372 ((U8) __BASE_TWO_BYTE_HI(c, NATIVE_TO_LATIN1)))
373#define UTF8_EIGHT_BIT_LO(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
374 ((U8) __BASE_TWO_BYTE_LO(c, NATIVE_TO_LATIN1)))
375
376/* These final two macros in the series are used when the source can be any
377 * code point whose UTF-8 is known to occupy 2 bytes; they are less efficient
378 * than the EIGHT_BIT versions on EBCDIC platforms. We use the logical '~'
379 * operator instead of "<=" to avoid getting compiler warnings.
380 * MAX_PORTABLE_UTF8_TWO_BYTE should be exactly all one bits in the lower few
381 * places, so the ~ works */
382#define UTF8_TWO_BYTE_HI(c) \
383 (__ASSERT_((sizeof(c) == 1) \
384 || !(((WIDEST_UTYPE)(c)) & ~MAX_PORTABLE_UTF8_TWO_BYTE)) \
d9759938 385 ((U8) __BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI)))
48ccf5e1
KW
386#define UTF8_TWO_BYTE_LO(c) \
387 (__ASSERT_((sizeof(c) == 1) \
388 || !(((WIDEST_UTYPE)(c)) & ~MAX_PORTABLE_UTF8_TWO_BYTE)) \
d9759938 389 ((U8) __BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI)))
d06134e5 390
e7214ce8
KW
391/* This is illegal in any well-formed UTF-8 in both EBCDIC and ASCII
392 * as it is only in overlongs. */
393#define ILLEGAL_UTF8_BYTE I8_TO_NATIVE_UTF8(0xC1)
394
7e2040f0 395/*
e3036cf4 396 * 'UTF' is whether or not p is encoded in UTF8. The names 'foo_lazy_if' stem
20df05f4
KW
397 * from an earlier version of these macros in which they didn't call the
398 * foo_utf8() macros (i.e. were 'lazy') unless they decided that *p is the
399 * beginning of a utf8 character. Now that foo_utf8() determines that itself,
400 * no need to do it again here
7e2040f0 401 */
e3036cf4
KW
402#define isIDFIRST_lazy_if(p,UTF) ((IN_BYTES || !UTF ) \
403 ? isIDFIRST(*(p)) \
404 : isIDFIRST_utf8((const U8*)p))
32636478
KW
405#define isWORDCHAR_lazy_if(p,UTF) ((IN_BYTES || (!UTF )) \
406 ? isWORDCHAR(*(p)) \
407 : isWORDCHAR_utf8((const U8*)p))
408#define isALNUM_lazy_if(p,UTF) isWORDCHAR_lazy_if(p,UTF)
1d72bdf6 409
89ebb4a3
JH
410#define UTF8_MAXLEN UTF8_MAXBYTES
411
8cb75cc8
KW
412/* A Unicode character can fold to up to 3 characters */
413#define UTF8_MAX_FOLD_CHAR_EXPAND 3
414
a98fe34d 415#define IN_BYTES (CopHINTS_get(PL_curcop) & HINT_BYTES)
0064a8a9 416#define DO_UTF8(sv) (SvUTF8(sv) && !IN_BYTES)
66cbab2c 417#define IN_UNI_8_BIT \
d6ded950
KW
418 (((CopHINTS_get(PL_curcop) & (HINT_UNI_8_BIT)) \
419 || (CopHINTS_get(PL_curcop) & HINT_LOCALE_PARTIAL \
420 /* -1 below is for :not_characters */ \
421 && _is_in_locale_category(FALSE, -1))) \
422 && ! IN_BYTES)
b36bf33f 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
a1776718
KW
488 *
489 * BE AWARE that this test doesn't rule out malformed code points, in
490 * particular overlongs */
7131f24d 491#ifdef EBCDIC /* Both versions assume well-formed UTF8 */
bc3632a8
KW
492# define UTF8_IS_SUPER(s) (NATIVE_UTF8_TO_I8(* (U8*) (s)) >= 0xF9 \
493 && (NATIVE_UTF8_TO_I8(* (U8*) (s)) > 0xF9 \
a3481822 494 || (NATIVE_UTF8_TO_I8(* ((U8*) (s) + 1)) >= 0xA2)))
7131f24d 495#else
0404c92a
KW
496# define UTF8_IS_SUPER(s) (*(U8*) (s) >= 0xF4 \
497 && (*(U8*) (s) > 0xF4 || (*((U8*) (s) + 1) >= 0x90)))
7131f24d
KW
498#endif
499
b96a92fb
KW
500/* These are now machine generated, and the 'given' clause is no longer
501 * applicable */
502#define UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s) \
503 cBOOL(is_NONCHAR_utf8(s))
504#define UTF8_IS_NONCHAR_(s) \
505 UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s)
7131f24d 506
c867b360
JH
507#define UNICODE_SURROGATE_FIRST 0xD800
508#define UNICODE_SURROGATE_LAST 0xDFFF
509#define UNICODE_REPLACEMENT 0xFFFD
510#define UNICODE_BYTE_ORDER_MARK 0xFEFF
1d72bdf6 511
b851fbc1 512/* Though our UTF-8 encoding can go beyond this,
c76687c5 513 * let's be conservative and do as Unicode says. */
b851fbc1
JH
514#define PERL_UNICODE_MAX 0x10FFFF
515
949cf498
KW
516#define UNICODE_WARN_SURROGATE 0x0001 /* UTF-16 surrogates */
517#define UNICODE_WARN_NONCHAR 0x0002 /* Non-char code points */
518#define UNICODE_WARN_SUPER 0x0004 /* Above 0x10FFFF */
519#define UNICODE_WARN_FE_FF 0x0008 /* Above 0x10FFFF */
520#define UNICODE_DISALLOW_SURROGATE 0x0010
521#define UNICODE_DISALLOW_NONCHAR 0x0020
522#define UNICODE_DISALLOW_SUPER 0x0040
523#define UNICODE_DISALLOW_FE_FF 0x0080
bb88be5f
KW
524#define UNICODE_WARN_ILLEGAL_INTERCHANGE \
525 (UNICODE_WARN_SURROGATE|UNICODE_WARN_NONCHAR|UNICODE_WARN_SUPER)
526#define UNICODE_DISALLOW_ILLEGAL_INTERCHANGE \
527 (UNICODE_DISALLOW_SURROGATE|UNICODE_DISALLOW_NONCHAR|UNICODE_DISALLOW_SUPER)
949cf498
KW
528
529/* For backward source compatibility, as are now the default */
530#define UNICODE_ALLOW_SURROGATE 0
531#define UNICODE_ALLOW_SUPER 0
532#define UNICODE_ALLOW_ANY 0
b851fbc1 533
1d72bdf6
NIS
534#define UNICODE_IS_SURROGATE(c) ((c) >= UNICODE_SURROGATE_FIRST && \
535 (c) <= UNICODE_SURROGATE_LAST)
a10ec373 536#define UNICODE_IS_REPLACEMENT(c) ((c) == UNICODE_REPLACEMENT)
872c91ae 537#define UNICODE_IS_BYTE_ORDER_MARK(c) ((c) == UNICODE_BYTE_ORDER_MARK)
7131f24d
KW
538#define UNICODE_IS_NONCHAR(c) ((c >= 0xFDD0 && c <= 0xFDEF) \
539 /* The other noncharacters end in FFFE or FFFF, which \
540 * the mask below catches both of, but beyond the last \
541 * official unicode code point, they aren't \
542 * noncharacters, since those aren't Unicode \
543 * characters at all */ \
544 || ((((c & 0xFFFE) == 0xFFFE)) && ! UNICODE_IS_SUPER(c)))
545#define UNICODE_IS_SUPER(c) ((c) > PERL_UNICODE_MAX)
546#define UNICODE_IS_FE_FF(c) ((c) > 0x7FFFFFFF)
1d72bdf6 547
ec34087a
KW
548#define LATIN_SMALL_LETTER_SHARP_S LATIN_SMALL_LETTER_SHARP_S_NATIVE
549#define LATIN_SMALL_LETTER_Y_WITH_DIAERESIS \
550 LATIN_SMALL_LETTER_Y_WITH_DIAERESIS_NATIVE
551#define MICRO_SIGN MICRO_SIGN_NATIVE
552#define LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE \
553 LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE_NATIVE
554#define LATIN_SMALL_LETTER_A_WITH_RING_ABOVE \
555 LATIN_SMALL_LETTER_A_WITH_RING_ABOVE_NATIVE
09091399
JH
556#define UNICODE_GREEK_CAPITAL_LETTER_SIGMA 0x03A3
557#define UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA 0x03C2
558#define UNICODE_GREEK_SMALL_LETTER_SIGMA 0x03C3
9dcbe121 559#define GREEK_SMALL_LETTER_MU 0x03BC
9e682c18
KW
560#define GREEK_CAPITAL_LETTER_MU 0x039C /* Upper and title case
561 of MICRON */
562#define LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS 0x0178 /* Also is title case */
563#define LATIN_CAPITAL_LETTER_SHARP_S 0x1E9E
564#define LATIN_SMALL_LETTER_LONG_S 0x017F
a9f50d33
KW
565#define LATIN_SMALL_LIGATURE_LONG_S_T 0xFB05
566#define LATIN_SMALL_LIGATURE_ST 0xFB06
9e682c18
KW
567#define KELVIN_SIGN 0x212A
568#define ANGSTROM_SIGN 0x212B
09091399 569
9e55ce06 570#define UNI_DISPLAY_ISPRINT 0x0001
c728cb41
JH
571#define UNI_DISPLAY_BACKSLASH 0x0002
572#define UNI_DISPLAY_QQ (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH)
573#define UNI_DISPLAY_REGEX (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH)
9e55ce06 574
5cd46e1f
KW
575#define ANYOF_FOLD_SHARP_S(node, input, end) \
576 (ANYOF_BITMAP_TEST(node, LATIN_SMALL_LETTER_SHARP_S) && \
137165a6 577 (ANYOF_NONBITMAP(node)) && \
39065660 578 (ANYOF_FLAGS(node) & ANYOF_LOC_NONBITMAP_FOLD) && \
07b6858f 579 ((end) > (input) + 1) && \
305b8651 580 isALPHA_FOLD_EQ((input)[0], 's'))
6302f837 581
ebc501f0 582#define SHARP_S_SKIP 2
3b0fc154 583
a4f7a67c
KW
584/* If you want to exclude surrogates, and beyond legal Unicode, see the blame
585 * log for earlier versions which gave details for these */
4d646140 586
6302f837
KW
587/* A helper macro for isUTF8_CHAR, so use that one, and not this one. This is
588 * retained solely for backwards compatibility and may be deprecated and
589 * removed in a future Perl version.
590 *
591 * regen/regcharclass.pl generates is_UTF8_CHAR_utf8() macros for up to these
d9f92374
KW
592 * number of bytes. So this has to be coordinated with that file */
593#ifdef EBCDIC
594# define IS_UTF8_CHAR_FAST(n) ((n) <= 3)
595#else
596# define IS_UTF8_CHAR_FAST(n) ((n) <= 4)
597#endif
598
4d646140 599#ifndef EBCDIC
6302f837
KW
600/* A helper macro for isUTF8_CHAR, so use that one instead of this. This was
601 * generated by regen/regcharclass.pl, and then moved here. The lines that
602 * generated it were then commented out. This was done solely because it takes
603 * on the order of 10 minutes to generate, and is never going to change, unless
604 * the generated code is improved.
605 *
606 * The EBCDIC versions have been cut to not cover all of legal Unicode, so
607 * don't take too long to generate, and there is a separate one for each code
608 * page, so they are in regcharclass.h instead of here */
39a0f513 609/*
5dca9278 610 UTF8_CHAR: Matches legal UTF-8 encoded characters from 2 through 4 bytes
39a0f513 611
5dca9278 612 0x80 - 0x1FFFFF
39a0f513 613*/
4d646140 614/*** GENERATED CODE ***/
5dca9278
KW
615#define is_UTF8_CHAR_utf8_no_length_checks(s) \
616( ( 0xC2 <= ((U8*)s)[0] && ((U8*)s)[0] <= 0xDF ) ? \
617 ( ( ( ((U8*)s)[1] & 0xC0 ) == 0x80 ) ? 2 : 0 ) \
618: ( 0xE0 == ((U8*)s)[0] ) ? \
619 ( ( ( ( ((U8*)s)[1] & 0xE0 ) == 0xA0 ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
620: ( 0xE1 <= ((U8*)s)[0] && ((U8*)s)[0] <= 0xEF ) ? \
621 ( ( ( ( ((U8*)s)[1] & 0xC0 ) == 0x80 ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
622: ( 0xF0 == ((U8*)s)[0] ) ? \
623 ( ( ( ( 0x90 <= ((U8*)s)[1] && ((U8*)s)[1] <= 0xBF ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) && ( ( ((U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\
624: ( ( ( ( 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 625#endif
3b0fc154 626
6302f837 627/*
5dca9278
KW
628=head1 Unicode Support
629
630=for apidoc Am|STRLEN|isUTF8_CHAR|const U8 *s|const U8 *e
631
632Returns the number of bytes beginning at C<s> which form a legal UTF-8 (or
633UTF-EBCDIC) encoded character, looking no further than C<e - s> bytes into
634C<s>. Returns 0 if the sequence starting at C<s> through C<e - 1> is not
635well-formed UTF-8
6302f837
KW
636
637Note that an INVARIANT character (i.e. ASCII on non-EBCDIC
5dca9278
KW
638machines) is a valid UTF-8 character.
639
640=cut
641*/
6302f837 642
dd9bc2b0 643#define isUTF8_CHAR(s, e) (UNLIKELY((e) <= (s)) \
6302f837
KW
644 ? 0 \
645 : (UTF8_IS_INVARIANT(*s)) \
646 ? 1 \
dd9bc2b0 647 : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \
6302f837 648 ? 0 \
dd9bc2b0 649 : LIKELY(IS_UTF8_CHAR_FAST(UTF8SKIP(s))) \
5dca9278 650 ? is_UTF8_CHAR_utf8_no_length_checks(s) \
6302f837
KW
651 : _is_utf8_char_slow(s, e))
652
3cedd9d9
KW
653#define is_utf8_char_buf(buf, buf_end) isUTF8_CHAR(buf, buf_end)
654
6302f837
KW
655/* Do not use; should be deprecated. Use isUTF8_CHAR() instead; this is
656 * retained solely for backwards compatibility */
657#define IS_UTF8_CHAR(p, n) (isUTF8_CHAR(p, (p) + (n)) == n)
e9a8c099 658
57f0e7e2
KW
659#endif /* H_UTF8 */
660
e9a8c099
MHM
661/*
662 * Local variables:
663 * c-indentation-style: bsd
664 * c-basic-offset: 4
14d04a33 665 * indent-tabs-mode: nil
e9a8c099
MHM
666 * End:
667 *
14d04a33 668 * ex: set ts=8 sts=4 sw=4 et:
e9a8c099 669 */