Commit | Line | Data |
---|---|---|
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 | ||
051a06d4 KW |
29 | /* For to_utf8_fold_flags, q.v. */ |
30 | #define FOLD_FLAGS_LOCALE 0x1 | |
31 | #define FOLD_FLAGS_FULL 0x2 | |
a0270393 | 32 | #define FOLD_FLAGS_NOMIX_ASCII 0x4 |
051a06d4 | 33 | |
83199d38 KW |
34 | /* For _core_swash_init(), internal core use only */ |
35 | #define _CORE_SWASH_INIT_USER_DEFINED_PROPERTY 0x1 | |
5d3d13d1 | 36 | #define _CORE_SWASH_INIT_RETURN_IF_UNDEF 0x2 |
87367d5f | 37 | #define _CORE_SWASH_INIT_ACCEPT_INVLIST 0x4 |
83199d38 | 38 | |
a0270393 | 39 | #define to_uni_fold(c, p, lenp) _to_uni_fold_flags(c, p, lenp, FOLD_FLAGS_FULL) |
051a06d4 KW |
40 | #define to_utf8_fold(c, p, lenp) _to_utf8_fold_flags(c, p, lenp, \ |
41 | FOLD_FLAGS_FULL, NULL) | |
42 | #define to_utf8_lower(a,b,c) _to_utf8_lower_flags(a,b,c,0, NULL) | |
43 | #define to_utf8_upper(a,b,c) _to_utf8_upper_flags(a,b,c,0, NULL) | |
44 | #define to_utf8_title(a,b,c) _to_utf8_title_flags(a,b,c,0, NULL) | |
36bb2ab6 | 45 | |
fd7cb289 RGS |
46 | /* Source backward compatibility. */ |
47 | #define uvuni_to_utf8(d, uv) uvuni_to_utf8_flags(d, uv, 0) | |
48 | #define is_utf8_string_loc(s, len, ep) is_utf8_string_loclen(s, len, ep, 0) | |
49 | ||
eda9cac1 KW |
50 | #define foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \ |
51 | foldEQ_utf8_flags(s1, pe1, l1, u1, s2, pe2, l2, u2, 0) | |
a33c29bc | 52 | #define FOLDEQ_UTF8_NOMIX_ASCII (1 << 0) |
5e64d0fa | 53 | #define FOLDEQ_UTF8_LOCALE (1 << 1) |
18f762c3 KW |
54 | #define FOLDEQ_S1_ALREADY_FOLDED (1 << 2) |
55 | #define FOLDEQ_S2_ALREADY_FOLDED (1 << 3) | |
a33c29bc | 56 | |
e6226b18 KW |
57 | /* |
58 | =for apidoc ibcmp_utf8 | |
59 | ||
60 | This is a synonym for (! foldEQ_utf8()) | |
61 | ||
62 | =cut | |
63 | */ | |
64 | #define ibcmp_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \ | |
65 | cBOOL(! foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2)) | |
66 | ||
1d72bdf6 NIS |
67 | #ifdef EBCDIC |
68 | /* The equivalent of these macros but implementing UTF-EBCDIC | |
69 | are in the following header file: | |
70 | */ | |
71 | ||
72 | #include "utfebcdic.h" | |
fd7cb289 | 73 | |
d06134e5 | 74 | #else /* ! EBCDIC */ |
73c4f7a1 GS |
75 | START_EXTERN_C |
76 | ||
a0ed51b3 | 77 | #ifdef DOINIT |
6f06b55f | 78 | EXTCONST unsigned char PL_utf8skip[] = { |
b2635aa8 KW |
79 | /* 0x00 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */ |
80 | /* 0x10 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */ | |
81 | /* 0x20 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */ | |
82 | /* 0x30 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */ | |
83 | /* 0x40 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */ | |
84 | /* 0x50 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */ | |
85 | /* 0x60 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */ | |
86 | /* 0x70 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */ | |
87 | /* 0x80 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */ | |
88 | /* 0x90 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */ | |
89 | /* 0xA0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */ | |
90 | /* 0xB0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */ | |
91 | /* 0xC0 */ 2,2, /* overlong */ | |
92 | /* 0xC2 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0080 to U+03FF */ | |
93 | /* 0xD0 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0400 to U+07FF */ | |
94 | /* 0xE0 */ 3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3, /* U+0800 to U+FFFF */ | |
95 | /* 0xF0 */ 4,4,4,4,4,4,4,4,5,5,5,5,6,6, /* above BMP to 2**31 - 1 */ | |
96 | /* 0xFE */ 7,13, /* Perl extended (never was official UTF-8). Up to 72bit | |
97 | allowed (64-bit + reserved). */ | |
a0ed51b3 LW |
98 | }; |
99 | #else | |
6f06b55f | 100 | EXTCONST unsigned char PL_utf8skip[]; |
a0ed51b3 LW |
101 | #endif |
102 | ||
73c4f7a1 | 103 | END_EXTERN_C |
7e2040f0 | 104 | |
b96a92fb | 105 | #include "regcharclass.h" |
ec34087a KW |
106 | #include "unicode_constants.h" |
107 | ||
59a449d5 KW |
108 | /* Native character to/from iso-8859-1. Are the identity functions on ASCII |
109 | * platforms */ | |
110 | #define NATIVE_TO_LATIN1(ch) (ch) | |
111 | #define LATIN1_TO_NATIVE(ch) (ch) | |
112 | ||
113 | /* I8 is an intermediate version of UTF-8 used only in UTF-EBCDIC. We thus | |
114 | * consider it to be identical to UTF-8 on ASCII platforms. Strictly speaking | |
115 | * UTF-8 and UTF-EBCDIC are two different things, but we often conflate them | |
116 | * because they are 8-bit encodings that serve the same purpose in Perl, and | |
117 | * rarely do we need to distinguish them. The term "NATIVE_UTF8" applies to | |
118 | * whichever one is applicable on the current platform */ | |
119 | #define NATIVE_UTF8_TO_I8(ch) (ch) | |
120 | #define I8_TO_NATIVE_UTF8(ch) (ch) | |
121 | ||
1d72bdf6 NIS |
122 | /* Transforms in wide UV chars */ |
123 | #define UNI_TO_NATIVE(ch) (ch) | |
124 | #define NATIVE_TO_UNI(ch) (ch) | |
d7578b48 | 125 | |
d06134e5 | 126 | /* As there are no translations, avoid the function wrapper */ |
1754c1a1 | 127 | #define utf8n_to_uvchr utf8n_to_uvuni |
6dd9dce9 | 128 | #define valid_utf8_to_uvchr valid_utf8_to_uvuni |
1754c1a1 | 129 | #define uvchr_to_utf8 uvuni_to_utf8 |
2b9d42f0 | 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 | 152 | Note the gaps before several of the byte entries above marked by '*'. These are |
37e2e78e KW |
153 | caused by legal UTF-8 avoiding non-shortest encodings: it is technically |
154 | possible to UTF-8-encode a single code point in different ways, but that is | |
155 | explicitly 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 | |
170 | As you can see, the continuation bytes all begin with C<10>, and the | |
e1b711da | 171 | leading bits of the start byte tell how many bytes there are in the |
8c007b5a JH |
172 | encoded character. |
173 | ||
65ab9279 TC |
174 | Perl'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 | |
207 | * character. | |
208 | * */ | |
22901f30 | 209 | #define UTF_START_MARK(len) (((len) > 7) ? 0xFF : (0xFE << (7-(len)))) |
b2635aa8 KW |
210 | |
211 | /* Masks out the initial one bits in a start byte, leaving the real data ones. | |
212 | * Doesn't work on an invariant byte */ | |
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 | |
1d68d6cd | 235 | #ifdef HAS_QUAD |
5bbb0b5a | 236 | #define UNISKIP(uv) ( (uv) < 0x80 ? 1 : \ |
1d68d6cd SC |
237 | (uv) < 0x800 ? 2 : \ |
238 | (uv) < 0x10000 ? 3 : \ | |
239 | (uv) < 0x200000 ? 4 : \ | |
240 | (uv) < 0x4000000 ? 5 : \ | |
241 | (uv) < 0x80000000 ? 6 : \ | |
9041c2e3 | 242 | (uv) < UTF8_QUAD_MAX ? 7 : 13 ) |
1d68d6cd SC |
243 | #else |
244 | /* No, I'm not even going to *TRY* putting #ifdef inside a #define */ | |
5bbb0b5a | 245 | #define UNISKIP(uv) ( (uv) < 0x80 ? 1 : \ |
1d68d6cd SC |
246 | (uv) < 0x800 ? 2 : \ |
247 | (uv) < 0x10000 ? 3 : \ | |
248 | (uv) < 0x200000 ? 4 : \ | |
249 | (uv) < 0x4000000 ? 5 : \ | |
250 | (uv) < 0x80000000 ? 6 : 7 ) | |
251 | #endif | |
252 | ||
03c76984 KW |
253 | /* How wide can a single UTF-8 encoded character become in bytes. */ |
254 | /* NOTE: Strictly speaking Perl's UTF-8 should not be called UTF-8 since UTF-8 | |
255 | * is an encoding of Unicode, and Unicode's upper limit, 0x10FFFF, can be | |
256 | * expressed with 4 bytes. However, Perl thinks of UTF-8 as a way to encode | |
257 | * non-negative integers in a binary format, even those above Unicode */ | |
258 | #define UTF8_MAXBYTES 13 | |
259 | ||
260 | /* The maximum number of UTF-8 bytes a single Unicode character can | |
261 | * uppercase/lowercase/fold into. Unicode guarantees that the maximum | |
262 | * expansion is 3 characters. On ASCIIish platforms, the highest Unicode | |
263 | * character occupies 4 bytes, therefore this number would be 12, but this is | |
264 | * smaller than the maximum width a single above-Unicode character can occupy, | |
265 | * so use that instead */ | |
266 | #if UTF8_MAXBYTES < 12 | |
267 | #error UTF8_MAXBYTES must be at least 12 | |
268 | #endif | |
269 | ||
270 | #define UTF8_MAXBYTES_CASE UTF8_MAXBYTES | |
271 | ||
d06134e5 KW |
272 | #endif /* EBCDIC vs ASCII */ |
273 | ||
274 | /* Rest of these are attributes of Unicode and perl's internals rather than the | |
275 | * encoding, or happen to be the same in both ASCII and EBCDIC (at least at | |
276 | * this level; the macros that some of these call may have different | |
277 | * definitions in the two encodings */ | |
278 | ||
59a449d5 KW |
279 | /* In domain restricted to ASCII, these may make more sense to the reader than |
280 | * the ones with Latin1 in the name */ | |
281 | #define NATIVE_TO_ASCII(ch) NATIVE_TO_LATIN1(ch) | |
282 | #define ASCII_TO_NATIVE(ch) LATIN1_TO_NATIVE(ch) | |
283 | ||
284 | /* More or less misleadingly-named defines, retained for back compat */ | |
285 | #define NATIVE_TO_UTF(ch) NATIVE_UTF8_TO_I8(ch) | |
286 | #define NATIVE_TO_I8(ch) NATIVE_UTF8_TO_I8(ch) | |
287 | #define UTF_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch) | |
288 | #define I8_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch) | |
289 | #define NATIVE8_TO_UNI(ch) NATIVE_TO_LATIN1(ch) | |
d06134e5 | 290 | |
537124e4 KW |
291 | /* Adds a UTF8 continuation byte 'new' of information to a running total code |
292 | * point 'old' of all the continuation bytes so far. This is designed to be | |
293 | * used in a loop to convert from UTF-8 to the code point represented */ | |
bb88be5f KW |
294 | #define UTF8_ACCUMULATE(old, new) (((old) << UTF_ACCUMULATION_SHIFT) \ |
295 | | (((U8)new) & UTF_CONTINUATION_MASK)) | |
d06134e5 | 296 | |
4ab10950 KW |
297 | /* This works in the face of malformed UTF-8. */ |
298 | #define UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(s, e) (UTF8_IS_DOWNGRADEABLE_START(*s) \ | |
299 | && ( (e) - (s) > 1) \ | |
300 | && UTF8_IS_CONTINUATION(*((s)+1))) | |
301 | ||
94bb8c36 | 302 | /* Convert a two (not one) byte utf8 character to a native code point value. |
2950f2a7 KW |
303 | * Needs just one iteration of accumulate. Should not be used unless it is |
304 | * known that the two bytes are legal: 1) two-byte start, and 2) continuation. | |
305 | * Note that the result can be larger than 255 if the input character is not | |
306 | * downgradable */ | |
94bb8c36 KW |
307 | #define TWO_BYTE_UTF8_TO_NATIVE(HI, LO) \ |
308 | UNI_TO_NATIVE(UTF8_ACCUMULATE((NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), \ | |
309 | NATIVE_UTF8_TO_I8(LO))) | |
310 | ||
311 | /* Should never be used, and be deprecated */ | |
312 | #define TWO_BYTE_UTF8_TO_UNI(HI, LO) NATIVE_TO_UNI(TWO_BYTE_UTF8_TO_NATIVE(HI, LO)) | |
2950f2a7 | 313 | |
15824458 KW |
314 | /* How many bytes in the UTF-8 encoded character whose first (perhaps only) |
315 | * byte is pointed to by 's' */ | |
d06134e5 KW |
316 | #define UTF8SKIP(s) PL_utf8skip[*(const U8*)(s)] |
317 | ||
15824458 KW |
318 | /* Is the byte 'c' the same character when encoded in UTF-8 as when not. This |
319 | * works on both UTF-8 encoded strings and non-encoded, as it returns TRUE in | |
320 | * each for the exact same set of bit patterns. (And it works on any byte in a | |
321 | * UTF-8 encoded string) */ | |
bc3632a8 | 322 | #define UTF8_IS_INVARIANT(c) UNI_IS_INVARIANT(NATIVE_UTF8_TO_I8(c)) |
15824458 | 323 | |
bc3632a8 | 324 | #define NATIVE_IS_INVARIANT(c) UNI_IS_INVARIANT(NATIVE_TO_LATIN1(c)) |
d06134e5 KW |
325 | |
326 | #define MAX_PORTABLE_UTF8_TWO_BYTE 0x3FF /* constrained by EBCDIC */ | |
327 | ||
328 | /* The macros in the next sets are used to generate the two utf8 or utfebcdic | |
329 | * bytes from an ordinal that is known to fit into two bytes; it must be less | |
330 | * than 0x3FF to work across both encodings. */ | |
67af0a71 KW |
331 | /* Nocast allows these to be used in the case label of a switch statement; |
332 | * however this doesn't won't work for ebcdic, and should be avoided. Use | |
333 | * regen/unicode_constants instead */ | |
334 | #define UTF8_TWO_BYTE_HI_nocast(c) I8_TO_NATIVE_UTF8((NATIVE_TO_UNI(c) \ | |
428921e5 | 335 | >> UTF_ACCUMULATION_SHIFT) | (0xFF & UTF_START_MARK(2))) |
67af0a71 KW |
336 | #define UTF8_TWO_BYTE_LO_nocast(c) I8_TO_NATIVE_UTF8((NATIVE_TO_UNI(c) \ |
337 | & UTF_CONTINUATION_MASK) \ | |
338 | | UTF_CONTINUATION_MARK) | |
d06134e5 KW |
339 | |
340 | #define UTF8_TWO_BYTE_HI(c) ((U8) (UTF8_TWO_BYTE_HI_nocast(c))) | |
341 | #define UTF8_TWO_BYTE_LO(c) ((U8) (UTF8_TWO_BYTE_LO_nocast(c))) | |
342 | ||
67af0a71 KW |
343 | /* This name is used when the source is a single byte. For EBCDIC these could |
344 | * be more efficiently written; the reason is that things above 0xFF have to be | |
345 | * special-cased, which is done by the EBCDIC version of NATIVE_TO_UNI() */ | |
d06134e5 KW |
346 | #define UTF8_EIGHT_BIT_HI(c) UTF8_TWO_BYTE_HI((U8)(c)) |
347 | #define UTF8_EIGHT_BIT_LO(c) UTF8_TWO_BYTE_LO((U8)(c)) | |
348 | ||
e7214ce8 KW |
349 | /* This is illegal in any well-formed UTF-8 in both EBCDIC and ASCII |
350 | * as it is only in overlongs. */ | |
351 | #define ILLEGAL_UTF8_BYTE I8_TO_NATIVE_UTF8(0xC1) | |
352 | ||
7e2040f0 | 353 | /* |
e3036cf4 | 354 | * 'UTF' is whether or not p is encoded in UTF8. The names 'foo_lazy_if' stem |
20df05f4 KW |
355 | * from an earlier version of these macros in which they didn't call the |
356 | * foo_utf8() macros (i.e. were 'lazy') unless they decided that *p is the | |
357 | * beginning of a utf8 character. Now that foo_utf8() determines that itself, | |
358 | * no need to do it again here | |
7e2040f0 | 359 | */ |
e3036cf4 KW |
360 | #define isIDFIRST_lazy_if(p,UTF) ((IN_BYTES || !UTF ) \ |
361 | ? isIDFIRST(*(p)) \ | |
362 | : isIDFIRST_utf8((const U8*)p)) | |
32636478 KW |
363 | #define isWORDCHAR_lazy_if(p,UTF) ((IN_BYTES || (!UTF )) \ |
364 | ? isWORDCHAR(*(p)) \ | |
365 | : isWORDCHAR_utf8((const U8*)p)) | |
366 | #define isALNUM_lazy_if(p,UTF) isWORDCHAR_lazy_if(p,UTF) | |
1d72bdf6 | 367 | |
89ebb4a3 JH |
368 | #define UTF8_MAXLEN UTF8_MAXBYTES |
369 | ||
8cb75cc8 KW |
370 | /* A Unicode character can fold to up to 3 characters */ |
371 | #define UTF8_MAX_FOLD_CHAR_EXPAND 3 | |
372 | ||
a98fe34d | 373 | #define IN_BYTES (CopHINTS_get(PL_curcop) & HINT_BYTES) |
0064a8a9 | 374 | #define DO_UTF8(sv) (SvUTF8(sv) && !IN_BYTES) |
66cbab2c | 375 | #define IN_UNI_8_BIT \ |
b36bf33f KW |
376 | (CopHINTS_get(PL_curcop) & (HINT_UNI_8_BIT|HINT_LOCALE_NOT_CHARS) \ |
377 | && ! IN_LOCALE_RUNTIME && ! IN_BYTES) | |
378 | ||
1d72bdf6 | 379 | |
c76687c5 KW |
380 | #define UTF8_ALLOW_EMPTY 0x0001 /* Allow a zero length string */ |
381 | ||
382 | /* Allow first byte to be a continuation byte */ | |
1d72bdf6 | 383 | #define UTF8_ALLOW_CONTINUATION 0x0002 |
c76687c5 KW |
384 | |
385 | /* Allow second... bytes to be non-continuation bytes */ | |
1d72bdf6 | 386 | #define UTF8_ALLOW_NON_CONTINUATION 0x0004 |
949cf498 KW |
387 | |
388 | /* expecting more bytes than were available in the string */ | |
389 | #define UTF8_ALLOW_SHORT 0x0008 | |
390 | ||
391 | /* Overlong sequence; i.e., the code point can be specified in fewer bytes. */ | |
392 | #define UTF8_ALLOW_LONG 0x0010 | |
393 | ||
394 | #define UTF8_DISALLOW_SURROGATE 0x0020 /* Unicode surrogates */ | |
395 | #define UTF8_WARN_SURROGATE 0x0040 | |
396 | ||
397 | #define UTF8_DISALLOW_NONCHAR 0x0080 /* Unicode non-character */ | |
398 | #define UTF8_WARN_NONCHAR 0x0100 /* code points */ | |
399 | ||
400 | #define UTF8_DISALLOW_SUPER 0x0200 /* Super-set of Unicode: code */ | |
401 | #define UTF8_WARN_SUPER 0x0400 /* points above the legal max */ | |
402 | ||
403 | /* Code points which never were part of the original UTF-8 standard, the first | |
404 | * byte of which is a FE or FF on ASCII platforms. */ | |
405 | #define UTF8_DISALLOW_FE_FF 0x0800 | |
406 | #define UTF8_WARN_FE_FF 0x1000 | |
407 | ||
408 | #define UTF8_CHECK_ONLY 0x2000 | |
409 | ||
410 | /* For backwards source compatibility. They do nothing, as the default now | |
411 | * includes what they used to mean. The first one's meaning was to allow the | |
412 | * just the single non-character 0xFFFF */ | |
413 | #define UTF8_ALLOW_FFFF 0 | |
414 | #define UTF8_ALLOW_SURROGATE 0 | |
415 | ||
33d9abfb | 416 | #define UTF8_DISALLOW_ILLEGAL_INTERCHANGE (UTF8_DISALLOW_SUPER|UTF8_DISALLOW_NONCHAR|UTF8_DISALLOW_SURROGATE|UTF8_DISALLOW_FE_FF) |
949cf498 | 417 | #define UTF8_WARN_ILLEGAL_INTERCHANGE \ |
33d9abfb | 418 | (UTF8_WARN_SUPER|UTF8_WARN_NONCHAR|UTF8_WARN_SURROGATE|UTF8_WARN_FE_FF) |
949cf498 KW |
419 | #define UTF8_ALLOW_ANY \ |
420 | (~(UTF8_DISALLOW_ILLEGAL_INTERCHANGE|UTF8_WARN_ILLEGAL_INTERCHANGE)) | |
421 | #define UTF8_ALLOW_ANYUV \ | |
422 | (UTF8_ALLOW_EMPTY \ | |
423 | & ~(UTF8_DISALLOW_ILLEGAL_INTERCHANGE|UTF8_WARN_ILLEGAL_INTERCHANGE)) | |
9f7f3913 TS |
424 | #define UTF8_ALLOW_DEFAULT (ckWARN(WARN_UTF8) ? 0 : \ |
425 | UTF8_ALLOW_ANYUV) | |
1d72bdf6 | 426 | |
7131f24d KW |
427 | /* Surrogates, non-character code points and above-Unicode code points are |
428 | * problematic in some contexts. This allows code that needs to check for | |
429 | * those to to quickly exclude the vast majority of code points it will | |
430 | * encounter */ | |
b96a92fb KW |
431 | #define UTF8_FIRST_PROBLEMATIC_CODE_POINT_FIRST_BYTE \ |
432 | FIRST_SURROGATE_UTF8_FIRST_BYTE | |
7131f24d | 433 | |
b96a92fb KW |
434 | #define UTF8_IS_SURROGATE(s) cBOOL(is_SURROGATE_utf8(s)) |
435 | #define UTF8_IS_REPLACEMENT(s, send) cBOOL(is_REPLACEMENT_utf8_safe(s,send)) | |
7131f24d KW |
436 | |
437 | /* ASCII EBCDIC I8 | |
438 | * U+10FFFF: \xF4\x8F\xBF\xBF \xF9\xA1\xBF\xBF\xBF max legal Unicode | |
439 | * U+110000: \xF4\x90\x80\x80 \xF9\xA2\xA0\xA0\xA0 | |
440 | * U+110001: \xF4\x90\x80\x81 \xF9\xA2\xA0\xA0\xA1 | |
441 | */ | |
442 | #ifdef EBCDIC /* Both versions assume well-formed UTF8 */ | |
bc3632a8 KW |
443 | # define UTF8_IS_SUPER(s) (NATIVE_UTF8_TO_I8(* (U8*) (s)) >= 0xF9 \ |
444 | && (NATIVE_UTF8_TO_I8(* (U8*) (s)) > 0xF9 \ | |
445 | || (NATIVE_UTF8_TO_I8(* (U8*) ((s)) + 1 >= 0xA2)))) | |
7131f24d | 446 | #else |
0404c92a KW |
447 | # define UTF8_IS_SUPER(s) (*(U8*) (s) >= 0xF4 \ |
448 | && (*(U8*) (s) > 0xF4 || (*((U8*) (s) + 1) >= 0x90))) | |
7131f24d KW |
449 | #endif |
450 | ||
b96a92fb KW |
451 | /* These are now machine generated, and the 'given' clause is no longer |
452 | * applicable */ | |
453 | #define UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s) \ | |
454 | cBOOL(is_NONCHAR_utf8(s)) | |
455 | #define UTF8_IS_NONCHAR_(s) \ | |
456 | UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s) | |
7131f24d | 457 | |
c867b360 JH |
458 | #define UNICODE_SURROGATE_FIRST 0xD800 |
459 | #define UNICODE_SURROGATE_LAST 0xDFFF | |
460 | #define UNICODE_REPLACEMENT 0xFFFD | |
461 | #define UNICODE_BYTE_ORDER_MARK 0xFEFF | |
1d72bdf6 | 462 | |
b851fbc1 | 463 | /* Though our UTF-8 encoding can go beyond this, |
c76687c5 | 464 | * let's be conservative and do as Unicode says. */ |
b851fbc1 JH |
465 | #define PERL_UNICODE_MAX 0x10FFFF |
466 | ||
949cf498 KW |
467 | #define UNICODE_WARN_SURROGATE 0x0001 /* UTF-16 surrogates */ |
468 | #define UNICODE_WARN_NONCHAR 0x0002 /* Non-char code points */ | |
469 | #define UNICODE_WARN_SUPER 0x0004 /* Above 0x10FFFF */ | |
470 | #define UNICODE_WARN_FE_FF 0x0008 /* Above 0x10FFFF */ | |
471 | #define UNICODE_DISALLOW_SURROGATE 0x0010 | |
472 | #define UNICODE_DISALLOW_NONCHAR 0x0020 | |
473 | #define UNICODE_DISALLOW_SUPER 0x0040 | |
474 | #define UNICODE_DISALLOW_FE_FF 0x0080 | |
bb88be5f KW |
475 | #define UNICODE_WARN_ILLEGAL_INTERCHANGE \ |
476 | (UNICODE_WARN_SURROGATE|UNICODE_WARN_NONCHAR|UNICODE_WARN_SUPER) | |
477 | #define UNICODE_DISALLOW_ILLEGAL_INTERCHANGE \ | |
478 | (UNICODE_DISALLOW_SURROGATE|UNICODE_DISALLOW_NONCHAR|UNICODE_DISALLOW_SUPER) | |
949cf498 KW |
479 | |
480 | /* For backward source compatibility, as are now the default */ | |
481 | #define UNICODE_ALLOW_SURROGATE 0 | |
482 | #define UNICODE_ALLOW_SUPER 0 | |
483 | #define UNICODE_ALLOW_ANY 0 | |
b851fbc1 | 484 | |
1d72bdf6 NIS |
485 | #define UNICODE_IS_SURROGATE(c) ((c) >= UNICODE_SURROGATE_FIRST && \ |
486 | (c) <= UNICODE_SURROGATE_LAST) | |
a10ec373 | 487 | #define UNICODE_IS_REPLACEMENT(c) ((c) == UNICODE_REPLACEMENT) |
872c91ae | 488 | #define UNICODE_IS_BYTE_ORDER_MARK(c) ((c) == UNICODE_BYTE_ORDER_MARK) |
7131f24d KW |
489 | #define UNICODE_IS_NONCHAR(c) ((c >= 0xFDD0 && c <= 0xFDEF) \ |
490 | /* The other noncharacters end in FFFE or FFFF, which \ | |
491 | * the mask below catches both of, but beyond the last \ | |
492 | * official unicode code point, they aren't \ | |
493 | * noncharacters, since those aren't Unicode \ | |
494 | * characters at all */ \ | |
495 | || ((((c & 0xFFFE) == 0xFFFE)) && ! UNICODE_IS_SUPER(c))) | |
496 | #define UNICODE_IS_SUPER(c) ((c) > PERL_UNICODE_MAX) | |
497 | #define UNICODE_IS_FE_FF(c) ((c) > 0x7FFFFFFF) | |
1d72bdf6 | 498 | |
f067b878 NA |
499 | #ifdef HAS_QUAD |
500 | # define UTF8_QUAD_MAX UINT64_C(0x1000000000) | |
501 | #endif | |
3bd709b1 | 502 | |
ec34087a KW |
503 | #define LATIN_SMALL_LETTER_SHARP_S LATIN_SMALL_LETTER_SHARP_S_NATIVE |
504 | #define LATIN_SMALL_LETTER_Y_WITH_DIAERESIS \ | |
505 | LATIN_SMALL_LETTER_Y_WITH_DIAERESIS_NATIVE | |
506 | #define MICRO_SIGN MICRO_SIGN_NATIVE | |
507 | #define LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE \ | |
508 | LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE_NATIVE | |
509 | #define LATIN_SMALL_LETTER_A_WITH_RING_ABOVE \ | |
510 | LATIN_SMALL_LETTER_A_WITH_RING_ABOVE_NATIVE | |
09091399 JH |
511 | #define UNICODE_GREEK_CAPITAL_LETTER_SIGMA 0x03A3 |
512 | #define UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA 0x03C2 | |
513 | #define UNICODE_GREEK_SMALL_LETTER_SIGMA 0x03C3 | |
9dcbe121 | 514 | #define GREEK_SMALL_LETTER_MU 0x03BC |
78a0d3cc KW |
515 | #define GREEK_CAPITAL_LETTER_MU 0x039C /* Upper and title case of MICRON */ |
516 | #define LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS 0x0178 /* Also is title case */ | |
97298f37 | 517 | #define LATIN_CAPITAL_LETTER_SHARP_S 0x1E9E |
8d64d87f KW |
518 | #define LATIN_SMALL_LETTER_LONG_S 0x017F |
519 | #define KELVIN_SIGN 0x212A | |
520 | #define ANGSTROM_SIGN 0x212B | |
09091399 | 521 | |
9e55ce06 | 522 | #define UNI_DISPLAY_ISPRINT 0x0001 |
c728cb41 JH |
523 | #define UNI_DISPLAY_BACKSLASH 0x0002 |
524 | #define UNI_DISPLAY_QQ (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH) | |
525 | #define UNI_DISPLAY_REGEX (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH) | |
9e55ce06 | 526 | |
5cd46e1f KW |
527 | #define ANYOF_FOLD_SHARP_S(node, input, end) \ |
528 | (ANYOF_BITMAP_TEST(node, LATIN_SMALL_LETTER_SHARP_S) && \ | |
137165a6 | 529 | (ANYOF_NONBITMAP(node)) && \ |
39065660 | 530 | (ANYOF_FLAGS(node) & ANYOF_LOC_NONBITMAP_FOLD) && \ |
07b6858f | 531 | ((end) > (input) + 1) && \ |
d22b930b KW |
532 | toFOLD((input)[0]) == 's' && \ |
533 | toFOLD((input)[1]) == 's') | |
ebc501f0 | 534 | #define SHARP_S_SKIP 2 |
3b0fc154 | 535 | |
a4f7a67c KW |
536 | /* If you want to exclude surrogates, and beyond legal Unicode, see the blame |
537 | * log for earlier versions which gave details for these */ | |
4d646140 KW |
538 | |
539 | #ifndef EBCDIC | |
540 | /* This was generated by regen/regcharclass.pl, and then moved here. The lines | |
541 | * that generated it were then commented out. This was done solely because it | |
542 | * takes on the order of 10 minutes to generate, and is never going to change. | |
543 | * The EBCDIC equivalent hasn't been commented out in regcharclass.pl, so it | |
544 | * should generate and run the correct stuff */ | |
39a0f513 YO |
545 | /* |
546 | UTF8_CHAR: Matches utf8 from 1 to 4 bytes | |
547 | ||
548 | 0x0 - 0x1FFFFF | |
549 | */ | |
4d646140 KW |
550 | /*** GENERATED CODE ***/ |
551 | #define is_UTF8_CHAR_utf8_safe(s,e) \ | |
552 | ( ((e)-(s) > 3) ? \ | |
553 | ( ( ( ((U8*)s)[0] & 0x80 ) == 0x00 ) ? 1 \ | |
554 | : ( 0xC2 <= ((U8*)s)[0] && ((U8*)s)[0] <= 0xDF ) ? \ | |
555 | ( ( ( ((U8*)s)[1] & 0xC0 ) == 0x80 ) ? 2 : 0 ) \ | |
556 | : ( 0xE0 == ((U8*)s)[0] ) ? \ | |
557 | ( ( ( ( ((U8*)s)[1] & 0xE0 ) == 0xA0 ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\ | |
558 | : ( 0xE1 <= ((U8*)s)[0] && ((U8*)s)[0] <= 0xEF ) ? \ | |
559 | ( ( ( ( ((U8*)s)[1] & 0xC0 ) == 0x80 ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\ | |
560 | : ( 0xF0 == ((U8*)s)[0] ) ? \ | |
561 | ( ( ( ( 0x90 <= ((U8*)s)[1] && ((U8*)s)[1] <= 0xBF ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) && ( ( ((U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\ | |
39a0f513 | 562 | : ( ( ( ( 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 |
563 | : ((e)-(s) > 2) ? \ |
564 | ( ( ( ((U8*)s)[0] & 0x80 ) == 0x00 ) ? 1 \ | |
565 | : ( 0xC2 <= ((U8*)s)[0] && ((U8*)s)[0] <= 0xDF ) ? \ | |
566 | ( ( ( ((U8*)s)[1] & 0xC0 ) == 0x80 ) ? 2 : 0 ) \ | |
567 | : ( 0xE0 == ((U8*)s)[0] ) ? \ | |
568 | ( ( ( ( ((U8*)s)[1] & 0xE0 ) == 0xA0 ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\ | |
39a0f513 | 569 | : ( ( ( 0xE1 <= ((U8*)s)[0] && ((U8*)s)[0] <= 0xEF ) && ( ( ((U8*)s)[1] & 0xC0 ) == 0x80 ) ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\ |
4d646140 KW |
570 | : ((e)-(s) > 1) ? \ |
571 | ( ( ( ((U8*)s)[0] & 0x80 ) == 0x00 ) ? 1 \ | |
39a0f513 | 572 | : ( ( 0xC2 <= ((U8*)s)[0] && ((U8*)s)[0] <= 0xDF ) && ( ( ((U8*)s)[1] & 0xC0 ) == 0x80 ) ) ? 2 : 0 )\ |
4d646140 KW |
573 | : ((e)-(s) > 0) ? \ |
574 | ( ( ((U8*)s)[0] & 0x80 ) == 0x00 ) \ | |
575 | : 0 ) | |
576 | #endif | |
3b0fc154 JH |
577 | |
578 | /* IS_UTF8_CHAR(p) is strictly speaking wrong (not UTF-8) because it | |
579 | * (1) allows UTF-8 encoded UTF-16 surrogates | |
580 | * (2) it allows code points past U+10FFFF. | |
581 | * The Perl_is_utf8_char() full "slow" code will handle the Perl | |
582 | * "extended UTF-8". */ | |
4d646140 | 583 | #define IS_UTF8_CHAR(p, n) (is_UTF8_CHAR_utf8_safe(p, (p) + (n)) == n) |
2f454f11 | 584 | |
4d646140 KW |
585 | /* regen/regcharclass.pl generates is_UTF8_CHAR_utf8_safe() macros for up to |
586 | * these number of bytes. So this has to be coordinated with it */ | |
587 | #ifdef EBCDIC | |
2f454f11 | 588 | # define IS_UTF8_CHAR_FAST(n) ((n) <= 5) |
4d646140 KW |
589 | #else |
590 | # define IS_UTF8_CHAR_FAST(n) ((n) <= 4) | |
591 | #endif | |
e9a8c099 | 592 | |
57f0e7e2 KW |
593 | #endif /* H_UTF8 */ |
594 | ||
e9a8c099 MHM |
595 | /* |
596 | * Local variables: | |
597 | * c-indentation-style: bsd | |
598 | * c-basic-offset: 4 | |
14d04a33 | 599 | * indent-tabs-mode: nil |
e9a8c099 MHM |
600 | * End: |
601 | * | |
14d04a33 | 602 | * ex: set ts=8 sts=4 sw=4 et: |
e9a8c099 | 603 | */ |