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 | ||
3cd96634 KW |
29 | #include "regcharclass.h" |
30 | #include "unicode_constants.h" | |
31 | ||
051a06d4 | 32 | /* For to_utf8_fold_flags, q.v. */ |
e4f4ef45 KW |
33 | #define FOLD_FLAGS_LOCALE 0x1 |
34 | #define FOLD_FLAGS_FULL 0x2 | |
35 | #define FOLD_FLAGS_NOMIX_ASCII 0x4 | |
051a06d4 | 36 | |
83199d38 KW |
37 | /* For _core_swash_init(), internal core use only */ |
38 | #define _CORE_SWASH_INIT_USER_DEFINED_PROPERTY 0x1 | |
5d3d13d1 | 39 | #define _CORE_SWASH_INIT_RETURN_IF_UNDEF 0x2 |
87367d5f | 40 | #define _CORE_SWASH_INIT_ACCEPT_INVLIST 0x4 |
83199d38 | 41 | |
7bbfa158 KW |
42 | /* |
43 | =head1 Unicode Support | |
8cca77bc KW |
44 | L<perlguts/Unicode Support> has an introduction to this API. |
45 | ||
46 | See also L</Character classification>, | |
47 | and L</Character case changing>. | |
48 | Various functions outside this section also work specially with Unicode. | |
49 | Search for the string "utf8" in this document. | |
7bbfa158 KW |
50 | |
51 | =for apidoc is_ascii_string | |
52 | ||
8871a094 | 53 | This is a misleadingly-named synonym for L</is_utf8_invariant_string>. |
7bbfa158 KW |
54 | On ASCII-ish platforms, the name isn't misleading: the ASCII-range characters |
55 | are exactly the UTF-8 invariants. But EBCDIC machines have more invariants | |
8871a094 KW |
56 | than just the ASCII characters, so C<is_utf8_invariant_string> is preferred. |
57 | ||
58 | =for apidoc is_invariant_string | |
59 | ||
60 | This is a somewhat misleadingly-named synonym for L</is_utf8_invariant_string>. | |
61 | C<is_utf8_invariant_string> is preferred, as it indicates under what conditions | |
62 | the string is invariant. | |
7bbfa158 KW |
63 | |
64 | =cut | |
65 | */ | |
8871a094 KW |
66 | #define is_ascii_string(s, len) is_utf8_invariant_string(s, len) |
67 | #define is_invariant_string(s, len) is_utf8_invariant_string(s, len) | |
7bbfa158 | 68 | |
de69f3af KW |
69 | #define uvchr_to_utf8(a,b) uvchr_to_utf8_flags(a,b,0) |
70 | #define uvchr_to_utf8_flags(d,uv,flags) \ | |
71 | uvoffuni_to_utf8_flags(d,NATIVE_TO_UNI(uv),flags) | |
72 | #define utf8_to_uvchr_buf(s, e, lenp) \ | |
842991ae | 73 | utf8n_to_uvchr(s, (U8*)(e) - (U8*)(s), lenp, \ |
de69f3af | 74 | ckWARN_d(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY) |
f9380377 KW |
75 | #define utf8n_to_uvchr(s, len, lenp, flags) \ |
76 | utf8n_to_uvchr_error(s, len, lenp, flags, 0) | |
de69f3af | 77 | |
a0270393 | 78 | #define to_uni_fold(c, p, lenp) _to_uni_fold_flags(c, p, lenp, FOLD_FLAGS_FULL) |
a239b1e2 KW |
79 | |
80 | #define to_utf8_fold(s, r, lenr) \ | |
607313a1 | 81 | _to_utf8_fold_flags (s, NULL, r, lenr, FOLD_FLAGS_FULL, __FILE__, __LINE__) |
a239b1e2 | 82 | #define to_utf8_lower(s, r, lenr) \ |
607313a1 | 83 | _to_utf8_lower_flags(s, NULL, r ,lenr, 0, __FILE__, __LINE__) |
a239b1e2 | 84 | #define to_utf8_upper(s, r, lenr) \ |
607313a1 | 85 | _to_utf8_upper_flags(s, NULL, r, lenr, 0, __FILE__, __LINE__) |
a239b1e2 | 86 | #define to_utf8_title(s, r, lenr) \ |
607313a1 | 87 | _to_utf8_title_flags(s, NULL, r, lenr ,0, __FILE__, __LINE__) |
36bb2ab6 | 88 | |
eda9cac1 KW |
89 | #define foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \ |
90 | foldEQ_utf8_flags(s1, pe1, l1, u1, s2, pe2, l2, u2, 0) | |
baa60164 | 91 | #define FOLDEQ_UTF8_NOMIX_ASCII (1 << 0) |
cea315b6 | 92 | #define FOLDEQ_LOCALE (1 << 1) |
18f762c3 KW |
93 | #define FOLDEQ_S1_ALREADY_FOLDED (1 << 2) |
94 | #define FOLDEQ_S2_ALREADY_FOLDED (1 << 3) | |
d635b710 KW |
95 | #define FOLDEQ_S1_FOLDS_SANE (1 << 4) |
96 | #define FOLDEQ_S2_FOLDS_SANE (1 << 5) | |
a33c29bc | 97 | |
e6226b18 KW |
98 | #define ibcmp_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \ |
99 | cBOOL(! foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2)) | |
100 | ||
1d72bdf6 NIS |
101 | #ifdef EBCDIC |
102 | /* The equivalent of these macros but implementing UTF-EBCDIC | |
103 | are in the following header file: | |
104 | */ | |
105 | ||
106 | #include "utfebcdic.h" | |
fd7cb289 | 107 | |
d06134e5 | 108 | #else /* ! EBCDIC */ |
73c4f7a1 GS |
109 | START_EXTERN_C |
110 | ||
111e8ed9 KW |
111 | /* How wide can a single UTF-8 encoded character become in bytes. */ |
112 | /* NOTE: Strictly speaking Perl's UTF-8 should not be called UTF-8 since UTF-8 | |
113 | * is an encoding of Unicode, and Unicode's upper limit, 0x10FFFF, can be | |
114 | * expressed with 4 bytes. However, Perl thinks of UTF-8 as a way to encode | |
115 | * non-negative integers in a binary format, even those above Unicode */ | |
116 | #define UTF8_MAXBYTES 13 | |
117 | ||
a0ed51b3 | 118 | #ifdef DOINIT |
6f06b55f | 119 | EXTCONST unsigned char PL_utf8skip[] = { |
b2635aa8 KW |
120 | /* 0x00 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */ |
121 | /* 0x10 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */ | |
122 | /* 0x20 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */ | |
123 | /* 0x30 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */ | |
124 | /* 0x40 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */ | |
125 | /* 0x50 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */ | |
126 | /* 0x60 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */ | |
127 | /* 0x70 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */ | |
128 | /* 0x80 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */ | |
129 | /* 0x90 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */ | |
130 | /* 0xA0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */ | |
131 | /* 0xB0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */ | |
132 | /* 0xC0 */ 2,2, /* overlong */ | |
1ff3baa2 | 133 | /* 0xC2 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0080 to U+03FF */ |
b2635aa8 KW |
134 | /* 0xD0 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0400 to U+07FF */ |
135 | /* 0xE0 */ 3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3, /* U+0800 to U+FFFF */ | |
136 | /* 0xF0 */ 4,4,4,4,4,4,4,4,5,5,5,5,6,6, /* above BMP to 2**31 - 1 */ | |
6937f885 KW |
137 | /* Perl extended (never was official UTF-8). Up to 36 bit */ |
138 | /* 0xFE */ 7, | |
139 | /* More extended, Up to 72 bits (64-bit + reserved) */ | |
111e8ed9 | 140 | /* 0xFF */ UTF8_MAXBYTES |
a0ed51b3 LW |
141 | }; |
142 | #else | |
6f06b55f | 143 | EXTCONST unsigned char PL_utf8skip[]; |
a0ed51b3 LW |
144 | #endif |
145 | ||
73c4f7a1 | 146 | END_EXTERN_C |
7e2040f0 | 147 | |
1a3756de | 148 | #if defined(_MSC_VER) && _MSC_VER < 1400 |
6f6d1bab TC |
149 | /* older MSVC versions have a smallish macro buffer */ |
150 | #define PERL_SMALL_MACRO_BUFFER | |
151 | #endif | |
152 | ||
59a449d5 KW |
153 | /* Native character to/from iso-8859-1. Are the identity functions on ASCII |
154 | * platforms */ | |
6f6d1bab | 155 | #ifdef PERL_SMALL_MACRO_BUFFER |
1a3756de TC |
156 | #define NATIVE_TO_LATIN1(ch) ((U8)(ch)) |
157 | #define LATIN1_TO_NATIVE(ch) ((U8)(ch)) | |
6f6d1bab | 158 | #else |
4c8cd605 KW |
159 | #define NATIVE_TO_LATIN1(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch))) |
160 | #define LATIN1_TO_NATIVE(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch))) | |
6f6d1bab | 161 | #endif |
59a449d5 KW |
162 | |
163 | /* I8 is an intermediate version of UTF-8 used only in UTF-EBCDIC. We thus | |
164 | * consider it to be identical to UTF-8 on ASCII platforms. Strictly speaking | |
165 | * UTF-8 and UTF-EBCDIC are two different things, but we often conflate them | |
166 | * because they are 8-bit encodings that serve the same purpose in Perl, and | |
167 | * rarely do we need to distinguish them. The term "NATIVE_UTF8" applies to | |
168 | * whichever one is applicable on the current platform */ | |
6f6d1bab | 169 | #ifdef PERL_SMALL_MACRO_BUFFER |
1d4ea287 DD |
170 | #define NATIVE_UTF8_TO_I8(ch) ((U8) (ch)) |
171 | #define I8_TO_NATIVE_UTF8(ch) ((U8) (ch)) | |
6f6d1bab | 172 | #else |
4c8cd605 KW |
173 | #define NATIVE_UTF8_TO_I8(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch))) |
174 | #define I8_TO_NATIVE_UTF8(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch))) | |
6f6d1bab | 175 | #endif |
59a449d5 | 176 | |
1d72bdf6 | 177 | /* Transforms in wide UV chars */ |
4c8cd605 KW |
178 | #define UNI_TO_NATIVE(ch) ((UV) (ch)) |
179 | #define NATIVE_TO_UNI(ch) ((UV) (ch)) | |
d7578b48 | 180 | |
877d9f0d | 181 | /* |
9041c2e3 | 182 | |
a14e0a36 KW |
183 | The following table is from Unicode 3.2, plus the Perl extensions for above |
184 | U+10FFFF | |
877d9f0d | 185 | |
a14e0a36 | 186 | Code Points 1st Byte 2nd Byte 3rd 4th 5th 6th 7th 8th-13th |
877d9f0d | 187 | |
375122d7 | 188 | U+0000..U+007F 00..7F |
e1b711da | 189 | U+0080..U+07FF * C2..DF 80..BF |
a14e0a36 KW |
190 | U+0800..U+0FFF E0 * A0..BF 80..BF |
191 | U+1000..U+CFFF E1..EC 80..BF 80..BF | |
192 | U+D000..U+D7FF ED 80..9F 80..BF | |
193 | U+D800..U+DFFF ED A0..BF 80..BF (surrogates) | |
194 | U+E000..U+FFFF EE..EF 80..BF 80..BF | |
195 | U+10000..U+3FFFF F0 * 90..BF 80..BF 80..BF | |
196 | U+40000..U+FFFFF F1..F3 80..BF 80..BF 80..BF | |
197 | U+100000..U+10FFFF F4 80..8F 80..BF 80..BF | |
198 | Below are above-Unicode code points | |
199 | U+110000..U+13FFFF F4 90..BF 80..BF 80..BF | |
200 | U+110000..U+1FFFFF F5..F7 80..BF 80..BF 80..BF | |
201 | U+200000..U+FFFFFF F8 * 88..BF 80..BF 80..BF 80..BF | |
202 | U+1000000..U+3FFFFFF F9..FB 80..BF 80..BF 80..BF 80..BF | |
203 | U+4000000..U+3FFFFFFF FC * 84..BF 80..BF 80..BF 80..BF 80..BF | |
204 | U+40000000..U+7FFFFFFF FD 80..BF 80..BF 80..BF 80..BF 80..BF | |
205 | U+80000000..U+FFFFFFFFF FE * 82..BF 80..BF 80..BF 80..BF 80..BF 80..BF | |
206 | U+1000000000.. FF 80..BF 80..BF 80..BF 80..BF 80..BF * 81..BF 80..BF | |
877d9f0d | 207 | |
e1b711da | 208 | Note the gaps before several of the byte entries above marked by '*'. These are |
37e2e78e KW |
209 | caused by legal UTF-8 avoiding non-shortest encodings: it is technically |
210 | possible to UTF-8-encode a single code point in different ways, but that is | |
211 | explicitly forbidden, and the shortest possible encoding should always be used | |
15824458 | 212 | (and that is what Perl does). The non-shortest ones are called 'overlongs'. |
8c007b5a | 213 | |
877d9f0d JH |
214 | */ |
215 | ||
8c007b5a JH |
216 | /* |
217 | Another way to look at it, as bits: | |
218 | ||
b2635aa8 | 219 | Code Points 1st Byte 2nd Byte 3rd Byte 4th Byte |
8c007b5a | 220 | |
b2635aa8 KW |
221 | 0aaa aaaa 0aaa aaaa |
222 | 0000 0bbb bbaa aaaa 110b bbbb 10aa aaaa | |
223 | cccc bbbb bbaa aaaa 1110 cccc 10bb bbbb 10aa aaaa | |
224 | 00 000d ddcc cccc bbbb bbaa aaaa 1111 0ddd 10cc cccc 10bb bbbb 10aa aaaa | |
8c007b5a JH |
225 | |
226 | As you can see, the continuation bytes all begin with C<10>, and the | |
e1b711da | 227 | leading bits of the start byte tell how many bytes there are in the |
8c007b5a JH |
228 | encoded character. |
229 | ||
df863e43 KW |
230 | Perl's extended UTF-8 means we can have start bytes up through FF, though any |
231 | beginning with FF yields a code point that is too large for 32-bit ASCII | |
232 | platforms. FF signals to use 13 bytes for the encoded character. This breaks | |
233 | the paradigm that the number of leading bits gives how many total bytes there | |
234 | are in the character. | |
65ab9279 | 235 | |
8c007b5a JH |
236 | */ |
237 | ||
eda91ad7 KW |
238 | /* Anything larger than this will overflow the word if it were converted into a UV */ |
239 | #if defined(UV_IS_QUAD) | |
240 | # define HIGHEST_REPRESENTABLE_UTF8 "\xFF\x80\x8F\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF" | |
241 | #else | |
242 | # define HIGHEST_REPRESENTABLE_UTF8 "\xFE\x83\xBF\xBF\xBF\xBF\xBF" | |
243 | #endif | |
244 | ||
6c88483e | 245 | /* Is the representation of the Unicode code point 'cp' the same regardless of |
15824458 | 246 | * being encoded in UTF-8 or not? */ |
2d1545e5 | 247 | #define OFFUNI_IS_INVARIANT(cp) isASCII(cp) |
15824458 | 248 | |
c2b32798 KW |
249 | /* |
250 | =for apidoc Am|bool|UVCHR_IS_INVARIANT|UV cp | |
251 | ||
252 | Evaluates to 1 if the representation of code point C<cp> is the same whether or | |
253 | not it is encoded in UTF-8; otherwise evaluates to 0. UTF-8 invariant | |
254 | characters can be copied as-is when converting to/from UTF-8, saving time. | |
255 | C<cp> is Unicode if above 255; otherwise is platform-native. | |
256 | ||
257 | =cut | |
258 | */ | |
259 | ||
cf1be84e | 260 | #define UVCHR_IS_INVARIANT(cp) OFFUNI_IS_INVARIANT(cp) |
38953e5a | 261 | |
c9264833 KW |
262 | /* This defines the bits that are to be in the continuation bytes of a multi-byte |
263 | * UTF-8 encoded character that mark it is a continuation byte. */ | |
264 | #define UTF_CONTINUATION_MARK 0x80 | |
265 | ||
a95ec4fb | 266 | /* Misleadingly named: is the UTF8-encoded byte 'c' part of a variant sequence |
5c06326b KW |
267 | * in UTF-8? This is the inverse of UTF8_IS_INVARIANT. The |0 makes sure this |
268 | * isn't mistakenly called with a ptr argument */ | |
269 | #define UTF8_IS_CONTINUED(c) (((U8)((c) | 0)) & UTF_CONTINUATION_MARK) | |
15824458 KW |
270 | |
271 | /* Is the byte 'c' the first byte of a multi-byte UTF8-8 encoded sequence? | |
272 | * This doesn't catch invariants (they are single-byte). It also excludes the | |
5c06326b KW |
273 | * illegal overlong sequences that begin with C0 and C1. The |0 makes sure |
274 | * this isn't mistakenly called with a ptr argument */ | |
275 | #define UTF8_IS_START(c) (((U8)((c) | 0)) >= 0xc2) | |
15824458 | 276 | |
858cd8ab KW |
277 | /* For use in UTF8_IS_CONTINUATION() below */ |
278 | #define UTF_IS_CONTINUATION_MASK 0xC0 | |
279 | ||
15824458 | 280 | /* Is the byte 'c' part of a multi-byte UTF8-8 encoded sequence, and not the |
5c06326b KW |
281 | * first byte thereof? The |0 makes sure this isn't mistakenly called with a |
282 | * ptr argument */ | |
858cd8ab | 283 | #define UTF8_IS_CONTINUATION(c) \ |
5c06326b | 284 | ((((U8)((c) | 0)) & UTF_IS_CONTINUATION_MASK) == UTF_CONTINUATION_MARK) |
0ae1fa71 | 285 | |
15824458 KW |
286 | /* Is the UTF8-encoded byte 'c' the first byte of a two byte sequence? Use |
287 | * UTF8_IS_NEXT_CHAR_DOWNGRADEABLE() instead if the input isn't known to | |
288 | * be well-formed. Masking with 0xfe allows the low bit to be 0 or 1; thus | |
5c06326b KW |
289 | * this matches 0xc[23]. The |0 makes sure this isn't mistakenly called with a |
290 | * ptr argument */ | |
291 | #define UTF8_IS_DOWNGRADEABLE_START(c) ((((U8)((c) | 0)) & 0xfe) == 0xc2) | |
4ab10950 | 292 | |
15824458 | 293 | /* Is the UTF8-encoded byte 'c' the first byte of a sequence of bytes that |
5c06326b KW |
294 | * represent a code point > 255? The |0 makes sure this isn't mistakenly |
295 | * called with a ptr argument */ | |
296 | #define UTF8_IS_ABOVE_LATIN1(c) (((U8)((c) | 0)) >= 0xc4) | |
8850bf83 | 297 | |
15824458 KW |
298 | /* This is the number of low-order bits a continuation byte in a UTF-8 encoded |
299 | * sequence contributes to the specification of the code point. In the bit | |
300 | * maps above, you see that the first 2 bits are a constant '10', leaving 6 of | |
301 | * real information */ | |
1d72bdf6 | 302 | #define UTF_ACCUMULATION_SHIFT 6 |
b2635aa8 | 303 | |
fed423a5 KW |
304 | /* ^? is defined to be DEL on ASCII systems. See the definition of toCTRL() |
305 | * for more */ | |
306 | #define QUESTION_MARK_CTRL DEL_NATIVE | |
307 | ||
308 | /* Surrogates, non-character code points and above-Unicode code points are | |
309 | * problematic in some contexts. This allows code that needs to check for | |
310 | * those to to quickly exclude the vast majority of code points it will | |
311 | * encounter */ | |
312 | #define isUTF8_POSSIBLY_PROBLEMATIC(c) ((U8) c >= 0xED) | |
313 | ||
784d4f31 KW |
314 | /* A helper macro for isUTF8_CHAR, so use that one instead of this. This was |
315 | * generated by regen/regcharclass.pl, and then moved here. Then it was | |
316 | * hand-edited to add some LIKELY() calls, presuming that malformations are | |
317 | * unlikely. The lines that generated it were then commented out. This was | |
318 | * done because it takes on the order of 10 minutes to generate, and is never | |
319 | * going to change, unless the generated code is improved, and figuring out | |
320 | * the LIKELYs there would be hard. | |
321 | * | |
322 | UTF8_CHAR: Matches legal UTF-8 variant code points up through 0x1FFFFFF | |
323 | ||
324 | 0x80 - 0x1FFFFF | |
325 | */ | |
326 | /*** GENERATED CODE ***/ | |
327 | #define is_UTF8_CHAR_utf8_no_length_checks(s) \ | |
9f2eed98 P |
328 | ( ( 0xC2 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xDF ) ? \ |
329 | ( LIKELY( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) ? 2 : 0 ) \ | |
330 | : ( 0xE0 == ((const U8*)s)[0] ) ? \ | |
331 | ( LIKELY( ( ( ((const U8*)s)[1] & 0xE0 ) == 0xA0 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\ | |
332 | : ( 0xE1 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xEF ) ? \ | |
333 | ( LIKELY( ( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\ | |
334 | : ( 0xF0 == ((const U8*)s)[0] ) ? \ | |
335 | ( LIKELY( ( ( 0x90 <= ((const U8*)s)[1] && ((const U8*)s)[1] <= 0xBF ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) && ( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\ | |
336 | : ( ( ( ( 0xF1 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xF7 ) && LIKELY( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) ) && LIKELY( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) && LIKELY( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 ) | |
784d4f31 KW |
337 | |
338 | /* The above macro handles UTF-8 that has this start byte as the maximum */ | |
339 | #define _IS_UTF8_CHAR_HIGHEST_START_BYTE 0xF7 | |
340 | ||
e23e8bc1 KW |
341 | /* A helper macro for isSTRICT_UTF8_CHAR, so use that one instead of this. |
342 | * Like is_UTF8_CHAR_utf8_no_length_checks(), this was moved here and LIKELYs | |
343 | * added manually. | |
344 | * | |
345 | STRICT_UTF8_CHAR: Matches legal Unicode UTF-8 variant code points, no | |
346 | surrrogates nor non-character code points | |
347 | */ | |
348 | /*** GENERATED CODE ***/ | |
349 | #define is_STRICT_UTF8_CHAR_utf8_no_length_checks(s) \ | |
9f2eed98 P |
350 | ( ( 0xC2 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xDF ) ? \ |
351 | ( LIKELY( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) ? 2 : 0 ) \ | |
352 | : ( 0xE0 == ((const U8*)s)[0] ) ? \ | |
353 | ( LIKELY( ( ( ((const U8*)s)[1] & 0xE0 ) == 0xA0 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\ | |
354 | : ( ( 0xE1 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xEC ) || 0xEE == ((const U8*)s)[0] ) ?\ | |
355 | ( ( ( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\ | |
356 | : ( 0xED == ((const U8*)s)[0] ) ? \ | |
357 | ( LIKELY( ( ( ((const U8*)s)[1] & 0xE0 ) == 0x80 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\ | |
358 | : ( 0xEF == ((const U8*)s)[0] ) ? \ | |
359 | ( ( ( 0x80 <= ((const U8*)s)[1] && ((const U8*)s)[1] <= 0xB6 ) || ( 0xB8 <= ((const U8*)s)[1] && ((const U8*)s)[1] <= 0xBE ) ) ?\ | |
360 | ( LIKELY( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ? 3 : 0 ) \ | |
361 | : ( 0xB7 == ((const U8*)s)[1] ) ? \ | |
362 | ( LIKELY( ( ((const U8*)s)[2] & 0xF0 ) == 0x80 || ( ((const U8*)s)[2] & 0xF0 ) == 0xB0 ) ? 3 : 0 )\ | |
363 | : ( ( 0xBF == ((const U8*)s)[1] ) && ( 0x80 <= ((const U8*)s)[2] && ((const U8*)s)[2] <= 0xBD ) ) ? 3 : 0 )\ | |
364 | : ( 0xF0 == ((const U8*)s)[0] ) ? \ | |
365 | ( ( ( 0x90 <= ((const U8*)s)[1] && ((const U8*)s)[1] <= 0x9E ) || ( 0xA0 <= ((const U8*)s)[1] && ((const U8*)s)[1] <= 0xAE ) || ( 0xB0 <= ((const U8*)s)[1] && ((const U8*)s)[1] <= 0xBE ) ) ?\ | |
366 | ( LIKELY( ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\ | |
367 | : ( ((const U8*)s)[1] == 0x9F || ( ( ((const U8*)s)[1] & 0xEF ) == 0xAF ) ) ? \ | |
368 | ( ( 0x80 <= ((const U8*)s)[2] && ((const U8*)s)[2] <= 0xBE ) ? \ | |
369 | ( LIKELY( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ? 4 : 0 ) \ | |
370 | : LIKELY( ( 0xBF == ((const U8*)s)[2] ) && ( 0x80 <= ((const U8*)s)[3] && ((const U8*)s)[3] <= 0xBD ) ) ? 4 : 0 )\ | |
e23e8bc1 | 371 | : 0 ) \ |
9f2eed98 P |
372 | : ( 0xF1 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xF3 ) ? \ |
373 | ( ( ( ( ((const U8*)s)[1] & 0xC8 ) == 0x80 ) || ( ( ((const U8*)s)[1] & 0xCC ) == 0x88 ) || ( ( ((const U8*)s)[1] & 0xCE ) == 0x8C ) || ( ( ((const U8*)s)[1] & 0xCF ) == 0x8E ) ) ?\ | |
374 | ( LIKELY( ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\ | |
375 | : ( ( ((const U8*)s)[1] & 0xCF ) == 0x8F ) ? \ | |
376 | ( ( 0x80 <= ((const U8*)s)[2] && ((const U8*)s)[2] <= 0xBE ) ? \ | |
377 | ( LIKELY( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ? 4 : 0 ) \ | |
378 | : LIKELY( ( 0xBF == ((const U8*)s)[2] ) && ( 0x80 <= ((const U8*)s)[3] && ((const U8*)s)[3] <= 0xBD ) ) ? 4 : 0 )\ | |
e23e8bc1 | 379 | : 0 ) \ |
9f2eed98 P |
380 | : ( 0xF4 == ((const U8*)s)[0] ) ? \ |
381 | ( ( 0x80 <= ((const U8*)s)[1] && ((const U8*)s)[1] <= 0x8E ) ? \ | |
382 | ( LIKELY( ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\ | |
383 | : ( 0x8F == ((const U8*)s)[1] ) ? \ | |
384 | ( ( 0x80 <= ((const U8*)s)[2] && ((const U8*)s)[2] <= 0xBE ) ? \ | |
385 | ( LIKELY( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ? 4 : 0 ) \ | |
386 | : LIKELY( ( 0xBF == ((const U8*)s)[2] ) && ( 0x80 <= ((const U8*)s)[3] && ((const U8*)s)[3] <= 0xBD ) ) ? 4 : 0 )\ | |
e23e8bc1 KW |
387 | : 0 ) \ |
388 | : 0 ) | |
389 | ||
a82be82b KW |
390 | /* Similarly, |
391 | C9_STRICT_UTF8_CHAR: Matches legal Unicode UTF-8 variant code | |
392 | points, no surrogates | |
393 | 0x0080 - 0xD7FF | |
394 | 0xE000 - 0x10FFFF | |
395 | */ | |
396 | /*** GENERATED CODE ***/ | |
397 | #define is_C9_STRICT_UTF8_CHAR_utf8_no_length_checks(s) \ | |
9f2eed98 P |
398 | ( ( 0xC2 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xDF ) ? \ |
399 | ( LIKELY( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) ? 2 : 0 ) \ | |
400 | : ( 0xE0 == ((const U8*)s)[0] ) ? \ | |
401 | ( LIKELY( ( ( ((const U8*)s)[1] & 0xE0 ) == 0xA0 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\ | |
402 | : ( ( 0xE1 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xEC ) || ( ((const U8*)s)[0] & 0xFE ) == 0xEE ) ?\ | |
403 | ( LIKELY( ( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\ | |
404 | : ( 0xED == ((const U8*)s)[0] ) ? \ | |
405 | ( LIKELY( ( ( ((const U8*)s)[1] & 0xE0 ) == 0x80 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\ | |
406 | : ( 0xF0 == ((const U8*)s)[0] ) ? \ | |
407 | ( LIKELY( ( ( 0x90 <= ((const U8*)s)[1] && ((const U8*)s)[1] <= 0xBF ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) && ( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\ | |
408 | : ( 0xF1 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xF3 ) ? \ | |
409 | ( LIKELY( ( ( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) && ( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\ | |
410 | : LIKELY( ( ( ( 0xF4 == ((const U8*)s)[0] ) && ( ( ((const U8*)s)[1] & 0xF0 ) == 0x80 ) ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) && ( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 ) | |
a82be82b | 411 | |
fed423a5 KW |
412 | #endif /* EBCDIC vs ASCII */ |
413 | ||
414 | /* 2**UTF_ACCUMULATION_SHIFT - 1 */ | |
415 | #define UTF_CONTINUATION_MASK ((U8) ((1U << UTF_ACCUMULATION_SHIFT) - 1)) | |
416 | ||
7028aeba KW |
417 | /* Internal macro to be used only in this file to aid in constructing other |
418 | * publicly accessible macros. | |
419 | * The number of bytes required to express this uv in UTF-8, for just those | |
420 | * uv's requiring 2 through 6 bytes, as these are common to all platforms and | |
421 | * word sizes. The number of bytes needed is given by the number of leading 1 | |
422 | * bits in the start byte. There are 32 start bytes that have 2 initial 1 bits | |
423 | * (C0-DF); there are 16 that have 3 initial 1 bits (E0-EF); 8 that have 4 | |
424 | * initial 1 bits (F0-F8); 4 that have 5 initial 1 bits (F9-FB), and 2 that | |
425 | * have 6 initial 1 bits (FC-FD). The largest number a string of n bytes can | |
426 | * represent is (the number of possible start bytes for 'n') | |
427 | * * (the number of possiblities for each start byte | |
428 | * The latter in turn is | |
429 | * 2 ** ( (how many continuation bytes there are) | |
430 | * * (the number of bits of information each | |
431 | * continuation byte holds)) | |
432 | * | |
433 | * If we were on a platform where we could use a fast find first set bit | |
434 | * instruction (or count leading zeros instruction) this could be replaced by | |
435 | * using that to find the log2 of the uv, and divide that by the number of bits | |
436 | * of information in each continuation byte, adjusting for large cases and how | |
437 | * much information is in a start byte for that length */ | |
72164d3a | 438 | #define __COMMON_UNI_SKIP(uv) \ |
7028aeba KW |
439 | (UV) (uv) < (32 * (1U << ( UTF_ACCUMULATION_SHIFT))) ? 2 : \ |
440 | (UV) (uv) < (16 * (1U << (2 * UTF_ACCUMULATION_SHIFT))) ? 3 : \ | |
441 | (UV) (uv) < ( 8 * (1U << (3 * UTF_ACCUMULATION_SHIFT))) ? 4 : \ | |
442 | (UV) (uv) < ( 4 * (1U << (4 * UTF_ACCUMULATION_SHIFT))) ? 5 : \ | |
443 | (UV) (uv) < ( 2 * (1U << (5 * UTF_ACCUMULATION_SHIFT))) ? 6 : | |
72164d3a KW |
444 | |
445 | /* Internal macro to be used only in this file. | |
446 | * This adds to __COMMON_UNI_SKIP the details at this platform's upper range. | |
fed423a5 | 447 | * For any-sized EBCDIC platforms, or 64-bit ASCII ones, we need one more test |
72164d3a KW |
448 | * to see if just 7 bytes is needed, or if the maximum is needed. For 32-bit |
449 | * ASCII platforms, everything is representable by 7 bytes */ | |
fed423a5 | 450 | #if defined(UV_IS_QUAD) || defined(EBCDIC) |
72164d3a | 451 | # define __BASE_UNI_SKIP(uv) (__COMMON_UNI_SKIP(uv) \ |
7028aeba | 452 | (UV) (uv) < ((UV) 1U << (6 * UTF_ACCUMULATION_SHIFT)) ? 7 : UTF8_MAXBYTES) |
1d68d6cd | 453 | #else |
72164d3a | 454 | # define __BASE_UNI_SKIP(uv) (__COMMON_UNI_SKIP(uv) 7) |
1d68d6cd SC |
455 | #endif |
456 | ||
fed423a5 KW |
457 | /* The next two macros use the base macro defined above, and add in the tests |
458 | * at the low-end of the range, for just 1 byte, yielding complete macros, | |
459 | * publicly accessible. */ | |
460 | ||
461 | /* Input is a true Unicode (not-native) code point */ | |
462 | #define OFFUNISKIP(uv) (OFFUNI_IS_INVARIANT(uv) ? 1 : __BASE_UNI_SKIP(uv)) | |
2084b489 | 463 | |
5352a763 KW |
464 | /* |
465 | ||
466 | =for apidoc Am|STRLEN|UVCHR_SKIP|UV cp | |
467 | returns the number of bytes required to represent the code point C<cp> when | |
468 | encoded as UTF-8. C<cp> is a native (ASCII or EBCDIC) code point if less than | |
469 | 255; a Unicode code point otherwise. | |
470 | ||
471 | =cut | |
472 | */ | |
fdb6583d | 473 | #define UVCHR_SKIP(uv) ( UVCHR_IS_INVARIANT(uv) ? 1 : __BASE_UNI_SKIP(uv)) |
5352a763 | 474 | |
b651802e KW |
475 | /* The largest code point representable by two UTF-8 bytes on this platform. |
476 | * As explained in the comments for __COMMON_UNI_SKIP, 32 start bytes with | |
fed423a5 | 477 | * UTF_ACCUMULATION_SHIFT bits of information each */ |
aa206fb7 KW |
478 | #define MAX_UTF8_TWO_BYTE (32 * (1U << UTF_ACCUMULATION_SHIFT) - 1) |
479 | ||
b651802e KW |
480 | /* The largest code point representable by two UTF-8 bytes on any platform that |
481 | * Perl runs on. This value is constrained by EBCDIC which has 5 bits per | |
482 | * continuation byte */ | |
aa206fb7 KW |
483 | #define MAX_PORTABLE_UTF8_TWO_BYTE (32 * (1U << 5) - 1) |
484 | ||
c03c0950 KW |
485 | /* The maximum number of UTF-8 bytes a single Unicode character can |
486 | * uppercase/lowercase/fold into. Unicode guarantees that the maximum | |
487 | * expansion is UTF8_MAX_FOLD_CHAR_EXPAND characters, but any above-Unicode | |
488 | * code point will fold to itself, so we only have to look at the expansion of | |
489 | * the maximum Unicode code point. But this number may be less than the space | |
490 | * occupied by a very large code point under Perl's extended UTF-8. We have to | |
491 | * make it large enough to fit any single character. (It turns out that ASCII | |
492 | * and EBCDIC differ in which is larger) */ | |
493 | #define UTF8_MAXBYTES_CASE \ | |
494 | (UTF8_MAXBYTES >= (UTF8_MAX_FOLD_CHAR_EXPAND * OFFUNISKIP(0x10FFFF)) \ | |
495 | ? UTF8_MAXBYTES \ | |
496 | : (UTF8_MAX_FOLD_CHAR_EXPAND * OFFUNISKIP(0x10FFFF))) | |
497 | ||
d06134e5 KW |
498 | /* Rest of these are attributes of Unicode and perl's internals rather than the |
499 | * encoding, or happen to be the same in both ASCII and EBCDIC (at least at | |
500 | * this level; the macros that some of these call may have different | |
501 | * definitions in the two encodings */ | |
502 | ||
59a449d5 KW |
503 | /* In domain restricted to ASCII, these may make more sense to the reader than |
504 | * the ones with Latin1 in the name */ | |
505 | #define NATIVE_TO_ASCII(ch) NATIVE_TO_LATIN1(ch) | |
506 | #define ASCII_TO_NATIVE(ch) LATIN1_TO_NATIVE(ch) | |
507 | ||
508 | /* More or less misleadingly-named defines, retained for back compat */ | |
509 | #define NATIVE_TO_UTF(ch) NATIVE_UTF8_TO_I8(ch) | |
510 | #define NATIVE_TO_I8(ch) NATIVE_UTF8_TO_I8(ch) | |
511 | #define UTF_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch) | |
512 | #define I8_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch) | |
513 | #define NATIVE8_TO_UNI(ch) NATIVE_TO_LATIN1(ch) | |
d06134e5 | 514 | |
c0236afe | 515 | /* This defines the 1-bits that are to be in the first byte of a multi-byte |
97d0ceda KW |
516 | * UTF-8 encoded character that mark it as a start byte and give the number of |
517 | * bytes that comprise the character. 'len' is the number of bytes in the | |
518 | * multi-byte sequence. */ | |
c0236afe KW |
519 | #define UTF_START_MARK(len) (((len) > 7) ? 0xFF : (0xFF & (0xFE << (7-(len))))) |
520 | ||
521 | /* Masks out the initial one bits in a start byte, leaving the real data ones. | |
522 | * Doesn't work on an invariant byte. 'len' is the number of bytes in the | |
523 | * multi-byte sequence that comprises the character. */ | |
524 | #define UTF_START_MASK(len) (((len) >= 7) ? 0x00 : (0x1F >> ((len)-2))) | |
525 | ||
537124e4 KW |
526 | /* Adds a UTF8 continuation byte 'new' of information to a running total code |
527 | * point 'old' of all the continuation bytes so far. This is designed to be | |
155d2738 KW |
528 | * used in a loop to convert from UTF-8 to the code point represented. Note |
529 | * that this is asymmetric on EBCDIC platforms, in that the 'new' parameter is | |
530 | * the UTF-EBCDIC byte, whereas the 'old' parameter is a Unicode (not EBCDIC) | |
531 | * code point in process of being generated */ | |
532 | #define UTF8_ACCUMULATE(old, new) (((old) << UTF_ACCUMULATION_SHIFT) \ | |
533 | | ((NATIVE_UTF8_TO_I8((U8)new)) \ | |
534 | & UTF_CONTINUATION_MASK)) | |
d06134e5 | 535 | |
4ab10950 KW |
536 | /* This works in the face of malformed UTF-8. */ |
537 | #define UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(s, e) (UTF8_IS_DOWNGRADEABLE_START(*s) \ | |
538 | && ( (e) - (s) > 1) \ | |
539 | && UTF8_IS_CONTINUATION(*((s)+1))) | |
540 | ||
5aaebcb3 | 541 | /* Number of bytes a code point occupies in UTF-8. */ |
5352a763 | 542 | #define NATIVE_SKIP(uv) UVCHR_SKIP(uv) |
bd18bd40 | 543 | |
5aaebcb3 KW |
544 | /* Most code which says UNISKIP is really thinking in terms of native code |
545 | * points (0-255) plus all those beyond. This is an imprecise term, but having | |
2accb712 | 546 | * it means existing code continues to work. For precision, use UVCHR_SKIP, |
5352a763 KW |
547 | * NATIVE_SKIP, or OFFUNISKIP */ |
548 | #define UNISKIP(uv) UVCHR_SKIP(uv) | |
5aaebcb3 | 549 | |
3c0792e4 KW |
550 | /* Longer, but more accurate name */ |
551 | #define UTF8_IS_ABOVE_LATIN1_START(c) UTF8_IS_ABOVE_LATIN1(c) | |
552 | ||
a62b247b KW |
553 | /* Convert a UTF-8 variant Latin1 character to a native code point value. |
554 | * Needs just one iteration of accumulate. Should be used only if it is known | |
555 | * that the code point is < 256, and is not UTF-8 invariant. Use the slower | |
556 | * but more general TWO_BYTE_UTF8_TO_NATIVE() which handles any code point | |
557 | * representable by two bytes (which turns out to be up through | |
558 | * MAX_PORTABLE_UTF8_TWO_BYTE). The two parameters are: | |
559 | * HI: a downgradable start byte; | |
560 | * LO: continuation. | |
561 | * */ | |
562 | #define EIGHT_BIT_UTF8_TO_NATIVE(HI, LO) \ | |
563 | ( __ASSERT_(UTF8_IS_DOWNGRADEABLE_START(HI)) \ | |
564 | __ASSERT_(UTF8_IS_CONTINUATION(LO)) \ | |
565 | LATIN1_TO_NATIVE(UTF8_ACCUMULATE(( \ | |
566 | NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), (LO)))) | |
567 | ||
94bb8c36 | 568 | /* Convert a two (not one) byte utf8 character to a native code point value. |
2950f2a7 KW |
569 | * Needs just one iteration of accumulate. Should not be used unless it is |
570 | * known that the two bytes are legal: 1) two-byte start, and 2) continuation. | |
571 | * Note that the result can be larger than 255 if the input character is not | |
572 | * downgradable */ | |
94bb8c36 | 573 | #define TWO_BYTE_UTF8_TO_NATIVE(HI, LO) \ |
b18b15ad | 574 | ( __ASSERT_(PL_utf8skip[HI] == 2) \ |
635e76f5 | 575 | __ASSERT_(UTF8_IS_CONTINUATION(LO)) \ |
94bb8c36 | 576 | UNI_TO_NATIVE(UTF8_ACCUMULATE((NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), \ |
635e76f5 | 577 | (LO)))) |
94bb8c36 KW |
578 | |
579 | /* Should never be used, and be deprecated */ | |
580 | #define TWO_BYTE_UTF8_TO_UNI(HI, LO) NATIVE_TO_UNI(TWO_BYTE_UTF8_TO_NATIVE(HI, LO)) | |
2950f2a7 | 581 | |
bd18bd40 KW |
582 | /* |
583 | ||
584 | =for apidoc Am|STRLEN|UTF8SKIP|char* s | |
585 | returns the number of bytes in the UTF-8 encoded character whose first (perhaps | |
586 | only) byte is pointed to by C<s>. | |
587 | ||
588 | =cut | |
589 | */ | |
2a70536e KW |
590 | #define UTF8SKIP(s) PL_utf8skip[*(const U8*)(s)] |
591 | #define UTF8_SKIP(s) UTF8SKIP(s) | |
d06134e5 | 592 | |
2d1545e5 KW |
593 | /* Most code that says 'UNI_' really means the native value for code points up |
594 | * through 255 */ | |
595 | #define UNI_IS_INVARIANT(cp) UVCHR_IS_INVARIANT(cp) | |
596 | ||
c2b32798 KW |
597 | /* |
598 | =for apidoc Am|bool|UTF8_IS_INVARIANT|char c | |
599 | ||
600 | Evaluates to 1 if the byte C<c> represents the same character when encoded in | |
601 | UTF-8 as when not; otherwise evaluates to 0. UTF-8 invariant characters can be | |
602 | copied as-is when converting to/from UTF-8, saving time. | |
603 | ||
604 | In spite of the name, this macro gives the correct result if the input string | |
605 | from which C<c> comes is not encoded in UTF-8. | |
606 | ||
607 | See C<L</UVCHR_IS_INVARIANT>> for checking if a UV is invariant. | |
608 | ||
609 | =cut | |
610 | ||
611 | The reason it works on both UTF-8 encoded strings and non-UTF-8 encoded, is | |
612 | that it returns TRUE in each for the exact same set of bit patterns. It is | |
613 | valid on a subset of what UVCHR_IS_INVARIANT is valid on, so can just use that; | |
614 | and the compiler should optimize out anything extraneous given the | |
615 | implementation of the latter. The |0 makes sure this isn't mistakenly called | |
616 | with a ptr argument. | |
617 | */ | |
5c06326b | 618 | #define UTF8_IS_INVARIANT(c) UVCHR_IS_INVARIANT((c) | 0) |
5fc230f1 KW |
619 | |
620 | /* Like the above, but its name implies a non-UTF8 input, which as the comments | |
621 | * above show, doesn't matter as to its implementation */ | |
38953e5a | 622 | #define NATIVE_BYTE_IS_INVARIANT(c) UVCHR_IS_INVARIANT(c) |
d06134e5 | 623 | |
48ccf5e1 KW |
624 | /* The macros in the next 4 sets are used to generate the two utf8 or utfebcdic |
625 | * bytes from an ordinal that is known to fit into exactly two (not one) bytes; | |
626 | * it must be less than 0x3FF to work across both encodings. */ | |
627 | ||
628 | /* These two are helper macros for the other three sets, and should not be used | |
629 | * directly anywhere else. 'translate_function' is either NATIVE_TO_LATIN1 | |
1ff3baa2 KW |
630 | * (which works for code points up through 0xFF) or NATIVE_TO_UNI which works |
631 | * for any code point */ | |
48ccf5e1 | 632 | #define __BASE_TWO_BYTE_HI(c, translate_function) \ |
2863dafa | 633 | (__ASSERT_(! UVCHR_IS_INVARIANT(c)) \ |
48ccf5e1 | 634 | I8_TO_NATIVE_UTF8((translate_function(c) >> UTF_ACCUMULATION_SHIFT) \ |
2863dafa | 635 | | UTF_START_MARK(2))) |
48ccf5e1 | 636 | #define __BASE_TWO_BYTE_LO(c, translate_function) \ |
2863dafa | 637 | (__ASSERT_(! UVCHR_IS_INVARIANT(c)) \ |
48ccf5e1 | 638 | I8_TO_NATIVE_UTF8((translate_function(c) & UTF_CONTINUATION_MASK) \ |
2863dafa | 639 | | UTF_CONTINUATION_MARK)) |
48ccf5e1 | 640 | |
48ccf5e1 KW |
641 | /* The next two macros should not be used. They were designed to be usable as |
642 | * the case label of a switch statement, but this doesn't work for EBCDIC. Use | |
9d0d3a03 | 643 | * regen/unicode_constants.pl instead */ |
48ccf5e1 KW |
644 | #define UTF8_TWO_BYTE_HI_nocast(c) __BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI) |
645 | #define UTF8_TWO_BYTE_LO_nocast(c) __BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI) | |
646 | ||
647 | /* The next two macros are used when the source should be a single byte | |
648 | * character; checked for under DEBUGGING */ | |
649 | #define UTF8_EIGHT_BIT_HI(c) (__ASSERT_(FITS_IN_8_BITS(c)) \ | |
4c8cd605 | 650 | ( __BASE_TWO_BYTE_HI(c, NATIVE_TO_LATIN1))) |
48ccf5e1 | 651 | #define UTF8_EIGHT_BIT_LO(c) (__ASSERT_(FITS_IN_8_BITS(c)) \ |
4c8cd605 | 652 | (__BASE_TWO_BYTE_LO(c, NATIVE_TO_LATIN1))) |
48ccf5e1 KW |
653 | |
654 | /* These final two macros in the series are used when the source can be any | |
655 | * code point whose UTF-8 is known to occupy 2 bytes; they are less efficient | |
656 | * than the EIGHT_BIT versions on EBCDIC platforms. We use the logical '~' | |
657 | * operator instead of "<=" to avoid getting compiler warnings. | |
d52b8576 | 658 | * MAX_UTF8_TWO_BYTE should be exactly all one bits in the lower few |
48ccf5e1 KW |
659 | * places, so the ~ works */ |
660 | #define UTF8_TWO_BYTE_HI(c) \ | |
661 | (__ASSERT_((sizeof(c) == 1) \ | |
d52b8576 | 662 | || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \ |
4c8cd605 | 663 | (__BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI))) |
48ccf5e1 KW |
664 | #define UTF8_TWO_BYTE_LO(c) \ |
665 | (__ASSERT_((sizeof(c) == 1) \ | |
d52b8576 | 666 | || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \ |
4c8cd605 | 667 | (__BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI))) |
d06134e5 | 668 | |
e7214ce8 KW |
669 | /* This is illegal in any well-formed UTF-8 in both EBCDIC and ASCII |
670 | * as it is only in overlongs. */ | |
671 | #define ILLEGAL_UTF8_BYTE I8_TO_NATIVE_UTF8(0xC1) | |
672 | ||
7e2040f0 | 673 | /* |
e3036cf4 | 674 | * 'UTF' is whether or not p is encoded in UTF8. The names 'foo_lazy_if' stem |
20df05f4 KW |
675 | * from an earlier version of these macros in which they didn't call the |
676 | * foo_utf8() macros (i.e. were 'lazy') unless they decided that *p is the | |
677 | * beginning of a utf8 character. Now that foo_utf8() determines that itself, | |
678 | * no need to do it again here | |
7e2040f0 | 679 | */ |
34aeb2e9 KW |
680 | #define isIDFIRST_lazy_if(p,UTF) \ |
681 | _is_utf8_FOO(_CC_IDFIRST, (const U8 *) p, "isIDFIRST_lazy_if", \ | |
682 | "isIDFIRST_lazy_if_safe", \ | |
683 | cBOOL(UTF && ! IN_BYTES), 0, __FILE__,__LINE__) | |
1d72bdf6 | 684 | |
da8c1a98 KW |
685 | #define isIDFIRST_lazy_if_safe(p, e, UTF) \ |
686 | ((IN_BYTES || !UTF) \ | |
687 | ? isIDFIRST(*(p)) \ | |
688 | : isIDFIRST_utf8_safe(p, e)) | |
689 | ||
34aeb2e9 KW |
690 | #define isWORDCHAR_lazy_if(p,UTF) \ |
691 | _is_utf8_FOO(_CC_IDFIRST, (const U8 *) p, "isWORDCHAR_lazy_if", \ | |
692 | "isWORDCHAR_lazy_if_safe", \ | |
693 | cBOOL(UTF && ! IN_BYTES), 0, __FILE__,__LINE__) | |
694 | ||
da8c1a98 KW |
695 | #define isWORDCHAR_lazy_if_safe(p, e, UTF) \ |
696 | ((IN_BYTES || !UTF) \ | |
697 | ? isWORDCHAR(*(p)) \ | |
698 | : isWORDCHAR_utf8_safe((U8 *) p, (U8 *) e)) | |
699 | ||
34aeb2e9 KW |
700 | #define isALNUM_lazy_if(p,UTF) \ |
701 | _is_utf8_FOO(_CC_IDFIRST, (const U8 *) p, "isALNUM_lazy_if", \ | |
702 | "isWORDCHAR_lazy_if_safe", \ | |
703 | cBOOL(UTF && ! IN_BYTES), 0, __FILE__,__LINE__) | |
da8c1a98 | 704 | |
89ebb4a3 JH |
705 | #define UTF8_MAXLEN UTF8_MAXBYTES |
706 | ||
8cb75cc8 KW |
707 | /* A Unicode character can fold to up to 3 characters */ |
708 | #define UTF8_MAX_FOLD_CHAR_EXPAND 3 | |
709 | ||
a98fe34d | 710 | #define IN_BYTES (CopHINTS_get(PL_curcop) & HINT_BYTES) |
bd18bd40 KW |
711 | |
712 | /* | |
713 | ||
714 | =for apidoc Am|bool|DO_UTF8|SV* sv | |
715 | Returns a bool giving whether or not the PV in C<sv> is to be treated as being | |
716 | encoded in UTF-8. | |
717 | ||
718 | You should use this I<after> a call to C<SvPV()> or one of its variants, in | |
719 | case any call to string overloading updates the internal UTF-8 encoding flag. | |
720 | ||
721 | =cut | |
722 | */ | |
0064a8a9 | 723 | #define DO_UTF8(sv) (SvUTF8(sv) && !IN_BYTES) |
1ff3baa2 KW |
724 | |
725 | /* Should all strings be treated as Unicode, and not just UTF-8 encoded ones? | |
726 | * Is so within 'feature unicode_strings' or 'locale :not_characters', and not | |
727 | * within 'use bytes'. UTF-8 locales are not tested for here, but perhaps | |
728 | * could be */ | |
729 | #define IN_UNI_8_BIT \ | |
d6ded950 KW |
730 | (((CopHINTS_get(PL_curcop) & (HINT_UNI_8_BIT)) \ |
731 | || (CopHINTS_get(PL_curcop) & HINT_LOCALE_PARTIAL \ | |
732 | /* -1 below is for :not_characters */ \ | |
733 | && _is_in_locale_category(FALSE, -1))) \ | |
734 | && ! IN_BYTES) | |
b36bf33f | 735 | |
1d72bdf6 | 736 | |
c76687c5 | 737 | #define UTF8_ALLOW_EMPTY 0x0001 /* Allow a zero length string */ |
2b5e7bc2 | 738 | #define UTF8_GOT_EMPTY UTF8_ALLOW_EMPTY |
c76687c5 KW |
739 | |
740 | /* Allow first byte to be a continuation byte */ | |
1d72bdf6 | 741 | #define UTF8_ALLOW_CONTINUATION 0x0002 |
2b5e7bc2 | 742 | #define UTF8_GOT_CONTINUATION UTF8_ALLOW_CONTINUATION |
c76687c5 | 743 | |
2b5e7bc2 | 744 | /* Unexpected continuation byte */ |
1d72bdf6 | 745 | #define UTF8_ALLOW_NON_CONTINUATION 0x0004 |
2b5e7bc2 | 746 | #define UTF8_GOT_NON_CONTINUATION UTF8_ALLOW_NON_CONTINUATION |
949cf498 KW |
747 | |
748 | /* expecting more bytes than were available in the string */ | |
749 | #define UTF8_ALLOW_SHORT 0x0008 | |
2b5e7bc2 | 750 | #define UTF8_GOT_SHORT UTF8_ALLOW_SHORT |
949cf498 | 751 | |
94953955 KW |
752 | /* Overlong sequence; i.e., the code point can be specified in fewer bytes. |
753 | * First one will convert the overlong to the REPLACEMENT CHARACTER; second | |
754 | * will return what the overlong evaluates to */ | |
949cf498 | 755 | #define UTF8_ALLOW_LONG 0x0010 |
94953955 | 756 | #define UTF8_ALLOW_LONG_AND_ITS_VALUE (UTF8_ALLOW_LONG|0x0020) |
2b5e7bc2 KW |
757 | #define UTF8_GOT_LONG UTF8_ALLOW_LONG |
758 | ||
d60baaa7 KW |
759 | #define UTF8_ALLOW_OVERFLOW 0x0080 |
760 | #define UTF8_GOT_OVERFLOW UTF8_ALLOW_OVERFLOW | |
949cf498 | 761 | |
f180b292 | 762 | #define UTF8_DISALLOW_SURROGATE 0x0100 /* Unicode surrogates */ |
2b5e7bc2 | 763 | #define UTF8_GOT_SURROGATE UTF8_DISALLOW_SURROGATE |
f180b292 | 764 | #define UTF8_WARN_SURROGATE 0x0200 |
949cf498 | 765 | |
f180b292 | 766 | #define UTF8_DISALLOW_NONCHAR 0x0400 /* Unicode non-character */ |
2b5e7bc2 | 767 | #define UTF8_GOT_NONCHAR UTF8_DISALLOW_NONCHAR |
f180b292 | 768 | #define UTF8_WARN_NONCHAR 0x0800 /* code points */ |
949cf498 | 769 | |
f180b292 | 770 | #define UTF8_DISALLOW_SUPER 0x1000 /* Super-set of Unicode: code */ |
2b5e7bc2 | 771 | #define UTF8_GOT_SUPER UTF8_DISALLOW_SUPER |
f180b292 | 772 | #define UTF8_WARN_SUPER 0x2000 /* points above the legal max */ |
949cf498 | 773 | |
d35f2ca5 KW |
774 | /* Code points which never were part of the original UTF-8 standard, which only |
775 | * went up to 2 ** 31 - 1. Note that these all overflow a signed 32-bit word, | |
776 | * The first byte of these code points is FE or FF on ASCII platforms. If the | |
777 | * first byte is FF, it will overflow a 32-bit word. */ | |
f180b292 | 778 | #define UTF8_DISALLOW_ABOVE_31_BIT 0x4000 |
2b5e7bc2 | 779 | #define UTF8_GOT_ABOVE_31_BIT UTF8_DISALLOW_ABOVE_31_BIT |
f180b292 | 780 | #define UTF8_WARN_ABOVE_31_BIT 0x8000 |
d35f2ca5 KW |
781 | |
782 | /* For back compat, these old names are misleading for UTF_EBCDIC */ | |
783 | #define UTF8_DISALLOW_FE_FF UTF8_DISALLOW_ABOVE_31_BIT | |
784 | #define UTF8_WARN_FE_FF UTF8_WARN_ABOVE_31_BIT | |
949cf498 | 785 | |
f180b292 | 786 | #define UTF8_CHECK_ONLY 0x10000 |
99a765e9 | 787 | #define _UTF8_NO_CONFIDENCE_IN_CURLEN 0x20000 /* Internal core use only */ |
949cf498 KW |
788 | |
789 | /* For backwards source compatibility. They do nothing, as the default now | |
790 | * includes what they used to mean. The first one's meaning was to allow the | |
791 | * just the single non-character 0xFFFF */ | |
792 | #define UTF8_ALLOW_FFFF 0 | |
793 | #define UTF8_ALLOW_SURROGATE 0 | |
794 | ||
ecc1615f KW |
795 | /* C9 refers to Unicode Corrigendum #9: allows but discourages non-chars */ |
796 | #define UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE \ | |
797 | (UTF8_DISALLOW_SUPER|UTF8_DISALLOW_SURROGATE) | |
798 | #define UTF8_WARN_ILLEGAL_C9_INTERCHANGE (UTF8_WARN_SUPER|UTF8_WARN_SURROGATE) | |
799 | ||
d35f2ca5 | 800 | #define UTF8_DISALLOW_ILLEGAL_INTERCHANGE \ |
ecc1615f | 801 | (UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE|UTF8_DISALLOW_NONCHAR) |
949cf498 | 802 | #define UTF8_WARN_ILLEGAL_INTERCHANGE \ |
ecc1615f KW |
803 | (UTF8_WARN_ILLEGAL_C9_INTERCHANGE|UTF8_WARN_NONCHAR) |
804 | ||
d60baaa7 KW |
805 | /* This is used typically for code that is willing to accept inputs of |
806 | * illformed UTF-8 sequences, for whatever reason. However, all such sequences | |
807 | * evaluate to the REPLACEMENT CHARACTER unless other flags overriding this are | |
808 | * also present. */ | |
2d532c27 KW |
809 | #define UTF8_ALLOW_ANY ( UTF8_ALLOW_CONTINUATION \ |
810 | |UTF8_ALLOW_NON_CONTINUATION \ | |
811 | |UTF8_ALLOW_SHORT \ | |
d60baaa7 KW |
812 | |UTF8_ALLOW_LONG \ |
813 | |UTF8_ALLOW_OVERFLOW) | |
2d532c27 KW |
814 | |
815 | /* Accept any Perl-extended UTF-8 that evaluates to any UV on the platform, but | |
816 | * not any malformed. This is the default. (Note that UVs above IV_MAX are | |
817 | * deprecated. */ | |
818 | #define UTF8_ALLOW_ANYUV 0 | |
819 | #define UTF8_ALLOW_DEFAULT UTF8_ALLOW_ANYUV | |
1d72bdf6 | 820 | |
89d986df KW |
821 | /* |
822 | =for apidoc Am|bool|UTF8_IS_SURROGATE|const U8 *s|const U8 *e | |
823 | ||
824 | Evaluates to non-zero if the first few bytes of the string starting at C<s> and | |
825 | looking no further than S<C<e - 1>> are well-formed UTF-8 that represents one | |
826 | of the Unicode surrogate code points; otherwise it evaluates to 0. If | |
827 | non-zero, the value gives how many bytes starting at C<s> comprise the code | |
828 | point's representation. | |
829 | ||
830 | =cut | |
831 | */ | |
832 | #define UTF8_IS_SURROGATE(s, e) is_SURROGATE_utf8_safe(s, e) | |
833 | ||
834 | ||
835 | #define UTF8_IS_REPLACEMENT(s, send) is_REPLACEMENT_utf8_safe(s,send) | |
836 | ||
837 | /* | |
838 | =for apidoc Am|bool|UTF8_IS_SUPER|const U8 *s|const U8 *e | |
839 | ||
840 | Recall that Perl recognizes an extension to UTF-8 that can encode code | |
841 | points larger than the ones defined by Unicode, which are 0..0x10FFFF. | |
842 | ||
843 | This macro evaluates to non-zero if the first few bytes of the string starting | |
844 | at C<s> and looking no further than S<C<e - 1>> are from this UTF-8 extension; | |
845 | otherwise it evaluates to 0. If non-zero, the value gives how many bytes | |
846 | starting at C<s> comprise the code point's representation. | |
847 | ||
848 | 0 is returned if the bytes are not well-formed extended UTF-8, or if they | |
849 | represent a code point that cannot fit in a UV on the current platform. Hence | |
850 | this macro can give different results when run on a 64-bit word machine than on | |
851 | one with a 32-bit word size. | |
0c58a72b | 852 | |
89d986df KW |
853 | Note that it is deprecated to have code points that are larger than what can |
854 | fit in an IV on the current machine. | |
7131f24d | 855 | |
89d986df KW |
856 | =cut |
857 | ||
858 | * ASCII EBCDIC I8 | |
7131f24d KW |
859 | * U+10FFFF: \xF4\x8F\xBF\xBF \xF9\xA1\xBF\xBF\xBF max legal Unicode |
860 | * U+110000: \xF4\x90\x80\x80 \xF9\xA2\xA0\xA0\xA0 | |
861 | * U+110001: \xF4\x90\x80\x81 \xF9\xA2\xA0\xA0\xA1 | |
89d986df KW |
862 | */ |
863 | #ifdef EBCDIC | |
a14e0a36 | 864 | # define UTF8_IS_SUPER(s, e) \ |
89d986df KW |
865 | (( LIKELY((e) > (s) + 4) \ |
866 | && NATIVE_UTF8_TO_I8(*(s)) >= 0xF9 \ | |
867 | && ( NATIVE_UTF8_TO_I8(*(s)) > 0xF9 \ | |
3d42f267 | 868 | || (NATIVE_UTF8_TO_I8(*((s) + 1)) >= 0xA2)) \ |
89d986df | 869 | && LIKELY((s) + UTF8SKIP(s) <= (e))) \ |
2b479609 | 870 | ? _is_utf8_char_helper(s, s + UTF8SKIP(s), 0) : 0) |
7131f24d | 871 | #else |
a14e0a36 | 872 | # define UTF8_IS_SUPER(s, e) \ |
89d986df KW |
873 | (( LIKELY((e) > (s) + 3) \ |
874 | && (*(U8*) (s)) >= 0xF4 \ | |
875 | && ((*(U8*) (s)) > 0xF4 || (*((U8*) (s) + 1) >= 0x90))\ | |
876 | && LIKELY((s) + UTF8SKIP(s) <= (e))) \ | |
2b479609 | 877 | ? _is_utf8_char_helper(s, s + UTF8SKIP(s), 0) : 0) |
7131f24d KW |
878 | #endif |
879 | ||
b96a92fb KW |
880 | /* These are now machine generated, and the 'given' clause is no longer |
881 | * applicable */ | |
0c58a72b | 882 | #define UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s, e) \ |
89d986df KW |
883 | cBOOL(is_NONCHAR_utf8_safe(s,e)) |
884 | ||
885 | /* | |
886 | =for apidoc Am|bool|UTF8_IS_NONCHAR|const U8 *s|const U8 *e | |
887 | ||
888 | Evaluates to non-zero if the first few bytes of the string starting at C<s> and | |
889 | looking no further than S<C<e - 1>> are well-formed UTF-8 that represents one | |
890 | of the Unicode non-character code points; otherwise it evaluates to 0. If | |
891 | non-zero, the value gives how many bytes starting at C<s> comprise the code | |
892 | point's representation. | |
893 | ||
894 | =cut | |
895 | */ | |
0c58a72b KW |
896 | #define UTF8_IS_NONCHAR(s, e) \ |
897 | UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s, e) | |
7131f24d | 898 | |
c867b360 JH |
899 | #define UNICODE_SURROGATE_FIRST 0xD800 |
900 | #define UNICODE_SURROGATE_LAST 0xDFFF | |
901 | #define UNICODE_REPLACEMENT 0xFFFD | |
902 | #define UNICODE_BYTE_ORDER_MARK 0xFEFF | |
1d72bdf6 | 903 | |
b851fbc1 | 904 | /* Though our UTF-8 encoding can go beyond this, |
c76687c5 | 905 | * let's be conservative and do as Unicode says. */ |
b851fbc1 JH |
906 | #define PERL_UNICODE_MAX 0x10FFFF |
907 | ||
d35f2ca5 KW |
908 | #define UNICODE_WARN_SURROGATE 0x0001 /* UTF-16 surrogates */ |
909 | #define UNICODE_WARN_NONCHAR 0x0002 /* Non-char code points */ | |
910 | #define UNICODE_WARN_SUPER 0x0004 /* Above 0x10FFFF */ | |
911 | #define UNICODE_WARN_ABOVE_31_BIT 0x0008 /* Above 0x7FFF_FFFF */ | |
912 | #define UNICODE_DISALLOW_SURROGATE 0x0010 | |
913 | #define UNICODE_DISALLOW_NONCHAR 0x0020 | |
914 | #define UNICODE_DISALLOW_SUPER 0x0040 | |
915 | #define UNICODE_DISALLOW_ABOVE_31_BIT 0x0080 | |
ecc1615f KW |
916 | #define UNICODE_WARN_ILLEGAL_C9_INTERCHANGE \ |
917 | (UNICODE_WARN_SURROGATE|UNICODE_WARN_SUPER) | |
bb88be5f | 918 | #define UNICODE_WARN_ILLEGAL_INTERCHANGE \ |
ecc1615f KW |
919 | (UNICODE_WARN_ILLEGAL_C9_INTERCHANGE|UNICODE_WARN_NONCHAR) |
920 | #define UNICODE_DISALLOW_ILLEGAL_C9_INTERCHANGE \ | |
921 | (UNICODE_DISALLOW_SURROGATE|UNICODE_DISALLOW_SUPER) | |
bb88be5f | 922 | #define UNICODE_DISALLOW_ILLEGAL_INTERCHANGE \ |
ecc1615f | 923 | (UNICODE_DISALLOW_ILLEGAL_C9_INTERCHANGE|UNICODE_DISALLOW_NONCHAR) |
949cf498 KW |
924 | |
925 | /* For backward source compatibility, as are now the default */ | |
926 | #define UNICODE_ALLOW_SURROGATE 0 | |
927 | #define UNICODE_ALLOW_SUPER 0 | |
928 | #define UNICODE_ALLOW_ANY 0 | |
b851fbc1 | 929 | |
2d6b3d38 KW |
930 | /* This matches the 2048 code points between UNICODE_SURROGATE_FIRST (0xD800) and |
931 | * UNICODE_SURROGATE_LAST (0xDFFF) */ | |
932 | #define UNICODE_IS_SURROGATE(uv) (((UV) (uv) & (~0xFFFF | 0xF800)) \ | |
933 | == 0xD800) | |
934 | ||
646d1759 KW |
935 | #define UNICODE_IS_REPLACEMENT(uv) ((UV) (uv) == UNICODE_REPLACEMENT) |
936 | #define UNICODE_IS_BYTE_ORDER_MARK(uv) ((UV) (uv) == UNICODE_BYTE_ORDER_MARK) | |
c149ab20 KW |
937 | |
938 | /* Is 'uv' one of the 32 contiguous-range noncharacters? */ | |
939 | #define UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv) ((UV) (uv) >= 0xFDD0 \ | |
940 | && (UV) (uv) <= 0xFDEF) | |
941 | ||
942 | /* Is 'uv' one of the 34 plane-ending noncharacters 0xFFFE, 0xFFFF, 0x1FFFE, | |
943 | * 0x1FFFF, ... 0x10FFFE, 0x10FFFF, given that we know that 'uv' is not above | |
944 | * the Unicode legal max */ | |
945 | #define UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv) \ | |
946 | (((UV) (uv) & 0xFFFE) == 0xFFFE) | |
947 | ||
948 | #define UNICODE_IS_NONCHAR(uv) \ | |
949 | ( UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv) \ | |
950 | || ( LIKELY( ! UNICODE_IS_SUPER(uv)) \ | |
951 | && UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv))) | |
952 | ||
953 | #define UNICODE_IS_SUPER(uv) ((UV) (uv) > PERL_UNICODE_MAX) | |
d35f2ca5 | 954 | #define UNICODE_IS_ABOVE_31_BIT(uv) ((UV) (uv) > 0x7FFFFFFF) |
1d72bdf6 | 955 | |
ec34087a KW |
956 | #define LATIN_SMALL_LETTER_SHARP_S LATIN_SMALL_LETTER_SHARP_S_NATIVE |
957 | #define LATIN_SMALL_LETTER_Y_WITH_DIAERESIS \ | |
958 | LATIN_SMALL_LETTER_Y_WITH_DIAERESIS_NATIVE | |
959 | #define MICRO_SIGN MICRO_SIGN_NATIVE | |
960 | #define LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE \ | |
961 | LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE_NATIVE | |
962 | #define LATIN_SMALL_LETTER_A_WITH_RING_ABOVE \ | |
963 | LATIN_SMALL_LETTER_A_WITH_RING_ABOVE_NATIVE | |
09091399 JH |
964 | #define UNICODE_GREEK_CAPITAL_LETTER_SIGMA 0x03A3 |
965 | #define UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA 0x03C2 | |
966 | #define UNICODE_GREEK_SMALL_LETTER_SIGMA 0x03C3 | |
9dcbe121 | 967 | #define GREEK_SMALL_LETTER_MU 0x03BC |
9e682c18 KW |
968 | #define GREEK_CAPITAL_LETTER_MU 0x039C /* Upper and title case |
969 | of MICRON */ | |
970 | #define LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS 0x0178 /* Also is title case */ | |
0766489e KW |
971 | #ifdef LATIN_CAPITAL_LETTER_SHARP_S_UTF8 |
972 | # define LATIN_CAPITAL_LETTER_SHARP_S 0x1E9E | |
973 | #endif | |
74894415 KW |
974 | #define LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE 0x130 |
975 | #define LATIN_SMALL_LETTER_DOTLESS_I 0x131 | |
9e682c18 | 976 | #define LATIN_SMALL_LETTER_LONG_S 0x017F |
a9f50d33 KW |
977 | #define LATIN_SMALL_LIGATURE_LONG_S_T 0xFB05 |
978 | #define LATIN_SMALL_LIGATURE_ST 0xFB06 | |
9e682c18 KW |
979 | #define KELVIN_SIGN 0x212A |
980 | #define ANGSTROM_SIGN 0x212B | |
09091399 | 981 | |
9e55ce06 | 982 | #define UNI_DISPLAY_ISPRINT 0x0001 |
c728cb41 JH |
983 | #define UNI_DISPLAY_BACKSLASH 0x0002 |
984 | #define UNI_DISPLAY_QQ (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH) | |
985 | #define UNI_DISPLAY_REGEX (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH) | |
9e55ce06 | 986 | |
5cd46e1f KW |
987 | #define ANYOF_FOLD_SHARP_S(node, input, end) \ |
988 | (ANYOF_BITMAP_TEST(node, LATIN_SMALL_LETTER_SHARP_S) && \ | |
137165a6 | 989 | (ANYOF_NONBITMAP(node)) && \ |
39065660 | 990 | (ANYOF_FLAGS(node) & ANYOF_LOC_NONBITMAP_FOLD) && \ |
07b6858f | 991 | ((end) > (input) + 1) && \ |
305b8651 | 992 | isALPHA_FOLD_EQ((input)[0], 's')) |
6302f837 | 993 | |
ebc501f0 | 994 | #define SHARP_S_SKIP 2 |
3b0fc154 | 995 | |
6302f837 | 996 | /* |
5dca9278 KW |
997 | |
998 | =for apidoc Am|STRLEN|isUTF8_CHAR|const U8 *s|const U8 *e | |
999 | ||
35f8c9bd | 1000 | Evaluates to non-zero if the first few bytes of the string starting at C<s> and |
2c6ed66c KW |
1001 | looking no further than S<C<e - 1>> are well-formed UTF-8, as extended by Perl, |
1002 | that represents some code point; otherwise it evaluates to 0. If non-zero, the | |
1003 | value gives how many bytes starting at C<s> comprise the code point's | |
2717076a KW |
1004 | representation. Any bytes remaining before C<e>, but beyond the ones needed to |
1005 | form the first code point in C<s>, are not examined. | |
6302f837 | 1006 | |
35f8c9bd KW |
1007 | The code point can be any that will fit in a UV on this machine, using Perl's |
1008 | extension to official UTF-8 to represent those higher than the Unicode maximum | |
1009 | of 0x10FFFF. That means that this macro is used to efficiently decide if the | |
2717076a KW |
1010 | next few bytes in C<s> is legal UTF-8 for a single character. |
1011 | ||
1012 | Use C<L</isSTRICT_UTF8_CHAR>> to restrict the acceptable code points to those | |
1013 | defined by Unicode to be fully interchangeable across applications; | |
1014 | C<L</isC9_STRICT_UTF8_CHAR>> to use the L<Unicode Corrigendum | |
1015 | #9|http://www.unicode.org/versions/corrigendum9.html> definition of allowable | |
1016 | code points; and C<L</isUTF8_CHAR_flags>> for a more customized definition. | |
1017 | ||
1018 | Use C<L</is_utf8_string>>, C<L</is_utf8_string_loc>>, and | |
1019 | C<L</is_utf8_string_loclen>> to check entire strings. | |
35f8c9bd KW |
1020 | |
1021 | Note that it is deprecated to use code points higher than what will fit in an | |
1022 | IV. This macro does not raise any warnings for such code points, treating them | |
1023 | as valid. | |
1024 | ||
1025 | Note also that a UTF-8 INVARIANT character (i.e. ASCII on non-EBCDIC machines) | |
1026 | is a valid UTF-8 character. | |
5dca9278 KW |
1027 | |
1028 | =cut | |
1029 | */ | |
6302f837 | 1030 | |
784d4f31 KW |
1031 | #define isUTF8_CHAR(s, e) \ |
1032 | (UNLIKELY((e) <= (s)) \ | |
1033 | ? 0 \ | |
1034 | : (UTF8_IS_INVARIANT(*s)) \ | |
df863e43 KW |
1035 | ? 1 \ |
1036 | : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \ | |
1037 | ? 0 \ | |
1038 | : LIKELY(NATIVE_UTF8_TO_I8(*s) <= _IS_UTF8_CHAR_HIGHEST_START_BYTE) \ | |
1039 | ? is_UTF8_CHAR_utf8_no_length_checks(s) \ | |
1040 | : _is_utf8_char_helper(s, e, 0)) | |
6302f837 | 1041 | |
3cedd9d9 KW |
1042 | #define is_utf8_char_buf(buf, buf_end) isUTF8_CHAR(buf, buf_end) |
1043 | ||
e23e8bc1 KW |
1044 | /* |
1045 | ||
1046 | =for apidoc Am|STRLEN|isSTRICT_UTF8_CHAR|const U8 *s|const U8 *e | |
1047 | ||
1048 | Evaluates to non-zero if the first few bytes of the string starting at C<s> and | |
1049 | looking no further than S<C<e - 1>> are well-formed UTF-8 that represents some | |
1050 | Unicode code point completely acceptable for open interchange between all | |
1051 | applications; otherwise it evaluates to 0. If non-zero, the value gives how | |
2717076a KW |
1052 | many bytes starting at C<s> comprise the code point's representation. Any |
1053 | bytes remaining before C<e>, but beyond the ones needed to form the first code | |
1054 | point in C<s>, are not examined. | |
e23e8bc1 KW |
1055 | |
1056 | The largest acceptable code point is the Unicode maximum 0x10FFFF, and must not | |
1057 | be a surrogate nor a non-character code point. Thus this excludes any code | |
1058 | point from Perl's extended UTF-8. | |
1059 | ||
1060 | This is used to efficiently decide if the next few bytes in C<s> is | |
2717076a KW |
1061 | legal Unicode-acceptable UTF-8 for a single character. |
1062 | ||
1063 | Use C<L</isC9_STRICT_UTF8_CHAR>> to use the L<Unicode Corrigendum | |
1064 | #9|http://www.unicode.org/versions/corrigendum9.html> definition of allowable | |
1065 | code points; C<L</isUTF8_CHAR>> to check for Perl's extended UTF-8; | |
1066 | and C<L</isUTF8_CHAR_flags>> for a more customized definition. | |
e23e8bc1 | 1067 | |
9f2abfde KW |
1068 | Use C<L</is_strict_utf8_string>>, C<L</is_strict_utf8_string_loc>>, and |
1069 | C<L</is_strict_utf8_string_loclen>> to check entire strings. | |
1070 | ||
e23e8bc1 KW |
1071 | =cut |
1072 | */ | |
1073 | ||
1074 | #define isSTRICT_UTF8_CHAR(s, e) \ | |
1075 | (UNLIKELY((e) <= (s)) \ | |
1076 | ? 0 \ | |
1077 | : (UTF8_IS_INVARIANT(*s)) \ | |
1078 | ? 1 \ | |
1079 | : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \ | |
1080 | ? 0 \ | |
1081 | : is_STRICT_UTF8_CHAR_utf8_no_length_checks(s)) | |
1082 | ||
a82be82b KW |
1083 | /* |
1084 | ||
1085 | =for apidoc Am|STRLEN|isC9_STRICT_UTF8_CHAR|const U8 *s|const U8 *e | |
1086 | ||
1087 | Evaluates to non-zero if the first few bytes of the string starting at C<s> and | |
1088 | looking no further than S<C<e - 1>> are well-formed UTF-8 that represents some | |
1089 | Unicode non-surrogate code point; otherwise it evaluates to 0. If non-zero, | |
1090 | the value gives how many bytes starting at C<s> comprise the code point's | |
2717076a KW |
1091 | representation. Any bytes remaining before C<e>, but beyond the ones needed to |
1092 | form the first code point in C<s>, are not examined. | |
a82be82b KW |
1093 | |
1094 | The largest acceptable code point is the Unicode maximum 0x10FFFF. This | |
1095 | differs from C<L</isSTRICT_UTF8_CHAR>> only in that it accepts non-character | |
1096 | code points. This corresponds to | |
1097 | L<Unicode Corrigendum #9|http://www.unicode.org/versions/corrigendum9.html>. | |
1098 | which said that non-character code points are merely discouraged rather than | |
1099 | completely forbidden in open interchange. See | |
1100 | L<perlunicode/Noncharacter code points>. | |
1101 | ||
2717076a KW |
1102 | Use C<L</isUTF8_CHAR>> to check for Perl's extended UTF-8; and |
1103 | C<L</isUTF8_CHAR_flags>> for a more customized definition. | |
1104 | ||
9f2abfde KW |
1105 | Use C<L</is_c9strict_utf8_string>>, C<L</is_c9strict_utf8_string_loc>>, and |
1106 | C<L</is_c9strict_utf8_string_loclen>> to check entire strings. | |
1107 | ||
a82be82b KW |
1108 | =cut |
1109 | */ | |
1110 | ||
1111 | #define isC9_STRICT_UTF8_CHAR(s, e) \ | |
1112 | (UNLIKELY((e) <= (s)) \ | |
1113 | ? 0 \ | |
1114 | : (UTF8_IS_INVARIANT(*s)) \ | |
1115 | ? 1 \ | |
1116 | : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \ | |
1117 | ? 0 \ | |
1118 | : is_C9_STRICT_UTF8_CHAR_utf8_no_length_checks(s)) | |
1119 | ||
25e3a4e0 KW |
1120 | /* |
1121 | ||
1122 | =for apidoc Am|STRLEN|isUTF8_CHAR_flags|const U8 *s|const U8 *e| const U32 flags | |
1123 | ||
1124 | Evaluates to non-zero if the first few bytes of the string starting at C<s> and | |
1125 | looking no further than S<C<e - 1>> are well-formed UTF-8, as extended by Perl, | |
1126 | that represents some code point, subject to the restrictions given by C<flags>; | |
1127 | otherwise it evaluates to 0. If non-zero, the value gives how many bytes | |
2717076a KW |
1128 | starting at C<s> comprise the code point's representation. Any bytes remaining |
1129 | before C<e>, but beyond the ones needed to form the first code point in C<s>, | |
1130 | are not examined. | |
25e3a4e0 KW |
1131 | |
1132 | If C<flags> is 0, this gives the same results as C<L</isUTF8_CHAR>>; | |
1133 | if C<flags> is C<UTF8_DISALLOW_ILLEGAL_INTERCHANGE>, this gives the same results | |
1134 | as C<L</isSTRICT_UTF8_CHAR>>; | |
1135 | and if C<flags> is C<UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE>, this gives | |
1136 | the same results as C<L</isC9_STRICT_UTF8_CHAR>>. | |
1137 | Otherwise C<flags> may be any combination of the C<UTF8_DISALLOW_I<foo>> flags | |
1138 | understood by C<L</utf8n_to_uvchr>>, with the same meanings. | |
1139 | ||
1140 | The three alternative macros are for the most commonly needed validations; they | |
1141 | are likely to run somewhat faster than this more general one, as they can be | |
1142 | inlined into your code. | |
1143 | ||
9f2abfde KW |
1144 | Use L</is_utf8_string_flags>, L</is_utf8_string_loc_flags>, and |
1145 | L</is_utf8_string_loclen_flags> to check entire strings. | |
1146 | ||
25e3a4e0 KW |
1147 | =cut |
1148 | */ | |
1149 | ||
1150 | #define isUTF8_CHAR_flags(s, e, flags) \ | |
1151 | (UNLIKELY((e) <= (s)) \ | |
1152 | ? 0 \ | |
1153 | : (UTF8_IS_INVARIANT(*s)) \ | |
1154 | ? 1 \ | |
1155 | : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \ | |
1156 | ? 0 \ | |
1157 | : _is_utf8_char_helper(s, e, flags)) | |
1158 | ||
6302f837 KW |
1159 | /* Do not use; should be deprecated. Use isUTF8_CHAR() instead; this is |
1160 | * retained solely for backwards compatibility */ | |
1161 | #define IS_UTF8_CHAR(p, n) (isUTF8_CHAR(p, (p) + (n)) == n) | |
e9a8c099 | 1162 | |
57f0e7e2 KW |
1163 | #endif /* H_UTF8 */ |
1164 | ||
e9a8c099 | 1165 | /* |
14d04a33 | 1166 | * ex: set ts=8 sts=4 sw=4 et: |
e9a8c099 | 1167 | */ |