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 KW |
74 | ckWARN_d(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY) |
75 | ||
a0270393 | 76 | #define to_uni_fold(c, p, lenp) _to_uni_fold_flags(c, p, lenp, FOLD_FLAGS_FULL) |
445bf929 KW |
77 | #define to_utf8_fold(c, p, lenp) _to_utf8_fold_flags(c, p, lenp, FOLD_FLAGS_FULL) |
78 | #define to_utf8_lower(a,b,c) _to_utf8_lower_flags(a,b,c,0) | |
79 | #define to_utf8_upper(a,b,c) _to_utf8_upper_flags(a,b,c,0) | |
80 | #define to_utf8_title(a,b,c) _to_utf8_title_flags(a,b,c,0) | |
36bb2ab6 | 81 | |
fd7cb289 | 82 | /* Source backward compatibility. */ |
fd7cb289 RGS |
83 | #define is_utf8_string_loc(s, len, ep) is_utf8_string_loclen(s, len, ep, 0) |
84 | ||
eda9cac1 KW |
85 | #define foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \ |
86 | foldEQ_utf8_flags(s1, pe1, l1, u1, s2, pe2, l2, u2, 0) | |
baa60164 | 87 | #define FOLDEQ_UTF8_NOMIX_ASCII (1 << 0) |
cea315b6 | 88 | #define FOLDEQ_LOCALE (1 << 1) |
18f762c3 KW |
89 | #define FOLDEQ_S1_ALREADY_FOLDED (1 << 2) |
90 | #define FOLDEQ_S2_ALREADY_FOLDED (1 << 3) | |
d635b710 KW |
91 | #define FOLDEQ_S1_FOLDS_SANE (1 << 4) |
92 | #define FOLDEQ_S2_FOLDS_SANE (1 << 5) | |
a33c29bc | 93 | |
e6226b18 KW |
94 | #define ibcmp_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \ |
95 | cBOOL(! foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2)) | |
96 | ||
1d72bdf6 NIS |
97 | #ifdef EBCDIC |
98 | /* The equivalent of these macros but implementing UTF-EBCDIC | |
99 | are in the following header file: | |
100 | */ | |
101 | ||
102 | #include "utfebcdic.h" | |
fd7cb289 | 103 | |
d06134e5 | 104 | #else /* ! EBCDIC */ |
73c4f7a1 GS |
105 | START_EXTERN_C |
106 | ||
111e8ed9 KW |
107 | /* How wide can a single UTF-8 encoded character become in bytes. */ |
108 | /* NOTE: Strictly speaking Perl's UTF-8 should not be called UTF-8 since UTF-8 | |
109 | * is an encoding of Unicode, and Unicode's upper limit, 0x10FFFF, can be | |
110 | * expressed with 4 bytes. However, Perl thinks of UTF-8 as a way to encode | |
111 | * non-negative integers in a binary format, even those above Unicode */ | |
112 | #define UTF8_MAXBYTES 13 | |
113 | ||
a0ed51b3 | 114 | #ifdef DOINIT |
6f06b55f | 115 | EXTCONST unsigned char PL_utf8skip[] = { |
b2635aa8 KW |
116 | /* 0x00 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */ |
117 | /* 0x10 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */ | |
118 | /* 0x20 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */ | |
119 | /* 0x30 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */ | |
120 | /* 0x40 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */ | |
121 | /* 0x50 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */ | |
122 | /* 0x60 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */ | |
123 | /* 0x70 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */ | |
124 | /* 0x80 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */ | |
125 | /* 0x90 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */ | |
126 | /* 0xA0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */ | |
127 | /* 0xB0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */ | |
128 | /* 0xC0 */ 2,2, /* overlong */ | |
1ff3baa2 | 129 | /* 0xC2 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0080 to U+03FF */ |
b2635aa8 KW |
130 | /* 0xD0 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0400 to U+07FF */ |
131 | /* 0xE0 */ 3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3, /* U+0800 to U+FFFF */ | |
132 | /* 0xF0 */ 4,4,4,4,4,4,4,4,5,5,5,5,6,6, /* above BMP to 2**31 - 1 */ | |
6937f885 KW |
133 | /* Perl extended (never was official UTF-8). Up to 36 bit */ |
134 | /* 0xFE */ 7, | |
135 | /* More extended, Up to 72 bits (64-bit + reserved) */ | |
111e8ed9 | 136 | /* 0xFF */ UTF8_MAXBYTES |
a0ed51b3 LW |
137 | }; |
138 | #else | |
6f06b55f | 139 | EXTCONST unsigned char PL_utf8skip[]; |
a0ed51b3 LW |
140 | #endif |
141 | ||
73c4f7a1 | 142 | END_EXTERN_C |
7e2040f0 | 143 | |
1a3756de | 144 | #if defined(_MSC_VER) && _MSC_VER < 1400 |
6f6d1bab TC |
145 | /* older MSVC versions have a smallish macro buffer */ |
146 | #define PERL_SMALL_MACRO_BUFFER | |
147 | #endif | |
148 | ||
59a449d5 KW |
149 | /* Native character to/from iso-8859-1. Are the identity functions on ASCII |
150 | * platforms */ | |
6f6d1bab | 151 | #ifdef PERL_SMALL_MACRO_BUFFER |
1a3756de TC |
152 | #define NATIVE_TO_LATIN1(ch) ((U8)(ch)) |
153 | #define LATIN1_TO_NATIVE(ch) ((U8)(ch)) | |
6f6d1bab | 154 | #else |
4c8cd605 KW |
155 | #define NATIVE_TO_LATIN1(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch))) |
156 | #define LATIN1_TO_NATIVE(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch))) | |
6f6d1bab | 157 | #endif |
59a449d5 KW |
158 | |
159 | /* I8 is an intermediate version of UTF-8 used only in UTF-EBCDIC. We thus | |
160 | * consider it to be identical to UTF-8 on ASCII platforms. Strictly speaking | |
161 | * UTF-8 and UTF-EBCDIC are two different things, but we often conflate them | |
162 | * because they are 8-bit encodings that serve the same purpose in Perl, and | |
163 | * rarely do we need to distinguish them. The term "NATIVE_UTF8" applies to | |
164 | * whichever one is applicable on the current platform */ | |
6f6d1bab | 165 | #ifdef PERL_SMALL_MACRO_BUFFER |
1d4ea287 DD |
166 | #define NATIVE_UTF8_TO_I8(ch) ((U8) (ch)) |
167 | #define I8_TO_NATIVE_UTF8(ch) ((U8) (ch)) | |
6f6d1bab | 168 | #else |
4c8cd605 KW |
169 | #define NATIVE_UTF8_TO_I8(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch))) |
170 | #define I8_TO_NATIVE_UTF8(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch))) | |
6f6d1bab | 171 | #endif |
59a449d5 | 172 | |
1d72bdf6 | 173 | /* Transforms in wide UV chars */ |
4c8cd605 KW |
174 | #define UNI_TO_NATIVE(ch) ((UV) (ch)) |
175 | #define NATIVE_TO_UNI(ch) ((UV) (ch)) | |
d7578b48 | 176 | |
877d9f0d | 177 | /* |
9041c2e3 | 178 | |
a14e0a36 KW |
179 | The following table is from Unicode 3.2, plus the Perl extensions for above |
180 | U+10FFFF | |
877d9f0d | 181 | |
a14e0a36 | 182 | Code Points 1st Byte 2nd Byte 3rd 4th 5th 6th 7th 8th-13th |
877d9f0d | 183 | |
375122d7 | 184 | U+0000..U+007F 00..7F |
e1b711da | 185 | U+0080..U+07FF * C2..DF 80..BF |
a14e0a36 KW |
186 | U+0800..U+0FFF E0 * A0..BF 80..BF |
187 | U+1000..U+CFFF E1..EC 80..BF 80..BF | |
188 | U+D000..U+D7FF ED 80..9F 80..BF | |
189 | U+D800..U+DFFF ED A0..BF 80..BF (surrogates) | |
190 | U+E000..U+FFFF EE..EF 80..BF 80..BF | |
191 | U+10000..U+3FFFF F0 * 90..BF 80..BF 80..BF | |
192 | U+40000..U+FFFFF F1..F3 80..BF 80..BF 80..BF | |
193 | U+100000..U+10FFFF F4 80..8F 80..BF 80..BF | |
194 | Below are above-Unicode code points | |
195 | U+110000..U+13FFFF F4 90..BF 80..BF 80..BF | |
196 | U+110000..U+1FFFFF F5..F7 80..BF 80..BF 80..BF | |
197 | U+200000..U+FFFFFF F8 * 88..BF 80..BF 80..BF 80..BF | |
198 | U+1000000..U+3FFFFFF F9..FB 80..BF 80..BF 80..BF 80..BF | |
199 | U+4000000..U+3FFFFFFF FC * 84..BF 80..BF 80..BF 80..BF 80..BF | |
200 | U+40000000..U+7FFFFFFF FD 80..BF 80..BF 80..BF 80..BF 80..BF | |
201 | U+80000000..U+FFFFFFFFF FE * 82..BF 80..BF 80..BF 80..BF 80..BF 80..BF | |
202 | U+1000000000.. FF 80..BF 80..BF 80..BF 80..BF 80..BF * 81..BF 80..BF | |
877d9f0d | 203 | |
e1b711da | 204 | Note the gaps before several of the byte entries above marked by '*'. These are |
37e2e78e KW |
205 | caused by legal UTF-8 avoiding non-shortest encodings: it is technically |
206 | possible to UTF-8-encode a single code point in different ways, but that is | |
207 | explicitly forbidden, and the shortest possible encoding should always be used | |
15824458 | 208 | (and that is what Perl does). The non-shortest ones are called 'overlongs'. |
8c007b5a | 209 | |
877d9f0d JH |
210 | */ |
211 | ||
8c007b5a JH |
212 | /* |
213 | Another way to look at it, as bits: | |
214 | ||
b2635aa8 | 215 | Code Points 1st Byte 2nd Byte 3rd Byte 4th Byte |
8c007b5a | 216 | |
b2635aa8 KW |
217 | 0aaa aaaa 0aaa aaaa |
218 | 0000 0bbb bbaa aaaa 110b bbbb 10aa aaaa | |
219 | cccc bbbb bbaa aaaa 1110 cccc 10bb bbbb 10aa aaaa | |
220 | 00 000d ddcc cccc bbbb bbaa aaaa 1111 0ddd 10cc cccc 10bb bbbb 10aa aaaa | |
8c007b5a JH |
221 | |
222 | As you can see, the continuation bytes all begin with C<10>, and the | |
e1b711da | 223 | leading bits of the start byte tell how many bytes there are in the |
8c007b5a JH |
224 | encoded character. |
225 | ||
65ab9279 TC |
226 | Perl's extended UTF-8 means we can have start bytes up to FF. |
227 | ||
8c007b5a JH |
228 | */ |
229 | ||
eda91ad7 KW |
230 | /* Anything larger than this will overflow the word if it were converted into a UV */ |
231 | #if defined(UV_IS_QUAD) | |
232 | # define HIGHEST_REPRESENTABLE_UTF8 "\xFF\x80\x8F\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF" | |
233 | #else | |
234 | # define HIGHEST_REPRESENTABLE_UTF8 "\xFE\x83\xBF\xBF\xBF\xBF\xBF" | |
235 | #endif | |
236 | ||
6c88483e | 237 | /* Is the representation of the Unicode code point 'cp' the same regardless of |
15824458 | 238 | * being encoded in UTF-8 or not? */ |
2d1545e5 | 239 | #define OFFUNI_IS_INVARIANT(cp) isASCII(cp) |
15824458 | 240 | |
38953e5a KW |
241 | /* Is the representation of the code point 'cp' the same regardless of |
242 | * being encoded in UTF-8 or not? 'cp' is native if < 256; Unicode otherwise | |
243 | * */ | |
cf1be84e | 244 | #define UVCHR_IS_INVARIANT(cp) OFFUNI_IS_INVARIANT(cp) |
38953e5a | 245 | |
c9264833 KW |
246 | /* This defines the bits that are to be in the continuation bytes of a multi-byte |
247 | * UTF-8 encoded character that mark it is a continuation byte. */ | |
248 | #define UTF_CONTINUATION_MARK 0x80 | |
249 | ||
a95ec4fb | 250 | /* Misleadingly named: is the UTF8-encoded byte 'c' part of a variant sequence |
5c06326b KW |
251 | * in UTF-8? This is the inverse of UTF8_IS_INVARIANT. The |0 makes sure this |
252 | * isn't mistakenly called with a ptr argument */ | |
253 | #define UTF8_IS_CONTINUED(c) (((U8)((c) | 0)) & UTF_CONTINUATION_MARK) | |
15824458 KW |
254 | |
255 | /* Is the byte 'c' the first byte of a multi-byte UTF8-8 encoded sequence? | |
256 | * This doesn't catch invariants (they are single-byte). It also excludes the | |
5c06326b KW |
257 | * illegal overlong sequences that begin with C0 and C1. The |0 makes sure |
258 | * this isn't mistakenly called with a ptr argument */ | |
259 | #define UTF8_IS_START(c) (((U8)((c) | 0)) >= 0xc2) | |
15824458 | 260 | |
858cd8ab KW |
261 | /* For use in UTF8_IS_CONTINUATION() below */ |
262 | #define UTF_IS_CONTINUATION_MASK 0xC0 | |
263 | ||
15824458 | 264 | /* Is the byte 'c' part of a multi-byte UTF8-8 encoded sequence, and not the |
5c06326b KW |
265 | * first byte thereof? The |0 makes sure this isn't mistakenly called with a |
266 | * ptr argument */ | |
858cd8ab | 267 | #define UTF8_IS_CONTINUATION(c) \ |
5c06326b | 268 | ((((U8)((c) | 0)) & UTF_IS_CONTINUATION_MASK) == UTF_CONTINUATION_MARK) |
0ae1fa71 | 269 | |
15824458 KW |
270 | /* Is the UTF8-encoded byte 'c' the first byte of a two byte sequence? Use |
271 | * UTF8_IS_NEXT_CHAR_DOWNGRADEABLE() instead if the input isn't known to | |
272 | * be well-formed. Masking with 0xfe allows the low bit to be 0 or 1; thus | |
5c06326b KW |
273 | * this matches 0xc[23]. The |0 makes sure this isn't mistakenly called with a |
274 | * ptr argument */ | |
275 | #define UTF8_IS_DOWNGRADEABLE_START(c) ((((U8)((c) | 0)) & 0xfe) == 0xc2) | |
4ab10950 | 276 | |
15824458 | 277 | /* Is the UTF8-encoded byte 'c' the first byte of a sequence of bytes that |
5c06326b KW |
278 | * represent a code point > 255? The |0 makes sure this isn't mistakenly |
279 | * called with a ptr argument */ | |
280 | #define UTF8_IS_ABOVE_LATIN1(c) (((U8)((c) | 0)) >= 0xc4) | |
8850bf83 | 281 | |
15824458 KW |
282 | /* This is the number of low-order bits a continuation byte in a UTF-8 encoded |
283 | * sequence contributes to the specification of the code point. In the bit | |
284 | * maps above, you see that the first 2 bits are a constant '10', leaving 6 of | |
285 | * real information */ | |
1d72bdf6 | 286 | #define UTF_ACCUMULATION_SHIFT 6 |
b2635aa8 | 287 | |
fed423a5 KW |
288 | /* ^? is defined to be DEL on ASCII systems. See the definition of toCTRL() |
289 | * for more */ | |
290 | #define QUESTION_MARK_CTRL DEL_NATIVE | |
291 | ||
292 | /* Surrogates, non-character code points and above-Unicode code points are | |
293 | * problematic in some contexts. This allows code that needs to check for | |
294 | * those to to quickly exclude the vast majority of code points it will | |
295 | * encounter */ | |
296 | #define isUTF8_POSSIBLY_PROBLEMATIC(c) ((U8) c >= 0xED) | |
297 | ||
298 | #endif /* EBCDIC vs ASCII */ | |
299 | ||
300 | /* 2**UTF_ACCUMULATION_SHIFT - 1 */ | |
301 | #define UTF_CONTINUATION_MASK ((U8) ((1U << UTF_ACCUMULATION_SHIFT) - 1)) | |
302 | ||
7028aeba KW |
303 | /* Internal macro to be used only in this file to aid in constructing other |
304 | * publicly accessible macros. | |
305 | * The number of bytes required to express this uv in UTF-8, for just those | |
306 | * uv's requiring 2 through 6 bytes, as these are common to all platforms and | |
307 | * word sizes. The number of bytes needed is given by the number of leading 1 | |
308 | * bits in the start byte. There are 32 start bytes that have 2 initial 1 bits | |
309 | * (C0-DF); there are 16 that have 3 initial 1 bits (E0-EF); 8 that have 4 | |
310 | * initial 1 bits (F0-F8); 4 that have 5 initial 1 bits (F9-FB), and 2 that | |
311 | * have 6 initial 1 bits (FC-FD). The largest number a string of n bytes can | |
312 | * represent is (the number of possible start bytes for 'n') | |
313 | * * (the number of possiblities for each start byte | |
314 | * The latter in turn is | |
315 | * 2 ** ( (how many continuation bytes there are) | |
316 | * * (the number of bits of information each | |
317 | * continuation byte holds)) | |
318 | * | |
319 | * If we were on a platform where we could use a fast find first set bit | |
320 | * instruction (or count leading zeros instruction) this could be replaced by | |
321 | * using that to find the log2 of the uv, and divide that by the number of bits | |
322 | * of information in each continuation byte, adjusting for large cases and how | |
323 | * much information is in a start byte for that length */ | |
72164d3a | 324 | #define __COMMON_UNI_SKIP(uv) \ |
7028aeba KW |
325 | (UV) (uv) < (32 * (1U << ( UTF_ACCUMULATION_SHIFT))) ? 2 : \ |
326 | (UV) (uv) < (16 * (1U << (2 * UTF_ACCUMULATION_SHIFT))) ? 3 : \ | |
327 | (UV) (uv) < ( 8 * (1U << (3 * UTF_ACCUMULATION_SHIFT))) ? 4 : \ | |
328 | (UV) (uv) < ( 4 * (1U << (4 * UTF_ACCUMULATION_SHIFT))) ? 5 : \ | |
329 | (UV) (uv) < ( 2 * (1U << (5 * UTF_ACCUMULATION_SHIFT))) ? 6 : | |
72164d3a KW |
330 | |
331 | /* Internal macro to be used only in this file. | |
332 | * This adds to __COMMON_UNI_SKIP the details at this platform's upper range. | |
fed423a5 | 333 | * For any-sized EBCDIC platforms, or 64-bit ASCII ones, we need one more test |
72164d3a KW |
334 | * to see if just 7 bytes is needed, or if the maximum is needed. For 32-bit |
335 | * ASCII platforms, everything is representable by 7 bytes */ | |
fed423a5 | 336 | #if defined(UV_IS_QUAD) || defined(EBCDIC) |
72164d3a | 337 | # define __BASE_UNI_SKIP(uv) (__COMMON_UNI_SKIP(uv) \ |
7028aeba | 338 | (UV) (uv) < ((UV) 1U << (6 * UTF_ACCUMULATION_SHIFT)) ? 7 : UTF8_MAXBYTES) |
1d68d6cd | 339 | #else |
72164d3a | 340 | # define __BASE_UNI_SKIP(uv) (__COMMON_UNI_SKIP(uv) 7) |
1d68d6cd SC |
341 | #endif |
342 | ||
fed423a5 KW |
343 | /* The next two macros use the base macro defined above, and add in the tests |
344 | * at the low-end of the range, for just 1 byte, yielding complete macros, | |
345 | * publicly accessible. */ | |
346 | ||
347 | /* Input is a true Unicode (not-native) code point */ | |
348 | #define OFFUNISKIP(uv) (OFFUNI_IS_INVARIANT(uv) ? 1 : __BASE_UNI_SKIP(uv)) | |
2084b489 | 349 | |
5352a763 KW |
350 | /* |
351 | ||
352 | =for apidoc Am|STRLEN|UVCHR_SKIP|UV cp | |
353 | returns the number of bytes required to represent the code point C<cp> when | |
354 | encoded as UTF-8. C<cp> is a native (ASCII or EBCDIC) code point if less than | |
355 | 255; a Unicode code point otherwise. | |
356 | ||
357 | =cut | |
358 | */ | |
fdb6583d | 359 | #define UVCHR_SKIP(uv) ( UVCHR_IS_INVARIANT(uv) ? 1 : __BASE_UNI_SKIP(uv)) |
5352a763 | 360 | |
b651802e KW |
361 | /* The largest code point representable by two UTF-8 bytes on this platform. |
362 | * As explained in the comments for __COMMON_UNI_SKIP, 32 start bytes with | |
fed423a5 | 363 | * UTF_ACCUMULATION_SHIFT bits of information each */ |
aa206fb7 KW |
364 | #define MAX_UTF8_TWO_BYTE (32 * (1U << UTF_ACCUMULATION_SHIFT) - 1) |
365 | ||
b651802e KW |
366 | /* The largest code point representable by two UTF-8 bytes on any platform that |
367 | * Perl runs on. This value is constrained by EBCDIC which has 5 bits per | |
368 | * continuation byte */ | |
aa206fb7 KW |
369 | #define MAX_PORTABLE_UTF8_TWO_BYTE (32 * (1U << 5) - 1) |
370 | ||
c03c0950 KW |
371 | /* The maximum number of UTF-8 bytes a single Unicode character can |
372 | * uppercase/lowercase/fold into. Unicode guarantees that the maximum | |
373 | * expansion is UTF8_MAX_FOLD_CHAR_EXPAND characters, but any above-Unicode | |
374 | * code point will fold to itself, so we only have to look at the expansion of | |
375 | * the maximum Unicode code point. But this number may be less than the space | |
376 | * occupied by a very large code point under Perl's extended UTF-8. We have to | |
377 | * make it large enough to fit any single character. (It turns out that ASCII | |
378 | * and EBCDIC differ in which is larger) */ | |
379 | #define UTF8_MAXBYTES_CASE \ | |
380 | (UTF8_MAXBYTES >= (UTF8_MAX_FOLD_CHAR_EXPAND * OFFUNISKIP(0x10FFFF)) \ | |
381 | ? UTF8_MAXBYTES \ | |
382 | : (UTF8_MAX_FOLD_CHAR_EXPAND * OFFUNISKIP(0x10FFFF))) | |
383 | ||
d06134e5 KW |
384 | /* Rest of these are attributes of Unicode and perl's internals rather than the |
385 | * encoding, or happen to be the same in both ASCII and EBCDIC (at least at | |
386 | * this level; the macros that some of these call may have different | |
387 | * definitions in the two encodings */ | |
388 | ||
59a449d5 KW |
389 | /* In domain restricted to ASCII, these may make more sense to the reader than |
390 | * the ones with Latin1 in the name */ | |
391 | #define NATIVE_TO_ASCII(ch) NATIVE_TO_LATIN1(ch) | |
392 | #define ASCII_TO_NATIVE(ch) LATIN1_TO_NATIVE(ch) | |
393 | ||
394 | /* More or less misleadingly-named defines, retained for back compat */ | |
395 | #define NATIVE_TO_UTF(ch) NATIVE_UTF8_TO_I8(ch) | |
396 | #define NATIVE_TO_I8(ch) NATIVE_UTF8_TO_I8(ch) | |
397 | #define UTF_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch) | |
398 | #define I8_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch) | |
399 | #define NATIVE8_TO_UNI(ch) NATIVE_TO_LATIN1(ch) | |
d06134e5 | 400 | |
c0236afe | 401 | /* This defines the 1-bits that are to be in the first byte of a multi-byte |
97d0ceda KW |
402 | * UTF-8 encoded character that mark it as a start byte and give the number of |
403 | * bytes that comprise the character. 'len' is the number of bytes in the | |
404 | * multi-byte sequence. */ | |
c0236afe KW |
405 | #define UTF_START_MARK(len) (((len) > 7) ? 0xFF : (0xFF & (0xFE << (7-(len))))) |
406 | ||
407 | /* Masks out the initial one bits in a start byte, leaving the real data ones. | |
408 | * Doesn't work on an invariant byte. 'len' is the number of bytes in the | |
409 | * multi-byte sequence that comprises the character. */ | |
410 | #define UTF_START_MASK(len) (((len) >= 7) ? 0x00 : (0x1F >> ((len)-2))) | |
411 | ||
537124e4 KW |
412 | /* Adds a UTF8 continuation byte 'new' of information to a running total code |
413 | * point 'old' of all the continuation bytes so far. This is designed to be | |
155d2738 KW |
414 | * used in a loop to convert from UTF-8 to the code point represented. Note |
415 | * that this is asymmetric on EBCDIC platforms, in that the 'new' parameter is | |
416 | * the UTF-EBCDIC byte, whereas the 'old' parameter is a Unicode (not EBCDIC) | |
417 | * code point in process of being generated */ | |
418 | #define UTF8_ACCUMULATE(old, new) (((old) << UTF_ACCUMULATION_SHIFT) \ | |
419 | | ((NATIVE_UTF8_TO_I8((U8)new)) \ | |
420 | & UTF_CONTINUATION_MASK)) | |
d06134e5 | 421 | |
7c560c3b KW |
422 | /* If a value is anded with this, and the result is non-zero, then using the |
423 | * original value in UTF8_ACCUMULATE will overflow, shifting bits off the left | |
424 | * */ | |
425 | #define UTF_ACCUMULATION_OVERFLOW_MASK \ | |
426 | (((UV) UTF_CONTINUATION_MASK) << ((sizeof(UV) * CHARBITS) \ | |
427 | - UTF_ACCUMULATION_SHIFT)) | |
428 | ||
4ab10950 KW |
429 | /* This works in the face of malformed UTF-8. */ |
430 | #define UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(s, e) (UTF8_IS_DOWNGRADEABLE_START(*s) \ | |
431 | && ( (e) - (s) > 1) \ | |
432 | && UTF8_IS_CONTINUATION(*((s)+1))) | |
433 | ||
5aaebcb3 | 434 | /* Number of bytes a code point occupies in UTF-8. */ |
5352a763 | 435 | #define NATIVE_SKIP(uv) UVCHR_SKIP(uv) |
bd18bd40 | 436 | |
5aaebcb3 KW |
437 | /* Most code which says UNISKIP is really thinking in terms of native code |
438 | * points (0-255) plus all those beyond. This is an imprecise term, but having | |
2accb712 | 439 | * it means existing code continues to work. For precision, use UVCHR_SKIP, |
5352a763 KW |
440 | * NATIVE_SKIP, or OFFUNISKIP */ |
441 | #define UNISKIP(uv) UVCHR_SKIP(uv) | |
5aaebcb3 | 442 | |
3c0792e4 KW |
443 | /* Longer, but more accurate name */ |
444 | #define UTF8_IS_ABOVE_LATIN1_START(c) UTF8_IS_ABOVE_LATIN1(c) | |
445 | ||
a62b247b KW |
446 | /* Convert a UTF-8 variant Latin1 character to a native code point value. |
447 | * Needs just one iteration of accumulate. Should be used only if it is known | |
448 | * that the code point is < 256, and is not UTF-8 invariant. Use the slower | |
449 | * but more general TWO_BYTE_UTF8_TO_NATIVE() which handles any code point | |
450 | * representable by two bytes (which turns out to be up through | |
451 | * MAX_PORTABLE_UTF8_TWO_BYTE). The two parameters are: | |
452 | * HI: a downgradable start byte; | |
453 | * LO: continuation. | |
454 | * */ | |
455 | #define EIGHT_BIT_UTF8_TO_NATIVE(HI, LO) \ | |
456 | ( __ASSERT_(UTF8_IS_DOWNGRADEABLE_START(HI)) \ | |
457 | __ASSERT_(UTF8_IS_CONTINUATION(LO)) \ | |
458 | LATIN1_TO_NATIVE(UTF8_ACCUMULATE(( \ | |
459 | NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), (LO)))) | |
460 | ||
94bb8c36 | 461 | /* Convert a two (not one) byte utf8 character to a native code point value. |
2950f2a7 KW |
462 | * Needs just one iteration of accumulate. Should not be used unless it is |
463 | * known that the two bytes are legal: 1) two-byte start, and 2) continuation. | |
464 | * Note that the result can be larger than 255 if the input character is not | |
465 | * downgradable */ | |
94bb8c36 | 466 | #define TWO_BYTE_UTF8_TO_NATIVE(HI, LO) \ |
b18b15ad | 467 | ( __ASSERT_(PL_utf8skip[HI] == 2) \ |
635e76f5 | 468 | __ASSERT_(UTF8_IS_CONTINUATION(LO)) \ |
94bb8c36 | 469 | UNI_TO_NATIVE(UTF8_ACCUMULATE((NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), \ |
635e76f5 | 470 | (LO)))) |
94bb8c36 KW |
471 | |
472 | /* Should never be used, and be deprecated */ | |
473 | #define TWO_BYTE_UTF8_TO_UNI(HI, LO) NATIVE_TO_UNI(TWO_BYTE_UTF8_TO_NATIVE(HI, LO)) | |
2950f2a7 | 474 | |
bd18bd40 KW |
475 | /* |
476 | ||
477 | =for apidoc Am|STRLEN|UTF8SKIP|char* s | |
478 | returns the number of bytes in the UTF-8 encoded character whose first (perhaps | |
479 | only) byte is pointed to by C<s>. | |
480 | ||
481 | =cut | |
482 | */ | |
2a70536e KW |
483 | #define UTF8SKIP(s) PL_utf8skip[*(const U8*)(s)] |
484 | #define UTF8_SKIP(s) UTF8SKIP(s) | |
d06134e5 | 485 | |
2d1545e5 KW |
486 | /* Most code that says 'UNI_' really means the native value for code points up |
487 | * through 255 */ | |
488 | #define UNI_IS_INVARIANT(cp) UVCHR_IS_INVARIANT(cp) | |
489 | ||
15824458 KW |
490 | /* Is the byte 'c' the same character when encoded in UTF-8 as when not. This |
491 | * works on both UTF-8 encoded strings and non-encoded, as it returns TRUE in | |
5fc230f1 KW |
492 | * each for the exact same set of bit patterns. It is valid on a subset of |
493 | * what UVCHR_IS_INVARIANT is valid on, so can just use that; and the compiler | |
494 | * should optimize out anything extraneous given the implementation of the | |
5c06326b KW |
495 | * latter. The |0 makes sure this isn't mistakenly called with a ptr argument. |
496 | * */ | |
497 | #define UTF8_IS_INVARIANT(c) UVCHR_IS_INVARIANT((c) | 0) | |
5fc230f1 KW |
498 | |
499 | /* Like the above, but its name implies a non-UTF8 input, which as the comments | |
500 | * above show, doesn't matter as to its implementation */ | |
38953e5a | 501 | #define NATIVE_BYTE_IS_INVARIANT(c) UVCHR_IS_INVARIANT(c) |
d06134e5 | 502 | |
48ccf5e1 KW |
503 | /* The macros in the next 4 sets are used to generate the two utf8 or utfebcdic |
504 | * bytes from an ordinal that is known to fit into exactly two (not one) bytes; | |
505 | * it must be less than 0x3FF to work across both encodings. */ | |
506 | ||
507 | /* These two are helper macros for the other three sets, and should not be used | |
508 | * directly anywhere else. 'translate_function' is either NATIVE_TO_LATIN1 | |
1ff3baa2 KW |
509 | * (which works for code points up through 0xFF) or NATIVE_TO_UNI which works |
510 | * for any code point */ | |
48ccf5e1 | 511 | #define __BASE_TWO_BYTE_HI(c, translate_function) \ |
2863dafa | 512 | (__ASSERT_(! UVCHR_IS_INVARIANT(c)) \ |
48ccf5e1 | 513 | I8_TO_NATIVE_UTF8((translate_function(c) >> UTF_ACCUMULATION_SHIFT) \ |
2863dafa | 514 | | UTF_START_MARK(2))) |
48ccf5e1 | 515 | #define __BASE_TWO_BYTE_LO(c, translate_function) \ |
2863dafa | 516 | (__ASSERT_(! UVCHR_IS_INVARIANT(c)) \ |
48ccf5e1 | 517 | I8_TO_NATIVE_UTF8((translate_function(c) & UTF_CONTINUATION_MASK) \ |
2863dafa | 518 | | UTF_CONTINUATION_MARK)) |
48ccf5e1 | 519 | |
48ccf5e1 KW |
520 | /* The next two macros should not be used. They were designed to be usable as |
521 | * the case label of a switch statement, but this doesn't work for EBCDIC. Use | |
9d0d3a03 | 522 | * regen/unicode_constants.pl instead */ |
48ccf5e1 KW |
523 | #define UTF8_TWO_BYTE_HI_nocast(c) __BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI) |
524 | #define UTF8_TWO_BYTE_LO_nocast(c) __BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI) | |
525 | ||
526 | /* The next two macros are used when the source should be a single byte | |
527 | * character; checked for under DEBUGGING */ | |
528 | #define UTF8_EIGHT_BIT_HI(c) (__ASSERT_(FITS_IN_8_BITS(c)) \ | |
4c8cd605 | 529 | ( __BASE_TWO_BYTE_HI(c, NATIVE_TO_LATIN1))) |
48ccf5e1 | 530 | #define UTF8_EIGHT_BIT_LO(c) (__ASSERT_(FITS_IN_8_BITS(c)) \ |
4c8cd605 | 531 | (__BASE_TWO_BYTE_LO(c, NATIVE_TO_LATIN1))) |
48ccf5e1 KW |
532 | |
533 | /* These final two macros in the series are used when the source can be any | |
534 | * code point whose UTF-8 is known to occupy 2 bytes; they are less efficient | |
535 | * than the EIGHT_BIT versions on EBCDIC platforms. We use the logical '~' | |
536 | * operator instead of "<=" to avoid getting compiler warnings. | |
d52b8576 | 537 | * MAX_UTF8_TWO_BYTE should be exactly all one bits in the lower few |
48ccf5e1 KW |
538 | * places, so the ~ works */ |
539 | #define UTF8_TWO_BYTE_HI(c) \ | |
540 | (__ASSERT_((sizeof(c) == 1) \ | |
d52b8576 | 541 | || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \ |
4c8cd605 | 542 | (__BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI))) |
48ccf5e1 KW |
543 | #define UTF8_TWO_BYTE_LO(c) \ |
544 | (__ASSERT_((sizeof(c) == 1) \ | |
d52b8576 | 545 | || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \ |
4c8cd605 | 546 | (__BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI))) |
d06134e5 | 547 | |
e7214ce8 KW |
548 | /* This is illegal in any well-formed UTF-8 in both EBCDIC and ASCII |
549 | * as it is only in overlongs. */ | |
550 | #define ILLEGAL_UTF8_BYTE I8_TO_NATIVE_UTF8(0xC1) | |
551 | ||
7e2040f0 | 552 | /* |
e3036cf4 | 553 | * 'UTF' is whether or not p is encoded in UTF8. The names 'foo_lazy_if' stem |
20df05f4 KW |
554 | * from an earlier version of these macros in which they didn't call the |
555 | * foo_utf8() macros (i.e. were 'lazy') unless they decided that *p is the | |
556 | * beginning of a utf8 character. Now that foo_utf8() determines that itself, | |
557 | * no need to do it again here | |
7e2040f0 | 558 | */ |
97d0ceda KW |
559 | #define isIDFIRST_lazy_if(p,UTF) ((IN_BYTES || !UTF) \ |
560 | ? isIDFIRST(*(p)) \ | |
e3036cf4 | 561 | : isIDFIRST_utf8((const U8*)p)) |
97d0ceda KW |
562 | #define isWORDCHAR_lazy_if(p,UTF) ((IN_BYTES || (!UTF)) \ |
563 | ? isWORDCHAR(*(p)) \ | |
32636478 KW |
564 | : isWORDCHAR_utf8((const U8*)p)) |
565 | #define isALNUM_lazy_if(p,UTF) isWORDCHAR_lazy_if(p,UTF) | |
1d72bdf6 | 566 | |
89ebb4a3 JH |
567 | #define UTF8_MAXLEN UTF8_MAXBYTES |
568 | ||
8cb75cc8 KW |
569 | /* A Unicode character can fold to up to 3 characters */ |
570 | #define UTF8_MAX_FOLD_CHAR_EXPAND 3 | |
571 | ||
a98fe34d | 572 | #define IN_BYTES (CopHINTS_get(PL_curcop) & HINT_BYTES) |
bd18bd40 KW |
573 | |
574 | /* | |
575 | ||
576 | =for apidoc Am|bool|DO_UTF8|SV* sv | |
577 | Returns a bool giving whether or not the PV in C<sv> is to be treated as being | |
578 | encoded in UTF-8. | |
579 | ||
580 | You should use this I<after> a call to C<SvPV()> or one of its variants, in | |
581 | case any call to string overloading updates the internal UTF-8 encoding flag. | |
582 | ||
583 | =cut | |
584 | */ | |
0064a8a9 | 585 | #define DO_UTF8(sv) (SvUTF8(sv) && !IN_BYTES) |
1ff3baa2 KW |
586 | |
587 | /* Should all strings be treated as Unicode, and not just UTF-8 encoded ones? | |
588 | * Is so within 'feature unicode_strings' or 'locale :not_characters', and not | |
589 | * within 'use bytes'. UTF-8 locales are not tested for here, but perhaps | |
590 | * could be */ | |
591 | #define IN_UNI_8_BIT \ | |
d6ded950 KW |
592 | (((CopHINTS_get(PL_curcop) & (HINT_UNI_8_BIT)) \ |
593 | || (CopHINTS_get(PL_curcop) & HINT_LOCALE_PARTIAL \ | |
594 | /* -1 below is for :not_characters */ \ | |
595 | && _is_in_locale_category(FALSE, -1))) \ | |
596 | && ! IN_BYTES) | |
b36bf33f | 597 | |
1d72bdf6 | 598 | |
c76687c5 KW |
599 | #define UTF8_ALLOW_EMPTY 0x0001 /* Allow a zero length string */ |
600 | ||
601 | /* Allow first byte to be a continuation byte */ | |
1d72bdf6 | 602 | #define UTF8_ALLOW_CONTINUATION 0x0002 |
c76687c5 KW |
603 | |
604 | /* Allow second... bytes to be non-continuation bytes */ | |
1d72bdf6 | 605 | #define UTF8_ALLOW_NON_CONTINUATION 0x0004 |
949cf498 KW |
606 | |
607 | /* expecting more bytes than were available in the string */ | |
608 | #define UTF8_ALLOW_SHORT 0x0008 | |
609 | ||
610 | /* Overlong sequence; i.e., the code point can be specified in fewer bytes. */ | |
611 | #define UTF8_ALLOW_LONG 0x0010 | |
612 | ||
613 | #define UTF8_DISALLOW_SURROGATE 0x0020 /* Unicode surrogates */ | |
614 | #define UTF8_WARN_SURROGATE 0x0040 | |
615 | ||
616 | #define UTF8_DISALLOW_NONCHAR 0x0080 /* Unicode non-character */ | |
617 | #define UTF8_WARN_NONCHAR 0x0100 /* code points */ | |
618 | ||
619 | #define UTF8_DISALLOW_SUPER 0x0200 /* Super-set of Unicode: code */ | |
620 | #define UTF8_WARN_SUPER 0x0400 /* points above the legal max */ | |
621 | ||
d35f2ca5 KW |
622 | /* Code points which never were part of the original UTF-8 standard, which only |
623 | * went up to 2 ** 31 - 1. Note that these all overflow a signed 32-bit word, | |
624 | * The first byte of these code points is FE or FF on ASCII platforms. If the | |
625 | * first byte is FF, it will overflow a 32-bit word. */ | |
626 | #define UTF8_DISALLOW_ABOVE_31_BIT 0x0800 | |
627 | #define UTF8_WARN_ABOVE_31_BIT 0x1000 | |
628 | ||
629 | /* For back compat, these old names are misleading for UTF_EBCDIC */ | |
630 | #define UTF8_DISALLOW_FE_FF UTF8_DISALLOW_ABOVE_31_BIT | |
631 | #define UTF8_WARN_FE_FF UTF8_WARN_ABOVE_31_BIT | |
949cf498 KW |
632 | |
633 | #define UTF8_CHECK_ONLY 0x2000 | |
634 | ||
635 | /* For backwards source compatibility. They do nothing, as the default now | |
636 | * includes what they used to mean. The first one's meaning was to allow the | |
637 | * just the single non-character 0xFFFF */ | |
638 | #define UTF8_ALLOW_FFFF 0 | |
639 | #define UTF8_ALLOW_SURROGATE 0 | |
640 | ||
d35f2ca5 | 641 | #define UTF8_DISALLOW_ILLEGAL_INTERCHANGE \ |
93e6dbd6 KW |
642 | ( UTF8_DISALLOW_SUPER|UTF8_DISALLOW_NONCHAR \ |
643 | |UTF8_DISALLOW_SURROGATE) | |
949cf498 | 644 | #define UTF8_WARN_ILLEGAL_INTERCHANGE \ |
93e6dbd6 KW |
645 | (UTF8_WARN_SUPER|UTF8_WARN_NONCHAR|UTF8_WARN_SURROGATE) |
646 | #define UTF8_ALLOW_ANY \ | |
647 | (~( UTF8_DISALLOW_ILLEGAL_INTERCHANGE|UTF8_DISALLOW_ABOVE_31_BIT \ | |
648 | |UTF8_WARN_ILLEGAL_INTERCHANGE|UTF8_WARN_ABOVE_31_BIT)) | |
949cf498 KW |
649 | #define UTF8_ALLOW_ANYUV \ |
650 | (UTF8_ALLOW_EMPTY \ | |
651 | & ~(UTF8_DISALLOW_ILLEGAL_INTERCHANGE|UTF8_WARN_ILLEGAL_INTERCHANGE)) | |
9f7f3913 TS |
652 | #define UTF8_ALLOW_DEFAULT (ckWARN(WARN_UTF8) ? 0 : \ |
653 | UTF8_ALLOW_ANYUV) | |
1d72bdf6 | 654 | |
89d986df KW |
655 | /* |
656 | =for apidoc Am|bool|UTF8_IS_SURROGATE|const U8 *s|const U8 *e | |
657 | ||
658 | Evaluates to non-zero if the first few bytes of the string starting at C<s> and | |
659 | looking no further than S<C<e - 1>> are well-formed UTF-8 that represents one | |
660 | of the Unicode surrogate code points; otherwise it evaluates to 0. If | |
661 | non-zero, the value gives how many bytes starting at C<s> comprise the code | |
662 | point's representation. | |
663 | ||
664 | =cut | |
665 | */ | |
666 | #define UTF8_IS_SURROGATE(s, e) is_SURROGATE_utf8_safe(s, e) | |
667 | ||
668 | ||
669 | #define UTF8_IS_REPLACEMENT(s, send) is_REPLACEMENT_utf8_safe(s,send) | |
670 | ||
671 | /* | |
672 | =for apidoc Am|bool|UTF8_IS_SUPER|const U8 *s|const U8 *e | |
673 | ||
674 | Recall that Perl recognizes an extension to UTF-8 that can encode code | |
675 | points larger than the ones defined by Unicode, which are 0..0x10FFFF. | |
676 | ||
677 | This macro evaluates to non-zero if the first few bytes of the string starting | |
678 | at C<s> and looking no further than S<C<e - 1>> are from this UTF-8 extension; | |
679 | otherwise it evaluates to 0. If non-zero, the value gives how many bytes | |
680 | starting at C<s> comprise the code point's representation. | |
681 | ||
682 | 0 is returned if the bytes are not well-formed extended UTF-8, or if they | |
683 | represent a code point that cannot fit in a UV on the current platform. Hence | |
684 | this macro can give different results when run on a 64-bit word machine than on | |
685 | one with a 32-bit word size. | |
0c58a72b | 686 | |
89d986df KW |
687 | Note that it is deprecated to have code points that are larger than what can |
688 | fit in an IV on the current machine. | |
7131f24d | 689 | |
89d986df KW |
690 | =cut |
691 | ||
692 | * ASCII EBCDIC I8 | |
7131f24d KW |
693 | * U+10FFFF: \xF4\x8F\xBF\xBF \xF9\xA1\xBF\xBF\xBF max legal Unicode |
694 | * U+110000: \xF4\x90\x80\x80 \xF9\xA2\xA0\xA0\xA0 | |
695 | * U+110001: \xF4\x90\x80\x81 \xF9\xA2\xA0\xA0\xA1 | |
89d986df KW |
696 | */ |
697 | #ifdef EBCDIC | |
a14e0a36 | 698 | # define UTF8_IS_SUPER(s, e) \ |
89d986df KW |
699 | (( LIKELY((e) > (s) + 4) \ |
700 | && NATIVE_UTF8_TO_I8(*(s)) >= 0xF9 \ | |
701 | && ( NATIVE_UTF8_TO_I8(*(s)) > 0xF9 \ | |
702 | || (NATIVE_UTF8_TO_I8(*(s) + 1) >= 0xA2)) \ | |
703 | && LIKELY((s) + UTF8SKIP(s) <= (e))) \ | |
704 | ? _is_utf8_char_slow(s, UTF8SKIP(s)) : 0) | |
7131f24d | 705 | #else |
a14e0a36 | 706 | # define UTF8_IS_SUPER(s, e) \ |
89d986df KW |
707 | (( LIKELY((e) > (s) + 3) \ |
708 | && (*(U8*) (s)) >= 0xF4 \ | |
709 | && ((*(U8*) (s)) > 0xF4 || (*((U8*) (s) + 1) >= 0x90))\ | |
710 | && LIKELY((s) + UTF8SKIP(s) <= (e))) \ | |
711 | ? _is_utf8_char_slow(s, UTF8SKIP(s)) : 0) | |
7131f24d KW |
712 | #endif |
713 | ||
b96a92fb KW |
714 | /* These are now machine generated, and the 'given' clause is no longer |
715 | * applicable */ | |
0c58a72b | 716 | #define UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s, e) \ |
89d986df KW |
717 | cBOOL(is_NONCHAR_utf8_safe(s,e)) |
718 | ||
719 | /* | |
720 | =for apidoc Am|bool|UTF8_IS_NONCHAR|const U8 *s|const U8 *e | |
721 | ||
722 | Evaluates to non-zero if the first few bytes of the string starting at C<s> and | |
723 | looking no further than S<C<e - 1>> are well-formed UTF-8 that represents one | |
724 | of the Unicode non-character code points; otherwise it evaluates to 0. If | |
725 | non-zero, the value gives how many bytes starting at C<s> comprise the code | |
726 | point's representation. | |
727 | ||
728 | =cut | |
729 | */ | |
0c58a72b KW |
730 | #define UTF8_IS_NONCHAR(s, e) \ |
731 | UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s, e) | |
7131f24d | 732 | |
c867b360 JH |
733 | #define UNICODE_SURROGATE_FIRST 0xD800 |
734 | #define UNICODE_SURROGATE_LAST 0xDFFF | |
735 | #define UNICODE_REPLACEMENT 0xFFFD | |
736 | #define UNICODE_BYTE_ORDER_MARK 0xFEFF | |
1d72bdf6 | 737 | |
b851fbc1 | 738 | /* Though our UTF-8 encoding can go beyond this, |
c76687c5 | 739 | * let's be conservative and do as Unicode says. */ |
b851fbc1 JH |
740 | #define PERL_UNICODE_MAX 0x10FFFF |
741 | ||
d35f2ca5 KW |
742 | #define UNICODE_WARN_SURROGATE 0x0001 /* UTF-16 surrogates */ |
743 | #define UNICODE_WARN_NONCHAR 0x0002 /* Non-char code points */ | |
744 | #define UNICODE_WARN_SUPER 0x0004 /* Above 0x10FFFF */ | |
745 | #define UNICODE_WARN_ABOVE_31_BIT 0x0008 /* Above 0x7FFF_FFFF */ | |
746 | #define UNICODE_DISALLOW_SURROGATE 0x0010 | |
747 | #define UNICODE_DISALLOW_NONCHAR 0x0020 | |
748 | #define UNICODE_DISALLOW_SUPER 0x0040 | |
749 | #define UNICODE_DISALLOW_ABOVE_31_BIT 0x0080 | |
bb88be5f KW |
750 | #define UNICODE_WARN_ILLEGAL_INTERCHANGE \ |
751 | (UNICODE_WARN_SURROGATE|UNICODE_WARN_NONCHAR|UNICODE_WARN_SUPER) | |
752 | #define UNICODE_DISALLOW_ILLEGAL_INTERCHANGE \ | |
753 | (UNICODE_DISALLOW_SURROGATE|UNICODE_DISALLOW_NONCHAR|UNICODE_DISALLOW_SUPER) | |
949cf498 KW |
754 | |
755 | /* For backward source compatibility, as are now the default */ | |
756 | #define UNICODE_ALLOW_SURROGATE 0 | |
757 | #define UNICODE_ALLOW_SUPER 0 | |
758 | #define UNICODE_ALLOW_ANY 0 | |
b851fbc1 | 759 | |
2d6b3d38 KW |
760 | /* This matches the 2048 code points between UNICODE_SURROGATE_FIRST (0xD800) and |
761 | * UNICODE_SURROGATE_LAST (0xDFFF) */ | |
762 | #define UNICODE_IS_SURROGATE(uv) (((UV) (uv) & (~0xFFFF | 0xF800)) \ | |
763 | == 0xD800) | |
764 | ||
646d1759 KW |
765 | #define UNICODE_IS_REPLACEMENT(uv) ((UV) (uv) == UNICODE_REPLACEMENT) |
766 | #define UNICODE_IS_BYTE_ORDER_MARK(uv) ((UV) (uv) == UNICODE_BYTE_ORDER_MARK) | |
c149ab20 KW |
767 | |
768 | /* Is 'uv' one of the 32 contiguous-range noncharacters? */ | |
769 | #define UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv) ((UV) (uv) >= 0xFDD0 \ | |
770 | && (UV) (uv) <= 0xFDEF) | |
771 | ||
772 | /* Is 'uv' one of the 34 plane-ending noncharacters 0xFFFE, 0xFFFF, 0x1FFFE, | |
773 | * 0x1FFFF, ... 0x10FFFE, 0x10FFFF, given that we know that 'uv' is not above | |
774 | * the Unicode legal max */ | |
775 | #define UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv) \ | |
776 | (((UV) (uv) & 0xFFFE) == 0xFFFE) | |
777 | ||
778 | #define UNICODE_IS_NONCHAR(uv) \ | |
779 | ( UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv) \ | |
780 | || ( LIKELY( ! UNICODE_IS_SUPER(uv)) \ | |
781 | && UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv))) | |
782 | ||
783 | #define UNICODE_IS_SUPER(uv) ((UV) (uv) > PERL_UNICODE_MAX) | |
d35f2ca5 | 784 | #define UNICODE_IS_ABOVE_31_BIT(uv) ((UV) (uv) > 0x7FFFFFFF) |
1d72bdf6 | 785 | |
ec34087a KW |
786 | #define LATIN_SMALL_LETTER_SHARP_S LATIN_SMALL_LETTER_SHARP_S_NATIVE |
787 | #define LATIN_SMALL_LETTER_Y_WITH_DIAERESIS \ | |
788 | LATIN_SMALL_LETTER_Y_WITH_DIAERESIS_NATIVE | |
789 | #define MICRO_SIGN MICRO_SIGN_NATIVE | |
790 | #define LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE \ | |
791 | LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE_NATIVE | |
792 | #define LATIN_SMALL_LETTER_A_WITH_RING_ABOVE \ | |
793 | LATIN_SMALL_LETTER_A_WITH_RING_ABOVE_NATIVE | |
09091399 JH |
794 | #define UNICODE_GREEK_CAPITAL_LETTER_SIGMA 0x03A3 |
795 | #define UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA 0x03C2 | |
796 | #define UNICODE_GREEK_SMALL_LETTER_SIGMA 0x03C3 | |
9dcbe121 | 797 | #define GREEK_SMALL_LETTER_MU 0x03BC |
9e682c18 KW |
798 | #define GREEK_CAPITAL_LETTER_MU 0x039C /* Upper and title case |
799 | of MICRON */ | |
800 | #define LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS 0x0178 /* Also is title case */ | |
0766489e KW |
801 | #ifdef LATIN_CAPITAL_LETTER_SHARP_S_UTF8 |
802 | # define LATIN_CAPITAL_LETTER_SHARP_S 0x1E9E | |
803 | #endif | |
74894415 KW |
804 | #define LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE 0x130 |
805 | #define LATIN_SMALL_LETTER_DOTLESS_I 0x131 | |
9e682c18 | 806 | #define LATIN_SMALL_LETTER_LONG_S 0x017F |
a9f50d33 KW |
807 | #define LATIN_SMALL_LIGATURE_LONG_S_T 0xFB05 |
808 | #define LATIN_SMALL_LIGATURE_ST 0xFB06 | |
9e682c18 KW |
809 | #define KELVIN_SIGN 0x212A |
810 | #define ANGSTROM_SIGN 0x212B | |
09091399 | 811 | |
9e55ce06 | 812 | #define UNI_DISPLAY_ISPRINT 0x0001 |
c728cb41 JH |
813 | #define UNI_DISPLAY_BACKSLASH 0x0002 |
814 | #define UNI_DISPLAY_QQ (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH) | |
815 | #define UNI_DISPLAY_REGEX (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH) | |
9e55ce06 | 816 | |
5cd46e1f KW |
817 | #define ANYOF_FOLD_SHARP_S(node, input, end) \ |
818 | (ANYOF_BITMAP_TEST(node, LATIN_SMALL_LETTER_SHARP_S) && \ | |
137165a6 | 819 | (ANYOF_NONBITMAP(node)) && \ |
39065660 | 820 | (ANYOF_FLAGS(node) & ANYOF_LOC_NONBITMAP_FOLD) && \ |
07b6858f | 821 | ((end) > (input) + 1) && \ |
305b8651 | 822 | isALPHA_FOLD_EQ((input)[0], 's')) |
6302f837 | 823 | |
ebc501f0 | 824 | #define SHARP_S_SKIP 2 |
3b0fc154 | 825 | |
a4f7a67c KW |
826 | /* If you want to exclude surrogates, and beyond legal Unicode, see the blame |
827 | * log for earlier versions which gave details for these */ | |
4d646140 | 828 | |
6302f837 KW |
829 | /* A helper macro for isUTF8_CHAR, so use that one, and not this one. This is |
830 | * retained solely for backwards compatibility and may be deprecated and | |
831 | * removed in a future Perl version. | |
832 | * | |
833 | * regen/regcharclass.pl generates is_UTF8_CHAR_utf8() macros for up to these | |
d9f92374 KW |
834 | * number of bytes. So this has to be coordinated with that file */ |
835 | #ifdef EBCDIC | |
836 | # define IS_UTF8_CHAR_FAST(n) ((n) <= 3) | |
837 | #else | |
838 | # define IS_UTF8_CHAR_FAST(n) ((n) <= 4) | |
839 | #endif | |
840 | ||
4d646140 | 841 | #ifndef EBCDIC |
6302f837 | 842 | /* A helper macro for isUTF8_CHAR, so use that one instead of this. This was |
c24e28a3 KW |
843 | * generated by regen/regcharclass.pl, and then moved here. Then it was |
844 | * hand-edited to add some LIKELY() calls, presuming that malformations are | |
845 | * unlikely. The lines that generated it were then commented out. This was | |
846 | * done because it takes on the order of 10 minutes to generate, and is never | |
847 | * going to change, unless the generated code is improved, and figuring out | |
848 | * there the LIKELYs would be hard. | |
6302f837 | 849 | * |
1ff3baa2 KW |
850 | * The EBCDIC versions have been cut to not cover all of legal Unicode, |
851 | * otherwise they take too long to generate; besides there is a separate one | |
852 | * for each code page, so they are in regcharclass.h instead of here */ | |
39a0f513 | 853 | /* |
5dca9278 | 854 | UTF8_CHAR: Matches legal UTF-8 encoded characters from 2 through 4 bytes |
39a0f513 | 855 | |
5dca9278 | 856 | 0x80 - 0x1FFFFF |
39a0f513 | 857 | */ |
4d646140 | 858 | /*** GENERATED CODE ***/ |
5dca9278 KW |
859 | #define is_UTF8_CHAR_utf8_no_length_checks(s) \ |
860 | ( ( 0xC2 <= ((U8*)s)[0] && ((U8*)s)[0] <= 0xDF ) ? \ | |
c24e28a3 | 861 | ( LIKELY( ( ((U8*)s)[1] & 0xC0 ) == 0x80 ) ? 2 : 0 ) \ |
5dca9278 | 862 | : ( 0xE0 == ((U8*)s)[0] ) ? \ |
c24e28a3 | 863 | ( LIKELY( ( ( ((U8*)s)[1] & 0xE0 ) == 0xA0 ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\ |
5dca9278 | 864 | : ( 0xE1 <= ((U8*)s)[0] && ((U8*)s)[0] <= 0xEF ) ? \ |
c24e28a3 | 865 | ( LIKELY( ( ( ((U8*)s)[1] & 0xC0 ) == 0x80 ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\ |
5dca9278 | 866 | : ( 0xF0 == ((U8*)s)[0] ) ? \ |
c24e28a3 KW |
867 | ( LIKELY( ( ( 0x90 <= ((U8*)s)[1] && ((U8*)s)[1] <= 0xBF ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) && ( ( ((U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\ |
868 | : ( ( ( ( 0xF1 <= ((U8*)s)[0] && ((U8*)s)[0] <= 0xF7 ) && LIKELY( ( ((U8*)s)[1] & 0xC0 ) == 0x80 ) ) && LIKELY( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) && LIKELY( ( ((U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 ) | |
4d646140 | 869 | #endif |
3b0fc154 | 870 | |
6302f837 | 871 | /* |
5dca9278 KW |
872 | |
873 | =for apidoc Am|STRLEN|isUTF8_CHAR|const U8 *s|const U8 *e | |
874 | ||
35f8c9bd KW |
875 | Evaluates to non-zero if the first few bytes of the string starting at C<s> and |
876 | looking no further than S<C<e - 1>> are well-formed UTF-8 that represents some | |
877 | code point; otherwise it evaluates to 0. If non-zero, the value gives how many | |
878 | many bytes starting at C<s> comprise the code point's representation. | |
6302f837 | 879 | |
35f8c9bd KW |
880 | The code point can be any that will fit in a UV on this machine, using Perl's |
881 | extension to official UTF-8 to represent those higher than the Unicode maximum | |
882 | of 0x10FFFF. That means that this macro is used to efficiently decide if the | |
883 | next few bytes in C<s> is legal UTF-8 for a single character. Use | |
884 | L</is_utf8_string>(), L</is_utf8_string_loclen>(), and | |
885 | L</is_utf8_string_loc>() to check entire strings. | |
886 | ||
887 | Note that it is deprecated to use code points higher than what will fit in an | |
888 | IV. This macro does not raise any warnings for such code points, treating them | |
889 | as valid. | |
890 | ||
891 | Note also that a UTF-8 INVARIANT character (i.e. ASCII on non-EBCDIC machines) | |
892 | is a valid UTF-8 character. | |
5dca9278 KW |
893 | |
894 | =cut | |
895 | */ | |
6302f837 | 896 | |
dd9bc2b0 | 897 | #define isUTF8_CHAR(s, e) (UNLIKELY((e) <= (s)) \ |
6302f837 KW |
898 | ? 0 \ |
899 | : (UTF8_IS_INVARIANT(*s)) \ | |
900 | ? 1 \ | |
dd9bc2b0 | 901 | : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \ |
6302f837 | 902 | ? 0 \ |
dd9bc2b0 | 903 | : LIKELY(IS_UTF8_CHAR_FAST(UTF8SKIP(s))) \ |
5dca9278 | 904 | ? is_UTF8_CHAR_utf8_no_length_checks(s) \ |
35f8c9bd | 905 | : _is_utf8_char_slow(s, UTF8SKIP(s))) |
6302f837 | 906 | |
3cedd9d9 KW |
907 | #define is_utf8_char_buf(buf, buf_end) isUTF8_CHAR(buf, buf_end) |
908 | ||
6302f837 KW |
909 | /* Do not use; should be deprecated. Use isUTF8_CHAR() instead; this is |
910 | * retained solely for backwards compatibility */ | |
911 | #define IS_UTF8_CHAR(p, n) (isUTF8_CHAR(p, (p) + (n)) == n) | |
e9a8c099 | 912 | |
57f0e7e2 KW |
913 | #endif /* H_UTF8 */ |
914 | ||
e9a8c099 | 915 | /* |
14d04a33 | 916 | * ex: set ts=8 sts=4 sw=4 et: |
e9a8c099 | 917 | */ |