| 1 | /* utf8.h |
| 2 | * |
| 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 | * |
| 9 | * Copyright (C) 2000, 2001, 2002, 2005, 2006, 2007, 2009, |
| 10 | * 2010, 2011 by Larry Wall and others |
| 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 | |
| 17 | #ifndef PERL_UTF8_H_ /* Guard against recursive inclusion */ |
| 18 | #define PERL_UTF8_H_ 1 |
| 19 | |
| 20 | /* Use UTF-8 as the default script encoding? |
| 21 | * Turning this on will break scripts having non-UTF-8 binary |
| 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 | |
| 29 | #include "regcharclass.h" |
| 30 | #include "unicode_constants.h" |
| 31 | |
| 32 | /* For to_utf8_fold_flags, q.v. */ |
| 33 | #define FOLD_FLAGS_LOCALE 0x1 |
| 34 | #define FOLD_FLAGS_FULL 0x2 |
| 35 | #define FOLD_FLAGS_NOMIX_ASCII 0x4 |
| 36 | |
| 37 | /* |
| 38 | =head1 Unicode Support |
| 39 | L<perlguts/Unicode Support> has an introduction to this API. |
| 40 | |
| 41 | See also L</Character classification>, |
| 42 | and L</Character case changing>. |
| 43 | Various functions outside this section also work specially with Unicode. |
| 44 | Search for the string "utf8" in this document. |
| 45 | |
| 46 | =for apidoc is_ascii_string |
| 47 | |
| 48 | This is a misleadingly-named synonym for L</is_utf8_invariant_string>. |
| 49 | On ASCII-ish platforms, the name isn't misleading: the ASCII-range characters |
| 50 | are exactly the UTF-8 invariants. But EBCDIC machines have more invariants |
| 51 | than just the ASCII characters, so C<is_utf8_invariant_string> is preferred. |
| 52 | |
| 53 | =for apidoc is_invariant_string |
| 54 | |
| 55 | This is a somewhat misleadingly-named synonym for L</is_utf8_invariant_string>. |
| 56 | C<is_utf8_invariant_string> is preferred, as it indicates under what conditions |
| 57 | the string is invariant. |
| 58 | |
| 59 | =cut |
| 60 | */ |
| 61 | #define is_ascii_string(s, len) is_utf8_invariant_string(s, len) |
| 62 | #define is_invariant_string(s, len) is_utf8_invariant_string(s, len) |
| 63 | |
| 64 | #define uvoffuni_to_utf8_flags(d,uv,flags) \ |
| 65 | uvoffuni_to_utf8_flags_msgs(d, uv, flags, 0) |
| 66 | #define uvchr_to_utf8(a,b) uvchr_to_utf8_flags(a,b,0) |
| 67 | #define uvchr_to_utf8_flags(d,uv,flags) \ |
| 68 | uvchr_to_utf8_flags_msgs(d,uv,flags, 0) |
| 69 | #define uvchr_to_utf8_flags_msgs(d,uv,flags,msgs) \ |
| 70 | uvoffuni_to_utf8_flags_msgs(d,NATIVE_TO_UNI(uv),flags, msgs) |
| 71 | #define utf8_to_uvchr_buf(s, e, lenp) \ |
| 72 | utf8_to_uvchr_buf_helper((const U8 *) (s), (const U8 *) e, lenp) |
| 73 | #define utf8n_to_uvchr(s, len, lenp, flags) \ |
| 74 | utf8n_to_uvchr_error(s, len, lenp, flags, 0) |
| 75 | #define utf8n_to_uvchr_error(s, len, lenp, flags, errors) \ |
| 76 | utf8n_to_uvchr_msgs(s, len, lenp, flags, errors, 0) |
| 77 | |
| 78 | #define to_uni_fold(c, p, lenp) _to_uni_fold_flags(c, p, lenp, FOLD_FLAGS_FULL) |
| 79 | |
| 80 | #define foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \ |
| 81 | foldEQ_utf8_flags(s1, pe1, l1, u1, s2, pe2, l2, u2, 0) |
| 82 | #define FOLDEQ_UTF8_NOMIX_ASCII (1 << 0) |
| 83 | #define FOLDEQ_LOCALE (1 << 1) |
| 84 | #define FOLDEQ_S1_ALREADY_FOLDED (1 << 2) |
| 85 | #define FOLDEQ_S2_ALREADY_FOLDED (1 << 3) |
| 86 | #define FOLDEQ_S1_FOLDS_SANE (1 << 4) |
| 87 | #define FOLDEQ_S2_FOLDS_SANE (1 << 5) |
| 88 | |
| 89 | #define ibcmp_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \ |
| 90 | cBOOL(! foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2)) |
| 91 | |
| 92 | #ifdef EBCDIC |
| 93 | /* The equivalent of these macros but implementing UTF-EBCDIC |
| 94 | are in the following header file: |
| 95 | */ |
| 96 | |
| 97 | #include "utfebcdic.h" |
| 98 | |
| 99 | #else /* ! EBCDIC */ |
| 100 | START_EXTERN_C |
| 101 | |
| 102 | /* |
| 103 | |
| 104 | =for apidoc AmnU|STRLEN|UTF8_MAXBYTES |
| 105 | |
| 106 | The maximum width of a single UTF-8 encoded character, in bytes. |
| 107 | |
| 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 | |
| 113 | =cut |
| 114 | */ |
| 115 | #define UTF8_MAXBYTES 13 |
| 116 | |
| 117 | #ifdef DOINIT |
| 118 | EXTCONST unsigned char PL_utf8skip[] = { |
| 119 | /* 0x00 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */ |
| 120 | /* 0x10 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */ |
| 121 | /* 0x20 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */ |
| 122 | /* 0x30 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */ |
| 123 | /* 0x40 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */ |
| 124 | /* 0x50 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */ |
| 125 | /* 0x60 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */ |
| 126 | /* 0x70 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */ |
| 127 | /* 0x80 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */ |
| 128 | /* 0x90 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */ |
| 129 | /* 0xA0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */ |
| 130 | /* 0xB0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */ |
| 131 | /* 0xC0 */ 2,2, /* overlong */ |
| 132 | /* 0xC2 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0080 to U+03FF */ |
| 133 | /* 0xD0 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0400 to U+07FF */ |
| 134 | /* 0xE0 */ 3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3, /* U+0800 to U+FFFF */ |
| 135 | /* 0xF0 */ 4,4,4,4,4,4,4,4,5,5,5,5,6,6, /* above BMP to 2**31 - 1 */ |
| 136 | /* Perl extended (never was official UTF-8). Up to 36 bit */ |
| 137 | /* 0xFE */ 7, |
| 138 | /* More extended, Up to 72 bits (64-bit + reserved) */ |
| 139 | /* 0xFF */ UTF8_MAXBYTES |
| 140 | }; |
| 141 | #else |
| 142 | EXTCONST unsigned char PL_utf8skip[]; |
| 143 | #endif |
| 144 | |
| 145 | END_EXTERN_C |
| 146 | |
| 147 | #if defined(_MSC_VER) && _MSC_VER < 1400 |
| 148 | /* older MSVC versions have a smallish macro buffer */ |
| 149 | #define PERL_SMALL_MACRO_BUFFER |
| 150 | #endif |
| 151 | |
| 152 | /* |
| 153 | |
| 154 | =for apidoc Am|U8|NATIVE_TO_LATIN1|U8 ch |
| 155 | |
| 156 | Returns the Latin-1 (including ASCII and control characters) equivalent of the |
| 157 | input native code point given by C<ch>. Thus, C<NATIVE_TO_LATIN1(193)> on |
| 158 | EBCDIC platforms returns 65. These each represent the character C<"A"> on |
| 159 | their respective platforms. On ASCII platforms no conversion is needed, so |
| 160 | this macro expands to just its input, adding no time nor space requirements to |
| 161 | the implementation. |
| 162 | |
| 163 | For conversion of code points potentially larger than will fit in a character, |
| 164 | use L</NATIVE_TO_UNI>. |
| 165 | |
| 166 | =for apidoc Am|U8|LATIN1_TO_NATIVE|U8 ch |
| 167 | |
| 168 | Returns the native equivalent of the input Latin-1 code point (including ASCII |
| 169 | and control characters) given by C<ch>. Thus, C<LATIN1_TO_NATIVE(66)> on |
| 170 | EBCDIC platforms returns 194. These each represent the character C<"B"> on |
| 171 | their respective platforms. On ASCII platforms no conversion is needed, so |
| 172 | this macro expands to just its input, adding no time nor space requirements to |
| 173 | the implementation. |
| 174 | |
| 175 | For conversion of code points potentially larger than will fit in a character, |
| 176 | use L</UNI_TO_NATIVE>. |
| 177 | |
| 178 | =for apidoc Am|UV|NATIVE_TO_UNI|UV ch |
| 179 | |
| 180 | Returns the Unicode equivalent of the input native code point given by C<ch>. |
| 181 | Thus, C<NATIVE_TO_UNI(195)> on EBCDIC platforms returns 67. These each |
| 182 | represent the character C<"C"> on their respective platforms. On ASCII |
| 183 | platforms no conversion is needed, so this macro expands to just its input, |
| 184 | adding no time nor space requirements to the implementation. |
| 185 | |
| 186 | =for apidoc Am|UV|UNI_TO_NATIVE|UV ch |
| 187 | |
| 188 | Returns the native equivalent of the input Unicode code point given by C<ch>. |
| 189 | Thus, C<UNI_TO_NATIVE(68)> on EBCDIC platforms returns 196. These each |
| 190 | represent the character C<"D"> on their respective platforms. On ASCII |
| 191 | platforms no conversion is needed, so this macro expands to just its input, |
| 192 | adding no time nor space requirements to the implementation. |
| 193 | |
| 194 | =cut |
| 195 | */ |
| 196 | |
| 197 | #ifdef PERL_SMALL_MACRO_BUFFER |
| 198 | # define NATIVE_TO_LATIN1(ch) ((U8)(ch)) |
| 199 | # define LATIN1_TO_NATIVE(ch) ((U8)(ch)) |
| 200 | #else |
| 201 | # define NATIVE_TO_LATIN1(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) ((ch) | 0))) |
| 202 | # define LATIN1_TO_NATIVE(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) ((ch) | 0))) |
| 203 | #endif |
| 204 | |
| 205 | /* I8 is an intermediate version of UTF-8 used only in UTF-EBCDIC. We thus |
| 206 | * consider it to be identical to UTF-8 on ASCII platforms. Strictly speaking |
| 207 | * UTF-8 and UTF-EBCDIC are two different things, but we often conflate them |
| 208 | * because they are 8-bit encodings that serve the same purpose in Perl, and |
| 209 | * rarely do we need to distinguish them. The term "NATIVE_UTF8" applies to |
| 210 | * whichever one is applicable on the current platform */ |
| 211 | #ifdef PERL_SMALL_MACRO_BUFFER |
| 212 | #define NATIVE_UTF8_TO_I8(ch) ((U8) (ch)) |
| 213 | #define I8_TO_NATIVE_UTF8(ch) ((U8) (ch)) |
| 214 | #else |
| 215 | #define NATIVE_UTF8_TO_I8(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) ((ch) | 0))) |
| 216 | #define I8_TO_NATIVE_UTF8(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) ((ch) | 0))) |
| 217 | #endif |
| 218 | |
| 219 | #define UNI_TO_NATIVE(ch) ((UV) ((ch) | 0)) |
| 220 | #define NATIVE_TO_UNI(ch) ((UV) ((ch) | 0)) |
| 221 | |
| 222 | /* |
| 223 | |
| 224 | The following table is from Unicode 3.2, plus the Perl extensions for above |
| 225 | U+10FFFF |
| 226 | |
| 227 | Code Points 1st Byte 2nd Byte 3rd 4th 5th 6th 7th 8th-13th |
| 228 | |
| 229 | U+0000..U+007F 00..7F |
| 230 | U+0080..U+07FF * C2..DF 80..BF |
| 231 | U+0800..U+0FFF E0 * A0..BF 80..BF |
| 232 | U+1000..U+CFFF E1..EC 80..BF 80..BF |
| 233 | U+D000..U+D7FF ED 80..9F 80..BF |
| 234 | U+D800..U+DFFF ED A0..BF 80..BF (surrogates) |
| 235 | U+E000..U+FFFF EE..EF 80..BF 80..BF |
| 236 | U+10000..U+3FFFF F0 * 90..BF 80..BF 80..BF |
| 237 | U+40000..U+FFFFF F1..F3 80..BF 80..BF 80..BF |
| 238 | U+100000..U+10FFFF F4 80..8F 80..BF 80..BF |
| 239 | Below are above-Unicode code points |
| 240 | U+110000..U+13FFFF F4 90..BF 80..BF 80..BF |
| 241 | U+110000..U+1FFFFF F5..F7 80..BF 80..BF 80..BF |
| 242 | U+200000..U+FFFFFF F8 * 88..BF 80..BF 80..BF 80..BF |
| 243 | U+1000000..U+3FFFFFF F9..FB 80..BF 80..BF 80..BF 80..BF |
| 244 | U+4000000..U+3FFFFFFF FC * 84..BF 80..BF 80..BF 80..BF 80..BF |
| 245 | U+40000000..U+7FFFFFFF FD 80..BF 80..BF 80..BF 80..BF 80..BF |
| 246 | U+80000000..U+FFFFFFFFF FE * 82..BF 80..BF 80..BF 80..BF 80..BF 80..BF |
| 247 | U+1000000000.. FF 80..BF 80..BF 80..BF 80..BF 80..BF * 81..BF 80..BF |
| 248 | |
| 249 | Note the gaps before several of the byte entries above marked by '*'. These are |
| 250 | caused by legal UTF-8 avoiding non-shortest encodings: it is technically |
| 251 | possible to UTF-8-encode a single code point in different ways, but that is |
| 252 | explicitly forbidden, and the shortest possible encoding should always be used |
| 253 | (and that is what Perl does). The non-shortest ones are called 'overlongs'. |
| 254 | |
| 255 | */ |
| 256 | |
| 257 | /* |
| 258 | Another way to look at it, as bits: |
| 259 | |
| 260 | Code Points 1st Byte 2nd Byte 3rd Byte 4th Byte |
| 261 | |
| 262 | 0aaa aaaa 0aaa aaaa |
| 263 | 0000 0bbb bbaa aaaa 110b bbbb 10aa aaaa |
| 264 | cccc bbbb bbaa aaaa 1110 cccc 10bb bbbb 10aa aaaa |
| 265 | 00 000d ddcc cccc bbbb bbaa aaaa 1111 0ddd 10cc cccc 10bb bbbb 10aa aaaa |
| 266 | |
| 267 | As you can see, the continuation bytes all begin with C<10>, and the |
| 268 | leading bits of the start byte tell how many bytes there are in the |
| 269 | encoded character. |
| 270 | |
| 271 | Perl's extended UTF-8 means we can have start bytes up through FF, though any |
| 272 | beginning with FF yields a code point that is too large for 32-bit ASCII |
| 273 | platforms. FF signals to use 13 bytes for the encoded character. This breaks |
| 274 | the paradigm that the number of leading bits gives how many total bytes there |
| 275 | are in the character. */ |
| 276 | |
| 277 | /* This is the number of low-order bits a continuation byte in a UTF-8 encoded |
| 278 | * sequence contributes to the specification of the code point. In the bit |
| 279 | * maps above, you see that the first 2 bits are a constant '10', leaving 6 of |
| 280 | * real information */ |
| 281 | #define UTF_ACCUMULATION_SHIFT 6 |
| 282 | |
| 283 | /* ^? is defined to be DEL on ASCII systems. See the definition of toCTRL() |
| 284 | * for more */ |
| 285 | #define QUESTION_MARK_CTRL DEL_NATIVE |
| 286 | |
| 287 | /* Surrogates, non-character code points and above-Unicode code points are |
| 288 | * problematic in some contexts. This allows code that needs to check for |
| 289 | * those to to quickly exclude the vast majority of code points it will |
| 290 | * encounter */ |
| 291 | #define isUTF8_POSSIBLY_PROBLEMATIC(c) (__ASSERT_(FITS_IN_8_BITS(c)) \ |
| 292 | (U8) c >= 0xED) |
| 293 | |
| 294 | #define UNICODE_IS_PERL_EXTENDED(uv) UNLIKELY((UV) (uv) > 0x7FFFFFFF) |
| 295 | |
| 296 | #endif /* EBCDIC vs ASCII */ |
| 297 | |
| 298 | /* 2**UTF_ACCUMULATION_SHIFT - 1. This masks out all but the bits that carry |
| 299 | * real information in a continuation byte. This turns out to be 0x3F in |
| 300 | * UTF-8, 0x1F in UTF-EBCDIC. */ |
| 301 | #define UTF_CONTINUATION_MASK ((U8) ((1U << UTF_ACCUMULATION_SHIFT) - 1)) |
| 302 | |
| 303 | /* For use in UTF8_IS_CONTINUATION(). This turns out to be 0xC0 in UTF-8, |
| 304 | * E0 in UTF-EBCDIC */ |
| 305 | #define UTF_IS_CONTINUATION_MASK ((U8) (0xFF << UTF_ACCUMULATION_SHIFT)) |
| 306 | |
| 307 | /* This defines the bits that are to be in the continuation bytes of a |
| 308 | * multi-byte UTF-8 encoded character that mark it is a continuation byte. |
| 309 | * This turns out to be 0x80 in UTF-8, 0xA0 in UTF-EBCDIC. (khw doesn't know |
| 310 | * the underlying reason that B0 works here) */ |
| 311 | #define UTF_CONTINUATION_MARK (UTF_IS_CONTINUATION_MASK & 0xB0) |
| 312 | |
| 313 | /* Is the byte 'c' part of a multi-byte UTF8-8 encoded sequence, and not the |
| 314 | * first byte thereof? */ |
| 315 | #define UTF8_IS_CONTINUATION(c) (__ASSERT_(FITS_IN_8_BITS(c)) \ |
| 316 | (((NATIVE_UTF8_TO_I8(c) & UTF_IS_CONTINUATION_MASK) \ |
| 317 | == UTF_CONTINUATION_MARK))) |
| 318 | |
| 319 | /* Is the representation of the Unicode code point 'cp' the same regardless of |
| 320 | * being encoded in UTF-8 or not? This is a fundamental property of |
| 321 | * UTF-8,EBCDIC */ |
| 322 | #define OFFUNI_IS_INVARIANT(c) (((WIDEST_UTYPE)(c)) < UTF_CONTINUATION_MARK) |
| 323 | |
| 324 | /* |
| 325 | =for apidoc Am|bool|UVCHR_IS_INVARIANT|UV cp |
| 326 | |
| 327 | Evaluates to 1 if the representation of code point C<cp> is the same whether or |
| 328 | not it is encoded in UTF-8; otherwise evaluates to 0. UTF-8 invariant |
| 329 | characters can be copied as-is when converting to/from UTF-8, saving time. |
| 330 | C<cp> is Unicode if above 255; otherwise is platform-native. |
| 331 | |
| 332 | =cut |
| 333 | */ |
| 334 | #define UVCHR_IS_INVARIANT(cp) (OFFUNI_IS_INVARIANT(NATIVE_TO_UNI(cp))) |
| 335 | |
| 336 | /* Internal macro to be used only in this file to aid in constructing other |
| 337 | * publicly accessible macros. |
| 338 | * The number of bytes required to express this uv in UTF-8, for just those |
| 339 | * uv's requiring 2 through 6 bytes, as these are common to all platforms and |
| 340 | * word sizes. The number of bytes needed is given by the number of leading 1 |
| 341 | * bits in the start byte. There are 32 start bytes that have 2 initial 1 bits |
| 342 | * (C0-DF); there are 16 that have 3 initial 1 bits (E0-EF); 8 that have 4 |
| 343 | * initial 1 bits (F0-F8); 4 that have 5 initial 1 bits (F9-FB), and 2 that |
| 344 | * have 6 initial 1 bits (FC-FD). The largest number a string of n bytes can |
| 345 | * represent is (the number of possible start bytes for 'n') |
| 346 | * * (the number of possiblities for each start byte |
| 347 | * The latter in turn is |
| 348 | * 2 ** ( (how many continuation bytes there are) |
| 349 | * * (the number of bits of information each |
| 350 | * continuation byte holds)) |
| 351 | * |
| 352 | * If we were on a platform where we could use a fast find first set bit |
| 353 | * instruction (or count leading zeros instruction) this could be replaced by |
| 354 | * using that to find the log2 of the uv, and divide that by the number of bits |
| 355 | * of information in each continuation byte, adjusting for large cases and how |
| 356 | * much information is in a start byte for that length */ |
| 357 | #define __COMMON_UNI_SKIP(uv) \ |
| 358 | (UV) (uv) < (32 * (1U << ( UTF_ACCUMULATION_SHIFT))) ? 2 : \ |
| 359 | (UV) (uv) < (16 * (1U << (2 * UTF_ACCUMULATION_SHIFT))) ? 3 : \ |
| 360 | (UV) (uv) < ( 8 * (1U << (3 * UTF_ACCUMULATION_SHIFT))) ? 4 : \ |
| 361 | (UV) (uv) < ( 4 * (1U << (4 * UTF_ACCUMULATION_SHIFT))) ? 5 : \ |
| 362 | (UV) (uv) < ( 2 * (1U << (5 * UTF_ACCUMULATION_SHIFT))) ? 6 : |
| 363 | |
| 364 | /* Internal macro to be used only in this file. |
| 365 | * This adds to __COMMON_UNI_SKIP the details at this platform's upper range. |
| 366 | * For any-sized EBCDIC platforms, or 64-bit ASCII ones, we need one more test |
| 367 | * to see if just 7 bytes is needed, or if the maximum is needed. For 32-bit |
| 368 | * ASCII platforms, everything is representable by 7 bytes */ |
| 369 | #if defined(UV_IS_QUAD) || defined(EBCDIC) |
| 370 | # define __BASE_UNI_SKIP(uv) (__COMMON_UNI_SKIP(uv) \ |
| 371 | (UV) (uv) < ((UV) 1U << (6 * UTF_ACCUMULATION_SHIFT)) ? 7 : UTF8_MAXBYTES) |
| 372 | #else |
| 373 | # define __BASE_UNI_SKIP(uv) (__COMMON_UNI_SKIP(uv) 7) |
| 374 | #endif |
| 375 | |
| 376 | /* The next two macros use the base macro defined above, and add in the tests |
| 377 | * at the low-end of the range, for just 1 byte, yielding complete macros, |
| 378 | * publicly accessible. */ |
| 379 | |
| 380 | /* Input is a true Unicode (not-native) code point */ |
| 381 | #define OFFUNISKIP(uv) (OFFUNI_IS_INVARIANT(uv) ? 1 : __BASE_UNI_SKIP(uv)) |
| 382 | |
| 383 | /* |
| 384 | |
| 385 | =for apidoc Am|STRLEN|UVCHR_SKIP|UV cp |
| 386 | returns the number of bytes required to represent the code point C<cp> when |
| 387 | encoded as UTF-8. C<cp> is a native (ASCII or EBCDIC) code point if less than |
| 388 | 255; a Unicode code point otherwise. |
| 389 | |
| 390 | =cut |
| 391 | */ |
| 392 | #define UVCHR_SKIP(uv) ( UVCHR_IS_INVARIANT(uv) ? 1 : __BASE_UNI_SKIP(uv)) |
| 393 | |
| 394 | #define UTF_MIN_START_BYTE \ |
| 395 | ((UTF_CONTINUATION_MARK >> UTF_ACCUMULATION_SHIFT) | UTF_START_MARK(2)) |
| 396 | |
| 397 | /* Is the byte 'c' the first byte of a multi-byte UTF8-8 encoded sequence? |
| 398 | * This doesn't catch invariants (they are single-byte). It also excludes the |
| 399 | * illegal overlong sequences that begin with C0 and C1 on ASCII platforms, and |
| 400 | * C0-C4 I8 start bytes on EBCDIC ones */ |
| 401 | #define UTF8_IS_START(c) (__ASSERT_(FITS_IN_8_BITS(c)) \ |
| 402 | (NATIVE_UTF8_TO_I8(c) >= UTF_MIN_START_BYTE)) |
| 403 | |
| 404 | #define UTF_MIN_ABOVE_LATIN1_BYTE \ |
| 405 | ((0x100 >> UTF_ACCUMULATION_SHIFT) | UTF_START_MARK(2)) |
| 406 | |
| 407 | /* Is the UTF8-encoded byte 'c' the first byte of a sequence of bytes that |
| 408 | * represent a code point > 255? */ |
| 409 | #define UTF8_IS_ABOVE_LATIN1(c) (__ASSERT_(FITS_IN_8_BITS(c)) \ |
| 410 | (NATIVE_UTF8_TO_I8(c) >= UTF_MIN_ABOVE_LATIN1_BYTE)) |
| 411 | |
| 412 | /* Is the UTF8-encoded byte 'c' the first byte of a two byte sequence? Use |
| 413 | * UTF8_IS_NEXT_CHAR_DOWNGRADEABLE() instead if the input isn't known to |
| 414 | * be well-formed. */ |
| 415 | #define UTF8_IS_DOWNGRADEABLE_START(c) (__ASSERT_(FITS_IN_8_BITS(c)) \ |
| 416 | inRANGE(NATIVE_UTF8_TO_I8(c), \ |
| 417 | UTF_MIN_START_BYTE, UTF_MIN_ABOVE_LATIN1_BYTE - 1)) |
| 418 | |
| 419 | /* The largest code point representable by two UTF-8 bytes on this platform. |
| 420 | * As explained in the comments for __COMMON_UNI_SKIP, 32 start bytes with |
| 421 | * UTF_ACCUMULATION_SHIFT bits of information each */ |
| 422 | #define MAX_UTF8_TWO_BYTE (32 * (1U << UTF_ACCUMULATION_SHIFT) - 1) |
| 423 | |
| 424 | /* The largest code point representable by two UTF-8 bytes on any platform that |
| 425 | * Perl runs on. This value is constrained by EBCDIC which has 5 bits per |
| 426 | * continuation byte */ |
| 427 | #define MAX_PORTABLE_UTF8_TWO_BYTE (32 * (1U << 5) - 1) |
| 428 | |
| 429 | /* The maximum number of UTF-8 bytes a single Unicode character can |
| 430 | * uppercase/lowercase/fold into. Unicode guarantees that the maximum |
| 431 | * expansion is UTF8_MAX_FOLD_CHAR_EXPAND characters, but any above-Unicode |
| 432 | * code point will fold to itself, so we only have to look at the expansion of |
| 433 | * the maximum Unicode code point. But this number may be less than the space |
| 434 | * occupied by a very large code point under Perl's extended UTF-8. We have to |
| 435 | * make it large enough to fit any single character. (It turns out that ASCII |
| 436 | * and EBCDIC differ in which is larger) */ |
| 437 | #define UTF8_MAXBYTES_CASE \ |
| 438 | (UTF8_MAXBYTES >= (UTF8_MAX_FOLD_CHAR_EXPAND * OFFUNISKIP(0x10FFFF)) \ |
| 439 | ? UTF8_MAXBYTES \ |
| 440 | : (UTF8_MAX_FOLD_CHAR_EXPAND * OFFUNISKIP(0x10FFFF))) |
| 441 | |
| 442 | /* Rest of these are attributes of Unicode and perl's internals rather than the |
| 443 | * encoding, or happen to be the same in both ASCII and EBCDIC (at least at |
| 444 | * this level; the macros that some of these call may have different |
| 445 | * definitions in the two encodings */ |
| 446 | |
| 447 | /* In domain restricted to ASCII, these may make more sense to the reader than |
| 448 | * the ones with Latin1 in the name */ |
| 449 | #define NATIVE_TO_ASCII(ch) NATIVE_TO_LATIN1(ch) |
| 450 | #define ASCII_TO_NATIVE(ch) LATIN1_TO_NATIVE(ch) |
| 451 | |
| 452 | /* More or less misleadingly-named defines, retained for back compat */ |
| 453 | #define NATIVE_TO_UTF(ch) NATIVE_UTF8_TO_I8(ch) |
| 454 | #define NATIVE_TO_I8(ch) NATIVE_UTF8_TO_I8(ch) |
| 455 | #define UTF_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch) |
| 456 | #define I8_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch) |
| 457 | #define NATIVE8_TO_UNI(ch) NATIVE_TO_LATIN1(ch) |
| 458 | |
| 459 | /* This defines the 1-bits that are to be in the first byte of a multi-byte |
| 460 | * UTF-8 encoded character that mark it as a start byte and give the number of |
| 461 | * bytes that comprise the character. 'len' is the number of bytes in the |
| 462 | * multi-byte sequence. */ |
| 463 | #define UTF_START_MARK(len) (((len) > 7) ? 0xFF : (0xFF & (0xFE << (7-(len))))) |
| 464 | |
| 465 | /* Masks out the initial one bits in a start byte, leaving the real data ones. |
| 466 | * Doesn't work on an invariant byte. 'len' is the number of bytes in the |
| 467 | * multi-byte sequence that comprises the character. */ |
| 468 | #define UTF_START_MASK(len) (((len) >= 7) ? 0x00 : (0x1F >> ((len)-2))) |
| 469 | |
| 470 | /* Adds a UTF8 continuation byte 'new' of information to a running total code |
| 471 | * point 'old' of all the continuation bytes so far. This is designed to be |
| 472 | * used in a loop to convert from UTF-8 to the code point represented. Note |
| 473 | * that this is asymmetric on EBCDIC platforms, in that the 'new' parameter is |
| 474 | * the UTF-EBCDIC byte, whereas the 'old' parameter is a Unicode (not EBCDIC) |
| 475 | * code point in process of being generated */ |
| 476 | #define UTF8_ACCUMULATE(old, new) (__ASSERT_(FITS_IN_8_BITS(new)) \ |
| 477 | ((old) << UTF_ACCUMULATION_SHIFT) \ |
| 478 | | ((NATIVE_UTF8_TO_I8(new)) \ |
| 479 | & UTF_CONTINUATION_MASK)) |
| 480 | |
| 481 | /* This works in the face of malformed UTF-8. */ |
| 482 | #define UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(s, e) \ |
| 483 | ( UTF8_IS_DOWNGRADEABLE_START(*(s)) \ |
| 484 | && ( (e) - (s) > 1) \ |
| 485 | && UTF8_IS_CONTINUATION(*((s)+1))) |
| 486 | |
| 487 | /* Number of bytes a code point occupies in UTF-8. */ |
| 488 | #define NATIVE_SKIP(uv) UVCHR_SKIP(uv) |
| 489 | |
| 490 | /* Most code which says UNISKIP is really thinking in terms of native code |
| 491 | * points (0-255) plus all those beyond. This is an imprecise term, but having |
| 492 | * it means existing code continues to work. For precision, use UVCHR_SKIP, |
| 493 | * NATIVE_SKIP, or OFFUNISKIP */ |
| 494 | #define UNISKIP(uv) UVCHR_SKIP(uv) |
| 495 | |
| 496 | /* Longer, but more accurate name */ |
| 497 | #define UTF8_IS_ABOVE_LATIN1_START(c) UTF8_IS_ABOVE_LATIN1(c) |
| 498 | |
| 499 | /* Convert a UTF-8 variant Latin1 character to a native code point value. |
| 500 | * Needs just one iteration of accumulate. Should be used only if it is known |
| 501 | * that the code point is < 256, and is not UTF-8 invariant. Use the slower |
| 502 | * but more general TWO_BYTE_UTF8_TO_NATIVE() which handles any code point |
| 503 | * representable by two bytes (which turns out to be up through |
| 504 | * MAX_PORTABLE_UTF8_TWO_BYTE). The two parameters are: |
| 505 | * HI: a downgradable start byte; |
| 506 | * LO: continuation. |
| 507 | * */ |
| 508 | #define EIGHT_BIT_UTF8_TO_NATIVE(HI, LO) \ |
| 509 | ( __ASSERT_(UTF8_IS_DOWNGRADEABLE_START(HI)) \ |
| 510 | __ASSERT_(UTF8_IS_CONTINUATION(LO)) \ |
| 511 | LATIN1_TO_NATIVE(UTF8_ACCUMULATE(( \ |
| 512 | NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), (LO)))) |
| 513 | |
| 514 | /* Convert a two (not one) byte utf8 character to a native code point value. |
| 515 | * Needs just one iteration of accumulate. Should not be used unless it is |
| 516 | * known that the two bytes are legal: 1) two-byte start, and 2) continuation. |
| 517 | * Note that the result can be larger than 255 if the input character is not |
| 518 | * downgradable */ |
| 519 | #define TWO_BYTE_UTF8_TO_NATIVE(HI, LO) \ |
| 520 | (__ASSERT_(FITS_IN_8_BITS(HI)) \ |
| 521 | __ASSERT_(FITS_IN_8_BITS(LO)) \ |
| 522 | __ASSERT_(PL_utf8skip[HI] == 2) \ |
| 523 | __ASSERT_(UTF8_IS_CONTINUATION(LO)) \ |
| 524 | UNI_TO_NATIVE(UTF8_ACCUMULATE((NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), \ |
| 525 | (LO)))) |
| 526 | |
| 527 | /* Should never be used, and be deprecated */ |
| 528 | #define TWO_BYTE_UTF8_TO_UNI(HI, LO) NATIVE_TO_UNI(TWO_BYTE_UTF8_TO_NATIVE(HI, LO)) |
| 529 | |
| 530 | /* |
| 531 | |
| 532 | =for apidoc Am|STRLEN|UTF8SKIP|char* s |
| 533 | returns the number of bytes in the UTF-8 encoded character whose first (perhaps |
| 534 | only) byte is pointed to by C<s>. |
| 535 | |
| 536 | =cut |
| 537 | */ |
| 538 | #define UTF8SKIP(s) PL_utf8skip[*(const U8*)(s)] |
| 539 | |
| 540 | /* |
| 541 | =for apidoc Am|STRLEN|UTF8_SKIP|char* s |
| 542 | This is a synonym for L</C<UTF8SKIP>> |
| 543 | |
| 544 | =cut |
| 545 | */ |
| 546 | |
| 547 | #define UTF8_SKIP(s) UTF8SKIP(s) |
| 548 | |
| 549 | /* |
| 550 | |
| 551 | =for apidoc Am|STRLEN|UTF8_SAFE_SKIP|char* s|char* e |
| 552 | returns 0 if S<C<s E<gt>= e>>; otherwise returns the number of bytes in the |
| 553 | UTF-8 encoded character whose first byte is pointed to by C<s>. But it never |
| 554 | returns beyond C<e>. On DEBUGGING builds, it asserts that S<C<s E<lt>= e>>. |
| 555 | |
| 556 | =cut |
| 557 | */ |
| 558 | #define UTF8_SAFE_SKIP(s, e) (__ASSERT_((e) >= (s)) \ |
| 559 | ((e) - (s)) <= 0 \ |
| 560 | ? 0 \ |
| 561 | : MIN(((e) - (s)), UTF8_SKIP(s))) |
| 562 | |
| 563 | /* Most code that says 'UNI_' really means the native value for code points up |
| 564 | * through 255 */ |
| 565 | #define UNI_IS_INVARIANT(cp) UVCHR_IS_INVARIANT(cp) |
| 566 | |
| 567 | /* |
| 568 | =for apidoc Am|bool|UTF8_IS_INVARIANT|char c |
| 569 | |
| 570 | Evaluates to 1 if the byte C<c> represents the same character when encoded in |
| 571 | UTF-8 as when not; otherwise evaluates to 0. UTF-8 invariant characters can be |
| 572 | copied as-is when converting to/from UTF-8, saving time. |
| 573 | |
| 574 | In spite of the name, this macro gives the correct result if the input string |
| 575 | from which C<c> comes is not encoded in UTF-8. |
| 576 | |
| 577 | See C<L</UVCHR_IS_INVARIANT>> for checking if a UV is invariant. |
| 578 | |
| 579 | =cut |
| 580 | |
| 581 | The reason it works on both UTF-8 encoded strings and non-UTF-8 encoded, is |
| 582 | that it returns TRUE in each for the exact same set of bit patterns. It is |
| 583 | valid on a subset of what UVCHR_IS_INVARIANT is valid on, so can just use that; |
| 584 | and the compiler should optimize out anything extraneous given the |
| 585 | implementation of the latter. The |0 makes sure this isn't mistakenly called |
| 586 | with a ptr argument. |
| 587 | */ |
| 588 | #define UTF8_IS_INVARIANT(c) UVCHR_IS_INVARIANT((c) | 0) |
| 589 | |
| 590 | /* Like the above, but its name implies a non-UTF8 input, which as the comments |
| 591 | * above show, doesn't matter as to its implementation */ |
| 592 | #define NATIVE_BYTE_IS_INVARIANT(c) UVCHR_IS_INVARIANT(c) |
| 593 | |
| 594 | /* Misleadingly named: is the UTF8-encoded byte 'c' part of a variant sequence |
| 595 | * in UTF-8? This is the inverse of UTF8_IS_INVARIANT. */ |
| 596 | #define UTF8_IS_CONTINUED(c) (__ASSERT_(FITS_IN_8_BITS(c)) \ |
| 597 | (! UTF8_IS_INVARIANT(c))) |
| 598 | |
| 599 | /* The macros in the next 4 sets are used to generate the two utf8 or utfebcdic |
| 600 | * bytes from an ordinal that is known to fit into exactly two (not one) bytes; |
| 601 | * it must be less than 0x3FF to work across both encodings. */ |
| 602 | |
| 603 | /* These two are helper macros for the other three sets, and should not be used |
| 604 | * directly anywhere else. 'translate_function' is either NATIVE_TO_LATIN1 |
| 605 | * (which works for code points up through 0xFF) or NATIVE_TO_UNI which works |
| 606 | * for any code point */ |
| 607 | #define __BASE_TWO_BYTE_HI(c, translate_function) \ |
| 608 | (__ASSERT_(! UVCHR_IS_INVARIANT(c)) \ |
| 609 | I8_TO_NATIVE_UTF8((translate_function(c) >> UTF_ACCUMULATION_SHIFT) \ |
| 610 | | UTF_START_MARK(2))) |
| 611 | #define __BASE_TWO_BYTE_LO(c, translate_function) \ |
| 612 | (__ASSERT_(! UVCHR_IS_INVARIANT(c)) \ |
| 613 | I8_TO_NATIVE_UTF8((translate_function(c) & UTF_CONTINUATION_MASK) \ |
| 614 | | UTF_CONTINUATION_MARK)) |
| 615 | |
| 616 | /* The next two macros should not be used. They were designed to be usable as |
| 617 | * the case label of a switch statement, but this doesn't work for EBCDIC. Use |
| 618 | * regen/unicode_constants.pl instead */ |
| 619 | #define UTF8_TWO_BYTE_HI_nocast(c) __BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI) |
| 620 | #define UTF8_TWO_BYTE_LO_nocast(c) __BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI) |
| 621 | |
| 622 | /* The next two macros are used when the source should be a single byte |
| 623 | * character; checked for under DEBUGGING */ |
| 624 | #define UTF8_EIGHT_BIT_HI(c) (__ASSERT_(FITS_IN_8_BITS(c)) \ |
| 625 | ( __BASE_TWO_BYTE_HI(c, NATIVE_TO_LATIN1))) |
| 626 | #define UTF8_EIGHT_BIT_LO(c) (__ASSERT_(FITS_IN_8_BITS(c)) \ |
| 627 | (__BASE_TWO_BYTE_LO(c, NATIVE_TO_LATIN1))) |
| 628 | |
| 629 | /* These final two macros in the series are used when the source can be any |
| 630 | * code point whose UTF-8 is known to occupy 2 bytes; they are less efficient |
| 631 | * than the EIGHT_BIT versions on EBCDIC platforms. We use the logical '~' |
| 632 | * operator instead of "<=" to avoid getting compiler warnings. |
| 633 | * MAX_UTF8_TWO_BYTE should be exactly all one bits in the lower few |
| 634 | * places, so the ~ works */ |
| 635 | #define UTF8_TWO_BYTE_HI(c) \ |
| 636 | (__ASSERT_((sizeof(c) == 1) \ |
| 637 | || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \ |
| 638 | (__BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI))) |
| 639 | #define UTF8_TWO_BYTE_LO(c) \ |
| 640 | (__ASSERT_((sizeof(c) == 1) \ |
| 641 | || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \ |
| 642 | (__BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI))) |
| 643 | |
| 644 | /* This is illegal in any well-formed UTF-8 in both EBCDIC and ASCII |
| 645 | * as it is only in overlongs. */ |
| 646 | #define ILLEGAL_UTF8_BYTE I8_TO_NATIVE_UTF8(0xC1) |
| 647 | |
| 648 | /* |
| 649 | * 'UTF' is whether or not p is encoded in UTF8. The names 'foo_lazy_if' stem |
| 650 | * from an earlier version of these macros in which they didn't call the |
| 651 | * foo_utf8() macros (i.e. were 'lazy') unless they decided that *p is the |
| 652 | * beginning of a utf8 character. Now that foo_utf8() determines that itself, |
| 653 | * no need to do it again here |
| 654 | */ |
| 655 | #define isIDFIRST_lazy_if(p,UTF) \ |
| 656 | _is_utf8_FOO(_CC_IDFIRST, (const U8 *) p, "isIDFIRST_lazy_if", \ |
| 657 | "isIDFIRST_lazy_if_safe", \ |
| 658 | cBOOL(UTF && ! IN_BYTES), 0, __FILE__,__LINE__) |
| 659 | |
| 660 | #define isIDFIRST_lazy_if_safe(p, e, UTF) \ |
| 661 | ((IN_BYTES || !UTF) \ |
| 662 | ? isIDFIRST(*(p)) \ |
| 663 | : isIDFIRST_utf8_safe(p, e)) |
| 664 | |
| 665 | #define isWORDCHAR_lazy_if(p,UTF) \ |
| 666 | _is_utf8_FOO(_CC_IDFIRST, (const U8 *) p, "isWORDCHAR_lazy_if", \ |
| 667 | "isWORDCHAR_lazy_if_safe", \ |
| 668 | cBOOL(UTF && ! IN_BYTES), 0, __FILE__,__LINE__) |
| 669 | |
| 670 | #define isWORDCHAR_lazy_if_safe(p, e, UTF) \ |
| 671 | ((IN_BYTES || !UTF) \ |
| 672 | ? isWORDCHAR(*(p)) \ |
| 673 | : isWORDCHAR_utf8_safe((U8 *) p, (U8 *) e)) |
| 674 | |
| 675 | #define isALNUM_lazy_if(p,UTF) \ |
| 676 | _is_utf8_FOO(_CC_IDFIRST, (const U8 *) p, "isALNUM_lazy_if", \ |
| 677 | "isWORDCHAR_lazy_if_safe", \ |
| 678 | cBOOL(UTF && ! IN_BYTES), 0, __FILE__,__LINE__) |
| 679 | |
| 680 | #define UTF8_MAXLEN UTF8_MAXBYTES |
| 681 | |
| 682 | /* A Unicode character can fold to up to 3 characters */ |
| 683 | #define UTF8_MAX_FOLD_CHAR_EXPAND 3 |
| 684 | |
| 685 | #define IN_BYTES UNLIKELY(CopHINTS_get(PL_curcop) & HINT_BYTES) |
| 686 | |
| 687 | /* |
| 688 | |
| 689 | =for apidoc Am|bool|DO_UTF8|SV* sv |
| 690 | Returns a bool giving whether or not the PV in C<sv> is to be treated as being |
| 691 | encoded in UTF-8. |
| 692 | |
| 693 | You should use this I<after> a call to C<SvPV()> or one of its variants, in |
| 694 | case any call to string overloading updates the internal UTF-8 encoding flag. |
| 695 | |
| 696 | =cut |
| 697 | */ |
| 698 | #define DO_UTF8(sv) (SvUTF8(sv) && !IN_BYTES) |
| 699 | |
| 700 | /* Should all strings be treated as Unicode, and not just UTF-8 encoded ones? |
| 701 | * Is so within 'feature unicode_strings' or 'locale :not_characters', and not |
| 702 | * within 'use bytes'. UTF-8 locales are not tested for here, but perhaps |
| 703 | * could be */ |
| 704 | #define IN_UNI_8_BIT \ |
| 705 | (( ( (CopHINTS_get(PL_curcop) & HINT_UNI_8_BIT)) \ |
| 706 | || ( CopHINTS_get(PL_curcop) & HINT_LOCALE_PARTIAL \ |
| 707 | /* -1 below is for :not_characters */ \ |
| 708 | && _is_in_locale_category(FALSE, -1))) \ |
| 709 | && (! IN_BYTES)) |
| 710 | |
| 711 | |
| 712 | #define UTF8_ALLOW_EMPTY 0x0001 /* Allow a zero length string */ |
| 713 | #define UTF8_GOT_EMPTY UTF8_ALLOW_EMPTY |
| 714 | |
| 715 | /* Allow first byte to be a continuation byte */ |
| 716 | #define UTF8_ALLOW_CONTINUATION 0x0002 |
| 717 | #define UTF8_GOT_CONTINUATION UTF8_ALLOW_CONTINUATION |
| 718 | |
| 719 | /* Unexpected non-continuation byte */ |
| 720 | #define UTF8_ALLOW_NON_CONTINUATION 0x0004 |
| 721 | #define UTF8_GOT_NON_CONTINUATION UTF8_ALLOW_NON_CONTINUATION |
| 722 | |
| 723 | /* expecting more bytes than were available in the string */ |
| 724 | #define UTF8_ALLOW_SHORT 0x0008 |
| 725 | #define UTF8_GOT_SHORT UTF8_ALLOW_SHORT |
| 726 | |
| 727 | /* Overlong sequence; i.e., the code point can be specified in fewer bytes. |
| 728 | * First one will convert the overlong to the REPLACEMENT CHARACTER; second |
| 729 | * will return what the overlong evaluates to */ |
| 730 | #define UTF8_ALLOW_LONG 0x0010 |
| 731 | #define UTF8_ALLOW_LONG_AND_ITS_VALUE (UTF8_ALLOW_LONG|0x0020) |
| 732 | #define UTF8_GOT_LONG UTF8_ALLOW_LONG |
| 733 | |
| 734 | #define UTF8_ALLOW_OVERFLOW 0x0080 |
| 735 | #define UTF8_GOT_OVERFLOW UTF8_ALLOW_OVERFLOW |
| 736 | |
| 737 | #define UTF8_DISALLOW_SURROGATE 0x0100 /* Unicode surrogates */ |
| 738 | #define UTF8_GOT_SURROGATE UTF8_DISALLOW_SURROGATE |
| 739 | #define UTF8_WARN_SURROGATE 0x0200 |
| 740 | |
| 741 | /* Unicode non-character code points */ |
| 742 | #define UTF8_DISALLOW_NONCHAR 0x0400 |
| 743 | #define UTF8_GOT_NONCHAR UTF8_DISALLOW_NONCHAR |
| 744 | #define UTF8_WARN_NONCHAR 0x0800 |
| 745 | |
| 746 | /* Super-set of Unicode: code points above the legal max */ |
| 747 | #define UTF8_DISALLOW_SUPER 0x1000 |
| 748 | #define UTF8_GOT_SUPER UTF8_DISALLOW_SUPER |
| 749 | #define UTF8_WARN_SUPER 0x2000 |
| 750 | |
| 751 | /* The original UTF-8 standard did not define UTF-8 with start bytes of 0xFE or |
| 752 | * 0xFF, though UTF-EBCDIC did. This allowed both versions to represent code |
| 753 | * points up to 2 ** 31 - 1. Perl extends UTF-8 so that 0xFE and 0xFF are |
| 754 | * usable on ASCII platforms, and 0xFF means something different than |
| 755 | * UTF-EBCDIC defines. These changes allow code points of 64 bits (actually |
| 756 | * somewhat more) to be represented on both platforms. But these are Perl |
| 757 | * extensions, and not likely to be interchangeable with other languages. Note |
| 758 | * that on ASCII platforms, FE overflows a signed 32-bit word, and FF an |
| 759 | * unsigned one. */ |
| 760 | #define UTF8_DISALLOW_PERL_EXTENDED 0x4000 |
| 761 | #define UTF8_GOT_PERL_EXTENDED UTF8_DISALLOW_PERL_EXTENDED |
| 762 | #define UTF8_WARN_PERL_EXTENDED 0x8000 |
| 763 | |
| 764 | /* For back compat, these old names are misleading for overlongs and |
| 765 | * UTF_EBCDIC. */ |
| 766 | #define UTF8_DISALLOW_ABOVE_31_BIT UTF8_DISALLOW_PERL_EXTENDED |
| 767 | #define UTF8_GOT_ABOVE_31_BIT UTF8_GOT_PERL_EXTENDED |
| 768 | #define UTF8_WARN_ABOVE_31_BIT UTF8_WARN_PERL_EXTENDED |
| 769 | #define UTF8_DISALLOW_FE_FF UTF8_DISALLOW_PERL_EXTENDED |
| 770 | #define UTF8_WARN_FE_FF UTF8_WARN_PERL_EXTENDED |
| 771 | |
| 772 | #define UTF8_CHECK_ONLY 0x10000 |
| 773 | #define _UTF8_NO_CONFIDENCE_IN_CURLEN 0x20000 /* Internal core use only */ |
| 774 | |
| 775 | /* For backwards source compatibility. They do nothing, as the default now |
| 776 | * includes what they used to mean. The first one's meaning was to allow the |
| 777 | * just the single non-character 0xFFFF */ |
| 778 | #define UTF8_ALLOW_FFFF 0 |
| 779 | #define UTF8_ALLOW_FE_FF 0 |
| 780 | #define UTF8_ALLOW_SURROGATE 0 |
| 781 | |
| 782 | /* C9 refers to Unicode Corrigendum #9: allows but discourages non-chars */ |
| 783 | #define UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE \ |
| 784 | (UTF8_DISALLOW_SUPER|UTF8_DISALLOW_SURROGATE) |
| 785 | #define UTF8_WARN_ILLEGAL_C9_INTERCHANGE (UTF8_WARN_SUPER|UTF8_WARN_SURROGATE) |
| 786 | |
| 787 | #define UTF8_DISALLOW_ILLEGAL_INTERCHANGE \ |
| 788 | (UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE|UTF8_DISALLOW_NONCHAR) |
| 789 | #define UTF8_WARN_ILLEGAL_INTERCHANGE \ |
| 790 | (UTF8_WARN_ILLEGAL_C9_INTERCHANGE|UTF8_WARN_NONCHAR) |
| 791 | |
| 792 | /* This is typically used for code that processes UTF-8 input and doesn't want |
| 793 | * to have to deal with any malformations that might be present. All such will |
| 794 | * be safely replaced by the REPLACEMENT CHARACTER, unless other flags |
| 795 | * overriding this are also present. */ |
| 796 | #define UTF8_ALLOW_ANY ( UTF8_ALLOW_CONTINUATION \ |
| 797 | |UTF8_ALLOW_NON_CONTINUATION \ |
| 798 | |UTF8_ALLOW_SHORT \ |
| 799 | |UTF8_ALLOW_LONG \ |
| 800 | |UTF8_ALLOW_OVERFLOW) |
| 801 | |
| 802 | /* Accept any Perl-extended UTF-8 that evaluates to any UV on the platform, but |
| 803 | * not any malformed. This is the default. */ |
| 804 | #define UTF8_ALLOW_ANYUV 0 |
| 805 | #define UTF8_ALLOW_DEFAULT UTF8_ALLOW_ANYUV |
| 806 | |
| 807 | /* |
| 808 | =for apidoc Am|bool|UTF8_IS_SURROGATE|const U8 *s|const U8 *e |
| 809 | |
| 810 | Evaluates to non-zero if the first few bytes of the string starting at C<s> and |
| 811 | looking no further than S<C<e - 1>> are well-formed UTF-8 that represents one |
| 812 | of the Unicode surrogate code points; otherwise it evaluates to 0. If |
| 813 | non-zero, the value gives how many bytes starting at C<s> comprise the code |
| 814 | point's representation. |
| 815 | |
| 816 | =cut |
| 817 | */ |
| 818 | #define UTF8_IS_SURROGATE(s, e) is_SURROGATE_utf8_safe(s, e) |
| 819 | |
| 820 | |
| 821 | #define UTF8_IS_REPLACEMENT(s, send) is_REPLACEMENT_utf8_safe(s,send) |
| 822 | |
| 823 | #define MAX_LEGAL_CP ((UV)IV_MAX) |
| 824 | |
| 825 | /* |
| 826 | =for apidoc Am|bool|UTF8_IS_SUPER|const U8 *s|const U8 *e |
| 827 | |
| 828 | Recall that Perl recognizes an extension to UTF-8 that can encode code |
| 829 | points larger than the ones defined by Unicode, which are 0..0x10FFFF. |
| 830 | |
| 831 | This macro evaluates to non-zero if the first few bytes of the string starting |
| 832 | at C<s> and looking no further than S<C<e - 1>> are from this UTF-8 extension; |
| 833 | otherwise it evaluates to 0. If non-zero, the value gives how many bytes |
| 834 | starting at C<s> comprise the code point's representation. |
| 835 | |
| 836 | 0 is returned if the bytes are not well-formed extended UTF-8, or if they |
| 837 | represent a code point that cannot fit in a UV on the current platform. Hence |
| 838 | this macro can give different results when run on a 64-bit word machine than on |
| 839 | one with a 32-bit word size. |
| 840 | |
| 841 | Note that it is illegal to have code points that are larger than what can |
| 842 | fit in an IV on the current machine. |
| 843 | |
| 844 | =cut |
| 845 | |
| 846 | * ASCII EBCDIC I8 |
| 847 | * U+10FFFF: \xF4\x8F\xBF\xBF \xF9\xA1\xBF\xBF\xBF max legal Unicode |
| 848 | * U+110000: \xF4\x90\x80\x80 \xF9\xA2\xA0\xA0\xA0 |
| 849 | * U+110001: \xF4\x90\x80\x81 \xF9\xA2\xA0\xA0\xA1 |
| 850 | */ |
| 851 | #ifdef EBCDIC |
| 852 | # define UTF8_IS_SUPER(s, e) \ |
| 853 | (( LIKELY((e) > (s) + 4) \ |
| 854 | && NATIVE_UTF8_TO_I8(*(s)) >= 0xF9 \ |
| 855 | && ( NATIVE_UTF8_TO_I8(*(s)) > 0xF9 \ |
| 856 | || (NATIVE_UTF8_TO_I8(*((s) + 1)) >= 0xA2)) \ |
| 857 | && LIKELY((s) + UTF8SKIP(s) <= (e))) \ |
| 858 | ? is_utf8_char_helper(s, s + UTF8SKIP(s), 0) : 0) |
| 859 | #else |
| 860 | # define UTF8_IS_SUPER(s, e) \ |
| 861 | (( LIKELY((e) > (s) + 3) \ |
| 862 | && (*(U8*) (s)) >= 0xF4 \ |
| 863 | && ((*(U8*) (s)) > 0xF4 || (*((U8*) (s) + 1) >= 0x90))\ |
| 864 | && LIKELY((s) + UTF8SKIP(s) <= (e))) \ |
| 865 | ? is_utf8_char_helper(s, s + UTF8SKIP(s), 0) : 0) |
| 866 | #endif |
| 867 | |
| 868 | /* These are now machine generated, and the 'given' clause is no longer |
| 869 | * applicable */ |
| 870 | #define UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s, e) \ |
| 871 | cBOOL(is_NONCHAR_utf8_safe(s,e)) |
| 872 | |
| 873 | /* |
| 874 | =for apidoc Am|bool|UTF8_IS_NONCHAR|const U8 *s|const U8 *e |
| 875 | |
| 876 | Evaluates to non-zero if the first few bytes of the string starting at C<s> and |
| 877 | looking no further than S<C<e - 1>> are well-formed UTF-8 that represents one |
| 878 | of the Unicode non-character code points; otherwise it evaluates to 0. If |
| 879 | non-zero, the value gives how many bytes starting at C<s> comprise the code |
| 880 | point's representation. |
| 881 | |
| 882 | =for apidoc AmnU|UV|UNICODE_REPLACEMENT |
| 883 | |
| 884 | Evaluates to 0xFFFD, the code point of the Unicode REPLACEMENT CHARACTER |
| 885 | |
| 886 | =cut |
| 887 | */ |
| 888 | #define UTF8_IS_NONCHAR(s, e) \ |
| 889 | UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s, e) |
| 890 | |
| 891 | #define UNICODE_SURROGATE_FIRST 0xD800 |
| 892 | #define UNICODE_SURROGATE_LAST 0xDFFF |
| 893 | #define UNICODE_REPLACEMENT 0xFFFD |
| 894 | #define UNICODE_BYTE_ORDER_MARK 0xFEFF |
| 895 | |
| 896 | /* Though our UTF-8 encoding can go beyond this, |
| 897 | * let's be conservative and do as Unicode says. */ |
| 898 | #define PERL_UNICODE_MAX 0x10FFFF |
| 899 | |
| 900 | #define UNICODE_WARN_SURROGATE 0x0001 /* UTF-16 surrogates */ |
| 901 | #define UNICODE_WARN_NONCHAR 0x0002 /* Non-char code points */ |
| 902 | #define UNICODE_WARN_SUPER 0x0004 /* Above 0x10FFFF */ |
| 903 | #define UNICODE_WARN_PERL_EXTENDED 0x0008 /* Above 0x7FFF_FFFF */ |
| 904 | #define UNICODE_WARN_ABOVE_31_BIT UNICODE_WARN_PERL_EXTENDED |
| 905 | #define UNICODE_DISALLOW_SURROGATE 0x0010 |
| 906 | #define UNICODE_DISALLOW_NONCHAR 0x0020 |
| 907 | #define UNICODE_DISALLOW_SUPER 0x0040 |
| 908 | #define UNICODE_DISALLOW_PERL_EXTENDED 0x0080 |
| 909 | #define UNICODE_DISALLOW_ABOVE_31_BIT UNICODE_DISALLOW_PERL_EXTENDED |
| 910 | |
| 911 | #define UNICODE_GOT_SURROGATE UNICODE_DISALLOW_SURROGATE |
| 912 | #define UNICODE_GOT_NONCHAR UNICODE_DISALLOW_NONCHAR |
| 913 | #define UNICODE_GOT_SUPER UNICODE_DISALLOW_SUPER |
| 914 | #define UNICODE_GOT_PERL_EXTENDED UNICODE_DISALLOW_PERL_EXTENDED |
| 915 | |
| 916 | #define UNICODE_WARN_ILLEGAL_C9_INTERCHANGE \ |
| 917 | (UNICODE_WARN_SURROGATE|UNICODE_WARN_SUPER) |
| 918 | #define UNICODE_WARN_ILLEGAL_INTERCHANGE \ |
| 919 | (UNICODE_WARN_ILLEGAL_C9_INTERCHANGE|UNICODE_WARN_NONCHAR) |
| 920 | #define UNICODE_DISALLOW_ILLEGAL_C9_INTERCHANGE \ |
| 921 | (UNICODE_DISALLOW_SURROGATE|UNICODE_DISALLOW_SUPER) |
| 922 | #define UNICODE_DISALLOW_ILLEGAL_INTERCHANGE \ |
| 923 | (UNICODE_DISALLOW_ILLEGAL_C9_INTERCHANGE|UNICODE_DISALLOW_NONCHAR) |
| 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 |
| 929 | |
| 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 | |
| 935 | #define UNICODE_IS_REPLACEMENT(uv) ((UV) (uv) == UNICODE_REPLACEMENT) |
| 936 | #define UNICODE_IS_BYTE_ORDER_MARK(uv) ((UV) (uv) == UNICODE_BYTE_ORDER_MARK) |
| 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) |
| 954 | |
| 955 | #define LATIN_SMALL_LETTER_SHARP_S LATIN_SMALL_LETTER_SHARP_S_NATIVE |
| 956 | #define LATIN_SMALL_LETTER_Y_WITH_DIAERESIS \ |
| 957 | LATIN_SMALL_LETTER_Y_WITH_DIAERESIS_NATIVE |
| 958 | #define MICRO_SIGN MICRO_SIGN_NATIVE |
| 959 | #define LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE \ |
| 960 | LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE_NATIVE |
| 961 | #define LATIN_SMALL_LETTER_A_WITH_RING_ABOVE \ |
| 962 | LATIN_SMALL_LETTER_A_WITH_RING_ABOVE_NATIVE |
| 963 | #define UNICODE_GREEK_CAPITAL_LETTER_SIGMA 0x03A3 |
| 964 | #define UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA 0x03C2 |
| 965 | #define UNICODE_GREEK_SMALL_LETTER_SIGMA 0x03C3 |
| 966 | #define GREEK_SMALL_LETTER_MU 0x03BC |
| 967 | #define GREEK_CAPITAL_LETTER_MU 0x039C /* Upper and title case |
| 968 | of MICRON */ |
| 969 | #define LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS 0x0178 /* Also is title case */ |
| 970 | #ifdef LATIN_CAPITAL_LETTER_SHARP_S_UTF8 |
| 971 | # define LATIN_CAPITAL_LETTER_SHARP_S 0x1E9E |
| 972 | #endif |
| 973 | #define LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE 0x130 |
| 974 | #define LATIN_SMALL_LETTER_DOTLESS_I 0x131 |
| 975 | #define LATIN_SMALL_LETTER_LONG_S 0x017F |
| 976 | #define LATIN_SMALL_LIGATURE_LONG_S_T 0xFB05 |
| 977 | #define LATIN_SMALL_LIGATURE_ST 0xFB06 |
| 978 | #define KELVIN_SIGN 0x212A |
| 979 | #define ANGSTROM_SIGN 0x212B |
| 980 | |
| 981 | #define UNI_DISPLAY_ISPRINT 0x0001 |
| 982 | #define UNI_DISPLAY_BACKSLASH 0x0002 |
| 983 | #define UNI_DISPLAY_QQ (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH) |
| 984 | #define UNI_DISPLAY_REGEX (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH) |
| 985 | |
| 986 | #define ANYOF_FOLD_SHARP_S(node, input, end) \ |
| 987 | (ANYOF_BITMAP_TEST(node, LATIN_SMALL_LETTER_SHARP_S) && \ |
| 988 | (ANYOF_NONBITMAP(node)) && \ |
| 989 | (ANYOF_FLAGS(node) & ANYOF_LOC_NONBITMAP_FOLD) && \ |
| 990 | ((end) > (input) + 1) && \ |
| 991 | isALPHA_FOLD_EQ((input)[0], 's')) |
| 992 | |
| 993 | #define SHARP_S_SKIP 2 |
| 994 | |
| 995 | #define is_utf8_char_buf(buf, buf_end) isUTF8_CHAR(buf, buf_end) |
| 996 | #define bytes_from_utf8(s, lenp, is_utf8p) \ |
| 997 | bytes_from_utf8_loc(s, lenp, is_utf8p, 0) |
| 998 | |
| 999 | /* |
| 1000 | |
| 1001 | =for apidoc Am|STRLEN|isUTF8_CHAR_flags|const U8 *s|const U8 *e| const U32 flags |
| 1002 | |
| 1003 | Evaluates to non-zero if the first few bytes of the string starting at C<s> and |
| 1004 | looking no further than S<C<e - 1>> are well-formed UTF-8, as extended by Perl, |
| 1005 | that represents some code point, subject to the restrictions given by C<flags>; |
| 1006 | otherwise it evaluates to 0. If non-zero, the value gives how many bytes |
| 1007 | starting at C<s> comprise the code point's representation. Any bytes remaining |
| 1008 | before C<e>, but beyond the ones needed to form the first code point in C<s>, |
| 1009 | are not examined. |
| 1010 | |
| 1011 | If C<flags> is 0, this gives the same results as C<L</isUTF8_CHAR>>; |
| 1012 | if C<flags> is C<UTF8_DISALLOW_ILLEGAL_INTERCHANGE>, this gives the same results |
| 1013 | as C<L</isSTRICT_UTF8_CHAR>>; |
| 1014 | and if C<flags> is C<UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE>, this gives |
| 1015 | the same results as C<L</isC9_STRICT_UTF8_CHAR>>. |
| 1016 | Otherwise C<flags> may be any combination of the C<UTF8_DISALLOW_I<foo>> flags |
| 1017 | understood by C<L</utf8n_to_uvchr>>, with the same meanings. |
| 1018 | |
| 1019 | The three alternative macros are for the most commonly needed validations; they |
| 1020 | are likely to run somewhat faster than this more general one, as they can be |
| 1021 | inlined into your code. |
| 1022 | |
| 1023 | Use L</is_utf8_string_flags>, L</is_utf8_string_loc_flags>, and |
| 1024 | L</is_utf8_string_loclen_flags> to check entire strings. |
| 1025 | |
| 1026 | =cut |
| 1027 | */ |
| 1028 | |
| 1029 | #define isUTF8_CHAR_flags(s, e, flags) \ |
| 1030 | (UNLIKELY((e) <= (s)) \ |
| 1031 | ? 0 \ |
| 1032 | : (UTF8_IS_INVARIANT(*s)) \ |
| 1033 | ? 1 \ |
| 1034 | : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \ |
| 1035 | ? 0 \ |
| 1036 | : is_utf8_char_helper(s, e, flags)) |
| 1037 | |
| 1038 | /* Do not use; should be deprecated. Use isUTF8_CHAR() instead; this is |
| 1039 | * retained solely for backwards compatibility */ |
| 1040 | #define IS_UTF8_CHAR(p, n) (isUTF8_CHAR(p, (p) + (n)) == n) |
| 1041 | |
| 1042 | #endif /* PERL_UTF8_H_ */ |
| 1043 | |
| 1044 | /* |
| 1045 | * ex: set ts=8 sts=4 sw=4 et: |
| 1046 | */ |