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 /* 148 149 =for apidoc Am|U8|NATIVE_TO_LATIN1|U8 ch 150 151 Returns the Latin-1 (including ASCII and control characters) equivalent of the 152 input native code point given by C<ch>. Thus, C<NATIVE_TO_LATIN1(193)> on 153 EBCDIC platforms returns 65. These each represent the character C<"A"> on 154 their respective platforms. On ASCII platforms no conversion is needed, so 155 this macro expands to just its input, adding no time nor space requirements to 156 the implementation. 157 158 For conversion of code points potentially larger than will fit in a character, 159 use L</NATIVE_TO_UNI>. 160 161 =for apidoc Am|U8|LATIN1_TO_NATIVE|U8 ch 162 163 Returns the native equivalent of the input Latin-1 code point (including ASCII 164 and control characters) given by C<ch>. Thus, C<LATIN1_TO_NATIVE(66)> on 165 EBCDIC platforms returns 194. These each represent the character C<"B"> on 166 their respective platforms. On ASCII platforms no conversion is needed, so 167 this macro expands to just its input, adding no time nor space requirements to 168 the implementation. 169 170 For conversion of code points potentially larger than will fit in a character, 171 use L</UNI_TO_NATIVE>. 172 173 =for apidoc Am|UV|NATIVE_TO_UNI|UV ch 174 175 Returns the Unicode equivalent of the input native code point given by C<ch>. 176 Thus, C<NATIVE_TO_UNI(195)> on EBCDIC platforms returns 67. These each 177 represent the character C<"C"> on their respective platforms. On ASCII 178 platforms no conversion is needed, so this macro expands to just its input, 179 adding no time nor space requirements to the implementation. 180 181 =for apidoc Am|UV|UNI_TO_NATIVE|UV ch 182 183 Returns the native equivalent of the input Unicode code point given by C<ch>. 184 Thus, C<UNI_TO_NATIVE(68)> on EBCDIC platforms returns 196. These each 185 represent the character C<"D"> on their respective platforms. On ASCII 186 platforms no conversion is needed, so this macro expands to just its input, 187 adding no time nor space requirements to the implementation. 188 189 =cut 190 */ 191 192 #define NATIVE_TO_LATIN1(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) ((ch) | 0))) 193 #define LATIN1_TO_NATIVE(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) ((ch) | 0))) 194 195 /* I8 is an intermediate version of UTF-8 used only in UTF-EBCDIC. We thus 196 * consider it to be identical to UTF-8 on ASCII platforms. Strictly speaking 197 * UTF-8 and UTF-EBCDIC are two different things, but we often conflate them 198 * because they are 8-bit encodings that serve the same purpose in Perl, and 199 * rarely do we need to distinguish them. The term "NATIVE_UTF8" applies to 200 * whichever one is applicable on the current platform */ 201 #define NATIVE_UTF8_TO_I8(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) ((ch) | 0))) 202 #define I8_TO_NATIVE_UTF8(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) ((ch) | 0))) 203 204 #define UNI_TO_NATIVE(ch) ((UV) ((ch) | 0)) 205 #define NATIVE_TO_UNI(ch) ((UV) ((ch) | 0)) 206 207 /* 208 209 The following table is from Unicode 3.2, plus the Perl extensions for above 210 U+10FFFF 211 212 Code Points 1st Byte 2nd Byte 3rd 4th 5th 6th 7th 8th-13th 213 214 U+0000..U+007F 00..7F 215 U+0080..U+07FF * C2..DF 80..BF 216 U+0800..U+0FFF E0 * A0..BF 80..BF 217 U+1000..U+CFFF E1..EC 80..BF 80..BF 218 U+D000..U+D7FF ED 80..9F 80..BF 219 U+D800..U+DFFF ED A0..BF 80..BF (surrogates) 220 U+E000..U+FFFF EE..EF 80..BF 80..BF 221 U+10000..U+3FFFF F0 * 90..BF 80..BF 80..BF 222 U+40000..U+FFFFF F1..F3 80..BF 80..BF 80..BF 223 U+100000..U+10FFFF F4 80..8F 80..BF 80..BF 224 Below are above-Unicode code points 225 U+110000..U+13FFFF F4 90..BF 80..BF 80..BF 226 U+110000..U+1FFFFF F5..F7 80..BF 80..BF 80..BF 227 U+200000..U+FFFFFF F8 * 88..BF 80..BF 80..BF 80..BF 228 U+1000000..U+3FFFFFF F9..FB 80..BF 80..BF 80..BF 80..BF 229 U+4000000..U+3FFFFFFF FC * 84..BF 80..BF 80..BF 80..BF 80..BF 230 U+40000000..U+7FFFFFFF FD 80..BF 80..BF 80..BF 80..BF 80..BF 231 U+80000000..U+FFFFFFFFF FE * 82..BF 80..BF 80..BF 80..BF 80..BF 80..BF 232 U+1000000000.. FF 80..BF 80..BF 80..BF 80..BF 80..BF * 81..BF 80..BF 233 234 Note the gaps before several of the byte entries above marked by '*'. These are 235 caused by legal UTF-8 avoiding non-shortest encodings: it is technically 236 possible to UTF-8-encode a single code point in different ways, but that is 237 explicitly forbidden, and the shortest possible encoding should always be used 238 (and that is what Perl does). The non-shortest ones are called 'overlongs'. 239 240 */ 241 242 /* 243 Another way to look at it, as bits: 244 245 Code Points 1st Byte 2nd Byte 3rd Byte 4th Byte 246 247 0aaa aaaa 0aaa aaaa 248 0000 0bbb bbaa aaaa 110b bbbb 10aa aaaa 249 cccc bbbb bbaa aaaa 1110 cccc 10bb bbbb 10aa aaaa 250 00 000d ddcc cccc bbbb bbaa aaaa 1111 0ddd 10cc cccc 10bb bbbb 10aa aaaa 251 252 As you can see, the continuation bytes all begin with C<10>, and the 253 leading bits of the start byte tell how many bytes there are in the 254 encoded character. 255 256 Perl's extended UTF-8 means we can have start bytes up through FF, though any 257 beginning with FF yields a code point that is too large for 32-bit ASCII 258 platforms. FF signals to use 13 bytes for the encoded character. This breaks 259 the paradigm that the number of leading bits gives how many total bytes there 260 are in the character. */ 261 262 /* This is the number of low-order bits a continuation byte in a UTF-8 encoded 263 * sequence contributes to the specification of the code point. In the bit 264 * maps above, you see that the first 2 bits are a constant '10', leaving 6 of 265 * real information */ 266 #define UTF_ACCUMULATION_SHIFT 6 267 268 /* ^? is defined to be DEL on ASCII systems. See the definition of toCTRL() 269 * for more */ 270 #define QUESTION_MARK_CTRL DEL_NATIVE 271 272 /* Surrogates, non-character code points and above-Unicode code points are 273 * problematic in some contexts. This allows code that needs to check for 274 * those to quickly exclude the vast majority of code points it will 275 * encounter */ 276 #define isUTF8_POSSIBLY_PROBLEMATIC(c) (__ASSERT_(FITS_IN_8_BITS(c)) \ 277 (U8) c >= 0xED) 278 279 #define UNICODE_IS_PERL_EXTENDED(uv) UNLIKELY((UV) (uv) > 0x7FFFFFFF) 280 281 #endif /* EBCDIC vs ASCII */ 282 283 /* 2**UTF_ACCUMULATION_SHIFT - 1. This masks out all but the bits that carry 284 * real information in a continuation byte. This turns out to be 0x3F in 285 * UTF-8, 0x1F in UTF-EBCDIC. */ 286 #define UTF_CONTINUATION_MASK ((U8) ((1U << UTF_ACCUMULATION_SHIFT) - 1)) 287 288 /* For use in UTF8_IS_CONTINUATION(). This turns out to be 0xC0 in UTF-8, 289 * E0 in UTF-EBCDIC */ 290 #define UTF_IS_CONTINUATION_MASK ((U8) (0xFF << UTF_ACCUMULATION_SHIFT)) 291 292 /* This defines the bits that are to be in the continuation bytes of a 293 * multi-byte UTF-8 encoded character that mark it is a continuation byte. 294 * This turns out to be 0x80 in UTF-8, 0xA0 in UTF-EBCDIC. (khw doesn't know 295 * the underlying reason that B0 works here) */ 296 #define UTF_CONTINUATION_MARK (UTF_IS_CONTINUATION_MASK & 0xB0) 297 298 /* Is the byte 'c' part of a multi-byte UTF8-8 encoded sequence, and not the 299 * first byte thereof? */ 300 #define UTF8_IS_CONTINUATION(c) (__ASSERT_(FITS_IN_8_BITS(c)) \ 301 (((NATIVE_UTF8_TO_I8(c) & UTF_IS_CONTINUATION_MASK) \ 302 == UTF_CONTINUATION_MARK))) 303 304 /* Is the representation of the Unicode code point 'cp' the same regardless of 305 * being encoded in UTF-8 or not? This is a fundamental property of 306 * UTF-8,EBCDIC */ 307 #define OFFUNI_IS_INVARIANT(c) (((WIDEST_UTYPE)(c)) < UTF_CONTINUATION_MARK) 308 309 /* 310 =for apidoc Am|bool|UVCHR_IS_INVARIANT|UV cp 311 312 Evaluates to 1 if the representation of code point C<cp> is the same whether or 313 not it is encoded in UTF-8; otherwise evaluates to 0. UTF-8 invariant 314 characters can be copied as-is when converting to/from UTF-8, saving time. 315 C<cp> is Unicode if above 255; otherwise is platform-native. 316 317 =cut 318 */ 319 #if defined(__m88k__) 320 /* XXX workaround: m88k gcc3 produces wrong code with NATIVE_TO_UNI() */ 321 #define UVCHR_IS_INVARIANT(cp) (OFFUNI_IS_INVARIANT(cp)) 322 #else /* the original one */ 323 #define UVCHR_IS_INVARIANT(cp) (OFFUNI_IS_INVARIANT(NATIVE_TO_UNI(cp))) 324 #endif 325 326 /* Internal macro to be used only in this file to aid in constructing other 327 * publicly accessible macros. 328 * The number of bytes required to express this uv in UTF-8, for just those 329 * uv's requiring 2 through 6 bytes, as these are common to all platforms and 330 * word sizes. The number of bytes needed is given by the number of leading 1 331 * bits in the start byte. There are 32 start bytes that have 2 initial 1 bits 332 * (C0-DF); there are 16 that have 3 initial 1 bits (E0-EF); 8 that have 4 333 * initial 1 bits (F0-F8); 4 that have 5 initial 1 bits (F9-FB), and 2 that 334 * have 6 initial 1 bits (FC-FD). The largest number a string of n bytes can 335 * represent is (the number of possible start bytes for 'n') 336 * * (the number of possiblities for each start byte 337 * The latter in turn is 338 * 2 ** ( (how many continuation bytes there are) 339 * * (the number of bits of information each 340 * continuation byte holds)) 341 * 342 * If we were on a platform where we could use a fast find first set bit 343 * instruction (or count leading zeros instruction) this could be replaced by 344 * using that to find the log2 of the uv, and divide that by the number of bits 345 * of information in each continuation byte, adjusting for large cases and how 346 * much information is in a start byte for that length */ 347 #define __COMMON_UNI_SKIP(uv) \ 348 (UV) (uv) < (32 * (1U << ( UTF_ACCUMULATION_SHIFT))) ? 2 : \ 349 (UV) (uv) < (16 * (1U << (2 * UTF_ACCUMULATION_SHIFT))) ? 3 : \ 350 (UV) (uv) < ( 8 * (1U << (3 * UTF_ACCUMULATION_SHIFT))) ? 4 : \ 351 (UV) (uv) < ( 4 * (1U << (4 * UTF_ACCUMULATION_SHIFT))) ? 5 : \ 352 (UV) (uv) < ( 2 * (1U << (5 * UTF_ACCUMULATION_SHIFT))) ? 6 : 353 354 /* Internal macro to be used only in this file. 355 * This adds to __COMMON_UNI_SKIP the details at this platform's upper range. 356 * For any-sized EBCDIC platforms, or 64-bit ASCII ones, we need one more test 357 * to see if just 7 bytes is needed, or if the maximum is needed. For 32-bit 358 * ASCII platforms, everything is representable by 7 bytes */ 359 #if defined(UV_IS_QUAD) || defined(EBCDIC) 360 # define __BASE_UNI_SKIP(uv) (__COMMON_UNI_SKIP(uv) \ 361 (UV) (uv) < ((UV) 1U << (6 * UTF_ACCUMULATION_SHIFT)) ? 7 : UTF8_MAXBYTES) 362 #else 363 # define __BASE_UNI_SKIP(uv) (__COMMON_UNI_SKIP(uv) 7) 364 #endif 365 366 /* The next two macros use the base macro defined above, and add in the tests 367 * at the low-end of the range, for just 1 byte, yielding complete macros, 368 * publicly accessible. */ 369 370 /* Input is a true Unicode (not-native) code point */ 371 #define OFFUNISKIP(uv) (OFFUNI_IS_INVARIANT(uv) ? 1 : __BASE_UNI_SKIP(uv)) 372 373 /* 374 375 =for apidoc Am|STRLEN|UVCHR_SKIP|UV cp 376 returns the number of bytes required to represent the code point C<cp> when 377 encoded as UTF-8. C<cp> is a native (ASCII or EBCDIC) code point if less than 378 255; a Unicode code point otherwise. 379 380 =cut 381 */ 382 #define UVCHR_SKIP(uv) ( UVCHR_IS_INVARIANT(uv) ? 1 : __BASE_UNI_SKIP(uv)) 383 384 #define UTF_MIN_START_BYTE \ 385 ((UTF_CONTINUATION_MARK >> UTF_ACCUMULATION_SHIFT) | UTF_START_MARK(2)) 386 387 /* Is the byte 'c' the first byte of a multi-byte UTF8-8 encoded sequence? 388 * This excludes invariants (they are single-byte). It also excludes the 389 * illegal overlong sequences that begin with C0 and C1 on ASCII platforms, and 390 * C0-C4 I8 start bytes on EBCDIC ones. On EBCDIC E0 can't start a 391 * non-overlong sequence, so we define a base macro and for those platforms, 392 * extend it to also exclude E0 */ 393 #define UTF8_IS_START_base(c) (__ASSERT_(FITS_IN_8_BITS(c)) \ 394 (NATIVE_UTF8_TO_I8(c) >= UTF_MIN_START_BYTE)) 395 #ifdef EBCDIC 396 # define UTF8_IS_START(c) \ 397 (UTF8_IS_START_base(c) && (c) != I8_TO_NATIVE_UTF8(0xE0)) 398 #else 399 # define UTF8_IS_START(c) UTF8_IS_START_base(c) 400 #endif 401 402 #define UTF_MIN_ABOVE_LATIN1_BYTE \ 403 ((0x100 >> UTF_ACCUMULATION_SHIFT) | UTF_START_MARK(2)) 404 405 /* Is the UTF8-encoded byte 'c' the first byte of a sequence of bytes that 406 * represent a code point > 255? */ 407 #define UTF8_IS_ABOVE_LATIN1(c) (__ASSERT_(FITS_IN_8_BITS(c)) \ 408 (NATIVE_UTF8_TO_I8(c) >= UTF_MIN_ABOVE_LATIN1_BYTE)) 409 410 /* Is the UTF8-encoded byte 'c' the first byte of a two byte sequence? Use 411 * UTF8_IS_NEXT_CHAR_DOWNGRADEABLE() instead if the input isn't known to 412 * be well-formed. */ 413 #define UTF8_IS_DOWNGRADEABLE_START(c) (__ASSERT_(FITS_IN_8_BITS(c)) \ 414 inRANGE(NATIVE_UTF8_TO_I8(c), \ 415 UTF_MIN_START_BYTE, UTF_MIN_ABOVE_LATIN1_BYTE - 1)) 416 417 /* The largest code point representable by two UTF-8 bytes on this platform. 418 * As explained in the comments for __COMMON_UNI_SKIP, 32 start bytes with 419 * UTF_ACCUMULATION_SHIFT bits of information each */ 420 #define MAX_UTF8_TWO_BYTE (32 * (1U << UTF_ACCUMULATION_SHIFT) - 1) 421 422 /* The largest code point representable by two UTF-8 bytes on any platform that 423 * Perl runs on. This value is constrained by EBCDIC which has 5 bits per 424 * continuation byte */ 425 #define MAX_PORTABLE_UTF8_TWO_BYTE (32 * (1U << 5) - 1) 426 427 /* 428 429 =for apidoc AmnU|STRLEN|UTF8_MAXBYTES_CASE 430 431 The maximum number of UTF-8 bytes a single Unicode character can 432 uppercase/lowercase/titlecase/fold into. 433 434 =cut 435 436 * Unicode guarantees that the maximum expansion is UTF8_MAX_FOLD_CHAR_EXPAND 437 * characters, but any above-Unicode code point will fold to itself, so we only 438 * have to look at the expansion of the maximum Unicode code point. But this 439 * number may be less than the space occupied by a very large code point under 440 * Perl's extended UTF-8. We have to make it large enough to fit any single 441 * character. (It turns out that ASCII and EBCDIC differ in which is larger) 442 * 443 =cut 444 */ 445 #define UTF8_MAXBYTES_CASE \ 446 MAX(UTF8_MAXBYTES, UTF8_MAX_FOLD_CHAR_EXPAND * OFFUNISKIP(0x10FFFF)) 447 448 /* Rest of these are attributes of Unicode and perl's internals rather than the 449 * encoding, or happen to be the same in both ASCII and EBCDIC (at least at 450 * this level; the macros that some of these call may have different 451 * definitions in the two encodings */ 452 453 /* In domain restricted to ASCII, these may make more sense to the reader than 454 * the ones with Latin1 in the name */ 455 #define NATIVE_TO_ASCII(ch) NATIVE_TO_LATIN1(ch) 456 #define ASCII_TO_NATIVE(ch) LATIN1_TO_NATIVE(ch) 457 458 /* More or less misleadingly-named defines, retained for back compat */ 459 #define NATIVE_TO_UTF(ch) NATIVE_UTF8_TO_I8(ch) 460 #define NATIVE_TO_I8(ch) NATIVE_UTF8_TO_I8(ch) 461 #define UTF_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch) 462 #define I8_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch) 463 #define NATIVE8_TO_UNI(ch) NATIVE_TO_LATIN1(ch) 464 465 /* This defines the 1-bits that are to be in the first byte of a multi-byte 466 * UTF-8 encoded character that mark it as a start byte and give the number of 467 * bytes that comprise the character. 'len' is the number of bytes in the 468 * multi-byte sequence. */ 469 #define UTF_START_MARK(len) (((len) > 7) ? 0xFF : ((U8) (0xFE << (7-(len))))) 470 471 /* Masks out the initial one bits in a start byte, leaving the real data ones. 472 * Doesn't work on an invariant byte. 'len' is the number of bytes in the 473 * multi-byte sequence that comprises the character. */ 474 #define UTF_START_MASK(len) (((len) >= 7) ? 0x00 : (0x1F >> ((len)-2))) 475 476 /* Adds a UTF8 continuation byte 'new' of information to a running total code 477 * point 'old' of all the continuation bytes so far. This is designed to be 478 * used in a loop to convert from UTF-8 to the code point represented. Note 479 * that this is asymmetric on EBCDIC platforms, in that the 'new' parameter is 480 * the UTF-EBCDIC byte, whereas the 'old' parameter is a Unicode (not EBCDIC) 481 * code point in process of being generated */ 482 #define UTF8_ACCUMULATE(old, new) (__ASSERT_(FITS_IN_8_BITS(new)) \ 483 ((old) << UTF_ACCUMULATION_SHIFT) \ 484 | ((NATIVE_UTF8_TO_I8(new)) \ 485 & UTF_CONTINUATION_MASK)) 486 487 /* This works in the face of malformed UTF-8. */ 488 #define UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(s, e) \ 489 ( UTF8_IS_DOWNGRADEABLE_START(*(s)) \ 490 && ( (e) - (s) > 1) \ 491 && UTF8_IS_CONTINUATION(*((s)+1))) 492 493 /* Number of bytes a code point occupies in UTF-8. */ 494 #define NATIVE_SKIP(uv) UVCHR_SKIP(uv) 495 496 /* Most code which says UNISKIP is really thinking in terms of native code 497 * points (0-255) plus all those beyond. This is an imprecise term, but having 498 * it means existing code continues to work. For precision, use UVCHR_SKIP, 499 * NATIVE_SKIP, or OFFUNISKIP */ 500 #define UNISKIP(uv) UVCHR_SKIP(uv) 501 502 /* Longer, but more accurate name */ 503 #define UTF8_IS_ABOVE_LATIN1_START(c) UTF8_IS_ABOVE_LATIN1(c) 504 505 /* Convert a UTF-8 variant Latin1 character to a native code point value. 506 * Needs just one iteration of accumulate. Should be used only if it is known 507 * that the code point is < 256, and is not UTF-8 invariant. Use the slower 508 * but more general TWO_BYTE_UTF8_TO_NATIVE() which handles any code point 509 * representable by two bytes (which turns out to be up through 510 * MAX_PORTABLE_UTF8_TWO_BYTE). The two parameters are: 511 * HI: a downgradable start byte; 512 * LO: continuation. 513 * */ 514 #define EIGHT_BIT_UTF8_TO_NATIVE(HI, LO) \ 515 ( __ASSERT_(UTF8_IS_DOWNGRADEABLE_START(HI)) \ 516 __ASSERT_(UTF8_IS_CONTINUATION(LO)) \ 517 LATIN1_TO_NATIVE(UTF8_ACCUMULATE(( \ 518 NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), (LO)))) 519 520 /* Convert a two (not one) byte utf8 character to a native code point value. 521 * Needs just one iteration of accumulate. Should not be used unless it is 522 * known that the two bytes are legal: 1) two-byte start, and 2) continuation. 523 * Note that the result can be larger than 255 if the input character is not 524 * downgradable */ 525 #define TWO_BYTE_UTF8_TO_NATIVE(HI, LO) \ 526 (__ASSERT_(FITS_IN_8_BITS(HI)) \ 527 __ASSERT_(FITS_IN_8_BITS(LO)) \ 528 __ASSERT_(PL_utf8skip[HI] == 2) \ 529 __ASSERT_(UTF8_IS_CONTINUATION(LO)) \ 530 UNI_TO_NATIVE(UTF8_ACCUMULATE((NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), \ 531 (LO)))) 532 533 /* Should never be used, and be deprecated */ 534 #define TWO_BYTE_UTF8_TO_UNI(HI, LO) NATIVE_TO_UNI(TWO_BYTE_UTF8_TO_NATIVE(HI, LO)) 535 536 /* 537 538 =for apidoc Am|STRLEN|UTF8SKIP|char* s 539 returns the number of bytes a non-malformed UTF-8 encoded character whose first 540 (perhaps only) byte is pointed to by C<s>. 541 542 If there is a possibility of malformed input, use instead: 543 544 =over 545 546 =item L</C<UTF8_SAFE_SKIP>> if you know the maximum ending pointer in the 547 buffer pointed to by C<s>; or 548 549 =item L</C<UTF8_CHK_SKIP>> if you don't know it. 550 551 =back 552 553 It is better to restructure your code so the end pointer is passed down so that 554 you know what it actually is at the point of this call, but if that isn't 555 possible, L</C<UTF8_CHK_SKIP>> can minimize the chance of accessing beyond the end 556 of the input buffer. 557 558 =cut 559 */ 560 #define UTF8SKIP(s) PL_utf8skip[*(const U8*)(s)] 561 562 /* 563 =for apidoc Am|STRLEN|UTF8_SKIP|char* s 564 This is a synonym for L</C<UTF8SKIP>> 565 566 =cut 567 */ 568 569 #define UTF8_SKIP(s) UTF8SKIP(s) 570 571 /* 572 =for apidoc Am|STRLEN|UTF8_CHK_SKIP|char* s 573 574 This is a safer version of L</C<UTF8SKIP>>, but still not as safe as 575 L</C<UTF8_SAFE_SKIP>>. This version doesn't blindly assume that the input 576 string pointed to by C<s> is well-formed, but verifies that there isn't a NUL 577 terminating character before the expected end of the next character in C<s>. 578 The length C<UTF8_CHK_SKIP> returns stops just before any such NUL. 579 580 Perl tends to add NULs, as an insurance policy, after the end of strings in 581 SV's, so it is likely that using this macro will prevent inadvertent reading 582 beyond the end of the input buffer, even if it is malformed UTF-8. 583 584 This macro is intended to be used by XS modules where the inputs could be 585 malformed, and it isn't feasible to restructure to use the safer 586 L</C<UTF8_SAFE_SKIP>>, for example when interfacing with a C library. 587 588 =cut 589 */ 590 591 #define UTF8_CHK_SKIP(s) \ 592 (s[0] == '\0' ? 1 : MIN(UTF8SKIP(s), \ 593 my_strnlen((char *) (s), UTF8SKIP(s)))) 594 /* 595 596 =for apidoc Am|STRLEN|UTF8_SAFE_SKIP|char* s|char* e 597 returns 0 if S<C<s E<gt>= e>>; otherwise returns the number of bytes in the 598 UTF-8 encoded character whose first byte is pointed to by C<s>. But it never 599 returns beyond C<e>. On DEBUGGING builds, it asserts that S<C<s E<lt>= e>>. 600 601 =cut 602 */ 603 #define UTF8_SAFE_SKIP(s, e) (__ASSERT_((e) >= (s)) \ 604 ((e) - (s)) <= 0 \ 605 ? 0 \ 606 : MIN(((e) - (s)), UTF8_SKIP(s))) 607 608 /* Most code that says 'UNI_' really means the native value for code points up 609 * through 255 */ 610 #define UNI_IS_INVARIANT(cp) UVCHR_IS_INVARIANT(cp) 611 612 /* 613 =for apidoc Am|bool|UTF8_IS_INVARIANT|char c 614 615 Evaluates to 1 if the byte C<c> represents the same character when encoded in 616 UTF-8 as when not; otherwise evaluates to 0. UTF-8 invariant characters can be 617 copied as-is when converting to/from UTF-8, saving time. 618 619 In spite of the name, this macro gives the correct result if the input string 620 from which C<c> comes is not encoded in UTF-8. 621 622 See C<L</UVCHR_IS_INVARIANT>> for checking if a UV is invariant. 623 624 =cut 625 626 The reason it works on both UTF-8 encoded strings and non-UTF-8 encoded, is 627 that it returns TRUE in each for the exact same set of bit patterns. It is 628 valid on a subset of what UVCHR_IS_INVARIANT is valid on, so can just use that; 629 and the compiler should optimize out anything extraneous given the 630 implementation of the latter. The |0 makes sure this isn't mistakenly called 631 with a ptr argument. 632 */ 633 #define UTF8_IS_INVARIANT(c) UVCHR_IS_INVARIANT((c) | 0) 634 635 /* Like the above, but its name implies a non-UTF8 input, which as the comments 636 * above show, doesn't matter as to its implementation */ 637 #define NATIVE_BYTE_IS_INVARIANT(c) UVCHR_IS_INVARIANT(c) 638 639 /* Misleadingly named: is the UTF8-encoded byte 'c' part of a variant sequence 640 * in UTF-8? This is the inverse of UTF8_IS_INVARIANT. */ 641 #define UTF8_IS_CONTINUED(c) (__ASSERT_(FITS_IN_8_BITS(c)) \ 642 (! UTF8_IS_INVARIANT(c))) 643 644 /* The macros in the next 4 sets are used to generate the two utf8 or utfebcdic 645 * bytes from an ordinal that is known to fit into exactly two (not one) bytes; 646 * it must be less than 0x3FF to work across both encodings. */ 647 648 /* These two are helper macros for the other three sets, and should not be used 649 * directly anywhere else. 'translate_function' is either NATIVE_TO_LATIN1 650 * (which works for code points up through 0xFF) or NATIVE_TO_UNI which works 651 * for any code point */ 652 #define __BASE_TWO_BYTE_HI(c, translate_function) \ 653 (__ASSERT_(! UVCHR_IS_INVARIANT(c)) \ 654 I8_TO_NATIVE_UTF8((translate_function(c) >> UTF_ACCUMULATION_SHIFT) \ 655 | UTF_START_MARK(2))) 656 #define __BASE_TWO_BYTE_LO(c, translate_function) \ 657 (__ASSERT_(! UVCHR_IS_INVARIANT(c)) \ 658 I8_TO_NATIVE_UTF8((translate_function(c) & UTF_CONTINUATION_MASK) \ 659 | UTF_CONTINUATION_MARK)) 660 661 /* The next two macros should not be used. They were designed to be usable as 662 * the case label of a switch statement, but this doesn't work for EBCDIC. Use 663 * regen/unicode_constants.pl instead */ 664 #define UTF8_TWO_BYTE_HI_nocast(c) __BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI) 665 #define UTF8_TWO_BYTE_LO_nocast(c) __BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI) 666 667 /* The next two macros are used when the source should be a single byte 668 * character; checked for under DEBUGGING */ 669 #define UTF8_EIGHT_BIT_HI(c) (__ASSERT_(FITS_IN_8_BITS(c)) \ 670 ( __BASE_TWO_BYTE_HI(c, NATIVE_TO_LATIN1))) 671 #define UTF8_EIGHT_BIT_LO(c) (__ASSERT_(FITS_IN_8_BITS(c)) \ 672 (__BASE_TWO_BYTE_LO(c, NATIVE_TO_LATIN1))) 673 674 /* These final two macros in the series are used when the source can be any 675 * code point whose UTF-8 is known to occupy 2 bytes; they are less efficient 676 * than the EIGHT_BIT versions on EBCDIC platforms. We use the logical '~' 677 * operator instead of "<=" to avoid getting compiler warnings. 678 * MAX_UTF8_TWO_BYTE should be exactly all one bits in the lower few 679 * places, so the ~ works */ 680 #define UTF8_TWO_BYTE_HI(c) \ 681 (__ASSERT_((sizeof(c) == 1) \ 682 || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \ 683 (__BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI))) 684 #define UTF8_TWO_BYTE_LO(c) \ 685 (__ASSERT_((sizeof(c) == 1) \ 686 || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \ 687 (__BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI))) 688 689 /* This is illegal in any well-formed UTF-8 in both EBCDIC and ASCII 690 * as it is only in overlongs. */ 691 #define ILLEGAL_UTF8_BYTE I8_TO_NATIVE_UTF8(0xC1) 692 693 /* 694 * 'UTF' is whether or not p is encoded in UTF8. The names 'foo_lazy_if' stem 695 * from an earlier version of these macros in which they didn't call the 696 * foo_utf8() macros (i.e. were 'lazy') unless they decided that *p is the 697 * beginning of a utf8 character. Now that foo_utf8() determines that itself, 698 * no need to do it again here 699 */ 700 #define isIDFIRST_lazy_if_safe(p, e, UTF) \ 701 ((IN_BYTES || !UTF) \ 702 ? isIDFIRST(*(p)) \ 703 : isIDFIRST_utf8_safe(p, e)) 704 #define isWORDCHAR_lazy_if_safe(p, e, UTF) \ 705 ((IN_BYTES || !UTF) \ 706 ? isWORDCHAR(*(p)) \ 707 : isWORDCHAR_utf8_safe((U8 *) p, (U8 *) e)) 708 #define isALNUM_lazy_if_safe(p, e, UTF) isWORDCHAR_lazy_if_safe(p, e, UTF) 709 710 #define UTF8_MAXLEN UTF8_MAXBYTES 711 712 /* A Unicode character can fold to up to 3 characters */ 713 #define UTF8_MAX_FOLD_CHAR_EXPAND 3 714 715 #define IN_BYTES UNLIKELY(CopHINTS_get(PL_curcop) & HINT_BYTES) 716 717 /* 718 719 =for apidoc Am|bool|DO_UTF8|SV* sv 720 Returns a bool giving whether or not the PV in C<sv> is to be treated as being 721 encoded in UTF-8. 722 723 You should use this I<after> a call to C<SvPV()> or one of its variants, in 724 case any call to string overloading updates the internal UTF-8 encoding flag. 725 726 =cut 727 */ 728 #define DO_UTF8(sv) (SvUTF8(sv) && !IN_BYTES) 729 730 /* Should all strings be treated as Unicode, and not just UTF-8 encoded ones? 731 * Is so within 'feature unicode_strings' or 'locale :not_characters', and not 732 * within 'use bytes'. UTF-8 locales are not tested for here, but perhaps 733 * could be */ 734 #define IN_UNI_8_BIT \ 735 (( ( (CopHINTS_get(PL_curcop) & HINT_UNI_8_BIT)) \ 736 || ( CopHINTS_get(PL_curcop) & HINT_LOCALE_PARTIAL \ 737 /* -1 below is for :not_characters */ \ 738 && _is_in_locale_category(FALSE, -1))) \ 739 && (! IN_BYTES)) 740 741 742 #define UTF8_ALLOW_EMPTY 0x0001 /* Allow a zero length string */ 743 #define UTF8_GOT_EMPTY UTF8_ALLOW_EMPTY 744 745 /* Allow first byte to be a continuation byte */ 746 #define UTF8_ALLOW_CONTINUATION 0x0002 747 #define UTF8_GOT_CONTINUATION UTF8_ALLOW_CONTINUATION 748 749 /* Unexpected non-continuation byte */ 750 #define UTF8_ALLOW_NON_CONTINUATION 0x0004 751 #define UTF8_GOT_NON_CONTINUATION UTF8_ALLOW_NON_CONTINUATION 752 753 /* expecting more bytes than were available in the string */ 754 #define UTF8_ALLOW_SHORT 0x0008 755 #define UTF8_GOT_SHORT UTF8_ALLOW_SHORT 756 757 /* Overlong sequence; i.e., the code point can be specified in fewer bytes. 758 * First one will convert the overlong to the REPLACEMENT CHARACTER; second 759 * will return what the overlong evaluates to */ 760 #define UTF8_ALLOW_LONG 0x0010 761 #define UTF8_ALLOW_LONG_AND_ITS_VALUE (UTF8_ALLOW_LONG|0x0020) 762 #define UTF8_GOT_LONG UTF8_ALLOW_LONG 763 764 #define UTF8_ALLOW_OVERFLOW 0x0080 765 #define UTF8_GOT_OVERFLOW UTF8_ALLOW_OVERFLOW 766 767 #define UTF8_DISALLOW_SURROGATE 0x0100 /* Unicode surrogates */ 768 #define UTF8_GOT_SURROGATE UTF8_DISALLOW_SURROGATE 769 #define UTF8_WARN_SURROGATE 0x0200 770 771 /* Unicode non-character code points */ 772 #define UTF8_DISALLOW_NONCHAR 0x0400 773 #define UTF8_GOT_NONCHAR UTF8_DISALLOW_NONCHAR 774 #define UTF8_WARN_NONCHAR 0x0800 775 776 /* Super-set of Unicode: code points above the legal max */ 777 #define UTF8_DISALLOW_SUPER 0x1000 778 #define UTF8_GOT_SUPER UTF8_DISALLOW_SUPER 779 #define UTF8_WARN_SUPER 0x2000 780 781 /* The original UTF-8 standard did not define UTF-8 with start bytes of 0xFE or 782 * 0xFF, though UTF-EBCDIC did. This allowed both versions to represent code 783 * points up to 2 ** 31 - 1. Perl extends UTF-8 so that 0xFE and 0xFF are 784 * usable on ASCII platforms, and 0xFF means something different than 785 * UTF-EBCDIC defines. These changes allow code points of 64 bits (actually 786 * somewhat more) to be represented on both platforms. But these are Perl 787 * extensions, and not likely to be interchangeable with other languages. Note 788 * that on ASCII platforms, FE overflows a signed 32-bit word, and FF an 789 * unsigned one. */ 790 #define UTF8_DISALLOW_PERL_EXTENDED 0x4000 791 #define UTF8_GOT_PERL_EXTENDED UTF8_DISALLOW_PERL_EXTENDED 792 #define UTF8_WARN_PERL_EXTENDED 0x8000 793 794 /* For back compat, these old names are misleading for overlongs and 795 * UTF_EBCDIC. */ 796 #define UTF8_DISALLOW_ABOVE_31_BIT UTF8_DISALLOW_PERL_EXTENDED 797 #define UTF8_GOT_ABOVE_31_BIT UTF8_GOT_PERL_EXTENDED 798 #define UTF8_WARN_ABOVE_31_BIT UTF8_WARN_PERL_EXTENDED 799 #define UTF8_DISALLOW_FE_FF UTF8_DISALLOW_PERL_EXTENDED 800 #define UTF8_WARN_FE_FF UTF8_WARN_PERL_EXTENDED 801 802 #define UTF8_CHECK_ONLY 0x10000 803 #define _UTF8_NO_CONFIDENCE_IN_CURLEN 0x20000 /* Internal core use only */ 804 805 /* For backwards source compatibility. They do nothing, as the default now 806 * includes what they used to mean. The first one's meaning was to allow the 807 * just the single non-character 0xFFFF */ 808 #define UTF8_ALLOW_FFFF 0 809 #define UTF8_ALLOW_FE_FF 0 810 #define UTF8_ALLOW_SURROGATE 0 811 812 /* C9 refers to Unicode Corrigendum #9: allows but discourages non-chars */ 813 #define UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE \ 814 (UTF8_DISALLOW_SUPER|UTF8_DISALLOW_SURROGATE) 815 #define UTF8_WARN_ILLEGAL_C9_INTERCHANGE (UTF8_WARN_SUPER|UTF8_WARN_SURROGATE) 816 817 #define UTF8_DISALLOW_ILLEGAL_INTERCHANGE \ 818 (UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE|UTF8_DISALLOW_NONCHAR) 819 #define UTF8_WARN_ILLEGAL_INTERCHANGE \ 820 (UTF8_WARN_ILLEGAL_C9_INTERCHANGE|UTF8_WARN_NONCHAR) 821 822 /* This is typically used for code that processes UTF-8 input and doesn't want 823 * to have to deal with any malformations that might be present. All such will 824 * be safely replaced by the REPLACEMENT CHARACTER, unless other flags 825 * overriding this are also present. */ 826 #define UTF8_ALLOW_ANY ( UTF8_ALLOW_CONTINUATION \ 827 |UTF8_ALLOW_NON_CONTINUATION \ 828 |UTF8_ALLOW_SHORT \ 829 |UTF8_ALLOW_LONG \ 830 |UTF8_ALLOW_OVERFLOW) 831 832 /* Accept any Perl-extended UTF-8 that evaluates to any UV on the platform, but 833 * not any malformed. This is the default. */ 834 #define UTF8_ALLOW_ANYUV 0 835 #define UTF8_ALLOW_DEFAULT UTF8_ALLOW_ANYUV 836 837 /* 838 =for apidoc Am|bool|UTF8_IS_SURROGATE|const U8 *s|const U8 *e 839 840 Evaluates to non-zero if the first few bytes of the string starting at C<s> and 841 looking no further than S<C<e - 1>> are well-formed UTF-8 that represents one 842 of the Unicode surrogate code points; otherwise it evaluates to 0. If 843 non-zero, the value gives how many bytes starting at C<s> comprise the code 844 point's representation. 845 846 =cut 847 */ 848 #define UTF8_IS_SURROGATE(s, e) is_SURROGATE_utf8_safe(s, e) 849 850 851 #define UTF8_IS_REPLACEMENT(s, send) is_REPLACEMENT_utf8_safe(s,send) 852 853 #define MAX_LEGAL_CP ((UV)IV_MAX) 854 855 /* 856 =for apidoc Am|bool|UTF8_IS_SUPER|const U8 *s|const U8 *e 857 858 Recall that Perl recognizes an extension to UTF-8 that can encode code 859 points larger than the ones defined by Unicode, which are 0..0x10FFFF. 860 861 This macro evaluates to non-zero if the first few bytes of the string starting 862 at C<s> and looking no further than S<C<e - 1>> are from this UTF-8 extension; 863 otherwise it evaluates to 0. If non-zero, the value gives how many bytes 864 starting at C<s> comprise the code point's representation. 865 866 0 is returned if the bytes are not well-formed extended UTF-8, or if they 867 represent a code point that cannot fit in a UV on the current platform. Hence 868 this macro can give different results when run on a 64-bit word machine than on 869 one with a 32-bit word size. 870 871 Note that it is illegal to have code points that are larger than what can 872 fit in an IV on the current machine. 873 874 =cut 875 876 * ASCII EBCDIC I8 877 * U+10FFFF: \xF4\x8F\xBF\xBF \xF9\xA1\xBF\xBF\xBF max legal Unicode 878 * U+110000: \xF4\x90\x80\x80 \xF9\xA2\xA0\xA0\xA0 879 * U+110001: \xF4\x90\x80\x81 \xF9\xA2\xA0\xA0\xA1 880 */ 881 #ifdef EBCDIC 882 # define UTF8_IS_SUPER(s, e) \ 883 (( LIKELY((e) > (s) + 4) \ 884 && NATIVE_UTF8_TO_I8(*(s)) >= 0xF9 \ 885 && ( NATIVE_UTF8_TO_I8(*(s)) > 0xF9 \ 886 || (NATIVE_UTF8_TO_I8(*((s) + 1)) >= 0xA2)) \ 887 && LIKELY((s) + UTF8SKIP(s) <= (e))) \ 888 ? is_utf8_char_helper(s, s + UTF8SKIP(s), 0) : 0) 889 #else 890 # define UTF8_IS_SUPER(s, e) \ 891 (( LIKELY((e) > (s) + 3) \ 892 && (*(U8*) (s)) >= 0xF4 \ 893 && ((*(U8*) (s)) > 0xF4 || (*((U8*) (s) + 1) >= 0x90))\ 894 && LIKELY((s) + UTF8SKIP(s) <= (e))) \ 895 ? is_utf8_char_helper(s, s + UTF8SKIP(s), 0) : 0) 896 #endif 897 898 /* These are now machine generated, and the 'given' clause is no longer 899 * applicable */ 900 #define UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s, e) \ 901 cBOOL(is_NONCHAR_utf8_safe(s,e)) 902 903 /* 904 =for apidoc Am|bool|UTF8_IS_NONCHAR|const U8 *s|const U8 *e 905 906 Evaluates to non-zero if the first few bytes of the string starting at C<s> and 907 looking no further than S<C<e - 1>> are well-formed UTF-8 that represents one 908 of the Unicode non-character code points; otherwise it evaluates to 0. If 909 non-zero, the value gives how many bytes starting at C<s> comprise the code 910 point's representation. 911 912 =for apidoc AmnU|UV|UNICODE_REPLACEMENT 913 914 Evaluates to 0xFFFD, the code point of the Unicode REPLACEMENT CHARACTER 915 916 =cut 917 */ 918 #define UTF8_IS_NONCHAR(s, e) \ 919 UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s, e) 920 921 #define UNICODE_SURROGATE_FIRST 0xD800 922 #define UNICODE_SURROGATE_LAST 0xDFFF 923 #define UNICODE_REPLACEMENT 0xFFFD 924 #define UNICODE_BYTE_ORDER_MARK 0xFEFF 925 926 /* Though our UTF-8 encoding can go beyond this, 927 * let's be conservative and do as Unicode says. */ 928 #define PERL_UNICODE_MAX 0x10FFFF 929 930 #define UNICODE_WARN_SURROGATE 0x0001 /* UTF-16 surrogates */ 931 #define UNICODE_WARN_NONCHAR 0x0002 /* Non-char code points */ 932 #define UNICODE_WARN_SUPER 0x0004 /* Above 0x10FFFF */ 933 #define UNICODE_WARN_PERL_EXTENDED 0x0008 /* Above 0x7FFF_FFFF */ 934 #define UNICODE_WARN_ABOVE_31_BIT UNICODE_WARN_PERL_EXTENDED 935 #define UNICODE_DISALLOW_SURROGATE 0x0010 936 #define UNICODE_DISALLOW_NONCHAR 0x0020 937 #define UNICODE_DISALLOW_SUPER 0x0040 938 #define UNICODE_DISALLOW_PERL_EXTENDED 0x0080 939 940 #ifdef PERL_CORE 941 # define UNICODE_ALLOW_ABOVE_IV_MAX 0x0100 942 #endif 943 #define UNICODE_DISALLOW_ABOVE_31_BIT UNICODE_DISALLOW_PERL_EXTENDED 944 945 #define UNICODE_GOT_SURROGATE UNICODE_DISALLOW_SURROGATE 946 #define UNICODE_GOT_NONCHAR UNICODE_DISALLOW_NONCHAR 947 #define UNICODE_GOT_SUPER UNICODE_DISALLOW_SUPER 948 #define UNICODE_GOT_PERL_EXTENDED UNICODE_DISALLOW_PERL_EXTENDED 949 950 #define UNICODE_WARN_ILLEGAL_C9_INTERCHANGE \ 951 (UNICODE_WARN_SURROGATE|UNICODE_WARN_SUPER) 952 #define UNICODE_WARN_ILLEGAL_INTERCHANGE \ 953 (UNICODE_WARN_ILLEGAL_C9_INTERCHANGE|UNICODE_WARN_NONCHAR) 954 #define UNICODE_DISALLOW_ILLEGAL_C9_INTERCHANGE \ 955 (UNICODE_DISALLOW_SURROGATE|UNICODE_DISALLOW_SUPER) 956 #define UNICODE_DISALLOW_ILLEGAL_INTERCHANGE \ 957 (UNICODE_DISALLOW_ILLEGAL_C9_INTERCHANGE|UNICODE_DISALLOW_NONCHAR) 958 959 /* For backward source compatibility, as are now the default */ 960 #define UNICODE_ALLOW_SURROGATE 0 961 #define UNICODE_ALLOW_SUPER 0 962 #define UNICODE_ALLOW_ANY 0 963 964 /* This matches the 2048 code points between UNICODE_SURROGATE_FIRST (0xD800) and 965 * UNICODE_SURROGATE_LAST (0xDFFF) */ 966 #define UNICODE_IS_SURROGATE(uv) (((UV) (uv) & (~0xFFFF | 0xF800)) \ 967 == 0xD800) 968 969 #define UNICODE_IS_REPLACEMENT(uv) ((UV) (uv) == UNICODE_REPLACEMENT) 970 #define UNICODE_IS_BYTE_ORDER_MARK(uv) ((UV) (uv) == UNICODE_BYTE_ORDER_MARK) 971 972 /* Is 'uv' one of the 32 contiguous-range noncharacters? */ 973 #define UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv) ((UV) (uv) >= 0xFDD0 \ 974 && (UV) (uv) <= 0xFDEF) 975 976 /* Is 'uv' one of the 34 plane-ending noncharacters 0xFFFE, 0xFFFF, 0x1FFFE, 977 * 0x1FFFF, ... 0x10FFFE, 0x10FFFF, given that we know that 'uv' is not above 978 * the Unicode legal max */ 979 #define UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv) \ 980 (((UV) (uv) & 0xFFFE) == 0xFFFE) 981 982 #define UNICODE_IS_NONCHAR(uv) \ 983 ( UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv) \ 984 || ( LIKELY( ! UNICODE_IS_SUPER(uv)) \ 985 && UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv))) 986 987 #define UNICODE_IS_SUPER(uv) ((UV) (uv) > PERL_UNICODE_MAX) 988 989 #define LATIN_SMALL_LETTER_SHARP_S LATIN_SMALL_LETTER_SHARP_S_NATIVE 990 #define LATIN_SMALL_LETTER_Y_WITH_DIAERESIS \ 991 LATIN_SMALL_LETTER_Y_WITH_DIAERESIS_NATIVE 992 #define MICRO_SIGN MICRO_SIGN_NATIVE 993 #define LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE \ 994 LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE_NATIVE 995 #define LATIN_SMALL_LETTER_A_WITH_RING_ABOVE \ 996 LATIN_SMALL_LETTER_A_WITH_RING_ABOVE_NATIVE 997 #define UNICODE_GREEK_CAPITAL_LETTER_SIGMA 0x03A3 998 #define UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA 0x03C2 999 #define UNICODE_GREEK_SMALL_LETTER_SIGMA 0x03C3 1000 #define GREEK_SMALL_LETTER_MU 0x03BC 1001 #define GREEK_CAPITAL_LETTER_MU 0x039C /* Upper and title case 1002 of MICRON */ 1003 #define LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS 0x0178 /* Also is title case */ 1004 #ifdef LATIN_CAPITAL_LETTER_SHARP_S_UTF8 1005 # define LATIN_CAPITAL_LETTER_SHARP_S 0x1E9E 1006 #endif 1007 #define LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE 0x130 1008 #define LATIN_SMALL_LETTER_DOTLESS_I 0x131 1009 #define LATIN_SMALL_LETTER_LONG_S 0x017F 1010 #define LATIN_SMALL_LIGATURE_LONG_S_T 0xFB05 1011 #define LATIN_SMALL_LIGATURE_ST 0xFB06 1012 #define KELVIN_SIGN 0x212A 1013 #define ANGSTROM_SIGN 0x212B 1014 1015 #define UNI_DISPLAY_ISPRINT 0x0001 1016 #define UNI_DISPLAY_BACKSLASH 0x0002 1017 #define UNI_DISPLAY_BACKSPACE 0x0004 /* Allow \b when also 1018 UNI_DISPLAY_BACKSLASH */ 1019 #define UNI_DISPLAY_QQ (UNI_DISPLAY_ISPRINT \ 1020 |UNI_DISPLAY_BACKSLASH \ 1021 |UNI_DISPLAY_BACKSPACE) 1022 1023 /* Character classes could also allow \b, but not patterns in general */ 1024 #define UNI_DISPLAY_REGEX (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH) 1025 1026 #define ANYOF_FOLD_SHARP_S(node, input, end) \ 1027 (ANYOF_BITMAP_TEST(node, LATIN_SMALL_LETTER_SHARP_S) && \ 1028 (ANYOF_NONBITMAP(node)) && \ 1029 (ANYOF_FLAGS(node) & ANYOF_LOC_NONBITMAP_FOLD) && \ 1030 ((end) > (input) + 1) && \ 1031 isALPHA_FOLD_EQ((input)[0], 's')) 1032 1033 #define SHARP_S_SKIP 2 1034 1035 #define is_utf8_char_buf(buf, buf_end) isUTF8_CHAR(buf, buf_end) 1036 #define bytes_from_utf8(s, lenp, is_utf8p) \ 1037 bytes_from_utf8_loc(s, lenp, is_utf8p, 0) 1038 1039 /* 1040 1041 =for apidoc Am|STRLEN|isUTF8_CHAR_flags|const U8 *s|const U8 *e| const U32 flags 1042 1043 Evaluates to non-zero if the first few bytes of the string starting at C<s> and 1044 looking no further than S<C<e - 1>> are well-formed UTF-8, as extended by Perl, 1045 that represents some code point, subject to the restrictions given by C<flags>; 1046 otherwise it evaluates to 0. If non-zero, the value gives how many bytes 1047 starting at C<s> comprise the code point's representation. Any bytes remaining 1048 before C<e>, but beyond the ones needed to form the first code point in C<s>, 1049 are not examined. 1050 1051 If C<flags> is 0, this gives the same results as C<L</isUTF8_CHAR>>; 1052 if C<flags> is C<UTF8_DISALLOW_ILLEGAL_INTERCHANGE>, this gives the same results 1053 as C<L</isSTRICT_UTF8_CHAR>>; 1054 and if C<flags> is C<UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE>, this gives 1055 the same results as C<L</isC9_STRICT_UTF8_CHAR>>. 1056 Otherwise C<flags> may be any combination of the C<UTF8_DISALLOW_I<foo>> flags 1057 understood by C<L</utf8n_to_uvchr>>, with the same meanings. 1058 1059 The three alternative macros are for the most commonly needed validations; they 1060 are likely to run somewhat faster than this more general one, as they can be 1061 inlined into your code. 1062 1063 Use L</is_utf8_string_flags>, L</is_utf8_string_loc_flags>, and 1064 L</is_utf8_string_loclen_flags> to check entire strings. 1065 1066 =cut 1067 */ 1068 1069 #define isUTF8_CHAR_flags(s, e, flags) \ 1070 (UNLIKELY((e) <= (s)) \ 1071 ? 0 \ 1072 : (UTF8_IS_INVARIANT(*s)) \ 1073 ? 1 \ 1074 : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \ 1075 ? 0 \ 1076 : is_utf8_char_helper(s, e, flags)) 1077 1078 /* Do not use; should be deprecated. Use isUTF8_CHAR() instead; this is 1079 * retained solely for backwards compatibility */ 1080 #define IS_UTF8_CHAR(p, n) (isUTF8_CHAR(p, (p) + (n)) == n) 1081 1082 #endif /* PERL_UTF8_H_ */ 1083 1084 /* 1085 * ex: set ts=8 sts=4 sw=4 et: 1086 */ 1087