| 1 | /* utf8.c |
| 2 | * |
| 3 | * Copyright (C) 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008 |
| 4 | * by Larry Wall and others |
| 5 | * |
| 6 | * You may distribute under the terms of either the GNU General Public |
| 7 | * License or the Artistic License, as specified in the README file. |
| 8 | * |
| 9 | */ |
| 10 | |
| 11 | /* |
| 12 | * 'What a fix!' said Sam. 'That's the one place in all the lands we've ever |
| 13 | * heard of that we don't want to see any closer; and that's the one place |
| 14 | * we're trying to get to! And that's just where we can't get, nohow.' |
| 15 | * |
| 16 | * [p.603 of _The Lord of the Rings_, IV/I: "The Taming of Sméagol"] |
| 17 | * |
| 18 | * 'Well do I understand your speech,' he answered in the same language; |
| 19 | * 'yet few strangers do so. Why then do you not speak in the Common Tongue, |
| 20 | * as is the custom in the West, if you wish to be answered?' |
| 21 | * --Gandalf, addressing Théoden's door wardens |
| 22 | * |
| 23 | * [p.508 of _The Lord of the Rings_, III/vi: "The King of the Golden Hall"] |
| 24 | * |
| 25 | * ...the travellers perceived that the floor was paved with stones of many |
| 26 | * hues; branching runes and strange devices intertwined beneath their feet. |
| 27 | * |
| 28 | * [p.512 of _The Lord of the Rings_, III/vi: "The King of the Golden Hall"] |
| 29 | */ |
| 30 | |
| 31 | #include "EXTERN.h" |
| 32 | #define PERL_IN_UTF8_C |
| 33 | #include "perl.h" |
| 34 | #include "inline_invlist.c" |
| 35 | |
| 36 | #ifndef EBCDIC |
| 37 | /* Separate prototypes needed because in ASCII systems these are |
| 38 | * usually macros but they still are compiled as code, too. */ |
| 39 | PERL_CALLCONV UV Perl_utf8n_to_uvchr(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen, U32 flags); |
| 40 | PERL_CALLCONV UV Perl_valid_utf8_to_uvchr(pTHX_ const U8 *s, STRLEN *retlen); |
| 41 | PERL_CALLCONV U8* Perl_uvchr_to_utf8(pTHX_ U8 *d, UV uv); |
| 42 | #endif |
| 43 | |
| 44 | static const char unees[] = |
| 45 | "Malformed UTF-8 character (unexpected end of string)"; |
| 46 | |
| 47 | /* |
| 48 | =head1 Unicode Support |
| 49 | |
| 50 | This file contains various utility functions for manipulating UTF8-encoded |
| 51 | strings. For the uninitiated, this is a method of representing arbitrary |
| 52 | Unicode characters as a variable number of bytes, in such a way that |
| 53 | characters in the ASCII range are unmodified, and a zero byte never appears |
| 54 | within non-zero characters. |
| 55 | |
| 56 | =cut |
| 57 | */ |
| 58 | |
| 59 | /* |
| 60 | =for apidoc is_ascii_string |
| 61 | |
| 62 | Returns true if the first C<len> bytes of the string C<s> are the same whether |
| 63 | or not the string is encoded in UTF-8 (or UTF-EBCDIC on EBCDIC machines). That |
| 64 | is, if they are invariant. On ASCII-ish machines, only ASCII characters |
| 65 | fit this definition, hence the function's name. |
| 66 | |
| 67 | If C<len> is 0, it will be calculated using C<strlen(s)>. |
| 68 | |
| 69 | See also L</is_utf8_string>(), L</is_utf8_string_loclen>(), and L</is_utf8_string_loc>(). |
| 70 | |
| 71 | =cut |
| 72 | */ |
| 73 | |
| 74 | bool |
| 75 | Perl_is_ascii_string(const U8 *s, STRLEN len) |
| 76 | { |
| 77 | const U8* const send = s + (len ? len : strlen((const char *)s)); |
| 78 | const U8* x = s; |
| 79 | |
| 80 | PERL_ARGS_ASSERT_IS_ASCII_STRING; |
| 81 | |
| 82 | for (; x < send; ++x) { |
| 83 | if (!UTF8_IS_INVARIANT(*x)) |
| 84 | break; |
| 85 | } |
| 86 | |
| 87 | return x == send; |
| 88 | } |
| 89 | |
| 90 | /* |
| 91 | =for apidoc uvuni_to_utf8_flags |
| 92 | |
| 93 | Adds the UTF-8 representation of the code point C<uv> to the end |
| 94 | of the string C<d>; C<d> should have at least C<UTF8_MAXBYTES+1> free |
| 95 | bytes available. The return value is the pointer to the byte after the |
| 96 | end of the new character. In other words, |
| 97 | |
| 98 | d = uvuni_to_utf8_flags(d, uv, flags); |
| 99 | |
| 100 | or, in most cases, |
| 101 | |
| 102 | d = uvuni_to_utf8(d, uv); |
| 103 | |
| 104 | (which is equivalent to) |
| 105 | |
| 106 | d = uvuni_to_utf8_flags(d, uv, 0); |
| 107 | |
| 108 | This is the recommended Unicode-aware way of saying |
| 109 | |
| 110 | *(d++) = uv; |
| 111 | |
| 112 | This function will convert to UTF-8 (and not warn) even code points that aren't |
| 113 | legal Unicode or are problematic, unless C<flags> contains one or more of the |
| 114 | following flags: |
| 115 | |
| 116 | If C<uv> is a Unicode surrogate code point and UNICODE_WARN_SURROGATE is set, |
| 117 | the function will raise a warning, provided UTF8 warnings are enabled. If instead |
| 118 | UNICODE_DISALLOW_SURROGATE is set, the function will fail and return NULL. |
| 119 | If both flags are set, the function will both warn and return NULL. |
| 120 | |
| 121 | The UNICODE_WARN_NONCHAR and UNICODE_DISALLOW_NONCHAR flags correspondingly |
| 122 | affect how the function handles a Unicode non-character. And, likewise for the |
| 123 | UNICODE_WARN_SUPER and UNICODE_DISALLOW_SUPER flags, and code points that are |
| 124 | above the Unicode maximum of 0x10FFFF. Code points above 0x7FFF_FFFF (which are |
| 125 | even less portable) can be warned and/or disallowed even if other above-Unicode |
| 126 | code points are accepted by the UNICODE_WARN_FE_FF and UNICODE_DISALLOW_FE_FF |
| 127 | flags. |
| 128 | |
| 129 | And finally, the flag UNICODE_WARN_ILLEGAL_INTERCHANGE selects all four of the |
| 130 | above WARN flags; and UNICODE_DISALLOW_ILLEGAL_INTERCHANGE selects all four |
| 131 | DISALLOW flags. |
| 132 | |
| 133 | |
| 134 | =cut |
| 135 | */ |
| 136 | |
| 137 | U8 * |
| 138 | Perl_uvuni_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags) |
| 139 | { |
| 140 | PERL_ARGS_ASSERT_UVUNI_TO_UTF8_FLAGS; |
| 141 | |
| 142 | /* The first problematic code point is the first surrogate */ |
| 143 | if (uv >= UNICODE_SURROGATE_FIRST |
| 144 | && ckWARN4_d(WARN_UTF8, WARN_SURROGATE, WARN_NON_UNICODE, WARN_NONCHAR)) |
| 145 | { |
| 146 | if (UNICODE_IS_SURROGATE(uv)) { |
| 147 | if (flags & UNICODE_WARN_SURROGATE) { |
| 148 | Perl_ck_warner_d(aTHX_ packWARN(WARN_SURROGATE), |
| 149 | "UTF-16 surrogate U+%04"UVXf, uv); |
| 150 | } |
| 151 | if (flags & UNICODE_DISALLOW_SURROGATE) { |
| 152 | return NULL; |
| 153 | } |
| 154 | } |
| 155 | else if (UNICODE_IS_SUPER(uv)) { |
| 156 | if (flags & UNICODE_WARN_SUPER |
| 157 | || (UNICODE_IS_FE_FF(uv) && (flags & UNICODE_WARN_FE_FF))) |
| 158 | { |
| 159 | Perl_ck_warner_d(aTHX_ packWARN(WARN_NON_UNICODE), |
| 160 | "Code point 0x%04"UVXf" is not Unicode, may not be portable", uv); |
| 161 | } |
| 162 | if (flags & UNICODE_DISALLOW_SUPER |
| 163 | || (UNICODE_IS_FE_FF(uv) && (flags & UNICODE_DISALLOW_FE_FF))) |
| 164 | { |
| 165 | return NULL; |
| 166 | } |
| 167 | } |
| 168 | else if (UNICODE_IS_NONCHAR(uv)) { |
| 169 | if (flags & UNICODE_WARN_NONCHAR) { |
| 170 | Perl_ck_warner_d(aTHX_ packWARN(WARN_NONCHAR), |
| 171 | "Unicode non-character U+%04"UVXf" is illegal for open interchange", |
| 172 | uv); |
| 173 | } |
| 174 | if (flags & UNICODE_DISALLOW_NONCHAR) { |
| 175 | return NULL; |
| 176 | } |
| 177 | } |
| 178 | } |
| 179 | if (UNI_IS_INVARIANT(uv)) { |
| 180 | *d++ = (U8)UTF_TO_NATIVE(uv); |
| 181 | return d; |
| 182 | } |
| 183 | #if defined(EBCDIC) |
| 184 | else { |
| 185 | STRLEN len = UNISKIP(uv); |
| 186 | U8 *p = d+len-1; |
| 187 | while (p > d) { |
| 188 | *p-- = (U8)UTF_TO_NATIVE((uv & UTF_CONTINUATION_MASK) | UTF_CONTINUATION_MARK); |
| 189 | uv >>= UTF_ACCUMULATION_SHIFT; |
| 190 | } |
| 191 | *p = (U8)UTF_TO_NATIVE((uv & UTF_START_MASK(len)) | UTF_START_MARK(len)); |
| 192 | return d+len; |
| 193 | } |
| 194 | #else /* Non loop style */ |
| 195 | if (uv < 0x800) { |
| 196 | *d++ = (U8)(( uv >> 6) | 0xc0); |
| 197 | *d++ = (U8)(( uv & 0x3f) | 0x80); |
| 198 | return d; |
| 199 | } |
| 200 | if (uv < 0x10000) { |
| 201 | *d++ = (U8)(( uv >> 12) | 0xe0); |
| 202 | *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80); |
| 203 | *d++ = (U8)(( uv & 0x3f) | 0x80); |
| 204 | return d; |
| 205 | } |
| 206 | if (uv < 0x200000) { |
| 207 | *d++ = (U8)(( uv >> 18) | 0xf0); |
| 208 | *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80); |
| 209 | *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80); |
| 210 | *d++ = (U8)(( uv & 0x3f) | 0x80); |
| 211 | return d; |
| 212 | } |
| 213 | if (uv < 0x4000000) { |
| 214 | *d++ = (U8)(( uv >> 24) | 0xf8); |
| 215 | *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80); |
| 216 | *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80); |
| 217 | *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80); |
| 218 | *d++ = (U8)(( uv & 0x3f) | 0x80); |
| 219 | return d; |
| 220 | } |
| 221 | if (uv < 0x80000000) { |
| 222 | *d++ = (U8)(( uv >> 30) | 0xfc); |
| 223 | *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80); |
| 224 | *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80); |
| 225 | *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80); |
| 226 | *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80); |
| 227 | *d++ = (U8)(( uv & 0x3f) | 0x80); |
| 228 | return d; |
| 229 | } |
| 230 | #ifdef HAS_QUAD |
| 231 | if (uv < UTF8_QUAD_MAX) |
| 232 | #endif |
| 233 | { |
| 234 | *d++ = 0xfe; /* Can't match U+FEFF! */ |
| 235 | *d++ = (U8)(((uv >> 30) & 0x3f) | 0x80); |
| 236 | *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80); |
| 237 | *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80); |
| 238 | *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80); |
| 239 | *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80); |
| 240 | *d++ = (U8)(( uv & 0x3f) | 0x80); |
| 241 | return d; |
| 242 | } |
| 243 | #ifdef HAS_QUAD |
| 244 | { |
| 245 | *d++ = 0xff; /* Can't match U+FFFE! */ |
| 246 | *d++ = 0x80; /* 6 Reserved bits */ |
| 247 | *d++ = (U8)(((uv >> 60) & 0x0f) | 0x80); /* 2 Reserved bits */ |
| 248 | *d++ = (U8)(((uv >> 54) & 0x3f) | 0x80); |
| 249 | *d++ = (U8)(((uv >> 48) & 0x3f) | 0x80); |
| 250 | *d++ = (U8)(((uv >> 42) & 0x3f) | 0x80); |
| 251 | *d++ = (U8)(((uv >> 36) & 0x3f) | 0x80); |
| 252 | *d++ = (U8)(((uv >> 30) & 0x3f) | 0x80); |
| 253 | *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80); |
| 254 | *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80); |
| 255 | *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80); |
| 256 | *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80); |
| 257 | *d++ = (U8)(( uv & 0x3f) | 0x80); |
| 258 | return d; |
| 259 | } |
| 260 | #endif |
| 261 | #endif /* Loop style */ |
| 262 | } |
| 263 | |
| 264 | /* |
| 265 | |
| 266 | Tests if the first C<len> bytes of string C<s> form a valid UTF-8 |
| 267 | character. Note that an INVARIANT (i.e. ASCII) character is a valid |
| 268 | UTF-8 character. The number of bytes in the UTF-8 character |
| 269 | will be returned if it is valid, otherwise 0. |
| 270 | |
| 271 | This is the "slow" version as opposed to the "fast" version which is |
| 272 | the "unrolled" IS_UTF8_CHAR(). E.g. for t/uni/class.t the speed |
| 273 | difference is a factor of 2 to 3. For lengths (UTF8SKIP(s)) of four |
| 274 | or less you should use the IS_UTF8_CHAR(), for lengths of five or more |
| 275 | you should use the _slow(). In practice this means that the _slow() |
| 276 | will be used very rarely, since the maximum Unicode code point (as of |
| 277 | Unicode 4.1) is U+10FFFF, which encodes in UTF-8 to four bytes. Only |
| 278 | the "Perl extended UTF-8" (the infamous 'v-strings') will encode into |
| 279 | five bytes or more. |
| 280 | |
| 281 | =cut */ |
| 282 | STATIC STRLEN |
| 283 | S_is_utf8_char_slow(const U8 *s, const STRLEN len) |
| 284 | { |
| 285 | dTHX; /* The function called below requires thread context */ |
| 286 | |
| 287 | STRLEN actual_len; |
| 288 | |
| 289 | PERL_ARGS_ASSERT_IS_UTF8_CHAR_SLOW; |
| 290 | |
| 291 | utf8n_to_uvuni(s, len, &actual_len, UTF8_CHECK_ONLY); |
| 292 | |
| 293 | return (actual_len == (STRLEN) -1) ? 0 : actual_len; |
| 294 | } |
| 295 | |
| 296 | /* |
| 297 | =for apidoc is_utf8_char_buf |
| 298 | |
| 299 | Returns the number of bytes that comprise the first UTF-8 encoded character in |
| 300 | buffer C<buf>. C<buf_end> should point to one position beyond the end of the |
| 301 | buffer. 0 is returned if C<buf> does not point to a complete, valid UTF-8 |
| 302 | encoded character. |
| 303 | |
| 304 | Note that an INVARIANT character (i.e. ASCII on non-EBCDIC |
| 305 | machines) is a valid UTF-8 character. |
| 306 | |
| 307 | =cut */ |
| 308 | |
| 309 | STRLEN |
| 310 | Perl_is_utf8_char_buf(const U8 *buf, const U8* buf_end) |
| 311 | { |
| 312 | |
| 313 | STRLEN len; |
| 314 | |
| 315 | PERL_ARGS_ASSERT_IS_UTF8_CHAR_BUF; |
| 316 | |
| 317 | if (buf_end <= buf) { |
| 318 | return 0; |
| 319 | } |
| 320 | |
| 321 | len = buf_end - buf; |
| 322 | if (len > UTF8SKIP(buf)) { |
| 323 | len = UTF8SKIP(buf); |
| 324 | } |
| 325 | |
| 326 | #ifdef IS_UTF8_CHAR |
| 327 | if (IS_UTF8_CHAR_FAST(len)) |
| 328 | return IS_UTF8_CHAR(buf, len) ? len : 0; |
| 329 | #endif /* #ifdef IS_UTF8_CHAR */ |
| 330 | return is_utf8_char_slow(buf, len); |
| 331 | } |
| 332 | |
| 333 | /* |
| 334 | =for apidoc is_utf8_char |
| 335 | |
| 336 | DEPRECATED! |
| 337 | |
| 338 | Tests if some arbitrary number of bytes begins in a valid UTF-8 |
| 339 | character. Note that an INVARIANT (i.e. ASCII on non-EBCDIC machines) |
| 340 | character is a valid UTF-8 character. The actual number of bytes in the UTF-8 |
| 341 | character will be returned if it is valid, otherwise 0. |
| 342 | |
| 343 | This function is deprecated due to the possibility that malformed input could |
| 344 | cause reading beyond the end of the input buffer. Use L</is_utf8_char_buf> |
| 345 | instead. |
| 346 | |
| 347 | =cut */ |
| 348 | |
| 349 | STRLEN |
| 350 | Perl_is_utf8_char(const U8 *s) |
| 351 | { |
| 352 | PERL_ARGS_ASSERT_IS_UTF8_CHAR; |
| 353 | |
| 354 | /* Assumes we have enough space, which is why this is deprecated */ |
| 355 | return is_utf8_char_buf(s, s + UTF8SKIP(s)); |
| 356 | } |
| 357 | |
| 358 | |
| 359 | /* |
| 360 | =for apidoc is_utf8_string |
| 361 | |
| 362 | Returns true if the first C<len> bytes of string C<s> form a valid |
| 363 | UTF-8 string, false otherwise. If C<len> is 0, it will be calculated |
| 364 | using C<strlen(s)> (which means if you use this option, that C<s> has to have a |
| 365 | terminating NUL byte). Note that all characters being ASCII constitute 'a |
| 366 | valid UTF-8 string'. |
| 367 | |
| 368 | See also L</is_ascii_string>(), L</is_utf8_string_loclen>(), and L</is_utf8_string_loc>(). |
| 369 | |
| 370 | =cut |
| 371 | */ |
| 372 | |
| 373 | bool |
| 374 | Perl_is_utf8_string(const U8 *s, STRLEN len) |
| 375 | { |
| 376 | const U8* const send = s + (len ? len : strlen((const char *)s)); |
| 377 | const U8* x = s; |
| 378 | |
| 379 | PERL_ARGS_ASSERT_IS_UTF8_STRING; |
| 380 | |
| 381 | while (x < send) { |
| 382 | /* Inline the easy bits of is_utf8_char() here for speed... */ |
| 383 | if (UTF8_IS_INVARIANT(*x)) { |
| 384 | x++; |
| 385 | } |
| 386 | else if (!UTF8_IS_START(*x)) |
| 387 | return FALSE; |
| 388 | else { |
| 389 | /* ... and call is_utf8_char() only if really needed. */ |
| 390 | const STRLEN c = UTF8SKIP(x); |
| 391 | const U8* const next_char_ptr = x + c; |
| 392 | |
| 393 | if (next_char_ptr > send) { |
| 394 | return FALSE; |
| 395 | } |
| 396 | |
| 397 | if (IS_UTF8_CHAR_FAST(c)) { |
| 398 | if (!IS_UTF8_CHAR(x, c)) |
| 399 | return FALSE; |
| 400 | } |
| 401 | else if (! is_utf8_char_slow(x, c)) { |
| 402 | return FALSE; |
| 403 | } |
| 404 | x = next_char_ptr; |
| 405 | } |
| 406 | } |
| 407 | |
| 408 | return TRUE; |
| 409 | } |
| 410 | |
| 411 | /* |
| 412 | Implemented as a macro in utf8.h |
| 413 | |
| 414 | =for apidoc is_utf8_string_loc |
| 415 | |
| 416 | Like L</is_utf8_string> but stores the location of the failure (in the |
| 417 | case of "utf8ness failure") or the location C<s>+C<len> (in the case of |
| 418 | "utf8ness success") in the C<ep>. |
| 419 | |
| 420 | See also L</is_utf8_string_loclen>() and L</is_utf8_string>(). |
| 421 | |
| 422 | =for apidoc is_utf8_string_loclen |
| 423 | |
| 424 | Like L</is_utf8_string>() but stores the location of the failure (in the |
| 425 | case of "utf8ness failure") or the location C<s>+C<len> (in the case of |
| 426 | "utf8ness success") in the C<ep>, and the number of UTF-8 |
| 427 | encoded characters in the C<el>. |
| 428 | |
| 429 | See also L</is_utf8_string_loc>() and L</is_utf8_string>(). |
| 430 | |
| 431 | =cut |
| 432 | */ |
| 433 | |
| 434 | bool |
| 435 | Perl_is_utf8_string_loclen(const U8 *s, STRLEN len, const U8 **ep, STRLEN *el) |
| 436 | { |
| 437 | const U8* const send = s + (len ? len : strlen((const char *)s)); |
| 438 | const U8* x = s; |
| 439 | STRLEN c; |
| 440 | STRLEN outlen = 0; |
| 441 | |
| 442 | PERL_ARGS_ASSERT_IS_UTF8_STRING_LOCLEN; |
| 443 | |
| 444 | while (x < send) { |
| 445 | const U8* next_char_ptr; |
| 446 | |
| 447 | /* Inline the easy bits of is_utf8_char() here for speed... */ |
| 448 | if (UTF8_IS_INVARIANT(*x)) |
| 449 | next_char_ptr = x + 1; |
| 450 | else if (!UTF8_IS_START(*x)) |
| 451 | goto out; |
| 452 | else { |
| 453 | /* ... and call is_utf8_char() only if really needed. */ |
| 454 | c = UTF8SKIP(x); |
| 455 | next_char_ptr = c + x; |
| 456 | if (next_char_ptr > send) { |
| 457 | goto out; |
| 458 | } |
| 459 | if (IS_UTF8_CHAR_FAST(c)) { |
| 460 | if (!IS_UTF8_CHAR(x, c)) |
| 461 | c = 0; |
| 462 | } else |
| 463 | c = is_utf8_char_slow(x, c); |
| 464 | if (!c) |
| 465 | goto out; |
| 466 | } |
| 467 | x = next_char_ptr; |
| 468 | outlen++; |
| 469 | } |
| 470 | |
| 471 | out: |
| 472 | if (el) |
| 473 | *el = outlen; |
| 474 | |
| 475 | if (ep) |
| 476 | *ep = x; |
| 477 | return (x == send); |
| 478 | } |
| 479 | |
| 480 | /* |
| 481 | |
| 482 | =for apidoc utf8n_to_uvuni |
| 483 | |
| 484 | Bottom level UTF-8 decode routine. |
| 485 | Returns the code point value of the first character in the string C<s>, |
| 486 | which is assumed to be in UTF-8 (or UTF-EBCDIC) encoding, and no longer than |
| 487 | C<curlen> bytes; C<*retlen> (if C<retlen> isn't NULL) will be set to |
| 488 | the length, in bytes, of that character. |
| 489 | |
| 490 | The value of C<flags> determines the behavior when C<s> does not point to a |
| 491 | well-formed UTF-8 character. If C<flags> is 0, when a malformation is found, |
| 492 | zero is returned and C<*retlen> is set so that (S<C<s> + C<*retlen>>) is the |
| 493 | next possible position in C<s> that could begin a non-malformed character. |
| 494 | Also, if UTF-8 warnings haven't been lexically disabled, a warning is raised. |
| 495 | |
| 496 | Various ALLOW flags can be set in C<flags> to allow (and not warn on) |
| 497 | individual types of malformations, such as the sequence being overlong (that |
| 498 | is, when there is a shorter sequence that can express the same code point; |
| 499 | overlong sequences are expressly forbidden in the UTF-8 standard due to |
| 500 | potential security issues). Another malformation example is the first byte of |
| 501 | a character not being a legal first byte. See F<utf8.h> for the list of such |
| 502 | flags. For allowed 0 length strings, this function returns 0; for allowed |
| 503 | overlong sequences, the computed code point is returned; for all other allowed |
| 504 | malformations, the Unicode REPLACEMENT CHARACTER is returned, as these have no |
| 505 | determinable reasonable value. |
| 506 | |
| 507 | The UTF8_CHECK_ONLY flag overrides the behavior when a non-allowed (by other |
| 508 | flags) malformation is found. If this flag is set, the routine assumes that |
| 509 | the caller will raise a warning, and this function will silently just set |
| 510 | C<retlen> to C<-1> and return zero. |
| 511 | |
| 512 | Certain code points are considered problematic. These are Unicode surrogates, |
| 513 | Unicode non-characters, and code points above the Unicode maximum of 0x10FFFF. |
| 514 | By default these are considered regular code points, but certain situations |
| 515 | warrant special handling for them. If C<flags> contains |
| 516 | UTF8_DISALLOW_ILLEGAL_INTERCHANGE, all three classes are treated as |
| 517 | malformations and handled as such. The flags UTF8_DISALLOW_SURROGATE, |
| 518 | UTF8_DISALLOW_NONCHAR, and UTF8_DISALLOW_SUPER (meaning above the legal Unicode |
| 519 | maximum) can be set to disallow these categories individually. |
| 520 | |
| 521 | The flags UTF8_WARN_ILLEGAL_INTERCHANGE, UTF8_WARN_SURROGATE, |
| 522 | UTF8_WARN_NONCHAR, and UTF8_WARN_SUPER will cause warning messages to be raised |
| 523 | for their respective categories, but otherwise the code points are considered |
| 524 | valid (not malformations). To get a category to both be treated as a |
| 525 | malformation and raise a warning, specify both the WARN and DISALLOW flags. |
| 526 | (But note that warnings are not raised if lexically disabled nor if |
| 527 | UTF8_CHECK_ONLY is also specified.) |
| 528 | |
| 529 | Very large code points (above 0x7FFF_FFFF) are considered more problematic than |
| 530 | the others that are above the Unicode legal maximum. There are several |
| 531 | reasons: they requre at least 32 bits to represent them on ASCII platforms, are |
| 532 | not representable at all on EBCDIC platforms, and the original UTF-8 |
| 533 | specification never went above this number (the current 0x10FFFF limit was |
| 534 | imposed later). (The smaller ones, those that fit into 32 bits, are |
| 535 | representable by a UV on ASCII platforms, but not by an IV, which means that |
| 536 | the number of operations that can be performed on them is quite restricted.) |
| 537 | The UTF-8 encoding on ASCII platforms for these large code points begins with a |
| 538 | byte containing 0xFE or 0xFF. The UTF8_DISALLOW_FE_FF flag will cause them to |
| 539 | be treated as malformations, while allowing smaller above-Unicode code points. |
| 540 | (Of course UTF8_DISALLOW_SUPER will treat all above-Unicode code points, |
| 541 | including these, as malformations.) Similarly, UTF8_WARN_FE_FF acts just like |
| 542 | the other WARN flags, but applies just to these code points. |
| 543 | |
| 544 | All other code points corresponding to Unicode characters, including private |
| 545 | use and those yet to be assigned, are never considered malformed and never |
| 546 | warn. |
| 547 | |
| 548 | Most code should use L</utf8_to_uvchr_buf>() rather than call this directly. |
| 549 | |
| 550 | =cut |
| 551 | */ |
| 552 | |
| 553 | UV |
| 554 | Perl_utf8n_to_uvuni(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen, U32 flags) |
| 555 | { |
| 556 | dVAR; |
| 557 | const U8 * const s0 = s; |
| 558 | U8 overflow_byte = '\0'; /* Save byte in case of overflow */ |
| 559 | U8 * send; |
| 560 | UV uv = *s; |
| 561 | STRLEN expectlen; |
| 562 | SV* sv = NULL; |
| 563 | UV outlier_ret = 0; /* return value when input is in error or problematic |
| 564 | */ |
| 565 | UV pack_warn = 0; /* Save result of packWARN() for later */ |
| 566 | bool unexpected_non_continuation = FALSE; |
| 567 | bool overflowed = FALSE; |
| 568 | bool do_overlong_test = TRUE; /* May have to skip this test */ |
| 569 | |
| 570 | const char* const malformed_text = "Malformed UTF-8 character"; |
| 571 | |
| 572 | PERL_ARGS_ASSERT_UTF8N_TO_UVUNI; |
| 573 | |
| 574 | /* The order of malformation tests here is important. We should consume as |
| 575 | * few bytes as possible in order to not skip any valid character. This is |
| 576 | * required by the Unicode Standard (section 3.9 of Unicode 6.0); see also |
| 577 | * http://unicode.org/reports/tr36 for more discussion as to why. For |
| 578 | * example, once we've done a UTF8SKIP, we can tell the expected number of |
| 579 | * bytes, and could fail right off the bat if the input parameters indicate |
| 580 | * that there are too few available. But it could be that just that first |
| 581 | * byte is garbled, and the intended character occupies fewer bytes. If we |
| 582 | * blindly assumed that the first byte is correct, and skipped based on |
| 583 | * that number, we could skip over a valid input character. So instead, we |
| 584 | * always examine the sequence byte-by-byte. |
| 585 | * |
| 586 | * We also should not consume too few bytes, otherwise someone could inject |
| 587 | * things. For example, an input could be deliberately designed to |
| 588 | * overflow, and if this code bailed out immediately upon discovering that, |
| 589 | * returning to the caller *retlen pointing to the very next byte (one |
| 590 | * which is actually part of of the overflowing sequence), that could look |
| 591 | * legitimate to the caller, which could discard the initial partial |
| 592 | * sequence and process the rest, inappropriately */ |
| 593 | |
| 594 | /* Zero length strings, if allowed, of necessity are zero */ |
| 595 | if (UNLIKELY(curlen == 0)) { |
| 596 | if (retlen) { |
| 597 | *retlen = 0; |
| 598 | } |
| 599 | |
| 600 | if (flags & UTF8_ALLOW_EMPTY) { |
| 601 | return 0; |
| 602 | } |
| 603 | if (! (flags & UTF8_CHECK_ONLY)) { |
| 604 | sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (empty string)", malformed_text)); |
| 605 | } |
| 606 | goto malformed; |
| 607 | } |
| 608 | |
| 609 | expectlen = UTF8SKIP(s); |
| 610 | |
| 611 | /* A well-formed UTF-8 character, as the vast majority of calls to this |
| 612 | * function will be for, has this expected length. For efficiency, set |
| 613 | * things up here to return it. It will be overriden only in those rare |
| 614 | * cases where a malformation is found */ |
| 615 | if (retlen) { |
| 616 | *retlen = expectlen; |
| 617 | } |
| 618 | |
| 619 | /* An invariant is trivially well-formed */ |
| 620 | if (UTF8_IS_INVARIANT(uv)) { |
| 621 | return (UV) (NATIVE_TO_UTF(*s)); |
| 622 | } |
| 623 | |
| 624 | /* A continuation character can't start a valid sequence */ |
| 625 | if (UNLIKELY(UTF8_IS_CONTINUATION(uv))) { |
| 626 | if (flags & UTF8_ALLOW_CONTINUATION) { |
| 627 | if (retlen) { |
| 628 | *retlen = 1; |
| 629 | } |
| 630 | return UNICODE_REPLACEMENT; |
| 631 | } |
| 632 | |
| 633 | if (! (flags & UTF8_CHECK_ONLY)) { |
| 634 | sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (unexpected continuation byte 0x%02x, with no preceding start byte)", malformed_text, *s0)); |
| 635 | } |
| 636 | curlen = 1; |
| 637 | goto malformed; |
| 638 | } |
| 639 | |
| 640 | #ifdef EBCDIC |
| 641 | uv = NATIVE_TO_UTF(uv); |
| 642 | #endif |
| 643 | |
| 644 | /* Here is not a continuation byte, nor an invariant. The only thing left |
| 645 | * is a start byte (possibly for an overlong) */ |
| 646 | |
| 647 | /* Remove the leading bits that indicate the number of bytes in the |
| 648 | * character's whole UTF-8 sequence, leaving just the bits that are part of |
| 649 | * the value */ |
| 650 | uv &= UTF_START_MASK(expectlen); |
| 651 | |
| 652 | /* Now, loop through the remaining bytes in the character's sequence, |
| 653 | * accumulating each into the working value as we go. Be sure to not look |
| 654 | * past the end of the input string */ |
| 655 | send = (U8*) s0 + ((expectlen <= curlen) ? expectlen : curlen); |
| 656 | |
| 657 | for (s = s0 + 1; s < send; s++) { |
| 658 | if (LIKELY(UTF8_IS_CONTINUATION(*s))) { |
| 659 | #ifndef EBCDIC /* Can't overflow in EBCDIC */ |
| 660 | if (uv & UTF_ACCUMULATION_OVERFLOW_MASK) { |
| 661 | |
| 662 | /* The original implementors viewed this malformation as more |
| 663 | * serious than the others (though I, khw, don't understand |
| 664 | * why, since other malformations also give very very wrong |
| 665 | * results), so there is no way to turn off checking for it. |
| 666 | * Set a flag, but keep going in the loop, so that we absorb |
| 667 | * the rest of the bytes that comprise the character. */ |
| 668 | overflowed = TRUE; |
| 669 | overflow_byte = *s; /* Save for warning message's use */ |
| 670 | } |
| 671 | #endif |
| 672 | uv = UTF8_ACCUMULATE(uv, *s); |
| 673 | } |
| 674 | else { |
| 675 | /* Here, found a non-continuation before processing all expected |
| 676 | * bytes. This byte begins a new character, so quit, even if |
| 677 | * allowing this malformation. */ |
| 678 | unexpected_non_continuation = TRUE; |
| 679 | break; |
| 680 | } |
| 681 | } /* End of loop through the character's bytes */ |
| 682 | |
| 683 | /* Save how many bytes were actually in the character */ |
| 684 | curlen = s - s0; |
| 685 | |
| 686 | /* The loop above finds two types of malformations: non-continuation and/or |
| 687 | * overflow. The non-continuation malformation is really a too-short |
| 688 | * malformation, as it means that the current character ended before it was |
| 689 | * expected to (being terminated prematurely by the beginning of the next |
| 690 | * character, whereas in the too-short malformation there just are too few |
| 691 | * bytes available to hold the character. In both cases, the check below |
| 692 | * that we have found the expected number of bytes would fail if executed.) |
| 693 | * Thus the non-continuation malformation is really unnecessary, being a |
| 694 | * subset of the too-short malformation. But there may be existing |
| 695 | * applications that are expecting the non-continuation type, so we retain |
| 696 | * it, and return it in preference to the too-short malformation. (If this |
| 697 | * code were being written from scratch, the two types might be collapsed |
| 698 | * into one.) I, khw, am also giving priority to returning the |
| 699 | * non-continuation and too-short malformations over overflow when multiple |
| 700 | * ones are present. I don't know of any real reason to prefer one over |
| 701 | * the other, except that it seems to me that multiple-byte errors trumps |
| 702 | * errors from a single byte */ |
| 703 | if (UNLIKELY(unexpected_non_continuation)) { |
| 704 | if (!(flags & UTF8_ALLOW_NON_CONTINUATION)) { |
| 705 | if (! (flags & UTF8_CHECK_ONLY)) { |
| 706 | if (curlen == 1) { |
| 707 | sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (unexpected non-continuation byte 0x%02x, immediately after start byte 0x%02x)", malformed_text, *s, *s0)); |
| 708 | } |
| 709 | else { |
| 710 | sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (unexpected non-continuation byte 0x%02x, %d bytes after start byte 0x%02x, expected %d bytes)", malformed_text, *s, (int) curlen, *s0, (int)expectlen)); |
| 711 | } |
| 712 | } |
| 713 | goto malformed; |
| 714 | } |
| 715 | uv = UNICODE_REPLACEMENT; |
| 716 | |
| 717 | /* Skip testing for overlongs, as the REPLACEMENT may not be the same |
| 718 | * as what the original expectations were. */ |
| 719 | do_overlong_test = FALSE; |
| 720 | if (retlen) { |
| 721 | *retlen = curlen; |
| 722 | } |
| 723 | } |
| 724 | else if (UNLIKELY(curlen < expectlen)) { |
| 725 | if (! (flags & UTF8_ALLOW_SHORT)) { |
| 726 | if (! (flags & UTF8_CHECK_ONLY)) { |
| 727 | sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (%d byte%s, need %d, after start byte 0x%02x)", malformed_text, (int)curlen, curlen == 1 ? "" : "s", (int)expectlen, *s0)); |
| 728 | } |
| 729 | goto malformed; |
| 730 | } |
| 731 | uv = UNICODE_REPLACEMENT; |
| 732 | do_overlong_test = FALSE; |
| 733 | if (retlen) { |
| 734 | *retlen = curlen; |
| 735 | } |
| 736 | } |
| 737 | |
| 738 | #ifndef EBCDIC /* EBCDIC allows FE, FF, can't overflow */ |
| 739 | if ((*s0 & 0xFE) == 0xFE /* matches both FE, FF */ |
| 740 | && (flags & (UTF8_WARN_FE_FF|UTF8_DISALLOW_FE_FF))) |
| 741 | { |
| 742 | /* By adding UTF8_CHECK_ONLY to the test, we avoid unnecessary |
| 743 | * generation of the sv, since no warnings are raised under CHECK */ |
| 744 | if ((flags & (UTF8_WARN_FE_FF|UTF8_CHECK_ONLY)) == UTF8_WARN_FE_FF |
| 745 | && ckWARN_d(WARN_UTF8)) |
| 746 | { |
| 747 | /* This message is deliberately not of the same syntax as the other |
| 748 | * messages for malformations, for backwards compatibility in the |
| 749 | * unlikely event that code is relying on its precise earlier text |
| 750 | */ |
| 751 | sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s Code point beginning with byte 0x%02X is not Unicode, and not portable", malformed_text, *s0)); |
| 752 | pack_warn = packWARN(WARN_UTF8); |
| 753 | } |
| 754 | if (flags & UTF8_DISALLOW_FE_FF) { |
| 755 | goto malformed; |
| 756 | } |
| 757 | } |
| 758 | if (UNLIKELY(overflowed)) { |
| 759 | |
| 760 | /* If the first byte is FF, it will overflow a 32-bit word. If the |
| 761 | * first byte is FE, it will overflow a signed 32-bit word. The |
| 762 | * above preserves backward compatibility, since its message was used |
| 763 | * in earlier versions of this code in preference to overflow */ |
| 764 | sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (overflow at byte 0x%02x, after start byte 0x%02x)", malformed_text, overflow_byte, *s0)); |
| 765 | goto malformed; |
| 766 | } |
| 767 | #endif |
| 768 | |
| 769 | if (do_overlong_test |
| 770 | && expectlen > (STRLEN)UNISKIP(uv) |
| 771 | && ! (flags & UTF8_ALLOW_LONG)) |
| 772 | { |
| 773 | /* The overlong malformation has lower precedence than the others. |
| 774 | * Note that if this malformation is allowed, we return the actual |
| 775 | * value, instead of the replacement character. This is because this |
| 776 | * value is actually well-defined. */ |
| 777 | if (! (flags & UTF8_CHECK_ONLY)) { |
| 778 | sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (%d byte%s, need %d, after start byte 0x%02x)", malformed_text, (int)expectlen, expectlen == 1 ? "": "s", UNISKIP(uv), *s0)); |
| 779 | } |
| 780 | goto malformed; |
| 781 | } |
| 782 | |
| 783 | /* Here, the input is considered to be well-formed , but could be a |
| 784 | * problematic code point that is not allowed by the input parameters. */ |
| 785 | if (uv >= UNICODE_SURROGATE_FIRST /* isn't problematic if < this */ |
| 786 | && (flags & (UTF8_DISALLOW_ILLEGAL_INTERCHANGE |
| 787 | |UTF8_WARN_ILLEGAL_INTERCHANGE))) |
| 788 | { |
| 789 | if (UNICODE_IS_SURROGATE(uv)) { |
| 790 | if ((flags & (UTF8_WARN_SURROGATE|UTF8_CHECK_ONLY)) == UTF8_WARN_SURROGATE |
| 791 | && ckWARN2_d(WARN_UTF8, WARN_SURROGATE)) |
| 792 | { |
| 793 | sv = sv_2mortal(Perl_newSVpvf(aTHX_ "UTF-16 surrogate U+%04"UVXf"", uv)); |
| 794 | pack_warn = packWARN2(WARN_UTF8, WARN_SURROGATE); |
| 795 | } |
| 796 | if (flags & UTF8_DISALLOW_SURROGATE) { |
| 797 | goto disallowed; |
| 798 | } |
| 799 | } |
| 800 | else if ((uv > PERL_UNICODE_MAX)) { |
| 801 | if ((flags & (UTF8_WARN_SUPER|UTF8_CHECK_ONLY)) == UTF8_WARN_SUPER |
| 802 | && ckWARN2_d(WARN_UTF8, WARN_NON_UNICODE)) |
| 803 | { |
| 804 | sv = sv_2mortal(Perl_newSVpvf(aTHX_ "Code point 0x%04"UVXf" is not Unicode, may not be portable", uv)); |
| 805 | pack_warn = packWARN2(WARN_UTF8, WARN_NON_UNICODE); |
| 806 | } |
| 807 | if (flags & UTF8_DISALLOW_SUPER) { |
| 808 | goto disallowed; |
| 809 | } |
| 810 | } |
| 811 | else if (UNICODE_IS_NONCHAR(uv)) { |
| 812 | if ((flags & (UTF8_WARN_NONCHAR|UTF8_CHECK_ONLY)) == UTF8_WARN_NONCHAR |
| 813 | && ckWARN2_d(WARN_UTF8, WARN_NONCHAR)) |
| 814 | { |
| 815 | sv = sv_2mortal(Perl_newSVpvf(aTHX_ "Unicode non-character U+%04"UVXf" is illegal for open interchange", uv)); |
| 816 | pack_warn = packWARN2(WARN_UTF8, WARN_NONCHAR); |
| 817 | } |
| 818 | if (flags & UTF8_DISALLOW_NONCHAR) { |
| 819 | goto disallowed; |
| 820 | } |
| 821 | } |
| 822 | |
| 823 | if (sv) { |
| 824 | outlier_ret = uv; |
| 825 | goto do_warn; |
| 826 | } |
| 827 | |
| 828 | /* Here, this is not considered a malformed character, so drop through |
| 829 | * to return it */ |
| 830 | } |
| 831 | |
| 832 | return uv; |
| 833 | |
| 834 | /* There are three cases which get to beyond this point. In all 3 cases: |
| 835 | * <sv> if not null points to a string to print as a warning. |
| 836 | * <curlen> is what <*retlen> should be set to if UTF8_CHECK_ONLY isn't |
| 837 | * set. |
| 838 | * <outlier_ret> is what return value to use if UTF8_CHECK_ONLY isn't set. |
| 839 | * This is done by initializing it to 0, and changing it only |
| 840 | * for case 1). |
| 841 | * The 3 cases are: |
| 842 | * 1) The input is valid but problematic, and to be warned about. The |
| 843 | * return value is the resultant code point; <*retlen> is set to |
| 844 | * <curlen>, the number of bytes that comprise the code point. |
| 845 | * <pack_warn> contains the result of packWARN() for the warning |
| 846 | * types. The entry point for this case is the label <do_warn>; |
| 847 | * 2) The input is a valid code point but disallowed by the parameters to |
| 848 | * this function. The return value is 0. If UTF8_CHECK_ONLY is set, |
| 849 | * <*relen> is -1; otherwise it is <curlen>, the number of bytes that |
| 850 | * comprise the code point. <pack_warn> contains the result of |
| 851 | * packWARN() for the warning types. The entry point for this case is |
| 852 | * the label <disallowed>. |
| 853 | * 3) The input is malformed. The return value is 0. If UTF8_CHECK_ONLY |
| 854 | * is set, <*relen> is -1; otherwise it is <curlen>, the number of |
| 855 | * bytes that comprise the malformation. All such malformations are |
| 856 | * assumed to be warning type <utf8>. The entry point for this case |
| 857 | * is the label <malformed>. |
| 858 | */ |
| 859 | |
| 860 | malformed: |
| 861 | |
| 862 | if (sv && ckWARN_d(WARN_UTF8)) { |
| 863 | pack_warn = packWARN(WARN_UTF8); |
| 864 | } |
| 865 | |
| 866 | disallowed: |
| 867 | |
| 868 | if (flags & UTF8_CHECK_ONLY) { |
| 869 | if (retlen) |
| 870 | *retlen = ((STRLEN) -1); |
| 871 | return 0; |
| 872 | } |
| 873 | |
| 874 | do_warn: |
| 875 | |
| 876 | if (pack_warn) { /* <pack_warn> was initialized to 0, and changed only |
| 877 | if warnings are to be raised. */ |
| 878 | const char * const string = SvPVX_const(sv); |
| 879 | |
| 880 | if (PL_op) |
| 881 | Perl_warner(aTHX_ pack_warn, "%s in %s", string, OP_DESC(PL_op)); |
| 882 | else |
| 883 | Perl_warner(aTHX_ pack_warn, "%s", string); |
| 884 | } |
| 885 | |
| 886 | if (retlen) { |
| 887 | *retlen = curlen; |
| 888 | } |
| 889 | |
| 890 | return outlier_ret; |
| 891 | } |
| 892 | |
| 893 | /* |
| 894 | =for apidoc utf8_to_uvchr_buf |
| 895 | |
| 896 | Returns the native code point of the first character in the string C<s> which |
| 897 | is assumed to be in UTF-8 encoding; C<send> points to 1 beyond the end of C<s>. |
| 898 | C<*retlen> will be set to the length, in bytes, of that character. |
| 899 | |
| 900 | If C<s> does not point to a well-formed UTF-8 character and UTF8 warnings are |
| 901 | enabled, zero is returned and C<*retlen> is set (if C<retlen> isn't |
| 902 | NULL) to -1. If those warnings are off, the computed value if well-defined (or |
| 903 | the Unicode REPLACEMENT CHARACTER, if not) is silently returned, and C<*retlen> |
| 904 | is set (if C<retlen> isn't NULL) so that (S<C<s> + C<*retlen>>) is the |
| 905 | next possible position in C<s> that could begin a non-malformed character. |
| 906 | See L</utf8n_to_uvuni> for details on when the REPLACEMENT CHARACTER is returned. |
| 907 | |
| 908 | =cut |
| 909 | */ |
| 910 | |
| 911 | |
| 912 | UV |
| 913 | Perl_utf8_to_uvchr_buf(pTHX_ const U8 *s, const U8 *send, STRLEN *retlen) |
| 914 | { |
| 915 | PERL_ARGS_ASSERT_UTF8_TO_UVCHR_BUF; |
| 916 | |
| 917 | assert(s < send); |
| 918 | |
| 919 | return utf8n_to_uvchr(s, send - s, retlen, |
| 920 | ckWARN_d(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY); |
| 921 | } |
| 922 | |
| 923 | /* Like L</utf8_to_uvchr_buf>(), but should only be called when it is known that |
| 924 | * there are no malformations in the input UTF-8 string C<s>. surrogates, |
| 925 | * non-character code points, and non-Unicode code points are allowed. A macro |
| 926 | * in utf8.h is used to normally avoid this function wrapper */ |
| 927 | |
| 928 | UV |
| 929 | Perl_valid_utf8_to_uvchr(pTHX_ const U8 *s, STRLEN *retlen) |
| 930 | { |
| 931 | const UV uv = valid_utf8_to_uvuni(s, retlen); |
| 932 | |
| 933 | PERL_ARGS_ASSERT_VALID_UTF8_TO_UVCHR; |
| 934 | |
| 935 | return UNI_TO_NATIVE(uv); |
| 936 | } |
| 937 | |
| 938 | /* |
| 939 | =for apidoc utf8_to_uvchr |
| 940 | |
| 941 | DEPRECATED! |
| 942 | |
| 943 | Returns the native code point of the first character in the string C<s> |
| 944 | which is assumed to be in UTF-8 encoding; C<retlen> will be set to the |
| 945 | length, in bytes, of that character. |
| 946 | |
| 947 | Some, but not all, UTF-8 malformations are detected, and in fact, some |
| 948 | malformed input could cause reading beyond the end of the input buffer, which |
| 949 | is why this function is deprecated. Use L</utf8_to_uvchr_buf> instead. |
| 950 | |
| 951 | If C<s> points to one of the detected malformations, and UTF8 warnings are |
| 952 | enabled, zero is returned and C<*retlen> is set (if C<retlen> isn't |
| 953 | NULL) to -1. If those warnings are off, the computed value if well-defined (or |
| 954 | the Unicode REPLACEMENT CHARACTER, if not) is silently returned, and C<*retlen> |
| 955 | is set (if C<retlen> isn't NULL) so that (S<C<s> + C<*retlen>>) is the |
| 956 | next possible position in C<s> that could begin a non-malformed character. |
| 957 | See L</utf8n_to_uvuni> for details on when the REPLACEMENT CHARACTER is returned. |
| 958 | |
| 959 | =cut |
| 960 | */ |
| 961 | |
| 962 | UV |
| 963 | Perl_utf8_to_uvchr(pTHX_ const U8 *s, STRLEN *retlen) |
| 964 | { |
| 965 | PERL_ARGS_ASSERT_UTF8_TO_UVCHR; |
| 966 | |
| 967 | return utf8_to_uvchr_buf(s, s + UTF8_MAXBYTES, retlen); |
| 968 | } |
| 969 | |
| 970 | /* |
| 971 | =for apidoc utf8_to_uvuni_buf |
| 972 | |
| 973 | Returns the Unicode code point of the first character in the string C<s> which |
| 974 | is assumed to be in UTF-8 encoding; C<send> points to 1 beyond the end of C<s>. |
| 975 | C<retlen> will be set to the length, in bytes, of that character. |
| 976 | |
| 977 | This function should only be used when the returned UV is considered |
| 978 | an index into the Unicode semantic tables (e.g. swashes). |
| 979 | |
| 980 | If C<s> does not point to a well-formed UTF-8 character and UTF8 warnings are |
| 981 | enabled, zero is returned and C<*retlen> is set (if C<retlen> isn't |
| 982 | NULL) to -1. If those warnings are off, the computed value if well-defined (or |
| 983 | the Unicode REPLACEMENT CHARACTER, if not) is silently returned, and C<*retlen> |
| 984 | is set (if C<retlen> isn't NULL) so that (S<C<s> + C<*retlen>>) is the |
| 985 | next possible position in C<s> that could begin a non-malformed character. |
| 986 | See L</utf8n_to_uvuni> for details on when the REPLACEMENT CHARACTER is returned. |
| 987 | |
| 988 | =cut |
| 989 | */ |
| 990 | |
| 991 | UV |
| 992 | Perl_utf8_to_uvuni_buf(pTHX_ const U8 *s, const U8 *send, STRLEN *retlen) |
| 993 | { |
| 994 | PERL_ARGS_ASSERT_UTF8_TO_UVUNI_BUF; |
| 995 | |
| 996 | assert(send > s); |
| 997 | |
| 998 | /* Call the low level routine asking for checks */ |
| 999 | return Perl_utf8n_to_uvuni(aTHX_ s, send -s, retlen, |
| 1000 | ckWARN_d(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY); |
| 1001 | } |
| 1002 | |
| 1003 | /* Like L</utf8_to_uvuni_buf>(), but should only be called when it is known that |
| 1004 | * there are no malformations in the input UTF-8 string C<s>. Surrogates, |
| 1005 | * non-character code points, and non-Unicode code points are allowed */ |
| 1006 | |
| 1007 | UV |
| 1008 | Perl_valid_utf8_to_uvuni(pTHX_ const U8 *s, STRLEN *retlen) |
| 1009 | { |
| 1010 | UV expectlen = UTF8SKIP(s); |
| 1011 | const U8* send = s + expectlen; |
| 1012 | UV uv = NATIVE_TO_UTF(*s); |
| 1013 | |
| 1014 | PERL_ARGS_ASSERT_VALID_UTF8_TO_UVUNI; |
| 1015 | |
| 1016 | if (retlen) { |
| 1017 | *retlen = expectlen; |
| 1018 | } |
| 1019 | |
| 1020 | /* An invariant is trivially returned */ |
| 1021 | if (expectlen == 1) { |
| 1022 | return uv; |
| 1023 | } |
| 1024 | |
| 1025 | /* Remove the leading bits that indicate the number of bytes, leaving just |
| 1026 | * the bits that are part of the value */ |
| 1027 | uv &= UTF_START_MASK(expectlen); |
| 1028 | |
| 1029 | /* Now, loop through the remaining bytes, accumulating each into the |
| 1030 | * working total as we go. (I khw tried unrolling the loop for up to 4 |
| 1031 | * bytes, but there was no performance improvement) */ |
| 1032 | for (++s; s < send; s++) { |
| 1033 | uv = UTF8_ACCUMULATE(uv, *s); |
| 1034 | } |
| 1035 | |
| 1036 | return uv; |
| 1037 | } |
| 1038 | |
| 1039 | /* |
| 1040 | =for apidoc utf8_to_uvuni |
| 1041 | |
| 1042 | DEPRECATED! |
| 1043 | |
| 1044 | Returns the Unicode code point of the first character in the string C<s> |
| 1045 | which is assumed to be in UTF-8 encoding; C<retlen> will be set to the |
| 1046 | length, in bytes, of that character. |
| 1047 | |
| 1048 | This function should only be used when the returned UV is considered |
| 1049 | an index into the Unicode semantic tables (e.g. swashes). |
| 1050 | |
| 1051 | Some, but not all, UTF-8 malformations are detected, and in fact, some |
| 1052 | malformed input could cause reading beyond the end of the input buffer, which |
| 1053 | is why this function is deprecated. Use L</utf8_to_uvuni_buf> instead. |
| 1054 | |
| 1055 | If C<s> points to one of the detected malformations, and UTF8 warnings are |
| 1056 | enabled, zero is returned and C<*retlen> is set (if C<retlen> doesn't point to |
| 1057 | NULL) to -1. If those warnings are off, the computed value if well-defined (or |
| 1058 | the Unicode REPLACEMENT CHARACTER, if not) is silently returned, and C<*retlen> |
| 1059 | is set (if C<retlen> isn't NULL) so that (S<C<s> + C<*retlen>>) is the |
| 1060 | next possible position in C<s> that could begin a non-malformed character. |
| 1061 | See L</utf8n_to_uvuni> for details on when the REPLACEMENT CHARACTER is returned. |
| 1062 | |
| 1063 | =cut |
| 1064 | */ |
| 1065 | |
| 1066 | UV |
| 1067 | Perl_utf8_to_uvuni(pTHX_ const U8 *s, STRLEN *retlen) |
| 1068 | { |
| 1069 | PERL_ARGS_ASSERT_UTF8_TO_UVUNI; |
| 1070 | |
| 1071 | return valid_utf8_to_uvuni(s, retlen); |
| 1072 | } |
| 1073 | |
| 1074 | /* |
| 1075 | =for apidoc utf8_length |
| 1076 | |
| 1077 | Return the length of the UTF-8 char encoded string C<s> in characters. |
| 1078 | Stops at C<e> (inclusive). If C<e E<lt> s> or if the scan would end |
| 1079 | up past C<e>, croaks. |
| 1080 | |
| 1081 | =cut |
| 1082 | */ |
| 1083 | |
| 1084 | STRLEN |
| 1085 | Perl_utf8_length(pTHX_ const U8 *s, const U8 *e) |
| 1086 | { |
| 1087 | dVAR; |
| 1088 | STRLEN len = 0; |
| 1089 | |
| 1090 | PERL_ARGS_ASSERT_UTF8_LENGTH; |
| 1091 | |
| 1092 | /* Note: cannot use UTF8_IS_...() too eagerly here since e.g. |
| 1093 | * the bitops (especially ~) can create illegal UTF-8. |
| 1094 | * In other words: in Perl UTF-8 is not just for Unicode. */ |
| 1095 | |
| 1096 | if (e < s) |
| 1097 | goto warn_and_return; |
| 1098 | while (s < e) { |
| 1099 | if (!UTF8_IS_INVARIANT(*s)) |
| 1100 | s += UTF8SKIP(s); |
| 1101 | else |
| 1102 | s++; |
| 1103 | len++; |
| 1104 | } |
| 1105 | |
| 1106 | if (e != s) { |
| 1107 | len--; |
| 1108 | warn_and_return: |
| 1109 | if (PL_op) |
| 1110 | Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), |
| 1111 | "%s in %s", unees, OP_DESC(PL_op)); |
| 1112 | else |
| 1113 | Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), "%s", unees); |
| 1114 | } |
| 1115 | |
| 1116 | return len; |
| 1117 | } |
| 1118 | |
| 1119 | /* |
| 1120 | =for apidoc utf8_distance |
| 1121 | |
| 1122 | Returns the number of UTF-8 characters between the UTF-8 pointers C<a> |
| 1123 | and C<b>. |
| 1124 | |
| 1125 | WARNING: use only if you *know* that the pointers point inside the |
| 1126 | same UTF-8 buffer. |
| 1127 | |
| 1128 | =cut |
| 1129 | */ |
| 1130 | |
| 1131 | IV |
| 1132 | Perl_utf8_distance(pTHX_ const U8 *a, const U8 *b) |
| 1133 | { |
| 1134 | PERL_ARGS_ASSERT_UTF8_DISTANCE; |
| 1135 | |
| 1136 | return (a < b) ? -1 * (IV) utf8_length(a, b) : (IV) utf8_length(b, a); |
| 1137 | } |
| 1138 | |
| 1139 | /* |
| 1140 | =for apidoc utf8_hop |
| 1141 | |
| 1142 | Return the UTF-8 pointer C<s> displaced by C<off> characters, either |
| 1143 | forward or backward. |
| 1144 | |
| 1145 | WARNING: do not use the following unless you *know* C<off> is within |
| 1146 | the UTF-8 data pointed to by C<s> *and* that on entry C<s> is aligned |
| 1147 | on the first byte of character or just after the last byte of a character. |
| 1148 | |
| 1149 | =cut |
| 1150 | */ |
| 1151 | |
| 1152 | U8 * |
| 1153 | Perl_utf8_hop(pTHX_ const U8 *s, I32 off) |
| 1154 | { |
| 1155 | PERL_ARGS_ASSERT_UTF8_HOP; |
| 1156 | |
| 1157 | PERL_UNUSED_CONTEXT; |
| 1158 | /* Note: cannot use UTF8_IS_...() too eagerly here since e.g |
| 1159 | * the bitops (especially ~) can create illegal UTF-8. |
| 1160 | * In other words: in Perl UTF-8 is not just for Unicode. */ |
| 1161 | |
| 1162 | if (off >= 0) { |
| 1163 | while (off--) |
| 1164 | s += UTF8SKIP(s); |
| 1165 | } |
| 1166 | else { |
| 1167 | while (off++) { |
| 1168 | s--; |
| 1169 | while (UTF8_IS_CONTINUATION(*s)) |
| 1170 | s--; |
| 1171 | } |
| 1172 | } |
| 1173 | return (U8 *)s; |
| 1174 | } |
| 1175 | |
| 1176 | /* |
| 1177 | =for apidoc bytes_cmp_utf8 |
| 1178 | |
| 1179 | Compares the sequence of characters (stored as octets) in C<b>, C<blen> with the |
| 1180 | sequence of characters (stored as UTF-8) in C<u>, C<ulen>. Returns 0 if they are |
| 1181 | equal, -1 or -2 if the first string is less than the second string, +1 or +2 |
| 1182 | if the first string is greater than the second string. |
| 1183 | |
| 1184 | -1 or +1 is returned if the shorter string was identical to the start of the |
| 1185 | longer string. -2 or +2 is returned if the was a difference between characters |
| 1186 | within the strings. |
| 1187 | |
| 1188 | =cut |
| 1189 | */ |
| 1190 | |
| 1191 | int |
| 1192 | Perl_bytes_cmp_utf8(pTHX_ const U8 *b, STRLEN blen, const U8 *u, STRLEN ulen) |
| 1193 | { |
| 1194 | const U8 *const bend = b + blen; |
| 1195 | const U8 *const uend = u + ulen; |
| 1196 | |
| 1197 | PERL_ARGS_ASSERT_BYTES_CMP_UTF8; |
| 1198 | |
| 1199 | PERL_UNUSED_CONTEXT; |
| 1200 | |
| 1201 | while (b < bend && u < uend) { |
| 1202 | U8 c = *u++; |
| 1203 | if (!UTF8_IS_INVARIANT(c)) { |
| 1204 | if (UTF8_IS_DOWNGRADEABLE_START(c)) { |
| 1205 | if (u < uend) { |
| 1206 | U8 c1 = *u++; |
| 1207 | if (UTF8_IS_CONTINUATION(c1)) { |
| 1208 | c = UNI_TO_NATIVE(TWO_BYTE_UTF8_TO_UNI(c, c1)); |
| 1209 | } else { |
| 1210 | Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), |
| 1211 | "Malformed UTF-8 character " |
| 1212 | "(unexpected non-continuation byte 0x%02x" |
| 1213 | ", immediately after start byte 0x%02x)" |
| 1214 | /* Dear diag.t, it's in the pod. */ |
| 1215 | "%s%s", c1, c, |
| 1216 | PL_op ? " in " : "", |
| 1217 | PL_op ? OP_DESC(PL_op) : ""); |
| 1218 | return -2; |
| 1219 | } |
| 1220 | } else { |
| 1221 | if (PL_op) |
| 1222 | Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), |
| 1223 | "%s in %s", unees, OP_DESC(PL_op)); |
| 1224 | else |
| 1225 | Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), "%s", unees); |
| 1226 | return -2; /* Really want to return undef :-) */ |
| 1227 | } |
| 1228 | } else { |
| 1229 | return -2; |
| 1230 | } |
| 1231 | } |
| 1232 | if (*b != c) { |
| 1233 | return *b < c ? -2 : +2; |
| 1234 | } |
| 1235 | ++b; |
| 1236 | } |
| 1237 | |
| 1238 | if (b == bend && u == uend) |
| 1239 | return 0; |
| 1240 | |
| 1241 | return b < bend ? +1 : -1; |
| 1242 | } |
| 1243 | |
| 1244 | /* |
| 1245 | =for apidoc utf8_to_bytes |
| 1246 | |
| 1247 | Converts a string C<s> of length C<len> from UTF-8 into native byte encoding. |
| 1248 | Unlike L</bytes_to_utf8>, this over-writes the original string, and |
| 1249 | updates C<len> to contain the new length. |
| 1250 | Returns zero on failure, setting C<len> to -1. |
| 1251 | |
| 1252 | If you need a copy of the string, see L</bytes_from_utf8>. |
| 1253 | |
| 1254 | =cut |
| 1255 | */ |
| 1256 | |
| 1257 | U8 * |
| 1258 | Perl_utf8_to_bytes(pTHX_ U8 *s, STRLEN *len) |
| 1259 | { |
| 1260 | U8 * const save = s; |
| 1261 | U8 * const send = s + *len; |
| 1262 | U8 *d; |
| 1263 | |
| 1264 | PERL_ARGS_ASSERT_UTF8_TO_BYTES; |
| 1265 | |
| 1266 | /* ensure valid UTF-8 and chars < 256 before updating string */ |
| 1267 | while (s < send) { |
| 1268 | U8 c = *s++; |
| 1269 | |
| 1270 | if (!UTF8_IS_INVARIANT(c) && |
| 1271 | (!UTF8_IS_DOWNGRADEABLE_START(c) || (s >= send) |
| 1272 | || !(c = *s++) || !UTF8_IS_CONTINUATION(c))) { |
| 1273 | *len = ((STRLEN) -1); |
| 1274 | return 0; |
| 1275 | } |
| 1276 | } |
| 1277 | |
| 1278 | d = s = save; |
| 1279 | while (s < send) { |
| 1280 | STRLEN ulen; |
| 1281 | *d++ = (U8)utf8_to_uvchr_buf(s, send, &ulen); |
| 1282 | s += ulen; |
| 1283 | } |
| 1284 | *d = '\0'; |
| 1285 | *len = d - save; |
| 1286 | return save; |
| 1287 | } |
| 1288 | |
| 1289 | /* |
| 1290 | =for apidoc bytes_from_utf8 |
| 1291 | |
| 1292 | Converts a string C<s> of length C<len> from UTF-8 into native byte encoding. |
| 1293 | Unlike L</utf8_to_bytes> but like L</bytes_to_utf8>, returns a pointer to |
| 1294 | the newly-created string, and updates C<len> to contain the new |
| 1295 | length. Returns the original string if no conversion occurs, C<len> |
| 1296 | is unchanged. Do nothing if C<is_utf8> points to 0. Sets C<is_utf8> to |
| 1297 | 0 if C<s> is converted or consisted entirely of characters that are invariant |
| 1298 | in utf8 (i.e., US-ASCII on non-EBCDIC machines). |
| 1299 | |
| 1300 | =cut |
| 1301 | */ |
| 1302 | |
| 1303 | U8 * |
| 1304 | Perl_bytes_from_utf8(pTHX_ const U8 *s, STRLEN *len, bool *is_utf8) |
| 1305 | { |
| 1306 | U8 *d; |
| 1307 | const U8 *start = s; |
| 1308 | const U8 *send; |
| 1309 | I32 count = 0; |
| 1310 | |
| 1311 | PERL_ARGS_ASSERT_BYTES_FROM_UTF8; |
| 1312 | |
| 1313 | PERL_UNUSED_CONTEXT; |
| 1314 | if (!*is_utf8) |
| 1315 | return (U8 *)start; |
| 1316 | |
| 1317 | /* ensure valid UTF-8 and chars < 256 before converting string */ |
| 1318 | for (send = s + *len; s < send;) { |
| 1319 | U8 c = *s++; |
| 1320 | if (!UTF8_IS_INVARIANT(c)) { |
| 1321 | if (UTF8_IS_DOWNGRADEABLE_START(c) && s < send && |
| 1322 | (c = *s++) && UTF8_IS_CONTINUATION(c)) |
| 1323 | count++; |
| 1324 | else |
| 1325 | return (U8 *)start; |
| 1326 | } |
| 1327 | } |
| 1328 | |
| 1329 | *is_utf8 = FALSE; |
| 1330 | |
| 1331 | Newx(d, (*len) - count + 1, U8); |
| 1332 | s = start; start = d; |
| 1333 | while (s < send) { |
| 1334 | U8 c = *s++; |
| 1335 | if (!UTF8_IS_INVARIANT(c)) { |
| 1336 | /* Then it is two-byte encoded */ |
| 1337 | c = UNI_TO_NATIVE(TWO_BYTE_UTF8_TO_UNI(c, *s++)); |
| 1338 | } |
| 1339 | *d++ = c; |
| 1340 | } |
| 1341 | *d = '\0'; |
| 1342 | *len = d - start; |
| 1343 | return (U8 *)start; |
| 1344 | } |
| 1345 | |
| 1346 | /* |
| 1347 | =for apidoc bytes_to_utf8 |
| 1348 | |
| 1349 | Converts a string C<s> of length C<len> bytes from the native encoding into |
| 1350 | UTF-8. |
| 1351 | Returns a pointer to the newly-created string, and sets C<len> to |
| 1352 | reflect the new length in bytes. |
| 1353 | |
| 1354 | A NUL character will be written after the end of the string. |
| 1355 | |
| 1356 | If you want to convert to UTF-8 from encodings other than |
| 1357 | the native (Latin1 or EBCDIC), |
| 1358 | see L</sv_recode_to_utf8>(). |
| 1359 | |
| 1360 | =cut |
| 1361 | */ |
| 1362 | |
| 1363 | /* This logic is duplicated in sv_catpvn_flags, so any bug fixes will |
| 1364 | likewise need duplication. */ |
| 1365 | |
| 1366 | U8* |
| 1367 | Perl_bytes_to_utf8(pTHX_ const U8 *s, STRLEN *len) |
| 1368 | { |
| 1369 | const U8 * const send = s + (*len); |
| 1370 | U8 *d; |
| 1371 | U8 *dst; |
| 1372 | |
| 1373 | PERL_ARGS_ASSERT_BYTES_TO_UTF8; |
| 1374 | PERL_UNUSED_CONTEXT; |
| 1375 | |
| 1376 | Newx(d, (*len) * 2 + 1, U8); |
| 1377 | dst = d; |
| 1378 | |
| 1379 | while (s < send) { |
| 1380 | const UV uv = NATIVE_TO_ASCII(*s++); |
| 1381 | if (UNI_IS_INVARIANT(uv)) |
| 1382 | *d++ = (U8)UTF_TO_NATIVE(uv); |
| 1383 | else { |
| 1384 | *d++ = (U8)UTF8_EIGHT_BIT_HI(uv); |
| 1385 | *d++ = (U8)UTF8_EIGHT_BIT_LO(uv); |
| 1386 | } |
| 1387 | } |
| 1388 | *d = '\0'; |
| 1389 | *len = d-dst; |
| 1390 | return dst; |
| 1391 | } |
| 1392 | |
| 1393 | /* |
| 1394 | * Convert native (big-endian) or reversed (little-endian) UTF-16 to UTF-8. |
| 1395 | * |
| 1396 | * Destination must be pre-extended to 3/2 source. Do not use in-place. |
| 1397 | * We optimize for native, for obvious reasons. */ |
| 1398 | |
| 1399 | U8* |
| 1400 | Perl_utf16_to_utf8(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen) |
| 1401 | { |
| 1402 | U8* pend; |
| 1403 | U8* dstart = d; |
| 1404 | |
| 1405 | PERL_ARGS_ASSERT_UTF16_TO_UTF8; |
| 1406 | |
| 1407 | if (bytelen & 1) |
| 1408 | Perl_croak(aTHX_ "panic: utf16_to_utf8: odd bytelen %"UVuf, (UV)bytelen); |
| 1409 | |
| 1410 | pend = p + bytelen; |
| 1411 | |
| 1412 | while (p < pend) { |
| 1413 | UV uv = (p[0] << 8) + p[1]; /* UTF-16BE */ |
| 1414 | p += 2; |
| 1415 | if (uv < 0x80) { |
| 1416 | #ifdef EBCDIC |
| 1417 | *d++ = UNI_TO_NATIVE(uv); |
| 1418 | #else |
| 1419 | *d++ = (U8)uv; |
| 1420 | #endif |
| 1421 | continue; |
| 1422 | } |
| 1423 | if (uv < 0x800) { |
| 1424 | *d++ = (U8)(( uv >> 6) | 0xc0); |
| 1425 | *d++ = (U8)(( uv & 0x3f) | 0x80); |
| 1426 | continue; |
| 1427 | } |
| 1428 | if (uv >= 0xd800 && uv <= 0xdbff) { /* surrogates */ |
| 1429 | if (p >= pend) { |
| 1430 | Perl_croak(aTHX_ "Malformed UTF-16 surrogate"); |
| 1431 | } else { |
| 1432 | UV low = (p[0] << 8) + p[1]; |
| 1433 | p += 2; |
| 1434 | if (low < 0xdc00 || low > 0xdfff) |
| 1435 | Perl_croak(aTHX_ "Malformed UTF-16 surrogate"); |
| 1436 | uv = ((uv - 0xd800) << 10) + (low - 0xdc00) + 0x10000; |
| 1437 | } |
| 1438 | } else if (uv >= 0xdc00 && uv <= 0xdfff) { |
| 1439 | Perl_croak(aTHX_ "Malformed UTF-16 surrogate"); |
| 1440 | } |
| 1441 | if (uv < 0x10000) { |
| 1442 | *d++ = (U8)(( uv >> 12) | 0xe0); |
| 1443 | *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80); |
| 1444 | *d++ = (U8)(( uv & 0x3f) | 0x80); |
| 1445 | continue; |
| 1446 | } |
| 1447 | else { |
| 1448 | *d++ = (U8)(( uv >> 18) | 0xf0); |
| 1449 | *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80); |
| 1450 | *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80); |
| 1451 | *d++ = (U8)(( uv & 0x3f) | 0x80); |
| 1452 | continue; |
| 1453 | } |
| 1454 | } |
| 1455 | *newlen = d - dstart; |
| 1456 | return d; |
| 1457 | } |
| 1458 | |
| 1459 | /* Note: this one is slightly destructive of the source. */ |
| 1460 | |
| 1461 | U8* |
| 1462 | Perl_utf16_to_utf8_reversed(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen) |
| 1463 | { |
| 1464 | U8* s = (U8*)p; |
| 1465 | U8* const send = s + bytelen; |
| 1466 | |
| 1467 | PERL_ARGS_ASSERT_UTF16_TO_UTF8_REVERSED; |
| 1468 | |
| 1469 | if (bytelen & 1) |
| 1470 | Perl_croak(aTHX_ "panic: utf16_to_utf8_reversed: odd bytelen %"UVuf, |
| 1471 | (UV)bytelen); |
| 1472 | |
| 1473 | while (s < send) { |
| 1474 | const U8 tmp = s[0]; |
| 1475 | s[0] = s[1]; |
| 1476 | s[1] = tmp; |
| 1477 | s += 2; |
| 1478 | } |
| 1479 | return utf16_to_utf8(p, d, bytelen, newlen); |
| 1480 | } |
| 1481 | |
| 1482 | /* for now these are all defined (inefficiently) in terms of the utf8 versions. |
| 1483 | * Note that the macros in handy.h that call these short-circuit calling them |
| 1484 | * for Latin-1 range inputs */ |
| 1485 | |
| 1486 | bool |
| 1487 | Perl_is_uni_alnum(pTHX_ UV c) |
| 1488 | { |
| 1489 | U8 tmpbuf[UTF8_MAXBYTES+1]; |
| 1490 | uvchr_to_utf8(tmpbuf, c); |
| 1491 | return is_utf8_alnum(tmpbuf); |
| 1492 | } |
| 1493 | |
| 1494 | bool |
| 1495 | Perl_is_uni_idfirst(pTHX_ UV c) |
| 1496 | { |
| 1497 | U8 tmpbuf[UTF8_MAXBYTES+1]; |
| 1498 | uvchr_to_utf8(tmpbuf, c); |
| 1499 | return is_utf8_idfirst(tmpbuf); |
| 1500 | } |
| 1501 | |
| 1502 | bool |
| 1503 | Perl_is_uni_alpha(pTHX_ UV c) |
| 1504 | { |
| 1505 | U8 tmpbuf[UTF8_MAXBYTES+1]; |
| 1506 | uvchr_to_utf8(tmpbuf, c); |
| 1507 | return is_utf8_alpha(tmpbuf); |
| 1508 | } |
| 1509 | |
| 1510 | bool |
| 1511 | Perl_is_uni_ascii(pTHX_ UV c) |
| 1512 | { |
| 1513 | return isASCII(c); |
| 1514 | } |
| 1515 | |
| 1516 | bool |
| 1517 | Perl_is_uni_blank(pTHX_ UV c) |
| 1518 | { |
| 1519 | U8 tmpbuf[UTF8_MAXBYTES+1]; |
| 1520 | uvchr_to_utf8(tmpbuf, c); |
| 1521 | return is_utf8_blank(tmpbuf); |
| 1522 | } |
| 1523 | |
| 1524 | bool |
| 1525 | Perl_is_uni_space(pTHX_ UV c) |
| 1526 | { |
| 1527 | U8 tmpbuf[UTF8_MAXBYTES+1]; |
| 1528 | uvchr_to_utf8(tmpbuf, c); |
| 1529 | return is_utf8_space(tmpbuf); |
| 1530 | } |
| 1531 | |
| 1532 | bool |
| 1533 | Perl_is_uni_digit(pTHX_ UV c) |
| 1534 | { |
| 1535 | U8 tmpbuf[UTF8_MAXBYTES+1]; |
| 1536 | uvchr_to_utf8(tmpbuf, c); |
| 1537 | return is_utf8_digit(tmpbuf); |
| 1538 | } |
| 1539 | |
| 1540 | bool |
| 1541 | Perl_is_uni_upper(pTHX_ UV c) |
| 1542 | { |
| 1543 | U8 tmpbuf[UTF8_MAXBYTES+1]; |
| 1544 | uvchr_to_utf8(tmpbuf, c); |
| 1545 | return is_utf8_upper(tmpbuf); |
| 1546 | } |
| 1547 | |
| 1548 | bool |
| 1549 | Perl_is_uni_lower(pTHX_ UV c) |
| 1550 | { |
| 1551 | U8 tmpbuf[UTF8_MAXBYTES+1]; |
| 1552 | uvchr_to_utf8(tmpbuf, c); |
| 1553 | return is_utf8_lower(tmpbuf); |
| 1554 | } |
| 1555 | |
| 1556 | bool |
| 1557 | Perl_is_uni_cntrl(pTHX_ UV c) |
| 1558 | { |
| 1559 | return isCNTRL_L1(c); |
| 1560 | } |
| 1561 | |
| 1562 | bool |
| 1563 | Perl_is_uni_graph(pTHX_ UV c) |
| 1564 | { |
| 1565 | U8 tmpbuf[UTF8_MAXBYTES+1]; |
| 1566 | uvchr_to_utf8(tmpbuf, c); |
| 1567 | return is_utf8_graph(tmpbuf); |
| 1568 | } |
| 1569 | |
| 1570 | bool |
| 1571 | Perl_is_uni_print(pTHX_ UV c) |
| 1572 | { |
| 1573 | U8 tmpbuf[UTF8_MAXBYTES+1]; |
| 1574 | uvchr_to_utf8(tmpbuf, c); |
| 1575 | return is_utf8_print(tmpbuf); |
| 1576 | } |
| 1577 | |
| 1578 | bool |
| 1579 | Perl_is_uni_punct(pTHX_ UV c) |
| 1580 | { |
| 1581 | U8 tmpbuf[UTF8_MAXBYTES+1]; |
| 1582 | uvchr_to_utf8(tmpbuf, c); |
| 1583 | return is_utf8_punct(tmpbuf); |
| 1584 | } |
| 1585 | |
| 1586 | bool |
| 1587 | Perl_is_uni_xdigit(pTHX_ UV c) |
| 1588 | { |
| 1589 | U8 tmpbuf[UTF8_MAXBYTES_CASE+1]; |
| 1590 | uvchr_to_utf8(tmpbuf, c); |
| 1591 | return is_utf8_xdigit(tmpbuf); |
| 1592 | } |
| 1593 | |
| 1594 | UV |
| 1595 | Perl__to_upper_title_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp, const char S_or_s) |
| 1596 | { |
| 1597 | /* We have the latin1-range values compiled into the core, so just use |
| 1598 | * those, converting the result to utf8. The only difference between upper |
| 1599 | * and title case in this range is that LATIN_SMALL_LETTER_SHARP_S is |
| 1600 | * either "SS" or "Ss". Which one to use is passed into the routine in |
| 1601 | * 'S_or_s' to avoid a test */ |
| 1602 | |
| 1603 | UV converted = toUPPER_LATIN1_MOD(c); |
| 1604 | |
| 1605 | PERL_ARGS_ASSERT__TO_UPPER_TITLE_LATIN1; |
| 1606 | |
| 1607 | assert(S_or_s == 'S' || S_or_s == 's'); |
| 1608 | |
| 1609 | if (UNI_IS_INVARIANT(converted)) { /* No difference between the two for |
| 1610 | characters in this range */ |
| 1611 | *p = (U8) converted; |
| 1612 | *lenp = 1; |
| 1613 | return converted; |
| 1614 | } |
| 1615 | |
| 1616 | /* toUPPER_LATIN1_MOD gives the correct results except for three outliers, |
| 1617 | * which it maps to one of them, so as to only have to have one check for |
| 1618 | * it in the main case */ |
| 1619 | if (UNLIKELY(converted == LATIN_SMALL_LETTER_Y_WITH_DIAERESIS)) { |
| 1620 | switch (c) { |
| 1621 | case LATIN_SMALL_LETTER_Y_WITH_DIAERESIS: |
| 1622 | converted = LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS; |
| 1623 | break; |
| 1624 | case MICRO_SIGN: |
| 1625 | converted = GREEK_CAPITAL_LETTER_MU; |
| 1626 | break; |
| 1627 | case LATIN_SMALL_LETTER_SHARP_S: |
| 1628 | *(p)++ = 'S'; |
| 1629 | *p = S_or_s; |
| 1630 | *lenp = 2; |
| 1631 | return 'S'; |
| 1632 | default: |
| 1633 | Perl_croak(aTHX_ "panic: to_upper_title_latin1 did not expect '%c' to map to '%c'", c, LATIN_SMALL_LETTER_Y_WITH_DIAERESIS); |
| 1634 | assert(0); /* NOTREACHED */ |
| 1635 | } |
| 1636 | } |
| 1637 | |
| 1638 | *(p)++ = UTF8_TWO_BYTE_HI(converted); |
| 1639 | *p = UTF8_TWO_BYTE_LO(converted); |
| 1640 | *lenp = 2; |
| 1641 | |
| 1642 | return converted; |
| 1643 | } |
| 1644 | |
| 1645 | /* Call the function to convert a UTF-8 encoded character to the specified case. |
| 1646 | * Note that there may be more than one character in the result. |
| 1647 | * INP is a pointer to the first byte of the input character |
| 1648 | * OUTP will be set to the first byte of the string of changed characters. It |
| 1649 | * needs to have space for UTF8_MAXBYTES_CASE+1 bytes |
| 1650 | * LENP will be set to the length in bytes of the string of changed characters |
| 1651 | * |
| 1652 | * The functions return the ordinal of the first character in the string of OUTP */ |
| 1653 | #define CALL_UPPER_CASE(INP, OUTP, LENP) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_toupper, "ToUc", "utf8::ToSpecUc") |
| 1654 | #define CALL_TITLE_CASE(INP, OUTP, LENP) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_totitle, "ToTc", "utf8::ToSpecTc") |
| 1655 | #define CALL_LOWER_CASE(INP, OUTP, LENP) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_tolower, "ToLc", "utf8::ToSpecLc") |
| 1656 | |
| 1657 | /* This additionally has the input parameter SPECIALS, which if non-zero will |
| 1658 | * cause this to use the SPECIALS hash for folding (meaning get full case |
| 1659 | * folding); otherwise, when zero, this implies a simple case fold */ |
| 1660 | #define CALL_FOLD_CASE(INP, OUTP, LENP, SPECIALS) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_tofold, "ToCf", (SPECIALS) ? "utf8::ToSpecCf" : NULL) |
| 1661 | |
| 1662 | UV |
| 1663 | Perl_to_uni_upper(pTHX_ UV c, U8* p, STRLEN *lenp) |
| 1664 | { |
| 1665 | dVAR; |
| 1666 | |
| 1667 | /* Convert the Unicode character whose ordinal is <c> to its uppercase |
| 1668 | * version and store that in UTF-8 in <p> and its length in bytes in <lenp>. |
| 1669 | * Note that the <p> needs to be at least UTF8_MAXBYTES_CASE+1 bytes since |
| 1670 | * the changed version may be longer than the original character. |
| 1671 | * |
| 1672 | * The ordinal of the first character of the changed version is returned |
| 1673 | * (but note, as explained above, that there may be more.) */ |
| 1674 | |
| 1675 | PERL_ARGS_ASSERT_TO_UNI_UPPER; |
| 1676 | |
| 1677 | if (c < 256) { |
| 1678 | return _to_upper_title_latin1((U8) c, p, lenp, 'S'); |
| 1679 | } |
| 1680 | |
| 1681 | uvchr_to_utf8(p, c); |
| 1682 | return CALL_UPPER_CASE(p, p, lenp); |
| 1683 | } |
| 1684 | |
| 1685 | UV |
| 1686 | Perl_to_uni_title(pTHX_ UV c, U8* p, STRLEN *lenp) |
| 1687 | { |
| 1688 | dVAR; |
| 1689 | |
| 1690 | PERL_ARGS_ASSERT_TO_UNI_TITLE; |
| 1691 | |
| 1692 | if (c < 256) { |
| 1693 | return _to_upper_title_latin1((U8) c, p, lenp, 's'); |
| 1694 | } |
| 1695 | |
| 1696 | uvchr_to_utf8(p, c); |
| 1697 | return CALL_TITLE_CASE(p, p, lenp); |
| 1698 | } |
| 1699 | |
| 1700 | STATIC U8 |
| 1701 | S_to_lower_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp) |
| 1702 | { |
| 1703 | /* We have the latin1-range values compiled into the core, so just use |
| 1704 | * those, converting the result to utf8. Since the result is always just |
| 1705 | * one character, we allow <p> to be NULL */ |
| 1706 | |
| 1707 | U8 converted = toLOWER_LATIN1(c); |
| 1708 | |
| 1709 | if (p != NULL) { |
| 1710 | if (UNI_IS_INVARIANT(converted)) { |
| 1711 | *p = converted; |
| 1712 | *lenp = 1; |
| 1713 | } |
| 1714 | else { |
| 1715 | *p = UTF8_TWO_BYTE_HI(converted); |
| 1716 | *(p+1) = UTF8_TWO_BYTE_LO(converted); |
| 1717 | *lenp = 2; |
| 1718 | } |
| 1719 | } |
| 1720 | return converted; |
| 1721 | } |
| 1722 | |
| 1723 | UV |
| 1724 | Perl_to_uni_lower(pTHX_ UV c, U8* p, STRLEN *lenp) |
| 1725 | { |
| 1726 | dVAR; |
| 1727 | |
| 1728 | PERL_ARGS_ASSERT_TO_UNI_LOWER; |
| 1729 | |
| 1730 | if (c < 256) { |
| 1731 | return to_lower_latin1((U8) c, p, lenp); |
| 1732 | } |
| 1733 | |
| 1734 | uvchr_to_utf8(p, c); |
| 1735 | return CALL_LOWER_CASE(p, p, lenp); |
| 1736 | } |
| 1737 | |
| 1738 | UV |
| 1739 | Perl__to_fold_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp, const bool flags) |
| 1740 | { |
| 1741 | /* Corresponds to to_lower_latin1(), <flags> is TRUE if to use full case |
| 1742 | * folding */ |
| 1743 | |
| 1744 | UV converted; |
| 1745 | |
| 1746 | PERL_ARGS_ASSERT__TO_FOLD_LATIN1; |
| 1747 | |
| 1748 | if (c == MICRO_SIGN) { |
| 1749 | converted = GREEK_SMALL_LETTER_MU; |
| 1750 | } |
| 1751 | else if (flags && c == LATIN_SMALL_LETTER_SHARP_S) { |
| 1752 | *(p)++ = 's'; |
| 1753 | *p = 's'; |
| 1754 | *lenp = 2; |
| 1755 | return 's'; |
| 1756 | } |
| 1757 | else { /* In this range the fold of all other characters is their lower |
| 1758 | case */ |
| 1759 | converted = toLOWER_LATIN1(c); |
| 1760 | } |
| 1761 | |
| 1762 | if (UNI_IS_INVARIANT(converted)) { |
| 1763 | *p = (U8) converted; |
| 1764 | *lenp = 1; |
| 1765 | } |
| 1766 | else { |
| 1767 | *(p)++ = UTF8_TWO_BYTE_HI(converted); |
| 1768 | *p = UTF8_TWO_BYTE_LO(converted); |
| 1769 | *lenp = 2; |
| 1770 | } |
| 1771 | |
| 1772 | return converted; |
| 1773 | } |
| 1774 | |
| 1775 | UV |
| 1776 | Perl__to_uni_fold_flags(pTHX_ UV c, U8* p, STRLEN *lenp, const U8 flags) |
| 1777 | { |
| 1778 | |
| 1779 | /* Not currently externally documented, and subject to change |
| 1780 | * <flags> bits meanings: |
| 1781 | * FOLD_FLAGS_FULL iff full folding is to be used; |
| 1782 | * FOLD_FLAGS_LOCALE iff in locale |
| 1783 | * FOLD_FLAGS_NOMIX_ASCII iff non-ASCII to ASCII folds are prohibited |
| 1784 | */ |
| 1785 | |
| 1786 | PERL_ARGS_ASSERT__TO_UNI_FOLD_FLAGS; |
| 1787 | |
| 1788 | if (c < 256) { |
| 1789 | UV result = _to_fold_latin1((U8) c, p, lenp, |
| 1790 | cBOOL(((flags & FOLD_FLAGS_FULL) |
| 1791 | /* If ASCII-safe, don't allow full folding, |
| 1792 | * as that could include SHARP S => ss; |
| 1793 | * otherwise there is no crossing of |
| 1794 | * ascii/non-ascii in the latin1 range */ |
| 1795 | && ! (flags & FOLD_FLAGS_NOMIX_ASCII)))); |
| 1796 | /* It is illegal for the fold to cross the 255/256 boundary under |
| 1797 | * locale; in this case return the original */ |
| 1798 | return (result > 256 && flags & FOLD_FLAGS_LOCALE) |
| 1799 | ? c |
| 1800 | : result; |
| 1801 | } |
| 1802 | |
| 1803 | /* If no special needs, just use the macro */ |
| 1804 | if ( ! (flags & (FOLD_FLAGS_LOCALE|FOLD_FLAGS_NOMIX_ASCII))) { |
| 1805 | uvchr_to_utf8(p, c); |
| 1806 | return CALL_FOLD_CASE(p, p, lenp, flags & FOLD_FLAGS_FULL); |
| 1807 | } |
| 1808 | else { /* Otherwise, _to_utf8_fold_flags has the intelligence to deal with |
| 1809 | the special flags. */ |
| 1810 | U8 utf8_c[UTF8_MAXBYTES + 1]; |
| 1811 | uvchr_to_utf8(utf8_c, c); |
| 1812 | return _to_utf8_fold_flags(utf8_c, p, lenp, flags, NULL); |
| 1813 | } |
| 1814 | } |
| 1815 | |
| 1816 | /* for now these all assume no locale info available for Unicode > 255; and |
| 1817 | * the corresponding macros in handy.h (like isALNUM_LC_uvchr) should have been |
| 1818 | * called instead, so that these don't get called for < 255 */ |
| 1819 | |
| 1820 | bool |
| 1821 | Perl_is_uni_alnum_lc(pTHX_ UV c) |
| 1822 | { |
| 1823 | return is_uni_alnum(c); /* XXX no locale support yet */ |
| 1824 | } |
| 1825 | |
| 1826 | bool |
| 1827 | Perl_is_uni_idfirst_lc(pTHX_ UV c) |
| 1828 | { |
| 1829 | return is_uni_idfirst(c); /* XXX no locale support yet */ |
| 1830 | } |
| 1831 | |
| 1832 | bool |
| 1833 | Perl_is_uni_alpha_lc(pTHX_ UV c) |
| 1834 | { |
| 1835 | return is_uni_alpha(c); /* XXX no locale support yet */ |
| 1836 | } |
| 1837 | |
| 1838 | bool |
| 1839 | Perl_is_uni_ascii_lc(pTHX_ UV c) |
| 1840 | { |
| 1841 | return is_uni_ascii(c); /* XXX no locale support yet */ |
| 1842 | } |
| 1843 | |
| 1844 | bool |
| 1845 | Perl_is_uni_blank_lc(pTHX_ UV c) |
| 1846 | { |
| 1847 | return is_uni_blank(c); /* XXX no locale support yet */ |
| 1848 | } |
| 1849 | |
| 1850 | bool |
| 1851 | Perl_is_uni_space_lc(pTHX_ UV c) |
| 1852 | { |
| 1853 | return is_uni_space(c); /* XXX no locale support yet */ |
| 1854 | } |
| 1855 | |
| 1856 | bool |
| 1857 | Perl_is_uni_digit_lc(pTHX_ UV c) |
| 1858 | { |
| 1859 | return is_uni_digit(c); /* XXX no locale support yet */ |
| 1860 | } |
| 1861 | |
| 1862 | bool |
| 1863 | Perl_is_uni_upper_lc(pTHX_ UV c) |
| 1864 | { |
| 1865 | return is_uni_upper(c); /* XXX no locale support yet */ |
| 1866 | } |
| 1867 | |
| 1868 | bool |
| 1869 | Perl_is_uni_lower_lc(pTHX_ UV c) |
| 1870 | { |
| 1871 | return is_uni_lower(c); /* XXX no locale support yet */ |
| 1872 | } |
| 1873 | |
| 1874 | bool |
| 1875 | Perl_is_uni_cntrl_lc(pTHX_ UV c) |
| 1876 | { |
| 1877 | return is_uni_cntrl(c); /* XXX no locale support yet */ |
| 1878 | } |
| 1879 | |
| 1880 | bool |
| 1881 | Perl_is_uni_graph_lc(pTHX_ UV c) |
| 1882 | { |
| 1883 | return is_uni_graph(c); /* XXX no locale support yet */ |
| 1884 | } |
| 1885 | |
| 1886 | bool |
| 1887 | Perl_is_uni_print_lc(pTHX_ UV c) |
| 1888 | { |
| 1889 | return is_uni_print(c); /* XXX no locale support yet */ |
| 1890 | } |
| 1891 | |
| 1892 | bool |
| 1893 | Perl_is_uni_punct_lc(pTHX_ UV c) |
| 1894 | { |
| 1895 | return is_uni_punct(c); /* XXX no locale support yet */ |
| 1896 | } |
| 1897 | |
| 1898 | bool |
| 1899 | Perl_is_uni_xdigit_lc(pTHX_ UV c) |
| 1900 | { |
| 1901 | return is_uni_xdigit(c); /* XXX no locale support yet */ |
| 1902 | } |
| 1903 | |
| 1904 | U32 |
| 1905 | Perl_to_uni_upper_lc(pTHX_ U32 c) |
| 1906 | { |
| 1907 | /* XXX returns only the first character -- do not use XXX */ |
| 1908 | /* XXX no locale support yet */ |
| 1909 | STRLEN len; |
| 1910 | U8 tmpbuf[UTF8_MAXBYTES_CASE+1]; |
| 1911 | return (U32)to_uni_upper(c, tmpbuf, &len); |
| 1912 | } |
| 1913 | |
| 1914 | U32 |
| 1915 | Perl_to_uni_title_lc(pTHX_ U32 c) |
| 1916 | { |
| 1917 | /* XXX returns only the first character XXX -- do not use XXX */ |
| 1918 | /* XXX no locale support yet */ |
| 1919 | STRLEN len; |
| 1920 | U8 tmpbuf[UTF8_MAXBYTES_CASE+1]; |
| 1921 | return (U32)to_uni_title(c, tmpbuf, &len); |
| 1922 | } |
| 1923 | |
| 1924 | U32 |
| 1925 | Perl_to_uni_lower_lc(pTHX_ U32 c) |
| 1926 | { |
| 1927 | /* XXX returns only the first character -- do not use XXX */ |
| 1928 | /* XXX no locale support yet */ |
| 1929 | STRLEN len; |
| 1930 | U8 tmpbuf[UTF8_MAXBYTES_CASE+1]; |
| 1931 | return (U32)to_uni_lower(c, tmpbuf, &len); |
| 1932 | } |
| 1933 | |
| 1934 | static bool |
| 1935 | S_is_utf8_common(pTHX_ const U8 *const p, SV **swash, |
| 1936 | const char *const swashname) |
| 1937 | { |
| 1938 | /* returns a boolean giving whether or not the UTF8-encoded character that |
| 1939 | * starts at <p> is in the swash indicated by <swashname>. <swash> |
| 1940 | * contains a pointer to where the swash indicated by <swashname> |
| 1941 | * is to be stored; which this routine will do, so that future calls will |
| 1942 | * look at <*swash> and only generate a swash if it is not null |
| 1943 | * |
| 1944 | * Note that it is assumed that the buffer length of <p> is enough to |
| 1945 | * contain all the bytes that comprise the character. Thus, <*p> should |
| 1946 | * have been checked before this call for mal-formedness enough to assure |
| 1947 | * that. */ |
| 1948 | |
| 1949 | dVAR; |
| 1950 | |
| 1951 | PERL_ARGS_ASSERT_IS_UTF8_COMMON; |
| 1952 | |
| 1953 | /* The API should have included a length for the UTF-8 character in <p>, |
| 1954 | * but it doesn't. We therefor assume that p has been validated at least |
| 1955 | * as far as there being enough bytes available in it to accommodate the |
| 1956 | * character without reading beyond the end, and pass that number on to the |
| 1957 | * validating routine */ |
| 1958 | if (!is_utf8_char_buf(p, p + UTF8SKIP(p))) |
| 1959 | return FALSE; |
| 1960 | if (!*swash) { |
| 1961 | U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST; |
| 1962 | *swash = _core_swash_init("utf8", swashname, &PL_sv_undef, 1, 0, NULL, &flags); |
| 1963 | } |
| 1964 | return swash_fetch(*swash, p, TRUE) != 0; |
| 1965 | } |
| 1966 | |
| 1967 | bool |
| 1968 | Perl_is_utf8_alnum(pTHX_ const U8 *p) |
| 1969 | { |
| 1970 | dVAR; |
| 1971 | |
| 1972 | PERL_ARGS_ASSERT_IS_UTF8_ALNUM; |
| 1973 | |
| 1974 | /* NOTE: "IsWord", not "IsAlnum", since Alnum is a true |
| 1975 | * descendant of isalnum(3), in other words, it doesn't |
| 1976 | * contain the '_'. --jhi */ |
| 1977 | return is_utf8_common(p, &PL_utf8_alnum, "IsWord"); |
| 1978 | } |
| 1979 | |
| 1980 | bool |
| 1981 | Perl_is_utf8_idfirst(pTHX_ const U8 *p) /* The naming is historical. */ |
| 1982 | { |
| 1983 | dVAR; |
| 1984 | |
| 1985 | PERL_ARGS_ASSERT_IS_UTF8_IDFIRST; |
| 1986 | |
| 1987 | if (*p == '_') |
| 1988 | return TRUE; |
| 1989 | /* is_utf8_idstart would be more logical. */ |
| 1990 | return is_utf8_common(p, &PL_utf8_idstart, "IdStart"); |
| 1991 | } |
| 1992 | |
| 1993 | bool |
| 1994 | Perl_is_utf8_xidfirst(pTHX_ const U8 *p) /* The naming is historical. */ |
| 1995 | { |
| 1996 | dVAR; |
| 1997 | |
| 1998 | PERL_ARGS_ASSERT_IS_UTF8_XIDFIRST; |
| 1999 | |
| 2000 | if (*p == '_') |
| 2001 | return TRUE; |
| 2002 | /* is_utf8_idstart would be more logical. */ |
| 2003 | return is_utf8_common(p, &PL_utf8_xidstart, "XIdStart"); |
| 2004 | } |
| 2005 | |
| 2006 | bool |
| 2007 | Perl__is_utf8__perl_idstart(pTHX_ const U8 *p) |
| 2008 | { |
| 2009 | dVAR; |
| 2010 | |
| 2011 | PERL_ARGS_ASSERT__IS_UTF8__PERL_IDSTART; |
| 2012 | |
| 2013 | return is_utf8_common(p, &PL_utf8_perl_idstart, "_Perl_IDStart"); |
| 2014 | } |
| 2015 | |
| 2016 | bool |
| 2017 | Perl_is_utf8_idcont(pTHX_ const U8 *p) |
| 2018 | { |
| 2019 | dVAR; |
| 2020 | |
| 2021 | PERL_ARGS_ASSERT_IS_UTF8_IDCONT; |
| 2022 | |
| 2023 | return is_utf8_common(p, &PL_utf8_idcont, "IdContinue"); |
| 2024 | } |
| 2025 | |
| 2026 | bool |
| 2027 | Perl_is_utf8_xidcont(pTHX_ const U8 *p) |
| 2028 | { |
| 2029 | dVAR; |
| 2030 | |
| 2031 | PERL_ARGS_ASSERT_IS_UTF8_XIDCONT; |
| 2032 | |
| 2033 | return is_utf8_common(p, &PL_utf8_idcont, "XIdContinue"); |
| 2034 | } |
| 2035 | |
| 2036 | bool |
| 2037 | Perl_is_utf8_alpha(pTHX_ const U8 *p) |
| 2038 | { |
| 2039 | dVAR; |
| 2040 | |
| 2041 | PERL_ARGS_ASSERT_IS_UTF8_ALPHA; |
| 2042 | |
| 2043 | return is_utf8_common(p, &PL_utf8_alpha, "IsAlpha"); |
| 2044 | } |
| 2045 | |
| 2046 | bool |
| 2047 | Perl_is_utf8_ascii(pTHX_ const U8 *p) |
| 2048 | { |
| 2049 | dVAR; |
| 2050 | |
| 2051 | PERL_ARGS_ASSERT_IS_UTF8_ASCII; |
| 2052 | |
| 2053 | /* ASCII characters are the same whether in utf8 or not. So the macro |
| 2054 | * works on both utf8 and non-utf8 representations. */ |
| 2055 | return isASCII(*p); |
| 2056 | } |
| 2057 | |
| 2058 | bool |
| 2059 | Perl_is_utf8_blank(pTHX_ const U8 *p) |
| 2060 | { |
| 2061 | dVAR; |
| 2062 | |
| 2063 | PERL_ARGS_ASSERT_IS_UTF8_BLANK; |
| 2064 | |
| 2065 | return is_utf8_common(p, &PL_utf8_blank, "XPosixBlank"); |
| 2066 | } |
| 2067 | |
| 2068 | bool |
| 2069 | Perl_is_utf8_space(pTHX_ const U8 *p) |
| 2070 | { |
| 2071 | dVAR; |
| 2072 | |
| 2073 | PERL_ARGS_ASSERT_IS_UTF8_SPACE; |
| 2074 | |
| 2075 | return is_utf8_common(p, &PL_utf8_space, "IsXPerlSpace"); |
| 2076 | } |
| 2077 | |
| 2078 | bool |
| 2079 | Perl_is_utf8_perl_space(pTHX_ const U8 *p) |
| 2080 | { |
| 2081 | dVAR; |
| 2082 | |
| 2083 | PERL_ARGS_ASSERT_IS_UTF8_PERL_SPACE; |
| 2084 | |
| 2085 | /* Only true if is an ASCII space-like character, and ASCII is invariant |
| 2086 | * under utf8, so can just use the macro */ |
| 2087 | return isSPACE_A(*p); |
| 2088 | } |
| 2089 | |
| 2090 | bool |
| 2091 | Perl_is_utf8_perl_word(pTHX_ const U8 *p) |
| 2092 | { |
| 2093 | dVAR; |
| 2094 | |
| 2095 | PERL_ARGS_ASSERT_IS_UTF8_PERL_WORD; |
| 2096 | |
| 2097 | /* Only true if is an ASCII word character, and ASCII is invariant |
| 2098 | * under utf8, so can just use the macro */ |
| 2099 | return isWORDCHAR_A(*p); |
| 2100 | } |
| 2101 | |
| 2102 | bool |
| 2103 | Perl_is_utf8_digit(pTHX_ const U8 *p) |
| 2104 | { |
| 2105 | dVAR; |
| 2106 | |
| 2107 | PERL_ARGS_ASSERT_IS_UTF8_DIGIT; |
| 2108 | |
| 2109 | return is_utf8_common(p, &PL_utf8_digit, "IsDigit"); |
| 2110 | } |
| 2111 | |
| 2112 | bool |
| 2113 | Perl_is_utf8_posix_digit(pTHX_ const U8 *p) |
| 2114 | { |
| 2115 | dVAR; |
| 2116 | |
| 2117 | PERL_ARGS_ASSERT_IS_UTF8_POSIX_DIGIT; |
| 2118 | |
| 2119 | /* Only true if is an ASCII digit character, and ASCII is invariant |
| 2120 | * under utf8, so can just use the macro */ |
| 2121 | return isDIGIT_A(*p); |
| 2122 | } |
| 2123 | |
| 2124 | bool |
| 2125 | Perl_is_utf8_upper(pTHX_ const U8 *p) |
| 2126 | { |
| 2127 | dVAR; |
| 2128 | |
| 2129 | PERL_ARGS_ASSERT_IS_UTF8_UPPER; |
| 2130 | |
| 2131 | return is_utf8_common(p, &PL_utf8_upper, "IsUppercase"); |
| 2132 | } |
| 2133 | |
| 2134 | bool |
| 2135 | Perl_is_utf8_lower(pTHX_ const U8 *p) |
| 2136 | { |
| 2137 | dVAR; |
| 2138 | |
| 2139 | PERL_ARGS_ASSERT_IS_UTF8_LOWER; |
| 2140 | |
| 2141 | return is_utf8_common(p, &PL_utf8_lower, "IsLowercase"); |
| 2142 | } |
| 2143 | |
| 2144 | bool |
| 2145 | Perl_is_utf8_cntrl(pTHX_ const U8 *p) |
| 2146 | { |
| 2147 | dVAR; |
| 2148 | |
| 2149 | PERL_ARGS_ASSERT_IS_UTF8_CNTRL; |
| 2150 | |
| 2151 | if (isASCII(*p)) { |
| 2152 | return isCNTRL_A(*p); |
| 2153 | } |
| 2154 | |
| 2155 | /* All controls are in Latin1 */ |
| 2156 | if (! UTF8_IS_DOWNGRADEABLE_START(*p)) { |
| 2157 | return 0; |
| 2158 | } |
| 2159 | return isCNTRL_L1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1))); |
| 2160 | } |
| 2161 | |
| 2162 | bool |
| 2163 | Perl_is_utf8_graph(pTHX_ const U8 *p) |
| 2164 | { |
| 2165 | dVAR; |
| 2166 | |
| 2167 | PERL_ARGS_ASSERT_IS_UTF8_GRAPH; |
| 2168 | |
| 2169 | return is_utf8_common(p, &PL_utf8_graph, "IsGraph"); |
| 2170 | } |
| 2171 | |
| 2172 | bool |
| 2173 | Perl_is_utf8_print(pTHX_ const U8 *p) |
| 2174 | { |
| 2175 | dVAR; |
| 2176 | |
| 2177 | PERL_ARGS_ASSERT_IS_UTF8_PRINT; |
| 2178 | |
| 2179 | return is_utf8_common(p, &PL_utf8_print, "IsPrint"); |
| 2180 | } |
| 2181 | |
| 2182 | bool |
| 2183 | Perl_is_utf8_punct(pTHX_ const U8 *p) |
| 2184 | { |
| 2185 | dVAR; |
| 2186 | |
| 2187 | PERL_ARGS_ASSERT_IS_UTF8_PUNCT; |
| 2188 | |
| 2189 | return is_utf8_common(p, &PL_utf8_punct, "IsPunct"); |
| 2190 | } |
| 2191 | |
| 2192 | bool |
| 2193 | Perl_is_utf8_xdigit(pTHX_ const U8 *p) |
| 2194 | { |
| 2195 | dVAR; |
| 2196 | |
| 2197 | PERL_ARGS_ASSERT_IS_UTF8_XDIGIT; |
| 2198 | |
| 2199 | return is_utf8_common(p, &PL_utf8_xdigit, "IsXDigit"); |
| 2200 | } |
| 2201 | |
| 2202 | bool |
| 2203 | Perl_is_utf8_mark(pTHX_ const U8 *p) |
| 2204 | { |
| 2205 | dVAR; |
| 2206 | |
| 2207 | PERL_ARGS_ASSERT_IS_UTF8_MARK; |
| 2208 | |
| 2209 | return is_utf8_common(p, &PL_utf8_mark, "IsM"); |
| 2210 | } |
| 2211 | |
| 2212 | bool |
| 2213 | Perl_is_utf8_X_regular_begin(pTHX_ const U8 *p) |
| 2214 | { |
| 2215 | dVAR; |
| 2216 | |
| 2217 | PERL_ARGS_ASSERT_IS_UTF8_X_REGULAR_BEGIN; |
| 2218 | |
| 2219 | return is_utf8_common(p, &PL_utf8_X_regular_begin, "_X_Regular_Begin"); |
| 2220 | } |
| 2221 | |
| 2222 | bool |
| 2223 | Perl_is_utf8_X_extend(pTHX_ const U8 *p) |
| 2224 | { |
| 2225 | dVAR; |
| 2226 | |
| 2227 | PERL_ARGS_ASSERT_IS_UTF8_X_EXTEND; |
| 2228 | |
| 2229 | return is_utf8_common(p, &PL_utf8_X_extend, "_X_Extend"); |
| 2230 | } |
| 2231 | |
| 2232 | bool |
| 2233 | Perl_is_utf8_X_prepend(pTHX_ const U8 *p) |
| 2234 | { |
| 2235 | /* If no code points in the Unicode version being worked on match |
| 2236 | * GCB=Prepend, this will set PL_utf8_X_prepend to &PL_sv_undef during its |
| 2237 | * first call. Otherwise, it will set it to a swash created for it. |
| 2238 | * swash_fetch() hence can't be used without checking first if it is valid |
| 2239 | * to do so. */ |
| 2240 | |
| 2241 | dVAR; |
| 2242 | bool initialized = cBOOL(PL_utf8_X_prepend); |
| 2243 | bool ret; |
| 2244 | |
| 2245 | PERL_ARGS_ASSERT_IS_UTF8_X_PREPEND; |
| 2246 | |
| 2247 | if (PL_utf8_X_prepend == &PL_sv_undef) { |
| 2248 | return FALSE; |
| 2249 | } |
| 2250 | |
| 2251 | if ((ret = is_utf8_common(p, &PL_utf8_X_prepend, "_X_GCB_Prepend")) |
| 2252 | || initialized) |
| 2253 | { |
| 2254 | return ret; |
| 2255 | } |
| 2256 | |
| 2257 | /* Here the code point being checked was not a prepend, and we hadn't |
| 2258 | * initialized PL_utf8_X_prepend, so we don't know if it is just this |
| 2259 | * particular input code point that didn't match, or if the table is |
| 2260 | * completely empty. The is_utf8_common() call did the initialization, so |
| 2261 | * we can inspect the swash's inversion list to find out. If there are no |
| 2262 | * elements in its inversion list, it's empty, and nothing will ever match, |
| 2263 | * so set things up so we can skip the check in future calls. */ |
| 2264 | if (_invlist_len(_get_swash_invlist(PL_utf8_X_prepend)) == 0) { |
| 2265 | SvREFCNT_dec(PL_utf8_X_prepend); |
| 2266 | PL_utf8_X_prepend = &PL_sv_undef; |
| 2267 | } |
| 2268 | |
| 2269 | return FALSE; |
| 2270 | } |
| 2271 | |
| 2272 | bool |
| 2273 | Perl_is_utf8_X_special_begin(pTHX_ const U8 *p) |
| 2274 | { |
| 2275 | dVAR; |
| 2276 | |
| 2277 | PERL_ARGS_ASSERT_IS_UTF8_X_SPECIAL_BEGIN; |
| 2278 | |
| 2279 | return is_utf8_common(p, &PL_utf8_X_special_begin, "_X_Special_Begin"); |
| 2280 | } |
| 2281 | |
| 2282 | bool |
| 2283 | Perl_is_utf8_X_L(pTHX_ const U8 *p) |
| 2284 | { |
| 2285 | dVAR; |
| 2286 | |
| 2287 | PERL_ARGS_ASSERT_IS_UTF8_X_L; |
| 2288 | |
| 2289 | return is_utf8_common(p, &PL_utf8_X_L, "_X_GCB_L"); |
| 2290 | } |
| 2291 | |
| 2292 | bool |
| 2293 | Perl_is_utf8_X_RI(pTHX_ const U8 *p) |
| 2294 | { |
| 2295 | dVAR; |
| 2296 | |
| 2297 | PERL_ARGS_ASSERT_IS_UTF8_X_RI; |
| 2298 | |
| 2299 | return is_utf8_common(p, &PL_utf8_X_RI, "_X_RI"); |
| 2300 | } |
| 2301 | |
| 2302 | /* These constants are for finding GCB=LV and GCB=LVT. These are for the |
| 2303 | * pre-composed Hangul syllables, which are all in a contiguous block and |
| 2304 | * arranged there in such a way so as to facilitate alorithmic determination of |
| 2305 | * their characteristics. As such, they don't need a swash, but can be |
| 2306 | * determined by simple arithmetic. Almost all are GCB=LVT, but every 28th one |
| 2307 | * is a GCB=LV */ |
| 2308 | #define SBASE 0xAC00 /* Start of block */ |
| 2309 | #define SCount 11172 /* Length of block */ |
| 2310 | #define TCount 28 |
| 2311 | |
| 2312 | #if 0 /* This routine is not currently used */ |
| 2313 | bool |
| 2314 | Perl_is_utf8_X_LV(pTHX_ const U8 *p) |
| 2315 | { |
| 2316 | /* Unlike most other similarly named routines here, this does not create a |
| 2317 | * swash, so swash_fetch() cannot be used on PL_utf8_X_LV. */ |
| 2318 | |
| 2319 | dVAR; |
| 2320 | |
| 2321 | UV cp = valid_utf8_to_uvchr(p, NULL); |
| 2322 | |
| 2323 | PERL_ARGS_ASSERT_IS_UTF8_X_LV; |
| 2324 | |
| 2325 | /* The earliest Unicode releases did not have these precomposed Hangul |
| 2326 | * syllables. Set to point to undef in that case, so will return false on |
| 2327 | * every call */ |
| 2328 | if (! PL_utf8_X_LV) { /* Set up if this is the first time called */ |
| 2329 | PL_utf8_X_LV = swash_init("utf8", "_X_GCB_LV", &PL_sv_undef, 1, 0); |
| 2330 | if (_invlist_len(_get_swash_invlist(PL_utf8_X_LV)) == 0) { |
| 2331 | SvREFCNT_dec(PL_utf8_X_LV); |
| 2332 | PL_utf8_X_LV = &PL_sv_undef; |
| 2333 | } |
| 2334 | } |
| 2335 | |
| 2336 | return (PL_utf8_X_LV != &PL_sv_undef |
| 2337 | && cp >= SBASE && cp < SBASE + SCount |
| 2338 | && (cp - SBASE) % TCount == 0); /* Only every TCount one is LV */ |
| 2339 | } |
| 2340 | #endif |
| 2341 | |
| 2342 | bool |
| 2343 | Perl_is_utf8_X_LVT(pTHX_ const U8 *p) |
| 2344 | { |
| 2345 | /* Unlike most other similarly named routines here, this does not create a |
| 2346 | * swash, so swash_fetch() cannot be used on PL_utf8_X_LVT. */ |
| 2347 | |
| 2348 | dVAR; |
| 2349 | |
| 2350 | UV cp = valid_utf8_to_uvchr(p, NULL); |
| 2351 | |
| 2352 | PERL_ARGS_ASSERT_IS_UTF8_X_LVT; |
| 2353 | |
| 2354 | /* The earliest Unicode releases did not have these precomposed Hangul |
| 2355 | * syllables. Set to point to undef in that case, so will return false on |
| 2356 | * every call */ |
| 2357 | if (! PL_utf8_X_LVT) { /* Set up if this is the first time called */ |
| 2358 | PL_utf8_X_LVT = swash_init("utf8", "_X_GCB_LVT", &PL_sv_undef, 1, 0); |
| 2359 | if (_invlist_len(_get_swash_invlist(PL_utf8_X_LVT)) == 0) { |
| 2360 | SvREFCNT_dec(PL_utf8_X_LVT); |
| 2361 | PL_utf8_X_LVT = &PL_sv_undef; |
| 2362 | } |
| 2363 | } |
| 2364 | |
| 2365 | return (PL_utf8_X_LVT != &PL_sv_undef |
| 2366 | && cp >= SBASE && cp < SBASE + SCount |
| 2367 | && (cp - SBASE) % TCount != 0); /* All but every TCount one is LV */ |
| 2368 | } |
| 2369 | |
| 2370 | bool |
| 2371 | Perl_is_utf8_X_T(pTHX_ const U8 *p) |
| 2372 | { |
| 2373 | dVAR; |
| 2374 | |
| 2375 | PERL_ARGS_ASSERT_IS_UTF8_X_T; |
| 2376 | |
| 2377 | return is_utf8_common(p, &PL_utf8_X_T, "_X_GCB_T"); |
| 2378 | } |
| 2379 | |
| 2380 | bool |
| 2381 | Perl_is_utf8_X_V(pTHX_ const U8 *p) |
| 2382 | { |
| 2383 | dVAR; |
| 2384 | |
| 2385 | PERL_ARGS_ASSERT_IS_UTF8_X_V; |
| 2386 | |
| 2387 | return is_utf8_common(p, &PL_utf8_X_V, "_X_GCB_V"); |
| 2388 | } |
| 2389 | |
| 2390 | bool |
| 2391 | Perl_is_utf8_X_LV_LVT_V(pTHX_ const U8 *p) |
| 2392 | { |
| 2393 | dVAR; |
| 2394 | |
| 2395 | PERL_ARGS_ASSERT_IS_UTF8_X_LV_LVT_V; |
| 2396 | |
| 2397 | return is_utf8_common(p, &PL_utf8_X_LV_LVT_V, "_X_LV_LVT_V"); |
| 2398 | } |
| 2399 | |
| 2400 | bool |
| 2401 | Perl__is_utf8_quotemeta(pTHX_ const U8 *p) |
| 2402 | { |
| 2403 | /* For exclusive use of pp_quotemeta() */ |
| 2404 | |
| 2405 | dVAR; |
| 2406 | |
| 2407 | PERL_ARGS_ASSERT__IS_UTF8_QUOTEMETA; |
| 2408 | |
| 2409 | return is_utf8_common(p, &PL_utf8_quotemeta, "_Perl_Quotemeta"); |
| 2410 | } |
| 2411 | |
| 2412 | /* |
| 2413 | =for apidoc to_utf8_case |
| 2414 | |
| 2415 | The C<p> contains the pointer to the UTF-8 string encoding |
| 2416 | the character that is being converted. This routine assumes that the character |
| 2417 | at C<p> is well-formed. |
| 2418 | |
| 2419 | The C<ustrp> is a pointer to the character buffer to put the |
| 2420 | conversion result to. The C<lenp> is a pointer to the length |
| 2421 | of the result. |
| 2422 | |
| 2423 | The C<swashp> is a pointer to the swash to use. |
| 2424 | |
| 2425 | Both the special and normal mappings are stored in F<lib/unicore/To/Foo.pl>, |
| 2426 | and loaded by SWASHNEW, using F<lib/utf8_heavy.pl>. The C<special> (usually, |
| 2427 | but not always, a multicharacter mapping), is tried first. |
| 2428 | |
| 2429 | The C<special> is a string like "utf8::ToSpecLower", which means the |
| 2430 | hash %utf8::ToSpecLower. The access to the hash is through |
| 2431 | Perl_to_utf8_case(). |
| 2432 | |
| 2433 | The C<normal> is a string like "ToLower" which means the swash |
| 2434 | %utf8::ToLower. |
| 2435 | |
| 2436 | =cut */ |
| 2437 | |
| 2438 | UV |
| 2439 | Perl_to_utf8_case(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, |
| 2440 | SV **swashp, const char *normal, const char *special) |
| 2441 | { |
| 2442 | dVAR; |
| 2443 | U8 tmpbuf[UTF8_MAXBYTES_CASE+1]; |
| 2444 | STRLEN len = 0; |
| 2445 | const UV uv0 = valid_utf8_to_uvchr(p, NULL); |
| 2446 | /* The NATIVE_TO_UNI() and UNI_TO_NATIVE() mappings |
| 2447 | * are necessary in EBCDIC, they are redundant no-ops |
| 2448 | * in ASCII-ish platforms, and hopefully optimized away. */ |
| 2449 | const UV uv1 = NATIVE_TO_UNI(uv0); |
| 2450 | |
| 2451 | PERL_ARGS_ASSERT_TO_UTF8_CASE; |
| 2452 | |
| 2453 | /* Note that swash_fetch() doesn't output warnings for these because it |
| 2454 | * assumes we will */ |
| 2455 | if (uv1 >= UNICODE_SURROGATE_FIRST) { |
| 2456 | if (uv1 <= UNICODE_SURROGATE_LAST) { |
| 2457 | if (ckWARN_d(WARN_SURROGATE)) { |
| 2458 | const char* desc = (PL_op) ? OP_DESC(PL_op) : normal; |
| 2459 | Perl_warner(aTHX_ packWARN(WARN_SURROGATE), |
| 2460 | "Operation \"%s\" returns its argument for UTF-16 surrogate U+%04"UVXf"", desc, uv1); |
| 2461 | } |
| 2462 | } |
| 2463 | else if (UNICODE_IS_SUPER(uv1)) { |
| 2464 | if (ckWARN_d(WARN_NON_UNICODE)) { |
| 2465 | const char* desc = (PL_op) ? OP_DESC(PL_op) : normal; |
| 2466 | Perl_warner(aTHX_ packWARN(WARN_NON_UNICODE), |
| 2467 | "Operation \"%s\" returns its argument for non-Unicode code point 0x%04"UVXf"", desc, uv1); |
| 2468 | } |
| 2469 | } |
| 2470 | |
| 2471 | /* Note that non-characters are perfectly legal, so no warning should |
| 2472 | * be given */ |
| 2473 | } |
| 2474 | |
| 2475 | uvuni_to_utf8(tmpbuf, uv1); |
| 2476 | |
| 2477 | if (!*swashp) /* load on-demand */ |
| 2478 | *swashp = _core_swash_init("utf8", normal, &PL_sv_undef, 4, 0, NULL, NULL); |
| 2479 | |
| 2480 | if (special) { |
| 2481 | /* It might be "special" (sometimes, but not always, |
| 2482 | * a multicharacter mapping) */ |
| 2483 | HV * const hv = get_hv(special, 0); |
| 2484 | SV **svp; |
| 2485 | |
| 2486 | if (hv && |
| 2487 | (svp = hv_fetch(hv, (const char*)tmpbuf, UNISKIP(uv1), FALSE)) && |
| 2488 | (*svp)) { |
| 2489 | const char *s; |
| 2490 | |
| 2491 | s = SvPV_const(*svp, len); |
| 2492 | if (len == 1) |
| 2493 | len = uvuni_to_utf8(ustrp, NATIVE_TO_UNI(*(U8*)s)) - ustrp; |
| 2494 | else { |
| 2495 | #ifdef EBCDIC |
| 2496 | /* If we have EBCDIC we need to remap the characters |
| 2497 | * since any characters in the low 256 are Unicode |
| 2498 | * code points, not EBCDIC. */ |
| 2499 | U8 *t = (U8*)s, *tend = t + len, *d; |
| 2500 | |
| 2501 | d = tmpbuf; |
| 2502 | if (SvUTF8(*svp)) { |
| 2503 | STRLEN tlen = 0; |
| 2504 | |
| 2505 | while (t < tend) { |
| 2506 | const UV c = utf8_to_uvchr_buf(t, tend, &tlen); |
| 2507 | if (tlen > 0) { |
| 2508 | d = uvchr_to_utf8(d, UNI_TO_NATIVE(c)); |
| 2509 | t += tlen; |
| 2510 | } |
| 2511 | else |
| 2512 | break; |
| 2513 | } |
| 2514 | } |
| 2515 | else { |
| 2516 | while (t < tend) { |
| 2517 | d = uvchr_to_utf8(d, UNI_TO_NATIVE(*t)); |
| 2518 | t++; |
| 2519 | } |
| 2520 | } |
| 2521 | len = d - tmpbuf; |
| 2522 | Copy(tmpbuf, ustrp, len, U8); |
| 2523 | #else |
| 2524 | Copy(s, ustrp, len, U8); |
| 2525 | #endif |
| 2526 | } |
| 2527 | } |
| 2528 | } |
| 2529 | |
| 2530 | if (!len && *swashp) { |
| 2531 | const UV uv2 = swash_fetch(*swashp, tmpbuf, TRUE /* => is utf8 */); |
| 2532 | |
| 2533 | if (uv2) { |
| 2534 | /* It was "normal" (a single character mapping). */ |
| 2535 | const UV uv3 = UNI_TO_NATIVE(uv2); |
| 2536 | len = uvchr_to_utf8(ustrp, uv3) - ustrp; |
| 2537 | } |
| 2538 | } |
| 2539 | |
| 2540 | if (len) { |
| 2541 | if (lenp) { |
| 2542 | *lenp = len; |
| 2543 | } |
| 2544 | return valid_utf8_to_uvchr(ustrp, 0); |
| 2545 | } |
| 2546 | |
| 2547 | /* Here, there was no mapping defined, which means that the code point maps |
| 2548 | * to itself. Return the inputs */ |
| 2549 | len = UTF8SKIP(p); |
| 2550 | Copy(p, ustrp, len, U8); |
| 2551 | |
| 2552 | if (lenp) |
| 2553 | *lenp = len; |
| 2554 | |
| 2555 | return uv0; |
| 2556 | |
| 2557 | } |
| 2558 | |
| 2559 | STATIC UV |
| 2560 | S_check_locale_boundary_crossing(pTHX_ const U8* const p, const UV result, U8* const ustrp, STRLEN *lenp) |
| 2561 | { |
| 2562 | /* This is called when changing the case of a utf8-encoded character above |
| 2563 | * the Latin1 range, and the operation is in locale. If the result |
| 2564 | * contains a character that crosses the 255/256 boundary, disallow the |
| 2565 | * change, and return the original code point. See L<perlfunc/lc> for why; |
| 2566 | * |
| 2567 | * p points to the original string whose case was changed; assumed |
| 2568 | * by this routine to be well-formed |
| 2569 | * result the code point of the first character in the changed-case string |
| 2570 | * ustrp points to the changed-case string (<result> represents its first char) |
| 2571 | * lenp points to the length of <ustrp> */ |
| 2572 | |
| 2573 | UV original; /* To store the first code point of <p> */ |
| 2574 | |
| 2575 | PERL_ARGS_ASSERT_CHECK_LOCALE_BOUNDARY_CROSSING; |
| 2576 | |
| 2577 | assert(! UTF8_IS_INVARIANT(*p) && ! UTF8_IS_DOWNGRADEABLE_START(*p)); |
| 2578 | |
| 2579 | /* We know immediately if the first character in the string crosses the |
| 2580 | * boundary, so can skip */ |
| 2581 | if (result > 255) { |
| 2582 | |
| 2583 | /* Look at every character in the result; if any cross the |
| 2584 | * boundary, the whole thing is disallowed */ |
| 2585 | U8* s = ustrp + UTF8SKIP(ustrp); |
| 2586 | U8* e = ustrp + *lenp; |
| 2587 | while (s < e) { |
| 2588 | if (UTF8_IS_INVARIANT(*s) || UTF8_IS_DOWNGRADEABLE_START(*s)) |
| 2589 | { |
| 2590 | goto bad_crossing; |
| 2591 | } |
| 2592 | s += UTF8SKIP(s); |
| 2593 | } |
| 2594 | |
| 2595 | /* Here, no characters crossed, result is ok as-is */ |
| 2596 | return result; |
| 2597 | } |
| 2598 | |
| 2599 | bad_crossing: |
| 2600 | |
| 2601 | /* Failed, have to return the original */ |
| 2602 | original = valid_utf8_to_uvchr(p, lenp); |
| 2603 | Copy(p, ustrp, *lenp, char); |
| 2604 | return original; |
| 2605 | } |
| 2606 | |
| 2607 | /* |
| 2608 | =for apidoc to_utf8_upper |
| 2609 | |
| 2610 | Convert the UTF-8 encoded character at C<p> to its uppercase version and |
| 2611 | store that in UTF-8 in C<ustrp> and its length in bytes in C<lenp>. Note |
| 2612 | that the ustrp needs to be at least UTF8_MAXBYTES_CASE+1 bytes since |
| 2613 | the uppercase version may be longer than the original character. |
| 2614 | |
| 2615 | The first character of the uppercased version is returned |
| 2616 | (but note, as explained above, that there may be more.) |
| 2617 | |
| 2618 | The character at C<p> is assumed by this routine to be well-formed. |
| 2619 | |
| 2620 | =cut */ |
| 2621 | |
| 2622 | /* Not currently externally documented, and subject to change: |
| 2623 | * <flags> is set iff locale semantics are to be used for code points < 256 |
| 2624 | * <tainted_ptr> if non-null, *tainted_ptr will be set TRUE iff locale rules |
| 2625 | * were used in the calculation; otherwise unchanged. */ |
| 2626 | |
| 2627 | UV |
| 2628 | Perl__to_utf8_upper_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, const bool flags, bool* tainted_ptr) |
| 2629 | { |
| 2630 | dVAR; |
| 2631 | |
| 2632 | UV result; |
| 2633 | |
| 2634 | PERL_ARGS_ASSERT__TO_UTF8_UPPER_FLAGS; |
| 2635 | |
| 2636 | if (UTF8_IS_INVARIANT(*p)) { |
| 2637 | if (flags) { |
| 2638 | result = toUPPER_LC(*p); |
| 2639 | } |
| 2640 | else { |
| 2641 | return _to_upper_title_latin1(*p, ustrp, lenp, 'S'); |
| 2642 | } |
| 2643 | } |
| 2644 | else if UTF8_IS_DOWNGRADEABLE_START(*p) { |
| 2645 | if (flags) { |
| 2646 | result = toUPPER_LC(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1))); |
| 2647 | } |
| 2648 | else { |
| 2649 | return _to_upper_title_latin1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)), |
| 2650 | ustrp, lenp, 'S'); |
| 2651 | } |
| 2652 | } |
| 2653 | else { /* utf8, ord above 255 */ |
| 2654 | result = CALL_UPPER_CASE(p, ustrp, lenp); |
| 2655 | |
| 2656 | if (flags) { |
| 2657 | result = check_locale_boundary_crossing(p, result, ustrp, lenp); |
| 2658 | } |
| 2659 | return result; |
| 2660 | } |
| 2661 | |
| 2662 | /* Here, used locale rules. Convert back to utf8 */ |
| 2663 | if (UTF8_IS_INVARIANT(result)) { |
| 2664 | *ustrp = (U8) result; |
| 2665 | *lenp = 1; |
| 2666 | } |
| 2667 | else { |
| 2668 | *ustrp = UTF8_EIGHT_BIT_HI(result); |
| 2669 | *(ustrp + 1) = UTF8_EIGHT_BIT_LO(result); |
| 2670 | *lenp = 2; |
| 2671 | } |
| 2672 | |
| 2673 | if (tainted_ptr) { |
| 2674 | *tainted_ptr = TRUE; |
| 2675 | } |
| 2676 | return result; |
| 2677 | } |
| 2678 | |
| 2679 | /* |
| 2680 | =for apidoc to_utf8_title |
| 2681 | |
| 2682 | Convert the UTF-8 encoded character at C<p> to its titlecase version and |
| 2683 | store that in UTF-8 in C<ustrp> and its length in bytes in C<lenp>. Note |
| 2684 | that the C<ustrp> needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the |
| 2685 | titlecase version may be longer than the original character. |
| 2686 | |
| 2687 | The first character of the titlecased version is returned |
| 2688 | (but note, as explained above, that there may be more.) |
| 2689 | |
| 2690 | The character at C<p> is assumed by this routine to be well-formed. |
| 2691 | |
| 2692 | =cut */ |
| 2693 | |
| 2694 | /* Not currently externally documented, and subject to change: |
| 2695 | * <flags> is set iff locale semantics are to be used for code points < 256 |
| 2696 | * Since titlecase is not defined in POSIX, uppercase is used instead |
| 2697 | * for these/ |
| 2698 | * <tainted_ptr> if non-null, *tainted_ptr will be set TRUE iff locale rules |
| 2699 | * were used in the calculation; otherwise unchanged. */ |
| 2700 | |
| 2701 | UV |
| 2702 | Perl__to_utf8_title_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, const bool flags, bool* tainted_ptr) |
| 2703 | { |
| 2704 | dVAR; |
| 2705 | |
| 2706 | UV result; |
| 2707 | |
| 2708 | PERL_ARGS_ASSERT__TO_UTF8_TITLE_FLAGS; |
| 2709 | |
| 2710 | if (UTF8_IS_INVARIANT(*p)) { |
| 2711 | if (flags) { |
| 2712 | result = toUPPER_LC(*p); |
| 2713 | } |
| 2714 | else { |
| 2715 | return _to_upper_title_latin1(*p, ustrp, lenp, 's'); |
| 2716 | } |
| 2717 | } |
| 2718 | else if UTF8_IS_DOWNGRADEABLE_START(*p) { |
| 2719 | if (flags) { |
| 2720 | result = toUPPER_LC(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1))); |
| 2721 | } |
| 2722 | else { |
| 2723 | return _to_upper_title_latin1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)), |
| 2724 | ustrp, lenp, 's'); |
| 2725 | } |
| 2726 | } |
| 2727 | else { /* utf8, ord above 255 */ |
| 2728 | result = CALL_TITLE_CASE(p, ustrp, lenp); |
| 2729 | |
| 2730 | if (flags) { |
| 2731 | result = check_locale_boundary_crossing(p, result, ustrp, lenp); |
| 2732 | } |
| 2733 | return result; |
| 2734 | } |
| 2735 | |
| 2736 | /* Here, used locale rules. Convert back to utf8 */ |
| 2737 | if (UTF8_IS_INVARIANT(result)) { |
| 2738 | *ustrp = (U8) result; |
| 2739 | *lenp = 1; |
| 2740 | } |
| 2741 | else { |
| 2742 | *ustrp = UTF8_EIGHT_BIT_HI(result); |
| 2743 | *(ustrp + 1) = UTF8_EIGHT_BIT_LO(result); |
| 2744 | *lenp = 2; |
| 2745 | } |
| 2746 | |
| 2747 | if (tainted_ptr) { |
| 2748 | *tainted_ptr = TRUE; |
| 2749 | } |
| 2750 | return result; |
| 2751 | } |
| 2752 | |
| 2753 | /* |
| 2754 | =for apidoc to_utf8_lower |
| 2755 | |
| 2756 | Convert the UTF-8 encoded character at C<p> to its lowercase version and |
| 2757 | store that in UTF-8 in ustrp and its length in bytes in C<lenp>. Note |
| 2758 | that the C<ustrp> needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the |
| 2759 | lowercase version may be longer than the original character. |
| 2760 | |
| 2761 | The first character of the lowercased version is returned |
| 2762 | (but note, as explained above, that there may be more.) |
| 2763 | |
| 2764 | The character at C<p> is assumed by this routine to be well-formed. |
| 2765 | |
| 2766 | =cut */ |
| 2767 | |
| 2768 | /* Not currently externally documented, and subject to change: |
| 2769 | * <flags> is set iff locale semantics are to be used for code points < 256 |
| 2770 | * <tainted_ptr> if non-null, *tainted_ptr will be set TRUE iff locale rules |
| 2771 | * were used in the calculation; otherwise unchanged. */ |
| 2772 | |
| 2773 | UV |
| 2774 | Perl__to_utf8_lower_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, const bool flags, bool* tainted_ptr) |
| 2775 | { |
| 2776 | UV result; |
| 2777 | |
| 2778 | dVAR; |
| 2779 | |
| 2780 | PERL_ARGS_ASSERT__TO_UTF8_LOWER_FLAGS; |
| 2781 | |
| 2782 | if (UTF8_IS_INVARIANT(*p)) { |
| 2783 | if (flags) { |
| 2784 | result = toLOWER_LC(*p); |
| 2785 | } |
| 2786 | else { |
| 2787 | return to_lower_latin1(*p, ustrp, lenp); |
| 2788 | } |
| 2789 | } |
| 2790 | else if UTF8_IS_DOWNGRADEABLE_START(*p) { |
| 2791 | if (flags) { |
| 2792 | result = toLOWER_LC(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1))); |
| 2793 | } |
| 2794 | else { |
| 2795 | return to_lower_latin1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)), |
| 2796 | ustrp, lenp); |
| 2797 | } |
| 2798 | } |
| 2799 | else { /* utf8, ord above 255 */ |
| 2800 | result = CALL_LOWER_CASE(p, ustrp, lenp); |
| 2801 | |
| 2802 | if (flags) { |
| 2803 | result = check_locale_boundary_crossing(p, result, ustrp, lenp); |
| 2804 | } |
| 2805 | |
| 2806 | return result; |
| 2807 | } |
| 2808 | |
| 2809 | /* Here, used locale rules. Convert back to utf8 */ |
| 2810 | if (UTF8_IS_INVARIANT(result)) { |
| 2811 | *ustrp = (U8) result; |
| 2812 | *lenp = 1; |
| 2813 | } |
| 2814 | else { |
| 2815 | *ustrp = UTF8_EIGHT_BIT_HI(result); |
| 2816 | *(ustrp + 1) = UTF8_EIGHT_BIT_LO(result); |
| 2817 | *lenp = 2; |
| 2818 | } |
| 2819 | |
| 2820 | if (tainted_ptr) { |
| 2821 | *tainted_ptr = TRUE; |
| 2822 | } |
| 2823 | return result; |
| 2824 | } |
| 2825 | |
| 2826 | /* |
| 2827 | =for apidoc to_utf8_fold |
| 2828 | |
| 2829 | Convert the UTF-8 encoded character at C<p> to its foldcase version and |
| 2830 | store that in UTF-8 in C<ustrp> and its length in bytes in C<lenp>. Note |
| 2831 | that the C<ustrp> needs to be at least UTF8_MAXBYTES_CASE+1 bytes since the |
| 2832 | foldcase version may be longer than the original character (up to |
| 2833 | three characters). |
| 2834 | |
| 2835 | The first character of the foldcased version is returned |
| 2836 | (but note, as explained above, that there may be more.) |
| 2837 | |
| 2838 | The character at C<p> is assumed by this routine to be well-formed. |
| 2839 | |
| 2840 | =cut */ |
| 2841 | |
| 2842 | /* Not currently externally documented, and subject to change, |
| 2843 | * in <flags> |
| 2844 | * bit FOLD_FLAGS_LOCALE is set iff locale semantics are to be used for code |
| 2845 | * points < 256. Since foldcase is not defined in |
| 2846 | * POSIX, lowercase is used instead |
| 2847 | * bit FOLD_FLAGS_FULL is set iff full case folds are to be used; |
| 2848 | * otherwise simple folds |
| 2849 | * bit FOLD_FLAGS_NOMIX_ASCII is set iff folds of non-ASCII to ASCII are |
| 2850 | * prohibited |
| 2851 | * <tainted_ptr> if non-null, *tainted_ptr will be set TRUE iff locale rules |
| 2852 | * were used in the calculation; otherwise unchanged. */ |
| 2853 | |
| 2854 | UV |
| 2855 | Perl__to_utf8_fold_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, U8 flags, bool* tainted_ptr) |
| 2856 | { |
| 2857 | dVAR; |
| 2858 | |
| 2859 | UV result; |
| 2860 | |
| 2861 | PERL_ARGS_ASSERT__TO_UTF8_FOLD_FLAGS; |
| 2862 | |
| 2863 | /* These are mutually exclusive */ |
| 2864 | assert (! ((flags & FOLD_FLAGS_LOCALE) && (flags & FOLD_FLAGS_NOMIX_ASCII))); |
| 2865 | |
| 2866 | assert(p != ustrp); /* Otherwise overwrites */ |
| 2867 | |
| 2868 | if (UTF8_IS_INVARIANT(*p)) { |
| 2869 | if (flags & FOLD_FLAGS_LOCALE) { |
| 2870 | result = toLOWER_LC(*p); |
| 2871 | } |
| 2872 | else { |
| 2873 | return _to_fold_latin1(*p, ustrp, lenp, |
| 2874 | cBOOL(flags & FOLD_FLAGS_FULL)); |
| 2875 | } |
| 2876 | } |
| 2877 | else if UTF8_IS_DOWNGRADEABLE_START(*p) { |
| 2878 | if (flags & FOLD_FLAGS_LOCALE) { |
| 2879 | result = toLOWER_LC(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1))); |
| 2880 | } |
| 2881 | else { |
| 2882 | return _to_fold_latin1(TWO_BYTE_UTF8_TO_UNI(*p, *(p+1)), |
| 2883 | ustrp, lenp, |
| 2884 | cBOOL((flags & FOLD_FLAGS_FULL |
| 2885 | /* If ASCII safe, don't allow full |
| 2886 | * folding, as that could include SHARP |
| 2887 | * S => ss; otherwise there is no |
| 2888 | * crossing of ascii/non-ascii in the |
| 2889 | * latin1 range */ |
| 2890 | && ! (flags & FOLD_FLAGS_NOMIX_ASCII)))); |
| 2891 | } |
| 2892 | } |
| 2893 | else { /* utf8, ord above 255 */ |
| 2894 | result = CALL_FOLD_CASE(p, ustrp, lenp, flags & FOLD_FLAGS_FULL); |
| 2895 | |
| 2896 | if ((flags & FOLD_FLAGS_LOCALE)) { |
| 2897 | return check_locale_boundary_crossing(p, result, ustrp, lenp); |
| 2898 | } |
| 2899 | else if (! (flags & FOLD_FLAGS_NOMIX_ASCII)) { |
| 2900 | return result; |
| 2901 | } |
| 2902 | else { |
| 2903 | /* This is called when changing the case of a utf8-encoded |
| 2904 | * character above the Latin1 range, and the result should not |
| 2905 | * contain an ASCII character. */ |
| 2906 | |
| 2907 | UV original; /* To store the first code point of <p> */ |
| 2908 | |
| 2909 | /* Look at every character in the result; if any cross the |
| 2910 | * boundary, the whole thing is disallowed */ |
| 2911 | U8* s = ustrp; |
| 2912 | U8* e = ustrp + *lenp; |
| 2913 | while (s < e) { |
| 2914 | if (isASCII(*s)) { |
| 2915 | /* Crossed, have to return the original */ |
| 2916 | original = valid_utf8_to_uvchr(p, lenp); |
| 2917 | Copy(p, ustrp, *lenp, char); |
| 2918 | return original; |
| 2919 | } |
| 2920 | s += UTF8SKIP(s); |
| 2921 | } |
| 2922 | |
| 2923 | /* Here, no characters crossed, result is ok as-is */ |
| 2924 | return result; |
| 2925 | } |
| 2926 | } |
| 2927 | |
| 2928 | /* Here, used locale rules. Convert back to utf8 */ |
| 2929 | if (UTF8_IS_INVARIANT(result)) { |
| 2930 | *ustrp = (U8) result; |
| 2931 | *lenp = 1; |
| 2932 | } |
| 2933 | else { |
| 2934 | *ustrp = UTF8_EIGHT_BIT_HI(result); |
| 2935 | *(ustrp + 1) = UTF8_EIGHT_BIT_LO(result); |
| 2936 | *lenp = 2; |
| 2937 | } |
| 2938 | |
| 2939 | if (tainted_ptr) { |
| 2940 | *tainted_ptr = TRUE; |
| 2941 | } |
| 2942 | return result; |
| 2943 | } |
| 2944 | |
| 2945 | /* Note: |
| 2946 | * Returns a "swash" which is a hash described in utf8.c:Perl_swash_fetch(). |
| 2947 | * C<pkg> is a pointer to a package name for SWASHNEW, should be "utf8". |
| 2948 | * For other parameters, see utf8::SWASHNEW in lib/utf8_heavy.pl. |
| 2949 | */ |
| 2950 | |
| 2951 | SV* |
| 2952 | Perl_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv, I32 minbits, I32 none) |
| 2953 | { |
| 2954 | PERL_ARGS_ASSERT_SWASH_INIT; |
| 2955 | |
| 2956 | /* Returns a copy of a swash initiated by the called function. This is the |
| 2957 | * public interface, and returning a copy prevents others from doing |
| 2958 | * mischief on the original */ |
| 2959 | |
| 2960 | return newSVsv(_core_swash_init(pkg, name, listsv, minbits, none, NULL, NULL)); |
| 2961 | } |
| 2962 | |
| 2963 | SV* |
| 2964 | Perl__core_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv, I32 minbits, I32 none, SV* invlist, U8* const flags_p) |
| 2965 | { |
| 2966 | /* Initialize and return a swash, creating it if necessary. It does this |
| 2967 | * by calling utf8_heavy.pl in the general case. The returned value may be |
| 2968 | * the swash's inversion list instead if the input parameters allow it. |
| 2969 | * Which is returned should be immaterial to callers, as the only |
| 2970 | * operations permitted on a swash, swash_fetch() and |
| 2971 | * _get_swash_invlist(), handle both these transparently. |
| 2972 | * |
| 2973 | * This interface should only be used by functions that won't destroy or |
| 2974 | * adversely change the swash, as doing so affects all other uses of the |
| 2975 | * swash in the program; the general public should use 'Perl_swash_init' |
| 2976 | * instead. |
| 2977 | * |
| 2978 | * pkg is the name of the package that <name> should be in. |
| 2979 | * name is the name of the swash to find. Typically it is a Unicode |
| 2980 | * property name, including user-defined ones |
| 2981 | * listsv is a string to initialize the swash with. It must be of the form |
| 2982 | * documented as the subroutine return value in |
| 2983 | * L<perlunicode/User-Defined Character Properties> |
| 2984 | * minbits is the number of bits required to represent each data element. |
| 2985 | * It is '1' for binary properties. |
| 2986 | * none I (khw) do not understand this one, but it is used only in tr///. |
| 2987 | * invlist is an inversion list to initialize the swash with (or NULL) |
| 2988 | * flags_p if non-NULL is the address of various input and output flag bits |
| 2989 | * to the routine, as follows: ('I' means is input to the routine; |
| 2990 | * 'O' means output from the routine. Only flags marked O are |
| 2991 | * meaningful on return.) |
| 2992 | * _CORE_SWASH_INIT_USER_DEFINED_PROPERTY indicates if the swash |
| 2993 | * came from a user-defined property. (I O) |
| 2994 | * _CORE_SWASH_INIT_RETURN_IF_UNDEF indicates that instead of croaking |
| 2995 | * when the swash cannot be located, to simply return NULL. (I) |
| 2996 | * _CORE_SWASH_INIT_ACCEPT_INVLIST indicates that the caller will accept a |
| 2997 | * return of an inversion list instead of a swash hash if this routine |
| 2998 | * thinks that would result in faster execution of swash_fetch() later |
| 2999 | * on. (I) |
| 3000 | * |
| 3001 | * Thus there are three possible inputs to find the swash: <name>, |
| 3002 | * <listsv>, and <invlist>. At least one must be specified. The result |
| 3003 | * will be the union of the specified ones, although <listsv>'s various |
| 3004 | * actions can intersect, etc. what <name> gives. |
| 3005 | * |
| 3006 | * <invlist> is only valid for binary properties */ |
| 3007 | |
| 3008 | dVAR; |
| 3009 | SV* retval = &PL_sv_undef; |
| 3010 | HV* swash_hv = NULL; |
| 3011 | const int invlist_swash_boundary = |
| 3012 | (flags_p && *flags_p & _CORE_SWASH_INIT_ACCEPT_INVLIST) |
| 3013 | ? 512 /* Based on some benchmarking, but not extensive, see commit |
| 3014 | message */ |
| 3015 | : -1; /* Never return just an inversion list */ |
| 3016 | |
| 3017 | assert(listsv != &PL_sv_undef || strNE(name, "") || invlist); |
| 3018 | assert(! invlist || minbits == 1); |
| 3019 | |
| 3020 | /* If data was passed in to go out to utf8_heavy to find the swash of, do |
| 3021 | * so */ |
| 3022 | if (listsv != &PL_sv_undef || strNE(name, "")) { |
| 3023 | dSP; |
| 3024 | const size_t pkg_len = strlen(pkg); |
| 3025 | const size_t name_len = strlen(name); |
| 3026 | HV * const stash = gv_stashpvn(pkg, pkg_len, 0); |
| 3027 | SV* errsv_save; |
| 3028 | GV *method; |
| 3029 | |
| 3030 | PERL_ARGS_ASSERT__CORE_SWASH_INIT; |
| 3031 | |
| 3032 | PUSHSTACKi(PERLSI_MAGIC); |
| 3033 | ENTER; |
| 3034 | SAVEHINTS(); |
| 3035 | save_re_context(); |
| 3036 | /* We might get here via a subroutine signature which uses a utf8 |
| 3037 | * parameter name, at which point PL_subname will have been set |
| 3038 | * but not yet used. */ |
| 3039 | save_item(PL_subname); |
| 3040 | if (PL_parser && PL_parser->error_count) |
| 3041 | SAVEI8(PL_parser->error_count), PL_parser->error_count = 0; |
| 3042 | method = gv_fetchmeth(stash, "SWASHNEW", 8, -1); |
| 3043 | if (!method) { /* demand load utf8 */ |
| 3044 | ENTER; |
| 3045 | errsv_save = newSVsv(ERRSV); |
| 3046 | /* It is assumed that callers of this routine are not passing in |
| 3047 | * any user derived data. */ |
| 3048 | /* Need to do this after save_re_context() as it will set |
| 3049 | * PL_tainted to 1 while saving $1 etc (see the code after getrx: |
| 3050 | * in Perl_magic_get). Even line to create errsv_save can turn on |
| 3051 | * PL_tainted. */ |
| 3052 | SAVEBOOL(PL_tainted); |
| 3053 | PL_tainted = 0; |
| 3054 | Perl_load_module(aTHX_ PERL_LOADMOD_NOIMPORT, newSVpvn(pkg,pkg_len), |
| 3055 | NULL); |
| 3056 | if (!SvTRUE(ERRSV)) |
| 3057 | sv_setsv(ERRSV, errsv_save); |
| 3058 | SvREFCNT_dec(errsv_save); |
| 3059 | LEAVE; |
| 3060 | } |
| 3061 | SPAGAIN; |
| 3062 | PUSHMARK(SP); |
| 3063 | EXTEND(SP,5); |
| 3064 | mPUSHp(pkg, pkg_len); |
| 3065 | mPUSHp(name, name_len); |
| 3066 | PUSHs(listsv); |
| 3067 | mPUSHi(minbits); |
| 3068 | mPUSHi(none); |
| 3069 | PUTBACK; |
| 3070 | errsv_save = newSVsv(ERRSV); |
| 3071 | /* If we already have a pointer to the method, no need to use |
| 3072 | * call_method() to repeat the lookup. */ |
| 3073 | if (method ? call_sv(MUTABLE_SV(method), G_SCALAR) |
| 3074 | : call_sv(newSVpvs_flags("SWASHNEW", SVs_TEMP), G_SCALAR | G_METHOD)) |
| 3075 | { |
| 3076 | retval = *PL_stack_sp--; |
| 3077 | SvREFCNT_inc(retval); |
| 3078 | } |
| 3079 | if (!SvTRUE(ERRSV)) |
| 3080 | sv_setsv(ERRSV, errsv_save); |
| 3081 | SvREFCNT_dec(errsv_save); |
| 3082 | LEAVE; |
| 3083 | POPSTACK; |
| 3084 | if (IN_PERL_COMPILETIME) { |
| 3085 | CopHINTS_set(PL_curcop, PL_hints); |
| 3086 | } |
| 3087 | if (!SvROK(retval) || SvTYPE(SvRV(retval)) != SVt_PVHV) { |
| 3088 | if (SvPOK(retval)) |
| 3089 | |
| 3090 | /* If caller wants to handle missing properties, let them */ |
| 3091 | if (flags_p && *flags_p & _CORE_SWASH_INIT_RETURN_IF_UNDEF) { |
| 3092 | return NULL; |
| 3093 | } |
| 3094 | Perl_croak(aTHX_ |
| 3095 | "Can't find Unicode property definition \"%"SVf"\"", |
| 3096 | SVfARG(retval)); |
| 3097 | Perl_croak(aTHX_ "SWASHNEW didn't return an HV ref"); |
| 3098 | } |
| 3099 | } /* End of calling the module to find the swash */ |
| 3100 | |
| 3101 | /* If this operation fetched a swash, and we will need it later, get it */ |
| 3102 | if (retval != &PL_sv_undef |
| 3103 | && (minbits == 1 || (flags_p |
| 3104 | && ! (*flags_p |
| 3105 | & _CORE_SWASH_INIT_USER_DEFINED_PROPERTY)))) |
| 3106 | { |
| 3107 | swash_hv = MUTABLE_HV(SvRV(retval)); |
| 3108 | |
| 3109 | /* If we don't already know that there is a user-defined component to |
| 3110 | * this swash, and the user has indicated they wish to know if there is |
| 3111 | * one (by passing <flags_p>), find out */ |
| 3112 | if (flags_p && ! (*flags_p & _CORE_SWASH_INIT_USER_DEFINED_PROPERTY)) { |
| 3113 | SV** user_defined = hv_fetchs(swash_hv, "USER_DEFINED", FALSE); |
| 3114 | if (user_defined && SvUV(*user_defined)) { |
| 3115 | *flags_p |= _CORE_SWASH_INIT_USER_DEFINED_PROPERTY; |
| 3116 | } |
| 3117 | } |
| 3118 | } |
| 3119 | |
| 3120 | /* Make sure there is an inversion list for binary properties */ |
| 3121 | if (minbits == 1) { |
| 3122 | SV** swash_invlistsvp = NULL; |
| 3123 | SV* swash_invlist = NULL; |
| 3124 | bool invlist_in_swash_is_valid = FALSE; |
| 3125 | |
| 3126 | /* If this operation fetched a swash, get its already existing |
| 3127 | * inversion list, or create one for it */ |
| 3128 | |
| 3129 | if (swash_hv) { |
| 3130 | swash_invlistsvp = hv_fetchs(swash_hv, "V", FALSE); |
| 3131 | if (swash_invlistsvp) { |
| 3132 | swash_invlist = *swash_invlistsvp; |
| 3133 | invlist_in_swash_is_valid = TRUE; |
| 3134 | } |
| 3135 | else { |
| 3136 | swash_invlist = _swash_to_invlist(retval); |
| 3137 | } |
| 3138 | } |
| 3139 | |
| 3140 | /* If an inversion list was passed in, have to include it */ |
| 3141 | if (invlist) { |
| 3142 | |
| 3143 | /* Any fetched swash will by now have an inversion list in it; |
| 3144 | * otherwise <swash_invlist> will be NULL, indicating that we |
| 3145 | * didn't fetch a swash */ |
| 3146 | if (swash_invlist) { |
| 3147 | |
| 3148 | /* Add the passed-in inversion list, which invalidates the one |
| 3149 | * already stored in the swash */ |
| 3150 | invlist_in_swash_is_valid = FALSE; |
| 3151 | _invlist_union(invlist, swash_invlist, &swash_invlist); |
| 3152 | } |
| 3153 | else { |
| 3154 | |
| 3155 | /* Here, there is no swash already. Set up a minimal one, if |
| 3156 | * we are going to return a swash */ |
| 3157 | if ((int) _invlist_len(invlist) > invlist_swash_boundary) { |
| 3158 | swash_hv = newHV(); |
| 3159 | retval = newRV_inc(MUTABLE_SV(swash_hv)); |
| 3160 | } |
| 3161 | swash_invlist = invlist; |
| 3162 | } |
| 3163 | } |
| 3164 | |
| 3165 | /* Here, we have computed the union of all the passed-in data. It may |
| 3166 | * be that there was an inversion list in the swash which didn't get |
| 3167 | * touched; otherwise save the one computed one */ |
| 3168 | if (! invlist_in_swash_is_valid |
| 3169 | && (int) _invlist_len(swash_invlist) > invlist_swash_boundary) |
| 3170 | { |
| 3171 | if (! hv_stores(MUTABLE_HV(SvRV(retval)), "V", swash_invlist)) |
| 3172 | { |
| 3173 | Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed"); |
| 3174 | } |
| 3175 | } |
| 3176 | |
| 3177 | if ((int) _invlist_len(swash_invlist) <= invlist_swash_boundary) { |
| 3178 | SvREFCNT_dec(retval); |
| 3179 | retval = newRV_inc(swash_invlist); |
| 3180 | } |
| 3181 | } |
| 3182 | |
| 3183 | return retval; |
| 3184 | } |
| 3185 | |
| 3186 | |
| 3187 | /* This API is wrong for special case conversions since we may need to |
| 3188 | * return several Unicode characters for a single Unicode character |
| 3189 | * (see lib/unicore/SpecCase.txt) The SWASHGET in lib/utf8_heavy.pl is |
| 3190 | * the lower-level routine, and it is similarly broken for returning |
| 3191 | * multiple values. --jhi |
| 3192 | * For those, you should use to_utf8_case() instead */ |
| 3193 | /* Now SWASHGET is recasted into S_swatch_get in this file. */ |
| 3194 | |
| 3195 | /* Note: |
| 3196 | * Returns the value of property/mapping C<swash> for the first character |
| 3197 | * of the string C<ptr>. If C<do_utf8> is true, the string C<ptr> is |
| 3198 | * assumed to be in utf8. If C<do_utf8> is false, the string C<ptr> is |
| 3199 | * assumed to be in native 8-bit encoding. Caches the swatch in C<swash>. |
| 3200 | * |
| 3201 | * A "swash" is a hash which contains initially the keys/values set up by |
| 3202 | * SWASHNEW. The purpose is to be able to completely represent a Unicode |
| 3203 | * property for all possible code points. Things are stored in a compact form |
| 3204 | * (see utf8_heavy.pl) so that calculation is required to find the actual |
| 3205 | * property value for a given code point. As code points are looked up, new |
| 3206 | * key/value pairs are added to the hash, so that the calculation doesn't have |
| 3207 | * to ever be re-done. Further, each calculation is done, not just for the |
| 3208 | * desired one, but for a whole block of code points adjacent to that one. |
| 3209 | * For binary properties on ASCII machines, the block is usually for 64 code |
| 3210 | * points, starting with a code point evenly divisible by 64. Thus if the |
| 3211 | * property value for code point 257 is requested, the code goes out and |
| 3212 | * calculates the property values for all 64 code points between 256 and 319, |
| 3213 | * and stores these as a single 64-bit long bit vector, called a "swatch", |
| 3214 | * under the key for code point 256. The key is the UTF-8 encoding for code |
| 3215 | * point 256, minus the final byte. Thus, if the length of the UTF-8 encoding |
| 3216 | * for a code point is 13 bytes, the key will be 12 bytes long. If the value |
| 3217 | * for code point 258 is then requested, this code realizes that it would be |
| 3218 | * stored under the key for 256, and would find that value and extract the |
| 3219 | * relevant bit, offset from 256. |
| 3220 | * |
| 3221 | * Non-binary properties are stored in as many bits as necessary to represent |
| 3222 | * their values (32 currently, though the code is more general than that), not |
| 3223 | * as single bits, but the principal is the same: the value for each key is a |
| 3224 | * vector that encompasses the property values for all code points whose UTF-8 |
| 3225 | * representations are represented by the key. That is, for all code points |
| 3226 | * whose UTF-8 representations are length N bytes, and the key is the first N-1 |
| 3227 | * bytes of that. |
| 3228 | */ |
| 3229 | UV |
| 3230 | Perl_swash_fetch(pTHX_ SV *swash, const U8 *ptr, bool do_utf8) |
| 3231 | { |
| 3232 | dVAR; |
| 3233 | HV *const hv = MUTABLE_HV(SvRV(swash)); |
| 3234 | U32 klen; |
| 3235 | U32 off; |
| 3236 | STRLEN slen; |
| 3237 | STRLEN needents; |
| 3238 | const U8 *tmps = NULL; |
| 3239 | U32 bit; |
| 3240 | SV *swatch; |
| 3241 | U8 tmputf8[2]; |
| 3242 | const UV c = NATIVE_TO_ASCII(*ptr); |
| 3243 | |
| 3244 | PERL_ARGS_ASSERT_SWASH_FETCH; |
| 3245 | |
| 3246 | /* If it really isn't a hash, it isn't really swash; must be an inversion |
| 3247 | * list */ |
| 3248 | if (SvTYPE(hv) != SVt_PVHV) { |
| 3249 | return _invlist_contains_cp((SV*)hv, |
| 3250 | (do_utf8) |
| 3251 | ? valid_utf8_to_uvchr(ptr, NULL) |
| 3252 | : c); |
| 3253 | } |
| 3254 | |
| 3255 | /* Convert to utf8 if not already */ |
| 3256 | if (!do_utf8 && !UNI_IS_INVARIANT(c)) { |
| 3257 | tmputf8[0] = (U8)UTF8_EIGHT_BIT_HI(c); |
| 3258 | tmputf8[1] = (U8)UTF8_EIGHT_BIT_LO(c); |
| 3259 | ptr = tmputf8; |
| 3260 | } |
| 3261 | /* Given a UTF-X encoded char 0xAA..0xYY,0xZZ |
| 3262 | * then the "swatch" is a vec() for all the chars which start |
| 3263 | * with 0xAA..0xYY |
| 3264 | * So the key in the hash (klen) is length of encoded char -1 |
| 3265 | */ |
| 3266 | klen = UTF8SKIP(ptr) - 1; |
| 3267 | off = ptr[klen]; |
| 3268 | |
| 3269 | if (klen == 0) { |
| 3270 | /* If char is invariant then swatch is for all the invariant chars |
| 3271 | * In both UTF-8 and UTF-8-MOD that happens to be UTF_CONTINUATION_MARK |
| 3272 | */ |
| 3273 | needents = UTF_CONTINUATION_MARK; |
| 3274 | off = NATIVE_TO_UTF(ptr[klen]); |
| 3275 | } |
| 3276 | else { |
| 3277 | /* If char is encoded then swatch is for the prefix */ |
| 3278 | needents = (1 << UTF_ACCUMULATION_SHIFT); |
| 3279 | off = NATIVE_TO_UTF(ptr[klen]) & UTF_CONTINUATION_MASK; |
| 3280 | } |
| 3281 | |
| 3282 | /* |
| 3283 | * This single-entry cache saves about 1/3 of the utf8 overhead in test |
| 3284 | * suite. (That is, only 7-8% overall over just a hash cache. Still, |
| 3285 | * it's nothing to sniff at.) Pity we usually come through at least |
| 3286 | * two function calls to get here... |
| 3287 | * |
| 3288 | * NB: this code assumes that swatches are never modified, once generated! |
| 3289 | */ |
| 3290 | |
| 3291 | if (hv == PL_last_swash_hv && |
| 3292 | klen == PL_last_swash_klen && |
| 3293 | (!klen || memEQ((char *)ptr, (char *)PL_last_swash_key, klen)) ) |
| 3294 | { |
| 3295 | tmps = PL_last_swash_tmps; |
| 3296 | slen = PL_last_swash_slen; |
| 3297 | } |
| 3298 | else { |
| 3299 | /* Try our second-level swatch cache, kept in a hash. */ |
| 3300 | SV** svp = hv_fetch(hv, (const char*)ptr, klen, FALSE); |
| 3301 | |
| 3302 | /* If not cached, generate it via swatch_get */ |
| 3303 | if (!svp || !SvPOK(*svp) |
| 3304 | || !(tmps = (const U8*)SvPV_const(*svp, slen))) { |
| 3305 | /* We use utf8n_to_uvuni() as we want an index into |
| 3306 | Unicode tables, not a native character number. |
| 3307 | */ |
| 3308 | const UV code_point = utf8n_to_uvuni(ptr, UTF8_MAXBYTES, 0, |
| 3309 | ckWARN(WARN_UTF8) ? |
| 3310 | 0 : UTF8_ALLOW_ANY); |
| 3311 | swatch = swatch_get(swash, |
| 3312 | /* On EBCDIC & ~(0xA0-1) isn't a useful thing to do */ |
| 3313 | (klen) ? (code_point & ~((UV)needents - 1)) : 0, |
| 3314 | needents); |
| 3315 | |
| 3316 | if (IN_PERL_COMPILETIME) |
| 3317 | CopHINTS_set(PL_curcop, PL_hints); |
| 3318 | |
| 3319 | svp = hv_store(hv, (const char *)ptr, klen, swatch, 0); |
| 3320 | |
| 3321 | if (!svp || !(tmps = (U8*)SvPV(*svp, slen)) |
| 3322 | || (slen << 3) < needents) |
| 3323 | Perl_croak(aTHX_ "panic: swash_fetch got improper swatch, " |
| 3324 | "svp=%p, tmps=%p, slen=%"UVuf", needents=%"UVuf, |
| 3325 | svp, tmps, (UV)slen, (UV)needents); |
| 3326 | } |
| 3327 | |
| 3328 | PL_last_swash_hv = hv; |
| 3329 | assert(klen <= sizeof(PL_last_swash_key)); |
| 3330 | PL_last_swash_klen = (U8)klen; |
| 3331 | /* FIXME change interpvar.h? */ |
| 3332 | PL_last_swash_tmps = (U8 *) tmps; |
| 3333 | PL_last_swash_slen = slen; |
| 3334 | if (klen) |
| 3335 | Copy(ptr, PL_last_swash_key, klen, U8); |
| 3336 | } |
| 3337 | |
| 3338 | switch ((int)((slen << 3) / needents)) { |
| 3339 | case 1: |
| 3340 | bit = 1 << (off & 7); |
| 3341 | off >>= 3; |
| 3342 | return (tmps[off] & bit) != 0; |
| 3343 | case 8: |
| 3344 | return tmps[off]; |
| 3345 | case 16: |
| 3346 | off <<= 1; |
| 3347 | return (tmps[off] << 8) + tmps[off + 1] ; |
| 3348 | case 32: |
| 3349 | off <<= 2; |
| 3350 | return (tmps[off] << 24) + (tmps[off+1] << 16) + (tmps[off+2] << 8) + tmps[off + 3] ; |
| 3351 | } |
| 3352 | Perl_croak(aTHX_ "panic: swash_fetch got swatch of unexpected bit width, " |
| 3353 | "slen=%"UVuf", needents=%"UVuf, (UV)slen, (UV)needents); |
| 3354 | NORETURN_FUNCTION_END; |
| 3355 | } |
| 3356 | |
| 3357 | /* Read a single line of the main body of the swash input text. These are of |
| 3358 | * the form: |
| 3359 | * 0053 0056 0073 |
| 3360 | * where each number is hex. The first two numbers form the minimum and |
| 3361 | * maximum of a range, and the third is the value associated with the range. |
| 3362 | * Not all swashes should have a third number |
| 3363 | * |
| 3364 | * On input: l points to the beginning of the line to be examined; it points |
| 3365 | * to somewhere in the string of the whole input text, and is |
| 3366 | * terminated by a \n or the null string terminator. |
| 3367 | * lend points to the null terminator of that string |
| 3368 | * wants_value is non-zero if the swash expects a third number |
| 3369 | * typestr is the name of the swash's mapping, like 'ToLower' |
| 3370 | * On output: *min, *max, and *val are set to the values read from the line. |
| 3371 | * returns a pointer just beyond the line examined. If there was no |
| 3372 | * valid min number on the line, returns lend+1 |
| 3373 | */ |
| 3374 | |
| 3375 | STATIC U8* |
| 3376 | S_swash_scan_list_line(pTHX_ U8* l, U8* const lend, UV* min, UV* max, UV* val, |
| 3377 | const bool wants_value, const U8* const typestr) |
| 3378 | { |
| 3379 | const int typeto = typestr[0] == 'T' && typestr[1] == 'o'; |
| 3380 | STRLEN numlen; /* Length of the number */ |
| 3381 | I32 flags = PERL_SCAN_SILENT_ILLDIGIT |
| 3382 | | PERL_SCAN_DISALLOW_PREFIX |
| 3383 | | PERL_SCAN_SILENT_NON_PORTABLE; |
| 3384 | |
| 3385 | /* nl points to the next \n in the scan */ |
| 3386 | U8* const nl = (U8*)memchr(l, '\n', lend - l); |
| 3387 | |
| 3388 | /* Get the first number on the line: the range minimum */ |
| 3389 | numlen = lend - l; |
| 3390 | *min = grok_hex((char *)l, &numlen, &flags, NULL); |
| 3391 | if (numlen) /* If found a hex number, position past it */ |
| 3392 | l += numlen; |
| 3393 | else if (nl) { /* Else, go handle next line, if any */ |
| 3394 | return nl + 1; /* 1 is length of "\n" */ |
| 3395 | } |
| 3396 | else { /* Else, no next line */ |
| 3397 | return lend + 1; /* to LIST's end at which \n is not found */ |
| 3398 | } |
| 3399 | |
| 3400 | /* The max range value follows, separated by a BLANK */ |
| 3401 | if (isBLANK(*l)) { |
| 3402 | ++l; |
| 3403 | flags = PERL_SCAN_SILENT_ILLDIGIT |
| 3404 | | PERL_SCAN_DISALLOW_PREFIX |
| 3405 | | PERL_SCAN_SILENT_NON_PORTABLE; |
| 3406 | numlen = lend - l; |
| 3407 | *max = grok_hex((char *)l, &numlen, &flags, NULL); |
| 3408 | if (numlen) |
| 3409 | l += numlen; |
| 3410 | else /* If no value here, it is a single element range */ |
| 3411 | *max = *min; |
| 3412 | |
| 3413 | /* Non-binary tables have a third entry: what the first element of the |
| 3414 | * range maps to */ |
| 3415 | if (wants_value) { |
| 3416 | if (isBLANK(*l)) { |
| 3417 | ++l; |
| 3418 | |
| 3419 | /* The ToLc, etc table mappings are not in hex, and must be |
| 3420 | * corrected by adding the code point to them */ |
| 3421 | if (typeto) { |
| 3422 | char *after_strtol = (char *) lend; |
| 3423 | *val = Strtol((char *)l, &after_strtol, 10); |
| 3424 | l = (U8 *) after_strtol; |
| 3425 | } |
| 3426 | else { /* Other tables are in hex, and are the correct result |
| 3427 | without tweaking */ |
| 3428 | flags = PERL_SCAN_SILENT_ILLDIGIT |
| 3429 | | PERL_SCAN_DISALLOW_PREFIX |
| 3430 | | PERL_SCAN_SILENT_NON_PORTABLE; |
| 3431 | numlen = lend - l; |
| 3432 | *val = grok_hex((char *)l, &numlen, &flags, NULL); |
| 3433 | if (numlen) |
| 3434 | l += numlen; |
| 3435 | else |
| 3436 | *val = 0; |
| 3437 | } |
| 3438 | } |
| 3439 | else { |
| 3440 | *val = 0; |
| 3441 | if (typeto) { |
| 3442 | /* diag_listed_as: To%s: illegal mapping '%s' */ |
| 3443 | Perl_croak(aTHX_ "%s: illegal mapping '%s'", |
| 3444 | typestr, l); |
| 3445 | } |
| 3446 | } |
| 3447 | } |
| 3448 | else |
| 3449 | *val = 0; /* bits == 1, then any val should be ignored */ |
| 3450 | } |
| 3451 | else { /* Nothing following range min, should be single element with no |
| 3452 | mapping expected */ |
| 3453 | *max = *min; |
| 3454 | if (wants_value) { |
| 3455 | *val = 0; |
| 3456 | if (typeto) { |
| 3457 | /* diag_listed_as: To%s: illegal mapping '%s' */ |
| 3458 | Perl_croak(aTHX_ "%s: illegal mapping '%s'", typestr, l); |
| 3459 | } |
| 3460 | } |
| 3461 | else |
| 3462 | *val = 0; /* bits == 1, then val should be ignored */ |
| 3463 | } |
| 3464 | |
| 3465 | /* Position to next line if any, or EOF */ |
| 3466 | if (nl) |
| 3467 | l = nl + 1; |
| 3468 | else |
| 3469 | l = lend; |
| 3470 | |
| 3471 | return l; |
| 3472 | } |
| 3473 | |
| 3474 | /* Note: |
| 3475 | * Returns a swatch (a bit vector string) for a code point sequence |
| 3476 | * that starts from the value C<start> and comprises the number C<span>. |
| 3477 | * A C<swash> must be an object created by SWASHNEW (see lib/utf8_heavy.pl). |
| 3478 | * Should be used via swash_fetch, which will cache the swatch in C<swash>. |
| 3479 | */ |
| 3480 | STATIC SV* |
| 3481 | S_swatch_get(pTHX_ SV* swash, UV start, UV span) |
| 3482 | { |
| 3483 | SV *swatch; |
| 3484 | U8 *l, *lend, *x, *xend, *s, *send; |
| 3485 | STRLEN lcur, xcur, scur; |
| 3486 | HV *const hv = MUTABLE_HV(SvRV(swash)); |
| 3487 | SV** const invlistsvp = hv_fetchs(hv, "V", FALSE); |
| 3488 | |
| 3489 | SV** listsvp = NULL; /* The string containing the main body of the table */ |
| 3490 | SV** extssvp = NULL; |
| 3491 | SV** invert_it_svp = NULL; |
| 3492 | U8* typestr = NULL; |
| 3493 | STRLEN bits; |
| 3494 | STRLEN octets; /* if bits == 1, then octets == 0 */ |
| 3495 | UV none; |
| 3496 | UV end = start + span; |
| 3497 | |
| 3498 | if (invlistsvp == NULL) { |
| 3499 | SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE); |
| 3500 | SV** const nonesvp = hv_fetchs(hv, "NONE", FALSE); |
| 3501 | SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE); |
| 3502 | extssvp = hv_fetchs(hv, "EXTRAS", FALSE); |
| 3503 | listsvp = hv_fetchs(hv, "LIST", FALSE); |
| 3504 | invert_it_svp = hv_fetchs(hv, "INVERT_IT", FALSE); |
| 3505 | |
| 3506 | bits = SvUV(*bitssvp); |
| 3507 | none = SvUV(*nonesvp); |
| 3508 | typestr = (U8*)SvPV_nolen(*typesvp); |
| 3509 | } |
| 3510 | else { |
| 3511 | bits = 1; |
| 3512 | none = 0; |
| 3513 | } |
| 3514 | octets = bits >> 3; /* if bits == 1, then octets == 0 */ |
| 3515 | |
| 3516 | PERL_ARGS_ASSERT_SWATCH_GET; |
| 3517 | |
| 3518 | if (bits != 1 && bits != 8 && bits != 16 && bits != 32) { |
| 3519 | Perl_croak(aTHX_ "panic: swatch_get doesn't expect bits %"UVuf, |
| 3520 | (UV)bits); |
| 3521 | } |
| 3522 | |
| 3523 | /* If overflowed, use the max possible */ |
| 3524 | if (end < start) { |
| 3525 | end = UV_MAX; |
| 3526 | span = end - start; |
| 3527 | } |
| 3528 | |
| 3529 | /* create and initialize $swatch */ |
| 3530 | scur = octets ? (span * octets) : (span + 7) / 8; |
| 3531 | swatch = newSV(scur); |
| 3532 | SvPOK_on(swatch); |
| 3533 | s = (U8*)SvPVX(swatch); |
| 3534 | if (octets && none) { |
| 3535 | const U8* const e = s + scur; |
| 3536 | while (s < e) { |
| 3537 | if (bits == 8) |
| 3538 | *s++ = (U8)(none & 0xff); |
| 3539 | else if (bits == 16) { |
| 3540 | *s++ = (U8)((none >> 8) & 0xff); |
| 3541 | *s++ = (U8)( none & 0xff); |
| 3542 | } |
| 3543 | else if (bits == 32) { |
| 3544 | *s++ = (U8)((none >> 24) & 0xff); |
| 3545 | *s++ = (U8)((none >> 16) & 0xff); |
| 3546 | *s++ = (U8)((none >> 8) & 0xff); |
| 3547 | *s++ = (U8)( none & 0xff); |
| 3548 | } |
| 3549 | } |
| 3550 | *s = '\0'; |
| 3551 | } |
| 3552 | else { |
| 3553 | (void)memzero((U8*)s, scur + 1); |
| 3554 | } |
| 3555 | SvCUR_set(swatch, scur); |
| 3556 | s = (U8*)SvPVX(swatch); |
| 3557 | |
| 3558 | if (invlistsvp) { /* If has an inversion list set up use that */ |
| 3559 | _invlist_populate_swatch(*invlistsvp, start, end, s); |
| 3560 | return swatch; |
| 3561 | } |
| 3562 | |
| 3563 | /* read $swash->{LIST} */ |
| 3564 | l = (U8*)SvPV(*listsvp, lcur); |
| 3565 | lend = l + lcur; |
| 3566 | while (l < lend) { |
| 3567 | UV min, max, val, upper; |
| 3568 | l = S_swash_scan_list_line(aTHX_ l, lend, &min, &max, &val, |
| 3569 | cBOOL(octets), typestr); |
| 3570 | if (l > lend) { |
| 3571 | break; |
| 3572 | } |
| 3573 | |
| 3574 | /* If looking for something beyond this range, go try the next one */ |
| 3575 | if (max < start) |
| 3576 | continue; |
| 3577 | |
| 3578 | /* <end> is generally 1 beyond where we want to set things, but at the |
| 3579 | * platform's infinity, where we can't go any higher, we want to |
| 3580 | * include the code point at <end> */ |
| 3581 | upper = (max < end) |
| 3582 | ? max |
| 3583 | : (max != UV_MAX || end != UV_MAX) |
| 3584 | ? end - 1 |
| 3585 | : end; |
| 3586 | |
| 3587 | if (octets) { |
| 3588 | UV key; |
| 3589 | if (min < start) { |
| 3590 | if (!none || val < none) { |
| 3591 | val += start - min; |
| 3592 | } |
| 3593 | min = start; |
| 3594 | } |
| 3595 | for (key = min; key <= upper; key++) { |
| 3596 | STRLEN offset; |
| 3597 | /* offset must be non-negative (start <= min <= key < end) */ |
| 3598 | offset = octets * (key - start); |
| 3599 | if (bits == 8) |
| 3600 | s[offset] = (U8)(val & 0xff); |
| 3601 | else if (bits == 16) { |
| 3602 | s[offset ] = (U8)((val >> 8) & 0xff); |
| 3603 | s[offset + 1] = (U8)( val & 0xff); |
| 3604 | } |
| 3605 | else if (bits == 32) { |
| 3606 | s[offset ] = (U8)((val >> 24) & 0xff); |
| 3607 | s[offset + 1] = (U8)((val >> 16) & 0xff); |
| 3608 | s[offset + 2] = (U8)((val >> 8) & 0xff); |
| 3609 | s[offset + 3] = (U8)( val & 0xff); |
| 3610 | } |
| 3611 | |
| 3612 | if (!none || val < none) |
| 3613 | ++val; |
| 3614 | } |
| 3615 | } |
| 3616 | else { /* bits == 1, then val should be ignored */ |
| 3617 | UV key; |
| 3618 | if (min < start) |
| 3619 | min = start; |
| 3620 | |
| 3621 | for (key = min; key <= upper; key++) { |
| 3622 | const STRLEN offset = (STRLEN)(key - start); |
| 3623 | s[offset >> 3] |= 1 << (offset & 7); |
| 3624 | } |
| 3625 | } |
| 3626 | } /* while */ |
| 3627 | |
| 3628 | /* Invert if the data says it should be. Assumes that bits == 1 */ |
| 3629 | if (invert_it_svp && SvUV(*invert_it_svp)) { |
| 3630 | |
| 3631 | /* Unicode properties should come with all bits above PERL_UNICODE_MAX |
| 3632 | * be 0, and their inversion should also be 0, as we don't succeed any |
| 3633 | * Unicode property matches for non-Unicode code points */ |
| 3634 | if (start <= PERL_UNICODE_MAX) { |
| 3635 | |
| 3636 | /* The code below assumes that we never cross the |
| 3637 | * Unicode/above-Unicode boundary in a range, as otherwise we would |
| 3638 | * have to figure out where to stop flipping the bits. Since this |
| 3639 | * boundary is divisible by a large power of 2, and swatches comes |
| 3640 | * in small powers of 2, this should be a valid assumption */ |
| 3641 | assert(start + span - 1 <= PERL_UNICODE_MAX); |
| 3642 | |
| 3643 | send = s + scur; |
| 3644 | while (s < send) { |
| 3645 | *s = ~(*s); |
| 3646 | s++; |
| 3647 | } |
| 3648 | } |
| 3649 | } |
| 3650 | |
| 3651 | /* read $swash->{EXTRAS} |
| 3652 | * This code also copied to swash_to_invlist() below */ |
| 3653 | x = (U8*)SvPV(*extssvp, xcur); |
| 3654 | xend = x + xcur; |
| 3655 | while (x < xend) { |
| 3656 | STRLEN namelen; |
| 3657 | U8 *namestr; |
| 3658 | SV** othersvp; |
| 3659 | HV* otherhv; |
| 3660 | STRLEN otherbits; |
| 3661 | SV **otherbitssvp, *other; |
| 3662 | U8 *s, *o, *nl; |
| 3663 | STRLEN slen, olen; |
| 3664 | |
| 3665 | const U8 opc = *x++; |
| 3666 | if (opc == '\n') |
| 3667 | continue; |
| 3668 | |
| 3669 | nl = (U8*)memchr(x, '\n', xend - x); |
| 3670 | |
| 3671 | if (opc != '-' && opc != '+' && opc != '!' && opc != '&') { |
| 3672 | if (nl) { |
| 3673 | x = nl + 1; /* 1 is length of "\n" */ |
| 3674 | continue; |
| 3675 | } |
| 3676 | else { |
| 3677 | x = xend; /* to EXTRAS' end at which \n is not found */ |
| 3678 | break; |
| 3679 | } |
| 3680 | } |
| 3681 | |
| 3682 | namestr = x; |
| 3683 | if (nl) { |
| 3684 | namelen = nl - namestr; |
| 3685 | x = nl + 1; |
| 3686 | } |
| 3687 | else { |
| 3688 | namelen = xend - namestr; |
| 3689 | x = xend; |
| 3690 | } |
| 3691 | |
| 3692 | othersvp = hv_fetch(hv, (char *)namestr, namelen, FALSE); |
| 3693 | otherhv = MUTABLE_HV(SvRV(*othersvp)); |
| 3694 | otherbitssvp = hv_fetchs(otherhv, "BITS", FALSE); |
| 3695 | otherbits = (STRLEN)SvUV(*otherbitssvp); |
| 3696 | if (bits < otherbits) |
| 3697 | Perl_croak(aTHX_ "panic: swatch_get found swatch size mismatch, " |
| 3698 | "bits=%"UVuf", otherbits=%"UVuf, (UV)bits, (UV)otherbits); |
| 3699 | |
| 3700 | /* The "other" swatch must be destroyed after. */ |
| 3701 | other = swatch_get(*othersvp, start, span); |
| 3702 | o = (U8*)SvPV(other, olen); |
| 3703 | |
| 3704 | if (!olen) |
| 3705 | Perl_croak(aTHX_ "panic: swatch_get got improper swatch"); |
| 3706 | |
| 3707 | s = (U8*)SvPV(swatch, slen); |
| 3708 | if (bits == 1 && otherbits == 1) { |
| 3709 | if (slen != olen) |
| 3710 | Perl_croak(aTHX_ "panic: swatch_get found swatch length " |
| 3711 | "mismatch, slen=%"UVuf", olen=%"UVuf, |
| 3712 | (UV)slen, (UV)olen); |
| 3713 | |
| 3714 | switch (opc) { |
| 3715 | case '+': |
| 3716 | while (slen--) |
| 3717 | *s++ |= *o++; |
| 3718 | break; |
| 3719 | case '!': |
| 3720 | while (slen--) |
| 3721 | *s++ |= ~*o++; |
| 3722 | break; |
| 3723 | case '-': |
| 3724 | while (slen--) |
| 3725 | *s++ &= ~*o++; |
| 3726 | break; |
| 3727 | case '&': |
| 3728 | while (slen--) |
| 3729 | *s++ &= *o++; |
| 3730 | break; |
| 3731 | default: |
| 3732 | break; |
| 3733 | } |
| 3734 | } |
| 3735 | else { |
| 3736 | STRLEN otheroctets = otherbits >> 3; |
| 3737 | STRLEN offset = 0; |
| 3738 | U8* const send = s + slen; |
| 3739 | |
| 3740 | while (s < send) { |
| 3741 | UV otherval = 0; |
| 3742 | |
| 3743 | if (otherbits == 1) { |
| 3744 | otherval = (o[offset >> 3] >> (offset & 7)) & 1; |
| 3745 | ++offset; |
| 3746 | } |
| 3747 | else { |
| 3748 | STRLEN vlen = otheroctets; |
| 3749 | otherval = *o++; |
| 3750 | while (--vlen) { |
| 3751 | otherval <<= 8; |
| 3752 | otherval |= *o++; |
| 3753 | } |
| 3754 | } |
| 3755 | |
| 3756 | if (opc == '+' && otherval) |
| 3757 | NOOP; /* replace with otherval */ |
| 3758 | else if (opc == '!' && !otherval) |
| 3759 | otherval = 1; |
| 3760 | else if (opc == '-' && otherval) |
| 3761 | otherval = 0; |
| 3762 | else if (opc == '&' && !otherval) |
| 3763 | otherval = 0; |
| 3764 | else { |
| 3765 | s += octets; /* no replacement */ |
| 3766 | continue; |
| 3767 | } |
| 3768 | |
| 3769 | if (bits == 8) |
| 3770 | *s++ = (U8)( otherval & 0xff); |
| 3771 | else if (bits == 16) { |
| 3772 | *s++ = (U8)((otherval >> 8) & 0xff); |
| 3773 | *s++ = (U8)( otherval & 0xff); |
| 3774 | } |
| 3775 | else if (bits == 32) { |
| 3776 | *s++ = (U8)((otherval >> 24) & 0xff); |
| 3777 | *s++ = (U8)((otherval >> 16) & 0xff); |
| 3778 | *s++ = (U8)((otherval >> 8) & 0xff); |
| 3779 | *s++ = (U8)( otherval & 0xff); |
| 3780 | } |
| 3781 | } |
| 3782 | } |
| 3783 | sv_free(other); /* through with it! */ |
| 3784 | } /* while */ |
| 3785 | return swatch; |
| 3786 | } |
| 3787 | |
| 3788 | HV* |
| 3789 | Perl__swash_inversion_hash(pTHX_ SV* const swash) |
| 3790 | { |
| 3791 | |
| 3792 | /* Subject to change or removal. For use only in one place in regcomp.c. |
| 3793 | * Can't be used on a property that is subject to user override, as it |
| 3794 | * relies on the value of SPECIALS in the swash which would be set by |
| 3795 | * utf8_heavy.pl to the hash in the non-overriden file, and hence is not set |
| 3796 | * for overridden properties |
| 3797 | * |
| 3798 | * Returns a hash which is the inversion and closure of a swash mapping. |
| 3799 | * For example, consider the input lines: |
| 3800 | * 004B 006B |
| 3801 | * 004C 006C |
| 3802 | * 212A 006B |
| 3803 | * |
| 3804 | * The returned hash would have two keys, the utf8 for 006B and the utf8 for |
| 3805 | * 006C. The value for each key is an array. For 006C, the array would |
| 3806 | * have a two elements, the utf8 for itself, and for 004C. For 006B, there |
| 3807 | * would be three elements in its array, the utf8 for 006B, 004B and 212A. |
| 3808 | * |
| 3809 | * Essentially, for any code point, it gives all the code points that map to |
| 3810 | * it, or the list of 'froms' for that point. |
| 3811 | * |
| 3812 | * Currently it ignores any additions or deletions from other swashes, |
| 3813 | * looking at just the main body of the swash, and if there are SPECIALS |
| 3814 | * in the swash, at that hash |
| 3815 | * |
| 3816 | * The specials hash can be extra code points, and most likely consists of |
| 3817 | * maps from single code points to multiple ones (each expressed as a string |
| 3818 | * of utf8 characters). This function currently returns only 1-1 mappings. |
| 3819 | * However consider this possible input in the specials hash: |
| 3820 | * "\xEF\xAC\x85" => "\x{0073}\x{0074}", # U+FB05 => 0073 0074 |
| 3821 | * "\xEF\xAC\x86" => "\x{0073}\x{0074}", # U+FB06 => 0073 0074 |
| 3822 | * |
| 3823 | * Both FB05 and FB06 map to the same multi-char sequence, which we don't |
| 3824 | * currently handle. But it also means that FB05 and FB06 are equivalent in |
| 3825 | * a 1-1 mapping which we should handle, and this relationship may not be in |
| 3826 | * the main table. Therefore this function examines all the multi-char |
| 3827 | * sequences and adds the 1-1 mappings that come out of that. */ |
| 3828 | |
| 3829 | U8 *l, *lend; |
| 3830 | STRLEN lcur; |
| 3831 | HV *const hv = MUTABLE_HV(SvRV(swash)); |
| 3832 | |
| 3833 | /* The string containing the main body of the table */ |
| 3834 | SV** const listsvp = hv_fetchs(hv, "LIST", FALSE); |
| 3835 | |
| 3836 | SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE); |
| 3837 | SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE); |
| 3838 | SV** const nonesvp = hv_fetchs(hv, "NONE", FALSE); |
| 3839 | /*SV** const extssvp = hv_fetchs(hv, "EXTRAS", FALSE);*/ |
| 3840 | const U8* const typestr = (U8*)SvPV_nolen(*typesvp); |
| 3841 | const STRLEN bits = SvUV(*bitssvp); |
| 3842 | const STRLEN octets = bits >> 3; /* if bits == 1, then octets == 0 */ |
| 3843 | const UV none = SvUV(*nonesvp); |
| 3844 | SV **specials_p = hv_fetchs(hv, "SPECIALS", 0); |
| 3845 | |
| 3846 | HV* ret = newHV(); |
| 3847 | |
| 3848 | PERL_ARGS_ASSERT__SWASH_INVERSION_HASH; |
| 3849 | |
| 3850 | /* Must have at least 8 bits to get the mappings */ |
| 3851 | if (bits != 8 && bits != 16 && bits != 32) { |
| 3852 | Perl_croak(aTHX_ "panic: swash_inversion_hash doesn't expect bits %"UVuf, |
| 3853 | (UV)bits); |
| 3854 | } |
| 3855 | |
| 3856 | if (specials_p) { /* It might be "special" (sometimes, but not always, a |
| 3857 | mapping to more than one character */ |
| 3858 | |
| 3859 | /* Construct an inverse mapping hash for the specials */ |
| 3860 | HV * const specials_hv = MUTABLE_HV(SvRV(*specials_p)); |
| 3861 | HV * specials_inverse = newHV(); |
| 3862 | char *char_from; /* the lhs of the map */ |
| 3863 | I32 from_len; /* its byte length */ |
| 3864 | char *char_to; /* the rhs of the map */ |
| 3865 | I32 to_len; /* its byte length */ |
| 3866 | SV *sv_to; /* and in a sv */ |
| 3867 | AV* from_list; /* list of things that map to each 'to' */ |
| 3868 | |
| 3869 | hv_iterinit(specials_hv); |
| 3870 | |
| 3871 | /* The keys are the characters (in utf8) that map to the corresponding |
| 3872 | * utf8 string value. Iterate through the list creating the inverse |
| 3873 | * list. */ |
| 3874 | while ((sv_to = hv_iternextsv(specials_hv, &char_from, &from_len))) { |
| 3875 | SV** listp; |
| 3876 | if (! SvPOK(sv_to)) { |
| 3877 | Perl_croak(aTHX_ "panic: value returned from hv_iternextsv() " |
| 3878 | "unexpectedly is not a string, flags=%lu", |
| 3879 | (unsigned long)SvFLAGS(sv_to)); |
| 3880 | } |
| 3881 | /*DEBUG_U(PerlIO_printf(Perl_debug_log, "Found mapping from %"UVXf", First char of to is %"UVXf"\n", valid_utf8_to_uvchr((U8*) char_from, 0), valid_utf8_to_uvchr((U8*) SvPVX(sv_to), 0)));*/ |
| 3882 | |
| 3883 | /* Each key in the inverse list is a mapped-to value, and the key's |
| 3884 | * hash value is a list of the strings (each in utf8) that map to |
| 3885 | * it. Those strings are all one character long */ |
| 3886 | if ((listp = hv_fetch(specials_inverse, |
| 3887 | SvPVX(sv_to), |
| 3888 | SvCUR(sv_to), 0))) |
| 3889 | { |
| 3890 | from_list = (AV*) *listp; |
| 3891 | } |
| 3892 | else { /* No entry yet for it: create one */ |
| 3893 | from_list = newAV(); |
| 3894 | if (! hv_store(specials_inverse, |
| 3895 | SvPVX(sv_to), |
| 3896 | SvCUR(sv_to), |
| 3897 | (SV*) from_list, 0)) |
| 3898 | { |
| 3899 | Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed"); |
| 3900 | } |
| 3901 | } |
| 3902 | |
| 3903 | /* Here have the list associated with this 'to' (perhaps newly |
| 3904 | * created and empty). Just add to it. Note that we ASSUME that |
| 3905 | * the input is guaranteed to not have duplications, so we don't |
| 3906 | * check for that. Duplications just slow down execution time. */ |
| 3907 | av_push(from_list, newSVpvn_utf8(char_from, from_len, TRUE)); |
| 3908 | } |
| 3909 | |
| 3910 | /* Here, 'specials_inverse' contains the inverse mapping. Go through |
| 3911 | * it looking for cases like the FB05/FB06 examples above. There would |
| 3912 | * be an entry in the hash like |
| 3913 | * 'st' => [ FB05, FB06 ] |
| 3914 | * In this example we will create two lists that get stored in the |
| 3915 | * returned hash, 'ret': |
| 3916 | * FB05 => [ FB05, FB06 ] |
| 3917 | * FB06 => [ FB05, FB06 ] |
| 3918 | * |
| 3919 | * Note that there is nothing to do if the array only has one element. |
| 3920 | * (In the normal 1-1 case handled below, we don't have to worry about |
| 3921 | * two lists, as everything gets tied to the single list that is |
| 3922 | * generated for the single character 'to'. But here, we are omitting |
| 3923 | * that list, ('st' in the example), so must have multiple lists.) */ |
| 3924 | while ((from_list = (AV *) hv_iternextsv(specials_inverse, |
| 3925 | &char_to, &to_len))) |
| 3926 | { |
| 3927 | if (av_len(from_list) > 0) { |
| 3928 | int i; |
| 3929 | |
| 3930 | /* We iterate over all combinations of i,j to place each code |
| 3931 | * point on each list */ |
| 3932 | for (i = 0; i <= av_len(from_list); i++) { |
| 3933 | int j; |
| 3934 | AV* i_list = newAV(); |
| 3935 | SV** entryp = av_fetch(from_list, i, FALSE); |
| 3936 | if (entryp == NULL) { |
| 3937 | Perl_croak(aTHX_ "panic: av_fetch() unexpectedly failed"); |
| 3938 | } |
| 3939 | if (hv_fetch(ret, SvPVX(*entryp), SvCUR(*entryp), FALSE)) { |
| 3940 | Perl_croak(aTHX_ "panic: unexpected entry for %s", SvPVX(*entryp)); |
| 3941 | } |
| 3942 | if (! hv_store(ret, SvPVX(*entryp), SvCUR(*entryp), |
| 3943 | (SV*) i_list, FALSE)) |
| 3944 | { |
| 3945 | Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed"); |
| 3946 | } |
| 3947 | |
| 3948 | /* For debugging: UV u = valid_utf8_to_uvchr((U8*) SvPVX(*entryp), 0);*/ |
| 3949 | for (j = 0; j <= av_len(from_list); j++) { |
| 3950 | entryp = av_fetch(from_list, j, FALSE); |
| 3951 | if (entryp == NULL) { |
| 3952 | Perl_croak(aTHX_ "panic: av_fetch() unexpectedly failed"); |
| 3953 | } |
| 3954 | |
| 3955 | /* When i==j this adds itself to the list */ |
| 3956 | av_push(i_list, newSVuv(utf8_to_uvchr_buf( |
| 3957 | (U8*) SvPVX(*entryp), |
| 3958 | (U8*) SvPVX(*entryp) + SvCUR(*entryp), |
| 3959 | 0))); |
| 3960 | /*DEBUG_U(PerlIO_printf(Perl_debug_log, "%s: %d: Adding %"UVXf" to list for %"UVXf"\n", __FILE__, __LINE__, valid_utf8_to_uvchr((U8*) SvPVX(*entryp), 0), u));*/ |
| 3961 | } |
| 3962 | } |
| 3963 | } |
| 3964 | } |
| 3965 | SvREFCNT_dec(specials_inverse); /* done with it */ |
| 3966 | } /* End of specials */ |
| 3967 | |
| 3968 | /* read $swash->{LIST} */ |
| 3969 | l = (U8*)SvPV(*listsvp, lcur); |
| 3970 | lend = l + lcur; |
| 3971 | |
| 3972 | /* Go through each input line */ |
| 3973 | while (l < lend) { |
| 3974 | UV min, max, val; |
| 3975 | UV inverse; |
| 3976 | l = S_swash_scan_list_line(aTHX_ l, lend, &min, &max, &val, |
| 3977 | cBOOL(octets), typestr); |
| 3978 | if (l > lend) { |
| 3979 | break; |
| 3980 | } |
| 3981 | |
| 3982 | /* Each element in the range is to be inverted */ |
| 3983 | for (inverse = min; inverse <= max; inverse++) { |
| 3984 | AV* list; |
| 3985 | SV** listp; |
| 3986 | IV i; |
| 3987 | bool found_key = FALSE; |
| 3988 | bool found_inverse = FALSE; |
| 3989 | |
| 3990 | /* The key is the inverse mapping */ |
| 3991 | char key[UTF8_MAXBYTES+1]; |
| 3992 | char* key_end = (char *) uvuni_to_utf8((U8*) key, val); |
| 3993 | STRLEN key_len = key_end - key; |
| 3994 | |
| 3995 | /* Get the list for the map */ |
| 3996 | if ((listp = hv_fetch(ret, key, key_len, FALSE))) { |
| 3997 | list = (AV*) *listp; |
| 3998 | } |
| 3999 | else { /* No entry yet for it: create one */ |
| 4000 | list = newAV(); |
| 4001 | if (! hv_store(ret, key, key_len, (SV*) list, FALSE)) { |
| 4002 | Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed"); |
| 4003 | } |
| 4004 | } |
| 4005 | |
| 4006 | /* Look through list to see if this inverse mapping already is |
| 4007 | * listed, or if there is a mapping to itself already */ |
| 4008 | for (i = 0; i <= av_len(list); i++) { |
| 4009 | SV** entryp = av_fetch(list, i, FALSE); |
| 4010 | SV* entry; |
| 4011 | if (entryp == NULL) { |
| 4012 | Perl_croak(aTHX_ "panic: av_fetch() unexpectedly failed"); |
| 4013 | } |
| 4014 | entry = *entryp; |
| 4015 | /*DEBUG_U(PerlIO_printf(Perl_debug_log, "list for %"UVXf" contains %"UVXf"\n", val, SvUV(entry)));*/ |
| 4016 | if (SvUV(entry) == val) { |
| 4017 | found_key = TRUE; |
| 4018 | } |
| 4019 | if (SvUV(entry) == inverse) { |
| 4020 | found_inverse = TRUE; |
| 4021 | } |
| 4022 | |
| 4023 | /* No need to continue searching if found everything we are |
| 4024 | * looking for */ |
| 4025 | if (found_key && found_inverse) { |
| 4026 | break; |
| 4027 | } |
| 4028 | } |
| 4029 | |
| 4030 | /* Make sure there is a mapping to itself on the list */ |
| 4031 | if (! found_key) { |
| 4032 | av_push(list, newSVuv(val)); |
| 4033 | /*DEBUG_U(PerlIO_printf(Perl_debug_log, "%s: %d: Adding %"UVXf" to list for %"UVXf"\n", __FILE__, __LINE__, val, val));*/ |
| 4034 | } |
| 4035 | |
| 4036 | |
| 4037 | /* Simply add the value to the list */ |
| 4038 | if (! found_inverse) { |
| 4039 | av_push(list, newSVuv(inverse)); |
| 4040 | /*DEBUG_U(PerlIO_printf(Perl_debug_log, "%s: %d: Adding %"UVXf" to list for %"UVXf"\n", __FILE__, __LINE__, inverse, val));*/ |
| 4041 | } |
| 4042 | |
| 4043 | /* swatch_get() increments the value of val for each element in the |
| 4044 | * range. That makes more compact tables possible. You can |
| 4045 | * express the capitalization, for example, of all consecutive |
| 4046 | * letters with a single line: 0061\t007A\t0041 This maps 0061 to |
| 4047 | * 0041, 0062 to 0042, etc. I (khw) have never understood 'none', |
| 4048 | * and it's not documented; it appears to be used only in |
| 4049 | * implementing tr//; I copied the semantics from swatch_get(), just |
| 4050 | * in case */ |
| 4051 | if (!none || val < none) { |
| 4052 | ++val; |
| 4053 | } |
| 4054 | } |
| 4055 | } |
| 4056 | |
| 4057 | return ret; |
| 4058 | } |
| 4059 | |
| 4060 | SV* |
| 4061 | Perl__swash_to_invlist(pTHX_ SV* const swash) |
| 4062 | { |
| 4063 | |
| 4064 | /* Subject to change or removal. For use only in one place in regcomp.c */ |
| 4065 | |
| 4066 | U8 *l, *lend; |
| 4067 | char *loc; |
| 4068 | STRLEN lcur; |
| 4069 | HV *const hv = MUTABLE_HV(SvRV(swash)); |
| 4070 | UV elements = 0; /* Number of elements in the inversion list */ |
| 4071 | U8 empty[] = ""; |
| 4072 | |
| 4073 | /* The string containing the main body of the table */ |
| 4074 | SV** const listsvp = hv_fetchs(hv, "LIST", FALSE); |
| 4075 | SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE); |
| 4076 | SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE); |
| 4077 | SV** const extssvp = hv_fetchs(hv, "EXTRAS", FALSE); |
| 4078 | SV** const invert_it_svp = hv_fetchs(hv, "INVERT_IT", FALSE); |
| 4079 | |
| 4080 | const U8* const typestr = (U8*)SvPV_nolen(*typesvp); |
| 4081 | const STRLEN bits = SvUV(*bitssvp); |
| 4082 | const STRLEN octets = bits >> 3; /* if bits == 1, then octets == 0 */ |
| 4083 | U8 *x, *xend; |
| 4084 | STRLEN xcur; |
| 4085 | |
| 4086 | SV* invlist; |
| 4087 | |
| 4088 | PERL_ARGS_ASSERT__SWASH_TO_INVLIST; |
| 4089 | |
| 4090 | /* read $swash->{LIST} */ |
| 4091 | if (SvPOK(*listsvp)) { |
| 4092 | l = (U8*)SvPV(*listsvp, lcur); |
| 4093 | } |
| 4094 | else { |
| 4095 | /* LIST legitimately doesn't contain a string during compilation phases |
| 4096 | * of Perl itself, before the Unicode tables are generated. In this |
| 4097 | * case, just fake things up by creating an empty list */ |
| 4098 | l = empty; |
| 4099 | lcur = 0; |
| 4100 | } |
| 4101 | loc = (char *) l; |
| 4102 | lend = l + lcur; |
| 4103 | |
| 4104 | /* Scan the input to count the number of lines to preallocate array size |
| 4105 | * based on worst possible case, which is each line in the input creates 2 |
| 4106 | * elements in the inversion list: 1) the beginning of a range in the list; |
| 4107 | * 2) the beginning of a range not in the list. */ |
| 4108 | while ((loc = (strchr(loc, '\n'))) != NULL) { |
| 4109 | elements += 2; |
| 4110 | loc++; |
| 4111 | } |
| 4112 | |
| 4113 | /* If the ending is somehow corrupt and isn't a new line, add another |
| 4114 | * element for the final range that isn't in the inversion list */ |
| 4115 | if (! (*lend == '\n' |
| 4116 | || (*lend == '\0' && (lcur == 0 || *(lend - 1) == '\n')))) |
| 4117 | { |
| 4118 | elements++; |
| 4119 | } |
| 4120 | |
| 4121 | invlist = _new_invlist(elements); |
| 4122 | |
| 4123 | /* Now go through the input again, adding each range to the list */ |
| 4124 | while (l < lend) { |
| 4125 | UV start, end; |
| 4126 | UV val; /* Not used by this function */ |
| 4127 | |
| 4128 | l = S_swash_scan_list_line(aTHX_ l, lend, &start, &end, &val, |
| 4129 | cBOOL(octets), typestr); |
| 4130 | |
| 4131 | if (l > lend) { |
| 4132 | break; |
| 4133 | } |
| 4134 | |
| 4135 | invlist = _add_range_to_invlist(invlist, start, end); |
| 4136 | } |
| 4137 | |
| 4138 | /* Invert if the data says it should be */ |
| 4139 | if (invert_it_svp && SvUV(*invert_it_svp)) { |
| 4140 | _invlist_invert_prop(invlist); |
| 4141 | } |
| 4142 | |
| 4143 | /* This code is copied from swatch_get() |
| 4144 | * read $swash->{EXTRAS} */ |
| 4145 | x = (U8*)SvPV(*extssvp, xcur); |
| 4146 | xend = x + xcur; |
| 4147 | while (x < xend) { |
| 4148 | STRLEN namelen; |
| 4149 | U8 *namestr; |
| 4150 | SV** othersvp; |
| 4151 | HV* otherhv; |
| 4152 | STRLEN otherbits; |
| 4153 | SV **otherbitssvp, *other; |
| 4154 | U8 *nl; |
| 4155 | |
| 4156 | const U8 opc = *x++; |
| 4157 | if (opc == '\n') |
| 4158 | continue; |
| 4159 | |
| 4160 | nl = (U8*)memchr(x, '\n', xend - x); |
| 4161 | |
| 4162 | if (opc != '-' && opc != '+' && opc != '!' && opc != '&') { |
| 4163 | if (nl) { |
| 4164 | x = nl + 1; /* 1 is length of "\n" */ |
| 4165 | continue; |
| 4166 | } |
| 4167 | else { |
| 4168 | x = xend; /* to EXTRAS' end at which \n is not found */ |
| 4169 | break; |
| 4170 | } |
| 4171 | } |
| 4172 | |
| 4173 | namestr = x; |
| 4174 | if (nl) { |
| 4175 | namelen = nl - namestr; |
| 4176 | x = nl + 1; |
| 4177 | } |
| 4178 | else { |
| 4179 | namelen = xend - namestr; |
| 4180 | x = xend; |
| 4181 | } |
| 4182 | |
| 4183 | othersvp = hv_fetch(hv, (char *)namestr, namelen, FALSE); |
| 4184 | otherhv = MUTABLE_HV(SvRV(*othersvp)); |
| 4185 | otherbitssvp = hv_fetchs(otherhv, "BITS", FALSE); |
| 4186 | otherbits = (STRLEN)SvUV(*otherbitssvp); |
| 4187 | |
| 4188 | if (bits != otherbits || bits != 1) { |
| 4189 | Perl_croak(aTHX_ "panic: _swash_to_invlist only operates on boolean " |
| 4190 | "properties, bits=%"UVuf", otherbits=%"UVuf, |
| 4191 | (UV)bits, (UV)otherbits); |
| 4192 | } |
| 4193 | |
| 4194 | /* The "other" swatch must be destroyed after. */ |
| 4195 | other = _swash_to_invlist((SV *)*othersvp); |
| 4196 | |
| 4197 | /* End of code copied from swatch_get() */ |
| 4198 | switch (opc) { |
| 4199 | case '+': |
| 4200 | _invlist_union(invlist, other, &invlist); |
| 4201 | break; |
| 4202 | case '!': |
| 4203 | _invlist_invert(other); |
| 4204 | _invlist_union(invlist, other, &invlist); |
| 4205 | break; |
| 4206 | case '-': |
| 4207 | _invlist_subtract(invlist, other, &invlist); |
| 4208 | break; |
| 4209 | case '&': |
| 4210 | _invlist_intersection(invlist, other, &invlist); |
| 4211 | break; |
| 4212 | default: |
| 4213 | break; |
| 4214 | } |
| 4215 | sv_free(other); /* through with it! */ |
| 4216 | } |
| 4217 | |
| 4218 | return invlist; |
| 4219 | } |
| 4220 | |
| 4221 | SV* |
| 4222 | Perl__get_swash_invlist(pTHX_ SV* const swash) |
| 4223 | { |
| 4224 | SV** ptr; |
| 4225 | |
| 4226 | PERL_ARGS_ASSERT__GET_SWASH_INVLIST; |
| 4227 | |
| 4228 | if (! SvROK(swash)) { |
| 4229 | return NULL; |
| 4230 | } |
| 4231 | |
| 4232 | /* If it really isn't a hash, it isn't really swash; must be an inversion |
| 4233 | * list */ |
| 4234 | if (SvTYPE(SvRV(swash)) != SVt_PVHV) { |
| 4235 | return SvRV(swash); |
| 4236 | } |
| 4237 | |
| 4238 | ptr = hv_fetchs(MUTABLE_HV(SvRV(swash)), "V", FALSE); |
| 4239 | if (! ptr) { |
| 4240 | return NULL; |
| 4241 | } |
| 4242 | |
| 4243 | return *ptr; |
| 4244 | } |
| 4245 | |
| 4246 | /* |
| 4247 | =for apidoc uvchr_to_utf8 |
| 4248 | |
| 4249 | Adds the UTF-8 representation of the Native code point C<uv> to the end |
| 4250 | of the string C<d>; C<d> should have at least C<UTF8_MAXBYTES+1> free |
| 4251 | bytes available. The return value is the pointer to the byte after the |
| 4252 | end of the new character. In other words, |
| 4253 | |
| 4254 | d = uvchr_to_utf8(d, uv); |
| 4255 | |
| 4256 | is the recommended wide native character-aware way of saying |
| 4257 | |
| 4258 | *(d++) = uv; |
| 4259 | |
| 4260 | =cut |
| 4261 | */ |
| 4262 | |
| 4263 | /* On ASCII machines this is normally a macro but we want a |
| 4264 | real function in case XS code wants it |
| 4265 | */ |
| 4266 | U8 * |
| 4267 | Perl_uvchr_to_utf8(pTHX_ U8 *d, UV uv) |
| 4268 | { |
| 4269 | PERL_ARGS_ASSERT_UVCHR_TO_UTF8; |
| 4270 | |
| 4271 | return Perl_uvuni_to_utf8_flags(aTHX_ d, NATIVE_TO_UNI(uv), 0); |
| 4272 | } |
| 4273 | |
| 4274 | U8 * |
| 4275 | Perl_uvchr_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags) |
| 4276 | { |
| 4277 | PERL_ARGS_ASSERT_UVCHR_TO_UTF8_FLAGS; |
| 4278 | |
| 4279 | return Perl_uvuni_to_utf8_flags(aTHX_ d, NATIVE_TO_UNI(uv), flags); |
| 4280 | } |
| 4281 | |
| 4282 | /* |
| 4283 | =for apidoc utf8n_to_uvchr |
| 4284 | |
| 4285 | Returns the native character value of the first character in the string |
| 4286 | C<s> |
| 4287 | which is assumed to be in UTF-8 encoding; C<retlen> will be set to the |
| 4288 | length, in bytes, of that character. |
| 4289 | |
| 4290 | C<length> and C<flags> are the same as L</utf8n_to_uvuni>(). |
| 4291 | |
| 4292 | =cut |
| 4293 | */ |
| 4294 | /* On ASCII machines this is normally a macro but we want |
| 4295 | a real function in case XS code wants it |
| 4296 | */ |
| 4297 | UV |
| 4298 | Perl_utf8n_to_uvchr(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen, |
| 4299 | U32 flags) |
| 4300 | { |
| 4301 | const UV uv = Perl_utf8n_to_uvuni(aTHX_ s, curlen, retlen, flags); |
| 4302 | |
| 4303 | PERL_ARGS_ASSERT_UTF8N_TO_UVCHR; |
| 4304 | |
| 4305 | return UNI_TO_NATIVE(uv); |
| 4306 | } |
| 4307 | |
| 4308 | bool |
| 4309 | Perl_check_utf8_print(pTHX_ register const U8* s, const STRLEN len) |
| 4310 | { |
| 4311 | /* May change: warns if surrogates, non-character code points, or |
| 4312 | * non-Unicode code points are in s which has length len bytes. Returns |
| 4313 | * TRUE if none found; FALSE otherwise. The only other validity check is |
| 4314 | * to make sure that this won't exceed the string's length */ |
| 4315 | |
| 4316 | const U8* const e = s + len; |
| 4317 | bool ok = TRUE; |
| 4318 | |
| 4319 | PERL_ARGS_ASSERT_CHECK_UTF8_PRINT; |
| 4320 | |
| 4321 | while (s < e) { |
| 4322 | if (UTF8SKIP(s) > len) { |
| 4323 | Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), |
| 4324 | "%s in %s", unees, PL_op ? OP_DESC(PL_op) : "print"); |
| 4325 | return FALSE; |
| 4326 | } |
| 4327 | if (UNLIKELY(*s >= UTF8_FIRST_PROBLEMATIC_CODE_POINT_FIRST_BYTE)) { |
| 4328 | STRLEN char_len; |
| 4329 | if (UTF8_IS_SUPER(s)) { |
| 4330 | if (ckWARN_d(WARN_NON_UNICODE)) { |
| 4331 | UV uv = utf8_to_uvchr_buf(s, e, &char_len); |
| 4332 | Perl_warner(aTHX_ packWARN(WARN_NON_UNICODE), |
| 4333 | "Code point 0x%04"UVXf" is not Unicode, may not be portable", uv); |
| 4334 | ok = FALSE; |
| 4335 | } |
| 4336 | } |
| 4337 | else if (UTF8_IS_SURROGATE(s)) { |
| 4338 | if (ckWARN_d(WARN_SURROGATE)) { |
| 4339 | UV uv = utf8_to_uvchr_buf(s, e, &char_len); |
| 4340 | Perl_warner(aTHX_ packWARN(WARN_SURROGATE), |
| 4341 | "Unicode surrogate U+%04"UVXf" is illegal in UTF-8", uv); |
| 4342 | ok = FALSE; |
| 4343 | } |
| 4344 | } |
| 4345 | else if |
| 4346 | ((UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s)) |
| 4347 | && (ckWARN_d(WARN_NONCHAR))) |
| 4348 | { |
| 4349 | UV uv = utf8_to_uvchr_buf(s, e, &char_len); |
| 4350 | Perl_warner(aTHX_ packWARN(WARN_NONCHAR), |
| 4351 | "Unicode non-character U+%04"UVXf" is illegal for open interchange", uv); |
| 4352 | ok = FALSE; |
| 4353 | } |
| 4354 | } |
| 4355 | s += UTF8SKIP(s); |
| 4356 | } |
| 4357 | |
| 4358 | return ok; |
| 4359 | } |
| 4360 | |
| 4361 | /* |
| 4362 | =for apidoc pv_uni_display |
| 4363 | |
| 4364 | Build to the scalar C<dsv> a displayable version of the string C<spv>, |
| 4365 | length C<len>, the displayable version being at most C<pvlim> bytes long |
| 4366 | (if longer, the rest is truncated and "..." will be appended). |
| 4367 | |
| 4368 | The C<flags> argument can have UNI_DISPLAY_ISPRINT set to display |
| 4369 | isPRINT()able characters as themselves, UNI_DISPLAY_BACKSLASH |
| 4370 | to display the \\[nrfta\\] as the backslashed versions (like '\n') |
| 4371 | (UNI_DISPLAY_BACKSLASH is preferred over UNI_DISPLAY_ISPRINT for \\). |
| 4372 | UNI_DISPLAY_QQ (and its alias UNI_DISPLAY_REGEX) have both |
| 4373 | UNI_DISPLAY_BACKSLASH and UNI_DISPLAY_ISPRINT turned on. |
| 4374 | |
| 4375 | The pointer to the PV of the C<dsv> is returned. |
| 4376 | |
| 4377 | =cut */ |
| 4378 | char * |
| 4379 | Perl_pv_uni_display(pTHX_ SV *dsv, const U8 *spv, STRLEN len, STRLEN pvlim, UV flags) |
| 4380 | { |
| 4381 | int truncated = 0; |
| 4382 | const char *s, *e; |
| 4383 | |
| 4384 | PERL_ARGS_ASSERT_PV_UNI_DISPLAY; |
| 4385 | |
| 4386 | sv_setpvs(dsv, ""); |
| 4387 | SvUTF8_off(dsv); |
| 4388 | for (s = (const char *)spv, e = s + len; s < e; s += UTF8SKIP(s)) { |
| 4389 | UV u; |
| 4390 | /* This serves double duty as a flag and a character to print after |
| 4391 | a \ when flags & UNI_DISPLAY_BACKSLASH is true. |
| 4392 | */ |
| 4393 | char ok = 0; |
| 4394 | |
| 4395 | if (pvlim && SvCUR(dsv) >= pvlim) { |
| 4396 | truncated++; |
| 4397 | break; |
| 4398 | } |
| 4399 | u = utf8_to_uvchr_buf((U8*)s, (U8*)e, 0); |
| 4400 | if (u < 256) { |
| 4401 | const unsigned char c = (unsigned char)u & 0xFF; |
| 4402 | if (flags & UNI_DISPLAY_BACKSLASH) { |
| 4403 | switch (c) { |
| 4404 | case '\n': |
| 4405 | ok = 'n'; break; |
| 4406 | case '\r': |
| 4407 | ok = 'r'; break; |
| 4408 | case '\t': |
| 4409 | ok = 't'; break; |
| 4410 | case '\f': |
| 4411 | ok = 'f'; break; |
| 4412 | case '\a': |
| 4413 | ok = 'a'; break; |
| 4414 | case '\\': |
| 4415 | ok = '\\'; break; |
| 4416 | default: break; |
| 4417 | } |
| 4418 | if (ok) { |
| 4419 | const char string = ok; |
| 4420 | sv_catpvs(dsv, "\\"); |
| 4421 | sv_catpvn(dsv, &string, 1); |
| 4422 | } |
| 4423 | } |
| 4424 | /* isPRINT() is the locale-blind version. */ |
| 4425 | if (!ok && (flags & UNI_DISPLAY_ISPRINT) && isPRINT(c)) { |
| 4426 | const char string = c; |
| 4427 | sv_catpvn(dsv, &string, 1); |
| 4428 | ok = 1; |
| 4429 | } |
| 4430 | } |
| 4431 | if (!ok) |
| 4432 | Perl_sv_catpvf(aTHX_ dsv, "\\x{%"UVxf"}", u); |
| 4433 | } |
| 4434 | if (truncated) |
| 4435 | sv_catpvs(dsv, "..."); |
| 4436 | |
| 4437 | return SvPVX(dsv); |
| 4438 | } |
| 4439 | |
| 4440 | /* |
| 4441 | =for apidoc sv_uni_display |
| 4442 | |
| 4443 | Build to the scalar C<dsv> a displayable version of the scalar C<sv>, |
| 4444 | the displayable version being at most C<pvlim> bytes long |
| 4445 | (if longer, the rest is truncated and "..." will be appended). |
| 4446 | |
| 4447 | The C<flags> argument is as in L</pv_uni_display>(). |
| 4448 | |
| 4449 | The pointer to the PV of the C<dsv> is returned. |
| 4450 | |
| 4451 | =cut |
| 4452 | */ |
| 4453 | char * |
| 4454 | Perl_sv_uni_display(pTHX_ SV *dsv, SV *ssv, STRLEN pvlim, UV flags) |
| 4455 | { |
| 4456 | PERL_ARGS_ASSERT_SV_UNI_DISPLAY; |
| 4457 | |
| 4458 | return Perl_pv_uni_display(aTHX_ dsv, (const U8*)SvPVX_const(ssv), |
| 4459 | SvCUR(ssv), pvlim, flags); |
| 4460 | } |
| 4461 | |
| 4462 | /* |
| 4463 | =for apidoc foldEQ_utf8 |
| 4464 | |
| 4465 | Returns true if the leading portions of the strings C<s1> and C<s2> (either or both |
| 4466 | of which may be in UTF-8) are the same case-insensitively; false otherwise. |
| 4467 | How far into the strings to compare is determined by other input parameters. |
| 4468 | |
| 4469 | If C<u1> is true, the string C<s1> is assumed to be in UTF-8-encoded Unicode; |
| 4470 | otherwise it is assumed to be in native 8-bit encoding. Correspondingly for C<u2> |
| 4471 | with respect to C<s2>. |
| 4472 | |
| 4473 | If the byte length C<l1> is non-zero, it says how far into C<s1> to check for fold |
| 4474 | equality. In other words, C<s1>+C<l1> will be used as a goal to reach. The |
| 4475 | scan will not be considered to be a match unless the goal is reached, and |
| 4476 | scanning won't continue past that goal. Correspondingly for C<l2> with respect to |
| 4477 | C<s2>. |
| 4478 | |
| 4479 | If C<pe1> is non-NULL and the pointer it points to is not NULL, that pointer is |
| 4480 | considered an end pointer beyond which scanning of C<s1> will not continue under |
| 4481 | any circumstances. This means that if both C<l1> and C<pe1> are specified, and |
| 4482 | C<pe1> |
| 4483 | is less than C<s1>+C<l1>, the match will never be successful because it can |
| 4484 | never |
| 4485 | get as far as its goal (and in fact is asserted against). Correspondingly for |
| 4486 | C<pe2> with respect to C<s2>. |
| 4487 | |
| 4488 | At least one of C<s1> and C<s2> must have a goal (at least one of C<l1> and |
| 4489 | C<l2> must be non-zero), and if both do, both have to be |
| 4490 | reached for a successful match. Also, if the fold of a character is multiple |
| 4491 | characters, all of them must be matched (see tr21 reference below for |
| 4492 | 'folding'). |
| 4493 | |
| 4494 | Upon a successful match, if C<pe1> is non-NULL, |
| 4495 | it will be set to point to the beginning of the I<next> character of C<s1> |
| 4496 | beyond what was matched. Correspondingly for C<pe2> and C<s2>. |
| 4497 | |
| 4498 | For case-insensitiveness, the "casefolding" of Unicode is used |
| 4499 | instead of upper/lowercasing both the characters, see |
| 4500 | L<http://www.unicode.org/unicode/reports/tr21/> (Case Mappings). |
| 4501 | |
| 4502 | =cut */ |
| 4503 | |
| 4504 | /* A flags parameter has been added which may change, and hence isn't |
| 4505 | * externally documented. Currently it is: |
| 4506 | * 0 for as-documented above |
| 4507 | * FOLDEQ_UTF8_NOMIX_ASCII meaning that if a non-ASCII character folds to an |
| 4508 | ASCII one, to not match |
| 4509 | * FOLDEQ_UTF8_LOCALE meaning that locale rules are to be used for code |
| 4510 | * points below 256; unicode rules for above 255; and |
| 4511 | * folds that cross those boundaries are disallowed, |
| 4512 | * like the NOMIX_ASCII option |
| 4513 | * FOLDEQ_S1_ALREADY_FOLDED s1 has already been folded before calling this |
| 4514 | * routine. This allows that step to be skipped. |
| 4515 | * FOLDEQ_S2_ALREADY_FOLDED Similarly. |
| 4516 | */ |
| 4517 | I32 |
| 4518 | Perl_foldEQ_utf8_flags(pTHX_ const char *s1, char **pe1, register UV l1, bool u1, const char *s2, char **pe2, register UV l2, bool u2, U32 flags) |
| 4519 | { |
| 4520 | dVAR; |
| 4521 | const U8 *p1 = (const U8*)s1; /* Point to current char */ |
| 4522 | const U8 *p2 = (const U8*)s2; |
| 4523 | const U8 *g1 = NULL; /* goal for s1 */ |
| 4524 | const U8 *g2 = NULL; |
| 4525 | const U8 *e1 = NULL; /* Don't scan s1 past this */ |
| 4526 | U8 *f1 = NULL; /* Point to current folded */ |
| 4527 | const U8 *e2 = NULL; |
| 4528 | U8 *f2 = NULL; |
| 4529 | STRLEN n1 = 0, n2 = 0; /* Number of bytes in current char */ |
| 4530 | U8 foldbuf1[UTF8_MAXBYTES_CASE+1]; |
| 4531 | U8 foldbuf2[UTF8_MAXBYTES_CASE+1]; |
| 4532 | |
| 4533 | PERL_ARGS_ASSERT_FOLDEQ_UTF8_FLAGS; |
| 4534 | |
| 4535 | /* The algorithm requires that input with the flags on the first line of |
| 4536 | * the assert not be pre-folded. */ |
| 4537 | assert( ! ((flags & (FOLDEQ_UTF8_NOMIX_ASCII | FOLDEQ_UTF8_LOCALE)) |
| 4538 | && (flags & (FOLDEQ_S1_ALREADY_FOLDED | FOLDEQ_S2_ALREADY_FOLDED)))); |
| 4539 | |
| 4540 | if (pe1) { |
| 4541 | e1 = *(U8**)pe1; |
| 4542 | } |
| 4543 | |
| 4544 | if (l1) { |
| 4545 | g1 = (const U8*)s1 + l1; |
| 4546 | } |
| 4547 | |
| 4548 | if (pe2) { |
| 4549 | e2 = *(U8**)pe2; |
| 4550 | } |
| 4551 | |
| 4552 | if (l2) { |
| 4553 | g2 = (const U8*)s2 + l2; |
| 4554 | } |
| 4555 | |
| 4556 | /* Must have at least one goal */ |
| 4557 | assert(g1 || g2); |
| 4558 | |
| 4559 | if (g1) { |
| 4560 | |
| 4561 | /* Will never match if goal is out-of-bounds */ |
| 4562 | assert(! e1 || e1 >= g1); |
| 4563 | |
| 4564 | /* Here, there isn't an end pointer, or it is beyond the goal. We |
| 4565 | * only go as far as the goal */ |
| 4566 | e1 = g1; |
| 4567 | } |
| 4568 | else { |
| 4569 | assert(e1); /* Must have an end for looking at s1 */ |
| 4570 | } |
| 4571 | |
| 4572 | /* Same for goal for s2 */ |
| 4573 | if (g2) { |
| 4574 | assert(! e2 || e2 >= g2); |
| 4575 | e2 = g2; |
| 4576 | } |
| 4577 | else { |
| 4578 | assert(e2); |
| 4579 | } |
| 4580 | |
| 4581 | /* If both operands are already folded, we could just do a memEQ on the |
| 4582 | * whole strings at once, but it would be better if the caller realized |
| 4583 | * this and didn't even call us */ |
| 4584 | |
| 4585 | /* Look through both strings, a character at a time */ |
| 4586 | while (p1 < e1 && p2 < e2) { |
| 4587 | |
| 4588 | /* If at the beginning of a new character in s1, get its fold to use |
| 4589 | * and the length of the fold. (exception: locale rules just get the |
| 4590 | * character to a single byte) */ |
| 4591 | if (n1 == 0) { |
| 4592 | if (flags & FOLDEQ_S1_ALREADY_FOLDED) { |
| 4593 | f1 = (U8 *) p1; |
| 4594 | n1 = UTF8SKIP(f1); |
| 4595 | } |
| 4596 | |
| 4597 | else { |
| 4598 | /* If in locale matching, we use two sets of rules, depending |
| 4599 | * on if the code point is above or below 255. Here, we test |
| 4600 | * for and handle locale rules */ |
| 4601 | if ((flags & FOLDEQ_UTF8_LOCALE) |
| 4602 | && (! u1 || UTF8_IS_INVARIANT(*p1) |
| 4603 | || UTF8_IS_DOWNGRADEABLE_START(*p1))) |
| 4604 | { |
| 4605 | /* There is no mixing of code points above and below 255. */ |
| 4606 | if (u2 && (! UTF8_IS_INVARIANT(*p2) |
| 4607 | && ! UTF8_IS_DOWNGRADEABLE_START(*p2))) |
| 4608 | { |
| 4609 | return 0; |
| 4610 | } |
| 4611 | |
| 4612 | /* We handle locale rules by converting, if necessary, the |
| 4613 | * code point to a single byte. */ |
| 4614 | if (! u1 || UTF8_IS_INVARIANT(*p1)) { |
| 4615 | *foldbuf1 = *p1; |
| 4616 | } |
| 4617 | else { |
| 4618 | *foldbuf1 = TWO_BYTE_UTF8_TO_UNI(*p1, *(p1 + 1)); |
| 4619 | } |
| 4620 | n1 = 1; |
| 4621 | } |
| 4622 | else if (isASCII(*p1)) { /* Note, that here won't be both |
| 4623 | ASCII and using locale rules */ |
| 4624 | |
| 4625 | /* If trying to mix non- with ASCII, and not supposed to, |
| 4626 | * fail */ |
| 4627 | if ((flags & FOLDEQ_UTF8_NOMIX_ASCII) && ! isASCII(*p2)) { |
| 4628 | return 0; |
| 4629 | } |
| 4630 | n1 = 1; |
| 4631 | *foldbuf1 = toLOWER(*p1); /* Folds in the ASCII range are |
| 4632 | just lowercased */ |
| 4633 | } |
| 4634 | else if (u1) { |
| 4635 | to_utf8_fold(p1, foldbuf1, &n1); |
| 4636 | } |
| 4637 | else { /* Not utf8, get utf8 fold */ |
| 4638 | to_uni_fold(NATIVE_TO_UNI(*p1), foldbuf1, &n1); |
| 4639 | } |
| 4640 | f1 = foldbuf1; |
| 4641 | } |
| 4642 | } |
| 4643 | |
| 4644 | if (n2 == 0) { /* Same for s2 */ |
| 4645 | if (flags & FOLDEQ_S2_ALREADY_FOLDED) { |
| 4646 | f2 = (U8 *) p2; |
| 4647 | n2 = UTF8SKIP(f2); |
| 4648 | } |
| 4649 | else { |
| 4650 | if ((flags & FOLDEQ_UTF8_LOCALE) |
| 4651 | && (! u2 || UTF8_IS_INVARIANT(*p2) || UTF8_IS_DOWNGRADEABLE_START(*p2))) |
| 4652 | { |
| 4653 | /* Here, the next char in s2 is < 256. We've already |
| 4654 | * worked on s1, and if it isn't also < 256, can't match */ |
| 4655 | if (u1 && (! UTF8_IS_INVARIANT(*p1) |
| 4656 | && ! UTF8_IS_DOWNGRADEABLE_START(*p1))) |
| 4657 | { |
| 4658 | return 0; |
| 4659 | } |
| 4660 | if (! u2 || UTF8_IS_INVARIANT(*p2)) { |
| 4661 | *foldbuf2 = *p2; |
| 4662 | } |
| 4663 | else { |
| 4664 | *foldbuf2 = TWO_BYTE_UTF8_TO_UNI(*p2, *(p2 + 1)); |
| 4665 | } |
| 4666 | |
| 4667 | /* Use another function to handle locale rules. We've made |
| 4668 | * sure that both characters to compare are single bytes */ |
| 4669 | if (! foldEQ_locale((char *) f1, (char *) foldbuf2, 1)) { |
| 4670 | return 0; |
| 4671 | } |
| 4672 | n1 = n2 = 0; |
| 4673 | } |
| 4674 | else if (isASCII(*p2)) { |
| 4675 | if ((flags & FOLDEQ_UTF8_NOMIX_ASCII) && ! isASCII(*p1)) { |
| 4676 | return 0; |
| 4677 | } |
| 4678 | n2 = 1; |
| 4679 | *foldbuf2 = toLOWER(*p2); |
| 4680 | } |
| 4681 | else if (u2) { |
| 4682 | to_utf8_fold(p2, foldbuf2, &n2); |
| 4683 | } |
| 4684 | else { |
| 4685 | to_uni_fold(NATIVE_TO_UNI(*p2), foldbuf2, &n2); |
| 4686 | } |
| 4687 | f2 = foldbuf2; |
| 4688 | } |
| 4689 | } |
| 4690 | |
| 4691 | /* Here f1 and f2 point to the beginning of the strings to compare. |
| 4692 | * These strings are the folds of the next character from each input |
| 4693 | * string, stored in utf8. */ |
| 4694 | |
| 4695 | /* While there is more to look for in both folds, see if they |
| 4696 | * continue to match */ |
| 4697 | while (n1 && n2) { |
| 4698 | U8 fold_length = UTF8SKIP(f1); |
| 4699 | if (fold_length != UTF8SKIP(f2) |
| 4700 | || (fold_length == 1 && *f1 != *f2) /* Short circuit memNE |
| 4701 | function call for single |
| 4702 | byte */ |
| 4703 | || memNE((char*)f1, (char*)f2, fold_length)) |
| 4704 | { |
| 4705 | return 0; /* mismatch */ |
| 4706 | } |
| 4707 | |
| 4708 | /* Here, they matched, advance past them */ |
| 4709 | n1 -= fold_length; |
| 4710 | f1 += fold_length; |
| 4711 | n2 -= fold_length; |
| 4712 | f2 += fold_length; |
| 4713 | } |
| 4714 | |
| 4715 | /* When reach the end of any fold, advance the input past it */ |
| 4716 | if (n1 == 0) { |
| 4717 | p1 += u1 ? UTF8SKIP(p1) : 1; |
| 4718 | } |
| 4719 | if (n2 == 0) { |
| 4720 | p2 += u2 ? UTF8SKIP(p2) : 1; |
| 4721 | } |
| 4722 | } /* End of loop through both strings */ |
| 4723 | |
| 4724 | /* A match is defined by each scan that specified an explicit length |
| 4725 | * reaching its final goal, and the other not having matched a partial |
| 4726 | * character (which can happen when the fold of a character is more than one |
| 4727 | * character). */ |
| 4728 | if (! ((g1 == 0 || p1 == g1) && (g2 == 0 || p2 == g2)) || n1 || n2) { |
| 4729 | return 0; |
| 4730 | } |
| 4731 | |
| 4732 | /* Successful match. Set output pointers */ |
| 4733 | if (pe1) { |
| 4734 | *pe1 = (char*)p1; |
| 4735 | } |
| 4736 | if (pe2) { |
| 4737 | *pe2 = (char*)p2; |
| 4738 | } |
| 4739 | return 1; |
| 4740 | } |
| 4741 | |
| 4742 | /* |
| 4743 | * Local variables: |
| 4744 | * c-indentation-style: bsd |
| 4745 | * c-basic-offset: 4 |
| 4746 | * indent-tabs-mode: nil |
| 4747 | * End: |
| 4748 | * |
| 4749 | * ex: set ts=8 sts=4 sw=4 et: |
| 4750 | */ |