| 1 | /* |
| 2 | * This file is part of DisOrder |
| 3 | * Copyright (C) 2007 Richard Kettlewell |
| 4 | * |
| 5 | * This program is free software; you can redistribute it and/or modify |
| 6 | * it under the terms of the GNU General Public License as published by |
| 7 | * the Free Software Foundation; either version 2 of the License, or |
| 8 | * (at your option) any later version. |
| 9 | * |
| 10 | * This program is distributed in the hope that it will be useful, but |
| 11 | * WITHOUT ANY WARRANTY; without even the implied warranty of |
| 12 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 13 | * General Public License for more details. |
| 14 | * |
| 15 | * You should have received a copy of the GNU General Public License |
| 16 | * along with this program; if not, write to the Free Software |
| 17 | * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 |
| 18 | * USA |
| 19 | */ |
| 20 | /** @file lib/unicode.c |
| 21 | * @brief Unicode support functions |
| 22 | * |
| 23 | * Here by UTF-8 and UTF-8 we mean the encoding forms of those names (not the |
| 24 | * encoding schemes). The primary encoding form is UTF-32 but convenience |
| 25 | * wrappers using UTF-8 are provided for a number of functions. |
| 26 | * |
| 27 | * The idea is that all the strings that hit the database will be in a |
| 28 | * particular normalization form, and for the search and tags database |
| 29 | * in case-folded form, so they can be naively compared within the |
| 30 | * database code. |
| 31 | * |
| 32 | * As the code stands this guarantee is not well met! |
| 33 | * |
| 34 | * Subpages: |
| 35 | * - @ref utf32props |
| 36 | * - @ref utftransform |
| 37 | * - @ref utf32iterator |
| 38 | * - @ref utf32 |
| 39 | * - @ref utf8 |
| 40 | */ |
| 41 | |
| 42 | #include "common.h" |
| 43 | |
| 44 | #include "mem.h" |
| 45 | #include "vector.h" |
| 46 | #include "unicode.h" |
| 47 | #include "unidata.h" |
| 48 | |
| 49 | /** @defgroup utf32props Unicode Code Point Properties */ |
| 50 | /*@{*/ |
| 51 | |
| 52 | static const struct unidata *utf32__unidata_hard(uint32_t c); |
| 53 | |
| 54 | /** @brief Find definition of code point @p c |
| 55 | * @param c Code point |
| 56 | * @return Pointer to @ref unidata structure for @p c |
| 57 | * |
| 58 | * @p c can be any 32-bit value, a sensible value will be returned regardless. |
| 59 | * The returned pointer is NOT guaranteed to be unique to @p c. |
| 60 | */ |
| 61 | static inline const struct unidata *utf32__unidata(uint32_t c) { |
| 62 | /* The bottom half of the table contains almost everything of interest |
| 63 | * and we can just return the right thing straight away */ |
| 64 | if(c < UNICODE_BREAK_START) |
| 65 | return &unidata[c / UNICODE_MODULUS][c % UNICODE_MODULUS]; |
| 66 | else |
| 67 | return utf32__unidata_hard(c); |
| 68 | } |
| 69 | |
| 70 | /** @brief Find definition of code point @p c |
| 71 | * @param c Code point |
| 72 | * @return Pointer to @ref unidata structure for @p c |
| 73 | * |
| 74 | * @p c can be any 32-bit value, a sensible value will be returned regardless. |
| 75 | * The returned pointer is NOT guaranteed to be unique to @p c. |
| 76 | * |
| 77 | * Don't use this function (although it will work fine) - use utf32__unidata() |
| 78 | * instead. |
| 79 | */ |
| 80 | static const struct unidata *utf32__unidata_hard(uint32_t c) { |
| 81 | if(c < UNICODE_BREAK_START) |
| 82 | return &unidata[c / UNICODE_MODULUS][c % UNICODE_MODULUS]; |
| 83 | /* Within the break everything is unassigned */ |
| 84 | if(c < UNICODE_BREAK_END) |
| 85 | return utf32__unidata(0xFFFF); /* guaranteed to be Cn */ |
| 86 | /* Planes 15 and 16 are (mostly) private use */ |
| 87 | if((c >= 0xF0000 && c <= 0xFFFFD) |
| 88 | || (c >= 0x100000 && c <= 0x10FFFD)) |
| 89 | return utf32__unidata(0xE000); /* first Co code point */ |
| 90 | /* Everything else above the break top is unassigned */ |
| 91 | if(c >= UNICODE_BREAK_TOP) |
| 92 | return utf32__unidata(0xFFFF); /* guaranteed to be Cn */ |
| 93 | /* Currently the rest is language tags and variation selectors */ |
| 94 | c -= (UNICODE_BREAK_END - UNICODE_BREAK_START); |
| 95 | return &unidata[c / UNICODE_MODULUS][c % UNICODE_MODULUS]; |
| 96 | } |
| 97 | |
| 98 | /** @brief Return the combining class of @p c |
| 99 | * @param c Code point |
| 100 | * @return Combining class of @p c |
| 101 | * |
| 102 | * @p c can be any 32-bit value, a sensible value will be returned regardless. |
| 103 | */ |
| 104 | static inline int utf32__combining_class(uint32_t c) { |
| 105 | return utf32__unidata(c)->ccc; |
| 106 | } |
| 107 | |
| 108 | /** @brief Return the combining class of @p c |
| 109 | * @param c Code point |
| 110 | * @return Combining class of @p c |
| 111 | * |
| 112 | * @p c can be any 32-bit value, a sensible value will be returned regardless. |
| 113 | */ |
| 114 | int utf32_combining_class(uint32_t c) { |
| 115 | return utf32__combining_class(c); |
| 116 | } |
| 117 | |
| 118 | /** @brief Return the General_Category value for @p c |
| 119 | * @param c Code point |
| 120 | * @return General_Category property value |
| 121 | * |
| 122 | * @p c can be any 32-bit value, a sensible value will be returned regardless. |
| 123 | */ |
| 124 | static inline enum unicode_General_Category utf32__general_category(uint32_t c) { |
| 125 | return utf32__unidata(c)->general_category; |
| 126 | } |
| 127 | |
| 128 | /** @brief Determine Grapheme_Break property |
| 129 | * @param c Code point |
| 130 | * @return Grapheme_Break property value of @p c |
| 131 | * |
| 132 | * @p c can be any 32-bit value, a sensible value will be returned regardless. |
| 133 | */ |
| 134 | static inline enum unicode_Grapheme_Break utf32__grapheme_break(uint32_t c) { |
| 135 | return utf32__unidata(c)->grapheme_break; |
| 136 | } |
| 137 | |
| 138 | /** @brief Determine Word_Break property |
| 139 | * @param c Code point |
| 140 | * @return Word_Break property value of @p c |
| 141 | * |
| 142 | * @p c can be any 32-bit value, a sensible value will be returned regardless. |
| 143 | */ |
| 144 | static inline enum unicode_Word_Break utf32__word_break(uint32_t c) { |
| 145 | return utf32__unidata(c)->word_break; |
| 146 | } |
| 147 | |
| 148 | /** @brief Determine Sentence_Break property |
| 149 | * @param c Code point |
| 150 | * @return Word_Break property value of @p c |
| 151 | * |
| 152 | * @p c can be any 32-bit value, a sensible value will be returned regardless. |
| 153 | */ |
| 154 | static inline enum unicode_Sentence_Break utf32__sentence_break(uint32_t c) { |
| 155 | return utf32__unidata(c)->sentence_break; |
| 156 | } |
| 157 | |
| 158 | /** @brief Return true if @p c is ignorable for boundary specifications |
| 159 | * @param wb Word break property value |
| 160 | * @return non-0 if @p wb is unicode_Word_Break_Extend or unicode_Word_Break_Format |
| 161 | */ |
| 162 | static inline int utf32__boundary_ignorable(enum unicode_Word_Break wb) { |
| 163 | return (wb == unicode_Word_Break_Extend |
| 164 | || wb == unicode_Word_Break_Format); |
| 165 | } |
| 166 | |
| 167 | /** @brief Return the canonical decomposition of @p c |
| 168 | * @param c Code point |
| 169 | * @return 0-terminated canonical decomposition, or 0 |
| 170 | */ |
| 171 | static inline const uint32_t *utf32__decomposition_canon(uint32_t c) { |
| 172 | const struct unidata *const data = utf32__unidata(c); |
| 173 | const uint32_t *const decomp = data->decomp; |
| 174 | |
| 175 | if(decomp && !(data->flags & unicode_compatibility_decomposition)) |
| 176 | return decomp; |
| 177 | else |
| 178 | return 0; |
| 179 | } |
| 180 | |
| 181 | /** @brief Return the compatibility decomposition of @p c |
| 182 | * @param c Code point |
| 183 | * @return 0-terminated decomposition, or 0 |
| 184 | */ |
| 185 | static inline const uint32_t *utf32__decomposition_compat(uint32_t c) { |
| 186 | return utf32__unidata(c)->decomp; |
| 187 | } |
| 188 | |
| 189 | /*@}*/ |
| 190 | /** @defgroup utftransform Functions that transform between different Unicode encoding forms */ |
| 191 | /*@{*/ |
| 192 | |
| 193 | /** @brief Convert UTF-32 to UTF-8 |
| 194 | * @param s Source string |
| 195 | * @param ns Length of source string in code points |
| 196 | * @param ndp Where to store length of destination string (or NULL) |
| 197 | * @return Newly allocated destination string or NULL on error |
| 198 | * |
| 199 | * If the UTF-32 is not valid then NULL is returned. A UTF-32 code point is |
| 200 | * invalid if: |
| 201 | * - it codes for a UTF-16 surrogate |
| 202 | * - it codes for a value outside the unicode code space |
| 203 | * |
| 204 | * The return value is always 0-terminated. The value returned via @p *ndp |
| 205 | * does not include the terminator. |
| 206 | */ |
| 207 | char *utf32_to_utf8(const uint32_t *s, size_t ns, size_t *ndp) { |
| 208 | struct dynstr d; |
| 209 | uint32_t c; |
| 210 | |
| 211 | dynstr_init(&d); |
| 212 | while(ns > 0) { |
| 213 | c = *s++; |
| 214 | if(c < 0x80) |
| 215 | dynstr_append(&d, c); |
| 216 | else if(c < 0x0800) { |
| 217 | dynstr_append(&d, 0xC0 | (c >> 6)); |
| 218 | dynstr_append(&d, 0x80 | (c & 0x3F)); |
| 219 | } else if(c < 0x10000) { |
| 220 | if(c >= 0xD800 && c <= 0xDFFF) |
| 221 | goto error; |
| 222 | dynstr_append(&d, 0xE0 | (c >> 12)); |
| 223 | dynstr_append(&d, 0x80 | ((c >> 6) & 0x3F)); |
| 224 | dynstr_append(&d, 0x80 | (c & 0x3F)); |
| 225 | } else if(c < 0x110000) { |
| 226 | dynstr_append(&d, 0xF0 | (c >> 18)); |
| 227 | dynstr_append(&d, 0x80 | ((c >> 12) & 0x3F)); |
| 228 | dynstr_append(&d, 0x80 | ((c >> 6) & 0x3F)); |
| 229 | dynstr_append(&d, 0x80 | (c & 0x3F)); |
| 230 | } else |
| 231 | goto error; |
| 232 | --ns; |
| 233 | } |
| 234 | dynstr_terminate(&d); |
| 235 | if(ndp) |
| 236 | *ndp = d.nvec; |
| 237 | return d.vec; |
| 238 | error: |
| 239 | xfree(d.vec); |
| 240 | return 0; |
| 241 | } |
| 242 | |
| 243 | /** @brief Convert UTF-8 to UTF-32 |
| 244 | * @param s Source string |
| 245 | * @param ns Length of source string in code points |
| 246 | * @param ndp Where to store length of destination string (or NULL) |
| 247 | * @return Newly allocated destination string or NULL on error |
| 248 | * |
| 249 | * The return value is always 0-terminated. The value returned via @p *ndp |
| 250 | * does not include the terminator. |
| 251 | * |
| 252 | * If the UTF-8 is not valid then NULL is returned. A UTF-8 sequence |
| 253 | * for a code point is invalid if: |
| 254 | * - it is not the shortest possible sequence for the code point |
| 255 | * - it codes for a UTF-16 surrogate |
| 256 | * - it codes for a value outside the unicode code space |
| 257 | */ |
| 258 | uint32_t *utf8_to_utf32(const char *s, size_t ns, size_t *ndp) { |
| 259 | struct dynstr_ucs4 d; |
| 260 | uint32_t c32; |
| 261 | const uint8_t *ss = (const uint8_t *)s; |
| 262 | int n; |
| 263 | |
| 264 | dynstr_ucs4_init(&d); |
| 265 | while(ns > 0) { |
| 266 | const struct unicode_utf8_row *const r = &unicode_utf8_valid[*ss]; |
| 267 | if(r->count <= ns) { |
| 268 | switch(r->count) { |
| 269 | case 1: |
| 270 | c32 = *ss; |
| 271 | break; |
| 272 | case 2: |
| 273 | if(ss[1] < r->min2 || ss[1] > r->max2) |
| 274 | goto error; |
| 275 | c32 = *ss & 0x1F; |
| 276 | break; |
| 277 | case 3: |
| 278 | if(ss[1] < r->min2 || ss[1] > r->max2) |
| 279 | goto error; |
| 280 | c32 = *ss & 0x0F; |
| 281 | break; |
| 282 | case 4: |
| 283 | if(ss[1] < r->min2 || ss[1] > r->max2) |
| 284 | goto error; |
| 285 | c32 = *ss & 0x07; |
| 286 | break; |
| 287 | default: |
| 288 | goto error; |
| 289 | } |
| 290 | } else |
| 291 | goto error; |
| 292 | for(n = 1; n < r->count; ++n) { |
| 293 | if(ss[n] < 0x80 || ss[n] > 0xBF) |
| 294 | goto error; |
| 295 | c32 = (c32 << 6) | (ss[n] & 0x3F); |
| 296 | } |
| 297 | dynstr_ucs4_append(&d, c32); |
| 298 | ss += r->count; |
| 299 | ns -= r->count; |
| 300 | } |
| 301 | dynstr_ucs4_terminate(&d); |
| 302 | if(ndp) |
| 303 | *ndp = d.nvec; |
| 304 | return d.vec; |
| 305 | error: |
| 306 | xfree(d.vec); |
| 307 | return 0; |
| 308 | } |
| 309 | |
| 310 | /** @brief Test whether [s,s+ns) is valid UTF-8 |
| 311 | * @param s Start of string |
| 312 | * @param ns Length of string |
| 313 | * @return non-0 if @p s is valid UTF-8, 0 if it is not valid |
| 314 | * |
| 315 | * This function is intended to be much faster than calling utf8_to_utf32() and |
| 316 | * throwing away the result. |
| 317 | */ |
| 318 | int utf8_valid(const char *s, size_t ns) { |
| 319 | const uint8_t *ss = (const uint8_t *)s; |
| 320 | while(ns > 0) { |
| 321 | const struct unicode_utf8_row *const r = &unicode_utf8_valid[*ss]; |
| 322 | if(r->count <= ns) { |
| 323 | switch(r->count) { |
| 324 | case 1: |
| 325 | break; |
| 326 | case 2: |
| 327 | if(ss[1] < r->min2 || ss[1] > r->max2) |
| 328 | return 0; |
| 329 | break; |
| 330 | case 3: |
| 331 | if(ss[1] < r->min2 || ss[1] > r->max2) |
| 332 | return 0; |
| 333 | if(ss[2] < 0x80 || ss[2] > 0xBF) |
| 334 | return 0; |
| 335 | break; |
| 336 | case 4: |
| 337 | if(ss[1] < r->min2 || ss[1] > r->max2) |
| 338 | return 0; |
| 339 | if(ss[2] < 0x80 || ss[2] > 0xBF) |
| 340 | return 0; |
| 341 | if(ss[3] < 0x80 || ss[3] > 0xBF) |
| 342 | return 0; |
| 343 | break; |
| 344 | default: |
| 345 | return 0; |
| 346 | } |
| 347 | } else |
| 348 | return 0; |
| 349 | ss += r->count; |
| 350 | ns -= r->count; |
| 351 | } |
| 352 | return 1; |
| 353 | } |
| 354 | |
| 355 | /*@}*/ |
| 356 | /** @defgroup utf32iterator UTF-32 string iterators */ |
| 357 | /*@{*/ |
| 358 | |
| 359 | struct utf32_iterator_data { |
| 360 | /** @brief Start of string */ |
| 361 | const uint32_t *s; |
| 362 | |
| 363 | /** @brief Length of string */ |
| 364 | size_t ns; |
| 365 | |
| 366 | /** @brief Current position */ |
| 367 | size_t n; |
| 368 | |
| 369 | /** @brief Last two non-ignorable characters or (uint32_t)-1 |
| 370 | * |
| 371 | * last[1] is the non-Extend/Format character just before position @p n; |
| 372 | * last[0] is the one just before that. |
| 373 | * |
| 374 | * Exception 1: if there is no such non-Extend/Format character then an |
| 375 | * Extend/Format character is accepted instead. |
| 376 | * |
| 377 | * Exception 2: if there is no such character even taking that into account |
| 378 | * the value is (uint32_t)-1. |
| 379 | */ |
| 380 | uint32_t last[2]; |
| 381 | |
| 382 | /** @brief Tailoring for Word_Break */ |
| 383 | unicode_property_tailor *word_break; |
| 384 | }; |
| 385 | |
| 386 | /** @brief Initialize an internal private iterator |
| 387 | * @param it Iterator |
| 388 | * @param s Start of string |
| 389 | * @param ns Length of string |
| 390 | * @param n Absolute position |
| 391 | */ |
| 392 | static void utf32__iterator_init(utf32_iterator it, |
| 393 | const uint32_t *s, size_t ns, size_t n) { |
| 394 | it->s = s; |
| 395 | it->ns = ns; |
| 396 | it->n = 0; |
| 397 | it->last[0] = it->last[1] = -1; |
| 398 | it->word_break = 0; |
| 399 | utf32_iterator_set(it, n); |
| 400 | } |
| 401 | |
| 402 | /** @brief Create a new iterator pointing at the start of a string |
| 403 | * @param s Start of string |
| 404 | * @param ns Length of string |
| 405 | * @return New iterator |
| 406 | */ |
| 407 | utf32_iterator utf32_iterator_new(const uint32_t *s, size_t ns) { |
| 408 | utf32_iterator it = xmalloc(sizeof *it); |
| 409 | utf32__iterator_init(it, s, ns, 0); |
| 410 | return it; |
| 411 | } |
| 412 | |
| 413 | /** @brief Tailor this iterator's interpretation of the Word_Break property. |
| 414 | * @param it Iterator |
| 415 | * @param pt Property tailor function or NULL |
| 416 | * |
| 417 | * After calling this the iterator will call @p pt to determine the Word_Break |
| 418 | * property of each code point. If it returns -1 the default value will be |
| 419 | * used otherwise the returned value will be used. |
| 420 | * |
| 421 | * @p pt can be NULL to revert to the default value of the property. |
| 422 | * |
| 423 | * It is safe to call this function at any time; the iterator's internal state |
| 424 | * will be reset to suit the new tailoring. |
| 425 | */ |
| 426 | void utf32_iterator_tailor_word_break(utf32_iterator it, |
| 427 | unicode_property_tailor *pt) { |
| 428 | it->word_break = pt; |
| 429 | utf32_iterator_set(it, it->n); |
| 430 | } |
| 431 | |
| 432 | static inline enum unicode_Word_Break utf32__iterator_word_break(utf32_iterator it, |
| 433 | uint32_t c) { |
| 434 | if(!it->word_break) |
| 435 | return utf32__word_break(c); |
| 436 | else { |
| 437 | const int t = it->word_break(c); |
| 438 | |
| 439 | if(t < 0) |
| 440 | return utf32__word_break(c); |
| 441 | else |
| 442 | return t; |
| 443 | } |
| 444 | } |
| 445 | |
| 446 | /** @brief Destroy an iterator |
| 447 | * @param it Iterator |
| 448 | */ |
| 449 | void utf32_iterator_destroy(utf32_iterator it) { |
| 450 | xfree(it); |
| 451 | } |
| 452 | |
| 453 | /** @brief Find the current position of an interator |
| 454 | * @param it Iterator |
| 455 | */ |
| 456 | size_t utf32_iterator_where(utf32_iterator it) { |
| 457 | return it->n; |
| 458 | } |
| 459 | |
| 460 | /** @brief Set an iterator's absolute position |
| 461 | * @param it Iterator |
| 462 | * @param n Absolute position |
| 463 | * @return 0 on success, non-0 on error |
| 464 | * |
| 465 | * It is an error to position the iterator outside the string (but acceptable |
| 466 | * to point it at the hypothetical post-final character). If an invalid value |
| 467 | * of @p n is specified then the iterator is not changed. |
| 468 | * |
| 469 | * This function works by backing up and then advancing to reconstruct the |
| 470 | * iterator's internal state for position @p n. The worst case will be O(n) |
| 471 | * time complexity (with a worse constant factor that utf32_iterator_advance()) |
| 472 | * but the typical case is essentially constant-time. |
| 473 | */ |
| 474 | int utf32_iterator_set(utf32_iterator it, size_t n) { |
| 475 | /* We can't just jump to position @p n; the @p last[] values will be wrong. |
| 476 | * What we need is to jump a bit behind @p n and then advance forward, |
| 477 | * updating @p last[] along the way. How far back? We need to cross two |
| 478 | * non-ignorable code points as we advance forwards, so we'd better pass two |
| 479 | * such characters on the way back (if such are available). |
| 480 | */ |
| 481 | size_t m; |
| 482 | |
| 483 | if(n > it->ns) /* range check */ |
| 484 | return -1; |
| 485 | /* Walk backwards skipping ignorable code points */ |
| 486 | m = n; |
| 487 | while(m > 0 |
| 488 | && (utf32__boundary_ignorable(utf32__iterator_word_break(it, |
| 489 | it->s[m-1])))) |
| 490 | --m; |
| 491 | /* Either m=0 or s[m-1] is not ignorable */ |
| 492 | if(m > 0) { |
| 493 | --m; |
| 494 | /* s[m] is our first non-ignorable code; look for a second in the same |
| 495 | way **/ |
| 496 | while(m > 0 |
| 497 | && (utf32__boundary_ignorable(utf32__iterator_word_break(it, |
| 498 | it->s[m-1])))) |
| 499 | --m; |
| 500 | /* Either m=0 or s[m-1] is not ignorable */ |
| 501 | if(m > 0) |
| 502 | --m; |
| 503 | } |
| 504 | it->last[0] = it->last[1] = -1; |
| 505 | it->n = m; |
| 506 | return utf32_iterator_advance(it, n - m); |
| 507 | } |
| 508 | |
| 509 | /** @brief Advance an iterator |
| 510 | * @param it Iterator |
| 511 | * @param count Number of code points to advance by |
| 512 | * @return 0 on success, non-0 on error |
| 513 | * |
| 514 | * It is an error to advance an iterator beyond the hypothetical post-final |
| 515 | * character of the string. If an invalid value of @p n is specified then the |
| 516 | * iterator is not changed. |
| 517 | * |
| 518 | * This function has O(n) time complexity: it works by advancing naively |
| 519 | * forwards through the string. |
| 520 | */ |
| 521 | int utf32_iterator_advance(utf32_iterator it, size_t count) { |
| 522 | if(count <= it->ns - it->n) { |
| 523 | while(count > 0) { |
| 524 | const uint32_t c = it->s[it->n]; |
| 525 | const enum unicode_Word_Break wb = utf32__iterator_word_break(it, c); |
| 526 | if(it->last[1] == (uint32_t)-1 |
| 527 | || !utf32__boundary_ignorable(wb)) { |
| 528 | it->last[0] = it->last[1]; |
| 529 | it->last[1] = c; |
| 530 | } |
| 531 | ++it->n; |
| 532 | --count; |
| 533 | } |
| 534 | return 0; |
| 535 | } else |
| 536 | return -1; |
| 537 | } |
| 538 | |
| 539 | /** @brief Find the current code point |
| 540 | * @param it Iterator |
| 541 | * @return Current code point or 0 |
| 542 | * |
| 543 | * If the iterator points at the hypothetical post-final character of the |
| 544 | * string then 0 is returned. NB that this doesn't mean that there aren't any |
| 545 | * 0 code points inside the string! |
| 546 | */ |
| 547 | uint32_t utf32_iterator_code(utf32_iterator it) { |
| 548 | if(it->n < it->ns) |
| 549 | return it->s[it->n]; |
| 550 | else |
| 551 | return 0; |
| 552 | } |
| 553 | |
| 554 | /** @brief Test for a grapheme boundary |
| 555 | * @param it Iterator |
| 556 | * @return Non-0 if pointing just after a grapheme boundary, otherwise 0 |
| 557 | * |
| 558 | * This function identifies default grapheme cluster boundaries as described in |
| 559 | * UAX #29 s3. It returns non-0 if @p it points at the code point just after a |
| 560 | * grapheme cluster boundary (including the hypothetical code point just after |
| 561 | * the end of the string). |
| 562 | */ |
| 563 | int utf32_iterator_grapheme_boundary(utf32_iterator it) { |
| 564 | uint32_t before, after; |
| 565 | enum unicode_Grapheme_Break gbbefore, gbafter; |
| 566 | /* GB1 and GB2 */ |
| 567 | if(it->n == 0 || it->n == it->ns) |
| 568 | return 1; |
| 569 | /* Now we know that s[n-1] and s[n] are safe to inspect */ |
| 570 | /* GB3 */ |
| 571 | before = it->s[it->n-1]; |
| 572 | after = it->s[it->n]; |
| 573 | if(before == 0x000D && after == 0x000A) |
| 574 | return 0; |
| 575 | gbbefore = utf32__grapheme_break(before); |
| 576 | gbafter = utf32__grapheme_break(after); |
| 577 | /* GB4 */ |
| 578 | if(gbbefore == unicode_Grapheme_Break_Control |
| 579 | || before == 0x000D |
| 580 | || before == 0x000A) |
| 581 | return 1; |
| 582 | /* GB5 */ |
| 583 | if(gbafter == unicode_Grapheme_Break_Control |
| 584 | || after == 0x000D |
| 585 | || after == 0x000A) |
| 586 | return 1; |
| 587 | /* GB6 */ |
| 588 | if(gbbefore == unicode_Grapheme_Break_L |
| 589 | && (gbafter == unicode_Grapheme_Break_L |
| 590 | || gbafter == unicode_Grapheme_Break_V |
| 591 | || gbafter == unicode_Grapheme_Break_LV |
| 592 | || gbafter == unicode_Grapheme_Break_LVT)) |
| 593 | return 0; |
| 594 | /* GB7 */ |
| 595 | if((gbbefore == unicode_Grapheme_Break_LV |
| 596 | || gbbefore == unicode_Grapheme_Break_V) |
| 597 | && (gbafter == unicode_Grapheme_Break_V |
| 598 | || gbafter == unicode_Grapheme_Break_T)) |
| 599 | return 0; |
| 600 | /* GB8 */ |
| 601 | if((gbbefore == unicode_Grapheme_Break_LVT |
| 602 | || gbbefore == unicode_Grapheme_Break_T) |
| 603 | && gbafter == unicode_Grapheme_Break_T) |
| 604 | return 0; |
| 605 | /* GB9 */ |
| 606 | if(gbafter == unicode_Grapheme_Break_Extend) |
| 607 | return 0; |
| 608 | /* GB10 */ |
| 609 | return 1; |
| 610 | |
| 611 | } |
| 612 | |
| 613 | /** @brief Test for a word boundary |
| 614 | * @param it Iterator |
| 615 | * @return Non-0 if pointing just after a word boundary, otherwise 0 |
| 616 | * |
| 617 | * This function identifies default word boundaries as described in UAX #29 s4. |
| 618 | * It returns non-0 if @p it points at the code point just after a word |
| 619 | * boundary (including the hypothetical code point just after the end of the |
| 620 | * string) and 0 otherwise. |
| 621 | */ |
| 622 | int utf32_iterator_word_boundary(utf32_iterator it) { |
| 623 | enum unicode_Word_Break twobefore, before, after, twoafter; |
| 624 | size_t nn; |
| 625 | |
| 626 | /* WB1 and WB2 */ |
| 627 | if(it->n == 0 || it->n == it->ns) |
| 628 | return 1; |
| 629 | /* WB3 */ |
| 630 | if(it->s[it->n-1] == 0x000D && it->s[it->n] == 0x000A) |
| 631 | return 0; |
| 632 | /* WB4 */ |
| 633 | /* (!Sep) x (Extend|Format) as in UAX #29 s6.2 */ |
| 634 | if(utf32__sentence_break(it->s[it->n-1]) != unicode_Sentence_Break_Sep |
| 635 | && utf32__boundary_ignorable(utf32__iterator_word_break(it, it->s[it->n]))) |
| 636 | return 0; |
| 637 | /* Gather the property values we'll need for the rest of the test taking the |
| 638 | * s6.2 changes into account */ |
| 639 | /* First we look at the code points after the proposed boundary */ |
| 640 | nn = it->n; /* <it->ns */ |
| 641 | after = utf32__iterator_word_break(it, it->s[nn++]); |
| 642 | if(!utf32__boundary_ignorable(after)) { |
| 643 | /* X (Extend|Format)* -> X */ |
| 644 | while(nn < it->ns |
| 645 | && utf32__boundary_ignorable(utf32__iterator_word_break(it, |
| 646 | it->s[nn]))) |
| 647 | ++nn; |
| 648 | } |
| 649 | /* It's possible now that nn=ns */ |
| 650 | if(nn < it->ns) |
| 651 | twoafter = utf32__iterator_word_break(it, it->s[nn]); |
| 652 | else |
| 653 | twoafter = unicode_Word_Break_Other; |
| 654 | |
| 655 | /* We've already recorded the non-ignorable code points before the proposed |
| 656 | * boundary */ |
| 657 | before = utf32__iterator_word_break(it, it->last[1]); |
| 658 | twobefore = utf32__iterator_word_break(it, it->last[0]); |
| 659 | |
| 660 | /* WB5 */ |
| 661 | if(before == unicode_Word_Break_ALetter |
| 662 | && after == unicode_Word_Break_ALetter) |
| 663 | return 0; |
| 664 | /* WB6 */ |
| 665 | if(before == unicode_Word_Break_ALetter |
| 666 | && after == unicode_Word_Break_MidLetter |
| 667 | && twoafter == unicode_Word_Break_ALetter) |
| 668 | return 0; |
| 669 | /* WB7 */ |
| 670 | if(twobefore == unicode_Word_Break_ALetter |
| 671 | && before == unicode_Word_Break_MidLetter |
| 672 | && after == unicode_Word_Break_ALetter) |
| 673 | return 0; |
| 674 | /* WB8 */ |
| 675 | if(before == unicode_Word_Break_Numeric |
| 676 | && after == unicode_Word_Break_Numeric) |
| 677 | return 0; |
| 678 | /* WB9 */ |
| 679 | if(before == unicode_Word_Break_ALetter |
| 680 | && after == unicode_Word_Break_Numeric) |
| 681 | return 0; |
| 682 | /* WB10 */ |
| 683 | if(before == unicode_Word_Break_Numeric |
| 684 | && after == unicode_Word_Break_ALetter) |
| 685 | return 0; |
| 686 | /* WB11 */ |
| 687 | if(twobefore == unicode_Word_Break_Numeric |
| 688 | && before == unicode_Word_Break_MidNum |
| 689 | && after == unicode_Word_Break_Numeric) |
| 690 | return 0; |
| 691 | /* WB12 */ |
| 692 | if(before == unicode_Word_Break_Numeric |
| 693 | && after == unicode_Word_Break_MidNum |
| 694 | && twoafter == unicode_Word_Break_Numeric) |
| 695 | return 0; |
| 696 | /* WB13 */ |
| 697 | if(before == unicode_Word_Break_Katakana |
| 698 | && after == unicode_Word_Break_Katakana) |
| 699 | return 0; |
| 700 | /* WB13a */ |
| 701 | if((before == unicode_Word_Break_ALetter |
| 702 | || before == unicode_Word_Break_Numeric |
| 703 | || before == unicode_Word_Break_Katakana |
| 704 | || before == unicode_Word_Break_ExtendNumLet) |
| 705 | && after == unicode_Word_Break_ExtendNumLet) |
| 706 | return 0; |
| 707 | /* WB13b */ |
| 708 | if(before == unicode_Word_Break_ExtendNumLet |
| 709 | && (after == unicode_Word_Break_ALetter |
| 710 | || after == unicode_Word_Break_Numeric |
| 711 | || after == unicode_Word_Break_Katakana)) |
| 712 | return 0; |
| 713 | /* WB14 */ |
| 714 | return 1; |
| 715 | } |
| 716 | |
| 717 | /*@}*/ |
| 718 | /** @defgroup utf32 Functions that operate on UTF-32 strings */ |
| 719 | /*@{*/ |
| 720 | |
| 721 | /** @brief Return the length of a 0-terminated UTF-32 string |
| 722 | * @param s Pointer to 0-terminated string |
| 723 | * @return Length of string in code points (excluding terminator) |
| 724 | * |
| 725 | * Unlike the conversion functions no validity checking is done on the string. |
| 726 | */ |
| 727 | size_t utf32_len(const uint32_t *s) { |
| 728 | const uint32_t *t = s; |
| 729 | |
| 730 | while(*t) |
| 731 | ++t; |
| 732 | return (size_t)(t - s); |
| 733 | } |
| 734 | |
| 735 | /** @brief Stably sort [s,s+ns) into descending order of combining class |
| 736 | * @param s Start of array |
| 737 | * @param ns Number of elements, must be at least 1 |
| 738 | * @param buffer Buffer of at least @p ns elements |
| 739 | */ |
| 740 | static void utf32__sort_ccc(uint32_t *s, size_t ns, uint32_t *buffer) { |
| 741 | uint32_t *a, *b, *bp; |
| 742 | size_t na, nb; |
| 743 | |
| 744 | switch(ns) { |
| 745 | case 1: /* 1-element array is always sorted */ |
| 746 | return; |
| 747 | case 2: /* 2-element arrays are trivial to sort */ |
| 748 | if(utf32__combining_class(s[0]) > utf32__combining_class(s[1])) { |
| 749 | uint32_t tmp = s[0]; |
| 750 | s[0] = s[1]; |
| 751 | s[1] = tmp; |
| 752 | } |
| 753 | return; |
| 754 | default: |
| 755 | /* Partition the array */ |
| 756 | na = ns / 2; |
| 757 | nb = ns - na; |
| 758 | a = s; |
| 759 | b = s + na; |
| 760 | /* Sort the two halves of the array */ |
| 761 | utf32__sort_ccc(a, na, buffer); |
| 762 | utf32__sort_ccc(b, nb, buffer); |
| 763 | /* Merge them back into one, via the buffer */ |
| 764 | bp = buffer; |
| 765 | while(na > 0 && nb > 0) { |
| 766 | /* We want ascending order of combining class (hence <) |
| 767 | * and we want stability within combining classes (hence <=) |
| 768 | */ |
| 769 | if(utf32__combining_class(*a) <= utf32__combining_class(*b)) { |
| 770 | *bp++ = *a++; |
| 771 | --na; |
| 772 | } else { |
| 773 | *bp++ = *b++; |
| 774 | --nb; |
| 775 | } |
| 776 | } |
| 777 | while(na > 0) { |
| 778 | *bp++ = *a++; |
| 779 | --na; |
| 780 | } |
| 781 | while(nb > 0) { |
| 782 | *bp++ = *b++; |
| 783 | --nb; |
| 784 | } |
| 785 | memcpy(s, buffer, ns * sizeof(uint32_t)); |
| 786 | return; |
| 787 | } |
| 788 | } |
| 789 | |
| 790 | /** @brief Put combining characters into canonical order |
| 791 | * @param s Pointer to UTF-32 string |
| 792 | * @param ns Length of @p s |
| 793 | * @return 0 on success, non-0 on error |
| 794 | * |
| 795 | * @p s is modified in-place. See Unicode 5.0 s3.11 for details of the |
| 796 | * ordering. |
| 797 | * |
| 798 | * Currently we only support a maximum of 1024 combining characters after each |
| 799 | * base character. If this limit is exceeded then a non-0 value is returned. |
| 800 | */ |
| 801 | static int utf32__canonical_ordering(uint32_t *s, size_t ns) { |
| 802 | size_t nc; |
| 803 | uint32_t buffer[1024]; |
| 804 | |
| 805 | /* The ordering amounts to a stable sort of each contiguous group of |
| 806 | * characters with non-0 combining class. */ |
| 807 | while(ns > 0) { |
| 808 | /* Skip non-combining characters */ |
| 809 | if(utf32__combining_class(*s) == 0) { |
| 810 | ++s; |
| 811 | --ns; |
| 812 | continue; |
| 813 | } |
| 814 | /* We must now have at least one combining character; see how many |
| 815 | * there are */ |
| 816 | for(nc = 1; nc < ns && utf32__combining_class(s[nc]) != 0; ++nc) |
| 817 | ; |
| 818 | if(nc > 1024) |
| 819 | return -1; |
| 820 | /* Sort the array */ |
| 821 | utf32__sort_ccc(s, nc, buffer); |
| 822 | s += nc; |
| 823 | ns -= nc; |
| 824 | } |
| 825 | return 0; |
| 826 | } |
| 827 | |
| 828 | /* Magic numbers from UAX #15 s16 */ |
| 829 | #define SBase 0xAC00 |
| 830 | #define LBase 0x1100 |
| 831 | #define VBase 0x1161 |
| 832 | #define TBase 0x11A7 |
| 833 | #define LCount 19 |
| 834 | #define VCount 21 |
| 835 | #define TCount 28 |
| 836 | #define NCount (VCount * TCount) |
| 837 | #define SCount (LCount * NCount) |
| 838 | |
| 839 | /** @brief Guts of the decomposition lookup functions */ |
| 840 | #define utf32__decompose_one_generic(WHICH) do { \ |
| 841 | const uint32_t *dc = utf32__decomposition_##WHICH(c); \ |
| 842 | if(dc) { \ |
| 843 | /* Found a canonical decomposition in the table */ \ |
| 844 | while(*dc) \ |
| 845 | utf32__decompose_one_##WHICH(d, *dc++); \ |
| 846 | } else if(c >= SBase && c < SBase + SCount) { \ |
| 847 | /* Mechanically decomposable Hangul syllable (UAX #15 s16) */ \ |
| 848 | const uint32_t SIndex = c - SBase; \ |
| 849 | const uint32_t L = LBase + SIndex / NCount; \ |
| 850 | const uint32_t V = VBase + (SIndex % NCount) / TCount; \ |
| 851 | const uint32_t T = TBase + SIndex % TCount; \ |
| 852 | dynstr_ucs4_append(d, L); \ |
| 853 | dynstr_ucs4_append(d, V); \ |
| 854 | if(T != TBase) \ |
| 855 | dynstr_ucs4_append(d, T); \ |
| 856 | } else \ |
| 857 | /* Equal to own canonical decomposition */ \ |
| 858 | dynstr_ucs4_append(d, c); \ |
| 859 | } while(0) |
| 860 | |
| 861 | /** @brief Recursively compute the canonical decomposition of @p c |
| 862 | * @param d Dynamic string to store decomposition in |
| 863 | * @param c Code point to decompose (must be a valid!) |
| 864 | * @return 0 on success, non-0 on error |
| 865 | */ |
| 866 | static void utf32__decompose_one_canon(struct dynstr_ucs4 *d, uint32_t c) { |
| 867 | utf32__decompose_one_generic(canon); |
| 868 | } |
| 869 | |
| 870 | /** @brief Recursively compute the compatibility decomposition of @p c |
| 871 | * @param d Dynamic string to store decomposition in |
| 872 | * @param c Code point to decompose (must be a valid!) |
| 873 | * @return 0 on success, non-0 on error |
| 874 | */ |
| 875 | static void utf32__decompose_one_compat(struct dynstr_ucs4 *d, uint32_t c) { |
| 876 | utf32__decompose_one_generic(compat); |
| 877 | } |
| 878 | |
| 879 | /** @brief Magic utf32__compositions() return value for Hangul Choseong */ |
| 880 | static const uint32_t utf32__hangul_L[1]; |
| 881 | |
| 882 | /** @brief Return the list of compositions that @p c starts |
| 883 | * @param c Starter code point |
| 884 | * @return Composition list or NULL |
| 885 | * |
| 886 | * For Hangul leading (Choseong) jamo we return the special value |
| 887 | * utf32__hangul_L. These code points are not listed as the targets of |
| 888 | * canonical decompositions (make-unidata checks) so there is no confusion with |
| 889 | * real decompositions here. |
| 890 | */ |
| 891 | static const uint32_t *utf32__compositions(uint32_t c) { |
| 892 | const uint32_t *compositions = utf32__unidata(c)->composed; |
| 893 | |
| 894 | if(compositions) |
| 895 | return compositions; |
| 896 | /* Special-casing for Hangul */ |
| 897 | switch(utf32__grapheme_break(c)) { |
| 898 | default: |
| 899 | return 0; |
| 900 | case unicode_Grapheme_Break_L: |
| 901 | return utf32__hangul_L; |
| 902 | } |
| 903 | } |
| 904 | |
| 905 | /** @brief Composition step |
| 906 | * @param s Start of string |
| 907 | * @param ns Length of string |
| 908 | * @return New length of string |
| 909 | * |
| 910 | * This is called from utf32__decompose_generic() to compose the result string |
| 911 | * in place. |
| 912 | */ |
| 913 | static size_t utf32__compose(uint32_t *s, size_t ns) { |
| 914 | const uint32_t *compositions; |
| 915 | uint32_t *start = s, *t = s, *tt, cc; |
| 916 | |
| 917 | while(ns > 0) { |
| 918 | uint32_t starter = *s++; |
| 919 | int block_starters = 0; |
| 920 | --ns; |
| 921 | /* We don't attempt to compose the following things: |
| 922 | * - final characters whatever kind they are |
| 923 | * - non-starter characters |
| 924 | * - starters that don't take part in a canonical decomposition mapping |
| 925 | */ |
| 926 | if(ns == 0 |
| 927 | || utf32__combining_class(starter) |
| 928 | || !(compositions = utf32__compositions(starter))) { |
| 929 | *t++ = starter; |
| 930 | continue; |
| 931 | } |
| 932 | if(compositions != utf32__hangul_L) { |
| 933 | /* Where we'll put the eventual starter */ |
| 934 | tt = t++; |
| 935 | do { |
| 936 | /* See if we can find composition of starter+*s */ |
| 937 | const uint32_t cchar = *s, *cp = compositions; |
| 938 | while((cc = *cp++)) { |
| 939 | const uint32_t *decomp = utf32__decomposition_canon(cc); |
| 940 | /* We know decomp[0] == starter */ |
| 941 | if(decomp[1] == cchar) |
| 942 | break; |
| 943 | } |
| 944 | if(cc) { |
| 945 | /* Found a composition: cc decomposes to starter,*s */ |
| 946 | starter = cc; |
| 947 | compositions = utf32__compositions(starter); |
| 948 | ++s; |
| 949 | --ns; |
| 950 | } else { |
| 951 | /* No composition found. */ |
| 952 | const int class = utf32__combining_class(*s); |
| 953 | if(class) { |
| 954 | /* Transfer the uncomposable combining character to the output */ |
| 955 | *t++ = *s++; |
| 956 | --ns; |
| 957 | /* All the combining characters of the same class of the |
| 958 | * uncomposable character are blocked by it, but there may be |
| 959 | * others of higher class later. We eat the uncomposable and |
| 960 | * blocked characters and go back round the loop for that higher |
| 961 | * class. */ |
| 962 | while(ns > 0 && utf32__combining_class(*s) == class) { |
| 963 | *t++ = *s++; |
| 964 | --ns; |
| 965 | } |
| 966 | /* Block any subsequent starters */ |
| 967 | block_starters = 1; |
| 968 | } else { |
| 969 | /* The uncombinable character is itself a starter, so we don't |
| 970 | * transfer it to the output but instead go back round the main |
| 971 | * loop. */ |
| 972 | break; |
| 973 | } |
| 974 | } |
| 975 | /* Keep going while there are still characters and the starter takes |
| 976 | * part in some composition */ |
| 977 | } while(ns > 0 && compositions |
| 978 | && (!block_starters || utf32__combining_class(*s))); |
| 979 | /* Store any remaining combining characters */ |
| 980 | while(ns > 0 && utf32__combining_class(*s)) { |
| 981 | *t++ = *s++; |
| 982 | --ns; |
| 983 | } |
| 984 | /* Store the resulting starter */ |
| 985 | *tt = starter; |
| 986 | } else { |
| 987 | /* Special-casing for Hangul |
| 988 | * |
| 989 | * If there are combining characters between the L and the V then they |
| 990 | * will block the V and so no composition happens. Similarly combining |
| 991 | * characters between V and T will block the T and so we only get as far |
| 992 | * as LV. |
| 993 | */ |
| 994 | if(utf32__grapheme_break(*s) == unicode_Grapheme_Break_V) { |
| 995 | const uint32_t V = *s++; |
| 996 | const uint32_t LIndex = starter - LBase; |
| 997 | const uint32_t VIndex = V - VBase; |
| 998 | uint32_t TIndex; |
| 999 | --ns; |
| 1000 | if(ns > 0 |
| 1001 | && utf32__grapheme_break(*s) == unicode_Grapheme_Break_T) { |
| 1002 | /* We have an L V T sequence */ |
| 1003 | const uint32_t T = *s++; |
| 1004 | TIndex = T - TBase; |
| 1005 | --ns; |
| 1006 | } else |
| 1007 | /* It's just L V */ |
| 1008 | TIndex = 0; |
| 1009 | /* Compose to LVT or LV as appropriate */ |
| 1010 | starter = (LIndex * VCount + VIndex) * TCount + TIndex + SBase; |
| 1011 | } /* else we only have L or LV and no V or T */ |
| 1012 | *t++ = starter; |
| 1013 | /* There could be some combining characters that belong to the V or T. |
| 1014 | * These will be treated as non-starter characters at the top of the loop |
| 1015 | * and thuss transferred to the output. */ |
| 1016 | } |
| 1017 | } |
| 1018 | return t - start; |
| 1019 | } |
| 1020 | |
| 1021 | /** @brief Guts of the composition and decomposition functions |
| 1022 | * @param WHICH @c canon or @c compat to choose decomposition |
| 1023 | * @param COMPOSE @c 0 or @c 1 to compose |
| 1024 | */ |
| 1025 | #define utf32__decompose_generic(WHICH, COMPOSE) do { \ |
| 1026 | struct dynstr_ucs4 d; \ |
| 1027 | uint32_t c; \ |
| 1028 | \ |
| 1029 | dynstr_ucs4_init(&d); \ |
| 1030 | while(ns) { \ |
| 1031 | c = *s++; \ |
| 1032 | if((c >= 0xD800 && c <= 0xDFFF) || c > 0x10FFFF) \ |
| 1033 | goto error; \ |
| 1034 | utf32__decompose_one_##WHICH(&d, c); \ |
| 1035 | --ns; \ |
| 1036 | } \ |
| 1037 | if(utf32__canonical_ordering(d.vec, d.nvec)) \ |
| 1038 | goto error; \ |
| 1039 | if(COMPOSE) \ |
| 1040 | d.nvec = utf32__compose(d.vec, d.nvec); \ |
| 1041 | dynstr_ucs4_terminate(&d); \ |
| 1042 | if(ndp) \ |
| 1043 | *ndp = d.nvec; \ |
| 1044 | return d.vec; \ |
| 1045 | error: \ |
| 1046 | xfree(d.vec); \ |
| 1047 | return 0; \ |
| 1048 | } while(0) |
| 1049 | |
| 1050 | /** @brief Canonically decompose @p [s,s+ns) |
| 1051 | * @param s Pointer to string |
| 1052 | * @param ns Length of string |
| 1053 | * @param ndp Where to store length of result |
| 1054 | * @return Pointer to result string, or NULL on error |
| 1055 | * |
| 1056 | * Computes NFD (Normalization Form D) of the string at @p s. This implies |
| 1057 | * performing all canonical decompositions and then normalizing the order of |
| 1058 | * combining characters. |
| 1059 | * |
| 1060 | * Returns NULL if the string is not valid for either of the following reasons: |
| 1061 | * - it codes for a UTF-16 surrogate |
| 1062 | * - it codes for a value outside the unicode code space |
| 1063 | * |
| 1064 | * See also: |
| 1065 | * - utf32_decompose_compat() |
| 1066 | * - utf32_compose_canon() |
| 1067 | */ |
| 1068 | uint32_t *utf32_decompose_canon(const uint32_t *s, size_t ns, size_t *ndp) { |
| 1069 | utf32__decompose_generic(canon, 0); |
| 1070 | } |
| 1071 | |
| 1072 | /** @brief Compatibility decompose @p [s,s+ns) |
| 1073 | * @param s Pointer to string |
| 1074 | * @param ns Length of string |
| 1075 | * @param ndp Where to store length of result |
| 1076 | * @return Pointer to result string, or NULL on error |
| 1077 | * |
| 1078 | * Computes NFKD (Normalization Form KD) of the string at @p s. This implies |
| 1079 | * performing all canonical and compatibility decompositions and then |
| 1080 | * normalizing the order of combining characters. |
| 1081 | * |
| 1082 | * Returns NULL if the string is not valid for either of the following reasons: |
| 1083 | * - it codes for a UTF-16 surrogate |
| 1084 | * - it codes for a value outside the unicode code space |
| 1085 | * |
| 1086 | * See also: |
| 1087 | * - utf32_decompose_canon() |
| 1088 | * - utf32_compose_compat() |
| 1089 | */ |
| 1090 | uint32_t *utf32_decompose_compat(const uint32_t *s, size_t ns, size_t *ndp) { |
| 1091 | utf32__decompose_generic(compat, 0); |
| 1092 | } |
| 1093 | |
| 1094 | /** @brief Canonically compose @p [s,s+ns) |
| 1095 | * @param s Pointer to string |
| 1096 | * @param ns Length of string |
| 1097 | * @param ndp Where to store length of result |
| 1098 | * @return Pointer to result string, or NULL on error |
| 1099 | * |
| 1100 | * Computes NFC (Normalization Form C) of the string at @p s. This implies |
| 1101 | * performing all canonical decompositions, normalizing the order of combining |
| 1102 | * characters and then composing all unblocked primary compositables. |
| 1103 | * |
| 1104 | * Returns NULL if the string is not valid for either of the following reasons: |
| 1105 | * - it codes for a UTF-16 surrogate |
| 1106 | * - it codes for a value outside the unicode code space |
| 1107 | * |
| 1108 | * See also: |
| 1109 | * - utf32_compose_compat() |
| 1110 | * - utf32_decompose_canon() |
| 1111 | */ |
| 1112 | uint32_t *utf32_compose_canon(const uint32_t *s, size_t ns, size_t *ndp) { |
| 1113 | utf32__decompose_generic(canon, 1); |
| 1114 | } |
| 1115 | |
| 1116 | /** @brief Compatibility compose @p [s,s+ns) |
| 1117 | * @param s Pointer to string |
| 1118 | * @param ns Length of string |
| 1119 | * @param ndp Where to store length of result |
| 1120 | * @return Pointer to result string, or NULL on error |
| 1121 | * |
| 1122 | * Computes NFKC (Normalization Form KC) of the string at @p s. This implies |
| 1123 | * performing all canonical and compatibility decompositions, normalizing the |
| 1124 | * order of combining characters and then composing all unblocked primary |
| 1125 | * compositables. |
| 1126 | * |
| 1127 | * Returns NULL if the string is not valid for either of the following reasons: |
| 1128 | * - it codes for a UTF-16 surrogate |
| 1129 | * - it codes for a value outside the unicode code space |
| 1130 | * |
| 1131 | * See also: |
| 1132 | * - utf32_compose_canon() |
| 1133 | * - utf32_decompose_compat() |
| 1134 | */ |
| 1135 | uint32_t *utf32_compose_compat(const uint32_t *s, size_t ns, size_t *ndp) { |
| 1136 | utf32__decompose_generic(compat, 1); |
| 1137 | } |
| 1138 | |
| 1139 | /** @brief Single-character case-fold and decompose operation */ |
| 1140 | #define utf32__casefold_one(WHICH) do { \ |
| 1141 | const uint32_t *cf = utf32__unidata(c)->casefold; \ |
| 1142 | if(cf) { \ |
| 1143 | /* Found a case-fold mapping in the table */ \ |
| 1144 | while(*cf) \ |
| 1145 | utf32__decompose_one_##WHICH(&d, *cf++); \ |
| 1146 | } else \ |
| 1147 | utf32__decompose_one_##WHICH(&d, c); \ |
| 1148 | } while(0) |
| 1149 | |
| 1150 | /** @brief Case-fold @p [s,s+ns) |
| 1151 | * @param s Pointer to string |
| 1152 | * @param ns Length of string |
| 1153 | * @param ndp Where to store length of result |
| 1154 | * @return Pointer to result string, or NULL on error |
| 1155 | * |
| 1156 | * Case-fold the string at @p s according to full default case-folding rules |
| 1157 | * (s3.13) for caseless matching. The result will be in NFD. |
| 1158 | * |
| 1159 | * Returns NULL if the string is not valid for either of the following reasons: |
| 1160 | * - it codes for a UTF-16 surrogate |
| 1161 | * - it codes for a value outside the unicode code space |
| 1162 | */ |
| 1163 | uint32_t *utf32_casefold_canon(const uint32_t *s, size_t ns, size_t *ndp) { |
| 1164 | struct dynstr_ucs4 d; |
| 1165 | uint32_t c; |
| 1166 | size_t n; |
| 1167 | uint32_t *ss = 0; |
| 1168 | |
| 1169 | /* If the canonical decomposition of the string includes any combining |
| 1170 | * character that case-folds to a non-combining character then we must |
| 1171 | * normalize before we fold. In Unicode 5.0.0 this means 0345 COMBINING |
| 1172 | * GREEK YPOGEGRAMMENI in its decomposition and the various characters that |
| 1173 | * canonically decompose to it. */ |
| 1174 | for(n = 0; n < ns; ++n) |
| 1175 | if(utf32__unidata(s[n])->flags & unicode_normalize_before_casefold) |
| 1176 | break; |
| 1177 | if(n < ns) { |
| 1178 | /* We need a preliminary decomposition */ |
| 1179 | if(!(ss = utf32_decompose_canon(s, ns, &ns))) |
| 1180 | return 0; |
| 1181 | s = ss; |
| 1182 | } |
| 1183 | dynstr_ucs4_init(&d); |
| 1184 | while(ns) { |
| 1185 | c = *s++; |
| 1186 | if((c >= 0xD800 && c <= 0xDFFF) || c > 0x10FFFF) |
| 1187 | goto error; |
| 1188 | utf32__casefold_one(canon); |
| 1189 | --ns; |
| 1190 | } |
| 1191 | if(utf32__canonical_ordering(d.vec, d.nvec)) |
| 1192 | goto error; |
| 1193 | dynstr_ucs4_terminate(&d); |
| 1194 | if(ndp) |
| 1195 | *ndp = d.nvec; |
| 1196 | return d.vec; |
| 1197 | error: |
| 1198 | xfree(d.vec); |
| 1199 | xfree(ss); |
| 1200 | return 0; |
| 1201 | } |
| 1202 | |
| 1203 | /** @brief Compatibility case-fold @p [s,s+ns) |
| 1204 | * @param s Pointer to string |
| 1205 | * @param ns Length of string |
| 1206 | * @param ndp Where to store length of result |
| 1207 | * @return Pointer to result string, or NULL on error |
| 1208 | * |
| 1209 | * Case-fold the string at @p s according to full default case-folding rules |
| 1210 | * (s3.13) for compatibility caseless matching. The result will be in NFKD. |
| 1211 | * |
| 1212 | * Returns NULL if the string is not valid for either of the following reasons: |
| 1213 | * - it codes for a UTF-16 surrogate |
| 1214 | * - it codes for a value outside the unicode code space |
| 1215 | */ |
| 1216 | uint32_t *utf32_casefold_compat(const uint32_t *s, size_t ns, size_t *ndp) { |
| 1217 | struct dynstr_ucs4 d; |
| 1218 | uint32_t c; |
| 1219 | size_t n; |
| 1220 | uint32_t *ss = 0; |
| 1221 | |
| 1222 | for(n = 0; n < ns; ++n) |
| 1223 | if(utf32__unidata(s[n])->flags & unicode_normalize_before_casefold) |
| 1224 | break; |
| 1225 | if(n < ns) { |
| 1226 | /* We need a preliminary _canonical_ decomposition */ |
| 1227 | if(!(ss = utf32_decompose_canon(s, ns, &ns))) |
| 1228 | return 0; |
| 1229 | s = ss; |
| 1230 | } |
| 1231 | /* This computes NFKD(toCaseFold(s)) */ |
| 1232 | #define compat_casefold_middle() do { \ |
| 1233 | dynstr_ucs4_init(&d); \ |
| 1234 | while(ns) { \ |
| 1235 | c = *s++; \ |
| 1236 | if((c >= 0xD800 && c <= 0xDFFF) || c > 0x10FFFF) \ |
| 1237 | goto error; \ |
| 1238 | utf32__casefold_one(compat); \ |
| 1239 | --ns; \ |
| 1240 | } \ |
| 1241 | if(utf32__canonical_ordering(d.vec, d.nvec)) \ |
| 1242 | goto error; \ |
| 1243 | } while(0) |
| 1244 | /* Do the inner (NFKD o toCaseFold) */ |
| 1245 | compat_casefold_middle(); |
| 1246 | /* We can do away with the NFD'd copy of the input now */ |
| 1247 | xfree(ss); |
| 1248 | s = ss = d.vec; |
| 1249 | ns = d.nvec; |
| 1250 | /* Do the outer (NFKD o toCaseFold) */ |
| 1251 | compat_casefold_middle(); |
| 1252 | /* That's all */ |
| 1253 | dynstr_ucs4_terminate(&d); |
| 1254 | if(ndp) |
| 1255 | *ndp = d.nvec; |
| 1256 | return d.vec; |
| 1257 | error: |
| 1258 | xfree(d.vec); |
| 1259 | xfree(ss); |
| 1260 | return 0; |
| 1261 | } |
| 1262 | |
| 1263 | /** @brief Order a pair of UTF-32 strings |
| 1264 | * @param a First 0-terminated string |
| 1265 | * @param b Second 0-terminated string |
| 1266 | * @return -1, 0 or 1 for a less than, equal to or greater than b |
| 1267 | * |
| 1268 | * "Comparable to strcmp() at its best." |
| 1269 | */ |
| 1270 | int utf32_cmp(const uint32_t *a, const uint32_t *b) { |
| 1271 | while(*a && *b && *a == *b) { |
| 1272 | ++a; |
| 1273 | ++b; |
| 1274 | } |
| 1275 | return *a < *b ? -1 : (*a > *b ? 1 : 0); |
| 1276 | } |
| 1277 | |
| 1278 | /** @brief Identify a grapheme cluster boundary |
| 1279 | * @param s Start of string (must be NFD) |
| 1280 | * @param ns Length of string |
| 1281 | * @param n Index within string (in [0,ns].) |
| 1282 | * @return 1 at a grapheme cluster boundary, 0 otherwise |
| 1283 | * |
| 1284 | * This function identifies default grapheme cluster boundaries as described in |
| 1285 | * UAX #29 s3. It returns non-0 if @p n points at the code point just after a |
| 1286 | * grapheme cluster boundary (including the hypothetical code point just after |
| 1287 | * the end of the string). |
| 1288 | * |
| 1289 | * This function uses utf32_iterator_set() internally; see that function for |
| 1290 | * remarks on performance. |
| 1291 | */ |
| 1292 | int utf32_is_grapheme_boundary(const uint32_t *s, size_t ns, size_t n) { |
| 1293 | struct utf32_iterator_data it[1]; |
| 1294 | |
| 1295 | utf32__iterator_init(it, s, ns, n); |
| 1296 | return utf32_iterator_grapheme_boundary(it); |
| 1297 | } |
| 1298 | |
| 1299 | /** @brief Identify a word boundary |
| 1300 | * @param s Start of string (must be NFD) |
| 1301 | * @param ns Length of string |
| 1302 | * @param n Index within string (in [0,ns].) |
| 1303 | * @return 1 at a word boundary, 0 otherwise |
| 1304 | * |
| 1305 | * This function identifies default word boundaries as described in UAX #29 s4. |
| 1306 | * It returns non-0 if @p n points at the code point just after a word boundary |
| 1307 | * (including the hypothetical code point just after the end of the string). |
| 1308 | * |
| 1309 | * This function uses utf32_iterator_set() internally; see that function for |
| 1310 | * remarks on performance. |
| 1311 | */ |
| 1312 | int utf32_is_word_boundary(const uint32_t *s, size_t ns, size_t n) { |
| 1313 | struct utf32_iterator_data it[1]; |
| 1314 | |
| 1315 | utf32__iterator_init(it, s, ns, n); |
| 1316 | return utf32_iterator_word_boundary(it); |
| 1317 | } |
| 1318 | |
| 1319 | /** @brief Split [s,ns) into multiple words |
| 1320 | * @param s Pointer to start of string |
| 1321 | * @param ns Length of string |
| 1322 | * @param nwp Where to store word count, or NULL |
| 1323 | * @param wbreak Word_Break property tailor, or NULL |
| 1324 | * @return Pointer to array of pointers to words |
| 1325 | * |
| 1326 | * The returned array is terminated by a NULL pointer and individual |
| 1327 | * strings are 0-terminated. |
| 1328 | */ |
| 1329 | uint32_t **utf32_word_split(const uint32_t *s, size_t ns, size_t *nwp, |
| 1330 | unicode_property_tailor *wbreak) { |
| 1331 | struct utf32_iterator_data it[1]; |
| 1332 | size_t b1 = 0, b2 = 0 ,i; |
| 1333 | int isword; |
| 1334 | struct vector32 v32[1]; |
| 1335 | uint32_t *w; |
| 1336 | |
| 1337 | vector32_init(v32); |
| 1338 | utf32__iterator_init(it, s, ns, 0); |
| 1339 | it->word_break = wbreak; |
| 1340 | /* Work our way through the string stopping at each word break. */ |
| 1341 | do { |
| 1342 | if(utf32_iterator_word_boundary(it)) { |
| 1343 | /* We've found a new boundary */ |
| 1344 | b1 = b2; |
| 1345 | b2 = it->n; |
| 1346 | /*fprintf(stderr, "[%zu, %zu) is a candidate word\n", b1, b2);*/ |
| 1347 | /* Inspect the characters between the boundary and form an opinion as to |
| 1348 | * whether they are a word or not */ |
| 1349 | isword = 0; |
| 1350 | for(i = b1; i < b2; ++i) { |
| 1351 | switch(utf32__iterator_word_break(it, it->s[i])) { |
| 1352 | case unicode_Word_Break_ALetter: |
| 1353 | case unicode_Word_Break_Numeric: |
| 1354 | case unicode_Word_Break_Katakana: |
| 1355 | isword = 1; |
| 1356 | break; |
| 1357 | default: |
| 1358 | break; |
| 1359 | } |
| 1360 | } |
| 1361 | /* If it's a word add it to the list of results */ |
| 1362 | if(isword) { |
| 1363 | w = xcalloc(b2 - b1 + 1, sizeof(uint32_t)); |
| 1364 | memcpy(w, it->s + b1, (b2 - b1) * sizeof (uint32_t)); |
| 1365 | vector32_append(v32, w); |
| 1366 | } |
| 1367 | } |
| 1368 | } while(!utf32_iterator_advance(it, 1)); |
| 1369 | vector32_terminate(v32); |
| 1370 | if(nwp) |
| 1371 | *nwp = v32->nvec; |
| 1372 | return v32->vec; |
| 1373 | } |
| 1374 | |
| 1375 | /*@}*/ |
| 1376 | /** @defgroup utf8 Functions that operate on UTF-8 strings */ |
| 1377 | /*@{*/ |
| 1378 | |
| 1379 | /** @brief Wrapper to transform a UTF-8 string using the UTF-32 function */ |
| 1380 | #define utf8__transform(FN) do { \ |
| 1381 | uint32_t *to32 = 0, *decomp32 = 0; \ |
| 1382 | size_t nto32, ndecomp32; \ |
| 1383 | char *decomp8 = 0; \ |
| 1384 | \ |
| 1385 | if(!(to32 = utf8_to_utf32(s, ns, &nto32))) goto error; \ |
| 1386 | if(!(decomp32 = FN(to32, nto32, &ndecomp32))) goto error; \ |
| 1387 | decomp8 = utf32_to_utf8(decomp32, ndecomp32, ndp); \ |
| 1388 | error: \ |
| 1389 | xfree(to32); \ |
| 1390 | xfree(decomp32); \ |
| 1391 | return decomp8; \ |
| 1392 | } while(0) |
| 1393 | |
| 1394 | /** @brief Canonically decompose @p [s,s+ns) |
| 1395 | * @param s Pointer to string |
| 1396 | * @param ns Length of string |
| 1397 | * @param ndp Where to store length of result |
| 1398 | * @return Pointer to result string, or NULL on error |
| 1399 | * |
| 1400 | * Computes NFD (Normalization Form D) of the string at @p s. This implies |
| 1401 | * performing all canonical decompositions and then normalizing the order of |
| 1402 | * combining characters. |
| 1403 | * |
| 1404 | * Returns NULL if the string is not valid; see utf8_to_utf32() for reasons why |
| 1405 | * this might be. |
| 1406 | * |
| 1407 | * See also: |
| 1408 | * - utf32_decompose_canon(). |
| 1409 | * - utf8_decompose_compat() |
| 1410 | * - utf8_compose_canon() |
| 1411 | */ |
| 1412 | char *utf8_decompose_canon(const char *s, size_t ns, size_t *ndp) { |
| 1413 | utf8__transform(utf32_decompose_canon); |
| 1414 | } |
| 1415 | |
| 1416 | /** @brief Compatibility decompose @p [s,s+ns) |
| 1417 | * @param s Pointer to string |
| 1418 | * @param ns Length of string |
| 1419 | * @param ndp Where to store length of result |
| 1420 | * @return Pointer to result string, or NULL on error |
| 1421 | * |
| 1422 | * Computes NFKD (Normalization Form KD) of the string at @p s. This implies |
| 1423 | * performing all canonical and compatibility decompositions and then |
| 1424 | * normalizing the order of combining characters. |
| 1425 | * |
| 1426 | * Returns NULL if the string is not valid; see utf8_to_utf32() for reasons why |
| 1427 | * this might be. |
| 1428 | * |
| 1429 | * See also: |
| 1430 | * - utf32_decompose_compat(). |
| 1431 | * - utf8_decompose_canon() |
| 1432 | * - utf8_compose_compat() |
| 1433 | */ |
| 1434 | char *utf8_decompose_compat(const char *s, size_t ns, size_t *ndp) { |
| 1435 | utf8__transform(utf32_decompose_compat); |
| 1436 | } |
| 1437 | |
| 1438 | /** @brief Canonically compose @p [s,s+ns) |
| 1439 | * @param s Pointer to string |
| 1440 | * @param ns Length of string |
| 1441 | * @param ndp Where to store length of result |
| 1442 | * @return Pointer to result string, or NULL on error |
| 1443 | * |
| 1444 | * Computes NFC (Normalization Form C) of the string at @p s. This implies |
| 1445 | * performing all canonical decompositions, normalizing the order of combining |
| 1446 | * characters and then composing all unblocked primary compositables. |
| 1447 | * |
| 1448 | * Returns NULL if the string is not valid; see utf8_to_utf32() for reasons why |
| 1449 | * this might be. |
| 1450 | * |
| 1451 | * See also: |
| 1452 | * - utf32_compose_canon() |
| 1453 | * - utf8_compose_compat() |
| 1454 | * - utf8_decompose_canon() |
| 1455 | */ |
| 1456 | char *utf8_compose_canon(const char *s, size_t ns, size_t *ndp) { |
| 1457 | utf8__transform(utf32_compose_canon); |
| 1458 | } |
| 1459 | |
| 1460 | /** @brief Compatibility compose @p [s,s+ns) |
| 1461 | * @param s Pointer to string |
| 1462 | * @param ns Length of string |
| 1463 | * @param ndp Where to store length of result |
| 1464 | * @return Pointer to result string, or NULL on error |
| 1465 | * |
| 1466 | * Computes NFKC (Normalization Form KC) of the string at @p s. This implies |
| 1467 | * performing all canonical and compatibility decompositions, normalizing the |
| 1468 | * order of combining characters and then composing all unblocked primary |
| 1469 | * compositables. |
| 1470 | * |
| 1471 | * Returns NULL if the string is not valid; see utf8_to_utf32() for reasons why |
| 1472 | * this might be. |
| 1473 | * |
| 1474 | * See also: |
| 1475 | * - utf32_compose_compat() |
| 1476 | * - utf8_compose_canon() |
| 1477 | * - utf8_decompose_compat() |
| 1478 | */ |
| 1479 | char *utf8_compose_compat(const char *s, size_t ns, size_t *ndp) { |
| 1480 | utf8__transform(utf32_compose_compat); |
| 1481 | } |
| 1482 | |
| 1483 | /** @brief Case-fold @p [s,s+ns) |
| 1484 | * @param s Pointer to string |
| 1485 | * @param ns Length of string |
| 1486 | * @param ndp Where to store length of result |
| 1487 | * @return Pointer to result string, or NULL on error |
| 1488 | * |
| 1489 | * Case-fold the string at @p s according to full default case-folding rules |
| 1490 | * (s3.13). The result will be in NFD. |
| 1491 | * |
| 1492 | * Returns NULL if the string is not valid; see utf8_to_utf32() for reasons why |
| 1493 | * this might be. |
| 1494 | */ |
| 1495 | char *utf8_casefold_canon(const char *s, size_t ns, size_t *ndp) { |
| 1496 | utf8__transform(utf32_casefold_canon); |
| 1497 | } |
| 1498 | |
| 1499 | /** @brief Compatibility case-fold @p [s,s+ns) |
| 1500 | * @param s Pointer to string |
| 1501 | * @param ns Length of string |
| 1502 | * @param ndp Where to store length of result |
| 1503 | * @return Pointer to result string, or NULL on error |
| 1504 | * |
| 1505 | * Case-fold the string at @p s according to full default case-folding rules |
| 1506 | * (s3.13). The result will be in NFKD. |
| 1507 | * |
| 1508 | * Returns NULL if the string is not valid; see utf8_to_utf32() for reasons why |
| 1509 | * this might be. |
| 1510 | */ |
| 1511 | char *utf8_casefold_compat(const char *s, size_t ns, size_t *ndp) { |
| 1512 | utf8__transform(utf32_casefold_compat); |
| 1513 | } |
| 1514 | |
| 1515 | /** @brief Split [s,ns) into multiple words |
| 1516 | * @param s Pointer to start of string |
| 1517 | * @param ns Length of string |
| 1518 | * @param nwp Where to store word count, or NULL |
| 1519 | * @param wbreak Word_Break property tailor, or NULL |
| 1520 | * @return Pointer to array of pointers to words |
| 1521 | * |
| 1522 | * The returned array is terminated by a NULL pointer and individual |
| 1523 | * strings are 0-terminated. |
| 1524 | */ |
| 1525 | char **utf8_word_split(const char *s, size_t ns, size_t *nwp, |
| 1526 | unicode_property_tailor *wbreak) { |
| 1527 | uint32_t *to32 = 0, **v32 = 0; |
| 1528 | size_t nto32, nv, n; |
| 1529 | char **v8 = 0, **ret = 0; |
| 1530 | |
| 1531 | if(!(to32 = utf8_to_utf32(s, ns, &nto32))) goto error; |
| 1532 | if(!(v32 = utf32_word_split(to32, nto32, &nv, wbreak))) goto error; |
| 1533 | v8 = xcalloc(sizeof (char *), nv + 1); |
| 1534 | for(n = 0; n < nv; ++n) |
| 1535 | if(!(v8[n] = utf32_to_utf8(v32[n], utf32_len(v32[n]), 0))) |
| 1536 | goto error; |
| 1537 | ret = v8; |
| 1538 | *nwp = nv; |
| 1539 | v8 = 0; /* don't free */ |
| 1540 | error: |
| 1541 | if(v8) { |
| 1542 | for(n = 0; n < nv; ++n) |
| 1543 | xfree(v8[n]); |
| 1544 | xfree(v8); |
| 1545 | } |
| 1546 | if(v32) { |
| 1547 | for(n = 0; n < nv; ++n) |
| 1548 | xfree(v32[n]); |
| 1549 | xfree(v32); |
| 1550 | } |
| 1551 | xfree(to32); |
| 1552 | return ret; |
| 1553 | } |
| 1554 | |
| 1555 | |
| 1556 | /*@}*/ |
| 1557 | |
| 1558 | /* |
| 1559 | Local Variables: |
| 1560 | c-basic-offset:2 |
| 1561 | comment-column:40 |
| 1562 | fill-column:79 |
| 1563 | indent-tabs-mode:nil |
| 1564 | End: |
| 1565 | */ |