2 * This file is part of DisOrder
3 * Copyright (C) 2007 Richard Kettlewell
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.
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.
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
20 /** @file lib/unicode.c
21 * @brief Unicode support functions
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.
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
32 * As the code stands this guarantee is not well met!
39 #include <stdio.h> /* TODO */
46 /** @defgroup utf32props Unicode Code Point Properties */
49 static const struct unidata *utf32__unidata_hard(uint32_t c);
51 /** @brief Find definition of code point @p c
53 * @return Pointer to @ref unidata structure for @p c
55 * @p c can be any 32-bit value, a sensible value will be returned regardless.
56 * The returned pointer is NOT guaranteed to be unique to @p c.
58 static inline const struct unidata *utf32__unidata(uint32_t c) {
59 /* The bottom half of the table contains almost everything of interest
60 * and we can just return the right thing straight away */
61 if(c < UNICODE_BREAK_START)
62 return &unidata[c / UNICODE_MODULUS][c % UNICODE_MODULUS];
64 return utf32__unidata_hard(c);
67 /** @brief Find definition of code point @p c
69 * @return Pointer to @ref unidata structure for @p c
71 * @p c can be any 32-bit value, a sensible value will be returned regardless.
72 * The returned pointer is NOT guaranteed to be unique to @p c.
74 * Don't use this function (although it will work fine) - use utf32__unidata()
77 static const struct unidata *utf32__unidata_hard(uint32_t c) {
78 if(c < UNICODE_BREAK_START)
79 return &unidata[c / UNICODE_MODULUS][c % UNICODE_MODULUS];
80 /* Within the break everything is unassigned */
81 if(c < UNICODE_BREAK_END)
82 return utf32__unidata(0xFFFF); /* guaranteed to be Cn */
83 /* Planes 15 and 16 are (mostly) private use */
84 if((c >= 0xF0000 && c <= 0xFFFFD)
85 || (c >= 0x100000 && c <= 0x10FFFD))
86 return utf32__unidata(0xE000); /* first Co code point */
87 /* Everything else above the break top is unassigned */
88 if(c >= UNICODE_BREAK_TOP)
89 return utf32__unidata(0xFFFF); /* guaranteed to be Cn */
90 /* Currently the rest is language tags and variation selectors */
91 c -= (UNICODE_BREAK_END - UNICODE_BREAK_START);
92 return &unidata[c / UNICODE_MODULUS][c % UNICODE_MODULUS];
95 /** @brief Return the combining class of @p c
97 * @return Combining class of @p c
99 * @p c can be any 32-bit value, a sensible value will be returned regardless.
101 static inline int utf32__combining_class(uint32_t c) {
102 return utf32__unidata(c)->ccc;
105 /** @brief Return the General_Category value for @p c
107 * @return General_Category property value
109 * @p c can be any 32-bit value, a sensible value will be returned regardless.
111 static inline enum unicode_General_Category utf32__general_category(uint32_t c) {
112 return utf32__unidata(c)->general_category;
115 /** @brief Determine Grapheme_Break property
116 * @param c Code point
117 * @return Grapheme_Break property value of @p c
119 * @p c can be any 32-bit value, a sensible value will be returned regardless.
121 static inline enum unicode_Grapheme_Break utf32__grapheme_break(uint32_t c) {
122 return utf32__unidata(c)->grapheme_break;
125 /** @brief Determine Word_Break property
126 * @param c Code point
127 * @return Word_Break property value of @p c
129 * @p c can be any 32-bit value, a sensible value will be returned regardless.
131 static inline enum unicode_Word_Break utf32__word_break(uint32_t c) {
132 return utf32__unidata(c)->word_break;
135 /** @brief Determine Sentence_Break property
136 * @param c Code point
137 * @return Word_Break property value of @p c
139 * @p c can be any 32-bit value, a sensible value will be returned regardless.
141 static inline enum unicode_Sentence_Break utf32__sentence_break(uint32_t c) {
142 return utf32__unidata(c)->sentence_break;
145 /** @brief Return true if @p c is ignorable for boundary specifications
146 * @param wb Word break property value
147 * @return non-0 if @p wb is unicode_Word_Break_Extend or unicode_Word_Break_Format
149 static inline int utf32__boundary_ignorable(enum unicode_Word_Break wb) {
150 return (wb == unicode_Word_Break_Extend
151 || wb == unicode_Word_Break_Format);
155 /** @defgroup utftransform Functions that transform between different Unicode encoding forms */
158 /** @brief Convert UTF-32 to UTF-8
159 * @param s Source string
160 * @param ns Length of source string in code points
161 * @param ndp Where to store length of destination string (or NULL)
162 * @return Newly allocated destination string or NULL on error
164 * If the UTF-32 is not valid then NULL is returned. A UTF-32 code point is
166 * - it codes for a UTF-16 surrogate
167 * - it codes for a value outside the unicode code space
169 * The return value is always 0-terminated. The value returned via @p *ndp
170 * does not include the terminator.
172 char *utf32_to_utf8(const uint32_t *s, size_t ns, size_t *ndp) {
180 dynstr_append(&d, c);
181 else if(c < 0x0800) {
182 dynstr_append(&d, 0xC0 | (c >> 6));
183 dynstr_append(&d, 0x80 | (c & 0x3F));
184 } else if(c < 0x10000) {
185 if(c >= 0xD800 && c <= 0xDFFF)
187 dynstr_append(&d, 0xE0 | (c >> 12));
188 dynstr_append(&d, 0x80 | ((c >> 6) & 0x3F));
189 dynstr_append(&d, 0x80 | (c & 0x3F));
190 } else if(c < 0x110000) {
191 dynstr_append(&d, 0xF0 | (c >> 18));
192 dynstr_append(&d, 0x80 | ((c >> 12) & 0x3F));
193 dynstr_append(&d, 0x80 | ((c >> 6) & 0x3F));
194 dynstr_append(&d, 0x80 | (c & 0x3F));
199 dynstr_terminate(&d);
208 /** @brief Convert UTF-8 to UTF-32
209 * @param s Source string
210 * @param ns Length of source string in code points
211 * @param ndp Where to store length of destination string (or NULL)
212 * @return Newly allocated destination string or NULL
214 * The return value is always 0-terminated. The value returned via @p *ndp
215 * does not include the terminator.
217 * If the UTF-8 is not valid then NULL is returned. A UTF-8 sequence
218 * for a code point is invalid if:
219 * - it is not the shortest possible sequence for the code point
220 * - it codes for a UTF-16 surrogate
221 * - it codes for a value outside the unicode code space
223 uint32_t *utf8_to_utf32(const char *s, size_t ns, size_t *ndp) {
224 struct dynstr_ucs4 d;
226 const uint8_t *ss = (const uint8_t *)s;
228 dynstr_ucs4_init(&d);
232 /* Acceptable UTF-8 is that which codes for Unicode Scalar Values
233 * (Unicode 5.0.0 s3.9 D76)
236 * 7 data bits gives 0x00 - 0x7F and all are acceptable
239 * 11 data bits gives 0x0000 - 0x07FF but only 0x0080 - 0x07FF acceptable
241 * 1110xxxx 10xxxxxx 10xxxxxx
242 * 16 data bits gives 0x0000 - 0xFFFF but only 0x0800 - 0xFFFF acceptable
243 * (and UTF-16 surrogates are not acceptable)
245 * 11110xxx 10xxxxxx 10xxxxxx 10xxxxxx
246 * 21 data bits gives 0x00000000 - 0x001FFFFF
247 * but only 0x00010000 - 0x0010FFFF are acceptable
249 * It is NOT always the case that the data bits in the first byte are
250 * always non-0 for the acceptable values, so we do a separate check after
256 if(ns < 1) goto error;
259 if((c & 0xC0) != 0x80) goto error;
260 c32 = (c32 << 6) | (c & 0x3F);
261 if(c32 < 0x80) goto error;
262 } else if(c <= 0xEF) {
263 if(ns < 2) goto error;
266 if((c & 0xC0) != 0x80) goto error;
267 c32 = (c32 << 6) | (c & 0x3F);
269 if((c & 0xC0) != 0x80) goto error;
270 c32 = (c32 << 6) | (c & 0x3F);
271 if(c32 < 0x0800 || (c32 >= 0xD800 && c32 <= 0xDFFF)) goto error;
272 } else if(c <= 0xF7) {
273 if(ns < 3) goto error;
276 if((c & 0xC0) != 0x80) goto error;
277 c32 = (c32 << 6) | (c & 0x3F);
279 if((c & 0xC0) != 0x80) goto error;
280 c32 = (c32 << 6) | (c & 0x3F);
282 if((c & 0xC0) != 0x80) goto error;
283 c32 = (c32 << 6) | (c & 0x3F);
284 if(c32 < 0x00010000 || c32 > 0x0010FFFF) goto error;
287 dynstr_ucs4_append(&d, c32);
289 dynstr_ucs4_terminate(&d);
299 /** @defgroup utf32iterator UTF-32 string iterators */
302 struct utf32_iterator_data {
303 /** @brief Start of string */
306 /** @brief Length of string */
309 /** @brief Current position */
312 /** @brief Last two non-ignorable characters or (uint32_t)-1
314 * last[1] is the non-Extend/Format character just before position @p n;
315 * last[0] is the one just before that.
317 * Exception 1: if there is no such non-Extend/Format character then an
318 * Extend/Format character is accepted instead.
320 * Exception 2: if there is no such character even taking that into account
321 * the value is (uint32_t)-1.
326 /** @brief Create a new iterator pointing at the start of a string
327 * @param s Start of string
328 * @param ns Length of string
329 * @return New iterator
331 utf32_iterator utf32_iterator_new(const uint32_t *s, size_t ns) {
332 utf32_iterator it = xmalloc(sizeof *it);
336 it->last[0] = it->last[1] = -1;
340 /** @brief Initialize an internal private iterator
342 * @param s Start of string
343 * @param ns Length of string
344 * @param n Absolute position
346 static void utf32__iterator_init(utf32_iterator it,
347 const uint32_t *s, size_t ns, size_t n) {
351 it->last[0] = it->last[1] = -1;
352 utf32_iterator_advance(it, n);
355 /** @brief Destroy an iterator
358 void utf32_iterator_destroy(utf32_iterator it) {
362 /** @brief Find the current position of an interator
365 size_t utf32_iterator_where(utf32_iterator it) {
369 /** @brief Set an iterator's absolute position
371 * @param n Absolute position
372 * @return 0 on success, non-0 on error
374 * It is an error to position the iterator outside the string (but acceptable
375 * to point it at the hypothetical post-final character). If an invalid value
376 * of @p n is specified then the iterator is not changed.
378 int utf32_iterator_set(utf32_iterator it, size_t n) {
379 /* TODO figure out how far we must back up to be able to re-synchronize; see
387 it->last[0] = it->last[1] = -1;
389 return utf32_iterator_advance(it, n);
392 /** @brief Advance an iterator
394 * @param count Number of code points to advance by
395 * @return 0 on success, non-0 on error
397 * It is an error to advance an iterator beyond the hypothetical post-final
398 * character of the string. If an invalid value of @p n is specified then the
399 * iterator is not changed.
401 * This function has O(n) time complexity: it works by advancing naively
402 * forwards through the string.
404 int utf32_iterator_advance(utf32_iterator it, size_t count) {
405 if(count <= it->ns - it->n) {
407 const uint32_t c = it->s[it->n];
408 const enum unicode_Word_Break wb = utf32__word_break(c);
409 if(it->last[1] == (uint32_t)-1
410 || !utf32__boundary_ignorable(wb)) {
411 it->last[0] = it->last[1];
422 /** @brief Find the current code point
424 * @return Current code point or 0
426 * If the iterator points at the hypothetical post-final character of the
427 * string then 0 is returned. NB that this doesn't mean that there aren't any
428 * 0 code points inside the string!
430 uint32_t utf32_iterator_code(utf32_iterator it) {
437 /** @brief Test for a grapheme boundary
439 * @return Non-0 if pointing just after a grapheme boundary, otherwise 0
441 int utf32_iterator_grapheme_boundary(utf32_iterator it) {
442 uint32_t before, after;
443 enum unicode_Grapheme_Break gbbefore, gbafter;
445 if(it->n == 0 || it->n == it->ns)
447 /* Now we know that s[n-1] and s[n] are safe to inspect */
449 before = it->s[it->n-1];
450 after = it->s[it->n];
451 if(before == 0x000D && after == 0x000A)
453 gbbefore = utf32__grapheme_break(before);
454 gbafter = utf32__grapheme_break(after);
456 if(gbbefore == unicode_Grapheme_Break_Control
461 if(gbafter == unicode_Grapheme_Break_Control
466 if(gbbefore == unicode_Grapheme_Break_L
467 && (gbafter == unicode_Grapheme_Break_L
468 || gbafter == unicode_Grapheme_Break_V
469 || gbafter == unicode_Grapheme_Break_LV
470 || gbafter == unicode_Grapheme_Break_LVT))
473 if((gbbefore == unicode_Grapheme_Break_LV
474 || gbbefore == unicode_Grapheme_Break_V)
475 && (gbafter == unicode_Grapheme_Break_V
476 || gbafter == unicode_Grapheme_Break_T))
479 if((gbbefore == unicode_Grapheme_Break_LVT
480 || gbbefore == unicode_Grapheme_Break_T)
481 && gbafter == unicode_Grapheme_Break_T)
484 if(gbafter == unicode_Grapheme_Break_Extend)
491 /** @brief Test for a word boundary
493 * @return Non-0 if pointing just after a word boundary, otherwise 0
495 int utf32_iterator_word_boundary(utf32_iterator it) {
496 enum unicode_Word_Break twobefore, before, after, twoafter;
500 if(it->n == 0 || it->n == it->ns)
503 if(it->s[it->n-1] == 0x000D && it->s[it->n] == 0x000A)
506 /* (!Sep) x (Extend|Format) as in UAX #29 s6.2 */
507 if(utf32__sentence_break(it->s[it->n-1]) != unicode_Sentence_Break_Sep
508 && utf32__boundary_ignorable(utf32__word_break(it->s[it->n])))
510 /* Gather the property values we'll need for the rest of the test taking the
511 * s6.2 changes into account */
512 /* First we look at the code points after the proposed boundary */
513 nn = it->n; /* <it->ns */
514 after = utf32__word_break(it->s[nn++]);
515 if(!utf32__boundary_ignorable(after)) {
516 /* X (Extend|Format)* -> X */
518 && utf32__boundary_ignorable(utf32__word_break(it->s[nn])))
521 /* It's possible now that nn=ns */
523 twoafter = utf32__word_break(it->s[nn]);
525 twoafter = unicode_Word_Break_Other;
527 /* We've already recorded the non-ignorable code points before the proposed
529 before = utf32__word_break(it->last[1]);
530 twobefore = utf32__word_break(it->last[0]);
533 if(before == unicode_Word_Break_ALetter
534 && after == unicode_Word_Break_ALetter)
537 if(before == unicode_Word_Break_ALetter
538 && after == unicode_Word_Break_MidLetter
539 && twoafter == unicode_Word_Break_ALetter)
542 if(twobefore == unicode_Word_Break_ALetter
543 && before == unicode_Word_Break_MidLetter
544 && after == unicode_Word_Break_ALetter)
547 if(before == unicode_Word_Break_Numeric
548 && after == unicode_Word_Break_Numeric)
551 if(before == unicode_Word_Break_ALetter
552 && after == unicode_Word_Break_Numeric)
555 if(before == unicode_Word_Break_Numeric
556 && after == unicode_Word_Break_ALetter)
559 if(twobefore == unicode_Word_Break_Numeric
560 && before == unicode_Word_Break_MidNum
561 && after == unicode_Word_Break_Numeric)
564 if(before == unicode_Word_Break_Numeric
565 && after == unicode_Word_Break_MidNum
566 && twoafter == unicode_Word_Break_Numeric)
569 if(before == unicode_Word_Break_Katakana
570 && after == unicode_Word_Break_Katakana)
573 if((before == unicode_Word_Break_ALetter
574 || before == unicode_Word_Break_Numeric
575 || before == unicode_Word_Break_Katakana
576 || before == unicode_Word_Break_ExtendNumLet)
577 && after == unicode_Word_Break_ExtendNumLet)
580 if(before == unicode_Word_Break_ExtendNumLet
581 && (after == unicode_Word_Break_ALetter
582 || after == unicode_Word_Break_Numeric
583 || after == unicode_Word_Break_Katakana))
590 /** @defgroup utf32 Functions that operate on UTF-32 strings */
593 /** @brief Return the length of a 0-terminated UTF-32 string
594 * @param s Pointer to 0-terminated string
595 * @return Length of string in code points (excluding terminator)
597 * Unlike the conversion functions no validity checking is done on the string.
599 size_t utf32_len(const uint32_t *s) {
600 const uint32_t *t = s;
604 return (size_t)(t - s);
607 /** @brief Stably sort [s,s+ns) into descending order of combining class
608 * @param s Start of array
609 * @param ns Number of elements, must be at least 1
610 * @param buffer Buffer of at least @p ns elements
612 static void utf32__sort_ccc(uint32_t *s, size_t ns, uint32_t *buffer) {
613 uint32_t *a, *b, *bp;
617 case 1: /* 1-element array is always sorted */
619 case 2: /* 2-element arrays are trivial to sort */
620 if(utf32__combining_class(s[0]) > utf32__combining_class(s[1])) {
627 /* Partition the array */
632 /* Sort the two halves of the array */
633 utf32__sort_ccc(a, na, buffer);
634 utf32__sort_ccc(b, nb, buffer);
635 /* Merge them back into one, via the buffer */
637 while(na > 0 && nb > 0) {
638 /* We want descending order of combining class (hence <)
639 * and we want stability within combining classes (hence <=)
641 if(utf32__combining_class(*a) <= utf32__combining_class(*b)) {
657 memcpy(s, buffer, ns * sizeof(uint32_t));
662 /** @brief Put combining characters into canonical order
663 * @param s Pointer to UTF-32 string
664 * @param ns Length of @p s
665 * @return 0 on success, -1 on error
667 * @p s is modified in-place. See Unicode 5.0 s3.11 for details of the
670 * Currently we only support a maximum of 1024 combining characters after each
671 * base character. If this limit is exceeded then -1 is returned.
673 static int utf32__canonical_ordering(uint32_t *s, size_t ns) {
675 uint32_t buffer[1024];
677 /* The ordering amounts to a stable sort of each contiguous group of
678 * characters with non-0 combining class. */
680 /* Skip non-combining characters */
681 if(utf32__combining_class(*s) == 0) {
686 /* We must now have at least one combining character; see how many
688 for(nc = 1; nc < ns && utf32__combining_class(s[nc]) != 0; ++nc)
693 utf32__sort_ccc(s, nc, buffer);
700 /* Magic numbers from UAX #15 s16 */
708 #define NCount (VCount * TCount)
709 #define SCount (LCount * NCount)
711 /** @brief Guts of the decomposition lookup functions */
712 #define utf32__decompose_one_generic(WHICH) do { \
713 const uint32_t *dc = utf32__unidata(c)->WHICH; \
715 /* Found a canonical decomposition in the table */ \
717 utf32__decompose_one_##WHICH(d, *dc++); \
718 } else if(c >= SBase && c < SBase + SCount) { \
719 /* Mechanically decomposable Hangul syllable (UAX #15 s16) */ \
720 const uint32_t SIndex = c - SBase; \
721 const uint32_t L = LBase + SIndex / NCount; \
722 const uint32_t V = VBase + (SIndex % NCount) / TCount; \
723 const uint32_t T = TBase + SIndex % TCount; \
724 dynstr_ucs4_append(d, L); \
725 dynstr_ucs4_append(d, V); \
727 dynstr_ucs4_append(d, T); \
729 /* Equal to own canonical decomposition */ \
730 dynstr_ucs4_append(d, c); \
733 /** @brief Recursively compute the canonical decomposition of @p c
734 * @param d Dynamic string to store decomposition in
735 * @param c Code point to decompose (must be a valid!)
736 * @return 0 on success, -1 on error
738 static void utf32__decompose_one_canon(struct dynstr_ucs4 *d, uint32_t c) {
739 utf32__decompose_one_generic(canon);
742 /** @brief Recursively compute the compatibility decomposition of @p c
743 * @param d Dynamic string to store decomposition in
744 * @param c Code point to decompose (must be a valid!)
745 * @return 0 on success, -1 on error
747 static void utf32__decompose_one_compat(struct dynstr_ucs4 *d, uint32_t c) {
748 utf32__decompose_one_generic(compat);
751 /** @brief Guts of the decomposition functions */
752 #define utf32__decompose_generic(WHICH) do { \
753 struct dynstr_ucs4 d; \
756 dynstr_ucs4_init(&d); \
759 if((c >= 0xD800 && c <= 0xDFFF) || c > 0x10FFFF) \
761 utf32__decompose_one_##WHICH(&d, c); \
764 if(utf32__canonical_ordering(d.vec, d.nvec)) \
766 dynstr_ucs4_terminate(&d); \
775 /** @brief Canonically decompose @p [s,s+ns)
776 * @param s Pointer to string
777 * @param ns Length of string
778 * @param ndp Where to store length of result
779 * @return Pointer to result string, or NULL
781 * Computes the canonical decomposition of a string and stably sorts combining
782 * characters into canonical order. The result is in Normalization Form D and
783 * (at the time of writing!) passes the NFD tests defined in Unicode 5.0's
784 * NormalizationTest.txt.
786 * Returns NULL if the string is not valid for either of the following reasons:
787 * - it codes for a UTF-16 surrogate
788 * - it codes for a value outside the unicode code space
790 uint32_t *utf32_decompose_canon(const uint32_t *s, size_t ns, size_t *ndp) {
791 utf32__decompose_generic(canon);
794 /** @brief Compatibility decompose @p [s,s+ns)
795 * @param s Pointer to string
796 * @param ns Length of string
797 * @param ndp Where to store length of result
798 * @return Pointer to result string, or NULL
800 * Computes the compatibility decomposition of a string and stably sorts
801 * combining characters into canonical order. The result is in Normalization
802 * Form KD and (at the time of writing!) passes the NFKD tests defined in
803 * Unicode 5.0's NormalizationTest.txt.
805 * Returns NULL if the string is not valid for either of the following reasons:
806 * - it codes for a UTF-16 surrogate
807 * - it codes for a value outside the unicode code space
809 uint32_t *utf32_decompose_compat(const uint32_t *s, size_t ns, size_t *ndp) {
810 utf32__decompose_generic(compat);
813 /** @brief Single-character case-fold and decompose operation */
814 #define utf32__casefold_one(WHICH) do { \
815 const uint32_t *cf = utf32__unidata(c)->casefold; \
817 /* Found a case-fold mapping in the table */ \
819 utf32__decompose_one_##WHICH(&d, *cf++); \
821 utf32__decompose_one_##WHICH(&d, c); \
824 /** @brief Case-fold @p [s,s+ns)
825 * @param s Pointer to string
826 * @param ns Length of string
827 * @param ndp Where to store length of result
828 * @return Pointer to result string, or NULL
830 * Case-fold the string at @p s according to full default case-folding rules
831 * (s3.13) for caseless matching. The result will be in NFD.
833 * Returns NULL if the string is not valid for either of the following reasons:
834 * - it codes for a UTF-16 surrogate
835 * - it codes for a value outside the unicode code space
837 uint32_t *utf32_casefold_canon(const uint32_t *s, size_t ns, size_t *ndp) {
838 struct dynstr_ucs4 d;
843 /* If the canonical decomposition of the string includes any combining
844 * character that case-folds to a non-combining character then we must
845 * normalize before we fold. In Unicode 5.0.0 this means 0345 COMBINING
846 * GREEK YPOGEGRAMMENI in its decomposition and the various characters that
847 * canonically decompose to it. */
848 for(n = 0; n < ns; ++n)
849 if(utf32__unidata(s[n])->flags & unicode_normalize_before_casefold)
852 /* We need a preliminary decomposition */
853 if(!(ss = utf32_decompose_canon(s, ns, &ns)))
857 dynstr_ucs4_init(&d);
860 if((c >= 0xD800 && c <= 0xDFFF) || c > 0x10FFFF)
862 utf32__casefold_one(canon);
865 if(utf32__canonical_ordering(d.vec, d.nvec))
867 dynstr_ucs4_terminate(&d);
877 /** @brief Compatibilit case-fold @p [s,s+ns)
878 * @param s Pointer to string
879 * @param ns Length of string
880 * @param ndp Where to store length of result
881 * @return Pointer to result string, or NULL
883 * Case-fold the string at @p s according to full default case-folding rules
884 * (s3.13) for compatibility caseless matching. The result will be in NFKD.
886 * Returns NULL if the string is not valid for either of the following reasons:
887 * - it codes for a UTF-16 surrogate
888 * - it codes for a value outside the unicode code space
890 uint32_t *utf32_casefold_compat(const uint32_t *s, size_t ns, size_t *ndp) {
891 struct dynstr_ucs4 d;
896 for(n = 0; n < ns; ++n)
897 if(utf32__unidata(s[n])->flags & unicode_normalize_before_casefold)
900 /* We need a preliminary _canonical_ decomposition */
901 if(!(ss = utf32_decompose_canon(s, ns, &ns)))
905 /* This computes NFKD(toCaseFold(s)) */
906 #define compat_casefold_middle() do { \
907 dynstr_ucs4_init(&d); \
910 if((c >= 0xD800 && c <= 0xDFFF) || c > 0x10FFFF) \
912 utf32__casefold_one(compat); \
915 if(utf32__canonical_ordering(d.vec, d.nvec)) \
918 /* Do the inner (NFKD o toCaseFold) */
919 compat_casefold_middle();
920 /* We can do away with the NFD'd copy of the input now */
924 /* Do the outer (NFKD o toCaseFold) */
925 compat_casefold_middle();
927 dynstr_ucs4_terminate(&d);
937 /** @brief Order a pair of UTF-32 strings
938 * @param a First 0-terminated string
939 * @param b Second 0-terminated string
940 * @return -1, 0 or 1 for a less than, equal to or greater than b
942 * "Comparable to strcmp() at its best."
944 int utf32_cmp(const uint32_t *a, const uint32_t *b) {
945 while(*a && *b && *a == *b) {
949 return *a < *b ? -1 : (*a > *b ? 1 : 0);
952 /** @brief Identify a grapheme cluster boundary
953 * @param s Start of string (must be NFD)
954 * @param ns Length of string
955 * @param n Index within string (in [0,ns].)
956 * @return 1 at a grapheme cluster boundary, 0 otherwise
958 * This function identifies default grapheme cluster boundaries as described in
959 * UAX #29 s3. It returns 1 if @p n points at the code point just after a
960 * grapheme cluster boundary (including the hypothetical code point just after
961 * the end of the string).
963 int utf32_is_grapheme_boundary(const uint32_t *s, size_t ns, size_t n) {
964 struct utf32_iterator_data it[1];
966 utf32__iterator_init(it, s, ns, n);
967 return utf32_iterator_grapheme_boundary(it);
970 /** @brief Identify a word boundary
971 * @param s Start of string (must be NFD)
972 * @param ns Length of string
973 * @param n Index within string (in [0,ns].)
974 * @return 1 at a word boundary, 0 otherwise
976 * This function identifies default word boundaries as described in UAX #29 s4.
977 * It returns 1 if @p n points at the code point just after a word boundary
978 * (including the hypothetical code point just after the end of the string).
980 int utf32_is_word_boundary(const uint32_t *s, size_t ns, size_t n) {
981 struct utf32_iterator_data it[1];
983 utf32__iterator_init(it, s, ns, n);
984 return utf32_iterator_word_boundary(it);
988 /** @defgroup utf8 Functions that operate on UTF-8 strings */
991 /** @brief Wrapper to transform a UTF-8 string using the UTF-32 function */
992 #define utf8__transform(FN) do { \
993 uint32_t *to32 = 0, *decomp32 = 0; \
994 size_t nto32, ndecomp32; \
997 if(!(to32 = utf8_to_utf32(s, ns, &nto32))) goto error; \
998 if(!(decomp32 = FN(to32, nto32, &ndecomp32))) goto error; \
999 decomp8 = utf32_to_utf8(decomp32, ndecomp32, ndp); \
1006 /** @brief Canonically decompose @p [s,s+ns)
1007 * @param s Pointer to string
1008 * @param ns Length of string
1009 * @param ndp Where to store length of result
1010 * @return Pointer to result string, or NULL
1012 * Computes the canonical decomposition of a string and stably sorts combining
1013 * characters into canonical order. The result is in Normalization Form D and
1014 * (at the time of writing!) passes the NFD tests defined in Unicode 5.0's
1015 * NormalizationTest.txt.
1017 * Returns NULL if the string is not valid; see utf8_to_utf32() for reasons why
1020 * See also utf32_decompose_canon().
1022 char *utf8_decompose_canon(const char *s, size_t ns, size_t *ndp) {
1023 utf8__transform(utf32_decompose_canon);
1026 /** @brief Compatibility decompose @p [s,s+ns)
1027 * @param s Pointer to string
1028 * @param ns Length of string
1029 * @param ndp Where to store length of result
1030 * @return Pointer to result string, or NULL
1032 * Computes the compatibility decomposition of a string and stably sorts
1033 * combining characters into canonical order. The result is in Normalization
1034 * Form KD and (at the time of writing!) passes the NFKD tests defined in
1035 * Unicode 5.0's NormalizationTest.txt.
1037 * Returns NULL if the string is not valid; see utf8_to_utf32() for reasons why
1040 * See also utf32_decompose_compat().
1042 char *utf8_decompose_compat(const char *s, size_t ns, size_t *ndp) {
1043 utf8__transform(utf32_decompose_compat);
1046 /** @brief Case-fold @p [s,s+ns)
1047 * @param s Pointer to string
1048 * @param ns Length of string
1049 * @param ndp Where to store length of result
1050 * @return Pointer to result string, or NULL
1052 * Case-fold the string at @p s according to full default case-folding rules
1053 * (s3.13). The result will be in NFD.
1055 * Returns NULL if the string is not valid; see utf8_to_utf32() for reasons why
1058 char *utf8_casefold_canon(const char *s, size_t ns, size_t *ndp) {
1059 utf8__transform(utf32_casefold_canon);
1062 /** @brief Compatibility case-fold @p [s,s+ns)
1063 * @param s Pointer to string
1064 * @param ns Length of string
1065 * @param ndp Where to store length of result
1066 * @return Pointer to result string, or NULL
1068 * Case-fold the string at @p s according to full default case-folding rules
1069 * (s3.13). The result will be in NFKD.
1071 * Returns NULL if the string is not valid; see utf8_to_utf32() for reasons why
1074 char *utf8_casefold_compat(const char *s, size_t ns, size_t *ndp) {
1075 utf8__transform(utf32_casefold_compat);
1085 indent-tabs-mode:nil