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;
229 dynstr_ucs4_init(&d);
231 const struct unicode_utf8_row *const r = &unicode_utf8_valid[*ss];
238 if(ss[1] < r->min2 || ss[1] > r->max2)
243 if(ss[1] < r->min2 || ss[1] > r->max2)
248 if(ss[1] < r->min2 || ss[1] > r->max2)
257 for(n = 1; n < r->count; ++n) {
258 if(ss[n] < 0x80 || ss[n] > 0xBF)
260 c32 = (c32 << 6) | (ss[n] & 0x3F);
262 dynstr_ucs4_append(&d, c32);
266 dynstr_ucs4_terminate(&d);
275 /** @brief Test whether [s,s+ns) is valid UTF-8
276 * @param s Start of string
277 * @param ns Length of string
278 * @return non-0 if @p s is valid UTF-8, 0 if it is not valid
280 * This function is intended to be much faster than calling utf8_to_utf32() and
281 * throwing away the result.
283 int utf8_valid(const char *s, size_t ns) {
284 const uint8_t *ss = (const uint8_t *)s;
286 const struct unicode_utf8_row *const r = &unicode_utf8_valid[*ss];
292 if(ss[1] < r->min2 || ss[1] > r->max2)
296 if(ss[1] < r->min2 || ss[1] > r->max2)
298 if(ss[2] < 0x80 || ss[2] > 0xBF)
302 if(ss[1] < r->min2 || ss[1] > r->max2)
304 if(ss[2] < 0x80 || ss[2] > 0xBF)
306 if(ss[3] < 0x80 || ss[3] > 0xBF)
321 /** @defgroup utf32iterator UTF-32 string iterators */
324 struct utf32_iterator_data {
325 /** @brief Start of string */
328 /** @brief Length of string */
331 /** @brief Current position */
334 /** @brief Last two non-ignorable characters or (uint32_t)-1
336 * last[1] is the non-Extend/Format character just before position @p n;
337 * last[0] is the one just before that.
339 * Exception 1: if there is no such non-Extend/Format character then an
340 * Extend/Format character is accepted instead.
342 * Exception 2: if there is no such character even taking that into account
343 * the value is (uint32_t)-1.
348 /** @brief Create a new iterator pointing at the start of a string
349 * @param s Start of string
350 * @param ns Length of string
351 * @return New iterator
353 utf32_iterator utf32_iterator_new(const uint32_t *s, size_t ns) {
354 utf32_iterator it = xmalloc(sizeof *it);
358 it->last[0] = it->last[1] = -1;
362 /** @brief Initialize an internal private iterator
364 * @param s Start of string
365 * @param ns Length of string
366 * @param n Absolute position
368 static void utf32__iterator_init(utf32_iterator it,
369 const uint32_t *s, size_t ns, size_t n) {
373 it->last[0] = it->last[1] = -1;
374 utf32_iterator_set(it, n);
377 /** @brief Destroy an iterator
380 void utf32_iterator_destroy(utf32_iterator it) {
384 /** @brief Find the current position of an interator
387 size_t utf32_iterator_where(utf32_iterator it) {
391 /** @brief Set an iterator's absolute position
393 * @param n Absolute position
394 * @return 0 on success, non-0 on error
396 * It is an error to position the iterator outside the string (but acceptable
397 * to point it at the hypothetical post-final character). If an invalid value
398 * of @p n is specified then the iterator is not changed.
400 int utf32_iterator_set(utf32_iterator it, size_t n) {
401 /* We can't just jump to position @p n; the @p last[] values will be wrong.
402 * What we need is to jump a bit behind @p n and then advance forward,
403 * updating @p last[] along the way. How far back? We need to cross two
404 * non-ignorable code points as we advance forwards, so we'd better pass two
405 * such characters on the way back (if such are available).
409 if(n > it->ns) /* range check */
411 /* Walk backwards skipping ignorable code points */
413 while(m > 0 && (utf32__boundary_ignorable(utf32__word_break(it->s[m-1]))))
415 /* Either m=0 or s[m-1] is not ignorable */
418 /* s[m] is our first non-ignorable code; look for a second in the same
420 while(m > 0 && (utf32__boundary_ignorable(utf32__word_break(it->s[m-1]))))
422 /* Either m=0 or s[m-1] is not ignorable */
426 it->last[0] = it->last[1] = -1;
428 return utf32_iterator_advance(it, n - m);
431 /** @brief Advance an iterator
433 * @param count Number of code points to advance by
434 * @return 0 on success, non-0 on error
436 * It is an error to advance an iterator beyond the hypothetical post-final
437 * character of the string. If an invalid value of @p n is specified then the
438 * iterator is not changed.
440 * This function has O(n) time complexity: it works by advancing naively
441 * forwards through the string.
443 int utf32_iterator_advance(utf32_iterator it, size_t count) {
444 if(count <= it->ns - it->n) {
446 const uint32_t c = it->s[it->n];
447 const enum unicode_Word_Break wb = utf32__word_break(c);
448 if(it->last[1] == (uint32_t)-1
449 || !utf32__boundary_ignorable(wb)) {
450 it->last[0] = it->last[1];
461 /** @brief Find the current code point
463 * @return Current code point or 0
465 * If the iterator points at the hypothetical post-final character of the
466 * string then 0 is returned. NB that this doesn't mean that there aren't any
467 * 0 code points inside the string!
469 uint32_t utf32_iterator_code(utf32_iterator it) {
476 /** @brief Test for a grapheme boundary
478 * @return Non-0 if pointing just after a grapheme boundary, otherwise 0
480 int utf32_iterator_grapheme_boundary(utf32_iterator it) {
481 uint32_t before, after;
482 enum unicode_Grapheme_Break gbbefore, gbafter;
484 if(it->n == 0 || it->n == it->ns)
486 /* Now we know that s[n-1] and s[n] are safe to inspect */
488 before = it->s[it->n-1];
489 after = it->s[it->n];
490 if(before == 0x000D && after == 0x000A)
492 gbbefore = utf32__grapheme_break(before);
493 gbafter = utf32__grapheme_break(after);
495 if(gbbefore == unicode_Grapheme_Break_Control
500 if(gbafter == unicode_Grapheme_Break_Control
505 if(gbbefore == unicode_Grapheme_Break_L
506 && (gbafter == unicode_Grapheme_Break_L
507 || gbafter == unicode_Grapheme_Break_V
508 || gbafter == unicode_Grapheme_Break_LV
509 || gbafter == unicode_Grapheme_Break_LVT))
512 if((gbbefore == unicode_Grapheme_Break_LV
513 || gbbefore == unicode_Grapheme_Break_V)
514 && (gbafter == unicode_Grapheme_Break_V
515 || gbafter == unicode_Grapheme_Break_T))
518 if((gbbefore == unicode_Grapheme_Break_LVT
519 || gbbefore == unicode_Grapheme_Break_T)
520 && gbafter == unicode_Grapheme_Break_T)
523 if(gbafter == unicode_Grapheme_Break_Extend)
530 /** @brief Test for a word boundary
532 * @return Non-0 if pointing just after a word boundary, otherwise 0
534 int utf32_iterator_word_boundary(utf32_iterator it) {
535 enum unicode_Word_Break twobefore, before, after, twoafter;
539 if(it->n == 0 || it->n == it->ns)
542 if(it->s[it->n-1] == 0x000D && it->s[it->n] == 0x000A)
545 /* (!Sep) x (Extend|Format) as in UAX #29 s6.2 */
546 if(utf32__sentence_break(it->s[it->n-1]) != unicode_Sentence_Break_Sep
547 && utf32__boundary_ignorable(utf32__word_break(it->s[it->n])))
549 /* Gather the property values we'll need for the rest of the test taking the
550 * s6.2 changes into account */
551 /* First we look at the code points after the proposed boundary */
552 nn = it->n; /* <it->ns */
553 after = utf32__word_break(it->s[nn++]);
554 if(!utf32__boundary_ignorable(after)) {
555 /* X (Extend|Format)* -> X */
557 && utf32__boundary_ignorable(utf32__word_break(it->s[nn])))
560 /* It's possible now that nn=ns */
562 twoafter = utf32__word_break(it->s[nn]);
564 twoafter = unicode_Word_Break_Other;
566 /* We've already recorded the non-ignorable code points before the proposed
568 before = utf32__word_break(it->last[1]);
569 twobefore = utf32__word_break(it->last[0]);
572 if(before == unicode_Word_Break_ALetter
573 && after == unicode_Word_Break_ALetter)
576 if(before == unicode_Word_Break_ALetter
577 && after == unicode_Word_Break_MidLetter
578 && twoafter == unicode_Word_Break_ALetter)
581 if(twobefore == unicode_Word_Break_ALetter
582 && before == unicode_Word_Break_MidLetter
583 && after == unicode_Word_Break_ALetter)
586 if(before == unicode_Word_Break_Numeric
587 && after == unicode_Word_Break_Numeric)
590 if(before == unicode_Word_Break_ALetter
591 && after == unicode_Word_Break_Numeric)
594 if(before == unicode_Word_Break_Numeric
595 && after == unicode_Word_Break_ALetter)
598 if(twobefore == unicode_Word_Break_Numeric
599 && before == unicode_Word_Break_MidNum
600 && after == unicode_Word_Break_Numeric)
603 if(before == unicode_Word_Break_Numeric
604 && after == unicode_Word_Break_MidNum
605 && twoafter == unicode_Word_Break_Numeric)
608 if(before == unicode_Word_Break_Katakana
609 && after == unicode_Word_Break_Katakana)
612 if((before == unicode_Word_Break_ALetter
613 || before == unicode_Word_Break_Numeric
614 || before == unicode_Word_Break_Katakana
615 || before == unicode_Word_Break_ExtendNumLet)
616 && after == unicode_Word_Break_ExtendNumLet)
619 if(before == unicode_Word_Break_ExtendNumLet
620 && (after == unicode_Word_Break_ALetter
621 || after == unicode_Word_Break_Numeric
622 || after == unicode_Word_Break_Katakana))
629 /** @defgroup utf32 Functions that operate on UTF-32 strings */
632 /** @brief Return the length of a 0-terminated UTF-32 string
633 * @param s Pointer to 0-terminated string
634 * @return Length of string in code points (excluding terminator)
636 * Unlike the conversion functions no validity checking is done on the string.
638 size_t utf32_len(const uint32_t *s) {
639 const uint32_t *t = s;
643 return (size_t)(t - s);
646 /** @brief Stably sort [s,s+ns) into descending order of combining class
647 * @param s Start of array
648 * @param ns Number of elements, must be at least 1
649 * @param buffer Buffer of at least @p ns elements
651 static void utf32__sort_ccc(uint32_t *s, size_t ns, uint32_t *buffer) {
652 uint32_t *a, *b, *bp;
656 case 1: /* 1-element array is always sorted */
658 case 2: /* 2-element arrays are trivial to sort */
659 if(utf32__combining_class(s[0]) > utf32__combining_class(s[1])) {
666 /* Partition the array */
671 /* Sort the two halves of the array */
672 utf32__sort_ccc(a, na, buffer);
673 utf32__sort_ccc(b, nb, buffer);
674 /* Merge them back into one, via the buffer */
676 while(na > 0 && nb > 0) {
677 /* We want descending order of combining class (hence <)
678 * and we want stability within combining classes (hence <=)
680 if(utf32__combining_class(*a) <= utf32__combining_class(*b)) {
696 memcpy(s, buffer, ns * sizeof(uint32_t));
701 /** @brief Put combining characters into canonical order
702 * @param s Pointer to UTF-32 string
703 * @param ns Length of @p s
704 * @return 0 on success, -1 on error
706 * @p s is modified in-place. See Unicode 5.0 s3.11 for details of the
709 * Currently we only support a maximum of 1024 combining characters after each
710 * base character. If this limit is exceeded then -1 is returned.
712 static int utf32__canonical_ordering(uint32_t *s, size_t ns) {
714 uint32_t buffer[1024];
716 /* The ordering amounts to a stable sort of each contiguous group of
717 * characters with non-0 combining class. */
719 /* Skip non-combining characters */
720 if(utf32__combining_class(*s) == 0) {
725 /* We must now have at least one combining character; see how many
727 for(nc = 1; nc < ns && utf32__combining_class(s[nc]) != 0; ++nc)
732 utf32__sort_ccc(s, nc, buffer);
739 /* Magic numbers from UAX #15 s16 */
747 #define NCount (VCount * TCount)
748 #define SCount (LCount * NCount)
750 /** @brief Guts of the decomposition lookup functions */
751 #define utf32__decompose_one_generic(WHICH) do { \
752 const uint32_t *dc = utf32__unidata(c)->WHICH; \
754 /* Found a canonical decomposition in the table */ \
756 utf32__decompose_one_##WHICH(d, *dc++); \
757 } else if(c >= SBase && c < SBase + SCount) { \
758 /* Mechanically decomposable Hangul syllable (UAX #15 s16) */ \
759 const uint32_t SIndex = c - SBase; \
760 const uint32_t L = LBase + SIndex / NCount; \
761 const uint32_t V = VBase + (SIndex % NCount) / TCount; \
762 const uint32_t T = TBase + SIndex % TCount; \
763 dynstr_ucs4_append(d, L); \
764 dynstr_ucs4_append(d, V); \
766 dynstr_ucs4_append(d, T); \
768 /* Equal to own canonical decomposition */ \
769 dynstr_ucs4_append(d, c); \
772 /** @brief Recursively compute the canonical decomposition of @p c
773 * @param d Dynamic string to store decomposition in
774 * @param c Code point to decompose (must be a valid!)
775 * @return 0 on success, -1 on error
777 static void utf32__decompose_one_canon(struct dynstr_ucs4 *d, uint32_t c) {
778 utf32__decompose_one_generic(canon);
781 /** @brief Recursively compute the compatibility decomposition of @p c
782 * @param d Dynamic string to store decomposition in
783 * @param c Code point to decompose (must be a valid!)
784 * @return 0 on success, -1 on error
786 static void utf32__decompose_one_compat(struct dynstr_ucs4 *d, uint32_t c) {
787 utf32__decompose_one_generic(compat);
790 /** @brief Guts of the decomposition functions */
791 #define utf32__decompose_generic(WHICH) do { \
792 struct dynstr_ucs4 d; \
795 dynstr_ucs4_init(&d); \
798 if((c >= 0xD800 && c <= 0xDFFF) || c > 0x10FFFF) \
800 utf32__decompose_one_##WHICH(&d, c); \
803 if(utf32__canonical_ordering(d.vec, d.nvec)) \
805 dynstr_ucs4_terminate(&d); \
814 /** @brief Canonically decompose @p [s,s+ns)
815 * @param s Pointer to string
816 * @param ns Length of string
817 * @param ndp Where to store length of result
818 * @return Pointer to result string, or NULL
820 * Computes the canonical decomposition of a string and stably sorts combining
821 * characters into canonical order. The result is in Normalization Form D and
822 * (at the time of writing!) passes the NFD tests defined in Unicode 5.0's
823 * NormalizationTest.txt.
825 * Returns NULL if the string is not valid for either of the following reasons:
826 * - it codes for a UTF-16 surrogate
827 * - it codes for a value outside the unicode code space
829 uint32_t *utf32_decompose_canon(const uint32_t *s, size_t ns, size_t *ndp) {
830 utf32__decompose_generic(canon);
833 /** @brief Compatibility decompose @p [s,s+ns)
834 * @param s Pointer to string
835 * @param ns Length of string
836 * @param ndp Where to store length of result
837 * @return Pointer to result string, or NULL
839 * Computes the compatibility decomposition of a string and stably sorts
840 * combining characters into canonical order. The result is in Normalization
841 * Form KD and (at the time of writing!) passes the NFKD tests defined in
842 * Unicode 5.0's NormalizationTest.txt.
844 * Returns NULL if the string is not valid for either of the following reasons:
845 * - it codes for a UTF-16 surrogate
846 * - it codes for a value outside the unicode code space
848 uint32_t *utf32_decompose_compat(const uint32_t *s, size_t ns, size_t *ndp) {
849 utf32__decompose_generic(compat);
852 /** @brief Single-character case-fold and decompose operation */
853 #define utf32__casefold_one(WHICH) do { \
854 const uint32_t *cf = utf32__unidata(c)->casefold; \
856 /* Found a case-fold mapping in the table */ \
858 utf32__decompose_one_##WHICH(&d, *cf++); \
860 utf32__decompose_one_##WHICH(&d, c); \
863 /** @brief Case-fold @p [s,s+ns)
864 * @param s Pointer to string
865 * @param ns Length of string
866 * @param ndp Where to store length of result
867 * @return Pointer to result string, or NULL
869 * Case-fold the string at @p s according to full default case-folding rules
870 * (s3.13) for caseless matching. The result will be in NFD.
872 * Returns NULL if the string is not valid for either of the following reasons:
873 * - it codes for a UTF-16 surrogate
874 * - it codes for a value outside the unicode code space
876 uint32_t *utf32_casefold_canon(const uint32_t *s, size_t ns, size_t *ndp) {
877 struct dynstr_ucs4 d;
882 /* If the canonical decomposition of the string includes any combining
883 * character that case-folds to a non-combining character then we must
884 * normalize before we fold. In Unicode 5.0.0 this means 0345 COMBINING
885 * GREEK YPOGEGRAMMENI in its decomposition and the various characters that
886 * canonically decompose to it. */
887 for(n = 0; n < ns; ++n)
888 if(utf32__unidata(s[n])->flags & unicode_normalize_before_casefold)
891 /* We need a preliminary decomposition */
892 if(!(ss = utf32_decompose_canon(s, ns, &ns)))
896 dynstr_ucs4_init(&d);
899 if((c >= 0xD800 && c <= 0xDFFF) || c > 0x10FFFF)
901 utf32__casefold_one(canon);
904 if(utf32__canonical_ordering(d.vec, d.nvec))
906 dynstr_ucs4_terminate(&d);
916 /** @brief Compatibilit case-fold @p [s,s+ns)
917 * @param s Pointer to string
918 * @param ns Length of string
919 * @param ndp Where to store length of result
920 * @return Pointer to result string, or NULL
922 * Case-fold the string at @p s according to full default case-folding rules
923 * (s3.13) for compatibility caseless matching. The result will be in NFKD.
925 * Returns NULL if the string is not valid for either of the following reasons:
926 * - it codes for a UTF-16 surrogate
927 * - it codes for a value outside the unicode code space
929 uint32_t *utf32_casefold_compat(const uint32_t *s, size_t ns, size_t *ndp) {
930 struct dynstr_ucs4 d;
935 for(n = 0; n < ns; ++n)
936 if(utf32__unidata(s[n])->flags & unicode_normalize_before_casefold)
939 /* We need a preliminary _canonical_ decomposition */
940 if(!(ss = utf32_decompose_canon(s, ns, &ns)))
944 /* This computes NFKD(toCaseFold(s)) */
945 #define compat_casefold_middle() do { \
946 dynstr_ucs4_init(&d); \
949 if((c >= 0xD800 && c <= 0xDFFF) || c > 0x10FFFF) \
951 utf32__casefold_one(compat); \
954 if(utf32__canonical_ordering(d.vec, d.nvec)) \
957 /* Do the inner (NFKD o toCaseFold) */
958 compat_casefold_middle();
959 /* We can do away with the NFD'd copy of the input now */
963 /* Do the outer (NFKD o toCaseFold) */
964 compat_casefold_middle();
966 dynstr_ucs4_terminate(&d);
976 /** @brief Order a pair of UTF-32 strings
977 * @param a First 0-terminated string
978 * @param b Second 0-terminated string
979 * @return -1, 0 or 1 for a less than, equal to or greater than b
981 * "Comparable to strcmp() at its best."
983 int utf32_cmp(const uint32_t *a, const uint32_t *b) {
984 while(*a && *b && *a == *b) {
988 return *a < *b ? -1 : (*a > *b ? 1 : 0);
991 /** @brief Identify a grapheme cluster boundary
992 * @param s Start of string (must be NFD)
993 * @param ns Length of string
994 * @param n Index within string (in [0,ns].)
995 * @return 1 at a grapheme cluster boundary, 0 otherwise
997 * This function identifies default grapheme cluster boundaries as described in
998 * UAX #29 s3. It returns 1 if @p n points at the code point just after a
999 * grapheme cluster boundary (including the hypothetical code point just after
1000 * the end of the string).
1002 int utf32_is_grapheme_boundary(const uint32_t *s, size_t ns, size_t n) {
1003 struct utf32_iterator_data it[1];
1005 utf32__iterator_init(it, s, ns, n);
1006 return utf32_iterator_grapheme_boundary(it);
1009 /** @brief Identify a word boundary
1010 * @param s Start of string (must be NFD)
1011 * @param ns Length of string
1012 * @param n Index within string (in [0,ns].)
1013 * @return 1 at a word boundary, 0 otherwise
1015 * This function identifies default word boundaries as described in UAX #29 s4.
1016 * It returns 1 if @p n points at the code point just after a word boundary
1017 * (including the hypothetical code point just after the end of the string).
1019 int utf32_is_word_boundary(const uint32_t *s, size_t ns, size_t n) {
1020 struct utf32_iterator_data it[1];
1022 utf32__iterator_init(it, s, ns, n);
1023 return utf32_iterator_word_boundary(it);
1027 /** @defgroup utf8 Functions that operate on UTF-8 strings */
1030 /** @brief Wrapper to transform a UTF-8 string using the UTF-32 function */
1031 #define utf8__transform(FN) do { \
1032 uint32_t *to32 = 0, *decomp32 = 0; \
1033 size_t nto32, ndecomp32; \
1034 char *decomp8 = 0; \
1036 if(!(to32 = utf8_to_utf32(s, ns, &nto32))) goto error; \
1037 if(!(decomp32 = FN(to32, nto32, &ndecomp32))) goto error; \
1038 decomp8 = utf32_to_utf8(decomp32, ndecomp32, ndp); \
1045 /** @brief Canonically decompose @p [s,s+ns)
1046 * @param s Pointer to string
1047 * @param ns Length of string
1048 * @param ndp Where to store length of result
1049 * @return Pointer to result string, or NULL
1051 * Computes the canonical decomposition of a string and stably sorts combining
1052 * characters into canonical order. The result is in Normalization Form D and
1053 * (at the time of writing!) passes the NFD tests defined in Unicode 5.0's
1054 * NormalizationTest.txt.
1056 * Returns NULL if the string is not valid; see utf8_to_utf32() for reasons why
1059 * See also utf32_decompose_canon().
1061 char *utf8_decompose_canon(const char *s, size_t ns, size_t *ndp) {
1062 utf8__transform(utf32_decompose_canon);
1065 /** @brief Compatibility decompose @p [s,s+ns)
1066 * @param s Pointer to string
1067 * @param ns Length of string
1068 * @param ndp Where to store length of result
1069 * @return Pointer to result string, or NULL
1071 * Computes the compatibility decomposition of a string and stably sorts
1072 * combining characters into canonical order. The result is in Normalization
1073 * Form KD and (at the time of writing!) passes the NFKD tests defined in
1074 * Unicode 5.0's NormalizationTest.txt.
1076 * Returns NULL if the string is not valid; see utf8_to_utf32() for reasons why
1079 * See also utf32_decompose_compat().
1081 char *utf8_decompose_compat(const char *s, size_t ns, size_t *ndp) {
1082 utf8__transform(utf32_decompose_compat);
1085 /** @brief Case-fold @p [s,s+ns)
1086 * @param s Pointer to string
1087 * @param ns Length of string
1088 * @param ndp Where to store length of result
1089 * @return Pointer to result string, or NULL
1091 * Case-fold the string at @p s according to full default case-folding rules
1092 * (s3.13). The result will be in NFD.
1094 * Returns NULL if the string is not valid; see utf8_to_utf32() for reasons why
1097 char *utf8_casefold_canon(const char *s, size_t ns, size_t *ndp) {
1098 utf8__transform(utf32_casefold_canon);
1101 /** @brief Compatibility case-fold @p [s,s+ns)
1102 * @param s Pointer to string
1103 * @param ns Length of string
1104 * @param ndp Where to store length of result
1105 * @return Pointer to result string, or NULL
1107 * Case-fold the string at @p s according to full default case-folding rules
1108 * (s3.13). The result will be in NFKD.
1110 * Returns NULL if the string is not valid; see utf8_to_utf32() for reasons why
1113 char *utf8_casefold_compat(const char *s, size_t ns, size_t *ndp) {
1114 utf8__transform(utf32_casefold_compat);
1124 indent-tabs-mode:nil