1 /* decomp.c - Character decomposition.
3 * Copyright (C) 1999, 2000 Tom Tromey
4 * Copyright 2000 Red Hat, Inc.
6 * The Gnome Library is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public License as
8 * published by the Free Software Foundation; either version 2 of the
9 * License, or (at your option) any later version.
11 * The Gnome Library is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with the Gnome Library; see the file COPYING.LIB. If not,
18 * write to the Free Software Foundation, Inc., 59 Temple Place - Suite 330,
19 * Boston, MA 02111-1307, USA.
27 #include "gunidecomp.h"
30 #include "gunicodeprivate.h"
33 #define CC_PART1(Page, Char) \
34 ((combining_class_table_part1[Page] >= G_UNICODE_MAX_TABLE_INDEX) \
35 ? (combining_class_table_part1[Page] - G_UNICODE_MAX_TABLE_INDEX) \
36 : (cclass_data[combining_class_table_part1[Page]][Char]))
38 #define CC_PART2(Page, Char) \
39 ((combining_class_table_part2[Page] >= G_UNICODE_MAX_TABLE_INDEX) \
40 ? (combining_class_table_part2[Page] - G_UNICODE_MAX_TABLE_INDEX) \
41 : (cclass_data[combining_class_table_part2[Page]][Char]))
43 #define COMBINING_CLASS(Char) \
44 (((Char) <= G_UNICODE_LAST_CHAR_PART1) \
45 ? CC_PART1 ((Char) >> 8, (Char) & 0xff) \
46 : (((Char) >= 0xe0000 && (Char) <= G_UNICODE_LAST_CHAR) \
47 ? CC_PART2 (((Char) - 0xe0000) >> 8, (Char) & 0xff) \
51 * g_unichar_combining_class:
52 * @uc: a Unicode character
54 * Determines the canonical combining class of a Unicode character.
56 * Return value: the combining class of the character
61 g_unichar_combining_class (gunichar uc)
63 return COMBINING_CLASS (uc);
66 /* constants for hangul syllable [de]composition */
74 #define NCount (VCount * TCount)
75 #define SCount (LCount * NCount)
78 * g_unicode_canonical_ordering:
79 * @string: a UCS-4 encoded string.
80 * @len: the maximum length of @string to use.
82 * Computes the canonical ordering of a string in-place.
83 * This rearranges decomposed characters in the string
84 * according to their combining classes. See the Unicode
85 * manual for more information.
88 g_unicode_canonical_ordering (gunichar *string,
98 last = COMBINING_CLASS (string[0]);
99 for (i = 0; i < len - 1; ++i)
101 int next = COMBINING_CLASS (string[i + 1]);
102 if (next != 0 && last > next)
105 /* Percolate item leftward through string. */
106 for (j = i + 1; j > 0; --j)
109 if (COMBINING_CLASS (string[j - 1]) <= next)
112 string[j] = string[j - 1];
116 /* We're re-entering the loop looking at the old
125 /* http://www.unicode.org/unicode/reports/tr15/#Hangul
126 * r should be null or have sufficient space. Calling with r == NULL will
127 * only calculate the result_len; however, a buffer with space for three
128 * characters will always be big enough. */
130 decompose_hangul (gunichar s,
134 gint SIndex = s - SBase;
136 /* not a hangul syllable */
137 if (SIndex < 0 || SIndex >= SCount)
145 gunichar L = LBase + SIndex / NCount;
146 gunichar V = VBase + (SIndex % NCount) / TCount;
147 gunichar T = TBase + SIndex % TCount;
166 /* returns a pointer to a null-terminated UTF-8 string */
168 find_decomposition (gunichar ch,
172 int end = G_N_ELEMENTS (decomp_table);
174 if (ch >= decomp_table[start].ch &&
175 ch <= decomp_table[end - 1].ch)
179 int half = (start + end) / 2;
180 if (ch == decomp_table[half].ch)
186 offset = decomp_table[half].compat_offset;
187 if (offset == G_UNICODE_NOT_PRESENT_OFFSET)
188 offset = decomp_table[half].canon_offset;
192 offset = decomp_table[half].canon_offset;
193 if (offset == G_UNICODE_NOT_PRESENT_OFFSET)
197 return &(decomp_expansion_string[offset]);
199 else if (half == start)
201 else if (ch > decomp_table[half].ch)
212 * g_unicode_canonical_decomposition:
213 * @ch: a Unicode character.
214 * @result_len: location to store the length of the return value.
216 * Computes the canonical decomposition of a Unicode character.
218 * Return value: a newly allocated string of Unicode characters.
219 * @result_len is set to the resulting length of the string.
221 * Deprecated: 2.30: Use the more flexible g_unichar_fully_decompose()
225 g_unicode_canonical_decomposition (gunichar ch,
232 /* Hangul syllable */
233 if (ch >= SBase && ch < SBase + SCount)
235 decompose_hangul (ch, NULL, result_len);
236 r = g_malloc (*result_len * sizeof (gunichar));
237 decompose_hangul (ch, r, result_len);
239 else if ((decomp = find_decomposition (ch, FALSE)) != NULL)
244 *result_len = g_utf8_strlen (decomp, -1);
245 r = g_malloc (*result_len * sizeof (gunichar));
247 for (p = decomp, i = 0; *p != '\0'; p = g_utf8_next_char (p), i++)
248 r[i] = g_utf8_get_char (p);
252 /* Not in our table. */
253 r = g_malloc (sizeof (gunichar));
261 /* L,V => LV and LV,T => LVT */
263 combine_hangul (gunichar a,
267 gint LIndex = a - LBase;
268 gint SIndex = a - SBase;
270 gint VIndex = b - VBase;
271 gint TIndex = b - TBase;
273 if (0 <= LIndex && LIndex < LCount
274 && 0 <= VIndex && VIndex < VCount)
276 *result = SBase + (LIndex * VCount + VIndex) * TCount;
279 else if (0 <= SIndex && SIndex < SCount && (SIndex % TCount) == 0
280 && 0 < TIndex && TIndex < TCount)
282 *result = a + TIndex;
289 #define CI(Page, Char) \
290 ((compose_table[Page] >= G_UNICODE_MAX_TABLE_INDEX) \
291 ? (compose_table[Page] - G_UNICODE_MAX_TABLE_INDEX) \
292 : (compose_data[compose_table[Page]][Char]))
294 #define COMPOSE_INDEX(Char) \
295 (((Char >> 8) > (COMPOSE_TABLE_LAST)) ? 0 : CI((Char) >> 8, (Char) & 0xff))
302 gushort index_a, index_b;
304 if (combine_hangul (a, b, result))
307 index_a = COMPOSE_INDEX(a);
309 if (index_a >= COMPOSE_FIRST_SINGLE_START && index_a < COMPOSE_SECOND_START)
311 if (b == compose_first_single[index_a - COMPOSE_FIRST_SINGLE_START][0])
313 *result = compose_first_single[index_a - COMPOSE_FIRST_SINGLE_START][1];
320 index_b = COMPOSE_INDEX(b);
322 if (index_b >= COMPOSE_SECOND_SINGLE_START)
324 if (a == compose_second_single[index_b - COMPOSE_SECOND_SINGLE_START][0])
326 *result = compose_second_single[index_b - COMPOSE_SECOND_SINGLE_START][1];
333 if (index_a >= COMPOSE_FIRST_START && index_a < COMPOSE_FIRST_SINGLE_START &&
334 index_b >= COMPOSE_SECOND_START && index_b < COMPOSE_SECOND_SINGLE_START)
336 gunichar res = compose_array[index_a - COMPOSE_FIRST_START][index_b - COMPOSE_SECOND_START];
349 _g_utf8_normalize_wc (const gchar *str,
357 gboolean do_compat = (mode == G_NORMALIZE_NFKC ||
358 mode == G_NORMALIZE_NFKD);
359 gboolean do_compose = (mode == G_NORMALIZE_NFC ||
360 mode == G_NORMALIZE_NFKC);
364 while ((max_len < 0 || p < str + max_len) && *p)
367 gunichar wc = g_utf8_get_char (p);
369 if (wc >= SBase && wc < SBase + SCount)
372 decompose_hangul (wc, NULL, &result_len);
377 decomp = find_decomposition (wc, do_compat);
380 n_wc += g_utf8_strlen (decomp, -1);
385 p = g_utf8_next_char (p);
388 wc_buffer = g_new (gunichar, n_wc + 1);
393 while ((max_len < 0 || p < str + max_len) && *p)
395 gunichar wc = g_utf8_get_char (p);
398 gsize old_n_wc = n_wc;
400 if (wc >= SBase && wc < SBase + SCount)
403 decompose_hangul (wc, wc_buffer + n_wc, &result_len);
408 decomp = find_decomposition (wc, do_compat);
413 for (pd = decomp; *pd != '\0'; pd = g_utf8_next_char (pd))
414 wc_buffer[n_wc++] = g_utf8_get_char (pd);
417 wc_buffer[n_wc++] = wc;
422 cc = COMBINING_CLASS (wc_buffer[old_n_wc]);
426 g_unicode_canonical_ordering (wc_buffer + last_start, n_wc - last_start);
427 last_start = old_n_wc;
431 p = g_utf8_next_char (p);
436 g_unicode_canonical_ordering (wc_buffer + last_start, n_wc - last_start);
442 /* All decomposed and reordered */
444 if (do_compose && n_wc > 0)
450 for (i = 0; i < n_wc; i++)
452 int cc = COMBINING_CLASS (wc_buffer[i]);
455 (last_cc == 0 || last_cc < cc) &&
456 combine (wc_buffer[last_start], wc_buffer[i],
457 &wc_buffer[last_start]))
459 for (j = i + 1; j < n_wc; j++)
460 wc_buffer[j-1] = wc_buffer[j];
467 last_cc = COMBINING_CLASS (wc_buffer[i-1]);
486 * @str: a UTF-8 encoded string.
487 * @len: length of @str, in bytes, or -1 if @str is nul-terminated.
488 * @mode: the type of normalization to perform.
490 * Converts a string into canonical form, standardizing
491 * such issues as whether a character with an accent
492 * is represented as a base character and combining
493 * accent or as a single precomposed character. The
494 * string has to be valid UTF-8, otherwise %NULL is
495 * returned. You should generally call g_utf8_normalize()
496 * before comparing two Unicode strings.
498 * The normalization mode %G_NORMALIZE_DEFAULT only
499 * standardizes differences that do not affect the
500 * text content, such as the above-mentioned accent
501 * representation. %G_NORMALIZE_ALL also standardizes
502 * the "compatibility" characters in Unicode, such
503 * as SUPERSCRIPT THREE to the standard forms
504 * (in this case DIGIT THREE). Formatting information
505 * may be lost but for most text operations such
506 * characters should be considered the same.
508 * %G_NORMALIZE_DEFAULT_COMPOSE and %G_NORMALIZE_ALL_COMPOSE
509 * are like %G_NORMALIZE_DEFAULT and %G_NORMALIZE_ALL,
510 * but returned a result with composed forms rather
511 * than a maximally decomposed form. This is often
512 * useful if you intend to convert the string to
513 * a legacy encoding or pass it to a system with
514 * less capable Unicode handling.
516 * Return value: a newly allocated string, that is the
517 * normalized form of @str, or %NULL if @str is not
521 g_utf8_normalize (const gchar *str,
525 gunichar *result_wc = _g_utf8_normalize_wc (str, len, mode);
528 result = g_ucs4_to_utf8 (result_wc, -1, NULL, NULL, NULL);
535 decompose_hangul_step (gunichar ch,
544 if (SIndex < 0 || SIndex >= SCount)
545 return FALSE; /* not a hangul syllable */
547 L = LBase + SIndex / NCount;
548 V = VBase + (SIndex % NCount) / TCount;
549 T = TBase + SIndex % TCount;
556 /* split LVT -> LV,T */
559 LV = SBase + (LIndex * VCount + VIndex) * TCount;
566 /* split LV -> L,V */
575 * g_unichar_decompose:
576 * @ch: a Unicode character
577 * @a: return location for the first component of @ch
578 * @b: return location for the second component of @ch
580 * Performs a single decomposition step of the
581 * Unicode canonical decomposition algorithm.
583 * This function does not include compatibility
584 * decompositions. It does, however, include algorithmic
585 * Hangul Jamo decomposition, as well as 'singleton'
586 * decompositions which replace a character by a single
587 * other character. In the case of singletons *@b will
590 * If @ch is not decomposable, *@a is set to @ch and *@b
593 * Note that the way Unicode decomposition pairs are
594 * defined, it is guaranteed that @b would not decompose
595 * further, but @a may itself decompose. To get the full
596 * canonical decomposition for @ch, one would need to
597 * recursively call this function on @a. Or use
598 * g_unichar_fully_decompose().
600 * See <ulink url="http://unicode.org/reports/tr15/">UAX#15</ulink>
603 * Returns: %TRUE if the character could be decomposed
608 g_unichar_decompose (gunichar ch,
613 gint end = G_N_ELEMENTS (decomp_step_table);
615 if (decompose_hangul_step (ch, a, b))
618 /* TODO use bsearch() */
619 if (ch >= decomp_step_table[start].ch &&
620 ch <= decomp_step_table[end - 1].ch)
624 gint half = (start + end) / 2;
625 const decomposition_step *p = &(decomp_step_table[half]);
632 else if (half == start)
649 * @a: a Unicode character
650 * @b: a Unicode character
651 * @ch: return location for the composed character
653 * Performs a single composition step of the
654 * Unicode canonical composition algorithm.
656 * This function does not perform algorithmic composition
657 * for Hangul characters, and does not include compatibility
658 * compositions. It does, however, include 'singleton'
659 * compositions which replace a character by a single
660 * other character. To obtain these, pass zero for @b.
662 * This function includes algorithmic Hangul Jamo composition.
664 * If @a and @b do not compose a new character, @ch is set to zero.
666 * See <ulink url="http://unicode.org/reports/tr15/">UAX#15</ulink>
669 * Returns: %TRUE if the characters could be composed
674 g_unichar_compose (gunichar a,
678 if (combine (a, b, ch))
686 * g_unichar_fully_decompose:
687 * @ch: a Unicode character.
688 * @compat: whether perform canonical or compatibility decomposition
689 * @result: location to store decomposed result, or %NULL
690 * @result_len: length of @result
692 * Computes the canonical or compatibility decomposition of a
693 * Unicode character. For compatibility decomposition,
694 * pass %TRUE for @compat; for canonical decomposition
695 * pass %FALSE for @compat.
697 * The decomposed sequence is placed in @result. Only up to
698 * @result_len characters are written into @result. The length
699 * of the full decomposition (irrespective of @result_len) is
700 * returned by the function. For canonical decomposition, a
701 * result buffer of length 4 is always enough, whereas for
702 * compatibility decomposition, a buffer of 18 is enough.
704 * See <ulink url="http://unicode.org/reports/tr15/">UAX#15</ulink>
707 * Return value: the length of the full decomposition.
712 g_unichar_fully_decompose (gunichar ch,
720 /* Hangul syllable */
721 if (ch >= SBase && ch < SBase + SCount)
725 decompose_hangul (ch, result ? buffer : NULL, &len);
727 for (i = 0; i < len && i < result_len; i++)
728 result[i] = buffer[i];
731 else if ((decomp = find_decomposition (ch, compat)) != NULL)
736 len = g_utf8_strlen (decomp, -1);
738 for (p = decomp, i = 0; i < len && i < result_len; p = g_utf8_next_char (p), i++)
739 result[i] = g_utf8_get_char (p);
744 /* Does not decompose */
745 if (result && result_len >= 1)