*
* You should have received a copy of the GNU Lesser General Public
* License along with the Gnome Library; see the file COPYING.LIB. If not,
- * write to the Free Software Foundation, Inc., 59 Temple Place - Suite 330,
- * Boston, MA 02111-1307, USA.
+ * see <http://www.gnu.org/licenses/>.
*/
-#include "glib.h"
+/**
+ * SECTION:unicode
+ * @Title: Unicode Manipulation
+ * @Short_description: functions operating on Unicode characters and
+ * UTF-8 strings
+ * @See_also: g_locale_to_utf8(), g_locale_from_utf8()
+ *
+ * This section describes a number of functions for dealing with
+ * Unicode characters and strings. There are analogues of the
+ * traditional `ctype.h` character classification and case conversion
+ * functions, UTF-8 analogues of some string utility functions,
+ * functions to perform normalization, case conversion and collation
+ * on UTF-8 strings and finally functions to convert between the UTF-8,
+ * UTF-16 and UCS-4 encodings of Unicode.
+ *
+ * The implementations of the Unicode functions in GLib are based
+ * on the Unicode Character Data tables, which are available from
+ * [www.unicode.org](http://www.unicode.org/).
+ * GLib 2.8 supports Unicode 4.0, GLib 2.10 supports Unicode 4.1,
+ * GLib 2.12 supports Unicode 5.0, GLib 2.16.3 supports Unicode 5.1,
+ * GLib 2.30 supports Unicode 6.0.
+ */
+
+#include "config.h"
+
+#include <stdlib.h>
+
+#include "gunicode.h"
#include "gunidecomp.h"
+#include "gmem.h"
#include "gunicomp.h"
+#include "gunicodeprivate.h"
-#include <config.h>
-#include <stdlib.h>
+#define CC_PART1(Page, Char) \
+ ((combining_class_table_part1[Page] >= G_UNICODE_MAX_TABLE_INDEX) \
+ ? (combining_class_table_part1[Page] - G_UNICODE_MAX_TABLE_INDEX) \
+ : (cclass_data[combining_class_table_part1[Page]][Char]))
-/* We cheat a bit and cast type values to (char *). We detect these
- using the &0xff trick. */
-#define CC(Page, Char) \
- ((((GPOINTER_TO_INT(combining_class_table[Page])) & 0xff) \
- == GPOINTER_TO_INT(combining_class_table[Page])) \
- ? GPOINTER_TO_INT(combining_class_table[Page]) \
- : (combining_class_table[Page][Char]))
+#define CC_PART2(Page, Char) \
+ ((combining_class_table_part2[Page] >= G_UNICODE_MAX_TABLE_INDEX) \
+ ? (combining_class_table_part2[Page] - G_UNICODE_MAX_TABLE_INDEX) \
+ : (cclass_data[combining_class_table_part2[Page]][Char]))
#define COMBINING_CLASS(Char) \
- (((Char) > (G_UNICODE_LAST_CHAR)) ? 0 : CC((Char) >> 8, (Char) & 0xff))
+ (((Char) <= G_UNICODE_LAST_CHAR_PART1) \
+ ? CC_PART1 ((Char) >> 8, (Char) & 0xff) \
+ : (((Char) >= 0xe0000 && (Char) <= G_UNICODE_LAST_CHAR) \
+ ? CC_PART2 (((Char) - 0xe0000) >> 8, (Char) & 0xff) \
+ : 0))
+
+/**
+ * g_unichar_combining_class:
+ * @uc: a Unicode character
+ *
+ * Determines the canonical combining class of a Unicode character.
+ *
+ * Returns: the combining class of the character
+ *
+ * Since: 2.14
+ **/
+gint
+g_unichar_combining_class (gunichar uc)
+{
+ return COMBINING_CLASS (uc);
+}
-/* Compute the canonical ordering of a string in-place. */
+/* constants for hangul syllable [de]composition */
+#define SBase 0xAC00
+#define LBase 0x1100
+#define VBase 0x1161
+#define TBase 0x11A7
+#define LCount 19
+#define VCount 21
+#define TCount 28
+#define NCount (VCount * TCount)
+#define SCount (LCount * NCount)
+
+/**
+ * g_unicode_canonical_ordering:
+ * @string: a UCS-4 encoded string.
+ * @len: the maximum length of @string to use.
+ *
+ * Computes the canonical ordering of a string in-place.
+ * This rearranges decomposed characters in the string
+ * according to their combining classes. See the Unicode
+ * manual for more information.
+ **/
void
g_unicode_canonical_ordering (gunichar *string,
gsize len)
{
gsize j;
/* Percolate item leftward through string. */
- for (j = i; j > 0; --j)
+ for (j = i + 1; j > 0; --j)
{
gunichar t;
- if (COMBINING_CLASS (string[j]) <= next)
+ if (COMBINING_CLASS (string[j - 1]) <= next)
break;
- t = string[j + 1];
- string[j + 1] = string[j];
- string[j] = t;
+ t = string[j];
+ string[j] = string[j - 1];
+ string[j - 1] = t;
swap = 1;
}
/* We're re-entering the loop looking at the old
}
}
-static guchar *
+/* http://www.unicode.org/unicode/reports/tr15/#Hangul
+ * r should be null or have sufficient space. Calling with r == NULL will
+ * only calculate the result_len; however, a buffer with space for three
+ * characters will always be big enough. */
+static void
+decompose_hangul (gunichar s,
+ gunichar *r,
+ gsize *result_len)
+{
+ gint SIndex = s - SBase;
+ gint TIndex = SIndex % TCount;
+
+ if (r)
+ {
+ r[0] = LBase + SIndex / NCount;
+ r[1] = VBase + (SIndex % NCount) / TCount;
+ }
+
+ if (TIndex)
+ {
+ if (r)
+ r[2] = TBase + TIndex;
+ *result_len = 3;
+ }
+ else
+ *result_len = 2;
+}
+
+/* returns a pointer to a null-terminated UTF-8 string */
+static const gchar *
find_decomposition (gunichar ch,
gboolean compat)
{
if (compat)
{
offset = decomp_table[half].compat_offset;
- if (offset == 0xff)
+ if (offset == G_UNICODE_NOT_PRESENT_OFFSET)
offset = decomp_table[half].canon_offset;
}
else
{
offset = decomp_table[half].canon_offset;
- if (offset == 0xff)
+ if (offset == G_UNICODE_NOT_PRESENT_OFFSET)
return NULL;
}
- return decomp_table[half].expansion + offset;
+ return &(decomp_expansion_string[offset]);
}
else if (half == start)
break;
return NULL;
}
+/**
+ * g_unicode_canonical_decomposition:
+ * @ch: a Unicode character.
+ * @result_len: location to store the length of the return value.
+ *
+ * Computes the canonical decomposition of a Unicode character.
+ *
+ * Returns: a newly allocated string of Unicode characters.
+ * @result_len is set to the resulting length of the string.
+ *
+ * Deprecated: 2.30: Use the more flexible g_unichar_fully_decompose()
+ * instead.
+ **/
gunichar *
g_unicode_canonical_decomposition (gunichar ch,
gsize *result_len)
{
- guchar *decomp = find_decomposition (ch, FALSE);
+ const gchar *decomp;
+ const gchar *p;
gunichar *r;
- if (decomp)
+ /* Hangul syllable */
+ if (ch >= SBase && ch < SBase + SCount)
+ {
+ decompose_hangul (ch, NULL, result_len);
+ r = g_malloc (*result_len * sizeof (gunichar));
+ decompose_hangul (ch, r, result_len);
+ }
+ else if ((decomp = find_decomposition (ch, FALSE)) != NULL)
{
/* Found it. */
- int i, len;
- /* We store as a double-nul terminated string. */
- for (len = 0; (decomp[len] || decomp[len + 1]);
- len += 2)
- ;
+ int i;
- /* We've counted twice as many bytes as there are
- characters. */
- *result_len = len / 2;
- r = malloc (len / 2 * sizeof (gunichar));
+ *result_len = g_utf8_strlen (decomp, -1);
+ r = g_malloc (*result_len * sizeof (gunichar));
- for (i = 0; i < len; i += 2)
- {
- r[i / 2] = (decomp[i] << 8 | decomp[i + 1]);
- }
+ for (p = decomp, i = 0; *p != '\0'; p = g_utf8_next_char (p), i++)
+ r[i] = g_utf8_get_char (p);
}
else
{
/* Not in our table. */
- r = malloc (sizeof (gunichar));
+ r = g_malloc (sizeof (gunichar));
*r = ch;
*result_len = 1;
}
- /* Supposedly following the Unicode 2.1.9 table means that the
- decompositions come out in canonical order. I haven't tested
- this, but we rely on it here. */
return r;
}
+/* L,V => LV and LV,T => LVT */
+static gboolean
+combine_hangul (gunichar a,
+ gunichar b,
+ gunichar *result)
+{
+ gint LIndex = a - LBase;
+ gint SIndex = a - SBase;
+
+ gint VIndex = b - VBase;
+ gint TIndex = b - TBase;
+
+ if (0 <= LIndex && LIndex < LCount
+ && 0 <= VIndex && VIndex < VCount)
+ {
+ *result = SBase + (LIndex * VCount + VIndex) * TCount;
+ return TRUE;
+ }
+ else if (0 <= SIndex && SIndex < SCount && (SIndex % TCount) == 0
+ && 0 < TIndex && TIndex < TCount)
+ {
+ *result = a + TIndex;
+ return TRUE;
+ }
+
+ return FALSE;
+}
+
#define CI(Page, Char) \
- ((((GPOINTER_TO_INT(compose_table[Page])) & 0xff) \
- == GPOINTER_TO_INT(compose_table[Page])) \
- ? GPOINTER_TO_INT(compose_table[Page]) \
- : (compose_table[Page][Char]))
+ ((compose_table[Page] >= G_UNICODE_MAX_TABLE_INDEX) \
+ ? (compose_table[Page] - G_UNICODE_MAX_TABLE_INDEX) \
+ : (compose_data[compose_table[Page]][Char]))
#define COMPOSE_INDEX(Char) \
- (((Char) > (G_UNICODE_LAST_CHAR)) ? 0 : CI((Char) >> 8, (Char) & 0xff))
+ (((Char >> 8) > (COMPOSE_TABLE_LAST)) ? 0 : CI((Char) >> 8, (Char) & 0xff))
-gboolean
+static gboolean
combine (gunichar a,
gunichar b,
gunichar *result)
{
gushort index_a, index_b;
+ if (combine_hangul (a, b, result))
+ return TRUE;
+
index_a = COMPOSE_INDEX(a);
+
if (index_a >= COMPOSE_FIRST_SINGLE_START && index_a < COMPOSE_SECOND_START)
{
if (b == compose_first_single[index_a - COMPOSE_FIRST_SINGLE_START][0])
return TRUE;
}
else
- return FALSE;
+ return FALSE;
}
index_b = COMPOSE_INDEX(b);
+
if (index_b >= COMPOSE_SECOND_SINGLE_START)
{
if (a == compose_second_single[index_b - COMPOSE_SECOND_SINGLE_START][0])
return TRUE;
}
else
- return FALSE;
+ return FALSE;
}
if (index_a >= COMPOSE_FIRST_START && index_a < COMPOSE_FIRST_SINGLE_START &&
- index_b >= COMPOSE_SECOND_START && index_a < COMPOSE_SECOND_SINGLE_START)
+ index_b >= COMPOSE_SECOND_START && index_b < COMPOSE_SECOND_SINGLE_START)
{
gunichar res = compose_array[index_a - COMPOSE_FIRST_START][index_b - COMPOSE_SECOND_START];
p = str;
while ((max_len < 0 || p < str + max_len) && *p)
{
+ const gchar *decomp;
gunichar wc = g_utf8_get_char (p);
- guchar *decomp = find_decomposition (wc, do_compat);
-
- if (decomp)
- {
- int len;
- /* We store as a double-nul terminated string. */
- for (len = 0; (decomp[len] || decomp[len + 1]);
- len += 2)
- ;
- n_wc += len / 2;
- }
- else
- n_wc++;
+ if (wc >= SBase && wc < SBase + SCount)
+ {
+ gsize result_len;
+ decompose_hangul (wc, NULL, &result_len);
+ n_wc += result_len;
+ }
+ else
+ {
+ decomp = find_decomposition (wc, do_compat);
+
+ if (decomp)
+ n_wc += g_utf8_strlen (decomp, -1);
+ else
+ n_wc++;
+ }
p = g_utf8_next_char (p);
}
while ((max_len < 0 || p < str + max_len) && *p)
{
gunichar wc = g_utf8_get_char (p);
- guchar *decomp;
+ const gchar *decomp;
int cc;
gsize old_n_wc = n_wc;
- decomp = find_decomposition (wc, do_compat);
-
- if (decomp)
- {
- int len;
- /* We store as a double-nul terminated string. */
- for (len = 0; (decomp[len] || decomp[len + 1]);
- len += 2)
- wc_buffer[n_wc++] = (decomp[len] << 8 | decomp[len + 1]);
- }
+ if (wc >= SBase && wc < SBase + SCount)
+ {
+ gsize result_len;
+ decompose_hangul (wc, wc_buffer + n_wc, &result_len);
+ n_wc += result_len;
+ }
else
- wc_buffer[n_wc++] = wc;
+ {
+ decomp = find_decomposition (wc, do_compat);
+
+ if (decomp)
+ {
+ const char *pd;
+ for (pd = decomp; *pd != '\0'; pd = g_utf8_next_char (pd))
+ wc_buffer[n_wc++] = g_utf8_get_char (pd);
+ }
+ else
+ wc_buffer[n_wc++] = wc;
+ }
if (n_wc > 0)
{
/* All decomposed and reordered */
-
if (do_compose && n_wc > 0)
{
gsize i, j;
int cc = COMBINING_CLASS (wc_buffer[i]);
if (i > 0 &&
- (last_cc == 0 || last_cc != cc) &&
+ (last_cc == 0 || last_cc < cc) &&
combine (wc_buffer[last_start], wc_buffer[i],
&wc_buffer[last_start]))
{
* @str: a UTF-8 encoded string.
* @len: length of @str, in bytes, or -1 if @str is nul-terminated.
* @mode: the type of normalization to perform.
- *
- * Convert a string into canonical form, standardizing
+ *
+ * Converts a string into canonical form, standardizing
* such issues as whether a character with an accent
* is represented as a base character and combining
- * accent or as a single precomposed characters. You
- * should generally call g_utf8_normalize before
- * comparing two Unicode strings.
+ * accent or as a single precomposed character. The
+ * string has to be valid UTF-8, otherwise %NULL is
+ * returned. You should generally call g_utf8_normalize()
+ * before comparing two Unicode strings.
*
* The normalization mode %G_NORMALIZE_DEFAULT only
* standardizes differences that do not affect the
* (in this case DIGIT THREE). Formatting information
* may be lost but for most text operations such
* characters should be considered the same.
- * For example, g_utf8_collate() normalizes
- * with %G_NORMALIZE_ALL as its first step.
*
* %G_NORMALIZE_DEFAULT_COMPOSE and %G_NORMALIZE_ALL_COMPOSE
* are like %G_NORMALIZE_DEFAULT and %G_NORMALIZE_ALL,
* useful if you intend to convert the string to
* a legacy encoding or pass it to a system with
* less capable Unicode handling.
- *
- * Return value: the string in normalized form
+ *
+ * Returns: a newly allocated string, that is the
+ * normalized form of @str, or %NULL if @str is not
+ * valid UTF-8.
**/
gchar *
g_utf8_normalize (const gchar *str,
{
gunichar *result_wc = _g_utf8_normalize_wc (str, len, mode);
gchar *result;
-
+
result = g_ucs4_to_utf8 (result_wc, -1, NULL, NULL, NULL);
g_free (result_wc);
return result;
}
+
+static gboolean
+decompose_hangul_step (gunichar ch,
+ gunichar *a,
+ gunichar *b)
+{
+ gint SIndex, TIndex;
+
+ if (ch < SBase || ch >= SBase + SCount)
+ return FALSE; /* not a hangul syllable */
+
+ SIndex = ch - SBase;
+ TIndex = SIndex % TCount;
+
+ if (TIndex)
+ {
+ /* split LVT -> LV,T */
+ *a = ch - TIndex;
+ *b = TBase + TIndex;
+ }
+ else
+ {
+ /* split LV -> L,V */
+ *a = LBase + SIndex / NCount;
+ *b = VBase + (SIndex % NCount) / TCount;
+ }
+
+ return TRUE;
+}
+
+/**
+ * g_unichar_decompose:
+ * @ch: a Unicode character
+ * @a: return location for the first component of @ch
+ * @b: return location for the second component of @ch
+ *
+ * Performs a single decomposition step of the
+ * Unicode canonical decomposition algorithm.
+ *
+ * This function does not include compatibility
+ * decompositions. It does, however, include algorithmic
+ * Hangul Jamo decomposition, as well as 'singleton'
+ * decompositions which replace a character by a single
+ * other character. In the case of singletons *@b will
+ * be set to zero.
+ *
+ * If @ch is not decomposable, *@a is set to @ch and *@b
+ * is set to zero.
+ *
+ * Note that the way Unicode decomposition pairs are
+ * defined, it is guaranteed that @b would not decompose
+ * further, but @a may itself decompose. To get the full
+ * canonical decomposition for @ch, one would need to
+ * recursively call this function on @a. Or use
+ * g_unichar_fully_decompose().
+ *
+ * See
+ * [UAX#15](http://unicode.org/reports/tr15/)
+ * for details.
+ *
+ * Returns: %TRUE if the character could be decomposed
+ *
+ * Since: 2.30
+ */
+gboolean
+g_unichar_decompose (gunichar ch,
+ gunichar *a,
+ gunichar *b)
+{
+ gint start = 0;
+ gint end = G_N_ELEMENTS (decomp_step_table);
+
+ if (decompose_hangul_step (ch, a, b))
+ return TRUE;
+
+ /* TODO use bsearch() */
+ if (ch >= decomp_step_table[start].ch &&
+ ch <= decomp_step_table[end - 1].ch)
+ {
+ while (TRUE)
+ {
+ gint half = (start + end) / 2;
+ const decomposition_step *p = &(decomp_step_table[half]);
+ if (ch == p->ch)
+ {
+ *a = p->a;
+ *b = p->b;
+ return TRUE;
+ }
+ else if (half == start)
+ break;
+ else if (ch > p->ch)
+ start = half;
+ else
+ end = half;
+ }
+ }
+
+ *a = ch;
+ *b = 0;
+
+ return FALSE;
+}
+
+/**
+ * g_unichar_compose:
+ * @a: a Unicode character
+ * @b: a Unicode character
+ * @ch: return location for the composed character
+ *
+ * Performs a single composition step of the
+ * Unicode canonical composition algorithm.
+ *
+ * This function includes algorithmic Hangul Jamo composition,
+ * but it is not exactly the inverse of g_unichar_decompose().
+ * No composition can have either of @a or @b equal to zero.
+ * To be precise, this function composes if and only if
+ * there exists a Primary Composite P which is canonically
+ * equivalent to the sequence <@a,@b>. See the Unicode
+ * Standard for the definition of Primary Composite.
+ *
+ * If @a and @b do not compose a new character, @ch is set to zero.
+ *
+ * See
+ * [UAX#15](http://unicode.org/reports/tr15/)
+ * for details.
+ *
+ * Returns: %TRUE if the characters could be composed
+ *
+ * Since: 2.30
+ */
+gboolean
+g_unichar_compose (gunichar a,
+ gunichar b,
+ gunichar *ch)
+{
+ if (combine (a, b, ch))
+ return TRUE;
+
+ *ch = 0;
+ return FALSE;
+}
+
+/**
+ * g_unichar_fully_decompose:
+ * @ch: a Unicode character.
+ * @compat: whether perform canonical or compatibility decomposition
+ * @result: (allow-none): location to store decomposed result, or %NULL
+ * @result_len: length of @result
+ *
+ * Computes the canonical or compatibility decomposition of a
+ * Unicode character. For compatibility decomposition,
+ * pass %TRUE for @compat; for canonical decomposition
+ * pass %FALSE for @compat.
+ *
+ * The decomposed sequence is placed in @result. Only up to
+ * @result_len characters are written into @result. The length
+ * of the full decomposition (irrespective of @result_len) is
+ * returned by the function. For canonical decomposition,
+ * currently all decompositions are of length at most 4, but
+ * this may change in the future (very unlikely though).
+ * At any rate, Unicode does guarantee that a buffer of length
+ * 18 is always enough for both compatibility and canonical
+ * decompositions, so that is the size recommended. This is provided
+ * as %G_UNICHAR_MAX_DECOMPOSITION_LENGTH.
+ *
+ * See
+ * [UAX#15](http://unicode.org/reports/tr15/)
+ * for details.
+ *
+ * Returns: the length of the full decomposition.
+ *
+ * Since: 2.30
+ **/
+gsize
+g_unichar_fully_decompose (gunichar ch,
+ gboolean compat,
+ gunichar *result,
+ gsize result_len)
+{
+ const gchar *decomp;
+ const gchar *p;
+
+ /* Hangul syllable */
+ if (ch >= SBase && ch < SBase + SCount)
+ {
+ gsize len, i;
+ gunichar buffer[3];
+ decompose_hangul (ch, result ? buffer : NULL, &len);
+ if (result)
+ for (i = 0; i < len && i < result_len; i++)
+ result[i] = buffer[i];
+ return len;
+ }
+ else if ((decomp = find_decomposition (ch, compat)) != NULL)
+ {
+ /* Found it. */
+ gsize len, i;
+
+ len = g_utf8_strlen (decomp, -1);
+
+ for (p = decomp, i = 0; i < len && i < result_len; p = g_utf8_next_char (p), i++)
+ result[i] = g_utf8_get_char (p);
+
+ return len;
+ }
+
+ /* Does not decompose */
+ if (result && result_len >= 1)
+ *result = ch;
+ return 1;
+}
projects / platform / upstream / glib.git / blobdiff