1 /* gunicollate.c - Collation
3 * Copyright 2001,2005 Red Hat, Inc.
5 * The Gnome Library is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU Lesser General Public License as
7 * published by the Free Software Foundation; either version 2 of the
8 * License, or (at your option) any later version.
10 * The Gnome Library is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * Lesser General Public License for more details.
15 * You should have received a copy of the GNU Lesser General Public
16 * License along with the Gnome Library; see the file COPYING.LIB. If not,
17 * write to the Free Software Foundation, Inc., 59 Temple Place - Suite 330,
18 * Boston, MA 02111-1307, USA.
25 #ifdef __STDC_ISO_10646__
30 #include <CoreServices/CoreServices.h>
35 #include "gunicodeprivate.h"
37 #include "gstrfuncs.h"
38 #include "gtestutils.h"
40 #ifndef __STDC_ISO_10646__
46 /* Workaround for bug in MSVCR80.DLL */
48 msc_strxfrm_wrapper (char *string1,
52 if (!string1 || count <= 0)
56 return strxfrm (&tmp, string2, 1);
58 return strxfrm (string1, string2, count);
60 #define strxfrm msc_strxfrm_wrapper
65 * @str1: a UTF-8 encoded string
66 * @str2: a UTF-8 encoded string
68 * Compares two strings for ordering using the linguistically
69 * correct rules for the <link linkend="setlocale">current locale</link>.
70 * When sorting a large number of strings, it will be significantly
71 * faster to obtain collation keys with g_utf8_collate_key() and
72 * compare the keys with strcmp() when sorting instead of sorting
73 * the original strings.
75 * Return value: < 0 if @str1 compares before @str2,
76 * 0 if they compare equal, > 0 if @str1 compares after @str2.
79 g_utf8_collate (const gchar *str1,
92 g_return_val_if_fail (str1 != NULL, 0);
93 g_return_val_if_fail (str2 != NULL, 0);
95 str1_utf16 = g_utf8_to_utf16 (str1, -1, NULL, &len1, NULL);
96 str2_utf16 = g_utf8_to_utf16 (str2, -1, NULL, &len2, NULL);
98 UCCompareTextDefault (kUCCollateStandardOptions,
99 str1_utf16, len1, str2_utf16, len2,
106 #elif defined(__STDC_ISO_10646__)
111 g_return_val_if_fail (str1 != NULL, 0);
112 g_return_val_if_fail (str2 != NULL, 0);
114 str1_norm = _g_utf8_normalize_wc (str1, -1, G_NORMALIZE_ALL_COMPOSE);
115 str2_norm = _g_utf8_normalize_wc (str2, -1, G_NORMALIZE_ALL_COMPOSE);
117 result = wcscoll ((wchar_t *)str1_norm, (wchar_t *)str2_norm);
122 #else /* !__STDC_ISO_10646__ */
124 const gchar *charset;
128 g_return_val_if_fail (str1 != NULL, 0);
129 g_return_val_if_fail (str2 != NULL, 0);
131 str1_norm = g_utf8_normalize (str1, -1, G_NORMALIZE_ALL_COMPOSE);
132 str2_norm = g_utf8_normalize (str2, -1, G_NORMALIZE_ALL_COMPOSE);
134 if (g_get_charset (&charset))
136 result = strcoll (str1_norm, str2_norm);
140 gchar *str1_locale = g_convert (str1_norm, -1, charset, "UTF-8", NULL, NULL, NULL);
141 gchar *str2_locale = g_convert (str2_norm, -1, charset, "UTF-8", NULL, NULL, NULL);
143 if (str1_locale && str2_locale)
144 result = strcoll (str1_locale, str2_locale);
145 else if (str1_locale)
147 else if (str2_locale)
150 result = strcmp (str1_norm, str2_norm);
152 g_free (str1_locale);
153 g_free (str2_locale);
159 #endif /* __STDC_ISO_10646__ */
164 #if defined(__STDC_ISO_10646__) || defined(HAVE_CARBON)
165 /* We need UTF-8 encoding of numbers to encode the weights if
166 * we are using wcsxfrm. However, we aren't encoding Unicode
167 * characters, so we can't simply use g_unichar_to_utf8.
169 * The following routine is taken (with modification) from GNU
170 * libc's strxfrm routine:
172 * Copyright (C) 1995-1999,2000,2001 Free Software Foundation, Inc.
173 * Written by Ulrich Drepper <drepper@cygnus.com>, 1995.
176 utf8_encode (char *buf, wchar_t val)
190 for (step = 2; step < 6; ++step)
191 if ((val & (~(guint32)0 << (5 * step + 1))) == 0)
197 *buf = (unsigned char) (~0xff >> step);
201 buf[step] = 0x80 | (val & 0x3f);
211 #endif /* __STDC_ISO_10646__ || HAVE_CARBON */
216 collate_key_to_string (UCCollationValue *key,
223 /* Pretty smart algorithm here: ignore first eight bytes of the
224 * collation key. It doesn't produce results equivalent to
225 * UCCompareCollationKeys's, but the difference seems to be only
226 * that UCCompareCollationKeys in some cases produces 0 where our
227 * comparison gets -1 or 1. */
229 if (key_len * sizeof (UCCollationValue) <= 8)
230 return g_strdup ("");
233 for (i = 8; i < key_len * sizeof (UCCollationValue); i++)
234 /* there may be nul bytes, encode byteval+1 */
235 result_len += utf8_encode (NULL, *((guchar*)key + i) + 1);
237 result = g_malloc (result_len + 1);
239 for (i = 8; i < key_len * sizeof (UCCollationValue); i++)
240 result_len += utf8_encode (result + result_len, *((guchar*)key + i) + 1);
242 result[result_len] = 0;
247 carbon_collate_key_with_collator (const gchar *str,
249 CollatorRef collator)
251 UniChar *str_utf16 = NULL;
254 UCCollationValue staticbuf[512];
255 UCCollationValue *freeme = NULL;
256 UCCollationValue *buf;
260 gchar *result = NULL;
262 str_utf16 = g_utf8_to_utf16 (str, len, NULL, &len_utf16, NULL);
263 try_len = len_utf16 * 5 + 2;
265 if (try_len <= sizeof staticbuf)
268 buf_len = sizeof staticbuf;
272 freeme = g_new (UCCollationValue, try_len);
277 ret = UCGetCollationKey (collator, str_utf16, len_utf16,
278 buf_len, &key_len, buf);
280 if (ret == kCollateBufferTooSmall)
282 freeme = g_renew (UCCollationValue, freeme, try_len * 2);
284 buf_len = try_len * 2;
285 ret = UCGetCollationKey (collator, str_utf16, len_utf16,
286 buf_len, &key_len, buf);
290 result = collate_key_to_string (buf, key_len);
292 result = g_strdup ("");
300 carbon_collate_key (const gchar *str,
303 static CollatorRef collator;
305 if (G_UNLIKELY (!collator))
307 UCCreateCollator (NULL, 0, kUCCollateStandardOptions, &collator);
311 static gboolean been_here;
313 g_warning ("%s: UCCreateCollator failed", G_STRLOC);
315 return g_strdup ("");
319 return carbon_collate_key_with_collator (str, len, collator);
323 carbon_collate_key_for_filename (const gchar *str,
326 static CollatorRef collator;
328 if (G_UNLIKELY (!collator))
330 /* http://developer.apple.com/qa/qa2004/qa1159.html */
331 UCCreateCollator (NULL, 0,
332 kUCCollateComposeInsensitiveMask
333 | kUCCollateWidthInsensitiveMask
334 | kUCCollateCaseInsensitiveMask
335 | kUCCollateDigitsOverrideMask
336 | kUCCollateDigitsAsNumberMask
337 | kUCCollatePunctuationSignificantMask,
342 static gboolean been_here;
344 g_warning ("%s: UCCreateCollator failed", G_STRLOC);
346 return g_strdup ("");
350 return carbon_collate_key_with_collator (str, len, collator);
353 #endif /* HAVE_CARBON */
356 * g_utf8_collate_key:
357 * @str: a UTF-8 encoded string.
358 * @len: length of @str, in bytes, or -1 if @str is nul-terminated.
360 * Converts a string into a collation key that can be compared
361 * with other collation keys produced by the same function using
364 * The results of comparing the collation keys of two strings
365 * with strcmp() will always be the same as comparing the two
366 * original keys with g_utf8_collate().
368 * Note that this function depends on the
369 * <link linkend="setlocale">current locale</link>.
371 * Return value: a newly allocated string. This string should
372 * be freed with g_free() when you are done with it.
375 g_utf8_collate_key (const gchar *str,
382 g_return_val_if_fail (str != NULL, NULL);
383 result = carbon_collate_key (str, len);
385 #elif defined(__STDC_ISO_10646__)
391 gsize result_len = 0;
393 g_return_val_if_fail (str != NULL, NULL);
395 str_norm = _g_utf8_normalize_wc (str, len, G_NORMALIZE_ALL_COMPOSE);
397 xfrm_len = wcsxfrm (NULL, (wchar_t *)str_norm, 0);
398 result_wc = g_new (wchar_t, xfrm_len + 1);
399 wcsxfrm (result_wc, (wchar_t *)str_norm, xfrm_len + 1);
401 for (i=0; i < xfrm_len; i++)
402 result_len += utf8_encode (NULL, result_wc[i]);
404 result = g_malloc (result_len + 1);
406 for (i=0; i < xfrm_len; i++)
407 result_len += utf8_encode (result + result_len, result_wc[i]);
409 result[result_len] = '\0';
415 #else /* !__STDC_ISO_10646__ */
418 const gchar *charset;
421 g_return_val_if_fail (str != NULL, NULL);
423 str_norm = g_utf8_normalize (str, len, G_NORMALIZE_ALL_COMPOSE);
427 if (g_get_charset (&charset))
429 xfrm_len = strxfrm (NULL, str_norm, 0);
430 if (xfrm_len >= 0 && xfrm_len < G_MAXINT - 2)
432 result = g_malloc (xfrm_len + 1);
433 strxfrm (result, str_norm, xfrm_len + 1);
438 gchar *str_locale = g_convert (str_norm, -1, charset, "UTF-8", NULL, NULL, NULL);
442 xfrm_len = strxfrm (NULL, str_locale, 0);
443 if (xfrm_len < 0 || xfrm_len >= G_MAXINT - 2)
451 result = g_malloc (xfrm_len + 2);
453 strxfrm (result + 1, str_locale, xfrm_len + 1);
461 xfrm_len = strlen (str_norm);
462 result = g_malloc (xfrm_len + 2);
464 memcpy (result + 1, str_norm, xfrm_len);
465 result[xfrm_len+1] = '\0';
469 #endif /* __STDC_ISO_10646__ */
474 /* This is a collation key that is very very likely to sort before any
475 collation key that libc strxfrm generates. We use this before any
476 special case (dot or number) to make sure that its sorted before
479 #define COLLATION_SENTINEL "\1\1\1"
482 * g_utf8_collate_key_for_filename:
483 * @str: a UTF-8 encoded string.
484 * @len: length of @str, in bytes, or -1 if @str is nul-terminated.
486 * Converts a string into a collation key that can be compared
487 * with other collation keys produced by the same function using strcmp().
489 * In order to sort filenames correctly, this function treats the dot '.'
490 * as a special case. Most dictionary orderings seem to consider it
491 * insignificant, thus producing the ordering "event.c" "eventgenerator.c"
492 * "event.h" instead of "event.c" "event.h" "eventgenerator.c". Also, we
493 * would like to treat numbers intelligently so that "file1" "file10" "file5"
494 * is sorted as "file1" "file5" "file10".
496 * Note that this function depends on the
497 * <link linkend="setlocale">current locale</link>.
499 * Return value: a newly allocated string. This string should
500 * be freed with g_free() when you are done with it.
505 g_utf8_collate_key_for_filename (const gchar *str,
521 * Split the filename into collatable substrings which do
522 * not contain [.0-9] and special-cased substrings. The collatable
523 * substrings are run through the normal g_utf8_collate_key() and the
524 * resulting keys are concatenated with keys generated from the
525 * special-cased substrings.
527 * Special cases: Dots are handled by replacing them with '\1' which
528 * implies that short dot-delimited substrings are before long ones,
535 * Numbers are handled by prepending to each number d-1 superdigits
536 * where d = number of digits in the number and SUPERDIGIT is a
537 * character with an integer value higher than any digit (for instance
538 * ':'). This ensures that single-digit numbers are sorted before
539 * double-digit numbers which in turn are sorted separately from
540 * triple-digit numbers, etc. To avoid strange side-effects when
541 * sorting strings that already contain SUPERDIGITs, a '\2'
542 * is also prepended, like this
548 * file\2::100 (file100)
549 * file:foo (file:foo)
551 * This has the side-effect of sorting numbers before everything else (except
552 * dots), but this is probably OK.
554 * Leading digits are ignored when doing the above. To discriminate
555 * numbers which differ only in the number of leading digits, we append
556 * the number of leading digits as a byte at the very end of the collation
559 * To try avoid conflict with any collation key sequence generated by libc we
560 * start each switch to a special cased part with a sentinel that hopefully
561 * will sort before anything libc will generate.
567 result = g_string_sized_new (len * 2);
568 append = g_string_sized_new (0);
572 /* No need to use utf8 functions, since we're only looking for ascii chars */
573 for (prev = p = str; p < end; p++)
580 collate_key = g_utf8_collate_key (prev, p - prev);
581 g_string_append (result, collate_key);
582 g_free (collate_key);
585 g_string_append (result, COLLATION_SENTINEL "\1");
603 collate_key = g_utf8_collate_key (prev, p - prev);
604 g_string_append (result, collate_key);
605 g_free (collate_key);
608 g_string_append (result, COLLATION_SENTINEL "\2");
612 /* write d-1 colons */
626 if (*p == '0' && !digits)
628 else if (g_ascii_isdigit(*p))
632 /* count an all-zero sequence as
633 * one digit plus leading zeros
646 g_string_append_c (result, ':');
650 if (leading_zeros > 0)
652 g_string_append_c (append, (char)leading_zeros);
653 prev += leading_zeros;
656 /* write the number itself */
657 g_string_append_len (result, prev, p - prev);
660 --p; /* go one step back to avoid disturbing outer loop */
664 /* other characters just accumulate */
671 collate_key = g_utf8_collate_key (prev, p - prev);
672 g_string_append (result, collate_key);
673 g_free (collate_key);
676 g_string_append (result, append->str);
677 g_string_free (append, TRUE);
679 return g_string_free (result, FALSE);
680 #else /* HAVE_CARBON */
681 return carbon_collate_key_for_filename (str, len);