[Xfce4-commits] <thunar:andrzejr/utf8_collate> Sorting files, switching to g_utf8_collate_key_for_filename()
Andrzej
noreply at xfce.org
Tue May 1 15:32:02 CEST 2012
Updating branch refs/heads/andrzejr/utf8_collate
to bc042483d3772de459f9eeb893d5f62580f2fa6c (commit)
from 61601dbced3e39f783dcfd501c67e7e6ac12dcae (commit)
commit bc042483d3772de459f9eeb893d5f62580f2fa6c
Author: Andrzej <ndrwrdck at gmail.com>
Date: Tue May 1 22:11:05 2012 +0900
Sorting files, switching to g_utf8_collate_key_for_filename()
This means number sort is gone but this should fix our issues with locales.
Sorting itself should be much faster but there might be some performance
impact of collation (but it's less frequent than comparison).
thunar/thunar-file.c | 164 ++++++++-----------------------------------------
thunar/thunar-file.h | 2 +
2 files changed, 29 insertions(+), 137 deletions(-)
diff --git a/thunar/thunar-file.c b/thunar/thunar-file.c
index 9509600..fd49c60 100644
--- a/thunar/thunar-file.c
+++ b/thunar/thunar-file.c
@@ -304,9 +304,11 @@ thunar_file_finalize (GObject *object)
/* free the custom icon name */
g_free (file->custom_icon_name);
- /* free display name and basename */
+ /* free display name, basename and collate keys*/
g_free (file->display_name);
g_free (file->basename);
+ g_free (file->collate_key);
+ g_free (file->collate_key_fc);
/* free the thumbnail path */
g_free (file->thumbnail_path);
@@ -630,6 +632,8 @@ thunar_file_get (GFile *gfile,
file->info = NULL;
file->custom_icon_name = NULL;
file->display_name = NULL;
+ file->collate_key = NULL;
+ file->collate_key_fc = NULL;
file->basename = NULL;
if (thunar_file_load (file, NULL, error))
@@ -731,13 +735,19 @@ thunar_file_load (ThunarFile *file,
g_free (file->custom_icon_name);
file->custom_icon_name = NULL;
- /* free display name and basename */
+ /* free display name, basename and collate keys */
g_free (file->display_name);
file->display_name = NULL;
g_free (file->basename);
file->basename = NULL;
+ g_free (file->collate_key);
+ file->collate_key = NULL;
+
+ g_free (file->collate_key_fc);
+ file->collate_key_fc = NULL;
+
/* free thumbnail path */
g_free (file->thumbnail_path);
file->thumbnail_path = NULL;
@@ -893,6 +903,11 @@ thunar_file_load (ThunarFile *file,
}
}
+ /* cache a collate keys for display name */
+ file->collate_key = g_utf8_collate_key_for_filename (file->display_name, -1);
+ file->collate_key_fc = g_utf8_collate_key_for_filename
+ (g_utf8_casefold (file->display_name, -1), -1);
+
/* set thumb state to unknown */
file->flags =
(file->flags & ~THUNAR_FILE_THUMB_STATE_MASK) | THUNAR_FILE_THUMB_STATE_UNKNOWN;
@@ -3292,12 +3307,7 @@ thunar_file_compare_by_name (const ThunarFile *file_a,
const ThunarFile *file_b,
gboolean case_sensitive)
{
- const gchar *ap;
- const gchar *bp;
- const gchar *filename_a;
- const gchar *filename_b;
- guchar ac;
- guchar bc;
+ gint result = 0;
#ifdef G_ENABLE_DEBUG
/* probably too expensive to do the instance check every time
@@ -3307,147 +3317,27 @@ thunar_file_compare_by_name (const ThunarFile *file_a,
_thunar_return_val_if_fail (THUNAR_IS_FILE (file_b), 0);
#endif
- /* we compare only the display names (UTF-8!) */
- filename_a = thunar_file_get_display_name (file_a);
- filename_b = thunar_file_get_display_name (file_b);
-
- /* start at the beginning of both strings */
- ap = filename_a;
- bp = filename_b;
-
/* check if we should ignore case */
- if (G_LIKELY (case_sensitive))
+ if (G_LIKELY (case_sensitive == FALSE))
{
- /* try simple (fast) ASCII comparison first */
- for (;; ++ap, ++bp)
- {
- /* check if the characters differ or we have a non-ASCII char */
- ac = *((const guchar *)ap);
- bc = *((const guchar *)bp);
- if (ac != bc || ac == 0 || ac > 127)
- break;
- }
-
- /* fallback to Unicode comparison */
- if (G_UNLIKELY (ac > 127 || bc > 127))
- {
- for (;; ap = g_utf8_next_char (ap), bp = g_utf8_next_char (bp))
- {
- /* check if characters differ or end of string */
- ac = g_utf8_get_char (ap);
- bc = g_utf8_get_char (bp);
- if (ac != bc || ac == 0)
- break;
- }
- }
- }
- else
- {
- /* try simple (fast) ASCII comparison first (case-insensitive!) */
- for (;; ++ap, ++bp)
- {
- /* check if the characters differ or we have a non-ASCII char */
- ac = *((const guchar *)ap);
- bc = *((const guchar *)bp);
- if (g_ascii_tolower (ac) != g_ascii_tolower (bc) || ac == 0 || ac > 127)
- break;
- }
-
- /* fallback to Unicode comparison (case-insensitive!) */
- if (G_UNLIKELY (ac > 127 || bc > 127))
- {
- for (;; ap = g_utf8_next_char (ap), bp = g_utf8_next_char (bp))
- {
- /* check if characters differ or end of string */
- ac = g_utf8_get_char (ap);
- bc = g_utf8_get_char (bp);
- if (g_unichar_tolower (ac) != g_unichar_tolower (bc) || ac == 0)
- break;
- }
- }
+ result = strcmp (file_a->collate_key_fc, file_b->collate_key_fc);
}
- /* if both strings are equal, we're done */
- if (G_UNLIKELY (ac == bc
- || (!case_sensitive
- && g_unichar_tolower (ac) == g_unichar_tolower (bc))))
+ /* if case sensitive or if ci comparison didn't find a difference */
+ if (G_UNLIKELY (case_sensitive == TRUE || result == 0))
{
- return 0;
+ result = strcmp (file_a->collate_key, file_b->collate_key);
}
- /* check if one of the characters that differ is a digit */
- if (G_UNLIKELY (g_ascii_isdigit (ac) || g_ascii_isdigit (bc)))
- {
- /* if both strings differ in a digit, we use a smarter comparison
- * to get sorting 'file1', 'file5', 'file10' done the right way.
- */
- if (g_ascii_isdigit (ac) && g_ascii_isdigit (bc))
- {
- return compare_by_name_using_number (ap, bp, filename_a, filename_b);
- }
-
- /* a second case is '20 file' and '2file', where comparison by number
- * makes sense if the previous char for both strings is a digit.
- */
- if (ap > filename_a
- && bp > filename_b
- && g_ascii_isdigit (*(ap - 1))
- && g_ascii_isdigit (*(bp - 1)))
- {
- /* go back one character to have both variables point to the numbers again */
- ap -= 1;
- bp -= 1;
-
- return compare_by_name_using_number (ap, bp, filename_a, filename_b);
- }
- }
-
- /* otherwise, if they differ in a unicode char, use the
- * appropriate collate function for the current locale (only
- * if charset is UTF-8, else the required transformations
- * would be too expensive)
- */
-#ifdef HAVE_STRCOLL
- if ((ac > 127 || bc > 127) && g_get_charset (NULL))
- {
- /* case-sensitive is easy, case-insensitive is expensive,
- * but we use a simple optimization to make it fast.
- */
- if (G_LIKELY (case_sensitive))
- {
- return strcoll (ap, bp);
- }
- else
- {
- /* we use a trick here, so we don't need to allocate
- * and transform the two strings completely first (8
- * byte for each buffer, so all compilers should align
- * them properly)
- */
- gchar abuf[8];
- gchar bbuf[8];
-
- /* transform the unicode chars to strings and
- * make sure the strings are nul-terminated.
- */
- abuf[g_unichar_to_utf8 (g_unichar_tolower(ac), abuf)] = '\0';
- bbuf[g_unichar_to_utf8 (g_unichar_tolower(bc), bbuf)] = '\0';
-
- /* compare the unicode chars (as strings) */
- return strcoll (abuf, bbuf);
- }
- }
+#ifdef G_ENABLE_DEBUG
+ /* if both strings are equal, we're screwed (two same filenames in one directory?) */
+ _thunar_return_val_if_fail (result != 0, 0);
#endif
- /* else, they differ in an ASCII character */
- if (G_UNLIKELY (!case_sensitive))
- return (g_unichar_tolower (ac) > g_unichar_tolower (bc)) ? 1 : -1;
- else
- return (ac > bc) ? 1 : -1;
+ return result;
}
-
gboolean
thunar_file_same_filesystem (const ThunarFile *file_a,
const ThunarFile *file_b)
diff --git a/thunar/thunar-file.h b/thunar/thunar-file.h
index d1a5704..4794bf6 100644
--- a/thunar/thunar-file.h
+++ b/thunar/thunar-file.h
@@ -117,6 +117,8 @@ struct _ThunarFile
GFile *gfile;
gchar *custom_icon_name;
gchar *display_name;
+ gchar *collate_key;
+ gchar *collate_key_fc;
gchar *basename;
gchar *thumbnail_path;
guint flags;
More information about the Xfce4-commits
mailing list