From 5308d12239b9896781e0f293791a410fc29f8f68 Mon Sep 17 00:00:00 2001
From: Aleksander Morgado <aleksander@lanedo.com>
Date: Wed, 12 Dec 2012 17:04:27 +0100
Subject: [PATCH] search: skip combining diacritical marks in search operations

https://bugzilla.gnome.org/show_bug.cgi?id=648587
---
 src/shell-app-system.c |  3 +-
 src/shell-app.c        |  8 ++---
 src/shell-util.c       | 78 ++++++++++++++++++++++++++++++++++++++++++
 src/shell-util.h       |  2 ++
 4 files changed, 86 insertions(+), 5 deletions(-)

diff --git a/src/shell-app-system.c b/src/shell-app-system.c
index 7f34c28b6..89a074cda 100644
--- a/src/shell-app-system.c
+++ b/src/shell-app-system.c
@@ -738,7 +738,8 @@ normalize_terms (GSList *terms)
   for (iter = terms; iter; iter = iter->next)
     {
       const char *term = iter->data;
-      normalized_terms = g_slist_prepend (normalized_terms, shell_util_normalize_and_casefold (term));
+      normalized_terms = g_slist_prepend (normalized_terms,
+                                          shell_util_normalize_casefold_and_unaccent (term));
     }
   return normalized_terms;
 }
diff --git a/src/shell-app.c b/src/shell-app.c
index 8ff53ec7d..1f8d6f01f 100644
--- a/src/shell-app.c
+++ b/src/shell-app.c
@@ -1319,16 +1319,16 @@ shell_app_init_search_data (ShellApp *app)
 
   appinfo = gmenu_tree_entry_get_app_info (app->entry);
   name = g_app_info_get_name (G_APP_INFO (appinfo));
-  app->casefolded_name = shell_util_normalize_and_casefold (name);
+  app->casefolded_name = shell_util_normalize_casefold_and_unaccent (name);
 
   generic_name = g_desktop_app_info_get_generic_name (appinfo);
   if (generic_name)
-    app->casefolded_generic_name = shell_util_normalize_and_casefold (generic_name);
+    app->casefolded_generic_name = shell_util_normalize_casefold_and_unaccent (generic_name);
   else
     app->casefolded_generic_name = NULL;
 
   exec = g_app_info_get_executable (G_APP_INFO (appinfo));
-  normalized_exec = shell_util_normalize_and_casefold (exec);
+  normalized_exec = shell_util_normalize_casefold_and_unaccent (exec);
   app->casefolded_exec = trim_exec_line (normalized_exec);
   g_free (normalized_exec);
 
@@ -1343,7 +1343,7 @@ shell_app_init_search_data (ShellApp *app)
       i = 0;
       while (keywords[i])
         {
-          app->casefolded_keywords[i] = shell_util_normalize_and_casefold (keywords[i]);
+          app->casefolded_keywords[i] = shell_util_normalize_casefold_and_unaccent (keywords[i]);
           ++i;
         }
       app->casefolded_keywords[i] = NULL;
diff --git a/src/shell-util.c b/src/shell-util.c
index 56ebd0219..3821b3aff 100644
--- a/src/shell-util.c
+++ b/src/shell-util.c
@@ -122,12 +122,90 @@ shell_util_normalize_and_casefold (const char *str)
   if (str == NULL)
     return NULL;
 
+  /* NOTE: 'ALL' is equivalent to 'NFKD'. If this is ever updated, please
+   * update the unaccenting mechanism as well. */
   normalized = g_utf8_normalize (str, -1, G_NORMALIZE_ALL);
   result = g_utf8_casefold (normalized, -1);
   g_free (normalized);
   return result;
 }
 
+/* Combining diacritical mark?
+ *  Basic range: [0x0300,0x036F]
+ *  Supplement:  [0x1DC0,0x1DFF]
+ *  For Symbols: [0x20D0,0x20FF]
+ *  Half marks:  [0xFE20,0xFE2F]
+ */
+#define IS_CDM_UCS4(c) (((c) >= 0x0300 && (c) <= 0x036F)  || \
+                        ((c) >= 0x1DC0 && (c) <= 0x1DFF)  || \
+                        ((c) >= 0x20D0 && (c) <= 0x20FF)  || \
+                        ((c) >= 0xFE20 && (c) <= 0xFE2F))
+
+/* Copied from tracker/src/libtracker-fts/tracker-parser-glib.c under the GPL
+ * Originally written by Aleksander Morgado <aleksander@gnu.org>
+ */
+char *
+shell_util_normalize_casefold_and_unaccent (const char *str)
+{
+  char *tmp;
+  gsize i = 0, j = 0, ilen;
+
+  if (str == NULL)
+    return NULL;
+
+  /* Get the NFKD-normalized and casefolded string */
+  tmp = shell_util_normalize_and_casefold (str);
+  ilen = strlen (tmp);
+
+  while (i < ilen)
+    {
+      gunichar unichar;
+      gchar *next_utf8;
+      gint utf8_len;
+
+      /* Get next character of the word as UCS4 */
+      unichar = g_utf8_get_char_validated (&tmp[i], -1);
+
+      /* Invalid UTF-8 character or end of original string. */
+      if (unichar == (gunichar) -1 ||
+          unichar == (gunichar) -2)
+        {
+          break;
+        }
+
+      /* Find next UTF-8 character */
+      next_utf8 = g_utf8_next_char (&tmp[i]);
+      utf8_len = next_utf8 - &tmp[i];
+
+      if (IS_CDM_UCS4 ((guint32) unichar))
+        {
+          /* If the given unichar is a combining diacritical mark,
+           * just update the original index, not the output one */
+          i += utf8_len;
+          continue;
+        }
+
+      /* If already found a previous combining
+       * diacritical mark, indexes are different so
+       * need to copy characters. As output and input
+       * buffers may overlap, need to use memmove
+       * instead of memcpy */
+      if (i != j)
+        {
+          memmove (&tmp[j], &tmp[i], utf8_len);
+        }
+
+      /* Update both indexes */
+      i += utf8_len;
+      j += utf8_len;
+    }
+
+  /* Force proper string end */
+  tmp[j] = '\0';
+
+  return tmp;
+}
+
 /**
  * shell_util_format_date:
  * @format: a strftime-style string format, as parsed by
diff --git a/src/shell-util.h b/src/shell-util.h
index 9dbf7239b..41ba96f4a 100644
--- a/src/shell-util.h
+++ b/src/shell-util.h
@@ -20,6 +20,8 @@ int      shell_util_get_week_start             (void);
 
 char    *shell_util_normalize_and_casefold     (const char       *str);
 
+char    *shell_util_normalize_casefold_and_unaccent (const char  *str);
+
 char    *shell_util_format_date                (const char       *format,
                                                 gint64            time_ms);