Merge pull request #514 from techee/linear_tag_remove

Add linear tag remove path for cases where not many files are open

tagmanager/src/tm_tag.c

@@ -828,6 +828,31 @@ gboolean tm_tags_sort(GPtrArray *tags_array, TMTagAttrType *sort_attributes,
 void tm_tags_remove_file_tags(TMSourceFile *source_file, GPtrArray *tags_array)
 {
 	guint i;
+
+	/* Now we choose between an algorithm with complexity O(tags_array->len) and
+	 * O(source_file->tags_array->len * log(tags_array->len)). The latter algorithm
+	 * is better when tags_array contains many times more tags than
+	 * source_file->tags_array so instead of trying to find the removed tags
+	 * linearly, binary search is used. The constant 20 is more or less random
+	 * but seems to work well. It's exact value isn't so critical because it's
+	 * the extremes where the difference is the biggest: when
+	 * source_file->tags_array->len == tags_array->len (single file open) and
+	 * source_file->tags_array->len << tags_array->len (the number of tags
+	 * from the file is a small fraction of all tags).
+	 */
+	if (source_file->tags_array->len != 0 &&
+		tags_array->len / source_file->tags_array->len < 20)
+	{
+		for (i = 0; i < tags_array->len; i++)
+		{
+			TMTag *tag = tags_array->pdata[i];
+
+			if (tag->file == source_file)
+				tags_array->pdata[i] = NULL;
+		}
+	}
+	else
+	{
 		GPtrArray *to_delete = g_ptr_array_sized_new(source_file->tags_array->len);
 
 		for (i = 0; i < source_file->tags_array->len; i++)
@@ -859,6 +884,7 @@ void tm_tags_remove_file_tags(TMSourceFile *source_file, GPtrArray *tags_array)
 			*tag = NULL;
 		}
 		g_ptr_array_free(to_delete, TRUE);
+	}
 
 	tm_tags_prune(tags_array);
 }