物件導向查詢模式

Abstract

近年來，物件導向技術正快速地成長，物件導向技術成長得愈快，物件導 向資料庫便愈形重要。與關連式資料庫比較起來，物件導向資料庫兼具優 點與缺點。今日物件導向資料庫缺點之一在於它的查詢能力，一個物件導 向資料庫不僅僅應支援一個物件導向資料模式，它必須能夠提供給使用者 一個強大的查詢功能。雖然在物件導向資料模式上已有合理程度的共識， 然而，很少物件導向資料庫系統能提供完整的查詢功能。原因在於物件導 向資料庫的查詢存在著許多特點，使得其明顯地不同於關聯式資料庫的查 詢，而這些特點對於物件導向資料庫之查詢的評估以及其複雜度，有著十 分明顯的影響。就成形於物件導向資料模式之物件導向觀念，本篇論文提 出一個物件導向查詢模式。首先，分析物件導向資料庫與關連式資料庫查 詢之間的基本差異。接著，根據這些差異提出一個物件導向查詢模式，該 模式根據B+_Tree的法則，提出一新的索引技術B+_Tree_Like索引透過B+ _Tree_Like索引技術的支援，該模式可提供類別階層、類別繼承、物件識 別子指定、VIEW功能，以及其他物件導向等特性。最後， B+_Tree_Like 索引技術的輸出將被集結為一組元件，這些元件可用來加強查詢的搜尋。 經由加入這些元件，將可修改已廣為運用的Genetic Algorithm ，並証實 可改善其搜尋效率。最重要的是，本物件導向查詢模式所介紹的這些元件 ，可以輕易地運用到其他搜尋邏輯上。 In recent years, the object_oriented technology is growing with a rapid speed. The faster the object_oriented technology grows, the more important the object_oriented database (OODB) becomes. Compared with the relational database system, OODB has both advantages and disadvantages . One of the shortcomings in today's OODB is its query capability. An OODB should not only support an object_oriented data model but also provide the user with a powerful query func tion. There is a reasonable degree of agreement about the object_oriented data model. However, few operational OODBs support the desired query function. A query under an object_oriented data mod el exhibits characteristics which clearly distinguish it from its relational counterpart. These charac teristics have significant impacts on the complexity of query evaluations in OODB. In this thesis, we present an object_oriented query model which is consistent with object_oriented concepts embo died in the object_oriented data model. First, we analyze the fundamental differences between the object_oriented and relational databases in term of the semantics of their queries. Then, based on these differences, we suggest an object_oriented query model. This model can support class hierar chy, class inheritance, object_identifier assignment, view operation, and other object_oriented char acteristics via the B+_Tree_like index technology we propose. The output of this query model are divided into a set of primitives, which can later be used to facilitate the query search. The well_known genetic algorithm is modified by including these primitives. Most important of all, these primitives we introduce can easily be incorporated into other search algorithms.