Cost-Based Object Query Optimization

The long-term goal of our research is to develop a top-down transformation-based and cost-based optimizer for object queries. Relational optimization techniques have been widely and successfully adopted. Thus, the primary task of object query optimization is to adapt relational techniques, and meanwhile to invent new techniques, to address the new features in object query languages such as method invocation, path expressions, user-defined data type, reference attributes, collection-valued attributes and multiple collection types. Previous work [GM93, H95, S98] dealt with path expressions, method invocation and user-defined data types. Multiple collection types and collection-valued attributes (CVAs), being important features, have not been investigated thoroughly regarding their impact on traditional optimization techniques. This thesis proposed an algebraic framework for object query optimization with attention paid to multiplecollection types and CVAs. In the following, we first observe the problems in building a costbased object query optimizer using the current body of knowledge. Then we present our solutions to these problems. Finally, we report the current status of this research.