Using an Object-Oriented Database System to Construct A Spatial Database Kernel for Gis Applications

A CHOI,WS LUK

1992-01-01

Abstract:In this Paper we show how a generic object-oriented database system can be extended into some domain-specific object-oriented database systems by building additional software layers on top of it. With each lower layer supporting the constructs and views provided by the upper layers, a high-level interface to the system is provided to the user, without having him/her to know about the details of the underlying database system. Specifically, the design of a spatial database kernel for GIS (Geographic Information System) applications which sits on top of a generic object-oriented database system is presented here. Bi-level data modelling approach is used as a methodology for its design and implementation. This data model consists at the top level the geographic object data model, and the geometric object data model at the bottom level The former consists of geographic objects, semantic spatial functions and non-spatial functions, whereas the geometric object data model consists of geometric objects and geometric functions. Geometric objects (e.g. polygons, lines, points) are the spatial representations of the geographic objects (e.g. countries, roads, rivers). Geometric functions directly support the implementation of the semantic spatial functions of the geographic object data model Furthermore, super-functions can be built using the existing functions in the kernel to build more specialized GIS domain-specific functions. Users can invoke these superfunctions, just like calling any other functions in the kernel, in writing GIS application programs that run on top of the kernel With this extensible set of application-oriented and domain-specific functions, a high-level query language called OFQL or the Object-oriented Functional Query Language, is designed to allow user to pose spatial and non-spatial queries of a geographic nature. High-level query Processing and optimization issues, which are among the major concerns in the design of the spatial database kernel, are also addressed. In particular, we show how IS-A and PART-OF abstraction hierarchies are handled. Using the concept of cascaded function execution, both cyclic and acyclic queries are processed in the same way. Chain processing is also introduced wherein a query processing graph (QP graph) can be decomposed into chains where depth-first search is applicable. Our QP graph is capable of handling more than one target object class. Tuple objects are also introduced to reflect new relationships among two or more geographic objects.