Spatial access methods

The main purpose of spatial access methods is to support efficient selection of objects based on spatial properties. For example, a range query selects objects lying within specified ranges of coordinates; a nearest neighbour query finds the object lying closest to a specified object (see Worboys, Chapter 26). Further, spatial access methods are also used to implement efficiently such spatial analyses as map overlay, and other types of spatial joins. Two characteristics of spatial datasets are that they are frequently large and that the data are quite often distributed in an irregular manner. A spatial access method needs to take into account both spatial indexing and clustering techniques. Without a spatial index, every object in the database has to be checked to see whether it meets the spatial selection criterion; a ‘full table scan’ in a relational database. As spatial datasets are usually very large, such checking is unacceptable in practice for interactive use and most other applications. Therefore, a spatial index is required, in order to find the required objects efficiently without looking at every object. In cases when the whole spatial dataset resides in main memory it is sufficient to know the addresses of the requested objects, as main memory storage allows random access and does not introduce significant delays. However, most spatial datasets are so large that they cannot reside in the main memory of the computer and must be stored in secondary memory, such as its hard disk. Clustering is needed to group those objects which are often requested together. Otherwise, many different disk pages will have to be fetched, resulting in slow response. In a spatial context, clustering implies that objects which are close together in reality are also stored close together in memory. Many strategies for clustering objects in spatial databases adopt some form of ‘space-filling curve’ by ordering objects according to their sequence along a path that traverses all parts of the space. In traditional database systems, sorting (or ordering) of the data forms the basis for efficient searching, as in the B-tree approach (Bayer and Creight 1973). Although there are obvious bases for sorting text strings, numbers, or dates (1-dimensional data), there are no such simple solutions for sorting higher-dimensional spatial data. Computer memory is 1-dimensional but spatial data is 2-dimensional or 3-dimensional (or even higher dimensioned), and must be organised somehow in memory. An intuitive solution is to use a regular grid just as on a paper map. Each grid cell has a unique name, e.g. ‘A3’, ‘C6’, or ‘D5’. The cells are stored in some order in memory and can each contain a (fixed) number of