Spatio-temporal Models and Languages: An Approach Based on Constraints

The introduction of spatio-temporal information in database systems presents us with an important data modelling challenge: the design of data models general and powerful enough to handle conventional thematic data, purely temporal or spatial concepts and spatio-temporal concepts. General purpose DBMS (e.g., relational) are not appropriate for storing and manipulating spatio-temporal data because of the complex structure of temporal and geometric information and the intricate temporal and spatial relationships among sets of related objects. Moreover, the costly operations involved in temporal and spatial object management seem at first glance to prevent the logical-level approach based on simple data structures (e.g., relations) manipulated through a limited set of simple operations (e.g., relational algebra). In the temporal world the main research trend has therefore been to introduce a variety of temporal data models with new temporal operations, and then worry about the semantics and appropriateness of these operations [45,44]. Similarly, in the spatial world the main approach has been to introduce extensions of conventional database models with abstract spatial data types encapsulating geometric structures and operations (see for example [39,35,40,19,20]). Chapter 4 of this book follows this school of thought for the modeling and querying of spatio-temporal information. Similarly, in the commercial world, DBMS such as Oracle [21] and Illustra [47] provide in their latest version separate modules devoted to temporal and spatial data. In this chapter we take an approach that differs significantly from the trends outlined above. We notice that the abstract data type approach lacks uniformity for the representation of data. To avoid this, we adopt the constraint data model, first introduced by Kanellakis, Kuper, and Revesz [25,33], and show that it is a very successful paradigm for the representation of spatio-temporal data in a unified framework. The constraint data model allows a uniform representation of all kinds of information present in spatio-temporal applications, and supports declarative query languages well-suited for complex spatio-temporal queries. Contrary to the temporal and spatial data models mentioned above, it also enables the straightforward modeling of indefinite information, a feature particularly useful in many spatio-temporal applications.