Performing database operations over mismatched domains
Linda G. DeMichiel · 1990
A database application must often combine data from a variety of sources so that decisions can be based on a broad range of information. When data must be obtained from diverse database systems, it may be represented in forms that are not directly comparable. Within an environment of autonomous databases, it is important to be able to perform operations over information obtained from mismatched data domains without requiring changes to foreign source databases. We present a solution to the problem of correctly performing relational database query operations despite data domain mismatch. We use domain mappings and virtual attributes to resolve domain differences and to map conflicting attribute values to common domains. When domains are semantically mismatched, however, it is not always possible to map a real attribute value to a single definite value of the target domain, and in such cases, we must generate partial values. In certain cases it is not possible for the extended operations to preserve all the information given in their operand relations. We identify these cases, analyze the meaning of this information loss, and present query optimization strategies and additional mechanisms to reduce or eliminate such information loss. We have implemented a system, DomainMatch, that demonstrates this extended relational algebra and uses it together with the mechanisms of domain mappings and virtual attributes to perform relational query operations over information obtained from mismatched domains. We describe this system and illustrate its use. This work enables a database system to present information derived from multiple sources in an integrated form. Our approach is particularly important in a distributed environment where individual databases are autonomous and all updating control resides with local databases, since it can be used with no modification of local schemas. The end user is presented with considerably more relevant information than would be obtained using conventional relational algebras.