Symmetric Relations and Cardinality-Bounded Multisets in Database Systems

摘要

In a binary symmetric relationship, A is related to B if and only if B is related to A. Symmetric relationships between k participating entities can be represented as multisets of cardinality k. Cardinality-bounded multisets are natural in several real-world applications. Conventional representations in relational databases suffer from several consistency and performance problems. We argue that the database system itself should provide native support for cardinality-bounded multisets. We provide techniques to be implemented by the database engine that avoid the drawbacks, and allow a schema designer to simply declare a table to be symmetric in certain attributes. We describe a compact data structure, and update methods for the structure. We describe an algebraic symmetric closure operator, and show how it can be moved around in a query plan during query optimization in order to improve performance. We describe indexing methods that allow efficient lookups on the symmetric columns. We show how to perform database normalization in the presence of symmetric relations. We provide techniques for inferring that a view is symmetric. We also describe a syntactic SQL extension that allows the succinct formulation of queries over symmetric relations.