Polynomial time designs toward both BCNF and efficient data manipulation

Abstract

We define the independence-reducibility based on a modification of key dependencies, which has better computational properties and is more practically useful than the original one based on key dependencies. Using this modification as a tool, we design BCNF databases that are highly desirable with respect to updates and/or query answering. In particular, given a set U of attributes and a set F of functional dependencies over U, we characterize when F can be embedded in a database scheme over U that is independent and is BCNF with respect to F, a polynomial time algorithm that tests this characterization and produces such a database scheme whenever possible is presented. The produced database scheme contains the fewest possible number of relation schemes. Then we show that designs of embedding constant-time-maintainable BCNF schemes and of embedding independence-reducible schemes share exactly the same method with the above design. Finally, a simple modification of this method yields a polynomial time algorithm for designing embedding separable BCNF schemes.