On the Computational Aspect of Rule Based Database Updates

Abstract

We develop a general framework of rule based database update on the basis of prioritized logic programs. This approach has advantages in representing and handling information conflict in updates. We then focus on the computational aspect of this approach. We show that under our framework, the model checking in an update is co-NP-complete, while the associated inference problem is II ^P ₂ -complete. From a practical viewpoint, we propose two strategies to simplify an update specification so that under some conditions, solving an update problem can be based on solving some smaller components of the original problem. Our experimental result shows that these strategies indeed significantly reduce the time cost of evaluating an update specification.