Programmable View Update Strategies on Relations

TL;DR

This paper introduces a Datalog-based language and validation framework for programmable, validatable view update strategies in relational databases, addressing ambiguity and improving efficiency.

Contribution

It proposes a novel Datalog approach with validation algorithms and an optimization method for practical, reliable view updates in databases.

Findings

Validation algorithm ensures well-behavedness of Datalog programs

Fragment of Datalog is both sound and complete for validation

Experimental results demonstrate feasibility and efficiency

Abstract

View update is an important mechanism that allows updates on a view by translating them into the corresponding updates on the base relations. The existing literature has shown the ambiguity of translating view updates. To address this ambiguity, we propose a robust language-based approach for making view update strategies programmable and validatable. Specifically, we introduce a novel approach to use Datalog to describe these update strategies. We propose a validation algorithm to check the well-behavedness of the written Datalog programs. We present a fragment of the Datalog language for which our validation is both sound and complete. This fragment not only has good properties in theory but is also useful for solving practical view updates. Furthermore, we develop an algorithm for optimizing user-written programs to efficiently implement updatable views in relational database management systems. We have implemented our proposed approach. The experimental results show that our framework is feasible and efficient in practice.