Abstract
Relational data may have access limitations, i.e., relations may require certain attributes to be selected when they are accessed; this happens, for instance, while querying web data sources (wrapped in relational form) or legacy databases. It is known that the evaluation of a conjunctive query under access limitations requires a recursive algorithm that is encoded into a Datalog program. In this paper we consider the problem of optimising query answering in this setting, where the query language is that of conjunctive queries. We review some optimisation techniques for this problem, that aim to reduce the number of accesses to the data in the query plan. Then we argue that checking query containment is necessary in this case for achieving effective query optimisation. Checking containment in the presence of access limitations would amount to check containment of recursive Datalog programs, which is undecidable in general. We show however that, due to the specific form of the Datalog programs resulting from encoding access limitations, the containment problem is indeed decidable. We present a decision procedure, first presented in a paper by the authors and based on chase techniques, and we analyse its computational complexity.