Ensuring Efficiently the Integrity of Persistent Object Systems via Abstract Interpretation

In this paper, we propose an eecient and reliable method to deal with integrity constraints in a persistent object system. First we provide the application programmer with the ability to express integrity constraints but we also give him the possibility to use high level language constructs to help him in writing safe transactions. The goal of our approach is to avoid the (run time) checking of constraints by proving formally that transactions preserve integrity constraints. We mainly use two abstract interpretation techniques to do that. Abstract interpretation is a semantics-based tool that yields some reliable information about the possible run-time behaviour of programs, with fully automatic algorithms. We present informally the methods that we use: a simple method, based on path reachability, and a more powerful and complex method that uses a predicate transformer. A predicate transformer is a function that, given a transaction and a formula describing its input data, yields a formula describing its output data. We nally describe the current prototype that applies those diierent techniques. It provides in fact the O 2 compiler with an integrity constraint manager.