For each view predicate, a Transaction Tree (T-tree) can be derived from a database schema. T-trees are data structures which serve to anticipate transactions by which view update requests can be translated to updates of underlying base relations. Since T-trees are obtained at schema specification time, i.e. in advance of update time, they provide a basis for planning strategies with which appropriate transactions can be chosen to satisfy view update requests, and also for strategies which support schema evolution. The main features of the approach are that it is simple, it offers solutions to the problem of missing values, it is correct and it is complete (in the case of nonrecursive views).

We compare and synthesize different methods for integrity checking in deductive databases. First, we state simplified integrity checking for deductive databases independently of the particular strategy used by different methods found in the literature. In accordance with this statement, we classify integrity checking methods into two main groups: methods with a generation phase without fact access and methods with a generation phase with fact access. Then, we propose an implementation scheme (a metaprogram) where the differences and similarities among the methods can be pointed out. In this common implementation framework, we compare the methods; this comparison is based on the number of facts accessed by each of them during integrity checking. Finally and from the analysis of the results, we define a convergence method which synthesizes some different features from several methods.>