Transformation of data traversals and operations in application programs to account for semantic changes of databases

Abstract

This paper addresses the problem of application program conversion to account for changes in database semantics that result in changes in the schema and database contents. With the observation that the existing data models can be viewed as alternative ways of modeling the same database semantics, a methodology of application program analysis and conversion based on an existing-DBMS-model-and schema-independent representation of both the database and programs is presented. In this methodology, the source and target databases are described in terms of the association types of a semantic association model. The structural properties, the integrity constraints, and the operational characteristics (storage operation behaviors) of the association types are more explicitly defined to reveal the semantics that is generally hidden in application programs. The explicit descriptions of the source and target databases are used as the basis for program analysis and conversion. Application programs are described in terms of a small number of "access patterns" which defme the data traversals and operations of the programs. In addition to the methodology, this paper (1) describes a model of a generalized application program conversion system that serves as a framework for research, (2) presents an analysis of access patterns that serve as the primitives for program description, (3) delineates some meaningful semantic changes to databases and their corresponding transformation rules for program conversion, (4) illustrates the application of these rules to two different approaches to program conversion problems, and (5) reports on the development effort undertaken at the University of Florida.

Reverse engineering of relational database applications

“…For general-purpose application code, this problem may be next to unsolvable in an automatic way. This is probably the reason why work on database application program analysis has been very scarce, and the semantic information extracted by approaches like the one of [9] is rather limited. In this paper, we present techniques for attacking the problem anyway, in which we exploit the particular characteristics of relational database application software.…”