Using Inheritance to Provide Subschema Views in a Shared Persistent Object Database

Object-oriented persistent programming reduces the ‘impedance mismatch’ between application programs and the stored data which they need to manipulate. This reduction is especially beneficial in applications such as CAD data management, where the data concerned has a complex and dynamic structure. The authors are using an implementation of persistent C++ to write multiple CAD applications sharing a single persistent store. One problem that has arisen from compile-time class definition in C++ is that not all classes may be compiled into all applications; this leads to problems because programs may access objects of unrecognised classes through polymorphic references. Other persistent C++ implementations require that an application program be recompiled with the appropriate class definition before objects of that class may be accessed. This is inconvenient when there are many programs and the global schema is frequently extended. This paper presents a mechanism by which objects of unrecognised classes may be loaded and treated as instances of a recognised superclass, allowing easy addition of new classes to the global schema without requiring recompilation of any existing application programs.