Developers of application software must often work with “legacy systems.” These are systems that have evolved over many years and are considered irreplaceable, either because it is thought that duplicating their function would be too expensive, or because they are trusted by users. Because of their age, such systems are likely to have been implemented in a conventional language with limited use of data abstraction or encapsulation. The lack of abstraction complicates adding new applications to such systems and the lack of encapsulation impedes modifying the system because applications depend on system internals. We describe our experience providing and using an object-oriented interface to a legacy system.

Abstract TGMS (tiered geometric modelling system) is an experimental system intended to reduce the difficulty and cost of developing new solid modelling applications while preserving investment in an existing, time‐tested, solid modeller. The application developer writes programs in the TGMS language, which consists of a base language augmented by data types (classes) for geometry. The base programming language is AML/X, an object‐oriented language intended for use in design and manufacturing applications. Solid modelling is done using an interface to GDP (geometric design processor), an existing, production‐quality solid modelling system. TGMS shows how a system written in a non‐object‐oriented language can be used as a base for an object‐oriented application programming environment. This paper presents the design of TGMS and discusses important aspects of its implementation. The TGMS language defines classes and methods that provide an interface to functions and data represented in a non‐object‐oriented language. To implement TGMS, issues related to communication, storage management and consistency were addressed. These, and other, implementation issues are discussed. The paper includes sample programs and a summary of feedback from TGMS users. It concludes that the use of object‐oriented programming for geometry has many benefits, and that an object‐oriented system can be built on top of a non‐object‐oriented system at less cost than building the equivalent object‐oriented system from scratch.