Copying and Comparing: Problems and Solutions

In object oriented programming, it is sometimes necessary to copy objects and to compare them for equality or inequality. We discuss some of the issues involved in copying and comparing objects and we address the problem of generating appropriate copying and comparing operations automatically, a service that is not provided by most object oriented languages and environments. Automatic generation appears to be not only desirable, because hand-coding these methods is mechanical and yet error-prone, but also feasible, because the form of the code is simple and largely predictable. Some languages and some object models presented in the literature do support generic copying and comparing, typically defining separate “shallow” and “deep” versions of both operations. A close examination of these definitions reveals inadequacies. If the objects involved are simple, copying and comparing them is straightforward. However, there are at least three areas in which insufficient attention has been given to copying and comparing complex objects: (1) values are not distinguished from objects; (2) aggregation is not distinguished from association; and (3) the correct handling of linked structures other than trees is neglected. Solving the third problem requires a mechanism built into the language, such as exists in Eiffel. Building such a mechanism without modifying the language requires a language with sufficient reflexive facilities, such as Smalltalk. Even then, the task is difficult and the result is likely to be insecure. We show that fully automatic generation of copying and comparing operations is not feasible because compilers and other software tools have access only to the structure of the objects and not to their semantics. Nevertheless, it is possible to provide default methods that do most of the work correctly and can be fine-tuned with a small additional amount of hand-coding. We include an example that illustrates the application of our proposals to C++. It is based on additional declarations handled by a preprocessor.