this paper we study the constrained equivalence of programs with effects. In particular, we present a formal system for deriving such equivalences. Constrained equivalence is defined via a model theoretic characterization of operational, or observational, equivalence called strong isomorphism. Operational equivalence, as introduced by Morris [23] and Plotkin [27], treats programs as black boxes. Two expressions are operationally equivalent if they are indistinguishable in all program contexts. This equivalence is the basis for soundness results for program calculi and program transformation theories. Strong isomorphism, as introduced by Mason [14], also treats programs as black boxes. Two expressions are strongly isomorphic if in all memory states they return the same value, and have the same effect on memory (modulo the production of garbage). Strong isomorphism implies operational equivalence. The converse is true for first-order languages; it is false for full higher-order languages. However, even in the higher-order case, it remains an useful tool for establishing equivalence. Since strong isomorphism is defined by quantifying over memory states, rather than program contexts, it is a simple matter to restrict this equivalence to those memory states which satisfy a set of constraints. It is for this reason that strong isomorphism is a useful relation, even in the higher-order case. The formal system we present defines a single-conclusion consequence relation \Sigma ` \Phi where \Sigma is a finite set of constraints and \Phi is an assertion. The semantics of the formal system is given by a semantic consequence relation, \Sigma j= \Phi, defined in terms of a class of memory models for assertions and constraints. The assertions we consider are of the following two forms...

In this paper we report progress in the development of methods for reasoning about the equivalence of objects with memory, and the use of these methods to describe sound operations on objects in terms of formal program transformations. We also formalize three different aspects of objects: their specification, their behavior, and their canonical representation. Formal connections among these aspects provide methods for optimization and reasoning about systems of objects. To illustrate these ideas we give a formal derivation of an optimized specialized window editor from generic specifications of its components. A new result in this paper enables one to make use of symbolic evaluation (with respect to a set of constraints) to establish the equivalence of objects. This form of evaluation is not only mechanizable, it is also generalizes the conditions under which partial evaluation usually takes place. 1 Overview In [19] a general challenge for partial evaluation technology was presented ...