Abstract

Object-oriented programs frequently deal with hierarchical forms of data, often called whole-part arrangements. When in such arrangements the methods in the parts must compute information about their container, the programmer is currently forced to manually maintain a two-way relationship, also known as friend pointers. The maintenance problem becomes particularly complex in the case of deeply nested hierarchies. To eliminate this drudgery of pointer maintainenace, Gil and Lorenz proposed a language construct, dubbed environmental acquisition, in 1996. The construct has been adapted and studied in several different contexts over the past decade. In this paper, we analyze Gil and Lorenz’s language design with a formal model. Specifically, we extend ClassicJava with environmental acquisition and determine what it takes to prove the revised type system sound. Then we study a number of design decisions in this framework and illustrate when a formal model can help (or not) make such decisions. We also discuss how this model helps us understand what a full-scale implementation of type-sound environmental acquisition would require. 1 From Patterns to Language Constructs Object-oriented programs often manipulate data in the form of nested containers, which of course are equivalent

Citations