Type Transformations

Current work on data structure encapsulation and abstraction focuses attention on individual structures and permits separate local optimizations, of these structures. We extend this work by developing the beginnings of an algebra for aggregating individual data types into larger more coordinated structures which can be more effectively optimized. The present work can well be viewed as the data equivalent of cross-procedural optimization. We believe aggregations of pure abstract types are both common and essential in practical programs and that techniques for building and manipulating them must be developed before abstract specification and program transformation can become a practical programming paradigm.