Optimization of Very High Level Languages - II. Deducing Relationships of Inclusion and Membership

SETL is a set-theoretically oriented language of very high level whose reportoire of semantic objects includes finite sets, ordered n-tuples, and sets of ordered n-tuples useable as mappings. This two-part paper studies the optimization problems associated with such a language. The first issue studied is that of copy optimization, i.e. the discovery of cases in which changes to a compound data object can be made ''differentially'' without recopying the whole of the object. The second part of the paper sets forth techniques for the automatic discovery of relationships of inclusion and membership on which automatic data-structure choice may be based. The methods used extend known techniques of data-flow analysis and of logical program analysis.