Dataflow computing and eager and lazy evaluations

Abstract

Eager and lazy evaluations in a dataflow model are proposed, Such evaluation enables nonstrict evaluation, structure data manipulation and nondeterminate computation. Several dataflow computation models are discussed from the viewpoint of their by-value and by-reference mechanisms, i. e., their token to data correspondence. It is shown that effective implementation is achieved by unifying both mechanisms. This implies the effective implementation of the lenient cons and lazy cons concept in list manipulation. Nonstrict list manipulation is shown to be useful for stream-oriented processing, and for nondeterminate computation combined with the nonstrict primitive operator, Arbiter. Several sample programs are included to show that concurrent processes and object-oriented programs can be intuitively described in functional language.