Static semantics, types, and binding time analysis

Part 1 of this paper is an analysis of problems concerning type systems and static semantics relevant to designing and implementing new programming languages. Part 2 proposes and exemplifies a way to derive a static semantics from a language's dynamic semantics, using a technique known as binding time analysis known from partial evaluation (Ershov's “mixed computation”) 20, 21]. The goal is to analyze the language's type structure during the design process without the need for hand construction of a type system. Ideally, one might automate the construction of a type checker, given only the dynamic semantics as input. In particular, binding time analysis of the interpreter program can be used to distinguish statically detectable errors from dynamic ones, and a static semantics is seen as a mechanism for deciding whether the program to be interpreted can give rise to any of the static errors. Our basic viewpoint is operational: to see how and where in the interpreter immediately observable type errors manifest themselves; to classify them as static or dynamic; and to design the static semantics to detect all static errors in a given program.