Functional Specification of Imperative Programs: An Alternative Point
of View of Functional Languages

Markus Mohnen

           In this paper we will show how to embed functional
           concepts in an imperative host language in a \emph{natural
           way}. This is achieved by interpreting functional programs
           as \emph{specifications} of imperative programs.  We will
           introduce a refined \emph{notion of laziness}, which allows
           us to interpret lazy computations as special constructors,
           which encapsulate the suspended computation. Even higher-order
           concepts can be represented efficiently, if we restrict
           the possibility to use partial applications. Essentially,
           we disallow the use of sequences of partial applications as
           hidden data structures. By considering a setting in which the
           unit of translation is a \emph{module}, this approach is the
           key to an \emph{integration} of the functional programming
           paradigm and the imperative programming paradigm on a
           \emph{module level}.  Functional components of the resulting
           \emph{hybrid} module architectures \emph{may use} and \emph{may
           be used by} imperative components without need for special
           \emph{inter-language interfaces}.  Beside the `best of both
           worlds', this concept offers the possibility to \emph{migrate}
           `module by module' from purely imperative systems to hybrid
           systems and to do \emph{rapid prototyping} using functional
           modules.  Experimental results with a prototype version of
           a translator will show that we can compete with traditional
           functional languages like \texttt{Haskell}.