Efficient lazy narrowing using demandedness analysis

Abstract

Functional logic programming languages have a functional syntax and use narrowing as operational semantics. Here we consider the efficient implementation of lazy narrowing, a strategy which only evaluates the arguments of a function application, if their evaluation is really demanded. For an efficient implementation of lazy narrowing it is crucial to evaluate the arguments as early as possible. Otherwise the arguments are frequently reevaluated. A demandedness analysis is used to detect which parts of the arguments can safely be evaluated before the call to the function. Several approaches (e.g. [HLW92, JMM92]) also use this idea, but they sacrifice laziness in order to avoid inefficiency. Our approach is more lazy than the previous approaches, and it uses a more powerful notion of demandedness, which allows to express infinite demand patterns like e.g. spine normal form. Moreover, in contrast to the previous approaches, we take into account dependencies between the arguments of a function in order to improve the efficiency.

This work has been partially supported by the Spanish PRONTIC project TIC93-0737-C02-02