A recursive interpreter for the Icon programming language

Abstract

Abstrac tThe implementation of the Icon programming language is more interestin g and difficult than the implementation o f many other programming languages because an expression i n Icon can generate a sequence of results . The implementatio n therefore must support contr ol backtracking in expressio n evaluation . There also are several novel control structure s related to g enerators . Because expression evaluation is limite d lexically . a full coroutine mechanism is not needed an d expression evaluation can be handled in a stack-like fashion .The implementation of Icon consists of a virtual machin e with it stack-based architecture and an interpreter that execute s the virtual machine instructions . There have been several different interpreters for Icon's virtual machine . This pape r describes a new approach in which the interpreter is calle d recursively whenever the context for expression evaluatio n changes . This recursive interpreter has the advantage of bein g conceptually clear and flexible without sacrificing efficiency .

Compiling Mercury to High-Level C Code

“…The idea of implementing nondeterminism by invoking a continuation on success and falling through on failure is not new. It has been proposed several times in the literature, in several contexts -for example Prolog meta-interpreters and implementing nondeterminism in languages such as Lisp [5] -and it is closely related to the idea of binarization of Prolog programs [15] and to the implementation of generators in languages such as Icon [13]. However, few of these papers have formal translation rules, and few consider the optimization opportunities presented by knowledge of determinism information, which can derived by program analysis even for languages such as Prolog in which determinism is not a fundamental concept.…”

A new implementation of the Icon language

“…O'Bagy and Griswold developed a technique for translating Icon that utilizes recursive interpreters [14]. The basic idea behind recursive interpreters for goal-directed evaluation is that each generator that produces a value does so by recursively invoking the interpreter.…”