Chapter

Domain-Specific Program Generation

Volume 3016 of the series Lecture Notes in Computer Science pp 239-253

Retrofitting the AutoBayes Program Synthesis System with Concrete Syntax

  • Bernd FischerAffiliated withRIACS / NASA Ames Research Center
  • , Eelco VisserAffiliated withInstitute of Information and Computing Sciences, Universiteit Utrecht

* Final gross prices may vary according to local VAT.

Get Access

Abstract

AutoBayes is a fully automatic, schema-based program synthesis system for statistical data analysis applications. Its core component is a schema library, i.e., a collection of generic code templates with associated applicability constraints which are instantiated in a problem-specific way during synthesis. Currently, AutoBayes is implemented in Prolog; the schemas thus use abstract syntax (i.e., Prolog terms) to formulate the templates. However, the conceptual distance between this abstract representation and the concrete syntax of the generated programs makes the schemas hard to create and maintain.

In this paper we describe how AutoBayes is retrofitted with concrete syntax. We show how it is integrated into Prolog and describe how the seamless interaction of concrete syntax fragments with AutoBayes’s remaining “legacy” meta-programming kernel based on abstract syntax is achieved. We apply the approach to gradually migrate individual schemas without forcing a disruptive migration of the entire system to a different meta-programming language. First experiences show that a smooth migration can be achieved. Moreover, it can result in a considerable reduction of the code size and improved readability of xthe code. In particular, abstracting out fresh-variable generation and second-order term construction allows the formulation of larger continuous fragments.