Hoogle⋆: Constants and λ-abstractions in Petri-net-based Synthesis using Symbolic Execution


Type-directed component-based program synthesis is the task of automatically building a function with applications of available components and whose type matches a given goal type. Existing approaches to component-based synthesis, based on classical proof search, cannot deal with large sets of components. Recently, Hoogle+, a component-based synthesizer for Haskell, overcomes this issue by reducing the search problem to a Petri-net reachability problem. However, Hoogle+ cannot synthesize constants nor λ-abstractions, which limits the problems that it can solve. We present Hoogle⋆, an extension to Hoogle+ that brings constants and λ-abstractions to the search space, in two independent steps. First, we introduce the notion of wildcard component, a component that matches all types. This enables the algorithm to produce incomplete functions, i.e., functions containing occurrences of the wildcard component. Second, we complete those functions, by replacing each occurrence with constants or custom-defined λ-abstractions. We have chosen to find constants by means of an inference algorithm: we present a new unification algorithm based on symbolic execution that uses the input-output examples supplied by the user to compute substitutions for the occurrences of the wildcard. When compared to Hoogle+, Hoogle⋆ can solve more kinds of problems, especially problems that require the generation of constants and λ-abstractions, without performance degradation.

In 37th European Conference on Object-Oriented Programming (ECOOP 2023), 2023
Awarded: Artifact Evaluation Award (Available and Reusable)
CORE A conference

Computer Scientist

My research interests include software reliability, software verification, and formal methods applied to software engineering. I am also interested in interactive storytelling. For more details, see some of my projects or my selected (or recent) publications. More posts are available in my blog. Follow me on Twitter or add me on LinkedIn.