Neural Sketch Learning for Conditional Program Generation

TL;DR

This paper introduces a neural sketch learning approach for conditional program generation, enabling the prediction of complete, type-safe Java methods from minimal input labels like API calls or data types.

Contribution

It proposes training neural models on program sketches to better handle syntax and semantic constraints in code generation tasks.

Findings

Effective prediction of Java methods from limited labels

Sketch-based neural training improves syntactic and semantic correctness

System successfully generates API-heavy Java code from minimal input

Abstract

We study the problem of generating source code in a strongly typed, Java-like programming language, given a label (for example a set of API calls or types) carrying a small amount of information about the code that is desired. The generated programs are expected to respect a "realistic" relationship between programs and labels, as exemplified by a corpus of labeled programs available during training. Two challenges in such conditional program generation are that the generated programs must satisfy a rich set of syntactic and semantic constraints, and that source code contains many low-level features that impede learning. We address these problems by training a neural generator not on code but on program sketches, or models of program syntax that abstract out names and operations that do not generalize across programs. During generation, we infer a posterior distribution over sketches, then concretize samples from this distribution into type-safe programs using combinatorial techniques. We implement our ideas in a system for generating API-heavy Java code, and show that it can often predict the entire body of a method given just a few API calls or data types that appear in the method.