Intelligent code completion with Bayesian networks

Abstract

Code completion is an integral part of modern Integrated Development Environments (IDEs). Intelligent code completion systems can reduce long lists of type-correct proposals to relevant items. In this work, we replace an existing code completion engine named Best-Matching Neighbor (BMN) by an approach using Bayesian Networks named Pattern-based Bayesian Network (PBN).We use additional context information for more precise recommendations and apply clustering techniques to improve model sizes and to increase speed. We compare the new approach with the existing algorithm and, in addition to prediction quality, we also evaluate model size and inference speed. Our results show that the additional context information we collect improves prediction quality, and that PBN can obtain comparable prediction quality to BMN, while model size and inference speed scale better with large input sizes.