RepairNet: Contextual Sequence-to-Sequence Network for Automated Program Repair

Compile-time errors can wreak havoc for programmers – seasoned and novice. Often developers spend a lot of time debugging them. An automated system to repair such errors can be a useful aid to the developers for their productivity. In this work, we propose a deep generative model, RepairNet, that au...

Full description

Saved in:
Bibliographic Details
Published inArtificial Intelligence in Education pp. 3 - 15
Main Authors Abhinav, Kumar, Sharvani, Vijaya, Dubey, Alpana, D’Souza, Meenakshi, Bhardwaj, Nitish, Jain, Sakshi, Arora, Veenu
Format Book Chapter
LanguageEnglish
Published Cham Springer International Publishing
SeriesLecture Notes in Computer Science
Subjects
Bug fixing
Program repair
Sequence modeling
Online AccessGet full text

Cover

More Information
Summary:Compile-time errors can wreak havoc for programmers – seasoned and novice. Often developers spend a lot of time debugging them. An automated system to repair such errors can be a useful aid to the developers for their productivity. In this work, we propose a deep generative model, RepairNet, that automatically repairs programs that fail at compile time. RepairNet is based on sequence-to-sequence modeling and uses both code and error messages to repair the program. We evaluated the effectiveness of our system on 6,971 erroneous submissions for 93 programming tasks. RepairNet outperforms the existing state-of-the-art technique, MACER, with 17% relative improvement of repair accuracy. Our approach can fix 66.4% of the erroneous submissions completely and 14.2% partially.
ISBN:9783030782917
3030782913
ISSN:0302-9743
1611-3349
DOI:10.1007/978-3-030-78292-4_1