Context Fusioning Multi Line Patch Generator

Manually fixing software bugs requires a significant amount of time and effort. Approximately 50% of a developer’s time is allocated to bug fixing. Automated Program Repair (APR) is a cutting-edge area of software engineering that focuses on developing techniques and tools for automatically locating and fixing software bugs without human intervention. Four primary categories of APR techniques can be identified with distinct approaches and methodologies. These categories encompass search-based, template-based, constraint-based, and learning-based techniques. Learning-based techniques focus on mining the templates for empirical information that can be used to fix bugs. These techniques can address the limitations found in other techniques such as huge search space, high human intervention, and being limited to defined templates. State-of-the-art learning-based techniques are deep learning-based techniques. Various categories of software bugs can be identified based on the code modification (patch) required, encompassing single-file single-line, multi-file single-line, single-file multi-line, and multi-file multi-line patches. The majority of APR tools available are limited in addressing bugs that need single-line patches. These APR tools fail to generate multi-line patches. However, a limited number of research has been conducted to advance the capability of generating multi-line patches for bugs.