Supporting Software Maintenance in Heterogeneous Contexts with SST-Based Framework

Abstract

Maintaining large-scale and legacy software systems is a complex and time-consuming activity, largely because relevant information is scattered across various tools and artifacts. Practitioners need to perform manual, error-prone correlation of heterogeneous data when performing essential maintenance tasks.

This thesis proposes a method to support software maintenance activities using the \emph{Universal Data Source (UDS)} framework based on the Single Source of Truth (SST) paradigm. It consists of a layer of reusable probes that extract targeted data from diverse sources (like source code, version control systems, issue trackers, and static analysis tools), integrating the collected heterogeneous data into a unified, queryable graph model managed by the SST layer, and a layer of tailored visualizations that address the specific needs of each maintenance task.

The feasibility and practical value of the proposed method are demonstrated through three real-world use cases: bug triaging, change impact analysis, and code quality enhancement. In each case, carefully designed probes and customized visualizers reduce manual efforts and help smarter decision-making. The use cases show that targeted, incremental analyses are practically achievable and deliver immediate benefits for real maintenance scenarios.