An architectural blueprint for a pluggable version control system for software (evolution) analysis

Abstract

Abstract-Current version control systems are not built to be systematically analyzed. They have greatly evolved since their first appearance, but their focus has always been towards supporting developers in forward engineering activities. Supporting the analysis of the development history has so far been neglected. A plethora of third party applications have been built to fill this gap. To extract the data needed, they use interfaces that were not built for that. Drawing from our experience in mining and analyzing version control repositories, we propose an architectural blueprint for a plug-in based version control system in which analyses can be directly plugged into it in a flexible and lightweight way, to support both developers and analysts. We show the potential of this approach in three usage scenarios and we also give some examples for these analysis plug-ins.

Evolution of technical debt remediation in Python: A case study on the Apache Software Ecosystem

“…One limitation of SonarQube (and all the other equivalent tools that we have investigated) is that the evolutionary analysis is not incremental. 28 This means that a new version (e.g., originated from a new commit) is fully parsed and analyzed, even if only a single line in a single file is changed. This imposes constraints on evolution analysis because it becomes prohibitive to analyze all the commits of a system.…”

Fostering Real-Time Software Analysis by Leveraging Heterogeneous and Autonomous Software Repositories

“…It was further observed that state-of-the-art tools in program comprehension are either unknown or rarely utilized. In addition to the question catalogs, Ghezzi and Gall [11] presented a set of usage scenarios in the context of software analysis.…”

“…The current practice is that the VCSs only keep track of the source code. It means that to calculate metrics, the entire source code needs to be fetched from the repository and parsed or even partially compiled [11]. For example, features like test coverage, test success, code clones, and complexity are critical in observ-ing the continuous code quality of a project.…”

Software Residence Application in the Versions of a Software Product Line

“…Second (Ghezzi, Wursch, Giger, & Gall, 2012) the actual concept of a version control system and its first implementation was introduced by (Rochkind, 1975). In this publication, the author called the source code control system as a tool to assist the project schedule controlling changes in the source code.…”