[Paper Review] Problems in Systematic Application of Software Metrics and Possible Solution

Gordana Rakić, Zoran Budimac|arXiv (Cornell University)|Nov 15, 2013

Software Engineering Research7 references9 citations

TL;DR

This paper identifies key limitations in existing software metrics tools—particularly their language-specific design and lack of extensibility—and proposes a new, language-agnostic metrics tool built on a modular architecture. The prototype demonstrates independence from input programming languages through a unified intermediate representation, enabling systematic, reusable application of software metrics across diverse codebases.

ABSTRACT

Systematic application of software metric techniques can lead to significant improvements of the quality of a final software product. However, there is still the evident lack of wider utilization of software metrics techniques and tools due to many reasons. In this paper we investigate some limitations of contemporary software metrics tools and then propose construction of a new tool that would solve some of the problems. We describe the promising prototype, its internal structure, and then focus on its independency of the input language.

Motivation & Objective

• Address the widespread underutilization of software metrics in practice due to tooling limitations.
• Identify core shortcomings in existing software metrics tools, including language dependency and poor extensibility.
• Develop a new tool architecture that supports systematic, repeatable application of software metrics across multiple programming languages.
• Ensure the tool's independence from input source code languages through a language-agnostic internal representation.
• Provide a prototype demonstrating feasibility and extensibility of the proposed architecture in real-world software quality assessment.

Proposed method

• Design a modular software metrics tool with a pluggable architecture to support various programming languages.
• Introduce a unified intermediate representation (IR) that decouples metric computation from source language parsing.
• Implement a front-end parser for each target language that translates source code into the common IR.
• Develop a back-end engine that applies metrics to the IR, ensuring language-agnostic metric evaluation.
• Use a plugin-based design to allow easy extension for new languages and metrics without modifying core logic.
• Validate the approach using a prototype that supports multiple languages and demonstrates consistent metric computation.

Experimental results

Research questions

• RQ1 What are the primary limitations of current software metrics tools that hinder their systematic adoption?
• RQ2 How can a software metrics tool be designed to be independent of the input programming language?
• RQ3 Can a unified intermediate representation enable consistent and reusable application of software metrics across different languages?
• RQ4 To what extent does the proposed architecture improve extensibility and maintainability of software metrics tools?
• RQ5 How does the prototype demonstrate the feasibility of language-agnostic software metrics analysis?

Key findings

• Existing software metrics tools suffer from strong language dependency, limiting their reusability and scalability across diverse codebases.
• The proposed tool architecture successfully decouples metric computation from source language parsing using a common intermediate representation.
• The prototype demonstrates that metrics can be consistently applied across multiple programming languages without rewriting core logic.
• The modular, plugin-based design enables easy integration of new languages and metrics, enhancing maintainability and extensibility.
• The tool's independence from input language is validated through implementation and testing on multiple language targets.
• The approach provides a viable foundation for systematic, repeatable, and scalable software quality assessment in industrial and academic settings.