AN ARCHITECTURE-BASED COMPLEXITY CLASSIFICATION FRAMEWORK AND METRICS FOR ANALYZING THE MODIFIABILITY OF LARAVEL SOFTWARE

Abstract

In the modern world of web development, frameworks like Laravel have become essential tools for building scalable and maintainable web applications. Laravel, a popular PHP framework, utilizes a Model-View-Controller (MVC) architecture, which provides significant benefits in terms of quality of Laravel software. However, as applications built with Laravel grow in size and complexity, modifying them becomes increasingly challenging. This thesis aims to address these challenges by developing a classification framework and metrics to analyzing and managing the inherent complexity within Laravel software. To tackle, an Architecture-Based Complexity Classification Framework for Laravel Software (ACCFLS) was then developed and implemented to identify and classify the unique attributes that causes complexity in Laravel software based on the MVC architecture. Using snowballing, 40 Laravel industry experts were engaged to validate the proposed framework on its functional suitability. The experts affirmed that the model is functionally appropriate, correct and complete to guide in the identification and classification of attributes that cause complexity in Laravel software. Next, a suite of nine new structural complexity metrics was defined based on the proposed framework. Among them were three composite metrics, namely Controller Complexity Metrics for Laravel (CCMLV), Model Complexity Metrics for Laravel (MCMLV), and View Complexity Metrics for Laravel (VCMLV). The metrics were theoretically validated successfully using Weyuker’s nine properties and Kaner framework proving their mathematical soundness and practicality respectively. A prototype metrics tool was designed and developed to automate the collection and calculation of these complexity metrics using a real-world Laravel software. An empirical validation of the metrics was also done through controlled laboratory within-the subject experiment involving a sample of 52 final year computing students from MUT reached through systematic random sampling. For the modifiability ratings, the results gave rs of -0.938**, p-values of < 0.001, -0.972**, p-values of < 0.001 and -0.800**, p-values of 0.005 and R2of 0.893, p-values of < 0.001, 0.993, p-values of < 0.001 and 0.594, p-values of 0.009, for the time taken to modify, the results have rs of 0.939**, p-value of < 0.001, 0.924**, p-value of < 0.001 and 0.794**, p-value of < 0.006and R2of 0.823, 0.831, 0.856 and a p-value of < 0.001 in all instances for the CCMLV, MCMLV and VCMLV respectively. The results ascertained that the defined metrics are predictors of modifiability of Laravel software. Results confirmed a strong correlation between higher complexity scores and longer modification times and reduced ease of modifying the Laravel software, with the rs. of -0 Finally, the Analytical Hierarchical Process (AHP) technique was developed to rank the metrics based on their impact on Laravel complexity, providing developers with a clear understanding of which aspects of Laravel contribute most to its complexity. Results show that metrics related to controller function complexity and model relationships had the greatest influence on modifiability. In conclusion, this research provides a comprehensive framework, novel Laravel structural complexity metrics and a set of practical tools that help developers assess and manage the complexity of Laravel software. By automating the measurement of complexity and providing a clear method for prioritizing refactoring efforts, this work offers valuable insights into improving the modifiability of Laravel applications. The findings not only contribute to the academic field of software engineering but also offer practical solutions that developers can apply in real-world Laravel projects to enhance code quality, streamline development, and reduce maintenance costs. In future, the proposed framework and metrics can be extended to capture additional complexity dimensions, also further validations can be done to the proposed metrics in industry settings.