Abstract:   

Aspect-oriented software product lines are not a new idea, but their application is facing two obstacles: establishing software product lines is challenging and aspect-oriented programming is not that widely accepted. In this paper, we address exactly these two obstacles by an approach to establishing lightweight aspect-oriented software product lines with automated product derivation. This is particularly relevant for data preprocessing systems, which are typically custom-built with respect to the data and its structure. They may involve data cleaning, reduction, profiling, validation, etc., which may have variant implementations and may be composed in different settings. The approach is simple and accessible because developers decide about variation points directly in the code without any assumption on development process and applied management. Also, it allows for variability management by making the code readable, configurable, and adaptable mainly to scripts and code fragments in a modular and concise way. The use of annotations helps preserve feature models in code. We presented the approach on the battleship game and data preprocessing pipeline product lines, which include the configuration of features, product derivation mechanism based on annotations applied by the user on certain classes and methods, implementation of features according to the feature model, and the possibility to generate all given software product derivations.

Abstract:   

The evolution of product families is complex due to the exponential growth of new products with a modified codebase. It is caused by the introduction of new such important or restrictive features that allow transforming an existing solution into a new product as the basis for another one. Specifically, both the structure of resulting components and semantic information need to be taken into account to provide in-depth supporting materials by capturing feature interactions and their capabilities. Additionally, data from heterogeneous applications usually differ and their handling needs account for different scoring. We tackled these problems by creating various supporting views based on structural and semantic information. Related applications are modeled as graphs, also instances of their components are optionally included, and various matrix based algorithms based on similarity metrics are integrated. The final integrated and automated approach is efficient because it runs in polynomial time, is extensively focused on dependencies/connections between nodes, can be adapted to big data, and is extendable and enhanced to support different metrics. Its capability to organize analyzed parts into hierarchies by applying hierarchic clustering helps to specify the context to support comprehension or find a related position of features based on their interaction, especially their coupling. With regards to possible design issues, each updated product should be checked by an expert or more autonomously against performed predictions, especially its complexity and coupling between components. Additionally, extendability can be measured from the chosen sequence of such updates. An approach including an automated matrix based feature recognition process is presented in the analysis of modular Angular applications. Our future work will be more focused on semantic enhancements, and its applications on big data by generating and analyzing various types of fractals including extracted knowledge from them.

Abstract:   

The evolution of product families is complex due to the exponential growth of new products with a modified codebase. It is caused by the introduction of new such important or restrictive features that allow transforming an existing solution into a new product as the basis for another one. Specifically, both the structure of resulting components and semantic information need to be taken into account to provide in-depth supporting materials by capturing feature interactions and their capabilities. Additionally, data from heterogeneous applications usually differ and their handling needs account for different scoring. We tackled these problems by creating various supporting views based on structural and semantic information. Related applications are modeled as graphs, also instances of their components are optionally included, and various matrix based algorithms based on similarity metrics are integrated. The final integrated and automated approach is efficient because it runs in polynomial time, is extensively focused on dependencies/connections between nodes, can be adapted to big data, and is extendable and enhanced to support different metrics. Its capability to organize analyzed parts into hierarchies by applying hierarchic clustering helps to specify the context to support comprehension or find a related position of features based on their interaction, especially their coupling. With regards to possible design issues, each updated product should be checked by an expert or more autonomously against performed predictions, especially its complexity and coupling between components. Additionally, extendability can be measured from the chosen sequence of such updates. An approach including an automated matrix based feature recognition process is presented in the analysis of modular Angular applications. Our future work will be more focused on semantic enhancements, and its applications on big data by generating and analyzing various types of fractals including extracted knowledge from them.

Abstract:   

Existing approaches in software product lines usually neglect knowledge modeling and simulation of the interaction between features capable of bringing dynamism and automation. Consequently, these solutions miss opportunities to resolve associated and emerging problems, including defect detection or quality assurance, which can be solved by effectively extracting and utilizing knowledge from data based on the differences between variants. We bring capabilities to seize them in introducing a fully automated and minimalistic approach to software product line evolution that strictly focuses on handling variability at low-level code fragments. It incorporates the autonomous modeling of emerging knowledge across preconfigured simulations. Specifically, fully automated knowledge-driven software product line evolution provides various views on an existing software product line, its variants, and their evolution through semantic and structural information accompanied by the time and order of the performed changes. We initially developed our approach for software product line evolution, followed by its successful application to the evolution of fractal scripts. We present it accordingly. Fractal products are much simpler because an application state does not span out of recursive behavior, making it manageable within the drawing. Each change is propagated into repetitive phases, causing the visual performance of the implemented feature to be infinitely detailed. Even minor changes in low code fragments tend to manifest as user-visible features owing to recursive behavior, allowing one to massively introduce new features and/or configure existing features and manage variability observable as infinitely detailed shapes. Future applications propose automated observation of more comprehensive in-code representation of feature trees.