Identifying and Visualizing Variability in Object-Oriented Variability-Rich Systems

Abstract

In many variability-intensive systems, variability is implemented in code units provided by a host language, such as classes or functions, which do not align well with the domain features. Annotating or creating an orthogonal decomposition of code in terms of features implies extra effort, as well as massive and cumbersome refactoring activities. In this paper, we introduce an approach for identifying and visualizing the variability implementation places within the main decomposition structure of object-oriented code assets in a single variability-rich system. First, we propose to use symmetry, as a common property of some main implementation techniques, such as inheritance or overloading, to identify uniformly these places. We study symmetry in different constructs (e.g., classes), techniques (e.g., subtyping, overloading) and design patterns (e.g., strategy, factory), and we also show how we can use such symmetries to find variation points with variants. We then report on the implementation and application of a toolchain, symfinder, which automatically identifies and visualizes places with symmetry. The publicly available application to several large open-source systems shows that symfinder can help in characterizing code bases that are variability-rich or not, as well as in discerning zones of interest w.r.t. variability. CCS CONCEPTS• Software and its engineering → Software product lines; Object oriented development; Reusability. KEYWORDSIdentifying software variability, visualizing software variability, object-oriented variability-rich systems, tool support for understanding software variability, software product line engineering

symfinder

“…For a targeted variability-rich system, local symmetries are identified according to the defined symmetry in each language construct, technique, and design pattern given in the companion research paper [18] and summarized in Table 1. Specifically, each interface, abstract class, extended class, overloaded constructor, and overloaded method is identified.…”

“…The visualization itself has five main features: (1) the visualized graph of vp-s can be zoomed in and out, (2) the vp label appears when hovering the node, (3) the out-of-scope vp-s [18] can be filtered out (e.g., java.awt.images when analysing the variability of Java AWT library), (4) the isolated nodes can also be filtered out, and (5) the number of identified vp-s and variants.…”

“…This visualization enables one to discern zones of interest where variation points are strongly present and to get relevant information over concerned classes. Details about the usage of symmetry, the identification approach, and its validation over a set of large projects are available in [18].…”

“…• symfinder source code is publicly available at https://github. com/DeathStar3/symfinder, including some usage guidelines; • a video demonstrating the visualization part is available at https://www.youtube.com/watch?v=wb3U6MJ7nAM; • experimental results from [18] and an online demo are available at https://deathstar3.github.io/symfinder-demo/.…”

“…A more complete study of the symmetry in object-oriented constructs can be found in the companion research paper [18], as well as the definition of the relationship between variation points and symmetry. Basically, as variation points and variants, respectively, mark the unchanged and changeable parts in a design, they are…”