Model-driven support for product line evolution on feature level

Abstract

Please cite this article in press as: Pleuss, A., et al., Model-driven support for product line evolution on feature level. J. Syst. Software (2011), doi:10.1016/j.jss.2011.08.008 Software Product Lines (SPL) are an engineering technique to efficiently derive a set of similar products from a set of shared assets. In particular in conjunction with model-driven engineering, SPL engineering promises high productivity benefits. There is however, a lack of support for systematic management of SPL evolution, which is an important success factor as a product line often represents a long term investment. In this article, we present a model-driven approach for managing SPL evolution on feature level. ARTICLE IN PRESSTo reduce complexity we use model fragments to cluster related elements. The relationships between these fragments are specified using feature model concepts itself leading to a specific kind of feature model called EvoFM. A configuration of EvoFM represents an evolution step and can be transformed to a concrete instance of the product line (i.e., a feature model for the corresponding point in time). Similarly, automatic transformations allow the derivation of an EvoFM from a given set of feature models. This enables retrospective analysis of historic evolution and serves as a starting point for introduction of EvoFM, e.g., to plan future evolution steps.

Extraction and evolution of architectural variability models in plugin-based systems

“…In [22,53], Dhungana et al report that evolution support becomes particularly important for engineering SPLs and other variabilityintensive systems. They propose model-driven support at the feature level, using FM concepts [53]. They developed a set of operators to make evolve FMs but no reasoning and differencing techniques are proposed to control the evolution of the FMs.…”

“…Moreover the integration of our extraction and reasoning procedures with existing model-driven approaches to SPL engineering and evolution [54] should be further investigated, with the ultimate goal to ease the evolution of large families of software systems. In this context, we intend to couple the evolution of the architectural feature models to software development tracking information, such as commits in version control systems.…”

“…In this context, the task of building an architectural FM is very arduous for software architects. They typically have to deal with lots of plugins (usual customizations of the Eclipse IDE are made with several hundreds of plugins, corresponding to dozens of high-level features [27,53]), for which safe composition is the topmost requirement [40].…”

Feature Model Differences

“…A number of formalizations (e.g., [25,10]), automated reasoning operations [7] and tools (e.g., [20,4,27]) have been developed to address this issue. The formalism of FMs is now at the core of many generative or model-based approaches [9,24,23,11].…”

“…In [11,23], the authors report that evolution support becomes particularly important for engineering SPLs and other variability-intensive systems. They propose model-driven support at the feature level, using FM concepts [23]. Lotufo et al study the evolution of the Linux kernel variability model [16].…”