A structural taxonomy for lifted software product line analyses

Abstract

A software product line (SPL) is a structured collection of distinct software products developed from a common set of artifacts. SPLs can encompass millions of products, so analysing each product in a brute-force manner is infeasible. To analyse SPLs directly, analyses must be lifted, i.e., redefined to accommodate the semantics of SPLs. Over the past two decades, many kinds of analyses have been lifted from products to SPLs. Looking at the landscape of lifted analyses, we observe various techniques for lifting which vary across numerous dimensions. To help engineers and research navigate this landscape, we propose a classification scheme for lifted analyses based on a set of features lifted analyses can exhibit. We then conduct a systematic literature review (SLR) analysing the landscape of lifted analyses produced over the last 20 years. We analyse 140 research papers which discuss the design and implementation of lifted analyses. We provide quantitative analysis of the types of analyses which have been lifted, and apply our taxonomy to clarify how lifting was accomplished. We discuss examples of how each of the lifting methods have been applied, and identify gaps in the research literature which may provide directions for future work.