Benchmarking bidirectional transformations: theory, implementation, application, and assessment
Published in Software and Systems Modeling (SoSym), 2020
DOI: 10.1007/s10270-019-00752-x
Abstract
Bidirectional transformations (bx) are relevant for a wide range of application domains. While bx problems may be solved with unidirectional languages and tools, maintaining separate implementations of forward and backward synchronizers with mutually consistent behavior can be difficult, laborious, and error-prone. To address the challenges involved in handling bx problems, dedicated languages and tools for bx have been developed. Due to their heterogeneity, however, the numerous and diverse approaches to bx are difficult to compare, with the consequence that fundamental differences and similarities are not yet well understood. This motivates the need for suitable benchmarks that facilitate the comparison of bx approaches. This paper provides a comprehensive treatment of benchmarking bx, covering theory, implementation, application, and assessment. At the level of theory, we introduce a conceptual framework that defines and classifies architectures of bx tools. At the level of implementation, we describe Benchmarx, an infrastructure for benchmarking bx tools which is based on the conceptual framework. At the level of application, we report on a wide variety of solutions to the well-known Families-to-Persons benchmark, which were developed and compared with the help of Benchmarx. At the level of assessment, we reflect on the usefulness of the Benchmarx approach to benchmarking bx, based on the experiences gained from the Families-to-Persons benchmark.