Metamodelling - State of the Art and Research Challenges

This chapter discusses the current state of the art, and emerging research challenges, for metamodelling. In the state-of-the-art review on metamodelling, we review approaches, abstractions, and tools for metamodelling, evaluate them with respect to their expressivity, investigate what role(s) metamodels may play at run-time and how semantics can be assigned to metamodels and the domain-specific modeling languages they could define. In the emerging challenges section on metamodelling we highlight research issues regarding the management of complexity, consistency, and evolution of metamodels, and how the semantics of metamodels impacts each of these. 3.1 Metamodelling: State of the Art Models are powerful tools to express the structure, behavior, and other properties in mathematics, each of the hard sciences and in all areas of engineering. While models are very common, an explicit definition of a modelling language and an explicit manipulation of its models is tightly connected to computer based tools. Additional power can be gained by explicit definition and computer based manipulation of models e.g. in CAD, control engineering, algebraic mathematics and of course computer science. To be able to manipulate models, their language needs to be specified as model of these models—metamodels. In this section, we describe the state of the art for metamodelling, including the metamodelling of data structures, as well as the metamodelling of languages systems where appropriate. 3.1.1 Concepts in Metamodelling Metamodelling (literally, “beyond Modelling”) is the Modelling of models. In their most common use, metamodels describe the permitted structure to which models must adhere [1]; although out of the scope of this chapter, meta-metamodels formally describe metamodels, as they define the core abstractions H. Giese et al. (Eds.): MBEERTS, LNCS 6100, pp. 57–76, 2010. Springer-Verlag Berlin Heidelberg 2010 www.se-rwth.de/publications 58 J. Sprinkle et al. permitted in metamodelling. In fact, some metamodelling languages are selfdescriptive [2, 3]. A metamodel therefore describes the syntax of the models [4]. Through various extension mechanisms and additional rules with this representation of the syntax of models metamodels can also help to define the semantics of models, as we discuss later. The layered approach to modelling (through metamodelling) is depicted in Figure 3.1. Model Metamodel Meta-metamodel Semantic Artifacts (e.g., code, simulation) defines