This thesis contributes to improving support for complex textual modeling languages. This support refers to the automatic generation of tools for the end user -- e.g. parsers, editors, views, etc. -- and (parts of) the standard specifications defined to describe the languages -- e.g. the Object Management Group open specifications for varied modeling languages.
The particular subset of languages considered in the thesis are textual, model-based and complex. They are considered textual when their concrete syntax is textual, in particular, defined by a grammar. They are considered model-based when their abstract syntax is defined by a meta-model. They are considered complex when there is a significant gap between the concrete and abstract syntax of the language; in other words, when the abstract syntax meta-model cannot directly be derived or inferred from the concrete syntax grammar.
The contributions of this thesis address the problem of bridging the concrete and abstract syntax of complex textual modeling languages. In particular, the contributions include (a) a gap analysis of the limitations of related work; (b) a domain-specific transformation language for defining and executing concrete syntax to abstract syntax bridges; (c) an experimental evaluation of the proposed solution including studies to compare with related work.
Existing related work presents different issues when working on complex textual modeling languages. Either sufficient automatic tooling generation is not provided (Gra2Mol), or model-based languages are not appropriately supported (Spoofax), or complex gaps between the concrete and abstract syntax cannot be bridged (Xtext).
This thesis identifies the different concerns that arise when bridging the concrete and abstract syntax of complex textual modeling languages. In addition, some limitations of relevant related work are shown. With the aim of addressing these identified concerns a new approach is proposed, showing how these concerns are particularly addressed. Specifically, the proposed approach consists of complementing relevant related work (Xtext) with a novel domain-specific transformation language to declare bridges between the concrete syntax and abstract syntax of complex textual modeling languages.
The domain-specific transformation language is the main contribution of this thesis and is evaluated by means of qualitative and quantitative studies. Subject to the presented examples, the conducted experiments show that the proposed approach brings measurable benefits -- in terms of size of specification artefacts and execution time of the underlying implementation -- when compared to the state-of-the-art.
