We outline a research proposal which goal is to contribute to methods of new Domain-Specific Languages (DSLs) definition and implementation. We propose the automata based method for DSLs definition that allows specifying new languages with various notations in such a way that the language definition can be treated as a ready-to-use language implementation already. The automata based method allows defining language by three components: language metamodel (which includes an abstract syntax), concrete syntax and operational semantics. We use Unified Modeling Language (UML) as description formalism for all three components. Namely, language metamodel is defined using class diagrams. Concrete syntax is defined as parser using state machine diagrams. Semantics is defined as metamodel interpreter using state machine diagrams as well.
Domain-specific Languages (DSLs) are considered to be very effective in software
engineering. They raise the level of abstraction, provide common domain notation,
and improve development process therefore. Per se, DSLs allow describing a problem
solution in terms of the field of the problem, rather than in terms of computer. The
most critical part of the whole software development process in the context of
language oriented programming paradigm is definition and implementation of a new
DSL [14]. All approaches to solve this problem could be divided into those, which
use traditional grammars, and those, which are based on modeling in context of
Model Driven Engineering (MDE) [
However, these two styles of language definition do not differ in essence. As was
noticed in [
Another issue is specification of language semantics. The brief overview of
methods of semantics definition is given in [
At last, specification of new DSL should be sufficient for receiving its implementation automatically.
A number of language workbenches, which support development of DSLs with not
only textual notation, were worked out recently. One of them is MetaEdit+ tool [
Another one, AMMA [
We appeal for possibility to define various notations using single technique and for possibility to define both concrete syntax and semantics using the same specification (meta)language.
In this work, we propose just another DSL definition method based on model driven
architecture (MDA) and executable UML approach [
We extend formalism used for language abstract syntax definition to UML class diagram in comparison with widely spread MOF, as UML class diagram can express more copious structures. We propose description both of concrete syntax and of operational semantics as algorithms through UML state machine diagrams. Namely, concrete syntax is defined as a parsing algorithm, which analyzes source program representation and constructs abstract program. Operational semantics is defined as an algorithm of abstract program interpretation. One of the key features of the proposed method is unified view on different language notations. Any abstract program representation is considered as chain of events, which are processed by automata system defined in concrete syntax specification. Each terminal representation is considered as an event sent by some source of events. For example, events might be typographical characters in text, geometrical figures in diagram, controls in dialogue window, cells in spreadsheet, or sounds of spoken commands. Accordingly, any source of events is acceptable: a lexical analyzer of text, a graphical editor of diagrams, an editor of formulas or a dialogue window (Fig. 2).
To achieve automatic receiving of DSL's implementation we use automata based
programming paradigm [
We have developed an initial candidate DSL meta-metamodel – the abstract syntax of the proposed method (Fig. 3). This meta-metamodel accumulates expressiveness both of grammar formalism and of UML class diagram. We are going to investigate it and compare it with formal grammars to find out the kind of languages that could be defined as instances of this meta-metamodel. Moreover, mapping between DSL metametamodel and formal grammars could be useful for the reuse of already defined languages. This mapping could be also useful for development of the algorithm of generation of concrete syntax automata system from DSL metamodel.
DSL Metamodel name : String
Taxonomy * 1 classification specific 1..* 1 root axiom Language conce1pts concept
* name : String
* component
*
Clause alternative 1..*
Definition name : String
Static semantics
Constraint condition : bool * crossreference
Relation
Occurance multiplicity: Num role: String
We have devised the automata model being used as formalism for definition of both concrete syntax and semantics. Two little DSLs have been specified using this model: nonlinear and graphical language of the chess position and the mini language for manipulations with sets. Further, the automata programming virtual machine should be developed to execute these specifications of DSLs. Definition of the automata programming virtual machine with the means of the proposed method would be the best use case. In other words, we are going to apply the idea of bootstrapping.
The proposed method for DSL definition allows defining different languages with various concrete representations. DSL definition can be used as its software implementation in terms of the automata based virtual machine.