Specification of Graph Translators with Triple Graph Grammars

To summarize, triple graph grammars are a new formalism for the specification of complex interdependencies between separate and, in general, quite different graph-like data structures. Comparing them with previously suggested formalisms for the specification of data transformations they have the following advantages: • The underlying data model are graphs and not trees as e. • The same specification may be used as the description of a unidirectional as well as a bi-directional transformation process in contrast to [1, 7, 13]. • Furthermore, correspondences are modeled explicitly and are not restricted to the case of 1-to-1 relationships between rather similar data structures as e.g. in [17, 22, 23]. Although being purely declarative, a triple graph grammar may be used as input for the production of a whole family of rather efficiently working tools of varying functionality as • batch-oriented translators that take one data structure as input and return another new related data structure as output, • incrementally working translators that propagate changes of one data structure into its related data structures and work in both directions, • or even consistency observing analyzers that infer or maintain correspondences between related parts of different data structures only. But note that due to lack of space we had to concentrate on batch-oriented transformation tools in section 3 and to omit the construction of incrementally working transformation tools as wells as consistency observing analyzers. For more details on these subjects see [12]. Furthermore , we have to admit that all extensions proposed in section 4 are still on a more or less informal level and that all necessary adjustments of definitions and propositions of section 3 have still to be worked out in detail. Finally, we would like to get rid of the somewhat artificial restriction to monotonic productions by combining the presented approach with a theory of graph parsers for context-sensitive productions, although we are able to specify all kinds graph-to-graph translations, studied up to now, by means of monotonic triple graph grammars only.tween graphs of the same class have to preserve vertex and edge labels, whereas additional application knowledge is necessary to define these relations between different label sets L and L' appropriately. Having a closer look onto the definitions and propositions of section 3, the introduction of labels raises one minor problem: Provided with certain sets of labels for left-and right-hand side graphs LL and LR , how can we construct …