Synchronization of Views and Loose Semantics of Typed Graph Productions

The concept of views is used on two levels. First, so-called design views are developed for structuring specifications, that is, a system is modeled according to different views (e.g., representing the needs of different kinds of users) which have to be synchronized afterwards in order to build the whole system. Views can be specified by means of typed graph transformation systems, where the type graph determines the visible types and the productions describe the known operations of that view. The synchronization of views is done by the construction of cooperative parallel composition of graph transformation systems, developed by Leila Ribeiro and presented at the same seminar. If the specification is complete, a view may describe an observation of the system in operation. In this case we speak of a user view. It turns out that the semantics of such a view cannot be described by computations (i.e., graph transformations), but just by observations of computations of the global system. Such observations of computations cannot be represented by graph transformations in the usual sense because a local view may lack operations (productions) of the global system, so that state changes may be observed that do not have a cause in the local view. Therefore, the notion of graph transition is introduced as loose semantics for productions, where the production specifies only a lower bound to the activities that are to happen during application. Contrastingly, in the classical doublepushout approach to graph rewriting, productions are interpreted as complete descriptions of the transformations to be performed. For typed graph transformation systems a transition sequence semantics is developed, comprising all finite and infinite sequences of transitions in a system. Moreover, this semantics is shown to be compositional w.r.t. the synchronization of views.