Dimensions of reengineering environment infrastructures

ion power of a logic programming language to do grouping. Hence it has support for intentional and extensional grouping, which is used to express design patterns and software architectures. Because of its integration with Smalltalk, the model is extensible. When the model needs to encompass new kinds of entities, only some reification rules have to be added to make those entities explicit in SOUL. It supports incremental loading by definition: during the logic inference process those parts of the system that are needed are reified on the fly. Extra model information can be added easily in the form of extra logic facts and rules. Referencing of elements in the model is done as in any other logic programming language, i.e., by the use of logic variables that reference Smalltalk entities. Finally, SOUL supports code refactoring. FAMIX and Moose. FAMIX is the meta-model we designed for modeling software [15] and theMoose Reengineering Environment is the platform we built with a repository based on this meta-model [20]. Moose supports reengineering of medium to large-sized industrial software systems in multiple objectoriented implementation languages. Its meta-model, FAMIX, defines a minimal language-independent core of information that is always available and on top of which additional levels of information can be added [43–46]. The supported languages are modeled at the program entity level. The meta-model defines a minimal core based on object-oriented concepts and can be extended to support new entities found in different languages or needed when supporting higher views of the applications. Model elements can be either referenced by unique name for the elements that have one (the entities), or by unique identifier which all elements have (both entities and relations). FAMIX illustrates the strong impact of choices for the referencing schema. The unique referencing schema supports grouping even over multiple models. The meta-model does not support grouping explicitly, but the environment requires that all entities have a unique name. Moose supports incremental loading. It uses standard exchange formats such as CDIF and XMI, but with the chunk referencing schema. Apart from the textual storage with CDIF and XMI, Moose can store its in-memory repository. The FAMIX metamodel does not support multiple models, but Moose does. The repository can hold multiple models, but textual information transfers (defined by the meta-model) do not contain information aboutmultiple models. Moose defines its own simple ER-like meta-meta model. This meta-meta model is used to generate meta-models and support generic tool facilities. FAMIX originally represented model information as part of its meta-model. Now we treat the meta-model information as an entity, because it avoids having a meta-model-specific section while loading models. However, the model and information about the model are mixed, which is conceptually a questionable design choice. Hence tools have to be aware that one entity representing the model information exists. Furthermore, entities can be annotated. Finally Moose provides experimental support for higher level refactorings based on program entity level information rather than the more complete AST level of information [25].