Bridging Technical Spaces: Model Translation from TA to XMI and Back Again

Abstract

There are many different techniques and notations for extracting architecturally interesting information from the source code of existing software systems. This process is known as reverse engineering. One current problem with reverse engineering techniques is that models of software systems cannot easily be transferred from one notation and storage format to another. We refer to this as the problem of bridging <em>technical spaces</em>. <br /><br /> In this work, we approach the issue of bridging between the SWAG technical space and the UML technical space. The SWAG technical space, named after the Software Architecture Group at the University of Waterloo, consists of fact extractors, fact manipulators, schemas, and a fact storage language - the Tuple-Attribute language (TA). The UML technical space consists of the UML metamodel, the XML Metadata Interchange (XMI) format for encoding UML models, and various UML modeling tools. We have designed and implemented a plugin for MagicDraw UML, which will import, export, and merge between XMI-encoded UML models and TA-encoded Function-Level Schema models. <br /><br /> We document evidence of what is referred to as a <em>Bridge Domain</em> - a technical space which exists between two encodable spaces. The metamodels of the two notation languages that we have focused on are very rich and flexible, but neither technical space is capable of fully expressing an accurate architectural model of any given software system; however, each technical space is capable of maintaining certain semantic information relevant to that technical space through multiple merge operations.