Component Model Interoperability for Scientific Computing

The ability to create connections between previously incompatible software is the motivation for the work of this dissertation. We focus on devising a paradigm for interoperability between software components, a popular and often used software methodology, while making the practical concerns of this endeavor our main driving force. Component software interoperability is the ability for two or more existing software components to cooperate despite differences in component model, interface or execution platform. It is also the ability to connect and use existing servers though they may be plug-incompatible to the clients. The capability of combining any two components regardless of implementation platform creates a new realm of possibility in using common off-the-shelf components. Unfortunately, solving the general problem of component interoperability using a fully automatic mechanism is daunting. In this work, we describe and analyze a set of techniques and tools that we developed to make component interoperability attainable in a practical sense. Our approach simplifies a difficult but necessary goal of component interoperability and our user-driven generative tools provide several abstractions enabling flexible and structured component integration. By leveraging a generative programming approach, we develop a programmable code generator that interposes a bridge to translate between components of disparate component models. We extend this bridge generation with tools aimed at different aspects of the problem to create a reliable and flexible method of performing this task. Focus is given to mapping of incompatible interfaces and data types and to manipulating data between two heterogeneous components. In order to include semantic understanding of component models into our design, we define a component meta-modeling method and a bridging language. The bridging language leverages component meta-models, which semantically model the individual features of each component model, to produce a translation between two such models. We combine these methods into our stack of tools that aim to bridge difference between heterogeneous component instances and provide component interoperability.