The Python interpreter as a framework for integrating scientific computing software-components

The focus of the Molecular Simulation Laboratory is to model molecularinteractions. In particular, we are working on automated docking and molecular visualization. Building and simulating complex molecular systems requires the tight interoperation of a variety of software tools originating from various scientific disciplines and usually developed independently of each other. Over the last ten years we have evolved a strategy for addressing the formidable software engineering problem ofintegrating such heterogeneous software tools. The basic idea is that the Python interpreter serves as the integration framework and provides a powerful and flexible glue for rapidly prototyping applications from reusable software components (i.e. Python packages). We no longer think in terms of programs, but rather in terms of packages which can be loaded dynamically into the interpreter when needed, and instantly extend our framework (i.e. the Python interpreter) with new functionality. We have written more than 30 packages (>2500 classes) providing support for applications ranging from scientific visualization and visual programming to molecular simulations and virtual reality. Moreover, some of our components have been reused successfully by otherlaboratories for their own research. Applications created from our software components have been distributed to over 15000 users around the world. In this paper we describe our approach and its various applications, discuss the reasons that make this approach so successful, and present lessons learns and pitfalls to avoid in order to maximize the reusability and interoperability of software components