Rapid realization of executable domain models via automatic code generation

The gap between design and implementation always exists because changes happen frequently throughout software development process, along with rapid release cycles, and accompanied by time constraints and limited resources. The focus of our work is to reduce this gap for service-oriented projects. We proposed an approach which considers both technical strategies and agile methods, trying to streamline the progression from design to implementation at a relatively early phase, and then throughout the whole development lifecycle. Automatic code generation has the potential to reduce above problems to a certain extent. This paper describes our efforts to enable rapid and continuous delivery while leveraging parallelism in development via automatic code generation — specifically making domain models instantly executable. We describe a code generator that has been built to enable parallel development of services. It uses UML class diagram to model the problem domain, then rapidly realize the domain model as a set of NoSQL database collections, automate the generation of common database access functions, and automate the wrapping of these database functions within a set of RESTful APIs. We also consider several common deployment scenarios (e.g. requirements for media-handling, security, scalability) to ensure the flexibility and reusability of the target source code for subsequent development iterations. Several empirical project instances have been built using this code generation technique. Combine with agile methods, we attempt to shorten development schedule in both design and implementation stages, and to eliminate the risks caused by evolutionary development. The result shows a great saving of effort on development and less issues in implementation stage.