Bidirectional by necessity: Data persistence and adaptability for evolving application development

Abstract

Database-backed applications are ubiquitous. They have common requirements for data access, including a bidirectional requirement that the application and database must have schemas and instances that are synchronized with respect to the mapping between them. That synchronization must hold under both data updates (when an application is used) and schema evolution (when an application is versioned). The application developer treats the collection of structures and constraints on application data - collectively called a virtual database - as indistinguishable from a persistent database. To have such indistinguishability, that virtual database must be mapped to a persistent database by some means. Most application developers resort to constructing such a mapping from custombuilt middleware because available solutions are unable to embody all of the necessary capabilities. This paper returns to first principles of database application development and virtual databases. It introduces a tool called a channel, comprised of incremental atomic transformations with known and provable bidirectional properties, that supports the implementation of virtual databases. It uses channels to illustrate how to provide a singular mapping solution that meets all of the outlined requirements for an example application. © Springer-Verlag Berlin Heidelberg 2013.