Abstract

Support for generic programming consists of three essential ingredients: support for overloaded functions, a run-time type representation, and a generic view on data. Different approaches to datatype-generic programming occupy different points in this design space. In this article, we revisit the “Scrap your boilerplate” approach and identify its location within the three-dimensional design space. The characteristic features of “Scrap your boilerplate” are its two generic views, the ‘spine’ view for consuming and transforming data, and the ‘type-spine’ view for producing data. We show how to combine these views with different overloading mechanisms and type representations.