Weaving generic programming and traversal performance

Abstract

Developing complex software requires that functions be implemented over a variety of recursively defined data structures. While the design (and modeling) of structures is itself difficult, complex data can require even more complex functions. In this paper, we introduce a declarative form of traversal-based generic programming that modularizes functions over a structure using function-objects. Our approach is supported by a library and set of generative tools, collectively called DemeterF, which are used to implement modular, adaptive functions. While our traversals support high-level abstractions resulting in modular and extensible functions, we retain genericity, flexibility and performance through traversal generation and inlining.