A library writer's guide to shortcut fusion

Abstract

There are now a variety of shortcut fusion techniques in the wild for removing intermediate data structures in Haskell. They are often presented, however, specialised to a specific data structure and interface. This can make it difficult to transfer these techniques to other settings. In this paper, we give a roadmap for a library writer who would like to implement fusion for his own library. We explain shortcut fusion without reference to any specific implementation by treating it as an instance of data refinement. We also provide an example application of our framework using the features available in the Glasgow Haskell Compiler.