Type systems evolve too slowly to keep up with the quick evolution of libraries—especially those that introduce new abstractions. Type tailoring offers a lightweight solution: equip the core language with an API for modifying the elaboration of surface code into the internal language of the typechecker. Tailoring rules can improve both the precision and the expressiveness of type analysis by fine-tuning the elaboration step. Furthermore, these rules cooperate with the host type system by expanding to typechecked code. In the context of a hygienic metaprogramming system, tailoring rules can even compose with one another to create a seamless interface for downstream clients.

Type tailoring has emerged as a theme across several languages and metaprogramming systems, but never with direct support and rarely in the same shape twice. Both OCaml and Racket enable forms of tailoring, for example, but in different ways. This paper identifies key dimensions of type tailoring systems and illustrates the tradeoffs along each dimension. It demonstrates the usefulness of tailoring with examples that cover sized vectors, database queries, and optional types. Finally, it outlines a vision for future research in type metaprogramming.