3.14 Summary

Typelists are an important generic programming technique. They add new capabilities for library writers: expressing and manipulating arbitrarily large collections of types, generating data structures and code from such collections, and more.

At compile time, typelists offer most of the primitive functions that lists of values typically implement: add, erase, search, access, replace, erase duplicates, and even partial ordering by inheritance relationship. The code that implements typelist manipulation is confined to a pure functional style because there are no compile-time mutable values—a type or compile-time constant, once defined, cannot be changed. For this reason, most typelist processing relies on recursive templates and pattern matching through partial template specialization.

Typelists are useful when you have to write the same code—either declarative or imperative—for a collection of types. They enable you to abstract and generalize entities that escape all other generic programming techniques. For this reason, typelists are the enabling means for genuinely new idioms and library implementations, as you will see in Chapters 9 and 10.

Loki provides two powerful primitives for automatic generation of class hierarchies from typelists: GenScatterHierarchy and GenLinearHierarchy. They generate two basic class structures: scattered (Figure 3.2) and linear (Figure 3.6). The linear class hierarchy structure is the more size efficient one. A scattered class hierarchy has a useful property: All instantiations of the user-defined template (passed as an argument to GenScatter Hierarchy) are roots of the final class, as Figure 3.2 shows.