6.4.11. Type-level data declarations

TypeData
Since:9.6.1

Allow type data declarations, which define constructors at the type level.

This extension facilitates type-level (compile-time) programming by allowing a type-level counterpart of data declarations, such as this definition of type-level natural numbers:

type data Nat = Zero | Succ Nat

This is similar to the corresponding data declaration, except that the constructors Zero and Succ it introduces belong to the type constructor namespace, so they can be used in types, such as the type of length-indexed vectors:

data Vec :: Type -> Nat -> Type where
  Nil  :: Vec a Zero
  Cons :: a -> Vec a n -> Vec a (Succ n)

TypeData is a more fine-grained alternative to the DataKinds extension, which defines all the constructors in all data declarations as both data constructors and type constructors.

A type data declaration has the same syntax as a data declaration, either an ordinary algebraic data type or a GADT, prefixed with the keyword type, except that it may not contain a datatype context (even with DatatypeContexts), labelled fields, strictness flags, or a deriving clause.

The only constraints permitted in the types of constructors are equality constraints, e.g.:

type data P :: Type -> Type -> Type where
  MkP :: (a ~ Natural, b ~~ Char) => P a b

Because type data declarations introduce type constructors, they do not permit constructors with the same names as types, so the following declaration is invalid:

type data T = T     // Invalid

The compiler will reject this declaration, because the type constructor T is defined twice (as the datatype being defined and as a type constructor).

The main type constructor of a type data declaration can be defined recursively, as in the Nat example above, but its constructors may not be used in types within the same mutually recursive group of declarations, so the following is forbidden:

type data T f = K (f (K Int))  // Invalid