Datatype : hol_type quotation -> unit
STRUCTURE
bossLib
SYNOPSIS
Define a concrete datatype.
DESCRIPTION
Many formalizations require the definition of new types. For example, ML-style datatypes are commonly used to model the abstract syntax of programming languages and the state-space of elaborate transition systems. In HOL, such datatypes (at least, those that are inductive, or, alternatively, have a model in an initial algebra) may be specified using the invocation Datatype `<spec>`, where <spec> should conform to the following grammar: 
   spec    ::= [ <binding> ; ]* <binding>

   binding ::= <ident> = [ <clause> | ]* <clause>
            |  <ident> = <| [ <ident> : <type> ; ]* <ident> : <type> |>

   clause  ::= <ident> <tyspec>*

   tyspec  ::= ( <type> )
            |  <atomic-type>
where <atomic-type> is a single token denoting a type. For example, num, bool and 'a.

When a datatype is successfully defined, a number of standard theorems are automatically proved about the new type: the constructors of the type are proved to be injective and disjoint, induction and case analysis theorems are proved, and each type also has a ‘size’ function defined for it. All these theorems are stored in the current theory and added to a database accessed via the functions in TypeBase.

The notation used to declare datatypes is, unfortunately, not the same as that of ML. If anything, the syntax is rather more like Haskell’s. For example, an ML declaration 

   datatype ('a,'b) btree = Leaf of 'a
                          | Node of ('a,'b) btree * 'b * ('a,'b) btree
would most likely be declared in HOL as 
   Datatype `btree = Leaf 'a
                   | Node btree 'b btree`
Note that any type parameters for the new type are not allowed; they are inferred from the right hand side of the binding. The type variables in the specification become arguments to the new type operator in alphabetic order.

When a record type is defined, the parser is adjusted to allow new syntax (appropriate for records), and a number of useful simplification theorems are also proved. The most useful of the latter are automatically stored in the TypeBase and can be inspected using the simpls_of function. For further details on record types, see the DESCRIPTION.

EXAMPLE
In the following, we shall give an overview of the kinds of types that may be defined by Datatype.

To start, enumerated types can be defined as in the following example: 

   Datatype `enum = A1  | A2  | A3  | A4  | A5
                  | A6  | A7  | A8  | A9  | A10
                  | A11 | A12 | A13 | A14 | A15
                  | A16 | A17 | A18 | A19 | A20
                  | A21 | A22 | A23 | A24 | A25
                  | A26 | A27 | A28 | A29 | A30`
Other non-recursive types may be defined as well: 
   Datatype `foo = N num
                 | B bool
                 | Fn ('a -> 'b)
                 | Pr ('a # 'b`)
Turning to recursive types, we can define a type of binary trees where the leaves are numbers. 
   Datatype `tree = Leaf num | Node tree tree`
We have already seen a type of binary trees having polymorphic values at internal nodes. This time, we will declare it in “paired” format. 
    Datatype `tree = Leaf 'a
                   | Node (tree # 'b # tree)`
This specification seems closer to the declaration that one might make in ML, but is more difficult to deal with in proof than the curried format used above.

The basic syntax of the named lambda calculus is easy to describe: 

    - load "stringTheory";
    > val it = () : unit

    - Datatype `lambda = Var string
                       | Const 'a
                       | Comb lambda lambda
                       | Abs lambda lambda`
The syntax for ‘de Bruijn’ terms is roughly similar: 
   Datatype `dB = Var string
                | Const 'a
                | Bound num
                | Comb dB dB
                | Abs dB`
Arbitrarily branching trees may be defined by allowing a node to hold the list of its subtrees. In such a case, leaf nodes do not need to be explicitly declared. 
   Datatype `ntree = Node of 'a (ntree list)`
A (tupled) type of ‘first order terms’ can be declared as follows: 
   Datatype `term = Var string
                  | Fnapp (string # term list)`
Mutally recursive types may also be defined. The following, extracted by Elsa Gunter from the Definition of Standard ML, captures a subset of Core ML. 
   Datatype
    `atexp = var_exp string
           | let_exp dec exp ;

       exp = aexp    atexp
           | app_exp exp atexp
           | fn_exp  match ;

     match = match  rule
           | matchl rule match ;

      rule = rule pat exp ;

       dec = val_dec   valbind
           | local_dec dec dec
           | seq_dec   dec dec ;

   valbind = bind  pat exp
           | bindl pat exp valbind
           | rec_bind valbind ;

       pat = wild_pat
           | var_pat string`
Simple record types may be introduced using the <| ... |> notation. 
    Datatype `state = <| Reg1 : num; Reg2 : num; Waiting : bool |>`
The use of record types may be recursive. For example, the following declaration could be used to formalize a simple file system. 
   Datatype
     `file = Text string
           | Dir directory
       ;
      directory = <| owner : string ;
                     files : (string # file) list |>`
FAILURE
Now we address some types that cannot be declared with Datatype. In some cases they cannot exist in HOL at all; in others, the type can be built in the HOL logic, but Datatype is not able to make the definition.

First, an empty type is not allowed in HOL, so the following attempt is doomed to fail. 

   Datatype `foo = A foo`
So called ‘nested types’, which are occasionally quite useful, cannot at present be built with Datatype: 
   Datatype `btree = Leaf 'a
                   | Node (('a # 'a) btree)`
Co-algebraic types may not currently be built with Datatype, not even by attempting to encode the remainder of the list as a function: 
   Datatype `lazylist = Nil
                      | Cons ('a # (one -> lazylist))`
Indeed, this specification corresponds to an algebraic type isomorphic to “standard” lists, but Datatype rejects it because it cannot handle recursion to the right of a function arrow. The type of co-algebraic lists can be built in HOL: see llistTheory.

Finally, for cardinality reasons, HOL does not allow the following attempt to model the untyped lambda calculus as a set (note the -> in the clause for the Abs constructor): 

    Datatype `lambda = Var string
                     | Const 'a
                     | Comb lambda lambda
                     | Abs (lambda -> lambda)`
Instead, one would have to build a theory of complete partial orders (or something similar) with which to model the untyped lambda calculus.
COMMENTS
The consequences of an invocation of Datatype are stored in the current theory segment and in TypeBase. The principal consequences of a datatype definition are the primitive recursion and induction theorems. These provide the ability to define simple functions over the type, and an induction principle for the type. For a type named ty, the primitive recursion theorem is stored under ty_Axiom and the induction theorem is put under ty_induction. Other consequences include the distinctness of constructors (ty_distinct), and the injectivity of constructors (ty_11). A ‘degenerate’ version of ty_induction is also stored under ty_nchotomy: it provides for reasoning by cases on the construction of elements of ty. Finally, some special-purpose theorems are stored: ty_case_cong gives a congruence theorem for “case” statements on elements of ty. These case statements are introduced by ty_case_def. Also, a definition of the “size” of the type is added to the current theory, under the name ty_size_def.

For example, invoking 

   Datatype `tree = Leaf num | Node tree tree`;
results in the definitions 
   tree_case_def =
     |- (!a f f1. tree_CASE (Leaf a) f f1 = f a) /\
        !a0 a1 f f1. tree_CASE (Node a0 a1) f f1 = f1 a0 a1

   tree_size_def
     |- (!a. tree_size (Leaf a) = 1 + a) /\
         !a0 a1. tree_size (Node a0 a1) = 1 + (tree_size a0 + tree_size a1)
being added to the current theory. The following theorems about the datatype are also stored in the current theory. 
   tree_Axiom
     |- !f0 f1.
          ?fn. (!a. fn (Leaf a) = f0 a) /\
               !a0 a1. fn (Node a0 a1) = f1 a0 a1 (fn a0) (fn a1)

   tree_induction
     |- !P. (!n. P (Leaf n)) /\
            (!t t0. P t /\ P t0 ==> P (Node t t0))
            ==>
            !t. P t

   tree_nchotomy  |- !t. (?n. t = Leaf n) \/ ?t' t0. t = Node t' t0

   tree_11
     |- (!a a'. (Leaf a = Leaf a') = (a = a')) /\
         !a0 a1 a0' a1'. (Node a0 a1 = Node a0' a1') = (a0=a0') /\ (a1=a1')

   tree_distinct  |- !a1 a0 a. Leaf a <> Node a0 a1

   tree_case_cong
     |- !M M' f f1.
          (M = M') /\
          (!a. (M' = Leaf a) ==> (f a = f' a)) /\
          (!a0 a1. (M' = Node a0 a1) ==> (f1 a0 a1 = f1' a0 a1))
          ==>
          (tree_CASE M f f1 = tree_CASE M' f' f1')
When a type involving records is defined, many more definitions are made and added to the current theory.

A definition of mutually recursives types results in the above theorems and definitions being added for each of the defined types.

SEEALSO
new_type_definition, Define, Hol_reln, TypeBase
HOL  Trindemossen-1