Re: [Caml-list] type_declaration in env.ml

[Xavier Leroy <Xavier.Leroy@inria.fr>]

Pietro Abate wrote:
> hi all,
> I'm trying to understand this data structure in env.ml in the ocaml src
> files.
>
> type type_declaration =
>   { type_params: type_expr list;
>     type_arity: int;
>     type_kind: type_kind;
>     type_manifest: type_expr option;
>     type_variance: (bool * bool * bool) list }
>
> and type_kind =
>     Type_abstract
>   | Type_variant of (string * type_expr list) list * private_flag
>   | Type_record of (string * mutable_flag * type_expr) list
>                  * record_representation * private_flag
>
>
> If I've a type  like :
> type t = A of int | B of float
>
> I'll get
>
> type_params = []
> type_arity = 0
> type_kind = Type_variant (
>                 [ "A", [ { desc = Tvar ; level = ?? ; id = ?? }] ;
>                   "B", [ { desc = Tvar ; level = ?? ; id = ?? }] ] , false )
> type_manifest = ??
> type_variance = []
>
> correct ?

Yes, with type_manifest = None.

> what if I have just
> type t = int ?

As above, except
 type_kind = Type_abstract
 type_manifest = Some <the representation of type "int">

And before you ask,
  type t = M.t = A of int | B of float
would be
  type_kind = Type_variant ( ["A", ... ; "B", ....], false)
  type_manifest = Some <the representation of type "M.t">

> where I store the type identifier "t" in both cases ?

The typing environment (type Env.t) maps type identifiers such as "t"
to type declarations such as the above.

> Can anybody explain me the semantic of these fields ? I can guess some,
> but I'm not very confident...

Section 6.8.1 of the reference manual could help you understand the
interaction between the "type_kind" and "type_manifest" fields, and in
particular the reason why these are two separate fields.

Hope this helps,

- Xavier Leroy