Instead of having Client and Server depend on Message, consider having
Check_message_types depend on both Client and Server, and in
check_message_types.ml you have:

let check_message_types_are_equal (x : Client.t) : Server.t = x

On 24 June 2015 at 01:25, Kenichi Asai <asai@is.ocha.ac.jp> wrote:

> Thanks for all the answers!  Let me ask one more (somewhat related)
> question.  I want to share a type of messages between a client and a
> server (in communicating games students write to avoid segmentation
> fault caused by marshaling or a runtime error caused by bin-prot).
> One natural way to do it is to have a file message.ml:
>
> message.ml:
> -----
> type t = int * int (* example type of messages *)
> -----
>
> and use it in both the client and the server:
>
> client.ml:
> -----
> let send (message : Message.t) =
>   ... marshal the message and send to the server ...
> -----
>
> server.ml:
> -----
> let receive () : Message.t =
>   ... unmarshal the message received from the client ...
> -----
>
> In this case, however, students always have to write message.ml and
> declare Message.t even if it can be automatically inferred from the
> code of client.ml and server.ml.  Ideally, I want to provide
> message.ml something like:
>
> message.ml:
> -----
> type t = '_a
> -----
>
> which signifies that the type Message.t can be anything as long as it
> is instantiated to exactly one type throughout the program (i.e., both
> in the client and the server).  Students could then write only
> client.ml and server.ml.  If they instantiate Message.t consistently,
> the program compiles fine; otherwise, compilation leads to a type error.
>
> Because the above definition of message.ml is not allowed, I defined:
>
> message.ml:
> -----
> type t = int (* dummy type *)
> -----
>
> and tried to override it:
>
> client.ml:
> -----
> include Message
> type t = int * int
> -----
>
> but it doesn't work.  Would there be any way to impose a constraint
> that two types in two different modules to be the same without
> specifying the type itself?
>
> Sincerely,
>
> --
> Kenichi Asai
>
> --
> Caml-list mailing list.  Subscription management and archives:
> https://sympa.inria.fr/sympa/arc/caml-list
> Beginner's list: http://groups.yahoo.com/group/ocaml_beginners
> Bug reports: http://caml.inria.fr/bin/caml-bugs
>