caml-list - the Caml user's mailing list
 help / color / mirror / Atom feed
From: Anthony Tavener <anthony.tavener@gmail.com>
To: Gabriel Scherer <gabriel.scherer@gmail.com>
Cc: caml-list@inria.fr
Subject: Re: [Caml-list] Nested module exposing type from parent?
Date: Wed, 2 Nov 2011 19:03:11 -0600	[thread overview]
Message-ID: <CAN=ouMSzM=GUkRDRX8kYVSL4d+Z6iAttegV_=hL7_KVm9G6+RQ@mail.gmail.com> (raw)
In-Reply-To: <CAPFanBEghH7ebE3gmbsv1SESv6Ur60G5deN8RAG_ryWSUNao1w@mail.gmail.com>

[-- Attachment #1: Type: text/plain, Size: 5996 bytes --]

I had tried this before, because the first thing I did was run into the
cyclic type due to types being implicitly recursive. And I found that
Janestreet article... I'm sure I felt I had the answer then. But it didn't
work for my needs, because I wanted access to the record fields when
opening the sub-module. This is where Vincent's suggestion of using
'include' made the difference. Thank-you though!


On Wed, Nov 2, 2011 at 5:01 PM, Gabriel Scherer
<gabriel.scherer@gmail.com>wrote:

> I see that you solved your problem in a way you find satisfying, but I
> would like to point out that the reason why your original code didn't
> work isn't exactly what you seem to think.
>
> When you define a submodule, the types defined before in the parent
> modules are perfectly accessible and can be referred, just as you
> would do when referring to types defined at the toplevel. You need not
> qualify the type with the outer module name (Vec.t in your example),
> as you are still *inside* this parent module.
>
>  module Vec = struct
>    type t = int
>    module Type = struct
>      type u = t
>    end
>  end
>
>  (1 : Vec.Type.u);;
>
> The problem in your case is that you wish to give the same name to the
> type in Vec and in Vec.Type. This would lead to the following:
>  ... module Type = struct type t = t end ...
>
> But this is ill-defined : it is a recursive type defined as being
> itself. The problem is that the OCaml syntax for type declarations
> always consider them recursive (for values you have "let" and "let
> rec", for types you have "type" which behaves like "type rec" with no
> opt-out way possible). This is a flaw of the OCaml syntax which is
> relatively well-known, see eg. http://ocaml.janestreet.com/?q=node/25
>
> A workaround is to define your inner type "t" in two steps, using an
> different intermediate name to break the cycle:
>
>  module Vec = struct
>    type t = int
>    module Type = struct
>      type u = t
>      type t = u
>    end
>  end
>
>  (1 : Vec.Type.t);;
>
>
> On Wed, Nov 2, 2011 at 10:14 PM, Anthony Tavener
> <anthony.tavener@gmail.com> wrote:
> > Oops, I didn't do a group-reply... so in case anyone is interested in
> what I
> > ended up with:
> >
> > ---------- Forwarded message ----------
> > From: Anthony Tavener <anthony.tavener@gmail.com>
> > Date: Wed, Nov 2, 2011 at 2:50 PM
> > Subject: Re: [Caml-list] Nested module exposing type from parent?
> > To: Vincent Aravantinos <vincent.aravantinos@gmail.com>
> >
> >
> > Actually, better than I initially thought...
> > I keep this as I have them defined already, except as you said: include
> > instead of open.
> >   module Vec = struct
> >     module Type = struct
> >       type t = { x: int; y: int }
> >     end
> >     include Type
> >     let make x y = {x;y}
> >     let add a b = {x=a.x+b.x; y=a.y+b.y}
> >   end
> > Before, I had instead of the include:
> >   type t = Type.t
> >   open Type
> > Which worked, but then the type used everywhere was Vec.Type.t
> > Thanks again! Simple and effective, and I was looking in all the wrong
> > places. :)
> > On Wed, Nov 2, 2011 at 2:36 PM, Anthony Tavener <
> anthony.tavener@gmail.com>
> > wrote:
> >>
> >> Thank-you Vincent!
> >> Though this requires a home for the "source type" module, at least the
> >> types come out right in the end. Thanks!
> >> And this led me to read specifically about include to understand what it
> >> really does. :)
> >>
> >> On Wed, Nov 2, 2011 at 2:19 PM, Vincent Aravantinos
> >> <vincent.aravantinos@gmail.com> wrote:
> >>>
> >>> Using "include" instead of "open" would work, ie. turning your example
> >>> into:
> >>>
> >>> module Vec_main = struct
> >>>   type t = { x: int; y: int }
> >>>   let make x y = {x;y}
> >>>   let add a b = {x=a.x+b.x; y=a.y+b.y}
> >>> end
> >>>
> >>> module Vec = struct
> >>>   include Vec_main
> >>>   module Type = struct
> >>>     include Vec_main
> >>>     ...
> >>>   end
> >>> end
> >>>
> >>> Then:
> >>> # let n = Vec.make 2 5;;
> >>> val n : Vec.t = {Vec.x = 2; Vec.y = 5}
> >>> # open Vec.Type;;
> >>> # let m = {x=1;y=2};;
> >>> val m : Vec.Type.t = {x = 1; y = 2}
> >>> # Vec.add m n;;
> >>> - : Vec.t = {Vec.x = 3; Vec.y = 7}
> >>>
> >>> Cheers
> >>>
> >>> --
> >>> Vincent Aravantinos - Postdoctoral Fellow, Concordia University,
> Hardware
> >>> Verification Group
> >>>
> >>> On 11/02/2011 03:41 PM, Anthony Tavener wrote:
> >>>
> >>> I've been struggling with this occasionally...
> >>> I'm using nested modules to "open" access to select features of a
> module.
> >>> My problem is I can't find a way to *expose* types in the parent module
> >>> through such nested modules.
> >>> A simplified example of what I'm looking at:
> >>>   module Vec = struct
> >>>     type t = { x: int; y: int }
> >>>     let make x y = {x;y}
> >>>     let add a b = {x=a.x+b.x; y=a.y+b.y}
> >>>     module Type =
> >>>       (* something which has type t = Vec.t,
> >>>        * with exposed structure when "open"ed.
> >>>        * Also note that Vec is not really an
> >>>        * explicit module like this; instead it
> >>>        * is implemented in vec.ml *)
> >>>   end
> >>> Example usage...
> >>>   let n = Vec.make 2 5
> >>>   open Vec.Type
> >>>   let m = {x=1;y=2}
> >>>   Vec.add m n
> >>>
> >>> To date, I've defined the type in the Type submodule, which is then
> used
> >>> by the parent module. The unsatisfactory quality of this is that
> Vec.Type.t
> >>> is the "true" type. Ideally the concrete type would live at Vec.t, with
> >>> "open Vec.Type" bringing the fields of the type into scope.
> >>> As background, here are examples of opening different features of the
> Vec
> >>> module:
> >>>   let c = Vec.add a b
> >>>   open Vec.Prefixed
> >>>   let c = vadd a b
> >>>   open Vec.Ops
> >>>   let c = a +| b
> >>>   open Vec.Type
> >>>   let c = Vec.add a {x;y;z=0.}
> >>> Apologies if this is really beginner-list material. It's minor, but has
> >>> been bugging me.
> >>> Thank-you for looking,
> >>>  Tony
> >>>
> >>
> >
> >
> >
>

[-- Attachment #2: Type: text/html, Size: 8194 bytes --]

  parent reply	other threads:[~2011-11-03  1:03 UTC|newest]

Thread overview: 8+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2011-11-02 19:41 Anthony Tavener
2011-11-02 20:19 ` Vincent Aravantinos
     [not found]   ` <CAN=ouMTApZjpU-CaZtdL4njXtmtRu++7fzJBJL3w3FRcHfjtSA@mail.gmail.com>
     [not found]     ` <CAN=ouMS9rqqNR3KgCBnnjC_HcMrUnftVw643mkmhC_vrpXfv1A@mail.gmail.com>
2011-11-02 21:14       ` Fwd: " Anthony Tavener
2011-11-02 23:01         ` Gabriel Scherer
2011-11-03  0:06           ` Vincent Aravantinos
2011-11-03  1:03           ` Anthony Tavener [this message]
2011-11-03  0:41 ` Martin Jambon
2011-11-03  1:04   ` Anthony Tavener

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to='CAN=ouMSzM=GUkRDRX8kYVSL4d+Z6iAttegV_=hL7_KVm9G6+RQ@mail.gmail.com' \
    --to=anthony.tavener@gmail.com \
    --cc=caml-list@inria.fr \
    --cc=gabriel.scherer@gmail.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).