Mailing list for all users of the OCaml language and system.
 help / color / mirror / Atom feed
From: "Francisco Valverde Albacete" <fva@tsc.uc3m.es>
Cc: caml-list@inria.fr
Subject: Re: mutually recursive types and modules
Date: Fri, 14 May 1999 13:11:19 +0200	[thread overview]
Message-ID: <373C04D6.DCF0F90E@tsc.uc3m.es> (raw)
In-Reply-To: <199905111632.RAA19705@byrd.sle.sharp.co.uk>


Andrew Kay wrote:

> Dear OCaml users,
>
> We have a problem with mutually recursive types.

> In our project we use a graph representation which could be reduced to:
>
> # type node = {
> #             node_id : int;
> #             mutable edges : node list;
> #             ... (other fields)
> #             }
>
> Node_ids are generated at node creation time, and are unique, so we defined:
>
> # type compare_nodes n1 n2 = Pervasive.compare n1.node_id n2.node_id
>
> Then we built sets of nodes:
>
> # module NodeSet = Set.Make(struct type t=node let compare=compare_nodes end)
>
> So far so good.  Next we realised that we don't care about the order of
> edges in the edge list, and we are always converting edge lists into sets
> to do union operations and so on, so we decided to recode the node type
> with edges as sets for efficiency (which is very important here).
>
> * type node = {
> *             node_id : int;
> *             mutable edges : NodeSet.t;
> *             ... (other fields)
> *             }
>
> At this point the world seemed to spin and make me dizzy, because we can't
> defined NodeSet without node, and we can't define node without NodeSet.
> I can't see any way to express this in OCaml.
>
> Am I missing something obvious here?  How should I proceed?  Is this kind
> of recursion unreasonable, or just syntactically hard to express in OCaml?
> Or am I merely incapable of understanding the manual?

I have come across this problem as well in trying to implement heuristic
searches...
As recursive modules are, for the time being, dangerous to type and thus
disallowed, you'll have to make do with:

a) defining both structures at the same time, with the obvious work of
recoding
sets, etc.

b) pass a *polymorphic* set structure to your nodes:

module Node
    ( Set : (* set signature *) )
    =
    struct
        type 'a set = 'a Set.t (* whatever its name in the Set module *)
        type node = {
             node_id : int;
             mutable edges : node set; (* <- Here's the trick *)
             ... (other fields)
             }
    ....

    end

But BEWARE, this is going to cost you almost always an extra parameter to any
constructor in the Set module, a "compare" function to guarantee a linear
order
that makes the Set implementation efficient, so balance very well this with
the
previous alternative...

Hope this helps,

        Fran

PS: I have a polymorphic n-ary tree using the same structure you are trying to
build here in case you want to have a look at it. But I warn you that it is
also
quite gory. F.






      parent reply	other threads:[~1999-05-14 16:22 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
1999-05-11 16:32 Andrew Kay
1999-05-12 16:55 ` Benoit deBoursetty
1999-05-12 18:04 ` Markus Mottl
1999-05-14 11:11 ` Francisco Valverde Albacete [this message]

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=373C04D6.DCF0F90E@tsc.uc3m.es \
    --to=fva@tsc.uc3m.es \
    --cc=caml-list@inria.fr \
    /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