[Haskell-cafe] Fwd: How to combine simulations

[David Sorokin]

Hi all,

I forgot to click on ReplyAll. I thought that my answer could be interested
to others as well. Therefore, I resend the letter to Haskell Cafe.

Thanks,
David


---------- Forwarded message ----------
From: David Sorokin <david.sorokin at gmail.com>
Date: Wed, Sep 2, 2015 at 9:39 AM
Subject: Re: [Haskell-cafe] How to combine simulations
To: martin <martin.drautzburg at web.de>


Hi Martin,

Nevertheless, regarding Aivika, did you look at the Signal type (inspired
by the IObservable interface of .NET)? This type is quite composable. If I
understood you correctly, the Signal is what you call Event.

Moreover, if you need a mutable state, then you could use the Ref reference
and its signal refChanged. For example, the reference could be used to
decide whether to emit a signal based on the intermediate state.

Furthermore, if you want to build a computation with linear-looking signal
processing then you could use the Process monad (inspired by the Async
workflow of F#) and the processAwait function that suspends the
current discontinuous process waiting for a signal to come.

I hope it helps. In case of need, these ideas can be re-implemented.

Thanks,
David

On Mon, Aug 31, 2015 at 7:48 PM, martin <martin.drautzburg at web.de> wrote:

> Hello all,
>
> I've been trying hard to come up with an idea how to build a DES from
> smaller parts. So far, I came to the conclusion,
> that somewhere there must be an operation which takes an Event and maybe
> emits an Event (and appends to a log and
> updates some state). Those Events whould come from and go to the
> "environment" the simulation runs in.
>
> My mental model is two billiard tables, which are connected through a hole
> in the cushion and which each have a player.
> When I look at one such table, it would have to respond to Events from its
> player and from the other table and it would
> send events to its player ("all balls at rest") and to the other table.
>
> If I add the other table and the two players then the combined simulation
> would not emit any events at all and it would
> not respond to any events except maybe as START event. It would only
> depend on its initial state.
>
> But if I add only the player, but not the other table, it would still send
> events to the other table and respond to
> events from that other table.
>
> My problem is the type of Events. I could create a type which encompasses
> all possible events, but that violates the
> idea of composablitly. Somehow I would need to be able to take a system
> which accepts "player events" and "other table
> events", compose it with an other table and end up with a system which
> only accepts "player events" but no more "other
> table events" and similarly for the emitted events. And I don't quite know
> how to do this.
>
> Hope this makes some sense.
>
> Any pointers (which go beyond "aivika has a simulation component") would
> also be much appreciated.
> _______________________________________________
> Haskell-Cafe mailing list
> Haskell-Cafe at haskell.org
> http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-cafe
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.haskell.org/pipermail/haskell-cafe/attachments/20150902/fca1e8e8/attachment.html>