NetNews Usenet Archive 1992 #16/NN_1992_16.iso/spool/comp/lang/prolog/1437

Internet Message Format | 1992-07-22 | 2.5 KB

Path: sparky!uunet!mcsun!uknet!qmw-dcs!dc
From: [e-mail] (Daniel Cohen)
Newsgroups: comp.lang.prolog
Subject: Re: Prolog-WIMP interfacing principles
Message-ID: <[e-mail]>
Date: 22 Jul 92 11:28:00 GMT
References: <[e-mail]>
Sender: [e-mail] (Usenet News System)
Organization: Computer Science Dept, QMW, University of London
Lines: 38
Nntp-Posting-Host: guinness.dcs.qmw.ac.uk

In <[e-mail]> [e-mail] (Paul Singleton) writes:

>Please can anyone with experience or insight suggest some principles
>for interfacing Prolog programs to WIMP interfaces, i.e. buttons, sliders,
>menus, dialog boxes and all that stuff.  I avoid asserting and retracting
>global state, and try to contain side-effects, and I can't see how to
>employ WIMP features without compromising this style.  Well, I can, but ...


How about using streams? The connection to the server is a (pair of)
stream(s) down which the program sends commands like
new_window(.....,S) where S will be used as a stream to send commands
to the new window created. In particular, dialog boxes might be built
by creating a window as above and then sending commands like
new_slider(.....,SS) down the stream to the window (S). Each input
object will have one of its arguments a stream with which it can send
messages back to the Prolog program. Of course, you'll need some sort
of parallelism (would co-routining be enough?) or else have to write
some horrible polling system to deal with the messages sent back from
input devices.

You'll have guessed I'm used to using a concurrent logic language - the
Strand X-Windows interface works this way. The nice thing about such an
interface is that the only implicit state information and side-effects
are in the X server - the program itself remains clean, with all access
to the WIMP system working via the streams. A procedure cannot access
the WIMP system without being given access to one of the WIMP streams
and no changes need be made to the Prolog database.

I have a Strand-Ingres interface that works in the same way - all state
is hidden from client programs. Although the interface doesn't fit
relation databases into the logic programming model as neatly as some
Prolog database interfaces do, it does seem to be a lot cleaner.

Daniel Cohen              Department of Computer Science
Email: [e-mail]   Queen Mary and Westfield College
Tel: [phone]  Mile End Road, London E1 4NS, UK
Fax: [phone]      ******* Hit the North!!! *******