Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!uunet!mcsun!ukc!mucs!logitek!hrc63!mrcu!paj
From: paj@mrcu (Paul Johnson)
Newsgroups: comp.lang.eiffel
Subject: Re: a problem in programming with Eiffel
Message-ID: <798@puck.mrcu>
Date: 23 Jan 91 14:54:40 GMT
References: <20022@hydra.gatech.EDU>
Reply-To: paj@uk.co.gec-mrc (Paul Johnson)
Organization: GEC-Marconi Research Centre, Great Baddow, UK
Lines: 47
Summary:
Expires:
Sender:
Followup-To:


>I have a curious problem.  I am writing an application in Eiffel that
>makes use of the graphics library. In particular, I need to use
>objects that are similar to the class COMPLEX_FIG in the library, yet
>a little different from it. My class is similar in the sense that it
>is to have a list of figures and the whole list has to behave like a
>figure. But it is different because the figures in the list are
>really going to be a special type of figure that is a descendant of
>class FIGURE.

>At first I thought that I could make my class a descendent of the class
>COMPLEX_FIG and proceed from there. This is the most natural solution.
>However, it turns out that COMPLEX_FIG inherits from FIG and LINKED_LIST,
>and the generic parameter of the linked list is FIGURE. This causes a
>problem because I need the generic parameter to be something that is 
>capable of responding to feature calls of my special figure. This 
>seemingly cannot be done because the generic parameter is hardwired,
>so to speak, to the class FIGURE.

>My question is: short of duplicating the code for COMPLEX_FIG with the
>generic parameter of the inherited linked list now as my special figure,
>is there any other better way to solve the problem?

>I suspect that the problem is not just a matter of good library design,
>but also a problem of language design. Could someone enlighten me?

I will try, although I have not checked the following against the
Eiffel type rules.  Hence it may not work.

When you produce your child of COMPLEX_FIG (say, MY_COMPLEX_FIG,
although this would be a bad name in practice) you should say
`redefine first; rename first as old_first' and then define `first' as
being of the appropriate type and returning (using the reverse
assignment operator `?=') the value of `old_first'.  All the other
features of LINKED_LIST are `like first', so this will make the change
to all of them.

This is the first time I have really appreciated the utility of `like'.

If this does not work then use the reverse assignment operator outside
of COMPLEX_FIG to undo the type loss before applying your features.

Paul.
-- 
Paul Johnson                               UUCP: <world>!mcvax!ukc!gec-mrc!paj
--------------------------------!-------------------------|-------------------
GEC-Marconi Research is not 	| Telex: 995016 GECRES G  | Tel: +44 245 73331
responsible for my opinions.	| Inet: paj@uk.co.gec-mrc | Fax: +44 245 75244