Path: utzoo!utgpu!attcan!uunet!lll-winken!lll-tis!helios.ee.lbl.gov!nosc!ucsd!rutgers!uwvax!oddjob!tank!nucsrl!morrison
From: morrison@eecs.nwu.edu (Vance Morrison)
Newsgroups: comp.lang.c++
Subject: Re: Reuseable data structures?
Message-ID: <8180017@eecs.nwu.edu>
Date: 2 Sep 88 18:06:24 GMT
References: <4800034@m.cs.uiuc.edu>
Organization: Northwestern U, Evanston IL, USA
Lines: 41


Hello

In general I have found the following rule to hold

	Reuse of the most basic data structures requires the
	most advanced features of a programming language.

That is reuse of a specialized piece of code is easy, but
reuse of a generalized 'list', 'array', 'set' is more difficult.

In your example of a linked list, yes you can do what you said,
but you will find that most (if not all) of the functions must
be defined in the derived class.  For example

	list.add(elem)
	list.setval(elem)
	list.getval()
	list.delete()

with perhaps the exception of the delete function, all of these
functions need to know how something about the type of 'elem'
because of this they have to be in the derived class.

This problem is characteristic of 'container classes', like
arrays, list, sets, trees (essentially the most important data
types).  

There is a solution to this problem, and it lies in the fact
that most if not all the information the container class needs
to know about its elements is compile time information.   Thus
with a little extra help from the compilier a generic list
can ask for the compilier the infomation it needs.  I am in
the process of developing a simple preprocessor to C++ that will
do this.

At present, however, with standard C++, I do not believe there
is a truely satisfactory solution to the problem of container
classes.

Vance Morrison
Northwestern Univ