Path: utzoo!utgpu!water!watmath!clyde!att!osu-cis!killer!convex!texsun!pitstop!sundc!seismo!uunet!munnari!mulga!ttey
From: ttey@mulga.oz (YEO Tee Thiam Eric)
Newsgroups: comp.lang.c++
Subject: Re: C++ considered disenchanting
Message-ID: <3006@mulga.oz>
Date: 18 Oct 88 11:58:37 GMT
References: <2010@cs.Buffalo.EDU>
Organization: Comp Sci, Melbourne Uni, Australia
Lines: 45

in article <2010@cs.Buffalo.EDU>, crovella@sunybcs.uucp (Mark Crovella) says:
> These are minor points, but the overall result seems to be a perception
> that the language is a bit "hairy".  

I agree. I strongly believe that a language should be as simple as
possible for 2 reasons: (a) a compiler/preprocessor can easily be
constructed and possibly proven correct, (b) it is easier to formalise
the semantics of the language and remove ambiguity. In the case of
C++, the syntax of the language is, in my opinion, too complicated.
It is so complex that it has been said that to parse C++ correctly,
a recursive decent parser with an infinite look ahead and some
heuristics is required (I believed it was mentioned in some
documentation for GNU G++).

> As for Dave's conclusion (i.e., a disassociation between himself and C++),
> I think that in large part this may not be the fault of C++, but rather
> in the toolset we use and in the orientation we have to programming in C++.
> The problems we have had using C++ would be well solved through the use
> of a more comprehensive programming environment for the language.
> 
> I propose a discussion of the necessary, appropriate, and/or ideal
> tools and techniques for C++ programming.
> For example, a tool to resolve virtual function scoping dynamically.
> A "Flavor Inspector" -- a class examination tool.  An object inspector.
> A dynamic debugger which runs C++ interpretively.  etc.
> 
> The questions Dave raises are important -- I think it would be interesting 
> to hear people's views on them.

A good programming environment similar to that of Smalltalk would
certainly be nice. (Much feasibility studies need to be done on this
since C++ is a static language compared to Smalltalk.) However, I
generally found that when programming in C++, people tend to be
carried away with object-orientedness of the language; using derived
classes, etc, etc at every opportunity to model things that are better
left as structures.

Object-oriented programming is a paradigm. We are now finding more
and more ways of using this paradigm to model our problems.
C++ helps in acheiving this but it is by no means the best and
only way of programming. I believe that the essence of object-
oriented programming is in choosing what to be modelled as objects
and what to be treated as pure data structure.

						Eric Yeo