In article <rfgCxuvHn.CrJ@netcom.com> rfg@netcom.com (Ronald F. Guilmette) writes:

...

>I'd prefer to be able to solve this problem _without_ having to invent
>new (and gratuitous) types.
>
>(I could say the same thing about a _lot_ of problems in C++.  Sadly, it
>seems that the ``C++ solution'' to an awful lot of relatively simple
>problems is to have the programmer create a new type. :-(

Now, this is interesting. I have the same feeling, and I don't create new
classes unless I really have to. But maybe this is the wrong view when
using C++ ? Your use of the word "gratuitous" is interestingly ambigous here.
My dictionary tells me it means either "free" (sort of) or "without enough
reason".

I have especially felt this way when creating GUI applications. Hesitating
to make a new kind of button, to be derived from the standard button, etc. I
usually prefer to have a button which is configurable by setting attributes.

There are at least three reasons I hesitate to create new classes:

* It is a "lot of" initial work. You need an extra file, adding "#ifdef",
  worrying about copy-constructors, asignment operators etc.

* I worry about performance. The program grows bigger, possibly slower.

* More classes and more files makes the program harder to maintain.

The first item can be somewhat overcome by some kind of tool. Several exists.
When I think about it, I realize I can easily make some kind of macro to
emacs.

The second is probably just wrong. I can always use inline if it really
matters. Anyway, I should not worry about performance at this stage.

The last item is maybe the most important ? However, with namespaces, it is
easier to structure the source code. I also find it difficult to decide
when to remove a class.


So why not look at the other side of the coin. C++ is very flexible, because
you can so easily make your own class...
------
Lars Pensjo