NetNews Offline 2/NetNews Offline Volume 2.iso/news/comp/lang/c++-part1/3261

Text File | 1996-08-06 | 3.4 KB | 72 lines

Newsgroups: comp.lang.ada,comp.lang.c++,comp.lang.c,comp.lang.modula3,comp.lang.modula2,comp.lang.eiffel
Path: shrike.irvine.com!adam
From: [e-mail] (Adam Beneschan)
Subject: Re: Avoiding typedefs (was: Hungarian notation)
In-Reply-To: Douglas Evan Cook's message of Mon, 22 Jan 1996 14:58:59 -0700
References: <[e-mail]> <4bd <4cc2b2$[e-mail]>
 <4cud8f$[e-mail]> <[e-mail]> <[e-mail]>
Sender: [e-mail] (News Administration)
Organization: Irvine Compiler Corp., Irvine, California, USA
Date: Tue, 23 Jan 1996 02:16:36 GMT
Message-ID: <[e-mail]>

Douglas Evan Cook <[e-mail]> writes:

> On 21 Jan 1996, Kazimir Kylheku wrote:
>
> > I say, don't bother. I avoid typedefs like the plague, even for structures. I
> > much rather write the longer form.
> >
> >     void myfunction(struct whatever *x);
> >
> > Using excessive typedefs will just occlude your code. What is the point of
> > having a separate type for MSEC? Later, when someone sees a declaration of a
> > variable of type MSEC, he will have to go hunting and pecking for the
> > definition of MSEC. It _obviously_ stands for milliseconds, but what the hell
> > is it?

If MSEC obviously stands for milliseconds, then the answer to the
question "what the heck is it" is "Milliseconds".  If you care at that
point how "milliseconds" is implemented, you need to go back and
relearn about "abstraction" and why it's important.

I'll grant that this isn't all that helpful in C, which doesn't
contain a whole lot of support for abstract data types; so you pretty
much *do* have to know what type of variable it is if you're going to
use it.  In C++ or Ada83 or Ada95 or UCSD Pascal or Modula or Eiffel,
which do provide constructs allow you to define all the operations on
a type in one place so you don't have to worry about it later, calling
things by abstract type names without worrying about how they're
implemented is an absolute MUST.  (By the way, in response to an
earlier post, THIS is the reason for having named types in those
languages---NOT lazy compiler implementors.)

> The way I was taught, you use typedefs only for variables that are likely
> to change type (and you want to change all of the instances of that type
> at once) and for cross-platform portability (in a multi-language
> environment like Windows or when you are saving data to disk).  Using it
> for most other things will, as you said, only hamper coding and debugging
> efforts.  However, to use those examples to induce "typedefs are useless"
> is jumping to a bad conclusion.  If you want to argue that the uses that
> I gave are bad, we can do that, but to generalize without looking at all
> potential uses isn't a good idea.

I must be missing something.  How does

        void myfunction (Whatever *x);

(where Whatever is a typedef) "occlude your code" or "hamper coding
and debugging efforts", while

        void myfunction (struct whatever *x);

doesn't?  The second example looks just like the first, in that you're
declaring a variable using some name that I'd have to go somewhere
else and look up if I wanted to see how it was implemented.  The only
difference is a keyword.

Or does the character sequence "struct" have some magical properties
that cause occlusion to break up and everything to become clear?

                                -- Adam