NetNews Offline 2/NetNews Offline Volume 2.iso/news/comp/lang/c++-part1/7610

Text File | 1996-08-05 | 3.8 KB | 90 lines

Newsgroups: comp.lang.c++
Path: netcom.com!marnold
From: [e-mail] (Matt Arnold)
Subject: Re: Is there a standard for * and & placement style?
Message-ID: <[e-mail]>
Organization: NETCOM On-line Communication Services (408 261-4700 guest)
References: <[e-mail]> <[e-mail]> <4gckd5$[e-mail]> <[e-mail]> <4gk1b8$[e-mail]>
Date: Sat, 24 Feb 1996 08:24:10 GMT
Sender: [e-mail]

[e-mail] (David White) writes:

>[e-mail] (Matt Arnold) writes:

>>[e-mail] (Harlan Messinger) writes:

>>>I think the syntax is dumb in the first place. One declaration statement
>>>should declare variables of one type. I think

>>>    char* a, b, c;

>>>should declare three (char*) variables. If someone also wants a char
>>>variable d, then that should be on two lines:

>>That would be better, but I still subscribe to the "one concept per
>>line" theory of programming and would never declare three variables on
>>a single line.

>The concept used in declarations is that a declaration mimics an
>expression, e.g., given char *p, if you see *p then it evaluates to a
>char. The above form is therefore a special case because it can't always
>be used. In the case of, for example, char (*p)[10], you can't apply it.
>I can see that in the trivial special cases, like char* foo(), it seems
>more sensible to some people (foo() returns a char*), but it is an oddity
>in terms of the general rule.

Hmm, thinking of function foo() as "returning a char*" doesn't seem to me
to be any more a special case than thinking of p as "_being_ a char*".

I still think pointer-ness or reference-ness is better associated with the
type.  When I see "p" (from your example) in some code, I think of it as
being type "char*", and that is the way I declare it.

In fact, I'd argue that the fact that a function declaration such as...

   char* foo();

...is a "special case" for the "char *"-style, simply highlights that the
style, IMO, is conceptually inconsistent.  Certainly, the function itself
doens't have pointer-ness (although, as for as the language it concerned, it
resovles to an expression that does) and so it doesn't seem quite right to
associate the * with the function name.  It seems inconsistent that you
would make this concession, but turn around and insist the "char *p" makes
sense for the declaration of p.

I have never had any inconsistencies like this with the "char*"-style.  It
even, to me, resolves nicely with the fact that you can typedef a "char
pointer type", PCHAR say, and substitute it for all the char*'s throughout
the code, whether they be found in casts or declarations or function return
types.

I don't think there any magic in the fact that C/C++ declarations look
like expressions.  What if, for example, one declared pointer variables
using a symbol completely different than for dereferencing?

   char@ p = // alloc storage to put char "at"

   *p = 'x'

It would still make sense to me to think of p's type as "char@", which
after all *is* its type.


I guess it all depends on where you imagine pointer-ness residing:  "in"
the variable, or "in" the type.  To me it makes more sense, and actually
is correct, to think of pointer-ness as a being part of type (we all know
this is how compilers work).  I therefore prefer to write "char* p", not
"char *p".


-------------------------------------------------------------------------
Matt Arnold                       |        | ||| | |||| |  | | || ||
[e-mail]                |        | ||| | |||| |  | | || ||
Boston, MA                        |      0 | ||| | |||| |  | | || ||
617.389.7384 (h) 617.576.2760 (w) |        | ||| | |||| |  | | || ||
C++, MIDI, Win32/95 developer     |        | ||| 4 3 1   0 8 3 || ||
-------------------------------------------------------------------------