NetNews Usenet Archive 1992 #30/NN_1992_30.iso/spool/comp/lang/cplus/18045

Text File | 1992-12-16 | 3.3 KB | 84 lines

Newsgroups: comp.lang.c++
Path: sparky!uunet!uunet.ca!frumious!pat
From: [e-mail] (Patrick Smith)
Subject: Re: The fate of my exponentiation operator proposal
Message-ID: <[e-mail]>
Date: Wed, 16 Dec 1992 08:37:33 GMT
References: <[e-mail]> <[e-mail]> <[e-mail]>
Organization: None
Lines: 73

[e-mail] (Jim Adcock) writes:
|In article <[e-mail]> uunet.ca!frumious!pat writes:
||[e-mail] (Jim Adcock) writes:
|||Where is the offsetting benefit in *requiring* implementations to make this
|||change breaking conforming programs?  ANSI-C *already* requires that
|||macro implementations only perform one evaluations of the macro parameters,
|||so what are you gaining by forcing implementations to now use inline
|||functions instead of macros for these things?
||
||If these functions may be implemented as macros, then their names may not
||be used in other ways (not by programmers who aspire to write portable
||code, anyway).  For example, consider
||
||   class Prisoner {
||   public:
||      // ...
||      void free();   // Let him out on parole; not related to malloc()
||   };
|
|Agreed.  You give an excellent example of why ANSI-C functions should
|continue to be implemented as macros not as inlines.


I take it you consider the above to be an unreasonable use of the
name free.  Perhaps it is and perhaps it isn't; people often have
different opinions as to what's reasonable.  In the absence of a
strong concensus that something is unreasonable, I would prefer to
allow those who believe it to be reasonable to do it.


Here are a few other examples you might find more convincing.

Consider the above in the context of a project where malloc() and the
normal free() are never used; all memory allocation is done through
new and delete.

Steve Clamage mentioned the possibility of using sin to refer to
immoral acts, instead of the sine function.

In some cases, an existing name from the standard library can
be quite intuitive as a name for a new function.  For example,

   class String {
      // A sequence of characters, possibly including null characters
      // in arbitrary positions.
   public:
      size_t length() const;
         // Length of the sequence, including the nulls (if any).
      size_t strlen() const;
         // The number of characters before the first null character
         // (not including the null); the same as length() if there
         // aren't any nulls in the string.
   };

Stepping to another language for a moment, there's a problem with
this C code, if memory serves me correctly:

   typedef whatever Razor;
   void strop(Razor r);

The problem?  The C standard reserves names beginning with str for
use in future versions of the standard.  So this code might not
compile a few years down the road.  If one takes this restriction
seriously, there are quite a few normal English words one shouldn't
use in C programs (strong, straight, stress, strict, etc., etc.).
It would be nice to be able to use these words in C++ programs
without having to worry about future versions of the standard,
even if only in some contexts (eg. as member function names).

--
Patrick Smith
uunet.ca!frumious!pat
[e-mail]