Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!usc!samsung!uunet!world!wmm
From: wmm@world.std.com (William M Miller)
Newsgroups: comp.lang.c++
Subject: Re: difference between friend and static member function
Message-ID: <1991Feb27.175343.6038@world.std.com>
Date: 27 Feb 91 17:53:43 GMT
References: <1991Feb27.153602.19495@kodak.kodak.com>
Organization: The World Public Access UNIX, Brookline, MA
Lines: 32

cok@islsun.Kodak.COM (David Cok) writes:
> 4) The compiler (Sun C++ 2.0 on a SparcStation) complains about
>
>         static myclass operator+(myclass,myclass)
>
> saying
>         error:  myclass::operator +() must take 0 or 1 arguments
>
> This seems to me to be a bug since there is no implied this argument for a
> static member.  Right or wrong?

E&S section 13.4.2, page 333, says: "A binary operator may be declared
either by a nonstatic member function (section 9.3) taking one argument or
by a nonmember function taking two arguments."  (A similar passage in 13.4.1
deals with unary prefix operators.)  There's no provision for static member
functions to be used for operator overloading.  The compiler was correct to
reject the declaration.

To understand why, think about the definitions of operator member functions
and static member functions.  An expression "a + b," where "a" is an object
of a class with a member operator+(), is, by definition, equivalent to
"a.operator+(b)."  However, if a static member function is invoked using the
member selection syntax "a.smf()," the expression on the left side of the
"." is not evaluated (E&S section 9.4, page 180).  That's clearly not what
you want to happen with operators.

Obviously the rules could be rewritten to make static operator member
functions behave differently from ordinary static member functions, but it's
not clear, at least to me, whether it's worth changing.

-- William M. Miller, Glockenspiel, Ltd.
   wmm@world.std.com