"Igor A. Goussarov" <igusarov@akella.com> wrote in message
news:3C75F9EC.96317A09@akella.com...
> Team Hyer wrote:
> >
> > There's one case no one seems to have mentioned:
> >
> >     int i;
> >     double z;
> >     auto j = i, x = z;
>
>    Well... Since "int x, y;" is equivalent to "int x; int y;" I'd like
> the declaration in your example be treated as "auto j = i; auto x = z;"
> Yes, this looks odd because one might expect that j and x are of the
> same type. But 'auto' is not a real _type_ it's a special keyword (or
> whatever).

In that case, how should this be interpreted?

    int i;
    double z;
    auto j = i, *k = &z;
    auto *l = i;

I think a general way of dealing with is these problems is by transforming
it to a case that the current language can handle:  function template
argument deduction.  So for each list of declarations create a function
declaration with an argument for each declarator in the declaration list,
and call each function with the initialiser:

    template <class __type_1>
    void __decl_1( __type_1 j, __type_1 *k );

    __decl_1( i, &z );

    template <class __type_2>
    void __decl_2( __type_2 *k );

    __decl_2( i );

This at least gives a consistent set of rules without requiring lots of new
type deduction rules to be added to the standard.  In the above example,
both declaration lists fail: deducing __type_1 fails because the first
argument deduces it as int and the second as double.  The second example
fails because i is an int, and cannot match a __type_2 * argument.

I think this is perhaps more useful that deducing the types of j and x (in
your example) independently.  This way, if you want to deduce two types
independently, you can just write

    auto j = i;
    auto x = z;

or if you want the same type to be used,

    auto j = i, x = z;

allows you to do that.  It seems to me that this gives the programmer more
flexibility.  This approach also has the advantage that the decision on
whether j should be an int or an int & are already made using an existing,
working mechanism.

--
Richard Smith


---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://www.research.att.com/~austern/csc/faq.html                ]

To me that says that j & x should have different types - something that is a
bit different, yes, but it seems to be the lesser of two evils - having to
remember ordering rules to figure if j as a double or if x was an int would
be bad, and trying to make the compiler do promotions so that j and x both
were doubles would trip people up a lot when you're working with classes --
and it would just make people want to write user defined conversions more...


"Team Hyer" <teamhyer@yahoo.com> wrote in message
news:20020219103527.45769.qmail@web12402.mail.yahoo.com...
> There's one case no one seems to have mentioned:
>
>     int i;
>     double z;
>     auto j = i, x = z;
>
> -- Tom Hyer
>
>
> __________________________________________________
> Do You Yahoo!?
> Yahoo! Sports - Coverage of the 2002 Olympic Games
> http://sports.yahoo.com
>
> ---
> [ comp.std.c++ is moderated.  To submit articles, try just posting with ]
> [ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
> [              --- Please see the FAQ before posting. ---               ]
> [ FAQ: http://www.research.att.com/~austern/csc/faq.html                ]
>

---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://www.research.att.com/~austern/csc/faq.html                ]

There's one case no one seems to have mentioned:

    int i;
    double z;
    auto j = i, x = z;

-- Tom Hyer


__________________________________________________
Do You Yahoo!?
Yahoo! Sports - Coverage of the 2002 Olympic Games
http://sports.yahoo.com

---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://www.research.att.com/~austern/csc/faq.html                ]