• =?ISO-8859-1?Q?=22Sz=2E_Horv=E1t=22?= Aug 31 2007
• Anton Sekeris <no.spam inter.nl.net> Aug 31 2007
• =?ISO-8859-1?Q?=22Sz=2E_Horv=E1t=22?= Aug 31 2007
• =?ISO-8859-1?Q?=22Sz=2E_Horv=E1t=22?= Aug 31 2007
• Anton Sekeris <no.spam inter.nl.net> Sep 01 2007
• Walter Bright <newshound1 digitalmars.com> Aug 31 2007
• Jan Knepper <jan smartsoft.us> Sep 01 2007

The following code gives an error in spite of c being const.

class A {
	static const double c = 1.0;
};


         static const double c = 1.0;
                               ^
te.cpp(5) : Error: initializer for non-const static member must be 
outside of class def
--- errorlevel 1

I've fallen into this trap as well on more than one occasion, but 
actually DMC is just following the standard here (you will see similar 
behaviour using compilers with the EDG front-end). You can only do what 
you are trying to do for integral or enumeration types.

9.4.2 Static data members [class.static.data] - item 4 says:

If a static data member is of const integral or const enumeration type, 
its declaration in the class definition can specify a 
constant-initializer which shall be an integral constant expression 
(5.19). In that case, the member can appear in integral constant 
expressions. The member shall still be defined in a namespace scope if 
it is used in the program and the namespace scope definition shall not 
contain an initializer.

Sz. Horvát wrote:
 The following code gives an error in spite of c being const.
 
 class A {
     static const double c = 1.0;
 };
 
 
         static const double c = 1.0;
                               ^
 te.cpp(5) : Error: initializer for non-const static member must be 
 outside of class def
 --- errorlevel 1

Anton Sekeris wrote:
 I've fallen into this trap as well on more than one occasion, but 
 actually DMC is just following the standard here (you will see similar 
 behaviour using compilers with the EDG front-end). You can only do what 
 you are trying to do for integral or enumeration types.
 
 9.4.2 Static data members [class.static.data] - item 4 says:
 
 If a static data member is of const integral or const enumeration type, 
 its declaration in the class definition can specify a 
 constant-initializer which shall be an integral constant expression 
 (5.19). In that case, the member can appear in integral constant 
 expressions. The member shall still be defined in a namespace scope if 
 it is used in the program and the namespace scope definition shall not 
 contain an initializer.


Hm ... you are right.  http://www.comeaucomputing.com/tryitout/ gives an 
error too.  But I really cannot see the reason for not allowing this, it 
just makes one's life more difficult ... did the people writing the 
standard simply forget about floating point types?

And DMC's error message is a little bit confusing (telling me that const 
double is not const).

 In that case, the member can appear in integral constant
 expressions.


 But I really cannot see the reason for not allowing this,


OK, so should I think about this as a special exception, allowed only to 
make things similar to the following code possible?

class A {
     static const int len = 3;
     int a[len];
};

But there is no special reason to allow doing the same with float or 
double (except maybe convenience? ...)

Sz. Horvát wrote:
  >> In that case, the member can appear in integral constant
 expressions.




  > But I really cannot see the reason for not allowing this,
 
 OK, so should I think about this as a special exception, allowed only to 
 make things similar to the following code possible?
 
 class A {
     static const int len = 3;
     int a[len];
 };
 
 But there is no special reason to allow doing the same with float or 
 double (except maybe convenience? ...)


I haven't a clue what the reason was for skipping non-integral built-in 
types, but your theory sounds quite plausible. From what I've seen in 
the past these types of standard strangeness are also rather often 
related to backwards compatibility with C, although in this scenario I 
don't readily see that.

Sz. Horvát wrote:
 But I really cannot see the reason for not allowing this, it 
 just makes one's life more difficult ... did the people writing the 
 standard simply forget about floating point types?


Ours is not to reason why, but to implement the Standard or cry. <g>

Sz. Horvát wrote:
 The following code gives an error in spite of c being const.
 
 class A {
     static const double c = 1.0;
 };
 
 
         static const double c = 1.0;
                               ^
 te.cpp(5) : Error: initializer for non-const static member must be 
 outside of class def
 --- errorlevel 1


h/hpp:

class  A
{
    static const double   c;
};

cpp:

const double       A :: c = 1.0;



-- 
ManiaC++
Jan Knepper

www.janknepper.com

But as for me and my household, we shall use Mozilla...
www.mozilla.org
Get legal - Get OpenOffice.org www.openoffice.org