Re: VC++ 6.0

[James W. <ja...@fr...>]

Jose' Cruanyes wrote:
> Il giorno 01/mar/07, alle ore 18:33, James W. Walker ha scritto:
> 
>> or we could change E3Num_Max to be a function template defined
>> exactly like std::max.
>>
> 
> taked... I'm leaving now (CET here) if tomorrow is still to do, I'll  
> do it.

So, where did you get that template implementation?  It's not the one 
used in Xcode or in CodeWarrior.  Those reference casts looked 
suspicious, so I tried it out...

	float	res1 = E3Num_Max( 23, 1.0f );
	float	res2 = E3Num_Max( 23, 100.0f );
	float	res3 = E3Num_Max( 1.0f, 23 );
	float	res4 = E3Num_Max( 100.0f, 23 );

The results I saw were 1e+9, 1e+9, 1.0, and 100.0 respectively.  Not good.

The answer to my original question "what is the advantage of using 
E3Num_Max rather than std::max?" seems to be: std::max does not allow 
operands of mixed type.  Not only can you not mix int and float, you 
can't mix int and unsigned long.  See 
<http://www.aristeia.com/Papers/C++ReportColumns/jan95.pdf> for a 
discussion of these issues.

When I tried a more standard definition of E3Num_Max and E3Num_Min, I 
was surprised to find that we weren't using them directly with mixed 
types, the only problem was with E3Num_Clamp.  So, I wrote a template 
version of E3Num_Clamp that casts its second and third parameters to 
match the type of the first parameter.

I've committed an update to E3Utils.h with the new template definitions.
-- 
   James W. Walker, Innoventive Software LLC
   <http://www.frameforge3d.com/>