Re: [boost] [Review] Fixed Strings review period extended

From: Andy Little (andy_at_[hidden])
Date: 2006-01-30 15:00:06

"Reece Dunn"  wrote

> But the solution is still using templates. You don't need to have a template
> to use std::[w]string. Thus, if you wanted to move:
>
>   __declspec(dllexport)
>   void some_function( const std::string & str );
>
> to use fixed_string with the implementation you mention, you would need to
> expose the implementation of some_function to your customers and thus cannot
> use it as part of a library (.lib) or DLL.

The limitations of templates  are widely known. The advantages are too. If the 
advantages dont appear in this case then there is little point in using 
templates.  If it needs to be implementable as part of a DLL then it obviously 
cant practically have compile time fixed capacity, but that is the trade off you 
have to make ( OTOH maybe you could instantiate a few useful sizes, so maybe its 
not impossible).   It strikes me that the design radically changed  midstream 
and you now have a strange hybrid. This is the source of a lot of confusion with 
the result that potential users throw up their hands and move on. Are you 
designing a string  with a compile time fixed size or are you dessigning a 
string with a construction time fixed capacity, if so then why not  a  runtime 
variable capacity?  What requirements must the class satisfy? What are the 
limitations imposed by these requirements? What then are the advantages over 
std::string etc, etc.

regards
Andy Little