Re: [boost] std::complex (Was: Re: multi_array management)

From: Greg Colvin (gcolvin_at_[hidden])
Date: 2002-01-12 17:18:40

From: "Beman Dawes" <bdawes_at_[hidden]>
> At 01:45 PM 1/10/2002, David Abrahams wrote:
> 
>  >What's needed here is a loosening of the restrictions on what constitutes 
> a
>  >POD.
> 
> I've run headlong into the "no constructors" limitation numerous 
> times.  I'd love to see a relaxation on the requirements.

As would I.  But if we relax it then PODs won't be aggregates.

> I'm worried that there was some serious reason for the restriction.  You 
> might want to ask someone in core why they are there before putting a lot 
> of effort into a proposal.

As I recall, the problem was drawing the line between objects with
a well defined, C-compatible layout and objects that had been, in
Mike Ball's phrase, "touched by C++ magic".  At the time I argued
for including as much of C++ as possible on the non-magical side
of the line, but implementors saw that as restricting their use of
magic.  So now the line is drawn very close to C, and as such the
concept of POD has proven useful in specifying the C-compatible
subset of C++.