Re: [boost] Re: Re: ublas type safety and bound check

From: boost (boost_at_[hidden])
Date: 2002-07-03 06:20:42

Salut,

On Wednesday 03 July 2002 11:51, Benedikt Weber wrote:
> "Peter Schmitteckert (boost)" <boost_at_[hidden]> wrote
>
> > I like the feature that I can assign a full matrix to a more specialized,
> > in my cas a hermitean matrix.
> >
> > Changing this would currently break my code.
>
> It's useful and dangerous at the same time. I did not try it, but what
> happens if your full matrix is not hermitean. How is it transformed to the
> hermitean matrix? There are also the adapters, although I am not sure what
> they do exactly and if they are applicable in this example.
>

At the point of assignment my rhs matrix has to be hermitean.-
i would expect that it just takes the upper or lower triangle, depending
on the declaration of the hermitean matrix (upper / lower), but I
havn't checked this.

I agree that this behaviour is dangerous, so an explicit conversion
should be preferred.

Best wishes,
Peter

P.S.
Being a theoretical physicist I'm used to assign anything to everything,
as long a correct results are obtained. :)