Re: [eigen] Specializing max_coeff_visitor for some number types

[Christoph Hertzberg <chtz@xxxxxxxxxxxxxxxxxxxxxxxx>]

On 26.05.2014 13:56, Marc Glisse wrote:
> On Mon, 26 May 2014, Christoph Hertzberg wrote:
> > I don't know what would be the least ugly and least intrusive way. I'm
> > afraid that
> >  internal::is_same<scalar_abs_op, scalar_score_coeff_op>::value
> > will not work.
>
> More like std::is_base_of in this case.

Well, even if we re-implement is_base_of (to stay C++98 compatible) this 
won't be a good idea, because users might inherit from abs but redefine 
the function.

> > Maybe add another flag to
> >  functor_traits<scalar_score_coeff_op<Scalar> >?
>
> I was thinking of adding some typedef (or static enum or whatever) in
> the default implementation of scalar_score_coeff_op and testing for the
> existence of that typedef. Unless some user defines his own
> scalar_score_coeff_op<Type> specialization by deriving from
> scalar_score_coeff_op<OtherType>, it should detect if the user
> specialized scalar_score_coeff_op without the risk of forgetting to
> update functor_traits. Does that seem ok?

Yes, directly adding an enum to scalar_score_coeff_op is certainly safer 
to avoid unexpected behavior if the user forgets to implement something.


Christoph


--
----------------------------------------------
Dipl.-Inf., Dipl.-Math. Christoph Hertzberg
Cartesium 0.049
Universität Bremen
Enrique-Schmidt-Straße 5
28359 Bremen

Tel: +49 (421) 218-64252
----------------------------------------------