Re: [boost] different matrix library?

Subject: Re: [boost] different matrix library?
From: joel (joel.falcou_at_[hidden])
Date: 2009-08-17 13:24:25

DE wrote:
> for expression classes only shapes to be concerned:
> rectangular, symmetric, upper- and lower-triangular
> what i forgot?
>   
I'm not sure you doesn't want to specialize on other thing.
But maybe we can have an hybrid solution : named expression for shape + 
tempalte policy for the remaining.
In the same vein, maybe either to have them be
expression<Xpr, diagonal>
than
diagonal_expression<Xpr>
so you can use partial tempalte specialization for matching them 'en masse'
> example?
Well see the Design rationale for boost::parameters
> but every unique entity can provide perfectly reasonable interface for
> just that entity
> e.g. a sparse matrix can provide non_zero_count() which will return
> the actual number of stored values
> furhtermore similar entities may (and must) be factored so there will
> be no code duplication
>   
Hmm, then we really need soemthign hybrid (CRTP+Policy)
CRTP leverage the raw things, Policy specialize them.
> arguments?
>   
See earlier but now it's maybe not that much valid

-- 
___________________________________________
Joel Falcou - Assistant Professor
PARALL Team - LRI - Universite Paris Sud XI
Tel : [phone]