Re: [ublas] [bindings][traits] matrix_traits

Subject: Re: [ublas] [bindings][traits] matrix_traits
From: Karl Meerbergen (karl.meerbergen_at_[hidden])
Date: 2009-03-13 09:17:20

Rutger,

Be careful with changes in the traits: other bindings code might have to
be modified.
Adding triangular_t and triangular_packed_t is a good suggestion.

I do not think unitary, orthogonal and positive definite properties can
be considered as structural properties. Those form a third category,
e.g. matrix_math_property.

Karl

Rutger ter Borg wrote:
> All, 
>
> after having a closer look at the current matrix_traits, I think we need one
> more concept that is orthogonal on the current matrix_structure. Right now,
> it has matrix_structure type which can be something along the line of 
>
>   struct general_t {};
>   struct symmetric_t {};
>   struct symmetric_packed_t {};
>   struct hermitian_t {};
>   struct hermitian_packed_t {};
>   struct banded_t {};
>   struct triangular_t {};
>   struct triangular_packed_t {};
>
> Now, what I think we could make it ourselves a bit easier if we split this
> in matrix structure (or property) and storage layout.
>
> E.g., 
>
> matrix_traits<T>::matrix_storage
> dense_t {};
> packed_t {};
> diagonal_t {};
> bidiagonal_t {};
> tridiagonal_t {};
> banded_t {};
>
> matrix_traits<T>::matrix_structure
> general_t {};
> symmetric_t {};
> hermitian_t {};
> triangular_t {};
> positive_definite_t {};
> hessenberg_t {};
> orthogonal_t {};
> unitary_t {};
>
> I even think this list is near exhaustive. What do you think?
>
> Regards,
>
> Rutger
>
>
>
>
> _______________________________________________
> ublas mailing list
> ublas_at_[hidden]
> http://lists.boost.org/mailman/listinfo.cgi/ublas
>