Re: [eigen] [Review] Pull request 66, Huge Tensor module improvements

[Christoph Hertzberg <chtz@xxxxxxxxxxxxxxxxxxxxxxxx>]

On 16.06.2014 16:11, Christian Seiler wrote:
> Hi,
>
> > > I generally like your new names, but I dont think I can reuse
> > > DenseBase at the moment as a base class, since EigenBase,
> > > DenseCoeffsBase and DenseBase currently only support two indices.
> > > Thats why I suggested copying but not reusing Eigens own class
> > > hierarchy.
> >
> > I dont see why this is a problem: MultiDimBase would add the needed
> > multi-dim accessor overloads.
>
> Let me just a list a couple of methods of these classes:
>
> [...]

Another thing, which at least would look strange is the order of 
template parameters of Array vs Tensor.
  Array<Scalar, Rows, Cols, Options, MaxRowsAtCompileTime, 
MaxColsAtCompileTime>
vs
  MultiDimArray<Scalar, NumOfDimension, Options, Dim1, Dim2, ...>


And another question I could not extract from the discussion so far:
Do you intend to support Tensors where the number of dimensions is not 
fixed at compile-time?
Using ArrayXi-based indexing it would not be too hard to implement and 
it might have some (very) rare use-cases (e.g., n'th derivations of 
m-variable, single-value functions are m^n-tensors).

(To be clear: I don't mean to request this feature :) )

Christoph

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

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