Re: [eigen] Recursion and block matrices

[Norman Goldstein <normvcr@xxxxxxxxx>]

Thanks for the good tips.  I have found that Ref is not working for me 
with Array (at least,
not as well as it works with Matrix).  Perhaps, the current devel has 
not caught up on the
Array side?  Here is source code to illustrate the problem:

////////////////////////////////////////////////////////////////////////////////////////
#include <Eigen/Core>
using namespace Eigen;

//typedef ArrayXXi D;
typedef MatrixXi D;

int main( int argc, char* argv[] )
{
  D exponents( 3, 4 );

  Ref< D > e1 =
    exponents.bottomLeftCorner( exponents.rows() - 1,
exponents.cols() );

  e1(0,0) = 3; // Avoid "unused variable" warning.

  return 0;
}// main
//////////////////////////////////////////////////////////////////////////////////////////

This compiles as is, but changing the typedef to the Array one, gives 
the error:

error: conversion from ‘
Eigen::Block<Eigen::Array<int, -1, -1>, -1, -1, false, true>
’ to non-scalar type ‘
Eigen::Ref<Eigen::Array<int, -1, -1> >
’ requested

Thank you,
Norm


On 09/21/2012 11:01 AM, Christoph Hertzberg wrote:
> On 21.09.2012 19:05, Norman Goldstein wrote:
> > -- What was it about the Eigen class designs that made it convenient to
> > restrict Ref to innerStride==1?
> > Would it be a major development to remove this restriction? (convenient
> > for transposing and even/odd
> > subscripting).
>
> It's only much more efficient at runtime. If you don't mind that, you 
> can allow dynamic inner stride this way:
>     Ref<MatrixXf,0,Stride<> >
>
> Read this for details:
> http://eigen.tuxfamily.org/dox-devel/classEigen_1_1Ref.html
>
> I'm not sure if the current implementation would allow passing a 
> transposed matrix to that yet (I guess not).
>
>
> > -- Is there a signature for incr() that would allow me to pass in the
> > MatrixXd directly, without first
> > converting it to a Ref in main():
> >    Ref< MatrixXd > sub( mat );
> >    incr( sub );
>
> void incr( Ref< MatrixXd > matV );
>
> and in C++11 it should also be possible to use an l-value reference:
> void incr( Ref< MatrixXd >&& matV )
>
> > -- Are the two expressions
> >    Ref< MatrixXd > sub = mat;
> >    Ref< MatrixXd > sub( mat );
> > supposed to be equavalent?  I think that a compiler is allowed to
> > interpret the
> > first one as a default constructor followed by an assignment operator.
>
> No, for a class A
>     A a = b; // not allowed if constructor is explicit ...
> and
>     A a(b);
> call the constructor, and
>     A a; a = b;
> calls the default constructor and then the assignment operator.
> Ref does not have a default constructor so the second interpretation 
> would not even be possible.
> Of course, there is (almost?) always a rule that the compiler can 
> interpret things differently, if the outcome is the same.
>
>
>
> Christoph