Is there a preferred way to create a matrix from a vector?

[Parsa Amini]

I don't see, say a DynamicMatrix constructor that accepts a dense vector, or something similar. Obviously, I can iterate over the elements of a vector to fill a matrix, but is there a better way of doing this?

[Klaus Iglberger]

Hi!

Thanks for creating this proposal. There is indeed no matrix constructor that accepts a vector. The preferred way to convert between vectors and matrices is via the row() and column() views:

DynamicMatrix<int> A( 3UL, 5UL );                    // Creating a 3x5 matrix
DynamicVector<int,rowVector>    a{ 1, 2, 3, 4, 5 };  // Creating a row vector
DynamicVector<int,columnVector> b{ 1, 2, 3 };        // Creating a column vector

row   ( A, 1UL ) = a;  // Assigning the row vector to the 1st row of A
column( A, 2UL ) = b;  // Assigning the column vector to the 2nd column of A

a = row   ( A, 2UL );  // Assigning the 2nd row of A to the row vector
b = column( A, 3UL );  // Assigning the 3rd column of A to the column vector

Please note that the interface distinguishes between row vectors and column vectors. The attempt to assign a row vector to a column vector or vice versa results in a compilation error.

The feature to assign a vector to all rows or all columns of a matrix is unfortunately not available yet, but has already been requested in issue #43. For now, the following two functions might help you:

template< typename MT, typename VT >
void fill( MT& matrix, const Vector<VT,columnVector>& vector )
{
   for( size_t j=0UL; j<matrix.columns(); ++j ) {
      column( matrix, j ) = ~vector;
   }
}

template< typename MT, typename VT >
void fill( MT& matrix, const Vector<VT,rowVector>& vector )
{
   for( size_t j=0UL; j<matrix.rows(); ++j ) {
      row( matrix, j ) = ~vector;
   }
}

If you pass a row vector to the fill() function, then it is assigned to all rows of the matrix, if you pass a column vector, it will be assigned to all columns of the matrix.

Hopefully this solves the problem.

Best regards,

Klaus!

---

Edit: As the matrix is a 3x5 matrix, the row vector a needs to be a vector with 5 elements and the column vector b needs to be a vector with 3 elements.