When will a member function for shifting vector values be supported in Eigen?

Thanks.

are you talking about bitewise shifting (e.g., apply operator<<(int) to each coeff) or element wise shifting (0 1 2 3 => _ _ 0 1) ?

If the former I realize that we won't be able to use operators like m << 2 because operator<<(scalar) is already overloaded for the comma initializer.

I am talking about shifting elements.

Ex: rowVector = [1, 2, 3, 4]
rowVector.leftShift(1) // shift elements by 1 to the left, adding a copy
// adding a copy of the last element at the left
// side.
so now rowVector is [2, 3, 4, 4]

I can do this myself, but I was wondering when the project will code methods like this in a possibly more efficient way.

Thanks.

ok, this would require a new expression class. I can imagine 3 variants:
- circular shift, [1, 2, 3, 4] => [2, 3, 4, 1], the overhead is one integer modulo (= circshift in matlab)
- shift with clamping (what you suggest), the overhead is one integer min or max.
- shift with insertion of zeros, requires one if

This feature is of low priority for us, so unless someone else is interested to do it, I cannot tell you when it will be available.

Shifting is apparently still not implemented, but is fairly easily accomplished using the Map class. This should be scalable to matrices depending on the storage order of the matrix.

// Initialize
int n = 4;
float x[4] = { };
Map<MatrixXf> mf(x,1,n);
mf << 1, 2, 3, 4;

// Shift the matrix
memcpy(x,&x[1],(n-1)*sizeof(&x[0]));