采用args特征稀疏矩阵的C ++函数

Question

Eigen library provides/suggests numerous ways to pass a dense matrix in a function, so that it works for a different types that share the same base, and avoids copying (ie Ref<>, template expressions).

However, I haven't found anything equivalent for sparse matrices in either eigen documentation or online.

I basically have the following question: How can I write a function with a generic interface so it can be called with both SparseMatrix or MappedSparseMatrix objects, of potentially different template arguments, without copying?

I have tried template expression of SparseMatrixBase withe derived arguments but I couldn't make it work.

A simple example code will be appreciated.

Answer 1

Simply write a template function taking any SparseMatrixBase<Derived> , for instance:

template<typename Derived>
void foo(const SparseMatrixBase<Derived> &a_mat) {
    const Derived &mat(a_mat.derived());
    SparseMatrix<typename Derived::Scalar> tr_mat = mat.transpose();
}