How to traverse the matix rows for index not in a special index vector P by Eigen?

Question

Today i met a trouble.I have a special index vector P,and a matrix A.for example:

    MatrixXd B = MatrixXd::Ones(10, 10);
    std::cout << B;
    std::vector<int> P = { 1, 6, 3, 7, 5 };

and I want to change values of rows of B.Which rows' indexs is not in P;In this case ,i want to do some implements to row 0,2,4,5,8,9.

But how to get that efficiently by eigen?

Answer 1

The use of Eigen is rather secondary to the question. The main issue is that you need to find all indices in the range [0, B.rows()) which are not in P. From the looks of things, P is not sorted, which is unfortunate. I assume it is at least unique as this saves us some memory.

Here is a version that is optimized to be fast, rather than pretty.

std::vector<int> indices_not_in(const std::vector<int>& P, int end)
{
    assert(P.size() <= end);
    std::vector<bool> inP(end);
    for(int pi: P) {
        assert(inP[pi] == false && "Duplicate entry");
        inP[pi] = true;
    }
    std::vector<int> notP(inP.size() - P.size() + 1);
    auto cur = notP.begin();
    int i = 0;
    for(bool inPi: inP) {
        *cur = i++;
        cur += ! inPi;
    }
    notP.erase(cur, notP.end());
    return notP;
}

int main()
{
    Eigen::MatrixXd B = Eigen::MatrixXd::Ones(10, 10);
    std::cout << B << "\n\n";
    std::vector<int> P = { 1, 6, 3, 7, 5 };
    std::vector<int> notP = indices_not_in(P, B.rows());
    B(notP, Eigen::all) *= 2.;
    std::cout << B << "\n";
}

If you cannot guarantee that the P vector contains unique entries, then here is a modified version that works with that:

std::vector<int> indices_not_in(const std::vector<int>& P, int end)
{
    std::vector<bool> inP(end);
    for(int pi: P)
        inP[pi] = true;
    std::vector<int> notP(end);
    auto cur = notP.begin();
    int i = 0;
    for(bool inPi: inP) {
        *cur = i++;
        cur += ! inPi;
    }
    notP.erase(cur, notP.end());
    return notP;
}