Hi,

I currently have a multidimensional std::vector called "vecs". It has k vectors, each with 10 elements. I would like to map each of these vectors to an Eigen::VectorXd. I have the following code but at some point in my code, I get a heap corruption error. I've used the special allocator for STL containers as mentioned in the documentation (http://eigen.tuxfamily.org/dox/TopicStlContainers.html)

Code: Select all

std::vector<Eigen::VectorXd, Eigen::aligned_allocator<Eigen::VectorXd> > t_vecs.at(k);

for (int i = 0; i < k; i++){
   new (&t_vecs.at(i)) Eigen::Map<Eigen::VectorXd>(vecs.at(i).at(0), 10);
}


Could you please help me in correcting any mistakes in this code?

Thank you for your help.

I've never used the placement new. But I think that if you write "new (&v) Map<...>(...)" to make v map another piece of memory, then v should be of the type Map. However, in your code, t_vecs.at(i) is of type VectorXd. So my guess would be that you need to change the type of t_vecs as follows:

Code: Select all

std::vector<Eigen::Map<Eigen::VectorXd>,
            Eigen::aligned_allocator<Eigen::Map<Eigen::VectorXd> > > t_vecs(k);


Thanks for your suggestion. However, it produces the compile time error 'no appropriate default constructor available'.

On another note, if I have k = 1 and change my code to use just one map of one std::vector, everything works well. It's only when k > 1 and I start mapping to multideminsional vectors that this heap corruption error occurs.

I ended up just using a vector of Eigen::VectorXd at the beginning so there's no need to map it. It works now.

Thank You.

Jitse's reply was correct, the problem is that Map does not have a default ctor so you connot simply create a std vector of Map. I don't know if using the vector(size,model) ctor will work:

Eigen::Map<Eigen::VectorXd> empty(0,0);

std::vector<Eigen::Map<Eigen::VectorXd> > t_vecs(k, empty);

no need for the aligned allocator here.