I think push_back does only make sense for vectors. I would prefer a free fuction such as 'cat' that can operate also on tensors.

Being also an R programmer, I like the functional approach with free functions. It also makes sense from a template point of view.

It's indeed a 'cat' or 'c' we need. In R for example, you also have rbind and cbind for adding rows or columns to matrices/data.frame and 'c' for anything which looks like a vector. It's very convenient to use.

For a ublas::vector, it seems pretty obvious. For a matrix? Or a tensor?
If I decide to represent all the vectors as matrices now, they become row and column vectors then.

Should I expect the same effect on doing a push_back to a row vector and a column vector?

In general, could we make ublas STL compatible in general (means vector/matrix/tensor are STL compatible)?
Does my question even make sense?

Makes totally sense. I have already sth prepared. We can do it very generic.

Awesome. Could you create a branch in the git and push your code on that branch so that we can start playing with it/

Cheers,

David