Is there a nice (and efficient) way of normalising an array of row or column vectors?

I miss is too, so I'll try to implement it asap. In the meantime I think you can still do:

mat = mat.array().rowwise() / mat.colwise().norm().eval().array();