I recommend taking a look at how eigen does things for ideas before any major design decisions, and consider the possibility of directly integrating with ublas as well once you start getting into real matrix operations.

You may also want to have a pose object that defines the relative position and orientation of the object in space, for both 2D and 3D. This is closely related to a transform but in many cases more useful/easier to use. Given an object with one pose you can simply provide a function with the object, pose and the new desired pose, rather than having to construct a transform from one to the other. Of course this is not exclusive to the transform functions, and they should exist as well.

https://en.wikipedia.org/wiki/Pose_(computer_vision)

On Tue, May 28, 2013 at 7:26 PM, Adam Wulkiewicz <adam.wulkiewicz@gmail.com> wrote:

Hi,

Adam Wulkiewicz wrote:

Hi,

I've been thinking about additional concepts like vector, matrix, etc.
and how they could be used in Boost.Geometry. In our context they
probably should be seen only as translation and rotation representations.

/.../

In the free time you may check extensions/algebra. For now there is only Vector concept and some algorithms (for now only for vectors defined in cartesian cs). Basically this is what is implemented now: