On Wed, Sep 16, 2009 at 7:50 AM, Rohit Garg
<rpg.314@xxxxxxxxx> wrote:
I have written the rigid transform class and it's unit test. The code
compiles, and executes correctly. The API is as close to eigen's
transform class as possible. I hope this is good enough to make into
eigen. The documentation is sketchy, but it was secondary till now.
The unit tests are not eigen'ish, but it is trivial to add
VERIFY_IS_APPROX, and I didn't want to add that to my code base.
NOTE: If you remove the V1.normalize() and V2.normalize() calls,
eigen's results change. I spent a lot of brain cycles fighting this.
:(
It may be a good idea to keep the axis vector in the angleaxis class,
and the direction vector in the parametrizedline class normalized.
Comments/questions/cribs/criticism/commendation (!) welcome.
Thanks for pushing this forward!. Definitely worth giving it a try and benching it with the existing Transform class (perfo-wise). Regarding the class name, I think it's important to clearly state that only 3D rigid transformations are being modeled, especially considering that the existing Transform class works for nD (and hence avoid confusion). Apart from that, and out of the top of my head, an inverse() function would be convenient, and maybe (?) a function for getting an equivalent Transform instance..
Cheers and thanks again!
Adolfo.
--
Rohit Garg
http://rpg-314.blogspot.com/
Senior Undergraduate
Department of Physics
Indian Institute of Technology
Bombay