Is there a numpy function to perform coordinate transformations?

Question

I'm developing some lecture notes on numerical methods in an ipython notebook, and I need to include some coordinate transformation functions. I've already coded these functions (below), but the details aren't exactly fundamental to the topic, and it would simplify the lecture if I could call pre-written functions.

I see the scipy image rotation function, but I just need coordinate translation and rotation. ie

# transform from global to panel coordinates
def transformXY( self, x, y ):
    xp = x-self.xc
    yp = y-self.yc
    xpp = xp*self.sx+yp*self.sy
    ypp = yp*self.sx-xp*self.sy
    return [ xpp, ypp ]

# rotate velocity back to global coordinates
def rotateUV( self, u, v):
    up = u*self.sx-v*self.sy
    vp = v*self.sx+u*self.sy
    return [ up, vp ]

In these, sx and sy are the unit tangent vector components.

you can also move your own functions into a separate .py file and import them from there. That way their internals are hidden in the notebook. — PeterE, Mar 28 '15 at 17:51
@weymouth [in this answer](http://stackoverflow.com/a/22311973/832621) it is shown how I implemented a general cartesian coordinate system transformation using NumPy — Saullo G. P. Castro, Mar 29 '15 at 09:04
@weymouth Are these methods of a class? If not, you probably should not be using the name `self`. — Roland Smith, Mar 29 '15 at 15:30

PeterE · Accepted Answer · 2015-03-29T09:48:11.337

1

Copied and expanded from previous comment:

You can also move your own functions into a separate .py file and import them from there. That way their internals are hidden in the notebook. If you place them in a file/module called e.g. transformation.py and place that file next to your notebook file , you can then import your functions with from transformation import *

As has been suggested in a now deleted answer, I would recommend using matrices for any transformation, especially rotations. This is in my opinion much clearer than element-wise modifications.

EDIT: Afaik there are no predefined coordinate-transformation functions in numpy. In sympy there is something, but I'm not sure how useful it is, it seams a bit much for a simple transformation. For transformations via matrices google found me the this module, which appears to be quite comprehensive.

edited Mar 29 '15 at 09:48

answered Mar 28 '15 at 18:35

PeterE

5,715
5
29
51

3

Using `import *` is usually a [bad idea](http://stackoverflow.com/a/2386740/1461210) – ali_m Mar 28 '15 at 18:50
@ali_m True enough. My point was more about the fact that the folder the notebook is in is searched for modules to import. – PeterE Mar 28 '15 at 20:50
Yes, that would work for me, I guess. For posterity - is the answer to my question, "no". – weymouth Mar 29 '15 at 01:41
1

You *can* also apply translations (basically *all* [affine transformations](http://en.wikipedia.org/wiki/Affine_transformation)) with matrices, if you use [homogeneous coordinates](http://en.wikipedia.org/wiki/Homogeneous_coordinates). – Roland Smith Mar 29 '15 at 15:37

Is there a numpy function to perform coordinate transformations?

1 Answers1