fftshift and cart2pol

[Gordon Haverland]

Following up on my best practices question.

(Yes, that function is from the mathworks exchange, but it looks 
like it needs extensive changes to actually work.)

If I produce a matrix of values, I will design it such that 1 row 
corresponds to Y=0 and 1 column corresponds to X=0.  And, I would 
expect something like cart2pol to work just fine.

If I take the FFT of 2D data and then shift it using fftshift, I 
have 2 rows representing x=0 and two columns representing y=0.  I 
would imagine that the data structures for Fourier transformed 
matrices are not special, and so cart2pol has no idea that the 
axes are actually doubled.  And I suspect this means there is an 
error in the radii computed for each point.

Or is there some magic in cart2pol to look out for this?

Thanks.
Gord