[Numpy-tickets] [NumPy] #269: scipy.take does not function correctly

[NumPy]

#269: scipy.take does not function correctly
--------------------+-------------------------------------------------------
 Reporter:  kantor  |       Owner:  somebody
     Type:  defect  |      Status:  new     
 Priority:  normal  |   Milestone:          
Component:  Other   |     Version:          
 Severity:  normal  |    Keywords:          
--------------------+-------------------------------------------------------
 I came across what I believe is a bug with the scipy.take function.  I was
 having trouble with optimize.leastsq. On one computer
 (powerpc) it was able to compute the covariance matrix of a least squares
 fit, however on a
 (x86) this failed with an error about inverse expecting a square matrix. I
 traced this to the fact that on the powerpc
 scipy.take(scipy.eye(2),[1,0]) correctly returns the matrix
 [[0,1],[1,0]], however on the x86 machine
 the same code only returns [0,1].
 Oddly directly using Numeric.take works correctly on x86 (gives the full
 matrix). This makes the leastsq function (when used with full_output) in
 scipy/optimize/minpack.py fail because the computed permutation matrix is
 just a vector on x86.

 These errors were on system running ubuntu in all cases. The scipy and
 numpy packages were built from the source on Andrew Staw's deb repository.

-- 
Ticket URL: <http://projects.scipy.org/scipy/numpy/ticket/269>
NumPy <http://projects.scipy.org/scipy/numpy>
The fundamental package needed for scientific computing with Python.