[NumPy-Tickets] [NumPy] #1732: Set complex128 array with a complex64 element fails

[NumPy Trac]

#1732: Set complex128 array with a complex64 element fails
--------------------------------------+-------------------------------------
 Reporter:  SevenThunders             |       Owner:  somebody
     Type:  defect                    |      Status:  new     
 Priority:  normal                    |   Milestone:  2.0.0   
Component:  numpy.core                |     Version:  1.5.0   
 Keywords:  complex array assignment  |  
--------------------------------------+-------------------------------------
 Observe the following code snippet:


 {{{
 import numpy as np
 z = np.empty((2),dtype=np.complex128)
 z[0] = np.complex64(1.+1.j)
 z[1] = np.complex64(1.+1.j)
 print z
 [ 1.+0.j  1.+0.j]
 }}}
 Numpy fails to make an obvious conversion from complex64 floats to
 complex128.  The imaginary parts are lost,  which can lead to surprising
 and difficult to find bugs.  Now if the intention is for this conversion
 to be explicit,  then the assignment should have generated an exception,
 not the bizarre behavior of only keeping the real part.

-- 
Ticket URL: <http://projects.scipy.org/numpy/ticket/1732>
NumPy <http://projects.scipy.org/numpy>
My example project