[NumPy-Tickets] [NumPy] #1633: Changeset 8438 breaks ldexp on 64bit/bigendian systems

NumPy Trac numpy-tickets@scipy....
Mon Oct 11 15:59:12 CDT 2010


#1633: Changeset 8438 breaks ldexp on 64bit/bigendian systems
--------------------+-------------------------------------------------------
 Reporter:  nics    |       Owner:  somebody    
     Type:  defect  |      Status:  needs_review
 Priority:  normal  |   Milestone:  1.5.1       
Component:  Other   |     Version:  1.5.0       
 Keywords:          |  
--------------------+-------------------------------------------------------

Comment(by nics):

 The problem of #866 is not with ldexp, but with the fact that numpy cannot
 convert np.intc to np.int.
 This is the real issue and in #866 it errors out because it cannot fulfill
 the ldexp-request with the input data it got. numpy is completely right
 about that: It can not. The np.int-array should be cast to an np.intc
 array before, either manually or automatically by numpy.

 In the meanwhile: If you ask me, getting wrong results is more serious to
 scientists than getting error messages. My personal opinion is that the
 attached diff is less problematic than #866. On the other side it might
 break existing python code...

 Just to be curious: Are casts implemented to the bigger types, that is
 float->double, intc->int?

-- 
Ticket URL: <http://projects.scipy.org/numpy/ticket/1633#comment:4>
NumPy <http://projects.scipy.org/numpy>
My example project


More information about the NumPy-Tickets mailing list