[NumPy-Tickets] [NumPy] #1982: In-place arithmetic operations: wrong calculation order?

[NumPy Trac]

#1982: In-place arithmetic operations: wrong calculation order?
------------------------+---------------------------------------------------
 Reporter:  ling        |       Owner:  somebody   
     Type:  defect      |      Status:  new        
 Priority:  normal      |   Milestone:  Unscheduled
Component:  numpy.core  |     Version:  1.6.0      
 Keywords:              |  
------------------------+---------------------------------------------------
 In-place operation like "subtract(b[:-1], b[1:], b[:-1])" used to work
 fine in Numpy 1.4.1, but sometimes fails with Numpy 1.6.0.  See the
 example below (the problematic output is Out[6]):

 {{{
 In [1]: import numpy; numpy.__version__
 Out[1]: '1.6.0'

 In [2]: a = numpy.arange(12).reshape((3, 4))

 In [3]: b = a[::-1].copy(); numpy.subtract(b[:-1], b[1:])
 Out[3]:
 array([[4, 4, 4, 4],
        [4, 4, 4, 4]])

 In [4]: numpy.subtract(b[:-1], b[1:], b[:-1])
 Out[4]:
 array([[4, 4, 4, 4],
        [4, 4, 4, 4]])

 In [5]: b = a[::-1]; numpy.subtract(b[:-1], b[1:])
 Out[5]:
 array([[4, 4, 4, 4],
        [4, 4, 4, 4]])

 In [6]: numpy.subtract(b[:-1], b[1:], b[:-1])
 Out[6]:
 array([[4, 5, 6, 7],
        [4, 4, 4, 4]])
 }}}

 Similar problem happens to other in-place operations (e.g., multiply,
 divide).

-- 
Ticket URL: <http://projects.scipy.org/numpy/ticket/1982>
NumPy <http://projects.scipy.org/numpy>
My example project