[NumPy-Tickets] [NumPy] #2087: In-place math functions like exp(A, out=A) giving wrong results with integer arguments

[NumPy Trac]

#2087: In-place math functions like exp(A, out=A) giving wrong results with
integer arguments
-------------------------+--------------------------------------------------
 Reporter:  chris        |        Type:  defect         
   Status:  new          |    Priority:  normal         
Milestone:  Unscheduled  |   Component:  numpy.matrixlib
  Version:  1.6.1        |    Keywords:                 
-------------------------+--------------------------------------------------
 Given an input matrix A with dtype 'int64', a mathematical function
 like exp incorrectly keeps the type of A as an integer type producing
 a wrong result if used as an in-place operation
 numpy.exp(A, out=A)

 The other version utilising a new result matrix works correctly:
 A = numpy.exp(A)

 Tested on MacPort (Mac OS X 10.6.8), Python 2.7.2 and Python 3.2.2.

 Example:

 {{{
 A = numpy.array([[1, 0],[0,1]])
 A.dtype
 ->dtype('int64')
 numpy.exp(A)
 ->array([[ 2.71828183,  1.        ],
 ->          [ 1.        ,          2.71828183]]) # correct result

 A = numpy.array([[1, 0],[0,1]])
 numpy.exp(A, out=A)
 ->array([[2, 1],
 ->         [1, 2]]) # wrong result
 A.dtype
 -> dtype('int64')
 A == numpy.array([[2, 1],[1,2]])
 -> array([[ True,  True],
 ->          [ True,  True]], dtype=bool)
 }}}

-- 
Ticket URL: <http://projects.scipy.org/numpy/ticket/2087>
NumPy <http://projects.scipy.org/numpy>
My example project