[NumPy-Tickets] [NumPy] #1975: [Mixing integers and floats] : wrong results in algebraic equation

[NumPy Trac]

#1975: [Mixing integers and floats] : wrong results in algebraic equation
--------------------+-------------------------------------------------------
 Reporter:  bde     |       Owner:  somebody   
     Type:  defect  |      Status:  new        
 Priority:  normal  |   Milestone:  Unscheduled
Component:  Other   |     Version:  1.5.1      
 Keywords:          |  
--------------------+-------------------------------------------------------
 I'm experiencing a strange behavior of numpy when using a ndarray filled
 with integers in an algebraic equation with floats. See code snippet below
 :


 {{{
 >>>import numpy
 >>> def A(ar): return ((12200**2)*(ar**4)) / ( (ar**2+20.6**2) *
 (ar**2+12200**2) * (ar**2+107.7**2)**0.5 * (ar**2+737.9**2)**0.5 )
 ...
 >>> A(numpy.array([1,2,3])) //array with integers
 array([  2.95808751e-08,  -3.77586449e-07,  -1.61632406e-07]) // negative
 numbers are wrong (A is never negative)
 >>> A(numpy.array([1,2.0,3.0])) //array of int and floats
 array([  2.95808751e-08,   4.69917244e-07,   2.35099089e-06]) // correct!
 >>> A(numpy.array([1.,2,3.0]))
 array([  2.95808751e-08,   4.69917244e-07,   2.35099089e-06]) // correct!
 >>> A(1)
 2.9580875078524171e-08
 >>> A(1.)
 2.9580875078524171e-08
 >>> A(2)
 4.6991724410274925e-07
 >>> A(2.0)
 4.6991724410274925e-07

 }}}

 Is it a problem related to the way numpy makes the cast from int to float
 ? This is strange because there is no need to have ALL the numbers
 of "ar" to be float in order to make it work, one can remain int as you
 can see below...

-- 
Ticket URL: <http://projects.scipy.org/numpy/ticket/1975>
NumPy <http://projects.scipy.org/numpy>
My example project