[NumPy-Tickets] [NumPy] #764: Inconsistent behavior of float32, float64

[NumPy Trac]

#764: Inconsistent behavior of float32, float64
------------------------+---------------------------------------------------
 Reporter:  charris     |       Owner:  somebody      
     Type:  defect      |      Status:  needs_decision
 Priority:  normal      |   Milestone:  Unscheduled   
Component:  numpy.core  |     Version:  devel         
 Keywords:              |  
------------------------+---------------------------------------------------
Changes (by pwil3058):

 * cc: pwil3058@… (added)


Comment:

 I'm having problems that may be related to this issue and I'd like an
 opinion, please.

 Basically what happens is I determine an angles (using either math.atan2
 or numpy.arctan2) and stick them in a numpy array and when I get them back
 they are not always in the range math.pi to -math.pi (or numpy.pi to
 -numpy.pi) if their absolute values were near pi (although they were when
 I put them in).

 I've tried this with a number of different dtypes for the array and always
 have problems:

 1. either my asserts re the values are triggered, or
 2. they aren't triggered and I get an incorrect result due the
 calculations going ahead in spite of being out of the bounds.

 Any ideas?

-- 
Ticket URL: <http://projects.scipy.org/numpy/ticket/764#comment:7>
NumPy <http://projects.scipy.org/numpy>
My example project