[NumPy-Tickets] [NumPy] #2122: numpy.take gives corrupted result with inplace operation

[NumPy Trac]

#2122: numpy.take gives corrupted result with inplace operation
-------------------------+--------------------------------------------------
 Reporter:  rainwoodman  |       Owner:  somebody   
     Type:  defect       |      Status:  new        
 Priority:  high         |   Milestone:  Unscheduled
Component:  numpy.core   |     Version:  1.6.1      
 Keywords:               |  
-------------------------+--------------------------------------------------

Comment(by david.warde-farley):

 I don't think this is a bug. AFAIK `numpy.take` isn't really intended to
 be used in this way. Furthermore, because arrays may not be contiguous in
 memory (e.g. smallest stride > 1 element), determining whether there is
 any overlap between input and output arrays is quite complicated, and I'm
 not sure it's practical to do so.

 There are lots of places in NumPy where inplace operations will
 unavoidably fail if you're not careful, e.g. you might think that `a =
 arange(50); a[:] += a[::-1]` will return an array of 49s, but it doesn't.

 I do agree that `take()` should at least document that it's not meant to
 be used when `a` and `out` alias each other, and to prefer fancy indexing
 in this case; ideally, an error would be raised, but I'm not sure that's
 practical for the reasons above.

-- 
Ticket URL: <http://projects.scipy.org/numpy/ticket/2122#comment:1>
NumPy <http://projects.scipy.org/numpy>
My example project