python - unexpected result in numpy array slicing (view vs copy) -

i'm trying reduce amount of copying in code , came across surprising behavior when dealing numpy array slicing , views, explained in:

scipy wiki page on copying numpy arrays

i've stumbled across following behavior, unexpected me:

case 1.:

import numpy np = np.ones((3,3)) b = a[:,1:2] b += 5 print print b.base 

as expected, outputs:

array([[ 1.,  6.,  1.],        [ 1.,  6.,  1.],        [ 1.,  6.,  1.]]) true 

case 2: when performing slicing , addition in 1 line, things different:

import numpy np = np.ones((3,3)) b = a[:,1:2] + 5 print print b.base 

the part that's surprising me a[:,1:2] not seem create view, used left hand side argument, so, outputs:

array([[ 1.,  1.,  1.],        [ 1.,  1.,  1.],        [ 1.,  1.,  1.]]) false 

maybe can shed light on why these 2 cases different, think i'm missing something.

solution: missed obvious fact "+" operator, other in-place operator "+=" create copy, it's in fact not related slicing other how in-place operators defined numpy arrays.

to illustrate this, following generates same output case 2:

import numpy np = np.ones((3,3)) b = a[:,1:2] b = b + 5 print print b.base 

the above no different than:

>>> a=np.arange(5) >>> b=a >>> b array([0, 1, 2, 3, 4])  >>> b+=5 >>> array([5, 6, 7, 8, 9]) >>> b array([5, 6, 7, 8, 9])  >>> b=b+5 >>> b array([10, 11, 12, 13, 14]) >>> array([5, 6, 7, 8, 9]) 

which, @ least me, seem expected behavior. b+=x operator calls __iadd__ importantly first tries modify array in place, update b still view of a. while b=b+x operator calls __add__ creates new temporary data , assigns b.

for a[i] +=b sequence (in numpy):

a.__setitem__(i, a.__getitem__(i).__iadd__(b))