Numpy - weird behavior with plus equal with slicing

Question

Plus equal is giving a different answer than assigning to the explicit sum (which is the answer one expects), when slicing is involved. Is there a reason for this? Should plus equal be avoided?

a = np.arange(10)
b = np.arange(10)
a[3:] += a[:-3]
b[3:] = b[3:] + b[:-3]
print a
#[ 0  1  2  3  5  7  9 12 15 18] 
print b
#[ 0  1  2  3  5  7  9 11 13 15]

Answer 1

As JBernardo commented, += change the array in place.

a[3:] += [a:-3] is similar to following:

>>> import numpy as np
>>> a = np.arange(10)
>>> 
>>> for i in range(3, 10):
...     print('a[{}] ({}) += a[{}] ({})'.format(i, a[i], i-3, a[i-3]))
...     a[i] += a[i-3]
...     print('  a[{}] -> {}'.format(i, a[i]))
... 
a[3] (3) += a[0] (0)
  a[3] -> 3
a[4] (4) += a[1] (1)
  a[4] -> 5
a[5] (5) += a[2] (2)
  a[5] -> 7
a[6] (6) += a[3] (3)
  a[6] -> 9
a[7] (7) += a[4] (5)  # NOTE: not (4)
  a[7] -> 12
a[8] (8) += a[5] (7)
  a[8] -> 15
a[9] (9) += a[6] (9)
  a[9] -> 18

To avoid that, use copy of the array:

>>> a = np.arange(10)
>>> a[3:] += np.copy(a[:-3]) # OR np.array(a[:-3])
>>> a
array([ 0,  1,  2,  3,  5,  7,  9, 11, 13, 15])