numpy.argsort() bug

[JerryZhou199]

the function argsort() returns a wrong indice that would sort an array.
for example:

from numpy import *
a=array([4,2,3,1])
argsort(a)
array([3, 1, 2, 0], dtype=int64)

so this is right, it returned the sorted index. but the follow example is wrong:

a=array([4,2,3])
argsort(a)
array([1, 2, 0], dtype=int64)

the right index should be array([2,0,1]),and the follow example is also wrong:

a=array([5,1,4,3,2])
argsort(a)
array([1, 4, 3, 2, 0], dtype=int64)

the right index should be array([4,0,3,2,1]),and the follow example is also wrong:

a=array([7,1,5,2,4,6,3])
argsort(a)
array([1, 3, 6, 4, 2, 5, 0], dtype=int64)

the right index should be array([6,0,4,1,3,5,2]),so I find a error rule that function argsort() will display the wrong index when an array that starts with a maximum and has an odd number of elements.

Reproducing code example:

>>>from numpy import *
>>> a=array([4,2,3,1])
>>> argsort(a)
array([3, 1, 2, 0], dtype=int64)
### so this is right, it returned the sorted index. but the follow example is wrong:
>>> a=array([4,2,3])
>>> argsort(a)
array([1, 2, 0], dtype=int64)
### the right index should be array([2,0,1]),and the follow example is also wrong:

>>> a=array([5,1,4,3,2])
>>> argsort(a)
array([1, 4, 3, 2, 0], dtype=int64)
### the right index should be array([4,0,3,2,1]),and the follow example is also wrong:
>>> a=array([7,1,5,2,4,6,3])
>>> argsort(a)
array([1, 3, 6, 4, 2, 5, 0], dtype=int64)

Error message:

Numpy/Python version information:

Numpy 1.13.3/Python 3.5.2

1.13.3 3.5.2 (v3.5.2:4def2a2901a5, Jun 25 2016, 22:18:55) [MSC v.1900 64 bit (AMD64)]

[pv]

The invariant satisfied by argsort is "a[np.argsort(a)] is sorted", so all of the above are fine.

[eric-wieser]

Duplicate of #8757