BUG: np.argsort not returning correct order

[RodrigoVillatoro]

Describe the issue:

The resulting order of np.argsort appears to be incorrect and does not match with the order returned by x[argsort_idx].

Specifically, when sorting in descending order, the highlighted number should be in position 11, not 9.

Strangely enough, this results in the correct ordering of the values:

Not entirely sure what can be going on here. Thanks!

Reproduce the code example:

import numpy as np

x = np.array([ 26.929932  ,  17.864126  ,   5.439558  ,  -0.10773016,
        -9.52839   ,  -7.620389  , -11.919527  , -15.451256  ,
       -16.36096   , -18.313873  , -19.791018  , -18.288052  ,
       -21.068373  , -24.030613  , -23.9124    , -25.630302  ,
       -29.34889   , -29.329416  , -29.221819  , -26.609049  ,
       -29.375376  , -28.213442  , -28.636854  , -28.140215  ,
       -27.100985  ])

# Returns incorrect order of indices
list(zip(x, np.argsort(-1*x)))

# Returns correct order of list
x[np.argsort(-1*x)]

Error message:

No response

NumPy/Python version information:

numpy 1.18.1
python 3.7.10

[seberg]

Argsort does not return ranks, it returns sorting indices to be applied to the input. So the zip() does not make any sense, because the order of the argsort result is sorted and not the original data order.

[RodrigoVillatoro]

You are completely right. Thank you!

[eric-wieser]

Duplicate of #8757