# ndarray fancy index

## Fancy index

Fancy indexing is a NumPy term, which refers to the use of integer arrays for indexing. Suppose we have an 8 × 4 array:

``````In [117]: arr = np.empty((8, 4))

In [118]: for i in range(8):
.....:     arr[i] = i

In [119]: arr
Out[119]:
array([[ 0.,  0.,  0.,  0.],
[ 1.,  1.,  1.,  1.],
[ 2.,  2.,  2.,  2.],
[ 3.,  3.,  3.,  3.],
[ 4.,  4.,  4.,  4.],
[ 5.,  5.,  5.,  5.],
[ 6.,  6.,  6.,  6.],
[ 7.,  7.,  7.,  7.]])
``````

To select a subset of rows in a specific order, simply pass in a list of integers or ndarray s for the specified order:

``````In [120]: arr[[4, 3, 0, 6]]
Out[120]:
array([[ 4.,  4.,  4.,  4.],
[ 3.,  3.,  3.,  3.],
[ 0.,  0.,  0.,  0.],
[ 6.,  6.,  6.,  6.]])
``````

This code really meets our requirements! Using a negative index will pick rows from the end:

``````In [121]: arr[[-3, -5, -7]]
Out[121]:
array([[ 5.,  5.,  5.,  5.],
[ 3.,  3.,  3.,  3.],
[ 1.,  1.,  1.,  1.]])
``````

Passing in more than one index array at a time is a little special. It returns a one-dimensional array in which the elements correspond to the index tuples:

``````In [122]: arr = np.arange(32).reshape((8, 4))

In [123]: arr
Out[123]:
array([[ 0,  1,  2,  3],
[ 4,  5,  6,  7],
[ 8,  9, 10, 11],
[12, 13, 14, 15],
[16, 17, 18, 19],
[20, 21, 22, 23],
[24, 25, 26, 27],
[28, 29, 30, 31]])

In [124]: arr[[1, 5, 7, 2], [0, 3, 1, 2]]
Out[124]: array([ 4, 23, 29, 10])
``````

The reshape method is described in detail in Appendix A.

Finally, elements (1,0), (5,3), (7,1) and (2,2) were selected. No matter how many dimensions an array is, a fancy index is always one-dimensional.

The behavior of this fancy index may be different from what some users expect (including me). The row column subset of the selection matrix should be in the form of a rectangular region. Here is a way to get this result:

``````In [125]: arr[[1, 5, 7, 2]][:, [0, 3, 1, 2]]
Out[125]:
array([[ 4,  7,  5,  6],
[20, 23, 21, 22],
[28, 31, 29, 30],
[ 8, 11,  9, 10]])
``````

Remember, unlike slicing, a fancy index always copies data into a new array.

Posted by jadeddog on Sun, 09 Feb 2020 08:15:07 -0800