Numpy Advanced Indexing with 1D array

Question

So given a 1D array like

x = np.array([0,1,2,3,4,5,6,7,8,9])

I want to index multiple elements at the same time. For example instead of

x[1]
x[2]

I want to use

x[(1,2)]
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
IndexError: too many indices for array

It works for 1 2D array like for example

x = np.array([[1,2,3,4],[5,6,7,8],[7,6,8,9]])
>>> x
array([[1, 2, 3, 4],
       [5, 6, 7, 8],
       [7, 6, 8, 9]])
>>> x[(1,2),(1,3)]
array([6, 9])
>>> x[(1,2),:]
array([[5, 6, 7, 8],
       [7, 6, 8, 9]])

So as you can see, for nd-arrays it works fine! Any way to do this kind of indexing for 1d-arrays?

Alex Riley · Accepted Answer · 2015-07-13 15:05:40Z

You need to wrap the indices in a list, not a tuple: x[[1,2]]. This triggers advanced indexing and NumPy returns a new array with the values at the indices you've written.

Whenever possible, NumPy implicitly assumes that each element of a tuple indexes a different dimension of the array. Your array has 1 dimension, not 2, hence x[(1,2)] raises an error.

The reason x[(1,2), :] succeeds with a 2D array is that you've explicitly told NumPy that the array has (at least) two dimensions and said what you want from the first two axes. The index is parsed as the 2-tuple ((1,2), :) so (1,2) is instead used for advanced indexing along the first axis. Had you simply used x[(1,2)] or x[1,2], you would get a single element at row 1, column 2.

Parsing the index is very complicated for NumPy because (unlike Python) there are several different indexing methods that can be used. To complicate things further, different methods may be used on different axes! You can study the exact implementation in NumPy's mapping.c file.

Kasravnd · Accepted Answer · 2015-07-13 14:45:18Z

3

You need to put the indices in a list not tuple,numpy use tuples for indexing multi-dimensional arrays :

>>> x[[1,2]]
array([1, 2])

>>> x[(1,2)]
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
IndexError: too many indices

Another example :

>>> x = np.array([6,4,0,8,77,11,2,12,67,90])
>>> x[[6,0]]
array([2, 6])

answered Jul 13, 2015 at 14:45

Kasravnd

108k19 gold badges167 silver badges195 bronze badges

1 Comment

Manfred Weis Over a year ago

to alleviate the problem with tuples vs lists, etc. writing x[[*idx]] that first "explodes" the index array and then "repackages" a list; works at least with idx being a tuple, a list and np.array

Collectives™ on Stack Overflow

Numpy Advanced Indexing with 1D array

2 Answers 2

Comments

1 Comment

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

Comments

1 Comment

Your Answer

Sign up or log in

Post as a guest

Related