numpy array integer indexing in more than one dimension

Question

I'm pretty sure I'm missing something with integer indexing and could use some help. Say that I create a 2D array:

>>> import numpy as np
>>> x=np.array(range(24)).reshape((4,6))
>>> x
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]])

I can then select row 1 and 2 with:

>>> x[[1,2],:]
array([[ 6,  7,  8,  9, 10, 11],
       [12, 13, 14, 15, 16, 17]])

Or the column 1 of rows 2 and 3 with:

>>> x[[1,2],1]
array([ 7, 13])

So it would makes sense to me that I can select columns 3, 4 and 5 of rows 1 and 2 with this:

>>> x[[1,2],[3,4,5]]
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
ValueError: shape mismatch: objects cannot be broadcast to a single shape

And instead I need to do it in two steps:

>>> a=x[[1,2],:]
>>> a
array([[ 6,  7,  8,  9, 10, 11],
       [12, 13, 14, 15, 16, 17]])
>>> a[:,[3,4,5]]
array([[ 9, 10, 11],
       [15, 16, 17]])

Coming from R, my expectations seem to be wrong. Can you confirm that this is indeed not possible in one step, or suggest a better alternative? Thanks!

EDIT: please note my choice of rows and columns in the example happen to be consecutive, but they don't have to be. In other words, slice indexing won't do for my case.

Jaime · Accepted Answer · 2014-01-25 15:27:09Z

5

You also have the option of using broadcasting among the indexing arrays, which is what I would normally do, rather than indexing twice, which creates an intermediate copy of your data:

>>> x[[[1], [2]],[[3, 4, 5]]]
array([[ 9, 10, 11],
       [15, 16, 17]])

To see a little better what is going on and how to handle larger numbers of indices:

>>> row_idx = np.array([1, 2])
>>> col_idx = np.array([3, 4, 5])
>>> x[row_idx.reshape(-1, 1), col_idx]
array([[ 9, 10, 11],
       [15, 16, 17]])

answered Jan 25, 2014 at 15:27

Jaime

67.7k19 gold badges128 silver badges164 bronze badges

Sign up to request clarification or add additional context in comments.

2 Comments

Miquel Over a year ago

Right you are, that intermediate copy was really bugging me. Thans a lot!

Eelco Hoogendoorn Over a year ago

nice; I wasn't aware of this!

Ashwini Chaudhary · Accepted Answer · 2014-01-25 10:14:42Z

2

Something like this:

In [28]: x[1:3, 3:6]                                                                             
Out[28]: 
array([[ 9, 10, 11],
       [15, 16, 17]])

answered Jan 25, 2014 at 10:14

Ashwini Chaudhary

252k60 gold badges478 silver badges519 bronze badges

2 Comments

Miquel Over a year ago

Thanks! I need to index with an index array instead of a slice, though . In other words, the rows and columns might not be consecutive or in odred

Ashwini Chaudhary Over a year ago

@Miquel I don't think it is possible to do indexing on multi-dimensional array that way. Read: docs.scipy.org/doc/numpy/user/…

Collectives™ on Stack Overflow

numpy array integer indexing in more than one dimension

2 Answers 2

2 Comments

2 Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

2 Comments

2 Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related