Numpy: Select row multiple times using different indices array

Question

Suppose I have the following array.

l = np.asarray([1,3,5,7])

Out[552]: array([1, 3, 5, 7])

I can select the row twice using a index array np.asarray([[0,1],[1,2]]):

l[np.asarray([[0,1],[1,2]])]
Out[553]: 
array([[1, 3],
       [3, 5]])

It doesn't work if the index array have different length on each row:

l[np.asarray([[1,3],[1,2,3]])]

Traceback (most recent call last):

  File "<ipython-input-555-3ec2ab141cd4>", line 1, in <module>
    l[np.asarray([[1,3],[1,2,3]])]
IndexError: arrays used as indices must be of integer (or boolean) type

My desired output for this example would be:

array([[3, 7],
       [3, 5, 7]])

Can someone please help?

numpy doesn't support ragged (non-rectangular arrays), so you might want to think of other datastructures you could use to store your output (like lists). — DSM
– DSM, Commented May 3, 2017 at 0:38

Chih-Hsu Jack Lin · Accepted Answer · 2017-05-03 00:50:50Z

1

I think this is the closest I can get.

import numpy as np
l = np.asarray([1, 3, 5, 7])
idx = [[1,3],[1,2,3]]
output = np.array([np.array(l[i]) for i in idx])
print output

Result:

[array([3, 7]) array([3, 5, 7])]

answered May 3, 2017 at 0:50

Chih-Hsu Jack Lin

4532 silver badges17 bronze badges

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

2 Comments

Allen Qin Over a year ago

Thank you. This is closer. Maybe it can't get any better if Numpy doesn't support ragged array as @DSM mentioned.

Chih-Hsu Jack Lin Over a year ago

Yeah, I think list is better for such indexing.

Stephen Rauch · Accepted Answer · 2017-05-03 00:32:59Z

0

You can get the result you are after if you build the lists seperately.

Code:

l = np.asarray([1, 3, 5, 7])

# Build a sample array
print(np.array([[3, 7], [3, 5, 7]]))   

# do the lookups into the original array 
print(np.array([list(l[[1, 3]]), list(l[[1, 2, 3]])]))

Results:

[[3, 7] [3, 5, 7]]
[[3, 7] [3, 5, 7]]

answered May 3, 2017 at 0:32

Stephen Rauch♦

50.1k32 gold badges118 silver badges143 bronze badges

2 Comments

Allen Qin Over a year ago

Thanks for the answer but my index array has many rows and this method won't scale.

Stephen Rauch Over a year ago

@Allen, Often a good idea to put these sorts of details into the question, and not left as happy surprises later.

Collectives™ on Stack Overflow

Numpy: Select row multiple times using different indices array

2 Answers 2

2 Comments

2 Comments

Your Answer

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

Related