Getting only particular columns in every row in a numpy array

Question

Possible Duplicate:
numpy: access an array by column

I have a numpy array (numpy is imported as np)

gona = np.array([[ 1,  2,  3],
       [ 4,  5,  6],
       [ 7,  8,  9],
       [10, 11, 12]])

I can get the values of entire column of 1th row by gona[1][:].

array([4, 5, 6])

But if I try to get all values of a particular column of all rows (say I want values of 1st column in every row) I would try the gona[:][1]. But the result I get from this is same as before.

What can be the reason for this? How do I do such a thing in numpy?

no. I wanted know why both ways give same results too – maheshakya Feb 01 '13 at 17:39 — maheshakya, Feb 01 '13 at 17:39

score 38 · Accepted Answer · edited Jan 03 '23 at 21:00

38

You actually want to do this:

>>> a
array([[ 1,  2,  3],
       [ 4,  5,  6],
       [ 7,  8,  9],
       [10, 11, 12]])
>>> a[:,1]
array([ 2,  5,  8, 11])

a[:] just returns the entire array, so then a[:][1] is returning the second row of a. I think that's where your confusion arises.

See this section of the Tentative Numpy Tutorial for more information on indexing multidimensional arrays.

edited Jan 03 '23 at 21:00

Glorfindel

21,988
13
81
109

answered Feb 01 '13 at 16:26

John Vinyard

12,997
3
30
43

2

`a[:][1]` selects the second row – Matti Lyra Feb 01 '13 at 16:31
Oops. You're right. I've corrected that. – John Vinyard Feb 01 '13 at 16:34
1

This did the job. Thank you – maheshakya Feb 01 '13 at 16:42
What does the comma mean? – Phil Nov 24 '17 at 21:06
The comma is the axis delimiter, so in this example, I'm selecting all elements along the first axis, and only the second element along the second axis. – John Vinyard Nov 25 '17 at 15:38

Matti Lyra · Answer 2 · 2015-03-27T16:18:27.760

There seems to be a slight confusion in terms of the positioning of the braces, gona[:][1] first selects everything from the array, and from that array then selects the second row. To select particular columns you put the indices within the same square brackets separated by a comma:

gona = np.array([[ 1,  2,  3],
       [ 4,  5,  6],
       [ 7,  8,  9],
       [10, 11, 12]])

gona[1,:]
Out[21]: array([4, 5, 6])

gona[:,1]
Out[22]: array([ 2,  5,  8, 11])

gona[:,0]
Out[23]: array([ 1,  4,  7, 10])

you can also just select a range of rows for instance

gona[0:2,0] # only take the two first rows of the first column
Out[24]: array([2, 5])

thank you separately for selecting a range of rows code :) – Andrii Kovalchuk Mar 18 '15 at 20:26 — Andrii Kovalchuk, Mar 18 '15 at 20:26

score 0 · Answer 3 · answered Feb 01 '13 at 16:22

Like this:

gona = numpy.array([[ 1,  2,  3],
       [ 4,  5,  6],
       [ 7,  8,  9],
       [10, 11, 12]])

# List comprehension, just get each element in 'gona', and then get first element in that list
out = [x[0] for x in gona]

print out

Output:

>>> 
[1, 4, 7, 10]
>>>

Getting only particular columns in every row in a numpy array

3 Answers3