Question

I have a large numpy array (8 by 30000) and I want to delete some rows according to some criteria. This criteria is only applicable in one column.

Example:

>>> p = np.array([[0, 1, 3], [1 , 5, 6], [4, 3, 56], [1, 34, 4]])
>>> p
array([[ 0,  1,  3],
   [ 1,  5,  6],
   [ 4,  3, 56],
   [ 1, 34,  4]])

here I would like to remove every row in which the value of the 3rd column is >30, ie. here row 3.

As the array is pretty large, I’d like to avoid for loops. I thought of this:

>>> a[~(a>30).any(1), :]
array([[0, 1, 3],
   [1, 5, 6]])

But there, it obviously removes the two last rows. Any ideas on how to do that in a efficient way?

Answer 1

p = p[~(p[:,2] > 30)]

or (if your condition is easily inversible):

p = p[p[:,2] <= 30]

returns

array([[ 0,  1,  3],
       [ 1,  5,  6],
       [ 1, 34,  4]])