Replacing elements in multidimensional numpy array based on condition

Question

I am puzzled why this is not working:

a = np.array([[1,2,3], [1,20, 30], [4, 5, 6]])

a[a[:, 0] == 1][:, 2] = [9999, 9999]

I expect:

a = [[1, 2, 9999], [1, 20, 9999], [4, 5, 6]]

But nothing changes. What is wrong?

AFAIK, u = a[a[:, 0] == 1] copies the data because a[:, 0] == 1 is an advanced index. Next, u[:, 2] indexes into that copy, so assignment affects the copy, not the original array. — ForceBru
– ForceBru, Commented Sep 19, 2022 at 13:27

mozway · Accepted Answer · 2022-09-19 13:29:59Z

1

As explained in the comments, you're currently updating an in-memory copy, not your original array.

You can use direct indexing:

a[a[:,0]==1, 2] = 9999

or with numpy.where (not necessary here):

a[np.where(a[:,0]==1)[0], 2] = 9999

updated a:

array([[   1,    2, 9999],
       [   1,   20, 9999],
       [   4,    5,    6]])

answered Sep 19, 2022 at 13:29

mozway

267k13 gold badges56 silver badges106 bronze badges

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

1 Answer 1