Numpy array remains unchanged even after assignment

Question

Suppose I have the matrix x. x has 3 columns and arbitrarily many rows. I want to grab all the rows that have a certain value and change this value for all these rows. Below is an example as well as my attempt to do this.

x = np.array(
 [[-3.1913035 , -1.34344639, 0],
 [-2.54438272, -1.88907741, 1],
 [ 2.12029563, 2.51443883, 3],
 [-2.98150865, -1.53789653, 3],
 [-1.94179128, -3.1429703, 3 ]])
x[x[:,2] == 3][:,2] = 5 # the values 3 and 5 are strictly examples.
print(x)

Printing x on the last line displays an unchanged matrix. I expect x to look like

np.array(
 [[-3.1913035 , -1.34344639, 0],
 [-2.54438272, -1.88907741, 1],
 [ 2.12029563, 2.51443883, 5],
 [-2.98150865, -1.53789653, 5],
 [-1.94179128, -3.1429703, 5 ]])

Can I please have some help? I have been searching up this problem for hours but I have not found the solution.

• Jérôme Richard: x[x[:,2] == 3][:,2] does not do what you think it does: x[x[:,2] == 3] creates a new temporary array and not a view. It is not possible for Numpy do return a view because it would not be efficient. You need to use np.where or np.argwhere so to create a new array based on the x[:,2] == 3 result.

Answer 1

A way to do what you want is to grab only the column you want to change and then change it.

y = x[:,2]
y[y==3] = 5

• Kookie: Wow, that is very unintuitive, but it makes sense.
• Stef: or (less readable) as x[:,2][x[:,2] == 3] = 5

Answer 2

If you want to replace all elements that matches 3 then this might be helpful.

x[x == 3] = 5

This might be useful to specify column criteria in

np.ix_(row_criteria, col_criteria)

x[np.ix_(x[:,2]==3, [2])] = 5

• Kookie: I think this is a good solution, but not for my purpose (which I should have further specified) because I don't want to edit the first two columns