Views and Copies

Basic slicing, such as A[0, :], returns a view. A view shares data with the original array.

That means changing the view can change the original.

Here, row is a view, so A changes too.

This behavior is useful for performance, but it can surprise you.

Use copy when you need separation

If you want a separate array, use .copy():

row = A[0, :].copy()

Now changing row does not change A.

The habit

When you slice and then mutate, ask:

Do I want this change to affect the original array?

If yes, a view can be fine. If no, copy first.

This is the same idea as list copying, but it matters more with arrays because large numerical data should not be copied accidentally.

Not every selection behaves like a basic slice. Some later NumPy operations return copies instead. Do not try to memorize every case yet; when mutation matters, use .copy() to state clearly that you want independent data.

row = A[0, :].copy()