Python List Multiplications — It’s a Trap!

An odd case I encountered while working on my project. I need a 3x3 grid represented by 2-dimensional array. To avoid typing the whole array by hand, list multiplications immediately came to mind.

So instead of writing:

arr = [[1, 1, 1], 
       [1, 1, 1], 
       [1, 1, 1]]

I initialized the array as follows:

arr = [[1] * 3] * 3

The list multiplication seemed so neat and simple at the time, until I ran into a problem later when trying to modify the values inside.

Changing value at 2nd row, 2nd column to zero: arr[1][1] = 0 , I was hoping to modity the grid as follows

arr = [[1, 1, 1], 
       [1, 0, 1], 
       [1, 1, 1]]

However, the array became as follows:

[[1, 0, 1], 
 [1, 0, 1], 
 [1, 0, 1]]

Explanation

The assignment arr[1][1] = 0 causing each row of the grid to change is because * in creation of the array is copying the address of the first dimension [1, 1, 1] . Instead of copying the items, list multiplication replicates references to the same item: a shallow copy. The list multiplication actually works in the first dimension but not the second:

lst = [1] * 3
lst[0] = 0
lst
>>> [0, 1, 1]

Other Possible Implementations

• Nested for-loops

arr = [[1 for i in range(3)] for j in range(3)]

• For loop with list multiplication in inner lists

arr = [[1] * 3 for i in range(3)]

• List multiplication but with deep copy for inner lists, here lst[:] is used to obtain a deepcopy of the original array hence the return array won’t be affected by any changes in the original objects.

arr = [x[:] for x in [[1] * 3] * 3]

The 3rd method list multiplication turns out to be the fasted among all three, method 2 quite similar and method 1 being the slowest:

Comparing the operations for creating a much larger grid, method 2 and 3 are clearly much more efficient than the first as well.

Final Note

Beware of the difference and usages for shallow & deep copy as a wrong copy in data might lead to series error in production. For more on shallow and deep copy, refer to my previous article under Section Copy: Shallow vs Deep.

Thanks for reading and stay tuned for more articles on Python!