Say I have the following np.array():

[1, 2, 3, 4, 5, 6, 7, 8, 9]

I would like to tile the values across with an offset. In this case we group by three and offset by 1:

[
  [1, 2, 3],
  [2, 3, 4],
  [3, 4, 5],
  [4, 5, 6],
  [5, 6, 7],
  [6, 7, 8],
  [7, 8, 9],
]

Is there a built in function to achieve this that leverages C internals of numpy? I'm working with very long arrays and using standard array manipulation with loops has been prohibitively slow.

CodePudding user response：

Based on the explanation in the question, if the expected array be as (not that is written now without [5, 6, 7], [6, 7, 8]):

[[1 2 3]
 [2 3 4]
 [3 4 5]
 [4 5 6]
 [5 6 7]
 [6 7 8]
 [7 8 9]]

The easiest way is mentioned by @ Michael Szczesny.
But, based on one of SO answers and by modifying that to adapt with this question:

a = np.array([1, 2, 3, 4, 5, 6, 7, 8, 9])
slice_ = 2
b = np.tile(a, (len(a) - slice_, 1))
rolling_val = np.arange(len(b))   slice_   1
rows, column_indices = np.ogrid[:b.shape[0], :b.shape[1]]
column_indices = column_indices - rolling_val[:, np.newaxis]
result = b[rows, column_indices]
result = result[::-1][:len(a) - slice_, :slice_   1]

CodePudding user response：

You can use numpy.lib.stride_tricks.sliding_window_view like below:

numpy.lib.stride_tricks.sliding_window_view is New in version 1.20.0. you need update your numpy if you numpy didn't update.

arr = [1, 2, 3, 4, 5, 6, 7, 8, 9]
np.lib.stride_tricks.sliding_window_view(arr , 3)

Output:

array([[1, 2, 3],
       [2, 3, 4],
       [3, 4, 5],
       [4, 5, 6],
       [5, 6, 7],
       [6, 7, 8],
       [7, 8, 9]])