Given arr1 = [[0,0,0,0],[0,0,0,0],[0,0,0,0] and arr2 = [[1,2],[3,4],[4,5]] and positions = [0,1,2] we wish to add elements of arr2 to elements of elements of arr1 at the indices specified by postions that is our answer should be [[1,2,0,0],[0,3,4,0],[0,0,4,5]] Note that we guarantee beforehand that the position would be valid.
A simple for loop over the outer array solution works but is pretty slow for my application. I tried using splicing and specifing positions but nothing seems to work.
You haven't really defined your existing solution, but the simple, straightforward way would be something like:
>>> arr1 = [[0,0,0,0],[0,0,0,0],[0,0,0,0]]
>>> arr2 = [[1,2],[3,4],[4,5]]
>>> positions = [0,1,2]
>>> for sub, items, pos in zip(arr1, arr2, positions):
... sub[pos: pos + len(items)] = items
...
>>> arr1
[[1, 2, 0, 0], [0, 3, 4, 0], [0, 0, 4, 5]]
If mutating the sublists in arr1 is fine, this should be pretty much the fastest solution in vanilla python.
If making it with numpy you'd still need to arrange indices (positions) and values (arr2) shapes. The following is solution using np.put_along_axis routine:
arr1 = np.array([[0,0,0,0],[0,0,0,0],[0,0,0,0]])
arr2 = [[1,2], [3,4], [4,5]]
pos = [0, 1, 2]
np.put_along_axis(arr1, np.array([np.arange(p, len(r)+p)
for p, r in zip(pos, arr2)]), arr2, axis=1)
print(arr1)
[[1 2 0 0]
[0 3 4 0]
[0 0 4 5]]
