I have an array a = np.arange(10). I want to use np.roll, where the shift can be positive or negative. I then want to use indexing to remove the elements which have been padded to the beginning (if the shift is positive) or end (if the shift is negative) of the array. So for example:
>>> a = np.arange(10)
>>> np.roll(a, 1)[1:]
array([0, 1, 2, 3, 4, 5, 6, 7, 8])
>>> np.roll(a, -1)[:-1]
array([1, 2, 3, 4, 5, 6, 7, 8, 9])
Is there a generic way to accomplish this in a single line?
I don't think its possible to do this in a generic (not hard-coded way). But its pretty easy to write a short function that does what you want.
def roll2(arr, i):
"""Roll and clip array"""
if i == 0:
return arr
elif i > 0:
s = slice(i, None)
else:
s = slice(None, i)
return np.roll(a, i)[s]
Why are you even using roll here? The whole point of roll is to create wrap-around behaviour, which is exactly the feature that you seem to be trying to eliminate. A simple slice is sufficient:
def shift_right(arr, i):
if i <= 0:
return arr[-i:]
else:
return arr[:-i]
This function behaves exactly as the one in David's answer
padding?