Suppose I have an array like:
a = np.arange(0,10)
Why does a[-1:9] give an empty result? I expected it to give a result containing a[-1], a[0], a[1], ... a[8].
Add a comment
|
1 Answer
The slice is interpreted as starting at a[-1], which is the same as a[len(a)-1], so a[-1:9] is equivalent to a[9:9], which is an empty list. Your expected result isn't a contiguous range, which is what a slice must produce.
2 Comments
Mastiff
I see. In my mind, it should expand the range first, so [-1,0,1,...,8], then alias the -1 to 9, resulting in [9,0,1,2,...,8]. All I was trying to do was recreate the np.roll() function using indexing.
hpaulj
np.roll code is complex because it can work on any axis. But basically, for your shift it does res[0]=a[-1] and res[1:] = a[:-1]. You can't do the roll with one slicing expression. slice produces a view, and a roll cannot be a view.