I want to do an assignment to a multidimensional array where each element of the array is 3 short integers:
a = ndarray([3,3,3], dtype='u2,u2,u2')
a[2,2,2] = [1,2,3]
Traceback (most recent call last): File "", line 1, in a[2,2,2] = [1,2,3] TypeError: expected a readable buffer object
I am going to be using a large array and would like to get direct indexing into the array for performance. What is a good way to do this in python?
Thanks for any insight into how to do this?
The elements of an array of dtype='u2,u2,u2' are a tuple of three shorts, not a list of three shorts. So:
a[2,2,2] = (1,2,3)
(The parens are of course not necessary, but I used them to make it obvious this is a tuple.)
You can also pass it an array if you want:
a[2,2,2] = np.array([1,2,3])
Of course the error message here certainly could be better… What it's actually complaining about is something deeper than what you'd expect, and it doesn't help you debug the problem.
a[x,y,z] = tuple(foo) is faster or slower than a[x,y,z] = np.array(foo); if it matters, test it. But I doubt it will matter. If it matters, it's probably because you're doing it in a big loop—in which case the fact that you're looping in Python instead of vectorizing is the real problem you need to solve.