Reconcile np.fromiter and multidimensional arrays in Python
By itself, np.fromiter only supports constructing 1D arrays, and as such, it expects an iterable that will yield individual values rather than tuples/lists/sequences etc. One way to work around this limitation would be to use itertools.chain.from_iterable to lazily 'unpack' the output of your generator expression into a single 1D sequence of values:
import numpy as npfrom itertools import chaindef fun(i): return tuple(4*i + j for j in range(4))a = np.fromiter(chain.from_iterable(fun(i) for i in range(5)), 'i', 5 * 4)a.shape = 5, 4print(repr(a))# array([[ 0, 1, 2, 3],# [ 4, 5, 6, 7],# [ 8, 9, 10, 11],# [12, 13, 14, 15],# [16, 17, 18, 19]], dtype=int32)