How can I create a numpy array with multiple fields per index?

There are at least two ways which look similar but are actually quite different:

The first way is to create a two dimensional array:

import numpy as np

foo = np.array([[0,1],[2,3],[4,5],[6,7]])
# foo = np.arange(4*2).reshape(4,2)   # this does the same thing
print(foo[0])    # the first row of `foo`
# [0 1]

print(foo[1])    # the second row of `foo`
# [2 3]

The second way is to create a one-dimensional array of dtype 'object', where the objects are Python lists:

bar = np.empty(4, dtype='object')
bar[0] = [0,1]
bar[1] = [2,3]
bar[2] = [4,5]
bar[3] = [6,7]

print(bar[0])
# [0, 1]

print(bar[1])
# [2, 3]

Note that in the first example, foo[0] is a NumPy array. In the second example, bar[0] is a Python list.

Numerical calculations done on foo will tend to be much quicker than similar operations done on bar. If the items stored in the Python lists are numeric and/or of a homogenous type (such as all strings), using a higher-dimensional NumPy array tends to be better choice than a NumPy array of dtype object. (Especially for numeric data, not only are there speed benefits to using homogenous arrays of non-object dtypes, but moreover, some NumPy functions do not work on arrays of object dtype.)

What type are 'item1' and 'item2'? Also are they the same type? See the example below.

your_array = [[5,3], [3,6], [4, 9]] # you could do this
print(your_array[0]) # would print [5,3]