python 字典到 numpy 数组切片

Question

I have a dictionary.

cars_dict= { 'cars_per_cap': [809, 731, 588, 18, 200, 70, 45], 'country': ['United States','Australia','Japan','India','Russia','Morocco','Egypt'], 'drives_right': [True, False, False, False, True, True, True] }

I changed it into NumPy by doing

cars_numpy = np.array(cars_dict)

cars_numpy = array({'cars_per_cap': [809, 731, 588, 18, 200, 70, 45], 'country': ['United States', 'Australia', 'Japan', 'India', 'Russia', 'Morocco', 'Egypt'], 'drives_right': [True, False, False, False, True, True, True]}, dtype=object)

I want to do slicing and get 'United states' from it

print(cars_numpy[1][0]) is not working.

error I get is this

---

IndexError                                Traceback (most recent call last)
<ipython-input-121-54a4195f7446> in <module>
----> 1 print(cars_numpy[1][0])

IndexError: too many indices for array

Answer 1

Dictionary is not a list, you cant pass it to numpy and expect the same results, its not ordered, it uses an internal data structure in order to be efficient.

What you could do is:

cars_numpy = np.array(list(cars_dict.values()))

Which will put them into an array, but you cannot guarantee the order between the keys using this method, eg the row of country might be before cars_per_cap .

Answer 2

I have a few things to suggest -

1. You need a nested list instead of a dictionary to create a multi-dim NumPy array. This is available using list(dictionary.values()) .
2. Using np.array(dictionary) will give you a NumPy array with a single entry that holds the dict. Therefore the error IndexError: too many indices for array because you are asking for a row and column, but it only has a single element at arr[0]
3. arr[1][0] is a highly inefficient way of using numpy. Instead, try arr[1,0]
4. Dictionaries are now insertion ordered. As of Python 3.6, for the CPython implementation of Python, dictionaries remember the order of items inserted. Changing them to numpy arrays (their values() ) will retain this order.

Therefore this would work -

np.array(list(cars_dict.values()))[1,0]

'United States'