Maintaining NaN value when going from python list to numpy array?

Question

In my code, I must go from a large nested list (python list of python lists) and in each sublist, some entries may be the numpy values NaN. When I create this nested list it looks something like this:

import numpy as np
nested_list = [[np.nan, 19], ['a', np.nan]]
>>> print(nested_list)
[[nan, 19], ['a', nan]]

I expect when I evaluate the equality of a NaN element of the sublist against itself, I should get False due to it being a null value which I do:

>>>print(nested_list[1][1] == nested_list[1][1])
False

Now, I want to turn this nested list into a 2d numpy array, but when I do, the NaN values just turn into strings and do not retain their null-ness:

arr = np.array(list_of_lists)
>>>print(arr)
[['nan' '19']
 ['a' 'nan']]
>>>print(arr[1][1] == arr[1][1])
True

How do I keep NaN from being turned into a string?

• hpaulj: Why do you want a numpy array? Why not stick with the list, where a mix of types is natural.
• cameronpoe: @hpaulj I had some previous code written when the object I'm dealing with is a numpy array from the start, but the code I'm writing now necessitates me starting with a list, so I wanted to convert back to what the code already knew how to deal with.
• hpaulj: Just beware that an object dtype array is quite different from a regular numeric array.

Answer 1

You can do it like this

import numpy as np
nested_list = [[np.nan, 19], ['a', np.nan]]
np.array(nested_list, dtype='object')