Converting a 3D List to a 3D NumPy array

Question

Currently, I have a 3D Python list in jagged array format.
A = [[[0, 0, 0], [0, 0, 0], [0, 0, 0]], [[0], [0], [0]]]

Is there any way I could convert

Answer 1

It we turn your list into an array, we get a 2d array of objects

In [1941]: A = [[[0, 0, 0], [0, 0, 0], [0, 0, 0]], [[0], [0], [0]]]
In [1942]: A = np.array(A)
In [1943]: A.shape
Out[1943]: (2, 3)
In [1944]: A
Out[1944]: 
array([[[0, 0, 0], [0, 0, 0], [0, 0, 0]],
       [[0], [0], [0]]], dtype=object)

When I try A+1 it iterates over the elements of A and tries to do +1 for each. In the case of numeric array it can do that in fast compiled code. With an object array it has to invoke the + operation for each element.

In [1945]: A+1
...
TypeError: can only concatenate list (not "int") to list

Lets try that again with a flat iteration over A:

In [1946]: for a in A.flat:
      ...:     print(a+1)
....
TypeError: can only concatenate list (not "int") to list

The elements of A are lists; + for a list is a concatenate:

In [1947]: for a in A.flat:
      ...:     print(a+[1])
      ...:     
[0, 0, 0, 1]
[0, 0, 0, 1]
[0, 0, 0, 1]
[0, 1]
[0, 1]
[0, 1]

If the elements of A were themselves arrays, I think the +1 would work.

In [1956]: for i, a in np.ndenumerate(A):
      ...:     A[i]=np.array(a)
      ...:     
In [1957]: A
Out[1957]: 
array([[array([0, 0, 0]), array([0, 0, 0]), array([0, 0, 0])],
       [array([0]), array([0]), array([0])]], dtype=object)
In [1958]: A+1
Out[1958]: 
array([[array([1, 1, 1]), array([1, 1, 1]), array([1, 1, 1])],
       [array([1]), array([1]), array([1])]], dtype=object)

And to get back to the pure list form, we have apply tolist to both the elements of the object array and to the array itself:

In [1960]: A1=A+1
In [1961]: for i, a in np.ndenumerate(A1):
      ...:     A1[i]=a.tolist()

In [1962]: A1
Out[1962]: 
array([[[1, 1, 1], [1, 1, 1], [1, 1, 1]],
       [[1], [1], [1]]], dtype=object)
In [1963]: A1.tolist()
Out[1963]: [[[1, 1, 1], [1, 1, 1], [1, 1, 1]], [[1], [1], [1]]]

This a rather round about way of adding a value to all elements of nested lists. I could have done that with one iteration:

In [1964]: for i,a in np.ndenumerate(A):
      ...:     A[i]=[x+1 for x in a]
      ...:     
In [1965]: A
Out[1965]: 
array([[[1, 1, 1], [1, 1, 1], [1, 1, 1]],
       [[1], [1], [1]]], dtype=object)

So doing math on object arrays is hit and miss. Some operations do propagate to the elements, but even those depend on how the elements behave.