[FIXED] Calculate mean of coordinates of a Pandas Series

Issue

I have a pandas Series like this, name place:

0    [[[1,2], [3,4], [5,6], [7,8]]]
1    [[[3,2], [6,4], [7,6], [2,8]]]
2    nan
...

the elements of each row are the coordinates of the vertices of a box.

I would like to find the center of the box, keeping the nan values as I want to concatenate this series to a dataframe and I need the same number of rows.

My initial idea was to write a function to calculate the mean:

def centroid(arr_val):
    if arr_val:
        center = np.mean(np.array(arr_val[0]), axis=0)
    else:
        center = arr_val
    return center

and apply it to each element of the series:

place.apply(lambda x: centroid(x))

but I have some problem with the dimention of the list in the series as it's a list of list of lists...

What I would like to have is:

0   [4., 5.]
1   [4.5, 5. ]
2   nan
...

Any suggestion?

Solution

Your issue is not the dimensions of the lists, it is that you are not properly checking whether the input is nan or not. Your test passes even when arr_val is nan:

arr_val = np.nan
if arr_val:
    print('nan is truthy')

Prints nan is truthy. See Why do "Not a Number" values equal True when cast as boolean in Python/Numpy?

Changing your test - given your data it's probably simplest just to check for a list - will resolve your problem:

def centroid(arr_val):
    if isinstance(arr_val, list):
        center = np.mean(np.array(arr_val[0]), axis=0)
    else:
        center = arr_val
    return center

place = pd.Series([
    [[[1,2], [3,4], [5,6], [7,8]]],
    [[[3,2], [6,4], [7,6], [2,8]]],
    np.nan
])
place.apply(centroid)

Output:

0    [4.0, 5.0]
1    [4.5, 5.0]
2           NaN
dtype: object

Answered By - Nick