Extract columns of numpy array consisting of lists-CodePudding

I have a variable var

When I print it in jupyter, it gives:

var

#array([list([9166855000000.0, 13353516.0]),
#       list([7818836000000.0, 11389833.0]),
#       list([20269756000000.0, 29527304.0]),
#       list([66886956000000.0, 97435384.0]),
#       list([58686560000000.0, 85489730.0]),
#       list([50809440000000.0, 74014984.0])], dtype=object)

print(var)

[list([9166855000000.0, 13353516.0]) 
 list([7818836000000.0, 11389833.0])
 list([20269756000000.0, 29527304.0])
 list([66886956000000.0, 97435384.0]) 
 list([58686560000000.0, 85489730.0])
 list([50809440000000.0, 74014984.0])]

The type is:

print(type(var))

#<class 'numpy.ndarray'>

How can I devide the second elements of the sublists by the first ones?

I want to get the following values as an array or list:

13353516.0/9166855000000.0
...
74014984.0/50809440000000.0

CodePudding user response：

The solution seems to be rather straightforward

var[:,1] / var[:,0]

would output

array([1.4567172710815215e-06, 1.4567172146851526e-06,
       1.4567172885554222e-06, 1.456717270853229e-06,
       1.4567173472086283e-06, 1.4567171769655403e-06], dtype=object)

To reproduce define var as follows:

import numpy as np
var = np.array([list([9166855000000.0, 13353516.0]),
       list([7818836000000.0, 11389833.0]),
       list([20269756000000.0, 29527304.0]),
       list([66886956000000.0, 97435384.0]),
       list([58686560000000.0, 85489730.0]),
       list([50809440000000.0, 74014984.0])], dtype=object)

CodePudding user response：

If I generate the structure like this:

import numpy as np

var = np.empty([6,], dtype=object)

ll = [list([9166855000000.0, 13353516.0]),
       list([7818836000000.0, 11389833.0]),
       list([20269756000000.0, 29527304.0]),
       list([66886956000000.0, 97435384.0]),
       list([58686560000000.0, 85489730.0]),
       list([50809440000000.0, 74014984.0])]

for i, l in enumerate(ll):
    var[i] = l

then var contains a 1D array (of lists)

array([list([9166855000000.0, 13353516.0]),
       list([7818836000000.0, 11389833.0]),
       list([20269756000000.0, 29527304.0]),
       list([66886956000000.0, 97435384.0]),
       list([58686560000000.0, 85489730.0]),
       list([50809440000000.0, 74014984.0])], dtype=object)

and the solution might be:

[var[i][1] / var[i][0] for i in range(len(var))]

returning a list:

[1.4567172710815215e-06,
 1.4567172146851526e-06,
 1.4567172885554222e-06,
 1.456717270853229e-06,
 1.4567173472086283e-06,
 1.4567171769655403e-06]

Or, indeed, a more elegant solution would be using @hpaulj's suggestion and @DmitriChubarov's solution:

var = np.stack(var)

var[:,1] / var[:,0]

returning an array:

array([1.45671727e-06, 1.45671721e-06, 1.45671729e-06, 1.45671727e-06,
       1.45671735e-06, 1.45671718e-06])