Accessing variables from one class to another in Python

Question

I've got couple of huge tables with data called bnds.data and densities.data . I've also got a class processing those data tables. That class is called in a loop and in order to avoid repetitive and time demanding loading those data tables into memory, I want to creat another class and initialise some variables in there. Here's a piece of code giving the idea of what I want to do:

import numpy as np

class Pixel:
    def  __init__(self):
        self.bounds = np.loadtxt('bnds.data')
        self.rho = np.loadtxt('densities.data')

class Density(Pixel):

    def __init__(self, pixel, phi, theta):
        self.phi = phi
        self.theta = theta
        latitude =  int(90 - self.phi +1)
        longitude = int(180 + self.theta + 1)
        n = (latitude -1)*360 + longitude -1
        self.rho = pixel.bounds[n]

    def print_rho(self):
        print (self.rho)

pixel = Pixel()
rho = Density(pixel, 10, 20) # phi = 10, theta = 20
rho.print_rho()

Here , the instance of Pixel is sent to class Density. The data loading is done in the Pixel class. The density class will be called in a loop. What I don't quite understand is whether the Pixel class will be initialized every time the Density class is called? If yes then how to avoid it? My guess is that the Pixel class is initialized once and for all regardless of how many time the Density class is called. Is it correct? phi and theta are variables and they take different values in a loop. What I need is the bnds.data & densities.data tables to be loaded once and for all.

Answer 1

In general - yes, with current code, Pixel's __init__ is only executed once. However, there is no need for Density to be a subclass of Pixel, it shares no resemblance or functionality.

A more sensible solution would be to have a method on Pixel, which returns a Density, which is required for initialization of Density, eg:

class Pixel:
    def  __init__(self):
        self.bounds = np.loadtxt('bnds.data')
        self.rho = np.loadtxt('densities.data')

    def get_density(self, phi, theta):
        return Density(self, phi, theta)

pixel = Pixel()
rho = pixel.get_density(10,20) # phi = 10, theta = 20
rho.print_rho()

That way you make your code more readable.

Note that classes can be quite expensive if performance is critical. If you're only calculating rho, you could also turn the whole Density class to a get_rho(phi, theta) method avoiding instantiating a new class at all.

Answer 2

First off, when you are using inheritance you don't need to pass an instance of parent class to child class. That's what you are using inheritance.

You can simply access to your parent class attributes within your child class after calling the parent constructor using super() in pyhton 3 and new_style classes in python 2 or directly call it in old_style classes. And then you'll be sure that after each instantiate bnds.data and densities.data tables will be loaded once.

class Pixel:
    def  __init__(self):
        self.bounds = np.loadtxt('bnds.data')
        self.rho = np.loadtxt('densities.data')

class Density(Pixel):
    def __init__(self,phi,theta):
        super(Density, self).__init__()
        self.phi = phi
        self.theta = theta
        latitude =  int(90 - self.phi +1)
        longitude = int(180 + self.theta + 1)
        n = (latitude -1)*360 + longitude -1
        self.new_rho = self.bounds[n]
    def print_rho(self):
        print (self.new_rho)

rho = Density(10,20) # phi = 10, theta = 20
rho.print_rho()