CODEWITHSUNDEEP
pythonoop
cs_stackX
cs_stackX

Reputation: 1527

Best way to call Python class methods automatically upon instantiation

I have a machine learning algorithm which involves a series of steps, such as cleaning data, preparing training data etc. Each step is stored in a separate method of a python class. I'm wondering what the best practice way to structure my class is so that the steps are automatically executed upon class instantiation.

Here's what I've done (the code is illustrative, but this approach works on the real algorithm). It feels a little clunky. Is there a more elegant way?

 class Kaggle():
    """
    An algorithm
    """

    def __init__( self ):
        self.bar = 1

    def step_one( self, some_text_data ):
        self.bar = 1 ** 2
        # Do some data cleaning
        # return processed data

    def step_two( self ):
        foo = step_one(baz)
        # do some more processing

    def step_three( self ):
        bar = step_two()
        # output results

    def run( self ):
        self.step_one()
        self.step_two()
        self.step_three()

if __name__ == "__main__":
    kaggle = Kaggle()
    kaggle.run()

Upvotes: 6

Views: 10850

Answers (3)

solly989
solly989

Reputation: 43

I was working with dataclasses and found this to work as well:

from dataclasses import dataclass

@dataclass
class MyClass:
    var1: str
    var2: str

    def __post_init__(self):
        self.func1()
        self.func2()

    def func1(self):
        print(self.var1)

    def func2(self):
        print(self.var2)

a = MyClass('Hello', 'World!')
Hello
World!

Upvotes: 0

Alexander
Alexander

Reputation: 109528

If your goal is for the object to be "automatically executed upon class instantiation", just put self.run() in __init__:

def __init__(self):
    self.bar = 1
    self.run()

As an aside, one should try to keep the __init__ method lightweight and just use it to instantiate the object. Although "clunky", your original Kaggle class is how I would design it (i.e. instantiate the object and then have a separate run method to run your pipeline). I might rename run to run_pipeline for better readability, but everything else looks good to me.

Upvotes: 11

Paul Rooney
Paul Rooney

Reputation: 21609

Put all the calls in your __init__ method. Is this not what you wanted to achieve? You could add a flag with a default value, that allows you to not run the tests if you want.

def __init__( self, runtests=True ):
    self.bar = 1
    if runtests:
        self.step_one()
        self.step_two()
        self.step_three()

Upvotes: 3

Related Questions