Initialize different classes in a similar way

Question

I have different Classes which I need to initialize partially in almost the same way. The init function takes in input a file path and/or some numpy data.

class InertialData1:
   def __init__(self, file=None, data=None):
      # --- this is the (almost) common part ---
      if file is None:
        self.file = 'Unknown path to file'
      else:
        self.file = file
      assert isinstance(self.file, str)

      if data is None:
        self.data = read_logfile(file, 'a_token')
      else:
        self.data = data
      assert isinstance(self.data, np.ndarray)
      # --- end of the common part ---

      # Here other stuff different from class to class

The token in the read_logfile() function (an external function) also changes from class to class.

class InertialData2:
   def __init__(self, file=None, data=None):
      # here the common code as above, but with 'another_token' instead of 'a_token'
      # here other stuff
...

Writing the same lines in all the classes definition is a no go, of course.

Possible solution.

I thought about defining an external function like this

def initialize(file, data, token):
    if file is None:
        file = 'Unknown path to file'
    else:
        file = file
    assert isinstance(file, str)

    if data is None:
        data = read_logfile(file, token)
    else:
        data = data
    assert isinstance(data, np.ndarray)
return file, data

Then I could use it inside the init method of each class like this:

class InertialData1:
   def __init__(self, file=None, data=None):
        self.file = file
        self.data = data
        self.file, self.data = initialize(self.file, self.data, token='a_token')

This way seems to be working. I just have the feeling it is not the best way of doing it, or not pythonic, and I hope to learn something from your answers :) Thanks

Feliks Montez · Accepted Answer · 2017-08-16T16:46:52.373

1

You should be able to DRY (Don't Repeat Yourself) up your code by using class inheritance:

class InertialData:
  def __init__(self, file=None, data=None):
    # --- this is the common part ---
    if file is None:
      self.file = 'Unknown path to file'
    else:
      self.file = file
    assert isinstance(self.file, str)

  def init_data(self, token, data):
    if data is None:
      self.data = read_logfile(self.file, token)
    else:
      self.data = data
    assert isinstance(self.data, np.ndarray)

class InertialData1(InertialData):
  def __init__(self, file=None, data=None):
    # --- this is the different part ---
    super().__init__(file=file, data=data);
    self.init_data('token_a', data)

Notice a few things:

You can make a "common" class, InertialData, that has the repeated functionality. Then All of your custom classes can inherit from the common one. See how the class InertialData1 takes InertialData as a parameter?

In all classes that inherit from InertialData, you can call super and this will call the __init__ method of the superclass, which in this case is InertialData.

The bit of code that initializes the data that has the differing token for each class has been pulled into a function that takes the token as a parameter. Now you can call initialize_data('token_a') in the __init__ function of any class that inherits from InertialData.

I renamed your classes to begin with capital letters (as is Python's convention).

edited Aug 16 '17 at 16:46

answered Aug 11 '17 at 13:51

Feliks Montez

548
3
15

Thanks for your reply. However it doesn't seem to work like this. TypeError: super does not take keyword arguments – Robyc Aug 14 '17 at 09:36
Sorry, I forgot that `super` automatically reuses the arguments that are passed in. I also realized that you will need to pass `data` into the `init_data` method, so I added that code too – Feliks Montez Aug 14 '17 at 17:55
`super` alone is also not working. `super().__init__(file, data)` is working. However, I'm unsure it's a good practice to call the base init function like this. – Robyc Aug 16 '17 at 14:32
I see my mistake now. I was calling super like you do in the language Ruby... Yes, it is conventional to use `super().__init__()` in python 3. See [this](https://stackoverflow.com/questions/10482953/python-extending-with-using-super-python-3-vs-python-2) answer. You will still want to use your keyword arguments though. I updated my answer accordingly. – Feliks Montez Aug 16 '17 at 16:45
@Robyc, does this solution seem better now? – Feliks Montez Aug 17 '17 at 20:53

Initialize different classes in a similar way

1 Answers1