Python – instance attribute defined outside __init__()

Question:

I get the warning in question for the following code:

from tkinter import *
from tkinter import ttk


class Autocomplete(Frame, object):
    def __init__(self, *args, **kwargs):
        super(Autocomplete, self).__init__(*args, **kwargs)
        self.list = []

    def build(self, width, height, entries):
        # I get the warning for the following 8 lines:
        self._entries = entries
        self.listbox_height = height
        self.entry_width = width
        self.text = StringVar()
        self.entry = ttk.Entry(self, textvariable=self.text, width=self.entry_width)
        self.frame = Frame(self)
        self.listbox = Listbox(self.frame, height=self.listbox_height, width=self.entry_width)
        self.dropdown = Listbox(self.frame, height=self.listbox_height, width=self.entry_width, background="#cfeff9",
                                takefocus=0)
        self.entry.pack()
        self.frame.pack()
        self.listbox.grid(column=0, row=0, sticky=N)
        self.dropdown.grid(column=0, row=0, sticky=N)
        self.dropdown.grid_forget()
        return self
root = Frame(Tk())
autocomplete = Autocomplete(root).build(74, 10, entries)
root.pack()
autocomplete.pack()
mainloop()

How should I fix this? I tried to move everything to init but then I had some errors for passing the arguments in the line I was creating the Autocompelete object. So please provide me with all the changes I have to make. not just smth like you have to move them.
I could fix the warning by adding 8 definition lines assigning None to all the variables but I think that’s such a stupid solution. So what is the right thing to do?

Asked By: yukashima huksay

||
Source

Answers:

It’s always important to keep in mind that not all Warnings require fixing. Warnings are just Warnings. They are supposed to point out a specific part of the code because it’s a “common” source of problems. But sometimes you need/want to do it that way.

I could fix the warning by adding 8 definition lines assigning None to all the variables

That’s just “silencing” the Warnings, in my opinion that’s just as good as ignoring the Warnings.

So what is the right thing to do?

The right way would be to just use __init__. I did a quick test and I don’t have any problems.

However that’s just an example how one could do it. I haven’t checked what Frame wants as arguments for __init__ so it could lead to conflicts:

from tkinter import *
from tkinter import ttk

class Autocomplete(Frame, object):
    def __init__(self, *args, **kwargs):
        width, height, entries = kwargs.pop('width'), kwargs.pop('height'), kwargs.pop('entries')
        super(Autocomplete, self).__init__(*args, **kwargs)
        self.list = []
        self._entries = entries
        self.listbox_height = height
        self.entry_width = width
        self.text = StringVar()
        self.entry = ttk.Entry(self, textvariable=self.text, width=self.entry_width)
        self.frame = Frame(self)
        self.listbox = Listbox(self.frame, height=self.listbox_height, width=self.entry_width)
        self.dropdown = Listbox(self.frame, height=self.listbox_height, width=self.entry_width, background="#cfeff9",
                                takefocus=0)
        self.entry.pack()
        self.frame.pack()
        self.listbox.grid(column=0, row=0, sticky=N)
        self.dropdown.grid(column=0, row=0, sticky=N)
        self.dropdown.grid_forget()

root = Frame(Tk())
autocomplete = Autocomplete(root, width=74, height=10, entries=entries)
root.pack()
autocomplete.pack()
mainloop()
Answered By: MSeifert

I would also add that as of python 3.3, this also hints about missing an optimization feature for each instance’s __dict__ when any attribute is set outside of the __init__ method.

As stated in Fluent Python 2nd Edition, Chapter 3, page 103:

Python’s default behavior is to store instance attributes in a special __dict__ attribute, which is a dict attached to each instance. Since PEP 412—Key-Sharing Dictionary was implemented in Python 3.3, instances of a class can share a common hash table, stored with the class. That common hash table is shared by the __dict__ of each new instance that has the same attributes names as the first instance of that class when __init__ returns. Each instance __dict__ can then hold only its own attribute values as a simple array of pointers. Adding an instance attribute after __init__ forces Python to create a new hash table just for the __dict__ of that one instance (which was the default behavior for all instances before Python 3.3). According to PEP 412, this opti‐ mization reduces memory use by 10% to 20% for object-oriented programs.

There’s further explanation of the internal optimizations at FluentPython: Internals of sets and dicts

Answered By: radyz