Python init return failure to create

Question

First off, I know that the __init__() function of a class in Python cannot return a value, so sadly this option is unavailable.

Due to the structure of my code, it makes sense to have data assertions (and prompts for the user to give information) inside the __init__ function of the class. However, this means that the creation of the object can fail, and I would like to be able to gracefully recover from this.

I was wondering what the best way to continue with this is. I've considered setting a global boolean as a 'valid construction' flag, but I'd prefer not to.

Any other ideas (besides restructuring so assertions can happen outside of the initialization and values are passed in as arguments)? I'm basically looking for a way to have return 0 on success and return -1 on failure during initialization. (Like most C system calls)

Returning 0 on success and -1 on failure is very un-Pythonic. Error returns are bad, because they're easy to ignore, and verbose to handle. On top of that, using 0 and -1 as boolean values just adds confusion with no benefit. Don't true to use Python to write C code. — abarnert, Jun 27 '13 at 01:23
As @jsbueno points out in his answer, class instances are _initialized_ in `__init__()`, not created, which is why it doesn't `return` a value. Instead of using a global variable to detect failures doing this or throwing an exception, you could add a data member to your class to indicate its state and then check its value directly or call a method that returns its value after `__init__()` has been called to determine if anything (and possibly what) when wrong. — martineau, Jun 27 '13 at 04:07

jsbueno · Accepted Answer · 2016-03-22T20:58:50.223

16

You could raise an exception when either assertio fail, or -, if you really don't want or can't work with exceptions, you can write the __new__ method in your classes - in Python, __init__ is technically an "initializer" method - and it should fill in he attributes and acquire some of the resources and others your object will need during its life cycle - However, Python does define a real constructor, the __new__ method, which is called prior to __init__- and unlike this, __new__ actually does return a value: the newly created (uninitialized) instance itself.

So you can place your checks inside __new__ and and simply return None is something fails - otherwise, return the result of the call to the superclass __new__ method (one can't do the actual memory allocation for the object in pure Python, so ultimately you have to call a constructor written in native code in a superclass - usually this is object.__new__ in the base of your class hierarchy.

NB: In Python 2, you must have object as the base class for your hierarchy - otherwise not only __new__ is not called, as a whole lot of features added later to Python objects will just not work. In short, class MyClass(object): , never class MyClass: - unless you are on Python3.

edited Mar 22 '16 at 20:58

answered Jun 27 '13 at 01:13

jsbueno

99,910
10
151
209

You say `__new__` is called before `__init__`. I just tried this, but `__new__` is NOT called when I create an instance of a class (Py2.7). – qwerty_so Mar 21 '16 at 15:53
I found it. You need to extend object in your class. Without that, new will not be called! – qwerty_so Mar 22 '16 at 10:40
No. If I omit the object base class,`__new__` is not called automatically. Go figure! – qwerty_so Mar 22 '16 at 17:10
No and again no. My print is not called, unless I add the object base class. – qwerty_so Mar 22 '16 at 17:13
Hehe. Not only me falling in traps ;-) Would you add a note to your answer, so we can delete this discussion? – qwerty_so Mar 22 '16 at 17:36

score 2 · Answer 2 · answered Jun 27 '13 at 01:07

2

Have you considered raising an exception? That is the usual way to signal a failure.

answered Jun 27 '13 at 01:07

D.Shawley

58,213
10
98
113

Andrew Leason · Answer 3 · 2022-01-26T15:33:38.100

0

What about writing a @classmethod function? You can then check all the conditions and either create and initialize a new class, or return None. The calling function then just has to check if the returned class is not None.

edited Jan 26 '22 at 15:33

answered Jan 26 '22 at 15:24

Andrew Leason

1
1

Hong승원 · Answer 4 · 2023-01-07T14:50:15.207

0

I returned __init__ using some tricks.

This is how to set the error to 'try' by making an error in __init__.

Below is an example code.

import sys

class MyClass(object):
    def __init__(self, test=None):
        if test is None:
            sys.exit(1)

        self.test = test
        print("do something __init__ here")

try:
    obj = MyClass()
except:
    obj = None

print(obj)

edited Jan 07 '23 at 14:50

answered Jan 07 '23 at 14:44

Hong승원

122
6

Python __init__ return failure to create

4 Answers4

Python init return failure to create