Question

I’m trying to make a simple derived class based on str, with the addition of an instance variable, flag. For reasons I do not understand, I get an error if I try to pass the flag to the constructor:

>>> class Strvalue(str):
        def __init__(self, content, flag=None):
                str.__init__(self, content)
                self.flag = flag

>>> Strvalue("No problem")
'No problem'

>>> Strvalue("Problem", flag=None)
Traceback (most recent call last):
  File "<pyshell#113>", line 1, in <module>
    Strvalue("Problem", flag=None)
TypeError: str() takes at most 1 argument (2 given)

I’ve checked that in the succesful calls, the Strvalue constructor really does get called– I haven’t mistyped __init__ or something of that sort. So what is going on?

Edit: According to this question (and @Martijn’s answer), the problem is avoided by overriding __new__ as well. The question was why this was happening.

Answer 1

You need to use __new__ instead of __init__ when subclassing str, see basic customization.

>>> class Strvalue(str):
...     def __new__(cls, content, flag=None):
...         inst = str.__new__(cls, content)
...         inst.flag = flag
...         return inst
... 
>>> Strvalue('foo', True)
'foo'
>>> foo = Strvalue('foo', True)
>>> foo
'foo'
>>> foo.flag
True

Your code doesn’t override str.__new__, so the original str.__new__ constructor is called with your two arguments, and it only accepts one.

str objects are immutable, they construct a new instance in __new__, which then cannot be changed anymore; by the time __init__ is called, self is an immutable object, so __init__ for str doesn’t make sense. You can still also define an __init__ method, but since you already have __new__, there is really no need to divide the work up across two methods.