Any idea if there is a way to make the following code to work
class Test(object):
def __init__(self, var):
self.var = var
def changeme(self):
self = Test(3)
t = Test(1)
assert t.var == 1
t.changeme()
assert t.var == 3
is something like the following safe to use for more complex objects (like django models, to hot swap the db entry the instance is referring to)
class Test(object):
def __init__(self, var):
self.var = var
def changeme(self):
new_instance = Test(3)
self.__dict__ = new_instance.__dict__
t = Test(1)
assert t.var == 1
t.changeme()
assert t.var == 3
self = Test(3)
is re-binding the local nameself
, with no externally observable effects.Assigning
self.__dict__
(unless you're talking about instances with__slots__
or from classes with non-trivial metaclasses) is usually OK, and so isself.__init__(3)
to re-initialize the instance. However I'd prefer to have a specific methodself.restart(3)
which knows it's being called on an already-initialized instance and does whatever's needed to cater for that specific and unusual case.No, and no.
Having said that, you can change the class, but don't do that either.
The former code works, except it won't do much, seeing as it just replaces the object named 'self' within the scope of changeme(). Python names aren't locked to values, they are always relative to their scope or namespace.
To do what you want you'd need to have access to a name outside the class, which you could assign to from within it:
This basically just overwrites the name 'myclass' to point to the new instance. It doesn't "overwrite" the first instance like you might think. The old instance still lives, and will be garbage collected unless referenced elsewhere.