Problem when subclass instance changes base class instance variable

[Gerry Sutton]

Hi All!

I have noticed a strange behavior when using a constant identifier to
initialize an instance list variable in a base class and then trying to
modifying the list in subclasses by using either the list.extend method or
even by having the subclass create a whole new list in the variable.

The following example illustrates the situation.

Lbase = ['Initial Base Data']

LSub1 = ['SubClass 1 data']

LSub2 = ['SubClass 2 data']

##

class BaseClass:

def __init__(self):

self.data = Lbase #<---- this fails??

# self.data = ['Initial Base Data'] #<---- this works

def printData(self):

print self.data

##

class SubClass1(BaseClass):

def __init__(self):

BaseClass.__init__(self)

self.data.extend(LSub1)

# self.data += LSub1

##

class SubClass2(BaseClass):

def __init__(self):

BaseClass.__init__(self)

self.data.extend(LSub2)

# self.data += LSub2

s1 = SubClass1()

s1.printData()

s2 = SubClass2()

s2.printData()

s11 = SubClass1()

s11.printData()

s21 = SubClass2()

s21.printData()

s1.printData()

['Initial Base Data', 'SubClass 1 data']

['Initial Base Data', 'SubClass 1 data', 'SubClass 2 data']

['Initial Base Data', 'SubClass 1 data', 'SubClass 2 data', 'SubClass 1
data']

['Initial Base Data', 'SubClass 1 data', 'SubClass 2 data', 'SubClass 1
data', 'SubClass 2 data']

['Initial Base Data', 'SubClass 1 data', 'SubClass 2 data', 'SubClass 1
data', 'SubClass 2 data']

It looks like the instance variable is acting like a class variable with all
the subclass changes accumulating as more objects are created.

Does anyone have an explanation for this behavior?

Are other mutable sequence types affected the same way?

To fix the problem you simply use a literal list to initialize the variable
in the base class instead of an identifier.

This problem does not exist if the data variable is a string, maybe because
it is immutable?

I am using Python Ver 2.3.

Does this code behave the same way in 2.4,

[Peter Otten]

Gerry Sutton wrote:

I have noticed a strange behavior when using a constant identifier to
initialize an instance list variable in a base class and then trying to
modifying the list in subclasses by using either the list.extend method or
even by having the subclass create a whole new list in the variable.

The following example illustrates the situation. Lbase = ['Initial Base Data'] class BaseClass:
def __init__(self):
self.data*=*Lbase*****************#<----**this*fails??

self.data and Lbase are now bound to the same variable. Consequently changes
to that variable through self.data affect Lbase and vice versa. This is
standard Python behaviour. If you want to use Lbase as a kind of initial
value, modify BaseClass to make a (shallow) copy:

class BaseClass:
def __init__(self):
self.data = list(Lbase)

Now every BaseClass instance gets its separate copy. If you have nested
mutables, e. g. a list of lists, look into the copy module for deepcopy().

Peter