Difference between class and instance variables in Python

Object oriented programming in Python can be tricky if your background comes from C++ or Java. First thing is that both C++ and Java are statically typed. This is different in Python that belongs to the class of dynamic scoped languages [4].

For Java and C++ for example if you forget to declare a variable before you try to access it it will generate an error during compilation phase. Below is an example code in Java that illustrates this.

Java

	public class JavaExample {
	// with the next line commented the code is erroring out when we try to compile it
	//public int number;
	public JavaExample (int _number) {
	number=_number;
	}
	public void show() {
	System.out.println("[test " + number + "] list=%s");
	}
	public static void main(String[] args) {
	JavaExample t=new JavaExample(1);
	t.show();
	}
	}

view raw JavaExample.java hosted with ❤ by GitHub

$ javac JavaExample.java 
JavaExample.java:7: cannot find symbol
symbol  : variable number
location: class JavaExample
        number=_number;
        ^
JavaExample.java:11: cannot find symbol
symbol  : variable number
location: class JavaExample
       System.out.println("[test " + number + "] list=%s");
                                     ^
2 errors

Python

When you start applying object oriented programming paradigms in Python and start to write classes and create instances you may encounter an interesting phenomena. The code below illustrate this.

	class Test1InstanceVariable:
	mylist=[]
	def __init__(self, number):
	self.number= number
	def add(self, i):
	self.mylist.append(i)
	def show(self):
	print "[test %d] list=%s" % (self.number, " ".join(self.mylist))
	class Test2ClassVariable:
	mylist=[]
	def __init__(self, number):
	self.number= number
	self.mylist=[]
	def add(self, i):
	self.mylist.append(i)
	def show(self):
	print "[test %d] list=%s" % (self.number, " ".join(self.mylist))
	if __name__ == '__main__':
	t1=Test1InstanceVariable(1)
	t1.add('one')
	t1.show()
	t1=Test1InstanceVariable(2)
	t1.add('two')
	t1.show()
	t1=Test2ClassVariable(3)
	t1.add('three')
	t1.show()
	t1=Test2ClassVariable(4)
	t1.add('four')
	t1.show()

view raw object_variables_example.py hosted with ❤ by GitHub

The difference between the 2 classes is that in the class Test2ClassVariable we have defined this variable self.mylist=[] additionally.

When you run this code this is the output:

$ python object_variables_example.py
[test 1] list=one
[test 2] list=one two
[test 3] list=three
[test 4] list=four

The interesting part of the output is in line #3. The variable list has 2 entries where at first look we would expect it should have only one. Not knowing anything about objects in Python this would be a natrual intuition for everyone with Java or C++ background.

Explanation

We see this behavoiur because Python uses a concept of class and instance variables. For Java/C++ programmers a python class variable is simply a Java/C++ static class variable. In our example variable list is a class variable and it is shared between all class instances. That explains the output we see.

References

1. http://stackoverflow.com/questions/68645/static-class-variables-in-python
2. http://docs.python.org/tutorial/classes.html#instance-objects
3. http://en.wikipedia.org/wiki/Class_variable
4. http://en.wikipedia.org/wiki/Python_(programming_language)