Python 201: What are descriptors?

Descriptors were introduced to Python way back in version 2.2. They provide the developer with the ability to add managed attributes to objects. The methods needed to create a descriptor are __get__, __set__ and __delete__. If you define any of these methods, then you have created a descriptor.

The idea behind the descriptor is to get, set or delete attributes from your object’s dictionary. When you access a class attribute, this starts the lookup chain. Should the looked up value be an object with one of our descriptor methods defined, then the descriptor method will be invoked.

Descriptors power a lot of the magic of Python’s internals. They are what make properties, methods and even the super function work. They are also used to implement the new style classes that were also introduced in Python 2.2.

The Descriptor Protocol

The protocol to create a descriptor is really quite easy. You only need to define one or more of the following:

  • __get__(self, obj, type=None), returns value
  • __set__(self, obj, value), returns None
  • __delete__(self, obj), returns None

Once you’ve defined at least one, you have created a descriptor. If you can you define both __get__ and __set__, you will have created a data descriptor. A descriptor with only __get__() defined are known as non-data descriptors and are usually used for methods. The reason for this distinction in descriptor types is that if an instance’s dictionary happens to have a data descriptor, the descriptor will take precedence during the lookup. If the instance’s dictionary has an entry that matches up with a non-data descriptor, then the dictionary’s own entry will take precedence over the descriptor.

You can also create a read-only descriptor if you define both __get__ and __set__, but raise an AttributeError when the __set__ method is called.

Calling a Descriptor

The most common method of calling a descriptor is for the descriptor to be invoked automatically when you access an attribute. A typical example would be my_obj.attribute_name. This will cause your object to look up attribute_name in the my_obj object. If your attribute_name happens to define __get__(), then attribute_name.__get__(my_obj) will get called. This all depends on whether your instance is an object or a class.

The magic behind this lies in the magic method known as __getattribute__, which will turn my_obj.a into this: type(my_obj).__dict__[‘a’].__get__(a, type(a)). You can read all about the implementation in Python’s documentation here:

According to said documentation, there are a few points to keep in mind in regards to calling a descriptor:

  • The descriptor is invoked via the default implementation of the __getattribute__ method
  • If you override __getattribute__, this will prevent the descriptor from getting automatically called
  • object.__getattribute__() and type.__getattribute__() don’t call __get__() the same way
  • A data descriptor will always, ALWAYS override instance dictionaries
  • The non-data descriptor can be overridden by instance dictionaries.

More information on how all this works can be found in Python’s data model, the Python source code and in Guido van Rossum’s document, “Unifying types and class in Python”.

Descriptor examples

At this point, you may be confused how you would even use a descriptor. I always find it helpful when I am learning a new concept if I have a few examples that demonstrate how it works. So in this section, we will look at some examples so you will know how to use descriptors in your own code!

Let’s start by writing a really simple data descriptor and then use it in a class. This example is based on one from Python’s documentation:

class MyDescriptor():
    A simple demo descriptor
    def __init__(self, initial_value=None, name='my_var'):
        self.var_name = name
        self.value = initial_value
    def __get__(self, obj, objtype):
        print('Getting', self.var_name)
        return self.value
    def __set__(self, obj, value):
        msg = 'Setting {name} to {value}'
        print(msg.format(name=self.var_name, value=value))
        self.value = value
class MyClass():
    desc = MyDescriptor(initial_value='Mike', name='desc')
    normal = 10
if __name__ == '__main__':
    c = MyClass()
    c.desc = 100

Here we create a class and define three magic methods:

  • __init__ – our constructor which takes a value and the name of our variable
  • __get__ – prints out the current variable name and returns the value
  • __set__ – prints out the name of our variable and the value we just assigned and sets the value itself

Then we create a class that creates an instance of our descriptor as a class attribute and also creates a normal class attribute. Then we run a few “tests” by creating an instance of our normal class and accessing our class attributes. Here is the output:

Getting desc
Setting desc to 100
Getting desc

As you can see, when we access c.desc, it prints out our “Getting” message and we print out what it returns, which is “Mike”. Next we print out the regular class attribute’s value. Finally we change the descriptor variable’s value, which causes our “Setting” message to be printed. We also double-check the current value to make sure that it was actually set, which is why you see that last “Getting” message.

Python uses descriptors underneath the covers to build properties, bound / unbound methods and class methods. If you look up the property class in Python’s documentation, you will see that it follows the descriptor protocol very closely:

property(fget=None, fset=None, fdel=None, doc=None)

It clearly shows that the property class has a getter, setter and a deleting method.

Let’s look at another example where we use a descriptor to do validation:

from weakref import WeakKeyDictionary
class Drinker:
    def __init__(self):
        self.req_age = 21
        self.age = WeakKeyDictionary()
    def __get__(self, instance_obj, objtype):
        return self.age.get(instance_obj, self.req_age)
    def __set__(self, instance, new_age):
        if new_age < 21:
            msg = '{name} is too young to legally imbibe'
            raise Exception(msg.format(
        self.age[instance] = new_age
        print('{name} can legally drink in the USA'.format(
    def __delete__(self, instance):
        del self.age[instance]
class Person:
    drinker_age = Drinker()
    def __init__(self, name, age): = name
        self.drinker_age = age
p = Person('Miguel', 30)
p = Person('Niki', 13)

Once again, we create a descriptor class. In this case, we use Python’s weakref library’s WeakKeyDictionary, which is a neat class that creates a dictionary that maps keys weakly. What this means is that when there are no strong references to a key in the dictionary, that key and its value will be discarded. We are using that in this example to prevent our Person instances from hanging around indefinitely.

Anyway, the part of the descriptor that we care most about is in our __set__ method. Here we check to see that the instance’s age parameter is greater than 21, which is what you would need to be in the USA if you wanted to drink an alcoholic beverage. If you’re age is lower, then it will raise an exception. Otherwise it will print out the name of the person and a message. To test out our descriptor, we create two instances with one that is greater than 21 in age and one that is less. If you run this code you should see the following output:

Miguel can legally drink in the USA
Traceback (most recent call last):
  File "", line 32, in <module>
    p = Person('Niki', 13)
  File "", line 28, in __init__
    self.drinker_age = age
  File "", line 14, in __set__
    raise Exception(msg.format(
Exception: Niki is too young to legally imbibe

That obviously worked the way it was supposed to, but it’s not really obvious how it worked. The reason this works the way it does is that when we go to set drinker_age, Python notices that it is a descriptor. Python knows that drinker_age is a descriptor because we defined it as such when we created it as a class attribute:

drinker_age = Drinker()

So when we go to set it, we actually call our descriptor’s __set__ method which passes in the instance and the age that we are trying to set. If the age is less than 21, then we raise an exception with a custom message. Otherwise we print out a message that you are old enough.

Getting back to how this all works, if we were to try to print out the drinker_age, Python would execute Person.drinker_age.__get__. Since drinker_age is a descriptor, its __get__ is what actually gets called. If you wanted to set the drinker_age, you would do this:

p.drinker_age = 32

Python would then call Person.drinker_age.__set__ and since that method is also implemented in our descriptor, then the descriptor method is the one that gets called. Once you trace your way through the code execution a few times, you will quickly see how this all works.

The main thing to remember is that descriptors are linked to classes and not to instances.


Descriptors are pretty important because of all the places they are used in Python’s source code. They can be really useful to you too if you understand how they work. However, their use cases are pretty limited and you probably won’t be using them very often. Hopefully this chapter will have helped you see the descriptor’s usefulness and when you might want to use one yourself.

Related Reading

Print Friendly, PDF & Email