Introduction to Python descriptors

Manage attribute access with Python descriptors

Learn to easily create and apply descriptors in Python.


Alex Starostin (, QA Engineer, IBM

Photo of Alexander StarostinAlex is a QA Engineer in IBM Australia development lab. His main focus is software testing and QA support for development team in Agile environment. Apart from regular work duties Alex likes to devote his time to learn new technologies and programming languages.

26 June 2012

Also available in Chinese Japanese Vietnamese Portuguese


Python descriptors were introduced in Python 2.2, along with new style classes, yet they remain widely unused. Python descriptors are a way to create managed attributes. Among their many advantages, managed attributes are used to protect an attribute from changes or to automatically update the values of a dependant attribute.

Descriptors increase an understanding of Python, and improve coding skills. This article introduces descriptor protocol and demonstrates how to create and use descriptors.

Descriptors protocol

Python descriptor protocol is simply a way to specify what happens when an attribute is referenced on a model. It allows a programmer to easily and efficiently manage attribute access:

  • set
  • get
  • delete

In other programming languages, descriptors are referred to as setter and getter, where public functions are used to Get and Set a private variable. Python doesn't have a private variables concept, and descriptor protocol can be considered as a Pythonic way to achieve something similar.

In general, a descriptor is an object attribute with a binding behavior, one whose attribute access is overridden by methods in the descriptor protocol. Those methods are __get__, __set__, and __delete__. If any of these methods are defined for an object, it is said to be a descriptor. Take a closer look at these methods in Listing 1.

Listing 1. Descriptor methods
__get__(self, instance, owner)
__set__(self, instance, value)
__delete__(self, instance)


__get__ accesses the attribute. It returns the value of the attribute, or raise the AttributeError exception if a requested attribute is not present.

__set__ is called in an attribute assignment operation. Returns nothing.

__delete__ controls a delete operation. Returns nothing.

It is important to note that descriptors are assigned to a class, not an instance. Modifying the class overwrites or deletes the descriptor itself, rather than triggering its code.

When descriptors are needed

Consider an email attribute. Verification of the correct email format is necessary before assigning a value to that attribute. This descriptor allows email to be processed through a regular expression and its format validated before assigning it to an attribute.

In many other cases, Python protocol descriptors control access to attributes, such as protection of the name attribute.

Creating descriptors

You can create a descriptor a number of ways:

  • Create a class and override any of the descriptor methods: __set__, __ get__, and __delete__. This method is used when the same descriptor is needed across many different classes and attributes, for example, for type validation.
  • Use a property type which is a simpler and more flexible way to create a descriptor.
  • Use the power of property decorators which are a combination of property type method and Python decorators.

All examples below are similar from an operational viewpoint. The difference lies in implementation.

Creating descriptors using class methods

Listing 2 demonstrates the simplicity of controlling attribute assignment in Python.

Listing 2. Creating descriptors using class methods
class Descriptor(object):

    def __init__(self):
        self._name = ''

    def __get__(self, instance, owner):
        print "Getting: %s" % self._name
        return self._name

    def __set__(self, instance, name):
        print "Setting: %s" % name
        self._name = name.title()

    def __delete__(self, instance):
        print "Deleting: %s" %self._name
        del self._name

class Person(object):
    name = Descriptor()

Use this code and see the output:

>>> user = Person()
>>> = 'john smith'
Setting: john smith
Getting: John Smith
'John Smith'
>>> del
Deleting: John Smith

A descriptor class was created overriding __set__(), __get__() and __delete__() methods of the parent class in such a way that

  • get will print Getting
  • delete will print Deleting
  • set will print Setting

and change the attribute value to title (first letter uppercase, other letters lowercase) before assignment. This is handy, for example, when storing and printing names.

Uppercase conversion can equally be moved to __get__() method. The _value will have the original value, and will be converted to title on get request.

Creating descriptors using property type

While the descriptor specified in Listing 2 is valid and functional, another method is through the property type. With the property(), it is easy to create a usable descriptor for any attribute. The syntax for creating property() is property(fget=None, fset=None, fdel=None, doc=None) where:

  • fget – attribute get method
  • fset – attribute set method
  • fdel – attribute delete method
  • doc – docstring

Rewrite the example using property, as in Listing 3.

Listing 3. Creating descriptor with property type
class Person(object):
    def __init__(self):
        self._name = ''

    def fget(self):
        print "Getting: %s" % self._name
        return self._name
    def fset(self, value):
        print "Setting: %s" % value
        self._name = value.title()

    def fdel(self):
        print "Deleting: %s" %self._name
        del self._name
    name = property(fget, fset, fdel, "I'm the property.")

Use this code and see the output:

>>> user = Person()
>>> = 'john smith'
Setting: john smith
Getting: John Smith
'John Smith'
>>> del
Deleting: John Smith

Clearly, the result is the same. Note here that fget, fset and fdel methods are optional, but if one is not specified, an exception of AttributeError is raised when the respective operation is attempted. For example, a name property is declared with None as fset, and then the developer tries to assign value to name attribute. An exception AttributeError is raised.

This can be used to define read-only attributes in the system.

name = property(fget, None, fdel, "I'm the property") = 'john smith'


Traceback (most recent call last):
File stdin, line 21, in mоdule = 'john smith'
AttributeError: can't set attribute

Creating descriptors using property decorators

Descriptors can be created with Python decorators, as in Listing 4. A Python decorator is a specific change to the Python syntax allowing a more convenient alteration of functions and methods. In this case, attribute management methods are altered. Find more information on the application of Python decorators in the developerWorks article, Decorators make magic easy.

Listing 4. Creating descriptors with property decorators
class Person(object):

    def __init__(self):
        self._name = ''

    def name(self):
        print "Getting: %s" % self._name
        return self._name

    def name(self, value):
        print "Setting: %s" % value
        self._name = value.title()

    def name(self):
        print ">Deleting: %s" % self._name
        del self._name

Creating descriptors at run time

All the preceding examples operate with the name attribute. The limitation of this approach is the necessity of separately overriding __set__(), __get__() and __delete__() for each attribute. Listing 5 provides a possible solution when a developer wishes to add property attributes at run time. This uses property type to build a data descriptor.

Listing 5. Creating descriptors at run time
class Person(object):

    def addProperty(self, attribute):
        # create local setter and getter with a particular attribute name 
        getter = lambda self: self._getProperty(attribute)
        setter = lambda self, value: self._setProperty(attribute, value)

        # construct property attribute and add it to the class
        setattr(self.__class__, attribute, property(fget=getter, \
                                                    fset=setter, \
                                                    doc="Auto-generated method"))

    def _setProperty(self, attribute, value):
        print "Setting: %s = %s" %(attribute, value)
        setattr(self, '_' + attribute, value.title())    

    def _getProperty(self, attribute):
        print "Getting: %s" %attribute
        return getattr(self, '_' + attribute)

Let's play with this code:

>>> user = Person()
>>> user.addProperty('name')
>>> user.addProperty('phone')
>>> = 'john smith'
Setting: name = john smith
>>> = '12345'
Setting: phone = 12345
Getting: name
'John Smith'
>>> user.__dict__
{'_phone': '12345', '_name': 'John Smith'}

This created name and phone attributes at run time. They are accessible by corresponding name, but they are stored in the object namespace dictionary as _name and _phone, as specified in _setProperty method. Basically, name and phone are accessors to internal _name and _phone attributes.

You might have one question regarding a _name attribute in the system when the developer tries to add name property attribute. The answer is that it will overwrite the existing _name attribute with the new property attribute. This code allows control of how attributes are handled inside a class.


Python descriptors allow for powerful and flexible attribute management with new style classes. Combined with decorators, they make for elegant programming, allowing creation of Setters and Getters, as well as read-only attributes. It also allows you to run attribute validation on request by value or type. You can apply descriptors in many areas, but use them with discretion to avoid unnecessary code complexity stemming from overriding the normal behaviors of an object.



Get products and technologies

  • Evaluate IBM products in the way that suits you best: Download a product trial, try a product online, use a product in a cloud environment, or spend a few hours in the SOA Sandbox learning how to implement Service Oriented Architecture efficiently.



developerWorks: Sign in

Required fields are indicated with an asterisk (*).

Need an IBM ID?
Forgot your IBM ID?

Forgot your password?
Change your password

By clicking Submit, you agree to the developerWorks terms of use.


The first time you sign into developerWorks, a profile is created for you. Information in your profile (your name, country/region, and company name) is displayed to the public and will accompany any content you post, unless you opt to hide your company name. You may update your IBM account at any time.

All information submitted is secure.

Choose your display name

The first time you sign in to developerWorks, a profile is created for you, so you need to choose a display name. Your display name accompanies the content you post on developerWorks.

Please choose a display name between 3-31 characters. Your display name must be unique in the developerWorks community and should not be your email address for privacy reasons.

Required fields are indicated with an asterisk (*).

(Must be between 3 – 31 characters.)

By clicking Submit, you agree to the developerWorks terms of use.


All information submitted is secure.

Dig deeper into Open source on developerWorks

Zone=Open source
ArticleTitle=Introduction to Python descriptors