Python Descriptors
Summary: in this tutorial, you’ll learn about Python descriptors, how descriptors work, and how to apply them more effectively.
Introduction to the Python descriptors
Suppose you have a class Person
with two instance attributes first_name
and last_name
:
class Person:
def __init__(self, first_name, last_name):
self.first_name = first_name
self.last_name = last_name
Code language: Python (python)
And you want the first_name
and last_name
attributes to be non-empty strings. These plain attributes cannot guarantee this.
To enforce the data validity, you can use property with a getter and setter methods, like this:
class Person:
def __init__(self, first_name, last_name):
self.first_name = first_name
self.last_name = last_name
def first_name(self):
return self._first_name
def first_name(self, value):
if not isinstance(value, str):
raise ValueError('The first name must a string')
if len(value) == 0:
raise ValueError('The first name cannot be empty')
self._first_name = value
def last_name(self):
return self._last_name
def last_name(self, value):
if not isinstance(value, str):
raise ValueError('The last name must a string')
if len(value) == 0:
raise ValueError('The last name cannot be empty')
self._last_name = value
Code language: Python (python)
In this Person
class, the getter returns the attribute value while the setter validates it before assigning it to the attribute.
This code works perfectly fine. However, it is redundant because the validation logic validates the first and last names is the same.
Also, if the class has more attributes that require a non-empty string, you need to duplicate this logic in other properties. In other words, this validation logic is not reusable.
To avoid duplicating the logic, you may have a method that validates data and reuse this method in other properties. This approach will enable reusability. However, Python has a better way to solve this by using descriptors.
First, define a descriptor class that implements three methods __set_name__
, __get__
, and __set__
:
class RequiredString:
def __set_name__(self, owner, name):
self.property_name = name def __get__(self, instance, owner):
if instance is None:
return self
return instance.__dict__[self.property_name] or None
def __set__(self, instance, value):
if not isinstance(value, str):
raise ValueError(f'The {self.property_name} must be a string')
if len(value) == 0:
raise ValueError(f'The {self.property_name} cannot be empty')
instance.__dict__[self.property_name] = value
Code language: Python (python)
Second, use the RequiredString
class in the Person
class:
class Person:
first_name = RequiredString()
last_name = RequiredString()
Code language: Python (python)
If you assign an empty string or a non-string value to the first_name
or last_name
attribute of the Person
class, you’ll get an error.
For example, the following attempts to assign an empty string to the first_name
attribute:
try:
person = Person()
person.first_name = ''
except ValueError as e:
print(e)
Code language: Python (python)
Error:
The first_name must be a string
Code language: Python (python)
Also, you can use the RequiredString
class in any class with attributes that require a non-empty string value.
Besides the RequiredString
, you can define other descriptors to enforce other data types like age, email, and phone. And this is just a simple application of the descriptors.
Let’s understand how descriptors work.
Descriptor protocol
In Python, the descriptor protocol consists of three methods:
__get__
gets an attribute value__set__
sets an attribute value__delete__
deletes an attribute
Optionally, a descriptor can have the __set_name__
method that sets an attribute on an instance of a class to a new value.
What is a descriptor
A descriptor is an object of a class that implements one of the methods specified in the descriptor protocol.
Descriptors have two types: data descriptor and non-data descriptor.
- A data descriptor is an object of a class that implements the
__set__
and/or__delete__
method. - A non-data descriptor is an object that implements the
__get__
method only.
The descriptor type specifies the property lookup resolution that we’ll cover in the next tutorial.
How descriptors work
The following modifies the RequiredString
class to include the print
statements that print out the arguments.
class RequiredString:
def __set_name__(self, owner, name):
print(f'__set_name__ was called with owner={owner} and name={name}')
self.property_name = name def __get__(self, instance, owner):
print(f'__get__ was called with instance={instance} and owner={owner}')
if instance is None:
return self
return instance.__dict__[self.property_name] or None
def __set__(self, instance, value):
print(f'__set__ was called with instance={instance} and value={value}')
if not isinstance(value, str):
raise ValueError(f'The {self.property_name} must a string')
if len(value) == 0:
raise ValueError(f'The {self.property_name} cannot be empty')
instance.__dict__[self.property_name] = value
class Person:
first_name = RequiredString()
last_name = RequiredString()
Code language: Python (python)
The __set_name__
method
When you compile the code, you’ll see that Python creates the descriptor objects for first_name
and last_name
and automatically call the __set_name__
method of these objects. Here’s the output:
__set_name__ was called with owner=<class '__main__.Person'> and name=first_name
__set_name__ was called with owner=<class '__main__.Person'> and name=last_name
Code language: plaintext (plaintext)
In this example, the owner argument of __set_name__
is set to the Person
class in the __main__
module, and the name
argument is set to the first_name
and last_name
attribute accordingly.
It means that Python automatically calls the __set_name__
when the owning class Person
is created. The following statements are equivalent:
first_name = RequiredString()
Code language: Python (python)
and
first_name.__set_name__(Person, 'first_name')
Code language: Python (python)
Inside, the __set_name__
method, we assign the name
argument to the property_name
instance attribute of the descriptor
object so that we can access it later in the __get__
and __set__
method:
self.property_name = name
Code language: Python (python)
The first_name
and last_name
are the class variables of the Person
class. If you look at the Person.__dict__
class attribute, you’ll see two descriptor objects first_name
and last_name
:
from pprint import pprintpprint(Person.__dict__)
Code language: Python (python)
Output:
mappingproxy({'__dict__': <attribute '__dict__' of 'Person' objects>,
'__doc__': None,
'__module__': '__main__',
'__weakref__': <attribute '__weakref__' of 'Person' objects>,
'first_name': <__main__.RequiredString object at 0x0000019D6AB947F0>,
'last_name': <__main__.RequiredString object at 0x0000019D6ACFBE80>})
Code language: Python (python)
The __set__
method
Here’s the __set__
method of the RequiredString
class:
def __set__(self, instance, value):
print(f'__set__ was called with instance={instance} and value={value}') if not isinstance(value, str):
raise ValueError(f'The {self.property_name} must be a string')
if len(value) == 0:
raise ValueError(f'The {self.property_name} cannot be empty')
instance.__dict__[self.property_name] = value
Code language: Python (python)
When you assign the new value to a descriptor, Python calls __set__
method to set the attribute on an instance of the owner class to the new value. For example:
person = Person()
person.first_name = 'John'
Code language: Python (python)
Output:
__set__ was called with instance=<__main__.Person object at 0x000001F85F7167F0> and value=John
Code language: Python (python)
In this example, the instance
argument is person
object and the value is the string 'John'
. Inside the __set__
method, we raise a ValueError
if the new value is not a string or if it is an empty string.
Otherwise, we assign the value to the instance attribute first_name
of the person
object:
instance.__dict__[self.property_name] = value
Code language: Python (python)
Note that Python uses instance.__dict__
dictionary to store instance attributes of the instance
object.
Once you set the first_name
and last_name
of an instance of the Person
object, you’ll see the instance attributes with the same names in the instance’s __dict__
. For example:
person = Person()
print(person.__dict__) # {}person.first_name = 'John'
person.last_name = 'Doe'
print(person.__dict__) # {'first_name': 'John', 'last_name': 'Doe'}
Code language: Python (python)
Output:
{}
{'first_name': 'John', 'last_name': 'Doe'}
Code language: plaintext (plaintext)
The __get__
method
The following shows the __get__
method of the RequiredString
class:
def __get__(self, instance, owner):
print(f'__get__ was called with instance={instance} and owner={owner}')
if instance is None:
return self return instance.__dict__[self.property_name] or None
Code language: Python (python)
Python calls the __get__
method of the Person
‘s object when you access the first_name
attribute. For example:
person = Person()person.first_name = 'John'
print(person.first_name)
Code language: Python (python)
Output:
__set__ was called with instance=<__main__.Person object at 0x000001F85F7167F0> and value=John
__get__ was called with instance=<__main__.Person object at 0x000001F85F7167F0> and owner=<class '__main__.Person'>
Code language: plaintext (plaintext)
The __get__
method returns the descriptor if the instance
is None
. For example, if you access the first_name
or last_name
from the Person
class, you’ll see the descriptor object:
print(Person.first_name)
Code language: Python (python)
Output:
<__main__.RequiredString object at 0x000001AF1DA147F0>
Code language: plaintext (plaintext)
If the instance
is not None
, the __get__()
method returns the value of the attribute with the name property_name
of the instance
object.
Summary
- Descriptors are objects of class that implements one of the method in the descriptor protocol including
__set__
,__get__
,__del__