Python dataclass

Summary: in this tutorial, you’ll learn about the Python dataclass decorator and how to use it effectively.

Introduction to the Python dataclass

Python introduced the dataclass in version 3.7 (PEP 557). The dataclass allows you to define classes with less code and more functionality out of the box.

The following defines a regular Person class with two instance attributes name and age:

class Person: def __init__(self, name, age): self.name = name self.age = age
Code language: Python (python)

This Person class has the __init__ method that initializes the name and age attributes.

If you want to have a string representation of the Person object, you need to implement the __str__ or __repr__ method. Also, if you want to compare two instances of the Person class by an attribute, you need to implement the __eq__ method.

However, if you use the dataclass, you’ll have all of these features (and even more) without implementing these dunder methods.

To make the Person class a data class, you follow these steps:

First, import the dataclass decorator from the dataclasses module:

from dataclasses import dataclass
Code language: Python (python)

Second, decorate the Person class with the dataclass decorator and declare the attributes:

@dataclass class Person: name: str age: int
Code language: Python (python)

In this example, the Person class has two attributes name with the type str and age with the type int. By doing this, the @dataclass decorator implicitly creates the __init__ method like this:

def __init__(name: str, age: int)
Code language: Python (python)

Note that the order of the attributes declared in the class will determine the orders of the parameters in the __init__ method.

And you can create the Person‘s object:

p1 = Person('John', 25)
Code language: Python (python)

When printing out the Person‘s object, you’ll get a readable format:

print(p1)
Code language: Python (python)

Output:

Person(name='John', age=25)
Code language: Python (python)

Also, if you compare two Person‘s objects with the same attribute value, it’ll return True. For example:

p1 = Person('John', 25) p2 = Person('John', 25) print(p1 == p2)
Code language: Python (python)

Output:

True
Code language: Python (python)

The following discusses other functions that a data class provides.

Default values

When using a regular class, you can define default values for attributes. For example, the following Person class has the iq parameter with the default value of 100.

class Person: def __init__(self, name, age, iq=100): self.name = name self.age = age self.iq = iq
Code language: Python (python)

To define a default value for an attribute in the dataclass, you assign it to the attribute like this:

from dataclasses import dataclass @dataclass class Person: name: str age: int iq: int = 100 print(Person('John Doe', 25))
Code language: Python (python)

Like the parameter rules, the attributes with the default values must appear after the ones without default values. Therefore, the following code will not work:

from dataclasses import dataclass @dataclass class Person: iq: int = 100 name: str age: int
Code language: Python (python)

Convert to a tuple or a dictionary

The dataclasses module has the astuple() and asdict() functions that convert an instance of the dataclass to a tuple and a dictionary. For example:

from dataclasses import dataclass, astuple, asdict @dataclass class Person: name: str age: int iq: int = 100 p = Person('John Doe', 25) print(astuple(p)) print(asdict(p))
Code language: Python (python)

Output:

('John Doe', 25, 100) {'name': 'John Doe', 'age': 25, 'iq': 100}
Code language: Python (python)

Create immutable objects

To create readonly objects from a dataclass, you can set the frozen argument of the dataclass decorator to True. For example:

from dataclasses import dataclass, astuple, asdict @dataclass(frozen=True) class Person: name: str age: int iq: int = 100
Code language: Python (python)

If you attempt to change the attributes of the object after it is created, you’ll get an error. For example:

p = Person('Jane Doe', 25) p.iq = 120
Code language: Python (python)

Error:

dataclasses.FrozenInstanceError: cannot assign to field 'iq'
Code language: Python (python)

Customize attribute behaviors

If don’t want to initialize an attribute in the __init__ method, you can use the field() function from the dataclasses module.

The following example defines the can_vote attribute that is initialized using the __init__ method:

from dataclasses import dataclass, field class Person: name: str age: int iq: int = 100 can_vote: bool = field(init=False)
Code language: Python (python)

The field() function has multiple interesting parameters such as repr, hash, compare, and metadata.

If you want to initialize an attribute that depends on the value of another attribute, you can use the __post_init__ method. As its name implies, Python calls the __post_init__ method after the __init__ method.

The following use the __post_init__ method to initialize the can_vote attribute based on the age attribute:

from dataclasses import dataclass, field @dataclass class Person: name: str age: int iq: int = 100 can_vote: bool = field(init=False) def __post_init__(self): print('called __post_init__ method') self.can_vote = 18 <= self.age <= 70 p = Person('Jane Doe', 25) print(p)
Code language: Python (python)

Output:

called the __post_init__ method Person(name='Jane Doe', age=25, iq=100, can_vote=True)
Code language: Python (python)

Sort objects

By default, a dataclass implements the __eq__ method.

To allow different types of comparisons like __lt__, __lte__, __gt__, __gte__, you can set the order argument of the @dataclass decorator to True:

@dataclass(order=True)
Code language: CSS (css)

By doing this, the dataclass will sort the objects by every field until it finds a value that’s not equal.

In practice, you often want to compare objects by a particular attribute, not all attributes. To do that, you need to define a field called sort_index and set its value to the attribute that you want to sort.

For example, suppose you have a list of Person‘s objects and want to sort them by age:

members = [ Person('John', 25), Person('Bob', 35), Person('Alice', 30) ]
Code language: Python (python)

To do that, you need to:

  • First, pass the order=True parameter to the @dataclass decorator.
  • Second, define the sort_index attribute and set its init parameter to False.
  • Third, set the sort_index to the age attribute in the __post_init__ method to sort the Person‘s object by age.

The following shows the code for sorting Person‘s objects by age:

from dataclasses import dataclass, field @dataclass(order=True) class Person: sort_index: int = field(init=False, repr=False) name: str age: int iq: int = 100 can_vote: bool = field(init=False) def __post_init__(self): self.can_vote = 18 <= self.age <= 70 # sort by age self.sort_index = self.age members = [ Person(name='John', age=25), Person(name='Bob', age=35), Person(name='Alice', age=30) ] sorted_members = sorted(members) for member in sorted_members: print(f'{member.name}(age={member.age})')
Code language: Python (python)

Output:

John(age=25) Alice(age=30) Bob(age=35)
Code language: Python (python)

Summary

  • Use the @dataclass decorator from the dataclasses module to make a class a dataclass. The dataclass object implements the __eq__ and __str__ by default.
  • Use the astuple() and asdict() functions to convert an object of a dataclass to a tuple and dictionary.
  • Use frozen=True to define a class whose objects are immutable.
  • Use __post_init__ method to initalize attributes that depends on other attributes.
  • Use sort_index to specify the sort attributes of the dataclass objects.
Did you find this tutorial helpful ?