How to use NamedTuple and Dataclass in Python?

Last Updated : 26 Jul, 2020

We have all worked with classes and objects for more than once while coding. But have you ever wondered how to create a class other than the naive methods we have all been taught. Don't worry in this article we are going to cover these alternate methods. There are two alternative ways to construct a class in Python.

First, of all let's create a class with the naive methods. Let's create a class transaction, where each payment transaction has a sender, a receiver, date, and the amount.

Example:

Python

class Transaction:
    
    def __init__(self, amount, sender, receiver, date):
      self.amount = amount
      self.sender = sender
      self.receiver = receiver
      self.date = date

This is pretty straight forward but let’s be honest this is a lot of code for us lazy developers. Now, this is where NamedTuple comes in.

Using NamedTuple

Named Tuple is essentially an extension of the Python built-in tuple data type. Python’s tuple is a simple data structure for grouping objects with different types. Its defining feature is being immutable i.e an object whose state cannot be modified after it is created.

We will start with the collections package. This package provides alternatives to Python’s general built-in types like dict, list, set, and tuple. Creating a NamedTuple object is as straightforward as creating the class.

Python3

import collections

Transaction = collections.namedtuple('Transaction',
                                     ['sender', 'amount',
                                      'receiver', 'date'])

# Creating object
record = Transaction(sender="Aryaman",
                     receiver="Ankur",
                     date="2020-06-18",
                     amount=1.0)

print(record)
print(record.receiver)

Output:

Transaction(sender='Aryaman', amount=1.0, receiver='Ankur', date='2020-06-18')

Ankur

Using DataClass

DataClass may be a new feature introduced since Python 3.7. It is used as a decorator. What it does under the hood is implementing __init__, __repr__, etc for us.

Example:

Python3

from dataclasses import dataclass


@dataclass
class Transaction:
    sender: str
    receiver: str
    date: str
    amount: float


# Creating object
record = Transaction(sender="Aryaman",
                     receiver="Ankur",
                     date="2020-06-18",
                     amount=1.0)

print(record)
print(record.receiver)

Output:

Transaction(sender='Aryaman', receiver='Ankur', date='2020-06-18', amount=1.0) Ankur

NamedTuple behaves like a tuple, while DataClass behaves more like a regular Python class because by default, the attributes are all mutable and they can only be accessed by name, not by index.