Python Namedtuple Example

Namedtuple. With tuples we store small groups of related fields together.

A Point tuple, for example will have two coordinates: X and Y. The coordinates are part of a single unit.

To use namedtuple, we import the collections module. We specify the Type name. And we pass a string list of all the field names—these can then be directly used as fields.

An example. Here we create an Employee namedtuple with three fields. We pass a list to collections.namedtuple to specify field names id, title and salary.

Create: We create a namedtuple by using the Employee class name returned by collections.namedtuple.

And: At the end of the program we access the "title" field with the syntax e.title: no index is required.

Based on:

Python 3

Python program that creates a namedtuple

import collections

# Specify the Employee namedtuple.
Employee = collections.namedtuple("Employee", ["id", "title", "salary"])

# Create Employee instance.
e = Employee(1, "engineer", 10000)

# Display Employee.
print(e)
print("Title is", e.title)

Output

Employee(id=1, title='engineer', salary=10000)
Title is engineer

Make method. Namedtuple offers extra methods—one of them is _make. With _make() we create a namedtuple from an iterable (like a list). The list elements are turned into tuple fields.

Tip: The make method fills namedtuple fields based on the order of the list's elements. So we must be careful with ordering.

List

Python program that uses _make method

import collections

# A namedtuple type.
Style = collections.namedtuple("Style", ["color", "size", "width"])

# A list containing three values.
values = ["red", 10, 15]

# Make a namedtuple from the list.
tuple = Style._make(values)
print(tuple)

Output

Style(color='red', size=10, width=15)

Creation time benchmark. This benchmark tests how long it takes to create a namedtuple versus a regular tuple. The results are not encouraging for namedtuple.

Note: I found that namedtuple is about four times slower in creation than a regular tuple.

Also: With a JIT compiler like PyPy, namedtuple was more competitive but still slower.

Python program that benchmarks namedtuple, tuple

import collections
import time

# The namedtuple instance.
Animal = collections.namedtuple("Animal", ["size", "color"])

print(time.time())

# Version 1: create namedtuple.
i = 0
while i < 10000000:
    a = Animal(100, "blue")
    if a[0] != 100:
	raise Exception()
    i += 1

print(time.time())

# Version 2: create tuple.
i = 0
while i < 10000000:
    a = (100, "blue")
    if a[0] != 100:
	raise Exception()
    i += 1

print(time.time())

Results

1403288922.718007
1403288936.606831  namedtuple = 13.89 s
1403288939.667796  tuple      =  3.06 s

Research. The Python documentation provides information about namedtuples. These are implemented in a similar way to regular tuples. So they do not use excess memory.

Named tuple instances do not have per-instance dictionaries, so they are lightweight and require no more memory than regular tuples.

Collections, namedtuple: python.org

A comparison. Even though namedtuple does not use excess memory, it involves more steps to initialize. And it requires a separate module (collections), making it harder to access.

In most programs, a tuple is a better choice than a named tuple. And for large groups of keys, a dictionary or set may be better. Namedtuple is limited in its use.

Related Links