What is `__init__.py` for in Python?

Wamaitha N.

The Problem

What is __init__.py used for in Python?

The Solution

In Python projects, if you create a file called __init__.py in a directory then Python will treat that directory as a package. A package in Python is a collection of modules (individual .py files) that can be imported into other Python files.

In older versions of Python (before 3.3), it was necessary to create an __init__.py file in a package before you could use import statements in the form from mypackage import mymodule. Since version 3.3 and the implementation of PEP 420, Python will automatically create “Namespace Packages” implicitly in many cases. This means that __init__.py is now often optional, but it’s still useful to structure your initialization code and define how statements like from mypackage import * should work.

Let’s take a look at what the __init__.py file does with an example Python package. Imagine we have a calculator script that imports a divider and multiplier module from an operations package.

Our directory structure looks like this:

├── calculator.py └── operations ├── divider.py └── multiplier.py

And our three files look like this:

# calculator.py from operations import multiplier, divider mymultiplier = multiplier.Multiplier() result = mymultiplier.multiply(2, 5) print(f"2 x 5 is {result}") mydivider = divider.Divider() result = mydivider.divide(10, 2) print(f"10 / 2 is {result}")
# operations/divider.py class Divider: def divide(self, a, b): return a / b
# operations/multiplier.py class Multiplier: def multiply(self, a, b): return a * b

In Python 3.3 and above, this works fine. When we run from operations import multiplier, Python treats the operations subdirectory as a Namespace Package and imports the multiplier module from that package. In previous versions of Python, this would have raised an error:

File "calculator.py", line 1, in <module> from operations import multiplier ImportError: No module named operations

And simply adding a blank file called __init__.py in the operations directory would fix the issue.

touch operations/__init__.py

Now running the calculator.py script would work as expected:

2 x 5 is 10 10 / 2 is 5.0

What does __init__.py do?

Even in Python 3.3 and above, creating an __init__.py file will change the behavior of your scripts, though these changes are more subtle than in older versions of Python. It can automatically define the __file__ metadata of your package, which is useful.

If you import a package and inspect it with Python’s built-in dir function, you’ll see a difference with and without having an __init__.py file.

Add the following to the top of calculator.py:

# calculator.py import operations print(dir(operations)) print(operations.__file__)

This shows us all the variables available in the operations package and prints out its path on the system. Without an __init__.py file in your operations subdirectory, you’ll see the following:

['__doc__', '__file__', '__loader__', '__name__', '__package__', '__path__', '__spec__'] None

You can see that Python has automatically added some metadata in the variables starting and ending with two underscores, like __file__.

However, the None on the next line shows that the __file__ variable is blank.

If you create an __init__.py file in the operations subdirectory, it is now an explicit package. If you run the calculator.py file again, you’ll now see some differences in the output:

['__builtins__', '__cached__', '__doc__', '__file__', '__loader__', '__name__', '__package__', '__path__', '__spec__'] /Users/g/python_init_example/operations/__init__.py

Now the full path to the __init__.py file has been used as the __file__ variable for the operations package. Python has also added the __builtins__ and __cached__ metadata variables for us.

Using __init__.py to run code and control * imports

Larger and more complicated packages often use __init__.py to better organize code and to run any initialization that should be run automatically when the package is imported.

For example, you could add the line print("Hi from __init__.py") to your operations/__init__.py file. Now if you import the operations package, it will run the print statement immediately:

python_init_example$ python3 Python 3.11.5 (main, Aug 24 2023, 15:09:45) [Clang 14.0.3 (clang-1403.0.22.14.1)] on darwin Type "help", "copyright", "credits" or "license" for more information. >>> import operations Hi from init

A common thing to define in the __init__.py is the __all__ variable. This overwrites what modules and functions should be imported when a user runs an import statement in the form from mypackage import *.

By default, Python does not import modules from a package. So if we changed the first line in our calculator.py file to import everything from operations, it would not work.

# calculator.py # broken import unless we explicitly define `__all__` in `__init__.py` from operations import * mymultiplier = multiplier.Multiplier() result = mymultiplier.multiply(2, 5) print(f"2 x 5 is {result}") mydivider = divider.Divider() result = mydivider.divide(10, 2) print(f"10 / 2 is {result}")

If we run calculator.py now, we’ll get an error:

Traceback (most recent call last): File "/Users/g/python_init_example/calculator.py", line 3, in <module> mymultiplier = multiplier.Multiplier() ^^^^^^^^^^ NameError: name 'multiplier' is not defined

But if we add the following line to operations/__init__.py:

__all__ = ['divider', 'multiplier']

the calculator.py script will run as expected:

2 x 5 is 10 10 / 2 is 5.0

Get Started With Sentry

Get actionable, code-level insights to resolve Python performance bottlenecks and errors.

  1. Create a free Sentry account

  2. Create a Python project and note your DSN

  3. Grab the Sentry Python SDK

pip install --upgrade sentry-sdk
  1. Configure your DSN
import sentry_sdk sentry_sdk.init( "https://<key>@sentry.io/<project>", # Set traces_sample_rate to 1.0 to capture 100% # of transactions for performance monitoring. # We recommend adjusting this value in production. traces_sample_rate=1.0, )

Check our documentation for the latest instructions.

Loved by over 4 million developers and more than 90,000 organizations worldwide, Sentry provides code-level observability to many of the world’s best-known companies like Disney, Peloton, Cloudflare, Eventbrite, Slack, Supercell, and Rockstar Games. Each month we process billions of exceptions from the most popular products on the internet.

Share on Twitter
Bookmark this page
Ask a questionJoin the discussion

Related Answers

A better experience for your users. An easier life for your developers.

    TwitterGitHubDribbbleLinkedinDiscord
© 2024 • Sentry is a registered Trademark
of Functional Software, Inc.