Demystifying Setup.py: The Blueprint Behind Python Packages 📦

Ever wondered what makes 'pip install' work? Let's break down the mechanics behind setup.py 📦

At its core, setup.py is a Python script that defines your package's identity card. It contains essential metadata about your package - from its name and version to dependencies and distribution settings.

Here's what a basic setup.py looks like:

setup(
    name="my-awesome-package",
    version="1.0.0",
    author="Your Name",
    description="A short description",
)

Here are key techniques that make package distribution seamless:

1. Smart Package Discovery:

setup(
    packages=find_packages(exclude=['tests*']),
# Include non-Python files
    package_data={
        'mypackage': ['data/*.csv', 'models/*.pkl']
    }
)

Pro tip: Don't forget to include init.py in all packages you want to distribute!

2. Environment-Specific Dependencies:

setup(
    install_requires=[
        'numpy>=1.20.0',
        'scipy; python_version>="3.8"',
        'pandas<2.0; python_version<"3.8"'
    ]
)

3. Package Structure:

pythonCopysetup(
    packages=find_packages(exclude=['tests*']),
    # Include non-Python files
    package_data={
        'mypackage': ['data/*.csv', 'models/*.pkl']
    }
)

4. Distribution Settings:

• Determines what gets included in your wheel/distribution files

• Controls how your package is built and installed

• Manages Python version compatibility

When you're ready to distribute, simply run: python3 setup.py bdist_wheel

This creates a wheel file - a binary distribution format that makes installation faster and more reliable.

🔍 Want to dive deeper? Here are resources I found useful:

Here’s a video that I had used long ago to understand the setup.py file better:

https://www.youtube.com/watch?v=QgZ7qv4Cd0Y

The source documentation:

https://docs.python.org/3.11/distutils/setupscript.html