Dealing with dependencies

Developing Python Packages

James Fulton

Climate informatics researcher

What are dependencies?

Other packages you import inside your package
Inside mymodule.py:

# These imported packages are dependencies
import numpy as np
import pandas as pd
...

Adding dependencies to setup.py

from setuptools import setup, find_packages

setup(
    ...
    install_requires=['pandas', 'scipy', 'matplotlib'],
)

Controlling dependency version

from setuptools import setup, find_packages

setup(
    ...
    install_requires=[

        'pandas>=1.0',

        'scipy==1.1',

        'matplotlib>=2.2.1,<3'

    ],
)

Controlling dependency version

from setuptools import setup, find_packages

setup(
    ...
    install_requires=[

        'pandas>=1.0',         # good
        'scipy==1.1',          # bad
        'matplotlib>=2.2.1,<3' # good
    ],
)

Allow as many package versions as possible
Get rid of unused dependencies

Python versions

from setuptools import setup, find_packages

setup(
    ...
    python_requires='>=2.7, !=3.0.*, !=3.1.*',
)

Choosing dependency and package versions

Check the package history or release notes
- e.g. the NumPy release notes
Test different versions

A sample of the NumPy release notes.

Making an environment for developers

pip freeze

alabaster==0.7.12
appdirs==1.4.4
argh==0.26.2
...
wrapt==1.11.2
yapf==0.29.0
zipp==3.1.0

Making an environment for developers

Save package requirements to a file

pip freeze > requirements.txt

Install requirements from file

pip install -r requirements.txt

mysklearn/
|-- mysklearn
|   |-- __init__.py
|   |-- preprocessing
|   |   |-- __init__.py
|   |   |-- normalize.py
|   |   |-- standardize.py
|   |-- regression
|   |   |-- __init__.py
|   |   |-- regression.py
|   |-- utils.py
|-- setup.py
|-- requirements.txt   <-- developer environment

Let's practice!

Developing Python Packages