Provide a simple way to install requirements
What
When you release your software (and while developing it), you need a simple way to install the dependencies and set up your application.
Why
This makes it easier for end-users to use your tool. This also helps make it simpler if you decide to package your application.
When
Inception
Implementations:
- conda environment.yml
- requirements.txt
- pip install .
Recommendation:
We recommend using conda (or one of its many derivatives) as it makes both installation of dependencies and packaging of your application accessible
A conda environment file (environment.yml) is particularly well-suited for scientific software as it can specify exact versions of both Python and non-Python dependencies, ensuring consistent behavior across different systems. It is important to note that this practice is common in the scientific community but not as common in the broader Python community. Most Python packages in general should focus on being pip-installable, as this is a more common practice.
Including installation instructions in your README and providing multiple options (such as conda and pip) increases the likelihood that users can successfully install your software regardless of their preferred package manager.