Skip to content

Conda meets Cargo. Keep your conda dev environments local, specific, and up-to-date.

License

Notifications You must be signed in to change notification settings

Yoshanuikabundi/snakes_on_a_plane

Repository files navigation

Snakes on a Plane

GitHub Actions Build Status codecov Documentation Status

Conda meets Cargo. SOAP lets you easily maintain Conda environments for individual projects.

Soap is configured in soap.toml. SOAP always looks for this file in the root of the git repository it was called from. It can also be configured in the tool.soap table of pyproject.toml. Specify environments with a name and a Conda environment YAML file:

[envs]
dev = "devtools/conda-envs/test_env.yml"
user = "devtools/conda-envs/user_env.yml"
docs = "devtools/conda-envs/docs_env.yml"

Then run commands in an environment with soap run:

soap run --env docs "sphinx-build -n auto docs docs/_build/html"

SOAP will check that the environment matches the specification with every call, so if you pull in an update to docs_env.yml and run soap run ... again, your environment will be updated. This won't necessarily update dependencies if the spec hasn't changed; to do this, run soap update:

soap update

You can also define your own aliases for commands. For simple commands, define the command as a value in the aliases table:

[aliases]
greet = "echo 'hello world'"

To configure an alias, define a table instead of a string:

[aliases.docs]
cmd = "sphinx-build -j auto docs docs/_build/html"
chdir = true # Run the command in the git repository root directory
env = "docs" # Use the docs environment by default
description = "Build the docs with Sphinx" # Description for --help

In either case, the alias becomes a sub-command:

soap greet

The environment used by an alias can be defined in the TOML file, but it can also be overridden on the command line:

soap docs --env user

SOAP will always check that the environment is correct before running aliases, just like for soap run!

Installation

I recommend installing pipx and micromamba with your system package manager, and then installing SOAP with pipx. This will give you a fast, easy to manage Conda installation and will avoid installing anything in your system Python distribution. For example, with the Paru package manager for Arch Linux:

paru -Syu python-pipx micromamba-bin
git clone https://github.com/Yoshanuikabundi/snakes_on_a_plane.git
cd snakes_on_a_plane
pipx install .

If you have an existing Mamba or Conda installation, SOAP should be able to detect and use it - just install pipx and SOAP.

If you're using Conda, I highly recommend Mamba - it's much faster than Conda, can solve environments that Conda can't, and has an identical API. Install it with:

conda install -n base mamba

Copyright

Copyright (c) 2021, Josh Mitchell (Yoshanuikabundi)

Acknowledgements

Project based on the Computational Molecular Science Python Cookiecutter version 1.6.

About

Conda meets Cargo. Keep your conda dev environments local, specific, and up-to-date.

Resources

License

Code of conduct

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages