Install and manage a standalone Python interpreter and dependencies.

Note: This project makes use of Anaconda, a Python distribution by Continuum Analytics. It is expected that the Anaconda End User License Agreement is agreed upon.

• Create per-project Python installations.
• Configure PyPi and Anaconda Cloud dependencies in a single package.json file.
• Create a standalone archive for offline installation on another machine.
• Supports a large number of 64bit Linux and OSX platforms.
• Supports a large number of Python versions.
• Supports a large number of precompiled binary dependencies (e.g. numpy, scipy) from Anaconda Cloud.

A small py.sh script is added to the root of your Python project. This is used to bootstrap a standalone Python interpreter in a hidden .pysh directory.

A standalone Python interpreter will be installed into .pysh.

curl -sL https://raw.githubusercontent.com/etianen/py.sh/master/install.sh | bash

Recommended: Add .pysh to your .gitignore file.

Advanced: The steps performed by the automated installer are documented in install.sh. You can perform them manually if preferred.

Hint: You can learn a lot from ./py.sh --help and ./py.sh <command name> --help.

Run any command with ./py.sh run <your command>.

For example, to fire up your standalone Python interpreter:

./py.sh run python

Or to install and run flake8:

./py.sh run pip install flake8
./py.sh run flake8 .

Rather than prefix every command with ./py.sh run, you can activate your environment with ./py.sh activate.

For example, to install and run flake8:

./py.sh activate
pip install flake8
flake8 .

You can specify the Python version, along with any Anaconda Cloud and PyPi dependencies, by creating a package.json file in the root of your project.

An example package.json file:

{
  "name": "awesome-package",
  "version": "0.1.0",
  "pysh": {
    "python": {
      "version": "3.5"
    },
    "conda": {
      "dependencies": {
        "psycopg2": "2.6.1"
      },
      "devDependencies": {
        "pytest": "2.8.5"
      }
    },
    "pip": {
      "dependencies": {
        "django": "1.9.2"
      },
      "devDependencies": {
        "flake8": "2.5.4"
      },
      "extra_index_urls": []
    },
    "install": [
      "my-post-install-command.sh",
      "python -c \"echo 'Everything installed OK'\""
    ]
  }
}

Whenever you update your package.json file, run ./py.sh install to rebuild your environment with the specified dependencies.

Note: Any commands in the install section will be run in your environment at the end of every ./py.sh install.

If you create a .env script in the root of your project, it will be sourced by the shell before any ./py.sh run commands are run, and at the start of every ./py.sh activate session.

Note: It's traditional to use this to configure environmental variables for your local environment. You'll probably want to add the .env script to your .gitignore file too.

It's possible to create a standalone archive of your entire project, including dependencies and Python interpreter, for offline installation on another machine.

./py.sh dist

You will find a zip archive in the ./dist folder of your project. Copy this to another machine of the same operating system and architecture, then unzip and install your project.

unzip your-project-1.0.0.zip -d your_project
cd your_project
./py.sh install --offline

Note: Offline installers work best if all binary dependencies (e.g. numpy, scipy) are obtained from Anaconda Cloud. Otherwise, large binary dependencies can take an extremely long time to compile during installation.

These things are otherwise difficult to do:

• Run a modern Python interpreter on an outdated version of Linux.
• Produce offline installers for a Python project.
• Run multiple version of Python on the same machine in production.

Probably, but I've never heard of XXX, or it didn't support all the features I wanted.

virtualenv is useful for managing a standalone bunch of PyPy dependencies. It doesn't help with installing Python interpreters not included in your operating system, and large binary dependencies can be a pain to install.

pyenv is useful for managing multiple Python versions in development, but it doesn't help with binary dependencies.

This project is just a convenience wrapper around Miniconda, so it hasn't really reinvented anything.

• Both pip and conda use completely different requirements file formats, and it's a pain to have to keep moving from one to the other.
• It's useful to be able to specify extra dependencies for development that shouldn't be included in production builds.
• There needs to be somewhere to specify the Python version.
• Being able to specify post-installation steps is really useful for reproducible installs.
• Many projects already have a package.json file for managing nodejs dependencies.

This project is built on every push using the Travis-CI service.

Downloads and bug tracking can be found at the main project website.

This project was developed by Dave Hall. You can get the code from the project site.

Dave Hall is a freelance web developer, based in Cambridge, UK. You can usually find him on the Internet:

• Website
• Google Profile