What are the Python equivalents to Ruby's bundler / Perl's carton?

Question

I know about virtualenv and pip. But these are a bit different from bundler/carton.

For instance:

• pip writes the absolute path to shebang or activate script
• pip doesn't have the exec sub command (bundle exec bar)
• virtualenv copies the Python interpreter to a local directory

Does every Python developer use virtualenv/pip? Are there other package management tools for Python?

Solution 1

From what i've read about bundler — pip without virtualenv should work just fine for you. You can think of it as something between regular gem command and bundler. Common things that you can do with pip:

1. Installing packages (gem install)

   pip install mypackage
   

2. Dependencies and bulk-install (gemfile)

   Probably the easiest way is to use pip's requirements.txt files. Basically it's just a plain list of required packages with possible version constraints. It might look something like:

   nose==1.1.2
   django<1.3
   PIL
   

   Later when you'd want to install those dependencies you would do:

   $ pip install -r requirements.txt
   

   A simple way to see all your current packages in requirements-file syntax is to do:

   $ pip freeze
   

   You can read more about it here.

3. Execution (bundler exec)

   All python packages that come with executable files are usually directly available after install (unless you have custom setup or it's a special package). For example:

   $ pip install gunicorn
   $ gunicorn -h 
   

4. Package gems for install from cache (bundler package)

   There is pip bundle and pip zip/unzip. But i'm not sure if many people use it.

p.s. If you do care about environment isolation you can also use virtualenv together with pip (they are close friends and work perfectly together). By default pip installs packages system-wide which might require admin rights.

Solution 2

You can use pipenv, which has similar interface with bundler.

$ pip install pipenv

Pipenv creates virtualenv automatically and installs dependencies from Pipfile or Pipfile.lock.

$ pipenv --three           # Create virtualenv with Python3
$ pipenv install           # Install dependencies from Pipfile
$ pipenv install requests  # Install `requests` and update Pipfile
$ pipenv lock              # Generate `Pipfile.lock`
$ pipenv shell             # Run shell with virtualenv activated

You can run command with virtualenv scope like bundle exec.

$ pipenv run python3 -c "print('hello!')"

Solution 3

There is a clone pbundler.

The version that is currently in pip simply reads the requirements.txt file you already have, but is much out of date. It's also not totally equivalent: it insists on making a virtualenv. Bundler, I notice, only installs what packages are missing, and gives you the option of giving your sudo password to install into your system dirs or of restarting, which doesn't seem to be a feature of pbundler.

However, the version on git is an almost complete rewrite to be much closer to Bundler's behaviour... including having a "Cheesefile" and now not supporting requirements.txt. This is unfortunate, since requirements.txt is the de facto standard in pythonland, and there's even Offical BDFL-stamped work to standardize it. When that comes into force, you can be sure that something like pbundler will become the de facto standard. Alas, nothing quite stable yet that I know of (but I would love to be proven wrong).

Solution 4

I wrote one — https://github.com/Deepwalker/pundler . On PIP its pundle, name was already taken.

It uses requirements(_\w+)?.txt files as your desired dependencies and creates frozen(_\w+)?.txt files with frozen versions.

About (_\w+)? thing — this is envs. You can create requirements_test.txt and then use PUNDLEENV=test to use this deps in your run with requirements.txt ones alongside.

And about virtualenv – you need not one, its what pundle takes from bundler in first head.

Solution 5

Python Poetry is the closest to Ruby bundler as of 2020 (and already since 2018). It's already more than two years old, still very active, has great documentation. One might complain about curl-pipe-python-style being the recommended way of installing, but there are alternatives, e.g. homebrew on macOS.

• Primary website: https://python-poetry.org/
• Github: https://github.com/python-poetry/poetry
• Documentation: https://python-poetry.org/docs/

It uses virtualenvs behind the scenes (in contrast to bundler), but it provides and uses a lock-file, takes care of sub dependencies, adheres to specified version constraints and allows automatically updating outdated packages. There's even autocompletion for your favorite shell.

With its use of a pyproject.toml file, it's also going a bit further than bundler (closer to a gemspec. It's also comparable to JavaScript's and TypeScript's npm and yarn).

Poetrify (a complementing project) helps converting projects from requirements.txt to pyproject.toml for Poetry.

The lock file can be exported to requirements.txt by poetry export -f requirements.txt > requirements.txt, if you need that for other tooling (or the unlikely case want to go back).

View more solutions