Rip: Ruby's Intelligent Packaging
Rip is an attempt to create a next generation packaging system for Ruby.
For more thorough documentation please see the Rip site:
Introduction
Let's get right to it then, shall we?
First we install rip.
$ sudo gem install rip
Did that work?
$ rip check
All systems go.
Yep. Let's see what libraries rip knows about.
$ rip list
ripenv: base
nothing installed
None. Really? Let's try to require Grit.
$ ruby -r grit -e 'puts Grit'
ruby: no such file to load -- grit (LoadError)
Whoops. Not found. Let's install the latest using rip.
$ rip install git://github.com/defunkt/grit.git v1.1.1b
Successfully installed grit v1.1.1
$ rip list
ripenv: base
diff-lcs (491fbc0)
mime-types (v1.16)
grit (v1.1.1)
Great, now we have Grit and all its dependencies.
$ ruby -r grit -e 'puts Grit'
Grit
$ ruby -r grit -e 'puts MIME::Types'
MIME::Types
And we don't need any magical require
statements to use them!
If we'd like, we can now move between rip environments to mix and match libraries.
$ rip env create my_fresh_env
ripenv: created my_fresh_env
$ ruby -r grit -e 'puts Grit'
ruby: no such file to load -- grit (LoadError)
And so much more.
Overview
Inspired by Python's virtualenv and pip, Rip aims to be a simple and powerful way to install and manage Ruby packages.
Multiple Package Support
Rip can install from a variety of sources: directories, single files, git repositories - even Rubygems.
Adding a new package is easy and we expect Rip to support more formats in the future.
Virtual Environments
Virtual environments ("ripenvs") can be created so multiple versions of a package may be installed and used by different applications concurrently.
ripenvs are easy to create, copy, delete, and share. Recipes for creating ripenvs are trivial to generate and publish.
Install-time Dependency Resolution
Dependency checking happens when Rip packages are installed, not when packages run.
A dependency graph is constructed for the entire virtual environment, making it easy to debug leaf-node transitive dependency issues.
Clear Error Messages
Installation and dependency errors should be clear and give as much information as possible in order to help you fix the problem.
Few Dependencies
Rip itself should work without anything but the Ruby standard library, for maximum portability.
Rip vs RubyGems
No building
Rip's support for a variety of package types means there is nothing to build and distribute.
Tag your Git repository and publicize the latest version, or just pass around Gists. Rip does not care.
Rip dependencies are listed as separate lines in a plaintext file and can reference any package type. As a result, Rip packages can depend on existing Rubygems that aren't available from any other source.
This means projects unaware of Rip can be installed by Rip and managed by ripenvs. Adding the dependencies yourself is easy.
Multiple Environments
Rip makes it easy to have multiple environments with different versions of libraries.
You could even clone a ripenv then upgrade a single library to test its impact on the environment as a whole. Installation not go smoothly? Delete the new ripenev then continue using the stable one.
Dependency Conflict Resolution
With Rip, version conflicts in dependencies are simpler to resolve: you know exactly what version of which libraries are requesting which versions of the same library at installation time. As a result conflicts are resolved when you're thinking about installing your code, not later on when you're thinking about running it.
Hands off
Rip requires no changes to your code, only an optional deps.rip
file added
to the root of your project. As a result you do not force Rip on anyone else
and individuals are free to re-package your code using other systems.
Distributed
There is no canonical server for Rip packages, which may be good or bad.
Installing Packages
Each Rip package optionally contains at the root a deps.rip
file
identifying it as a Rip package and listing its dependencies.
Let's take the ambition project as an example. This is its
deps.rip
in full:
git://github.com/drnic/rubigen.git REL-1.3.0
git://github.com/seattlerb/ruby2ruby.git e3cf57559 # 1.1.8
git://github.com/seattlerb/parsetree.git 480ede9d9 # 2.1.1
If you were to run rip install git://github.com/defunkt/ambition
the following steps would occur:
- The source would be fetched and unpacked as
ambition
in the cwd - The source of each dependency in
deps.rip
would be fetched - Each dependency would be unpacked into the current ripenv at the revision or tag specified
- Each dependency's
deps.rip
would be fetched and unpacked into the ripenv, etc
As this process unfolds, a mapping of libraries and versions is kept in memory. When a library is declared multiple times at different versions the process is halted and the error reported.
If you've cloned ambition
on your own you can still install the
dependencies using rip install deps.rip
Uninstalling Packages
The easiest way to mass uninstall packages is to delete your ripenv
and create a new one. Otherwise, rip uninstall package
will do the trick.
Rip will complain if you attempt to uninstall a package that others depend
on. To remove the package anyway, use -y
. To remove the package and the
dependents, use -d
(for dependents).
Extensions
Installing a package that is not Rip aware but needs an extension
built? Use rip build
.
Rip will attempt to run rake rip:install
in your library if a
Rakefile is found. If you need to compile your C extension or do
any other work, this is the place.
Rip Directory Structure
Rip is currently user-specific.
Here is a typical directory structure for Rip:
rip/
- rip-packages/
- active/
- bin/
- lib/
- base/
- bin/
- lib/
- base.ripenv
- cheat/
- bin/
- lib/
- cheat.ripenv
- thunderhorse/
- bin/
- lib/
- thunderhorse.ripenv
The above contains three ripenvs: base
, cheat
, and thunderhorse
. Each
ripenv contains directories for executable binaries and Ruby source files.
They also include a generated .ripenv
file containing metadata about
the ripenev and its packages.
This individual may use base
for general tomfoolery (it's the
default), cheat
for developing their Cheat application, and
thunderhorse
for working on their new Thunderhorse project.
active
is a symlink to the current, active ripenv. We also see a
rip-packages
directory. This is where Rip stores the raw repositories.
Let us focus on the cheat
ripenv:
rip/
- cheat/
- bin/
- camping
- lib/
- markaby/
- builder.rb
- cssproxy.rb
- metaid.rb
- rails.rb
- tags.rb
- template.rb
- markaby.rb
- camping/
- db.rb
- fastcgi.rb
- reloader.rb
- session.rb
- webrick.rb
- camping-unabridged.rb
- camping.rb
- cheat.ripenv
When using the cheat
ripenv, a camping
binary will be in our PATH
.
When running Ruby scripts, or even executing the camping
binary,
rip/cheat/lib
will be in our $LOAD_PATH. Therefor we may, for instance,
require "markaby"
in a Ruby script and it will succeed.
In any other ripenv, the cheat
ripenv's binaries and libraries are as
good as non-existant.
Deployment
Want to get a copy of your local environment on your deployment
server? Generate a .rip
file with rip freeze
then upload and
install it.
Shortcomings
Currently it's UNIX-only. This is because Rip needs to manipulate the RUBYLIB and PATH environment variables so that Ruby knows where to find installed Rip packages.
As a result, the setup script expects you to be running bash or zshell.
Contributors
- Chris Wanstrath (that's me!)
- Jeff Hodges
- Tom Preston-Werner
- John Barnette
- Blake Mizerany
- Ryan Tomayko
- Pat Nakajima
Special Thanks
- Coda Hale for the phrase "leaf-node transitive dependency issues"
- GitHub for sponsoring development