If you don't use gemsets and are just interested in my bundler flow, skip to the Bundler workflow section.
In my previous post about switching from rvm to ruby-install and chruby, I mentioned that my primary motivation for switching away from rvm is runaway gemset disk usage:
$ rvm disk-usage gemsets
Gemsets Usage: 5.4G
I'm generally working on a bunch of different ruby projects at once, and like every good rvm user, I would create a new gemset when I started or cloned the project because ... well, that's what you do! But the problem is that there's no straightforward way to eliminate orphan gemsets when
rming the project directory, and so you're left with 5.4GB of gemsets and you have no idea whether those are even relevant anymore.
Enter bundler, the tool that seems to become more and more magical with each version. You're probably already using bundler (if you aren't, WHAT?!) but letting it install into rvm's gemset folder hierarchy rather than directly into your project's directory. It's time to change that in just a few easy steps!
Inspired by Ryan McGeary's excellent "Vendor Everything" post from 2011, I set out to come up with my own bundler workflow.
I've configured bundler to install gems to my project's
vendor/bundle directory. That way, if I remove the project directory, all of the gems that I've installed specifically for it are gone as well. No more hunting down orphaned gemsets.
bundle config --global path vendor/bundle
This tells bundler to install gems to
vendor/bundle by default (rather than your global gem store). You can of course always install gems that you need globally (like bundler) by using
Rails 4+ will generate binstubs for
bin/. With binstubs, you can run
bin/rake instead of
bundle exec rake, for example. If you want to make it really simple, just add
./bin to the front of your $PATH and your shell will always look there first. To generate binstubs for any other gems, like spring, just run
bundle binstub <gemname> and it will create one in
bin/ for you.
This is a little bit tricky with chruby because it always puts its gem bin directory at the front of $PATH so you need to add it in an after hook, which isn't officially supported. This is how I accomplished that with my zsh/chruby setup.
You'll also want to add the following line to your
.gitignore if it isn't there already:
Bonus points: parallelize bundler
First, make sure you update to bundler >= 1.5.0
If you're using Linux, run
nproc. If you're using a Mac, run
sysctl -n hw.ncpu. If you're using Windows, then god help you...
Take the output of that command and replace the X below with it. This will tell bundler to always spawn X number of processes to install gems in parallel. Your
bundle install runs will be much faster.
bundle config --global jobs X
You might also be interested in these articles...