How Do Gems Work?

Most of the time, gems in Ruby Just Work. But there’s a big problem with Ruby magic: when things go wrong, it’s hard to find out why.

You won’t often run into problems with your gems. But when you do, Google is surprisingly unhelpful. The error messages are generic, so they could have one of many different causes. And if you don’t understand how gems actually work with Ruby, you’re going to have a tough time debugging these problems on your own.

Gems might seem magical. But with a little investigation, they’re pretty easy to understand.

What does gem install do?

A Ruby gem is just some code zipped up with a little extra data. You can see the code inside a gem with gem unpack:

~/Source/playground jweiss$ gem unpack resque_unit
Fetching: resque_unit-0.4.8.gem (100%)
Unpacked gem: '/Users/jweiss/Source/playground/resque_unit-0.4.8'
~/Source/playground jweiss$ cd resque_unit-0.4.8 
~/Source/playground/resque_unit-0.4.8 jweiss$ find .
~/Source/playground/resque_unit-0.4.8 jweiss$

gem install, in its simplest form, does something kind of like this. It grabs the gem and puts its files into a special directory on your system. You can see where gem install will install your gems if you run gem environment (look for the INSTALLATION DIRECTORY: line):

~ jweiss$ gem environment
RubyGems Environment:
  - RUBY VERSION: 2.1.2 (2014-05-08 patchlevel 95) [x86_64-darwin14.0]
  - INSTALLATION DIRECTORY: /usr/local/Cellar/ruby/2.1.2/lib/ruby/gems/2.1.0
~ jweiss$ ls /usr/local/Cellar/ruby/2.1.2/lib/ruby/gems/2.1.0
bin		    bundler		doc		    gems
build_info	cache		extensions	specifications

All of your installed gem code will be there, under the gems directory.

These paths vary from system to system, and they also depend on how you installed Ruby (rvm is different from Homebrew, which is different from rbenv, and so on). So gem environment will be helpful when you want to know where your gems’ code lives.

How does gem code get required?

To help you use the code inside of your gems, RubyGems overrides Ruby’s require method. (It does this in core_ext/kernel_require.rb). The comment is pretty clear:

  # When RubyGems is required, Kernel#require is replaced with our own which
  # is capable of loading gems on demand.
  # When you call <tt>require 'x'</tt>, this is what happens:
  # * If the file can be loaded from the existing Ruby loadpath, it
  #   is.
  # * Otherwise, installed gems are searched for a file that matches.
  #   If it's found in gem 'y', that gem is activated (added to the
  #   loadpath).

For example, say you wanted to load active_support. RubyGems will try to require it using Ruby’s require method. That gives you this error:

LoadError: cannot load such file -- active_support
	from (irb):17:in `require'
	from (irb):17
	from /usr/local/bin/irb:11:in `<main>'

RubyGems looks at that error message, and sees that it needs to find active_support.rb inside of a gem, instead. To do that, it’ll scan through its gems’ metadata, looking for a gem that contains active_support.rb:

irb(main):001:0> spec = Gem::Specification.find_by_path('active_support')
=> #<Gem::Specification:0x3fe366874324 activesupport-4.2.0.beta1>

Then, it activates the gem, which adds the code inside the gem to Ruby’s load path (the directories you can require files from):

irb(main):002:0> $LOAD_PATH
=> ["/usr/local/Cellar/ruby/2.1.2/lib/ruby/site_ruby/2.1.0", "/usr/local/Cellar/ruby/2.1.2/lib/ruby/site_ruby/2.1.0/x86_64-darwin14.0", "/usr/local/Cellar/ruby/2.1.2/lib/ruby/site_ruby", "/usr/local/Cellar/ruby/2.1.2/lib/ruby/vendor_ruby/2.1.0", "/usr/local/Cellar/ruby/2.1.2/lib/ruby/vendor_ruby/2.1.0/x86_64-darwin14.0", "/usr/local/Cellar/ruby/2.1.2/lib/ruby/vendor_ruby", "/usr/local/Cellar/ruby/2.1.2/lib/ruby/2.1.0", "/usr/local/Cellar/ruby/2.1.2/lib/ruby/2.1.0/x86_64-darwin14.0"]
irb(main):003:0> spec.activate
=> true
irb(main):004:0> $LOAD_PATH
=> ["/usr/local/Cellar/ruby/2.1.2/lib/ruby/gems/2.1.0/gems/i18n-0.7.0.beta1/lib", "/usr/local/Cellar/ruby/2.1.2/lib/ruby/gems/2.1.0/gems/thread_safe-0.3.4/lib", "/usr/local/Cellar/ruby/2.1.2/lib/ruby/gems/2.1.0/gems/activesupport-4.2.0.beta1/lib", "/usr/local/Cellar/ruby/2.1.2/lib/ruby/site_ruby/2.1.0", "/usr/local/Cellar/ruby/2.1.2/lib/ruby/site_ruby/2.1.0/x86_64-darwin14.0", "/usr/local/Cellar/ruby/2.1.2/lib/ruby/site_ruby", "/usr/local/Cellar/ruby/2.1.2/lib/ruby/vendor_ruby/2.1.0", "/usr/local/Cellar/ruby/2.1.2/lib/ruby/vendor_ruby/2.1.0/x86_64-darwin14.0", "/usr/local/Cellar/ruby/2.1.2/lib/ruby/vendor_ruby", "/usr/local/Cellar/ruby/2.1.2/lib/ruby/2.1.0", "/usr/local/Cellar/ruby/2.1.2/lib/ruby/2.1.0/x86_64-darwin14.0"]

Now that active_support is on the load path, you can require the files in the gem just like any other piece of Ruby code. You can even use the original version of require, the one that was overwritten by RubyGems:

irb(main):005:0> gem_original_require 'active_support'
=> true


A little knowledge goes a long way

RubyGems might seem complicated. But at its most basic level, it’s just managing Ruby’s load path for you. That’s not to say it’s all easy. I didn’t cover how RubyGems manages version conflicts between gems, gem binaries (like rails and rake), C extensions, and lots of other stuff.

But knowing RubyGems even at this surface level will help out a lot. With a little code reading and playing in irb, you’ll be able to dive into the source of your gems. You can understand where your gems live, so you can make sure that RubyGems knows about them. And once you know how gems are loaded, you can dig into some crazy-looking loading problems.

Did you like this article? You should read these:

Finished another Rails tutorial and still don't know how to start?

Have you slogged through the same guide three times and still can't retain enough to write apps on your own?

In my free 7-part course, you’ll discover the fastest way to learn and remember new Rails ideas, so you can use them when you need them. And you'll learn to use what you already know to build your own Rails project.