class Bundler::Installer

Attributes

post_install_messages[RW]

Public Class Methods

install(root, definition, options = {}) click to toggle source

Begins the installation process for Bundler. For more information see the run method on this class.

# File lib/bundler/installer.rb, line 12
def self.install(root, definition, options = {})
  installer = new(root, definition)
  installer.run(options)
  installer
end

Public Instance Methods

run(options) click to toggle source

Runs the install procedures for a specific Gemfile.

Firstly, this method will check to see if Bundler.bundle_path exists and if not then will create it. This is usually the location of gems on the system, be it RVM or at a system path.

Secondly, it checks if Bundler has been configured to be “frozen” Frozen ensures that the Gemfile and the Gemfile.lock file are matching. This stops a situation where a developer may update the Gemfile but may not run `bundle install`, which leads to the Gemfile.lock file not being correctly updated. If this file is not correctly updated then any other developer running `bundle install` will potentially not install the correct gems.

Thirdly, Bundler checks if there are any dependencies specified in the Gemfile using Bundler::Environment#dependencies. If there are no dependencies specified then Bundler returns a warning message stating so and this method returns.

Fourthly, Bundler checks if the default lockfile (Gemfile.lock) exists, and if so then proceeds to set up a defintion based on the default gemfile (Gemfile) and the default lock file (Gemfile.lock). However, this is not the case if the platform is different to that which is specified in Gemfile.lock, or if there are any missing specs for the gems.

Fifthly, Bundler resolves the dependencies either through a cache of gems or by remote. This then leads into the gems being installed, along with stubs for their executables, but only if the –binstubs option has been passed or Bundler.options has been set earlier.

Sixthly, a new Gemfile.lock is created from the installed gems to ensure that the next time that a user runs `bundle install` they will receive any updates from this process.

Finally: TODO add documentation for how the standalone process works.

# File lib/bundler/installer.rb, line 49
def run(options)
  # Create the BUNDLE_PATH directory
  begin
    Bundler.bundle_path.mkpath unless Bundler.bundle_path.exist?
  rescue Errno::EEXIST
    raise PathError, "Could not install to path `#{Bundler.settings[:path]}` " +
      "because of an invalid symlink. Remove the symlink so the directory can be created."
  end

  if Bundler.settings[:frozen]
    @definition.ensure_equivalent_gemfile_and_lockfile(options[:deployment])
  end

  if dependencies.empty?
    Bundler.ui.warn "The Gemfile specifies no dependencies"
    lock
    return
  end

  if Bundler.default_lockfile.exist? && !options["update"]
    begin
      tmpdef = Definition.build(Bundler.default_gemfile, Bundler.default_lockfile, nil)
      local = true unless tmpdef.new_platform? || tmpdef.missing_specs.any?
    rescue BundlerError
    end
  end

  # Since we are installing, we can resolve the definition
  # using remote specs
  unless local
    options["local"] ?
      @definition.resolve_with_cache! :
      @definition.resolve_remotely!
  end

  # Must install gems in the order that the resolver provides
  # as dependencies might actually affect the installation of
  # the gem.
  Installer.post_install_messages = {}
  specs.each do |spec|
    install_gem_from_spec(spec, options[:standalone])
  end

  lock
  generate_standalone(options[:standalone]) if options[:standalone]
end