Class: Gem::Indexer

Relationships & Source Files
Super Chains via Extension / Inclusion / Inheritance
Instance Chain: self, UserInteraction, DefaultUserInteraction
Inherits:	Object
Defined in:	lib/rubygems/indexer.rb

Overview

Top level class for building the gem repository index.

Class Method Summary

.new(directory, options = {}) ⇒ Indexer constructor

Create an indexer that will index the gems in #directory.

Instance Attribute Summary

#build_modern rw

Build indexes for RubyGems 1.2.0 and newer when true.
#dest_directory readonly

Index install location.
#dest_latest_specs_index readonly

Latest specs index install location.
#dest_prerelease_specs_index readonly

Prerelease specs index install location.
#dest_specs_index readonly

Specs index install location.
#directory readonly

Index build directory.

DefaultUserInteraction - Included

#ui	See DefaultUserInteraction.ui
#ui=	See DefaultUserInteraction.ui=

Instance Method Summary

#abbreviate(spec)

Abbreviate the spec for downloading.
#build_indicies

Build various indicies.
#build_marshal_gemspecs

Builds Marshal quick index gemspecs.
#build_modern_index(index, file, name)

Build a single index for RubyGems 1.2 and newer.
#build_modern_indicies

Builds indicies for RubyGems 1.2 and newer.
#compact_specs(specs)

Compacts Marshal output for the specs index data source by using identical objects as much as possible.
#compress(filename, extension)

Compress filename with extension.
#compress_indicies

Compresses indicies on disk.
#gem_file_list

List of gem file names to index.
#generate_index

Builds and installs indicies.
#gzip(filename)

Zlib::GzipWriter wrapper that gzips filename on disk.
#install_indicies

Install generated indicies into the destination directory.
#make_temp_directories

Make directories for index generation.
#map_gems_to_specs(gems)
#paranoid(path, extension)

Ensure path and path with extension are identical.
#sanitize(spec)

Sanitize the descriptive fields in the spec.
#sanitize_string(string)

Sanitize a single string.
#update_index

Perform an in-place update of the repository from newly added gems.
#update_specs_index(index, source, dest)

Combines specs in index and source then writes out a new copy to dest.

UserInteraction - Included

#alert	Displays an alert `statement`.
#alert_error	Displays an error `statement` to the error output location.
#alert_warning	Displays a warning `statement` to the warning output location.
#ask	Asks a `question` and returns the answer.
#ask_for_password	Asks for a password with a `prompt`
#ask_yes_no	Asks a yes or no `question`.
#choose_from_list	Asks the user to answer `question` with an answer from the given `list`.
#say	Displays the given `statement` on the standard output (or equivalent).
#terminate_interaction	Terminates the RubyGems process with the given `exit_code`
#verbose	Calls `say` with `msg` or the results of the block if really_verbose is true.

DefaultUserInteraction - Included

#use_ui

See DefaultUserInteraction.use_ui

Constructor Details

.new(directory, options = {}) ⇒ `Indexer`

Create an indexer that will index the gems in #directory.

[ GitHub ]

# File 'lib/rubygems/indexer.rb', line 51


def initialize(directory, options = {})
  require 'fileutils'
  require 'tmpdir'
  require 'zlib'

  unless defined?(Builder::XChar) then
    raise "Gem::Indexer requires that the XML Builder library be installed:" +
          "\n\tgem install builder"
  end

  options = { :build_modern => true }.merge options

  @build_modern = options[:build_modern]

  @dest_directory = directory
  @directory = File.join(Dir.tmpdir, "gem_generate_index_#{$$}")

  marshal_name = "Marshal.#{Gem.marshal_version}"

  @master_index = File.join @directory, 'yaml'
  @marshal_index = File.join @directory, marshal_name

  @quick_dir = File.join @directory, 'quick'
  @quick_marshal_dir = File.join @quick_dir, marshal_name
  @quick_marshal_dir_base = File.join "quick", marshal_name # FIX: UGH

  @quick_index = File.join @quick_dir, 'index'
  @latest_index = File.join @quick_dir, 'latest_index'

  @specs_index = File.join @directory, "specs.#{Gem.marshal_version}"
  @latest_specs_index =
    File.join(@directory, "latest_specs.#{Gem.marshal_version}")
  @prerelease_specs_index =
    File.join(@directory, "prerelease_specs.#{Gem.marshal_version}")
  @dest_specs_index =
    File.join(@dest_directory, "specs.#{Gem.marshal_version}")
  @dest_latest_specs_index =
    File.join(@dest_directory, "latest_specs.#{Gem.marshal_version}")
  @dest_prerelease_specs_index =
    File.join(@dest_directory, "prerelease_specs.#{Gem.marshal_version}")

  @files = []
end

Instance Attribute Details

#build_modern (rw)

Build indexes for RubyGems 1.2.0 and newer when true

[ GitHub ]

# File 'lib/rubygems/indexer.rb', line 21


attr_accessor :build_modern

#dest_directory (readonly)

Index install location

[ GitHub ]

# File 'lib/rubygems/indexer.rb', line 26


attr_reader :dest_directory

#dest_latest_specs_index (readonly)

Latest specs index install location

[ GitHub ]

# File 'lib/rubygems/indexer.rb', line 36


attr_reader :dest_latest_specs_index

#dest_prerelease_specs_index (readonly)

Prerelease specs index install location

[ GitHub ]

# File 'lib/rubygems/indexer.rb', line 41


attr_reader :dest_prerelease_specs_index

#dest_specs_index (readonly)

Specs index install location

[ GitHub ]

# File 'lib/rubygems/indexer.rb', line 31


attr_reader :dest_specs_index

#directory (readonly)

Index build directory

[ GitHub ]

# File 'lib/rubygems/indexer.rb', line 46


attr_reader :directory

Instance Method Details

#abbreviate(spec)

Abbreviate the spec for downloading. Abbreviated specs are only used for searching, downloading and related activities and do not need deployment specific information (e.g. list of files). So we abbreviate the spec, making it much smaller for quicker downloads.

[ GitHub ]

# File 'lib/rubygems/indexer.rb', line 103


def abbreviate(spec)
  spec.files = []
  spec.test_files = []
  spec.rdoc_options = []
  spec.extra_rdoc_files = []
  spec.cert_chain = []
  spec
end

#build_indicies

Build various indicies

[ GitHub ]

# File 'lib/rubygems/indexer.rb', line 115


def build_indicies
  Gem::Specification.dirs = []
  Gem::Specification.add_specs(*map_gems_to_specs(gem_file_list))

  build_marshal_gemspecs
  build_modern_indicies if @build_modern

  compress_indicies
end

#build_marshal_gemspecs

Builds Marshal quick index gemspecs.

[ GitHub ]

# File 'lib/rubygems/indexer.rb', line 128


def build_marshal_gemspecs
  count = Gem::Specification.count { |s| not s.default_gem? }
  progress = ui.progress_reporter count,
                                  "Generating Marshal quick index gemspecs for #{count} gems",
                                  "Complete"

  files = []

  Gem.time 'Generated Marshal quick index gemspecs' do
    Gem::Specification.each do |spec|
      next if spec.default_gem?
      spec_file_name = "#{spec.original_name}.gemspec.rz"
      marshal_name = File.join @quick_marshal_dir, spec_file_name

      marshal_zipped = Gem.deflate Marshal.dump(spec)
      open marshal_name, 'wb' do |io| io.write marshal_zipped end

      files << marshal_name

      progress.updated spec.original_name
    end

    progress.done
  end

  @files << @quick_marshal_dir

  files
end

#build_modern_index(index, file, name)

Build a single index for RubyGems 1.2 and newer

[ GitHub ]

# File 'lib/rubygems/indexer.rb', line 161


def build_modern_index(index, file, name)
  say "Generating #{name} index"

  Gem.time "Generated #{name} index" do
    open(file, 'wb') do |io|
      specs = index.map do |*spec|
        # We have to splat here because latest_specs is an array, while the
        # others are hashes.
        spec = spec.flatten.last
        platform = spec.original_platform

        # win32-api-1.0.4-x86-mswin32-60
        unless String === platform then
          alert_warning "Skipping invalid platform in gem: #{spec.full_name}"
          next
        end

        platform = Gem::Platform::RUBY if platform.nil? or platform.empty?
        [spec.name, spec.version, platform]
      end

      specs = compact_specs(specs)
      Marshal.dump(specs, io)
    end
  end
end

#build_modern_indicies

Builds indicies for RubyGems 1.2 and newer. Handles full, latest, prerelease

[ GitHub ]

# File 'lib/rubygems/indexer.rb', line 191


def build_modern_indicies
  specs = Gem::Specification.reject { |s| s.default_gem? }

  prerelease, released = specs.partition { |s|
    s.version.prerelease?
  }
  latest_specs =
    Gem::Specification.latest_specs.reject { |s| s.default_gem? }

  build_modern_index(released.sort, @specs_index, 'specs')
  build_modern_index(latest_specs.sort, @latest_specs_index, 'latest specs')
  build_modern_index(prerelease.sort, @prerelease_specs_index,
                     'prerelease specs')

  @files += [@specs_index,
             "#{@specs_index}.gz",
             @latest_specs_index,
             "#{@latest_specs_index}.gz",
             @prerelease_specs_index,
             "#{@prerelease_specs_index}.gz"]
end

#compact_specs(specs)

Compacts Marshal output for the specs index data source by using identical objects as much as possible.

[ GitHub ]

# File 'lib/rubygems/indexer.rb', line 271


def compact_specs(specs)
  names = {}
  versions = {}
  platforms = {}

  specs.map do |(name, version, platform)|
    names[name] = name unless names.include? name
    versions[version] = version unless versions.include? version
    platforms[platform] = platform unless platforms.include? platform

    [names[name], versions[version], platforms[platform]]
  end
end

#compress(filename, extension)

Compress filename with extension.

[ GitHub ]

# File 'lib/rubygems/indexer.rb', line 288


def compress(filename, extension)
  data = Gem.read_binary filename

  zipped = Gem.deflate data

  open "#{filename}.#{extension}", 'wb' do |io|
    io.write zipped
  end
end

#compress_indicies

Compresses indicies on disk

[ GitHub ]

# File 'lib/rubygems/indexer.rb', line 255


def compress_indicies
  say "Compressing indicies"

  Gem.time 'Compressed indicies' do
    if @build_modern then
      gzip @specs_index
      gzip @latest_specs_index
      gzip @prerelease_specs_index
    end
  end
end

#gem_file_list

List of gem file names to index.

[ GitHub ]

# File 'lib/rubygems/indexer.rb', line 301


def gem_file_list
  Dir[File.join(@dest_directory, "gems", '*.gem')]
end

#generate_index

Builds and installs indicies.

[ GitHub ]

# File 'lib/rubygems/indexer.rb', line 308


def generate_index
  make_temp_directories
  build_indicies
  install_indicies
rescue SignalException
ensure
  FileUtils.rm_rf @directory
end

#gzip(filename)

Zlib::GzipWriter wrapper that gzips filename on disk.

[ GitHub ]

# File 'lib/rubygems/indexer.rb', line 320


def gzip(filename)
  Zlib::GzipWriter.open "#{filename}.gz" do |io|
    io.write Gem.read_binary(filename)
  end
end

#install_indicies

Install generated indicies into the destination directory.

[ GitHub ]

# File 'lib/rubygems/indexer.rb', line 329


def install_indicies
  verbose = Gem.configuration.really_verbose

  say "Moving index into production dir #{@dest_directory}" if verbose

  files = @files
  files.delete @quick_marshal_dir if files.include? @quick_dir

  if files.include? @quick_marshal_dir and not files.include? @quick_dir then
    files.delete @quick_marshal_dir

    dst_name = File.join(@dest_directory, @quick_marshal_dir_base)

    FileUtils.mkdir_p File.dirname(dst_name), :verbose => verbose
    FileUtils.rm_rf dst_name, :verbose => verbose
    FileUtils.mv(@quick_marshal_dir, dst_name,
                 :verbose => verbose, :force => true)
  end

  files = files.map do |path|
    path.sub(/^#{Regexp.escape @directory}\/?/, '') # HACK?
  end

  files.each do |file|
    src_name = File.join @directory, file
    dst_name = File.join @dest_directory, file

    FileUtils.rm_rf dst_name, :verbose => verbose
    FileUtils.mv(src_name, @dest_directory,
                 :verbose => verbose, :force => true)
  end
end

#make_temp_directories

Make directories for index generation

[ GitHub ]

# File 'lib/rubygems/indexer.rb', line 365


def make_temp_directories
  FileUtils.rm_rf @directory
  FileUtils.mkdir_p @directory, :mode => 0700
  FileUtils.mkdir_p @quick_marshal_dir
end

#map_gems_to_specs(gems)

[ GitHub ]

# File 'lib/rubygems/indexer.rb', line 213


def map_gems_to_specs gems
  gems.map { |gemfile|
    if File.size(gemfile) == 0 then
      alert_warning "Skipping zero-length gem: #{gemfile}"
      next
    end

    begin
      spec = Gem::Package.new(gemfile).spec
      spec.loaded_from = gemfile

      # HACK: fuck this shit - borks all tests that use pl1
      # if File.basename(gemfile, ".gem") != spec.original_name then
      #   exp = spec.full_name
      #   exp << " (#{spec.original_name})" if
      #     spec.original_name != spec.full_name
      #   msg = "Skipping misnamed gem: #{gemfile} should be named #{exp}"
      #   alert_warning msg
      #   next
      # end

      abbreviate spec
      sanitize spec

      spec
    rescue SignalException
      alert_error "Received signal, exiting"
      raise
    rescue Exception => e
      msg = ["Unable to process #{gemfile}",
             "#{e.message} (#{e.class})",
             "\t#{e.backtrace.join "\n\t"}"].join("\n")
      alert_error msg
    end
  }.compact
end

#paranoid(path, extension)

Ensure Gem.path and path with extension are identical.

[ GitHub ]

# File 'lib/rubygems/indexer.rb', line 374


def paranoid(path, extension)
  data = Gem.read_binary path
  compressed_data = Gem.read_binary "#{path}.#{extension}"

  unless data == Gem.inflate(compressed_data) then
    raise "Compressed file #{compressed_path} does not match uncompressed file #{path}"
  end
end

#sanitize(spec)

Sanitize the descriptive fields in the spec. Sometimes non-ASCII characters will garble the site index. Non-ASCII characters will be replaced by their XML entity equivalent.

[ GitHub ]

# File 'lib/rubygems/indexer.rb', line 388


def sanitize(spec)
  spec.summary              = sanitize_string(spec.summary)
  spec.description          = sanitize_string(spec.description)
  spec.post_install_message = sanitize_string(spec.post_install_message)
  spec.authors              = spec.authors.collect { |a| sanitize_string(a) }

  spec
end

#sanitize_string(string)

Sanitize a single string.

[ GitHub ]

# File 'lib/rubygems/indexer.rb', line 400


def sanitize_string(string)
  return string unless string

  # HACK the #to_s is in here because RSpec has an Array of Arrays of
  # Strings for authors.  Need a way to disallow bad values on gemspec
  # generation.  (Probably won't happen.)
  string = string.to_s

  begin
    Builder::XChar.encode string
  rescue NameError, NoMethodError
    string.to_xs
  end
end

#update_index

Perform an in-place update of the repository from newly added gems.

[ GitHub ]

# File 'lib/rubygems/indexer.rb', line 418


def update_index
  make_temp_directories

  specs_mtime = File.stat(@dest_specs_index).mtime
  newest_mtime = Time.at 0

  updated_gems = gem_file_list.select do |gem|
    gem_mtime = File.stat(gem).mtime
    newest_mtime = gem_mtime if gem_mtime > newest_mtime
    gem_mtime >= specs_mtime
  end

  if updated_gems.empty? then
    say 'No new gems'
    terminate_interaction 0
  end

  specs = map_gems_to_specs updated_gems
  prerelease, released = specs.partition { |s| s.version.prerelease? }

  Gem::Specification.dirs = []
  Gem::Specification.add_specs(*specs)

  files = build_marshal_gemspecs

  Gem.time 'Updated indexes' do
    update_specs_index released, @dest_specs_index, @specs_index
    update_specs_index released, @dest_latest_specs_index, @latest_specs_index
    update_specs_index(prerelease,
                       @dest_prerelease_specs_index,
                       @prerelease_specs_index)
  end

  compress_indicies

  verbose = Gem.configuration.really_verbose

  say "Updating production dir #{@dest_directory}" if verbose

  files << @specs_index
  files << "#{@specs_index}.gz"
  files << @latest_specs_index
  files << "#{@latest_specs_index}.gz"
  files << @prerelease_specs_index
  files << "#{@prerelease_specs_index}.gz"

  files = files.map do |path|
    path.sub(/^#{Regexp.escape @directory}\/?/, '') # HACK?
  end

  files.each do |file|
    src_name = File.join @directory, file
    dst_name = File.join @dest_directory, file # REFACTOR: duped above

    FileUtils.mv src_name, dst_name, :verbose => verbose,
                 :force => true

    File.utime newest_mtime, newest_mtime, dst_name
  end
end

#update_specs_index(index, source, dest)

Combines specs in index and source then writes out a new copy to dest. For a latest index, does not ensure the new file is minimal.

[ GitHub ]

# File 'lib/rubygems/indexer.rb', line 483


def update_specs_index(index, source, dest)
  specs_index = Marshal.load Gem.read_binary(source)

  index.each do |spec|
    platform = spec.original_platform
    platform = Gem::Platform::RUBY if platform.nil? or platform.empty?
    specs_index << [spec.name, spec.version, platform]
  end

  specs_index = compact_specs specs_index.uniq.sort

  open dest, 'wb' do |io|
    Marshal.dump specs_index, io
  end
end

Class: Gem::Indexer

Overview

Class Method Summary

Instance Attribute Summary

DefaultUserInteraction - Included

Instance Method Summary

UserInteraction - Included

DefaultUserInteraction - Included

Constructor Details

.new(directory, options = {}) ⇒ Indexer

Instance Attribute Details

#build_modern (rw)

#dest_directory (readonly)

#dest_latest_specs_index (readonly)

#dest_prerelease_specs_index (readonly)

#dest_specs_index (readonly)

#directory (readonly)

Instance Method Details

#abbreviate(spec)

#build_indicies

#build_marshal_gemspecs

#build_modern_index(index, file, name)

#build_modern_indicies

#compact_specs(specs)

#compress(filename, extension)

#compress_indicies

#gem_file_list

#generate_index

#gzip(filename)

#install_indicies

#make_temp_directories

#map_gems_to_specs(gems)

#paranoid(path, extension)

#sanitize(spec)

#sanitize_string(string)

#update_index

#update_specs_index(index, source, dest)

.new(directory, options = {}) ⇒ `Indexer`