class Gem::Indexer

Top level class for building the gem repository index.

Attributes

build_modern[RW]

Build indexes for RubyGems 1.2.0 and newer when true

dest_directory[R]

Index install location

dest_latest_specs_index[R]

Latest specs index install location

dest_prerelease_specs_index[R]

Prerelease specs index install location

dest_specs_index[R]

Specs index install location

directory[R]

Index build directory

Public Class Methods

new(directory, options = {}) click to toggle source

Create an indexer that will index the gems in directory.

# 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

Public Instance Methods

abbreviate(spec) click to toggle source

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.

# 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() click to toggle source

Build various indicies

# 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() click to toggle source

Builds Marshal quick index gemspecs.

# 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) click to toggle source

Build a single index for RubyGems 1.2 and newer

# 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() click to toggle source

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

# 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) click to toggle source

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

# 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) click to toggle source

Compress filename with extension.

# 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() click to toggle source

Compresses indicies on disk

# 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() click to toggle source

List of gem file names to index.

# File lib/rubygems/indexer.rb, line 301
def gem_file_list
  Dir[File.join(@dest_directory, "gems", '*.gem')]
end
generate_index() click to toggle source

Builds and installs indicies.

# 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) click to toggle source

Zlib::GzipWriter wrapper that gzips filename on disk.

# 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() click to toggle source

Install generated indicies into the destination directory.

# 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() click to toggle source

Make directories for index generation

# 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) click to toggle source
# 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 => e
      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) click to toggle source

Ensure path and path with extension are identical.

# 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) click to toggle source

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.

# 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) click to toggle source

Sanitize a single string.

# 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() click to toggle source

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

# 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) click to toggle source

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.

# 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