hollownest-ruby-aws 0.0.0

data/README.rdoc

@@ -0,0 +1,138 @@
+#--
+# $Id: README.rdoc,v 1.21 2009/02/20 00:45:17 ianmacd Exp $
+#++
+#
+#
+# = Ruby/AWS - A Ruby interface to the Amazon Associates Web Services API.
+# 
+# == Introduction
+# 
+# Ruby/AWS is a Ruby language library that allows programmatic access to
+# the popular Amazon Web sites via the AWS v4 API. It is the successor to the
+# now obsolete Ruby/Amazon.
+#
+# In addition to the original
+# amazon.com[http://www.amazon.com/exec/obidos/redirect-home/calibanorg-20]
+# site, the local sites
+# amazon.co.uk[http://www.amazon.co.uk/exec/obidos/redirect-home/caliban-21],
+# amazon.de[http://www.amazon.de/exec/obidos/redirect-home/calibanorg0a-21],
+# amazon.fr[http://www.amazon.fr/exec/obidos/redirect-home/caliban08-21],
+# amazon.ca[http://www.amazon.ca/exec/obidos/redirect-home/caliban-20] and
+# amazon.co.jp[http://www.amazon.co.jp/exec/obidos/redirect-home/calibanorg-20]
+# are also supported.
+#
+# Although the library is still in development, the AWS v4 API is now more or
+# less fully supported, with only tiny gaps in the functionality of some
+# operations.
+#
+# The following operations are supported:
+#
+#  BrowseNodeLookup
+#  CustomerContentLookup
+#  CustomerContentSearch
+#  Help
+#  ItemLookup
+#  ItemSearch
+#  ListLookup
+#  ListSearch
+#  SellerListingLookup
+#  SellerListingSearch
+#  SellerLookup
+#  SimilarityLookup
+#  TagLookup
+#  TransactionLookup
+#  VehiclePartLookup
+#  VehiclePartSearch
+#  VehicleSearch
+#
+# Remote shopping-carts are also supported. This adds the following operations:
+#
+#  CartCreate
+#  CartAdd
+#  CartModify
+#  CartClear
+#  CartGet
+# 
+# In addition, multiple operations and batch requests are also supported.
+#
+# Ruby/AWS also offers advanced features not directly available in the AWS
+# API, such as the ability to retrieve *all* results pages for a particular
+# search, rather than having to manually deal with AWS responses of 10 results
+# per page.
+#
+# You can also retrieve product images and optionally overlay them with
+# percentage discount icons.
+#
+# Another advanced feature is the ability to cache responses returned by AWS.
+# If the cache is used (as it is by default), the results of each unique
+# query will be cached and used for 24 hours. The cache can be manually
+# flushed of all or just the expired entries.
+#
+# One other useful advanced feature is the ability to determine the
+# appropriate Amazon locale for a given client, based on its IP address or
+# host name. This allows you to perform AWS operations using the correct
+# geographical Amazon site for any given client. German and Austrian clients
+# can be made to interact with amazon.de, British and Irish clients with
+# amazon.co.uk, etc.
+#
+#
+# == Installation
+#
+# Please see the +INSTALL+ file supplied with the software for details of how 
+# to install Ruby/AWS. You can choose between an installation script and a
+# RubyGems[http://www.rubygems.org/] installation.
+#
+# Note, however, if choosing the gem installation, that whilst Ruby/AWS's
+# RubyForge UNIX name is now ruby-aaws. The ruby-aws name was taken by
+# {another project}[http://rubyforge.org/projects/ruby-aws/] and this clash
+# prevented remote installation of the Ruby/AWS gem.
+#
+#
+# == Prerequisites
+#
+# Before you can use this library, you need to obtain an Amazon Web Services
+# {access key
+# ID}[https://aws-portal.amazon.com/gp/aws/developer/registration/index.html].
+#
+# You should also apply for an {Associates
+# account}[http://docs.amazonwebservices.com/AWSECommerceService/2009-01-06/GSG/BecominganAssociate.html],
+# although this isn't strictly necessary. If you do not explicitly provide an
+# Associates tag in your calls through Ruby/AWS, the tag of the Ruby/AWS
+# author will be used by default.
+#
+# 
+# == See Also
+# 
+# Ultimately, the way to get the most from this library is to read the AWS
+# documentation to get a feel for what is possible, and then experiment with
+# this library to see how the AWS calls are mapped into the Ruby world. You
+# should also review this library's
+# RDoc[http://www.ruby-doc.org/core/classes/RDoc.html]
+# documentation[http://www.caliban.org/ruby/ruby-aws/] as well as the
+# plain-text +README+ file that came with the archive.
+#
+# Additionally, there's a
+# {mailing-list}[http://www.caliban.org/mailman/listinfo/ruby-aws] available,
+# where you can discuss all Ruby/AWS-related subjects and issues.
+#
+# Please see the Amazon Web Services
+# documentation[http://developer.amazonwebservices.com/connect/kbcategory.jspa?categoryID=5]
+# for definitive information on the capabilities and inner workings of the AWS
+# API.
+#
+#
+# == Download
+#
+# Version 0.5.0
+# === {gzip'ed tar archive}[http://www.caliban.org/files/ruby/ruby-aws-0.5.0.tar.gz]
+# === {Ruby Gem}[http://www.caliban.org/files/ruby/ruby-aaws-0.5.0.gem]
+# === {Fedora 9 RPM}[http://www.caliban.org/files/redhat/RPMS/noarch/ruby-aws-0.5.0-1.fc9.noarch.rpm]
+# === {Fedora 9 doc RPM}[http://www.caliban.org/files/redhat/RPMS/noarch/ruby-aws-doc-0.5.0-1.fc9.noarch.rpm]
+# === {Fedora 9 source RPM}[http://www.caliban.org/files/redhat/SRPMS/ruby-aws-0.5.0-1.fc9.src.rpm]
+# 
+# 
+# ---
+# Author::   Ian Macdonald <mailto:ian@caliban.org>
+# Version::  0.5.0
+# Copyright:: (C) 2008-2009 Ian Macdonald
+# Licence::  GPL[http://www.gnu.org/copyleft/gpl.html]

data/lib/amazon.rb

@@ -0,0 +1,153 @@
+# $Id: amazon.rb,v 1.26 2009/01/19 16:45:11 ianmacd Exp $
+#
+
+module Amazon
+
+  # A top-level exception container class.
+  #
+  class AmazonError < StandardError; end
+
+  NAME = 'Ruby/Amazon'
+  @@config = {}
+
+  # Prints debugging messages and works like printf, except that it prints
+  # only when Ruby is run with the -d switch.
+  #
+  def Amazon.dprintf(format='', *args)
+    $stderr.printf( format + "\n", *args ) if $DEBUG
+  end
+
+
+  # Encode a string, such that it is suitable for HTTP transmission.
+  #
+  def Amazon.url_encode(string)
+
+    # Shamelessly plagiarised from Wakou Aoyama's cgi.rb.
+    #
+    string.gsub( /([^ a-zA-Z0-9_.-]+)/n ) do
+      '%' + $1.unpack( 'H2' * $1.size ).join( '%' ).upcase
+    end.tr( ' ', '+' )
+  end
+
+
+  # Convert a string from CamelCase to ruby_case.
+  #
+  def Amazon.uncamelise(str)
+    # Avoid modifying by reference.
+    #
+    str = str.dup
+
+    # Don't mess with string if all caps.
+    #
+    str.gsub!( /(.+?)(([A-Z][a-z]|[A-Z]+$))/, "\\1_\\2" ) if str =~ /[a-z]/
+
+    # Convert to lower case.
+    #
+    str.downcase
+  end
+
+
+  # A Class for dealing with configuration files, such as
+  # <tt>/etc/amazonrc</tt> and <tt>~/.amazonrc</tt>.
+  #
+  class Config < Hash
+
+    require 'stringio'
+
+    # Exception class for configuration file errors.
+    #
+    class ConfigError < AmazonError; end
+
+    # A configuration may be passed in as a string. Otherwise, the files
+    # <tt>/etc/amazonrc</tt> and <tt>~/.amazonrc</tt> are read if they exist
+    # and are readable.
+    #
+    def initialize(config_str=nil)
+      locale = nil
+
+      if config_str
+
+	# We have been passed a config file as a string.
+	#
+        config_files = [ config_str ]
+	config_class = StringIO
+
+      else
+
+	# Perform the usual search for the system and user config files.
+	#
+	config_files = [ File.join( '', 'etc', 'amazonrc' ) ]
+
+	# Figure out where home is. The locations after HOME are for Windows.
+	# [ruby-core:12347]
+	#
+	home = ENV['AMAZONRCDIR'] ||
+	       ENV['HOME'] || ENV['HOMEDRIVE'] + ENV['HOMEPATH'] ||
+	       ENV['USERPROFILE']
+	user_rcfile = ENV['AMAZONRCFILE'] || '.amazonrc'
+
+	if home
+	  config_files << File.expand_path( File.join( home, user_rcfile ) )
+	end
+
+	config_class = File
+      end
+
+      config_files.each do |cf|
+
+	if config_class == StringIO
+	  readable = true
+	else
+	  # We must determine whether the file is readable.
+	  #
+	  readable = File.exists?( cf ) && File.readable?( cf )
+	end
+
+	if readable
+
+	  Amazon.dprintf( 'Opening %s ...', cf ) if config_class == File
+
+	  config_class.open( cf ) { |f| lines = f.readlines }.each do |line|
+	    line.chomp!
+
+	    # Skip comments and blank lines.
+	    #
+	    next if line =~ /^(#|$)/
+
+	    Amazon.dprintf( 'Read: %s', line )
+
+	    # Determine whether we're entering the subsection of a new locale.
+	    #
+	    if match = line.match( /^\[(\w+)\]$/ )
+	      locale = match[1]
+	      Amazon.dprintf( "Config locale is now '%s'.", locale )
+	      next
+	    end
+
+	    # Store these, because we'll probably find a use for these later.
+	    #
+	    begin
+      	      match = line.match( /^\s*(\S+)\s*=\s*(['"]?)([^'"]+)(['"]?)/ )
+	      key, begin_quote, val, end_quote = match[1, 4]
+	      raise ConfigError if begin_quote != end_quote
+
+	    rescue NoMethodError, ConfigError
+	      raise ConfigError, "bad config line: #{line}"
+	    end
+
+	    if locale && locale != 'global'
+	      self[locale] ||= {}
+	      self[locale][key] = val
+	    else
+	      self[key] = val
+	    end
+
+	  end
+	end
+
+      end
+
+    end
+  end
+
+end

data/lib/amazon/aws.rb

@@ -0,0 +1,1339 @@
+# $Id: aws.rb,v 1.84 2009/02/19 16:01:11 ianmacd Exp $
+#
+#:include: ../../README.rdoc
+
+module Amazon
+
+  module AWS
+
+    require 'uri'
+    require 'amazon'
+    require 'amazon/aws/cache'
+    require 'rexml/document'
+
+    NAME = '%s/%s' % [ Amazon::NAME, 'AWS' ]
+    VERSION = '0.4.4'
+    USER_AGENT = '%s %s' % [ NAME, VERSION ]
+
+    # Default Associate tags to use per locale.
+    #
+    DEF_ASSOC  = {
+      'ca' => 'caliban-20',
+      'de' => 'calibanorg0a-21',
+      'fr' => 'caliban08-21',
+      'jp' => 'calibanorg-20',
+      'uk' => 'caliban-21',
+      'us' => 'calibanorg-20'
+    }
+
+    # Service name and API version for AWS. The version of the API used can be
+    # changed via the user configuration file.
+    #
+    SERVICE = { 'Service' => 'AWSECommerceService',
+		'Version' => '2009-01-06'
+    }
+
+    # Maximum number of 301 and 302 HTTP responses to follow, should Amazon
+    # later decide to change the location of the service.
+    #
+    MAX_REDIRECTS = 3
+
+    # Maximum number of results pages that can be retrieved for a given
+    # search operation, using whichever pagination parameter is relevant to
+    # that type of operation.
+    #
+    PAGINATION = {
+      'ItemSearch'	      => { 'parameter' => 'ItemPage',
+						  'max_page' => 400 },
+      'ItemLookup'	      => { 'paraneter' => 'OfferPage',
+						  'max_page' => 100 },
+      'ListLookup'	      => { 'parameter' => 'ProductPage',
+						  'max_page' =>  30 },
+      'ListSearch'	      => { 'parameter' => 'ListPage',
+						  'max_page' =>  20 },
+      'CustomerContentLookup' => { 'parameter' => 'ReviewPage',
+						  'max_page' =>  10 },
+      'CustomerContentSearch' => { 'parameter' => 'CustomerPage',
+						  'max_page' =>  20 },
+      'VehiclePartLookup'     => { 'parameter' => 'FitmentPage',
+						  'max_page' =>  10 }
+    }
+    # N.B. ItemLookup can also use the following two pagination parameters
+    #
+    #		      max. page
+    #		      ---------
+    # VariationPage   150
+    # ReviewPage       20
+  
+
+    # Exception class for HTTP errors.
+    #
+    class HTTPError < AmazonError; end
+
+
+    # Exception class for faulty batch operations.
+    #
+    class BatchError < AmazonError; end
+
+
+    class Endpoint
+
+      attr_reader :host, :path
+
+      def initialize(endpoint)
+	uri = URI.parse( endpoint )
+	@host = uri.host
+	@path = uri.path
+      end
+    end
+
+    ENDPOINT = {
+      'ca' => Endpoint.new( 'http://ecs.amazonaws.ca/onca/xml' ),
+      'de' => Endpoint.new( 'http://ecs.amazonaws.de/onca/xml' ),
+      'fr' => Endpoint.new( 'http://ecs.amazonaws.fr/onca/xml' ),
+      'jp' => Endpoint.new( 'http://ecs.amazonaws.jp/onca/xml' ),
+      'uk' => Endpoint.new( 'http://ecs.amazonaws.co.uk/onca/xml' ),
+      'us' => Endpoint.new( 'http://ecs.amazonaws.com/onca/xml' )
+    }
+
+    # Fetch a page, either from the cache or by HTTP. This is used internally.
+    #
+    def AWS.get_page(request, query)  # :nodoc:
+
+      url = ENDPOINT[request.locale].path + query
+      cache_url = ENDPOINT[request.locale].host + url
+
+      # Check for cached page and return that if it's there.
+      #
+      if request.cache && request.cache.cached?( cache_url )
+	body = request.cache.fetch( cache_url )
+	return body if body
+      end
+
+      # Get the existing connection. If there isn't one, force a new one.
+      #
+      conn = request.conn || request.reconnect.conn
+      user_agent = request.user_agent
+
+      Amazon.dprintf( 'Fetching http://%s%s ...', conn.address, url )
+
+      begin
+	response = conn.get( url, { 'user-agent' => user_agent } )
+
+      # If we've pulled and processed a lot of pages from the cache (or
+      # just not passed by here recently), the HTTP connection to the server
+      # will probably have timed out.
+      #
+      rescue Errno::ECONNRESET
+	conn = request.reconnect.conn
+	retry
+      end
+
+      redirects = 0
+      while response.key? 'location'
+	if ( redirects += 1 ) > MAX_REDIRECTS
+	  raise HTTPError, "More than #{MAX_REDIRECTS} redirections"
+	end
+
+	old_url = url
+	url = URI.parse( response['location'] )
+	url.scheme = old_url.scheme unless url.scheme
+	url.host = old_url.host unless url.host
+	Amazon.dprintf( 'Following HTTP %s to %s ...', response.code, url )
+	response = Net::HTTP::start( url.host ).
+		     get( url.path, { 'user-agent' => user_agent } )
+      end
+
+      if response.code != '200'
+	raise HTTPError, "HTTP response code #{response.code}"
+      end
+
+      # Cache the page if we're using a cache.
+      #
+      if request.cache
+	request.cache.store( cache_url, response.body )
+      end
+
+      response.body
+    end
+
+
+    def AWS.assemble_query(items)  # :nodoc:
+      query = ''
+
+      # We must sort the items into an array to get reproducible ordering
+      # of the query parameters. Otherwise, URL caching would not work. We
+      # must also convert the keys to strings, in case Symbols have been used
+      # as the keys.
+      #
+      items.sort { |a,b| a.to_s <=> b.to_s }.each do |k, v|
+	query << '&%s=%s' % [ k, Amazon.url_encode( v.to_s ) ]
+      end
+
+      # Replace initial ampersand with question-mark.
+      #
+      query[0] = '?'
+
+      query
+    end
+
+
+    # Everything returned by AWS is an AWSObject.
+    #
+    class AWSObject
+
+      include REXML
+
+      # This method can be used to load AWSObject data previously serialised
+      # by Marshal.dump.
+      #
+      # Example:
+      #
+      #  File.open( 'aws.dat' ) { |f| Amazon::AWS::AWSObject.load( f ) }
+      #
+      # Marshal.load cannot be used directly, because subclasses of AWSObject
+      # are dynamically defined as needed when AWS XML responses are parsed.
+      #
+      # Later attempts to load objects instantiated from these classes cause a
+      # problem for Marshal, because it knows nothing of classes that were
+      # dynamically defined by a separate process.
+      #
+      def AWSObject.load(io)
+	begin
+	  Marshal.load( io )
+	rescue ArgumentError => ex
+	  m = ex.to_s.match( /Amazon::AWS::AWSObject::([^ ]+)/ )
+	  const_set( m[1], Class.new( AWSObject ) )
+
+	  io.rewind
+	  retry
+	end
+      end
+ 
+
+      # This method can be used to load AWSObject data previously serialised
+      # by YAML.dump.
+      #
+      # Example:
+      #
+      #  File.open( 'aws.yaml' ) { |f| Amazon::AWS::AWSObject.yaml_load( f ) }
+      #
+      # The standard YAML.load cannot be used directly, because subclasses of
+      # AWSObject are dynamically defined as needed when AWS XML responses are
+      # parsed.
+      #
+      # Later attempts to load objects instantiated from these classes cause a
+      # problem for YAML, because it knows nothing of classes that were
+      # dynamically defined by a separate process.
+      #
+      def AWSObject.yaml_load(io)
+        io.each do |line|
+    
+	  # File data is external, so it's deemed unsafe when $SAFE > 0, which
+	  # is the case with mod_ruby, for example, where $SAFE == 1.
+	  #
+	  # YAML data isn't eval'ed or anything dangerous like that, so we
+	  # consider it safe to untaint it. If we don't, mod_ruby will complain
+	  # when Module#const_defined? is invoked a few lines down from here.
+	  #
+	  line.untaint
+	  
+	  m = line.match( /Amazon::AWS::AWSObject::([^ ]+)/ )
+	  if m
+	    cl_name = [ m[1] ]
+	  
+	    # Module#const_defined? takes 2 parameters in Ruby 1.9.
+	    #
+	    cl_name << false if Object.method( :const_defined? ).arity == -1
+	  
+	    unless AWSObject.const_defined?( *cl_name )
+	      AWSObject.const_set( m[1], Class.new( AWSObject ) )
+	    end
+	  
+	  end
+	end
+    
+	io.rewind
+	YAML.load( io )
+      end
+
+
+      def initialize(op=nil)
+	# The name of this instance variable must never clash with the
+	# uncamelised name of an Amazon tag.
+	#
+	# This is used to store the REXML::Text value of an element, which
+	# exists only when the element contains no children.
+	#
+	@__val__ = nil
+	@__op__ = op if op
+      end
+
+
+      def method_missing(method, *params)
+	iv = '@' + method.id2name
+
+	if instance_variables.include?( iv )
+	  instance_variable_get( iv )
+	elsif instance_variables.include?( iv.to_sym )
+
+	  # Ruby 1.9 Object#instance_variables method returns Array of Symbol,
+	  # not String.
+	  #
+	  instance_variable_get( iv.to_sym )
+	else
+	  nil
+	end
+      end
+      private :method_missing
+ 
+
+      def remove_val
+	remove_instance_variable( :@__val__ )
+      end
+      private :remove_val
+
+
+      # Iterator method for cycling through an object's properties and values.
+      #
+      def each  # :yields: property, value
+	self.properties.each do |iv|
+	  yield iv, instance_variable_get( "@#{iv}" )
+	end
+      end
+
+      alias :each_property :each
+
+
+      def inspect  # :nodoc:
+	remove_val if instance_variable_defined?( :@__val__ ) && @__val__.nil?
+	str = super
+	str.sub( /@__val__=/, 'value=' ) if str
+      end
+
+
+      def to_s	# :nodoc:
+	if instance_variable_defined?( :@__val__ )
+	  return @__val__ if @__val__.is_a?( String )
+	  remove_val
+	end
+
+	string = ''
+
+	# Assemble the object's details.
+	#
+	each { |iv, value| string << "%s = %s\n" % [ iv, value ] }
+
+	string
+      end
+
+      alias :to_str :to_s
+
+
+      def to_i	# :nodoc:
+	@__val__.to_i
+      end
+
+
+      def ==(other)  # :nodoc:
+        @__val__.to_s == other
+      end
+
+
+      def =~(other)  # :nodoc:
+	@__val__.to_s =~ other
+      end
+
+
+      # This alias makes the ability to determine an AWSObject's properties a
+      # little more intuitive. It's pretty much just an alias for the
+      # inherited <em>Object#instance_variables</em> method, with a little
+      # tidying.
+      #
+      def properties
+	# Make sure we remove the leading @.
+	#
+	iv = instance_variables.collect { |v| v = v[1..-1] }
+	iv.delete( '__val__' )
+	iv
+      end
+
+
+      # Provide a shortcut down to the data likely to be of most interest.
+      # This method is experimental and may be removed.
+      #
+      def kernel  # :nodoc: 
+	# E.g. Amazon::AWS::SellerListingLookup -> seller_listing_lookup
+	#
+	stub = Amazon.uncamelise( @__op__.class.to_s.sub( /^.+::/, '' ) )
+
+	# E.g. seller_listing_response
+	#
+	level1 = stub + '_response'
+
+	# E.g. seller_listing
+	#
+	level3 = stub.sub( /_[^_]+$/, '' )
+
+	# E.g. seller_listings
+	#
+	level2 = level3 + 's'
+
+	# E.g.
+	# seller_listing_search_response[0].seller_listings[0].seller_listing
+	#
+	self.instance_variable_get( "@#{level1}" )[0].
+	     instance_variable_get( "@#{level2}" )[0].
+	     instance_variable_get( "@#{level3}" )
+      end
+
+
+      # Convert an AWSObject to a Hash.
+      #
+      def to_h
+	hash = {}
+
+	each do |iv, value|
+	  if value.is_a? AWSObject
+	    hash[iv] = value.to_h
+	  elsif value.is_a?( AWSArray ) && value.size == 1
+	    hash[iv] = value[0]
+	  else
+	    hash[iv] = value
+	  end
+	end
+
+	hash
+      end
+
+
+      # Fake the appearance of an AWSObject as a hash. _key_ should be any
+      # attribute of the object and can be a String, Symbol or anything else
+      # that can be converted to a String with to_s.
+      #
+      def [](key)
+	instance_variable_get( "@#{key}" )
+      end
+
+
+      # Recursively walk through an XML tree, starting from _node_. This is
+      # called internally and is not intended for user code.
+      #
+      def walk(node)  # :nodoc:
+    
+	if node.instance_of?( REXML::Document )
+	  walk( node.root )
+    
+	elsif node.instance_of?( REXML::Element )
+	  name = Amazon.uncamelise( node.name )
+    
+	  cl_name = [ node.name ]
+
+	  # Module#const_defined? takes 2 parameters in Ruby 1.9.
+	  #
+	  cl_name << false if Object.method( :const_defined? ).arity == -1
+
+	  # Create a class for the new element type unless it already exists.
+	  #
+	  unless AWS::AWSObject.const_defined?( *cl_name )
+	    cl = AWS::AWSObject.const_set( node.name, Class.new( AWSObject ) )
+
+	    # Give it an accessor for @attrib.
+	    #
+	    cl.send( :attr_accessor, :attrib )
+	  end
+    
+	  # Instantiate an object in the newly created class.
+	  #
+	  obj = AWS::AWSObject.const_get( node.name ).new
+
+	  sym_name = "@#{name}".to_sym
+    
+	  if instance_variable_defined?( sym_name)
+    	    instance_variable_set( sym_name,
+    	      instance_variable_get( sym_name ) << obj )
+	  else
+	    instance_variable_set( sym_name, AWSArray.new( [ obj ] ) )
+	  end
+    
+	  if node.has_attributes?
+	    obj.attrib = {}
+	    node.attributes.each_pair do |a_name, a_value|
+	      obj.attrib[a_name.downcase] =
+		a_value.to_s.sub( /^#{a_name}=/, '' )
+	    end
+	  end
+
+	  node.children.each { |child| obj.walk( child ) }
+    
+	else # REXML::Text
+	  @__val__ = node.to_s
+	end
+      end
+
+
+      # For objects of class AWSObject::.*Image, fetch the image in question,
+      # optionally overlaying a discount icon for the percentage amount of
+      # _discount_ to the image.
+      #
+      def get(discount=nil)
+	if self.class.to_s =~ /Image$/ && @url
+          url = URI.parse( @url[0] )
+          url.path.sub!( /(\.\d\d\._)/, "\\1PE#{discount}" ) if discount
+
+	  # FIXME: All HTTP in Ruby/AWS should go through the same method.
+	  #
+          Net::HTTP.start( url.host, url.port ) do |http|
+	    http.get( url.path )
+	  end.body
+
+	else
+	  nil
+	end
+      end
+
+    end
+
+
+    # Everything we get back from AWS is transformed into an array. Many of
+    # these, however, have only one element, because the corresponding XML
+    # consists of a parent element containing only a single child element.
+    #
+    # This class consists solely to allow single element arrays to pass a
+    # method call down to their one element, thus obviating the need for lots
+    # of references to <tt>foo[0]</tt> in user code.
+    #
+    # For example, the following:
+    #
+    #  items = resp.item_search_response[0].items[0].item
+    #
+    # can be reduced to:
+    #
+    #  items = resp.item_search_response.items.item
+    #
+    class AWSArray < Array
+
+      def method_missing(method, *params)
+	self.size == 1 ? self[0].send( method, *params ) : super
+      end
+      private :method_missing
+
+
+      # In the case of a single-element array, return the first element,
+      # converted to a String.
+      #
+      def to_s  # :nodoc:
+	self.size == 1 ? self[0].to_s : super
+      end
+
+      alias :to_str :to_s
+
+
+      # In the case of a single-element array, return the first element,
+      # converted to an Integer.
+      #
+      def to_i  # :nodoc:
+	self.size == 1 ? self[0].to_i : super
+      end
+
+
+      # In the case of a single-element array, compare the first element with
+      # _other_.
+      #
+      def ==(other)  # :nodoc:
+	self.size == 1 ? self[0].to_s == other : super
+      end
+
+
+      # In the case of a single-element array, perform a pattern match on the
+      # first element against _other_.
+      #
+      def =~(other)  # :nodoc:
+	self.size == 1 ? self[0].to_s =~ other : super
+      end
+
+    end
+
+ 
+    # This is the base class of all AWS operations.
+    #
+    class Operation
+
+      # These are the types of AWS operation currently implemented by Ruby/AWS.
+      #
+      OPERATIONS = %w[
+	BrowseNodeLookup      CustomerContentLookup   CustomerContentSearch
+	Help		      ItemLookup	      ItemSearch
+	ListLookup	      ListSearch	      SellerListingLookup
+	SellerListingSearch   SellerLookup	      SimilarityLookup
+	TagLookup	      TransactionLookup	      VehiclePartLookup
+	VehiclePartSearch     VehicleSearch
+
+	CartAdd		      CartClear		      CartCreate
+	CartGet		      CartModify
+      ]
+
+      attr_reader :kind
+      attr_accessor :params
+
+      def initialize(parameters)
+
+	op_kind = self.class.to_s.sub( /^.*::/, '' )
+	unless OPERATIONS.include?( op_kind ) || op_kind == 'MultipleOperation'
+	  raise "Bad operation: #{op_kind}"
+	end
+	#raise 'Too many parameters' if parameters.size > 10
+
+	@kind = op_kind
+	@params = { 'Operation' => op_kind }.merge( parameters )
+      end
+
+
+      # Group together operations of the same class in a batch request.
+      # _operations_ should be either an operation of the same class as *self*
+      # or an array of such operations.
+      #
+      # If you need to batch operations from different classes, use a
+      # MultipleOperation instead.
+      #
+      # Example:
+      #
+      #  is = ItemSearch.new( 'Books', { 'Title' => 'ruby programming' } )
+      #  is2 = ItemSearch.new( 'Music', { 'Artist' => 'stranglers' } )
+      #	 is.batch( is2 )
+      #
+      #	Please see MultipleOperation.new for details of a couple of
+      #	restrictions that also apply to batched operations.
+      #
+      def batch(*operations)
+
+	# Remove the Operation parameter to avoid batch syntax being applied.
+	# We'll readd it at the end.
+	#
+	op_type = @params.delete( 'Operation' )
+
+	operations.flatten.each do |op|
+
+	  unless self.class == op.class
+	    raise BatchError, "You can't batch different classes of operation. Use class MultipleOperation."
+	  end
+
+	  # Remove the Operation parameter.
+	  #
+	  op.params.delete( 'Operation' )
+
+	  # Apply batch syntax.
+	  #
+	  @params = batch_parameters( @params, op.params )
+	end
+
+	# Reinstate the Operation parameter.
+	#
+	@params.merge!( { 'Operation' => op_type } )
+      end
+
+
+      # Convert parameters to batch format, e.g. ItemSearch.1.Title.
+      #
+      def batch_parameters(params, *b_params)  # :nodoc:
+
+	@index ||= 1
+
+	unless b_params.empty?
+	  op_str = self.class.to_s.sub( /^.+::/, '' )
+
+	  # Fudge the operation string if we're dealing with a shopping cart.
+	  #
+	  op_str = 'Item' if op_str =~ /^Cart/
+
+	  all_parameters = [ params ].concat( b_params )
+	  params = {}
+
+	  all_parameters.each_with_index do |hash, index|
+
+	    # Don't batch an already batched hash.
+	    #
+	    if ! hash.empty? && hash.to_a[0][0] =~ /^.+\..+\..+$/
+	      params = hash
+	      next
+	    end
+
+	    hash.each do |tag, val|
+	      shared_param = '%s.%d.%s' % [ op_str, @index + index, tag ]
+	      params[shared_param] = val
+	    end
+	  end
+
+	  @index += b_params.size
+
+	end
+
+	params
+      end
+
+    end
+
+
+    # This class can be used to merge multiple operations into a single
+    # operation for greater efficiency.
+    #
+    class MultipleOperation < Operation
+
+      # This will allow you to take two Operation objects and combine them to
+      # form a single object, which can then be used to perform a single
+      # request to AWS. This allows for greater efficiency, reducing the
+      # number of requests sent to AWS.
+      #
+      # AWS currently imposes a limit of two combined operations in a multiple
+      # operation.
+      #
+      # <em>operation1</em> and <em>operation2</em> are both objects from a
+      # subclass of Operation, such as ItemSearch, ItemLookup, etc.
+      #
+      # There are currently a few restrictions in the Ruby/AWS implementation
+      # of multiple operations:
+      #
+      # - ResponseGroup objects used when calling AWS::Search::Request#search
+      #   apply to both operations. You cannot use a different ResponseGroup
+      #   with each operation.
+      #
+      # - One or both operations may have multiple results pages available,
+      #   but only the first page is returned. If you need the subsequent
+      #   pages, perform the operations separately, not as part of a
+      #   MultipleOperation.
+      #
+      # Example:
+      #
+      #  is = ItemSearch.new( 'Books', { 'Title' => 'Ruby' } )
+      #  il = ItemLookup.new( 'ASIN', { 'ItemId' => 'B0013DZAYO',
+      #					'MerchantId' => 'Amazon' } )
+      #	 mo = MultipleOperation.new( is, il )
+      #
+      # As you can see, the operations that are combined as a
+      # MultipleOperation do not have to belong to the same class. In the
+      # above example, we compose a multiple operation consisting of an
+      # ItemSearch and an ItemLookup.
+      #
+      # If you want to batch operations belonging to the same class,
+      # Operation#batch provides an alternative.
+      # 
+      def initialize(operation1, operation2)
+
+	# Safeguard against changing original Operation objects in place. This
+	# is to protect me, not for user code.
+	#
+	operation1.freeze
+	operation2.freeze
+
+	op_kind = '%s,%s' % [ operation1.kind, operation2.kind ]
+
+	# Duplicate Operation objects and remove their Operation parameter.
+	# 
+	op1 = operation1.dup
+	op1.params = op1.params.dup
+	op1.params.delete( 'Operation' )
+
+	op2 = operation2.dup
+	op2.params = op2.params.dup
+	op2.params.delete( 'Operation' )
+
+	if op1.class == op2.class
+
+	  # If both operations are of the same type, we combine the parameters
+	  # of both.
+	  #
+	  b_params = op1.batch_parameters( op1.params, op2.params )
+	else
+
+	  # We have to convert the parameters to batch format.
+	  #
+	  bp1 = op1.batch_parameters( op1.params, {} )
+	  bp2 = op2.batch_parameters( op2.params, {} )
+	  b_params = bp1.merge( bp2 )
+	end
+
+	params = { 'Operation' => op_kind }.merge( b_params )
+	super( params )
+
+      end
+      
+    end
+
+
+    # This class of operation aids in finding out about AWS operations and
+    # response groups.
+    #
+    class Help < Operation
+
+      # Return information on AWS operations and response groups.
+      #
+      # For operations, required and optional parameters are returned, along
+      # with information about which response groups the operation can use.
+      #
+      # For response groups, The list of operations that can use that group is
+      # returned, as well as the list of response tags returned by the group.
+      #
+      # _help_type_ is the type of object for which help is being sought, such
+      # as *Operation* or *ResponseGroup*. _about_ is the name of the
+      # operation or response group you need help with, and _parameters_ is an
+      # optional hash of parameters that further refine the request for help.
+      #
+      def initialize(help_type, about, parameters={})
+	super( { 'HelpType' => help_type,
+		 'About'    => about
+	       }.merge( parameters ) )
+      end
+
+    end
+
+
+    # This is the class for the most common type of AWS look-up, an
+    # ItemSearch. This allows you to search for items that match a set of
+    # broad criteria. It returns items for sale by Amazon merchants and most
+    # types of seller.
+    #
+    class ItemSearch < Operation
+
+      # Not all search indices work in all locales. It is the user's
+      # responsibility to ensure that a given index is valid within a given
+      # locale.
+      #
+      # According to the AWS documentation:
+      #
+      # - *All* searches through all indices (but currently exists only in the
+      #   *US* locale).
+      # - *Blended* combines Apparel, Automotive, Books, DVD, Electronics,
+      #   GourmetFood, Kitchen, Music, PCHardware, PetSupplies, Software,
+      #   SoftwareVideoGames, SportingGoods, Tools, Toys, VHS and VideoGames.
+      # - *Merchants* combines all search indices for a merchant given with
+      #   MerchantId.
+      # - *Music* combines the Classical, DigitalMusic, and MusicTracks
+      #   indices.
+      # - *Video* combines the DVD and VHS search indices.
+      #
+      # Note that {page
+      # 53}[http://docs.amazonwebservices.com/AWSECommerceService/2009-01-06/DG/SearchIndices.html]
+      # of the PDF of the AWS Developer Guide (revision 2009-01-06) contains
+      # an outdated description of *Blended* with too few subindices. {Page
+      # 95}[http://docs.amazonwebservices.com/AWSECommerceService/2009-01-06/DG/CommonItemSearchParameters.html]
+      # of the PDF contains the correct list.
+      #
+      SEARCH_INDICES = %w[
+	All
+	Apparel
+	Automotive
+	Baby
+	Beauty
+	Blended
+	Books
+	Classical
+	DigitalMusic
+	DVD
+	Electronics
+	ForeignBooks
+	GourmetFood
+	Grocery
+	HealthPersonalCare
+	Hobbies
+	HomeGarden
+	Industrial
+	Jewelry
+	KindleStore
+	Kitchen
+	Magazines
+	Merchants
+	Miscellaneous
+	MP3Downloads
+	Music
+	MusicalInstruments
+	MusicTracks
+	OfficeProducts
+	OutdoorLiving
+	PCHardware
+	PetSupplies
+	Photo
+	SilverMerchant
+	Software
+	SoftwareVideoGames
+	SportingGoods
+	Tools
+	Toys
+	VHS
+	Video
+	VideoGames
+	Watches
+	Wireless
+	WirelessAccessories
+      ]
+
+
+      # Search AWS for items. _search_index_ must be one of _SEARCH_INDICES_
+      # and _parameters_ is an optional hash of parameters that further refine
+      # the scope of the search.
+      #
+      # Example:
+      #
+      #  is = ItemSearch.new( 'Books', { 'Title' => 'ruby programming' } )
+      #
+      # In the above example, we search for books with <b>Ruby Programming</b>
+      # in the title.
+      #
+      def initialize(search_index, parameters)
+	unless SEARCH_INDICES.include? search_index.to_s
+	  raise "Invalid search index: #{search_index}"
+	end
+
+	super( { 'SearchIndex' => search_index }.merge( parameters ) )
+      end
+
+    end
+
+
+    # This class of look-up deals with searching for *specific* items by some
+    # uniquely identifying attribute, such as the ASIN (*A*mazon *S*tandard
+    # *I*tem *N*umber).
+    #
+    class ItemLookup < Operation
+
+      # Look up a specific item in the AWS catalogue. _id_type_ is the type of
+      # identifier and  _parameters_ is a hash that identifies the item to be
+      # located and narrows the scope of the search.
+      #
+      # Example:
+      #
+      #  il = ItemLookup.new( 'ASIN', { 'ItemId' => 'B000AE4QEC'
+      #					'MerchantId' => 'Amazon' } )
+      #
+      # In the above example, we search for an item, based on its ASIN. The
+      # use of _MerchantId_ restricts the offers returned to those for sale
+      # by Amazon (as opposed to third-party sellers).
+      #
+      def initialize(id_type, parameters)
+	super( { 'IdType' => id_type }.merge( parameters ) )
+      end
+
+    end
+
+
+    # Search for items for sale by a particular seller.
+    #
+    class SellerListingSearch < Operation
+
+      # Search for items for sale by a particular seller. _seller_id_ is the
+      # Amazon seller ID and _parameters_ is an optional hash of parameters
+      # that further refine the scope of the search.
+      #
+      # Example:
+      #
+      #  sls = SellerListingSearch.new( 'A33J388YD2MWJZ',
+      #					{ 'Keywords' => 'Killing Joke' } )
+      #
+      # In the above example, we search seller <b>A33J388YD2MWJ</b>'s listings
+      # for items with the keywords <b>Killing Joke</b>.
+      #
+      def initialize(seller_id, parameters)
+	super( { 'SellerId' => seller_id }.merge( parameters ) )
+      end
+
+    end
+
+
+    # Return specified items in a seller's store.
+    #
+    class SellerListingLookup < ItemLookup
+
+      # Look up a specific item for sale by a specific seller. _id_type_ is
+      # the type of identifier and _parameters_ is a hash that identifies the
+      # item to be located and narrows the scope of the search.
+      #
+      # Example:
+      #
+      #  sll = SellerListingLookup.new( 'AP8U6Y3PYQ9VO', 'ASIN',
+      #					{ 'Id' => 'B0009RRRC8' } )
+      #
+      # In the above example, we search seller <b>AP8U6Y3PYQ9VO</b>'s listings
+      # to find items for sale with the ASIN <b>B0009RRRC8</b>.
+      #
+      def initialize(seller_id, id_type, parameters)
+	super( id_type, { 'SellerId' => seller_id }.merge( parameters ) )
+      end
+
+    end
+
+
+    # Return information about a specific seller.
+    #
+    class SellerLookup < Operation
+
+      # Search for the details of a specific seller. _seller_id_ is the Amazon
+      # ID of the seller in question and _parameters_ is an optional hash of
+      # parameters that further refine the scope of the search.
+      #
+      # Example:
+      #
+      #  sl = SellerLookup.new( 'A3QFR0K2KCB7EG' )
+      #
+      # In the above example, we look up the details of the seller with ID
+      # <b>A3QFR0K2KCB7EG</b>.
+      #
+      def initialize(seller_id, parameters={})
+	super( { 'SellerId' => seller_id }.merge( parameters ) )
+      end
+
+    end
+
+
+    # Obtain the information an Amazon customer has made public about
+    # themselves.
+    #
+    class CustomerContentLookup < Operation
+
+      # Search for public customer data. _customer_id_ is the unique ID
+      # identifying the customer on Amazon and _parameters_ is an optional
+      # hash of parameters that further refine the scope of the search.
+      #
+      # Example:
+      #
+      #  ccl = CustomerContentLookup.new( 'AJDWXANG1SYZP' )
+      #
+      # In the above example, we look up public data about the customer with
+      # the ID <b>AJDWXANG1SYZP</b>.
+      #
+      def initialize(customer_id, parameters={})
+	super( { 'CustomerId' => customer_id }.merge( parameters ) )
+      end
+
+    end
+
+
+    # Retrieve basic Amazon customer data.
+    #
+    class CustomerContentSearch < Operation
+
+      # Retrieve customer information, using an e-mail address or name.
+      #
+      # If _customer_id_ contains an '@' sign, it is assumed to be an e-mail
+      # address. Otherwise, it is assumed to be the customer's name.
+      #
+      # Example:
+      #
+      #  ccs = CustomerContentSearch.new( 'ian@caliban.org' )
+      #
+      # In the above example, we look up customer information about
+      # <b>ian@caliban.org</b>. The *CustomerInfo* response group will return,
+      # amongst other things, a _customer_id_ property, which can then be
+      # plugged into CustomerContentLookup to retrieve more detailed customer
+      # information.
+      #
+      def initialize(customer_id)
+	id = customer_id =~ /@/ ? 'Email' : 'Name'
+	super( { id => customer_id } )
+      end
+
+    end
+
+
+    # Find wishlists, registry lists, etc. created by users and placed on
+    # Amazon. These are items that customers would like to receive as
+    # presnets.
+    #
+    class ListSearch < Operation
+
+      # Search for Amazon lists. _list_type_ is the type of list to search for
+      # and _parameters_ is an optional hash of parameters that narrow the
+      # scope of the search.
+      #
+      # Example:
+      #
+      #  ls = ListSearch.new( 'WishList', { 'Name' => 'Peter Duff' }
+      #
+      # In the above example, we retrieve the wishlist for the Amazon user,
+      # <b>Peter Duff</b>.
+      #
+      def initialize(list_type, parameters)
+	super( { 'ListType' => list_type }.merge( parameters ) )
+      end
+
+    end
+
+
+    # Find the details of specific wishlists, registries, etc.
+    #
+    class ListLookup < Operation
+
+      # Look up and return details about a specific list. _list_id_ is the
+      # Amazon list ID, _list_type_ is the type of list and _parameters_ is an
+      # optional hash of parameters that narrow the scope of the search.
+      #
+      # Example:
+      #
+      #  ll = ListLookup.new( '3P722DU4KUPCP', 'Listmania' )
+      #
+      # In the above example, a *Listmania* list with the ID
+      # <b>3P722DU4KUPCP</b> is retrieved from AWS.
+      #
+      def initialize(list_id, list_type, parameters={})
+        super( { 'ListId'   => list_id,
+	         'ListType' => list_type
+	       }.merge( parameters ) )
+      end
+
+    end
+
+
+    # Amazon use browse nodes as a means of organising the millions of items
+    # in their inventory. An example might be *Carving Knives*. Looking up a
+    # browse node enables you to determine that group's ancestors and
+    # descendants.
+    #
+    class BrowseNodeLookup < Operation
+
+      # Look up and return the details of an Amazon browse node. _node_ is the
+      # browse node to look up and _parameters_ is an optional hash of
+      # parameters that further refine the scope of the search. _parameters_
+      # is currently unused.
+      #
+      # Example:
+      #
+      #  bnl = BrowseNodeLookup.new( '11232', {} )
+      #
+      # In the above example, we look up the browse node with the ID
+      # <b>11232</b>. This is the <b>Social Sciences</b> browse node.
+      #
+      def initialize(node, parameters={})
+	super( { 'BrowseNodeId' => node }.merge( parameters ) )
+      end
+
+    end
+
+
+    # Similarity look-up is for items similar to others.
+    #
+    class SimilarityLookup < Operation
+
+      # Look up items similar to _asin_, which can be a single item or an
+      # array. _parameters_ is an optional hash of parameters that further
+      # refine the scope of the search.
+      #
+      # Example:
+      #
+      #  sl = SimilarityLookup.new( 'B000051WBE' )
+      #
+      # In the above example, we search for items similar to the one with ASIN
+      # <b>B000051WBE</b>.
+      #
+      def initialize(asin, parameters={})
+	super( { 'ItemId' => asin.to_a.join( ',' ) }.merge( parameters ) )
+      end
+
+    end
+
+
+    # Search for entities based on user-defined tags. A tag is a descriptive
+    # word that a customer uses to label entities on Amazon's Web site.
+    # Entities can be items for sale, Listmania lists, guides, etc.
+    #
+    class TagLookup < Operation
+
+      # Look up entities based on user-defined tags. _tag_name_ is the tag to
+      # search on and _parameters_ is an optional hash of parameters that
+      # further refine the scope of the search.
+      #
+      # Example:
+      #
+      #  tl = TagLookup.new( 'Awful' )
+      #
+      # In the example above, we search for entities tagged by users with the
+      # word *Awful*.
+      #
+      def initialize(tag_name, parameters={})
+	super( { 'TagName' => tag_name }.merge( parameters ) )
+      end
+
+    end
+
+
+    # Search for information on previously completed purchases.
+    #
+    class TransactionLookup < Operation
+
+      # Return information on an already completed purchase. _transaction_id_
+      # is actually the order number that is created when you place an order
+      # on Amazon.
+      #
+      # Example:
+      #
+      #  tl = TransactionLookup.new( '103-5663398-5028241' )
+      #
+      # In the above example, we retrieve the details of order number
+      # <b>103-5663398-5028241</b>.
+      #
+      def initialize(transaction_id)
+	super( { 'TransactionId' => transaction_id } )
+      end
+
+    end
+
+
+    # Look up individual vehicle parts.
+    #
+    class VehiclePartLookup < Operation
+
+      # Look up a particular vehicle part. _item_id_ is the ASIN of the part
+      # in question and _parameters_ is an optional hash of parameters that
+      # further refine the scope of the search.
+      #
+      # Although the _item_id_ alone is enough to locate the part, providing
+      # _parameters_ can be useful in determining whether the part looked up
+      # is a fit for a particular vehicle type, as with the *VehiclePartFit*
+      # response group.
+      # 
+      # Example:
+      #
+      #  vpl = VehiclePartLookup.new( 'B000C1ZLI8',
+      #				      { 'Year' => 2008,
+      #				        'MakeId' => 73,
+      #				        'ModelId' => 6039,
+      #				        'TrimId' => 20 } )
+      #
+      #	Here, we search for a <b>2008</b> model *Audi* <b>R8</b> with *Base*
+      #	trim. The required Ids can be found using VehiclePartSearch.
+      #
+      def initialize(item_id, parameters={})
+	super( { 'ItemId' => item_id }.merge( parameters ) )
+      end
+
+    end
+
+
+    # Search for parts for a given vehicle.
+    #
+    class VehiclePartSearch < Operation
+
+      # Find parts for a given _year_, _make_id_ and _model_id_ of vehicle.
+      # _parameters_ is an optional hash of parameters that further refine the
+      # scope of the search.
+      #
+      # Example:
+      #
+      #  vps = VehiclePartSearch.new( 2008, 73, 6039,
+      #				      { 'TrimId' => 20,
+      #				        'EngineId' => 8914 } )
+      #
+      # In this example, we look for parts that will fit a <b>2008</b> model
+      # *Audi* <b>R8</b> with *Base* trim and a <b>4.2L V8 Gas DOHC
+      # Distributorless Naturally Aspirated Bosch Motronic Electronic FI
+      # MFI</b> engine.
+      #
+      # Note that pagination of VehiclePartSearch results is not currently
+      # supported.
+      #
+      # Use VehicleSearch to learn the MakeId and ModelId of the vehicle in
+      # which you are interested.
+      #
+      def initialize(year, make_id, model_id, parameters={})
+	super( { 'Year'	   => year,
+		 'MakeId'  => make_id,
+		 'ModelId' => model_id }.merge( parameters ) )
+      end
+
+    end
+
+
+    # Search for vehicles.
+    #
+    class VehicleSearch < Operation
+
+      # Search for vehicles, based on one or more of the following
+      # _parameters_: Year, MakeId, ModelId and TrimId.
+      #
+      # This method is best used iteratively. For example, first search on
+      # year with a response group of *VehicleMakes* to return all makes for
+      # that year.
+      #
+      # Next, search on year and make with a response group of *VehicleModels*
+      # to find all models for that year and make.
+      #
+      # Then, search on year, make and model with a response group of
+      # *VehicleTrims* to find all trim packages for that year, make and model.
+      #
+      # Finally, if required, search on year, make, model and trim package
+      # with a response group of *VehicleOptions* to find all vehicle options
+      # for that year, make, model and trim package.
+      #
+      # Example:
+      #
+      #  vs = VehicleSearch.new( { 'Year' => 2008,
+      #				   'MakeId' => 20,
+      #				   'ModelId' => 6039,
+      #				   'TrimId' => 20 } )
+      #
+      # In this example, we search for <b>2008 Audi R8</b> vehicles with a
+      # *Base* trim package. Used with the *VehicleOptions* response group,
+      # a list of vehicle options would be returned.
+      #
+      def initialize(parameters={})
+	super
+      end
+
+    end
+
+    # Response groups determine which data pertaining to the item(s) being
+    # sought is returned. They can strongly influence the amount of data
+    # returned, so you should always use the smallest response group(s)
+    # containing the data of interest to you, to avoid masses of unnecessary
+    # data being returned.
+    #
+    class ResponseGroup
+
+      attr_reader :list, :params
+
+      # Define a set of one or more response groups to be applied to items
+      # retrieved by an AWS operation.
+      #
+      # If no response groups are given in _rg_ when instantiating an object,
+      # *Small* will be used by default.
+      #
+      # Example:
+      #
+      #  rg = ResponseGroup.new( 'Medium', 'Offers', 'Reviews' )
+      #
+      def initialize(*rg)
+	rg << 'Small' if rg.empty?
+	@list = rg
+	@params = { 'ResponseGroup' => @list.join( ',' ) }
+      end
+
+    end
+
+
+    # All dynamically generated exceptions occur within this namespace.
+    #
+    module Error
+
+      # The base exception class for errors that result from AWS operations.
+      # Classes for these are dynamically generated as subclasses of this one.
+      #
+      class AWSError < AmazonError; end
+
+      def Error.exception(xml)
+        err_class = xml.elements['Code'].text.sub( /^AWS.*\./, '' )
+        err_msg = xml.elements['Message'].text
+
+	# Dynamically define a new exception class for this class of error,
+	# unless it already exists.
+	#
+	unless Amazon::AWS::Error.const_defined?( err_class )
+	  Amazon::AWS::Error.const_set( err_class, Class.new( AWSError ) )
+	end
+
+	# Generate and return a new exception from the relevant class.
+	#
+	Amazon::AWS::Error.const_get( err_class ).new( err_msg )
+      end
+
+    end
+
+  end
+
+end