bare-ruby-aws 0.1

data/lib/amazon.rb

@@ -0,0 +1,144 @@
+# $Id: amazon.rb,v 1.33 2010/03/19 17:20:46 ianmacd Exp $
+#
+
+module Amazon
+
+  # A top-level exception container class.
+  #
+  class AmazonError < StandardError; end
+
+  NAME = 'Ruby/Amazon'
+
+  @@config = {}
+
+  # We're going to have to use String#size if String#bytesize isn't available.
+  # This is for Ruby pre-1.8.7.
+  #
+  unless String.instance_methods.include? 'bytesize'
+    String.module_eval( 'alias :bytesize :size' )
+  end
+
+  # 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, but then altered
+    # slightly to please AWS.
+    #
+    string.gsub( /([^a-zA-Z0-9_.~-]+)/ ) do
+      '%' + $1.unpack( 'H2' * $1.bytesize ).join( '%' ).upcase
+    end
+  end
+
+  # Convert a string from CamelCase to ruby_case.
+  #
+  def Amazon.uncamelise(string)
+    # Avoid modifying by reference.
+    #
+    string = string.dup
+
+    # Don't mess with string if all caps.
+    #
+    if string =~ /[a-z]/
+      string.gsub!( /(.+?)(([A-Z][a-z]|[A-Z]+$))/, "\\1_\\2" )
+    end
+
+    # Convert to lower case.
+    #
+    string.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)
+      locale = nil
+
+      if config_str
+
+	# We have been passed a config file as a string.
+	#
+        config_files = [ config_str ]
+	config_class = File
+
+      else
+	puts 'No config file specified'
+
+      end
+
+      config_files.each do |cf|
+
+	# We must determine whether the file is readable.
+	#
+	readable = File.exists?( cf ) && File.readable?( cf )
+
+	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
+	else
+	  puts "could not open file"
+	end
+
+      end
+
+    end
+  end
+
+end

data/lib/amazon/aws.rb

@@ -0,0 +1,963 @@
+# $Id: aws.rb,v 1.130 2010/03/20 11:58:50 ianmacd Exp $
+#
+#:include: ../../README.rdoc
+
+module Amazon
+
+  module AWS
+
+    require 'amazon'
+    require 'amazon/aws/cache'
+    require 'enumerator'
+    require 'iconv'
+    require 'rexml/document'
+    require 'uri'
+
+    NAME = '%s/%s' % [ Amazon::NAME, 'AWS' ]
+    VERSION = '0.8.1'
+    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' => '2009-11-01'
+    }
+
+    # 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 appropriate
+    # for that kind of operation.
+    #
+    PAGINATION = {
+      'ItemSearch'	      => { 'parameter' => 'ItemPage',
+						  'max_page' => 400 },
+      'ItemLookup'	      => { 'parameter' => '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
+
+
+    # A hash to store character encoding converters.
+    #
+    @@encodings = {}
+
+
+    # Exception class for HTTP errors.
+    #
+    class HTTPError < AmazonError; end
+
+
+    # Exception class for faulty batch operations.
+    #
+    class BatchError < AmazonError; end
+
+
+    # Exception class for obsolete features.
+    #
+    class ObsolescenceError < 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)  # :nodoc:
+
+      url = ENDPOINT[request.locale].path + request.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
+
+      # Check whether we have a secret key available for signing the request.
+      # If so, sign the request for authentication.
+      #
+      if request.config['secret_key_id']
+	unless request.sign
+	  Amazon.dprintf( 'Warning! Failed to sign request. No OpenSSL support for SHA256 digest.' )
+	end
+
+	url = ENDPOINT[request.locale].path + request.query
+      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 EOFError,		Errno::ECONNABORTED, Errno::ECONNREFUSED,
+	     Errno::ECONNRESET, Errno::EPIPE,	     Errno::ETIMEDOUT,
+	     Timeout::Error	=> error
+	Amazon.dprintf( 'Connection to server lost: %s. Retrying...', error )
+	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, encoding=nil)  # :nodoc:
+
+      query = ''
+      @@encodings[encoding] ||= Iconv.new( 'utf-8', encoding ) if encoding
+
+      # 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 parameter values to strings, in case Symbols
+      # have been used as the values.
+      #
+      items.sort { |a,b| a.to_s <=> b.to_s }.each do |k, v|
+	if encoding
+	  query << '&%s=%s' %
+	    [ k, Amazon.url_encode( @@encodings[encoding].iconv( v.to_s ) ) ]
+	else
+	  query << '&%s=%s' % [ k, Amazon.url_encode( v.to_s ) ]
+	end
+      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 RUBY_VERSION >= '1.9.0'
+
+	    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 )
+
+	  # Return the instance variable that matches the method called.
+	  #
+	  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 )
+	elsif @__val__.respond_to?( method.id2name )
+
+	  # If our value responds to the method in question, call the method
+	  # on that.
+	  #
+	  @__val__.send( method.id2name )
+	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 ==(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 RUBY_VERSION >= '1.9.0'
+
+	  # 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[
+	ItemSearch
+
+
+      ]
+
+      attr_reader :kind
+      attr_accessor :params, :response_group
+
+      def initialize(parameters)
+
+	op_kind = self.class.to_s.sub( /^.*::/, '' )
+
+	raise "Bad operation: #{op_kind}" unless OPERATIONS.include?( op_kind )
+
+	if ResponseGroup::DEFAULT.key?( op_kind )
+	  response_group =
+	    ResponseGroup.new( ResponseGroup::DEFAULT[op_kind] )
+	else
+	  response_group = nil
+	end
+
+	if op_kind =~ /^Cart/
+	  @params = parameters
+	else
+	  @params = Hash.new { |hash, key| hash[key] = [] }
+	  @response_group = Hash.new { |hash, key| hash[key] = [] }
+
+	  unless op_kind == 'MultipleOperation'
+	    @params[op_kind] = [ parameters ]
+	    @response_group[op_kind] = [ response_group ]
+	  end
+	end
+
+	@kind = op_kind
+      end
+
+
+      # Make sure we can still get to the old @response_group= writer method.
+      #
+      alias :response_group_orig= :response_group=
+
+      # If the user assigns to @response_group, we need to set this response
+      # group for any and all operations that may have been batched.
+      #
+      def response_group=(rg) # :nodoc:
+        @params.each_value do |op_arr|
+          op_arr.each do |op|
+            op['ResponseGroup'] = rg
+          end
+        end
+      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 of different classes, use a
+      # MultipleOperation instead.
+      #
+      # Example:
+      #
+      #  is = ItemSearch.new( 'Books', { 'Title' => 'ruby programming' } )
+      #  is2 = ItemSearch.new( 'Music', { 'Artist' => 'stranglers' } )
+      #  is.response_group = ResponseGroup.new( :Small )
+      #  is2.response_group = ResponseGroup.new( :Tracks )
+      #  is.batch( is2 )
+      #
+      # Please see MultipleOperation.new for implementation details that also
+      # apply to batched operations.
+      #
+      def batch(*operations)
+
+	operations.flatten.each do |op|
+
+	  unless self.class == op.class
+	    raise BatchError, "You can't batch operations of different classes. Use class MultipleOperation."
+	  end
+
+	  # Add the operation's single element array containing the parameter
+	  # hash to the array.
+	  #
+	  @params[op.kind].concat( op.params[op.kind] )
+
+	  # Add the operation's response group array to the array.
+	  #
+	  @response_group[op.kind].concat( op.response_group[op.kind] )
+	end
+
+      end
+
+
+      # Return a hash of operation parameters and values, possibly converted to
+      # batch syntax, suitable for encoding in a query.
+      #
+      def query_parameters  # :nodoc:
+        query = {}
+
+        @params.each do |op_kind, ops|
+
+          # If we have only one type of operation and only one operation of
+	  # that type, return that one in non-batched syntax.
+          #
+          if @params.size == 1 && @params[op_kind].size == 1
+            return { 'Operation' => op_kind,
+          	   'ResponseGroup' => @response_group[op_kind][0] }.
+          	   merge( @params[op_kind][0] )
+          end
+
+          # Otherwise, use batch syntax.
+          #
+          ops.each_with_index do |op, op_index|
+
+            # Make sure we use a response group of some kind.
+            #
+            shared = '%s.%d.ResponseGroup' % [ op_kind, op_index + 1 ]
+            query[shared] = op['ResponseGroup'] ||
+          		  ResponseGroup::DEFAULT[op_kind]
+
+            # Add all of the parameters to the query hash.
+            #
+            op.each do |k, v|
+              shared = '%s.%d.%s' % [ op_kind, op_index + 1, k ]
+              query[shared] = v
+            end
+          end
+        end
+
+        # Add the operation list.
+        #
+        { 'Operation' => @params.keys.join( ',' ) }.merge( query )
+      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.
+      # - *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.
+      #
+      SEARCH_INDICES = %w[
+	All
+	Apparel
+	Automotive
+	Baby
+	Beauty
+	Blended
+	Books
+	Classical
+	DigitalMusic
+	DVD
+	Electronics
+	ForeignBooks
+	GourmetFood
+	Grocery
+	HealthPersonalCare
+	Hobbies
+	HomeGarden
+	HomeImprovement
+	Industrial
+	Jewelry
+	KindleStore
+	Kitchen
+	Lighting
+	Magazines
+	Merchants
+	Miscellaneous
+	MP3Downloads
+	Music
+	MusicalInstruments
+	MusicTracks
+	OfficeProducts
+	OutdoorLiving
+	Outlet
+	PCHardware
+	PetSupplies
+	Photo
+	Shoes
+	SilverMerchants
+	Software
+	SoftwareVideoGames
+	SportingGoods
+	Tools
+	Toys
+	UnboxVideo
+	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
+
+
+
+
+    # Response groups determine which data pertaining to the item(s) being
+    # sought is returned. They 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
+
+      # The default type of response group to use with each type of operation.
+      #
+      DEFAULT = {
+		  'ItemSearch'		  => :Large
+      }
+
+      # Define a set of one or more response groups to be applied to items
+      # retrieved by an AWS operation.
+      #
+      # Example:
+      #
+      #  rg = ResponseGroup.new( 'Medium', 'Offers', 'Reviews' )
+      #
+      def initialize(*rg)
+	@list = rg.join( ',' )
+      end
+
+
+      # We need a form we can interpolate into query strings.
+      #
+      def to_s	# :nodoc:
+	@list
+      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.
+	#
+	# Note that Ruby 1.9's Module.const_defined? needs a second parameter
+	# of *false*, or it will also search AWSError's ancestors.
+	#
+	cd_params = [ err_class ]
+	cd_params << false if RUBY_VERSION >= '1.9.0'
+
+	unless Amazon::AWS::Error.const_defined?( *cd_params )
+	  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
+
+
+    # Create a shorthand module method for each of the AWS operations. These
+    # can be used to create less verbose code at the expense of flexibility.
+    #
+    # For example, we might normally write the following code:
+    #
+    #  is = ItemSearch.new( 'Books', { 'Title' => 'Ruby' } )
+    #  rg = ResponseGroup.new( 'Large' )
+    #  req = Request.new
+    #  response = req.search( is, rg )
+    #
+    # but we could instead use ItemSearch's associated module method as
+    # follows:
+    #
+    #  response = Amazon::AWS.item_search( 'Books', { 'Title' => 'Ruby' } )
+    #
+    # Note that these equivalent module methods all attempt to use the *Large*
+    # response group, which may or may not work. If an
+    # Amazon::AWS::Error::InvalidResponseGroup is raised, we will scan the
+    # text of the error message returned by AWS to try to glean a valid
+    # response group and then retry the operation using that instead.
+
+
+    # Obtain a list of all subclasses of the Operation class.
+    #
+    classes =
+      ObjectSpace.enum_for( :each_object, class << Operation; self; end ).to_a
+
+    classes.each do |cl|
+      # Convert class name to Ruby case, e.g. ItemSearch => item_search.
+      #
+      class_name = cl.to_s.sub( /^.+::/, '' )
+      uncamelised_name = Amazon.uncamelise( class_name )
+
+      # Define the module method counterpart of each operation.
+      #
+      module_eval %Q(
+	def AWS.#{uncamelised_name}(*params)
+	  # Instantiate an object of the desired operational class.
+	  #
+	  op = #{cl.to_s}.new( *params )
+
+	  # Attempt a search for the given operation using its default
+	  # response group types.
+	  #
+	  results = Search::Request.new.search( op )
+	  yield results if block_given?
+	  return results
+
+	end
+      )
+    end
+
+  end
+
+end