#-- # Copyright (c) 2010 Herryanto Siatono # # Permission is hereby granted, free of charge, to any person obtaining # a copy of this software and associated documentation files (the # "Software"), to deal in the Software without restriction, including # without limitation the rights to use, copy, modify, merge, publish, # distribute, sublicense, and/or sell copies of the Software, and to # permit persons to whom the Software is furnished to do so, subject to # the following conditions: # # The above copyright notice and this permission notice shall be # included in all copies or substantial portions of the Software. # # THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, # EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF # MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. # IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY # CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, # TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE # SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #++ require 'net/http' require 'nokogiri' require 'cgi' require 'hmac-sha2' require 'base64' require 'openssl' module Amazon class RequestError < StandardError; end class Ecs VERSION = '2.2.5' SERVICE_URLS = { :us => 'http://ecs.amazonaws.com/onca/xml', :uk => 'http://ecs.amazonaws.co.uk/onca/xml', :ca => 'http://ecs.amazonaws.ca/onca/xml', :de => 'http://ecs.amazonaws.de/onca/xml', :jp => 'http://ecs.amazonaws.jp/onca/xml', :fr => 'http://ecs.amazonaws.fr/onca/xml', :it => 'http://webservices.amazon.it/onca/xml', :cn => 'http://webservices.amazon.cn/onca/xml', :es => 'http://webservices.amazon.es/onca/xml' } OPENSSL_DIGEST_SUPPORT = OpenSSL::Digest.constants.include?( 'SHA256' ) || OpenSSL::Digest.constants.include?( :SHA256 ) OPENSSL_DIGEST = OpenSSL::Digest.new( 'sha256' ) if OPENSSL_DIGEST_SUPPORT @@options = { :version => "2011-08-01", :service => "AWSECommerceService" } @@debug = false # Default search options def self.options @@options end # Set default search options def self.options=(opts) @@options = opts end # Get debug flag. def self.debug @@debug end # Set debug flag to true or false. def self.debug=(dbg) @@debug = dbg end def self.configure(&proc) raise ArgumentError, "Block is required." unless block_given? yield @@options end # Search amazon items with search terms. Default search index option is 'Books'. # For other search type other than keywords, please specify :type => [search type param name]. def self.item_search(terms, opts = {}) opts[:operation] = 'ItemSearch' opts[:search_index] = opts[:search_index] || 'Books' type = opts.delete(:type) if type opts[type.to_sym] = terms else opts[:keywords] = terms end self.send_request(opts) end # Search an item by ASIN no. def self.item_lookup(item_id, opts = {}) opts[:operation] = 'ItemLookup' opts[:item_id] = item_id self.send_request(opts) end # Search a browse node by BrowseNodeId def self.browse_node_lookup(browse_node_id, opts = {}) opts[:operation] = 'BrowseNodeLookup' opts[:browse_node_id] = browse_node_id self.send_request(opts) end # Generic send request to ECS REST service. You have to specify the :operation parameter. def self.send_request(opts) opts = self.options.merge(opts) if self.options # Include other required options opts[:timestamp] = Time.now.utc.strftime("%Y-%m-%dT%H:%M:%SZ") request_url = prepare_url(opts) log "Request URL: #{request_url}" res = Net::HTTP.get_response(URI::parse(request_url)) unless res.kind_of? Net::HTTPSuccess raise Amazon::RequestError, "HTTP Response: #{res.code} #{res.message}" end Response.new(res.body) end def self.validate_request(opts) raise Amazon::RequestError, "" if opts[:associate_tag] end # Response object returned after a REST call to Amazon service. class Response # XML input is in string format def initialize(xml) @doc = Nokogiri::XML(xml, nil, 'UTF-8') @doc.remove_namespaces! # @doc.xpath("//*").each { |elem| elem.name = elem.name.downcase } # @doc.xpath("//@*").each { |att| att.name = att.name.downcase } end # Return Nokogiri::XML::Document object. def doc @doc end # Return true if request is valid. def is_valid_request? Element.get(@doc, "//IsValid") == "True" end # Return true if response has an error. def has_error? !(error.nil? || error.empty?) end # Return error message. def error Element.get(@doc, "//Error/Message") end # Return error code def error_code Element.get(@doc, "//Error/Code") end # Return an array of Amazon::Element item objects. def items @items ||= (@doc/"Item").collect { |item| Element.new(item) } end # Return the first item (Amazon::Element) def first_item items.first end # Return current page no if :item_page option is when initiating the request. def item_page @item_page ||= Element.get(@doc, "//ItemPage").to_i end # Return total results. def total_results @total_results ||= Element.get(@doc, "//TotalResults").to_i end # Return total pages. def total_pages @total_pages ||= Element.get(@doc, "//TotalPages").to_i end def marshal_dump @doc.to_s end def marshal_load(xml) initialize(xml) end end protected def self.log(s) return unless self.debug if defined? RAILS_DEFAULT_LOGGER RAILS_DEFAULT_LOGGER.error(s) elsif defined? LOGGER LOGGER.error(s) else puts s end end private def self.prepare_url(opts) country = opts.delete(:country) country = (country.nil?) ? 'us' : country request_url = SERVICE_URLS[country.to_sym] raise Amazon::RequestError, "Invalid country '#{country}'" unless request_url secret_key = opts.delete(:AWS_secret_key) request_host = URI.parse(request_url).host qs = '' opts = opts.collect do |a,b| [camelize(a.to_s), b.to_s] end opts = opts.sort do |c,d| c[0].to_s <=> d[0].to_s end opts.each do |e| log "Adding #{e[0]}=#{e[1]}" next unless e[1] e[1] = e[1].join(',') if e[1].is_a? Array # v = URI.encode(e[1].to_s, Regexp.new("[^#{URI::PATTERN::UNRESERVED}]")) v = self.url_encode(e[1].to_s) qs << "&" unless qs.length == 0 qs << "#{e[0]}=#{v}" end signature = '' unless secret_key.nil? request_to_sign="GET\n#{request_host}\n/onca/xml\n#{qs}" signature = "&Signature=#{sign_request(request_to_sign, secret_key)}" end "#{request_url}?#{qs}#{signature}" end def self.url_encode(string) string.gsub( /([^a-zA-Z0-9_.~-]+)/ ) do '%' + $1.unpack( 'H2' * $1.bytesize ).join( '%' ).upcase end end def self.camelize(s) s.to_s.gsub(/\/(.?)/) { "::" + $1.upcase }.gsub(/(^|_)(.)/) { $2.upcase } end def self.sign_request(url, key) return nil if key.nil? if (OPENSSL_DIGEST_SUPPORT) signature = OpenSSL::HMAC.digest(OPENSSL_DIGEST, key, url) signature = [signature].pack('m').chomp else signature = Base64.encode64( HMAC::SHA256.digest(key, url) ).strip end signature = URI.escape(signature, Regexp.new("[+=]")) return signature end end # Internal wrapper class to provide convenient method to access Nokogiri element value. class Element class << self # Return the text value of an element. def get(element, path='.') return unless element result = element.at_xpath(path) result = result.inner_html if result result end # Return an unescaped text value of an element. def get_unescaped(element, path='.') result = self.get(element, path) CGI::unescapeHTML(result) if result end # Return an array of values based on the given path. def get_array(element, path='.') return unless element result = element/path if (result.is_a? Nokogiri::XML::NodeSet) || (result.is_a? Array) result.collect { |item| self.get(item) } else [self.get(result)] end end # Return child element text values of the given path. def get_hash(element, path='.') return unless element result = element.at_xpath(path) if result hash = {} result = result.children result.each do |item| hash[item.name] = item.inner_html end hash end end end # Pass Nokogiri::XML::Element object def initialize(element) @element = element end # Returns Nokogiri::XML::Element object def elem @element end # Returns a Nokogiri::XML::NodeSet of elements matching the given path. Example: element/"author". def /(path) elements = @element/path return nil if elements.size == 0 elements end # Return an array of Amazon::Element matching the given path def get_elements(path) elements = self./(path) return unless elements elements = elements.map{|element| Element.new(element)} end # Similar with search_and_convert but always return first element if more than one elements found def get_element(path) elements = get_elements(path) elements[0] if elements end # Get the text value of the given path, leave empty to retrieve current element value. def get(path='.') Element.get(@element, path) end # Get the unescaped HTML text of the given path. def get_unescaped(path='.') Element.get_unescaped(@element, path) end # Get the array values of the given path. def get_array(path='.') Element.get_array(@element, path) end # Get the children element text values in hash format with the element names as the hash keys. def get_hash(path='.') Element.get_hash(@element, path) end def attributes return unless self.elem self.elem.attributes end def to_s elem.to_s if elem end end end