always send back SOAP responses as UTF-8, we can't guarantee that SOAP4R

supports any encoding sent by caller.

add documentation describing how to ensure :string types don't get converted
into :base64 by SOAP4R when containing non-ASCII chars and $KCODE is not set
to a value.


git-svn-id: http://svn-commit.rubyonrails.org/rails/trunk@1822 5ecf4fe2-1ee6-0310-87b1-e25e094e27de

actionwebservice/README

@@ -215,6 +215,25 @@ custom types and message definition types:
 The default namespace used is 'urn:ActionWebService', if you don't supply
 one.
 
+
+== ActionWebService and UTF-8
+
+If you're going to be sending back strings containing non-ASCII UTF-8
+characters using the <tt>:string</tt> data type, you need to make sure that
+Ruby is using UTF-8 as the default encoding for its strings.
+
+The default in Ruby to use US-ASCII encoding for strings, which causes a string
+validation check in the Ruby SOAP library to fail and your string to be sent
+back as a Base-64 value, which may confuse clients that expected strings
+because of the WSDL.
+
+Two ways of doing setting the default string encoding are:
+
+* Start Ruby using the <tt>-Ku</tt> command-line option to the Ruby executable
+* Set the <tt>$KCODE</tt> flag in <tt>config/environment.rb</tt> to the
+  string <tt>'UTF8'</tt>
+
+
 == Testing your APIs
 
 

actionwebservice/lib/action_web_service/protocol/soap_protocol.rb

@@ -11,7 +11,8 @@ module ActionWebService # :nodoc:
       end
       
       class SoapProtocol < AbstractProtocol # :nodoc:
-        DefaultEncoding = 'utf-8'
+        AWSEncoding = 'UTF-8'
+        XSDEncoding = 'UTF8'
 
         attr :marshaler
 
@@ -27,10 +28,9 @@ module ActionWebService # :nodoc:
         def decode_action_pack_request(action_pack_request)
           return nil unless soap_action = has_valid_soap_action?(action_pack_request)
           service_name = action_pack_request.parameters['action']
-          charset = parse_charset(action_pack_request.env['HTTP_CONTENT_TYPE'])
           protocol_options = { 
             :soap_action => soap_action,
-            :charset  => charset
+            :charset  => AWSEncoding
           }
           decode_request(action_pack_request.raw_post, service_name, protocol_options)
         end
@@ -42,8 +42,7 @@ module ActionWebService # :nodoc:
         end
 
         def decode_request(raw_request, service_name, protocol_options={})
-          charset = protocol_options[:charset] || DefaultEncoding
-          envelope = SOAP::Processor.unmarshal(raw_request, :charset => charset)
+          envelope = SOAP::Processor.unmarshal(raw_request, :charset => AWSEncoding)
           unless envelope
             raise ProtocolError, "Failed to parse SOAP request message"
           end
@@ -109,8 +108,17 @@ module ActionWebService # :nodoc:
             end
           end
           envelope = create_soap_envelope(response)
-          charset = protocol_options[:charset] || DefaultEncoding
-          Response.new(SOAP::Processor.marshal(envelope, :charset => charset), "text/xml; charset=#{charset}", return_value)
+
+          # FIXME: This is not thread-safe, but StringFactory_ in SOAP4R only
+          #        reads target encoding from the XSD::Charset.encoding variable.
+          #        This is required to ensure $KCODE strings are converted
+          #        correctly to UTF-8 for any values of $KCODE.
+          previous_encoding = XSD::Charset.encoding
+          XSD::Charset.encoding = XSDEncoding
+          response_body = SOAP::Processor.marshal(envelope, :charset => AWSEncoding)
+          XSD::Charset.encoding = previous_encoding
+
+          Response.new(response_body, "text/xml; charset=#{AWSEncoding}", return_value)
         end
 
         def protocol_client(api, protocol_name, endpoint_uri, options={})
@@ -138,15 +146,6 @@ module ActionWebService # :nodoc:
             soap_action
           end
 
-          def parse_charset(content_type)
-            return DefaultEncoding if content_type.nil?
-            if /^text\/xml(?:\s*;\s*charset=([^"]+|"[^"]+"))$/i =~ content_type
-              $1
-            else
-              DefaultEncoding
-            end
-          end
-
           def create_soap_envelope(body)
             header = SOAP::SOAPHeader.new
             body = SOAP::SOAPBody.new(body)

actionwebservice/test/abstract_dispatcher.rb

@@ -4,6 +4,7 @@ require 'stringio'
 class ActionController::Base; def rescue_action(e) raise e end; end
 
 module DispatcherTest
+  Utf8String = "One World Caf\303\251"
   WsdlNamespace = 'http://rubyonrails.com/some/namespace'
 
   class Node < ActiveRecord::Base
@@ -58,6 +59,7 @@ module DispatcherTest
     api_method :hash_struct_return, :returns => [[Person]]
     api_method :thrower
     api_method :void
+    api_method :test_utf8, :returns => [:string]
     api_method :hex, :expects => [:base64], :returns => [:string]
     api_method :unhex, :expects => [:string], :returns => [:base64]
   end
@@ -223,6 +225,10 @@ module DispatcherTest
       @void_called = @method_params
     end
 
+    def test_utf8
+      Utf8String
+    end
+
     def hex(s)
       return s.unpack("H*")[0]
     end
@@ -398,12 +404,6 @@ module DispatcherCommonTests
       raise NotImplementedError
     end
 
-    def update_request(ap_request)
-    end
-
-    def check_response(ap_response)
-    end
-
     def protocol
       @protocol
     end
@@ -453,11 +453,10 @@ module DispatcherCommonTests
       # puts body
       ap_request = protocol.encode_action_pack_request(service_name, public_method_name, body, :request_class => ActionController::TestRequest)
       ap_request.env.update(request_env)
-      update_request(ap_request)
       ap_response = ActionController::TestResponse.new
       container.process(ap_request, ap_response)
       # puts ap_response.body
-      check_response(ap_response)
+      @response_body = ap_response.body
       public_method_name, return_value = protocol.decode_response(ap_response.body)
       unless is_exception?(return_value) || virtual
         return_value = method.cast_returns(return_value)

actionwebservice/test/dispatcher_action_controller_soap_test.rb

@@ -67,16 +67,23 @@ class TC_DispatcherActionControllerSoap < Test::Unit::TestCase
     assert_equal(["bloggerCat1", "bloggerCat2"], blogger_cats)
   end
 
+  def test_utf8
+    @direct_controller.web_service_exception_reporting = true
+    $KCODE = 'u'
+    assert_equal(Utf8String, do_method_call(@direct_controller, 'TestUtf8'))
+
+    # If $KCODE is not set to UTF-8, any strings with non-ASCII UTF-8 data
+    # will be sent back as base64 by SOAP4R. By the time we get it here though,
+    # it will be decoded back into a string. So lets read the base64 value
+    # from the message body directly.
+    $KCODE = 'NONE'
+    do_method_call(@direct_controller, 'TestUtf8')
+    retval = SOAP::Processor.unmarshal(@response_body).body.response
+    assert retval.is_a?(SOAP::SOAPBase64)
+    assert_equal "T25lIFdvcmxkIENhZsOp", retval.data.to_s
+  end
+
   protected
-    def update_request(ap_request)
-      ap_request.env.update('HTTP_CONTENT_TYPE' => 'text/xml; charset=us-ascii')
-    end
-
-    def check_response(ap_response)
-      assert_equal 'text/xml; charset=us-ascii', ap_response.headers['Content-Type']
-      assert_match /xml.*?encoding="us-ascii"/, ap_response.body
-    end
-
     def exception_message(soap_fault_exception)
       soap_fault_exception.detail.cause.message
     end