Try to escape invalid UTF-8 characters during harvesting.

git-svn-id: http://oai.rubyforge.org/svn/trunk@948 4dc5e89f-90f6-0310-ab54-a6a856e7c30e

lib/oai/client.rb

@@ -155,7 +155,7 @@ def list_sets(opts={})
     def do_request(verb, opts = nil)
       # fire off the request and return appropriate DOM object
       uri = build_uri(verb, opts)
-      xml = get(uri)
+      xml = strip_invalid_utf_8_chars(get(uri))
       if @parser == 'libxml' 
         # remove default namespace for oai-pmh since libxml
         # isn't able to use our xpaths to get at them 
@@ -184,36 +184,20 @@ def encode(value)
     end
 
     def load_document(xml)
-      retried = false
       case @parser
       when 'libxml'
         begin
           parser = XML::Parser.new()
           parser.string = xml
           return parser.parse
         rescue XML::Parser::ParseError => e
-          if retried
-            raise OAI::Exception, 'response not well formed XML: '+e, caller
-          end
-          ic = Iconv.new('UTF-8//IGNORE', 'UTF-8')
-          xml2 = ic.iconv(xml << ' ')[0..-2]
-          puts "equal? #{xml == xml2}"
-          retried = true
-          retry
+          raise OAI::Exception, 'response not well formed XML: '+e, caller
         end
       when 'rexml'
         begin
           return REXML::Document.new(xml)
         rescue REXML::ParseException => e
-          if retried
-            puts xml
-            raise OAI::Exception, 'response not well formed XML: '+e, caller
-          end
-          puts "RETRYING"
-          ic = Iconv.new('UTF-8//IGNORE', 'UTF-8')
-          xml = ic.iconv(xml << ' ')[0..-2]
-          retried = true
-          retry
+          raise OAI::Exception, 'response not well formed XML: '+e, caller
         end
       end
     end
@@ -296,5 +280,19 @@ def parse_date(value)
       dt.utc
     end
 
+
+    # Strip out invalid UTF-8 characters.  Regex from the W3C, inverted.
+    # http://www.w3.org/International/questions/qa-forms-utf-8.en.php
+    def strip_invalid_utf_8_chars(xml)
+      simple_bytes = xml.gsub(/[\x00-\x08\x10\x0B\x0C\x0E-\x19\x7F]
+                             | [\x00-\x7F][\x80-\xBF]+
+                             | ([\xC0\xC1]|[\xF0-\xFF])[\x80-\xBF]*
+                             | [\xC2-\xDF]((?![\x80-\xBF])|[\x80-\xBF]{2,})
+                             | [\xE0-\xEF](([\x80-\xBF](?![\x80-\xBF]))
+                             | (?![\x80-\xBF]{2})|[\x80-\xBF]{3,})/x, '?')
+      simple_bytes.gsub(/\xE0[\x80-\x9F][\x80-\xBF]
+                       | \xED[\xA0-\xBF][\x80-\xBF]/,'?')
+    end
+
   end
 end

lib/oai/harvester.rb

@@ -9,6 +9,7 @@
 require 'chronic'
 require 'socket'
 
+require 'oai/client'
 require 'oai/harvester/config'
 require 'oai/harvester/harvest'
 require 'oai/harvester/logging'

lib/oai/harvester/harvest.rb

@@ -107,7 +107,7 @@ def call(url, opts)
       end
 
       def get_records(doc)
-        doc.find("/OAI-PMH/ListRecords/record").to_a
+        doc.doc.root.elements.to_a("/OAI-PMH/ListRecords/record")
       end
 
       def build_options_hash(site)

lib/oai/harvester/shell.rb

@@ -47,8 +47,8 @@ def start
             end
           rescue 
             puts "Not a recognized command, or bad options.  Type 'help' for clues."
-            #puts $!
-            #puts $!.backtrace.join("\n")
+            puts $!
+            puts $!.backtrace.join("\n")
           end
         end
       end

test/client/tc_list_records.rb

@@ -1,6 +1,7 @@
 require 'test_helper'
 
 class GetRecordsTest < Test::Unit::TestCase
+
   def test_get_records
     client = OAI::Client.new 'http://localhost:3333/oai'
     response = client.list_records 

test/client/tc_utf8_escaping.rb

@@ -0,0 +1,11 @@
+require 'test_helper'
+
+class UTF8Test < Test::Unit::TestCase
+
+  def test_escaping_invalid_utf_8_characters
+    client = OAI::Client.new 'http://localhost:3333/oai', :parser => 'libxml'
+    invalid_utf_8 = [2, 3, 4, 104, 5, 101, 6, 108, 66897, 108, 66535, 111, 1114112, 33, 55234123, 33].pack("U*")
+    assert_equal("hello!!", client.send(:strip_invalid_utf_8_chars, invalid_utf_8).gsub(/\?/, ''))
+  end
+
+end