Implement error handling in Nokogiri parser

If a parse error occurred anywhere inside the XML, the Nokogiri parser just returns whatever happens to be on top of the stack. This makes it very hard to troubleshoot the error down the line.

To avoid this we need to implement the `error` method and raise an error. This introduces a new exception type Nori::ParseError which we also need to raise in the REXML parser to make it symmetric.

Author: stenlarsson

lib/nori.rb

@@ -3,6 +3,7 @@
 require "nori/xml_utility_node"
 
 class Nori
+  class ParseError < StandardError; end
 
   def self.hash_key(name, options = {})
     name = name.tr("-", "_") if options[:convert_dashes_to_underscores]

lib/nori/parser/nokogiri.rb

@@ -10,6 +10,7 @@ module Nokogiri
 
       class Document < ::Nokogiri::XML::SAX::Document
         attr_accessor :options
+        attr_accessor :last_error
 
         def stack
           @stack ||= []
@@ -44,13 +45,17 @@ def characters(string)
 
         alias cdata_block characters
 
+        def error(message)
+          @last_error = message
+        end
       end
 
       def self.parse(xml, options)
         document = Document.new
         document.options = options
         parser = ::Nokogiri::XML::SAX::Parser.new document
         parser.parse xml
+        raise ParseError, document.last_error if document.last_error
         document.stack.length > 0 ? document.stack.pop.to_hash : {}
       end
 

lib/nori/parser/rexml.rb

@@ -15,7 +15,11 @@ def self.parse(xml, options)
         parser = ::REXML::Parsers::BaseParser.new(xml)
 
         while true
-          raw_data = parser.pull
+          begin
+            raw_data = parser.pull
+          rescue ::REXML::ParseException => error
+            raise Nori::ParseError, error.message
+          end
           event = unnormalize(raw_data)
           case event[0]
           when :end_document

spec/nori/nori_spec.rb

@@ -640,6 +640,10 @@
         expect(parse(' ')).to eq({})
       end
 
+      it "raises error on missing end tag" do
+        expect { parse('<foo><bar>foo bar</foo>') }.to raise_error(Nori::ParseError)
+      end
+
     end
   end