nokogiri 1.19.4-java → 1.19.5-java

data/lib/nokogiri/css/parser.rb

@@ -1,7 +1,7 @@
 # frozen_string_literal: true
 #
 # DO NOT MODIFY!!!!
-# This file is automatically generated by Racc 1.8.0
+# This file is automatically generated by Racc 1.8.1
 # from Racc grammar file "parser.y".
 #
 
@@ -23,19 +23,36 @@ module Nokogiri
     class Parser < Racc::Parser
 
 
-def unescape_css_identifier(identifier)
-  identifier.gsub(/\\(?:([^0-9a-fA-F])|([0-9a-fA-F]{1,6})\s?)/){ |m| $1 || [$2.hex].pack('U') }
+def self.unescape_css_identifier(identifier)
+  identifier.gsub(/\\(?:([^0-9a-fA-F])|([0-9a-fA-F]{1,6})(?:\r\n|[ \t\n\r\f])?)/){ |m| $1 || unescape_css_code_point($2) }
 end
 
-def unescape_css_string(str)
-  str.gsub(/\\(?:([^0-9a-fA-F])|([0-9a-fA-F]{1,6})\s?)/) do |m|
-    if $1=="\n"
+def self.unescape_css_string(str)
+  str.gsub(/\\(?:(\r\n|[^0-9a-fA-F])|([0-9a-fA-F]{1,6})(?:\r\n|[ \t\n\r\f])?)/) do |m|
+    if ["\n", "\r\n", "\r", "\f"].include?($1)
       ''
     else
-      $1 || [$2.hex].pack('U')
+      $1 || unescape_css_code_point($2)
     end
   end
 end
+
+# https://www.w3.org/TR/css-syntax-3/#consume-escaped-code-point
+def self.unescape_css_code_point(hex)
+  code_point = hex.hex
+  if code_point == 0 || code_point.between?(0xD800, 0xDFFF) || code_point > 0x10FFFF
+    "\u{FFFD}"
+  else
+    [code_point].pack('U')
+  end
+end
+
+# A decoded "*" would turn into the universal selector.
+def unescape_css_element_name(identifier)
+  name = Parser.unescape_css_identifier(identifier)
+  raise SyntaxError, "Invalid element name '#{identifier}'" if name == "*"
+  name
+end
 ##### State transition tables begin ###
 
 racc_action_table = [
@@ -402,7 +419,7 @@ def _reduce_8(val, _values, result)
 end
 
 def _reduce_9(val, _values, result)
- result = val[1]
+ result = Parser.unescape_css_identifier(val[1])
     result
 end
 
@@ -458,7 +475,7 @@ end
 # reduce 20 omitted
 
 def _reduce_21(val, _values, result)
- result = Node.new(:CLASS_CONDITION, [unescape_css_identifier(val[1])])
+ result = Node.new(:CLASS_CONDITION, [Parser.unescape_css_identifier(val[1])])
     result
 end
 
@@ -470,14 +487,13 @@ def _reduce_23(val, _values, result)
 end
 
 def _reduce_24(val, _values, result)
- result = Node.new(:ELEMENT_NAME, [val[0], val[2]])
+      result = Node.new(:ELEMENT_NAME, [val[0] && unescape_css_element_name(val[0]), unescape_css_element_name(val[2])])
+
     result
 end
 
 def _reduce_25(val, _values, result)
-      name = val[0]
-      result = Node.new(:ELEMENT_NAME, [name])
-
+ result = Node.new(:ELEMENT_NAME, [unescape_css_element_name(val[0])])
     result
 end
 
@@ -507,12 +523,13 @@ def _reduce_30(val, _values, result)
 end
 
 def _reduce_31(val, _values, result)
- result = Node.new(:ATTRIB_NAME, [[val[0], val[2]].compact.join(':')])
+      result = Node.new(:ATTRIB_NAME, [val[0], val[2]].compact.map { |name| Parser.unescape_css_identifier(name) })
+
     result
 end
 
 def _reduce_32(val, _values, result)
- result = Node.new(:ATTRIB_NAME, [val[0]])
+ result = Node.new(:ATTRIB_NAME, [Parser.unescape_css_identifier(val[0])])
     result
 end
 
@@ -691,17 +708,17 @@ end
 # reduce 63 omitted
 
 def _reduce_64(val, _values, result)
- result = Node.new(:ID, [unescape_css_identifier(val[0])])
+ result = Node.new(:ID, [Parser.unescape_css_identifier(val[0])])
     result
 end
 
 def _reduce_65(val, _values, result)
- result = [val[0], unescape_css_identifier(val[1])]
+ result = [val[0], Parser.unescape_css_identifier(val[1])]
     result
 end
 
 def _reduce_66(val, _values, result)
- result = [val[0], unescape_css_string(val[1])]
+ result = [val[0], Parser.unescape_css_string(val[1][1..-2])]
     result
 end
 

data/lib/nokogiri/css/parser.y

@@ -22,7 +22,7 @@ rule
   ;
 
   xpath_attribute_name:
-    '@' IDENT { result = val[1] }
+    '@' IDENT { result = Parser.unescape_css_identifier(val[1]) }
   ;
 
   xpath_attribute:
@@ -55,7 +55,7 @@ rule
   ;
 
   class:
-    '.' IDENT { result = Node.new(:CLASS_CONDITION, [unescape_css_identifier(val[1])]) }
+    '.' IDENT { result = Node.new(:CLASS_CONDITION, [Parser.unescape_css_identifier(val[1])]) }
   ;
 
   element_name:
@@ -64,11 +64,10 @@ rule
   ;
 
   namespaced_ident:
-    namespace '|' IDENT { result = Node.new(:ELEMENT_NAME, [val[0], val[2]]) }
-  | IDENT {
-      name = val[0]
-      result = Node.new(:ELEMENT_NAME, [name])
+    namespace '|' IDENT {
+      result = Node.new(:ELEMENT_NAME, [val[0] && unescape_css_element_name(val[0]), unescape_css_element_name(val[2])])
     }
+  | IDENT { result = Node.new(:ELEMENT_NAME, [unescape_css_element_name(val[0])]) }
   ;
 
   namespace:
@@ -89,8 +88,10 @@ rule
   ;
 
   attrib_name:
-    namespace '|' IDENT { result = Node.new(:ATTRIB_NAME, [[val[0], val[2]].compact.join(':')]) }
-  | IDENT { result = Node.new(:ATTRIB_NAME, [val[0]]) }
+    namespace '|' IDENT {
+      result = Node.new(:ATTRIB_NAME, [val[0], val[2]].compact.map { |name| Parser.unescape_css_identifier(name) })
+    }
+  | IDENT { result = Node.new(:ATTRIB_NAME, [Parser.unescape_css_identifier(val[0])]) }
   | xpath_attribute
   ;
 
@@ -209,12 +210,12 @@ rule
   ;
 
   attribute_id:
-    HASH { result = Node.new(:ID, [unescape_css_identifier(val[0])]) }
+    HASH { result = Node.new(:ID, [Parser.unescape_css_identifier(val[0])]) }
   ;
 
   attrib_val_0or1:
-    eql_incl_dash IDENT { result = [val[0], unescape_css_identifier(val[1])] }
-  | eql_incl_dash STRING { result = [val[0], unescape_css_string(val[1])] }
+    eql_incl_dash IDENT { result = [val[0], Parser.unescape_css_identifier(val[1])] }
+  | eql_incl_dash STRING { result = [val[0], Parser.unescape_css_string(val[1][1..-2])] }
   | eql_incl_dash NUMBER { result = [val[0], val[1]] }
   |
   ;
@@ -262,16 +263,33 @@ end
 
 ---- inner
 
-def unescape_css_identifier(identifier)
-  identifier.gsub(/\\(?:([^0-9a-fA-F])|([0-9a-fA-F]{1,6})\s?)/){ |m| $1 || [$2.hex].pack('U') }
+def self.unescape_css_identifier(identifier)
+  identifier.gsub(/\\(?:([^0-9a-fA-F])|([0-9a-fA-F]{1,6})(?:\r\n|[ \t\n\r\f])?)/){ |m| $1 || unescape_css_code_point($2) }
 end
 
-def unescape_css_string(str)
-  str.gsub(/\\(?:([^0-9a-fA-F])|([0-9a-fA-F]{1,6})\s?)/) do |m|
-    if $1=="\n"
+def self.unescape_css_string(str)
+  str.gsub(/\\(?:(\r\n|[^0-9a-fA-F])|([0-9a-fA-F]{1,6})(?:\r\n|[ \t\n\r\f])?)/) do |m|
+    if ["\n", "\r\n", "\r", "\f"].include?($1)
       ''
     else
-      $1 || [$2.hex].pack('U')
+      $1 || unescape_css_code_point($2)
     end
   end
 end
+
+# https://www.w3.org/TR/css-syntax-3/#consume-escaped-code-point
+def self.unescape_css_code_point(hex)
+  code_point = hex.hex
+  if code_point == 0 || code_point.between?(0xD800, 0xDFFF) || code_point > 0x10FFFF
+    "\u{FFFD}"
+  else
+    [code_point].pack('U')
+  end
+end
+
+# A decoded "*" would turn into the universal selector.
+def unescape_css_element_name(identifier)
+  name = Parser.unescape_css_identifier(identifier)
+  raise SyntaxError, "Invalid element name '#{identifier}'" if name == "*"
+  name
+end

data/lib/nokogiri/css/xpath_visitor.rb

@@ -9,6 +9,11 @@ module Nokogiri
     class XPathVisitor
       WILDCARD_NAMESPACES = Nokogiri.libxml2_patches.include?("0009-allow-wildcard-namespaces.patch") # :nodoc:
 
+      # The ASCII characters allowed in an XML NCName. Every XPath delimiter and operator is ASCII,
+      # so non-ASCII characters are left for the XPath parser to accept or reject.
+      XPATH_NAME = /\A(?:[A-Za-z_]|[^\x00-\x7F])(?:[A-Za-z0-9_.-]|[^\x00-\x7F])*\z/
+      private_constant :XPATH_NAME
+
       # Enum to direct XPathVisitor when to use Nokogiri builtin XPath functions.
       module BuiltinsConfig
         # Never use Nokogiri builtin functions, always generate vanilla XPath 1.0 queries. This is
@@ -136,7 +141,7 @@ module Nokogiri
         when /^(last|last-of-type)\(/
           "position()=last()"
         when /^contains\(/
-          "contains(.,#{node.value[1]})"
+          "contains(.,#{xpath_function_argument(node.value[1])})"
         when /^gt\(/
           "position()>#{node.value[1]}"
         when /^only-child\(/
@@ -151,9 +156,7 @@ module Nokogiri
 
           # xpath function call, let's marshal those arguments
           args = ["."]
-          args += node.value[1..-1].map do |n|
-            n.is_a?(Nokogiri::CSS::Node) ? n.accept(self) : n
-          end
+          args += node.value[1..-1].map { |arg| xpath_function_argument(arg) }
           "nokogiri:#{node.value.first}#{args.join(",")})"
         end
       end
@@ -168,24 +171,15 @@ module Nokogiri
       end
 
       def visit_id(node)
-        node.value.first =~ /^#(.*)$/
-        "@id='#{Regexp.last_match(1)}'"
+        "@id=#{xpath_string_literal(node.value.first.delete_prefix("#"))}"
       end
 
       def visit_attribute_condition(node)
         attribute = node.value.first.accept(self)
         return attribute if node.value.length == 1
 
-        value = node.value.last
-        value = "'#{value}'" unless /^['"]/.match?(value)
-
-        # quoted values - see test_attribute_value_with_quotes in test/css/test_parser.rb
-        if (value[0] == value[-1]) && %q{"'}.include?(value[0])
-          str_value = value[1..-2]
-          if str_value.include?(value[0])
-            value = 'concat("' + str_value.split('"', -1).join(%q{",'"',"}) + '","")'
-          end
-        end
+        str_value = node.value.last
+        value = xpath_string_literal(str_value)
 
         case node.value[1]
         when :equal
@@ -199,8 +193,7 @@ module Nokogiri
         when :dash_match
           "#{attribute}=#{value} or starts-with(#{attribute},concat(#{value},'-'))"
         when :includes
-          value = value[1..-2] # strip quotes
-          css_class(attribute, value)
+          css_class(attribute, str_value)
         when :suffix_match
           "substring(#{attribute},string-length(#{attribute})-string-length(#{value})+1,string-length(#{value}))=#{value}"
         else
@@ -265,17 +258,21 @@ module Nokogiri
       end
 
       def visit_element_name(node)
-        if @doctype == DoctypeConfig::HTML5 && html5_element_name_needs_namespace_handling(node)
+        html5_local_name = @doctype == DoctypeConfig::HTML5 && html5_element_name_needs_namespace_handling(node)
+        node.value.each { |name| validate_xpath_element_name(name) } unless html5_local_name
+
+        if html5_local_name
           # HTML5 has namespaces that should be ignored in CSS queries
           # https://github.com/sparklemotion/nokogiri/issues/2376
+          name = node.value.first
           if @builtins == BuiltinsConfig::ALWAYS || (@builtins == BuiltinsConfig::OPTIMAL && Nokogiri.uses_libxml?)
-            if WILDCARD_NAMESPACES
-              "*:#{node.value.first}"
+            if WILDCARD_NAMESPACES && xpath_name?(name)
+              "*:#{name}"
             else
-              "*[nokogiri-builtin:local-name-is('#{node.value.first}')]"
+              "*[nokogiri-builtin:local-name-is(#{xpath_string_literal(name)})]"
             end
           else
-            "*[local-name()='#{node.value.first}']"
+            "*[local-name()=#{xpath_string_literal(name)}]"
           end
         elsif node.value.length == 2 # has a namespace prefix
           if node.value.first.nil? # namespace prefix is empty
@@ -292,7 +289,18 @@ module Nokogiri
       end
 
       def visit_attrib_name(node)
-        "@#{node.value.first}"
+        if node.value.length == 2 # has a namespace prefix
+          unless node.value.all? { |name| xpath_name?(name) }
+            raise Nokogiri::CSS::SyntaxError, "Invalid attribute name '#{node.value.join("|")}'"
+          end
+
+          "@#{node.value.join(":")}"
+        elsif xpath_name?(node.value.first)
+          "@#{node.value.first}"
+        else
+          # HTML parsers allow attribute names like "@click" and ":class". On JRuby, local-name() of ":class" is "class".
+          "@*[namespace-uri()='' and name()=#{xpath_string_literal(node.value.first)}]"
+        end
       end
 
       def accept(node)
@@ -307,6 +315,27 @@ module Nokogiri
         end
       end
 
+      def validate_xpath_element_name(name)
+        unless name.nil? || name == "*" || xpath_name?(name)
+          raise Nokogiri::CSS::SyntaxError, "Invalid element name '#{name}'"
+        end
+      end
+
+      def xpath_name?(name)
+        XPATH_NAME.match?(name)
+      end
+
+      # Function arguments reach the visitor as raw tokens, and only a STRING token starts with a quote.
+      def xpath_function_argument(arg)
+        if arg.is_a?(Nokogiri::CSS::Node)
+          arg.accept(self)
+        elsif arg.is_a?(String) && arg.start_with?("'", '"')
+          xpath_string_literal(Parser.unescape_css_string(arg[1..-2]))
+        else
+          arg
+        end
+      end
+
       def html5_element_name_needs_namespace_handling(node)
         # if there is already a namespace (i.e., it is a prefixed QName), use it as normal
         node.value.length == 1 &&
@@ -326,7 +355,9 @@ module Nokogiri
           options[:last] ? "(last()-position()+1)" : "position()"
         end
 
-        if b.zero?
+        if a.zero?
+          "#{position}=#{b}"
+        elsif b.zero? && a.positive?
           "(#{position} mod #{a})=0"
         else
           compare = a < 0 ? "<=" : ">="
@@ -345,7 +376,7 @@ module Nokogiri
           b = values[3].to_i
         elsif op == "-"
           a = values[0].to_i
-          b = a - (values[3].to_i % a)
+          b = a.zero? ? -values[3].to_i : a - (values[3].to_i % a)
         else
           raise ArgumentError, "expected an+b node to have either + or - as the operator, but is #{op.inspect}"
         end
@@ -365,10 +396,22 @@ module Nokogiri
       def css_class(hay, needle)
         if @builtins == BuiltinsConfig::ALWAYS || (@builtins == BuiltinsConfig::OPTIMAL && Nokogiri.uses_libxml?)
           # use the builtin implementation
-          "nokogiri-builtin:css-class(#{hay},'#{needle}')"
+          "nokogiri-builtin:css-class(#{hay},#{xpath_string_literal(needle)})"
         else
           # use only ordinary xpath functions
-          "contains(concat(' ',normalize-space(#{hay}),' '),' #{needle} ')"
+          "contains(concat(' ',normalize-space(#{hay}),' '),#{xpath_string_literal(" #{needle} ")})"
+        end
+      end
+
+      # XPath 1.0 string literals have no escape mechanism, so a string containing both quote
+      # characters has to be assembled with concat().
+      def xpath_string_literal(string)
+        if !string.include?("'")
+          "'#{string}'"
+        elsif !string.include?('"')
+          %("#{string}")
+        else
+          "concat('" + string.split("'", -1).join(%q{',"'",'}) + "')"
         end
       end
     end

data/lib/nokogiri/html5/document.rb

@@ -48,6 +48,11 @@ module Nokogiri
       # Since v1.14.0
       attr_reader :quirks_mode
 
+      # Whether the parser treated the content of +noscript+ elements as text, or +nil+ before the
+      # first parse. Fragments parsed into this document must match it, because the serializer emits
+      # +noscript+ text literally only when this is true.
+      attr_accessor :parse_noscript_content_as_text # :nodoc:
+
       class << self
         # :call-seq:
         #   parse(input) { |options| ... } → HTML5::Document
@@ -150,8 +155,11 @@ module Nokogiri
           options[:max_errors] ||= options.delete(:max_parse_errors) || Nokogiri::Gumbo::DEFAULT_MAX_ERRORS
           options[:max_tree_depth] ||= Nokogiri::Gumbo::DEFAULT_MAX_TREE_DEPTH
 
+          noscript_content_as_text = options[:parse_noscript_content_as_text] ? true : false
+
           doc = Nokogiri::Gumbo.parse(string, url, self, **options)
           doc.encoding = "UTF-8"
+          doc.parse_noscript_content_as_text = noscript_content_as_text
           doc
         end
       end
@@ -160,6 +168,7 @@ module Nokogiri
         super
         @url = nil
         @quirks_mode = nil
+        @parse_noscript_content_as_text = nil
       end
 
       # :call-seq:

data/lib/nokogiri/html5/document_fragment.rb

@@ -130,7 +130,9 @@ module Nokogiri
       #   element. (default +Nokogiri::Gumbo::DEFAULT_MAX_ATTRIBUTES+)
       #
       # - +parse_noscript_content_as_text:+ (Boolean) Whether to parse the content of +noscript+
-      #   elements as text. (default +false+)
+      #   elements as text. (default: the value +document+ was parsed with, or +false+) Raises
+      #   ArgumentError if +document+ was parsed with a different value, because the serializer
+      #   escapes +noscript+ text according to the document's value.
       #
       # See rdoc-ref:HTML5@Parsing+options for a complete description of these parsing options.
       #
@@ -162,8 +164,21 @@ module Nokogiri
         options[:max_attributes] ||= Nokogiri::Gumbo::DEFAULT_MAX_ATTRIBUTES
         options[:max_errors] ||= options.delete(:max_parse_errors) || Nokogiri::Gumbo::DEFAULT_MAX_ERRORS
         options[:max_tree_depth] ||= Nokogiri::Gumbo::DEFAULT_MAX_TREE_DEPTH
+        noscript_content_as_text = options[:parse_noscript_content_as_text] = noscript_parsing_mode(doc, options)
 
         Nokogiri::Gumbo.fragment(self, input, context, **options)
+        doc.parse_noscript_content_as_text = noscript_content_as_text if doc.parse_noscript_content_as_text.nil?
+      end
+
+      private def noscript_parsing_mode(doc, options)
+        doc_mode = doc.parse_noscript_content_as_text
+        mode = options.fetch(:parse_noscript_content_as_text, doc_mode) ? true : false
+
+        if !doc_mode.nil? && mode != doc_mode
+          raise ArgumentError, "parse_noscript_content_as_text: #{mode} does not match the document's value #{doc_mode}"
+        end
+
+        mode
       end
 
       def serialize(options = {}, &block) # :nodoc:

data/lib/nokogiri/html5/node.rb

@@ -61,7 +61,7 @@ module Nokogiri
           native_write_to(io, encoding, indent_string, config_options)
         else
           # Serialize including the current node.
-          html = html_standard_serialize(options[:preserve_newline] || false)
+          html = html_standard_serialize(options[:preserve_newline] || false, document.parse_noscript_content_as_text)
           encoding ||= document.encoding || Encoding::UTF_8
           io << html.encode(encoding, fallback: lambda { |c| "&#x#{c.ord.to_s(16)};" })
         end

data/lib/nokogiri/html5.rb

@@ -181,6 +181,22 @@ module Nokogiri
   #   #          })]
   #   #      })
   #
+  # ⚠ HTML parsed with <tt>parse_noscript_content_as_text: false</tt> and then serialized may
+  # parse differently in a browser that supports scripting. A comment or a raw-text element (such
+  # as +style+ or +xmp+) inside +noscript+ may contain the text <tt></noscript></tt>. The
+  # serializer emits it unchanged, as the HTML standard requires, and the browser ends the
+  # +noscript+ element there and parses the rest as markup.
+  #
+  #   frag = Nokogiri::HTML5.fragment("<noscript><!--</noscript><img src=x onerror=alert(1)>--></noscript>")
+  #   frag.css("img") # => []
+  #   puts frag.to_html
+  #   # => <noscript><!--</noscript><img src=x onerror=alert(1)>--></noscript>
+  #   # A browser that supports scripting parses this as a noscript element followed by an img element.
+  #
+  # Applications that sanitize HTML and allow +noscript+ elements should parse with
+  # <tt>parse_noscript_content_as_text: true</tt>, which exposes that +img+ element to the
+  # sanitizer, or should remove +noscript+ elements.
+  #
   # == HTML Serialization
   #
   # After parsing HTML, it may be serialized using any of the Nokogiri::XML::Node serialization

data/lib/nokogiri/version/constant.rb

@@ -2,5 +2,5 @@
 
 module Nokogiri
   # The version of Nokogiri you are using
-  VERSION = "1.19.4"
+  VERSION = "1.19.5"
 end

data/lib/nokogiri/xml/attr.rb

@@ -8,6 +8,8 @@ module Nokogiri
       alias_method :to_s, :content
       alias_method :content=, :value=
 
+      undef_method :line= if method_defined?(:line=)
+
       #
       #  :call-seq: deconstruct_keys(array_of_names) → Hash
       #

data/lib/nokogiri/xml/attribute_decl.rb

@@ -10,7 +10,8 @@ module Nokogiri
       undef_method :content
       undef_method :namespace
       undef_method :namespace_definitions
-      undef_method :line if method_defined?(:line)
+      undef_method :line
+      undef_method :line= if method_defined?(:line=)
 
       private
 

data/lib/nokogiri/xml/builder.rb

@@ -314,7 +314,7 @@ module Nokogiri
 
         @context = nil
         @arity = nil
-        @ns = nil
+        @ns_prefix = nil
 
         options = DEFAULT_DOCUMENT_OPTIONS.merge(options)
         options.each do |k, v|
@@ -355,20 +355,8 @@ module Nokogiri
       ###
       # Build a tag that is associated with namespace +ns+.  Raises an
       # ArgumentError if +ns+ has not been defined higher in the tree.
-      def [](ns)
-        if @parent != @doc
-          @ns = @parent.namespace_definitions.find { |x| x.prefix == ns.to_s }
-        end
-        return self if @ns
-
-        @parent.ancestors.each do |a|
-          next if a == doc
-
-          @ns = a.namespace_definitions.find { |x| x.prefix == ns.to_s }
-          return self if @ns
-        end
-
-        @ns = { pending: ns.to_s }
+      def [](ns_prefix)
+        @ns_prefix = ns_prefix.to_s
         self
       end
 
@@ -395,29 +383,33 @@ module Nokogiri
         if @context&.respond_to?(method)
           @context.send(method, *args, &block)
         else
-          node = @doc.create_element(method.to_s.sub(/[_!]$/, ""), *args) do |n|
-            # Set up the namespace
-            if @ns.is_a?(Nokogiri::XML::Namespace)
-              n.namespace = @ns
-              @ns = nil
-            end
-          end
-
-          if @ns.is_a?(Hash)
-            node.namespace = node.namespace_definitions.find { |x| x.prefix == @ns[:pending] }
-            if node.namespace.nil?
-              raise ArgumentError, "Namespace #{@ns[:pending]} has not been defined"
-            end
-
-            @ns = nil
-          end
-
+          node = @doc.create_element(method.to_s.sub(/[_!]$/, ""), *args)
+          bind_ns(node)
           insert(node, &block)
         end
       end
 
       private
 
+      def bind_ns(node)
+        return if @ns_prefix.nil?
+
+        ancestors = [node, parent, parent.ancestors].flatten
+        ancestors.each do |ancestor|
+          break if ancestor.nil? || ancestor == @doc
+
+          if (ns = ancestor.namespace_definitions.find { |x| x.prefix == @ns_prefix })
+            @ns_prefix = nil
+            node.namespace = ns
+            break
+          end
+        end
+
+        return if @ns_prefix.nil?
+
+        raise ArgumentError, "Namespace prefix #{@ns_prefix.inspect} has not been defined"
+      end
+
       ###
       # Insert +node+ as a child of the current Node
       def insert(node, &block)

data/lib/nokogiri/xml/document.rb

@@ -434,6 +434,7 @@ module Nokogiri
       undef_method :swap, :parent, :namespace, :default_namespace=
       undef_method :add_namespace_definition, :attributes
       undef_method :namespace_definitions, :line, :add_namespace
+      undef_method :line= if method_defined?(:line=)
 
       def add_child(node_or_tags)
         raise "A document may not have multiple root nodes." if (root && root.name != "nokogiri_text_wrapper") && !(node_or_tags.comment? || node_or_tags.processing_instruction?)

data/lib/nokogiri/xml/dtd.rb

@@ -8,7 +8,8 @@ module Nokogiri
       undef_method :content
       undef_method :namespace
       undef_method :namespace_definitions
-      undef_method :line if method_defined?(:line)
+      undef_method :line
+      undef_method :line= if method_defined?(:line=)
 
       def keys
         attributes.keys

data/lib/nokogiri/xml/element_decl.rb

@@ -5,7 +5,8 @@ module Nokogiri
     class ElementDecl < Nokogiri::XML::Node
       undef_method :namespace
       undef_method :namespace_definitions
-      undef_method :line if method_defined?(:line)
+      undef_method :line
+      undef_method :line= if method_defined?(:line=)
 
       private
 

data/lib/nokogiri/xml/entity_decl.rb

@@ -7,7 +7,8 @@ module Nokogiri
       undef_method :attributes
       undef_method :namespace
       undef_method :namespace_definitions
-      undef_method :line if method_defined?(:line)
+      undef_method :line
+      undef_method :line= if method_defined?(:line=)
 
       def self.new(name, doc, *args)
         doc.create_entity(name, *args)