node.rb

# encoding: UTF-8
# frozen_string_literal: true
require "stringio"
require "nokogiri/xml/node/save_options"

module Nokogiri
  module XML
    ####
    # Nokogiri::XML::Node is your window to the fun filled world of dealing
    # with XML and HTML tags.  A Nokogiri::XML::Node may be treated similarly
    # to a hash with regard to attributes.  For example (from irb):
    #
    #   irb(main):004:0> node
    #   => <a href="#foo" id="link">link</a>
    #   irb(main):005:0> node['href']
    #   => "#foo"
    #   irb(main):006:0> node.keys
    #   => ["href", "id"]
    #   irb(main):007:0> node.values
    #   => ["#foo", "link"]
    #   irb(main):008:0> node['class'] = 'green'
    #   => "green"
    #   irb(main):009:0> node
    #   => <a href="#foo" id="link" class="green">link</a>
    #   irb(main):010:0>
    #
    # See Nokogiri::XML::Node#[] and Nokogiri::XML#[]= for more information.
    #
    # Nokogiri::XML::Node also has methods that let you move around your
    # tree.  For navigating your tree, see:
    #
    # * Nokogiri::XML::Node#parent
    # * Nokogiri::XML::Node#children
    # * Nokogiri::XML::Node#next
    # * Nokogiri::XML::Node#previous
    #
    #
    # When printing or otherwise emitting a document or a node (and
    # its subtree), there are a few methods you might want to use:
    #
    # * content, text, inner_text, to_str: emit plaintext
    #
    #   These methods will all emit the plaintext version of your
    #   document, meaning that entities will be replaced (e.g., "&lt;"
    #   will be replaced with "<"), meaning that any sanitizing will
    #   likely be un-done in the output.
    #
    # * to_s, to_xml, to_html, inner_html: emit well-formed markup
    #
    #   These methods will all emit properly-escaped markup, meaning
    #   that it's suitable for consumption by browsers, parsers, etc.
    #
    # You may search this node's subtree using Searchable#xpath and Searchable#css
    class Node
      include Nokogiri::XML::PP::Node
      include Nokogiri::XML::Searchable
      include Enumerable

      # Element node type, see Nokogiri::XML::Node#element?
      ELEMENT_NODE = 1
      # Attribute node type
      ATTRIBUTE_NODE = 2
      # Text node type, see Nokogiri::XML::Node#text?
      TEXT_NODE = 3
      # CDATA node type, see Nokogiri::XML::Node#cdata?
      CDATA_SECTION_NODE = 4
      # Entity reference node type
      ENTITY_REF_NODE = 5
      # Entity node type
      ENTITY_NODE = 6
      # PI node type
      PI_NODE = 7
      # Comment node type, see Nokogiri::XML::Node#comment?
      COMMENT_NODE = 8
      # Document node type, see Nokogiri::XML::Node#xml?
      DOCUMENT_NODE = 9
      # Document type node type
      DOCUMENT_TYPE_NODE = 10
      # Document fragment node type
      DOCUMENT_FRAG_NODE = 11
      # Notation node type
      NOTATION_NODE = 12
      # HTML document node type, see Nokogiri::XML::Node#html?
      HTML_DOCUMENT_NODE = 13
      # DTD node type
      DTD_NODE = 14
      # Element declaration type
      ELEMENT_DECL = 15
      # Attribute declaration type
      ATTRIBUTE_DECL = 16
      # Entity declaration type
      ENTITY_DECL = 17
      # Namespace declaration type
      NAMESPACE_DECL = 18
      # XInclude start type
      XINCLUDE_START = 19
      # XInclude end type
      XINCLUDE_END = 20
      # DOCB document node type
      DOCB_DOCUMENT_NODE = 21

      def initialize(name, document) # :nodoc:
        # ... Ya.  This is empty on purpose.
      end

      ###
      # Decorate this node with the decorators set up in this node's Document
      def decorate!
        document.decorate(self)
      end

      # @!group Searching via XPath or CSS Queries

      ###
      # Search this node's immediate children using CSS selector +selector+
      def >(selector)
        ns = document.root.namespaces
        xpath CSS.xpath_for(selector, :prefix => "./", :ns => ns).first
      end

      # @!endgroup

      # @!group Manipulating Document Structure

      ###
      # Add +node_or_tags+ as a child of this Node.
      # +node_or_tags+ can be a Nokogiri::XML::Node, a ::DocumentFragment, a ::NodeSet, or a string containing markup.
      #
      # Returns the reparented node (if +node_or_tags+ is a Node), or NodeSet (if +node_or_tags+ is a DocumentFragment, NodeSet, or string).
      #
      # Also see related method +<<+.
      def add_child(node_or_tags)
        node_or_tags = coerce(node_or_tags)
        if node_or_tags.is_a?(XML::NodeSet)
          node_or_tags.each { |n| add_child_node_and_reparent_attrs n }
        else
          add_child_node_and_reparent_attrs node_or_tags
        end
        node_or_tags
      end

      ###
      # Add +node_or_tags+ as the first child of this Node.
      # +node_or_tags+ can be a Nokogiri::XML::Node, a ::DocumentFragment, a ::NodeSet, or a string containing markup.
      #
      # Returns the reparented node (if +node_or_tags+ is a Node), or NodeSet (if +node_or_tags+ is a DocumentFragment, NodeSet, or string).
      #
      # Also see related method +add_child+.
      def prepend_child(node_or_tags)
        if first = children.first
          # Mimic the error add_child would raise.
          raise RuntimeError, "Document already has a root node" if document? && !(node_or_tags.comment? || node_or_tags.processing_instruction?)
          first.__send__(:add_sibling, :previous, node_or_tags)
        else
          add_child(node_or_tags)
        end
      end

      ###
      # Add html around this node
      #
      # Returns self
      def wrap(html)
        new_parent = document.parse(html).first
        add_next_sibling(new_parent)
        new_parent.add_child(self)
        self
      end

      ###
      # Add +node_or_tags+ as a child of this Node.
      # +node_or_tags+ can be a Nokogiri::XML::Node, a ::DocumentFragment, a ::NodeSet, or a string containing markup.
      #
      # Returns self, to support chaining of calls (e.g., root << child1 << child2)
      #
      # Also see related method +add_child+.
      def <<(node_or_tags)
        add_child node_or_tags
        self
      end

      ###
      # Insert +node_or_tags+ before this Node (as a sibling).
      # +node_or_tags+ can be a Nokogiri::XML::Node, a ::DocumentFragment, a ::NodeSet, or a string containing markup.
      #
      # Returns the reparented node (if +node_or_tags+ is a Node), or NodeSet (if +node_or_tags+ is a DocumentFragment, NodeSet, or string).
      #
      # Also see related method +before+.
      def add_previous_sibling(node_or_tags)
        raise ArgumentError.new("A document may not have multiple root nodes.") if (parent && parent.document?) && !(node_or_tags.comment? || node_or_tags.processing_instruction?)

        add_sibling :previous, node_or_tags
      end

      ###
      # Insert +node_or_tags+ after this Node (as a sibling).
      # +node_or_tags+ can be a Nokogiri::XML::Node, a ::DocumentFragment, a ::NodeSet, or a string containing markup.
      #
      # Returns the reparented node (if +node_or_tags+ is a Node), or NodeSet (if +node_or_tags+ is a DocumentFragment, NodeSet, or string).
      #
      # Also see related method +after+.
      def add_next_sibling(node_or_tags)
        raise ArgumentError.new("A document may not have multiple root nodes.") if (parent && parent.document?) && !(node_or_tags.comment? || node_or_tags.processing_instruction?)

        add_sibling :next, node_or_tags
      end

      ####
      # Insert +node_or_tags+ before this node (as a sibling).
      # +node_or_tags+ can be a Nokogiri::XML::Node, a ::DocumentFragment, a ::NodeSet, or a string containing markup.
      #
      # Returns self, to support chaining of calls.
      #
      # Also see related method +add_previous_sibling+.
      def before(node_or_tags)
        add_previous_sibling node_or_tags
        self
      end

      ####
      # Insert +node_or_tags+ after this node (as a sibling).
      # +node_or_tags+ can be a Nokogiri::XML::Node, a Nokogiri::XML::DocumentFragment, or a string containing markup.
      #
      # Returns self, to support chaining of calls.
      #
      # Also see related method +add_next_sibling+.
      def after(node_or_tags)
        add_next_sibling node_or_tags
        self
      end

      ####
      # Set the inner html for this Node to +node_or_tags+
      # +node_or_tags+ can be a Nokogiri::XML::Node, a Nokogiri::XML::DocumentFragment, or a string containing markup.
      #
      # Returns self.
      #
      # Also see related method +children=+
      def inner_html=(node_or_tags)
        self.children = node_or_tags
        self
      end

      ####
      # Set the inner html for this Node +node_or_tags+
      # +node_or_tags+ can be a Nokogiri::XML::Node, a Nokogiri::XML::DocumentFragment, or a string containing markup.
      #
      # Returns the reparented node (if +node_or_tags+ is a Node), or NodeSet (if +node_or_tags+ is a DocumentFragment, NodeSet, or string).
      #
      # Also see related method +inner_html=+
      def children=(node_or_tags)
        node_or_tags = coerce(node_or_tags)
        children.unlink
        if node_or_tags.is_a?(XML::NodeSet)
          node_or_tags.each { |n| add_child_node_and_reparent_attrs n }
        else
          add_child_node_and_reparent_attrs node_or_tags
        end
        node_or_tags
      end

      ####
      # Replace this Node with +node_or_tags+.
      # +node_or_tags+ can be a Nokogiri::XML::Node, a ::DocumentFragment, a ::NodeSet, or a string containing markup.
      #
      # Returns the reparented node (if +node_or_tags+ is a Node), or NodeSet (if +node_or_tags+ is a DocumentFragment, NodeSet, or string).
      #
      # Also see related method +swap+.
      def replace(node_or_tags)
        raise("Cannot replace a node with no parent") unless parent

        # We cannot replace a text node directly, otherwise libxml will return
        # an internal error at parser.c:13031, I don't know exactly why
        # libxml is trying to find a parent node that is an element or document
        # so I can't tell if this is bug in libxml or not. issue #775.
        if text?
          replacee = Nokogiri::XML::Node.new "dummy", document
          add_previous_sibling_node replacee
          unlink
          return replacee.replace node_or_tags
        end

        node_or_tags = parent.coerce(node_or_tags)

        if node_or_tags.is_a?(XML::NodeSet)
          node_or_tags.each { |n| add_previous_sibling n }
          unlink
        else
          replace_node node_or_tags
        end
        node_or_tags
      end

      ####
      # Swap this Node for +node_or_tags+
      # +node_or_tags+ can be a Nokogiri::XML::Node, a ::DocumentFragment, a ::NodeSet, or a string containing markup.
      #
      # Returns self, to support chaining of calls.
      #
      # Also see related method +replace+.
      def swap(node_or_tags)
        replace node_or_tags
        self
      end

      ####
      # Set the Node's content to a Text node containing +string+. The string gets XML escaped, not interpreted as markup.
      def content=(string)
        self.native_content = encode_special_chars(string.to_s)
      end

      ###
      # Set the parent Node for this Node
      def parent=(parent_node)
        parent_node.add_child(self)
        parent_node
      end

      ###
      # Adds a default namespace supplied as a string +url+ href, to self.
      # The consequence is as an xmlns attribute with supplied argument were
      # present in parsed XML.  A default namespace set with this method will
      # now show up in #attributes, but when this node is serialized to XML an
      # "xmlns" attribute will appear. See also #namespace and #namespace=
      def default_namespace=(url)
        add_namespace_definition(nil, url)
      end

      ###
      # Set the default namespace on this node (as would be defined with an
      # "xmlns=" attribute in XML source), as a Namespace object +ns+. Note that
      # a Namespace added this way will NOT be serialized as an xmlns attribute
      # for this node. You probably want #default_namespace= instead, or perhaps
      # #add_namespace_definition with a nil prefix argument.
      def namespace=(ns)
        return set_namespace(ns) unless ns

        unless Nokogiri::XML::Namespace === ns
          raise TypeError, "#{ns.class} can't be coerced into Nokogiri::XML::Namespace"
        end
        if ns.document != document
          raise ArgumentError, "namespace must be declared on the same document"
        end

        set_namespace ns
      end

      ###
      # Do xinclude substitution on the subtree below node. If given a block, a
      # Nokogiri::XML::ParseOptions object initialized from +options+, will be
      # passed to it, allowing more convenient modification of the parser options.
      def do_xinclude(options = XML::ParseOptions::DEFAULT_XML)
        options = Nokogiri::XML::ParseOptions.new(options) if Integer === options

        # give options to user
        yield options if block_given?

        # call c extension
        process_xincludes(options.to_i)
      end

      alias :next :next_sibling
      alias :previous :previous_sibling
      alias :next= :add_next_sibling
      alias :previous= :add_previous_sibling
      alias :remove :unlink
      alias :name= :node_name=
      alias :add_namespace :add_namespace_definition

      # @!endgroup

      alias :text :content
      alias :inner_text :content
      alias :name :node_name
      alias :type :node_type
      alias :to_str :text
      alias :clone :dup
      alias :elements :element_children

      # @!group Working With Node Attributes

      ###
      # Get the attribute value for the attribute +name+
      def [](name)
        get(name.to_s)
      end

      ###
      # Set the attribute value for the attribute +name+ to +value+
      def []=(name, value)
        set name.to_s, value.to_s
      end

      ####
      # Returns a hash containing the node's attributes.  The key is
      # the attribute name without any namespace, the value is a Nokogiri::XML::Attr
      # representing the attribute.
      # If you need to distinguish attributes with the same name, with different namespaces
      # use #attribute_nodes instead.
      def attributes
        attribute_nodes.each_with_object({}) do |node, hash|
          hash[node.node_name] = node
        end
      end

      ###
      # Get the attribute values for this Node.
      def values
        attribute_nodes.map(&:value)
      end

      ###
      # Does this Node's attributes include <value>
      def value?(value)
        values.include? value
      end

      ###
      # Get the attribute names for this Node.
      def keys
        attribute_nodes.map(&:node_name)
      end

      ###
      # Iterate over each attribute name and value pair for this Node.
      def each
        attribute_nodes.each { |node|
          yield [node.node_name, node.value]
        }
      end

      ###
      # Remove the attribute named +name+
      def remove_attribute(name)
        attr = attributes[name].remove if key? name
        clear_xpath_context if Nokogiri.jruby?
        attr
      end

      # Get the CSS class names of a Node.
      #
      # This is a convenience function and is equivalent to:
      #   node.kwattr_values("class")
      #
      # @see #kwattr_values
      # @see #add_class
      # @see #append_class
      # @see #remove_class
      #
      # @return [Array<String>]
      #
      #   The CSS classes present in the Node's +class+ attribute. If
      #   the attribute is empty or non-existent, the return value is
      #   an empty array.
      #
      # @example
      #   node         # => <div class="section title header"></div>
      #   node.classes # => ["section", "title", "header"]
      #
      def classes
        kwattr_values("class")
      end

      # Ensure HTML CSS classes are present on a +Node+. Any CSS
      # classes in +names+ that already exist in the +Node+'s +class+
      # attribute are _not_ added. Note that any existing duplicates
      # in the +class+ attribute are not removed. Compare with
      # {#append_class}.
      #
      # This is a convenience function and is equivalent to:
      #   node.kwattr_add("class", names)
      #
      # @see #kwattr_add
      # @see #classes
      # @see #append_class
      # @see #remove_class
      #
      # @param names [String, Array<String>]
      #
      #   CSS class names to be added to the Node's +class+
      #   attribute. May be a string containing whitespace-delimited
      #   names, or an Array of String names. Any class names already
      #   present will not be added. Any class names not present will
      #   be added. If no +class+ attribute exists, one is created.
      #
      # @return [Node] Returns +self+ for ease of chaining method calls.
      #
      # @example Ensure that a +Node+ has CSS class "section"
      #   node                      # => <div></div>
      #   node.add_class("section") # => <div class="section"></div>
      #   node.add_class("section") # => <div class="section"></div> # duplicate not added
      #
      # @example Ensure that a +Node+ has CSS classes "section" and "header", via a String argument.
      #   node                             # => <div class="section section"></div>
      #   node.add_class("section header") # => <div class="section section header"></div>
      #   # Note that the CSS class "section" is not added because it is already present.
      #   # Note also that the pre-existing duplicate CSS class "section" is not removed.
      #
      # @example Ensure that a +Node+ has CSS classes "section" and "header", via an Array argument.
      #   node                                  # => <div></div>
      #   node.add_class(["section", "header"]) # => <div class="section header"></div>
      #
      def add_class(names)
        kwattr_add("class", names)
      end

      # Add HTML CSS classes to a +Node+, regardless of
      # duplication. Compare with {#add_class}.
      #
      # This is a convenience function and is equivalent to:
      #   node.kwattr_append("class", names)
      #
      # @see #kwattr_append
      # @see #classes
      # @see #add_class
      # @see #remove_class
      #
      # @param names [String, Array<String>]
      #
      #   CSS class names to be appended to the Node's +class+
      #   attribute. May be a string containing whitespace-delimited
      #   names, or an Array of String names. All class names passed
      #   in will be appended to the +class+ attribute even if they
      #   are already present in the attribute value. If no +class+
      #   attribute exists, one is created.
      #
      # @return [Node] Returns +self+ for ease of chaining method calls.
      #
      # @example Append "section" to a +Node+'s CSS +class+ attriubute
      #   node                         # => <div></div>
      #   node.append_class("section") # => <div class="section"></div>
      #   node.append_class("section") # => <div class="section section"></div> # duplicate added!
      #
      # @example Append "section" and "header" to a +Node+'s CSS +class+ attribute, via a String argument.
      #   node                                # => <div class="section section"></div>
      #   node.append_class("section header") # => <div class="section section section header"></div>
      #   # Note that the CSS class "section" is appended even though it is already present.
      #
      # @example Append "section" and "header" to a +Node+'s CSS +class+ attribute, via an Array argument.
      #   node                                     # => <div></div>
      #   node.append_class(["section", "header"]) # => <div class="section header"></div>
      #   node.append_class(["section", "header"]) # => <div class="section header section header"></div>
      #
      def append_class(names)
        kwattr_append("class", names)
      end

      # Remove HTML CSS classes from a +Node+. Any CSS classes in +names+ that
      # exist in the +Node+'s +class+ attribute are removed, including any
      # multiple entries.
      #
      # If no CSS classes remain after this operation, or if +names+ is
      # +nil+, the +class+ attribute is deleted from the node.
      #
      # This is a convenience function and is equivalent to:
      #   node.kwattr_remove("class", names)
      #
      # @see #kwattr_remove
      # @see #classes
      # @see #add_class
      # @see #append_class
      #
      # @param names [String, Array<String>]
      #
      #   CSS class names to be removed from the Node's +class+ attribute. May
      #   be a string containing whitespace-delimited names, or an Array of
      #   String names. Any class names already present will be removed. If no
      #   CSS classes remain, the +class+ attribute is deleted.
      #
      # @return [Node] Returns +self+ for ease of chaining method calls.
      #
      # @example
      #   node                         # => <div class="section header"></div>
      #   node.remove_class("section") # => <div class="header"></div>
      #   node.remove_class("header")  # => <div></div> # attribute is deleted when empty
      #
      def remove_class(names = nil)
        kwattr_remove("class", names)
      end

      # Retrieve values from a keyword attribute of a Node.
      #
      # A "keyword attribute" is a node attribute that contains a set
      # of space-delimited values. Perhaps the most familiar example
      # of this is the HTML +class+ attribute used to contain CSS
      # classes. But other keyword attributes exist, for instance
      # [`rel`](https://developer.mozilla.org/en-US/docs/Web/HTML/Attributes/rel).
      #
      # @see #classes
      # @see #kwattr_add
      # @see #kwattr_append
      # @see #kwattr_remove
      #
      # @param attribute_name [String] The name of the keyword attribute to be inspected.
      #
      # @return [Array<String>]
      #
      #   The values present in the Node's +attribute_name+
      #   attribute. If the attribute is empty or non-existent, the
      #   return value is an empty array.
      #
      # @example
      #   node                      # => <a rel="nofollow noopener external">link</a>
      #   node.kwattr_values("rel") # => ["nofollow", "noopener", "external"]
      #
      # @since v1.11.0
      #
      def kwattr_values(attribute_name)
        keywordify(get_attribute(attribute_name) || [])
      end

      # Ensure that values are present in a keyword attribute.
      #
      # Any values in +keywords+ that already exist in the +Node+'s
      # attribute values are _not_ added. Note that any existing
      # duplicates in the attribute values are not removed. Compare
      # with {#kwattr_append}.
      #
      # A "keyword attribute" is a node attribute that contains a set
      # of space-delimited values. Perhaps the most familiar example
      # of this is the HTML +class+ attribute used to contain CSS
      # classes. But other keyword attributes exist, for instance
      # [`rel`](https://developer.mozilla.org/en-US/docs/Web/HTML/Attributes/rel).
      #
      # @see #add_class
      # @see #kwattr_values
      # @see #kwattr_append
      # @see #kwattr_remove
      #
      # @param attribute_name [String] The name of the keyword attribute to be modified.
      #
      # @param keywords [String, Array<String>]
      #
      #   Keywords to be added to the attribute named
      #   +attribute_name+. May be a string containing
      #   whitespace-delimited values, or an Array of String
      #   values. Any values already present will not be added. Any
      #   values not present will be added. If the named attribute
      #   does not exist, it is created.
      #
      # @return [Node] Returns +self+ for ease of chaining method calls.
      #
      # @example Ensure that a +Node+ has "nofollow" in its +rel+ attribute.
      #   node                               # => <a></a>
      #   node.kwattr_add("rel", "nofollow") # => <a rel="nofollow"></a>
      #   node.kwattr_add("rel", "nofollow") # => <a rel="nofollow"></a> # duplicate not added
      #
      # @example Ensure that a +Node+ has "nofollow" and "noreferrer" in its +rel+ attribute, via a String argument.
      #   node                                          # => <a rel="nofollow nofollow"></a>
      #   node.kwattr_add("rel", "nofollow noreferrer") # => <a rel="nofollow nofollow noreferrer"></a>
      #   # Note that "nofollow" is not added because it is already present.
      #   # Note also that the pre-existing duplicate "nofollow" is not removed.
      #
      # @example Ensure that a +Node+ has "nofollow" and "noreferrer" in its +rel+ attribute, via an Array argument.
      #   node                                               # => <a></a>
      #   node.kwattr_add("rel", ["nofollow", "noreferrer"]) # => <a rel="nofollow noreferrer"></a>
      #
      # @since v1.11.0
      #
      def kwattr_add(attribute_name, keywords)
        keywords = keywordify(keywords)
        current_kws = kwattr_values(attribute_name)
        new_kws = (current_kws + (keywords - current_kws)).join(" ")
        set_attribute(attribute_name, new_kws)
        self
      end

      # Add keywords to a Node's keyword attribute, regardless of
      # duplication. Compare with {#kwattr_add}.
      #
      # A "keyword attribute" is a node attribute that contains a set
      # of space-delimited values. Perhaps the most familiar example
      # of this is the HTML +class+ attribute used to contain CSS
      # classes. But other keyword attributes exist, for instance
      # [`rel`](https://developer.mozilla.org/en-US/docs/Web/HTML/Attributes/rel).
      #
      # @see #append_class
      # @see #kwattr_values
      # @see #kwattr_add
      # @see #kwattr_remove
      #
      # @param attribute_name [String] The name of the keyword attribute to be modified.
      #
      # @param keywords [String, Array<String>]
      #
      #   Keywords to be added to the attribute named
      #   +attribute_name+. May be a string containing
      #   whitespace-delimited values, or an Array of String
      #   values. All values passed in will be appended to the named
      #   attribute even if they are already present in the
      #   attribute. If the named attribute does not exist, it is
      #   created.
      #
      # @return [Node] Returns +self+ for ease of chaining method calls.
      #
      # @example Append "nofollow" to the +rel+ attribute.
      #   node                                  # => <a></a>
      #   node.kwattr_append("rel", "nofollow") # => <a rel="nofollow"></a>
      #   node.kwattr_append("rel", "nofollow") # => <a rel="nofollow nofollow"></a> # duplicate added!
      #
      # @example Append "nofollow" and "noreferrer" to the +rel+ attribute, via a String argument.
      #   node                                             # => <a rel="nofollow"></a>
      #   node.kwattr_append("rel", "nofollow noreferrer") # => <a rel="nofollow nofollow noreferrer"></a>
      #   # Note that "nofollow" is appended even though it is already present.
      #
      # @example Append "nofollow" and "noreferrer" to the +rel+ attribute, via an Array argument.
      #   node                                                  # => <a></a>
      #   node.kwattr_append("rel", ["nofollow", "noreferrer"]) # => <a rel="nofollow noreferrer"></a>
      #
      # @since v1.11.0
      #
      def kwattr_append(attribute_name, keywords)
        keywords = keywordify(keywords)
        current_kws = kwattr_values(attribute_name)
        new_kws = (current_kws + keywords).join(" ")
        set_attribute(attribute_name, new_kws)
        self
      end

      # Remove keywords from a keyword attribute. Any matching
      # keywords that exist in the named attribute are removed,
      # including any multiple entries.
      #
      # If no keywords remain after this operation, or if +keywords+
      # is +nil+, the attribute is deleted from the node.
      #
      # A "keyword attribute" is a node attribute that contains a set
      # of space-delimited values. Perhaps the most familiar example
      # of this is the HTML +class+ attribute used to contain CSS
      # classes. But other keyword attributes exist, for instance
      # [`rel`](https://developer.mozilla.org/en-US/docs/Web/HTML/Attributes/rel).
      #
      # @see #remove_class
      # @see #kwattr_values
      # @see #kwattr_add
      # @see #kwattr_append
      #
      # @param attribute_name [String] The name of the keyword attribute to be modified.
      #
      # @param keywords [String, Array<String>]
      #
      #   Keywords to be removed from the attribute named
      #   +attribute_name+. May be a string containing
      #   whitespace-delimited values, or an Array of String
      #   values. Any keywords present in the named attribute will be
      #   removed. If no keywords remain, or if +keywords+ is nil, the
      #   attribute is deleted.
      #
      # @return [Node] Returns +self+ for ease of chaining method calls.
      #
      # @example
      #   node                                    # => <a rel="nofollow noreferrer">link</a>
      #   node.kwattr_remove("rel", "nofollow")   # => <a rel="noreferrer">link</a>
      #   node.kwattr_remove("rel", "noreferrer") # => <a>link</a> # attribute is deleted when empty
      #
      # @since v1.11.0
      #
      def kwattr_remove(attribute_name, keywords)
        if keywords.nil?
          remove_attribute(attribute_name)
          return self
        end

        keywords = keywordify(keywords)
        current_kws = kwattr_values(attribute_name)
        new_kws = current_kws - keywords
        if new_kws.empty?
          remove_attribute(attribute_name)
        else
          set_attribute(attribute_name, new_kws.join(" "))
        end
        self
      end

      alias :delete :remove_attribute
      alias :get_attribute :[]
      alias :attr :[]
      alias :set_attribute :[]=
      alias :has_attribute? :key?

      # @!endgroup

      ###
      # Returns true if this Node matches +selector+
      def matches?(selector)
        ancestors.last.search(selector).include?(self)
      end

      ###
      # Create a DocumentFragment containing +tags+ that is relative to _this_
      # context node.
      def fragment(tags)
        type = document.html? ? Nokogiri::HTML : Nokogiri::XML
        type::DocumentFragment.new(document, tags, self)
      end

      ###
      # Parse +string_or_io+ as a document fragment within the context of
      # *this* node.  Returns a XML::NodeSet containing the nodes parsed from
      # +string_or_io+.
      def parse(string_or_io, options = nil)
        ##
        # When the current node is unparented and not an element node, use the
        # document as the parsing context instead. Otherwise, the in-context
        # parser cannot find an element or a document node.
        # Document Fragments are also not usable by the in-context parser.
        if !element? && !document? && (!parent || parent.fragment?)
          return document.parse(string_or_io, options)
        end

        options ||= (document.html? ? ParseOptions::DEFAULT_HTML : ParseOptions::DEFAULT_XML)
        if Integer === options
          options = Nokogiri::XML::ParseOptions.new(options)
        end
        # Give the options to the user
        yield options if block_given?

        contents = string_or_io.respond_to?(:read) ?
          string_or_io.read :
          string_or_io

        return Nokogiri::XML::NodeSet.new(document) if contents.empty?

        # libxml2 does not obey the `recover` option after encountering errors during `in_context`
        # parsing, and so this horrible hack is here to try to emulate recovery behavior.
        #
        # Unfortunately, this means we're no longer parsing "in context" and so namespaces that
        # would have been inherited from the context node won't be handled correctly. This hack was
        # written in 2010, and I regret it, because it's silently degrading functionality in a way
        # that's not easily prevented (or even detected).
        #
        # I think preferable behavior would be to either:
        #
        # a. add an error noting that we "fell back" and pointing the user to turning off the `recover` option
        # b. don't recover, but raise a sensible exception
        #
        # For context and background: https://github.com/sparklemotion/nokogiri/issues/313
        # FIXME bug report: https://github.com/sparklemotion/nokogiri/issues/2092
        error_count = document.errors.length
        node_set = in_context(contents, options.to_i)
        if (node_set.empty? && (document.errors.length > error_count))
          if options.recover?
            fragment = Nokogiri::HTML::DocumentFragment.parse contents
            node_set = fragment.children
          else
            raise document.errors[error_count]
          end
        end
        node_set
      end

      ###
      # Returns a Hash of +{prefix => value}+ for all namespaces on this
      # node and its ancestors.
      #
      # This method returns the same namespaces as #namespace_scopes.
      #
      # Returns namespaces in scope for self -- those defined on self
      # element directly or any ancestor node -- as a Hash of
      # attribute-name/value pairs. Note that the keys in this hash
      # XML attributes that would be used to define this namespace,
      # such as "xmlns:prefix", not just the prefix. Default namespace
      # set on self will be included with key "xmlns". However,
      # default namespaces set on ancestor will NOT be, even if self
      # has no explicit default namespace.
      def namespaces
        namespace_scopes.each_with_object({}) do |ns, hash|
          prefix = ns.prefix
          key = prefix ? "xmlns:#{prefix}" : "xmlns"
          hash[key] = ns.href
        end
      end

      # Returns true if this is a Comment
      def comment?
        type == COMMENT_NODE
      end

      # Returns true if this is a CDATA
      def cdata?
        type == CDATA_SECTION_NODE
      end

      # Returns true if this is an XML::Document node
      def xml?
        type == DOCUMENT_NODE
      end

      # Returns true if this is an HTML::Document node
      def html?
        type == HTML_DOCUMENT_NODE
      end

      # Returns true if this is a Document
      def document?
        is_a? XML::Document
      end

      # Returns true if this is a ProcessingInstruction node
      def processing_instruction?
        type == PI_NODE
      end

      # Returns true if this is a Text node
      def text?
        type == TEXT_NODE
      end

      # Returns true if this is a DocumentFragment
      def fragment?
        type == DOCUMENT_FRAG_NODE
      end

      ###
      # Fetch the Nokogiri::HTML::ElementDescription for this node.  Returns
      # nil on XML documents and on unknown tags.
      def description
        return nil if document.xml?
        Nokogiri::HTML::ElementDescription[name]
      end

      ###
      # Is this a read only node?
      def read_only?
        # According to gdome2, these are read-only node types
        [NOTATION_NODE, ENTITY_NODE, ENTITY_DECL].include?(type)
      end

      # Returns true if this is an Element node
      def element?
        type == ELEMENT_NODE
      end

      alias :elem? :element?

      ###
      # Turn this node in to a string.  If the document is HTML, this method
      # returns html.  If the document is XML, this method returns XML.
      def to_s
        document.xml? ? to_xml : to_html
      end

      # Get the inner_html for this node's Node#children
      def inner_html(*args)
        children.map { |x| x.to_html(*args) }.join
      end

      # Get the path to this node as a CSS expression
      def css_path
        path.split(/\//).map { |part|
          part.length == 0 ? nil : part.gsub(/\[(\d+)\]/, ':nth-of-type(\1)')
        }.compact.join(" > ")
      end

      ###
      # Get a list of ancestor Node for this Node.  If +selector+ is given,
      # the ancestors must match +selector+
      def ancestors(selector = nil)
        return NodeSet.new(document) unless respond_to?(:parent)
        return NodeSet.new(document) unless parent

        parents = [parent]

        while parents.last.respond_to?(:parent)
          break unless ctx_parent = parents.last.parent
          parents << ctx_parent
        end

        return NodeSet.new(document, parents) unless selector

        root = parents.last
        search_results = root.search(selector)

        NodeSet.new(document, parents.find_all { |parent|
          search_results.include?(parent)
        })
      end

      ####
      # Yields self and all children to +block+ recursively.
      def traverse(&block)
        children.each { |j| j.traverse(&block) }
        block.call(self)
      end

      ###
      # Accept a visitor.  This method calls "visit" on +visitor+ with self.
      def accept(visitor)
        visitor.visit(self)
      end

      ###
      # Test to see if this Node is equal to +other+
      def ==(other)
        return false unless other
        return false unless other.respond_to?(:pointer_id)
        pointer_id == other.pointer_id
      end

      ###
      # Compare two Node objects with respect to their Document.  Nodes from
      # different documents cannot be compared.
      def <=>(other)
        return nil unless other.is_a?(Nokogiri::XML::Node)
        return nil unless document == other.document
        compare other
      end

      # @!group Serialization and Generating Output

      ###
      # Serialize Node using +options+.  Save options can also be set using a
      # block. See SaveOptions.
      #
      # These two statements are equivalent:
      #
      #  node.serialize(:encoding => 'UTF-8', :save_with => FORMAT | AS_XML)
      #
      # or
      #
      #   node.serialize(:encoding => 'UTF-8') do |config|
      #     config.format.as_xml
      #   end
      #
      def serialize(*args, &block)
        options = args.first.is_a?(Hash) ? args.shift : {
          :encoding => args[0],
          :save_with => args[1],
        }

        encoding = options[:encoding] || document.encoding
        options[:encoding] = encoding

        outstring = String.new
        outstring.force_encoding(Encoding.find(encoding || "utf-8"))
        io = StringIO.new(outstring)
        write_to io, options, &block
        io.string
      end

      ###
      # Serialize this Node to HTML
      #
      #   doc.to_html
      #
      # See Node#write_to for a list of +options+.  For formatted output,
      # use Node#to_xhtml instead.
      def to_html(options = {})
        to_format SaveOptions::DEFAULT_HTML, options
      end

      ###
      # Serialize this Node to XML using +options+
      #
      #   doc.to_xml(:indent => 5, :encoding => 'UTF-8')
      #
      # See Node#write_to for a list of +options+
      def to_xml(options = {})
        options[:save_with] ||= SaveOptions::DEFAULT_XML
        serialize(options)
      end

      ###
      # Serialize this Node to XHTML using +options+
      #
      #   doc.to_xhtml(:indent => 5, :encoding => 'UTF-8')
      #
      # See Node#write_to for a list of +options+
      def to_xhtml(options = {})
        to_format SaveOptions::DEFAULT_XHTML, options
      end

      ###
      # Write Node to +io+ with +options+. +options+ modify the output of
      # this method.  Valid options are:
      #
      # * +:encoding+ for changing the encoding
      # * +:indent_text+ the indentation text, defaults to one space
      # * +:indent+ the number of +:indent_text+ to use, defaults to 2
      # * +:save_with+ a combination of SaveOptions constants.
      #
      # To save with UTF-8 indented twice:
      #
      #   node.write_to(io, :encoding => 'UTF-8', :indent => 2)
      #
      # To save indented with two dashes:
      #
      #   node.write_to(io, :indent_text => '-', :indent => 2)
      #
      def write_to(io, *options)
        options = options.first.is_a?(Hash) ? options.shift : {}
        encoding = options[:encoding] || options[0]
        if Nokogiri.jruby?
          save_options = options[:save_with] || options[1]
          indent_times = options[:indent] || 0
        else
          save_options = options[:save_with] || options[1] || SaveOptions::FORMAT
          indent_times = options[:indent] || 2
        end
        indent_text = options[:indent_text] || " "

        # Any string times 0 returns an empty string. Therefore, use the same
        # string instead of generating a new empty string for every node with
        # zero indentation.
        indentation = indent_times.zero? ? "" : (indent_text * indent_times)

        config = SaveOptions.new(save_options.to_i)
        yield config if block_given?

        native_write_to(io, encoding, indentation, config.options)
      end

      ###
      # Write Node as HTML to +io+ with +options+
      #
      # See Node#write_to for a list of +options+
      def write_html_to(io, options = {})
        write_format_to SaveOptions::DEFAULT_HTML, io, options
      end

      ###
      # Write Node as XHTML to +io+ with +options+
      #
      # See Node#write_to for a list of +options+
      def write_xhtml_to(io, options = {})
        write_format_to SaveOptions::DEFAULT_XHTML, io, options
      end

      ###
      # Write Node as XML to +io+ with +options+
      #
      #   doc.write_xml_to io, :encoding => 'UTF-8'
      #
      # See Node#write_to for a list of options
      def write_xml_to(io, options = {})
        options[:save_with] ||= SaveOptions::DEFAULT_XML
        write_to io, options
      end

      def canonicalize(mode = XML::XML_C14N_1_0, inclusive_namespaces = nil, with_comments = false)
        c14n_root = self
        document.canonicalize(mode, inclusive_namespaces, with_comments) do |node, parent|
          tn = node.is_a?(XML::Node) ? node : parent
          tn == c14n_root || tn.ancestors.include?(c14n_root)
        end
      end

      # @!endgroup

      protected

      def coerce(data)
        case data
        when XML::NodeSet
          return data
        when XML::DocumentFragment
          return data.children
        when String
          return fragment(data).children
        when Document, XML::Attr
          # unacceptable
        when XML::Node
          return data
        end

        raise ArgumentError, <<-EOERR
Requires a Node, NodeSet or String argument, and cannot accept a #{data.class}.
(You probably want to select a node from the Document with at() or search(), or create a new Node via Node.new().)
        EOERR
      end

      private

      def keywordify(keywords)
        case keywords
        when Enumerable
          return keywords
        when String
          return keywords.scan(/\S+/)
        else
          raise ArgumentError.new("Keyword attributes must be passed as either a String or an Enumerable, but received #{keywords.class}")
        end
      end

      def add_sibling(next_or_previous, node_or_tags)
        raise("Cannot add sibling to a node with no parent") unless parent

        impl = (next_or_previous == :next) ? :add_next_sibling_node : :add_previous_sibling_node
        iter = (next_or_previous == :next) ? :reverse_each : :each

        node_or_tags = parent.coerce(node_or_tags)
        if node_or_tags.is_a?(XML::NodeSet)
          if text?
            pivot = Nokogiri::XML::Node.new "dummy", document
            send impl, pivot
          else
            pivot = self
          end
          node_or_tags.send(iter) { |n| pivot.send impl, n }
          pivot.unlink if text?
        else
          send impl, node_or_tags
        end
        node_or_tags
      end

      USING_LIBXML_WITH_BROKEN_SERIALIZATION = Nokogiri.uses_libxml?("~> 2.6.0").freeze
      private_constant :USING_LIBXML_WITH_BROKEN_SERIALIZATION

      def to_format(save_option, options)
        return dump_html if USING_LIBXML_WITH_BROKEN_SERIALIZATION

        options[:save_with] = save_option unless options[:save_with]
        serialize(options)
      end

      def write_format_to(save_option, io, options)
        return (io << dump_html) if USING_LIBXML_WITH_BROKEN_SERIALIZATION

        options[:save_with] ||= save_option
        write_to io, options
      end

      def inspect_attributes
        [:name, :namespace, :attribute_nodes, :children]
      end

      # @private
      IMPLIED_XPATH_CONTEXTS = [".//".freeze].freeze

      def add_child_node_and_reparent_attrs(node)
        add_child_node node
        node.attribute_nodes.find_all { |a| a.name =~ /:/ }.each do |attr_node|
          attr_node.remove
          node[attr_node.name] = attr_node.value
        end
      end
    end
  end
end