Nokogiri end_element demo

gistfile1.rb

xml = %{
<doc>
  <key1>value1</key1>
  <key2>
Value with
line breaks
</key2>
</doc>
}

require 'nokogiri'
class BadParser < Nokogiri::XML::SAX::Document

  attr_reader :response

  def initialize
    reset
  end

  def reset
    @response = {}
  end

  def characters(string)
    @value ||= ''
    @value << string
  end

  def start_element(name, attrs = [])
    @value = nil
  end

  def end_element(name)
    # Does not change @value. This is bad.
    case name
    when 'key1', 'key2'
      @response[name] = @value
    end
  end
end

class GoodParser < Nokogiri::XML::SAX::Document

  attr_reader :response

  def initialize
    reset
  end

  def reset
    @response = {}
  end

  def characters(string)
    @value ||= ''
    @value << string
  end

  def start_element(name, attrs = [])
    @value = nil
  end

  def end_element(name)
    case name
    when 'key1', 'key2'
      @response[name] = @value
    end

    # Resets @value. This is good
    @value = nil
  end
end

bad = BadParser.new
good = GoodParser.new

Nokogiri::XML::SAX::Parser.new(bad).parse(xml)
Nokogiri::XML::SAX::Parser.new(good).parse(xml)

# Notice that the bad parser has extra whitespace at the end of "value1"
# because #characters is called with the whitespace between </key1> and <key2>.
# Since @value was not reset with end_element,
# @response['key1'] points to the same @value
bad.response

# => {"key1"=>"value1\n  ", "key2"=>"\nValue with\nline breaks\n\n"}

# A good solution is to reset @value from end_element
good.response

# => {"key1"=>"value1", "key2"=>"\nValue with\nline breaks\n"}