xmllexer.rb

require 'nokogiri'

# XMLLexer - SAX with scanner like API
class XMLLexer
  def initialize(path)
    @f = Fiber.new do
      parser = Nokogiri::HTML::SAX::Parser.new(DocProxy.new)
      parser.parse_file(path)
      raise EOFError
    end
    @buf = []
  end

  # Returns an event
  def next
    return @buf.pop unless @buf.empty?
    @f.resume
  end

  # Returns an event without moving the pointer
  def peek
    return @buf[0] unless @buf.empty?
    @buf[0] = @f.resume
  end

  # Returns true if given condition is correct, otherwise raise error
  def verify(*arg)
    ary = self.next
    validate(ary, arg)
    nil
  end

  # Returns true and consumes the item if given condition is correct.
  # Otherwise returns false and keep the item in the buffer.
  def check(*arg)
    ary = self.peek
    begin
      validate(ary, arg)
      self.next
      true
    rescue RuntimeError
      false
    end
  end

  # get text from the XML's text node
  def gets
    ary = self.next
    raise "'characters' is expected, but '#{ary[0]}'" if ary[0] != :characters
    str = ary[1]
    while true
      ary = self.peek
      case ary[0]
      when :characters
        self.next
        str << ary[1]
        next
      when :start_element
        case ary[1]
        when 'br'
          self.next
          verify :end_element, 'br'
          next
        end
      end
      break
    end
    str
  end

  private
  def validate(ary, arg) # :nodoc:
    type = arg[0]
    raise "'#{type}' is expected, but '#{ary.inspect}'" if type != ary[0]
    case arg[0]
    when :characters
      case arg[1]
      when :blank
        unless /\A\s*\z/ =~ ary[1]
          raise "blank is expected but '#{ary[1].inspect}'"
        end
      when Regexp
        unless arg[1] =~ ary[1]
          raise "characters:'#{ary[1].inspect}' is expected to match '#{arg[1]}'"
        end
      when String
        if arg[1] != ary[1]
          raise "'#{arg[1]}' is expected but '#{ary[1].inspect}'"
        end
      when nil
      else
        raise 'invalid expectation'
      end
    when :start_element
      return if arg.size == 1
      raise "start_element: '#{arg[1]}' is expected, but '#{ary[1]}'" if arg[1] != ary[1]
      return if arg.size == 2
      case arg[2]
      when Hash
        attrs = ary[2]
        arg[2].each_pair do |k, v|
          case v
          when Regexp
            vs = ary[2].assoc(k)
            raise unless vs
            val = vs[1]
            raise unless val
            unless v =~ val
              raise "start_element:#{arg[1]} '#{k}'='#{val}' is not match /#{v}/"
            end
          when String
            vs = ary[2].assoc(k)
            raise unless vs
            val = vs[1]
            raise unless val
            if v != val
              raise "start_element:#{arg[1]} '#{k}'='#{val}' != '#{v}'"
            end
          end
        end
      end
    else
      result = ary[0, arg.size]
      raise "'#{arg}' is expected, but '#{result}'" if arg != result
    end
  end
end

class DocProxy < Nokogiri::XML::SAX::Document
  def cdata_block(str) Fiber.yield __method__, str end
  def characters(str) Fiber.yield __method__, str end
  def comment(str) Fiber.yield __method__, str end
  def end_document; Fiber.yield [__method__] end
  def end_element(name) Fiber.yield __method__, name end
  def end_element_namespace(name,prefix,uri) Fiber.yield __method__, name, prefix, uri end
  def error(str) Fiber.yield __method__, str end
  def start_document; Fiber.yield [__method__] end
  def start_element(name, attrs) Fiber.yield __method__, name, attrs end
  def start_element_namespace(name, attrs, prefix,url, ns) Fiber.yield __method__, name, attrs end
  def warning(str) Fiber.yield __method__, str end
  def xmldecl(version, encoding, standalone) Fiber.yield __method__, version, encoding, standalone end
end