class Sunbro::Page

Attributes

body[R]

The raw HTTP response body of the page

code[RW]

Integer response code of the page

depth[RW]

Depth of this page from the root of the crawl. This is not necessarily the shortest path; use PageStore#shortest_paths! to find that value.

error[R]

Exception object, if one was raised during HTTP#fetch_page

headers[R]

Headers of the HTTP response

redirect_from[RW]
redirect_to[R]

URL of the page this one redirected to, if any

referer[RW]

URL of the page that brought us to this page

response_time[RW]

Response time of the request for this page in milliseconds

url[RW]

The URL of the page

visited[RW]

Boolean indicating whether or not this page has been visited in PageStore#shortest_paths!

Public Class Methods

from_hash(hash) click to toggle source
# File lib/sunbro/page.rb, line 210
def self.from_hash(hash)
  page = self.new(URI(hash['url']))
  {'@headers'       => JSON.load(hash['headers']),
   '@body'          => hash['body'],
   '@code'          => hash['code'].to_i,
   '@error'         => hash['error'],
   '@visited'       => hash['visited'],
   '@referer'       => hash['referer'],
   '@redirect_to'   => (hash['redirect_to'].present?) ? URI(hash['redirect_to']) : nil,
   '@redirect_from' => (hash['redirect_from'].present?) ? URI(hash['redirect_from']) : nil,
   '@response_time' => hash['response_time'].to_i,
   '@fetched'       => hash['fetched']
  }.each do |var, value|
    page.instance_variable_set(var, value)
  end
  page
end
new(url, params = {}) click to toggle source

Create a new page

# File lib/sunbro/page.rb, line 32
def initialize(url, params = {})
  @url = url

  @code = params[:code]
  @headers = params[:headers] || {}
  @headers['content-type'] ||= ['']
  @aliases = Array(params[:aka]).compact
  @referer = params[:referer]
  @depth = params[:depth] || 0
  @redirect_to = to_absolute(params[:redirect_to])
  @response_time = params[:response_time]
  @error = params[:error]
  @fetched = !params[:code].nil?
  @force_format = params[:force_format]
  @body = params[:body]
  @redirect_from = params[:redirect_from]
end

Public Instance Methods

base() click to toggle source

Base URI from the HTML doc head element www.w3.org/TR/html4/struct/links.html#edef-BASE

# File lib/sunbro/page.rb, line 149
def base
  @base = if doc
    href = doc.search('//head/base/@href')
    URI(href.to_s) unless href.nil? rescue nil
  end unless @base

  return nil if @base && @base.to_s().empty?
  @base
end
content_type() click to toggle source

The content-type returned by the HTTP request for this page

# File lib/sunbro/page.rb, line 100
def content_type
  return headers['content-type'].first if headers['content-type'].first.present?
  headers['content_type']
end
cookies() click to toggle source

Array of cookies received with this page as WEBrick::Cookie objects.

# File lib/sunbro/page.rb, line 93
def cookies
  WEBrick::Cookie.parse_set_cookies(@headers['Set-Cookie']) rescue []
end
discard_doc!() click to toggle source

Delete the Nokogiri document and response body to conserve memory

# File lib/sunbro/page.rb, line 78
def discard_doc!
  @doc = @body = nil
end
doc() click to toggle source

Nokogiri document for the HTML body

# File lib/sunbro/page.rb, line 53
def doc
  @doc ||= begin
    if image?
      nil
    elsif should_parse_as?(:xml)
      Nokogiri::XML(@body, @url.to_s)
    elsif should_parse_as?(:html)
      Nokogiri::HTML(@body, @url.to_s)
    elsif @body
      Nokogiri.parse(@body, @url.to_s)
    end
  end
end
fetched?() click to toggle source

Was the page successfully fetched? true if the page was fetched with no error, false otherwise.

# File lib/sunbro/page.rb, line 86
def fetched?
  @fetched
end
html?() click to toggle source

Returns true if the page is a HTML document, returns false otherwise.

# File lib/sunbro/page.rb, line 117
def html?
  !!(content_type =~ %r{^(text/html|application/xhtml+xml)\b})
end
image?() click to toggle source

Returns true if the page is an image, returns false otherwise.

# File lib/sunbro/page.rb, line 109
def image?
  !!(content_type =~ %r{^(image/)\b})
end
in_domain?(uri) click to toggle source

Returns true if uri is in the same domain as the page, returns false otherwise

# File lib/sunbro/page.rb, line 182
def in_domain?(uri)
  uri.host == @url.host
end
marshal_dump() click to toggle source
# File lib/sunbro/page.rb, line 186
def marshal_dump
  [@url, @headers, @body, @links, @code, @visited, @depth, @referer, @redirect_to, @response_time, @fetched]
end
marshal_load(ary) click to toggle source
# File lib/sunbro/page.rb, line 190
def marshal_load(ary)
  @url, @headers, @body, @links, @code, @visited, @depth, @referer, @redirect_to, @response_time, @fetched = ary
end
not_found?() click to toggle source

Returns true if the page was not found (returned 404 code), returns false otherwise.

# File lib/sunbro/page.rb, line 141
def not_found?
  404 == @code
end
present?() click to toggle source
# File lib/sunbro/page.rb, line 71
def present?
  !error && code && body.present? && doc
end
redirect?() click to toggle source

Returns true if the page is a HTTP redirect, returns false otherwise.

# File lib/sunbro/page.rb, line 133
def redirect?
  (300..307).include?(@code)
end
to_absolute(link) click to toggle source

Converts relative URL link into an absolute URL based on the location of the page

# File lib/sunbro/page.rb, line 164
def to_absolute(link)
  return nil if link.nil?

  # remove anchor
  link = URI.encode(URI.decode(link.to_s.gsub(/#[a-zA-Z0-9_-]*$/,'')))

  relative = URI(link)
  absolute = base ? base.merge(relative) : @url.merge(relative)

  absolute.path = '/' if absolute.path.empty?

  return absolute
end
to_hash() click to toggle source
# File lib/sunbro/page.rb, line 194
def to_hash
  {
    'url'           => @url.to_s,
    'headers'       => headers.to_json,
    'body'          => @body,
    'code'          => @code,
    'error'         => (@error ? @error.to_s : nil),
    'visited'       => @visited,
    'referer'       => (@referer ? @referer.to_s : nil),
    'redirect_to'   => (@redirect_to ? @redirect_to.to_s : nil),
    'redirect_from' => (@redirect_from ? @redirect_from.to_s : nil),
    'response_time' => @response_time,
    'fetched'       => @fetched
  }.reject { |k, v| v.nil? }
end
valid?() click to toggle source
# File lib/sunbro/page.rb, line 67
def valid?
  (url != "about:blank") && !not_found? && present?
end
xml?() click to toggle source

Returns true if the page is a XML document, returns false otherwise.

# File lib/sunbro/page.rb, line 125
def xml?
  !!(content_type =~ %r{^(text/xml|application/xml)\b})
end

Private Instance Methods

cleanup_encoding(source) click to toggle source
# File lib/sunbro/page.rb, line 230
def cleanup_encoding(source)
  return source unless source && (html? || xml? || @force_format)
  text = source.dup
  text.encode!('UTF-16', 'UTF-8', {:invalid => :replace, :undef => :replace, :replace => '?'})
  text.encode('UTF-8', 'UTF-16')
end
should_parse_as?(format) click to toggle source
# File lib/sunbro/page.rb, line 237
def should_parse_as?(format)
  return false unless @body
  return @force_format == format if @force_format
  send("#{format}?")
end