Why doesn't my web-crawling method find all the links? - ruby

Question

I'm trying to create a simple web-crawler, so I wrote this:
(Method get_links take a parent link from which we will seek)
require 'nokogiri'
require 'open-uri'
def get_links(link)
link = "http://#{link}"
doc = Nokogiri::HTML(open(link))
links = doc.css('a')
hrefs = links.map {|link| link.attribute('href').to_s}.uniq.delete_if {|href| href.empty?}
array = hrefs.select {|i| i[0] == "/"}
host = URI.parse(link).host
links_list = array.map {|a| "#{host}#{a}"}
end
(Method search_links, takes an array from get_links method and search at this array)
def search_links(urls)
urls = get_links(link)
urls.uniq.each do |url|
begin
links = get_links(url)
compare = urls & links
urls << links - compare
urls.flatten!
rescue OpenURI::HTTPError
warn "Skipping invalid link #{url}"
end
end
return urls
end
This method finds most of links from the website, but not all.
What did I do wrong? Which algorithm I should use?

Answer 1

Some comments about your code:
def get_links(link)
link = "http://#{link}"
# You're assuming the protocol is always http.
# This isn't the only protocol on used on the web.
doc = Nokogiri::HTML(open(link))
links = doc.css('a')
hrefs = links.map {|link| link.attribute('href').to_s}.uniq.delete_if {|href| href.empty?}
# You can write these two lines more compact as
# hrefs = doc.xpath('//a/#href').map(&:to_s).uniq.delete_if(&:empty?)
array = hrefs.select {|i| i[0] == "/"}
# I guess you want to handle URLs that are relative to the host.
# However, URLs relative to the protocol (starting with '//')
# will also be selected by this condition.
host = URI.parse(link).host
links_list = array.map {|a| "#{host}#{a}"}
# The value assigned to links_list will implicitly be returned.
# (The assignment itself is futile, the right-hand-part alone would
# suffice.) Because this builds on `array` all absolute URLs will be
# missing from the return value.
end
Explanation for
hrefs = doc.xpath('//a/#href').map(&:to_s).uniq.delete_if(&:empty?)
.xpath('//a/#href') uses the attribute syntax of XPath to directly get to the href attributes of a elements
.map(&:to_s) is an abbreviated notation for .map { |item| item.to_s }
.delete_if(&:empty?) uses the same abbreviated notation
And comments about the second function:
def search_links(urls)
urls = get_links(link)
urls.uniq.each do |url|
begin
links = get_links(url)
compare = urls & links
urls << links - compare
urls.flatten!
# How about using a Set instead of an Array and
# thus have the collection provide uniqueness of
# its items, so that you don't have to?
rescue OpenURI::HTTPError
warn "Skipping invalid link #{url}"
end
end
return urls
# This function isn't recursive, it just calls `get_links` on two
# 'levels'. Thus you search only two levels deep and return findings
# from the first and second level combined. (Without the "zero'th"
# level - the URL passed into `search_links`. Unless off course if it
# also occured on the first or second level.)
#
# Is this what you intended?
end

Answer 2

You should probably be using mechanize:
require 'mechanize'
agent = Mechanize.new
page = agent.get url
links = page.search('a[href]').map{|a| page.uri.merge(a[:href]).to_s}
# if you want to remove links with a different host (hyperlinks?)
links.reject!{|l| URI.parse(l).host != page.uri.host}
Otherwise you'll have trouble converting relative urls to absolute properly.