url_scraper.rb 2.4 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100
  1. module Docs
  2. class UrlScraper < Scraper
  3. class << self
  4. attr_accessor :params
  5. attr_accessor :headers
  6. def inherited(subclass)
  7. super
  8. subclass.params = params.deep_dup
  9. subclass.headers = headers.deep_dup
  10. end
  11. end
  12. self.params = {}
  13. self.headers = { 'User-Agent' => 'devdocs.io' }
  14. private
  15. def request_one(url)
  16. Request.run url, request_options
  17. end
  18. def request_all(urls, &block)
  19. Requester.run urls, request_options: request_options, &block
  20. end
  21. def request_options
  22. { params: self.class.params, headers: self.class.headers }
  23. end
  24. def process_response?(response)
  25. if response.error?
  26. raise "Error status code (#{response.code}): #{response.url}"
  27. end
  28. response.success? && response.html? && base_url.contains?(response.effective_url)
  29. end
  30. module FixRedirectionsBehavior
  31. def self.included(base)
  32. base.extend ClassMethods
  33. end
  34. module ClassMethods
  35. attr_accessor :fix_redirections
  36. attr_reader :redirections
  37. def store_pages(store)
  38. return super unless fix_redirections
  39. instrument 'info.doc', msg: 'Fetching redirections...'
  40. with_redirections do
  41. instrument 'info.doc', msg: 'Building pages...'
  42. super
  43. end
  44. end
  45. private
  46. def with_redirections
  47. @redirections = new.fetch_redirections
  48. yield
  49. ensure
  50. @redirections = nil
  51. end
  52. end
  53. def fetch_redirections
  54. result = {}
  55. with_filters 'container', 'normalize_urls', 'internal_urls' do
  56. build_pages do |page|
  57. next if page[:response_effective_path] == page[:response_path]
  58. result[page[:response_path].downcase] = page[:response_effective_path]
  59. end
  60. end
  61. result
  62. end
  63. private
  64. def process_response(response)
  65. super.merge! response_effective_path: response.effective_path, response_path: response.path
  66. end
  67. def additional_options
  68. { redirections: self.class.redirections }
  69. end
  70. def with_filters(*filters)
  71. stack = FilterStack.new
  72. stack.push(*filters)
  73. pipeline.instance_variable_set :@filters, stack.to_a.freeze
  74. yield
  75. ensure
  76. @pipeline = nil
  77. end
  78. end
  79. include FixRedirectionsBehavior
  80. end
  81. end