diff --git a/app/lib/link_details_extractor.rb b/app/lib/link_details_extractor.rb
index bd78aef7a9f..bebc9c9fd72 100644
--- a/app/lib/link_details_extractor.rb
+++ b/app/lib/link_details_extractor.rb
@@ -157,7 +157,7 @@ class LinkDetailsExtractor
end
def title
- html_entities_decode(structured_data&.headline || opengraph_tag('og:title') || document.xpath('//title').map(&:content).first)&.strip
+ html_entities_decode(structured_data&.headline || opengraph_tag('og:title') || head.at_xpath('title')&.content)&.strip
end
def description
@@ -205,11 +205,11 @@ class LinkDetailsExtractor
end
def language
- valid_locale_or_nil(structured_data&.language || opengraph_tag('og:locale') || document.xpath('//html').pick('lang'))
+ valid_locale_or_nil(structured_data&.language || opengraph_tag('og:locale') || document.root.attr('lang'))
end
def icon
- valid_url_or_nil(structured_data&.publisher_icon || link_tag('apple-touch-icon') || link_tag('shortcut icon'))
+ valid_url_or_nil(structured_data&.publisher_icon || link_tag('apple-touch-icon') || link_tag('icon'))
end
private
@@ -237,11 +237,11 @@ class LinkDetailsExtractor
end
def link_tag(name)
- document.xpath("//link[@rel=\"#{name}\"]").pick('href')
+ head.xpath('//link[@rel][@href]').find { |el| el.kwattr_values('rel').any? { |v| name.casecmp?(v) } }&.attr('href')
end
def opengraph_tag(name)
- document.xpath("//meta[@property=\"#{name}\" or @name=\"#{name}\"]").pick('content')
+ head.at_xpath("//meta[@property='#{name}' or @name='#{name}'][@content]")&.attr('content')
end
def meta_tag(name)
@@ -249,6 +249,8 @@ class LinkDetailsExtractor
end
def structured_data
+ return @structured_data if defined?(@structured_data)
+
# Some publications have more than one JSON-LD definition on the page,
# and some of those definitions aren't valid JSON either, so we have
# to loop through here until we find something that is the right type
@@ -273,6 +275,10 @@ class LinkDetailsExtractor
@document ||= detect_encoding_and_parse_document
end
+ def head
+ @head ||= document.at_xpath('/html/head')
+ end
+
def detect_encoding_and_parse_document
[detect_encoding, nil, header_encoding].uniq.each do |encoding|
document = Nokogiri::HTML(@html, nil, encoding)
diff --git a/spec/lib/link_details_extractor_spec.rb b/spec/lib/link_details_extractor_spec.rb
index b1e5cedced3..e7f22eddabb 100644
--- a/spec/lib/link_details_extractor_spec.rb
+++ b/spec/lib/link_details_extractor_spec.rb
@@ -42,6 +42,7 @@ RSpec.describe LinkDetailsExtractor do
Man bites dog
+