better domain/desc splitting in quackit script
This commit is contained in:
parent
03da6a79fa
commit
78814f7e09
2
Rakefile
2
Rakefile
@ -71,7 +71,7 @@ task :quackit_import do
|
||||
new_domains = Set.new
|
||||
doc = Nokogiri::HTML(open('http://www.quackit.com/domain-names/country_domain_extensions.cfm'))
|
||||
doc.css('#content li').each do |li|
|
||||
desc = li.content.split(/\s+-\s+/)[1]
|
||||
desc = li.content.split(/\s+-\s+/, 2)[1]
|
||||
if desc =~ /academic|education|school/i
|
||||
domain_el = li.at_css('b')
|
||||
# some lines have more than one domain listed
|
||||
|
||||
Loading…
Reference in New Issue
Block a user