Remove Postrank::URI

PostRank::URI couldn’t handle umlauts.

We will lose the feature to detect urls without protocol “twingly.com”,
but we don’t see the need for this feature.

On the plus side, lots of runtime dependencies are removed (nokogiri!).

Gemfile.lock

@@ -4,7 +4,6 @@ PATH
     twingly-url-normalizer (0.0.1)
       addressable
       domainatrix
-      postrank-uri
 
 GEM
   remote: https://rubygems.org/
@@ -13,12 +12,6 @@ GEM
     ansi (1.4.3)
     domainatrix (0.0.11)
       addressable
-    nokogiri (1.5.11)
-    postrank-uri (1.0.17)
-      addressable (~> 2.3.0)
-      nokogiri (~> 1.5.5)
-      public_suffix (~> 1.1.3)
-    public_suffix (1.1.3)
     rake (10.1.1)
     shoulda-context (1.1.6)
     turn (0.9.6)

lib/twingly-url-normalizer.rb

@@ -1,4 +1,3 @@
-require 'postrank-uri'
 require 'domainatrix'
 require 'addressable/uri'
 
@@ -12,7 +11,7 @@ def self.normalize(potential_urls)
       end
 
       def self.extract_urls(potential_urls)
-        PostRank::URI.extract(potential_urls)
+        Array(potential_urls).map(&:split).flatten
       end
 
       def self.normalize_url(url)

test/unit/normalization_test.rb

@@ -86,5 +86,10 @@ class NormalizerTest < Test::Unit::TestCase
 
       assert_equal "http://www.twingly.com/", result
     end
+
+    should "not be able to normalize url without protocol" do
+      url = "twingly.com/"
+      assert_raises(Addressable::URI::InvalidURIError) { @normalizer.normalize_url(url) }
+    end
   end
 end

twingly-url-normalizer.gemspec

@@ -12,7 +12,6 @@ Gem::Specification.new do |s|
   s.summary     = "Ruby library for URL normalization"
   s.required_ruby_version = ">= 1.9.3"
 
-  s.add_dependency "postrank-uri"
   s.add_dependency "domainatrix"
   s.add_dependency "addressable"