air18n 0.1.29 → 0.1.30

data/lib/air18n/phrase_translation.rb

@@ -370,10 +370,10 @@ module Air18n
           compare_to = translation_pair[:previous_translation].size
         else
           distance = levenshtein_distance_fast(
-            PhraseTranslation.segment(translation_pair[:previous_translation]),
-            PhraseTranslation.segment(translation_pair[:translation]),
+            translation_pair[:previous_translation].scan(/[[:alnum:]]+/),
+            translation_pair[:translation].scan(/[[:alnum:]]+/),
             1, 0, 1)
-          compare_to = PhraseTranslation.segment(translation_pair[:previous_translation]).size
+          compare_to = translation_pair[:previous_translation].scan(/[[:alnum:]]+/).size
         end
 
         proportion_translated = [distance.to_f / compare_to.to_f, 1.0].min
@@ -400,7 +400,6 @@ module Air18n
         phrase_to_phrase_translations = pt_scope.all.group_by { |pt| [pt.locale, pt.phrase_id] }
         phrase_to_phrase_translations.each do |(locale, phrase_id), phrase_translations|
           phrase_translations.sort_by! { |pt| pt.created_at }
-          LoggingHelper.info "Getting previous translation!"
           previous_translation = PhraseTranslation.where("created_at < ?", phrase_translations.first.created_at - 1.second).
                                                    where(:locale => locale).
                                                    where(:phrase_id => phrase_id).
@@ -453,6 +452,8 @@ module Air18n
     # Break text into words for purposes of word count.
     # First uses to_s to convert text to a string; so for a nil input, returns
     # empty array.
+    # Should only be used for English text, because it only looks at words
+    # composed of regexp "\w".
     def self.segment(text)
       text.to_s.scan(/\w+/)
     end

data/lib/air18n/version.rb

@@ -1,3 +1,3 @@
 module Air18n
-  VERSION = "0.1.29"
+  VERSION = "0.1.30"
 end

data/spec/lib/air18n/phrase_translation_spec.rb

@@ -513,6 +513,23 @@ describe Air18n::PhraseTranslation do
         words_translated, words_verified = Air18n::PhraseTranslation.word_counts_from_translation_pair(pair)
         words_translated.should == 1 # 3 / 17 * 5 = 0.88, rounds to 1
         words_verified.should == 4
+
+        # Test that Greek works. There was once a bug where word segmentation
+        # didn't work for non-Latin alphabets.
+        pair = {
+          :translation=>"ακύρωση",
+          :previous_translation=>"ακύρωση",
+          :was_stale=>false,
+          :locale=>"el",
+          :user_id=>2895375,
+          :previous_user_id=>2462280,
+          :datetime=>Date.new(2012, 11, 2),
+          :source_word_count=>1,
+          :phrase_key=>"cancel"
+        }
+        words_translated, words_verified = Air18n::PhraseTranslation.word_counts_from_translation_pair(pair)
+        words_translated.should == 0
+        words_verified.should == 1
       end
 
       it 'should compute correct monthly activities' do

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: air18n
 version: !ruby/object:Gem::Version
-  version: 0.1.29
+  version: 0.1.30
   prerelease: 
 platform: ruby
 authors:
@@ -13,7 +13,7 @@ authors:
 autorequire: 
 bindir: bin
 cert_chain: []
-date: 2012-11-13 00:00:00.000000000 Z
+date: 2012-11-14 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: i18n