GC classifier db

lib/linguist/classifier.rb

@@ -34,6 +34,21 @@ module Linguist
       @languages_total += 1
     end
 
+    def gc
+      @tokens.each do |language, tokens|
+        if @language_tokens[language] > 20
+          tokens.each do |name, count|
+            if count == 1
+              @tokens[language].delete(name)
+              @language_tokens[language] -= 1
+              @tokens_total -= 1
+            end
+          end
+        end
+      end
+      self
+    end
+
     def classify(data)
       tokens = Tokenizer.new(data).tokens
 

lib/linguist/sample.rb

@@ -29,7 +29,7 @@ module Linguist
     def self.classifier
       classifier = Classifier.new
       each { |sample| classifier.train(sample.language, sample.data) }
-      classifier
+      classifier.gc
     end
 
     def initialize(path, language)

test/test_classifier.rb

@@ -15,7 +15,7 @@ class TestClassifier < Test::Unit::TestCase
     File.read(File.join(fixtures_path, name))
   end
 
-  def test_train_and_classify
+  def test_classify
     classifier = Classifier.new
     classifier.train Language["Ruby"], fixture("ruby/foo.rb")
     classifier.train Language["Objective-C"], fixture("objective-c/Foo.h")
@@ -30,6 +30,10 @@ class TestClassifier < Test::Unit::TestCase
     assert results.first[1] < 0.5, results.first.inspect
   end
 
+  def test_gc
+    Classifier.instance.gc
+  end
+
   # def test_instance_classify
   #   Sample.each do |sample|
   #     results = Classifier.instance.classify(sample.data)