Bayes integration test of Memory and Redis backends with real data

.gitignore

@@ -1,2 +1,3 @@
 Gemfile.lock
 pkg
+*.rdb

test/bayes/bayesian_common_tests.rb

@@ -2,7 +2,7 @@
 
 module BayesianCommonTests
   def test_good_training
-     assert_equal ['love'], @classifier.train_interesting('love')
+    assert_equal ['love'], @classifier.train_interesting('love')
   end
 
   def test_training_with_utf8

test/bayes/bayesian_integration_test.rb

@@ -0,0 +1,54 @@
+# encoding: utf-8
+
+require File.dirname(__FILE__) + '/../test_helper'
+
+class BayesianIntegrationTest < Minitest::Test
+  def setup
+    begin
+      @memory_classifier = ClassifierReborn::Bayes.new 'Ham', 'Spam'
+      @redis_backend = ClassifierReborn::BayesRedisBackend.new
+      @redis_classifier = ClassifierReborn::Bayes.new 'Ham', 'Spam', backend: @redis_backend
+    rescue Redis::CannotConnectError => e
+      skip(e)
+    end
+    sms_spam_collection = File.expand_path(File.dirname(__FILE__) + '/../data/corpus/SMSSpamCollection.tsv')
+    @training_set = File.read(sms_spam_collection).force_encoding("utf-8").split("\n")
+    @testing_set = @training_set.pop(1000)
+  end
+
+  def teardown
+    @redis_backend.instance_variable_get(:@redis).flushdb
+  end
+
+  def test_equality_of_backends
+    train_model @memory_classifier
+    train_model @redis_classifier
+    assert_equal classification_scores(@memory_classifier).hash, classification_scores(@redis_classifier).hash
+    untrain_model @memory_classifier, 2000
+    untrain_model @redis_classifier, 2000
+    assert_equal classification_scores(@memory_classifier).hash, classification_scores(@redis_classifier).hash
+  end
+
+  def train_model(classifier)
+    @training_set.each do |line|
+      parts = line.strip.split("\t")
+      classifier.train(parts.first, parts.last)
+    end
+  end
+
+  def untrain_model(classifier, limit=Float::INFINITY)
+    @training_set.each_with_index do |line, i|
+      break if i >= limit
+      parts = line.strip.split("\t")
+      classifier.untrain(parts.first, parts.last)
+    end
+  end
+
+  def classification_scores(classifier)
+    @testing_set.collect do |line|
+      parts = line.strip.split("\t")
+      result, score = classifier.classify_with_score(parts.last)
+      "#{result}:#{score}"
+    end
+  end
+end

test/bayes/bayesian_redis_test.rb

@@ -8,25 +8,28 @@ class BayesianRedisTest < Minitest::Test
 
   def setup
     begin
-      @classifier = ClassifierReborn::Bayes.new 'Interesting', 'Uninteresting', backend: ClassifierReborn::BayesRedisBackend.new
+      @redis_backend = ClassifierReborn::BayesRedisBackend.new
+      @alternate_redis_backend = ClassifierReborn::BayesRedisBackend.new(db: 1)
+      @classifier = ClassifierReborn::Bayes.new 'Interesting', 'Uninteresting', backend: @redis_backend
     rescue Redis::CannotConnectError => e
       skip(e)
     end
   end
 
   def teardown
-    @classifier.instance_variable_get(:@backend).instance_variable_get(:@redis).flushall
+    @redis_backend.instance_variable_get(:@redis).flushdb
+    @alternate_redis_backend.instance_variable_get(:@redis).flushdb
   end
 
   def another_classifier
-    ClassifierReborn::Bayes.new %w(Interesting Uninteresting), backend: ClassifierReborn::BayesRedisBackend.new(db: 1)
+    ClassifierReborn::Bayes.new %w(Interesting Uninteresting), backend: @alternate_redis_backend
   end
 
   def auto_categorize_classifier
-    ClassifierReborn::Bayes.new 'Interesting', 'Uninteresting', auto_categorize: true, backend: ClassifierReborn::BayesRedisBackend.new(db: 1)
+    ClassifierReborn::Bayes.new 'Interesting', 'Uninteresting', auto_categorize: true, backend: @alternate_redis_backend
   end
 
   def threshold_classifier(category)
-    ClassifierReborn::Bayes.new category, backend: ClassifierReborn::BayesRedisBackend.new(db: 1)
+    ClassifierReborn::Bayes.new category, backend: @alternate_redis_backend
   end
 end

test/data/corpus/README.md

@@ -0,0 +1,5 @@
+# Sample dtasets for training and testing
+
+## SMSSpamCollection.tsv
+
+The [SMS Spam Collection v.1](http://dcomp.sor.ufscar.br/talmeida/smsspamcollection/) is a public set of SMS labeled messages that have been collected for mobile phone spam research. It has one collection composed by 5,574 English, real and non-enconded messages, tagged according being `ham` (legitimate) or `spam`.