Refactoring of chrome speech recognition architecture (CL1.3)

content/browser/speech/speech_recognizer_impl_unittest.cc

 // Copyright (c) 2012 The Chromium Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 #include <vector>
 
 #include "content/browser/browser_thread_impl.h"
+#include "content/browser/speech/google_ssfe_remote_engine.h"
 #include "content/browser/speech/speech_recognizer_impl.h"
 #include "content/public/browser/speech_recognition_event_listener.h"
 #include "content/test/test_url_fetcher_factory.h"
 #include "media/audio/audio_manager.h"
 #include "media/audio/fake_audio_input_stream.h"
 #include "media/audio/fake_audio_output_stream.h"
 #include "media/audio/test_audio_input_controller_factory.h"
 #include "net/base/net_errors.h"
 #include "net/url_request/url_request_status.h"
 #include "testing/gtest/include/gtest/gtest.h"
 
+using base::MessageLoopProxy;
 using content::BrowserThread;
 using content::BrowserThreadImpl;
 using media::AudioInputController;
 using media::TestAudioInputController;
 using media::TestAudioInputControllerFactory;
 
 namespace {
 
 class MockAudioManager : public AudioManagerBase {
  public:
@@ skipping 45 matching lines @@
   virtual void UnMuteAll() OVERRIDE {}
   virtual bool IsRecordingInProcess() OVERRIDE { return false; }
   virtual scoped_refptr<base::MessageLoopProxy> GetMessageLoop() OVERRIDE {
     return audio_thread_->message_loop_proxy();
   }
   virtual void Init() OVERRIDE {};
  private:
   scoped_ptr<base::Thread> audio_thread_;
   DISALLOW_COPY_AND_ASSIGN(MockAudioManager);
 };
-}  // namespace
+} // namespace

[hans, 2012/03/16 11:12:56]

the extra space should be there

[Primiano Tucci (use gerrit), 2012/03/16 15:03:42]

Done.

 
 
 namespace speech {
 
 class SpeechRecognizerImplTest : public content::SpeechRecognitionEventListener,
                                  public testing::Test {
  public:
   SpeechRecognizerImplTest()
       : io_thread_(BrowserThread::IO, &message_loop_),
         audio_manager_(new MockAudioManager()),
         audio_ended_(false),
         recognition_ended_(false),
         result_received_(false),
         audio_started_(false),
         error_(content::SPEECH_RECOGNITION_ERROR_NONE),
         volume_(-1.0f) {
-    recognizer_ = new SpeechRecognizerImpl(
-        this, 1, std::string(), std::string(), NULL, false, std::string(),
-        std::string());
+    GoogleSSFERemoteEngine* sr_engine =
+        new GoogleSSFERemoteEngine(NULL /* URLRequestContextGetter */);
+    GoogleSSFERemoteEngineConfig config;
+    config.audio_num_bits_per_sample =
+        SpeechRecognizerImpl::kNumBitsPerAudioSample;
+    config.audio_sample_rate = SpeechRecognizerImpl::kAudioSampleRate;
+    config.filter_profanities = false;
+    sr_engine->SetConfiguration(config);
+
+    recognizer_ = new SpeechRecognizerImpl(this, 1, sr_engine);
     recognizer_->SetAudioManagerForTesting(audio_manager_.get());
+
     int audio_packet_length_bytes =
         (SpeechRecognizerImpl::kAudioSampleRate *
-         SpeechRecognizerImpl::kAudioPacketIntervalMs *
+         GoogleSSFERemoteEngine::kAudioPacketIntervalMs *
          ChannelLayoutToChannelCount(SpeechRecognizerImpl::kChannelLayout) *
          SpeechRecognizerImpl::kNumBitsPerAudioSample) / (8 * 1000);
     audio_packet_.resize(audio_packet_length_bytes);
   }
 
   // Overridden from content::SpeechRecognitionEventListener:
   virtual void OnAudioStart(int caller_id) OVERRIDE {
     audio_started_ = true;
   }
 
   virtual void OnAudioEnd(int caller_id) OVERRIDE {
     audio_ended_ = true;
   }
 
   virtual void OnRecognitionResult(
       int caller_id, const content::SpeechRecognitionResult& result) OVERRIDE {
     result_received_ = true;
   }
 
   virtual void OnRecognitionError(
-      int caller_id,
-      const content::SpeechRecognitionErrorCode& error) OVERRIDE {
-    error_ = error;
+      int caller_id, const content::SpeechRecognitionError& error) OVERRIDE {
+    error_ = error.code;
   }
 
   virtual void OnAudioLevelsChange(int caller_id, float volume,
-                                   float noise_volume) OVERRIDE {
+                                    float noise_volume) OVERRIDE {
     volume_ = volume;
     noise_volume_ = noise_volume;
   }
 
   virtual void OnRecognitionEnd(int caller_id) OVERRIDE {
     recognition_ended_ = true;
   }
 
   virtual void OnRecognitionStart(int caller_id) OVERRIDE {}
   virtual void OnEnvironmentEstimationComplete(int caller_id) OVERRIDE {}
@@ skipping 37 matching lines @@
   content::SpeechRecognitionErrorCode error_;
   TestURLFetcherFactory url_fetcher_factory_;
   TestAudioInputControllerFactory audio_input_controller_factory_;
   std::vector<uint8> audio_packet_;
   float volume_;
   float noise_volume_;
 };
 
 TEST_F(SpeechRecognizerImplTest, StopNoData) {
   // Check for callbacks when stopping record before any audio gets recorded.
-  EXPECT_TRUE(recognizer_->StartRecognition());
+  recognizer_->StartRecognition();
   recognizer_->AbortRecognition();
+  MessageLoop::current()->RunAllPending();
   EXPECT_FALSE(audio_ended_);
-  EXPECT_FALSE(recognition_ended_);
+  EXPECT_TRUE(recognition_ended_);

[Primiano Tucci (use gerrit), 2012/03/15 09:14:36]

The event interface is slightly changed. AudioEnded and RecognitionEnded now are
always raised, in order to guarantee start/end closure.

   EXPECT_FALSE(result_received_);
   EXPECT_FALSE(audio_started_);
   EXPECT_EQ(content::SPEECH_RECOGNITION_ERROR_NONE, error_);
 }
 
 TEST_F(SpeechRecognizerImplTest, CancelNoData) {
   // Check for callbacks when canceling recognition before any audio gets
   // recorded.
-  EXPECT_TRUE(recognizer_->StartRecognition());
+  recognizer_->StartRecognition();
   recognizer_->StopAudioCapture();
-  EXPECT_TRUE(audio_ended_);
+  MessageLoop::current()->RunAllPending();
+  EXPECT_FALSE(audio_ended_);
   EXPECT_TRUE(recognition_ended_);
   EXPECT_FALSE(result_received_);
   EXPECT_FALSE(audio_started_);
   EXPECT_EQ(content::SPEECH_RECOGNITION_ERROR_NONE, error_);
 }
 
 TEST_F(SpeechRecognizerImplTest, StopWithData) {
   // Start recording, give some data and then stop. This should wait for the
   // network callback to arrive before completion.
-  EXPECT_TRUE(recognizer_->StartRecognition());
+  recognizer_->StartRecognition();
+  MessageLoop::current()->RunAllPending();
   TestAudioInputController* controller =
       audio_input_controller_factory_.controller();
   ASSERT_TRUE(controller);
 
   // Try sending 5 chunks of mock audio data and verify that each of them
   // resulted immediately in a packet sent out via the network. This verifies
   // that we are streaming out encoded data as chunks without waiting for the
   // full recording to complete.
   const size_t kNumChunks = 5;
   for (size_t i = 0; i < kNumChunks; ++i) {
     controller->event_handler()->OnData(controller, &audio_packet_[0],
                                         audio_packet_.size());
     MessageLoop::current()->RunAllPending();
     TestURLFetcher* fetcher = url_fetcher_factory_.GetFetcherByID(0);
     ASSERT_TRUE(fetcher);
     EXPECT_EQ(i + 1, fetcher->upload_chunks().size());
   }
 
   recognizer_->StopAudioCapture();
+  MessageLoop::current()->RunAllPending();
   EXPECT_TRUE(audio_started_);
   EXPECT_TRUE(audio_ended_);
   EXPECT_FALSE(recognition_ended_);
   EXPECT_FALSE(result_received_);
   EXPECT_EQ(content::SPEECH_RECOGNITION_ERROR_NONE, error_);
 
   // Issue the network callback to complete the process.
   TestURLFetcher* fetcher = url_fetcher_factory_.GetFetcherByID(0);
   ASSERT_TRUE(fetcher);
 
   fetcher->set_url(fetcher->GetOriginalURL());
   net::URLRequestStatus status;
   status.set_status(net::URLRequestStatus::SUCCESS);
   fetcher->set_status(status);
   fetcher->set_response_code(200);
   fetcher->SetResponseString(
       "{\"status\":0,\"hypotheses\":[{\"utterance\":\"123\"}]}");
   fetcher->delegate()->OnURLFetchComplete(fetcher);
-
+  MessageLoop::current()->RunAllPending();
   EXPECT_TRUE(recognition_ended_);
   EXPECT_TRUE(result_received_);
   EXPECT_EQ(content::SPEECH_RECOGNITION_ERROR_NONE, error_);
 }
 
 TEST_F(SpeechRecognizerImplTest, CancelWithData) {
-  // Start recording, give some data and then cancel. This should create
-  // a network request but give no callbacks.
-  EXPECT_TRUE(recognizer_->StartRecognition());
+  // Start recording, give some data and then cancel.
+  recognizer_->StartRecognition();
+  MessageLoop::current()->RunAllPending();
   TestAudioInputController* controller =
       audio_input_controller_factory_.controller();
   ASSERT_TRUE(controller);
   controller->event_handler()->OnData(controller, &audio_packet_[0],
                                       audio_packet_.size());
   MessageLoop::current()->RunAllPending();
   recognizer_->AbortRecognition();
+  MessageLoop::current()->RunAllPending();
   ASSERT_TRUE(url_fetcher_factory_.GetFetcherByID(0));
   EXPECT_TRUE(audio_started_);
-  EXPECT_FALSE(audio_ended_);
-  EXPECT_FALSE(recognition_ended_);
+  EXPECT_TRUE(audio_ended_);
+  EXPECT_TRUE(recognition_ended_);
   EXPECT_FALSE(result_received_);
   EXPECT_EQ(content::SPEECH_RECOGNITION_ERROR_NONE, error_);
 }
 
 TEST_F(SpeechRecognizerImplTest, ConnectionError) {
   // Start recording, give some data and then stop. Issue the network callback
   // with a connection error and verify that the recognizer bubbles the error up
-  EXPECT_TRUE(recognizer_->StartRecognition());
+  recognizer_->StartRecognition();
+  MessageLoop::current()->RunAllPending();
   TestAudioInputController* controller =
       audio_input_controller_factory_.controller();
   ASSERT_TRUE(controller);
   controller->event_handler()->OnData(controller, &audio_packet_[0],
                                       audio_packet_.size());
   MessageLoop::current()->RunAllPending();
   TestURLFetcher* fetcher = url_fetcher_factory_.GetFetcherByID(0);
   ASSERT_TRUE(fetcher);
 
   recognizer_->StopAudioCapture();
+  MessageLoop::current()->RunAllPending();
   EXPECT_TRUE(audio_started_);
   EXPECT_TRUE(audio_ended_);
   EXPECT_FALSE(recognition_ended_);
   EXPECT_FALSE(result_received_);
   EXPECT_EQ(content::SPEECH_RECOGNITION_ERROR_NONE, error_);
 
   // Issue the network callback to complete the process.
   fetcher->set_url(fetcher->GetOriginalURL());
   net::URLRequestStatus status;
   status.set_status(net::URLRequestStatus::FAILED);
   status.set_error(net::ERR_CONNECTION_REFUSED);
   fetcher->set_status(status);
   fetcher->set_response_code(0);
   fetcher->SetResponseString("");
   fetcher->delegate()->OnURLFetchComplete(fetcher);
-
-  EXPECT_FALSE(recognition_ended_);
+  MessageLoop::current()->RunAllPending();
+  EXPECT_TRUE(recognition_ended_);
   EXPECT_FALSE(result_received_);
   EXPECT_EQ(content::SPEECH_RECOGNITION_ERROR_NETWORK, error_);
 }
 
 TEST_F(SpeechRecognizerImplTest, ServerError) {
   // Start recording, give some data and then stop. Issue the network callback
   // with a 500 error and verify that the recognizer bubbles the error up
-  EXPECT_TRUE(recognizer_->StartRecognition());
+  recognizer_->StartRecognition();
+  MessageLoop::current()->RunAllPending();
   TestAudioInputController* controller =
       audio_input_controller_factory_.controller();
   ASSERT_TRUE(controller);
   controller->event_handler()->OnData(controller, &audio_packet_[0],
                                       audio_packet_.size());
   MessageLoop::current()->RunAllPending();
   TestURLFetcher* fetcher = url_fetcher_factory_.GetFetcherByID(0);
   ASSERT_TRUE(fetcher);
 
   recognizer_->StopAudioCapture();
+  MessageLoop::current()->RunAllPending();
   EXPECT_TRUE(audio_started_);
   EXPECT_TRUE(audio_ended_);
   EXPECT_FALSE(recognition_ended_);
   EXPECT_FALSE(result_received_);
   EXPECT_EQ(content::SPEECH_RECOGNITION_ERROR_NONE, error_);
 
   // Issue the network callback to complete the process.
   fetcher->set_url(fetcher->GetOriginalURL());
   net::URLRequestStatus status;
   status.set_status(net::URLRequestStatus::SUCCESS);
   fetcher->set_status(status);
   fetcher->set_response_code(500);
   fetcher->SetResponseString("Internal Server Error");
   fetcher->delegate()->OnURLFetchComplete(fetcher);
-
-  EXPECT_FALSE(recognition_ended_);
+  MessageLoop::current()->RunAllPending();
+  EXPECT_TRUE(recognition_ended_);
   EXPECT_FALSE(result_received_);
   EXPECT_EQ(content::SPEECH_RECOGNITION_ERROR_NETWORK, error_);
 }
 
 TEST_F(SpeechRecognizerImplTest, AudioControllerErrorNoData) {
   // Check if things tear down properly if AudioInputController threw an error.
-  EXPECT_TRUE(recognizer_->StartRecognition());
+  recognizer_->StartRecognition();
+  MessageLoop::current()->RunAllPending();
   TestAudioInputController* controller =
       audio_input_controller_factory_.controller();
   ASSERT_TRUE(controller);
   controller->event_handler()->OnError(controller, 0);
   MessageLoop::current()->RunAllPending();
   EXPECT_FALSE(audio_started_);
   EXPECT_FALSE(audio_ended_);
-  EXPECT_FALSE(recognition_ended_);
+  EXPECT_TRUE(recognition_ended_);
   EXPECT_FALSE(result_received_);
   EXPECT_EQ(content::SPEECH_RECOGNITION_ERROR_AUDIO, error_);
 }
 
 TEST_F(SpeechRecognizerImplTest, AudioControllerErrorWithData) {
   // Check if things tear down properly if AudioInputController threw an error
   // after giving some audio data.
-  EXPECT_TRUE(recognizer_->StartRecognition());
+  recognizer_->StartRecognition();
+  MessageLoop::current()->RunAllPending();
   TestAudioInputController* controller =
       audio_input_controller_factory_.controller();
   ASSERT_TRUE(controller);
   controller->event_handler()->OnData(controller, &audio_packet_[0],
                                       audio_packet_.size());
   controller->event_handler()->OnError(controller, 0);
   MessageLoop::current()->RunAllPending();
   ASSERT_TRUE(url_fetcher_factory_.GetFetcherByID(0));
   EXPECT_TRUE(audio_started_);
-  EXPECT_FALSE(audio_ended_);
-  EXPECT_FALSE(recognition_ended_);
+  EXPECT_TRUE(audio_ended_);
+  EXPECT_TRUE(recognition_ended_);
   EXPECT_FALSE(result_received_);
   EXPECT_EQ(content::SPEECH_RECOGNITION_ERROR_AUDIO, error_);
 }
 
 TEST_F(SpeechRecognizerImplTest, NoSpeechCallbackIssued) {
   // Start recording and give a lot of packets with audio samples set to zero.
   // This should trigger the no-speech detector and issue a callback.
-  EXPECT_TRUE(recognizer_->StartRecognition());
+  recognizer_->StartRecognition();
+  MessageLoop::current()->RunAllPending();
   TestAudioInputController* controller =
       audio_input_controller_factory_.controller();
   ASSERT_TRUE(controller);
-  controller = audio_input_controller_factory_.controller();
-  ASSERT_TRUE(controller);
 
-  int num_packets = (SpeechRecognizerImpl::kNoSpeechTimeoutSec * 1000) /
-                     SpeechRecognizerImpl::kAudioPacketIntervalMs;
+  int num_packets = (SpeechRecognizerImpl::kNoSpeechTimeoutMs) /
+                     GoogleSSFERemoteEngine::kAudioPacketIntervalMs + 1;
   // The vector is already filled with zero value samples on create.
   for (int i = 0; i < num_packets; ++i) {
     controller->event_handler()->OnData(controller, &audio_packet_[0],
                                         audio_packet_.size());
   }
   MessageLoop::current()->RunAllPending();
   EXPECT_TRUE(audio_started_);
-  EXPECT_FALSE(audio_ended_);
-  EXPECT_FALSE(recognition_ended_);
+  EXPECT_TRUE(audio_ended_);
+  EXPECT_TRUE(recognition_ended_);
   EXPECT_FALSE(result_received_);
   EXPECT_EQ(content::SPEECH_RECOGNITION_ERROR_NO_SPEECH, error_);
 }
 
 TEST_F(SpeechRecognizerImplTest, NoSpeechCallbackNotIssued) {
   // Start recording and give a lot of packets with audio samples set to zero
   // and then some more with reasonably loud audio samples. This should be
   // treated as normal speech input and the no-speech detector should not get
   // triggered.
-  EXPECT_TRUE(recognizer_->StartRecognition());
+  recognizer_->StartRecognition();
+  MessageLoop::current()->RunAllPending();
   TestAudioInputController* controller =
       audio_input_controller_factory_.controller();
   ASSERT_TRUE(controller);
   controller = audio_input_controller_factory_.controller();
   ASSERT_TRUE(controller);
 
-  int num_packets = (SpeechRecognizerImpl::kNoSpeechTimeoutSec * 1000) /
-                     SpeechRecognizerImpl::kAudioPacketIntervalMs;
+  int num_packets = (SpeechRecognizerImpl::kNoSpeechTimeoutMs) /
+                     GoogleSSFERemoteEngine::kAudioPacketIntervalMs;
 
   // The vector is already filled with zero value samples on create.
   for (int i = 0; i < num_packets / 2; ++i) {
     controller->event_handler()->OnData(controller, &audio_packet_[0],
                                         audio_packet_.size());
   }
 
   FillPacketWithTestWaveform();
   for (int i = 0; i < num_packets / 2; ++i) {
     controller->event_handler()->OnData(controller, &audio_packet_[0],
                                         audio_packet_.size());
   }
 
   MessageLoop::current()->RunAllPending();
   EXPECT_EQ(content::SPEECH_RECOGNITION_ERROR_NONE, error_);
   EXPECT_TRUE(audio_started_);
   EXPECT_FALSE(audio_ended_);
   EXPECT_FALSE(recognition_ended_);
   recognizer_->AbortRecognition();
+  MessageLoop::current()->RunAllPending();
 }
 
 TEST_F(SpeechRecognizerImplTest, SetInputVolumeCallback) {
   // Start recording and give a lot of packets with audio samples set to zero
   // and then some more with reasonably loud audio samples. Check that we don't
   // get the callback during estimation phase, then get zero for the silence
   // samples and proper volume for the loud audio.
-  EXPECT_TRUE(recognizer_->StartRecognition());
+  recognizer_->StartRecognition();
+  MessageLoop::current()->RunAllPending();
   TestAudioInputController* controller =
       audio_input_controller_factory_.controller();
   ASSERT_TRUE(controller);
   controller = audio_input_controller_factory_.controller();
   ASSERT_TRUE(controller);
 
   // Feed some samples to begin with for the endpointer to do noise estimation.
   int num_packets = SpeechRecognizerImpl::kEndpointerEstimationTimeMs /
-                    SpeechRecognizerImpl::kAudioPacketIntervalMs;
+                    GoogleSSFERemoteEngine::kAudioPacketIntervalMs;
   FillPacketWithNoise();
   for (int i = 0; i < num_packets; ++i) {
     controller->event_handler()->OnData(controller, &audio_packet_[0],
                                         audio_packet_.size());
   }
   MessageLoop::current()->RunAllPending();
   EXPECT_EQ(-1.0f, volume_);  // No audio volume set yet.
 
   // The vector is already filled with zero value samples on create.
   controller->event_handler()->OnData(controller, &audio_packet_[0],
                                       audio_packet_.size());
   MessageLoop::current()->RunAllPending();
   EXPECT_FLOAT_EQ(0.74939233f, volume_);
 
   FillPacketWithTestWaveform();
   controller->event_handler()->OnData(controller, &audio_packet_[0],
                                       audio_packet_.size());
   MessageLoop::current()->RunAllPending();
   EXPECT_FLOAT_EQ(0.89926866f, volume_);
   EXPECT_FLOAT_EQ(0.75071919f, noise_volume_);
 
   EXPECT_EQ(content::SPEECH_RECOGNITION_ERROR_NONE, error_);
   EXPECT_FALSE(audio_ended_);
   EXPECT_FALSE(recognition_ended_);
   recognizer_->AbortRecognition();
+  MessageLoop::current()->RunAllPending();
 }
 
 }  // namespace speech