Introducing the basic functionality to the speech input experimental API.

chrome/browser/extensions/extension_speech_input_manager.cc

 // Copyright (c) 2011 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 "chrome/browser/extensions/extension_speech_input_manager.h"
 
+#include "base/bind.h"
+#include "base/json/json_writer.h"
+#include "base/utf_string_conversions.h"
+#include "base/values.h"
+#include "chrome/browser/extensions/extension_event_router.h"
+#include "chrome/browser/extensions/extension_speech_input_api_constants.h"
+#include "chrome/browser/profiles/profile.h"
 #include "chrome/browser/profiles/profile_dependency_manager.h"
 #include "chrome/browser/profiles/profile_keyed_service.h"
 #include "chrome/browser/profiles/profile_keyed_service_factory.h"
 #include "chrome/common/chrome_notification_types.h"
 #include "chrome/common/extensions/extension.h"
+#include "content/browser/browser_thread.h"
 #include "content/common/notification_service.h"
 
+using namespace speech_input;
+
+namespace constants = extension_speech_input_api_constants;
+
 namespace {
 
 // Wrap an ExtensionSpeechInputManager using scoped_refptr to avoid
 // assertion failures on destruction because of not using release().
 class ExtensionSpeechInputManagerWrapper : public ProfileKeyedService {
  public:
   explicit ExtensionSpeechInputManagerWrapper(
       ExtensionSpeechInputManager* manager)
       : manager_(manager) {}
 
@@ skipping 62 matching lines @@
 ProfileKeyedService*
     ExtensionSpeechInputManager::Factory::BuildServiceInstanceFor(
     Profile* profile) const {
   scoped_refptr<ExtensionSpeechInputManager> manager(
       new ExtensionSpeechInputManager(profile));
   return new ExtensionSpeechInputManagerWrapper(manager);
 }
 
 ExtensionSpeechInputManager::ExtensionSpeechInputManager(Profile* profile)
     : profile_(profile),
-      extension_(NULL) {
+      state_(kIdle) {
   registrar_.Add(this, chrome::NOTIFICATION_EXTENSION_UNLOADED,
                  Source<Profile>(profile_));
 }
 
 ExtensionSpeechInputManager::~ExtensionSpeechInputManager() {
 }
 
 ExtensionSpeechInputManager* ExtensionSpeechInputManager::GetForProfile(
     Profile* profile) {
   ExtensionSpeechInputManagerWrapper *wrapper =
@@ skipping 10 matching lines @@
 void ExtensionSpeechInputManager::Observe(int type,
     const NotificationSource& source,
     const NotificationDetails& details) {
   if (type == chrome::NOTIFICATION_EXTENSION_UNLOADED)
     ExtensionUnloaded(Details<UnloadedExtensionInfo>(details)->extension->id());
   else
     NOTREACHED();
 }
 
 void ExtensionSpeechInputManager::ShutdownOnUIThread() {
-  // TODO(leandrogracia): Force stop to speech recognition if active.
+  DCHECK(BrowserThread::CurrentlyOn(BrowserThread::UI));
+  VLOG(1) << "Profile shutting down.";
+
+  base::AutoLock auto_lock(state_lock_);
+  DCHECK(state_ != kShutdown);
+  if (state_ != kIdle) {
+    BrowserThread::PostTask(BrowserThread::IO, FROM_HERE,
+        base::Bind(&ExtensionSpeechInputManager::ForceStopOnIOThread, this));
+  }
+  state_ = kShutdown;
+  VLOG(1) << "Entering the shutdown sink state.";
   registrar_.RemoveAll();
   profile_ = NULL;
 }
 
-void ExtensionSpeechInputManager::ExtensionUnloaded(const std::string& id) {
-  // TODO(leandrogracia): Force stop to speech recognition if the extension
-  // is currently using the API.
-}
-
-void ExtensionSpeechInputManager::Start(const Extension *extension) {
-  // TODO(leandrogracia): Start speech recognition.
-}
-
-void ExtensionSpeechInputManager::Stop(const Extension *extension) {
-  // TODO(leandrogracia): Stop speech recognition.
-}
-
+void ExtensionSpeechInputManager::ExtensionUnloaded(
+    const std::string& extension_id) {
+  DCHECK(BrowserThread::CurrentlyOn(BrowserThread::UI));
+
+  base::AutoLock auto_lock(state_lock_);
+  if (state_ == kShutdown)
+    return;
+
+  VLOG(1) << "Extension unloaded. Requesting to enforce stop...";
+  if (extension_id_in_use_ == extension_id) {
+    if (state_ != kIdle) {
+      BrowserThread::PostTask(BrowserThread::IO, FROM_HERE,
+          base::Bind(&ExtensionSpeechInputManager::ForceStopOnIOThread, this));
+    }
+  }
+}
+
+void ExtensionSpeechInputManager::SetRecognitionResult(
+    int caller_id,
+    const SpeechInputResult& result) {
+  DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
+  DCHECK_EQ(caller_id, kSpeechCallerId);
+
+  BrowserThread::PostTask(BrowserThread::UI, FROM_HERE,
+      base::Bind(&ExtensionSpeechInputManager::SetRecognitionResultOnUIThread,
+      this, result));
+}
+
+void ExtensionSpeechInputManager::SetRecognitionResultOnUIThread(
+    const SpeechInputResult& result) {
+  DCHECK(BrowserThread::CurrentlyOn(BrowserThread::UI));
+
+  {
+    base::AutoLock auto_lock(state_lock_);
+    if (state_ == kShutdown)
+      return;
+    state_ = kIdle;
+  }
+
+  ListValue args;
+  DictionaryValue* js_event = new DictionaryValue();
+  args.Append(js_event);
+
+  ListValue* js_hypothesis_array = new ListValue();
+  js_event->Set(constants::kHypothesesKey, js_hypothesis_array);
+
+  for (size_t i = 0; i < result.hypotheses.size(); ++i) {
+    const SpeechInputHypothesis& hypothesis = result.hypotheses[i];
+
+    DictionaryValue* js_hypothesis_object = new DictionaryValue();
+    js_hypothesis_array->Append(js_hypothesis_object);
+
+    js_hypothesis_object->SetString(constants::kUtteranceKey,
+        UTF16ToUTF8(hypothesis.utterance));
+    js_hypothesis_object->SetDouble(constants::kConfidenceKey,
+        hypothesis.confidence);
+  }
+
+  std::string json_args;
+  base::JSONWriter::Write(&args, false, &json_args);
+  VLOG(1) << "Results: " << json_args;
+  DispatchEventToExtension(constants::kOnResultEvent, json_args);
+}
+
+void ExtensionSpeechInputManager::DidStartReceivingAudio(int caller_id) {
+  VLOG(1) << "DidStartReceivingAudio";
+  DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
+  DCHECK_EQ(caller_id, kSpeechCallerId);
+
+  BrowserThread::PostTask(BrowserThread::UI, FROM_HERE,
+      base::Bind(&ExtensionSpeechInputManager::DidStartReceivingAudioOnUIThread,
+      this));
+}
+
+void ExtensionSpeechInputManager::DidCompleteRecording(int caller_id) {
+  DCHECK_EQ(caller_id, kSpeechCallerId);
+}
+
+void ExtensionSpeechInputManager::DidCompleteRecognition(int caller_id) {
+  DCHECK_EQ(caller_id, kSpeechCallerId);
+}
+
+void ExtensionSpeechInputManager::DidStartReceivingAudioOnUIThread() {
+  DCHECK(BrowserThread::CurrentlyOn(BrowserThread::UI));
+
+  base::AutoLock auto_lock(state_lock_);
+  if (state_ == kShutdown)
+    return;
+
+  DCHECK_EQ(state_, kStarting);
+  VLOG(1) << "State changed to recording";
+  state_ = kRecording;
+
+  VLOG(1) << "Sending start notification";
+  NotificationService::current()->Notify(
+      chrome::NOTIFICATION_EXTENSION_SPEECH_INPUT_RECORDING_STARTED,
+      Source<Profile>(profile_),
+      Details<std::string>(&extension_id_in_use_));
+}
+
+void ExtensionSpeechInputManager::OnRecognizerError(
+    int caller_id, SpeechInputError error) {
+  DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
+  DCHECK_EQ(caller_id, kSpeechCallerId);
+  VLOG(1) << "OnRecognizerError: " << error;
+
+  base::AutoLock auto_lock(state_lock_);
+  if (state_ == kShutdown)
+    return;
+
+  std::string event_error_code;
+  switch (error) {
+    case kErrorNone:
+      break;
+
+    case kErrorAudio:
+      if (state_ == kStarting) {
+        event_error_code = constants::kErrorUnableToStart;
+      } else {
+        event_error_code = constants::kErrorCaptureError;
+      }
+      break;
+
+    case kErrorNetwork:
+      event_error_code = constants::kErrorNetworkError;
+      break;
+
+    case kErrorBadGrammar:
+      // No error is returned on invalid language, for example.
+      // To avoid confusion about when this is would be fired, the invalid
+      // params error is not being exposed to the onError event.
+      event_error_code = constants::kErrorUnableToStart;
+      break;
+
+    case kErrorNoSpeech:
+      event_error_code = constants::kErrorNoSpeechHeard;
+      break;
+
+    case kErrorNoMatch:
+      event_error_code = constants::kErrorNoResults;
+      break;
+
+    // The remaining kErrorAborted case should never be returned by the server.
+    default:
+      NOTREACHED();
+  }
+
+  VLOG(1) << "State changed to idle";
+  state_ = kIdle;
+
+  if (!event_error_code.empty()) {
+    BrowserThread::PostTask(BrowserThread::UI, FROM_HERE,
+        base::Bind(&ExtensionSpeechInputManager::DispatchError,
+        this, event_error_code));
+  }
+}
+
+void ExtensionSpeechInputManager::DidCompleteEnvironmentEstimation(
+    int caller_id) {
+  DCHECK_EQ(caller_id, kSpeechCallerId);
+}
+
+void ExtensionSpeechInputManager::DidStartReceivingSpeech(int caller_id) {
+  DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
+  DCHECK_EQ(caller_id, kSpeechCallerId);
+  VLOG(1) << "DidStartReceivingSpeech";
+
+  std::string json_args;
+  BrowserThread::PostTask(BrowserThread::UI, FROM_HERE,
+      base::Bind(&ExtensionSpeechInputManager::DispatchEventToExtension,
+      this, std::string(constants::kOnSoundStartEvent), json_args));
+}
+
+void ExtensionSpeechInputManager::DidStopReceivingSpeech(int caller_id) {
+  DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
+  DCHECK_EQ(caller_id, kSpeechCallerId);
+  VLOG(1) << "DidStopReceivingSpeech";
+
+  std::string json_args;
+  BrowserThread::PostTask(BrowserThread::UI, FROM_HERE,
+      base::Bind(&ExtensionSpeechInputManager::DispatchEventToExtension,
+      this, std::string(constants::kOnSoundEndEvent), json_args));
+}
+
+void ExtensionSpeechInputManager::DispatchEventToExtension(
+    const std::string& event, const std::string& json_args) {
+  DCHECK(BrowserThread::CurrentlyOn(BrowserThread::UI));
+
+  base::AutoLock auto_lock(state_lock_);
+  if (state_ == kShutdown)
+    return;
+
+  if (profile_ && profile_->GetExtensionEventRouter()) {
+    std::string final_args;
+    if (json_args.empty()) {
+      ListValue args;
+      base::JSONWriter::Write(&args, false, &final_args);
+    } else {
+      final_args = json_args;
+    }
+
+    profile_->GetExtensionEventRouter()->DispatchEventToExtension(
+        extension_id_in_use_, event, final_args, profile_, GURL());
+  }
+}
+
+void ExtensionSpeechInputManager::DispatchError(
+    const std::string& error) {
+  DCHECK(BrowserThread::CurrentlyOn(BrowserThread::UI));
+
+  {
+    base::AutoLock auto_lock(state_lock_);
+    if (state_ == kShutdown)
+      return;
+
+    ExtensionError details(extension_id_in_use_, error);
+    NotificationService::current()->Notify(
+        chrome::NOTIFICATION_EXTENSION_SPEECH_INPUT_FAILED,
+        Source<Profile>(profile_),
+        Details<ExtensionError>(&details));
+  }
+
+  ListValue args;
+  DictionaryValue *js_error = new DictionaryValue();
+  args.Append(js_error);
+  js_error->SetString(constants::kErrorCodeKey, error);
+  std::string json_args;
+  base::JSONWriter::Write(&args, false, &json_args);
+  DispatchEventToExtension(constants::kOnErrorEvent, json_args);
+}
+
+bool ExtensionSpeechInputManager::Start(const std::string& extension_id,
+    const std::string& language, const std::string& grammar,
+    bool filter_profanities, std::string* error) {
+  DCHECK(BrowserThread::CurrentlyOn(BrowserThread::UI));
+  DCHECK(error);
+  VLOG(1) << "Requesting start (UI thread)";
+
+  base::AutoLock auto_lock(state_lock_);
+  if (state_ == kShutdown ||
+      (!extension_id_in_use_.empty() && extension_id_in_use_ != extension_id)) {
+    *error = constants::kErrorRequestDenied;
+    return false;
+  }
+
+  if (state_ != kIdle) {
+    *error = constants::kErrorRequestOngoing;

[Satish, 2011/10/11 13:37:16]

'RequestOngoing' - This error message is cryptic. Can we rename it to
'RequestInProgress' or 'AlreadyInUse' ? Same for the string sent to javascript

[Leandro Graciá Gil, 2011/10/11 18:08:00]

Done.

+    return false;
+  }
+
+  extension_id_in_use_ = extension_id;
+  VLOG(1) << "State changed to starting";
+  state_ = kStarting;
+
+  BrowserThread::PostTask(BrowserThread::IO, FROM_HERE,
+      base::Bind(&ExtensionSpeechInputManager::StartOnIOThread, this,
+      profile_->GetRequestContext(), language, grammar, filter_profanities));
+  return true;
+}
+
+void ExtensionSpeechInputManager::StartOnIOThread(
+    net::URLRequestContextGetter* context_getter,
+    const std::string& language, const std::string& grammar,
+    bool filter_profanities) {
+  DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
+  VLOG(1) << "Requesting start (IO thread)";
+
+  // Everything put inside the lock to ensure the validity of context_getter,
+  // guaranteed while not in the shutdown state. Any ongoing or recognition
+  // request will be requested to be aborted when entering the shutdown state.
+  base::AutoLock auto_lock(state_lock_);
+  if (state_ == kShutdown)
+    return;
+
+  if (!AudioManager::GetAudioManager()->HasAudioInputDevices()) {
+    BrowserThread::PostTask(BrowserThread::UI, FROM_HERE,
+        base::Bind(&ExtensionSpeechInputManager::DispatchError, this,
+        std::string(constants::kErrorNoDevicesAvailable)));
+    return;
+  }
+
+  recognizer_ = new SpeechRecognizer(this, kSpeechCallerId, language, grammar,
+      context_getter, filter_profanities, "", "");
+  recognizer_->StartRecording();
+}
+
+bool ExtensionSpeechInputManager::Stop(const std::string& extension_id,
+    std::string* error) {
+  DCHECK(BrowserThread::CurrentlyOn(BrowserThread::UI));
+  DCHECK(error);
+  VLOG(1) << "Requesting stop (UI thread)";
+
+  base::AutoLock auto_lock(state_lock_);
+  if (state_ == kShutdown ||
+      (!extension_id_in_use_.empty() && extension_id_in_use_ != extension_id)) {
+    *error = constants::kErrorRequestDenied;
+    return false;
+  }
+
+  if (state_ != kRecording) {
+    *error = constants::kErrorRequestOngoing;

[Satish, 2011/10/11 13:37:16]

Is this the correct error to return when Stop is called and no session is
active?

[Leandro Graciá Gil, 2011/10/11 18:08:00]

Fixed in later modifications of the code.

On 2011/10/11 13:37:16, Satish wrote:

+    return false;
+  }
+
+  DCHECK(recognizer_);
+
+  VLOG(1) << "State changed to stopping";
+  state_ = kStopping;
+
+  BrowserThread::PostTask(BrowserThread::IO, FROM_HERE,
+      base::Bind(&ExtensionSpeechInputManager::ForceStopOnIOThread, this));
+  return true;
+}
+
+void ExtensionSpeechInputManager::ForceStopOnIOThread() {
+  DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
+  VLOG(1) << "Requesting forced stop (IO thread)";
+
+  base::AutoLock auto_lock(state_lock_);
+  DCHECK(state_ != kIdle);
+
+  if (recognizer_) {
+    recognizer_->CancelRecognition();
+    recognizer_.release();
+  }
+
+  if (state_ == kShutdown)
+    return;
+
+  BrowserThread::PostTask(BrowserThread::UI, FROM_HERE,
+      base::Bind(&ExtensionSpeechInputManager::StopSucceededOnUIThread, this));
+}
+
+void ExtensionSpeechInputManager::StopSucceededOnUIThread() {
+  DCHECK(BrowserThread::CurrentlyOn(BrowserThread::UI));
+  VLOG(1) << "Stop succeeded (UI thread)";
+
+  base::AutoLock auto_lock(state_lock_);
+  if (state_ == kShutdown)
+    return;
+
+  state_ = kIdle;
+  VLOG(1) << "State changed to idle";
+
+  NotificationService::current()->Notify(
+      chrome::NOTIFICATION_EXTENSION_SPEECH_INPUT_RECORDING_STOPPED,
+      Source<Profile>(profile_), // Guarded by the state_ == kShutdown check.
+      Details<std::string>(&extension_id_in_use_));
+
+  extension_id_in_use_.clear();
+}