src/chrome/browser/speech/extension_api/tts_extension_api.cc

   1 // Copyright (c) 2012 The Chromium Authors. All rights reserved.
   2 // Use of this source code is governed by a BSD-style license that can be
   3 // found in the LICENSE file.
   4
   5 #include "chrome/browser/speech/extension_api/tts_extension_api.h"
   6
   7 #include <string>
   8
   9 #include "base/lazy_instance.h"
  10 #include "base/memory/weak_ptr.h"
  11 #include "base/values.h"
  12 #include "chrome/browser/profiles/profile.h"
  13 #include "chrome/browser/speech/extension_api/tts_engine_extension_api.h"
  14 #include "chrome/browser/speech/extension_api/tts_extension_api_constants.h"
  15 #include "chrome/browser/speech/tts_controller.h"
  16 #include "extensions/browser/event_router.h"
  17 #include "extensions/browser/extension_function_registry.h"
  18 #include "ui/base/l10n/l10n_util.h"
  19
  20 namespace constants = tts_extension_api_constants;
  21
  22 namespace events {
  23 const char kOnEvent[] = "tts.onEvent";
  24 };  // namespace events
  25
  26 const char *TtsEventTypeToString(TtsEventType event_type) {
  27   switch (event_type) {
  28     case TTS_EVENT_START:
  29       return constants::kEventTypeStart;
  30     case TTS_EVENT_END:
  31       return constants::kEventTypeEnd;
  32     case TTS_EVENT_WORD:
  33       return constants::kEventTypeWord;
  34     case TTS_EVENT_SENTENCE:
  35       return constants::kEventTypeSentence;
  36     case TTS_EVENT_MARKER:
  37       return constants::kEventTypeMarker;
  38     case TTS_EVENT_INTERRUPTED:
  39       return constants::kEventTypeInterrupted;
  40     case TTS_EVENT_CANCELLED:
  41       return constants::kEventTypeCancelled;
  42     case TTS_EVENT_ERROR:
  43       return constants::kEventTypeError;
  44     case TTS_EVENT_PAUSE:
  45       return constants::kEventTypePause;
  46     case TTS_EVENT_RESUME:
  47       return constants::kEventTypeResume;
  48     default:
  49       NOTREACHED();
  50       return constants::kEventTypeError;
  51   }
  52 }
  53
  54 TtsEventType TtsEventTypeFromString(const std::string& str) {
  55   if (str == constants::kEventTypeStart)
  56     return TTS_EVENT_START;
  57   if (str == constants::kEventTypeEnd)
  58     return TTS_EVENT_END;
  59   if (str == constants::kEventTypeWord)
  60     return TTS_EVENT_WORD;
  61   if (str == constants::kEventTypeSentence)
  62     return TTS_EVENT_SENTENCE;
  63   if (str == constants::kEventTypeMarker)
  64     return TTS_EVENT_MARKER;
  65   if (str == constants::kEventTypeInterrupted)
  66     return TTS_EVENT_INTERRUPTED;
  67   if (str == constants::kEventTypeCancelled)
  68     return TTS_EVENT_CANCELLED;
  69   if (str == constants::kEventTypeError)
  70     return TTS_EVENT_ERROR;
  71   if (str == constants::kEventTypePause)
  72     return TTS_EVENT_PAUSE;
  73   if (str == constants::kEventTypeResume)
  74     return TTS_EVENT_RESUME;
  75
  76   NOTREACHED();
  77   return TTS_EVENT_ERROR;
  78 }
  79
  80 namespace extensions {
  81
  82 // One of these is constructed for each utterance, and deleted
  83 // when the utterance gets any final event.
  84 class TtsExtensionEventHandler
  85     : public UtteranceEventDelegate,
  86       public base::SupportsWeakPtr<TtsExtensionEventHandler> {
  87  public:
  88   virtual void OnTtsEvent(Utterance* utterance,
  89                           TtsEventType event_type,
  90                           int char_index,
  91                           const std::string& error_message) OVERRIDE;
  92 };
  93
  94 void TtsExtensionEventHandler::OnTtsEvent(Utterance* utterance,
  95                                           TtsEventType event_type,
  96                                           int char_index,
  97                                           const std::string& error_message) {
  98   if (utterance->src_id() < 0) {
  99     if (utterance->finished())
 100       delete this;
 101     return;
 102   }
 103
 104   const std::set<TtsEventType>& desired_event_types =
 105       utterance->desired_event_types();
 106   if (desired_event_types.size() > 0 &&
 107       desired_event_types.find(event_type) == desired_event_types.end()) {
 108     if (utterance->finished())
 109       delete this;
 110     return;
 111   }
 112
 113   const char *event_type_string = TtsEventTypeToString(event_type);
 114   scoped_ptr<base::DictionaryValue> details(new base::DictionaryValue());
 115   if (char_index >= 0)
 116     details->SetInteger(constants::kCharIndexKey, char_index);
 117   details->SetString(constants::kEventTypeKey, event_type_string);
 118   if (event_type == TTS_EVENT_ERROR) {
 119     details->SetString(constants::kErrorMessageKey, error_message);
 120   }
 121   details->SetInteger(constants::kSrcIdKey, utterance->src_id());
 122   details->SetBoolean(constants::kIsFinalEventKey, utterance->finished());
 123
 124   scoped_ptr<base::ListValue> arguments(new base::ListValue());
 125   arguments->Set(0, details.release());
 126
 127   scoped_ptr<extensions::Event> event(
 128       new extensions::Event(events::kOnEvent, arguments.Pass()));
 129   event->restrict_to_browser_context = utterance->profile();
 130   event->event_url = utterance->src_url();
 131   extensions::EventRouter::Get(utterance->profile())
 132       ->DispatchEventToExtension(utterance->src_extension_id(), event.Pass());
 133
 134   if (utterance->finished())
 135     delete this;
 136 }
 137
 138 bool TtsSpeakFunction::RunAsync() {
 139   std::string text;
 140   EXTENSION_FUNCTION_VALIDATE(args_->GetString(0, &text));
 141   if (text.size() > 32768) {
 142     error_ = constants::kErrorUtteranceTooLong;
 143     return false;
 144   }
 145
 146   scoped_ptr<base::DictionaryValue> options(new base::DictionaryValue());
 147   if (args_->GetSize() >= 2) {
 148     base::DictionaryValue* temp_options = NULL;
 149     if (args_->GetDictionary(1, &temp_options))
 150       options.reset(temp_options->DeepCopy());
 151   }
 152
 153   std::string voice_name;
 154   if (options->HasKey(constants::kVoiceNameKey)) {
 155     EXTENSION_FUNCTION_VALIDATE(
 156         options->GetString(constants::kVoiceNameKey, &voice_name));
 157   }
 158
 159   std::string lang;
 160   if (options->HasKey(constants::kLangKey))
 161     EXTENSION_FUNCTION_VALIDATE(options->GetString(constants::kLangKey, &lang));
 162   if (!lang.empty() && !l10n_util::IsValidLocaleSyntax(lang)) {
 163     error_ = constants::kErrorInvalidLang;
 164     return false;
 165   }
 166
 167   std::string gender_str;
 168   TtsGenderType gender;
 169   if (options->HasKey(constants::kGenderKey))
 170     EXTENSION_FUNCTION_VALIDATE(
 171         options->GetString(constants::kGenderKey, &gender_str));
 172   if (gender_str == constants::kGenderMale) {
 173     gender = TTS_GENDER_MALE;
 174   } else if (gender_str == constants::kGenderFemale) {
 175     gender = TTS_GENDER_FEMALE;
 176   } else if (gender_str.empty()) {
 177     gender = TTS_GENDER_NONE;
 178   } else {
 179     error_ = constants::kErrorInvalidGender;
 180     return false;
 181   }
 182
 183   double rate = 1.0;
 184   if (options->HasKey(constants::kRateKey)) {
 185     EXTENSION_FUNCTION_VALIDATE(
 186         options->GetDouble(constants::kRateKey, &rate));
 187     if (rate < 0.1 || rate > 10.0) {
 188       error_ = constants::kErrorInvalidRate;
 189       return false;
 190     }
 191   }
 192
 193   double pitch = 1.0;
 194   if (options->HasKey(constants::kPitchKey)) {
 195     EXTENSION_FUNCTION_VALIDATE(
 196         options->GetDouble(constants::kPitchKey, &pitch));
 197     if (pitch < 0.0 || pitch > 2.0) {
 198       error_ = constants::kErrorInvalidPitch;
 199       return false;
 200     }
 201   }
 202
 203   double volume = 1.0;
 204   if (options->HasKey(constants::kVolumeKey)) {
 205     EXTENSION_FUNCTION_VALIDATE(
 206         options->GetDouble(constants::kVolumeKey, &volume));
 207     if (volume < 0.0 || volume > 1.0) {
 208       error_ = constants::kErrorInvalidVolume;
 209       return false;
 210     }
 211   }
 212
 213   bool can_enqueue = false;
 214   if (options->HasKey(constants::kEnqueueKey)) {
 215     EXTENSION_FUNCTION_VALIDATE(
 216         options->GetBoolean(constants::kEnqueueKey, &can_enqueue));
 217   }
 218
 219   std::set<TtsEventType> required_event_types;
 220   if (options->HasKey(constants::kRequiredEventTypesKey)) {
 221     base::ListValue* list;
 222     EXTENSION_FUNCTION_VALIDATE(
 223         options->GetList(constants::kRequiredEventTypesKey, &list));
 224     for (size_t i = 0; i < list->GetSize(); ++i) {
 225       std::string event_type;
 226       if (list->GetString(i, &event_type))
 227         required_event_types.insert(TtsEventTypeFromString(event_type.c_str()));
 228     }
 229   }
 230
 231   std::set<TtsEventType> desired_event_types;
 232   if (options->HasKey(constants::kDesiredEventTypesKey)) {
 233     base::ListValue* list;
 234     EXTENSION_FUNCTION_VALIDATE(
 235         options->GetList(constants::kDesiredEventTypesKey, &list));
 236     for (size_t i = 0; i < list->GetSize(); ++i) {
 237       std::string event_type;
 238       if (list->GetString(i, &event_type))
 239         desired_event_types.insert(TtsEventTypeFromString(event_type.c_str()));
 240     }
 241   }
 242
 243   std::string voice_extension_id;
 244   if (options->HasKey(constants::kExtensionIdKey)) {
 245     EXTENSION_FUNCTION_VALIDATE(
 246         options->GetString(constants::kExtensionIdKey, &voice_extension_id));
 247   }
 248
 249   int src_id = -1;
 250   if (options->HasKey(constants::kSrcIdKey)) {
 251     EXTENSION_FUNCTION_VALIDATE(
 252         options->GetInteger(constants::kSrcIdKey, &src_id));
 253   }
 254
 255   // If we got this far, the arguments were all in the valid format, so
 256   // send the success response to the callback now - this ensures that
 257   // the callback response always arrives before events, which makes
 258   // the behavior more predictable and easier to write unit tests for too.
 259   SendResponse(true);
 260
 261   UtteranceContinuousParameters continuous_params;
 262   continuous_params.rate = rate;
 263   continuous_params.pitch = pitch;
 264   continuous_params.volume = volume;
 265
 266   Utterance* utterance = new Utterance(GetProfile());
 267   utterance->set_text(text);
 268   utterance->set_voice_name(voice_name);
 269   utterance->set_src_extension_id(extension_id());
 270   utterance->set_src_id(src_id);
 271   utterance->set_src_url(source_url());
 272   utterance->set_lang(lang);
 273   utterance->set_gender(gender);
 274   utterance->set_continuous_parameters(continuous_params);
 275   utterance->set_can_enqueue(can_enqueue);
 276   utterance->set_required_event_types(required_event_types);
 277   utterance->set_desired_event_types(desired_event_types);
 278   utterance->set_extension_id(voice_extension_id);
 279   utterance->set_options(options.get());
 280   utterance->set_event_delegate(
 281       (new TtsExtensionEventHandler())->AsWeakPtr());
 282
 283   TtsController* controller = TtsController::GetInstance();
 284   controller->SpeakOrEnqueue(utterance);
 285   return true;
 286 }
 287
 288 bool TtsStopSpeakingFunction::RunSync() {
 289   TtsController::GetInstance()->Stop();
 290   return true;
 291 }
 292
 293 bool TtsPauseFunction::RunSync() {
 294   TtsController::GetInstance()->Pause();
 295   return true;
 296 }
 297
 298 bool TtsResumeFunction::RunSync() {
 299   TtsController::GetInstance()->Resume();
 300   return true;
 301 }
 302
 303 bool TtsIsSpeakingFunction::RunSync() {
 304   SetResult(base::Value::CreateBooleanValue(
 305       TtsController::GetInstance()->IsSpeaking()));
 306   return true;
 307 }
 308
 309 bool TtsGetVoicesFunction::RunSync() {
 310   std::vector<VoiceData> voices;
 311   TtsController::GetInstance()->GetVoices(GetProfile(), &voices);
 312
 313   scoped_ptr<base::ListValue> result_voices(new base::ListValue());
 314   for (size_t i = 0; i < voices.size(); ++i) {
 315     const VoiceData& voice = voices[i];
 316     base::DictionaryValue* result_voice = new base::DictionaryValue();
 317     result_voice->SetString(constants::kVoiceNameKey, voice.name);
 318     result_voice->SetBoolean(constants::kRemoteKey, voice.remote);
 319     if (!voice.lang.empty())
 320       result_voice->SetString(constants::kLangKey, voice.lang);
 321     if (voice.gender == TTS_GENDER_MALE)
 322       result_voice->SetString(constants::kGenderKey, constants::kGenderMale);
 323     else if (voice.gender == TTS_GENDER_FEMALE)
 324       result_voice->SetString(constants::kGenderKey, constants::kGenderFemale);
 325     if (!voice.extension_id.empty())
 326       result_voice->SetString(constants::kExtensionIdKey, voice.extension_id);
 327
 328     base::ListValue* event_types = new base::ListValue();
 329     for (std::set<TtsEventType>::iterator iter = voice.events.begin();
 330          iter != voice.events.end(); ++iter) {
 331       const char* event_name_constant = TtsEventTypeToString(*iter);
 332       event_types->Append(base::Value::CreateStringValue(event_name_constant));
 333     }
 334     result_voice->Set(constants::kEventTypesKey, event_types);
 335
 336     result_voices->Append(result_voice);
 337   }
 338
 339   SetResult(result_voices.release());
 340   return true;
 341 }
 342
 343 TtsAPI::TtsAPI(content::BrowserContext* context) {
 344   ExtensionFunctionRegistry* registry =
 345       ExtensionFunctionRegistry::GetInstance();
 346   registry->RegisterFunction<ExtensionTtsEngineSendTtsEventFunction>();
 347   registry->RegisterFunction<TtsGetVoicesFunction>();
 348   registry->RegisterFunction<TtsIsSpeakingFunction>();
 349   registry->RegisterFunction<TtsSpeakFunction>();
 350   registry->RegisterFunction<TtsStopSpeakingFunction>();
 351   registry->RegisterFunction<TtsPauseFunction>();
 352   registry->RegisterFunction<TtsResumeFunction>();
 353 }
 354
 355 TtsAPI::~TtsAPI() {
 356 }
 357
 358 static base::LazyInstance<BrowserContextKeyedAPIFactory<TtsAPI> > g_factory =
 359     LAZY_INSTANCE_INITIALIZER;
 360
 361 BrowserContextKeyedAPIFactory<TtsAPI>* TtsAPI::GetFactoryInstance() {
 362   return g_factory.Pointer();
 363 }
 364
 365 }  // namespace extensions