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/values.h"
  11 #include "chrome/browser/profiles/profile.h"
  12 #include "chrome/browser/speech/extension_api/tts_engine_extension_api.h"
  13 #include "chrome/browser/speech/extension_api/tts_engine_extension_observer.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:
  87   explicit TtsExtensionEventHandler(const std::string& src_extension_id);
  88
  89   void OnTtsEvent(Utterance* utterance,
  90                   TtsEventType event_type,
  91                   int char_index,
  92                   const std::string& error_message) override;
  93
  94  private:
  95   // The extension ID of the extension that called speak() and should
  96   // receive events.
  97   std::string src_extension_id_;
  98 };
  99
 100 TtsExtensionEventHandler::TtsExtensionEventHandler(
 101     const std::string& src_extension_id)
 102     : src_extension_id_(src_extension_id) {
 103 }
 104
 105 void TtsExtensionEventHandler::OnTtsEvent(Utterance* utterance,
 106                                           TtsEventType event_type,
 107                                           int char_index,
 108                                           const std::string& error_message) {
 109   if (utterance->src_id() < 0) {
 110     if (utterance->finished())
 111       delete this;
 112     return;
 113   }
 114
 115   const std::set<TtsEventType>& desired_event_types =
 116       utterance->desired_event_types();
 117   if (desired_event_types.size() > 0 &&
 118       desired_event_types.find(event_type) == desired_event_types.end()) {
 119     if (utterance->finished())
 120       delete this;
 121     return;
 122   }
 123
 124   const char *event_type_string = TtsEventTypeToString(event_type);
 125   scoped_ptr<base::DictionaryValue> details(new base::DictionaryValue());
 126   if (char_index >= 0)
 127     details->SetInteger(constants::kCharIndexKey, char_index);
 128   details->SetString(constants::kEventTypeKey, event_type_string);
 129   if (event_type == TTS_EVENT_ERROR) {
 130     details->SetString(constants::kErrorMessageKey, error_message);
 131   }
 132   details->SetInteger(constants::kSrcIdKey, utterance->src_id());
 133   details->SetBoolean(constants::kIsFinalEventKey, utterance->finished());
 134
 135   scoped_ptr<base::ListValue> arguments(new base::ListValue());
 136   arguments->Set(0, details.release());
 137
 138   scoped_ptr<extensions::Event> event(
 139       new extensions::Event(::extensions::events::TTS_ON_EVENT,
 140                             ::events::kOnEvent, arguments.Pass()));
 141   event->restrict_to_browser_context = utterance->browser_context();
 142   event->event_url = utterance->src_url();
 143   extensions::EventRouter::Get(utterance->browser_context())
 144       ->DispatchEventToExtension(src_extension_id_, event.Pass());
 145
 146   if (utterance->finished())
 147     delete this;
 148 }
 149
 150 bool TtsSpeakFunction::RunAsync() {
 151   std::string text;
 152   EXTENSION_FUNCTION_VALIDATE(args_->GetString(0, &text));
 153   if (text.size() > 32768) {
 154     error_ = constants::kErrorUtteranceTooLong;
 155     return false;
 156   }
 157
 158   scoped_ptr<base::DictionaryValue> options(new base::DictionaryValue());
 159   if (args_->GetSize() >= 2) {
 160     base::DictionaryValue* temp_options = NULL;
 161     if (args_->GetDictionary(1, &temp_options))
 162       options.reset(temp_options->DeepCopy());
 163   }
 164
 165   std::string voice_name;
 166   if (options->HasKey(constants::kVoiceNameKey)) {
 167     EXTENSION_FUNCTION_VALIDATE(
 168         options->GetString(constants::kVoiceNameKey, &voice_name));
 169   }
 170
 171   std::string lang;
 172   if (options->HasKey(constants::kLangKey))
 173     EXTENSION_FUNCTION_VALIDATE(options->GetString(constants::kLangKey, &lang));
 174   if (!lang.empty() && !l10n_util::IsValidLocaleSyntax(lang)) {
 175     error_ = constants::kErrorInvalidLang;
 176     return false;
 177   }
 178
 179   std::string gender_str;
 180   TtsGenderType gender;
 181   if (options->HasKey(constants::kGenderKey))
 182     EXTENSION_FUNCTION_VALIDATE(
 183         options->GetString(constants::kGenderKey, &gender_str));
 184   if (gender_str == constants::kGenderMale) {
 185     gender = TTS_GENDER_MALE;
 186   } else if (gender_str == constants::kGenderFemale) {
 187     gender = TTS_GENDER_FEMALE;
 188   } else if (gender_str.empty()) {
 189     gender = TTS_GENDER_NONE;
 190   } else {
 191     error_ = constants::kErrorInvalidGender;
 192     return false;
 193   }
 194
 195   double rate = 1.0;
 196   if (options->HasKey(constants::kRateKey)) {
 197     EXTENSION_FUNCTION_VALIDATE(
 198         options->GetDouble(constants::kRateKey, &rate));
 199     if (rate < 0.1 || rate > 10.0) {
 200       error_ = constants::kErrorInvalidRate;
 201       return false;
 202     }
 203   }
 204
 205   double pitch = 1.0;
 206   if (options->HasKey(constants::kPitchKey)) {
 207     EXTENSION_FUNCTION_VALIDATE(
 208         options->GetDouble(constants::kPitchKey, &pitch));
 209     if (pitch < 0.0 || pitch > 2.0) {
 210       error_ = constants::kErrorInvalidPitch;
 211       return false;
 212     }
 213   }
 214
 215   double volume = 1.0;
 216   if (options->HasKey(constants::kVolumeKey)) {
 217     EXTENSION_FUNCTION_VALIDATE(
 218         options->GetDouble(constants::kVolumeKey, &volume));
 219     if (volume < 0.0 || volume > 1.0) {
 220       error_ = constants::kErrorInvalidVolume;
 221       return false;
 222     }
 223   }
 224
 225   bool can_enqueue = false;
 226   if (options->HasKey(constants::kEnqueueKey)) {
 227     EXTENSION_FUNCTION_VALIDATE(
 228         options->GetBoolean(constants::kEnqueueKey, &can_enqueue));
 229   }
 230
 231   std::set<TtsEventType> required_event_types;
 232   if (options->HasKey(constants::kRequiredEventTypesKey)) {
 233     base::ListValue* list;
 234     EXTENSION_FUNCTION_VALIDATE(
 235         options->GetList(constants::kRequiredEventTypesKey, &list));
 236     for (size_t i = 0; i < list->GetSize(); ++i) {
 237       std::string event_type;
 238       if (list->GetString(i, &event_type))
 239         required_event_types.insert(TtsEventTypeFromString(event_type.c_str()));
 240     }
 241   }
 242
 243   std::set<TtsEventType> desired_event_types;
 244   if (options->HasKey(constants::kDesiredEventTypesKey)) {
 245     base::ListValue* list;
 246     EXTENSION_FUNCTION_VALIDATE(
 247         options->GetList(constants::kDesiredEventTypesKey, &list));
 248     for (size_t i = 0; i < list->GetSize(); ++i) {
 249       std::string event_type;
 250       if (list->GetString(i, &event_type))
 251         desired_event_types.insert(TtsEventTypeFromString(event_type.c_str()));
 252     }
 253   }
 254
 255   std::string voice_extension_id;
 256   if (options->HasKey(constants::kExtensionIdKey)) {
 257     EXTENSION_FUNCTION_VALIDATE(
 258         options->GetString(constants::kExtensionIdKey, &voice_extension_id));
 259   }
 260
 261   int src_id = -1;
 262   if (options->HasKey(constants::kSrcIdKey)) {
 263     EXTENSION_FUNCTION_VALIDATE(
 264         options->GetInteger(constants::kSrcIdKey, &src_id));
 265   }
 266
 267   // If we got this far, the arguments were all in the valid format, so
 268   // send the success response to the callback now - this ensures that
 269   // the callback response always arrives before events, which makes
 270   // the behavior more predictable and easier to write unit tests for too.
 271   SendResponse(true);
 272
 273   Utterance* utterance = new Utterance(GetProfile());
 274   utterance->set_text(text);
 275   utterance->set_voice_name(voice_name);
 276   utterance->set_src_id(src_id);
 277   utterance->set_src_url(source_url());
 278   utterance->set_lang(lang);
 279   utterance->set_gender(gender);
 280   utterance->set_continuous_parameters(rate, pitch, volume);
 281   utterance->set_can_enqueue(can_enqueue);
 282   utterance->set_required_event_types(required_event_types);
 283   utterance->set_desired_event_types(desired_event_types);
 284   utterance->set_extension_id(voice_extension_id);
 285   utterance->set_options(options.get());
 286   utterance->set_event_delegate(new TtsExtensionEventHandler(extension_id()));
 287
 288   TtsController* controller = TtsController::GetInstance();
 289   controller->SpeakOrEnqueue(utterance);
 290   return true;
 291 }
 292
 293 bool TtsStopSpeakingFunction::RunSync() {
 294   TtsController::GetInstance()->Stop();
 295   return true;
 296 }
 297
 298 bool TtsPauseFunction::RunSync() {
 299   TtsController::GetInstance()->Pause();
 300   return true;
 301 }
 302
 303 bool TtsResumeFunction::RunSync() {
 304   TtsController::GetInstance()->Resume();
 305   return true;
 306 }
 307
 308 bool TtsIsSpeakingFunction::RunSync() {
 309   SetResult(
 310       new base::FundamentalValue(TtsController::GetInstance()->IsSpeaking()));
 311   return true;
 312 }
 313
 314 bool TtsGetVoicesFunction::RunSync() {
 315   std::vector<VoiceData> voices;
 316   TtsController::GetInstance()->GetVoices(GetProfile(), &voices);
 317
 318   scoped_ptr<base::ListValue> result_voices(new base::ListValue());
 319   for (size_t i = 0; i < voices.size(); ++i) {
 320     const VoiceData& voice = voices[i];
 321     base::DictionaryValue* result_voice = new base::DictionaryValue();
 322     result_voice->SetString(constants::kVoiceNameKey, voice.name);
 323     result_voice->SetBoolean(constants::kRemoteKey, voice.remote);
 324     if (!voice.lang.empty())
 325       result_voice->SetString(constants::kLangKey, voice.lang);
 326     if (voice.gender == TTS_GENDER_MALE)
 327       result_voice->SetString(constants::kGenderKey, constants::kGenderMale);
 328     else if (voice.gender == TTS_GENDER_FEMALE)
 329       result_voice->SetString(constants::kGenderKey, constants::kGenderFemale);
 330     if (!voice.extension_id.empty())
 331       result_voice->SetString(constants::kExtensionIdKey, voice.extension_id);
 332
 333     base::ListValue* event_types = new base::ListValue();
 334     for (std::set<TtsEventType>::iterator iter = voice.events.begin();
 335          iter != voice.events.end(); ++iter) {
 336       const char* event_name_constant = TtsEventTypeToString(*iter);
 337       event_types->Append(new base::StringValue(event_name_constant));
 338     }
 339     result_voice->Set(constants::kEventTypesKey, event_types);
 340
 341     result_voices->Append(result_voice);
 342   }
 343
 344   SetResult(result_voices.release());
 345   return true;
 346 }
 347
 348 TtsAPI::TtsAPI(content::BrowserContext* context) {
 349   ExtensionFunctionRegistry* registry =
 350       ExtensionFunctionRegistry::GetInstance();
 351   registry->RegisterFunction<ExtensionTtsEngineSendTtsEventFunction>();
 352   registry->RegisterFunction<TtsGetVoicesFunction>();
 353   registry->RegisterFunction<TtsIsSpeakingFunction>();
 354   registry->RegisterFunction<TtsSpeakFunction>();
 355   registry->RegisterFunction<TtsStopSpeakingFunction>();
 356   registry->RegisterFunction<TtsPauseFunction>();
 357   registry->RegisterFunction<TtsResumeFunction>();
 358
 359   // Ensure we're observing newly added engines for the given context.
 360   TtsEngineExtensionObserver::GetInstance(Profile::FromBrowserContext(context));
 361 }
 362
 363 TtsAPI::~TtsAPI() {
 364 }
 365
 366 static base::LazyInstance<BrowserContextKeyedAPIFactory<TtsAPI> > g_factory =
 367     LAZY_INSTANCE_INITIALIZER;
 368
 369 BrowserContextKeyedAPIFactory<TtsAPI>* TtsAPI::GetFactoryInstance() {
 370   return g_factory.Pointer();
 371 }
 372
 373 }  // namespace extensions