src/chrome/browser/speech/tts_controller_impl.cc

   1 // Copyright 2014 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/tts_controller_impl.h"
   6
   7 #include <string>
   8 #include <vector>
   9
  10 #include "base/float_util.h"
  11 #include "base/values.h"
  12 #include "chrome/browser/browser_process.h"
  13 #include "chrome/browser/speech/tts_platform.h"
  14
  15 namespace {
  16 // A value to be used to indicate that there is no char index available.
  17 const int kInvalidCharIndex = -1;
  18
  19 // Given a language/region code of the form 'fr-FR', returns just the basic
  20 // language portion, e.g. 'fr'.
  21 std::string TrimLanguageCode(std::string lang) {
  22   if (lang.size() >= 5 && lang[2] == '-')
  23     return lang.substr(0, 2);
  24   else
  25     return lang;
  26 }
  27
  28 }  // namespace
  29
  30 bool IsFinalTtsEventType(TtsEventType event_type) {
  31   return (event_type == TTS_EVENT_END ||
  32           event_type == TTS_EVENT_INTERRUPTED ||
  33           event_type == TTS_EVENT_CANCELLED ||
  34           event_type == TTS_EVENT_ERROR);
  35 }
  36
  37 //
  38 // UtteranceContinuousParameters
  39 //
  40
  41
  42 UtteranceContinuousParameters::UtteranceContinuousParameters()
  43     : rate(-1),
  44       pitch(-1),
  45       volume(-1) {}
  46
  47
  48 //
  49 // VoiceData
  50 //
  51
  52
  53 VoiceData::VoiceData()
  54     : gender(TTS_GENDER_NONE),
  55       remote(false),
  56       native(false) {}
  57
  58 VoiceData::~VoiceData() {}
  59
  60
  61 //
  62 // Utterance
  63 //
  64
  65 // static
  66 int Utterance::next_utterance_id_ = 0;
  67
  68 Utterance::Utterance(content::BrowserContext* browser_context)
  69     : browser_context_(browser_context),
  70       id_(next_utterance_id_++),
  71       src_id_(-1),
  72       gender_(TTS_GENDER_NONE),
  73       can_enqueue_(false),
  74       char_index_(0),
  75       finished_(false) {
  76   options_.reset(new base::DictionaryValue());
  77 }
  78
  79 Utterance::~Utterance() {
  80   DCHECK(finished_);
  81 }
  82
  83 void Utterance::OnTtsEvent(TtsEventType event_type,
  84                            int char_index,
  85                            const std::string& error_message) {
  86   if (char_index >= 0)
  87     char_index_ = char_index;
  88   if (IsFinalTtsEventType(event_type))
  89     finished_ = true;
  90
  91   if (event_delegate_)
  92     event_delegate_->OnTtsEvent(this, event_type, char_index, error_message);
  93   if (finished_)
  94     event_delegate_.reset();
  95 }
  96
  97 void Utterance::Finish() {
  98   finished_ = true;
  99 }
 100
 101 void Utterance::set_options(const base::Value* options) {
 102   options_.reset(options->DeepCopy());
 103 }
 104
 105 TtsController* TtsController::GetInstance() {
 106   return TtsControllerImpl::GetInstance();
 107 }
 108
 109 //
 110 // TtsControllerImpl
 111 //
 112
 113 // static
 114 TtsControllerImpl* TtsControllerImpl::GetInstance() {
 115   return Singleton<TtsControllerImpl>::get();
 116 }
 117
 118 TtsControllerImpl::TtsControllerImpl()
 119     : current_utterance_(NULL),
 120       paused_(false),
 121       platform_impl_(NULL),
 122       tts_engine_delegate_(NULL) {
 123 }
 124
 125 TtsControllerImpl::~TtsControllerImpl() {
 126   if (current_utterance_) {
 127     current_utterance_->Finish();
 128     delete current_utterance_;
 129   }
 130
 131   // Clear any queued utterances too.
 132   ClearUtteranceQueue(false);  // Don't sent events.
 133 }
 134
 135 void TtsControllerImpl::SpeakOrEnqueue(Utterance* utterance) {
 136   // If we're paused and we get an utterance that can't be queued,
 137   // flush the queue but stay in the paused state.
 138   if (paused_ && !utterance->can_enqueue()) {
 139     Stop();
 140     paused_ = true;
 141     return;
 142   }
 143
 144   if (paused_ || (IsSpeaking() && utterance->can_enqueue())) {
 145     utterance_queue_.push(utterance);
 146   } else {
 147     Stop();
 148     SpeakNow(utterance);
 149   }
 150 }
 151
 152 void TtsControllerImpl::SpeakNow(Utterance* utterance) {
 153   // Ensure we have all built-in voices loaded. This is a no-op if already
 154   // loaded.
 155   bool loaded_built_in =
 156       GetPlatformImpl()->LoadBuiltInTtsExtension(utterance->browser_context());
 157
 158   // Get all available voices and try to find a matching voice.
 159   std::vector<VoiceData> voices;
 160   GetVoices(utterance->browser_context(), &voices);
 161   int index = GetMatchingVoice(utterance, voices);
 162
 163   VoiceData voice;
 164   if (index != -1) {
 165     // Select the matching voice.
 166     voice = voices[index];
 167   } else {
 168     // However, if no match was found on a platform without native tts voices,
 169     // attempt to get a voice based only on the current locale without respect
 170     // to any supplied voice names.
 171     std::vector<VoiceData> native_voices;
 172
 173     if (GetPlatformImpl()->PlatformImplAvailable())
 174       GetPlatformImpl()->GetVoices(&native_voices);
 175
 176     if (native_voices.empty() && !voices.empty()) {
 177       // TODO(dtseng): Notify extension caller of an error.
 178       utterance->set_voice_name("");
 179       // TODO(gaochun): Replace the global variable g_browser_process with
 180       // GetContentClient()->browser() to eliminate the dependency of browser
 181       // once TTS implementation was moved to content.
 182       utterance->set_lang(g_browser_process->GetApplicationLocale());
 183       index = GetMatchingVoice(utterance, voices);
 184
 185       // If even that fails, just take the first available voice.
 186       if (index == -1)
 187         index = 0;
 188       voice = voices[index];
 189     } else {
 190       // Otherwise, simply give native voices a chance to handle this utterance.
 191       voice.native = true;
 192     }
 193   }
 194
 195   GetPlatformImpl()->WillSpeakUtteranceWithVoice(utterance, voice);
 196
 197   if (!voice.native) {
 198 #if !defined(OS_ANDROID)
 199     DCHECK(!voice.extension_id.empty());
 200     current_utterance_ = utterance;
 201     utterance->set_extension_id(voice.extension_id);
 202     if (tts_engine_delegate_)
 203       tts_engine_delegate_->Speak(utterance, voice);
 204     bool sends_end_event =
 205         voice.events.find(TTS_EVENT_END) != voice.events.end();
 206     if (!sends_end_event) {
 207       utterance->Finish();
 208       delete utterance;
 209       current_utterance_ = NULL;
 210       SpeakNextUtterance();
 211     }
 212 #endif
 213   } else {
 214     // It's possible for certain platforms to send start events immediately
 215     // during |speak|.
 216     current_utterance_ = utterance;
 217     GetPlatformImpl()->clear_error();
 218     bool success = GetPlatformImpl()->Speak(
 219         utterance->id(),
 220         utterance->text(),
 221         utterance->lang(),
 222         voice,
 223         utterance->continuous_parameters());
 224     if (!success)
 225       current_utterance_ = NULL;
 226
 227     // If the native voice wasn't able to process this speech, see if
 228     // the browser has built-in TTS that isn't loaded yet.
 229     if (!success && loaded_built_in) {
 230       utterance_queue_.push(utterance);
 231       return;
 232     }
 233
 234     if (!success) {
 235       utterance->OnTtsEvent(TTS_EVENT_ERROR, kInvalidCharIndex,
 236                             GetPlatformImpl()->error());
 237       delete utterance;
 238       return;
 239     }
 240   }
 241 }
 242
 243 void TtsControllerImpl::Stop() {
 244   paused_ = false;
 245   if (current_utterance_ && !current_utterance_->extension_id().empty()) {
 246 #if !defined(OS_ANDROID)
 247     if (tts_engine_delegate_)
 248       tts_engine_delegate_->Stop(current_utterance_);
 249 #endif
 250   } else {
 251     GetPlatformImpl()->clear_error();
 252     GetPlatformImpl()->StopSpeaking();
 253   }
 254
 255   if (current_utterance_)
 256     current_utterance_->OnTtsEvent(TTS_EVENT_INTERRUPTED, kInvalidCharIndex,
 257                                    std::string());
 258   FinishCurrentUtterance();
 259   ClearUtteranceQueue(true);  // Send events.
 260 }
 261
 262 void TtsControllerImpl::Pause() {
 263   paused_ = true;
 264   if (current_utterance_ && !current_utterance_->extension_id().empty()) {
 265 #if !defined(OS_ANDROID)
 266     if (tts_engine_delegate_)
 267       tts_engine_delegate_->Pause(current_utterance_);
 268 #endif
 269   } else if (current_utterance_) {
 270     GetPlatformImpl()->clear_error();
 271     GetPlatformImpl()->Pause();
 272   }
 273 }
 274
 275 void TtsControllerImpl::Resume() {
 276   paused_ = false;
 277   if (current_utterance_ && !current_utterance_->extension_id().empty()) {
 278 #if !defined(OS_ANDROID)
 279     if (tts_engine_delegate_)
 280       tts_engine_delegate_->Resume(current_utterance_);
 281 #endif
 282   } else if (current_utterance_) {
 283     GetPlatformImpl()->clear_error();
 284     GetPlatformImpl()->Resume();
 285   } else {
 286     SpeakNextUtterance();
 287   }
 288 }
 289
 290 void TtsControllerImpl::OnTtsEvent(int utterance_id,
 291                                         TtsEventType event_type,
 292                                         int char_index,
 293                                         const std::string& error_message) {
 294   // We may sometimes receive completion callbacks "late", after we've
 295   // already finished the utterance (for example because another utterance
 296   // interrupted or we got a call to Stop). This is normal and we can
 297   // safely just ignore these events.
 298   if (!current_utterance_ || utterance_id != current_utterance_->id()) {
 299     return;
 300   }
 301   current_utterance_->OnTtsEvent(event_type, char_index, error_message);
 302   if (current_utterance_->finished()) {
 303     FinishCurrentUtterance();
 304     SpeakNextUtterance();
 305   }
 306 }
 307
 308 void TtsControllerImpl::GetVoices(content::BrowserContext* browser_context,
 309                               std::vector<VoiceData>* out_voices) {
 310 #if !defined(OS_ANDROID)
 311   if (browser_context && tts_engine_delegate_)
 312     tts_engine_delegate_->GetVoices(browser_context, out_voices);
 313 #endif
 314
 315   TtsPlatformImpl* platform_impl = GetPlatformImpl();
 316   if (platform_impl) {
 317     // Ensure we have all built-in voices loaded. This is a no-op if already
 318     // loaded.
 319     platform_impl->LoadBuiltInTtsExtension(browser_context);
 320     if (platform_impl->PlatformImplAvailable())
 321       platform_impl->GetVoices(out_voices);
 322   }
 323 }
 324
 325 bool TtsControllerImpl::IsSpeaking() {
 326   return current_utterance_ != NULL || GetPlatformImpl()->IsSpeaking();
 327 }
 328
 329 void TtsControllerImpl::FinishCurrentUtterance() {
 330   if (current_utterance_) {
 331     if (!current_utterance_->finished())
 332       current_utterance_->OnTtsEvent(TTS_EVENT_INTERRUPTED, kInvalidCharIndex,
 333                                      std::string());
 334     delete current_utterance_;
 335     current_utterance_ = NULL;
 336   }
 337 }
 338
 339 void TtsControllerImpl::SpeakNextUtterance() {
 340   if (paused_)
 341     return;
 342
 343   // Start speaking the next utterance in the queue.  Keep trying in case
 344   // one fails but there are still more in the queue to try.
 345   while (!utterance_queue_.empty() && !current_utterance_) {
 346     Utterance* utterance = utterance_queue_.front();
 347     utterance_queue_.pop();
 348     SpeakNow(utterance);
 349   }
 350 }
 351
 352 void TtsControllerImpl::ClearUtteranceQueue(bool send_events) {
 353   while (!utterance_queue_.empty()) {
 354     Utterance* utterance = utterance_queue_.front();
 355     utterance_queue_.pop();
 356     if (send_events)
 357       utterance->OnTtsEvent(TTS_EVENT_CANCELLED, kInvalidCharIndex,
 358                             std::string());
 359     else
 360       utterance->Finish();
 361     delete utterance;
 362   }
 363 }
 364
 365 void TtsControllerImpl::SetPlatformImpl(
 366     TtsPlatformImpl* platform_impl) {
 367   platform_impl_ = platform_impl;
 368 }
 369
 370 int TtsControllerImpl::QueueSize() {
 371   return static_cast<int>(utterance_queue_.size());
 372 }
 373
 374 TtsPlatformImpl* TtsControllerImpl::GetPlatformImpl() {
 375   if (!platform_impl_)
 376     platform_impl_ = TtsPlatformImpl::GetInstance();
 377   return platform_impl_;
 378 }
 379
 380 int TtsControllerImpl::GetMatchingVoice(
 381     const Utterance* utterance, std::vector<VoiceData>& voices) {
 382   // Make two passes: the first time, do strict language matching
 383   // ('fr-FR' does not match 'fr-CA'). The second time, do prefix
 384   // language matching ('fr-FR' matches 'fr' and 'fr-CA')
 385   for (int pass = 0; pass < 2; ++pass) {
 386     for (size_t i = 0; i < voices.size(); ++i) {
 387       const VoiceData& voice = voices[i];
 388
 389       if (!utterance->extension_id().empty() &&
 390           utterance->extension_id() != voice.extension_id) {
 391         continue;
 392       }
 393
 394       if (!voice.name.empty() &&
 395           !utterance->voice_name().empty() &&
 396           voice.name != utterance->voice_name()) {
 397         continue;
 398       }
 399       if (!voice.lang.empty() && !utterance->lang().empty()) {
 400         std::string voice_lang = voice.lang;
 401         std::string utterance_lang = utterance->lang();
 402         if (pass == 1) {
 403           voice_lang = TrimLanguageCode(voice_lang);
 404           utterance_lang = TrimLanguageCode(utterance_lang);
 405         }
 406         if (voice_lang != utterance_lang) {
 407           continue;
 408         }
 409       }
 410       if (voice.gender != TTS_GENDER_NONE &&
 411           utterance->gender() != TTS_GENDER_NONE &&
 412           voice.gender != utterance->gender()) {
 413         continue;
 414       }
 415
 416       if (utterance->required_event_types().size() > 0) {
 417         bool has_all_required_event_types = true;
 418         for (std::set<TtsEventType>::const_iterator iter =
 419                  utterance->required_event_types().begin();
 420              iter != utterance->required_event_types().end();
 421              ++iter) {
 422           if (voice.events.find(*iter) == voice.events.end()) {
 423             has_all_required_event_types = false;
 424             break;
 425           }
 426         }
 427         if (!has_all_required_event_types)
 428           continue;
 429       }
 430
 431       return static_cast<int>(i);
 432     }
 433   }
 434
 435   return -1;
 436 }
 437
 438 void TtsControllerImpl::VoicesChanged() {
 439   for (std::set<VoicesChangedDelegate*>::iterator iter =
 440            voices_changed_delegates_.begin();
 441        iter != voices_changed_delegates_.end(); ++iter) {
 442     (*iter)->OnVoicesChanged();
 443   }
 444 }
 445
 446 void TtsControllerImpl::AddVoicesChangedDelegate(
 447     VoicesChangedDelegate* delegate) {
 448   voices_changed_delegates_.insert(delegate);
 449 }
 450
 451 void TtsControllerImpl::RemoveVoicesChangedDelegate(
 452     VoicesChangedDelegate* delegate) {
 453   voices_changed_delegates_.erase(delegate);
 454 }
 455
 456 void TtsControllerImpl::SetTtsEngineDelegate(
 457     TtsEngineDelegate* delegate) {
 458   tts_engine_delegate_ = delegate;
 459 }
 460
 461 TtsEngineDelegate* TtsControllerImpl::GetTtsEngineDelegate() {
 462   return tts_engine_delegate_;
 463 }