Initial support for native TTS on linux through tts extension API.

chrome/browser/speech/extension_api/tts_extension_api_linux.cc

Index: chrome/browser/speech/extension_api/tts_extension_api_linux.cc
diff --git a/chrome/browser/speech/extension_api/tts_extension_api_linux.cc b/chrome/browser/speech/extension_api/tts_extension_api_linux.cc
index f289f50edc706e46002d5b9a240b4774a0cc2bc9..b592f806a062eeddc79f2c7f27dae167f0894924 100644
--- a/chrome/browser/speech/extension_api/tts_extension_api_linux.cc
+++ b/chrome/browser/speech/extension_api/tts_extension_api_linux.cc
@@ -2,18 +2,164 @@
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
+#include <dlfcn.h>
+#include <libspeechd.h>
+#include <math.h>
+
 #include "base/memory/singleton.h"
 #include "chrome/browser/speech/extension_api/tts_extension_api_platform.h"
 
 namespace {
 const char kNotSupportedError[] =
     "Native speech synthesis not supported on this platform.";
+
+// Speech dispatcher exports.
+typedef   SPDConnection* (*spd_open_func)(const char* client_name,

[dmazzoni, 2012/09/09 21:40:18]

Extra spaces

+    const char* connection_name,
+    const char* user_name,
+    SPDConnectionMode mode);
+typedef int (*spd_say_func)(SPDConnection* connection,
+                            SPDPriority priority,
+                            const char* text);
+typedef void (*spd_stop_func)(SPDConnection* connection);
+typedef void (*spd_close_func)(SPDConnection* connection);
+typedef int (*spd_set_notification_on_func)(SPDConnection* connection,
+                                       SPDNotification notification);

[dmazzoni, 2012/09/09 21:40:18]

Fix indentation

+typedef int (*spd_set_voice_rate_func)(SPDConnection* connection, int rate);
+typedef int (*spd_set_voice_pitch_func)(SPDConnection* connection, int pitch);
+};
+
+class SpeechDispatcherWrapper {
+ public:
+  static SPDNotificationType current_notification;
+
+  SpeechDispatcherWrapper() : loaded_(false) {

[dmazzoni, 2012/09/09 21:40:18]

This is long, so move the implementation below rather than inlined in the class
definition.

+    library_ = dlopen("libspeechd.so", RTLD_LAZY);
+    if (!library_)
+      return;
+
+    spd_open = reinterpret_cast<spd_open_func>(dlsym(library_, "spd_open"));
+    if (!spd_open)
+      return;
+
+    spd_say = reinterpret_cast<spd_say_func>(dlsym(library_, "spd_say"));
+    if (!spd_say)
+      return;
+
+    spd_stop = reinterpret_cast<spd_stop_func>(dlsym(library_, "spd_stop"));
+    if (!spd_stop)
+      return;
+
+    spd_close = reinterpret_cast<spd_close_func>(dlsym(library_, "spd_close"));
+    if (!spd_close)
+      return;
+
+    conn_ = spd_open("chrome", "extension_api", NULL, SPD_MODE_THREADED);
+    if (!conn_)
+      return;
+
+    spd_set_notification_on = reinterpret_cast<spd_set_notification_on_func>(
+        dlsym(library_, "spd_set_notification_on"));
+    if (!spd_set_notification_on)
+      return;
+
+    spd_set_voice_rate = reinterpret_cast<spd_set_voice_rate_func>(
+        dlsym(library_, "spd_set_voice_rate"));
+    if (!spd_set_voice_rate)
+      return;
+
+    spd_set_voice_pitch = reinterpret_cast<spd_set_voice_pitch_func>(
+        dlsym(library_, "spd_set_voice_pitch"));
+    if (!spd_set_voice_pitch)
+      return;
+
+    // Register callbacks for all events.
+    conn_->callback_begin =
+        conn_->callback_end =
+        conn_->callback_cancel =
+        conn_->callback_pause =
+        conn_->callback_resume =
+        &SpeechDispatcherWrapper::NotificationCallback;
+
+    spd_set_notification_on(conn_, SPD_BEGIN);
+    spd_set_notification_on(conn_, SPD_END);
+    spd_set_notification_on(conn_, SPD_CANCEL);
+    spd_set_notification_on(conn_, SPD_PAUSE);
+    spd_set_notification_on(conn_, SPD_RESUME);
+
+    loaded_ = true;
+  }
+
+  ~SpeechDispatcherWrapper() {
+    if (conn_) {
+      spd_close(conn_);
+      conn_ = NULL;
+    }
+
+    if (library_) {
+      dlclose(library_);
+      library_ = NULL;
+    }
+  }
+
+  bool Speak(const char* text) {
+    if (!loaded())
+      return false;
+
+    spd_say(conn_, SPD_TEXT, text);
+    return true;
+  }
+
+  bool IsSpeaking() {
+    return SpeechDispatcherWrapper::current_notification == SPD_EVENT_BEGIN;
+  }
+
+  bool StopSpeaking() {
+    if (!loaded())
+      return false;
+    spd_stop(conn_);
+    return true;
+  }
+
+  void SetRate(int rate) {
+    spd_set_voice_rate(conn_, rate);
+  }
+
+  void SetPitch(int pitch) {
+    spd_set_voice_pitch(conn_, pitch);
+  }
+
+  // States whether Speech Dispatcher loaded successfully.
+  bool loaded() {
+    return loaded_;
+  }
+
+ private:
+  static void NotificationCallback(size_t msg_id,
+                                   size_t client_id,
+                                   SPDNotificationType type);
+
+  // Interface bindings.
+  spd_open_func spd_open;
+  spd_say_func spd_say;
+  spd_stop_func spd_stop;
+  spd_close_func spd_close;
+  spd_set_notification_on_func spd_set_notification_on;
+  spd_set_voice_rate_func spd_set_voice_rate;
+  spd_set_voice_pitch_func spd_set_voice_pitch;
+
+  bool loaded_;
+  void* library_;
+  SPDConnection* conn_;
+  DISALLOW_COPY_AND_ASSIGN(SpeechDispatcherWrapper);
 };
+SPDNotificationType SpeechDispatcherWrapper::current_notification =

[dmazzoni, 2012/09/09 21:40:18]

Add a newline before this, and add a comment that says "static" just before it.
Also, it should end in an underscore just like other member variables.

+    SPD_EVENT_END;
 
 class ExtensionTtsPlatformImplLinux : public ExtensionTtsPlatformImpl {
  public:
   virtual bool PlatformImplAvailable() {
-    return false;
+    return spd_.loaded();
   }
 
   virtual bool Speak(
@@ -21,22 +167,33 @@ class ExtensionTtsPlatformImplLinux : public ExtensionTtsPlatformImpl {
       const std::string& utterance,
       const std::string& lang,
       const UtteranceContinuousParameters& params) {
-    error_ = kNotSupportedError;
-    return false;
+    if (!spd_.loaded()) {
+      error_ = kNotSupportedError;
+      return false;
+    }
+
+    if (params.rate >= 0.0 && params.rate <= 10.0)
+      spd_.SetRate(100 * log10(params.rate));

[dmazzoni, 2012/09/09 21:40:18]

Documentation for how you came up with this mapping?

It looks symmetrical, so that's good - but is there a reason why a rate of 2.0x
gets mapped to +30 and 0.5 gets mapped to -30? Do those roughly correspond to
twice and half the speaking rate of normal?

Try installing the TTS Debug app from the web store, it does some timing
measurements to make sure that your speech engine is calibrated in the right
ballpark.

+    if (params.pitch >= 0.0 && params.pitch <= 10.0)
+      spd_.SetPitch(100 * log10(params.pitch));
+
+    utterance_ = utterance;
+    utterance_id_ = utterance_id;
+
+    spd_.Speak(utterance.c_str());
+    return true;
   }
 
   virtual bool StopSpeaking() {
-    error_ = kNotSupportedError;
-    return false;
+    return spd_.StopSpeaking();
   }
 
   virtual bool IsSpeaking() {
-    error_ = kNotSupportedError;
-    return false;
+    return spd_.IsSpeaking();
   }
 
   virtual bool SendsEvent(TtsEventType event_type) {
-    return false;
+    return (event_type == TTS_EVENT_START || event_type == TTS_EVENT_END);
   }
 
   // Get the single instance of this class.
@@ -44,10 +201,36 @@ class ExtensionTtsPlatformImplLinux : public ExtensionTtsPlatformImpl {
     return Singleton<ExtensionTtsPlatformImplLinux>::get();
   }
 
+  void OnSpeechEvent(SPDNotificationType type) {
+    ExtensionTtsController* controller = ExtensionTtsController::GetInstance();
+    switch (type) {
+      case SPD_EVENT_BEGIN:
+        controller->OnTtsEvent(
+            utterance_id_, TTS_EVENT_START, 0, std::string());
+        break;
+      case SPD_EVENT_END:
+        controller->OnTtsEvent(
+            utterance_id_, TTS_EVENT_END, utterance_.size(), std::string());
+        break;
+      // TODO(dtseng): Unimplemented.

[dmazzoni, 2012/09/09 21:40:18]

Can you implement these as part of this CL? They shouldn't be too bad, and then
it'd be complete...

+      case SPD_EVENT_INDEX_MARK:
+      case SPD_EVENT_CANCEL:
+      case SPD_EVENT_PAUSE:
+      case SPD_EVENT_RESUME:
+        break;
+    }
+  }
+
  private:
-  ExtensionTtsPlatformImplLinux() {}
+  ExtensionTtsPlatformImplLinux(): utterance_id_(0) {}
   virtual ~ExtensionTtsPlatformImplLinux() {}
 
+  SpeechDispatcherWrapper spd_;
+
+  // These apply to the current utterance only.
+  std::string utterance_;
+  int utterance_id_;
+
   friend struct DefaultSingletonTraits<ExtensionTtsPlatformImplLinux>;
 
   DISALLOW_COPY_AND_ASSIGN(ExtensionTtsPlatformImplLinux);
@@ -57,3 +240,10 @@ class ExtensionTtsPlatformImplLinux : public ExtensionTtsPlatformImpl {
 ExtensionTtsPlatformImpl* ExtensionTtsPlatformImpl::GetInstance() {
   return ExtensionTtsPlatformImplLinux::GetInstance();
 }
+
+// static
+void SpeechDispatcherWrapper::NotificationCallback(
+    size_t msg_id, size_t client_id, SPDNotificationType type) {
+  SpeechDispatcherWrapper::current_notification = type;
+  ExtensionTtsPlatformImplLinux::GetInstance()->OnSpeechEvent(type);
+}