Add support for speak properties such as rate to platform specific TTS.

BUG=none.
TEST=Use extension to drive the tts engines on Windows and Mac.  Had key bindings so that verification of:
  decrease rate, increase rate, decrease pitch, increase pitch, decrease volume, and increase volume worked.  Verified that the limits, when reached, did not have any adverse effects.
Review URL: http://codereview.chromium.org/3325021

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@59572 0039d316-1c4b-4281-b951-d872f2087c98

7 files changed, 287 insertions, 173 deletions

diff --git a/chrome/browser/extensions/extension_tts_api.cc b/chrome/browser/extensions/extension_tts_api.cc
index 47168af..ba9c2f8 100644
--- a/chrome/browser/extensions/extension_tts_api.cc
+++ b/chrome/browser/extensions/extension_tts_api.cc
@@ -12,43 +12,13 @@
 #include "chrome/browser/chromeos/cros/cros_library.h"
 #include "chrome/browser/chromeos/cros/speech_synthesis_library.h"
 
-using base::DoubleToString;
+namespace util = extension_tts_api_util;
 
-const char kNameKey[] = "name";
-const char kLanguageNameKey[] = "languageName";
-const char kGenderKey[] = "gender";
-const char kRateKey[] = "rate";
-const char kPitchKey[] = "pitch";
-const char kVolumeKey[] = "volume";
-const char kEqualStr[] = "=";
-const char kDelimiter[] = ";";
+using base::DoubleToString;
 
 namespace {
   const char kCrosLibraryNotLoadedError[] =
       "Cros shared library not loaded.";
-
-  bool ReadNumberByKey(DictionaryValue* dict, const char* key,
-      double* ret_value) {
-    Value* value;
-    dict->Get(key, &value);
-    if (value->IsType(Value::TYPE_INTEGER)) {
-      int int_value;
-      if (!dict->GetInteger(key, &int_value))
-        return false;
-      *ret_value = int_value;
-    } else if (value->IsType(Value::TYPE_REAL)) {
-      if (!dict->GetReal(key, ret_value))
-        return false;
-    } else {
-      return false;
-    }
-    return true;
-  }
-
-  void AppendSpeakOption(std::string key, std::string value,
-      std::string* options) {
-    *options += key + kEqualStr + value + kDelimiter;
-  }
 };
 
 bool ExtensionTtsSpeakFunction::RunImpl() {
@@ -59,26 +29,31 @@ bool ExtensionTtsSpeakFunction::RunImpl() {
   if (args_->GetDictionary(1, &speak_options)) {
     std::string str_value;
     double real_value;
-    if (speak_options->HasKey(kLanguageNameKey) &&
-        speak_options->GetString(kLanguageNameKey, &str_value)) {
-      AppendSpeakOption(std::string(kNameKey), str_value, &options);
+    if (speak_options->HasKey(util::kLanguageNameKey) &&
+      speak_options->GetString(util::kLanguageNameKey, &str_value)) {
+        util::AppendSpeakOption(
+            std::string(util::kNameKey), str_value, &options);
     }
-    if (speak_options->HasKey(kGenderKey) &&
-        speak_options->GetString(kGenderKey, &str_value)) {
-      AppendSpeakOption(std::string(kGenderKey), str_value, &options);
+    if (speak_options->HasKey(util::kGenderKey) &&
+      speak_options->GetString(util::kGenderKey, &str_value)) {
+        util::AppendSpeakOption(
+            std::string(util::kGenderKey), str_value, &options);
     }
-    if (ReadNumberByKey(speak_options, kRateKey, &real_value))
+    if (util::ReadNumberByKey(speak_options, util::kRateKey, &real_value)) {
       // The TTS service allows a range of 0 to 5 for speech rate.
-      AppendSpeakOption(std::string(kRateKey),
-                        DoubleToString(real_value * 5), &options);
-    if (ReadNumberByKey(speak_options, kPitchKey, &real_value))
+      util::AppendSpeakOption(std::string(util::kRateKey),
+          DoubleToString(real_value * 5), &options);
+    }
+    if (util::ReadNumberByKey(speak_options, util::kPitchKey, &real_value)) {
       // The TTS service allows a range of 0 to 2 for speech pitch.
-      AppendSpeakOption(std::string(kPitchKey),
-                        DoubleToString(real_value * 2), &options);
-    if (ReadNumberByKey(speak_options, kVolumeKey, &real_value))
+      util::AppendSpeakOption(std::string(util::kPitchKey),
+          DoubleToString(real_value * 2), &options);
+    }
+    if (util::ReadNumberByKey(speak_options, util::kVolumeKey, &real_value)) {
       // The TTS service allows a range of 0 to 5 for speech volume.
-      AppendSpeakOption(std::string(kVolumeKey),
-                        DoubleToString(real_value * 5), &options);
+      util::AppendSpeakOption(std::string(util::kVolumeKey),
+          DoubleToString(real_value * 5), &options);
+    }
   }
   if (chromeos::CrosLibrary::Get()->EnsureLoaded()) {
     if (!options.empty()) {
diff --git a/chrome/browser/extensions/extension_tts_api.h b/chrome/browser/extensions/extension_tts_api.h
index d5f578d..84fa384 100644
--- a/chrome/browser/extensions/extension_tts_api.h
+++ b/chrome/browser/extensions/extension_tts_api.h
@@ -5,8 +5,8 @@
 #ifndef CHROME_BROWSER_EXTENSIONS_EXTENSION_TTS_API_H_
 #define CHROME_BROWSER_EXTENSIONS_EXTENSION_TTS_API_H_
 
-#include "base/singleton.h"
 #include "chrome/browser/extensions/extension_function.h"
+#include "chrome/browser/extensions/extension_tts_api_util.h"
 
 class ExtensionTtsSpeakFunction : public SyncExtensionFunction {
   ~ExtensionTtsSpeakFunction() {}
diff --git a/chrome/browser/extensions/extension_tts_api_mac.mm b/chrome/browser/extensions/extension_tts_api_mac.mm
index 98e49e3..1b45380 100644
--- a/chrome/browser/extensions/extension_tts_api_mac.mm
+++ b/chrome/browser/extensions/extension_tts_api_mac.mm
@@ -1,39 +1,69 @@
-// Copyright (c) 2010 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 "extension_tts_api.h"

-

-#include <string>

-

-#include "base/values.h"

-#include "chrome/browser/extensions/extension_function.h"

-

-#import <cocoa/cocoa.h>

-

-static NSSpeechSynthesizer* speech_synthesizer_;

-

-void InitializeSpeechSynthesizer() {

-  if (!speech_synthesizer_)

-    speech_synthesizer_ = [[NSSpeechSynthesizer alloc] init];

-}

-

-bool ExtensionTtsSpeakFunction::RunImpl() {

-  InitializeSpeechSynthesizer();

-  std::string utterance;

-  EXTENSION_FUNCTION_VALIDATE(args_->GetString(0, &utterance));

-  return

-      [speech_synthesizer_ startSpeakingString:

-           [NSString stringWithUTF8String: utterance.c_str()]];

-}

-

-bool ExtensionTtsStopSpeakingFunction::RunImpl() {

-    InitializeSpeechSynthesizer();

-  [speech_synthesizer_ stopSpeaking];

-  return true;

-}

-

-bool ExtensionTtsIsSpeakingFunction::RunImpl() {

-    InitializeSpeechSynthesizer();

-  return [speech_synthesizer_ isSpeaking];

-}

+// Copyright (c) 2010 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 "extension_tts_api.h"
+
+#include <string>
+
+#include "base/values.h"
+#include "chrome/browser/extensions/extension_function.h"
+
+#import <Cocoa/Cocoa.h>
+
+namespace util = extension_tts_api_util;
+
+static NSSpeechSynthesizer* speech_synthesizer_;
+
+void InitializeSpeechSynthesizer() {
+  if (!speech_synthesizer_)
+    speech_synthesizer_ = [[NSSpeechSynthesizer alloc] init];
+}
+
+bool ExtensionTtsSpeakFunction::RunImpl() {
+  InitializeSpeechSynthesizer();
+  std::string utterance;
+  EXTENSION_FUNCTION_VALIDATE(args_->GetString(0, &utterance));
+  DictionaryValue* speak_options = NULL;
+
+  // Parse speech properties.
+  if (args_->GetDictionary(1, &speak_options)) {
+    std::string str_value;
+    double real_value;
+    // NSSpeechSynthesizer equivalents for kGenderKey and kLanguageNameKey do
+    // not exist and thus are not supported.
+    if (util::ReadNumberByKey(speak_options, util::kRateKey, &real_value)) {
+      // The TTS api defines rate via words per minute.
+      [speech_synthesizer_
+          setObject:[NSNumber numberWithInt:real_value*400]
+          forProperty:NSSpeechRateProperty error:nil];
+    }
+    if (util::ReadNumberByKey(speak_options, util::kPitchKey, &real_value)) {
+      // The TTS api allows an approximate range of 30 to 65 for speech pitch.
+      [speech_synthesizer_
+          setObject: [NSNumber numberWithInt:(real_value*35 + 30)]
+          forProperty:NSSpeechPitchBaseProperty error:nil];
+    }
+    if (util::ReadNumberByKey(speak_options, util::kVolumeKey, &real_value)) {
+      // The TTS api allows a range of 0.0 to 1.0 for speech volume.
+      [speech_synthesizer_
+          setObject: [NSNumber numberWithFloat:real_value]
+          forProperty:NSSpeechVolumeProperty error:nil];
+    }
+  }
+
+  return
+      [speech_synthesizer_ startSpeakingString:
+           [NSString stringWithUTF8String: utterance.c_str()]];
+}
+
+bool ExtensionTtsStopSpeakingFunction::RunImpl() {
+    InitializeSpeechSynthesizer();
+  [speech_synthesizer_ stopSpeaking];
+  return true;
+}
+
+bool ExtensionTtsIsSpeakingFunction::RunImpl() {
+    InitializeSpeechSynthesizer();
+  return [speech_synthesizer_ isSpeaking];
+}
diff --git a/chrome/browser/extensions/extension_tts_api_util.cc b/chrome/browser/extensions/extension_tts_api_util.cc
new file mode 100644
index 0000000..ac5d53b
--- /dev/null
+++ b/chrome/browser/extensions/extension_tts_api_util.cc
@@ -0,0 +1,38 @@
+// Copyright (c) 2010 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_tts_api_util.h"
+
+namespace extension_tts_api_util {
+
+// Static.
+bool ReadNumberByKey(DictionaryValue* dict,
+                     const char* key,
+                     double* ret_value) {
+  Value* value;
+  if (!dict->Get(key, &value))
+    return false;
+
+  if (value->IsType(Value::TYPE_INTEGER)) {
+    int int_value;
+    if (!dict->GetInteger(key, &int_value))
+      return false;
+    *ret_value = int_value;
+  } else if (value->IsType(Value::TYPE_REAL)) {
+    if (!dict->GetReal(key, ret_value))
+      return false;
+  } else {
+    return false;
+  }
+  return true;
+}
+
+// Static.
+void AppendSpeakOption(std::string key,
+                       std::string value,
+                       std::string* options) {
+  *options += key + kEqualStr + value + kDelimiter;
+}
+
+}  // namespace extension_tts_api_util.
diff --git a/chrome/browser/extensions/extension_tts_api_util.h b/chrome/browser/extensions/extension_tts_api_util.h
new file mode 100644
index 0000000..4b41871
--- /dev/null
+++ b/chrome/browser/extensions/extension_tts_api_util.h
@@ -0,0 +1,32 @@
+// Copyright (c) 2010 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.
+
+#ifndef CHROME_BROWSER_EXTENSIONS_EXTENSION_TTS_API_UTIL_H_
+#define CHROME_BROWSER_EXTENSIONS_EXTENSION_TTS_API_UTIL_H_
+
+#include <string>
+
+#include "base/values.h"
+
+namespace extension_tts_api_util {
+
+const char kNameKey[] = "name";
+const char kLanguageNameKey[] = "languageName";
+const char kGenderKey[] = "gender";
+const char kRateKey[] = "rate";
+const char kPitchKey[] = "pitch";
+const char kVolumeKey[] = "volume";
+const char kEqualStr[] = "=";
+const char kDelimiter[] = ";";
+
+bool ReadNumberByKey(DictionaryValue* dict,
+                     const char* key,
+                     double* ret_value);
+
+void AppendSpeakOption(std::string key,
+                       std::string value,
+                       std::string* options);
+
+}  // namespace extension_tts_api_util.
+#endif  // CHROME_BROWSER_EXTENSIONS_EXTENSION_TTS_API_UTIL_H_
diff --git a/chrome/browser/extensions/extension_tts_api_win.cc b/chrome/browser/extensions/extension_tts_api_win.cc
index 260d047..480880b 100644
--- a/chrome/browser/extensions/extension_tts_api_win.cc
+++ b/chrome/browser/extensions/extension_tts_api_win.cc
@@ -1,86 +1,123 @@
-// Copyright (c) 2010 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 "extension_tts_api.h"

-

-#include <atlbase.h>

-#include <atlcom.h>

-#include <sapi.h>

-

-#include "base/scoped_comptr_win.h"

-#include "base/singleton.h"

-#include "base/values.h"

-

-class SpeechSynthesizerWrapper {

- public:

-  SpeechSynthesizerWrapper() : speech_synthesizer_(NULL),

-                               paused_(false),

-                               permanent_failure_(false) {

-    InitializeSpeechSynthesizer();

-  }

-

-  bool InitializeSpeechSynthesizer() {

-    if (!SUCCEEDED(CoCreateInstance(CLSID_SpVoice, NULL, CLSCTX_SERVER,

-        IID_ISpVoice, reinterpret_cast<void**>(&speech_synthesizer_)))) {

-      permanent_failure_ = true;

-      return false;

-    }

-

-    if (paused_)

-      speech_synthesizer_->Resume();

-    return true;

-  }

-

-  ScopedComPtr<ISpVoice> speech_synthesizer() {

-    return speech_synthesizer_;

-  }

-

-  bool paused() {

-    return paused_;

-  }

-

-  void paused(bool state) {

-    paused_ = state;

-  }

-

- private:

-  ScopedComPtr<ISpVoice> speech_synthesizer_;

-  bool paused_;

-  // Indicates an error retrieving the SAPI COM interface.

-  bool permanent_failure_;

-};

-

-typedef Singleton<SpeechSynthesizerWrapper> SpeechSynthesizerSingleton;

-

-bool ExtensionTtsSpeakFunction::RunImpl() {

-  ScopedComPtr<ISpVoice> speech_synthesizer =

-      SpeechSynthesizerSingleton::get()->speech_synthesizer();

-  if (speech_synthesizer) {

-    std::wstring utterance;

-    EXTENSION_FUNCTION_VALIDATE(args_->GetString(0, &utterance));

-    if (SpeechSynthesizerSingleton::get()->paused())

-      speech_synthesizer->Resume();

-    speech_synthesizer->Speak(

-        utterance.c_str(), SPF_ASYNC | SPF_PURGEBEFORESPEAK, NULL);

-    return true;

-  }

-

-  return false;

-}

-

-bool ExtensionTtsStopSpeakingFunction::RunImpl() {

-  // We need to keep track of the paused state since SAPI doesn't have a stop

-  // method.

-  ScopedComPtr<ISpVoice> speech_synthesizer =

-      SpeechSynthesizerSingleton::get()->speech_synthesizer();

-  if (speech_synthesizer && !SpeechSynthesizerSingleton::get()->paused()) {

-    speech_synthesizer->Pause();

-    SpeechSynthesizerSingleton::get()->paused(true);

-  }

-  return true;

-}

-

-bool ExtensionTtsIsSpeakingFunction::RunImpl() {

-  return false;

-}

+// Copyright (c) 2010 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_tts_api.h"
+
+#include <atlbase.h>
+#include <atlcom.h>
+#include <sapi.h>
+
+#include "base/scoped_comptr_win.h"
+#include "base/singleton.h"
+#include "base/string_number_conversions.h"
+#include "base/values.h"
+
+namespace util = extension_tts_api_util;
+
+class SpeechSynthesizerWrapper {
+ public:
+  SpeechSynthesizerWrapper() : speech_synthesizer_(NULL),
+                               paused_(false),
+                               permanent_failure_(false) {
+    InitializeSpeechSynthesizer();
+  }
+
+  bool InitializeSpeechSynthesizer() {
+    if (!SUCCEEDED(CoCreateInstance(CLSID_SpVoice,
+                                    NULL,
+                                    CLSCTX_SERVER,
+                                    IID_ISpVoice,
+                                    reinterpret_cast<void**>(
+                                        &speech_synthesizer_)))) {
+      permanent_failure_ = true;
+      return false;
+    }
+
+    if (paused_)
+      speech_synthesizer_->Resume();
+    return true;
+  }
+
+  ScopedComPtr<ISpVoice> speech_synthesizer() {
+    return speech_synthesizer_;
+  }
+
+  bool paused() {
+    return paused_;
+  }
+
+  void paused(bool state) {
+    paused_ = state;
+  }
+
+ private:
+  ScopedComPtr<ISpVoice> speech_synthesizer_;
+  bool paused_;
+  // Indicates an error retrieving the SAPI COM interface.
+  bool permanent_failure_;
+};
+
+typedef Singleton<SpeechSynthesizerWrapper> SpeechSynthesizerSingleton;
+
+bool ExtensionTtsSpeakFunction::RunImpl() {
+  ScopedComPtr<ISpVoice> speech_synthesizer =
+      SpeechSynthesizerSingleton::get()->speech_synthesizer();
+  if (speech_synthesizer) {
+    std::wstring utterance;
+    EXTENSION_FUNCTION_VALIDATE(args_->GetString(0, &utterance));
+
+    std::string options = "";
+    DictionaryValue* speak_options = NULL;
+
+    // Parse speech properties.
+    if (args_->GetDictionary(1, &speak_options)) {
+      std::string str_value;
+      double real_value;
+      // Speech API equivalents for kGenderKey and kLanguageNameKey do not
+      // exist and thus are not supported.
+      if (util::ReadNumberByKey(speak_options, util::kRateKey, &real_value)) {
+        // The TTS api allows a range of -10 to 10 for speech rate.
+        speech_synthesizer->SetRate(static_cast<int32>(real_value*20 - 10));
+      }
+      if (util::ReadNumberByKey(speak_options, util::kPitchKey, &real_value)) {
+        // The TTS api allows a range of -10 to 10 for speech pitch.
+        // TODO(dtseng): cleanup if we ever
+        // use any other properties that require xml.
+        std::wstring pitch_value =
+            base::IntToString16(static_cast<int>(real_value*20 - 10));
+        utterance = L"<pitch absmiddle=\"" + pitch_value + L"\">" +
+            utterance + L"</pitch>";
+      }
+      if (util::ReadNumberByKey(
+          speak_options, util::kVolumeKey, &real_value)) {
+        // The TTS api allows a range of 0 to 100 for speech volume.
+        speech_synthesizer->SetVolume(static_cast<uint16>(real_value * 100));
+      }
+    }
+
+    if (SpeechSynthesizerSingleton::get()->paused())
+      speech_synthesizer->Resume();
+    speech_synthesizer->Speak(
+        utterance.c_str(), SPF_ASYNC | SPF_PURGEBEFORESPEAK, NULL);
+    return true;
+  }
+
+  return false;
+}
+
+bool ExtensionTtsStopSpeakingFunction::RunImpl() {
+  // We need to keep track of the paused state since SAPI doesn't have a stop
+  // method.
+  ScopedComPtr<ISpVoice> speech_synthesizer =
+      SpeechSynthesizerSingleton::get()->speech_synthesizer();
+  if (speech_synthesizer && !SpeechSynthesizerSingleton::get()->paused()) {
+    speech_synthesizer->Pause();
+    SpeechSynthesizerSingleton::get()->paused(true);
+  }
+  return true;
+}
+
+bool ExtensionTtsIsSpeakingFunction::RunImpl() {
+  return false;
+}
diff --git a/chrome/chrome_browser.gypi b/chrome/chrome_browser.gypi
index e7f007b..5032f92 100644
--- a/chrome/chrome_browser.gypi
+++ b/chrome/chrome_browser.gypi
@@ -1468,6 +1468,8 @@
         'browser/extensions/extension_tts_api.h',
         'browser/extensions/extension_tts_api_gtk.cc',
         'browser/extensions/extension_tts_api_mac.mm',
+        'browser/extensions/extension_tts_api_util.cc',
+        'browser/extensions/extension_tts_api_util.h',
         'browser/extensions/extension_tts_api_win.cc',
         'browser/extensions/extension_toolbar_model.cc',
         'browser/extensions/extension_toolbar_model.h',