Speech JavaScript API: Use the MediaStreamManager to ask for user permission.

content/browser/speech/speech_recognition_manager_impl.h

 // Copyright (c) 2012 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 CONTENT_BROWSER_SPEECH_SPEECH_RECOGNITION_MANAGER_IMPL_H_
 #define CONTENT_BROWSER_SPEECH_SPEECH_RECOGNITION_MANAGER_IMPL_H_
 
 #include <map>
 #include <string>
 
 #include "base/basictypes.h"
 #include "base/callback.h"
 #include "base/compiler_specific.h"
 #include "base/memory/weak_ptr.h"
+#include "content/browser/renderer_host/media/media_stream_requester.h"
 #include "content/public/browser/speech_recognition_event_listener.h"
 #include "content/public/browser/speech_recognition_manager.h"
 #include "content/public/browser/speech_recognition_session_config.h"
 #include "content/public/browser/speech_recognition_session_context.h"
 #include "content/public/common/speech_recognition_error.h"
 
 namespace content {
 class BrowserMainLoop;
 class SpeechRecognitionManagerDelegate;
 }
 
+namespace media_stream {
+class MediaStreamManager;
+}
+
 namespace speech {
 
 class SpeechRecognizer;
 
 // This is the manager for speech recognition. It is a single instance in
 // the browser process and can serve several requests. Each recognition request
 // corresponds to a session, initiated via |CreateSession|.
-// In every moment the manager has at most one session capturing audio, which
-// is identified by |session_id_capturing_audio_|. However, multiple sessions
-// can live in parallel in respect of the aforementioned constraint, i.e. while
-// waiting for results.
-// This class does not handle user interface objects (bubbles, tray icon).
-// Those are managed by the delegate, which receives a copy of all sessions
-// events.
+//
+// In any moment, the manager has a single session known as the primary session,
+// |primary_session_id_|.
+// This is the session that is capturing audio, waiting for user permission,
+// etc. There may also be other, non-primary, sessions living in parallel that
+// are waiting for results but not recording audio.
 //
 // The SpeechRecognitionManager has the following responsibilities:
 //  - Handles requests received from various render views and makes sure only
 //    one of them accesses the audio device at any given time.
 //  - Handles the instantiation of SpeechRecognitionEngine objects when
 //    requested by SpeechRecognitionSessions.
 //  - Relays recognition results/status/error events of each session to the
 //    corresponding listener (demuxing on the base of their session_id).
 //  - Relays also recognition results/status/error events of every session to
 //    the catch-all snoop listener (optionally) provided by the delegate.
@@ skipping 66 matching lines @@
     EVENT_RECOGNITION_ENDED,
     EVENT_MAX_VALUE = EVENT_RECOGNITION_ENDED
   };
 
   struct Session {
     Session();
     ~Session();
 
     int id;
     bool listener_is_active;
+    bool is_waiting_for_permission;

[Satish, 2012/09/12 06:15:09]

instead of this flag in each Session object, how about storing the session_id in
the PermissionRequest class and check with the other session_id value when
necessary? That would reduce a few lines of code

[hans, 2012/09/12 09:46:58]

Done.

     content::SpeechRecognitionSessionConfig config;
     content::SpeechRecognitionSessionContext context;
     scoped_refptr<SpeechRecognizer> recognizer;
   };
 
   // Callback issued by the SpeechRecognitionManagerDelegate for reporting
   // asynchronously the result of the CheckRecognitionIsAllowed call.
-  void RecognitionAllowedCallback(int session_id, bool is_allowed);
+  void RecognitionAllowedCallback(int session_id,
+                                  bool ask_user,
+                                  bool is_allowed);
 
   // Entry point for pushing any external event into the session handling FSM.
   void DispatchEvent(int session_id, FSMEvent event);
 
   // Defines the behavior of the session handling FSM, selecting the appropriate
   // transition according to the session, its current state and the event.
   void ExecuteTransitionAndGetNextState(
       const Session& session, FSMState session_state, FSMEvent event);
 
   // Retrieves the state of the session, enquiring directly the recognizer.
   FSMState GetSessionState(int session_id) const;
 
   // The methods below handle transitions of the session handling FSM.
   void SessionStart(const Session& session);
   void SessionAbort(const Session& session);
   void SessionStopAudioCapture(const Session& session);
   void ResetCapturingSessionId(const Session& session);
   void SessionDelete(const Session& session);
   void NotFeasible(const Session& session, FSMEvent event);
 
   bool SessionExists(int session_id) const;
   const Session& GetSession(int session_id) const;
   content::SpeechRecognitionEventListener* GetListener(int session_id) const;
   content::SpeechRecognitionEventListener* GetDelegateListener() const;
   int GetNextSessionID();
 
   typedef std::map<int, Session> SessionsTable;
   SessionsTable sessions_;
-  int session_id_capturing_audio_;
+  int primary_session_id_;
   int last_session_id_;
   bool is_dispatching_event_;
   scoped_ptr<content::SpeechRecognitionManagerDelegate> delegate_;
 
   // Used for posting asynchronous tasks (on the IO thread) without worrying
   // about this class being destroyed in the meanwhile (due to browser shutdown)
   // since tasks pending on a destroyed WeakPtr are automatically discarded.
   base::WeakPtrFactory<SpeechRecognitionManagerImpl> weak_factory_;
+
+  class PermissionRequest : public media_stream::MediaStreamRequester {

[Satish, 2012/09/12 06:15:09]

can this be forward declared in this header and the actual definition be moved
to the .cc file?

[hans, 2012/09/12 09:46:58]

No, because it inherits from another class it can't be forward declared. I did
try, though :)

[hans, 2012/09/12 09:46:58]

Done.

+   public:
+    PermissionRequest(media_stream::MediaStreamManager* media_stream_manager,
+                      const base::Callback<void(bool is_allowed)>& callback);
+    virtual ~PermissionRequest();
+
+    void Start(int render_process_id, int render_view_id, const GURL& origin);
+    void Abort();
+
+    // MediaStreamRequester methods.
+    virtual void StreamGenerated(
+        const std::string& label,
+        const media_stream::StreamDeviceInfoArray& audio_devices,
+        const media_stream::StreamDeviceInfoArray& video_devices) OVERRIDE;
+    virtual void StreamGenerationFailed(const std::string& label) OVERRIDE;
+
+    // The callbacks below are ignored.
+    virtual void AudioDeviceFailed(const std::string& label,
+                                   int index) OVERRIDE {}
+    virtual void VideoDeviceFailed(const std::string& label,
+                                   int index) OVERRIDE {}
+    virtual void DevicesEnumerated(
+        const std::string& label,
+        const media_stream::StreamDeviceInfoArray& devices) OVERRIDE {}
+    virtual void DeviceOpened(
+        const std::string& label,
+        const media_stream::StreamDeviceInfo& device_info) OVERRIDE {}
+
+   private:
+    media_stream::MediaStreamManager* media_stream_manager_;
+    base::Callback<void(bool is_allowed)> callback_;
+    std::string label_;
+  };
+  scoped_ptr<PermissionRequest> permission_request_;
 };
 
 }  // namespace speech
 
 #endif  // CONTENT_BROWSER_SPEECH_SPEECH_RECOGNITION_MANAGER_IMPL_H_