Fix some spelling mistakes (mostly in comments).
[dcpomatic.git] / src / lib / video_decoder.cc
index 508ed90b71c71f62e874f0a763b4b19f2d9db8d7..cf21f885a3fbc6af45ba2efde1cfc18ad645eb9b 100644 (file)
@@ -1,5 +1,5 @@
 /*
-    Copyright (C) 2012-2018 Carl Hetherington <cth@carlh.net>
+    Copyright (C) 2012-2021 Carl Hetherington <cth@carlh.net>
 
     This file is part of DCP-o-matic.
 
 
 */
 
-#include "video_decoder.h"
-#include "image.h"
-#include "raw_image_proxy.h"
+
+#include "compose.hpp"
 #include "film.h"
+#include "frame_interval_checker.h"
+#include "image.h"
+#include "j2k_image_proxy.h"
 #include "log.h"
-#include "compose.hpp"
-#include <boost/foreach.hpp>
+#include "raw_image_proxy.h"
+#include "video_decoder.h"
 #include <iostream>
 
 #include "i18n.h"
 
+
 using std::cout;
-using std::list;
-using std::max;
-using std::back_inserter;
-using boost::shared_ptr;
-using boost::optional;
+using std::dynamic_pointer_cast;
+using std::shared_ptr;
 using namespace dcpomatic;
 
+
 VideoDecoder::VideoDecoder (Decoder* parent, shared_ptr<const Content> c)
        : DecoderPart (parent)
        , _content (c)
+       , _frame_interval_checker (new FrameIntervalChecker())
 {
 
 }
 
+
 /** Called by decoder classes when they have a video frame ready.
  *  @param frame Frame index within the content; this does not take into account 3D
  *  so for 3D_ALTERNATE this value goes:
@@ -60,98 +63,116 @@ VideoDecoder::emit (shared_ptr<const Film> film, shared_ptr<const ImageProxy> im
                return;
        }
 
-       /* Before we `re-write' the frame indexes of these incoming data we need to check for
-          the case where the user has some 2D content which they have marked as 3D.  With 3D
-          we should get two frames for each frame index, but in this `bad' case we only get
-          one.  We need to throw an exception if this happens.
-       */
-
-       if (_content->video->frame_type() == VIDEO_FRAME_TYPE_3D) {
-               if (_last_threed_frames.size() > 4) {
-                       _last_threed_frames.erase (_last_threed_frames.begin());
-               }
-               _last_threed_frames.push_back (decoder_frame);
-               if (_last_threed_frames.size() == 4) {
-                       if (_last_threed_frames[0] != _last_threed_frames[1] || _last_threed_frames[2] != _last_threed_frames[3]) {
-                               boost::throw_exception (
-                                       DecodeError(
-                                               String::compose(
-                                                       _("The content file %1 is set as 3D but does not appear to contain 3D images.  Please set it to 2D.  "
-                                                         "You can still make a 3D DCP from this content by ticking the 3D option in the DCP video tab."),
-                                                       _content->path(0)
-                                                       )
+       auto const afr = _content->active_video_frame_rate(film);
+       auto const vft = _content->video->frame_type();
+
+       auto frame_time = ContentTime::from_frames (decoder_frame, afr);
+
+       /* Do some heuristics to try and spot the case where the user sets content to 3D
+        * when it is not.  We try to tell this by looking at the differences in time between
+        * the first few frames.  Real 3D content should have two frames for each timestamp.
+        */
+       if (_frame_interval_checker) {
+               _frame_interval_checker->feed (frame_time, afr);
+               if (_frame_interval_checker->guess() == FrameIntervalChecker::PROBABLY_NOT_3D && vft == VideoFrameType::THREE_D) {
+                       boost::throw_exception (
+                               DecodeError(
+                                       String::compose(
+                                               _("The content file %1 is set as 3D but does not appear to contain 3D images.  Please set it to 2D.  "
+                                                 "You can still make a 3D DCP from this content by ticking the 3D option in the DCP video tab."),
+                                               _content->path(0)
                                                )
-                                       );
-                       }
+                                       )
+                               );
                }
-       }
 
-       double const afr = _content->active_video_frame_rate(film);
+               if (_frame_interval_checker->guess() != FrameIntervalChecker::AGAIN) {
+                       _frame_interval_checker.reset ();
+               }
+       }
 
        Frame frame;
-       Eyes eyes = EYES_BOTH;
+       Eyes eyes = Eyes::BOTH;
        if (!_position) {
                /* This is the first data we have received since initialisation or seek.  Set
                   the position based on the frame that was given.  After this first time
-                  we just cound frames, since (as with audio) it seems that ContentTimes
+                  we just count frames, since (as with audio) it seems that ContentTimes
                   are unreliable from FFmpegDecoder.  They are much better than audio times
                   but still we get the occasional one which is duplicated.  In this case
                   ffmpeg seems to carry on regardless, processing the video frame as normal.
                   If we drop the frame with the duplicated timestamp we obviously lose sync.
                */
-               _position = ContentTime::from_frames (decoder_frame, afr);
-               if (_content->video->frame_type() == VIDEO_FRAME_TYPE_3D_ALTERNATE) {
+
+               if (vft == VideoFrameType::THREE_D_ALTERNATE) {
                        frame = decoder_frame / 2;
-                       _last_emitted_eyes = EYES_RIGHT;
+                       eyes = (decoder_frame % 2) ? Eyes::RIGHT : Eyes::LEFT;
                } else {
                        frame = decoder_frame;
+                       if (vft == VideoFrameType::THREE_D) {
+                               auto j2k = dynamic_pointer_cast<const J2KImageProxy>(image);
+                               /* At the moment only DCP decoders producers VideoFrameType::THREE_D, so only the J2KImageProxy
+                                * knows which eye it is.
+                                */
+                               if (j2k && j2k->eye()) {
+                                       eyes = j2k->eye().get() == dcp::Eye::LEFT ? Eyes::LEFT : Eyes::RIGHT;
+                               }
+                       }
                }
+
+               _position = ContentTime::from_frames (frame, afr);
        } else {
-               VideoFrameType const ft = _content->video->frame_type ();
-               if (ft == VIDEO_FRAME_TYPE_3D_ALTERNATE || ft == VIDEO_FRAME_TYPE_3D) {
+               if (vft == VideoFrameType::THREE_D) {
+                       auto j2k = dynamic_pointer_cast<const J2KImageProxy>(image);
+                       if (j2k && j2k->eye()) {
+                               if (j2k->eye() == dcp::Eye::LEFT) {
+                                       frame = _position->frames_round(afr) + 1;
+                                       eyes = Eyes::LEFT;
+                               } else {
+                                       frame = _position->frames_round(afr);
+                                       eyes = Eyes::RIGHT;
+                               }
+                       } else {
+                               /* This should not happen; see above */
+                               frame = _position->frames_round(afr) + 1;
+                       }
+               } else if (vft == VideoFrameType::THREE_D_ALTERNATE) {
                        DCPOMATIC_ASSERT (_last_emitted_eyes);
-                       if (_last_emitted_eyes.get() == EYES_RIGHT) {
+                       if (_last_emitted_eyes.get() == Eyes::RIGHT) {
                                frame = _position->frames_round(afr) + 1;
-                               eyes = EYES_LEFT;
+                               eyes = Eyes::LEFT;
                        } else {
                                frame = _position->frames_round(afr);
-                               eyes = EYES_RIGHT;
+                               eyes = Eyes::RIGHT;
                        }
                } else {
                        frame = _position->frames_round(afr) + 1;
                }
        }
 
-       switch (_content->video->frame_type ()) {
-       case VIDEO_FRAME_TYPE_2D:
-               Data (ContentVideo (image, frame, EYES_BOTH, PART_WHOLE));
+       switch (vft) {
+       case VideoFrameType::TWO_D:
+       case VideoFrameType::THREE_D:
+               Data (ContentVideo (image, frame, eyes, Part::WHOLE));
                break;
-       case VIDEO_FRAME_TYPE_3D:
+       case VideoFrameType::THREE_D_ALTERNATE:
        {
-               Data (ContentVideo (image, frame, eyes, PART_WHOLE));
-               _last_emitted_frame = frame;
+               Data (ContentVideo (image, frame, eyes, Part::WHOLE));
                _last_emitted_eyes = eyes;
                break;
        }
-       case VIDEO_FRAME_TYPE_3D_ALTERNATE:
-       {
-               Data (ContentVideo (image, frame, eyes, PART_WHOLE));
-               _last_emitted_eyes = eyes;
+       case VideoFrameType::THREE_D_LEFT_RIGHT:
+               Data (ContentVideo (image, frame, Eyes::LEFT, Part::LEFT_HALF));
+               Data (ContentVideo (image, frame, Eyes::RIGHT, Part::RIGHT_HALF));
                break;
-       }
-       case VIDEO_FRAME_TYPE_3D_LEFT_RIGHT:
-               Data (ContentVideo (image, frame, EYES_LEFT, PART_LEFT_HALF));
-               Data (ContentVideo (image, frame, EYES_RIGHT, PART_RIGHT_HALF));
-               break;
-       case VIDEO_FRAME_TYPE_3D_TOP_BOTTOM:
-               Data (ContentVideo (image, frame, EYES_LEFT, PART_TOP_HALF));
-               Data (ContentVideo (image, frame, EYES_RIGHT, PART_BOTTOM_HALF));
+       case VideoFrameType::THREE_D_TOP_BOTTOM:
+               Data (ContentVideo (image, frame, Eyes::LEFT, Part::TOP_HALF));
+               Data (ContentVideo (image, frame, Eyes::RIGHT, Part::BOTTOM_HALF));
                break;
-       case VIDEO_FRAME_TYPE_3D_LEFT:
-               Data (ContentVideo (image, frame, EYES_LEFT, PART_WHOLE));
+       case VideoFrameType::THREE_D_LEFT:
+               Data (ContentVideo (image, frame, Eyes::LEFT, Part::WHOLE));
                break;
-       case VIDEO_FRAME_TYPE_3D_RIGHT:
-               Data (ContentVideo (image, frame, EYES_RIGHT, PART_WHOLE));
+       case VideoFrameType::THREE_D_RIGHT:
+               Data (ContentVideo (image, frame, Eyes::RIGHT, Part::WHOLE));
                break;
        default:
                DCPOMATIC_ASSERT (false);
@@ -160,10 +181,11 @@ VideoDecoder::emit (shared_ptr<const Film> film, shared_ptr<const ImageProxy> im
        _position = ContentTime::from_frames (frame, afr);
 }
 
+
 void
 VideoDecoder::seek ()
 {
-       _position = boost::optional<ContentTime>();
-       _last_emitted_frame.reset ();
+       _position = boost::none;
        _last_emitted_eyes.reset ();
+       _frame_interval_checker.reset (new FrameIntervalChecker());
 }