Merge 1.0-seek and subtitle-content.
[dcpomatic.git] / src / lib / ffmpeg_decoder.cc
index a051bcd1283d9f120b51516e00fa24f013e65eb6..55f4eb5c69b63cc97c0c5ce6ba1bb44db619b0ff 100644 (file)
@@ -68,9 +68,7 @@ FFmpegDecoder::FFmpegDecoder (shared_ptr<const Film> f, shared_ptr<const FFmpegC
        , _subtitle_codec (0)
        , _decode_video (video)
        , _decode_audio (audio)
-       , _video_pts_offset (0)
-       , _audio_pts_offset (0)
-       , _just_sought (false)
+       , _pts_offset (0)
 {
        setup_subtitle ();
 
@@ -83,27 +81,25 @@ FFmpegDecoder::FFmpegDecoder (shared_ptr<const Film> f, shared_ptr<const FFmpegC
           Then we remove big initial gaps in PTS and we allow our
           insertion of black frames to work.
 
-          We will do:
-            audio_pts_to_use = audio_pts_from_ffmpeg + audio_pts_offset;
-            video_pts_to_use = video_pts_from_ffmpeg + video_pts_offset;
+          We will do pts_to_use = pts_from_ffmpeg + pts_offset;
        */
 
        bool const have_video = video && c->first_video();
-       bool const have_audio = audio && c->audio_stream() && c->audio_stream()->first_audio;
+       bool const have_audio = _decode_audio && c->audio_stream () && c->audio_stream()->first_audio;
 
        /* First, make one of them start at 0 */
 
        if (have_audio && have_video) {
-               _video_pts_offset = _audio_pts_offset = - min (c->first_video().get(), c->audio_stream()->first_audio.get());
+               _pts_offset = - min (c->first_video().get(), c->audio_stream()->first_audio.get());
        } else if (have_video) {
-               _video_pts_offset = - c->first_video().get();
+               _pts_offset = - c->first_video().get();
        } else if (have_audio) {
-               _audio_pts_offset = - c->audio_stream()->first_audio.get();
+               _pts_offset = - c->audio_stream()->first_audio.get();
        }
 
        /* Now adjust both so that the video pts starts on a frame */
        if (have_video && have_audio) {
-               double first_video = c->first_video().get() + _video_pts_offset;
+               double first_video = c->first_video().get() + _pts_offset;
                double const old_first_video = first_video;
                
                /* Round the first video up to a frame boundary */
@@ -111,8 +107,7 @@ FFmpegDecoder::FFmpegDecoder (shared_ptr<const Film> f, shared_ptr<const FFmpegC
                        first_video = ceil (first_video * c->video_frame_rate()) / c->video_frame_rate ();
                }
 
-               _video_pts_offset += first_video - old_first_video;
-               _audio_pts_offset += first_video - old_first_video;
+               _pts_offset += first_video - old_first_video;
        }
 }
 
@@ -143,12 +138,10 @@ FFmpegDecoder::flush ()
                decode_audio_packet ();
        }
 
-       /* Stop us being asked for any more data */
-       _video_position = _ffmpeg_content->video_length ();
-       _audio_position = _ffmpeg_content->audio_length ();
+       AudioDecoder::flush ();
 }
 
-void
+bool
 FFmpegDecoder::pass ()
 {
        int r = av_read_frame (_format_context, &_packet);
@@ -164,7 +157,7 @@ FFmpegDecoder::pass ()
                }
 
                flush ();
-               return;
+               return true;
        }
 
        shared_ptr<const Film> film = _film.lock ();
@@ -181,6 +174,7 @@ FFmpegDecoder::pass ()
        }
 
        av_free_packet (&_packet);
+       return false;
 }
 
 /** @param data pointer to array of pointers to buffers.
@@ -295,68 +289,135 @@ FFmpegDecoder::bytes_per_audio_sample () const
        return av_get_bytes_per_sample (audio_sample_format ());
 }
 
-void
-FFmpegDecoder::seek (VideoContent::Frame frame, bool accurate)
+int
+FFmpegDecoder::minimal_run (boost::function<bool (optional<ContentTime>, optional<ContentTime>, int)> finished)
 {
-       double const time_base = av_q2d (_format_context->streams[_video_stream]->time_base);
+       int frames_read = 0;
+       optional<ContentTime> last_video;
+       optional<ContentTime> last_audio;
 
-       /* If we are doing an accurate seek, our initial shot will be 5 frames (5 being
-          a number plucked from the air) earlier than we want to end up.  The loop below
-          will hopefully then step through to where we want to be.
-       */
-       int initial = frame;
+       while (!finished (last_video, last_audio, frames_read)) {
+               int r = av_read_frame (_format_context, &_packet);
+               if (r < 0) {
+                       /* We should flush our decoders here, possibly yielding a few more frames,
+                          but the consequence of having to do that is too hideous to contemplate.
+                          Instead we give up and say that you can't seek too close to the end
+                          of a file.
+                       */
+                       return frames_read;
+               }
 
-       if (accurate) {
-               initial -= 5;
+               ++frames_read;
+
+               double const time_base = av_q2d (_format_context->streams[_packet.stream_index]->time_base);
+
+               if (_packet.stream_index == _video_stream) {
+
+                       avcodec_get_frame_defaults (_frame);
+                       
+                       int finished = 0;
+                       r = avcodec_decode_video2 (video_codec_context(), _frame, &finished, &_packet);
+                       if (r >= 0 && finished) {
+                               last_video = rint (
+                                       (av_frame_get_best_effort_timestamp (_frame) * time_base + _pts_offset) * TIME_HZ
+                                       );
+                       }
+
+               } else if (_ffmpeg_content->audio_stream() && _packet.stream_index == _ffmpeg_content->audio_stream()->index (_format_context)) {
+                       AVPacket copy_packet = _packet;
+                       while (copy_packet.size > 0) {
+
+                               int finished;
+                               r = avcodec_decode_audio4 (audio_codec_context(), _frame, &finished, &_packet);
+                               if (r >= 0 && finished) {
+                                       last_audio = rint (
+                                               (av_frame_get_best_effort_timestamp (_frame) * time_base + _pts_offset) * TIME_HZ
+                                               );
+                               }
+                                       
+                               copy_packet.data += r;
+                               copy_packet.size -= r;
+                       }
+               }
+               
+               av_free_packet (&_packet);
        }
 
-       if (initial < 0) {
-               initial = 0;
+       return frames_read;
+}
+
+bool
+FFmpegDecoder::seek_overrun_finished (ContentTime seek, optional<ContentTime> last_video, optional<ContentTime> last_audio) const
+{
+       return (last_video && last_video.get() >= seek) || (last_audio && last_audio.get() >= seek);
+}
+
+bool
+FFmpegDecoder::seek_final_finished (int n, int done) const
+{
+       return n == done;
+}
+
+void
+FFmpegDecoder::seek_and_flush (ContentTime t)
+{
+       int64_t s = ((double (t) / TIME_HZ) - _pts_offset) /
+               av_q2d (_format_context->streams[_video_stream]->time_base);
+
+       if (_ffmpeg_content->audio_stream ()) {
+               s = min (
+                       s, int64_t (
+                               ((double (t) / TIME_HZ) - _pts_offset) /
+                               av_q2d (_ffmpeg_content->audio_stream()->stream(_format_context)->time_base)
+                               )
+                       );
        }
 
-       /* Initial seek time in the stream's timebase */
-       int64_t const initial_vt = ((initial / _ffmpeg_content->video_frame_rate()) - _video_pts_offset) / time_base;
+       /* Ridiculous empirical hack */
+       s--;
 
-       av_seek_frame (_format_context, _video_stream, initial_vt, AVSEEK_FLAG_BACKWARD);
+       av_seek_frame (_format_context, _video_stream, s, AVSEEK_FLAG_BACKWARD);
 
        avcodec_flush_buffers (video_codec_context());
+       if (audio_codec_context ()) {
+               avcodec_flush_buffers (audio_codec_context ());
+       }
        if (_subtitle_codec_context) {
                avcodec_flush_buffers (_subtitle_codec_context);
        }
+}
 
-       _just_sought = true;
-       _video_position = frame;
+void
+FFmpegDecoder::seek (ContentTime time, bool accurate)
+{
+       Decoder::seek (time, accurate);
+       AudioDecoder::seek (time, accurate);
        
-       if (frame == 0 || !accurate) {
-               /* We're already there, or we're as close as we need to be */
-               return;
+       /* If we are doing an accurate seek, our initial shot will be 200ms (200 being
+          a number plucked from the air) earlier than we want to end up.  The loop below
+          will hopefully then step through to where we want to be.
+       */
+
+       ContentTime pre_roll = accurate ? (0.2 * TIME_HZ) : 0;
+       ContentTime initial_seek = time - pre_roll;
+       if (initial_seek < 0) {
+               initial_seek = 0;
        }
 
-       while (1) {
-               int r = av_read_frame (_format_context, &_packet);
-               if (r < 0) {
-                       return;
-               }
+       /* Initial seek time in the video stream's timebase */
 
-               if (_packet.stream_index != _video_stream) {
-                       av_free_packet (&_packet);
-                       continue;
-               }
-               
-               int finished = 0;
-               r = avcodec_decode_video2 (video_codec_context(), _frame, &finished, &_packet);
-               if (r >= 0 && finished) {
-                       _video_position = rint (
-                               (av_frame_get_best_effort_timestamp (_frame) * time_base + _video_pts_offset) * _ffmpeg_content->video_frame_rate()
-                               );
+       seek_and_flush (initial_seek);
 
-                       if (_video_position >= (frame - 1)) {
-                               av_free_packet (&_packet);
-                               break;
-                       }
-               }
-               
-               av_free_packet (&_packet);
+       if (!accurate) {
+               /* That'll do */
+               return;
+       }
+
+       int const N = minimal_run (boost::bind (&FFmpegDecoder::seek_overrun_finished, this, time, _1, _2));
+
+       seek_and_flush (initial_seek);
+       if (N > 0) {
+               minimal_run (boost::bind (&FFmpegDecoder::seek_final_finished, this, N - 1, _3));
        }
 }
 
@@ -373,6 +434,7 @@ FFmpegDecoder::decode_audio_packet ()
 
                int frame_finished;
                int const decode_result = avcodec_decode_audio4 (audio_codec_context(), _frame, &frame_finished, &copy_packet);
+
                if (decode_result < 0) {
                        shared_ptr<const Film> film = _film.lock ();
                        assert (film);
@@ -381,31 +443,17 @@ FFmpegDecoder::decode_audio_packet ()
                }
 
                if (frame_finished) {
-                       
-                       if (_audio_position == 0) {
-                               /* Where we are in the source, in seconds */
-                               double const pts = av_q2d (_format_context->streams[copy_packet.stream_index]->time_base)
-                                       * av_frame_get_best_effort_timestamp(_frame) + _audio_pts_offset;
-
-                               if (pts > 0) {
-                                       /* Emit some silence */
-                                       shared_ptr<AudioBuffers> silence (
-                                               new AudioBuffers (
-                                                       _ffmpeg_content->audio_channels(),
-                                                       pts * _ffmpeg_content->content_audio_frame_rate()
-                                                       )
-                                               );
-                                       
-                                       silence->make_silent ();
-                                       audio (silence, _audio_position);
-                               }
-                       }
+                       ContentTime const ct = (
+                               av_frame_get_best_effort_timestamp (_frame) *
+                               av_q2d (_ffmpeg_content->audio_stream()->stream (_format_context)->time_base)
+                               + _pts_offset
+                               ) * TIME_HZ;
                        
                        int const data_size = av_samples_get_buffer_size (
                                0, audio_codec_context()->channels, _frame->nb_samples, audio_sample_format (), 1
                                );
-                       
-                       audio (deinterleave_audio (_frame->data, data_size), _audio_position);
+
+                       audio (deinterleave_audio (_frame->data, data_size), ct);
                }
                        
                copy_packet.data += decode_result;
@@ -454,45 +502,9 @@ FFmpegDecoder::decode_video_packet ()
                }
                
                if (i->second != AV_NOPTS_VALUE) {
-
-                       double const pts = i->second * av_q2d (_format_context->streams[_video_stream]->time_base) + _video_pts_offset;
-
-                       if (_just_sought) {
-                               /* We just did a seek, so disable any attempts to correct for where we
-                                  are / should be.
-                               */
-                               _video_position = rint (pts * _ffmpeg_content->video_frame_rate ());
-                               _just_sought = false;
-                       }
-
-                       double const next = _video_position / _ffmpeg_content->video_frame_rate();
-                       double const one_frame = 1 / _ffmpeg_content->video_frame_rate ();
-                       double delta = pts - next;
-
-                       while (delta > one_frame) {
-                               /* This PTS is more than one frame forward in time of where we think we should be; emit
-                                  a black frame.
-                               */
-
-                               /* XXX: I think this should be a copy of the last frame... */
-                               boost::shared_ptr<Image> black (
-                                       new Image (
-                                               static_cast<AVPixelFormat> (_frame->format),
-                                               libdcp::Size (video_codec_context()->width, video_codec_context()->height),
-                                               true
-                                               )
-                                       );
-                               
-                               black->make_black ();
-                               video (image, false, _video_position);
-                               delta -= one_frame;
-                       }
-
-                       if (delta > -one_frame) {
-                               /* This PTS is within a frame of being right; emit this (otherwise it will be dropped) */
-                               video (image, false, _video_position);
-                       }
-                               
+                       double const pts = i->second * av_q2d (_format_context->streams[_video_stream]->time_base) + _pts_offset;
+                       VideoFrame const f = rint (pts * _ffmpeg_content->video_frame_rate ());
+                       video (image, false, f);
                } else {
                        shared_ptr<const Film> film = _film.lock ();
                        assert (film);
@@ -525,14 +537,6 @@ FFmpegDecoder::setup_subtitle ()
        }
 }
 
-bool
-FFmpegDecoder::done () const
-{
-       bool const vd = !_decode_video || (_video_position >= _ffmpeg_content->video_length());
-       bool const ad = !_decode_audio || !_ffmpeg_content->audio_stream() || (_audio_position >= _ffmpeg_content->audio_length());
-       return vd && ad;
-}
-       
 void
 FFmpegDecoder::decode_subtitle_packet ()
 {
@@ -555,11 +559,11 @@ FFmpegDecoder::decode_subtitle_packet ()
        /* Subtitle PTS in seconds (within the source, not taking into account any of the
           source that we may have chopped off for the DCP)
        */
-       double const packet_time = (static_cast<double> (sub.pts ) / AV_TIME_BASE) + _video_pts_offset;
-
+       double const packet_time = (static_cast<double> (sub.pts) / AV_TIME_BASE) + _pts_offset;
+       
        /* hence start time for this sub */
-       Time const from = (packet_time + (double (sub.start_display_time) / 1e3)) * TIME_HZ;
-       Time const to = (packet_time + (double (sub.end_display_time) / 1e3)) * TIME_HZ;
+       ContentTime const from = (packet_time + (double (sub.start_display_time) / 1e3)) * TIME_HZ;
+       ContentTime const to = (packet_time + (double (sub.end_display_time) / 1e3)) * TIME_HZ;
 
        AVSubtitleRect const * rect = sub.rects[0];