Rename MD5Digester -> Digester.
[dcpomatic.git] / src / lib / ffmpeg_decoder.cc
index 72845a133eb0b11be7a5cfc307e0d0d5ef4addd4..df6b7416b35a2fc521846ed91ec490e6faaef45d 100644 (file)
@@ -1,19 +1,20 @@
 /*
     Copyright (C) 2012-2016 Carl Hetherington <cth@carlh.net>
 
-    This program is free software; you can redistribute it and/or modify
+    This file is part of DCP-o-matic.
+
+    DCP-o-matic is free software; you can redistribute it and/or modify
     it under the terms of the GNU General Public License as published by
     the Free Software Foundation; either version 2 of the License, or
     (at your option) any later version.
 
-    This program is distributed in the hope that it will be useful,
+    DCP-o-matic is distributed in the hope that it will be useful,
     but WITHOUT ANY WARRANTY; without even the implied warranty of
     MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
     GNU General Public License for more details.
 
     You should have received a copy of the GNU General Public License
-    along with this program; if not, write to the Free Software
-    Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+    along with DCP-o-matic.  If not, see <http://www.gnu.org/licenses/>.
 
 */
 
 #include "util.h"
 #include "log.h"
 #include "ffmpeg_decoder.h"
+#include "subtitle_decoder.h"
 #include "ffmpeg_audio_stream.h"
 #include "ffmpeg_subtitle_stream.h"
 #include "video_filter_graph.h"
 #include "audio_buffers.h"
 #include "ffmpeg_content.h"
 #include "raw_image_proxy.h"
+#include "video_decoder.h"
 #include "film.h"
-#include "md5_digester.h"
+#include "audio_decoder.h"
 #include "compose.hpp"
+#include "subtitle_content.h"
 #include <dcp/subtitle_string.h>
 #include <sub/ssa_reader.h>
 #include <sub/subtitle.h>
@@ -72,14 +76,30 @@ using boost::split;
 using dcp::Size;
 
 FFmpegDecoder::FFmpegDecoder (shared_ptr<const FFmpegContent> c, shared_ptr<Log> log, bool fast)
-       : VideoDecoder (c)
-       , AudioDecoder (c, fast)
-       , SubtitleDecoder (c)
-       , FFmpeg (c)
+       : FFmpeg (c)
        , _log (log)
-       , _pts_offset (pts_offset (c->ffmpeg_audio_streams(), c->first_video(), c->video_frame_rate()))
 {
+       if (c->video) {
+               video.reset (new VideoDecoder (this, c, log));
+               _pts_offset = pts_offset (c->ffmpeg_audio_streams(), c->first_video(), c->active_video_frame_rate());
+       } else {
+               _pts_offset = ContentTime ();
+       }
+
+       if (c->audio) {
+               audio.reset (new AudioDecoder (this, c->audio, fast, log));
+       }
 
+       if (c->subtitle) {
+               subtitle.reset (
+                       new SubtitleDecoder (
+                               this,
+                               c->subtitle,
+                               bind (&FFmpegDecoder::image_subtitles_during, this, _1, _2),
+                               bind (&FFmpegDecoder::text_subtitles_during, this, _1, _2)
+                               )
+                       );
+       }
 }
 
 void
@@ -92,10 +112,12 @@ FFmpegDecoder::flush ()
 
        /* XXX: should we reset _packet.data and size after each *_decode_* call? */
 
-       while (decode_video_packet ()) {}
+       while (video && decode_video_packet ()) {}
 
-       decode_audio_packet ();
-       AudioDecoder::flush ();
+       if (audio) {
+               decode_audio_packet ();
+               audio->flush ();
+       }
 }
 
 bool
@@ -122,7 +144,7 @@ FFmpegDecoder::pass (PassReason reason, bool accurate)
        int const si = _packet.stream_index;
        shared_ptr<const FFmpegContent> fc = _ffmpeg_content;
 
-       if (si == _video_stream && !_ignore_video && (accurate || reason != PASS_REASON_SUBTITLE)) {
+       if (_video_stream && si == _video_stream.get() && !video->ignore() && (accurate || reason != PASS_REASON_SUBTITLE)) {
                decode_video_packet ();
        } else if (fc->subtitle_stream() && fc->subtitle_stream()->uses_index (_format_context, si)) {
                decode_subtitle_packet ();
@@ -207,7 +229,7 @@ FFmpegDecoder::deinterleave_audio (shared_ptr<FFmpegAudioStream> stream) const
                int sample = 0;
                int channel = 0;
                for (int i = 0; i < total_samples; ++i) {
-                       audio->data(channel)[sample] = static_cast<float>(*p++) / (1 << 31);
+                       audio->data(channel)[sample] = static_cast<float>(*p++) / 2147483648;
 
                        ++channel;
                        if (channel == stream->channels()) {
@@ -218,6 +240,17 @@ FFmpegDecoder::deinterleave_audio (shared_ptr<FFmpegAudioStream> stream) const
        }
        break;
 
+       case AV_SAMPLE_FMT_S32P:
+       {
+               int32_t** p = reinterpret_cast<int32_t **> (_frame->data);
+               for (int i = 0; i < stream->channels(); ++i) {
+                       for (int j = 0; j < frames; ++j) {
+                               audio->data(i)[j] = static_cast<float>(p[i][j]) / 2147483648;
+                       }
+               }
+       }
+       break;
+
        case AV_SAMPLE_FMT_FLT:
        {
                float* p = reinterpret_cast<float*> (_frame->data[0]);
@@ -270,9 +303,17 @@ FFmpegDecoder::bytes_per_audio_sample (shared_ptr<FFmpegAudioStream> stream) con
 void
 FFmpegDecoder::seek (ContentTime time, bool accurate)
 {
-       VideoDecoder::seek (time, accurate);
-       AudioDecoder::seek (time, accurate);
-       SubtitleDecoder::seek (time, accurate);
+       if (video) {
+               video->seek (time, accurate);
+       }
+
+       if (audio) {
+               audio->seek (time, accurate);
+       }
+
+       if (subtitle) {
+               subtitle->seek (time, accurate);
+       }
 
        /* If we are doing an `accurate' seek, we need to use pre-roll, as
           we don't really know what the seek will give us.
@@ -285,11 +326,18 @@ FFmpegDecoder::seek (ContentTime time, bool accurate)
           http://www.mjbshaw.com/2012/04/seeking-in-ffmpeg-know-your-timestamp.html
        */
 
+       DCPOMATIC_ASSERT (_video_stream);
+
        ContentTime u = time - _pts_offset;
        if (u < ContentTime ()) {
                u = ContentTime ();
        }
-       av_seek_frame (_format_context, _video_stream, u.seconds() / av_q2d (_format_context->streams[_video_stream]->time_base), AVSEEK_FLAG_BACKWARD);
+       av_seek_frame (
+               _format_context,
+               _video_stream.get(),
+               u.seconds() / av_q2d (_format_context->streams[_video_stream.get()]->time_base),
+               AVSEEK_FLAG_BACKWARD
+               );
 
        avcodec_flush_buffers (video_codec_context());
 
@@ -358,7 +406,7 @@ FFmpegDecoder::decode_audio_packet ()
                        }
 
                        if (data->frames() > 0) {
-                               audio (*stream, data, ct);
+                               audio->give (*stream, data, ct);
                        }
                }
 
@@ -370,6 +418,8 @@ FFmpegDecoder::decode_audio_packet ()
 bool
 FFmpegDecoder::decode_video_packet ()
 {
+       DCPOMATIC_ASSERT (_video_stream);
+
        int frame_finished;
        if (avcodec_decode_video2 (video_codec_context(), _frame, &frame_finished, &_packet) < 0 || !frame_finished) {
                return false;
@@ -400,10 +450,10 @@ FFmpegDecoder::decode_video_packet ()
                shared_ptr<Image> image = i->first;
 
                if (i->second != AV_NOPTS_VALUE) {
-                       double const pts = i->second * av_q2d (_format_context->streams[_video_stream]->time_base) + _pts_offset.seconds ();
-                       video (
+                       double const pts = i->second * av_q2d (_format_context->streams[_video_stream.get()]->time_base) + _pts_offset.seconds ();
+                       video->give (
                                shared_ptr<ImageProxy> (new RawImageProxy (image)),
-                               llrint (pts * _ffmpeg_content->video_frame_rate ())
+                               llrint (pts * _ffmpeg_content->active_video_frame_rate ())
                                );
                } else {
                        LOG_WARNING_NC ("Dropping frame without PTS");
@@ -484,12 +534,21 @@ FFmpegDecoder::decode_bitmap_subtitle (AVSubtitleRect const * rect, ContentTimeP
        */
        shared_ptr<Image> image (new Image (AV_PIX_FMT_RGBA, dcp::Size (rect->w, rect->h), true));
 
+#ifdef DCPOMATIC_HAVE_AVSUBTITLERECT_PICT
+       /* Start of the first line in the subtitle */
+       uint8_t* sub_p = rect->pict.data[0];
+       /* sub_p looks up into a BGRA palette which is here
+          (i.e. first byte B, second G, third R, fourth A)
+       */
+       uint32_t const * palette = (uint32_t *) rect->pict.data[1];
+#else
        /* Start of the first line in the subtitle */
        uint8_t* sub_p = rect->data[0];
        /* sub_p looks up into a BGRA palette which is here
           (i.e. first byte B, second G, third R, fourth A)
        */
        uint32_t const * palette = (uint32_t *) rect->data[1];
+#endif
        /* And the stream has a map of those palette colours to colours
           chosen by the user; created a `mapped' palette from those settings.
        */
@@ -520,11 +579,15 @@ FFmpegDecoder::decode_bitmap_subtitle (AVSubtitleRect const * rect, ContentTimeP
                        /* XXX: this seems to be wrong to me (isn't the output image RGBA?) but it looks right on screen */
                        *out_line_p++ = (p.a << 24) | (p.r << 16) | (p.g << 8) | p.b;
                }
+#ifdef DCPOMATIC_HAVE_AVSUBTITLERECT_PICT
+               sub_p += rect->pict.linesize[0];
+#else
                sub_p += rect->linesize[0];
+#endif
                out_p += image->stride()[0] / sizeof (uint32_t);
        }
 
-       dcp::Size const vs = _ffmpeg_content->video_size ();
+       dcp::Size const vs = _ffmpeg_content->video->size ();
        dcpomatic::Rect<double> const scaled_rect (
                static_cast<double> (rect->x) / vs.width,
                static_cast<double> (rect->y) / vs.height,
@@ -532,7 +595,7 @@ FFmpegDecoder::decode_bitmap_subtitle (AVSubtitleRect const * rect, ContentTimeP
                static_cast<double> (rect->h) / vs.height
                );
 
-       image_subtitle (period, image, scaled_rect);
+       subtitle->give_image (period, image, scaled_rect);
 }
 
 void
@@ -550,52 +613,8 @@ FFmpegDecoder::decode_ass_subtitle (string ass, ContentTimePeriod period)
 
        sub::RawSubtitle base;
        list<sub::RawSubtitle> raw = sub::SSAReader::parse_line (base, bits[9]);
-       list<sub::Subtitle> subs = sub::collect<list<sub::Subtitle> > (raw);
-
-       /* XXX: lots of this is copied from TextSubtitle; there should probably be some sharing */
-
-       /* Highest line index in this subtitle */
-       int highest = 0;
-       BOOST_FOREACH (sub::Subtitle i, subs) {
-               BOOST_FOREACH (sub::Line j, i.lines) {
-                       DCPOMATIC_ASSERT (j.vertical_position.reference && j.vertical_position.reference.get() == sub::TOP_OF_SUBTITLE);
-                       DCPOMATIC_ASSERT (j.vertical_position.line);
-                       highest = max (highest, j.vertical_position.line.get());
-               }
-       }
 
-       list<dcp::SubtitleString> ss;
-
-       BOOST_FOREACH (sub::Subtitle i, sub::collect<list<sub::Subtitle> > (sub::SSAReader::parse_line (base, bits[9]))) {
-               BOOST_FOREACH (sub::Line j, i.lines) {
-                       BOOST_FOREACH (sub::Block k, j.blocks) {
-                               ss.push_back (
-                                       dcp::SubtitleString (
-                                               boost::optional<string> (),
-                                               k.italic,
-                                               dcp::Colour (255, 255, 255),
-                                               /* 48pt is 1/22nd of the screen height */
-                                               48,
-                                               1,
-                                               dcp::Time (i.from.seconds(), 1000),
-                                               dcp::Time (i.to.seconds(), 1000),
-                                               0,
-                                               dcp::HALIGN_CENTER,
-                                               /* This 1.015 is an arbitrary value to lift the bottom sub off the bottom
-                                                  of the screen a bit to a pleasing degree.
-                                               */
-                                               1.015 - ((1 + highest - j.vertical_position.line.get()) * 1.5 / 22),
-                                               dcp::VALIGN_TOP,
-                                               k.text,
-                                               static_cast<dcp::Effect> (0),
-                                               dcp::Colour (255, 255, 255),
-                                               dcp::Time (),
-                                               dcp::Time ()
-                                               )
-                                       );
-                       }
-               }
+       BOOST_FOREACH (sub::Subtitle const & i, sub::collect<list<sub::Subtitle> > (raw)) {
+               subtitle->give_text (period, i);
        }
-
-       text_subtitle (period, ss);
 }