src/lib/dcp_decoder.cc

   1 /*
   2     Copyright (C) 2014-2020 Carl Hetherington <cth@carlh.net>
   3
   4     This file is part of DCP-o-matic.
   5
   6     DCP-o-matic is free software; you can redistribute it and/or modify
   7     it under the terms of the GNU General Public License as published by
   8     the Free Software Foundation; either version 2 of the License, or
   9     (at your option) any later version.
  10
  11     DCP-o-matic is distributed in the hope that it will be useful,
  12     but WITHOUT ANY WARRANTY; without even the implied warranty of
  13     MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  14     GNU General Public License for more details.
  15
  16     You should have received a copy of the GNU General Public License
  17     along with DCP-o-matic.  If not, see <http://www.gnu.org/licenses/>.
  18
  19 */
  20
  21 #include "dcp_decoder.h"
  22 #include "dcp_content.h"
  23 #include "audio_content.h"
  24 #include "video_decoder.h"
  25 #include "audio_decoder.h"
  26 #include "j2k_image_proxy.h"
  27 #include "text_decoder.h"
  28 #include "ffmpeg_image_proxy.h"
  29 #include "image.h"
  30 #include "config.h"
  31 #include "frame_interval_checker.h"
  32 #include <dcp/dcp.h>
  33 #include <dcp/cpl.h>
  34 #include <dcp/reel.h>
  35 #include <dcp/mono_picture_asset.h>
  36 #include <dcp/mono_picture_asset_reader.h>
  37 #include <dcp/stereo_picture_asset.h>
  38 #include <dcp/stereo_picture_asset_reader.h>
  39 #include <dcp/reel_picture_asset.h>
  40 #include <dcp/reel_sound_asset.h>
  41 #include <dcp/reel_subtitle_asset.h>
  42 #include <dcp/reel_closed_caption_asset.h>
  43 #include <dcp/mono_picture_frame.h>
  44 #include <dcp/stereo_picture_frame.h>
  45 #include <dcp/sound_frame.h>
  46 #include <dcp/sound_asset_reader.h>
  47 #include <dcp/subtitle_image.h>
  48 #include <boost/foreach.hpp>
  49 #include <iostream>
  50
  51 #include "i18n.h"
  52
  53 using std::list;
  54 using std::cout;
  55 using boost::shared_ptr;
  56 using boost::dynamic_pointer_cast;
  57 using boost::optional;
  58
  59 DCPDecoder::DCPDecoder (shared_ptr<const Film> film, shared_ptr<const DCPContent> c, bool fast)
  60         : DCP (c)
  61         , Decoder (film)
  62         , _decode_referenced (false)
  63 {
  64         if (c->can_be_played()) {
  65                 if (c->video) {
  66                         video.reset (new VideoDecoder (this, c));
  67                 }
  68                 if (c->audio) {
  69                         audio.reset (new AudioDecoder (this, c->audio, fast));
  70                 }
  71                 BOOST_FOREACH (shared_ptr<TextContent> i, c->text) {
  72                         /* XXX: this time here should be the time of the first subtitle, not 0 */
  73                         text.push_back (shared_ptr<TextDecoder> (new TextDecoder (this, i, ContentTime())));
  74                 }
  75         }
  76
  77         list<shared_ptr<dcp::CPL> > cpl_list = cpls ();
  78
  79         if (cpl_list.empty()) {
  80                 throw DCPError (_("No CPLs found in DCP."));
  81         }
  82
  83         shared_ptr<dcp::CPL> cpl;
  84         BOOST_FOREACH (shared_ptr<dcp::CPL> i, cpl_list) {
  85                 if (_dcp_content->cpl() && i->id() == _dcp_content->cpl().get()) {
  86                         cpl = i;
  87                 }
  88         }
  89
  90         if (!cpl) {
  91                 /* No CPL found; probably an old file that doesn't specify it;
  92                    just use the first one.
  93                 */
  94                 cpl = cpls().front ();
  95         }
  96
  97         set_decode_referenced (false);
  98
  99         _reels = cpl->reels ();
 100
 101         _reel = _reels.begin ();
 102         _offset = 0;
 103         get_readers ();
 104 }
 105
 106
 107 bool
 108 DCPDecoder::pass ()
 109 {
 110         if (!_dcp_content->can_be_played()) {
 111                 return true;
 112         }
 113
 114         if (_reel == _reels.end()) {
 115                 if (audio) {
 116                         audio->flush ();
 117                 }
 118                 return true;
 119         }
 120
 121         double const vfr = _dcp_content->active_video_frame_rate (film());
 122
 123         /* Frame within the (played part of the) reel that is coming up next */
 124         int64_t const frame = _next.frames_round (vfr);
 125
 126         shared_ptr<dcp::PictureAsset> picture_asset = (*_reel)->main_picture()->asset();
 127         DCPOMATIC_ASSERT (picture_asset);
 128
 129         /* We must emit texts first as when we emit the video for this frame
 130            it will expect already to have the texts.
 131         */
 132         pass_texts (_next, picture_asset->size());
 133
 134         if ((_mono_reader || _stereo_reader) && (_decode_referenced || !_dcp_content->reference_video())) {
 135                 int64_t const entry_point = (*_reel)->main_picture()->entry_point ();
 136                 if (_mono_reader) {
 137                         video->emit (
 138                                 film(),
 139                                 shared_ptr<ImageProxy> (
 140                                         new J2KImageProxy (
 141                                                 _mono_reader->get_frame (entry_point + frame),
 142                                                 picture_asset->size(),
 143                                                 AV_PIX_FMT_XYZ12LE,
 144                                                 _forced_reduction
 145                                                 )
 146                                         ),
 147                                 _offset + frame
 148                                 );
 149                 } else {
 150                         video->emit (
 151                                 film(),
 152                                 shared_ptr<ImageProxy> (
 153                                         new J2KImageProxy (
 154                                                 _stereo_reader->get_frame (entry_point + frame),
 155                                                 picture_asset->size(),
 156                                                 dcp::EYE_LEFT,
 157                                                 AV_PIX_FMT_XYZ12LE,
 158                                                 _forced_reduction
 159                                                 )
 160                                         ),
 161                                 _offset + frame
 162                                 );
 163
 164                         video->emit (
 165                                 film(),
 166                                 shared_ptr<ImageProxy> (
 167                                         new J2KImageProxy (
 168                                                 _stereo_reader->get_frame (entry_point + frame),
 169                                                 picture_asset->size(),
 170                                                 dcp::EYE_RIGHT,
 171                                                 AV_PIX_FMT_XYZ12LE,
 172                                                 _forced_reduction
 173                                                 )
 174                                         ),
 175                                 _offset + frame
 176                                 );
 177                 }
 178         }
 179
 180         if (_sound_reader && (_decode_referenced || !_dcp_content->reference_audio())) {
 181                 int64_t const entry_point = (*_reel)->main_sound()->entry_point ();
 182                 shared_ptr<const dcp::SoundFrame> sf = _sound_reader->get_frame (entry_point + frame);
 183                 uint8_t const * from = sf->data ();
 184
 185                 int const channels = _dcp_content->audio->stream()->channels ();
 186                 int const frames = sf->size() / (3 * channels);
 187                 shared_ptr<AudioBuffers> data (new AudioBuffers (channels, frames));
 188                 float** data_data = data->data();
 189                 for (int i = 0; i < frames; ++i) {
 190                         for (int j = 0; j < channels; ++j) {
 191                                 data_data[j][i] = static_cast<int> ((from[0] << 8) | (from[1] << 16) | (from[2] << 24)) / static_cast<float> (INT_MAX - 256);
 192                                 from += 3;
 193                         }
 194                 }
 195
 196                 audio->emit (film(), _dcp_content->audio->stream(), data, ContentTime::from_frames (_offset, vfr) + _next);
 197         }
 198
 199         _next += ContentTime::from_frames (1, vfr);
 200
 201         if ((*_reel)->main_picture ()) {
 202                 if (_next.frames_round (vfr) >= (*_reel)->main_picture()->duration()) {
 203                         next_reel ();
 204                         _next = ContentTime ();
 205                 }
 206         }
 207
 208         return false;
 209 }
 210
 211 void
 212 DCPDecoder::pass_texts (ContentTime next, dcp::Size size)
 213 {
 214         list<shared_ptr<TextDecoder> >::const_iterator decoder = text.begin ();
 215         if (decoder == text.end()) {
 216                 /* It's possible that there is now a main subtitle but no TextDecoders, for example if
 217                    the CPL has just changed but the TextContent's texts have not been recreated yet.
 218                 */
 219                 return;
 220         }
 221
 222         if ((*_reel)->main_subtitle()) {
 223                 pass_texts (
 224                         next,
 225                         (*_reel)->main_subtitle()->asset(),
 226                         _dcp_content->reference_text(TEXT_OPEN_SUBTITLE),
 227                         (*_reel)->main_subtitle()->entry_point(),
 228                         *decoder,
 229                         size
 230                         );
 231                 ++decoder;
 232         }
 233
 234         BOOST_FOREACH (shared_ptr<dcp::ReelClosedCaptionAsset> i, (*_reel)->closed_captions()) {
 235                 pass_texts (
 236                         next, i->asset(), _dcp_content->reference_text(TEXT_CLOSED_CAPTION), i->entry_point(), *decoder, size
 237                         );
 238                 ++decoder;
 239         }
 240 }
 241
 242 void
 243 DCPDecoder::pass_texts (
 244         ContentTime next, shared_ptr<dcp::SubtitleAsset> asset, bool reference, int64_t entry_point, shared_ptr<TextDecoder> decoder, dcp::Size size
 245         )
 246 {
 247         double const vfr = _dcp_content->active_video_frame_rate (film());
 248         /* Frame within the (played part of the) reel that is coming up next */
 249         int64_t const frame = next.frames_round (vfr);
 250
 251         if (_decode_referenced || !reference) {
 252                 list<shared_ptr<dcp::Subtitle> > subs = asset->subtitles_during (
 253                         dcp::Time (entry_point + frame, vfr, vfr),
 254                         dcp::Time (entry_point + frame + 1, vfr, vfr),
 255                         true
 256                         );
 257
 258                 list<dcp::SubtitleString> strings;
 259
 260                 BOOST_FOREACH (shared_ptr<dcp::Subtitle> i, subs) {
 261                         shared_ptr<dcp::SubtitleString> is = dynamic_pointer_cast<dcp::SubtitleString> (i);
 262                         if (is) {
 263                                 if (!strings.empty() && (strings.back().in() != is->in() || strings.back().out() != is->out())) {
 264                                         dcp::SubtitleString b = strings.back();
 265                                         decoder->emit_plain (
 266                                                 ContentTimePeriod (
 267                                                         ContentTime::from_frames(_offset - entry_point, vfr) + ContentTime::from_seconds(b.in().as_seconds()),
 268                                                         ContentTime::from_frames(_offset - entry_point, vfr) + ContentTime::from_seconds(b.out().as_seconds())
 269                                                         ),
 270                                                 strings
 271                                                 );
 272                                         strings.clear ();
 273                                 }
 274
 275                                 strings.push_back (*is);
 276                         }
 277
 278                         /* XXX: perhaps these image subs should also be collected together like the string ones are;
 279                            this would need to be done both here and in DCPSubtitleDecoder.
 280                         */
 281
 282                         shared_ptr<dcp::SubtitleImage> ii = dynamic_pointer_cast<dcp::SubtitleImage> (i);
 283                         if (ii) {
 284                                 emit_subtitle_image (
 285                                         ContentTimePeriod (
 286                                                 ContentTime::from_frames (_offset - entry_point, vfr) + ContentTime::from_seconds (i->in().as_seconds ()),
 287                                                 ContentTime::from_frames (_offset - entry_point, vfr) + ContentTime::from_seconds (i->out().as_seconds ())
 288                                                 ),
 289                                         *ii,
 290                                         size,
 291                                         decoder
 292                                         );
 293                         }
 294                 }
 295
 296                 if (!strings.empty()) {
 297                         dcp::SubtitleString b = strings.back();
 298                         decoder->emit_plain (
 299                                 ContentTimePeriod (
 300                                         ContentTime::from_frames(_offset - entry_point, vfr) + ContentTime::from_seconds(b.in().as_seconds()),
 301                                         ContentTime::from_frames(_offset - entry_point, vfr) + ContentTime::from_seconds(b.out().as_seconds())
 302                                         ),
 303                                 strings
 304                                 );
 305                         strings.clear ();
 306                 }
 307         }
 308 }
 309
 310 void
 311 DCPDecoder::next_reel ()
 312 {
 313         _offset += (*_reel)->main_picture()->duration();
 314         ++_reel;
 315         get_readers ();
 316 }
 317
 318 void
 319 DCPDecoder::get_readers ()
 320 {
 321         if (_reel == _reels.end() || !_dcp_content->can_be_played ()) {
 322                 _mono_reader.reset ();
 323                 _stereo_reader.reset ();
 324                 _sound_reader.reset ();
 325                 return;
 326         }
 327
 328         if ((*_reel)->main_picture()) {
 329                 shared_ptr<dcp::PictureAsset> asset = (*_reel)->main_picture()->asset ();
 330                 shared_ptr<dcp::MonoPictureAsset> mono = dynamic_pointer_cast<dcp::MonoPictureAsset> (asset);
 331                 shared_ptr<dcp::StereoPictureAsset> stereo = dynamic_pointer_cast<dcp::StereoPictureAsset> (asset);
 332                 DCPOMATIC_ASSERT (mono || stereo);
 333                 if (mono) {
 334                         _mono_reader = mono->start_read ();
 335                         _stereo_reader.reset ();
 336                 } else {
 337                         _stereo_reader = stereo->start_read ();
 338                         _mono_reader.reset ();
 339                 }
 340         } else {
 341                 _mono_reader.reset ();
 342                 _stereo_reader.reset ();
 343         }
 344
 345         if ((*_reel)->main_sound()) {
 346                 _sound_reader = (*_reel)->main_sound()->asset()->start_read ();
 347         } else {
 348                 _sound_reader.reset ();
 349         }
 350 }
 351
 352 void
 353 DCPDecoder::seek (ContentTime t, bool accurate)
 354 {
 355         if (!_dcp_content->can_be_played ()) {
 356                 return;
 357         }
 358
 359         Decoder::seek (t, accurate);
 360
 361         _reel = _reels.begin ();
 362         _offset = 0;
 363         get_readers ();
 364
 365         int const pre_roll_seconds = 2;
 366
 367         /* Pre-roll for subs */
 368
 369         ContentTime pre = t - ContentTime::from_seconds (pre_roll_seconds);
 370         if (pre < ContentTime()) {
 371                 pre = ContentTime ();
 372         }
 373
 374         /* Seek to pre-roll position */
 375
 376         while (_reel != _reels.end() && pre >= ContentTime::from_frames ((*_reel)->main_picture()->duration(), _dcp_content->active_video_frame_rate(film()))) {
 377                 ContentTime rd = ContentTime::from_frames ((*_reel)->main_picture()->duration(), _dcp_content->active_video_frame_rate(film()));
 378                 pre -= rd;
 379                 t -= rd;
 380                 next_reel ();
 381         }
 382
 383         /* Pass texts in the pre-roll */
 384
 385         double const vfr = _dcp_content->active_video_frame_rate (film());
 386         for (int i = 0; i < pre_roll_seconds * vfr; ++i) {
 387                 pass_texts (pre, (*_reel)->main_picture()->asset()->size());
 388                 pre += ContentTime::from_frames (1, vfr);
 389         }
 390
 391         /* Seek to correct position */
 392
 393         while (_reel != _reels.end() && t >= ContentTime::from_frames ((*_reel)->main_picture()->duration(), _dcp_content->active_video_frame_rate(film()))) {
 394                 t -= ContentTime::from_frames ((*_reel)->main_picture()->duration(), _dcp_content->active_video_frame_rate(film()));
 395                 next_reel ();
 396         }
 397
 398         _next = t;
 399 }
 400
 401 void
 402 DCPDecoder::set_decode_referenced (bool r)
 403 {
 404         _decode_referenced = r;
 405
 406         if (video) {
 407                 video->set_ignore (_dcp_content->reference_video() && !_decode_referenced);
 408         }
 409         if (audio) {
 410                 audio->set_ignore (_dcp_content->reference_audio() && !_decode_referenced);
 411         }
 412 }
 413
 414 void
 415 DCPDecoder::set_forced_reduction (optional<int> reduction)
 416 {
 417         _forced_reduction = reduction;
 418 }
 419
 420 ContentTime
 421 DCPDecoder::position () const
 422 {
 423         return ContentTime::from_frames(_offset, _dcp_content->active_video_frame_rate(film())) + _next;
 424 }