src/lib/dcp_decoder.cc

   1 /*
   2     Copyright (C) 2014-2022 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
  22 #include "atmos_decoder.h"
  23 #include "audio_content.h"
  24 #include "audio_decoder.h"
  25 #include "config.h"
  26 #include "constants.h"
  27 #include "dcp_content.h"
  28 #include "dcp_decoder.h"
  29 #include "digester.h"
  30 #include "ffmpeg_image_proxy.h"
  31 #include "frame_interval_checker.h"
  32 #include "image.h"
  33 #include "j2k_image_proxy.h"
  34 #include "text_decoder.h"
  35 #include "video_decoder.h"
  36 #include <dcp/cpl.h>
  37 #include <dcp/dcp.h>
  38 #include <dcp/decrypted_kdm.h>
  39 #include <dcp/mono_picture_asset.h>
  40 #include <dcp/mono_picture_asset_reader.h>
  41 #include <dcp/mono_picture_frame.h>
  42 #include <dcp/reel.h>
  43 #include <dcp/reel_atmos_asset.h>
  44 #include <dcp/reel_closed_caption_asset.h>
  45 #include <dcp/reel_picture_asset.h>
  46 #include <dcp/reel_sound_asset.h>
  47 #include <dcp/reel_subtitle_asset.h>
  48 #include <dcp/search.h>
  49 #include <dcp/sound_asset_reader.h>
  50 #include <dcp/sound_frame.h>
  51 #include <dcp/stereo_picture_asset.h>
  52 #include <dcp/stereo_picture_asset_reader.h>
  53 #include <dcp/stereo_picture_frame.h>
  54 #include <dcp/subtitle_image.h>
  55 #include <iostream>
  56
  57 #include "i18n.h"
  58
  59
  60 using std::cout;
  61 using std::dynamic_pointer_cast;
  62 using std::list;
  63 using std::make_shared;
  64 using std::map;
  65 using std::shared_ptr;
  66 using std::string;
  67 using std::vector;
  68 using boost::optional;
  69 using namespace dcpomatic;
  70
  71
  72 DCPDecoder::DCPDecoder (shared_ptr<const Film> film, shared_ptr<const DCPContent> content, bool fast, bool tolerant, shared_ptr<DCPDecoder> old)
  73         : Decoder (film)
  74         , _dcp_content (content)
  75 {
  76         if (content->can_be_played()) {
  77                 if (content->video) {
  78                         video = make_shared<VideoDecoder>(this, content);
  79                 }
  80                 if (content->audio) {
  81                         audio = make_shared<AudioDecoder>(this, content->audio, fast);
  82                 }
  83                 for (auto i: content->text) {
  84                         text.push_back (make_shared<TextDecoder>(this, i));
  85                         /* We should really call maybe_set_position() on this TextDecoder to set the time
  86                          * of the first subtitle, but it probably doesn't matter since we'll always
  87                          * have regularly occurring video (and maybe audio) content.
  88                          */
  89                 }
  90                 if (content->atmos) {
  91                         atmos = make_shared<AtmosDecoder>(this, content);
  92                 }
  93         }
  94
  95         /* We try to avoid re-scanning the DCP's files every time we make a new DCPDecoder; we do this
  96            by re-using the _reels list.  Before we do this we need to check that nothing too serious
  97            has changed in the DCPContent.
  98
  99            We do this by storing a digest of the important bits of the DCPContent and then checking that's
 100            the same before we re-use _reels.
 101         */
 102
 103         _lazy_digest = calculate_lazy_digest (content);
 104
 105         if (old && old->lazy_digest() == _lazy_digest) {
 106                 _reels = old->_reels;
 107         } else {
 108
 109                 auto cpl_list = dcp::find_and_resolve_cpls(content->directories(), tolerant);
 110
 111                 if (cpl_list.empty()) {
 112                         throw DCPError (_("No CPLs found in DCP."));
 113                 }
 114
 115                 shared_ptr<dcp::CPL> cpl;
 116                 for (auto i: cpl_list) {
 117                         if (_dcp_content->cpl() && i->id() == _dcp_content->cpl().get()) {
 118                                 cpl = i;
 119                         }
 120                 }
 121
 122                 if (!cpl) {
 123                         /* No CPL found; probably an old file that doesn't specify it;
 124                            just use the first one.
 125                         */
 126                         cpl = cpl_list.front();
 127                 }
 128
 129                 if (content->kdm()) {
 130                         cpl->add (decrypt_kdm_with_helpful_error(content->kdm().get()));
 131                 }
 132
 133                 _reels = cpl->reels ();
 134         }
 135
 136         set_decode_referenced (false);
 137
 138         _reel = _reels.begin ();
 139         get_readers ();
 140
 141         _font_id_allocator.add_fonts_from_reels(_reels);
 142         _font_id_allocator.allocate();
 143 }
 144
 145
 146 bool
 147 DCPDecoder::pass ()
 148 {
 149         if (!_dcp_content->can_be_played()) {
 150                 return true;
 151         }
 152
 153         if (_reel == _reels.end()) {
 154                 if (audio) {
 155                         audio->flush ();
 156                 }
 157                 return true;
 158         }
 159
 160         auto const vfr = _dcp_content->active_video_frame_rate (film());
 161
 162         /* Frame within the (played part of the) reel that is coming up next */
 163         auto const frame = _next.frames_round (vfr);
 164
 165         auto picture_asset = (*_reel)->main_picture()->asset();
 166         DCPOMATIC_ASSERT (picture_asset);
 167
 168         /* We must emit texts first as when we emit the video for this frame
 169            it will expect already to have the texts.
 170         */
 171         pass_texts (_next, picture_asset->size());
 172
 173         if ((_mono_reader || _stereo_reader) && (_decode_referenced || !_dcp_content->reference_video())) {
 174                 auto const entry_point = (*_reel)->main_picture()->entry_point().get_value_or(0);
 175                 if (_mono_reader) {
 176                         video->emit (
 177                                 film(),
 178                                 std::make_shared<J2KImageProxy>(
 179                                         _mono_reader->get_frame (entry_point + frame),
 180                                         picture_asset->size(),
 181                                         AV_PIX_FMT_XYZ12LE,
 182                                         _forced_reduction
 183                                         ),
 184                                 _offset + frame
 185                                 );
 186                 } else {
 187                         video->emit (
 188                                 film(),
 189                                 std::make_shared<J2KImageProxy>(
 190                                         _stereo_reader->get_frame (entry_point + frame),
 191                                         picture_asset->size(),
 192                                         dcp::Eye::LEFT,
 193                                         AV_PIX_FMT_XYZ12LE,
 194                                         _forced_reduction
 195                                         ),
 196                                 _offset + frame
 197                                 );
 198
 199                         video->emit (
 200                                 film(),
 201                                 std::make_shared<J2KImageProxy>(
 202                                         _stereo_reader->get_frame (entry_point + frame),
 203                                         picture_asset->size(),
 204                                         dcp::Eye::RIGHT,
 205                                         AV_PIX_FMT_XYZ12LE,
 206                                         _forced_reduction
 207                                         ),
 208                                 _offset + frame
 209                                 );
 210                 }
 211         }
 212
 213         if (_sound_reader && (_decode_referenced || !_dcp_content->reference_audio())) {
 214                 auto const entry_point = (*_reel)->main_sound()->entry_point().get_value_or(0);
 215                 auto sf = _sound_reader->get_frame (entry_point + frame);
 216                 auto from = sf->data ();
 217
 218                 int const channels = _dcp_content->audio->stream()->channels ();
 219                 int const frames = sf->size() / (3 * channels);
 220                 auto data = make_shared<AudioBuffers>(channels, frames);
 221                 auto data_data = data->data();
 222                 for (int i = 0; i < frames; ++i) {
 223                         for (int j = 0; j < channels; ++j) {
 224                                 data_data[j][i] = static_cast<int> ((from[0] << 8) | (from[1] << 16) | (from[2] << 24)) / static_cast<float> (INT_MAX - 256);
 225                                 from += 3;
 226                         }
 227                 }
 228
 229                 audio->emit (film(), _dcp_content->audio->stream(), data, ContentTime::from_frames (_offset, vfr) + _next);
 230         }
 231
 232         if (_atmos_reader) {
 233                 DCPOMATIC_ASSERT (_atmos_metadata);
 234                 auto const entry_point = (*_reel)->atmos()->entry_point().get_value_or(0);
 235                 atmos->emit (film(), _atmos_reader->get_frame(entry_point + frame), _offset + frame, *_atmos_metadata);
 236         }
 237
 238         _next += ContentTime::from_frames (1, vfr);
 239
 240         if ((*_reel)->main_picture ()) {
 241                 if (_next.frames_round (vfr) >= (*_reel)->main_picture()->duration()) {
 242                         next_reel ();
 243                         _next = ContentTime ();
 244                 }
 245         }
 246
 247         return false;
 248 }
 249
 250
 251 void
 252 DCPDecoder::pass_texts (ContentTime next, dcp::Size size)
 253 {
 254         auto decoder = text.begin ();
 255         if (decoder == text.end()) {
 256                 /* It's possible that there is now a main subtitle but no TextDecoders, for example if
 257                    the CPL has just changed but the TextContent's texts have not been recreated yet.
 258                 */
 259                 return;
 260         }
 261
 262         if ((*_reel)->main_subtitle()) {
 263                 pass_texts (
 264                         next,
 265                         (*_reel)->main_subtitle()->asset(),
 266                         _dcp_content->reference_text(TextType::OPEN_SUBTITLE),
 267                         (*_reel)->main_subtitle()->entry_point().get_value_or(0),
 268                         *decoder,
 269                         size
 270                         );
 271                 ++decoder;
 272         }
 273
 274         for (auto i: (*_reel)->closed_captions()) {
 275                 pass_texts (
 276                         next, i->asset(), _dcp_content->reference_text(TextType::CLOSED_CAPTION), i->entry_point().get_value_or(0), *decoder, size
 277                         );
 278                 ++decoder;
 279         }
 280 }
 281
 282 void
 283 DCPDecoder::pass_texts (
 284         ContentTime next, shared_ptr<dcp::SubtitleAsset> asset, bool reference, int64_t entry_point, shared_ptr<TextDecoder> decoder, dcp::Size size
 285         )
 286 {
 287         auto const vfr = _dcp_content->active_video_frame_rate (film());
 288         /* Frame within the (played part of the) reel that is coming up next */
 289         auto const frame = next.frames_round (vfr);
 290
 291         if (_decode_referenced || !reference) {
 292                 auto subs = asset->subtitles_during (
 293                         dcp::Time (entry_point + frame, vfr, vfr),
 294                         dcp::Time (entry_point + frame + 1, vfr, vfr),
 295                         true
 296                         );
 297
 298                 vector<dcp::SubtitleString> strings;
 299
 300                 for (auto i: subs) {
 301                         auto is = dynamic_pointer_cast<const dcp::SubtitleString>(i);
 302                         if (is) {
 303                                 if (!strings.empty() && (strings.back().in() != is->in() || strings.back().out() != is->out())) {
 304                                         auto b = strings.back();
 305                                         decoder->emit_plain (
 306                                                 ContentTimePeriod (
 307                                                         ContentTime::from_frames(_offset - entry_point, vfr) + ContentTime::from_seconds(b.in().as_seconds()),
 308                                                         ContentTime::from_frames(_offset - entry_point, vfr) + ContentTime::from_seconds(b.out().as_seconds())
 309                                                         ),
 310                                                 strings,
 311                                                 asset->subtitle_standard()
 312                                                 );
 313                                         strings.clear ();
 314                                 }
 315
 316                                 dcp::SubtitleString is_copy = *is;
 317                                 if (is_copy.font()) {
 318                                         is_copy.set_font(_font_id_allocator.font_id(_reel - _reels.begin(), asset->id(), is_copy.font().get()));
 319                                 } else {
 320                                         is_copy.set_font(_font_id_allocator.default_font_id());
 321                                 }
 322                                 strings.push_back(is_copy);
 323                         }
 324
 325                         /* XXX: perhaps these image subs should also be collected together like the string ones are;
 326                            this would need to be done both here and in DCPSubtitleDecoder.
 327                         */
 328
 329                         auto ii = dynamic_pointer_cast<const dcp::SubtitleImage>(i);
 330                         if (ii) {
 331                                 emit_subtitle_image (
 332                                         ContentTimePeriod (
 333                                                 ContentTime::from_frames (_offset - entry_point, vfr) + ContentTime::from_seconds (i->in().as_seconds ()),
 334                                                 ContentTime::from_frames (_offset - entry_point, vfr) + ContentTime::from_seconds (i->out().as_seconds ())
 335                                                 ),
 336                                         *ii,
 337                                         size,
 338                                         decoder
 339                                         );
 340                         }
 341                 }
 342
 343                 if (!strings.empty()) {
 344                         auto b = strings.back();
 345                         decoder->emit_plain (
 346                                 ContentTimePeriod (
 347                                         ContentTime::from_frames(_offset - entry_point, vfr) + ContentTime::from_seconds(b.in().as_seconds()),
 348                                         ContentTime::from_frames(_offset - entry_point, vfr) + ContentTime::from_seconds(b.out().as_seconds())
 349                                         ),
 350                                 strings,
 351                                 asset->subtitle_standard()
 352                                 );
 353                         strings.clear ();
 354                 }
 355         }
 356 }
 357
 358
 359 void
 360 DCPDecoder::next_reel ()
 361 {
 362         _offset += (*_reel)->main_picture()->actual_duration();
 363         ++_reel;
 364         get_readers ();
 365 }
 366
 367
 368 void
 369 DCPDecoder::get_readers ()
 370 {
 371         if (_reel == _reels.end() || !_dcp_content->can_be_played ()) {
 372                 _mono_reader.reset ();
 373                 _stereo_reader.reset ();
 374                 _sound_reader.reset ();
 375                 _atmos_reader.reset ();
 376                 return;
 377         }
 378
 379         if (video && !video->ignore() && (*_reel)->main_picture()) {
 380                 auto asset = (*_reel)->main_picture()->asset ();
 381                 auto mono = dynamic_pointer_cast<dcp::MonoPictureAsset> (asset);
 382                 auto stereo = dynamic_pointer_cast<dcp::StereoPictureAsset> (asset);
 383                 DCPOMATIC_ASSERT (mono || stereo);
 384                 if (mono) {
 385                         _mono_reader = mono->start_read ();
 386                         _mono_reader->set_check_hmac (false);
 387                         _stereo_reader.reset ();
 388                 } else {
 389                         _stereo_reader = stereo->start_read ();
 390                         _stereo_reader->set_check_hmac (false);
 391                         _mono_reader.reset ();
 392                 }
 393         } else {
 394                 _mono_reader.reset ();
 395                 _stereo_reader.reset ();
 396         }
 397
 398         if (audio && !audio->ignore() && (*_reel)->main_sound()) {
 399                 _sound_reader = (*_reel)->main_sound()->asset()->start_read ();
 400                 _sound_reader->set_check_hmac (false);
 401         } else {
 402                 _sound_reader.reset ();
 403         }
 404
 405         if ((*_reel)->atmos()) {
 406                 auto asset = (*_reel)->atmos()->asset();
 407                 _atmos_reader = asset->start_read();
 408                 _atmos_reader->set_check_hmac (false);
 409                 _atmos_metadata = AtmosMetadata (asset);
 410         } else {
 411                 _atmos_reader.reset ();
 412                 _atmos_metadata = boost::none;
 413         }
 414 }
 415
 416
 417 void
 418 DCPDecoder::seek (ContentTime t, bool accurate)
 419 {
 420         if (!_dcp_content->can_be_played ()) {
 421                 return;
 422         }
 423
 424         Decoder::seek (t, accurate);
 425
 426         _reel = _reels.begin ();
 427         _offset = 0;
 428         get_readers ();
 429
 430         int const pre_roll_seconds = 2;
 431
 432         /* Pre-roll for subs */
 433
 434         auto pre = t - ContentTime::from_seconds (pre_roll_seconds);
 435         if (pre < ContentTime()) {
 436                 pre = ContentTime ();
 437         }
 438
 439         /* Seek to pre-roll position */
 440
 441         while (
 442                 _reel != _reels.end() &&
 443                 pre >= ContentTime::from_frames ((*_reel)->main_picture()->actual_duration(), _dcp_content->active_video_frame_rate(film()))
 444                 ) {
 445
 446                 auto rd = ContentTime::from_frames ((*_reel)->main_picture()->actual_duration(), _dcp_content->active_video_frame_rate(film()));
 447                 pre -= rd;
 448                 t -= rd;
 449                 next_reel ();
 450         }
 451
 452         /* Pass texts in the pre-roll */
 453
 454         if (_reel != _reels.end()) {
 455                 auto const vfr = _dcp_content->active_video_frame_rate (film());
 456                 for (int i = 0; i < pre_roll_seconds * vfr; ++i) {
 457                         pass_texts (pre, (*_reel)->main_picture()->asset()->size());
 458                         pre += ContentTime::from_frames (1, vfr);
 459                 }
 460         }
 461
 462         /* Seek to correct position */
 463
 464         while (
 465                 _reel != _reels.end() &&
 466                 t >= ContentTime::from_frames ((*_reel)->main_picture()->actual_duration(), _dcp_content->active_video_frame_rate(film()))
 467                 ) {
 468
 469                 t -= ContentTime::from_frames ((*_reel)->main_picture()->actual_duration(), _dcp_content->active_video_frame_rate(film()));
 470                 next_reel ();
 471         }
 472
 473         _next = t;
 474 }
 475
 476
 477 void
 478 DCPDecoder::set_decode_referenced (bool r)
 479 {
 480         _decode_referenced = r;
 481
 482         if (video) {
 483                 video->set_ignore (_dcp_content->reference_video() && !_decode_referenced);
 484         }
 485         if (audio) {
 486                 audio->set_ignore (_dcp_content->reference_audio() && !_decode_referenced);
 487         }
 488 }
 489
 490
 491 void
 492 DCPDecoder::set_forced_reduction (optional<int> reduction)
 493 {
 494         _forced_reduction = reduction;
 495 }
 496
 497
 498 string
 499 DCPDecoder::calculate_lazy_digest (shared_ptr<const DCPContent> c) const
 500 {
 501         Digester d;
 502         for (auto i: c->paths()) {
 503                 d.add (i.string());
 504         }
 505         if (_dcp_content->kdm()) {
 506                 d.add(_dcp_content->kdm()->id());
 507         }
 508         d.add (static_cast<bool>(c->cpl()));
 509         if (c->cpl()) {
 510                 d.add (c->cpl().get());
 511         }
 512         return d.get ();
 513 }
 514
 515
 516 ContentTime
 517 DCPDecoder::position () const
 518 {
 519         return ContentTime::from_frames(_offset, _dcp_content->active_video_frame_rate(film())) + _next;
 520 }
 521