git.sesse.net Git - casparcg/blob - modules/ffmpeg/producer/util/util.cpp

   1 /*\r
   2 * Copyright (c) 2011 Sveriges Television AB <info@casparcg.com>\r
   3 *\r
   4 * This file is part of CasparCG (www.casparcg.com).\r
   5 *\r
   6 * CasparCG is free software: you can redistribute it and/or modify\r
   7 * it under the terms of the GNU General Public License as published by\r
   8 * the Free Software Foundation, either version 3 of the License, or\r
   9 * (at your option) any later version.\r
  10 *\r
  11 * CasparCG is distributed in the hope that it will be useful,\r
  12 * but WITHOUT ANY WARRANTY; without even the implied warranty of\r
  13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the\r
  14 * GNU General Public License for more details.\r
  15 *\r
  16 * You should have received a copy of the GNU General Public License\r
  17 * along with CasparCG. If not, see <http://www.gnu.org/licenses/>.\r
  18 *\r
  19 * Author: Robert Nagy, ronag89@gmail.com\r
  20 */\r
  21 \r
  22 #include "../../stdafx.h"\r
  23 \r
  24 #include "util.h"\r
  25 \r
  26 #include "flv.h"\r
  27 \r
  28 #include "../tbb_avcodec.h"\r
  29 #include "../../ffmpeg_error.h"\r
  30 \r
  31 #include <tbb/concurrent_unordered_map.h>\r
  32 #include <tbb/concurrent_queue.h>\r
  33 \r
  34 #include <core/producer/frame/frame_transform.h>\r
  35 #include <core/producer/frame/frame_factory.h>\r
  36 #include <core/producer/frame_producer.h>\r
  37 #include <core/mixer/write_frame.h>\r
  38 #include <core/mixer/audio/audio_util.h>\r
  39 \r
  40 #include <common/exception/exceptions.h>\r
  41 #include <common/utility/assert.h>\r
  42 #include <common/memory/memcpy.h>\r
  43 \r
  44 #include <tbb/parallel_for.h>\r
  45 \r
  46 #include <boost/filesystem.hpp>\r
  47 #include <boost/lexical_cast.hpp>\r
  48 #include <boost/algorithm/string.hpp>\r
  49 \r
  50 #if defined(_MSC_VER)\r
  51 #pragma warning (push)\r
  52 #pragma warning (disable : 4244)\r
  53 #endif\r
  54 extern "C" \r
  55 {\r
  56         #include <libswscale/swscale.h>\r
  57         #include <libavcodec/avcodec.h>\r
  58         #include <libavformat/avformat.h>\r
  59 }\r
  60 #if defined(_MSC_VER)\r
  61 #pragma warning (pop)\r
  62 #endif\r
  63 \r
  64 namespace caspar { namespace ffmpeg {\r
  65                 \r
  66 std::shared_ptr<core::audio_buffer> flush_audio()\r
  67 {\r
  68         static std::shared_ptr<core::audio_buffer> audio(new core::audio_buffer());\r
  69         return audio;\r
  70 }\r
  71 \r
  72 std::shared_ptr<core::audio_buffer> empty_audio()\r
  73 {\r
  74         static std::shared_ptr<core::audio_buffer> audio(new core::audio_buffer());\r
  75         return audio;\r
  76 }\r
  77 \r
  78 std::shared_ptr<AVFrame>                        flush_video()\r
  79 {\r
  80         static std::shared_ptr<AVFrame> video(avcodec_alloc_frame(), av_free);\r
  81         return video;\r
  82 }\r
  83 \r
  84 std::shared_ptr<AVFrame>                        empty_video()\r
  85 {\r
  86         static std::shared_ptr<AVFrame> video(avcodec_alloc_frame(), av_free);\r
  87         return video;\r
  88 }\r
  89 \r
  90 core::field_mode::type get_mode(const AVFrame& frame)\r
  91 {\r
  92         if(!frame.interlaced_frame)\r
  93                 return core::field_mode::progressive;\r
  94 \r
  95         return frame.top_field_first ? core::field_mode::upper : core::field_mode::lower;\r
  96 }\r
  97 \r
  98 core::pixel_format::type get_pixel_format(PixelFormat pix_fmt)\r
  99 {\r
 100         switch(pix_fmt)\r
 101         {\r
 102         case CASPAR_PIX_FMT_LUMA:       return core::pixel_format::luma;\r
 103         case PIX_FMT_GRAY8:                     return core::pixel_format::gray;\r
 104         case PIX_FMT_BGRA:                      return core::pixel_format::bgra;\r
 105         case PIX_FMT_ARGB:                      return core::pixel_format::argb;\r
 106         case PIX_FMT_RGBA:                      return core::pixel_format::rgba;\r
 107         case PIX_FMT_ABGR:                      return core::pixel_format::abgr;\r
 108         case PIX_FMT_YUV444P:           return core::pixel_format::ycbcr;\r
 109         case PIX_FMT_YUV422P:           return core::pixel_format::ycbcr;\r
 110         case PIX_FMT_YUV420P:           return core::pixel_format::ycbcr;\r
 111         case PIX_FMT_YUV411P:           return core::pixel_format::ycbcr;\r
 112         case PIX_FMT_YUV410P:           return core::pixel_format::ycbcr;\r
 113         case PIX_FMT_YUVA420P:          return core::pixel_format::ycbcra;\r
 114         default:                                        return core::pixel_format::invalid;\r
 115         }\r
 116 }\r
 117 \r
 118 core::pixel_format_desc get_pixel_format_desc(PixelFormat pix_fmt, size_t width, size_t height)\r
 119 {\r
 120         // Get linesizes\r
 121         AVPicture dummy_pict;   \r
 122         avpicture_fill(&dummy_pict, nullptr, pix_fmt == CASPAR_PIX_FMT_LUMA ? PIX_FMT_GRAY8 : pix_fmt, width, height);\r
 123 \r
 124         core::pixel_format_desc desc;\r
 125         desc.pix_fmt = get_pixel_format(pix_fmt);\r
 126                 \r
 127         switch(desc.pix_fmt)\r
 128         {\r
 129         case core::pixel_format::gray:\r
 130         case core::pixel_format::luma:\r
 131                 {\r
 132                         desc.planes.push_back(core::pixel_format_desc::plane(dummy_pict.linesize[0], height, 1));                                               \r
 133                         return desc;\r
 134                 }\r
 135         case core::pixel_format::bgra:\r
 136         case core::pixel_format::argb:\r
 137         case core::pixel_format::rgba:\r
 138         case core::pixel_format::abgr:\r
 139                 {\r
 140                         desc.planes.push_back(core::pixel_format_desc::plane(dummy_pict.linesize[0]/4, height, 4));                                             \r
 141                         return desc;\r
 142                 }\r
 143         case core::pixel_format::ycbcr:\r
 144         case core::pixel_format::ycbcra:\r
 145                 {               \r
 146                         // Find chroma height\r
 147                         size_t size2 = dummy_pict.data[2] - dummy_pict.data[1];\r
 148                         size_t h2 = size2/dummy_pict.linesize[1];                       \r
 149 \r
 150                         desc.planes.push_back(core::pixel_format_desc::plane(dummy_pict.linesize[0], height, 1));\r
 151                         desc.planes.push_back(core::pixel_format_desc::plane(dummy_pict.linesize[1], h2, 1));\r
 152                         desc.planes.push_back(core::pixel_format_desc::plane(dummy_pict.linesize[2], h2, 1));\r
 153 \r
 154                         if(desc.pix_fmt == core::pixel_format::ycbcra)                                          \r
 155                                 desc.planes.push_back(core::pixel_format_desc::plane(dummy_pict.linesize[3], height, 1));       \r
 156                         return desc;\r
 157                 }               \r
 158         default:                \r
 159                 desc.pix_fmt = core::pixel_format::invalid;\r
 160                 return desc;\r
 161         }\r
 162 }\r
 163 \r
 164 int make_alpha_format(int format)\r
 165 {\r
 166         switch(get_pixel_format(static_cast<PixelFormat>(format)))\r
 167         {\r
 168         case core::pixel_format::ycbcr:\r
 169         case core::pixel_format::ycbcra:\r
 170                 return CASPAR_PIX_FMT_LUMA;\r
 171         default:\r
 172                 return format;\r
 173         }\r
 174 }\r
 175 \r
 176 safe_ptr<core::write_frame> make_write_frame(const void* tag, const safe_ptr<AVFrame>& decoded_frame, const safe_ptr<core::frame_factory>& frame_factory, int hints, const core::channel_layout& audio_channel_layout)\r
 177 {                       \r
 178         static tbb::concurrent_unordered_map<int64_t, tbb::concurrent_queue<std::shared_ptr<SwsContext>>> sws_contexts_;\r
 179         \r
 180         if(decoded_frame->width < 1 || decoded_frame->height < 1)\r
 181                 return make_safe<core::write_frame>(tag, audio_channel_layout);\r
 182 \r
 183         const auto width  = decoded_frame->width;\r
 184         const auto height = decoded_frame->height;\r
 185         auto desc                 = get_pixel_format_desc(static_cast<PixelFormat>(decoded_frame->format), width, height);\r
 186         \r
 187         if(hints & core::frame_producer::ALPHA_HINT)\r
 188                 desc = get_pixel_format_desc(static_cast<PixelFormat>(make_alpha_format(decoded_frame->format)), width, height);\r
 189 \r
 190         std::shared_ptr<core::write_frame> write;\r
 191 \r
 192         if(desc.pix_fmt == core::pixel_format::invalid)\r
 193         {\r
 194                 auto pix_fmt = static_cast<PixelFormat>(decoded_frame->format);\r
 195                 auto target_pix_fmt = PIX_FMT_BGRA;\r
 196 \r
 197                 if(pix_fmt == PIX_FMT_UYVY422)\r
 198                         target_pix_fmt = PIX_FMT_YUV422P;\r
 199                 else if(pix_fmt == PIX_FMT_YUYV422)\r
 200                         target_pix_fmt = PIX_FMT_YUV422P;\r
 201                 else if(pix_fmt == PIX_FMT_UYYVYY411)\r
 202                         target_pix_fmt = PIX_FMT_YUV411P;\r
 203                 else if(pix_fmt == PIX_FMT_YUV420P10)\r
 204                         target_pix_fmt = PIX_FMT_YUV420P;\r
 205                 else if(pix_fmt == PIX_FMT_YUV422P10)\r
 206                         target_pix_fmt = PIX_FMT_YUV422P;\r
 207                 else if(pix_fmt == PIX_FMT_YUV444P10)\r
 208                         target_pix_fmt = PIX_FMT_YUV444P;\r
 209                 \r
 210                 auto target_desc = get_pixel_format_desc(target_pix_fmt, width, height);\r
 211 \r
 212                 write = frame_factory->create_frame(tag, target_desc, audio_channel_layout);\r
 213                 write->set_type(get_mode(*decoded_frame));\r
 214 \r
 215                 std::shared_ptr<SwsContext> sws_context;\r
 216 \r
 217                 //CASPAR_LOG(warning) << "Hardware accelerated color transform not supported.";\r
 218                 \r
 219                 int64_t key = ((static_cast<int64_t>(width)                      << 32) & 0xFFFF00000000) | \r
 220                                           ((static_cast<int64_t>(height)                 << 16) & 0xFFFF0000) | \r
 221                                           ((static_cast<int64_t>(pix_fmt)                <<  8) & 0xFF00) | \r
 222                                           ((static_cast<int64_t>(target_pix_fmt) <<  0) & 0xFF);\r
 223                         \r
 224                 auto& pool = sws_contexts_[key];\r
 225                                                 \r
 226                 if(!pool.try_pop(sws_context))\r
 227                 {\r
 228                         double param;\r
 229                         sws_context.reset(sws_getContext(width, height, pix_fmt, width, height, target_pix_fmt, SWS_BILINEAR, nullptr, nullptr, &param), sws_freeContext);\r
 230                 }\r
 231                         \r
 232                 if(!sws_context)\r
 233                 {\r
 234                         BOOST_THROW_EXCEPTION(operation_failed() << msg_info("Could not create software scaling context.") << \r
 235                                                                         boost::errinfo_api_function("sws_getContext"));\r
 236                 }       \r
 237                 \r
 238                 safe_ptr<AVFrame> av_frame(avcodec_alloc_frame(), av_free);     \r
 239                 avcodec_get_frame_defaults(av_frame.get());                     \r
 240                 if(target_pix_fmt == PIX_FMT_BGRA)\r
 241                 {\r
 242                         auto size = avpicture_fill(reinterpret_cast<AVPicture*>(av_frame.get()), write->image_data().begin(), PIX_FMT_BGRA, width, height);\r
 243                         CASPAR_VERIFY(size == write->image_data().size()); \r
 244                 }\r
 245                 else\r
 246                 {\r
 247                         av_frame->width  = width;\r
 248                         av_frame->height = height;\r
 249                         for(size_t n = 0; n < target_desc.planes.size(); ++n)\r
 250                         {\r
 251                                 av_frame->data[n]               = write->image_data(n).begin();\r
 252                                 av_frame->linesize[n]   = target_desc.planes[n].linesize;\r
 253                         }\r
 254                 }\r
 255 \r
 256                 sws_scale(sws_context.get(), decoded_frame->data, decoded_frame->linesize, 0, height, av_frame->data, av_frame->linesize);      \r
 257                 pool.push(sws_context);\r
 258 \r
 259                 write->commit();                \r
 260         }\r
 261         else\r
 262         {\r
 263                 write = frame_factory->create_frame(tag, desc, audio_channel_layout);\r
 264                 write->set_type(get_mode(*decoded_frame));\r
 265 \r
 266                 for(int n = 0; n < static_cast<int>(desc.planes.size()); ++n)\r
 267                 {\r
 268                         auto plane            = desc.planes[n];\r
 269                         auto result           = write->image_data(n).begin();\r
 270                         auto decoded          = decoded_frame->data[n];\r
 271                         auto decoded_linesize = decoded_frame->linesize[n];\r
 272                         \r
 273                         CASPAR_ASSERT(decoded);\r
 274                         CASPAR_ASSERT(write->image_data(n).begin());\r
 275 \r
 276                         if(decoded_linesize != static_cast<int>(plane.width))\r
 277                         {\r
 278                                 // Copy line by line since ffmpeg sometimes pads each line.\r
 279                                 tbb::parallel_for<size_t>(0, desc.planes[n].height, [&](size_t y)\r
 280                                 {\r
 281                                         fast_memcpy(result + y*plane.linesize, decoded + y*decoded_linesize, plane.linesize);\r
 282                                 });\r
 283                         }\r
 284                         else\r
 285                         {\r
 286                                 fast_memcpy(result, decoded, plane.size);\r
 287                         }\r
 288 \r
 289                         write->commit(n);\r
 290                 }\r
 291         }\r
 292 \r
 293         if(decoded_frame->height == 480) // NTSC DV\r
 294         {\r
 295                 write->get_frame_transform().fill_translation[1] += 2.0/static_cast<double>(frame_factory->get_video_format_desc().height);\r
 296                 write->get_frame_transform().fill_scale[1] = 1.0 - 6.0*1.0/static_cast<double>(frame_factory->get_video_format_desc().height);\r
 297         }\r
 298         \r
 299         // Fix field-order if needed\r
 300         if(write->get_type() == core::field_mode::lower && frame_factory->get_video_format_desc().field_mode == core::field_mode::upper)\r
 301                 write->get_frame_transform().fill_translation[1] += 1.0/static_cast<double>(frame_factory->get_video_format_desc().height);\r
 302         else if(write->get_type() == core::field_mode::upper && frame_factory->get_video_format_desc().field_mode == core::field_mode::lower)\r
 303                 write->get_frame_transform().fill_translation[1] -= 1.0/static_cast<double>(frame_factory->get_video_format_desc().height);\r
 304 \r
 305         return make_safe_ptr(write);\r
 306 }\r
 307 \r
 308 bool is_sane_fps(AVRational time_base)\r
 309 {\r
 310         double fps = static_cast<double>(time_base.den) / static_cast<double>(time_base.num);\r
 311         return fps > 20.0 && fps < 65.0;\r
 312 }\r
 313 \r
 314 AVRational fix_time_base(AVRational time_base)\r
 315 {\r
 316         if(time_base.num == 1)\r
 317                 time_base.num = static_cast<int>(std::pow(10.0, static_cast<int>(std::log10(static_cast<float>(time_base.den)))-1));    \r
 318                         \r
 319         if(!is_sane_fps(time_base))\r
 320         {\r
 321                 auto tmp = time_base;\r
 322                 tmp.den /= 2;\r
 323                 if(is_sane_fps(tmp))\r
 324                         time_base = tmp;\r
 325         }\r
 326 \r
 327         return time_base;\r
 328 }\r
 329 \r
 330 double read_fps(AVFormatContext& context, double fail_value)\r
 331 {                                               \r
 332         auto video_index = av_find_best_stream(&context, AVMEDIA_TYPE_VIDEO, -1, -1, 0, 0);\r
 333         auto audio_index = av_find_best_stream(&context, AVMEDIA_TYPE_AUDIO, -1, -1, 0, 0);\r
 334         \r
 335         if(video_index > -1)\r
 336         {\r
 337                 const auto video_context = context.streams[video_index]->codec;\r
 338                 const auto video_stream  = context.streams[video_index];\r
 339                                                 \r
 340                 AVRational time_base = video_context->time_base;\r
 341 \r
 342                 if(boost::filesystem2::path(context.filename).extension() == ".flv")\r
 343                 {\r
 344                         try\r
 345                         {\r
 346                                 auto meta = read_flv_meta_info(context.filename);\r
 347                                 return boost::lexical_cast<double>(meta["framerate"]);\r
 348                         }\r
 349                         catch(...)\r
 350                         {\r
 351                                 return 0.0;\r
 352                         }\r
 353                 }\r
 354                 else\r
 355                 {\r
 356                         time_base.num *= video_context->ticks_per_frame;\r
 357 \r
 358                         if(!is_sane_fps(time_base))\r
 359                         {                       \r
 360                                 time_base = fix_time_base(time_base);\r
 361 \r
 362                                 if(!is_sane_fps(time_base) && audio_index > -1)\r
 363                                 {\r
 364                                         auto& audio_context = *context.streams[audio_index]->codec;\r
 365                                         auto& audio_stream  = *context.streams[audio_index];\r
 366 \r
 367                                         double duration_sec = audio_stream.duration / static_cast<double>(audio_context.sample_rate);\r
 368                                                                 \r
 369                                         time_base.num = static_cast<int>(duration_sec*100000.0);\r
 370                                         time_base.den = static_cast<int>(video_stream->nb_frames*100000);\r
 371                                 }\r
 372                         }\r
 373                 }\r
 374                 \r
 375                 double fps = static_cast<double>(time_base.den) / static_cast<double>(time_base.num);\r
 376 \r
 377                 double closest_fps = 0.0;\r
 378                 for(int n = 0; n < core::video_format::count; ++n)\r
 379                 {\r
 380                         auto format = core::video_format_desc::get(static_cast<core::video_format::type>(n));\r
 381 \r
 382                         double diff1 = std::abs(format.fps - fps);\r
 383                         double diff2 = std::abs(closest_fps - fps);\r
 384 \r
 385                         if(diff1 < diff2)\r
 386                                 closest_fps = format.fps;\r
 387                 }\r
 388         \r
 389                 return closest_fps;\r
 390         }\r
 391 \r
 392         return fail_value;      \r
 393 }\r
 394 \r
 395 void fix_meta_data(AVFormatContext& context)\r
 396 {\r
 397         auto video_index = av_find_best_stream(&context, AVMEDIA_TYPE_VIDEO, -1, -1, 0, 0);\r
 398 \r
 399         if(video_index > -1)\r
 400         {\r
 401                 auto video_stream   = context.streams[video_index];\r
 402                 auto video_context  = context.streams[video_index]->codec;\r
 403                                                 \r
 404                 if(boost::filesystem2::path(context.filename).extension() == ".flv")\r
 405                 {\r
 406                         try\r
 407                         {\r
 408                                 auto meta = read_flv_meta_info(context.filename);\r
 409                                 double fps = boost::lexical_cast<double>(meta["framerate"]);\r
 410                                 video_stream->nb_frames = static_cast<int64_t>(boost::lexical_cast<double>(meta["duration"])*fps);\r
 411                         }\r
 412                         catch(...){}\r
 413                 }\r
 414                 else\r
 415                 {\r
 416                         auto stream_time = video_stream->time_base;\r
 417                         auto duration    = video_stream->duration;\r
 418                         auto codec_time  = video_context->time_base;\r
 419                         auto ticks               = video_context->ticks_per_frame;\r
 420 \r
 421                         if(video_stream->nb_frames == 0)\r
 422                                 video_stream->nb_frames = (duration*stream_time.num*codec_time.den)/(stream_time.den*codec_time.num*ticks);     \r
 423                 }\r
 424         }\r
 425 }\r
 426 \r
 427 safe_ptr<AVPacket> create_packet()\r
 428 {\r
 429         safe_ptr<AVPacket> packet(new AVPacket, [](AVPacket* p)\r
 430         {\r
 431                 av_free_packet(p);\r
 432                 delete p;\r
 433         });\r
 434         \r
 435         av_init_packet(packet.get());\r
 436         return packet;\r
 437 }\r
 438 \r
 439 safe_ptr<AVCodecContext> open_codec(AVFormatContext& context, enum AVMediaType type, int& index)\r
 440 {       \r
 441         AVCodec* decoder;\r
 442         index = THROW_ON_ERROR2(av_find_best_stream(&context, type, -1, -1, &decoder, 0), "");\r
 443         //if(strcmp(decoder->name, "prores") == 0 && decoder->next && strcmp(decoder->next->name, "prores_lgpl") == 0)\r
 444         //      decoder = decoder->next;\r
 445 \r
 446         THROW_ON_ERROR2(tbb_avcodec_open(context.streams[index]->codec, decoder), "");\r
 447         return safe_ptr<AVCodecContext>(context.streams[index]->codec, tbb_avcodec_close);\r
 448 }\r
 449 \r
 450 safe_ptr<AVFormatContext> open_input(const std::wstring& filename)\r
 451 {\r
 452         AVFormatContext* weak_context = nullptr;\r
 453         THROW_ON_ERROR2(avformat_open_input(&weak_context, narrow(filename).c_str(), nullptr, nullptr), filename);\r
 454         safe_ptr<AVFormatContext> context(weak_context, av_close_input_file);                   \r
 455         THROW_ON_ERROR2(avformat_find_stream_info(weak_context, nullptr), filename);\r
 456         fix_meta_data(*context);\r
 457         return context;\r
 458 }\r
 459 \r
 460 std::wstring print_mode(size_t width, size_t height, double fps, bool interlaced)\r
 461 {\r
 462         std::wostringstream fps_ss;\r
 463         fps_ss << std::fixed << std::setprecision(2) << (!interlaced ? fps : 2.0 * fps);\r
 464 \r
 465         return boost::lexical_cast<std::wstring>(width) + L"x" + boost::lexical_cast<std::wstring>(height) + (!interlaced ? L"p" : L"i") + fps_ss.str();\r
 466 }\r
 467 \r
 468 bool is_valid_file(const std::wstring filename, const std::vector<std::wstring>& invalid_exts)\r
 469 {\r
 470         static std::vector<std::wstring> valid_exts = boost::assign::list_of(L".m2t")(L".mov")(L".mp4")(L".dv")(L".flv")(L".mpg")(L".wav")(L".mp3")(L".dnxhd")(L".h264")(L".prores");\r
 471 \r
 472         auto ext = boost::to_lower_copy(boost::filesystem::wpath(filename).extension());\r
 473                 \r
 474         if(std::find(invalid_exts.begin(), invalid_exts.end(), ext) != invalid_exts.end())\r
 475                 return false;   \r
 476 \r
 477         if(std::find(valid_exts.begin(), valid_exts.end(), ext) != valid_exts.end())\r
 478                 return true;    \r
 479 \r
 480         auto filename2 = narrow(filename);\r
 481 \r
 482         if(boost::filesystem::path(filename2).extension() == ".m2t")\r
 483                 return true;\r
 484 \r
 485         std::ifstream file(filename);\r
 486 \r
 487         std::vector<unsigned char> buf;\r
 488         for(auto file_it = std::istreambuf_iterator<char>(file); file_it != std::istreambuf_iterator<char>() && buf.size() < 2048; ++file_it)\r
 489                 buf.push_back(*file_it);\r
 490 \r
 491         if(buf.empty())\r
 492                 return nullptr;\r
 493 \r
 494         AVProbeData pb;\r
 495         pb.filename = filename2.c_str();\r
 496         pb.buf          = buf.data();\r
 497         pb.buf_size = buf.size();\r
 498 \r
 499         int score = 0;\r
 500         return av_probe_input_format2(&pb, true, &score) != nullptr;\r
 501 }\r
 502 \r
 503 bool is_valid_file(const std::wstring filename)\r
 504 {\r
 505         static const std::vector<std::wstring> invalid_exts = boost::assign::list_of(L".png")(L".tga")(L".bmp")(L".jpg")(L".jpeg")(L".gif")(L".tiff")(L".tif")(L".jp2")(L".jpx")(L".j2k")(L".j2c")(L".swf")(L".ct");\r
 506         \r
 507         return is_valid_file(filename, invalid_exts);\r
 508 }\r
 509 \r
 510 std::wstring probe_stem(const std::wstring stem, const std::vector<std::wstring>& invalid_exts)\r
 511 {\r
 512         auto stem2 = boost::filesystem2::wpath(stem);\r
 513         auto dir = stem2.parent_path();\r
 514         for(auto it = boost::filesystem2::wdirectory_iterator(dir); it != boost::filesystem2::wdirectory_iterator(); ++it)\r
 515         {\r
 516                 if(boost::iequals(it->path().stem(), stem2.filename()) && is_valid_file(it->path().file_string(), invalid_exts))\r
 517                         return it->path().file_string();\r
 518         }\r
 519         return L"";\r
 520 }\r
 521 \r
 522 core::channel_layout get_audio_channel_layout(\r
 523                 const AVCodecContext& context, const std::wstring& custom_channel_order)\r
 524 {\r
 525         if (!custom_channel_order.empty())\r
 526         {\r
 527                 auto layout = core::create_custom_channel_layout(\r
 528                                 custom_channel_order,\r
 529                                 core::default_channel_layout_repository());\r
 530 \r
 531                 layout.num_channels = context.channels;\r
 532 \r
 533                 return layout;\r
 534         }\r
 535 \r
 536         int64_t ch_layout = context.channel_layout;\r
 537 \r
 538         if (ch_layout == 0)\r
 539                 ch_layout = av_get_default_channel_layout(context.channels);\r
 540 \r
 541         switch (ch_layout) // TODO: refine this auto-detection\r
 542         {\r
 543         case AV_CH_LAYOUT_MONO:\r
 544                 return core::default_channel_layout_repository().get_by_name(L"MONO");\r
 545         case AV_CH_LAYOUT_STEREO:\r
 546                 return core::default_channel_layout_repository().get_by_name(L"STEREO");\r
 547         case AV_CH_LAYOUT_5POINT1:\r
 548         case AV_CH_LAYOUT_5POINT1_BACK:\r
 549                 return core::default_channel_layout_repository().get_by_name(L"SMPTE");\r
 550         case AV_CH_LAYOUT_7POINT1:\r
 551                 return core::default_channel_layout_repository().get_by_name(L"DOLBYE");\r
 552         }\r
 553 \r
 554         return core::create_unspecified_layout(context.channels);\r
 555 }\r
 556 \r
 557 //\r
 558 //void av_dup_frame(AVFrame* frame)\r
 559 //{\r
 560 //      AVFrame* new_frame = avcodec_alloc_frame();\r
 561 //\r
 562 //\r
 563 //      const uint8_t *src_data[4] = {0};\r
 564 //      memcpy(const_cast<uint8_t**>(&src_data[0]), frame->data, 4);\r
 565 //      const int src_linesizes[4] = {0};\r
 566 //      memcpy(const_cast<int*>(&src_linesizes[0]), frame->linesize, 4);\r
 567 //\r
 568 //      av_image_alloc(new_frame->data, new_frame->linesize, new_frame->width, new_frame->height, frame->format, 16);\r
 569 //\r
 570 //      av_image_copy(new_frame->data, new_frame->linesize, src_data, src_linesizes, frame->format, new_frame->width, new_frame->height);\r
 571 //\r
 572 //      frame =\r
 573 //}\r
 574 \r
 575 }}