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

   1 /*
   2 * Copyright 2013 Sveriges Television AB http://casparcg.com/
   3 *
   4 * This file is part of CasparCG (www.casparcg.com).
   5 *
   6 * CasparCG 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 3 of the License, or
   9 * (at your option) any later version.
  10 *
  11 * CasparCG 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 CasparCG. If not, see <http://www.gnu.org/licenses/>.
  18 *
  19 * Author: Robert Nagy, ronag89@gmail.com
  20 */
  21
  22 #include "../../StdAfx.h"
  23
  24 #include "audio_filter.h"
  25
  26 #include "../../ffmpeg_error.h"
  27 #include "../../ffmpeg.h"
  28 #include "../util/util.h"
  29
  30 #include <common/assert.h>
  31 #include <common/except.h>
  32
  33 #include <boost/algorithm/string.hpp>
  34 #include <boost/thread.hpp>
  35 #include <boost/format.hpp>
  36 #include <boost/rational.hpp>
  37
  38 #include <cstdio>
  39 #include <sstream>
  40 #include <string>
  41 #include <algorithm>
  42
  43 #if defined(_MSC_VER)
  44 #pragma warning (push)
  45 #pragma warning (disable : 4244)
  46 #endif
  47 extern "C"
  48 {
  49         #include <libavutil/avutil.h>
  50         #include <libavutil/imgutils.h>
  51         #include <libavutil/opt.h>
  52         #include <libavfilter/avfilter.h>
  53         #include <libavfilter/avcodec.h>
  54         #include <libavfilter/buffersink.h>
  55         #include <libavfilter/buffersrc.h>
  56 }
  57 #if defined(_MSC_VER)
  58 #pragma warning (pop)
  59 #endif
  60
  61 namespace caspar { namespace ffmpeg {
  62
  63 std::string create_sourcefilter_str(const audio_input_pad& input_pad, std::string name)
  64 {
  65         const auto asrc_options = (boost::format("abuffer=time_base=%1%/%2%:sample_rate=%3%:sample_fmt=%4%:channel_layout=0x%|5$x| [%6%]")
  66                 % input_pad.time_base.numerator() % input_pad.time_base.denominator()
  67                 % input_pad.sample_rate
  68                 % av_get_sample_fmt_name(input_pad.sample_fmt)
  69                 % input_pad.audio_channel_layout
  70                 % name).str();
  71
  72         return asrc_options;
  73 }
  74
  75 std::string create_filter_list(const std::vector<std::string>& items)
  76 {
  77         return boost::join(items, "|");
  78 }
  79
  80 std::string channel_layout_to_string(uint64_t channel_layout)
  81 {
  82         return (boost::format("0x%|1$x|") % channel_layout).str();
  83 }
  84
  85 std::string create_sinkfilter_str(const audio_output_pad& output_pad, std::string name)
  86 {
  87         const auto asink_options = (boost::format("[%4%] abuffersink")//=sample_fmts=%1%:channel_layouts=%2%:sample_rates=%3%")
  88                 % create_filter_list(cpplinq::from(output_pad.sample_fmts)
  89                                 .select(&av_get_sample_fmt_name)
  90                                 .select([](const char* str) { return std::string(str); })
  91                                 .to_vector())
  92                 % create_filter_list(cpplinq::from(output_pad.sample_fmts)
  93                                 .select(&channel_layout_to_string)
  94                                 .to_vector())
  95                 % create_filter_list(cpplinq::from(output_pad.sample_rates)
  96                                 .select([](int samplerate) { return boost::lexical_cast<std::string>(samplerate); })
  97                                 .to_vector())
  98                 % name).str();
  99
 100         return asink_options;
 101 }
 102
 103 struct audio_filter::implementation
 104 {
 105         std::string                                             filtergraph_;
 106
 107         std::shared_ptr<AVFilterGraph>  audio_graph_;
 108         std::vector<AVFilterContext*>   audio_graph_inputs_;
 109         std::vector<AVFilterContext*>   audio_graph_outputs_;
 110
 111         std::vector<audio_input_pad>    input_pads_;
 112
 113         implementation(
 114                 std::vector<audio_input_pad> input_pads,
 115                 std::vector<audio_output_pad> output_pads,
 116                 const std::string& filtergraph)
 117                 : filtergraph_(boost::to_lower_copy(filtergraph))
 118                 , input_pads_(std::move(input_pads))
 119         {
 120                 if (input_pads_.empty())
 121                         CASPAR_THROW_EXCEPTION(invalid_argument() << msg_info("input_pads cannot be empty"));
 122
 123                 if (output_pads.empty())
 124                         CASPAR_THROW_EXCEPTION(invalid_argument() << msg_info("output_pads cannot be empty"));
 125
 126                 audio_graph_.reset(
 127                         avfilter_graph_alloc(),
 128                         [](AVFilterGraph* p)
 129                 {
 130                         avfilter_graph_free(&p);
 131                 });
 132
 133                 std::vector<std::string> complete_filter_graph;
 134
 135                 {
 136                         int i = 0;
 137                         for (auto& input_pad : input_pads_)
 138                                 complete_filter_graph.push_back(create_sourcefilter_str(input_pad, "a:" + boost::lexical_cast<std::string>(i++)));
 139                 }
 140
 141                 if (filtergraph_.empty())
 142                         complete_filter_graph.push_back("[a:0] anull [aout:0]");
 143                 else
 144                         complete_filter_graph.push_back(filtergraph_);
 145
 146                 {
 147                         int i = 0;
 148                         for (auto& output_pad : output_pads)
 149                                 complete_filter_graph.push_back(create_sinkfilter_str(output_pad, "aout:" + boost::lexical_cast<std::string>(i++)));
 150                 }
 151
 152                 configure_filtergraph(
 153                                 *audio_graph_,
 154                                 boost::join(complete_filter_graph, ";"),
 155                                 audio_graph_inputs_,
 156                                 audio_graph_outputs_);
 157
 158                 if (is_logging_quiet_for_thread())
 159                         CASPAR_LOG(trace)
 160                                 <<      u16(std::string("\n")
 161                                         + avfilter_graph_dump(
 162                                                         audio_graph_.get(),
 163                                                         nullptr));
 164                 else
 165                         CASPAR_LOG(debug)
 166                                 << u16(std::string("\n")
 167                                         + avfilter_graph_dump(
 168                                                 audio_graph_.get(),
 169                                                 nullptr));
 170         }
 171
 172         void configure_filtergraph(
 173                         AVFilterGraph& graph,
 174                         const std::string& filtergraph,
 175                         std::vector<AVFilterContext*>& source_contexts,
 176                         std::vector<AVFilterContext*>& sink_contexts)
 177         {
 178                 AVFilterInOut* outputs  = nullptr;
 179                 AVFilterInOut* inputs   = nullptr;
 180
 181                 FF(avfilter_graph_parse2(
 182                                 &graph,
 183                                 filtergraph.c_str(),
 184                                 &inputs,
 185                                 &outputs));
 186
 187                 // Workaround because outputs and inputs are not filled in for some reason
 188                 for (unsigned i = 0; i < graph.nb_filters; ++i)
 189                 {
 190                         auto filter = graph.filters[i];
 191
 192                         if (std::string(filter->filter->name) == "abuffer")
 193                                 source_contexts.push_back(filter);
 194
 195                         if (std::string(filter->filter->name) == "abuffersink")
 196                                 sink_contexts.push_back(filter);
 197                 }
 198
 199                 for (AVFilterInOut* iter = inputs; iter; iter = iter->next)
 200                         source_contexts.push_back(iter->filter_ctx);
 201
 202                 for (AVFilterInOut* iter = outputs; iter; iter = iter->next)
 203                         sink_contexts.push_back(iter->filter_ctx);
 204
 205                 FF(avfilter_graph_config(
 206                         &graph,
 207                         nullptr));
 208         }
 209
 210         void push(int input_pad_id, const std::shared_ptr<AVFrame>& src_av_frame)
 211         {
 212                 FF(av_buffersrc_add_frame(
 213                         audio_graph_inputs_.at(input_pad_id),
 214                         src_av_frame.get()));
 215         }
 216
 217         void push(int input_pad_id, const boost::iterator_range<const int32_t*>& frame_samples)
 218         {
 219                 auto& input_pad                         = input_pads_.at(input_pad_id);
 220                 auto num_samples                        = frame_samples.size() / av_get_channel_layout_nb_channels(input_pad.audio_channel_layout);
 221                 auto input_frame                        = ffmpeg::create_frame();
 222
 223                 input_frame->channels           = av_get_channel_layout_nb_channels(input_pad.audio_channel_layout);
 224                 input_frame->channel_layout     = input_pad.audio_channel_layout;
 225                 input_frame->sample_rate                = input_pad.sample_rate;
 226                 input_frame->nb_samples         = static_cast<int>(num_samples);
 227                 input_frame->format                     = input_pad.sample_fmt;
 228                 input_frame->pts                                = 0;
 229
 230                 av_samples_fill_arrays(
 231                                 input_frame->extended_data,
 232                                 input_frame->linesize,
 233                                 reinterpret_cast<const std::uint8_t*>(frame_samples.begin()),
 234                                 input_frame->channels,
 235                                 input_frame->nb_samples,
 236                                 static_cast<AVSampleFormat>(input_frame->format),
 237                                 16);
 238
 239                 push(input_pad_id, input_frame);
 240         }
 241
 242         std::shared_ptr<AVFrame> poll(int output_pad_id)
 243         {
 244                 auto filt_frame = create_frame();
 245
 246                 const auto ret = av_buffersink_get_frame(
 247                         audio_graph_outputs_.at(output_pad_id),
 248                         filt_frame.get());
 249
 250                 if(ret == AVERROR_EOF || ret == AVERROR(EAGAIN))
 251                         return nullptr;
 252
 253                 FF_RET(ret, "poll");
 254
 255                 return filt_frame;
 256         }
 257 };
 258
 259 audio_filter::audio_filter(
 260                 std::vector<audio_input_pad> input_pads,
 261                 std::vector<audio_output_pad> output_pads,
 262                 const std::string& filtergraph)
 263         : impl_(new implementation(std::move(input_pads), std::move(output_pads), filtergraph))
 264 {
 265 }
 266 audio_filter::audio_filter(audio_filter&& other) : impl_(std::move(other.impl_)){}
 267 audio_filter& audio_filter::operator=(audio_filter&& other){impl_ = std::move(other.impl_); return *this;}
 268 void audio_filter::push(int input_pad_id, const std::shared_ptr<AVFrame>& frame){impl_->push(input_pad_id, frame);}
 269 void audio_filter::push(int input_pad_id, const boost::iterator_range<const int32_t*>& frame_samples) { impl_->push(input_pad_id, frame_samples); }
 270 std::shared_ptr<AVFrame> audio_filter::poll(int output_pad_id){return impl_->poll(output_pad_id);}
 271 std::wstring audio_filter::filter_str() const{return u16(impl_->filtergraph_);}
 272 std::vector<spl::shared_ptr<AVFrame>> audio_filter::poll_all(int output_pad_id)
 273 {
 274         std::vector<spl::shared_ptr<AVFrame>> frames;
 275         for(auto frame = poll(output_pad_id); frame; frame = poll(output_pad_id))
 276                 frames.push_back(spl::make_shared_ptr(frame));
 277         return frames;
 278 }
 279
 280 }}