// // FFMPEGDecoder.m // FFMPEG // // Created by Andre Reffhaug on 2/26/08. // Copyright 2008 __MyCompanyName__. All rights reserved. // // test #import "FFMPEGDecoder.h" #include #import "Logging.h" #define ST_BUFF 2048 int ffmpeg_read(void *opaque, uint8_t *buf, int buf_size) { id source = (__bridge id)opaque; long sizeRead = [source read:buf amount:buf_size]; if(sizeRead == 0) return AVERROR_EOF; return (int)sizeRead; } int ffmpeg_write(void *opaque, uint8_t *buf, int buf_size) { return -1; } int64_t ffmpeg_seek(void *opaque, int64_t offset, int whence) { id source = (__bridge id)opaque; if(whence & AVSEEK_SIZE) { if([source seekable]) { int64_t curOffset = [source tell]; [source seek:0 whence:SEEK_END]; int64_t size = [source tell]; [source seek:curOffset whence:SEEK_SET]; return size; } return -1; } whence &= ~(AVSEEK_SIZE | AVSEEK_FORCE); return [source seekable] ? ([source seek:offset whence:whence] ? [source tell] : -1) : -1; } @implementation FFMPEGDecoder + (void)initialize { if(self == [FFMPEGDecoder class]) { av_log_set_flags(AV_LOG_SKIP_REPEATED); av_log_set_level(AV_LOG_ERROR); } } - (id)init { self = [super init]; if(self) { lastReadPacket = NULL; lastDecodedFrame = NULL; codecCtx = NULL; formatCtx = NULL; ioCtx = NULL; buffer = NULL; } return self; } - (BOOL)open:(id)s { int errcode, i; AVStream *stream; source = s; formatCtx = NULL; totalFrames = 0; framesRead = 0; BOOL isStream = NO; // register all available codecs NSURL *url = [s url]; if([[url scheme] isEqualToString:@"http"] || [[url scheme] isEqualToString:@"https"]) { source = nil; [s close]; isStream = YES; formatCtx = avformat_alloc_context(); if(!formatCtx) { ALog(@"Unable to allocate AVFormat context"); return NO; } NSString *urlString = [url absoluteString]; if((errcode = avformat_open_input(&formatCtx, [urlString UTF8String], NULL, NULL)) < 0) { char errDescr[4096]; av_strerror(errcode, errDescr, 4096); ALog(@"Error opening file, errcode = %d, error = %s", errcode, errDescr); return NO; } } else { buffer = av_malloc(32 * 1024); if(!buffer) { ALog(@"Out of memory!"); return NO; } ioCtx = avio_alloc_context(buffer, 32 * 1024, 0, (__bridge void *)source, ffmpeg_read, ffmpeg_write, ffmpeg_seek); if(!ioCtx) { ALog(@"Unable to create AVIO context"); return NO; } formatCtx = avformat_alloc_context(); if(!formatCtx) { ALog(@"Unable to allocate AVFormat context"); return NO; } formatCtx->pb = ioCtx; if((errcode = avformat_open_input(&formatCtx, "", NULL, NULL)) < 0) { char errDescr[4096]; av_strerror(errcode, errDescr, 4096); ALog(@"Error opening file, errcode = %d, error = %s", errcode, errDescr); return NO; } } if((errcode = avformat_find_stream_info(formatCtx, NULL)) < 0) { char errDescr[4096]; av_strerror(errcode, errDescr, 4096); ALog(@"Can't find stream info, errcode = %d, error = %s", errcode, errDescr); return NO; } streamIndex = -1; AVCodecParameters *codecPar; for(i = 0; i < formatCtx->nb_streams; i++) { stream = formatCtx->streams[i]; codecPar = stream->codecpar; if(streamIndex < 0 && codecPar->codec_type == AVMEDIA_TYPE_AUDIO) { DLog(@"audio codec found"); streamIndex = i; } else { stream->discard = AVDISCARD_ALL; } } if(streamIndex < 0) { ALog(@"no audio codec found"); return NO; } stream = formatCtx->streams[streamIndex]; codecPar = stream->codecpar; codecCtx = avcodec_alloc_context3(NULL); if(!codecCtx) { ALog(@"could not allocate codec context"); return NO; } if((errcode = avcodec_parameters_to_context(codecCtx, codecPar)) < 0) { char errDescr[4096]; av_strerror(errcode, errDescr, 4096); ALog(@"Can't copy codec parameters to context, errcode = %d, error = %s", errcode, errDescr); return NO; } enum AVCodecID codec_id = codecCtx->codec_id; const AVCodec *codec = NULL; AVDictionary *dict = NULL; switch(codec_id) { case AV_CODEC_ID_MP3: codec = avcodec_find_decoder_by_name("mp3float"); break; case AV_CODEC_ID_MP2: codec = avcodec_find_decoder_by_name("mp2float"); break; case AV_CODEC_ID_MP1: codec = avcodec_find_decoder_by_name("mp1float"); break; case AV_CODEC_ID_AAC: codec = avcodec_find_decoder_by_name("libfdk_aac"); av_dict_set_int(&dict, "drc_level", -2, 0); // disable DRC av_dict_set_int(&dict, "level_limit", 0, 0); // disable peak limiting break; case AV_CODEC_ID_ALAC: codec = avcodec_find_decoder_by_name("alac"); break; case AV_CODEC_ID_AC3: codec = avcodec_find_decoder_by_name("ac3"); break; case AV_CODEC_ID_EAC3: codec = avcodec_find_decoder_by_name("eac3"); break; default: break; } if(!codec) codec = avcodec_find_decoder(codec_id); if(@available(macOS 10.15, *)) { } else { if(codec && codec->name) { const char *name = codec->name; size_t name_len = strlen(name); if(name_len > 3) { name += name_len - 3; if(!strcmp(name, "_at")) { ALog(@"AudioToolbox decoder picked on old macOS, disabling: %s", codec->name); codec = NULL; // Disable AudioToolbox codecs on Mojave and older } } } } if(!codec) { ALog(@"codec not found"); av_dict_free(&dict); return NO; } if((errcode = avcodec_open2(codecCtx, codec, &dict)) < 0) { char errDescr[4096]; av_dict_free(&dict); av_strerror(errcode, errDescr, 4096); ALog(@"could not open codec, errcode = %d, error = %s", errcode, errDescr); return NO; } av_dict_free(&dict); // Bah, their skipping is broken codecCtx->flags2 |= AV_CODEC_FLAG2_SKIP_MANUAL; lastDecodedFrame = av_frame_alloc(); av_frame_unref(lastDecodedFrame); lastReadPacket = malloc(sizeof(AVPacket)); av_new_packet(lastReadPacket, 0); readNextPacket = YES; bytesConsumedFromDecodedFrame = INT_MAX; seekFrame = -1; frequency = codecCtx->sample_rate; channels = codecCtx->channels; channelConfig = (uint32_t)codecCtx->channel_layout; floatingPoint = NO; switch(codecCtx->sample_fmt) { case AV_SAMPLE_FMT_U8: case AV_SAMPLE_FMT_U8P: bitsPerSample = 8; break; case AV_SAMPLE_FMT_S16: case AV_SAMPLE_FMT_S16P: bitsPerSample = 16; break; case AV_SAMPLE_FMT_S32: case AV_SAMPLE_FMT_S32P: bitsPerSample = 32; break; case AV_SAMPLE_FMT_FLT: case AV_SAMPLE_FMT_FLTP: bitsPerSample = 32; floatingPoint = YES; break; case AV_SAMPLE_FMT_DBL: case AV_SAMPLE_FMT_DBLP: bitsPerSample = 64; floatingPoint = YES; break; default: return NO; } lossy = NO; if(floatingPoint) lossy = YES; if(!floatingPoint) { switch(codec_id) { case AV_CODEC_ID_MP2: case AV_CODEC_ID_MP3: case AV_CODEC_ID_AAC: case AV_CODEC_ID_AC3: // case AV_CODEC_ID_DTS: // lossy will return float, caught above, lossless will be integer case AV_CODEC_ID_VORBIS: case AV_CODEC_ID_DVAUDIO: case AV_CODEC_ID_WMAV1: case AV_CODEC_ID_WMAV2: case AV_CODEC_ID_MACE3: case AV_CODEC_ID_MACE6: case AV_CODEC_ID_VMDAUDIO: case AV_CODEC_ID_MP3ADU: case AV_CODEC_ID_MP3ON4: case AV_CODEC_ID_WESTWOOD_SND1: case AV_CODEC_ID_GSM: case AV_CODEC_ID_QDM2: case AV_CODEC_ID_COOK: case AV_CODEC_ID_TRUESPEECH: case AV_CODEC_ID_SMACKAUDIO: case AV_CODEC_ID_QCELP: case AV_CODEC_ID_DSICINAUDIO: case AV_CODEC_ID_IMC: case AV_CODEC_ID_MUSEPACK7: case AV_CODEC_ID_MLP: case AV_CODEC_ID_GSM_MS: case AV_CODEC_ID_ATRAC3: case AV_CODEC_ID_NELLYMOSER: case AV_CODEC_ID_MUSEPACK8: case AV_CODEC_ID_SPEEX: case AV_CODEC_ID_WMAVOICE: case AV_CODEC_ID_WMAPRO: case AV_CODEC_ID_ATRAC3P: case AV_CODEC_ID_EAC3: case AV_CODEC_ID_SIPR: case AV_CODEC_ID_MP1: case AV_CODEC_ID_TWINVQ: case AV_CODEC_ID_MP4ALS: case AV_CODEC_ID_ATRAC1: case AV_CODEC_ID_BINKAUDIO_RDFT: case AV_CODEC_ID_BINKAUDIO_DCT: case AV_CODEC_ID_AAC_LATM: case AV_CODEC_ID_QDMC: case AV_CODEC_ID_CELT: case AV_CODEC_ID_G723_1: case AV_CODEC_ID_G729: case AV_CODEC_ID_8SVX_EXP: case AV_CODEC_ID_8SVX_FIB: case AV_CODEC_ID_BMV_AUDIO: case AV_CODEC_ID_RALF: case AV_CODEC_ID_IAC: case AV_CODEC_ID_ILBC: case AV_CODEC_ID_OPUS: case AV_CODEC_ID_COMFORT_NOISE: case AV_CODEC_ID_METASOUND: case AV_CODEC_ID_PAF_AUDIO: case AV_CODEC_ID_ON2AVC: case AV_CODEC_ID_DSS_SP: case AV_CODEC_ID_CODEC2: case AV_CODEC_ID_FFWAVESYNTH: case AV_CODEC_ID_SONIC: case AV_CODEC_ID_SONIC_LS: case AV_CODEC_ID_EVRC: case AV_CODEC_ID_SMV: case AV_CODEC_ID_4GV: case AV_CODEC_ID_INTERPLAY_ACM: case AV_CODEC_ID_XMA1: case AV_CODEC_ID_XMA2: case AV_CODEC_ID_ATRAC3AL: case AV_CODEC_ID_ATRAC3PAL: case AV_CODEC_ID_DOLBY_E: case AV_CODEC_ID_APTX: case AV_CODEC_ID_SBC: case AV_CODEC_ID_ATRAC9: case AV_CODEC_ID_HCOM: case AV_CODEC_ID_ACELP_KELVIN: case AV_CODEC_ID_MPEGH_3D_AUDIO: case AV_CODEC_ID_SIREN: case AV_CODEC_ID_HCA: case AV_CODEC_ID_FASTAUDIO: lossy = YES; break; default: break; } } // totalFrames = codecCtx->sample_rate * ((float)formatCtx->duration/AV_TIME_BASE); AVRational tb = { .num = 1, .den = codecCtx->sample_rate }; totalFrames = isStream ? 0 : av_rescale_q(stream->duration, stream->time_base, tb); bitrate = (int)((codecCtx->bit_rate) / 1000); framesRead = 0; endOfStream = NO; endOfAudio = NO; if(!isStream) { if(stream->start_time && stream->start_time != AV_NOPTS_VALUE) skipSamples = av_rescale_q(stream->start_time, stream->time_base, tb); if(skipSamples < 0) skipSamples = 0; } else { skipSamples = 0; } seekFrame = skipSamples; // Skip preroll if necessary if(totalFrames < 0) totalFrames = 0; seekable = [s seekable]; return YES; } - (void)close { if(lastReadPacket) { av_packet_unref(lastReadPacket); free(lastReadPacket); lastReadPacket = NULL; } if(lastDecodedFrame) { av_free(lastDecodedFrame); lastDecodedFrame = NULL; } if(codecCtx) { avcodec_close(codecCtx); avcodec_free_context(&codecCtx); codecCtx = NULL; } if(formatCtx) { avformat_close_input(&(formatCtx)); formatCtx = NULL; } if(ioCtx) { buffer = ioCtx->buffer; av_free(ioCtx); ioCtx = NULL; } if(buffer) { av_free(buffer); buffer = NULL; } } - (void)dealloc { [self close]; } - (int)readAudio:(void *)buf frames:(UInt32)frames { if(totalFrames && framesRead >= totalFrames) return 0; int frameSize = channels * (bitsPerSample / 8); int dataSize = 0; int bytesToRead = frames * frameSize; int bytesRead = 0; int seekBytesSkip = 0; int errcode; int8_t *targetBuf = (int8_t *)buf; memset(buf, 0, bytesToRead); while(bytesRead < bytesToRead) { // buffer size needed to hold decoded samples, in bytes int planeSize; int planar = av_sample_fmt_is_planar(codecCtx->sample_fmt); dataSize = av_samples_get_buffer_size(&planeSize, codecCtx->channels, lastDecodedFrame->nb_samples, codecCtx->sample_fmt, 1); if(dataSize < 0) dataSize = 0; while(readNextPacket && !endOfAudio) { // consume next chunk of encoded data from input stream if(!endOfStream) { av_packet_unref(lastReadPacket); if((errcode = av_read_frame(formatCtx, lastReadPacket)) < 0) { if(errcode == AVERROR_EOF) { DLog(@"End of stream"); endOfStream = YES; } if(formatCtx->pb && formatCtx->pb->error) break; } if(lastReadPacket->stream_index != streamIndex) continue; } if((errcode = avcodec_send_packet(codecCtx, endOfStream ? NULL : lastReadPacket)) < 0) { if(errcode != AVERROR(EAGAIN)) { char errDescr[4096]; av_strerror(errcode, errDescr, 4096); ALog(@"Error sending packet to codec, errcode = %d, error = %s", errcode, errDescr); return 0; } } readNextPacket = NO; // we probably won't need to consume another chunk } if(dataSize <= bytesConsumedFromDecodedFrame) { if(endOfStream && endOfAudio) break; bytesConsumedFromDecodedFrame = 0; if((errcode = avcodec_receive_frame(codecCtx, lastDecodedFrame)) < 0) { if(errcode == AVERROR_EOF) { endOfAudio = YES; break; } else if(errcode == AVERROR(EAGAIN)) { // Read another packet readNextPacket = YES; continue; } else { char errDescr[4096]; av_strerror(errcode, errDescr, 4096); ALog(@"Error receiving frame, errcode = %d, error = %s", errcode, errDescr); return 0; } } // Something has been successfully decoded dataSize = av_samples_get_buffer_size(&planeSize, codecCtx->channels, lastDecodedFrame->nb_samples, codecCtx->sample_fmt, 1); if(dataSize < 0) dataSize = 0; // FFmpeg seeking by packet is usually inexact, so skip up to // target sample using packet timestamp // New: Moved here, because sometimes preroll packets also // trigger EAGAIN above, so ask for the next packet's timestamp // instead if(seekFrame >= 0 && errcode >= 0) { DLog(@"Seeking to frame %lld", seekFrame); AVRational tb = { .num = 1, .den = codecCtx->sample_rate }; int64_t packetBeginFrame = av_rescale_q( lastReadPacket->dts, formatCtx->streams[streamIndex]->time_base, tb); if(packetBeginFrame < seekFrame) { seekBytesSkip += (int)((seekFrame - packetBeginFrame) * frameSize); } seekFrame = -1; } int minSkipped = FFMIN(dataSize, seekBytesSkip); bytesConsumedFromDecodedFrame += minSkipped; seekBytesSkip -= minSkipped; } int _channels = codecCtx->channels; uint32_t _channelConfig = (uint32_t)codecCtx->channel_layout; float _frequency = codecCtx->sample_rate; if(_channels != channels || _channelConfig != channelConfig || _frequency != frequency) { if(bytesRead > 0) { break; } else { channels = _channels; channelConfig = _channelConfig; frequency = _frequency; [self willChangeValueForKey:@"properties"]; [self didChangeValueForKey:@"properties"]; } } int toConsume = FFMIN((dataSize - bytesConsumedFromDecodedFrame), (bytesToRead - bytesRead)); // copy decoded samples to Cog's buffer if(!planar || channels == 1) { memmove(targetBuf + bytesRead, (lastDecodedFrame->data[0] + bytesConsumedFromDecodedFrame), toConsume); } else { uint8_t *out = (uint8_t *)targetBuf + bytesRead; int bytesPerSample = bitsPerSample / 8; int bytesConsumedPerPlane = bytesConsumedFromDecodedFrame / channels; int toConsumePerPlane = toConsume / channels; for(int s = 0; s < toConsumePerPlane; s += bytesPerSample) { for(int ch = 0; ch < channels; ++ch) { memcpy(out, lastDecodedFrame->extended_data[ch] + bytesConsumedPerPlane + s, bytesPerSample); out += bytesPerSample; } } } bytesConsumedFromDecodedFrame += toConsume; bytesRead += toConsume; } int framesReadNow = bytesRead / frameSize; if(totalFrames && (framesRead + framesReadNow > totalFrames)) framesReadNow = (int)(totalFrames - framesRead); framesRead += framesReadNow; return framesReadNow; } - (long)seek:(long)frame { if(!totalFrames) return -1; if(frame >= totalFrames) { framesRead = totalFrames; endOfStream = YES; endOfAudio = YES; return -1; } AVRational tb = { .num = 1, .den = codecCtx->sample_rate }; int64_t ts = av_rescale_q(frame, tb, formatCtx->streams[streamIndex]->time_base); int ret = avformat_seek_file(formatCtx, streamIndex, ts - 1000, ts, ts, 0); avcodec_flush_buffers(codecCtx); if(ret < 0) { framesRead = totalFrames; endOfStream = YES; endOfAudio = YES; return -1; } readNextPacket = YES; // so we immediately read next packet bytesConsumedFromDecodedFrame = INT_MAX; // so we immediately begin decoding next frame framesRead = frame; seekFrame = frame + skipSamples; endOfStream = NO; endOfAudio = NO; return frame; } - (NSDictionary *)properties { return @{@"channels": [NSNumber numberWithInt:channels], @"channelConfig": [NSNumber numberWithUnsignedInt:channelConfig], @"bitsPerSample": [NSNumber numberWithInt:bitsPerSample], @"Unsigned": [NSNumber numberWithBool:(bitsPerSample == 8)], @"sampleRate": [NSNumber numberWithFloat:frequency], @"floatingPoint": [NSNumber numberWithBool:floatingPoint], @"totalFrames": [NSNumber numberWithDouble:totalFrames], @"bitrate": [NSNumber numberWithInt:bitrate], @"seekable": [NSNumber numberWithBool:seekable], @"codec": [NSString stringWithUTF8String:avcodec_get_name(codecCtx->codec_id)], @"endian": @"host", @"encoding": lossy ? @"lossy" : @"lossless"}; } + (NSArray *)fileTypes { return @[@"wma", @"asf", @"tak", @"mp4", @"m4a", @"aac", @"mp3", @"mp2", @"m2a", @"mpa", @"ape", @"ac3", @"dts", @"dtshd", @"wav", @"tta", @"vqf", @"vqe", @"vql", @"ra", @"rm", @"rmj", @"mka", @"weba"]; } + (NSArray *)mimeTypes { return @[@"application/wma", @"application/x-wma", @"audio/x-wma", @"audio/x-ms-wma", @"audio/x-tak", @"application/ogg", @"audio/aacp", @"audio/mpeg", @"audio/mp4", @"audio/x-mp3", @"audio/x-mp2", @"audio/x-matroska", @"audio/x-ape", @"audio/x-ac3", @"audio/x-dts", @"audio/x-dtshd", @"audio/x-at3", @"audio/wav", @"audio/tta", @"audio/x-tta", @"audio/x-twinvq", @"application/vnd.apple.mpegurl"]; } + (NSArray *)fileTypeAssociations { return @[ @[@"Windows Media Audio File", @"song.icns", @"wma", @"asf"], @[@"TAK Audio File", @"song.icns", @"tak"], @[@"MPEG-4 Audio File", @"m4a.icns", @"mp4", @"m4a"], @[@"MPEG-4 AAC Audio File", @"song.icns", @"aac"], @[@"MPEG Audio File", @"mp3.icns", @"mp3", @"m2a", @"mpa"], @[@"Monkey's Audio File", @"ape.icns", @"ape"], @[@"AC-3 Audio File", @"song.icns", @"ac3"], @[@"DTS Audio File", @"song.icns", @"dts"], @[@"DTS-HD MA Audio File", @"song.icns", @"dtshd"], @[@"True Audio File", @"song.icns", @"tta"], @[@"TrueVQ Audio File", @"song.icns", @"vqf", @"vqe", @"vql"], @[@"Real Audio File", @"song.icns", @"ra", @"rm", @"rmj"], @[@"Matroska Audio File", @"song.icns", @"mka"], @[@"WebM Audio File", @"song.icns", @"weba"] ]; } + (float)priority { return 1.5; } @end