TEST(API, Test2SilenceRawFp)
{
short zeroes[1024];
std::fill(zeroes, zeroes + 1024, 0);
ChromaprintContext *ctx = chromaprint_new(CHROMAPRINT_ALGORITHM_TEST2);
ASSERT_NE(nullptr, ctx);
SCOPE_EXIT(chromaprint_free(ctx));
ASSERT_EQ(1, chromaprint_start(ctx, 44100, 1));
for (int i = 0; i < 130; i++) {
ASSERT_EQ(1, chromaprint_feed(ctx, zeroes, 1024));
}
uint32_t *fp;
int length;
ASSERT_EQ(1, chromaprint_finish(ctx));
ASSERT_EQ(1, chromaprint_get_raw_fingerprint(ctx, &fp, &length));
SCOPE_EXIT(chromaprint_dealloc(fp));
ASSERT_EQ(3, length);
EXPECT_EQ(627964279, fp[0]);
EXPECT_EQ(627964279, fp[1]);
EXPECT_EQ(627964279, fp[2]);
}
TEST(API, Test2SilenceFp)
{
short zeroes[1024];
std::fill(zeroes, zeroes + 1024, 0);
ChromaprintContext *ctx = chromaprint_new(CHROMAPRINT_ALGORITHM_TEST2);
ASSERT_NE(nullptr, ctx);
SCOPE_EXIT(chromaprint_free(ctx));
ASSERT_EQ(1, chromaprint_start(ctx, 44100, 1));
for (int i = 0; i < 130; i++) {
ASSERT_EQ(1, chromaprint_feed(ctx, zeroes, 1024));
}
char *fp;
uint32_t fp_hash;
ASSERT_EQ(1, chromaprint_finish(ctx));
ASSERT_EQ(1, chromaprint_get_fingerprint(ctx, &fp));
SCOPE_EXIT(chromaprint_dealloc(fp));
ASSERT_EQ(1, chromaprint_get_fingerprint_hash(ctx, &fp_hash));
ASSERT_EQ(18, strlen(fp));
EXPECT_EQ(std::string("AQAAA0mUaEkSRZEGAA"), std::string(fp));
ASSERT_EQ(627964279, fp_hash);
}
static int write_header(AVFormatContext *s)
{
ChromaprintMuxContext *cpr = s->priv_data;
AVStream *st;
avpriv_lock_avformat();
cpr->ctx = chromaprint_new(cpr->algorithm);
avpriv_unlock_avformat();
if (!cpr->ctx) {
av_log(s, AV_LOG_ERROR, "Failed to create chromaprint context.\n");
return AVERROR(ENOMEM);
}
if (cpr->silence_threshold != -1) {
#if CPR_VERSION_INT >= AV_VERSION_INT(0, 7, 0)
if (!chromaprint_set_option(cpr->ctx, "silence_threshold", cpr->silence_threshold)) {
av_log(s, AV_LOG_ERROR, "Failed to set silence threshold.\n");
goto fail;
}
#else
av_log(s, AV_LOG_ERROR, "Setting the silence threshold requires Chromaprint "
"version 0.7.0 or later.\n");
goto fail;
#endif
}
if (s->nb_streams != 1) {
av_log(s, AV_LOG_ERROR, "Only one stream is supported\n");
goto fail;
}
st = s->streams[0];
if (st->codecpar->channels > 2) {
av_log(s, AV_LOG_ERROR, "Only up to 2 channels are supported\n");
goto fail;
}
if (st->codecpar->sample_rate < 1000) {
av_log(s, AV_LOG_ERROR, "Sampling rate must be at least 1000\n");
goto fail;
}
if (!chromaprint_start(cpr->ctx, st->codecpar->sample_rate, st->codecpar->channels)) {
av_log(s, AV_LOG_ERROR, "Failed to start chromaprint\n");
goto fail;
}
return 0;
fail:
cleanup(cpr);
return AVERROR(EINVAL);
}
static GstFlowReturn
gst_chromaprint_transform_ip (GstBaseTransform * trans, GstBuffer * buf)
{
GstChromaprint *chromaprint = GST_CHROMAPRINT (trans);
GstAudioFilter *filter = GST_AUDIO_FILTER (trans);
GstMapInfo map_info;
guint nsamples;
gint rate, channels;
rate = GST_AUDIO_INFO_RATE (&filter->info);
channels = GST_AUDIO_INFO_CHANNELS (&filter->info);
if (G_UNLIKELY (rate <= 0 || channels <= 0))
return GST_FLOW_NOT_NEGOTIATED;
if (!chromaprint->record)
return GST_FLOW_OK;
if (!gst_buffer_map (buf, &map_info, GST_MAP_READ))
return GST_FLOW_ERROR;
nsamples = map_info.size / (channels * 2);
if (nsamples == 0)
goto end;
if (chromaprint->nsamples == 0) {
chromaprint_start (chromaprint->context, rate, channels);
}
chromaprint->nsamples += nsamples;
chromaprint->duration = chromaprint->nsamples / rate;
chromaprint_feed (chromaprint->context, map_info.data,
map_info.size / sizeof (guint16));
if (chromaprint->duration >= chromaprint->max_duration
&& !chromaprint->fingerprint) {
gst_chromaprint_create_fingerprint (chromaprint);
}
end:
gst_buffer_unmap (buf, &map_info);
return GST_FLOW_OK;
}
static sout_stream_id_t *Add( sout_stream_t *p_stream, es_format_t *p_fmt )
{
sout_stream_sys_t *p_sys = p_stream->p_sys;
sout_stream_id_t *id = NULL;
if ( p_fmt->i_cat == AUDIO_ES && !p_sys->id )
{
if( p_fmt->i_codec != VLC_CODEC_S16N || p_fmt->audio.i_channels > 2 )
{
msg_Warn( p_stream, "bad input format: need s16l, 1 or 2 channels" );
goto error;
}
id = malloc( sizeof( sout_stream_id_t ) );
if ( !id ) goto error;
id->i_channels = p_fmt->audio.i_channels;
id->i_samplerate = p_fmt->audio.i_rate;
id->i_samples = p_sys->i_duration * id->i_samplerate;
if ( !chromaprint_start( p_sys->p_chromaprint_ctx, p_fmt->audio.i_rate, id->i_channels ) )
{
msg_Err( p_stream, "Failed starting chromaprint on %uHz %uch samples",
p_fmt->audio.i_rate, id->i_channels );
goto error;
}
else
{
p_sys->id = id;
msg_Dbg( p_stream, "Starting chromaprint on %uHz %uch samples",
p_fmt->audio.i_rate, id->i_channels );
}
return id;
}
error:
free( id );
return NULL;
}
TEST(API, TestFp) {
std::vector<short> data = LoadAudioFile("data/test_stereo_44100.raw");
ChromaprintContext *ctx = chromaprint_new(CHROMAPRINT_ALGORITHM_TEST2);
ASSERT_NE(nullptr, ctx);
SCOPE_EXIT(chromaprint_free(ctx));
ASSERT_EQ(1, chromaprint_get_num_channels(ctx));
ASSERT_EQ(11025, chromaprint_get_sample_rate(ctx));
ASSERT_EQ(1, chromaprint_start(ctx, 44100, 1));
ASSERT_EQ(1, chromaprint_feed(ctx, data.data(), data.size()));
char *fp;
uint32_t fp_hash;
ASSERT_EQ(1, chromaprint_finish(ctx));
ASSERT_EQ(1, chromaprint_get_fingerprint(ctx, &fp));
SCOPE_EXIT(chromaprint_dealloc(fp));
ASSERT_EQ(1, chromaprint_get_fingerprint_hash(ctx, &fp_hash));
EXPECT_EQ(std::string("AQAAC0kkZUqYREkUnFAXHk8uuMZl6EfO4zu-4ABKFGESWIIMEQE"), std::string(fp));
ASSERT_EQ(3732003127, fp_hash);
}
QString Chromaprinter::CreateFingerprint() {
Q_ASSERT(QThread::currentThread() != qApp->thread());
buffer_.open(QIODevice::WriteOnly);
GMainContext* context = g_main_context_new();
g_main_context_push_thread_default(context);
event_loop_ = g_main_loop_new(context, FALSE);
pipeline_ = gst_pipeline_new("pipeline");
GstElement* src = CreateElement("filesrc", pipeline_);
GstElement* decode = CreateElement("decodebin2", pipeline_);
GstElement* convert = CreateElement("audioconvert", pipeline_);
GstElement* resample = CreateElement("audioresample", pipeline_);
GstElement* sink = CreateElement("appsink", pipeline_);
if (!src || !decode || !convert || !resample || !sink) {
return QString();
}
convert_element_ = convert;
// Connect the elements
gst_element_link_many(src, decode, nullptr);
gst_element_link_many(convert, resample, nullptr);
// Chromaprint expects mono floats at a sample rate of 11025Hz.
GstCaps* caps = gst_caps_new_simple(
"audio/x-raw-int", "width", G_TYPE_INT, 16, "channels", G_TYPE_INT,
kDecodeChannels, "rate", G_TYPE_INT, kDecodeRate, nullptr);
gst_element_link_filtered(resample, sink, caps);
gst_caps_unref(caps);
GstAppSinkCallbacks callbacks;
memset(&callbacks, 0, sizeof(callbacks));
callbacks.new_buffer = NewBufferCallback;
gst_app_sink_set_callbacks(reinterpret_cast<GstAppSink*>(sink), &callbacks,
this, nullptr);
g_object_set(G_OBJECT(sink), "sync", FALSE, nullptr);
g_object_set(G_OBJECT(sink), "emit-signals", TRUE, nullptr);
// Set the filename
g_object_set(src, "location", filename_.toUtf8().constData(), nullptr);
// Connect signals
CHECKED_GCONNECT(decode, "new-decoded-pad", &NewPadCallback, this);
gst_bus_set_sync_handler(gst_pipeline_get_bus(GST_PIPELINE(pipeline_)),
BusCallbackSync, this);
guint bus_callback_id = gst_bus_add_watch(
gst_pipeline_get_bus(GST_PIPELINE(pipeline_)), BusCallback, this);
QTime time;
time.start();
// Start playing
gst_element_set_state(pipeline_, GST_STATE_PLAYING);
g_main_loop_run(event_loop_);
g_main_loop_unref(event_loop_);
g_main_context_unref(context);
int decode_time = time.restart();
buffer_.close();
QByteArray data = buffer_.data();
ChromaprintContext* chromaprint =
chromaprint_new(CHROMAPRINT_ALGORITHM_DEFAULT);
chromaprint_start(chromaprint, kDecodeRate, kDecodeChannels);
chromaprint_feed(chromaprint, reinterpret_cast<void*>(data.data()),
data.size() / 2);
chromaprint_finish(chromaprint);
void* fprint = nullptr;
int size = 0;
int ret = chromaprint_get_raw_fingerprint(chromaprint, &fprint, &size);
QByteArray fingerprint;
if (ret == 1) {
void* encoded = nullptr;
int encoded_size = 0;
chromaprint_encode_fingerprint(fprint, size, CHROMAPRINT_ALGORITHM_DEFAULT,
&encoded, &encoded_size, 1);
fingerprint.append(reinterpret_cast<char*>(encoded), encoded_size);
chromaprint_dealloc(fprint);
chromaprint_dealloc(encoded);
}
chromaprint_free(chromaprint);
int codegen_time = time.elapsed();
qLog(Debug) << "Decode time:" << decode_time
<< "Codegen time:" << codegen_time;
// Cleanup
callbacks.new_buffer = nullptr;
gst_app_sink_set_callbacks(reinterpret_cast<GstAppSink*>(sink), &callbacks,
this, nullptr);
gst_bus_set_sync_handler(gst_pipeline_get_bus(GST_PIPELINE(pipeline_)),
nullptr, nullptr);
g_source_remove(bus_callback_id);
gst_element_set_state(pipeline_, GST_STATE_NULL);
gst_object_unref(pipeline_);
return fingerprint;
}
int decode_audio_file(ChromaprintContext *chromaprint_ctx, int16_t *buffer1, int16_t *buffer2, const char *file_name, int max_length, int *duration)
{
int i, ok = 0, remaining, length, consumed, buffer_size, codec_ctx_opened = 0;
AVFormatContext *format_ctx = NULL;
AVCodecContext *codec_ctx = NULL;
AVCodec *codec = NULL;
AVStream *stream = NULL;
AVPacket packet, packet_temp;
#ifdef HAVE_AV_AUDIO_CONVERT
AVAudioConvert *convert_ctx = NULL;
#endif
int16_t *buffer;
if (!strcmp(file_name, "-")) {
file_name = "pipe:0";
}
#if LIBAVFORMAT_VERSION_INT < AV_VERSION_INT(53, 2, 0)
if (av_open_input_file(&format_ctx, file_name, NULL, 0, NULL) != 0) {
#else
if (avformat_open_input(&format_ctx, file_name, NULL, NULL) != 0) {
#endif
fprintf(stderr, "ERROR: couldn't open the file\n");
goto done;
}
if (av_find_stream_info(format_ctx) < 0) {
fprintf(stderr, "ERROR: couldn't find stream information in the file\n");
goto done;
}
for (i = 0; i < format_ctx->nb_streams; i++) {
codec_ctx = format_ctx->streams[i]->codec;
if (codec_ctx && codec_ctx->codec_type == AVMEDIA_TYPE_AUDIO) {
stream = format_ctx->streams[i];
break;
}
}
if (!stream) {
fprintf(stderr, "ERROR: couldn't find any audio stream in the file\n");
goto done;
}
codec = avcodec_find_decoder(codec_ctx->codec_id);
if (!codec) {
fprintf(stderr, "ERROR: unknown codec\n");
goto done;
}
if (avcodec_open(codec_ctx, codec) < 0) {
fprintf(stderr, "ERROR: couldn't open the codec\n");
goto done;
}
codec_ctx_opened = 1;
if (codec_ctx->channels <= 0) {
fprintf(stderr, "ERROR: no channels found in the audio stream\n");
goto done;
}
if (codec_ctx->sample_fmt != AV_SAMPLE_FMT_S16) {
#ifdef HAVE_AV_AUDIO_CONVERT
convert_ctx = av_audio_convert_alloc(AV_SAMPLE_FMT_S16, codec_ctx->channels,
codec_ctx->sample_fmt, codec_ctx->channels, NULL, 0);
if (!convert_ctx) {
fprintf(stderr, "ERROR: couldn't create sample format converter\n");
goto done;
}
#else
fprintf(stderr, "ERROR: unsupported sample format\n");
goto done;
#endif
}
*duration = stream->time_base.num * stream->duration / stream->time_base.den;
av_init_packet(&packet);
av_init_packet(&packet_temp);
remaining = max_length * codec_ctx->channels * codec_ctx->sample_rate;
chromaprint_start(chromaprint_ctx, codec_ctx->sample_rate, codec_ctx->channels);
while (1) {
if (av_read_frame(format_ctx, &packet) < 0) {
break;
}
packet_temp.data = packet.data;
packet_temp.size = packet.size;
while (packet_temp.size > 0) {
buffer_size = BUFFER_SIZE;
#if LIBAVCODEC_VERSION_INT < AV_VERSION_INT(52, 23, 0)
consumed = avcodec_decode_audio2(codec_ctx,
buffer1, &buffer_size, packet_temp.data, packet_temp.size);
#else
consumed = avcodec_decode_audio3(codec_ctx,
buffer1, &buffer_size, &packet_temp);
#endif
if (consumed < 0) {
break;
}
packet_temp.data += consumed;
packet_temp.size -= consumed;
if (buffer_size <= 0) {
if (buffer_size < 0) {
fprintf(stderr, "WARNING: size returned from avcodec_decode_audioX is too small\n");
}
continue;
}
if (buffer_size > BUFFER_SIZE) {
fprintf(stderr, "WARNING: size returned from avcodec_decode_audioX is too large\n");
continue;
}
#ifdef HAVE_AV_AUDIO_CONVERT
if (convert_ctx) {
const void *ibuf[6] = { buffer1 };
void *obuf[6] = { buffer2 };
#if LIBAVUTIL_VERSION_INT < AV_VERSION_INT(51, 8, 0)
int istride[6] = { av_get_bits_per_sample_format(codec_ctx->sample_fmt) / 8 };
#else
int istride[6] = { av_get_bytes_per_sample(codec_ctx->sample_fmt) };
#endif
int ostride[6] = { 2 };
int len = buffer_size / istride[0];
if (av_audio_convert(convert_ctx, obuf, ostride, ibuf, istride, len) < 0) {
break;
}
buffer = buffer2;
buffer_size = len * ostride[0];
}
else {
buffer = buffer1;
}
#else
buffer = buffer1;
#endif
length = MIN(remaining, buffer_size / 2);
if (!chromaprint_feed(chromaprint_ctx, buffer, length)) {
fprintf(stderr, "ERROR: fingerprint calculation failed\n");
goto done;
}
if (max_length) {
remaining -= length;
if (remaining <= 0) {
goto finish;
}
}
}
if (packet.data) {
av_free_packet(&packet);
}
}
finish:
if (!chromaprint_finish(chromaprint_ctx)) {
fprintf(stderr, "ERROR: fingerprint calculation failed\n");
goto done;
}
ok = 1;
done:
if (codec_ctx_opened) {
avcodec_close(codec_ctx);
}
if (format_ctx) {
av_close_input_file(format_ctx);
}
#ifdef HAVE_AV_AUDIO_CONVERT
if (convert_ctx) {
av_audio_convert_free(convert_ctx);
}
#endif
return ok;
}
int fpcalc_main(int argc, char **argv)
{
int i, j, max_length = 120, num_file_names = 0, raw = 0, raw_fingerprint_size, duration;
int16_t *buffer1, *buffer2;
int32_t *raw_fingerprint;
char *file_name, *fingerprint, **file_names;
ChromaprintContext *chromaprint_ctx;
int algo = CHROMAPRINT_ALGORITHM_DEFAULT;
file_names = malloc(argc * sizeof(char *));
for (i = 1; i < argc; i++) {
char *arg = argv[i];
if (!strcmp(arg, "-length") && i + 1 < argc) {
max_length = atoi(argv[++i]);
}
else if (!strcmp(arg, "-version") || !strcmp(arg, "-v")) {
printf("fpcalc version %s\n", chromaprint_get_version());
return 0;
}
else if (!strcmp(arg, "-raw")) {
raw = 1;
}
else if (!strcmp(arg, "-algo") && i + 1 < argc) {
const char *v = argv[++i];
if (!strcmp(v, "test1")) { algo = CHROMAPRINT_ALGORITHM_TEST1; }
else if (!strcmp(v, "test2")) { algo = CHROMAPRINT_ALGORITHM_TEST2; }
else if (!strcmp(v, "test3")) { algo = CHROMAPRINT_ALGORITHM_TEST3; }
else if (!strcmp(v, "test4")) { algo = CHROMAPRINT_ALGORITHM_TEST4; }
else {
fprintf(stderr, "WARNING: unknown algorithm, using the default\n");
}
}
else if (!strcmp(arg, "-set") && i + 1 < argc) {
i += 1;
}
else {
file_names[num_file_names++] = argv[i];
}
}
if (!num_file_names) {
printf("usage: %s [OPTIONS] FILE...\n\n", argv[0]);
printf("Options:\n");
printf(" -version print version information\n");
printf(" -length SECS length of the audio data used for fingerprint calculation (default 120)\n");
printf(" -raw output the raw uncompressed fingerprint\n");
printf(" -algo NAME version of the fingerprint algorithm\n");
return 2;
}
av_register_all();
av_log_set_level(AV_LOG_ERROR);
buffer1 = av_malloc(BUFFER_SIZE + 16);
buffer2 = av_malloc(BUFFER_SIZE + 16);
chromaprint_ctx = chromaprint_new(algo);
for (i = 1; i < argc; i++) {
char *arg = argv[i];
if (!strcmp(arg, "-set") && i + 1 < argc) {
char *name = argv[++i];
char *value = strchr(name, '=');
if (value) {
*value++ = '\0';
chromaprint_set_option(chromaprint_ctx, name, atoi(value));
}
}
}
for (i = 0; i < num_file_names; i++) {
file_name = file_names[i];
if (!decode_audio_file(chromaprint_ctx, buffer1, buffer2, file_name, max_length, &duration)) {
fprintf(stderr, "ERROR: unable to calculate fingerprint for file %s, skipping\n", file_name);
continue;
}
if (i > 0) {
printf("\n");
}
printf("FILE=%s\n", file_name);
printf("DURATION=%d\n", duration);
if (raw) {
if (!chromaprint_get_raw_fingerprint(chromaprint_ctx, (void **)&raw_fingerprint, &raw_fingerprint_size)) {
fprintf(stderr, "ERROR: unable to calculate fingerprint for file %s, skipping\n", file_name);
continue;
}
printf("FINGERPRINT=");
for (j = 0; j < raw_fingerprint_size; j++) {
printf("%d%s", raw_fingerprint[j], j + 1 < raw_fingerprint_size ? "," : "");
}
printf("\n");
chromaprint_dealloc(raw_fingerprint);
}
else {
if (!chromaprint_get_fingerprint(chromaprint_ctx, &fingerprint)) {
fprintf(stderr, "ERROR: unable to calculate fingerprint for file %s, skipping\n", file_name);
continue;
}
printf("FINGERPRINT=%s\n", fingerprint);
chromaprint_dealloc(fingerprint);
}
}
chromaprint_free(chromaprint_ctx);
av_free(buffer1);
av_free(buffer2);
free(file_names);
return 0;
}
Chroma::Result Chroma::operator() (const QString& filename)
{
std::shared_ptr<AVFormatContext> formatCtx;
{
AVFormatContext *formatCtxRaw = nullptr;
if (avformat_open_input (&formatCtxRaw, filename.toLatin1 ().constData (), nullptr, nullptr))
throw std::runtime_error ("error opening file");
formatCtx.reset (formatCtxRaw,
[] (AVFormatContext *ctx) {
avformat_close_input (&ctx);
});
}
{
QMutexLocker locker (&CodecMutex_);
if (avformat_find_stream_info (formatCtx.get (), nullptr) < 0)
throw std::runtime_error ("could not find stream");
}
AVCodec *codec = nullptr;
const auto streamIndex = av_find_best_stream (formatCtx.get (), AVMEDIA_TYPE_AUDIO, -1, -1, &codec, 0);
if (streamIndex < 0)
throw std::runtime_error ("could not find audio stream");
auto stream = formatCtx->streams [streamIndex];
bool codecOpened = false;
std::shared_ptr<AVCodecContext> codecCtx (stream->codec,
[&codecOpened] (AVCodecContext *ctx) {
if (codecOpened) avcodec_close (ctx);
});
{
QMutexLocker locker (&CodecMutex_);
if (avcodec_open2 (codecCtx.get (), codec, nullptr) < 0)
throw std::runtime_error ("couldn't open the codec");
}
codecOpened = true;
if (codecCtx->channels <= 0)
throw std::runtime_error ("no channels found");
std::shared_ptr<SwrContext> swr;
if (codecCtx->sample_fmt != AV_SAMPLE_FMT_S16)
{
swr.reset (swr_alloc (), [] (SwrContext *ctx) {
if (ctx) swr_free (&ctx);
});
av_opt_set_int (swr.get (), "in_channel_layout", codecCtx->channel_layout, 0);
av_opt_set_int (swr.get (), "out_channel_layout", codecCtx->channel_layout, 0);
av_opt_set_int (swr.get (), "in_sample_rate", codecCtx->sample_rate, 0);
av_opt_set_int (swr.get (), "out_sample_rate", codecCtx->sample_rate, 0);
av_opt_set_sample_fmt (swr.get (), "in_sample_fmt", codecCtx->sample_fmt, 0);
av_opt_set_sample_fmt (swr.get (), "out_sample_fmt", AV_SAMPLE_FMT_S16, 0);
swr_init (swr.get ());
}
AVPacket packet;
av_init_packet (&packet);
const int maxLength = 120;
auto remaining = maxLength * codecCtx->channels * codecCtx->sample_rate;
chromaprint_start (Ctx_, codecCtx->sample_rate, codecCtx->channels);
std::shared_ptr<AVFrame> frame (av_frame_alloc (),
[] (AVFrame *frame) {
av_frame_free (&frame);
});
auto maxDstNbSamples = 0;
uint8_t *dstData [1] = { nullptr };
std::shared_ptr<void> dstDataGuard (nullptr,
[&dstData] (void*) {
if (dstData [0]) av_freep (&dstData [0]);
});
while (true)
{
if (av_read_frame (formatCtx.get (), &packet) < 0)
break;
std::shared_ptr<void> guard (nullptr,
[&packet] (void*) {
if (packet.data) av_free_packet (&packet);
});
if (packet.stream_index != streamIndex)
continue;
av_frame_unref (frame.get ());
int gotFrame = false;
auto consumed = avcodec_decode_audio4 (codecCtx.get (), frame.get (), &gotFrame, &packet);
if (consumed < 0 || !gotFrame)
continue;
uint8_t **data = nullptr;
if (swr)
{
if (frame->nb_samples > maxDstNbSamples)
{
if (dstData [0])
av_freep (&dstData [0]);
int linesize = 0;
if (av_samples_alloc (dstData, &linesize, codecCtx->channels, frame->nb_samples, AV_SAMPLE_FMT_S16, 1) < 0)
throw std::runtime_error ("cannot allocate memory for resampling");
}
if (swr_convert (swr.get (), dstData, frame->nb_samples, const_cast<const uint8_t**> (frame->data), frame->nb_samples) < 0)
throw std::runtime_error ("cannot resample audio");
data = dstData;
}
else
data = frame->data;
auto length = std::min (remaining, frame->nb_samples * codecCtx->channels);
if (!chromaprint_feed (Ctx_, data [0], length))
throw std::runtime_error ("cannot feed data");
bool finished = false;
if (maxLength)
{
remaining -= length;
if (remaining <= 0)
finished = true;
}
if (finished)
break;
}
if (!chromaprint_finish (Ctx_))
throw std::runtime_error ("fingerprint calculation failed");
char *fingerprint = 0;
if (!chromaprint_get_fingerprint (Ctx_, &fingerprint))
throw std::runtime_error ("unable to get fingerprint");
QByteArray result (fingerprint);
chromaprint_dealloc (fingerprint);
const double divideFactor = 1. / av_q2d (stream->time_base);
const double duration = stream->duration / divideFactor;
return { result, static_cast<int> (duration) };
}
int decode_audio_file(ChromaprintContext *chromaprint_ctx, const char *file_name, int max_length, int *duration)
{
int ok = 0, remaining, length, consumed, codec_ctx_opened = 0, got_frame, stream_index;
AVFormatContext *format_ctx = NULL;
AVCodecContext *codec_ctx = NULL;
AVCodec *codec = NULL;
AVStream *stream = NULL;
AVFrame *frame = NULL;
#if defined(HAVE_SWRESAMPLE)
SwrContext *convert_ctx = NULL;
#elif defined(HAVE_AVRESAMPLE)
AVAudioResampleContext *convert_ctx = NULL;
#else
void *convert_ctx = NULL;
#endif
int max_dst_nb_samples = 0, dst_linsize = 0;
uint8_t *dst_data[1] = { NULL };
uint8_t **data;
AVPacket packet;
if (!strcmp(file_name, "-")) {
file_name = "pipe:0";
}
if (avformat_open_input(&format_ctx, file_name, NULL, NULL) != 0) {
fprintf(stderr, "ERROR: couldn't open the file\n");
goto done;
}
if (avformat_find_stream_info(format_ctx, NULL) < 0) {
fprintf(stderr, "ERROR: couldn't find stream information in the file\n");
goto done;
}
stream_index = av_find_best_stream(format_ctx, AVMEDIA_TYPE_AUDIO, -1, -1, &codec, 0);
if (stream_index < 0) {
fprintf(stderr, "ERROR: couldn't find any audio stream in the file\n");
goto done;
}
stream = format_ctx->streams[stream_index];
codec_ctx = stream->codec;
codec_ctx->request_sample_fmt = AV_SAMPLE_FMT_S16;
if (avcodec_open2(codec_ctx, codec, NULL) < 0) {
fprintf(stderr, "ERROR: couldn't open the codec\n");
goto done;
}
codec_ctx_opened = 1;
if (codec_ctx->channels <= 0) {
fprintf(stderr, "ERROR: no channels found in the audio stream\n");
goto done;
}
if (codec_ctx->sample_fmt != AV_SAMPLE_FMT_S16) {
int64_t channel_layout = codec_ctx->channel_layout;
if (!channel_layout) {
channel_layout = get_default_channel_layout(codec_ctx->channels);
}
#if defined(HAVE_SWRESAMPLE)
convert_ctx = swr_alloc_set_opts(NULL,
channel_layout, AV_SAMPLE_FMT_S16, codec_ctx->sample_rate,
channel_layout, codec_ctx->sample_fmt, codec_ctx->sample_rate,
0, NULL);
if (!convert_ctx) {
fprintf(stderr, "ERROR: couldn't allocate audio converter\n");
goto done;
}
if (swr_init(convert_ctx) < 0) {
fprintf(stderr, "ERROR: couldn't initialize the audio converter\n");
goto done;
}
#elif defined(HAVE_AVRESAMPLE)
convert_ctx = avresample_alloc_context();
av_opt_set_int(convert_ctx, "out_channel_layout", channel_layout, 0);
av_opt_set_int(convert_ctx, "out_sample_fmt", AV_SAMPLE_FMT_S16, 0);
av_opt_set_int(convert_ctx, "out_sample_rate", codec_ctx->sample_rate, 0);
av_opt_set_int(convert_ctx, "in_channel_layout", channel_layout, 0);
av_opt_set_int(convert_ctx, "in_sample_fmt", codec_ctx->sample_fmt, 0);
av_opt_set_int(convert_ctx, "in_sample_rate", codec_ctx->sample_rate, 0);
if (!convert_ctx) {
fprintf(stderr, "ERROR: couldn't allocate audio converter\n");
goto done;
}
if (avresample_open(convert_ctx) < 0) {
fprintf(stderr, "ERROR: couldn't initialize the audio converter\n");
goto done;
}
#else
fprintf(stderr, "ERROR: unsupported audio format (please build fpcalc with libswresample)\n");
goto done;
#endif
}
if (stream->duration != AV_NOPTS_VALUE) {
*duration = stream->time_base.num * stream->duration / stream->time_base.den;
}
else if (format_ctx->duration != AV_NOPTS_VALUE) {
*duration = format_ctx->duration / AV_TIME_BASE;
}
else {
fprintf(stderr, "ERROR: couldn't detect the audio duration\n");
goto done;
}
remaining = max_length * codec_ctx->channels * codec_ctx->sample_rate;
chromaprint_start(chromaprint_ctx, codec_ctx->sample_rate, codec_ctx->channels);
frame = av_frame_alloc();
while (1) {
if (av_read_frame(format_ctx, &packet) < 0) {
break;
}
if (packet.stream_index == stream_index) {
av_frame_unref(frame);
got_frame = 0;
consumed = avcodec_decode_audio4(codec_ctx, frame, &got_frame, &packet);
if (consumed < 0) {
fprintf(stderr, "WARNING: error decoding audio\n");
continue;
}
if (got_frame) {
data = frame->data;
if (convert_ctx) {
if (frame->nb_samples > max_dst_nb_samples) {
av_freep(&dst_data[0]);
if (av_samples_alloc(dst_data, &dst_linsize, codec_ctx->channels, frame->nb_samples, AV_SAMPLE_FMT_S16, 1) < 0) {
fprintf(stderr, "ERROR: couldn't allocate audio converter buffer\n");
goto done;
}
max_dst_nb_samples = frame->nb_samples;
}
#if defined(HAVE_SWRESAMPLE)
if (swr_convert(convert_ctx, dst_data, frame->nb_samples, (const uint8_t **)frame->data, frame->nb_samples) < 0)
#elif defined(HAVE_AVRESAMPLE)
if (avresample_convert(convert_ctx, dst_data, 0, frame->nb_samples, (uint8_t **)frame->data, 0, frame->nb_samples) < 0)
#endif
{
fprintf(stderr, "ERROR: couldn't convert the audio\n");
goto done;
}
data = dst_data;
}
length = MIN(remaining, frame->nb_samples * codec_ctx->channels);
if (!chromaprint_feed(chromaprint_ctx, data[0], length)) {
goto done;
}
if (max_length) {
remaining -= length;
if (remaining <= 0) {
goto finish;
}
}
}
}
av_free_packet(&packet);
}
finish:
if (!chromaprint_finish(chromaprint_ctx)) {
fprintf(stderr, "ERROR: fingerprint calculation failed\n");
goto done;
}
ok = 1;
done:
if (frame) {
av_frame_free(&frame);
}
if (dst_data[0]) {
av_freep(&dst_data[0]);
}
if (convert_ctx) {
#if defined(HAVE_SWRESAMPLE)
swr_free(&convert_ctx);
#elif defined(HAVE_AVRESAMPLE)
avresample_free(&convert_ctx);
#endif
}
if (codec_ctx_opened) {
avcodec_close(codec_ctx);
}
if (format_ctx) {
avformat_close_input(&format_ctx);
}
return ok;
}
QString chromaprinter::calcFingerPrint(SoundSourceProxy& soundSource){
soundSource.open();
m_SampleRate = soundSource.getSampleRate();
unsigned int length = soundSource.length();
if (m_SampleRate == 0 ){
qDebug() << "Skipping invalid file:" << soundSource.getFilename();
return QString();
}
// this is worth 2min of audio, multiply by 2 because we have 2 channels
// AcoustID only stores a fingerprint for the first two minutes of a song
// on their server so we need only a fingerprint of the first two minutes
// --kain88 July 2012
m_NumSamples = 120*2*m_SampleRate;
// check that the song is actually longer then the amount of audio we use
if (m_NumSamples > length) {
m_NumSamples = length;
}
SAMPLE *pData = new SAMPLE[m_NumSamples];
QTime timerReadingFile;
timerReadingFile.start();
unsigned int read = soundSource.read(m_NumSamples, pData);
if (read!=m_NumSamples) {
qDebug() << "oh that's embarrasing I couldn't read the track";
return QString();
}
qDebug("reading file took: %d ms" , timerReadingFile.elapsed());
ChromaprintContext* ctx = chromaprint_new(CHROMAPRINT_ALGORITHM_DEFAULT);
// we have 2 channels in mixxx always
chromaprint_start(ctx, m_SampleRate, 2);
QTime timerGeneratingFingerPrint;
timerGeneratingFingerPrint.start();
int success = chromaprint_feed(ctx, pData, m_NumSamples);
if (!success) {
qDebug() << "could not generate fingerprint";
delete [] pData;
return QString();
}
chromaprint_finish(ctx);
void* fprint = NULL;
int size = 0;
int ret = chromaprint_get_raw_fingerprint(ctx, &fprint, &size);
QByteArray fingerprint;
if (ret == 1) {
void* encoded = NULL;
int encoded_size = 0;
chromaprint_encode_fingerprint(fprint, size,
CHROMAPRINT_ALGORITHM_DEFAULT,
&encoded,
&encoded_size, 1);
fingerprint.append(reinterpret_cast<char*>(encoded), encoded_size);
chromaprint_dealloc(fprint);
chromaprint_dealloc(encoded);
}
chromaprint_free(ctx);
delete [] pData;
qDebug("generating fingerprint took: %d ms" , timerGeneratingFingerPrint.elapsed());
return fingerprint;
}
static void *print_thread(void *arg) {
struct GrooveFingerprinterPrivate *p = (GrooveFingerprinterPrivate *)arg;
struct GrooveFingerprinter *printer = &p->externals;
struct GrooveBuffer *buffer;
while (!p->abort_request.load()) {
pthread_mutex_lock(&p->info_head_mutex);
if (p->info_queue_count >= printer->info_queue_size) {
pthread_cond_wait(&p->drain_cond, &p->info_head_mutex);
pthread_mutex_unlock(&p->info_head_mutex);
continue;
}
// we definitely want to unlock the mutex while we wait for the
// next buffer. Otherwise there will be a deadlock when sink_flush or
// sink_purge is called.
pthread_mutex_unlock(&p->info_head_mutex);
int result = groove_sink_buffer_get(p->sink, &buffer, 1);
pthread_mutex_lock(&p->info_head_mutex);
if (result == GROOVE_BUFFER_END) {
// last file info
emit_track_info(p);
// send album info
struct GrooveFingerprinterInfo *info = allocate<GrooveFingerprinterInfo>(1);
if (info) {
info->duration = p->album_duration;
groove_queue_put(p->info_queue, info);
} else {
av_log(NULL, AV_LOG_ERROR, "unable to allocate album fingerprint info\n");
}
p->album_duration = 0.0;
p->info_head = NULL;
p->info_pos = -1.0;
pthread_mutex_unlock(&p->info_head_mutex);
continue;
}
if (result != GROOVE_BUFFER_YES) {
pthread_mutex_unlock(&p->info_head_mutex);
break;
}
if (buffer->item != p->info_head) {
if (p->info_head) {
emit_track_info(p);
}
if (!chromaprint_start(p->chroma_ctx, 44100, 2)) {
av_log(NULL, AV_LOG_ERROR, "unable to start fingerprint\n");
}
p->track_duration = 0.0;
p->info_head = buffer->item;
p->info_pos = buffer->pos;
}
double buffer_duration = buffer->frame_count / (double)buffer->format.sample_rate;
p->track_duration += buffer_duration;
p->album_duration += buffer_duration;
if (!chromaprint_feed(p->chroma_ctx, buffer->data[0], buffer->frame_count * 2)) {
av_log(NULL, AV_LOG_ERROR, "unable to feed fingerprint\n");
}
pthread_mutex_unlock(&p->info_head_mutex);
groove_buffer_unref(buffer);
}
return NULL;
}