ref: ba0ba1066d77d8c8c1ddcf16c2661ed65a52375d
parent: 3d1ca811984309dd36162eb92444b1e921b9b3da
author: Paul Brossier <piem@piem.org>
date: Wed Dec 4 16:40:44 EST 2013
src/io/source_avcodec.c: added first draft
--- /dev/null
+++ b/src/io/source_avcodec.c
@@ -1,0 +1,414 @@
+/*
+ Copyright (C) 2013 Paul Brossier <piem@aubio.org>
+
+ This file is part of aubio.
+
+ aubio is free software: you can redistribute it and/or modify
+ it under the terms of the GNU General Public License as published by
+ the Free Software Foundation, either version 3 of the License, or
+ (at your option) any later version.
+
+ aubio is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ GNU General Public License for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with aubio. If not, see <http://www.gnu.org/licenses/>.
+
+*/
+
+
+#include "config.h"
+
+#ifdef HAVE_AVCODEC
+
+#include <sndfile.h>
+#include <libavcodec/avcodec.h>
+#include <libavformat/avformat.h>
+#include <libavresample/avresample.h>
+#include <libavutil/opt.h>
+#include <stdlib.h>
+
+#include "aubio_priv.h"
+#include "fvec.h"
+#include "fmat.h"
+#include "source_avcodec.h"
+
+#define MAX_CHANNELS 6
+#define MAX_SIZE 4096
+#define MAX_SAMPLES MAX_CHANNELS * MAX_SIZE
+
+#define SHORT_TO_FLOAT(x) (smpl_t)(x * 3.0517578125e-05)
+
+struct _aubio_source_avcodec_t {+ uint_t hop_size;
+ uint_t samplerate;
+ uint_t channels;
+
+ // some data about the file
+ char_t *path;
+ int input_samplerate;
+ int input_channels;
+
+ // avcodec stuff
+ AVFormatContext *avFormatCtx;
+ AVCodecContext *avCodecCtx;
+ AVFrame *avFrame;
+ AVPacket avPacket;
+ AVAudioResampleContext *avr;
+ int16_t *output;
+ int read_samples;
+ int read_index;
+};
+
+aubio_source_avcodec_t * new_aubio_source_avcodec(char_t * path, uint_t samplerate, uint_t hop_size) {+ aubio_source_avcodec_t * s = AUBIO_NEW(aubio_source_avcodec_t);
+ int err;
+
+ if (path == NULL) {+ AUBIO_ERR("Aborted opening null path\n");+ return NULL;
+ }
+
+ s->hop_size = hop_size;
+ s->channels = 1;
+ s->path = path;
+
+ // try opening the file and get some info about it
+ // register all formats and codecs
+ av_register_all();
+
+ // open file
+ AVFormatContext *avFormatCtx = s->avFormatCtx;
+ avFormatCtx = NULL;
+ if ( (err = avformat_open_input(&avFormatCtx, s->path, NULL, NULL) ) < 0 ) {+ uint8_t errorstr_len = 128;
+ char errorstr[errorstr_len];
+ if (av_strerror (err, errorstr, errorstr_len) == 0) {+ AUBIO_ERR("Failed opening %s (%s)\n", s->path, errorstr);+ } else {+ AUBIO_ERR("Failed opening %s (unknown error)\n", s->path);+ }
+ goto beach;
+ }
+
+ // retrieve stream information
+ if ( (err = avformat_find_stream_info(avFormatCtx, NULL)) < 0 ) {+ uint8_t errorstr_len = 128;
+ char errorstr[errorstr_len];
+ if (av_strerror (err, errorstr, errorstr_len) == 0) {+ AUBIO_ERR("Could not find stream information for %s (%s)\n", s->path, errorstr);+ } else {+ AUBIO_ERR("Could not find stream information for %s (unknown error)\n", s->path);+ }
+ goto beach;
+ }
+
+ // Dump information about file onto standard error
+ //av_dump_format(avFormatCtx, 0, s->path, 0);
+
+ uint_t i;
+ sint_t selected_stream = -1;
+ for (i = 0; i < avFormatCtx->nb_streams; i++) {+ if (avFormatCtx->streams[i]->codec->codec_type == AVMEDIA_TYPE_AUDIO) {+ if (selected_stream == -1) {+ selected_stream = i;
+ } else {+ AUBIO_WRN("More than one audio stream in %s, taking the first one\n", s->path);+ }
+ }
+ }
+ if (selected_stream == -1) {+ AUBIO_ERR("No audio stream in %s\n", s->path);+ goto beach;
+ }
+
+ //AUBIO_DBG("Taking stream %d in file %s\n", selected_stream, s->path);+
+ AVCodecContext *avCodecCtx = s->avCodecCtx;
+ avCodecCtx = avFormatCtx->streams[selected_stream]->codec;
+ AVCodec *codec = avcodec_find_decoder(avCodecCtx->codec_id);
+ if (codec == NULL) {+ AUBIO_ERR("Could not find decoder for %s", s->path);+ goto beach;
+ }
+
+ if ( ( err = avcodec_open2(avCodecCtx, codec, NULL) ) < 0) {+ uint8_t errorstr_len = 128;
+ char errorstr[errorstr_len];
+ if (av_strerror (err, errorstr, errorstr_len) == 0) {+ AUBIO_ERR("Could not load codec for %s (%s)\n", s->path, errorstr);+ } else {+ AUBIO_ERR("Could not load codec for %s (unknown error)\n", s->path);+ }
+ goto beach;
+ }
+
+ /* get input specs */
+ s->input_samplerate = avCodecCtx->sample_rate;
+ s->input_channels = avCodecCtx->channels;
+
+ //AUBIO_DBG("input_samplerate: %d\n", s->input_samplerate);+ //AUBIO_DBG("input_channels: %d\n", s->input_channels);+
+ if (samplerate == 0) {+ samplerate = s->input_samplerate;
+ //AUBIO_DBG("sampling rate set to 0, automagically adjusting to %d\n", samplerate);+ }
+ s->samplerate = samplerate;
+
+ int64_t input_layout = av_get_default_channel_layout(s->input_channels);
+ int64_t mono_layout = av_get_default_channel_layout(1);
+
+ AVAudioResampleContext *avr = s->avr;
+ avr = avresample_alloc_context();
+ av_opt_set_int(avr, "in_channel_layout", input_layout, 0);
+ av_opt_set_int(avr, "out_channel_layout", mono_layout, 0);
+ av_opt_set_int(avr, "in_sample_rate", s->input_samplerate, 0);
+ av_opt_set_int(avr, "out_sample_rate", s->samplerate, 0);
+ av_opt_set_int(avr, "in_sample_fmt", avCodecCtx->sample_fmt, 0);
+ av_opt_set_int(avr, "out_sample_fmt", AV_SAMPLE_FMT_S16, 0);
+ if ( ( err = avresample_open(avr) ) < 0) {+ uint8_t errorstr_len = 128;
+ char errorstr[errorstr_len];
+ if (av_strerror (err, errorstr, errorstr_len) == 0) {+ AUBIO_ERR("Could not open AVAudioResampleContext for %s (%s)\n", s->path, errorstr);+ } else {+ AUBIO_ERR("Could not open AVAudioResampleContext for %s (unknown error)\n", s->path);+ }
+ goto beach;
+ }
+
+ AVFrame *avFrame = s->avFrame;
+ avFrame = avcodec_alloc_frame();
+ if (!avFrame) {+ AUBIO_ERR("Could not allocate frame for (%s)\n", s->path);+ }
+ AVPacket avPacket = s->avPacket;
+ av_init_packet(&avPacket);
+
+ s->output = (int16_t *)av_malloc(FF_MIN_BUFFER_SIZE * sizeof(int16_t));
+
+
+ s->read_samples = 0;
+ s->read_index = 0;
+
+ //goto beach;
+
+ /* allocate data for de/interleaving reallocated when needed. */
+ //s->scratch_size = s->input_hop_size * s->input_channels;
+ //s->scratch_data = AUBIO_ARRAY(float,s->scratch_size);
+ s->avFormatCtx = avFormatCtx;
+ s->avCodecCtx = avCodecCtx;
+ s->avFrame = avFrame;
+ s->avPacket = avPacket;
+ s->avr = avr;
+
+ //av_log_set_level(AV_LOG_QUIET);
+
+ return s;
+
+beach:
+ AUBIO_ERR("can not read %s at samplerate %dHz with a hop_size of %d\n",+ s->path, s->samplerate, s->hop_size);
+ del_aubio_source_avcodec(s);
+ return NULL;
+}
+
+void aubio_source_avcodec_readframe(aubio_source_avcodec_t *s, int * read_samples) {+ AVFormatContext *avFormatCtx = s->avFormatCtx;
+ AVCodecContext *avCodecCtx = s->avCodecCtx;
+ AVFrame *avFrame = s->avFrame;
+ AVPacket avPacket = s->avPacket;
+ AVAudioResampleContext *avr = s->avr;
+ int16_t *output = s->output;
+
+ uint_t i;
+ int err = av_read_frame (avFormatCtx, &avPacket);
+ if (err != 0) {+ //AUBIO_ERR("Could not read frame for (%s)\n", s->path);+ *read_samples = 0;
+ return;
+ }
+
+ int got_frame = 0;
+ int len = avcodec_decode_audio4(avCodecCtx, avFrame, &got_frame, &avPacket);
+
+ if (len < 0) {+ AUBIO_ERR("Error while decoding %s\n", s->path);+ //goto beach;
+ return;
+ }
+ if (got_frame == 0) {+ AUBIO_ERR("Could not get frame for (%s)\n", s->path);+ } else {+ //int data_size =
+ av_samples_get_buffer_size(NULL,
+ avCodecCtx->channels, avFrame->nb_samples,
+ avCodecCtx->sample_fmt, 1);
+ //AUBIO_WRN("Got data_size %d frame for (%s)\n", data_size, s->path);+ }
+
+ int in_samples = avFrame->nb_samples;
+ for (i = 0; i < in_samples; i ++) {+ //AUBIO_WRN("%d\n", avFrame->data[i]);+ }
+
+
+#if 1
+ int in_plane_size = 0; //avFrame->linesize[0];
+ int out_plane_size = 0; //sizeof(float); //in_samples * sizeof(float);
+ int max_out_samples = FF_MIN_BUFFER_SIZE;
+ if (avresample_convert ( avr,
+ (void **)&output, out_plane_size, max_out_samples,
+ (void **)avFrame->data, in_plane_size, in_samples) < 0) {+ AUBIO_ERR("Could not convert frame (%s)\n", s->path);+ }
+ //AUBIO_ERR("Got in_plane_size %d frame for (%s)\n", in_plane_size, s->path);+ //AUBIO_WRN("Delay is %d for %s\n", avresample_get_delay(avr), s->path);+ //AUBIO_WRN("max_out_samples is %d for FF_MIN_BUFFER_SIZE %d\n", + // max_out_samples, FF_MIN_BUFFER_SIZE);
+
+ int out_samples = avresample_available(avr) + (avresample_get_delay(avr)
+ + in_samples) * s->samplerate / s->input_samplerate;
+ //AUBIO_WRN("Converted %d to %d samples\n", in_samples, out_samples);+ for (i = 0; i < out_samples; i ++) {+ //AUBIO_DBG("%f\n", SHORT_TO_FLOAT(output[i]));+ }
+#else
+ int in_plane_size = 0; //avFrame->linesize[0];
+ int out_plane_size = 0; //sizeof(float); //in_samples * sizeof(float);
+ int max_out_samples = FF_MIN_BUFFER_SIZE;
+ if (avresample_convert ( avr,
+ NULL, out_plane_size, max_out_samples,
+ (uint8_t **)avFrame->data, in_plane_size, in_samples) < 0) {+ AUBIO_ERR("Could not convert frame (%s)\n", s->path);+ }
+ AUBIO_ERR("Got in_plane_size %d frame for (%s)\n", in_plane_size, s->path);+ AUBIO_WRN("Delay is %d for %s\n", avresample_get_delay(avr), s->path);+ AUBIO_WRN("max_out_samples is %d for FF_MIN_BUFFER_SIZE %d\n", + max_out_samples, FF_MIN_BUFFER_SIZE);
+
+ //int out_samples = avresample_available (avr) +
+ // (avresample_get_delay(avr) + in_samples) * s->samplerate / s->input_samplerate;
+ int out_samples = avresample_available(avr);
+ AUBIO_WRN("Found %d samples available for %s\n", avresample_available(avr), s->path);+ //
+ err = avresample_read (avr, (uint8_t **)output, out_samples);
+ if (err != 0) {+ AUBIO_WRN("Error %d while reading %s\n", err, s->path);+ }
+ AUBIO_WRN("Converted %d to %d samples\n", in_samples, out_samples);+
+ for (i = 0; i < out_samples; i ++) {+ AUBIO_WRN("%f\n", SHORT_TO_FLOAT(output[i]));+ }
+#endif
+ s->avFormatCtx = avFormatCtx;
+ s->avCodecCtx = avCodecCtx;
+ s->avFrame = avFrame;
+ s->avPacket = avPacket;
+ s->avr = avr;
+ s->output = output;
+
+ *read_samples = out_samples;
+}
+
+
+void aubio_source_avcodec_do(aubio_source_avcodec_t * s, fvec_t * read_data, uint_t * read){+ uint_t i;
+ //AUBIO_DBG("entering 'do' with %d, %d\n", s->read_samples, s->read_index);+ // begin reading
+ if (s->read_samples == 0) {+ int avcodec_read = 0;
+ aubio_source_avcodec_readframe(s, &avcodec_read);
+ s->read_samples += avcodec_read;
+ s->read_index = 0;
+ }
+ if (s->read_samples < s->hop_size) {+ // write the end of the buffer to the beginning of read_data
+ int partial = s->read_samples;
+ for (i = 0; i < partial; i++) {+ read_data->data[i] = SHORT_TO_FLOAT(s->output[i + s->read_index]);
+ }
+ s->read_samples = 0;
+ s->read_index = 0;
+ // get more data
+ int avcodec_read = 0;
+ aubio_source_avcodec_readframe(s, &avcodec_read);
+ s->read_samples += avcodec_read;
+ s->read_index = 0;
+ // write the beginning of the buffer to the end of read_data
+ int end = MIN((int)(s->hop_size), s->read_samples);
+ if (avcodec_read == 0) { + end = partial;
+ }
+ for (i = partial; i < end; i++) {+ read_data->data[i] = SHORT_TO_FLOAT(s->output[i - partial + s->read_index]);
+ }
+ if (end < s->hop_size) {+ for (i = end; i < s->hop_size; i++) {+ read_data->data[i] = 0.;
+ }
+ }
+ s->read_index += partial;
+ s->read_samples -= partial;
+ *read = end;
+ } else {+ for (i = 0; i < s->hop_size; i++) {+ read_data->data[i] = SHORT_TO_FLOAT(s->output[i + s->read_index]);
+ }
+ s->read_index += s->hop_size;
+ s->read_samples -= s->hop_size;
+ *read = s->hop_size;
+ }
+}
+
+void aubio_source_avcodec_do_multi(aubio_source_avcodec_t * s, fmat_t * read_data, uint_t * read){+ //uint_t i,j, input_channels = s->input_channels;
+}
+
+uint_t aubio_source_avcodec_get_samplerate(aubio_source_avcodec_t * s) {+ return s->samplerate;
+}
+
+uint_t aubio_source_avcodec_get_channels(aubio_source_avcodec_t * s) {+ return s->input_channels;
+}
+
+uint_t aubio_source_avcodec_seek (aubio_source_avcodec_t * s, uint_t pos) {+ //uint_t resampled_pos = (uint_t)ROUND(pos * s->input_samplerate * 1. / s->samplerate);
+ return 0; //sf_seek (s->handle, resampled_pos, SEEK_SET);
+}
+
+void del_aubio_source_avcodec(aubio_source_avcodec_t * s){+ if (!s) return;
+ if (s->output != NULL) {+ av_free(s->output);
+ }
+ if (s->avr != NULL) {+ avresample_close( s->avr );
+ av_free ( s->avr );
+ }
+ s->avr = NULL;
+ if (s->avFrame != NULL) {+ avcodec_free_frame( &(s->avFrame) );
+ }
+ s->avFrame = NULL;
+ if ( &(s->avPacket) != NULL) {+ av_free_packet( &(s->avPacket) );
+ }
+ if (s->avCodecCtx != NULL) {+ avcodec_close ( s->avCodecCtx );
+ }
+ s->avCodecCtx = NULL;
+ if (s->avFormatCtx != NULL) {+ avformat_close_input ( &(s->avFormatCtx) );
+ }
+ s->avFrame = NULL;
+ s->avFormatCtx = NULL;
+ AUBIO_FREE(s);
+}
+
+#endif /* HAVE_SNDFILE */
--- /dev/null
+++ b/src/io/source_avcodec.h
@@ -1,0 +1,134 @@
+/*
+ Copyright (C) 2013 Paul Brossier <piem@aubio.org>
+
+ This file is part of aubio.
+
+ aubio is free software: you can redistribute it and/or modify
+ it under the terms of the GNU General Public License as published by
+ the Free Software Foundation, either version 3 of the License, or
+ (at your option) any later version.
+
+ aubio is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ GNU General Public License for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with aubio. If not, see <http://www.gnu.org/licenses/>.
+
+*/
+
+#ifndef _AUBIO_SOURCE_AVCODEC_H
+#define _AUBIO_SOURCE_AVCODEC_H
+
+/** \file
+
+ Read from file using [libavcodec](http://libav.org/)
+
+ Avoid including this file directly! Prefer using ::aubio_source_t instead to
+ make your code portable.
+
+ To write to file, use ::aubio_sink_t.
+
+ \example io/test-source_avcodec.c
+
+*/
+
+#ifdef __cplusplus
+extern "C" {+#endif
+
+/** avcodec media source object */
+typedef struct _aubio_source_avcodec_t aubio_source_avcodec_t;
+
+/**
+
+ create new ::aubio_source_avcodec_t
+
+ \param uri the file path or uri to read from
+ \param samplerate sampling rate to view the fie at
+ \param hop_size the size of the blocks to read from
+
+ Creates a new source object. If `0` is passed as `samplerate`, the sample
+ rate of the original file is used.
+
+ The samplerate of newly created source can be obtained using
+ ::aubio_source_avcodec_get_samplerate.
+
+*/
+aubio_source_avcodec_t * new_aubio_source_avcodec(char_t * uri, uint_t samplerate, uint_t hop_size);
+
+/**
+
+ read monophonic vector of length hop_size from source object
+
+ \param s source object, created with ::new_aubio_source_avcodec
+ \param read_to ::fvec_t of data to read to
+ \param read upon returns, equals to number of frames actually read
+
+ Upon returns, `read` contains the number of frames actually read from the
+ source. `hop_size` if enough frames could be read, less otherwise.
+
+*/
+void aubio_source_avcodec_do(aubio_source_avcodec_t * s, fvec_t * read_to, uint_t * read);
+
+/**
+
+ read polyphonic vector of length hop_size from source object
+
+ \param s source object, created with ::new_aubio_source_avcodec
+ \param read_to ::fmat_t of data to read to
+ \param read upon returns, equals to number of frames actually read
+
+ Upon returns, `read` contains the number of frames actually read from the
+ source. `hop_size` if enough frames could be read, less otherwise.
+
+*/
+void aubio_source_avcodec_do_multi(aubio_source_avcodec_t * s, fmat_t * read_to, uint_t * read);
+
+/**
+
+ get samplerate of source object
+
+ \param s source object, created with ::new_aubio_source_avcodec
+ \return samplerate, in Hz
+
+*/
+uint_t aubio_source_avcodec_get_samplerate(aubio_source_avcodec_t * s);
+
+/**
+
+ get number of channels of source object
+
+ \param s source object, created with ::new_aubio_source_avcodec
+ \return number of channels
+
+*/
+uint_t aubio_source_avcodec_get_channels (aubio_source_avcodec_t * s);
+
+/**
+
+ seek source object
+
+ \param s source object, created with ::new_aubio_source_avcodec
+ \param pos position to seek to, in frames
+
+ \return 0 if sucessful, non-zero on failure
+
+*/
+uint_t aubio_source_avcodec_seek (aubio_source_avcodec_t *s, uint_t pos);
+
+/**
+
+ close source and cleanup memory
+
+ \param s source object, created with ::new_aubio_source_avcodec
+
+*/
+void del_aubio_source_avcodec(aubio_source_avcodec_t * s);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _AUBIO_SOURCE_AVCODEC_H */
--- a/src/wscript_build
+++ b/src/wscript_build
@@ -5,6 +5,10 @@
uselib += ['FFTW3', 'FFTW3F']
uselib += ['SAMPLERATE']
uselib += ['SNDFILE']
+uselib += ['AVCODEC']
+uselib += ['AVFORMAT']
+uselib += ['AVRESAMPLE']
+uselib += ['AVUTIL']
uselib += ['JACK']
uselib += ['LASH']
--- /dev/null
+++ b/tests/src/io/test-source_avcodec.c
@@ -1,0 +1,47 @@
+#define AUBIO_UNSTABLE 1
+#include <aubio.h>
+#include "utils_tests.h"
+#include "config.h"
+
+// this file uses the unstable aubio api, please use aubio_source instead
+// see src/io/source.h and tests/src/source/test-source.c
+
+int main (int argc, char **argv)
+{+ uint_t err = 0;
+ if (argc < 2) {+ err = 2;
+ PRINT_ERR("not enough arguments\n");+ PRINT_MSG("usage: %s <source_path> [samplerate]\n", argv[0]);+ return err;
+ }
+
+#ifdef HAVE_AVCODEC
+ uint_t samplerate = 0;
+ uint_t hop_size = 256;
+ uint_t n_frames = 0, read = 0;
+ if ( argc == 3 ) samplerate = atoi(argv[2]);
+
+ char_t *source_path = argv[1];
+
+ fvec_t *vec = new_fvec(hop_size);
+ aubio_source_avcodec_t * s = new_aubio_source_avcodec(source_path, samplerate, hop_size);
+ if (samplerate == 0 ) samplerate = aubio_source_avcodec_get_samplerate(s);
+
+ if (!s) { err = 1; goto beach; }+
+ do {+ aubio_source_avcodec_do(s, vec, &read);
+ fvec_print (vec);
+ n_frames += read;
+ } while ( read == hop_size );
+
+beach:
+ del_aubio_source_avcodec (s);
+ del_fvec (vec);
+#else
+ err = 3;
+ PRINT_ERR("aubio was not compiled with aubio_source_avcodec\n");+#endif /* HAVE_AVCODEC */
+ return err;
+}
--- a/wscript
+++ b/wscript
@@ -56,6 +56,8 @@
help_str = 'compile with LASH (auto)', help_disable_str = 'disable LASH' )
add_option_enable_disable(ctx, 'sndfile', default = None,
help_str = 'compile with sndfile (auto)', help_disable_str = 'disable sndfile')
+ add_option_enable_disable(ctx, 'avcodec', default = None,
+ help_str = 'compile with libavcodec (auto)', help_disable_str = 'disable libavcodec')
add_option_enable_disable(ctx, 'samplerate', default = None,
help_str = 'compile with samplerate (auto)', help_disable_str = 'disable samplerate')
add_option_enable_disable(ctx, 'memcpy', default = True,
@@ -198,6 +200,16 @@
if (ctx.options.enable_lash != False):
ctx.check_cfg(package = 'lash-1.0', atleast_version = '0.5.0',
args = '--cflags --libs', uselib_store = 'LASH', mandatory = False)
+
+ if (ctx.options.enable_avcodec != False):
+ ctx.check_cfg(package = 'libavcodec', atleast_version = '54.35.0',
+ args = '--cflags --libs', uselib_store = 'AVCODEC', mandatory = False)
+ ctx.check_cfg(package = 'libavformat', atleast_version = '52.3.0',
+ args = '--cflags --libs', uselib_store = 'AVFORMAT', mandatory = False)
+ ctx.check_cfg(package = 'libavutil', atleast_version = '52.3.0',
+ args = '--cflags --libs', uselib_store = 'AVUTIL', mandatory = False)
+ ctx.check_cfg(package = 'libavresample', atleast_version = '1.0.1',
+ args = '--cflags --libs', uselib_store = 'AVRESAMPLE', mandatory = False)
# write configuration header
ctx.write_config_header('src/config.h')