source: src/io/source_avcodec.c @ 9d8cdb55

feature/autosinkfeature/cnnfeature/cnn_orgfeature/constantqfeature/crepefeature/crepe_orgfeature/pitchshiftfeature/pydocstringsfeature/timestretchfix/ffmpeg5sampleryinfft+
Last change on this file since 9d8cdb55 was 1504b7c, checked in by Paul Brossier <piem@piem.org>, 8 years ago

src/io/source_avcodec.c: fix for old libavcodec54

  • Property mode set to 100644
File size: 16.4 KB
Line 
1/*
2  Copyright (C) 2013 Paul Brossier <piem@aubio.org>
3
4  This file is part of aubio.
5
6  aubio 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  aubio 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 aubio.  If not, see <http://www.gnu.org/licenses/>.
18
19*/
20
21
22#include "config.h"
23
24#ifdef HAVE_LIBAV
25
26#include <libavcodec/avcodec.h>
27#include <libavformat/avformat.h>
28#include <libavresample/avresample.h>
29#include <libavutil/opt.h>
30#include <stdlib.h>
31
32// determine whether we use libavformat from ffmpeg or from libav
33#define FFMPEG_LIBAVFORMAT (LIBAVFORMAT_VERSION_MICRO > 99 )
34// max_analyze_duration2 was used from ffmpeg libavformat 55.43.100 through 57.2.100
35#define FFMPEG_LIBAVFORMAT_MAX_DUR2 FFMPEG_LIBAVFORMAT && ( \
36      (LIBAVFORMAT_VERSION_MAJOR == 55 && LIBAVFORMAT_VERSION_MINOR >= 43) \
37      || (LIBAVFORMAT_VERSION_MAJOR == 56) \
38      || (LIBAVFORMAT_VERSION_MAJOR == 57 && LIBAVFORMAT_VERSION_MINOR < 2) \
39      )
40
41// backward compatibility with libavcodec55
42#if LIBAVCODEC_VERSION_INT < AV_VERSION_INT(55,28,1)
43#warning "libavcodec55 is deprecated"
44#define HAVE_AUBIO_LIBAVCODEC_DEPRECATED 1
45#define av_frame_alloc  avcodec_alloc_frame
46#define av_frame_free avcodec_free_frame
47#define av_packet_unref av_free_packet
48#endif
49
50#include "aubio_priv.h"
51#include "fvec.h"
52#include "fmat.h"
53#include "source_avcodec.h"
54
55#define AUBIO_AVCODEC_MAX_BUFFER_SIZE FF_MIN_BUFFER_SIZE
56
57struct _aubio_source_avcodec_t {
58  uint_t hop_size;
59  uint_t samplerate;
60  uint_t channels;
61
62  // some data about the file
63  char_t *path;
64  uint_t input_samplerate;
65  uint_t input_channels;
66
67  // avcodec stuff
68  AVFormatContext *avFormatCtx;
69  AVCodecContext *avCodecCtx;
70  AVFrame *avFrame;
71  AVPacket avPacket;
72  AVAudioResampleContext *avr;
73  smpl_t *output;
74  uint_t read_samples;
75  uint_t read_index;
76  sint_t selected_stream;
77  uint_t eof;
78  uint_t multi;
79};
80
81// hack to create or re-create the context the first time _do or _do_multi is called
82void aubio_source_avcodec_reset_resampler(aubio_source_avcodec_t * s, uint_t multi);
83void aubio_source_avcodec_readframe(aubio_source_avcodec_t *s, uint_t * read_samples);
84
85uint_t aubio_source_avcodec_has_network_url(aubio_source_avcodec_t *s);
86
87uint_t aubio_source_avcodec_has_network_url(aubio_source_avcodec_t *s) {
88  char proto[20], authorization[256], hostname[128], uripath[256];
89  int proto_size = 20, authorization_size = 256, hostname_size = 128,
90      *port_ptr = 0, path_size = 256;
91  av_url_split(proto, proto_size, authorization, authorization_size, hostname,
92      hostname_size, port_ptr, uripath, path_size, s->path);
93  if (strlen(proto)) {
94    return 1;
95  }
96  return 0;
97}
98
99
100aubio_source_avcodec_t * new_aubio_source_avcodec(const char_t * path, uint_t samplerate, uint_t hop_size) {
101  aubio_source_avcodec_t * s = AUBIO_NEW(aubio_source_avcodec_t);
102  int err;
103  if (path == NULL) {
104    AUBIO_ERR("source_avcodec: Aborted opening null path\n");
105    goto beach;
106  }
107  if ((sint_t)samplerate < 0) {
108    AUBIO_ERR("source_avcodec: Can not open %s with samplerate %d\n", path, samplerate);
109    goto beach;
110  }
111  if ((sint_t)hop_size <= 0) {
112    AUBIO_ERR("source_avcodec: Can not open %s with hop_size %d\n", path, hop_size);
113    goto beach;
114  }
115
116  s->hop_size = hop_size;
117  s->channels = 1;
118
119  if (s->path) AUBIO_FREE(s->path);
120  s->path = AUBIO_ARRAY(char_t, strnlen(path, PATH_MAX) + 1);
121  strncpy(s->path, path, strnlen(path, PATH_MAX) + 1);
122
123  // register all formats and codecs
124  av_register_all();
125
126  if (aubio_source_avcodec_has_network_url(s)) {
127    avformat_network_init();
128  }
129
130  // try opening the file and get some info about it
131  AVFormatContext *avFormatCtx = s->avFormatCtx;
132  avFormatCtx = NULL;
133  if ( (err = avformat_open_input(&avFormatCtx, s->path, NULL, NULL) ) < 0 ) {
134    char errorstr[256];
135    av_strerror (err, errorstr, sizeof(errorstr));
136    AUBIO_ERR("source_avcodec: Failed opening %s (%s)\n", s->path, errorstr);
137    goto beach;
138  }
139
140  // try to make sure max_analyze_duration is big enough for most songs
141#if FFMPEG_LIBAVFORMAT_MAX_DUR2
142  avFormatCtx->max_analyze_duration2 *= 100;
143#else
144  avFormatCtx->max_analyze_duration *= 100;
145#endif
146
147  // retrieve stream information
148  if ( (err = avformat_find_stream_info(avFormatCtx, NULL)) < 0 ) {
149    char errorstr[256];
150    av_strerror (err, errorstr, sizeof(errorstr));
151    AUBIO_ERR("source_avcodec: Could not find stream information " "for %s (%s)\n", s->path,
152        errorstr);
153    goto beach;
154  }
155
156  // dump information about file onto standard error
157  //av_dump_format(avFormatCtx, 0, s->path, 0);
158
159  // look for the first audio stream
160  uint_t i;
161  sint_t selected_stream = -1;
162  for (i = 0; i < avFormatCtx->nb_streams; i++) {
163#if FF_API_LAVF_AVCTX
164    if (avFormatCtx->streams[i]->codecpar->codec_type == AVMEDIA_TYPE_AUDIO) {
165#else
166    if (avFormatCtx->streams[i]->codec->codec_type == AVMEDIA_TYPE_AUDIO) {
167#endif
168      if (selected_stream == -1) {
169        selected_stream = i;
170      } else {
171        AUBIO_WRN("source_avcodec: More than one audio stream in %s, "
172            "taking the first one\n", s->path);
173      }
174    }
175  }
176  if (selected_stream == -1) {
177    AUBIO_ERR("source_avcodec: No audio stream in %s\n", s->path);
178    goto beach;
179  }
180  //AUBIO_DBG("Taking stream %d in file %s\n", selected_stream, s->path);
181  s->selected_stream = selected_stream;
182
183  AVCodecContext *avCodecCtx = s->avCodecCtx;
184#if FF_API_LAVF_AVCTX
185  AVCodecParameters *codecpar = avFormatCtx->streams[selected_stream]->codecpar;
186  if (codecpar == NULL) {
187    AUBIO_ERR("source_avcodec: Could not find decoder for %s", s->path);
188    goto beach;
189  }
190  AVCodec *codec = avcodec_find_decoder(codecpar->codec_id);
191
192  /* Allocate a codec context for the decoder */
193  avCodecCtx = avcodec_alloc_context3(codec);
194  if (!avCodecCtx) {
195    AUBIO_ERR("source_avcodec: Failed to allocate the %s codec context for path %s\n",
196        av_get_media_type_string(AVMEDIA_TYPE_AUDIO), s->path);
197    goto beach;
198  }
199#else
200  avCodecCtx = avFormatCtx->streams[selected_stream]->codec;
201  AVCodec *codec = avcodec_find_decoder(avCodecCtx->codec_id);
202#endif
203  if (codec == NULL) {
204    AUBIO_ERR("source_avcodec: Could not find decoder for %s", s->path);
205    goto beach;
206  }
207
208#if FF_API_LAVF_AVCTX
209  /* Copy codec parameters from input stream to output codec context */
210  if ((err = avcodec_parameters_to_context(avCodecCtx, codecpar)) < 0) {
211    AUBIO_ERR("source_avcodec: Failed to copy %s codec parameters to decoder context for %s\n",
212       av_get_media_type_string(AVMEDIA_TYPE_AUDIO), s->path);
213    goto beach;
214  }
215#endif
216
217  if ( ( err = avcodec_open2(avCodecCtx, codec, NULL) ) < 0) {
218    char errorstr[256];
219    av_strerror (err, errorstr, sizeof(errorstr));
220    AUBIO_ERR("source_avcodec: Could not load codec for %s (%s)\n", s->path, errorstr);
221    goto beach;
222  }
223
224  /* get input specs */
225  s->input_samplerate = avCodecCtx->sample_rate;
226  s->input_channels   = avCodecCtx->channels;
227  //AUBIO_DBG("input_samplerate: %d\n", s->input_samplerate);
228  //AUBIO_DBG("input_channels: %d\n", s->input_channels);
229
230  if (samplerate == 0) {
231    s->samplerate = s->input_samplerate;
232  } else {
233    s->samplerate = samplerate;
234  }
235
236  if (s->samplerate >  s->input_samplerate) {
237    AUBIO_WRN("source_avcodec: upsampling %s from %d to %d\n", s->path,
238        s->input_samplerate, s->samplerate);
239  }
240
241  AVFrame *avFrame = s->avFrame;
242  avFrame = av_frame_alloc();
243  if (!avFrame) {
244    AUBIO_ERR("source_avcodec: Could not allocate frame for (%s)\n", s->path);
245  }
246
247  /* allocate output for avr */
248  s->output = (smpl_t *)av_malloc(AUBIO_AVCODEC_MAX_BUFFER_SIZE * sizeof(smpl_t));
249
250  s->read_samples = 0;
251  s->read_index = 0;
252
253  s->avFormatCtx = avFormatCtx;
254  s->avCodecCtx = avCodecCtx;
255  s->avFrame = avFrame;
256
257  // default to mono output
258  aubio_source_avcodec_reset_resampler(s, 0);
259
260  s->eof = 0;
261  s->multi = 0;
262
263  //av_log_set_level(AV_LOG_QUIET);
264
265  return s;
266
267beach:
268  //AUBIO_ERR("can not read %s at samplerate %dHz with a hop_size of %d\n",
269  //    s->path, s->samplerate, s->hop_size);
270  del_aubio_source_avcodec(s);
271  return NULL;
272}
273
274void aubio_source_avcodec_reset_resampler(aubio_source_avcodec_t * s, uint_t multi) {
275  // create or reset resampler to/from mono/multi-channel
276  if ( (multi != s->multi) || (s->avr == NULL) ) {
277    int64_t input_layout = av_get_default_channel_layout(s->input_channels);
278    uint_t output_channels = multi ? s->input_channels : 1;
279    int64_t output_layout = av_get_default_channel_layout(output_channels);
280    AVAudioResampleContext *avr = avresample_alloc_context();
281    AVAudioResampleContext *oldavr = s->avr;
282
283    av_opt_set_int(avr, "in_channel_layout",  input_layout,           0);
284    av_opt_set_int(avr, "out_channel_layout", output_layout,          0);
285    av_opt_set_int(avr, "in_sample_rate",     s->input_samplerate,    0);
286    av_opt_set_int(avr, "out_sample_rate",    s->samplerate,          0);
287    av_opt_set_int(avr, "in_sample_fmt",      s->avCodecCtx->sample_fmt, 0);
288#if HAVE_AUBIO_DOUBLE
289    av_opt_set_int(avr, "out_sample_fmt",     AV_SAMPLE_FMT_DBL,      0);
290#else
291    av_opt_set_int(avr, "out_sample_fmt",     AV_SAMPLE_FMT_FLT,      0);
292#endif
293    // TODO: use planar?
294    //av_opt_set_int(avr, "out_sample_fmt",     AV_SAMPLE_FMT_FLTP,      0);
295    int err;
296    if ( ( err = avresample_open(avr) ) < 0) {
297      char errorstr[256];
298      av_strerror (err, errorstr, sizeof(errorstr));
299      AUBIO_ERR("source_avcodec: Could not open AVAudioResampleContext for %s (%s)\n",
300          s->path, errorstr);
301      //goto beach;
302      return;
303    }
304    s->avr = avr;
305    if (oldavr != NULL) {
306      avresample_close( oldavr );
307      av_free ( oldavr );
308      oldavr = NULL;
309    }
310    s->multi = multi;
311  }
312}
313
314void aubio_source_avcodec_readframe(aubio_source_avcodec_t *s, uint_t * read_samples) {
315  AVFormatContext *avFormatCtx = s->avFormatCtx;
316  AVCodecContext *avCodecCtx = s->avCodecCtx;
317  AVFrame *avFrame = s->avFrame;
318  AVPacket avPacket = s->avPacket;
319  av_init_packet (&avPacket);
320  AVAudioResampleContext *avr = s->avr;
321  smpl_t *output = s->output;
322  *read_samples = 0;
323
324  do
325  {
326    int err = av_read_frame (avFormatCtx, &avPacket);
327    if (err == AVERROR_EOF) {
328      s->eof = 1;
329      goto beach;
330    }
331    if (err != 0) {
332      char errorstr[256];
333      av_strerror (err, errorstr, sizeof(errorstr));
334      AUBIO_ERR("source_avcodec: could not read frame in %s (%s)\n", s->path, errorstr);
335      goto beach;
336    }
337  } while (avPacket.stream_index != s->selected_stream);
338
339  int got_frame = 0;
340#if FF_API_LAVF_AVCTX
341  int ret = avcodec_send_packet(avCodecCtx, &avPacket);
342  if (ret < 0 && ret != AVERROR_EOF) {
343    AUBIO_ERR("source_avcodec: error when sending packet for %s\n", s->path);
344    goto beach;
345  }
346  ret = avcodec_receive_frame(avCodecCtx, avFrame);
347  if (ret >= 0) {
348    got_frame = 1;
349  }
350  if (ret < 0) {
351    if (ret == AVERROR(EAGAIN)) {
352      AUBIO_WRN("source_avcodec: output is not available right now - user must try to send new input\n");
353    } else if (ret == AVERROR_EOF) {
354      AUBIO_WRN("source_avcodec: the decoder has been fully flushed, and there will be no more output frames\n");
355    } else {
356      AUBIO_ERR("source_avcodec: decoding errors on %s\n", s->path);
357      goto beach;
358    }
359  }
360#else
361  int len = avcodec_decode_audio4(avCodecCtx, avFrame, &got_frame, &avPacket);
362
363  if (len < 0) {
364    AUBIO_ERR("source_avcodec: error while decoding %s\n", s->path);
365    goto beach;
366  }
367#endif
368  if (got_frame == 0) {
369    AUBIO_WRN("source_avcodec: did not get a frame when reading %s\n", s->path);
370    goto beach;
371  }
372
373  int in_linesize = 0;
374  av_samples_get_buffer_size(&in_linesize, avCodecCtx->channels,
375      avFrame->nb_samples, avCodecCtx->sample_fmt, 1);
376  int in_samples = avFrame->nb_samples;
377  int out_linesize = 0;
378  int max_out_samples = AUBIO_AVCODEC_MAX_BUFFER_SIZE;
379  int out_samples = avresample_convert ( avr,
380        (uint8_t **)&output, out_linesize, max_out_samples,
381        (uint8_t **)avFrame->data, in_linesize, in_samples);
382  if (out_samples <= 0) {
383    AUBIO_WRN("source_avcodec: no sample found while converting frame (%s)\n", s->path);
384    goto beach;
385  }
386
387  *read_samples = out_samples;
388
389beach:
390  s->avFormatCtx = avFormatCtx;
391  s->avCodecCtx = avCodecCtx;
392  s->avFrame = avFrame;
393  s->avr = avr;
394  s->output = output;
395
396  av_packet_unref(&avPacket);
397}
398
399void aubio_source_avcodec_do(aubio_source_avcodec_t * s, fvec_t * read_data, uint_t * read){
400  uint_t i;
401  uint_t end = 0;
402  uint_t total_wrote = 0;
403  // switch from multi
404  if (s->multi == 1) aubio_source_avcodec_reset_resampler(s, 0);
405  while (total_wrote < s->hop_size) {
406    end = MIN(s->read_samples - s->read_index, s->hop_size - total_wrote);
407    for (i = 0; i < end; i++) {
408      read_data->data[i + total_wrote] = s->output[i + s->read_index];
409    }
410    total_wrote += end;
411    if (total_wrote < s->hop_size) {
412      uint_t avcodec_read = 0;
413      aubio_source_avcodec_readframe(s, &avcodec_read);
414      s->read_samples = avcodec_read;
415      s->read_index = 0;
416      if (s->eof) {
417        break;
418      }
419    } else {
420      s->read_index += end;
421    }
422  }
423  if (total_wrote < s->hop_size) {
424    for (i = total_wrote; i < s->hop_size; i++) {
425      read_data->data[i] = 0.;
426    }
427  }
428  *read = total_wrote;
429}
430
431void aubio_source_avcodec_do_multi(aubio_source_avcodec_t * s, fmat_t * read_data, uint_t * read){
432  uint_t i,j;
433  uint_t end = 0;
434  uint_t total_wrote = 0;
435  // switch from mono
436  if (s->multi == 0) aubio_source_avcodec_reset_resampler(s, 1);
437  while (total_wrote < s->hop_size) {
438    end = MIN(s->read_samples - s->read_index, s->hop_size - total_wrote);
439    for (j = 0; j < read_data->height; j++) {
440      for (i = 0; i < end; i++) {
441        read_data->data[j][i + total_wrote] =
442          s->output[(i + s->read_index) * s->input_channels + j];
443      }
444    }
445    total_wrote += end;
446    if (total_wrote < s->hop_size) {
447      uint_t avcodec_read = 0;
448      aubio_source_avcodec_readframe(s, &avcodec_read);
449      s->read_samples = avcodec_read;
450      s->read_index = 0;
451      if (s->eof) {
452        break;
453      }
454    } else {
455      s->read_index += end;
456    }
457  }
458  if (total_wrote < s->hop_size) {
459    for (j = 0; j < read_data->height; j++) {
460      for (i = total_wrote; i < s->hop_size; i++) {
461        read_data->data[j][i] = 0.;
462      }
463    }
464  }
465  *read = total_wrote;
466}
467
468uint_t aubio_source_avcodec_get_samplerate(const aubio_source_avcodec_t * s) {
469  return s->samplerate;
470}
471
472uint_t aubio_source_avcodec_get_channels(const aubio_source_avcodec_t * s) {
473  return s->input_channels;
474}
475
476uint_t aubio_source_avcodec_seek (aubio_source_avcodec_t * s, uint_t pos) {
477  int64_t resampled_pos = (uint_t)ROUND(pos * (s->input_samplerate * 1. / s->samplerate));
478  int64_t min_ts = MAX(resampled_pos - 2000, 0);
479  int64_t max_ts = MIN(resampled_pos + 2000, INT64_MAX);
480  int seek_flags = AVSEEK_FLAG_FRAME | AVSEEK_FLAG_ANY;
481  int ret = avformat_seek_file(s->avFormatCtx, s->selected_stream,
482      min_ts, resampled_pos, max_ts, seek_flags);
483  if (ret < 0) {
484    AUBIO_ERR("Failed seeking to %d in file %s", pos, s->path);
485  }
486  // reset read status
487  s->eof = 0;
488  s->read_index = 0;
489  s->read_samples = 0;
490  // reset the AVAudioResampleContext
491  avresample_close(s->avr);
492  avresample_open(s->avr);
493  return ret;
494}
495
496uint_t aubio_source_avcodec_get_duration (aubio_source_avcodec_t * s) {
497  if (s && &(s->avFormatCtx) != NULL) {
498    int64_t duration = s->avFormatCtx->duration;
499    return s->samplerate * ((uint_t)duration / 1e6 );
500  }
501  return 0;
502}
503
504uint_t aubio_source_avcodec_close(aubio_source_avcodec_t * s) {
505  if (s->avr != NULL) {
506    avresample_close( s->avr );
507    av_free ( s->avr );
508  }
509  s->avr = NULL;
510  if (s->avCodecCtx != NULL) {
511    avcodec_close ( s->avCodecCtx );
512  }
513  s->avCodecCtx = NULL;
514  if (s->avFormatCtx != NULL) {
515    avformat_close_input(&s->avFormatCtx);
516#ifndef HAVE_AUBIO_LIBAVCODEC_DEPRECATED // avoid crash on old libavcodec54
517    avformat_free_context(s->avFormatCtx);
518#endif
519    s->avFormatCtx = NULL;
520  }
521  av_packet_unref(&s->avPacket);
522  return AUBIO_OK;
523}
524
525void del_aubio_source_avcodec(aubio_source_avcodec_t * s){
526  if (!s) return;
527  aubio_source_avcodec_close(s);
528  if (s->output != NULL) {
529    av_free(s->output);
530  }
531  s->output = NULL;
532  if (s->avFrame != NULL) {
533    av_frame_free( &(s->avFrame) );
534  }
535  if (s->path) AUBIO_FREE(s->path);
536  s->avFrame = NULL;
537  AUBIO_FREE(s);
538}
539
540#endif /* HAVE_LIBAV */
Note: See TracBrowser for help on using the repository browser.