source: src/io/source_avcodec.c @ ccdc51b

feature/autosinkfeature/cnnfeature/cnn_orgfeature/constantqfeature/crepefeature/crepe_orgfeature/pitchshiftfeature/pydocstringsfeature/timestretchfix/ffmpeg5sampler
Last change on this file since ccdc51b was 4d44531, checked in by Paul Brossier <piem@piem.org>, 8 years ago

src/io/source_avcodec.c: make sure seek position is >= 0

  • Property mode set to 100644
File size: 16.8 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#include "aubio_priv.h"
22
23#ifdef HAVE_LIBAV
24
25#include <libavcodec/avcodec.h>
26#include <libavformat/avformat.h>
27#include <libavresample/avresample.h>
28#include <libavutil/opt.h>
29#include <stdlib.h>
30
31// determine whether we use libavformat from ffmpeg or from libav
32#define FFMPEG_LIBAVFORMAT (LIBAVFORMAT_VERSION_MICRO > 99 )
33// max_analyze_duration2 was used from ffmpeg libavformat 55.43.100 through 57.2.100
34#define FFMPEG_LIBAVFORMAT_MAX_DUR2 FFMPEG_LIBAVFORMAT && ( \
35      (LIBAVFORMAT_VERSION_MAJOR == 55 && LIBAVFORMAT_VERSION_MINOR >= 43) \
36      || (LIBAVFORMAT_VERSION_MAJOR == 56) \
37      || (LIBAVFORMAT_VERSION_MAJOR == 57 && LIBAVFORMAT_VERSION_MINOR < 2) \
38      )
39
40// backward compatibility with libavcodec55
41#if LIBAVCODEC_VERSION_INT < AV_VERSION_INT(55,28,1)
42#warning "libavcodec55 is deprecated"
43#define HAVE_AUBIO_LIBAVCODEC_DEPRECATED 1
44#define av_frame_alloc  avcodec_alloc_frame
45#define av_frame_free avcodec_free_frame
46#define av_packet_unref av_free_packet
47#endif
48
49#include "aubio_priv.h"
50#include "fvec.h"
51#include "fmat.h"
52#include "source_avcodec.h"
53
54#define AUBIO_AVCODEC_MAX_BUFFER_SIZE FF_MIN_BUFFER_SIZE
55
56struct _aubio_source_avcodec_t {
57  uint_t hop_size;
58  uint_t samplerate;
59  uint_t channels;
60
61  // some data about the file
62  char_t *path;
63  uint_t input_samplerate;
64  uint_t input_channels;
65
66  // avcodec stuff
67  AVFormatContext *avFormatCtx;
68  AVCodecContext *avCodecCtx;
69  AVFrame *avFrame;
70  AVPacket avPacket;
71  AVAudioResampleContext *avr;
72  smpl_t *output;
73  uint_t read_samples;
74  uint_t read_index;
75  sint_t selected_stream;
76  uint_t eof;
77  uint_t multi;
78};
79
80// hack to create or re-create the context the first time _do or _do_multi is called
81void aubio_source_avcodec_reset_resampler(aubio_source_avcodec_t * s, uint_t multi);
82void aubio_source_avcodec_readframe(aubio_source_avcodec_t *s, uint_t * read_samples);
83
84uint_t aubio_source_avcodec_has_network_url(aubio_source_avcodec_t *s);
85
86uint_t aubio_source_avcodec_has_network_url(aubio_source_avcodec_t *s) {
87  char proto[20], authorization[256], hostname[128], uripath[256];
88  int proto_size = 20, authorization_size = 256, hostname_size = 128,
89      *port_ptr = 0, path_size = 256;
90  av_url_split(proto, proto_size, authorization, authorization_size, hostname,
91      hostname_size, port_ptr, uripath, path_size, s->path);
92  if (strlen(proto)) {
93    return 1;
94  }
95  return 0;
96}
97
98
99aubio_source_avcodec_t * new_aubio_source_avcodec(const char_t * path, uint_t samplerate, uint_t hop_size) {
100  aubio_source_avcodec_t * s = AUBIO_NEW(aubio_source_avcodec_t);
101  int err;
102  if (path == NULL) {
103    AUBIO_ERR("source_avcodec: Aborted opening null path\n");
104    goto beach;
105  }
106  if ((sint_t)samplerate < 0) {
107    AUBIO_ERR("source_avcodec: Can not open %s with samplerate %d\n", path, samplerate);
108    goto beach;
109  }
110  if ((sint_t)hop_size <= 0) {
111    AUBIO_ERR("source_avcodec: Can not open %s with hop_size %d\n", path, hop_size);
112    goto beach;
113  }
114
115  s->hop_size = hop_size;
116  s->channels = 1;
117
118  if (s->path) AUBIO_FREE(s->path);
119  s->path = AUBIO_ARRAY(char_t, strnlen(path, PATH_MAX) + 1);
120  strncpy(s->path, path, strnlen(path, PATH_MAX) + 1);
121
122  // register all formats and codecs
123  av_register_all();
124
125  if (aubio_source_avcodec_has_network_url(s)) {
126    avformat_network_init();
127  }
128
129  // try opening the file and get some info about it
130  AVFormatContext *avFormatCtx = s->avFormatCtx;
131  avFormatCtx = NULL;
132  if ( (err = avformat_open_input(&avFormatCtx, s->path, NULL, NULL) ) < 0 ) {
133    char errorstr[256];
134    av_strerror (err, errorstr, sizeof(errorstr));
135    AUBIO_ERR("source_avcodec: Failed opening %s (%s)\n", s->path, errorstr);
136    goto beach;
137  }
138
139  // try to make sure max_analyze_duration is big enough for most songs
140#if FFMPEG_LIBAVFORMAT_MAX_DUR2
141  avFormatCtx->max_analyze_duration2 *= 100;
142#else
143  avFormatCtx->max_analyze_duration *= 100;
144#endif
145
146  // retrieve stream information
147  if ( (err = avformat_find_stream_info(avFormatCtx, NULL)) < 0 ) {
148    char errorstr[256];
149    av_strerror (err, errorstr, sizeof(errorstr));
150    AUBIO_ERR("source_avcodec: Could not find stream information " "for %s (%s)\n", s->path,
151        errorstr);
152    goto beach;
153  }
154
155  // dump information about file onto standard error
156  //av_dump_format(avFormatCtx, 0, s->path, 0);
157
158  // look for the first audio stream
159  uint_t i;
160  sint_t selected_stream = -1;
161  for (i = 0; i < avFormatCtx->nb_streams; i++) {
162#if FF_API_LAVF_AVCTX
163    if (avFormatCtx->streams[i]->codecpar->codec_type == AVMEDIA_TYPE_AUDIO) {
164#else
165    if (avFormatCtx->streams[i]->codec->codec_type == AVMEDIA_TYPE_AUDIO) {
166#endif
167      if (selected_stream == -1) {
168        selected_stream = i;
169      } else {
170        AUBIO_WRN("source_avcodec: More than one audio stream in %s, "
171            "taking the first one\n", s->path);
172      }
173    }
174  }
175  if (selected_stream == -1) {
176    AUBIO_ERR("source_avcodec: No audio stream in %s\n", s->path);
177    goto beach;
178  }
179  //AUBIO_DBG("Taking stream %d in file %s\n", selected_stream, s->path);
180  s->selected_stream = selected_stream;
181
182  AVCodecContext *avCodecCtx = s->avCodecCtx;
183#if FF_API_LAVF_AVCTX
184  AVCodecParameters *codecpar = avFormatCtx->streams[selected_stream]->codecpar;
185  if (codecpar == NULL) {
186    AUBIO_ERR("source_avcodec: Could not find decoder for %s", s->path);
187    goto beach;
188  }
189  AVCodec *codec = avcodec_find_decoder(codecpar->codec_id);
190
191  /* Allocate a codec context for the decoder */
192  avCodecCtx = avcodec_alloc_context3(codec);
193  if (!avCodecCtx) {
194    AUBIO_ERR("source_avcodec: Failed to allocate the %s codec context for path %s\n",
195        av_get_media_type_string(AVMEDIA_TYPE_AUDIO), s->path);
196    goto beach;
197  }
198#else
199  avCodecCtx = avFormatCtx->streams[selected_stream]->codec;
200  AVCodec *codec = avcodec_find_decoder(avCodecCtx->codec_id);
201#endif
202  if (codec == NULL) {
203    AUBIO_ERR("source_avcodec: Could not find decoder for %s", s->path);
204    goto beach;
205  }
206
207#if FF_API_LAVF_AVCTX
208  /* Copy codec parameters from input stream to output codec context */
209  if ((err = avcodec_parameters_to_context(avCodecCtx, codecpar)) < 0) {
210    AUBIO_ERR("source_avcodec: Failed to copy %s codec parameters to decoder context for %s\n",
211       av_get_media_type_string(AVMEDIA_TYPE_AUDIO), s->path);
212    goto beach;
213  }
214#endif
215
216  if ( ( err = avcodec_open2(avCodecCtx, codec, NULL) ) < 0) {
217    char errorstr[256];
218    av_strerror (err, errorstr, sizeof(errorstr));
219    AUBIO_ERR("source_avcodec: Could not load codec for %s (%s)\n", s->path, errorstr);
220    goto beach;
221  }
222
223  /* get input specs */
224  s->input_samplerate = avCodecCtx->sample_rate;
225  s->input_channels   = avCodecCtx->channels;
226  //AUBIO_DBG("input_samplerate: %d\n", s->input_samplerate);
227  //AUBIO_DBG("input_channels: %d\n", s->input_channels);
228
229  if (samplerate == 0) {
230    s->samplerate = s->input_samplerate;
231  } else {
232    s->samplerate = samplerate;
233  }
234
235  if (s->samplerate >  s->input_samplerate) {
236    AUBIO_WRN("source_avcodec: upsampling %s from %d to %d\n", s->path,
237        s->input_samplerate, s->samplerate);
238  }
239
240  AVFrame *avFrame = s->avFrame;
241  avFrame = av_frame_alloc();
242  if (!avFrame) {
243    AUBIO_ERR("source_avcodec: Could not allocate frame for (%s)\n", s->path);
244  }
245
246  /* allocate output for avr */
247  s->output = (smpl_t *)av_malloc(AUBIO_AVCODEC_MAX_BUFFER_SIZE * sizeof(smpl_t));
248
249  s->read_samples = 0;
250  s->read_index = 0;
251
252  s->avFormatCtx = avFormatCtx;
253  s->avCodecCtx = avCodecCtx;
254  s->avFrame = avFrame;
255
256  // default to mono output
257  aubio_source_avcodec_reset_resampler(s, 0);
258
259  s->eof = 0;
260  s->multi = 0;
261
262  //av_log_set_level(AV_LOG_QUIET);
263
264  return s;
265
266beach:
267  //AUBIO_ERR("can not read %s at samplerate %dHz with a hop_size of %d\n",
268  //    s->path, s->samplerate, s->hop_size);
269  del_aubio_source_avcodec(s);
270  return NULL;
271}
272
273void aubio_source_avcodec_reset_resampler(aubio_source_avcodec_t * s, uint_t multi) {
274  // create or reset resampler to/from mono/multi-channel
275  if ( (multi != s->multi) || (s->avr == NULL) ) {
276    int64_t input_layout = av_get_default_channel_layout(s->input_channels);
277    uint_t output_channels = multi ? s->input_channels : 1;
278    int64_t output_layout = av_get_default_channel_layout(output_channels);
279    AVAudioResampleContext *avr = avresample_alloc_context();
280    AVAudioResampleContext *oldavr = s->avr;
281
282    av_opt_set_int(avr, "in_channel_layout",  input_layout,           0);
283    av_opt_set_int(avr, "out_channel_layout", output_layout,          0);
284    av_opt_set_int(avr, "in_sample_rate",     s->input_samplerate,    0);
285    av_opt_set_int(avr, "out_sample_rate",    s->samplerate,          0);
286    av_opt_set_int(avr, "in_sample_fmt",      s->avCodecCtx->sample_fmt, 0);
287#if HAVE_AUBIO_DOUBLE
288    av_opt_set_int(avr, "out_sample_fmt",     AV_SAMPLE_FMT_DBL,      0);
289#else
290    av_opt_set_int(avr, "out_sample_fmt",     AV_SAMPLE_FMT_FLT,      0);
291#endif
292    // TODO: use planar?
293    //av_opt_set_int(avr, "out_sample_fmt",     AV_SAMPLE_FMT_FLTP,      0);
294    int err;
295    if ( ( err = avresample_open(avr) ) < 0) {
296      char errorstr[256];
297      av_strerror (err, errorstr, sizeof(errorstr));
298      AUBIO_ERR("source_avcodec: Could not open AVAudioResampleContext for %s (%s)\n",
299          s->path, errorstr);
300      //goto beach;
301      return;
302    }
303    s->avr = avr;
304    if (oldavr != NULL) {
305      avresample_close( oldavr );
306      av_free ( oldavr );
307      oldavr = NULL;
308    }
309    s->multi = multi;
310  }
311}
312
313void aubio_source_avcodec_readframe(aubio_source_avcodec_t *s, uint_t * read_samples) {
314  AVFormatContext *avFormatCtx = s->avFormatCtx;
315  AVCodecContext *avCodecCtx = s->avCodecCtx;
316  AVFrame *avFrame = s->avFrame;
317  AVPacket avPacket = s->avPacket;
318  av_init_packet (&avPacket);
319  AVAudioResampleContext *avr = s->avr;
320  smpl_t *output = s->output;
321  *read_samples = 0;
322
323  do
324  {
325    int err = av_read_frame (avFormatCtx, &avPacket);
326    if (err == AVERROR_EOF) {
327      s->eof = 1;
328      goto beach;
329    }
330    if (err != 0) {
331      char errorstr[256];
332      av_strerror (err, errorstr, sizeof(errorstr));
333      AUBIO_ERR("source_avcodec: could not read frame in %s (%s)\n", s->path, errorstr);
334      goto beach;
335    }
336  } while (avPacket.stream_index != s->selected_stream);
337
338  int got_frame = 0;
339#if FF_API_LAVF_AVCTX
340  int ret = avcodec_send_packet(avCodecCtx, &avPacket);
341  if (ret < 0 && ret != AVERROR_EOF) {
342    AUBIO_ERR("source_avcodec: error when sending packet for %s\n", s->path);
343    goto beach;
344  }
345  ret = avcodec_receive_frame(avCodecCtx, avFrame);
346  if (ret >= 0) {
347    got_frame = 1;
348  }
349  if (ret < 0) {
350    if (ret == AVERROR(EAGAIN)) {
351      AUBIO_WRN("source_avcodec: output is not available right now - user must try to send new input\n");
352    } else if (ret == AVERROR_EOF) {
353      AUBIO_WRN("source_avcodec: the decoder has been fully flushed, and there will be no more output frames\n");
354    } else {
355      AUBIO_ERR("source_avcodec: decoding errors on %s\n", s->path);
356      goto beach;
357    }
358  }
359#else
360  int len = avcodec_decode_audio4(avCodecCtx, avFrame, &got_frame, &avPacket);
361
362  if (len < 0) {
363    AUBIO_ERR("source_avcodec: error while decoding %s\n", s->path);
364    goto beach;
365  }
366#endif
367  if (got_frame == 0) {
368    AUBIO_WRN("source_avcodec: did not get a frame when reading %s\n", s->path);
369    goto beach;
370  }
371
372  int in_linesize = 0;
373  av_samples_get_buffer_size(&in_linesize, avCodecCtx->channels,
374      avFrame->nb_samples, avCodecCtx->sample_fmt, 1);
375  int in_samples = avFrame->nb_samples;
376  int out_linesize = 0;
377  int max_out_samples = AUBIO_AVCODEC_MAX_BUFFER_SIZE;
378  int out_samples = avresample_convert ( avr,
379        (uint8_t **)&output, out_linesize, max_out_samples,
380        (uint8_t **)avFrame->data, in_linesize, in_samples);
381  if (out_samples <= 0) {
382    AUBIO_WRN("source_avcodec: no sample found while converting frame (%s)\n", s->path);
383    goto beach;
384  }
385
386  *read_samples = out_samples;
387
388beach:
389  s->avFormatCtx = avFormatCtx;
390  s->avCodecCtx = avCodecCtx;
391  s->avFrame = avFrame;
392  s->avr = avr;
393  s->output = output;
394
395  av_packet_unref(&avPacket);
396}
397
398void aubio_source_avcodec_do(aubio_source_avcodec_t * s, fvec_t * read_data, uint_t * read){
399  uint_t i;
400  uint_t end = 0;
401  uint_t total_wrote = 0;
402  // switch from multi
403  if (s->multi == 1) aubio_source_avcodec_reset_resampler(s, 0);
404  while (total_wrote < s->hop_size) {
405    end = MIN(s->read_samples - s->read_index, s->hop_size - total_wrote);
406    for (i = 0; i < end; i++) {
407      read_data->data[i + total_wrote] = s->output[i + s->read_index];
408    }
409    total_wrote += end;
410    if (total_wrote < s->hop_size) {
411      uint_t avcodec_read = 0;
412      aubio_source_avcodec_readframe(s, &avcodec_read);
413      s->read_samples = avcodec_read;
414      s->read_index = 0;
415      if (s->eof) {
416        break;
417      }
418    } else {
419      s->read_index += end;
420    }
421  }
422  if (total_wrote < s->hop_size) {
423    for (i = total_wrote; i < s->hop_size; i++) {
424      read_data->data[i] = 0.;
425    }
426  }
427  *read = total_wrote;
428}
429
430void aubio_source_avcodec_do_multi(aubio_source_avcodec_t * s, fmat_t * read_data, uint_t * read){
431  uint_t i,j;
432  uint_t end = 0;
433  uint_t total_wrote = 0;
434  // switch from mono
435  if (s->multi == 0) aubio_source_avcodec_reset_resampler(s, 1);
436  while (total_wrote < s->hop_size) {
437    end = MIN(s->read_samples - s->read_index, s->hop_size - total_wrote);
438    for (j = 0; j < read_data->height; j++) {
439      for (i = 0; i < end; i++) {
440        read_data->data[j][i + total_wrote] =
441          s->output[(i + s->read_index) * s->input_channels + j];
442      }
443    }
444    total_wrote += end;
445    if (total_wrote < s->hop_size) {
446      uint_t avcodec_read = 0;
447      aubio_source_avcodec_readframe(s, &avcodec_read);
448      s->read_samples = avcodec_read;
449      s->read_index = 0;
450      if (s->eof) {
451        break;
452      }
453    } else {
454      s->read_index += end;
455    }
456  }
457  if (total_wrote < s->hop_size) {
458    for (j = 0; j < read_data->height; j++) {
459      for (i = total_wrote; i < s->hop_size; i++) {
460        read_data->data[j][i] = 0.;
461      }
462    }
463  }
464  *read = total_wrote;
465}
466
467uint_t aubio_source_avcodec_get_samplerate(const aubio_source_avcodec_t * s) {
468  return s->samplerate;
469}
470
471uint_t aubio_source_avcodec_get_channels(const aubio_source_avcodec_t * s) {
472  return s->input_channels;
473}
474
475uint_t aubio_source_avcodec_seek (aubio_source_avcodec_t * s, uint_t pos) {
476  int64_t resampled_pos = (uint_t)ROUND(pos * (s->input_samplerate * 1. / s->samplerate));
477  int64_t min_ts = MAX(resampled_pos - 2000, 0);
478  int64_t max_ts = MIN(resampled_pos + 2000, INT64_MAX);
479  int seek_flags = AVSEEK_FLAG_FRAME | AVSEEK_FLAG_ANY;
480  int ret = AUBIO_FAIL;
481  if (s->avFormatCtx != NULL && s->avr != NULL) {
482    ret = AUBIO_OK;
483  } else {
484    AUBIO_ERR("source_avcodec: failed seeking in %s (file not opened?)", s->path);
485    return ret;
486  }
487  if ((sint_t)pos < 0) {
488    AUBIO_ERR("source_avcodec: could not seek %s at %d (seeking position"
489       " should be >= 0)\n", s->path, pos);
490    return AUBIO_FAIL;
491  }
492  ret = avformat_seek_file(s->avFormatCtx, s->selected_stream,
493      min_ts, resampled_pos, max_ts, seek_flags);
494  if (ret < 0) {
495    AUBIO_ERR("source_avcodec: failed seeking to %d in file %s", pos, s->path);
496  }
497  // reset read status
498  s->eof = 0;
499  s->read_index = 0;
500  s->read_samples = 0;
501  // reset the AVAudioResampleContext
502  avresample_close(s->avr);
503  avresample_open(s->avr);
504  return ret;
505}
506
507uint_t aubio_source_avcodec_get_duration (aubio_source_avcodec_t * s) {
508  if (s && &(s->avFormatCtx) != NULL) {
509    int64_t duration = s->avFormatCtx->duration;
510    return s->samplerate * ((uint_t)duration / 1e6 );
511  }
512  return 0;
513}
514
515uint_t aubio_source_avcodec_close(aubio_source_avcodec_t * s) {
516  if (s->avr != NULL) {
517    avresample_close( s->avr );
518    av_free ( s->avr );
519  }
520  s->avr = NULL;
521  if (s->avCodecCtx != NULL) {
522    avcodec_close ( s->avCodecCtx );
523  }
524  s->avCodecCtx = NULL;
525  if (s->avFormatCtx != NULL) {
526    avformat_close_input(&s->avFormatCtx);
527#ifndef HAVE_AUBIO_LIBAVCODEC_DEPRECATED // avoid crash on old libavcodec54
528    avformat_free_context(s->avFormatCtx);
529#endif
530    s->avFormatCtx = NULL;
531  }
532  av_packet_unref(&s->avPacket);
533  return AUBIO_OK;
534}
535
536void del_aubio_source_avcodec(aubio_source_avcodec_t * s){
537  if (!s) return;
538  aubio_source_avcodec_close(s);
539  if (s->output != NULL) {
540    av_free(s->output);
541  }
542  s->output = NULL;
543  if (s->avFrame != NULL) {
544    av_frame_free( &(s->avFrame) );
545  }
546  s->avFrame = NULL;
547  if (s->path) {
548    AUBIO_FREE(s->path);
549  }
550  s->path = NULL;
551  AUBIO_FREE(s);
552}
553
554#endif /* HAVE_LIBAV */
Note: See TracBrowser for help on using the repository browser.