view libmpdemux/demux_audio.c @ 16946:47c5e9846cd3

ultra simple&slow pp filter, yes yet another spp like filter :) this one does actually compress&decompress the video at various shifts with lavc while the other spp filters are doing optimized intra only filtering limitations: mpeg4 is hardcoded, all options too, pretty trivial to change though, even filtering with non dct codecs like snow could be tried ... the qscale/qp is only taken fron the first MB of each image and then used for the whole image (would needs some small changes to lavc to let the user set the qscales for the mbs themselfs but iam to lazy ...) this needs ALOT of cpu time and memory especially at uspp=8 ...

#include "config.h"
#include "../mp_msg.h"

#include <stdlib.h>
#include <stdio.h>
#include "stream.h"
#include "demuxer.h"
#include "stheader.h"
#include "genres.h"
#include "mp3_hdr.h"

#include <string.h>
#ifdef MP_DEBUG
#include <assert.h>
#endif

#define MP3 1
#define WAV 2
#define fLaC 3


#define HDR_SIZE 4

typedef struct da_priv {
  int frmt;
  float last_pts;
} da_priv_t;

// how many valid frames in a row we need before accepting as valid MP3
#define MIN_MP3_HDRS 5

//! Used to describe a potential (chain of) MP3 headers we found
typedef struct mp3_hdr {
  off_t frame_pos; // start of first frame in this "chain" of headers
  off_t next_frame_pos; // here we expect the next header with same parameters
  int mp3_chans;
  int mp3_freq;
  int mpa_spf;
  int mpa_layer;
  int mpa_br;
  int cons_hdrs; // if this reaches MIN_MP3_HDRS we accept as MP3 file
  struct mp3_hdr *next;
} mp3_hdr_t;

extern void free_sh_audio(sh_audio_t* sh);
extern void print_wave_header(WAVEFORMATEX *h);

int hr_mp3_seek = 0;

/**
 * \brief free a list of MP3 header descriptions
 * \param list pointer to the head-of-list pointer
 */
static void free_mp3_hdrs(mp3_hdr_t **list) {
  mp3_hdr_t *tmp;
  while (*list) {
    tmp = (*list)->next;
    free(*list);
    *list = tmp;
  }
}

/**
 * \brief add another potential MP3 header to our list
 * If it fits into an existing chain this one is expanded otherwise
 * a new one is created.
 * All entries that expected a MP3 header before the current position
 * are discarded.
 * The list is expected to be and will be kept sorted by next_frame_pos
 * and when those are equal by frame_pos.
 * \param list pointer to the head-of-list pointer
 * \param st_pos stream position where the described header starts
 * \param mp3_chans number of channels as specified by the header (*)
 * \param mp3_freq sampling frequency as specified by the header (*)
 * \param mpa_spf frame size as specified by the header
 * \param mpa_layer layer type ("version") as specified by the header (*)
 * \param mpa_br bitrate as specified by the header
 * \param mp3_flen length of the frame as specified by the header
 * \return If non-null the current file is accepted as MP3 and the
 * mp3_hdr struct describing the valid chain is returned. Must be
 * freed independent of the list.
 *
 * parameters marked by (*) must be the same for all headers in the same chain
 */
static mp3_hdr_t *add_mp3_hdr(mp3_hdr_t **list, off_t st_pos,
                               int mp3_chans, int mp3_freq, int mpa_spf,
                               int mpa_layer, int mpa_br, int mp3_flen) {
  mp3_hdr_t *tmp;
  int in_list = 0;
  while (*list && (*list)->next_frame_pos <= st_pos) {
    if (((*list)->next_frame_pos < st_pos) || ((*list)->mp3_chans != mp3_chans)
         || ((*list)->mp3_freq != mp3_freq) || ((*list)->mpa_layer != mpa_layer) ) {
      // wasn't valid!
      tmp = (*list)->next;
      free(*list);
      *list = tmp;
    } else {
      (*list)->cons_hdrs++;
      (*list)->next_frame_pos = st_pos + mp3_flen;
      (*list)->mpa_spf = mpa_spf;
      (*list)->mpa_br = mpa_br;
      if ((*list)->cons_hdrs >= MIN_MP3_HDRS) {
        // copy the valid entry, so that the list can be easily freed
        tmp = malloc(sizeof(mp3_hdr_t));
        memcpy(tmp, *list, sizeof(mp3_hdr_t));
        tmp->next = NULL;
        return tmp;
      }
      in_list = 1;
      list = &((*list)->next);
    }
  }
  if (!in_list) { // does not belong into an existing chain, insert
    // find right position to insert to keep sorting
    while (*list && (*list)->next_frame_pos <= st_pos + mp3_flen)
      list = &((*list)->next);
    tmp = malloc(sizeof(mp3_hdr_t));
    tmp->frame_pos = st_pos;
    tmp->next_frame_pos = st_pos + mp3_flen;
    tmp->mp3_chans = mp3_chans;
    tmp->mp3_freq = mp3_freq;
    tmp->mpa_spf = mpa_spf;
    tmp->mpa_layer = mpa_layer;
    tmp->mpa_br = mpa_br;
    tmp->cons_hdrs = 1;
    tmp->next = *list;
    *list = tmp;
  }
  return NULL;
}

static int demux_audio_open(demuxer_t* demuxer) {
  stream_t *s;
  sh_audio_t* sh_audio;
  uint8_t hdr[HDR_SIZE];
  int frmt = 0, n = 0, step;
  off_t st_pos = 0, next_frame_pos = 0;
  // mp3_hdrs list is sorted first by next_frame_pos and then by frame_pos
  mp3_hdr_t *mp3_hdrs = NULL, *mp3_found = NULL;
  da_priv_t* priv;
#ifdef MP_DEBUG
  assert(demuxer != NULL);
  assert(demuxer->stream != NULL);
#endif
  
  s = demuxer->stream;

  stream_read(s, hdr, HDR_SIZE);
  while(n < 30000 && !s->eof) {
    int mp3_freq, mp3_chans, mp3_flen, mpa_layer, mpa_spf, mpa_br;
    st_pos = stream_tell(s) - HDR_SIZE;
    step = 1;

    if( hdr[0] == 'R' && hdr[1] == 'I' && hdr[2] == 'F' && hdr[3] == 'F' ) {
      stream_skip(s,4);
      if(s->eof)
	break;
      stream_read(s,hdr,4);
      if(s->eof)
	break;
      if(hdr[0] != 'W' || hdr[1] != 'A' || hdr[2] != 'V'  || hdr[3] != 'E' )
	stream_skip(s,-8);
      else
      // We found wav header. Now we can have 'fmt ' or a mp3 header
      // empty the buffer
	step = 4;
    } else if( hdr[0] == 'I' && hdr[1] == 'D' && hdr[2] == '3' && (hdr[3] >= 2)) {
      int len;
      stream_skip(s,2);
      stream_read(s,hdr,4);
      len = (hdr[0]<<21) | (hdr[1]<<14) | (hdr[2]<<7) | hdr[3];
      stream_skip(s,len);
      step = 4;
    } else if( hdr[0] == 'f' && hdr[1] == 'm' && hdr[2] == 't' && hdr[3] == ' ' ) {
      frmt = WAV;
      break;      
    } else if((mp3_flen = mp_get_mp3_header(hdr, &mp3_chans, &mp3_freq,
                                &mpa_spf, &mpa_layer, &mpa_br)) > 0) {
      mp3_found = add_mp3_hdr(&mp3_hdrs, st_pos, mp3_chans, mp3_freq,
                              mpa_spf, mpa_layer, mpa_br, mp3_flen);
      if (mp3_found) {
        frmt = MP3;
        break;
      }
    } else if( hdr[0] == 'f' && hdr[1] == 'L' && hdr[2] == 'a' && hdr[3] == 'C' ) {
      frmt = fLaC;
      stream_skip(s,-4);
      break;
    }
    // Add here some other audio format detection
    if(step < HDR_SIZE)
      memmove(hdr,&hdr[step],HDR_SIZE-step);
    stream_read(s, &hdr[HDR_SIZE - step], step);
    n++;
  }

  free_mp3_hdrs(&mp3_hdrs);

  if(!frmt)
    return 0;

  sh_audio = new_sh_audio(demuxer,0);

  switch(frmt) {
  case MP3:
    sh_audio->format = (mp3_found->mpa_layer < 3 ? 0x50 : 0x55);
    demuxer->movi_start = mp3_found->frame_pos;
    next_frame_pos = mp3_found->next_frame_pos;
    sh_audio->audio.dwSampleSize= 0;
    sh_audio->audio.dwScale = mp3_found->mpa_spf;
    sh_audio->audio.dwRate = mp3_found->mp3_freq;
    sh_audio->wf = malloc(sizeof(WAVEFORMATEX));
    sh_audio->wf->wFormatTag = sh_audio->format;
    sh_audio->wf->nChannels = mp3_found->mp3_chans;
    sh_audio->wf->nSamplesPerSec = mp3_found->mp3_freq;
    sh_audio->wf->nAvgBytesPerSec = mp3_found->mpa_br * (1000 / 8);
    sh_audio->wf->nBlockAlign = mp3_found->mpa_spf;
    sh_audio->wf->wBitsPerSample = 16;
    sh_audio->wf->cbSize = 0;    
    sh_audio->i_bps = sh_audio->wf->nAvgBytesPerSec;
    free(mp3_found);
    mp3_found = NULL;
    if(s->end_pos) {
      char tag[4];
      stream_seek(s,s->end_pos-128);
      stream_read(s,tag,3);
      tag[3] = '\0';
      if(strcmp(tag,"TAG"))
	demuxer->movi_end = s->end_pos;
      else {
	char buf[31];
	uint8_t g;
	demuxer->movi_end = stream_tell(s)-3;
	stream_read(s,buf,30);
	buf[30] = '\0';
	demux_info_add(demuxer,"Title",buf);
	stream_read(s,buf,30);
	buf[30] = '\0';
	demux_info_add(demuxer,"Artist",buf);
	stream_read(s,buf,30);
	buf[30] = '\0';
	demux_info_add(demuxer,"Album",buf);
	stream_read(s,buf,4);
	buf[4] = '\0';
	demux_info_add(demuxer,"Year",buf);
	stream_read(s,buf,30);
	buf[30] = '\0';
	demux_info_add(demuxer,"Comment",buf);
	if(buf[28] == 0 && buf[29] != 0) {
	  uint8_t trk = (uint8_t)buf[29];
	  sprintf(buf,"%d",trk);
	  demux_info_add(demuxer,"Track",buf);
	}
	g = stream_read_char(s);
	demux_info_add(demuxer,"Genre",genres[g]);
      }
    }
    break;
  case WAV: {
    unsigned int chunk_type;
    unsigned int chunk_size;
    WAVEFORMATEX* w;
    int l;
    l = stream_read_dword_le(s);
    if(l < 16) {
      mp_msg(MSGT_DEMUX,MSGL_ERR,"[demux_audio] Bad wav header length: too short (%d)!!!\n",l);
      free_sh_audio(sh_audio);
      return 0;
    }
    sh_audio->wf = w = (WAVEFORMATEX*)malloc(l > sizeof(WAVEFORMATEX) ? l : sizeof(WAVEFORMATEX));
    w->wFormatTag = sh_audio->format = stream_read_word_le(s);
    w->nChannels = sh_audio->channels = stream_read_word_le(s);
    w->nSamplesPerSec = sh_audio->samplerate = stream_read_dword_le(s);
    w->nAvgBytesPerSec = stream_read_dword_le(s);
    w->nBlockAlign = stream_read_word_le(s);
    w->wBitsPerSample = sh_audio->samplesize = stream_read_word_le(s);
    w->cbSize = 0;
    sh_audio->i_bps = sh_audio->wf->nAvgBytesPerSec;
    l -= 16;
    if (l > 0) {
    w->cbSize = stream_read_word_le(s);
    l -= 2;
     if (w->cbSize > 0) {
      if (l < w->cbSize) {
        mp_msg(MSGT_DEMUX,MSGL_ERR,"[demux_audio] truncated extradata (%d < %d)\n",
	l,w->cbSize);
        stream_read(s,(char*)((char*)(w)+sizeof(WAVEFORMATEX)),l);
        l = 0;
      } else {
        stream_read(s,(char*)((char*)(w)+sizeof(WAVEFORMATEX)),w->cbSize);
        l -= w->cbSize;
      }
     }
    }

    if(verbose>0) print_wave_header(w);
    if(l)
      stream_skip(s,l);
    do
    {
      chunk_type = stream_read_fourcc(demuxer->stream);
      chunk_size = stream_read_dword_le(demuxer->stream);
      if (chunk_type != mmioFOURCC('d', 'a', 't', 'a'))
        stream_skip(demuxer->stream, chunk_size);
    } while (chunk_type != mmioFOURCC('d', 'a', 't', 'a'));
    demuxer->movi_start = stream_tell(s);
    demuxer->movi_end = s->end_pos;
//    printf("wav: %X .. %X\n",(int)demuxer->movi_start,(int)demuxer->movi_end);
    // Check if it contains dts audio
    if((w->wFormatTag == 0x01) && (w->nChannels == 2) && (w->nSamplesPerSec == 44100)) {
	unsigned char buf[16384]; // vlc uses 16384*4 (4 dts frames)
	unsigned int i;
	stream_read(s, buf, sizeof(buf));
	for (i = 0; i < sizeof(buf); i += 2) {
	    // DTS, 14 bit, LE
	    if((buf[i] == 0xff) && (buf[i+1] == 0x1f) && (buf[i+2] == 0x00) &&
	       (buf[i+3] == 0xe8) && ((buf[i+4] & 0xfe) == 0xf0) && (buf[i+5] == 0x07)) {
		sh_audio->format = 0x2001;
		mp_msg(MSGT_DEMUX,MSGL_V,"[demux_audio] DTS audio in wav, 14 bit, LE\n");
		break;
	    }
	    // DTS, 14 bit, BE
	    if((buf[i] == 0x1f) && (buf[i+1] == 0xff) && (buf[i+2] == 0xe8) &&
	       (buf[i+3] == 0x00) && (buf[i+4] == 0x07) && ((buf[i+5] & 0xfe) == 0xf0)) {
		sh_audio->format = 0x2001;
		mp_msg(MSGT_DEMUX,MSGL_V,"[demux_audio] DTS audio in wav, 14 bit, BE\n");
		break;
	    }
	    // DTS, 16 bit, BE
	    if((buf[i] == 0x7f) && (buf[i+1] == 0xfe) && (buf[i+2] == 0x80) &&
	       (buf[i+3] == 0x01)) {
		sh_audio->format = 0x2001;
		mp_msg(MSGT_DEMUX,MSGL_V,"[demux_audio] DTS audio in wav, 16 bit, BE\n");
		break;
	    }
	    // DTS, 16 bit, LE
	    if((buf[i] == 0xfe) && (buf[i+1] == 0x7f) && (buf[i+2] == 0x01) &&
	       (buf[i+3] == 0x80)) {
		sh_audio->format = 0x2001;
		mp_msg(MSGT_DEMUX,MSGL_V,"[demux_audio] DTS audio in wav, 16 bit, LE\n");
		break;
	    }
	}
	if (sh_audio->format == 0x2001)
	    mp_msg(MSGT_DEMUX,MSGL_DBG2,"[demux_audio] DTS sync offset = %u\n", i);

    }
    stream_seek(s,demuxer->movi_start);
  } break;
  case fLaC:
	    sh_audio->format = mmioFOURCC('f', 'L', 'a', 'C');
	    demuxer->movi_start = stream_tell(s);
	    demuxer->movi_end = s->end_pos;
	    break;
  }

  priv = (da_priv_t*)malloc(sizeof(da_priv_t));
  priv->frmt = frmt;
  priv->last_pts = -1;
  demuxer->priv = priv;
  demuxer->audio->id = 0;
  demuxer->audio->sh = sh_audio;
  sh_audio->ds = demuxer->audio;
  sh_audio->samplerate = sh_audio->audio.dwRate;

  if(stream_tell(s) != demuxer->movi_start)
  {
    mp_msg(MSGT_DEMUX, MSGL_V, "demux_audio: seeking from 0x%X to start pos 0x%X\n",
            (int)stream_tell(s), (int)demuxer->movi_start);
    stream_seek(s,demuxer->movi_start);
    if (stream_tell(s) != demuxer->movi_start) {
      mp_msg(MSGT_DEMUX, MSGL_V, "demux_audio: seeking failed, now at 0x%X!\n",
              (int)stream_tell(s));
      if (next_frame_pos) {
        mp_msg(MSGT_DEMUX, MSGL_V, "demux_audio: seeking to 0x%X instead\n",
                (int)next_frame_pos);
        stream_seek(s, next_frame_pos);
      }
    }
  }

  mp_msg(MSGT_DEMUX,MSGL_V,"demux_audio: audio data 0x%X - 0x%X  \n",(int)demuxer->movi_start,(int)demuxer->movi_end);

  return DEMUXER_TYPE_AUDIO;
}


static int demux_audio_fill_buffer(demuxer_t *demuxer, demux_stream_t *ds) {
  int l;
  demux_packet_t* dp;
  sh_audio_t* sh_audio;
  demuxer_t* demux;
  da_priv_t* priv;
  stream_t* s;
#ifdef MP_DEBUG
  assert(ds != NULL);
  assert(ds->sh != NULL);
  assert(ds->demuxer != NULL);
#endif
  sh_audio = ds->sh;
  demux = ds->demuxer;
  priv = demux->priv;
  s = demux->stream;

  if(s->eof)
    return 0;

  switch(priv->frmt) {
  case MP3 :
    while(1) {
      uint8_t hdr[4];
      stream_read(s,hdr,4);
      if (s->eof)
        return 0;
      l = mp_decode_mp3_header(hdr);
      if(l < 0) {
	if (demux->movi_end && stream_tell(s) >= demux->movi_end)
	  return 0; // might be ID3 tag, i.e. EOF
	stream_skip(s,-3);
      } else {
	dp = new_demux_packet(l);
	memcpy(dp->buffer,hdr,4);
	if (stream_read(s,dp->buffer + 4,l-4) != l-4)
	  return 0;
	priv->last_pts = priv->last_pts < 0 ? 0 : priv->last_pts + sh_audio->audio.dwScale/(float)sh_audio->samplerate;
	break;
      }
    } break;
  case WAV : {
    l = sh_audio->wf->nAvgBytesPerSec;
    dp = new_demux_packet(l);
    l = stream_read(s,dp->buffer,l);
    priv->last_pts = priv->last_pts < 0 ? 0 : priv->last_pts + l/(float)sh_audio->i_bps;
    break;
  }
  case fLaC: {
    l = 65535;
    dp = new_demux_packet(l);
    l = stream_read(s,dp->buffer,l);
    priv->last_pts = priv->last_pts < 0 ? 0 : priv->last_pts + l/(float)sh_audio->i_bps;
    break;
  }
  default:
    printf("Audio demuxer : unknown format %d\n",priv->frmt);
    return 0;
  }

  resize_demux_packet(dp, l);
  ds->pts = priv->last_pts - (ds_tell_pts(demux->audio) -
              sh_audio->a_in_buffer_len)/(float)sh_audio->i_bps;
  ds_add_packet(ds, dp);
  return 1;
}

static void high_res_mp3_seek(demuxer_t *demuxer,float time) {
  uint8_t hdr[4];
  int len,nf;
  da_priv_t* priv = demuxer->priv;
  sh_audio_t* sh = (sh_audio_t*)demuxer->audio->sh;

  nf = time*sh->samplerate/sh->audio.dwScale;
  while(nf > 0) {
    stream_read(demuxer->stream,hdr,4);
    len = mp_decode_mp3_header(hdr);
    if(len < 0) {
      stream_skip(demuxer->stream,-3);
      continue;
    }
    stream_skip(demuxer->stream,len-4);
    priv->last_pts += sh->audio.dwScale/(float)sh->samplerate;
    nf--;
  }
}

static void demux_audio_seek(demuxer_t *demuxer,float rel_seek_secs,int flags){
  sh_audio_t* sh_audio;
  stream_t* s;
  int base,pos;
  float len;
  da_priv_t* priv;

  if(!(sh_audio = demuxer->audio->sh))
    return;
  s = demuxer->stream;
  priv = demuxer->priv;

  if(priv->frmt == MP3 && hr_mp3_seek && !(flags & 2)) {
    len = (flags & 1) ? rel_seek_secs - priv->last_pts : rel_seek_secs;
    if(len < 0) {
      stream_seek(s,demuxer->movi_start);
      len = priv->last_pts + len;
      priv->last_pts = 0;
    }
    if(len > 0)
      high_res_mp3_seek(demuxer,len);
    sh_audio->delay = priv->last_pts -  (ds_tell_pts(demuxer->audio)-sh_audio->a_in_buffer_len)/(float)sh_audio->i_bps;
    return;
  }

  base = flags&1 ? demuxer->movi_start : stream_tell(s);
  if(flags&2)
    pos = base + ((demuxer->movi_end - demuxer->movi_start)*rel_seek_secs);
  else
    pos = base + (rel_seek_secs*sh_audio->i_bps);

  if(demuxer->movi_end && pos >= demuxer->movi_end) {
     pos = demuxer->movi_end;
    //sh_audio->delay = (stream_tell(s) - demuxer->movi_start)/(float)sh_audio->i_bps;
    //return;
  } else if(pos < demuxer->movi_start)
    pos = demuxer->movi_start;

  priv->last_pts = (pos-demuxer->movi_start)/(float)sh_audio->i_bps;
  sh_audio->delay = priv->last_pts - (ds_tell_pts(demuxer->audio)-sh_audio->a_in_buffer_len)/(float)sh_audio->i_bps;
  
  switch(priv->frmt) {
  case WAV:
    pos -= (pos - demuxer->movi_start) %
            (sh_audio->wf->nBlockAlign ? sh_audio->wf->nBlockAlign :
             (sh_audio->channels * sh_audio->samplesize));
    // We need to decrease the pts by one step to make it the "last one"
    priv->last_pts -= sh_audio->wf->nAvgBytesPerSec/(float)sh_audio->i_bps;
    break;
  }

  stream_seek(s,pos);
}

static void demux_close_audio(demuxer_t* demuxer) {
  da_priv_t* priv = demuxer->priv;

  if(!priv)
    return;
  free(priv);
}

static int demux_audio_control(demuxer_t *demuxer,int cmd, void *arg){
    sh_audio_t *sh_audio=demuxer->audio->sh;
    int audio_length = demuxer->movi_end / sh_audio->i_bps;
    da_priv_t* priv = demuxer->priv;
	    
    switch(cmd) {
	case DEMUXER_CTRL_GET_TIME_LENGTH:
	    if (audio_length<=0) return DEMUXER_CTRL_DONTKNOW;
	    *((double *)arg)=(double)audio_length;
	    return DEMUXER_CTRL_GUESS;

	case DEMUXER_CTRL_GET_PERCENT_POS:
	    if (audio_length<=0) 
    		return DEMUXER_CTRL_DONTKNOW;
    	    *((int *)arg)=(int)( (priv->last_pts*100)  / audio_length);
	    return DEMUXER_CTRL_OK;

	default:
	    return DEMUXER_CTRL_NOTIMPL;
    }
}


demuxer_desc_t demuxer_desc_audio = {
  "Audio demuxer",
  "audio",
  "Audio file",
  "?",
  "Audio only files",
  DEMUXER_TYPE_AUDIO,
  0, // unsafe autodetect
  demux_audio_open,
  demux_audio_fill_buffer,
  NULL,
  demux_close_audio,
  demux_audio_seek,
  demux_audio_control
};