k3b/libk3b/plugin/k3baudiodecoder.cpp

/* 
 *
 * $Id: k3baudiodecoder.cpp 619556 2007-01-03 17:38:12Z trueg $
 * Copyright (C) 2003 Sebastian Trueg <trueg@k3b.org>
 *
 * This file is part of the K3b project.
 * Copyright (C) 1998-2007 Sebastian Trueg <trueg@k3b.org>
 *
 * This program 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 2 of the License, or
 * (at your option) any later version.
 * See the file "COPYING" for the exact licensing terms.
 */

#include <config.h>

#include <k3bcore.h>

#include "k3baudiodecoder.h"
#include "k3bpluginmanager.h"

#include <kdebug.h>
#include <kfilemetainfo.h>

#include <qmap.h>

#include <math.h>

#ifdef HAVE_LIBSAMPLERATE
#include <samplerate.h>
#else
#include "libsamplerate/samplerate.h"
#endif

#if !(HAVE_LRINT && HAVE_LRINTF)
#define lrint(dbl)              ((int) (dbl))
#define lrintf(flt)             ((int) (flt))
#endif

// use a one second buffer
static const int DECODING_BUFFER_SIZE = 75*2352;

class K3bAudioDecoder::Private
{
public:
  Private()
    : metaInfo(0),
      resampleState(0),
      resampleData(0),
      inBuffer(0),
      inBufferPos(0),
      inBufferFill(0),
      outBuffer(0),
      monoBuffer(0),
      decodingBufferPos(0),
      decodingBufferFill(0),
      valid(true) {
  }

  // the current position of the decoder
  // This does NOT include the decodingBuffer
  K3b::Msf currentPos;

  // since the current position above is measured in frames
  // there might be a little offset since the decoded data is not
  // always a multiple of 2353 bytes
  int currentPosOffset;

  // already decoded bytes from last init or last seek
  // TODO: replace alreadyDecoded with currentPos
  unsigned long alreadyDecoded;

  K3b::Msf decodingStartPos;

  KFileMetaInfo* metaInfo;

  // set to true once decodeInternal() returned 0
  bool decoderFinished;

  // resampling
  SRC_STATE* resampleState;
  SRC_DATA* resampleData;

  float* inBuffer;
  float* inBufferPos;
  int inBufferFill;

  float* outBuffer;

  int samplerate;
  int channels;

  // mono -> stereo conversion
  char* monoBuffer;

  char decodingBuffer[DECODING_BUFFER_SIZE];
  char* decodingBufferPos;
  int decodingBufferFill;

  QMap<QString, QString> technicalInfoMap;
  QMap<MetaDataField, QString> metaInfoMap;

  bool valid;
};



K3bAudioDecoder::K3bAudioDecoder( QObject* parent, const char* name )
  : QObject( parent, name )
{
  d = new Private();
}


K3bAudioDecoder::~K3bAudioDecoder()
{
  cleanup();

  if( d->inBuffer ) delete [] d->inBuffer;
  if( d->outBuffer ) delete [] d->outBuffer;
  if( d->monoBuffer ) delete [] d->monoBuffer;

  delete d->metaInfo;
  delete d->resampleData;
  if( d->resampleState )
    src_delete( d->resampleState );
  delete d;
}


void K3bAudioDecoder::setFilename( const QString& filename )
{
  m_fileName = filename;
  delete d->metaInfo;
  d->metaInfo = 0;
}


bool K3bAudioDecoder::isValid() const
{
  return d->valid;
}


bool K3bAudioDecoder::analyseFile()
{
  d->technicalInfoMap.clear();
  d->metaInfoMap.clear();
  delete d->metaInfo;
  d->metaInfo = 0;

  cleanup();

  bool ret = analyseFileInternal( m_length, d->samplerate, d->channels );
  if( ret && ( d->channels == 1 || d->channels == 2 ) && m_length > 0 ) {
    d->valid = initDecoder();
    return d->valid;
  }
  else {
    d->valid = false;
    return false;
  }
}


bool K3bAudioDecoder::initDecoder( const K3b::Msf& startOffset )
{
  if( initDecoder() ) {
    if( startOffset > 0 )
      return seek( startOffset );
    else
      return true;
  }
  else
    return false;
}


bool K3bAudioDecoder::initDecoder()
{
  cleanup();

  if( d->resampleState )
    src_reset( d->resampleState );

  d->alreadyDecoded = 0;
  d->currentPos = 0;
  d->currentPosOffset = 0;
  d->decodingBufferFill = 0;
  d->decodingBufferPos = 0;
  d->decodingStartPos = 0;
  d->inBufferFill = 0;

  d->decoderFinished = false;

  return initDecoderInternal();
}


int K3bAudioDecoder::decode( char* _data, int maxLen )
{
  unsigned long lengthToDecode = (m_length - d->decodingStartPos).audioBytes();

  if( d->alreadyDecoded >= lengthToDecode )
    return 0;

  if( maxLen <= 0 )
    return 0;

  int read = 0;

  if( d->decodingBufferFill == 0 ) {
    //
    // now we decode into the decoding buffer
    // to ensure a minimum buffer size
    //
    d->decodingBufferFill = 0;
    d->decodingBufferPos = d->decodingBuffer;

    if( !d->decoderFinished ) {
      if( d->samplerate != 44100 ) {

	// check if we have data left from some previous conversion
	if( d->inBufferFill > 0 ) {
	  read = resample( d->decodingBuffer, DECODING_BUFFER_SIZE );
	}
	else {
	  if( !d->inBuffer ) {
	    d->inBuffer = new float[DECODING_BUFFER_SIZE/2];
	  }

	  if( (read = decodeInternal( d->decodingBuffer, DECODING_BUFFER_SIZE )) == 0 )
	    d->decoderFinished = true;

	  d->inBufferFill = read/2;
	  d->inBufferPos = d->inBuffer;
	  from16bitBeSignedToFloat( d->decodingBuffer, d->inBuffer, d->inBufferFill );
      
	  read = resample( d->decodingBuffer, DECODING_BUFFER_SIZE );
	}
      }
      else if( d->channels == 1 ) {
	if( !d->monoBuffer ) {
	  d->monoBuffer = new char[DECODING_BUFFER_SIZE/2];
	}

	// we simply duplicate every frame
	if( (read = decodeInternal( d->monoBuffer, DECODING_BUFFER_SIZE/2 )) == 0 )
	  d->decoderFinished = true;

	for( int i = 0; i < read; i+=2 ) {
	  d->decodingBuffer[2*i] = d->decodingBuffer[2*i+2] = d->monoBuffer[i];
	  d->decodingBuffer[2*i+1] = d->decodingBuffer[2*i+3] = d->monoBuffer[i+1];
	}

	read *= 2;
      }
      else {
	if( (read = decodeInternal( d->decodingBuffer, DECODING_BUFFER_SIZE )) == 0 )
	  d->decoderFinished = true;
      }
    }
  
    if( read < 0 ) {
      return -1;
    }
    else if( read == 0 ) {
      // check if we need to pad
      int bytesToPad = lengthToDecode - d->alreadyDecoded;
      if( bytesToPad > 0 ) {
	kdDebug() << "(K3bAudioDecoder) track length: " << lengthToDecode
		  << "; decoded module data: " << d->alreadyDecoded
		  << "; we need to pad " << bytesToPad << " bytes." << endl;
	
	if( DECODING_BUFFER_SIZE < bytesToPad )
	  bytesToPad = DECODING_BUFFER_SIZE;
	
	::memset( d->decodingBuffer, 0, bytesToPad );
	
	kdDebug() << "(K3bAudioDecoder) padded " << bytesToPad << " bytes." << endl;
	
	read = bytesToPad;
      }
      else {
	kdDebug() << "(K3bAudioDecoder) decoded " << d->alreadyDecoded << " bytes." << endl;
	return 0;
      }
    }
    else {
      
      // check if we decoded too much
      if( d->alreadyDecoded + read > lengthToDecode ) {
	kdDebug() << "(K3bAudioDecoder) we decoded too much. Cutting output by " 
		  << (read + d->alreadyDecoded - lengthToDecode) << endl;
	read = lengthToDecode - d->alreadyDecoded;
      }
    }

    d->decodingBufferFill = read;
  }


  // clear out the decoding buffer
  read = QMIN( maxLen, d->decodingBufferFill );
  ::memcpy( _data, d->decodingBufferPos, read );
  d->decodingBufferPos += read;
  d->decodingBufferFill -= read;

  d->alreadyDecoded += read;
  d->currentPos += (read+d->currentPosOffset)/2352;
  d->currentPosOffset = (read+d->currentPosOffset)%2352;

  return read;
}


// resample data in d->inBufferPos and save the result to data
// 
// 
int K3bAudioDecoder::resample( char* data, int maxLen )
{
  if( !d->resampleState ) {
    d->resampleState = src_new( SRC_SINC_MEDIUM_QUALITY, d->channels, 0 );
    if( !d->resampleState ) {
      kdDebug() << "(K3bAudioDecoder) unable to initialize resampler." << endl;
      return -1;
    }
    d->resampleData = new SRC_DATA;
  }

  if( !d->outBuffer ) {
    d->outBuffer = new float[DECODING_BUFFER_SIZE/2];
  }

  d->resampleData->data_in = d->inBufferPos;
  d->resampleData->data_out = d->outBuffer;
  d->resampleData->input_frames = d->inBufferFill/d->channels;
  d->resampleData->output_frames = maxLen/2/2;  // in case of mono files we need the space anyway
  d->resampleData->src_ratio = 44100.0/(double)d->samplerate;
  if( d->inBufferFill == 0 )
    d->resampleData->end_of_input = 1;  // this should force libsamplerate to output the last frames
  else
    d->resampleData->end_of_input = 0;

  int len = 0;
  if( (len = src_process( d->resampleState, d->resampleData ) ) ) {
    kdDebug() << "(K3bAudioDecoder) error while resampling: " << src_strerror(len) << endl;
    return -1;
  }

  if( d->channels == 2 )
    fromFloatTo16BitBeSigned( d->outBuffer, data, d->resampleData->output_frames_gen*d->channels );
  else {
    for( int i = 0; i < d->resampleData->output_frames_gen; ++i ) {
      fromFloatTo16BitBeSigned( &d->outBuffer[i], &data[4*i], 1 );
      fromFloatTo16BitBeSigned( &d->outBuffer[i], &data[4*i+2], 1 );
    }
  }
   
  d->inBufferPos += d->resampleData->input_frames_used*d->channels;
  d->inBufferFill -= d->resampleData->input_frames_used*d->channels;
  if( d->inBufferFill <= 0 ) {
    d->inBufferPos = d->inBuffer;
    d->inBufferFill = 0;
  }

  // 16 bit frames, so we need to multiply by 2
  // and we always have two channels
  return d->resampleData->output_frames_gen*2*2;
}


void K3bAudioDecoder::from16bitBeSignedToFloat( char* src, float* dest, int samples )
{
  while( samples ) {
    samples--;
    dest[samples] = static_cast<float>( Q_INT16(((src[2*samples]<<8)&0xff00)|(src[2*samples+1]&0x00ff)) / 32768.0 );
  }
}


void K3bAudioDecoder::fromFloatTo16BitBeSigned( float* src, char* dest, int samples )
{
  while( samples ) {
    samples--;

    float scaled = src[samples] * 32768.0;
    Q_INT16 val = 0;

    // clipping
    if( scaled >= ( 1.0 * 0x7FFF ) )
      val = 32767;
    else if( scaled <= ( -8.0 * 0x1000 ) )
      val = -32768;
    else
      val = lrintf(scaled);

    dest[2*samples]   = val>>8;
    dest[2*samples+1] = val;
  }
}


void K3bAudioDecoder::from8BitTo16BitBeSigned( char* src, char* dest, int samples )
{
  while( samples ) {
    samples--;

    float scaled = static_cast<float>(Q_UINT8(src[samples])-128) / 128.0 * 32768.0;
    Q_INT16 val = 0;
      
    // clipping
    if( scaled >= ( 1.0 * 0x7FFF ) )
      val = 32767;
    else if( scaled <= ( -8.0 * 0x1000 ) )
      val = -32768;
    else
      val = lrintf(scaled);

    dest[2*samples]   = val>>8;
    dest[2*samples+1] = val;
  }
}


bool K3bAudioDecoder::seek( const K3b::Msf& pos )
{
  kdDebug() << "(K3bAudioDecoder) seek from " << d->currentPos.toString() << " (+" << d->currentPosOffset
	    << ") to " << pos.toString() << endl;

  if( pos > length() )
    return false;

  d->decoderFinished = false;

  if( pos == d->currentPos && d->currentPosOffset == 0 )
    return true;

  if( pos == 0 )
    return initDecoder();

  bool success = false;

  //
  // First check if we may do a "perfect seek".
  // We cannot rely on the decoding plugins to seek perfectly. Especially
  // the mp3 decoder does not. But in case we want to split a live recording
  // it is absolutely nesseccary to perform a perfect seek.
  // So if we did not already decode past the seek position and the difference
  // between the current position and the seek position is less than some fixed
  // value we simply decode up to the seek position.
  //
  if( ( pos > d->currentPos || 
	( pos == d->currentPos && d->currentPosOffset == 0 ) )
      &&
      ( pos - d->currentPos < K3b::Msf(0,10,0) ) ) {  // < 10 seconds is ok
    kdDebug() << "(K3bAudioDecoder) performing perfect seek from " << d->currentPos.toString()
	      << " to " << pos.toString() << ". :)" << endl;
   
    unsigned long bytesToDecode = pos.audioBytes() - d->currentPos.audioBytes() - d->currentPosOffset;
    kdDebug() << "(K3bAudioDecoder) seeking " << bytesToDecode << " bytes." << endl;
    char buffi[10*2352];
    while( bytesToDecode > 0 ) {
      int read = decode( buffi, QMIN(10*2352, bytesToDecode) );
      if( read <= 0 )
	return false;
      
      bytesToDecode -= read;
    }

    kdDebug() << "(K3bAudioDecoder) perfect seek done." << endl;

    success = true;
  }
  else {
    //
    // Here we have to reset the resampling stuff since we restart decoding at another position.
    //
    if( d->resampleState )
      src_reset( d->resampleState );
    d->inBufferFill = 0;

    //
    // And also reset the decoding buffer to not return any garbage from previous decoding.
    //
    d->decodingBufferFill = 0;
    
    success = seekInternal( pos );
  }

  d->alreadyDecoded = 0;
  d->currentPos = d->decodingStartPos = pos;
  d->currentPosOffset = 0;

  return success;
}


void K3bAudioDecoder::cleanup()
{
}


QString K3bAudioDecoder::metaInfo( MetaDataField f )
{
  if( d->metaInfoMap.contains( f ) )
    return d->metaInfoMap[f];

  // fall back to KFileMetaInfo
  if( !d->metaInfo )
    d->metaInfo = new KFileMetaInfo( filename() );

  if( d->metaInfo->isValid() ) {
    QString tag;
    switch( f ) {
    case META_TITLE:
      tag = "Title";
      break;
    case META_ARTIST:
      tag = "Artist";
      break;
    case META_SONGWRITER:
      tag = "Songwriter";
      break;
    case META_COMPOSER:
      tag = "Composer";
      break;
    case META_COMMENT:
      tag = "Comment";
      break;
    }

    KFileMetaInfoItem item = d->metaInfo->item( tag );
    if( item.isValid() )
      return item.string();
  }

  return QString::null;
}


void K3bAudioDecoder::addMetaInfo( MetaDataField f, const QString& value )
{
  if( !value.isEmpty() )
    d->metaInfoMap[f] = value;
  else
    kdDebug() << "(K3bAudioDecoder) empty meta data field." << endl;
}


QStringList K3bAudioDecoder::supportedTechnicalInfos() const
{
  QStringList l;
  for( QMap<QString, QString>::const_iterator it = d->technicalInfoMap.begin();
       it != d->technicalInfoMap.end(); ++it )
    l.append( it.key() );
  return l;
}


QString K3bAudioDecoder::technicalInfo( const QString& key ) const
{
  return d->technicalInfoMap[key];
}


void K3bAudioDecoder::addTechnicalInfo( const QString& key, const QString& value )
{
  d->technicalInfoMap[key] = value;
}


K3bAudioDecoder* K3bAudioDecoderFactory::createDecoder( const KURL& url )
{
  kdDebug() << "(K3bAudioDecoderFactory::createDecoder( " << url.path() << " )" << endl;
  QPtrList<K3bPlugin> fl = k3bcore->pluginManager()->plugins( "AudioDecoder" );

  // first search for a single format decoder
  for( QPtrListIterator<K3bPlugin> it( fl ); it.current(); ++it ) {
    K3bAudioDecoderFactory* f = dynamic_cast<K3bAudioDecoderFactory*>( it.current() );
    if( f && !f->multiFormatDecoder() && f->canDecode( url ) ) {
      kdDebug() << "1" << endl; return f->createDecoder();}
  }
  
  // no single format decoder. Search for a multi format decoder
  for( QPtrListIterator<K3bPlugin> it( fl ); it.current(); ++it ) {
    K3bAudioDecoderFactory* f = dynamic_cast<K3bAudioDecoderFactory*>( it.current() );
    if( f && f->multiFormatDecoder() && f->canDecode( url ) ) {
      kdDebug() << "2" << endl; return f->createDecoder();}
  }

  kdDebug() << "(K3bAudioDecoderFactory::createDecoder( " << url.path() << " ) no success" << endl;

  // nothing found
  return 0;  
}

#include "k3baudiodecoder.moc"