path: root/libkmime/kmime_codec_qp.cpp

/*  -*- c++ -*-
    kmime_codec_qp.cpp

    This file is part of KMime, the KDE internet mail/usenet news message library.
    Copyright (c) 2002 Marc Mutz <mutz@kde.org>

    KMime is free software; you can redistribute it and/or modify it
    under the terms of the GNU General Public License, version 2, as
    published by the Free Software Foundation.

    KMime 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 this library; if not, write to the Free Software
    Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA

    In addition, as a special exception, the copyright holders give
    permission to link the code of this library with any edition of
    the Qt library by Trolltech AS, Norway (or with modified versions
    of Qt that use the same license as Qt), and distribute linked
    combinations including the two.  You must obey the GNU General
    Public License in all respects for all of the code used other than
    Qt.  If you modify this file, you may extend this exception to
    your version of the file, but you are not obligated to do so.  If
    you do not wish to do so, delete this exception statement from
    your version.
*/

#include "kmime_codec_qp.h"

#include "kmime_util.h"

#include <kdebug.h>

#include <cassert>

using namespace KMime;

namespace KMime {

// some helpful functions:

static inline char binToHex( uchar value ) {
  if ( value > 9 )
    return value + 'A' - 10;
  else
    return value + '0';
}

static inline uchar highNibble( uchar ch ) {
  return ch >> 4;
}

static inline uchar lowNibble( uchar ch ) {
  return ch & 0xF;
}

static inline bool keep( uchar ch ) {
  // no CTLs, except HT and not '?'
  return !( ch < ' ' && ch != '\t' || ch == '?' );
}

//
// QuotedPrintableCodec
//

class QuotedPrintableEncoder : public Encoder {
  char mInputBuffer[16];
  uchar mCurrentLineLength; // 0..76
  uchar mAccu;
  uint mInputBufferReadCursor  : 4; // 0..15
  uint mInputBufferWriteCursor : 4; // 0..15
  enum {
    Never, AtBOL, Definitely
  } mAccuNeedsEncoding    : 2;
  bool mSawLineEnd        : 1;
  bool mSawCR             : 1;
  bool mFinishing         : 1;
  bool mFinished          : 1;
protected:
  friend class QuotedPrintableCodec;
  QuotedPrintableEncoder( bool withCRLF=false )
    : Encoder( withCRLF ), mCurrentLineLength(0), mAccu(0),
      mInputBufferReadCursor(0), mInputBufferWriteCursor(0),
      mAccuNeedsEncoding(Never),
      mSawLineEnd(false), mSawCR(false), mFinishing(false),
      mFinished(false) {}

  bool needsEncoding( uchar ch ) {
    return ( ch > '~' || ch < ' ' && ch != '\t' || ch == '=' );
  }
  bool needsEncodingAtEOL( uchar ch ) {
    return ( ch == ' ' || ch == '\t' );
  }
  bool needsEncodingAtBOL( uchar ch ) {
    return ( ch == 'F' || ch == '.' || ch == '-' );
  }
  bool fillInputBuffer( const char* & scursor, const char * const send );
  bool processNextChar();
  void createOutputBuffer( char* & dcursor, const char * const dend );
public:
  virtual ~QuotedPrintableEncoder() {}

  bool encode( const char* & scursor, const char * const send,
	       char* & dcursor, const char * const dend );

  bool finish( char* & dcursor, const char * const dend );
};


class QuotedPrintableDecoder : public Decoder {
  const char mEscapeChar;
  char mBadChar;
  /** @p accu holds the msb nibble of the hexchar or zero. */
  uchar mAccu;
  /** @p insideHexChar is true iff we're inside an hexchar (=XY).
      Together with @ref mAccu, we can build this states:
      @li @p insideHexChar == @p false:
          normal text
      @li @p insideHexChar == @p true, @p mAccu == 0:
          saw the leading '='
      @li @p insideHexChar == @p true, @p mAccu != 0:
          saw the first nibble '=X'
   */
  const bool mQEncoding;
  bool mInsideHexChar;
  bool mFlushing;
  bool mExpectLF;
  bool mHaveAccu;
protected:
  friend class QuotedPrintableCodec;
  friend class Rfc2047QEncodingCodec;
  friend class Rfc2231EncodingCodec;
  QuotedPrintableDecoder( bool withCRLF=false,
			  bool aQEncoding=false, char aEscapeChar='=' )
    : Decoder( withCRLF ),
      mEscapeChar(aEscapeChar),
      mBadChar(0),
      mAccu(0),
      mQEncoding(aQEncoding),
      mInsideHexChar(false),
      mFlushing(false),
      mExpectLF(false),
      mHaveAccu(false) {}
public:
  virtual ~QuotedPrintableDecoder() {}

  bool decode( const char* & scursor, const char * const send,
	       char* & dcursor, const char * const dend );
  // ### really no finishing needed???
  bool finish( char* &, const char * const ) { return true; }
};


class Rfc2047QEncodingEncoder : public Encoder {
  uchar      mAccu;
  uchar      mStepNo;
  const char mEscapeChar;
  bool       mInsideFinishing : 1;
protected:
  friend class Rfc2047QEncodingCodec;
  friend class Rfc2231EncodingCodec;
  Rfc2047QEncodingEncoder( bool withCRLF=false, char aEscapeChar='=' )
    : Encoder( withCRLF ),
      mAccu(0), mStepNo(0), mEscapeChar( aEscapeChar ),
      mInsideFinishing( false )
  {
    // else an optimization in ::encode might break.
    assert( aEscapeChar == '=' || aEscapeChar == '%' );
  }

  // this code assumes that isEText( mEscapeChar ) == false!
  bool needsEncoding( uchar ch ) {
    if ( ch > 'z' ) return true; // {|}~ DEL and 8bit chars need
    if ( !isEText( ch ) ) return true; // all but a-zA-Z0-9!/*+- need, too
    if ( mEscapeChar == '%' && ( ch == '*' || ch == '/' ) )
      return true; // not allowed in rfc2231 encoding
    return false;
  }

public:
  virtual ~Rfc2047QEncodingEncoder() {}

  bool encode( const char* & scursor, const char * const send,
	       char* & dcursor, const char * const dend );
  bool finish( char* & dcursor, const char * const dend );
};

// this doesn't access any member variables, so it can be defined static
// but then we can't call it from virtual functions
static int QuotedPrintableDecoder_maxDecodedSizeFor( int insize, bool withCRLF ) {
  // all chars unencoded:
  int result = insize;
  // but maybe all of them are \n and we need to make them \r\n :-o
  if ( withCRLF )
    result += insize;

  // there might be an accu plus escape
  result += 2;

  return result;
}

Encoder * QuotedPrintableCodec::makeEncoder( bool withCRLF ) const {
  return new QuotedPrintableEncoder( withCRLF );
}

Decoder * QuotedPrintableCodec::makeDecoder( bool withCRLF ) const {
  return new QuotedPrintableDecoder( withCRLF );
}

int QuotedPrintableCodec::maxDecodedSizeFor( int insize, bool withCRLF ) const {
    return QuotedPrintableDecoder_maxDecodedSizeFor(insize, withCRLF);
}

Encoder * Rfc2047QEncodingCodec::makeEncoder( bool withCRLF ) const {
  return new Rfc2047QEncodingEncoder( withCRLF );
}

Decoder * Rfc2047QEncodingCodec::makeDecoder( bool withCRLF ) const {
  return new QuotedPrintableDecoder( withCRLF, true );
}

int Rfc2047QEncodingCodec::maxDecodedSizeFor( int insize, bool withCRLF ) const {
    return QuotedPrintableDecoder_maxDecodedSizeFor(insize, withCRLF);
}

Encoder * Rfc2231EncodingCodec::makeEncoder( bool withCRLF ) const {
  return new Rfc2047QEncodingEncoder( withCRLF, '%' );
}

Decoder * Rfc2231EncodingCodec::makeDecoder( bool withCRLF ) const {
  return new QuotedPrintableDecoder( withCRLF, true, '%' );
}

int Rfc2231EncodingCodec::maxDecodedSizeFor( int insize, bool withCRLF ) const {
    return QuotedPrintableDecoder_maxDecodedSizeFor(insize, withCRLF);
}

  /********************************************************/
  /********************************************************/
  /********************************************************/

bool QuotedPrintableDecoder::decode( const char* & scursor, const char * const send,
				     char* & dcursor, const char * const dend ) {
  if ( mWithCRLF )
    kdWarning() << "CRLF output for decoders isn't yet supported!" << endl;

  while ( scursor != send && dcursor != dend ) {
    if ( mFlushing ) {
      // we have to flush chars in the aftermath of an decoding
      // error. The way to request a flush is to
      // - store the offending character in mBadChar and
      // - set mFlushing to true.
      // The supported cases are (H: hexchar, X: bad char):
      // =X, =HX, CR
      // mBadChar is only written out if it is not by itself illegal in
      // quoted-printable (e.g. CTLs, 8Bits).
      // A fast way to suppress mBadChar output is to set it to NUL.
      if ( mInsideHexChar ) {
	// output '='
	*dcursor++ = mEscapeChar;
	mInsideHexChar = false;
      } else if ( mHaveAccu ) {
	// output the high nibble of the accumulator:
	*dcursor++ = binToHex( highNibble( mAccu ) );
	mHaveAccu = false;
	mAccu = 0;
      } else {
	// output mBadChar
	assert( mAccu == 0 );
	if ( mBadChar ) {
	  if ( mBadChar >= '>' && mBadChar <= '~' ||
	       mBadChar >= '!' && mBadChar <= '<' )
	    *dcursor++ = mBadChar;
	  mBadChar = 0;
	}
	mFlushing = false;
      }
      continue;
    }
    assert( mBadChar == 0 );

    uchar ch = *scursor++;
    uchar value = 255;

    if ( mExpectLF && ch != '\n' ) {
      kdWarning() << "QuotedPrintableDecoder: "
	"illegally formed soft linebreak or lonely CR!" << endl;
      mInsideHexChar = false;
      mExpectLF = false;
      assert( mAccu == 0 );
    }

    if ( mInsideHexChar ) {
      // next char(s) represent nibble instead of itself:
      if ( ch <= '9' ) {
	if ( ch >= '0' ) {
	  value = ch - '0';
	} else {
	  switch ( ch ) {
	  case '\r':
	    mExpectLF = true;
	    break;
	  case '\n':
	    // soft line break, but only if mAccu is NUL.
	    if ( !mHaveAccu ) {
	      mExpectLF = false;
	      mInsideHexChar = false;
	      break;
	    }
	    // else fall through
	  default:
	    kdWarning() << "QuotedPrintableDecoder: "
	      "illegally formed hex char! Outputting verbatim." << endl;
	    mBadChar = ch;
	    mFlushing = true;
	  }
	  continue;
	}
      } else { // ch > '9'
	if ( ch <= 'F' ) {
	  if ( ch >= 'A' ) {
	    value = 10 + ch - 'A';
	  } else { // [:-@]
	    mBadChar = ch;
	    mFlushing = true;
	    continue;
	  }
	} else { // ch > 'F'
	  if ( ch <= 'f' && ch >= 'a' ) {
	    value = 10 + ch - 'a';
	  } else {
	    mBadChar = ch;
	    mFlushing = true;
	    continue;
	  }
	}
      }

      assert( value < 16 );
      assert( mBadChar == 0 );
      assert( !mExpectLF );

      if ( mHaveAccu ) {
	*dcursor++ = char( mAccu | value );
	mAccu = 0;
	mHaveAccu = false;
	mInsideHexChar = false;
      } else {
	mHaveAccu = true;
	mAccu = value << 4;
      }
    } else { // not mInsideHexChar
      if ( ch <= '~' && ch >= ' ' || ch == '\t' ) {
	if ( ch == mEscapeChar ) {
	  mInsideHexChar = true;
	} else if ( mQEncoding && ch == '_' ) {
	  *dcursor++ = char(0x20);
	} else {
	  *dcursor++ = char(ch);
	}
      } else if ( ch == '\n' ) {
	*dcursor++ = '\n';
	mExpectLF = false;
      } else if ( ch == '\r' ) {
	mExpectLF = true;
      } else {
	kdWarning() << "QuotedPrintableDecoder: " << ch <<
	  " illegal character in input stream! Ignoring." << endl;
      }
    }
  }

  return (scursor == send);
}

bool QuotedPrintableEncoder::fillInputBuffer( const char* & scursor,
					      const char * const send ) {
  // Don't read more if there's still a tail of a line in the buffer:
  if ( mSawLineEnd )
    return true;

  // Read until the buffer is full or we have found CRLF or LF (which
  // don't end up in the input buffer):
  for ( ; ( mInputBufferWriteCursor + 1 ) % 16 != mInputBufferReadCursor
	  && scursor != send ; mInputBufferWriteCursor++ ) {
    char ch = *scursor++;
    if ( ch == '\r' ) {
      mSawCR = true;
    } else if ( ch == '\n' ) {
      // remove the CR from the input buffer (if any) and return that
      // we found a line ending:
      if ( mSawCR ) {
	mSawCR = false;
	assert( mInputBufferWriteCursor != mInputBufferReadCursor );
	mInputBufferWriteCursor--;
      }
      mSawLineEnd = true;
      return true; // saw CRLF or LF
    } else {
      mSawCR = false;
    }
    mInputBuffer[ mInputBufferWriteCursor ] = ch;
  }
  mSawLineEnd = false;
  return false; // didn't see a line ending...
}

bool QuotedPrintableEncoder::processNextChar() {

  // If we process a buffer which doesn't end in a line break, we
  // can't process all of it, since the next chars that will be read
  // could be a line break. So we empty the buffer only until a fixed
  // number of chars is left (except when mFinishing, which means that
  // the data doesn't end in newline):
  const int minBufferFillWithoutLineEnd = 4;

  assert( mOutputBufferCursor == 0 );

  int bufferFill = int(mInputBufferWriteCursor) - int(mInputBufferReadCursor) ;
  if ( bufferFill < 0 )
    bufferFill += 16;

  assert( bufferFill >=0 && bufferFill <= 15 );

  if ( !mFinishing && !mSawLineEnd &&
       bufferFill < minBufferFillWithoutLineEnd )
    return false;

  // buffer is empty, return false:
  if ( mInputBufferReadCursor == mInputBufferWriteCursor )
    return false;

  // Real processing goes here:
  mAccu = mInputBuffer[ mInputBufferReadCursor++ ];
  if ( needsEncoding( mAccu ) ) // always needs encoding or
    mAccuNeedsEncoding = Definitely;
  else if ( ( mSawLineEnd || mFinishing )  // needs encoding at end of line
	    && bufferFill == 1             // or end of buffer
	    && needsEncodingAtEOL( mAccu ) )
    mAccuNeedsEncoding = Definitely;
  else if ( needsEncodingAtBOL( mAccu ) )
    mAccuNeedsEncoding = AtBOL;
  else
    // never needs encoding
    mAccuNeedsEncoding = Never;

  return true;
}

// Outputs processed (verbatim or hex-encoded) chars and inserts soft
// line breaks as necessary. Depends on processNextChar's directions
// on whether or not to encode the current char, and whether or not
// the current char is the last one in it's input line:
void QuotedPrintableEncoder::createOutputBuffer( char* & dcursor,
						 const char * const dend )
{
  const int maxLineLength = 76; // rfc 2045

  assert( mOutputBufferCursor == 0 );

  bool lastOneOnThisLine = mSawLineEnd
    && mInputBufferReadCursor == mInputBufferWriteCursor;

  int neededSpace = 1;
  if ( mAccuNeedsEncoding == Definitely)
    neededSpace = 3;

  // reserve space for the soft hyphen (=)
  if ( !lastOneOnThisLine )
    neededSpace++;

  if ( mCurrentLineLength > maxLineLength - neededSpace ) {
    // current line too short, insert soft line break:
    write( '=', dcursor, dend );
    writeCRLF( dcursor, dend );
    mCurrentLineLength = 0;
  }

  if ( Never == mAccuNeedsEncoding ||
       AtBOL == mAccuNeedsEncoding && mCurrentLineLength != 0 ) {
    write( mAccu, dcursor, dend );
    mCurrentLineLength++;
  } else {
    write( '=', dcursor, dend );
    write( binToHex( highNibble( mAccu ) ), dcursor, dend );
    write( binToHex( lowNibble( mAccu ) ), dcursor, dend );
    mCurrentLineLength += 3;
  }
}


bool QuotedPrintableEncoder::encode( const char* & scursor, const char * const send,
				     char* & dcursor, const char * const dend )
{
  // support probing by the caller:
  if ( mFinishing ) return true;

  while ( scursor != send && dcursor != dend ) {
    if ( mOutputBufferCursor && !flushOutputBuffer( dcursor, dend ) )
      return (scursor == send);

    assert( mOutputBufferCursor == 0 );

    // fill input buffer until eol has been reached or until the
    // buffer is full, whatever comes first:
    fillInputBuffer( scursor, send );

    if ( processNextChar() )
      // there was one...
      createOutputBuffer( dcursor, dend );
    else if ( mSawLineEnd &&
	      mInputBufferWriteCursor == mInputBufferReadCursor ) {
      // load a hard line break into output buffer:
      writeCRLF( dcursor, dend );
      // signal fillInputBuffer() we are ready for the next line:
      mSawLineEnd = false;
      mCurrentLineLength = 0;
    } else
      // we are supposedly finished with this input block:
      break;
  }

  // make sure we write as much as possible and don't stop _writing_
  // just because we have no more _input_:
  if ( mOutputBufferCursor ) flushOutputBuffer( dcursor, dend );

  return (scursor == send);

} // encode

bool QuotedPrintableEncoder::finish( char* & dcursor,
				     const char * const dend ) {
  mFinishing = true;

  if ( mFinished )
    return flushOutputBuffer( dcursor, dend );

  while ( dcursor != dend ) {
    if ( mOutputBufferCursor && !flushOutputBuffer( dcursor, dend ) )
      return false;

    assert( mOutputBufferCursor == 0 );

    if ( processNextChar() )
      // there was one...
      createOutputBuffer( dcursor, dend );
    else if ( mSawLineEnd &&
	      mInputBufferWriteCursor == mInputBufferReadCursor ) {
      // load a hard line break into output buffer:
      writeCRLF( dcursor, dend );
      mSawLineEnd = false;
      mCurrentLineLength = 0;
    } else {
      mFinished = true;
      return flushOutputBuffer( dcursor, dend );
    }
  }

  return mFinished && !mOutputBufferCursor;

} // finish


bool Rfc2047QEncodingEncoder::encode( const char* & scursor, const char * const send,
				      char* & dcursor, const char * const dend )
{
  if ( mInsideFinishing ) return true;

  while ( scursor != send && dcursor != dend ) {
    uchar value;
    switch ( mStepNo ) {
    case 0:
      // read the next char and decide if and how do encode:
      mAccu = *scursor++;
      if ( !needsEncoding( mAccu ) ) {
	*dcursor++ = char(mAccu);
      } else if ( mEscapeChar == '=' && mAccu == 0x20 ) {
	// shortcut encoding for 0x20 (latin-1/us-ascii SPACE)
	// (not for rfc2231 encoding)
	*dcursor++ = '_';
      } else {
	// needs =XY encoding - write escape char:
	*dcursor++ = mEscapeChar;
	mStepNo = 1;
      }
      continue;
    case 1:
      // extract hi-nibble:
      value = highNibble(mAccu);
      mStepNo = 2;
      break;
    case 2:
      // extract lo-nibble:
      value = lowNibble(mAccu);
      mStepNo = 0;
      break;
    default: assert( 0 );
    }

    // and write:
    *dcursor++ = binToHex( value );
  }

  return (scursor == send);
} // encode

#include <qstring.h>

bool Rfc2047QEncodingEncoder::finish( char* & dcursor, const char * const dend ) {
  mInsideFinishing = true;

  // write the last bits of mAccu, if any:
  while ( mStepNo != 0 && dcursor != dend ) {
    uchar value;
    switch ( mStepNo ) {
    case 1:
      // extract hi-nibble:
      value = highNibble(mAccu);
      mStepNo = 2;
      break;
    case 2:
      // extract lo-nibble:
      value = lowNibble(mAccu);
      mStepNo = 0;
      break;
    default: assert( 0 );
    }

    // and write:
    *dcursor++ = binToHex( value );
  }

  return mStepNo == 0;
}




} // namespace KMime