// Copyright (C) 2003 Davis E. King (davis@dlib.net) // License: Boost Software License See LICENSE.txt for the full license. #ifndef DLIB_COMPRESS_STREAM_KERNEl_2_ #define DLIB_COMPRESS_STREAM_KERNEl_2_ #include "../algs.h" #include #include #include "compress_stream_kernel_abstract.h" namespace dlib { template < typename fce, typename fcd, typename lz77_buffer, typename sliding_buffer, typename fce_length, typename fcd_length, typename fce_index, typename fcd_index, typename crc32 > class compress_stream_kernel_2 { /*! REQUIREMENTS ON fce is an implementation of entropy_encoder_model/entropy_encoder_model_kernel_abstract.h the alphabet_size of fce must be 257. fce and fcd share the same kernel number. REQUIREMENTS ON fcd is an implementation of entropy_decoder_model/entropy_decoder_model_kernel_abstract.h the alphabet_size of fcd must be 257. fce and fcd share the same kernel number. REQUIREMENTS ON lz77_buffer is an implementation of lz77_buffer/lz77_buffer_kernel_abstract.h REQUIREMENTS ON sliding_buffer is an implementation of sliding_buffer/sliding_buffer_kernel_abstract.h is instantiated with T = unsigned char REQUIREMENTS ON fce_length is an implementation of entropy_encoder_model/entropy_encoder_model_kernel_abstract.h the alphabet_size of fce must be 513. This will be used to encode the length of lz77 matches. fce_length and fcd share the same kernel number. REQUIREMENTS ON fcd_length is an implementation of entropy_decoder_model/entropy_decoder_model_kernel_abstract.h the alphabet_size of fcd must be 513. This will be used to decode the length of lz77 matches. fce_length and fcd share the same kernel number. REQUIREMENTS ON fce_index is an implementation of entropy_encoder_model/entropy_encoder_model_kernel_abstract.h the alphabet_size of fce must be 32257. This will be used to encode the index of lz77 matches. fce_index and fcd share the same kernel number. REQUIREMENTS ON fcd_index is an implementation of entropy_decoder_model/entropy_decoder_model_kernel_abstract.h the alphabet_size of fcd must be 32257. This will be used to decode the index of lz77 matches. fce_index and fcd share the same kernel number. REQUIREMENTS ON crc32 is an implementation of crc32/crc32_kernel_abstract.h INITIAL VALUE this object has no state CONVENTION this object has no state !*/ const static unsigned long eof_symbol = 256; public: class decompression_error : public dlib::error { public: decompression_error( const char* i ) : dlib::error(std::string(i)) {} decompression_error( const std::string& i ) : dlib::error(i) {} }; compress_stream_kernel_2 ( ) {} ~compress_stream_kernel_2 ( ) {} void compress ( std::istream& in, std::ostream& out ) const; void decompress ( std::istream& in, std::ostream& out ) const; private: // restricted functions compress_stream_kernel_2(compress_stream_kernel_2&); // copy constructor compress_stream_kernel_2& operator=(compress_stream_kernel_2&); // assignment operator }; // ---------------------------------------------------------------------------------------- // ---------------------------------------------------------------------------------------- // member function definitions // ---------------------------------------------------------------------------------------- // ---------------------------------------------------------------------------------------- template < typename fce, typename fcd, typename lz77_buffer, typename sliding_buffer, typename fce_length, typename fcd_length, typename fce_index, typename fcd_index, typename crc32 > void compress_stream_kernel_2:: compress ( std::istream& in_, std::ostream& out_ ) const { std::streambuf::int_type temp; std::streambuf& in = *in_.rdbuf(); typename fce::entropy_encoder_type coder; coder.set_stream(out_); fce model(coder); fce_length model_length(coder); fce_index model_index(coder); const unsigned long LOOKAHEAD_LIMIT = 512; lz77_buffer buffer(15,LOOKAHEAD_LIMIT); crc32 crc; unsigned long count = 0; unsigned long lz77_count = 1; // number of times we used lz77 to encode unsigned long ppm_count = 1; // number of times we used ppm to encode while (true) { // write out a known value every 20000 symbols if (count == 20000) { count = 0; coder.encode(150,151,400); } ++count; // try to fill the lookahead buffer if (buffer.get_lookahead_buffer_size() < buffer.get_lookahead_buffer_limit()) { temp = in.sbumpc(); while (temp != EOF) { crc.add(static_cast(temp)); buffer.add(static_cast(temp)); if (buffer.get_lookahead_buffer_size() == buffer.get_lookahead_buffer_limit()) break; temp = in.sbumpc(); } } // compute the sum of ppm_count and lz77_count but make sure // it is less than 65536 unsigned long sum = ppm_count + lz77_count; if (sum >= 65536) { ppm_count >>= 1; lz77_count >>= 1; ppm_count |= 1; lz77_count |= 1; sum = ppm_count+lz77_count; } // if there are still more symbols in the lookahead buffer to encode if (buffer.get_lookahead_buffer_size() > 0) { unsigned long match_index, match_length; buffer.find_match(match_index,match_length,6); if (match_length != 0) { // signal the decoder that we are using lz77 coder.encode(0,lz77_count,sum); ++lz77_count; // encode the index and length pair model_index.encode(match_index); model_length.encode(match_length); } else { // signal the decoder that we are using ppm coder.encode(lz77_count,sum,sum); ++ppm_count; // encode the symbol using the ppm model model.encode(buffer.lookahead_buffer(0)); buffer.shift_buffers(1); } } else { // signal the decoder that we are using ppm coder.encode(lz77_count,sum,sum); model.encode(eof_symbol); // now write the checksum unsigned long checksum = crc.get_checksum(); unsigned char byte1 = static_cast((checksum>>24)&0xFF); unsigned char byte2 = static_cast((checksum>>16)&0xFF); unsigned char byte3 = static_cast((checksum>>8)&0xFF); unsigned char byte4 = static_cast((checksum)&0xFF); model.encode(byte1); model.encode(byte2); model.encode(byte3); model.encode(byte4); break; } } // while (true) } // ---------------------------------------------------------------------------------------- template < typename fce, typename fcd, typename lz77_buffer, typename sliding_buffer, typename fce_length, typename fcd_length, typename fce_index, typename fcd_index, typename crc32 > void compress_stream_kernel_2:: decompress ( std::istream& in_, std::ostream& out_ ) const { std::streambuf& out = *out_.rdbuf(); typename fcd::entropy_decoder_type coder; coder.set_stream(in_); fcd model(coder); fcd_length model_length(coder); fcd_index model_index(coder); unsigned long symbol; unsigned long count = 0; sliding_buffer buffer; buffer.set_size(15); // Initialize the buffer to all zeros. There is no algorithmic reason to // do this. But doing so avoids a warning from valgrind so that is why // I'm doing this. for (unsigned long i = 0; i < buffer.size(); ++i) buffer[i] = 0; crc32 crc; unsigned long lz77_count = 1; // number of times we used lz77 to encode unsigned long ppm_count = 1; // number of times we used ppm to encode bool next_block_lz77; // decode until we hit the marker symbol while (true) { // make sure this is the value we expect if (count == 20000) { if (coder.get_target(400) != 150) { throw decompression_error("Error detected in compressed data stream."); } count = 0; coder.decode(150,151); } ++count; // compute the sum of ppm_count and lz77_count but make sure // it is less than 65536 unsigned long sum = ppm_count + lz77_count; if (sum >= 65536) { ppm_count >>= 1; lz77_count >>= 1; ppm_count |= 1; lz77_count |= 1; sum = ppm_count+lz77_count; } // check if we are decoding a lz77 or ppm block if (coder.get_target(sum) < lz77_count) { coder.decode(0,lz77_count); next_block_lz77 = true; ++lz77_count; } else { coder.decode(lz77_count,sum); next_block_lz77 = false; ++ppm_count; } if (next_block_lz77) { unsigned long match_length, match_index; // decode the match index model_index.decode(match_index); // decode the match length model_length.decode(match_length); match_index += match_length; buffer.rotate_left(match_length); for (unsigned long i = 0; i < match_length; ++i) { unsigned char ch = buffer[match_index-i]; buffer[match_length-i-1] = ch; crc.add(ch); // write this ch to out if (out.sputc(static_cast(ch)) != static_cast(ch)) { throw std::ios::failure("error occurred in compress_stream_kernel_2::decompress"); } } } else { // decode the next symbol model.decode(symbol); if (symbol != eof_symbol) { buffer.rotate_left(1); buffer[0] = static_cast(symbol); crc.add(static_cast(symbol)); // write this symbol to out if (out.sputc(static_cast(symbol)) != static_cast(symbol)) { throw std::ios::failure("error occurred in compress_stream_kernel_2::decompress"); } } else { // this was the eof marker symbol so we are done. now check the checksum // now get the checksum and make sure it matches unsigned char byte1; unsigned char byte2; unsigned char byte3; unsigned char byte4; model.decode(symbol); byte1 = static_cast(symbol); model.decode(symbol); byte2 = static_cast(symbol); model.decode(symbol); byte3 = static_cast(symbol); model.decode(symbol); byte4 = static_cast(symbol); unsigned long checksum = byte1; checksum <<= 8; checksum |= byte2; checksum <<= 8; checksum |= byte3; checksum <<= 8; checksum |= byte4; if (checksum != crc.get_checksum()) throw decompression_error("Error detected in compressed data stream."); break; } } } // while (true) } // ---------------------------------------------------------------------------------------- } #endif // DLIB_COMPRESS_STREAM_KERNEl_2_