Szip

szip, sunzip: (c) 1997-1999 Michael Schindler, szip@compressconsult.com http://www.compressconsult.com/szip/ (R),SM: szip and the data compression logo are "geschuetztes Markenzeichen" of Michael Schindler US and other patents pending. The program szip performs data compression/decompression, the current version is 1.12a. There were several previous incompatible versions of szip; version 1.00 to 1.04 use one format, version 1.05X another one. version 1.10 was available as alpha test only but had a bug in encoding. If you have szip files from that versions you need to keep the old decompression program around; if you lost it you can still download them from the website. 1.11 and 1.12 differ only in internals (IO macros instead of function calls, more memory allocations) from this version. Usage: szip [options] [inputfile [outputfile]] option meaning default -d decompress -b<blocksize> blocksize in 100kB -b17 -o<order> order of context -o6 -r<recordsize> recordsize -r1 -i incremental coding (differences to previous value) -v<level> turn on messages -v0 options may be grouped like -b14o10r3 if outputfile is omitted output is written to standardoutput. if inputfile is omitted too input is read from standardinput. I recommend using .sz for szipped files and .tar.sz or .tsz for szipped tarfiles. Future versions will produce these extensions. option effects: decompress: tells the program to decompress; default operation mode is compression. If present all other options except v are ignored. blocksize: larger blocks usually give better compression, but if your system gets into paging it will be slow. No effect on speed if enough memory is available. 1-41 possible. order: higher order gives better compression (and increased time). 3-255 possible. There is special code for order 4; this will give a faster (even faster than order 3) compression. order 0 makes a BWT transform (unlimited order) Decompression of -o0 is fastest, so use it for distribution. fast compression but larger: -o4 fast decompression and probably smaller: -o0 Some files compress better with a small order like 3 or 4. recordsize: tells what size (in bytes) the elementary datatype is. getting this one right will improve compression. 24-bit graphics: use -r3 2-channel 8-bit audio: use -r2 4-byte words: use -r4 1-byte chars: use -r1 (default) the recordsize need not be in sync with the real record; if you have a 4-byte header on an -r3 file still choose -r3. 1-127 possible incremental: use differences to the last value (after recordsize reordering) instead of the actual value. Good for sounds. verbosity level: output progress messages. OPERATING SYSTEMS SUPPORTED: The code is plain C; please check out the webpage for available compilations. It can be compiled for other platforms upon request, please ask. The produced files are platform independent, and all versions 1.1x produce this format. COPYING: This Program can be freely distributed as unchanged executeable, as long as this file accompanies them unchanged. The program itself may not be sold, however you may collect fees for copying, distribution or bundled items. It MUST be clear to your customer that he can get the same free of charge from other sources; mentioning the website http://www.compressconsult.com/szip/ and "Freeware" will fulfill this requirement. CAVEATS: DOS/Windows does NOT support binary pipes, so SPECIFY BOTH FILES. Depending on the compilation and the stdio C library there may be limitations of compressable filesizes; on 32-bit computers this limit is often 2 or 4GB. If you see this problem and you can recompile C-code please get in touch with me. The intention is mainly demonstration; I do not consider them production versions. This free program comes with ABSOLUTELY NO WARRANTY OF ANY KIND. Ask me about compression for your data; see http://www.compressconsult.com for more information. Szip is a free program of > d a t a < / / / / compression consulting -------------------------------- History June 2, 1997 MS release of version 1.00 June 9, 1997 MS release 1.01 minor bugfix in compression; sometimes got last byte wrong. Format unchanged July 22, 1997 MS 1.02: decompression has no more hash-table overflows and is faster. I also rearranged modules. Other compiler. Format unchanged July 29, 1997 MS 1.03 UUPS - a little pice of byte order dependent code made it into 1.02 decode, which caused lots of failures. Didn't do testing on PC. Also added an alpha test (sunzip_f) for new fast decode method. Format unchanged. Aug 10, 1997 MS 1.04 Bugfix in fast decompression; distributing fast decompression; first linux distribution Sept 23, 1997 MS 1.05X inofficial release, will not be further supported demonstrates command line options Sept 28, 1997 MS 1.05Xa bugfix in recordsize. Oct 31, 1997 MS 1.05Xd some parameters added, revived old code for order 4 Nov 18, 1997 MS 1.05Xe bugfix when compressing large random files, different compiler options. Feb-Jun 1998 MS 1.10 introduces a new, incompatible fileformat: Header (SZ\012\004). Restart after each block for error tolerance. Changed entropy coder. Support for unlimited order (-o0). Complete rewrite of the probability estimation. The compressed format produced by this version can be processed faster than the old one; even through this program version does a lot of unnecessary and inefficient things it is as fast as the old one. Aug 1998 MS New format again to fix a bug in szip 1.10 and to use a new model (cache) for encoding the model used. Sep 1999 MS 1.12 for some operating systems - uses macros instead of functions for IO Mar 2000 MS keeps memory allocated across blocks to avoid some memory managements that will not allow a working set greater than the memory currently used. MS: Michael Schindler, michael@compressconsult.com